INTERPRAEVENT 2016

13 th Congress 

INTERPRAEVENT 2016 

30 May to 2 June 2016 

Lucerne, Switzerland 

Conference Proceedings 

www.interpraevent.at 

Living 

with natural 

risks

Published by the International Research Society INTERPRAEVENT, Klagenfurt, Austria 

INTERPRAEVENT 2016 – Conference Proceedings © 2016 International Research Society INTERPRAEVENT 

The authors bear full responsibility for the contents of their contributions. 

Edited by Gernot Koboltschnig 

Designed by duoo - visuelle kommunikation, luzern 

All rights reserved; no parts of this publication may be reproduced, stored or retrieved 

in any form or by any means without the prior written permission of the authors. 

The International Research Society INTERPRAEVENT received the right for online and 

printed publishing. Printed in Switzerland.

13 th Congress 

INTERPRAEVENT 2016 

30 May to 2 June 2016 

Lucerne, Switzerland 

Conference Proceedings 

www.interpraevent.at 

Living 

with natural 

risks

13 TH INTERPRAEVENT CONGRESS 

The Science and Technology Advisory Board 

The internationally staffed Science and Technology Advisory Board (STAB) of Interpraevent 

already starts in an early stage of preparation to specify the orientation of the congress. 

The main topic of this congress is “living with natural hazards”. Therefore the STAB intended 

that the call for papers should cover all sectors of the risk cycle including all alpine natural 

hazards processes. And: scientists as well as practitioners should be attracted equally. Finally 

oral presentations were selected out of all submitted contributions based on their quality. 

In this matter quality is determined in an independent assessment process involving several 

experts. So, although the frame can be set in the beginning the final content and orientation 

of the congress is a result of the contribution of each single author. 

Besides their special function in the preparation of the congress, all STAB members served 

with their special expertise as reviewers and therefore had to read most of the written 

contributions. For all of them an Interpraevent Congress is always a gain in knowledge and 

a query of the actual state of the art in natural hazards prevention. 

Niki Beyer Portner (Switzerland) 

Florian Rudolf-Miklau (Austria) 

Ulrik Domaas (Norway) 

Dominique Laigle (France) 

Paolo Simonini (Italy) 

Markus Stoffel (Switzerland) 

Andreas von Poschinger (Germany) 

Matjaz Mikos (Slovenia) 

Yoshiharu Ishikawa (Japan) 

Hans Kienholz (Switzerland) 

Josef Schneider (Austria) 

Johannes Hübl (Austria) 

Bruno Mazzorana (Italy) 

Didier Richard (France) 

Fritz Zollinger (Switzerland) 

Head of STAB 

Editor of Section 1, Risk governance and policies 

2 nd editor of Section 1, Risk governance and policies 

Editor of Section 2, Data acquisition and modelling 

2 nd editor of Section 2, Data acquisition and modelling 

Editor of Section 3, Hazard and risk assessment, 

organization of post excursion 

2 nd editor of Section 3, Hazard and risk assessment 

Editor of Section 4, Hazard and risk mitigation 

2 nd editor of Section 4, Hazard and risk mitigation 

Editor of Section 5, Emergency management, 

organization of field trip 

2 nd editor of Section 5, Emergency management 

Member of STAB and design of 

IP 2016 Editorial Manager 

Member and secretary of STAB 

Member of STAB 

Member of STAB, organization of field trip 

4 | INTERPRAEVENT 2016 – Conference Proceedings

Preface 

Interpraevent is considered to be “the expert network for the protection from natural 

hazards”. In this frame we always tried to attract the experts from science and from 

en gin eering and administration as well. Therefore the call for papers was spread out 

to this huge community to let the experts participate already in the preparation of the 

13th Congress in Lucerne, Switzerland. To cover the entire risk circle the call comprised 

five main topics: 

Section 1: Risk governance and policies (objectives, strategies, communication) 

Section 2: Data acquisition and modelling (monitoring, processes, technologies, models) 

Section 3: Hazard and risk assessment (analysis, evaluation) 

Section 4: Hazard and risk mitigation (structural, nonstructural measures, insurance) 

Section 5: Emergency management (emergency planning, early warning, intervention, 

recovery) 

At none of the earlier Interpraevent congresses we received as many as 370 Extended 

Abstracts. All Extended Abstracts were double reviewed. Out of all acceptable Extended 

Abstracts 147 authors were invited to submit a Full Paper, whereas the decision for 

an invitation was primarily based on the quality of the submitted Extended Abstract. 

The country-wise distribution of Full Papers was adequate to the country-wise distribution 

of submitted Extended Abstracts. 116 authors followed the invitation and finally 108 Full 

Papers that were in line with the congress topics were accepted for publication. Furthermore 

the editorial board decided that only Full Papers can be published if at least one of 

the authors attends the congress. The Full Papers are well distributed over the five sections 

(from section 1 to 5: 18%, 25%, 25%, 17% and 15%) with a higher interest in sections 

2 and 3. Concerning natural hazards processes we are very glad to say that we were able to 

cover all typical alpine natural hazards processes: 19 Full Papers are dealing with avalanches, 

33 with debris flow, 60 with floods, 18 with landslides and 21 with rock fall (as the total 

number is higher than the number of published Full Papers, there are several Full Papers 

dealing with more than one natural hazards process). Similar to all published Extended 

Abstracts 50% of all Full Papers contents are case studies, which means the Interpraevent 

is still able to attract practitioners. 

Concerning the country-wise distribution we have a clearly focus and main interest from 

the country hosting the 13th Congress, because out of all 108 published Full Papers 60 come 

from Switzerland followed by 23 from Austria, 9 from Italy, 5 from France, 4 from Germany, 

3 from Japan, 2 from Taiwan, 1 from Norway and 1 from Slovenia. 

INTERPRAEVENT 2016 – Conference Proceedings | 5

For sure, it is hard work for an author to do an investigation, to prepare and finally to write 

a paper. But after submission editors need to select potential reviewers and to convince them 

to do a review. All 108 Full Papers have been carefully and critically read. Therefore we are 

grateful to all reviewers and would like to especially name those who did at least 4 reviews. 

Many, many thanks to: Franziska Schmid, Hans Romang, Helmut Knoblauch, Andreas 

Zischg, Richard Koschuch, Daniel Trappmann, Matjaž Četina, Catrin Promper, Roland Kaitna, 

Andreas Rimböck and Klaus Pukall. 

Many thanks to all authors, reviewers, editors and the designing team duoo! 

Niki Beyer Portner 

Chair of the Interpraevent Science & Technology Advisory Board (STAB) 

Gernot Koboltschnig 

Managing Editor of the IP2016 congress 

Business Manager of the International Research Society Interpraevent 


CONTENTS 

RISK GOVERNANCE AND POLICIES 

Integrated Risk Management – Identify, Evaluate and Manage Natural Risks 18 

Gian Reto Bezzola; Roberto Loat 

Evaluating the Effectiveness and the Efficiency of Mitigation Measures against Natural Hazards 27 

Michael Bründl; Reto Baumann; André Burkard; Fabian Dolf; André Gauderon; Eva Gertsch; Peter Gutwein; Bernhard Krummenacher; 

Bernard Loup; Adrian Schertenleib; Nicole Oggier; Linda Zaugg-Ettlin 

From the restoration of French mountainous areas to their global management: historical overview 

of the Water and Forestry Administration actions in public forests 34 

Simon Carladous; Guillaume Piton; Jean-Marc Tacnet; Félix Philippe; Régis Nepote-Vesino; Yann Quefféléan; Olivier Marco 

A Gender-sensitive Analysis of Natural Disasters - The Case of St. Lorenzen in Austria 43 

Doris Damyanovic; Karin Weber; Britta Fuchs; Christiane Brandenburg 

Audit „Flood – how prepared are we" – a method for local authorities to improve local 

risk management 52 

Paul Geisenhofer 

Risk-Dialogue – a challenge 61 

Rene Graf 

Examples of urban Flood Risk Management in Styria 70 

Rudolf Hornich; Tanja Schriebl 

Flood Risk Reduction through Object Protection Measures - Benefit Analysis on Catchment 

Scale in Upper Austria 79 

Matthias Huttenlau; Stefan Achleitner; Benjamin Winter; Manuel Plörer; Michael Hofer; Felix Weingartner 

Risk-based spatial planning - findings from two case studies 87 

Roberto Loat; Reto Camenzind; Esther Casanova; Eva Frick 

Public prevention of natural hazards in Swiss law: responsibilities and fields of action 

in legislative and executive bodies 96 

Roland Norer; Mark Govoni; Gregor Kost; Cornel Quinto; Barbara Schielein; Lukas Widmer; Christian Wulz 

Gender mainstreaming in disaster risk reduction – a step towards the visibility of women in DRR 104 

Catrin Promper; Maria Patek 

The central challenge of climate change adaptation for Alpine natural hazard management: 

Incorporation of future change of the damage potential 112 

Klaus Pukall; Sylvia Kruse 

Local Conditions and the Quality of Expert Networks: A Case Study of Avalanche 

Risk Prevention Practices 121 

Renate Renner; Gerhard Lieb 


CONTENTS 

Dealing with gravitational natural hazards: challenges for risk management and 

development planning 131 

Florian Rudolf-Miklau; Kanonier Arthur; Thomas Glade; Stix Elisabeth 

Water management as a part of civil engineering sector in Slovenia 141 

Jošt Sodnik; Blaž Kogovšek; Matjaž Mikoš 

Method for a risk based and transparent allocation formula of the remaining costs applied 

to the protection measures project "Laui Sörenberg", municipality of Flühli, Switzerland. 150 

Roland Stalder; Karl Grunder; Guido Küng 

What makes a successful flood control project? - An evaluation of project procedure 

and risk based on the perspectives of Swiss communes. 159 

Hannes Suter; Luzius Thomi; Raoul Kern; Matthias Künzler; Conny Gusterer; Andreas Zischg; Rolf Weingartner; 

Olivia Martius; Margreth Keiler 

The Zurich flood resilience alliance. A new approach for partnership to effective 

disaster risk reduction 168 

Michael Szoenyi; Linda Freiner 

Chances and Challenges in the Field of Residual Flood-Risk and Risk Communication: 

Ideas from Bavaria 176 

Ronja Wolter-Krautblatter; Andreas Rimböck; Tobias Hafner; Christian Wanger; Christoph Oberacker 

Practical Management of Debris-flow-prone Torrents in Taiwan 186 

Hsiao-Yuan Yin; Chen-Yang Lee; Chyan-Deng Jan; Meei-Ling Lin 

DATA ACQUISITION AND MODELLING 

Deciphering dynamics and magnitude of a recent debris-flow disaster in Vratna Dolina 196 

Juan Antonio Ballesteros Canovas; Milan Lehotsky; Karel Silhan; Sladek Jan; Anna Kidova; Radek Tichavsky; Pavel Stastny; Markus Stoffel 

Modeling rockfall trajectories with non-smooth contact/impact mechanics 203 

Perry Bartelt; Werner Gerber; Marc Christen; Yves Bühler 

A comparison of physical and computer-based debris flow modelling of a deflection 

structure at Illgraben, Switzerland 212 

Catherine Berger; Marc Christen; Jürg Speerli; Guido Lauber; Melanie Ulrich; Brian W. McArdell 

Exploiting damage claim records of public insurance companies for buildings to 

Increase knowledge about the occurrence of overland flow in Switzerland 221 

Daniel Benjamin Bernet; Rolf Weingartner; Volker Prasuhn 


Distributed acoustic monitoring to secure transport infrastructure against 

natural hazards - requirements and new developments 231 

Michael Brauner; Arnold Kogelnig; Ulrich Koenig; Günther Neunteufel; Hanna Schilcher 

Terrestrial radar interferometry for snow glide activity monitoring and its potential 

as precursor of wet snow avalanches. 239 

Rafael Caduff; Andreas Wiesmann; Yves Bühler; Claudia Bieler; Philippe Limpach 

Earthquake-triggered landslides in Switzerland: from historical observations to the actual hazard 249 

Donat Fäh; Jan Burjanek; Carlo Cauzzi; Remo Grolimund; Stefan Fritsche; Ulrike Kleinbrod 

Modelling of individual debris flows using Flow-R: A case study in four Swiss torrents 257 

Florian von Fischer; Margreth Keiler; Markus Zimmermann 

Slide-induced impulse waves in mountainous regions 265 

Helge Fuchs; Robert M. Boes 

Monitoring and management of a huge landslide in a built-up area induced by the 

excavation of deep highway tunnels 274 

Guido Gottardi; Giuseppe Ricceri; Alberto Selleri; Paolo Simonini; Daniela Salucci; Paola Torsello 

Adaption and further development of the numerical solution in the avalanche 

simulation model SamosAT 284 

Matthias Granig; Peter Sampl; Andreas Kofler; Jan-Thomas Fischer; Philipp Jörg 

An Effective Camera Based Water level recording Technology for Flood Monitoring 290 

Issa Hasan; Thomas Hies; Ebi Jose; Rudolf Duester; Marcus Sattler; Matthias Satzger 

Flood protection at Zurich main station: physical model experiments 296 

Florian Hinkelammert; Volker Weitbrecht; Robert M. Boes 

Characteristics of debris flow vibration signals in Shenmu, Taiwan 306 

Yi-Min Huang; Chung-Ray Chu; Yao-Min Fang; Ming-Chang Tsai; Bing-Jean Lee; Tien-Yin Chou; 

Chen-Yang Lee; Chen-Yu Chen; Hsiao-Yuan Yin 

Analysis of flood related processes at confluences of steep tributary channels and 

their receiving streams – 2d numerical modelling application 319 

Johannes Kammerlander; Bernhard Gems; Michael Sturm; Markus Aufleger 

Analysis and classification of bedload transport events with variable process characteristics 327 

Andrea Kreisler; Markus Moser; Johann Aigner; Rolf Rindler; Michael Tritthart; Helmut Habersack 

Integration of remote and terrestrial monitoring data for analysing alpine 

geomorphic processes – examples from Switzerland and Italy 336 

Volkmar Mair; David Mosna; Robert Kenner; Giulia Chinellato; Marcia Phillips; Claudia Strada; Benni Thiebes 


CONTENTS 

Nationwide assessment of the climate sensitivity of natural hazard processes in Switzerland 

A fuzzy logic approach 345 

Peter Mani; Stéphane Losey 

Defining sample size and strategy for dendrogeomorphic rockfall reconstructions 356 

Pauline Morel; Daniel Trappmann; Christophe Corona; Markus Stoffel 

Critical Rainfall Conditions Triggering Shallow Landslides or Debris Flows in Torrents – 

Analysis of Debris Flow events 2012, 2013 and 2014 in Austria 370 

Markus Moser; Stefan Janu; Susanne Mehlhorn 

Experimental Study on Effect of Houses on Debris-Flow Flooding and Deposition in 

Debris Flow Fan Areas 378 

Kana Nakatani; Megumi Kosugi; Yuji Hasegawa; Yoshifumi Satofuka; Takahisa Mizuyama 

Bedload transport simulation with the model sedFlow: application to mountain rivers in Switzerland 387 

Dieter Rickenmann; Martin Böckli; Florian U.M. Heimann; Alexandre Badoux; Jens M. Turowski 

Determining future evolution of landslides from the past: the historical evolution 

of shallow landslides in the upper Cassarate catchment (Southern Swiss Alps) 396 

Christian Ambrosi; Samuel Arrigo; Claudio Castelletti; Cristian Scapozza 

A specially developed side weir in the framework of an integral flood protection 

concept including a hydrological monitoring system 406 

Josef Schneider; Rudolf Schmidt; Matthias Redtenbacher; Franz Brenner 

Powder Snow Avalanche Engineering: New Methods to Calculate Air-Blast Pressures 

for Hazard Mapping 416 

Lukas Stoffel; Stefan Margreth; Mark Schaer; Marc Christen; Yves Bühler; Perry Bartelt 

Monitoring sediment fluxes in alpine rivers: the AQUASED project 426 

Gianluca Vignoli; Silvia Simoni; Francesco Comiti; Andrea Dell'Agnese; Walter Bertoldi; Roberto Dinale; Rudi Nadalet; 

Pierpaolo Macconi; Julius Staffler; Rudolf Pollinger 

Monitoring unstable parts in the ice-covered Weissmies northwest face 434 

Lukas E Preiswerk; Fabian Walter; Sridhar Anandakrishnan; Giulia Barfucci; Jan Beutel; Peter G Burkett; Pierre Dalban Canassy; 

Martin Funk; Philippe Limpach; Emanuele Marchetti; Lorenz Meier; Fabian Neyer 

M-AARE - Coupling atmospheric, hydrological, hydrodynamic and damage models 

in the Aare river basin, Switzerland 444 

Andreas Paul Zischg; Guido Felder; Rolf Weingartner; Juan José Gómez-Navarro; Veronika Röthlisberger; Daniel Bernet; Ole Rössler; 

Christoph Raible; Margreth Keiler; Olivia Martius 


HAZARD AND RISK ASSESSMENT 

Understanding the impact of climate change on debris-flow risk in a managed torrent: 

expected future damage versus maintenance costs 454 

Juan Antonio Ballesteros Canovas; Markus Stoffel; Klaus Schraml; Christophe Corona; Andreas Gobiet; Satyanarayana Tani; 

Sven Fuchs; Franz Sinabell; Roland Kaitina 

Unraveling the spatio-temporal debris-flow activity on a forested cone in the Kyrgyz Range: 

implications for hazard assessment 461 

Juan Antonio Ballesteros Canovas; Vitalii Zaginaev; Markus Stoffel; Sergei Erokhin 

Flood volume estimation in Switzerland using synthetic design hydrographs – a multivariate 

statistical approach 468 

Manuela Irene Brunner; Olivier Vannier; Anne-Catherine Favre; Daniel Viviroli; Paul Meylan; Anna Sikorska; Jan Seibert, 

Channel widening during extreme floods: how to integrate it within river corridor planning ? 477 

Francesco Comiti; Margherita Righini; Laura Nardi; Ana Lucìa; William Amponsah; Marco Cavalli; Nicola Surian; Lorenzo Marchi; 

Massimo Rinaldi; Marco Borga 

Key results of the Swiss wide natural hazard risk assessment on national roads 487 

Luuk Dorren; Philippe Arnold 

Investigation of the flood hazard of the Nuclear Power Plant KKG by earthquake induced 

dam breaks waves at the River Aare 494 

Davood Farshi; Michael Ballmer; Donat Job; Brigitte Faust 

Spatial and temporal exposure of elements at risk in Austria 503 

Sven Fuchs; Andreas Zischg; Margreth Keiler 

Event-based rapid landslide mapping including estimation of potential human impacts 

on landslide occurrence: a case study in Lower Austria 513 

Karin Gokesch; Thomas Glade; Joachim Schweigl 

Integral protection concept "Bielzug" 525 

Nicole Oggier; Christoph Graf; Reynald Delaloye; André Burkard 

Integrated natural hazards protection concept Vitznau LU – Case study Plattenbach 535 

Benjamin Hohermuth; Christoph Graf; Jörn Heilig 

Unveiling the avalanche activity in the Upper Goms Valley (Switzerland) over the 

past 400 years using tree-ring records 544 

Sébastien Guillet; Markus Stoffel; Christophe Corona 

Management of glacier floods in the Bernese Oberland 553 

Nils Hählen; Oliver Hitz; Damian Stoffel 


CONTENTS 

New recommendations for the assessment of river bank erosion hazards 561 

Lukas Hunzinger; Annette Bachmann; Ralph Brändle; Paul Dändliker; David Jud; Mario Koksch 

Human induced risk dynamics - a quantitative analysis of debris flow risks in Sörenberg, 

Switzerland (1950 to 2014) 571 

Margreth Keiler; Benjamin Fischer 

Maps of pluvial floods and their consequences: a case study 580 

Martin Mergili; Andreas Tader; Thomas Glade; Stefan Jäger; Clemens Neuhold; Heinz Stiefelmeyer 

Rockfall Susceptibility Maps in Styria considering the protective effect of forest 592 

Herwig Proske; Christian Bauer 

Sediment input from debris flows into mountain rivers: an event-based perspective 601 

Dieter Rickenmann; Markus Gerber; Martin Böckli 

Potential large wood-related hazards at bridges: the Długopole bridge in the 

Czarny Dunajec River, Polish Carpathians 610 

Virginia Ruiz-Villanueva; Bartłomiej Wyżga; Pawel Mikuś; Maciej Hajdukiewiczd; Markus Stoffel 

ProtectBio - Evaluation of the effects of protection forests on natural hazards due to gravity 619 

Arthur Sandri; Benjamin Lange; Stéphane Losey; Bernhard Perren 

Backwater rise due to driftwood accumulation 628 

Isabella Schalko; Dieter Brändli; Lukas Schmocker; Volker Weitbrecht; Robert Michael Boes 

Natural hazard induced risk: a dynamic individualised approach for calculating hit 

probability on networks 638 

Esther Schönthal; Margreth Keiler 

Flood Risk Map for the Canton of Zurich 647 

Christian Schuler; Thomas Egli; Mirco Heidemann; Manuela Häni 

A probabilistic approach to flood hazard assessment and risk management in floodplains 

considering levee failures 657 

Silvia Simoni; Gianluca Vignoli; Bruno Mazzorana; Claudio Volcan; Francesco Maria Cesari 

Landslide, flood and snow avalanche risk assessment for the safety management system 

of the railway Trento - Malè - Marilleva 670 

Gherardo Sonzio; Fulvio Bassetti; Ezio Facchin; Ettore Salgemma; Lucia Simeoni 

Efficient risk assessment of Norwegian railways combining GIS and field studies 678 

Heidi Hefre; Kjetil Sverdrup-Thygeson; Unni Eidsvig; Øyvind Armand Høydal 

Periglazial Hazard Indication Map: A Basic Instrument in Prospective Hazard Management 688 

Daniel Tobler; Peter Mani; Rachel Riner; Serena Liener; Nils Hählen; Ricarda Bender-Gàl 


The case study of Badouzih rockfall in northern Taiwan: mechanism, numerical simulation 

and hazard assessment 698 

Ching-Fang Lee; Ting-Chi Tsao; Lun-Wei Wei; Wei-Kai Huang 

HAZARD AND RISK MITIGATION 

Integrated bed-load and driftwood retention in Kien - Findings from model-based testing and 

the 2011 flood 708 

Warin Bertschi; Guido Lauber; Jürg Speerli; Armin Hemmi 

Freeboard computation of near critical flows 718 

Niki Antonina Beyer Portner; Patrick Fellay; Karim Laribi; Jean-Louis Boillat 

Planning of risk-based rockfall mitigation measures using 3D rockfall simulations 726 

Thomas Bickel; Markus Hodel 

Protection structures against natural hazards: from failure analysis to effectiveness assessment 735 

Simon Carladous; Jean-Marc Tacnet; Mireille Batton-Hubert 

Study of landslide run-out and impact on protection structures with the Material Point Method 744 

Francesca Ceccato; Paolo Simonini 

Property protection against flooding shown on a complex practical example 753 

Thomas Egli; Daniel Sturzenegger; Pierre Vanomsen 

Flood corridors to handle residual risk 763 

Markus W. Klauser; Christian Willi; Werner Fessler; Josef L.A. Eberli 

Flood protection concept against enourmous debris flow scenarios 771 

Rolf Künzi; Martin Amacher; Oliver Markus Hitz; Serena Liener; Christian Tognacca; Markus Zimmermann 

The impact of debris flows on structures: practice revisited in light of new scientific results 782 

Dominique Laigle; Mathieu Labbé 

People and buildings vulnerability to floods in mountain areas 791 

Luca Milanesi; Marco Pilotti; Roberto Ranzi 

Modern flood protection and rehabilitation concepts at pre-alpine alluvial rivers 799 

Michael Mueller; Peter Billeter; Matthias Mende; Manuel Zahno; Adrian Fahrni 

Bedload trapping in open check dam basins - measurements of flow velocities 

and depositions patterns 809 

Guillaume Piton; Ségolène Mejean; Costanza Carbonari; Jules Le Guern; Hervé Bellot; Alain Recking 


CONTENTS 

Sediment Management in Alpine Catchments: Area of conflicts between protection needs, 

complexity of waste law and quality goals for water bodies 818 

Florian Rudolf-Miklau; Susanne Mehlhorn 

Physical modelling optimization of a filter check dam in Switzerland 828 

Sebastian Schwindt; Giovanni De Cesare; Jean-Louis Boillat; Philippe Bianco; Anton J. Schleiss 

Settlement dynamics in floodplains: from assessing future flood hazard exposure 

to developing spatial adaptation measures 837 

Walter Seher; Lukas Löschner 

Flood losses in Switzerland compared with hazard maps 845 

Luzius Thomi, PhD; Matthias Künzler; Raoul Kern 

Controlled and efficient bed load management by means of variable drain locks embedded 

in a crownclosed large drain sediment control dam 853 

Arthur Vogl; Mathias H. Luxner; Hubert Agerer 

Actions for the Maintenance and Lifespan prolongation of SABO Facilities 862 

Hisashi Watanabe; Toshio Mori 

EMERGENCY MANAGEMENT 

Outflow forecast on a mountain river! (Emergency Management in the Canton of Nidwalden) 872 

Werner Fessler; Markus Klauser; Peter Seitz; Josef Eberli 

Intervention planning as a preventive tool for integral natural hazard management 

in South Tyrol/Italy 882 

Willigis Gallmetzer; Martin Eschgfäller; Roland Fasolo; Peter Egger 

Multiple early warning system on rock walls above a railway line in the Bernese Oberland 

(Switzerland) 891 

Ueli Gruner; Hans-Heini Utelli 

Applied flood-risk-management in the Machland-Nord, Upper Austria 900 

Raimund Heidrich 

Flood forecasting system for the Tyrolean Inn River (Austria): current state and 

furtherenhancements of a modular forecasting system for alpine catchments 909 

Matthias Huttenlau; Johannes Bellinger; Paul Schattan; Kristian Förster; Felix Oesterle; Katrin Schneider; Stefan Achleitner; 

Johannes Schöber; Georg Raffeiner; Robert Kirnbauer 

Advanced Flood Forecasting for Switzerland 917 

Karsten Jasper; Martin Ebel 


Can Twitter catch precursory phenomena before sediment disasters? 927 

Masaru Kunitomo; Joko Kamiyama; Kazuki Matsushita; Yuzuru Yamakage; Kunihiro Takeda; Cheng Qiu; Akiko Ito; Takeru Araki 

The flood warning service of the Austrian Federal Railways 935 

Günther Kundela; Ines Fordinal; Florian Mühlböck; Christian Rachoy 

Early Flood Warning for the City of Zurich: Evaluation of real-time Operations since 2010 944 

Katharina Liechti; Matthias Oplatka; Natascha Eisenhut; Massimiliano Zappa 

A new data management infrastructure for improved analysis and real-time publication 

of flood events in the Canton of Aargau 952 

Christophe Lienert 

Radar-based Warning and Alarm Systems for Alpine Mass Movements 960 

Lorenz Meier; Mylène Jacquemart; Bernhard Blattmann; Sam Wyssen; Bernhard Arnold; Martin Funk 

Practical experience with the Flood scenario catalogue Carinthia, a handbook for flood 

forecast and warning. 969 

Johannes Moser; Christian Kopeinig; Kurt Rohner 

Avalanche detection systems: A state-of-the art overview on selected operational radar 

and infrasound systems 978 

Walter Steinkogler; Lorenz Meier; Stian Langeland; Sam Wyssen 

Integrated risk management of natural hazards by the railway Company BLS Netz AG 988 

Hans-Heini Utelli; Christian Pfammatter; Franz Kuster 

Storm and Flood Warnings issued by Switzerland's Specialist Federal Agencies 997 

David Volken; Coralie Amiguet; Therese Buergi; Daniel Murer; Christoph Schmutz; 

Strategies for the reduction of natural hazards damages by optimized warning, alarming 

and intervention in Switzerland 1005 

Lilith Wernli-Schärer; Roland Bialek; Martin Buser; Christoph Flury; Bruno Gerber; Florian Haslinger; Christoph Hegg; Birgit Ottmer; 

Olivier Overney; Hans Romang; Christoph Schmutz; Jürg Schweizer 



RISK GOVERNANCE AND POLICIES 

(OBJECTIVES, STRATEGIES, COMMUNICATION) 


RISK GOVERNANCE AND POLICIES (OBJECTIVES, STRATEGIES, COMMUNICATION) 

Integrated Risk Management – Identify, Evaluate and 

Manage Natural Risks 

Integrales Risikomanagement – Naturrisiken erfassen, 

bewerten und steuern 

Gian Reto Bezzola, Dipl. Bauing. ETH¹; Roberto Loat, dipl. phil. nat. Geograph² 

ABSTRACT 

In Switzerland about 20 % of all inhabitants live in areas prone to natural hazards. Since 1972, 

on the average 2.5 persons per year lost their lives due to floods, debris flows, landslides and 

rockfalls. The mean annual damage arising from these processes is around CHF 310 million. 

In the past decades there has been a shift from technical defense measures towards an 

integrated risk management. Though, all prevailing natural hazards are considered, preparedness, 

response and recovery measures are combined in an optimal way, all responsible actors 

and those directly affected are involved and all dimensions of sustainability are respected. 

An ongoing communication about risks among all actors is a precondition for successful risk 

management. The main aim is to achieve and maintain a comparable security level for all 

natural hazards throughout Switzerland. The evaluation of actual security, the identification 

of need for action and the definition of priorities must therefore rely on country wide 

standardised risk overviews. A concept for such overviews is presented. 

ZUSAMMENFASSUNG 

In der Schweiz leben rund 20% der Bevölkerung in potentiellen Gefahrengebieten. Durch 

Hochwasser, Murgänge, Rutschungen und Sturzprozesse kamen seit 1972 im Mittel jedes 

Jahr 2.5 Menschen zu Tode und es entstanden Schäden von durchschnittlich 310 Mio. CHF 

pro Jahr. Die früher übliche technische Gefahrenabwehr ist durch das integrale Risikomanagement 

(IRM) abgelöst worden. Integral bedeutet, dass alle Naturgefahren betrachtet 

werden, alle Massnahmen aus den Bereichen Vorbeugung, Intervention und Regeneration 

optimal kombiniert werden, alle Akteure und die direkt Betroffenen beteiligt sowie alle 

Aspekte der Nachhaltigkeit berücksichtigt werden. Nicht mehr die Reduktion der Gefahr 

allein steht im Fokus, sondern vielmehr die Erfassung, Bewertung und Steuerung der 

Risiken, begleitet von einem aktiven Risikodialog. Angestrebt wird ein schweizweit vergleichbares 

Sicherheitsniveau für alle Naturgefahren. Die Beurteilung der aktuellen Sicherheit, 

des Handlungsbedarfs sowie die Definition von Prioritäten müssen sich deshalb auf 

schweizweit standardisierten Risikoübersichten stützen. Ein Konzept für die Erarbeitung 

solcher übergeordneter Risikoübersichten wird im vorliegenden Artikel präsentiert. 

1 Bundesamt für Umwelt BAFU, Bern, SWITZERLAND, gianreto.bezzola@bafu.admin.ch 

2 Bundesamt für Umwelt BAFU 

18 | INTERPRAEVENT 2016 – Conference Proceedings 

IP_2016_FP104

KEYWORDS 

risk governance; integrated risk management; risk overview; natural hazards 

EINFÜHRUNG 

Naturereignisse und der Umgang mit den damit verbundenen Gefahren und Risiken haben 

im Alpenland Schweiz eine grosse Bedeutung und Tradition. Naturereignisse verursachen 

immer wieder Schäden und fordern Menschenleben. Der Umgang mit Naturgefahren be - 

dingt bedeutende Investitionen und durch die Intensivierung der Raumnutzung sowie die 

Folgen des Klimawandels nehmen die Risiken zu. Die mittleren jährlichen Schäden seit 1972 

durch Hochwasser, Murgänge, Rutschungen und Sturzprozesse betragen rund 310 Mio. CHF. 

Im gleichen Zeitraum kamen durch die genannten Prozesse im Durchschnitt jedes Jahr 

2.5 Menschen zu Tode. Allein durch Hochwasser waren seit 1972 vier von fünf Schweizer 

Gemeinden ein- oder mehrmals betroffen und rund 20 % der Menschen in der Schweiz 

leben in potenziellen Überflutungsgebieten. 

PARADIGMENWECHSEL NACH DEM HOCHWASSER 1987 

Die Entwicklung im Umgang mit Naturgefahren und Risiken ist stark durch einzelne Grossereignisse 

geprägt. Ein verantwortungsvoller Umgang verlangt, dass Ereignisse analysiert und 

daraus Lehren gezogen werden. Denn Ereignisanalysen bieten die Möglichkeit, die Wirksamkeit 

von Massnahmen im konkreten Fall zu überprüfen. Solche Analysen haben in der 

Schweiz Tradition. 

Das Jahr 1987 gilt im schweizerischen Hochwasserschutz als Wendepunkt. Nach den 

schweren Unwettern im zentralen Alpenraum mit hohen Sachschäden und 8 Toten setzte 

sich die Einsicht durch, dass Schutzbauten allein nicht genügen, um Schäden durch Hochwasser 

zu verhindern. Die Ereignisse führten die Grenzen der damaligen Hochwasserschutzphilosophie, 

die auf bauliche Eingriffe an den Gewässern fokussiert war, vor Augen. 

Die Erkenntnisse aus der Ereignisanalyse (BWW, LHG 1991) flossen in das Bundesgesetz über 

den Wasserbau (Wasserbaugesetz, WBG) von 1991 ein und leiteten einen Paradigmenwechsel 

ein. Das Motto lautet: „von der Gefahrenabwehr zur Risikokultur“. 

GRUNDSÄTZE DES INTEGRALEN RISIKOMANAGEMENTS (IRM) 

Das Wasserbaugesetz gibt der Gefahrenanalyse, der 

Differenzierung der Schutzziele, den raumplanerischen 

Massnahmen, der zweckmässigen Massnahmenplanung 

und dem Umgang mit dem verbleibenden Risiko 

(Notfallplanung) eine umfassende gesetzliche Grundlage. 

Mit der von der Nationalen Plattform Naturgefahren 

erarbeiteten und vom Schweizerischen Bundesrat 

genehmigten Strategie Naturgefahren Schweiz (PLANAT 

Abbildung 1: Tätigkeiten im IRM, in 

Anlehnung an die Norm ISO 31000. 


2004) wurde der Weg zu einem integralen Risikomanagement vorgegeben. Integral bedeutet, 

dass alle Naturgefahren betrachtet werden, sich alle Verantwortlichen an der Planung und 

Umsetzung beteiligen, alle Betroffen einbezogen werden, alle Arten von Massnahmen 

(planerische, technische, biologische, organisatorische) zu beachten und alle Aspekte der 

Nachhaltigkeit zu berücksichtigen sind. Das integrale Risikomanagement stützt sich auf 

umfassende und aktuelle Gefahren- und Risikogrundlagen und ist begleitet von einem 

aktiven Dialog zu Risiken, Chancen und Handlungsoptionen mit den betroffenen Akteuren. 

Das Vorgehen bei der Umsetzung des integralen Risikomanagements lehnt sich an die allgemein 

gültigen Standards der ISO-Norm 31000 an (Abb. 1). 

ENTWICKLUNG DES IRM 

Der Wechsel von der Gefahrenabwehr zum integralen Risikomanagement erfolgte in der 

Schweiz hauptsächlich im Rahmen der Projektierung von Schutzmassnahmen. Durch die 

Identifizierung vorhandener Defizite, gestützt auf Gefahrenkarten und Schutzziele sowie 

mittels detaillierter Ausweisung der Risiken bei Nutzen-Kostenbetrachtung werden einerseits 

ein vergleichbarer Sicherheitsstandard und andererseits ein effizienter Mitteleinsatz gewährleistet. 

Die integrale Massnahmenplanung reduziert sich jedoch nicht auf eine rein ökonomische 

Betrachtung, sondern ist ein Optimierungsprozess (Abb. 2). Dabei werden die Auswirkungen 

beurteilt, Risiken und Chancen abgewogen, die Verhältnismässigkeit bezüglich 

aller Aspekte der Nachhaltigkeit beurteilt und entschieden in welchem Umfang die Risiken 

gemieden, gemindert und getragen werden. Möglichst früh müssen dabei auch Massnahmen 

der risikobasierten Raumplanung zur Vermeidung neuer inakzeptabler Risiken geplant und 

ergriffen werden um die erreichte Sicherheit zu halten. 

Abbildung 2: Vorgehen, um das angestrebte Sicherheitsniveau zu erreichen und zu halten (ergänzt nach PLANAT 2013). 


Der risikobasierte Ansatz bei Schutzprojekten ist im Naturgefahrenbereich heute weitgehend 

Standard. Einerseits stehen geeignete Methoden zur quantitativen Erfassung der Risiken und 

zur Beurteilung des Nutzen-Kostenverhältnisses risikomindernder Massnahmen im Bereich 

Naturgefahren zur Verfügung (Borter 1999a, 1999b; Bründl 2009). Andererseits bestehen mit 

den seit 2008 mit der Neugestaltung des Finanzausgleichs und der Aufgabenteilung zwischen 

Bund und Kantonen eingeführten Leistungsvereinbarungen mit messbaren Zielen zur 

Verbesserung der Sicherheit entsprechende Anreize und Verpflichtungen. Risikoorientierung 

und Nachhaltigkeit von grösseren Einzelprojekten werden über ein Anreizmodell gefördert. 

Kantone können zusätzliche Bundesmittel beantragen, wenn sie in den Bereichen „Integrales 

Risikomanagement“, „Technische Aspekte“ und „Partizipative Planung“ den Nachweis von 

bestimmten Mehrleistungen erbringen (BAFU 2015a). Spätestens auf Stufe Bauprojekt 

müssen die Kantone den Nachweis der Wirkung (Risikoreduktion) und der Wirtschaftlichkeit 

des Projektes (Nutzen-Kostenverhältnis) erbringen. Das BAFU hat dazu das Berechnungsprogramm 

EconoMe (BAFU 2015b) entwickelt. 

LÜCKEN 

Eine aktuelle Standortbestimmung im Bereich Naturgefahren zeigt, dass bezüglich der Umsetzung 

des integralen Risikomanagements noch Lücken bestehen (BAFU 2016). Diese 

betreffen sowohl die Massnahmenplanung als auch die Grundlagen für das übergeordnete 

Management. 

Auf der Ebene der Massnahmenplanung besteht bezüglich raumplanerischer Massnahmen 

ein grosses Potenzial, da die Risiken wesentlich durch die Nutzungen geprägt sind. Die 

zu nehmend intensivere Raumnutzung darf nicht dazu führen, dass die Risiken unkontrolliert 

zunehmen. Planerische Massnahmen zur Vermeidung neuer inakzeptabler Risiken müssen 

deshalb stärker beachtet werden. Um diese Lücke zu schliessen, laufen momentan verschiedene 

Arbeiten zum Thema der risikobasierten Raumplanung (Camenzind und Loat 2014). 

Weiter müssen im Rahmen der Massnahmenplanung die Risikoträger zur Beurteilung der 

Tragbarkeit der verbleibenden Risiken verstärkt mit einbezogen werden. 

Auf der Ebene des übergeordneten Managements sind heute erst ansatzweise Grundlagen für 

die risikobasierte Ausweisung des Handlungsbedarfs und Prioritätensetzung im räumlichen 

Gesamtkontext der Naturrisiken vorhanden. Auch Instrumente und Grundlagen für einen 

risikoorientierten Mitteleinsatz fehlen heute noch weitgehend. 

LÖSUNGSANSÄTZE 

Die Erfassung und Bewertung der Risiken hat sich auf Stufe Projekt grundsätzlich etabliert. 

Es bedarf jedoch einer Ergänzung dieses „bottom-up“-Ansatzes durch übergeordnete 

„top-down“-Betrachtungen, um innerhalb grösserer räumlicher Systeme und für übergreifende 

Massnahmen den Handlungsbedarf beurteilen und Prioritäten festlegen zu können. 

Bezüglich eines risikobasierten Mitteleinsatzes, einer risikobasierten Raumentwicklung oder 

überregionaler Grossprojekte gilt dies gleichermassen für Bund und Kantone. Sowohl beim 

Bund als auch bei den Kantonen sind deshalb bereits Bestrebungen im Gange Risikoübersich- 


ten zu erarbeiten (siehe beispielsweise: Elsener Metz et al. 2013; Kanton Bern 2014; Kanton 

Zürich 2015). 

Mit der Strategie Naturgefahren Schweiz wird ein schweizweit vergleichbares Sicherheitsniveau 

für alle Naturgefahren angestrebt (PLANAT 2004; 2013, 2015). Das angestrebte Sicher - 

heitsniveau entspricht dem von allen Verantwortungsträgern gemeinsam angestrebten 

Zustand. Schutzziele beschreiben dabei in quantitativer Form den Beitrag einzelner Verantwortungsträgers 

an das angestrebte Sicherheitsniveau und dienen auch als Überprüfungskriterien. 

Um eine optimale Wirkung zu erzielen, müssen die Schutzziele der einzelnen 

Akteure aufeinander abgestimmt sein. Gemäss der Departementsstrategie des Eidgenössischen 

Departements für Umwelt, Verkehr, Energie und Kommunikation UVEK ist ein 

Optimum zwischen den Ansprüchen an das Sicherheitsniveau und der finanziellen Tragbarkeit 

anzustreben (UVEK 2012). Die Beurteilung der Sicherheit, der vorhandenen Defizite 

und des Handlungsbedarfs sowie die Definition von Prioritäten müssen sich aus Gründen 

der Vergleichbarkeit auf schweizweit einheitliche Risikogrundlagen stützen. 

Diese für die übergeordneten Betrachtungen benötigten Risikogrundlagen, nachfolgend 

als „Risikoübersichten“ bezeichnet, müssen verschiedenen Anforderungen erfüllen: 

– Aus Gründen der Übersicht und des Aufwands dürfen sie nicht zu detailliert sein; 

– Aus Gründen der Aussagekraft dürfen sie nicht zu sehr generalisiert sein und müssen 

eine Unterscheidung nach Prozess und Art der betroffenen Schutzgüter erlauben; 

– Aus Gründen der Vergleichbarkeit müssen sie sich auf schweizweit verfügbare, homogene 

Daten stützen; 

– Schutzziele (ARE et al. 2005) sollen für eine erste Überprüfung hinsichtlich Defiziten in 

die Betrachtung einfliessen. 

KONZEPT ÜBERGEORDNETE RISIKOÜBERSICHT 

Das nachfolgend präsentierte Konzept für die Erarbeitung solcher übergeordneter Risikoübersichten 

ist ein Versuch, die bisherigen Überlegungen des Bundes sowie die ersten 

Erfahrungen der Kantone in diesem Bereich zusammenzufassen. Es gliedert sich in die 

nachfolgend beschriebenen 4 Schritte (Abb. 3). 

1. Verschnitt der Grundlagen zu Gefahren und Nutzungen 

Der Verschnitt der Grundlagen zu Gefahren und Nutzungen liefert eine Übersicht der potenziell 

betroffenen Schutzgüter (Schadenpotenzialübersicht). 

Bezüglich Gefahren stehen heute Grundlagen mit unterschiedlichem Detaillierungsgrad und 

Flächendeckung zur Verfügung. Eine umfassende Übersicht der Risiken bedingt, dass die 

Gefahrengrundlagen alle Prozesse abdecken, flächendeckend vorliegen, nach einheitlichen 

Standards erarbeitet sind, für Szenarien unterschiedlicher Wahrscheinlichkeit vorliegen und 

die auftretenden Intensitäten (z.B. Wassertiefe, Fliessgeschwindigkeit, Staudruck) aufzeigen. 

Diesen Forderungen am nächsten kommen die Intensitätskarten, die heute primär für 

Siedlungsgebiete und Hauptverkehrswege (Bahn, Strasse) vorliegen. 


Abbildung 3: Konzept und Schritte zur Erarbeitung übergeordneter Risikoübersichten. 


Schweizweit homogene Daten zur Raumnutzung werden durch verschiedene Bundesstellen 

bereitgestellt, z.B. durch das Bundesamt für Statistik BFS, das Bundesamt für Landestopografie 

swisstopo, das Bundesamt für Raumentwicklung ARE oder das Bundesamt für 

Bevölkerungsschutz BABS. Für Risikoübersichten werden diese Nutzungsinformationen 

sinnvollerweise entsprechend den Schutzgut-Kategorien (ARE et al. 2005; PLANAT 2013) 

gegliedert, indem z.B. unterschieden wird nach: 

– Personen (z.B. am Ort wohnend bzw. am Ort arbeitend) 

– Gebäuden (z.B. Wohnzonen, Industrieareale) 

– Infrastrukturen (z.B. Bahn, Strasse, Anlagen zur Ver- und Entsorgung) 

– natürliche Lebensgrundlagen (z.B. Wald, Landwirtschaftsland) 

– Kulturgütern 

– Sonderobjekten (z.B. Spitäler und Schulen) 

Die Schadenpotenzialübersicht gibt Auskunft darüber, wo welche Schutzgüter bei welchen 

Szenarien mit welchen Intensitäten betroffen sind. 

Da Risikomanagement zukunftsgerichtet ist, sind in diesem Schritt nicht nur die bestehenden 

Nutzungen sondern auch – als weitere Szenarien – geplante Nutzungen und zukünftigen 

Entwicklungen zu berücksichtigen. Informationen zur zukünftigen Entwicklung liefern 

kommunale Nutzungspläne, regionale Entwicklungskonzepte, kantonale Richtpläne sowie 

nationale Sachpläne. 

2. Bewertung des Schadenpotenzials 

Mit Hilfe der Schutzziele (ARE et al. 2005) erfolgt eine erste Überprüfung hinsichtlich Schutz - 

defiziten. Bei dieser Bewertung der Schadenpotenzialübersicht wird z.B. berücksichtigt: 

– wie viele potenziell betroffene Personen 

– schwachen Intensitäten ausgesetzt sind (in der Regel kein Schutzdefizit vorhanden); 

– mittleren oder starken Intensitäten ausgesetzt sind (in der Regel ein Schutzdefizit 

vorhanden). 

– für wie viele potenziell betroffene Sachwerte 

– kein Schutzdefizit vorhanden ist; 

– ein Schutzdefizit vorhanden ist. 

– wie viele Sonderobjekte welcher Art potenziell betroffen sind. 

Das Ergebnis ist eine bewertete Schadenpotenzialübersicht. 

3. Aggregierung zu räumlichen Risikoübersichten 

Aus Gründen der Übersichtlichkeit wird die bewertete Schadenpotenzialübersicht über 

geeignete räumliche Einheiten und durch Aggregierung einzelner Objektkategorien zu einem 

Index zusammengefasst. Zur Sicherstellung der Vergleichbarkeit, sollten dabei jeweils 

mindestens die Indices für die Hauptkategorien Personen, Sachwerte und Sonderobjekte 

ausgewiesen werden: 

– Index Personen 

Summe der bei Ereignissen mit sehr geringer Wahrscheinlichkeit betroffenen Personen; 


mit Unterscheidung des Anteils der durch schwache, mittlere und starke Intensitäten 

betroffenen Personen. 

– Index Sachwerte 

Abschätzung des Schadenerwartungswerts anhand der Wahrscheinlichkeit und standardisierten 

Werten für die betroffenen Objektkategorien (ohne Berücksichtigung von Exposition 

und Verletzlichkeit); mit Unterscheidung des Anteils der Objekte im Bereich ohne bzw. 

mit Schutzdefizit. 

– Index Sonderobjekte 

Summe der bei Ereignissen mit sehr geringer Wahrscheinlichkeit betroffenen Sonderobjekte; 

inklusive Angabe über deren Typ. 

4. Ableitung von Handlungsbedarf und Prioritäten 

Die Übersicht der Risikoindices für verschiedene Prozesse bzw. für verschiedene räumliche 

Einheiten bildet als Risikoübersicht eine weitere und nachvollziehbare Grundlage zum 

Aufzeigen des Handlungsbedarfs sowie zur Definition von Prioritäten in Raum und Zeit. 

Die Betrachtung der Gesamtrisiken (nicht nur des Anteils im Bereich der Schutzdefizite) 

ist notwendig, denn Handlungsbedarf ist grundsätzlich dort gegeben, 

– wo Schutzdefizite vorhanden sind; 

– wo grosse Gesamtrisiken vorhanden sind. 

Die höchste Priorität besteht dort, wo 

– bestehende hohe Risiken rasch und kostengünstig gemindert werden können; 

– neue inakzeptable Risiken gemieden werden können. 

Weitere Kriterien, die auf die Definition von Prioritäten Einfluss haben können: 

– die erzielbare Risikominderung; 

– die zeitliche Realisierbarkeit der möglichen Massnahmen; 

– die Umsetzbarkeit der möglichen Massnahmen (Akzeptanz, Finanzierbarkeit); 

– bereits laufende Planungen (Nutzung von Synergien); 

– eine erste grobe Abschätzung von Nutzen und Kosten. 

Risiken sind immer vorhanden – Handlungsbedarf im Umgang mit Naturrisiken ist 

deshalb grundsätzlich immer gegeben. 

FAZIT 

Ein effektives Risikomanagement im Bereich Naturgefahren bedingt eine umfassende 

Gesamtsicht der Risiken auf Stufe Gemeinde, Kanton und Bund. Entsprechende Übersichten 

sind einerseits notwendig zur Ausweisung des Handlungsbedarfs und zur Definition von 

Prioritäten und andererseits Voraussetzung für den Risikodialog. Solche Gesamtsichten 

basieren aus Gründen des Aufwands und der Übersichtlichkeit sinnvollerweise auf Indices. 

Das hier vorgestellte Konzept stützt sich auf schweizweit verfügbare Gefahren- und Nutzungsdaten 

sowie auf erste Erfahrungen in einzelnen Kantonen. Es gilt nun, die vorhandenen 

Ansätze gemeinsam zu einem Standard und zu einem über die verschiedenen Staatsebenen 

hinweg durchgängigen System weiter zu entwickeln. 


LITERATUR 

- ARE, BWG, BUWAL (Hrsg.) (2005): Empfehlung Raumplanung und Naturgefahren. 

Bundesamt für Raumentwicklung (ARE), Bundesamt für Wasser und Geologie (BWG), 

Bundesamt für Umwelt, Wald und Landschaft (BUWAL), Bern. 48 S. 

- BAFU (Hrsg.) (2015a): Handbuch Programmvereinbarungen im Umweltbereich 2016–2019. 

Umwelt-Vollzug Nr. 1501, Bundesamt für Umwelt BAFU, Bern. 266 S. 

- BAFU (2015b): Wirtschaftlichkeit von Schutzmassnahmen gegen Naturgefahren: EconoMe 

3.0, Bundesamt für Umwelt BAFU, http://www.econome.admin.ch (abgefragt 6.9.2015) 

- BAFU (2016): Umgang mit Naturgefahren in der Schweiz: Bericht des Bundesrats in 

Erfüllung des Postulats 12.4271 Darbellay vom 14.12.2012, Bundesamt für Umwelt BAFU, 

Bern (in Vorbereitung) 

- Borter P. (1999a): Risikoanalyse bei gravitativen Naturgefahren: Methode. Umwelt- 

Materialien Nr. 107/I, Bundesamt für Umwelt, Wald und Landschaft BUWAL, Bern. 115 S. 

- Borter P. (1999b): Risikoanalyse bei gravitativen Naturgefahren: Fallbeispiele und Daten. 

Umwelt-Materialien Nr. 107/II, Bundesamt für Umwelt, Wald und Landschaft BUWAL, Bern. 

129 S. 

- Bründl M. (Ed.) (2009): Risikokonzept für Naturgefahren. Nationale Plattform Naturgefahren 

PLANAT, Bern. 420 S. 

- BWW, LHG (Hrsg.) (1991): Ursachenanalyse der Hochwasser 1987: Ergebnisse der Untersuchungen. 

Mitteilung des Bundesamtes für Wasserwirtschaft BWW Nr. 4, Mitteilung der 

Landeshydrologie und –geologie LHG Nr. 14, Bern. 184 S. 

- Camenzind R., Loat R. (2014): Risikobasierte Raumplanung – Synthesebericht zu zwei 

Testplanungen auf Stufe kommunaler Nutzungsplanung. Nationale Plattform Naturgefahren 

PLANAT, Bundesamt für Raumentwicklung ARE, Bundesamt für Umwelt BAFU, Bern. 21 S. 

- Elsener Metz J., Schulthess J., Schneider A., Willi C., Stocker S., Rauber M. (2013): 

Hochwasser-Risikokarten für den Risikodialog in den Gemeinden - Risikoübersicht für den 

kommunalen Risikodialog im Kanton Schaffhausen. «Wasser Energie Luft» 105(2), 

S. 111-116. 

- Kanton Bern (2014): Naturgefahren im Kanton Bern - Eine Analyse der gefährdeten 

Gebiete und Schadenpotenziale sowie der daraus abgeleiteten Risiken. Arbeitsgruppe 

Naturgefahren des Kantons Bern, Interlaken. 41 S. 

- Kanton Zürich (2015): Risikokarte Hochwasser Kanton Zürich. Amt für Abfall, Wasser, 

Energie und Luft AWEL, http://www.awel.zh.ch/internet/baudirektion/awel/de/wasser/ 

hochwasserschutz/risikokarte.html (abgefragt 6.9.2015) 

- PLANAT (2004): Sicherheit vor Naturgefahren: Vision und Strategie. PLANAT-Reihe 1/2004, 

Nationale Plattform Naturgefahren PLANAT, Bern. 40 S. 

- PLANAT (2013): Sicherheitsniveau für Naturgefahren. Nationale Plattform Naturgefahren 

PLANAT, Bern. 15 S. 

- PLANAT (2015): Sicherheitsniveau für Naturgefahren – Materialien. Nationale Plattform 

Naturgefahren PLANAT, Bern. 68 S. 

- UVEK (2012): Departementsstrategie UVEK 2012, Bern, 31 S. 



Evaluating the Effectiveness and the Efficiency of 

Mitigation Measures against Natural Hazards 

Michael Bründl¹; Reto Baumann²; André Burkard³; Fabian Dolf 4 ; André Gauderon 4 ; Eva Gertsch²; Peter Gutwein 5 ; 

Bernhard Krummenacher 4 ; Bernard Loup²; Adrian Schertenleib²; Nicole Oggier³; Linda Zaugg-Ettlin 6 

ABSTRACT 

The Federal Office for the Environment in Switzerland (FOEN) introduced EconoMe in 2008 

to compare and prioritize mitigation projects against natural hazards. Mitigation projects have 

to be assessed regarding their effectiveness and economic efficiency with EconoMe in order 

to apply for financial subsidies. Experiences and user comments from practitioners have led 

to continuous improvement and the introduction of additional modules. As such, EconoMe- 

Light was introduced in 2015 for assessing projects with low investment costs and for a first, 

rapid estimation of the benefit-cost-ratio of a mitigation project. Analyses of 104 projects 

show that annual risk reduction of most projects exceeds the annual mitigation cost of 

mitigation measures by a factor two. The optimization of the effectiveness and the efficiency 

of mitigation measures is illustrated by the case study Rubi-/Chienbach in the Canton of 

Lucerne. In the next years, the performance and the data basis for risk calculations such as 

vulnerability and lethality curves will be improved and further tests on the robustness of 

the system for decision-making will be conducted. 

KEYWORDS 

risk assessment; Risk concept; Cost-Benefit-Analysis; Optimization; Decision-Making 

INTRODUCTION 

In 2008, the Federal Office for the Environment (FOEN) introduced the Online-Tool 

EconoMe (hereafter EconoMe) for a comparable evaluation of the effectiveness and the 

economic efficiency of mitigation measures against gravitational natural hazards. The legal 

background was a change of the subsidization practice, which requires that mitigation 

projects submitted to FOEN have to be assessed with comparable criteria of economic 

efficiency all over Switzerland. EconoMe is based on the general risk concept for natural 

hazards (Bründl, 2009; Bründl et al., 2009; Tobler and Krummenacher, 2013) and the 

up-to-date version 4.0 (release in April 2016) is available as Online and Offline-Version. 

It guides the user step-by-step through a complete risk assessment to compare the calculated 

risk with protection goals in order to check whether protection measures are needed (Dolf et 

1 WSL-Institute for Snow and Avalanche Research SLF, Davos Dorf, SWITZERLAND, bruendl@slf.ch 

2 Federal Office for the Environment 

3 wasser/schnee/lawinen. Ingenieurbüro Burkard AG 

4 GEOTEST AG 

5 Gutwein IT-Service 

6 WSL Institute for Snow and Avalanche Research SLF 

IP_2016_FP099 


al., 2014) and finally through the evaluation of an analyzed mitigation project by its benefitcost-ratio 

(Fig. 1; Bründl, 2009). EconoMe is well established with cantonal authorities and 

private engineering companies and additional tools with specific purposes were integrated 

in the EconoMe software platform during the last years (Bründl et al., 2012; Bründl, 2012). 

Recently, the overall framework of EconoMe served as a basis for the development of 

Prevent-Building, a tool to evaluate the efficiency of local structural protection measures 

(Bründl and Ettlin, 2014). In the following sections, we present recent developments of 

EconoMe by the example of EconoMe-Light before we show experiences gained with 

EconoMe in the last years and benefits of this tool with a case study. We conclude with an 

outlook on future developments. 

Figure 1: Screenshot EconoMe 3.0 (release version 4.0 in April 2016) showing the working step “consequence analysis”. On the left side, 

links to other tools of the “EconoMe-Family” are shown (1); in this step, the results of considered scenarios could be analyzed in detail 

(2), showing all parameters of a calculation (3). Source: www.econome.admin.ch. 


ECONOME-LIGHT – EXAMPLE OF RECENT DEVELOPMENTS IN ECONOME 

The application of EconoMe requires a sound analysis of the local hazard situation using 

intensity maps for all selected scenarios and an assessment of objects at risk regarding 

location, type and value. Damage susceptibility values for all categories of objects at risk are 

integrated as average values in the software (Bründl, 2009; Bründl et al., 2009). Comments 

of practitioners indicate that the effort for a full benefit-cost-analysis with EconoMe is not 

always appropriate for situations with a low damage potential and low mitigation costs; 

thus, practitioners asked for a simple tool for a quick assessment. In response to this, the 

tool EconoMe-Light was developed; it allows a rapid, simplified risk analysis and a simplified 

benefit-cost-estimation for a given risk situation following the steps (Fig. 2): 

– selecting the hazard process (e.g. debris flow); denoting the mitigation measure and its 

annual costs; 

– selecting one or several hazard scenarios with return periods of 30-, 100-, 300-years 

and/or two scenarios with adjustable return periods; 

– selecting categories (e.g. building) and types (e.g. residential house) for objects at risk; 

– attributing the number of objects at risk to an intensity class in all selected scenarios 

without and with consideration of mitigation measures; 

– interpretation of results: collective (societal) and individual risk to persons is calculated 

automatically; a benefit-cost-ratio greater than one indicates that the selected mitigation 

measure might provide an economically efficient solution; results can be either exported 

as XML-File or printed as pdf-document. 

EconoMe-Light was developed as a tool for a first, rough estimation and not for proofing 

the validity of subsidy payments. Basics like standard values of objects adapted for Switzerland 

(e.g. 650’000 CHF for a one-family residential building) or damage susceptibility values 

are identical to those in EconoMe. First experiences of practitioners indicate that EconoMe-Light 

is applied either for small projects with low damage potential and low costs of 

mitigation measure or as a first hint whether a detailed benefit-cost-analysis with EconoMe 

is justified. EconoMe-Light is available as Online- (with Internet access) or Offline-Tool 

(no Internet access). Registered EconoMe users have access to the tool since April 2015 

via the EconoMe-Platform (www.econome.admin.ch). 


Figure 2: Screenshot of an assessment with EconoMe-Light. (1) input of description, process type, mitigation measure and annual costs; 

(2) definition of scenarios; (3) definition of objects at risk; (4) selection of scenarios for consequence analysis; (5) results; (6) individual 

risk to persons. Source: www.econome-admin.ch. 

EXPERIENCES AND BENEFIT OF ECONOME-TOOLS 

For FOEN, EconoMe has become an indispensable tool in everyday operations concerning 

project steering and financing. Additionally to judging a project‘s general cost-effectiveness, 

various alternatives of protection measures can be compared to each other, which improves 

decision-making. Moreover, EconoMe allows calculating separately the risks and benefits for 

each stakeholder in a project and each process area and thereby cost-splitting among these 

stakeholders. Finally, EconoMe supports the FOEN in an efficient, transparent use of subsidies 

in context of technical protection measures against natural hazards. 

Communities and cantons which apply for subsidies for mitigation projects from the federal 

state (i.e. the FOEN as responsible authority) are required to analyze the effectiveness and the 

economic efficiency of measures using EconoMe. In a recent study, FOEN analyzed projects, 

which received financial support according to Swiss legal regulations (i.e. they were decreed), 

as is the Forestry Act (in German Waldgesetz, FA hereafter) and the Hydraulic Engineering 

Act (in German Wasserbaugesetz, HEA hereafter) for the years 2011-2013. Projects subsidized 

by FA include measures against avalanches, rock fall, and landslides; measures against 

hydrological driven processes like floods and debris flow are subsidized according to the HEA. 

In total, 104 projects were decreed and analysed with EconoMe, as of which 65 projects 

according to the HEA and 39 projects according to the FA. 


Table 1 shows, that 75% of the federal subsidies are invested in projects with a benefitcost-factor 

higher than 1.3 and 50% in projects with a benefit-cost-factor of at least 2.1. 

The mean value of the benefit-cost-factors is 3.6. The benefit-cost-factors of FA-projects are 

slightly higher, because these projects reduce more risks to persons, which are monetized 

with 5 million CHF (app. 4.6 million Euro) per averted fatality (Rheinberger, 2011; Leiter 

et al., 2013). 

Table 1: Benefit-Cost-Factors of mitigation projects decreed by FOEN in the years 2011-2013. 

benefit-cost- 

Projects decreed 



ratio 

according to Hydraulic 

according to Forestry 

according to HEA and FA 

Engineering Act (HEA) 

Act (FA) 

mean 3.3 4.1 3.6 

25% quantile 1.3 1.4 1.3 

median 1.9 2.4 2.1 

75% quantile 3.3 4.0 3.7 

Figure 3 shows the distribution of classified benefit-cost-ratios and the sum of investment 

costs. It can be seen that: 

– investment costs for projects subsidized according to HEA are in average higher than those 

subsidized by FA; 

– investment costs are highest for projects with a benefit-cost-factor between 1 and 2; this 

distribution of investment costs also correlates to the number of projects (not shown in 

the figure); 

– the range of benefit-cost-factors is high (minimum class 0.5 – 1, maximum class > 15); 

– projects with benefit-cost-factors lower than one were also decreed, as a result of a 

weighting of the social, political or ecological interests of the public. 

CASE STUDY RUBI-/CHIENBACH 

With EconoMe it is possible, to optimize a protection project during the planning process in 

order to achieve a better cost-effectiveness of a protection project, which can be illustrated 

with the flood protection project Rubi-/Chienbach in the community of Weggis, Canton 

Lucerne. After the flood in 2005, a set of different technical protection measures including 

a retention basin, dikes, a flood bypass channel as well as additional technical measures for 

managing the over load case was planned. During the participatory planning process with the 

community of Weggis, the Canton of Lucerne and FOEN, a detailed cost-effectiveness study 

for each part of the protection concept was conducted with EconoMe. It showed that the 

combination of a retention basin, dikes and a flood bypass channel with investment costs of 

CHF 6’000’000 would reduce risk by 96%, resulting in a benefit-cost factor of 4.9. The study 

also showed impressively that the additionally planned measures to manage over load case 

with costs of CHF 4’500’000 would reduce the risk by only additional 1%, resulting in a total 


140'000'000 

Sum of investment cost for decreed projects [CHF] 

120'000'000 

100'000'000 

80'000'000 

60'000'000 

40'000'000 

HEA + FA 

HEA 

FA 

20'000'000 

0 

0-0.5 

0.5-1 

1-1.5 

1.5-2 

2-2.5 

2.5-3 

3-3.5 

3.5-4 

4-4.5 

Classes of Benefit-Cost-Factor 

4.5-5 

5-5.5 

5.5-6 

6-15 

>15 

Figure 3: Distribution of sum of investment costs for decreed projects in 2011-2013 in classes of benefit-cost-factors. Projects with a 

benefit-cost-factor between 1 and 2 show the highest sum of investment costs. The range of benefit-cost-factors lies between 0.5 and 

larger than 15. The vertical line marks a change of the class range. 

risk reduction of 97%. Consequently, the community of Weggis decided to renounce these 

additional, non cost-effective technical measures. Instead, remaining risks were planned to be 

reduced with an early warning system and a professional emergency planning. In summary, 

this case study shows that EconoMe can assist planners in financially optimizing a whole 

project in the sense of cost-effectiveness, for the community, the Canton Lucerne and FOEN. 

However, cost-effectiveness is not the only element in the assessment of appropriate, 

sustainable and effective mitigation projects. Other criteria, like incidental costs, ecological 

worth, other superior interests and the acceptance by the involved population and authorities 

in charge are also decisive for the realization of mitigation projects. Therefore, every project 

has to be assessed individually by considering all influencing factors. 

CONCLUSIONS AND OUTLOOK 

Since its introduction in 2008, EconoMe served as a tool for prioritizing mitigation projects 

against gravitational natural hazards by benefit-cost criteria. An analysis of 104 projects 

decreed in 2011-2013 revealed that reduced annual risks overpass annual mitigation costs. 

The largest investments are made with projects with a benefit-cost-factor ranging between 

1 and 2; overall, the benefit-cost-factor varies from 0.5 to larger than 15 with at least 50% 

of the projects exceeding 2.1. 


In spring 2015, EconoMe 3.0 was released and introduced to practitioners in a workshop. 

In this workshop, EconoMe-Light was introduced for a simplified analysis of mitigation 

measures. Experiences over the last years have shown that regular knowledge exchange with 

practitioners in workshops generates valuable feedback helping to streamline the tool to 

users’ needs and thereby strengthens the commitment of users. We recommend to take the 

continuous maintenance of a tool and the education and exchange with users into account 

before the development of such a tool is considered. 

In 2016, version 4.0 was introduced, with enhancements in the performance and the user 

interface. However, a key issue in the next years will remain the improvement of the data 

basis for risk calculations such as vulnerability and lethality curves and further tests on the 

robustness of the system for decision-making. 

REFERENCES 

- Bründl M. (Ed.) (2009). Risikokonzept für Naturgefahren. Einzelprojekt A1.1: Leitfaden. 

Nationale Plattform Naturgefahren PLANAT, Bern: 420 p. 

- Bründl M., Romang H.E., Bischof N., Rheinberger C.M. (2009). The risk concept and its 

application in natural hazard risk management in Switzerland. Nat. Hazards Earth Syst. 

Sci., 9(3): 801-813. 

- Bründl M., Winkler C., Baumann R. (2012). “EconoMe-Railway”. A new calculation 

method and tool for comparing the effectiveness and the cost-efficiency of protective 

measures along railways. In: G. Koboltschnig J., Hübl J., Braun (eds), 12th Congress 

INTERPRAEVENT. International Research Society INTERPRAEVENT, Grenoble: 933-943. 

- Bründl M. (2012). EconoMe-Develop – a software tool for assessing natural hazard risk 

and economic optimisation of mitigation measures, Proceedings International Snow Science 

Workshop “a merging of theory and practice”. ISSW, Anchorage, Alaska: 639-643. 

- Bründl M., Ettlin L. (2014). Assessing the Economic Efficiency of Local Structural Protection 

Measures – Prevent-Building – A Tool for Building Insurances. In: M. Fujita et al. (Editors), 

Interpraevent 2014. International Research Society Interpraevent, Nara, Japan. 

- Dolf F., Krummenacher B., Aller D., Kuhn B., Gauderon A., Schwab S. (2014). Risikoanalyse 

für ein Sihl-Hochwasser in der Stadt Zürich. Wasser Energie Luft 105(4): 23-27. 

- Leiter A. M., Rheinberger C., Bründl, M. (2013). Zur Bewertung von Personenschäden - 

Grundsätze, Methoden und Anwendungen. Zeitschrift für Wildbach-, Lawinen-, Erosionsund 

Steinschlagschutz 77(172): 136-144. 

- Rheinberger C. M. (2011). A Mixed Logit Approach to Study Preferences for Safety on 

Alpine Roads. Environmental Resource Economics 49(1): 121-146. 

- Tobler D., Krummenacher B. (2013). Risk concept Switzerland hazard analysis, risk 

evaluation and protection measures. In: C. Margottini et al. (eds), The second world landslide 

forum, Rome 2011 (WLFF2-2011-0287), Landslide Science and Practice, Vol.7, DOI 10 

1007/978-3-642-31313-4_2, Springer-Verlag, Berlin, Heidelberg. 



From the restoration of French mountainous areas to 

their global management: historical overview of the 

Water and Forestry Administration actions in public 

forests 

Simon Carladous, PhD Student, Eng.¹; Guillaume Piton, PhD Student, Eng.²; Jean-Marc Tacnet, Dr, Eng.²; Félix Philippe, M.D.²; 

Régis Nepote-Vesino, Eng.³; Yann Quefféléan, Eng.³; Olivier Marco, Dr, Eng.³ 

ABSTRACT 

To protect against natural hazards in mountainous areas, the French government has implemented 

a number of forestry and civil engineering works such as check dams in public forests 

since the 19th century. Specifying each dam's objective and protective functions is the first 

requirement for their continued maintenance. The potential technical functions of check 

dams have been clarified in a recent publication (Piton et al. 2016). In the first part of a series 

of papers on the analysis of the context and objectives, this paper focuses on how they were 

implemented in public areas by the Water and Forestry Administration from the end of the 

19th century to the 1960s. We detail the objectives over time and the geographical locations. 

This national overview will help managers consider their present local protection structure 

management problems within an historical perspective. 

KEYWORDS 

Rehabilitation of mountainous areas; mitigation objectives; historical policy analysis; decision 

context 

INTRODUCTION 

In mountainous areas, natural phenomena put people and buildings at risk. Protection systems 

aim at mitigating this risk. In the French mountainous areas, a large number of protective 

structures and forestry works have been implemented since the 19th century. 

As a consequence, more than 21,000 civil protection structures are presently registered in 

approximatively 3,800 km² of protective public forests. 

The Water and Forestry Administration (WFA) carried out afforestation works and built these 

structures in public-owned areas from 1860 to 1964, focusing on torrential phenomena. 

The National Forestry Office (ONF) has managed existing public forests, whereas the national 

government has been responsible for maintaining civil structures since 1964. These decision-makers 

must decide on actions to implement on protective systems. With its expertise, 

the present Rehabilitation of Mountainous Lands department (RTM) of the ONF assists them. 

1 Irstea - ETGR, Univ. Grenoble Alpes, Ecole Nationale des Mines de Saint-Etienne, AgroParisTech, Saint-Martin-d'Hères, FRANCE, 

simon.carladous@irstea.fr 

2 Irstea-ETGR, Univ. Grenoble Alpes, FRANCE 

3 ONF-DTRTM, Grenoble, FRANCE 



Such decisions depend on the decision context, which includes regulatory obligations, 

technical limitations and the budget available that impose limitations and changes over time 

in relation to the sociopolitical context and technical knowledge. Making these decisions 

requires choosing the appropriate actions to implement. Protective actions on mountainous 

systems are compared based on several criteria such as the objective of the system, its 

effectiveness in achieving the objective, and its cost, all of which change over space and time. 

The objective of existing protection systems should be specified to decide on maintenance 

actions, but this also depends on the context in which it has been implemented. It can be 

difficult to understand it without overall knowledge of the changes in the decision context 

over time and space. To assist practitioners, we have undertaken an historical and geographical 

analysis to describe these changes examining several aspects at the French scale. 

– Historical and sociopolitical contexts have already been thoroughly analyzed (Fesquet 

1997). 

– Scientific and technical knowledge on protection works has been reviewed in recent 

scientific papers and technical guidelines. For torrents, this knowledge has been operationally 

developed in France since the 19th century, but there was no summary of how it had 

evolved over time. Therefore, we analyzed the archives in detail to summarize the potential 

functions of check dams in torrential streams (Piton et al. 2016). 

– Actual implementations of RTM laws provide a factual background for the decisions that 

have already been made. Analyzing these examples can help to describe decision contexts 

over time and space using several criteria: public-owned and afforestation areas as well as 

number and cost of civil protection structures. Up to the 1970s, three reports had summarized 

these aspects (Direction générale des eaux et forêts 1911; Messines du Sourbier 1964; 

Mougin 1931). There was, so far, no chronological comparison. 

The present paper focuses on implementations of RTM actions by the WFA from the 19th 

century to the 1960s. In the Methods section, we detail the organization of the database. In 

the Results section, we briefly review the main elements concerning scientific, technical, and 

regulatory changes. We then provide an analysis of the implementations in public-owned 

areas. Overall, this analysis reminds the historical evolution of the RTM actions in France and 

helps to better understand some regional specificities or similarities. 

METHODS 

National archives reports were examined following three axes: i) local decision contexts 

within general, regulatory, and management contexts, ii) technical functions of protection 

structures such as check dams, and iii) the main implementations and resource distribution 

over time including the building of a database registered in a Geographic Information System 

(GIS). The data can be listed as follow: 

works implemented by the WFA in public-owned areas 

– technical aspects: 

– area (ha): public-owned, to be acquired, artificially afforested, naturally forested, 

impossible to afforest, to be afforested; 


– structures (against torrential floods): types, number of civil engineering and rustic 

check dams, channels (km), wattlings and fascines (km), drainage networks (km); 

– structures (against avalanches): types, supporting walls (m), benches (m), diversion 

dams (m); 

– cost in current monetary value: 

– cost of public acquisition; 

– cost of works (against torrential floods): forestry, correction (civil engineering and 

rustic check dams), auxiliary (surveillance paths, fences, etc.) and others (studies); 

– cost of works (against avalanches): new, maintenance; 

– contextual aspects: 

– exposed elements: type and number; 

– specified objectives of works; 

– natural phenomena involved; 

forestry works implemented by local municipalities or by private owners, funded 

by the WFA 

– Technical aspects: 

– area (ha): artificially afforested; 

– Cost in current monetary value: 

– local municipality subsidies; 

– department subsidies; 

– WFA subsidies; 

– WFA subsidies to improve pastures and develop dairy cooperatives. 

These aspects have been registered at the RTM perimeter scale (independent geographical 

entity where afforestation and civil engineering works have been carried out according to the 

laws), and they have been clustered within the administrative departments and the mountainous 

massifs. The data come from three sources (Direction générale des eaux et forêts 

1911, Mougin 1931, Messines du Sourbier 1964) which summarized the WFA actions from 

1860 to their publication (Fig. 1, 2a & 3a), at the national scale for the first and the third 

sources, and at the French Alps scale for the second. A chronological view can be extracted 

(Fig. 2b) that is completed by a new analysis of the existing RTM database. Overall, it 

describes the present protection systems that are now managed in the French RTM public-owned 

areas (Fig. 3b) and how we inherited these thousands of structures with their 

regionally specific types, locations and objectives. 

RESULTS 

Scientific and technical changes and debates 

During the first part of the 19th century, the civil engineers Fabre (1748–1834) in 1797 and 

Surell (1813–1887) in 1841 advocated mountain afforestation as a national concern to 

control soil erosion in the mountains and limit solid transport in rivers and torrents. To 

reforest the headwater areas, Surell proposed using check dams, if needed, to stabilize the 


stream bed before planting trees and bushes on their banks. Even if he also considered 

afforestation as an effective long-term solution, the mining engineer Gras (1806–1873), 

followed by Breton (1811–1892), first theorized specific check dam functions in 1850 and 

1857: to curtail sediment recruitment from the bed and banks within a short-term objective, 

to consolidate cliffs and highly unstable areas, to definitively trap sediments upstream of 

dams, and to regulate sediment transport in torrent beds (Piton et al. 2016). 

Everywhere the government owned the areas to reforest after 1860, works were tested. First 

check dams were built to support afforestation, stabilizing the stream-bed at its current level, 

applying Surell’s theory. They were mainly local construction initiatives, rustic, lower than 

2 m in height, and made of diverse materials: dry stones, wattlings and fascines, brush 

mattresses, sods with or without stones, and wood. Larger check dams in masonry or cut 

stones were also built in fewer numbers, e.g., since 1868 in the Saint-Marthe torrent 

(Hautes-Alpes) under the management of Demontzey (1831–1898), a renowned forestry 

engineer (Direction générale des eaux et forêts 1911). 

Based on the aforementioned preliminary theoretical books and empirical implementations, 

in 1882 and 1897 the forestry engineers Demontzey and Thiéry (1841–1918) published 

technical guidelines. They distinguished forestry works, correction works (civil engineering 

and rustic check dams), auxiliary works, and others (studies) (Fig. 1). For torrents, higher 

check dams (more than 2 m high) were needed in addition to rustic dams. They generally 

were less than 4 m high and made of dry stones when sufficiently large stones were available. 

Masonry was used when stones were too small and for higher check dams. Other restoration 

works were also newly implemented, e.g., temporary retention dams, groynes, and embankments 

to center flow, diversion channels to bypass unstable banks, or dry stone drainage 

systems in landslide areas. For avalanches, techniques were limited to cut-stone walls and 

benches. 

After the period of intense WFA actions (1886–1914), their utility, notably afforestation, was 

debated during the 1920s. This issue was explained in the scientific discourse between the 

geographer Lenoble and forestry engineers (Mougin 1931). Since the 1950s, technical 

developments have broadened actions. Steel and reinforced concrete have helped to develop 

sediment traps and higher check dams as well as new avalanche and rock-fall protection 

structures (snow bridges, netting fences). 

Decisions concerning the management of protection systems in public areas have been 

progressively integrated into a new global natural hazard prevention policy since the 1970s, 

notably including land-use plans (Brugnot & Cassayre 2002). At the global watershed scale, 

current management must also take into account sediment continuity problems and sediment 

starving of valley fluvial systems, even though the initial objective was to curtail sediment 

production. 

Regulatory context and management organization 

The first law on afforestation (1860) aimed at extensively reforesting land to curtail sediment 

production in headwater areas and to limit flood peaks. Overly ambitious, it raised pastoralists’ 

ire, leading to local armed revolts in some regions, such as in the Southern Alps. The law 


on afforestation and grass seeding (1864) attempted to reconcile pastoral activities and soil 

protection using grass’s stabilizing effect. They introduced public management in mountainous 

areas (Fesquet 1997). Afforestation and grass-seeding works were decreed to be in the 

public interest within designated perimeters. Within these areas, private owners and local 

municipalities had to build the structures on their land or the government would impose 

their construction. Moreover, private and local municipalities could carry out voluntary 

afforestation works with state subsidies out of the nationally defined afforestation perimeters. 

In 11 mountainous forestry districts, a specific afforestation department was associated with 

the local WFA administration. 

The law on mountain area conservation and restoration (1882), also called the RTM law, 

reduced the afforestation ambition: the torrent control measures were concentrated in active 

areas (torrents, erosion, and avalanche release zones). The objective was to stop destructive 

events through restoration works. In these areas, the law attempted to reconcile mountain 

agriculture and sustainable behavior. Restoration works were declared of public utility within 

a given perimeter through a specific law. Previously treated areas could be integrated if 

restoration was needed: 703 km² in 1886 (Fig. 2B). Restoration actions could be implemented 

by: (i) the WFA in perimeters acquired by the French national government, (ii) local 

municipalities, or (iii) private owners through subsidized works, which were implemented 

mainly in the Cevennes and in the Northern Alps. Simultaneously, grants were awarded to 

support and improve land, pastures, and cheese cooperatives, mainly in the Pyrenees and the 

Northern Alps. 

The law on the regulation of the water regime (1913) made it possible to declare new 

perimeters of public utility to protect areas from erosion processes even if they were not 

active. This had been partially anticipated in some regions (Cevennes, Southern Alps). 

Since 1966, the WFA has been divided into local agricultural services and the ONF, which 

implemented actions according to the RTM laws in 25 departments (Fig. 2A). Ten specialized 

RTM departments within the ONF were created in 1971 (Fig. 3B). 

Afforestation and civil engineering as technical tools to implement RTM laws 

The 1860 and 1864 laws were applied to limited reforested and grass-seeded areas. The 1882 

law aimed at controlling active areas subjected to erosion, landslides, avalanches, etc. Its 

implementation had different objectives depending on the areas’ geography (Direction 

générale des eaux et forêts 1911). 

– On slopes, stopping the loss of pastoral lands aimed at limiting rural exodus. Curtailing 

sediment production in the headwaters also aimed at limiting distant sedimentation 

damage such as in the Bordeaux harbor receiving sediments from the Mont-Aigoual massif 

(Lozère - 48). Notably in the Southern Alps and the Cevennes, forestry works were favored 

(Fig. 1) in extensive areas that had already been acquired (Fig. 2A), mainly using rustic 

correction works to stabilize torrents and gullies (Fig. 3A). 

– Curtailing sediment production in the headwaters aimed at limiting increases in riverbed 

deposits and bed-shifting of torrential rivers, which aggravated floods and damaged fertile 


agricultural lands, roads, and housing such as on the Var River’s banks (Alpes- 

Maritimes - 06). 

– In local torrent valleys, stabilizing materials or snow in the headwater areas aimed at 

limiting direct damage on productive agricultural lands, housing, industrial areas, roads, 

and railroads such as on the Arc River’s watershed (Savoie - 73). In the Northern Alps, 

the dairy industry purchased limited afforestation areas (Fig. 2A) and favored civil 

structures such as check dams (Fig. 3A). 

Figure 1: Cost distribution of works implemented in public forests in application of RTM laws between 1860 and 1909 depending 

on the administrative department. Correction structures were the priority in the Northern Alps. Forestry works were mainly used in 

the Cevennes. Correction and forestry works were more evenly distributed in the Southern Alps and the Pyrenees. 

Since World War I, the number of new projects has decreased even if land acquisitions were 

not discontinued (Fig. 2B). Following World War II, the decrease in funding to maintain 

structures continued. Messines du Sourbier (1900–1989) provided a detailed national survey 

of implementations in 1964 (Fig. 2, Fig. 3A). Acquired areas at the national scale reached 

approximately 3,700 km², mainly in the Southern Alps and the Cevennes (Fig. 2) (Messines 

du Sourbier 1964). 


Figure 2: Distribution of RTM-acquired lands in 1965, their afforestation rate (A) and their evolution in the 25 RTM departments and the 

mountain massifs (B). 


In application of the RTM laws in publically owned areas, more than 100,000 check dams, 

mainly rustic dams made of dry stones (Fig. 3A), 28 km of channels and tunnels (exclusively 

in the Northern Alps), 663 km of drains, and 68 km of avalanche protection structures 

(mainly in the Alps) were registered (Messines du Sourbier 1964). 

Figure 3: Geographical distribution of (A) works in 1964 and (B) managed works in 2014. The present distribution of managed avalanche 

structures (B) is explained by the historical preferential implementation of these works in the Northern Alps and the Pyrenees (A). The 

RTM civil engineering actions have been mainly implemented in headwater areas and in the Alps (B). In 1965, more than 100,000 check 

dams were registered, but they were mainly rustic dams. The proportion of usual dams was higher in the Northern Alps (A). In April 

2014, more than 14,000 check dams were registered in the RTM database, covering 10 departments. Around 50% of them are lower 

than 2 m in height (B). 

Many rustic check dams built before 1914 have not been maintained as planned in initial 

guidelines (Demontzey 1882), due to the lack of grants but also for multiple reasons such as 

technical evolutions (e.g. open check dams advent), decrease in expectations concerning the 

hydrological role of forests, artificial and spontaneous afforestation, rural depopulation and 

changes in the decision makers' priorities. Even if they were registered in 1964, the new 

organization of the RTM reduced the number of managed structures, focusing on the most 

relevant ones in the current decision context and technical comprehension of natural hazards 


(Fig. 3). Today, as a result of all the past implementations, 85% of managed structures in the 

public forests maintained by the 10 RTM departments are located in headwaters and seek to 

limit the impact of torrents, including erosion processes. Eighty-five percent of torrential 

protection structures are check dams (Fig. 3B). 

CONCLUSIONS 

This paper overviews the changes in WFA actions according to the RTM laws from the 19th 

century to the 1960s. Several objectives have been assigned to protection systems within their 

geographical contexts: from extended afforestation in headwater areas to local protection. 

Protection structures have specific functions designed to meet their objectives. Forestry works 

and check dams have been the most widely implemented. 

Presently, RTM experts are expected to manage protection structures, mainly check dams, 

both technically and strategically. For the former, the local functions of the existing check 

dams must be specified with their objective in mind. For the latter, the objectives should be 

reviewed. Maintenance decisions are taken according to the present context and protection 

objectives but must optimally be aware of the past historical decisions. 

In the second part of this historical analysis (in preparation), we will describe the development 

of the overall management of mountainous areas since the 1970s, highlighting several 

key points such as the role played by avalanches, the public finance policy in favor of 

municipally-managed structures and of land-use plans, as well as the need for forest renewal. 

REFERENCES 

- Brugnot G., Cassayre, Y. (2002). De la politique française de restauration des terrains en 

montagne à la prévention des risques naturels. In FAO, ed. XII World Forestry Congress 

Conference proceedings. 

- Demontzey P. (1882). Traité pratique du reboisement et du gazonnement des montagnes. 

- Direction générale des eaux et forêts (1911). Restauration et conservation des terrains en 

montagne. Trois parties. 

- Fesquet F. (1997). Un corps quasi-militaire dans l’aménagement du territoire : le corps 

forestier et le reboisement des montagnes méditerranéennes en France et en Italie aux XIX et 

XXèmes siècles. PhD Thesis, Université Paul Valéry, Montpellier III. 

- Messines du Sourbier J., (1964). Enquête sur la conservation et la restauration des terrains 

en montagne. 

- Mougin M.P. (1931). La restauration des Alpes. 

- Piton, G. et al. (2016). Why do we build check dams? An historical perspective from the 

French experience. Earth Surface Processes and Landforms. 



A Gender-sensitive Analysis of Natural Disasters - 

The Case of St. Lorenzen in Austria 

Karin Weber, DI¹; Doris Damyanovic, Ass.Prof. Dipl.-Ing. Dr.¹; Britta Fuchs, DI Dr.¹; Christiane Brandenburg, Ao. Prof. DI. Dr.² 

ABSTRACT 

In this paper, we present a gender-sensitive analysis of a natural disaster from a planning 

perspective, illustrated by a debris flow in Styria, Austria, in 2012. The introduced findings 

were based on twenty semi-structured interviews with the residents considering the cycle of 

integrated risk management. The gender-sensitive research design identified various aspects 

that can enhance or decrease the capacity and vulnerability on the individual level. The 

gender-sensitive analysis showed that the vulnerabilities and capacities of people vary during 

the different phases of the risk cycle (prevention, coping, recovery). The ability to cope with 

natural hazards and risks varies highly, and this variability can only be revealed and be 

understood differentiated by a gender+ approach. At the same time, socio-economic factors 

and age of the community members have to be taken into account. Further research is 

needed to analyse and integrate gender+ specific needs and capacities in the field of disaster 

risk management effectively. 

KEYWORDS 

gender-specific approaches; natural hazard; capacities; Disaster Risk Management; 

vulnerability 

INTRODUCTION 

Social vulnerability can be defined as “the characteristics of a person or group and their 

situation that influence their capacity to anticipate, cope with, resist and recover from the 

impact of a natural hazard” (Wisner et al., 2004, p. 11). These social aspects are paramount 

for understanding different behaviour in dealing with natural hazards and include the 

socio-economic status of a given community and their members, as well as age structure, and 

- often neglected - aspects of gender. Surveys in the context of development cooperation 

(Bradshaw, 2015; JERA & eS4W, 2012; UNISDR et al., 2009) and recent research (e.g. 

Bacanovic, 2015; Chávez Rodríguez, 2013; Damyanovic et al., 2014; LeMasson, 2013) 

demonstrate that the consequences of natural disasters are not gender-neutral and need to 

be considered to reduce vulnerability effectively. 

In the context of natural hazards in alpine regions, the gender+ approach has so far only been 

applied in the research project GIAClim (Damyanovic et al., 2014) and in a Master’s thesis 

(Weber, 2015). Both analyses dealt with the major debris flow of 2012 in St. Lorenzen/ 

Austria. They indicate further discussions and research questions for a deeper understanding 

1 Universität für Bodenkultur Wien, Institut für Landschaftsplanung, Vienna, AUSTRIA, doris.damyanovic@boku.ac.at 

2 Universität für Bodenkultur Wien, Institut für Naturschutz, Erholungsplanung und Landschaftsentwicklung, Vienna, AUSTRIA 



of gender+ aspects in the disaster risk management in a European context. Damyanovic et al. 

(2014) and Weber (2015) designed their research as a case study. That entails that they 

provide insight into people’s experience with natural hazards and challenges in daily life from 

a particular area. However, case studies allow for collecting and investigating “contemporary 

phenomena in depth with real-life context” (Yin, 2009, p. 240) and contribute to the 

cumulative development of knowledge. 

The main research question of this paper is: 

How does the social context, including aspects of gender+ (gender, age, socioeconomic 

status) influence the individual behaviour of people within the cycle 

of integrated risk management? 

The authors embarked on answering this research question through a case study on the debris 

flow in July 2012 in St. Lorenzen in Paltental (district Trieben, provincial state of Styria in 

Austria), a village of ca. 300 permanent residents. St. Lorenzen is situated on an alluvial fan 

that was formed by the river Lorenzerbach. The Lorenzerbach is defined as a bedload carrying 

torrent. The actual hazard zone map (adopted in 2009) indicates that most of the residential 

zones of St. Lorenzen are situated in hazard-prone areas, which were also flooded during this 

event (see Figure 1). Intense precipitation events between June and August 2012 triggered 

the debris flow in St. Lorenzen. Numerous buildings were destroyed. The debris flow, was 

rated as an extreme event due to its magnitude and high amount and level of debris. 

Figure 1: Location of the case study, the hazard zone map of 2009 and the debris flow 2012, source: BEV (2014), Weber (2015). 


THEORETICAL FRAMEWORK 

GENDER+ AND VULNERABILITY 

The equation Disaster Risk = Hazard x Vulnerability underlines the duality of a natural (or 

human-induced) phenomenon and the concept of vulnerability (LeMasson, 2013). This 

structural view, which is regarded as a dominant view in vulnerability research (Hufschmidt, 

2011), reflects the social, economic, cultural, and political context people live in and their 

everyday living conditions, which are embedded in socially constructed modes of living 

(Fordham, 1999). A gender-specific conception of “risk” includes the assumption that 

“gender-based differences and inequalities have a strong negative or positive effect on the 

vulnerability and capacities of people exposed to hazards” (UNISDR et al., 2009). Consequently, 

the research design of this case study is based on the premise that specific sets of 

inequalities (class, race/ethnicity, sexual orientation and gender) need to be addressed to 

highlight the differentiated character and dynamics of inequalities, as indicated in Verloo 

(2006). The concept of intersectionality aims to reveal how gender-inequalities are connected 

to other structural inequalities. These structural inequalities have an impact on livelihoods 

of people and affect their everyday lives and capacities to cope with natural hazards. 

A livelihood consists of different bundles of resources or assets (both material and social 

resources) that are needed to support everyday lives. This perspective suggests that households 

with a larger bundle of assets will be more resilient to a hazard than a relatively 

asset-less household (FAO, 2008). In the context of livelihoods, gender+ is understood as a 

structural category which effects the everyday life situation and the power-relations between 

women and men (Hofmeister et al., 2013). In this research it is used as an analytic tool to 

understand the gender dimension in the risk perception of natural hazards. 

METHODOLOGICAL APPROACH TO THE CASE STUDY 

The diversity of methods applied in this case study reflects this concept of vulnerability and 

the interdisciplinary approach that has become standard in studies on disaster (e.g. Mercer 

et al., 2010). A set of methods was used to evaluate coping capacities and vulnerabilities of 

affected community members, following the aim to propose the implementation of gender+ 

in planning strategies. The research design consisted of spatial surveys, semi-structured 

interviews and document analyses. 

The method of semi-structured interviews was chosen in order to allow new ideas and individual 

points of view to be brought up during the interview as a result of what the interviewee 

says. The interview guideline (Damyanovic et al., 2014; Weber, 2015) was structured along 

the cycle of integrated risk management (BABS, 2013). This framework is used frequently in 

hazard and risk assessments to describe an ideal sequence of phases in dealing with natural 

hazards. The interview guideline included questions concerning knowledge and information 

about emergency provisions and prevention before, and after the hazard event. The affected 

community members were asked how they experienced the event and how they prepared for 

intervention. In addition, they were asked how they commu ni cated after the debris flow and 


organised reconstruction work in terms of the assignment of roles between women and men. 

The interviewees were contacted mainly through door-to-door requests or word-of-mouth 

recommendations (snowball sampling). To ensure anonymity, the interviewees were grouped 

according their age in ten annual steps (30-39 years; here 30+). The sample of 20 residents 

(men 60%: 3x 30+, 1x 40+, 3x 50+,4x 60+, 1x 80+; women 40%: 1x 30+, 3x 50+, 3x 60+, 1x 

70+) (Weber, 2015) was selected with the aim of high variability concerning socio-economic 

characteristics and severity of losses due to the debris flow, respectively location in the hazard 

zone. The interviews (duration: 1 to 2 hours average) were analysed chronologically along 

the cycle of integrated risk management using qualitative content analysis (Mayring, 2007). 

FINDINGS OF THE CASE STUDY 

The findings of the case study are structured along the risk cycle: preparedness, response, 

recovery. 

PHASE OF PREPAREDNESS 

St. Lorenzen and its surroundings have been affected by hazard events in the past. Before 

the debris flow occurred in 2012, five interviewees (25%) had own experiences in coping 

with hazard events, gained within their paid and voluntary work (3 men 2x 30+, 60+) and 

own experience (2 women 50+, 70+). For men, this experience within work offers them basic 

knowledge of how to respond during emergencies, rescue and evacuation. Furthermore, 

close ties to the local fire brigade offered them advantages in getting information at first hand 

(during the entire risk cycle). 

Almost all interviewee know the existence of the hazard zone plan which is provided by 

the Austrian Service for Torrent and Avalanche Control, a federal institution of the Federal 

Republic of Austria. More in-depth knowledge of the specific location and type of hazard 

zones was available concerning their own property. But still, four interviewees (20%) 

(3 men 30+, 40+, 60+; woman 60+) were not aware that their property was located in a 

hazard zone before the purchase of land or moving into the house, although a hazard zone 

plan existed since 1978. 

There is no specific knowledge of an emergency plan for St. Lorenzen. Two persons (10%, 

2 men 30+, 40+) mentioned that they know about risk management in general, like the 

planning of an evacuation and organisations responsible, as they are active members of 

the local fire brigade. 

PHASE OF RESPONSE 

Women initiated individual preparations for themselves and other family members, like 

sleeping in the upper floor, packing of belongings and documents needed in case of evac u- 

ation measurements in advance (3 women 30+, 50+, 60+). During the phase of response, 

women (4 women 2x 30+, 60+, 70+) were responsible for preparing the evacuation (coordinating 

children and family, assembling goods). Undertaking preparations for intervention, 


men were exposed to risky situations. Men (belonging to every age group) reported of 

clearing the channel from woody debris and stones to enable the runoff or installing 

protection measures on private property. These results show similarities to e.g. Pfister (2009) 

where a gender-specific division of tasks was observed too. 

The elderly (70+, 80+) were vulnerable at the phase of evacuation, because they were 

dependent on external help (local fire brigade and Red Cross) due to constricted agility and 

mobility. These results confirm similar findings of Birkmann et al. (2012) and Bacanovic 

(2015). More than half of the interviewed women (63%) described the evacuation in detail, 

as it affected them emotionally (similar to findings of Paech, 2013). One woman (50+) 

described the risky self-evacuation as traumatising. 

Similar to the phase of prevention, men (all age groups) and one woman (50+) were exposed 

to hazardous situations, as they went out to maintain protective measures (similar as in 

Bacanovic, 2015) or to help family members living in the neighbourhood. This high exposure 

lasted, because 20% of the interviewed men (30+, 2x 50+, 60+) refused to follow evacuation 

orders and stayed in the destroyed area to guard their devastated homes and care for the 

animals. 

Women (5 out of 8 - 63%) were seeking information at the evacuation centre, whereas men 

were told to be actively communicating (8 out of 12 – 66%) with the rescue service (similar 

to Bacanovic, 2015) and the local fire brigade. 

PHASE OF RECOVERY 

In this study, the elderly (60+, 70+, 80+) had minor problems financing reconstruction 

measures. Still, some households had to use most of their savings (men 50+, 80+), severance 

indemnity (man 50+), or even take out a loan (man, woman 50+) to finance building 

materials in advance. Financial support was organised for most of the persons affected. Not 

every interviewee could deal with the end of the application deadline. Elderly women (70+) 

and men (80+) needed help to fill in the applications. One young family (30+) was confronted 

with other vicissitudes of lives that slowed down their decision whether to stay in St. 

Lorenzen or not and applying for financial support, leading to lack of financial support. 

Answering questions about the division of labour during the phase of recovery, 25% (3 men 

2x 30+, 80+; women 50+, 60+) mentioned a stereotypical division of labour (similar to 

Chávez Rodríguez, 2013) where men did physical and women did care and subsistence work 

(IRP et al., 2010). 

Women followed up paid work sooner than men (with one exception, man 30+ – who felt 

the need to earn money) (similar findings in JERA & eS4W, 2012). Either they took a few 

days off (woman 60+), couldn´t get any unpaid leave (50+) or worked in own business (60+). 

This increased pressure added to the wearing situation of recovery, e.g. for a woman (50+) 

who was traumatised, suffered from insomnia, but had to go to work. On the contrast, men 

applied for leave of absence (either paid or unpaid) from their occupation to help with 

reconstruction works. 


PHASE OF PREPAREDNESS (AFTER THE HAZARD EVENT) 

Women (5 out of 8-63%) said they were concerned about possible risks, although edificial 

preventive structures had been built soon after the debris flow occurred. Whereas men 

(8 out of 12-66%) said they felt save. Only women (3 out of 8: 30+, 50+, 80+) are prepared 

for (eventually) recurring events, e.g. emergency package of belongings in the upper floor. 

Private protective measures include reinforced concrete walls to protect the property of the 

inundated area from future damages, expressing security needs. There were no gender-specific 

differences found concerning the degree or realization of protective measurements, which 

was rather determined by legal allowances. 

DISCUSSION 

The results highlight that a gender+-sensitive approach leads to a more differentiated and 

detailed knowledge about risk perception, coping capacities and vulnerability. First of all 

the results of the interviews show that women and men, according to their age and socioeconomic 

situation have different abilities to cope with and deal with natural hazards. 

Along the phases of the cycle of integrated risk management (see Figure 2). The findings 

indicate that the characteristics of local social relations, relationships with emergency services, 

and the type of housing and age profile of residents are playing an important role in terms of 

the capacities (Walker et al., 2006, p. 778). The findings highlight the need for a gender- and 

group-specific approach in the analysis of risk perception (Fleischhauer et al., 2012) as the 

survey revealed that men and women, young and old behaved and reacted differently according 

to their role and responsibilities in the family and community. It also revealed group- and 

gender-specific capacities that should be taken into account in DRR. 

Some studies (JERA & eS4W, 2012; Paech, 2013; Bacanovic, 2015) deal with gender-specific 

aspects of natural hazards within industrialised countries, but especially in the Alpine regions 

empirical studies are rare (Weber 2015). The gender-specific approach as discussed and 

applied in the planning disciplines and in sustainability science since the 1970ties (Hofmeister 

et al., 2013) can be taken as an example for integrating gender as an analytic category in 

DRR. Furthermore, gender understood as epistemological category contributes to a critical 

discussion of (feminist) methodologies and methods in sciences, engineering sciences and 

planning sciences (e.g. Fox Keller, 1986; Althoff et al., 2001; Hofmeister et al., 2013). 

CONCLUSION 

In the conclusion, we summarize aspects of a gender-sensitive approach in DRR that need to 

be developed further, especially in the context of Alpine regions. 

To support people´s livelihoods and strengthen their capacities through a gender-sensitive 

planning processes and natural hazards management, the entire risk cycle has to be taken 

into account. The integration of more comprehensive gender-sensitive surveys and sex and 

gender disaggregated data (see also in JERA & eS4W, 2012) would be necessary to gain 

deeper knowledge about gender-specific needs. However, the results of this case study 

indicate possible hinges for a gender-sensitive DRR and management of natural hazards: 


Figure 2: Findings from the semi-structured interviews, organised along the phases of the risk cycle. 

The members of the community have local knowledge, their specific experience and 

competences with regard to hazards, which are of great value for an efficient and fair DRR. 

Gender+ aspects need to be considered in all phases and at all scales of planning and 

communication processes concerning natural hazards (Damyanovic et al., 2014; Fuchs et al., 

forthcoming). The findings of this paper highlight starting points for supporting coping 

capacities. First of all, gender-specific data (on a local level) should be integrated in emergency 

plans in order to identify high priority areas in case of emergency (Weber, 2015). 

The identified lack of information about emergency provisions and prevention measures 

(hazard map included), should be tackled, using gender-specific information materials and 

participation on all levels and in all phases (Fleischhauer et al., 2012). As women do not have 

as much experience with emergency situations as man, due to rare participation in the local 

fire brigade, increased participation of women in trainings should be encouraged (also 

suggested in Bacanovic, 2015). 


REFERENCES 

- Althoff, M., Bereswill, M., Riegraf, B. (2001). Feministische Methodologien und Methoden. 

Wiesbaden: VS Verlag für Sozialwissenschaften. 

- Bacanovic, V. (2015). Gender Analysis of the Impact of the 2014 Floods in Serbia. Organization 

for Security and Co-operation in Europe - OSCE (publisher). 

- Birkmann, J., Bach, C., Vollmer, M. (2012). Tools for Resilience Building and Adaptive 

Spatial Governance Challenges for Spatial and Urban Planning in Dealing with Vulnerability. 

Raumforschung Raumordnung, 70, pp. 293–308. 

- Bradshaw, S. (2015). Engendering development and disasters. Disasters, 39, pp. 54–75. 

Retrieved December 8, 2015, 

- Bundesamt für Bevölkerungsschutz (BABS) (Ed.). (2013). Leitfaden KATAPLAN. Grundlage 

für kantonale Gefährungsanalysen und Massnahmenplanungen. Bern. 

- Chávez Rodríguez, L. (2013). Klimawandel und Gender: Untersuchung der Bedeutung von 

Geschlecht für die soziale Vulnerabilität in überflutungsgefährdeten Gebieten. 

- Damyanovic, D., Fuchs, B., Reinwald, F., Pircher, E. M., Eisl, J., Allex, B., Brandenburg, C., 

Hübl, J. (2014). GIAKlim - Gender Impact Assessment im Kontext der Klimawandelanpassung 

und Naturgefahren. Auftraggeber: BMLFUW, BMWFW, ÖBF, Land Oberösterreich. 

- FAO - Food and Agriculture Organization of the United Nations (Ed.) (2008). Disaster risk 

management systems analysis: a guide book. Rome. 

- Fleischhauer, M., Greiving, S., Flex, F., Scheibel, M., Stickler, T., Sereinig, N., Koboltschnig, 

G., Malvati, P., Vitale, V., Grifoni, P., Firus, K. (2012). Improving the active involvement of 

stakeholders and the public in flood risk management. Tools of an involvement strategy and 

case study results from Austria, Germany and Italy. Natural Hazards and Earth System 

Science, 12(9), pp. 2785–2798. 

- Fordham, M. (1999).The intersection of gender and social class in disaster: balancing 

resilience and vulnerability. International Journal of Mass Emergencies and Disasters. 17(1). 

pp 15-36. 

- Fox Keller, E. (1986). Liebe, Macht und Erkenntnis: männliche oder weibliche Wissenschaft?. 

München: Hanser. 

- Fuchs, B., Damyanovic, D., & Reinwald, F. (forthcoming). A Gender-sensitive Analysis of 

Spatial Planning Instruments Related to The Management of Natural Hazards in Austria., In: 

Buckingham, S., LeMasson, V. (Eds.), Understanding Climate Change through Gender 

Relations. Routledge. 

- Hofmeister, S., Katz, C., Mölders, T. (Eds.) (2013). Geschlechterverhältnisse und Nachhaltigkeit: 

die Kategorie Geschlecht in den Nachhaltigkeitswissenschaften. Opladen: Budrich. 

- Hufschmidt, G. (2011). A comparative analysis of several vulnerability concepts. 

Natural Hazards.58(2). pp. 621–643. 

- IRP, UNISDR, UNDP (Eds.) (2010). Guidance note on recovery: Gender. 

- Justice Equality Rights Access International (JERA), Economic Security4Women (eS4W) 

(2012). Women´s Voices From The Flood Plains. An economic gender lens on responses in 

disaster affected areas in Queensland and Victoria. North Sydney, NSW. 


- LeMasson, V. (2013). Exploring Disaster Risk Reduction and Climate Change Adaption 

from a gender perspective. Insights from Ladakh, India. 

- Mayring, P. (2007). Qualitative Inhaltsanalyse: Grundlagen und Techniken. Weinheim: 

Beltz. 

- Mercer, J., Kelman, I., Taranis, L., Suchet-Pearson, S. (2010). Framework for integrating 

indigenous and scientific knowledge for disaster risk reduction. Disasters, 34(1), pp. 214–239. 

- Paech, F. (2013). Eine Naturkatastrophe gendern?! Erzählungen über die Hamburger 

Sturmflut von 1962. ARIADNE: Forum für Frauen- und Geschlechtergeschichte, 64, pp. 

58–65. 

- Pfister, C. (2009). Learning from Nature-induced Disasters. Considerations from Historical 

Case Studies in Western Europe. In: C. Mauch & C. Pfister (Eds.), Natural Disasters, Cultural 

Responses. Case Studies Toward a Global Environmental History, pp. 17–40. Lanham: 

Lexington Books. 

- UNISDR, UNDP, & IUCN. (2009). Making Disaster Risk Reduction Gender-Sensitive. 

Policy and Practical Guidelines. Genf. Swiss: United Nations. 

- Verloo, M. (2006). Multiple Inequalities, Intersectionality and the European Union. 

European Journal of Women’s Studies, 13(3), pp. 211–228. 

- Walker, G. P., Great Britain, & Environment Agency. (2006). Addressing environmental 

inequalities: flood risk. Bristol: Environment Agency. 

- Weber, K. (2015). Landschaftsplanerische Betrachtung des Umgangsmit Naturgefahren 

aus genderspezifischer Perspektive am Fallbeispiel des Murenabganges in St. Lorenzen im 

Paltental 2012. Masterarbeit, Universität für Bodenkultur, Wien. 

- Yin, R. K. (2009). Case Study Research. Design and Methods. Sage Publications, 

Thousand Oaks, (4th ed.). 



Audit „Flood – how prepared are we" – a method for 

local authorities to improve local risk management 

Das Audit „Hochwasser – wie gut sind wir vorbereitet“ 

– ein strukturiertes Verfahren für Kommunen zur 

Verbesserung des kommunalen Risikomanagements 

Paul Geisenhofer, Dipl.-Ing.¹ 

ABSTRACT 

"Flood-Audit" is a service provided by the German Association for Water, Wastewater and 

Waste (DWA) to help local authorities achieve improvements in local flood precaution. 

"Flood-Audit" is a tool to raise local awareness and systematical knowledge about different 

aspects of risk management. It's not about "checking up" on people it's about giving advice. 

The assessment regards 35 indicators in seven fields of action. The ranges of action have to 

be assessed separately for river floods and flash floods and the three flood scenarios according 

to the European Floods Directive. To gain an overview, the indicators are summarized and 

visualized in two "precaution traffic lights". Each shows four sectors: spatial prevention, 

constructive prevention, precautionary behavior and risk prevention. This way the strengths 

and weaknesses of "flood preparedness" become powerfully visible for citizens and especially 

the elected municipal officers. This provides a solid basis for future precaution actions. 

So far about 30 audits were successfully carried out in different landscapes. Individual results 

are presented using the example of one Bavarian alpine municipality. 


Das Hochwasser-Audit ist ein Angebot der Deutschen Vereinigung für Wasserwirtschaft, Abwasser 

und Abfall (DWA) für Gemeinden zur nachhaltigen Verbesserung der lokalen Hochwasservorsorge. 

Ziel des Audits ist es, in einem vertrauensvollen Beratungsprozess (nicht 

Prüf-Prozess) unabhängig von aktuellen Schadensereignissen den Status der Hochwasservorsorge 

allumfassend zu erheben, das Risikobewusstsein zu schärfen und die Risikomanagementplanung 

systematisch zu vertiefen. Bewertet werden insgesamt 35 Indikatoren in sieben 

Handlungsbereichen der nicht-baulichen Hochwasservorsorge (non-structural Meassures), 

getrennt für Flusshochwasser und Sturzfluten und jeweils für die drei Hochwasserszenarien 

HQ häufig 

, HQ 100 

und HQ extrem 

. Um eine plakative Gesamtübersicht über den umfangreichen 

Ergebnisbericht zu erhalten, werden die Indikatoren zu zwei Hochwasservorsorge-Ampeln 

zusammengefasst. Jede Ampel bildet vier Bewertungssektoren ab: Flächenwirksame Vorsorge, 

Bauvorsorge, Verhaltenswirksame Vorsorge und Risikovorsorge. Stärken und Schwächen in 

1 Watermanagement Office Rosenheim, Rosenheim, Bavaria GERMANY, paul.geisenhofer@wwa-ro.bayern.de; 

paul.geisenhofer@web.de 



der Hochwasservorsorge werden so für die Bürgerschaft und vor allem die politischen 

Mandatsträgern eindringlich sichtbar und bilden die Basis für weiteres ziel gerichtetes 

Vorsorgehandeln. Bisher wurden rund 30 Audits in unterschiedlichen Naturräumen 

durchgeführt. Die Ergebnisse werden von den Gemeinden durchwegs positiv bewertet. 

Einzelne Ergebnisse werden am Beispiel einer bayerischen Alpengemeinde vorgestellt. 

KEYWORDS 

flood-audit; risk management; municipalities; non-structural meassures 

EINFÜHRUNG 

Seit Jahrzehnten wurden in Mitteleuropa größte Anstrengungen und Investitionen unternommen, 

um Siedlungen und Infrastruktur durch technische Maßnahmen vor Hochwasser 

zu schützen. Dennoch entstehen nach wie vor bei Hochwasser häufig immense Schäden und 

noch während des Hochwasserereignisses beginnt in den Medien die Suche, wer schuld daran 

sei, dass der (vielleicht sogar neu gebaute) Deich überströmt wurde. Trotz der Warnungen 

der Fachleute herrscht in der Öffentlichkeit nach wie vor die Illusion vor, dass Hochwasser 

beherrschbar sei und Schutzanlagen absolute Sicherheit schaffen. Gleichwohl hat schon 1995 

die Deutsche Länder-Arbeitsgemeinschaft Wasser (LAWA) mit ihren "Hochwasser-Leitlinien" 

für Klarheit gesorgt und dem technischen Hochwasserschutz als nur einer Komponente eines 

umfassenden Risikomanagements den angemessenen Stellenwert zugewiesen, der ihm 

gebührt. Mit Erlass der EU-Hochwasserrisikomanagement-Richtlinie wurde auch die Rechtsgrundlage 

für den Paradigmenwechsel vom Hochwasserschutz (mit Absolutheitsanspruch) 

zum Risikomanagement geschaffen. 

Risikomanagement ist nicht mehr in erster Linie die Aufgabe der Wasserwirtschaftsverwaltung 

Hochwasser fernzuhalten bzw. zu reduzieren, sondern vielmehr Aufgabe der Gesellschaft 

Hochwasserschäden gering zu halten. Dabei sind Extremabflüsse, die den technischen 

Hochwasserschutz überlasten können, regelmäßig mit zu betrachten. Schäden können in 

diesen Fällen nur noch in der überschwemmungsgefährdeten Fläche hinter den Schutzanlagen 

reduziert werden. Damit haben Kommunen mit ihrer Planungshoheit eine hohe 

Verantwortung aber auch vielfältige Handlungsmöglichkeiten. An dieser Stelle entsteht der 

Ruf nach einem Instrument, das Kommunen hilft, dieser Verantwortung gerecht zu werden 

und ihren individuellen Handlungsbedarf zu erkennen. 

Mit dem "Audit Hochwasser – wie gut sind wir vorbereitet" hat die deutsche Vereinigung für 

Wasserwirtschaft Abwasser und Abfall (DWA) ein solches Instrument zur Verfügung gestellt. 

WESEN UND INHALT DES AUDITS 

Allgemeines 

Das Audit „Hochwasser – wie gut sind wir vorbereitet“ ist ein Angebot speziell für kommunale 

Gebietskörperschaften zur nachhaltigen Verbesserung der lokalen Hochwasservorsorge. 

Es bietet der Gemeinde unabhängig von aktuellen Schadensereignissen eine Plattform, den 


Status ihrer Hochwasservorsorge ohne Zeitdruck umfassend zu erheben und in breitem 

Konsens zu begründeten und priorisierten Handlungskonzepten zu kommen. 

Das Hochwasser-Audit versteht sich als Hilfsmittel in einem Beratungsprozess (nicht Prüf- 

Prozess), der das Ziel hat, das lokale Risikobewusstsein zu schärfen und die Themen der 

Risikomanagementplanung systematisch zu vertiefen. Wesentliche Grundlage ist das gegen - 

seitige Vertrauen zwischen Auditor und Gemeinde. Als Auditoren werden von der DWA 

speziell geschulte Fachkräfte der Wasserwirtschaft beauftragt. 

Mit dem Audit schafft die Gemeinde auch eine hervorragende Grundlage für die von der 

EU geforderte öffentliche Kommunikation von Hochwasserrisiken. 

Inhalte des Hochwasser-Audits 

Allgemeines und Bezugsraum 

Das Hochwasser-Audit stellt die Informationslage aller Beteiligten über die Risiken und die 

möglichen Maßnahmen zu ihrer Verminderung in den Mittelpunkt. Bewertet wird in erster 

Linie die Güte der Information über die Risiken, nicht das Risiko selbst. Hintergrund dieser 

Schwerpunktsetzung ist die Erwartung, dass gut informierte Verwaltungen und Bürger sich 

richtiger verhalten werden, wenn ihnen aussagekräftige Informationen und praktikable 

Lösungen zur Risikoreduktion zur Verfügung stehen. 

Bezugsraum des Audits ist die Risiko- und Verantwortungsgemeinschaft einer Kommune im 

gesamten Gemeindegebiet ungeachtet tatsächlicher Sach- und Fachzuständigkeiten anderer 

Stellen. Das Audit konzentriert sich dabei auf die Bewertung lokal zu verantwortender 

Maßnahmen der nicht technischen Hochwasservorsorge (non-structural measures). Maßnahmen 

des technischen Hochwasserschutzes (Deiche, Rückhaltebecken etc.), werden als 

gegebene Randbedingungen z.B. für die Beschreibung von Risiken berücksichtigt, sind aber 

nicht Gegenstand der Bewertung des Audits. 

Szenarien 

Das Audit thematisiert neben den Risiken aus Flusshochwasser auch das Risiko von lokalen 

Sturzfluten, die häufig und insbesondere im alpinen Raum von mindestens ebenso großer 

Bedeutung sind. Gleichwohl steht eine konsequente planerische bzw. logistische Auseinandersetzung 

mit Sturzfluten bei Gemeinden oft noch im Hintergrund. Aufgrund der unterschiedlichen 

Randbedingungen und auch Handlungsinstrumente werden die Szenarien 

Flusshochwasser und Sturzfluten im Audit getrennt betrachtet. 

Entsprechend der EU-Hochwasserrichtlinie werden alle Betrachtungen jeweils für den Lastfall 

häufiges Hochwasser (HQ häufig 

), Hochwasser mit mittlerer Wahrscheinlichkeit (HQ 100 

) und 

außerordentlich seltenes Hochwasser (HQ extrem 

) durchgeführt. Somit ergeben sich insgesamt 

6 Szenarien. 

Bewertungsfelder, Handlungsbereiche, Indikatoren 

Auch die Struktur des Hochwasseraudits orientiert sich wiederum an den anerkannten 

Strukturen der LAWA-Leitlinie bzw. der EU-HWRM-RL. So werden zunächst die vier 

Bewertungsfelder 


– Flächenwirksame Vorsorge 

– Bauvorsorge 

– Verhaltenswirksame Vorsorge und 

– Risikovorsorge 

definiert. Diesen werden sieben Handlungsbereiche der nicht-baulichen Hochwasservorsorge 

zugeordnet, die dem im Bereich der Naturgefahren allg. anerkannten Risikokreislauf 

entnommen sind. 

Abbildung 1: Das Arbeitsfeld des Hochwasser-Audits im Risikokreislauf sind die nicht-technischen Handlungsfelder der Hochwasservorsorge 

(DWA) 

Den Handlungsbereichen wiederum werden die der Auditierung zugrundeliegenden 

Indikatoren zugeordnet. Bei Bedarf wird ein Indikator durch mehrere Merkmale beschrieben. 

Insgesamt ergeben sich 35 konkrete Fragestellungen. 

Bewertungsmethoden 

Die Bewertung zielt darauf ab, den Vorsorge-Status der Gemeinde übersichtlich und kompakt 

aber dennoch ausreichend differenziert und aussagekräftig darzustellen. Gegenstand der 

Bewertung sind die 35 Einzelmerkmale in den sieben Handlungsbereichen. Für jedes 

Merkmal werden getrennt für Flusshochwasser und Sturzfluten die drei Hochwasserszenarien 

untersucht. 

Die Einzelmerkmale werden formalisiert mit Punkten bewertet und für die vier Bewertungsfelder 

für jedes Szenario zu einem Gesamtergebnis aggregiert. Im Sinne einer einheitlichen 


Handhabung stehen dafür jeweils maximal 250 Punkte zur Verfügung, die mit einer vorgegebenen 

Gewichtung auf die Handlungsbereiche innerhalb des Feldes verteilt werden. 

Einzelmerkmale innerhalb eines Handlungsbereiches erhalten jeweils die gleiche Maximalpunktzahl. 

Eine stärkere Differenzierung würde sicherlich zu keinem aussagekräftigeren 

Ergebnis führen. Dieses pragmatische Bewertungssystem ist insbesondere deshalb vertretbar, 

weil die 4 Bewertungsfelder getrennt betrachtet werden und keine Gesamtsumme gebildet 

wird. 

Die Bewertungskriterien für alle Einzelmerkmale sind in einem Handbuch beschrieben, das 

von den Auditoren verbindlich anzuwenden ist. Dem Charakter des Audits als einem von 

Vertrauen getragenem Beratungsprozess innerhalb einer Kommune entsprechend ist eine 

Weitergabe und ein Vergleich von Auditergebnissen generell nicht vorgesehen. 

Eine Besonderheit des Bewertungssystems ist noch erwähnenswert: Von der Gemeinde 

konkret geplante aber zum Zeitpunkt des Audits noch nicht durchgeführte Maßnahmen 

werden vorab mit der halben Punktzahl gewertet. Damit soll nicht nur ein gewisser Handlungsanreiz 

geschaffen, es soll auch deutlich werden, dass das Audit der Einstieg in einen auf 

Dauer angelegten Prozess des Risikomanagements ist und eben nicht nur eine Bewertung 

zu einem bestimmten Stichtag im Sinne von „bestanden“ oder „durchgefallen“. Werden die 

geplanten Maßnahmen durchgeführt, erhält die Gemeinde beim Folgeaudit die volle Punktzahl, 

andernfalls keine Punkte. 

Darstellung der Ergebnisse - Dokumentation 

Im Rahmen eines Audits wird von der Gemeinde eine Vielzahl von Informationen zusammengetragen 

und dem Auditor präsentiert. Dessen Aufgabe ist es, die Bewertung durchzuführen 

und die Ergebnisse für unterschiedliche Zwecke geeignet darzustellen. Er fertigt dazu 

einen Bericht, in dem der Status aller Einzelmerkmale im Detail beschrieben und für alle 

Szenarien bewertet wird. Die Bewertungen werden außerdem tabellarisch zusammengefasst. 

Der Bericht hebt auch die von der Gemeinde geplanten Maßnahmen hervor. Diese können 

in der Gemeinde anschließend als Maßnahmenkatalog mit Terminen und Verantwortlichen 

in Eigenregie weiter bearbeitet werden. 

Eine weitere besonders plakative Ergebnisdarstellung sind die sog. Hochwasserampeln. Es 

handelt sich um Kreisdiagramme mit vier Quadranten entsprechend der Bewertungsfelder. 

Die drei Hochwasserlastfälle werden als konzentrische Ringe dargestellt. Die jeweils erreichte 

Punktzahl wird einem Farbsystem zugeordnet, das von grün („alle Hausaufgaben sind 

gemacht“) über orange und gelb nach rot („Vorsorgewüste, noch keine tragenden Ansätze 

erkennbar“) reicht. Weil die Ergebnisse für Flusshochwasser und für Sturzfluten oft sehr 

unterschiedlich sind, werden zwei Ampeln erstellt. 

Während der ausführliche Bericht vor allem für die Verwaltungsebene der Gemeinde wertvoll 

ist, bietet das Ampelsystem der Verwaltungsspitze, den politischen Mandatsträgern und der 

Öffentlichkeit die gewünschte Übersicht mit einer allgemein verständlichen Symbolik. Die 

Ampeln machen Stärken und Schwächen in der Hochwasservorsorge sichtbar und können 

so der Ausgangspunkt für weiteres zielgerichtetes Vorsorgehandeln sein. 


Abbildung 2: Hochwasserampeln – eine Stärken-Schwächenanalyse auf einen Blick für Bürgermeister, politische Mandatsträger und 

Öffentlichkeit (DWA) 

ABLAUF DES AUDITS 

Anfrage, Beauftragung 

Ein Hochwasser-Audit wird von der interessierten Gemeinde bei der DWA zunächst formlos 

angefragt. Die DWA übermittelt daraufhin ein Angebot über die im Audit angebotenen 

Dienstleistungen. Wird der Auftrag erteilt, beauftragt die DWA einen federführenden Auditor 

und falls erforderlich einen Co-Auditor. 

Vorbereitung 

Der Auditor stimmt mit der Gemeinde die zu beteiligenden Verwaltungseinheiten und den 

Audit-Termin vor Ort ab. Durch ein Vorgespräch kann sich die Kommune sachgerecht auf 

das Audit vorbereiten. Hierzu ist es hilfreich, wenn die Kommune Informationen zu den 

Indikatoren anhand einer Check-Liste zusammenstellt und – soweit vorhanden – Planmaterial 

und insb. Hochwassergefahrenkarten bereithält. 

Notwendige Unterlagen 

Im Rahmen eines Erst-Audits werden keinerlei spezifische Unterlagen zur Einsicht gefordert 

oder gar überprüft. Das heißt, dass die Gemeinde ihr individuelles Auditergebnis allein 

aufgrund ihrer eigenen Informationen generiert – je zutreffender und realistischer diese 

Informationen sind, desto zutreffender und „wertvoller“ wird auch das Auditergebnis sein. 

Bei Folge-Audits werden dann die maßnahmenbezogenen Unterlagen und Dokumentationen 

des Erstaudits als Nachweis erforderlich sein, ob geplante Maßnahmen tatsächlich durchgeführt 

wurden. 


Ablauf des Audits vor Ort 

Das Audit wird in der Regel an zwei Tagen vor Ort durchgeführt und protokolliert. Es läuft als 

Gespräch der Auditoren mit den Vertretern des Auftraggebers ab, die für die verschiedenen 

Aufgabenfelder der Hochwasservorsorge verantwortlich sind. Als Teilnehmer des Auftraggebers 

kommen Entscheider und Fachkräfte der folgenden Bereiche in Frage: 

– Ordnungsamt, Verwaltungsspitze; 

– Hochbau, Tiefbau, Bauverwaltung, Bauordnung, Baurecht, Liegenschaftsverwaltung; 

– Umweltverwaltung, Wasserwirtschaftsverwaltung, Grün- und Freiflächenverwaltung, 

Forstbehörde; 

– Wirtschaftsförderung, Gewerbeordnung; 

– Gesundheitsbehörden; 

– Katastrophenschutz, Feuerwehr, Rotes Kreuz; 

– Denkmalschutzverwaltung; 

– Weitere nach Einschätzung des Auftraggebers 

Protokoll, Urkunde, Folgeaudit 

Eine erste Erfassung der Bewertungspunkte wird den Teilnehmern bereits am Ende des Vor- 

Ort-Termins vorgestellt. Die ausführliche Dokumentation wird der Gemeinde vom Auditor 

zur Schlussabstimmung zugeschickt. Zum Abschluss des Audits erhält die Kommune von der 

DWA-Bundesgeschäftsstelle die Reinschrift der Dokumentation des gesamten Auditprozesses 

und eine Urkunde. Dem prozessunterstützenden Charakter des Audits folgend sollte 

spätestens nach 6 Jahren ein Folgeaudit durchgeführt werden, um den Fortschritt in der 

Hochwasservorsorge zu dokumentieren. 

Das Audit selbst gibt keine konkreten Maßnahmenempfehlungen. Die DWA wird die Weiterentwicklung 

der Hochwasservorsorge jedoch durch eine Liste von Maßnahmen unterstützten, 

die sich in der Praxis besonders bewährt haben. 

PRAXISERFAHRUNGEN 

Bisher wurden ca. 30 Audits in unterschiedlichsten Naturräumen Deutschlands (Flachland, 

Mittelgebirge, alpiner Raum) durchgeführt. Dabei wurden Kommunen zwischen ca. 1500 

und 1 Mio. Einwohner auditiert. Die Erfahrung in der Abwicklung zeigte, dass der Aufwand 

sowohl auf der Seite der Auditoren als auch der Kreis der beteiligten Mitarbeiter der 

Gemeinde mit der Gemeindegröße zunimmt. Vor Ort sind i.d.R. zwei Tage ausreichend. 

Generell war festzustellen, dass die Ergebnisse beim Szenario Flusshochwasser meist besser 

sind als beim Szenario Sturzfluten, das erst in der jüngeren Vergangenheit (vielleicht auch 

bedingt durch die Klimadiskussion?) an Bedeutung gewonnen hat. 

Die größten Vorsorgedefizite sind erwartungsgemäß beim Szenario HQ extrem 

festzustellen. 

Hier spiegelt sich noch eindrucksvoll wieder, dass das Denken im Hochwasserschutz früher 

regelmäßig beim HQ 100 

endete. 


Regelmäßig kündigen die Gemeinden angesichts der im Audit erkannten Defizite zahlreiche 

Vorsorgemaßnahmen an, die der Bewertungssystematik folgend mit der halben Punktzahl ins 

Ergebnis eingehen. Die Zukunft wird zeigen, was aus den guten Vorsätzen geworden ist. 

Häufige Einschätzung der Gemeinden war, dass allein schon die Durchführung des Audits vor 

Ort bereits erheblich zur Sensibilisierung der Verantwortlichen beigetragen habe. Außerdem 

werden vor allem in den Bereichen Informationsgewinnung und Informationswege aufgrund 

der vollständigen Betrachtungsweise von 35 Indikatoren wertvolle neue Erkenntnisse 

gewonnen. In den Bereichen Bauordnungs- und Bauplanungsrecht werden vielfach 

Vorsorgemöglichkeiten erst erkannt. 

Nachfolgend einige Ergebnisse einer alpinen Gemeinde mit 4500 Einwohnern, die durch das 

Hochwasser eines Sees sowie durch Wildbäche bedroht ist. Es hat sich dort als zweckmäßig 

erwiesen, die Wildbachgefahren dem Szenario Sturzfluten zuzuordnen. Originalzitate aus 

dem Auditbericht werden kursiv wiedergegeben: 

– In der Gemeinde gibt es zwar bereits eine Beratung der Bauherren, die Gemeinde hat aber 

erkannt, dass weit mehr Informationsmaterial zum Thema Bauvorsorge verfügbar ist und 

wird „auf der Homepage einen separaten Informationsbereich zum Thema Hochwasserangepasste 

Nutzung / Objektschutzmaßnahmen einrichten und dort die einschlägigen 

Broschüren gebündelt bereitstellen. Sie wird ihr Beratungsangebot entsprechend 

auswei ten.“ Allgemeine Informationen und eine spezielle Bauherrenberatung sind auch 

zum Thema Sturzfluten geplant. 

– Die Gemeinde erkannte, dass über die HQ 100 

-Linie hinaus die Betroffenheit von Gewerbeund 

Kurbetrieben nicht bekannt ist. „Die Kommune wird daher prüfen, ob für den Fall von 

HQ extrem 

weitere Betriebsstätten betroffen sein könnten und deren Notfallpläne anzu passen 

wären.“ 

– Gefahrenbereiche von Sturzfluten sind in der Gemeinde bisher nicht bekannt. Eine 

Grundlage im Sinne von Fließwegeplänen und Flächenausweisungen, die potentielle 

Gefahrenbereiche aufzeigen, gibt es noch nicht. „Die Kommune wird solche Planungsgrundlagen 

(Fließwegepläne) für die 3 Wildbäche erarbeiten (lassen) und Auswertungen 

vornehmen, in welchen Bereichen die menschliche Gesundheit akut gefährdet sein kann.“ 

– Die Gefahrenbereiche der Wildbäche konnten – da nicht bekannt – bisher nicht in 

Bauleitplänen dargestellt werden. „Für das Szenario HQ häufig 

ist dies voraussichtlich 

verzichtbar. Für HQ 100 

und HQ extrem 

wird die Verwertung der Erkenntnisse in der Bauleitplanung 

von der Kommune eingeplant.“ 

– Die Gemeinde hat außerdem erkannt, dass aufgrund der kleinräumigen Wildbachereignisse 

und der sehr kurzen Vorwarnzeiten weder eine zutreffende Hochwasservorhersage noch 

eine gegenüber dem derzeitigen Zustand verbesserte Vorwarnung der Bürgerschaft möglich 

ist. „In der Kommune wird … die Auswertung von Unwetterwarnungen des Deutschen 

Wetterdienstes bereits verfolgt; die Feuerwehr erhält entsprechende Unwetterwarnungen. 

Hier besteht ein Vorlauf von etwa 1 Stunde, die verfügbare Zeit ist also nicht ausreichend, 

um auch Bürger zu erreichen. Von der Kommune sollte deshalb in allgemeiner Form auf die 


Starkniederschlagsthematik und auch auf die Eigenverantwortung hingewiesen werden 

insbesondere in der Form, dass die Bürger die bestehenden Unwetterwarnsysteme 

beispielsweise der Versicherer nutzen.“ Da bei diesen Merkmalen keine weitergehende 

Vorsorge möglich ist, wurde die volle Punktzahl vergeben, damit nicht der Eindruck 

entsteht, dass es Handlungsbedarf geben könnte. 

FAZIT 

Das Hochwasser-Audit der DWA kann als eine der ersten vorbereitenden Maßnahmen einer 

wirkungsvollen Hochwasserrisikomanagementplanung eingesetzt werden. Die Gemeinde 

erhält eine individuelle, sorgfältig dokumentierte und strukturierte Defizitanalyse über 

35 Indikatoren, bei der auch die individuelle Verwaltungsstruktur berücksichtigt wird. Darin 

enthalten ist auch eine Dokumentation aller geplanten Maßnahmen einschließlich eines 

Handlungszeitraums und der in der Gemeinde zuständigen Akteure. Analyse und Maßnahmen 

befassen sich nicht nur mit Flusshochwasserereignissen sondern auch mit Starkregenereignissen, 

die in besiedelten Bereichen immer mehr in den Vordergrund treten. 

Die in der Gemeinde geplanten Maßnahmen können nach dem Audit in einem kontinuierlichen 

Prozess weiter bearbeitet werden. Die Durchführung des Audits ist ein Beratungsprozess 

bei dem die Gemeinde auf neue Handlungsfelder und geeignete Maßnahmen aufmerksam 

gemacht wird, über deren Zweckmäßigkeit entscheiden und diese an die individuelle 

Situation der Gemeinde anpassen kann. Mit den sog. „Hochwasserampeln“ erhält der 

Bürgermeister einen plakativen, aber trotzdem aussagekräftigen Überblick über die Stärken 

und Schwächen der Gemeinde beim Risikomanagement und somit auch über die vordringlichsten 

Handlungsfelder. 

Das „Audit Hochwasser – wie gut sind wir vorbereitet?“ leistet somit auf kommunaler Ebene 

einen entscheidenden Beitrag zu einer nachhaltigen Vorsorge. 

LITERATUR 

- LAWA (1999): Leitlinien für einen zukunftsweisenden Hochwasserschutz 

- EU (2007): Richtlinie über die Bewertung und das Management von Hochwasserrisiken 

(2007/60/EG) 

- DWA (2010): Merkblatt M 551 Audit „Hochwasser – wie gut sind wir vorbereitet“ 

- DWA (2014 – 2015): Div. Mitteilungen aus dem DWA-Arbeitskreis 4.6 Audit-Hochwasser 



Risk-Dialogue – a challenge 

Risiko-Dialog als Herausforderung 

Rene Graf, Dipl. Ing. ETH¹ 

ABSTRACT 

A successful Risk-Dialogue between experts and persons at risk fosters the quality of decisions 

and raises their acceptance. This article contributes to successful dialogues by combining 

observations from practice with findings from social sciences. The focus concentrates on 

a diverging notion of risk and a tension felt between science and subjective perception. 

Risk-experts try to mitigate loss whereas people at risk tend to optimise nature risks and 

integral life-opportunities. Their behaviour is influenced by external reasons like change 

and their individual risk-mentality. They can't always be convinced by "objective", scientific 

findings 

as on the one hand these results represent simplifications, on the other hand they don't 

correspond to the constructivist perception of the dialogue-partners. The chance for experts 

to approve their communication is not to perfect their scientific argumentation. It's the 

interest for the actors and their decision-making structure. Empathy cannot be acquired 

but in courses: It is necessary to get to know unconversant ways of living directly. Concrete 

suggestions can be found at the end of this article. 


Ein erfolgreicher Risiko-Dialog zwischen Fachleuten und Betroffenen trägt zur Qualität von 

Entscheidungen bei und erhöht deren Akzeptanz. Dieser Beitrag liefert Anregungen zur 

erfolgreichen Gestaltung von Dialogen, indem er Beobachtungen aus der Praxis mit Erkenntnissen 

aus den Sozialwissenschaften verknüpft. Im Mittelpunkt stehen Überlegungen zu 

einem divergierenden Risiko-Begriff und zum Spannungsfeld Wissenschaftlichkeit versus 

subjektive Wahrnehmung. Risiko-Experten wollen primär Schäden abwehren, Betroffene 

streben nach einer Optimierung von Risiken aus Naturgefahren und Chancen für ihr 

integrales Leben. Ihr Verhalten unterliegt äusseren Einflüssen wie dem Wandel sowie einer 

persönlichkeitsimmanenten Risikomentalität. "Objektive", wissenschaftliche Erkenntnisse 

vermögen nicht immer zu überzeugen, da diese einerseits Vereinfachungen darstellen, 

andererseits nicht der konstruktivistischen Wahrnehmung der Gesprächspartner entsprechen. 

Das Potenzial für Experten, ihre Chancen im Dialog zu verbessern, liegt somit nicht in der 

Perfektionierung ihrer wissenschaftlichen Argumentation, sondern in ihrem Interesse für die 

Akteure und deren Entscheidungssituation. Empathie kann aber nicht nur in Kursen erlernt 

werden, sondern erfordert auch, sich "live" mit Lebensweisen zu befassen, die einem nicht 

vertraut sind. Konkrete Empfehlungen, wie dies geschehen kann, finden sich am Schluss des 

Beitrags. 

1 Verkehr und Infrastruktur Kanton Luzern, Kriens 2 Sternmatt, SWITZERLAND, rene.graf@lu.ch;rene.graf@riskcoach.ch 



KEYWORDS 

riskmanagement; communication; risk perception; social sciences 

EINLEITUNG 

Fachkräfte für Naturgefahren erarbeiten Grundlagen und Projekte von hoher Qualität. Trotzdem 

stossen diese bei Betroffenen zuweilen auf Ablehnung. Bei der Umsetzung fällt deshalb 

dem Dialog eine Schlüsselrolle zu. Viele Fachleute meistern diese Klippe gut, anderen bereitet 

sie Mühe. In Diskussionen mit Ingenieuren und Naturwissenschaftlern erhielt der Autor den 

Eindruck, Erkenntnisse aus den Sozialwissenschaften würden zuwenig wahrgenommen. Oft 

ging es dabei um den Begriff "Risiko" oder um den Umgang mit Unschärfen und Unsicherheiten. 

In diesem Praxisbeitrag werden alte und neue Erkenntnisse in einen Zusammenhang 

gestellt, der einen anregenden Blick über den Zaun der eigenen Kernkompetenzen ermöglicht. 

Er soll Naturgefahren-Fachleuten helfen, Gespräche noch erfolgreicher zu gestalten. 

BEDEUTUNG DES RISIKO-DIALOGS 

Risikomanagement "umfasst die laufende systematische Erfassung und Bewertung von 

Risiken sowie die Planung und Realisierung von Massnahmen" (PLANAT 2015). 

Die Erfassung von Risiken ist eine Domäne der Fachleute: Die auftraggebende Instanz grenzt 

das System ab. Innerhalb dieser inhaltlichen und räumlichen Grenzen folgen Identifikation 

und Analyse der Risiken aber fachtechnischen Fragestellungen. Anders die Bewertung der 

Risiken: Das Aushandeln der Grenzen kollektiver Akzeptanz - was darf passieren, was nicht? 

- ist nicht ein fachtechnischer, sondern ein politischer Vorgang: Individuen und Gruppen 

suchen einen Konsens, der die gesellschaftspolitischen Realitäten spiegelt. Bei der Wahl der 

Massnahmen kommt den Experten wieder eine stärkere Bedeutung zu: Sie schlagen 

mögliche Lösungen vor und liefern Aussagen zu deren Wirkungsgrad, Umweltverträglichkeit, 

Kosteneffizienz usw. 

Ein fruchtbarer Dialog zwischen Fachleuten und Betroffenen trägt einerseits zur Qualität der 

Entscheidungen bei, andererseits erhöht er deren Akzeptanz. In der Schweiz gewährt der 

Bund höhere Beiträge an die Erstellung von Schutzbauten, wenn die Bauherrschaft Mehrleistungen 

für ein partizipatives Vorgehen nachweisen kann (BAFU 2012). 

DIVERGENTE RISIKOBEGRIFFE 

Das Bundesamt für Umwelt definiert Risiko als "Grösse und Wahrscheinlichkeit eines 

möglichen Schadens" (BAFU 2015). Naturgefahren-Experten sind dafür ausgebildet und 

angestellt, Risiken zu vermindern. Aus dieser Optik erscheint Risiko als etwas Unerwünschtes. 

Aber Risiko ist auch etwas Essentielles, für das menschliche Leben Unverzichtbares: Was 

immer ein Mensch unternimmt kann in einen Erfolg münden oder eben mit einem Schaden 

enden. So betrachtet gleicht Risiko einer Währung, die ein Mensch investieren muss, wenn er 

etwas erreichen will. Walter Munter, Risikoforscher und Lawinenexperte, sagt, das Risiko sei 

dazu da, dass der Mensch seine Fähigkeiten überhaupt entwickeln könne (KNECHT 2015). 

Lessing bestätigt: „Es wäre wenig in der Welt unternommen worden, wenn man immer nur 


auf den Ausgang gesehen hätte.“ (LESSING zit. 2014) Diese intuitive Wahrnehmung von 

Risiko durch Betroffene erzeugt eine Spannung zum Grundauftrag von Naturgefahren-Fachleuten, 

Risiken zu vermindern: Die Betroffenen gehen - oft unbewusst - Wagnisse ein mit 

dem Ziel, im Leben zu reussieren. 

Zu Wagemut ruft bereits die Bibel auf: Ein Mann geht auf Reisen. Sein Vermögen vertraut 

er seinen Dienern an. Nach seiner Rückkehr lobt er die beiden Diener, die etwas gewagt und 

sein Geld vermehrt haben. Den dritten Diener jedoch lässt er hart bestrafen, weil dieser kein 

Risiko eingehen wollte und das ihm anvertraute Geld vergraben hatte. Zaudern gilt bis heute 

als unattraktiv: „Don’t be a Maybe!“ lautet ein Werbespruch für Zigaretten. Ein Gerichts - 

urteil bestätigt, dass ein gewisses Risiko "sozialadäquat" sei: Eine Frau war in einem Selbstbedienungsrestaurant 

auf einer Kartoffel ausgeglitten war und hatte sich dabei den Arm 

gebrochen. Das Gericht wertete den Nutzen einer effizienten Verpflegungsmöglichkeit höher 

als das Risiko, sich dabei zu verletzen. (OBERGERICHT KT. ZÜRICH 2014) 

Die divergierenden Prämissen wirken sich im Risiko-Dialog aus: Experten tendieren darauf, 

den "Preis" möglichst tief zu halten. Betroffene dagegen versuchen, den "richtigen Preis" zu 

finden, um Chancen und Risiken über das diskutierte Naturgefahren-Problem hinaus in der 

Balance zu halten. Dieses "Optimierungsprogramm" läuft im Hintergrund immer mit - unbewusst, 

aber mit einer dem Leben verbundenen Kraft. 

EINFLÜSSE AUFS RISIKOVERHALTEN 

Ein Individuum verhält sich nicht in jeder Situation gleich risikofreudig. Wissenschaftler 

weisen aber auf Einflüsse hin, die die Risikofreudigkeit grundsätzlich beeinflussen können: 

Einen möglichen Einfluss von aussen benennt der Historiker Philipp Blom: Wandel kann 

dazu führen, dass das Leben mit einer neuen, existenziellen Unsicherheit verbunden wird. 

Blom erklärt dies am Beispiel eines südfranzösischen Bauern anfangs des 20. Jahrhunderts: 

Dieser hat den gleichen Beruf wie sein Vater und sein Grossvater. Er heisst gleich, wohnt im 

selben Haus, gehört derselben Konfession an und hört auf die gleichen Respektpersonen. 

Die wirtschaftliche Lage zwingt ihn, nach Paris zu ziehen und Fabrikarbeiter zu werden. 

Damit verliert er den vertrauten Boden unter seinen Füssen. Jeder Entscheid wird damit 

zu einem Wagnis (BLOM 2014). 

Auf persönlichkeitsimmanente Einflüsse verweist Michael Hampe. In Anlehnung an den 

Philosophen Nicholas Rescher beschreibt er drei Mentalitäten, die bereits mit einfachen 

psychologischen Tests unterscheidbar seien (ARNSWALD/SCHÜTT 2011): 

– Risiko-Vermeider betrachten Risiken als Gefahren, die man neutralisieren muss. 

Diese Haltung deckt sich oft mit dem Berufsverständnis von Naturgefahren-Experten. 

– Risiko-Kalkulierer versuchen, Gefahren gegen Nutzen abzugleichen. 

– Risiko-Sucher sehen im Unvorhersehbaren Chancen für wünschenswerte Veränderungen. 


"Der springende Punkt bei Hampes Überlegungen ist, dass jede der drei Mentalitäten für sich 

vernünftig ist, dass sie einander aber irrational erscheinen, wenn man sie (…) miteinander 

konfrontiert." (TOBLER 2014) 

Daraus folgert für den Dialog: 

– Kommt z.B. die Information, in einem gefährdeten Gebiet zu leben, überraschend, können 

die Betroffenen dies als Wandel ihrer Lebensbedingungen empfinden. Ihr Verhalten kann 

von einer Verunsicherung geprägt werden, die über das formale Thema eines Gesprächs 

hinaus reicht. 

– Treten Differenzen auf, sind diese möglicherweise nicht (nur) sachlich bedingt, sondern 

auch durch Persönlichkeitsmerkmale oder das Berufsverständnis von Experten. 

Im Gespräch ist es deshalb wichtig, auf Hinweise zu achten, wonach das Verhalten der 

Betroffenen möglicherweise nicht primär von Sachfragen geprägt ist sondern durch andere, 

der kognitiv-rationalen Wahrnehmung schwer zugängliche Einflüsse. 

WISSENSCHAFTLICHKEIT VS. ERLEBEN 

In langjähriger Berufstätigkeit traf der Autor immer wieder Fachleute, die Mühe hatten, 

natur- und ingenieurwissenschaftlich hergeleitete, aber mit Unschärfe behaftete Erkenntnisse 

so umzusetzen, dass sie als Diskussionsbasis taugten. So weigerte sich ein Experte, die Pixel 

einer Murgangmodellierung zu Wirkungsräumen zu arrondieren: Eine derart vereinfachte 

Intensitätskarte würde die gewonnenen Erkenntnisse nicht genau genug wiedergeben. 

Verwandt damit ist das Festhalten an hochexakten hydrologischen Kennziffern oder 

metergenau abgegrenzten Gefahrengebieten. Ein solches berufliches Selbstverständnis deckt 

sich mit der Ansicht von Nobelpreisträger Francis Crick, wonach die naturwissenschaftlich-experimentelle 

Methodik „die einzig sinnvolle Herangehensweise an das Bewusstsein“ 

sei. Die Naturwissenschaft habe noch immer Recht behalten. (BECKER 2009) Naturwissenschaftlich-mathematische 

Methoden sind unverzichtbar, Statistik etwa zur Auswertung von 

Messreihen, Algorithmen für Modellierungen. So gewonnene Erkenntnisse müssen aber 

immer kritisch gewürdigt werden. Internetpionier Jaron Lanier: „Wir Menschen sind Genies 

darin, uns durch den Gebrauch von Computern verwirren zu lassen. Das beste Beispiel dafür 

ist, dass Computer so tun, als wäre Statistik eine adäquate Beschreibung der Realität. (…) 

Es gibt eine allgemeine statistische Vorhersehbarkeit, aber sie gilt nur für begrenzte Zeitabschnitte, 

und ihre Beschränkungen lassen sich nicht universell vorhersagen.“ (LANIER 2014) 

Mathematisch-naturwissenschaftliche Methoden helfen, eine Problematik besser zu verstehen. 

„Damit wird zwar immer feiner berechenbar, was berechenbar ist. Aber dem Erleben des 

Einzelnen kommt man damit nicht wirklich näher“ betont Psychiatrieprofessor Daniel Hell 

(HELL 2014). Ein Beispiel: 1999 hatten Murgänge in Sörenberg markante Schäden verursacht. 

An einer Versammlung wurde den Betroffenen erläutert, was über die Entstehung der 

Murgänge bekannt war, wie Fachleute die weiteren Risiken einschätzten und was dagegen 

unternommen werden könnte. Unmittelbar danach sprachen zwei Teilnehmer den Autor an. 


Der eine bezeichnete die Einschätzung der Risiken als völlig übertrieben, der andere die 

Bedrohung als riesig. Beide hatten zuvor die identischen, wissenschaftlich untermauerten 

Informationen erhalten. Es zeigt sich ein „Spannungsfeld zwischen der räumlichen Konkretheit 

von Gefahrenzonen oder Katastrophenorten, und andererseits der Ortlosigkeit des 

«gefühlten» Risikos.“ (Müller-Mahn 2009) 

Verstehen und Erleben prägen ein Gespräch mit. Das Erleben lässt sich aber viel schwieriger 

in Worte fassen oder darstellen als Ergebnisse wissenschaftlicher Analysen. Im Dialog muss 

auf beides geachtet werden, gleichzeitig und gleichwertig. 

UNUMGÄNGLICHE VEREINFACHUNGEN 

Crick betont, um zu (richtigen) Erkenntnissen zu gelangen müsse reduktionistisch gearbeitet 

werden (BECKER 2009). Reduktionisten glaubten aber, „dass auch die komplexesten 

Systeme aus atomaren und subatomaren Entsprechungen von Federn, Zahnrädchen und 

Hebeln bestehen, die die Natur auf unendliche, vielfältige, geniale Art kombiniere“, so der 

Psychologe John Briggs und der Physiker David Peat. (Briggs/Peat 1993) Wer ein System 

vereinfacht, verzichtet auf Teile desselben: Ein Bohrkern gibt Auskunft über einen Punkt, 

aber nicht über eine Fläche. Karten und Modelle stellen reduzierte Abbilder einer komplexen 

Wirklichkeit dar. Reduktionistische Methoden sind unverzichtbar. Die Vereinfachungen 

können aber dazu führen, einem Argument mehr Gewicht beizumessen als ihm bei einer 

ganzheitlichen Betrachtung zufallen würde. Reduktionistisch erarbeitete Erkenntnisse 

müssen entsprechend achtsam vorgelegt werden. Eine hilfreiche Grundhaltung beschreibt 

der Ökonom Thomas Piketty: „Ökonomen verfügen nicht über unfehlbares Wissen. Aber sie 

haben mehr Zeit als andere Bürger, über wirtschaftliche Probleme nachzudenken. (…) Mein 

wichtigstes Ziel ist es, einen nützlichen Diskussionsbeitrag zu leisten." (BINSWANGER 2014) 

OBJEKTIVITÄT VS. KONSTRUKTIVISMUS 

Gemäss dem Geographen Müller-Mahn stehen sich „in der Vielfalt von Auffassungen über 

den Bedeutungsgehalt von Risiko (…) zwei erkenntnistheoretische Grundpositionen" 

diametral gegenüber, "die objektivistische und die konstruktivistische" (Müller-Mahn 2009). 

Objektivistische Ansätze lägen der gesamten natur- und ingenieurwissenschaftlichen 

Risikoforschung zu Grunde. "Risiko bezieht sich demnach auf ein von aussen (aus der Natur) 

über die betroffenen Menschen oder Gesellschaften hereinbrechendes Ereignis, das über den 

damit verbundenen Schaden gesellschaftlich relevant wird. An dieser Stelle wird der 

diametrale Gegensatz der sozialwissenschaftlichen Perspektiven erkennbar, die Risiko als 

etwas sehen, das nicht von aussen bzw. aus «der Natur» heraus den Menschen überfällt, 

sondern das letztlich von ihm selbst durch seine Wahrnehmung und sein Handeln hergestellt 

wird." (MÜLLER-MAHN 2009) 

Die Neurobiologen Humberto Maturana und Francisco Varela zeigen auf, dass Menschen, 

die dasselbe anschauen, unterschiedliche „Wirklichkeiten“ wahrnehmen. Was das Auge sieht 

führt nicht zu einer 1:1-Abbildung im Hirn. Die Projektion eines Bildes auf die Netzhaut 

„wirkt vielmehr wie eine Stimme (Perturbation), welche zu den vielen Stimmen bei einer 


heftigen Diskussion in einer grossen Familie hinzukommt.“ (MATURANA/VARELA 1987) 

Die anderen Stimmen sind all die Assoziationen, Hoffnungen, Ängste usw., die sich bereits 

im Hirn tummeln, „wobei der schliesslich erreichte Konsens (…) nicht Ausdruck dessen ist, 

was die Familienmitglieder im einzelnen vorgebracht haben.“ Die beiden Neurobiologen 

schliessen daraus: „Das Phänomen der Kommunikation hängt nicht von dem ab, was 

übermittelt wird, sondern von dem, was im Empfänger geschieht. Und dies hat wenig zu tun 

mit «übertragener Information».“ 

Auch die Psychologen Arist von Schlippe und Jochen Schweitzer halten „Wirklichkeit als 

nicht loslösbar vom Beobachter“. Wenn aber „ein Sachverhalt aus unterschiedlichen 

Perspektiven unterschiedlich gesehen werden kann und [dies] zu unterschiedlichen Konsequenzen, 

Urteilen, Entscheidungen führt, dann verliert die Sache (selber) zunehmend an 

Bedeutung. (…) Statt dessen verlagert sich das Interesse auf die Art und Weise, wie soziale 

Gruppen die Sache sehen, benennen und kategorisieren.“ (VON SCHLIPPE/SCHWEITZER 

1998) 

Müller-Mahn folgert für den Dialog: "Von einem solchen konstruktivistischen Risikoverständnis 

ausgehend muss sich das (…) Interesse auf die Akteure und ihre Entscheidungssituation 

richten, d.h. auf die der Entscheidung zugrunde liegenden Kenntnisse, Erfahrungen, Wert - 

massstäbe und Bedürfnisse (…) Ziel dieser subjekt- und handlungszentrierten Ansätze muss 

es daher sein, die Handlungslogiken der Akteure im Kontext ihrer konkreten Handlungs- und 

Lebensbedingungen zu verstehen.“ (MÜLLER-MAHN 2009) 

EMPATHIE IST LERNBAR 

Matthias Haller, Gründer der Stiftung Risiko-Dialog, hält fest, Risiko bürge „für ein Spannungsfeld, 

in dem objektive Daten, aber auch emotionale Reaktionen angelegt sind.“ 

Der Risiko-Dialog werde so zu einer „Auseinandersetzung um mögliche Zukünfte, von den 

einen als Chance, von den anderen ebenso sehr als Gefahr wahrgenommen. Damit ist 

Risikodialog schon im Ursprung Konfliktmanagement und Mediation.“ (HALLER 2014) 

Naturwissenschaftler Matthias Holenstein ergänzt: „Auch die Empathie spielt in der Auseinandersetzung 

eine wichtige Rolle.“ (HOLENSTEIN 2008) 

Empathie bezeichnet die Fähigkeit, sich in Lebensentwürfe, Empfindungen und Verhaltensweisen 

einzufühlen. Jede Fachperson führt ihr eigenes, vielseitiges Leben und lernt dabei 

vieles über das Mensch-Sein. Eine besondere Herausforderung stellt dabei das dar, was einem 

nicht à priori vertraut ist. Es macht deshalb Sinn, sich in Felder zu begeben, in denen man 

mehr erfährt über Bedingungen, Gefühle und Denkweisen, die einem auf Grund der eigenen 

Lebensführung fremd sind. Die Philosophin Susan Neiman empfiehlt Reisen, um sich und 

andere besser kennenzulernen. Vom Blumenbinder zu lernen unten auf dem Markt. 

Oder schlicht Orte zu meiden, wo man die schlauste Person im Raum ist. (SCHMID 2015) 

Auch ein Blick auf das eigene Leseverhalten mag sich lohnen. Gut belegt sind Unterschiede 

zwischen den Geschlechtern: So seien männliche Jugendliche "eher an sachbezogener 


Information interessiert, während die Mädchen (…) fiktionale Geschichten bevorzugen. 

Den Mädchen fällt es leichter, sich lesend in die Erfahrung anderer Menschen einzuleben." 

(HURRELMANN 1994) Der erwachsene Mann meide "die Lektüre dessen, was die Leseforschung 

«human interest stories» nennt und was Frauenlektüre ist. Er wendet sich stattdessen 

Fachzeitschriften und Tageszeitungen zu, die sich mit seiner Berufswirklichkeit beschäftigen." 

Die literarische Autorität des erwachsenen Mannes sei der Geschichts- und Naturwissenschafter 

(SCHLAFFER 2010). Und "Mädchen bevorzugen Belletristik (…) sowie Problemliteratur. 

Jungen dagegen lesen am liebsten Sachbücher" (KATZ 1994) 

Diesen allgemeinen Aussagen steht die Streubreite des individuellen Verhaltens gegenüber. 

Zwei Aussagen sind dennoch bedenkenswert: 

– Mit der Lektüre von Belletristik verbinden Fachleute Begriffe wie "human interest stories" 

und "sich in die Erfahrung anderer Menschen einleben". Romane ermöglichen es demnach, 

sich in das Wesen von Menschen einzufühlen, die den Lesenden grundsätzlich fremd 

sind. 

– Frauen wenden sich diesem Genre häufiger zu als Männer. In den Ingenieurwissenschaften 

sind sie aber in der Unterzahl. In einem schweizerischen Forschungsprogramms wurde 

festgestellt: "So werden zum Beispiel (…) nur wenige Frauen Ingenieurinnen.“ (Maihofer 

2013) 

EMPATHIE IN DER PRAXIS 

2010-13 erarbeiteten alle 61 Feuerwehren des Kantons Luzern eine Notfallplanung. 

Dabei wurden sie von Naturgefahren-Fachleuten unterstützt. Bewerber für diese Mandate 

mussten sich ausweisen über persönliche Erfahrungen im Sozialen Feld Feuerwehr oder 

einem vergleichbaren Sozialen Feld. Die Ausschreibung verwies auf das Konzept von Habitus 

und Sozialen Feldern des Soziologen Pierre Bourdieu. Vom Experten, der angab, er sei 

Spielführer eines Eishockeyvereins der 3. Liga, durfte angenommen werden, dass er mit 

dem Habitus typischer Feuerwehrleute vertraut war und es ihm leicht fallen würde, sich 

mit diesen zu verständigen. Ein anderer jedoch empfahl sich für die Zusammenarbeit mit 

der Feuerwehr mit der Referenz, dem Komitee des Rotary Clubs anzugehören… Das Projekt 

„Notfallplanungen“ wurde ein voller Erfolg. Die Feuerwehren nannten dafür zwei Gründe: 

– die Bereitschaft, den Erfahrungen der Feuerwehr aus deren Einsätzen das gleiche Gewicht 

beizumessen wie den Gefahrenkarten 

– das "unakademische" Auftreten der Experten 

FOLGERUNGEN 

Natur- und ingenieurwissenschaftliche Methoden ermöglichen es, einen Dialog mit Hilfe 

analytisch zugänglicher Grundlagen anzustossen. Betroffene werden darauf aber nicht immer 

"rational" reagieren: 

– Ihre Wahrnehmung ist konstruktivistisch. Damit entzieht sie sich zumindest teilweise 

einem reduktionistisch-logischen Zugriff. 


– Ihre Haltung gegenüber dem Risiko ist zwiespältig: Im Fokus steht nicht nur ein möglicher 

Schaden, sondern auch eine "Währung", mit der im Leben wichtige Ziele "erkauft" werden 

können. 

– Ihr Verhalten wird von schlecht erkennbaren Einflüssen mitgeprägt - externen, aber auch 

persönlichen. 

Das Potenzial für Experten, ihre Chancen im Dialog zu verbessern, liegt somit nicht in der 

Perfektionierung ihrer hochwertigen wissenschaftlichen Argumentation. Gefragt ist vielmehr 

ein wirkungsvoller Einsatz derselben im Gespräch mit Menschen, deren Haltungen und 

Verhalten einem nicht immer vertraut sind. 

EMPFEHLUNGEN FÜR DEN RISIKO-DIALOG 

Ein Weg, Erkenntnisse klar darzulegen und gleichzeitig Unsicherheiten zu benennen, 

besteht darin, die Aussagen entlang der folgenden Fragen zu kategorisieren: 

1. Was "wissen" wir? Worauf basiert dieses Wissen? 

2. Worüber bestehen Vermutungen? Worauf stützen sich diese? 

3. Wo tappen wir im Dunkeln? 

4. Was wird unternommen, um Erkenntnislücken zu schliessen? 

Schwieriger ist es - entsprechend der ganzheitlichen Art der Herausforderung - Empfehlungen 

abzugeben, wie die Empathie gegenüber Dialogpartnern verbessert werden kann: Deren 

Verhalten wird beeinflusst wird von einer "unwissenschaftlichen", aber stark lebensverbundenen 

Wahrnehmung von Risiko, von sich wandelnden Lebensbedingungen und einer 

persönlichen Risiko-Mentalität. Einerseits bietet sich der Besuch entsprechender Kurse an. 

Die Herausforderung besteht aber nicht primär darin, mehr über das Funktionieren von 

Menschen zu wissen, sondern sich noch besser in diese einfühlen zu können. Ein Weg dazu 

kann sein, dorthin zu gehen, wo es uns nicht in erster Linie hinzieht: Physisch, z.B. auf 

Reisen oder an Anlässe, die ausserhalb des eigenen Erfahrungsbereichs liegen. Oder virtuell 

durch ein entsprechendes Leseverhaltens oder die Auswahl von Filmen, die bisher nicht zu 

den Favoriten zählten. 

Institutionen können sich verbessern, indem sie vermehrt sozialwissenschaftliche Kompetenzen 

aufbauen, theoretische und anwendungsbezogene. Mögliche Wege sind: 

– das Weiterbildungskonzept anpassen 

– den Mitarbeitenden die nötigen Ressourcen zur Verfügung stellen 

– den Anteil an Sozialwissenschaftlern und Frauen erhöhen 

REFERENCES 

- Arnswald, U./Schütt H.-P. (Hg. 2011): Rationalität und Irrationalität in den Wissenschaften 

- BAFU, Bundesamt für Umwelt (2011). Fachspezifische Erläuterungen zur Programmvereinbarung 

im Bereich Schutzbauten und Gefahrengrundlagen. 


- BAFU, Bundesamt für Umwelt (2015): Glossar zu EconoMe 3.0. 

http://www.econome.admin.ch/glossar.php 

- Becker, P. (2009). In der Bewusstseinsfalle? 

- Binswanger, D. (2014). Triumph des Faktenhubers. Das Magazin 18/2014 

- Blom P. (2014). Was hat 1914 mit unserer Zeit zu tun? Interview Das Magazin 1-2/2014 

- Briggs, J., Peat F.D. (1993): Die Entdeckung des Chaos 

- Haller, M. (2014). Fokus Risiko: Gefahr oder Chance? riskBRIEF 1/2014 

- Hell, D. (2014). Nicht das Gehirn ist bedrückt, sondern der Mensch. Tages-Anzeiger 

15.3.2014 (Hervorhebung durch den Autor) 

- Holenstein, M. (2008). Reden, wie darüber geredet wird. Interview Schweizerische 

Technische Zeitschrift Nr.10/2008 

- Hurrelmann, B. (1994):Leseförderung. Praxis Deutsch, H 127 

- Katz, D. (1994). Leseverhalten von Berufsschülern. 

- Knecht, N. (2015). Risiko ist ein Menschenrecht. Tages-Anzeiger 29.1.2015 

- Lanier, J. (2014). Wir geben die Verantwortung ab. Tages-Anzeiger 13.10.2014 

- Lessing, G.E. (zit. 2014): zitiert in riskBRIEF 1/2014 

- Maihofer A. et al (2013): Kontinuität und Wandel von Geschlechterungleichheiten 

in Ausbildungs- und Berufsverläufen junger Erwachsener in der Schweiz. 

- Maturana, H. R., Varela F. J. (1987): Der Baum der Erkenntnis. 

- Müller-Mahn, D. (2007): Perspektiven der geografischen Risikoforschung. Geographische 

Rundschau 59/10 

- Obergericht Kt. Zürich (2014): Beschluss vom 8.8.2014 

- PLANAT, Nationale Plattform Naturgefahren (2015). Zyklus: Risikomanagement. 

http://www.planat.ch/de/fachleute/risikomanagement/ 

- Schlaffer, H. (2010): Lektüre und Geschlecht. Neue Zürcher Zeitung 31.7. 2010 

- Schmid, B. (2015). Werdet erwachsen! Das Magazin Nr. 7/2015 

- Tobler, A. (2014). Unser Risiko gib uns heute. Tages-Anzeiger 8.5.2014 

- Von Schlippe, A., Schweitzer, J. (1998). Lehrbuch der systemischen Therapie und Beratung 



Examples of urban Flood Risk Management in Styria 

Hochwasserrisikomanagement in städtischen 

Bereichen – Beispiele aus der Steiermark 

Rudolf Hornich, Dipl.-Ing.¹; Tanja Schriebl, Dipl.-Ing.² 

ABSTRACT 

Under the pressure of demographic, infrastructural and spatial problems the issue of flood risk 

has often been dodged in urban areas; therefore the damage potential has risen enormously 

in some cities. The flooding of many streams in the last years caused great damage in urban 

areas. Beside technical flood protection measures non-structural measures are to be adopted 

directly at the site of the structures and further detailed alert and intervention plans are to be 

worked out for civil defence forces. Particular attention is being devoted to public relations. 

Watercourses were straightened, narrowed or forced into canals and the urgently-needed 

retention spaces were sacrificed for the sake of other uses. The room required to safely receive 

the incoming floodwaters is, therefore, no longer there. The number and intensity of flash 

and pluvial floods as consequences of heavy rain events is increasing. This means a need for 

new tools and adaptation measures. Using the example of 4 cities the strategy in urban flood 

risk management in Styria shall be illustrated. 


Siedlungsdruck, Infrastruktur- und Standortprobleme führten dazu, dass man in urbanen 

Lebensräumen der Thematik Hochwassergefährdung vielfach ausgewichen ist. Dadurch ist 

das Schadenspotential in einigen Städten enorm gestiegen. Hochwasserereignis an mehreren 

Bächen haben in den letzten Jahren enorme Schäden in städtischen Bereichen verursacht. 

Neben technischen Schutzmaßnahmen sind verstärkt nicht technische Maßnahmen 

anzuwenden und darüber hinaus detaillierte Alarm- und Einsatzpläne auszuarbeiten. 

Besonderes Augenmerk ist der Öffentlichkeitsarbeit zu widmen. Bachläufe wurden begradigt, 

eingeengt und in Kanäle eingeleitet und die erforderlichen Retentionsräume wurden für 

andere Zwecke geopfert. Die erforderlichen Räume zur gesicherten Abfuhr von Hochwässern 

sind nicht mehr vorhanden. Die Anzahl und die Intensität der Starkregenereignisse ist 

zunehmend, diese bedeutet die Notwendigkeit für neuen Strategien und Anpassungsmassnahmen. 

Anhand der Beispiele von 4 Städten wird die Strategie des urbanen Hochwasserrisikomanagements 

in der Steiermark dargestellt. 

KEYWORDS 

flood risk management, flood protection, river restoration, public relations, urban ares 

1 Office of the Styrian Government, Graz, AUSTRIA, rudolf.hornich@stmk.gv.at 

2 Office of the Styrian Government, Department 14 



EINLEITUNG 

Siedlungsdruck, Infrastruktur- und Standortprobleme in Verbindung mit langen Perioden 

ohne markante Hochwasserereignisse Ende des letzten Jahrhunderts und das Fehlen von 

entsprechenden Unterlagen über Hochwasserabflussbereiche führten dazu, dass bei städtebaulichen 

Planungen die Thematik Hochwassergefährdung vielfach vernachlässigt wurde. 

Versäumnisse auf diesem Gebiet haben im Ereignisfall verheerende Folgen nach sich gezogen. 

Das Schadenspotential ist in einigen dicht besiedelten Gemeinden in den letzten Jahren 

enorm gestiegen. In der Steiermark sind davon exemplarisch die Städte Feldbach, Voitsberg 

und Bad Radkersburg, besonders aber die Hauptstadt Graz betroffen. Anhand dieser Beispiele 

sollen die Schwierigkeiten und Probleme und mögliche Lösungen zur Thematik „Hochwassermanagement 

im städtischen Bereich“ unter Berücksichtigung der Anforderungen der 

Hochwasserrichtlinie (RL 2007/60/EG) und der Wasserrahmenrichtlinie (2000/60/EG) 

dargestellt werden. 

Abbildung 1: Projektgebiete - Übersichtskarte 

PROBLEMSTELLUNG 

Starkregenereignisse verursachten in den letzten Jahren Überflutungen und große Schäden 

in urbanen Bereichen. Durch das immer weitere Heranrücken von Bebauungen und 

höherwertigen Nutzungen an die Fließgewässer sind oft in den dicht bebauten Unterlaufabschnitten 

erforderliche Abflussräume für Hochwasser kaum mehr vorhanden. Die Abflussquerschnitte 

nehmen eher ab als zu. Der für die schadlose Abfuhr von Hochwässern 


enötigte Platz ist zumeist bebaut und somit nicht mehr verfügbar. Verrohrungen, Überdeckungen 

und das Verlegen von Gerinnen aus der Tiefenlinie haben zusätzlich dazu geführt, 

dass der Hochwasserabfluss vom Bachbett völlig abgetrennt ist. Die Wassermassen fließen 

unkontrolliert innerhalb des Siedlungsgebietes ab und verursachen große Schäden. Technische 

Schutzmaßnahmen sind auf Grund beengter Platzverhältnisse und fehlender Räume 

oft nicht ausreichend, um einen adäquaten Hochwasserschutz herzustellen. 

STADT GRAZ, SACHPROGRAMM HOCHWASSER GRAZER BÄCHE 

Im Stadtgebiet von Graz findet man über 50 Bäche vor, wobei 10 davon als Wildbäche 

ausgewiesen sind. Zahllose historische Hochwasserereignisse sind in Graz überliefert. 

Nach der Hochwasserkatastrophe 1975 wurden erste konkrete Hochwasserschutzkonzepte 

erarbeitet. Eine Hochwasserabflussuntersuchung im Jahre 1997 mit Ausweisung der 

Überflutungsflächen für das 30- und 100-jährliche Hochwasser für die Stadtbäche hat 

ergeben, dass rund 1000 Objekte hochwassergefährdet sind. 

Beim Hochwasser am 21. August 2005 haben die Überflutungen an fast allen Grazer Bächen 

ein derartiges Ausmaß erreicht, dass für das Stadtgebiet Katastrophenalarm ausgelöst wurde. 

Hunderte Keller und Erdgeschoße sowie mehrere Tiefgaragen wurden unter Wasser gesetzt. 

Die Schäden betrugen rund 5 Millionen Euro. Im Sommer 2009 musste wiederum zweimal, 

im Juli und im August, Hochwasserkatastrophenalarm ausgerufen werden. 

In enger Kooperation zwischen der Stadt Graz, dem Land Steiermark und Bundesministerium 

für Land- und Forstwirtschaft, Umwelt und Wasserwirtschaft wurde im Jahr 2006 ein 

gemeinsames Strategiepapier „Sachprogramm Grazer Bäche“ ausgearbeitet. Unter Berücksichtigung 

der Erfordernisse der Fachbereiche Raumordnung, Stadtentwicklung, Gewässerökologie, 

Siedlungswasserwirtschaft und Katastrophenschutz wurde das vordringliche Ziel 

mit dem „Erreichen eines nachhaltigen Hochwasserschutzes für die gefährdeten Objekte 

und Infrastruktureinrichtungen innerhalb der Stadt Graz“ formuliert. 

Der vorgeschlagene Maßnahmenkatalog beinhaltet: 

– die Freihaltung und zusätzliche Schaffung von Überflutungsräumen 

– die Verbesserung der Abflussleistungen durch Querschnittserweiterungen und Aufweitungen 

– die Errichtung von 29 Hochwasserrückhaltebecken 

– die Erreichung des guten morphologischen Zustandes bzw. des guten ökologischen 

Potentials. 

Trotz optimaler Ausnutzung der örtlichen Möglichkeiten und umfangreichen Ankauf von 

Grundflächen, ist es nicht möglich, für alle gefährdeten Siedlungen einen Hochwasserschutz 

bis zu einem HQ 100 

zu erreichen. Im Vergleich zum Ist-Zustand wird jedoch überall eine 

deutliche Verbesserung erzielt. Für jene Abschnitte, wo mit technischen Schutzmaßnahmen 

kein ausreichender Hochwasserschutz zu erzielen ist, sind ergänzend Objektschutzmaßnahmen 

vorzusehen und zusätzlich detaillierte Alarm- und Einsatzpläne für die Einsatzkräfte 

auszuarbeiten. 


Abbildung 2: Petersbach Stadt Graz, Mittellauf nach Umsetzung der Hochwasserschutzmaßnahmen 

Bei der Umsetzung von Hochwasserschutzmaßnahmen im Stadtgebiet ist man mit einer 

Unzahl von Problemen konfrontiert. Neben der Koordination verschiedener Leitungsträger 

oder der Aufrechterhaltung der Verkehrsströme stellen die Verhandlungen für die Bereitstellung 

der erforderlichen Grundflächen die größte Herausforderung dar. Ein besonderes 

Augenmerk wird daher der Information und der Beteiligung der Öffentlichkeit gewidmet. 

Durch Informationsveranstaltungen, Webseiten (z.B. www.wasser.graz.at) und Broschüren 

soll der Bevölkerung aufgezeigt werden, dass in Ergänzung zu den Schutzmaßnahmen durch 

die Öffentliche Hand auch jeder Betroffene in Eigenverantwortung seinen Beitrag leisten 

kann (muss), indem er an passive Schutzmaßnahmen wie hochwasserangepasstes Bauen, 

Leben und Wohnen, Einsatz mobiler Elemente, Hochwasserschadensversicherungen und 

persönliche Notfallpläne denkt und dafür auch die notwendigen Vorbereitungen trifft. 

Die Gesamtkosten für die Umsetzung im Rahmen eines Zehnjahres-Bauprogrammes (2006 

– 2015) wurden auf der Preisbasis August 2006 mit € 65,0 Millionen geschätzt. Die Finanzierung 

wird vom Bund, dem Land Steiermark und der Stadt Graz getragen. 

Bisher wurden rund 75% der geplanten Maßnahmen (Hochwasserrückhaltebecken und 

Linearausbau) umgesetzt. Diese haben sich bereits bei einigen Hochwasserereignissen positiv 

bewährt und größere Schäden verhindert. Bei allen Projekten konnten auch ökologische 

Verbesserungen erzielt werden. An mehreren Stellen wurden Zugänge zum Gewässer 

ermöglicht, um der Bevölkerung auch neue Freizeit- und Erholungsräume zu erschließen. 

Für die potentiell signifikanten Hochwasserrisikogebiete im Stadtgebiet von Graz wurden 

2014 in Entsprechung der EU-Hochwasserrichtlinie 2007/60/EG Hochwasserrisikomanagementpläne 

erstellt. 

Neben den strukturellen Maßnahmen, die im Sachprogramm bereits definiert wurden, liegen 

Schwerpunkte auch auf nichtstrukturellen Maßnahmen. Als Teil der Öffentlichkeitsarbeit 


wurde gemeinsam mit Vertretern der Bürgerinitiativen eine Broschüre mit dem Titel 

„Hochwasser – ich sorge vor“ (www.katastrophenschutz.graz.at) erstellt. Die Feuerwehr der 

Stadt Graz hat ein eigenes Hochwasser App entwickelt und neue „Katastrophenschutzpläne 

Hochwasser“ ausgearbeitet. 

STADT FELDBACH, HOCHWASSERSCHUTZ RAAB - STÄDTEBAULICHE AKZENTE 

Die Ergebnisse einer Hochwasserabflussuntersuchung an der Raab im Jahre 2008 haben 

gezeigt, dass der bestehende Hochwasserschutz in der Stadt Feldbach auf Grund von 

geänderten Hydrodaten und geänderten topografischen Verhältnissen für ein 100-jährliches 

Ereignis - mit einer Abflussmenge von 260m³/s - nicht mehr ausreichend war. Daher waren 

neue Ufermauern zu errichten und die vorhandenen Dämme zu erhöhen bzw. an den Stand 

der Technik anzupassen. Zur Verbesserung der Standsicherheit und zur Minimierung der 

Durchströmung im Hochwasserfall wurden Stahlspundwänden in die Dämme eingerammt. 

In Verbindung mit der Optimierung des Hochwasserschutzes war es aber auch ein Ziel des 

Projektes, den Fluss im Stadtgebiet von Feldbach mit Freizeit- und Erholungsbereichen 

auszustatten und aufzuwerten. Die alte Raabregulierung mit monotonen und befestigten 

Trapezprofilen war nur auf die Hochwasserabflussleistung ausgelegt – eine Naherholungsfunktion 

war nicht gegeben. Flussmorphologische und ökologische Verbesserungen wurden 

durch Auflösung der Profilmonotonie in Form von lokalen Aufweitungen, Einbringung von 

Totholz und Strukturierung der Böschungsbereiche sowie dem Umbau der vorhandenen 

Sohlstufen erreicht. 

Hauptziel des Projektes war die Erschließung des Flussraumes für die Bevölkerung und 

Verbesserung der Wahrnehmung des Flusses als attraktives städtisches Element. Dazu wurde 

das Gehwege- und Radwegenetz an den Fluss und in die Böschungsbereiche der Raab verlegt 

und Furten aus Betonelementen errichtet, die an zwei Stellen im Stadtgebiet eine Querung 

des Flusses, aber auch eine Aufenthaltsmöglichkeit am Fluss bieten. Ab dem HQ 1 

werden die 

Abgänge zum Fluss durch die Feuerwehr gesperrt. Eine Plattform, die im Bereich eines Cafés 

oberhalb des Hochwasserabflussraumes in das Flussprofil ragt, soll den Aufenthalt am Fluss 

attraktivieren. 

Als Beitrag zur Öffentlichkeitsarbeit und zur Information der Bevölkerung wurde im 

Stadtgebiet am Raabufer direkt neben dem Geh- und Radweg ein „gläserner Pegel“ mit 

entsprechenden Erläuterungen errichtet. 

Die Zugangsmöglichkeiten zum Fluss und die Nutzung des Flussraumes für Freizeit- und 

Erholungsaktivitäten wurden sofort nach Fertigstellung des Projektes von der Bevölkerung 

sehr gut angenommen. Vor allem Kinder und Jugendliche haben diese neuen Räume 

entdeckt und erobert. 


Abbildung 3: Raab Stadt Feldbach, erlebbarer städtischer Flussraum nach dem Ausbau 

STADT VOITSBERG, HOCHWASSERSCHUTZ KAINACH – ANPASSUNG AN BESTEHENDE STRUKTUREN 

Die Kainach weist bis zur Stadt Voitsberg ein Einzugsgebiet von 210 km² auf. Das HQ 100 

beträgt 190 m³/s, bereits bei einem HQ 30 

gibt es Ausuferungen im Stadtgebiet. 

Erste Dokumente über eine Regulierung der Kainach stammen aus dem Jahr 1803. Diese 

diente damals noch dem Zweck, die landwirtschaftlichen Kulturen rund um Voitsberg vor 

wiederkehrenden Hochwässern zu schützen und die Bewirtschaftung durch eine Begradigung 

des Flusses zu erleichtern. Erst ab der 2. Hälfte des 20. Jahrhundert diente ein Ausbau der 

Kainach dem Schutz von Siedlungsstrukturen. Die Ausbauwassermenge lag bei 160 m³/s. 

In den Jahren 1996-1998 wurde an der Kainach eine Abflussuntersuchung durchgeführt und 

die Hochwasserüberflutungsflächen für das HQ 30 

und das HQ 100 

ausgewiesen. Dabei kristallisierte 

sich heraus, dass Teile des Stadtgebiets im Hochwasserabflussbereich der Kainach 

liegen. Im HQ 100 

Ereignisfall sind rund 108 Objekte und Infrastruktureinrichtungen betroffen. 

Somit bestand erneut die Notwendigkeit zur Umsetzung von Hochwasserschutzmaßnahmen 

im Stadtbereich. 

Von der Planung bis zur Umsetzung der Hochwasserschutzmaßnahmen war es ein weiter 

Weg. Dabei spielten viele Faktoren wie z.B. kein Auftreten eines HQ 30 

Ereignisses an der 

Kainach seit dem letzten Ausbau, kaum freie Flächen für Hochwasserschutzmaßnahmen 

sowie keine Wahrnehmung mehr des Gewässers durch die Stadtbevölkerung eine große 


Rolle. Dadurch dauerten die Grundverhandlungen beinahe 7 Jahre. Im Jahr 2006 konnte der 

wasser- und naturschutzrechtliche Bescheid für die Hochwasserschutzmaßnahmen erlangt 

werden. 

Derzeit werden die Maßnahmen zur Erzielung eines 100-jährlichen Hochwasserschutzes 

realisiert. Aufgrund der beengten Platzverhältnisse handelt es sich bei den Hochwasserschutzmaßnahmen 

in erster Linie um einen linearen Ausbau mit einem Freibord von 50 cm, 

möglichst wenigen Eingriffen direkt am Gewässer und der Errichtung von Mauern nur in 

Bereichen, in denen aus Platzgründen keine Dämme errichtet werden können. Dennoch ist 

es notwendig die Kainach auf einer Länge von mehreren hundert Metern einzutiefen, um 

einen geeigneten Hochwasserschutz zu erreichen. 

Abbildung 4: Kainach Stadt Voitsberg, Strukturierungen der Flusssohle 

Des Weiteren werden auch Maßnahmen zur Verbesserung der Gewässerökologie umgesetzt. 

Dazu gehören die Auflösung von 3 Sohlstufen in Form der Errichtung von Sohlrampen, die 

Strukturierung der Gewässersohle an geeigneten Stellen und der Umbau der Mündung des 

Tregistbaches in eine stufenlosen Anbindung. Durch die Umwandlung der Sohlstufen in 

Sohlrampen und der damit verbundenen Entfernung von riesigen Kolken, ist es auch 

notwendig, durch Sohlstrukturen wieder die Möglichkeit von Wintereinständen für die 

Fische zu schaffen. 

Eine neue, an den Stand der Technik angepasste, 2D-Abflussuntersuchung bestätigte im Jahr 

2011 die bestehenden Abflussflächen aus der 1D- Untersuchung Ende der 90-iger Jahre, doch 

auch neue, bis dahin unbekannte Überflutungsflächen kamen hinzu. 

Es wurde festgestellt, dass die Kainach flussauf bereits in der Nachbargemeinde Bärnbach 

ausufert und in das Stadtzentrum abfließt. Demnach besteht die Gefahr, dass das Hochwasser 


aufgrund der nun errichteten Linearmaßnahmen, die zum Großteil aus Dämmen und 

Mauern bestehen, nicht mehr in die Kainach zurückfließen kann. Daher sind die Planungsarbeiten 

zu erweitern und anzupassen. Die Umsetzung der entsprechenden Hochwasserschutzmaßnahmen 

erfolgt in einem 3. Bauabschnitt, wobei zusätzlich Aufweitungen des Flussbettes 

vorgesehen sind. 

Da es sich beim Stadtbereich Voitsberg um ein potentiell signifikantes Hochwasserrisikogebiet 

gemäß der Europäischen Hochwasserrichtlinie 2007/60/EG handelt, wurde im Jahr 2014 

der zu erstellende Hochwasserrisikomanagementplan gemeinsam mit der Stadtgemeinde 

Voitsberg und allen betroffenen öffentlichen Stellen erarbeitet. 

STADT BAD RADKERSBURG, HOCHWASSERSCHUTZ MUR – GRENZÜBERSCHREITENDER ANSATZ 

Im Auftrag der ständigen Österreichisch-Slowenischen Kommission für die Mur wurde im 

Jahr 2000 eine aktuelle Hochwasserabflussuntersuchung für die Mur-Grenzstrecke erstellt. 

Dabei hat sich herausgestellt, dass der im Jahre 1972 errichtete Hochwasserschutz für die 

Stadt Bad Radkersburg mit einem Schutzgrad für ein 100-jährliches Hochwasserereignis nicht 

mehr gegeben war. Darüber hinaus entsprechen die vor mehr als 40 Jahren errichteten 

Dämme nicht mehr den heutigen technischen Anforderungen. Neben der Herstellung der 

erforderlichen Damm-Nivellette ist daher auch die Standsicherheit der bestehenden Dämme 

zu gewährleisten. Als Kerndichtung des Dammes wurde im Projekt eine Schmalwand 

gewählt, die bis 6 m in den Untergrund einbindet, um die Unterströmung des Dammes zu 

verringern. 

Der Hochwasserrisikomanagementplan für diesen Abschnitt sieht neben den strukturellen 

Maßnahmen auch ein Bündel an nicht strukturellen Maßnahmen wie Verbesserung des 

bestehenden Hochwasserprognosemodells, Ausarbeitung von Katastrophenschutzplänen 

Hochwasser und eine intensive Öffentlichkeitsarbeit zur Verbesserung des Hochwasser-Bewusstseins 

in der Bevölkerung vor. 

Neben der Hochwassersicherheit sind beim Projekt in Bad Radkersburg auch städtebauliche 

Aspekte von hoher Bedeutung. Im Bereich des Thermenareals werden landschaftsökologische 

Maßnahmen bei der Gestaltung des Hochwasserschutzdammes berücksichtigt. Bei der 

Murbrücke wurden auf österreichischer Seite Abgänge und eine Plattform in den Fluss 

errichtet, wodurch ein Zugang zur Mur und direkte „Flusseindrücke“ ermöglicht werden. 

Auf slowenischer Seite wurden zwei Flussplattformen und eine „Murtribüne“ in Form von 

Betonstufen im Böschungsbereich errichtet. Beide Projekte wurden im Rahmen des Projektes 

Europäische territoriale Zusammenarbeit (ETZ) mit finanziellen Mitteln der EU kofinanziert. 


FAZIT 

Die Hochwasserkatastrophen der letzten Jahre haben ganz klar aufgezeigt, dass künftig vor 

allem im urbanen Bereich einem gezielten Hochwasserrisikomanagement wesentlich mehr 

Augenmerk geschenkt werden muss, als dies bisher der Fall war. Neben den rein technischen, 

baulichen Hochwasserschutzmaßnahmen sind daher verstärkt nicht technische Maßnahmen 

wie zum Beispiel optimierte Einsatz- und Alarmplänen und Prognosemodelle für kleine, 

städtische Einzugsgebiete auszuarbeiten. Ein besonderer Schwerpunkt ist auch auf eine 

intensive Öffentlichkeitsarbeit mit gezielten Informationen und auf die Einbindung und 

Beteiligung der Öffentlichkeit zu legen. Die Ergebnisse der „vorläufigen Bewertung des 

Hochwasserrisikos“ gemäß Richtlinie 2007/60/EG zeigten auch sehr deutlich auf, dass in den 

nächsten Jahren der Schwerpunkt eindeutig in den Bereichen mit dichter Besiedelung, also 

in den städtischen Bereichen, liegt. Ein integraler Ansatz und eine intensive interdisziplinäre 

Zusammenarbeit bei Hochwasserschutzprojekten in städtischen Bereichen führen zu einer 

erheblichen Reduktion des Hochwasserrisikos und des Schadenspotentials. Gutes Hochwasserrisikomanagement 

wird in Zukunft noch mehr als bisher die Lebensqualität urbaner 

Räume mitbestimmen. 

LITERATUR 

- Amt der Steiermärkischen Landesregierung, Fachabteilung 19B, Forsttechnischer Dienst für 

Wildbach- und Lawinenverbauung, Sektion Steiermark, Magistrat Graz, A 10/5 – Abteilung 

für Grünraum und Gewässer (2007). Sachprogramm Grazer Bäche, Maßnahmenprogramm. 



Flood Risk Reduction through Object Protection 

Measures - Benefit Analysis on Catchment Scale 

in Upper Austria 

Reduzierung des Hochwasserrisikos durch Objektschutzmaßnahmen 

– Nutzenanalyse auf Einzugsgebietsskala 

in Oberösterreich 

Matthias Huttenlau¹; Stefan Achleitner²; Benjamin Winter³; Manuel Plörer²; Michael Hofer 4 ; Felix Weingartner 5 

ABSTRACT 

Object-specific protective measures can play a vital role in the framework of flood risk 

management to reduce flood risk. From a technical point of view, those measures can be 

divided into waterproof and elevated construction technics. The temporal development of 

flood risk and the potential benefit of two types of technical protection measures to reduce 

damages was analysed and assessed within a case study conducted in Upper Austria on 

mesoscale. The study frameworks applies on the one side the common scenario-based 

approach considering events with certain return periods and on the other side a stochastic 

approach. It is shown, that both types of object protection measures are an efficient instrument 

until the protection goal is reached. If intensities exceed the protection goal of inundation 

depths with a recurrence interval of 100 years plus 20cm, a significant and swift decrease 

of the protective effect can be identified, especially with waterproof construction technics. 

In contrast, the protective effect of elevated construction technics persists significantly longer. 


Objektschutzmaßnahmen können bei der Umsetzung von Maßnahmen zur Reduzierung des 

Hochwasserrisikos im Rahmen des integrierten Hochwasserrisiko-Managements einen 

effizienten Beitrag leisten. Dabei stehen unterschiedliche Objektschutzmaßnahmen der 

wasserdichten Bauweise und der erhöhten Bauweise zur Verfügung. Im vorliegenden Beitrag 

wurde in einem mesoskaligen Einzugsgebiet in Oberösterreich der Frage nachgegangen, wie 

sich das Hochwasserrisiko über die Zeit möglicherweise verändert und welcher monetäre 

schadenreduzierende Nutzen von Objektschutzmaßnahmen im Rahmen der baurechtlichen 

Genehmigungsverfahren zu erwarten ist. Hierbei wurde zum einen der in der Praxis gängige 

1 alpS Centre for Climate Change Adaptation, Innsbruck, AUSTRIA, huttenlau@alps-gmbh.com 

2 Arbeitsbereich Wasserbau, Institut für Infrastruktur, Universität Innsbruck, AUSTRIA 

3 alpS GmbH, Innsbruck und Institut für Geographie, Universität Innsbruck, AUSTRIA 

4 Ingenieurbüro Dipl.-Ing. Günter Humer GmbH, Geboltskirchen, AUSTRIA 

5 Abteilung Oberflächengewässerwirtschaft, Gruppe Schutzwasserwirtschaft, Amt der Oberösterreichischen Landesregierung, Linz, 

AUSTRIA 



szenarien-basierte Ansatz über Ereignisse mit bestimmten Eintretenswahrscheinlichkeiten 

gewählt, zum anderen kam ein stochastischer Ansatz zur Anwendung. Es kann mit der 

Untersuchung gezeigt werden, dass beide Arten von Objektschutzmaßnahmen ein effizientes 

Instrument bis zur Erreichung des Schutzzieles (Hochwasserabflussbereich eines 100-jährlichen 

Ereignisses plus 20cm) darstellen. Ab dem Überschreiten des Schutzzieles ist eine rasche 

und deutliche Abnahme der Schutzwirkung bis hin zur Unwirksamkeit bei Maßnahmen der 

wasserdichten Bauweise erkennbar, wohingegen erhöhte Bauweisen ihre positive Schutzwirkung 

deutlich länger aufrechterhalten und damit generell eine effizientere Maßnahme 

darstellen. 

KEYWORDS 

flood risk; risk management; object-specific protective measure 

EINFÜHRUNG 

Risikoanalysen und –bewertungen sind eine wesentliche Entscheidungsgrundlage bei der 

Entwicklung und Umsetzung von Maßnahmen zur Reduzierung des Hochwasserrisikos. Die 

EU Richtlinie 2007/60/EG über die Bewertung und das Management von Hochwasserrisiken 

(ABI L 288/27) stellt dabei den rechtlichen Rahmen zu einer langfristigen Reduzierung des 

Hochwasserrisikos dar. In diesem Zusammenhang können Objektschutzmaßnahmen (OSM) 

– neben Maßnahmen zur Abflussreduzierung, technischen sowie planerischen Maßnahmen 

und Bewältigungsmaßnahmen – einen effektiven Beitrag zur Reduzierung von potenziellen 

Schäden beitragen (z.B. Holub et al., 2012). Dieser Beitrag stellt die Ergebnisse einer 

risikobasierten Fallstudie dar, bei der die zeitliche Entwicklung des Schadenpotenzials und der 

schadenmindernde Effekt von strukturellen Objektschutzmaßnahmen (erhöhte Bauweise 

und wasserdichte Bauweise) exemplarisch in einem mesoskaligen Einzugsgebiet untersucht 

wurde. 

Das Untersuchungsgebiet umfasst das hydrologische Einzugsgebiet des sogenannten Ottnanger 

Redls in Oberösterreich (A). Die Siedlungsgebiete der acht in diesem Einzugsgebiet 

liegenden Gemeinden mit ca. 18.500 Einwohnern sind in regelmäßigen Abständen von 

Hochwasser betroffen, wobei das Hochwasser im Jahr 2002 den bisher größten Schaden 

verursachte. Das Einzugsgebiet umfasst eine Fläche von ca. 60 km² in Höhenlagen zwischen 

400 und 720 m ü. Adria. 

Das Oberösterreichische Bautechnikgesetz 2013 (Oö. BauTG 2013) (LBGl. Nr. 35/2013) sieht 

in § 47 Hochwassergeschützte Gestaltung von Gebäuden vor, dass Neu-, Zu- und Umbauten 

von Gebäuden im 100-jährlichen Hochwasserabflussbereich hochwassergeschützt zu planen 

und auszuführen sind. Unter hochwassergestützter Gestaltung sind insbesondere folgenden 

Maßnahmen zu sehen: (i) gegenüber dem Untergrund abgedichtete oder aufgeständerte 

Bauweise, (ii) Abdichtungs- und Schutzmaßnahmen gegen einen Wassereintritt, (iii) 

auftriebssichere Ausführung aus wasserbeständigen Materialien, (iv) die Fußbodenoberkante 

von Wohnräumen, Stallungen, Wirtschaftsräumen, etc. muss mindestens 20cm über dem 

Niveau des 100-jährlichen Hochwasserabflussbereiches liegen. Durch die mit 01.01.2015 in 


Kraft getretene Novelle des Oö- BauTG 2013 wurde dieser Grenzwert von 20cm auf 50cm 

erhöht. Die Auswirkungen dieser Novelle wurde in der vorliegende Studie nicht mehr 

berücksichtigt. 

Auf dem Hintergrund der vorliegenden baurechtlichen Rahmenbedingungen wurden 

folgende Fragen verfolgt: Welche Schadenpotentiale sind durch Siedlungsentwicklung im 

Untersuchungsgebiet ausgehend von den 1990er Jahren über 2012 bis 2030 denkbar und 

wie hoch ist die potenzielle schadenmindernde Wirkung von strukturellen OSM. 

METHODEN 

Das generelle Untersuchungskonzept folgt dem risikobasierten Ansatz, bei dem sich das 

Risiko aus der Eintretenswahrscheinlichkeit eines gefährlichen Ereignisses und dem Schadenausmaß 

zusammensetzt. Der Analyseablauf untergliedert sich in mehrere, aufeinander 

aufbauende Schritte, basierend auf der grundlegenden Gliederung in Gefahrenanalyse, 

Expositionsanalyse und Folgenanalyse (z.B. Kienholz, 2005). Ein Überblick über den 

gesamten Analyseablauf ist in Abbildung 1 gegeben, für weiterführende Sensitivitätsanalysen 

zu den einzelnen Analysekomponenten wird auf Achleitner et al. (in press) verwiesen. 

Abbildung 1: Analysekonzept und Ablaufdiagramm der Bearbeitung. 

Aufbauend auf der aktuellen Siedlungsstruktur (Szenario Sz2012) wurde die Siedlungsstruktur 

Ende der 1990er Jahre (Szenario Sz1999) v.a. durch Fernerkundungsdaten rekonstruiert 

und zwei Szenarien für das Jahr 2030 (Szenarien Sz_mod und Sz_max) mit einem partizipativen 

Ansatz erarbeitet (Abbildung 1, (2)). Die aktuelle Siedlungsstruktur wurde durch die 

Kombination einer Vielzahl an Geodaten (Adressdaten inkl. genauer Gebäudeattributierung, 

Gebäudegrundrisse, Kataster, Flächenwidmungspläne, Geländemodelle DGM/DOM, etc.) und 

weiterführenden Kartierungen erfasst. Somit konnte neben der Verortung der Gebäude auch 


deren Funktionalität und Größe (Bruttogeschoßfläche sowie Bruttokubatur) zur weiterführenden 

monetären Bewertung detailliert erfasst werden. Das Siedlungsszenario Sz_mod stellt 

ein mittleres, aus heutiger Sicht moderates Entwicklungsszenario dar. Hierzu wurde die 

statistische Entwicklung in den einzelnen Gemeinden (Statistik Austria), Szenarien der 

Raumentwicklung Österreichs bis in das Jahr 2030 (ÖROK, 2009), kleinräumige Bevölkerungsprognosen 

für Österreich bis 2030 (Hainka, 2010), überörtliche Raumentwicklungskonzepte 

(Land Oberösterreich) bzw. örtliche Entwicklungskonzepte der einzelnen Gemeinden 

sowie Ausschlusskriterien durch unterschiedliche Zonierungen analysiert und aufbereitet. 

Die Szenarienentwicklung erfolgte in einem zweistufigen Verfahren unter Einbeziehung 

lokaler Expertenmeinungen (Gemeindeamtsleiter und Raumplaner). Das Siedlungsszenario 

Sz_max geht dagegen von einer vollständigen Bebauung der ausreichend vorhandenen 

Baulandreserven unter Berücksichtigung von Bebauungsverbotskriterien aus, wobei auf ein 

einheitliches Ortsbild durch Nachbarschaftskriterien geachtet wurde. 

Die berücksichtigten hydrologischen Szenarien basierend auf dem rekonstruierten Niederschlagsfeld 

des Ereignisses 2002, welches zur Generierung weiterer Ereignisse skaliert wurde. 

Hierbei wurde die räumlich-zeitliche Niederschlagsverteilung beibehalten und die Niederschlagssumme 

mit einem Faktor variiert (Abbildung 1, (1.2)). Die Kalibrierung des hydrologischen 

Modells (HEC-HMS) erfolgte für das Ereignis 2002 mit einem neuntägigen Zeitraum 

(Abbildung 1, (1.1)). Die berücksichtigten Gefahrenkarten (Überflutungstiefen mit einem 

Meter Auflösung) wurde mit einer 2d-hydraulischen Modellierung (Hydra_AS_2d) berechnet, 

wobei für alle verwendeten Siedlungsszenarien separate hydraulische Netze erstellt 

wurden (Abbildung 1, (1.3)). 

Die monetäre Bewertung der Siedlungsstruktur erfolgte anhand der zur Verfügung gestellten 

Richtlinien der Oberösterreichischen Landesversicherung zur Gebäude und Inventarbewertung 

für das Jahr 2012; Werte für das Jahr 2012 stellen allgemein die Referenzwerte für die 

Untersuchung dar. Die vorhandenen Gebäudeinformationen wurden durch Kartierungen 

vervollständigt, sodass die Anwendung der versicherungsinternen Richtlinien für private 

Wohngebäude, für landwirtschaftliche Anwesen sowie für Gewerbe und Industrie durchgeführt 

werden konnte (Abbildung 1, (3)). Zur Berechnung der potenziellen Schäden wurden 

die Schadenfunktionen nach BUWAL (Borter et al., 1999) berücksichtigt. Um die Wirkung 

der OSM (i) wasserdichte Bauweise (bzw. mobile Maßnahmen zur Erreichung der Wasserdichtigkeit) 

(folgend OSM 1) und (ii) erhöhte Bauweise (folgend OSM 2) zu berücksichtigen, 

erfolgte ein Modifikation der verwendeten Schadenfunktionen (Abbildung 1, (4) bzw. 

Abbildung 2). Als Referenzmarke bis zu der die OSM ihre volle Wirksamkeit erfüllen, wurde 

der in der Gefahrenzonenplanung ermittelte Wasserstand des 100-jährlichen Hochwasserabflussbereiches 

plus 20 cm herangezogen. 

Die abschließende Schadenmodellierung und Risikoanalyse sowie die vergleichende Gegen - 

überstellung der (i) zeitlichen Entwicklung und (ii) Wirksamkeit von OSM, erfolgte mit zwei 

Ansätzen. Zum einen mit dem gängigen Ansatz unter Berücksichtigung von Szenarien mit 

unterschiedlichen Eintretenswahrscheinlichkeiten, zum anderen mit einem stochastischen 

Ansatz. Für den stochastischen Ansatz wird (i) eine 10.000-jährliche Zeitreihe des jährlichen 


Maximalabflusses (AMS) für den Abflusspegel mit einer Monte-Carlo-Simulation generiert, 

(ii) den jährlichen Maximalabflüssen potenzielle Schäden aus einer Schaden-Wahrscheinlichkeit-Beziehung 

(Regression aus skalierten Events mit zugeordneten Eintretenswahrscheinlichkeiten 

und den analysierten Schäden) zugewiesen und (iii) diese Zeitreihe mit einem 

gleitenden Zeitfenster von 20 Jahren statistisch kumulativ ausgewertet. 

Abbildung 2: Schematische Darstellung der berücksichtigten Objektschutzmaßnahmen (a) wasserdichte Bauweise und (b) erhöhte 

Bauweise sowie die daraus resultierenden Modifikationen von Schadenfunktionen. Unter erhöhter Bauweise sind sowohl Aufschüttungen 

als auch aufgeständerte Bauweisen zu verstehen. 

ERGEBNISSE 

Im Rahmen der Gefahrenanalyse wurden ausgehend vom räumlich differenzierten Niederschlagsereignisses 

2002 (Skalierungsfaktor 1,0) fünf weitere synthetische Niederschlagsereignisse 

skaliert. Die Skalierungsfaktoren erstreckten sich von 0,4 bis 1,4, wobei der Skalierungsfaktor 

0,4 in etwa einem 1-jährlichen Abflussereignis und der Skalierungsfaktor 1,4 einem 

fast 400-jährlichen Abflussereignis entspricht. Aus den daraus abgeleiteten hydrologischen 

Szenarien wurden 24 hochaufgelöste Hochwassergefahrenkarten erstellt, je sechs Gefahrenkarten 

für die vier unterschiedlichen Siedlungsszenarien Sz1999, Sz2012, Sz_mod und 

Sz_max. 

Die Ergebnisse zeigen, dass es im Untersuchungsgebiet ohne OSM, oder weiteren Maßnahmen 

des Hochwasserrisikomanagements, zu einer Zunahme der Schäden über die Zeit 

kommen wird. Mit den zugrunde gelegten Analyseannahmen treten bei der Analyse eines 

100-jährlichen Ereignisses kaum abweichende Schäden zwischen Sz1999 und Sz2012 in 

Höhe von etwa EUR 8,8 Mio. auf, diese steigen bis ins Jahr 2030 auf EUR 10,1 Mio. (Sz_mod) 

bzw. EUR 14,4 Mio. (SZ_max) an. Dies entspricht einer Zunahme zwischen 2012 und 2030 

von 15 % bei Sz_mod und 64 % bei Sz_max. Für ein Ereignis mit einer Eintretenswahrscheinlichkeit 

von beinahe 400 Jahren steigt der analysierte Schaden von EUR 22,9 Mio. im 

Jahr 2012 auf (i) EUR 26 Mio. im Jahr 2030 bei Sz_mod und (ii) EUR 34 Mio. bei Sz_max an. 

Dies entspricht einer Zunahme von 13 % bzw. 48 % (siehe auch Abbildung 3, Ergebnisdarstellung 

OSM0). 

Da erst seit dem Jahr 2006 Gefahrenzonenpläne vorliegen, wurden in der weiterführenden 

Auswertung OSM auch erst ab 2006 berücksichtigt. Die Ergebnisse der szenarien-basierten 

Auswertung (Einzelereignisse definiert über Eintretenswahrscheinlichkeiten) sind in 

Abbildung 3 dargestellt. Die signifikantesten Nutzeneffekte sind bis zur Erreichung des 


Abbildung 3: Darstellung der Analyseergebnisse für den szenarien-basierten Ansatz für die Siedlungsszenarien (a) Sz1999, (b) Sz2012, 

(c) Sz_mod und (d) Sz_max unter Berücksichtigung von Objektschutzmaßnahmen. 

Schutzzieles (HQ 100 

+20 cm) zu verzeichnen. Wie zu erwarten, kommt es ab dem Zeitpunkt 

der Umsetzung von OSM zu keiner weiteren Erhöhung der Schäden unabhängig der Sied - 

lungsentwicklung bei Ereignissen mit einer Eintretenswahrscheinlichkeit bis zu 100 Jahren. 

Bei intensiveren aber selteneren Ereignissen (größer HQ 100 

) wird die Ausführung der OSM 

immer ausschlaggebender. Je intensiver ein Ereignis (ab HQ 100 

), desto geringer wird die 

potenzielle Wirkung der OSM1 (wasserdichte Bauweise), bei Ereignissen von beinahe HQ 400 

kann diese Wirkung nur noch als sehr gering eingestuft werden. Dahingegen behält die 

OSM2 (erhöhte Bauweise) ihre positive Wirkung relativ zu Bautätigkeiten ohne OSM auch 

bei Ereignissen größer HQ 100 

bei. Der potenzielle Nutzen unterschiedlicher OSM ab der 

Überschreitung des Schutzzieles ist somit verschieden. Beispielsweise sind die analysierten 

Schäden bei einem Ereignis von beinahe HQ 400 

(i) für das Szenario Sz_max, keine OSM, 

EUR 33,9 Mio hoch, (ii) für das Szenario SZ_max, OSM1, EUR 31,8. Mio. hoch und (iii) für 

das Szenario SZ_max, OSM2, EUR 24,6 Mio. hoch. Die kumulativen Schäden der stochastischen 

Auswertung an der aktuellen Siedlungsstruktur (Sz2012) zeigen eine große Bandbreite 

mit einer Standardabweichung von ± EUR 5 Mio. auf (siehe Abbildung 4), ein Ausdruck 

des zeitlich stochastischen Auftretens von Hochwasserereignissen. Ohne OSM ist mit einer 


entsprechenden Zunahme der Schäden über die Zeit bzw. in Abhängigkeit der betrachteten 

Siedlungsszenarien zu rechnen. Neben der Berücksichtigung von OSM ab 2006 (Abbildung 4 

(a)) wurde auch der (theoretische) Effekt von OSM im gesamten Gebäudebestand (Abbildung 

4 (b)) ausgewertet. Unabhängig von der Zunahme des Schadenpotenzials über die Zeit, ist ein 

konstanter Median der potenziellen Schäden bei Berücksichtigung der beiden betrachteten 

Varianten von OSM erkennbar. Dies begründet sich in der statistischen Auswirkung von 

vielen Ereignissen mit geringen Schäden im Vergleich zu wenigen Ereignisse mit sehr hohen 

Schäden. Die schadenmindernde Wirkung der beiden OSM-Varianten (OSM1 und OSM2) ist 

in der stochastischen Auswertung (Abbildung 4), ähnlich der Auswertung von Einzelszenarien 

(Abbildung 3), nachweisbar, der unterschiedliche Effekt jedoch nicht so deutlich akzentuiert. 

Abbildung 4: Darstellung der Analyseergebnisse für den stochastischen Ansatz mit einem Betrachtungszeitraum von 20 Jahren unter 

Berücksichtigung von Objektschutzmaßnahmen (a) ab 2006 bzw. (b) bei allen Gebäuden. 

FAZIT 

Mit der vorliegenden Arbeit kann aufgezeigt werden, dass OSM bis zu einem Grenzwert der 

Bemessung als sehr effiziente Hochwasserschutzmaßnahmen angesehen werden können. 

Ähnliche Schlussfolgerungen wurden bereits für das Risikomanagement alpiner Naturgefahren 

gezogen (z.B. Holub et al. 2012). Bei seltenen Extremereignissen ist die Ausführung 

entscheidend, wobei eine erhöhte Bauweise gegenüber einer wasserdichten Bauweise zu 

einer deutlich höheren Schadenminderung beiträgt. Bei der erhöhten Bauweise durch Aufschüttung 

ist jedoch zu beachten, dass es durch diese Maßnahme zu keiner Verschlechterung 

der Hochwasserabflusssituation für weitere Grundstücke kommen darf. Einschränkungen in 

der Aussagegüte der vorgestellten Ergebnisse ergeben sich aus dem vereinfachten Ansatz bei 

der Entwicklung von Siedlungsszenarien, den Unsicherheiten betreffend der Analysekomponenten 

und der Annahme, das sich hydro-meteorologische Erkenntnisse aus der Vergangenheit 

in die Zukunft übertragen lassen. Die Unsicherheiten umfassen vor allem (i) statistische 

Unsicherheiten in der hydrologischen Bearbeitung (Zeitreihen, statistische Methoden) sowie 


die den Siedlungsszenarien zugrunde gelegten Prognosen, (ii) Unsicherheiten im Hinblick auf 

das angewandte Gebäudebewertungsverfahren und (iii) Unsicherheiten in Bezug zu den 

berücksichtigten Schadenfunktionen. Die vorgestellten absoluten monetären Summen sind 

deshalb mit nicht-quantifizierten Unsicherheiten behaftet sind, ein relativer Vergleich und die 

Bewertung von Maßnahmen ist jedoch unabhängig davon möglich. 

Allgemein ist anzumerken, dass ausschließlich OSM nicht geeignet sind, um das Hochwasserrisiko 

bei gleichzeitiger Zunahme des Schadenpotenzials auf einem konstanten Niveau zu 

halten oder zur Minderung des Hochwasserrisikos beitragen zu können. Schäden können 

jedoch relativ zur Siedlungsentwicklung reduziert werden. Jegliche Siedlungsentwicklung in 

potenziell hochwassergefährdeten Gebiet trägt zuerst einmal zu einer Erhöhung des Risikos 

bei. Eine nachhaltige Reduzierung des Hochwasserrisikos kann nur im Rahmen des integralen 

Hochwasserrisikomanagements erfolgen, in einem Bündel aus Maßnahmen und 

politischen Instrumenten (Abflussminderung, Schutzmaßnahmen, Minderung/Reduzierung 

des Schadenpotenzials und der Vulnerabilität) (siehe z.B. Klijn et al. 2015). 

LITERATUR 

- ABI L288/27 (Amtsblatt der Europäischen Union): Richtlinie 2007/60/EG des Europäischen 

Parlaments und des Rates über die Bewertung und das Management von Hochwasserrisiken. 

- Achleitner, S., Huttenlau, M., Winter, B., Reiss, J., Plörer, M., Hofer, M. (in press): Temporal 

development of flood risk considering settlement dynamics and local flood protection 

measures on catchemnt scale: An Austrian case study, in: International Journal of River 

Basin Management. 

- Bundesamt für Umwelt, Wald und Landschaften BUWAL (Hrsg.) (1999): Risikoanalyse bei 

gravitativen Naturgefahren – Fallbeispiele und Daten, Umwelt-Materialien Nr. 107/II, Bern. 

- Hainka, A. (2010): Kleinräumige Bevölkerungsprognosen für Österreich2010 bis 2030 mit 

Ausblick bis 2050, Endbericht zur Bevölkerungsprognose, Österreichische Raumordnungs - 

konferenz (ÖROK), Wien. 

- Holub, M., Suda, J., Fuchs, S. (2012): Mountain hazards: reducing vulnerability by adapted 

building design, in: Environmental Earth Sciences, Volume 66, Issue 7, 1853-1870. 

- Kienholz, H. (2005). Analyse und Bewertung alpiner Naturgefahren – eine Daueraufgabe 

im Rahmen des integralen Risikomanagements, in: Geographica Helvetica, Issue 1/2005, 

3-15. 

- Klijn, F., Kreibich, H., de Moel, H., Penning-Rowsell, E. (2015): Adaptive flood risk 

management planning based on a comprehensive flood risk conceptualisation, in: Mitigation 

and Adaptation Strategies for Global Change, Volume 20, Issue 6, 845-864. 

- LBGL Nr. 35/2013: Landesgesetz über die bautechnischen Anforderungen an Bauwerke und 

Bauprodukte (Oö. Bautechnikgesetz 2013 - Oö. BauTG 2013). 

- Österreichische Raumordnungskonferenz ÖROK (2009): Szenarien der Raumentwicklung 

Österreichs 2030. Regionale Herausforderungen & Handlungsstrategien, Schriftenreihe 

Nr. 176/II, Wien. 



Risk-based spatial planning - findings from two 

case studies 

Roberto Loat¹; Reto Camenzind²; Esther Casanova³; Eva Frick 4 

ABSTRACT 

Event analysis shows that damages can be reduced by considering the risks involved with 

natural hazard zones, in particularly those exposed to lower levels of hazard. The purpose 

of risk-based spatial planning is to guide settlement development such that any risks are 

contained at an acceptable level to society in the long term. Current risks should be known 

and new, unacceptable risks should be avoided. It is possible to avoid hazardous areas when 

authorising new building zones or infrastructures, but there are great challenges in dealing 

with existing settlements. In such cases, greater care and attention must be paid to present 

and future risks. The earlier the stage at which risk-based spatial planning is incorporated into 

the planning process, the greater the potential for effective negotiations within the project, 

and the sooner effective counter-measures can be taken. However, the rationale behind 

risk-based spatial planning can only be applied successfully in practice through close col laboration 

between all of the parties involved, such as spatial planners, land-owners, natural 

hazard specialists and insurers. 

KEYWORDS 

risk governance; spatial planning; risk based spatial planning; natural hazards; natural hazards 

zones 

INTRODUCTION 

The risks associated with natural hazards are on the increase 

In Switzerland, a series of devastating floods has resulted in extensive damage in recent years. 

Other factors have also contributed to the increase in recorded damage. These include 

substantial population growth, the more intensive use of space as well as the increase in the 

value of buildings and infrastructures. Often, the greater risk and thus greater damage is to 

be found not in regions exposed to substantial and medium hazard levels, but in regions of 

intensive land use which face only low or residual hazard levels (marked in yellow and 

yellow-white hatched on hazard maps). In Switzerland, around one fifth of building zones 

are in at-risk areas. Approaches to handling these risks therefore play a key role in sustainable 

spatial development. 

1 Federal Office for the Environment FOEN, Bern, SWITZERLAND, roberto.loat@bafu.admin.ch 

2 Federal Office for Spatial Development ARE, Bern, SWITZERLAND 

3 Esther Casanova Raumplanung, Chur, SWITZERLAND 

4 tur gmbh, Davos Dorf, SWITZERLAND 



Implementing hazard maps demands risk awareness 

Hazard maps are available for 95% of the settled areas of Switzerland. They have been produced 

according to unified standards issued by the Federal Office for the Environment FOEN. 

Hazard maps display four hazard levels: red and blue areas denote a substantial or medium 

hazard respectively, while yellow illustrates a low hazard and yellow-white hatched areas a 

residual hazard. The hazard maps and the associated regulations must be implemented by the 

responsible authorities. In spatial planning, hazard maps offer a key instrument enabling 

communes to manage natural hazards and land use. In zones which are subject to a substantial 

hazard, it is standard practice to forbid the construction of new and the extension of 

existing buildings. Specific construction regulations are in place for zones facing a medium 

hazard. In zones with either a low or residual hazard, there is still no obligation to provide 

protection measures, although they could remain at considerable risk should a large-scale 

hazard event occur in densely populated areas, causing extensive and costly damage. 

Risk-based spatial planning therefore goes a step further. Alongside an assessment of hazard 

levels, this approach is designed to take current and future land use, and the associated risks, 

into consideration in spatial planning decision-making. The goal is to avoid any new, 

intolerable risks. This in turn means that protection measures in yellow and yellow-white 

zones should also be considered and implemented. The risk-based approach discussed herein 

is based on the ‘Security Level for Natural Hazards’ report (PLANAT, 2013). The report calls 

for the recommended security level to be achieved primarily by land use management, and 

demands in particular that new, unacceptable risks are avoided (Fig. 1). 

Figure 1: Method by which to achieve and maintain the recommended security level (PLANAT, 2013). 


Identifying conflict early and managing the development of risk 

Risk-based spatial planning does not stringently impose risk avoidance, but focuses on developing 

risk-awareness. The aim is not to block land use entirely, but to manage risk in 

a way that is transparent to those affected. In doing so it is possible to find meaningful and 

reasonable solutions to mitigate risk (Fig. 2). These solutions are specific to each case and 

may differ. In this respect, spatial planning plays a crucial role in providing solutions. Where 

new land use is concerned, alternative locations can be planned at a sufficiently early stage. 

In the case of established settlements, the existing risks can be identified and the relevant 

land use restrictions can be defined in partnership with those affected. 

Figure 2: Land use analysis: appropriate action to control the development of risk is determined by the initial conditions presented 

for spatial planning, land use potential and the specific natural hazard situation. 

Risk-based spatial planning relies not only on identifying the existing hazards in a given area, 

but also on pinpointing the risks that may arise from new or more intensive land use. 

When balancing interests, spatial planning should ensure that the frequency and impact of 

natural disasters affecting the people and property of the future are minimised. In this 

instance the role of spatial planning is to ensure that the demand for land use is balanced 

with the appropriate protection requirements. This requires all stakeholders to play an active 

part in the process. 


LAND USE PLANNING CASE STUDIES 

Evaluation of a broad spectrum of solutions 

In order to consolidate the concept of risk-based planning, the Federal Office for the Environment 

FOEN and the Federal Office for Spatial Development ARE commissioned two land use 

planning case studies to be carried out collaboratively by spatial planning and natural hazard 

experts, Casanova and tur gmbh (Casanova Raumplanung/tur gmbh, 2013), and Strittmatter 

and Partner AG (Strittmatter und Partner AG, 2012). Two communes which are affected by 

different hazard processes and hazard levels, and have a broad range of land uses, were 

selected for the case studies. To ensure that a wide range of solutions could be explored, 

fictitious but realistic examples of land use demands were adopted (Tables 1 and 2). 

The following questions were formulated by the FOEN and the ARE to be investigated in the 

case studies: 

– How can spatial planning tools be applied to achieve risk-appropriate land use in accordance 

with the hazard process and level in a given case? 

– What spatial planning tools are available to the selected communes, and how can they 

be deployed to ensure that known risks are respected in the planning process? 

What synergies exist between protection strategies and other tools of risk prevention? 

Table 1: The comparative case studies for each canton, investigating both natural hazards and given types of land use. 

Case studies Commune in Canton St. Gallen Commune in Canton Graubünden 

Hazard process 

Process intensity 

Advance warning time 

Current land use ‐ 

Decision-making tree facilitates a systematic approach 

To ensure a systematic approach during the process of spatial planning, the experts executing 

the case studies developed a decision-making tree featuring the relevant decision-making 

criteria and options for action (Fig. 3). The complete decision tree can be found in the 

summary publication issued by the bodies involved in the case studies (PLANAT/BAFU/ARE 

(2014). This report also sets out key information on the principal concepts of risk-based 

spatial planning, decision-making criteria, and the appropriate courses of action. 

Certain aspects of a planning project must be assessed before the decision-making tree is 

applied. These are whether a planned land use encroaches on a hazard zone, whether 

necessary information about hazard processes (hazard maps, intensity maps) is available and 

up to date, and finally whether this information contains sufficient detail, or must be 

supplemented. Current and planned land use must be examined in the context of land use 

demand although, depending on the project, this information may not be available until the 

(special) land use planning process or the building permit procedure. The only effective 

means of assessing potential damage and risks is to overlay project plans with hazard maps 

and detailed hazard assessments. 


Table 2: The comparative case studies with their respective spatial planning procedures applied, points of conflict and suggested 

solutions. 

 

 

 

 

‐ 

 

 

 

 

 

 

‐ 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

‐ 

 

 

 

 

 

 

‐ 

‐ 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

‐ 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

‐ 

 

 

 

Following this situation analysis, the decision-making tree guides the user through all of the 

necessary decisions. In addition to questions concerning risk analysis, it is essential to balance 

different spatial planning interests. It may be that a protection measure is technically feasible 

and cost-effective, but it must still meet design and acceptance criteria. Aspects of land use 

such as water protection, as well as landscape and nature conservation, must also be considered. 

The spectrum of possible implications for spatial planning is very broad, ranging from 


Figure 3: Schematic representation of the most relevant decision-making criteria and options for action of the decision-making tree 

(extract from PLANAT/BAFU/ARE, 2014). 


providing information to the landowner, without further action on the part of the authorities, 

through to a full ban on land use. 

Generally speaking, the sooner in the planning process that risk-based spatial planning is 

applied, the greater the leeway for negotiation within the project as a whole. For example, 

the following steps can still be taken in the early planning stages: 

– New zoning in at-risk areas can be avoided; 

– Risks can be assessed and alternative locations considered; 

– Land use can be adapted in the best way to the prevailing risks and land use restrictions; 

– There can be an early discussion of residual risks and their acceptability. This is particularly 

important for land use entailing high risks to individuals or other special risks. 

At a later stage in the process, such as when a building permit has already been issued, the 

room for manoeuvre is often significantly smaller. This is mainly because changes to the 

planned project can involve a disproportionate amount of effort. It is true, however, that in 

some cases it is only at this time that the true land use is known in detail. Ultimately, it is the 

type of land use that dictates the measures required to mitigate risks. Risk-based spatial 

planning may still be of value in such situations. Those affected may be made aware of the 

associated risks and will thus be able to take meaningful action. The decision-making tree 

was designed to offer specific, practicable recommendations for different situations. The three 

examples described below show possible counter-measures that can be implemented in 

different phases of spatial planning: 

– Single property protection: protection measures applied directly to a property can reduce its 

vulnerability, and thus the damage caused during an event (e.g. reinforced construction); 

– Multiple-property protection: if protection measures can be applied to multiple objects 

simultaneously (e.g. a diversion dam installed along several buildings), there may be 

advantages in terms of both design and cost-effectiveness. Special land use planning is a 

tool that allows binding measures to be imposed on land-owners; 

– Emergency planning: with sufficient advance warning time, clearly defined intervention 

measures taken by emergency personnel or local residents can minimise property damage. 

FINDINGS FROM THE CASE STUDIES 

The case studies have shown that the scope for negotiation in spatial planning depends 

heavily upon whether the case involves new building zones, or the intensification or 

modification of existing land use. For new land use projects, it may be possible to negotiate 

an alternative local or regional location, for example. There is less opportunity to influence 

the situation where land use is to be intensified within an existing land use zone. 

It is essential that current and complete hazard maps and information are available to facilitate 

an adequate assessment of the risks involved. Furthermore, since risk-based planning should 


factor in a variety of information in addition to hazard maps, the process also requires further 

data such as scaled intensity maps and risk maps. 

To ensure that risk-based planning is executed effectively, close cooperation between spatial 

planners and natural hazard experts must be established at an early stage. This is crucial both 

at this planning stage, and in the later implementation process. 

The action which should be taken in a particular situation depends on various factors. 

The specific hazard process plays an important role of course. Where this is gradual, there is 

usually sufficient advance warning time to evacuate persons and, if possible, property. 

Furthermore, intervention measures can reduce the extent of damage should an event occur. 

In the case of sudden events, there is little or no advance warning time. Here, it is important 

to consider the effectiveness of protection measures to cope with the intensity of such events. 

The key factor to remember is that protection measures must always be selected according to 

how the land is used. 

CONCLUSIONS 

The two land use planning case studies have been used to take initial development steps 

towards a systematic approach to risk-based spatial planning. The new method facilitates the 

effective implementation of measures designed to reduce risk at both the level of local land 

use planning, and as part of the building permit procedure. 

At the same time, interest in risk-based spatial planning must be generated. The need for 

greater awareness in handling hazards and risks must be demonstrated to representatives of 

communal and cantonal authorities, planning and engineering offices and insurance 

companies. Furthermore, greater support must be given for close cooperation between spatial 

planners and natural hazard experts, with the inclusion of those who are affected. 

The concept of risk-based spatial planning should not be applied to land use planning alone, 

but to every stage of the spatial planning process, i.e. cantonal structural plans, land use 

planning and the building permit procedure. 

The next step is to take the findings from this and other ongoing risk-based spatial planning 

projects and use them as input into a new working guide or the revision of the ‘Spatial 

Planning and Natural Hazards’ recommendation (ARE / BWG / BUWAL, 2005). First of all, 

however, any outstanding issues should be examined in more depth, and the current 

methodology should be applied and tested in further practical examples. 


REFERENCES 

- ARE/BWG/BUWAL (2005). Recommendation. Spatial Planning and Natural Hazards. 

Federal Office for Spatial Development ARE / Federal Office for Water and Geology BWG / 

Swiss Agency for the Environment, Forests and Landscape BUWAL, Bern. 

- Bundesgesetz vom 21. Juni 1991 über den Wasserbau (Wasserbaugesetz, WBG 1991, 

SR 721.100) 

- Casanova Raumplanung/tur gmbh (2013). Risikobasierte Raumplanung in der kommunalen 

Nutzungsplanung. Testplanung 2 (PLANAT Project A6) 

- PLANAT (2013). Security Level for Natural Hazards. National Platform for Natural Hazards 

PLANAT, Bern. 

- PLANAT/BAFU/ARE (2014). Risikobasierte Raumplanung. National Platform for Natural 

Hazards. PLANAT, Federal Office for the Environment FOEN, Federal Office for Spatial 

Planning ARE, Bern. 

- Strittmatter und Partner AG (2012). Risikobasierte Raumplanung in der kommunalen 

Nutzungsplanung. Testplanung 1 (PLANAT Project A6) 



Public prevention of natural hazards in Swiss law: 

responsibilities and fields of action in legislative and 

executive bodies 

Staatliche Naturgefahrenprävention im Schweizer 

Recht: Verantwortlichkeiten, legislative und 

verwaltungsorganisatorische Handlungsfelder in 

rechtlicher Perspektive 

Roland Norer, Prof. Dr. iur.¹; Mark Govoni, lic. iur.²; Gregor Kost, Attorny at Law; Cornel Quinto, Attorney at Law, LLM³; 

Barbara Schielein, MLaw 1 ; Lukas Widmer, Attorney at Law; Christian Wulz, MLaw 

ABSTRACT 

The presented research project at hand on natural hazards law deals with the manifold legal 

aspects in context with the handling of risks of natural hazards. The questions of law to be 

answered focus on the grounds and the extent of state responsibility based on the legal 

doctrine on the duty to protect derived from the fundamental rights at the one hand, and on 

the margin of action of the state when it comes to legislation and implementation with 

particular regard to integrated risk management and aspects of insurance law at the other 

hand. 


Das hier vorgestellte Forschungsprojekt zum Naturgefahrenrecht beschäftigt sich mit den 

vielfältigen rechtlichen Aspekten des Umgangs mit Naturgefahrenrisiken. Die zu klärenden 

Rechtsfragen fokussieren zum einen auf Begründung und Umfang staatlicher Verantwortlichkeit 

aus der Schutzpflichtendogmatik zu den Grundrechten, zum anderen auf die staatlichen 

Handlungsfelder in Gesetzgebung und Vollzug unter besonderer Berücksichtigung des 

integralen Risikomanagements und versicherungsrechtlichen Aspekten. 

KEYWORDS 

government liability; doctrine on the duty to protect; Integrated risk management; division of 

competences; building insurance 

1 University of Lucerne, SWITZERLAND, roland.norer@unilu.ch 

2 Federal Office for the Environment FOEN, SWITZERLAND 

3 Lustenberger Attorneys at Law, Zurich, SWITZERLAND 



EINFÜHRUNG 

Das durch den Schweizerischen Nationalfonds geförderte Forschungsprojekt „Naturgefahrenrecht. 

Grundlagen für Rechtshomogenität und -effizienz im Naturgefahrenmanagement“ 

an der Rechtswissenschaftlichen Fakultät der Universität Luzern beschäftigt sich mit den 

rechtlichen Aspekten des Umgangs mit Naturgefahrenrisiken in der Schweiz. Der Schwerpunkt 

liegt dabei auf der Frage nach den juristischen Handlungsfeldern staatlicher Gefahrenund 

Risikoprävention bzw. -reduktion. 

METHODEN 

Innerhalb des Forschungsprojektes kommen verschiedene juristische Auslegungsmethoden 

[wie die grammatikalische (nach dem Wortlaut), systematische (nach dem Regelungszusammenhang) 

und historische (nach dem Willen des Gesetzgebers)] zur Anwendung, mit denen 

zugrundeliegenden Rechtsprinzipien und möglichen Interpretationen unbestimmter 

Rechtsbegriffe nachgespürt wird. Einen besonderen Stellenwert nimmt der (kantonale) 

Rechtsvergleich ein. 

1. Umfang staatlicher Verantwortlichkeit 

In einem ersten Schritt wird nach den Aufgaben des Staates gefragt und inwieweit er für den 

Schutz vor Naturgefahren verantwortlich gemacht werden kann. Die Grenze zwischen der 

staatsrechtlichen Verpflichtung [EGMR Budayeva u.a. gegen Russland, Urteil vom 20.3.2008, 

Kolyadenko u.a. gegen Russland, Urteil vom 28.2.2012] und der legitimen Eigenvorsorge der 

Bürger ist schwierig zu bestimmen. 

Weitgehend unbestritten ist, dass Grundrechte nicht nur als Abwehrrechte fungieren, 

sondern darüber hinaus den Staat dazu verpflichten, seiner grundrechtlichen Schutzpflicht 

aktiv nachzukommen. Es ist herrschende Lehre, dass prinzipiell alle Grundrechte für die 

Schutzpflichtendogmatik fruchtbar gemacht werden können. Für Naturgefahren relevant 

sind insbesondere das Recht auf Leben und körperliche Unversehrtheit und die Eigentumsgarantie 

[Art. 10 und Art. 26 der Schweizerischen Bundesverfassung]. 

Fraglich ist dagegen, wie die Schutzpflichten umzusetzen sind und wo deren Grenzen liegen. 

Klärungsbedarf besteht zudem beim Schutzbereich und dessen Tragweite. Leitet man – wie 

dies überwiegend getan wird – die Schutzpflichten objektiv-rechtlich her, ist der Schutz 

kommender Generationen innerhalb der Schutzpflichtendogmatik. Dies ist insbesondere für 

die Problematik des Klimawandels eine wichtige Feststellung. 

Die deutsche Rechtsprechung, die die Schutzpflichtendogmatik neben dem EGMR maßgeblich 

geprägt hat, spricht wiederholt davon, dass die Schutzpflichten „vor allem“ und „insbesondere“ 

vor Übergriffen Dritter gelten [BVerfGE 39, 1 (42); 53, 30 (57)]. Folgerichtig sind 

menschverursachte Naturgefahren wie Hochwasser – durch den Klimawandel oder die 

Verbauung von Retentionsräumen bedingt – von den Schutzpflichten erfasst. Weniger klar 


ist, ob natürliche Gefahren [also die, die nicht vom Menschen verursacht oder beeinflusst 

werden wie z.B. Erdbeben] innerhalb der Schutzpflichtendogmatik liegen. Diejenigen, die 

dies verneinen [Krings, S. 219], sehen darin eine faktische Überforderung des Staates, der 

ohnehin nichts gegen Naturgewalten ausrichten könne. Die Gegenmeinung [Schmitz, 

S. 109f.] fokussiert zu Recht auf die Tatsache, dass es nicht darum gehe, die Natur zu be herr - 

schen, sondern darum, die möglichen Schadensauswirkungen mit entsprechenden Maßnahmen 

abzuschwächen oder zu verhindern. 

Auch zur Frage, ob die grundrechtlichen Schutzpflichten uneingeschränkt gelten, teilen sich 

die Meinungen [zum Ganzen: Dietlein, S. 108, 118f.]. Die Befürworter sehen ausgehend von 

der Struktur der Schutzpflichten als Optimierungsgebote einen umfassenden Geltungsbereich 

und relativieren diesen an anderer Stelle. Der Gegenmeinung nach bedarf es bereits im 

Schutzbereich gewisser Voraussetzungen, um die Schutzpflichten aktivieren zu können. 

Das entscheidende Kriterium für die Anerkennung grundrechtlicher Schutzpflichten liegt 

gemäß schweizerischem Bundesgericht [BGE 119 Ia 28 E. 2 S. 31 Rue des Eaux-Vives] in der 

Intensität der Betroffenheit. Der EGMR bewertet Schutzpflichten dort als umfassend, wo mit 

besonders schweren Verletzungen zu rechnen ist; dies ist regelmäßig dann der Fall, wenn das 

Recht auf Leben tangiert ist. Kriterium ist das potentielle Schadensausmaß und die Wahrscheinlichkeit 

des Schadenseintritts. Je gefährdeter ein gewichtiges Rechtsgut ist, desto 

weniger wahrscheinlich muss der Schadenseintritt sein. Für die Naturgefahrenprävention 

bedeutet dies, dass bereits bei der Möglichkeit, dass Gefahr für Leib und Leben besteht, auch 

wenn diese äußerst unwahrscheinlich ist, die Schutzpflichten greifen. 

Staatliche Präventionsleistungen bemessen sich daher nach Art und Qualität des Schutzgutes 

und müssen im Rahmen des rechtlich und tatsächlich Möglichen liegen. Ein Anspruch auf 

einen bestimmten Schutz existiert nicht. Bezüglich der Umsetzung der Schutzpflichten 

besteht vielmehr ein weiter Gestaltungsspielraum. Beispielsweise ist die Bevölkerung über 

Naturgefahren zu informieren, was auch einen erheblichen Beitrag zur Stärkung der Eigenvorsorge 

bedeutet. Was vom Staat genau als Schutzleistung zu erbringen ist, ist letztlich in 

einem politischen Diskurs zu ermitteln. 

2. Staatliche Handlungsfelder 

In einem zweiten Schritt sind die staatlichen Handlungsfelder und damit die Kompetenzen 

und Rechtsmaterien zu definieren. Die vorhandenen hauptsächlichen Instrumente werden 

im Wesentlichen im Raumplanungsrecht, Baurecht, Wasserrecht, Forstrecht, Infrastrukturrecht 

und Katastrophenrecht geregelt. 

2.1. Integrales Risikomanagement 

Als übergeordnetes Konzept wird das moderne Naturgefahrenrecht insbesondere an seiner 

Kompatibilität mit staatlichen Rechtsetzungs- und Vollzugsaktivitäten mittels des für die 

Verwaltungsrechtswissenschaft neuartigen Instruments des Integralen Risikomanagements 

[IRM] zu prüfen sein. Dieser risikobasierte Ansatz, der treffend mit dem Leitsatz „Von der 


Gefahrenabwehr zur Risikokultur“ umschrieben wird, beschäftigt sich neben der grundlegenden 

Frage der Effizienz von ordnungsrechtlichen und freiwilligen Handlungsformen mit 

Maßnahmen der Prävention, Intervention und Regeneration. 

Befasst man sich mit den rechtlichen Aspekten der IRM-Strategie, fällt auf, dass das Begriffspaar 

Gefahr und Risiko bzw. deren Abgrenzung voneinander im Rahmen des IRM nicht mit 

dem klassischen polizeirechtlichen Begriffsverständnis, das auf die „hinreichende Wahrscheinlichkeit” 

eines Schadenseintritts abstellt, übereinstimmt. Während dort der Unterschied 

zwischen Gefahr und Risiko ein quantitativer ist, kommt im Rahmen der IRM-Strategie ein 

qualitativer hinzu. Das Risiko setzt sich demnach aus der Gefahr einerseits und der Verletzlichkeit 

andererseits zusammen. Im Unterschied zum Risiko sagt die Gefahr somit nichts über 

die Verletzlichkeit bzw. das Schadensausmaß aus. Von praktischer Bedeutung ist dieses 

Begriffsverständnis insbesondere im Zusammenhang mit der Erstellung von Gefahren- und 

Risikokarten, auf deren Grundlage die Schutzmaßnahmen geplant und umgesetzt werden 

sollen. Während sich die Gefahrenkarten auf die Darstellung von Ausmaß und Intensität der 

Prozesse beschränken, machen die Risikokarten auch Aussagen über das Schadenspotenzial 

[vgl. dazu Art. 6 EU-Hochwasserschutzrichtlinie 2007/60/EG]. Damit spricht vieles dafür, im 

Naturgefahrenrecht ein von der polizeirechtlichen Dogmatik abweichendes Verständnis der 

Begriffe „Gefahr“ und „Risiko“ einzuführen. Voraussetzung dafür ist jedoch, dass die 

betreffenden Begriffe im Naturgefahrenrecht definiert und in den Materialien zu den 

betreffenden Gesetzesanpassungen erläutert werden. 

2.2. Rechtsetzung 

Was die Rechtsetzung betrifft, ist auch im Schweizer Recht eine Rechtszersplitterung zu 

konstatieren. Da eine zentrale Kompetenznorm für Naturgefahren fehlt, sind Rechtsmaterien 

unterschiedlicher Zielsetzung und Struktur zum Teil gemeinsam anzuwenden. Dazu kommt 

noch die Verteilung der relevanten Rechtsetzung auf bis zu drei Ebenen [Bund – Kantone 

– Gemeinden]. Das Projekt identifiziert das System dieser Bestimmungen, fragt nach 

Normenhierarchien und -derogationen, verbindenden Elementen [z.B. Berücksichtigung 

kantonaler Gefahrenkarten in kommunalen Raumplänen] und erarbeitet Legislativvorschläge. 

Im Ergebnis kann auf eine zentrale bundesverfassungsrechtliche Norm und ein Naturgefahrengesetz 

verzichtet werden, wenn die Rechtsbereiche sich klar voneinander abgrenzen 

lassen und ein homogenes legislatives Gesamtsystem im Mehrebenensystem mit der 

Respektierung regionaler und kommunaler eigenständiger Lösungen gelingt. 

Dabei richtet sich die Kompetenzverteilung nach dem Grundsatz der Subsidiarität. 

Die Rechtsetzungskompetenzen des Bundes beschränken sich somit grundsätzlich auf den 

Schutz vor denjenigen Gefahren, vor denen sich die Kantone nicht alleine schützen können. 

Das sind insbesondere die Gefahren des Wassers [Art. 76 Abs. 1 BV] und die Gefahren, 

welche über die Schutzfunktion des Waldes abgewehrt werden können [Art. 77 Abs. 1 BV]. 

Was den Schutz vor Erdbeben und Hagel sowie allfälligen weiteren Naturgefahren betrifft, 

hat der Bund jedoch keine Rechtsetzungskompetenzen. Allfällige Schutzmaßnahmen gegen 

diese Gefahren entziehen sich somit grundsätzlich einer einheitlichen Bundesregelung. 


Stufe BV 

Art. 61 

(Zivilschutz) 

Art. 75 

(Raumplanung) 

Art. 76 

(Wasser) 

Art. 77 

(Wald) 

Stufe Bund 

BG 

Bevölkerungs- und 

Zivilschutz 

BG 

Raumplanung + VO 

BG 

Wasserbau + VO 

BG 

BG 

Wald 

Gewschutz + VO 

Stufe Kanton 

Kant. 

Bevölkerungsschutz 

gesetz 

Kant. 

Planungs- und 

Baugesetz 

Kant. 

Wasserbau- und 

Gewässerschutz- 

Kant. 

Waldgesetz 

gesetz 

Abbildung 1: Diese Abbildung zeigt die wesentlichen Rechtsgrundlagen auf den tangierten Rechtsetzungsebenen auf 

Wichtig erscheint eine Abkehr vom gefahrenabwehrbasierten Sicherheitsansatz hin zu einem 

legislativen Risikokonzept, wie das die EU-Hochwasserschutzrichtlinie 2007/60/EG vorexerziert 

hat. Der risikobasierte und probabilistische Regelungsansatz beschränkt sich im 

Gegensatz zum präskriptiven und deterministischen Ansatz darauf, ein maximal zulässiges 

Risiko in quantitativer Form [Schutzziel] festzulegen. Die Normadressaten sind grundsätzlich 

frei, mit welchen Mitteln sie diese Vorgaben einhalten. Im Gegensatz zum deterministischen 

Ansatz geht er davon aus, dass Risiken nie völlig eliminiert, wohl aber optimiert werden 

können und nimmt gewisse Risiken deshalb bewusst in Kauf, versucht diese aber nach 

rationalen Kriterien zu begrenzen. Mit dem im Rahmen der Neugestaltung des Finanzausgleichs 

und der Aufgabenteilung zwischen Bund und Kantonen [NFA] auch im Umweltschutzbereich 

eingeführten Instrument der Programmvereinbarung [Art. 20a Subventionsgesetz], 

wurde die Grundlage geschaffen, aufgrund welcher der Bund, die Schutzmaßnahmen 

der Kantone anhand ziel- und risikobasierter Kriterien subventionieren kann. Die rechtliche 

Verankerung der IRM-Strategie wurde vom Gesetzgeber demnach bereits teilweise vorgespurt. 

Letztlich ermöglicht erst der Übergang zum Paradigma des Risikomanagements eine kontinuierliche 

und ganzheitliche gesellschaftliche Analyse, Bewertung und Verringerung der 

Risiken. 

2.3. Vollzug 

Was den Vollzug anbelangt, untersucht das Forschungsprojekt verschiedene organisatorische 

und verwaltungsverfahrensrechtliche Modelle auf ihren Beitrag zu einer effizienten Umsetzung. 

Der Schutz vor Naturgefahren ist eine Verbundaufgabe, d.h., dass sich regelmäßig 

zahlreiche Akteure aller drei Staatsebenen, weitere öffentlich-rechtliche Körperschaften des 


kantonalen Rechts, Werk-, Anlage- und Konzessionsinhaber sowie weitere Privatpersonen 

[z.B. Grundeigentümer und Anstößer] mit dieser Thematik befassen und bestimmte Aufgaben 

wahrnehmen. Es ist somit zwingend erforderlich, dass Organisation bzw. Zusammenwirken 

der verschiedenen Akteure im Rahmen der Aufgabenerfüllung so gut wie möglich optimiert 

werden. 

Der eigentliche Vollzug des Naturgefahrenrechts erfolgt schwergewichtig auf kantonaler 

Ebene. Dabei präsentiert sich das kantonale Organisations- und Verfahrensrecht im Bereich 

der Naturgefahrenprävention sehr heterogen. Es bestehen in der Schweiz in diesem Bereich 

daher 26 unterschiedliche Zuständigkeitsordnungen und völlig verschiedene Verfahrensabläufe. 

Dabei können einerseits die horizontale [Organisation und Kompetenzverteilung auf 

kantonaler Ebene] und andererseits die vertikale Aufgabenverteilung [Nebeneinander von 

kantonalen, kommunalen Kompetenzen sowie von Kompetenzen von weiteren Personen des 

öffentlichen und privaten Rechts] unterschieden werden. Daneben spielen auch die Verfahren, 

in welche die Maßnahmen der Naturgefahrenabwehr eingebettet sind, sowie die 

bestehenden Koordinationsinstrumente für die Umsetzung des materiellen Naturgefahrenrechts 

eine zentrale Rolle. Sie sollen praktikabel, transparent und nicht unnötig kompliziert 

sein, eine umfassende Interessenabwägung und Koordination ermöglichen sowie die 

Bevölkerung möglichst frühzeitig mittels entsprechender Mitwirkungsrechte in die Entscheidfindung 

involvieren. Des Weiteren ist der Rechtsschutz umfassend sicherzustellen. 

Grundsätzlich ist eine erhebliche Vielzahl unterschiedlicher organisationsrechtlicher Möglichkeiten 

denkbar. So besteht die Möglichkeit, das kantonale Organisationsrecht im Bereich der 

Naturgefahrenprävention und somit namentlich in den vier Teilbereichen Gefahrengrundlagen 

[inklusive Berücksichtigung in der Raumplanung sowie im Baubereich (I)], baulicher 

Hochwasserschutz und Gewässerunterhalt (II), bauliche Schutzmaßnahmen gemäß Waldgesetz 

(III) und Gewässerräume (IV) „zentralistisch“ (1) auszugestalten und sämtliche oder die 

Mehrheit der entsprechenden Kompetenzen dem Kanton selber bzw. kantonalen Behörden 

zuzuweisen. Ferner ist es auch denkbar, den Gemeinden die größtmögliche Autonomie zu 

gewähren und ihnen einen Großteil der Aufgaben auf dem Gebiet der Naturgefahrenabwehr 

zuzuweisen [sog. „gemeindebezogener Ansatz“ (2)]. Als weitere Möglichkeit fällt die 

Zuweisung diverser Aufgaben im Bereich des Naturgefahrenrechts an die unmittelbaren 

Nutznießer von Maßnahmen gegen Naturgefahren, die Grundeigentümer und Werk-, 

Anlage- oder Konzessionsinhaber (sog. „nutznießerbezogener Ansatz“ (3)]. Diesfalls würde 

die einhellige Auffassung, dass die Naturgefahrenabwehr eine Staatsaufgabe darstellt, 

bedeutend relativiert. Ein „differenziertes“ Organisationsmodell (4) ist zudem die Verteilung 

der Kompetenzen an verschiedene Träger entsprechend deren Eignung für die betreffende 

Aufgabe. 

Je nach Größe sowie Topografie des Gebiets sowie aufgrund der dort vorhandenen Gefahren 

vor Naturgewalten legen die Kantone die Organisation im Bereich der Naturgefahrenabwehr 

fest. Des Weiteren spielen oftmals historische Entwicklungen eine bedeutende Rolle für die 


Ausgestaltung der Zuständigkeiten. Zudem kann auch die politische Grundhaltung im Kanton 

ein maßgebender Faktor sein [z.B. Haltung, dass der Staat umfassende Schutzpflichten inne 

hat oder die gegenteilige Auffassung, dass die Privaten selber für ihre Sicherheit zu sorgen 

haben]. Es ist mehr als fraglich, ob es ein „richtiges“ Organisationsmodell im Bereich der 

Naturgefahrenabwehr gibt. Vielmehr kann je nach den kantonalen Verhältnissen das eine 

oder das andere Modell als sachgerecht bezeichnet werden. Übergeordneten Grundsätzen 

[Subsidiaritätsprinzip, effizientes Staatshandeln, Koordinationsprinzip etc.] ist stets Rechnung 

zu tragen. 

Die Koordination zwischen den verschiedenen involvierten Trägerschaften kann jedenfalls 

insbesondere durch behördenverbindliche Grundlagen in der kantonalen Richtplanung, 

andere übergeordnete Planungen, kantonale Nutzungspläne, spezifische gesetzliche Vorschriften, 

konzentrierte Bewilligungsverfahren oder kantonale Prüfungs- und Genehmigungserfordernisse 

optimiert werden. Einen Schwerpunkt des Projekts bildet dabei auch die Frage, 

inwieweit die obere Ebene [also Bund im Verhältnis zu den Kantonen oder Kantone im 

Verhältnis zu den Gemeinden] vollzugslenkende Instrumente einsetzt. Hier ist zu bemerken, 

dass sehr oft nicht zu rechtlich verbindlichen Vorgaben gegriffen wird sondern untergesetzliche 

Elementen wie Richtlinien, Empfehlungen, Merkblätter oder technischen Normen zum 

Einsatz kommen. Den unbestreitbaren Vorteilen solcher Instrumente im Bereich der 

Aktualität und Flexibilität stehen aber grundlegende rechtliche Bedenken gegenüber. 

2.4. Versicherung 

In der Schweiz wird die Gebäudeversicherung gegen Elementarschäden durch zwei Systeme 

sichergestellt. In 19 Kantonen bei den kantonalen Gebäudeversicherern KGV, die selbständige, 

öffentlich-rechtliche kantonale Anstalten sind, welche über ein rechtliches Monopol 

verfügen und in kantonalen Gesetzen, teilweise zusätzlich in Kantonsverfassungen verankert 

sind. In 7 Kantonen mittels des GUSTAVO-Systems [Akronym der Anfangsbuchstaben der 

jeweiligen Kantone], welches den Schutz mittels Privatversicherungen erbringt und bezüglich 

Elementarschäden im Versicherungsaufsichtsgesetz des Bundes und der Aufsichtsverordnung 

verankert ist. 

Volkswirtschaftlich sowie sozial- und ordnungspolitisch ist die Gebäudeversicherung von 

großer Bedeutung, denn sie sichert den raschen Wiederaufbau, die Erhaltung der Wirtschaftsleistung 

[Gewerbe- und Industriebauten], wesentliche wenn nicht sogar existentielle 

Vermögensteile [v.a. Einfamilienhäuser Privater] sowie geeignete bauliche Maßnahmen in 

Risikogebieten. 

Ein Versicherungssystem ist der ad-hoc Staatshilfe im Katastrophenfall entschieden vorzuziehen. 

Neben den obigen Vorteilen schont es auch den allgemeinen Staatshaushalt, da nicht auf 

allgemeine Steuermittel zur Entschädigung von beispielsweise Hochwasseropfern zurückge- 


griffen werden muss. Diese haben zudem einen Rechtsanspruch auf Versicherungsleistung 

und nicht bloß eine vage und letztlich unsichere Aussicht auf staatliche Hilfeleistung. 

Für die Zukunft ist eine weitere Stärkung der Prävention wünschbar und nötig, um die 

Versicherbarkeit auch angesichts der Herausforderung Klimawandel langfristig sicherzustellen. 

So muss der Naturgefahr Starkregen in Zusammenhang mit dem Oberflächenabfluss 

vermehrt Beachtung geschenkt werden. Weiter sollten den KGV weitere Instrumente in die 

Hand gegeben werden wie der Einfluss auf Baubewilligungen oder konkrete Auflagen bei 

solchen [was heute nur in vereinzelten Kantonen möglich ist]. 

ERGEBNISSE 

Den Staat treffen unter gewissen Umständen zum Schutz vor Naturgefahren Handlungspflichten, 

wobei verschiedene Parameter wie Intensität der Betroffenheit, potentielles 

Schadensausmaß oder Wahrscheinlichkeit des Schadenseintritts zu berücksichtigen sind. 

In der Rechtsetzung wird angesichts der Rechtszersplitterung ein homogenes legislatives 

Gesamtsystem sowie ein vollständiger Paradigmenwechsel vom bisher vorherrschenden 

Sicherheitsansatz zum Risikoansatz zu fordern sein. Schließlich werden exemplarisch diverse 

Modelle aus der verwaltungsorganisatorischen Praxis untersucht, die tendenziell nicht 

zwingend für eine zentrale Behörde, in der alle Vollzugskompetenzen gebündelt werden, 

sondern zumindest für eine überschaubare kleinere Zahl an fachlich spezialisierten oder 

regional strukturierten Behördeneinrichtungen sprechen, die sich nach klaren Regeln selbst 

koordinieren oder von einer Koordinationsbehörde gesteuert werden. Das schweizerische 

System der Gebäudeversicherung ist nicht einforderbaren staatlichen Unterstützungen 

vorzuziehen. 

FAZIT 

Die Behandlung grundlegender juristischer Fragestellungen in Bezug auf staatliche Handlungspflichten, 

Legislative und Exekutive sowie Versicherungsrecht kann einen entscheidenden 

Beitrag zu einem effizienten Naturgefahrenmanagement leisten. 

LITERATUR 

- Dietlein J. (2005), Die Lehre von den grundrechtlichen Schutzpflichten 

- Fuchs S./Khakzadeh L./Weber K. (Hrsg.) (2006), Recht im Naturgefahrenmanagement 

- Hess J. (2011), Schutzziele im Umgang mit Naturrisiken in der Schweiz 

- Jaeckel L./Janssen G. (Hrsg.) (2012), Risikodogmatik im Umwelt- und Technikrecht 

- Krings G. (2003), Grund und Grenzen grundrechtlicher Schutzansprüche 

- PLANAT (2013), Sicherheitsniveau für Naturgefahren 

- Quinto C. (2010), Versicherungssysteme in Zeiten des Klimawandels 

- Rudolf-Miklau F. (2009), Naturgefahren-Management in Österreich 

- Schmitz S. C. (2010), Grundrechtliche Schutzpflichten 

- Seiler H. (2000), Risikobasiertes Recht. Wieviel Sicherheit wollen wir? 



Gender mainstreaming in disaster risk reduction – 

a step towards the visibility of women in DRR 

Catrin Promper, Dr. 1,2 ; Maria PatekPatek 1 

ABSTRACT 

Women and men perceive risks differently and therefore gender is an important part of 

disaster risk reduction requiring further consideration. Experiences of large disasters have 

shown that the role of women is significantly different from the one of men in the phases of 

prevention and disaster response. To further enhance resilience it is important to proactively 

include women in socio-political decision-making processes and to ensure the inclusion of 

women and youth-specific issues in disaster risk management plans. Therefore, the platform 

on Natural Hazards of the Alpine Convention - PLANALP - has included this topic in the 

current mandate 2015-2016. The focus is to raise awareness of gender issues in disaster risk 

reduction and to foster women's engagement by enhancing the visibility of female experts in 

this area of work. 

KEYWORDS 

disaster risk reduction;gender mainstreaming; policy; PLANALP 

INTRODUCTION 

Natural hazard risk is perceived differently by women and men and, consequently, differences 

of gender are a significant part of disaster risk reduction. This is also underlined by the 

recently adopted Sendai Framework for Disaster Risk Reduction 2015-2030 (UN 2015). 

With respect to natural hazards, often women are more vulnerable compared to men because 

of socially constructed roles and prerequisites; biological or biophysical factors; and existing 

patterns of discrimination, such as domestic violence against women, that become more 

evident in extreme situations, thus emphasizing the importance of this topic. This subsequently 

also affects children and elderlies who are frequently dependent on women, especially in 

developing countries where women often are responsible for the subsistence of the family 

members. 

Experiences from large disasters have shown that the role of women is significantly different 

from men in the phases of prevention and disaster response. In the case of the 2004 tsunami, 

it was discovered that more women and children than men died in the worst affected areas 

(OXFAM 2005). Reasons are, for example, that in the disaster-affected areas men were more 

capable of swimming or climbing trees than women (OXFAM 2005). Furthermore, a study 

1 Austrian Federal Ministry for Agriculture, Forestry, Environment and Water Management, Vienna, AUSTRIA, 

catrin.promper@bmlfuw.gv.at 

2 Food and Agricultural Organization of the United Nations, Rome, ITALY 



carried out in Serbia following the flood of 2014 (Baćanović 2015) indicated that elderly 

women and women living alone were more inclined to have suffered damages from the 

calamity. 

To further enhance resilience it is important to proactively include women in socio-political 

decision processes and to ensure the inclusion of women- and youth-specific topics in disaster 

risk management plans. Therefore, the empowerment of women at local levels is crucial for 

influencing local protection and response. However, it is important to highlight not only the 

need to strengthen the training on gender issues for political representatives and institutions 

that deal with disaster risk reduction and reconstruction, but also the need for increased 

representation of women in decision-making processes in these institutions (Damyanovic et 

al. 2014, UNISDR 2010). Based on this, the platform on Natural Hazards of the Alpine 

Convention – PLANALP – included this topic in the current mandate 2015-2016. The focus 

of the mandate is to enhance the visibility and foster the engagement of women in decisionmaking 

processes and to raise awareness on gender issues in the field of disaster risk 

reduction. 

DIMENSIONS OF DISASTER PERCEPTION 

One of the guiding principles of the recently adopted Sendai Framework for Disaster Risk 

Reduction 2015-2030 (UN 2015) is an all-of-society engagement and partnership which 

includes a gender perspective in all policies and practices. Social interactions between men 

and women result in socially constructed roles, responsibilities and identities that are 

reflected in a combination of physical and behavioural characteristics – gendered identities 

(Madhavi 2010). These gendered identities comprise: perception, attitudes, and status 

separating boys from girls and men from women. Perceptions are the subsequent views 

resulting from these gender identities, and the attitudes differ as these are guided by their 

respective perception (Madhavi 2010). Another reflection on social interactions of men and 

women is the resulting positions in family, community, or society (Madhavi 2010). These 

relations come into effect in all spheres of life including disasters, which is highlighted in 

various studies conducted in different regions of the world. To detail with these differences, 

the approach of Neumayer and Plümper (2007), distinguishing three causes for gender 

differences in mortality resulting from natural disasters, is applied in the following paragraphs. 

The first dimension can be summarized as a physiological or biological dimension (Neumayer 

& Plümper 2007) that incorporates limitations, such as the ability to run away from a wave 

or withstand a storm. The 2004 tsunami that affected Southeast Asia, South Asia, and East 

Africa is a good example. In this event more women and children died than men. One reason 

for that was missing the physical strength necessary to hold on or float, which subsequently 

also applies to men that were not strong enough. 


In terms of recovery, including rebuilding infrastructure and health risks, there is also 

evidence of physiological differences between men and women. A big issue in the recovery 

phase is that women often do not have the strength to help clear debris, but engage in 

cleaning after the heavy-duty jobs were done (Baćanović 2015). Therefore, they are sometimes 

not as proactive as men when it comes to returning to their homes. In addition, single 

women or families without a male family member are not able to carry out post-disaster 

recovery work (Baćanović 2015). Subsequently, these women and families have to pay 

somebody to assist them with their return and rebuilding of their homes. Another issue 

where differences between men and women is evident: men and women have different 

propensities to die from various diseases. This in combination with discriminated access to 

resources after disasters, like food and hygiene products, can have tremendous impact on 

women. For example, women in less-developed countries are often more likely to suffer from 

malnutrition (IUCN 2009), which constitutes to their vulnerability to diseases. However, 

the implications for gender-specific disaster mortality are ambiguous (Neumayer & Plümper 

2007). 

The second aspect comprises behaviours and social prerequisites, meaning that the role one 

has in society can have an impact on vulnerability (OXFAM 2005, Neumayer & Plümper 

2007). These preconditions range from clothing that might hinder protecting oneself or from 

escaping, to the obligation or duty of providing protection or care for children or the elderly. 

An example is given in the case of the 2004 tsunami where women stayed behind with 

children or elderly, and consequently were not able to save themselves (OXFAM 2005). 

This hints at another important aspect with respect to social constraints: time and location of 

a disastrous event. Again, examples are provided by the 2004 tsunami where women in India 

were waiting at the sea shore for the fish to be processed when the tsunami struck, whereas 

in Sri Lanka the tsunami hit in the hour the local women usually take their bath (OXFAM 

2005). Additionally, in Indonesia the men tended to work away from their homes in bigger 

cities resulting in proportionally more deaths among women (OXFAM 2005). 

Another aspect that is related to social structures are long term impacts on the economic 

situation. A study that was carried out after the Great Hanshin Earthquake in 1995 details the 

impacts that disasters can have on women. It was shown that companies let go of part time 

workers first, most of whom were women (Masai et al. 2010), who often take care of the 

children and, therefore, were not able to work full time. Another long term effect of disasters 

can be a changed situation after the disaster, when men have to, for example, fend for 

themselves or take over unfamiliar tasks as they have to take care of their remaining family 

(OXFAM 2005). 

The third dimension focusses on the recovery phase where Neumayer and Plümper (2007) 

underline that existing gender patterns, depending on the respective societal context, become 


exacerbated through increased competition of individuals resulting from the breakdown of 

social order and limited access to resources. 

A transfer of existing gender patterns was also highlighted by Baćanović (2015), indicating 

that in the displacement centres after the flood in Serbia, men took over leadership and 

organizational roles and stepped out into the public sphere, whereas women were more 

isolated and passive, often looked after the children and did not always get the psychological 

support needed. Another example from St. Lorenzen in Austria, in 2012 indicated a continuation 

of traditional gender roles after a debris flow event. Women took over the supply and 

provision roles such as cooking, whereas men were more active in reconstruction works 

(Damyanovic et al. 2014). Nevertheless, in this case the degree to which individuals were 

affected was not based solely on their gender, but more on the individual situation and the 

subsequent social network within the municipality (Damyanovic et al. 2014). 

Differences between men and women are also evident in the context of early warning and 

evacuation, depending on their specific social context. A study from the Save flood (Baćanović 

2015) showed a difference in response and actions after warnings. In this case, women 

preferred official warnings, whereas men were prepared to obtain informal information. 

Another example, from Japan, showed that women often were talking to and then evacuating 

with people around them, whereas men more often evacuated alone (Gender Equality 

Bureau 2014). The example of Hurricane Katrina showed that women were often the 

decision-makers when it came to preparedness and evacuation: deciding if, when, where, 

and how long a family would evacuate (Peek L. & Fothergill 2010). The reason was that 

approximately 40 percent of the households before the storm were headed by females, due to 

the absence of husbands, and therefore they were also responsible for evacuation (Willinger 

2008, Barber & Deitz 2015). However, the fact that a large proportion of these female-headed 

households lived in poverty exacerbated this situation because of difficult preconditions like 

limited access to transportation and other necessities for evacuation, or insecure employment 

(IWPR 2006, Barber & Deitz 2015). 

Another aspect is the increased health and security risk for women after an event. Violence 

against women is commonly observed. This is especially referred to after big events, e.g. after 

the tsunami in Sri Lanka where threats of rape and sexual assault increased (OXFAM 2005). 

However, domestic violence can also increase due to such extreme situations. For example, 

in New Zealand following the Whakatane Flood 2004, a doubling to tripling in domestic 

violence was observed by several agencies (Houghton 2010). Another example is the Great 

East Japan earthquake of 2011, where the evacuation operation of the centres was mainly 

carried out by men without considering the women's point of view, thus causing women 

inconveniences and discomforts (Gender Equality Bureau 2014). Examples would be special 

health needs or security issues due to poorly light areas. 


GENDER MAINSTREAMING IN DISASTER RISK REDUCTION 

The above examples clearly indicate several gender differences in the context of natural 

disasters and this is especially evident in the response phase. The different roles and perception 

of men and women in situations such as preparation, early warnings, or decisions on 

evacuation, underline the importance of meeting the different needs in all phases of a 

disaster. Gender mainstreaming involves ensuring that gender perspectives and attention to 

the goal of gender equality are central to all activities – policy development, research, 

advocacy/dialogue, legislation, resource allocation, planning, implementation and monitoring 

of programs and projects (UN Women). The focus therefore is on the differences in perception 

of men and women in all phases of the disaster risk cycle and how to address these differences. 

The all-of-society engagement should also account for age, disability, and cultural perspective 

in all policies and practices (UN 2015) related to disaster risk reduction at all levels. 

This further increases the preparedness and coping capacity among all groups of society. 

In order to increase the resilience of women, it is important to ensure that women are 

specifically addressed in disaster risk management plans and that they are also included in 

socio-political decision-making processes. The United Nations Development Programme 

(UNDP) also aims to ensure women’s participation in all dialogues and in solution-generation 

for disaster risk reduction (UNISDR & GDN 2009). This subsequently requires empowerment 

of women, which also means ensuring opportunities for women in science and technology; 

building capacity in women’s groups and community based organizations; and including 

gender mainstreaming in communication, training, and education, etc. (UNISDR & GDN 

2009). This is also underlined by Damyanovic et al. (2014) and UNISDR (2010) who 

emphasize that in addition to strengthening the training on gender issues for political 

representatives and institutions in disaster risk management, it is also key to set up trainings 

for representatives from institutions dealing with disaster relief or reconstruction. 

The examples in the previous sections show that in the context of a disaster the differences 

between men and women are highly evident in developing countries, especially when 

referring to impact, death rates, etc., but are also evident in developed countries. Gender 

mainstreaming therefore might enhance success rates of disaster risk reduction and, considering 

knowledge transfer via best practice examples, influence community resilience in 

developing countries. Especially in disaster risk reduction, among other issues, alpine 

countries have strong expertise and, therefore, have a significant influence on methods and 

policies adopted in other mountainous areas of the world. The examples above show that 

vulnerability and resilience are highly dependent on the social structure, however, it also 

underlines the need for taking responsibility and acting as a role model to enhance gender 

mainstreaming in the disaster risk context. 

Raising the awareness of experts and stakeholders, acting as multipliers in society, and 

tailoring information concepts for women, youth and men are examples towards main- 


streaming gender in disaster risk reduction. The inclusion of women in different levels of the 

disaster risk reduction processes – from expert to stakeholder levels in participatory processes 

– increases perspectives, and subsequently accounts for different groups of society. In the long 

term this aims at an overall increase of resilience for all levels of society. Women have a 

different perspective on disaster risk reduction and therefore it is important to add this to the 

discourse, both at a decision-making and societal level. As a transnational body, PLANALP 

takes on this responsibility and has started by raising awareness for gender mainstreaming by 

featuring female experts in the field of disaster risk reduction. 

INCREASING THE VISIBILITY OF WOMEN IN DISASTER RISK REDUCTION 

PLANALP integrated the aspect of gender in the context of natural hazard and disaster risk 

management in its mandate for 2015-2016 with the item: “The role of women in natural 

hazard management focusing on the Alpine region”. The objective is to raise awareness and 

increase the visibility of women in the field of disaster risk reduction and to show their 

vocational diversity. To implement this, PLANALP took the opportunity to contribute to an 

exhibition in the frauenmuseum Hittisau, Austria: “Ich, am Gipfel. Eine Frauenalpingeschichte”. 

This exhibition on women and alpinism was therefore complemented with 

“talking-heads”, showcasing the potential careers for young women in the field, thereby 

fostering mixed teams in disaster risk reduction in the long run. 

To highlight the different options and career paths, as well as challenges and opportunities 

of female experts in disaster risk reduction, structured interviews were conducted and filmed. 

These interviews were displayed as “talking-heads” – showing the head and upper body of 

the person interviewed – on a screen in the exhibition. The questions in these structured 

interviews ranged from personal motivation for the job, daily business to observable 

differences between men and women in disaster situations. The women interviewed had a 

broad range of obligations in the field of natural hazard management: cross-cutting functions 

such as spatial planning, or the management of prevention and protection of hydro-meteorological 

hazards like floods or avalanches. A geologist and engineer in the field of torrent and 

avalanche control represented some of the women who contributed to this “talking-head” 

project. To integrate this element in an exhibition on women and alpinism was beneficial for 

two reasons: 1) the exhibition raised awareness for gender in similar contexts and 2) it 

reached out to different levels of society including experts, but also the general public and 

youth. 

CONCLUSION 

In conclusion, gender mainstreaming is an important part of disaster risk reduction for 

increased resilience at all levels of society to natural hazards. This entails higher awareness, 

better understanding and further implementation and replication of tailored disaster risk 

reduction concepts that are gender-inclusive. The exhibition of PLANALP in the frauenmuseum 

was one step towards higher awareness among the general public but also among experts 


which act as multipliers within society. Increased awareness among experts further contributes 

to tailored concepts for natural hazard risk management mainstreaming the gender 

element. To design these concepts, as indicated above, open questions like in-depth knowledge 

on the interrelationship of gender and social factors and related consequences in disaster 

situations have to be investigated. Further, regional differences have to be explored as shown 

by the examples of Serbia and Japan, where the reaction to formal versus informal warning 

was very different. Overall, evaluation of existing disaster risk reduction plans, documenting 

best practices and further research on the questions posed above are concrete steps towards 

a successful accomplishment of gender mainstreaming in all phases of the disaster risk cycle. 

ACKNOWLEDGEMENTS 

The authors like to thank PLANALP and the frauenmuseum Hittisau for the good cooperation 

and the female experts from Austria, Germany, Italy and Switzerland for their contribution 

as “talking-heads”. Further the authors thank the reviewers for their valuable and extensive 

comments on the first version of this paper. The exhibition “Ich, am Gipfel. Eine Frauenalpingeschichte” 

is running from 14 June 2015 to 26 October 2016 in the frauenmuseum 

Hittisau, Austria. 

REFERENCES: 

- Baćanović V. (2015). Gender Analysis of the Impact of the 2014 Floods in Serbia. Organization 

for Security and Co-operation in Europe. 

- Barber K. and Deitz S. (2015). Missing in the Storm: The Gender Gap in Hurricane Katrina 

Reserach and Disaster Management Efforts. In: Haubert J. (edt.), 2015, Rethinking Disaster 

Recovery A Hurricane Katrina Perspective. Lexington Books. 

- Damyanovic, D., Fuchs B., Reinwald F., Pircher E., Allex, B., Eisl J. Brandenburg, C., Hübl, 

C. (2014): GIAKlim - Gender Impact Assessment im Kontext der Klimawandelanpassung und 

Naturgefahren. Endbericht von StartClim2013.F in StartClim2013: Anpassung an den 

Klimawandel in Österreich - Themenfeld Wasser, Auftraggeber: BMLFUW, BMWF, ÖBF, Land 

Oberösterreich. 

- Gender Equality Bureau (2014). Natural Disasters and Gender Statistics: Lessons from the 

Great East Japan Earthquake and Tsunami, From the “White Paper on Gender Equality 

2012”. Cabinet Office, Government of Japan. 

- Houghton R. (2010). Sex, Gender and Gender Relations in Disasters. In: Enarson E. and 

Dhar Chakrabarti P G, 2015 Women, Gender and Disaster Global Issues and Initiatives. 

Sage Publications. 

- IUCN (2009). Disaster and gender statistics. 

- Madhavi Malalgoda Ariyabandu (2010). "Everything Became a Struggle, Absolute Struggle": 

Post-flood Increases in Domestic Violence in New Zealand. In: Enarson E. and Dhar 

Chakrabarti P G, 2010 Women, Gender and Disaster Global Issues and Initiatives. Sage 

Publications. 


- Masai R., Kuzunishi L. and Kondo T. (2010). Women in the Great Hanshin Earthquake. 

In: Enarson E. and Dhar Chakrabarti P G, 2010 Women, Gender and Disaster Global Issues 

and Initiatives. Sage Publications. 

- Neumayer E. & Plümper T. (2007). The Gendered Nature of Natural Disasters: The Impact 

of Catastrophic Events on the Gender Gap in Life Expectancy, 1981–2002. Annals of the 

Association of American Geographers, 97:3, 551-566. 

- OXFAM International (2005). The tsunami's impact on women. Ofxam Briefing Note. 

- Peek L. and Fothergill A. (2010). Parenting in the Wake of Disaster: Mothers and Fathers 

Respond to Hurricane. In: Enarson E. and Dhar Chakrabarti P G, 2010, Women, Gender and 

Disaster Global Issues and Initiatives. Sage Publications. 

- UN (2015). Sendai Framework for Disaster Risk Reduction 2015-2030. 

- UNISDR (2010). Guidance note on recovery: gender 

- UNISDR and Gender and Disasters Network (2009). The Disaster Risk Reduction Process: 

A Gender Perspective A Contribution to the 2009 ISDR Global Assessment Report on Disaster 

Risk Reduction. 

- UN Women, www.unwomen.org. accessed 25.03.2015. 

- Willinger B. (edt.) (2008). Katrina and the Women of New Orleans – Executive report and 

summary of findings. Tulane Univeristy. 



The central challenge of climate change adaptation 

for Alpine natural hazard management: Incorporation 

of future change of the damage potential 

Die zentrale Herausforderung der Klimawandelanpassung 

für das Naturgefahrenmanagement im 

Alpenraum: Berücksichtigung des zukünftigen 

Wandels des Schadenpotentials 

Klaus Pukall, Dr.¹; Sylvia Kruse, Dr.² 

ABSTRACT 

Climate change adaptation within the field of alpine hazards management is not sufficiently 

taking future changes of the damage potential into account. Climatic and societal changes are 

expected to influence hazard exposure and social vulnerability considerably. Feedback mechanism 

between hazard management, settlement development and climate change adaptation 

need to be incorporated. The Safe-Development-Paradox proofs also to apply for the 

Alpine Space, i.e. that protection measures often lead to building activities and thus increases 

societal and financial values in areas potentially at risk. Because hazard zones can grow due 

to climatic changes, these interconnections between realized and planned protection and 

climate adaptation measures as well as future societal changes should be considered in hazard 

management and planning already today. Starting points for alpine hazard management are 

therefore first, to link climate and societal scenarios on local and supraregional scale; second, 

to include stakeholders from natural hazard management when negotiating future settlement 

development and spatial uses; and third, to account for uncertainty of future development 

by implementing flexible and adaptive measures. 


Die Klimaanpassungsstrategien für den Bereich des Naturgefahrenmanagement im Alpenraum 

berücksichtigen bisher unzureichend die zukünftigen Veränderungen des Schadenspotentials. 

Es ist jedoch davon auszugehen, dass sowohl der Klimawandel die naturwissenschaftlich 

beschreibbaren Gefahrenprozesse als auch gesellschaftliche Veränderungen die 

Verwundbarkeit der Bevölkerung beeinflussen werden. Rückkopplungsmechanismen 

zwischen Naturgefahrenmanagement, Siedlungsentwicklung und Klimaanpassung sind hier 

zu berücksichtigen. Das Safe-Development-Paradox zeigt auch im Alpenraum, dass gerade 

technische Schutzbauwerke zu einer stärkeren Siedlungsentwicklung in potenziellen 

1 Technische Universität München, Freising, GERMANY, klaus.pukall@tum.de 

2 Chair of Forest and Environmental Policy, University of Freiburg, GERMANY 



Gefahrenbereichen führen und damit das Schadenspotential zunimmt. Da sich diese Gefahrenbereiche 

durch den Klimawandel noch ausweiten können, sollten diese Zusammenhänge 

zwischen bereits umgesetzten und geplanten Schutz- und Klimaanpassungsmaßnahmen 

sowie der zukünftigen gesellschaftlichen Entwicklung bereits heute bei der Planung berücksichtig 

werden. Ansatzpunkte für das alpine Naturgefahrenmanagement bilden daher erstens 

die Verknüpfung von Klima- und Gesellschaftsszenarien auf lokaler und überregionaler 

Ebene, zweitens der Einbezug von Akteuren des Naturgefahrenmanagements in die Aushandlung 

der Siedlungsentwicklung und der Raumnutzungen, und drittens die systematische 

Berücksichtigung von Unsicherheit in Form von flexiblen und damit anpassungsfähigen 

Maßnahmen. 

KEYWORDS 

Climate change adaptation; spatial planning; safe development paradox; uncertainty 

EINLEITUNG 

Anpassung an den Klimawandel ist ein wichtiges Thema für das Naturgefahrenmanagement 

(NGM) im Alpenraum. Entsprechende Konzepte wurden ab 2000 in Klimaanpassungsstrategien 

(KAS) in Bayern (deutscher Anteil am Alpenraum), Österreich und der Schweiz 

erarbeitet. In diesem Beitrag vertreten wir die These, dass die zentrale Herausforderung der 

Klimaanpassung (KA) für das NGM die Berücksichtigung des zukünftigen Wandels nicht nur 

der Klimaparameter sondern vor allem des gesellschaftlichen Systems darstellt. Genauso 

wie bei den Klimaprojektionen des IPCC müssen soziale Zukünfte berücksichtigt und in das 

heutige Management von alpinen Naturgefahren integriert werden. 

Ziel dieses Artikels ist es die Notwendigkeit für ein zukunftsgerichtetes NGM herauszuarbeiten 

und mögliche Ansatzpunkte für die systematische Berücksichtigung des gesellschaftlichen 

und klimatischen Wandels im NGM zu benennen. 

Basierend auf einer Analyse der bestehenden KAS diskutieren wir, inwieweit zukünftige 

Veränderungen in natürlichen und sozialen Systemen berücksichtigt werden und identifizieren 

„blinde Flecken“, die bisher keine Berücksichtigung in der KA im NGM finden. Darauf 

aufbauend entwickeln wir Vorschläge für ein zukunftsgerichtetes NGM, das besonders die 

Steigerung des Schadenspotentials berücksichtigt. 

METHODIK 

Innerhalb des BMBF-Projekts „Alpine Naturgefahren im Klimawandel“ (Vertrag 01UV1004B, 

Deutsches Bundesministerium für Bildung und Forschung) wurden auf nationaler und 

sub-nationaler Ebene (Bayern, Graubünden, Tirol) die bis Ende 2013 erarbeiteten KAS sowie 

vorbereitende Dokumente (Gesetzestexte, in den Ländern aufgelegte Forschungsprogramme 

zur Klimafolgenabschätzung, sektorale Programme des NGM) mit Methoden der qualitativen 

Inhaltsanalyse untersucht. Außerdem wurden Leitfadeninterviews mit 41 Experten des NGM 


aus Österreich, Deutschland und der Schweiz über die KA im NGM geführt und ein internationaler 

Experten-Workshops durchgeführt, im dem Lösungsansätze für die Herausforderungen 

der KA erarbeitet wurden. 

STAND DER KLIMAANPASSUNG IN DEN UNTERSUCHUNGSLÄNDERN 

Die KAS, die in den untersuchten Ländern in den Jahren 2000-2014 erstmals erstellt wurden, 

basieren überwiegend auf länderspezifischen Forschungsprojekten, die sowohl historische 

klimabezogene Datenreihen analysierten als auch die globalen Klimamodelle durch unterschiedliche 

Downscaling-Mechanismen auf regionaler Ebene auswerteten. Die Ergebnisse 

dieser Forschungsprojekte flossen als Informationsgrundlage in die nationalen und regionalen 

Anpassungsstrategien ein. 

Eine vergleichende Analyse der KAS zeigt, dass KA in den Strategien der untersuchten 

Länder primär als eine Anpassung an ein verändertes regionales Klima verstanden wird. 

Eine Anpassung an sich verändernde gesellschaftliche Gegebenheiten (z. B. Siedlungs- und 

Wirtschaftsentwicklung, demographische Struktur etc.) findet bisher nicht statt. Anpassung 

wird zudem als routinierte Veränderung der etablierten Handlungsstrategien bzw. Maßnahmen 

empfunden, wie stellvertretend dieses Zitat aus der Schweiz zeigt: „Die zusätzlichen 

Herausforderungen infolge des Klimawandels können deshalb durch eine konsequente 

Umsetzung der PLANAT-Strategie und des integralen Risikomanagements gemeistert werden“ 

(Schweiz. Bundesrat 2012: 3808) 

BLINDE FLECKEN DER KA: FEHLENDE ZUKUNFTSORIENTIERUNG BEZÜGLICH SOZIALER PROZESSE 

Das derzeitige NGM ist vergangenheits- bzw. gegenwartsorientiert. Das für die meisten Maßnahmen 

des NGM zentrale Konzept der Jährlichkeit ist auf eine Betrachtung möglichst langer 

historischer Zeitreihen ausgelegt. Je länger die Beobachtung zurückreicht, desto besser kann 

das Magnitude-Frequenz-Verhältnis der betrachteten Gefahren bestimmt werden. Maßnahmenplanungen, 

die das Schadenspotential berücksichtigen, gehen vom aktuellen Zustand aus 

und vergleichen somit z. B. durch Kosten-Nutzen-Abwägungen die Reduktion des aktuellen 

Schadenspotentials durch unterschiedliche Maßnahmen. 

Dieser Vergangenheitsbezug wird innerhalb der KA durch Szenarien über die zukünftigen 

Auswirkungen der Klimaveränderung auf die Gefahrensituationen ergänzt. Die Einführung 

eines einheitlichen Klimaänderungsfaktors von 15% (d.h. das aus historischen Daten 

errechnete Bemessungsereignis wird um den Faktor 1,15 erhöht) für Hochwasserschutzbauten 

in Bayern im Jahr 2004 oder der 2003 fertig gestellte Lawinen- und Murgangschutzdamm 

in Pontresina (Graubünden) sind ein Beispiel dafür, dass bereits vereinzelt heutige Maßnahmen 

zukünftige Veränderungen antizipieren. 

Wie bereits in der Einleitung ausgeführt, basiert die Arbeit der Klimawandelforschung auf der 

Verwendung von Emissionsszenarien, die Annahmen über gesellschaftliche Entwicklungen 


einhalten. Die KAS berücksichtigen diese gesellschaftlichen Szenarien aber häufig nicht, um 

Maßnahmen der KA zu formulieren. KA sollte daher nicht nur die Projektionen zur Klimaentwicklung 

berücksichtigen sondern besonders bei der Formulierung der KA-Maßnahmen 

auch systematisch die unterschiedlichen möglichen Gesellschaftszukünfte berücksichtigen. 

Für das NGM müssten explizit folgende Themen beachtet werden. 

Rückkopplungsmechanismen aufgrund durchgeführter Schutzmaßnahmen: 

Das Safe-Development-Paradox 

Burby (2006) erklärt einen Großteil der Schäden in New Orleans, die nach Hurrikan Katrina 

entstanden sind, auf Basis des von ihm so benannten „Safe Development Paradox“: Jede 

Schutzmaßnahme, die zu einem Gefühl 100%iger Sicherheit führt, verursacht langfristig ein 

erhöhtes Schadenspotential. Burby (2006) erweitert damit das Verständnis für den sog. 

„levee effect”, den zuerst Segoe (1937) beschrieb. Falls ein Deich gebaut wird, empfinden 

die Bewohner des durch den Deich geschützten Gebiets als auch die lokalen Behörden dieses 

Gebiet als „sicher“. Die Siedlungsentwicklung findet in dem durch den Deich geschützten 

Gebiet so statt, als wäre das Hochwasser vollkommen gebannt – der besonders in Bayern 

verwendete Begriff der Hochwasserfreilegung unterstützt z.B. diese Fehlwahrnehmung. 

Dieser Prozess führt zu einer deutlichen Steigerung des Schadenspotentials im Vergleich zu 

nicht geschützten Flächen. Burby (2006) betont, dass besonders solche Maßnahmen, die 

neue Entwicklungsmöglichkeiten eröffnen und nicht bestehende Bebauung schützen, zum 

„Safe Development Paradox“ beitragen. 

Auch die Gefahrenzonenplanung im Alpenraum kann zu einem falschen Gefühl der Sicherheit 

führen. Keiler (2004) und Fuchs et al. (2004) wiesen für die Gemeinden Galtür und 

Davos nach, dass die höchsten Steigerungsraten der Schadenspotentialentwicklung (besonders 

durch die Erweiterung bzw. den Ausbau bestehender Bebauung) an der Grenze der 

Bauverbotszone stattfanden, da ja der Plan die relative Sicherheit des Bauplatzes darstellte. 

Die Gefahrenzonenpläne können die graduelle Abnahme der Gefährdung nicht angemessen 

wiedergeben (siehe dazu auch Kap. 5.3). Falls es durch den Klimawandel nun zu einer 

Verschärfung der natürlichen Prozesse kommt, wäre das Schadenspotential durch die 

Siedlungsentwicklung in den vermeintlich sicheren Bereichen deutlich erhöht. 

Wie Merz et al. (2009) darstellen, ist die empirische Evidenz für die quantitativen Auswirkungen 

des Safe Development Paradox derzeit noch relativ gering. Sie selbst gehen auf Basis 

ihrer Literaturauswertung von 5-10% Erhöhung des Schadenspotentials bei Flusshochwasser 

aufgrund des Safe Development Paradoxes aus. Für die Niederlande kommen Jongmann et 

al. (2014) zu deutlich höheren Steigerungsraten. Eine Analyse von Fuchs et al. (2015), die 

die Entwicklung des Gebäudebestands in Österreich seit 1919 untersuchten, zeigt dagegen 

für Flusshochwasser, Wildbach- und Lawinengefahren keine klaren Hinweise für ein Safe 

Development Paradox. Dies liegt auch daran, dass die Studie nicht darauf ausgelegt war, das 

Safe Development Paradox zu beschreiben. 


Interaktionen zwischen Raumordnung, Klimaschutz- sowie Klimaanpassungsmaßnahmen 

und dem NGM 

Die Raumordnung ist bezüglich des NGM durch das Top-Down-Instrument der Gefahrenzonenplanung 

bzw. in Deutschland der festgesetzten Überschwemmungsgebiete geprägt. 

Auf der Basis einer nationalen Risikoabwägung (siehe Kap. 5.3) werden pauschale Baubeschränkungen 

ausgesprochen, von denen im Einzelfall nur marginal abgewichen werden 

kann (vgl. Höferl 2013, Pukall 2014). Der Raumplanung stehen (im Vergleich zur forstlichen 

und wasserwirtschaftlichen Fachplanung des NGM) wenig personelle und finanzielle 

Ressourcen zur Verfügung (Kruse und Pütz 2013). Eine aktive Auseinandersetzung über 

die Entwicklung des zukünftigen Schadenspotentials findet daher kaum statt. Gerade im 

inneralpinen Gebiet mit seinem beschränkten Siedlungsraum wären Planungen notwendig, 

wie der verbliebene Siedlungsbereich optimal genutzt wird, damit bei einer Berücksichtigung 

von Extremereignissen ein möglichst geringes Schadenspotential entsteht. Hierbei müssen 

neben der allgemeinen wirtschaftlichen und räumlichen Entwicklung auch die Interaktionen 

mit dem Klimaschutz und der KA berücksichtigt werden, durch deren Maßnahmen zusätzliches 

Schadenspotential entsteht (Pukall und Kruse 2014). Für das NGM wären daher vor 

allem unterschiedliche Szenarien für die Siedlungs- und Infrastrukturentwicklung notwendig. 

Diese Themen werden nicht ausreichend im derzeitigen NGM und den KAS behandelt. 

Eine gewisse Problemwahrnehmung ist in den KAS erkennbar: „Vermehrte raumwirksame 

Klimafolgenrisiken und gleichzeitig zunehmende Raumansprüche der Gesellschaft führen 

insbesondere in den alpin geprägten Teilräumen Österreichs mit naturbedingt knappem 

Dauersiedlungsraum zu zunehmender Flächenverknappung und damit zur Einengung 

zukünftiger wirtschaftlicher Entwicklungsmöglichkeiten“ (BMLFUW 2012, 302). 

SCHLUSSFOLGERUNGEN: ANKNÜPFUNGSPUNKTE FÜR EIN ZUKUNFTSORIENTIERTES NATUR- 

GEFAHREN MANAGEMENT 

Aus unserer Sicht gibt es drei zentrale Anknüpfungspunkte, um die Zukunftsorientierung 

des NGM und den Einbezug von sowohl klimatischen als auch gesellschaftlichen Zukünften 

zu unterstützen. 

Berücksichtigung der zukünftigen Klima- und Gesellschaftsentwicklung bei 

Maßnahmen des NGM 

Die Bewertung unterschiedlicher Vorsorgestrategien muss unter Berücksichtigung der zukünftigen 

Klima- und Gesellschaftsentwicklung erfolgen. Wie in der Klimaforschung sollten 

hier gesellschaftliche Szenarien erstellt und räumlich modelliert werden. Das österreichische 

Projekt RiskAdapt hat hierfür am Beispiel Hochwasser eine geeignete Methodik erarbeitet. 

Neben einer landesweiten Verknüpfung von Klima- und Bevölkerungsszenarien bedarf es 

insbesondere auf lokaler Ebene Aushandlungsprozesse, welche Entwicklungsszenarien in 

einer Gemeinde erwünscht sind und welche Schutzstrategien dafür die beste Lösung darstellen 

(RiskAdapt 2015). 


Kosten-Nutzen-Analysen von geplanten Schutzmaßnahmen, wie z.B. das in der Schweiz für 

alle Naturgefahren vorgeschrieben Werkzeug “EconoMe“ oder die in Bayern und Österreich 

bestehenden verwaltungsinternen Wirtschaftlichkeitsrichtlinien, sollten Annahmen über die 

erwartete Veränderungen von Siedlungsentwicklung und Klimafolgen beinhalten. Hierbei 

wäre eine enge Zusammenarbeit mit Raumordnungsakteuren notwendig. Entscheidend ist 

dabei auch, ob der Faktor der Risikoaversion, d.h. das gesellschaftliche Bedürfnis besonders 

seltene Großschadenereignissen zu vermeiden, berücksichtigt wird. Merz et al. (2009) 

konnten z.B. an drei Fallstudien in Deutschland zeigen, dass bei einer starken Ausprägung 

der Risikoaversion, Warnungskonzepte gegenüber technischen Schutzmaßnahmen mit 

Deichen das Gesamtrisiko am stärksten minimieren würden. Ohne den Faktor Risikoaversion 

wären trotz angenommener Effekte des Safe Development Paradoxes, die wie oben dargestellt 

zur Entstehung der seltenen Großschadensereignisse beitragen, die Deiche die beste 

Risikominimierungsmethode gewesen. 

Aktive Beeinflussung der Siedlungs- und Infrastrukturentwicklung durch Akteure 

des NGM 

Aushandlungsprozesse über die zukünftige lokale und regionale Raumnutzung, Tourismuskonzepte 

und auch die notwendigen Maßnahmen zur KA laufen häufig informal ab und 

binden staatliche Träger öffentlicher Belange, wie z. B. die mit dem NGM betrauten Forstund 

Wasserbauverwaltungen, nicht bereits in der Vorplanung, sondern erst später in den 

formalen Planungs- oder Genehmigungsprozessen ein. So können die staatlichen Träger 

öffentlicher Belange erst spät mit ihren Vorstellungen zum Schutz vor Naturgefahren die 

Prozesse der Raumentwicklung beeinflussen. Sie finden sich daher häufig in der Rolle des 

Verhinderers von Entwicklungsideen wieder oder es wird Druck ausgeübt, regionalwirtschaftliche 

Entwicklungsprojekte mit Hilfe von technischen Schutzmaßnahmen zu ermöglichen. 

Aus unserer Sicht sollten alle Akteure des NGM daran mitwirken, sich aktiv in lokale 

Raumentwicklungsprozesse einzubringen. Für ein zukunftsgerichtetes NGM ist es nicht nur 

wichtig, die Siedlungs- und Infrastrukturentwicklung mit Hilfe der Gefahrenzonenplanung 

in den hoch gefährdeten Gebieten zu verhindern, sondern die Raumentwicklung so zu 

beeinflussen, dass hoch vulnerable Infrastrukturen (z.B. Tourismuseinrichtungen, die viele 

Personen anziehen) in besonders sichere Gebiete gelenkt werden. 

Entscheidend hierbei ist, dass die Akteure gut in lokale Netzwerke eingebunden sind und 

somit frühzeitig mit neuen Entwicklungsideen in Kontakt kommen. Sie könnten auf der 

einen Seite versuchen, dem Naturgefahrenthema eine gewichtige Stimme in den lokalen 

Aushandlungen zu geben. Auf der anderen Seite könnten sie eine Mittlerposition zu den 

staatlichen Behörden einnehmen, so dass sich im Bedarfsfall die Behörden frühzeitig mit 

ihrer fachlichen Kompetenz in informelle Planungen einbringen können. 

Hierbei kann zwischen zwei Personengruppen differenziert werden: 

– Personen, die keinen unmittelbaren Bezug zu staatlichen Verwaltungen haben, wie 

z.B. Mitglieder der Lawinenkommissionen, der Feuerwehren sowie die in der Schweiz 


geschaffenen Naturgefahrenberater. Alle diese Personen sollten in ihrer Aus- und Fortbildung 

für diese, eigentlich nicht zu ihrem Aufgabengebiet gehörende Rolle sensibilisiert 

werden. 

– Personen, die Teil der staatlichen Verwaltungen sind. Hierbei dürfen nicht nur für die 

Raumplanung formal zuständigen Mitarbeiter betrachtet werden, sondern insbesondere 

Mitarbeiter wie z.B. die Arbeiter der Flussmeisterstellen der bayerischen Wasserwirtschaftsverwaltung 

oder die Waldaufseher in Tirol. Diese sind bei den Gemeinden angestellt, 

fachlich werden sie aber von der Forstverwaltung betreut. Die oben erwähnte Mittlerposition 

würde hier durch eine gute behördeninterne Kommunikation ausgefüllt. 

Dieser Vorschlag hat natürlich auch seine Beschränkungen, da die betrachteten Personengruppen 

Eigeninteressen haben können, die dazu führen, dass sie die ihnen zugedachte Rolle 

nicht ausfüllen. Z.B. kann ein Mitglied der Lawinenkommission gleichzeitig Hotelbesitzer 

sein, der somit eine geplante Erweiterung des Skigebiets eher fördern als verhindern möchte. 

Flexibler Umgang mit Unsicherheit 

Einigkeit herrscht in der KA-Literatur darüber, dass flexibel und adaptiv auf den Klimawandel 

und die damit einhergehende Unsicherheit reagiert werden sollte (z.B. Etkin et al. 2012). 

Aufgrund der zunehmenden Unsicherheit, wenn sowohl die Folgen des Klimawandels als 

auch Annahmen über gesellschaftliche Zukünfte in die Planungen des NGM einfließen sollen, 

kommt somit ein ausschließlich auf errechenbaren Wahrscheinlichkeiten basierendes 

Risikomanagement an seine Grenzen. Etkin et al. (2012) argumentieren, dass der Umgang 

mit Unsicherheit im Zuge von klimatischen und gesellschaftlichen Veränderungen eine große 

Bedeutung im Risikomanagement erhalten müsste. Dies bedeutet, dass die etablierten 

Routinen des Risikomanagements, die von wohldefinierten und wenig veränderlichen 

Systemen ausgehen, mit dem Vorsorgeprinzip, das besonders bei Unsicherheit angewendet 

wird, sinnvoll verknüpft werden müssen (Etkin et al. 2012: 588; vgl. auch Kuhlicke und 

Kruse 2009). Beispielsweise wird aufgrund des Vorsorgeprinzips in der Gefahrenzonenplanung 

in Hochrisikogebieten, den roten Gefahrenzonen, jegliche Bebauung untersagt. 

Im Sinne des Vorsorgeprinzips sollte die Bauentwicklung nicht nur in Gebieten hoher 

Gefährdung beschränkt werden, deren Grenzen aufgrund des Klimawandels nur mit größerer 

Unsicherheit bestimmt werden können, sondern in besonders sichere Gebiete gelenkt 

werden. Dies gilt insbesondere für teure Infrastrukturprojekte und Gefahren, die Menschenleben 

bedrohen (z.B. Steinschlag, Lawinen). 

Bereits im bestehenden NGM besteht für die Gefahrenzonenplanung die Problematik, dass 

der fließende Übergang von hoher zur niedriger Gefährdung in den rechtlich normierten 

Gefahrenzonenplänen nicht abgebildet werden kann, da Grenzen zwischen unterschiedlichen 

Gefährdungsstufen nicht graduell dargestellt werden. Während in der Schweiz und in Österreich 

zumindest 2-3 Zonen mit höherer Gefährdung ausgeschieden werden, ist die diskrete 

Logik bei den festgesetzten Überschwemmungsgebieten in Deutschland besonders stark 


ausgeprägt. Es gibt nur Gebiete innerhalb der Überschwemmungsgebiete, in denen ein 

Bauverbot gilt und Gebiete außerhalb, in denen keinerlei Baubeschränkungen bestehen. 

Im Zuge eines zukunftsgerichteten NGM sollten die staatlichen Mindeststandards, die auf 

Basis des Vorsorgeprinzips gerechtfertigt sind, um flexible Anpassungsmöglichkeiten der 

Gemeinden ergänzt werden. So könnte sich selbst eine Bebauung in einer roten Lawinenzone 

als sinnvoll herausstellen, wenn ausschließlich eine Nutzung in den Sommermonaten 

stattfindet und der Investor z.B. aufgrund einer kurzen Amortisationzeit seiner Investition 

bereit ist, das Risiko zu tragen. Es böte sich zudem der Einsatz finanzieller Instrumente 

(e.g. Versicherungspflicht, risiko-abhängige Grundsteuer usw.) an, die wie die Gefährdungslage 

graduell steigen und flexibel an die bestehende bzw. sich verändernde Gefährdungslage 

angepasst werden könnten (Filatova 2014). Die Berücksichtigung aktueller und zukünftiger 

Gefährdungslagen könnten so in Markt- und Bodenpreise integriert werden und damit die 

Schadenspotentialentwicklung steuern. 

LITERATUR 

- BMLFUW (Bundesministerium für für Land- und Forstwirtschaft, Umwelt und Wasserwirtschaft, 

2012): Die österreichische Strategie zur Anpassung an den Klimawandel: 

Teil 2 – Aktionsplan Handlungsempfehlungen für die Umsetzung, Wien. 

- Burby R.J. (2006): Hurricane Katrina and the Paradoxes of Government Disaster Policy: 

Bringing about Wise Governmental Decisions for Hazardous Areas, in: Ann Am Acad 

Political Social Sci 604, p. 171-191. 

- Etkin D., Medalye J., Higuchi, K. (2012): Climate warming and natural disaster management: 

An exploration of the issues, in: Climatic Change 112, p. 585-599. 

- Filatova T. (2014): Market-based instruments for flood risk management: A review of 

theory, practice and perspectives for climate adaptation policy, in: Environ Sci & Policy 37, 

p. 227-242. 

- Fuchs S., Bründl M., Stötter J. (2004): Development of Avalanche Risk between 1950 and 

2000 in the Municipality of Davos, Switzerland, in: Nat Hazards Earth Syst Sci 4, p. 263-275. 

- Fuchs S., Keiler M., Zischg A. (2015): A spatiotemporal multi-hazard exposure assessment 

based on property data, in: Nat Hazards Earth Syst Sci 15, p. 2127-2142. 

- Höferl K. (2010): ‚Von der Gefahrenabwehr zur Risikokultur’ - Diskurse zum raumplanerischen 

Umgang mit Hochwasser in (Nieder-)Österreich, Dissertation an der Universität 

für Bodenkultur, Wien. 

- Jongman B., Koks E.E., Husby T.G., Ward P.J. (2014): Increasing flood exposure in the 

Netherlands: implications for risk financing, in: Nat Hazards Earth Syst Sci 14, p. 1245-1255. 

- Keiler M. (2004): Development of the Damage Potential resulting from Avalanche Risk in 

the period 1950–2000, Case Study Galtür, in: Nat Hazards Earth Syst Sci 4, p. 249-256. 

- Kuhlicke C., Kruse S. (2009): Nichtwissen und Resilienz in der lokalen Klimaanpassung- 

Widersprüche zwischen theoriegeleiteten Handlungsempfehlungen und empirischen 

Befunden am Beispiel des Sommerhochwassers 2002, in: Gaia 18, p. 247-254. 


- Merz B., Elmer F., Thieken A.H. (2009): Significance of “high probability/low damage” 

versus “low probability/high damage” flood events, in: Nat Hazards Earth Syst Sci 9, p. 

1033–1046. 

- Pukall K. (2014): Von Top-Down zu Bottom-Up: Berücksichtigung von regionalen 

Entwicklungsprozessen im staatlichen Naturgefahrenmanagement, in: Böschen S., Gill B., 

Kropp C., Vogel K. (Hrsg.): Klima von unten – Regionale Governance und gesellschaftlicher 

Wandel, Frankfurt / New York, p. 287-308. 

- Pukall K., Kruse S. (2014): Entwicklungslinien für das Management alpiner Naturgefahren 

im Klimawandel (Essay), in: Schweiz Z Forstwes 165, p. 37-42. 

- RiskAdapt (2015): https://riskadapt.boku.ac.at/home.htm (überprüft 3.7.2015) 

- Schweizerischer Bundesrat (2012): Anpassung an den Klimawandel in der Schweiz – Ziele, 

Herausforderungen und Handlungsfelder: Erster Teil der Strategie des Bundesrates vom 

2. März 2012, in: BBl 2012: p. 3777- 3858. 

- Segoe L. (1937): Flood control and the cities, in: American City 52, p. 55-56. 



Local Conditions and the Quality of Expert Networks: 

A Case Study of Avalanche Risk Prevention Practices 

Renate Renner, Ph.D¹; Gerhard Lieb¹ 

ABSTRACT 

As natural hazards and risk always represent an interaction of natural and social systems, 

avalanche risk prevention is especially important in heavily-touristed alpine regions. This 

research compares the local risk prevention practices of Nordkette, Tirol and Planneralm, 

Styria by considering the influence of local conditions and the quality of expert networks. 

This in-depth analysis demonstrates that local conditions influence the intensity and 

frequency of necessary prevention measurements and the level of pressure on the decision 

makers. Local conditions influence the level of professionalization, the incorporation of 

systematic data analysis and personnel competence. This study aligns with general findings 

about the role of social capital, emphasizing that trust within the avalanche commission team 

and between the avalanche commission and external experts increases the quality of local 

risk prevention. Although the study provides important insight into the scope of risk 

prevention practices, further research is necessary to understand how coping capacity could 

be improved through optimizing the use of social networks. 

KEYWORDS 

Local conditions; quality of networks; avalanche risk prevention; risk communication 

INTRODUCTION 

It is commonly assumed that due to climate change, the number of extreme weather events 

resulting in gravitational processes is increasing (e.g. IPCC, 2014), although some critics 

emphasise the missing observational records for that assumption (Stoffel and Huggel, 2012). 

What is certain, however, is that the Austrian economy relies heavily on tourism and is 

vulnerable to climate change, especially in alpine regions (e.g. OECD, 2007). Nowadays 

winter sport regions are facing the challenge of ensuring the safety of local and visiting 

populations, safeguarding infrastructure, and at the same time, gaining economic profit. 

Therefore, efficient and professional avalanche risk prevention in Austria is a topic of great 

importance and leads us to focus on internal risk communication (Renn, 2008) and risk 

prevention practices in this paper. Höppner et al. (2012) assume that through risk communication, 

the social capacity –competence to cope with hazard events - can be increased at 

individual, communal and organisational (the risk managing) level. Social networks are 

considered to be a key social capacity because of their role in transmitting other capacity 

types like motivation, knowledge or financial resources (Kuhlicke and Steinführer, 2010). 

1 Institute of Geography and Regional Science, University of Graz, AUSTRIA, renate.renner@uni-graz.at 



Based on Höppner’s insights into social capacity, we first seek to shed light on the local expert 

network in order to consider its role on the network’s quality. This is important because 

permanent protection measures cannot replace human services and temporary protection 

measures. Therefore, local avalanche risk prevention significantly depends on the engagement 

and know-how of local avalanche commissions. Second, we assume an interaction 

between local conditions and the way risk prevention is practiced. 

In brief summary, the case study aims to deepen the limited understanding of the nature of 

local avalanche risk prevention practices. Through explorative case studies in Tirol and Styria, 

we consider local and network internal conditions and aim to address a set of interrelated 

questions: 1.) How do communication and decision processes between the study sites differ? 

2.) How are the practices of avalanche risk prevention influenced by local conditions? 3.) 

How do local conditions and the quality of networks support or hinder local risk prevention? 

RESEARCH DESIGN 

The theory of structuration (Giddens, 1995) serves as a meta-theory to consider the context 

dependence of social action, as local risk prevention can never be understood isolated from 

where it takes place. Giddens explains the relation between “action” and “structure” as 

interdependent. “Structure” includes rules and resources and is a spatiotemporal phenomenon. 

While in our research “action” is represented by communication and decision processes 

within the experts network, “structure” includes local conditions. The latter refers to, for 

instance, legal requirements for avalanche commissions, personnel and financial resources 

and the economic relevance of risk management in the respective areas. 

In order to consider the interaction between local conditions and the way risk prevention is 

practiced, we applied maximum variation sampling (Quinn Patton, 2002). The two selected 

study sites in Austria differ in terms of geographical characteristics (see Table 1) such as snow 

climate, population and economic activity and therefore offer different local conditions for 

avalanche risk prevention. 

Study sites 

The case of the Nordkette in Tirol represents a densely populated central region that is 

threatened by avalanches. The mountain range in question is the main ski-area for the 

approximately 120,000 inhabitants of the city of Innsbruck, located at the foot of the 

Nordkette. Economically, the region is strongly diversified and characterised by agriculture 

and moderate tourism. In contrast, the second study site, Planneralm, is a sparsely populated 

peripheral area; permanent settlements are not at risk of avalanches. Nevertheless, the only 

access road to the Planneralm is jeopardised by 12 avalanche paths, which significantly 

influences the competitiveness of the Planneralm as a tourist destination. The mountain 

pasture forms part of the municipality of Irdning Donnersbachtal with a population of about 

4000. 


Table 1: Comparison of selected geographical characteristics of our study sites. 

Data material 

In accordance with the interpretative paradigm (Wilson, 1973) we developed problem-centered 

interviews (Witzel, 1982) to identify the core communication and decision-making 

process, and to better understand supporting and hindering factors of risk prevention. 

As a conceptional framework we applied the qualitative social network approach that allows 

understanding both who is connected to whom, and what quality of the relation exists 

(Hollstein, 2006). We focus on the intensity of risk communication (Renn, 2006), also asking 

interviewees from whom they use documents, whose information is of importance and who 

is involved in the discussion and decision-making process. 

As we are considering the local level, our analysis is derived from the viewpoint of local 

avalanche commissions. Besides considering their reflections on their own work we also use 

those sequences from interviews with other avalanche experts that consider their quality of 

relations to the local avalanche service. We interviewed members of the avalanche warning 

service, the avalanche commission, external consultants and members of the Wildbach- und 

Lawinenverbauung (WLV) which is the service for torrent and avalanche control in Austria. 

All together, 12 face-to-face interviews ranging from 50 to 90 minutes were conducted and 

analysed applying Mayrings (2010) content analysis. We inductively analysed indications of 

local conditions that influence avalanche risk prevention practice. In addition, the data are 

analysed deductively with the above mentioned focus on networks. 


RESULTS 

While in Tirol, rights and obligations of local avalanche commissions are regulated by law 

(LGBl. Nr. 104/1991), there are only official recommendations in Styria. Despite a different 

regulatory intensity, the composition and appointment of the members, the areas of responsibility 

and the avalanche commissions’ duties are to a great extend identical in content. The 

mayor of a region exposed to avalanches is primarily responsible for founding an avalanche 

commission in his/her municipality. Commission members need to have professional 

experience and must be available on-site during the winter season. The area of responsibility 

is the organized ski area (cross country skiing trails, ski slopes), traffic routes and the 

settlement area of the respective municipality. Local avalanche commissions exercise an 

advisory role; hence they are responsible for continuous evaluation of avalanche risk. 

Communication and decision processes of local avalanche commissions in the 

two study sites 

The assessment process at Nordkette in Tirol takes place daily in the winter season in the form 

of a meeting between at least 3 commission members. On ordinary days, this team is composed 

of 3 rotating employees from the lift operating company of the Nordkette. 

In dangerous situations, more members of the avalanche commission are involved, primarily 

forestry employees and members of the city government. Members of the avalanche 

commission exchange their private phone number so that everyone is reachable day and 

night, even on days off. Every commission member (altogether about 9) has e-mail access 

and receives the avalanche report daily. Those who are at work are in radio contact for a 

rapid assessment: “within minutes, we review the situation” (avalanche commission 

member). Through a logging software (LWD-KIP), not only the sequence of data research is 

documented, but also assessments, decisions and recommended measurements are documented 

digitally. Through LWD-KIP, avalanche commissions have access to all relevant snow 

and weather data of the respective region. Decision-making is based on a variety of different 

data (see figure 1). The data sources include: a meteorological station located at the summit 

lift station of the Nordkette, the analysis of snow and weather conditions through LWD-KIP, 

personal observations made by on-site assessment, and snow-profiles in early winter before 

many sections receive human traffic. Existing data is analysed daily, based on both practical 

knowledge and a discursive process of avalanche danger. 

In general, an unanimous decision is preferred regarding how to act in response to avalanche 

risk. The dominant attitude favors closing the skiing area as opposed to disagreement within 

the commission and an awkward feeling between commission members: “we think if one of 

us names a good argument to not open the ski-slope than it makes sense to keep it closed” 

(avalanche commission member). Numerous statements of the commission members refer to 

professional competence and a huge pool of experience (e.g. the ability to assess similar slope 

exposures and to transfer this knowledge to the area of responsibility, or to know where 

snow transporting can be expected if south or north wind picks up). The know-how of other 

informants (avalanche warning service etc.) is considered even though the commissions own 


Figure 1: Communication and decision-making process of the local avalanche commission Nordkette, Tirol. Networks used and content 

of relation from the avalanche commissions` perspective. 

daily and seasonal observations receive the greatest attention. Despite levels of professionalism 

and objective data, the commission members emphasise that the final decision remains 

a question of “gut feeling.” Members of WLV and the avalanche warning service in Tirol 

perceive the avalanche commission of the Nordkette to be very professional and independently 

working. The commission itself explains their relationship to the avalanche 

warning service to be good. They meet regularly and their relationship is open to constructive 

criticism. Once a year, all members of the commission and the local authority meet and 

discuss optimisation possibilities. 

At the Planneralm in Styria, the assessment process takes place on demand: “if it becomes 

dangerous, we meet each other” (chairman and his deputy). Consultations are rare and 

happen only between the chairman and his deputy, though occasionally third person is 

involved in the discussion. “We have a small frame that allows us to work effectively. It is 

never good if there are too many” (chairman). Primarily the avalanche danger is assessed by 

the chairman: “I say that’s how it looks like. Or my deputy informs me about his observations 

and asks me how I assess the situation and I say how it is.“ (chairman) The decision is 

made through one authoritative voice, and often without a discussion process between 

members. The fundamental attitude prioritizes rapid decision-making, whereas reflection and 

critical discussion are subordinated. Different statements from the chairman and his deputy 


efer to their professional experience and local knowledge (long term observations over many 

seasons, responsible for ski slopes preparation in the past, etc.). However, subjective perceptions 

are rarely verified or supplemented by standardized weather and snow data analysis. 

The latter are partly considered, but not systematically. Since the sports hostel of the University 

of Graz has been closed at the Planneralm, the avalanche commission has to proceed 

without observations and snow profiles from mountain guides who visited the alpine terrain 

daily. Currently, the chairman works instead with imagined snow profiles (“and I make a 

picture in my head about the existing snow profile, I am able to do that, I feel that”) and 

observes mountain slopes with field glasses only. Above all, subjective perception, personal 

experience and instinct determine risk assessment at the Planneralm. 

Figure 2: Communication and decision-making process of the local avalanche commission Planneralm, Styria. Networks used and 

content of relation from the avalanche commissions` perspective. 

A closer look at the quality of relations within the avalanche commission shows a variety of 

contradictory statements and disputes between the chairman and other official members of 

the commission. On the one hand, the chairman and his deputy emphasise having a good 

working relationship within the avalanche commission; on the other hand, conflicts within 

the team are mentioned in interviews from different sites. As a consequence, the organized 

ski area at the Planneralm is more or less excluded from the evaluation of the avalanche commissions 

chairman and his deputy. They see the manager of the cable cars as soley responsible 

for evaluating avalanche danger. Although the manager is an official member of the 


avalanche commission, he is not involved in regularly consultations between the chairman 

and his deputy. Statements from the chairman as well as from externals refer to the strained 

relationship between the chairman and the manager of the cable cars. Besides, the official 

recording clerk of the commission even refused the interview and mentioned that he was not 

actively involved in the avalanche commissions’ work. All together, of the 8 official avalanche 

commission members, only two regularly meet each other and represent the active avalanche 

commission of the Planneralm. Occasionally, one additional person is asked to assist in the 

case of avalanche dispersion. The others are not actively involved, except through their 

participation in a general meeting with the mayor once a year. External people critically stress 

that trust is lacking between the chairman and nearly all entrepreneurs at the Planneralm. 

While the chairman sees his relationship to the avalanche warning service as quite good, the 

contrary is true if considering the statements of the experts concerned. 

Influence of local conditions on avalanche risk prevention practices 

Table 2: Summary of inductively analysed indications of local conditions that influence risk prevention practice. 


Supporting and hindering local risk prevention 

The comparison of the study sites in this research has shown that well-functioning and 

trusting internal and external relationships improve both data quality and quantity of data 

(see figure 1 and 2) used for decision-making. Constructive team work allows critical 

reflection of personal opinions and perceptions, thus improving both final decisions and the 

quality of risk prevention. This aspect aligns with previous work on social capital in which it 

is understood to be embedded in social networks (Lin, 2001) and increases access to social 

support and information (Buckland and Rahman, 1999). It also corresponds to the so called 

“social and organisational capacities” (Höppner et al. 2012: 1757) or “network capacities” 

(Kuhlicke et al. 2011: 806) which emphasizes the importance of skills for communication, 

cooperation and building up trustful relationships. A transparent decision-making process can 

protect the commission in the case of misjudging avalanche risk, which in the past, has led to 

impeachment charges against members. Furthermore, exogenous factors such as a lack of 

financial resources and competitive neighboring regions, can significantly increase pressure 

on the decision-makers who often have a dual role as risk manager and owner of a tourist 

enterprise in the respective area. In avalanche commissions with full-time employees, 

personnel recruitment can be selective and attention can be given to professional knowledge 

and social competence. To the contrary, however, volunteer associations often have general 

difficulties in finding members. The case of the Planneralm has shown that this could 

exacerbate risk assessment measures in a context where distrust and conflicting interests 

among the avalanche commission team are already present. As a consequence, existing 

talents within team members are not tapped as resources, succession planning goes unnoticed, 

and risk prevention proceeds in far from optimal conditions. The results of this study 


ing us back to the coping capacity discussion (Kuhlicke et al., 2012) and other research in 

the context of climate change in which it was shown that local resilience can be fostered by 

strong social networks (Ford et al. 2006). 

This finding proves even more interesting when we discovered that social competence and 

the role of risk communication is neither part of the desired profile of commission members, 

nor it is considered in any of the training courses. Furthermore, this research has shown that 

professional working practices are not only a question of personnel resources but also of the 

culture of innovation. The existence and implementation of LWD-KIP allows systematic data 

analysis, logging and rapid risk communication. While at Planneralm not even information 

transfer between commission members is guaranteed, the digital logging software automatically 

informs all relevant persons, thus reducing human errors. 

CONCLUSION 

The importance of professional risk prevention in alpine regions highlights the need for more 

qualitative research assessing different practices and their influence factors. The main finding 

of the case studies in Tirol and Styria is that both local conditions and the quality of social 

relations within the expert team influence risk prevention. Although tasks and regulations of 

the commission teams are similar, the case studies show a considerable range of local risk 

prevention practices and the gap between an officially-presented picture and its practical 

reality. This finding points out the need to consider how to assure quality in the future. 

Closely linked with the quality assurance is the importance of communication skills and trusting 

relationships, which have been proven to be significant but understudied components in 

risk assessment. As this study presented only two contrasting examples, a larger study could 

provide evidence as to the type and quality of social networks through which reliable and 

optimized risk assessment decision-making occurs. 

LITERATURE 

- Buckland J., Rahman M. (1999): Community-based disaster management during 

- the 1997 Red River flood in Canada, in: Disasters 23, 3, 174-191. 

- Ford J., Smit B., Wandel J. (2006): Vulnerability to climate change in the Arctic: 

A case study from Arctic Bay, Canada, in: Global Environmental Change 16, 2, 145-160. 

- Giddens A. (1995): Die Konstitution der Gesellschaft, Frankfurt/ Main [u.a.], Campus. 

- Hollstein B. (2006): Qualitative Methoden und Netzwerkanalyse – ein Widerspruch?, in: 

B. Hollstein und F. Straus (Hrsg.), Qualitative Netzwerkanalyse. Konzepte, Methoden, 

Anwendungen, Wiesbaden, VS Verlag für Sozialwissenschaften, 11-36. 

- Höppner C., Whittle R., Bründl M., Buchecker M. (2012) Linking social capacities and risk 

communication in Europe: a gap between theory and practice? Nat. Hazards 64, 1753-1778. 

- IPCC (2014): Climate Change Synthesis Report, Summary for Policymakers, 

URL http://www.ipcc.ch/pdf/assessment-report/ar5/syr/AR5_SYR_FINAL_SPM.pdf. 

14.09.2015. 


- Kuhlicke C., Steinführer A. (2015): Preface: Building social capacities for natural hazards: 

an emerging field for research and practice in Europe. Nat. Hazards Earth Syst. Sci., 15, 

2359-2367. 

- Kuhlicke C., Steinführer A., Begg C., Bianchizza C., Bründl M., Buchhecker M., De Marchi 

D., Di Masso Tarditti M., Höppner C., Komac B., Lemkow L., Luther J, McCarthy S., Pellizzoni 

L., Renn O., Scolobig A., Supramaniam M., Tapsell S., Wachinger G., Walker G., Whittle R., 

Zorn M., Faulkner H. (2011): Perspectives on social capacity building for natural hazards: 

outlining an emerging field of research and practice in Europe. Environmental Science & 

Policy 14, 804-814. 

- Lin N. (2001): Building a network theory of social capital, in: Lin N., K.S. Cook, R.S. Burt 

(Hrsg.) Social capital: Theory and Research, New Brunswick, NJ, Transaction Press, 3-30. 

- Mayring P. (2010): Qualitative Inhaltsanalyse, Grundlagen und Techniken. 11. aktualisierte 

und überarbeitete Auflage. Weinheim und Basel, Beltz. 

- OECD (2007): Klimawandel in den Alpen. Anpassung des Wintertourismus und des 

Naturgefahrenmanagements, Agrawalla S. (Hrsg.), URL http://www.oecd.org/env/ 

cc/38002265.pdf. 14.09.2015. 

- Quinn Patton M. (2002): Qualitative research and evaluation methods, Thousand Oaks, 

California, SAGE. 

- Renn O. (2006): Risk communication: Consumers between information and irritation, in: 

Journal of Risk Research, 9, 8, 833-850. 

- Renn O. (2008): Risk Governance. Coping with Uncertainty in a Complex World. London, 

New York. Earthscan. 

- Stoffel M., Huggel C. (2012): Effects of climate change on mass movements in mountain 

environments, Prog Phys Geogr 36, 421-39. 

- Wilson T. P. (1973): Theorien der Interaktion und Modelle soziologischer Erklärung, in: 

Arbeitsgruppe Bielefelder Soziologen (Hrsg.), Alltagswissen, Interaktion und gesellschaftliche 

Wirklichkeit, Bd. 1 Symbolischer Interaktionismus und Ethnomethodologie, Reinbeck bei 

Hamburg, Rowohlt, 54-79. 

- Witzel A. (1982): Verfahren der qualitativen Sozialforschung. Überblick und Alternativen, 

Frankfurt/Main, New York, Campus. 



Dealing with gravitational natural hazards: challenges 

for risk management and development planning 

Umgang mit gravitativen Naturgefahren: 

Herausforderungen für das Risikomanagement und 

die Raumplanung 

Florian Rudolf-Miklau, Priv. Doz. Dr.¹; Kanonier Arthur, Univ.-Prof. Dr.²; Thomas Glade, Univ.-Prof. Dr.³; Stix Elisabeth, Mag. 4 

ABSTRACT 

In contrast to the flood and avalanche hazards, in Austria there is no comprehensive legal 

system of risk management for gravitational natural hazards (rock fall, landslide) existent so 

far. Furthermore there are no general standards available for the assessment and mapping of 

theses hazards as a basis for decision-making in areal development. However, significant 

changes in the Alpine land use have significantly increased the vulnerability of settlements 

and traffic routes in endangered areas. The article deals with the backgrounds and the results 

of a partnership established in the framework of the Austrian Spatial Development Strategy 

2011 (OeREK) which was dedicated to define the principles of hazard mapping and pave the 

path for a closer alignment of effective decisions with spatial relevance to the results of hazard 

assessment and risk-based protection goals. The OeREK-partnership represents a successful 

governance process in order to bridge the deficits of this legally fragmented policy field. 


Im Gegensatz zu den Hochwasser- und Lawinengefahren existiert in Österreich bisher kein 

geschlossenes Rechtssystem des Risikomanagements für gravitative Naturgefahren (Steinschlag, 

Rutschungen) und bestehen auch keine Standards für eine differenzierte Gefahrenanalyse 

und -darstellung als Grundlage raumwirksamer Entscheidungen. Bedeutende Veränderungen 

in der alpinen Raumnutzung haben jedoch die Verletzlichkeit von Siedlungsraum 

und Verkehrswegen in Gefahrenzonen deutlich erhöht. Der Beitrag behandelt die 

Hintergründe und Ergebnisse der Partnerschaft im Rahmen des Österreichischen Raumentwicklungskonzepts 

(ÖREK) 2011, welche die Grundlagen für Ausarbeitung von Gefahrenhinweiskarten 

und Gefahrenzonenplänen definiert hat und so eine stärker Ausrichtung 

raumwirksamer Entscheidungen an den Ergebnissen der Gefahrenanalyse und den risikoorientierten 

Schutzzielen ermöglicht. Die ÖREK-Partnerschaft stellt einen erfolgreichen 

Governance-Prozess zur Überbrückung der Defizite dieses kompetenzrechtlich zersplitterten 

und nur fragmentarisch normierten Politikfeldes dar. 

1 BMLFUW, Vienna, AUSTRIA, florian.rudolf-miklau@die-wildbach.at 

2 Technical University of Vienna, Department for Land Policy and Management, AUSTRIA 

3 University of Vienna, Department of Geography and Regional Research, AUSTRIA 

4 Austrian Conference on Spatial Planning, AUSTRIA 



KEYWORDS 

Gravitational natural hazards; hazard mapping; risk management;development planning; 

legal system 

EINLEITUNG UND HINTERGRUND 

Gravitative Naturgefahren (insb. Steinschlag, Rutschungen, Hangmuren) haben einen maßgeblichen 

Einfluss auf die Raumentwicklung in Österreich. Aufgrund der überwiegend gebirgigen 

Topografie des Landes sind nur rund 37 Prozent (Tirol: 12 %) des Staatsgebietes als 

Dauersiedlungsraum geeignet (BMLFUW, 2011). Zusätzlich entstehen weitere Einschränkungen 

durch Naturgefahren über entsprechende Gefahrenzonen; so waren Muren, Felsstürze 

und Lawinen viele Jahrhunderte sprichwörtlich die eigentlichen „Raumplaner“ alpiner 

Gemeinden (Mattle, 2015) und prägen noch heute die alpine Raumentwicklung. Prominente 

Beispiele der letzten Jahre sind die Rutschungen in Gasen und Haslau (Steiermark), die 

Hangbewegung in Doren (Vorarlberg) (Abbildung 1) oder der Felssturz Eiblschrofen (Tirol). 

Gravitative Gefahrenprozesse sind häufig mit intensiven und sich verändernden Landnutzungen 

gekoppelt und lösen gesellschaftliche Anpassungsprozesse aus (Papathoma-Köhle & 

Glade 2013; Promper & Rudolf-Miklau, 2015). 

Die Gefahrenprozesse können als Einzelereignisse auftreten, können aber auch ausgedehnte 

Regionen mit mehreren Hundert, manchmal sogar vielen Tausend Einzelereignisse umfassen. 

Im Wirkungsbereich entfalten sie eine meist räumlich begrenzte, jedoch mit hohen Intensitä- 

Abbildung 1 : Rutschung Doren (Vorarlberg) 


ten verbundene Schadenwirkung für Personen, Baubestand und Verkehrsinfrastruktur und 

stehen daher mit den maßgeblichen regionalen Entwicklungstrends (Baumann et al., 2000; 

Bätzig, 2003; Psenner, 2006; Rudolf- Miklau et al., 2014; ÖROK-Atlas, 2014) in vielfältiger 

Interaktion. Einige der folgenden Wechselwirkungen können bezogen auf gravitative 

Naturgefahren stark risikoerhöhend wirken: 

Der im Alpenraum ablaufende demografische Wandel weist in Richtung der Konzentration 

der Siedlungs- und Wirtschaftsentwicklung in den zentralen Kernstädten und führt zu einer 

Suburbanisierung der Talräume (u.a. Inntal, Rheintal). Im Gegensatz dazu weisen strukturell 

benachteiligte, ländliche Gebiete negative Entwicklungen (z.B. Landflucht, Entvölkerung 

entlegener Bergtäler) auf. Die zunehmende Entkoppelung von Wohn- und Wirtschaftsräumen 

erfordert eine steigende Mobilität und verstärkt die Abhängigkeit der Bevölkerung von 

der Nutzbarkeit der Verkehrswege und der Funktionsfähigkeit der Versorgungslinien. 

Besonders kritisch sind die signifikante Zunahme von Flächennutzung in Hanglagen (Gunstlagen, 

Baulandverknappung), die exponentielle Wertzunahme des Gebäudebestandes, 

der Flächenverbrauch und die „Versiegelung“ der Landschaft (Hartflächen, touristische 

Erschließung in den Alpentälern: Schipisten, Golfplätze), der Wegebau in Hanglagen sowie 

der Verlust traditioneller land- und forstwirtschaftlicher Nutzungsformen (De-Agrarisierung) 

zu bewerten. Dadurch kommt es zu risikoerhöhende Veränderungen der Geländetopographie 

mit Auswirkungen auf das Hangwasserregime bzw. im Falle von Bauland auf die Veränderung 

des Baugrundrisikos. Ein für die Raumplanung zentraler Effekt der Veränderungen ist 

ein zunehmendes strukturelles Ungleichgewicht zwischen Gebieten mit ausreichend sicherer 

Baulandreserve und jenen mit akutem Baulandmangel außerhalb von Gefahrenzonen. Diese 

Disparitäten verzerren auch die Verteilung des regionalen Schutzbedarfs und erhöht unter 

anderem auch den Siedlungsdruck auf Hanglagen, wo vielfach noch keine Gefahren(hinweis) 

karten, Gefahrenzonenpläne oder Risikokarten verfügbar sind (Glade et al., 2013). Ebenso 

gravierend ist die saisonale Verlagerung des Personenrisikos von den Wohn- und Arbeitsstätten 

hin zu den Freizeit- und Urlaubsgebieten oder auf die Verkehrs- und Transitsachsen 

(Promper & Rudolf-Miklau, 2015). 

Dem komplexen Zusammenhang zwischen Raumnutzung und den Risiken durch gravitative 

Naturgefahren kann am besten mit einem gesamtheitlichen Risikomanagement entsprochen 

werden. (PLANAT, 2009; Loat, 2015) Aus semantischer Sicht ist dazu anzumerken, dass der 

„Risikobegriff“ in der österreichischen Raumplanung bisher weder in den Rechtsgrundsätzen 

noch im Vollzug etabliert ist und daher gegenüber dem eindeutig definierten Gefahrenbegriff 

unscharf oder nur abstrakt verwendet wird. Das grundlegende Schutzziel (siehe Begriffsdefinition 

in: Promper et al., 2015; abweichend von Schweizer Definition nach Camenzind und 

Loat, 2014) des Risikomanagements ist die Reduktion der Verletzlichkeit des Lebensraums für 

Naturkatastrophen und die Erhöhung der Resilienz der Gesellschaft (Ammann, 2006; Rudolf- 

Miklau, 2009). Die Raumplanung nimmt die Berücksichtigung von Naturgefahren und 

Risiken eine zentrale Funktion ein. Dabei geht es nicht nur um die kartografische Darstellung 


der Flächenwirkung von Gefahrenprozessen (Gefahrenplanung) und damit verbundenen 

Risiken (Risikoplanung), sondern auch um die Möglichkeit, die resultierenden Risiken durch 

planerische Maßnahmen zu verringern (präventive Raumplanung) oder drohenden Schäden 

vorzubeugen (Rudolf-Miklau, 2009). Die Raumplanung hat im Zusammenhang mit Naturgefahren 

zwei grundlegende Anforderungen zu erfüllen (Kanonier, 2012; 2015): 

– Anpassung der Raumnutzung an die Gefahren einschließlich der Beschränkung der 

Nutzung in gefährdeten Gebieten; 

– Anpassung der Raumnutzung an die Erfordernisse der Gefahrenprävention, z. B. durch 

Freihaltung von Ablagerungsräumen oder gezielte Flächenbewirtschaftung. 

Grundsätzlich steht im Rahmen einer risikoangepassten Regionalentwicklung eine Palette 

von wirkungsvollen Instrumenten zur Verfügung. Dazu zählen die Festlegung von Entwicklungszielen 

in Regionalplänen und Raumentwicklungsprogrammen, die Reservierung von 

Freiräumen (schutzwirksame Vorbehaltsflächen, Retentionsflächen), die Steuerung der 

Siedlungsentwicklung in der örtlichen Raumordnung (Flächenwidmungsplan), rechtsverbindliche 

Festlegungen der Planungsbehörde (Widmungsbeschränkungen, Widmungsverbote) 

sowie die Risikokommunikation mit dem Ziel der Information und Bewusstseinsbildung 

(Rudolf-Miklau, 2009; Kanonier, 2012). Die Anwendung dieser Instrumente ist im Risikomanagement 

nur in enger Abstimmung und Kombination mit den Planungen und Maßnahmen 

der übrigen Sektoren (Geologie, Wildbach- und Lawinenverbauung, Verkehrswesen, 

Katastrophenschutz, Land- und Forstwirtschaft) effektiv möglich. 

RECHTSPOLITISCHE RAHMENBEDINGUNGEN UND DEFIZITE 

Im Bereich des Schutzes vor gravitativen Naturgefahren besteht in Österreich eine starke 

Kompetenzsplitterung, die auf das Fehlen eine konkreten Kompetenztatbestandes in der 

Bundesverfassung sowie eines geschlossenen Rechtssystem für das Risikomanagement 

zurückzuführen ist (Wagner & Jandl, 2013). Die Kompetenzzersplitterung spiegelt sich auch 

auf organisationsrechtlicher Ebene wieder, da zahlreiche Institutionen der Gebietskörperschaften 

sowie privater Institutionen Präventions- und Sicherheitsaufgaben erfüllen 

(Rudolf-Miklau, 2009). Im Gegensatz zu den Hochwasser- und Lawinengefahren sind 

Rechtsgrundlagen und Kompetenzverteilung für das Risikomanagements gravitativer 

Naturgefahren wesentlich lückenhafter und diffuser ausgeprägt (Kanonier, 2015). Aus der 

Erkenntnis des Europäischen Gerichtshofs für Menschenrechte (EGMR) „Budayeva gegen 

Russland“ ist nicht generell zu entnehmen, dass die Staaten konkret zur Setzung präventiver 

Schutzmaßnahmen gegen gravitative Naturgefahren verpflichtet wären. Es liegt vielmehr in 

deren politischen Ermessen der Staaten, in welchem Umfang sie Präventionsmaßnahmen 

durchführen bzw. dafür die entsprechenden rechtlichen und organisatorischen Rahmenbedingungen 

schaffen (Wagner, 2008). Mit Ausnahme des Bodenschutzprotokolls der Alpenkonvention 

fehlen spezifische rechtliche Bestimmungen für gravitative Naturgefahren auch 

im internationalen und europäischen Recht, die staatliche Handlungspflichten begründen 

würden (Kanonier, 2015). 


Eine allgemein gültige Rechtsnorm für die Prävention gravitativer Naturgefahren (insb. zum 

Schutz des Siedlungsraums) besteht somit im nationalen österreichischen Recht nicht. 

Gesetzliche Verpflichtungen zur Vornahme von Schutzvorkehrungen ergeben sich lediglich 

aus dem Straßenrecht, dem Eisenbahnrecht und dem Seilbahnrecht (Hattenberger, 2004). 

Sonstige einschlägige Rechtsnormen finden sich auch in anderen Materiengesetzen, beispielsweise 

im Wasserrecht, im Forstrecht, im Raumordnungs- und Baurecht sowie im Katastrophenschutzrecht 

(Rudolf-Miklau, 2009). Im Zusammenhang mit der Darstellung und 

Bewertung von Naturgefahren sind vor allem die bundesrechtlichen Vorschriften im § 11 

ForstG und § 42a WRG anzuwenden. Als wesentliche Grundlagen für kommunale Planungsund 

Bauentscheidungen sind Gefahren(hinweis)karten bzw. Gefahrenzonenpläne (GZP) 

bedeutend. Von einigen Ausnahmen abgesehen (u.a. für die Gefahrenhinweiskarten für 

Niederösterreich: Petschko et al. 2013) erfolgt für den Großteil der gravitativen Gefahren 

(Steinschlag, Rutschung) – im Unterschied etwa zur Schweiz (Loat, 2015) und anders als für 

Hochwasser- und Lawinengefahren – keine Abstufung nach Gefährdungsgraden (rote/ 

rot-gelbe und gelbe Gefahrenzonen). Eine abgestufte Berücksichtigung der Gefahren in der 

Raumplanung – etwa für bebaubare, bedingt bebaubare und nicht bebaubare Flächen – ist 

allein durch die kommunale Flächenwidmungsplanung kaum möglich, zumal „bedingte“ 

Baulandwidmungen selten sind (Ausnahmen stellen in einzelnen Bundesländern Aufschließungszonen 

dar). (Kanonier, 2015) Die Gefahrenzonenpläne der Wildbach- und 

Lawinenverbauung (WLV) enthalten allerdings die Darstellungskategorie „Brauner Hinweisbereich“ 

für Gebiete, die von Steinschlag- und Rutschungsgefahren potenziell betroffen sind, 

ohne eine konkrete Aussage über Ausdehnung, Häufigkeit oder Intensität zu geben (Rudolf-Miklau, 

2009). Davon abweichend werden im Bundesland Vorarlberg im WLV-GZP 

abgestufte Gefahrenbereiche („Braun“, „Braun intensiv“) für Steinschlag und Rutschungen 

ausgewiesen. Die Bundesländer Niederösterreich (Abbildung 2) und Oberösterreich haben 

jeweils Gefahrenhinweiskarten ausgearbeitet, welche als Grundlage für die einzelfallbezogene 

Gefahrenbewertung und die Konkretisierung des erforderlichen geologischen Untersuchungsaufwandes 

herangezogen werden (Glade et al. 2013; Bell et al. 2013). Die Gefahrenhinweiskarten 

enthalten jedoch keine genauen Angaben über Frequenz oder Magnitude (Prozessintensität) 

der gravitativen Massenbewegungen und können daher nicht für die Detailbegutachtung 

von Widmungs- und Bauvorhaben herangezogen werden (Glade, 2015). 

Kenntnisse über räumliche Abgrenzungen, Häufigkeit und Intensität von Naturgefahren von 

gravitativen Naturgefahren stellen zentrale Kriterien für behördliche Widmungs- und Bauentscheidungen 

dar. Die meisten Raumordnungsgesetze der Länder enthalten deshalb allgemeine 

Verpflichtungen zur Kenntlichmachung von Gefahren in den Flächenwidmungsplänen. 

Aufgrund des weitgehenden Fehlens einheitlicher Rechtsnormen über die Art der 

Analysemethode und die kartographische Gefahrendarstellung der Analyseergebnisse liegt es 

meist im Ermessen der Behörden, welche Informationen im Zusammenhang mit gravitativen 

Massenbewegungen bei konkreten Planungs- und Baumaßnahmen verwendet werden. Soweit 

nicht vereinzelt detaillierte Gefahrenkarten oder Gefahrenzonenpläne verfügbar sind, 

bleibt es also den Gemeinden überlassen, im Rahmen der allgemeinen Verpflichtung zur 


Abbildung 2: Beispiel eine Gefahrenhinweiskarte für Steinschlag der Gemeinde Dürnstein (Niederösterreich) 

vollständigen Erhebung der räumlichen Gegebenheiten auch die von Steinschlag und 

Rutschungen verursachten Gefahrenbereiche soweit wie möglich zu erfassen (Kanonier, 

2012; 2015). Allgemein ist iSd meisten Raumordnungsgesetze eine Baulandwidmung 

unzulässig, wenn sich die betreffende Fläche wegen gravitativen Naturgefahren für eine 

zweckmäßige Bebauung nicht eignet. Für den Fall, dass gravitative Naturgefahren nicht 

räumlich ausgeschlossen werden können, erfordern Widmungs- und Baubewilligungsentscheidungen 

daher spezifische Untersuchungen und Erhebungen. Generelle Regelungen 

gefahrenangepasster Nutzungen auf der Ebene der Raumplanung erschiene zielführen, 

widerspricht allerdings einerseits dem Grundgedanken des „Gefahren meiden“ und wäre 

andererseits wohl zu komplex, um auf kommunaler Ebene ohne umfangreich Ausnahmen 

und Auslegungsregeln umsetzbar zu sein. Daher gibt es in den einzelnen Raumordnungsund 

Baugesetzen dafür unterschiedliche (im Detail dazu: Kanonier, 2015), wobei sich diese 

Regelungen meist an den Hochwassergefahren orientieren. Eine Anpassung dieser Regelungen 

für die besonderen Eigenschaften der gravitativen Naturgefahren scheint zielführend. 

ÖREK-PARTNERSCHAFT „GRAVITATIVE NATURGEFAHREN“: RISIKO GOVERNANCE PROZESS UND 

POLITISCHE EMPFEHLUNGEN 

ÖREK-Partnerschaften stellen ein wesentliches Umsetzungsinstrument des Österreichischen 

Raumentwicklungskonzeptes (ÖREK, 2011) dar. In diesem Rahmen wurde die ÖREK-Partnerschaft 

„Risikomanagement für gravitative Naturgefahren in der Raumplanung“ eingerich- 


tet, um für diesen kompetenzrechtlich zersplitterten und fachlich segmentierten Bereich eine 

neue Kooperations- und Entwicklungsplattform zu schaffen, die der hohen naturwissenschaftlichen, 

technischen und rechtlichen Komplexität der Fragestellungen gerecht wurde 

und die Herausforderung einer fachübergreifenden Harmonisierung leisten konnte. Neben 

der Schaffung fundierter fachlicher Grundlagen für die Bereiche der geologischen Gefahrenanalyse 

und -darstellung, der fachplanerischen Umsetzung und der Berücksichtigung in der 

Raumplanung (ÖROK-Schriftenreihe 193) bestand das wichtigste Ziel der Partnerschaft in 

der Erstellung von politischen ÖROK-Empfehlungen (ÖROK, 2015). 

Als neuer Ansatz wird im Rahmen dieser ÖREK-Partnerschaft „Risiko“ in die Bewertung der 

Folgen von gravitativen Naturgefahren für die Raumplanung und andere Sektoren in 

Österreich eingeführt. Der Risikobegriff ist jedoch kein traditioneller Rechtsbegriff wie der 

Gefahrenbegriff (Kanonier, 2015). In der Logik der Risikoforschung ist ein Risiko raumplanungsrelevant, 

wenn mithilfe raumplanerischer Instrumente Eintrittswahrscheinlichkeit oder 

Konsequenz eines Ereignisses für bestimmte, hinlänglich sicher identifizierbare Entstehungsund/oder 

Gefährdungsräume beeinflussbar sind (Glade, 2015). Risikoabschätzungen sind 

jedoch längst nicht Standard in der Raumplanung, sondern werden nur in speziellen Fällen 

durchgeführt (Kanonier, 2015). Ziel der Partnerschaft war es daher, durch die Entwicklung 

von Grundlagen des Risikomanagements die Basis für einen risikoorientierten und differenzierten 

Umgang mit Steinschlag und Rutschungen in der Raumplanung zu legen. 

Eine weitere Herausforderung des Prozesses stellten die offensichtlichen Grenzen der 

bestehenden bundesstaatlichen Strukturen für eine Festlegung allgemeiner Schutzziele, die 

formale Abstimmung von Grundsätzen der kartographischen Gefahrendarstellung (im Detail 

siehe: Schwarz et al., 2014) sowie für abgestimmte Anpassungen des Raumordnungs- und 

Baurechts dar. Auch wenn in der Partnerschaft direkte Einflussnahme auf kompetenzrechtliche 

Tatbestände nicht intendiert war, ist es doch gelungen, durch einen Governance-Prozess 

alle relevanten Akteure in Österreich auf freiwilliger Basis einzubinden und nach gemeinsam 

festgelegten Regeln zu gemeinsam beschlossenen, fachlichen und politischen Empfehlungen 

anzuleiten. Mit Rücksichtnahme auf die föderalen Rahmenbedingungen wurden Empfehlungen 

geschaffen, die einerseits ein bundesweite Vergleichbarkeit von Planungen und Sicherheitsentscheidungen 

ermöglichen und andererseits auf regionale und bundesländerspezifische 

Besonderheiten Rücksicht nehmen. Auch konnten bereits bestehende Pilotprojekte und 

Planungen (Glade & Krause, 2015) erfolgreich in die Ergebnisse der ÖREK-Partnerschaft 

integriert werden. 

Aus den fachlichen Ergebnissen der Partnerschaft wurden von einer Redaktionsgruppe 

politische ÖROK-Empfehlungen entwickelt, die von den politischen Partnern der ÖROK 

genehmigt und im Februar 2016 publiziert wure (ÖROK, 2016). Folgende Empfehlungen für 

das „Risikomanagement für gravitative Naturgefahren in der Raumplanung“ wurden 

vorgeschlagen: 


1. Integriertes Naturgefahrenmanagement: Durch ein integriertes Naturgefahrenmanagement 

soll langfristig eine möglichst große Sicherheit vor allen Naturgefahren erzielt werden. 

2. Risikoorientierte Raumplanung: Die räumliche Verteilung von Nutzungen und Bautätigkeiten 

soll so gesteuert werden, dass Beeinträchtigungen durch gravitative Massenbewegungen 

möglichst gering gehalten werden. 

3. Präventive Aufgabe der Raumplanung und des Bauwesens: Im Raumordnungs- und 

Baurecht sind die spezifischen Gegebenheiten gravitativer Naturgefahren durch Nutzungsbeschränkungen 

und Bauverbote verstärkt zu berücksichtigen. 

4. Zusammenwirken vielfältiger Fachmaterien: Eine effiziente Verknüpfung der unterschiedlichen 

Instrumente, Schutzmaßnahmen und Finanzmittel ist verstärkt anzustreben, wobei 

aktuelle Informationen über gravitative Naturgefahren die fachliche Grundlage bilden 

sollen. 

5. Raumbezogene Daten und Informationen: Informationen über gravitative Gefahrenbereiche 

sind möglichst umgehend für den raumrelevanten Bereich zu erheben, in Karten 

darzustellen und regelmäßig anzupassen. 

6. Generelle Systematik der kartographischen Gefahrendarstellung: Gravitative Naturgefahren 

sind systematisch in unterschiedlichen Karten für verschiedene Planungsebenen darzustellen, 

wobei auch das Modell der Gefahrenzonenplanung Anwendung finden soll. 

7. Definition von Sicherheitsniveaus: Unter Berücksichtigung der raumordnungsrechtlichen 

Schutzziele sind einheitliche Sicherheitsniveaus bezüglich gravitativer Naturgefahren 

festzulegen. 

8. Risikokommunikation und Risk Governance: Eine verbesserte Risikokommunikation soll 

über gravitative Naturgefahren, insb. auch über die langfristigen Wirkungszusammenhänge 

und das Restrisiko, sowie über die spezifischen Karten und Maßnahmen informieren. 

SCHLUSSFOLGERUNGEN UND AUSBLICK 

Eine risikoorientierte Raumnutzung soll dazu beitragen, dass keine wesentliche Erhöhung des 

Schadenpotentials bzw. eine Reduktion möglicher Schäden durch Naturgefahren erfolgt 

sowie durch eine frühzeitige Berücksichtigung von Naturgefahren im Planungsprozess keine 

untragbaren Risiken vorhanden sind und sich entwickeln können. In der Raumplanung sind 

die spezifischen Gegebenheiten gravitativer Naturgefahren verstärkt zu berücksichtigen, 

wobei grundsätzlich Gebiete mit hohem Gefahrenpotential nicht bebaut werden sollen. Für 

bestehende Bauwerke und Nutzungen ist auf Basis einer Risikobewertung eine Erhöhung der 

Sicherheit (Risikoreduktion) anzustreben. Durch eine risikoorientierte Raumplanung soll 

weiters der Ressourcenaufwand für technische Schutz- und allfällige Wiederherstellungsmaßnahmen 

nach Ereignissen künftig deutlich reduziert werden, wobei planerische, 

nicht-bauliche Maßnahmen technischen Eingriffen vorzuziehen sind. 

Auch wenn ÖROK-Empfehlungen aus formaler Sicht unverbindlichen Charakter aufweisen, 

so wurden doch im Rahmen der ÖREK-Partnerschaft fachliche Grundlagen geschaffen und 

Standards definiert, die eine Umsetzung in der Raumordnungspolitik nahelegen und eine 


stärker Ausrichtung raumwirksamer Entscheidungen an den Ergebnissen der Gefahrenanalyse 

und den risikoorientierten Schutzzielen ermöglicht. Jedenfalls konnte im Kielwasser der 

ÖROK ein Sektor übergreifendes Fachnetzwerk etabliert werden, welches geeignet ist auf 

Basis der Ergebnisse der Partnerschaft einheitliche fachliche Grundsätze und Methoden 

voranzutreiben. Eine normative Umsetzung der entsprechenden Empfehlungen insb. im 

Raumordnung- und Baurecht würde wesentlich zur Verbesserung des Risikomanagement für 

gravitative Naturgefahren beitragen. 

LITERATUR 

- Ammann W. J. (2006): Risk concept, integral risk management and risk governance. 

In: Ammann W. J., Dannenmann S., Vulliet L. (Hrsg.): RISK21 – Coping with Risks due to 

Natural Hazards in the 21st Century (Proceedings of the Risk21 Workshop, Ascona, Switzerland), 

Taylor & Francis: 3–24. 

- Baumann R., Guggisberger C., Loat R., Diethelm I. (2000): Empfehlung Raumplanung und 

Naturgefahren. Bundesamt für Raumplanung/BRP, Bundesamt für Wasser und Geologie/ 

BWG, Bundesamt für Wald/BUWAL, Bern. 

- BMLFUW (2011): Grund genug? Flächenmanagement in Österreich - Fortschritte und 

Perspektiven. Umweltbundesamt Wien. 

- Bätzing W. (2003): Die Alpen: Geschichte und Zukunft einer europäischen Kulturlandschaft. 

CH Beck. 

- Camenzind R., Loat R. (2014): Risikobasierte Raumplanung - Synthesebericht zu zwei 

Testplanungen auf Stufe kommunaler Nutzungsplanung. Bern, Nationale Plattform Naturgefahren 

PLANAT, Bundesamt für Raumentwicklung, Bundesamt für Umwelt. 

- Glade, T. (2015): Gravitative Naturgefahren: Entstehung, Wirkungen und Risikomanagement. 

In: ÖROK (Hrsg.): Risikomanagement für gravitative Naturgefahren in der Raumplanung, 

ÖROK-Schriftenreihe Bd. 193. 

- Glade T., Krause M. (2015). Gefahrenzonenkarten und Gefahrenhinweiskarten: Erhebungsansätze, 

Anwendung und Bedeutung für die Raumplanung. In: ÖROK (Hrsg.): Risikomanagement 

für gravitative Naturgefahren in der Raumplanung, ÖROK-Schriftenreihe Bd. 

193. 

- Glade T., Petschko H., Bell R., Leopold P., Proske H. (2013): Karten zu gravitativen Massenbewegungen: 

Möglichkeiten und Grenzen ihrer Aussagekraft.- Raumdialog - Magazin für 

Raumplanung und Regionalentwicklung in Niederösterreich 2: 10-13. 

- Hattenberger D. (2004): Rechtliche Aspekte betreffend Lawinenschutzbauten (Teil I). 

Bau rechtliche Blätter: 221f. 

- Kanonier (2012): Umsetzung von Gefahrenkarten und Gefahrenzonenplänen in der 

Raumordnung und im Bauwesen. In: Suda J. und Rudolf-Miklau F. (Hrsg.): Bauen und 

Naturgefahren, Verlag Ambra (vormals Springer) Wien: 199–225. 

- Kanonier A. (2015): Rechtsgrundlagen des Schutzes vor gravitativen Prozessen (Muren, 

Lawinen, Steinschlag, Rutschungen) im Bundesrecht sowie Raumordnungs- und Baurecht 


der Länder. In: ÖROK (Hrsg.): Risikomanagement für gravitative Naturgefahren in der 

Raumplanung, ÖROK-Schriftenreihe Bd. 193. 

- Loat R. (2015): Risikoorientierter Umgang mit gravitativen Naturgefahrenrisiken in der 

Raumplanung am Beispiel der Schweiz. In: ÖROK (Hrsg.): Risikomanagement für gravitative 

Naturgefahren in der Raumplanung, ÖROK-Schriftenreihe Bd. 193. 

- Mattle A. (2015): Gravitative Naturgefahren aus Sicht einer Gemeinde. In: ÖROK (Hrsg.): 

Risikomanagement für gravitative Naturgefahren in der Raumplanung, ÖROK-Schriftenreihe 

Bd. 193. 

- Nationale Plattform Naturgefahren PLANAT (2009): Risikokonzept für Naturgefahren, 

Leitfaden. Strategie Naturgefahren Schweiz, Schlussbericht 2. Phase. 

- Österreichische Raumordnungskonferenz (2011): Österreichisches Raumentwicklungskonzept 

(ÖREK) Österreichische Raumordnungskonferenz (2014): ÖROK-Atlas. 

- Österreichische Raumordnungskonferenz (2016). ÖROK-Empfehlungen Nr. 54: Risikomanagement 

für gravitative Naturgefahren in der Raumplanung. Broschüre, Wien: 

http://www.oerok.gv.at/fileadmin/Bilder/2.Reiter-Raum_u._Region/1.OEREK/OEREK_2011/ 

PS_Risikom/2016-02-24-Brosch%C3%BCre_Risikomanagment_FINAL-54_Internetversion. 

pdf. 

- Papathoma-Köhle M. & Glade T. (2013): The role of vegetation cover change for landslide 

hazard and risk. In: Renaud G., Sudmeier-Rieux K. & Estrella M. (Eds.): The Role of Ecosystems 

in Disaster Risk Reduction. UNU-Press, Tokyo: 293-320. 

- Promper C., Rudolf-Miklau F. (2015): Die ÖREK-Partnerschaft „Risikomanagement für 

gravitative Naturgefahren“: Problemstellung und fachpolitische Ziel. In: ÖROK (Hrsg.): 

Risikomanagement für gravitative Naturgefahren in der Raumplanung, ÖROK-Schriftenreihe 

Bd. 193. 

- Promper C., Rudolf-Miklau F., Hinterleitner R. (2015): Begriffe und Definitionen (Anhang). 

In: ÖROK (Hrsg.): Risikomanagement für gravitative Naturgefahren in der Raumplanung, 

ÖROK-Schriftenreihe Bd. 193. 

- Psenner R. (Hrsg.) (2006): Die Alpen im Jahr 2020. IUP Innsbruck University Press. 

- Rudolf-Miklau F., Stix E., Mehlhorn S., Sauermoser S. (2014): Strategic Partnership for 

“Risk Management in Spatial Planning concerning Gravitational Hazards”. Proceedings of 

World Landslide Forum 3, 2-6 June 2014, Beijing. 

- Schwarz L., Kociu A., Hagen K., Rudolf-Miklau F. (2014): Landslide assessment for spatial 

planning – the new Austrian ACSP-Standards. 13th Congress INTERPRAEVENT 2016, Luzern 

(Beitrag eingereicht). 

- Wagner E. (2008): Katastrophenprävention: Optionen de lege lataund de lege ferenda.- In: 

Kerschner (Hrsg.): Handbuch Naturkatastrophenrecht. Schriftenreihe Recht der Umwelt Bd. 

24. Verlag Manz Wien. 

- Wagner E., Jandl C. (2013): Zivil- und verwaltungsrechtliche Fragestellungen des Baus und 

der Instandhaltung von Steinschlagschutzbauwerken. Schriftenreihe Umweltrecht und 

Umwelttechnikrecht Bd. 4, Linz. 



Water management as a part of civil engineering 

sector in Slovenia 

Jošt Sodnik, MSc in Civ. Eng. 1,2 ; Blaž Kogovšek, Univ.Diploma in WMCE²; Matjaž Mikoš, Prof.Dr. MSc in Civ. Eng.² 

ABSTRACT 

The construction industry including water management sector represents an important part 

of the Slovenian economy and during the recession that started in 2009, this industry has 

shrunk for ~10-20% annually, and in 2012 it dropped below 50% of the 2008 level. 

Compared to Austria, Slovenia invests up to more than three times less into the water 

management sector, including regular maintenance. Flood hazard in Slovenia is increasing 

rather than decreasing, and organisation of the water management sector was claimed to a 

large extent to be one of the major reasons for the present deteriorating situation. In 2016, 

a newly established Directorate for Waters of the Republic of Slovenia should bring fresh air 

into the scene. Considering rather poor condition of water infrastructure in Slovenia and 

very high flood hazard, water management must become one of the top priorities in years 

to come. 

KEYWORDS 

financial investments; regular maintenance; Slovenia; water infrastructure; water management 

INTRODUCTION 

A healthy status of the Slovenian construction sector is not only the responsibility of the 

sector itself, part of which is also the water management sector, but of the whole society. 

The construction industry represents an important part of the Slovenian economy (around 

6% GDP) and before the recession that started in 2009, in Slovenian construction sector 

there were more than 90.000 employees in 4600 constructional companies, plus 10.000 

self-employed workers directly working in construction business. In 2009, the construction 

sector started to shrink for about 10-20% annually, and in 2012 it was already well below 

50% of the 2008 level. Engineering works (including water management and flood protection 

measures) represents 50% of Slovenian construction sector activities (the rest is 

construction of buildings). In 2012, five out of ten biggest Slovenian construction companies 

were bankrupt and three more were in serious financial problems. Revival of the construction 

sector as a part of the industrial sector is crucial for the revival of Slovenian economy in 

general to come of the financial and economic crisis started in 2009. In the last two years 

construction sector was mainly growing due to realisation of the EU Cohesion Fund projects 

for building sewage systems in rural areas. With the end of 2015 the majority of these 

1 Water Management Company, Kranj, Kranj, SLOVENIA, jost.sodnik@gmail.com 

2 University of Ljubljana, Faculty of Civil and Geodetic Engineering, SLOVENIA 



projects will finish and 2016 will be crucial for many of the construction companies (including 

water management sector) in Slovenia since there are no new infrastructural projects in 

sight in the near future. 

WATER MANAGEMENT IN SLOVENIA 

In Slovenia, water management is governed by the Ministry of the environment and spatial 

planning that prepares all national legislation (harmonized with EU directives). The operational 

body on the national level is Slovenian Environment Agency (ARSO). ARSO is the 

water manager on the national level and is in charge for all spatial planning permits, 

maintenance works on water infrastructure, water infrastructure cadastre, restoration plans 

after natural disasters, supervision, etc. Slovenia is divided into eight river sub-basins 

(watersheds). Each watershed has a concessionaire for maintenance works on water 

infrastructure and other tasks holding a concession contract with the Ministry. Concessionaires 

are local Water management companies which are responsible for carrying out 

maintenance works, according to plans, prepared by the Agency. 

In the last two decades all these companies have become private firms and the Republic of 

Slovenia owns only 25% + 1 share or even less in some cases. In the last strategic plan for 

final privatisation of selected state capital investments, the Government of the Republic of 

Slovenia recognized Water management companies as strategic property and decided not to 

fully privatise them. On the other hand, state owned institutions are not acceptable in 

Slovenia since we are still in process of privatisation. Water management “in the field” is 

carried out through these companies. Scope of these measures is limited with funding 

possibilities of the Agency and the Ministry. Funding of maintenance works was shrinking 

and today it represents only about 10-20% of yearly business of Water management 

companies. These companies were forced to start building other structures on the free market 

(canalization systems, roads, other infrastructure). Two of eight water management companies 

went bankrupt due to lack of business. Analysis show that in the period from 1986 to 

1998 funding of water management in Slovenia shrank from 0.71% to 0.07% of GDP (Umek 

and Banovec, 1998). This trend lead slowed down the education process and gaining 

operational experiences of professionals. Civil engineering study programs have become 

non-attractive and new engineers have hard time to gain operational experiences. 

WATER MANAGEMENT FINANCING IN SLOVENIA AND A COMPARISON WITH AUSTRIA 

Slovenia and Austria are comparable in terms of topography, climate, torrents and rivers. 

Population density is also comparable (105 p/km 2 ). Austria has 100,000km of waterways with 

a density of 1.2km/km 2 (BMLFUW, 2010; 2012a) and Slovenia has 28,000km of rivers with a 

density of 1.4 km/km 2 (Bat et al., 2003). Due to low financing of water managements sector 

in Slovenia in the last two decades, water infrastructure is in a very poor condition and 

therefore the flood hazard level of many densely populated areas is very high. The analysis 

of financial investments in the water management sector (water infrastructure for river 

engineering and in general for the protection against floods, torrents, landslides, avalanches, 


and erosion) of Austria and Slovenia in the period 2002-2014 was prepared. Given the 

differences between the two countries in gross domestic product, population and length of 

the hydrographic network, in all cases investments in Slovenia are significantly lagging 

behind Austria. 

Figure 1: Comparison of “water management investments” as % GDP between Slovenia and Austria. 

Slovenia invests in the water management sector for the mentioned purposes annually on 

average between 0.02 and 0.03% of GDP, Austria, on average, 0.055% (Fig. 1). Slovenia 

invests comparably to Austria only in years with extreme floods that have caused damages of 

several 100 million EUR (2008 and 2009 after massive flooding in September 2007). 

Figure 2: Comparison of “water management investments” as €/capita between Slovenia and Austria. 


With regard to the number of inhabitants, the annual investment into water management in 

Slovenia is 5 EUR per capita and in Austria is 17 EUR per capita (Fig. 2). Also the investments 

for a kilometre of a watercourse in Slovenia were on average only about 400 EUR per km and 

in Austria nearly 1500 EUR per km in the investigated period (Fig. 3). 

Figure 3: Comparison of “water management investments” as €/km between Slovenia and Austria. 

The ratios for Slovenia are even worse, because we used in the presented analysis for Austria 

only data on its state funding. On average, every year the Austrian Federal Government 

makes available funds (subsidies are granted subject to the provisions of the Hydraulic 

Engineering Assistance Act) from the Disaster Relief Fund of the Federal State to the amount 

of € 69.9 million for the torrents, avalanche, and erosion control. Together with contributions 

from the Federal Provinces and stake holders (municipalities, water corporations, others) 

funds to the amount of almost € 122 million are thus available annually for investments into 

active control measures (BMLFUW, 2012a;b). Since the 2002 Floods, Austria invested all 

together 2.9 billion EUR into protection against natural hazards. According to an action plan 

called „Flood Safe Austria“ that was accepted in Austria after the 2013 Floods, will the 

Austrian Federal State invest around 1 billion EUR in the period 2014-2019 into protection 

of their citizens (BMLFUW, 2016). 

Water management funding in Slovenia is based only on national funds. Municipalities do 

not have their own funds to invest in that sector. There have been some changes in that field, 

and some municipalities realized that national funds are not sufficient to ensure a good state 

of water infrastructure and they started to ensure their own budget to finance parts of the 

maintenance works, mostly parts closely connected with their own infrastructure (municipality 

roads, sewage systems, water supply systems). But these cases are rather exceptions and 

such funds do not present amounts worth mentioning in the overall statistics. This is also one 

of the steps forward that could improve the state in Water management sector in Slovenia. 


Figures 1-3 show that Slovenia lags behind Austria, except in years 2008 and 2009 when all 

the mitigation measures and restoration of water infrastructure after massive flooding in 

September 2007 were carried out. After 2007 floods occurred in 2009, 2010, 2012 and 2014. 

Damage exceeded 0.05% of national GDP in 2012. GDP data were taken from the World 

Bank (data.worldbank.org). 

Table 1: Damage on water infrastructure in Slovenia (% of national GDP). 

After last large flood events in Slovenia the maintenance funding remained on the same level 

as before (8.5 mil € on the national level); instead the Ministry prepared restoration plans and 

provided additional financing. None of these restoration programs reached the planned 

amount due to lack of financing in the following fiscal years and due to new flood events. 

Despite financing problems these restorations programs improved state of water infrastructure 

in flooded areas. Numerous check dams and retention dams were built on small 

torrential tributaries, since sediment transport and woody debris were one of the most 

pressing problems during past flood events. Nevertheless, mostly were these measures rather 

a first-aid than actually improving the flood-safety situation. 

In the end of 2014, after the fifth massive flood in 8 years, the new government set the flood 

protection and good condition of water infrastructure as its top priorities. Government 

prepared so called action plans which should improve the state of water infrastructure in 

Slovenia (see Fig. 6 for an example of a neglected and therefore damaged check dam) and 

reduce flood hazard. This action plan was divided into four major chapters (MOP, 2014): 

– The first part was short term measures to compensate lack of maintenance in the past years 

with amount of 11.9mil €. In 2015, only 11.1mil€ of the action plan was carried out due to 

lack of annual financing. Majority of the financing was used to carry out the restoration 

measures in the recently flooded areas. So, the first part of the action plan turned into 

another remediation program, as we know them in the past to remediate devastated 

flooded areas. 

– The second part of the action plan was an increase of the regular maintenance funding for 

water management to at least 25 mil€ on the national level per year. Also according to our 

interpretation, the annual maintenance budget in Slovenia in the field of water manage- 


ment should raise from 8.5 to 25 million € per year – this is ~3% of the estimated value of 

water infrastructure in Slovenia of around 800 million €. In 2015 that was not applied and 

the maintenance budget stayed the same as in the last few years (7.6 mil €/year). The third 

part (the last part related to water management, since the fourth part is related to road 

infrastructure damaged in floods) of the action plan was long term measures including 

administrative changes, changes of national funding and greater focus on EU funding of 

measures to reduce flood hazard. 

Figure 4: A degraded check dam on the Belca torrent in NW Slovenia. 

On administration level the Parliament of Slovenia in July 2015 accepted an amendment to 

the Water Law (ZV-1E, 2015) and established the Directorate for Waters of the Republic of 

Slovenia (DRSV) that will start its work in January 2016 as a constitutional part of the 

Ministry of the Environment and Spatial Planning. The situation could have been better, if 

only we have succeeded to raise more than only 22.5% of the available EU structural funds 

to be spent in Slovenia for these purposes in the period of 2007-2013. The Directorate for 

Waters will be a stand-alone fiscal user of the ministry budget, and will be in charge for 

investments in water management in Slovenia – the staff will be recruited from employees of 

the Slovenian Environment Agency (ARSO - http://www.arso.gov.si/en/), Institute for Water 

of the Republic of Slovenia (IzVRS - http://www.izvrs.si/?lang=en), and internal units of the 

Ministry of the Environment and Spatial Planning (MOP - http://www.mop.gov.si/en/), 


especially the Water and Investments Directorate. The main reason for the directorate 

establishment was reaching findings that flood hazard in Slovenia is increasing rather 

decreasing (after the last few major floods), and the water management organisation was 

claimed to a large extent to be one of the major reasons for the present unsatisfactory 

situation. What concerns about new Directorate is that it brings only administrative changes 

with existing experts and mostly officers and does not bring necessary professional strengthening 

to water management sector. 

As a basis for a more effective work of the future WD will be a Water Management Strategy 

– to be prepared before the end of 2015 . Part of it is also a Flood Risk Management Plan 2016 

– 2021 (MOP, 2015) and Basin Management Plan for the same period, among other documents. 

The Flood Risk Management plan (National plan for flood risk reduction) recognizes 

61 locations with high flood risk where implementation of mitigation measures should be a 

priority. In last 25 years flood caused 1,800 million € of direct damage, only in last 10 years 

1,200 million € of damage. On average in Slovenia floods cause 150 million € of damage. 

The main question that remains is the scope of funding in the future. No administrative 

change can basically ensure reduced flood risk without increased and effectively spent 

funding. An important step in the past was done by establishing a web platform called eVode 

(http://evode.arso.gov.si/), where stakeholders can find different evidences, studies and other 

data on water use and water management. A part of these publically available data are now 

also LiDAR data for the whole area of the Republic of Slovenia, which is a very advanced 

approach and enables wide usage of high resolution data for flood hazard mapping, spatial 

planning etc. Also an important ongoing activity is the determination of water areas. 

Based on the Water Act, a water area is an area with permanent or occasional water presence 

(stream channels) including flood areas (flood plains) and areas close to watercourses till first 

geomorphological change – tough, water areas are not covering full river corridors. 

This project includes cooperation of water management and geodetic experts and will provide 

very valuable data layer which will be important in the process of spatial planning. 

WATER MANAGEMENT AND CONSTRUCTION SECTOR IN SLOVENIA 

According to the Standard Classification of Activities (version 2008), water management is 

part of F. Construction (in former Yugoslavia, water management was a separate activity 

outside construction) – Construction sector is subdivided to: 

– F41. Construction of buildings, 

– F42. Civil engineering (among others F42.91 Construction of water projects) and 

– F.43 Specialised construction activities. 

The peak number of employed workers in construction sector in Slovenia was in October 

2008 with over 92,000 (more than 11% of all employed persons in Slovenia) and the value 

of the executed construction works in 2008 in Slovenia was estimated at 3,551 billion EUR 

(1,727 billion EUR for civil engineering) – this dropped down to 1,927 billion EUR in 2014 

(1,269 billion EUR for civil engineering). In 2015, Slovenia has already signed first 5 projects 

in the field of drinking water supply and collecting and treating of waste water for over 


150 million EUR to be partially co-financed from the European Structural Funds and the 

Cohesion Fund in the 2014-2020 programming period. In the field of flood protection 

projects, the available funds in this period are estimated to 600 million EUR; an amount to 

be of importance to revive the construction sector in Slovenia, where large investments into 

highway construction has recently stopped (the value of close to 800 km highways and 

motorways in Slovenia is estimated at 6+ billion EUR – for comparison purpose: the value of 

close to 10,000 water management structures in Slovenia is estimated to close to 800 million 

EUR). The realization of these Cohesion Fund projects still depends on the success of 

Slovenian bureaucracy which has prevented the realisation of 2010-2015 period projects for 

the first three years and thus the majority of projects started only in the second part of 2013; 

a fact that additionally hit the Slovenian construction sector in the past 5 years. 

CONCLUSIONS 

Considering very poor condition of water infrastructure in Slovenia and very high flood 

hazard, water management and flood mitigation measures must become one of the top 

priorities in years to come. Previously described administrative changes in Slovenia are a step 

forward and rule out the administrative obstacles for efficient water management in Slovenia. 

But for now (early 2016) all these changes still did not bring the necessary changes in 

funding. Maintenance is still inadequate and projects for EU funding are still not a top priority 

of the Government. Increased funding of water management will reduce flood hazard and 

consequently also help to a certain level Slovenian construction sector to recover after 

recession in 2009 when the Slovenian national highway project was terminated. The results 

of our study were presented on two Slovenian national congresses on water management 

and natural disasters in Slovenia and published in the national journal on Civil engineering 

(Sodnik and Mikoš, 2013; Sodnik etal., 2015). Authors of the study were actively involved in 

the preparation of the administrative changes in water management in Slovenia. The state 

of the mind of the authorities and decision makers must change and maintenance works and 

flood protection measures must be carried out instead of remediation projects after each 

flood event. 

REFERENCES 

- Bat, M., Dobnikar Tehovnik, M., Mihorko, P., Grbović, J. (2003). Tekoče vode (Running 

waters). In: Uhan, J. (Ed.), Bat, M. (Ed.), Vodno bogastvo Slovenije. Agencija Republike 

Slovenije za okolje, Ministrstvo za okolje, prostor in energijo Republike Slovenije, Ljubljana, 

27-37. http://www.arso.gov.si/vode/publikacije%20in%20poro%c4%8dila/Vodno_bogastvo_2tekoce_vode.pdf 

- BMLFUW (2010). Nationaler Gewässerbewirtschaftungsplan 2009 – NGP 2009. Bundesministerium 

für Land- und Forstwirtschaft, Umwelt und Wasserwirtschaft, Wien, 225 p. 

- http://www.lebensministerium.at/publikationen/wasser/wasserwirtschaft_wasserpolitik/ 

nationaler_gewaesserbewirtschaftungsplan_2009.html 


- BMLFUW (2012a). Wildbach- und Lawinenverbauung in Österreich. Bundesministerium 

für Land- und Forstwirtschaft, Umwelt und Wasserwirtschaft, Sektion Forstwesen, Wien, 

Austria, 27p. 

- https://www.bmlfuw.gv.at/forst/schutz-naturgefahren/wildbach-lawinen/broschueren/ 

WildbachundLawinen.html 

- BMLFUW (2012b). Schutz vor Naturgefahren in Österreich – 2002 - 2011. Bundesministerium 

für Land- und Forstwirtschaft, Umwelt und Wasserwirtschaft, Sektion Forstwesen, Wien, 

Austria, 51p. 

- http://www.wasseraktiv.at/resources/files/2012/8/13/2798/schutz-vor-naturgefahren-2002- 

2011-end-aug2012.pdf 

- BMLFUW (2016). Investitionen in den Schutz von Naturgefahren. Bundesministerium 

für Land- und Forstwirtschaft, Umwelt und Wasserwirtschaft, Wien, Austria, https://www. 

bmlfuw.gv.at/forst/schutz-naturgefahren/wlv2014-2019.html 

- MOP (2014). Akcijski načrt interventnih aktivnosti zaradi poplav. Ministrstvo za okolje in 

prostor Republike Slovenije, Ljubljana, 81 p. 

- http://www.mop.gov.si/fileadmin/mop.gov.si/pageuploads/medijsko_sredisce/2014/11_november/28_AN_poplave/Akcijski_nacrt_poplave_MOP.pdf 

- MOP (2015). Načrt zmanjševanja poplavne ogroženosti (NZPO Si) – predlog ver 1.04, 

Ministrstvo za okolje in prostor Republike Slovenije, Ljubljana, 357 p. 

- http://www.mop.gov.si/fileadmin/mop.gov.si/pageuploads/podrocja/voda/nzpo/NZPO_ 

SLO_2015_12_08.pdf 

- Sodnik J., Mikoš M. (2006). Estimation of magnitudes of debris flows in selected torrential 

watersheds in Slovenia. Acta geographica Slovenica 46/1: 93-123. 

- Sodnik J., Mikoš M. (2013). Vodarstvo in vzdrževanje vodne infrastrukture v Sloveniji = 

Water Management and Maintenance of Water Infrastructure in Slovenia. Gradbeni vestnik 

62/8, 166-173. 

- Sodnik J., Kogovšek B., Mikoš M. (2015). Vlaganja v vodno infrastrukturo v Sloveniji in v 

Avstriji = Investments into Water Infrastructure in Slovenia and in Austria. Gradbeni vestnik 

64/1, 3-12. 

- Umek, T., Banovec, P. (1998). Trenutna institucionalna ureditev gospodarjenja z vodami 

v Republiki Sloveniji in trendi nadaljnjega razvoja, Mišičev vodarski dan 1998, 112-120. 

http://mvd20.com/LETO1998/R20.pdf 

- ZV-1E (2015). Zakon o spremembah in dopolnitvah Zakona o vodah (ZV-1E). 

Official Gazette of the Republic of Slovenia, Nr. 56 (of July 29, 2015), p. 6551 



Method for a risk based and transparent allocation 

formula of the remaining costs applied to the 

protection measures project "Laui Sörenberg", 

municipality of Flühli, Switzerland. 

Methode für einen risikobasierten und transparenten 

Restkostenteiler beim Schutzprojekt "Laui Sörenberg", 

Gemeinde Flühli, Schweiz. 

Roland Stalder, Engineer in forestry¹; Karl Grunder, Engineer in Forestry¹; Guido Küng² 

ABSTRACT 

The landowners benefiting from the realization of protection measures are required to contribute 

to the costs. The allocation of the costs is legally based on the Ordinance on perimeter. 

It stipulates an allocation formula assessing the added value for the property after implementing 

the protection measures. In the following, a method is presented which equitably 

allocates the remaining costs based on official and scientific facts and which is transparent and 

comprehensible for the landowners who must contribute. The quantifying of the added value 

is derived from the benefit regarding safety and spatial planning. The focus is on the safety 

benefit, therefore it is counted double. The benefit is multiplied by the land and building 

values for each piece of property. These resulting figures lead to the allocation formula for the 

contributions of the remaining costs. This method was developed and successfully applied for 

the significant and complex protection project «Laui Sörenberg», municipality of Flühli, 

canton of Lucerne. The Federal Supreme Court confirmed the decision of the local authority. 


Bei der Realisierung von Schutzmassnahmen gegen gravitative Naturgefahren sind die 

profitierenden Grundeigentümer an den Kosten der Massnahmen zu beteiligen. Die Verteilung 

der Kosten basiert rechtlich auf der Kantonalen Verordnung über die Grundeigentümer- 

Beiträge an öffentliche Werke (Perimeterverordnung). Diese fordert einen Verteilschlüssel, 

der sich an den Vorteilen orientiert, die den Grundstücken aus den Schutzmassnahmen 

entstehen. Nachfolgend wird eine Methode vorgestellt, die einerseits die Restkosten auf der 

Basis von amtlichen und naturwissenschaftlichen Fakten gerecht verteilt und andererseits für 

die Beitragspflichtigen transparent und nachvollziehbar ist. Zur Quantifizierung der Vorteile 

wird der Nutzen bezüglich Sicherheit und Raumplanung hergeleitet. Der Sicherheitsnutzen 

steht dabei im Vordergrund und wird höher gewichtet. Der Nutzen wird mit Grundstück- und 

1 oeko-b ag, Schüpfheim, SWITZERLAND, roland.stalder@oeko-b.ch 

2 Municipal government Flühli, SWITZERLAND 



Gebäudewert zur Gesamtteilerzahl pro Grundstück multipliziert. Diese ermittelten Teilerzahlen 

bilden den Verteilschlüssel für die zu leistenden Beiträge an die Baukosten. Diese 

Methode wurde für das grosse und komplexe Schutzprojekt Laui Sörenberg, Gemeinde 

Flühli, Kanton Luzern, entwickelt und erfolgreich umgesetzt. Die von den Beschwerdeführern 

angerufenen Gerichte stützten die Entscheide der lokalen Behörde. 

KEYWORDS 

protection measures; method; allocation formula; risk based; transparent 

EINLEITUNG 

Das Gebiet Laui Sörenberg in der Gemeinde Flühli, Kanton Luzern, ist ein prähistorisches 

Bergsturzgebiet. Im Jahr 1910 sackte eine grosse Bergmasse um rund vierzig Höhenmeter ab 

(Einwohnergemeinde Flühli 2011, S. 1f.). Der Talboden wurde meterhoch mit Schutt und 

grossen Steinbrocken übersart und die Waldemme vorübergehend zu einem See aufgestaut. 

In den Jahren 1922, 1986 und 1999 ereigneten sich erneut grössere Murgänge, wobei insbesondere 

das jüngste Ereignis zu grossen Schäden im Siedlungsgebiet führte. Nur dank 

glücklichen Umständen kamen keine Personen zu Schaden. 

Das gefährdete Gebiet wurde seit ca. 1960 vorwiegend mit Ferienhäusern überbaut. 

Mit der Einführung der Raumplanung wurde es 1973 als provisorisches Gefahrengebiet 

ausgeschieden. Auf der Basis der Gefahrenkarte 1996 wurde ein Wasserbauprojekt ausgearbeitet, 

welches zwischen 2009 und 2015 realisiert wurde. Die Schutzmassnahmen sollen 

die darunterliegenden rund 700 Wohneinheiten mit einem Versicherungswert von 

ca. 200 Millionen Franken vor Zerstörung oder grösseren Schäden bewahren. Das Personenrisiko 

wird fast vollständig beseitigt und das Sachwertrisiko auf sehr seltene Ereignisse 

begrenzt. Die Baukosten betrugen rund 18 Millionen Franken. Der Regierungsrat des 

Kantons Luzern hat bei der Projektbewilligung (Regierungsrat des Kantons Luzern 2006, 

S. 18) festgelegt, dass die interessierten Grundeigentümer 13% dieser Baukosten selber zu 

tragen haben (CHF 2’340’000.-). Diese Kosten wurden basierend auf der kantonalen 

Perimeterverordnung mittels Perimeterverfahren auf die Beitragspflichtigen verteilt. 

Die übrigen 87% der Baukosten wurden durch die öffentliche Hand finanziert (Bund 43%, 

Kanton 30.7% und Gemeinde 13.3%). 

Gegen den Perimeterentscheid des Gemeinderats Flühli gingen 38 Einsprachen ein. 

Die letzte wurde 2015 durch das Bundesgericht abgewiesen. 

KONTEXT 

Das Schutzprojekt Laui Sörenberg dauerte von der Vorstudie bis zum Abschluss rund 

16 Jahre. Dabei war das Perimeterverfahren eines unter zahlreichen Verfahren. Die Komplexität 

des Projekts und die Interaktion der damit verbundenen politischen, projekt- und 

verfahrenstechnischen sowie gesellschaftlichen Prozesse forderte vom Perimeterverfahren 

weit mehr als eine arithmetische Berechnung der Kostenanteile. 


Den rechtlichen Rahmen bildet die Perimeterverordnung, welche auch einige Vorgaben zur 

Methodik macht (z.B. Klassenbildung und Grundmass, siehe Kapitel „Methode“). Die Perimeterverordnung 

verlangt, dass sich der Verteilschlüssel an den Vorteilen orientiert, die den 

Grundstücken aus den Schutzmassnahmen entstehen. Dabei haben das Gerechtigkeits- und 

das Zumutbarkeitsprinzip oberste Priorität. 

Die Bevölkerung setzte sich intensiv mit dem Schutzprojekt auseinander, bisweilen flossen 

emotionale, ideelle und ethische Aspekte in die Diskussion ein. Die Gewährung des rechtlichen 

Gehörs erfolgte bereits vorgängig zum Perimeterverfahren im Rahmen der Projektauflage 

(mit 60 Einsprachen) und der Auflage des Zonenplans (mit 40 Einsprachen). Dies 

veranlasste die Projektgruppe, für den Kostenteiler die bereits vertrauten und anerkannten 

Grundlagen „Intensitätskarten“ und „Gefahrenzonenplan“ zu verwenden. Zur Herleitung des 

Risikos wurde deshalb eine vereinfachte Methode auf der Basis der schon eingeführten 

Grundlagen gewählt (in Anlehnung an Heinimann 1999, S.36; siehe Kapitel „Methode“). 

Der Kostenteiler basiert auf Sachrisiken, die Personenrisiken wurden nicht berücksichtigt. 

Einerseits verläuft der Gradient der Personenrisiken innerhalb des Perimeters gleichartig wie 

derjenige der Sachrisiken, daher ist dieser Analogieschluss für eine Abschätzung der relativen 

Risikoverteilung anwendbar. Andererseits befürchtete die Projektgruppe, dass der Dialog mit 

der Bevölkerung durch den Einbezug der Personenrisiken emotionaler würde, aufgrund 

ethischer Grundsatzfragen (Wie viel ist ein Menschenleben wert?) und wegen zusätzlichen 

variablen Parametern (Was ist eine „gerechte“ Aufenthaltsdauer in einem Ferienhaus im 

Gegensatz zum Wohnhaus?). 

Die Vorteile aus den Schutzmassnahmen wurden sowohl im Hinblick auf die Sicherheit, 

wie auch aus raumplanerischer Sicht hergeleitet (Einwohnergemeinde Flühli 2011, S. 4). 

Die Basis für den Sicherheitsnutzen bilden die Intensitätskarten, für den raumplanerischen 

Nutzen wurde der Gefahrenzonenplan verwendet (siehe Kapitel „Methode“). Der Sicherheitsnutzen 

ist mit dem Hauptziel des Schutzprojekts begründet, nämlich dem Schutz des 

Dorfes Sörenberg vor gravitativen Naturgefahren. Die Projektgruppe entschied sich für die 

zusätzliche Berücksichtigung des raumplanerischen Nutzens, da die Bevölkerung gegenüber 

Gefahrenzonen sehr sensibel war und den ideellen Wert von Entwicklungsmöglichkeiten auf 

den zu schützenden Parzellen als bedeutend einstufte. Der raumplanerische Nutzen beschreibt 

also die Verringerung risikobedingter Eigentumsbeschränkungen durch das Schutzprojekt. 

Der raumplanerische Nutzen wurde als sekundärer Aspekt halb so stark gewichtet, 

wie der Sicherheitsnutzen. 

Die Projektgruppe hat im Perimeterverfahren nach dem Grundsatz „Fakten statt behördliches 

Ermessen“ gearbeitet. Die Erarbeitung der naturwissenschaftlichen Fakten (Intensitätskarten 

und Gefahrenzonenplan) erfolgte nach Stand des Wissens und wurde durch Gemeinde und 

kantonale Fachstelle begleitet. Die amtlichen Fakten lassen ebenfalls keinen Spielraum für 

Ermessen. Zu erwähnen ist hierbei, dass die Verkehrswertschätzung des Bodens durch 

anerkannte Spezialisten durchgeführt wurde. 


Die Projektgruppe war also mit der Herausforderung konfrontiert, die Methode zur Erarbeitung 

des Kostenteilers so zu gestalten, dass a) die Perimeterverordnung korrekt umgesetzt 

wird, b) der Perimeterentscheid für die Betroffenen nachvollziehbar, transparent, gerecht und 

zumutbar ist und c) anstelle behördlichen Ermessens naturwissenschaftliche und amtliche 

Fakten verwendet werden. Es galt, diese unterschiedlichen Anforderungen gesamthaft 

bestmöglich zu erfüllen. 

METHODE ZUR ERARBEITUNG EINES KOSTENTEILERS 

Rechtlich basiert das Perimeterverfahren auf dem kantonalen Wasserbaugesetz (SRL 760), 

dem kantonalen Planungs- und Baugesetz (SRL 735) sowie der Verordnung über Grundeigentümer-Beiträge 

an öffentliche Werke (Perimeterverordnung) des Kantons Luzern (SRL 

732). Die Perimeterverordnung (PV) enthält bedeutende methodische Rahmenbedingungen. 

Die Grundlagen für die Berechnung der Höhe des einzelnen Beitrags bilden geeignete 

Grundmasse sowie die Klassenzahl (gemäss PV §7). Das Grundmass für die Grundstücke wird 

mit der Grundstückfläche, der Grundnutzung, dem Bodenwert sowie dem betroffenen Anteil 

hergeleitet. Das Grundmass für die Gebäude wird mit dem Gebäudegrundriss, dem Gebäudeversicherungswert 

sowie dem betroffenen Anteil hergeleitet (Einwohnergemeinde Flühli 

2011, S. 4f). Da den Grundstücken sowie den Gebäuden aus sicherheitsrelevanter wie auch 

aus raumplanerischer Sicht unterschiedliche Vorteile erwachsen, werden diese Vorteile 

jeweils separat berechnet und in einer Punktzahl ausgedrückt. Dabei stellen der raumplanerische 

Nutzen sowie der Sicherheitsnutzen je eine Kostengruppe im Sinne von PV §8 Abs. 2 

dar. Für die Kostengruppe „Raumplanerischer Nutzen“ werden die Gefahrenzonen vor und 

nach Massnahmen verwendet, für die Kostengruppe „Sicherheit“ die Intensitätskarten vor 

und nach Massnahmen (Einwohnergemeinde Flühli 2011, S. 5). Die resultierende Punktzahl 

(maximal 24 Punkte) entspricht der Einteilung in 24 Klassen im Sinne von PV §9. 

Die Erarbeitung des Kostenteilers erfolgt in 5 Schritten. Eine grafische Übersicht dazu findet 

sich in Fig. 1. 

Im 1. Schritt werden durch Vergleich der Intensitätskarten sowie der Gefahrenzonenpläne 

– jeweils vor und nach Massnahmen - diejenigen Grundstücke identifiziert, welche dank den 

Schutzbauten einen Sicherheits- oder raumplanerischen Nutzen erhalten. Nur diese Grundstücke 

sind beitragspflichtig. 

Im 2. Schritt wird für jedes beitragspflichtige Grundstück der Wert von Boden und Gebäude 

ermittelt (vgl. Fig. 1, blau umrahmtes Textfeld). Für die Berechnung des Bodenwerts werden 

auf der Basis einer Verkehrswertschatzung die Werte für die Grundnutzungen Bauzone (in 

Sörenberg CHF 200.- / m 2 ), Landwirtschaft (CHF 5.- / m 2 ), Grünzone (CHF 40.- / m 2 ) sowie 

Wald (CHF 0.50 / m 2 ) ermittelt. Durch Multiplikation mit der Grundstücksfläche wird der 

Bodenwert berechnet. Der Gebäudewert wird durch die Gebäudeversicherung Luzern zur 

Verfügung gestellt. Hierfür wird ein Stichdatum festgelegt, auf ausserordentliche Schatzungen 

wird verzichtet. Die Gebäudefläche lässt sich aus der Amtlichen Vermessung ermitteln. 


Fig. 1: Übersicht der Methode zur Erarbeitung eines Kostenteilers. 


Im 3. Schritt wird der raumplanerische Nutzen quantifiziert. Zur Herleitung werden die 

Gefahrenzonen vor und nach Massnahmen verglichen, sowohl für den Boden, als auch für 

die Gebäude (vgl. Fig. 1, grün umrahmtes Textfeld, oben). Das Grundstück erfährt dann einen 

Vorteil, wenn es durch die Massnahmen einer schwächeren Gefahrenzone zugeteilt oder 

gänzlich aus der Gefahrenzone entlassen wird. Der Nutzen wird in Punkten ausgedrückt. 

Der Wertebereich der Gesamtpunktzahl reicht von 0 Punkten (keine Zuteilung in schwächere 

Gefahrenzone) bis zu 24 Punkten (vollständige Entlassung aus roter Gefahrenzone) und 

entspricht den 24 Klassen im Sinne von Perimeterverordnung §9. Der Gesamtpunktzahl des 

raumplanerischen Nutzens wird ein Faktor zugeteilt, damit die Spannweite vom kleinsten 

zum grössten Beitragszahler verhältnismässig bleibt (vgl. Fig. 1, grün umrahmtes Textfeld, 

links unten). Der Faktor für das Gebäude wird mit dem Gebäudewert (2. Schritt) zur TeilerzahlGebäude, 

Raumplanung multipliziert, der Faktor für den Boden mit dem Bodenwert zur 

TeilerzahlBoden, Raumplanung. Die Herleitung der Punktetabellen und Faktoren wird am 

Ende des Kapitels beschrieben. 

Im 4. Schritt wird der Sicherheitsnutzen ermittelt. Zur Herleitung werden für drei Wiederkehrperioden 

(häufig, selten und sehr selten) die Intensitätskarten vor und nach Massnahmen 

verglichen, sowohl für den Boden, als auch für die Gebäude (vgl. Fig. 1, gelb umrahmtes 

Textfeld, links oben). Kann das Grundstück aufgrund der Schutzmassnahmen einer schwächeren 

Intensitätsklasse zugeordnet werden, erfährt dieses einen Nutzen bezüglich Sicherheit. 

Je grösser die Reduktion der Intensität, desto höher wird der Nutzen und damit die Punktzahl. 

Der Wertebereich der Gesamtpunktzahl und die Zuteilung zu einem Faktor sind analog 

dem raumplanerischen Nutzen (vgl. Fig. 1, gelb umrahmtes Feld, rechts). Der Faktor für das 

Gebäude wird mit dem Gebäudewert (2. Schritt) zur TeilerzahlGebäude, Sicherheit multipliziert, 

der Faktor für den Boden mit dem Bodenwert zur TeilerzahlBoden, Sicherheit. 

Im 5. Schritt werden die aus dem 3. und 4. Schritt resultierenden Teilerzahlen zur Gesamtteilerzahl 

pro Grundstück addiert (vgl. Fig. 1, schwarz umrahmtes Textfeld). Diese wird mit 

dem Perimeterentscheid erlassen und nach Erledigung allfälliger Einsprachen rechtskräftig. 

Der effektiv geschuldete Beitrag in CHF lässt sich erst anhand der Bauabrechnung ermitteln. 

Die verwendeten Tabellen wurden im Rahmen des Perimeterverfahrens erarbeitet. Auf der 

Basis einer vereinfachten Risikoanalyse über das ganze Projektgebiet und unter Einbezug der 

gefahrentechnischen Grundlagen wurden plausible Punktzahlen für die verschiedenen 

Ausprägungen von Nutzen ermittelt (vgl. Fig. 1, grün umrahmtes Textfeld, Punktetabelle). 

Grundstücke, die aus der roten Gefahrenzone entlassen werden, haben den grössten 

raumplanerischen Nutzen und erhalten somit eine höhere Punktzahl. Beim Sicherheitsnutzen 

haben diejenigen Grundstücke mit der grössten Reduktion die höchste Punktzahl. Dabei 

erhält eine Reduktion der Intensität bei kürzeren Wiederkehrperioden eine höhere Punktzahl 

als bei längeren Wiederkehrperioden (vgl. Fig. 1, gelb umrahmtes Textfeld, Punktetabelle). 

Der Faktor wurde zunächst linear dem Wertebereich der Punktetabelle zugeordnet: Faktor 0 

für 0 Punkte, Faktor 1 für 24 Punkte. Aus Gründen der Verhältnismässigkeit wurde der 

Faktor bei den kleinsten und grössten Punktwerten angepasst. Die Festlegung der reduzierten 


Gewichtung des raumplanerischen Nutzens auf 50% erfolgte nach denselben Grundsätzen. 

Alle verwendeten Tabellen wurden in einem iterativen Verfahren mit zahlreichen Plausibilitätschecks 

bereinigt, immer gestützt auf das Zumutbarkeits- und Gerechtigkeitsprinzip. 

RESULTATE 

Für den Kostenteiler Laui Sörenberg resultierte eine Gesamtteilerzahl aller Grundstücke von 

202‘129‘908 Teilern. Die Teilerzahl ist für jedes Grundstück absolut, d.h. falls es für ein Grundstück 

eine Anpassung der Teilerzahl gibt, hat dies keine Auswirkungen auf die Teilerzahlen 

der anderen Grundstücke zur Folge. Im Gegensatz dazu sind der prozentuale Anteil und der 

effektive Betrag in Franken eine relative Grösse. Anhand des prozentualen Anteils an der 

Gesamtteilerzahl aller Grundstücke und der projektierten Kosten kann sich ein Beitragspflichtiger 

beim Erlass des Perimeters über den zu bezahlenden Beitrag informieren. 

Prozentualer Anteil und Betrag in Franken sind aber bis zur definitiven Baukostenabrechnung 

nur orientierend. Im Perimeterverfahren wird der Kostenteiler pro Grundstück rechtskräftig 

veranlagt. Die definitive Festlegung des Frankenbetrages erfolgt in einem zweiten 

Verfahren – ebenfalls unter Gewährung des rechtlichen Gehörs. 

Fig. 2: Klassierung der Gesamtteilerzahlen pro Grundstück mit Farbabstufungen für den Restkostenteiler Laui Sörenberg. 


Für die Schutzmassnahmen Laui Sörenberg bezahlt der überwiegende Anteil der Beitragspflichtigen 

deutlich weniger als 1% der zu verteilenden Kosten. Der mit Abstand grösste 

Beitrag liegt bei 6.53%. Angesichts der projektierten Baukosten von 18 Millionen Franken 

respektive zu verteilenden Kosten von 2.34 Millionen Franken sind Beiträge zwischen 

CHF 100.- und CHF 150‘000.- zu erwarten (Fig. 2). 

Ändern sich bei einem beitragspflichtigen Grundstück die für die Beitragsberechnung 

massgebenden Verhältnisse wesentlich (z.B. durch Neubauten oder bauliche Veränderungen, 

Änderungen der Nutzungsvorschriften, etc.), so ist eine nachträgliche Beitragspflicht nach 

derselben Methode zu bezahlen. Ein solcher Beitrag vermindert sich für jedes Jahr, das seit 

Fertigstellung der Massnahmen vergangen ist, um linear fünf Prozent (PV §12a, Ziff. 3). 

Der nachträgliche Beitrag kann auf die bisher beitragspflichtigen Grundeigentümer verteilt 

oder für den künftigen Unterhalt der Schutzbauten verwendet werden. 

Gegen den Perimeterentscheid des Gemeinderats Flühli gingen 38 Einsprachen ein. Die letzte 

wurde 2015 durch das Bundesgericht abgewiesen. Sowohl das Urteil des Kantonsgerichts 

Luzern (23. Juni 2014, 7H 13 153) wie auch der Bundesgerichtsentscheid (29. Mai 2015, 

BGE 2C_672/2014) fielen zugunsten des Gemeinderats und damit auch der gewählten 

Methode zur Herleitung des Kostenteilers aus. 

SCHLUSSFOLGERUNGEN 

Die Methode hat sich in Sörenberg sehr gut bewährt. Der gewählte Weg im Spannungsfeld 

der verschiedenen Anforderungen ist zielführend: der Kostenteiler wird durch die Beitragspflichtigen 

– mit vergleichsweise wenigen Ausnahmen – verstanden und akzeptiert, also als 

rechtsgleich und verhältnismässig empfunden. Gleichzeitig hielt die vorgestellte Methode 

allen Einsprachen stand und wurde von den angerufenen Gerichten (Kantonsgericht und 

Bundesgericht) vollumfänglich bestätigt. Hierbei gilt es anzumerken, dass insbesondere das 

Kantonsgericht die Methode detailliert auf ihre Rechtsstaatlichkeit geprüft hat. 

Aus Sicht der Autoren sind die nachfolgenden Erfolgsfaktoren für ein erfolgreiches Perimeterverfahren 

massgeblich: 

Fakten statt Erwägungen: Die Berechnung des Kostenteilers muss zwingend auf amtlichen 

und naturwissenschaftlichen Fakten basieren. Im Perimeterverfahren darf es keinen 

Ermessens- respektive Verhandlungsspielraum geben. 

Rechtsstaatlichkeit und Verhältnismässigkeit: Bei der Entwicklung der Methode ist dauernd 

ein detaillierter Abgleich mit den gesetzlichen Grundlagen zu machen. Im Kanton Luzern ist 

dies die Perimeterverordnung. Denn vor Gericht wird in erster Linie deren Einhaltung 

geprüft. 


Transparenz: Die Beitragspflichtigen müssen ihren Beitrag nachvollziehen können. Hierfür 

wurden Faktenblätter pro Grundstück mit der Berechnung der Gesamtteilerzahl erarbeitet 

und im direkten Gespräch erläutert. 

Politische Führung: Die Gemeinde Flühli als Projektträgerin hat die politische Führung zu 

jeder Zeit vorbildlich wahrgenommen. Ein Gemeinderatsmitglied wurde hierzu während der 

intensivsten Phase mit einem 60%-Pensum mandatiert. 

Keine Vermischung von Perimeterverfahren und Einschränkungen aus Projekt: Der Kostenteiler 

soll ohne Einbezug allfälliger Einschränkungen aus dem Projekt (z.B. Landverlust) 

berechnet werden. Entschädigungen für derartige Einschränkungen sind in einem separaten 

Verfahren auszuhandeln. 

SCHLUSSBEMERKUNGEN 

Mit der Beteiligung der interessierten Grundstückbesitzer an den Kosten wird das Schutzprojekt 

für diese fassbar und sensibilisiert für weitere Aufgaben wie Unterhalt und Notfallplanung. 

In Sörenberg hat dieser transparente Einbezug die Solidarität unter den Betroffenen 

gefördert. Es ist unerlässlich, bei einer allfälligen Übertragung der Methode auf andere 

Objekte deren Eignung fallweise kritisch zu überprüfen. Insbesondere die Punktetabellen und 

die Faktoren müssen sich in erster Linie an den Rechtsgrundsätzen der Rechtsgleichheit und 

der Verhältnismässigkeit orientieren. 

LITERATUR 

- Einwohnergemeinde Flühli (2011): Schutzbauten Wasserbauprojekt Laui Sörenberg, 

Perimeter, Entscheid des Gemeinderates Flühli zur Festsetzung der Beitragspflicht der 

interessierten Grundeigentümer (Perimeterentscheid), 38 S. 

- Heinimann H.R. et al. (1999): Risikoanalyse bei gravitativen Naturgefahren, Umwelt-Materialien 

Nr. 107, Teil 1, Bundesamt für Umwelt, Wald und Landschaft BUWAL, 117 S. 

- Oeko-B AG, GEOTEST AG, J. Auchli AG, geo7 AG (2003): Integralprojekt Laui Sörenberg, 

Genehmigungsprojekt Nr. 431.1-LU-0000/0002, Technischer Bericht mit Beilagen. 

- Bundesgericht, II. öffentlich-rechtliche Abteilung (2015): Bundesgerichtsentscheid BGer 

2C_672/2014 vom 29.05.2015. 

- Kantonsgericht Luzern, 4. Abteilung (2014): Urteil des Kantonsgericht Luzern 7H 13 153 

vom 23.06.2015. 

- Regierungsrat des Kantons Luzern (2006): Regierungsratsentscheid zum Wasserbauprojekt 

Laui Sörenberg, Protokoll Nr. 248 vom 17.02.2006, 20 S. 

- Planungs- und Baugesetz des Kantons Luzern (PBG) vom 07.03.1989 (SRL 735). 

- Verordnung über Grundeigentümer-Beiträge an öffentliche Werke des Kantons Luzern 

(Perimeterverordnung, PV) vom 16.10.1969, Stand 01.01.2014 (SRL 732). 

- Wasserbaugesetz des Kantons Luzern (WBG) vom 30.01.1979 (SRL 760). 



What makes a successful flood control project? - An 

evaluation of project procedure and risk based on the 

perspectives of Swiss communes. 

Was macht Hochwasserschutzprojekte erfolgreich? 

Eine Evaluation von Projektablauf und Risiko basierend 

auf den Perspektiven Schweizer Gemeinden. 

Hannes Suter, MSc 1,3 ; Luzius Thomi, Dr. 2 ; Raoul Kern, MSc 2 ; Matthias Künzler, MSc 2 ; Conny Gusterer, BSc 3 ; 

Andreas Zischg, Dr. 1,3 ; Rolf Weingartner, Prof. Dr. 1,3 ; Olivia Martius, Prof. Dr. 1,3 ; Margreth Keiler, PD Dr. 3 

ABSTRACT 

In this study, we evaluate the role of Swiss communes in the process of planning and 

implementing flood control projects. We analyzed 71 Swiss flood control projects by 

evaluating technical reports, online-surveys and interviews with communal project managers. 

Flood control measures are mostly being planned in response to flood events. Generally, 

the planning of the structural flood control measures is not linked with organizational or 

spatial planning measures. The assessment indicates that flood control measures lead to a 

risk reduction in the protected perimeter in the short term. However, the risk evolution in 

the long term is uncertain. Due to current socio-economic and climatic developments in 

Switzerland flood risk is likely to increase in the future. To be able to move from an eventbased 

to a risk-based strategy, an appropriate basis such as a risk monitoring is required. 


Die präsentierte Studie untersucht die Rolle Schweizer Gemeinden in der Initiierung, Planung 

und Umsetzung von Hochwasserschutzprojekten. Dazu wurden 71 Hochwasserschutzprojekte 

durch eine Auswertung der technischen Projektberichte, einer Onlineumfrage und Interviews 

mit kommunalen Projektverantwortlichen evaluiert. Hochwasserschutzprojekte werden 

mehrheitlich nach Überschwemmungsereignissen initiiert und umgesetzt. Eine systematische 

Koordination der zentralen wasserbaulichen Massnahmen mit organisatorischen oder 

raumplanerischen Massnahmen zur Risikominimierung findet nicht grundsätzlich statt. 

Kurzfristig reduzieren die untersuchten Massnahmen das Risiko nachweislich. Die zukünftige 

Risikoentwicklung ist ungewiss. Das Risiko dürfte sich aufgrund der sozio-ökonomischen und 

klimatischen Entwicklung langfristig jedoch trotz Schutzmassnahmen erhöhen. 

1 University of Bern, Oeschger Center for Climate Change Research, Mobiliar Lab for Natural Risks, Bern, SWITZERLAND, 

hannes.suter@gmail.com 

2 Swiss Mobiliar Insurance Company, Bern, SWITZERLAND 

3 University of Bern, Institute of Geography, Bern, SWITZERLAND 



Um die Notwendigkeit und die Auswirkungen von Hochwasserschutzprojekten abzuschätzen, 

braucht es ein Risikomonitoring. 

KEYWORDS 

Flood control; risk; project procedure; Switzerland 

EINFÜHRUNG 

Seit dem Hochwasserereignis im August 2005, das in der Schweiz Schäden im Gesamtwert 

von rund 3 Mrd. CHF verursacht hatte (Bezzola & Hegg, 2007), unterstützte die Schweizerische 

Mobiliar Versicherungsgesellschaft gut 80 Präventionsprojekte zum Schutz vor Naturgefahren. 

Das Mobiliar Lab für Naturrisiken der Universität Bern hat 71 dieser Projekte 

evaluiert und die Resultate im Bericht „Was macht Hochwasserschutzprojekte erfolgreich“ 

(Thomi et al., 2015) veröffentlicht. Dieser Artikel greift spezifische Teilaspekte dieser 

Forschungsresultate heraus, die im breiteren Kontext des gesellschaftlichen Umgangs mit 

Hochwasserrisiken stehen. Die folgenden Fragen stehen im Vordergrund: 

1. Wie wird die Notwendigkeit von Hochwasserschutzprojekten erkannt und wer sind die 

wichtigsten Akteure für die Erkennung der Notwendigkeit? 

2. Wer legt die Schutzziele fest und wie werden diese definiert? 

3. Welche Kombination von Massnahmen führt zu einer langfristigen Minderung des Risikos? 

4. Wie wird die langfristige Risikoentwicklung im geschützten Perimeter eingeschätzt? 

5. Wie können Risiken frühzeitig erkannt werden? 

METHODE 

Die Evaluation umfasst sechs Phasen, wobei die die ersten vier auf die Gemeindesicht 

fokussieren 

1. Auswertung der technischen Projektberichte anhand von 151 Indikatoren mit Schwerpunkt 

auf die Forschungsfragen 1, 2 und 3 (Anzahl untersuchte Projekte: n=71). 

2. Schriftliche Umfrage unter den Projektverantwortlichen der Gemeinden (n=57) mit 

Schwerpunkt auf die Forschungsfragen 1, 2, 3 und 4. 

3. Halbstandardisierte Experteninterviews mit Projektverantwortlichen der Gemeinden (n=6) 

mit Schwerpunkt auf die Forschungsfragen 1, 2 und 4. 

4. GIS gestützte Auswertung mit Schwerpunkt auf die Forschungsfrage 5. 

In den Phasen 5 und 6 werden die Ergebnisse durch den Einbezug weiterer im Hochwasserschutz 

involvierter Akteure erweitert und bieten die Basis für einen Vergleich im Kapitel 

„Diskussion”: 

5. Halbstandardisierte Experteninterviews mit kantonalen Projektverantwortlichen (n=2) mit 

Schwerpunkt auf die Forschungsfragen 3, 4 und 5. 

6. Workshop mit zentralen Akteuren im Schweizer Hochwasserschutz (Bund, Kanton, 

Gemeinde, Ingenieurbüro, Versicherung, Wissenschaft) (Teilnehmer: 15) mit Schwerpunkt 

auf die Forschungsfragen 1, 2, 3, 4 und 5. 


ERGEBNISSE 

Wer entwickelt Hochwasserschutzprojekte und wie wird deren Notwendigkeit erkannt? 

In 48 von 57 Projekten nannten die befragten Projektverantwortlichen, dass ein Hochwasserereignis 

Auslöser für das Projekt war. Weitere Gründe für Projektlancierungen sind ein 

durch eine Gefahrenkarte aufgezeigtes Schutzdefizit, eine Revitalisierung sowie die Sanierung 

bestehender Wasserbauanlagen oder Infrastrukturprojekte (z. B. Strassenbau; vgl. Abbildung 

1). 

Abbildung 1. Grund für die Lancierung von Hochwasserschutzprojekten (Mehrfachnennungen möglich). 

Auf die Frage, wer den Anstoss zum Hochwasserschutzprojekt gab, nannten die Projektverantwortlichen 

auf Gemeindeebene in der Umfrage am häufigsten die Gemeinde. 

Wer legt die Schutzziele fest und wie werden diese definiert? 

Bei der Ausgestaltung der Massnahmen besteht ein gewisser Spielraum, d.h. die Schutzziele 

können zwischen den betroffenen Akteuren bis zu einem gewissen Mass ausgehandelt 

werden. Allerdings scheint dies in vielen Projekten nicht der Fall gewesen zu sein: In 

mehreren Interviews wird darauf verwiesen, dass die Schutzziele nicht grundsätzlich 

diskutiert wurden. Auch die Umfrage bei den Projektverantwortlichen auf Gemeindeebene 

zeigt, dass nur in wenigen Fällen eine eingehende Schutzzieldiskussion stattfand. In den 

meisten Fällen sind die Empfehlungen von Bund und Kanton diskussionslos übernommen 

worden. 

Welche Kombination von Massnahmen führt zu einer langfristigen Minderung des Risikos? 

Die untersuchten Projekte umfassen in erster Linie wasserbauliche Massnahmen zur 

Reduktion der Hochwassergefährdung. In 67 von 71 Projekten wurden Massnahmen zur 

Verbesserung der Kapazität umgesetzt. Dazu gehören etwa der Gerinneausbau oder der 

Wasserrückhalt. Weiter wurden in den technischen Berichten Massnahmen zum Schutz vor 

Ufer- und Sohlenerosion, Geschieberückhalt und Rückhalt für Schwemmholz genannt. 


In vielen Fällen ergänzen Massnahmen raumplanerischer oder organisatorischer Art sowie 

Objektschutzmassnahmen die baulichen Hochwasserschutzmassnahmen. Zwar wird in 

einigen technischen Berichten auf diese ergänzenden Massnahmen verwiesen, eine inhaltliche 

Verzahnung – oder gar ein Gesamtkonzept – von baulichen und zusätzlichen Massnahmen 

konnte jedoch nur selten festgestellt werden. 

In 54 % aller untersuchten Projektunterlagen wurde der Überlastfall (Hochwasser übersteigt 

die Dimensionierungsgrösse der Schutzmassnahmen) erwähnt, wenn auch oft eher knapp. 

In 46 % der Projekte stehen keine Informationen zum Überlastfall zur Verfügung. 

Einschätzung der Risikoentwicklung 

Kurzfristig senken die umgesetzten Hochwasserschutzmassnahmen das Hochwasserrisiko. 

In 18 Fällen fand laut Umfrage nach der Umsetzung der Massnahmen ein weiteres Hochwasserereignis 

statt; bei keinem dieser Ereignisse sind ausserhalb des Gerinnes Schäden aufgetreten. 

Die Gemeinden setzten sich jedoch mehrheitlich nicht mit den langfristigen Auswirkungen 

von Hochwasserschutzprojekten und der langfristigen Risikoentwicklung auseinander. 

Dies zeigt sich auch darin, dass in keinem der Projekte ein systematisches Monitoring der 

Risikoentwicklung nach Umsetzung der Massnahmen stattfindet. Zudem nimmt die Bautätigkeit 

in den durch das Projekt geschützten Zonen nach Fertigstellung der Massnahmen in 

einem Drittel aller untersuchten Fälle zu, insbesondere dann, wenn Projekte aufgrund eines 

erkannten Schutzdefizits aus der Gefahrenkarte lanciert wurden. In drei Fällen sind nach der 

Fertigstellung der Hochwasserprojekte neue Einzonungen im geschützten Perimeter geplant. 

Diese Tendenzen bezüglich Bautätigkeit und Einzonungen sind insofern bemerkenswert, als 

sie bereits kurze Zeit nach Realisierung der Massnahmen (sämtliche untersuchten Projekte 

sind in den letzten zehn Jahren projektiert bzw. umgesetzt worden) festgestellt werden 

können. 

Abbildung 2 zeigt die Bevölkerungsveränderung (x-Achse) und den Anteil der gefährdeten 

Gebiete an den Bauzonen von 1435 Schweizer Gemeinden (y-Achse). Ein starkes Bevölkerungswachstum 

wird als Indikator für eine zunehmende Bautätigkeit interpretiert. Es wird 

angenommen, dass die Bautätigkeit innerhalb der Bauzonen stattfinden wird. Befinden sich 

diese im gefährdeten Gebiet (rot, blau, gelb) ist mit einem zunehmenden Risiko zu rechnen. 

Gemeinden im oberen rechten Quadranten zeichnen sich durch ein überdurchschnittliches 

Bevölkerungswachstum aus und gleichzeitig sind deren Bauzonen zu einem erheblichen Teil 

Wassergefahren ausgesetzt. Das Risiko, definiert nach BUWAL (1998) als „Funktion der 

Wahrscheinlichkeit eines Schadensereignisses und des möglichen Schadensausmasses“, wird 

in diesen Gemeinden mittelfristig vermutlich zunehmen, sofern die bedrohten Objekte nicht 

gegenüber Überschwemmungen geschützt werden bzw. die fraglichen Objekte eine geringe 

Verletzlichkeit gegenüber Überschwemmungen aufweisen. 


DISKUSSION 

Wer entwickelt Hochwasserschutzprojekte und wie wird deren Notwendigkeit erkannt? 

Abbildung 2. Gemeinden (n=1435) in Abhängigkeit der Bevölkerungsveränderung 2013 bis 2017 (berechnet durch die Firma bwv its 

GmbH auf der Basis von Daten des Bundesamts für Statistik) und des Anteils der Gefahrenzonen gelb, blau und rot an der Bauzone 

(IKGEO, 2015). Berücksichtigt sind nur Gemeinden mit einer Bauzone > 1 ha und einer Abdeckung der Bauzonen durch eine 

Gefahrenkarte von mindestens 50 %. Dunkelgraue Dreiecke: Gemeinden mit einem untersuchten Hochwasserschutzprojekt. Graue Linien: 

Median von Bevölkerungsveränderung und Anteil der Gefahrenzonen an der Bauzone. 

In den untersuchten Projekten sind die Gemeinden meist wasserbaupflichtig und nehmen 

sich deshalb als zentrale Akteure wahr. Allerdings stossen sie aufgrund der hohen Komplexität 

von Hochwasserschutzprojekten häufig an ihre Grenzen. In der Projekterarbeitung kommt 

deshalb den Kantonen oft die Schlüsselrolle zu, auch wenn sie das Projekt formell nicht 

führen (Thomi, 2005). Die starke Einflussnahme der Kantone – und indirekt des Bundes – 

wird von den kommunalen Behörden teilweise als einengend wahrgenommen. 

Hochwasserereignisse sind der wichtigste Auslöser für Schutzprojekte. Trotz der grossen 

Anstrengungen der letzten zwei Jahrzehnte auf strategischer und konzeptueller Ebene zum 

präventiven Umgang mit Naturgefahren (z. B. integrales Risikomanagement, Risikokultur; 

vgl. z. B. PLANAT, 2004) sowie der Schaffung von Grundlagen zur Gefahrenerkennung (z. B. 


Gefahrenkartierung), sind in vielen Fällen Hochwasserereignisse nötig damit Massnahmen 

umgesetzt werden. Überschwemmungen machen Risiken greifbar und bauen so den nötigen 

politischen Druck zur Umsetzung auf, wie dies auch Zaugg (2006), Thomi (2010) oder 

Scheuchzer et al. (2012) beschrieben haben. 

Wer definiert Schutzziele und wie werden diese definiert? 

Eine Schutzzieldiskussion kann die Sensibilisierung für Naturgefahren und somit das 

Risikobewusstsein stärken (ARE et al., 2005). Schutzziele werden aber auf Gemeindeebene 

wenig diskutiert, sondern vom Kanton bzw. Bund übernommen. Diese Aussage trifft auf die 

Mehrheit der 71 untersuchten Projekte zu. Demgegenüber wiesen Fachspezialisten am 

Workshop darauf hin, dass in gewissen Hochwasserschutzprojekten Schutzziele durchaus 

intensiv zwischen Bund, Kanton und Gemeinde diskutiert würden. Woher rührt diese 

Diskrepanz? Möglicherweise verstehen nicht alle Akteure dasselbe unter den Begriffen 

„Schutzziel“ und „Schutzzieldiskussion“, die gerade für Nicht-Fachspezialisten sehr abstrakt 

erscheinen. Hess (2008: 158) bemerkte in einer Analyse eines Fallbeispiels, dass „die 

Schutzzieldiskussion die meisten Teilnehmenden überfordert“. Zudem wurde in der Umfrage 

gefragt, wer sich an der Schutzzieldiskussion beteiligte und nicht wie stark diese Beteiligung 

war. Denkbar ist auch, dass sich die Gemeinden bei der Frage nach den Schutzzielen stark auf 

das mandatierte Planungsbüro stützen und dessen Vorschläge nicht hinterfragen. 

Welche Kombination von Massnahmen führt zu einer langfristigen Minderung des Risikos? 

Wasserbauliche Massnahmen werden oft losgelöst von organisatorischen und raumplanerischen 

umgesetzt, im Fokus steht klar der bauliche Hochwasserschutz am Gewässer. In der 

Umfrage unter den kommunalen Projektverantwortlichen, welche in vielen Fällen nach 

Abschluss der Massnahmen stattfand, werden tendenziell mehr organisatorische und 

raumplanerische Massnahmen erwähnt als in den Projektunterlagen. Möglicherweise findet 

während des Projekts eine gewisse Sensibilisierung der Gemeindeverantwortlichen betreffend 

weiteren Massnahmen statt. 

Das Bundesamt für Umwelt (BAFU) prüft anhand bestimmter Kriterien die Umsetzung des 

integralen Risikomanagements und fördert dieses mit zusätzlichen 6 % Bundesbeiträgen 

(BAFU, 2011, 2015b). Die zusätzlichen Bundesmittel werden allerdings nur bei Einzelprojekten 

direkt vom BAFU gesprochen. In den durch das sogenannte Grundangebot abgegoltenen 

Projekten (dazu gehören in der Regel Projekte mit Gesamtkosten von max. 1 Mio. CHF bis 

2011 bzw. max. 5 Mio. CHF seit 2012; vgl. BAFU, 2011, 2015b) sind die Kriterien der 

Kantone massgebend. Ob und in welcher Form die Kantone effektiv Anreize für ein 

integrales Risikomanagement schufen, konnte im Rahmen der vorliegenden Untersuchung 

nicht geklärt werden. Offensichtlich ist, dass nicht alle Kantone solche Anreize kennen. 

Einschätzung der Risikoentwicklung 

Aus einer Risikoperspektive wird die durch das Projekt gewonnene Risikoreduktion durch 

das anwachsende Schadenpotenzial in vielen Fällen langfristig wieder zunichte gemacht (vgl. 


Keiler et al. 2006). Handlungsbedarf entsteht vor allem beim Überlastfall und in der gelben 

Gefahrenzone. In beiden Fällen rechnen betroffene Akteure vermutlich kaum mit einer Überschwemmung, 

was zu einer unangepassten Landnutzung und somit zu einer erhöhten 

Vulnerabilität führen kann, wie dies bereits Tobin (1995) und Zaugg (2006) festgestellt haben. 

Lösungsansätze bestehen in einer Reduktion der Verletzlichkeit, z. B. durch Objektschutz 

(Egli, 2002), oder der Verminderung des Anstiegs der Sachwerte innerhalb des gefährdeten 

Perimeters. Zentral für das Risikomanagement des Überlastfalls ist gemäss Petrasckeck et al. 

(2002) eine Notfallplanung. Ein weiterer wichtiger Lösungsansatz stellt die risikobasierte 

Raumplanung dar. Der Handlungsbedarf sollte dabei nicht nur gefahrenbasiert geprüft 

werden, sondern es sollen stärker die Raumnutzungen und das Schadenpotenzial berücksichtigt 

werden (Camenzind & Loat, 2014). 

Grundsätzlich muss jedoch beachtet werden, dass eine Abschätzung der langfristigen 

Risikoentwicklung aufgrund der kurzen Zeitdauer zwischen Projektrealisierung und der 

vorliegenden Untersuchung nur bedingt möglich ist. 

Frühzeitige Erkennung von Risiken 

Für die frühzeitige Erkennung von Risiken stehen verschiedene Grundlagen zur Verfügung, 

wie z. B. Gefahren- und vereinzelt Risiko- oder Schutzdefizitkarten (BAFU, 2015a). Diese 

stellen jedoch nur eine Momentaufnahme dar und lassen zukünftige Veränderungen der 

Gefährdung (z. B. durch den Klimawandel), der Raumnutzung, der Wertekonzentrationen, 

der Verletzlichkeit oder der Bautätigkeit im bedrohten Perimeter in der Regel ausser Acht. 

Ein vorausschauendes und systematisches Überwachen der risikorelevanten Parameter 

(Risikomonitoring) findet heute kaum statt, wäre aber für einen nachhaltigen Umgang mit 

Naturrisiken dringend notwendig (Petrascheck et al., 2002). Das Monitoring müsste die 

risikorelevanten Parameter (z. B. Bevölkerungswachstum, Bautätigkeit, gefahrenseitige 

Prozessveränderungen) berücksichtigen und zeitlich-räumliche Veränderungen abbilden. 

Um punkto Risikoentwicklung stark gefährdete Gemeinden frühzeitig zu eruieren, könnte 

das in der Methodenphase 4 beschriebene Vorgehen einen ersten Ansatz bieten. 

FAZIT 

In den untersuchten Projekten dominiert das Handeln als Reaktion auf Überschwemmungen. 

Ein risikobasiertes und proaktives Agieren würde hingegen die Prävention stärken und 

Schäden frühzeitig verhindern. Gefahrenkarten alleine sind kein ausreichendes Instrument, 

um Risiken langfristig zu senken. Mögliche Ansätze sind eine risikobasierte Raumplanung 

(vgl. Camenzind & Loat, 2014) und die Priorisierung von Massnahmen basierend auf einer 

Risikoübersicht. Dieses Wissen zum risikobasierten Umgang mit Naturrisiken ist seit längerer 

Zeit vorhanden (Egli, 1996). Möglicherweise werden diese Ansätze zur Risikoreduktion aber 

schlichtweg vom raschen Anstieg der potentiellen Schäden im gefährdeten Gebiet überholt 

(White et al, 2001). 

Um eine optimale Risikoreduktion zu erzielen und Fehlinvestitionen zu verhindern, ist eine 

frühe Koordination aller Massnahmen auf lokaler (z. B. Gemeinde), aber vor allem auch auf 


egionaler Ebene (z. B. Einzugsgebiet) unabdingbar (vgl. Scheuchzer et al., 2012). Hier 

besteht Bedarf für ein Instrument, das erlaubt, die verschiedenen baulichen, organisatorischen 

und raumplanerischen Massnahmen besser aufeinander abzustimmen. Dabei ist 

wichtig, dass die Massnahmen zeitlich, räumlich sowie in Bezug auf die Gefahr (Intensität, 

Frequenz) koordiniert werden und sich gegenseitig ergänzen. 

Schliesslich sind es die betroffenen Personen, welche die zukünftige Entwicklung ihrer 

Gemeinde bestimmen. Durch einen partizipativen Einbezug dieser Akteure in die strategische 

Planung (Hostmann et al., 2005, KOHS, 2004) in das Hochwasserschutzprojekt kann das 

Risikobewusstsein einer ganzen Region nachhaltig positiv beeinflusst werden. 

LITERATUR 

- Bezzola G.R., Hegg C. (Hrsg.) (2007). Ereignisanalyse Hochwasser 2005, Teil 1 – Prozesse, 

Schäden und erste Einordnung. Bundesamt für Umwelt, Eidgenössische Forschungsanstalt 

WSL. Umwelt-Wissen Nr. 0707. 215 S. 

- Bundesamt für Umwelt BAFU (Hrsg.) (2006). Realisierungsgrad pro Kanton, alle 

Gefahrenarten, 2006. Show Me. Bern, Bundesamt für Umwelt (BAFU). 

- Bundesamt für Umwelt BAFU (Hrsg.) (2011). Handbuch Programmvereinbarungen im 

Umweltbereich. Mitteilung des BAFU als Vollzugsbehörde an Gesuchsteller. 

Bundesamt für Umwelt, Bern. Umwelt-Vollzug Nr. 1105. 

- Bundesamt für Umwelt BAFU (Hrsg.) (2015a). Realisierungsgrad pro Kanton, alle 

Gefahrenarten, 2015. Show Me. Bern, Bundesamt für Umwelt (BAFU). 

- Bundesamt für Umwelt BAFU (Hrsg.) (2015b). Handbuch Programmvereinbarungen im 

Umweltbereich 2016–2019. Mitteilung des BAFU als Vollzugsbehörde an Gesuchsteller. 

Bundesamt für Umwelt, Bern. Umwelt-Vollzug Nr. 1501: 266 S. 

- Bundesamt für Umwelt, Wald und Landschaft BUWAL (Hrsg.) (1998). Begriffsdefinitionen 

zu den Themen: Geomorphologie, Naturgefahren, Forstwesen, Sicherheit, Risiko. Bundesamt 

für Umwelt, Wald und Landschaft, Bern. 

- Bundesamt für Raumentwicklung ARE; Bundesamt für Wasser und Geologie BWG; 

Bundesamt für Umwelt, Wald und Landschaft BUWAL (Hrsg.) (2005). Empfehlung – Raumplanung 

und Naturgefahren. Bern. 

- Camenzind R., Loat R. (2014). Risikobasierte Raumplanung – Synthesebericht zu zwei 

Testplanungen auf Stufe kommunaler Nutzungsplanung. Nationale Plattform Naturgefahren / 

Bundesamt für Raumentwicklung / Bundesamt für Umwelt, Bern. 21 S. 

- Egli T. (1996). Hochwasserschutz und Raumplanung. Schutz vor Naturgefahren mit 

Instrumenten der Raumplanung — dargestellt am Beispiel von Hochwasser und Murgängen. 

ETH Zürich. 

- Egli T. (2002). Hochwasserschutz durch nachhaltiges Schadenpotenzialmanagement. 

Internationales Symposium 2002 in Zürich: „Moderne Methoden und Konzepte im Wasserbau“. 

Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie der ETH Zürich und dem 

Schweizerischen Wasserwirtschaftsbund. 

- Hess J.T. (2008). Schutzziele im Umgang mit Naturrisiken in der Schweiz. ETH, Zürich. 


- Hostmann M., Buchecker M., Ejderyan O., Geiser U., Junker B., Schweizer S., Truffer B., 

Zaugg Stern M. (2005). Wasserbauprojekte gemeinsam planen. Handbuch für die Partizipation 

und Entscheidungsfindung bei Wasserbauprojekten. Eawag, WSL, LCH-EPFL, 

VAW-ETHZ. 48 pp. 

- Interkantonale Koordination in der Geoinformation IKGEO (2015). „Bauzonen Schweiz 

(harmonisiert)“, Bundesamt für Raumentwicklung ARE, Stand 1.1.2012). http://www.ikgeo. 

ch/dokumentation/geodaten-bauzonen-schweiz.html (Zugriff: 01.07.2015). 

- Keiler M., Sailer R., Jörg P., Weber, C., Fuchs S., Zischg A., Sauermoser S. (2006). 

Avalanche risk assessment – a multi-temporal approach, results from Galtür, Austria. 

Natural Hazards and Earth System Sciences 6 (4), 637-651. 

- Kommission für Hochwasserschutz KOHS (2004). Qualitätssicherung bei der Planung von 

Hochwasserschutzmassnahmen. Leitfaden für Auftraggeber und -nehmer von Hochwasserschutzprojekten. 

Schweizerischer Wasserwirtschaftsverband, Baden. 

- Petrascheck, A.; Hegg, C.; Schmid, F. (2002): Zusammenfassung und Schlussfolgerungen. 

WSL, Eidgenössische Forschungsanstalt für Wald Schnee und Landschaft & BWG, Bundesamt 

für Wasser und Geologie: Hochwasser 2000. Ereignisanalyse/Fallbeispiele. Bern: Bundesamt 

für Wasser und Geologie. S. 7-14. 

- PLANAT (Hrsg.) (2004). Sicherheit vor Naturgefahren. Vision und Strategie. PLANAT Reihe 

1/2004. Biel, Nationale Plattform Naturgefahren (PLANAT). 

- Scheuchzer P., Walter F., Truffer B., Balsiger J., Chaix O., Kempter T., Klinke A., Menzel S., 

Wehse H., Zysset A. (2012): Auf dem Weg zu einer integrierten Wasserwirtschaft. Synthese 

zum Projekt IWAGO – Integrated Water Governance with Adaptive Capacity in Switzerland 

Projekt im Rahmen des NFP 61 „Nachhaltige Wassernutzung“. 

- Thomi, L. (2005). La gestion de l'aménagement des cours d'eau dans les cantons suisses 

de Glaris, de Berne et du Valais. Geographica Helvetica, 60/1, 35-43. 

- Thomi L. (2010). Rôle des paramètres sociopolitiques et des connaissances dans la gestion 

des risques hydrologiques. Université de Lausanne, Lausanne 

- Thomi L., Zischg A., Suter H. (2015). Was macht Hochwasserschutzprojekte erfolgreich? – 

Eine Evaluation der Risikoentwicklung, des Nutzens und der Rolle privater Geldgeber. 

Universität Bern Geographisches Institut, Mobiliar Lab. 

- Tiefbauamt des Kantons Bern (2010). Grundsätze der Kommunikation im Wasserbau. 2-4. 

- Tobin G.A. (1995). The levee love affair: A stormy relationship. Water Resources Bulletin. 

JAWRA, 3. 

- White G.F., Kates R.W., Burton I. (2001). Knowing better and losing even more: the use of 

knowledge. Hazards management Glob. Environ. Change: Part B – Environ. Hazards, 

3, pp. 81–92 

- Zaugg M. (2006). Philosophiewandel im schweizerischen Wasserbau - Zur Vollzugspraxis 

des nachhaltigen Hochwasserschutzes 1. Auflage. Geographisches Institut. Abteilung 

Humangeographie. Universität Zürich, Zürich. 



The Zurich flood resilience alliance: A new approach to 

partnership for effective disaster risk reduction 

Michael Szoenyi, MAS NatHaz, MSc¹; Linda Freiner² 

ABSTRACT 

An introduction to the Zurich flood resilience alliance. This paper provides an overview of the 

global challenges caused by flooding and how we are tackling them through the work of the 

Zurich flood resilience alliance. Risks of floods are increasing because of population growth, 

more people living near water and growing prosperity. There are several ways to enhance 

flood resilience. We believe preventive action results in benefits far in excess of those 

recovery can provide. Flood risks are increasingly interconnected and interdependent. 

A holistic approach is needed to address them. We have created a pioneering collaboration 

through which we can tackle the challenges communities face. We work as an alliance 

bringing together organizations with complementary skills, launched by Zurich Insurance 

Group in 2013. The alliance includes two humanitarian organizations - the Red Cross and 

Practical Action - and two leading research institutions - Wharton and IIASA. The program 

is based on a new approach to cross-sector collaboration. It brings together flood risk 

research, community-based programs and risk expertise. 

KEYWORDS 

flood; resilience; measurement; alliance; partnership 

INTRODUCTION 

Floods affect more people globally than any other natural hazard. They cause some of the 

largest economic, social and humanitarian losses, involving on average some 250 million 

people per year (UNISDR 2013). While floods are natural, their disastrous consequences are 

not. Often the poorest and least-prepared communities suffer most. Evidence shows that 

repeated disasters like floods undermine societies’ and economies’ potential to develop and 

trap them in a poverty cycle. We tend to think of these events as happening in other places to 

other people, but floods also cause devastation in developed countries. The reasons are 

surprisingly similar in both developing and developed countries. To address the need for a 

proactive approach to flood risks, Zurich Insurance Group (Zurich) launched a dedicated 

flood resilience program in 2013. It includes two humanitarian organizations – the International 

Federation of Red Cross and Red Crescent Societies (IFRC), and Practical Action – and 

two leading research institutions: the Wharton School of the University of Pennsylvania 

(Wharton), and the International Institute of Applied Systems Analysis (IIASA) in Austria. 

The program is based on a new approach to cross-sector collaboration. It brings together flood 

1 Zurich Insurance Group, Zurich, SWITZERLAND, michael.szoenyi@zurich.com 

2 Zurich Insurance Company, SWITZERLAND 



isk research, community-based programs and hazard and risk assessment expertise. It looks 

for, and shares ways that community flood resilience can be measured and improved. 

We define resilience as the ability to continue to thrive in the face of disasters. The program 

directly helps about 125,000 people through projects in flood-prone communities in 

Bangladesh, Indonesia, Mexico, Nepal, Peru and the U.S. 

Risks of floods are increasing. By some estimates, river flooding alone (not counting other 

floods such as surface water, ocean flooding, storm surge etc.) could annually affect 

54 million people worldwide by 2030, more than double the number currently affected 

(Scientific American, 2015). There are several reasons why floods are having a greater 

impact: 

Growing populations, more people living in cities. The world’s urban population increased 

fivefold from 700 million in 1950 to 3.9 billion in 2014 (United Nations, 2014). Urban growth 

is particularly strong in developing countries, where cities’ disaster plans and emergency 

facilities are often unable to cope with major floods. Constructing buildings on flood plains, 

paving over land that provided drainage and lack of waste infrastructure all add to the risks. 

More people living near water: Where land is at a premium, developers may ignore warnings 

and build in places exposed to river floods and storm surges. Cities in coastal regions on or 

near rivers such as Jakarta, Lima, Shanghai, Dhaka, and Manila are highly vulnerable. But 

floods can also shut down a metropolis like New York City, as Superstorm Sandy showed in 

2012. 

Greater prosperity – more to lose: Development of countries depending on agriculture is more 

vulnerable to drought, while countries reliant on industrial development for growth suffer 

more from floods (Collier et al, 2013). A higher standard of living that brings with it more 

manufacturing and production increases the value of property at risk. Such risk might be 

alleviated by protection measures (for example, raising buildings). Even so, a catastrophic 

flood can threaten not only the economy of a region, but entire global supply chains: severe 

flooding in manufacturing sites in Thailand in 2011 led to global shortages of components 

needed by major car makers and severely hit electronics production. 

Climate change: Flood risk could continue to increase significantly in many parts of the world 

due to expected changes in climate patterns. Warmer temperatures affect weather patterns 

and sea levels. Tropical cyclones common in the northwest Pacific included Typhoon Haiyan 

in 2013, perhaps the strongest tropical cyclone in recorded history to make landfall, which 

killed over 7,000 people. 

METHODS 

We have thus taken the following approach to increasing flood resilience: 

Understanding flood resilience 

It is hard to change habits or convince people to move out of harm’s way. Improving resilience 

is doubly important because it helps people to anticipate and cope with floods. It not only 

allows them to reduce the flood exposure to lives and property. Resilience also helps them 


ecover more quickly. It keeps people’s lives intact before, during and after floods. It helps 

communities become more prosperous and stable. Resilience is frequently described as a 

system or even a system of systems, one that is holistic in nature. ”A system-wide approach 

to resilience needs to capture a range of activities, actors and processes that are part of a 

resilience building system,” according to a UN study (Winderl, 2014). 

Ways resilience can be increased 

There are several different ways we can enhance flood resilience: through better assessment 

of flood hazards and communicating the risk to residents; taking measures to lessen the 

severity of floods and mitigate their impact, including first-aid and health training, community 

planning, and setting up emergency shelters; gaining a better understanding of how 

decisions are made in the face of risks and uncertainty to make the most effective solutions 

easier to find; improving warning systems and helping communities adopt emergency 

protocols; supporting efforts to rebuild better, to safer standards, after floods; ensuring people 

have opportunities to secure an income during floods, for example, by providing skills 

training so farmers have alternative revenues when cropland is under water; developing ways 

to safeguard assets exposed to flood risk at an individual or community level; and working 

with local officials and other policymakers and the private sector to help make communities 

more flood resilient. 

Increasing resilience makes economic sense 

We believe proactive action, reducing flood risk before an event, brings benefits far in excess 

of those recovery can provide. On average, for every dollar spent on targeted flood-risk 

reduction measures, five dollars can be saved by avoiding and reducing losses (Mechler et al, 

2014). Despite the advantages of acting before floods happen to improve resilience, over the 

past two decades, only 13 percent of aid went to reducing and eliminating risks. The 

remaining 87 percent were used for emergency response, reconstruction and post-disaster 

rehabilitation (Kellett and Caravani 2013). This emphasis on relief as opposed to resilience is 

neither logical nor efficient. Psychology plays a major role in flood resilience. By better 

understanding how people think, we can address the reasons why, despite its high cost-effectiveness, 

some communities and even international donors do not invest enough in pre-flood 

mitigation. One common misconception is the ‘it will never happen to us’ syndrome: 

decision-makers underestimate flood risk, preferring to see floods as an unlikely event. There 

is also the ‘there is nothing we can do, anyway’ syndrome – people become fatalistic when 

they feel powerless to control the outcome of events. People also procrastinate: even when 

they know that investing in flood protection is necessary, they avoid making decisions. They 

might fall into the trap of refusing to invest in flood protection because they assume the 

government or donors will step in. There is also a ‘gambler’ mentality: people believe that 

because a flood recently occurred, there won’t be another one any time soon, forgetting that 

disasters occur independent of one another. There are very real budget constraints that must 

be overcome to convince people to take action. 


Creating a flood resilience alliance 

Flood risks are increasingly interconnected and interdependent. Through our pioneering 

collaboration (Figure 1), we can tackle the challenges communities face. This effort works as 

an alliance that brings together organizations with complementary skills and expertise. We 

are seeing the advantages of a combined approach in community programs where the IFRC 

and Practical Action use their extensive experience working with communities to identify and 

implement solutions. Programming is an iterative process typically starting by assessing and 

analyzing the situation; innovative solutions are devised, and then work begins with 

communities to assess, select and implement the best solutions. The impact of the actions 

implemented is then evaluated. Research by Wharton and IIASA confirms the advantages of 

investing in pre-event mitigation as opposed to post-event relief. The research also provides 

objective evidence that can influence policymakers’ decisions. Further, it can create an 

environment in which insurance and other risk transfer mechanisms can be part of the 

solution. As an insurer, Zurich acts as a catalyst in providing human, technical and financial 

resources. The Z Zurich Foundation has made an initial five-year commitment of USD 35.6 

million to the alliance. This is in addition to contributions of time, expertise and resources of 

Zurich employees around the world. 

Figure 1: Overview of the five organizations forming the Zurich flood resilience alliance. 


RESULTS 

How we improve flood resilience 

Buildings can be rebuilt after a flood. It is harder to rebuild many aspects of life, especially 

when dealing with repeated floods. With each flood, social bonds are tested, people lose 

income, family ties strained. Increasing a community’s resilience to floods is the best way to 

counter these destructive forces. We can improve resilience, even when the task is sometimes 

difficult. Below, we show some examples of the specific challenges that we face and how 

we have addressed them. 

We need to invest more in flood resilience 

Images of flood disasters in the media resonate with the public. People respond, and at least 

for a time, give generously. But such aid, while invaluable for the recovery phase, may fail to 

provide long-term solutions. Therefore, we need to direct investments toward increasing 

communities’ flood resilience for the long term, not just when disaster strikes. In Mexico, we 

provide better facilities and alternative livelihoods: Work led by the IFRC and the Mexican 

Red Cross began in 2014 in 11 communities around Jonuta, a municipality in the state of 

Tabasco. Tabasco includes one of the world’s largest wetlands through which the Usumacinta 

River flows. The river can rise as much as 3 meters during the rainy season. The people in 

these communities must live with floods several months out of the year and therefore need 

innovative solutions to earn a living, even during periods of high water. Work is underway to 

build multi-purpose community buildings that can double as school rooms or medical 

facilities if necessary. Work is also being done to train communities on how to catch and 

prepare the invasive devil fish, benefiting not only the environment, but possibly also 

providing them with a sustainable food source and a new source of income. 

In Nepal, we improve early warning systems: Nepal is still recovering from the devastating 

earthquake in April 2015. But even as recovery begins, some communities in Nepal face a 

new threat: the onset of the monsoon season. Monsoon rains often trigger flash floods and 

mudslides, posing significant risks. Timely warnings about imminent floods can save lives and 

help people protect their possessions. Under the lead of Practical Action, we are working in 

the Karnali River basin which begins in the southern slopes of the Himalayas and flows 

through Nepal to India. A major focus is on improving early warning systems implemented in 

2010. In particular, this includes improving weather forecasting, keeping live-saving 

technology working even under extreme conditions, and training communities to act on the 

information received from measuring stations along the Karnali River. 

In Indonesia, we are connecting upstream and downstream communities. Bukit Duri in the 

southern part of Jakarta city is bordered on one side by the Ciliwung River. When it floods, 

garbage and sewage block the river and end up flowing into the community. Much of this 

waste comes from upstream communities such as Tugu Utara. The IFRC and the Red Cross 

society in Indonesia, PMI, are leading our work with both Bukit Duri and Tugu Utara to 

improve waste management practices. By helping remove the waste from the river, the 

impact of periodic flooding is reduced. Not only that, but the waste management process 


also provides a valuable chance for these communities to add paying jobs, as the waste can 

be converted to compost and sold locally. 

How to measure flood resilience 

Good data and statistics tell us if an approach works, and also let us know if one approach 

works better than another. The information derived from measurements lets us identify 

successful actions. It tells us why measures succeed. It allows us to better understand which 

particular actions work not just in one community, but in others as well. There is no 

one-size-fits-all solution or tool to measure resilience. Any system used to measure resilience 

should find answers to specific questions, in our case related to an individual community, and 

in the face of a specific peril, in this case flooding. Useful, empirical measures of flood 

resilience offer clear, unbiased insights, and eliminate the need to make decisions based solely 

on subjective impressions or anecdotal evidence. However, a recent survey conducted for the 

United Nations Development Programme concluded that “no general measurement framework 

for disaster resilience has been empirically verified yet” (Winderl, 2014). We understand 

resilience as an outcome that ensures a community can continue to thrive and develop in 

face of a shock. However, resilience can come from many sources. We are developing a 

community-based flood resilience measurement tool based on the five categories (financial, 

physical, human, social and natural) of sustainable livelihoods (the ‘Five Cs’) framework 

established by the UK’s Department for International Development (DFID, 1999), and the 

four properties (robustness, redundancy, rapidity and resourcefulness) of resilience (the 4R) 

formulated by MCEER (Renschler et al, 2010). We will produce community resilience 

measurements based on a set of factors falling into a capital and one or several properties of 

the 4R. These sources of resilience allow us to assess the level of resilience, using Zurich Risk 

Engineering’s Risk Grading approach to combine them into a joint framework. If resilience 

cannot be empirically verified, how do you empirically measure whether a community is 

more resilient as a result of your work? The approach brings together quantitative and 

qualitative data about the sources contributing to resilience, allowing us not only to assess 

them but also to use the measurement results as a support for decision making to identify 

actions for enhancing resilience. By combining the expertise of all our members of the Zurich 

alliance, this challenge is what we have set out to address. 

Enabling communities to control their own future 

For communities that must regularly deal with floods, change seems particularly daunting. 

Very often these communities struggle to think beyond the immediate present. To keep 

people engaged in finding long-term solutions, community members must be part of the 

dialogue and the solution. Communities need structures that support dialogue: every member 

of a community must be part of discussions when looking for solutions. Following Superstorm 

Sandy, a survey of over 1,000 New York City residents showed that people often tend 

to misjudge the risks floods pose to their own lives and property. A survey (Botzen et al, 

2015) found that homeowners may generally be aware of flood risks, but they often fail to 


ecognize the risks they face as individuals. People tend to underestimate potential losses. 

This might explain why 80 percent of residents in the area inundated by Sandy’s storm surge 

had no flood insurance, and 90 percent of small business had no flood protection, either 

– despite the fact that flood insurance is highly subsidized by the U.S. federal government. 

If people had a clearer understanding of what they stood to lose, they would be more likely 

to take protective measures. Based on these findings, the alliance has recommended that the 

U.S. Federal Emergency Management Agency (FEMA) provide flood maps showing not just 

where floods might occur, but also the damage floods could cause. 

Policymakers need information and insights 

Limiting development in areas with flood hazards is difficult. People want to – or must – live 

and work near water. But they can still be alerted to the risks of building on a flood plain. 

Policymakers and local officials should encourage better planning. Wherever possible, this 

should be done without compromising development that could benefit a community’s 

long-term well-being and prosperity. We are engaging with the government in Peru. The El 

Niño phenomenon occasionally leads to heavy rains that cause floods in some parts of the 

world. Many years may pass between floods, but when they do come, the floods can be 

devastating. In communities in the Piura region in northern Peru and in the Rimac River 

Basin northeast of Lima, our program is working to make it easier for people to be better 

prepared for these types of floods. One factor elevating the risk is that, because floods tend to 

be rare in these areas, people forget the danger, as the buildings built in risky areas since the 

last flood attest. A big part of the program involves working closely with local authorities. 

It is also important to increase confidence in the government’s ability to provide assistance. 

CONCLUSIONS 

Our flood resilience program offers a platform to advocate learning, share knowledge, and 

apply what we learn in individual communities to help others, while serving as a catalyst for 

innovation and policy dialogue. A number of activities will support this approach in the 

future: We are developing an ‘open-source’ solutions catalogue. This makes it easier to share 

knowledge to build flood resilience. The catalogue, drawing on the flood resilience alliance’s 

community programs, will provide vulnerable communities with access to information on 

flood mitigation measures and solutions. It could also include research, processes and tools 

developed by the alliance. We are also testing a tool to understand the flood resilience system. 

The flood resilience measurement will help us assess the impact of resilience-building 

interventions and the community resilience development over time. We are also developing 

a tool to increase the understanding of the interactions between the sources of resilience and 

how they are driving flood risk and wellbeing. It will be tested in 2015 in pilot communities 

in Nepal and Peru. The tool will make it easier to identify key elements, map risk systems and 

sources, and spot potential problems and interdependencies. We are exploring different ways 

the alliance could build further resources. This might include establishing a ‘Flood Resilience 

Academy’. The Academy would be one way to share our knowledge and support practitioners 


in achieving more sustainable, replicable solutions. It could also be supporting efforts to 

identify and develop innovations to address flood risk. We will continue to share our insights 

and findings. We acknowledge that there are many different approaches to resilience. 

Ours includes flood resilience measurements, testing our risk approaches in different settings 

in both low-income and developed economies, in rural and urban settings, to make sure our 

approach is replicable and scalable to help enhancing flood resilience beyond our own 

activities. 

REFERENCES 

- Botzen Walter, Kunreuther Howard and Erwann Michel-Kerjan (2015). Divergence 

between individual perceptions and objective indicators of tail risks: Evidence from floodplain 

residents in New York City. Judgment and Decision Making, 10, 4, 365 – 385. 

- Collier Ben, Miranda Mario and Jerry Skees. Natural Disasters and Credit Supply Shocks in 

Developing and Emerging Economies. Wharton Working Paper 2013-03. 

- DFID. Department for International Development. Sustainable Livelihoods Guidance Sheets. 

http://www.eldis.org/vfile/upload/1/document/0901/section2.pdf, last retrieved August 4, 

2015. 

- Kellett Jan and Alice Caravani. Financing Disaster Risk Reduction. A 20 year story of 

international aid. Global Facility for Disaster Reduction and Recovery (GFDRR), 2013. 

- Lehmann Evan and Climatewire (2015). Extreme Rain May Flood 54 Million People by 

2030. Scientific American, March 5, 2015. 

- Mechler Reinhard and the Zurich Flood Resilience Alliance (2014). Making Communities 

More Flood Resilient: The Role of Cost Benefit Analysis and Other Decision-Support Tools in 

Disaster Risk Reduction. September 9, 2014: http://opim.wharton.upenn.edu/risk/library/ 

ZAlliance-decisiontools-WP.pdf 

- Renschler Chris, Frazier Amy, Arendt Lucy, Cimellaro Gian-Paolo, Reinhorn Anrei and 

Michel Bruneau (2010). A Framework for Defining and Measuring Resilience at the 

Community Scale. MCEER Technical Report 10-0006. 

- United Nations Office for Disaster Risk Reduction, UNISDR (2013). www.unisdr.org/ 

archive/33693 . Last retrieved Aug 4, 2015. 

- United Nations (2014). Department for Economic and Social Affairs. Factsheet: Population 

Facts – Our Urbanizing World. 

- Winderl, Thomas (2014). Disaster Resilience Measurements: Stocktaking of Ongoing Efforts 

in Developing Systems for Measuring Resilience, 59pp. United Nations Development 

Programme 



Chances and Challenges in the Field of Residual Flood- 

Risk and Risk Communication: Ideas from Bavaria 

Chancen und Herausforderungen im Bereich des Hochwasser-Restrisikos 

und der Risikokommunikation: 

Ideen aus Bayern 

Ronja Wolter-Krautblatter, Dipl.-Geogr.¹; Andreas Rimböck, Dr.-Ing.²; Tobias Hafner, Dr.-Ing.³; Christian Wanger, Dipl.-Ing. ³; 

Christoph Oberacker, M.Sc. Geogr.² 

ABSTRACT 

Every flood-protection system is limited and a residual risk always remains. To reduce these 

residual risks, we need responsible citizens and municipalities. But how to motivate them for 

measures of personal provision and precautionary land use? The topic of residual flood risks 

is often suppressed, underestimated and steps back behind day-to-day lives. This article shows 

communication ideas to activate people for measures that reduce the residual risks. 


Jedes Hochwasser-Schutzsystem hat seine Grenzen, ein Restrisiko bleibt immer bestehen. Um 

dieses Restrisiko zu reduzieren brauchen wir verantwortungsbewusste Bürger und Kommunen. 

Wie können wir sie für Maßnahmen der Eigenvorsorge und Flächenvorsorge motivieren? 

Das Thema Restrisiko bei Hochwasser wird häufig verdrängt, falsch eingeschätzt und 

tritt hinter anderen Alltagsproblemen zurück. Dieser Artikel zeigt Kommunikationsideen auf, 

um Menschen für Maßnahmen zur Reduzierung des Restrisikos zu motivieren. 

KEYWORDS 

residual risks; risk-communication; damage potential. 

EINFÜHRUNG 

In der Regel werden Siedlungsbereiche in Bayern bis zu einem hundertjährlichen Hochwasser 

geschützt. Schutzmaßnahmen können jedoch keinen absoluten Schutz vor Naturgefahren 

gewährleisten, ein Restrisiko bleibt immer bestehen: Es kann niemals ausgeschlossen werden, 

dass ein Extremereignis den Schutzgrad übersteigt oder technisches Versagen eintritt 

(Überlastfall) (BUWAL 1999). Es sind gerade die extremen Naturereignisse mit geringer 

Eintrittswahrscheinlichkeit, die als Restrisiko hingenommen werden, die eine Gesellschaft 

jedoch vor die größten Herausforderungen stellen und zur Katastrophe führen können. Das 

1 Bavarian Environment Agency, Augsburg, GERMANY, Ronja.Wolter-Krautblatter@lfu.bayern.de 

2 Bavarian Environment Agency 

3 Bavarian State Ministry of the Environment and Consumer Protection 



weltweit ansteigende Schadensausmaß im Zusammenhang mit Naturereignissen (Munich Re 

2015) zeigt, dass das Thema Restrisiko zukünftig stärker berücksichtigt und ein größeres 

Bewusstsein hierfür geschaffen werden muss. Nur so können die potentiell betroffenen 

Menschen und Kommunen für Maßnahmen motiviert werden, die mögliche Schäden 

begrenzen und das Restrisiko reduzieren. 

Auch wenn Restrisiken nicht komplett vermieden werden können, so kann zukünftiger 

Schaden sehr wohl begrenzt werden. Restrisiken können mitunter von Privatpersonen (z.B. 

mit Eigenvorsorge), aber auch von Kommunen (z.B. mit Flächenvorsorge, Alarm-, Einsatzplanung) 

reduziert werden. Der Inhalt dieses Aufsatzes beschränkt sich jedoch darauf, die 

Eigenvorsorge von Privatpersonen sowie die erweiterte Flächenvorsorge von Kommunen mit 

zielgerichteter Kommunikation nachhaltig zu stärken. Unter Eigenvorsorge werden im 

Folgenden Maßnahmen der Bauvorsorge, Verhaltens- und Informationsvorsorge sowie 

Risikovorsorge verstanden. 


An folgenden Faktoren müssen Ideen zur Minimierung des Restrisikos ansetzen. 

a) Restrisiko hinter Schutzanlagen 

Gewisse sozio-ökonomische Entwicklungen können zu einer Erhöhung des Restrisikos hinter 

Schutzanlagen führen (Rother 2014): Mit dem Bau von Schutzanlagen, deren Weiterentwicklung 

und technische Verbesserung Sicherheit versprechen, steigt der Wert und die 

Attraktivität der geschützten Flächen. Dies führt häufig zu intensiveren Nutzungen, womit 

die Schadenspotentiale und damit auch das Restrisiko ansteigen. Beim Eintritt eines Extremhochwassers, 

das das Bemessungsereignis übertrifft und zu Überflutungen führt, sind die 

Schäden deutlich höher als sie vor der Errichtung der Schutzanlagen und der darauf 

folgenden Nutzungsintensivierung gewesen wäre (siehe Abb. 1). Investitionen in den 

Hochwasserschutz, die eigentlich eine Absenkung der Hochwasserrisiken bewirken sollten, 

können durch die zunehmende Nutzung der Gefahrenbereiche das Gegenteil bewirken 

(Seifert 2012). 

Daraus sollte jedoch keine grundsätzliche Kritik am Bau von Schutzanlagen abgeleitet 

werden, da Schäden bis zum Bemessungsereignis dadurch vermieden werden können. Die 

Siedlungsentwicklung hinter Schutzbauwerken sollte aufgrund der Restrisiken nicht komplett 

verhindert werden. Doch gilt es, die Nutzung an die vorhandene Gefahr anzupassen und z.B. 

auf besonders sensible Einrichtungen (wie z.B. Krankenhäuser, Feuerwehrstationen usw.) 

sowie kritische Infrastrukturen auch in Bereichen mit Restgefährdung zu verzichten. Bei 

bereits bestehenden Einrichtungen dieser Art sollten über den Schutz vor einem hundertjährlichen 

Hochwasser hinaus weitere Maßnahmen realisiert werden. So hat sich zum Beispiel 

ein Bankinstitut in Rosenheim freiwillig entschieden, Vorsorgemaßnahmen über die 

staatlichen HQ 100 

-Schutzmaßnahmen hinaus einzuleiten, um im Katastrophenfall für die 

betroffene Rosenheimer Bevölkerung einen voll funktionierenden „Zahlungsverkehr“ 

aufrechterhalten zu können. Die staatlichen und kommunalen Hochwasserschutzmaßnah- 


men wie Deiche und Mauern finden vorrangig beim Flusshochwasser Anwendung. Daneben 

können Hochwasserschäden jedoch auch durch Grundwasser, Starkniederschläge, wild 

abfließendes Hangwasser oder Rückstau aus der Kanalisation entstehen. Hier liegt die 

Verantwortung bei Bauherrn und Hausbesitzern mit einer verantwortungsvollen Bauweise 

und dem Abschluss von Elementarversicherungen. 

Abbildung 1: Die „ökonomische Wirkungsumkehr“ (nach Seifert 2012). 

Problematisch ist die Diskrepanz rechtlicher Ver- und Gebote: 

– Bis zum hundertjährlichen Hochwasser besteht grundsätzliches Bauverbot sowie weitere 

Auflagen, z.B. die Sicherung von Öltanks in den vorläufig gesicherten und festgesetzten 

Überschwemmungsgebieten. 

– Darüber hinaus gibt es keinerlei gesetzlichen Vorgaben. Doch gerade hier kann durch 

vorausschauendes planerisches Handeln viel erreicht werden. 


) Risikowahrnehmung 

Die individuelle Risikowahrnehmung bestimmt darüber, ob Eigenvorsorge für notwendig 

erachtet und ergriffen werden oder nicht. Da Menschen gemäß ihrer subjektiven Erfahrung, 

Überzeugung und selektiven Wahrnehmung handeln, wird Risiko zu dem, was Menschen für 

bedrohlich halten (Zwick & Renn 2008). „As long as men live by what they believe to be so, 

their beliefs are real in their consequences” (Bendix in Zwick & Renn 2008). Dass diese 

Risiko-Einschätzung der Bevölkerung selten deckungsgleich mit der Einschätzung von 

Experten ist, zeigt sich auch beim Hochwasserrisiko. Die Risikowahrnehmung kann durch 

verschiedene Faktoren verringert werden (Smith & Patley 2009): Für freiwillig eingegangene 

Risiken (z. B. Bebauung überschwemmungsgefährdeter Flächen) ist die Risikobereitschaft in 

der Bevölkerung sehr viel größer als bei zugemuteten Risiken. Weiterhin wird Flusshochwasser 

als bekannte, kontrollierte Gefahr wahrgenommen. Auch die Berichterstattung in den 

Medien (insbesondere das Fernsehen und zunehmend das Internet) beeinflusst die Wahrnehmung 

von Risiken und kann zu Verzerrungen führen. Mediale Inszenierung kann auch zur 

Resignation oder Abstumpfung des Einzelnen führen. Individuelle Erfahrungen mit einer 

Gefahr können ein Anreiz sein, sich mit Eigenvorsorge auseinanderzusetzen. Ob und 

inwieweit dies stattfindet, hängt jedoch stark von der jeweiligen Persönlichkeit ab. Manchmal 

sind Schadensereignisse aber auch zu lange her, um Teil des Bewusstseins der potentiell 

betroffenen Bevölkerung zu sein. Wagner (2004) kommt auf eine „Halbwertszeit“ der 

Erinnerung von 14 Jahren. 

Im Risikomanagement muss die subjektive Risikowahrnehmung der Bevölkerung Berücksichtigung 

finden. 

METHODEN UND ERGEBNISSE 

Aus dieser Problemstellung heraus ergeben sich für uns einige Kommunikationsziele im 

Risikodialog. Insbesondere wollen wir vermitteln, dass: 

– Risikomanagement als Gemeinschaftsaufgabe nur optimal funktionieren kann, wenn alle 

Akteure ihre Verantwortung wahrnehmen; hier ist jeder Einzelne gefragt, 

– jeder Einzelne sein persönliches Risiko beeinflussen kann, 

– eine angepasste Nutzung hinter Schutzbauten (Flächenvorsorge) notwendig ist, um den 

Schaden zu begrenzen. 

Darüber hinaus wollen wir die Grundlagen schaffen, dass 

– jeder Wahrscheinlichkeiten realistisch einschätzen kann, 

– Naturgefahren und daraus resultierende Risiken objektiv eingeschätzt werden können, z.B. 

durch Bereitstellung seriöser Informationsquellen. 

Auf dieser Grundlage soll auch die potentiell betroffene Bevölkerung für Maßnahmen der 

Eigenvorsorge sensibilisiert und für Maßnahmen im Bereich der Bau-, Verhaltens- und 

Risikovorsorge motiviert werden (siehe LAWA 1995). 

Die im Folgenden aufgeführten Botschaften sind die ersten Schritte zur Erreichung dieser 

Ziele und wurden bereits umgesetzt. Einige davon werden im Internetauftritt und in 


Veröffentlichungen des Bayerischen Landesamtes für Umwelt und des Bayerischen Staatsministeriums 

für Umwelt und Verbraucherschutz kommuniziert sowie in öffentlichen Vorträgen 

aufgegriffen. Die Erfahrungen daraus sollen in ein umfassendes bayernweites Kommunikationskonzept 

einfließen und die Themen weiter ausgearbeitet und verbreitet werden. 

a) Jeder “managt” (Alltags-)Risiken 

Das Leben ist voller Risiken und der Umgang damit gehört zum Alltag. Menschen betreiben 

jeden Tag, oft unbewusst, Risikomanagement und wägen Kosten und Nutzen einer Handlung 

ab. Ein Beispiel hierfür ist das Autofahren: Zugunsten der Vorteile, die uns die Mobilität 

bietet, nehmen wir die Risiken in Kauf und haben gelernt damit umzugehen. Der Mensch hat 

das Risikomanagement im Bereich des Autofahrens nahezu optimiert. Wie auch beim 

Hochwasserschutz, wird das Risiko handhabbar, wenn alle Komponenten des Risikokreislaufs 

ineinander greifen. Hierzu muss jeder Akteur seine individuelle Verantwortung wahrnehmen: 

Von Staat und Kommunen, über den Rettungsdienst und die Wirtschaft, bis hin zu 

jedem einzelnen Autofahrer, der sein Risiko, z. B. durch korrektes Verhalten im Straßenverkehr, 

beeinflusst (siehe Abb. 2). 

Abbildung 2: Risikomanagement am Beispiel des Straßenverkehrs. 

Umfangreiche Maßnahmen, die das Risiko auf allen Ebenen des Autofahrens reduzieren, 

verdeutlichen auch, dass hier das Risikomanagement gesellschaftlich breit verankert und 

akzeptiert ist. Ein wesentlicher Unterschied zur Kommunikation von Hochwasserrisiken ist 

es jedoch, dass das Autofahren grundsätzlich als Teil des Alltags positiv besetzt ist und die 


Vorteile der Mobilität klar im Vordergrund stehen. An Hochwasser denkt jedoch kaum einer 

gerne. 

b) Wahrscheinlichkeiten greifbarer machen: Vergleiche mit bekannten Größen 

Der Begriff des hundertjährlichen Hochwassers (HQ 100 

) kann für den Laien irreführend sein, 

da es sich lediglich um einen statistischen Wert handelt. Die Verwendung des Jährlichkeitsbegriffs 

in der Außenkommunikation hat bereits Wagner (2004) und Hagemeier-Klose (2011) 

kritisiert. Der Begriff führt häufig zu Fehlassoziationen bei der Bevölkerung ist darüber 

hinaus auch wegen der erwarteten Veränderungen aufgrund des Klimawandels problematisch. 

Vorgeschlagen wird die Verwendung folgender Begriffe in der Außenkommunikation: 

„häufiges Ereignis“ anstelle von HQ 5-20 

, „mittleres Ereignis“ anstelle von HQ 100 

und „seltenes 

Ereignis“ anstelle von HQ extrem 

. 

Beim Vergleich der Eintrittswahrscheinlichkeit eines Hochwassers mit unterschiedlichen 

Alltagsrisiken wird deutlich, dass unsere persönliche Risikowahrnehmung nicht immer dem 

tatsächlichen „objektiven“ Risiko entspricht (siehe Tab. 1). Beispielweise ist es wahrscheinlicher, 

dass ein Flussanwohner einmal im Leben ein 150-jährliches Hochwasser erlebt, als dass 

ein Autofahrer einmal im Leben bei einem Autounfall verunglückt. Während die meisten 

Menschen wahrscheinlich wissen, wie sie sich bei einem Autounfall zu verhalten haben, 

bleibt fraglich, ob allen potentiell Betroffenen das richtige Verhalten im Hochwasserfall 

bekannt ist. 

Auch das einfache Beispiel eines Würfelbeispiels verdeutlicht die Bedeutung des Restrisikos: 

Die Wahrscheinlichkeit, als Flussanwohner im Laufe eines 80-jährigen Lebens mindestens 

einmal ein hundertjährliches oder noch größeres Hochwasserereignis zu erleben, das ein 

Schutzbauwerk möglicherweise überströmt, ist höher als die Wahrscheinlichkeit, beim 

Würfeln eine 1, 2 oder 3 zu würfeln (siehe Abb. 3). Die tatsächliche Gefhar von einem 

seltenen Hochwasser betroffen zu werden wird also von den allermeisten Bedrohten deutlich 

unterschätzt mit der fatalen Folge, dass persönliche Vorsorgemaßnahmen unterbleiben. 

Wahrscheinlichkeit: 55 % 

mit einem Wurf 

Erleben eines (mind.) 

hundertjährlichen Hochwasserereignisses 

innerhalb eines 

Menschenlebens von 80 Jahren als 

Flussanwohner 

oder 

würfeln 

Abbildung 3: Die Wahrscheinlichkeit in einem Menschenleben von 80 Jahren mindestens ein hundertjährliches Hochwasserereignis zu 

erleben, überschreitet die Wahrscheinlichkeit beim Würfeln eine 1, 2 oder 3 zu würfeln. 


Tabelle 1: Alltagsrisiken im Vergleich. 

c) Schadensentwicklung hinter Schutzbauwerken: Flächenvorsorge für Kommunen 

Um die Bedeutung der Flächenvorsorge zu verdeutlichen, soll hier aufgezeigt werden, wie 

sich Schadenspotenziale nach der Umsetzung von Schutzmaßnahmen entwickeln können. 

Dazu wurden in einem ersten Schritt verschiedene Szenarien in einer Beispiel-Gemeinde in 

Bayern betrachtet (siehe Abb. 4). Die Werte in der Grafik beruhen auf Berechnungen, ergänzt 

durch Annahmen. Im ersten Fall besitzt die Siedlung lediglich einen Schutzgrad für ein 

zehnjähriges Hochwasser (HQ 10 

), beim zweiten dagegen einen Schutzgrad für ein achtzigjähriges 

Hochwasser (HQ 80 

). Das nachfolgende Diagramm zeigt die entsprechenden Schadenseinheiten 

[SE] je Hochwasserereignis, welche bis zum HQ 200 

zunehmen. Ab dem HQ 200 

wird 

angenommen, dass die Schadenseinheiten konstant bleiben, also bildlich „bereits alle Häuser 

unter Wasser“ stehen. Insgesamt wäre für diese Fälle mit einem durchschnittlichen jährlichen 

Schaden von 4,12 SE, bzw. 0,92 SE zu rechnen. Nun wird z. B. ein Hochwasserrückhaltebecken 

gebaut, welches die bestehende Siedlung bis zum hundertjährlichen Hochwasser 

(HQ 100 

) schützt („Vollschutz“). Bei größeren Ereignissen läuft das Becken über. Nachdem 

jedoch mit dem Becken ein HQ 100 

-Schutz erzielt wurde, entfallen jegliche bauliche Einschränkungen 

im geschützten Bereich. Die bestehende Siedlung entwickelt sich weiter und es wird 

ein neues Baugebiet realisiert. Es zeigen sich folgende Ergebnisse: 


– Eine Verbesserung des Schutzgrads einer Siedlung gegenüber Hochwasser bringt eine 

deutliche Verbesserung für den Einzelnen mit sich, da es seltener zum Schaden kommt. 

Während der durchschnittliche jährliche Schaden bei einem Schutzgrad für ein HQ 10 

4,12 SE/Jahr beträgt, fällt dieser auf 0,92 SE/Jahr bei einer Erhöhung des Schutzgrads auf 

ein HQ 80 

, bzw. 0,74 SE/Jahr bei einer Erhöhung des Schutzgrads auf ein HQ 100 

. 

– Für die Gemeinde insgesamt kann der Bau des Rückhaltebeckens mittelfristig möglicherweise 

sogar zu höheren mittleren jährlichen Schäden führen. Dies ist in diesem Beispiel der 

Fall, wenn einerseits bereits ein hoher Schutzgrad (HQ 80 

) bestand und mit der Realisierung 

des Hochwassersschutzes (HQ 100 

) neue Baugebiete entstehen. In dieser Kombination 

nehmen die mittleren jährlichen Schäden von 0,92 SE/Jahr auf 0,98 SE/Jahr zu. 

Abbildung 4: Die Schadensentwicklung hinter Schutzbauwerken. Verschiedene Szenarien in einer Beispiel-Gemeinde in Bayern (Werte in 

der Grafik beruhen auf Berechnungen, ergänzt durch Annahmen). 

Zugleich kann der Bau von Schutzbauwerken als Anstoß genommen werden, bei Bürgern 

und Vorteilsziehenden für individuelle Vorsorge zu werben. Hier spielt auch der Solidargedanke 

eine entscheidende Rolle: Die Hochwasserschutzanlagen werden von der Gemeinschaft 

(Land, Kommune) und somit mit Steuergeldern von Personen, die nicht von Hochwasser 

betroffen sind, finanziert. Von den vorteilsziehenden Bürgern kann im Gegenzug verlangt 

werden, dass sie sich bezüglich der verbleibenden Restrisiken absichern. Beispielweise 

werden Elementarschadensversicherungen deutlich günstiger, wenn die Gefährdungsklasse 

sich aufgrund von Schutzanlagen verbessert. Im Überlastfall können über Versicherungen 

existenzbedrohende Schäden aufgefangen und die Fluthilfeprogramme der Gemeinschaft 

entlastet werden. 


AUSBLICK: KOMMUNIKATIONSKONZEPT „HOCHWASSER UND HOCHWASSERRISIKOMANAGEMENT“ 

Die aufgezeigten Ideen und ersten Schritte sollen künftig verstärkt umgesetzt werden. 

Dazu sollen sie in ein umfassendes Kommunikationskonzept zum Thema „ Hochwasser und 

Hochwasserrisikomanagement“ einfließen. Ziele dieses Konzeptes sind neben den oben 

dargestellten vor allem 

– die Förderung der Akzeptanz für staatliche Hochwasserschutzmaßnahmen, u.a. durch die 

Verstärkung offener gesellschaftlicher Dialogprozesse und anderer partizipativer Verfahren, 

– Steigerung des Risikobewusstseins für Hochwasser innerhalb der kommunalen Verwaltung 

und der breiten Öffentlichkeit. 

Der Prozess der nachhaltigen Minderung von Hochwasserrisiken ist gemeinsame Aufgabe für 

die Wasserwirtschaftsverwaltung, Kommunen, Unternehmen sowie Bürger und wird durch 

das neue Kommunikationskonzept begleitend unterstützt werden. 

DISKUSSION UND FAZIT 

Bei allen Ideen, das Bewusstsein in der Bevölkerung für das Restrisiko zu erhöhen, sehen wir 

auch die Grenzen dieser Möglichkeiten. Verdrängungsmechanismen sind nützliche Teile der 

menschlichen Natur und ermöglichen einen angstfreien Alltag in einem Leben voller 

Gefahren. Unsere Ideen sollen nicht zu einer Verängstigung der Bevölkerung führen, sondern 

aufzeigen, dass Maßnahmen der Eigenvorsorge den Einzelnen entlasten: Wenn beispielsweise 

der Hochwasserschutz beim Neubau eines Hauses einmal umgesetzt ist oder eine Elementarschadensversicherung 

abgeschlossen ist, muss man sich im Alltag weniger Sorgen um das 

Hochwasserrisiko machen. Es muss berücksichtigt werden, dass wir in einer Gesellschaft 

leben, in der zum Teil Informationsüberfluss herrscht. Unsere Botschaften stehen neben dem 

Informations-Input und der Werbung vieler anderer Interessensgruppen. Hinzu kommen 

alltägliche Nöte und Sorgen, die oft Priorität vor der Restrisiko-Problematik im Hochwasserfall 

haben. Daher kommt es sicherlich auch auf den richtigen Zeitpunkt an, um mit unseren 

Botschaften an die Bevölkerung heranzutreten. Die aktuellen Entwicklungen des Schadenspotentials 

zeigen einen großen Handlungsbedarf im Hinblick auf die Kommunikation von 

(Rest-)Risiken im Bereich Naturgefahren. Mit klaren Botschaften können Menschen für die 

Grenzen von Schutzbauten, die Verzerrung menschlicher Risikowahrnehmung, ihre eigene 

Verantwortung innerhalb des Risikomanagements und die Vorzüge der Flächenvorsorge 

sensibilisiert werden. Risikokommunikation darf nicht dazu führen, dass die Bevölkerung vor 

dem Restrisiko resigniert, sondern muss sich zum Ziel setzen, Menschen für das vorhandene 

Restrisiko zu sensibilisieren und das Restrisiko durch konkrete Maßnahmen beherrschbar zu 

machen, um besser mit möglichen Folgen umgehen und leben zu können. 


LITERATUR 

- BUWAL (1999): Risikoanalyse bei gravitativen Naturgefahren, Methode. Umwelt-Materialien 

Nr. 107/I, Naturgefahren. Bundesamt für Umwelt, Wald und Landschaft, Bern. 

- Hagemeir-Klose, Maria (2011): Hochwasserrisikokommunikation zwischen Wasserwirtschaftsverwaltung 

und Öffentlichkeit – eine Evaluation der Wahrnehmung und Wirkung 

behördlicher Informationsinstrumente in Bayern (Dissertation). Fakultät für Wirtschaftswissenschaften 

der Technischen Universität München (http://mediatum.ub.tum.de/ 

node?id=1006044, abgerufen am 18.11.2015). 

- LAWA (Länderarbeitsgemeinschaft Hochwasser) (1995): Leitlinien für einen zukunftsweisenden 

Hochwasserschutz: Hochwasser - Ursachen und Konsequenzen (http://lawa.de/ 

documents/Leitlinien_d59.pdf, abgerufen am 18.11.2015). 

- Munich Re (2015): Topics Geo, Naturkatastrophen 2014. Analysen, Bewertungen, Positionen. 

München. 

- Rother, K.-H. (2014): Zur Abschätzung des Risikos hinter Hochwasserschutzanlagen unter 

sich verändernden gesellschaftlichen Randbedingungen. KW Korrespondenz Wasserwirtschaft 

2014 (7), Nr. 11: S 659 – 666. 

- Seifert, P. (2012): Mit Sicherheit wächst der Schaden. Geschäftsstelle des Regionalen 

Planungsverbandes Oberes Elbtal/Osterzgebirge. Radebeul. 

- Smith, K. & D. Patley (2009): Assessing Risk and Reducing Disaster. Routledge. 

- Wagner, Klaus (2004): Naturgefahrenbewusstsein und –kommunikation am Beispiel von 

Sturzfluten und Rutschungen in vier Gemeinden des bayerischen Alpenraums (Dissertation). 

Fakultät Wissenschaftszentrum Weihenstephan für Ernährung, Landnutzung und Umwelt 

der Technischen Universität München (http://mediatum.ub.tum.de/node?id=603521, 

abgerufen am 18.11.2015). 

- Zwick, M. & O. Renn (2008): Risikokonzepte jenseits von Eintrittswahrscheinlichkeit und 

Schadenserwartung. In: Felgentreff, C. & T. Glade (Hrsg.): Naturrisiken und Sozialkatastrophen. 

Berlin, Heidelberg. 



Practical Management of Debris-flow-prone Torrents 

in Taiwan 

Hsiao-Yuan Yin, Ph.D.¹; Chen-Yang Lee, Master¹; Chyan-Deng Jan, Ph.D.²; Meei-Ling Lin, Ph.D. 3 

ABSTRACT 

In Taiwan, owing to steep topography, weak geological structure, frequent earthquakes and 

severe rainfalls, especially brought by typhoons, debris flow has become one of the most 

serious disasters causing enormous losses of lives and properties. The Soil and Water 

Conservation Bureau (SWCB), Council of Agriculture is the national-level department for the 

preparedness on the reductions of debris flow disasters. To deal with debris flow hazards, the 

first step for mitigation works is to identify the locations of debris-flow-prone torrents (i.e., 

potential debris flow torrents). So far, there are 1,673 potential debris flow torrents in Taiwan, 

and about 48-thousand people living in the debris-flow endangered flooding areas. In this 

paper, we will introduce the practical management measures of debris flow hazards including 

the identification and evaluation of potential debris flow torrents, administrative operation 

process, watershed investigation, risk potential assessment, flooding zone mapping and 

evacuation route map. Good preparedness and management of potential debris flow torrents 

have been proved to be an effective measure in reducing debris flow hazards. 

KEYWORDS 

debris flow hazards; debris-flow-prone torrent; potential debris flow torrent; evacuation route 

map 

INTRODUCTION 

Debris flows are rapid, gravity-induced flows of mixtures of rocks, mud and water. Due to the 

steep topography, young and weak geological formations, earthquakes, erodible soils and 

heavy rainfall brought by typhoons, Taiwan is subject to debris flows and other sedimentrelated 

disasters, causing enormous losses of lives and properties in mountain areas (Jan and 

Chen, 2005). According to the "Disaster Prevention and Protection Act" of Taiwan, the SWCB 

is the national-level department for the preparedness and mitigation works of debris flow 

hazards. As we know that reliable and accurate debris-flow watches and warnings must be 

based on sound identification of areas susceptible to debris flows and recognition of the 

conditions that will result in their occurrence. Therefore, the first step for debris-flow hazards 

mitigation is to identify the locations of debris-flow-prone torrents. In this paper, we will 

introduce our practical experiences of debris-flow managements in Taiwan. 

1 Soil and Water Conservation Bureau, Council of Agriculture, Nantou, TAIWAN, sammya@mail.swcb.gov.tw 

2 National Cheng Kung University, Taiwan 

3 National Taiwan University, Taiwan 



POTENTIAL DEBRIS FLOW TORRENTS 

Since1990, the SWCB has devoted to establish the inventory of potential debris flow torrents. 

Potential debris flow torrents are debris-flow-prone torrents (or gullies) associated with 

important targets to be protected, and are identified after comprehensive evaluation of the 

natural occurrence conditions and the on-site human activities. A debris-flow-prone torrent 

without any residents or infrastructures to be protected will not be classified as a potential 

debris flow torrent, because the zoning of potential debris flow torrents is for human-oriented 

disaster management. In 1996, there were 485 potential debris flow torrents recognized and 

managed. After a catastrophic Chi-Chi earthquake (magnitude 7.3 on the Richter scale) in 

1999, the number of potential debris flow torrents increased to 722 due to significant 

landslides in mountainous areas. However, the number dramatically increased to 1,673 after 

severe actions of rainfalls, especially brought by typhoons during the last 15 years. In order 

to mitigate the possible debris flow hazards, it is crucial for the government to update the 

inventory database of potential debris flow torrents every year before the flood season from 

May to October in Taiwan. 

IDENTIFICATION AND EVALUATION OF POTENTIAL DEBRIS FLOW TORRENTS 

Lin et al. (2010) suggested that the potential degree of debris-flow-prone torrents can be determined 

primarily by watershed area, drainage slope, landslide ratio, and geological structure. 

The comprehensive identification and evaluation procedure of potential debris flow torrents 

used in Taiwan is summarized in an operation direction, named as "Zoning operation 

directions of potential debris flow torrents" published by SWCB in 2013. Two major processes 

of the operation direction are presented in detail as follows. 

Preliminary identification 

In order to identify potential debris flow torrents, two preliminary conditions should be 

considered i.e., the existence of protected targets and the characteristics of valley landform. 

The possible protected targets contain the residents, buildings, roads and other infrastructures 

which are potentially endangered by debris flows, especially in or nearby the downstream 

area of debris-flow-prone torrents. The geomorphologic characteristics of valley landform are 

identified mainly by using aerial orthophoto base map and/or high resolution DTM (digital 

terrian model). It also has a further requirement with the catchment area (only considering 

the watershed areas above the 10 degree gradient point in the main channel) is larger than 

3 hectares. After satisfying the above-mentioned two prerequisites, if the torrent also has 

historic debris flow events or geological disaster potential according to the geological map 

from the Central Geological Survey of Taiwan, then the torrent could be classified as the 

candidate of potential debris flow torrents. 

Potential evaluation 

Once a torrent has been considered as a potential debris flow torrent, the next step is to 

evaluate its potential level. The protected targets and the debris-flow occurrence potential are 


the two major factors used to evaluate and classify the degrees of potential debris flow 

torrents into three categories, i.e., high, medium and low potential debris flow torrents. 

Protected targets 

The vast majority of loss or damage occurs in the depositional zone of debris flows, which is 

referred to as the creek fan. Fans are preferred locations for urban development because they 

are well drained, gently sloping, and often provide good aquifers (Jakob 2005). Generally 

speaking, a torrent considered as the potential debris flow torrent has its downstream 

endangered fan area. From the historic events, people who live in the alluvial fan are 

vulnerable by debris flows meaning that the fan area roughly equates to the so-called affected 

area. The evaluation of protected targets is to estimate the vulnerability of those residents, 

buildings, roads, infrastructures and mitigation measures within the affected area as shown in 

Table 1. The mitigation measures mainly depend on the weighting of watershed management 

performance especially the levels of engineering constructions and ecological treatments 

through the field investigation. Three levels of the weighting values are applied consisting 

poor (weighting:1), fair (weighting:0.8) and good or unnecessary (weighting:0.6). The final 

score of protected targets evaluation is equal to the weighting times the sum scores of 

buildings and transportation facilities. Therefore, the degree of potential damage on protected 

targets can be classified into three risk grades as low (≤40 scores), l medium(40~60 ee la scores) Table and Table 1 

high(≥60 scores). 

Table 1: Potential evaluation of protected targets. 

Protected targets 

(100 scores) 

Buildings (65) 

Transportation 

facilities (35) 

Watershed 

management 

performance 

Final scores 

Classifications 

Scores 

Public infrastructures relating to disaster mitigation 

(Schools, hospitals and public shelters) 

65 

Above 5 houses 60 

1-4 houses 30 

None 0 

Bridges 35 

Roads 20 

None 0 

Weighting 

Poor 1.0 

Fair 0.8 

Good or unnecessary 0.6 

Scores of (Buildings+ Transportation facilities)×Weighting 


le 2 

Occurrence degree 

The occurrence degree is referred to the probability of debris flow events. In this paper, the 

major factors relating to the debris flow occurrence comprise the landslide ratio, drainage 

slope, sedimentation amount, geological structure and vegetation condition as shown in Table 

2. According to the evaluation scores in Table 2, the degree of potential damage on occurrence 

degree also can be classified into three risk grades as low (≤46 scores), medium 

(46~62 scores) and high(≥62 scores). Among them, the scores of 46 and 62 are the 30% and 

70% probability of cumulative distribution diagram during statistic analysis respectively. 

By using the potential damage evaluation on inhabitation including both protected targets 

and occurrence degree, the identified potential debris flow torrents in Taiwan can be classified 

into three risk grades as low, medium and high through the following risk matrix shown in 

Table 3. Those high risk torrents have the priority of hazards mitigation measures such as 

evacuation propaganda and drills, engineering construction and landuse replaning. 

l ee la Table Table 2 

Table 2: Potential evaluation of occurrence degree. 

Occurrence degree 

(100 scores) 

Landslides ratio (25) 

Drainage slope (25) 

Sedimentation (20) 

Geology (15) 

Vegetation (15) 

Classifications 

Scores 

Obvious large landslide areas 

(landslide ratio≧5%) 

25 

Small scale landslide areas 

(1%< landslide ratio

Table 3: Risk matrix used in potential debris flow torrent classification. 

Risk Degree 

Degree of Hazards 

on Protected 

Targets 

Degree of Occurrence Potential 

Low Medium High 

Low Low Low Medium 

Medium Low Medium High 

High Medium High High 

PRACTICAL MANAGEMENT FOR DISASTERS MITIGATION 

Every year before flood season, the information of potential debris flow torrents especially the 

affected areas and the inventory of affected targets will be released to local governments for 

proofreading and revision if necessary. So far in Taiwan, the 1,673 potential debris flow 

torrents are distributed among 17 cities (counties), 159 townships or 684 villages. The 

number of total protected people accumulates to 47,830. Based on the detailed name-list 

inventory information, local governments can easily help the residents to establish their own 

evacuation route maps as shown in Figure 1. Each map contains a lot of emergency response 

information such as important agencies (police station, fire department and hospital) and 

contact persons, shelter information (capacity, address, phone numbers), landing area of 

helicopter, evacuation directions and routes, and the allocation of potential debris flow 

torrents. During the ordinary times, people who live in the affected areas of debris flows will 

follow their own evacuation route maps to have drills and training activities. While during 

the emergency period of typhoons or torrential rains, they will be evacuated by local 

governments according to the debris flow warnings issued by the SWCB. The debris flow 

warning model is based on the rainfall criteria including the effective accumulated rainfall 

and rainfall intensity (Jan et al., 2013). During Typhoon Morakot in 2009 

(caused 665 casualties and 34 missing in Taiwan), 9,100 residents who lived in the affected 

areas of potential debris flow torrents were evacuated. Among them, at least 1,046 people 

escaped from the possible casualties in terms of the post disaster investigation (Yin et al. 

2014). Another successful evacuation example was in Holiu, Fuxing district, Taoyuan city 

during Typhoon Soudelor in August, 2015. Figure 2 shows the rainfall hyetograph progress of 

debris flow event. At 20:30 on Aug. 6, the Central Weather Bureau issued the Land and Sea 

typhoon warnings due to the threat of Typhoon Soudelor. At 12:00 on Aug. 7, the local 

government encouraged the residents to carry on autonomous evacuation according to the 

evacuation route map as shown in Figure 1. At 17:00, the SWCB issued the debris flow 

yellow warning for evacuation advisement and the local residents accomplished the evacuation 

of the whole village before 19:00. In the early morning at 5:00 on Aug. 8, the SWCB 

again issued the debris flow red warning for enforcement evacuation, two hours and 

45 minutes earlier than the occurrence of debris flow at 7:45. The debris flow rushed down 

the Holiu village and buried 14 houses as shown in Figure 3. Due to autonomous evacuation 


in advance, all 48 residents were safely in the shelter. The preparedness management of 

potential debris flow torrents has been proved to make a good contribution to debris flow 

disaster mitigation, especially timely evacuation operation. 

Figure 1: Evacuation route map of Holiu village near potential debris flow torrent (blue line, No. TaoyuanDF034). Shelters, moving 

directions, helicopter landing area, and other information of hospital, police station, fire department, emergency operation center, and 

emergency contact person are all indicated in the map. 


Figure 2: Hyetograph and warning time associated with the debris flow event in Holiu during Typhoon Soudelor on August 8, 2015. 

Figure 3: Debris flow disasters in Holiu village. 


CONCLUSIONS 

Identification of potential debris flow torrents and their risk evaluation play important roles 

in hazards mitigation. Good management on debris flow torrents can effectively reducing 

debris flow hazards. The practical watershed management includes the identification 

procedure, evaluation measures, administrative operation process, watershed investigation, 

risk potential assessment, downstream fan area zoning, and evacuation route map. Through 

good management and annual review of the database of potential debris flow torrents as well 

as the real-time information update, protected people inventory, evacuation operation and 

public awareness, our government has effectively reduce debris flow hazards. After the 

challenges of severe rainfalls brought by tens of typhoon events in the last one and half 

decades, the preparedness management of potential debris flow torrents has been proved 

to make a good contribution to debris flow hazard mitigation. 

REFERENCES 

- Jakob M. (2005). Debris-flow hazard analysis, Chapter 17 in the book of Debris-Flow 

Hazards and Related Phenomena, edited by M. Jakob and O. Hungr, pp. 411-443, Springer, 

published by Praxis Publishing Ltd., UK. 

- Jan, C.D., and Chen, C.L. (2005). Debris flow caused by Typhoon Herb in Taiwan, Chapter 

21 in the book of Debris-Flow Hazards and Related Phenomena, edited by Matthias Jakob 

and Oldrich Hungr, pp. 539-563, Springer, published by Praxis Publishing Ltd., UK. 

- Jan, C.D. Kuo, F.H., Wang, J.S. (2013). Early warning criteria for debris flows and their 

application in Taiwan, of ICL Landslide Teaching Tools, edited by Kyoji Sassa, Bin He Mauri 

McSaveney and Osamu Nagai, TXT-tool 2.886-1.3, published by International Consortium on 

Landslides, Japan. 

- Lin M.L., Chen T.C. (2010). Zoning procedure modification of potential debris flow torrents, 

contract research report of Soil and Water Conservation Bureau, Council of Agriculture, in 

Chinese. 

- Soil and Water Conservation Bureau, (2013). Zoning operation directions of potential debris 

flow torrents, Council of Agriculture, in Chinese. 

- Yin H.Y., Lee C.Y., Jan C.D. (2014). A web-based decision support system for debris flow 

disaster management in Taiwan, 12th International IAEG Congress, Sep. 15-19, 2014, Torino, 

Italy, volume 3, pp. 109-113. 



Living 

with natural 

risks

Data Acquisition and Modelling 

(monitoring, processes, technologies, models) 


DATA ACQUISITION AND MODELLING (MONITORING, PROCESSES, TECHNOLOGIES, MODELS) 

Deciphering dynamics and magnitude of a recent 

debris-flow disaster in Vratna Dolina 

Juan Antonio Ballesteros Canovas, PhD. 1 ; Milan Lehotsky 2 ; Karel Silhan 3 ; Sladek Jan 2 ; Anna Kidova 2 ; Radek Tichavsky 3 ; 

Pavel Stastny 4 ; Markus Stoffel 5 

ABSTRACT 

The capacity to describe extreme debris-flow events responsible of large disasters is clearly 

limited by the usual lack of records and direct observations. Post-event field recognition is a 

powerful tool to overcome these limitations and to deliver baseline data for a better process 

understanding. In this communication, we combine post-event field recognition and 

dendrogeomorphic approaches to describe an extreme debris-flow event which took place in 

Vratna valley (Slovakia) in 2014 and to quantify its magnitude. Additionally, we analyse the 

meteorological triggers of the flow. Results provide insights for an improved characterization 

of extreme events in this region, and are thought to be useful to calibrate physically-based 

models in order to implement risk reduction strategies. 

KEYWORDS 

debris flow, post-event field recognition, tree-ring, peak discharge, Vratna dolina 

INTRODUCTION 

Rainfall-induced landsliding represents one of the most common triggers of massive debris 

flow in many mountain environments. Soil moisture saturation commonly results in a 

decrease of soil cohesion, which in turn can result in intense sediment transfer into steep 

channels. As a result, powerful mass movements can be one of the downstream consequences 

(Jakob and Hungr, 2005) with recorded transfers of up to ~109 m³ and intense geomorphic 

changes, mostly related with scour erosion and sediment deposition on flatter terrains. 

In order to reduce disaster risk, an appropriate description, analysis and interpretation of 

extreme debris-flow events will be essential. However, in the Slovakian context, but also in 

other mountain regions, limited knowledge still exists on how big such extreme events are 

and how they are affected by ongoing and future climatic changes. This question is not trivial, 

as limited records generally exist in mountain areas, thus avoiding to draw reliable conclusions 

about their expected magnitude, triggers and geomorphic impacts. 

Systematic post-event field recognition is a valuable approach to document the affected areas, 

magnitude and flow type of recent high-magnitude events (Marchi and Cavalli, 2007). This 

1 University of Bern, Bern, SWITZERLAND, juan.ballesteros@dendrolab.ch 

2 Slovak Academy of Sciences, Bratislava 

3 University of Ostrava, Department of Physical Geography and Geoecology 

4 Slovak Hydrometeorological Institute 

5 dendrolab.ch - The Swiss Tree-Ring Lab. Institute for Geological Sciences. Baltzerstrasse 1+3, 3012 Bern Climate Change an 

Climate Impacts (C3i) Institute for Environmental Sciences 7 route de Drize, 1227 Carouge-Geneva 



approach combines several data sources, including fresh botanical or geological evidence, 

gauging records, direct observation or technical documents. By carrying out post-field 

recognitions, researchers and practitioners will gain new insights about catchment response 

as well as about the effectiveness of existing mitigation measures. Beyond these advantages, 

outcomes from post-field event recognition and indirect proxies are extremely valuable to 

calibrate physically-based models and define scenarios so as to delimit hazard zones. 

In this communication, we report observation and measurements from post-event field 

recognition where we aimed at quantifying the extraordinary debris-flow event of 21 July 

2014 in Vranta Dolina (Malá Fatra National Park, Slovakia) after intense rainfalls. The 

extraordinary nature of the July 2014 debris flow in Vratna valley is underlined by its large 

geomorphic imprint, and direct economic losses in the order of 6 million US$, mainly due to 

fan deposition in a ski resort and erosion processes of transport infrastructures downstream of 

the resort. During the event, 122 tourists remained isolated and had to be rescued the next 

day. 

METHODS 

Study site: The Vratna valley is located in the northern part of the Malá Fatra Mts (Slovakia). 

The catchment has an area of almost 18.4 km 2 with high relief energy of up to 1000 m. The 

siteis characterized by complex geology (limestone, dolomites, sandstones, and quartzite). 

The upper slopes (up to 30 – 40°) are covered by alpine meadows, dwarf mountain pines and 

rock outcrops. In the central part, the valley is covered with spruce and beech forests. At the 

lower part of the valley, the valley bottom is defined by a narrow gorge (ca. 20 m). Gauging 

records do not exist for Vrátňanka Brook. The valley belongs to the Malá Fatra National park 

and is considered as an important all-season tourist centre. 

Methodological steps: Two field surveys were carried out after the debris-flow event. The 

first field survey was performed immediately after the event on 23 July 2014, and focused on 

the middle and lower sections of the catchment. During this first survey, debris-flow damage, 

the extension of debris deposits, and flood stage based on fresh indicators has been analysed 

and reported. Eyewitness information related with the event was also recorded. In addition, 

we collected aerial pictures taken from a helicopter of the upper part of the catchment to 

determine fresh geomorphic imprints of erosion at the hillslopes. The second field survey was 

performed during September 2014. During this survey, we (i) described the geomorphic 

imprints related with the debris-flow event in the middle and upper parts of the catchment, 

(ii) perform a survey for peak discharge estimation and modelling, and (iii) investigated past 

debris-flow activity based on tree-ring analyses (Stoffel and Corona, 2014). 

For the geomorphic characterization, we delineated the upper failure zones, the middle 

transport (and erosion) zone, and the lower deposition zone, as well as the bedload sediment-laden 

floodwater zone. Field mapping was combined with free access videos and aerial 

imagery, as well as with field measurements using a total station and laser range finder. In the 

middle zone affected by both transport and erosion, four channel reaches connected with the 

main sediment sources and presenting stable cross-section were surveyed in detail. Two more 


iver reaches located in the bedload sediment-laden floodwater zone was surveyed as well in 

order to determine the evolution of peak discharge along the entire catchment. Cross-sections 

were characterized by presenting fresh paleostage indicators on trees (PSI) and high water 

marks (HWMs; Ballesteros-Cánovas et al., 2015) related with the debris-flow event. 

Past debris-flow activity was investigated based on tree-ring analyses of scarred trees growing 

on the cone (Stoffel and Corona, 2014). Event detection was based on injuries, tangential 

rows of traumatic resin ducts (TRD) and abrupt growth suppression/releases and reaction 

wood. For cross dating and the detection of growth anomalies, we also sampled undisturbed 

Norway spruce (Picea abies) trees growing in nearby areas. 

We then used the continuity equation: 

Q=A·V (m 3 s-1) 

… to estimate peak discharge of the debris-flow event, where A is the cross-sectional area 

(m 2 ) and V is the mean velocity of the flow (m s-1). At the middle transport/erosion zone, the 

mean debris flow velocity (VDF) was estimated by the empirical equation as a function of the 

flow depth (H, given by HWMs) and channel slope (I in %) (Rickenmann, 1999): 

VDF = 4.83 x H0.5 x I0.25 

At the sediment-laden floodwater zone, the mean flow velocity (VF) was alternatively 

obtained by using the critical-depth method (Webb and Jarret, 2002). This method is specially 

recommended to determine peak discharge of paleoflood events in mountain streams up to 

7% slope subjected presenting critical cross-section, because the peaks discharge estimation is 

independently of roughness. Mathematically, the flow velocity is estimated by equalling the 

Froude Number to 1 (F=1): 

VF = (g x dh) 0.5 

where g is the gravitational acceleration and dh the flow depth provided by the height of 

HWMs or PSIs (dh=A/T , with A= cross sectional area of the flow and T= the width of the free 

surface). 

Finally, we also characterized the meteorological conditions related with the triggering of this 

extreme debris-flow event. To this end, we used gridded observation fields from the freely 

available E-OBS and TRMM datasets as well as rain gauge records from the Štefanová station 

(49o13´57´´, 19o03´41´´, 632 m asl) located 3 km far from the ski resort and 5 km from the 

source area of the debris flow at Chleb peak. 

RESULTS 

Geomorphic imprint of the disaster and debris-flow quantification 

The methodology has allowed description of the geomorphic imprint and quantification of 

the magnitude of the extreme debris-flow in Vratna valley in 2014. 

The failure zone was characterized by 33 patches, ca. 136 000 m 2 in total, on north, northwest 

and west-facing slopes with angles between 22 and 33°. Distinct scarps of up to 1.5 m in 


height along the upper and lateral margins of the failure zones distinguish the failure surfaces 

from adjacent undisturbed slopes. Their upper margins are usually developed at the contact 

with the downslope lines of dwarf mountain pine patches. Undisturbed sediment exposed in 

the scarps indicates that failed material consists primarily of a soil body and clast-supported 

regolith and that the initial mechanism of failure in these debris flows was sliding. 

The transport/erosion zone is composed by nine first- and second-order channels with lengths 

between 260 and 950 m, and average gradients of 22° to34°. Debris-flow erosion washed out 

the colluvium or exposed fresh outcrops of the bedrock. Downstream, the debris flows 

removed most of the forest vegetation along the channels, which supposes an important 

input of large woody debris (LWD) to the system. The vertical erosion and transport of debris 

flows was intense, mostly manifested on the heads of the original fans as fan head trenches 

(example zone 5: analysed channel 100 m long with 10 m wide, an average slope by almost 

20° defining between 3-4 m of vertical erosion). A large amount of fresh sediment marks and 

scars on trees defines the maximum flow heights were observed in channel reaches where 

quartzite outcrops. At these sites, we also estimated peak discharges using the methods 

described above (see Figure 1). The total discharge contributing to the main fan was 

419.2m³/s. 

The sediment zone: The main deposition is located at the junction of four feeder valleys and 

fan heads covering the upper and middle parts of the main valley. The 1-3 m thick deposition 

covered a surface 500 m in length and 20-50 m in width of an old debris fan, defining a 

volume of roughly 75,000m 3 . Clasts do not display impact marks, suggesting that the flow 

was laminar. Sand and pebbles occupy the interstices between the boulders and cobbles. 

The existing ski resort operated as an obstacle to the flow which led to its deceleration and 

mass deposition as well as to flow bifurcation. This infrastructure was seriously damaged by 

the flowing mass which penetrated inside and deposited there an up to 1-2 m thick layer. The 

fan toe, used as a parking area, was impacted by the mixed gravel-sandy and LWD deposition 

up to 0.5 m thick and 39 cars have been removed and damaged by the debris-flow current. 

The main very coarse (boulder-cobble) debris-flow mass followed the main, fourth-order 

channel and moved downstream. On the fan-toe with an average of 11° gradient, the already 

entrenched channel (ca. 3 m) incised again into bedrock (up to 1 m), doubled its width (up to 

15 m) and undercut the access road. Several cars from the parking area were thereby 

damaged. 

Suspended load sediment-laden floodwater zone: This area exhibits a different response to 

the debris-flow event in comparison with those previously mentioned. The upper part of this 

zone is ca. 1,700 m long with a relatively wide valley floor (100-150 m), unconfined channel, 

relatively low banks and low slope (on average 1.5°) inhibited further transport of coarse 

mass within the channel. However, the trash lines, avulsions and sandy deposition reflect the 

flood demine of the suspended sediment load during the prevailing time of the debris flow 

event. Downstream along the ca. 800 m gorge the flow eroded the route at several places so 

that it was unpassable for cars. At this level, peak discharge was reconstructed at 103 and 

117.5 m³/s for sites 5 and 6, respectively. According to official reports, 20,000 m³ of the 


deposited material has been removed, ca. 5 km of the Vrátňanka Brook including the reaches 

in Terchová village had to be channelized again after the event resulting in overall costs of 

restoration works of 12 M€. 

Meteorological context and trigger 

Figure 1: Geomorphological sketch of the Vratna Dolina catchment. The cross section 1 to 6 represent the location were the peak 

discharges were reconstructed. 

From a meteorological perspective, the synoptic situation related with this extreme debrisflow 

event is characterized by a shallow trough of low pressure spread from the North Sea 

over the Alpine and Carpathian areas on 20 July 2014, which moved over SE Moravian and 

W Slovakia the subsequent day (Figure 2). At the same time, warm and humid air masses 

crossed Slovakia, and a low pressure system at high atmospheric layers in N Italy left an 

unstable vertical distribution of air pressure. The co-occurrence of convergent air circulation, 

air humidity, and convection systems left intense and short-lived precipitation over the Malá 

Fatra region. According to rainfall data from the local gauge station, the core of the downpour 

occurred from 15:30 to 17:00 with 52 mm (daily total 62 mm). The course of the downpour 

indicates two rainfall events/waves: a first one 15:30 - 16:05 with 18 mm and a second one 

from 16:05 to 17:00 with 34 mm. At short term, the maximum rainfall intensity that took 

place during the first wave was between 50–60 mm/h, and during the second wave almost 

60–80 mm/h. 


Figure 2: Surface field pressure retrieved from E-OBS dataset and daily area-average precipitation at the study site for the entire month 

(TRMM dataset). 

Past debris flow activity at the study site 

Despite of the lack of gauge station records, historical archives suggest that similar large flow 

events could have taken place in the catchment in 1848 (Kapasný, 2008). Tree-ring analysis 

suggests that in recent decades, at least six events occurred on the two cones located in the 

upper ski resort. The oldest event on cone 1 was dated at 1959 (1965 on cone 2) and the 

youngest to 1998 (2008 on cone 2). Synchronous process activity on both cones occurred in 

1965 and 1991. The event chronologies on both cones are shown in Figure 3. This paleoreconstruction 

suggests a recurrence interval of 9.2 years for cone 1, and 7 years for cone 2 

over the last ~50 yrs. 

Figure 3: The chronologies of past events on both studied cones. Full black line – sure event (Wit ≥ 1 and at least two trees with signal 

in tree ring series), dashed black line – probably event (at least two trees with signal but Wit < 1) 

CONCLUSIONS 

The field surveys carried out after the event coupled with available aerial pictures and 

different climatological dataset have allowed to quantify and describe the extreme debris flow 

in Vratna Dolina during July 2014. We have described the geomorphic imprint of this event, 

and evaluated its erosion, transport and sedimentation capacity. Our observation and 

quantification revels that both large amount of sediment and woody debris were transported 


during the event, with a major deposition zone that matches with the location of the ski 

resort facilities. The subset of observations here reported can be used to calibrate debris flow 

models (i.e. RAMMS model) and provide reliable scenarios for future extreme events, which 

could then be used to help the planning and implementation of future infrastructure 

development in the region. Moreover, our observation related with climate-linkages and 

meteorological triggers could be used to improve our understanding of rainfall threshold for 

triggering extreme debris flow events in the region, and consequently useful for early 

warning system proposes. Therefore, despite the exceptional weather conditions during the 

event, we could report historical and tree-ring records suggesting that debris-flow events 

have taken place frequently in the area. 

Our observations can be used to provide inputs for the modeling of extreme scenarios which 

could then be used to help the planning and implementation of future infrastructure 

development in the region. Moreover, our outcomes can be used to improve our understanding 

of climate-extreme event linkages in European mountains. 

REFERENCES 

- Ballesteros-Cánovas, J.A. Stoffel, M. St. George, S. Hirschboeck, K. (2015). A review of 

flood records from tree rings. Progress in Phisical Geogrpahy. In press. 

- Jakob M., Hungr O., Jakob D.M. (2005). Debris-flow hazards and related phenomena. 

Berlin, Springer. 

- Marchi, L., Cavalli, M. (2007). Procedures for the documentation of historical debris flows: 

application to the Chieppena Torrent (Italian Alps). Environmental management 40(3): 

493-503. 

- Rickenmann, D. (1999). Empirical relationships for debris flows. Natural hazards 19(1): 

47-77. 

- Stoffel, M., Corona, C., 2014. Dendroecological dating of geomorphic disturbance in tress. 

Tree Ring Res. 70, 3-20. 

- Webb RH., Jarrett RD. (2002). One Dimensional Estimation Techniques for Discharges of 

Paleofloods and Historical Floods. Ancient Floods, Modern Hazards: 111-125. 



Modeling rockfall trajectories with non-smooth 

contact/impact mechanics 

Perry Bartelt 1 ; Werner Gerber 2 ; Marc Christen 2 ; Yves Bühler 1 

ABSTRACT 

Rockfall is an increasing problem in many mountainous regions throughout the world. 

In Switzerland during the summer of 2015 we observed escalating rockfall activity due to 

exceptionally warm temperatures and degradation of permafrost. Consistent and reproducible 

calculation methods are urgently required to predict rockfall inundation areas for hazard 

mapping and mitigation measure planning. In this paper we examine the use of non-smooth 

contact/impact mechanics to model the trajectory of falling rocks. Non-smooth mechanics 

parameterize the mechanical resistance of ground with consistent frictional parameters. 

These parameters account for rock shape and different rock propagation modes such as 

rolling, skipping, jumping and sliding; enabling a the characterization of different ground 

types, varying from extra soft to extra hard. We apply the new method to back-calculate a 

well-documented rockfall event demonstrating the important role of ground parameterization 

and rock shape in rockfall intensity mapping. 

KEYWORDS 

rockfall dynamics; modeling; hazard assessment 

INTRODUCTION 

The preparation of rockfall hazard maps and the planning of mitigation measures require 

reliable and consistent methods to predict rockfall runout, velocities and energies (Dorren, 

2003; Volkwein et al., 2011). To date, rockfall modeling is based primarily on empirical 

shadow angle type methods (Glover, 2014) or lumped-mass/rigid-body type approaches 

(Schweizer, 2015). Discrete element methods that model the rock impact-rebound process 

using elastic-inelastic contact stiffness have also been applied to the rockfall problem (An and 

Tannant, 2007; Thoeni et al., 2014). These methods have been calibrated to a significant 

degree such that they can be applied to solve many practical problems. 

But a significant problem with existing rigid-body rockfall models is that the rock-ground 

interaction is based on simple rebound physics. Coefficients of restitution are used to define 

the relationship between the incoming rock velocity and the post-impact velocity vector. The 

restitution coefficients therefore define the jump direction, height and length. This information 

is used directly in the preparation of hazard maps. 

1 WSL Institute for Snow and Avalanche Research SLF, Davos Dorf, SWITZERLAND, bartelt@slf.ch 

2 Swiss Federal Institute for Forest, Landscape and Snow Research WSL 




The implicit assumption of rebound physics, however, is that rocks jump, a-priori. The entire 

rock-ground interaction is condensed to a single time point. Rocks do not slide or roll and 

remain in contact with the ground. Some rigid-body models account for rock rotations in the 

in-flight phase, but the influence of rock rotations during impact is generally ignored as the 

restitution coefficients are defined with respect to the translational velocity. The restitution 

coefficients are independent of the impact configuration, which depends on the rock shape 

and rock orientation at the time of impact (Glover, 2015). 

To overcome this problem, many rockfall models use stochastic methods to describe the 

rock-ground interaction. Random number generators that account for the variability of the 

impact process essentially produce restitution coefficients. It is argued that the ground is 

variable and therefore the restitution coefficients must likewise vary. The variability of the 

impact configuration (rock size, shape, orientation, rotation) is not considered. Because rocks 

roll and slide before stopping (Fig. 1), many rigid-body approaches must also adopt ad-hoc 

stopping procedures to reproduce the observed runout distance. 

The use of simple rebound physics coupled with ad-hoc stopping procedures limits the 

practical relevance and use of rockfall models. For example, important information from field 

observations (rock size, shape, scar length, scar depth, jumping, sliding, rolling motion) 

cannot be used to define consistent rockfall model parameters. It is therefore difficult to transfer 

modeling results from one geomorphological setting to another. 

Figure 1: The rockfall model RAMMS::ROCKFALL applies non-smooth contact impact mechanics to the rockfall problem. Rock trajectories 

in three-dimensional terrain are calculated with variable rock shapes. Rock rotations and ground sliding are considered. 


METHODS: NON-SMOOTH CONTACT/IMPACT MECHANICS 

To overcome this problem, newly developed methods to treat the contact and impact were 

applied to the rockfall problem (Schweizer, 2015). The methods were introduced in the 

natural hazard program RAMMS (Christen et al. 2012) and applied to model observed 

rockfall events. The components of the model are: 

– Rock shape, size and impact configuration. Rocks are described by point clouds that 

represent the complex surface topology (Fig. 2). Rocks can therefore have different forms, 

varying from equant to plate like. Rock size can vary from 0.5 m 3 (rockfall) up to 100 m 3 

(blockfall). Impact forces are applied on the rock surface, depending on the impact 

configuration, leading to rock rotations and a natural modeling of lateral dispersion. 

– Rock rotations, quaternions and gyroscopic forces. Rock rotations are considered not 

only in the in-flight phase, but also during the contact/impact phase. Thus, rolling and 

sliding can be modeled. Computationally efficient quaternion algebra is used to define the 

orientation of the rock. Gyroscopic forces resulting from the rock rotations will upright 

plate-like rock shapes producing dangerous wheel-type trajectories. 

– Set-valued contact laws with friction. Set-valued force laws are used to describe 

stick-slip type phenomena (Fig. 2). They allow sliding with friction. These laws are essential 

to initiate rock-jumping and therefore long rock runout. Set-valued force laws are 

non-stochastic and therefore can be transferred to similar terrain. Restitution coefficients 

are not used to parameterize the rock-ground interaction, rather the sliding friction 

parameters. Soft ground allows more sliding and therefore true rock stopping without 

ad-hoc conditions. 

PARAMETERIZATION OF GROUND FRICTION 

The most significant difference between rebound models and non-smooth contact/impact 

models is the parameterization of ground friction. Rebound models simulate ground hardness 

with apparent restitution coefficients that describe the bounciness of the terrain for a single 

impact and point. In non-smooth models, ground impact is considered over a finite sliding 

distance s , which we term the scarring length (Fig. 2). The scarring length s is defined when 

any point on the surface of the rock is in contact with the ground. Contact is not defined with 

respect to the rock’s center-of-mass, rather with respect to the complex surface geometry of 

the rock. Contact forces, including friction, are therefore introduced at the surface of the rock 

and cause rock rotations. The magnitude of the rotation at exit (when the rock departs the 

scar) are therefore dependent on the orientation of the when it hits the ground. A restitution 

coefficient ε is applied at the rock surface to account for energy dissipation in the normal 

direction. The model uses ε = 0 (fully plastic) for all terrain types. Bounce heights are purely 

a function of the impact orientation and the friction in the tangential direction. Hard ground 

does not allow sliding and therefore produces bouncing modes of propagation. The ground 

scar is being created at the speed and given by 


⎧ v 

s = ⎨ 

⎩− 

βs 

if contact 

if no contact 

Equation 1 

where ⎧ v is the sliding speed of the rock and β is a parameter controlling ramping effects, 

⎨ 

created when ground material is being displaced during the scar formation. When the ground 

is soft β is small because the ground is softer and the rock remains in contact with the ground 

longer (Fig. 2). The scar parameter β serves to extend the time that ground friction, or any 

other drag process, operates on the rock. It is therefore a useful parameter to describe 

low-lying vegetation layers in forests. 

Figure 2: a) The interaction of the rock with the ground interface involves prolonged frictional sliding. This complex process is not 

governed by apparent restitution coefficients, rather mechanical properties of the ground under large deformation and strain-rates. 

Impulsive rebound forces coupled with frictional torques induce rock rotations leading to skipping/spinning-type propagation modes. b) 

Coulomb friction is parameterized as a function of the sliding distance using four parameters: μ min , μ max , κ and β. An additional 

velocity dependent drag ν accounts for viscous strain-rate effects during ground penetration/deformation. 

When the rock is in contact with the ground friction is described by a combination of 

Coulomb friction S N and viscous friction S V . These forces act in the tangential direction; that is, 

in the direction opposite to the rock velocity. The Coulomb friction coefficient μ(s) for s > 0 is 

given by the transcendental function 

2 

= . 

µ ( s) 

µ 

min 

+ ( µ 

max 

− µ 

min 

) arctan( κs) 

π 

Equation 2 

This function contains three parameters: μ min , μ max and κ , see Fig. 2. The parameter μ min 

defines the initial friction at the beginning of the ground interaction. Softer ground has lower 

μ min values. We then assume that during the ground interaction that the rock penetrates the 

ground and because of confining pressures ground material cannot be easily displaced out of 

the scar. This leads to an increase of friction, or scar hardening. The maximum friction is 

defined by μ max which we regard as a limit friction value, constant for all soils, but very low for 

easily deformable and porous ground materials like snow. An important parameter is κ which 

defines how a quickly the ground material changes from μ min to μ max which is a function of 

how easily a ground material compresses (in the sliding direction) during ground penetration. 


/ 

/ 

/ 

96 

97 

98 

99 

100 

101 

102 

103 

104 

105 

106 

107 

108 

109 

110 

111 

112 

113 

114 

115 

116 

117 

118 

119 

120 

121 

122 

123 

124 

compresses (in the sliding direction) during ground penetration. These parameters may well be 

seasonally dependent, for example, a function of whether the ground is froen or unfroen, or a 

These function parameters of the water content. may well be seasonally dependent, for example, a function of whether the 

ground is frozen or unfrozen, or a function of the water content. 

The Coulomb friction friction term is supplemented term is supplemented with a viscous, rate-dependent with a viscous, friction, rate-dependent friction 

S V 

= −νM v 

Equation 3 

Equation 3 

where M is the total mass of the rock and ν is the viscous drag coefficient. The softer the 

where M is the total mass of the rock and ν is the viscous drag coefficient. The softer the ground, 

ground, the larger is the viscous coefficient.With this five parameter frictional model it is 

the larger is the viscous coefficient. 

Table TABE 1: 1 Ground categories and associated friction parameters in the RAMMS::ROCKFALL program 

Category Picture Description Parameters Category Picture Description Parameters 

With this five parameter frictional model it is possible to establish seven general ground 

Extra Soft 

Very wet ground. µmin: 0.2 Medium 

Rocks jump. 

µmin: 0.4 

Moor categories: Extra Soft, Soft, Cannot edium cross Soft, without edium, µmax: edium 2 Hard, Hard and Extra Hard, see Table Penetration depths are µmax: 2 

deep sink-in. 

β: 50 Shallow 

small. Ground is flat. β: 175 

1. These categories differ in the degree of a) sliding 

κ: 

friction, 

1 

b) bounciness, Meadow c) ground contact time Rocky debris is present. κ: 2.5 

ν: 0.9 

Non-paved mountain 

and d) viscous drag. epending on the rock-ground impact configuration, different propagation 

v: 0.5 

roads. 

Soft modes can result, varying from Soft ground extreme with braking many in soft µmin: terrain 0.25 (Extra Hard Soft) to extreme umps on Rocks jump over 

Moist, deep 

deep soil layers. Ground µmax: 2 Rock scree 

ground. Mixture of large 

Meadow hard ground (Extra Hard). An contains eighth no category, large rock Snow, β: was 100 especially introduced to account for and small rocks. 

fragments. Often very κ: 1.25 

Usually without 

extremely low friction sliding moist. modes. Foot indundations ν: 0.8 

vegetation. 

/ 

remain and are visible. 

/ 

The parameter values were determined using a combination of labortory tests, experimental 

Medium 

Rocks penetrate µmin: 0.3 Extra Hard 

Ground is very hard 

Soft investigations in the field and meadow application surface on leaving documented µmax: 2 case studies Rockface (Glover, / 2014). Clearly, and is marginally 

Deep 

impact scars. Soil is β: 125 Bedrock 

deformed by rocks. 

Meadow more work will be performed deep, in this few direction rock in the near future. 

κ: 1.5 

Rockface and paved 

fragments. 

ν: 0.7 

roads. 

/ / 

Medium FIGURE 3 

Meadow is deep, but µmin: 0.35 Snow 

Rocks slide on snow 

Meadow 

contains rock fragments. µmax: 2 

surface. 

As shown in Fig. 3, the different The meadow ground can categories be result β: not 150 only in different runout distances, but 

covered with vegetation. κ: 2 

also propagation velocities and spreading behavior. v: In this 0.6 example an ideal 30 o slope was used to 

compare the different hardness categories. An equant shaped rock was used and started with 

different orientations. The large spread in traectories, especially in hard terrain, is therefore 

possible obtained by to changing establish the initial seven conditions general and ground not by random categories: rebound coefficients Extra Soft, during Soft, the Medium Soft, 

Medium, process modeling. Medium Hard, Hard and Extra Hard, see Table 1. These categories differ in the 

degree of a) sliding friction, b) bounciness, c) ground contact time and d) viscous drag. 

Depending 

Results: The 

on 

Brien 

the rock-ground 

(GR) case study 

impact configuration, different propagation modes can result, 

varying from extreme braking in soft terrain (Extra Soft) to extreme jumps on hard ground 

The village of Brien, Canton Grisons (GR), Switerland is the site of much on-going rockfall activity. 

(Extra Hard). An eighth category, Snow, was especially introduced to account for extremely 

n uly 24, 2014 a 94 ton, 34 m 3 plate-like rock detached from the upper limestone layer (1650 m 

low friction sliding modes. 

The parameter values were determined using a combination of labortory tests, experimental 

investigations in the field and application on documented case studies (Glover, 2014). 

Clearly, more work will be performed in this direction in the near future. 

As shown in Fig. 3, the different ground categories result not only in different runout 

distances, but also propagation velocities and spreading behavior. In this example an ideal 30° 

slope was used to compare the different hardness categories. An equant shaped rock was used 

and started with different orientations. The large spread in trajectories, especially in hard 

terrain, is therefore obtained by changing the initial conditions and not by random rebound 

coefficients during the process modeling. 

µmin: 0.55 

µmax: 2 

β: 185 

κ: 3 

v: 0.4 

µmin: 0.8 

µmax: 2 

β: 200 

κ: 4 

v: 0.3 

µmin: 0.1 

µmax: 0.35 

β: 150 

κ: 2 

v: 0.7 


Figure 3: Simulation results using equant shaped rocks for three different ground categories: soft, medium and hard. Runout distances 

and calculated velocities. Note that spreading increasing with increasing hardness. The statistics are based on the maximum velocity 

in traversed cell. 

Figure 4: Disposition of the Brienz case study. On July 24, 2014 a large, plate-shaped limestone rock released and travelled to the 

road. Insets: the release zone, destroyed larch forest and toppled rock. The scar width indicated a wheel-like rolling propagation mode. 


RESULTS: THE BRIENZ (GR) CASE STUDY 

The village of Brienz, Canton Grisons (GR), 

Switzerland is the site of much on-going 

rockfall activity. On July 24, 2014 a 94 ton, 

34 m 3 plate-like rock detached from the 

upper limestone layer (1650 m a.s.l.) and 

descended down 800 m to the outskirts of 

the village (1150 m a. s.l.), barely stopping at 

the road connecting Davos and Lenzerheide 

(Fig. 4). The instable limestone layer is 

located above a deforming Schist stratum of 

Bündnerschiefer. The rock accelerated on the 

steep, transition zone covered with rock 

debris, possibly reaching peak velocities up to 

50 m/s. The rock crashed into an old larch 

forest, knocking over several trees before 

decelerating and stopping on a flat meadow, 

located next to the road (Fig. 4). 

Immediately after the event, a laser scanning 

of the visible portion of the rock was 

performed and ground scarring on the 

runout meadow documented. The width and 

length of the penetration scars indicated a 

skipping, rotating wheel-like propagation 

mode before the rock lost velocity and 

toppled to its final resting place (Fig. 4). 

Simulations of the Brienz rockfall event using 

RAMMS::ROCKFALL were used to parameterize 

ground friction. The laser scanned rock 

was used to define the exact shape and 

dimensions of the rock (Fig. 5). In the 

simulations 500 rocks with different 

orientations were released from the point 

release zone identified in the field observations. 

Figure 5: Role of rock shape on rockfall runout. Equant shaped 

rocks have, in general, the farthest runout distances. However, the 

real plate-shaped rock can also propagate large distances when it 

attains “wheel-rolling” modes. 

The simulations revealed that long runout was only possible only if the rock managed to 

become upright and rotate around the wheel axis. For the terrain parameterization, category 

Hard was used in the release zone (limestone and schist stratums), Medium Hard in the 

transition zone (rock debris) and Medium Soft in the runout zone (Meadow). With this 

ground parameterization it was possible to model the observed runout, as well as the 

skipping propagation mode before the rock stopped. As in reality the simulated rock loses 


otational stability at the end, wobbles and then topples on it side. Although the rock reached 

peak velocities of 50 m/s in the transition, mean velocities were considerably lower, approximately 

30 m/s. Rotational velocities in the runout zone are only 4 rad/s (approximately half 

a rotation per second). 

To demonstrate the importance of rock shape, elongated and equant rock shapes were tested 

with the same ground parameterization (Fig. 5). Elongated rocks had similar runout 

distances, but exhibited larger lateral spreading. Similar to the real (platy) form, many rocks 

stopped shortly after transgressing the transition zone. This was not the case for the equant 

shaped rocks: almost all rocks reached the runout zone and travelled slightly farther than 

both the platy and elongated forms. 

CONCLUSIONS 

Rockfall dynamics has traditionally relied on restitution coefficients to model rock-ground 

interactions. In contrast, non-smooth contact/impact models invoke friction laws to describe 

rock penetration and sliding. The frictional forces are applied on the surface of the rock, 

allowing the inclusion of rockshape into rockfall analysis. The main consequences of this new 

mechanical description are: 

Firstly, and most importantly, non-smooth models simulate the complexity of the impact 

configuration that lead to the multitude of different rock propagation modes that govern 

rockfall: skipping, sliding, jumping and rolling. Rotational speeds and jump heights can 

increase or decrease, depending on the impact configuration, a fact supported by numerous 

observations (Glover, 2014). Modeling this behavior is required to determine the onset of 

stopping, which in stochastic rebound models is based on ad-hoc and user-defined cutoff 

criteria to bring to a halt the unlimited process of elastic rock bouncing. 

The second consequence of non-smooth contact/impact mechanics is the possibility to model 

the mechanical properties of ground with consistent parameters associated with physical 

processes such as (1) material strength (μ min ,μ max ), (2) material hardening leading to ramping 

and rock ejection (κ), (3) viscous, strain-rate dependent material behavior (ν) and (4) 

material weakening/softening at release (β). With these parameters it is possible to characterize 

different ground types, varying from extra soft (highly dissipative ground that can stop all 

rock shapes) to extra hard (where only rocks of non-equant shapes can be stopped realistically). 

Non-smooth contact/impact mechanics offers the possibility of modeling tree and stump 

impact with new physical models. 

These two features of non-smooth contact/impact mechanics facilitate a realistic modeling 

of the variance of rockfall. It is necessary to understand this variance in order to develop 

better and more consistent methods of risk based hazard mapping. 

The hard contact approach is more computationally demanding that simple rebound 

methods. The calculation of a single trajectory requires approximately 1 second on a 

standard desk-top computer. However, the computation time can increase when rocks slide 

and remain in contact with the ground. In our applications we are finding that the statistical 

distributions do not vary after several thousand rocks are used. In fact sometimes only 


500 rocks are needed to obtain stable statistical values. For single slope domains the 

computational time is therefore not prohibitative; for large scale hazard mapping shadow 

angle methods will continue to provide the best computation times for large area analysis. 

The application of non-smooth contact/impact mechanics to the rockfall problem is now in its 

initial phase. Even at this stage, however, we believe the future impact on rockfall science, 

engineering and practice will be significant. One development that we are observing is an 

increased interest in rockfall experiments and examination of rockfall events in the field. 

Experimental measurements (e.g.to obtain rotational speed as a function of ground type) and 

field observations (e.g. documentation of scar lengths, rock shapes and jump lengths) can be 

used directly to calibrate non-smooth model parameters. Key is the modeling of rock 

jumping, rolling and sliding in three-dimensional terrain without introducing stochastic and 

non-physical rebound coefficients. 

REFERENCES 

- Dorren, L. K. (2003). A review of rockfall mechanics and modeling approaches. 

Progress in Physical Geography, 27(1), 67-87. 

- Leine, R. et al. (2015). Simulation of rockfall trajectories with consideration of rock shape, 

Multibody Syst Dyn (2014) 32:241–271, DOI 10.1007/s11044-013-9393-4. 

- Schweizer, A. (2015). Ein nichtglattes mechanisches Modell für Steinschlag, Dissertation, 

ETH Zürich, Switzerland. 

- Glover, J. (2014). Rock-shape and its role in rockfall dynamics, Doctoral Thesis, 

Durham University. 

- Christen, M. et al. (2012). Integral hazard management using a unified software 

environment: numerical simulation tool „RAMMS“ for gravitational natural hazards. 

In: G. - Koboltschnig, J. Hübl and J. Braun (Editors), Interpraevent, pp. 77 - 86. 

Volkwein et al. (2011). Rockfall characterisation and structural protection-a review, 

Natural Hazards and Earth System Sciences, 11. 

- An, B. and D. Tannant (2007), Discrete element method contact model for dynamic 

simulation of inelastic rock impact, Computers & Geosciences 33 (2007) 513–521. 

Thoeni, K. et al., (2014), A 3D discrete element modelling approach for rockfall analysis with 

drapery systems, International Journal of Rock Mechanics and Mining Sciences, 68, 107-119. 



A comparison of physical and computer-based debris 

flow modelling of a deflection structure at Illgraben, 

Switzerland 

Catherine Berger, Ph.D. 1 ; Marc Christen, Dipl. Ing. 2 ; Jürg Speerli, Prof. Ph.D. 3 ; Guido Lauber, Ph.D. 4 ; Melanie Ulrich, MSc FHO 

in Engineering 5 ; Brian W. McArdell, Ph.D. 6 

ABSTRACT 

On the fan of the Illgraben (Switzerland), settlements and infrastructure are endangered by 

large debris flows. A protection concept was therefore elaborated to partially deflect large 

debris flows into a forest for deposition and in order to reduce discharge and flow volume 

in the channel on the fan. Because deflection structures for debris flows are uncommon, 

a physical model was built at a scale of 1:60 to test functionality and optimize the design. 

Afterwards, the computer-based model RAMMS::DEBRIS FLOW was compared with the 

physical model results and to further analyze functionality and robustness especially for rare 

debris flows. In general, the numerical model runs with RAMMS showed a similar flow 

behavior as in the physical experiments, and separation effects of the deflection structure 

were confirmed. The standard Voellmy model was not able to reproduce the constant 

velocities found in the experiments before the dosing structure. Therefore, a version of the 

Voellmy model including cohesion was used and showed good agreement with the physical 

modelling. 

KEYWORDS 

Debris Flow; physical modelling; computer-based modelling; protection measures 

INTRODUCTION 

The Illgraben (Switzerland) experiences several debris flows per year and is one of the most 

active torrents in the Alps. The village Susten located on the Illgraben fan is endangered by 

large debris flows and protection measures are required. In the present protection concept, 

large debris flows shall be partially deflected into the Pfyn Forest for deposition. Smaller 

debris flows, not exceeding the channel capacity on the fan, will continue to flow into the 

River Rhone. Due to the complex processes, the functionality and geometry of the deflection 

structure were studied in a physical model. Afterwards, the initial and final geometry of the 

1 Emch+Berger AG, Spiez, Switzerland, spiez@emchberger.ch; today working for geo7 AG, Bern, Switzerland 

2 WSL Institute for Snow and Avalanche Research SLF, Davos, Switzerland 

3 HSR Hochschule für Technik, Rapperswil, Switzerland 

4 Emch+Berger AG, Spiez, Switzerland 

5 HSR Hochschule für Technik, Rapperswil, Switzerland 

6 WSL Institute for Forest, Snow and Landscape Research, Birmensdorf, Switzerland 



deflection structure were evaluated using the computational runout model RAMMS::DEBRIS 

FLOW to compare flow behavior and functionality of both model approaches. 

Figure 1: Overview of the Illgraben catchment and fan including physical model perimeter, instrumentation of the observation station, 

event scenarios and overview on the performed modellings (marked with “x” where the method was applied). 

STUDY SITE 

The Illgraben catchment (10.3 km 2 , Figure 1) located in southwestern Switzerland in the 

community of Leuk extends from 2716 m asl at the Illhorn to the outlet of the Illgraben into 

the Rhone River at 610 m asl. The climate is temperate-humid and influenced by the rain 

shadow effect within an interalpine valley which generates relatively low annual precipitation. 

Several debris flows occur every year and are initiated in the steep sub-catchment 

(4.6 km 2 ) on the north face of the Illhorn where abundant sediment is available (Berger et 

al., 2011) and annual sediment transport into the Rhone River is about 70,000 m³ (Badoux et 

al., 2009). The Illgraben fan has a radius of about 2 km with channel slopes of 6 % to 10 %. 

In the catchment and upper third of the fan, the channel is deeply incised and normally has 

a double-trapezoidal shape with a low and high water zone. On the lowest part of the fan, a 

single trapezoidal-shaped channel is observed and discharge capacity is reduced considerably. 

About 30 check dams stabilize the channel on the fan and of the lower trunk channel in the 

catchment. The village Susten and a settlement below the fan apex are located on the right 


side of the fan and are partially within the highly (red) and intermediate (blue) danger areas 

according to Swiss guidelines for hazard zonation (BUWAL, 1998). The left side of the fan is 

covered by the Pfyn Forest which is part of the Federal Inventory of Landscapes and Natural 

Monuments of National Importance of Switzerland. A debris flow observation station is run 

by the Swiss Federal Institute for Forest, Snow and Landscape Research WSL since 2000 and 

a large variety of flow types have been observed (McArdell et al., 2007; Badoux et al., 2009). 

DESCRIPTION OF PROTECTION MEASURES 

After defining event scenarios (Figure 1), hazard maps and initial protection concepts were 

elaborated. Due to very frequent debris flows and large expected volumes, retention of the 

debris in the catchment is inappropriate. Conveying very large debris flows into the River 

Rhone by enlarging the present channel may cause undesired backwater effects in the Rhone 

and stopping of debris flows on the fan, and consequent outbreak and flooding of settlements 

and infrastructure could be expected as a consequence. Therefore, we focused on a partial 

deflection of large, endangering debris flows at the fan apex in order to reduce peak discharge 

in the present channel. Smaller debris flows not exceeding the capacity of the channel on the 

fan will remain in the channel and flow into the River Rhone. 

Within the protection concept, all debris flows are led to the deflection structure through a 

broad channel (discharge area 300 m 2 ) stabilized with check dams (Figure 2). The deflection 

structure is located in a tight curve (radius about 30 m) and consists of a crossover duct / 

breach joining the channel with the Pfyn Forest and a dosing dam. Because of their inertia 

and front height, large debris flows are intended to overshoot the curve and flow into the 

crossover duct and are thereby partially deflected into the forest for deposition. Smaller debris 

flows pass the dosing structure and are conveyed to the present streambed through a smaller 

channel (discharge area 45 m 2 ) stabilized with a slightly steeper block ramp and check dams. 

A simplified geometry with trapezoidal channel shapes and channel slopes at 6 % were used 

for modelling and initial planning. Minimum channel capacity on the fan and back-calculations 

of events where backwater effects in the River Rhone were dominant, were used for 

dimensioning and to estimate the activation point of the deflection structure: annual and 

frequent debris flows are conveyed back to the present channel and flow into the River 

Rhone, whereas for medium events (Figure 1) the deflection is activated. The activation 

threshold was set at a discharge area of 45 m 2 and equals the capacity of the lower trapezoid 

in the broad channel or of the channel returning to the present streambed. 

The volume of debris-flow material deflected into the Pfyn Forest will increase with front 

speed and front height. However, it is important that dosage of debris flows is governed by 

maximum discharge or wetted area and a direct control on the deflected volume is not 

possible. Therefore, debris flows with a comparatively small front or several smaller surges 

would not be deflected and could transport large debris volumes into the River Rhone and 

backwater effects therefore cannot be excluded. 


PHYSICAL MODELLING OF THE DEFLECTION 

Because deflection structures for debris flows are uncommon, a physical model was built at 

a scale of 1:60 (Figure 2) to test functionality and optimize the design (Berger et al., 2014). 

Due to the large event volumes and physical limitations, modelling was limited to events 

within the activation range of the deflection structure and therefore to frequent events 

(redirection of the flows to the present channel / no activation) and medium events (deflection 

activated, Figure 1). For these events, target values based on event scenarios, derived 

from measured events at the observation station and field estimates were defined for model 

Figure 2: Initial model (left) and final model (right) of the deflection structure with indication of the structural elements and performance 

at medium events. Circle A: impact and surging on the left wall of the dosing structure, circle B: superelevation of the surface in the 

curve. 


calibration and analysis of the model results. Sediment mixture, volume and discharge were 

scaled using Froude similarity, and flow properties (height, velocity, volume deflected) were 

measured at several locations. Reduced volumes (Figure 1) were used because the tail of a 

debris flow is not controlling the main functionality of the deflection structure, and target 

peak discharges were obtained with smaller volumes. 

For the initial geometry (Figure 2), the following behavior was observed: 

– Smaller debris flows (frequent events) below the activation threshold of the deflection 

were reverted to the present channel without overtopping of the banks downstream of the 

dosing structure. However, debris material was deposited in the breach by spillover. 

– Larger debris flows (medium events) activated the deflection. However, the flow impacted 

and surged on the left wall of the dosing structure (Figure 2, circle A). 

Improvements were therefore needed, mainly with respect to flow behavior and 

separation effects. Consequently, the geometry was optimized iteratively in six model setups 

and a total of 40 experiments, to optimize the functionality and robustness of the structure. 

In the final geometry, the following features were changed compared to the initial 

model (Figure 2): 

– The breach was rotated downstream to reduce impact on the dosing structure. 

– The leading part of the left side of the dosing structure was designed without banks to 

reduce impact and surging. 

– The channel bed in the breach was raised by 1 m to account for superelevation and obtain 

less deposition in the breach during events below the activation threshold of the deflection. 

COMPUTER-BASED MODELLING OF THE DEFLECTION 

A computational model was used for comparison with the physical modelling and to analyze 

functionality and robustness for rare and extreme debris flows which could not be modelled 

in the laboratory. RAMMS::DEBRIS FLOW is a 2D numerical simulation tool developed by 

WSL. The core of the program is a second-order numerical solution of the depth-averaged 

equations of motion for granular flows (Christen et al., 2012). 

The debris rheology was described by an extended Voellmy-model which includes cohesion 

(Bartelt et al., 2015). In this model shear stress S is calculated according to the relation, 

S N 

 

N 

 

N 

0 

1 

N 

exp 

0 

1 

2 

U 

N 

0 

g 

 

where N is the basal normal stress, ║U║ is the norm of the velocity U, ρ the debris flow density, 

g the acceleration due to gravity and N 0 is the cohesion. When N 0 =0, the model reduces to 

the standard two-parameter (μ, ξ) Voellmy model. When μ=0, the model describes an ideal 

plastic material where N 0 acts similar to a yield stress, see Figure 3. Moreover, the stress never 

exceeds N 0 for large flow heights. 

The model was calibrated using data from the Illgraben observation station and target values 

defined for the physical model and 3-point hydrographs were defined at the inlet into the 

model perimeter. Model runs were performed at two spatial extents and grid resolutions: 


Figure 3: Shear vs. normal stress relation used to model debris flow motion in confined and unconfined terrain. In the channel we set N0 

= 500 Pa and μ = 0.00. This produces the ideal plastic behavior depicted above. In unconfined flow, wet set μ =0.07. Note the similarity 

to actual debris flow measurements shown in McArdell et al. (2007). 

a smaller extent (1 m grid) representing the perimeter of the physical model for the initial 

and final geometry of the deflection and a larger extent (2 m grid) covering the fan apex at 

a grid resolution of 2 m using the final geometry of the deflection and present topography 

without deflection. A summary of modelled scenarios and applied spatial extents is listed in 

Figure 1. For each model run, maximum flow height, maximum velocity and final deposition 

were exported to ESRI ArcGIS for analysis. Furthermore, cross-sections at different locations 

were extracted in RAMMS::DEBRIS FLOW to estimate maximum discharge from maximum 

values of flow height and velocity (Figure 4) and final deposition was used to estimate the 

deflected debris volume. 

RESULTS AND DISCUSSION 

The standard Voellmy model was not able to reproduce the constant velocities found in the 

experiments before the dosing structure. The modified Voellmy model (including cohesion) 

predicted both the debris flow velocities and flow heights with the parameter combination 

μ = 0.00, ξ = 600 m/s 2 and N0 = 500 Pa. This suggests that in the channel the debris is fluidized 

and the flow is well lubricated (μ = 0). The cohesion describes the visco-plastic behavior of the 

fluidized debris mixture. This approach could not be applied to model debris motion outside 

the channel. To take into account unconfined, open terrain outside the channel it was 

necessary to increase the Coulomb friction from μ=0.00 (visco-plastic) to μ = 0.07, thus 

introducing a type of plastic hardening into the flow rheology. This allows plausible modeling 

of the flow stopping behavior outside the channel. 

In general, the simulations with RAMMS showed a similar flow behavior as in the physical 

experiments. The reduction of debris-flow discharge due to the diversion structure also 


Figure 4: Maximum flow height and maximum velocity for the initial and final geometry of the deflection using three event scenarios and 

indicating maximum discharge at cross-sections (numbers in red). See Figure 5 for the color legend. 

provided plausible results for events with discharges > 950 m³/s and the channel capacity 

below the deflection structure was not exceeded (Figure 4). The impact of the flow front on 

the left wall of the dosing structure was reduced by the improvements from the initial to the 

final geometry. As in the laboratory experiments, superelevation of the debris flow surface in 

the narrow curve was observed (see Figures 2 and 4) and overtopping into the breach 

occurred at frequent events (Q = 250 m³/s). However, discharge at the edge of the breach was 

reduced from 45 to 25 m³/s (see Figure 4) due to the raised channel surface in the final 

geometry. 

With respect to the larger spatial extent (Figure 5), overtopping at the planned location of the 

breach and deposition in the Pfyn Forest was observed in the numerical model runs on the 

present topography without deflection. This partly corresponds with recent observations of 

debris flow deposits and indicates a natural tendency of debris flows to break out at the 

channel curve and flow towards Pfyn Forest. However, the magnitude and extend of the 

outbreak in the computer model partly is attributed to overtopping, as explained in Berger et 

al. (2012). Using the final geometry of the deflection, about 115,000 m³ or about 25 % of the 

initial volume were deposited in the Pfyn Forest. The representation of the deflection 

structure was good and showed a similar behavior to the results with the 1 m grid. 


Figure 5: Maximum flow height on the large model extent with and without the deflection structure for a rare event and indication of 

model settings. 


Functionality and geometry of the deflection structure in-planning at the apex of the 

Illgraben fan were tested and improved using physical scale-model experiments and afterwards 

verified using the debris-flow runout simulation model RAMMS::DEBRIS FLOW. 

In general, the functionality of the deflection structure, i.e. reducing peak discharge and the 

total volume of large debris flows, could be demonstrated. However, the work also indicates 

that maintenance of the structure is essential to minimize separation effects of the deflection 

due to overtopping and deposition of debris in the breach during smaller events. This is 

important to ensure that the deflection structure only directly controls maximum discharge 

of debris flows. Because events with a comparatively small discharge yet a large total volume 

could pass the deflection structure and might cause backwater effects and flooding of 

settlements and infrastructure on the fan, overall risk management remains an essential part 

of the mitigation, including e.g. periodical observation of the catchment and the channel, 

land use planning or emergency plans. 

Using two different models, further questions arise on the limitations of both approaches, 

model artifacts in the results and finally on the representation of nature. However, no direct 

comparison with nature is available and corroborating results only can be achieved in a 

multi-method approach using different models and estimates. 


REFERENCES 

- Badoux A., Graf C., Rhyner J., Kuntner R., McArdell, B.W. (2009). A debris-flow alarm 

system for the Alpine Illgraben catchment: design and performance. Natural Hazards 49(3): 

517-539. 

- Bartelt P., Vera Valero C., Feistl T., Christen M., Bühler Y., Buser O. (2015), Modelling 

cohesion in snow avalanche flow, Journal of Glaciology, Vol. 61, No. 229, 2015 doi: 

10.3189/2015JoG14J126. 

- Berger C., McArdell B.W., Schlunegger F. (2011). Sediment transfer patterns at the Illgraben 

catchment, Switzerland: Implications for the scales of debris flow activities. Geomorphology 

125: 421-432. 

- Berger C., McArdell B.W., Lauber G. (2012). Murgangmodellierung im Illgraben, Schweiz, 

mit dem numerischen 2D-Modell RAMMS. Interpraevent Conference Proceedings. 

- Berger C., Ulrich M., Lauber G., Speerli J. (2014). Hochwasserschutz Illgraben: Ausleitbauwerk 

für grosse Murgänge. Tagungsband Internationales Symposium Wasser- und Flussbau 

im Alpenraum. 

- BWW, BRP, BUWAL (1998). Empfehlungen, Berücksichtigung der Hochwassergefahren bei 

raumwirksamen Tätigkeiten. Biel, Switzerland. 

- Christen M., Bühler Y., Bartelt P., Leine R., Glover J., Schweizer A., Graf C., McArdell B.W., 

Gerber W., Deubelbeiss Y., Feistl T., Volkwein A. (2012). Integral hazard management using a 

unified software environment: numerical simulation tool „RAMMS“ for gravitational natural 

hazards. Interpraevent Conference Proceedings. 

- McArdell B.W., Bartelt P., Kowalski J. (2007). Field observations of basal forces and 

fluid pore pressure in a debris flow. Geophys. Res. Lett. 34, L07406, 

doi:10.1029/2006GL029183: 4 pp. 



Exploiting damage claim records of public insurance 

companies for buildings to increase knowledge about 

the occurrence of overland flow in Switzerland 

Daniel Benjamin Bernet, MSc 1,2 ; Rolf Weingartner, Prof. Dr. 2 ; Volker Prasuhn, Dr. 3 

ABSTRACT 

Overland flow is difficult to assess because direct data is missing. As Swiss public insurance 

companies for buildings cover overland flow along with other hazards, we exploited their 

records to investigate the occurrence of overland flow indirectly. With a novel classification 

scheme, it is possible for the first time, to distinguish claims related to overland flow from 

inundations caused by watercourses. We analyzed gapless data records from 1991 to 2013 of 

the cantons Neuchâtel, Fribourg, Nidwalden and Graubünden, each representing a different 

typical Swiss landscape. Altogether, roughly 40-50 % of the damage claims can be associated 

with overland flow, which account for 20-30 % of total loss in that period. However, the 

inter-cantonal differences are large and reflect the embedment of overland flow in the 

landscape’s geographic setting. Finally, looking at averages per km 2 and year, we found that 

pre-alpine Fribourg is affected most by overland flow. As an outlook, we are confident that 

the presented methodology can be used to start studying overland flow from a more 

process-oriented perspective. 

KEYWORDS 

overland flow; inundation and flooding; damage claims; public insurance companies for 

buildings 

INTRODUCTION 

Post-damage analyses in the field of flood hydrology highlight that not only overtopping 

rivers and lakes cause a substantial amount of loss. Reportedly, about fifty percent of all 

damages to buildings are caused by overland flow (Bezzola and Hegg, 2008). Overland flow 

propagates over the land surface as thin sheet flow or anastomosing braids of rivulets and 

trickles, until the flow reaches or is concentrated into recognizable channels (Chow et al., 

1988; Ward and Robinson, 2000). Thus, overland flow is not constrained to a riverbed, but 

occurs diffusely in the landscape. Furthermore, overland flow is generally associated with a 

short response time and practically no advance warning, which makes it difficult to observe 

and study the process directly. Maybe that is the reason why, in spite of the existence of 

several practical tools to assess the hazard of overland flow (Kipfer et al., 2012; Rüttimann 

1 University of Bern, Bern, SWITZERLAND, daniel.bernet@giub.unibe.ch 

2 Institute of Geography & Oeschger Centre for Climate Change Research & Mobiliar Lab for Natural Risks, University of Bern, Bern, 

Switzerland 

3 Agroscope, Institute for Sustainability Sciences ISS, Zurich, Switzerland 



and Egli, 2010; Bernet, 2013), little is known about where and when overland flow occurred 

in the past and will occur in the future. 

As there is no direct information about the occurrence of overland flow in space and time, 

traces of the flow’s propagation can be found wherever the process has caused detectable 

damages, or claims thereof. These can be used as a proxy for the occurrence of overland flow. 

Data sources implicitly containing such information are house owners’ damage claims 

recorded by Swiss Public Insurance Companies for Buildings (PICB). 

Our overarching goal is to improve the process understanding of overland flow. In this paper, 

we want to demonstrate that damage claim records can provide very useful, indirect 

information about the occurrence of overland flow in space and time. Furthermore, by 

looking at the total number of claims as well as total loss related to overland flow and 

inundation from watercourses respectively, we want to highlight the relative relevance of 

these processes. For these purposes, we have analyzed damage claim records of the PICB 

of Neuchâtel (NE), Fribourg (FR), Nidwalden (NW) and Graubünden (GR). These cantons 

approved our data enquiry and are chosen for this pilot study, as they cover different 

landscape patterns typical for the whole Switzerland. 

TERMS AND DEFINITIONS 

Terms used by scientists, insurers or practitioners to describe processes that can lead to water 

related damages to buildings may differ. Thus, hereafter, some important terms are defined: 

– In accordance to the definition above, overland flow is understood as surface runoff that 

propagates unchannelled over the land surface until it reaches the next river or lake. 

– Damages to buildings caused by water entering the structure at ground level (this excludes 

penetrating groundwater, backwater from the sewer system and rainfall directly entering 

the building through its envelope) are, hereafter, referred to as water damages. 

– For reasons of readability, we abbreviate floods and inundations from rivers and lakes, 

explicitly excluding overland flow, by simply referring to inundations from watercourses. 

Thus, the term watercourse always refers to both rivers and lakes. 

– In this paper, we make use of two different flood hazard maps (Swiss flood hazard maps 

and Aquaprotect, see methodology). The former includes assessment perimeters, whereas 

the latter does not. Both indicate areas that are at hazard complemented by hazard-free 

zones. All hazardous areas, regardless of the hazard level and the map source, are referred 

to as flood zones. 

DATA 

In Switzerland, PICB are present in 19 out of the total 26 cantons, within which they each 

hold a monopoly position. In addition, it is (with a few exceptions) mandatory for all house 

owners to insure their buildings against natural hazards including avalanches, snow pressure 

and -load, hail, storm, land- and rockslides, falling rocks and inundation processes. Concerning 

the latter, hazards associated with water entering the building at ground level are covered. 

Consequently, all damages caused by overland flow are insured and recorded by one single 


institution. Unfortunately, PICB generally do not distinguish different inundation causes and, 

thus, the responsible process for the claimed damages, namely overland flow or inundation 

from watercourses, must be identified. 

DATA HARMONIZATION 

For the present pilot study, we have analyzed records of house owners’ damage claims related 

to flood processes of the PICB of Neuchâtel (NE), Fribourg (FR), Nidwalden (NW) and 

Graubünden (GR), representing the most typical landscapes of Switzerland. The data 

delivered by the different PICB were quite heterogeneous and, thus, needed to be harmonized. 

The most important variables used in this study were the address, geocode, total loss, 

processing status of the damage claim, as well as the occurrence date of the claimed damage: 

– All four PICB provided geocoded damage claims. Claims with missing spatial reference 

were geocoded using the provided addresses, whenever possible. 

– The loss of each damage claim was calculated by adding the payout and the corresponding 

deductible. Then, the losses were indexed to 2014. Note that in case of a covered damage, 

all PICB calculate the payout according to the reinstatement costs (value as new), except 

Fribourg. In the latter canton, reinstatement costs are determined according to (for us 

untraceable) depreciated values. 

– The damage dates were checked for plausibility manually. The last year with complete 

records is 2013 for all four PICB. The complete records start in 1983 in Fribourg (31 y), 

1987 in Nidwalden (27 y), 1988 in Neuchâtel (26 y) and 1991 in Graubünden (23 y), 

respectively. 

– The status of the damage claims were categorized commonly. For this paper, only completed 

damage claims, i.e. claims with an actual payout, are considered. This ensures that only 

damage claims of an insured risk (i.e. overland flow and inundation from watercourses) 

are analyzed. 

Overall, the data comprises 15’200 inundation related damage claims, of which 11’239 are 

justified and geocoded. For analyses concerning each canton separately (i.e. the compilation 

of percentiles, see methodology), the geocoded, justified claims are used. For the comparison 

between the cantons (i.e. number and total loss of the different classes per canton, see 

results), only the overlapping period from 1991 to 2013 is considered (23 y), counting a total 

of 9’451 damage claims. 

METHODOLOGY 

To differentiate damage claims that are associated either with overland flow or inundations 

from watercourses, we have developed a classification scheme (Figure 1). The scheme is not 

directly applicable to single damage claims, as it neglects important influencing factors such as 

micro and macro topography, the circumstances of a loss, etc. However, by applying it to a 

large dataset and computing summary statistics, the methodology is robust and produces 

productive results. 


Figure 1: Each geocoded damage claim is categorized according to the displayed classification scheme. The white rhombuses each 

represents a yes or no test, checking whether the coordinates of the affected building are within 25 m of the corresponding spatial 

object (that takes the uncertainty of the buildings represented as point objects into account). Only coordinates located outside of the 

flood zones (hazard map and Aquaprotect) reach either of the green rhombuses. For each of these claims, the distance d between the 

claim’s point location and the closest watercourse is compared to the 25th, 50th, 75th and the 99th percentiles of the corresponding 

cumulative distribution of the claims within flood zones (Figure 2, blue curve and percentiles). In this way, all claims outside of the 

flood zones are classified depending on the percentile range they fall in. All possible paths of the classification scheme are 

schematically shown in Figure 3 the colors of the ellipses are used throughout the paper to denote overland flow (A: red and B: orange), 

inundation from watercourses (E: dark blue and D: light blue) or damage claims that could not be associated with either process 

(C: gray). 

The scheme makes use of existing flood hazard maps, as mentioned in the introduction. 

The rationale of the scheme is to use these maps directly for claims located within indicated 

flood zones. From this claim subset, we can then infer characteristics and use them for the 

classification of claims outside of the flood zones. For the latter claims, we make the following 

assumption: The distance to the closest watercourse from a flooded building determines how 

likely that particular building has been affected by inundation from a watercourse or by 

overland flow. Understandably, if the building is located close to a watercourse, the responsible 

process has most likely been inundation from that particular watercourse. On contrary, 

if the building is located far away from any watercourse, overland flow has most likely caused 

the damage. 


The Swiss flood hazard maps provide detailed information about hazardous zones related to 

inundation from watercourses (Petrascheck and Loat, 1997). However, these maps were 

compiled for a predefined perimeter, which is generally constricted to construction zones. 

Moreover, the hazard maps were compiled with differing methodologies in each canton. 

To smooth out these differences and to increase the spatial coverage, Aquaprotect, a flood 

zone map covering the whole Switzerland, provided by the Swiss Federal Office for the 

Environment (FOEN), is used in addition to the hazard maps. Assessed with a coarse but 

standardized methodology, Aquaprotect indicates flood zones of the larger watercourses 

associated with return periods of 50, 100, 250 and 500 years, while neglecting existing flood 

control measures. The dataset referring to a return period of 250 years is chosen, as the Swiss 

flood hazard maps consider a return period of maximal 300 years. Although it is possible that 

overland flow occurs within a mapped flood zone, it can be assumed that the predominant 

damage causing process in these areas are inundations from watercourses. Thus, the 

distribution of affected buildings located within flood zones in relation to the closest watercourse 

is used to classify damage claims located outside of these flood zones. Thereby we 

assume, that the patterns of damages caused by inundations from watercourses within the 

mapped flood zones are the same for damages located outside of these zones. 

Clearly, the zone of influence of a watercourse depends on the geographical and geological 

properties of the landscape, as well as on the size of the river. For that reason, we compiled 

distance distributions for each canton and river size class separately. The latter is feasible, as 

the FOEN provides a dataset (referred to as FLOZ) that can be linked to the Swiss hydrological 

network of the product VECTOR25 provided by swisstopo. In that way, each river section is 

assigned to the corresponding Strahler Stream Order SSO (Strahler, 1964), which can be used 

as a proxy for the river’s size. The SSO takes discrete numbers, which range from 1 to 9 in 

Switzerland. According to Weissmann et al. (2009) a SSO of 1 refers to small, 2-3 to medium 

and 4-9 to large Swiss rivers. 

Figure 2 displays the cumulative distribution of the damage claims within or outside of the 

mapped flood zones, depending on the distance to the next river of a certain SSO. The blue 

curve, corresponding to the claims within the flood zones, rises sharply within the vicinity of 

watercourses but levels off quickly with increasing distance. On the other hand, the green 

curve, referring to damage claims outside the flood zones, rises more gradually and levels off 

at a much higher distance. We interpret this behavior as the superposition of damages caused 

by inundation from watercourses (small distances) and by overland flow (farther away from 

the watercourses). To obtain an objective way to disentangle these processes we take the 

distance to the next watercourse that correspond to the 25th, 50th, 75th and the 99th 

percentiles of the cumulated damage claims located within the flood zones (Figure 2). 

As mentioned before, such curves are compiled for each canton and river class separately. 


Figure 2: Exemplary cumulative distribution of all damage claims of Graubünden within flood zones that are closest to a medium sized 

river (SSO: 2-3), plotted against the distance to the corresponding river (blue curve). As a comparison, the cumulative distribution of all 

damages outside of the flood zones are displayed (green curve). Additionally, the corresponding percentiles (25th, 50th, 75th and 99th) 

of both curves are indicated (dotted lines). Note that for each of the three SSO classes (1-2: small rivers; 2-3: medium rivers; 4-9: large 

rivers) as well as for each canton, the percentiles of the claims within flood zones are computed and applied separately to reflect the 

different geographical and hydrological setting of each canton. 

The coordinate pair of each damage claim is classified using the scheme shown in Figure 1. 

As a depiction thereof, all possible cases are illustrated in Figure 3. We distinguish five classes, 

each referring to the dominant process responsible for the claimed water damage with a 

qualitative indication of how certain the classification is: 

– A: most likely overland flow 

– B: likely overland flow 

– C: damage causing process uncertain 

– D: likely inundation from a watercourse 

– E: most likely inundation from a watercourse 

RESULTS 

Our analyses show that 43 % of all claims are likely and most likely associated with overland 

flow and 47 % with inundations from watercourses (Figure 4). The remaining 10 % cannot 

be associated with either process. Looking at the numbers from the individual cantons, it 

becomes apparent that the fractions differ greatly from canton to canton. In Neuchâtel and 

Fribourg, more than half of the damage claims relate to overland flow. However, in Fribourg 

the classification is associated with a larger uncertainty, i.e. 18 % could not be classified as 

either overland flow or inundation from watercourses. 


Figure 3: Each displayed point corresponds to a distinct path in the classification scheme (Figure 1) and denotes the corresponding 

process responsible for the caused damage (A: most likely overland flow; B: likely overland flow; C: damage causing process uncertain; 

D: likely inundation from a watercourse; E: most likely inundation from a watercourse). The arrows show how the percentiles of the 

cumulative distribution of damage claims within flood zones to the next watercourse are applied in practice. Furthermore, the display 

indicates that the percentiles are computed for each river class formed by the Strahler Stream Orders (SSO; 1-2: small rivers; 2-3: 

medium rivers; 4-9: large rivers) as well as for each canton separately. 

In Graubünden 59 % of the claims are associated with inundation from watercourses, while 

more than every third claim relates to overland flow. The classification in Nidwalden shows a 

completely different picture. The vast share of 92 % of all claims is caused most likely by 

inundations from watercourses, while the remaining classes each account for only a few 

percent of all claims. 

Although Nidwalden has the lowest amount of damage claims amongst the cantons in 

numbers, the total loss is almost as high as in Graubünden in the same period (Figure 4). 

The opposite is the case for Fribourg. It ranks highest amongst the cantons in terms of 

number of claims, but has a relatively low associated cumulated loss. Neuchâtel, although 

ranking in the same range as Nidwalden and Graubünden in terms of damage claim numbers, 

it had to cope with the smallest amount of loss. 

To get an idea about the density of these processes in space and time, the numbers can be 

related to the size of each canton (Table 1). For this simple assessment, we have neglected 

the vulnerability, elements at risk and other possibly relevant factors: In Fribourg, damage 

claims are most frequently associated with overland flow and cause a yearly average loss of 

CHF 451.00 per km 2 , followed by Neuchâtel. Although in Graubünden overland flow 


accounts for almost 40 % of all damage 

claims (Figure 4) the density, both in terms of 

occurrence but also in terms of yearly loss per 

area, is by far the lowest. Graubünden also 

ranks last when looking at inundations from 

watercourses, however unlike the average 

occurrence of overland flow, the loss density 

is in the same order as in Fribourg, and 

Neuchâtel. Nidwalden, on the other hand, 

clearly stands out. With more than 0.2 

damage claims amounting to more than 7000 

CHF per km 2 and year, the values are by far 

the highest. 

Figure 4: Relative distributions of the number of damage claims, 

as well as total water damage loss (indexed as per 2014) for each 

canton separately and for the cantons in total. Note that the 

number of claims (n) as well as the total loss (s) in million Swiss 

Francs correspond to the statistics of the overlapping period from 

1991 to 2013. 

DISCUSSION 

Based on several case studies, it is stated that 

about half of all (justified) water damage 

claims are caused by overland flow (e.g. 

Bezzola and Hegg, 2008). Our analyses based 

on gapless claim records of the last 23 years 

show that with 43 %, less than half of the 

claims can be associated clearly with overland 

flow. Nevertheless, the share is highly 

significant. In terms of loss, it also underlines 

previous case studies and assumptions 

respectively, revealing that on average, the 

loss associated with overland flow is lower 

than loss associated with inundation from 

watercourses. However, our results also show 

that this is not the case for all areas. The 

regional differences, as illustrated by Figure 

4, can be explained by the different landscapes 

of the studied cantons. The geographical 

and geological patterns are reflected by 

the results: 

– In Fribourg, representative for the Pre-Alps, overland flow occurs most frequently 

(Table 1), which can also be explained by the geological features, favoring overland flow 

(Weingartner, 1999). Further, the prevalence of many underground rivers that go partly 

back to extensive melioration in the past century promote overland flow. Inundation from 

watercourses are frequent as well, but less intense than in more alpine regions. 


– As expected, in Neuchâtel with its typical karstic landscape, damage claims are more 

frequently caused by overland flow than by inundations from watercourses. 

– Nidwalden, a canton with steep slopes, mostly less permeable soils, resulting in a dense 

river network, together with densely populated valley floors, is exposed heavily to 

inundation processes from watercourses (Figure 4). Although, overland flow may be 

responsible for certain damage claims, the dominant process is inundation from watercourses 

by far. 

– Graubünden, the largest canton in Switzerland, is mountainous and overall loosely 

populated. Thus, the relative occurrence of claims related to inundations from watercourses, 

but even more so for overland flow, are very low. However, the associated losses 

are very high due to the devastating floods of mountain torrents. 

Table 1: Yearly rates at which each canton is affected by overland flow (class A + B + ½ C) or inundation from watercourses (class ½ C 

+ D + E, see also Figure 4), obtained by dividing the absolute numbers by the area of the respective canton and the record length of 23 

years (1991 -2013). The rows are ordered according to the yearly rate of buildings affected by overland flow. 

CONCLUSIONS 

Indubitably, overland flow causes frequent damages to buildings in Switzerland. For the first 

time, we can support this with a gapless data record covering representative areas of 

Switzerland, as we have collected and harmonized damage claim records of four Swiss PICB 

each representative for typical Swiss regions and covering the last 23 years. Due to the large 

dataset, the numbers are robust. However, it has to be noted that our novel methodology to 

disentangle water damages to buildings only works for large numbers and cannot be applied 

to single damage claims. 

We have demonstrated that it is feasible and worthwhile to analyze damage claim data from 

public insurance companies, even more so, as it is, to our best knowledge, the only source 

that indicates overland flow over a large part of Switzerland within a longer period. The next 

step forward is to use the disentangled dataset, in order to analyze spatial and temporal 

patterns of the occurrence of overland flow. In this way, we can move towards more 

process-based investigations that are required to better understand and, ultimately predict, 

overland flow in the future. 


ACKNOWLEDGMENTS 

The authors would like to thank the Public Insurance Companies for Buildings of the cantons 

Neuchâtel, Fribourg, Nidwalden and Graubünden for their support and for providing the data 

used in this study. Moreover, we would like to thank the two anonymous referees for their 

valuable comments and suggestions. 

REFERENCES 

- Bernet, L.: Gebäudeschäden durch Schlammeintrag infolge von Bodenerosion durch Wasser 

in der Schweiz: Entwicklung eines GIS-Modells zur Gefahrenabschätzung auf der Basis der 

Erosionsrisikokarte, Masterarbeit, Geographisches Institut der Universität Bern, Universität 

Bern, Bern, 2013. 

- Bezzola, G. R. and Hegg, C. (Eds.): Ereignisanalyse Hochwasser 2005: Teil 2 - Analyse von 

Prozessen, Massnahmen und Gefahrengrundlagen, Umweltwissen, Nr. 0825, Bundesamt für 

Umwelt (BAFU); Eidgenössische Forschungsanstalt (WSL), Bern, 2008. 

- Chow, V. T., Maidment, D. R., and Mays, L. W.: Applied hydrology, McGraw-Hill series in 

water resources and environmental engineering, McGraw-Hill, New York, xiii, 572, 1988. 

- Kipfer, A., Kienholz, C., and Liener, S.: Ein neuer Ansatz zur Modellierung von Oberflächenabfluss, 

in: INTERPRAEVENT 2012 - Conference Proceedings, Koboltschnig, G., Hübl, 

J., Braun, J. (Eds.), 12th Interpraevent Congress, Grenoble, France, 23-26 April 2012, 

International Research Society INTERPRAEVENT, Klagenfurt, 179–189, 2012. 

- Petrascheck, A. and Loat, R.: Berücksichtigung der Hochwassergefahren bei raumwirksamen 

Tätigkeiten: Empfehlungen 1997, Bundesamt für Wasserwirtschaft (BWW); Bundesamt für 

Raumplanung (BRP); Bundesamt für Umwelt, Wald und Landschaft (BUWAL), Biel, 1997. 

- Rüttimann, D. and Egli, T.: Wegleitung Punktuelle Gefahrenabklärung Oberflächenwasser, 

Naturgefahrenkommission des Kantons St. Gallen, 2010. 

- Strahler, A. N.: Quantitative geomorphology of drainage basin and channel networks, 

Handbook of Applied Hydrology, 1964. 

- Ward, R. C. and Robinson, M.: Principles of hydrology, 4th ed, McGraw-Hill, London, 

Burr Ridge, Ill., xiv, 450, 2000. 

- Weingartner, R.: Regionalhydrologische Analysen–Grundlagen und Anwendungen, Beiträge 

zur Hydrologie der Schweiz, Nr. 37, Schweizerische Gesellschaft für Hydrologie und 

Limnologie, Bern, 1999. 

- Weissmann, H. Z., Könitzer, C., and Bertiller, A.: Strukturen der Fliessgewässer in der 

Schweiz: Zustand von Sohle, Ufer und Umland (Ökomorphologie); Ergebnisse der ökomorphologischen 

Kartierung. Stand: April 2009, Bundesamt für Umwelt (BAFU), Bern, 

Umwelt-Zustand, Nr. 0926, 100 S., 2009. 



Distributed acoustic monitoring to secure transport 

infrastructure against natural hazards - requirements 

and new developments 

Michael Brauner, Dr DI 1 ; Arnold Kogelnig, Dr DDI 2 ; Ulrich Koenig, DI 2 ; Günther Neunteufel, Ing 3 ; Hanna Schilcher 2 

ABSTRACT 

The Austrian Federal Railways (OEBB) rail-network operate several trans-European 

connections, thereof approximately 20% are exposed to natural hazards. For trains, obstacles 

in the tracks, damage to the track bed and direct train hits are the most relevant hazard 

scenarios. Classical permanent protections structures are sometimes hard to realize due to 

technical reasons and the limited time windows available for construction work. Therefore 

detection technologies gain more importance as an alternative to classical protection 

measures. Distributed acoustic sensing (DAS) with fiber optic cables is in use worldwide for 

monitoring pipelines or e.g. securing industrial and military facilities. This study evaluates if 

reliable detection of hazardous processes in the track area is possible by using existing fiber 

optic communication cables. We show that processes such as rock fall or tree-throws can be 

detected well, but due to the high sensitivity of the monitoring system, interfering signals still 

produce a relative high number of false alarms. 

KEYWORDS 

railway; rock fall detection; DAS monitoring; OTDR fiber cable; natural hazard 

INTRODUCTION 

In Austria the Austrian Federal Railways (OEBB) operate several important trans-European 

railway routes. Thereof, especially transalpine railway lines show long slope-traverses, which 

are often exposed to gravitational processes such as rock falls. Protection nets and frequent 

rock wall clearing and track inspection typically cover the resulting protection needs. 

However, in implementing these expensive measures, the demand for high track availability 

and thus the small number of maintenance windows is becoming a challenge. Therefore, the 

future focus must be on cost effective alternatives with low maintenance intervals. 

One approach is to directly monitor protection nets or install slider fences. With them rock 

fall activity is detected promptly and maintenance work or even track closure can be arranged 

in time (Schorno et al 2011). But this monitoring technology is costly and requires the 

installation of protection nets or slider fences in advance. 

1 OEBB Infrastructure AG, Vienna, AUSTRIA, michael.brauner@oebb.at;michael.brauner@live.at 

2 PULSE Engineering GmbH Kajetan Schellmanngasse 8, 2352 Gumpoldskirchen, Austria 

3 NBG Fosa GmbH Zweiländerstraße 1 3950 Gmünd 



Here distributed acoustic sensing (DAS) using fiber optic cables might be an alternative. 

This technology is frequently used in different professional fields such as pipeline monitoring, 

securing industrial facilities or structural health monitoring. Recently it is also applied to 

other purposes such as the detection of avalanche activity (Prokop et al 2014) or railway 

tracking (Zeni et al 2013). 

AIM OF THE STUDY AND MAIN ACTIVITIES 

This study was carried out in close cooperation between NBG FOSA GmbH and PULSE 

Engineering GmbH on behalf of OEBB Infrastructure AG. It was defined the main goal to 

evaluate if the technology of distributed acoustic sensing (DAS) can meet the OEBB standards 

for automatic rock fall and tree-throw detection as follows: 

– Automatic detection of a minimum rock fall impact of 1.5 kJ on the railway track with a 

detection certainty of 97%. 1.5 kJ was defined as the lowest impact level, which may cause 

initial damage to track or train material. 

– The responsible operators shall receive up to four simple alarm thresholds with clear action 

plans following up. 

– For maintenance reason no direct installation of fiber cables in track ballast is acceptable. 

Standard fiber cable material and standard cable troughs or cable mounting techniques 

have to be used. 

– For secure operation and easy maintenance the main equipment should be installed in 

railway control centers and not wayside of the tracks. 

First tests in 2012 were promising and build the basis for the feasibility - study which is 

presented in this paper. This study is structured as follows: 

As different parameters (meteorology, track bed, cable properties, etc.) may have influence on 

the monitoring quality with fiber optic cables, a basic parameter assessment was carried out at 

the beginning of the study. Field experiments under laboratory conditions followed. They had 

the aim to analyze and understand the rock fall induced signal pattern and its frequency 

band. Then full-scale tests on an operating railway track during maintenance windows were 

used to gather signals of artificially produced rock falls and tree-throws and to develop 

recognition patterns. To test system stability and improve alarm parameters, finally the system 

was applied on an operating railway track for four months. 

BACKGROUND ON DISTRIBUTED ACOUSTIC MONITORING 

A so-called interrogator unit attached to one fiber of the cable induces the pulses of light and 

analyses the natural backscatter (e.g. Rayleigh, Brillouin and Raman scattering) of each point 

of the fiber (Akkerman et al 2013). Thus it can be interpreted like an array of countless 

microphones along the length of the fiber (Inaudi et al 2007). 

The detection principle of fiber optic cable is the Optical-Time-Domain-Reflectometry 

(ODTR). It is used to determine and analyze travel time and reflection characteristics in the 

electromagnetic spectrum of light and hence to define location and kind of disturbance. In 

this study the principle of Rayleigh backscattering is applied, using a commercially available 


interrogator unit (iDAS, Silixa Ltd.). Imperfections of the fiber but also movements or 

vibrations along the cable (e.g. induced by rock fall activity) result in different backscattering 

characteristics. By assessing the attenuation of the pulse, the type of interference can be 

characterized and traced with a spatial resolution of approximately 10 m (Fig 1). 

Figure 1: System sketch of a distributed acoustic monitoring system using fiber optic cables. 

BASIC ASSESSMENT 

The first test results clearly showed that for automatic event detection many different factors 

need to be considered and investigated along the railway track. The most important ones are 

summarized below: 

– Natural Hazards and the lack of protection along the railway: To achieve an efficient and 

accurate monitoring system the following questions need to be addressed: i.) Where are 

the potential risk areas and active protection measures located along the track? ii.) 

Where are the boundaries between different hazardous processes and secure area? 

– Properties of the fiber optic cable: Especially when operating telecom fiber optic cables 

are employed, it is crucial to be aware of their present location and condition. Cables, 

earth-buried or located in a cable trough, underground or overhead, show different 

sensibility to rock fall events. If multiple cables are installed in a cable trough, their 


placement may influence the sensing quality. Cable crossings and overlaying produce noise 

in the data. 

– Properties of the track bed: The construction of track and track ballast has substantial 

influence on signal attenuation and propagation. The track bed is originally developed to 

obtain optimal load transfer and to minimize vibration. Tests revealed that the elastic pads 

between the rails and ties and - more importantly - the elasticity of the ties and track 

ballast lead to significant signal attenuation and in consequence to different sensing results, 

depending on the location of the fiber optic cable. Therefore it has to be answered in 

advance: How does the construction of the track and track ballast look like? 

– Interfering external signal sources: To eliminate interferences from other transport carriers 

(e.g. cars, planes, helicopters, etc.) their influence has to be quantified. Knowing source, 

amplitude and distance of possible interfering signals, these signals can be characterized. 

Therefore, the following questions need to be addressed: i.) Are there roads or crossings 

nearby? ii.) Is there air traffic? iii.) Is construction work going on and can it be registered 

in advance? 

– Meteorology: The influence of meteorological events such as wind or rain can have severe 

impact on the monitored signal. They may induce vibrations in the cable trough or the 

cable itself. Consequently, the forecast of gusting wind or rain is crucial in order to prevent 

false alarms. 

FIELD EXPERIMENTS UNDER LABORATORY CONDITIONS (SPRING 2013) 

The objective of these experiments was to analyze the signal and in particular the frequency 

band of different kinds of rock fall impacts under known conditions and to evaluate the 

performance of different cable placements. Using this information, a preliminary definition of 

detection parameters was carried out and preliminary recognition patterns, so called “events”, 

were defined. For this purpose a 150 m long privately owned shunting track was adapted. 

Both sides of the track were equipped with different cable layout according to OEBB 

regulations. One was set up as underground cable, one was placed inside the newly build 

cable trough and a third one was directly attached to the rail. Then impacts by objects with 

various weights and materials were induced and their different signals were analyzed (Fig.2). 

The field experiments lead to the result that the optimal frequency range for rock fall 

detection is between 100 Hz and 500 Hz. For these frequencies, a pulse rate of 4 kHz delivers 

an optimal relation between detection range and sensitivity. The variable resonance behavior 

and frequency domain of rails, ties, ballast and foundation can be utilized for improved event 

discrimination along the track. However, with a single cable layout no reliable detection 

crosswise to the track axis is possible. That is because at the same distance to the fiber, direct 

impacts on the rail may result in higher signals than collisions on foundation or adjacent 

ground. Here fiber optic cables directly mounted at the rail may be a promising alternative as 

they enable a better discrimination between direct rail-track impact and ballast impact due 

to different signal attenuation. 


Figure 2: Left: impact on rails with a granite bowl from a predefined drop height. Right: Installation of cable trough according to OEBB 

regulations for the field experiments. 

TEST SITE HIEFLAU ADMONT – FULL SCALE FIELD TESTS (SPRING AND FALL 2013) 

After the field experiments full-scale tests on an operating railway line (Hieflau – Admont, 

Styria, Austria) were carried out. This line is a 40 km single track, with approx. 280 trains/ 

month. This alpine type track is characterized by 4 bridges and 4 tunnels as well as different 

types of fiber optic cables and different laying systems (earth-based or overhead). One fiber of 

an operating telecom fiber optic cable was connected to the interrogator unit and signals of 

artificially produced tree- and rock falls were analyzed. The goal was to gather data on a 

railway track in operating state and redefine the preliminary detection patterns under realistic 

conditions. Therefore 30 rocks of different size (up to 50 kg) and shape were thrown down 

a wall and an excavator was used to simulate around 20 tree-falls (Fig. 3). 

Figure 3: Left: Rocks thrown down on railway tracks. Right: Simulation of tree-fall using an excavator. 

The full-scale tests showed that rocks of 20 kg or more and a dropping height above 10 m can 

be easily identified along the railway axis, but a precise definition of drop height and weight 

cannot be achieved. For the tree-throws the characteristic frequency spectrum is similar to 

rock fall but has less energy. They can be identified, if the drop height is higher than 4 m. 


During the experiment blasting operations were carried out on a construction site close to the 

test site, which produced strong interfering signals. It can be concluded that the automatic 

detection system must be deactivated in case of nearby blasting or construction works. Similar 

the influence of traffic nearer than 10 m becomes dominant and hampers automatic detection 

based on defined event patterns. These influences fade out for distances above 50 m. 

TEST SITE HIEFLAU ADMONT – LONG-TERM TEST (SPRING 2014) 

After the full-scale tests and in-depth analysis of the gathered data the detection parameters 

were modified and three recognition patterns were defined. For events with an impact energy 

lower than 1.5 kJ, we defined the category “rock fall manual”, meaning interaction of a 

specialist is needed to define whether this was a rock fall or an interfering signal. For impact 

energy higher than 1.5 kJ we defined the category “rock fall automatic”, meaning that an 

automatic alarm will be triggered. The last category was “train” in order to identify trains on 

the track. 

In a final step a 6 km long section of the railway track Hieflau-Admont (Styria, Austria) was 

monitored for four months during operating state. The goal was to test the stability of the 

detection parameters and to test the stability of the whole system setup (interrogator unit, 

data storage, automatic data analysis) during constant operation over a longer period. The 

interrogator unit and the alarm server were installed in the railway control center at the 

railway station Hieflau. The server transmitted alarms to a web-interface, enabling the 

geographically referenced visualization of train positions, event-location and alarm type. 

It also enables email notifications and remote modification of the alarm parameters in case 

of malfunction or maintenance windows. 

After a few weeks it was clear that the monitoring system itself runs stable, however the 

existing cable placement in a concrete trough produced more false alarms than expected and 

the alarm parameters therefore needed adaption. Here the remote access to the permanent 

data logging enabled quick improvement of the detection parameters. Problems were caused 

by several unexpected incidents. For example, a train passing the monitored area can be 

easily tracked and identified with less than 0.5% error rate, but e.g. maintenance trains 

stopping in the area for several minutes and then starting again were hard to identify. 

The Implementation of train detection to differentiate from rock fall, the masking of bridges, 

crossings, overhead troughs, or save-guarded track section, improved the recognition quality 

tremendously. 

Moreover thunderstorms passing the area of our test operation produced a lot of false alarms. 

After signal analysis and site investigations we found out that the source of the interfering 

signals was not the wind or rain itself, but rather signposts, cable-masts and trough covers 

that induced vibrations into the ground, the cable through and the cable itself. To reduce false 

alarms they need to be acoustically uncoupled from the fiber cable. 


It can be concluded that within the 4-months period we were able to reduce the false alarms 

for rock fall events higher than 1.5 kJ from 35/day down to 5/day, which is still too high, to 

meet the error rate of 3%, defined as criterion for automated detection. Up to now our 

experiences suggest, that this error rate may be met for rock fall events larger than 15 kJ. 

calAt the moment, operator’s expertise is still required for events smaller than 15 kJ to 

prevent false alarms. 

The long-term test further showed that the system must allow for quick and easy entry of 

maintenance action where automatic alarms are disabled, and the integration of the 

monitoring system into the OEBB IT architecture can be achieved via TCP/IP interface and 

alarm notification can be triggered via SMTP-interface. 

CONCLUSIONS AND FUTURE OUTLOOK 

With standard telecom fiber optic cables the impact of potentially hazardous processes such as 

rock fall on the railway track can be detected. However, for the development of an automatic 

detection system still some challenges exist. Above all, the sound installation of existing fiber 

optic cables along the track and even the placement of the cable itself in the trough is of high 

importance. Harsh meteorological conditions, especially strong wind and rain, can cause 

additional noise in the signals and need to be considered. In the long-term test operation false 

alarm rates of 35/day could be reduced to 5/day. These values are still too high and far away 

from the predefined goal of detecting a rock fall event of 1.5 kJ with a probability of 97 %. 

As an alternative to further reduce the false alarms we tested a fiber optic cable mounted 

directly to the rails. Thereby, interfering signals caused by variable underground or inhomogeneous 

cable installations can be removed. At the moment it is unknown if this setup will 

cause other interfering signals. Akkerman et al (2013) tested a similar setup with a cable 

installed in the track ballast and reached a minimum false alarm rate of 1/day. Therefore 

alternative mounting setups of fiber optic cables seem promising but on the other hand a 

reduction of the false alarm rate below 1/day at the moment is not realistic. 

The detection of rock fall or other natural hazards along many kilometers of railway tracks 

using fiber optic cables has huge potential but also still needs a lot of research work. In this 

feasibility study we were able to demonstrate the advantage of a linear acoustic sensing 

system providing gapless monitoring of many kilometers of railway tracks but also faced 

many challenges with the automatic detection due to interfering signals. We believe that the 

potentials of the cable itself, the cable installation and the detection algorithm are not fully 

exploited yet. Therefore, a great need for research in this area still exists. 

The authors hope that this study will stimulate interest in this field and encourage other 

engineers, scientists and supporters to get active in this fascinating work of distributed 

acoustic sensing. 


REFERENCE 

- Akkerman J., Prahl F. (2013). Fiber Optic Sensing for Detecting Rock Falls on Rail Rights 

of Way. AREAM annual conference 2013, Indianapolis, USA. 

- Inaudi D.,Glisic B. (2007). Distributed fiber-optic sensing for long-range monitoring of 

pipelines. The 3rd International Conference on Structural Health Monitoring of Intelligent 

Infrastructure, Vancouver, British Columbia, Canada90: 112-124. 

- Prokop A., Schön P., Wirbel A. and Jungmayr M. (2014). Monitoring Avalanche Activity 

using Distributed Acoustic Fiber Optic Sensing, Proceedings International Snow Science 

Workshop 2014, S. 129-133. 

- Schorno R., Schmidt C., Nietlispach U., (2011). Zugkontrolleinrichtungen in der Schweiz. 

Eb - Elektrische Bahnen. 109 (2011) Heft 9. 

- Zeni L., Minardo A., Porcaro G., Giannetta D. and Bernini R. (2013). Monitoring railways 

with optical fibers, Applied Optics, Vol. 52, No. 16, S. 3770-3776. 



Terrestrial radar interferometry for snow glide activity 

monitoring and its potential as precursor of wet snow 

avalanches. 

Rafael Caduff 1 ; Andreas Wiesmann 2 ; Yves Bühler 3 ; Claudia Bieler 3 ; Philippe Limpach 4 

ABSTRACT 

Snowpack displacement and resulting full depth gliding snow avalanches are a widespread 

problem in alpine regions during springtime after snow-rich winters. This is a major threat for 

infrastructures nearby. Today field observers detect gliding snow visually. However, they can 

only detect suspicious snowpack displacement after gliding cracks opened. Many wet snow 

avalanches release without formation of visible tension cracks in an early stage so that they 

appear to happen spontaneously. Terrestrial radar interferometry has recently proven to be an 

effective method for the detection and monitoring of snow glide activity on a slope scale. 

Movements of the snowpack less than 0.5 mm/h can be detected as validated with a total 

station. Detection of snow glide activity is therefore achievable at a very early stage. Continuous 

measurements with rates of up to one scene per minute allow the immediate assessment 

of snow glide activity on the monitored slope. Therefore this method might be applied in 

future to detect precursors for full depth gliding snow avalanches. 

KEYWORDS 

snow glide monitoring; terrestrial radar interferometry; wet snow avalanche; data validation 

INTRODUCTION 

Different gravitational deformation processes such as settlement, creep and full depth snow 

glide are acting on snow slopes. Once full depth gliding is present, it can have a twofold 

impact on infrastructures within the snow slope. The impact is either direct, i.e. through 

immediate contact of the gliding snow column with the infrastructure, or the impact is 

indirect, i.e. the infrastructure gets impacted by a full depth snow glide avalanche from above. 

In both cases infrastructure such as buildings, roads, railways and power lines can be 

damaged severely (Bartelt et al., 2012). Gliding snow avalanches are hard to predict because 

they depend on surface roughness (Feistl et al., 2014) as well as on the moisture penetration 

into the snow pack which can change within very short time (Höller, 2013; Mitterer et al., 

2011). Full depth snow glide is not limited to extraordinary snow glide winters with high 

snow columns but is common as well during wet snow conditions with low snow height. Full 

depth snow glide is usually identified on a slope by the presence of clearly visible morphologi- 

1 GAMMA Remote Sensing AG, Gümligen, SWITZERLAND, caduff@gamma-rs.ch 

2 Gamma Remote Sensing, Switzerland 

3 WSL Institute for Snow and Avalanche Research SLF, Switzerland 

4 ETH Zürich, Institute of Geodesy and Photogrammetry, Switzerland 



cal features such as extensional cracks or compression structures at the toe of a snow glide 

area (stauchwall) that can be detected e.g. by time lapse photography (Van Herwijen et 

al., 2013). Besides these morphological signs, detection and quantification of very small snow 

glide movements was possible only point-wise. All measurements leading to information on 

the movement of the snow column had to be instrumented on site (Höller, 2014). Terrestrial 

radar interferometry was recently introduced as an innovative method that allows very high 

spatio-temporal sampling on a slope-wide scale during day and night, foggy conditions and 

independent from the existence of visible deformation related features in the snowpack 

(Caduff et al. 2015a; Wiesmann et al. 2015). 

Here we shortly summarize the technical preconditions necessary to monitor the snow pack 

displacements on a slope wide scale. We then discuss the potential and the limitations of 

interferometric measurements of snow to determine the rate of glide movement with a 

sensitivity of less than 0.5 mm/h. It was possible for the first time to validate measurements 

from the terrestrial radar interferometer with survey data obtained from a total station. 

Finally, we compare the snow glide activity with the avalanche activity on the Dorfberg test 

site in Davos, Switzerland during the winter 2014/2015. 

METHODS 

The GAMMA Portable Radar Interferometer (GPRI), operating at 17.2 GHz (Ku-band) was 

used for the campaign (Werner et al. 2012). The GPRI was deployed for continuous monitoring 

of the Dorfberg area at the valley bottom of Davos Dorf and was protected with a radome 

of 2.4 m diameter (Figure 1). From this position, the instrument illuminates the target area 

with an azimuthal scan (mechanical rotation of the antennas with 10°/sec). The received 

backscatter signal (amplitude and phase) forms a 2-dimensional image of the area with a 

pixel resolution of 0.75 m in range and a nominal azimuth resolution of 8m in 1 km distance. 

The final georeferenced maps were resampled to a pixel resolution of 1×1 m. The distance to 

the scatterer influences the relative phase value, therefore movement towards or away from 

the sensor induces phase changes. This effect is used to detect displacement in the target area 

in the order of a fraction of the instrument wavelength (1.74 cm). A recent review of the 

method with examples in the geoscientific field is given by Caduff et al. (2015b). 

It is necessary for the conversion of phase changes to 1-dimensional displacement in 

line-of-sight (LOS) that patches show high interferometric coherence. Coherence is a value 

that expresses similarity of surrounding raster cells (0: no similarity, 1 high similarity). The 

temporal behavior of the scatterers determines the amount of coherence of the interferogram 

(Figure 2). There are mechanical changes on the target surface, vegetation movement due to 

wind, snow drift or avalanches that lead to almost immediate decorrelation. In our study, 

avalanches were detected and mapped using coherence maps as shown in Figure 2. Avalanche 

outlines were drawn and verified manually. An operator decision led to a qualitative 

classification of the degree of decorrelation as seen in Figure 2 that allows a rough separation 


Figure 1: Overview of the monitored slope of the Dorfberg located in Davos Dorf, Switzerland. All detected avalanches during the observation 

period are shown in map (A). The radome (B left), containing the GAMMA Portable Radar Interferometer GPRI (B right) is located 

at the spot indicated with “GPRI”. The wet snow avalanches of a 48 h period are drawn on the image (C) The total daily sum of the area 

of the detected avalanches is shown in D. The corresponding daily mean temperature and snow height measurements from the weather 

station located 100 m apart from the GPRI are shown in plot E. 

of full depth avalanches (high degree of decorrelation) and superficial impacts on the snow 

(low degree of decorrelation). 

Changes in the physical state of the snow (dry/wet) lead to changes in the dielectric properties 

of the surface and have an influence as well on interferometric coherence especially 

during transition times (e.g. immediately after sunrise). In Figure 2 this effect is expressed in 

the maps showing the distribution of the coherence after 2 min and after 21 min during wet 

snow conditions. Comparisons to dry snow conditions show that the decorrelation time 

ranges from several hours during dry snow conditions to less than ten minutes in wet snow 

conditions without visible mechanical impact. In conclusion, the sampling interval must be 

below the decorrelation time to be able to convert the interferometric phase to displacement 

information. 


The campaign was performed using acquisitions with 2 or 3 minute intervals to retain high 

coherence. After each acquisition an automated processing calculated interferograms, 

coherence maps and stacks of temporal averages (30 min) of the displacement to reduce 

atmospheric phase noise. A mobile data connection allowed remote access to the instrument 

and the data through a web-interface. 

Figure 2: Coherence maps with 3 min and 21 min temporal baselines show different effects reducing the coherence on the slope. Snow 

surface is completely decorrelated after 21 min during wet snow conditions. 

RESULTS 

Evaluation of optimal temporal sampling rates 

Even though the interferometric coherence can be kept high during wet snow conditions, 

transition effects may still be visible in the final displacement diagrams. Figure 3 shows such 

effects and the attempt to reduce their impact in order to separate them from the actual snow 

displacement. The matrix shows the averaged displacement of stacks with different counts of 

observations prior to an avalanche release from point d1. However, since for early warning 

purposes, the time for the detection should be as short as possible, different temporal 

sampling intervals were tested. 

Although the deformation hot-spot d1 is visible already in the 2 min interferograms of 14:02 

and 14:14, those scenes suffer from atmospheric effects that impact the entire scene. Small 

changes in the physical parameters of the air (temperature, pressure, humidity) can induce 

phase delays in the interferogram. The patches, if not detected and filtered properly could 

lead to a misinterpretation of atmospheric disturbances as deformation signals. Patches of 

atmospheric effects with similar size and shape as snow glide deformation signals show 

randomness in time and therefore get reduced significantly with increasing number of 

observations (stacking). 


Figure 3: Effects of different averaging intervals on the displacement maps in a small subset. The used dataset shows the preceding 

situation of an avalanche triggered by the release of the snow glide area d1. A second snow glide area is marked as d2. For this area the 

validation with the total station was performed (see Figure 5). 

For comparison, the single interferogram scenes in Figure 3 were normalized to a stable point 

(rock) close to the area since patches of atmospheric influences are present over the entire 

scene. In the 12 h stack, the velocity is less visible since both areas d1 and d2 were slower the 

days before and during the night. The temporal evolution of d1 and d2 is shown in Figure 4. 

The enlarged map in Figure 3 shows that the sensitivity of the measurements can reach less 

than 1 mm. Areas with LOS velocities of 0.25 mm/h can clearly be separated. However, such 

a precision is achieved only if the snow surface is not disturbed by snowfall or snowdrift. 

Another effect that may hinder the detection of snow glide hot-spots is the settlement of 


freshly fallen snow. The deformation in the vertical is partly seen as deformation in LOS. 

Since the deformation values are in the same order or even exceed deformation induced by 

snow gliding, a clear distinction between both processes is not possible. 

The LOS displacement during the transition of dry to wet snow (10:14) is strongly disturbed 

by effects described in the method section. During this short time period, a high precision 

measurement of snow glide motion is difficult. 

Snow glide activity maps 

Snow glide activity on the slope was assessed by creating daily average LOS velocity (vLOS) 

maps. The maps were spatially filtered (5x5 pixel average) to calculate isotachs. The isotachs 

were grouped in 3 classes (1, 2 and 3 mm) for which the total area and number of isolated 

snow glide hot-spots were calculated. The results are plotted in Figure 4. Information about 

the number of avalanches and the total area affected by the avalanches are plotted for the 

same period. The period between the 15th and 29th of March is considered as a wet snow 

avalanche period. Here the avalanches are often preceded by snow glide movement. The 

displacement diagram of different selected points shows the temporal evolution of the snow 

glide displacement. Snow-melt and other snow related effects were corrected for each point 

with the subtraction of those signatures using a local non-gliding reference. Point d1 shows 

movement on a cm-scale only before the avalanche release, while the release area of d2 

showed long term continuous and constant snow glide activity beginning more than 8 days 

before the release. A strong acceleration was recorded a few hours before the avalanche 

release. Other points (d3, d4, d5) show clear acceleration cycles during day/night. 

VALIDATION 

A validation of the displacement measurements with the GPRI was performed with a 

combined approach of total station and experimental single frequency GPS sensors. The GPS 

phase data is processed differentially with respect to a nearby reference station. The field 

setup is shown in Figure 5. Reflective foil targets were put on lightweight styrofoam plates 

equipped with plastic anchors for fixation on the snow surface. The targets were installed 

early in the morning, when a firm snow surface was present on a target area previously 

detected with the GPRI. 

The survey was started at 10:30. A second measurement was acquired at 14:00. Shortly after 

this measurement, a small avalanche was released above the test field and went through the 

middle of the field, destroying some targets and sweeping away both GPS boxes. The 

avalanche released from point d1 (Figure 3 and Figure 4). It is a very small area with a local 

slope angle of 40-45°. It is to mention, that the avalanche from d1 was not able to trigger the 

release of the entire area of d2. However, the total station measurements were continued on 

the remaining targets. After the campaign, the total deformation of each point was calculated 

(small image in Figure 5). To compare the measurements with the GPRI LOS velocity map, it 


Figure 4: Synthesis of the results of the wet snow avalanche period in March 2015. Top left: Snow glide activity map. Top right: 

avalanche activity map. Bottom left: Line of side snow glide displacement for selected points together with temperature (air and snow 

surface 30 min) and reflected short wave radiation (RSWR). Bottom left: Comparison of snow glide activity (number and area) with 

avalanche activity (number of avalanches and affected area). 

was necessary to convert the total station results to LOS. The LOS fraction was determined by 

calculating the angle between the displacement vector and the look vector of the GPRI for 

each point. The total displacement was then scaled with the LOS fraction and plotted on top 

of the GPRI LOS velocity map using the same color scale. The results of the LOS-corrected 

values of both, the GPRI and the total station measurements are in very good agreement 

(within 0.5 mm/hour) for the relatively short observation period. 

After their installation at around 09:30, the GPS boxes were slowly gliding along the slope 

and deforming the snow surface due to their weight, mainly coming from the battery 


included in the boxes. Unfortunately, the GPS boxes were swept away by the avalanche 

before they reached an equilibrium that would allow detecting the movement of the snow 

pack. 

Figure 5: Field setup of the validation with total station and GPS boxes. The total station is located at (782494/188159). The results of 

the total station measurements were calculated for the observation period after the avalanche and were converted to LOS velocity. The 

results are plotted as color coded points on top of the GPRI LOS velocity map of the same observation period (14:16-17:00 UTC) with the 

same color coding. Below the map, the uncorrected total station displacement values and an example of the displacement vector are 

given. 

CONCLUSIONS 

The presented results confirm the feasibility of snow glide displacement measurements using 

terrestrial radar interferometry. With proper stacking, determination of the slope wide snow 

glide activity with relative sensitivity down to 0.25 mm/h and an absolute accuracy of 

0.5 mm is reached during weather conditions without precipitation and with absent strong 

winds. This is the case for parts of the observed slope that are free of shrub vegetation that 

emerges from the snowpack. The results and the tests performed during the campaign clearly 

show that the atmospheric influence is negligible when the data is averaged in periods of 

12 min to 4 h. This simplifies the in-situ data processing and allows faster acquisitions. 

However, due to the high sensitivity of the instrument, small changes in the snow column 

as induced by snow-fall or wind drift can lead to phase changes and exceed the displacement 

signatures. During these conditions, signatures seen by the radar could not be clearly 

identified as either snow glide or some other processes acting on the snow pack. A careful 


interpretation of the interferometric data is as well necessary during times of transition from 

dry to wet snow (e.g. after sunrise). The transition of the liquid water content in the 

snowpack may shift the phase center from within the snowpack to the surface (Caduff et al. 

2015a) and therefore induce phase changes. Once snow glide hot-spots are located, tracking 

of the point displacement is possible in shorter intervals using simple normalization to 

suppress atmospheric or snow property induced phase effects. Interpretation of the temporal 

evolution of snow glide hot-spots led to the conclusion that most release zones of wet snow 

avalanches show signature of movement hours to days before the release and in most of the 

cases, strong acceleration of the movement hours before the avalanche release was recorded. 

Since many of the released wet snow avalanches show acceleration hours to minutes before 

the event, acceleration might be valuable indicator for the timing of the release. However, as 

the deformation signature of d1 in Figure 4 shows, additional investigations of the influence 

of the local topography are needed since the trigger-acceleration of d1 was apparently lower 

due to its steep release zone. 

For the first time, an absolute validation of the snow pack displacement measured using 

terrestrial radar interferometry was performed using total station measurements. The 

resulting LOS velocities show very good agreement within less than 0.5 mm/h. However, 

since the validation campaign lasted only a few hours and the temperature conditions did not 

allow coverage of a full freeze/thaw cycle of the snow-surface, the quantification of the phase 

effects during the transition are not yet fully understood. 

The campaign helped to solve major open question regarding the use of terrestrial radar 

interferometry rapid detection and quantification on snow glide on a slope. The technique 

opens new possibilities in the research on snow gliding and wet-snow avalanches. We also 

expect it to prove valuable in the management of high-risk situations during elevated snow 

glide and wet-snow avalanche activity. 


Part of the funding for this research has been provided through the Interreg project 

STRADA 2. 

REFERENCES 

- Bartelt, P., Pielmeier, C., Margreth, S., Harvey, S., Stucki, T. (2012). The underestimated role 

of the Stauchwall in full-depth avalanche release. Proceedings of the International Snow 

Science Workshop, September 16-21, 2012, Anchorage, AK, USA. 127-133. 

- Caduff R., Wiesmann A., Bühler Y., Pielmeier C. (2015a). Continuous monitoring of 

snowpack displacement at high spatial and temporal resolution with terrestrial radar 

interferometry. Snowpack displacement monitoring. Geophysical Research Letters 42(3), 

813-820. 

- Caduff R., Schlunegger F., Kos A., Wiesmann A. (2015b). A review of terrestrial radar 

interferometry for measuring surface change in the geosciences. Earth Surface Processes and 

Landforms 40(2), 208-228. 


- Feistl T., Bebi P., Dreier L., Hanewinkel M., Bartelt P. (2014). Quantification of basal friction 

for technical and silvicultural glide-snow avalanche mitigation measures. Natural Hazards and 

Earth System Science 14, 2921-2931. 

- Höller P. (2013). Snow gliding and glide avalanches: a review. Natural Hazards 71, 1259- 

1288. 

- Mitterer C., Hirashima H., and Schweizer J. (2011). Wet-snow instabilities: Comparison of 

measured and modelled liquid water content and snow stratigraphy. Annals of Glaciology 52, 

201-208. 

- Van Herwijen A., Berthod N., Simenhois R., Mitterer C. (2013). Using time-lapse photography 

in avalanche research, Proceedings of the International Snow Science Workshop, October 

7-11, 2013, Grenoble, F, 950-954. 

- Wiesmann A., Caduff R., Mätzler C. (2015). Terrestrial Radar Observations of Dynamic 

Changes in Alpine Snow. IEEE Journal of Selected Topics in Applied Earth Observations and 

Remote Sensing 8(7), 3665-3671. 

- Werner C., Wiesmann A., Strozzi T., Kos A., Caduff R., Wegmüller U. (2012). The GPRI 

multi-mode differential interferometric radar for ground-based observations. Proceedings of 

the 9th European Conference on Synthetic Aperture Radar, April, 23-26, 2012, Nuremberg, 

D, 304-307. 



Earthquake-triggered landslides in Switzerland: from 

historical observations to the actual hazard 

Donat Fäh, Dr. 1 ; Jan Burjanek, Dr. 2 ; Carlo Cauzzi, Dr. 2 ; Remo Grolimund, Msc 2 ; Stefan Fritsche, Dr. 3 ; Ulrike Kleinbrod, Msc 2 

ABSTRACT 

The building inventory in the Alpine area multiplied over the last century, increasing exposure 

to potential earthquake impacts. This is particularly critical in relation to earthquaketriggered 

mass movements. In Switzerland, it is possible to assess the spatial extent and 

impact of such secondary earthquake effects from historical and paleo-seismological analysis 

of past damaging earthquakes. After calibration against observations of these events, scenario 

modelling has been recently implemented within ShakeMap, a tool applicable to near 

real-time estimates. The spatial resolution could be improved using additional information 

like rock-slope hazard maps or detailed studies on particular unstable slopes. In order to 

improve our understanding of the dynamic response of unstable rock slopes, an extensive 

measurement campaign is presently being performed to determine eigenfrequencies, groundmotion 

polarization and amplification features, and to estimate the volume of the instability. 

A semi-permanent seismic installation was set up at the Alpe di Roscioro (Preonzo) site, 

a slope that is close to collapse, to continuously monitor seasonal variations of its dynamic 

behavior and how it changes over time. 

KEYWORDS 

earthquakes; induced phenomena; landslides; rock fall; tsunamis 

INTRODUCTION 

Moderate to high seismic risk in Switzerland results from moderate seismicity combined with 

high population density, high degree of industrialization and relatively low preparedness level 

due to the comparatively high return periods of damaging events. 28 events of moment 

magnitude Mw ≥ 5.5 have been identified over the past 700 years, twelve of which caused 

severe damage (Intensity of VIII or higher). The epicentral areas of the strongest events 

experienced extensive damage from the earthquake ground motion and different induced 

phenomena. Such phenomena are liquefaction in the river plains, reactivation of landslides, 

extended rock-fall and tsunamis in lakes generated by induced mass movements. Nevertheless, 

we expect that smaller but more frequent earthquakes may induce large ground motions 

locally, as well as small-scale movements and failures of critically stressed slopes. 

Due to engineering progress in the last two centuries, seismically unfavorable sites have 

become attractive for the establishment of settlements and industries. Many villages have 

1 Swiss Seismological Service ETH Zurich, Zürich, SWITZERLAND, faeh@sed.ethz.ch 

2 Swiss Seismological Service ETH Zurich 

3 Qatar Reinsurance Company Zurich 



expanded into the river plains and reached the toes of hazardous slopes in the valleys. Future 

earthquakes will therefore cause more damage than in the past. Focused studies might 

therefore help to identify areas at risk and to develop strategies for the mitigation of possible 

impacts. 

EARTHQUAKE INDUCED EFFECTS DURING PAST EVENTS IN SWITZERLAND 

Several events with significant induced effects have been reassessed in the course of the 

historical revision of the Earthquake Catalogue of Switzerland ECOS-09 (Fäh et al., 2011). 

A wide variety of historical documents, e.g. chronicles, administration documents, newspaper 

articles, diaries and scientific articles, describing such effects have been found and analyzed 

(e.g. Fritsche et al., 2012). Events with significant induced effects are the ones of 1584 at 

Aigle Mw=5.9 (tsunami and landslides), 1601 in Obwalden Mw=5.9 (landslides and tsunami), 

1755 at Brig Mw=5.7 (landslides), 1855 at Visp Mw=6.2 (liquefaction and landslides), 

and 1946 at Sierre Mw=5.8 (liquefaction, landslides, and avalanches). 

The main shock of the 1584 earthquake series at Aigle on March 11 had an epicentral intensity 

of VIII. It triggered subaquatic slope failures resulting in a small tsunami and seiche in Lake 

Geneva. The strongest of the aftershocks triggered a destructive rock-fall covering the villages 

Corbeyrier and Yvorne, VD, killing about 320 people (e.g. Fritsche et al., 2012). The earthquake 

of 1601 in Unterwalden produced several rock-falls in central Switzerland. The 

rock-fall at mount Bürgenstock collapsed into the lake and earthquake-triggered subaquatic 

landslides produced a high wave and a seiche in Lake Lucerne causing heavy destruction in 

the proximity of the shorelines (e.g. Schwarz-Zanetti and Fäh, 2011). The historical accounts 

on the destructive wave could be complemented with sedimentological studies that indicated 

a number of similar events in the past 15,000 years (e.g. Strasser et al., 2013). Among the 

relatively rare reports on earthquake-induced effects of the 1755 Brig event are an earth 

slump near Brig and accounts about fissures and cracks in the ground that can be interpreted 

as evidences for liquefaction. The event probably also triggered a large rockslide with an 

approximate volume of 1.5 million m³ destroying parts of the village of Niedergrächen (Gisler 

and Fäh, 2011; Fritsche et al., 2012). A large number of induced effects are documented in 

the case of the strongest earthquake of the last 300 years in Switzerland, which occurred in 

1855 at Visp. Among them are rock-falls and landslides, as well as liquefaction-related ground 

deformations, lateral spreading and ground settlements (Fritsche et al., 2012). Similar effects 

are documented for the 1946 earthquake sequence at Sierre (Figure 1). Since the main shock 

occurred in wintertime, avalanches are also among the documented effects (Fritsche and Fäh, 

2009; Fritsche et al., 2012). 

Paleo-seismology provides information about severe events that occurred in pre-historic 

times. The available data is mostly related to potentially earthquake-triggered mass movements 

in lakes and corresponding induced tsunamis. One of the most significant potentially 

earthquake-related event is the Tauredunum rock-fall in AD 563, with a destructive tsunami 


in Lake Geneva. This event is documented in medieval chronicles as well as in the natural 

archive of lake sediments (e.g. Kremer et al., 2012). Several coeval slope failures occurring 

around 2200 BP in different Swiss lakes, such as Lake Zurich, Lake Lucerne, Lake Geneva and 

Lake Neuchâtel suggest another very large seismic event in Switzerland (Strasser et al., 2013; 

Kremer, 2014; Reusch et al., 2015). Another potentially earthquake-triggered flooding event 

has been dated in Late Iron Age documented by the destruction of a Celtic wooden bridge in 

Cornaux, NE (e.g. Grolimund et al., 2015). 

While some sedimentary and archaeological evidence in Switzerland could be attributed to 

historically documented seismic events, it is a more difficult matter for events prior to 1000 

AD. To reduce the uncertainties inherent to paleo-seismological events, e.g. due to dating, it is 

indispensable to integrate all available data. Therefore, the Swiss Seismological Service (SED) 

is currently collecting the available datasets in a common database with homogeneous quality 

criteria (Gassner et al., 2015). 

Figure 1: Earthquake-induced effects of the 1946 main shock at Sierre and its largest aftershock (after Fritsche et al., 2009). 

EARTHQUAKE-INDUCED EFFECTS IN FUTURE EARTHQUAKES 

Quantitative estimates of co-seismic land sliding (distribution patterns, magnitude-frequency 

relationships) and subsequent impacts are still difficult today. This is due to many input 

parameters they require. We therefore presently follow two different strategies: investigations 


on unstable rock slopes to understand their dynamic behaviour, and the development of 

regional models for earthquake triggering of landslides. 

At first, estimating the likelihood of potentially catastrophic landslides requires thorough 

understanding of the mechanisms driving slope dynamic behaviour. Long-term strength 

degradation of brittle rock masses is typically related to progressive failure. This is due to the 

propagation of new fractures through intact rock bridges and shearing along pre-existing 

discontinuities. A fundamental difficulty lies in experimental quantification of this process 

and the strength of fractured rock masses in relationship to the driving forces, i.e. the 

assessment of how close a given slope is to failure. This difficulty has major practical 

consequences, as the factor of safety of a slope at a given time defines the amplitude of the 

short-term disturbance required to trigger failure. 

A recent study demonstrated a very specific seismic response of unstable rock slopes. This 

might be considered as their unique signature (Burjanek et al., 2012). In particular, recorded 

ground motions (earthquakes or ambient vibrations) are highly directional in the unstable 

part of the rock slope, and significantly amplified with respect to the stable areas. These 

characteristics allow mapping stable and unstable portions of the rock mass. The predominant 

directions of ground motion match the past or on-going displacement directions (Burjanek et 

al., 2010). These effects are strongest at certain frequencies, which were identified as the 

eigenfrequencies of the unstable rock mass. The relative ground-motion amplifications with 

respect to the stable part of the slope reach typically peak amplitudes of 7 – 10 (see Figure 2). 

These effects are presently being assessed in a project supported by ETHZ on a large number 

of unstable rock slopes in Switzerland. On the one hand, such strong amplification levels and 

polarized ground motion may directly influence the potential for earthquake-triggered 

failure. On the other hand, the seismic response adds additional valuable information to the 

characterization of such unstable rock slopes. This will allow to further reduce the uncertainties 

related to stability, volume size and hazard assessment. 

Semi-permanent installations of seismic stations on unstable sites allow for long-term 

monitoring of eigenfrequencies and relative amplifications. It has been shown, that the 

fundamental frequency of the unstable slope could change with the time, and even could be 

seen as a precursor of the collapse (Levy et al, 2010). Two seismic stations have been recently 

installed by SED at the Alpe di Roscioro (Preonzo) site, a slope that is close to collapse. 

The relative amplification is monitored in time and shows seasonal variations (Figure 2). 

Strong variability is observed in wintertime (especially during freezing-thawing periods), 

whereas response is stable during summer. Such analyses are necessary for the correct 

interpretation of the dynamic response of instable slopes. 

The second research strategy deals with the use of geospatial proxies and rapid earthquake 

shaking information through Swiss ShakeMaps (Cauzzi et al., 2014) for the assessment of the 


Figure 2: Time variation of the ground-motion amplification (in color) of the rock instability at Alpe di Roscioro (Preonzo, Ticino). Changes 

in the amplification are due to seasonal variations affecting the global stiffness of the instability in the frequency band 3.7 - 5 Hz, as 

well as the dynamic response of sub-blocks at frequencies above 7 Hz. 

likelihood of landslides, rock-falls and liquefaction in near real-time. Our current prototype 

implementation relies on the customisation of the global and regional approaches of Nowicki 

et al. (2014) and Zhu et al. (2015) to Swiss conditions by comparing the predictions with the 

observations of damaging events. Notable amongst the aforementioned historical earthquakes 

is the 1946 Sierre event (Mw 5.8) and its strongest aftershock (Mw 5.5) for which observations 

of the induced effects are abundant (Fritsche and Fäh, 2009). Our best landslide 

prediction model for the Sierre 1946 event is shown in Figure 3. With a few exceptions, the 

model can satisfactorily represent the geographic distribution of the observations. The model 

seems to work better for rock-falls (where just steepness of the slope and ground shaking 

are the dominating factors) than for soil landslides (where the water content can play a 

significant role as well). The white diamond close to the epicenter in Figure 3 is the major 

(4-5 million cubic meters) Rawylhorn rock-fall (e.g. Fritsche and Fäh, 2009). There are large 

areas in Figure 3, N and SW of the epicenter, where mass movements were not observed. 

It is not clear if these areas constitute “false positives”. According to Fritsche and Fäh (2009), 

the secondary effects are well described in contemporary newspaper articles. However, a 

contemporary damage assessment carried out on behalf of the Canton Valais has survived in 

fragments implying only partial completeness of the historical dataset. It could be that the 

dip and state of the rock layers played a significant role in the distribution of the observed 

mass movements in the area. In general, the distribution of historically known earth- 


Figure 3: Landslide likelihood scenario for the 1946 Sierre event. The colour on the map from white to red reflects the likelihood from 0 to 

1. Symbols indicate the epicentre location (Black star), the observed rock-falls (diamonds), landslides (circles) and avalanches 

(triangles) triggered by the main shock (pink symbols) and the largest aftershock (white symbols). 

quake-triggered mass movements can be modelled satisfactorily. In a further step, cantonal 

hazard maps for landslide and rock-fall hazards could be included in the procedure to improve 

the spatial resolution of the prediction model. 


CONCLUSIONS 

The past earthquakes show impressively, that the Alpine region faces considerable seismic 

hazard and risk, in particular related to earthquake-induced phenomena such as landslides, 

rock-fall and liquefaction. Research on earthquake-triggered mass movements is addressed 

through historical and paleo-seismological analysis of past events, scenario modelling using 

ShakeMap, and studies of the dynamic response of a large number of unstable rock slopes. 

The ongoing research on the micro-vibrations of rock instabilities includes extensive 

measurement campaign to determine eigenfrequencies, ground-motion polarization features, 

ground-motion amplification levels, and to estimate the volume of the instabilities. This is 

resulting in a worldwide unique database of unstable-slope seismic responses. It is an 

investment for the future. It will serve for the monitoring of potential changes in the unstable 

rock masses. It is a base for the interpretation of potential future landslides triggered by 

earthquakes. For example, it could allow quantifying the slope damage after strong shaking 

(if the slope has not collapsed), and it might be used to improve the ShakeMap approach. 

For the ShakeMap approach we presently use georeferenced susceptibility proxies and 

ground shaking parameters for scenario modelling. However, the primary use of ShakeMaps 

is rapid estimates of earthquake impacts immediately after an event. This approach is based 

on accessible explanatory variables combined through simple functional forms with coefficients 

calibrated against the observations of past events. It is optimized for near real-time 

estimates based on USGS-style ShakeMaps implemented at SED since 2007. 

The presented research has a high practical relevance to Swiss ShakeMap end-users and 

stakeholders managing lifeline systems. Practitioners and cantonal authorities can run 

scenarios from which danger zones can be mapped and possible impacts estimated. After 

a strong earthquake information about possible landslides and rock-fall will be available 

within minutes. Future improvement might include cantonal hazard maps for landslide and 

rock-fall. This can be combined with real-time information such as precipitation and temperature, 

data from deformation devices or seismic stations located on particular instabilities. 

This research additionally facilitates future investigations on earthquake induced lake 

tsunamis triggered by subaquatic mass movements. 

REFERENCES 

- Burjánek J., Stamm G., Poggi V., Moore J.R., Fäh D. (2010): Ambient vibration analysis 

of an unstable mountain slope, Geophys. J. Int., 180, 820–828. 

- Burjánek J., Moore J.R., Yugsi Molina F.X., Fäh D. (2012): Instrumental evidence of normal 

mode rock slope vibration, Geophys. J. Int. 188:559–569. 

- Cauzzi C., Edwards B., Fäh D., et al. (2014): New predictive equations and site amplification 

estimates for the next-generation Swiss ShakeMaps, Geophys. J. Int. 200:421–438. 

- Fäh D. et al. (2011): ECOS-09 Earthquake Catalogue of Switzerland Release 2011 Report 

and Database, Public catalogue, 17.4.2011, Swiss Seismological Service ETH Zurich. 

- Fritsche S., Fäh D. (2009): The 1946 magnitude 6.1 earthquake in the Valais: site-effects 

as contributor to the damage, Swiss Journal of Geosciences, 102(3), 423–439. 


- Fritsche S., Fäh D., Schwarz-Zanetti G. (2012): Historical intensity VIII earthquakes along 

the Rhone valley (Valais, Switzerland): primary and secondary effects, Swiss Journal of 

Geosciences, 105(1), 1–18. 

- Gassner G., Fäh D., Strasser M., Grolimund R., Wirth S. (2015): Possible large earthquakes 

in prehistorical times: An overview of palaeoseismological data for Switzerland, Swiss Journal 

of Geosciences, submitted. 

- Gisler M., Fäh D. (2011): Grundlagen des makroseismischen Erdbebenkatalogs der Schweiz, 

1681–1878, vdf Hochschulverlag AG an der ETH Zürich, Zürich. 

- Grolimund R., Strasser M., Fäh D. (2015): Combining Natural and Historical Archives: 

What can we Learn about Large Alpine Earthquakes from the Late Iron Age to the Early 

Middle Ages?, Swiss Journal of Geosciences, submitted. 

- Kremer K. (2014): Reconstructing 4000 years of event history in deep Lake Geneva 

(Switzerland – France): insights from the sedimentary record, University of Geneva. 

- Kremer K., Simpson G., Girardclos S. (2012): Giant Lake Geneva tsunami in AD 563, 

Nature Geoscience, doi:10.1038/ngeo1618. 

- Levy C., Baillet L., Jongmans D., Mourot P., Hantz D. (2010): Dynamic response of the 

Chamousset rock column (Western Alps, France), J. geophys. Res., 115, F04043, doi:10.1029/ 

2009JF001606. 

- Nowicki M.A., Wald D.J., Hamburger M.W., et al. (2014): Development of a globally 

applicable model for near real-time prediction of seismically induced landslides. Eng Geol 

173:54–65. 

- Reusch A. et al. (2015): Giant lacustrine pockmarks with subaqueous groundwater 

discharge and subsurface sediment mobilization, Geophys. Res. Lett., 42(9), 2015GL064179, 

doi:10.1002/2015GL064179. 

- Schwarz-Zanetti G., Fäh D. (2011): Grundlagen des makroseismischen Erdbebenkatalogs 

der Schweiz, 1300–1680, vdf Hochschulverlag AG an der ETH Zürich, Zürich. 

- Strasser M., Monecke K., Schnellmann M., Anselmetti F.S. (2013): Lake sediments as 

natural seismographs: A compiled record of Late Quaternary earthquakes in Central 

Switzerland and its implication for Alpine deformation, Sedimentology, 60(1), 319–341. 

- Zhu J., Daley D., Baise L.G., et al. (2014): A Geospatial Liquefaction Model for Rapid 

Response and Loss Estimation. Earthq Spectra 141208072728004. doi: 

10.1193/121912EQS353M. 



Modelling of individual debris flows using Flow-R: 

A case study in four Swiss torrents 

Florian von Fischer, MSc 1 ; Margreth Keiler, PD Dr. 2 ; Markus Zimmermann, Dr. 2 

ABSTRACT 

Recently, an empirical debris flow model called Flow-R has been added to the broad range 

of existing debris flow modelling software. While Flow-R’s applicability on regional scale has 

been confirmed in several studies, it’s potential in local debris flow modelling has not yet 

been evaluated. In this study, Flow-R’s potential in debris flow modelling on local scale is tested 

via an application in four debris flow torrents in Switzerland. Obtained results are 

validated by comparison to documented debris flow events as well as model results from the 

commonly used hydraulic debris flow model RAMMS. Results show that due to the nonhydraulic 

model conception, the potential of Flow-R in individual debris flow modelling is 

limited. Plausible debris flow patterns are only achieved on torrents showing a low to 

moderate channelization, i.e. incision on the fan. Computed process velocities are unreliable 

due to the inability of the model to account for debris flow mass. Moreover, modelled debris 

flow magnitudes are biased by the compulsory non-volumetric definition of magnitudes. 

KEYWORDS 

debris flow modelling; Flow-R; susceptibility; debris flow hazard; case study 

INTRODUCTION 

In order to assess debris flow hazard, a variety of computer based debris flow models have 

been developed in the past decades. Recently, the empirical debris flow model Flow-R, 

designed for “susceptibility mapping of debris flows at regional scale” (Horton et al. 2013:1) 

was published. It allows the identification of potential source areas for debris flows as well as 

their propagation extent, essentially based on a DEM. Flow-R has been proved suitable for 

the production of regional debris flow “[…] susceptibility maps with satisfying accuracy” 

(Horton et al. 2013:870). Although Flow-R is not specifically designed for debris flow 

modelling on local scale, it is advisable to “compare the assessed susceptible zone with specific 

events in order to evaluate the accuracy of the results“ (Horton et al. 2013: 870). However, a 

systematic comparison of individual debris flow events and Flow-R modelling results gained 

on a highly resolved DEM (2m) has never been con-ducted. Therefore, the assessment of the 

accuracy and the allegedly limited applicability of Flow-R on local scale forms the main 

objective of this study. It is hypothesized that when applied on a highly resolved digital 

elevation model, Flow-R represents an alternative to sophisticated hydraulic models, such 

as RAMMS (Christen et al. 2012), in the modelling of individual debris flow events. 

1 Universität Bern, Bern, SWITZERLAND, flo.v.fischer@gmail.com 

2 Universität Bern 



This hypothesis is mainly tested with respect to modelled affected areas but also considers 

computed process velocities and model output for different event magnitudes. 

METHODS 

Flow-R’s potential in local debris flow modelling is evaluated by means of a reconstruction of 

four well-documented debris flow events in the Swiss Alps. Due to the diverse event 

characteristics the four torrents Richleren (UR, 1987 event: ~4’000 m³), Minstigerbach (VS, 

1987 event: ~30’000 m³), Glyssibach (BE, 2005 event: ~70’000 m³) and Varuna (GR, 1987 

event: ~185’000 m³) are selected for this study. Debris-flow model runs are conducted on the 

2m-resolved digital elevation model SwissALTI 3D . The calibration of the debris-flow model is 

essentially based on: 

– all available information from event documentations (VAW 1992, LLE 2006) comprising 

debris flow velocities, maximum discharge rates, event volumes and affected areas 

– empirical benchmark calibration values for the model’s friction parameters from literature 

(Horton et al. 2013, Gamma 2000, Zimmermann et al. 1997, Rickenmann & Zimmermann 

1993, Christen et al. 2012). 

In an iterative modelling process, the model’s parameters are adjusted until the obtained 

output fits the observed debris flow extent as well as possible (Fig. 1). For reasons of 

comparison, the documented debris flows are reconstructed with the hydraulic debris flow 

model RAMMS following the same iterative procedure. Among all input variables, the 

adjustment of applied friction parameters plays a predominant role in the calibration process 

for both debris flow models applied. For a better understanding, the following section shortly 

summarizes the basic functional principles of Flow-R and RAMMS based on Horton et al. 

(2013) and Christen et al. (2012). 

Flow-R calculates debris flows starting from either predefined source cells or from source cells 

empirically derived from the DEM. The direction and spreading of the debris flow is computed 

with a flow direction algorithm including a spreading exponent, whereas the assessment of 

the runout distance is based on one of two simple physical approaches. The model after Perla 

et al. (1980) calculates the velocity of the flow including friction parameters (μ) and mass-todrag 

ratio (M/D). Alternatively the SFLM, determining the maximum runout distance based 

on a minimum travel angle (TA) and maximum velocity (v lim ) can be selected. Flow-R does 

not consider debris flow volume but works with a unitary mass of 1. Flow-R model outputs 

include raster-data of spatial susceptibility as well as relative kinetic energy. 

The RAMMS debris flow model employs a modified Voellmy-fluid friction model, which 

includes a dry-Coulomb friction parameter (μ) and a turbulent friction parameter (ξ). In 

contrast to Flow-R, RAMMS requires an input debris flow volume. As applied in this study, 

debris flow volume and velocity can be included in a user-defined input hydrograph, which 

specifies maximum discharge rates and debris flow and velocity over time for given location 

in the torrent. The outputs generated with RAMMS comprise maximum flow height, velocity 

and impact pressure. 


Figure 1: Schematic representation of the research design for the assessment of Flow-R’s applicabil-ity in local debris-flow modelling. 

In addition to the event reconstruction, the selected debris flow models are also used to 

simulate further debris flow magnitudes, which are based on predefined scenarios. In 

RAMMS, additional debris flow magnitudes are defined by means of increasing or decreasing 

event volumes. With respect to friction parameters, the best-fit values determined in the prior 

event reconstruction are adopted, as rheological properties are assumed constant. Due to 

the model conception, this approach is not applicable to Flow-R. Alternatively, debris-flow 

magnitudes are defined based on a method applied by Gamma (2000) who defines magnitude 

scenarios according to relative travel distances on the fan, assuming longer travel distances 

with increasing debris flow magnitude. In Flow-R different magnitudes are thus modelled by 

adjusting friction parameters or travel angle, until the model results roughly correspond to 

the expected runout distance. 

For the appraisal of Flow-R’s performance, model outputs of Flow-R and RAMMS are 

compared mutually as well as to the documented debris flow events. This analysis includes 

both qualitative and quantitative aspects. The qualitative assessment focuses on the plausibility 

of the obtained debris flow pattern from a geomorphologic point of view. In other words, 

model output is evaluated with respect to its representativity of natural process behaviour, 

especially in connection with outbreaks of the flow from the channel. The quantitative assessment 

of model performance is based on the confusion matrix methodology by Beguerìa 

(2006). Confusion matrices allow an evaluation of the predictive power of models based on 

an overlay of model results and according validation datasets (Begueria 2006:315ff). 

The intersection of debris flow model output and documented debris flow extent results in 

four different classes of areas, composing the confusion matrix (Fig. 2). The evaluation of 

the debris flow models used in this study is based on the three following indices: 

– Sensitivity: Proportion of area correctly modelled as affected. 

– Specificity: Proportion of area correctly modelled as unaffected (Degree of underestimation 

or “cautiousness” of the model). 

– Efficiency: Proportion of the sum of all areas correctly predicted by the model (affected 

and unaffected). 


Figure 2: Subarea classification and model performance indices according to the confusion matrix method by Beguerìa (2006). 


Best-fit calibration parameters 

For comparability to other studies, the calibration parameters for Flow-R and RAMMS based 

on which the best-fit model output was achieved are listed below (Table 1). 

Table 1: Employed best-fit calibration values for Flow-R and RAMMS debris flow event modelling. 

Debris-flow extents 

The areas affected by debris flows as modelled with Flow-R show a respectable agreement 

with observed debris flow extents as well as with the results obtained with RAMMS 

(c.f. Richleren torrent, Fig. 3). Although not perfectly matching the extent of the calibration 

event, the modelled flow paths are certainly plausible. Model outputs from both RAMMS and 

Flow-R show a surprisingly high agreement despite the great difference in model conception. 

However, this agreement is restricted to sites, where a low degree of channelization i.e. 

incision of the torrent can be observed on the debris flow fan. In moderately to strongly 


channelized torrents, the non-hydraulic concep-tion of Flow-R composes an important 

disadvantage. Backwater effects occurring due to con-strictions in the channel or due to 

a confinement of the flow path cannot be simulated, resulting in a significantly reduced 

outbreak from the channel (c.f. Glyssibach, Fig. 4). Compared to RAMMS, Flow-R thus 

shows a higher sensitivity towards this “channelization” in the topography. 

Figure 3: Best-fit modelling results (Flow-R & RAMMS) for the 1987 event in the Richleren. 

The quantitative comparison of debris flow modelling results based on confusion matrices 

confirms the good agreement between model output and observed debris flow pattern. 

As visible in Fig. 5, both models reach respectable efficiency values in all study areas. Quite 

contrarily to what might be expected, the sensitivity index (amount of correctly predicted 

affected areas) shows an inverse correlation with achieved efficiency: the model showing a 

lower efficiency uniformly achieves the higher sensitivity value. Additionally the variability 

in the sensitivity index is generally larger. Other than the efficiency, the sensitivity apparently 

depends on the degree of channelization of the tor-rent on the fan. As discussed earlier, this 

condition especially applies for Flow-R. The obtained sensitivity values confirm these findings 

as low sensitivity indices correspond with a higher degree of channelization. 

Since the efficiency values obviously cannot be explained by the sensitivity indices due to 

the inverse proportions, an inclusion of the specificity index is required. The specificity, which 

can also be referred to as the degree of underestimation or the “cautiousness” of a model, 

is higher for RAMMS in larger catchments, while Flow-R shows higher values in smaller 


Figure 4: Best-fit debris flow modelling results (Flow-R & RAMMS) for the 2005 event in the Glyssibach torrent. 

catchments. Furthermore, differences between RAMMS and Flow-R are larger in the 

Minstigerbach and the Glyssibach. The specificity index thus depends both on the catchment 

area as well as on the degree of channel-ization in fan topography. 

In each of the four research areas, a lower specificity index goes hand in hand with a higher 

sensitivity. In other words, the less cautious a model, the more generously it classifies cells as 

affected by the process, which in turn leads to a higher probability of reconstructing areas 

actually affected in reality (true positives in the confusion matrix). However, due to this high 

generosity, the amount of false positives increases too. Thus, the overall efficiency (percentage 

of overall correctly classified cells) does not necessarily increase. Quite contrarily, the 

comparison between efficiency and specificity indices in all four catchments shows that 

higher efficiency values are achieved by the more cautious, i.e. the more underestimating 

model. 

Debris-flow velocity 

Since Flow-R does not consider debris flow mass, obtained debris-flow velocities may only be 

considered relatively. Based on the various model runs conducted in the course of this study 

it can be stated that the spatial pattern of modelled velocities is highly sensitive towards local 

changes in terrain slope. Even small changes in the slope gradient lead to a pronounced 

decline or increase of modelled velocities, which may exceed reasonable ranges. Additionally, 

obtained patterns of ve-locities show a regular pattern over the complete width of the flow on 


Figure 5: Efficiency, sensitivity and specificity indices for best-fit Flow-R & RAMMS model output. 

uniform debris flow fans, while only in pronounced channels, a decrease of velocities towards 

the flow margins can be ob-served. 

Debris-flow magnitudes 

Concerning the modelling of different event magnitudes with Flow-R it needs to be stated 

that without a calibration event, a nonbiased modelling of different magnitudes is practically 

impossible due to the necessarily non-volumetric definition of event scenarios. A definition 

of magnitudes solely based on expected runout distances does not take into account channel 

topography (especially incision) on the debris flow fan, which majorly influences debris flow 

behaviour. The extent of the modelled debris flows therefore mostly reflects an expected 

runout pattern or –distance and does not correspond well with RAMMS model output. 

CONCLUSIONS 

Based on the obtained results it can be stated that due to the non-hydraulic model conception, 

the potential of Flow-R in individual debris flow modelling is limited. Plausible debris 

flow patterns are only achieved on torrents showing a low to moderate channelization, i.e. 

incision on the fan. Computed process velocities are unusable due to the inability of the 

model to account for debris flow mass. Moreover, the modelling output for different debris 


flow magnitudes is biased by the compulsory non-volumetric definition of magnitudes. It 

therefore needs to be emphasized that Flow-R is not suitable for hazard mapping in a narrow 

sense, but it provides first-hand information on debris-flow runout. Despite the restricted 

potential of Flow-R in local debris flow event modelling, an application may thus still provide 

valuable preliminary information on debris flow susceptibility. Based on insights achieved 

with Flow-R debris-flow modelling, the financial expenses connected with a hydraulic debris 

flow simulation may thus be substantially reduced. 

REFERENCES 

- Beguerìa S. (2006). Validation and Evaluation of Predictive Models in Hazard Assessment 

and Risk Management. Natural Hazards 37: 315-329. 

- Christen, M.; Bühler, Y.; Bartelt, P.; Leine, R.; Glover, J.; Schweizer, A.; Graf, C.; McArdell, 

B. W.; Gerber, W.; Deubelbeiss, Y.; Feistl, T.; Volkwein, A. (2012). Integral Hazard Management 

Using a Unified Software Environment. Numerical Simulation Tool “RAMMS” for 

Gravitational Natural Hazards. 12th Congress INTERPRAEVENT 2012, Grenoble. 

- Gamma, P. (2000). dfwalk – Ein Murgang-Simulationsprogramm zur Gefahrenzonierung. 

Geographica Bernensia G66. Geographisches Institut der Universität Bern, Bern. 

- Horton, P.; Jaboyedoff, M.; Rudaz, B.; Zimmermann, M. (2013). Flow-R, a Model for 

Susceptibility Mapping of Debris Flows and Other Gravitational Hazards at Regional Scale. 

Natural Hazards and Earth System Sciences 13: 869- 885. 

- Jakob, M. & Hungr, O. (2005). Introduction. In: Jakob M. & Hungr O. (Eds.). Debris Flow 

Hazard and Related Phenomena. Springer & Praxis Publishing Ltd, Berlin. 

- LLE Glyssibach (2006). Lokale lösungsorientierte Ereignisanalyse Glyssibach. Bericht zum 

Vorprojekt. NDR Consulting Zimmermann & Niederer + Pozzi Umwelt AG, im Auftrag des 

Tiefbauamts des Kantons Bern, Oberingenieurkreis I, Thun. 

- Perla, R.; Cheng, T. T.; McClung, D. M. (1980). A Two-Parameter Model of Snow Avalanche 

Motion. Journal of Glaciology 26: 197-207. 

- Rickenmann, D. & Zimmermann, M. (1993). The 1987 Debris Flows in Switzerland: 

Documentation and Analysis. Geomorphology 8: 175-189. 

- VAW (1992). Murgänge 1987 – Dokumentation und Analyse. Unveröffentlichter Bericht Nr. 

97.6. Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie (VAW), Eidg. Technische 

Hochschule ETH Zürich, Zürich. 

- Zimmermann, M.; Mani, P.; Gamma, P. (1997). Murganggefahr und Klimaänderung – ein 

GIS-Basierter Ansatz. Schlussbericht NFP31. Vdf Hochschulverlag, Zürich. 



Slide-induced impulse waves in mountainous regions 

Helge Fuchs, Dr. 1 ; Robert M. Boes, Prof. Dr. 2 

ABSTRACT 

Impulse waves occur as a natural hazard particularly in mountainous regions with lakes 

surrounded by a steep shoreline. Landslides, rockslides, glacier calvings or avalanches then 

transfer its momentum to the water body thereby generating large tsunamigenic waves. 

At lakes, damages are expected for the shore vegetation or possible infrastructure. In presence 

of a reservoir, potential dam overtopping could lead to downstream damages due to (a) direct 

wave impact, (b) float impact, (c) deposited float and debris or even (d) complete dam failure. 

Using physical model tests conducted within various PhD theses at the Laboratory of 

Hydraulics, Hydrology and Glaciology (VAW) of ETH Zurich, the impulse wave generation 

and the wave-shore interaction were intensively investigated. Based on a computation 

guideline published in 2009, the possible hazard potential may be assessed. The present 

contribution provides an overview on impulse waves and an evaluation of underwater slide 

propagation and deposition features. These data are useful for the calibration of numerical 

models or the above mentioned assessment approach. 

KEYWORDS 

Landslide; Impulse wave; Physical modelling; Wave height; Slide deposition 

INTRODUCTION 

Impulse waves are considered a severe danger in mountainous regions. They are generated 

in lakes or reservoirs following a landslide, rockslide, rockfall, avalanche, or glacier impact. 

The usually short propagation distance and the negligible wave attenuation due to the 

tsunamigenic long wave behavior lead to a massive damage potential. Damages at the 

opposite shore are generated due to (a) direct wave impact on buildings and structures, (b) 

driftwood and float, and (c) their deposits after water retreat. Possible dam overtopping may 

lead to structural damages or even (d) a total dam failure. The generated dam break wave 

may then propagate downstream endangering distant settlements due to widespread flooding. 

The proglacial lake at the Lower Grindelwald Glacier in Switzerland was created due to a 

combination of glacier melt and glacier retreat. The lake level varies over the year with an 

increase starting in spring with rising temperature and thus, glacier melt, a maximum still 

water depth of ≈ 34 m and a decrease in autumn up to complete drainage at the end of a year 

(Hählen 2010). The steep and unstable rock flanks provoke frequent minor slide events. 

A larger 100,000 m³ landslide entered the lake in May 2009 thereby generating impulse 

1 ETH Zürich, Zürich, SWITZERLAND, fuchs@vaw.baug.ethz.ch 

2 ETH Zürich, Zürich, SWITZERLAND 



waves (Fig. 1). According to video records and photographs, the initial splash height was 

determined to ≈ 100 m and wave run-up to ≈ 10 m. 

Figure 1. 2010 Grindelwald glacier lake event with (a) slide impact, (b) run-up traces 

In 2010 a ≈ 300,000 m³ rock-ice avalanche impacted a glacier lake near Carhuaz, Peru. 

Despite a 20 m freeboard the wave overtopped the natural rock-dam with ≈ 1×10 6 m³ of 

water (Schneider et al. 2014). The generated flood wave transported an increasing amount of 

material with increasing propagation distance and finally caused significant damages to the 

village of Carhuaz, located 15 km downstream and 2000 m below the glacier lake (Fig. 2). 

In the course of global warming, glacier retreat creates additional glacier lakes thereby 

increasing the hazard risk for mountainous regions. 

Figure 2. Process scheme of 2010 Carhuaz impulse wave flood event (adapted from Schneider et al. 2014) 

During the planning phase of a reservoir, potential impulse wave events may be physically 

modelled in a laboratory. This procedure requires certain cost and duration and thus is not 

suitable for a quick hazard assessment in case of observed rock or slope instabilities at existing 

reservoirs or lakes. The VAW assessment guideline (Heller et al. 2009) summarizing the past 

impulse wave research at VAW and accounting for a literature review, allows for a quick and 

adequate assessment of slide-induced impulse waves in reservoirs. 

The present contribution provides a general overview on impulse waves and details a recent 

evaluation of underwater slide propagation and deposition features. These data are particularly 

useful for the calibration of numerical models or the above mentioned assessment approach 

using generally applicable equations. 


RESULTS OF PAST VAW IMPULSE WAVE RESEARCH 

Impulse wave events may be modelled either as solid body slides, for which the slide 

characteristics are easily controlled and measured (e.g. Russell 1837, Wiegel 1955, or 

Kamphuis and Bowering 1972) or using granular material, thereby significantly complicating 

the modelling procedure (e.g. Huber 1980, Fritz 2002; Zweifel 2004; Heller 2007; Mohammed 

2010; Bregoli et al. 2013; or Evers and Hager 2015). The simplified use of solid bodies has to 

be carefully selected depending on the prototype slide granulometry. Given a granular slide 

impacts the water with a small slide velocity, a 3-phase flow may be generated, incorporating 

the slide material, the water and air, thereby leading to significant energy dissipation. 

Investigations may either involve a simple 2D test-setup corresponding to a wave channel, 

or a more advanced three-dimensional (3D) setup accounting for complex wave propagation 

features, e.g. reflection, refraction or diffraction. Given such a wave basin (3D setup) the slide 

protrudes radially below the water surface and waves propagate radially from the slide impact 

location. With the wave energy being distributed over an increasing area, the resulting wave 

attenuation is larger as compared with the 2D case. The waves observed in laboratory were 

mainly of Stokes-type, cnoidal-like or solitary-like types. A more detailed review of the 

impulse wave generation process is provided by Heller et al. (2009). 

Figure 3. VAW impulse wave channel (Fritz 2002) 

Starting in 1998, the PhD research cycle on impulse waves at VAW involved a wave channel 

equipped with a pneumatic landslide generator (Fig. 3). This unique setup allowed for the 

independent variation of all basic parameters, e.g. the still water depth h, slide thickness s, 


slide impact velocity V s , bulk slide volume V s , bulk slide density ρ s 

, and slide impact angle α 

(Fig. 4a). The resulting wave height H was then determined for various propagation distances 

x. 

Figure 4. Parameter definition scheme for (a) slide induced impulse wave generation, (b) underwater slide characteristics 

As a result of more than 400 tests, Heller (2007) identified the impulse product parameter P 

involving the slide Froude number F = V s /(gh) ½ , the relative slide thickness S = s/h, and the 

relative slide mass M = m s 

/(ρ w 

Bh 2 ) as the governing parameters for impulse wave generation 

(1) 

with ρ w 

as the water density and B as the slide width. The maximum wave height HM 

representing the vertical distance between the wave through and the wave crest in the near 

field is independent of the propagation direction and results as 

(2) 

The subsequent wave-shore interaction can be treated either as run-up on a vertical wall for 

arch and gravity dams, wave run-up on linearly inclined embankment dams or shore slopes 

(e.g. Hall and Watts 1953; Synolaki 1987; Teng et al. 2000; Goseberg 2011; Baldock et al. 

2012) or overland flow on a connected horizontal plane (Zelt and Raichlen 1991; Schüttrumpf 

and Oumeraci 2005; Sælevik et al. 2013; or Fuchs and Hager 2015). According to the 

latter, the maximum solitary wave run-up r on a linearly inclined shore in terms of the 

vertical distance between still water level and the maximum onshore water surface elevation 

is 

(3) 

The maximum wave run-up height therefore mainly depends on the wave type, the wave 

height H in front of the shore, and the shore slope β. Due to scale limitations in physical 

models additional effects e.g. surface roughness in terms of shore vegetation or shore 

permeability are hard to address. 


OBJECTIVE AND METHODOLOGY OF CURRENT STUDY 

The underwater motion of subaerial granular slides has hardly received any attention in the 

past although the slide deposition might be important for infrastructure and dam safety 

elements, e.g. a reservoir bottom outlet. In addition, valuable information following a 

back-analysis of past slide events is useful to calibrate the calculation procedure or possible 

numerical models. Testing in the VAW impulse wave channel mostly involved video recordings 

for documentation purposes. Videos of 41 tests selected so that the range of initial basic 

parameters is covered best were analyzed regarding the underwater slide dynamics and their 

final deposition patterns. Video still images were corrected for distortion and the coordinates 

of distinctive slide points, namely the slide front position along bottom f 0 

, the maximum slide 

front position f 1 

, the maximum slide thickness s and s′, respectively, and the rear slide 

position b (Fig. 4b) were evaluated over the test duration. The coordinate accuracy of 

±30 mm for the largest underwater slide velocities of V s 

≈ 7.5 m/s was mainly determined by 

the camera exposure time and thus blurry still images (Fig. 5a). For smaller slide velocities 

the accuracy was ±5 mm. 

RESULTS ON UNDERWATER SLIDE PROPAGATION 

Three distinctive tests were selected to demonstrate the variety of underwater slide motion of 

impulse wave events (Fig. 5). Test A corresponds to a slide impact angle of α = 45° and 

therefore represents a high velocity slide at a typical steep mountain shore. Whereas Test B 

corresponds to a vertical slide impact occurring e.g. for rock-fall scenarios, Test C represents 

a moderate α = 30° angle in combination with a small slide velocity, i.e. a slide that is initially 

located only slightly above the still water surface. The main slide parameters at water surface 

contact are listed in Tab. 1. 

Table 1. Initial parameters of Tests A, B, and C 

Due to the high impact velocity, the slide of Test A is compacted when reaching the water 

surface, thereby generating a large splash (Fig. 5 a-b). Subsequently the slide transfers its 

momentum to the water body, thereby creating the impulse wave. Note the vertical water 

column for Tests A (Fig. 5b) and B (Fig. 5f). The generated water crater collapses in outward 

direction (Fig. 5c-d) and the granular slide protrudes underwater and comes to rest at its final 

deposition pattern characterized by the maximum deposition length and thickness (Fig. 5d). 

With the reduced impact velocity of Test B, the splash at water contact is much reduced 

(Fig. 5e) as compared to Test A. Before reaching the channel bottom, the slide is deflected in 


the channel direction (Fig. 5f) and the water crater collapses in inward direction (Fig. 5g-h), 

thereby pushing the slide backwards and thus affecting the deposition shape. 

Due to the small impact velocity of Test C, the slide front is not compacted and almost no 

splash is generated (Fig. 5i). The fluid does not separate from the slide so that no impact 

crater is formed but fluid enters the slide pores leading to a disintegrated slide (Fig. 5j-k). 

The slide back does not reach the water but deposits on the slide ramp (Fig. 5l). 

Figure 5. Still Image sequences for (a-d) Test A, (e-h) Test B, (i-l) Test C; time increment between images is ∆t ≈ 6/25 s 

The underwater slide motion therefore shows large differences depending on the slide impact 

characteristics. However, given the slide front motion is normalized using values of the final 

deposition pattern, i.e. xf0, end as the maximum front position and tend as the time when 

the slide comes to rest and the final deposition is reached, the underwater slide trajectory 

follows in good agreement by (Fig. 6) 

(4) 

where T = t/t end 

In addition to the underwater slide kinematics, characteristic values for the final deposition 

pattern were evaluated. Figure 7 shows the normalized slide front position x′ f0,end 

/h, the 

duration to reach the final position T end 

, and the final deposition thickness s′ end 

/h. A good 

correlation was found to the impact angle-corrected impulse product parameter 


P s 

= P · sin[(6/7)α] (Fig. 7a,c,e) which almost 

corresponds to the governing parameter 

combination for the impulse wave generation 

(Heller et al. 2009). However, a similarly 

good correlation was found for the impact 

angle-corrected slide mass M s 

= M sin[(6/7)α] 

(Fig. 7b,d,f) without including a kinematic 

parameter in the impact characteristics. 

The corresponding equations read 

(R 2 = 0.71) (5) 

Figure 6. Normalized underwater slide propagation 

(R 2 = 0.66) (6) 

(R 2 = 0.26) (7) 

(R 2 = 0.31) (8) 

(R 2 = 0.59) (9) 

(R 2 = 0.71) (10) 

From these observations follows that the larger the slide impact, the larger is the underwater 

slide protrusion, the smaller is the deposition thickness, and the shorter is the duration until 

the final deposition pattern is reached. A more detailed analysis of underwater slide features 

is presented by Fuchs et al. (2013). 

CONCLUSIONS 

Impulse waves may constitute a severe danger to the safety of dams or infrastructure in the 

surroundings of mountainous lakes. Based on the long and systematic expertise of the 

Laboratory of Hydraulics, Hydrology and Glaciology (VAW) including physical model testing 

and a literature review, an assessment guideline for landslide generated impulse waves in 

reservoirs was published in 2009 (Heller et al. 2009). Recently, video recordings of selected 

physical model tests were analyzed to investigate both the underwater slide kinematics and 

the resulting deposition characteristics. Based on estimates for specific slide deposit features, 

e.g. their length and thickness, it is demonstrated that final deposition patterns can be 


Figure 7. Normalized deposition characteristics of (a+b) slide front position, (c+d) propagation duration, and (e+f) deposition thickness 

versus P s 

and M s 

, respectively (*) tests with h ≤ 0.2 m potentially affected by scale effects; (- -) = ±30% 

described using only static input parameters but neglecting the slide kinematics. These results 

apply to calibrate a calculation procedure or to validate numerical models. This work thus 

contributes to the estimation of the risk assessment and to safety aspects relating to waveshore 

interaction mainly in the Alpine environment. 

REFERENCES 

- Baldock, T. E., Peiris, D., and Hogg, A. J. (2012). “Overtopping of solitary waves and solitary 

bores on a plane beach.” Proc. R. Soc. London Ser. A, 468, 3494–3516. 

- Bregoli, F., Bateman Pinzón, A., Medina Iglesias, V., and Gómez Cortéz, D. A. (2013). 

“Experimental studies on 3D impulse waves generated by rapid landslides and debris flow.” 

Ital. J. Eng. Geol. Environ. 6, 115–122. 

- Evers, F.M., Hager, W.H. (2015). “Impulse wave generation: Comparison of free granular 

with mesh-packed slides”. J. Mar. Sci. Eng., 3(1), 100–110. 

- Fuchs, H., Winz, E., and Hager, W. (2013). ”Underwater Landslide Characteristics from 

2D Laboratory Modeling.” J. Waterway, Port, Coastal, Ocean Eng., 139(6), 480–488. 

- Fuchs, H, Hager, W.H. (2015). ”Solitary impulse wave transformation to overland flow”. 

J. Waterway, Port, Coastal, and Ocean Engineering 141(5):04015004 (DOI 10.1061/(ASCE) 

WW.1943-5460.0000294). 

- Fritz, H.M. (2002). “Initial phase of landslide generated impulse waves”. VAW-Mitteilung 

178, H.-E. Minor, ed., ETH Zürich, Zurich, Switzerland. 


- Goseberg, N. (2011) “The run-up of long waves: Laboratory-scaled geophysical reproduction 

and onshore interaction with macro-roughness elements”. Ph.D. thesis, Leibniz Universität 

Hannover, Germany. 

- Hählen, N. (2010) “Moränenabbruch 22. Mai 2009 – Kurzbericht (Moraine slide May 22, 

2009 – Short report)”. Tiefbauamt des Kantons Bern, Thun (unpublished, in German). 

- Hall, J.V., Watts, G. M. (1953) “Laboratory investigation of the vertical rise of solitary waves 

on impermeable slopes.” Technical Memo Rep. No. 33, U.S. Army Corps of Engineers, Beach 

Erosion Board, Washington DC, USA. 

- Heller, V. (2007) “Landslide generated impulse waves - Prediction of near field characteristics.” 

VAW-Mitteilung 204, H.-E. Minor, ed., ETH Zurich, Zürich, Switzerland. 

- Heller, V., Hager, W.H., and Minor H.-E. (2009). “Landslide generated impulse waves in 

reservoirs: Basics and computation”. VAW-Mitteilung 211, R. Boes, ed., ETH Zurich, Zürich, 

Switzerland. 

- Huber, A. (1980) “Schwallwellen in Seen als Folge von Felsstürzen (Impulse waves 

following rockfalls)”. VAW-Mitteilung 47, Vischer, D., ed., ETH Zurich, Switzerland 

(in German). 

- Kamphuis, J. W., Bowering, R. J. (1972). “Impulse waves generated by landslides.” 

Proc., 12th Coastal Engineering Conf., ASCE, New York, 575–588. 

- Mohammed, F. (2010). “Physical modeling of tsunamis generated by three-dimensional 

deformable granular landslides.” Ph.D. thesis, Georgia Institute of Technology, Atlanta, GA. 

- Russell, J.S. (1837) “Report of the committee on waves”. Report of the 7th Meeting of the 

British Association for the Advancement of Science Liverpool 7:417–496, Murray, London. 

- Sælevik, G., Jensen, A., and Pedersen, G. (2013). “Runup of solitary waves on a straight 

and a composite beach.” Coastal Engineering, 77(1): 40–48. 

- Schneider, D., Huggel, C., Cochachin, A., Guillén, S., and García, J. (2014). “Mapping 

hazards from glacier lake outburst floods based on modelling of process cascades at 

Lake 513, Carhuaz, Peru”. Advances of Geosciences 35: 145–155. 

- Schüttrumpf, H.F.R., Oumeraci, H. (2005) “Layer thickness and velocities of wave overtopping 

flow at seadikes”. Coastal Engineering 52(6): 473–495. 

- Synolakis, C.E. (1987) “The run-up of solitary waves“. J. Fluid Mechanics 185: 523–545. 

- Teng, M.H., Feng, K., Liao, T.I. (2000). Experimental study on long wave run-up on plane 

beaches. Proc. 10th Intl. Offshore and Polar Engineering Conf. 3: 660–664. Chung, J.S., 

Olagnon, M., Kim, C.H. eds. ISOPE: California, USA. 

- Wiegel, R.L. (1955) “Laboratory studies of gravity waves generated by the movement of 

a submerged body.” Trans. American Geophysical Union 36(5): 759–774. 

- Zelt, J.A., Raichlen, F. (1991) “Overland flow from solitary waves”. J. Waterway, Port, 

Coastal, and Ocean Engineering 117(3): 247–263. 

- Zweifel, A. (2004) “Impulswellen: Effekte der Rutschdichte und der Wassertiefe (Impulse 

waves: effect of slide density and still water depth)”. VAW-Mitteilung 186, H.-E. Minor, ed., 

ETH Zurich, Zürich, Switzerland (in German). 



Monitoring and management of a huge landslide in a 

built-up area induced by the excavation of deep 

highway tunnels 

Monitoraggio e gestione di una grande frana in area 

urbanizzata indotta dallo scavo di gallerie autostradali 

profonde 

Guido Gottardi 1 ; Giuseppe Ricceri 2 ; Alberto Selleri 3 ; Paolo Simonini 4 ; Daniela Salucci 2 ; Paola Torsello 2 

ABSTRACT 

In 2011 a huge quiescent landside was suddenly reactivated by the excavation of two deep 

parallel tunnels belonging to the new highway, now completed, connecting Bologna to 

Firenze. This involved potentially hazardous situations for the inhabitants of a village lying 

on top of the landslide, inducing increasing anxiety in the population, who repeatedly asked 

for stopping the excavation and looking for new possible highway paths. Such circumstances 

induced the Regional Authorities and the Contractor to install a comprehensive displacement 

real-time monitoring system along the slope. In addition, the structural conditions of each 

building in the area were carefully checked, establishing suitable thresholds on structural 

movements in order to decide about their possible evacuation. At the same time, a group of 

experts was especially appointed to manage the evolving situation. The installed instruments 

allowed for a careful and constant control of the displacement rates of soil and structures and 

of their evolution with time, enabling the construction of both tunnels to proceed, even at 

reduced excavation rate, until their successful completion. 

RIASSUNTO 

Nel 2011 una grande frana quiescente fu improvvisamente riattivata dallo scavo di due 

gallerie profonde e parallele, facenti parte del nuovo tratto autostradale – ora completato - 

tra Bologna e Firenze, noto come Variante di Valico. Ciò implicò situazioni di potenziale 

pericolo e di preoccupazione per gli abitanti del paese soprastante, che chiesero ripetutamente 

la sospensione delle operazioni di scavo e l’individuazione di un tracciato alternativo. 

Tali circostanze indussero quindi gli enti competenti e le autorità regionali a richiedere 

all’impresa appaltatrice la progettazione e l’installazione di un articolato sistema di monitoraggio 

in tempo reale degli spostamenti del versante. L’assetto strutturale di ogni edificio 

è stato altresì tenuto sotto stretta osservazione, individuando delle soglie di riferimento sui 

1 University of Bologna, ITALY, guido.gottardi2@unibo.it 

2 Osservatorio Ambientale 

3 Autostrade per l‘Italia 

4 University of Padova 



elativi movimenti al raggiungimento delle quali procedere con la delocalizzazione degli 

abitanti. L’attento e costante controllo delle velocità di spostamento e della loro evoluzione 

nel tempo ha così consentito alle operazioni di scavo di procedere in sicurezza, anche a 

ritmi ridotti, fino al completamento dell’opera. 

KEYWORDS 

landslide; monitoring; emergency plan 

INTRODUZIONE 

Nel 2002 la Società Autostrade per l’Italia avviava i lavori di realizzazione della variante 

all’autostrada A1 tra Sasso Marconi e Barberino di Mugello, la cosiddetta Variante di Valico 

(VAV), il cui tracciato si compone di 23 viadotti e 22 gallerie, per uno sviluppo totale di circa 

58 km (Autostrade per l’Italia, 2006). La Variante attraversa per 43 km l’Emilia-Romagna 

e per 15 km la Toscana (Figura 1). 

Figura 1. Ubicazione del progetto della Variante di Valico e della Galleria Val di Sambro. 

Contestualmente all’inizio dei lavori veniva istituito, tramite un accordo ufficiale sottoscritto 

dal Ministero dell’Ambiente e dagli Enti locali, un Osservatorio Ambientale (OA) col compito 

di controllare l’operatività e l’efficacia del Piano di Monitoraggio Ambientale (PMA) attivato 

da Autostrade e di verificare l’impatto prodotto dai cantieri sul contesto ambientale e socioeconomico. 

La VAV si inserisce infatti all’interno di un territorio antropizzato, vario ed articolato, carat- 


terizzato dalla presenza dei massicci montuosi dell’Appennino tosco-emiliano, separati in 

genere da valli strette e solcate da corsi d’acqua a carattere torrentizio. Tale contesto geomorfologico 

è contraddistinto dalla presenza di innumerevoli frane, alcune delle quali di 

tipo complesso e di notevoli dimensioni (Bertolini et al., 2001). A presidio di questo scenario 

il PMA prevedeva dunque il monitoraggio di 10 componenti distribuite tra i settori antropico, 

idrico, geotecnico e naturale, con campagne di misura in continuo ovvero a frequenza 

variabile, a seconda della matrice monitorata. Dal punto di vista specifico dell’assetto del 

territorio, il PMA ha distinto i siti degni di attenzione in: “aree sensibili”, interessate da 

fenomeni franosi attivi, inattivi o quiescenti, e “opere sensibili”, ossia gli imbocchi delle 

gallerie naturali, gli scavi sui versanti o in adiacenza a strutture e/o infrastrutture 

(Barla et al., 2006). 

Il presente contributo si propone di presentare, sebbene a grandi linee, il complesso piano di 

emergenza messo a punto in occasione della riattivazione di un movimento di versante profondo 

indotto dalla realizzazione delle due canne affiancate denominate “galleria Val di 

Sambro”, che ha interessato le frazioni soprastanti di Santa Maria Maddalena e di Serrucce, 

in Comune di San Benedetto Val di Sambro (BO). Tale piano di emergenza ha preso l’avvio 

dall’installazione di un esteso sistema di monitoraggio topografico e geotecnico - costituito 

da strumentazione profonda, quali piezometri ed inclinometri - e dalla definizione di predefiniti 

valori di soglia in corrispondenza dei quali attivare una specifica procedura mediante 

l’azione congiunta di un appositamente costituito Collegio dei Tecnici, Amministratori locali 

e Autorità preposte. 

INQUADRAMENTO GEOLOGICO DEL SITO E CARATTERISTICHE DELL’OPERA 

La galleria Val di Sambro si sviluppa per 3.8 km circa con due corsie di marcia ed una corsia di 

emergenza. Lo scavo, avviato nel 2007 e terminato nel novembre 2014, è stato eseguito con 

metodo tradizionale. Il tracciato interessa in profondità il versante su cui poggiano i villaggi 

di Santa Maria Maddalena e di Serrucce (Figura 2), con coperture medie dell’ordine di 100 m. 

L’area di Santa Maria Maddalena è caratterizzata da unità geologiche costituite da depositi di 

frana perlopiù quiescenti e dal substrato roccioso costituito dalla formazione del flysch di 

Monghidoro. Dai rilevamenti geologici e dai sondaggi geomeccanici eseguiti in fase progettuale, 

l’asse della galleria attraversa un substrato costituito da torbiditi arenaceo-pelitiche in strati 

da sottili a molto spessi, che nell’area in esame sono prevalentemente arenacee, con granulometrie 

da medie o fini passanti localmente a grossolane, di colore grigio chiaro o bruno, 

passanti a peliti, spesso siltose, di colore grigio scuro (Autostrade per l’Italia, 2007). 

Le Figure 3a e 3b mostrano le sezioni geologiche denominate A-A e B-B, con la posizione 

delle due canne della galleria. 

Nello stesso tratto le coperture superficiali sono definite, in base alla loro dinamica geomorfologica, 

come corpo di frana attiva sovrapposto a un corpo di frana quiescente, precedute da 

una zona intensamente tettonizzata. 


Figura 2. Planimetria dell’area oggetto di monitoraggio. 

IMPLEMENTAZIONE DEL PMA 

Per la galleria Val di Sambro, il Piano di Monitoraggio Ambientale (PMA) prevedeva, all’inizio 

dei lavori di scavo, l’attivazione di strumentazione geotecnica prevalentemente a controllo 

dell’imbocco meridionale, dove la galleria attraversa una zona in frana. Per la restante parte 

dello scavo, il progetto indicava che la galleria fosse sufficientemente profonda da sottopassare 

i movimenti franosi quiescenti presenti lungo il tracciato. 

Con l’avanzare dello scavo da nord si evidenziarono però i primi significativi dissesti in alcuni 

edifici isolati che sorgevano nella zona denominata come ‘frana attiva’ in Figura 2. A seguito 

di questi primi movimenti di versante veniva redatto un piano di monitoraggio geotecnico 

specifico per l’abitato di Santa Maria Maddalena, anche in vista dell’approssimarsi al paese 

dello scavo della galleria. 

In particolare, la strumentazione prevista a più riprese per tale monitoraggio integrativo del 

versante comprendeva complessivamente: 

– n. 31 verticali inclinometriche con lettura manuale, spinte fino a 120 m di profondità; 

– n. 11 verticali inclinometriche strumentate con catene inclinometriche, profonde fino 

a 70 m; 

– n. 3 verticali multiparametriche tipo “DMS”, profonde 100 m; 

– n. 22 piezometri tipo Casagrande più profondi e numerosi piezometri a tubo aperto più 

superficiali; 

– n. 397 miniprismi installati su edifici e viadotto autostradale esistente e n. 255 pilastrini 

sul terreno, per un totale di 652 mire ottiche installate; 

– n. 15 stazioni GPS. 


Figura 3. Sezioni geologiche trasversali: a) A-A; e b) B-B; in prossimità di Santa Maria Maddalena. 

L’ubicazione degli strumenti per il monitoraggio geotecnico è riportato in Figura 2. Le misure, 

interpretate in termini di velocità di spostamento medie mensili, venivano costantemente 

raffrontate con predefiniti valori di soglia. Il controllo dei superamenti di soglia previsti ai fini 

di monitoraggio ambientale veniva attuato, sia per le misure sui fabbricati che per le verticali 

inclinometriche e le misure topografiche, mediante un sistema di monitoraggio automatico 

in tempo reale. Il raggiungimento od il superamento di tali valori di soglia comportava 

l’incremento della frequenza anche delle letture in manuale, fino ad una o più misure alla 

settimana, e la convocazione di incontri tecnici aventi il compito di definire eventuali azioni 

da intraprendere, quali la modifica delle modalità di avanzamento degli scavi in galleria o 

interventi integrativi a partire dalla superficie del versante. 

Nel luglio 2011 il fronte di scavo del tunnel nord della galleria Val di Sambro entrava nell’area 

di influenza della frazione di Santa Maria Maddalena, in una zona classificata come “frana 

in blocco” nella cartografia geologica regionale, caratterizzata da elementi analoghi alle 

deformazioni gravitative profonde di versante. L’originario modello geologico di progetto, 

revisionato sulla base dei primi risultati del monitoraggio, individuava la presenza di una 

paleofrana riattivata dal passaggio dei fronti di scavo della galleria (Autostrade per l’Italia, 

2011). Ne conseguiva una stima degli spostamenti attesi inevitabilmente superiori rispetto 


a quelli inizialmente previsti. Il movimento risultava comunque caratterizzato da velocità 

di spostamento relativamente modeste, che non facevano presupporre possibili fasi parossistiche. 

Con l’avvicinarsi del fronte di scavo al villaggio di Santa Maria Maddalena, le misure 

convergevano nel segnalare un movimento, coinvolgente più di 10 milioni di m³ di terreno, 

situato perlopiù a grandi profondità (intorno ai 50-60 m ed oltre), dai contorni geometrici 

non ancora completamente definiti ed in progressiva evoluzione, caratterizzato da persistenti 

velocità di spostamento intorno a pochi mm/mese, in parte chiaramente influenzate dalle 

operazioni di scavo in corso. In Figura 4, che mostra una sezione trasversale al tracciato, 

è possibile osservare la forma della superficie di scivolamento, i cui limiti (indicati in figura 

con linea tratteggiata) sono ben individuati dalle verticali dove sono stati installati gli 

inclinometri. Le misure inclinometriche definivano superfici di scivolamento caratterizzate 

da localizzazioni piuttosto nette su bande di taglio sottili, sulle quali la resistenza mobilitata – 

valutata tramite verifiche di stabilità a ritroso - appariva in linea con quanto emerso dalla 

determinazione sperimentale della resistenza al taglio di tipo residuo. 

Figure 4. Sezione C-C (cfr. figura 2) con l’indicazione della strumentazione installata. 

PIANO DI EMERGENZA COMUNALE 

Considerate le possibili conseguenze sulla sicurezza degli abitanti di Santa Maria Maddalena, 

la prosecuzione dei lavori venne subordinata all’integrazione del sistema di monitoraggio con 

uno specifico piano di emergenza. 

Nell’agosto 2012 tale piano veniva formalizzato nel “Piano di emergenza comunale per 

l’abitato di Santa Maria Maddalena” redatto sulla base di specifiche soglie di attenzione, 

preallarme e allarme definite in relazione agli esiti del monitoraggio degli edifici e del relativo 

versante. I valori di soglia individuati ai fini del piano di emergenza sono riportati in Tabella 1: 

per gli edifici, le soglie riguardano i gradienti di spostamento verticale ed orizzontale (espressi 

in millesimi) e sono state ricavate dalla letteratura tecnica con intenti particolarmente 


Tabella 1. Valori soglia di spostamento per gli edifici e delle velocità per il movimento di versante. 

Soglie relative al gradiente degli spostamenti per gli edifici 

Tipo di soglia 

gradiente di 

spostamento verticale 

differenziale 

gradiente di spostamento 

orizzontale differenziale 

Attenzione 1.00 0.45 

Preallarme 1.50 0.70 

Allarme 3.35 1.50 

Soglie di velocità profonda e superficiale per il movimento di versante 

Tipo di soglia Profonda Superficiale 

cautelativi in base alla classe di danno associata, mentre per il movimento franoso le soglie 

riguardano le velocità di spostamento sia a piano campagna che in profondità, nelle zone 

dove è localizzata la superficie di scivolamento della frana. Quest’ultime sono state definite in 

base alla natura dei terreni interessati e sulla scorta delle evidenze registrate in casi analoghi; 

i loro valori tengono evidentemente conto delle azioni conseguenti e differiscono sostanzialmente 

dalle soglie utilizzate ai fini di monitoraggio ambientale. 

Lo scopo del Piano di Emergenza Comunale era quello di definire le disposizioni organizzative 

per la preparazione e la gestione di eventuali situazioni di emergenza riferite al versante da 

parte di tutti gli enti e le strutture tecniche coinvolte. 

Il Piano Comunale di Emergenza individua per ciascun soggetto coinvolto le azioni da 

intraprendere a seconda dei tre scenari possibili: di attenzione, di preallarme o di allarme. 

Lo scenario di rischio più probabile considera la possibilità di un’evacuazione controllata 

e pianificata di poche o singole unità. Il numero totale degli edifici oggetto di monitoraggio 

è pari a 125, che ospitano 160 nuclei familiari e 339 persone abitualmente presenti. 

La procedura prevista in caso di superamento delle soglie di “allarme” è stata sintetizzata 

e schematizzata in Figura 6, nella quale sono state introdotte solamente le azioni ritenute più 

significative. In caso di superamento delle soglie di “allarme” per il versante (verifica a cura 

della Direzione Lavori) o per gli edifici (verifica a cura dell’impresa esecutrice), il Sindaco 

rinvia al Collegio dei Tecnici la verifica dello stato del versante e dell’entità dei danni alle 

abitazioni. Il Collegio dei Tecnici informa il Sindaco e la Protezione Civile delle verifiche 

eseguite e, in caso di accertato pericolo per la popolazione, comunica la necessità di delocalizzare 

la popolazione stessa. 

velocità di spostamento 

cm/giorno 

velocità di spostamento 

cm/giorno 

Attenzione 0.50 1.00 

Preallarme 1.20 2.00 

Allarme 3.00 4.00 

Al termine, completate le operazioni previste dalla procedura di “allarme”, sarà il Servizio 

Geologico Regionale a proporre la cessazione dello stato di allarme al Sindaco, il quale ne darà 

informazione a tutti gli Enti interessati ed alla popolazione delle località interessate. 


Figure 5. Spostamenti differenziali locali nella fascia di deformazione. 

CONCLUSIONI 

Nel presente lavoro è stato presentato un caso di riattivazione di un movimento profondo 

di versante, innescatosi a seguito dello scavo per la realizzazione della galleria Val di Sambro, 

e la sua interazione con gli abitati soprastanti. La frana riattivata è caratterizzata da un 

movimento molto profondo, con superficie di scorrimento che si attesta anche oltre 50-60 m 

di profondità. Date le dimensioni del corpo di frana e l’elevato volume di terreno coinvolto, 

era pressoché impossibile ricorrere a interventi di stabilizzazione del versante che arrestassero 

completamente il fenomeno in atto. Pertanto, al fine di preservare l’integrità degli edifici 

e l’incolumità della popolazione degli abitati interessati, è stato implementato un complesso 

piano di emergenza, basato sulla misurazione in campo di parametri ritenuti significativi, sul 

confronto di tali misure con valori di soglia predefiniti ed infine sull’avvio di una procedura 

di emergenza a seconda del livello di pericolo riscontrato. 

In prima fase si è quindi proceduto all’installazione di un accurato ed articolato sistema di 

monitoraggio geotecnico, costituito da strumentazione profonda per il controllo dell’evoluzione 

dei movimenti di versante, e topografico mediante strumentazione di precisione per il 

controllo dei danni strutturali agli edifici. In seconda fase si sono definiti i livelli di soglia, 

individuando due parametri significativi: la velocità di spostamento del versante, distinguendo 

tra movimento profondo e superficiale, ed il gradiente di spostamento differenziale, sia 

orizzontale che verticale, per gli edifici. Per entrambi i parametri sono stati definiti, ai fini 

della redazione del piano di emergenza, tre diversi valori soglia corrispondenti a tre livelli 

di pericolo denominati di attenzione, preallarme e allarme. In terza ed ultima fase è stato 


Figure 6. Piano di emergenza: procedura in caso di superamento della soglia di allarme. 

implementato un Piano di Emergenza Comunale, diversificato in funzione del livello di 

pericolo registrato, con lo scopo di definire le disposizioni organizzative necessarie per la 

preparazione e la pianificazione di eventuali situazioni di emergenza, come ad esempio 

l’evacuazione dei nuclei abitativi e la loro ricezione in strutture di accoglienza già predisposte. 

L’acquisizione delle registrazioni in automatico, per 24 ore al giorno, ha consentito, 

e consente tuttora, una valutazione in tempo reale della situazione in atto e l’eventuale 

immediata attivazione del piano di emergenza programmato, a seconda del livello di 

emergenza riscontrabile. Nel corso dei lavori, tuttavia, non si sono mai verificate condizioni 

effettive di superamento delle soglie predefinite, fatta eccezione per segnalazioni isolate 

e sporadiche che, dopo attenta verifica, sono risultate prive di significato. 

Con l’abbattimento dell’ultimo diaframma della canna sud della galleria Val di Sambro 

avvenuto nel novembre 2014, lo scavo della galleria Val di Sambro risulta oggi interamente 

completato. Il passaggio dei fronti in corrispondenza dell’abitato di Santa Maria Maddalena 

ha determinato velocità medie di spostamento del versante fino a 8-10 mm/mese e oltre nel 

2011-2012, poi ridottesi fino a 1-2 mm/mese nel biennio 2013-2014. Attualmente, la fitta 

rete di monitoraggio attiva sul versante conferma, in generale, la tendenza al sostanziale 

rallentamento dei movimenti profondi. 


BIBLIOGRAFIA 

- Autostrade per l’Italia (2006). Variante di Valico, Newsletter (https://www.autostrade.it/ 

variante-di-valico/newsletter-variante-di-valico/index.html). 

- Autostrade per l’Italia (2007). Autostrada A1 Milano – Napoli. Adeguamento del tratto di 

attraversamento appenninico tra Sasso Marconi e Barberino di Mugello. Tratto: La Quercia – 

Badia Nuova. Sub-tratta: Lagaro – Badia Nuova. Lotto 6-7, Progetto Esecutivo., Galleria Val 

di Sambro, Relazione Geomeccanica. 

- Autostrade per l’Italia (2011). Autostrada A1 Milano – Napoli. Adeguamento del tratto di 

attraversamento appenninico tra Sasso Marconi e Barberino di Mugello. Tratto: La Quercia – 

Badia Nuova. Sub-tratta: Lagaro – Badia Nuova. Lotto 5B, Progetto Esecutivo, Galleria Val 

di Sambro, Perizia di variante tecnica 2. 

- Barla G., Chiappone A., Vai L. (2006). Slope monitoring system. Instabilità di versante. 

Interazioni con le infrastrutture, i centri abitati e l’ambiente. XI Ciclo di Conferenze di 

Meccanica e Ingegneria delle Rocce (MIR 2006), Cap. 8: 177–202. 

- Bertolini G., Pellegrini M. (2001). The landslides of the Emilia Apennines (northern Italy) 

with reference to those which resumed activity in 1994–1999 period and required Civil 

Protection interventions. Quad. Geol. Appl. 8: 27–74. 



Adaption and further development of the numerical 

solution in the avalanche simulation model SamosAT 

Matthias Granig, DI 1 ; Peter Sampl, DI Dr. 2 ; Andreas Kofler, DI 3 ; Jan-Thomas Fischer, DI 3 ; Philipp Jörg, Mag 4 

ABSTRACT 

The avalanche simulation model SamosAT is in practical application for the assessment of 

potential avalanche hazards since 2008. Step by step the tool has been further developed. 

The current software release 2015 of SamosAT contains new improvements, especially to 

optimize the modeling of the dense flow part. The new release focuses on enhancements 

in the numerical solution as well as on the modeling of avalanches by implementing new 

boundary conditions. Additional features are optimizing the handling of the program. 

This paper gives an overview of the main new features and procedures of the SamosAT 

update including first work experiences and recommendations for the practical application. 

KEYWORDS 

avalanche modelling, SamosAT, avalanche simulation, model improvements 

INTRODUCTION 

The avalanche simulation software SamosAT (Snow Avalanche MOdelling and Simulation 

- Advanced Technology) is an approved tool for the assessment of potential avalanche hazards 

for many years. The SamosAT modeling recommendations were updated according to the 

actual findings (see Granig, Sauermoser, 2009; Joerg et al., 2010). The program release of 

SamosAT in March 2015 gives new possibilities in the simulation. In this paper we concentrate 

on the major enhancements of the model. Therefore we explain the newly implemented 

Smoothed Particle Hydrodynamics (SPH) function in more detail. So far the SamosAT model 

worked with a predefined end time criterion. Now a stop criterion based on the flow energy 

has been introduced to optimise the simulation time and to avoid premature simulation cut 

offs. Furthermore the mountain snow cover (MSC) approach (Fischer, 2013), that provides a 

continuous initial snow distribution, which can be used as boundary condition for avalanche 

release and entrainment is implemented in the SamosAT software. This feature allows an easy 

application and quick comparisons with the standard avalanche release for the analysis of 

the results. 

1 Avalanche and Torrent Control (WLV), Innsbruck, AUSTRIA, matthias.granig@die-wildbach.at 

2 AVL List GmbH, Graz 

3 Federal Research and Training Centre for Forests, Natural Hazards and Landscape, Innsbruck 

4 tur gmbh, Davos 



METHODS 

The lateral pressure forces within the SamosAT dense flow avalanche (DFA), acting tangentially 

to the terrain surface and in direction against the flow depth gradient, have been 

calculated in a simplified way in SamosAT. Under certain circumstances this has resulted in 

artificial “fingers” in the predicted DFA deposition. Figure 1 shows an example of such a 

finger on the orographic left side of the simulation. This problematic behaviour has been 

corrected by an improved calculation of these forces. The improved result is shown in figure 

2. Also the overrun of the ridge in the avalanche runout on the orographic right side is 

reduced in the SPH mode. 

Figure 1: SamosAT DFA standard simulation with lateral spreading 

Figure 2: SamosAT DFA simulation with SPH mode 

In order to explain what has been changed, some background information on the SamosAT 

simulation needs to be given. 

SamosAT, like for instance the model of Savage and Hutter (1989) or Hungr (in Harbitz et al. 

1998), employs a Lagrangian method to compute the DFA movement instead of the more 

common Eulerian method, as used for instance in RAMMS (Christen et al. 2008). The 

essence of the Lagrangian method is that the control volumes (CVs), used to apply the 

fundamental equations of conservation of mass and momentum, are attached to the fluid 

mass and move with the fluid, while in the Eulerian method the CVs are grid cells attached 

to the terrain model and hence have a fixed position. Since the avalanche typically covers a 

small part of the terrain only at a given time, the Eulerian method wastes storage and 

computational effort by filling the entire terrain with CVs. The Lagrangian CVs are always 

concentrated in the avalanche, effectively rendering a higher numerical resolution with the 

same number of CVs. Furthermore, the Lagrangian method is simpler to implement since 

there is no mass transfer between the CVs. With the Euler method, the mass flow between 

the CVs has to be computed for each time step. The disadvantage of the Lagrangian method, 

however, is that the momentum transfer between the CVs due to lateral friction and pressure 

is much harder to compute, since neither the shapes of the CVs nor their boundary surfaces 

are actually known precisely, only the centres of gravity of the CVs are. Fortunately, for 

shallow flows like the DFA, it can be shown (see Savage & Hutter 1991) that lateral friction 

and pressure forces are much smaller than the main forces, which are gravitational acceleration, 

bottom pressure and bottom friction. Indeed lateral friction is so small that it is ignored 

by most of the avalanche models altogether (see Harbitz et al. 1998), and lateral pressure 


needs not be calculated with full accuracy. This fact finally makes the Lagrangian method 

very attractive for DFA simulations. 

In the momentum balance for a CV, the pressure force Fp may be formulated as 

F p = -mg∇h 

with m the mass of the CV, g the gravitational acceleration and ∇h the gradient (vector) of the 

flow depth. SamosAT used a simple and fast method to compute the depth gradient. First, the 

masses of all CVs were assigned to points of an auxiliary uniform grid according to an inverse 

bilinear interpolation. Then the flow depth was computed from the summed masses at each 

grid point and finally the depth gradient was computed from a bilinear interpolation from the 

depths at the grid points. This works fine as long as the flow depth is not too small or the 

main forces are large compared to the lateral pressure. But for a single CV in the runout zone, 

where the main forces tend to zero, this method always gives a depth gradient and pressure 

force which drives the CV towards the centre of the containing auxiliary grid cell. The 

gradient is zero at the cell centre only, which is clearly unphysical. It should be zero everywhere 

for an isolated single CV. In presence of shallow channels in the runout which are 

aligned with the grid lines, this computation method even leads to the formation of “chains” 

of CVs. To resolve this problem, the depth gradient is now computed according to the 

Smoothed Particle Hydrodynamics (SPH), a Lagrangian method, which is well known in fluid 

dynamics (Monaghan 1992). According to SPH, the flow depth h at any terrain point x may 

be computed as the sum over all CVs, 

h(x)=∑m p ρ -1 w(r xp ) 

with m p 

the mass of the p-th CV, ρ the flow density and w a “kernel function” dependent on 

the distance r xp 

between x and the centre of the CV. The kernel function (figure 3, left) has 

the dimension m -2 and is chosen such that it drops to zero for distances larger than a 

smoothing length r k 

, which is determined to be identical to the grid size of the terrain model 

in SamosAT (usually 5 m). Only CVs within this distance need to be considered in the sum 

(figure 3, right). 

Taking the gradient of this depth function yields 

∇h(x)=∑m p ρ -1 ∇w(r xp ) 

This relation is used in SamosAT to compute the depth gradient at each CV-centre and the 

pressure force acting on it. This method avoids any auxiliary grid and the formation of 

grid-aligned “chains” and gives a zero depth gradient for a single CV with no other CVs closer 

than the smoothing length. It generally leads to a more homogenous distribution of the CVs. 

The method is slower than the original one, because all CVs closer than the smoothing radius 

have to be determined for each CV at each time step. Using fast searching algorithms this 


Figure 3: SPH kernel function used in SamosAT and CVs contributing to the flow depth 

overhead can be reduced such that DFA simulation times increases by about 20% to 50%. 

The SPH method gives a substantially more homogeneous distribution which is numerically 

desirable. 

Another improvement in SamosAT is the implemented stop criterion, such that the simulation 

is automatically terminated as soon as the avalanche comes to a rest. The model 

optionally stops as soon as the kinetic energy of the entire avalanche drops below a user 

defined fraction (typically 2-3%) of the maximal kinetic energy that occurred in the running 

simulation up to this point. Using the running maximum as reference ensures that the 

simulation is neither stopped prematurely in the release phase nor continued too long due 

to trailing avalanche parts still slowly moving when the main part has already stopped. 

The criterion is checked separately for the DFA-part and the powder-snow-part (if active), 

and the stop is performed only if the condition holds for both parts at the same time. The 

energy threshold values are derived pragmatically to ensure sufficient calculation steps and to 

save calculation time. A 2 % threshold value is good to be on the rather safe side, especially 

for smaller avalanches with a lower energy maximum to avoid early cut offs. In the previous 

versions it was necessary to predefine an end time (usually 200 sec.). Consequently after the 

simulation it was required to check, whether the assumed end time was sufficient. This 

procedure was time consuming and led in some cases to too short avalanche runout 

simulations. 

The mountain snow cover (MSC) approach was implemented in SamosAT as an additional 

feature to provide an alternative boundary condition for the snow mass input. This approach 

assumes a smooth snow cover distribution (hMSC) over the whole digital terrain model. It 

has the advantage that the definition of especially the lower end of an avalanche release area 

has less importance, because the potential snow entrainment provides similar snow heights 

along the avalanche path. Though the concept works properly it needs careful handling as a 


consequence of the shift of release mass to rather more snow entrainment. With the new 

implementation of this concept it can be easily tested and studied as an alternative simulation 

result. More details about the idea and the simulation of MSC can be derived from Fischer 

(2013) and Fischer et al. (2014). 

RESULTS AND CONCLUSIONS 

The tests with the SPH mode in SamosAT displayed a compact runout behavior as shown in 

figure 1 also for avalanches in complex terrain. This is in some cases resulting in slightly 

shorter avalanche (DFA) runout calculations. In a next step the avalanche simulation with 

SPH needs a comparison with the reference avalanche data pool for systematic verification. 

Therefore we recommend the practical application of the SPH mode as an additional scenario 

to the standard simulations in conjunction with a careful plausibility check to use the 

simulation for further assessments. 

So far the MSC approach usually calculates a rather big avalanche runout behavior in 

comparison to the documented avalanche events. A possibility is to combine the mountain 

snow cover approach with the SPH module. The first step is done to introduce the feature of 

the mountain snow cover simulation in the SamosAT program. Now further analysis can be 

done to verify and to develop a practical application procedure also for the standard simulation 

routine. 

The stop criterion is working as planned and can already be used in the practical application. 

We recommend a defined fraction of the kinetic energy of 2% for both dense flow and 

powder flow avalanches. Hence premature runout cut offs can be minimized with the 

stopping criterion. The computational time benefit is though smaller as expected. Still for 

computing powder snow avalanches which takes several hours the time saving is in order of 

10-15% of the total simulation time. Now with this new feature early cut offs can be avoided. 

Furthermore it reduces a potential source of error. 


LITERATURE 

- Christen, M., Bartelt, P., Kowalski, J., Stoffel, L., 2008: Calculation of dense snow avalanches 

in three-dimensional terrain with the numerical simulation program RAMMS. In: 

International Snow Science Workshop 2008, Proceedings. September 21-27, 2008. Whistler, 

BC, CAN. 709-716. 

- Fischer JT., Kofler A., Fellin W., Granig M., Kleemayr K. (2014): Optimization of Computational 

Snow Avalanche Simulation Tools. Proceedings, International Snow Science 

Workshop (ISSW), Banff, Canada, p. 665-669. 

- Fischer, J.-T. (2013): A novel approach to evaluate and compare computational snow 

avalanche simulation. Natural Hazards and Earth System Science, 13(6):1655–1667. 

Granig, M., Sauermoser, S. (2009): Ein Erfahrungsbericht über die Lawinenmodellierung aus 

der aktuellen praktischen Arbeit der WLV. Wildbach und Lawinenverbau, 73. Jahrgang, 

Heft Nr. 163., pp. 142-151, Salzburg. 

- Harbitz C.B. et al. (1998): A Survey of Computational Models for Snow Avalanche Motion, 

Deliverable No. 4 of the EU Programme SAME (Snow Avalanche Modeling, Mapping and 

Warning in Europe), Norwegian Geotechnical Institute, Oslo. 

- Joerg P. et al. (2010): Implementation of a new suspension model and calibration of the 

powder snow layer in SamosAT, European Geosciences Union, General Assembly 2010, Wien. 

Monaghan J.J. (1992): Smoothed Particle Hydrodynamics. Ann. Rev. Astron. Astrophys 

(1992). 30 : 543-74. 

- Sampl P. (2015): Abschlussbericht: Erweiterung der Lawinensimulationssoftware SamosAT, 

AVL, Graz. 

- Savage S. B., Hutter, K. (1991) The dynamics of avalanches of granular materials from 

initiation to runout. Part I: Analysis, Acta Mechanica 86, 201-223. 

- Savage S.B. & Hutter K. (1989): The motion of a finite mass of granular material down 

a rough incline. J. Fluid Mech. 199, 177-215. 



An Effective Camera Based Water level recording 

Technology for Flood Monitoring 

Issa Hasan, Dr. Eng. 1 ; Thomas Hies, Dr. 2 ; Ebi Jose 2 ; Rudolf Duester 3 ; Marcus Sattler 3 ; Matthias Satzger 3 

ABSTRACT 

Beside the technical flood protection constructions, a timely warning is a crucial factor for 

preven-tion or reduction of damages caused by flood disasters. The monitoring of water level 

at dams and dikes provides information to determine their water balance, which is one of the 

crucial factors for their stability. A continuous monitoring of surface water level is therefore 

very necessary and can be carried out by different methods. The optical water level measurement 

is a new method for this purpose. The aim of this work is to develop a new effective 

method for a continuous contactless water level measurement even under critical conditions 

like floods and hydraulic jumps, under which other localised gauging methods could be 

temporal out of operation or not sufficiently rep-resentative. In particular the access to 

images of the site, which are the basis of the new gauge method, will allow investigating the 

situation and judging on the quality of the water level meas-urement. The developed method 

was successfully validated at several sites and under different conditions. 

KEYWORDS 

water level;continuous monitoring;contactless measurement;optical method;floods 

INTRODUCTION 

Flood risk management has an increasing importance in many areas of the world, especially 

under the consideration of the potential influence of climate change and the reduction of 

natural river meadows caused by human activities. “Floods have the greatest damage 

potential of all natural dis-asters worldwide and affect the greatest number of people” (unsidr, 

2002). A recent confirmation of this fact is the June flood, which was with 15.2 billion US$ 

the most expensive natural disaster in 2013 in Germany (Munich RE, 2014). Very large 

amounts of precipitation caused flooding in most river basins in Germany (BfG, 2013). 

Therefore, an effective and timely monitoring of surface water level is a very crucial factor of 

the quantitative flood risk management for the reduction of potential damages. Especially in 

potential flooding areas, e.g. near rivers and dams, is a continuous surface water monitoring 

with an effective alarm system very necessary. The monitoring of surface water level at dams 

and dikes provides information to determine their water balance, which is one of the crucial 

factors for their stability. Physical and numerical models have shown that even partially 

1 SEBA Hydrometrie, Kaufbeuren, GERMANY, hasan@seba.de 

2 2DHI Water & Environment (S) Pte Ltd, Singapore 

3 SEBA Hydrometrie GmbH & Co. KG, Germany 



saturated conditions on the air side of dams and dikes could cause mechanical instabilities 

(Hasan et al., 2012). 

Monitoring of surface water level can be carried out by different methods. Pressure sensors, 

floaters, radar sensors and ultrasonic techniques are some of the most applied methods 

providing accurate and reliable data acquisition for water level measurement. A disadvantage 

of these methods is the regular visual inspection of the site required due to environmental 

changes. The optical water level measurement is a new method for this purpose, which has 

been investigated more widely in the recent years. A significant example of water level 

detecting based on image processing is the camera system of HydroPix Monitoring, which is 

installed in a closed sewer system having the ad-vantage that light conditions are stable 

without any day and night time changes and that electrical power connection is available 

(Nguyen, 2009). Further optical gauge system is the GaugeCam, which is designed for open 

channels and tested in the laboratory and at one site and proposes the usage of infrared light 

for night applications (www.gaugecam.com). 

Other applications like RiverBoard (www.tenevia.com) use server based image processing 

which al-low to connect different camera technologies but require available and stable 

internet connections for real-time processing. 

The aim of the present work is to develop a new effective method for a continuous contactless 

water level measurement even under critical conditions like floods and hydraulic jumps, 

under which other localized gauging methods could be temporal out of operation or not 

sufficiently representative. In particular, the access to images of the site, which are the basis of 

the new water level detection method, will allow investigating the situation and judging on 

the quality of the water level measurement. The image processing and, as a result, the 

obtaining of water level data should be occur on site. 

METHODS 

SEBA Hydrometrie GmbH & Co. KG has developed a new gauge for enhanced water Level 

detection by image processing, which is based on the edge detection principle. The new 

instrument for optical water level measurement is called GaugeKeeper. The detection 

algorithm is running in real-time on the device. 

As it is shown in the Figure 1, GaugeKeeper system consists of the following components: 

day and night camera, infrared projector, a white board, a processor unit, a data logger and 

data transmission unit. All units are integrated in one system but it is also possible to apply 

only the GaugeKeeper-Algorithm for an existing system (e.g. an installed camera). 

Using this technology it is easy to survey, measure, and verify the water level data (Hies et 

al., 2012). Independently, alarm limits can be defined for the case the water level reaches a 

critical limit or hardware warnings (i.e. battery capacity). SMS alarms can be sent to up to 


Figure 1: This figure shows the GaugeKeeper components and an 

on-site picture of the installed system. 

8 different mobile phone numbers as well as 

to a facsimile. Prior to using the system for 

the first time, the Gauge-Keeper needs to be 

calibrated to the site‘s specific conditions. 

That is done via a graphical user interface 

(GUI). Thereafter the software calibrates the 

system and defines an individual measuring 

scale. The surveillance camera is ruggedized, 

equipped with special, non-visible infrared 

illumination for night-time measurement and 

uses an integrated powerful processor to 

automatically convert data to measurement 

values. The obtained data are available as 

hydrographs, images and time-lapse movies. 

The frequency is configurable and the images 

are saved to a local SD-card for preservation 

of evidence. The water level is measured and 

converted inside the processor unit and 

stored on the data logger. Those data and the 

images can be downloaded via remote access 

(e.g. GSM/GPRS, satellite, landline, radio 

transmission, DSL, Ethernet). The major steps 

for the robust detection of the water level are 

summarized below: 

– Automatic, adaptive selection of the region-of-interest (ROI); 

– An image processing technique, the so-called edge detection, is applied to obtain the edge 

image of the ROI image edge-detection assuming discontinuities of intensities of pixels of 

images are linked to physical changes e.g. material changes, depth changes, surface 

orientation changes etc. (Barrow et al., 1981), (Lindeberg, 2001), (Jaehne, 2002); 

– Hough-transformation calculating the longest straight line in the edge image of the water 

level (the red line). 

Figure 2 and 3 show the transformation of the captured images to undistorted frontal-view 

images with the region of interest and red line indicating the water level. 

After the water level is detected through the above methods, the water depth of the channel 

is interpolated based on the field measurement done at the ROI (Hies et al., 2012). 


Figure 2: This figure shows the water level detection for an image captured by the day-camera (left: original image, right: transformed 

image). 

Figure 3: This figure shows the water level detection for an image captured by the night-camera (left: original image, right: transformed 

image). 

RESULTS 

The developed optical gauge was tested at several sites and under different conditions (rain 

and heavy tropical rain events, snow, fog, day and night) in Germany and Singapore. The 

validation was carried out by comparing the water levels determined by the new developed 

gauge against the levels measured by conventional sensors like radar and pressure sensor at 

the same location. The agreement between the measurements carried out by the optical 

gauge and by the two reference sensors was very good (Fig. 4 and 5). The accuracy of the 

camera system was 1.1 % for the test period between April and December 2011. A second 

test, with a pressure sensor as a reference gauge, was carried out in the period between 

December 2014 and February 2015. The relative deviation was in this case around 0.95 % 

and the root-mean-square error (RMSE) was about 1 cm (the average water depth of the site 

was around 76 cm). 


Figure 4: This figure shows a water level measurement comparison between GaugeKeeper and a radar sensor at a site in Singapore. 

Figure 5: This figure shows a water level measurement comparison between GaugeKeeper and a pressure sensor at a site in Germany. 

Due to the optical image processing method, the developed gauge system requires no contact 

with the measured medium. Therefore, silting or flotsam have no influence on the operation 

of Gauge-Keeper and it can work even in extreme events such as floods and hydraulic jumps. 

The distance can be up to 70 m be-tween the camera and ROI (the white board). 

The special lighting of the presented gauge system enables its operation under many different 

weather and light conditions. The low energy consumption of the measuring allows a 

permanent autonomous battery or solar operation. The implemented low power components 

allow continuous 24 hours, 7 days operation with real-time monitoring access based on 

battery power supply for a minimum duration of 14 days (Hies et al., 2012). 

The activation of the measuring system can alternatively be controlled by trigger pulses. This 

can be very useful in cases where the water doesn’t flow permanently or if the water level 

exceeds a pre-defined value (e.g. flood events). Applying a water detection sensor, Gauge- 

Keeper automatically will be switched from a sleep mode (e.g. every 24 hours) to a dynamic 

measurement mode with higher data recording frequency (e.g. every 2 minutes). This feature 

provides a detailed visual representation of the observed events and saves energy significantly. 

The user of this gauge system can be immediately informed by SMS and/ or e-mail when 


critical system states (e.g. in case of low battery voltage or sensor drift) are reached or when 

definable thresholds are exceeded/ fallen be-low. 

CONCLUSIONS 

The presented optical gauge was successfully verified and validated under various weather 

and lighting conditions. The new developed gauge has proven to be a reliable method for 

on-site water level measurement. In addition to the measured water level value, the 

associated photographic evidence can be provided by the gauge system. The application of 

this technology within flood protection measures can give a real time overview about the 

observed sites to undertake timely the suitable actions especially in urban areas and where 

landslides and dike-breaches are potential. The GaugeKeeper is maintenance-free. The 

camera lenses and the glass of the illumination unit must be clean and free of dirt. The white 

board is coated with special coating materials, which minimizes the cleaning work. 

The developed optical gauge will be tested in the future in industrial and agricultural sectors 

to detect the filling level of liquids and solids. 

REFERENCES 

- Barrow, H. G. and Tenenbaum, J. M. (1981). Interpreting line drawings as three-dimensional 

surfac-es, Artificial Intelligence, vol 17, issues 1-3, pages 75-116. 

- Hasan, I., M. Meyer, J. Guo , P.-W. Gräber, (2012). Simulation of the hydrological regime 

in earth dams and dikes as a basis for stability analysis using the software PCSiWaPro ® . 

ISBN 978-83-60455-66-1. Opole/ Polen (2012). 

- Hies, T. Parasuraman, S.B., Wang, Y., Duester, R., Eikaas, H.S, Tan, K.M. (2012): 

Enhanced Wa-ter-level Detection by Image Processing. Proceedings 10th International 

Conference on Hydroin-formatics, HIC 2012, Hamburg, Germany 

- Jaehne, B. (2002). Digital Image Processing, Springer, ISBN 3-540-67754-2. 

- Lindeberg, T. (2001). Edge detection, in M. Hazewinkel (editor), Encyclopedia of Mathematics, 

Kluwer/Springer, ISBN 1402006098. (http://eom.springer.de/E/e120040.htm) 

Munich Reinsurance (2014). http://www.munichre.com/de/media-relations/publications/ 

press-releases/2014/2014-01-07-press-release/index.html 

- Nguyen, L. S., Schaeli, B., Sage, D., Kayal, S., Jeanbourquin, D., Barry, D. a, & Rossi, L. 

(2009). Vision-based system for the control and measurement of wastewater flow rate in 

sewer systems. Water science and technology, 60(9), 2281-9. 

- The United Nations office for disaster risk reduction (2002). 

http://www.unisdr.org/files/558_7639.pdf 

- The Federal Institute of Hydrology (Bundesanstalt für Gewässerkunde, BfG) 2013. 

Annual report 2012/213. ISSN 0170–5156. http://www.bafg.de/EN/03_The_%20BfG/ 

AnnualReport1213.pdf?__blob=publicationFile 

- http://www.gaugecam.com 

- http://www.hydropix.ch 

- www.tenevia.com 



Flood protection at Zurich main station: physical model 

experiments 

Hochwasserschutz Hauptbahnhof Zürich: Hydraulische 

Modellversuche 

Florian Hinkelammert, Msc. 1 ; Dr. Volker Weitbrecht 2 ; Prof. Dr. Robert M. Boes 2 

ABSTRACT 

Essential parts of downtown Zurich are located on the alluvial fan of the Sihl River. Due to 

high sediment loads and considerable amounts of driftwood during extreme flood events, 

the potential damage for a PMF is estimated at about 5 billion CHF. The main bottleneck 

concerning the discharge capacity is located at Zurich's main railway station, where the Sihl 

River crosses the station in an intermediate floor through five culverts. For the successful 

design of flood control measures, reliable data on the discharge capacity with the influence 

of driftwood and sediment transport are essential. To this end, VAW of ETH Zurich performed 

physical model investigations at a scale of 1:30. In parallel, a numerical 2D-model was 

operated. The experiments showed that the consideration of backwater effects triggered by 

the confluence of Sihl River and Limmat River 400 m downstream of Zurich main station 

is crucial. Therefore, the discharge of the Limmat River has to be considered for the specification 

of the discharge capacity of the five culverts. 


Grosse Bereiche der Stadt Zürich liegen auf dem Schwemmkegel des Voralpenflusses Sihl und 

sind stark hochwassergefährdet. Aufgrund der starken Verbauung wird das Schadenspotential 

bei EHQ auf bis zu 5 Milliarden Franken geschätzt. Die massgebende Schlüsselstelle für die 

Abflusskapazität der Sihl im Stadtgebiet ist die Position der Gleishalle des Hauptbahnhofs 

Zürich, welche von der Sihl in einem Zwischengeschoss in fünf Durchlässen durchquert wird. 

Diese Situation begünstigt Verklausungen durch Schwemmholz und anderes Treibgut und 

kann zu einer Reduktion der Abflusskapazität bei Materialablagerungen in den Durchlässen 

führen. Der Ausbau des Hochwasserschutzes im Stadtgebiet bedingt eine möglichst exakte 

Kenntnis der Abflusskapazität der Sihldurchlässe. Hybride Untersuchungen der Versuchsanstalt 

für Wasserbau, Hydrologie und Glaziologie (VAW) der ETH Zürich mittels eines 

physikalischen Modells im Massstab 1:30 sowie einem hydronumerischen 2D-Modell zeigten, 

dass der Einfluss des 400 m flussab gelegenen Zusammenflusses von Sihl und Limmat 

massgebenden Abfluss auf die Wasserspiegellagen in den Sihldurchlässen hat. Angaben zur 

1 VAW, ETHZ, Zürich, SWITZERLAND, hinkelammert@vaw.baug.ethz.ch 

2 VAW, ETHZ 



Abflusskapazität der Sihl können daher nur in Zusammenhang mit dem jeweiligen Limmat-Abfluss 

eindeutig interpretiert werden. 

KEYWORDS 

Flood protection; physical model; hybrid modelling; debris flows; sediment transport 

HINTERGRUND 

Grosse Teile des Zürcher Stadtgebiets wurden auf dem natürlichen Schwemmkegel der Sihl 

errichtet und sind bei ausserordentlichen Hochwasserereignissen stark überflutungsgefährdet. 

Aufgrund des stetigen Ausbaus dieser Gebiete im Zuge der Stadtentwicklung wird das 

Schadenspotential bei einer Abflussspitze von 550 m³/s auf bis zu 5 Milliarden Franken 

geschätzt (Marti et al. 2014). Eine Schlüsselstelle für die Abflusskapazität der Sihl im Stadtgebiet 

nimmt der Hauptbahnhof Zürich („HB Zürich“) ein, dessen Gleishalle von der Sihl in 

einem Zwischengeschoss mit fünf Durchlässen durchquert wird. 

Das Hochwasserereignis von August 2005 entsprach im Stadtgebiet Zürich mit einem Sihl- 

Spitzenabfluss von 280 m³/s knapp einem HQ 30 

. Aufgrund einer günstigen Wetterlage kam 

es zu keinen Überflutungen im Stadtgebiet, das Ereignis verdeutlichte jedoch die im Hochwasserschutz 

bestehenden Defizite. Aufgrund des dringenden Handlungsbedarfs starteten die 

Kantone Zürich und Schwyz unter Leitung des Amts für Abfall, Wasser, Energie und Luft des 

Kantons Zürich (AWEL) ein umfangreiches Projekt zum Hochwasserschutz im Einzugsgebiet 

Sihl – Zürichsee - Limmat (Kanton Zürich 2012). Neben der Umsetzung diverser kurz- und 

mittelfristiger Massnahmen in den Jahren 2005 bis 2012 werden bis 2017 zwei Konzepte für 

den langfristigen Hochwasserschutz geprüft. Variante 1 sieht die Errichtung eines Entlastungsstollens 

vor, welcher ab einem Abfluss von 150 m³/s die Hochwasserspitzen der Sihl in den 

Zürichsee ableiten würde. Die 2. Variante „Kombilösung Energie“ beinhaltet den Ausbau des 

bestehenden Pumpspeicherkraftwerks Etzelwerk am Sihlsee und vereint Aspekte des 

Hochwasserschutzes und der Energieproduktion (Marti et al. 2014). 

ZIELSETZUNG 

Beide Varianten für den langfristigen Hochwasserschutz bedingen eine möglichst exakte 

Kenntnis der Abflusskapazität der Sihldurchlässe im Bereich des HB Zürich. Numerische 

Simulationen stossen bei solchen Fragestellungen an ihre Grenzen, vor allem die Prognose 

der Sohllagen sowie die Untersuchung von Verklausungsszenarien sind nicht in ausreichender 

Genauigkeit möglich. Die Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie 

(VAW) der ETH Zürich wurde durch das AWEL sowie die Schweizerischen Bundesbahnen 

(SBB) im Juli 2012 mit der Durchführung von hydraulischen Modellversuchen beauftragt. 

Die Resultate dienen unter anderem der Festlegung des Ausbaugrads der Sihl in der Stadt 

Zürich sowie der Massnahmenplanung im Überlastfall (Marti et al. 2014). 

Folgende Aspekte wurden untersucht: 

– Abflusskapazität der Sihldurchlässe mit sowie ohne gesetzliche Freiborde („Q max 

“) 


– Systemverhalten bei Hochwasserszenarien entsprechend der Gefahrenkartierung 

(HQ 100 

, HQ 300 

, EHQ) 

– Geschiebetransport mit Variantenuntersuchungen zu Sohlaufbau in Sihldurchlässen 

und Gerinne 

– Verklausungsgefahr durch Schwemmholz 

– Optimierungsmöglichkeiten für zukünftige Gestaltung der Durchlässe 

Für die hydraulischen Modellversuche wurden die Szenarien der Gefahrenkartierung („GK“, 

Basler & Hofmann 2008) als Grundlage verwendet. Die Spitzenabflüsse, Hydrographen sowie 

Abflusskombinationen wurden dementsprechend gewählt. Für die Bestimmung der maximalen 

Wasserspiegel wurde ein Szenario gewählt, bei dem einer der 5 Sihldurchlässe vollständig 

mit Schwemmholz verklaust ist. Während der Durchführung der Modellversuche wurden 

zusätzliche Untersuchungen zur Hydrologie des Einzugsgebiets abgeschlossen (Scherrer 2013), 

welche zu einer Erhöhung der Spitzenabflüsse der verschiedenen Jährlichkeiten führte. 

Diese höheren statistischen Abflusswerte wurden für die Hauptszenarien im hydraulischen 

Modell nicht berücksichtigt, jedoch im Rahmen einer Sensitivitätsanalyse untersucht. Auf das 

Hauptziel der Modellversuche, die Bestimmung der maximalen Abflusskapazität, hatte die 

Erhöhung der statistischen Hochwasserwerte keinen Einfluss. 

UNTERSUCHUNGSGEBIET SIHLDURCHLÄSSE 

Das Untersuchungsgebiet erstreckt sich vom Pegel Sihlhölzli an der Sihl bis zum Pegel Unterhard 

an der Limmat (Abbildung 1). 

Die Gleishalle des HB Zürich liegt 400 m flussauf des Zusammenflusses von Sihl und Limmat 

und wird von der Sihl in einem Zwischengeschoss durch 5 Durchlässe auf einer Länge von 

ca. 190 m gequert (Masse pro Durchlass: lichte Weite ca. 12 m, lichte Höhe ca. 3 m, Länge 

ca. 190 m). Für die Gewährleistung des Geschiebetransports in der Sihl ist diese vom flussauf 

des HB Zürich einmündenden Schanzengraben mittels einer ca. 1 m hohen, ca. 200 m langen 

Mauer getrennt. Bei mittleren Abflüssen fliesst die Sihl durch die linken drei Durchlässe 

unter der Gleishalle, während der Schanzengraben die rechten zwei Durchlässe für sich 

beansprucht. Erst ab einem Sihl-Abfluss von ca. 140 m³/s wird die Trennmauer überströmt, 

wodurch der Sihl bei Hochwasserereignissen der gesamte Abflussquerschnitt zur Verfügung 

steht. 

Knapp oberstrom des Zusammenflusses von Sihl und Limmat befindet sich im Gerinne der 

Limmat das Lettenwehr, welches den Wasserstand des Zürichsees regelt und als Ausleitungsbauwerk 

für das Kraftwerk Letten dient. Aufgrund dieser Anordnung haben die Wehrsteuerung 

sowie die Abflussmenge in der Restwasserstrecke (RWS) der Limmat einen direkten 

Einfluss auf das Abflussgeschehen der Sihl sowie die Wasserspiegellagen in den Sihldurchlässen. 

Das derzeitige Wehrreglement beschränkt den Abfluss in der Limmat unterstrom des 

Zusammenflusses auf max. 600 m³/s, wobei bei EHQ von einem Gesamtabfluss von 800 m³/s 

ausgegangen wird. 


Abbildung 1: Untersuchungsgebiet Sihldurchlässe (VAW, mod. nach EWZ) 

Für die Szenarien der GK wurde angenommen, dass bei Extremereignissen das KW Letten 

abgeschaltet wird und der gesamte Abfluss über die RWS fliesst. Tabelle 1 zeigt für die 

Szenarien der GK die entsprechenden Abflusskombinationen sowie die Geschiebe- und 

Schwemmholzfrachten während der Ereignisse gemäss den Festlegungen für die Modellversuche 

der VAW sowie den zur Verfügung stehenden relevanten Studien (Flussbau AG 2009, 

2010). 


Tabelle 1: Abflusskombinationen gemäss GK Stadt Zürich sowie Geschiebe- und Schwemmholzfrachten 

MODELLZUSTÄNDE 

Die Bauarbeiten für die Erstellung der SBB-Durchmesserlinie Altstetten – Zürich HB – Oerlikon 

(„DML“) führten im Untersuchungsgebiet zu massgebenden baulichen Veränderungen. 

Für die Untersuchung der Abflusskapazität der Sihldurchlässe wurde daher zwischen dem 

Zustand 2006 – vor dem Bau der DML – sowie dem Zustand 2014 – nach Abschluss der 

Bauarbeiten – unterschieden. 

Zustand 2006 

Bis 2006 verfügte die Sihl unterhalb des HB Zürich über eine grösstenteils natürliche Flusssohle, 

in deren letztem Drittel mit einer Überdeckung von ca. 1.3 m der Tiefbahnhof 

Museumsstrasse lag (Abbildung 2, Ziffer 1). Die Decke der Durchlässe setzt sich aus den 

Gleisbrücken der 16 Hauptgleise sowie den Bahnsteigen zusammen und ist von der Sihl frei 

zugänglich (Abbildung 2, Ziffer 2). Im Einlaufbereich unterhalb der Postbrücke nimmt die 

lichte Höhe von ca. 4.5 m auf etwa 3 m ab. Erst unterhalb der ersten Gleisbrücke ist somit 

der kritische Querschnitt im Einlaufbereich vorhanden (Abbildung 2, Ziffer 3). 

Die Durchlässe werden von Zwischenmauern unterteilt, welche als Widerlager für die Gleisbrücken 

dienen. Das untere Ende der Durchlässe wird von der Zollbrücke überspannt. Kurz 

vor diesem Bauwerk liegt bei sämtlichen Durchlässen der tiefste Punkt der Deckenkonstruktion 

(Abbildung 2, Ziffer 4). Ober- sowie unterstrom des HB Zürich befanden sich bis 2006 im 

Einflussbereich der Trennmauer ausgedehnte, teils stark bewachsene Schotterinseln. 

In diesem Zustand kam es in den Jahren 2005, 2006 und 2007 zu grösseren Hochwasserereignissen, 

welche anhand von Hochwasserspuren („HW-Spuren“) dokumentiert wurden. 

Zustand 2014 

Zur Erhöhung der Hochwassersicherheit während der Bauarbeiten für die DML wurde die 

Sohle in den Sihldurchlässen gegenüber der mittleren Sohle von 2006 um ca. 0.6 m abgesenkt 

und auf dieser Höhenlage flächig mit armiertem Beton versiegelt (Abbildung 2, Ziffer 

5). Der Tiefbahnhof Löwenstrasse wurde neu erstellt und liegt ca. 0.8 m bis 1.3 m unterhalb 

der Sohlversiegelung (Abbildung 2, Ziffer 6). Die Deckenplatten der Tiefbahnhöfe befinden 

sich in Trögen, welche mit Sohlmaterial der Sihl aufgefüllt wurden (Abbildung 2, Ziffern 7). 

Die Deckenstruktur der Sihldurchlässe wurde im Rahmen der Bauarbeiten für die DML nicht 

verändert. Neben der Absenkung der Sohle in den Durchlässen wurden flussauf sowie flussab 


Abbildung 2: Sihldurchlässe: Querschnitt und Längsschnitte in den Modellzuständen 2006 und 2014 (VAW, mod. nach SBB) 

des HB Zürich in der Sihl Baggerungen von bis zu 1 m Tiefe vorgenommen. Die Trennmauer 

oberhalb des HB wurde teilweise abgebrochen. 

HYBRIDE MODELLIERUNG 

Numerisches Modell 

Aufgrund der komplexen Randbedingungen für den Betrieb des hydraulischen Modells 

wurden von der VAW mit der Simulationsumgebung BASEMENT für beide Modellzustände 

hydronumerische 2D-Modelle erstellt. Der Perimeter der hydronumerischen Modelle 

erstreckte sich von der Pegelstation Sihlhölzli bis zur Kornhausbrücke (Abbildung 1), der 

Betrieb erfolgte rein hydraulisch mit fester Sohle. Die Kalibrierung des numerischen Modells 

erfolgte mit Hilfe von 15 HW-Spuren des Hochwasserereignisses von August 2005 an Sihl und 

Limmat und erlaubte speziell im Bereich des Zusammenflusses eine korrekte Abbildung der 

Wasserspiegellagen. Die Ergebnisse des numerischen Modells wurden in den Elementen des 

Berechnungsnetzes entsprechend der Lage der HW-Spuren bestimmt und den Naturwerten 

gegenübergestellt (Tabelle 2). Die Kalibrierung erfolgte über Optimierungen der Randbedingungen, 

der Netzauflösung sowie durch Anpassung der Rauhigkeitsbeiwerte. Die Validierung 


des Modells wurde anhand der Hochwasserereignisse in 2006 und 2007 durchgeführt, wobei 

nur für den Nahbereich des HB Zürich Hochwasserspuren vorlagen (Tabelle 2). 

Laborversuch 

Zur Untersuchung der Abflussverhältnisse in den Sihldurchlässen wurde in der Versuchshalle 

der VAW ein hydraulisches Modell im Massstab 1:30 nach Froude’scher Modellähnlichkeit 

errichtet. Das Modell bildete 900 m Fliessstrecke der Sihl ab, die maximale Breite des Modellperimeters 

betrug 200 m. Der Modelleinlauf befand sich im Prototyp 400 m oberhalb des HB 

Zürich, das Modellende lag 50 m flussauf des Zusammenflusses von Sihl und Limmat. Die 

korrekte Wiedergabe des Wasserstandes beim Modellauslauf wurde durch ein gestautes Auslaufbecken 

mit automatischer Regulierung des Wasserstands sichergestellt. 

Das Hauptgerinne der Sihl war mit beweglicher Sohle ausgestattet, während der Schanzengraben 

sowie die Ufer im Modell befestigt wurden. Für einen optimalen optischen Zugang 

wurden die Sihldurchlässe sowie die Seitenmauern des Modells in diesem Bereich in Acrylglas 

ausgeführt. Die Wasserspiegel wurden durch Ultraschallsensoren sowie Messzylinder, 

welche über kommunizierende Röhren mit der Flusssohle verbunden waren, aufgezeichnet. 

Die Sohllagen im Modell wurden mittels Laserdistanzmessung aufgenommen. Die Geschiebezugabe 

erfolgte über eine automatisierte Beschickungsanlage, Schwemmholz wurde manuell 

zugegeben. 

Tabelle 2: Abweichungen für Kalibrierung und Validierung von numerischer Modellierung und Laborversuch 

Die Kalibrierung des hydraulischen Modells erfolgte anhand der Hochwasserereignisse in den 

Jahren 2005 / 2006 / 2007 und in Gegenüberstellung mit den Ergebnissen der parallel 

durchgeführten numerischen Simulationen. Tabelle 2 gibt einen Überblick über die mittleren 

sowie maximalen Abweichungen der Kalibrierungs- und Validierungsereignisse. 

Die Ergebnisse des numerischen Modells waren bei sämtlichen Szenarien massgebend für die 

Festlegung der unteren Randbedingung im Laborversuch. Die Wasserspiegellagen wurden in 

einem Zielelement aus der hydronumerischen Simulation des jeweiligen Szenarios ausgelesen, 

massstäblich umgerechnet und mittels eines Korrekturwerts an die Strömungsbedingun- 


gen im Auslaufbecken des hydraulischen Modells angepasst. Diese korrigierten Zielwerte 

wurden für die automatisierte Steuerung des Wasserstands im Auslaufbecken verwendet. 

ERGEBNISSE 

Maximale Abflusskapazität 

Die Versuche im hydraulischen Modell verdeutlichten den massgebenden Einfluss des 

Rückstaus der Limmat auf die Abflussverhältnisse und die Wasserspiegellagen im Bereich der 

Sihldurchlässe. Angaben zur Abflusskapazität der Sihl können daher nur in Kombination mit 

dem jeweiligen Abfluss in der Limmat eindeutig interpretiert werden. Als maximale Abflusskapazität 

Qmax wurde durch das AWEL und die VAW der Abflusszustand festgelegt, bei dem 

ein „gleichmässiges Anschlagen des Wasserspiegels an Gleisbrücken bzw. Brückenbauwerken“ 

zu beobachten ist. Die Versuche im hydraulischen Modell zeigten, dass Q max 

gemäss dieser 

Definition bei einem Spitzenabfluss von 490 m³/s erreicht ist und in Kombination mit einem 

Limmatabfluss von 190 m³/s gilt („Q max 

490/190“). Dieser Wert wurde im Laborversuch für 

die Zustände 2006 und 2014 sowie für Versuche mit und ohne Schwemmholzzugabe 

bestätigt. 

Die Laborversuche zeigten, dass Q max 

auch durch eine Absenkung des Wasserspiegels beim 

Zusammenfluss Sihl / Limmat nicht signifikant vergrössert werden kann, obwohl die 

Strömungsprozesse in der Sihl stark durch den Limmatwasserstand beeinflusst werden. 

Grund dafür ist das gewählte Kriterium, bei dem der maximale Abfluss durch ein gleichmässiges 

Anschlagen des Wasserspiegels an der Deckenstruktur der Durchlässe definiert ist. Ein 

geringerer Limmatabfluss führte zwar im Mittel zu einer Absenkung der Wasserspiegel im 

Bereich des HB Zürich, aufgrund der grösseren Fliessgeschwindigkeiten bzw. Froudezahlen 

kam es jedoch zu verstärkter Wellenbildung. 

Prozesse im Zustand 2006 

Im Zustand 2006 kam es bei Qmax 490/190 im Einlaufbereich der Durchlässe zum Anschlagen 

des Wasserspiegels, das mittlere Freibord im kritischen Querschnitt betrug 0.2 bis 0.3 m. 

Aufgrund des Stammdurchmessers des Schwemmholzes von 0.3 bis 0.4 m passierte dieses die 

Sihldurchlässe zum grössten Teil, nur vereinzelt blieben Stämme an Widerlagern hängen. Bei 

den beobachteten Prozessen wurden jedoch lokal die Grenzen der Modellähnlichkeit erreicht. 

Das Schwemmholz wurde im Laborversuch als entastet modelliert und teilweise während der 

Versuche an den idealisierten Gleisbrücken aus Acrylglas entlangeschoben. Überlastversuche 

mit vollständiger Abdichtung des mittleren Durchlasses zeigten jedoch, dass es auch bei 

Vorhandensein von massiven Verklausungskörpern im Einlaufbereich zu keiner plötzlichen 

Reduktion der Abflusskapazität kommt. Unterhalb der Verklausungskörper wurden in der 

natürlichen Flusssohle tiefe Kolke gebildet, welche die Abflusskapazität aufrechterhielten. 

Im Auslaufbereich der Sihldurchlässe wurde hinsichtlich der Wasserspiegellagen der Rückstau 

der Limmat als massgebender Parameter identifiziert. Die Sohllagen flussab des HB Zürich 


sowie die Schotterinseln hatten nur einen geringen Einfluss auf die Abflussverhältnisse in 

diesem Bereich. 

Prozesse im Zustand 2014 

Aufgrund des dominierenden Rückstaus der Limmat führten die Baggerungen sowie die 

Sohlversiegelung in den Durchlässen im Zustand 2014 zu keiner signifikanten Erhöhung der 

Abflusskapazität. Durch die tieferen Sohllagen lagen die Wasserspiegel bei Qmax 490/190 um 

0.3 bis 0.4 m unterhalb des kritischen Querschnitts im Einlaufbereich, wodurch das Verklausungsrisiko 

reduziert wurde. Die Schwachstelle der Abflusskapazität lag im Zustand 2014 

somit im unteren Drittel der Sihldurchlässe, an dieser Stelle wurde bei Qmax 490/190 das 

erste Anschlagen des Wasserspiegels festgestellt. Auch in diesem Bereich blieb aufgrund der 

beschriebenen Effekte nur vereinzelt Schwemmholz hängen. Die Überlastversuche mit 

Abdichtung des mittleren Durchlasses zeigten, dass die Sohlversiegelung die Bildung von 

Kolken unterhalb von Verklausungen verhindert und in diesem Fall einen negativen Einfluss 

auf die Abflusskapazität hat. 

Sohlentwicklung 

Der Gerinnequerschnitt der Sihl verbreitert sich oberstrom des HB Zürich von 55 m auf 70 m. 

Gleichzeitig ist die Sohlneigung in diesem Flussabschnitt geringer, so dass der Bereich 

oberstrom des HB Zürich eine Auflandungsstrecke darstellt. Durch den Rückstau der Limmat 

im Hochwasserfall wird dieser Effekt verstärkt und es kommt auch flussab des HB Zürich zu 

Auflandungen. Das Vorhandensein grosser Schotterinseln im Zustand 2006 bestätigt diese 

Beobachtungen. Bei sämtlichen untersuchten Hochwasserszenarien kam es oberstrom der 

Sihldurchlässe zu Auflandungen. In den Durchlässen wurde bei diesen Szenarien die Bildung 

von Geschiebeansammlungen beobachtet, deren Fronten sich bei fortschreitender Versuchsdauer 

mit gleichbleibender Mächtigkeit durch die Durchlässe bewegten. Dieser Prozess wurde 

in beiden Modellzuständen beobachtet, ist in Zustand 2014 aufgrund der tieferen Sohllagen 

jedoch stärker ausgeprägt. Signifikanter Geschiebetransport aus der Sihl in die Limmat 

konnte bei keinem der untersuchten Hochwasserszenarien beobachtet werden. 

Bestvariante 

Das AWEL erteilte für die Projektierung der Durchmesserlinie die Auflage, dass das Sihlgerinne 

nach Abschluss der Bauarbeiten wieder in einen natürlichen Zustand versetzt wird. Vor 

diesem Hintergrund wurde von der VAW eine Bestvariante für das künftige Erscheinungsbild 

der Sihl im Bereich des HB Zürich entwickelt. Untersuchungen des Einflusses der Trennmauer 

zwischen Sihl und Schanzengraben zeigten, dass deren Entfernung keine negativen 

Auswirkungen auf die Abflusskapazität hat, jedoch eine dynamischere Entwicklung der Sohle 

zulässt. 

Eine natürliche Kiessohle in den Durchlässen ist für die Fischgängigkeit der Sihl sowie die 

Interaktion mit dem Grundwasserkörper von Vorteil. Vor diesem Hintergrund und unter 


Berücksichtigung der ermittelten Abflusskapazitäten empfiehlt die VAW, die Sohlversiegelung 

in den Sihldurchlässen zu entfernen und die natürliche Kiessohle wieder herzustellen. Durch 

die natürliche Auflandungstendenz im Untersuchungsgebiet werden auf Dauer die Sohlhöhen 

des Zustands 2006 wieder erreicht werden. 

Die Laborversuche zur maximalen Abflusskapazität im Zustand 2006 zeigten im Einlaufbereich 

der Sihldurchlässe Kolktiefen zwischen -1.5 und -2 m. Eintiefungen dieser Grössenordnung 

gefährden die Fundationen der Zwischenmauern der Sihldurchlässe und dementsprechend 

die Stabilität der darauf gelagerten Gleisbrücken. Im hydraulischen Modell wurde 

daher für den Einlaufbereich die Wirksamkeit eines Kolkschutzes mittels Blockteppich 

(Natursteinblöcke à 1.5 t) überprüft und bei Qmax 490/190 bestätigt. 

FAZIT 

Gemäss der beschriebenen Definition liegt die maximale Abflusskapazität der Sihl beim HB 

Zürich in beiden Modellzuständen bei einem Sihlabfluss von 490 m³/s in Kombination mit 

einem Limmatabfluss von 190 m³/s. Höhere bzw. tiefere Wasserspiegel in der Restwasserstrecke 

der Limmat beeinflussen die maximale Abflusskapazität der Sihldurchlässe entsprechend 

der Definition als Anschlagen des Wasserspiegels an Bauwerksunterkanten nicht 

signifikant. Bei Spitzenabflüssen bis HQ 300 

ist der Einfluss der Limmat jedoch massgebend 

und verlangt eine ganzheitliche Interpretation und Untersuchung der Abflussverhältnisse im 

Bereich des HB Zürich. Die Untersuchungen im hydraulischen Modell haben gezeigt, dass 

das Verklausungsrisiko bei diesen Maximalabflüssen gering ist, jedoch von Modelleffekten 

ausgegangen werden muss. Für die zukünftige Gestaltung der Sihl im Untersuchungsgebiet 

empfiehlt die VAW die Wiederherstellung der natürlichen Kiessohle mit den Sohllagen von 

2006. Die Sohlversiegelung in den Durchlässen bringt keine massgebende Erhöhung der 

Abflusskapazität und kann vor diesem Hintergrund rückgebaut werden. Aufgrund der 

Kolkbildung bei Extremereignissen im Einlaufbereich der Sihldurchlässe sind lokale Schutzmassnahmen 

der Sohle anzuraten. Die Entfernung der Trennmauer zwischen Sihl und 

Schanzengraben führen zu einer dynamischeren und natürlicheren Sohlentwicklung. 

LITERATUR 

- Basler und Hofmann AG (2008). Gefahrenkartierung Hochwasser Stadt Zürich, Technischer 

Bericht. Im Auftrag des AWEL. Stand: 24. November 2008 (unveröffentlicht) 

- Flussbau AG (2009). Schwemmholzstudie Sihl. Stand: 15. Dezember 2009 (unveröffentlicht) 

- Flussbau AG (2010). Geschiebehaushaltsstudie Sihl - Limmat. Stand: 20. August 2010 

(unveröffentlicht) 

- Kanton Zürich (2012). Wasserbau, Hochwasserschutz Sihl-Zürichsee-Limmat (Ausgabenbewilligung). 

Regierungsratsbeschluss 925. Stand: 12. September 2012 

- Marti C. et al. (2014). Hochwasserschutz Sihl-Zürichsee-Limmat. Tagungsband „Wasserbau 

und Flussbau im Alpenraum“, VAW-Mitteilung 227/228 (R. Boes, ed.), VAW, ETH Zürich 

- Scherrer AG (2013). Hochwasser-Hydrologie der Sihl - Hochwasserabschätzung unterhalb 

des Sihlsees bis Zürich. Im Auftrag des AWEL. Stand: Juli 2013 (unveröffentlicht) 



Characteristics of debris flow vibration signals in 

Shenmu, Taiwan 

Yi-Min Huang, Ph.D. 1 ; Chung-Ray Chu, Ph.D. 2 ; Yao-Min Fang, Ph.D 3 ; Ming-Chang Tsai 3 ; Bing-Jean Lee, Ph.D. 4 ; Tien-Yin Chou, 

Ph.D. 4 ; Chen-Yang Lee 5 ; Chen-Yu Chen, Ph.D. 6 ; Hsiao-Yuan Yin, Ph.D. 6 

ABSTRACT 

Influenced by the climate change and the extreme weather, debris flow has become a common 

disaster in Taiwan in recent years. To protect people from the impacts of debris flows, 

monitoring and warning system were established in Taiwan since 2002. Most of the warning 

systems or models are based on the analysis of rainfall, the major cause of debris flow (Jan et 

al., 2003; Jan and Lee, 2004; Lee 2006). The measurement of rainfall, however, is an indirect 

option of debris flow monitoring (Huang et al., 2013), resulting in "false alarms" most of the 

time. One direct measure is to use geophone and broadband seismograph (Chu et al., 2014; 

Huang et al., 2012) to record the vibration signals of debris flows. This study analyzed the 

vibration signals of a site in Shenmu, Taiwan, where debris flows frequently occurred (Lee et 

al., 2014), and discussed the performance of using vibration signals for debris flow warning. 

The results of using vibration signals were practically promising for debris flow monitoring. 

KEYWORDS 

debris flow; geophone; seismograph; ground vibration; monitoring 

INTRODUCTION 

Influenced by the climate change and the extreme weather, debris flow has become a 

common disaster in Taiwan in recent years. To protect people from the impacts of debris 

flows, monitoring and warning system were established by the Soil and Water Conservation 

Bureau (SWCB) in Taiwan since 2002. Currently there are 19 debris flow monitoring stations 

in Taiwan (Fig. 1). Most of the stations are located at the central part of Taiwan. 

Most of the warning systems or models are based on the analysis of rainfall, the major cause 

of debris flow (Jan et al., 2003; Jan and Lee, 2004; Lee 2006). The warnings based on the 

rainfall thresholds for debris flows were useful from the past experience. The measurement of 

rainfall, however, is an indirect option of debris flow monitoring (Huang et al., 2013), which 

is useful for disaster response but usually results in "false alarms". In contrast to the rainfall 

1 GIS Research Center, Taichung, TAIWAN, niner@gis.tw;ninerh99@gmail.com 

2 Fortemedia Inc. 

3 Department of Urban Planning and Spatial Information, Feng Chia University, Taiwan, R.O.C. 

4 Department of Civil Engineering, Feng Chia University, Taiwan, R.O.C. 

5 Soil and Water Conservation Bureau, Council of Agriculture, Taiwan R.O.C. 

6 Debris Flow Disaster Prevention Center, Soil and Water Conservation Bureau, Council of Agriculture, Taiwan, R.O.C. 



Figure 1: Debris flow monitoring stations in Taiwan 


measurement, using geophone (short period seismograph) and broadband seismograph 

(Chu et al., 2014; Huang et al., 2012) are direct options of debris flow monitoring, and they 

were used to record the vibration signals generated by debris flows. The vibration signals of 

cases in Shenmu, Taiwan, were analyzed to understand the characteristics of debris flow, and 

the possibility of applying geophones and broadband seismographs to debris flow warnings 

was discussed. 

STUDY AREA AND VIBRATION DETECTION OF DEBRIS FLOW 

The Shenmu area is located in the central Taiwan, where debris flows frequently occurred 

(Lee et al., 2014). A debris flow monitoring station was built in 2002. The local village is 

adjacent to the confluence of three streams: Aiyuzi Stream (DF226), Huosa Stream (DF227) 

and Chushuei Stream (DF199). These streams are classified as high-potential debris flow 

torrents. Table 1 summarized the environment of Shenmu, Table 2 and Fig. 2 indicate the 

area and locations of landslides along these three streams after 2009, and Table 3 lists the 

debris flow events in this area. It should be noted that the debris flows usually occurred at 

the Aiyuzi Stream in the past 5 years, due to its shorter length and large landslide area in 

its upstream (Huang, et al., 2013). 

Table 1: Environment of Shenmu Station (Huang, et al., 2013). 

Location Shenmu Village, Nantou County Debris Flow No. DF199, DF227, DF226 

Catchment Zhuoshui River Streams Chusuei, Huosa, Aiyuzi 

Debris Flow Warning 

Threshold 

250 mm Hazard Type Channelized debris flow 

Monitored Length 5.518 km Catchment Area 7,216.45 ha (Shenmu) 

Geology neogene sedimentary rock Slope at Source 30~50° 

Landslide area Large, 1%≦ landslide ratio ≦5% Sediment 

Average debris material size: 

3”-12” 

Vegetation Natural woods, medium sparse Damaged by debris, overflow 

Engineering Practice None Priority of Mitigation High 

Station Elevation 1,187 m Coordinate (TWD97) X: 235367 Y: 2602749 

Protected Targets 

Residents Facility Transportation 

> 5 households school roads, bridges 

Table 2: The landslide area in Shenmu after 2009 (Huang, et al., 2013). 

Debris Flow No. Stream Length (km) Catchment Area (ha) Landslide Area (ha) 

DF199 Chusuei Stream 7.16 861.56 33.29 

DF227 Huosa Stream 17.66 2,620 149.32 

DF226 Aiyuzi Stream 3.30 400.64 99.85 


Figure 2: The landslide areas of Shenmu site (image taken in 2009 after Typhoon Morakot). 

The sensors and instrument of rain gauges, geophones, broadband seismographs, wire sensors, 

and CCD camera were installed in the Shenmu Monitoring Station. The rain gauge is classified 

as the indirect option in the monitoring system to measure rainfall data. Most of researches 

about debris flow had been contributed in developing the relationship between the debris 

flow occurrence and rainfall. Rainfall is easy to measure and if a reasonable prediction model 

is using rainfall as a variable, it can provide a time window that can greatly improve the 

disaster warning and evacuation. The comparison of rainfall-based debris flow monitoring 

models can be made by model’s rainfall indices: accumulative rainfall (R), rainfall intensity 

(I), and the duration (T) (Lee, 2006). Each rainfall index has its characteristics in modeling 

debris flow. Therefore, Jan and Lee (2004) promoted a model considering the influence of all 

rainfall indices (R, T, and I), and used debris flow cases to determine the rainfall thresholds of 

a given location. This method was adopted by SWCB and works for debris flow warning in 

Taiwan. 

The direct option in debris flow monitoring system refers to the instruments and sensors that 

response when a debris flow actually occurs. Wire sensors, geophones and broadband 

seismographs are classified as this type. Wire sensors are simply steel wires crossing the river 

with signal transducers at ends. When a debris flow occurs, the flowing mass of debris will 

break the wire and results in a signal being sent out. Geophones and broadband seismographs 

are sensors installed along the river banks to detect ground surface vibration which comes 


Table 3: Debris flow hazard history of Shenmu (after Huang, et al., 2013). 

Location 

Geophone Hazard 

Date 

Event 

Occurrence 

(stream) 

Warning* Type 

2004/5/20 - Aiyuzi 14:53 NA debris flow 




2004/7/2 Typhoon Mindulle Aiyuzi 16:41 NA debris flow 

2005/7/19 Typhoon Haitang Chusuei, Aiyuzi - NA flood 

2005/8/4 Typhoon Matsa Chusuei, Aiyuzi - NA flood 

2005/9/1 Typhoon Talim Chusuei, Aiyuzi - NA flood 

2006/6/9 0609 Rainfall Chusuei, Aiyuzi about 08:00 NA debris flow 

2007/8/13 0809 Rainfall Chusuei - NA flood 

2007/8/18 Typhoon Sepat Chusuei - NA flood 

2007/10/6 Typhoon Krosa Chusuei - NA flood 

2008/7/17 Typhoon Kalmaegi Chusuei - NA flood 

2008/7/18 Typhoon Kalmaegi Aiyuzi - NA flood 

2009/8/8 Typhoon Morakot 

Chusuei,Aiyuzi, 08:00 (landslide) 

landslide, 

NA 

Huosa 16:57 (debris flow) 

debris flow 

2010/9/19 Typhoon Fanapi Huosa - NA flood 

2011/7/13 - Aiyuzi 14:33 No debris flow 

2011/7/19 0719 Rainfall Aiyuzi 3:19 No debris flow 

2011/11/10 - Aiyuzi 13:17 

Yes 

13:18 (Nov 10) 

debris flow 

2012/5/4 - Aiyuzi 

15:56 

16:09 

No debris flow 

2012/5/20 - Aiyuzi 8:15 No flood 

2012/6/10 0610 Rainfall Aiyuzi 

10:34 No (unstable 

15:14 communication) 

debris flow 

2012/6/11 0610 Rainfall Chusuei 17:08 No flood 

2013 0517 Rainfall Aiyuzi 7:02 (May 19) No flood 

2013 Typhoon Saulik Aiyuzi 6:54 (July 13) 

Yes 

6:47 (July 13) 

debris flow 

2013 Typhoon Trami Aiyuzi 22:41 (Aug. 21) 

NA (under 

repair) 

flood 

2014 0520 Rainfall Aiyuzi 12:53 (May 20) 

NA (unstable 

communication) 

debris flow 

*: the warning was issued when the max. accumulated wavelet energy was 5 times the background wavelet energy. 

from the movement of debris in the channel. The warning level for debris flow by geophone 

signals was predefined based on its accumulated wavelet energy (Lee et. al., 2012), and is still 

under testing. When the warning was issued by the geophone, it was checked with the status 

of nearby wire sensors (broken or not) and the CCD image (in-situ conditions). As shown in 

Table 3, the warnings were not issued every time when a debris flow occurred in the past 5 

years. This may because of the scale of debris flows and the geophone warning criteria. More 

study is needed on geophone warning criteria, and this paper will focus on the characteristics 

of debris flow vibration signals. 

The layout of monitoring sensors is shown in Fig. 3. One geophone and two seismographs are 

installed at the Aiyuzi Stream. The geophone of GS-20 DX made by Geospace Tech. was used, 


which has clean band pass of 250 Hz and intrinsic sensitivity of 0.7 V/in/sec. The broadband 

of Yardbird DF-2 was first installed at the site in 2013, made by Institute of Earth Science, 

Academia Sinica Taiwan, with frequency measurement of 0.13~200 Hz, intrinsic sensitivity of 

150 V/m/s, and 24 bit digital resolution. The signal data of z-direction (vertical) recorded by 

the geophone at the upstream and the broadband seismograph at the downstream of Aiyuzi 

Stream were used for analysis in this study. 

Figure 3: The monitoring layout of Shenmu Station. 

METHODS AND STUDY CASES 

To analyze the vibration signals of debris flows, the signal was processed to convert the data 

into a time sequence of velocity. Fast Fourier Transform (FFT) was first applied to obtain the 

distribution of vibration frequencies. The filter method of Haar wavelet transform (Fang et al., 

2008) was used in the signal analysis for geophone data, as well as to estimate the wavelet 

energy. The spectrograms of signal were plotted to illustrate the frequency characteristics 

from the broadband seismograph, and used to compare the signals of geophone and broadband 

seismograph. 

Four cases in the past 5 years were used for analysis and comparison (Table 4). All of them 

were debris flow events. The scale of debris flow, small, medium, and large, was determined 


Table 4: Wavelet energy change and time widow of warning of 4 events. 

Yea 

r 

2011 

2013 

2013 

Event 

1110 

Rainfall 

0530 

Heavy 

Rainfall 

Typhoon 

Saulik 

based on the maximum hourly rainfall, average flow speed, and the maximum accumulated 

wavelet energy. 

Max. 

hourly 

rainfall 

(mm) 

Flow 

speed 

(m/s) 

Warning 

announced * 

DF a 

arrived ** 

Max. 

accumulated 

wavelet 

energy (Jmax) 

Background 

wavelet 

energy (Jb) 

17 1.77 13:18 13:29 242.5 19.7 Y 

15 ~1.0 NA ~15:24 NA 31.45 Y 

51.5 8.52 6:47 6:54 7243.78 31.45 Y 

2014 

0520 

Heavy 

Rainfall 

39.5 4.87 NA *** 12:53 180.10 30.27 Y 

*, **: the time recorded based on the geophone at the upper stream of Aiyuzi River. 

***: communication unstable, no record. 

****: G for geophone (GS-20 DX) and BS for broadband seismograph (Yardbird DF-2) 

a: DF = Debris Flow 

DF? DF mass DF Scale 

mainly 

sediment (∅ 

50cm) 

medium 

small 

large 

medium 

to large 

Used for 

analysis 

(G or 

BS) **** 

G 

BS 

G & BS 

G & BS 

RESULTS 

The observation of three cases by the upstream geophone in the Aiyuzi Stream is shown in 

Fig. 4. It was noted that there peaks occurred in the signal before the arrive of debris flows. 

Fig. 5 shows the results of FFT, and it indicates that the significant frequency range of debris 

flow was less than 30 Hz in all three cases. The finding was in agreement with other 

researches (Fang et al., 2008). In order to understand the signature frequency, the Haar 

wavelet transform was applied to extract the signals of 0~31.25 Hz, as shown in Fig. 6. The 

signals of 0~31.25 Hz from the figure had more clear on peaks than those in Fig. 4. Therefore, 

the characteristic frequency of debris flow in Shenmu was about 0~31.25 Hz. In order to 

further analyze the relationship of signature frequency and the scale of debris flow, the 

spectrograms was used to find out the initial major frequency of debris flow front end. 

The spectrograms are shown in Fig. 7. Although the noise appeared in the figure, the energy 

distribution with time was still recognizable. In this figure, the frequency at which the energy 

propagated by the debris flow started to increase (the color changed from blue to red) was 

about 10 Hz or less in the cases, except the case of 1110 Rainfall in 2011. When considered 

with the arrival time of debris flow (Table 4), it was noted that the energy of frequency about 

5 Hz or less started to increase about 5 minutes (6:49) and 1 minutes (12:52) before the 

debris flow arrived at the location of geophone in the case of Typhoon Saulik 2013 and 0520 

Heavy Rainfall 2014, respectively. However, due the sensitivity of the geophone, the signals at 

low frequencies (

debris flows of large and medium to large scales, respectively. In contrast, the small debris 

flow of 0530 Rainfall 2013 had front-end major frequency of about 10 Hz (Table 5). Thus, 

the scale of a debris flow may be related to its front-end major frequency. 

Figure 4: The vibration signals of the upstream geophone. (a) 1110 Rainfall (Nov. 10, 2011) (b) Typhoon Saulik (July 13, 2013) (c) 0520 

Heavy Rainfall (May 20, 2014) 

In addition to the rainfall, the warning to a debris flow in Shenmu was also considered by 

applying geophone data. The test warnings were announced based on the accumulated 

wavelet energy estimated by the signals of the upstream geophone in Aiyuzi Stream. When 

the accumulated wavelet energy is 5 times the background value (Table 4), the warnings will 

be sent. From the past events, the testing warning rule of 5 times the background wavelet 

energy was considerably applicable. The time window between the warning and the arrival of 


debris flows at the upstream geophone was about 9 minutes in average at the Aiyuzi Steam. 

The 9-minute time window may not useful for evacuation during disaster response, but the 

direct measure of vibration signals provided distinct evidence of debris flow occurrence. 

Compared with the geophone data, the time window between the detected front-end signal 

and the arrival at the downstream of Aiyuzi Stream was about 4 minutes in average by the 

signals of broadband seismograph (Table 6). Both geophones and broadband seismographs 

were practically useful. 

Figure 5: The FFT results of the geophone signals. (a) 1110 Rainfall (Nov. 10, 2011) (b) Typhoon Saulik (July 13, 2013) 

(c) 0520 Heavy Rainfall (May 20, 2014) 


Figure 6: The signals of frequency 0~31.25 Hz of the upstream geophone. (a) 1110 Rainfall (Nov. 10, 2011) (b) Typhoon Saulik 

(July 13, 2013) (c) 0520 Heavy Rainfall (May 20, 2014) 


Figure 7: The spectrograms of signals at the upstream geophone. (a) 1110 Rainfall (Nov. 10, 2011) (b) Typhoon Saulik (July 13, 2013) 

(c) 0520 Heavy Rainfall (May 20, 2014) 

Table 5: Broadband major frequency and the scale of debris flows. 

Event Date Scale Flow speed Major frequency 

1 2013/5/30 small ~ 1 m/s 11 Hz 

2 2013/7/13 large ~ 10 m/s 2 Hz 

3 2014/5/20 median ~ 5 m/s 5 Hz 


Figure 8: The spectrograms of broadband seismograph in Aiyuzi River during (a) 0530 Rainfall 2013 (b) Typhoon Saulik in 2013 

(c) 0520 Heavy Rainfall in 2014. 

Table 6: Time earlier of observed frequency by the broadband seismograph at downstream to the occurrence of debris flow (6:57:39) 

during Typhoon Saulik 2013. 

Observed Frequency Recorded Time Time earlier 

2 Hz 6:53:25 4 m 14 s 

8 Hz 6:57:02 37 s 

20 Hz 6:57:25 14 s 

Peak wave 6:57:39 0 


CONCLUSION 

The data of broadband seismographs and geophones along the Aiyuzi River in Shenmu was 

used in this study, as well as 4 debris flow events. The frequency range of 0~10 Hz (broadband 

seismograph) and 0~31.25 Hz (short period seismograph) were practically suitable for 

debris flow warning in terms of vibration characteristics. It was noted that large debris flows 

had signature frequency less than 10 Hz, and the small to median debris flows usually had 

signature frequency greater than 10 Hz. The time window of response to the debris flow was 

about 4 minutes for broadband seismographs and 9 minutes in average for geophone from 

the cases in Shenmu. Warning levels of geophone signal were suggested to be 5 times the 

background energy (in joule) for debris flow warning. Overall, the study showed that the 

direct method of geophones and broadband seismograph was practically promising for debris 

flow monitoring. 

ACKNOWLEDGEMENT 

This research was partly supported by the Soil and Water Conservation Bureau, Council of 

Agriculture, Executive Yuan, Taiwan. Thanks to all people providing help on this study. 

REFERENCES 

- Chu C.-R., Huang C.-J., Lin C.-R., Wang C.-C., Kuo B.-Y., Yin H.-Y. (2014) “The Time- 

Frequency Signatures of Advanced Seismic Signals Generated by Debris Flows”, Abstract 

NH51B-3853 presented at 2014 Fall Meeting, AGU, San Francisco, Calif., 15-19 Dec. 

Fang Y.M. et al. (2008) Analysis of Debris Flow Underground Sound by Wavelet Transform- 

A Case Study of Events in Aiyuzih River, Journal of Chinese Soil and Water Conservation, 

CSWCS, 39(1), 27-44. 

- Huang C.-J., Chu C.-R., Tien T.-M., Yin H.-Y., and Chen P.-S. (2012) Calibration and 

deployment of a fiber-optic sensing system for monitoring debris flows, Sensors, 12(5), 

5835-5849. 

- Huang Y.M., Chen W.C., Fang Y.M., Lee B.J., Chou T.Y., Yin H.Y. (2013) Debris Flow 

Monitoring – A Case Study of Shenmu Area in Taiwan, Disaster Advances, 6(11), 1-9. 

- Jan C.D., Lee M.H., Huang T.H. (2003) Effect of Rainfall on Debris Flows in Taiwan, 

Proceedings of the International Conference on Slope Engineering, Hong Kong, 2, 741-751. 

- Jan C.D. and Lee M.H. (2004) A Debris-Flow Rainfall-Based Warning Model, Journal of 

Chinese Soil and Water Conservation, CSWCS, 35(3), 275-285 (in Chinese). 

- Lee B.J. et al. (2014) The 2012 On-site Data gathering and Monitoring Station Maintenance 

Program, Project Report, Soil and Water Conservation Bureau, 476 (in Chinese). 

- Lee M.H. (2006) A Rainfall-Based Debris Flow Warning Analysis and Its Application, Ph.D. 

Thesis, National Cheng Kung University, Taiwan, (in Chinese). 



Analysis of flood related processes at confluences of 

steep tributary channels and their receiving streams 

– 2d numerical modelling application 

Johannes Kammerlander, DI 1 ; Bernhard Gems, DI Dr. 2 ; Michael Sturm, DI 2 ; Markus Aufleger, Dr.-Ing. habil. 2 

ABSTRACT 

The work presented within this paper deals with the crucial hydraulic and morphologic 

processes at confluences of steep torrent channels and receiving streams in case of exceptional 

extreme events. 2d numerical modelling is accomplished with the BASEMENT software for 

the confluence of Schnannerbach torrent channel and Rosanna River. There, the damage 

causing flood event from August 2005 is reconstructed. Processes of bedload deposition, 

flooding and overbank sedimentation, as they could be observed in August 2005 and 

analysed within a physical model at the University of Innsbruck, are simulated. The modelling 

results provide a valuable insight into the processes being crucial for the damages on the 

adjacent flood plain. Compared to the laboratory analysis, which delivers a very reliable and 

vivid process representation but is restricted to a rather small spatial extent, numerical 

modelling allows for an analysis of bedload transport and deposition further downstream in 

the Rosanna River and backwater effects upstream. 

KEYWORDS 

bedload transport, event reconstruction, confluence zone, 2D model 

INTRODUCTION 

Experiences from recent torrential hazard events in the Alps reveal that confluences of steep 

tributaries and their receiving waters are often critical spots concerning flood risk. Due to 

massive supply of sediments from the torrent catchments and insufficient transport capacities 

in the receiving streams, processes of regressive aggradation appear, possibly leading to overbank 

flooding and sedimentation on the alluvial fan. In this respect, physical scale modelling 

of bedload transport processes proved to be a suitable tool for optimizing mitigation measures 

at confluence zones (Gems et al. 2014). 

However, to accurately reproduce natural conditions, the scale, and with it the similarity law, 

have to be defined carefully. In case of bedload pulses entering receiving waters, experimental 

modelling is typically restricted to the proximity of the confluence zone (Gems et al., 2014). 

In contrast, hydraulic and morphologic responses of the receiving streams extend far in 

upstream and downstream directions. The optimization of geometric patterns at the conflu- 

1 Austrian Service for Torrent and Avalanche Control, Salzburg, Salzburg AUSTRIA, johannes.kammerlander@die-wildbach.at; 

j.kammerlander@gmx.at 

2 Unit of Hydraulic Engineering, University of Innnsbruck 



ence zone yields a significant onward movement of sediment (Gems et al., 2014), but critical 

spots, featuring aggradation and flooding, may appear further downstream in the receiving 

water. In this regard, the area of risk is rather spacious and its extent strongly depends on the 

bedload transport capacity of the whole system rather than at the confluence only. In order 

to assess their impact on flood safety and to determine areas of risk, the application of a 

numerical tool is suggested to be an adequate method, since it easily copes with the large 

areal extent of the area of interest. 

METHODS 

General Remarks 

Numerical simulations of bedload transport in steep mountain streams are commonly 

accomplished by means of 1d hydrodynamic approaches. As long as water flow is laterally 

confined, these simplifications are of minor effects. In contrast, fluvial fans typically feature 

convex terrains with multiple flow paths in case of overbank flooding. Additionally, confluence 

zones exhibit complex flow patterns, since water enters from different directions. 

In order to account for these conditions, at least a 2d hydrodynamic model needs to be applied 

for the simulation of hydraulic and involved bedload transport processes. 

However, the set of equations in 2d hydrodynamic models is rather sophisticated (e.g. Vetsch 

et al., 2011) and its applicability to high gradient streams with huge sediment loads is hardly 

tested so far. To make a contribution to that, the simulation tool BASEMENT (© ETH Zurich) 

was applied to a case study in order to examine its performance. 

Case Study Event 

The case study comprises an extreme event in the Schnannerbach Torrent (Tyrol), which was 

intensively investigated and comprehensively documented concerning both, hydrologic 

(Chiari, 2008) and morphologic characteristics (Gems et al., 2014; Chiari, 2008; Hübl et al., 

2006; Figure 1); in these literature references the reader also finds a detailed overview of 

catchments characteristics. 

Summarizing, a heavy rain storm occurred in August 2005, which caused run-off generation 

and erosion along the steep scree slopes which are composed of lime stone rock and located 

in the upper catchment of the Schnannerbach Torrent. The total bedload that was transported 

to the fan apex was in the range of 36,000 m³ (Hübl et al., 2006) and 59,000 m³ (Gems et al., 

2014). However, about 25,000 m³ of sediment (volume including pores) deposited on the 

alluvial fan (Hübl et al., 2006), which means that up to 34,000 m³ of bedload has entered the 

receiving Rosanna River. 

However, in this study the event reconstruction of Gems et al. (2014) is used as boundary 

condition for the numerical investigations. In their study, the sediment flux that entered the 

lined trench at the fan was reconstructed by determining the system’s capacity (critical load in 

the confluence zone and transport capacity of the Schnannerbach channel) by means of 


Figure 1: The extent of the 2d model used for the simulation including the hydro- and sedigraphs at the upstream boundaries and the 

location of downstream boundaries which are defined by constant friction slopes; the purple points are used as spatial reference only. 

physical scale modelling. The boundary conditions as highlighted in Figure 1 sufficiently 

reproduced the spatial and temporal location of the system failures (overbank flooding) that 

were observed during the event in 2005. 

Model Extent 

The 2d numerical modelling tool BASEMENT (2006-2015) is used to simulate the hydraulics 

and the bedload transport of the considered flood event. Thereby, the numerical model covers 

the lower half of the alluvial fan of the Schnannerbach Torrent and extends over two 

kilometres of the receiving mountain stream (Rosanna River), with the confluence zone 

located in the middle (Figure 1). The computational grid of the modelling area (0.37 km²) is 

characterized by an unstructured mesh consisting of approx. 11,600 nodes which are either 

based on high resolution LiDAR data (floodplain) or bathymetric survey data (water courses). 

Summarizing, the alluvial fan features a mean gradient of 0.1 m/m and the channel is 

constructed as a lined trench with a sequence of small check dams (artificial steps) which 

prevents from channel bed erosion. The flow section of the artificial channel has a base and 

top width of approx. 5 m and 6 m and is on average 3 m deep. In contrast, the gradient of the 

receiving mountain stream is only 0.008 m/m and features a trapezoidal cross section with an 

average bed and top width of 16 m and 26 m, respectively. Although the geometrics of the 


channels and the floodplain are of high resolution, local structures crossing the channels 

(bridges) could not be considered. Hence, the model does not enable to simulate clogging of 

bridge cross sections with sediment due to pressurized flow, which had been observed during 

the prototype event (Gems et al., 2014). 

Except for the artificial steps and walls of the tributary channel, the water course of both, the 

Schnannerbach Torrent (tributary channel) and the Rosanna River (receiving mountain 

stream), was chosen to be mobile, allowing for river bed erosion. In addition, deposition and 

remobilization of bedload is considered within the entire computational domain. 

Parameter Settings 

The BASEMENT software simulates flow hydraulics by solving the shallow water equations 

that are commonly used to model a wide variety of physical phenomena (Vetsch et al., 2015). 

Next to the assumptions of a hydrostatic pressure distribution and steady-state resistance 

laws, this approach is only valid in case of small channel slopes (θ), with cos(θ)~1. The alluvial 

fan features a mean slope of 0.1 m/m, but the error is assumed to insignificantly influence the 

outcomes of this study. However, hydraulics in the overall steep and artificially stepped 

Schnannerbach Torrent are highly non uniform, with a permanent transition of super- and 

subcritical flow. Hence, solving the governing shallow water equations numerically might 

generate instabilities due to problems in convergence. In order to minimize these error 

sources, emphasis was put on (i) the specifications of parameters at the hydraulic boundaries 

(e.g. friction slope, weighting type) which were defined carefully by trial and error and (ii) a 

regularly distributed computation mesh consisting of triangular elements on the channel bed. 

As recommended by Vetsch et al. (2015), the element size was small in regions of abrupt 

changes in flow conditions, while it was coarse on the alluvial fan and the floodplain. 

Sensitivity tests confirmed the importance of these issues. 

A variety of flow resistance equations are available in BASEMENT. In this study, the approach 

of Manning-Strickler is used with spatially variable but temporally constant roughness scales 

(Strickler coefficients) of 23 m 0.33 s -1 for the channel bed of the Schnannerbach and 30 m 0.33 s -1 

for the Rosanna River. Thus, the friction attributed to the river bed does not differ in case the 

formerly bed surface is covered by deposited bedload, but its impact on flow hydraulics is 

assessed by changes of geometric patterns (e.g. the burial of the artificial steps causes the 

longitudinal profile to smooth). 

Bedload transport is calculated according the Meyer-Peter and Mueller (1949) equation 

which is extended to a fractionized approach in BASEMENT (Vetsch et al., 2015). 

The mobility of single grain sizes is determined by the hiding function that assumes equal 

mobility of all fractions finer than the D40 (grain size for which 40 % are finer by weight) 

and size selective mobility for coarser ones. Therefore, the grain size distributions were 

determined separately regarding both, the source (bed sediment and bedload) and the stream 

(Schnannerbach torrent and Rosanna River). 

In steep mountain streams, only a fraction of total shear stress is available for bedload 

transport. The form and spill drag around macro roughness elements, which are typically 


present in steep streams, act as momentum sinks. However, there is no approach available in 

the simulation tool BASEMENT to account for momentum losses due to macro roughness 

elements. But the artificial steps of the tributary channel (Figure 1) have a similar effect on 

the shear stress. While the energy gradient (and hence the shear stress) is very large at a 

certain step, it is comparatively small along the mobile channel between two steps. Hence, 

form resistance is assumed to be adequately reproduced by the high resolution of the channel 

geometrics only, rather than by additional, empirical equations. 

Next to the bedload transport that originates due to flow hydraulics, BASEMENT provides an 

option to account for gravitational transport, which is primarily attributed to river bank 

failures. However, gravitational induced relocations of sediment might also appear during the 

formation of large deposit cones due to slope failures or shallow landslides on the downwards 

facing slopes (Zollinger, 1983). In order to activate gravitational transport in BASEMENT, 

three critical failure angles must be defined beforehand. These distinguish for dry or wetted 

embankments or deposited sediment and are set to 35 °, 20° and 10°, respectively. 

RESULTS 

Basically, the results of numerical simulations reveal the capability of a 2d hydrodynamic 

modelling tool to accurately reproduce field and laboratory observations (Figure 2). However, 

there are still some areas where modelling results do not correspond to the aerial photograph. 

For instance the areal extent of overbank sedimentation on the alluvial fan is too small 

(Figure 2a), while deposit heights are generally too large when compared to the event 

reconstruction of Hübl et al. (2006). The cause of these differences is mainly attributed to 

local structures (e.g. walls, railings, etc.) that are insufficiently included in the model. In this 

respect, overbank flow might had been laterally confined by these structures and thus, the 

flow’s competence to transport sediment was higher than calculated. Additionally, numerical 

simulations insufficiently reproduce the overbank sedimentation of the Rosanna River 

(Figure 2a), although all except of the most upstream part is flooded with water (Figure 2b). 

Probably, these sedimentations refer to suspended load which is not accounted for in the 

numerical simulations. 

Despite these uncertainties, the failure mechanisms can be assessed in more detail and the 

results enable an event history analysis regarding the processes of bedload aggradation and 

their feedback on hydraulics and flood risk. 

Within the first few hours of the event, artificial steps in the tributary channel were filled up 

and thus, the longitudinal profile was smoothed which minimizes form resistance and 

maximizes flow competence to transport bedload (Figure 3a and 3b). However, the bedload 

that passed the tributary channel initially accumulated in the confluence zone (Figures 3b 

and 3c). There, an abrupt change in bed gradient from 0.07 m/m at the lowermost reach on 

the alluvial fan to less than 0.01 m/m in the Rosanna River appears. Despite the fact, that 

water discharge in the receiving stream is almost 7 times larger, transport capacity is less than 

in the small Schnannerbach channel. 


Figure 2: a) Magnitudes of deposition and erosion after the torrential hazard event (at the end of the simulation) and b) maximum flow 

depth of each node. The aerial photo (© Bundesamt für Eich- und Vermessungswesen) in the background of the figures shows the study 

area a few days after the extreme event (see also Figure 1 and Hübl et al., 2006). 

Figure 3: Evolution of the bed level in the lowermost reach of the tributary channel at certain times (a-d) during the case study event 

Consequently, the entering bedload forms a deposit cone. Thereby, the level of the cone’s 

crest defines the downstream base level of the tributary channel which contributes to a 

decrease of slope with growing bedload accumulation in the confluence zone (Figure 3c). 


By that, the transport capacity decreases as well, resulting in severe channel bed aggradation 

(Figure 3d). Hence, bedload accumulation successively propagates against the flow direction 

of the tributary channel and the channel’s flow capacity decreases until it is insufficient to 

cope with the discharge from the tributary catchment. Due to overbank flooding on the 

alluvial fan, the channel’s transport capacity rapidly decreases once more with most of the 

entering sediment depositing on the alluvial fan. In addition, modelling results further 

highlight the impact on flow hydraulics in the upstream reach of the receiving stream. 

Backwater effects caused a dramatic increase of the water level which was accompanied with 

the flooding of agricultural areas nearby the receiving stream (Figure 2b). 

According to the boundary condition in the Schnannerbach Torrent (Figure 1), the sediment 

load diminished after about 14 hours (Figure 1). As a consequence, the water cuts into the 

sediment filled channel by remobilizing the accumulated bedload. Corresponding to eye 

witness observations of the event in August 2005, the lined trench of the tributary channel is 

almost free from sediment at the end of the simulation. It is worth to note that channel 

incision propagates forward, starting at the upstream end and thus, differs from aggradation. 

However, there remain large sediment accumulations on the alluvial fan and in the confluence 

zone (Figure 2). 

Although the transport capacity of the receiving stream is far less than the sediment input 

from the tributary catchment, a significant amount of bedload is transported further 

downstream during the event. According to the numerical simulations, this is accompanied 

Figure 4: Time series of a) bedload transport rate and b) accumulated bedload transport (volumes refer to solid volumes) at certain 

locations in the receiving Rosanna River 

with an obvious river bed aggradation downstream of the confluence zone leading to a 

reduction of flow capacity, followed by overbank flooding and sedimentation. According to 

Gems et al. (2014) about 15,400 m³ of bedload passed the downstream end of their physical 

scale model (120 m downstream of the confluence zone) within the first 10.5 hours of the 

event that properly matches with the simulation results (11,700 m³; Figure 4). In total, a 

considerable fraction (approx. 25,000 m³) of the entire sediment pulse that originated from 


the tributary channel (59,000 m³ according to Gems et al., 2014) is remobilized at the 

confluence zone during the event. But the propagation distance is limited to a few hundred 

meters (Figure 4) with severe river bed aggradation and overbank flooding in this region 

(Figure 2). 

CONCLUSION 

The application of a 2d hydrodynamic simulation tool is suitable to reproduce the complex 

interactions of bed morphology (aggradation / incision) and flow hydraulics present at a river 

confluence zone during an exceptional event in the tributary catchment. It is worth to note 

that simulation results properly matched with observations although most parameters 

referred to their default values. 

Results reveal that the magnitude of bedload accumulation in the confluence zone majorly 

controls channel bed aggradation and overbank flooding in both, the tributary and the 

receiving stream; at least in this case study event. In terms of modelling, the growing rate of 

the deposit cone strongly depends on the angle of response of the deposited bedload, which 

was defined by 10°. However, there is a lack of knowledge regarding this parameter and thus, 

more emphasis is needed to evaluate the performance of 2d hydrodynamic modelling under 

different geometric configurations. 

LITERATUR 

- BASEMENT – Basic Simulation Environment for Computations of Environmental Flow and 

Natural Hazard Simulation. Version 2.5. ©ETH Zurich, VAW, Vetsch, D., Siviglia A., Ehrbar D., 

Facchini M., Gerber M., Kammerer S., Peter S., Vonwiller L., Volz C., Farshi D., Mueller, R. 

Rousselot P., Veprek R., Faeh R., 2006-2015. 

- Chiari M. 2008. Numerical modelling of bedload transport in torrents and mountain streams. 

PhD Thesis at the University of Natural Resources and Applied Life Sciences (BOKU), Vienna. 

Gems B., Sturm M., Vogl A., Weber C., Aufleger M. 2014. Analysis of damage causing hazard 

processes on a torrent fan – scale model tests of the Schnannerbach Torrent channel and its 

entry to the receiving water. Digital Proceedings of the Interpraevent 2014 in the Pacific Rim, 

Nara. 

- Hübl, J., Ganahl, E., Bacher, M., Chiari, M., Holub, M., Kaitna, R., Prokop, A., Dunwoody, 

G., Forster, A., Schneidenbauer, S., 2006. Dokumentation der Wildbachereignisse vom 22./23. 

August 2005 in Tirol, Band 2: Detaillierte Aufnahme (5W+ Standard); IAN Report 109 Band 2, 

Institute of Natural Hazards, University of Resources and Life Sciences Wien. 

- Vetsch D., Siviglia A., Ehrbar D., Facchini M., Gerber M., Kammerer S., Peter S., Vonwiller L., 

Volz C., Farshi D., Mueller, R. Rousselot P., Veprek R., Faeh R. 2015. System Manuals of 

BASEMENT, Version 2.5. Laboratory of Hydraulics, Glaciology and Hydrology (VAW), 

ETH Zurich. 

- Zollinger, F. 1983. Die Vorgänge in einem Geschiebeablagerungsplatz: ihre Morphologie und 

die Möglichkeiten einer Steuerung, PhD Thesis, ETH Zurich, No. 7419, 265 p (in German). 



Analysis and classification of bedload transport events 

with variable process characteristics 

Analyse und Klassifikation von Geschiebetransportereignissen 

mit unterschiedlichen Prozesscharakter 

istika 

Andrea Kreisler 1 ; Markus Moser²; Johann Aigner³; Rolf Rindler³; Michael Tritthart³; Helmut Habersack³ 

ABSTRACT 

In Salzburg/Austria an integrative monitoring system has been installed in the year 2011, 

which combines direct (mobile basket sampler, bedload trap) and indirect (geophone plates) 

measuring devices. Comprehensive results have been achieved in the monitoring period. 

The application of a geophone device enables the permanent recording of the intensity and 

distribution of bedload transport within the channel cross-section in high spatial and 

temporal resolution. The calibration of the geophone data is performed using results of direct 

bedload measurements and gives us the opportunity to calculate bedload rates and bedload 

yields in selected time periods. This work presents various bedload transport events at the 

Urslau stream and offers insights concerning measured bedload rates and bedload yields. 

The assessment of the monitoring data further shows that sediment availabilty is a determining 

factor for prevalent bedload fluxes at the Urslau stream. 


In Salzburg/Österreich wurde 2011 ein integratives Monitoringsystem installiert, bei dem 

direkte (mobiler Geschiebefänger, Geschiebefalle) und indirekte (Geophonanlage) Messgeräte 

kombiniert werden. Umfassende Ergebnisse konnten im Monitoringzeitraum gewonnen 

werden. Durch den Einsatz der Geophone wird die Intensität und Verteilung des Geschiebetransportes 

im Bachquerschnitt in hoher räumlicher und zeitlicher Auflösung erfasst. 

Die Kalibrierung der Geophondaten erfolgt über die Ergebnisse der direkten Geschiebemessungen 

und ermöglicht uns die Berechnung von Geschiebetransportraten und –frachten 

in beliebigen Zeiträumen. Diese Arbeit präsentiert unterschiedliche Geschiebetransportereignisse 

an der Urslau und zeigt Ergebnisse über gemessene Transportraten und Geschiebefrachten 

auf. Die Analyse der Monitoringergebnisse hat weiters ergeben, dass die Sedimentverfügbarkeit 

einen entscheidenden Einfluss auf vorherrschende Geschiebetransportraten 

an der Urslau hat. 

1 Universität für Bodenkultur, Institut für Wasserwirtschaft, Hydrologie und konstruktiver Wasserbau, Vienna, Österreich, 

andrea.kreisler@boku.ac.at 

2 Wildbach- und Lawinenverbauung, Österreich 

3 Universität für Bodenkultur, Institut für Wasserwirtschaft, Hydrologie und konstruktiver Wasserbau, Vienna, Österreich 



KEYWORDS 

bedload transport; integrative monitoring system; bedload transport events 

EINFÜHRUNG 

Die Bedeutung eines umfangreichen Geschiebemonitorings für Wissenschaft und Praxis, 

durch welches Geschiebetransportraten, -frachten und Korngrößen gewonnen werden 

können, ist unumstritten. Die Messung des Transportprozesses ist oftmals herausfordernd, 

wird jedoch vermehrt von unterschiedlichen Einrichtungen betrieben (z.B. Bunte et al., 

2004, Garcia et al., 2000, Habersack et al., 2001, Lenzi et al., 2004, Rickenmann et al., 2012). 

Naturmessdaten des Geschiebetransportes zeigen häufig eine hohe zeitliche und räumliche 

Variabilität auf, wodurch eine starke Streuung der Geschiebetransportraten bei vergleichbaren 

hydraulisches Bedingungen beobachtet werden kann (Habersack et al., 2008, Mao et 

al.,2014, Turowski et al., 2009). Wir präsentieren in dieser Arbeit das integrative Geschiebemesssystem, 

durch welches seit 2011 kontinuierliche Daten des Geschiebetransportprozesses 

an der Urslau gewonnen werden. Die Messstation wurde vom Institut für Wasserwirtschaft, 

Hydrologie und konstruktiven Wasserbau / Universität für Bodenkultur Wien in enger 

Zusammenarbeit mit der Wildbach- und Lawinenverbauung in Salzburg errichtet. Es werden 

in dieser Arbeit Geschiebefrachten und Transportraten von unterschiedlichen Ereignistypen 

angegeben. Die Auswertung der Daten zeigt unterschiedliche Geschiebetransportraten je nach 

Ereignistyp und Abflussmenge. Es wird dargestellt, dass die Geschiebeverfügbarkeit einen 

großen Einfluss auf aktuelle Transportraten hat. 

PROJEKTGEBIET 

Das Projektgebiet befindet sich in Salzburg/Österreich, nahe Maria Alm (Abb.1a). Die Urslau 

entspringt zwischen dem Hochkönig und dem Steinernen Meer (Abb.1b) und mündet nach 

18,8km Lauflänge in die Saalach. Das mittlere Gefälle ist 4% und die Größe des Einzugsgebietes 

beträgt 122 km². Am Pegel Saalfelden beträgt der Mittelwasserdurchfluss (MQ) 4,71 m³/s 

(Hydrografisches Jahrbuch 2012, Reihe 1951-2011). Am Unterlauf der Urslau, rund 7.5 km 

flussauf der Mündung in die Saalach, wurde im Jahr 2011 eine Geschiebemessstation 

errichtet (Abb.1c). Hier ist die Bachbreite 8m. Die charakteristische Korngröße D50 des 

Sohlsubstrats ist 0,05 m und D84 ist 0,09 m. 


Abbildung 1: (a) geografische Übersicht; (b) Aufnahme des Entstehungsgebiets; (c) Fotografie der Messstation; (d) Geschiebefalle und 

Geophonanlage; (e) Behälter Geschiebefalle; (f) Fangkorb 

GESCHIEBEMONITORING 

An der Station kommt ein Integratives Geschiebemesssystem zum Einsatz, in welchem 

direkte (mobiler Geschiebefänger, Geschiebefalle) und indirekte (Geophone) Messmethoden 

kombiniert werden. Im Folgenden wird ein Überblick über die Messgeräte gegeben, genauere 

Informationen werden in Kreisler et al. (in Bearbeitung) zur Verfügung gestellt. In Habersack 

et al. (in Bearbeitung) werden detaillierte technische Informationen und Anwendungsmöglichkeiten 

präsentiert. 

Das Konzept des Geschiebefängers (Abb. 1f) basiert auf den mobilen Geschiebefallen, welche 

in Bunte et al. (2004) präsentiert werden. Geschiebemessungen werden mit Hilfe eines 

Kranwagens durchgeführt. Die Messung mit dem Fangkorb ist auf niedrige Durchflüsse 

beschränkt. In höheren Durchflussbereichen wird die Geschiebefalle eingesetzt (Abb. 1d und 

1e). Die Geschiebefalle ist in der Mitte des Profils in der Sohle eingebaut. Im Ereignisfall kann 

ein Messschlitz über einen Hydraulikmechanismus geöffnet werden und das Geschiebe fällt in 

einen Sammelbehälter, der auf Wägezellen gelagert ist. Über die automatisch aufgezeichnete 

Massenzunahme wird der Geschiebetrieb erfasst. Durch den Einsatz der direkten Messgeräte 

kann der Geschiebetransport im gesamten Profil und die Korngrößen erfasst werden. 

Zusätzlich ist der Einsatz der direkten Messgeräte erforderlich, um die Geophone zu kalibrieren. 

Flussauf der Geschiebefalle sind sieben Geophone, die an Stahlplatten montiert sind, 

regelmäßig über den Querschnitt verteilt eingebaut (Abb.1d). Geschiebepartikel, die über die 

Platte transportiert werden, erzeugen Vibrationen, die vom Geophon registriert werden. 


Dieses Signal wird in ein elektrisches Spannungssignal transformiert und von einem Computersystem 

weiterverarbeitet. Bei der Verarbeitung des Rohdatensignals werden über eine 

Software unter anderem jede Minute die Anzahl der Impulse (Überschreiten eines Schwellenwertes 

von 0,1 Volt) aufgezeichnet. Durch diese automatisierte Messmethode wird der 

Geschiebetransport sowohl über den gesamten Querschnitt als auch zeitlich kontinuierlich 

(Minutenwerte) aufgezeichnet. 

Die Kalibrierung der Geophone, und damit die Möglichkeit die registrierten Geophonimpulse 

in Massen des transportierten Geschiebes umzurechnen, erfolgt durch eine Gegenüberstellung 

der Ergebnisse der direkten Messmethoden und der aufgezeichneten Geophonimpulse. 

Geophonimpulse werden nur von Geschiebekörnern größer als 10-20mm ausgelöst (Rickenmann 

et al., 2012). Daher wird für die Erstellung der Kalibrierungsfunktion nur Geschiebematerial 

>10mm berücksichtigt. Daraus folgt, dass berechnete Geschiebetransportraten und 

–frachten sich nur auf Geschiebematerial >10mm beziehen. In Kreisler et al. (in Bearbeitung) 

wird die Kalibrierungsfunktion präsentiert. Es wird gezeigt, dass ein linearer Zusammenhang 

zwischen dem gemessenen Geschiebetrieb und den registrierten Geophonimpulsen besteht 

und mit der in Formel 1 angegebenen Funktion beschrieben werden kann: 

Formel 1: Kalibrierungsfunktion nach Kreisler et al. (in Bearbeitung) 

wobei IMP den Mittelwert der registrierten Geophonimpulse im Messzeitraum [imp m -1 s -1 ] 

und qb den direkt gemessenen Geschiebetrieb D>10mm [kgm -1 s -1 ] bezeichnet. 

DATENGRUNDLAGE 

Die Auswertung der hier präsentierten Analyse umfasst die Messdaten des Jahres 2012. 

Es werden basierend auf 15 min Mittelwerten der Geophonimpulse Geschiebetransportraten 

und -frachten über die Kalibrierungsfunktion in Formel 1 berechnet. Die Durchflussdaten 

werden in 15 min Intervallen vom hydrografischen Dienst bezogen. Niederschlags- und 

Temperaturdaten werden von der Messstation 11137 in Maria Alm von der Zentralanstalt 

für Meteorologie und Geodynamik zur Verfügung gestellt. 

GESCHIEBETRANSPORTEFFIZIENZ 

Neben der Ermittlung von Geschiebetransportraten und Frachten kann durch die kontinuierliche 

Erfassung des Transportprozesses auch der Zusammenhang zwischen Geschiebetransport 

und Durchfluss untersucht werden. Wir haben in der Monitoringperiode beobachtet, dass die 

Beziehung zwischen Geschiebetransport und Durchfluss bei den unterschiedlichen Ereignissen 

verschieden ist. Um diesen Zusammenhang darzustellen, wurde für alle analysierten 

Geschiebetransportereignisse die Geschiebetransporteffizienz (in Folge mit ep bezeichnet) 

berechnet. Diese wird in Bagnold (1966) als Verhältnis zwischen „bedload work rate“ und 

„stream power“ definiert und berechnet sich wie folgt: 


Formel 2: Geschiebetransporteffizienz ep nach Bagnold (1966) 

wobei ib die Geschiebetransportrate unter Wasser darstellt: i b =q v g(ρ s -ρ) mit q v – volumetrische 

spezifische Geschiebetransportrate, g – Erdbeschleunigung, ρ s – Feststoffdichte, q – Fluiddichte. 

In Formel 2 ist ω - stream power mit ω = ρgqS, worin q – spezifischer Durchfluss und 

S – Gefälle. Tan α ist der Reibungswinkel und wurde, wie in Recking (2012) vorgeschlagen, 

in dieser Analyse mit α = 52° gewählt. Prinzipiell stellt e p das Verhältnis aus transportiertem 

Geschiebematerial und hydraulischen Verhältnissen dar und spiegelt im Zuge dieser Analyse 

die Intensität eines Ereignisses wieder. 

ERGEBNISSE 

Abbildung 2a zeigt die Ganglinie der Tagesmittelwerte des Geschiebetransportes und des 

Durchflusses des Jahres 2012 an der Urslau. Zusätzlich sind die Tagesssummen des Niederschlags 

und die Monatsmittelwerte der Temperatur an der Station 11137 in Maria Alm 

(Daten der Zentralanstalt für Meteorologie und Geodynamik) abgebildet. Für die folgende 

Analyse werden drei Bereiche (B1-B3) der Ganglinie herausgehoben. 

Bei Bereich B1 handelt es sich um die Schneeschmelzperiode. Die Niederschlagssummen sind 

verglichen mit den Sommermonaten gering und die Tagesmittelwerte des Durchflusses hoch. 

Die Ganglinie in Abbildung 2a veranschaulicht sehr gut, dass in dieser Schneeschmelzperiode 

die höchsten Tagesmittelwerte des Durchflusses vorherrschten. Ein exemplarischer Bereich 

(3 Tage) der Schneeschmelzperiode ist detailliert (15 min Mittelwerte) in Abbildung 2b 

dargestellt. Ein deutlicher Tagesgang der beiden Messwerte ist ersichtlich. Für die weitere 

Analyse wurde der Bereich B1 in Teilereignisse gegliedert. Diese Gliederung erfolgte nach 

hydrologischen Verhältnissen (Durchflussspitzen). Insgesamt umfasst Bereich B1 10 Teilereignisse, 

wobei davon TE2, TE3 und TE4 in Abbildung 2b dargestellt sind. 

Der zweite Bereich B2 umfasst ein lang andauerndes Geschiebetransportereignis im Sommer. 

Der Anfang des Ereignisses ist in 15 Minuten Auflösung in Abbildung 2b dargestellt. 

Bei diesem Ereignis wurde über einen Zeitraum von ungefähr 40 Tagen kontinuierlich 

Geschiebetransport erfasst. Dieser Bereich B2 wurde im Zuge der folgenden Analyse in 

10 Teilereignisse geteilt. 

Auf das langandauernde Geschiebeereignis (Bereich B2) folgen im Jahr 2012 7 weitere 

kleinere Ereignisse. Diese Folgeereignisse werden im Bereich B3 zusammengefasst. Das 

höchste dieser Folgeereignisse trat in der Nacht vom 23.08.2012 auf den 24.08.2012 auf und 

dauerte nur 10 Stunden. Abbildung 2b zeigt dieses Ereignis. Die Geschiebefracht des Bereichs 

B1 (Schneeschmelze) liegt bei 6500 t (27.04.12-06.05.12). Die mittlere Fracht von den Teilereignissen 

ist 450 t. Die mittlere Geschiebetransportrate von B1 beträgt 4.5 kgs -1 . Im Zuge 

des in Bereich B2 dargestellten Ereignisses wurde in einem Zeitraum von 40 Tagen eine 

Geschiebefracht von ca. 19800 t transportiert. In diesem Zeitraum betrug die mittlere Transportrate 

5.8 kgs -1 . Die Geschiebefracht des Ereignisses, welches in Abbildung 3c dargestellt ist, 

beträgt 274 t. Hierbei wurde ein maximaler Geschiebetransport von 63 kgs -1 gemessen. 


Abbildung 2: (a) Jahresganglinie von Durchfluss, Geschiebetransportrate, Niederschlagssumme und Temperatur; (b) ausgewählte 

Ereignisse 

Wie im Kapitel „Geschiebetransporteffizienz“ beschrieben, wurde für die analysierten Ereignisse 

der Wert e p nach Bagnold (1966) (Formel 2) berechnet und in Abbildung 3a aufgetragen. 

In Abbildung 3b werden jeweils die Mittelwerte des Geschiebetransportes und des 

Durchflusses der jeweiligen Teilereignisse dargestellt. Es zeigt sich in dieser Auswertung, dass 

der mittlere Durchfluss im Zeitraum der Schneeschmelze vergleichbar hoch ist. Die Abbildung 

veranschaulicht weiter, dass unter den analysierten Ereignissen die mittleren Geschiebetransportraten 

von TE1 und TE2 in Bereich 2 am größten waren. Es ist auffallend, dass bei TE10 

am Ende des Bereichs B2 der mittlere Geschiebetransport abnimmt, obwohl der Durchfluss, 

im Vergleich zu den vorherigen Werten, relativ konstant bleibt. Selbiges gilt für die Teilereignisse 

in Bereich B3. Der mittlere Durchfluss variiert hier zwischen 3.9 und 5.6 m 3 s -1 , jedoch 

nimmt die mittlere transportierte Geschieberate ab. Diese Beobachtung wird auch durch die 

Ergebnisse für e p in Abbildung 3a bestätigt. Hier werden die berechneten Werte e p pro 

Teilereignis in einem Box-Plot Diagramm dargestellt. Berechnete e p Ergebnisse für die 

Teilereignisse ‚TE1 und TE2 in Bereich 2 sind im Vergleich zu den anderen Werten sehr hoch, 


woraus die Autoren in diesem Teilbereich auf intensiven Geschiebetransport schließen. Von 

TE3 bis TE9 (Bereich 2) liegen die Werte in der gleichen Größenordnung und werden erst 

bei dem Teilereignis TE10 (Bereich 2) niedriger, da bei vergleichbaren Durchflussraten ein 

geringerer Geschiebetransport registriert wurde. Die Werte für e p der Ereignisse in Bereich 3 

liegen in derselben Größenordnung wie Ereignis TE10 (Bereich B2). Es zeigt sich in Abbildung 

2 deutlich, dass die Beziehung zwischen Geschiebetransport und Durchfluss bei den 

unterschiedlichen Ereignissen nicht konstant ist. Wir interpretieren für die hier dargestellten 

Ereignisse, dass während der Durchflussspitze von Bereich 2 (TE1 und TE2) viel Geschiebematerial 

mobilisiert und im weiteren Verlauf des Ereignisses transportiert wurde. Erst im 

Teilereignis TE10 kommt es zu einer verminderten Geschiebeverfügbarkeit und dadurch zu 

einem geringeren Wert von e p für die folgenden Ereignisse. Im Zeitraum der Schneeschmelze 

konnten sehr hohe Durchflussraten beobachtet werden, jedoch vergleichbar moderate 

Geschiebetransportraten. Die berechneten e p Werte liegen im Bereich der Teilereignisse von 

Bereich 3. Wir folgern daraus, dass im Zeitraum der Schneeschmelze die lang anhaltenden 

hohen Abflüsse zu einem Ausräumen des Gerinnes und damit reduziertem Geschiebedargebot 

an der Gewässersohle geführt haben und durch die bestehende Schneedecke der 

Geschiebeeintrag aus dem Einzugsgebiet eingeschränkt war. Zusätzlich fehlen dem Abflussprozess, 

der in diesem Zeitraum durch kontinuierliche hohe Werte charakerisiert ist, intensive 

und rasch ansteigende Durchflussspitzen. Verminderte Geschiebetransportraten in der 

Schnee schmelzperiode bei vergleichbaren Durchflusswerten wurden unter anderem von Mao 

et al., 2014 und Rickenmann et al., 1998 beobachtet. 

Abbildung 3: (a) Berechnete Geschiebetransporteffizienz; (b) Mittlerer Geschiebetransport und Durchfluss für die ausgewählten 

Teilereignisse. 


FAZIT 

Durch den Einsatz des integrativen Geschiebemesssystems an der Urslau können wichtige 

Messdaten über den Geschiebetransport in einem Wildbachunterlauf gewonnen werden. 

Durch das langjährige Monitoring konnte eine Kalibrierungsfunktion für die Geophone 

erstellt werden. Es wird somit ermöglicht, aus den kontinuierlichen und automatisch 

erfassten Geophondaten Geschiebetransportraten und -frachten in beliebigen Zeiträumen zu 

ermitteln. In dieser Arbeit werden die Daten eines Messjahres präsentiert, und es werden 

Geschiebefachten und Transportraten von ausgewählten Ereignissen an der Urslau angegeben. 

Diese Werte sind nicht nur für wildbachtechnische Fragestellungen in der Praxis wichtig, 

sondern ermöglichen auch ein erweitertes Prozessverständnis und bilden die Basis für die 

Kalibrierung und Weiterentwicklung von Transportformeln sowie numerischen Feststofftransportmodellen. 

Wir haben in dieser Arbeit dargestellt, dass an der Urslau hinsichtlich des 

Geschiebetransportverhaltens unterschiedliche Ereignisse auftreten. Es zeigt sich, dass die 

Beziehung zwischen Durchfluss und Geschiebetransport bei den unterschiedlichen Ereignissen 

nicht konstant ist. Schmelzbedingte Ereignisse liefern andere Transportraten als Ereignisse 

die durch Niederschlag bedingt sind. Bei langandauernden Ereignissen und Folgeereignissen 

konnten wir eine Abnahme des Geschiebetransportes bei vergleichbaren hydraulischen 

Verhältnissen beobachten. Die Vorgeschichte dieser Folgeereignisse führte zu einem „Ausräumen“ 

des leicht mobilisierbaren Materials an der Sohle und den Erosionsquellen am Ufer. 

Auch die Ausbildung einer Deckschicht wird von den Autoren angenommen. Wir folgern auf 

Basis des hier dargestellten Messjahres, dass an der Urslau unterschiedliche Geschiebeverfügbarkeiten 

und die Vorgeschichte eines Ereignisses den vorherrschenden Geschiebetransport 

beeinflussen. 

LITERATUR 

- Bagnold R. A. (1966): An approach to the sediment transport problem from general 

pyhsics, Geological survey Professional Paper. Washington. 

- Bunte K., Abt S., Potyondy J., Ryan S. (2004): Measurement of Coarse Gravel and Cobble 

Transport Using Portable Bedload Traps, in: Journal of Hydraulic Engineering, 130, S. 

879-893. 

- Garcia C., Larone J.B., Sala, M. (2000): Continuous monitoring of bedload flux in a 

mountain gravel-bed river, in: Geomorphology, 34, S. 23-31. 

- Habersack H., Kreisler A., Aigner J., Rindler R., Seitz H., Liedermann, M. (in Bearbeitung). 

Integrative automatic bedload transport monitoring 

- Habersack H., Nachtnebel P., Laronne J. (2001): The continuous measurement of bedload 

discharge in a large alpine gravel bed river, in: Journal of Hydraulic Research, 39, S. 125-133. 

- Habersack H., Seitz H., Laronne J. (2008): Spatio-temporal variability of bedload transport 

rate: Analysis and 2D modelling approach, in: Geodinamica Acta, 21, S. 67-79. 

- Kreisler A., Moser M., Aigner J., Rindler R., Tritthart M., Habersack H. (in Bearbeitung): 

Analysis and classification of bedload transport events with variable process characteristics 


- Lenzi M. A., Mao L., Comiti F. (2004). Magnitude-frequency analysis of bed load data in an 

Alpine boulder bed stream, in: Water Resources Research, 40, W07201. 

- Mao L., Dell'Agnese A., Huincache C., Penna D., Engel M., Niedrist G., Comiti F. (2014). 

Bedload hysteresis in a glacier-fed mountain river, in: Earth Surface Processes and Landforms, 

39, S. 964-976. 

- Recking A. (2012). Influence of sediment supply on mountain streams bedload transport, 

in: Geomorphology, 175–176, S. 139-150. 

- Rickenmann D., D’Agostino V., Fontanta G., Lenzi M., Marchi L. (1998). New results from 

sediment transport measurements in two Alpine torrents. IAHS Publication no. 248, 

S.283-289. 

- Rickenmann D., Turowski J. M., Bruno F., Angela K., Andreas L. (2012). Bedload transport 

measurements at the Erlenbach stream with geophones and automated basket samplers, in: 

Earth Surface Processes and Landforms. 

- Turowski J. M., Yager E. M., Badoux A., Rickenmann D., Molnar P. (2009). The impact of 

exceptional events on erosion, bedload transport and channel stability in a step-pool channel, 

in: Earth Surface Processes and Landforms, 34, S. 1661-1673. 



Integration of remote and terrestrial monitoring data 

for analysing alpine geomorphic processes – examples 

from Switzerland and Italy 

Volkmar Mair, Ph.D. 1 ; David Mosna, MSc 1 ; Robert Kenner, Ph.D 2 ; Giulia Chinellato, MSc 3 ; Marcia Phillips, Ph.D 2 ; 

Claudia Strada, Ph.D 1 ; Benni Thiebes, Ph.D. 3 

ABSTRACT 

Radar satellite systems represent a viable solution for monitoring geomorphic processes and 

slope instabilities in alpine environments as they cover large areas and have a sufficient 

spatial and temporal resolution. In this paper, we present some results of the Interreg project 

SloMove that focused on monitoring slow mass movements in high alpine areas using 

terrestrial and remote sensing techniques. We found that DInSAR is well suited for monitoring 

of geomorphic processes if the test sites are carefully selected, matching the geographical 

situation of the slope with the geometric conditions of the satellites. Movement trends and 

areas of increased activity can then be identified with high reliability. However, satellite 

remote sensing needs to be supported by terrestrial measurements, particularly in the case 

where the results aim for supporting civil protection purposes. The validation with groundbased 

methods, in our case DGNSS and TLS, showed that the magnitude of displacement 

cannot be assessed with the same accuracy and that punctual data represent good reference 

points for comparisons of the different measurement techniques. 

KEYWORDS 

InSAR; GNSS; TLS; rock glaciers; monitoring 

INTRODUCTION 

Over the last ten years, space-borne methods have become more frequently applied for the 

monitoring of landslides, rockslides and active rock glaciers. Differential interferometric 

synthetic aperture radar (DInSAR) has particularly gained in popularity due to its ability to 

detect and monitor ground displacements along the line of sight with very high accuracy (in 

the range of cm to mm). The main benefit of DInSAR is its possibility to investigate large 

areas at relatively low costs by analysing the phase differences from sets of SAR images. 

Several examples of SAR applications to geomorphic processes have been presented recently, 

e.g. for landslides and rock glaciers (Calò et al., 2014; Liu et al., 2012; Mair et al., 2008; 

Papke et al., 2012; Schlögel et al., 2015). 

1 Office for Geological Surveys and building material test, Autonomous Province of Bozen/Bolzano, Italy, david.mosna@provinz.bz.it 

2 WSL Institute for Snow and Avalanche Research SLF, Switzerland 

3 Institute for Applied Remote Sensing, European Academy of Bozen/Bolzano (EURAC), Italy 



Nevertheless, DInSAR has some limitations, in particular if the method is applied to alpine 

environments. These are related to the recording conditions of the acquiring satellite such as 

ascending and descending tracks, look direction, shadowing or the conditions at the study 

sites. In addition, temporal changes of surface reflectivity, e.g. related to vegetation or snow, 

can make it difficult to correctly correlate the images. However, the use of natural persistent 

scatterers may help to enhance the correlation and therefore allows analysing a greater set of 

images and larger time spans. Areas far from settlements and without infrastructure often 

only have few fix points that can be identified or presumed to be persistent or permanent 

scatterers (PS). The installation of artificial corner reflectors can provide ideal PS, as it allows 

to detect a known point and should thereby help to minimise de-correlation (see Mair et al. 

Interpraevent 2016). To have an independent control for the slope movements, we analysed 

two DInSAR test sites with terrestrial laser scanning (TLS) which required the installation of 

reference targets. Moreover, differential GPS (DGPS) measurements were carried out utilising 

the corner reflectors which were specifically designed to be used in combination with GNSS 

systems. Reference measurements were not only taken within the actively moving areas but 

also included measurements in the areas known to be not affected by displacements 

TEST SITES 

Test sites have to be selected very carefully due to the limitations given by the slopes and the 

restrictions implicated by the satellite system. For slopes factors such as orientation, geometry, 

the assumed movement rates and the vegetation and snow cover have to be considered very 

carefully, and in particular the presence and the distribution of natural persistent scatterers. 

For the satellite system the orbital period, the direction of the tracks, the line of sight and the 

look direction (geometrical distortion, shadowing) have to be determined, amongst other 

factors. In addition especially for the TLS-monitoring and the maintenance of the corner 

reflectors and the GPS, the logistics must be considered, including accessibility and, last but 

not least, the expenditure of time for every single measuring campaign. Careful planning is 

necessary for a long-term monitoring system that takes into account the potential of these 

methods. 

Within the SloMove project presented here, two test sites were identified, located between 

2500 and 3000 m asl, in South Tyrol, Italy, and Grisons Canton, Switzerland. The Italian site 

is located in the NE of Schnalstal (Val Senales), in the Kurzras (Maso Corto) ski resort, 

Schnals Municipality, Autonomous Province of Bolzano. The eastern flank of the Steinschlagspitze 

(Punta delle Frane) is affected by a rockslide that evolves into different rock glacier 

lobes, while the northern part features a large active rock glacier. The Swiss test site Foura da 

l’amd Ursina is located on the Schafberg above Pontresina, Upper Engadine, Canton Grisons, 

and includes three active rock glaciers named Ursina I-III. 


Figure 1: Sketch map of the two test sites in Graubünden and South Tyrol 

METHODS 

Monitoring of the test sites was carried out in 2012, 2013 and 2014 using the following 

methods: (i) DInSAR, (ii) TLS, and (iii) DGPS. 

DInSAR processing relied on SAR images acquired by COSMO SkyMed CSK ® and used the 

Small BAseline Subset (SBAS) algorithm (Berardino et al., 2002). The SBAS approach uses a 

stack of SAR acquisitions and implements an easy combination of a properly chosen set of 

multi-look DInSAR interferograms computed from these data. It allows the generation of 

mean deformation velocity maps and displacement time series. As each interferogram is 

calibrated with respect to a single pixel located in an area that can be assumed stable or at 

least, with known deformation behaviour, this point is referred to as reference SAR pixel 

(Casu et al., 2006). This requires well known natural reflectors or, in addition, the use of 

artificial corner reflectors provide the possibility to detect a known point and thereby 

minimise de-correlation. We designed and installed new specific artificial reflectors with an 

additional arm attached to accommodate a GPS antenna for periodic surveys. To facilitate the 

data cross validation and, in the perspective of an operational use, to be able to consistently 

switch from one system to the other, we also installed reference targets for TLS measurements. 

In this way, the artificial corner reflectors can be used as reference stations for all three 

monitoring technologies. 


Due to seasonal limitations (snow cover in winter causes de-correlation) and internal management 

decisions of the SAR image provider (i.e. the Italian Space Agency), only 12 images 

were available for the Italian test site, and 22 for the Swiss study area in an acquisition period 

of 3 years (2011-2013). According to Colesanti et al. (2003), accurate processing with the 

SBAS algorithm requires at least 20 consecutive images. Consequently, no advanced 

interpretations can be provided for the Italian test site. 

TLS campaigns were carried out annually using Riegl long range scanners (LPM321 and 

VZ-6000) and artificial reflectors for improved merging of point clouds acquired from 

different scan positions. The scans were performed with a resolution of < 10cm. The acquired 

point clouds were filtered in order to homogenize the spatial resolution and to remove 

outliers. Following Chen and Medioni (1991) the iterative closest point (ICP) algorithm was 

applied to match unchanged terrain parts in the scan and to achieve an optimal relative 

referencing of the multi-temporal scans. The set of relative registered point clouds was then 

transformed into global coordinates. Finally, the point clouds were transformed into grid 

based digital elevation models (DEM) with 20cm resolution. Based on similar surface 

patterns, individual patches of multi-temporal DEMs were correlated to obtain horizontal 

displacement information. 

Seven differential GNSS campaigns with double frequency sensors (Leica Viva GS10 and 

Leica GS530) were carried out in the summer months of 2012 - 2014. In total, 18 and 14 

fixed points were measured with rapid static surveys for at least 30 minutes and a sampling 

rate of 5 seconds in the Italian and Swiss study areas, respectively. Every measurement 

campaign utilised a mobile rover and two GPS base stations. The post-processing of the 

collected GNSS data was performed using the Leica Geo Office software by Leica Geosystems. 

In order to refer the local net to the global reference system the base stations were connected 

to permanent reference stations belonging to the geodetic network of Switzerland (SWIPOS) 

and South Tyrol (STPOS). The results between two measurement campaigns were referred 

against each other by correcting the point coordinates using the local reference stations. 

The accuracy was defined by evaluating the deviation of the corrected reference station 

coordinates between two measurement campaigns. 


Figure 2: Test sites monitoring line-up; a) test site “Steinschlagspitze Schnals Valley, South Tyrol; b) test site “Schafberg” Pontresina, 

Switzerland; c) new type of artificial corner reflector with GPS-antenna and TLS reflector at the base. 

DISCUSSION 

Data acquisition by different monitoring techniques of the same process within the same area 

in general leads to dissimilar information. The evaluation of the data obtained by these independent 

basic principles, techniques and calculation procedures, e.g. remote data acquisition 

compared to punctual on site data measurements, does not allow a direct comparison of 

the single results. The concept of real data integration to obtain a unique and coherent final 

product not only requires a straightforward comparison between single data sets or a mutual 

validation of them, but also the assimilation of the various input information. 

In all cases, it is necessary to have carefully georeferenced data sets that are attributable to the 

same spatial and temporal dimension. This requires extensive calculations and projections of 

the different vectors on a Cartesian coordinate system where in some cases the results are not 

distinct, e.g. when the movements are perpendicular to the line of sight of the satellite. This 

problem may be solved by analysing longer time-periods and by utilising larger data sets. 

With the described set-up, detailed information on the movement of the geomorphic features 

was acquired for both test sites. In general, the methods provide similar results and assess 

comparable movement trends. However, at some points the magnitude of displacements 

differs remarkably, or even shows a reverse trend. This is related to the fact that GPS and 

DInSAR analysis of the artificial corner reflectors do only measure single points and not 

changes of larger areas like TLS or DInSAR based on extensive natural scatterers. 


Figure 3: Comparison of horizontal TLS creep rates and DInSAR displacements. Different sources for data gaps are evident: red areas – 

shadow effects of TLS; blue areas – de-correlation of SAR images due to fast creep processes. 


CONCLUSIONS 

The integration of the data collected with the three applied monitoring techniques allows to 

draw the following conclusions: 

– Punctual information obtained from accurate selective data such as recorded with DGPS 

measurements or PS interferometry, represents a good reference point for comparison 

between different measurement techniques and principles. 

– Spatial information, collected by TLS- and SAR-images, is less accurate with respect to 

single points, but better represents the overall trends of the investigated mass movement. 

– Even with a limited number of SAR images, it was possible to obtain comparable information 

to the other planar monitoring techniques using the SBAS-algorithm. 

The suitability of the various methods for applications in the field of hazard management can 

be determined on the basis of our experiences with the different technologies: 

– DGPS only provides sparse information but accurate punctual results on the deformation 

of a large sliding mass. Very fast motions or non-uniform acceleration can be monitored 

efficiently. This technique requires a greater effort on site because of the installation of 

measuring devices and due to the longer data acquisition times. It is economically 

applicable only in specific cases for local civil protection matters. DGPS is thereby well 

suited for automatic alerting systems, however, the equipment remains expensive. 

– The TLS-technique allows the monitoring of larger areas in short periods of time. 

The use depends on two major factors: 

(i) Finding a suitable position for the measuring instrument and the reflectors (view angle 

and distance), from which it is possible to record the entire study area. 

(ii) The accessibility of the monitoring site and the weather conditions, because it works 

only with good visibility. 

This technology may be strengthened using new sensors and the improvement of the 

automatic data processing. An application for automatic alerting systems would be possible 

and some research activities in that direction have already been reported (Canli et al., 

2015). 

– DInSAR allows for the monitoring of larger areas that must be well pre-selected matching 

the geographical situation of the slope with the geometric conditions of the satellites. 

However, SAR image acquisition is only being carried out at relatively long time intervals 

and assessing surface deformation is a lengthy process. 

– The shortened intervals between two data acquisitions by satellites of the last generation 

like Cosmo Sky Med ® and Terra SAR X ® are already not suitable for monitoring of 

processes that require rapid data acquisition and evaluation and also for automatic alerting 

systems. 

At present, this technology is only suitable for science and land-use planning, as there is no 

guarantee for the delivery of SAR-images and coherent data sets. To use the technology for 

continuous monitoring and for civil defence affairs, it is necessary that this technology moves 


from a purely strategic and military application to a civil use. It should be ensured that data 

acquisitions are carried out following a user-defined program and that data processing and 

interpretation can be done continuously and – ideally - by technicians with local experience. 

In order to exploit the full potential of this technology, the following should be enhanced: 

– Establishment of research and evaluation institutions near the monitoring areas; 

– Reduction of revisiting times and costs for the data analysis 

– Development of new algorithms that allow good data processing and interpretation of 

a small number of SAR images. 


The described research was co-funded by the Interreg IV Italia-Svizzera Program, European 

Regional Development Fund.A9 and by the IFFI-programme of the Autonomous Province 

of Bozen, Italy. 

REFERENCES 

- Berardino P., Fornaro G., Lanari R., Sansosti E. (2002): A new algorithm for surface 

deformation monitoring based on small baseline differential SAR interferograms, in: IEEE 

Transactions on Geoscience and Remote Sensing, v. 40, no. 11, p. 2375–2383 

- Bollmann E., Rieg L., Spross M., Sailer R., Bucher K., Maukisch M., Monreal M., Zischg A., 

Mair V., Lang K., Stötter J. (2012): Blockgletscherkataster Südtirol: Erstellung und Analyse, 

in: Innsbrucker Geographische Studien Bd. 39: Permafrost in Südtirol, p. 147-171 

- Calò F., Ardizzone F., Castaldo R., Lollino P., Tizzani P., Guzzetti F., Lanari R., Angeli M.-G., 

Pontoni F., and Manunta M. (2014): Enhanced landslide investigations through advanced 

DInSAR techniques: The Ivancich case study, Assisi, Italy: Remote Sensing of Environment, 

v. 142, p. 69–82 

- Canli E., Thiebes B., Höfle B., Glade T. (2015b): Permanent 3D laser scanning system for 

alpine hillslope instabilities, in: 6th International Conference on Debris-Flow Hazards 

Mitigation. Presented at the 6th International Conference on Debris-Flow Hazards Mitigation, 

Tokyo, Japan 

- Casu F., Manzo M., Lanari R. (2006): A quantitative assessment of the SBAS algorithm 

performance for surface deformation retrival from DInSAR data, in: Remot Sensing of 

Environment 102, p.195-210 

- Colesanti C., Ferretti A., Prati C., Rocca F. (2003): Multi-Image Satellite SAR Interferometry: 

State of the Art and Future Trends, in: Proc Radar Conference, Adelaide. 2003. p. 239-244 

- Liu L., Millar C.I., Westfall R.D., Zebker H.A., and others (2012): Surface motion of active 

rock glaciers in the Sierra Nevada, California, USA: inventory and a case study using InSAR: 

The Cryosphere Discuss, v. 7, p. 343–371 

- Mair V., Zischg A., Krainer K., Stötter J., Zilger J., Belitz K., Schenk A., Damm B., Kleindienst 

H., Bucher K., Lang K., Tagnin S., Munari M. (2008): PROALP – Rilevamento e 

monitoraggio dei fenomeni permafrost. Esperienze della Provincia di Bolzano, in: AINEVA, 

64, p. 50-59 


- Papke J., Strozzi T., Wiesmann A., Wegmueller U., Tate N.J. (2012): Rock glacier monitoring 

with spaceborne SAR in Graechen, Valais, Switzerland, in: Geoscience and Remote Sensing 

Symposium (IGARSS), 2012 IEEE International, IEEE, p. 3911–3914. 

- Schlögel R., Doubre C., Malet J.-P., Masson F. (2015): Landslide deformation monitoring 

with ALOS/PALSAR imagery: A D-InSAR geomorphological interpretation method, in: 

Geomorphology, v. 231, p. 314–330. 



Nationwide assessment of the climate sensitivity of 

natural hazard processes in Switzerland A fuzzy logic 

approach 

Peter Mani, lic. phil. nat. 1 ; Stéphane Losey, Dipl. Ing. forest ETH 2 

ABSTRACT 

Information about the changes that may be expected with respect to natural hazard processes 

is crucial to the development of climate change adaptation strategies. To assess the impacts of 

climate change on the magnitude and frequency of natural hazard processes, a methodology 

was developed, which is based on existing spatial data, literature analyses and expert knowledge 

and uses fuzzy logic to assess the sensitivity of natural hazard processes to changes in 

the climate. Evaluation rules were defined for an intermediate and an extreme climate 

scenario in the case of gravitational processes, avalanches, debris flows and torrent processes. 

The results of the study present a differentiated view of the changes to be expected as a 

function of the regions and processes involved. The greatest changes may be expected in the 

Alps. Permafrost degradation and changes in water availability result in an increase in the 

frequency and magnitude of rock fall. The frequency of small and medium-sized torrent 

processes will increase significantly for the same reason. 

KEYWORDS 

natural hazards; climate change; GIS; fuzzy logic 

INTRODUCTION 

Changes in precipitation and temperature arising from climate change can influence not only 

the magnitude and frequency of natural hazard processes but also their spatial distribution 

(Schweiz. Eidgenossenschaft 2012). However, the impacts vary according to the features of 

the natural landscape. Hence, the knowledge of where such changes may be expected is a 

central factor in the development of climate change adaptation strategies. There are many 

studies which analyze the impact of climate change on natural hazards whereof many refer to 

the Alpine region (c.f. Stoffel et al. 2013, Stoffel et al. 2014, Gobiet et al. 2014, Pavlova et al. 

2014). But there are few that relate to lower areas in Switzerland. A nationwide, spatially 

differentiated overview about the climate change impact on natural hazards is still missing. 

In the context of a project on the climate sensitivity of natural hazard processes, which was 

commissioned by the Federal Office for the Environment (FOEN), a methodology was 

developed that enables the nationwide assessment of the sensitivity of natural hazard process- 

1 geo7 AG, geowissenschaftliches Büro, Bern, SWITZERLAND, peter.mani@geo7.ch 

2 Bundesamt für Umwelt BAFU, Bern 



es to changes in the climate in Switzerland. The focus of this paper will be on the methodological 

approach. In the second part a brief overview of the results is presented. 

METHODOLOGICAL APPROACH 

The links between the climate and process systems on the Earth’s surface are complex. The 

reasons for this include nonlinearities, retardation and hysteresis effects (Knight & Harrison 

2013). As a result, analyzing the changes in natural hazard processes due to climate change 

using numerical models is very complicated and, due to the data situation, only possible for 

small investigation areas and individual processes. Hence an alternative methodological 

approach, which is described below, was selected for the project described here. 

The availability of spatially resolved, comprehensive and homogeneous base data on hazard 

processes and their influencing parameters is a precondition for such a nationwide analysis. 

Such data are available in the form of the intermediate and final results of process simulations, 

which were carried out for all of Switzerland in the context of the development of the 

protective forest index maps (SilvaprotectCH) (BAFU 2008). These data cover block and rock 

falls, snow avalanches, shallow landslides and torrent processes (debris flows, bed load 

transport). However, in some instances they are encumbered by significant inaccuracies, 

which must be taken into account in their further use. This is possible using fuzzy logic that 

is based on Zadhe’s fuzzy set theory (1965). Unlike the boolean logic, where a value can only 

belong to one class, in the fuzzy logic the classes, the so-called linguistic variables, can overlap 

to incorporate the fuzziness of the data. The variables are combined logically, using fuzzy 

rules. These rules can be derived from literature, data analysis or expert knowledge. Fuzzy 

logic is applied in many fields, amongst others in expert systems. Fuzzy logic based expert 

systems are also used for the analysis of natural hazard processes. Several examples are 

related to landslides (Brauner & Ganahl 1999, Regmi et al. 2010, Thiery et al. 2014) and 

Zischg et al. (2005) used it for assessing the risk posed to transport routes by wet snow 

avalanches. 

For the evaluation of the climate sensitivity of natural hazard processes, the data were linked 

with the help of fuzzy rules. The rules were defined on the basis of literature research and 

expert hearings. Based on the concept of disposition (BUWAL 1998), evaluation rules are 

defined for each process for the basic disposition, the variable disposition and the triggering 

process. These rules were applied on the basis of grid data with a resolution of 10 meters for 

the detailed analyses over the whole of Switzerland and on the level of the 22,400 catchment 

areas from the division of Switzerland into catchment areas (BAFU, year not specified) for the 

spatial aggregation of the final results (Figure 1). This required the efficient implementation 

of the fuzzy analyses in the GIS environment. This was achieved by integrating Matlab‘s fuzzy 

logic tools into ArcGIS as geoprocessing tools. The calculations were then implemented as 

Python Script. This means that new calculations can be carried out efficiently if better base 

data or new insights relating to the evaluation become available. 


CLIMATE SCENARIOS 

Two climate scenarios, one medium and one extreme, were deduced for the sensitivity 

analysis from the Swiss Climate Change Scenarios CH2011 (2011). The ensembles from 

10 model chains for 188 temperature stations and 565 precipitation stations were used. 

The A1B emissions scenario (Nakicenovic & Swart 2000) provided the basis for these 

scenarios. The median values for the time horizon 2060 were used for the medium scenario, 

and the 97.5 % percentiles for the time horizon 2085 were used for the extreme scenario. 

The scenarios distinguish three regions. The scenarios for seasonal temperatures and seasonal 

precipitation are presented in Table 1 and Table 2 respectively. In addition, scenarios for heavy 

precipitation were defined on the basis of the extended CH2011 scenarios (CH2011, under 

review), and scenarios for the snow/rain distribution for different elevation ranges were 

derived from transient simulations. 

Table 1: Decisive seasonal temperature scenarios for medium and extreme scenario 

Season Central Plateau, Jura Pre-Alps, Alps South Switzerland 

medium 

[Δ°C] 

extreme 

[Δ°C] 

medium 

[Δ°C] 

extreme 

[Δ°C] 

medium 

[Δ°C] 

extreme 

[Δ°C] 

DJF +2 +3 +2 +3 +2 +4 

MAM +2 +3 +2 +3 +2 +4 

JJA +2 +5 +3 +5 +3 +5 

SON +2 +4 +2 +4 +2 +4 

Table 2: Decisive seasonal precipitation scenarios for the medium and extreme scenario 

Saison Central Plateau, Jura Pre-Alps, Alps South Switzerland 

medium 

[Δ%] 

extreme 

[Δ%] 

medium 

[Δ%] 

extreme 

[Δ%] 

medium 

[Δ%] 

extreme 

[Δ%] 

DJF +5 +25 0 +20 +10 +10 

MAM +5 +20 +5 +10 0 0 

JJA -10 -5 -10 -5 -15 -15 

SON -5 +25 0 +20 0 0 


IMPLEMENTATION OF THE METHODOLOGY: EXAMPLE ROCK FALL 

The implementation of the methodological approach is described in greater detail below using 

the example of the rock fall process. According to the National Platform of Natural Hazards 

(PLANAT) we define rock fall as a large volume of rock (> 100 m³) detached en masse. 

The hypothesis that changes in the permafrost and in the availability of water are the main 

factors was adopted as a basis for the rock fall sensitivity analysis. Thawing permafrost 

results in a reduction in the stability of rock faces and increases the disposition for rock fall 

(c.f. Gruber & Haeberli 2007). Changes in water availability influence the water pressure in 

fissures, which is an important triggering factor (c.f. Gruner 2008, Schneuwly & Stoffel 2008). 

Both factors can alter the magnitude and frequency of rock falls. In addition, new rock fall 

areas can evolve from glacier retreat, such as the Eiger rock slide (Oppikofer et al., 2008) 

(this point is not further developed). The structural characteristics of the rock body are also 

significant. However, nationwide data are lacking on this parameter, thus it could not be 

included in the evaluation. 

The assessment of the permafrost degradation incorporated, first, the permafrost index from 

the Permafrost Index Map of Switzerland (BAFU 2005), and, second, the exposition. The 

permafrost index provides qualitative data on the areal distribution, thickness and temperature 

of permafrost bodies. The higher the index, the greater the likelihood that permafrost 

is present and the thicker and colder the permafrost bodies will be. The exposition is used as 

an indicator of the influence of short-wave radiation or sensitivity to temperature increase. 

South-facing rock faces react less sensitively than north-facing ones as the radiation towards 

the south already gives rise to higher energy input today and this will not change significantly 

in the future. However, the change in temperature will have a stronger impact in north-facing 

rock faces (Salzmann et al. 2007). 

The elevation range and size of the catchment area above the area under consideration were 

taken into account for the evaluation of the water availability. For the calculation of the 

catchment area a multiple flow direction approach has been applied to account for both, 

surface runoff and subsurface flow (Quinn et al. 1991). If water can only flow from a small 

area above a rock face, sensitivity is lower than it would be if it can flow from a larger area. 

Elevation range was used as an indicator for the change in the distribution of the precipitation 

in snow and rain and the seasonal change in the total water availability (Table 3). Due 

to the greater winter precipitation and shift in snow melt in locations above 1,200 m asl, 

according to the extreme scenario, increased water inflow may be expected in the Central 

Plateau and Jura. The decline in the snow melt is overcompensated by the increased 

precipitation at this altitude in spring. The water inflow is unchanged or declines in summer. 

It increases strongly at all altitudes in autumn. The situation is largely comparable in the Alps 

and Pre-Alps. In spring alone, the water availability does not increase below 2,500 m asl. 

In south Switzerland, a – possibly strong – increase in water inflow may be expected at lower 


and medium altitudes. The water inflow declines in spring (exception altitudes > 2,500 m asl) 

and summer and remains unchanged in autumn. 

Table 3: Combination of the seasonal changes in precipitation and snow melt for the extreme scenario (R: rain, SM: snow melt, -: 

decrease, +/-: unchanged, +: increase, ++: strong increase) 

Altitude DJF MAM JJA SON 

Central Plateau, 

Jura 

R SM Total 

R SM Total 

R SM Total 

R SM Total 

< 1200 ++ +/- ++ ++ - + +/- +/- +/- ++ +/- ++ 

> 1200 + + ++ ++ - + +/- +/- +/- ++ +/- ++ 

Pre-Alps, Alps 

< 1200 ++ +/- ++ + - +/- +/- +/- +/- ++ +/- ++ 

1200 - 2500 + + ++ + - +/- +/- +/- +/- ++ +/- ++ 

> 2500 +/- +/- +/- +/- ++ ++ +/- - - ++ +/- ++ 

South Switzerland 

< 1200 + +/- + +/- - - - +/- - +/- +/- +/- 

1200 - 2500 + + ++ +/- - - - +/- - +/- +/- +/- 

> 2500 +/- +/- +/- +/- + + - - - +/- +/- +/- 

The permafrost degradation and water inflow were evaluated using fuzzy rules. The sensitivity 

of the rock fall process in relation to climate change was then deduced from these parameters. 

To this end, statements were made on changes in magnitude and frequency. Fuzzy rules 

were also applied here. Up to this point, the evaluation of sensitivity was still carried out 

on the level of the individual grid cells without taking into account whether rock fall process 

areas are present or not. In the concluding step, the grid data were aggregated to the 

catchment areas from the subdivision of Switzerland into catchment areas (BAFU, year not 

specified). This involved, first, sampling based on the process areas (Fig. 1 a). The proportion 

of process areas per square kilometer was calculated from this. The sensitivity level was then 

determined from the grid for each sample point. The 25 %, 50 % and 75 % percentiles of the 

sensitivity evaluation were calculated on the basis of these data (Fig. 1 b). The sensitivity was 

deduced from the percentiles and proportion of process area using fuzzy rules (Fig. 1 c and d). 

The maps in Figure 2 show that, according to the extreme scenario, an increase or strong 

increase in frequency may be expected for most of Switzerland at grid level. In the case of 

magnitude, a reduction or no change may be expected in the Central Plateau and Jura, and in 

the Pre-Alps and lower altitudes of the Alps. However, a strong increase may be expected at 

higher altitudes in the Alps. The increase in frequency may be explained by the greater water 

inflow in autumn, winter and spring in the Central Plateau and Jura. Due to the increase in 


Figure 1: Spatial aggregation concept at catchment area level 

frequency and limited rock material in areas without permafrost magnitude declines 

according to the magnitude-frequency relationship (c.f. Hungr et al. 1999). However, this 

relationship applies only with respect to a larger area. In south Switzerland, a reduction 

in frequency may be expected at lower altitudes while no change may be expected in the 

situation at higher altitudes. In the case of magnitude, with the exception of areas at the 

highest altitudes, an unchanged situation may be expected. The decline in precipitation is 

the main factor here. 

According to the aggregation to the level of the catchment areas, the changes are limited to 

the higher locations in the Alps and scattered areas in the Pre-Alps. The density of rockfall 

areas is low in other areas, which explains why no major changes may be expected here. 

A strong increase in rock fall frequency may be expected at the higher attitudes across the 

Bernese, Valais and Grison Alps, and also in the Uri and Glarus Alps. The same applies to 

magnitude. The reason for this is the increasing instability of rock faces due to permafrost 

degradation and the increased water availability. A decrease may be expected in some areas 

of the Pre-Alps as there no permafrost occurs. 


Figure 2: Results for rock fall at grid level (above) and catchment area level (below) 

RESULTS 

The results for all of the evaluated processes are presented in a highly generalized form in 

Figure 3. The overview presents a differentiated picture of the changes to be expected 

according to the region and processes involved. 

In the case of slope processes in the Jura and the Central Plateau no major changes may be 

expected with the exception of an increase at shallow landslides on steep slopes. With regard 

to avalanches in the Pre-Alps the reduction in snowfall in this region will lead to a decline in 

frequency and magnitude but also to unchanged conditions in some cases. The frequency 

and magnitude of block fall in the Pre-Alps will decrease or remain unchanged as a decline 

in freeze-thaw days causes less weathering and, therefore, less material is available for such 

events. An increase in shallow landslides on steep slopes may be expected as precipitation 

falls increasingly as rain, particularly in the winter, when evapotranspiration is low. 

The biggest changes may be expected in the Alps. According to both scenarios, permafrost 

degradation and the increase in water availability will prompt an increase in the frequency 

and magnitude of rock falls. Moreover, according to the extreme scenario in particular, an 

increase in avalanche activity may be expected at high elevation where precipitation 

continues to fall as snow while precipitation in winter will increase. In the case of hillslope 


Process Medium scenario Extreme scenario 

Block fall 

Frequency Magnitude Frequency Magnitude 

Rock fall 

Dry snow 

avalanche 

Wet snow 

avalanche 

Shallow 

landslides 

from steep 

slopes 

Debris flows 

from poorly 

consolidate 

d talus 

Torrent, 

small events 

Torrent 

large events 

Torrent very 

lange events 

Legend 

Figure 3: Results 3: Results of the of sensitivity the sensitivity analysis for analysis both scenarios for and both the scenarios different processes and the different processes 

debris flows from poorly consolidated talus, an increase in frequency may be expected due 

to the thawing of the permafrost. The results in south Switzerland are inconsistent. 


A decrease or no change may be expected for the majority of processes. Shallow landslides 

from steep slopes and wet snow avalanches (extreme scenario only) are the exceptions here. 

In both cases an increase of frequency may be expected. 

For the sensitivity assessment of torrent processes sediment yield and triggering events are 

taken into account. Changes in the sediment yield were derived from the assessment of the 

slope processes, in which only areas relevant for the sediment yield were considered. The 

assessment of the sensitivity of triggering events was derived from the climate scenarios. 

The sensitivity of torrent processes was assessed for three magnitudes. 

For the medium scenario in all regions an increase or strong increase of small events may 

be expected. The reason for this is the increase in the triggering events. In the Jura and the 

Central Plateau, this results in a decrease or unchanged conditions in large or very large 

events, as available bed load is limited. In the Pre-Alps, where many ancient deposits exist 

and in the Alps, where degradation of permafrost and glacier retreat cause an increase of sediment 

yield, an increase or a strong increase for large or very large events may be expected. 

In the extreme scenario in almost all regions a strong increase in triggering events must be 

expected. This results in an increase or a strong increase in small events. The limited 

availability of loos material in the Central Plateau leads to a decrease of large and very large 

torrent events. In the Jura increased sediment yield from shallow landslides leads partly to an 

increase of large events. With regard to very large events, this increase of shallow landslides 

cannot compensate the increase in frequency. In the Pre-Alps and Alps with its large 

additional sediment yield an increase or strong increase of large und very large events may 

be expected, despite the increase in frequency. In south Switzerland the situation is heterogeneous. 

With regard to large events an increase may be expected, due to several ancient 

deposits in higher regions. But very large events will decrease. 

DISCUSSION AND CONCLUSIONS 

In the context of the above-described project it was possible to develop a methodology which 

enables the efficient assessment of the sensitivity of natural hazard process related to climate 

change. Recent events in Switzerland (e.g. August 2005, October 2011) indicate an acceleration 

of processes, especially in the periglacial area, but also in the Pre-Alps. This is in correspondence 

with the results from this study. The new periglacial hazard index map, which was 

elaborated based on detailed input data for the Bernese Oberland (Tobler et al. in rev.) also 

coincide in most part. Effects of climate change as described in Stoffel & Huggel (2012) or in 

a detailed study on rock fall (Perret et al. 2006) are reproduced in the here presented results. 

But there are also some limits which must be considered when using the result from this 

study. Thus, the study is limited to the analysis of the disposition and the triggering events. 

The reach of hazard processes is not included. Furthermore, new process chains that could 

become relevant in the context of climate change (e.g. flood waves, triggered by a rock 

avalanche falling into a new glacier lake) are not considered. However, the results are 


valuable to determine areas for more detailed studies, they can be used for the planning of 

monitoring concepts and they show where climate change effects must be taken into account 

when hazard maps are revised. For this the data produced in this study will be made available 

to the cantonal authorities. 

BIBLIOGRAPHY 

- BAFU (2005). Hinweiskarte der potenziellen Permafrostverbreitung in der Schweiz. 

- BAFU (2008). SilvaProtect-CH – Phase I. Projektdokumentation. 

- BAFU (o. Jahr). Einzugsgebietsgliederung Schweiz EZGG-CH. 

- Brauner M. Ganahl E. (1999). GIS-basiertes Expertensystem zur Risikobewertung 

von Hanggleitungen in Wildbacheinzugsgebieten. Österreichische Zeitschrift für 

Vermessung & Geoinformation. 87. Jg. Heft 2+3: 93-100. 

- BUWAL (1998). Begriffsdefinition zu den Themen: Geomorphologie, Naturgefahren, 

Forstwesen, Sicherheit, Risiko. Arbeitsbericht. 

- CH2011 (2011). Swiss Climate Change Scenarios CH2011. Published by C2SM, MeteoSwiss, 

ETH, NCCR Climate, and OcCC. Zurich, Switzerland. 

- CH2011 (in rev). Projections of Extreme Precipitation in Switzerland. CH2011 Extension 

Series. 

- Gobiet A., Kotlarski S., Beniston M., Heinirch G., Rajczak J., Stoffel M. (2014). 21st century 

climate change in the European Alpts – A review. Sci. Total Environ., 493: 1138–1151. 

- Gruber S., Haeberli W. (2007) Permafrost in steep bedrock slopes and its temperaturerelated 

destabilization following climate change. J. Geophys. Res. 112: F02S18. 

- Gruner U. (2008). Klimatische und meteorologische Einflüsse auf Sturzprozesse. 

Interpraevent, 147–158. 

- Hungr O., Evans S.G., Hazzard J. (1999). Magnitude and frequency of rock falls and rock 

slides along the main transportation corridors of southwestern British Columbia. Can. 

Geotech. J., 36(2), 224-238. 

- Jasper K., Harrison S. (2013). The impacts of climate change on terrestrial Earth surface 

systems. Nature Climate Change, 3, 24-29. 

- Oppikofer T., Jaboyedoff M., and Keusen H.-R. (2008). Collapse at the eastern Eiger flank 

in the Swiss Alps, Nat. Geosci., 1, 531–535. 

- Pavlova I., Jomelli V., Brunstein D., Grancher D., Martin E., Déqué M. (2014). Debris flow 

activity related to recent climate conditions in the French Alps: A regional investigation. 

Geomorphology, 219: 248–259. 

- Perret S., Stoffel M., Kienholz, H. (2006). Spatial and temporal rockfall activity in a forest 

stand in the Swiss Prealps—a dendrogeomorphological case study. Geomorphology, 74(1), 

219-231. 

- Quinn P.F., Beven K.J., Chevallier P., Plancon O. (1991). The prediction of hillslope flow 

paths for distributed hydrological modelling using digital terrain models. Hydrol. Process, 

5:59-79. 


- Regmi N.R., Giardino J.R., Vitek J.D. (2010). Assessing susceptibility to landslides: 

Using models to understand observed changes in slopes. Geomorphology 122:25-38. 

- Salzmann N., Nötzli J., Hauck C., Gruber S., Hoelzle M., Haeberli W. (2007). Ground surface 

temperature scenarios in complex high-mountain topography based on regional climate 

model results. J. Geophys. Res., 112, F02S12, doi: 10.1029/2006JF000527. 

- Schneuwly D.M., Stoffel M. (2008). Tree-ring based reconstruction of the seasonal timing, 

major events and origin of rockfall on a case-study slope in the Swiss Alps, Nat. Hazards Earth 

Syst. Sci., 8, 203–211. 

- Stoffel M., Huggel, C. (2012). Effects of climate change on mass movements in mountain 

environments. Progress in Physical Geography, 36(3), 421-439. 

- Stoffel M., Mendlik T., Schneuwly-Bollschweiler M., Gobiet A. (2013). Possible impacts of 

climate change on debris flow activity in the Swiss Alps. Clim. Chang. 122(1-2), 141-155. 

- Stoffel M., Tiranti D., Huggel C. (2014). Climate change impacts on mass movements – 

Case studies from the European Alps. Sci. Total Environ., 493: 1255–1266. 

- Schweizerische Eidgenossenschaft (2012). Anpassung an den Klimawandel in der Schweiz 

– Ziele, Herausforderungen und Handlungsfelder. Erster Teil der Strategie des Bundesrates 

vom 2. März 2012. 

- Thiery Y., Maquaire O., Fressard M. (2014). Application of expert rules in indirect approaches 

for landslide susceptibility assessment. Landslides, 11(3), 411-424. 

- Tobler D., Mani P., RinerR., Liener S., Haehlen N., Bender-Gàl R., Graf K., Raetzo H. 

(in rev.). Gefahrenhinweiskarte Periglazial. Ein Grundlageninstrument für das präventive 

Gefahrenmanagement. Interpaevent 2016. 

- Zadhe L. A. (1965). Fuzzy Set. Information and Control 8: 338-353. 

- Zischg A., Fuchs S., Keiler M., Meissl G. (2005). Modelling the system behaviour of wet 

snow avalanches using an expert system approach for risk management on high alpine 

traffic roads. Nat. Hazards Earth Syst. Sci. 5: 821-832. 



Defining sample size and strategy for 

dendrogeomorphic rockfall reconstructions 

Pauline Morel 1 ; Daniel Trappmann 2 ; Christophe Corona 2 ; Markus Stoffel 2 

ABSTRACT 

Optimized sampling strategies have been recently proposed for dendrogeomorphic reconstructions 

of mass movements with a large spatial footprint, such as landslides, snow 

avalanches and debris flows. Such guidelines have been missing for rockfalls and cannot be 

transposed owing to the sporadic nature of this process and the occurrence of individual rocks 

and boulders. Based on a dataset of 314 European larch (Larix deciduaMill.) trees (64 trees/ 

ha), growing on an active rockfall slope, this study bridges this gap and proposes an optimized 

sampling strategy for the spatial and temporal reconstruction of rockfall activity. Spatially, our 

results demonstrate that the sampling of only 6 representative trees/ha can be sufficient to 

yield a reasonable mapping of the spatial distribution of rockfall frequencies on a slope, 

especially if the oldest and most heavily affected individuals are included in the analysis. 

Temporally, we demonstrate that at least 40 trees/ha are needed to obtain reliable rockfall 

chronologies. 

KEYWORDS 

Rockfall, Dendrogeomorphology, Simulation, Frequency, Methodology 

INTRODUCTION 

Rockfall represents one of the most frequent natural mass movement processes in mountainous 

areas and can be defined as the free falling, bouncing, or rolling of rocks downslope that 

typically originate from cliffs or rockwalls (Varnes, 1978; Berger et al., 2002). On forested 

slopes, each rock impact on trees dissipates kinetic energy and may change the rock's 

trajectory and velocity, thus reducing runout distances as compared to non-forested slopes 

(Jahn, 1988; Dorren et al., 2005, 2007). Impacts also leave characteristic scars on tree trunks 

and growth disturbances (GD) in tree-ring series that have been proven to be reliable, 

accurate and precise indicators to reconstruct past rockfall activity through dendrogeomorphic 

analysis (Alestalo, 1971; Stahle et al., 2003; Stoffel et al., 2005a, 2013; Stoffel and 

Bollschweiler, 2010; Stoffel and Corona, 2014). While in the earliest studies a limited number 

of 25–30 samples was used for rockfall reconstructions (Gsteiger, 1989; Schweingruber, 

1996), laterwork generally was based on much larger numbers of samples (135 - 283 trees; 

e.g., Stoffel et al., 2005b; Moya et al., 2010; Šilhán et al., 2013). A clear guideline regarding 

the sample size needed to obtain reliable results still does not exist. 

1 DendroLab, Bern, SWITZERLAND, pauline.morel@ymail.com;pauline.morel@geo.unibe.ch 

2 Dendrolab.ch, Institute of Geological Sciences, University of Berne, Baltzerstrasse 1+3, CH-3012 Berne, Switzerland 



A suite of recent studies concluded that an appropriate sampling design and size is a fundamental 

requirement to improve the reliability of dendrogeomorphic reconstructions (Schneuwly-Bollschweiler 

et al., 2013; Trappmann et al., 2013). In the case of mass movements with 

a large spatial footprint, such as snow avalanches (Corona et al., 2012), landslides (Corona et 

al., 2014), and debris flows (Schneuwly-Bollschweiler et al., 2013), it has been demonstrated 

that a definition of sample size thresholds is possible and that such values permit assessment 

of realistic event frequencies with optimized cost–benefit ratios. In contrast, rockfall does not 

typically leave a clear spatial footprint as it damages a limited number of individual trees 

along its trajectory (Stoffel and Perret, 2006;Moya et al., 2010). Therefore, the thresholds 

established for snow avalanches, landslides, and debris flows cannot be applied in this case as 

different thresholds and approaches need to be defined to obtain more reliable rockfall 

reconstructions and better input data for hazard zoning. In order to fill this gap, this study 

aims to determine optimal sampling sizes and strategy for dendrogeomorphic rockfall studies. 

Based on an unusually large dataset of rockfall induced GD in trees growing on a slope in the 

Swiss Alps, we (i) test results based on different subsets of trees and (ii) characterize the 

optimal spatial configuration of trees to be sampled on the slope using random bootstrap 

extraction of trees from the dataset. The same subsets were then used to (iii) explore the 

effect of sample size and tree selection on the reliability of reconstructed rockfall chronologies. 

Finally, (iv) the random extractions of trees were compared with a stratified sampling 

strategy based on an arbitrary selection of trees so as to propose clear guidelines for the 

selection of optimal trees in terms of tree location, age, number and frequency of GD. 

REGIONAL SETTINGS 

Located in the Saas valley (Valais, Switzerland, 46° 05′41 N; 7° 57′17 E; Fig. 1) between 

1670-1800 m asl., the investigated slope (5 ha) has a northeastern exposure and a slope 

ranging from 14-49°. The rockfall source area is formed by an active rock glacier at >2570m 

asl. at the lower permafrost boundary as well as by subvertical rock faces downslope of the 

rock glacier. The rocks deposited in the study area have mean axes lengths of 0.57 m and a 

volume of 0.31 m³. The tree stand at the site is mainly composed of Larix decidua Mill., 

intermixed with young Pinus cembra L. and Picea abies L. Karst. The tree age distribution 

(Fig.1D) shows relatively young individuals (11–40 yr) in the upper part of the slope, thus 

reflecting the influence of former cattle grazing. Rare avalanches may potentially have 

influenced the age distribution on the site as well, especially in the uppermost part where the 

rockfall couloir opens to form a relatively homogeneous talus slope. Based on geomorphic 

mapping and tree morphology at the site, however, rockfall is clearly the only relevant 

process causing damage to the sampled trees. In the lower part of the slope and on its 

northern part, older trees can be found. 


Figure 1: Overview of the study site. (A) Aerial photograph of the study area and location of sampled trees (green dots); (B) overview of 

the forest stand; C)View of the study site from the release areas located at the front of the Plattjen rock glacier towards the study site 

delimited by dotted box; D)and E) geographical location. 

METHODS 

Dendrogeomorphic reconstruction of rockfall activity 

In total, 616 increment cores were taken from 314 L. decidua trees. Systematic sampling was 

favored to a preferential selection of visibly impacted trees as L. decidua is known to heal and 

completely hide scars (Stoffel and Perret, 2006). Increment cores were extracted as close to 

the injury as possible, where the vascular cambium remained intact, providing complete 

tree-ring series and strong signals. In cases where no injury was visible, increment cores were 

extracted from the upslope side of the stem to maximize chances for impact detection at 0.5, 

1 and 1.5 m. More cores were taken from trees with larger diameters (and thicker, more 

structured bark) because more hidden scars can be expected. Trees were sampled every 12m 

along transects across the slope, but we excluded trees growing in the fall line of close 

neighbors. 

Standard dendrogeomorphic procedures were used during tree-ring analysis (Stoffel and 

Bollschweiler, 2008, 2010). Growth disturbances (see Astrade et al. (2012) or Stoffel and 

Corona (2014) for response typologies) were dated with annual precision. Years with rockfall 

activity were determined from injuries present on the increment cores, tangential rows of 

traumatic resin ducts (TRD), the presence of callus tissue, abrupt tree growth suppression or 


elease. Only strong GD (Stoffel and Corona, 2014) were kept for further analyses. Also, 

when more than one GD is noticed within a year, we only kept the strongest GD for the 

analyses. Details on the GD detected and kept for the analyses in the 314 trees used to date 

rockfall impacts can be seen in Table 1. 

Testing the optimal sampling strategy for spatial reconstruction 

We adapted the routine used in Corona et al. (2013a, see article for a detailed description) 

designed for snow avalanches, which allows the computation of random extractions (REs) of 

trees from the dendrogeomorphic dataset. The routine is based on the RE of n subsamples for 

m iterations. The frequency of rockfall activity was computed for each individual tree. 

A reference map (Refmap) was interpolated with the frequency values derived from 

individual trees of the entire dataset using the Inverse Distance Weighting (IDW) method. 

RE was performed with 30, 50, 100, 150, 200, 250, and 300 trees from the complete dataset. 

This extraction was repeated 100 times for each threshold in order to reduce dependence of 

results on sampling location. Interpolated frequency raster maps were automatically 

generated (based on IDW interpolation) for each subsample in the form of 100 “RESubmaps”. 

For each submap the Root Mean Square Error (RMSE) was computed from the reference 

map to quantify discrepancy between the submap and the reference map. 

Maps representing the lowest error (best sampling) and the highest error (worst sampling) for 

each step of random extraction were plotted and investigated in further detail via comparison 

with the Refmap. 

Testing the optimal sampling strategy for temporal reconstructions 

Rockfall time series were derived for the 30-300 tree subsets corresponding to best and worst 

RESubmaps based on the “range corrected impacts” concept (RCI) (see Trappmann et al., 

2013 for a detailed description). The RCI uses impact probability for the forest stand in each 

year to correct the number of recorded tree impacts. The concept thus includes a range of 

uncertainty and a quantitative estimation of events missed by dendrogeomorphic analyses. 

The RCI also permits definition of an adequate sample size, as the approach yields indications 

on the quality and the reliability of the reconstructed rockfall series. 

Stratified sampling strategy based on an arbitrary selection of trees 

Frequency maps and chronologies derived from random extractions (REs) were compared 

with results from the stratified sampling design based on an arbitrary sampling (AS) of trees 

so as to establish rules for future sampling design. The AS is based on the assumption that 

fewer samples are needed in areas with trees suggesting similar frequencies, meaning that on 

a slope segment with identical rockfall frequency, sampling one representative tree could 

hypothetically yield a reliable rockfall frequency for the sector. For this purpose, (i) heterogeneity 

maps were computed using the ArcGis spatial statistics tool “slope” (ESRI, 1998) which 

calculates the maximum rate of value change from one cell to its neighbors on the reference 

rockfall frequency map. Then, (ii) sample size was weighted in several compartments of the 


heterogeneity map according to the degree of heterogeneity (i.e. less trees in homogeneous 

areas). Finally, and for each compartment, (iii) trees were arbitrary selected as a function of 

their age (old vs. young trees) and of the number of GD (severely vs. little damaged trees). 

RESULTS 

Event frequency reconstruction 

The oldest tree sampled had 421 tree rings at sampling height and was present at the site 

since at least AD 1590, whereas the youngest tree reached sampling height only in AD 2000. 

The mean age of the tree population was 60 yr. Based on the analysis of GD in the tree-ring 

series, a total of 372 rockfall impacts could be detected (Table 1), resulting in a rockfall 

chronology spanning the last 106 years. The mean frequency of rockfalls at the level of 

individual trees is 0.031 impacts yr-1. The reference map (Fig.2) reflects the channelizing 

topography and also shows that the highest activity occurs at the outlet of the rockfall couloir 

descending from the rock glacier and the steep rockwall (south-west, and upper part of the 

study site). The northernmost part of the study site exhibits the lowest activity with a mean 

Table1: Overview on growth disturbances used to date rockfall injuries. 

Growth Traumatic resin 

Growth 

Callus tissue Injuries 

disturbances ducts 

suppression 

Growth release 

Number 336 20 6 6 4 

% 90 5 2 2 1 

Figure 2: Reference frequency map (Refmap) of rockfall derived from the 314 sampled trees. The interpolation was performed the using 

inverse distance weighted method. 


Figure 3.1: Best (lowest RMSE, left panel) and worst (highest RMSE, central panel) rockfall frequency maps interpolated for each of the 

100 subsamples of 30 to 150 randomly extracted trees. Maps on the right panel represent the differences in absolute value between 

the best and the worst frequency maps computed for each subsample. 


frequency of 0.0067 impacts yr-1 on individual trees. A downslope decline of rockfall activity 

also becomes apparent and illustrates the breaking effect of the forest stand on rockfall. 

On a temporal scale, the rockfall chronology is reliable after the RCI criterion for the period 

1950-2011, earlier years suffer from low numbers of available trees for reconstruction. 

Testing optimized random sampling strategy for spatial reconstruction 

Figures 3-4 illustrate differences between the reference frequency map (Refmap), computed 

with all sampled trees, and the best (min. mean RMSE) and worst (max. mean RMSE) 

RESubmaps obtained for the different RE subsets. Visual comparison of results shown in Fig.3 

suggests that the best RESubmaps derived from small sample sizes(30-50 trees) properly 

reproduce the W-E gradient in rockfall frequency. Conversely, the worst RESubmaps (30 and 

Figure 3.2: Best (lowest RMSE, left panel) and worst (highest RMSE, central panel) rockfall frequency maps interpolated for each of the 

100 subsamples of 200 to 300 randomly extracted trees. Maps on the right panel represent the differences in absolute value between 

the best and the worst frequency maps computed for each subsample. 


50) lead to significant under- and overestimations of rockfall frequencies in large compartments 

(south westernmost and northeastern parts) of the slope where trees were absent in 

the subset, thereby pointing to clear dependencies between mean RMSE and sampling design, 

i.e. the spatial distribution of trees selected for the interpolation. 

Figure 4 illustrates that the mean RMSE of RESubmaps decreases by >80% with increasing 

sample size, and varies between 0.015±0.01for RESubmap30 to 0.005±0.002 impacts yr–1 

for RESubmap300. Noteworthy, the best RESubmap50 and 100 have a RMSE comparable 

to the average RMSE obtained for RESubmaps100 and 150, respectively. 

Figure 4: Boxplots for the Root Mean Square Error (RMSE) between the reference frequency map computed from 314 trees and 

100 frequency maps computed with 30 to 300 randomly extracted trees. Boxplots show minimum, lower (Q 0.25), median (Q 0.5), 

upper quantile (Q 0.75) and maximum values for each sub-sample. 

Optimized sample size for rockfall chronologies 

The rockfall chronologies of the best and worst RESubmaps were then analyzed with the 

RCI concept. The chronologies obtained show significant differences depending on the 

number of trees used. Table 2 shows that with a sampling size of 30 trees and for 50 trees 

(worst sampling), not even a short part of the chronologies can be considered as reliable 

because more impacts are assumed to be missed in the trees. With increasing sample sizes 

(≥150 trees), RCI chronologies become more stable. According to the considerations mentioned 

above, Table 2 suggests a threshold of at least 150-200 trees to obtain short, yet 

reliable rockfall chronologies. 

Table 2: Time period covered by the rockfall time series determined from the Range Corrected Impact (RCI) for best and worst 

subsamples (30-300 trees). Annotation: NR: Not Reliable 

Subsample 

30 50 100 150 200 250 300 

Best 

Random 

extraction 

sampling 

Worst 

sampling 

NR 1994 1994 1991 1985 1950 1950 

NR NR 1994 1991 1991 1985 1950 


Chronology reliability is further explored in Fig.6 where the proportion of reconstructed event 

years is illustrated for each best and worst sampling of the RESubmaps. Event years are here 

defined as years with at least one rockfall impact detected in the given year of the time series. 

As could be expected, a larger number of sampled trees will lead to a more complete record 

of event years. However, by using a subset of only 30 (10%) trees, it is possible to reconstruct 

already 20% of the event years. Starting with 150 sampled trees, more than 50% (61% and 

74% for the best and worst sampling, respectively) of the event years can be reconstructed. 

When over 250 trees are sampled, the rate of reconstructed events years remains stable as 

well as the confidence interval. 

Figure 6: Percentage of reconstructed events for the worst and best random extractions as a function of sample sizes. 

Random extraction (RE) vs. arbitrary selection (AS) 

The heterogeneity map (Fig.7), reveals three compartments (A, B, C) corresponding to 

increasing levels of heterogeneity. According to our hypothesis that more sampled trees in 

areas with heterogeneous activity will yield better reconstructions, weights of 0.1, 0.3, and 

0.6 were attributed to each compartment, respectively. In a first test, 10% of trees were 

selected from homogeneous compartment C, 30% from the transition compartment B and 

60% from the heterogeneous compartment A for all subsamples from 30 to 300 trees. In a 

second test, we inverted the weights of the compartments (A: 0.6; B: 0.3; C: 0.1) to test the 

influence of stratified sampling. In each compartment, trees were again arbitrarily selected 

according to their age and to the number of visible scars. In total, eight different datasets were 


finally produced (old vs. young, severely vs. lightly injured trees for both preferential 

sampling in areas with homogeneous and heterogeneous activity) for sample size varying 

from 30-200 trees. The characteristics of each dataset are summarized in Table 3. 

Figure 7: Heterogeneity map computed from Refmap using the ArcGis spatial statistics tool “slope” (ESRI, 1998) that calculates the 

maximum rate of change in value from one cell to its neighbors. Rectangle delineates the three compartments (A) homogeneous, (B) 

intermediate and (C) heterogeneous used respectively weighted 0.1, 0.3 and 0.6 in the stratified sampling strategy. 

At the spatial scale, the comparison amongst ASsubmaps (Fig.8) clearly demonstrates that 

lower RMSEs are obtained when trees are sampled preferentially in areas with heterogeneous 

rockfall activity. Lower errors are also achieved when older trees are selected, and this finding 

is regardless of sample size. By contrast, 1.3 (30 trees) to 9 times (200 trees) larger discrepancies 

are observed when trees are selected in homogeneous areas and when trees without 

visible impact are arbitrarily selected. When comparing the AS and RE submaps, even lower 

RMSEs can be found for old trees and for a subset 150, the best 

RE results show lower RMSE than any AS reconstruction, even though similarly low levels of 

RMSE can be achieved if sampling preferentially focuses on older and frequently impacted 

trees in the heterogeneous areas. By comparing the AS and REsubmaps, it also becomes 

obvious that preferential sampling of trees without visible impacts results in higher RMSE 

values. Errors are lower if old trees are preferably sampled, over younger trees. 


Figure 8: Root Mean Square Error (RMSE) measuring the deviation of frequency maps generated with different subsets from the 

reference frequency map. RMSE is given for various sample sizes of best and worst random extractions as well as for different arbitrary 

selections. Preferential sampling in areas with heterogeneous rockfall activity is presented by ‘E’ and in homogeneous by ‘O’. 

Temporally, Table 3 demonstrates that in many cases, chronologies from AS are more reliable 

than those from RE, regardless of the sample size. For sample size

CONCLUSION 

Evaluating the potential of tree-ring analysis on an extensively sampled slope in the Valais 

region (Swiss Alps) reveals that the optimal sampling size and strategies will depend strongly 

on the aim of the reconstructions. We demonstrate that for a site with frequent rockfalls 

composed of individual rocks, as little as 6-10 trees/ha can be sufficient to obtain frequency 

maps similar to those obtained with the full dataset containing 63 trees/ha. Temporally, 

results show that 40 trees/ha can be sufficient to reconstruct 80% of past rockfall event years 

and that the chronologies obtained appear balanced. Although the thresholds provide very 

valuable indications on optimal sample sizes needed for reliable reconstructions, they should 

not be seen as absolute values. Instead, sample design and the number of investigated trees 

will need to remain flexible, as the nature of the process, topography, availability and ability 

of trees to record events will differ from site to site. In addition, representative trees will need 

to be selected with great care, even more so in case of small sample sizes. We thus suggest 

that trees should be selected after a preliminary assessment of process activity at the study site 

and based on the degree of heterogeneity in rockfall frequency. With respect to the selection 

of trees, we encourage a balanced choice of different age classes including old and heavily 

affected trees. Optimized minimum sampling sizes and design will ultimately facilitate 

fieldwork, and render analyses and interpretation more reliable, less time consuming, and 

will also improve cost-benefit ratios. 

REFERENCES 

- Alestalo, J., 1971. Dendrochronological interpretation of geomorphic processes. 

Fennia 105: 1-139 

- Astrade, L., Stoffel, M., Corona, C., Lopez Saez, J., 2012. L'utilisation des cernes de 

croissance des arbres pour l'étude des événements et des changements morphologiques : 

intérêts, méthodes et apports des recherches alpines à la dendrogéomorphologie. 

Géomorphologie: Relief, Processus, Environnement3: 295–316. 

- Berger, F., Quetel, C., Dorren, L., 2002. Forest: A natural protection mean against rockfall, 

but withwich efficiancy? The objectives and methodology of the ROCKFOR project. In: 

Proceedings International Congress Interpraevent 2002 in Pacific Rim, Matsumoto, Japan, 

pp. 815-826. 

- Corona, C., Lopez-Saez, J., Stoffel, M., 2014. Defining optimal sample size, sampling design 

and thresholds for dendrogeomorphic landslide reconstruction. Quaternary Geochronology 

22: 72-84. 

- Corona, C., Lopez Saez, J., Stoffel, M., Rovéra, G., Edouard, J.L., Berger, F., 2013a. 

Seven centuries of avalanche activity at Echalp (Queyras massif, southern French Alps) 

as inferred from tree rings. The Holocene 23:2: 292-304 

- Corona, C., Lopez-Saez, J., Stoffel, M., Bonnefoy, M., Richard, D., Astrade, L., Berger, F., 

2012. How much of the real avalanche activity can be captured with tree ring ? An evaluation 

of classic dendrogeomorphic approaches and comparison with historical archives. 

Cold Regions Science and Technology 74-75: 31-42. 


- Dorren, L.K.A., Berger, F., Le Hir, C., Mermin, E. and Tardif, P., 2005. Mechanisms, effects 

and management implications of rockfall in forests. Forest Ecology and Management 

215(1-3): 183-195. 

- Dorren, L., Berger, F., Jonsson, M., Krautblatter, M., Moelk, M., Stoffel, M., Wehrli, A., 

2007. State of the art in rockfall–forest interaction. Schweizerische Zeitschrift für Forstwesen158 

(6): 128–141. 

- Gsteiger, P., 1989. Steinschlag, Wald, Relief. Empirische Grundlagen zur Steinschlagmodellierung. 

Geographisches Institut, Universität Bern. 

- Jahn, J., 1988. Entwaldung und Steinschlag. Proceedings of the International Congress 

Interpraevent 1988, 1: 185-198. 

- Moya, J., Corominas, J., Pérez Arcas, J., 2010. Tree-ring based assessment of rockfall 

frequency on talus slopes at Solà d’Andorra, Eastern Pyrenees. Geomorphology 118 (3-4): 

393-408. 

- Schneuwly-Bollschweiler, M., Corona, C. Stoffel, M., 2013. How to improve dating quality 

and reduce noise in tree-ring based debrisflow reconstructions, Quaternary Geochronology 

18: 110-118. 

- Schweingruber F.H., 1996. Tree rings and environnment: Dendroecology. Paul Haupt, Bern, 

Stuttgart, Wien, 609 pp. 

- Šilhán, K., Pánek, T., Hradecký, J., 2013. Implications of spatial distribution of rockfall 

reconstructed by dendrogeomorphological methods. Nat. Hazards Earth Syst. Sci. 13: 

1817–1826. doi:10.5194/nhess-13-1817-2013 

- Stahle, DW., Fye, FK., Therrell, MD., 2003. Interannual to decadal climate and streamflow 

variability estimated from tree rings. Development in Quaternary Sciences 1: 491-504. 

- Stoffel, M., Corona, C., 2014. Dendroecological dating of geomorphic disturbance in trees. 

Tree-Ring Research 70, 3-20. 

- Stoffel, M., Bollschweiler, M., 2010. Tree-ring analysis in natural hazards research – preface. 

Natural Hazards and Earth System Sciences 10: 2355–2357. 

- Stoffel, M., Bollschweiler, M., 2008. Tree-ring analysis in natural hazards research – an 

overview. Natural Hazards and Earth System Sciences8: 187–202. 

- Stoffel, M., Perret, S., 2006. Reconstructing past rockfall activity with tree rings: some 

methodological considerations. Dendrochronologia24(1): 1-15. 

- Stoffel, M., Butler, D.R., Corona, C., 2013. Mass movements and tree rings: A guide for 

dendrogeomorphic field sampling and dating. Geomorphology 200: 106-120. 

- Stoffel, M., Lièvre I., Monbaron, M., Perret, S., 2005a. Seasonal timing of rockfall activity 

on forested slope at Täschgufer (Swiss Alps)- a dendeogeomorphical approach. Zeitschrift 

fuer Geomorphologie 49: 89-106. 

- Stoffel, M., Scheuwly, D., Bollschweiler, M., Lièvre I., Delaloye, R., Myint, M., Monbaron, 

M., 2005b. Analyzing rockfall activity (1600-2002) in a protection forest- a case study using 

dendrogeomorphology. Geomorphology 68: 224-241. 


- Trappmann, D., Corona, C., Stoffel, M., 2013. Rolling stones and tree rings: a state of 

research on dendrogeomorphic reconstructions of rockfall. Progress in Physical Geography 

37(5): 701-716. 

- Varnes, D.J., 1978. Slope movements: types and processes. In Schuster, R.L., Krizek, R.J., 

(eds), Landslide analysis and control. Transportation Research Board, Special Report 176, 

Washington D.C., pp 11-33. 



Critical Rainfall Conditions Triggering Shallow 

Landslides or Debris Flows in Torrents - Analysis of 

Debris Flow events 2012, 2013 and 2014 in Austria 

Markus Moser, DI 1 ; Stefan Janu, DI 2 ; Susanne Mehlhorn, Dipl. Geogr. 2 

ABSTRACT 

Generally, debris flows are caused by both small-scale intensive precipitation and long lasting 

rainfalls with lower intensity but high pre-wetting or both combined. The triggering mechanism 

of the debris flow events in Austria 2012, 2013 and 2014 were mass movements on 

steep slopes in the upper catchments. Those masses slide with very high velocity into the 

torrent beds provoking hyperconcentrated flows or debris flows. In areas of the geologically 

unstable Greywacke zone, the torrents were cleared up onto the bedrock and the debris was 

deposited in the storage areas of existing debris flow breakers or in torrents without technical 

protection measures the debris caused catastrophic damage on the alluvial fan. Following the 

events, comprehensive documentations and analyses were undertaken to support the 

understanding of the occurred processes to mitigate future hazards. Unfortunately, the 

small-scale heavy rain events are not detected by the precipitation stations. Therefore, 

weather radar data (INCA-Data) analysis was used to determine the - usually very local - intensities 

which caused those catastrophic landslides and debris flows. 

KEYWORDS 

critical rainfall conditions; debris flow; event analysis 

INTRODUCTION 

The knowledge of which precipitation intensities and durations are capable of triggering 

debris flows and landslides are of decisive importance to the effort of optimizing integrated 

protection concepts in torrent control. This knowledge is won through thorough event 

documentation and detailed analysis, here, of debris flow events that occurred in Austria in 

2012, 2013 and 2014. These events were characterized by very localized precipitation events 

of high intensities, both with and without a high degree of pre-wetting. Most of the debris 

flow events were triggered by slope failures in the uppermost catchment areas, which in turn 

were caused by a high water saturation of the soil and a temporary increase in the pore water 

pressure. The slopes were therefore destabilized through a reduction in the soil’s shear 

strength. 

1 Austrian Service for Torrent and Avalanche Control, Vienna, AUSTRIA, markus.moser@die-wildbach.at 

2 Austrian Service for Torrent and Avalanche Control 



PROBLEM STATEMENT 

Determining the precipitation sums per unit of time which are capable of triggering debris 

flow events such as those considered here are problematic for two reasons. Due to the very 

small spatial extent of the precipitation events, an analysis of data from precipitation stations 

(which are often not situated directly in the catchment area in question) is difficult and needs 

rely on interpolations. In addition to this, factors such as the event‘s prehistory (e.g. degree of 

pre-wetting) and the area’s disposition due to its geology play a deciding role. Numerous 

published studies, beginning with Caine (1980), Crozier (1986) and Zimmermann (1997), 

can be found that report empirically determined threshold values based on the precipitation 

parameters of intensity i [mm/h] and duration D [h]. The data used to investigate the 

relationships of intensity and duration were drawn from events which occurred in areas that 

have a great diversity of climatological and geological characteristics, and as such, the 

databases must be considered as rather inhomogeneous. Crozier extended the method in a 

1997 study to include precipitation data from rainfall events that did not trigger mass 

movement events. Guzzeti et al. (2008) employed the same method, but constrained the 

threshold values for a variety of climatic regions. MLIT (2004) expanded the methodology 

employed by Crozier (1986) by considering the prehistory (pre-wetting) of the events 

analysed, as well as the effective rainfall necessary to trigger slides and debris flow events. 

Strenger (2009) attempted to calculate precipitation threshold values for the triggering of 

debris flow events in the Vinschgau (South Tyrol, Italy) using the Antecedent Daily Rainfall 

Models after Glade et al. (2000). He was able to prove that although the model is generally 

well suited for such calculations, it is inadequate to capture the heavy rainfall events, which 

are usually of a very localized spatial extent, due to the poor spatial resolution of weather 

station data. Also Braun 2014 referred in his study to the inadequate detection of local heavy 

rainfall events with weather station data. Analyses of rainfall events using the highest 

available resolution for the whole dataset of one day lack the short term component. 

METHOD 

For this study, the analysis of a precipitation intensity and duration relationship was carried 

out using weather radar data (INCA data), limited spatially to a single geological unit in 

Austria, the Greywacke zone. The dataset was furthermore divided into mass movement 

events that occurred either with or without pre-wetting to take into account the catchment’s 

prehistory and disposition. The threshold values were empirically determined using the 

relationship found in measured precipitation data (weather station as well as weather radar 

data) between precipitation intensity i [mm/h] and duration D [h] (I = α× Dβ). The smallest 

precipitation sum which triggered a landslide or debris flow was selected as the threshold 

value. 


DATA SOURCES 

Weather radar data (INCA) 

The INCA (Integrated Nowcasting through Comprehensive Analysis) System of the ZAMG 

(Zentralanstalt für Meteorologie und Geodynamik; Eng: Central Institute for Meteorology and 

Geodynamics) is used for spatial and temporal analyses of a high resolution, as well as for 

weather predictions on the order of a few hours, taking particular account of regional and 

local topographic effects. The precipitation analysis of the INCA System comprises a combination 

of interpolated weather station data (with elevation effects taken into account) and 

weather radar data. The combined use of radar and rain gauges provides the benefit of each 

of these instruments’ respective strengths: the exactness of a point measurement on the 

ground (rain gauge) and the acquisition of a precipitation system’s spatial structure with the 

aid of weather radar. On the other hand, the methodology is not without weaknesses or 

possible shortfalls: unrepresentative locations and/or low spatial density of weather stations 

and rain gauges on the one hand, as well as the uncertainty of indirect precipitation measurement 

via weather radar on the other hand. The available data raster has a horizontal 

resolution of 1 x 1 km and a temporal resolution of 15 minutes. 

Figure 1: Geological map with catchment areas 


Geological conditions 

All of the investigated events occurred in the Greywacke zone (from Greywacke, the name of 

a Paleozoic sandstone). Geologically, this zone is located between the Northern Limestone 

Alps towards the north, of which it forms the geological base, and the Central Eastern Alps in 

the south. The bedrock consists of phyllite and the Paleozoic slate of the Greywacke zone, 

which are mostly extensively shattered and as a consequence prone to mechanical and 

chemical weathering. These rocks, especially the clay layers, produce a laminar, creeping 

scree material that covers the slopes in a substantial mantle of weathering debris of various 

thicknesses up to the ridges. This unconsolidated rock already becomes unstable at minimal 

levels of water uptake, which, upon further wetting produces large mass movements. 

DATA SOURCES 

Disposition – with and without pre-wetting 

The following two examples serve to illustrate the importance of a catchment area’s prehistory. 

The event which occurred in Klemmgraben in 2014 was a single event without any 

antecedent precipitation (0 mm precipitation in 3 days), whereas the event that occurred in 

the Lorenzerbach in 2012 followed three days of antecedent rain( 200 mm precipitation 

in 3 days) with the resulting high degree of pre-wetting making for a higher disposition. 

The different wetting-situations result in different behavior of the catchments. 

Figure 2: Precipitation and total precipitation – Klemmgraben 2014 


Figure 3: Precipitation and total precipitation – Lorenzerbach 2012 

STUDY SITES AND EVENTS 

Lorenzerbach, Paltental 2012 

The muddy debris flow event of 21. July 2012 in St. Lorenzen was caused by a combination 

of high sums of precipitation in the preceding 4 weeks (430 mm, this equals to a return 

period of more than 300 years) and heavy rain on the night of the event. Since there are no 

precipitation stations inside the catchment area, the areal precipitation was determined with 

the aid of weather radar data (INCA analysis, raster data 1 x 1 km, 15 minute intervals.). 

The highest 15 minute precipitation rate within the catchment area was found to be nearly 

40.4 mm/15 min and occurred on the 19th of July. In sum, from 14:00 UTC of the day before 

the event up to shortly after the debris flow, 96 mm of precipitation was registered in the 

catchment area of the Lorenzerbach torrent. These extreme precipitation sums led to a 

complete saturation of the groundwater reservoirs and to an activation of flow- and slide 

processes along the gorge sections. 

Sattelbach 2013 

A similar disposition could be reconstructed for the debris flow event in the Sattelbach 

torrent. A high pre-wetting since the beginning of May 2013, in combination with intensive 

rainfall toward the end of that month and on the 1st and 2nd of July, resulted in numerous 

debris flow and landslide events in the torrential catchment areas of Salzburg’s Pinzgau and 

Pongau regions. A catastrophic debris flow occurred in the municipality of Hüttau on the 

morning of 2. June 2013. Following a three day period of rain, a marked increase in 

precipitation rates occurred in the night of 1. June to the morning of 2. June. Evaluation of 

the precipitation rates obtained from weather radar data revealed a precipitation sum of 

80 mm in the catchment area up to the time of the debris flow. The trigger of the debris flow 

could be identified as a 2000 m² large landslide in the upper part of the catchment. The 


steepness of the ravine consequently enabled the development of a debris flow, which in its 

passing cleared the fore-filled sediment down to the bedrock. According to expert estimates, 

the fully developed debris flow transported 12,000 m³ of material. 

Klemmgraben 2014 

In terms of pre-wetting, a different disposition existed for the debris flows of 2014. Very short 

but intensive precipitation events were responsible for triggering these debris flows, a 

circumstance that was substantiated through analyses of weather radar data as well as data 

from surrounding weather stations. The Klemmgraben event on the 1st of August was 

likewise triggered by a very short but high intensity precipitation event. According to eyewitness 

reports, the heavy rain cell was situated above the upper portions of the catchment 

area. The analysis of the local precipitation was complemented by processing of INCA data, 

the sum of the precipitation was 28.3 mm/45 min with a peak of 12 mm/ 15 min. The 

surrounding precipitation stations registered from 16.9 mm (Flachau) to 33.2 mm (Radstadt) 

of rainfall on that day. The station Hüttau, in the north west of the catchment area, registered 

60 mm/45 min. 

RESULTS - CRITICAL RAINFALL CONDITIONS 

A number of studies (Austria/BFW 1972 – 2000, Caine 1980, Caine in Dodt 2007, Giannecchi 

2006, Braun 2014) have described critical rainfall conditions as a relationship between the 

intensity i [mm/h] and duration D [h] of a precipitation event that acts as a trigger for mass 

movements. 

The precipitation intensities – values in 15 minute intervals, derived from INCA analyses 

– that were associated with the debris flow events were classified according to the type of 

event and displayed graphically in a precipitation intensity-duration diagram. 

The following two event types were defined: 

– Events with a high level of pre-wetting 

– Events that occurred without any pre-wetting 

A comparison reveals the high intensities associated with events that occurred without 

pre-wetting and the low intensities associated with events with pre-wetting. 

The lowest precipitation values which triggered a debris flow were selected in order to obtain 

functions of rainfall intensity and duration thresholds for each event type in the Greywacke 

zone. Regression analysis yields the following relationship between threshold precipitation 

intensity and duration for events which occurred without pre-wetting: 

Formula 1: Regression equation for the relationship between 

precipitation intensity and duration for events without pre-wetting 

For events with a high degree of pre-wetting, these precipitation parameters are related thus: 

Formula 2: Regression equation for the relationship between 

precipitation intensity and duration for events with pre-wetting 


Figure 4 Rainfall intensity-duration thresholds of 8 debris flow events occurred 2012 - 2014 in the Greywacke zone, Austria 

Table 1 Precipitation threshold values per duration, event type and difference between event types in % 

This comparison also confirms the generally valid statement that relatively low precipitation 

intensities are sufficient to trigger debris flows if a high degree of pre-wetting is given. The 

estimated threshold value for the short duration for example of 0.5 hours only 58 % from the 

threshold value without pre-wetting is needed. The highest precipitation intensities found for 

the events described above (in mm/h, mm/2h, and so forth) correspond well to the values 

quoted in literature, as being critical, debris flow triggering rainfall conditions. 


DISCUSSION 

The advantage of the method introduced here, using weather radar data, is that it also allows 

precipitation threshold values to be calculated for very locally occurring intense rainfall 

events which cannot be sufficiently resolved using the existing weather station network. It is 

however important to not only consider the immediate time period of the mass movement 

in question, but to also include and analyse the event’s pre-history within its catchment area. 

Those different pre-wetting situations are the reason for a different behavior of the catchments 

and result in a higher disposition for catchments with a high degree of pre-wetting. 

Dividing a given dataset into events with and without pre-wetting is a first step and the 

examples provided here clearly illustrate that the results may differ quite substantially. Since 

different catchment areas may have significantly different dispositions for producing 

landslides and debris flows, based on their geology. Therefore only events in geological similar 

catchment areas were selected. An even more detailed data acquisition will be necessary for 

further investigations; this example may constitute an applicable approach for such research 

ventures. 

REFERENCES 

- Braun, M. 2014: Hydrometeorological Triggers of Debris Flows. Evolution of the temporal 

occurrence of debris flows between 1900 and 2008. Diplomarbeit / Masterarbeit - Institut für 

Alpine Naturgefahren (IAN), BOKU-Universität für Bodenkultur. 

- Caine, N., 1980: The rainfall intensity – duration control of shallow landslides and debris 

flows. Geografiska Annaler Series a-Physical Geography, 62(1-2): 23-27. 

Crozier, M. J.: Landslides: causes, consequences and environment, Routledge, London- 

New York, 252, 1986. 

- Guzzetti, F., Peruccacci, S., Rossi, M. and Stark, C.P., 2008. The rainfall intensity-duration 

control of shallow landslides and debris flows: an update. Landslides, 5(1): 3-17. 

- Zimmermann, M., Mani, P., Gamma, P., 1997. Murganggefahr und Klimaänderung - 

Ein GIS-basierter Ansatz. Nationales Forschungsprogramm "Klimaänderungen und Naturkatastrophen" 

(NFP 31), Schlussbericht. vdf Hochschulverlag AG an der ETH Zürich. 

- Glade T. (2000): Modelling landslide-triggering rainfalls in different regions of New Zealand 

the soil water status model. In: Zeitschrift f. Geomorphologie, Suppl.-Bd.122, S.63-84. 

- Godt. J. W., Coe J. A ., 2006: Alpine debris flows triggered by a 28 July 1999 thunderstorm 

in the central Front Range, Colorado, U.S. Geological Survey, Denver Federal Center, Box 

25046, M.S. 966, Denver, CO 80225-0046, USA 

- MLIT, 2004: Guidelines For Development of Warning and Evacuation System Against 

Sediment Disaster in Developing Countries, Ministry of Land, Infrastructure and Transport - 

Infrastructure Development Institute. Japan (2004). 

- Strenger, Mark-Philip (2009) Niederschlagsschwellenwerte bei der Auslösung von Muren. 

Diplomarbeit, Universität Wien. Fakultät für Geowissenschaften, Geographie und Astronomie; 

BetreuerIn: Glade, Thomas 



Experimental Study on Effect of Houses on Debris-Flow 

Flooding and Deposition in Debris Flow Fan Areas 

Kana Nakatani, Dr. 1 ; Megumi Kosugi 2 ; Yuji Hasegawa 3 ; Yoshifumi Satofuka 4 ; Takahisa Mizuyama 5 

ABSTRACT 

This study conducted model experiments to determine the influence of houses on debris flow 

flooding and deposition. We applied uniform and also coarse-grained sediment in the 

following scenario cases: without houses; with houses; with houses and fences; and with 

houses that can be destroyed. The model experiments showed that when houses are present, 

the debris flow spreads widely in the cross direction immediately upstream of the houses. 

Houses located in the debris fan also influence the deposition area. Especially when fences 

exist around the houses, flow moves down between the fences, as in the real disaster cases, 

where flow moves down towards the roads. Finally, when houses are destroyed, flow moves 

down through the destroyed houses, and changes the flooding and deposition process 

compared with the non-destroyed houses case. With debris flow containing coarse-grained 

sediment, most of it deposits in the upstream area of the houses when houses exist. 

KEYWORDS 

debris flow; debris flow fan area; house influence; model experiment 

INTRODUCTION 

Debris flow causes flooding and sediment deposition when it reaches the debris flow fan area. 

In Japan, houses are usually built in debris flow fan areas through urban development, and 

these buildings affect flow and deposition during a debris flow event. As the number of 

people living in debris flow fan areas has increased because of rapid urban development, 

debris flow disasters, such as the one that occurred in August 2014 in Hiroshima Prefecture, 

Japan, have caused widespread damage to human lives and infrastructure. However, few 

studies have examined the effect of houses on debris flow disasters. This study conducted 

model experiments to determine the influence of houses on debris flow flooding and 

deposition. 

METHODS 

First, we considered typical landform and debris flow-scale conditions. For landform 

conditions, we set a mountainous torrent in the upstream area as a rectangular straight open 

channel with a length, width, and slope of 5 m, 0.1 m, and 15°, and debris flow fans as 

1 Kyoto University, Kyoto, Kyoto JAPAN, kana2151@kais.kyoto-u.ac.jp 

2 Ministry of Land, Infrastructure and Transport, Chubu Regional Bureau, Tenryugawa Upstream River Office 

3 Disaster Prevention Research Institute, Kyoto University 

4 Department of Civil Engineering, Ritsumeikan University 

5 National Graduate Institute for Policy Studies 



esidential areas in the downstream area (Fig.1). The model scale was 1/50, and the transverse 

direction was set as flat. We covered the experimental channel with 0.1-m-thick 

sediment and supplied a steady flow of water from upstream to generate a debris flow. 

The peak discharge and continuance time were set by considering that small-scale debris flow 

occurs at a high frequency of 2.0 L/s for 30 s and 5.0 L/s for 20 s. We examined cases of 

uniform sediment (diameter 3 mm) and uniform sediment with 5% coarse-grained sediment 

(diameter 20 mm). The sediment density was 2.58 g/cm 3 and supplied sediment volume was 

45 L (including void). As shown in Table 1, we considered the scenarios of without houses, 

with houses, with houses and fences, and with destructible houses. To model actual two-story 

6-m-high houses and 2-m-high fences, 0.12-m-high house and 0.04-m-high fence models 

were used. In this study, we set destructible houses to see how the flow depth and deposition 

caused by debris flow change from cases with non-destructible houses. Therefore, we 

modeled the destructible house which can be destroyed with debris flow caused by supplied 

discharge 5.0L/s, and didn’t consider the strength of house. 

Figure 1: Cross-sectional (top left) and longitudinal (bottom left) profiles of the debris flow experiment model (top right), houses and 

sensors location (middle right) and destructible houses (bottom right) 


Table 1: Experiment cases 

To see the time series difference of debris flow direction in debris fan due to houses and 

fences existence, we recorded the experiments in video and the time series of flow depth and 

deposition thickness using ultrasonic wave sensors (see Fig.1). The sediment thickness and 

range were also measured after the experiments. 

RESULTS 

The model experiments showed that when houses were present, the debris flow spread 

widely in the transverse direction immediately upstream of the houses (Fig. 2 left). Houses 

located in the debris flow fan also influenced the deposition area. In the case of houses with 

fences, the flow moved down between fences, which is similar to actual disasters where 

debris flow moves down toward the roads. 

Fig. 3 shows the results of deposition thickness distribution. From results with no houses and 

uniform sediment cases, 2 L/s (Case 1) showed deposition in the upstream side where the 

incline changed from 12° to 9°. With 6 L/s (Case 4), results showed deposition in the 

downstream side where the incline changed from 9° to 6°. It is assumed that the deposition 

was affected by the change in the vertical section-like incline. When compared with the 

uniform sediment case and the case with 5% coarse-grained sediment, the result that 

included coarse-grained sediment showed deposition widely in the downstream area. 

When houses exist, results show deposition especially in the upstream of the house in the 

steepest location (Case 2). When fences exist around the houses, deposition becomes more 

remarkable around the houses and fences (Case 3). Results in 6 L/s cases show the same 

trend. When houses and fences exist, flow spreads in the transverse direction and deposition 

occurs in the upstream area compared with the cases without houses. When fences exist, 

flows are obstructed and deposition seems to occur because the flow move between the 

fences. When houses are destroyed by the debris flow (Case 6 and 10), flow pass through the 

inside of the destroyed houses, and results show deposition in the downstream area compared 

with the results of the cases with non-destructible houses. 


Figure 2: Deposition after the experiment (left) and sensors results (right) for debris flow discharge of 2.0 L/s (Case1-3) 

Fig. 4 shows the results of coarse sediment distribution and surface deposition. The plot 

shown in a circle in the figure represents the position of the coarse sediment in the surface; 

coarse sediment inside the deposition layer is not plotted as a circle. 

In the cases without houses (Case 8), little coarse sediment exists in the center line of the 

alluvial fan in the surface, in the 12–6° step-gradient. This seems to happen because when the 

flow reaches the fan from the channel, flow spreads widely, and then coarse sediment is left 

on the side where flow depths become small. Meanwhile, in the low-gradient (6–3°) area, 

coarse sediment deposits in the center of the surface. 


Figure 3: Sediment deposition thickness distribution for all the cases 


Figure 4: Coarse-sediment distribution and surface deposition (Cases 8-10) 

In the case of houses without destruction (Case 9), a larger deposition of the gravel occurs 

on the upstream side because of the influence of the houses. In the upstream section in the 

figure, no deposition can be seen, but in the 2nd and 3rd sections from the upstream, 

deposition of the gravel can be noted. Passing through the area with the houses, gravel moves 

downstream to the low-gradient 6–3°area. In Case 10 (houses with destruction), the 

distribution of the gravel is slightly different compared with Case 9. In Case 10, deposition 

of gravel can be seen in the upstream section because a large amount of gravel deposited 

upstream of the houses and on the inside of the destroyed houses. 

Fig.5 shows the data measured by the ultrasonic sensor. Note that in 6 L/s cases, flow 

splashed on the ch2 sensor due to the existence of houses, and data could not be acquired. 

During the experiment, the sensor measured the flow depth and the deposition thickness, 

and after the experiment it showed the deposition thickness. 


Figure 5: Sensor results of height in Cases 4, 5, 7(left), in Cases 4, 5, 8, 9 (center) and in Cases 5, 6, 9, 10(right) 

From Fig. 2 right, Case 2 showed high value compared to Case1 in ch1. Case 3 data was not 

recorded due to sensor trouble in ch1. Since the ch1 sensor is placed on the right bank side, 

the sensor reaction shows that flow expanded from the house located most upstream, and it 

reached the position of ch1. After 60 s passed, Case 2 still showed a large height, which reveals 

that deposition occurred in the front of the house. In the ch2 sensor, no major changes 

can be seen at the start of the value increase. Case 3 showed a slightly larger value at the 

start. After the increase, Case 1 height seems to decrease. However, values of Cases 2 and 3 

remained same and did not decrease. This seems to happen because sediment was not eroded. 

In ch3, Case 1 sensor did not react. In Cases 2 and 3, the sensor reacted and showed that flow 

and deposition had occurred. The extension of flooding and deposition occurred because of 

the existence of houses, which was more significant in Case 3. In ch4, Case 1 showed the 

largest value, second largest was Case 3, and smallest was the Case 2. The reason why Case 3 

showed a lager value than Case 2 is that when fences exist, flow seems to be inhibited 

between the fences and deposition occurs in that position. 


Fig. 5 left show the cases of 6 L/s comparing the existence of houses and fences. Case 7 

showed the largest value, second largest was Case 5, and the smallest was Case 4 in ch1. 

These results are due to the effect that flows spread in the transverse direction when houses 

and fences exist. In ch3, Case 4 did not react, Case 5 value increased to 5 mm, and Case 7 

increased to 30 mm, then decreased and showed 5 mm deposition at 60 s. In Case 7, two 

peaks of the value can be seen. From the experiment observation, the first peak happened 

when flow reached the alluvial fan, flowed down the center of the fan, and moved toward 

ch1, then ch3 from the outer side. The second peak happened when the flow reached the 

alluvial fan, moved down the center of the fan, passed through the area between the fences, 

and moved towards ch3. In ch4, Case 4 showed the largest value, followed by Case 7, and the 

smallest was Case 5’s. 

Fig. 5 center show the cases of 6 L/s comparing the influence of the existence of houses and 

the coarse-sediment. Case 5 and Case 9 results showed a larger value than cases without 

houses in ch1. Case 9 result including coarse-sediment showed the largest value. In Cases 4 

and 8, flow reached the alluvial fan, spread in the transverse direction, then reached the ch1 

sensor and showed reaction. Case 4 showed a larger value than Case 8. In ch3, the sensor did 

not react in cases without houses because flow did not reach this area. Cases 5 and 9 showed 

reaction in ch3, and the largest value was shown in Case 9, as in ch1. In ch4, the sensor 

showed a high value at the peak in Cases 4 and 8. In the latter half of the experiment, the 

value decreased in all cases and no significant change could be seen. 

Fig. 5 right show the cases of 6 L/s comparing the influence of the existence of destructible 

houses and the coarse-sediment. In ch1, cases of destructible houses showed a smaller value 

compared with the cases of non-destroyed houses. Case 6 showed the smallest value, and the 

next smallest was Case 10. This difference seems to happen because of the house destroyed 

time. The house located at the top was destroyed earlier in Case 6 than in Case 10, therefore 

flow passing through to the downstream became larger and flow moving in the transverse 

direction to ch1 decreased. In ch3, the sensor reacted in Cases 5 and 9. In cases with 

destroyed houses, the value is small and also the duration is short in ch3. In ch4, almost all 

cases show the same start of the data increasing to the peak value. Cases 9 and 10 showed 

almost the same peak value, and Case 10 showed a slightly earlier peak time. This was 

because the flow passed through the destroyed house and flowed downstream in Case 10. 

On the other hand, Case 9 took more time because flow moved around the houses. In 

uniform sediment Cases 5 and 6, initially flow reached ch4 and showed the same reaction. 

After 8 s, Cases 6 and 5 showed different reactions. This was because house destruction in 

Case 6 occurred at the top, and flow moved down and deposition occurred around and inside 

the destroyed house. The flow from the upstream to ch4 was temporarily reduced. 


CONCLUSIONS 

We investigated the effect of the presence of houses and fences, and the destruction of houses 

in flooding and deposition of debris flow by a model experiment. Results showed that when 

houses exist, the flooding and deposition area changes, and remarkable deposition occurs 

upstream of the houses. When houses and fences exist, flow spreads in the transverse 

direction and deposition occurs in the upstream area compared with cases without houses. 

When fences exist around the houses, flow moves down between the fences toward the road. 

When houses can be destroyed, water and sediment pass through the destroyed houses; 

therefore sediment can move downstream compared with cases with non-destroyed houses. 

When including coarse-sediment, a large amount of gravel deposits upstream and inside of 

the destroyed houses. 

According to the time-series changes, when fences exist around the houses, flow direction 

can be changed and more than one peak may occur from the flow passing through the roads. 

When house destruction happens, the timing of the destruction causes a difference of flow 

direction and peak time compared with the non-destroyed houses. 

On engineering hazard assessment, the area affected by debris flow have been estimated 

without considering houses influence until now. However, from this study, we confirmed that 

houses, fences, and the destruction of houses affect the debris flow behavior in an alluvial 

fan. Therefore, to estimate the hazard area with accuracy and to plan the effective mitigation 

measurement, it is necessary to consider the houses influence. For future studies, we need to 

gather more information about the impact of houses, fences, and destruction of houses in a 

real disaster. Based on these experimental results, we are planning to propose numerical 

simulation models and programs that can describe the effect of houses on the debris flow 

behavior in the alluvial fan area. 

ACKNOWLEDGEMENT 

This study was supported by the JSPS KAKENHI, Grant Number 15K16312, Grant-in-Aid for 

Young Scientists (B). 

REFERENCES 

- Nakatani, K., Okuyama, Y., Hasegawa, Y., Satofuka, Y., Mizuyama, T. (2013): Influence of 

housing and urban development on debris flow flooding and deposition, Journal of Mountain 

Science, Volume 10, Issue 2, pp. 273-280. 

- Nakatani, K., Kosugi, M., Hasegawa, Y., Satofuka, Y., Mizuyama, T. (2015): Experimental 

study on debris flow behavior in debris fan area: considering influence of houses. Journal of 

the Japan Society of Erosion Control Engineering 67(6): 22-32. (in Japanese) 



Bedload transport simulation with the model sedFlow: 

application to mountain rivers in Switzerland 

Dieter Rickenmann, Dr. 1 ; Martin Böckli 2 ; Florian U.M. Heimann, Dr. 2 ; Alexandre Badoux, Dr. 2 ; Jens M. Turowski, Dr. 3 

ABSTRACT 

The sedFlow simulation model for one-dimensional calculations of discharge and bedload 

transport was applied to five Swiss mountain rivers. After calibration it was capable of 

reproducing the observed bedload transport behavior reasonably well in the study rivers. 

For most simulations, the variable power equation (VPE) was used together with a reduced 

energy slope for the bedload transport calculations. The simulations with the Rickenmann 

(2001) transport equation were found to be suitable for simulation periods including a major 

flood event. The Wilcock and Crowe (2003) equation, a reference shear stress based approach, 

resulted in better agreement with long-term observations of bedload transport with 

only moderate peak flows. Using scenario simulations for a flood event, the effect of 

important lateral sediment input from tributaries due to debris-flow activity was investigated, 

indicating that bedload transport along the main river is mainly altered near the tributary 

locations. 

KEYWORDS 

bedload transport; 1D model; mountain river; Switzerland 

INTRODUCTION 

In alpine environments, bedload transport processes are dynamic and complex compared to 

lowland streams. Although there is a strong need for modelling tools in scientific and 

engineering applications, bedload transport in mountain streams is still relatively poorly 

understood. Field observations indicate that river bed morphology and thus hydraulic 

processes become increasingly complex as channel gradients become steeper. The range of 

observed grain diameters becomes larger, which entails more complex grain–grain and 

grain–flow interactions as well. Examining a large dataset on flow velocity measurements in 

steep streams, Rickenmann and Recking (2011) concluded that a considerable part of the 

river’s shear stress is consumed by turbulence due to complex bed morphology, summarized 

as macro-roughness. They proposed an procedure to quantify the impact of macro-roughness 

on bedload transport, which resulted in a better agreement between calculated and observed 

bedload transport in mountain streams and torrents (Nitsche et al., 2011). 

The sedFlow model for the one-dimensional simulation of discharge and bedload in a 

rectangular channel was developed at the Swiss Federal Research Institute WSL to assess river 

1 Swiss Federal Research Institute WSL, Birmensdorf, SWITZERLAND, dieter.rickenmann@wsl.ch 

2 Swiss Federal Research Institute WSL 

3 Helmholtz Centre Potsdam, GFZ German Research Centre for Geosciences, Germany 



ed morphodynamics and bedload transport on the catchment and reach scale in mountain 

rivers (Heimann et al. 2015a, 2015b; Heimann, 2014). The sedFlow model was applied in five 

Swiss mountain rivers, for which information on past bedload transport was available. The 

objective of this contribution is to present selected results and to discuss our experience with 

sedFlow. 

MODEL CHARACTERISTICS AND INPUT REQUIREMENTS 

The modelling tool includes the following main features: (1) consideration of state of the art 

approaches for the calculation of bedload transport in steep channels accounting for macroroughness; 

(2) bedload transport calculation for several grain diameter fractions separately i.e. 

fractional transport; (3) fast calculations for modelling transport in complete catchments and 

for performing scenario studies with automated calculation of many variations. 

A detailed description of the model structure and the numerical implementation is given by 

Heimann et al. (2015a). Using a simplified calculation procedure for the flow hydraulics, 

the sedFlow model can also simulate the effect of large sediment inputs by debris flows from 

tributaries. 

The required input parameters are the longitudinal channel profile, channel widths along the 

study reach, grain size distributions of the bed surface and subsurface sediment, and stream 

discharge. Typically, a calibration is made by varying the channel parameters such as grain 

size distribution and possibly channel width, if the width of a river reach is not naturally 

incised or constrained by lateral levees. In addition, different modelling options have to be 

defined: selecting a flow resistance equation, a bedload transport formula, the threshold for 

initiation of transport and the thickness of the sediment exchange layer at the bed surface. 

In sedFlow the stream channel is approximated by a rectangular profile; together with the 

Manning-Strickler equation, an analytical solution for the implicit flow routing using the 

kinematic wave approach is used, which results in fast calculation times. For steep streams 

with shallow flows it is recommended to use the variable power equation (VPE) of Ferguson 

(2007) (Rickenmann and Recking, 2011), and then the analytical solution for the implicit 

flow routing cannot be applied. Therefore, the VPE flow resistance calculation must be 

combined for example with explicit routing, but this requires relatively long computation 

times. An alternative option in sedFlow is to use of VPE together with a simplified hydraulic 

calculation, assuming constant flow per time step in channel reaches without lateral inflow. 

This model version results again in fast calculation times, and for (strong) lateral sediment 

inputs adverse channel slopes in the longitudinal profile are possible. For the application of 

sedFlow in the Brenno River it was found that a simplified backwater calculation produced 

similar bedload transport results as when using an implicit flow routing based on the 

kinematic wave approach. For the Kleine Emme, however this method of calculation resulted 

in substantially different, less plausible simulation results than the calculation with kinematic 

wave. The generally steeper channel reaches in the Brenno could be the reason that the 


simplified hydraulic calculation produced plausible results there. The model version combining 

the VPE with the simplified hydraulic calculation was used for all study rivers except for 

the Kleine Emmer River. 

The flow resistance calculations with the VPE used coefficients a 1 = 6.5 and a 2 = 2.5 as 

proposed both by Ferguson (2007) and Rickenmann and Recking (2011). A shear stress based 

bedload transport equation proposed by Rickenmann (2001) was used, modified for fractional 

transport calculations (Heimann et al. 2014a). To account for macro-roughness energy losses, 

a reduced energy slope was applied, based on an approach of Rickenmann and Recking 

(2011) and following a procedure described in Nitsche et al. (2011). Some calculations were 

also performed using the Wilcock and Crowe (2003) bedload transport equation, in combination 

with a reduced energy slope. 

STUDY CATCHMENTS AND SELECTED RESULTS OF MODEL APPLICATIONS 

The sedFlow model was applied to five Swiss mountain river catchments (Table 1). The best 

observations related to bedload transport were available for the Kleine Emme and the Brenno 

Rivers (Böckli et al., 2015c, 2015e). From consecutive surveys of river cross-sections, the net 

erosion and deposition along the longitudinal profile was derived, and additional information 

on gravel extraction and/or important sediment delivery from tributaries during the study 

period helped to constrain the total sediment budget (Rickenmann et al., 2014a; Heimann et 

al. 2015b). For the other three study catchments information of past bedload transport was 

less detailed (Böckli and Rickenmann, 2015; Böckli et al., 2015b, 2015d), for example with 

observed bedload volumes in a hydropower reservoir (Ferden at the Lonza River) or a sediment 

retention basin (Grosse Schliere). Below we summarize some results of our modelling 

experience. 

Effect of changing input parameters and transport equations 

The sedFlow model was first calibrated for the Kleine Emme and the Brenno mountain rivers 

over a study reach length of 20 km and 22 km, respectively. It produced a reasonable 

replication of the observed bedload transport for a five and ten year period, respectively 

(Rickenmann et al., 2014a; Heimann et al., 2015b). The Kleine Emme simulations included 

the extreme 2005 flood event (Rickenmann and Koschni, 2010) which resulted in substantial 

bedload transport due to important sediment input by bank erosion. 

The simulations with sedFlow showed that, in addition to choosing a suitable bed load 

transport formula, the minimum value for the critical dimensionless shear stress (Shields 

number) for beginning of bedload transport is also important for calibration. The critical 

Shields number is used with the transport formula of Rickenmann (2001), whereas a suitable 

reference shear stress has to be selected with the transport formula of Wilcock and Crowe 

(2003). The reference shear stress, which is comparable to the critical Shields number, can be 

varied by choosing the sand fraction of the surface bed material. Overall, the choice of the 


Table 1: Characteristics of the five study catchments. 

Kl. Emme Brenno Hasliaare Lonza Gr. Schliere 

Catchment area (A) [km 2 ] 478 (A) 397 (A) 554 (A) / 531 (B) 78 (A) / 131 (B) 27 (B) 

Mean eleva6on [m a.s.l.] 1050 1820 2150 2630 – 

Length simula6on reach 19.4 21.8 17.2 9.5 11.6 

Mean 

[km] 

channel slope [%] 0.8 2.6 3.8 4.7 7.7 

Sills or check dams yes no no no yes 

Channelized reaches yes no few no few 

Hydropower use 

liZle strong strong no no 

Tributaries 

(simula6on 

with 

reach) 

sed. input no yes yes yes yes 

debris-flow 

Gauging sta6ons 

ac6vity 

2 1 1 1 0 (C) 

(A) area upstream of FOEN gauging sta6on (Kleine Emme, Brenno, Hasliaare) 

(B) area upstream of most distal point of simula6on reach (Hasliaare, Lonza, Grosse Schliere) 

(C) discharge data from the neighboring catchment Kleine Schliere was used and adapted 

Shields number was important for the model calibration, particularly regarding the general 

level of bedload transport. The finally selected (optimized) Shields numbers were found to be 

in a plausible range, when compared with data from a recent field study on the beginning of 

bedload transport (Bunte et al. 2013). To best replicate the longitudinal pattern of net erosion 

and net deposition, only the grain size distribution along the river was optimized (in a 

plausible range) in the case of the Kleine Emmer River, and only the representative channel 

width was optimized (in a plausible range) for the depositional reaches in the Brenno River. 

For the simulation of the flood event 2011 in the Lonza, the transport formula of Rickenmann 

(2001) resulted in plausible results, also when using different hiding functions (which 

govern the fractional transport behavior). Here is an exponent of 1.5 was used in the 

calculation of the reduced energy slope. For the Lonza long-term simulations were made for 

the period 1976 to 2013, for which the cumulative sediment transport into the Ferden 

reservoir is known for 19 separate survey periods. During this period only one major flood 

event occurred in 2011. For the sedFlow simulations using the formula of Rickenmann 

(2001), the bedload transport was clearly overestimated (by about a factor of 10), even when 

varying the critical Shields number and the energy-slope exponent (Fig. 1). 

Using the transport formula of Wilcock and Crowe (2003) with an energy-slope exponent of 

1.5, resulted in similarly plausible results for the 2011 event as the formula of Rickenmann 

(2001). For the long-term simulations of Lonza, the transport formula of Wilcock and Crowe 

(2003) also yielded plausible results when either the sand fraction was set to 0.05 (high 

reference shear stress), or when a sand content of 0.20 (low reference shear stress) was used 

with an energy-slope exponent of 2 (Fig. 1). This result is not so surprising in the sense that a 

reference-based bedload transport equation can be expected to be better suited for weak to 


moderate bedload transport conditions. In fact, the bedload transport formula of Wilcock and 

Crowe (2003) in combination with a reduced energy slope and an exponent of 1.5 was found 

to provide reasonable agreement with field measurements of weak to moderate bedload 

transport over a wide range of channel conditions (Schneider et al., 2015). 

Effect of lateral sediment input 

The sedFlow model was applied to a 17 km long study reach of the Hasliaare River. There, 

several important debris-flow events occurred in tributaries during the last decade. In August 

2005 a large debris-flow event occurred in the Rotlauibach, a torrent catchment upstream of 

the village of Guttannen, and deposited a total volume of 500,000 m³ on the fan and in the 

main valley (Böckli et al., 2015d). After the 2005 event, sediment deposits with a thickness 

of about 1 m and a volume of about 15,000 m³ were observed in the Hasliaare channel at 

Innertkirchen, which is situated some 8 km downstream of the Rotlauibach. After calibration, 

the sedFlow model was able to replicate the observed deposition at Innertkirchen. 

Figure 1: Long-term simulations for the period 1976-2013 in the Lonza River without any major flood event. (GFsim/GFobs) is the ratio 

of simulated to observed bedload volume in the reservoir Ferden. Shown are results for different combinations of transport formulas 

(Ri: Rickenmann, 2001; W&C: Wilcock and Crowe, 2003), different Shields numbers (θc50), different sand fractions (Fs) and different 

energy-slope exponents (Exp). Simulations labelled “SL” were performed with a coarser layer of surface material. The box plots represent 

the variation over the 19 survey periods. The boxes include the median value and are limited by the upper and lower quartile. 

The whiskers indicate the maximum and minimum value. 

Extremely high debris-flow activity was observed in the Spreitgraben tributary in the period 

2009 to 2011, delivering a total of about 590’000 m³ into the Hasliaare River. Surprisingly 

only a very small part of this sediment was transported further downstream. Over six 

kilometers downstream of the Spreitgraben tributary, the sedFlow model predicted a strong 

decrease in bedload transport capacity, in qualitative agreement with the observations 

(Rickenmann et al., 2014b). 


The delta of the Hasliaare was surveyed repeatedly from 1905 to 1938 (Böckli et al., 2015d). 

From these data and accounting for fine material, we estimate the bedload supply to the Lake 

of Brienz to be about 40,000 to 70,000 m³/a. This amount is supported by geochronologic 

dating of the sediments in the alluvial plain between Meiringen and Lake of Brienz for the 

past 500 years (Carvalho and Schulte, 2013). Simulation results of bedload transport in the 

Hasliaare for entire years predicted an annual sediment delivery to the lake of Brienz which is 

in agreement with these observations (Rickenmann et al., 2014b). 

Scenarios of increased sediment input due to future debris flows from five tributaries were 

defined for further simulations. The basis was a flood hydrograph as in August 2005 with a 

peak flow of about 350 m³/s at the downstream end of the simulation reach, and a channel 

width of the order of 10 to 20 m. The simulation results show that both continuous (Fig. 2) 

and instantaneous (Fig. 3) lateral sediment input changes the bedload transport in the 

Hasliaare over a length of about 1 to 2 km both upstream and downstream of the active 

tributary. It can be observed that both the magnitude and the timing of the debris-flow inputs 

with regard to the duration of the flood event in the Hasliaare River have an important effect 

on the local bedload transport along the main river. An early input results in a larger decrease 

upstream and a larger increase downstream of the active tributary; more time is available for 

a reduction or increase of the bedload transport. The effect on the debris-flow deposit in the 

immediate confluence area tends to be greater, the later the sediment input occurs during the 

floods, while some distance further upstream an early entry may lead to more deposition. 

CONCLUSIONS 

After calibration, the sedFlow simulation model was capable of reproducing the observed 

bedload transport behavior reasonably well in five Swiss mountain rivers. For most simulations, 

the variable power equation (VPE) was used together with a reduced energy slope for 

the bedload transport calculations. The simulations with the Rickenmann (2001) transport 

equation were found to be suitable for simulation periods including a major flood event. 

The Wilcock and Crowe (2003) equation, a reference shear stress based approach, resulted 

in reasonable agreement with long-term observations of bedload transport for a 37 year 

period in the Lonza River with only moderate peak flows. Using scenario simulations for a 

flood event, the effect of important lateral sediment input from tributaries due to debris-flow 

activity was investigated. These simulations indicated that bedload transport along the main 

river is mainly altered near the tributary locations, suggesting that the transport further 

downstream may require multiple floods or longer flow durations. 


Figure 2: Simulated bedload transport in the Hasliaare River for the case of continuous sediment input from five tributaries. 

The simulation reach extends from Handegg (km 17) to Meiringen (km 0). A flood event as in August 2005 was simulated, and a total 

sediment input of 31000 m³ (low), 90000 m³ (avg.), 165000 m³ (high) due to debris flows from five tributaries (indicated by vertical 

solid lines in the graph) was assumed. Results are shown for different magnitudes of the debris-flow inputs during the flood event 

in the Hasliaare River. The vertical dotted lines indicate the location of the village Innertkirchen. 

Figure 3: Simulated bedload transport in the Hasliaare River for the case of instantaneous sediment input from five tributaries. 

The simulation reach extends from Handegg (km 17) to Meiringen (km 0). A flood event as in August 2005 was simulated, and a total 

sediment input of 165,000 m³ due to debris flows from five tributaries (indicated by vertical solid lines in the graph) was assumed. 

Results are shown for different timings of the debris-flow inputs during the flood event in the Hasliaare River. The vertical dotted lines 

indicate the location of the village Innertkirchen. 



The model development and the performance of the five case studies was supported by the 

following two projects: (1) «Sediment transport in mountain catchments» (Contract no. 

11.0026.PJ/ K154-7241) of the Swiss Federal Office for the Environment (FOEN), and (2) 

«SEDRIVER» (SNF Project no. 4061-125975) as part of the National Research Program NFP61 

of the Swiss National Science Foundation. 

REFERENCES 

- Böckli, M., Rickenmann, D. (2015a). Lonza – Geschiebetransport-Simulationen mit 

sedFlow. Eidg. Forschungsanstalt WSL, Birmensdorf, Schweiz. (*) 

- Böckli, M., Greber, C., Rickenmann, D. (2015b). Grosse Schliere – Geschiebetransport- 

Simulationen mit sedFlow. Eidg. Forschungsanstalt WSL, Birmensdorf, Schweiz. (*) 

- Böckli, M., Rickenmann, D., Heimann, F.U.M., Badoux, A. (2015c). Brenno – Geschiebetransport-Simulationen 

mit sedFlow. Eidg. Forschungsanstalt WSL, Birmensdorf, Schweiz. (*) 

- Böckli, M., Bieler, C., Rickenmann, D., Heimann, F.U.M., Badoux, A. (2015d). Hasliaare 

– Geschiebe-transport-Simulationen mit sedFlow. Eidg. Forschungsanstalt WSL, Birmensdorf, 

Schweiz. (*) 

- Böckli, M., Rickenmann, D., Heimann, F.U.M., Bieler, C., Burkhard, L., Badoux, A. (2015e). 

Kleine Emme – Geschiebetransport-Simulationen mit sedFlow. Eidg. Forschungsanstalt WSL, 

Birmensdorf, Schweiz. (*) 

- Bunte, K.B., Abt, S.R., Swingle, K.W., Cenderelli, D.A., Schneider, J.M. (2013). Critical 

Shields values in coarse-bedded steep streams. Water Resources Research 49: 1–21. 

- Carvalho, F., Schulte, L. (2013). Morphological control on sedimentation rates and patterns 

of delta floodplains in the Swiss Alps. Geomorphology 198: 163-176. 

- Ferguson, R. (2007). Flow resistance equations for gravel- and boulder-bed streams. 

Water Resources Research, 43, W05427, doi: 10.1029/2006WR005422. 

- Heimann, F.U.M. (2014). sedFlow, User manual. Swiss Federal Research Institute WSL, 

June 2014, Version 1.00. 

- Heimann, F.U.M., Rickenmann, D., Turowski, J.M. (2015a). sedFlow – a tool for simulating 

fractional bedload transport and longitudinal profile evolution in mountain streams. 

Earth Surface Dynamics 3: 15-34. 

- Heimann, F.U.M., Rickenmann, D., Böckli, M., Badoux, A., Turowski, J.M., Kirchner, J.W. 

(2015b). Calculation of bedload transport in Swiss mountain rivers using the model sedFlow: 

proof of concept. Earth Surface Dynamics 3: 35–54. 

- Nitsche, M., Rickenmann, D., Turowski, J.M., Badoux, A., Kirchner, J.W. (2011). Evaluation 

of bedload transport predictions using flow resistance equations to account for macroroughness 

in steep mountain streams. Water Resources Research, 47: W08513, doi: 

10.1029/2011wr010645. 

- Rickenmann, D. (2001). Comparison of bed load transport in torrent and gravel bed 

streams. Water Resources Research, 37, 3295–3305. 


- Rickenmann, D., Koschni, A. (2010). Sediment loads due to fluvial transport and debris 

flows during the 2005 flood events in Switzerland. Hydrological Processes, 24: 993–1007. 

doi: 10.1002/hyp.7536. 

- Rickenmann, D., Recking, A. (2011). Evaluation of flow resistance in gravel-bed rivers 

through a large field data set. Water Resources Research, 47, W07538, 

doi:10.1029/2010WR009793. 

- Rickenmann D., Heimann F., Böckli M., Turowski J.M., Bieler C., Badoux A. (2014a). 

Geschiebetransport-Simulationen mit sedFlow in zwei Gebirgsflüssen der Schweiz. 

Wasser Energie Luft 106(3): 187-199. 

- Rickenmann, D., Heimann, F.U.M., Turowski, J.M., Bieler, C., Böckli, M., Badoux, A. 

(2014b). Simulation of bedload transport in the Hasliaare River with increased sediment 

input. In: A.J. Schleiss, G. De Cesare, M.J. Franca, M. Pfister (eds.), River Flow 2014, 

CRC Press, Balkema, pp. 2273-2281 (pdf version). 

- Rickenmann, D., Böckli, M., Heimann, F.U.M., Badoux, A., Turowski, J.M. (2015). 

Das Modell sedFlow und Erfahrungen aus Simulationen des Geschiebetransportes in fünf 

Gebirgsflüssen der Schweiz. Synthesebericht. WSL Berichte, Heft 24, 68p. (*) 

(www.wsl.ch/publikationen/pdf/14594.pdf) 

- Schneider, J.M., Rickenmann, D., Turowski, J.M., Bunte, K., Kirchner, J.W. (2015). 

Applicability of bedload transport models for mixed size sediments in steep streams 

considering macro-roughness. Water Resources Research 51: 5260–5283, 

doi:10.1002/2014WR016417. 

- Wilcock P., Crowe J. (2003). Surface-based Transport Model for Mixed-Size Sediment. 

Journal of Hydraulic Engineering 129(2): 120-128. 

(*)these reports are available as pdf from the website: www.wsl.ch/sedFlow 



Determining future evolution of landslides from the 

past: the historical evolution of shallow landslides in 

the upper Cassarate catchment (Southern Swiss Alps) 

Christian Ambrosi, Dipl.-Geol., Dr., Prof. 2 ; Samuel Arrigo, MSc Geogr. 2 ; Claudio Castelletti, MSc Geol. 2 ; 

Cristian Scapozza, MSc Geogr., Dr. 1 

ABSTRACT 

In mountain environments that were subject to intensive pastoralism, the abandonment 

of alpine pastures and the natural reforestation caused the main modifications in land use. 

These two factors may have had a significant effect on evolution of shallow landslides. In the 

upper Cassarate catchment, the evolution of the number and surface of shallow landslides 

between 1900 and 2014 was studied with diachronical mapping on historical analysis. 

Quantitative analysis of precipitation and antecedent standardized precipitation index (IPAS) 

allowed triggering thresholds of 55 shallow landslide events to be calculated. The good 

statistical fits between the difference of IPAS and, respectively, the minimal value of IPAS and 

the sum of precipitations, allows defining the shallow landslide triggering thresholds. Two 

scenarios of shallow landslide triggering were implemented thanks to numerical modeling 

with the TRIGRS program and allowed a validation of the triggering thresholds determined 

by the historical analysis. A forecasting of future shallow landslides evolution was possible by 

defining 1) potential aggravation zones of observed landslides; 2) potential triggering zones 

of new instabilities. 

KEYWORDS 

digital photogrammetry;historical analysis;numerical modeling;shallow landslide; Swiss Alps 

INTRODUCTION 

In mountain environments subject to intensive pastoralism since the second half of 20th 

century, modifications in land use derived usually from the abandonment of alpine pastures 

and the natural reforestation. These two factors may have had a significant effect on the 

number and surface of shallow landslides. A diachronic mapping was carried out in four 

study sites in order to quantify their evolution in the upper Cassarate catchment (Southern 

Swiss Alps) in the last century. This catchment is particularly affected by deep landslides, 

which cover 43% of its surface. The 160 deep rotational landslides inventoried by the 

photo-interpretative mapping carried out by Castelletti et al. (2014) cover 27% of the entire 

surface and characterize not only the loose materials but also the bedrock until 50–60 m 

depth. These landslides are generally comprised within the perimeter of 8 deep seated gravita- 

1 University of Applied Sciences and Arts of Southern Switzerland (SUPSI), Canobbio, SWITZERLAND, cristian.scapozza@supsi.ch 

2 Institute of Earth Sciences University of Applied Sciences and Arts of Southern Switzerland (SUPSI) 



tional slope deformations (DSGSD, Figure 1), controlled by the geological and structural 

conditions of the valley, in particular by the dip slope conformation of the right side. These 

instabilities occupy the whole slope and often extend to the valley floor by mass transport 

phenomena (debris flows, earth flows, etc.) in ravines (Castelletti et al. 2014). In order to 

manage these effects, the upper part of the watershed was the target of major reforestation 

programs between 1880 and 2000 resulting in a reduction of the total areas affected by 

shallow landslides (Mariotta 2004). However, there are some areas where important slope 

erosion persists, even trends to increase. 

Figure 1: Geographical location and main slope instabilities in the upper Cassarate catchment (DSGSD: Deep Seated Gravitational Slope 

Deformation). Basemap: ©swisstopo. 

Considering the framework described above, the Consortium “Valle del Cassarate e golfo di 

Lugano” (CVC) has mandated the University of Applied Sciences and Arts of Southern 

Switzerland (SUPSI) to evaluate the evolution of shallow landslides and regressive erosion in 

the upper part of the watershed. The aim of this study is to provide essential basic data for an 

assessment of the present situation and an evaluation of the future evolution. The goal is the 

production of numerical scenarios for the evolution of shallow landslides. In this work, the 


identification and mapping of existent or potential slope instabilities (instability mapping; e.g. 

Guzzetti et al. 2012), the analysis of their evolution with time (diachronical mapping), the 

calculation of triggering thresholds based on historical events (e.g. Frattini et al. 2009), and 

the production of maps of potential landslide evolution (numerical modeling) are presented. 

The focus will be put on predicting future landslides by validation of triggering thresholds 

determined by correlation of shallow landslides with precipitation data, for better understand 

the relation between landslide occurrence and land use. 

METHODS 

Diachronical mapping 

In order to map shallow and deep seated landslides, we applied 3D digital stereoscopic 

photogrammetry to analogical and numerical aerial photographs taken between 1950 and 

2012 using the ArcGDS extension of the software ESRI® ArcGIS (Ambrosi and Scapozza 

2015). The stereoscopic vision allows obtaining precise results and collects a large amount of 

data such as perimeters, surfaces and volumes of Quaternary formations and/or landforms. 

The ArcGDS tool make it possible the direct exploitation, visualization and digitization of 

stereoscopic digital linear scanned images (e.g. Digital Image Strips). Combined with 

high-resolution digital elevation models, ArcGDS is a powerful tool, particularly over large 

areas, as well as under forest cover and on very steep slopes (Ambrosi and Scapozza 2015). 

Digital monophotogrammetry (or monoplotting) on ancient oblique non-metric photographs 

make it possible the shallow landslides identification for the period 1923–1950 (Figure 2A), 

moving back in time the 3D digital mapping of several decennia. Monoplotting allows the 

georeferentiation and orthorectification of single oblique unrectified photographs, which are 

related to a high-resolution digital elevation model (DEM; in this work the swissALTI3D, 

©swisstopo). The monoplotting relates each pixel of the photograph to the corresponding real 

world pixel on the DEM (Bozzini et al. 2012; Conedera et al. 2013; Scapozza et al. 2014). By 

means of this technique, it was possible to recuperate spatial data obtained by the georeferentiation 

and orthorectification of photographs of the first half of the 20th century in a digital 

format and in a GIS environment (Figure 2B and C). 

The diachronical mapping was based on the analysis and interpretation of terrestrial oblique 

photographs between 1923 and 1941, and of aerial photographs and orthophotos between 

1950 and 2012. Because the lack of images, the period between 1942 and 1949 could not be 

analysed. This kind of mapping based on historical analysis allowed quantifying the evolution 

of the number and surface of shallow landslides. This approach based on the join analysis of 

both oblique terrestrial and vertical aerial photographs makes it possible in particular a 

differentiation between shallow landslides that: 1) have developed since the beginning of the 

20th century; 2) remained stable across the decades; 3) have stabilized by means of reforestation 

programs. 


Figure 2: Example of mapping on an oblique terrestrial photograph. A) Georeferentiation and orthorectification of a terrestrial oblique 

photograph with the WSL-Monoplotting-tool: example of the Alpe Rompiago study area in 1923. B) Shallow landslides in the Alpe 

Rompiago study area in September 1941. C) The landslides shown in B) reported by the WSL-Monoplotting-tool on a 2012 orthophoto. 

Photographs ©CVC; orthophoto ©swisstopo. 


Calculation of triggering thresholds 

Evolution of shallow landslides was compared to a historical database of all mass movements 

occurred in the upper Cassarate catchment between 1900 and 2014. This database was compiled 

based on information on historical landslide events provided by S. Mariotta (unpublished 

data) and by our own work. This compilation was based on identification and 

verification of historical events reported in written press and in cantonal, municipal and 

consortium archives. The analysis of historical data revealed 62 events of which 58 were 

classified as shallow landslides and 4 as rockslides. Three events were excluded because their 

date and location were uncertain. Based on the statistical analysis of rainfall duration and 

intensity during the 55 shallow landslide events retained, triggering thresholds were 

calculated (e.g. Pedrozzi 2004). 

For every event, a climatic quantitative analysis was performed considering the precipitation 

sum, the precipitation return times and an antecedent standardized precipitation index 

(IPAS), allowing the characterization of the state of moisture/drought of the near surface 

(Seiler et al. 2002). IPAS allows a classification of the ground moisture from extremely wet 

(IPAS ≥ 2.0) to extremely dry (IPAS ≤ -2.0). Normal values are comprised between 1 and -1. 

The regressions were computed considering 1σ- and 2σ-error (the probability coefficient of 

0.68 and 0.95, respectively) and took into account: 1) the difference of IPAS before and after 

the three days prior to the event [∆IPAS]; 2) the minimal value of IPAS [MIN-IPAS] during 

the three days preceding the event; 3) the sum of precipitations [ΣP] of the 3 days prior to 

the analyzed shallow landslide event. 

Numerical modeling 

Numerical modeling of potential instability zones was performed using the TRIGRS program 

(Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability; see Baum et al. 

2008). This model is able to evaluate the effect of rainfall events on the temporal evolution of 

the slope stability considering local geotechnical characteristics and infiltration processes. 

Input data of TRIGRS program concerns topography and hydrography, derived from a DEM 

(in this work the swissALTI3D, ©swisstopo), thickness of loose materials (derived from direct 

observation and geophysical prospecting), water table depth, pressure diffusion below the 

water table, cohesion and friction angle (determined from laboratory analysis). 

This makes it possible to model the rainfall infiltration by the Richards equation as proposed 

by Iverson (2000), and the subsequent groundwater runoff in a spatial matrix defining the 

shallow subsurface. The slope stability is evaluated by the analysis of an undefined slope 

(Taylor 1948). 

RESULTS 

Quantification of variations in relative surface of shallow landslides was carried out based on 

the surface mapped for 1950 (Figure 3). For two study sites (Alpe Rompiago and Alpe 

Pietrarossa), surface variations derived from digital monophotogrammetry analysis show a 

slight increase in the area affected by shallow landslides in the decades before 1950. For the 


second half of 20th century, the data show a slow and gradual decrease in landslide area, 

which is in 2012 less than a half compared to 1950 (47% for Alpe Rompiago and 48% for 

Alpe Pietrarossa). In the two other study sites (Alpe Cottino and Cima di Fojorina) the 

increase in shallow landslides surface since 1923 is significant (+186% for Alpe Cottino and 

+389% for Cima di Fojorina with respect to 1950). And this despite an incomplete set of 

images (landslide surfaces from 1967 to 1977 are missing due to the poor coverage by aerial 

photographs for both sites). For Alpe Cottino, the main increase in landslide area was 

registered between 1983 and 1989 (+463% with respect to 1950), and a subsequent decrease 

until 2001. In the Cima di Fojorina, we registered a significant deterioration before 1983, 

where the total surface of instabilities was of +756% with respect to 1950. 

Figure 3: Evolution of absolute and standardized relative landslide surface and frequency diagram of the historical landslide events 

recorded in the upper Cassarate catchment. Standardisation was carried out by subtracting the mean and then dividing the standard 

deviation for obtaining dimensionless landslide surfaces and then compares the four study sites between them. In grey, data calculated 

by monophotogrammetry. 


The increase observed between 1958 and 1983, particularly, was related to increasing erosion 

zones due to sheep herding. 

Quantitative analysis of precipitation and IPAS allowed triggering thresholds of the 55 shallow 

landslide events occurred in the upper Cassarate catchment between 1900 and 2014 to 

be calculated (Figure 4). The statistical relationship between ∆IPAS and the shallow landslides 

historical dataset is very high, with 91.1% of the events that were predicted by this 

parameter. 

Triggering thresholds based on an exponential relationship between the sum of precipitation 

and the IPAS minimal value provides the highest correlation coefficients. The difference of 

IPAS before and after the three days prior to the event [∆IPAS3] and the minimal value of 

IPAS during the three days preceding the event [MIN-IPAS3] are correlated exponentially, 

with a good statistical fit (R = -0.79, R2 = 0.62, Figure 4A). This relation allows estimating the 

increase of IPAS with time and therefore quantifying the increase of moisture in the ground 

in the days before the landslide event. In addition, ∆IPAS has a good linear relationship with 

the sum of precipitations of the 3 days prior to the analyzed shallow landslide event [ΣP3] 

(R=0.74, R2 = 0.55, Figure 4B). By means of the two previous relationships, it was possible to 

calculate the final regression between ΣP3 and MIN-IPAS3 and then define the shallow 

landslide triggering thresholds for the upper Cassarate catchment (Figure 4C). This last 

relationship was used for a classification of all the days between 01.01.1900 and 31.12.2014 

(Figure 4D), in order to quantify positive and negative predictive values (PPV and NPV 

respectively). PPV of the triggering thresholds are lower than 5%, probably due to the very 

Figure 4: Statistical relationships between precipitations (data of the MeteoSwiss station of Bellinzona), IPAS and the occurrence of 

shallow landslides. A) Relationship between the difference of IPAS before and after the three days prior to the event [∆IPAS3] and the 

minimal value of IPAS during the three days preceding the event [MIN-IPAS3]. B) Relationship between the sum of precipitations of the 

3 days prior to the analyzed shallow landslide event [ΣP3] and ∆IPAS3. C) Relationship between ΣP3 and MIN-IPAS3 based on regressions 

presented in A and B. D) Classification of all the days between 01.01.1900 and 31.12.2014 based on regressions presented in C. 


Figure 5: Safety factor maps obtained for Alpe Pietrarossa by numerical modeling with TRIGRS. Modeling was performed with a sum of 

precipitation of 100 mm in 3 days. According to calculated triggering thresholds, triggering probability is very high for Scenario 1 A and 

very low for Scenario 1B. Under Scenario 1A, we would simulate a situation of probable landslide triggering, whereas under Scenario 1B, 

triggering must to be very limited or even absent. 

limited number of days with landslides (55) with respect to the total number of days analyzed 

(42’003 = 115 years). Despite this point, NPV are higher than 99.96%, indicating that it is 

very unlikely to have a landslide if this was not predicted. 

Two scenarios of shallow landslide triggering were finally implemented in the TRIGRS 

program for numerical modeling of triggering thresholds calculated. Validation of the model 

was performed based on scenarios related to potential conditions of instability (safety factor 

lower than 1) under the same rainfall conditions but with different ground moisture (Figure 5). 

The adjustment of cohesion and friction angle parameters makes it possible the final calib- 


ation of the numerical model. By means of this calibration, it was possible to integrate the 

results of numerical modeling with those of diachronical shallow landslide mapping for producing 

maps of: 1) potential aggravation zones of observed landslides; 2) potential triggering 

zones of new instabilities. The surface of potential unstable zones with a MIN-IPAS value of 

-2 (extremely dry), respectively 3 (extremely wet) is completely different for the same rainfall 

intensity and duration. This result allows a validation of triggering thresholds and confirms 

by numerical modeling that the state of ground moisture before the rainfall event has more 

influence on the landslide triggering that the same duration of rainfall. 

CONCLUSIONS 

Four main conclusions can be drawn from the observations, measurements and numerical 

models carried out at the four study sites: 

1. In the Alpe Rompiago and Alpe Pietrarossa study sites, the diachronical mapping illustrate 

a situation of relatively rapid closure of shallow landslide, produced by the significant 

decrease of intensive pastoralism and, as a consequence, the natural reforestation of these 

areas. 

2. Increase in shallow landslide surface for Alpe Cottino and Cima di Fojorina, on the other 

hand, is related to the geotechnical conditions of the near surface, which favor the 

triggering of new surface instabilities in relation to intense rainfall episodes occurred since 

the 1980s. 

3. Numerical modeling allowed the determination of areas with a potential safety factor 

below 1 to be determined on the basis of calculated triggering thresholds. This makes it 

possible to define two kinds of future instability zones: 1) potential zones that could 

present an increase of landslide area; 2) potential zones without an actual landslide that 

could generate a new instability. 

4. From a methodological point of view, the diachronical mapping highlights the integration 

of several techniques of 2D and 3D digital stereo- and mono-photogrammetry, allowing the 

collection of a lot of information about natural phenomena and also the identification and 

mapping of shallow landslides evolution of during time. The monoplotting technique 

makes it possible, in particular, to going back in time in this kind of mapping of several 

decades with respect to the classical aerial photographs analysis. 


This study was funded by the Ufficio dei pericoli naturali, degli incendi e dei progetti of the 

Canton of Ticino. A special thanks to the two anonymous reviewers for their useful feedback, 

as well as to Filippo Schenker for the English proofreading. 


REFERENCES 

- Ambrosi C., Scapozza C. (2015). Improvements in 3-D digital mapping for geomorphological 

and Quaternary geological cartography. Geographica Helvetica 70: 121-133. 

Baum R.L., Savage W.Z., Godt J.W. (2008). TRIGRS – A Fortran program for Transient 

Rainfall Infiltration and Grid-based Regional Slope-stability analysis, Version 2.0. U.S. 

Geological Survey Open-File Report 2008-1159, 75 pp. 

- Bozzini C., Conedera M., Krebs P. (2012). A new monoplotting tool to extract georeferenced 

vector data and orthorectified raster data from oblique non-metric photographs. International 

Journal of Heritage in the Digital Era 1: 499-518. 

- Castelletti C., Scapozza C., Ambrosi C. (2014). Cartographie de l’évolution des glissements 

de terrain peu profonds grâce à la stéréo- et mono-photogrammétrie digitale dans le haut 

bassin du Cassarate (Val Colla, Tessin). FAN-Agenda 2/2014: 10-13. 

Conedera M., Bozzini C., Scapozza C., Rè L., Ryter U., Krebs P. (2013). Anwendungspotenzial 

des WSL-Monoplotting-Tool im Naturgefahrenmanagement. Schweizerische Zeitschrift für 

Forstwesen 164: 173-180. 

- Frattini P., Crosta G., Sosio G. (2009). Approaches for defining thresholds and return periods 

for rainfall-triggered shallow landslides. Hydrological Processes 23: 1444-1460. 

Guzzetti F., Mondini A.C., Cardinali M., Fiorucci F., Santangelo M., Chang H.-T. (2012). 

Landslide inventory maps: New tools for an old problem. Earth-Science Reviews 112: 42-66.- 

Iverson R.M. (2000). Landslide triggering by rain infiltration. Water Resources Research 36: 

1897-1910. 

- Mariotta S. (2004). Il bacino del Cassarate. Sintesi di 120 anni di interventi forestali volti a 

garantire la sicurezza del territorio. Schweizerische Zeitschrift für Forstwesen 155: 278-285. 

Pedrozzi G. (2004). Triggering of landslides in Canton Ticino (Switzerland) and prediction by 

the rainfall intensity and duration method. Bulletin of Engineering Geology and the 

Environment 63: 281–291. 

- Scapozza C., Lambiel C., Bozzini C., Mari S., Conedera M. (2014). Assessing the rock glacier 

kinematics on three different timescales: a case study from the southern Swiss Alps. 

Earth Surface Processes and Landforms 39: 2056-2069. 

- Seiler R.A., Hayes M., Bressan L. (2002). Using the standardized precipitation index for 

flood risk monitoring. International Journal of Climatology 22: 1365-1376. 

Taylor D.W. (1948). Fundamentals of soil mechanics. New York: Wiley, 700 pp. 



A specially developed side weir in the framework of an 

integral flood protection concept including a hydrological 

monitoring system 

Entwicklung eines Hochwasserentlastungsbauwerkes 

und Installation eines Monitoringsystems im Rahmen 

eines integralen Hochwasserschutzkonzeptes 

Josef Schneider, DI Dr. 1 ; Rudolf Schmidt, DI Dr. 2 ; Matthias Redtenbacher, DI 3 ; Franz Brenner, DI 2 

ABSTRACT 

This contribution describes the findings of studies of a planned overflow structure within the 

integral flood protection concept of municipality Thalgau, located in the province Salzburg, 

Austria. A part of the project will be the installation of a specific structure work in combination 

with a lateral weir in the river Brunnbach. The intention of this construction and the 

side weir is to guarantee that in case of a flood only a part of the discharge will be released 

downstream of the weir into the river. Physical model tests were performed in the laboratory 

of the Institute of Hydraulic Engineering and Water Resources Management of Graz University 

of Technology to develop and optimize these structures. The tests, scaled 1:20 and 

performed regarding Froude's law of similarity, result in an intelligent combination of two 

culverts with an optimized side weir to fulfil the required flow situation. First results 

concerning planned design criteria as well as information regarding the monitoring concept 

complete this contribution. 


Dieser Beitrag beschreibt die Ergebnisse der Untersuchungen eines geplanten Entlastungsbauwerkes 

im Rahmen des integrierten Hochwasserschutzes der Gemeinde Thalgau im Bundesland 

Salzburg/Österreich. Dabei ist u.a. die Errichtung eines Streichwehres in Kombination 

mit einem Regulierbauwerk am Brunnbach vorgesehen. Der Zweck dieses Bauwerkes ist eine 

kontrollierte Abgabe von Wasser aus dem Brunnbach in ein seitlich gelegenes Retentionsbecken 

zu gewährleisten, damit nur eine klar definierte maximale Restwassermenge im 

Brunnbach unterwasserseitig des Wehres abfließen kann. Ein physikalischer Modellversuch 

des Bauwerkes am Institut für Wasserbau und Wasserwirtschaft der TU Graz resultierte in 

einer optimierten Gestaltung dieses Bauwerkes. Mittels des Modelles, das im Maßstab 1:20 

errichtet und nach dem Froude’schen Ähnlichkeitskriterium betrieben wurde, konnten die 

1 Graz University of Technology, Graz, AUSTRIA, schneider@tugraz.at 

2 Wildbach- und Lawinenverbauung 

3 Graz University of Technology 



geforderten Anforderungen an das Bauwerk erfolgreich getestet und somit ein erfolgsversprechender 

Ausführungsvorschlag erarbeitet werden. Erste Ergebnisse hinsichtlich geplanter 

Bemessungsgrundlagen, die in weiterer Folge entwickelt werden sollen, sowie Angaben 

hinsichtlich des vorgesehenen Monitoringkonzeptes runden diesen Beitrag ab. 

KEYWORDS 

flood protection; physical model test; side weir; monitoring; design criterium 

EINFÜHRUNG 

Die im Bundesland Salzburg gelegene Gemeinde Thalgau wird immer wieder von größeren 

Hochwässern heimgesucht. Im Sommer 2002 ereigneten sich drei Hochwasserereignisse an 

der Fuschler Ache sowie deren beiden Zubringern Fischbach und Brunnbach mit dem 

Ergebnis, dass große Flächen des gesamten Tales überflutet wurden. 

Die potenziell überflutete Fläche beträgt ca. 320 ha. Etwa 5000 Einwohner leben in diesem 

hochwassergefährdeten Bereich. Dies entspricht rund 12% des gesamten ständig bewohnten 

Gebietes der Gemeinde Thalgau. Der historische Rückblick zeigt, dass dieses Gebiet schon 

immer gefährdet war und auch in ziemlich regelmäßigen Abständen überflutet wurde. 

Vor allem die für dieses Gebiet typischen lang anhaltenden advektiven Niederschläge, die 

durch kurze konvektive Ereignisse zusätzlich überlagert werden, können kritische Situationen 

hinsichtlich Hochwassergefahr bewirken (Salzburger Landesregierung, 2006). 

Aus diesem Grund wurde für die Gemeinde Thalgau seitens der Österreichischen Bundeswasserbauverwaltung 

(zuständig für den Fischbach) und der Wildbach- und Lawinenverbauung 

(zuständig für Fischbach und Brunnbach) ein integrales Hochwasserschutzkonzept gestartet. 

Dieser integrale Ansatz bedeutet, dass der Schutz gegen Naturgefahren durch den Einsatz 

aktiver und passiver Maßnahmen unter Einbeziehung aller Interessensbeteiligter erreicht 

wird. Der aktive Hochwasserschutz umfasst sowohl technische Maßnahmen im Gewässer als 

auch einzugsgebietsbezogene Schritte wie den gezielten Rückhalt von Wasser in der Fläche 

durch technische Eingriffe. Andererseits kommen weiters passive Schutzmaßnahmen zum 

Einsatz. Dazu zählen die Gefahrenzonenplanung, die Risikovermeidung durch das gezielte 

Bewusstmachen von Gefahren der Beteiligten, sowie eine geeignete und vorausschauende 

Raumplanung (Salzburger Landesregierung, 2006). 

Durch die linearen Verbauungsmaßnahmen an den Bächen in der Vergangenheit in Verbindung 

mit einer starken Siedlungstätigkeit sind großräumige Überflutungsflächen als wesentliche 

Hochwasserabfluss- und Retentionsgebiete in Verlust geraten. Betreffend technischer, 

aktiver Hochwasserschutzmaßnahmen kommt deshalb im vorliegenden Hochwasserschutzkonzept 

der Errichtung von Rückhaltebecken große Bedeutung zu. Im Zuge dessen ist u.a. 

die Errichtung eines Streichwehres in Kombination mit einem Regulierbauwerk am Brunnbach 

vorgesehen. Der Zweck dieses Bauwerkes ist eine kontrollierte Abgabe von Wasser aus 

dem Brunnbach in ein seitlich gelegenes Retentionsbecken (Nebenschluss) zu gewährleisten, 

damit nur eine klar definierte maximale Restwassermenge im Brunnbach unterwasserseitig 

des Wehres abfließen kann. Somit ist dieser Bereich der Gemeinde Thalgau (dem Ortszent- 


um vorgelagertes Industrie- und Gewerbegebiet) vor Hochwasser geschützt. Das Regulierbauwerk 

stellte eine große Herausforderung dar, da ein herkömmliches Wehr vor allem 

hinsichtlich beschränkter räumlicher Rahmenbedingungen in der dort geschützten Landschaft 

keine Option darstellte. Des Weiteren soll das Regulierbauwerk keine beweglichen Teile 

aufweisen, was eine zusätzliche Erschwernis bedeutet. 

Der Brunnbach weist bis zu seiner Mündung in den Fischbach (im Ortszentrum von Thalgau) 

eine Einzugsgebietsfläche von 18,9 km² auf. Die hundertjährliche Hochwasserspitze bei 

einem 6- stündigen Bemessungsregen von 158 mm beträgt 52 m³/s. Diese soll auf eine 

Ausbauwassermenge von 32 m³/s reduziert werden. Dies stellt ein Minimalziel dar, ein 

Freibord konnte dabei aufgrund der vorherrschenden Verhältnisse im Ortszentrum von 

Thalgau, die keinerlei Aufweitungen zulassen, nicht mehr berücksichtigt werden. 

Die insgesamt 4 (3 neu errichtet im Hauptschluss und 1 bestehendes im Nebenschluss, 

welches mit gegenständlichem Streichwehr adaptiert wurde) Rückhaltebecken besitzen ein 

Retentionsvolumen von insgesamt 550.000 m³ inklusive Fließretention bei Überschreiten 

der Stauziele. 

Für die Entwicklung und Optimierung des Entlastungsbauwerkes (z.B. Länge und Form des 

Streichwehres, Art und Abmessungen des Regulierbauwerkes) am Brunnbach wurde ein 

physikalischer Modellversuch durchgeführt. Das Institut für Wasserbau und Wasserwirtschaft 

der Technischen Universität Graz wurde seitens der Wildbach- und Lawinenverbauung, 

Gebietsbauleitung Pongau, Flachgau und Tennengau (WLV) damit beauftragt einen hydraulischen 

Modellversuch des Streichwehres Brunnbach durchzuführen. 

DAS MODELL 

Es wurde ein physikalisches Modell eines Ausschnittes des Brunnbaches im Wasserbaulabor 

der Technischen Universität Graz errichtet (siehe Abbildung 1, Ausgangszustand). Die 

Umfassungsmauer des Modelles wurde klassisch mittels Ziegeln errichtet und die Bachsohle 

aus vermörteltem Grobkies, der naturähnlich skaliert wurde, eingebaut. Die relevanten 

Bauteile wie die Wehrkrone, das Regulierbauwerk und ähnliches wurden aus Kunststoff 

hergestellt. Das hydraulische Modell wurde nach dem Froude’schen Ähnlichkeitskriterium 

betrieben. Froude’sche Ähnlichkeit bedeutet, dass das Verhältnis der Trägheits- und Schwerekräfte 

in der Natur und im Modell gleich ist. Das Modell für das Streichwehr wurde nach den 

vorgegebenen Profilen des Auftraggebers mit fester Flusssohle im Maßstab 1:20 ohne Überhöhung 

aufgebaut. Es entspricht einer Naturlänge von etwa 128 m. Im Labor hat das Modell 

eine Länge von etwa 6,4 m, eine max. Breite von 2 m und ist 0,8 m hoch (Zenz et al., 2014). 

Die Aufgabenstellung umfasst die Sicherstellung eines maximalen Abflusses im bachab vom 

Streichwehr gelegenen Abschnitt des Brunnbaches von 18 m³/s bei einer zufließenden 

Wassermenge von bis zu 60 m³/s. Der Zufluss beinhaltet eine gewisse Sicherheit und ist daher 

etwas größer als das HQ 100 

. Die geforderte Ausbauwassermenge von 32 m³/s im Ortszentrum 

von Thalgau ergibt sich aus dem Abfluss des Streichwehres sowie weiteren Zuflüssen 


unterwasserseitig des Bauwerkes über eine Strecke von etwa 2 km. Der seitliche Abwurf 

der überschüssigen Wassermenge von 42 m³/s bei Maximalabfluss hat über das untersuchte 

Streichwehr in ein Rückhaltebecken zu erfolgen. Zusammenfassend kann man folgende 

limitierenden Faktoren für den Modellversuch anführen: 

– Maximaler Zufluss in das Modell: 60 m³/s 

– Beginnende seitliche Hochwasserentlastung bei einem Zufluss von 16-18m³/s 

– Maximale verbleibende Restwassermenge im Brunnbach: 18m³/s 

– Keine beweglichen Teile/Verschlüsse 

Abbildung 1: Physikalisches Modell im Maßstab 1:20, Ausgangszustand, trocken, schematische Darstellung der limitierenden Zu- und 

Abflüsse 

Vor allem die Vorgabe, dass keine beweglichen Teile bei instationären Zuflussbedingungen 

Verwendung finden sollen, stellte eine gewisse Herausforderung dar. Eine Steuerung des 


Abflusses mit Hilfe eines beweglichen Verschlusses wäre trivial gewesen. Ausgehend vom 

Ausgangszustand, der eine schräg angeströmte Tauchwand bei einem relativ kurzen Streichwehr 

abbildete (siehe auch Abbildung 1), wurde in weiterer Folge eine Vielzahl an Varianten 

getestet (Zenz et al., 2014). 

ERGEBNISSE 

Schlussendlich wurde folgende Lösung gefunden. Ein Regulierbauwerk, bestehend aus zwei 

hintereinander, normal auf die Hauptfließrichtung angeordneten Durchlässen in Kombination 

mit einem 46 m langen Streichwehr erfüllt die Vorgaben. Der Abstand zwischen den 

beiden Durchlässen beträgt 5,1 m, die Öffnungsweite des oberwasserseitigen Durchlasses ist 

3 m breit und 1,96 m hoch, des unterwasserseitig gelegenen Durchlasses 3,48 m breit und 

1,26 m hoch. Die beiden hintereinander angeordneten Durchlässe bewirken, dass sowohl der 

erste wie auch der zweite Durchlass eingestaut werden. Der Zufluss zu den Durchlässen 

erfolgt strömend, innerhalb der Kammer tritt eine hochturbulente Abflusssituation ein und 

nach der zweiten Öffnung verlässt das Wasser den Bereich im schießenden Strömungszustand. 

Das bedeutet, dass durch den Aufstau in der „Kammer“ (Bereich zwischen den beiden 

Durchlässen), die Spiegeldifferenz oberwasserseitig des Durchlasses 1 zur „Kammer“ reduziert 

wird und daher eine Drosselung des ersten Durchlasses auftritt. Die Kombination dieser 

beiden Durchlässe mit der dazwischen angeordneten „Kammer“ ist von entscheidender 

Bedeutung für die Funktionsfähigkeit des Bauwerkes. In Abbildung 2 wird der maximale 

Abfluss im Modell fotografisch dargestellt. 

Abbildung 2: Ausführungsvorschlag des Streichwehres und Regulierorganes bei maximalem Zufluss 

Wie diese spezielle Kombination des Streichwehres mit dem Regulierbauwerk wirkt, ist näher 

in Abbildung 3 nachvollziehbar. Die horizontale blaue Linie symbolisiert die Höhe der 

Streichwehrkrone. Die grüne Linie kennzeichnet den Abfluss im Brunnbach, gemessen direkt 

unterhalb des Regulierbauwerkes. Die Zunahme des Wasserstandes im Bereich des Wehres 

(gemessen direkt oberwasserseitig des Regulierbauwerkes) ist durch die rote Linie dargestellt. 

Auf der Abszisse wird die in das Modell zufließende Wassermenge dargestellt. Bei zunehmendem 

Zufluss in das Modell wird die gesamte Wassermenge bis zu etwa 16 m³/s durch die 


Abbildung 3: Ergebnisse des Ausführungsvorschlages 

Öffnungen des Regulierbauwerkes ohne einen Rückstau durchgeleitet. Ab 16 m³/s beginnt 

der Einstau des Regulierorgans, das heisst, dass ab diesem Zeitpunkt Wasser über das Streichwehr 

entlastet wird. Damit steigt naturgemäß auch der Wasserstand im Bereich des Streichwehres 

nicht mehr so stark an (Abflachen des Gradienten beim Knick). Bei der weiteren 

Zunahme des Zuflusses konnte eine leichte Abnahme des Abflusses durch das Regulierbauwerk 

beobachtet werden, was durch die auftretenden Verluste in den Durchlässen erklärbar 

ist. Ab einem Zufluss von etwa 19 m³/s steigt der Abfluss im unterwasserseitig gelegenen 

Bereich des Brunnbaches wieder. Bei der maximal zufließenden Wassermenge von 60 m³/s 

kann ein Ausfluss durch das Regulierorgan von 18 m³/s und somit eine Entlastung von 

42m³/s erreicht werden. Der Wasserstand direkt oberwasserseitig des Regulierorganes liegt bei 

580,4 m über Meeresniveau, es werden keine unerwünschten Ausuferungen in diesem 

Bereich beobachtet. 

Als nächster Schritt werden aufbauend auf diese Versuche Bemessungsrichtlinien für ähnlich 

gelagerte Probleme erarbeitet. In diesem Beitrag sollen die ersten vorläufigen Untersuchungen 

vorgestellt werden, die Werte sind qualitativ zu verstehen. In nächster Zeit werden 

sowohl standardisierte Laborversuche als auch numerische Untersuchungen durchgeführt, 

um hier gesicherte Grundlagen zu schaffen. 

Schritt 1 ist die Erstellung eines numerischen 1D Modells (HEC-RAS), um mögliche Variationen 

der Geometrie zu untersuchen. Das numerische Modell wurde analog zum physikali- 


schen Modellversuch aufgesetzt und auf Basis der Messungen im Versuch kalibriert. In Abbildung 

4 ist ein Längsschnitt mit den maßgebenden, kalibrierten Wasserständen dargestellt. 

Es sind der Oberwasserbereich mit dem Streichwehr und den beiden Durchlässen sowie der 

Unterwasserbereich dargestellt. Die beiden abgebildeten Wasserspiegellinien stellen sich bei 

den Zuflüssen „18 m³/s“ bzw. „60 m³/s“ ein. Die gelben Quadrate symbolisieren die Wasserstands-Punktmessungen 

im physikalischen Modell bei einem Zufluss von 60 m³/s. 

Auf diesen Abfluss wurde somit das numerische Modell kalibriert. Auf Basis dieses Modelles 

wurden unterschiedliche Variationen berechnet. So wurde einerseits die Länge des Wehres 

adaptiert, aber auch die jeweiligen Öffnungen in den Durchlässen verändert. Damit konnte 

das Verhalten des Abflusses durch die Änderung der hydraulischen Komponenten beurteilt 

werden. 

Abbildung 4: Längsschnitt durch das numerische 1D Modell, kalibrierte Wasserstände 

In Abbildung 5 ist beispielhaft ein Ergebnis solch einer Berechnung dargestellt. Es wurden 

hierfür beide Durchlassöffnungen prozentuell verkleinert, alle anderen Randbedingungen 

jedoch gleich gelassen. Die 100% Öffnung entspricht somit dem Ausführungsvorschlag, wie 

er oben beschrieben wurde. Die grüne horizontale Linie symbolisiert den Zufluss zum Modell, 

der bei allen Varianten konstant bei 60 m³/s (100%) gehalten wurde. Eine Verkleinerung der 

Öffnung bewirkt einerseits eine Reduktion des Abflusses im Brunnbach unterwasserseitig des 

Regulierorganes (violette Linie), andererseits naturgemäß eine Erhöhung der seitlich 

abgeworfenen Wassersmenge (blaue Linie) mit einer einhergehenden Erhöhung des 

Wasserstandes direkt oberhalb des Regulierorganes (rote Linie). Beide Abflüsse müssen in 

Summe dem Gesamtzufluss entsprechen. 

MONITORINGSYSTEM 

Der Bau und die Instandhaltung von Rückhaltebecken sind aus dem modernen Hochwassermanagement 

nicht mehr wegzudenken und halten zunehmend auch Einzug bei kleineren 


Abbildung 5: Änderung der Querschnittsfläche der Durchlässe, Auswirkung auf Abflüsse und Wasserstände 

Einzugsgebieten. Eine besondere Herausforderung stellt dabei die Minimierung des Betriebsrisikos 

und die Reduzierung der Betriebskosten dar. Automatisierte Monitoringsysteme 

können dabei von großem Nutzen sein. Abbildung 6 zeigt die in diesem Projekt eingesetzte 

mobile Monitoringanlage MOSES (Mobiles Sicherheits-Einsatzsystem), bestehend aus einem 

Datenlogger, Energieversorgung (Solar- und Batterie) und Übertragungseinrichtungen (GSM 

Modem). Es werden mittels Radar Geschwindigkeit und Abfluss gemessen. Zusätzlich 

Optische Informationen mittels Kamera (Standbilder alle drei Stunden, ab Überschreiten 

eines Schwellenwertes alle 5 min) und Infrarotscheinwerfer. 

Retentionsbauwerke stellen in Österreich Schlüsselbauwerke gemäß ONR Richtlinien 24800 

(ONR, 2008a) und 24803 (ONR, 2008b) dar. Sie müssen in jährlichen Intervallen auf ihre 

Funktionstüchtigkeit kontrolliert werden. Hydrologische Messsysteme können die operationale 

Arbeit erleichtern und die Sicherheit der Anlage erhöhen. Mit seinen im Endausbau 

7 großen Rückhaltebecken und einem Retentionsvolumen von etwa 1 Mio. m³ stellt das 

Projekt Thalgau diesbezüglich ein Pilotprojekt dar. Noch im Sommer 2015 werden die ersten 

drei Becken mit automatisierten Pegel- und Abflussmessungen ausgerüstet. Das Zusammenspiel 

der einzelnen Becken in Abhängigkeit von Niederschlagsereignissen soll Aufschlüsse 

über Projekterfolg und Optimierungsbedarf sowie Erkenntnisse für zukünftige Projektierungen 

liefern. 


Abbildung 6: Mobiles Sicherheits-Einsatzsystem MOSES zur Baustellenabsicherung am Brunnbach 

FAZIT UND AUSBLICK 

Dieser Beitrag zeigt ein Beispiel physikalischer Modelluntersuchungen für eine technische 

Lösung einer Hochwasserschutzmaßnahme im Rahmen eines integralen Hochwasserschutzkonzeptes 

sowie dessen zeitgemäßes Monitoringsystem. 

Die besondere Herausforderung in diesem Falle bestand, dass vorgegeben war, keine 

beweglichen Teile anzuwenden. Die Kombination eines Streichwehres mit einem Regulierbauwerk, 

das aus zwei normal auf die Hauptströmung angeordneten Durchlässen besteht, 

stellt eine äußerst zufriedenstellende Ausführungsform dar. Nur die besondere Anordnung 

dieser beiden Durchlässe, die sich gegenseitig beeinflussen, ermöglicht die Erfüllung der 

Vorgaben. Somit konnte für den Hochwasserschutz des unterliegenden Gebietes eine sehr 

gute Lösung gefunden werden. Hinsichtlich Schwemmholzeintrag in den Bereich des 

Bauwerkes kann durch bereits oberwasserseitig angeordnete Auswurfmöglichkeiten davon 

ausgegangen werden, dass keine Verklausungsgefahr besteht. Eine äußerst unwahrscheinliche 

Verklausung hätte aber auch keine schwerwiegenden Auswirkungen zur Folge, da dabei 

beim zu schützenden Bereich in der Gemeinde Thalgau mit noch geringerem Abfluss zu 

rechnen wäre und das Rückhaltebecken sehr groß und mit einer ausreichenden Entlastungsvorrichtung 

versehen ist. Ebenfalls ist mit überschaubaren Sedimentmengen in diesem 

Bereich zu rechnen, da oberwasserseitig ausreichende Rückhaltemaßnahmen vorgesehen 

sind. 


Daraus resultierend sollen Bemessungsrichtlinien für ähnlich gelagerte Fälle entwickelt 

werden. Es wurden erste qualitative Ergebnisse erzielt. Dafür ist ein mit Hilfe der Modellversuche 

kalibriertes 1D numerisches Modell entwickelt worden, um Variationen der 

Geometrie beurteilen zu können. Sowohl die Verkürzung des Streichwehres als auch die 

Verringerung der Querschnittsöffnungen bedingen demensprechende Reaktionen des 

Abflussbildes. Hier sind weiterführende Grundlagenversuche an physikalischen Modellen 

notwendig, um vor allem darauf basierend numerische Modelle kalibrieren zu können, die 

der Entstehung von Bemessungsrichtlinien zugrunde liegen sollen. 

LITERATUR 

- ONR (2008a). Schutzbauwerke der Wildbachverbauung – Begriffe und ihre Definitionen 

und Klassifizierung 

- ONR (2008b). Schutzbauwerke der Wildbachverbauung – Betrieb, Überwachung und 

Instandhaltung 

- Salzburger Landesregierung (2006). Integraler Hochwasserschutz Thalgau – ein innovatives 

Schutzkonzept für die Fluss- und Wildbacheinzugsgebiete in der Marktgemeinde Thalgau / 

Land Salzburg, Amt der Salzburger Landesregierung, Fachabteilung 6/6, Wasserwirtschaft, 

Bundeswasserbauverwaltung. 

- Zenz G., Schneider J., Redtenbacher M., Lazar F. (2014). Hydraulischer Modellversuch – 

Streichwehr Brunnbach, nicht publizierter Schlussbericht. 



Powder Snow Avalanche Engineering: New Methods to 

Calculate Air-Blast Pressures for Hazard Mapping 

Lukas Stoffel 1 ; Stefan Margreth 2 ; Mark Schaer 2 ; Marc Christen 2 ; Yves Bühler 2 ; Perry Bartelt 2 

ABSTRACT 

Powder snow avalanches are a common hazard in high alpine regions. Steep tracks and cold 

snow temperatures facilitate the formation of the powder suspension cloud developing from 

the dense, fast-moving avalanche core. Considerable damage to buildings and power lines is 

possible even when the dense core of the avalanche stops before reaching the infrastructure. 

An efficient simulation tool that calculates powder pressures in three-dimensional terrain 

would help engineers plan mitigation measures. We present a novel two-layer powder 

avalanche model that couples the cloud to the flowing core, allowing the simulation of mixed 

flowing/powder avalanches. The model predicts endangered areas outside the reach of the 

dense core. We apply this model to a well-documented case study in southern Switzerland 

and discuss the potential for future engineering applications. 

KEYWORDS 

avalanche;avalanche dynamics;Modeling;Hazard Assessment. 

INTRODUCTION: HAZARD MAPPING AND POWDER SNOW AVALANCHE DYNAMICS 

The yellow hazard zone was introduced into Swiss snow avalanche hazard mapping procedures 

to account for powder avalanches with a 300 year return period and pressures smaller 

than 3 kPa (RL, 1984). The yellow zone is placed between the blue and white hazard zones 

and allows engineers to consider destructive pressures beyond the reach of the dense flowing 

avalanche core. For example, in the yellow zone, exposed building components must be 

dimensioned to withstand the wind-like blast arising from the powder cloud. Although 

powder avalanche pressures are small in comparison to the pressure of the dense avalanche 

core, they can work over the entire height of a tall structure, leading to significant overturning 

moments and failure. Power line cables are especially vulnerable to the air-blast which 

can reach significant heights above the ground. Hazard engineers use the yellow zone to 

signal that there is danger for persons outside buildings that might be hit by hard snow 

granules, woody debris or rocks that are transported by the powder cloud. 

Numerical dense flow avalanche models are now in widespread application throughout the 

world to predict impact pressures of the flowing core, e.g. Christen (2010) and Sampl and 

Zwinger (2004). Flowing avalanche pressures primarily define the red and blue danger zone. 

If the hazard engineer, however, expects extreme powder snow avalanche activity, there are 

1 WSL Institute for Snow and Avalanche Research SLF, Davos Dorf, SWITZERLAND, stoffell@slf.ch 




few practical (and computationally efficient) models that can be employed to predict powder 

avalanche air-blast pressures. 

Powder avalanche engineering/modeling is especially difficult because many questions center 

around the influence of terrain features. As the momentum of the powder cloud is acquired 

from the avalanche core, terrain features in the transition zone are of great importance. 

For example, steep terrain segments in the transition zone serve to accelerate the avalanche 

core and therefore provide sufficient forward inertia to the cloud. Steep terrain is therefore 

the first indicator of powder cloud formation and that the air-blast calculations are required. 

Terrain features that divert the direction of the avalanche core are also considered, especially 

when defining the lateral dispersion of the cloud (Bozhinskiy and Losev, 1998; Bühler et al., 

2011). The momentum imparted to the cloud will carry it forward in the direction of the 

core. As the cloud is less sensitive to ground features at the avalanche base, terrain features 

serve to separate the core from the cloud. Significant powder avalanche pressures can therefore 

be generated in the transition and runout zone at the lateral edges of the avalanche track 

– in both the transition and runout zones (Grigoryan et al., 1982). The hazard engineer 

therefore must identify terrain feature that potentially cause a flow separation. This is very 

difficult to do without models that consider both the motion of avalanche core and powder 

cloud in three-dimensional terrain. 

Snow climatology, avalanche volume, amount of snow entrainment and track elevation 

should also be considered when deciding if the powder avalanche pressures are relevant. 

Cold snow temperatures facilitate powder avalanche formation. Dry-cold-snow granules 

produce avalanche cores that fluidize and are less dense, resulting in faster flows. This allows 

the intake of air into the avalanche. The air becomes loaded with ice dust and is eventually 

blown-out of the avalanche to create the powder cloud (Bartelt et al., 2015). The temperature 

of the avalanche core is in large part determined by the temperature of the entrained snow 

(Vera Valero et al., 2015). The snowcover is, in general, colder at higher elevations. At warm 

temperatures, the liquid water film at the surface of snow crystals assists the bonding of stray 

crystals and the formation of large, heavy snow clumps (Bartelt and McArdell, 2011). 

Therefore, at colder temperatures it is easier to produce powder avalanches. Steep avalanche 

tracks with large release volumes and with potential for snow entrainment are therefore 

primary candidates for powder snow avalanches in cold climates and high elevations. 

In this paper we present a new two-layer powder avalanche model that facilitates the 

calculations of air-blast pressures. A salient feature of the model is the consideration of both 

the avalanche core and the suspension cloud (Bartelt et al., 2015). The model is threedimensional 

and accounts for snow entrainment. We apply the model to simulate the 

well-documented 1999 All’Acqua avalanche to demonstrate how air-blast pressures can 

be calculated outside the domain of the dense avalanche core. 


METHODS : TWO-LAYER MIXED FLOWING/POWDER AVALANCHE DYNAMICS MODELS 

To model powder snow avalanches, we apply the two-layer model of Bartelt et al. (2015) 

which has been implemented in the research version of the RAMMS (Christen et al., 2010) 

avalanche dynamics program. Two-layer models consist of an avalanche core Φ and a powder 

cloud Π (Figure 1). Two-layer models were first proposed by Russian scientists in the former 

Soviet Union (Bozhinskiy and Losev, 1998). They have also been applied to calculate powder 

avalanche pressures in the European alps (Naaim and Gurer, 1998; Issler 1998). The cloud Π 

is treated as an inertial flow arising from avalanche core Φ. The yellow zone is therefore 

defined by the attenuation of the powder cloud velocity after it detaches from the core. The 

density of the core is not constant – allowing dilute, disperse and dense flow densities. At the 

avalanche front, the core is dilute (“saltation layer”), allowing the intake of air Λ. This air is 

mixed with ice dust and blown out to create the powder cloud. For more details about the 

avalanche model, see Bartelt et al. (2015). 

65 out to create the powder cloud. For more details about the avalanche model, see Bartelt et al. 

65 

66 

out to create the powder cloud. For more details about the avalanche model, see Bartelt et al. 

(2015). 

66 (2015). 65 out to create the powder cloud. For more details about the avalanche model, see Bartelt et al. 

67 

Figure 1: The structure 1 66 (2015). 

odel, see Bartelt et al. of a mixed flowing/powder avalanche. The avalanche contains a dilute core Φ (often termed the “saltation 

67 layer”) Figure and the 1 powder cloud Π. The cloud can decouple from the core and move independently. Snow Σ and air Λ are entrained by the 

68 avalanche. The model state 67 variables Figure 1 for the core are, 

68 The model state variables for the core are, 

68 The model state variables for the core are, 

The model state variables for the core are, 

( ) T 

U Φ 

M = 

Φ 

, M 

ΦuΦ 

, M 

ΦvΦ 

, RhΦ 

, hΦ 

, M 

ΦwΦ 

, N 69 ( M , M u , M v 

Rh , h , M w , N ) K 

( T ) T 

Φ 

M 

Φ 

, M 

ΦuΦ 

, M 

ΦvΦ 

, RhΦ 

, hΦ 

, M 

ΦwΦ 

, N 

Φ Φ Φ Φ Φ 

Φ Φ Φ Φ K (1) 

K 

(1) 

and (1) 

70 for the cloud, 70 

70 

and for the cloud, 

71 U = ( M 

(2) 

( M 

u 

M 

( M 

v 

M u 

h ) T Π Π 

M Π Π 

M u Π Π 

M v Π 

h ) M v ) T 

Π Π Π Π Π Π 

hΠ 

(2) 

(2) 

T 

Π Π Π Π Π Π Π 

The slope parallel The slope parallel flow velocity of the core is given by the two-dimensional vector 

flow velocity of the core is given by the two-dimensional vector 

( u 

The slope parallel flow velocity of the core is given by the two-dimensional vector ( ) T 

Φ Φ 

vΦ 

72 u v ) T 

72 similarly for the of the avalanche core 

The slope parallel flow velocity of the core is given by the two-dimensional vector ( al vector ( u v ) 

u v ) T 

in the slope perpendicular direction is w Φ . This be the initial blow-out 

Φ = Φ 

, 

Φ 

Φ Φ Φ 

73 similarly for the powder cloud, u ( ) T 

Π 

= uΠ 

, vΠ 

. The velocity of the avalanche core in the slope 

Φ Φ Φ 

anche 73 core similarly in the slope for the 74 powder perpendicular cloud, direction u = is( uw . v ) T 

73 velocity similarly of for the the plume powder structures cloud, of Π 

the Π 

powder 

( This velocity 

u Π 

. The can velocity be considered of the the avalanche initial blow-out core velocity in the of slope the 

Φ 

v cloud. 

) T This upward velocity should be 

blow-out velocity of the 75 plume structures of Π the powder Π cloud. Π 

. The This upward velocity velocity of the should avalanche be carefully core studied in the to predict slope 

74 perpendicular direction is w . This velocity can be considered the initial blow-out velocity of the 

Φ 

perpendicular direction is w . This velocity can be considered the initial blow-out velocity of the 

Φ 

plume structures of the powder cloud. This upward velocity should be carefully studied to predict 

76 forces acting on powder cables suspended well-above the avalanche core. Slope-perpendicular 

arefully 74 studied to predict 


75 

77 velocities are generated when the free mechanical energy R of the avalanche is deflected at the 

e. Slope-perpendicular

74 

75 

76 

77 

78 

79 

80 

81 

82 

83 

84 

85 

86 

87 

88 

89 

90 

perpendicular 74 perpendicular direction direction is w . This w velocity . This can velocity be considered can be considered the initial the blow-out initial blow-out velocity of velocity the of the 

Φ 

Φ 

plume 75 structures plume structures of the powder of the cloud. powder This cloud. upward This velocity upward should velocity be should carefully be carefully studied to studied predict to predic 

forces 76 acting forces on acting powder on cables powder suspended cables suspended well-above well-above the avalanche the avalanche core. Slope-perpendicular 

core. Slope-perpendicular 

velocities 77 velocities are generated are generated when the when free mechanical the free mechanical energy R energy of the R avalanche of the avalanche is deflected is deflected at the at the 

basal 78 boundary, basal boundary, creating creating the dispersive the dispersive pressure pressure N (Buser N and (Buser Bartelt, and 2015). Bartelt, The 2015). dispersive The dispersive 

carefully studied to predict forces acting on powder K cables K suspended well-above the 

pressure 79 avalanche pressure is associated is core. associated with Slope-perpendicular the with inelastic the inelastic scattering velocities scattering of are snow generated granules of snow when granules and the fluidization free and mechanical fluidization of the core of the core 

energy R of the avalanche is deflected at the basal boundary, creating the dispersive pressure 

and 80 therefore and therefore formation the formation of so-called of so-called “saltation “saltation layers”. layers”. Core and Core cloud and masses cloud per masses unit area per unit area 

N K (Buser and Bartelt, 2015). The dispersive pressure is associated with the inelastic scattering 

are 81 denoted of are snow denoted M granules and 

and and 

Φ fluidization M , respectively. of the core Flow and heights therefore measured the formation from the of so-called basal boundary for the 

Φ 

M , respectively. Flow heights measured from the basal boundary for the 

Π Π 

“saltation layers”. Core and cloud masses per unit area are denoted M Φ and M Π , respectively. 

core 82 and core the cloud and the are cloud designated are designated h and 

and 

Φ 

h . 

Φ 

h . 

Π Π 

Flow heights measured from the basal boundary for the core and the cloud are designated 

h Φ and h Π . 

The 83 motion The of motion the mixed of of the the avalanche mixed avalanche is found is is by found found solving by by solving the solving differential the the differential vector vector equations: vector equations: equations: 

84 

∂U 

U ∂ 

Φ 

Φ 

y 

x + = G 

Φ 

t 

x ∂y 

and and U ∂ 

Π 

Π 

Φ 

∂Φ 

∂Φ 

x y 

85 + 

x y 

+ = G 

Φ 

+ = G (3) 

∂t 

∂x 

∂y 

and ∂U 

Π 

∂Π 

∂Π 

x y 

+ + = G (3) 

∂t 

∂ 

xt 

∂ 

yx 

∂ 

Π 

y (3) 

Π 

86 

for 87 the core for for the the and core core cloud and and layers, cloud cloud layers, respectively. layers, respectively. respectively. The flux The The components flux flux components components for the for core the for the core 

( Φ core (Φ x, Φ( 

) y 

Φ T ) T ) T 

cloud (Π x, Π y ) T x 

, Φ y 

and 

x 

, Φ y 

and 

are defined in Bartelt et al. (2015) and are not of interest here. Of greater 

cloud 88 ( Π practical cloud () Π T interest ) are T the components of the vectors G Φ G Π describing the friction and 

x 

, Π y 

are 

x 

, Πdefined y 

are defined Bartelt in et Bartelt al. (2015) et al. and (2015) are not and of are interest not of here. interest Of here. greater Of greater 

entrainment processes acting on the core Φ and cloud Π. These are, 

practical 89 practical interest are interest the components are the components of the vectors of the Gvectors 

and 

and 

Φ 

G describing the friction and 

Φ 

G describing the friction and 

Π Π 

⎛ 

M 

− 

M 

⎞ 

entrainment 90 processes ⎜ acting on the core ⎟ Φ and cloud Π. These are, 

91 

92 

93 

94 

95 

96 

97 

98 

99 

100 

101 

102 

entrainment 

⎛ 

Σ→Φ 

Φ→Π 

M 

⎛ processes M acting on the core Φ and cloud Π. These are, 

−⎛ 

⎞ 

Σ→Φ 

− M 

Φ→Π 

⎜ 

⎞ 

Σ→Φ 

M 

M 

⎞ 

Φ→Π 

⎜ 

⎛ 

⎛ 

⎜ 

⎟Σ→Φ 

− M 

M 

⎞ 

Σ→Φ 

⎛ 

Φ→Π 

⎜ 

⎞ 

Σ→Φ 

Φ→Π ⎟ 

⎜ 

G 

− 

S 

u 

⎟ 

⎟ 

Φ→Π 

⎜⎜G 

⎞ 

Σ→Φ 

− M 

⎟ 

x 

− 

Φ→Π 

⎜ 

⎜Gx 

SΦx 

− M 

Φ→Π 

⎟ 

uΦ 

x ⎜ x 

Φ 

Φ→Π 

u ⎟ 

⎛ 

⎜ 

x x ⎟ Φ 

⎛ 

M 

⎞ 

M 

⎛ 

+ 

Φ→Π 

+ M 

⎞ 

Φ→Π 

Λ→Π 

⎜ 

G 

S 

ΦΦ 

x x 

− 

− − 

S 

MM 

Φ→Π Φ→Π 

x 

Φ 

x 

−⎜ 

GM 

u 

Φ→Π 

− 

⎟u 

Φ Φ ⎟ 

⎜ 

S 

Φu 

Φ− 

M 

Φ→Πu 

⎜ 

x 

⎟ 

⎜ 

⎜ 

G 

− S 

− 

⎜ 

⎟ 

x 

⎟ 

Φ ⎛ M 

Φ 

M 

⎟ 

⎛ M 

⎞ 

Φ→Π 

+ M⎛ 

⎞ 

Φ→Π 

+ M⎛ 

M 

⎞ 

Λ→ΠΦ→Π 

+ M 

M 

⎞ 

Λ→Π Φ→Π 

+ 

Λ→Π 

Λ→Π 

⎜ 

Φ 

− M 

Φ→Π 

⎜ 

⎟ 

y y 

Λ→Π 

⎞ 

+ M 

Λ→Π 

⎜ 

G −⎜ 

S⎜ 

G 

v 

⎜ 

− SΦ 

Φ 

MΦ 

M 

− 

M 

Φ→Πv 

⎜ 

⎟ 

y 

Φ 

⎜ 

G − SΦ 

− M 

Φ→Π y 

⎟ v 

⎜ 

⎟ 

y y 

Φy 

Φ 

y y 

Φ→Πv 

Φ→ΠvΦ 

⎟ 

⎜ 

⎜ 

⎟ 

⎟ 

Φ 

= 

⎜y 

α 

( S 

) 

⎟ ⎟ ⎜ M 

Φ→Π 

⎜ M 

uΦ 

− 

Φ→Πu 

SΠx 

Φ ⎜− 

⎟ 

91 G MS 

ΠΦ→Π 

x ⎟ uΦ 

91 Φ 

= 

G 

⎜ 

91 

Φ 

⋅ u 

G 

Φ 

y 

⋅ 

− 

y 

− 

u 

β 

SΦy 

− M 

Φ 

( 

( R 

= 

) 

−K 

Φ ⎜Φ 

α( S 

β 

⎟ 

Φ 

⎟⋅ 

u 

and Φ→Π 

(4) 

( 

) 

) 

) − 

and 

Φ 

βR 

and 

G = ⎜ M 

⎟ 

G ⎜ 

Φ→Πu 

K 

Πand 

= 

⎜ 

⎜ M 

Φ→Πu 

Φ→Πu 

Φ 

− S 

Φ 

− SΠx 

⎟ 

91 G Πx 

) 

u 

⎟ 

91 G = 

Φ 

= 

S 

Π 

x 

91 

Φ ⎜ α S 

⎜ α S 

Φ 

⋅ u 

Φ 

⋅ u 

Φ 

− βRΦ 

− βRK 

⎟ and G ⎜ G = 

= 

K 

G = 

⋅ − 

Φ 

α S 

u 

⎟ and G = 

Π ⎜ 

⎟ 

(4) 

βRK 

and 

Π ⎜G Π 

= ⎟ 

(4) 

⎜ 

⎟ 

⎟ 

⎜ 

w 

⎜ 

M 

Φ→Πv 

⎜ 

⎟ Φ 

− S 

Π 

(4) Πx 

⎟ 

G 

Φ 

= ⎜ α( S 

Π 

Φ 

− Sy 

(4) (4) 

⎜ Φ 

⋅ u 

Φ 

) − βR 

⎜ 

⎟K 

⎟ and 

w 

⎟ ⎜ 

MG Φ→Π Π v= 

Φ ⎜− 

Sy 

(4) 

⎜ 

w 

⎟ 

⎜ 

M ⎜ 

M 

Φ→Πv 

Φ→Πv 

− 

Φ 

− SyΦ 

Sy 

⎜ 

Φ 

w 

⎜ 

⎟M 

⎟ 

vΦ 

⎟ 

(4) 

⎜⎜ 

Φ 

w ⎜ 

⎟ ⎟ 

Φ 

⎟ 

⎟ 

⎜ 

Φ 

w 

M 

vΦ 

− 

Sy 

⎜ 

Φ 

⎟ 

Φ 

⎟ 

⎟ 

⎜ N 

⎝ VΦ→Π 

+ 

⎜ N 

Λ→Π 

⎜ 

⎟ 

⎝ VΦ→Π 

+ V⎜ 

M 

Φ→ΠvΦ 

− Sy 

⎜ N ⎟ 

⎜ 

⎟ 

⎜ 

⎟ 

⎝ V ⎝ V 

Φ→Π 

+ 

Φ→Π 

+ 

K 

N ⎟ 

⎝ 

⎠ 

⎟ 

Λ→Π 

⎜ 

Φ 

VΦ→Π 

⎠ + V 

K ⎟ 

K 

V 

Λ→Π ⎠ 

⎜⎜ 

⎜ ⎜ 

⎟ 

K 

Λ→Π 

⎝ 2γP 

− 

N 

K 

⎝ VΦ→Π 

+ V⎠ 

Λ→Π ⎠ 

⎜ 

⎝ ⎜ 

2γ 

2 

P 

Nw N 

⎟ 

− 

Φ2 

Nw 

/ h 

⎟ 

⎟ 

⎝ 

⎝ 2γP 

⎝ 

Φ2γ 

/ P 

h⎠ 

⎟ 

⎝ VΦ→Π 

+ VΛ→Π 

⎠ 

K Φ 

Φ− 

2 

− 2Nw 

⎜ 

/ 

h 

Φ 

/ h 

Nw ⎠ Φ 

/ hΦ 

⎠ 

92 ⎜ 

⎟ 

Φ 

Φ 

⎠ 

92 92 

⎟ 

⎝ 2γP 

− 2Nw 

⎠ 

93 The 92 core ⎝ 2γP 

− 2Nw 

of the avalanche Φ 

/ h 

is driven Φ 

/ hby ⎠Φ 

the ⎠ gravitational acceleration in the slope-parallel directions 

92 

93 

The core 93 

of The the avalanche core of the is avalanche driven by is the driven gravitational by the gravitational acceleration acceleration in the slope-parallel in the slope-parallel directions directions 

93 92 The core of the avalanche is driven by the gravitational acceleration in the slope-parallel directions 

9493 

G The core of the avalanche is driven by the gravitational acceleration in the slope-parallel directions 

93 = ( 

The core of the avalanche is driven by the gravitational in the slope-parallel directions 

The core G 

94 

x 

of = 

, G 

94 ( y 

) 

the x 

, 

= ( Gavalanche G 

M 

y is driven by the the slope-parallel 

94 G = ( ( ) = 

Φ 

g 

( MG 

x 

, M 

xΦ, 

Gg 

Φ 

x 

, 

g 

94 G = directions 94 

x 

, G 

y 

= Gravitational is decomposed into three 

() = 

x 

, G( ) ( 

) My 

). ravitational acceleration is decomposed into three 

y 

= 

Φ( Mg 

yΦ 

) 

. g 

ravitational 

x 

, M 

Φ 

g 

y 

My 

= M g 

Φ 

g 

Gx 

, G 

y 

) 

x 

, 

Φ 

x 

, 

= ( 

Φ y 

) ) 

). ravitational acceleration acceleration is decomposed is decomposed into three into three 

. ravitational 

Φ y 

. ravitational acceleration is decomposed into three 

95 gravitational components, 

acceleration is decomposed into three 

95 

gravitational 95 gravitational components, M 

Φ 

g 

x 

, M 

Φ 

g 

y 

g = ( g 

components, x 

= 

, 

g( g 

y 

, 

). ravitational acceleration is decomposed into three 

95 gravitational components, g = 

g 

x 

, 

gz 

). The mass exchanges in the mixed flowing avalanche 

G = ( G 

y 

g , 

g= 

z 

95 

gravitational 

95 gravitational components, g = ( ( )( . g The 

x 

, g 

mass 

y 

, g 

z 

components, 

x 

, g 

y 

, gx 

g = ( 

z 

) ) 

) exchanges . The mass in exchanges the mixed in flowing the mixed avalanche flowing avalanche 

x 

, G 

y 

) = ( M 

Φ 

g 

x 

, M 

Φ 

g 

y 

). ravitational 

. 

y 

, 

The 

z 

. The mass acceleration exchanges in is the decomposed mixed flowing into avalanche three 

mass exchanges in the in mixed flowing avalanche 

96 system are: snow entrainment into the avalanche flowing avalanche 

x 

, g 

y 

, gcore 

z 

. The M mass exchanges in the mixed flowing avalanche 

96 

system 96 

are: system snow entrainment are: snow entrainment into the avalanche into the core avalanche 96 

system 96 

system are: system 

are: snow are: 

snow entrainment snow entrainment 

entrainment into into the into the and mass blow-out of 

avalanche the avalanche Σ→ M 

Φ 

, volume and mass blow-out of ice- 

Σ→Φcore 

, volume M 

Σ→and Φ 

, volume mass blow-out and mass of blow-out ice- of icegravitational 

components, g = ( g core M 

core M 

96 system are: snow entrainment into the avalanche core 

Σ→Φ 

, volume 

Φ 

and mass blow-out of ice- 

M 

Σ→ , volume and mass blow-out of icex 

, g 

y 

, g 

z 

) 

97 dust from the core V 

Φ 

ice-dust 97 

dust from 97 

the dust core 

air Σ→ , volume and mass blow-out of ice- 

97 dust from the Φ→ from Π 

and 

V 

M 

Φ→ 

the 

Πcore 

and Φ→V M 

Π 

and direct air entrainment into the powder cloud 

Φ→ 

Π 

and M 

V 

direct 

Φ→Πair and entrainment direct air entrainment into the 

core V 

into powder into the the cloud Λ→ powder Π 

V 

Λ→ 

Πcloud V 

Λ→Π 

97 

dust 97 

from dust the core 

from the V core 

Φ→Π 

and 

Φ→Π 

and M 

M 

V Φ→Π 

and 

Φ→Π 

and 

Φ→direct Π 

and direct air entrainment into the powder cloud V 

air entrainment into the powder cloud V 

Λ→Π 

98 system (volume) are: and snow M entrainment M 

Φ→Π 

and direct air entrainment into the powder cloud 

Λ→Π 

98 

(volume) 98 

and (volume) Λ→ 

M 

Π 

(mass). into the avalanche core M 

V 

Λ→Π 

Λ→ 

Π 

and 

(mass). M 

Λ→Π 

(mass). 

Σ→Φ 

98 (volume) and M 

98 Central (volume) to the and model M equations is the inclusion of the free mechanical energy R of the 

98 (volume) and 

Λ→Π 

(mass). 

Π 

M Λ→ (mass). 

avalanche core which is Λ→ (mass). 

99 dust Central from to the model core equations V 

Φ→ 

Π 

and is sum the inclusion of Φ→ the Π 

energy of the free of mechanical the random energy granule R of the motions avalanche V 

K 

and 

Λ→ 

99 

Central 99 

to the Central model to equations the model is equations the inclusion is the of inclusion the free mechanical of the free mechanical energy R of energy the avalanche R of the avalanche 

99 Central to the model equations is the inclusion of the free mechanical energy R of the avalanche 

10099 

configurational core Central to (density) model equations changes is the Rinclusion V 

(Buser of and the free Bartelt, mechanical 2015). energy The R total of the free avalanche mechanical 

100 99 

which core Central 

is the which is to 

sum the the 

of sum model 

the energy of equations 

of the energy is 

random 

of the inclusion 

granule random granule of 

motions 

100 core which is the sum of the energy of the the random free motions mechanical granule 

R and configurational 

(volume) and M 

K 

R motions 

and energy configurational R of the avalanche 

K 

R and configurational 

100 energy 

100 

core is which produced 

core which Λ→is Π 

(mass). 

K 

is the sum from 

the sum 

of the of 

energy rate 

the energy of the working 

of the random 

random of granule the 

granule 

total motions shear 

motions 

R S R 

and Φ 

· configurational 

u 

and 

K Φ 

, the 

configurational 

101 (density) model parameter α 

100 core changes which is the sum of the energy of the random granule motions K 

101 

(density) 101 

R (Buser and Bartelt, 2015). The total free mechanical energy changes 

R (Buser and Bartelt, 2015). The total free mechanical R and is produced configurational from 

(density) V changes 

Kenergy is produced from 

V 

R (Buser and Bartelt, 2015). The total free mechanical energy is produced from 

V 

describing 101 (density) the production changes R rate (Buser 0 ≤ and α ≤ Bartelt, 0.20 (Buser 2015). The and total Bartelt, free mechanical 2015). The energy kinetic is produced part of from the 

101 (density) changes R (Buser V 

102 the 

and Bartelt, 2015). The total free mechanical energy is produced from 

101 rate of (density) woring changes of the V total 

fluctuation 102 

the rate 102 energy of woring the Rrate K 

decays of the woring R shear (Buser and Bartelt, 2015). The total free mechanical energy is produced from 

V 

total 

102 the rate of woring of the 

according of shear 

S 

the Φ 

⋅u , the model parameter α describing the production 

total S Φ 

total shear 

the parameter β because of irreversible Φ 

⋅ushear 

, the model parameter α describing the production 

Φ 

S 

Φ 

⋅u , the model parameter α describing 

Φ 

collisional/ 

the production 

102 the rate of woring of the total shear S 

frictional 102 interactions the rate of woring the of the avalanche total shear Φ 

⋅u 

S 

, Φ 

⋅u , the model parameter α describing the production 

the model Φ 

103 rate 0 ≤ parameter α describing the production 

core. 

ΦS 

The parameter β is very dependent on the snow 

103 

rate 103 

α 

0 

≤ 0. 

α 

20 (Buser and Bartelt, 2015). The inetic ≤rate 0. 

200 ≤(Buser α ≤ 0. 

and 20 

Bartelt, (Buser and 2015). Bartelt, Φ The ⋅u , part the model of the parameter fluctuation energy α describing the production 

Φ 

inetic 2015). part The 

103 rate 0 ≤ α ≤ 0. 20 (Buser and Bartelt, 2015). The inetic inetic of the fluctuation 

part of the part of the energy fluctuation 

R decays 

K 

core which is the sum of the energy of the random granule motions fluctuation R energy decays 

K 

R decays 

R K 

103 

energy and configurational 

temperature. rate 0 ≤ α ≤ 0. 20 (Buser and Bartelt, 2015). The inetic part of the fluctuation energy 

K 

103 rate 0 ≤ α ≤ 0. 20 

R 

R decays 

decays K 

104 according to the parameter 104 

according 104 

to according the parameter to (Buser β because 

the parameter and Bartelt, 

of irreversible 

2015). The 

collisionalfrictional 

inetic part of the 

interactions 

fluctuation 

in 

energy 

the K 

β because of β irreversible because of irreversible collisionalfrictional collisionalfrictional interactions interactions the R decays 

K in the 

104 

104 

according according to the parameter β because of irreversible collisionalfrictional interactions in the 

105 (density) avalanche changes to the 104 according 

core. The 

to 

parameter 

the V β because of irreversible collisionalfrictional interactions in the 

105 

avalanche 105 

core. avalanche The parameter β is very 

core. The parameter β is because 

dependent very dependent of β is irreversible 

on the snow 

very on dependent the collisionalfrictional 

temperature. 

snow on temperature. 

the snow temperature. 

interactions in the 

105 avalanche core. The parameter β is very dependent on the snow 

105 avalanche core. The parameter β is very dependent on the snow temperature. 

INTERPRAEVENT temperature. 2016 – Conference Proceedings | 419 

the rate 105 

of avalanche woring core. of the The parameter total shear β is S very dependent on the snow temperature. 

106 

S = S , S we use 

Φ 

⋅u 

a oellmy-type Φ 

ansatz 

The core of the avalanche is driven by the gravitational acceleration in the slope-parallel directions 

. The mass exchanges in the mixed flowing avalanche 

, volume and mass blow-out of ice- 

M and direct air entrainment into the powder cloud Π 

Central to the model equations is the inclusion of the free mechanical energy R of the avalanche 

R (Buser and Bartelt, 2015). The total free mechanical energy is produced from 

For shearing in the core ( ) 

, the model parameter α describing the production

103 

104 

105 

106 

107 

108 

109 

110 

111 

112 

113 

114 

115 

116 

117 

118 

119 

120 

121 

122 

123 

124 

125 

126 

127 

128 

129 

130 

131 

132 

133 

134 

102101 

101 the (density) rate (density) of woring changes changes of R the (Buser 

V 

R total (Buser and shear and Bartelt, Bartelt, 

Φ 

⋅u2015). , the 2015). model The total The parameter total free free mechanical describing energy energy the is production 

produced is produced from from 

V 

Φ 

102 rate the 0 ≤rate α of ≤ woring 0. 20 (Buser of the total and shear Bartelt, S 

Φ 

⋅u2015). , the model The inetic parameter part α describing of the fluctuation the production 

Φ 

energy R decays 

K 

103 102 102 rate the rate the rate of 0. of 20woring (Buser of the and of total the Bartelt, total shear shear 2015). S The inetic part of the fluctuation energy decays 

Φ 

⋅u S , the model parameter α describing the Φ ⋅u , the model parameter α describing the 

K 

production 

Φ 

103 according rate 0 ≤to α the ≤ 0. 20 parameter (Buser and Bartelt, 

β because 2015). The of irreversible inetic part of the collisionalfrictional fluctuation energy R interactions decays 

K 

in the 

104 103 103 according rate rate 0 ≤to 0 α the ≤ 0. α parameter 20 ≤ 0. (Buser 20 (Buser and β because and Bartelt, Bartelt, of 2015). irreversible 2015). The inetic The collisionalfrictional inetic part part of the of the fluctuation interactions energy energy in R the decays 

K 

R decays 

104 according to the parameter β because of irreversible collisionalfrictional interactions in the K 

105 avalanche 104 104 avalanche according according core. core. to the to The parameter the parameter β because β is β very because is very dependent of of dependent irreversible on the snow collisionalfrictional on temperature. the snow temperature. 

interactions the in the 

105 avalanche core. The parameter β is very dependent on the snow temperature. 

105 105 avalanche avalanche core. core. The parameter The parameter β is very β is very dependent on the on snow the snow temperature. 

106 For shearing in the core Φ Φy 

For in Φ 

= ( we use a oellmy-type ansatz 

106 For shearing in the core S 

Φ 

= ( SΦxS 

, ΦS 

Φx 

y, 

) Swe Φy 

use ) 

we a oellmy-type a use Voellmy-type a oellmy-type ansatz ansatz ansatz 

2 

106 106 For shearing For Πshearing the in core the core S 

Φ 

= S 

Φ 

( S= Φ 

2 

x( , S 

Φ 

107 

2 

xy, 

) Swe Φy 

) use we a use a oellmy-type ansatz ansatz 

Φ u ⎡ 

u ⎡ 

V 

ρΦ 

u ⎤ 

. 

u ⎤ 

(5) 

Π 

Φ 

107 S Π 

Π 

ξ Φ 

S 

Φ 

= ⎢µ 

Φ 

V 

2 2 

Φ 

= ⎢µ 

( 

⎡( 

RV 

u RV 

⎡) 

N + ρΦ 

g ⎥ . ρΦ 

u 

⎤ 

⎥ 

⎤ 

(5) 

u Π 

u 

Φ 

u 

Π 

Φ . (5) 

107 107 

108 

S ⎢ 

⎥ 

uΦ 

= S 

Π 

= ⎢ µ ( R⎢ 

ξ 

Φ ⎣ Vµ 

)( 

NR 

V+ 

) Nρ 

Φ+ 

gΦ 

( R 

ρΦ 

. (5) 

Π ⎢ 

gV 

) ⎥ 

⎣u 

⎢ 

( R⎦ 

⎥ . (5) 

V⎥ 

) 

108 

⎥ 

Π 

u 

ξ ⎦ . (5) 

109 The normal pressure ⎣ 

Π ⎢ 

Φ 

( Rξ 

V 

) 

⎣ 

Φ 

( R 

includes the self-weight ⎦ 

V 

) ⎥⎦ 

of the avalanche, centripetal and dispersive 

109 108 108 The normal pressure N includes the self-weight of the avalanche, centripetal and dispersive 

110 pressures. Both the Coulomb friction The 109 109 

normal The normal The normal pressure pressure 

pressure N N includes N includes the the 

the the V self-weight and turbulent of friction ξ 

of the of the the 

of the avalanche, avalanche, V centripetal depend on the 

and and dispersive 

centripetal and dispersive 

110 pressures. Both the Coulomb friction µ ( R V 

) and turbulent friction ξ ( R V 

) depend on the dispersive and dispersive 

111 pressures. 110 110 configurational pressures. pressures. Both Both the energy Both the Coulomb the content Coulomb of friction the friction core µ ( R 

V 

µ , 

V) 

( R 

and see V 

) (Bartelt and turbulent turbulent et friction al., 2015). friction ξ ( R V 

ξ) 

( depend R V 

) depend on the on on the the configurational 

energy content of the core R V 

111 pressures. configurational Both the energy Coulomb content of friction the core R µ ( R, see (Bartelt 

, see V 

and turbulent et al., 2015). 

V 

friction ξ ( R ) 

(Bartelt et al., V 

depend on the 

111 111 

configurational energy energy content content of the of core the core R , see (Bartelt et al., 2015). 

V 

R , see (Bartelt et al., 2015). 

V 

112 configurational Drag rag the the powder energy cloud content Π of Πxthe Πy 

core 

is given by 

R , by a see velocity a (Bartelt squared squared et type al., 2015). law type law 

112 rag the powder cloud S 

Π 

= ( SΠx 

, SΠy 

) is given Vby a velocity squared type law 

112 112 rag rag on the on powder the powder cloud 2 cloud S 

Π 

= S 

Π 

( S= Πx( , S 

Π 

xy, 

) Sis Πygiven ) is given by a by velocity a velocity squared squared type type law law 

u 

⎡ 

u 

⎤ 

Π 

Π 

rag S Π 

= 

on the ⎢ 

powder ρ 

Π 

g 

cloud ⎥ 

. 

S 

Π 

= ( SΠx 

, SΠy 

) is given by a velocity squared type law 

(6) 

u 

Π 

⎢ ξ 

⎣ 

Π 

⎥⎦ 

. (6) 

The impact pressure of the powder cloud is given by 

The impact pressure 

p of the powder cloud is given by 

Π 

2 

⎡ 

u 

⎤ 

Π 

p Π 

= 

C D 

⎢ 

ρ 

Π 

⎥ 

(7). 

⎢⎣ 

2 

⎥⎦ 

. (7) 

In the example calculations we consider only the impact on walls with C D 

=2. 

n the example calculations we consider only the impact on walls with D 

= 

2 

RESULTS: esults: POWDER owder PRESSURE ressure FIELD SIMULATION Field OF Simulation THE 1999 ALL’ACQUA of the AVALANCHE 

1999 llcqua valanche 

To demonstrate how powder pressures can be deduced using two-layer avalanche dynamics 

models, we consider the well-documented 1999 All’Acqua avalanche (SLF, 1999; SLF, 2015). 

To demonstrate how powder pressures can be deduced using two-layer avalanche dynamics 

A large powder avalanche released spontaneously from the Poncione di Cassina Baggio 

(Bedretto models, we Valley) consider the the 20th well-documented of February, 1999. 1999 The llcqua blast of avalanche the powder (SF, cloud 1999 damaged SF, 2015). the 

large 

chimney powder avalanche and solar panels released of the spontaneously Pianseccohütte from (Fig. the 2 oncione ) as well as di several Cassina cables Baggio of (Bedretto an electric 

alley) 

power line (Fig. 2). Several buildings in the runout zone were slightly damaged. 

on the 20 th of February, 1999. The blast of the powder cloud damaged the chimney and solar 

Furthermore up to 12 ha of 200 year old forest were destroyed on the lateral edges of the 

track panels and of in the the ianseccohtte runout zone (Fig. 2). 

2 ) as well as several cables of an electric power line (Fig. 2). 

A back analysis of this event revealed that (SLF, 1999; SLF, 2015): 

Several buildings in the runout zone were slightly damaged. Furthermore up to 12 ha of 200 year 

– A 150 m wide release zone with average release depth of 2 m was observed on the east 

old 

slope forest 

of the were 

Poncione destroyed 

di Cassina on the lateral 

Baggio. edges 

These of 

observations the trac and 

were in the 

used runout 

to delineate zone (Fig. 

the 

2). 

starting zone of the avalanche and define the release volume. 

– bac The powder analysis pressures of this event at the revealed Pianseccohütte that (SF, were 1999 estimated SF, 2015): to be 2 – 3 kPa. 

The core did not reach the hut. 

• 150 m wide release zone with average release depth of 2 m was observed on the east 

– The powder pressure in the forest between the Pianseccohütte and the avalanche track 

were estimated slope of the to be oncione 5 kPa. 

di Cassina Baggio. These observations were used to delineate the 

– At 1900 starting m a.s.l. zone the of cables the of avalanche the BKW and power define line the located release 50 m volume. above ground were not 

damaged by the powder cloud. Powder pressures are estimated to be less than 2 kPa. 

• The powder pressures at the ianseccohtte were estimated to be 2 3 a. The core did 


C . 

not reach the hut. 

• The powder pressure in the forest between the ianseccohtte and the avalanche trac 

were estimated to be 5 a.

– At 1700 m a.s.l. the cables of the ATEL powder line located 15 m above ground were cut 

and three masts overturned. The estimated pressures were calculated to be between 

2 kPa and 5 kPa. 

– Moderate forest damage at the valley bottom and at run-up counter slope indicate that 

powder pressures were approximately 2 kPa between 5 m and 15 m above ground. 

Figure 2: The All’ Acqua avalanche track with the calculated maximum core velocity. The figure depicts the starting zone (2500 m), 

the Pianseccohütte (1985 m), the BFW power line located 50 m above ground (1900 m), the ATEL power line, located 15 m above ground 

(1700 m), the runout zone, the stream and counter slope and the buildings of All’ Acqua (1600 m). 

The simulations reveal that after release the core of the avalanche split into two distinct flow 

arms (Fig. 2). At 2200 m a.s.l. the avalanche descended over steep cliffs, reaching a maximum 

flow velocity of some 35 m/s. The two flow arms combined at 2000 m a.s.l. and managed 

to traverse a 300 m long flat track segment (slope inclination 10o), slowing to a velocity of 

20 m/s. The avalanche core passed 120 m from the Pianseccohütte (Fig 2). 

The avalanche then reached a steeper channel section and accelerated to 35 m/s, passing 

under the two power lines (Fig. 2). In the runout zone the avalanche core split into two flow 

fingers, one finger crossed the road and reached the stream at the valley bottom, 100 m after 

the road (Fig. 2). The core did not travel up the counter slope. One flow arm passed only 

30 m – 50 m from several buildings located in All’Acqua. 


Figure 3: a) Maximum powder pressures. Location of profiles. Profile a: Pianseccohütte. Profile b: ATEL power line. Profile c: Road, 

All’Acqua. b) Calculated maximum powder cloud heights and final extent of powder cloud. Model parameters: h 0 

= 1.5m, ρ0 = 250 kg/ 

m³, V0 = 38’000m 3 , h Σ 

= 0.5m, ρ Σ 

= 200 kg/m 3 , μ0 = 0.55, ξ0 = 2500 m/s², α=0.07, β=0.65 1/s, R0 = 2 kJ/m³. 

A major result in the model calculations is the difference between the simulated pressure at 

the Pianseccohütte (approximately 3 kPa) and the buildings located in the runout zone at 

All’Acqua (less than 1 kPa). The calculated powder snow avalanche pressures at the 

Pianseccohütte were larger, although the building is located 120 m from the flowing core. 

The calculated air-blast pressures at All’Acqua are smaller although the buildings are located 

only 30 m from the runout finger of the avalanche core. This result is due to terrain effects. 

The hut is located directly in the flow direction of the core as it descends over the cliffs. 

At this point the core reaches velocities up to 40 m/s. The core is later deflected but the 

powder cloud travels towards the Pianseccohütte. At the valley bottom in All’ Acqua, the 

houses are not in the primary direction of the core. A comparison of the pressures profiles a 

and c (defined in Fig. 3), reveals a large difference in the lateral attenuation of powder 

pressure: at the Pianseccohütte the air-blast pressure attenuates over a distance of approximately 

160 m (Fig. 4a); at All’Acqua the pressure attenuates over a distance of 50 m (Fig. 4c). 

Large air-blast pressures (between 6 kPa and 10 kPa) were predicted at the location where 

the avalanche crossed the ATEL power line (Fig. 4d). Maximum flow densities of the powder 

cloud varied between 4.5 kg/m³ and 5.0 kg/m³. In two-layer models these pressures 

represent the mean pressure exerted by the powder cloud. When the avalanche is a pure 

powder avalanche this region extends from the ground to the top of the cloud. The calculated 

powder cloud heights at the front of the avalanche (where the pressures are highest) are 

between 15 m and 20 m (Fig. 4d). The heights eventually increase to heights over 50 m 


Figure 4: a) Lateral attenuation of powder pressure at the profile a (Fig. 3a), Pianseccohütte. b) Attenuation of powder pressures at the 

ATEL power line, profile b (Fig. 3a). c) Powder profile at the road, profile c (Fig. 3a). d) Calculated powder cloud height and pressure at 

Pianseccohütte, ATEL power line and ATEL power line mast. 

(Fig. 3b). Therefore, the model predicts the ATEL cables are clearly vulnerable, although 

the two-layer model does not predict the attenuation of pressure with increasing height. 

At present we assume uniform depth profiles for velocity and density. 

Calculated pressures at the power mast are 2 kPa; calculated heights are 10 m, quickly rising 

to 25 m (Fig. 4d). These, coupled with the forces on the cables, would be sufficient to 

overturn the mast. At the upper power line (BKW) the calculated heights at the front of the 

avalanche are between 20 m and 30 m, lower than the 50 m suspension height of the cables 

(Fig. 3b). 

Finally, we note that the calculated pressures in the forested regions (Fig. 2) are enough to 

blow-down trees (Feistl et al., 2015). The regions of highest pressures (over 3 kPa) are 

located in the forest regions near the Pianseccohütte and between the two power lines 

(Fig. 3b). Calculated pressures on the counter slope (between 1 kPa and 2 kPa) are enough 

to cause isolated tree blow-downs (Fig. 3b). 

CONCLUSIONS 

In this paper we applied the research version of RAMMS, a two-layer numerical avalanche 

dynamics model, to simulate the spatial distribution of powder avalanche pressure in 

three-dimensional terrain. We are particularly interested in the pressures beyond the reach 

of the flowing avalanche core, as these pressures are used (1) to delineate yellow hazard 

zones, (2) identify regions of powder avalanche destruction (forests, buildings) and (3) 


calculate powder pressures on objects such as power lines, located some vertical distance 

above the flowing avalanche. We treat the powder cloud as an inertial flow: air is taken in by 

the avalanche core, loaded with ice-dust and blown-out with the speed of the flowing core. 

Once formed the powder cloud moves independently of the avalanche core. The primary 

flow direction and speed of the powder cloud is defined by the velocity of the core. The 

attenuation of the powder cloud speed (drag, Eq. 6) defines the magnitude of the powder 

cloud pressure outside the domain of the flowing core. 

Our simulation of the All’Acqua avalanche event indicate that because two-layer models 

simulate the movement of the core and cloud they allow the inclusion of important terrain 

effects which determine the area endangered by powder pressures, especially at the lateral 

edges of the avalanche. One-dimensional models are not able to reproduce these features. 

The model calculates the powder cloud height, speed and mean pressure, but does not predict 

the attenuation of pressure with height. For this purpose, we do not believe that threedimensional 

calculations are necessary, as these are computationally expensive; however, an 

engineering method is still out-standing. For this purpose, it will be necessary to redesign the 

experimental test sites to obtain measurements that validate an analytical pressure function. 

REFERENCES 

- Bartelt P., Buser O., Vera Valero C., Bühler Y. (2015) Configurational energy and the 

formation of mixed flowing/powder snow and ice avalanches, Annals of Glaciology, 71. 

- Bartelt P. and McArdell B. (2009). Granulometric investigations of snow avalanches. J. 

Glaciol., 55(193), 829–833 (doi: 10.3189/002214309790152384). 

- Bozhinskiy AN and Losev KS. (1998). The fundamentals of avalanche science [transl. 

CE Bartelt]. Eidg. Inst. Schnee- und Lawinenforsch. Mitt. 55. 

- Bühler Y., Christen M., Kowalski J., Bartelt P. (2011) Sensitivity of snow avalanche 

simulations to digital elevation model quality and resolution. Ann. Glaciol. 52, 58, 72-80. 

- Buser O. and Bartelt P. (2015). An energy-based method to calculate streamwise density 

variations in snow avalanches, Journal of Glaciology, 61(227), doi: 10.3189/2015JoG14J054. 

- Christen M., Kowalski J., Bartelt P. (2010). RAMMS: Numerical simulation of dense snow 

avalanches in three-dimensional terrain, Cold Regions Science and Technology, doi:10.1016/j. 

coldregions.2010.04.005. 

- Dreier L., Bühler Y., Ginzler C., Bartelt P. (2015). Comparison of simulated powder snow 

velocities, volumes and flow widths with photogrammetric measurements, Annals of 

Glaciology, in press. 

- Feistl, T., Bebi, P., Christen, M., Margreth, S., Diefenbach, L., and Bartelt, P. (2015). 

Forest damage and snow avalanche flow regime, Nat. Hazards Earth Syst. Sci., 15, 

1275-1288, doi:10.5194/nhess-15-1275-2015. 

- Grigoryan S., Urubayev N. and Nekrasov I .(1982). Experimental’noye issledovaniye lav 

innoy vozdushnoy volny [Experimental investigation of an avalanche air blast]. Mater. 

Glyatsiol. Issled., 


44, 87–93 [in Russian]. 

- Issler, D. (1998) Modelling of snow entrainment and deposition in powder snow 

avalanches, Annals of Glaciology, 26, 253 – 258. 

- Naaim, M. Gurer, I. (1998). Two-phase numerical model of powder avalanche: 

theory and application, Natural Hazards, 117, 129-145. 

- RL (1984). Richtlinien zur Berücksichtigung der Lawinengefahr bei raumwirksamen 

Tätigkeiten. Bundesamt für Forstwesen, Eidg. Institut für Schnee- und Lawinenforschung. 

Sampl, P., Zwinger, T., (2004). Avalanche simulation with SAMOS. Annals of Glaciology, 

38, 393-398. 

- SLF (1999). Lawinengefährdung der ATEL 380/220kV Leitung Airolo–Ponte(I) im Bereich 

All’Acqua Bedretto (TI), SLF Gutachten G 99.13. 

- SLF (2015). Lawineneinwirkungen auf die Pianseccohütte CAS/SAC, Bedretto (TI), 

SLF Gutachten G 2015.06. 

- Vera Valero C., Wikstroem Jones K., Bühler Y., Batelt, P. (2015). Release temperature, 

snow-cover entrainment and the thermal flow regime of snow avalanches, Journal of 

Glaciology, Vol. 61, No. 225, doi: 10.3189/2015JoG14J117. 



Monitoring sediment fluxes in Alpine rivers: 

the AQUASED project 

Gianluca Vignoli, Ph.D. 1 ; Silvia Simoni, Ph.D. 2 ; Francesco Comiti, Ph.D. 3 ; Andrea Dell'Agnese, Ph.D. 3 ; Walter Bertoldi, Ph.D. 4 ; 

Roberto Dinale, M.S. 5 ; Rudi Nadalet, M.S. 5 ; Pierpaolo Macconi, M.S. 6 ; Julius Staffler, M.S. 6 ; Rudolf Pollinger, M.S. 6 

ABSTRACT 

Monitoring activities in Alpine rivers is normally focused on water discharge measurements; 

sediment transport measurements have recently gained interest for several reasons, such as 

flood protection, research purposes, sediment budget and continuity. A new monitoring 

station has been designed and deployed in the Sulden/Solda river (South Tyrol, Italy) to 

address these needs. The watershed is characterized by large glacier areas and steep slopes 

feeding the river with sediments. The station is equipped with a 4m-rack of geophone plates 

and an acoustic pipe-hydrophone. Bedload is measured along with suspended load, water 

stage, water conductivity and temperature. This work presents the installation and the results 

collected during a high flood event and compares them to values derived from bedload 

equations obtained by other authors. 

KEYWORDS 

monitoring, bed load, geophone, alpine rivers 

INTRODUCTION 

The traditional monitoring in Alpine rivers is focused on water discharge measurements. 

Recently the monitoring activities have focused both on water discharge and on sediment 

load (bedload + suspended load). Sediments move across the channel network transported by 

water according to two main mechanisms: as suspended sediment and as bed load. The first 

mechanism involves small size sediments (silt and sand), which are lifted by the water 

turbulence; the second involves the larger sediments (pebbles and cobbles), which move by 

rolling, sliding and saltation close to the riverbed. Mountain river hydraulic and morphology 

are strictly linked through bedload transport processes (Gomez, 2006); monitoring bedload 

is important to engineers, scientists, public authorities and people/institutions/enterprises 

operative in water resources. Measuring and understanding sediment transport phenomena is 

the key point for further investigation on the morphodynamic evolution of alpine rivers and 

on their response to natural and anthropic forcing. Suspended sediment transport monitoring 

is a consolidated activity and requires one or more turbidity meter to be installed in a river 

1 CISMA srl, Bolzano, ITALY, gianluca.vignoli@cisma.it 

2 Mountain-eering srl 

3 Facoltà di Scienze e Tecnologie-Libera Università di Bolzano 

4 Dipartimento di Ingegneria Civile Ambientale e Meccanica-Università degli Studi di Trento 

5 Ufficio Idrografico della Provincia Autonoma di Bolzano 

6 Agenzia per la protezione civile – Provincia Autonoma di Bolzano 



cross section. Bedload monitoring is a more challenging activity in mountain streams because 

of the complexity of the process. In the last decades both direct and indirect approaches for 

bedload monitoring have been applied in Europe (Lenzi et al., 1999; Rickenmann et al., 2012; 

Rickenmann et al., 2014; Dell'Agnese et al., 2014; Mao et al., 2014) and in Japan (Mizuyama 

et al.,2010). Within the AQUASED (which stands for the latin AQUA=water and SEDiments) 

project a new monitoring station has been set up along the Solda river to measure water and 

sediment fluxes. This work describes the set up of the new station and presents the instruments 

calibration and results related to bedload monitoring. 

THE AQUASED PROJECT 

Within the AQUASED project a new monitoring station (Figure 1) has been designed and 

set up in South Tyrol (Italy), along the Solda river. The project partners encompass two SMEs, 

two universities and the local authorities. The Solda river drains a 145 km 2 wide watershed 

with many glacial areas (18 km 2 ), which strongly affect the hydrological regime and the 

sediment availability. The new station has been designed with the aim of monitoring both 

water and sediment fluxes (suspended- and bedload). Suspended sediment monitoring is 

performed using a standard turbidity meter, bedload is indirectly measured by a rack of 

8 Swiss-plate geophones, installed on the downstream end of a check dam. The minimum 

plates sensitivity is 2, 3 cm-grain size . The plates rack is 4m long and covers half the river 

width (8m), from the river center to the right bank. Water discharge is measured by means of 

water level measurements and the estimation of a suitable rating curve, derived from several 

direct water discharge measurements, performed using the salt dilution method. This consists 

in dumping a known amount of salt (NaCl) into the stream so that its base conductivity is 

increased. This increment is indirectly measured with a conductivity meter and a thermometer 

installed on the river bank. The whole station is the result of a large cooperation: the 

SMEs provided part of the measuring instruments and the design activities, the universities 

provided the scientific support and the local authorities (Water Department of the Province of 

Bolzano) provided the necessary support to make the project be operative. 

Figure 1: (left) the measuring station: the array of geophones is installed on half width on the downstream end of the check dam and 

(right) a detail of the turbidimeter installation system 


The monitoring station has been installed during winter 2013-14 and completed at the beginning 

of May 2014; the collected data include: water level, turbidity, temperature and conductivity 

and geophone impact data. One of the novelties of this installation is the geophones data 

recording system, which saves continuously the whole raw plate velocity (cm/s) data sampled 

at 5kHz, without loss of information, adopting a signal/noise detecting algorithm and an 

on-the-fly compression. This system can save a year of 5 kHz data in about 20 GB of disk 

space for each plate. The saved data can be elaborated both counting the number of impulses 

above given thresholds (i.e. the “traditional” plate signal elaboration described in Rickenmann 

et al. 2014) and applying any type of signal process algorithm. The traditional signal 

elaboration consists in fixing a threshold and counting the number of recurrences (called 

impulses) the signal exceeds it. Here six different thresholds are used; the signal intensity 

depends on the sediment grain size, vibrations due to larger cobbles are typically higher, 

therefore the upper threshold can be related to large-size sediments, while the lower are 

related to smaller sediments. In the following the number of impulses are collected by six 

channels, the first (channel n. 1) refers to the lower threshold and channel n.6 refers to 

the higher threshold. 

In 2014 the monitoring station was equipped with an acoustic pipe sensor (“Japanese pipe 

hydrophone”, Mizuyama et al., 2010) for bedload sampling, with the main aim of performing 

inter-comparison between the two indirect bedload measuring systems (plate geophone and 

pipe hydrophone), but unfortunately on August 13th 2014 a high intensity flood heavily 

damaged the pipe. 

Figure 2: bedload sampling using a crane mounted trap (left), the sediment trap (right) 

CALIBRATION 

The geophone plates carry out indirect measurement of the bedload discharge. Basically, 

recorded data are a measure of the vibrations induced by the sediments moving on the plates. 

Suitable rating curves are needed to correlate the specific bedload transport (measured in kg/ 

(m s)) to the number of impulses recorded by the system. Calibration measurements have 


een performed using a heavy bedload trap, handled by a crane. The trap is made by a net 

characterized by an opening size of 3.6 mm anchored to a metal frame. It is lowered into the 

stream flow just downstream of a plate and held by a crane and four ropes, which help in the 

trap positioning. The trap is kept in the sampling position for 5 to 10 minute, depending on 

the bedload transport intensity. The net retains sediment particles larger than 3.6 mm in diameter. 

The collected material is then sieved using four sieves (Figure 3), characterized by net 

dimensions of 64mm, 45mm, 32mm and 22mm. The five separated samples are weighed, the 

four larger size samples are then released back into the stream and the grain size distribution 

of the finest size fraction (

RESULTS 

Since its installation the station worked regularly, recording a high intensity flood occurred 

on August 13th 2014. Figure 5 (upper part) shows the water discharge data recorded in the 

summer season 2015: the Solda river is characterized by a snow melting period during May 

and June, followed by a glacier melting period, during July and August. Figure 5 (lower part) 

shows the plate signal measured by plate n.2, located close to the river center. The number 

of impulses recorded on channel 2 is high, especially during high flows and during the 

glacier melting period. During the low flow periods channel 2 impulses never drop below 

100 impulses/h, since during the summer bedload is always active. The number of impulses 

recorded on channel 4 refers to larger sediments and during low flow conditions (end of May, 

June and August) it is very low, showing that in those periods large-size sediment transport 

is not observed. 

100000 

30 

20 

10 

0 

discharge [m 3 /s] 

pulses/h 

10000 

1000 

100 

10 

1 

05/16 05/30 06/13 06/27 07/11 07/25 08/08 08/22 

Channel 2 Channel 4 water discharge 

Figure 5: water discharge (above) and bedload impulses (below) during summer 2015, data from plate 2, located near the river center 

Figure 6 shows the flood of August 13th 2014, which was characterized by a peak of 72 m 3 /s 

(specific discharge 0.5 m³/s/km 2 ), and by an intense sediment transport, which attains values 

up to 1000 kg/s, (preliminarily estimate using data from only one geophone plate). The 

computation of the total amount of sediments transported across the measuring station 

during the entire flood event was 7100 tons. Impulses on channel 6 were recorded mainly 

during this intense flood event, characterized by large boulder transport. Water and sediment 

fluxes seems to be well correlated at the beginning of the flood; on the contrary the water 


discharge dropped rapidly at the end of the event whereas the sediment discharge kept high 

values for several hours. Unfortunately the acoustic pipe was damaged during this event 

because it was impacted by large boulders (up to 1m) transported by the flow. In contrast, the 

geophone plates were not damaged. Data show that the actual bedload is significantly lower 

compared to values estimated applying the Schocklitsch (1962) and Rickenmann (1990 and 

2001) equations, even after accounting for form drag correction. Sediment transport 

equations approximately match the bedload rates recorded by the geophone plates only at the 

peak of the transport event. This confirms that using formulas to estimate bedload in steep 

gravel-bed rivers can lead to large errors (Gray and Simões, 2008) especially at normal flows. 

Figure 6: bedload rate (thick black line) and water discharge data (grey line) data collected during the August 13th 2014 event 

Figure 7 shows the number of impulses recorded during summer 2015 on channel 4 versus 

the water discharge. Data show a complex pattern: the bedload transport intensity seems to 

be not well correlated to water discharge. High bedload rates have been measured whereas 

water discharge values never exceeded 7-8 m 3 /s especially in May; high bedload rates have 

been recorded during high flow conditions in June; in July the bedload transport rates did not 

reach the values observed in June-even though the water discharge reached a similar peak 

value; finally in August, high bedload transport intensities have been measured for intermediate 

flow conditions. Bedload intensity reflects both sediment availability (in the catchment 

as well as within the riverbed) and the sediment transport capacity of a given river reach. For 

example the high bedload transport rate measured in May could be related to high sediment 

availability within the riverbed; the high bedload transport rate measured in August is 


probably related to the sediments coming from the glaciers located in the upper part of the 

catchment. 

100000 

10000 

1000 

pulses/h 

100 

10 

1 

5 10 15 20 25 30 

water discharge [m 3 /s] 

May 

June 

July 

August 

Figure 7: 2015 measured channel 4-impulses are plotted versus water discharge, data from plate 2, located near the river center 

CONCLUSIONS 

A new sediment transport monitoring station in a glacierized and active watershed has been 

designed and installed. The collected parameters are bedload, suspended load, water stage, 

water conductivty and temperature. Bedload is indirectly measured using a rack of 8 geophone- 

plates. Data are recorded using an innovative signal/noise recognition system and 

a compression algorithm, which allows to save the whole high-frequency signal, with an 

acceptable amount of storage usage. Calibration was performed using a bedload sampling 

device with a mobile crane support. 

A complete data set of bedload and water discharge values was collected during 2014 and 

2015. The Solda river behaviour is characterized by a glacial runoff regime, with a very low 

water discharge during winter, a snow melting period between March and June and a glacier 

melting period from July to August-September. Sediment transport data show higher bedload 

values during the glacier melting period and very low sediment transport rates during the 

snow melting phase, probably due to considerable difference in the sediment availability. 

Flood event data analyses and the comparison with bedload formulas confirm the complexity 

of bedload estimation in steep gravel bed rivers. 


REFERENCES 

- Dell'Agnese, A., Mao, L., Comiti, F. 2014: Calibration of an acoustic pipe sensor through 

bedload traps in a glacierized basin. Catena 121:222-231. 

- Gomez, B., 2006: The potential rate of bedload transport, Proceedings of the National 

Accademy od Sciences, V. 103, no.46, p. 17170-17173. 

- Gray, J.R., and Simões, F.J.M., 2008: Estimating sediment discharge, appendix D of 

Garcia,- Marcelo, ed., Sedimentation engineering: ASCE Manuals and Reports on Eng. 

Practice No. 110, p. 1067–1088 

- Lenzi, M.A.; D'Agostino, V.; and Billi, P., 1999: Bedload transport in the instrumented 

catchment of the Rio Cordon: Part 1. Analysis of bedload records, conditions and threshold 

of bedload entrainment. Catena 36 (3), 171-190. 

- Mao, L., Dell'Agnese, A., Huincache, C., Penna, D., Engel, M., Niedrist, G., Comiti, F. 2014: 

Bedload hysteresis in a glacier-fed mountain river. Earth Surface Processes And Landforms 

39(7):964-976. 

- Mizuyama, T.; Laronne, J.B; Nonaka, M.; Sawada, T.; Satofuka, Y; Matsuoka, M.; Yamashita, 

S.; Sako, Y; Tamaki, S.; Watari, M; Yamaguchi, S. and Tsuruta, K, 2010: Calibration of a 

Passive Acoustic Bedload Monitoring System in Japanese Mountain Rivers, part of U.S. 

Geological Survey Scientific Investigations Report 2010-5091. 

- Rickenmann, D. 1990: Bedload transport capacity of slurry flows at steep slopes, Diss. 

Techn. Wiss. ETH Zürich, Nr. 9065. http://dx.doi.org/10.3929/ethz-a-000555802 

- Rickenmann, D., 2001: Comparison of bed load transport in torrents and gravel bed streams. 

Water Resour. Res. 37, 12: 3295-3305. 

- Rickenmann, D.; Turowski, J.M.; Fritschi, B.; Klaiber, A.; Ludwig, A., 2012: Bedload 

transport measurements at the Erlenbach stream with geophones and automated basket 

samplers. Earth Surf. Process. Landf. 37: 1000-1011. 

- Rickenmann, D., Turowski, J.M., Fritschi, B., Wyss, C., Laronne, J., Barzilai, R., Reid, I., 

Kreisler, A., Aigner, J., Seitz, H., Habersack, H. (2014): Bedload transport measurements with 

impact plate geophones: comparison of sensor calibration in different gravel-bed streams. 

Earth Surface Processes and Landforms, 39, 928–942, doi: 10.1002/esp.3499. 

- Schocklitsch, A, 1962: Handbuch des Wasserbaues, Speringer, Vienne, Vol 1, 173-177. 



Monitoring unstable parts in the ice-covered 

Weissmies northwest face 

Lukas E Preiswerk 2 ; Fabian Walter 1 ; Sridhar Anandakrishnan 3 ; Giulia Barfucci 4 ; Jan Beutel 5 ; Peter G Burkett 3 ; 

Pierre Dalban Canassy 2 ; Martin Funk 2 ; Philippe Limpach 6 ; Emanuele Marchetti 4 ; Lorenz Meier 7 ; Fabian Neyer 6 

ABSTRACT 

The glacierized northwest face of Weissmies in the Saas valley (Switzerland) recently became 

unstable due to climate-induced glacier thinning of the supporting Triftgletscher below. 

In the case of a large break-off of ice, human infrastructure in the Saas valley is exposed to 

the danger of an ice/snow avalanche. A monitoring campaign was initiated with the goal 

of detecting precursory signals to the break-off. Interferometric and Doppler radar, optical 

imaging as well as GPS sensors provide measurements of surface displacements. 

Infrasound and seismometer arrays monitor acoustic and seismic emissions of ice avalanches 

and englacial fracture development. Here we discuss the monitoring methods and the results 

obtained so far. The unstable glacier mass did not undergo a large-scale break-off event, 

in fact it decelerated during the unusually warm summer months. An explanation remains 

elusive but likely involves subglacial processes and bedrock topography. Nevertheless, 

our results allow us to draw important conclusions regarding the suitability of different approaches 

to monitoring unstable glaciers. 

KEYWORDS 

glacier monitoring; ice avanlache; early warning; natural hazards 

INTRODUCTION 

The hazard potential of glaciers ranges from relatively small ice falls (Röthlisberger, 1978) to 

substantial glacier break-offs, which have killed thousands of people in the past (Lliboutry, 

1975). To monitor such hazards, previous studies have focused on surface deformation and 

icequake activity (e.g. Failletaz et al., 2011; Dalban Canassy et al., 2012). Moreover, damage 

evolution (Pralong and Funk, 2006) and slider block models (Failletaz et al., 2010) have 

provided theoretical insights into glacier instabilities. 

The thermal regime of glaciers plays a key role in processes leading to instabilities. Cold-based 

glaciers are frozen to their beds and fail via fracture growth. Temperate-based glaciers slide on 

1 Laboratory of Hydraulics, Hydrology and Glaciology, ETH Zürich, Switzerland, fwalter@vaw.baug.ethz.ch 

2 Laboratory of Hydraulics, Hydrology and Glaciology, ETH Zürich, Switzerland 

3 Department of Geosciences and Earth and Environmental Systems Institute, Pennsylvania State University, USA 

4 Department of Earth Sciences, University of Firenze, Italy 

5 Computer Engineering and Networks Laboratory, ETH Zürich, Switzerland 

6 Institute of Geodesy and Photogrammetry, ETH Zürich, Switzerland 

7 GEOPRAEVENT AG, Zürich, Switzerland 



the bedrock and undergo “active phases”, which may lead to large-scale ruptures (Failletaz 

et al., 2015). These acceleration phases usually occur late in the melting season and are 

caused by elevated subglacial water pressure. Ongoing climate changes affect glacier stability 

as previously cold glaciers may transition to temperate thermal regimes (e.g. Glacier de 

Taconnaz (F); Gilbert et al., 2015). 

Here, we study the ice-covered northwest face of Weissmies in the Saas valley (Switzerland). 

Until recently, the adjacent Triftgletscher buttressed this ice cover from below (Fig. 1A). 

However, Triftgletscher's thinning has almost entirely removed this support (Fig. 1B). 

In addition, the glacier is likely in a transition from cold to temperate as surface meltwater 

warms the previously cold bedrock, weakening the ice-bed interface and further promoting 

instability. The situation is critical, because tourist activity and – in the case of a large event – 

human infrastructure in the Saas valley are exposed to the danger of a glacier break-off. 

Therefore, a monitoring campaign was initiated to determine the frequency and volumes of 

icefalls, to improve our understanding of processes leading to glacier instabilities, and to 

detect break-off precursors. An interferometric radar and an optical camera were installed in 

October 2014, followed by GPS sensors, Doppler radar, an infrasound array and seismometers 

in June 2015 (Fig. 1C). 

A Weissmies 4017 masl B 

200 m 

+ 

C 

Instabil1 Instabil2 + Doppler Radar Interferometric Radar GPS Photogrammetry Infrasound Array Seismometers 

Interferometric Radar 

Doppler Radar 

GPS 

October 2014 

March 2015 

break-off 

break-off 

Photogrammetry 

October 2014 

Infrasound 

Seismology 

11.6. 18.6. 25.6. 2.7. 5.7. 9.7. 16.7. 23.7. 30.7. 6.8. 13.8. 

Date (2015) 

Figure 1: A) Oblique view of the unstable glaciers on the northwest face of Weissmies, between 2005 and 2009. (Source: swisstopo) B) 

The same view in 2014, with the position of our sensors. Note the substantial reduction of supporting ice below “Instabil1”. C) A 

temporal overview of the measurements during summer 2015. 


Here, we discuss the observations in terms of processes affecting glacier stability. We focus on 

two icefalls, one of which occurred on 25 June 2015, when part of the steep ice mass 

“Instabil1” (Fig. 1) broke off (http://youtu.be/0aVyTfqafzg). The second ice fall originated 

from “Instabil2” on 5 July 2015 (http://youtu.be/m3nEzT9TIYE; a radar animation, as this 

event was not captured by the video camera). The estimated volumes are a few thousand m³ 

per event, which is only a tiny fraction of the 750 000 m³ that could potentially break off. 

In winter, events of more than 30 000 m³ of ice may endanger the ski resort, and a break-off 

of more than 200 000 m³ of ice could reach the town of Saas-Grund (information obtained 

from the company wasser/schnee/lawinen - A. Burkard AG, Brig). 

METHODS AND RESULTS 

Surface deformation 

The goal of monitoring of surface deformation is to detect hyperbolically increasing surface 

velocities, which typically precede large break-off events (Flotron, 1977; Röthlisberger, 1978). 


A 

Velocity, cm d -1 

30 

20 

10 

break-off 

break-off 


Glacier 

Instabil1 

Instabil2 

fastest 5% 

B 

C 

Velocity, cm d -1 

Hours above 0° C 

0 

20 

15 

10 

5 

0 

11.6. 18.6. 25.6. 2.7. 9.7. 16.7. 23.7. 30.7. 6.8. 

Date (2015) 

25 

20 

15 

10 

5 

Glacier 

Instabil1 

Instabil2 

GPS Instabil1 

0 

0 

1.6.14 1.8.14 1.10.14 1.12.14 1.2.15 1.4.15 1.6.15 1.8.15 

Date 

Figure 2: A) Radar line-of-sight velocities of Triftgletscher and the unstable zones. Note the increase in velocity prior to the second 

break-off event. B) GPS 3D velocities on Instabil1. The increase around 9 July is mostly an increase in vertical direction, likely due the 

GPS antenna stakes sinking into the ice. C) Long-term velocities (in color) and temperature (in hours above 0° C) clearly show that the 

summer of 2015 was warmer than the previous one, leading to more melt. 

25 

20 

15 

10 

5 

Radar velocities, cm d -1 


The interferometric radar emits microwaves (17 GHz, wavelength ≈ 1.7 cm) and evaluates the 

signal reflected by the glacier surface (Rödelsperger et al., 2010). Our IBIS-L system produces 

2D images (azimuth and range) using the Synthetic Aperture Radar (SAR) technique. Since 

microwaves are affected by air humidity, temperature and pressure, atmospheric effects have 

to be removed. Reflection data are projected onto a topographic model to obtain a 3D image 

of glacier surface velocities. Uncertainties are estimated at 10% - 20% based on daily signal 

fluctuations, which are attributed to radar processing and not to actual glacier motion. Radar 

measurements are possible in low visibility weather and at night, and can be made from a 

safe distance (~ 2 km in this case). However, only the target motion component parallel to 

the line of sight is recorded. 

Fig. 2A shows the average velocity of the glacier and the unstable parts during summer 2015. 

The movement is around 10 cm d -1 and fairly stable in this period. Long-term velocities 

(Fig. 2C) indicate that the unstable parts are significantly slower than in October 2014, when 

Instabil1 was moving at more than 20 cm d -1 . 

A clear velocity increase to up to 25 cm d -1 precedes the icefall on 5 July (Fig. 2A). The area 

affected by the acceleration is readily identified in the projection onto an elevation model 

(Fig. 3C&D). In contrast, no elevated velocities were registered prior to the 25 June event 

(Fig. 2A, 3A&B). 

Doppler Radar 

A Range-Doppler Radar operating at X band (10 GHz), with the ability to measure azimuth 

angle, range and velocity was installed to detect ice avalanches. It has opening angles of 90° 

horizontally and 10° vertically and captures all movements up to a range of 2000 m within 

this area, hence also snow avalanches. 

GPS 

Two low-power, low-cost GPS sensors that measure the L1 frequency only were installed on 

Instabil1 (Fig. 1B) measuring in-situ surface velocities. A local long-haul WLAN enables 

wireless data transmission. Differential processing provides accuracies in the cm range (Fig. 

2B). GPS sensors yield valuable ground truth data for the radar measurements. The difference 

between GPS and radar velocities can be attributed to projection and local site effects. 

However, they constitute point measurements and neither predicted nor detected the two icefalls, 

which occurred a few 100 m from the antennas. 

Photogrammetry 

In contrast to radar measurements, this method is sensitive to displacements perpendicular to 

the line of sight. Subsequent image templates (small sections around a feature in the image) 

are matched using least squares optimization (e.g. Grün, 1985). We used an optical off-theshelf 

digital camera with 70 mm focal length resulting in a pixel resolution of ~15 cm, slightly 

varying due to the various target distances in the field of view. The accuracy of the technique 

is on the order of 0.05 pixel (e.g. Mass and Hampel, 2006). In our case, pronounced snow 

height and ice topography changes, as well as strong variations in scene illumination lead to 

relatively high inaccuracies of about 0.5 pixel, equivalent to 7.5 cm. The event on 5 July was 

captured in the image sequence, with observed peak velocities of around 30 cm d -1 in 


A 

before break-off 

B 

after break-off 

25 June 

25 June 

C 

5 July 

0 12 24 cm d -1 

0 12 24 cm d -1 

0 12 24 cm d -1 0 12 24 cm d -1 

before break-off 

D 

after break-off 

5 July 

horizontal component break-off 

0.35 0.35 vertical component break-off 

E F 

0.30 

0.30 

0.25 

0.25 

0.20 

0.20 

0.15 

0.15 

0.10 

0.10 

0.05 

0.05 

0.00 

0.00 

-0.05 

-0.05 

-0.10 

28.6. 30.6. 2.7. 4.7. 6.7. 8.7. 10.7. 28.6. 30.6. 2.7. 4.7. 6.7. 8.7. 10.7. 

Date (2015) 

Date (2015) 

Relative motion, m 

Figure 3: A) - D) Radar velocities before and after the break-off events. Note the almost identical radar images A) and B), but the clear 

difference between C) and D). E) & F) Displacement estimates derived from optical image sequences. The transition of image pixel to 

metric displacements is based on the camera-target distance. The data gap after the event is due to the loss of coherent image 

structures after the break-off. 

horizontal and 20 cm d-1 in vertical direction (Fig. 3E&F). There was no sign in the image 

sequence for the earlier event (25 June). 

Acoustic and Seismic Emission 

Acoustic and seismic waves originate from icequakes, which are mostly tensile dislocations in 

the ice, or arguably arise from stick-slip motion of the glacier (Walter et al., 2008). Changes 

in icequake activity reflect the evolution of englacial damage prior to a glacier break-off 

(Failletaz et al., 2011). 

Relative motion, m 


Infrasound 

We measured low-frequency (1-20 Hz) acoustic waves with a small aperture (150 m) 4 

element infrasound array installed ~2.5 km away from the unstable glacier tongue (Fig 1B). 

Each array element was equipped with a differential air pressure transducer with a sensitivity 

of 25 mV/Pa and lower corner frequency of 0.01 Hz. 

Array processing was performed on the infrasonic data applying multi-channel correlation 

among all elements of the array. Signal correlation is exploited to calculate arrival time 

differences of incoming sound waves, which are then used to determine wave parameters 

(event back-azimuth and apparent propagation velocity) (Ulivieri et al., 2011). Analysis was 

performed over 5 s windows, leading to a total of 52283 detections of coherent signals during 

the observation period. A variety of continuous background sources (e.g. nearby melt water 

streams, cultural activity) are responsible for the large number and diurnal variability of infrasound 

detections (Fig. 4A). 

Requiring wave parameter stability reduces the detection list to 19 events (yellow crosses in 

Fig. 4A). Among those, the two break-offs can be unambiguously identified by their 

back-azimuth (Fig. 4B). During the 25 June break-off, the decreasing back-azimuth reflects 

the northward motion of the ice avalanche (Fig. 4D). The decrease in back-azimuth for the 

event of 5 July is less pronounced, as this event was moving due west, towards the array 

(Fig. 4F). Despite the reliable detection capability, the large 2.5 km distance between array 

and glacier implied a low signal-to-noise ratio. Any signals from precursory fracture activity 

are likely hidden in the background noise. 

A 

B 

C 

D 

Pressure, Pa 

Back Azimuth, °N 

Pressure, Pa 


8 

6 

4 

2 

0 

360 

270 

180 

90 

0 

0.1 

0 

-0.1 

115 

110 

break-off 

18.6. 25.6. 2.7. 9.7. 16.7. 23.7. 

340 

330 

105 

320 

05:58:00 05:58:15 05:58:30 05:58:45 05:59:00 

Time (25.6.2015) 

Apparent Velocity, m s -1 

E 

F 

Weissmies NW-face 

Infrasound events 

Doppler radar events 

105 

320 

17:17:30 17:17:45 17:18:00 17:18:15 17:18:30 

Time (5.7.2015) 

Figure 4: A) Pressure and B) back-azimuth of infrasound detections. From all events (yellow crosses), the two break-offs can be 

identified from their back-azimuth pointing towards Weissmies’ northwest face. The Doppler radar picked up smaller events as well 

(green asterisks). C) - F) Pressure, back-azimuth and apparent sound velocities of the two break-off events. 

Pressure, Pa 


break-off 

0.1 

0 

-0.1 

115 

110 

340 

330 

Apparent Velocity, m s -1 


Seismology 

We installed three seismometers (sensitivity from 10 Hz to 500 Hz) on Instabil1, two of them 

collocated with the GPS sensors (Fig 1B). The seismometers were recovered after two weeks, 

as wireless data transmission did not work reliably. Therefore, our interpretation is limited 

to the break-off event of 25 June. In the hours before the break-off, no precursory activity 

stands out (Fig. 5A&B). Moreover, automatic icequake detections (Walter et al., 2008) 

indicate a steady decrease in fracture event activity (Fig. 5C). 

A 

B 

Frequency (Hz) 

C 

counts 

5000 

0 

-5000 

250 

100 

50 

break-off 

10 

00:00 01:00 02:00 03:00 04:00 05:00 06:00 

Time (25 June) 

40 

× 10 4 

5 

4 

3 

2 

1 

break-off 

4000 

Power Spectral Density 

events per hour 

D 

30 

20 

10 

0 

0 

13.06. 15.06. 17.06. 19.06. 21.06. 23.06. 25.06. 

Date (2015) 

3000 

2000 

1000 

30.9.2014 E 

8.8.2015 

cumulative number of events 

Figure 5: A) Seismogram of the 6 hours before break-off. Data gaps amounting to less than 0.01% were filled via interpolation. B) 

Corresponding spectrogram. Vertical spectral lines indicate icequake occurrences and highlight their broadband (10-100 Hz) character. 

Note the clear signal of the break-off event. The straight horizontal lines are electronic noise. C) Evolution of events in the two weeks 

before the break-off. D) 2014 aerial photograph of Instabil 1 showing widespread wetted bedrock and thus sheet-like subglacial melt 

discharge. (Source: U. Andenmatten) E) Photograph from 2015 showing a single channel of subglacial discharge. 


DISCUSSION 

Monitoring Methods 

No large break-off events occurred during our monitoring period. Nevertheless, the interferometric 

radar successfully forecasted the break-off of 5 July. On the other hand, the radar 

detected no acceleration prior to the 25 June event. Photogrammetric analysis of this event 

also failed to detect changes in surface motion at least 15 minutes before break-off. Both techniques 

are sensitive to glacier surface motion in different (perpendicular) directions. We 

therefore suggest that the 25 June break-off had a different failure mechanism, which unlike 

the 5 July event was not preceded by an acceleration phase. Thus, instabilities of a few 

thousand m³ of ice can develop and fail without speed-up. 

Infrasound provided reliable detection of both break-off events and no false alarms. The 

Doppler radar detected the two avalanches as well as some smaller events. Similarly, the 

25 June break-off generated the strongest detected seismic signal. Further analysis of the 

preceding icequakes is needed to clarify if these events provide some information about 

precursory fracture activity leading to break-off. 

In conclusion, none of the presented techniques by itself is sufficient for automatic detection 

and forecasting of break-off events. On the other hand, break-offs of at least a few thousand 

m³ of ice with precursory acceleration can be reliably forecasted using interferometric radar 

and photogrammetry. However, at this stage manual review of radar and photographs is 

needed. Infrasound, Doppler radar and seismic measurements produce clear break-off signals, 

which require minimal human interaction and could potentially be fully automated. A 

large-scale break-off of the entire unstable ice mass may produce stronger signals, including 

precursors, which all presented techniques are capable of detecting. 

Glacier dynamics 

The surface velocities on the unstable part of the glacier decreased from ~20 cm d -1 in October 

2014 to 5 cm d -1 in February 2015 (Fig. 2C). During July 2015, it even decreased to an 

unexpected low of 3 cm d -1 . 

In 2015, long periods above freezing substantially promoted meltwater production (Fig. 2C). 

Surface meltwater most likely accessed the glacier bed, where it partly refroze, warming the 

ice/bed contact and promoting basal sliding. Therefore, contrary to the observed slowdown, 

the unstable glacier part was expected to move faster during the 2015 high-melt periods. 

This slowdown can be explained by a change in subglacial hydraulics. Comparing images 

from 2014 and 2015 reveals a widespread zone of wet rock below the unstable ice in 2014 

(Fig. 5D). Subglacial water was thus present under an extended part of the glacier. Conversely, 

a single stream exited the unstable part in 2015 (Fig. 5E). Higher amounts of available 

meltwater in 2015 probably lead to the channelization of subglacial meltwater flow making 

the drainage system more efficient during summer 2015 than in 2014. This reduced basal 


water pressures, strengthening the basal contact (Kamb et al., 1985), and leading to limited 

basal motion. Another explanation for glacier deceleration is the presence of a subglacial rock 

barrier, against which the ice mass has recently stabilized. There is no further evidence for 

this theory, but at this stage this possibility cannot be excluded. Our seismological results can 

explain both the sliding decrease and subglacial barrier hypothesis in terms of reduced 

stick-slip icequakes and less extensional crevasse icequakes, respectively. 

By the time of writing, it is unclear how long the various measurements will continue in the 

future. This will largely depend on the future hazard assessment by the responsible authorities. 

Additional investigations on the englacial thermal regime with borehole measurements 

would help to understand the dynamical behavior of this steep ice covered face. A numerical 

ice flow model incorporating material damage would help to identify critical regions where 

crevasse formation indicates imminent failure. Finally, bedrock topography and the total ice 

volume, which could be obtained by ground-penetrating radar, would be an asset for future 

hazard assessment. 


The Swiss National Science Foundation financed the salaries of LP and FW and part of the 

instrument deployment (GlaHMSeis Project PP00P2_157551). We received funding from the 

US National Science Foundation (OPP1039982) for development of the geopebbles. We thank 

the editor and two anonymous reviewers for constructive comments. We are also indebted to 

U. Andenmatten, S. Bilén, R. Bock, Ch. Marty, J. Portelli, M. Pusateri, the Municipality of 

Saas-Grund, the Canton of Valais, the Swiss Federal Office for the Environment and 

Bergbahnen Hohsaas AG. 

REFERENCES 

- Dalban Canassy, P., Faillettaz, J., Walter, F., & Huss, M. (2012). Seismic activity and surface 

motion of a steep temperate glacier: a study on Triftgletscher, Switzerland. J. Glac., 58(209). 

- Faillettaz, J., Sornette, D., & Funk, M. (2010). Gravity-driven instabilities: Interplay between 

state- and velocity-dependent frictional sliding and stress corrosion damage cracking. 

JGR:SE, 115(B3), B03409. 

- Faillettaz, J., Funk, M., & Sornette, D. (2011). Icequakes coupled with surface displacements 

for predicting glacier break-off. J. Glac., 57(203), 453–460. 

- Faillettaz, J., Funk, M., & Vincent, C. (2015). Avalanching glacier instabilities: Review on 

processes and early warning perspectives. Rev. Geophys., 53. 

- Flotron, A. (1977). Movement studies on a hanging glacier in relation with an ice 

avalanche. J. Glac., 19(81), 671–672. 

- Gilbert, A., Vincent, C., Gagliardini, O., Krug, J., & Berthier, E. (2015). Assessment of 

thermal change in cold avalanching glaciers in relation to climate warming. Geophys. 

Res. Lett., 42. 


- Grün, A. W. (1985). Adaptive Least Squares Correlation: A Powerful Image Matching 

Technique. S. Afr. J. Photogramm., Rem. Sens & Cart., 14, 175–187. 

- Kamb, B., et al. (1985). Glacier Surge Mechanism: 1982-1983 Surge of Variegated Glacier, 

Alaska. Science 227(4686), 469–479. 

- Lliboutry, L. (1975). La catastrophe du Yungay (Pérou). Proc. Snow & Ice Symp., Moscow, 

August 1971, 353–363. 

- Maas, H.-G., & Hampel, U. (2006). Photogrammetric Techniques in Civil Engineering 

Material Testing and Structure Monitoring. Photogramm. Eng. Rem. Sens., 72(1), 39–45. 

- Pralong, A., & Funk, M. (2006). On the instability of avalanching glaciers. J. Glac., 

52(176), 31–48. 

- Rödelsperger, S., Läufer, G., Gerstenecker, C., & Becker, M. (2010). Monitoring of displacements 

with ground-based microwave interferometry: IBIS-S and IBIS-L. J. Appl. Geod., 

4(1), 41–54. 

- Röthlisberger, H. (1978). Eislawinen und Ausbrüche von Gletscherseen. Jahrb. Schweiz. 

nat.forsch. Ges., 158, 170–212. 

- Ulivieri, G., Marchetti, E., Ripepe, M., Chiambretti, I., De Rosa, G., & Segor, V. (2011). 

Monitoring snow avalanches in Northwestern Italian Alps using an infrasound array. 

CRST, 69(2-3), 177–183. 

- Walter, F., Deichmann, N., & Funk, M. (2008). Basal icequakes during changing subglacial 

water pressures beneath Gornergletscher, Switzerland. J. Glac., 54(186), 511–521. 



M-AARE - Coupling atmospheric, hydrological, hydrodynamic 

and damage models in the Aare river basin, 

Switzerland 

Andreas Paul Zischg, PhD 1, 2 ; Guido Felder, Msc 2 ; Rolf Weingartner, Prof. 2 ; Juan José Gómez-Navarro, PhD 2 ; Veronika 

Röthlisberger, Msc 2 ; Daniel Bernet, Msc 2 ; Ole Rössler, PhD 3 ; Christoph Raible, Prof. 4 ; Margreth Keiler, PD 5 ; Olivia Martius, Prof. 2 

ABSTRACT 

The triggering mechanism and the temporal evolution of large flood events, especially of 

worst-case scenarios, are not yet fully understood. Consequently, the cumulative losses of 

extreme floods are unknown. To study the link between weather conditions, discharges and 

flood losses it is necessary to couple atmospheric, hydrological, hydrodynamic and damage 

models. The objective of the M-AARE project is to test the potentials and opportunities of 

a model chain that relates atmospheric conditions to flood losses or risks. The M-AARE model 

chain is a set of coupled models consisting of four main components: the precipitation 

module, the hydrology module, the hydrodynamic module, and the damage module. 

The models are coupled in a cascading framework with harmonized time-steps. First 

exploratory applications show that the one way coupling of the WRF-PREVAH-BASEMENT 

models has been achieved and provides promising new insights for a better understanding 

of key aspects in flood risk analysis. 

KEYWORDS 

model coupling; worst-case flood; flood risk; Aare river; Switzerland 

INTRODUCTION 

In Switzerland, floods are the major cause of significant economic losses. The amplitude of 

flood peaks and the flood volume depend on the intensity and track of the triggering 

precipitation events, the topography and geology of the catchments, the wetness of the 

catchments prior to precipitation events as well as the hydro-morphologic conditions in the 

floodplains. However, the detailed triggering mechanism and the temporal evolution of 

large flood events, especially of worst-case scenarios, are not yet fully understood. Regarding 

mesoscale catchments, insights on the precipitation patterns leading to the most extreme 

floods are missing. Consequently, the cumulative losses of worst-case floods are unknown. 

The knowledge of the worst-case flood or of flood discharge return periods of up to 

10’000 years are important for managing critical infrastructures as well as financial risks, e.g., 

1 University of Bern, Bern, SWITZERLAND, andreas.zischg@giub.unibe.ch 

2 Mobiliar Lab for Natural Risks, Oeschger Centre for Climate Change Research, Institute of Geography, University of Bern 

3 Oeschger Centre for Climate Change Research, Institute of Geography, University of Bern 

4 Physics Institute, Oeschger Centre for Climate Change Research, University of Bern 

5 Institute of Geography, University of Bern 



for portfolio management of insurance companies. On a longer time scale, the question how 

the expected changes in precipitation intensities due to climatic changes influence the flood 

risk is of special interest. To study the link between weather conditions, discharges and flood 

losses it is necessary to couple atmospheric, hydrological, hydrodynamic and damage models. 

An attempt for coupling hydrologic with hydraulic models for flash flood predictions has been 

shown by Laganier et al. (2014). Various examples of coupling process models and vulnerability 

models were elaborated in the CRISMA project (Heikkilä et al. 2015). The focus in this 

project laid on the improvement of crisis management by simulating complex crisis scenarios 

of winter storm events, coastal submersion processes, earthquakes and for forest fires. 

An example of coupling hydrologic, hydrodynamic and damage models is given by Kourgialas 

and Karatzas (2013). 

The objective of the project “M-AARE – Coupling atmospheric, hydrological, hydrodynamic 

and damage models in the Aare river basin” is to test the potentials and opportunities of a 

model chain that relates atmospheric conditions to flood risks. Thus, the main question is 

whether the coupling of atmospheric, hydrologic, hydrodynamic and damage models could 

potentially contribute to a better understanding of the formation of flood events and their 

consequences. Another aim of this explorative study is to quantify the resources needed for 

simulating these processes in a model chain. 

With the model chain, the discharges from the catchments to the floodplains for selected 

precipitation scenarios and the retention effects in lakes and floodplains should be quantified. 

This allows to predict the flooded areas and the related losses to exposed residential buildings. 

Beside the hydro-meteorological characteristics, a key aspect of the method is to characterize 

and capture the non-linear effects of flood retention in the valley bottom and in the lakes. 

Furthermore, the model chain will allow the quantification of potential losses for given 

scenarios based on flow depths and flow velocities and therefore provide a sound basis for 

risk analysis. 

The model chain was developed and tested in the watershed of the river Aare upstream of 

Bern, Switzerland, with an area of approx. 3000 km 2 . This river basin is a complex network 

of sub-catchments with different runoff characteristics including two larger regulated lakes. 

Most of the rivers are trained since the 18th and 19th century. 

METHODS 

The M-AARE model chain is a set of coupled models consisting of four main components: 

the precipitation module, the hydrology module, the hydrodynamic module, and the damage 

module (see Fig. 1). The selected models in each module will be inter-changeable or can be 

used in an ensemble-framework for further sensitivity and uncertainty assessments. 

Precipitation module 

The precipitation module provides precipitation scenarios as inputs for the rainfall-runoff 

model. The latter is set up for each tributary and delivers the input hydrographs for the 

hydrodynamic model. The precipitation scenarios are formulated using two different 

approaches: a) by defining representative spatio-temporal precipitation patterns represented 


in gridded datasets or b) by selecting extreme precipitation events from a long climate simulation 

of a Global Circulation Model (CESM1) and downscaling these selected cases with 

a Regional Climate Model (WRF). The first approach was used to estimate the probable 

maximum precipitation (PMP), which is done by applying a Monte Carlo approach. The 

identified spatio-temporal distributions with the most severe impacts feed subsequent models. 

For the second approach, a long-term climate simulation (a control simulation spanning more 

than 500 years) with the Earth System Model (ESM) provides a coarse-resolution dataset of 

several centuries of precipitation. From this data set, a number of case studies corresponding 

to extreme situations are selected as candidates for further analysis. However, the global 

model employs a coarse spatial resolution (1 degree) that precludes the accurate simulation of 

the precipitation in areas of complex topography such as Switzerland. Hence, these cases 

need to be dynamically downscaled with a Regional Climate Model (RCM). The applied RCM 

WRF implements a spatial resolution of 2 km over the entire alpine area, which allows a 

more realistic representation of precipitation induced by interactions between the large-scale 

forcing and orography. Outputs of both approaches in this module consist of gridded time 

series of temperature and precipitation of a selected number of scenarios. 

Figure 1. Conceptional setup of the model chain M-AARE in the Aare river basin. 

Hydrologic module 

For the rainfall-runoff modelling, we apply the hydrological model PREVAH (Viviroli et al. 

2009). The model is set up for 15 sub-catchments that are located within the Aare basin 

with a spatial resolution of 1 km and hourly time steps. The delimitation of the catchments 

are presented in Fig. 2. The model is fed by the precipitation scenarios described above. 

The model output of the hydrologic module is used as the upper boundary condition of the 

hydrodynamic model. 


Hydrodynamic module 

The generated hydrographs are then routed with the hydrodynamic model BASEMENT-ETH 

(Vetsch et al. 2015) that accounts for the retention effects of lakes and floodplains. The 

hydraulic model was set up in two ways, a 1D- hydrodynamic model for all research 

questions regarding the flow routing only and a 2D-model used for coupling the damage 

module. The 1D model consists of cross sections along the whole valley bottom and considers 

the characteristics of the two lakes (lake Thun and lake Brienz) in modulating the flood 

hydrographs from the upper catchments. All of the main rivers considered in the 1D model 

were coupled in one integrated hydrodynamic model. This model consists of the Aare river 

from Meiringen to Bern including the two lakes and the area between the two lakes, the 

Gürbe valley, the Lütschine valley downstream from Gsteig, the Kander river downstream 

from the confluence with the Simme river. The river reaches in the main floodplains are 

implemented also in a 2D hydrodynamic model. Therefore, depending on the research 

question or on the damage model applied, these river reaches can either be modelled in 1D 

or in 2D, respectively. The 1D model was used for studying worst case discharges at the basin 

outlet in Bern. The 2D model was used to delimitate the flooded areas and as an input for the 

damage module. The spatial setup of the interface between the hydrologic and the hydraulic 

models is shown in Fig. 2. 

Figure 2. Spatial setup of the interface between hydrologic and hydrodynamic models. The black lines show the catchment delimitations. 

The calibrated catchments are indicated by labels. The black triangles indicate where the output of the hydrologic model will be used as 

input for the hydrodynamic model. The hatched areas represent the floodplains modelled in a 2D hydrodynamic model. The damage 

model is applied only in these areas. 


Damage module 

The hydrodynamic model – run in 2D mode – provides the basis for the damage module. This 

module consists of a dataset of buildings, each object classified by type, functionality, construction 

period, volume, reconstruction costs, and number of residents. The flood intensity maps 

resulting from the hydrodynamic module lead to the calculation of the object-specific 

vulnerability and therefore to the estimation of object-specific losses. The cumulative losses of 

a simulated precipitation scenario are summed up in a second step. Currently, a vulnerability 

function based on insurance data and reconstructed flood events is implemented. The method 

for the elaboration of the vulnerability curve follows the approach of Papathoma-Köhle et al. 

(2015), adopted to flood damages based on insurance data in Switzerland. The loss of life is 

calculated after Jonkman et al. (2008). 

Coupling strategy 

After Laganier et al. (2014), the model coupling strategy can either be of unidirectional or 

of bidirectional type. The first case is also called external coupling or a model cascade; the 

information is exchanged in one direction only. In the second case, the coupled sub-models 

interact between each other. In our case, the models are coupled in a cascading framework 

with harmonized time-steps. The coupling of the modules is controlled in a central timing 

and control device. 

Calibration and validation 

Each of the sub-models is calibrated and validated separately. The precipitation module is 

bias-corrected against gridded data sets of observations of precipitation in Switzerland. The 

hydrologic model is calibrated, if available, with observation data at the outflow of each 

sub-catchment (8 gauged sub-catchments). The models for the ungauged sub-catchments are 

regionalized by applying the parameter regionalization method proposed by Viviroli (2011). 

The hydrodynamic model was calibrated by empirically adjusting the friction coefficients. 

The values were adjusted by reconstructing observed flood events with particular regard to 

the water surface elevation in the main channel at peak discharge and the runtime of a peak 

discharge from one gauging station to another. The hydrodynamic model was validated based 

on watermarks along the rivers measured during the flood event in June 2014. The computed 

water surface elevations are within +/- 30 cm at nearly bankfull discharge. The 2D 

hydrodynamic model is calibrated in terms of reproducing the known channel capacity of the 

river reaches and in terms of reproducing the flooded areas of known flood events of 2005. 

The validation of the modelled flooded areas could not be quantified directly because the 

river geometry changed remarkably in some river reaches since the last observed floodings 

due to river training works. The damage model was validated in terms of reproducing the 

order of dimension of observed cumulative losses in past flood events. A direct validation of 

the damages to buildings was not possible because of lacking data at single objects level. 


Figure 3. Flooded areas and losses to buildings in the subcatchment Gürbe related to a hydrograph resulting from one probable 

maximum precipitation scenario. The map at the left shows the flow depths at peak discharge, the diagram at the right shows the inflow 

hydrograph (continuous line) and the computed losses (grey area) over the time axis (hours). 

RESULTS 

The main result of the M-AARE project is the set up of a modelling chain of one-way coupled 

deterministic models. The first simulations of the model chain show that the chosen settings 

are suitable for modelling these natural processes, from precipitation to floods and flood 

losses. The meteo module provided numerous precipitation scenarios with different spatio-temporal 

distributions. These scenarios provide the input for the hydrologic model and the 

resulting discharges feed into the hydrodynamic model. The use of the downscaled global 

circulation model with a regional climate model showed that the latter is able to improve the 

simulation of precipitation compared to the GCM alone. Although, the large-scale flow and 

the location of the precipitation maxima is very similar at continental scales (as it is driven by 

the boundary conditions provided by the GCM) the spatial structure of the precipitation is 

refined at mesoscale scales, producing stronger precipitation gradients that allow to identify 

the main orographic barriers. Furthermore, much higher precipitation rates occur in some 

river catchments, which are indicative of potential disastrous situations at localised regions. 

The setting of the hydrodynamic model is able to consider retention capacities of lakes and 

floodplains, and to investigate the relationship between characteristics of process intensities 

and the related damages to residential buildings. The coupling between hydrologic and 

hydrodynamic models indicate that the relation between the input peak discharge and the 

modelled peak discharge at the outlet of the floodplains shows non-linear effects, which are 

usually neglected in extreme value statistical analyses. Further results of coupling the 

hydrologic and hydrodynamic models in probable maximum flood analyses are described in 

Felder et al. (subm.). Fig. 3 shows exemplarily the result of one of the simulated worst case 


scenarios. The map in Fig. 3 shows the modelled flow depths at peak discharge. The diagram 

in Fig. 3 shows exemplarily a hydrograph of one probable maximum precipitation scenario in 

one of the subcatchments and the related losses on buildings. The evolvement of the losses 

during the flood event is shown on the same time axis as the discharge. 


The results show that the one way coupling of the modules has been achieved and provides 

promising new insights for a better understanding of key aspects in flood risk analysis. 

The described model configuration allows to route the precipitation through different states 

of the river system and to take the retention effects of lakes and floodplains into account. 

The modelled magnitude of the effect of retention areas highlights the importance of 

considering such effects in extreme discharge estimations. First exploratory applications with 

the coupling of the WRF-PREVAH-BASEMENT models show the importance of clearly 

defined interfaces between the models. The coupling entails that all of the models are flexible 

enough to meet the requirements for the exchange of data, especially taking into account the 

different temporal resolution of each model. It is shown, that both the PREVAH model and 

the BASEMENT model are suited to be chained together and both are flexible enough to be 

operated by an external controller. In our case, the most important aspect is to harmonise the 

different time steps by using one framework. Therefore, the controller module is of strategic 

importance. Another important point is the definition of the location in space where the 

rainfall-runoff model delivers the computed discharges to the hydrodynamic model. These 

interfaces are located on the upper edges of floodplains in which remarkable effects of flood 

retention are to be supposed. The first applications of the damage model show that the 

vulnerability functions are crucial for calculating the damages. This module needs to be 

further improved and validated. 

In conclusion, the M-AARE model chain has shown that the coupling of deterministic models 

offers a high potential to address further research questions and offers opportunities to 

provide a sound framework for different tasks in flood risk management. However, a successful 

implementation requires a high demand on specific knowledge and an interdisciplinary 

approach. Each of the modules needs knowledge and the coupling itself requires a rigorous 

definition of the interfaces between the models and an expertise in setting up of the controller 

module. Overall, the described model chain may provide the basis for further investigations: 

– The model chain will simulate more scenarios of physically plausible peak discharges in the 

study area that are determined by the most extreme situations leaded by the large-scale 

circulation within the GCM. This will enable the analysis and characterization of worst-case 

floodings whose return period exceeds several centuries. 

– With this model chain, it is possible to quantify the cumulative effects of all river training 

works or to assess the sensitivity of river reaches to the effects of climatic changes. This 

allows analysing the lake regulation procedures in case of a worst case flood. 


– It allows forecasting flood damages on the basis of discharge forecasts in selected river 

reaches due to the analysed discharge-damage-relationships. 

– The implementation of a multi-modelling approach in the hydrologic, hydrodynamic and 

damage modules will provide the possibility to quantify and describe the uncertainties 

in more detail. 

– The M-AARE model chain may also provide a platform for planning of flood corridors and 

studying their effects in terms of flood hydrograph modulation on basin scale. 

– Up to now, only buildings are considered in the computation of losses. The damage module 

has to be extended to other categories, e.g. losses to infrastructures etc. 

REFERENCES 

- Felder, G.; Zischg, A.; Weingartner, R. (subm.). The effect of a coupling of hydrologic 

and hydrodynamic models on PMF estimation. 

- Heikkilä, A.-M.; Havlik, D.; Schlobinski, S. (2015). Modelling crisis management for 

improved action and preparedness. Espoo: VTT. 

- Jonkman, S. N.; Vrijling, J. K.; Vrouwenvelder, A. C. W. M. (2008). Methods for the 

estimation of loss of life due to floods: a literature review and a proposal for a new method. 

In: Nat Hazards 46 (3), 353–389. DOI: 10.1007/s11069-008-9227-5. 

- Kourgialas, Nektarios N.; Karatzas, George P. (2013). A hydro-economic modelling 

framework for flood damage estimation and the role of riparian vegetation. In: Hydrol. 

Process. 27 (4), S. 515–531. DOI: 10.1002/hyp.9256. 

- Laganier, O.; Ayral, P. A.; Salze, D.; Sauvagnargues, S. (2014). A coupling of hydrologic and 

hydraulic models appropriate for the fast floods of the Gardon River basin (France). 

In: Nat. Hazards Earth Syst. Sci. 14 (11), 2899–2920. DOI: 10.5194/nhess-14-2899-2014. 

- Papathoma-Köhle, M.; Zischg, A.; Fuchs, S.; Glade, T.; Keiler, M. (2015). Loss estimation 

for landslides in mountain areas – An integrated toolbox for vulnerability assessment and 

damage documentation. In: Environmental Modelling & Software 63, 156–169. 

DOI: 10.1016/j.envsoft.2014.10.003. 

- Viviroli, D.; Zappa, M.; Gurtz, J.; Weingartner, R. (2009). An introduction to the hydrological 

modelling system PREVAH and its pre- and post-processing-tools. In: Environmental 

Modelling & Software 24 (10), 1209–1222. DOI: 10.1016/j.envsoft.2009.04.001. 

- Viviroli, D.; Weingartner, R. (2011). Umfassende hochwasserhydrologische Beurteilung 

ungemessener mesoskaliger Einzugsgebiete im Schweizerischen Rheineinzugsgebiet durch 

prozessorientierte Modellierung. In: Hydrologie und Wasserbewirtschaftung, 55(5), 258-272. 

- Vetsch, D.; Siviglia, A.; Ehrbar, D.; Facchini, M.; Gerber, M.; Kammerer, S.; Peter, S.; 

Vonwiler, L.; Volz, C.; Farshi, D.; Mueller, R.; Rousselot, P.; Veprek, R.; Faeh, R. (2015). 

BASEMENT – Basic Simulation Environment for Computation of Environmental Flow 

and Natural Hazard Simulation. ETH Zürich. 



Convivere 

con i rischi 

naturali

Hazard and Risk Assessment 

(analysis, evaluation) 


HAZARD AND RISK ASSESSMENT (ANALYSIS, EVALUATION) 

Understanding the impact of climate change on 

debris-flow risk in a managed torrent: expected future 

damage versus maintenance costs 

Juan Antonio Ballesteros Canovas, PhD. 1 ; Markus Stoffel 2 ; Klaus Schraml 3 ; Christophe Corona 4 ; Andreas Gobiet 5 ; 

Satyanarayana Tani 5 ; Sven Fuchs 6 ; Franz Sinabell 7 ; Roland Kaitina 6 

ABSTRACT 

In this communication, we evaluate the role of maintenance costs of hydraulic infrastructures 

in the risk analysis of torrential channels under the impact of process and/or climate changes. 

We combine stochastic life-cycle analysis (LCA), debris-flow modelling and risk assessments 

to understand the cumulative effects of extreme debris-flow events and potential progressive 

degradation of infrastructure on mitigation structures. We compare two scenarios to assess 

the reliability of check dams in reducing debris flow risks, with a focus on their performance 

and maintainability. We detect that maintenance works will play an important role in the 

next decades to maintain the reliability of infrastructure at a high level of confidence, which 

will however result in high economic costs. 

KEYWORDS 

Debris flow, risk assessment, stochastic life-cycle analysis, cost-benefit 

INTRODUCTION 

Debris flows are categorized as one of the costliest natural hazard in mountain environments, 

causing repeated damage to infrastructures, urban development and even loss of life (Jakob 

and Hungr, 2005). In mountain areas, debris flows risk assessment is an important issue for 

practitioners and it could even become more crucial in the next decades due to (i) the 

expected changes in magnitude-frequency of processes due to climate, and (ii) the rapid 

socio-economic development of such environments (Totschnig and Fuchs, 2013). 

Disaster risk managers aim to reduce the expected losses based on passive (i.e. land-use 

management, hazard mapping) as well as active mitigation strategies (i.e. structural measurement, 

protection forest) (Holub and Fuchs, 2008). Active-based structural measures, such as 

retention basins, check dams and channel canalization are therefore common in Central 

1 University of Bern, Bern, SWITZERLAND, juan.ballesteros@dendrolab.ch 

2 Climate Change an Climate Impacts (C3i) Institute for Environmental Sciences 7 route de Drize, 1227 Carouge-Geneva 

3 University of Natural Resources and Applied Life Sciences (BOKU), Institute of Mountain Risk, Engineering (IAN), Vienna 

4 Centre National de la Recherche Scientifique (CNRS) UMR6042 Geolab, 4 rue Ledru, F-63057 Clermont-Ferrand Cedex, France 

5 University of Graz, Wegener Center for Climate and Global Change (WegCenter), Graz 

6 University of Natural Resources and Life Sciences, Institute of Mountain Risk Engineering, Austria 

7 Österreichisches Institut für Wirtschaftsforschung 



Europe (Mazzorana et al., 2012) and they have played an important role in vulnerability 

reduction. However, the reliability of these structures can be affected by their previous state. 

Given that major investments in active measures were realized over the last decades, it is 

hypothesized that structures may soon lose or may already have lost performance of 

hydraulic function due to attrition. Consequently, some of them may no longer present 

optimal states (Romang et al., 2003), which would however be crucial for efficient risk 

reduction (Sánchez-Silva, 2004). Under this premise, two main factors will be therefore 

critical in the next decades: (i) the possible increase in frequency and magnitude of debrisflow 

hazards related to more frequent extreme climatic conditions (precipitation, snowmelt 

events) and/or changes in land cover and land use; and (ii) the current and future state of 

reliability of existing infrastructures depending on their maintenance, repair, and potential 

system failures. 

In this communication, we present a coupled framework based on a time-dependent 

performance analysis of infrastructures in managed torrents affected by recurrent debris flow, 

their maintenance and operability and classical risk assessment procedures. By combining 

stochastic life-cycle analysis (LCA, Sanchez-Silva et al., 2011), debris flow modelling and risk 

assessment, we aim at understanding the cumulative effects of extreme events and progressive 

degradation on the design capacity of structures. We also aim at exploring the applicability 

of stochastic LCA to debris-flow risk assessment in a managed torrent watershed located in 

the southern part of the Austrian Alps where severe debris flows have been recorded in 1997. 

By comparing two alternatives (‘pre- and post-1997 event’), we assess the reliability and 

sustainability of check dams in reducing debris-flow risks, with a focus on their performance 

and maintainability. 

METHODS 

Study site: The study site is the Wartschenbach torrent, located in the southern Austrian 

Alps (Lienz district, province of East Tyrol). This torrent is characterized by a catchment area 

of 2.7 km 2 , with altitudes ranging from 670 to 2113 m a.s.l. The main channel is 3.6 km long 

and has an average slope gradient of 0.18 m/m, and maximum values up to 0.4 m/m in the 

central part (Melton index = 1,01). The apex of the sediment fan is located at 1460 m a.s.l 

and it has an average slope of 16°. The two villages of Nußdorf-Debant and Gaiming, located 

on the fan of the Wartschenbach torrent, have suffered severe damage from debris-flow 

activity in the past (Hübl et al., 2002). To achieve more effective protection of the exposed 

community, local authorities adopted a protection strategy focused on the successive 

implementation of a range of active measures (i.e. three open-slit check dams in the upper 

part of the fan). The historical record of events suggests that deposition heavily reduced the 

capacity of the dams between 1995 and 2000, requiring maintenance and cleaning work with 

an annual average cost of about 280,000 €. 


Methodological steps: We quantify expected losses due to debris-flow activity by using a 

coupled stochastic LCA risk assessment model (Figure 1; Ballesteros-Cánovas et al., in 

review). Expected losses were estimated as the sum of the costs related to (i) regular 

maintenance work of check dams to keep their original retention capacity after a debris-flow 

event; and (ii) damage downstream of the structure in case that a debris flow exceeds the 

maximum capacity of the dams. We considered two different alternatives for the development 

of infrastructures: i) pre-1997 event and, ii) post-1997 event. Additionally, we also 

considered three climate change (CC) scenarios (i.e. S1, S2 and S3) integrating expected 

changes in debris-flow frequency as a result of the expected increase in temperatures and 

extreme precipitation events (Gobiet et al., 2014). 

Figure 1: General description of the proposed method (Ballesteros-Cánovas et al., in review). 

We used the catalogue of historical natural disasters from the Lienz region (Hübl et al., 2011) 

to characterize the occurrence of debris-flow events. Then, we determined daily rainfall 

intensities prior to and at the day of occurrence of each event based on the Iselsberg-Penzelberg 

meteorological series. Debris-flow frequencies (i.e. number of events per decade) were 

then analyzed in terms of threshold exceedance (from 10 to 100 mm) by using the distribution 

of past daily rainfall triggers as a guide and basis for the assessment of how these 

thresholds could be exceeded in the future (Guzzetti et al., 2008). Finally, debris-flow 

frequencies were connected to statistically downscaled station data and error-corrected 

climate change scenarios available for fixed daily precipitation thresholds (comprised between 

10 to 100 mm) averaged over two months until the mid-21st century. For the purpose of this 

study, we used three contrasting climate scenarios derived from regional climate model. 

In the subsequent step, a stochastic LCA model (Sánchez-Silva et al., 2011) was implemented 

to determine the performance of existing dams during their life time (t=80 yrs). Two different 

alternatives (A) were considered: 


– (i) A1 (past situation, pre-1997 event): one dam (maximum retention capacity: 25,000 m 3 ). 

– (ii) A2 (current situation, post-1997 event): three check dams (maximum retention 

capacity: 77,000 m 3 ). 

Exponential distributions were used to characterize the average recurrence time of events for 

each scenario. In addition, LogNormal (LN) distributions were used to model expected 

magnitudes of events (i.e. the volume of deposition; m 3 ) based on historical information. 

Age-related degradation of check dam capacity was assessed with a monotonous-decrease 

time-dependent function by using observations from other torrents (Romang et al., 2003). 

During the LCA, two different check dam capacity limits can be taken into account namely 

Smin and Kmin. . Here, we have considered Smin as the average event capacity > 10,000 m 3 

resulting in a capacity loss requiring system maintenance according to historical records (Hübl 

et al., 2002), and Kmin as the maximum designed retention capacity volume for A1 = 25,000 

m 3 and A2 = 77,000m 3 .Building cost was assessed at 1 million € per dam and clearance (i.e. 

maintenance costs) at 5 € per m³ sediment retained behind the dam. This average value 

includes the cleaning of the dam as well as the transport, and further deposition of sediments; 

however other costs related with technical works during maintenance are not explicitly here 

considered (i.e. condition assessment, attendance, corrective maintenance). The stochastic 

LCA analysis has been performed using Monte-Carlo simulations with 1,000 iterations and 

takes account of a subsequent recovery of retention capacity up to 100 % when retention 

capacity of the check dams D[t] falls below Smin. In addition, the impact of performing 

periodic maintenance works was analyzed and different maintenance periods have been 

implemented in the model (i.e. each 2, 5, or 10 years). Their impact on the reliability of check 

dams has been evaluated in terms of number of recoveries, time to recovery, and average 

recovery cost. Here it was assumed that periodic maintenance works can improve check dam 

capacity by up to 10%. 

In case that Kmin is exceeded, deposition outside the channel causes damage to properties 

located on the fan. To estimate related cost with this scenario, we quantified damage 

functions based on available vulnerability curves (Fuchs et al., 2007; Totschnig and Fuchs, 

2013), intensity of the event (deposition depth in m), and economic value of houses. We then 

used the numerical simulation model Flo-2D to estimate expected debris-flow deposition (m) 

to each expected volume. Flo-2D is a bi-dimensional quadratic shear stress model describing 

regimes from viscous to turbulent/dispersive flow. The model was based on a 1-m digital 

elevation model and previously calibrated based on the well-documented event in August 

1997. ArcGis was used to determine the exposure degree of each building, whereas vulnerability 

curves and average economic properties values were used to define damage functions. 

Finally, a cost-based debris-flow risk assessment was based on the expected total annual cost 

incurred by the entire system, i.e. the sum of expected maintaining costs of check dams 

throughout the life cycle of works, and expected annual costs of debris-flow damage in terms 


of fatalities after a potential failure of the retention system was carried out. We compared 

the reliability of the entire system (different alternatives) under different scenarios based on 

the expected annual cost. In order to include all uncertainties from all modelling steps, we 

performed Monte Carlo simulations based on 5,000 iterations. 

RESULTS 

Analysis of the historical database suggests that more than 121 debris-flow and related events 

occurred in the Lienz region between 1950 and 2000 (mean: 21 events per decade). Threshold 

intensities extracted from the database for debris-flow initiation vary between 0 mm and 

123.3 mm day–1. Mean rainfall intensity for debris-flow triggering is 59 mm day–1. Maximum 

debris flow frequency was observed in the past during summer (S1: 5.2 events per 

decade in June-July) for daily rainfall intensities comprised between 10 and 20 mm. In the 

future, according to scenarios S2 (S3), the frequency of 10-20 mm rainfall will decrease 

(increase) by 30% (5%). As a consequence, the expected frequency of debris-flow events per 

decade will change accordingly: 

– i. S1 (reference scenario): 21 events/decade, 

– ii. S2 (best scenario): 14 events/decade (-33 %) 

– iii. S3 (worst scenario): 29 events/decade (+38 %). 

Figure 2 provides a graphical example of check dam performance as a result of debris-flow 

occurrences and ageing degradation along structure life (example of 1 and 10 simulations for 

A2). Results indicate that the expected damage resulting from debris flows in the Wartschenbach 

in A1 is almost 98 % lower than in A2. The mean expected annual damage for the 

scenarios were 56,955 € (S1), 78,653 € (S2) and 37,970 € (S3) for A1, and 1,236,753 € (S1), 

1,659,085 € (S2) and 790,179 € (S3) for A2. Changes in debris-flow frequency as a result of 

climate change are predicted to induce a range of variations in expected annual costs which 

are comprised between +38 % (S2) and -33 % (S3). 

Figure 2: Example of check dam performance analysis. Left (right) panel represents 1 (10) simulations. In both cases, the evolution 

of check dam performance has been evaluated for the mitigation scenario A1 (3 check dams) and the climate change scenario S1 

(current situation, 21 debris flow per decade (Ballesteros-Cánovas et al., in review). 


The cost related with system performance under climate change forcing vary between 

2,336,988 € (3,126,953 €) and 3,423,836 € (5,988,019 €) for A1 (A2), respectively. 

The total expected annual costs (expected damage including expected maintenance costs) 

are on average lower in A1 (S1=82,690 €; S2=114,459 €; S3=59,227 €) as compared to A2 

(S1 = 1,236,753 €; S2 = 1,659,085 €; S3 = 790,179 €) regardless of the scenario considered. 

As a consequence, we interpret that the performance of A1 is more economical than A2. 

However, when the initial installation cost of each check dam (estimated at 1 million € per 

dam) is weighted against the recovery factor (80-yr lifetime with a 5 % interest rate), the 

expected annual cost for A1 will vary between 212,315 and 267,547 €; whereas for A2 it is 

ranked between 1,710,114 € and 841,208 €. As a consequence (Figure 3), the annual net 

benefit without considering the installation cost is optimal for the current alternative 

(+14 % vs -9.5 % ), however, if the installation cost is included in the analysis, the current 

alternative seems much less cost-effective (-185 % versus -9.5 %). 

Figure 3: Annual net benefit comparison (in %) taking account of or ignoring the installation cost of the check dams. (*) indicates the 

reference level at which the appraisal was performed (Ballesteros-Cánovas et al., in review). 

CONCLUSIONS 

The coupled stochastic LCA risk analysis allowed for an estimation of expected annual costs 

related to debris-flow hazards in a managed torrent catchment under future climate change 

conditions. Through the use of a stochastic analysis, uncertainties related to the performance 

of the existing infrastructures and the quantification of damage have been considered. 

Despite the large and uncertainties step involve in the risk system analysis, the coupled 

stochastic LCA risk analysis has been demonstrated as being a powerful tool to integrate most 

of processes involved in the economic analysis. We conclude that integrated analysis 


incorporating deterioration and maintenance operation cost should be included in future 

analysis of debris flow on managed torrents. Therefore, the modeling approaches clearly 

highlighted the prominent role of maintenance operation works needed to maintain the 

reliability of infrastructure at a high level of confidence for our case study site. 


This project received financial support from the Austrian Climate and Energy Fund and is 

carried out within the framework of the `ACRP´ Program (project number B060372). 

REFERENCES 

- Ballesteros Cánovas J.A., Stoffel M., Schraml K. Corona C., Gobiet A., Sinabell F., Fuchs S. 

,Kaitna R. (in review). Debris flow risk analysis in a managed torrent based on a stochastic 

life-cycle assessment. Landslide. 

- Fuchs S., Heiss K., Hübl J. (2007). Towards an empirical vulnerability function for use in 

debris flow risk assessment. Natural Hazards Earth System Science 7: 495-506. 

- Guzzetti F., Peruccacci S., Rossi M., Stark C.P. (2008). The rainfall intensity-duration control 

of shallow landslides and debris flows: an update. Landslides 5: 3–17. 

- Gobiet A., Kotlarski S., Beniston M., Heinrich G., Rajczak J., Stoffel M. (2014). 21st century 

climate change in the European Alps – A review. Science of Total Environment 493: 1138– 

1151. 

- Holub M., Fuchs S. (2008). Benefits of local structural protection to mitigate torrent-related 

hazards. WIT Trans. Info. Comm. 39: 401-411. 

- Hübl J., Ganahl E. Schnetzer I. (2002). Dokumentation Wartschenbach, IAN Report, 52, 

Institut für Alpine Naturgefahren, Universität für Bodenkultur, Wien. 

- Hübl J., Fuchs S., Sitter F., Totschnig R. (2011). Towards a frequency-magnitude relationship 

for torrent events in Austria. In: Genevois, R., Hamilton, D., Prestininzi, A. (eds) Proceedings 

of the 5th International Conference on Debris-flow hazards mitigation: mechanics, prediction 

and assessment. Casa Editrice Università La Sapienza, Padova, pp. 895-902. 

- Jakob, M., Hungr, O. (2005). Debris-flow hazards and related phenomena. Springer, Berlin. 

- Mazzorana B., Levaggi L., Keiler M., Fuchs S. (2012). Towards dynamics in flood risk 

assessment. Nat. Hazards Earth Syst. Sci. 12 (11): 3571-3587 

- Romang H., Kienholz H., Kimmerle R., Böll A. (2003). Control structures, vulnerability, 

cost-effectiveness – a contribution to the management of risks from debris torrents. In: - Rickenmann, 

D., Chen, C., (eds.) Debris-flow hazards mitigation: mechanics, prediction and 

assessment. Millpress, Rotterdam, pp. 1303-1313. 

- Sanchez-Silva M., Klutke G. A., Rosowsky D.V. (2011). Life-cycle performance of structures 

subject to multiple deterioration mechanisms. Struct. Safety. 3: 206-217. 

- Totschnig R., Fuchs S. (2013). Mountain torrents: quantifying vulnerability and assessing 

uncertainties. Eng. Geol.: 155, 31-44 



Unraveling the spatio-temporal debris-flow activity on 

a forested cone in the Kyrgyz Range: implications for 

hazard assessment 

Vitalii Zaginaev 1 ; Juan Antonio Ballesteros Canovas, PhD. 2 ; Markus Stoffel, Prof. Dr. 2,3 ; Sergei Erokhin 1 

ABSTRACT 

Ongoing climate change has recently resulted in an increase of risks related with glacial lake 

outburst flows (GLOFs) and debris flows in Northern Tien Shan, Kyrgyzstan. In this communication, 

we analyze recent process activity by applying tree-ring analysis to contribute to a 

better understanding of the past debris-flow activity. Based on the analysis of 96 disturbed 

trees, we reconstruct spatio-temporal patterns of events going back to 1877. A total of 26 

events have been reconstructed, revealing high activity in the 1960s and 1970s. Our results 

are in agreement with the existing, but very fragmentary historical records, and can be used 

as a baseline for both risk assessment and for the understanding of glacier-climate-debris flow 

linkages in the region. 

KEYWORDS 

debris flow, glacial lake outburst flows, GLOFs, tree-rings, Northern Tien Shan 

INTRODUCTION 

Debris flows are rapid mass movements in which a combination of loose soil, rock, organic 

matter, air, and water mobilize as slurry flows downslope. This natural process is considered 

one of the most common natural hazards in mountain environments and is often responsible 

for large damage to infrastructure and even loss of life. Especially powerful events are the 

so-called glacial lake outburst flows (GLOFs), which are formed through the breaching of 

moraine dammed or supraglacial lakes or subglacial reservoirs as a consequence of glacier 

dynamics. These extreme events shape debris cones in the Tien Shan Mountains of Central 

Asia and produce intense changes in depositional forms, affecting long distances downstream 

of fans and thereby disrupting inhabited valleys (Erochin et al., 2009a). 

The creation of new unstable glacier lakes in many high mountain environments and the 

destabilization of the cryosphere is explained by changes in extreme precipitation and 

temperature patterns (Stoffel and Huggel, 2012). Given the projected temperature increase 

in Northern Tien Shan, one can expect these processes to become an even greater challenge 

for local authorities and populations in the next future. 

1 Institute of water problems and hydropower, National Academy of Science, Bishkek, KYRGYZSTAN 

2 Dendrolab.ch, Institute of Geological Sciences, University of Berne, SWITZERLAND juan.ballesteros@dendrolab.ch 

3 Climatic Change and Climate Impacts, Institute for Environmental Sciences, University of Geneva, SWITZERLAND 



In this communication, we contribute to the understanding of past spatio-temporal occurrences 

of GLOFs at a case-study site in northern Tien Shan. We use tree-ring data to provide 

the longest annually-resolved GLOF record existing to date. Results are expected to be useful 

to re-define debris-flow hazard at the level of the cone level where touristic activities become 

increasingly important as well as provide findings about long-term changes in debris-flow 

frequency. 

METHODS 

Study site 

The study site is located on the Aksay cone (42˚33 Ń; 74˚29 ́E; Figure 1). This is the largest 

debris-flow cone of northern Tien Shan (Kyrgyz Mountain Range). Catchment size is 28.3 

km 2 with a relief of 2645 m. There are two glaciers systems located in the accumulation area 

on the western slopes of Semenov Tianshanskiy (4895 m asl) and Korona (4691 m asl) peaks, 

respectively. Vegetation on the cone is formed by Picea abies and Betula sp.. The mean annual 

precipitation is 517 mm (ranging between 18.1 mm in January and 87.1 mm in May), mean 

annual temperature is 6.5 °C. Snow is present from September to April. At this site, debris-flow 

(GLOF) activity has been observed in the recent past; generally during the summer 

season (Erochin et al., 2009a). Archival records from the Kyrgyz Hydrometeorogical Survey 

and State Agency of Geology contains data on nine debris flows triggered by GLOFs and one 

event triggered by rainfall between 1960 and 1999 (Erochin et al., 2009b). 

Figure 1: A. Location of study area, North slope of Kyrgyz range, National Ala-Archa park, Aksay valley. B. View on cone from west to east. 

C. View on cone from east to west. D. Sampled site on satellite image (2014). 

METHODOLOGICAL STEPS 

Initially, a geomorphic description combining both aerial images interpretation (from 1960, 

1971, 1978, 2014) and field surveys was carried out to characterize past and current debris- 


flow channels, as well as the main depositional areas and forest cover. Then, all disturbed 

trees located on the fan were sampled following standard dendrogeomorphic procedures 

(Stoffel and Corona, 2014). At least two increment cores were extracted per tree. Additionally, 

undisturbed trees were sampled on adjacent, unaffected sites to build a reference chronology 

to identify pointer years for cross-dating. 

Samples were prepared and measured following standard dendrochronological procedures. 

Tree rings were first counted and then measured with a precision of 0.01mm using a digital 

LINTAB positioning table connected to a Leica stereomicroscope and TSAPWin Scientific 

software. Growth disturbance (GD) related with past debris flow activity were then identified 

on the increment cores and included injuries, callus tissue, compression wood, abrupt growth 

increase and/or growth suppression. In samples from Picea abies, the occurrence of tangential 

rows of traumatic resin ducts (TRD) was used as an indirect indicator of scars (Stoffel, 2008). 

Debris-flow definition was based on the weighted index value (Wit) developed by Kogelnig-Mayer 

et al. (2011). This index considers the number and the intensity of GDs within 

each tree-ring series and the total number of trees available for the reconstruction. We 

applied detection thresholds based on previous experience form debris flow in Alps (Schneuwly-Bollschweiler 

et al., 2013). The criteria here used for debris flow definition was: if Wit > 

1,2; then we considered a sure event; if 1,2>Wit>0,8 then we considered a potential event; if 

Wit < 0,8, then we reject the possibility of event. During this process, a visual analysis based 

on the spatial distribution of affected trees on the fan was also performed to detect potential 

incongruences. The representation and visualization of disturbed trees during specific 

debris-flow events also allows interpretation of spatial patterns of past events in the current 

and past channels. During this analysis we exclusively focused on the location of breakout 

sites, the extension of affected areas, and the dynamic of existing channels. 

RESULTS 

Eleven past debris flow channels as well as the main fan deposits and the forest cover 

evolution were identified based on aerial pictures (Figures 2 and 3). In the field, very large 

deposits with boulder sizes of up to 10-14 m in diameter were found at the cone apex. Most 

affected areas were located in the central and southern parts of the cone, where the forest 

cover was completely removed on older images (this site now bears a 20-30-yr old forest). 

By contrast, the northern part of the cone was less affected. 

A total of 320 GD were identified in the 156 samples from 96 P. abies trees affected by past 

debris-flow activity. Based on tree-ring analysis, the older trees date back to AD 1850 and are 

located in the northern part of cone, whereas the youngest trees grow next to the channel 

and in southern part of the cone. Our tree-ring data reveals that the 6% of the analyzed GDs 

corresponded to injuries, 25% of the reactions were strong, 33% medium, and 36% weak. 


Figure 2: Aksay debris cone changes. A. aerial image 11.08.1960; B. aerial image 01.09.1971; C. aerial image 08.08.1978; D satellite 

image 30.08.2014 

Figure 3: Number of debris flow channels. For each aerial image counted a number of active and passive channels. B. Forest cover for 

each year, area in 2014 – 100% 

Based on this dataset, the temporal reconstruction of debris-flow events at Aksay cone is 

presented in Figure 4. The oldest reconstructed event was in 1877, whereas the more recent 

event took place in 1999. Based on historical records, nine dated debris flow events were 

related with lake outbursts floods being transformed into debris flows (1960, 1961, 1965, 

1966, 1968, 1969, 1970, 1975, and 1980), and only in one case, the dated event was related 

with intense rainfall (i.e. 1999). Higher Wit indexes were observed for the events recorded in 

1980 (Wit=22,3), 1969 (Wit=10,6), 1960 (Wit=12,8) as well as for the newly documented 

events in 1928 (Wit=15,6), 1936 (Wit=9,1), and 1950 (Wit=11,8). The more recent events 

were, by contrast, characterized by lower Wit indices (e.g. in 1993 – Wit=1,42, and 1999 – 


Wit=2,12). The most recent event occurred on 23 and 25 July 2015, i.e. during our fieldwork 

for this study, but did not yet leave evidence (GD) in the tree-ring series. Our results show 

increased debris-flow activity on the cone between the 1960 and 1970 (0.54 event/year; i.e. 

1960, 1961, 1965, 1968, 1969, and 1970), and a significant decrease from 1970s to the end of 

the 20th century (0.16 event/year: 1973, 1975, 1977, 1980, 1993 and 1999). 

Figure 4: Reconstructed frequency of debris flows on the Aksay debris cone. 

Spatial analysis of the positions of disturbed trees shows that channels 1, 2, 3, and 6 have 

been repeatedly active during past events (up to 13 events). The most recent events in 

channels 3 and 8 were dated to 1999 and observed in 2015 (23 and 25 July). In contrast, less 

activity (only 3 events) was recorded in channel 9, with a last event in 1969. Based on these 

observations, we identify two main spatial patterns at Aksay cone (Figure 5): 

Pattern A. 

Debris flows affecting the entire cone: Reconstructed events in 1885, 1922, 1928, 1936, 1950, 

1955, 1960, 1961, 1965, 1966, 1968, 1969, 1970, 1977, and 1980. Generally, these events are 

related with high-magnitude flow energy and large sediment transport where boulders with 

more than 1 m diameter can occur, with the breakout of surges from the main channels (1, 2, 

4) at the apex of the cone. Based the historical records, at least the events of 1960, 1961, 

1965, 1966, 1968, 1969, 1970, and 1980 were caused by the outburst of proglacial Lake 

Aksay. The same pattern has also been observed for the previously unrecorded events in 

1922, 1928, 1936, 1950, and 1955. This pattern is therefore considered congruent with the 

imprint of past GLOFs events. Under this pattern, existing infrastructure can be affected 

during future events. 

Pattern B. 

Debris flows affecting the southern part of the cone: Reconstructed events: 1877, 1916, 1918, 

1924, 1934, 1941, 1943, 1973, 1975, 1993, 1999, and 2015. Existing records suggest that 

these events may be triggered by rainfall (1999) or intensive snow and ice melting in the 

absence of rainfall (2015). In this pattern, debris flows are affecting exclusively channels 12 

and 13 located in the southern part of the cone. In comparison with pattern A, pattern B is 

characterized by a transport of smaller amounts of sediments insufficient to produce 


outbreaks at the cone apex. Consequently, channels located in the northern of the cone are 

not affected. Also, the capacity of sediment to produce temporal dams at Ala-Archa River is 

minimal. Most sediment is deposited in the un-vegetated area on the cone. Infrastructures are 

not affected under this pattern. 

Figure 5: Spatial patterns of past debris-flow events: (A) Sketch of the Aksay cone with the localization of the identified channels, 

forest cover and sampled trees. (B) Debris flows affecting a north and south parts of cone an event in 1936 (pattern A). (C) Debris 

flows in 1968 affected almost cone, using channels in north, central and south parts of cone. In some cases flows reached a main 

Ala-Archa river following channels 10, 9, 1 (pattern A). It was registered event after outbursting glacier water pocket (D) In pattern 

B trees with growth disturbances by the 1999 event found in channel 8 in a central part and channels 4 and 3 in a current Aksay river 

channel (pattern B). 

CONCLUSIONS 

The coupled tree-ring analysis and classical geomorphic inspection has allowed to provide the 

longest annually-resolved debris flow history of Northern Tien Shan. Our results suggest that 

at least 26 events took place between AD 1877 and 2015, allowing tracking process dynamics 

in 11 mapped channels which are inactive under normal conditions. Results here provided 

are in agreement with historical records from 1960 to 1970, and therefore point out the 

reliability of this reconstruction. Our results are the basis to understand the glacier-climate- 

debris flow linkages in the region, as well as for the design of more reliable hazard maps and 

subsequent integrated risk management. Therefore, it is expected that both defined pattern 

will be useful to calibrate numerical model outcomes based on accurate topography to define 

future GLOFs/debris flow event scenarios. Moreover, the average occurrence of events here 

reported may be useful to define the probability of occurrence, and consequently the hazard 

level at the cone. Taken into account that the national natural park Ala-Archa is a popular 

tourist place, especially in summer season, our event pattern may be used to re-define hazard 

areas. This is specially the case in the central part of a cone (channels 4 and 8), where many 

tourist facilities have been constructed in hazard areas during last years. 



This study was funded by the Swiss National Science Foundation (SNF, Schweizerischer 

Nationalfonds zur Forderung der Wissenschaftlichen Forschung) in the framework of the 

DEFENCC (n° 152301; Future DEbris Flows and lake outburst floods in Tian Shan: possible 

impacts of projected Climate Change) project. 

REFERENCES 

- Erochin, S.A., Mamatkanov, D.M., Tuzova, T.V., 2009 a. Monitoring of Kyrgyz lakes at risk 

of outburst floods. Int. Conf. Floods. Tbilisi, Georgia, pp.130-147. 

- Erochin, S.A., Cerny M., 2009 b. Monitoring of outbursting lakes of Kyrgyzstan. Final Proc. 

Int. Conf. Mauntain-hazards, Bishkek, Kyrgyzstan, pp. 30–34. 

- Kogelnig-Mayer, B., Stoffel, M., Bollschweller, M., Hubl, J., Rudolf-Miklau, F., 2011. 

Possibilities and imitations of dendrogeomorphic time-series reconstructions on sites 

influenced by debris flows and frequent snow avalanche activity, Arct. Antarc. Alp. Res. 43 

(3), 649–658. 

- Schneuwly-Bollschweiler, M., Corona, C., Stoffel, M., 2013. How to improve dating quality 

and reduce noise in tree-ring based debris-flow reconstructions. Quat. Geochronol. 18, 

110–118. 

- Stoffel, M., 2008. Dating past geomorphic processes with tangential rows of traumatic resin 

ducts. Dendrochronologia 26 (1), 53-60. 

- Stoffel, M., Huggel, C., 2012. Effects of climate change on mass movements in mountain 

environments. Progress in Physical Geography 36, 421–439. 

- Stoffel, M., Corona, C., 2014. Dendroecological dating of geomorphic disturbance in tress. 

Tree Ring Res. 70, 3-20. 



Flood volume estimation in Switzerland using 

synthetic design hydrographs - a multivariate 

statistical approach 

Manuela Irene Brunner 1,2 ; Olivier Vannier, Dr. 1 ; Anne-Catherine Favre, Prof. 1 ; Daniel Viviroli, Dr. 2,3 ; Paul Meylan 4 ; 

Anna Sikorska 2 , Dr.; Jan Seibert, Prof. 2,5 

ABSTRACT 

Accurate estimations of flood peaks, volumes and hydrographs are needed to design safe and 

cost-effective hydraulic structures. In this study, we propose a statistical approach for the 

estimation of the design variables peak and volume by constructing a synthetic design 

hydrograph. Our approach is based on fitting probability density functions to observed flood 

hydrographs and takes the dependence between the two variables peak and volume into 

account. The method consists of the following six steps: sampling of flood events, baseflow 

separation, normalization of the hydrographs, fitting of the hydrographs with statistical 

density functions, modeling of peak and volume considering their dependence, and construction 

of the synthetic design hydrograph. The method was developed and tested based on data 

from nine meso-scale catchments in Switzerland, and has been shown to provide reliable synthetic 

design hydrographs for all of these catchments. While the method has so far been 

applied to gauged catchments, it is foreseen to make it applicable to ungauged catchments 

using regionalization approaches. 

KEYWORDS 

Synthetic design hydrographs; flood volume estimation; bivariate analysis, copulas 

INTRODUCTION 

An accurate flood estimation is needed for resilient flood risk management, to design 

hydraulic structures such as dam spillways, bridges, road culverts, and levees, and to manage 

residential area zoning, floodplains, and urban design. The major quantity of interest in flood 

estimation is the magnitude of the flood peak for a specific return period (Rosbjerg et al., 

2013). However, flood peaks provide only a limited description of a flood event. For the 

prevention of flood damage and for designing hydraulic structures, it is also important to 

know the flood volume and the shape of the entire flood hydrograph. The aim of this study 

was to develop a flood volume estimation method for meso-scale catchments (i.e. 20-1000 

km 2 ) which can easily be applied by practitioners. Although different methods to derive 

1 Université Grenoble-Alpes, Grenoble INP, LTHE, Grenoble, FRANCE, manuela.brunner@geo.uzh.ch 

2 Department of Geography, University of Zurich, SWITZERLAND 

3 belop gmbh, Sarnen, SWITZERLAND 

4 AlC Ingénieurs Conseil SA, Lausanne, SWITZERLAND 

5 Department of Earth Sciences, Uppsala University, Uppsala, SWEDEN 



design flood hydrographs are described in the literature, most of them are of limited use for 

practitioners because of their complexity (Yue et al., 2002). 

Therefore, we propose a statistical approach for the estimation of not only flood peaks but 

also flood volumes and the entire flood hydrographs using synthetic design hydrographs. 

The method has been developed as a simple tool that can support practitioners in consulting 

or engineering companies with reasonable efforts. The focus lies on meso-scale catchments 

with natural runoff conditions, i.e. without significant human alterations, and with an 

insignificant degree of glacierized areas. The basic idea relies on fitting probability density 

functions to observed flood hydrographs, while considering the dependence between flood 

peak and flood volume. This is essential because flood events are multivariate and a univariate 

frequency analysis does not allow for a complete assessment of the probability of their 

occurrence (Yue et al., 2002). The approach does not rely on a rainfall-runoff model, which 

makes it less demanding with respect to input data and methods. The transfer to ungauged 

catchments is foreseen in a next step. 

DATA 

The proposed method has been developed and tested using data from a representative set of 

nine meso-scale study catchments in Switzerland. The selected catchments cover different 

sizes, elevations, and regime types. To avoid hydrograph shapes modified by direct human 

impacts, we selected only catchments with flow conditions neither altered through hydropower 

plants nor lake regulation. For the development of the method, long term observations 

were required. We also exclude highly glacierized catchments because unimpaired flow 

records are scarce. Moreover, they are usually only sparsely populated and therefore exhibit a 

low damage potential. The characteristics of the nine study catchments selected are listed in 

Table 1. 

Table 1: List of study catchments and their characteristics. 


METHODS 

The method for the construction of synthetic design hydrographs (SDHs) relies on the fitting 

of statistical density functions to observed flood hydrographs considering the dependence 

between the design variables Q max 

and V. This is crucial, because a univariate frequency 

analysis of Q max 

or V alone cannot provide an accurate evaluation of the corresponding 

probabilities. The bivariate analysis was implemented using a copula model (Genest and 

Favre, 2007). The entire method of the SDH construction is divided into six main steps 

(Figure 1). 

Figure 1: Overview of the method proposed for the construction of synthetic design hydrographs. 

These steps include: 

1. Flood sampling: The SDH construction method is based on observed runoff data. Therefore, 

historical flood event hydrographs were selected in nine Swiss catchments (Table 1) 

independently from precipitation information. We used a peak-over-threshold (POT) 

approach to sample flood events. The threshold for the peak discharge was chosen 

iteratively to fulfill a target condition of a defined number of events per year (here four). 

The independence between successive events was ensured by prescribing a minimum time 

interval between them (here 72 hours). According to the extreme value theory, POT values 

follow a generalized Pareto distribution (GPD) (Coles, 2001). Therefore, we used a GPD 

to fit the peak discharge values. The volume values, on the contrary, follow a generalized 

extreme value distribution (GEV) because the threshold was only applied to the peak 

discharge and not to the volume. The goodness-of-fit of the GPD to the peak discharges 

and the GEV to the volumes was found to be good using the Akaike and the Bayesian 

information criteria (Meylan et al., 2012). 


2. Baseflow separation: The SDH approach describes only the quick flow component of the 

event hydrograph. Thus, it is necessary to distinguish between the slow and the fast runoff 

components to analyze the statistical properties of flood hydrographs (Yue et al., 2002). In 

this study, we applied a recursive digital filter (Eckhardt, 2005), whose two parameters 

need to be estimated for each catchment. This method allows for the separation of the 

baseflow from the quick flow and is easily applicable to a wide variety of catchments and 

provides reliable results. 

3. Normalization of the hydrographs: The quick flow component of the hydrographs was 

normalized so that both the base width and the volume of the modified hydrographs were 

equal to one. This was done by dividing the base width of each flood hydrograph by its 

duration D and then dividing the ordinate of each hydrograph by the mean runoff (V/D). 

4. Fitting of the normalized hydrographs with statistical density functions: The shape of a 

normalized hydrograph can be fitted by a probability density function (PDF) because both 

the area under the normalized hydrograph and the area under the PDF are equal to one. 

To select the best density, eight different PDFs were fitted to all of the normalized hydrographs 

in the nine study catchments: Normal, Lognormal, Fréchet, Weibull, Logistic, 

Gamma, inverse Gamma, and Beta (Nadarajah, 2007; Serinaldi and Grimaldi, 2011). The 

parameters of the distributions were estimated so that the PDFs approximate the shape of 

the normalized hydrographs as well as possible. The goodness-of-fit of each distribution 

function was ranked according to the three performance criteria: Nash-Sutcliffe efficiency 

(NSE), volumetric efficiency (VE), and correlation coefficient (R2). In each catchment, a 

representative normalized hydrograph was determined as the PDF that best fitted the median 

normalized hydrograph. We chose the median normalized hydrograph instead of the 

mean of the normalized hydrographs because it refers to a real event, which is not the case 

for the mean of the normalized hydrographs. 

5. Modelling the statistical dependence between Q max 

and V using copulas: The pair of design 

variables, Q max 

and V associated with a defined return period T was estimated using the 

marginal distributions of the variables and a copula to model their dependence. For the 

marginal distributions, we assumed a GEV distribution for the flood volumes and a GPD for 

the peak discharges. Copulas are multivariate distribution functions whose marginal 

distributions are uniform. The main advantage of this approach is that the selection of an 

appropriate model for the dependence between variables, represented by the copula, can 

then proceed independently from the choice of the marginal distributions (Genest and 

Favre, 2007). In contrast to standard multivariate distributions, copula models thus allow 

the variables to be characterized by different marginal distributions. The dependence 

between the two variables Q max 

and V was tested graphically by plotting all pairs of Q max 

and 

V and numerically by computing two rank correlation coefficients, Kendall's Tau and 

Spearman's Rho. Six copula models of the Archimedean copula family, namely the 

Gumbel, Clayton, Joe, Frank, Ali-Mikhail-Haq (AMH), and the independence copula plus 

the normal copula, were fitted and tested using both graphical approaches and a goodnessof-fit 

test based on the Cramér-von Mises statistic (Genest and Favre, 2007). A p-value for 


the Cramér-von Mises statistic of each copula was estimated using a statistical bootstrap 

procedure (Genest et al.,2009). The copula model with the best performance was used 

to estimate the two design variables for a given return period. 

6. Construction of the SDH: The SDH was obtained from scaling the representative normalized 

hydrograph by the two estimated design variables using: 

Formula 1 

where f(t) is the representative normalized hydrograph, and Q max 

and V are the design 

variables for a given return period T. In bivariate frequency analysis, in contrast to the 

univariate case, the definition of an event with a given return period is not clear (Yue and 

Rasmussen, 2002). The return period used to describe bivariate events can be defined in 

two different ways. The first of these ap proaches uses the conditional probability to 

determine a conditional return period, while the second approach uses joint probability 

distributions to calculate joint return periods (Gräler et al., 2013). Here, we relied on a 

conditional probability approach to estimate the design variable pairs Q max 

and V because 

the potential end-users of our approach are more familiar with conditional probabilities 

than joint probabilities. Thereby, the estimation of the variable V for the given return 

period was calculated according to its marginal distribution. Once the volume was 

estimated, the estimation of the second variable peak discharge could take place using the 

conditional cumulative distribution of the copula (Salvadori et al., 2011), as described in 

Step 5. The SDH was then calculated using the representative normalized hydrograph, 

scaled by the computed estimates of the design variables Q max 

and V. At the end of the 

procedure, the baseflow, removed from the hydrograph in Step 2 of the procedure, needs 

to be added to the SDH. 

RESULTS 

The proposed method was applied to a sample of nine meso-scale study catchments in 

Switzerland with an average record length of 40 years (Table 1). The results are illustrated 

by one of these study catchments, the Birse catchment at Moutier-la-Charrue. 

Eight PDFs were fitted to the normalized hydrographs that were derived from the raw data. 

In the Birse catchment, the Fréchet and Logistic distributions were most often ranked as the 

best PDFs for all considered performance criteria (Figure 2). Interestingly, the Fréchet 

distribution together with the normal distribution was also most often ranked as the worst 

PDF. The fact that the same statistical distribution was ranked both as the best and as the 

worst PDF could be due to differences in the causative mechanism of the flood events. 

The tendency of the Fréchet and Logistic distribution to provide good results was also 

observed in the other eight study catchments. 


Figure 2: Distributions of the ranks obtained by the eight statistical densities fitted to the normalized hydrographs in the Birse 

catchment. For each event, the eight densities were ranked according to the three performance criteria NSE, VE, and R2. 

Q max 

and V were clearly dependent in the Birse catchment with a Kendall's Tau of 0.323 and a 

Spearman’s Rho of 0.477. Figure 3 shows the observed pairs of Q max 

and V (red crosses) and 

10 000 pairs simulated using the seven fitted copula models (black crosses). 

Figure 3: Representation of the bivariate distribution of the variables Q max 

and V for the Birse catchment. The observations are plotted 

as red crosses. The black crosses represent 10 000 pairs of Q max 

and V values simulated according to one of the seven copulas which 

were fitted to the data and tested subsequently. 

Different behaviours among the modelled bivariate distributions can be observed. The 

Gumbel and Joe copulas tend to model a clear dependence between Q max 

and V with extreme 

values expanding towards the upper-right corner of the plot while other copulas do not 


indicate a dependence and rather behave in a way similar to the independence copula. As a 

quantitative assessment of a copula’s ability to represent the dependence between Q max 

and V, 

a goodness-of-fit test based on the Cramér-von Mises statistic was applied. This confirmed 

that the Gumbel and Joe copulas are best suited to model the dependence between Q max 

and 

V. The other tested copulas were rejected at a significance level of 0.05. This tendency is also 

visible in most of the other study catchments. 

The SDHs were computed using seven pairs of design variables, Q max 

and V, for the 100-year 

return period, computed with the seven different copulas mentioned above (Figure 4). The 

SDHs are associated with uncertainty envelopes which represent all the traces resulting of the 

Figure 4: SDHs for an event of a 100-year return period in the Birse catchment. The SDHs were calculated with seven copulas. The 

envelopes correspond to the eight statistical densities used for the fit. The hydrograph of the highest event (in terms of peak discharge) 

observed in the Birse catchment is shown as a reference. 

fits made with different statistical densities. There was a clear difference between all the SDHs 

derived using different copula models (Figure 4). All calculated values of Q max 

for a 100-year 

return period are larger than the value obtained by a univariate analysis (SDH independence). 

A univariate analysis would neglect the dependence between the two variables Q max 

and V. This tendency was observed in all nine study catchments. In addition, most of the 

computed values of the pair Q max 

and V exceed the highest recorded value. 


CONCLUSIONS & OUTLOOK 

The method presented proved to provide reliable SDHs for all nine study catchments. 

Although the method was developed using a small number of catchments, it can potentially 

be applied in any meso-scale catchment with no or small glacierized areas and unaltered flow 

conditions for which observational data are available. The major advantages of the method 

are its ease of application, its independence from a rainfall-runoff model, and the possibility 

to account for the dependence between peak discharge and volume. The importance of 

considering the dependence between peak discharge and volume was clearly shown in all 

study catchments, where the use of a classical univariate approach would have resulted in 

a clear underestimation of the magnitude of the design flood hydrograph compared to the 

approaches where the dependence between the two design variables peak discharge and 

flood volume is considered. 

In the next step, this method will be regionalized to ungauged catchments to also allow for 

the estimation of flood volume and hydrograph without any runoff measurements. Thus, 

several parameters used in the method need to be regionalized. To aid this regionalization, 

relationships between the modelled shape of hydrographs and a typology of causative flood 

mechanisms will be explored because different flood types are usually characterized by typical 

hydrograph shapes. Furthermore, relationships between catchment characteristics and 

parameters of the SDH will be quantified. 


We thank the Swiss Federal Office for the Environment (FOEN) for funding the project under 

contract 13.0028.KP / M285-0623 and for providing observed runoff data. We also thank the 

reviewers for their constructive comments. 

REFERENCES 

- Coles S. (2001). An introduction to statistical modeling of extreme values. Springer, 

London, 208. 

- Eckhardt K. (2005). How to construct recursive digital filters for baseflow separation. 

Hydrological Processes 19: 507-515. 

- Genest C. and Favre A.-C. (2007). Everything you always wanted to know about copula 

modeling but were afraid to ask. Journal of Hydrologic Engineering 12: 347-367. 

- Genest C., Rémillard B. and Beaudoin D. (2009). Goodness-of-fit tests for copulas: A review 

and a power study. Insurance: Mathematics and Economics 44: 199-213. 

- Gräler B., van den Berg M. J., Vandenberghe S., Petroselli A., Grimaldi S., Baets B. D. and - 

Verhoest N. E. C. (2013). Multivariate return periods in hydrology: a critical and practical 

review focusing on synthetic design hydrograph estimation. Hydrological Earth System 

Sciences 17: 1281-1296. 

- Meylan P., Favre A.-C. and Musy A. (2012). Predictive hydrology. A frequency analysis 

approach. Science Publishers. St. Helier, Jersey, British Channel Islands, 212. 


- Nadarajah S. (2007). Probability models for unit hydrograph derivation. Journal of 

Hydrology 344: 185-189. 

- Rosbjerg D., Blöschl G., Burn D. H., Castellarin A., Croke B., Baldassarre G. D., Iacobellis V., 

Kjeldsen T. R., Kuczera G., Merz R., Montanari A., Morris D., Ouarda T. B. M. J., Ren L., 

Rogger M., Salinas J. L., Toth E. and Viglione A. (2013). Prediction of floods in ungauged 

basins. In Runoff prediction in ungauged basins. A synthesis across processes, places and 

scales (G. Blöschl, M. Sivapalan, T. Wagener, A. Viglione, and H. Savenije, eds.), pp. 189-226. 

Cambridge University Press, Cambridge. 

- Salvadori G., DeMichele C. and Durante F. (2011). On the return period and design in a 

multivariate framework. Hydrological Earth System Sciences 15: 3293-3305. 

- Serinaldi F. and Grimaldi S. (2011). Synthetic design hydrographs based on distribution 

functions with finite support. Journal of Hydrologic Engineering 16: 434-446. 

- Yue S., Ouarda T. B. M. J., Bobée B., Legendre P. and Bruneau P. (2002). Approach for 

describing statistical properties of flood hydrograph. Journal of Hydrologic Engineering 7: 

147-153. 

- Yue S. and Rasmussen P. (2002). Bivariate frequency analysis: discussion of some useful 

concepts in hydrological application. Hydrological Processes 16: 2881-2898. 



Channel widening during extreme floods: how to 

integrate it within river corridor planning ? 

Francesco Comiti 1 ; Margherita Righini 2 ; Laura Nardi 3 ; Ana Lucía 4 ; William Amponsah 5 ; Marco Cavalli 5 ; Nicola Surian 2 ; 

Lorenzo Marchi 5 ; Massimo Rinaldi 3 ; Marco Borga 6 

ABSTRACT 

Channel widening taking place during large flood events can be substantial in mountain 

rivers, with consequent great potential damages to infrastructures and buildings. The purpose 

of this work is twofold: i) to provide a quantitative assessment of geomorphic effects of an 

extreme flood event (recurrence interval > 100 years); ii) to test on this study case a new 

hydromorphological methodological framework (IDRAIM) developed to guide river corridor 

planning and management. 

As to the first objective, field surveys were integrated with remote sensing, GIS and statistical 

analyses for a flood event occurred in 2011 in Northwestern Italy. Channel widening ratios 

(width after / width before the flood) were calculated and then correlated with different 

controlling factors, and envelope relationships were then obtained. 

The tool of the IDRAIM framework used for the second objective was the Event Dynamics 

Classification (EDC) applied to selected study reaches, whose widening ratios turned out to 

correspond well with the EDC classes. 

Based on the results obtained, a practical procedure for predicting the expected widening is 

finally proposed. 

KEYWORDS 

large floods; channel changes; stream power; bank erosion; river corridor 

INTRODUCTION 

Infrequent, high-magnitude floods can lead to sudden, dramatic channel changes in alluvial 

and semi-alluvial channels. Indeed, the geomorphic role of large floods has long been debated 

(e.g. Wolman and Miller, 1960; Costa and O'Connor, 1995; Phillips, 2002; Magilligan et al., 

2015). Nonetheless, rather few are the studies which have analyzed in detail the magnitude 

of channel widening determined by extreme floods in Alpine rivers (Krapesch et al., 2011), 

despite the paramount relevance of such process on flood hazard. In fact, a sound river 

corridor planning should include – beside flood inundation depth and velocities – the 

expected channel dynamics (bank erosion and bed incision/aggradation) occurring during 

1 Free University of Bozen-Bolzano, Bolzano, ITALY, francesco.comiti@unibz.it 

2 Dept. Geosciences - University of Padova, ITALY 

3 Dept. of Earth Sciences - University of Florence, ITALY 

4 Center for Applied Geoscience, Eberhard Karls University of Tübingen, GERMANY 

5 CNR IRPI - Padova, ITALY 

6 Dept. Land and Agroforest Environments - University of Padova, ITALY 



flood events, as these can both substantially modify the flooding pattern and cause direct 

damage to buildings and infrastructures. 

Unfortunately, the expected sudden and notable changes in channel width during floods are 

typically not included in flood hazard mapping. Indeed, our current understanding of the 

factors controlling channel widening is very limited. In fact, hydrodynamic forces were found 

to be not sufficient to explain geomorphic effects (e.g. Heritage et al., 2004; Nardi and Rinaldi, 

2015), and thus other factors, such as bedload supply and pre-flood channel planform (Dean 

and Schmidt, 2013), lateral confinement (Thompson and Croke, 2013), and channel 

curvature (Buraas et al., 2014) should also be also accounted for in the prediction of 

widening. However, field data available to build reliable statistical models or to validate 

numerical morphodynamic models are very limited. On the other hand, easy-to-use tools 

applicable by practitioners of river management agencies are much needed to predict the 

reach-scale morphological response of the channel network to extreme (recurrence intervals 

RI>100 yr) flood events. 

The purpose of this work is twofold: i) to provide a quantitative assessment of the channel 

widening associated to an extreme flood event which occurred in 2011 in Northwestern Italy 

(Magra River basin); and ii) to test on this study case a new methodological framework 

(IDRAIM) developed to guide river corridor planning and management which focuses on 

channel adjustments occurring both in the long-term and during extreme events. 

METHODS 

Study case: the 2011 flood event in the Magra River basin 

Figure 1: Location map of the Magra River basin and of the analyzed tributaries (from Surian et al., 2016). 


The Magra River basin is located in the northern Apennines (Italy) and covers an area of 

1717 km 2 , ranging from sea level to a maximum elevation of 1901 m a.s.l. (Fig. 1). Sedimentary 

rocks (mostly sandstones and mudstones) prevail in the basin, but some outcrops of 

magmatic (ophiolites) and metamorphic rocks are also present. The climate is Mediterranean, 

with dry summers and the most abundant precipitation occurring in autumn. The mean 

annual precipitation is about 1700 mm, and maximum values (up to about 3000 mm) are 

observed in the upper part of the catchment. Forests cover 66% of total basin area and 

occupy most of catchments slopes. Agricultural areas, urban areas and transportation 

structures mostly lie in the valley floors and on the lower sectors of the slopes. 

An intense precipitation event took place within the river basin on October 25th, 2011, and 

originated a flash flood both in the Magra River, with a peak discharge having a RI of about 

100 yr (Nardi and Rinaldi, 2015) and along several tributaries, there with extremely high 

peak flows (up to RI>300 yr, based on rainfall time series), where enormous volumes of large 

wood were eroded from the floodplains (Lucía et al., 2015) causing extensive bridge clogging. 

Rainfall maps for the study event were obtained based on data from a rain gauges network 

and the Monte Settepani radar, and the estimates show that maximum hourly rates were up 

to 149 mm/hr, whereas three-hours maximum and event-accumulation maxima were up to 

326 mm and 500 mm, respectively. Antecedent moisture conditions in the basins were 

intermediate. An integrated approach was adopted to investigate the geomorphic effects of 

the 2011 flood in the Magra catchment, and the whole approach is described in detail by 

Rinaldi et al. (2016). 

Six catchments chosen among those where rainfall was most intense were selected to analyze 

channel response: Pogliaschina, Gravegnola, Osca, Teglia, Geriola and Mangiola (Fig. 1). 

Figure 2: Example of the channel widening observed in the Mangiola River. 


In these rivers, unit peak discharge estimates range from 12.8 m 3 s -1 km -2 (Osca) to 

23.7m 3 s -1 km -2 (Pogliaschina). Streams within the six catchments are characterized by average 

channel slope ranging from 4.1% (Osca) to 8.8% (Geriola), coarse sediments (mainly gravels 

and cobbles), and a wide range of conditions in terms of lateral confinement, but only 

partly- and unconfined reaches were analyzed for channel widening. Importantly, artificial 

structures (bed and bank protections) were very limited before the flood. 

The morphological changes induced by the 2011 flood were assessed by field surveys and 

interpretation of aerial photographs. To assess changes in channel width, channel banks and 

islands (i.e. in-channel surfaces covered by woody vegetation) were digitized on pre- and 

post-flood orthophotos (Fig. 2). The term “channel” refers to the active channel, which 

includes low-flow channels and unvegetated or sparsely vegetated bars (i.e. exposed 

sediments). Then, channel width was calculated dividing channel area by the length of the 

reach, and changes in channel width were expressed as width ratio Wr, i.e. channel width 

after / channel width before the flood, as in Krapesch et al. (2011). 

The IDRAIM methodological framework 

The IDRAIM methodological framework (Rinaldi et al., 2014, 2015) includes the following 

four phases: (1) catchment-wide characterization of the fluvial system; (2) evolutionary 

trajectory reconstruction and assessment of current river conditions; (3) description of future 

trends of channel evolution; (4) identification of management options. A series of specific 

tools have been developed for the assessment of river conditions, in terms of morphological 

quality and channel dynamics. These latter include the “Morphological Dynamics Index”, the 

“Event Dynamics Classification”, and the “River Morphodynamic Corridors”. 

The present work focuses specifically on the “Event Dynamics Classification” (EDC), applying 

it to selected study reaches and comparing them with observed geomorphic changes. In 

particular, EDC leads to classify each reach into one of four classes of expected event 

dynamics (very high, high, medium, low), adopting a guided logical procedure based on flow 

charts. The assessment is carried out by combining two aspects (Table 1): 

Table 1: Classification of EDC (Event Dynamics Classification, from low to very high) based on the expected morphological changes 

(from small to very relevant) coupled to the clogging probability (From Rinaldi et al., 2015). 

i) the expected magnitude of morphological changes (4 classes) and ii) the clogging conditions 

(2 classes, i.e. likely or not likely occurrence of clogging, mostly by wood elements) at critical 

cross-sections (bridges and culverts). EDC provide information on the expected magnitude of 


94 

95 

96 

94 

97 

95 

98 

99 

96 

97 

100 

98 

101 99 

102 

100 

103 

104 

101 

105 

102 

106 

103 

107 

104 

108 

105 

109 

106 

107 

110 

108 

109 111 

112 

channel dynamics in a given reach on a one-dimensional scale. This information has to be 

integrated with a 2-D analysis to define the areas of the fluvial corridor that will be affected 

by such dynamics (“Event Morphodynamic Corridor”, EMC). The procedure suggested in 

IDRAIM for the delineation of the event river morphodynamic corridor (EMC) includes (i) 

reconstruction of historical channel changes; (ii) determining the expected flood spatial 

dynamics based on EDC class; iii) identification of natural elements of confinement (e.g., 

hillslopes, old terraces); and (iv) identification of reliable protection works preventing lateral 

lead to better statistical results compared to its calculation through the post-ev 

channel mobility. All the details of the methodology for both EDC and EMC can be found in 

Rinaldi et al. (2014 and 2015). 

al., 2016). 

RESULTS 

Channel widening in the Magra River basin 

The widening data (n=35) measured in the main channel of the Magra River (Nard 

The “Width Ratio” (Wr) measured in all the analyzed reaches (n=157) of the six basins 

lead to better statistical results compared to its calculation through the post-ev 

described are reported above are as plotted well. The in Figure large 3 against variability the estimated (log-log unit scale stream graph) power of the flood channel resp 

peak, al., 2016). calculated using the pre-event channel width (as in Krapesch et al., 2011) as 

energy is clearly apparent, especially at intermediate unit stream power (1000 

ω =γ QS/W, where γ is the specific weight of water, Q is the flood peak discharge, S is channel 

slope, and W is the channel width measured before the flood. The unit stream power 

best fit (R 2 = 0.44) power equation including all data is the following: 

estimated The widening using the data pre-event (n=35) width measured was found in to the lead main to better channel statistical of results the Magra compared River to (Nard 

its calculation through the post-event width (Surian et al., 2016). 

The 

are 

widening 

reported 

data (n=35) measured in the main channel of the Magra River (Nardi and 

WW rr = 0.07ωω 0.44 as well. The large variability (log-log scale graph) of the channel resp 

Rinaldi, 2015) are reported as well. The large variability (log-log scale graph) of the channel 

energy is clearly apparent, especially at intermediate unit stream power (1000 

response to similar flow energy is clearly apparent, especially at intermediate unit stream 

power best Based fit (1000-10000 on (R 2 this = 0.44) regression, Wmpower -2 ). The equation best the fit minimum (R 2 = including 0.44) power unit all stream equation data is power including the following: required all data is the to cause so 

following: 

out to be about 400 Wm -2 . However, one reach of the Magra River exhibited a w 

WW rr = 0.07ωω 0.44 

a unit stream power as low as 173 Wm -2 . On the other side, very limited wid 

Equation 1 

observed in some reaches for ω up to about 5000 Wm -2 . Very intense channel w 

Based Based on on this this regression, regression, the minimum the minimum unit stream power unit stream required power to cause some required widening to cause so 

turns instead out to observed be for ω>2000 -2 . Wmone -2 in reach other of the reaches Magra River (Gravegnola, exhibited a Pogliaschina, width O 

out to be about 400 Wm -2 . However, one reach of the Magra River exhibited a w 

ratio of 1.08 for a unit stream power as low as 173 Wm -2 . On the other side, very limited 

consequence of the enormous scatter in the relationship, indications on the 

widening a unit stream (Wr10) sounder was instead to plan observed on the for safety ω>2000 side. Wm -2 To in other this reaches aim, the (Gravegnola, 

observed Pogliaschina, in some Osca reaches channels). for As ω a consequence up to about of the 5000 enormous Wm -2 . scatter Very in intense the channel w 

following eq 

relationship, in Fig. 3) represents indications on the upper maximum envelope expected of width the ratios plot, are fitted sounder to to a plan linear on the trend excl 

instead observed for ω>2000 Wm -2 in other reaches (Gravegnola, Pogliaschina, O 

safety side. To this aim, the following equation (dashed line in Fig. 3) represents the upper 

outliers (9 out of 192 reaches, i.e. about 5%): 

envelope consequence of the plot, of fitted the enormous to a linear trend scatter excluding the seemingly relationship, outliers (9 indications out of 192 on the 

reaches, i.e. about 5%): 

width ratios are sounder to plan on the safety side. To this aim, the following eq 

WW rr ≈ 0.002ωω 

in Fig. 3) represents the upper envelope of the plot, fitted to a linear trend excl 

Equation 2 

outliers Nonetheless, (9 out the of 192 lateral reaches, confinement i.e. about (expressed 5%): by the confined index calcula 

INTERPRAEVENT 2016 – Conference Proceedings | 481 

width divided by pre-event channel width) plays a relevant role on the maxim

Figure 3: Scatterplot of width ratio vs unit stream power (calculated based on the pre-event channel width) for the study reaches. Data 

from Nardi and Rinaldi (2015) are also included. The solid line represents the best fit equation (Eq. 1), the dashed line the upper 

envelope curve (Eq. 2). 

Figure 4: Scatterplot of width ratio vs confinement index (i.e. alluvial plain width / pre-event channel width) for the study reaches. Data 

from Nardi and Rinaldi (2015) are not available. The dashed line represents the 1:1 relationship. 


Nonetheless, the lateral confinement (expressed by the confinement index calculated as 

alluvial plain width divided by pre-event channel width) plays a relevant role on the 

maximum width ratios, as shown in Figure 4, where most of the data plot below – in some 

cases considerably – the equality line. However, 10 reaches feature widening ratios exceeding 

the confinement index, an indication that part of the hillslopes was also eroded during the 

flood (this was verified in the field). Therefore, in most of cases the widening was not limited 

by the lateral extent of the erodible corridor, although a positive correlation exists between 

the two variables. 

When plotting the widening magnitude against the pre-flood channel width, an inverse 

relationship is shown (Fig. 5). The plot shows how large is the variability in the width ratios 

observed in the narrower (1-10 m) reaches, and that for large channels (>100 m) width ratios 

> 1.5 should not be expected. 

Figure 5: Scatterplot of width ratio vs pre-event channel width for the study reaches. Data from Nardi and Rinaldi (2015) are also 

included. The dashed line represents equation 3 (envelope curve). 

The envelope curve (fitted by eye to a simple power form, excluding 5% of the dataset) is 

given by following equation: 

124 

125 

126 

127 

WW rr ≈ 2WW 

 

Equation 3 

A more complete, in-depth statistical analysis – entailing multiple factors – for t 

the variability in the observed width ratios is presented in Surian et al. (2016), w 

prediction of the widening extent is discussed in the next section. 


A more complete, in-depth statistical analysis – entailing multiple factors – for the understanding 

of the variability in the observed width ratios is presented in Surian et al. (2016), 

whereas the practical prediction of the widening extent is discussed in the next section. 

Table 2: EDC (Event Dynamics Classification), width ratio and basic characteristics of some reaches of the Magra and Mangiola rivers.. 

Predicting channel widening based on morphological analysis 

The “Event Dynamics Classification” (EDC) was applied to some reaches of the Magra and 

Mangiola rivers, as a preliminary test. Table 2 presents the EDC classification along with the 

width ratio measured after the 2011 flood event. 

The partly confined reach classified as “Very high” event dynamics (the highest, indicating 

that flood waters and bedload transport are expected to abandon the existing channel as a 

result of avulsion processes related to relevant aggradation, landslide damming or bridge 

clogging) featured a widening of 3.5, much higher than the one falling into “High” class 

(Wr=1.3). For the unconfined reaches, the lowest slope Magra reach (close to the river outlet) 

classified with “Medium” dynamics (a class indicating that avulsions are not expected, and 

only local bank erosions and limited aggradation or incision are expected) showed no 

widening during the 2011 event, whereas a “Very high” dynamics was attributed to the 

steeper reach in the Mangiola which was characterized by Wr=5.3. 

DISCUSSION AND CONCLUSIONS 

The magnitude of the channel widening occurred during the 2011 flood in the Magra River 

basin is hugely variable, and the relevance and significance of the different factors (including 

unit stream power, confinement index and sediment supply as well) are mediated by channel 

slope, as statistically demonstrated in Surian et al. (2016). A simplified approach to a first 

order estimation of the average extent of the widening expected during a large (RI>100 yr) 

flood in mountain basins could be based on the upper envelope curves for the observed width 

ratios, with the caution that our database is limited to one river basin (although large and 

diverse) and to one event. The procedure should entail: i) application of eqs. 2-3 and 

averaging of their results, in order to derive the maximum potential width ratio based on 

flow energy/channel size; ii) assessment of the confinement index; iii) evaluation of EDC 

class. 


In case the Wr estimated by the integrated application of eqs. 2-3 is larger than the confinement 

index, and the EDC class is “high” or “very high” (meaning that artificial bank 

protections are not present, not reliable or not relevant for the event dynamics), the Wr 

should be assumed equal to the confinement index, or even slightly higher in case of poorly 

resistant hillslope substrates. In case EDC is high or very high, but the confinement index is 

larger than what predicted by Eqs. 2-3, then the latter values should be adopted. Finally, if 

EDC classes are low to medium (due to either stable bank protections, or cohesive banks/very 

low slopes), the potential for widening is limited, and the width ratios predicted by eqs. 2-3 

should be considered unrealistic, and a very limited widening (even null) could be assumed. 

Further post-event analysis will be obviously of great importance to broaden the dataset and 

to test the suggested simplified approach. 

REFERENCES 

- Buraas E.M., Renshaw C.E., Magilligan F.J., Dade W.B. (2014). Impact of reach geometry on 

stream channel sensitivity to extreme floods. Earth Surface Processes and Landforms 39: 

1778–1789. 

- Costa J.E., O'Connor J.E. (1995). Geomorphically effective floods. In: Costa, J.E., Miller, 

A.J., Potter, K.W., Wilcock, P. (Eds.), Natural and Anthropogenic Influences in Fluvial 

Geomorphology. Monograph, vol. 89. American Geophysical Union, Washington, D.C.: 

45–56. 

- Dean D.J., Schmidt J.C. (2013). The geomorphic effectiveness 693 of a large flood on the 

Rio Grande in the Big Bend region: Insights on geomorphic controls and post-flood geomorphic 

response. Geomorphology 201: 183–198. 

- Heritage G.L., Large A.R.G., Moon B.P., Jewitt G. (2004). Channel hydraulics and geomorphic 

effects of an extreme flood event on the Sabie River, South Africa. Catena 58: 151–181. 

- Krapesch G., Hauer C., Habersack H. (2011). Scale orientated analysis of river width 

changes due to extreme flood hazard. Natural Hazards and Earth System Sciences 11: 

2137–2147. 

- Lucía A., Comiti F., Borga M., Cavalli M., Marchi L. (2015). Dynamics of large wood during 

a flash flood in two mountain catchments. Nat. Hazards Earth Syst. Sci., 15, 1741–1755. 

- Magilligan F.J., Buraas E.M., Renshaw C.E. (2015). The efficacy of stream power and flow 

duration on geomorphic responses to catastrophic flooding. Geomorphology 228: 175-188. 

- Nardi L., Rinaldi M. (2015). Spatio-temporal patterns of channel changes in response to a 

major flood event: the case of the Magra River (central-northern Italy). Earth Surface - Processes 

and Landforms 40: 326-339. 

- Phillips J. D. (2002). Geomorphic impacts of flash flooding in a forested headwater basin. 

Journal of Hydrology 269: 236–250. 

- Rinaldi M., Surian N., Comiti F., Bussettini M. (2014) IDRAIM – Sistema di valutazione 

idromorfologica, analisi e monitoraggio dei corsi d'acqua. Istituto Superiore per la Protezione 

e la Ricerca Ambientale, Roma, 113/2014, ISBN 978-88-448-0661-3 


- Rinaldi M., Surian N., Comiti F., Bussettini M. (2015). A methodological framework for 

hydromorphological assessment, analysis and monitoring (IDRAIM) aimed at promoting 

integrated river management. Geomorphology 251: 122-136. 

- Rinaldi M., Amponsah W., Benvenuti M., Borga M., Comiti F., Lucía, A., Marchi L., Nardi, 

L., Righini M., Surian N. (2016). An integrated approach for investigating geomorphic 

response to an extreme flood event: methodological framework and application to the 

October 2011 flood in the Magra River catchment, Italy. Earth Surface Processes and 

Landforms 41: 835-846.. 

-Surian N., Righini M., Lucía A., Nardi L., Amponsah M., Benvenuti M., Borga M., Cavalli 

M., Comiti F., Marchi L., Rinaldi M., Viero A. (2016). Channel response to extreme floods: 

insights on controlling factors from six mountain rivers in northern Apennines, Italy. 

Geomorphology (in press). 

- Thompson C., Croke J. (2013). Geomorphic effects, flood power, and channel competence 

of a catastrophic flood in confined and unconfined reaches of the upper Lockyer valley, 

southeast Queensland, Australia. Geomorphology 197: 156–169. 

- Wolman M.G, Miller J.P. (1960). Magnitude and frequency of forces in geomorphic 

processes. Journal of Geology 68: 54–74. 



Key results of the Swiss wide natural hazard risk 

assessment on national roads 

Luuk Dorren, PhD 1 ; Philippe Arnold, MSc. 2 

ABSTRACT 

Since 2008, the Federal roads office FEDRO is owner and manager of all national roads in 

Switzerland, which comprise the highways as well as most of the mountain passes of key 

importance. As a result, Swiss-wide, standardized information on natural hazards that threaten 

national roads (highways) was not available. The FEDRO therefore decided to initiate a 

four year project aiming at quantifying and mapping all risks due to natural hazards threatening 

Swiss national roads. This paper presents the methodology used in this project and 

presents a summary of the monetarised risks of the evaluated road sections. The natural 

hazards that are assessed are snow avalanches, rock- and icefall, flooding, debris flows, 

landslides (permanent, spontaneous and slope type debris flows) and collapse dolines. Risk 

hot spots mainly occur due to road closure related to rockfall or bank erosion. Damage to 

infrastructure represents generally only up to 20 % of the total calculated risk; person risks 

add up to 8 % of the total risk. 

KEYWORDS 

risk evaluation; natural hazards, road management, hazard assessment 

INTRODUCTION 

Snow avalanches, landslides and flooding repeatedly pose a threat to Alpine regions (e.g., 

BUWAL 1999; Rudolf-Miklau et al. 2006; Bezzola and Hegg 2007). To reduce the risk posed 

by these hazards, about 775 million CHF is yearly invested in hazard prevention, by the Swiss 

Confederation, cantons, communes and privates (PLANAT, 2007). Protection against flooding 

and debris flows takes up 54% of this amount, 25% is used for landslides and rockfall 

prevention, and 21% for prevention of snow avalanches. Nevertheless, a 100% safety cannot 

be guaranteed. Examples such as the rockfall event of June 2006 on the Gotthard highway, 

the road destructing flooding and landsliding events in August 2005 or the numerous snow 

avalanches in the winter of 1999 show that road infrastructure and its users are susceptible to 

the impact of natural hazards. 

Since January 2008, the federal roads office (FEDRO) is responsible for the Swiss national 

road network (highways and the main alpine passes). Before then, the national roads were 

managed by Cantonal road services. As a result, Swiss-wide, standardized information on the 

type, frequency, intensity and location of natural hazards that threaten national roads, as well 

1 HAFL, Bern University of Applied Sciences, Zollikofen, SWITZERLAND 

2 Federal Roads Office FEDRO, SWITZERLAND 



as the costs of required protective measures, was not available. The FEDRO therefore decided 

to initiate a Swiss wide project, called “natural hazards on national roads – NHNR” with the 

technical support of the Federal Office for the Environment (FOEN), aiming at quantifying 

and mapping all risks due to natural hazards threatening Swiss national road network (total 

length = 1892 km). At the moment of writing, 75% of the national road network, including 

all alpine national roads, has been analysed, which allows to assess and evaluate the 

importance of natural hazard related risks on national roads. This paper will present the 

methodology used in the project and presents a summary of the monetarised risks of the 

evaluated road sections. 

METHODOLOGY 

The detailed project methodology of the project NHNR can be found online (FEDRO, 2012). 

This methodology describes in detail the following 4 main parts: 1) hazard assessment, 2) risk 

analysis, 3) risk evaluation and 4) planning of protective measures. As such the methodology 

defines the natural hazards to be studied, the study perimeter, the standards to be used, the 

risk equations, and parameter values to be used, as well as the products to be delivered. The 

project steering committee decided not to prescribe models for simulating the different 

natural hazards to be assessed, but rather to define the required products in detail, in order to 

guarantee a maximum transparency and traceability of the methods, models and assumptions 

used. 

For this project, the national road network has been split in road sections of 30 – 70 km. 

On each of these sections, the following natural hazards (if present), also called gravitational 

natural hazards, have to be assessed: 

– snow avalanches 

– rock- and icefall, rock avalanches 

– flooding and debris flows 

– landslides (permanent, spontaneous and slope type debris flows) 

– collapse dolines 

The field and modeling studies needed for the hazard and risk analysis are being done by 

consortiums of collaborating geotechnical bureaus. In general, each consortium consists of an 

interdisciplinary project leader with experience in natural hazards, an avalanche expert, one 

or two geological experts, a hydraulic engineering/flooding expert and a risk analysis expert. 

On average, they need approximately one and a half year to complete the natural hazard 

assessment per road section. The risk analysis is subsequently finalised within three months. 

We used hazard indication maps based on the projects Aquaprotect (FOEN, 2015a) and 

SilvaProtect-CH (FOEN, 2015b) to determine which hazards had to be studied in detail in 

which area. To obtain a homogeneous and comparable dataset for the entire Swiss national 

road network, 4 return period scenarios (0 – 10 yrs, 10 – 30 yrs, 30 – 100 yrs, 100 – 300 yrs 


and intensity classes low, medium and high) have to be defined for each the potential hazard 

source area. The so called damage potential perimeter that is to be taken into account is the 

area covered by the highway with a 10 m buffer, as well as surrounding facilities (e.g., 

parking places, technical tunnel installations, ...). The risk analysis is carried out on one or 

two lines that represent the road axes and on surrounding facilities. 

The risk calculation is finally done by an internet based risk calculation tool called RoadRisk 

(www.roadrisk.ch). This tool is WebGIS-based (Fig. 1) and intersects the intensity maps of all 

studied natural hazards for the defined return periods with the damage potential. The 

underlying attributes of the intensity polygons and the road axes allow the calculation of the 

following risk types: 

– Direct impact (Risk direct impact), 

– Collision with deposits on the road or with cars that are impacted by natural hazards (Risk 

collision), 

– Damage to infrastructure (Risk infra damage), 

– Precautionary road closure (Risk pre-closure), 

– Road closure after an event (Risk post-closure). 

Figure 1: A screen shot of the internet tool RoadRisk. 

Casualties are expressed in costs using a value of 5 million CHF per human life. Variables 

required for calculating person risks (direct impact and collision risk) are: 

– the probability of the a precautionary road closure during the hazard event. 

– the mean daily traffic. 


– the mean occupancy rate per vehicle (= 1.76 person per vehicle). 

– the maximum driving speed per road section. 

– the occurrence probability of traffic jams. 

– the lethality of the people in a car being impacted as function of 

– the type of hazard process and its intensity. 

The main variables for calculating the risks of infrastructure damages and road closure are 

infrastructure construction costs (defined by the FEDRO), as well as the daily costs for road 

closure (varying between 150’000 and 4’000’000 CHF/day, after Erath 2011) and the 

estimated duration of the road closure. As mentioned before, all risk calculation algorithms 

can be found in FEDRO (2012). 

RESULTS 

The results of the 22 completed national road sections show that the sum of the yearly 

expected damage due to gravitational natural hazards can add up to several 100’000 CHF/ 

year, and in extreme cases up to several millions CHF/year per road section. Fig. 2 shows a 

summary of the risk hot spots on the completed road sections. 

Figure 2: Overview of the risk hot spots on the national road sections that have been analysed between September 2008 and August 

2015. 

The most prominent risks are due to road closure in areas that are strongly affected by rock - 

fall or bank erosion/underscouring (cf. the damaged highway viaduct in canton Uri in 1987). 

Figure 2 also shows that high risks on natural roads due to natural hazard in Switzerland do 


not occur in alpine terrain only. An example is the risk hot-spot west of the city of Biel (on 

the N5 parallel to lake Biel), which is related to high frequency rockfall from a Jura-type 

limestone cliff and a relative high mean daily traffic (~15’500 vehicles per day). A typical risk 

hot-spot is the double highway viaduct of Chillon, close to Montreux at the eastern side of 

Lake Geneva. Although the occurrence probability if low, the pillars of this viaduct can be 

heavily damaged by falling blocks with volumes of 5m 3 , which would lead to a road closure 

of several months. The lack of suitable detour possibilities leads to a high risk of road closure 

after an event. 

Details of the expected damages (Fig. 3) show that direct damage to infrastructure represents 

generally only adds up to 16% of the total potential damages, but can reach up to 40% for 

some individual hazard sources. Looking at the sums of the expected damage values of the 

road sections that are currently finished, road closure after an event represents 69%, 

precautionary road closure 7%. Person risks (casualties) are mostly to be expected due to 

falling rocks, rock avalanches, debris flows and dense snow avalanches, which add up to 8% 

of the total risk. 

The results show that rockfall is responsible for 31% of the total calculated risk, bank erosion 

for 17%, flooding for 14%, snow avalanches for 13%, landslides for 9%, rock avalanches for 

Figure 3: Risk details per hazard and damage type for all analysed road sections. 


8%, debris flows for 6%, and slope-type debris flows for 1%. Finally, subsidence processes 

due to doline collapse, falling ice and snow gliding account for 0.7% resp. 0.3% and 0.1%. 

When calculating the expected number of casualties per year caused by rockfall (incl. rock 

mass falls) on national roads, by dividing the total of the rockfall related person risks by 5 

million CHF, the result is 0.25 casualties per year. In the last thirty years, rockfall caused 0.13 

casualties per year on average. The total expected number of casualties per year caused by all 

gravitational natural hazards on national roads is 0.6 

CONCLUSIONS 

The currently available results of the project NHNR show that rockfall and bank erosion are 

mainly responsible for the expected damage due to natural hazards on national roads in 

Switzerland. A large part of this is due to road closure after hazard events. 

The threshold for the individual risk of death due to natural hazards defined by the FEDRO is 

1 * 10 -5 . If an individual would pass all currently evaluated endangered national road sections 

every day, his or her individual risk of death would be 1 * 10 -4 . In the meantime, the project 

results regarding the number of casualties seem to be too pessimistic in comparison to reality 

of the last thirty years. 

The currently known total sum of expected yearly damage by natural hazards on national 

roads (~40 million CHF per year) are equal to 15% of the mean yearly amount for the 

prevention of gravitational hazards invested by the Swiss confederation between 2000 and 

2005 (PLANAT, 2007). This amount covers risk prevention on all roads, railways, residential 

areas and other critical infrastructure. At present, the mentioned expected yearly damage 

justifies continuous investments in risk reduction on national roads. 

REFERENCES 

- Bezzola, G.-R., Hegg, C. (2007). Ereignisanalyse Hochwasser 2005. Teil 1 – Prozesse, 

Schäden und erste Einordnung. BAFU & WSL. 

- BUWAL (1999). Leben mit dem Lawinenrisiko - Die Lehren aus dem Lawinenwinter 1999. 

BUWAL. 

- Erath, A. (2011). Vulnerability of road transport infrastructure, Ph.D. Thesis, ETH Zürich, 

Zürich. 

- FEDRO (2012). Natural hazards on national roads: Risk concept. Methodology for risk-based 

assessment, prevention and response to gravitative natural hazards on national roads. ASTRA 

Documentation 89001. Federal Roads Office FEDRO. http://www.astra.admin.ch/dienstleistungen/00129/00183/01156/index.html?lang=en 

- FOEN (2015a) Aquaprotect. Federal office for the environment FOEN. http://www.bafu. 

admin.ch/naturgefahren/01916/06598/index.html?lang=en 

- FOEN (2015b) SilvaProtect-CH - Schutzwald in der Schweiz. Federal office for the environment 

FOEN. http://www.bafu.admin.ch/naturgefahren/01920/01964/index.html?lang=de 


- PLANAT, (2007). Jährliche Aufwendungen für den Schutz vor Naturgefahren in der 

Schweiz (www.planat.ch). 41 p. 

- PLANAT, (2009). Risk concept for natural hazards, Strategy Natural Hazards Switzerland (in 

German; www.planat.ch). 

- Rudolf-Miklau, F., Ellmer, A., Gruber, H., Hübl, J., Kleemayr, K., Lang, E., Markart, G., 

Scheuringer, E., Schmid, F., Schnetzer, I., Weber, C., Wöhrer-Alge, M., (2006). Hochwasser 

2005 - Ereignisdokumentation. Bundesministerium für Land- und Forstwirtschaft, Umwelt 

und Wasserwirtschaft, Sektion Forst, Abteilung Wildbach- und Lawinenverbauung, Wien. 



Investigation of the flood hazard of the Nuclear Power 

Plant KKG by earthquake induced dam breaks waves at 

the River Aare 

Davood Farshi, Ph.D. 1 ; Michael Ballmer 2 ; Donat Job 3 ; Brigitte Faust 4 

ABSTRACT 

The flood safety of the Nuclear Power plant KKG in terms of dam breaks scenarios at the 

River Aare was investigated through using a deterministic method. Based on the requirements 

of the Swiss Federal Nuclear Safety Inspectorate (ENSI), four dam break scenarios were 

identified for the investigation. The scenarios were based on conservative assumptions of 

different combinations of weirs' breaks. 

The propagation of the flood waves were computed by means a hybrid model as a combination 

of one (1D) and two-dimensional (2D) models. 

Based on the simulations it was possible to estimate the risk of flooding of the KKG area and 

if needed to suggest some measures to prevent such flooding. Additionally it has been 

investigated, if important water intakes were at the risk in regard to the allowable maximum 

and minimum water level. 

The sediment transport was studied qualitatively, if there were high erosion risks. According 

to the calculated bottom shear stresses the river bed erosion is possible only locally and over a 

very short time period. 

KEYWORDS 

Dam break; Hybrid model; Flood wave; NPP 

INTRODUCTION 

After Fukushima, the question arises, how strongly the nuclear power plants (NPP) could be 

impacted by the natural extreme events in Switzerland. The Swiss Federal Nuclear Safety 

Inspectorate (ENSI) requires new investigation from all plant operators in Switzerland. Flood 

waves which are due to dam breaks caused by extreme earthquakes, are categorized under 

extreme events against which Swiss nuclear plants have to be protected. In order to ensure 

such protection, the NPP operators are obliged to review the hazard at regular intervals on 

the basis of experience and the latest developments in science and technology [ENSI 2014]. 

Based on previous ENSI’s requirements, the flood safety of the nuclear power plant Gösgen-Däniken 

AG (KKG) has been already evaluated and ensured for an extreme flood with 

an annual probability of occurrence of 10 -4 (HQ 10000 

). Additionally ENSI requested KKG to 

1 EnHydro GmbH, Zürich, SWITZERLAND, davood.farshi@enhydro.ch 

2 TK Consult AG, Zürich, Switzerland 

3 AF Consult Switzerland AG, Baden, Switzerland 

4 Kernkraftwerk Gösgen-Däniken AG, Däniken, Switzlerand 


FP055

Figure 1: Layout of the KKG area. 

Figure 2: Layout of the investigated reach of the river Aare. 


investigate the impacts of the flood waves due to the extreme earthquake-induced dam 

breaks (annual probability of occurrence of 10 -4 ) on KKG area and especially on important 

cooling water intakes (Figure 1). The break of all weirs upstream the KKG till the lake Biel 

has to be considered for the investigation (Figure 2). 

BASIC DATA 

Hydrology 

The important inflow and discharge measurement stations between the lake Biel and KKG 

are shown in Figure 3. The significant hydrological parameters are represented in Table 1. 

Figure 3: Important inflows and stations of the river Aare between the lake Biel and KKG [BAFU 2014]. 

Table 1: Important discharge values of the river Aare. 

Parameter 

Value 

(m 3 /s) 

Mean Discharge (Q m ) of Aare at Brügg 244.0 

Mean Discharge (Q m ) of Aare at 

Murgenthal 

287.0 

Low discharge (Q 347 ) of Aare at Brügg 106.0 

Low discharge (Q 347 ) of Aare at 

Murgenthal 

128.0 

TOPOGRAPHY 

The topography data of the investigated Area consisted of different forms of data. There were 

more than 850 cross sections measurements of the river Aare for a distance of 170 km and 

also a detailed DTM of the area close to KKG. 


METHODOLOGY 

For the modeling and computing of the flood waves due to dam breaks the investigation area 

has been subdivided into perimeters (Figure 4): 

Figure 4: Hybrid model of the investigation perimeter. 

A) 1D-Perimeter: the area from the three lakes (lake Neuenburg, Murten and Biel) to weir 

Ruppoldingen. 

B) 2D-Perimeter: the area from weir Ruppoldingen to the connection of hydropower channel 

and the old Aare. 

All dam breaks have been considered as sudden failure of the whole weir, which is based on a 

conservative assumption and is a well-established method in Switzerland [BFE 2014]. In case 

the hydropower house and weir are in the same place, the sudden break has been considered 

for the whole structure. For regulation of weirs the condition (n-1) has been adopted, in 

which the capable weir segment has been considered to be closed. The programs Flux/Floris, 

HYDRO_AS-2D and BASEMENT have been used for the simulation. 

SCENARIOS 

On the basis of ENSI’s requirements four different worst-cases have been defined as modeling 

scenarios (Table 2). 

The two mean and low discharges (Q 347 

, Q m 

) have been considered as basic discharges in the 

river Aare, because the peak discharge in case of any dam break depends on the water surface 


Table 2: Scenarios for the assessment of the flood hazard of KKG due to the dam breaks. 

Scenario Event 

1 

2 

3 

4 

difference between upstream and downstream of a weir. A combination of extreme flood 

event and earthquake-induced dam break was beyond the investigation criteria and has not 

been considered. 

The following assumptions have been considered in all scenarios: 

– The Weir Port breaks first due to the closeness to the epicenter of the earthquake 

– The downstream weirs are heavily damaged by the earthquake and can’t be regulated 

anymore. 

– All turbines are out of service after the earthquake (No discharge through Hydropower 

house). 

Sequential dam breaks at the River Aare 

(Weir Port to Weir Winznau) 


(Weir Port to Weir Winznau) 


(Weir Port to Weir Ruppoldingen) and WKW 

Gösgen 

Simultaneous dam breaks at the River Aare 

(Weir Port to Weir Ruppoldingen) und 

Sequential break of the Weir Winznau 

Discharge in 

River Aare [m 3 /s] 

Q 347 

– The downstream weirs break in arrival time of maximum wave (Sequential breaks). 

Q m 

Q m 

Q m 

INITIAL CONDITIONS 

The initial conditions in the river Aare has been calculated as steady state for the both 

discharges Q 347 

and Q m 

, with an assumption that all turbines of the power houses are in 

service. The water elevation upstream of the weirs has been set to the regulated storage level. 

Table 3 shows the regulated storage level for both discharges. 

Table 3: Regulated water elevation upstream head of the weirs (Initial condition). 

Weir Q 347 Q m 

429.00 

429.50 

Port 

(lake level) (lake level) 

Flumenthal 426.00 426.00 

Bannwil 417.30 417.30 

Wynau 408.08 408.08 

Ruppoldingen 397.20 398.20 

Winznau 388.14 388.14 


RESULTS 

The modeling of dam breaks has been run in two phases. In the first phase the break of weirs 

from Weir Port to Ruppoldingen has been simulated through the 1D-Model (see Methodology). 

In the second phase the outflow of the 1D-Model has been defined as a boundary 

condition for the 2D-Model and the break of the Weir Winznau und WKW Gösgen has been 

simulated. The time point 0.00 corresponds to the break of the Weir Port in all results. 

1The results show that sequential dam breaks induce very large peak flow at Weir Ruppoldingen 

and in the case of the scenario 2 the peak flows are larger (Figure 5 & 6). The peak flow 

discharge at Ruppoldingen is almost 4 times larger than the peak flood discharge with an 

annual probability of occurrence of 10 -2 (HQ 100 

). Contrary the peak discharge reduces at the 

time of the dam break of the Weir Winznau due to the overflow of the arriving flood waves, 

which reduces the water elevation difference between upstream and downstream of the weir 

(Figure 7 & 8). 

In Figure 9 and 11 the time history of the water surface elevation at the intake ZM05 is 

illustrated. As it is shown the critical elevation of the intake will never be touched by the 

flood waves. 

CONCLUSIONS 

The flood hazard of the KKG area has been investigated through different dam break 

scenarios of the weirs between the lake Biel and KKG. This investigation has been carried out 

Figure 5: Time history of the discharge at weirs for the scenario 1. 


Figure 6: Time history of the discharge at weirs for the scenario 2. 

Figure 7: Time history of the discharge at the Weir Winznau and at KKG for the scenario 1. 


Figure 8: Time history of the discharge at the Weir Winznau and at KKG for the scenario 2. 

Figure 9: Time history of the water elevation at the intake ZM05 for the scenario 1. 


Figure 10: Time history of the water elevation at the intake ZM05 for the scenario 2. 

based on a hybrid simulation model, which consists of 1D and 2D-model. With the simulations 

it can be shown, to what extent the KKG area and also the cool water intakes are 

imperiled through the flood waves of dam breaks. The scenarios are based on conservative 

assumptions such as sudden break of weirs and can be considered as worst cases. The 

simulations show that the scenario 2 induces large flood waves and has the worst impact on 

the area close to the KKG. 

REFERENCES 

- BAFU (2014). Homepage des Bundesamtes für Umwelt (BAFU), Bern, www.hydrodaten. 

admin.ch 

- ENSI (2014). The Safety of Swiss Nuclear Plants in case of External Flooding, www.ensi.ch. 

- BFE (2014). Richtlinie über die Sicherheit der Stauanlagen, Teil B: Besonderes Gefährdungspotenzial 

als Unterstellungskriterium. 



Spatial and temporal exposure of elements at risk in 

Austria 

Räumliche und zeitliche Exponiertheit von Gebäuden 

in Österreich 

Sven Fuchs, PD Dr. 1 ; Andreas Zischg, Dr. 2 ; Margreth Keiler, PD Dr. 3 

ABSTRACT 

The paper presents a nation-wide spatially explicit object-based assessment of buildings exposed 

to natural hazards in Austria. The assessment was based on two different datasets, (a) 

hazard information providing input to the exposure of elements at risk, and (b) information 

on the building stock combined from different spatial data available on the national level. It is 

shown that the repeatedly-stated assumption of increasing exposure due to continued population 

growth and related increase in assets has to be carefully evaluated by the local development 

of building stock. While some regions have shown a clearly above-average increase in 

assets, other regions were characterised by a below-average development. In sum, around 5 

% of all buildings are exposed to torrential flooding, and around 9 % to river flooding, with 

around 1 % of the buildings stock being multi-exposed. In conclusion, the presented objectbased 

assessment is an important and suitable tool for nation-wide exposure assessment and 

may be used in operational risk management. 


Aussagen zur Exposition von Wertobjekten gegenüber Naturgefahren hängen von der Verfügbarkeit 

entsprechender Datengrundlagen ab, insbesondere von einer genauen Bewertung der 

so genannten Risikoelemente und ihrer räumlichen und zeitlichen Dynamik. Bislang waren 

derartige Daten flächendeckend nicht verfügbar, und in Folge waren Untersuchungen zur Exposition 

auf die lokale Skalenebene einzelner Fallstudien beschränkt. Für alpine Naturgefahren 

wurden langfristige Änderungen des Schadenpotentials einem signifikanten Anstieg 

der Anzahl und Werte gefährdeter Gebäude zugeschrieben und können sowohl in Agglomerationen 

als auch im ländlichen Raum nachgewiesen werden (Fuchs et al. 2013). Kurzfristige 

Veränderungen im Schadenpotential variieren den beobachteten langfristigen Trend, vor 

allem in Bezug auf gefährdete Personen (Keiler et al. 2005; Zischg et al. 2005). Die meisten 

der aktuellen Studien basieren auf lokalen Daten oder auf aggregierter Information zur 

Landnutzung (Zischg et al. 2013), was zu substantiellen Unsicherheiten bei der Risikoberechnung 

führt (de Moel & Aerts 2011). Im Folgenden wird aufgezeigt, wie eine objektbasierte 

1 University of Natural Resources and Life Sciences, Institute of Mountain Risk Engineering, Vienna, AUSTRIA, 

sven.fuchs@boku.ac.at 

2 Universität Bern Geographisches Institut, Mobiliar Lab, Bern, SWITZERLAND 

3 Universität Bern Geographisches Institut, Bern, SWITZERLAND 



Expositionsanalyse auf nationalem Maßstab in hoher räumlicher Auflösung durchgeführt 

werden kann, und wie sich die Exposition gegenüber Naturgefahren räumlich und zeitlich 

entwickelt hat. 

KEYWORDS 

Exposure; building stock; spatial analysis; temporal analysis; risk management 

EINFÜHRUNG 

Aussagen zur Exposition von Wertobjekten gegenüber Naturgefahren hängen von der Verfügbarkeit 

entsprechender Datengrundlagen ab, insbesondere von einer genauen Bewertung der 

so genannten Risikoelemente und ihrer räumlichen und zeitlichen Dynamik. Bislang waren 

derartige Daten flächendeckend nicht verfügbar, und in Folge waren Untersuchungen zur 

Exposition auf die lokale Skalenebene einzelner Fallstudien beschränkt. Für alpine Naturgefahren 

wurden langfristige Änderungen des Schadenpotentials einem signifikanten Anstieg 

der Anzahl und Werte gefährdeter Gebäude zugeschrieben und können sowohl in Agglomerationen 

als auch im ländlichen Raum nachgewiesen werden (Fuchs et al. 2013). Kurzfristige 

Veränderungen im Schadenpotential variieren den beobachteten langfristigen Trend, vor 

allem in Bezug auf gefährdete Personen (Keiler et al. 2005; Zischg et al. 2005). Die meisten 

der aktuellen Studien basieren auf lokalen Daten oder auf aggregierter Information zur Landnutzung 

(Zischg et al. 2013), was zu substantiellen Unsicherheiten bei der Risikoberechnung 

führt (de Moel & Aerts 2011). Im Folgenden wird aufgezeigt, wie eine objektbasierte Expositionsanalyse 

auf nationalem Maßstab in hoher räumlicher Auflösung durchgeführt werden 

kann, und wie sich die Exposition gegenüber Naturgefahren räumlich und zeitlich entwickelt 

hat. 

METHODEN 

Die zeitlich-räumliche Analyse der Exposition folgt im Ansatz dem Risikokonzept für die 

Analyse und Bewertung von Naturgefahren, und basiert auf einer Verschneidung zwischen 

den gefährdeten Flächen und der Gebäudeinformation. Zwei unterschiedliche Datensätze 

wurden verwendet, (a) Gefahreninformation als Input für die Exposition, und (b) Informationen 

aus dem digitalen Gebäude- und Wohnungsregister (GWR II). Um ein vollständigeres 

Bild zur Exposition zu bekommen, wurden die Prozesse „Überschwemmung“, „Wildbachprozesse“ 

und „Lawinen“ analysiert und die Ergebnisse miteinander verglichen, da diese Prozessgruppen 

– neben Sturmwind und Hagel, die in vorliegender Studie jedoch nicht weiter analysiert 

werden – für die Mehrheit der Schäden im Alpenraum verantwortlich zeichnen (Hilker 

et al. 2009). 

Für alpine Naturgefahren (Wildbäche und Lawinen) wurden vorhandene Gefahrenzonenpläne 

verwendet. Gefahrenzonenpläne beziehen sich in der Regel auf eine einzelne Gemeinde, 

und auf den von einem Bemessungsereignis mit einer Wiederkehrperiode von 150 Jahren 

betroffenen Bereich (Republik Österreich 1976). Die roten und gelben Gefahrenzonen 


enthalten mit Stand 09/2014 rund 92% aller Gemeinden mit einer Verpflichtung zur 

Gefahrenzonenplanerstellung nach ForstG 1975 (Republik Österreich 1975). 

Zur Darstellung der von Überschwemmung betroffenen Flächen wurden Daten aus der 

digitalen eHORA Plattform (http://www.hochwasserrisiko.at/) verwendet. Diese Daten zeigen 

überschwemmungsgefährdete Bereiche und wurden gemeinsam vom Bundesministerium für 

Land- und Forstwirtschaft, Umwelt und Wasserwirtschaft und dem österreichischen Versicherungsverband 

auf mehr als 25.000 Flusskilometern modelliert (Stiefelmeyer & Hlatky 2008). 

Aufgrund der Diskussion über die Harmonisierung der Gefahrenkartierung in Österreich 

(Rudolf-Miklau & Sereinig 2009) wurde das modellierte hundertjährliche Hochwasser-Szenario 

für die Studie verwendet. 

Die Exposition wurde mit Hilfe des GWR II ermittelt. Seit der Umsetzung des Bundesgesetzes 

über das Gebäude und Wohnungsregister (Republik Österreich 2009, 2013) sind Gemeinden 

in Österreich für die Sammlung und Verarbeitung von digitalen Informationen für den 

gesamten Gebäudebestand verantwortlich. Diese Informationen werden zentral in einer 

Datenbank gespeichert und enthalten Angaben über die Lage und Größe der einzelnen 

Gebäude, sowie die Baukategorie und die Bauperiode (vor 1919, 1920-1944, 1945-1960, 

1961-1970, 1971-1980, 1981-1990, 1991-2000) bzw. das Baujahr (seit 2001). 

Es erfolgte eine räumliche Verschneidung zwischen der Gefahreninformation und dem GWR 

II-Datensatz. Mit Ausnahme von Sakralbauten wurde ein ökonomisches Modell zur Berechnung 

des Wertes der Gebäude angewendet unter Verwendung von (a) den vorliegenden 

Informationen des GWR II in Bezug auf Gebäudetyp, Anzahl der Geschosse und Nutzung, 

und (b) regional gemittelten Baukosten beruhend auf einer Kombination der Ansätze aus 

Keiler et al. (2006) und Kranewitter (2002). Die Baukosten wurden auf Neuwerte anstelle 

von Marktwerten gerechnet, was allgemeinen Versicherungsgrundsätzen entspricht, und 

wurden mit dem jeweiligen Index der Baukosten (Statistik Austria 2013) inflationsangepasst. 

Die räumlichen und zeitlichen Analysen des Schadenpotentials wurden aufgrund der 

Information zu Bauperiode und Baujahr der aktuell bestehenden Gebäude getätigt. Die 

Analyseergebnisse können demnach nicht dazu verwendet werden, um beispielsweise die 

tatsächliche (historische) Bevölkerungsentwicklung oder die tatsächliche (historische) 

Werteentwicklung in gefährdeten Bereichen abzuleiten, da diese Information nicht im 

Datensatz enthalten ist. Die Analyse stützt sich lediglich auf den Bauzeitpunkt der heute 

bestehenden Gebäude ab. 

ERGEBNISSE 

Im Folgenden werden Ergebnisse präsentiert, zunächst in Bezug auf die räumliche und 

anschließend in Bezug auf die zeitliche Analyse. Auffallend ist die generell bestehende 

Heterogenität zwischen einzelnen Gemeinden und Gebäudekategorien. 

In Österreich konnten 2.399.500 Gebäude eindeutig einer Information im GWR II zugeordnet 

werden, davon sind 319.026 (13,3 %) gegenüber den untersuchten Naturgefahren exponiert 


Tabelle 1: Exponierte Gebäude in Österreich (Einfach- und Mehrfachgefährdung). 

Einfachgefährdung 

Mehrfachgefährdung 

Bundesland 

Gebäude 

[N] 

Nichtexponierte 

Gebäude 

[N] 

Exponierte 

Gebäude 

[N] 

Exponierte 

Gebäude 

[%] 

Überschwemmung 

Wildbach 

[N] 

[N] 

Lawine 

[N] 


und Wildbach 

[N] 

Wildbach 

und 

Lawine 

[N] 


Über- 

und Lawine 

[N] 

schwemmung, 

Wildbach und 

Lawine [N] 

Burgenland 133.482 123.905 9.577 7,2 9.439 140 0 2 0 0 0 

Kärnten 185.693 161.782 23.911 12,9 17.012 8.466 188 1.660 95 10 10 

Niederösterreich 648.693 569.085 79.608 12,3 73.239 8.381 6 2.018 0 0 0 

Oberösterreich 425.718 378.307 47.411 11,1 37.836 12.471 137 2.950 22 71 10 

Salzburg 139.377 99.662 39.715 28,5 20.360 23.800 594 4.684 319 128 92 

Steiermark 381.484 331.065 50.419 13,2 27.953 25.695 460 3.530 130 52 23 

Tirol 192.381 141.735 50.646 26,3 25.635 24.631 4.465 2.975 924 276 90 

Vorarlberg 106.098 91.910 14.188 13,4 4.334 8.089 3.159 270 1.105 31 12 

Wien 186.574 183.023 3.551 1,9 3.551 0 0 0 0 0 0 

Summe 2.399.500 2.080.474 319.026 13,3 219.359 111.673 9.009 18.089 2.595 568 237 

(Tabelle 1). Von diesen knapp 2,4 Mio. Gebäuden sind 9 % (219.359) überschwemmungsgefährdet, 

und 5 % gegenüber alpinen Naturgefahren exponiert (Wildbach 111.673 und Lawine 

9.009). Insgesamt sind 298.248 Gebäude (93,5 % der exponierten Gebäude und 12,4 % des 

gesamten Gebäudebestandes) gegenüber einem Naturgefahrentyp exponiert, und 20.778 

Gebäude (6.5 % der exponierten Gebäude und 0,9 % des gesamten Gebäudebestandes) 

gegenüber mehr als einem Naturgefahrentyp (multi-exposure): 18.089 Gebäude sind von 

Flusshochwasser und Wildbächen betroffen, 2.595 von Wildbach- und Lawinengefahren, 

568 von Lawinen und Flusshochwasser, und 237 von allen drei Prozesstypen. 

Wird der Datensatz nach den unterschiedlichen Gebäudekategorien analysiert, ergibt sich 

folgendes Bild (vgl. Tabelle 2): Der Großteil der exponierten Gebäude entfällt auf Wohngebäude 

(Kategorien 1-3), aber es ist auch ein erheblicher Anteil an Hotels und ähnlichen 

Gebäuden (Gastgewerbe, Kategorie 4) und Gewerbebauten (Kategorien 5-8) exponiert. 

– Die insgesamt 2.056.322 Wohngebäude entsprechen 85,7 % des Gebäudebestandes in 

Österreich, und 12,62 % davon (259.687) sind exponiert. 

– Die insgesamt 140.470 Gewerbebauten entsprechen 5,86 % des Gebäudebestandes in 

Österreich, und 21,06 % davon (29.593) sind exponiert. 

– Die insgesamt 37.272 Gebäude des Gastgewerbes entsprechen 1,55 % des Gebäudebestandes 

in Österreich, und 23,04 % davon (8,589) sind exponiert. 

Die räumliche Abfrage auf Gemeindeebene ist in Abbildung 1 wiedergegeben. Die Referenzgröße 

der Abbildung sind die von den jeweiligen Prozessen betroffenen Gemeinden. In Bezug 

auf Lawinen beträgt die mittlere Anzahl exponierter Gebäude 30,4 pro Gemeinde (für von 

Lawinen gefährdete Gemeinden). Die höchste Exposition mit > 3 Standardabweichungen 

findet sich in Gemeinden Westösterreichs nahe des Alpenhauptkamms (Vorarlberg und Tirol). 

Die mittlere Anzahl wildbach-exponierter Gebäude beträgt 87,7 pro Gemeinde (für wildbachgefährdete 

Gemeinden). Gemeinden mit einer hohen Exposition finden sich vor allem im 

Bundesland Salzburg, sowie in einigen Gemeinden des Tiroler Unterlandes. Überschwemmung 

entlang der Flüsse betrifft größere Landesflächen, und im Mittel sind 97,1 Gebäude je 


Tabelle 2: Exponierte Gebäude in Österreich (nach Gebäudekategorien). 

Gebäudekategorie 

Gebäude 

[N] 

Gebäud 

e [%] 

Nichtexponiert 

e 

Gebäude 

[N] 

Exponi 

erte 

Gebäu 

de [N] 

Exponie 

rte 

Gebäud 

e [%] 

Einfachgefährdung 

Überschw 

emmung 

[N] 

Wildbac 

h [N] 

Lawine 

[N] 

Mehrfachgefährdung 

Übersch 

wemmu 

ng und 

Wildbac 

h [N] 

Wildbach 

und 

Lawine 

[N] 

Überschw 

emmung 

und 

Lawine 

[N] 

Überschw 

emmung, 

Wildbach 

und 

Lawine 

[N] 

Gebäude mit einer Wohnung (1) 1.510.151 62,94 1.335.938 174.213 11,54 119.189 60.424 4.607 8.600 1.280 221 94 

Gebäude mit zwei oder mehr 

Wohnungen (2) 

Wohngebäude für Gemeinschaften 

(3) 

542.118 22,59 457.359 84.759 15,63 56.195 32.477 2.308 5.421 681 177 58 

4.053 0,17 3.338 715 17,64 528 204 38 37 18 3 3 

Hotels und ähnliche Gebäude (4) 37.272 1,55 28.683 8.589 23,04 4.217 4.622 994 895 302 82 35 

Bürogebäude (5) 31.420 1,31 25.551 5.869 18,68 4.815 1.325 63 315 17 5 3 

Groß- und Einzelhandelsgebäude 

(6) 

Gebäude des Verkehrs- und 

Nachrichtenwesens (7) 

32.583 1,36 25.646 6.937 21,29 5.612 1.761 73 481 25 5 2 

4.319 0,18 3.525 794 18,38 544 295 53 73 24 9 8 

Industrie- und Lagergebäude (8) 72.148 3,01 56.155 15.993 22,17 12.874 4.113 248 1.139 86 30 13 

Gebäude für Kultur- 

/Freizeitzwecke, Bildungs- und 

Gesundheitswesen (9) 

Landwirtschaftliches Nutzgebäude 

(10) 

21.082 0,88 17.041 4.041 19,17 3.142 1.113 90 264 35 11 6 

18.496 0,77 17.341 1.155 6,24 624 501 121 66 24 4 3 

Privatgarage (11) 48.819 2,03 43.412 5.407 11,08 3.686 1.811 136 193 31 5 3 

Kirchen, sonstige Sakralbauten 

(12) 

4.384 0,18 3.896 488 11,13 289 200 47 33 15 2 2 

Pseudobaulichkeit (13) 4.536 0,19 3.683 853 18,81 797 71 3 18 0 0 0 

Sonstiges Bauwerk (14) 68.119 2,84 58.906 9.213 13,52 6.847 2.756 228 554 57 14 7 

Summe 2.399.500 100 2.080.474 319.026 13,30 219.359 111.673 9.009 18.089 2.595 568 237 


Abbildung 1: Anzahl der gegenüber Überschwemmung, Wildbächen und Lawinen exponierten Gebäude, dargestellt auf Gemeindeebene 

und in Form der Abweichung vom Mittelwert. 


Gemeinde exponiert. Aufgrund der hohen Anzahl an Gebäuden im Bereich der HORA-Modellierung 

sticht die Stadt Wien heraus, ebenso wie die Gemeinden entlang der Donau. 

2,500,000 

2,000,000 

a) 

Non-exposed buildings 

Exposed to snow avalanches 

Exposed to torrential flooding 

Exposed to river flooding 

3,500 

3,000 

2,500 

c) 

Buildings [N] 

1,500,000 

1,000,000 

Buildings [N] 

2,000 

1,500 

1,000 

500,000 

500 

0 

1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 

0 

1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 

Year 

Year 

Exposed to river flooding [N] 

Exposed to torrential flooding [N] 

Exposed to snow avalanches [N] 

7 

6 

5 

b) 

Total amount of buildings 

Exposed to snow avalanches 

Exposed to torrential flooding 

Exposed to river flooding 

12.0 

10.0 

d) 

Factor of growth [-] 

4 

3 

Buildings [%] 

8.0 

6.0 

4.0 

2 

1 

2.0 

0 

1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 

Year 

0 

1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 

Year 

Abbildung 2: Zeitliche Kennzahlen des Gebäudebestandes in Österreich. 

In Abbildung 2 ist die zeitliche Analyse des Gebäudebestandes in Österreich zusammengefasst. 

Insgesamt steigen über den Untersuchungszeitraum sowohl der Gesamtgebäudebestand, 

als auch die Anzahl der in den drei Gefahrengruppen exponierten Gebäude kontinuierlich an. 

Weder im Bereich der Überschwemmungen, noch im Bereich der Wildbäche sind besonders 

hohe Steigerungsraten auffallend (Abbildung 2a), hingegen ist seit den 1950er Jahren die 

Anzahl der nicht-exponierten Gebäude stark gestiegen. Darüber hinaus zeigt sich, dass die 

Anzahl der gegenüber Lawinen exponierten Gebäude gemessen am Gesamtbestand der 

Gebäude vernachlässigbar ist, diese Gruppe wurde aus diesem Grund in vorliegendem Bericht 

nicht weiter beschrieben. Seit 1919 ist die Anzahl der Gebäude in Österreich um den Faktor 

6,4 von 373.067 auf 2,4 Millionen gestiegen. Da für rund 4,2 % der Gebäude das Baujahr in 

der Datengrundlage fehlt, wurden diese von der weiteren Analyse ausgeschlossen. 

Die Anzahl der gegenüber Überschwemmung exponierten Gebäude ist um den Faktor 6,5 

gestiegen (von 33.697 auf 219.359), siehe auch Abbildung 2b, die Anzahl der gegenüber 

Wildbachgefahren exponierten Gebäude um den Faktor 5.9 (von 18.797 auf 111.673). 

In Abbildung 2b ist die kumulative Steigerungsrate des Gebäudebestandes dargestellt, zum 

einen für den Gesamtgebäudebestand und zum anderen für die gegenüber den drei Prozessgruppen 

exponierten Gebäude. Während die Steigerungsrate für die überschwemmungsge- 


fährdeten Gebäude leicht über jener des Gesamtbestandes liegt, ist die Rate für die wildbachgefährdeten 

Gebäude seit 1990 abnehmend und liegt unterhalb jener des Gesamtbestandes. 

In Abbildung 2c ist die durchschnittliche jährliche Summe der Neubauten für die Exposition 

gegenüber den drei Gefahrengruppen dargestellt. Bis in die 1970er Jahre ist diese deutlich 

angestiegen, ab 1980 nimmt diese leicht ab und in der letzten Dekade ist wiederum eine 

leichte Zunahme zu verzeichnen. Auffallend ist, dass die Kurven für Überschwemmung und 

Wildbachgefährdung einem ähnlichen Muster folgen. Im letzten untersuchten Jahr (2012) 

wurden 78 Gebäude in lawinengefährdeten, 1.028 in wildbachgefährdeten, und 2.172 in 

überschwemmungsgefährdeten Bereichen errichtet. Die zeitliche Entwicklung der Bautätigkeit 

in gefährdeten Bereichen ist in Abbildung 2d dargestellt als Verhältnis zwischen den 

jährlichen Neubauten innerhalb von Gefahrenzonen und der Gesamtzahl jährlicher Neubauten. 

Über den gesamten Untersuchungszeitraum nahm der Prozentsatz von 10,6 auf 8,1 % 

(Überschwemmung), und von 4,2 auf 3,8 % (Wildbach) ab. 

FAZIT 

Eine detaillierte und räumlich verortete Bewertung der gegenüber Naturgefahren exponierten 

Gebäude in Österreich wurde durchgeführt für überschwemmungs-, wildbach- und 

lawinengefährdete Bauwerke. Insgesamt ist im Zeitverlauf ein Anstieg des Schadenpotentials 

nachweisbar. Die räumliche Analyse zeigt, dass einige Gemeinden einen deutlich überdurchschnittlichen 

Anstieg und andere einen unterdurchschnittlichen Anstieg der Gebäudezahlen 

aufweisen (vgl. auch Fuchs et al. 2015). Dies spiegelt einerseits die Topographie des Landes 

wider, aber auch die unterschiedlichen wirtschaftlichen Aktivitäten: Beherbergungsbetriebe 

sind vor allem in den westlichen Bundesländern gegenüber alpinen Naturgefahren exponiert, 

während Gewerbebauten und Gebäude der Freizeitnutzung vor allem von Überschwemmung 

betroffen sind. Wohngebäude zeigen demgegenüber eine durchschnittliche Exposition. 

Eine flächendeckende objektbasierte Bewertung hat Vorteile im Vergleich zu kleinräumigen 

fallstudienbasierten Analyseansätzen, da kleinskalige Verteilungsmuster und unterschiedliche 

Auftretensmuster von gefährlichen Prozessen zu einer Verzerrung der Ergebnisse führen 

können. Die dargestellten Ergebnisse können genutzt werden, um flächendeckende Risikoabschätzungen 

(beispielsweise im Rahmen der Umsetzung der EU-Hochwasserrisiko-Managementrichtlinie) 

durchzuführen. Darüber hinaus können flächendeckende Aussagen zum 

Schadenausmaß zukünftiger Naturgefahrenereignisse getätigt werden. 


LITERATUR 

- de Moel H., Aerts J. (2011). Effect of uncertainty in land use, damage models and 

inundation depth on flood damage estimates. Natural Hazards 58: 407-425. 

- Fuchs S., Keiler M., Sokratov S., Shnyparkov A. (2013). Spatiotemporal dynamics: 

the need for an innovative approach in mountain hazard risk management. 

Natural Hazards 68: 1217-1241. 

- Fuchs S., Keiler M., Zischg A. (2015). A spatiotemporal multi-hazard exposure assessment 

based on property data. Natural Hazards and Earth System Sciences 15: 2127-2142. 

- Hilker N., Badoux A., Hegg C. (2009). The Swiss flood and landslide damage database 

1972-2007. Natural Hazards and Earth System Sciences 9: 913-925. 

- Jongman B., Koks E., Husby T., Ward P. (2014). Increasing flood exposure in the 

Netherlands: implications for risk financing. Natural Hazards and Earth System Sciences 14: 

1245-1255. 

-Keiler M., Zischg A., Fuchs S., Hama M., Stötter J. (2005). Avalanche related damage 

potential – changes of persons and mobile values since the mid-twentieth century, case study 

Galtür. Natural Hazards and Earth System Sciences 5: 49-58. 

- Keiler M., Zischg A., Fuchs S. (2006b). Methoden zur GIS-basierten Erhebung des 

Schadenpotenzials für naturgefahreninduzierte Risiken. In: Strobl J., Roth C. (eds) GIS 

und Sicherheitsmanagement. Wichmann, Heidelberg: 118-128. 

- Kranewitter H. (2002). Liegenschaftbewertung. GESCO, Wien. 

- Republik Österreich (1975). Forstgesetz 1975. BGBl 440/1975. 

- Republik Österreich (1976). Verordnung des Bundesministers für Land- und Forstwirtschaft 

vom 30. Juli 1976 über die Gefahrenzonenpläne. BGBl 436/1976. 

- Republik Österreich (2009). Bundesgesetz, mit dem das Registerzählungsgesetz, das 

Bundesgesetz über das Gebäude- und Wohnungsregister, das Bundesstatistikgesetz 2000 

und das E-Government-Gesetz geändert werden. BGBl 125/2009. 

- Republik Österreich (2013). Bundesgesetz über das Gebäude- und Wohnungsregister 

(GWR-Gesetz). BGBl 9/2004 i.d.F. 1/2013. 

- Rudolf-Miklau F., Sereinig N. (2009). Festlegung des Bemessungshochwassers: 

Prozessorientierte Harmonisierung für Flüsse und Wildbäche. Österreichische 

Wasser- und Abfallwirtschaft 61: 27-32. 

- Statistik Austria (2013). Baupreisindex für den Hoch- und Tiefbau. Statistik Austria, Wien. 

- Stiefelmeyer H., Hlatky T. (2008). HORA - An Austrian platform for natural hazards as a 

new way in risk communication. In: Mikoš M., Hübl J., Koboltschnig G. (eds) Internationales 

Symposion Interpraevent, Band 1, Dornbirn: 229-236. 

- Zischg A., Fuchs S., Keiler M., Stötter J. (2005). Temporal variability of damage potential on 

roads as a conceptual contribution towards a short-term avalanche risk simulation. Natural 

Hazards and Earth System Sciences 5: 235-242. 


- Zischg A., Schober S., Sereinig N., Rauter M., Seymann C., Goldschmidt F., Bäk R., 

Schleicher E. (2013). Monitoring the temporal development of natural hazard risks as a basis 

indicator for climate change adaptation. Natural Hazards 67: 1045-1058. 

DANKSAGUNG 

Die Studie wurde vom Bundesministerium für Land- und Forstwirtschaft, Umwelt und 

Wasserwirtschaft sowie vom Versicherungsverband Österreich unterstützt. 



Event-based rapid landslide mapping including 

estimation of potential human impacts on landslide 

occurrence: a case study in Lower Austria 

Karin Gokesch 1 ; Thomas Glade 1 ; Joachim Schweigl 2 

ABSTRACT 

Within any landslide susceptibility, hazard and risk assessment, landslide inventories play a 

crucial part determining the quality of further analyses. The applied methodology describes a 

fast way of assessing landslides related to a heavy rainfall event in Lower Austria in May 

2014. Based on reported damages and aerial photographs a first inventory was created. 

Information on the location, extent and human activities potentially influencing landslide 

occurrence were assessed. The resulting event-based landslide inventory showed that a quick 

landslide mapping can provide an overview of affected areas following a rainfall event. 

The extent of the landslides was assessed allowing a better estimation of potential hazards 

accompanying heavy precipitation in the study area. Analyzing the potential human impact 

on landslide occurrence showed that numerous indirect and direct influences, such as slope 

undercutting, forest roads or drainage, can serve as preparatory and predisposing factors for 

landslide occurrence. Although the direct human impact on landslide-triggering could not be 

determined, landslide mapping immediately after the event provided detailed documentation 

of human influences on landslide occurrence. 

KEYWORDS 

landslide mapping; landslide inventory; event documentation; human impact 

INTRODUCTION AND BACKGROUND 

Landslide investigations are important when assessing hazard and risk in any region. 

Especially in alpine areas with often highly varying lithological, morphological and environmental 

settings, landslides are relevant phenomena influencing the landscape (Schweigl & 

Hervás 2009). Landslide mapping, as an essential part of any landslide susceptibility, hazard 

and risk assessment, is a crucial element determining the outcome and applicability of any 

final product such as detailed slope stability information or any kind of inventory, susceptibility, 

hazard or risk maps (Glade et al. 2005). Especially in the context of specific triggering 

factors, the assessment of all associated landslides triggered by respective hydro-meteorological 

events is of particular interest. An inventory of event-related landslides, including debris 

slides and flows, translational or rotational slides or rock falls (according to Cruden & Varnes 

1996), not only provides detailed information on the time of occurrence, but also on the 

1 University of Vienna, Geomorphological Systems and Risk Research, Vienna, AUSTRIA, karin.gokesch@univie.ac.at 

2 Geological Survey, Federal State Government of Lower Austria, AUSTRIA 



specific landslide types triggered by the particular heavy rainfall event (Guzzetti et al. 2012). 

A landslide inventory in general represents the total number of landslides and their location 

in a certain area and can also include information on the type of movement and/or the 

time-scale of each event (Guzzetti et al. 2000). Often such inventories cannot be generated 

directly following a triggering event because of insufficient resources to perform the immediate 

investigations, which results in restricted data availability (Guzzetti et al. 2012). Several 

methods for creating landslide inventories can be applied, depending on the specific aim of 

the study and data availability (Hervás 2013). For example, creating an inventory based on 

remote sensing data has, in the past few years, become a very common method of generating 

new and extending existing inventories (Petschko et al. 2010). Although the interpretation of 

aerial photographs or landform models based on laserscanning is a rather quick way of 

generating a landslide inventory, there are often no such data available directly after a 

triggering rainfall event (Ghosh et al. 2012). Furthermore, remote sensing data do usually not 

include exact information on the time of occurrence for each landside, but only the date on 

which the photograph or laserscan data were taken, thus only showing all landslides that 

occurred in the period prior to that date. 

Another method of creating landslide inventories is the interpretation of historical data from 

different archives (e. g. literature sources, chronicles, newspaper reports; Glade 2001, 2005) 

or expert knowledge (Guzzetti 2005). But certain issues of data availability and accuracy can 

also arise when analyzing different data sources (Petschko et al. 2014b). Compiling a 

landslide inventory can also be done by field-mapping. Even though field-mapping can be 

very time-consuming and often landslides cannot directly be identified in the field (due to 

e.g. vegetation, removal of dislocated material; Bardi et al. 2014), it also shows numerous 

advantages compared to other methods (Guzzetti et al. 2012). Field-mapping can be done 

directly after a certain triggering event, thus providing information on the time of occurrence 

and also the direct relation to the triggering event can be assessed. 

Furthermore field-mapping allows for a detailed investigation of potential human impacts on 

predisposing factors of landslide occurrence, such as slope undercutting, deforestation, 

drainage etc. to determine the role of humans within in landslide assessment. During 

field-work it is possible to examine, and often quite clearly to determine, which kind of 

factors might have influenced each single landslide leading to the slope failure. Therefore the 

method applied here presents a rapid way to record landslide occurrences to better estimate 

and further analyze the underlying factors and consequent process dynamics. The presented 

study aims to apply a quick way of mapping specific landslides related to a certain triggering 

event in order to create a detailed landslide inventory. Furthermore this inventory also 

includes information on potential human impact on landslide occurrence, thus providing a 

first estimation of potential human influences on slope stability and the subsequent effects on 

the potentially related hazards and risks. A fast inventory of landslides related to certain 


triggering event, as presented within this study, can further serve as valuable input for future 

risk assessment and management (Guzzetti et al. 2012). 

STUDY AREA 

The study area is located in the western part of Lower Austria (Figure 1) within the districts 

of Amstetten, Waidhofen/Ybbs, Scheibbs, Lilienfeld and St. Pölten. The landscape is characterized 

by forests (mainly in the southern part) and farmland (Petschko et al. 2014a) with 

settlements situated predominantly in the valley bottoms. The area is based mainly within the 

lithological units of the Flyschzone and the Northern Calcerous Alps, in particular including 

Limestone and Dolomite (Figure 1). 

Figure 1: Lithological units and general location of the study area in Lower Austria, Austria. 

The study area is situated in a transition zone between marine and continental climate 

(Petschko 2014), with an average annual precipitation of 1200-2000 mm including extreme 

single rainfall events (Hydrographic Service of Lower Austria 2014). Throughout the whole 

region the highest precipitation and temperature values are usually recorded during the late 

spring and summer months, from May to August (Figure 2). In 2014 a heavy rainfall event 

occurred during this generally wet season, with a maximum of total precipitation in May 

of 453 mm in Lunz/See (district Scheibbs) or 348 mm in Waidhofen/Ybbs (ZAMG 2014). 

This rainfall event represents an event with a 30-year recurrence interval (Hydrographic 

Service of Lower Austria 2014), and mainly occurred from 15th to 18th of May followed by 


further precipitation peaks from 23rd to 28th of May, leading to numerous floods and 

landslides. This event thus served as a basis for the mapping and development of the 

event-based landslide inventory presented in this study. 

Figure 2: Average precipitation and temperature values in the districts of Amstetten, Lilienfeld, Scheibbs, St. Pölten and Waidhofen/Ybbs 

(modified after http://de.climate-data.org/index.html) 

METHODS 

Data acquisition 

The database used for generating the landslide inventory consists of several datasets provided 

by the provincial government of Lower Austria and the Geological Survey of Austria (Table 

1). 24 damage reports were included, that were already available shortly after the rainfall 

event. Furthermore aerial images of landslides, acquired during a helicopter flight carried out 

by the Geological Survey of Lower Austria, were used as a basis for the field investigations 

resulting in a total of 56 landslides. Each of these events were geographically located using 

Table 1: Available data. 

Level Year Resolution 

DGM & derivates District 2006-2009 1 m 

Orthophoto Federal state 2002/2005/2007 25 cm 

District boundaries District 2004 1:50 000 

Municipality boundaries Municipality 2004 1:50 000 

Rivers Federal state 2001 1:10 000 

Forest-cover Federal state 1986 1:50 000 

Road network Federal state a. n. 1:50 000 

Lithology Federal state 2011 1:200 000 


GPS coordinates, thus resulting in a first inventory of landslides including the location and 

number of events related to the May 2014 heavy rainfall event. 

Mapping 

Using these datasets as a basis, a detailed mapping to create the landslide inventory was 

carried out (Figure 3). Several field surveys were performed between May and September 

2014. Within the field surveys a specific mapping form was developed based on the work of 

Cruden and Varnes (1996) and Turner and Schuster (1996). This form includes information 

on the type of movement (rock fall, debris slide, debris flow or complex movements), 

geographical coordinates of landslide location, detailed extent and other relevant criteria such 

as land use, lithology, exposition and slope angle. Further parameters like the distance of the 

landslide to roads, buildings and watercourses as well as records of slope undercutting, 

deforestation or drainage were added in order to be able to estimate the potential human 

impact on landslide occurrence. In addition some information on planned and/or already 

performed mitigation measures were included to establish a highly detailed dataset for each 

landslide. 

Using the already available information on landslides reported by the Geological Survey of 

Lower Austria, a first mapping route was defined (Figure 3). This route served as a basis for 

the field work and was expanded adding surrounding areas and additional landslides based 

on conversations with locals and visible slope movements in the field, which had not been 

reported prior. The extent of all landslides near this route was measured, slope-profiles of 

each landslide were generated and numerous photographs were taken as further accompanying 

information. A specific code was assigned to each landslide to allow a precise identification 

within the landslide inventory. 

Figure 3: Flow-chart of the methodology for the mapping procedure. 


Based on this mapping procedure, the landslide extent was digitized taking also into account 

the different types of movement. In case of combined processes and complex movements, e. 

g. a debris slide resulting in a debris flow further downslope, the different process-types were 

separately mapped, measured and digitized to allow a more precise analysis of the process 

distributions. 

Delineation of mapped area 

For potential further analyses the completely mapped area, i. e. the area where a completeness 

of the inventory can be assumed, had to be delineated (Figure 3). Since not all areas 

within the whole study region of western Lower Austria could be fully mapped due to 

resource restrictions, certain adjustments to confine the mapped area were needed. Due to 

missing digital imagery and satellite data in the period directly after the May 2014 event and 

because of the urgent need to rapidly assess the landslide information, it was decided to 

specifically include only the areas that were actually investigated during the field surveys. 

This limitation is especially important, since the inventory and the related data on landslides 

in May 2014 might serve as input for further studies, for example for landslide susceptibility 

Figure 4: Detail of delineation of the mapped area, showing (a) the mapping route, (b) with a 1000 m radius representing maximum 

visibility, (c) with exclusion of forest areas and shadowing effects and (d) after including visibility of 50 m for forested areas 

representing the final mapped area. 


and hazard modeling. In such a case it is of major importance to only include these landslides 

and surrounding areas in the modeling analysis, which were covered within the field survey 

in order to get valid and therefore reliable results. 

Starting from the mapping route based on damage reports (Figure 4a) a maximum visibility 

of the landscape of 1000 m was assumed representing the average visible area taking into 

account a minimum landslide size of 100 m². This 1000 m buffer was added on all sides of the 

mapping route (Figure 4b). Additionally, densely forested parts were excluded due to reduced 

visibility. Within the mountainous area, shadowing effects of mountain ridges or valley 

corners were eliminated by reducing the total mapped area at the respective locations to the 

visible regions (Figure 4c). Finally within the remaining forest zones, a visibility of 50 m was 

assumed and implemented, representing the minimum visibility from the mapping route 

(Figure 4d). After these adjustments, the resulting study area represents the mapped area 

(Figure 5), where one can assume that all landslides that occurred during the May 2014 

rainfall event have been recognized, investigated and mapped. Thus the final landslide 

inventory can be regarded as complete in this area. 

RESULTS 

The investigated road network extends to a total of 292.76 km. After the abovementioned 

restriction (as deduced in Figure 4) the total mapped area covers a region of 112.57 km² 

(Figure 5). From the preliminary inventory of 56 already known events, 18 of these needed 

to be excluded prior to or in course of the field work. This had to be done because not all 

landslides identified on the aerial photographs could be specifically related to the May 2014 

rainfall event or because mapping in the field was not possible due to restricted access to the 

sites (e. g. non-accessible private properties). In some cases no landslide tracks could be found 

in the field, which also led to an elimination of these landslides from the inventory. In total 

52 sites were visited, but as mentioned above, not all could be mapped and thus have not 

been included in the inventory. 

With an addition of four landslides that were identified in the field but not formerly being 

recognized, a total of 42 landslides were investigated in detail. The landslide types included 28 

debris slides, one distinct debris flow, two rock falls and eleven complex movements where 

earth and debris slides turned into debris flows (Table 2). 

Table 2: Mapped landslides for each district distinguished by process types. 

Process type 

District 

Amstetten Lilienfeld Scheibbs St. Pölten 

Debris slide 7 3 2 11 

Debris flow 1 – – – 

Rock fall 1 – 1 – 

Complex movements 5 – 2 3 

∑ 14 3 5 14 


Figure 5: Mapped area during the field investigations and landslide inventory related to the May 2014 heavy rainfall event in the 

investigated study region. 

The final landslide inventory (Figure 5) showed not only the dimensions and general 

geographic information for each landslide, but also the differences of landslide occurrence 

related to land use and the potential influences of human activities such as drainage, deforestation 

or road construction. Most of the landslides could be detected in forest areas (45 %), 

followed by grassland (24 %) as well as pasture (21 %). Especially the rather high abundance 

in forest areas indicates the importance to account for completeness of the inventory in 

forested zones thus only including areas that actually were mapped when delineating the 

study area for potential further analyses. 

Regarding the analysis of potential human impact on landslide occurrence, especially the 

scarp and source as well as the surrounding area was investigated for each landslide. Within 

the complex movements, only the combination of earth and debris slides and flows have 

been examined. Rock falls show different triggering dynamics and therefore did not contribute 

to complex movements. The vicinity to drainages or wells seems to act as a preparatory 

factor influencing the hydrology of a slope thus potentially changing its stability. Forest roads 

and slope undercutting near the scarp area might also have a potential effect on stability as 

well as deforestation. For all landslides all these factors were recorded and also the multiple 

occurrence of different potentially influencing factors was registered. Table 34 shows the 

results of this investigation in relation to the land use. The total numbers indicate that within 


the investigated study area forest roads might be regarded as the main potential influence on 

slope stability. The highest number of landslides occurring directly on or in the vicinity of 

forest roads was found in forest areas where the roads directly cut through the tree-cover and 

change the topography through cuts in upslope and fills in downslope directions. This can be 

interpreted as a potential decrease of local slope stability leading to potential failures. But also 

drainage outlets could be identified as potential influences on slope stability in seven cases. 

These outlets were located above or, in some cases, directly within the landslide-mass, which 

is why the drainage influence on the movement at the specific sites cannot be fully excluded. 

Table 3: Potential human influences on landslide occurrence for each land use type. 

Forest roads Deforestation Wells Drainage Slope undercutting 

Forest 13 1 1 1 1 

Grassland 4 – 2 2 3 

Pasture 5 – 1 4 – 

Garden 1 – – – 1 

∑ 23 1 4 7 5 

CONCLUSIONS 

To rapidly create a landslide inventory following a heavy rainfall event is often necessary to 

better implement response strategies and estimate the effects of such extreme precipitation. 

The methodology applied within this study gives the indication, that it is possible to implement 

a detailed inventory on the basis of damages reported directly after an event with rather 

limited resources (e.g. data and time restrictions). Nevertheless, this method has proven to 

provide valuable information, not only for direct reconstruction and subsequent mitigation 

measures, but also for a first approximation of the underlying process-dynamics and a 

preliminary overview of potential impacts of human activities on slope stability. 

The delineation of the study area, i. e. the completely mapped area, carried out within this 

study has proven to be a necessary step towards creating a database for potential further 

analyses. It is crucial to explicitly define the investigated area within any susceptibility, hazard 

or risk mapping in order to provide useful information for further modeling or validation of 

existing models. The method for defining the mapped area applied here did not consider 

administrative boundaries or river catchments, as it is often the case in landslide analyses. 

On the contrary these boundaries were completely neglected to create an inventory related 

to the May 2014 rainfall event, which spread over several districts and river catchments but 

did not cover them all. Still the tight clipping of the study area alongside the mapping route 

shows the restrictions of the applied methodology with only a small area being completely 

mapped and large forested areas not being included in the inventory. The applied field 

mapping based on known events showed, that a fast way of creating an event-related 

landslide-database is to investigate these events first. As an immediate second step, this 

information has to be expanded by designing a mapping route and then extending the 

existing data base entries in order to receive a final detailed and most complete inventory. 


It has to be mentioned that not all landslides that occurred during the May 2014 event could 

be mapped, because such wide-spread mapping would have required much more time. When 

mapping landslides it is often crucial to act quickly within the given resources, since remedial 

works along the transport corridors often have to be done immediately following the event. 

Therefore it is necessary to investigate the events directly after their initiation and not wait 

until sufficient data is available. Even though not all landslides in the area could be mapped, 

a detailed inventory within the defined study area was established. This inventory contains, 

besides the specific time of occurrence for each recorded landslide, also detailed information 

on single landslide locations and the human activities that have a potential impact on their 

occurrence. 

The evaluation of the potential human impact showed that numerous activities including 

slope undercutting, drainage or deforestation have likely influenced slope stability. But still 

these relations are just field-based observations and have not been assessed in full detail. 

Indeed, a more profound analysis is required. Furthermore, to better understand the exact 

dynamics between the different processes leading to a landslide, further analyses are 

indispensable. The inventory established within this research only outlines the different 

potential impacts on slope stability within the study area and does not give a definite relation 

of human activities to landslide occurrence. However, it is a valuable source for practitioners 

and for further more detailed analysis. 


The authors thank the Provincial Government of Lower Austria and the Geological Survey of 

Lower Austria for the support and the provision of data as well as the municipalities and the 

governing mayors for their support during the field work. Further we would like to thank our 

colleagues from the research project “MoNOE – Method development for landslide susceptibility 

maps for Lower Austria” for their assistance and the reviewers for constructive 

comments. 


REFERENCES 

- Bardi F., Frodella W., Ciampalini A., Bianchini S., Del Ventisette C., Gigli G., Fanti R., Moretti 

S., Basile G. & Casagli N. (2014): Integration between ground based and satellite SAR data 

in landslide mapping: The San Fratello case study, in: Geomorphology 223, S. 45-60. 

- Cruden D. M. & Varnes D. J. (1996): Landslide types and processes. TRB Special Report, 

National Academy Press, Washington, S. 36-75. 

- Glade T. (2001): Landslide hazard assessment and historical landslide data - an inseparable 

couple? In: Glade T., Albini P. & Frances F. (Hrsg.): The use of historical data in natural hazard 

assessments, in: Advances of Technological and Natural Hazard Research, Kluwer, S. 153-169. 

- Glade T. (2005): Der Nutzen der historischen Daten für die naturwissenschaftliche Gefahrenzonierung, 

in: Schenk W. & Dix A. (Hrsg.): Naturkatastrophen und Naturrisiken in der 

vorindustriellen Zeit und ihre Auswirkungen auf Siedlungen und Kulturlandschaft.- Siedlungsforschung. 

Archäologie - Geschichte - Geographie 23, Selbstverlag ARKUM e.V. Bonn, S. 

9-18. 

- Glade T., Anderson M. & Crozier M.J. (Hrsg..) (2005): Landslide hazard and risk. Wiley. 

- Ghosh S., van Westen C. J., Carranza E. J. M., Jetten V. G., Cardinali M., Rossi M. & 

Guzzetti F. (2012): Generating event-based landslide maps in a data-scarce Himalayan 

environment for estimating temporal and magnitude probabilities, in: Engineering Geology 

128, S. 49-62. 

- Guzzetti F., Mondini A. C., Cardinali M., Fiorucci F., Santangelo M. & Chang K.-T. (2012): 

Landslide inventory maps: New tool for an old problem, in: Earth-Science Reviews 112, S. 

42-66. 

- Guzzetti F. (2005): Landslide hazard and risk assessment. Dissertation, Rheinisch Friedrich-Wilhelms 

Universität, Bonn. 

- Guzzetti F., Cardinali M., Reichenbach P. & Carrara A. (2000): Comparing Landslide Maps: A 

Case Study in the Upper Tiber River Basin, Central Italy, in: Environmental Management 25 

(3), S. 247-263. 

- Hervás J. (2013): Landslide Inventory, in: Bobrowsky P. T. (Hrsg.): Encyclopedia of Natural 

Hazards. Springer-Verlag, Dordrecht, S. 610-611. 

- Hydrographical Survey of Austria (Hydrographischer Dienst Österreich) (2014): N2-Auswertung 

– Niederschlag 15. Mai 2014, 16. Mai 2014, 28. Mai 2104.(online: http://www.noel.gv. 

at/Externeseiten/wasserstand/analysen/nlv/ereignisse_nlv.htm) 

- Petschko H. (2014): Challenges and solutions of modelling landslide susceptibility in 

heterogeneous regions – preparing maps for spatial planning in Lower Austria. Dissertation, 

Universität Wien. 

- Petschko H., Bell R. & Glade T. (2014a): Relative age estimation at landslide mapping on - 

LiDAR derivates: revealing the applicability of land cover data in statistical susceptibility 

modeling, in: Sassa K. et al. (Hrsg.): Landslide Science for a Safer Geoenvironment, Vol. 2, 

Springer, S. 337-343. 


- Petschko H., Brenning A., Bell R., Goetz J. & Glade T. (2014b): Assessing the quality of 

landslide susceptibility maps – case study Lower Austria, in: Natural Hazards and Earth 

System Sciences 14, S. 95-118. 

- Petschko H., Glade T., Bell R., Schweigl J. & Pomaroli G. (2010): Landslide inventories for 

regional early warning systems, in: Malet J. P., Glade T. & Casagli N. (Hrsg.): Proceedings of 

the International Conference “Mountain Risks: Bringing Science to Society”, CERG Editions, 

Strasbourg, S. 277-282. 

- Schweigl J. & Hervás J. (2009): Landslide Mapping in Austria. JRC Scientific and Technical 

Reports, European Commission, Luxemburg. 

- Turner A. K. & Schuster R. L. (Hrsg.) (1996): Landslides – Investigation and Mitigation. 

Special Report 247, Transportation Research Board, National Research Council, National 

Academy of Sciences. 

- ZAMG (2014): Witterungsübersicht Mai 2014 [online: http://www.zamg.ac.at/zamgWeb/ 

klima/klimarueckblick/archive/2014/05/wiewars05-14.pdf, Zugriff 26.02.2015, 12.45 Uhr] 



Integral protection concept "Bielzug" 

Integrales Schutzkonzept Bielzug 

Nicole Oggier, MSc. 2 ; Christoph Graf, dipl. Geogr. 1 ; Reynald Delaloye, Prof. Dr. 3 ; André Burkard, dipl. Kult.ing. 2 

ABSTRACT 

The basic disposition to natural hazards and the resulting risk situation in many Alpine 

catchments has changed significantly in recent years, partially due to climate change. Six 

torrents located in the highly-affected Matter Valley in Canton Valais, Switzerland (communities 

of St. Niklaus and Randa), the Grosse Grabe, Fallzug, Geisstriftbach, Birchbach and 

Dorfbach, were studied in 2013. During analysis, different debris flow runout scenarios were 

developed and modelled using the RAMMS::DEBRIS FLOW 1.5 model. Furthermore, a 

revised hazard map was created. The combination of computational modelling and field 

verification was quite useful in the project, representing an important and fruitful link 

between scientific research and practice. Based on the identified protection deficits, additional 

protection measures were developed for settlement areas, individual objects (e.g. power lines) 

and transportation routes (road and railway). The Bielzug example shows that structural 

measures are not always sufficient to protect settlements or transportation routes. A com - 

prehensive risk management that accounts for maintenance, spatial planning, organisational 

and structural measures is also essential. 


Die Grunddisposition für Naturgefahren und die daraus resultierende Gefahrenlage in vielen 

Alpinen Wildbacheinzugsgebieten hat sich in den letzten Jahren deutlich verändert, auch 

wegen des Klimawandels. In einer stark davon betroffenen Region in den Schweizer Alpen 

wurden 2013 die sechs Wildbäche Grosse Grabe, Bielzug, Fallzug, Geisstriftbach, Birchbach 

und Dorfbach (Gemeinden St. Niklaus und Randa im Mattertal (VS)) detailliert untersucht. 

Im Rahmen der Gefahrenanalyse wurden Szenarien für Murgangereignisse erarbeitet, diese 

mit dem Murgangmodell RAMMS::DEBRIS FLOW 1.5 modelliert und eine überarbeitete 

Gefahrenkarte erstellt. Die Kombination Modellierungen mit Simulationsprogrammen und 

Verifizierung durch Feldbegehungen hat sich im Rahmen des vorgestellten Projekts sehr gut 

bewährt. Sie stellt somit eine wichtige und gewinnbringende Verbindung zwischen wissenschaftlicher 

Grundlagenforschung und Praxis dar. Basierend auf den erkannten Schutzdefiziten 

wurden ergänzende Schutzmassnahmen zur Sicherung von Siedlungsgebieten, besonders 

schadenanfälligen Einzelobjekten (u.a. Stromleitungen) und Verkehrswegen (Strasse und 

Bahn) erarbeitet. Das Beispiel Bielzug zeigt auf, dass bauliche Massnahmen zum Schutz von 

Siedlungen und Verkehrswegen alleine nicht immer ausreichen. Es braucht ein umfassendes 

1 WSL Swiss Federal Research Institute Birmensdorf, SWITZERLAND, christoph.graf@wsl.ch 

2 wasser/schnee/lawinen, Ingenieurbüro A. Burkard AG, Brig-Glis, SWITZERLAND 

3 University of Fribourg, Departement of Geosciences, Unit of Geography, Fribourg, SWITZERLAND 



Risikomanagement, welches Unterhalts-, raumplanerische, organisatorische und bauliche 

Massnahmen beinhaltet. 

KEYWORDS 

debris flow; rock glacier;hazard assessment; integrated protection concept; emergency 

management 

EINLEITUNG 

In den letzten Jahrzehnten hat sich die Grunddisposition in vielen Alpinen Wildbacheinzugsgebieten 

aufgrund des Klimawandels und der damit verbundenen steigenden Temperaturen 

schleichend verändert. Bedingt durch das Auftauen von Permafrost in hochalpinem Gelände 

sind Siedlungsgebiete und Verkehrswege einer sich ständig verändernden neuen Gefahrenlage 

ausgesetzt, so auch im Mattertal (Delaloye et al., 2013) in den Schweizer Alpen (Abb. 1A). 

Lockermaterialeinträge aus destabilisierten Blockgletschern in den Einzugsgebieten steiler 

Wildbäche an der Westflanke der Mischabelgruppe führten seit 2010 vermehrt zu Murgangereignissen, 

welche in den letzten Jahren temporäre Evakuierungen von Anwohnern und 

Sperrungen von Verkehrswegen nötig machten (Graf et al., 2013). Um möglichst fundierte 

Informationen über die Veränderungen in solchen Gebieten sammeln zu können, diese zu 

analysieren und als Grundlage für zukünftige Entscheide bereitzustellen, sind verschiedene 

Forschergruppen mit Untersuchungen im Mattertal (VS) beschäftigt (Delaloye et al., 2013, 

Graf et al., 2013). 

Aufgrund der veränderten Gefahrensituation wurden 2013 im Auftrag der Gemeinden St. 

Niklaus und Randa und in enger Zusammenarbeit mit der kantonalen Dienststelle für Wald 

und Landschaft die sechs Wildbäche Grosse Grabe, Bielzug, Fallzug, Geisstriftbach, Birchbach 

und Dorfbach detailliert untersucht (Abb. 1B). Im Rahmen der Vorstudie Sicherheitskonzept 

Wildbäche Mattertal (arge wasser/schnee/lawinen, Geoplan AG, Bumann Reinhold GmbH, 

2013) wurden Szenarien für Murgangereignisse erarbeitet, diese mit dem Murgangmodell 

RAMMS::DEBRIS FLOW 1.5 (Graf et al., 2013, Christen et al., 2012) modelliert und eine 

überarbeitete Gefahrenkarte erstellt. Zur Zielsetzung gehörte die Beurteilung möglicher 

baulicher Schutzmassnahmen sowie temporärer Vorkehrungen zur Sicherung von Siedlungsgebieten, 

besonders schadenanfälligen Einzelobjekten (u.a. Stromleitungen) und Verkehrswegen 

(Strasse und Bahn). Das Vorgehen und die Methoden sind auf andere Wildbäche 

anwendbar. 

Basierend auf den Resultaten der oben beschriebenen Vorstudie wurden in der Zwischenzeit 

in einzelnen Bächen bereits detailliertere Projekte ausgearbeitet oder sind in Planung. Im 

Herbst 2014 wurde z. B. im Bielzug (Abb. 1B/C) ein Alarmsystem für die Siedlung und 

Verkehrsachsen in Betrieb genommen. Zudem werden im Bielzug ab Herbst 2015 bauliche 

Massnahmen innerhalb von ca. 2 Jahren umgesetzt (Abb. 1D). Ein Murgang im Fallzug 

verschüttete 2014 die Gleise der Matterhorn Gotthard Bahn (MGB) meterhoch und verur- 


sachte einen Schaden von ca. 400‘000 CHF. Der Fallzug hat bei der MGB eine hohe Dringlichkeit 

und der Vorschlag zur Ausarbeitung eines Wasserbauprojekts wurde 2014 an die 

zuständige kantonale Dienststelle weitergeleitet. In den restlichen Bächen sind zurzeit keine 

weiteren Massnahmen geplant. 

Abbildung 1: A: Lage Bielzug. B: Die sechs Einzugsgebiete der Vorstudie Sicherheitskonzept Wildbäche Mattertal. C: Übersicht 

Einzugsgebiet Bielzug. D: Geplante Schutzmassnahmen im Kegelbereich des Bielzugs. 


In der Folge werden die angewandten Methoden (Abb. 2) und Resultate der Vorstudie am 

Beispiel des Bielzugs (Abb. 1 C/D) erläutert. 

METHODEN 

Das Vorgehen und die dazugehörigen Referenzen sind in Abbildung 2 schematisch zusammengefasst. 

Um sich einen Überblick über die Aktivität in den Wildbächen zu verschaffen, 

wurde für jeden Bach ein Ereigniskataster zusammengestellt. Als Quellen dienten die 

Ereigniskataster des Kantons, der Gemeinden St. Niklaus und Randa, der Forschungsanstalt 

für Wald, Schnee und Landschaft (WSL) und der MGB. Alle Schutzbauten im Kegelbereich 

wurden aufgenommen und hinsichtlich ihres Zustandes gemäss der Methodik PROTECT 

(Romang et al., 2008) beurteilt. Um die massgebenden Szenarien für jeden Wildbach 

definieren zu können, wurde anschliessend die Charakteristik der Wildbäche bestimmt. 

Diese beinhaltet die Morphometrie, die Hochwasserabflüsse, die Karte der Phänomene und 

die Gefahrenpotentiale wie z. B. Gletscher, Blockgletscher, geologische Prozesse oder das 

Schwemmholzpotential. Den Unsicherheiten bei der Beurteilung und den Berechnungen 

wurde mit der Angabe von Bandbreiten Rechnung getragen, wie dies in der Ereignisanalyse 

zum Hochwasser 2005 (Bezzola und Hegg, 2008) empfohlen wird. Die Hochwasserabflüsse 

wurden mit HAKESCH (BAFU, 2003) bestimmt. In der Karte der Phänomene wurden die im 

Gelände sichtbaren Spuren und morphologischen Merkmale (stumme Zeugen) sowie 

Angaben zur Disposition von gefährlichen Prozessen erfasst und dargestellt. Damit liefert die 

Karte wichtige Hinweise bezüglich der Aktivität und Gefährlichkeit des untersuchten 

Gebietes. Die aktuelle Gefährdung durch Gletscher wurde in Zusammenarbeit mit der 

Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie VAW (ETHZ) beurteilt. Zudem 

dienten die Angaben aus dem „Inventar der gefährlichen Gletscher in der Schweiz“ als 

Grundlage (Raymond et al., 2003). Informationen über die jährliche Schuttproduktion, die 

Bewegungsgeschwindigkeiten, die Abbruchgefahr von Blockgletscherfronten, die Mächtigkeit 

etc. diverser Blockgletscher im Mattertal lieferten laufende Studien der Universität Fribourg 

(Delaloye et al., 2013). Für die Beurteilung der aktuellen geologischen Gefährdung durch 

Hanginstabilitäten (Rutschungen) aus Gebieten oberhalb der Waldgrenze dienten teilweise 

unveröffentlichte Interpretationskarten von InSAR-Daten (Satelliten-Daten) des Bundesamts 

für Umwelt (BAFU). Für die Abschätzung der potentiellen Materialmobilisierungen aus 

Hangzonen unterhalb der Permafrostzone durch Spontanrutschungen und Hangmuren 

wurde projektintern eine Methodik entwickelt. Aufgrund von Erfahrungswerten wurde 

davon ausgegangen, dass lediglich jene Hangbereiche Material in das Hauptgerinne liefern 

können, welche in einer Pufferzone von 150 m um das Hauptgerinne liegen. In dieser 

Pufferzone wurden die Mächtigkeit der Lockermaterialbedeckung aufgrund der Hangneigung 

für alle Flächen ohne Felsen abgeschätzt, welche eine Neigung > 20° aufweisen oder nicht 

stark bewaldet sind. Die daraus erhaltene Karte mit Angaben zu Vorkommen und Mächtigkeit 

des Lockergesteins diente als Grundlage zur Abschätzung der potentiellen Materialmobilisierung 

aus den Hangzonen. Die Abschätzung der Geschiebemengen wurde mit der 

Methodik SEDEX (Frick et al., 2011) durchgeführt. Die Geschiebefracht für ein 100-jährliches 


Ereignis (GF100) wurde zudem gemäss dem Flussdiagramm nach Spreafico et al. (1996) 

abgeschätzt und das Ergebnis mit den Resultaten nach SEDEX verglichen und beurteilt. 

Ausserdem wurde auch das mobilisierbare Schwemmholzpotential abgeschätzt. 

Ereigniskataster 

-Kanton 

-Gemeinde 

-WSL 

-MGB 

Schutzbauten 

gemäss Methodik PROTECT - Wirkung 

von technischen Schutzmassnahmen, 

Teil E, Romang et al. 2008 

Charakteristik Wildbäche 

-Morphometrie 

-Längenprofil 

-Hochwasserabflüsse 

(gemäss HAKESCH, BWG 2003) 

-Karte der Phänomene 

(gemäss Empfehlung „Symbolbaukasten 

...zur Kartierung der Phänomene“, BWG et 

...al. 1995) 

Gefahrenpotentiale 

-Gletscher (Inventar der gefährlichen Gletscher in 

...der Schweiz, VAW, ETHZ) 

-Blockgletscher (Uni Fribourg) 

-Hanginstabilitäten(InSAR-Daten, BAFU, 

...unveröffentlicht) 

-Schwemmholz (gemäss „Schwemmholz und 

...Hochwasser“, Rickenmann 1997) 

-Geschiebemengen (SEDEX, Frick et al. 2011) 

Definierung Szenarien 

-Auslösebedingungen 

-Mobilisierung 

-Materialeintrag 

-Ereigniskubaturen 

-Ereignisabläufe 

-Murgangtyp 

-Prozesslimitierende Faktoren 

-zukünftige Entwicklung 

Simulationen 

-mit RAMMS::DEBRIS FLOW 1.5 

-anschliessende Kontrolle im Feld 

Erstellung Intensitätskarten (Ik) 

IK’s für ein 30-, 100- und 300- jährliches sowie ein extremes 

Ereignis (EHQ) unter Berücksichtigung der bestehenden 

Schutzbauten 

gemäss Empfehlung „Berücksichtigung der Hochwassergefahren bei 

raumwirksamen Tätigkeiten“, BWW et al.1997 

Erstellung Gefahrenkarte (Gfk) 

gemäss Empfehlung „Berücksichtigung der Hochwassergefahren 

bei raumwirksamen Tätigkeiten“, BWW et al.1997 

Definierung Schutzziele 

Grundlage Schutzzielmatrix gemäss Empfehlung „Raumplanung 

und Naturgefahren“, ARE et al. 2005 

Keine Massnahmen 

ja 

Schutzziel 

erreicht? 

nein 

Massnahmen: 

-Unterhalt 

-raumplanerisch 

-organisatorisch 

-baulich 

Abbildung 2: Angewandte Methoden 

Basierend auf diesen Grundlagen wurden die massgebenden Gefahrenszenarien für jeden 

Wildbach bestimmt. Für die Szenarien wurden die Ereigniskubatur, die Auslösebedingungen, 

der Materialeintrag, die Mobilisierung, die Ereignisabläufe (Anzahl Schübe, Ablagerungen im 

Gerinne), der Murgangtyp (granular oder flüssig), die prozesslimitierenden Faktoren (Wasser, 

Geschiebe) und ein qualitativer Beschrieb der zukünftigen Entwicklung vom Geschiebepoten- 


tial im Einzugsgebiet definiert. Anhand der Murgangszenarien und mit Hilfe des Murgangmodells 

RAMMS::DEBRIS FLOW 1.5 (Christen et al., 2012) wurden für alle Bäche Simulationen 

durchgeführt. Das Modell wurde mit Hilfe von kartierten Ereignissen und den Erkenntnissen 

und Erfahrungen aus Graf et al. (2013) kalibriert. Hierbei wurden die bestehenden Schutzmassnahmen 

berücksichtigt. Schwachstellen entlang des Gerinnes wurden im Feld aufgenommen, 

mit den Ausbruchstellen der RAMMS-Modellierungen verglichen und deren 

Relevanz für Gerinneausbrüche in Abhängigkeit der Wiederkehrperiode beurteilt. Die 

Erstellung der Intensitätskarten Hochwasser und Murgang erfolgte anhand der Modellierungsresultate 

sowie einer Plausibilisierung im Feld. Aus den erarbeiteten Intensitätskarten 

wurden schliesslich die Gefahrenkarten erstellt. 

Die Schutzziele wurden projektspezifisch festgelegt. Als Grundlage diente die Schutzzielmatrix 

gemäss der Bundesempfehlung „Raumplanung und Naturgefahren“ (ARE et al., 2005). 

Bauzonen (Dorfbereiche), Freizeitanlagen und Campingplätze sollen einen vollständigen 

Schutz bis zu einem 100-jährlichen Murgang- oder Hochwasserereignis (M 100 

/HQ 100 

) 

aufweisen. Die Kantonsstrasse, Einzelgebäude, Infrastrukturanlagen sowie Hochspannungsleitungen 

sollen bei einem 30-jährlichen Ereignis (M 30 

/HQ 30 

) einen vollständigen Schutz 

aufweisen. Durch die Überlagerung der Intensitätskarten mit der Schutzzielkarte wurden 

Gebiete mit einem Schutzdefizit ermittelt. 

Aufgrund der durchgeführten Untersuchungen in den Einzugsgebieten und den vorhandenen 

Schutzdefiziten resultierte ein Massnahmenkatalog zur Verbesserung des Schutzes. 

Unterhalts-, raumplanerische, organisatorische und bauliche Massnahmen wurden dabei 

berücksichtigt. 

RESULTATE 

Im folgenden Kapitel werden am Beispiel des Bielzugs die Resultate der Untersuchung 

zusammengefasst und dargestellt. Gemäss dem Ereigniskataster ist im Bielzug durchschnittlich 

alle 7 Jahre mit Murgängen zu rechnen, welche einen Teil des Geschiebes im Gerinne 

mobilisieren. Die jüngsten Ereignisse haben gezeigt, dass sich im Bielzug Murgänge auch 

ohne Niederschläge bilden können. Aufgrund der überdurchschnittlichen Schneeschmelze 

und der gleichzeitig hohen Aktivität des Blockgletschers Gugla ereignete sich im Juni 2013 

eine mehrtägige Murgangserie. Die Blockgletschergeschwindigkeit hat sich rezent sehr stark 

beschleunigt und lag 2013 zwischen 4 und 17 m/Jahr in dem 100 m breiten und 100 - 150 m 

langen, jetzt gespaltenen Stirnbereich (Tab. 1). Durch die aktuell (2013-2014) hohe jährliche 

Schuttproduktion des Blockgletschers, die rund 8‘000 m 3 /Jahr erreichte (Delaloye et al., 

2013, Kummert and Delaloye, 2015), kann sich Geschiebe rasch wieder in Zwischendepots 

ansammeln (Tab. 1). 


1 

Tabelle 1: Entwicklung der mittleren Geschwindigkeit der Stirn [m/Jahr] und der Schuttproduktion des Blockgletschers Gugla [m 3 /Jahr]. 

Die Geschwindigkeiten wurden bis 2009 mittels Luftbildanalysen bestimmt. Seit 2007 werden GPS-Messungen durchgeführt. Die 

Schuttproduktion ist das abgeschätzte jährliche Volumen von Lockermaterial, das durch Blockgletscherbewegung und Eisschmelze an 

der Blockgletscherfront frei geworden ist. 

2 

1 

Mittlere 

Geschwindigkeit 

Blockgletscherstirn 

[m/Jahr] 

Schuttproduktion 

[m 3 /Jahr] 

1968- 

1982 

Die Untersuchungen im Bielzug ergaben die in Tabelle 2 aufgeführten, massgebenden 

Murgangszenarien. Grössere Abbrüche von Geschiebepaketen aus dem Blockgletscher 

wurden im EHQ-Szenario (>300-jährlich) berücksichtigt. Die Plausibilisierung der Geschiebemengen 

des Bielzugs hat gezeigt, dass die Feststofffracht für ein 100-jährliches Ereignis 

(GF100), welche gemäss Spreafico et al. (1996) abgeschätzt wurde (38‘000 m 3 ), unter der 

Bandbreite der mit SEDEX bestimmten Ereigniskubatur liegt (Tab. 2). Massgebend für diesen 

Unterschied sind die Art der Materialeinträge durch den Blockgletscher Gugla, welche bei 

Spreafico et al. (1996) nicht berücksichtigt werden. Die abgeschätzten Kubaturen werden 

daher als plausibel eingestuft. 

1982- 

1995 

1995- 

2005 

2005- 

2009 

2009- 

2013 

0.35 0.5 1.4 3.0 6.0 10.8 

250 400 1‘100 2‘300 4‘700 8‘500 

2013 Tendenz (2014-2033) Tendenz 

(nach 

2033) 

Möglicherweise zuerst 

ansteigend, dann 

konstant bis abnehmend 

Tabelle 2: Abfluss, massgebende/maximale Geschiebemengen/Murgangvolumina und grösster erwarteter Murgangschub für die 

untersuchten Szenarien. 

Szenario 

Hochwasserabfluss 

(Reinwasser) 

max. Geschiebefracht / 

Murgangvolumen 

2 

Grösster 

Murgangschub 

30-jährlich 9-12 m 3 /s 10‘000-20‘000 m 3 10‘000 m 3 

100-jährlich 13-18 m 3 /s 40‘000-60‘000 m 3 20‘000 m 3 

300-jährlich 19-26 m 3 /s 60‘000-80‘000 m 3 30‘000 m 3 

> 300-jährlich 26-36 m 3 /s 80‘000-120’000 m 3 90‘000 m 3 

abnehmend 

Die in Tabelle 2 aufgelisteten Szenarien dienten als Grundlage für die Simulationen mit 

RAMMS::DEBRIS FLOW 1.5 (Christen et al., 2012). Basierend auf den Erfahrungen im 

Dorfbach (Graf et al. 2013) und anhand von Ereignissen wurde das Modell numerisch 

kalibriert. Im Rahmen der Vorstudie wurden die repräsentativsten Simulationen pro 

Wiederkehrperiode und die entsprechenden Fliesswege, Fliesshöhen und Fliessgeschwindigkeiten 

dargestellt. Demnach ist ab einem 30-jährlichen Ereignis mit Gerinneausbrüchen 

unterhalb der bestehenden Schutzmauer beim Kegelhals zu rechnen, falls sich dort Geschiebe 

zwischenlagert. Aufgrund des zu geringen Retentionsvolumens des Geschiebesammlers muss 


auch hier mit einem Überströmen gerechnet werden. Ab dem 100-jährlichen Ereignis muss 

zudem mit Ausbrüchen am Kegelhals gerechnet werden. Die Resultate wurden im Rahmen 

einer Feldbegehung gutachterlich verifiziert. Basierend auf den beiden Beurteilungen wurde 

für jede Wiederkehrperiode eine Intensitätskarte erstellt. Gemäss Empfehlung (BWW et al., 

1997) werden für deren Erstellung die Ablagerungshöhe und die Fliessgeschwindigkeit 

berücksichtigt. In Absprache mit dem BAFU wurde beschlossen, bei der Intensitätsklassierung 

nicht die Ablagerungshöhe zu berücksichtigen, sondern die maximale Fliesshöhe, da diese 

massgebend für die Gefährdung ist. 

Die überarbeitete Gefahrenkarte zeigt nur eine geringe Abweichung gegenüber der früheren 

Version. Einzig die Flächenanteile der verschiedenen Gefahrenstufen haben sich leicht 

geändert. So ist das nördliche Siedlungsgebiet von Herbriggen stärker durch Murgänge 

betroffen als bisher angenommen. Dies ist vor allem auf die Berücksichtigung der Entwicklung 

des Blockgletschers Gugla zurückzuführen. 

1 

Tabelle 3: Resultierende Schutzdefizite (gekennzeichnet mit ‚ x ‘) und die erreichten Schutzziele (gekennzeichnet mit ‚ √ ‘) aufgrund der 

Intensitätskarten vor Massnahmen basierend auf der Bundesempfehlung „Raumplanung und Naturgefahren“ (ARE, BWG, BUWAL, 

2005). 

Kategorie 

30-jährliches 

Szenario 


Szenario 


Szenario 

EHQ 

(>300-jährlich) 

Verkehrswege und Leitungen (Kat 2.3) 

Kantonsstrasse ! ! ! " 

Bahnlinie ! ! ! " 

Hochspannungsleitung a " ! ! " 

Hochspannungsleitung b ! ! ! " 

Siedlungsgebiet (Kat. 3.2) 

2 

Herbriggen " ! ! ! 

Basierend auf den Intensitätskarten und der Schutzzielmatrix aus der Bundesempfehlung 

„Raumplanung und Naturgefahren“ (ARE et al., 2005) wurden die Schutzdefizite des 

Bielzugs bestimmt (Tab. 3). Um die Schutzdefizite beheben zu können, sind Massnahmen 

notwendig (Abb. 1D). Aus Platzgründen können im Bielzug bauliche Massnahmen nur für 

30- bis 100-jährliche Ereignisse dimensioniert werden. Sie umfassen eine Erhöhung der 

bestehenden Leitmauer am Kegelhals, einen zusätzlichen Ablenkdamm, eine Vergrösserung 

des bestehenden Geschiebesammlers und eine Verschalung des MGB-Durchlasses (vgl. Abb. 

1D). Diese Massnahmen alleine genügend jedoch nicht zum Schutz vor seltenen und sehr 

seltenen Ereignissen und schützen insbesondere die Verkehrswege ungenügend. Der 

Überlastfall wird daher mit organisatorischen Massnahmen abgedeckt. Für die Reduktion des 

verbleibenden Risikos wurde eine Warnanlage für Bahn und Strasse erstellt (Abb. 1D). Mit 

einer Kombination verschiedener Sensoren (Geophone, Murgangkombisensor, Pegelsonde) 

werden mittlere und grosse Ereignisse detektiert und die Verkehrswege werden im Ereignis- 


fall mittels Lichtsignalen automatisch gesperrt. Das Risikomanagement der Gemeinde St. 

Niklaus beinhaltet eine Notfallplanung, in welcher einerseits Beobachtungspunkte im 

Einzugsgebiet für eine verbesserte Frühwarnung und andererseits Interventionsmassnahmen 

für rasches und effizientes Handeln festgelegt worden sind. 


Die Vorstudie hat aufgezeigt, dass sich das Geschiebeangebot aufgrund der Aktivität eines 

Blockgletschers massgebend verändern kann. Die Gefahrenkarte in solchen Wildbächen sollte 

daher regelmässig überprüft und wenn nötig nachgeführt werden. Dies zeigt auch, dass eine 

Gefahrenkarte grundsätzlich nicht als statisch angesehen werden darf. Die Kombination von 

numerischer Modellierungen und Plausibilisierung der Resultate durch Feldbegehungen hat 

sich im Rahmen der Vorstudie Mattertal sehr gut bewährt. Dies stellt im vorliegenden Fall 

auch ein Bindeglied zwischen wissenschaftlicher Grundlagenforschung und der Praxis dar. 

Oft bieten sich aufgrund der örtlichen Verhältnisse bei Wildbächen primär bauliche Massnahmen 

an. Das Beispiel Bielzug hat jedoch aufgezeigt, dass bauliche Massnahmen zum Schutz 

von Siedlungen und Verkehrswegen nicht immer ausreichen. Es braucht daher ein Risikomanagement, 

welches Unterhalts-, raumplanerische, organisatorische und bauliche Massnahmen 

beinhaltet. 

REFERENZEN 

- arge wasser/schnee/lawinen, Geoplan AG, Bumann Reinhold GmbH (2013). Vorstudie 

Sicherheitskonzept Wildbäche im Matteral, Brig-Glis, Steg, Naters, Dezember 2013. 

- Bezzola G.R., Hegg C. (eds.) (2008). Ereignisanalyse Hochwasser 2005, Teil 2 - Analyse von 

Prozessen, Massnahmen und Gefahrengrundlagen. Bundesamt für Umwelt BAFU, Eidgenössische 

Forschungsanstalt WSL. Umwelt-Wissen Nr. 0825: 429 S. 

- Bundesamt für Umwelt, Wald und Landschaft BUWAL, Bundesamt für Wasser und Geologie 

BWG (1995). Symbolbaukasten zur Kartierung der Phänomene, 41 S. 

- Bundesamt für Raumentwicklung ARE, Bundesamt für Wasser und Geologie BWG, 

Bundesamt für Umwelt, Wald und Landschaft BUWAL (2005). Empfehlung - Raumplanung 

und Naturgefahren. 48 S. 

- Bundesamt für Wasser und Geologie BWG (2003). Hochwasserabschätzung in schweizerischen 

Einzugsgebieten, Berichte des BWG, Serie Wasser Nr. 4, Bern 2003, 118 S. 

- Bundesamt für Wasserwirtschaft BWW, Bundesamt für Raumplanung BRP, Bundesamt für 

Umwelt, Wald und Landschaft BUWAL (1997). Berücksichtigung der Hochwassergefahren bei 

raumwirksamen Tätigkeiten, Empfehlung Naturgefahren, Biel, 1997, 32 S. 

- Christen, M., Bühler, Y., Bartelt, P., Leine, R., Glover, J., Schweizer, A., Graf, C., McArdell, 

B.W., Gerber, W., Deubelbeiss, Y., Feistl, T. & Volkwein, A. (2012). Integral hazard management 

using a unified software environment: numerical simulation tool "RAMMS" for 

gravitational natural hazards. In: Koboltschnig, G.; Hübl, J.; Braun, J. (eds.) 12th Congress 

INTERPRAEVENT, 23-26 April 2012 Grenoble - France. Proceedings. Vol. 1. Klagenfurt, 

International Research Society INTERPRAEVENT. 77-86. 


- Delaloye, R., Morard, S., Barboux, C., Abbet, D., Gruber, V., Riedo, M., & Gachet, S., (2013). 

Rapidly moving rock glaciers in Mattertal. In: Graf, C.(Red.) Mattertal - ein Tal in Bewegung. 

Publikation zur Jahrestagung der Schweizerischen Geomorphologischen Gesellschaft, 9. Juni 

bis 1. Juli 2011, St. Niklaus. Birmensdorf, Eidg. Forschungsanstalt WSL. 21-31. 

- Frick, E., Kienholz, H., Romang, H. (2011). SEDEX. Anwenderhandbuch. Geographica 

Bernensia, P 42 128 S. 

- Graf, C., Deubelbeiss, Y., Bühler, Y., Meier, L., McArdell, B., Christen, M., Bartelt, P. (2013). 

Gefahrenkartierung Mattertal: Grundlagenbeschaffung und numerische Modellierung von 

Murgängen. In: Graf, C. (Red.) Mattertal - ein Tal in Bewegung. Publikation zur Jahrestagung 

der Schweizerischen Geomorphologischen Gesellschaft 29. Juni - 1. Juli 2011, St. Niklaus. 

Birmensdorf, Eidg. Forschungsanstalt WSL. 85-112. 

- Kummert, M. & Delaloye, R. (2015). Quantifying sediment transfer between the front of an 

active alpine rock glacier and a torrential gully. In: Jasiewicz J., Zwolinski Zb., Mitasova H., 

Hengl T. (eds.) Geomorphometry for Geosciences. Adam Mickiewicz University in Poznan - 

Institute of Geoecology and Geoinformation, International Society for Geomorphometry, 

Poznan, 2015. 193 - 196. 

- Raymond, M., Wegmann, M. und Funk, M. (2003). Inventar der gefährlichen Gletscher in 

der Schweiz. Mitteilungen Nr. 182 der Versuchsanstalt für Wasserbau, Hydrologie und 

Glaziologie, der ETH Zürich. Herausgeber: Prof. Dr.-Ing. H.-E. Minor. 

- Rickenmann, D. (1997). Empirical Relationships for Debris Flows. Natural Hazards. 47 - 77. 

- Romang, H. (Ed.) (2008). Wirkung von Schutzmassnahmen. Nationale Plattform für 

Naturgefahren PLANAT, Bern. 289 S. 

- Spreafico, M., Lehmann, CH., Naef, O. (1996). Empfehlung zur Abschätzung von Feststofffrachten 

in Wildbächen, Teil 1, Handbuch, Mitteilung Nr. 4, GHO, Bern. 



Integrated natural hazards protection concept Vitznau 

LU - Case study Plattenbach 

Benjamin Hohermuth, dipl. Umwelting. 2 ; Christoph Graf, dipl. Geogr. 1 ; Jörn Heilig, dipl. Ing. (TU) 3 

ABSTRACT 

Vitznau (LU) is located at the foot of the south flank of the pre-alpine Rigi mountain. Eight 

debris-flow and flood prone torrents run through the village into the Lake Lucerne. The work 

presented herein focuses on the Plattenbach. Within the revision of the integrated protection 

concept numerical simulations for debris flow and floods were incorporated into the planning 

and design of protection measures. This case study illustrates this approach which is still not 

standard practice in Switzerland. The effectiveness of debris-flow mitigation structures was 

successfully evaluated using the new No-Flux feature of the RApid Mass Movement 

Simulation (RAMMS). The original layout could be optimized in a hazard protection and 

economical aspect. Flood events with less sediment mobilization were simulated using the 

hydro-numerical model BASEMENT. 

The final design of the retention dam provides protection against debris flows with up to 

300-year return period. Additional measures on the fan allow for the conveyance of up to a 

100-year flood. In case of extreme events a robust system behaviour is expected. 

KEYWORDS 

integrated hazard management; protection concept; debris flow; RAMMS; BASEMENT 

INTRODUCTION 

Present-day numerical simulation tools allow for in-depth evaluations of hazard scenarios. 

Their use lately gained increasing importance in hazard and risk assessment. While the 

number of simulation tools has enlarged, the need for improvement of their practical 

application still exists. Especially the often extensive calibration of model parameters and the 

interpretation and further use of simulation results remains challenging. This paper presents a 

case study of an integrated natural hazard protection concept in which numerical simulations 

were used in an early project stage to evaluate existing measures and assist the design of new 

structures. 

The village of Vitznau (LU) is located at the bottom of the pre-alpine Rigi mountain in Central 

Switzerland. Eight steep mountain torrents run through the village. All torrents are prone to 

debris flows. Reworked hazards maps revealed protection deficits, showing weak points along 

the torrents. Based on these findings a risk-based prioritization by the Canton of Lucerne 

1 WSL Swiss Federal Research Institute, Birmensdorf, SWITZERLAND, christoph.graf@wsl.ch 

2 ETH Zurich Laboratory of Hydraulics, Hydrology and Glaciology, Zürich, SWITZERLAND 

3 HOLINGER AG, Liestal, SWITZERLAND 



educed the project to Altdorfbach, Kalibach, Widibach and Plattenbach. This paper focuses 

on the Plattenbach. The project perimeter also includes the adjoining Mühlebach (Fig. 1) for 

flood processes. 

Figure 1: Overview of Plattenbach catchment area (red line); landslide areas marked in red. 

The Plattenbach drains part of the steep southern flank of the Rigi. The bedded psephites on 

the southern flank are covered by detritus of variable thickness. Three instable zones 

“Stollenegg”, “Brand” and “Glinge” (Fig. 1) are prone to landslide activity. Main and most 

active slide area is “Stollenegg”. In case of long-lasting precipitation an acceleration of the 

sliding process is expected. The sediments transform to channelized debris-flows, estimated 

surge volume and maximum discharge are shown in Tab. 1. The channel runs predominantly 

on bedrock, is partially deeply incised, and passes through several narrow points. After 

crossing the railway embankment of the Rigibahn the Plattenbach runs for about 120 m in a 

culvert before the confluence with the Mühlebach. The last bottleneck is the culvert under 

the main road shortly before it flows into the Lake of Lucerne (Fig. 1). 

A total of three culverts on the fan and natural narrow points in the transit section limit the 

conveyance of floods and debris-flow surges. Debris or driftwood clogging at these narrow 

points poses a large damage potential for the today densely populated fan. As observed in the 


Table 1: Flood and debris-flow data for the Plattenbach. Values in brackets are Mühlebach flood discharges and bed-loads. Data from 

Holinger AG & NDR Consult (2013) 

Return period 30-year 100-year 300-year Extreme 

Flood event 

Discharge [m 3 /s] 12 (+10) 20 (+19) 32 (+26) 42 (+34) 

Bed-load volume [m 3 ] 10 (+1'000) 200 (+2'000) 800 (+7'000) 4000 (+20'000) 

Debris flow 

Total volume [m 3 ] 5'040 16'700 31'300 79'000 

Surges [-] 2 3 4 3 

Surge volume [m 3 ] 2'520 5'567 7'825 26'333 

Peak Discharge [m 3 /s] 70 130 170 420 

2005 event discharge form the Mühlebach and backwater effects from Lake Lucerne need to 

be considered as well. 

Only a few historical events are known for the Plattenbach. Two of them are documented. 

The one from 1910 is the largest recorded event. In 2005 a similar sequence of events with 

lower debris volumes and minor consequences took place. In both cases intense and 

long-lasting rainfall with a 100-year return period caused slope destabilization in the 

“Stollenegg” area. In 1910, about 15’000 m 3 of the total landslide mass of 30 – 40’000 m 3 

formed a debris flow and deposited on the western part of the fan. No debris or log jam at the 

already existing railroad embankment was reported. Information on flow depth and flow 

velocity is missing. At this time the western part of the fan was almost not inhabited expect of 

the old school building that was damaged and the protestant church that was eventually 

touched but not damaged because of elevated and distal location. In August 2005 a volume of 

500-1000 m 3 debris material was destabilized in the “Stollenegg” area. Toe erosion led to 

fluvial sediment transport and deposition in the culvert under the main road and at the outlet 

to the lake. 

METHODS 

In this project numerical models were used to assess different hazard processes in the initial 

state. In a second step the same numerical models were employed to evaluate the effectiveness 

of different natural hazard protection measures. The models were also used to assist the 

design of the protection measures. 

NUMERICAL MODELS 

The numerical models used in this work are the RApid Mass Movement Simulation 

(RAMMS) for debris-flow computations and BASEMENT for hydraulic flow and sediment 

transport simulations. RAMMS is based on the 2D shallow water equations and allows the 

computation of debris-flow runout on complex terrain (Christen et al. 2012). Debris flows are 

modelled as one single phase and the well-known rheological friction law of Voellmy-Salm is 

employed. The two empirical friction parameters were calibrated with the documented 

debris-flow event in 1910 (Hohermuth 2014). To account for different debris-flow mixtures 


the evaluation of protection measures was performed with a parameter range rather than 

single values (Hohermuth & Graf 2014). Digital Elevation Model (DEM) quality and resolution 

characterize the natural terrain and are therefore the key input parameters. The DEM used 

for the simulations is based on LIDAR data collected within the project and was verified in 

the field (Hohermuth 2014). The simulations in RAMMS were performed on a grid with 

1x1m spatial resolution. 

RAMMS 1.6.20 features so-called No-Flux cells, which allow to define impermeable flow 

areas such as buildings and retention dams. This feature was used for the first time in the 

work presented herein to evaluate protection measures i.e. deflection walls in a densely 

populated area. Fig. 2 shows the comparison of two different deflection wall designs. The 

simulations also allow to estimate the impact pressure and thus help with adequate design. 

Figure 2: Comparison of debris-flow runout with and without deflection walls. Simulation results from RAMMS 1.6.20 with No-Flux cells. 

Deflection walls were dismissed in a later project stage. 

BASEMENT is a simulation environment developed at the Laboratory of Hydraulics, Hydrology 

and Glaciology (Vetsch et al. 2013). For the computations within this project the 

2D module BASEPlane V2.4 was used. Bed load transport can be modelled with various 

empirical equations. Gravitational transport is included with a simple geometric approach 

based on critical slope angles. There are no flood marks available for calibration and the 

hydrology of the 2005 flood event is only poorly documented. However, a rough calibration 

was performed to match the general sediment deposition behavior observed in 2005. 

Despite lacking a thorough calibration the simulations allow for a relative comparison of 

different flood protection measures. 

OPTIMIZATION OF EXISTING DEBRIS-FLOW PROTECTION CONCEPT 

Preliminary studies (Holinger AG & NDR Consult 2013) propose the following set of 

measures for debris-flow protection: 

– Small check dams and slope drainage in the upper catchment “Stollenegg” (realized 

as emergency measure in 2012) 

– Sediment trap with V = 10’600 m 3 downstream of the “Brand” slope instability 

– Reinforcement of railway embankment to allow for sediment retention, V = 2’600 m 3 


– Emergency corridor for events with more than a 100-year return period 

The proposed measures were tested in RAMMS. The analysis showed that the original 

sediment retention volume is insufficient. In a 300-year debris flow –despite other measures 

– the remaining intensity would be above the acceptable limit. Based on simulation results 

the retention capacity of the check dam “Brand” was tripled to 33’000 m 3 , what allows for 

a complete retention of sediments for the “Brand” and “Stollenegg” slope instability up to a 

300-year return period. 

A sensitivity analysis of the hydrograph shape and maximum peak discharge showed that 

the capacity of the channel in the transitional zone is sufficient even for extreme events. 

This superseded measures to increase the channel capacity which were considered in an 

earlier project stage. 

The structural condition of the railway embankment makes a reinforcement to withstand 

debris-flow impact expensive. Backed up by numerical simulations it was concluded that 

debris flows originating from the “Glinge” slope instability can mostly be conveyed by the 

Plattenbach channel and do not lead to unacceptable intensities. Therefore measures in the 

transitional zone and sediment retention at the embankment were rejected. However to 

avoid clogging of the culvert a drift wood rack is intended upstream of the railway line. 

The optimized concept consists of: 

– Sediment trap with V = 33’000 m 3 downstream of the “Brand” slope instability 

– Driftwood rack at the embankment culvert (no sediment retention) 

– Emergency corridor for extreme events (R > 300-years) 

In contrast to the original concept, the optimized concept provides protection up to a 

300-year debris flow. 

EVALUATION OF DIFFERENT FLOOD PROTECTION CONCEPTS 

For regular flood events with moderate sediment mobilization additional measures are 

needed. The capacity of the culvert under the main road is 30 m 3 /s for clear water flow. 

This is insufficient in a 100-year event (concurrence of flood events in both Plattenbach and 

Mühlebach). After the hydraulic jump at the confluence of Platten- and Mühlebach sediment 

deposition occurs. Additional deposition takes place in the “Seestrasse” culvert after the 

confluence due to the smaller channel slope. This could be observed during the flood event in 

2005. The four different concepts shown in Fig. 3 were evaluated with the help of 2D 

simulations in BASEMENT. All concepts feature lateral walls (red in Fig. 3) and additional 

measures as follows: 

– V1: Capacity increase: The conveyance of the culvert under the main road is increased. The 

culvert width is increased from 2.2 m to 7.5 m, this creates enough capacity even with 

large sediment deposits. Capacity increase at the schoolhouse bridge. 


– V2: Side weir and diversion tunnel Mühlebach: Up to 10 m 3 /s are diverted from the 

Mühlebach during a flood event. 

– V3: Weir and diversion tunnel Plattenbach: A diversion weir inside the existing culvert 

“Zihlstrasse” diverts up to 15 m 3 /s. Capacity increase at the schoolhouse bridge. 

– V4: Flood depression overflow corridor: An flood corridor is created by an abatement of the 

main street. Additional object protection measures guide the water through the small park 

into the lake. Capacity increase at the schoolhouse bridge 

The concepts were assessed based on the criteria hydraulics, performance in extreme events, 

ecological impact and feasibility and costs. The large land acquisition is the biggest downside 

of concept V1. Numerical simulations and sediment transport calculations showed that in V2 

the maximum diverted discharge is limited to 10 m 3 /s to avoid sediment deposition after the 

diversion. Thus an increase of the discharge capacity of the “Seestrasse” culvert is still needed 

what makes this the most expensive concept. The approach flow of the side weir in V3 is 

supercritical. In general, side weirs are not recommended for approach flow Froude numbers 

Figure 3: Illustration of four different protection concepts (versions V1 - V4). 

F o 

> 0.75 due to their poor discharge characteristics (Bühlmann & Boes 2014). Sediment 

transport and the location inside the culvert further complicate the situation. Hydraulic model 

test would be required and the excess capacity during an extreme event is limited. Additionally 

concept V3 features high construction costs. The flood corridor in V4 will be in operation 

every 50-100 years. Discharges exceeding the capacity of the “Seestrasse” culvert are routed 


through the flood corridor into the lake. The low construction costs and the robust behavior 

during extreme events make V4 the best option. 

FINAL DESIGN 

Measures against debris flow 

The main requirements of the check dam “Brand” are retention, filtration and dehydration of 

sediments in case of debris flows and large floods. Due to spatial restrictions a conventional 

20 m high concrete structure was chosen over a series of net barriers. 

The retention volume is 33'000 m 3 for a sediment deposition slope of 7%. A 10 m wide 

overflow section is included in the middle of the 74 m long dam crest. To facilitate filtration 

and dehydration of debris a slot will be set up in the middle of the dam. Smaller flood events 

are conveyed by a gully at the bottom of the slot. An upstream driftwood / debris rack 

prevents early clogging of the slot during small events. The construction of a 700 m long new 

forest road allows access for construction and maintenance. 

To account for potential driftwood from the lower catchment a new driftwood rack upstream 

of the railway culvert is planned. A total of 15 bars with a max. height of 2 m and a bar 

spacing of 0.8 m are planned. The right turn of the Plattenbach just upstream of the culvert 

allows for almost parallel approach flow. Thus backwater effects at the rack can be minimized. 

Flood protection measures 

The final design consists of lateral walls along the Plattenbach and the Mühlebach (max. 

height 1.5 m) as well as of object protection measures. The abatement of the main road to 

generate a flood corridor illustrated in Fig.4 exhibits good synergies with a simultaneous 

project to slow down transit traffic. 

The capacity of the schoolhouse bridge across the Mühlebach has to be increased by the 

relocation of a small step and a local abatement of the river bed. 

Management of the overload event 

Although the measures provide protection up to a 300-year debris-flow event or a 100-year 

flood respectively, the overload scenario (extreme event) has to be evaluated. Numerical 

simulations show comparable flow depth and extent with and without the sediment 

retention “Brand” for extreme debris flows. It is assumed that the check dam is completely 

filled after the second surge and has no effect on third and last surge. Even the failure of the 

completely filled check dam does not lead to significantly higher intensities, because the 

intensities are in both cases mainly caused by the failure of the railway embankment. It can 

be concluded, that the measures do not aggravate the situation in an extreme debris-flow 

event. The project suggests the creation of an “overload zone” in which special regulations 

(building regulations, evacuation plans) apply. 


Figure 4: Schematic illustration of the flood corridor. Diversion of discharge exceeding the culvert capacity into Lake Lucerne. Deflection 

walls drawn in red (Picture Holinger 2015). 

For overload flood scenarios the capacity of the flood corridor is exceeded and low flood 

intensities occur along the main road. The flood corridor still has an attenuating effect during 

extreme events. 

Cost Benefit Analysis 

The total construction cost of all sediment and flood protection measures was estimated to 

12.8 Mio. CHF whereas the risk reduction per annum was (based on EconoMe 2.0) approximated 

to 740’000 CHF/a (Plattenbach) and 232’000 CHF/a (Mühlebach). Thus a benefit-cost 

ratio from 2.9 (Plattenbach) and 6.0 (Mühlebach) is achieved with a design lifespan of 100 

years and an interest rate of 2% (Holinger AG 2015). 

CONCLUSIONS 

Numerical debris-flow simulations in RAMMS were successfully employed to evaluate and 

optimize an existing protection concept. The tool allows to assess the effectiveness of 

protection measures. The optimizations allow for a higher protection level (up to 300-year 

return period) with the same cost-benefit ratio as the initial concept. The simulations help to 

test the robustness and resilience of the measures in an extreme event even though some 

limitations apply (Hohermuth & Graf 2014). 


BASEMENT was used to investigate bed level changes during flood events. The simulations 

were used to assist the design of a flood corridor and additional measures which form a 

robust system that can handle floods with a broad range of sediment volumes. 

REFERENCES 

- Bühlmann, M., Boes, R.M. (2014). Lateral flood discharge at rivers: Concepts and challenges. 

Proc. Intl. River Flow Conference (Schleiss, A.J., De Cesare, G., Franca, M.J., Pfister, M., 

eds.), ISBN 978-1-138-02674-2, Taylor & Francis Group, London, UK: 1799-1806. 

- Christen M., Bühler Y., Bartelt P., Leine R., Glover J., Schweizer A., Graf C., McArdell B.W., 

Gerber W., Deubelbeiss Y., Feistl T., Volkwein A. (2012). Integral Hazard Management Using a 

Unified Software Environment-Numerical Simulation Tool „RAMMS“ for Gravitational 

Natural Hazards. 12th Congress Interpraevent 2012. 

- Hohermuth B. (2014). Integrales Schutzkonzept Plattenbach Vitznau – Murgangsimulationen 

mit RAMMS (Integrated Natural Hazard Protection Concept Plattenbach Vitznau – Debris-flow 

Simulations with RAMMS). Master Thesis, Swiss Federal Institute for Forest, Snow 

and Landscape Research WSL and Laboratory of Hydraulics Hydrology and Glaciology (VAW), 

ETH Zurich (in German, unpublished). 

- Hohermuth B., Graf C. (2014). Einsatz numerischer Murgangsimulationen am Beispiel des 

integralen Schutzkonzepts Plattenbach Vitznau (The use of numerical debris-flow 

simulations – case study Plattenbach). Wasser Energie Luft 106(4), 285-290 (in German) 

- Holinger AG, NDR Consult. (2013). Integrales Schutzkonzept Vitznauer Bäche (Integral 

Natural Hazard Protection Concept „Vitznauer Bäche“). Technical Report (in German). 

- Holinger AG (2015). Integrales Schutzkonzept Plattenbach / Mühlebach (Integral Natural 

Hazard Protection Concept Plattenbach / Mühlebach. Technical Report (in German) 

- Vetsch D., Siviglia A., Ehrbar D., Faccini M., Gerber M., Kammerer S., Peter S., Vonwiller L., 

Volz C., Farshi D., Mueller R., Rousselot P., Veprek R., Faeh R. (2006-2013). BASEMENT-Basic 

Simulation Environment for Computation of Environmental Flow and Natural Hazard 

Simulation. Version 2.4. ETH Zurich. Available from http://www.basement.ethz.ch 



Unveiling the avalanche activity in the Upper Goms 

Valley (Switzerland) over the past 400 years using 

tree-ring records 

Sébastien Guillet 1 ; Markus Stoffel 1 ; Christophe Corona 2 

ABSTRACT 

Knowing the spatial extent of past high magnitude snow-avalanches on poorly documented 

forested slopes is essential for land planners as it may help to define with better accuracy 

avalanche hazards maps. Dendrogeomorphology has proven to be an efficient method to 

reconstruct spatio-temporal activity of snow avalanches on forested slopes for which limited 

historical avalanche records are available. Based on tree-ring records, we reconstructed 

avalanche activity over the past 400 years for two slopes located above the village of 

Oberwald in the Upper Goms Valley (Valais, Switzerland). This high resolution chronology is, 

to date, one of the longest records ever elaborated for the Swiss Alps. 566 trees presenting 

growth disturbances related to avalanche activity have been sampled and analyzed. Analyses 

of tree-ring growth disturbances allowed identification of 38 events and the mapping of their 

spatial extent for the period 1600-2014. 12 snow-avalanches traveled through the slopes and 

stopped at or very close to the valley bottom in 1689, 1720, 1793, 1813, 1880, 1937, 1951, 

1999, and 2003. 

KEYWORDS 

snow-avalanches; tree-rings; dendrogeomorphology; forested slopes 

INTRODUCTION 

Snow avalanches are a major natural hazard in the Swiss Alps. Every year, they affect 

transport infrastructure and may endanger settlements and threaten human life. Over the last 

century, urban sprawl in mountains areas of Switzerland in combination with growing 

demands for mobility and recreational activities have increased avalanche risk significantly. In 

a society with ever increasing avalanche risk and safety expectations, risk acceptance of 

modern societies has been equally decreasing over recent decades. Substantial efforts have 

therefore been deployed over the last decades to build databases listing past avalanche events 

and providing accurate information regarding their magnitude, spatial extent and return 

period. However, for most of the Swiss Alps historical records of past avalanche activity are 

sparse and scarce before the 20th century. Past avalanche activity is especially poorly 

1 Dendrolab.ch, Institute of Geological Sciences, University of Berne, SWITZERLAND, sebastien,guillet@dendrolab.ch 

2 Geolab, Université Blaise Pascal, 4 rue Ledru F-63057 Clermont-Ferrand, FRANCE 



documented in the Upper Goms valley (canton of Valais) where no long chronologies of snow 

avalanches currently exist. 

Dendrogeomorphology is the science using information contained in tree rings to reconstruct 

the activity of past natural hazards in time and space. On forested slopes, this approach has 

proven successful to compensate for the scarcity of written sources (Corona et al., 2012). 

Over the past 40 years, several dendrogeomorphic studies have indeed demonstrated that 

trees impacted by mass movements, such as avalanches, landslides, rockfall or debris flows, 

record the event in the form of growth disturbances (Stoffel and Bollschweiler, 2008). Since 

trees form one increment ring per year in temperate climates it is possible to date the 

occurrence of the geomorphic process with an annual resolution. 

Trees growing in the upper Goms valley are particularly suitable for dendrogeomorphic analyses. 

Indeed the upper Goms valley holds some of the oldest forest stands of Switzerland, thus 

offering a unique opportunity and first-hand information to document past avalanche activity 

over several centuries. In 2013 the Canton of Valais mandated the dendrolab.ch to perform 

an important study with the goal of reconstructing spatio-temporal avalanche activity on 

several slopes located on the south-facing side of the Goms Valley (Oberwald, Geschinen, 

Münster). Here we present the results of the dendrogeomorphic study for two avalanche 

paths located above the village of Oberwald. 

METHODS 

Study site: The Oberwald avalanche paths (46°32’N, 8°20’E) are located on the south-facing 

slope of the upper part of the Goms Valley in the Swiss Central Alps (Canton of Valais, 

Switzerland, Figure 1A-D). According to the nearby weather station of Ulrichen (46°5’N, 

8°31’E, 1346 m asl), annual temperature is 3.3°C for the period 1981-2010 and annual 

precipitation amounts to 1212 mm. Winter temperature (DJF) is -6.6°C while winter 

precipitation amounts to 300 mm. Between November and April precipitation falls primarily 

as snow and average annual snowfall reaches 578 cm for the period 1999–2010 (MeteoSwiss, 

2014). At the study site, snow avalanches are commonly triggered from several non-forested 

release zones located between 1680 and 2200 m asl. The forested slope located underneath 

the starting zones is mainly composed of European larch (Larix decidua) and Norway spruce 

(Picea abies) and can be subdivided into two units. 

The first unit (OB1) is a vast area (68 ha) extending from the Jostbach torrent to the 

Rätischbach torrent. The canton of Valais has defined 6 main avalanche tracks within this 

slope (Figure 1C). OB1 is characterized by a double segmentation. Between 1900 and 1680 

m, the slope profile is characterized by gentle slopes (15° on average). Below 1680 m, slope 

angles increase significantly and exceed 20°. We therefore assume that new avalanches can 

be triggered from these locations. This hypothesis is further supported by several rows of 

avalanche barriers put in place by cantonal authorities in this portion of the slope to stabilize 

snowpack and to prevent the release of avalanches. 


The second unit (OB2) extends from the Rätischbach torrent to the Restaurant Rhonequelle 

(61 ha). Three main avalanche paths have been defined by the Canton of Valais for this sector 

(Figure 1C). In 1951, due to the damages caused to three houses after the release of an 

avalanche, a deflecting barrier was built in the runout zone of the three avalanches paths to 

protect the village. The Oberwald avalanche tracks do not only pose a direct threat to the 

village of Oberwald, they also menace the Matterhorn-Gotthard Bahn (MGB) railway line 

connecting Brig (Canton of Valais) to Andermatt (Canton of Uri). 

Figure 1: Location of the study site in (A) Switzerland and (B) in the Goms Valley. (C) Spatial distribution of the sampled trees. The black 

and red squares respectively refer to the spatial extent of the sites OB1 and OB2. The colored areas represent the hazard map defined by 

the canton of Valais. Red indicates an area that is exposed to considerable danger with frequent avalanches (average return period of 

30 years or less). In the blue colored area, avalanches are less frequent (more than 30 year average return period) and accompanied by 

only small compression forces of less than 30 kN/m². The danger in the yellow area is low. Yellow is typically used to designate the 

runout zone of powder avalanches. (D) Photograph illustrating the investigated avalanche paths. 

Sampling strategy: A total of 566 trees (1140 increment cores, 10 cross sections) have been 

sampled during the summers of 2013 and 2014 (Figure 1C). A minimum of two cores were 

extracted per tree, one perpendicular to the slope and one in the downslope direction. GPS 

coordinates were recorded for each tree with 5 m precision. Trees presenting obvious 

evidence of snow avalanches, such as decapitation, tilting and injury, were preferentially 

targeted. Following the recent recommendations made by Stoffel et al. (2013) and Stoffel and 

Corona (2014), we selected old trees in order to extend the reconstruction of past avalanche 

events as far as possible but also considered younger trees to take account of the loss in 

sensitivity of older trees to record damage (Šilhán and Stoffel, 2015). Trees located close to 


sectors with intense forest management activities (logging) were not sampled to avoid the 

inclusion of growth disturbances not related with avalanche activity in the final avalanche 

reconstruction. 

Laboratory analyses: All samples were prepared following standard dendrochronological 

procedures (Bräker, 2002). Cores were mounted on wooden sticks and then polished with 

sandpaper. Tree rings were counted and analyzed using a LINTAB-5 positioning table 

connected to a Leica stereomicroscope. Individual tree-ring series were cross-dated using two 

local reference chronologies (Büntgen et al., 2006, Schweingruber, 1974) so as to correct our 

series for possibly missing rings. In a second step, all cores were visually inspected under a 

binocular to identify growth disturbances (GD) induced by snow avalanches. Typically the 

reactions most commonly observed in tree rings following an avalanche event are: (1) an 

abrupt growth suppression (GS) following apex loss or the break-off of branches, (2) the 

formation of compression wood (CW) after tilting of the trunk, (3) the production of callus 

tissue (CT) and tangential rows of traumatic resin ducts (TRD) after an impact . We assigned a 

score to each GD following the recommendations made by Kogelnig-Mayer et al. (2011) to 

distinguish between weak (intensity class 1), medium (intensity class 2), and strong (intensity 

class 3) reactions and clear evidence of injuries (intensity class 4). 

Reconstruction of avalanche events: After the identification and the dating of the GD, we 

developed the avalanche reconstruction using the weighted index factor (Wit) developed by 

Kogelnig-Mayer et al. (2011). The Wit gives a weight to each GD based on its intensity. It is 

expressed as follows: 

nn 

WWWWWW = [(∑ GGGGGG4 ∗ 7) + (∑ GGGGGG3 ∗ 5) + (∑ GGGGGG2 ∗ 3) + (∑ GGGGGG1 ∗ 1)] ∗ ∑ nn 

ii=1 RR tt 

∑nn 

ii=1 

where GDI (1 to 4)= Sum of trees showing clear evidence of injury (intensity class 4), strong 

reactions (intensity class 3), medium signals (intensity class 2), weak signals (intensity 

class 1), respectively. 

nn 

ii=1 

nn 

ii=1 

nn 

ii=1 

ii=1 AA tt 

R t 

= Total amount of trees reacting in year t. 

A t 

= Number of sampled trees being alive (sample size) in year t. 

All the years with Wit ≥1 were considered as avalanche events. In addition, and in order to be 

sure that the thresholds chosen were not too restrictive, we also carefully examined all years 

with a 0.8≥ Wit ≤1. The inclusion of avalanche events in the database did not only rely on the 

Wit index, but was also based on the spatial distribution of impacted trees. When the spatial 

pattern of affected trees is not meaningful (in terms of avalanche trajectory), we rejected 

events even if the Wit ≥ 1 criteria was clearly fulfilled. 


In a last methodological step, we attempted to assess the detection of past events and to 

separate signal from noise. Noise is defined here as all growth disturbances that cannot be 

related to avalanche activity but to insect outbreaks. In the Goms valley and more widely in 

the Alps, larch trees have been documented to suffer from larch budmoth (Zeiraphera diniana 

Gn.) outbreaks such that tree growth remains suppressed for several years (Esper et al., 

2007). Since growth suppression is also typical in trees with decapitation, a careful analysis of 

the nature of the damage and the distribution of affected trees needs to be performed. We 

used several databases recording larch budmoth outbreaks in the Swiss Alps and in the Goms 

valley in particular to this end (Weber et al., 1997; Baltenschweiler and Ruby, 1999; Esper et 

al., 2007). All the years with a Wit ≥ 0.8, matching with documented larch budmoth year and 

presenting 80% of GS but almost no evidence of TRD, CW, or injuries were thus considered 

as dubious and removed from the final tree-ring based avalanche reconstruction. 

RESULTS 

Age structure of the stand: The average age of trees sampled at OB1 amounts to 217 years 

(SD: 105 yrs) with the oldest individual dating back to AD 1429 and the youngest tree 

reaching sampling height in AD 1987. A total of 51 (16) trees were at least 300 (400) yrs old. 

At OB2, the mean age of sampled trees is 196 yrs (SD: 120 yrs). A total of 54 (10) trees were 

at least 300 (500) years old. The oldest trees sampled at OB2 date back to 1495, 1478, 1464, 

1452, 1449, 1434, and 1429, respectively. The youngest tree dates back to 1998. For both 

sites, the distribution of the oldest larches suggests that there was no event of sufficient 

magnitude to destroy larger parts of the forest stands for at least 300 years. 

Reconstruction of snow avalanche events: Analysis of the 566 trees revealed 2590 GD. 

Abrupt growth suppressions and TRD represent the most frequently observed source of 

evidence of past events with 1270 (49%) and 1170 (45%) occurrences in the tree-ring series. 

By contrast, compression wood and injuries were by far less frequently observed on the increment 

cores with respectively 94 (3%) and 55 (2%) observations. 

At OB1, analysis allowed reconstruction of 14 snow avalanche events for the period 1689– 

2014, namely in 1982, 1948, 1945, 1937, 1935, 1925, 1906, 1894, 1880, 1870, 1842, 1800, 

1793, and 1689. The years with the largest amount of responses and/or the highest Wit 

values were 1793 (Wit=4.7), 1880 (Wit=3.3), 1842 (Wit=2.2), 1906 (Wit=2.19), and 1689 

(Wit=1.97), (Figure 2A-B). As a result of the limited number of samples reaching back 

beyond 1670 (

At OB2, we identified a total of 18 snow avalanches between AD 1605 and 2014, namely in 

2003, 1999, 1975, 1966, 1945, 1937, 1919, 1912, 1882, 1880, 1846, 1813, 1803, 1793, 1692, 

1650, 1641, and 1605. The largest number of disturbances and/or the highest Wit values were 

recorded in 2003 (Wit=5.3), 1999 (Wit=3.3), 1937 (Wit=8.7), 1882 (Wit=3.2), 1880 (Wit=12), 

1813 (Wit=2.7), and 1641 (Wit=2.7). Again, four snow avalanches could not be determined 

as event years with the same level of confidence and were therefore recorded as possible 

events (1961, 1885, 1734 and 1681), (Figure 2C-D). 

Based on the criteria introduced earlier and the fact that they match with periods of known 

larch budmoth outbreak activity, the years 1581, 1599, 1909, 1910, 1953, 1954, 1955 and 

1981 were disregarded in the snow avalanche reconstruction. In addition, we also decided to 

reject 1986 from the OB1 avalanche reconstruction as the spatial distribution of the trees 

impacted during this year exhibited a pattern which is unlikely to stem from an avalanche. 

TD 

A 

4DD 

TD 

C 

T88D 

T955 

4DD 

W it Threshold=T 

Wit=T2=T 

Wit=T6=5 

W it Threshold=D=8D 

8 

8 

3DD 

3DD 

W it 

6 

4 

2DD 

Sample depth 

W it 

6 

4 

2DD 

Sample depth 

TDD 

TDD 

2 

2 

D 

D 

D 

D 

T4DD T45D T5DD T55D T6DD T65D T7DD T75D T8DD T85D T9DD T95D 2DDD 

Years 


Years 

B 

D 

Event 

T 

D 

Previously undocumented events 

Confirmed historical events 

Known avalanches not identified 

in tree-ring records 

Uncertains events 

Rejected events 

4DD 

3DD 

2DD 

TDD 

D 

Sample depth 

Event 

T 

D 

4DD 

3DD 

2DD 

TDD 

D 

Sample depth 


Years 

Avalanche reconstruction - Site OBT 


Years 

Avalanche reconstruction - Site OB2 

Figure 2: Event-response histograms showing avalanche induced growth responses from sampled trees at OB1 and OB2: (A) Preliminary 

avalanche chronology obtained at OB1 using the Weighted Index method (Wit). After additional analysis (see methods), 3 years (1954, 

1955 and 1986) were removed from the final tree-ring based avalanche reconstruction. (B) Final avalanche chronology covering the 

period 1605-2014 at OB1. (C) Preliminary avalanche chronology obtained at OB2 using the Wit method. After additional analyses, 7 

years (1581, 1599, 1909, 1910, 1954, 1955 and 1981) were removed from the final tree-ring based avalanche reconstruction. (D) Final 

avalanche chronology covering the period 1689-2014 at OB2. 

Spatial extent of avalanches events: The spatial distribution of trees affected by the same 

event was used to determine the minimum lateral spread and the minimum runout elevation 

of reconstructed avalanches. At OB1, the mean runout elevation is calculated at 1625 m asl. 

Runout elevation varies between 1755 m asl in 1880 and 1450 m asl in 1935 (1370 m asl in 

1951 if we include historical records; SLF 1952). Large avalanches stopped next to or at the 

valley bottom in at least 5 cases over the last 400 years, namely in 1689, 1793, 1894, 1937, 

and 1951. Two other events (1935, 1948) reached the valley floor but were released from the 

middle part of the slope. We also reconstructed 9 events which stopped in the upper portions 


of the slope (and at elevations above 1680 m asl), namely in 1800, 1810, 1870, 1880, 1906, 

1925, 1945, 1990 and 1995. 

At OB2, the mean runout elevation is 1625 m asl, varying between 1840 m asl in 1690 and 

1460 m asl in 2003 (1370 m in 1720 if we include historical records). Since 1580, at least 7 

snow avalanches developed sufficient energy to reach the valley floor, namely in 1720, 1793, 

1813, 1880, 1937, 1951, and 2003. According to our reconstruction, 15 events were, by 

contrast, restricted the upper part of the slope and would have stopped at elevations > 1680 

m asl, namely in 1605, 1641, 1650, 1681, 1692, 1734, 1803, 1880, 1882, 1885, 1912, 1919, 

1945, 1966, and 1975. 

Temporal and spatial accuracy of the reconstruction: In this study, a total of 38 

avalanche events were reconstructed for the period 1600-2014 based on a dendrogeomorphic 

approach. To assess the temporal and spatial reliability of our reconstruction, we compared 

our results with local records of past avalanches. We observe that 4 out of the 5 documented 

snow avalanches that occurred on the avalanche forested slopes of Oberwald could be 

reconstructed with dendrogeomorphic techniques. The most recent avalanche which 

occurred in 2003 at OB2 and destroyed about 500 m 3 of forest, is successfully recorded in our 

reconstruction. Its spatial extent is also very well captured by tree rings, as can been seen 

from visual comparison between two aerial photographs taken in 1999 and 2003 (Figure 

3A-C). The events of 1999 (OB2), 1961 (OB2) and 1935 (OB1), known from archival records, 

were also successfully identified. Only one event occurring during the winter 1920-1921 

could not be identified. The success rate is slightly higher than for tree-ring reconstructions of 

avalanches performed in the Oisans massif in France (Corona et al., 2010) and in the 

Mont-Blanc massif at Chamonix (Corona et al., 2012). 

Figure 3: (A, B) Diachronic evolution of the forested slope of Oberwald (OB2) between 1999 and 2005. (C), Tree-ring-based 

reconstruction of the 2003 avalanche event. 


Three other large avalanches were documented in Oberwald (one event in 1720, and two in 

1951). However none of them occurred on the forested slopes of OB1 and OB2. The 

avalanches occurred and stayed within the incised paths of the Jostbach and the Rätischbach 

torrents. Dendrogeomorphic techniques have more limitations when it comes to reconstruct 

avalanche activity in these deeply incised paths, as avalanches tend to break, uproot and 

remove trees, leaving no datable evidence of past avalanche events. As a consequence, most 

of the vegetation found in these paths consists of young broadleaved trees, often in the form 

of large shrubs such as Alnus viridis or Betula pendula Roth. While these specimens can be 

used to assess the most recent past, they will not yield any information on more ancient snow 

avalanches. For this reason, we could not find evidence for the 1720 and 1951 avalanches 

which reportedly occurred in the Jostbach and Rätischbach torrents. 

CONCLUSIONS 

The analyses of growth disturbances in trees growing on the forested slopes located above the 

village of Oberwald allowed identification of 38 events as well as the mapping of their spatial 

extent over the past 400 years. In total, 32 of the avalanches documented through tree-ring 

analyses have not been known prior to this study. At least 12 snow avalanches traveled 

through the slopes and stopped at or very close to the valley floor. The 1689, 1720, 1793, 

1813, 1880, 1937, 1951, 1999, and 2003 events were the largest avalanches observed over 

the last 400 years. The tree-ring based chronology of past avalanches presented in this paper 

is, to date, one of the longest records ever elaborated for the Swiss Alps, and the third longest 

record existing at the level of the European Alps (Luzian et al., 2011, Corona et al., 2013). 

This study highlights how beneficial can the dendrogeomorphic approach be to land use 

planners, as it provides first-hand information for the assessment of runout distances and 

return periods of extreme avalanche events on forested slopes for which only scarce historical 

archives are available. This approach may help practitioners in the future to define with 

better accuracy avalanche hazard zones. 

REFERENCES 

- Baltensweiler W., Rubli D. (1999). Dispersal, an important driving force of the cyclic 

population dynamics of the larch-bud-moth Zeiraphera diniana Gn. Forest Snow Landscape 

Research 74: 1-153. 

- Bräker O. U. (2002). Measuring and data processing in tree-ring research: a methodological 

introduction. Dendrochronologia 20: 203-216. 

- Büntgen U., Esper J., Frank D., Nicolussi K., Schmidhalter M. (2005). A 1052-Year tree-ring 

proxy for alpine summer temperatures. Climate Dynamics 25: 41-153. 

- Esper J., Büntgen U., Frank D.C., Nievergelt D., Liebhold A. (2007). 1200 years of regular 

outbreaks in alpine insects. Proceedings of the Royal Society B 274: 671-679. 

- Corona C., Rovéra G., Lopez Saez J., Stoffel M., Perfettini P. (2010). Spatio-temporal 

reconstruction of snow avalanche activity using tree rings: Jean Jeanne avalanche talus, 

Massif de l'Oisans, France. Catena 83: 107-118. 


- Corona C., Lopez Saez J., Stoffel M., Bonnefoy M., Richard D., Astrade L., Berger F. (2012). 

How much of the real avalanche activity can be captured with tree rings? An evaluation of 

classic dendrogeomorphic approaches and comparison with historical archives. Cold Regions 

Science and Technology 74-75: 31-42. 

- Corona C., Lopez Saez J., Stoffel M., Rovéra G., Edouard J.-P., Berger F. (2013). Seven 

centuries of avalanche activity at Echalp (Queyras massif, southern French Alps) as inferred 

from tree rings. The Holocene 23: 292-304. 

- Kogelnig-Mayer B., Stoffel M., Bollschweiler M., Hübl J., Rudolf-Miklau F. (2011). 

Possibilities and limitations of dendrogeomorphic time-series reconstructions on sites 

influenced by debris flows and frequent snow avalanche activity. Arctic, Antarctic, and Alpine 

Research 43: 619-658. 

- Luzian R., Pindur P., Nicolussi K., Sailer R., Haas J.N., Zwerger P. (2011). Holozänes 

Lawinengeschehen und „Global Warming". Zeitschrift für Wildbach, Lawinen, Erosions und 

Steinschlagschutz 167: 102-117. 

- MeteoSwiss (2014). http://www.meteoswiss.admin.ch/product/output/climate-data/ 

climate-diagrams-normal-values-station-processing/ULR/climsheet_ULR_np8110_e.pdf. 

Accessed on 02.04.2014. 

- SLF (1952). Schnee und Lawinen in den Schweizeralpen im Winter 1950/51, Winterbericht 

des Eidg. Instituts für Schnee und Lawinenforschung, Davos, Nr. 15, SLF Davos, 232 p. 

Schweingruber F.H. (1974). Riederalp VS Aletschwald, Swit140. NOAA International 

Tree-Ring Databank (ITRDB). 

- Šilhán K, Stoffel M. (2015). Impacts of age-dependent tree sensitivity and dating approaches 

on dendrogeomorphic time series of landslides. Geomorphology 236: 34-43. 

- Stoffel M., Bollschweiler M. (2008). Tree-ring analysis in natural hazards research – an 

overview. Natural Hazards and Earth System Sciences 8: 187-202. 

- Stoffel M., Butler D.R., Corona C. (2013). Mass movements and tree rings: A guide to 

dendrogeomorphic field sampling and dating. Geomorphology 200: 106-120. 

- Stoffel M., Corona C. (2014). Dendroecological dating of geomorphic disturbance in trees. 

Tree-Ring Research 70: 3-20. 

- Weber U. M. (1997). Dendroecological reconstruction and interpretation of larch budmoth 

(Zeiraphera diniana) outbreaks in two central alpine valleys of Switzerland from 1470-1990. 

Trees 11: 277-290. 



Management of glacier floods in the Bernese Oberland 

Umgang mit Gletscherhochwasser im Berner Oberland 

Nils Hählen 1 ; Oliver Hitz 2 ; Damian Stoffel 2 

ABSTRACT 

Glacier floods are an important part of natural hazards in the Bernese Oberland. As a result 

of climate change they will probably increase significantly in future. The approach of at least 

two scenarios with different characteristics (e.g. an optimistic and a pessimistic) is useful to 

handle the uncertainty in risk estimation. These scenarios can then be compared with the 

magnitude of floods caused by rainfall which are mapped on the hazard maps. This step 

allows a classification of the glacier flood in comparison to "normal" floods. Glaciers and their 

predisposition for hazardous processes change over time. As a consequence a periodical check 

of the situation is needed and if necessary the scenarios have to be adapted. Therefore a systematic 

monitoring and the sensitization of the persons responsible for security in the communes 

are essential for a successful hazard management of glacier risks. 


Gletscherhochwasser sind ein wichtiger Teil der Naturgefahren im Berner Oberland und 

dürften als Folgen des Klimawandels in Zukunft noch zunehmen. Für deren Risikoabschätzungen 

sind mindestens zwei Szenarien mit verschiedenen Ausprägungen (z.B. ein optimistisches 

und ein pessimistisches) hilfreich, um mit den Unsicherheiten umzugehen. Diese 

Szenarien können mit der Magnitude von niederschlagsverursachten Hochwassern verglichen 

werden, die in den Gefahrenkarten abgebildet sind. Dieser Schritt erlaubt eine Einordnung 

von Gletscherhochwasser im Vergleich zu “normalen” Hochwassern. Gletscher und ihre 

Disposition für gefährliche Prozesse verändern sich laufend. Deshalb ist eine periodische 

Kontrolle der aktuellen Situation wichtig und die Szenarien sind, wenn nötig, periodisch 

anzupassen. Für ein erfolgreiches Risikomanagement sind daher ein systematisches Monitoring 

der Gletscher und die Sensibilisierung der Sicherheitsverantwortlichen entscheidend. 

KEYWORDS 

glacier; flood; risk management; Bernese Oberland 

EINFÜHRUNG 

Rund 200 km 2 oder 7% der Fläche des Berner Oberlandes sind vergletschert. Schon immer 

haben Hochwasser aus Gletschern im Berggebiet Siedlungen bedroht. Im Berner Oberland 

gelten die Ereignisse Grüöbengletscher Guttannen (1921, 1942), Steingletscher Gadmen 

(1956, 1998) und Unterer Grindelwaldgletscher (1951, 2008) als die grössten in den letzten 

1 Amt für Wald des Kantons Bern, Interlaken, SWITZERLAND, nils.haehlen@vol.be.ch 

2 Oberingenieurkreis I, Tiefbauamt des Kantons Bern 



100 Jahren. Zu Beginn des 21. Jahrhunderts häuften sich solche Ereignisse als Folge des 

beschleunigten Gletscherrückgangs durch die Klimaänderung. Die Herausforderung im 

Management von Gletscherhochwasser liegt in der frühzeitigen Erkennung kritischer 

Situationen, der Herleitung realistischer Szenarien und dem geeigneten Umgang mit Risiken, 

die zwar nur vorübergehend bestehen, aber oft ein immenses Gefahrenpotential beinhalten. 

Gletscher können verschiedene Gefahrenprozesse verursachen. Weil die Reichweite von 

Hochwasser sehr gross ist, sind Gletscherhochwasser die am weitesten reichenden Prozesse 

unter den Gletschergefahren (Raymond et al. 2003) und bilden somit ein erhebliches 

Gefahrenpotential. Auslöser von Gletscherhochwassern sind plötzliche Entleerungen von 

Wasseransammlungen am, auf oder in einem Gletscher. Bei Gletscherhochwasser ist zwischen 

supraglazialen Gletschersee- und subglazialen Wassertaschenausbrüchen zu unterscheiden. 

Oft sind Gletscherseeausbrüche die Folge einer Verkettung von Prozessen, so z. B. wenn 

grosse Massenbewegungen in Gletscherseen, diese zum Überlaufen bringen. Weiterführende 

Informationen zu Gletscherhochwasser geben beispielsweise Raymond et al. (2003) oder 

Huggel et al. (2004). Vertiefte Informationen über die Hydraulik von Gletscherhochwasser 

sind in Worni et al. (2014) zu finden. 

GEFAHRENBEURTEILUNG BEI GLETSCHERHOCHWASSER 

Die Gefahrenbeurteilung besteht in der Festlegung von Szenarien sowie der Ausscheidung 

von Wirkungsräumen und dazugehörigen Prozessintensitäten. Ein pragmatisches Vorgehen 

für eine grobe Gefahrenbeurteilung bei Gletschern ist in Huggel et al. (2004) beschrieben. 

Diese Methode bewährt sich für eine erste, einfache Einschätzung. Falls danach eine 

relevante Gefährdung nicht ausgeschlossen werden kann, sind vertiefte Analysen nötig. 

Angaben zur Wahrscheinlichkeit von Ereignissen, zum Speichervolumen und zur Ausbruchsart 

mit dazugehöriger Ganglinie stellen die grossen Herausforderungen bei der Szenarienbildung 

für Gletscherhochwasser dar. Die verschiedenen Einflussfaktoren auf Szenarien und 

Prozessablauf sind in Abbildung 1 dargestellt und qualitativ nach ihrem Grad an Unsicherheit 

klassiert. Die Einteilung in die drei Kategorien erfolgt aufgrund des Kriteriums Qualität der 

Wissens- resp. Datenbasis und des Kriteriums der Subjektivität der nötigen Beurteilung. 

Unterliegen beide Kriterien einem grossen Streubereich, wird die Unsicherheit als gross, 

wenn nur eines einen grossen Streubereich hat, als mittel und wenn beide einen kleinen 

Streubereich haben als klein klassiert. Bei der Gefahrenbeurteilung ist auf die Faktoren mit 

grossen Unsicherheiten besonderes Augenmerk zu richten. 

Häufig wiederholen sich Gefahrenprozesse an einem Gletscher. Informationen zu früheren 

Ereignissen sind daher sehr hilfreich. Liegen keine Informationen zu früheren Ereignissen vor 

oder hat sich die Situation seit den letzten Ereignissen relevant verändert, sind längerfristige 

Prognosen mit grossen Unsicherheiten verbunden. Dann muss auf Analysen, Annahmen und 

Analogieschlüsse zu ähnlichen Fällen abgestützt werden. Die Unsicherheiten sind entsprechend 

gross; oft bedeutend grösser als bei anderen Prozessarten. Weil Gefahrenbeurteilungen 

bei Gletscherhochwasser meist erst gemacht werden, wenn eine Gefahrenquelle neu entsteht, 


ist die Wahrscheinlichkeit gross, dass sie in kurzer Zeit anhand tatsächlicher Ereignisse 

überprüft werden kann - Chance für die Experten, ihre Überlegungen zu verifizieren, aber 

auch Fluch, weil Abweichungen Erklärungsbedarf gegenüber der Bevölkerung und Behörden 

nötig machen und die Glaubwürdigkeit der Experten leiden kann. Die grössten Unsicherheiten 

in der Gefahrenbeurteilung von Gletscherhochwasser liegen in der Festlegung der 

Szenarien. Die aus den Szenarien abgeleitete Wirkungsbeurteilung beinhaltet bedeutend 

weniger Unsicherheiten, weil es für die Ausbreitung der Prozesse heute gute Modellierungssoftware 

gibt. Auch die dazu notwendigen Höhenmodelle liegen in der Schweiz in hoher 

Genauigkeit vor. 

Abbildung 1: Phasen in der Gefahrenbeurteilung und im Risikomanagement von Gletschergefahren mit dazugehörigen Teilschritten und 

Klassierung der Unsicherheit. 

Um die Unsicherheiten in der Szenarienbildung zu reduzieren, hat es sich bewährt, eine 

Bandbreite an möglichen Szenarien festzulegen, was am Giesengletscher an der Nordwestflanke 

der Jungfrau exemplarisch gezeigt werden soll. An diesem hat sich seit 2011 an der 

Gletscherzunge ein Eispaket abgespaltet. Bei einem Abbruch ist davon auszugehen, dass die 

Eistrümmer in der darunterliegenden Schlucht eine Wasserstauung verursachen. Je nach 

Ausbildung der Ablagerung sind im Falle eines nachfolgenden Seedurchbruchs unterschiedliche 

Abflussspitzen möglich. Ein realistisches Szenario mit 100‘000 m 3 Wasser würde eine 

Abflussspitze von 80 – 130 m 3 /s verursachen, im pessimistischen Szenario ist bei 400‘000 m 3 

Wasser mit einer Abflussspitze von 350 – 440 m 3 /s zu rechnen. 

RISIKOBEURTEILUNG VON GLETSCHERHOCHWASSER 

Das Gefahrenpotential von Gletschern ändert sich über die Zeit, womit Gletscherhochwasser 

nicht immer gleich wahrscheinlich sind. Dies unterscheidet Gletscher von anderen Gefahrenquellen, 

deren Grunddisposition über längere Zeiträume keiner relevanten Veränderung 

unterworfen ist. Diese Variation hat vor allem mit der kontinuierlichen, klimabedingten 

Geometrieänderung (Vorstoss- und Rückzugsphasen) der Gletscher zu tun: Neue Senken 

entstehen und bestehende vergrössern sich oder lösen sich auf. Ein Gletscher hat Phasen, in 


denen Gletscherhochwasser gehäuft und Phasen, in denen solche Prozesse nicht auftreten. 

Oft besteht eine Gefahrenquelle auch nur vorübergehend für einige Monate oder wenige 

Jahre. 

Abbildung 2: Vergleich der Bandbreite der niederschlagsverursachten Hochwasser (HQ 30 

bis HQ 300 

) aus der Gefahrenkarte, den effektiv 

aufgetretenen Abflussspitzen durch Gletscherhochwasser und der Bandbreite der Prognose für Abflussspitzen durch Wasserausbrüche 

für verschiedene Gletscher im Berner Oberland. 

Die Risikobeurteilung setzt voraus, dass neben Ort und Intensität eines möglichen Gletscherhochwassers 

auch die Eintretenswahrscheinlichkeit und das Ausmass des daraus resultierenden 

Schadens bekannt sind. Risiko ist allgemein das Produkt aus Schadenausmass und Eintretenswahrscheinlichkeit. 

Wegmann et al. (2004) haben eine Methode zum Risikomanagement 

bei Gletschergefahren entwickelt. Diese beruht darauf, dass im Rahmen eines partizipativen 

Verfahrens mit Fachspezialisten und Lokalkennern die Risiken durch Gletschergefahren 

quantitativ abgeschätzt werden können. In der Anwendung dieser Methode bildet v.a. die 

Festlegung der Eintretenswahrscheinlichkeit aufgrund der oft sehr grossen Unsicherheiten 

eine Schwierigkeit. Dies kann am Beispiel des Oberen Grindelwaldgletschers gut gezeigt 

werden: Dort wurde von 2009 bis 2011 durch über dreissig Wassertaschenausbrüche dreimal 

ein 100-jährliches und sechsmal ein 30-jährliches Hochwasser überschritten. In einem 

solchen Fall ist es nicht sinnvoll, die Hochwasserstatistik, welche sich auf niederschlagsbedingte 

Hochwasser bezieht, anzupassen und das bisherige 100-jährliche neu als 10-jährliches 

Hochwasser zu klassieren, da absehbar ist, dass die Gefahrensituation nur solange anhält, bis 

sich der Gletscher in wenigen Jahren aus der engen, flachen Schlucht zurückgezogen hat. 

Hier eine Wiederkehrperiode für Gletscherhochwasser zu definieren, ist nicht möglich. Der 

Unsicherheitsbereich einer so festgelegten Wiederkehrperiode würde um ein Mehrfaches 

stärker variieren als alle übrigen Einflussgrössen auf das Risiko. 


Bezüglich Eintretenswahrscheinlichkeit bei Gletscherhochwasser ergeben sich zwei Fragestellungen: 

– Interannuelle Wahrscheinlichkeit: Wie gross ist die Wahrscheinlichkeit für einen oder 

mehrere Wasserausbrüche mit einer bestimmten Abflussspitze innerhalb eines Jahres? 

– Dekadische Wahrscheinlichkeit: Welche Wiederkehrperiode ist einer solchen Abflussspitze 

in der Hochwasserstatistik des betroffenen Gewässers innerhalb einer Zeitspanne von 

einem oder mehreren Jahrzehnten zuzuordnen? 

Die interannuelle Wahrscheinlichkeit lässt sich nur durch eine situative, gutachterliche 

Beurteilung der Gefahrendisposition anhand des aktuellen Zustands des Gletschers und seines 

Umfelds beantworten und beinhaltet die beschriebenen Schritte der Gefahrenbeurteilung 

(siehe oben). Wird die Abflussspitze eines Gletscherhochwassers in Relation zu einem 

„normalen“ niederschlagsverursachten Hochwasser mit einer bekannten Wiederkehrperiode 

von beispielsweise 30, 100 oder 300 Jahren gesetzt, lässt sich die dekadische Wahrscheinlichkeit 

eingrenzen. Dies setzt aber voraus, dass Gletscherhochwasser über längere Zeiträume 

betrachtet nicht zu einer wesentlichen Veränderung der Hochwasserstatistik führen. Damit 

kann das potentielle Ereignis je nach Ausprägung in den Bereich eines häufigen oder eines 

aussergewöhnlichen Prozessablaufs eingeordnet werden (vgl. Abbildung 2). Gleichzeitig lässt 

sich die Eintretenswahrscheinlichkeit in der Risikoformel grob eingrenzen. 

RISIKOMANAGEMENT VON GLETSCHERHOCHWASSER 

Das Risikomanagement hat zum Ziel, Schäden an Personen und Sachwerten zu minimieren 

oder im besten Fall zu verhindern. Die Konzepte und Instrumente des integralen Risikomanagements 

bei gravitativen Naturgefahren in der Schweiz haben sich bewährt und lassen sich 

auch bei Gletscherhochwasser erfolgreich einsetzen. 

Da Gletscherhochwasser oft nur vorübergehend ein Risiko darstellen und die Wahrscheinlichkeit 

eines Grossereignisses meist gering ist, konzentrieren sich Präventionsmassnahmen 

üblicherweise auf organisatorische Massnahmen, d.h. den Schutz von Personen. Schäden an 

Sachwerten werden dabei eher in Kauf genommen. Das Risikomanagement setzt voraus, dass 

ein mögliches Gefahrenpotential erkannt wird. Da sich die Gletscher innerhalb kurzer Zeit 

stark verändern können, müssen sämtliche Schritte von der Szenarienbildung bis zur 

Wirkungsbeurteilung laufend überprüft und gegebenenfalls angepasst werden. Ein systematisches 

Monitoring ist daher eine essentielle Daueraufgabe. Dieses ermöglicht es, die Bildung 

oder Vergrösserung von Gletscherseen frühzeitig zu erkennen. Im Berner Oberland basiert 

dieses hauptsächlich auf periodischen Auswertungen von Luftbildern und Beobachtungsmeldungen 

durch Sicherheitsverantwortliche aus den Gemeinden, Bergführer oder Berggänger. 


RAUMPLANERISCHE MASSNAHMEN 

Präventive raumplanerische Massnahmen gegen Gletscherhochwasser sind aufgrund der 

grossen Unsicherheiten oft nicht realisierbar. Sofern die Wirkungsräume und Intensitäten 

eines Gletscherhochwassers im Bereich des in der Gefahrenkarte abgebildeten, niederschlagsverursachten 

Hochwassers liegen (vgl. Abbildung 2), besteht keine Notwendigkeit, diese 

speziell in der Gefahrenkarte zu berücksichtigen. Sobald sie aber ein bedeutend grösseres 

Ausmass annehmen, sind sie bei der Beurteilung von Bauvorhaben über Bauverbote oder 

Bauauflagen zu berücksichtigen. Dazu werden die erwarteten Wirkungsräume in einer 

separaten Karte als ergänzendes Gefahrengebiet mit unbestimmter Gefahrenstufe dargestellt. 

Eine solche Karte hat oft nur wenige Jahre Gültigkeit. Sie wird aufgehoben, sobald die 

Gefährdung nicht mehr besteht. So war aufgrund des Gletschersees auf dem Unteren 

Grindelwaldgletscher von 2008 bis 2010 in acht unterliegenden Gemeinden eine solche Karte 

in Kraft und hat sich als Präventionsmittel sehr bewährt. Grössere Bauvorhaben in kritischen 

Gebieten wurden gestützt auf dieses Instrument zurückgestellt. 

ORGANISATORISCHE MASSNAHMEN 

In Bezug auf Schutzmassnahmen ist die Art und Exposition der Schutzgüter entscheidend. In 

Fällen, in denen v.a. Personen gefährdet sind, ist es zweckmässig, die gefährdeten Räume 

abzusperren, wenn sich eine Gefahrensituation einstellt. Um diese Entscheidungen treffen zu 

können, sind Überwachungs- und Frühwarnsysteme hilfreich, welche die Sicherheitsverantwortlichen 

mit den notwendigen Informationen versorgen. Frühwarnsysteme haben zum 

Ziel, die Reaktionszeit im Vorfeld von Ereignissen zu erhöhen. Im Berner Oberland sind 

einige dieser Systeme erfolgreich im Einsatz. Voraussetzung dafür ist, dass die Prozesse 

verstanden sind und eine Notfallorganisation mit Pikettdienst besteht, welche die notwendigen 

Massnahmen bei kritischen Entwicklungen zeitgerecht umsetzt. Auf der Plaine Morte in 

der Gemeinde Lenk besteht ein solches Frühwarnsystem, das auf automatischen Kameras 

beim Gletschersee, einer Pegelmessung im See und einer Abflussmessung beim Gletschertor 

basiert. Durch die Gletscherseeausbrüche sind in erster Linie eine Alp, ein viel begangener 

Wanderweg in einem beliebten Ausflugsgebiet sowie ein Hotel und ein Campingplatz 

gefährdet. Die Seeausbrüche verliefen bisher sehr gutmütig: Ein Ausbruch war dank des 

Frühwarnsystems Tage im Voraus in der Stagnation und im anschliessenden gemächlichen 

Absinken des Seepegels erkennbar, bevor ein deutlicher Anstieg des Abflusses in der Simme 

auftrat. Damit besteht eine komfortable Reaktionszeit, um gefährdete Wanderwege präventiv 

zu sperren. Für den Fall eines grösseren Ausbruchs liegt eine Notfallplanung zum Schutz von 

Personen vor. 

BAULICHE SCHUTZMASSNAHMEN 

Bauliche Schutzmassnahmen an Gletschern sind selten und werden nur dann ergriffen, wenn 

das Ausmass eines Ereignisses sehr gross werden kann und der Schutz mit organisatorischen 

Massnahmen ungenügend ist. Beim Gletschersee auf dem Unteren Grindelwaldgletscher 

waren die Personenrisiken mit dem 2006 eingerichteten Frühwarnsystem und einer 


umfangreichen Notfallplanung in den Griff zu bekommen. Für den Stollenbau war dann aber 

das erwartete Schadenausmass im Lütschinental bis nach Interlaken von rund Fr. 200 Mio. 

(berechnet gemäss Methode Econome 1.0 BAFU 2007) im Falle eines Seeausbruchs sowie der 

absehbare mehrwöchige Verkehrsunterbruch durch Schäden an der Verkehrsinfrastruktur 

nach Grindelwald entscheidend (Fechtig, Hählen 2013). Daher wurden die Investitionen von 

Fr. 15 Mio. für den Stollenbau in einer aufwändigen Variantenevaluation durch ein Expertengremium 

als verhältnismässig angesehen. Der Stollen wurde 2009 gebaut und hat seither 

zusammen mit dem weiteren Abschmelzen des Gletschers die Gefahrenlage am Unteren 

Grindelwaldgletscher markant entschärft. 

VERZICHT AUF MASSNAHMEN 

Nicht bei jedem Gletschersee sind aufwändig Massnahmen nötig. Aus dem proglazialen 

Gletschersee am Hubelgletscher im hinteren Lauterbrunnental hat sich 2004 ein grosser 

Murgang ereignet, der eine Brücke zerstört, sechs parkierte Autos weggerissen und grosse 

Teile der Alp- und Forststrasse beschädigt hat. Ursache des Ausbruchs war der Bruch einer 

Verklausung des Seeausflusses durch Eisblöcke aus Eiskalbungen des hinterliegenden 

Gletschers. Ein solches Ereignis kann immer noch auftreten. Da jedoch nur Alpgebiete und 

Wanderwege betroffen sind, wurde auf umfangreiche Schutzmassnahmen verzichtet. Die 

gefährdeten Abschnitte des Wanderwegs sind mit Warntafeln mit Verhaltensanweisungen 

versehen. Daneben wird der Zustand des Gletschers periodisch durch die Sicherheitsverantwortlichen 

beurteilt. 

AUSBLICK 

Gemäss NELAK (2013) können im Berner Oberland im Zuge des Gletscherschwundes 

innerhalb der nächsten Jahrzehnte über 100 neue Seen auf insgesamt 39 Gletschern 

entstehen. Zusammen mit den tendenziell zunehmenden Instabilitäten durch degradierenden 

Permafrost, woraus sich Massenbewegungen in neue Seen ergeben können, dürfte das 

Gefahrenpotential in Zukunft weiter ansteigen. 2015 wurde im Berner Oberland eine Studie 

im Auftrag des Kantons abgeschlossen (vgl. Tobler et al. in Vorbereitung), welche u.a. die 

Entwicklung des Gefahrenpotentials der Gletscher bis ins Jahr 2060 grob beurteilt. Damit 

besteht neben dem Inventar der gefährlichen Gletscher der Schweiz (Raymond et al. 2003), 

welches auf einer retrospektiven Sicht beruht, auch eine Übersicht über mögliche Gefahren, 

die künftige Entwicklungen miteinschliesst. Diese Studie erlaubt es, die potentiell kritischen 

Gebiete zu identifizieren und mit einem geeigneten Monitoring ungünstige Entwicklungen 

frühzeitig erkennen zu können. Da die Bildung von gefährlichen Situationen bei Gletschern 

umso wahrscheinlicher ist, je stärker und schneller die Veränderung der Gletscher abläuft 

(GLACIORISK 2003), muss in den kommenden Jahren mit weiteren anspruchsvollen 

Situationen gerechnet werden. Die Bedeutung von anpassungsfähigen und verhältnismässigen 

Massnahmen wird dabei immer mehr zunehmen. 


LITERATUR 

- BAFU (2007): EconoMe 1.0. Online-Berechnungsprogramm zur Bestimmung der Wirtschaftlichkeit 

von Schutzmassnahmen gegen Naturgefahren. Handbuch/Dokumentation, 

Bern, 27 s. 

- Fechtig R., Hählen N. (2013): Entwässerungsstollen Gletschersee Unterer Grindelwaldgletscher. 

In: Tunnelling Switzerland, vdf Hochschulverlag Zürich, S. 386-391. 

GLACIORISK (2003): GLACIORISK. Survey and prevention of extreme glaciological hazards 

in European mountainous regions. EVG1 2000 00512 Final report (01.01.2001–31.12.2003), 

compiled by Richard D. and Gay M., Cemagref Grenoble, 62 S. 

- Huggel Chr., Haeberli W., Kääb A., Bieri D., Richardson S. (2004): An assessment procedure 

for glacial hazards in the Swiss Alps. In: Canadian Geotechnical Journal, 41, S. 1068-1083. 

- NELAK (2013): Neue Seen als Folge des Gletscherschwundes im Hochgebirge – Chancen 

und Risiken. Forschungsbericht NFP 61. Haeberli W., Bütler M., Huggel Chr., Müller H., 

Schleiss A. (Hrsg.), vdf Hochschulverlag Zürich. 290 S. 

- Raymond M., Wegmann M., Funk M. (2003): Inventar gefährlicher Gletscher in der 

Schweiz, VAW Mitteilungen 182, ETH Zürich, 368 S. 

- Tobler D., Mani P., Riner R., Liener S., Hählen N., Bender R., Graf K., Raetzo H. (in 

Vorbereitung): Periglazial Hazard Indication Map: A Basic Instrument in Prospective Hazard 

Management. Interpraevent 2016, Tagungsbeiträge, Klagenfurt. 

- Wegmann M., Bruderer A., Funk M., Wuilloud Ch. (2004): Partizipatives Verfahren zum 

Risikoma-nagement bei Naturgefahren. Angewendet für Gletschergefahren. In: Interpraevent 

2004 Band IX, Klagenfurt, S. 297-308. 

- Worni R., Huggel Chr., Clague J.J., Schaub Y., Stoffel M. (2014): Coupling glacial lake 

impact, dam breach, and flood processes: A modeling perspective. In: Geomorphology 224. S. 

161-176. 



New recommendations for the assessment of river 

bank erosion hazards 

Eine neue Empfehlung zur Beurteilung der Gefahr 

von Ufererosion an Fließgewässern 

Lukas Hunzinger, Dr. 1 ; Annette Bachmann 2 ; Ralph Brändle 3 ; Paul Dändliker 4 ; David Jud 5 ; Mario Koksch 6 

ABSTRACT 

An important prerequisite to protect settlements and lifelines from bank erosion effectively is 

a sound hazard assessment. This recommendation describes a step by step approach for such 

an assessment. In a first step basic scenarios are defined. The second step consists of a weak 

point analysis, in which the relevant load cases are identified. Load cases are: erosion of the 

toeslope, direct bank erosion and erosion of the slope crest. For each basic scenario the load 

parameters corresponding to the relevant load cases are determined as a function of discharge, 

bed load and estimated bed level changes and are compared to the resistance 

parameters of the bank. Bank erosion is assumed if the load parameters are larger than the 

resistance parameters. Finally, the dimension of the bank erosion is estimated. The erosion 

width may measure a multiple of the original channel width, depending whether changes in 

the channel pattern lead to large scale channel migration or whether bank erosion is triggered 

by local irregularities in the channel topography only. 


Eine wichtige Grundlage, um Siedlungen und Verkehrswege wirksam vor Ufererosion 

schützen zu können, ist eine qualitativ und quantitativ nachvollziehbare Gefahrenbeurteilung. 

Die vorliegende Empfehlung beschreibt die wichtigsten Schritte dazu. Nach der 

Definition von Grundszenarien werden im Rahmen der Schwachstellenanalyse die maßgeblichen 

Gefährdungsbilder für das Ufer identifiziert: Erosion am Böschungsfuß, direkter 

Strömungsangriff und Erosion an der Böschungsoberkante. Für alle zu untersuchenden 

Hochwasserszenarien werden für diese Gefährdungsbilder die Belastungsgrößen in Abhängigkeit 

von Abfluss, Geschiebetransport und Sohlenveränderung während eines Hochwassers 

bestimmt und dem Widerstand der Böschung gegenüber gestellt. Ist die Belastung größer als 

der Widerstand muss auf dem betreffenden Gewässerabschnitt mit Erosion gerechnet werden. 

Die Breite der Erosion kann ein mehrfaches der Gerinnebreite betragen. Sie hängt davon ab, 

1 Flussbau AG SAH, Bern, SWITZERLAND, lukas.hunzinger@flussbau.ch 

2 CSD Ingenieure AG, SWITZERLAND 

3 Kanton St. Gallen, Naturgefahrenkommission, SWITZERLAND 

4 Bundesamt für Umwelt BAFU, SWITZERLAND 

5 Meier und Partner AG, SWITZERLAND 

6 Kanton Luzern, vif, SWITZERLAND 



ob übergeordnete morphologische Prozesse zu einer großräumigen Verlagerung des Gerinnes 

führen oder ob die Seitenerosion durch eine lokale Unregelmäßigkeit im Gerinnequerschnitt 

verursacht wird. 

KEYWORDS 

bank erosion; hazard analysis; weak point analysis 

EINLEITUNG 

Die Ufererosion ist ein Gefahrenprozess, der im Vergleich zur Überflutung weniger häufig 

und auf kleineren Flächen auftritt. Ihr Gefahrenpotenzial wurde bislang allerdings unterschätzt. 

Dies hat insbesondere das Hochwasserereignis im Jahr 2005 in der Schweiz und in 

den angrenzenden Ländern aufgezeigt, bei welchem durch den Abtrag von Uferböschungen 

viele Gebäude und Infrastrukturanlagen beschädigt oder zerstört wurden (Hilker et al., 2007). 

Eine wichtige Grundlage, um Siedlungen und Verkehrswege wirksam vor Ufererosion 

schützen zu können, ist eine qualitativ und quantitativ nachvollziehbare Gefahrenbeurteilung. 

Zur Beurteilung des Prozesses Ufererosion gibt es bis heute keine allgemein anerkannten 

Methoden oder Berechnungsgrundlagen. Um diese Lücke zu schließen, haben die Fachleute 

Naturgefahren Schweiz (FAN) und die Kommission für Hochwasserschutz, Wasserbau und 

Gewässerpflege (KOHS) eine Empfehlung zur Beurteilung der Gefahr von Ufererosion 

erarbeiten lassen (FAN und KOHS 2015). Die Empfehlung richtet sich an Wasserbau- und 

Naturgefahrenfachleute aus Praxis und Verwaltung. Sie schlägt Vorgehensweisen vor, nach 

denen die Gefahr von Ufererosion Schritt für Schritt beurteilt werden soll. Die Wahl der 

geeigneten Berechnungsansätze, mit denen die maßgeblichen Größen in den einzelnen 

Bearbeitungsschritten bestimmt werden, ist hingegen dem Anwender überlassen. 

DER PROZESS DER SEITENEROSION 

Unter Seitenerosion versteht man den lateralen Abtrag des Böschungsmaterials und die 

daraus resultierende Verbreiterung eines Gerinnes. Die Erosion wird durch die Strömung 

verursacht, die wirkenden Kräfte werden mit der Schleppspannung beschrieben. Damit es zu 

einer Erosion kommt, muss die angreifende Schleppspannung größer sein als die Grenzschleppspannung 

der Böschung. Diese hängt unter anderem von der Art des Ufermaterials 

(Korngröße, Kohäsion, etc.), dem Bewuchs (Beck 2006, Requena 2008) sowie einer 

allfälligen Uferverbauung ab (Romang 2008). 

In der Literatur wird zwischen primärer und sekundärer Seitenerosion unterschieden. Gemäß 

Anderson et al. (1975) bewirkt die primäre Seitenerosion eine Verbreiterung des Gerinnes auf 

der gesamten Länge, ohne dass sich die Lage und die Form des Gerinnes verändern. Bei der 

sekundären Seitenerosion hingegen verlagert und verformt sich das Gerinne aufgrund von 

Querströmungen. Abhängig von Flussbettbreite, Korngrößen, Gerinneneigung und Sedimen- 


teintrag bilden sich Mäander oder ein verzweigtes Gerinne. Ahmari und da Silva (2011) 

haben Kriterien definiert, um das Auftreten verschiedener Gerinneformen voneinander 

abzugrenzen. 

Im vorliegenden Artikel wird die Seitenerosion als Gefahrenprozess verstanden, welcher 

außerhalb des Gerinnes beobachtet wird. Er wird deshalb als Ufererosion bezeichnet. 

DAS GENERELLE VORGEHEN ZUR BEURTEILUNG DER EROSIONSGEFAHR 

Das generelle Vorgehen zur Beurteilung der Gefahr von Ufererosion ist in die drei Phasen 

Definition von Grundszenarien, Schwachstellenanalyse und Wirkungsanalyse unterteilt 

(Abbildung 1). Mit den Grundszenarien werden Abflussmengen, Sedimenteintrag und 

Schwemmholzaufkommen bei möglichen Hochwasserereignissen festgelegt. In der Schwachstellenanalyse 

wird die Frage beantwortet, ob es auf dem untersuchten Gewässerabschnitt 

überhaupt zu einer Ufererosion kommen kann und in der Wirkungsanalyse wird das Ausmaß 

der Erosion bestimmt. Die Schwachstellen- und die Wirkungsanalyse werden für jeden 

untersuchten Gewässerabschnitt und für jedes Grundszenario einzeln durchlaufen. 

DIE SCHWACHSTELLENANALYSE 

Bei der Schwachstellenanalyse werden auf der Grundlage der aktuellen Morphologie und 

möglicher Veränderungen während eines Hochwassers verschiedene Gefährdungsbilder 

untersucht: a) Erosion am Böschungsfuß, b) direkter Strömungsangriff und c) Erosion an der 

Böschungsoberkante (Abbildung 2). 

Mit der Beschreibung der aktuellen Morphologie und der Prognose von morphologischen 

Veränderungen während eines Hochwasserereignisses können für einen Gerinneabschnitt die 

maßgeblichen und die weniger wahrscheinlichen Gefährdungsbilder identifiziert werden. So 

muss zum Beispiel das Gefährdungsbild Erosion am Böschungsfuß sicherlich berücksichtigt 

werden, wenn während eines Hochwasserereignisses Sohlenerosion erwartet wird. Werden 

hingegen Geschiebeablagerungen im Gerinne erwartet, ist das Gefährdungsbild Erosion an 

der Böschungsoberkante wahrscheinlicher. Der direkte Strömungsangriff muss am Prallhang 

und in geraden Fließstrecken berücksichtigt werden, nicht aber am Gleitufer von Flusskrümmungen. 

Ein nicht zu unterschätzender Aspekt der Morphologie sind Unregelmäßigkeiten im 

Abflussquerschnitt wie zum Beispiel Einbauten (Buhnen, Schwellen) oder Abflusshindernisse 

im Querschnitt (Bäume, Wurzelstöcke). Diese können Erosionsprozesse initiieren, wenn sie 

zu einer Strömungskonzentration am Ufer oder zur Bildung von Kolken führen. 

Für die als maßgeblich betrachteten Gefährdungsbilder werden die Belastungsgrößen in 

Abhängigkeit von Abfluss, Geschiebetransport und Sohlenveränderung während eines 

Hochwassers bestimmt. Je nach Gefährdungsbild ist die Schleppspannung oder die Kolktiefe 

die entscheidende Belastungsgröße. In vielen Fällen müssen die Belastungsgrößen gutachterlich 

und qualitativ bestimmt werden. In Gewässern in denen die hydraulischen Verhältnisse 

bei Hochwasser mit ausreichender Genauigkeit vorhergesagt werden können, sollen die 


Grundszenarien 

Abfluss Q 

Geschiebezufuhr G 

Holzzufuhr H 

Schwachstellenanalyse 

Bekannte Erosionsstellen 

Morphologie 

- aktuell 

- Veränderungen bei Hochwasser 

Gefährdungsbilder 

a) Erosion am Böschungsfuß 

b) Direkter Strömungsangriff 

c) Erosion an der 

Böschungsoberkante 

Belastungsgrößen 

- Schleppspannungen 

- Kolktiefen 

Erosionswiderstand 

- Uferbeschaffenheit 

- Bauhöhe, Baulänge, Fundation 

- Zustand 

- Grenzbelastung 

Fazit: 

Seitenerosion möglich? 

nein 

Beurteilung Ufererosion für 

das Szenario abgeschlossen 

ja 

Wirkungsanalyse 

Ausdehnung und Intensität 

- Flächige Ausdehnung 

- Intensität 

Räumliche 

Auftretenswahrscheinlichkeit 

Folgeprozesse untersuchen 

Abbildung 1: Flussdiagramm zum generelles Vorgehen zur Beurteilung der Gefahr von Ufererosion. 


a) Erosion am Böschungsfuß 

b) Direkter Strömungsangriff 

B UE 

B UE 

c) Erosion an der Böschungsoberkante 

B UE 

Abbildung 2: Gefährdungsbilder bei der Beurteilung der Gefahr von Ufererosion. BUE bezeichnet die Erosionsbreite. Die Pfeile 

symbolisieren die maßgebliche Belastung. 


Belastungsgrößen jedoch berechnet werden. Für die Berechnung von Kolktiefen sind aus der 

Literatur verschieden Ansätze bekannt, z.B. von Tschopp und Bisaz (1972) bei Überfällen, 

von Peter (1986) oder Kikkawa et al. (1976) bei Flusskrümmungen und von Zarn (1997) in 

Gerinnen mit Bänken. Die maßgebliche Belastungsgröße beim direkten Strömungsangriff 

(Abbildung 2b) und bei der Erosion an der Böschungskante (Abbildung 2c) ist die Schleppspannung 

auf der Böschung bzw. dem darüber liegenden Terrain. 

Die Belastungsgrößen werden dem Erosionswiderstand des Ufers gegenüber gestellt. Um den 

Widerstand zu bestimmen, müssen die Beschaffenheit und der Zustand des Ufers bzw. des 

umliegenden Terrains erfasst werden. Zu den Parametern, welche den Widerstand beeinflussen, 

zählen die Art des Ufers (Fels oder Lockermaterial, verbaut oder nicht verbaut), seine 

Kornzusammensetzung (Kies, Feinsedimente; kohäsiv oder nicht), sein Bewuchs (flachgründig, 

tiefgründig), seine Neigung und gegebenenfalls die Fundationstiefe von Verbauungen. In 

vielen Fällen ist auch hier eine gutachterliche und qualitative Beurteilung notwendig. 

Insbesondere die Fundationstiefe einer bestehenden Uferverbauung ist nur selten bekannt 

oder kann nur mit verhältnismäßig großem Aufwand ermittelt werden. Bei Uferverbauungen 

mit Blockwürfen lässt sich der Erosionswiderstand z.B. mit dem Ansatz von Stevens und 

Simons (1971) bestimmen. Für ingenieurbiologische Verbauungen sind Grenzwerte der 

Schleppspannung zum Beispiel bei Pasche (2000) zu finden. 

Zum Schluss der Schwachstellenanalyse wird aufgrund der in den vorangehenden Schritten 

gewonnenen Erkenntnisse entschieden, ob an einer bestimmten Stelle des Gewässers 

Ufererosion anzunehmen ist oder nicht. 

DIE WIRKUNGSANALYSE 

Im Rahmen der Wirkungsanalyse wird das Ausmaß der möglichen Erosion festgelegt. Es wird 

durch die Erosionsbreite, Erosionshöhe und Erosionslänge beschrieben (Abbildung 3). Die 

Erosionslänge L UE 

wird entlang der ursprünglichen Uferlinie gemessen. Als Erosionsbreite B UE 

wird die maximale Ausdehnung der Erosionsnische senkrecht zur Fließrichtung betrachtet. 

Die Erosionshöhe h UE 

bezeichnet den Höhenunterschied zwischen der ursprünglichen 

Gewässersohle und der Anrisskante. Die Erosionshöhe wird verwendet, um die Intensität des 

Gefahrenprozesses zu beschreiben. 

Die Erfahrungen aus dem Hochwasser 2005 und aus jüngeren Hochwasserereignissen lehren, 

dass die Erosionsbreite ein Mehrfaches der ursprünglichen Gerinnebreite aufweisen kann 

(Hunzinger und Durrer 2008, Krapesch et al. 2011, Bachmann 2012,). Verändert sich die 

Morphologie eines Gewässers großräumig, d.h. bilden sich neue Mäander oder Gerinneverzweigungen, 

haben Ufererosionen eine größere Ausdehnung als wenn sie durch eine lokale 

Unregelmäßigkeit im Abflussquerschnitt verursacht werden. 

Werden die Ergebnisse der Beurteilung für eine Risikoanalyse verwendet, wird der Ufererosion 

eine räumliche Auftretenswahrscheinlichkeit zugewiesen, welche zwischen 0.1 (in 

geraden Gerinnen) und 1.0 (in Außenkurven eines Mäanderbogens) liegt. Schließlich darf 

man nicht vergessen, dass durch die Erosion eines Ufers Folgeprozesse ausgelöst werden 


Abbildung 3: Bezeichnungen für das Ausmaß der Erosion. Beispiel an der Theiss, Ukraine. Foto: Flussbau AG SAH, 2001. 

können, wie z.B eine Hanginstabilität. Als letzter Schritt der Beurteilung müssen deshalb 

mögliche Folgeprozesse identifiziert werden. 

EIN ANWENDUNGSBEISPIEL 

Das folgende Anwendungsbeispiel soll die oben erwähnte Empfehlung zum Vorgehen veranschaulichen. 

Es werden Erosionsprozesse bei einem 300-jährlichen Hochwasserereignis am 

Ticino bei Chiggiogna, einem Gebirgsfluss in der Südschweiz, untersucht. Das Flussbett ist im 

Abbildung 4: Blocksatz am Ticino bei Chiggiogna. Foto: Flussbau AG SAH, 2012. 


untersuchten Abschnitt zwischen 40 m und 80 m breit und verzweigt. Das Längengefälle 

beträgt 1.7 %. Das Ufer ist mit einem intakten Blocksatz gesichert (Abbildung 4). 

Das betrachtete Grundszenario hat eine Abflussspitze von 610 m 3 /s und es werden rund 

60'000 m 3 Geschiebe zugeführt. Der Transport von Schwemmholz ist in diesem Beispiel für 

die Beurteilung der Ufererosion ohne Bedeutung. 

Die Schwachstellenanalyse zeigt, dass die Gefährdungsbilder Erosion am Böschungsfuß 

(Abbildung 2a) und direkter Strömungsangriff (Abbildung 2b) näher untersucht werden 

müssen: Für das 300-jährliche Ereignis ist mit großräumigen Ablagerungen auf der Sohle zu 

rechnen und das Gerinne wird sich stärker verzweigen. Die möglichen Kolke werden mit 

dem Ansatz von Zarn (1997) berechnet. Sie sind tiefer, als das Ufer fundiert ist. Für das 

Gefährdungsbild direkter Strömungsangriff wird die Schleppspannung auf der Böschung 

berechnet. Sie berücksichtigt eine Strömungskonzentration an der Kurvenaußenseite. Die 

Grenzschleppspannung wird nach Stevens und Simons (1971) bestimmt. Sie ist dank des 

hohen Blockgewichtes höher als die Belastung (Tabelle 1). Das Gefährdungsbild Erosion an 

der Böschungsoberkante ist wenig wahrscheinlich, weil auf dem Abschnitt bei Hochwasser 

keine Ausuferungen erwartet werden. Das Fazit der Schwachstellenanalyse lautet: durch 

Erosion am Böschungsfuß kann bei einem 300-jährichen Hochwasserereignis das Ufer erodieren 

(Tabelle 1). 

1 

Tabelle 1: Schwachstellenanalyse für das Anwendungsbeispiel Ticino bei Chiggiogna: Gefährdungsbilder, Belastung, Widerstand und 

Fazit. 

2 

3 

4 

5 

Gefährdungsbild Belastung Widerstand Fazit 

Erosion am Böschungsfuß Kolktiefe t = –2.5 m Fundationstiefe -1.5 m Seitenerosion möglich 

Direkter Strömungsangriff τ = 370 N/m 2 Blocksatz, 45° geneigt, Keine Seitenerosion 

Blockgewicht 4 t 

τ crit = 390 N/m 2 

Erosion an der 

keine – Keine Seitenerosion 

Böschungsoberkante 

Im Tabelle Rahmen 1: Schwachstellenanalyse der Wirkungsanalyse wird für postuliert, das Anwendungsbeispiel: dass einem 300-jährlichen Gefährdungsbilder, Ereignis das 

Gewässer Belastung, die Widerstand gesamte Talebene und Fazit. beansprucht, weil das Terrain hinter dem Blocksatz aus leicht 

erodierbaren alluvialen Schottern aufgebaut ist. Die Erosion kann auf der gesamten Länge des 

betrachteten Abschnittes auftreten. Die Erosionshöhe entspricht der Böschungshöhe und 

misst zwischen 2.5 m bis 3 m. Die räumliche Auftretenswahrscheinlichkeit wird zu 0.25 

gesetzt, d.h. es wird angenommen, dass sich wegen der Verzweigung lokale Prallhangsituationen 

und Abschnitte ohne Ufererosion abwechseln. Auf dem untersuchten Abschnitt werden 

keine Folgeprozesse der Erosion erwartet. 


Zum Vergleich: Auf einem Abschnitt 5 km oberstrom des Untersuchungsabschnittes wurden 

während des Hochwassers von 1987 mit einer Abflussspitze um 450 m 3 /s Seitenerosionsprozesse 

mit einer Erosionsbreite von 10 - 20 m beobachtet (Ufficio dei corsi d'acqua, 1987). Auf 

dem betreffenden Abschnitt war das Gerinne aber nicht verzweigt. 

FAZIT 

Die Belastung durch ein Hochwasser und der Erosionswiderstand einer Uferböschung lassen 

sich nicht immer oder nur mit großem Aufwand quantitativ erfassen. Oftmals ist man auf 

eine gutachterliche Schätzung angewiesen. Die Beurteilung der Gefahr der Seitenerosion 

erfordert deshalb große Erfahrung des Gutachters. Wie jede andere Gefahrenbeurteilung ist 

auch die Beurteilung der Ufererosion mit Unsicherheiten verbunden. Unsicherheiten können 

insbesondere im Rahmen der Schwachstellenanalyse bei der Bestimmung von Belastungsgrößen 

und bei der Bestimmung des Erosionswiderstandes auftreten. Aus den Unsicherheiten 

resultiert eine Unschärfe im Ergebnis der Beurteilung. Ein Mittel, um diese Unschärfe zu 

erkennen und möglicherweise zu vermindern, ist die Anwendung verschiedener Methoden 

bei der Bearbeitung der einzelnen Schritte. In manchen Fällen kann die Unschärfe in der 

Beurteilung vermindert werden, indem Unsicherheiten in den Grundlagen, insbesondere 

jene zur Beschaffenheit des Ufers, durch detaillierte Abklärungen beseitigt werden. Dabei gilt 

es aber, die Balance zwischen dem erforderlichen Mehraufwand und der Verbesserung der 

Aussagekraft zu wahren. 

DANK 

Die Erarbeitung der Empfehlung wurde durch das Bundesamt für Umwelt (Schweiz) 

finanziert. Eine erweiterte Arbeitsgruppe aus Mitgliedern von FAN und KOHS hat die 

Erarbeitung der Empfehlung fachlich begleitet. Ihnen gilt der besondere Dank der Autorenschaft. 

LITERATUR 

- Ahmari H. & da Silva A.M.F. (2011). Regions of bars, meandering and braiding in da Silva 

and Yalin's plan. Journal of Hydraulic Research. Vol. 49, No. 6: 718-727. 

- Anderson A.G., Parker G. & Wood A. (1975). The flow and stability characteristics of alluvial 

river channels. St. Anthony Falls Hydraulic Laboratory. Rep. No. 161 

- Bachmann A. (2012). Ausmass und Auftreten von Seitenerosionen bei Hochwasserereignissen. 

Geographisches Institut der Universität Bern, Bern. 

- Beck J.R. (2006). Streambank erosion hazard mapping: concepts, methodology and 

application on the Venoge river (Switzerland). Ecole polytechnique fédérale de Lausanne, 

Lausanne. 

- Hilker N., Aller D. & Hegg C. (2007): Schäden. in Bezzola G. R. & Hegg C. (Ed.) 2007: 

Ereignisanalyse Hochwasser 2005, Teil 1 – Prozesse, Schäden und erste Einordnung. 

Bundesamt für Umwelt BAFU, Eidgenössische Forschungsanstalt WSL. Umwelt- Wissen Nr. 

0707. 215 S. 


- FAN & KOHS (2015). Empfehlung zur Beurteilung der Gefahr von Ufererosion an Fliessgewässern. 

www.fan-info.ch bzw. www.swv.ch. 

- Hunzinger L. & Durrer S. (2008): Seitenerosion. in Bezzola G.R. & Hegg C. (Ed.) 2008. 

Ereignisanalyse Hochwasser 2005, Teil 2 – Analyse von Prozessen, Massnahmen und 

Gefahrengrundlagen. Bundesamt für Umwelt BAFU, Eidgenössische Forschungsanstalt für 

Wald, Schnee und Landschaft WSL. Umwelt-Wissen Nr. 0825: 429 S. 

- Kikkawa H., Ikeda S. & Kitagawa A. (1976). Flow and Bed Topography in Curved Open 

Channels. Proc. ASCE, J. of Hydr. Div. 102(HY9): 1327-1342. 

- Krapesch G., Hauer C. & Habersack H. (2011). Scale orientated analysis of river width 

changes due to extreme flood hazards. Natural Hazards and Earth System Sciences 11: 

2137-2147. 

- Pasche E. (2000). Wasserbau. Skript zur Vorlesung, Technische Universität Hamburg. 

- Peter W. (1986). Kurvenkolk - Untersuchung über die Sohlenausbildung in Flusskrümmungen. 

Mitteilung Nr. 85 der Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie der 

ETH Zürich. 

- Requena R. (2008). Seitenerosion in kiesführenden Flüssen – Prozessverständnis und 

quantitative Beschreibung. Mitteilung Nr. 210 der Versuchsanstalt für Wasserbau, Hydrologie 

und Glaziologie der ETH Zürich. 

- Romang H. (Ed.) 2008: Wirkung von Schutzmassnahmen. Teil F Flüsse. Nationale Plattform 

für Naturgefahren PLANAT, Bern. 289 S. 

- Stevens M.A. & Simons D.B. (1971). Stability Analysis for Coarse Granular Material on 

Slopes. River Mechanics, Shen H.W. (ed.), Fort Collins, Colorado, 17-1: 17-27. 

- Tschopp J. & Bisaz E. (1972). Profundidad de erosión al pie de un vertedero para la 

aplicación de corrección de arroyos en quebradas empinados (Erosion depth at weir toe for 

the application of river corrections at steep canyons). 5. Congreso Latinoamericano de 

Hidráulica, IAHR, Lima, PE. 

- Ufficio dei corsi d'acqua (1987). Alluvione Agosto 1987 Leventina + Bedretto, Fotografie. 

Repubblica e Cantone Ticino, Uffico dei corsi d'acqua. Unveröffentlicht. 

- Zarn B. (1997). Einfluss der Flussbettbreite auf die Wechselwirkung zwischen Abfluss, 

Morphologie und Geschiebetransportkapazität. Mitteilung der Versuchsanstalt für Wasserbau, 

Hydrologie und Glaziologie der ETH Zürich, 154 S. 



Human induced risk dynamics - a quantitative analysis 

of debris flow risks in Sörenberg, Switzerland 

(1950 to 2014) 

Benjamin Fischer, MSc 1,2 ; Margreth Keiler, PD Dr. 1 

ABSTRACT 

Settlement extension into endangered areas has led to increased losses through natural 

disasters in recent years. Despite the significant role of human activity for the development of 

losses, only few studies focus on the quantitative evolution of natural hazard risk over time. 

In this study, a quantitative multi-temporal risk approach is applied to analyse the debris flow 

risk evolution from 1950 to 2014 in Sörenberg, Switzerland. Three hazard scenarios are 

modelled with RAMMS debris flow 1.6.20. The analysis of elements at risk focuses on 

physical economic damage to building structures while the vulnerability is calculated based 

on the empirical vulnerability curve by Papathoma-Köhle et al. (2012). The results show that 

a massive building boom caused a risk increase between factor 41.1 and 65.6 from 1950 to 

2000. The implementation of structural mitigation measures in 2014 reduced the risk in all 

scenarios but the risk of scenario C was still 14-times higher compared to the risk in 1950. 

KEYWORDS 

risk; risk evolution; RAMMS debris flow; vulnerability; vulnerability curve 

INTRODUCTION 

The losses related to natural disasters considerably increased worldwide within the last 

decades. In recent literature it is widely accepted that human activity plays a key role for this 

development (e.g. Fuchs & Keiler 2013). This induced that the concept of risk has become the 

common approach to assess the impact of natural hazards on settlement areas (Fuchs et al. 

2004). Fuchs & Keiler (2013) emphasized that every risk parameter shows its own dynamics 

in time and space with increasing complexity between the different parameters. However, 

only few studies exist which quantify the risk evolution over a longer period of time (e.g. 

Keiler et al. 2006, Schwendtner et al. 2013, Kallen 2015). Thus, the objective of this study is 

to analyse quantitatively the debris flow risk evolution of Sörenberg, Switzerland, from 1950 

to 2014. 

The case study site is a small tourist resort in the Swiss Prealps (Fig. 1). The boom of winter 

sports boosted the touristic development in Sörenberg and caused a building boom starting in 


2 wasser/schnee/lawinen Ingenieurbüro A. Burkard AG, Brig-Glis, SWITZERLAND, b.fischer@wasserschneelawinen.ch 



the 1960s in the settlements Laui and Flüehütte, which are located on an ancient debris fan. 

The slopes above the settlements belong to a well-documented deep-seated sagging process 

(red in Fig. 1) which affects the three torrents Satzgraben, Lauigraben and Lauibach. The 

geology in the study area is Schlierenflysch. Six landslide events with subsequent debris flows 

occurred in the 20th century, whereas the last event dates from 14 May 1999 and occurred in 

the Lauibach (Zimmermann 2006). Since then, extensive mitigation measures were taken 

including a contingency plan and structural measures with two debris collectors in the 

Satzgraben, a debris collector in the Lauigraben and protection dams in all three torrents 

which were completed in autumn 2014 (Fig. 1; Fischer 2014). 

Figure 1: Study area with mitigation measures (finished in 2014) in Sörenberg. 


In terms of natural hazards, risk can be defined as R i,j 

= f (p Si 

, A Oj 

, v Oj,Si 

, p Oj,Si 

). Risk is a 

function of the probability of occurrence of the hazard scenario i (p Si 

), the value of object j at 

risk (A oj 

), the vulnerability of object j in dependence on scenario i (v Oj,Si 

) and of the probability 

of exposure of object j in scenario i (p Oj,Si 

) (Fuchs & Keiler 2013). In this study, vulnerability is 

defined according to Fell et al. (2008: 86):„The degree of loss to a given element or set of 

elements within the area affected by the landslide. It is expressed on a scale of 0 (no loss) to 1 

(total loss). For property, the loss will be the value of the damage relative to the value of the 

property.” 

METHODS 

Similar to recently conducted risk evolution analyses (e.g. Schwendtner et al. 2013), a 

quantitative multi-temporal risk approach is applied which consists of four work steps: the 

hazard analysis, the analysis of elements at risk, the vulnerability analysis and the risk 

calculation. 

Hazard analysis 

The hazard analysis aims to define event scenarios of the three torrents and generate 

quantitative intensity maps based on modelling with RAMMS debris flow 1.6.20 (Christen et 

al. 2012). In this study, the hazard scenarios are assumed to be constant over time. A pair of 

scenarios including a scenario without mitigation measures (situation from 1950 to 2000, 

scenarios 1a-3a) and a scenario with implemented mitigation measures (situation since 2014, 

scenarios 1b-3b) is defined for each torrent. They are defined according to the bed load 

scenario with a recurrence interval of 100 years, derived from the official hazard map. While 

the calibration of RAMMS in the Lauibach is based on the event analysis from the event in 

1999, there is no recent debris flow event, which would reflect the current hazard situation 

in the other two torrents. Thus, the model calibration in these cases are based on previous 

modelling, the semi-quantitative intensity maps for flood hazards and the described characteristics 

of the debris flow hazard in these torrents (Fischer 2014). The modelling is carried 

out with a digital terrain model with a resolution of 2 m. The input parameters of every event 

scenario including the modelled debris flow volume, the applied friction parameters μ and ξ 

and the maximum flow discharge Q max 

are presented in Tab. 1. The chronological sequence of 

the hazard scenarios 1a/1b, 2a/2b and 3a/3b results in the hazard evolution scenarios A-C. 

The modelling results have been verified on-site. 

Analysis of elements at risk 

The analysis of elements at risk focuses on physical economic damage to building structures 

based on data from the cantonal building insurance of Lucerne. Seven elements at risk layers 

were generated which represent the situation of the settlement in the years 1950, 1960, 

1970, 1980, 1990, 2000 and 2014. As a simplification of the method, the insurance values 


from 2014 are considered as stable for the entire investigation period. Consequently, the 

results of the different decades are inflation-adjusted and can be directly compared. 

Table 1: Applied modelling parameters in RAMMS debris flow 1.6.20 for the event scenarios 1a-3a and 1b-3b (Fischer 2014). 

Event 

scenario 

Scenario 1a Scenario 2a Scenario 3a Scenario 1b Scenario 2b Scenario 3b 

Torrent 

Satzgraben 

Lauigraben 

Satzgraben 

Lauibach Satzgraben Lauigraben Lauibach 

Mitigation 

measures 

(m 3 ) 

no no no no 26.000 34.500 40.000 

Debris flow 

volume 

(m 3 ) 

25.000 25.000 10.000 100.000 0 0 60'000 

µ / ξ 0.16 / 200 

0.25 / 

200 

0.16 / 

200 

0.3 / 100 - - 0.3 / 100 

Q max 80 50 20 50 - - 50 

Vulnerability analysis 

Papathoma-Köhle et al. (2012) provides a methodology for the development of a site-specific 

quantitative vulnerability curve. The methodology cannot be applied due to insufficient data 

on damaged buildings in Sörenberg. The vulnerability analysis is thus conducted quantitatively 

according to the empirical vulnerability function by Papathoma-Köhle et al. (2012) which 

was developed based on several case studies in the South Tyrol, Italy. This vulnerability curve 

describes the vulnerability as the ratio of the intensity (I) expressed as deposition height to 

the degree of loss (Fig. 2). 

Risk calculation 

As the three scenarios have a recurrence interval of 100 years, the potential loss of each 

scenario is divided by 100 to calculate the risk of each scenario expressed in CHF/a. 

RESULTS 

Hazard analysis 

The modelling of the event scenarios resulted in six intensity maps, which are used as a basis 

for the multi-temporal risk analysis. Fig. 3 shows the modelled intensity maps in the Lauibach 

before (scenario 3a, on the left) and after the implementation of mitigation measures 

(scenario 3b, on the right). The protection dams and debris collectors diminish the process 

intensities in scenario 3b, but according to the model results many parts of the settlements are 

still affected because the protection dams only retain 60’000 m 3 of 100’000 m 3 in the 


1 

0.9 

0.8 

0.7 

Degree of loss 

0.6 

0.5 

0.4 

0.3 

0.2 

0.1 

0 

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 

Intensity [m] 

Figure 2: Empirical vulnerability curve by Papathoma-Köhle et al. (2012). 

Figure 3: The hazard situation in the Lauibach before (scenario 3a; left) and after the implementation of mitigation measures (scenario 

3b; right), modelled with RAMMS debris flow 1.6.20 (Christen et al. 2012). 


Lauibach (Fischer 2014). The measures are more effective in the scenarios 1b and 2b, where 

they prevent the debris flows from reaching the settlement area. 

Evolution of the elements at risk 

Driven by a building boom with over 100 new buildings in each decade from the 1960s to the 

1980s, the values at risk considerably increased from less than 3 million CHF in 1950 to a 

range between 59.7 million (scenario A) and 188.6 million (scenario C) in 2000. This 

corresponds to a proportional development of factor 45.3 to 142.5. The implemented 

structural measures cut the values at risk to 0 in the scenarios A and B. In scenario C, the 

values at risk only decreased to 162.2 million CHF in 2014 because the extent of the affected 

area was slightly reduced and 20 new buildings were constructed between 2000 and 2014. 

Evolution of the vulnerability parameter 

The mean vulnerability per building increased from average values between 0.029 (standard 

deviation: 0.038; scenario A) and 0.034 (standard deviation: 0.050; scenario C) in 1960 to 

values between 0.043 (standard deviation: 0.058; scenario C) and 0.074 (standard deviation: 

0.098; scenario A) in 2000. Scenario A presents the most distinct proportional development 

with factor 2.6. In the last time step, the average vulnerability in scenario C is reduced to 

0.015. This implies that the average vulnerability decreased by 56 % (equal to factor 0.44) in 

scenario C from 1960 to 2014 while it drops to 0 in scenario 1 and 2. 

Risk evolution 

Obviously, the risk situation has changed substantially in the study area within the investigated 

time period from 1950 to 2014. Tab. 2 presents the proportional risk evolution of the 

hazard evolution scenarios A-C and event scenarios 1a-3a (without mitigation measures), 

whereas the oldest value of 1950 was set to factor 1. 

Table 2: Proportional development of risk in the scenarios A-C (including mitigation measures in 2014) and 1a-3a (excluding mitigation 

measures) in Sörenberg, with the values from 1950 serving as basis (Fischer 2014, © Geoinformation Kanton Luzern). 

Decade Scenario A Scenario B Scenario C Scenario 1a Scenario 2a Scenario 3a 

1950 1,0 1,0 1,0 1,0 1,0 1,0 

1960 2,7 3,4 2,1 2,7 3,4 2,1 

1970 17,1 22,5 11,8 17,1 22,5 11,8 

1980 26,7 32,6 20,8 26,7 32,6 20,8 

1990 59,8 47,7 37,3 59,8 47,7 37,3 

2000 65,6 49,9 41,1 65,6 49,9 41,1 

2014 0,0 0,0 14,3 65,8 52,7 50,0 


The risk evolution from 1950 to 2000 was dominated by the building boom since the 1960s 

which led to a proportional risk increase of factor 41.1 to factor 65.6. In monetary values, the 

risk reached 39’000 CHF/a in scenario A in 2000 and 92’700 CHF/a in scenario C in the same 

time step. While the risk would have further increased without the implementation of 

mitigation measures in the scenarios 1a-3a it was reduced to 0 in the scenarios A and B due 

to structural measures. In scenario C, it still shows an increase of factor 14 compared to the 

situation of 1950. 

DISCUSSION 

The calculated risk evolution depends on the applied methodology and on the definition of 

hazard scenarios and their probabilities, which feature several uncertainties. The event 

history shows that debris flow events always took place after an increased activity of the 

sagging mass and the occurrence of a landslide. Debris flows are thus tertiary processes and 

their volumes depend on how much slided material is remobilized. In the event of 1999, a 

landslide with 250’000 m 3 was released from the sagging mass whereas only 50’000 m 3 were 

remobilized by debris flows (Zimmermann 2006). Another uncertainty arises from the 

simplification of the probability of the hazard scenarios, which are treated as constant within 

the investigated time period. It should be further considered that the difference between the 

maximum flow height – the output of the modelling – and the deposition height, which is 

used as a parameter in the vulnerability curve, is neglected in this study. 

The evolution of the elements at risk in Sörenberg is very pronounced compared to the risk 

evolution analyses in Martell (factor 3.3; Schwendtner et al. 2013), and Galtür (factor 5; 

Keiler et al. 2006). In Sörenberg, the proportional development factors from 1950 to 2000 are 

thus 9 to 47 times higher than those in similar studies. However, building renovations (e.g. 

change of the building structure material) and changes of the building sizes, which were 

important factors for increasing values at risk in Martell and Galtür, were neglected in this 

case study. 

The mean vulnerability values in this study are very low. This is the consequence of the use 

of unprocessed modelling results, which enhances the comparability of the methodology. 

Although the empirical vulnerability curve by Papathoma-Köhle et al. (2012) has the 

advantage that it prevents inaccuracies due to defined vulnerability values for wide intensity 

ranges, it faces major methodological constraints. The vulnerability curve does not differ 

between different building structure types, which may cause inaccurate results for specific 

buildings and which would be important for the analysis of the vulnerability evolution over 

time. 

The risk evolution of scenario C indicates that the implementation of structural measures do 

not implicitly lead to a risk decrease over time, which is consistent with recent literature 

(Keiler et al. 2006, Schwendtner et al. 2013). Compared to other debris flow risk evolution 


analyses by Schwendtner et al. (2013) and Kallen (2015), which showed a risk increase of 

factor 5.8 between 1954 and 2006 in Martell, Italy, and an increase of factor 2.6 to 3.9 in 

Reichenbach, Switzerland, from 1890 to 2010, respectively, the analysed risk evolution in 

Sörenberg is however disproportionately fast and pronounced. Other case studies (e.g. Galtür 

(Keiler et al. 2006) or Davos (Fuchs et al. 2004)) showed a risk increase in the years after the 

implementation of the mitigation measures. This rebound effect of risk is the consequence of 

an improved hazard situation that encourages a more intensive settlement extension, which 

again increases the risk. The probability of a rebound effect is higher if there are considerable 

uncertainties in the definition of the hazard scenarios. As this is the case in Sörenberg, further 

investigations are necessary at a later date. 

CONCLUSION 

The quantitative risk evolution analysis in Sörenberg indicates the seriousness of the risk 

increase due to settlement extension into an active process area. A further investigation of the 

long-term effect of the mitigation measures on the risk evolution in Sörenberg will be 

necessary. Additionally, the study indicates methodological challenges as the development of 

specific vulnerability curves for different building structure types and the consideration of the 

variability of the hazard parameter into risk evolution analyses. 


Special thanks to the two anonymous reviewers whose comments improved the quality of 

the paper significantly. 

REFERENCES 

- Christen, M., Bühler, Y., Bartelt, P., Leine, R., Glover, J., Schweizer, A., Graf, C., McArdell, 

B.W., Gerber, W., Deubelbeiss, Y., Feistl, T., Volkwein, A., 2012. Integral hazard management 

using a unified software environment - Numerical simulation tool “RAMMS” for gravitational 

natural hazards. 12th Congr. Interpraevent 2012 - Grenoble. 

- Fell, R., Corominas, J., Bonnard, C., Cascini, L., Leroi, E., Savage, W.Z., 2008. Guidelines for 

landslide susceptibility, hazard and risk zoning for land use planning. Eng. Geol. 102 (2008), 

85–98. 

- Fischer, B., 2014. Risk evolution in debris flow prone areas from 1950 - 2014 – two case 

studies conducted in Lai-Ji (), Taiwan, and Sörenberg, Switzerland (Master thesis). 

Universität Bern. 

- Fuchs, S., Bründl, M., Stötter, J., 2004. Development of avalanche risk between 1950 and 

2000 in the municipality of Davos, Switzerland. Nat. Hazards Earth Syst. Sci. 4 (2), 263–275. 

- Fuchs, S., Keiler, M., 2013. Space and time: coupling dimensions in natural hazard risk 

management?, in: Müller-Mahn, D. (Ed.), The Spatial Dimension of Risk - How Geography 

Shapes the Emergence of Riskscapes. Earthscan publishers, London, pp. 189–201. 


- Kallen, I., 2015. Risikoentwicklung im Schatten des Geschiebesammlers - Die Veränderung 

des Risikos über die Zeit des Geschiebesammlerbaus im Richebach in Reichenbach i.K. bis 

heute (Masterarbeit). Universität Bern, Bern. 

- Keiler, M., Sailer, R., Jörg, P., Weber, C., Fuchs, S., Zischg, A., Sauermoser, S., 2006. 

Avalanche risk assessment - a multi-temporal approach, results from Galtür, Austria. Nat. 

Hazards Earth Syst. Sci. 6 (2006), 637–651. 

- Papathoma-Köhle, M., Totschnig, R., Keiler, M., Glade, T., 2012. A new vulnerability 

function for debris flow - the importance of physical vulnerability assessment in alpine areas. 

12th Congr. Interpraevent 2012 - Grenoble. 

- Schwendtner, B., Papathoma-Köhle, M., Glade, T., 2013. Risk evolution: how can changes 

in the built environment influence the potential loss of natural hazards? Nat. Hazards Earth 

Syst. Sci. 13, 2195–2207. 

- Zimmermann, M., 2006. Die Rutschung Sörenberg - nicht nur ein Studienobjekt, in: 

Beiträge Zur Angewandten Geologie - Festband Jean F. Schneider, Schriftenreihe des 

Depatments. Wien, pp. 119–130. 



Maps of pluvial floods and their consequences: 

a case study 

Martin Mergili 1 ; Andreas Tader 1 ; Thomas Glade 1 ; Stefan Jäger 2 ; Clemens Neuhold 3 ; Heinz Stiefelmeyer 3 

ABSTRACT 

GIS based methods represent a helpful tool set for analyzing pluvial floods and their consequences. 

Focusing on the consequences for buildings, we use a two-step procedure to 

exemplify such an analysis with a well-documented event in the city of Graz (Austria) and to 

highlight possible methodological pitfalls and limitations. (i) We compute spatially distributed 

inundation depths using the software FloodArea. (ii) Based on inundation depths and a set of 

rules and functions, we derive the exposure, vulnerability and risk for each building. (iii) We 

scale the results to the level of postal code zones. The official cadastre is used as input in 

combination with the building register, a DEM and soil and land cover data. Verification is 

based on loss reports, photos and videos. We demonstrate a certain potential of the suggested 

procedure to reproduce the documented damages at the level of postal code zones. However, 

the results are highly sensitive to the model assumptions and parameter settings, and a 

satisfactory back-calculation of even well-documented events remains a major challenge. 

KEYWORDS 

pluvial floods; flood routing; risk analysis; scaling 

INTRODUCTION 

Short but intense rainfall events often trigger heavy short-time flooding causing severe 

damage to residential housing, roads, critical infrastructures, agricultural lands and other 

types of private and public assets. These damages are not only related to assets concentrated 

in flood plain areas, but also to those located in hilly terrain, thus where water is running 

along pathways such as slope cuts or through villages along roads. Therefore, besides the 

generally well established frameworks for fluvial floods, much more attention has to be paid 

to pluvial floods (Maksimović et al., 2009; Henonin et al., 2013), which should not be 

confused with flash floods. 

Every year society suffers from financial losses according to such pluvial flooding (Zhou et al., 

2012, Smith et al., 2001 and Richard, 1995). The choice of appropriate adaptation and 

mitigation strategies often relies of GIS-supported analyses of expected inundation depths and 

consequences thereof. Such models have been used increasingly in the previous years 

(Leandro et al., 2009; Henonin et al., 2013). The effects of various key parameters on the 

1 Department of Geography and Regional Research, University of Vienna, AUSTRIA, martin.mergili@univie.ac.at 

2 geomer GmbH, Heidelberg, GERMANY 

3 Austrian Federal Ministry of Agriculture, Forestry, Environment and Water, Vienna, AUSTRIA 



model results were explored. Examples of such parameters include infiltration and surface 

roughness (Aronica et al., 1998; Singh et al., 2002; Casa et al., 2006; Choromanski et al., 

2008; Sangati et al., 2009), and the influence of sewer systems (Masimović et al., 2009; Saul 

et al., 2010; Sun, 2011). 

The present paper focuses on the GIS based back-calculation of a well-documented pluvial 

flood event in the city of Graz (Austria). The aim is to: 

– propose a consistent and flexible work flow to derive the inundation depth and its possible 

consequences for buildings at varying levels of spatial aggregation, to 

– identify the possibilities, but also the challenges related to such an approach, and to 

– highlight the sensitivity of the analysis to some key parameters. 

STUDY AREA 

Graz is the capital of the province of Styria. With a population of approx. 270,000 it represents 

the second largest metropolitan area in Austria. In the present study we consider the 

city of Graz (Austria) along with some neighbouring municipalities. The area surveyed covers 

302 km², ranging between 321 m and 1339 m a.s.l. (Figure 1). Four control points for 

evaluation of the modelled inundation depths are chosen according to the availability of 

reference data (photo and video recordings during the event). 

Figure 1: The study area. The green dots show the control points evaluated in detail in Table 3. 


The annual rainfall for Graz in the period 1971–2010 is on average 819 mm, for the month of 

May it averaged at 86 mm (Provincial Government of Styria, 2015). In May 2013, 219 mm of 

precipitation were measured. An event between May 6, 2013 at 8 pm and 8 am the next day 

with a total amount of 82.3 mm is considered in the present study. The recurrence interval 

for this event ranges between five and ten years (Austrian Federal Ministry of Agriculture, 

Forestry, Environment and Water, 2015). 

METHODS 

General work flow 

An integrated approach to compute inundation depths H (cm) and their possible consequences 

is applied. H is derived through flow routing and applied to each building in the study area 

in order to derive the vulnerability V (-) through a vulnerability function. The risk R 

(expressed as the expected cost of reconstruction in €) is computed by combining V and the 

associated building values (exposure E in €). The results are scaled to postal code zones and 

evaluated against point observations of inundation depths and the documentation of 

damages. 

Inundation depth 

We use the flood routing software FloodArea developed by Geomer to compute spatially 

distributed values of H from an input pluviograph, a digital elevation model (DEM) and 

further derivatives of the data listed in Table 1. A detailed description of the software 

FloodArea is provided by Rodriguez et al. (2003); Assmann et. al. (2007) and Geomer & Ruiz 

Rodriguez+Zeisler+Blank (2014). 

A set of nine model runs is used to quantify the effects of every single determinant on the 

simulation results with regard to H (Table 2). For all model runs, the amount of the twelve 

hours rainfall event (82.3 mm) is evenly distributed over twelve equal time steps, each of 

them representing one hour with an amount of approx. 6.9 mm. The surface roughness is 

defined on the basis of the land use classes provided by the national digital cadastral map for 

Austria (Federal Office of Metrology and Surveying, 2012; Table 1) and Manning-Strickler 

roughness coefficients associated to each land use class (Fritsch, 2011; Rössert, 1996). 

A 2 m DEM is used as reference. As the pixel size strongly influences the computational 

times, we repeat the computation with a pixel size of 4 m (Model runs 1 and 2 in Table 2). 

The present paper only considers pluvial flooding. In order to exclude the influence of 

possible fluvial flooding and potential backwater effects, the stream network is deeply incised 

into the DEM. Fritsch (2011) already pointed out the advantage of this approach. We perform 

this step based on an automatically generated stream network (Model run 3 in Table 2). 

Model run 4 evaluates the influence of infiltration on the inundation depth (Table 2). In the 

model runs 1–3, inundation is neglected (runoff coefficient r = 1). For further improvement, r 


Table 1: Key data used for the analyses. 

Description Source Inundation Consequences Evaluation 

depth 

Pluviograph for Austrian Federal Ministry of ● – – 

the event Agriculture, Forestry, 

Environment and Water 

Digital Elevation Federal Office of Metrology ● – – 

Model (DEM) and Surveying 

Official cadastre Federal Office of Metrology ● ● – 

(DKM) 

and Surveying 

Building register Statistik Austria – ● – 

(GWR) 

CORINE Land Environmental Agency ● – – 

cover 

Austria 

eBod 

Federal Research and ● – – 

Training Centre for Forests, 

Natural Hazards and 

Landscape (BFW) 

Photo 

Fire department of the city of – – ● 

documentation Graz 

Damage Austrian Insurance 

– – ● 

database Association 

Postal code 

zones 

Austrian Post – ● ● 

Table 2: Set of nine model runs. N = not considered; Y = considered; B = before; A = after; I = increased capacity of sewer system. 

Model run Pixel size (m) Automatic 

incision 

Infiltration Manual 

incision 

Sewer 

system 

1 2 N N N N 

2 4 N N N N 

3 4 Y N N N 

4 4 Y Y N N 

5 4 Y Y Y N 

6 4 Y Y Y A 

7 4 Y Y Y B 

8 4 Y Y Y A, I 

9 4 Y Y Y B, I 

in the range 0–1 is introduced, based on soil properties and vegetation cover. Building on the 

code of practice of the Federal Research and Training Centre for Forests, Natural Hazards and 

Landscape (Markart et al., 2011) a rough estimation of r is attempted. For this purpose we 

firstly use the Austrian soil map eBod (Federal Research and Training Centre for Forests, 

Natural Hazards and Landscape, 2013). The only free available release of this data set builds 

on raster of 1x1 km. For those raster cells where no data are available we assume r = 1. 

Secondly, we explore the CORINE land cover map (Environmental Agency Austria, 2012). 

Small water courses not depicted by the incised stream network layer can still cause unrealistically 

high modelled water levels when dammed by bridges or other flow obstacles. Manual 


model run 5 (activation of the sewer system after flood routing), whilst in model run 7 

wer capacity is integrated in r (activation of the sewer system before flood routing). 

order to explore the effects of an increased sewer capacity (as it is available in some o 

deepening of the DEM is necessary at selected locations in order to avoid impoundments 

ties) the capacity of the sewer system is expanded to 73.5 what equals an event with a 

(model run 5). 

turn period of five years (Austrian Federal Ministry of Agriculture, Forestry, Environme 

In urban areas the sewer system represents a key factor for runoff. The capacity of the sewer 

d Water, 2015; model runs 8 and 9). 

system in Graz is designed for the discharge of an event with a return period of one year for 

15 minutes duration with 111 l/ha/s (pers. communication W. Sprung, 03.09.2015). In consequence 

a corresponding sewer discharge for an event of twelve hours with a return period of 

nsequences 

one year is estimated (Austrian Federal Ministry of Agriculture, Forestry, Environment and 

Water, 2015), leading to an assumed sewer capacity of 43.8 l/m². Two approaches are 

compared: in model run 6 the sewer capacity is subtracted from the results of model run 5 

(activation of the sewer system after flood routing), whilst in model run 7 the sewer capacity 

e possible consequences of pluvial flooding are computed separately for each of the n 

is integrated in r (activation of the sewer system before flood routing). 

odel runs summarized in Table 2. 

In order to explore the effects of an increased sewer capacity (as it is available in some other 

cities) the capacity of the sewer system is expanded to 73.5 what equals an event with a 

e extent return of period each of building five years (Austrian is extracted Federal Ministry from the of Agriculture, DKM, the Forestry, surface Environment area of each building 

and Water, 2015; model runs 8 and 9). 

2 ) from the GWR (Table 1). We use the sum of the areas of all floors. As a first estimat 

CONSEQUENCES 

sume a cost of reconstruction E0 of 2,000 €/m² (modified after Austrian Economic 

The possible consequences of pluvial flooding are computed separately for each of the nine 

model runs summarized in Table 2. 

ambers, 2014) independent on the function or building material. E is computed as E0 

The extent of each building is extracted from the DKM, the surface area of each building A 

(m 2 ) from the GWR (Table 1). We use the sum of the areas of all floors. As a first estimate we 

n approximation to the vulnerability function presented by Hsu et al. (2011) is used to 

assume a cost of reconstruction E 0 

of 2,000 €/m² (modified after Austrian Economic Chambers, 

2014) independent on the function or building material. E is computed as E 0 

· A. 

mpute V of each building: 

An approximation to the vulnerability function presented by Hsu et al. (2011) is used to 

compute V of each building: 

Equation 1 

1.5 

I 

0.575 

tan( R 

) 

2 

V 0.8 e 

, Equatio 

here IR is the ratio between the relevant inundation depth HR (m) and the (in our case 

where I R 

is the ratio between the relevant inundation depth H R 

(m) and the (in our case 

estimated) height of the building (m) (relative intensity; Totschnig et al., 2011). H R 

is 

timated) height of the building (m) (relative intensity; Totschnig et al., 2011). HR is 

approximated by the mean value of H derived with the software FloodArea within a buffer 

area of five metres around each building. Finally, R of each building is computed as V · E. 

proximated by the mean value of H derived with the software FloodArea within a buf 

ea of 

SCALING 

five metres around each building. Finally, R of each building is computed as V · E 

An automated procedure building on the Python programming language and the R package 

for Statistical Computing (R Core Team, 2015) is employed for scaling the object-based values 

aling 


of H R 

, E, V and R to any desired spatial unit (large pixels, administrative units, catchments 

etc.). Diagrams displaying for defined threshold levels of each variable the fraction of objects 

equal or above the value of the variable are produced for each spatial unit. Further, the 

number and the fraction of objects with H R 

, E, V or R > 0 are displayed as well as the 

zone-specific averages and (for E and R) sums. Ranges are given for E and R. In the present 

study we scale the results to postal code zones. 

EVALUATION 

To verify the results of the simulation, we qualitatively compare the modelled inundation 

depths to observed inundation depths documented by photographs (Table 1). Further, we use 

damages related to the simulated event reported to the insurance business in order to validate 

the expected cost of recovery at the level of postal code zones. 


The key results for the nine model runs – inundation depths at four selected points, and the 

sums of R for two selected districts – are summarized in Table 3. H displays a certain degree 

of levelling when increasing the pixel size rather than a general change of the results. The 

calculation time for a single model run can be reduced from 6 days to 36 hours. In contrast, 

the incision of the stream channels significantly reduces H, and the values of R drop from 

severe overestimates to fairly realistic estimates. Whilst infiltration appears to exert a rather 

moderate effect on the results, the manual removal of artefacts at selected locations signifi- 

Figure 2: Modelled sums of R and documented damages for the postal code zones of Graz. Left: Sum of R for each zone yielded with 

model run 6; Right: sum of documented damages. The labels of the zones considered in detail in Table 3 are highlighted. 


cantly reduces the inundation depth at point 4 whilst it does not affect R in the central city 

and in the district of Andritz (Figure 1). Assuming the sewer system to alleviate flooding leads 

to significant reductions of both H and R. Particularly with regard to R, this effect dramatically 

increases when (i) activating the system before instead of after the flood routing or (ii) 

increasing the capacity of the sewer system. In summary, the incision of the stream network 

and the activation of the sewer system exert the most significant impacts on H and R for the 

event under investigation. 

Figure 2 illustrates the documented damages and the modelled distribution of R aggregated to 

the level of postal code zones. The summed values of R, as displayed in Figure 2, are 

problematic for model evaluation as they are strongly influenced by the number of objects in 

a given postal code zone, so that a certain level of correspondence with the documented 

damages is likely. Figure 3 therefore shows the average (out of all buildings properly 

registered in the DKM and the GWR) of the documented damages and the modelled 

Figure 3: Modelled averages (out of all buildings properly registered in the DKM and the GWR) of R and documented damages for the 

postal code zones of Graz. (a) Average of R for each zone yielded with model run 6; (b) average of documented damages. Note that also 

those buildings with R = 0 or with no documented damages are included in the averaging procedure. 

distribution of R. Both figures refer to the model run 6 yielding the most realistic results for 

the central city. From a visual comparison we deduce that the model results show some 

reasonable correlation with the documented patterns in general, but that there are a number 

of postal code zones with significant disagreements. 


Table 3: Key results of the nine model runs. H = inundation depth (cm), the subscripts in the column headers indicate the dots shown in 

Figure 1 the numbers refer to. R = risk (expected cost of recovery in million Euro) for the central district of Graz (postal code 8010) and 

Andritz (postal code 8045). The numbers given in brackets in the header refer to the documentation. 

Model 

run 

H 1 (≤6) H 2 (≤7) H 3 (≤12) H 4 (-) R 8010 (0.18) R 8045 

(0.24) 

1 10 184 1.5 854 23.64 79.03 

2 11 105 1.4 770 48.43 44.24 

3 3.5 0.1 0.3 770 0.86 0.04 

4 6.0 0.1 0.5 737 0.80 0.06 

5 6.0 0.1 0.6 193 0.80 0.06 

6 4.7 0.1 0.3 155 0.26 0.01 

7 2.9 0.1 0.3 93 0.01 0.00 

8 2.5 0.1 0.1 119 0.00 0.00 

9 0.6 0.0 0.1 21 0.00 0.00 

Whilst the modelled inundation depth corresponds reasonably well to the depth observed at 

point 1 (except for model run 9 where it is rather underestimated), it is underestimated at 

point 3 and obviously strongly governed by the stream network at point 2 (Table 3). The 

tendency of the model to underestimate H at the points 2 and 3 represents an unwanted local 

effect of the incision of the stream network (both points are located closely to incised 

streams). Further, the incision of the stream network is the likely cause for the underestimate 

of R for Andritz, where large buildings are located closely to a water course, compared to the 

central city (Table 3). Therefore, the usefulness of incising the stream network remains 

controversial. With regard to the modelled inundation depths (Table 3) one may conclude 

that the incision leads to a decreased quality of the results and other strategies have to be 

developed to separate the effects of fluvial and pluvial floods, as far as a clear separation is 

possible at all. A particular challenge consists in dealing with underground sections of water 

courses. In contrast, manual lowering of selected spots definitely leads to the disappearance of 

some artefacts in the model results. 

The sewer system appears to represent a key factor for flooding and its consequences (Ettrich, 

2007; Illgen and Niemann, 2011). It is therefore essential (i) to obtain more information on 

its real capacity to alleviate flooding and (ii) to develop improved generalization strategies to 

include its effects in flood routing algorithms. With regard to soil infiltration, Schumann et al. 

(2007) pointed out that it is certainly possible to reach locally acceptable results by generating 

clustered roughness parameters conditioned on remote sensing. However, the regionalization 

of infiltration remains a critical issue due to the poor spatial resolution of the eBod and 

CORINE data. Due to the possibly limited degree of validity for fine scales, the estimation of 

the runoff coefficient can be considered as a rather coarse approximation to reality only. 

However, surprisingly, infiltration seems not to affect the model results significantly under the 

assumptions taken. 


The interpretation of the modelled values of R requires utmost care. Firstly, not all the 

damages are necessarily reported to the insurance companies. Secondly, also the DKM and 

the GWR are not complete at all, so that the modelled values of R may represent underestimates. 

Strategies to deal with those issues of incompleteness have to be applied in the future. 

Thirdly, due to the exponential character of the function even a moderate misestimation of 

inundation depths may strongly affect the estimate of V and, therefore, R. Fourthly, we use 

the mean inundation depth as H R 

, and not the maximum which might be more appropriate. 

However, using the maximum, few artefacts in the raster map of H may result in very high 

vulnerabilities of selected buildings and cause extreme overestimates of R for the corresponding 

zone. 

Furthermore it has to be considered that the vulnerability function represents a generalization, 

considering some features of the vertical structure of a building, such as a cellar, only in 

an indirect way. Consequently, we build on the assumption that damages occur also at low 

inundation depths, which is often not the case in reality. The estimates of the buildings’ 

values, in contrast, only affect the model results in a linear way and are therefore considered 

less critical. 

Finally it shall be emphasized that the displayed evaluation of R against the documented 

damages (Figs. 2 and 3) allows to draw conclusions at the level of postal code zones only. Due 

to the lacking availability of suitable reference data there is currently no possibility to 

evaluate the model results at the level of objects. 

CONCLUSIONS 

We have shown that the modelled inundation depths of pluvial floods and – in particular 

– their consequences in terms of the expected losses react in a highly sensitive and nonlinear 

way to changes in the model assumptions and parameter settings. Even though the suggested 

procedure shows a certain potential to reproduce the patterns in the documented damages at 

the level of postal code zones, it remains a challenge to back-calculate well-documented 

events in a satisfactory way. More parameter studies as well as the modification of some of 

the model functions and additional validation efforts will be necessary to allow reasonably 

reliable forward calculations of possible future events of defined frequency and magnitude, 

and the costs of such events. In this context it is highly important to provide uncertainty 

estimates associated to the results. 


The support of the Austrian Federal Ministry of Agriculture, Forestry, Environment and 

Water, the Austrian Insurance Association (Dr. Thomas Hlatky), the fire department of the 

city of Graz and the Austrian Post is acknowledged. 


REFERENCES 

- Aronica G., Hankin B., Beven K. (1998). Uncertainty and equifinality in calibrating 

distributed roughness coefficients in a flood propagation model with limited data. Advances 

in Water Resources 22 (4): 349-365. 

- Assmann A., Schroeder M., Hristov M. (2007). High Performance Computing für die 

rasterbasierte Modellierung. Angewandte Geoinformatik: 19–24. 

- Austrian Economic Chambers (2014). Immobilienpreisspiegel. Wien. 

- Austrian Federal Ministry of Agriculture, Forestry, Environment and Water: eHyd, Kennwerte 

und Bemessung, under: http://ehyd.gv.at/# (link called 02.09.2015) 

- Casa A., Benito G., Thorndycraft V.R., Rico M. (2006). The topographic data source of digital 

terrain models as a key element in the accuracy of hydraulic flood modelling. Earth Surface 

Processes and Landforms 31 (4): 444-456. 

- Chormansky J., Van de Voorde T., De Roeck T., Batelaan O., Canters F. (2008). Improving 

Distributed Runoff Prediction in Urbanized Catchments with Remote Sensing based Estimates 

of Impervious Surface Cover. Sensors 8(2): 910-932. 

- Environmental Agency Austria (2012). CORINE Landbedeckung 2012, http://www. 

umweltbundesamt.at/umweltsituation/umweltinfo/opendata/oed_landbedeckung/?cgiproxy_ 

url=http%3A%2F%2Fhttpapp5.umweltbundesamt.at%2Fdata%2Fdataset.jsf%3Bjsessionid%3D8AA5AF9639F75096A56D570685A8F3DE%3Fid%3D29 

(link called 11.09.2015). 

- Ettrich, N. (2007). Hydrodynamische Kanalnetzberechnung – Oberflächenabflussmodellierung 

bei Starkregenereignissen in urbanen Gebieten. DWA Landesverbandstagung 

Baden-WürttembergFrauenhofer-Institut für Techno- und Wirtschaftsmathematik, Kaiserslautern. 

- Federal Office of Metrology and Surveying (2012). Digitale Katastralmappe Österreich 

Stichtagsdaten 10/2012. 

Federal Research and Training Centre for Forests, Natural Hazards and Landscape (2013). 

Bodenkarte eBod, http://bfw.ac.at/rz/bfwcms2.web?dok=7066 (link called 03.09.2015). 

- Fritsch K. (2011). Framework for economic pluvial flood risk assessment considering climate 

change effects and adaptation benefits, am Beispiel von Teileinzugsgebieten an der Murr. 

Diplomarbeit, Institut für Geographie, Technische Universität Dresden. 

- Geomer GmbH / Ruiz Rodriguez+Zeisler+Blank (2014). FloodArea HPC-Desktop Anwenderhandbuch. 

Vers. 10.2. 

- Henonin J., Russo B., Mark O., Gourbesville P. (2013). Real-time urban flood forecasting 

and modelling – a state of the art. Journal of Hydroinformatics 15 (3): 717-736. 

Hsu W.K., Huang P.C., Chang C.C., Chen C.W., Hung D.M., Chiang W.L. (2011). An integrated 

flood risk assessment model for property insurance industry in Taiwan. Natural Hazards 

58(3): 1295–1309. 

- Illgen, M., Niemann, A. (2011). Urbane Überflutungsvorsorge – was die Siedlungsentwässerung 

vom gewässerseitigen Hochwasserschutz lernen kann; 10. Regenwassertage, 10./11. Mai 

2011, Bad Soden. DWA Deutsche Vereinigung für Wasserwirtschaft, Abwasser und Abfall. 


- Leandro J., Chen A., Djordjević S., Savić D. (2009). Comparison of 1D/1D and 1D/2D 

Coupled (Sewer/Surface) Hydraulic Models for Urban Flood Simulation. Journal of Hydraulic 

Engineering 135 (6): 495–504. 

- Maksimović Č., Prodanović D., Boonya-Aroonnets S., Leitão J.P., Djordjević S., Allit R. 

(2009). Overland flow and pathway analysis for modelling of urban pluvial flooding. Journal 

of Hydraulic Research 47 (4): 512-523. 

- Markart G., Kohl B., Sotier B., Klebindert K., Schauert T., Bunza G., Pirkl H., Stern R. 

(2011). A Simple Code of Practice for the Assessment of Surface Runoff Coefficients for 

Alpine Soil-/Vegetation Units in Torrential Rain (Version 2.0). Department of Natural Hazards, 

Federal Research and Training Centre for Forests, Natural Hazards and Landscaper (BFW), 

Innsbruck. 

- Provincial Government of Styria: Klimadaten Graz-Universität (from January 2003 

onwards), under: http://www.statistik.steiermark.at/cms/dokumente/11682476_103034960/53afceaf/Graz-Universit%C3%A4t.pdf 

(link called 02.09.2015). 

- R Core Team (2015). R: A Language and Environment for Statistical Computing. R 

Foundation for Statistical Computing, Vienna, Austria, http://www.R-project.org (link called 

11.09.2015). 

- Richard S.J.T. (1995). The damage costs of climate change toward more comprehensive 

calculations. Inst. For Environmental Studies, Vrije Universiteit Amsterdam. 

- Rodriguez E.R., Zeisler P., Assmann A. (2003). GIS-Einsatz zur Gefahrenabwehr im 

Hochwasserfall. Hochwasserschutz und Katastrophenmanagement. Ernst & Sohn Spezial 4. 

Rössert R. (1996). Hydraulik im Wasserbau. 9. verbesserte Auflage, R. Oldenbourg Verlag 

München, Wien. 

- Sangati M., Borga M., Rabuffetti D., Bechini R. (2009). Influence of rainfall and soil 

properties spatial aggregation on extreme flash flood response modelling: An evaluation 

based on the Sesia river basin, North Western Italy. Advances in Water Resources 31: 

1090-1106. 

- Saul A., Djordjević S., Maksimović Č., Blanksby J. (2010). Integrated Urban Flood Modelling. 

In: Pender G., Faulkner H., (eds.). Flood Risk Science and Management: 258-288. 

Schumann G., Matgen P., Hoffmann L., Hostache R., Pappenberger F., Pfister L. (2007). 

Deriving distributed roughness values from satellite radar data for flood inundation modelling. 

Journal of Hydrology 334 (1-2): 96-111. 

Singh V. and Woolhiser D. (2002). Mathematical Modeling of Watershed Hydrology. Journal 

of Hydrologic Engineering 7 (4): 270-292. 

- Smith J.B., Schellenhuber H.-J., Mirza M.M.Q., Frankhauser S., Leemans R., Lin E., Ogallo 

L., Pittock B., Richels R., Rosenzweig C., Safriel U., Tol R.S.J., Weyant J., Yohe G. (2001). 

Vulnerability to climate change and reasons for concern: a synthesis: 913-967. 

- Sprung W., Holding Graz Services: Pers. Communication 03.09.2015. 

Sun S., Djordjević S., Khu S. (2011). A general framework for flood risk-based storm sewer 

network design. Urban Water Journal 8 (1): 13-27. 


- Totschnig, R., Sedlacek, W., Fuchs, S. (2011). A quantitative vulnerability function for 

fluvial sediment transport. Natural Hazards 58(2): 681-703. 

- Zhou Q., Mikkelsen P.S., Halsnæs K., Arnbjerg-Nielsen K. (2012). Framework for economic 

pluvial flood risk assessment considering climate change effects and adaptation benefits. 

Journal of Hydrology 414-415: 539-549. 



Rockfall Susceptibility Maps in Styria considering the 

protective effect of forest 

Herwig Proske, Mag. 1 ; Christian Bauer, Dr. 2 

ABSTRACT 

The study presents a GIS-based approach aiming at the identification of rockfall propagation 

areas considering the protective effects of forests in the Province of Styria, Austria. An 

empirical approach was selected to identify potential rockfall source zones. Based on a Digital 

Terrain Model derived from LiDAR data, threshold slope values for potential release areas 

were attributed to geotechnical units. The run-out distances were estimated by velocity 

calculation based on a one parameter friction model. The method was applied for the first 

time in such large study area. Friction coefficients were attributed to forestal units based on 

the detailed characterisation of relevant forest parameters derived from laserscanning data. 

Finally, two scenarios based on different friction coefficients were modelled: one considering 

the current forest cover and one assuming no forest cover. The results demonstrate the strong 

protective effect of the forest cover. Furthermore, an indicative hazard map was generated by 

classifying the transition frequencies and kinetic energies of blocks. 

KEYWORDS 

Rockfall; Forest effects; GIS based modelling; Indicative hazard maps; Styria 

INTRODUCTION 

Within the frame of the project “Indicative Natural Hazard Maps in the Province of Styria” 

the applicability of a GIS-based approach aiming at the identification of areas which are 

potentially endangered by rockfall has been analysed. Furthermore, the protective effect of 

forests was taken into account. This work was performed at JOANNEUM RESEARCH, 

Institute for Information and Communication Technologies (Remote Sensing and Geoinformation). 

The study area is the Province of Styria, located in the southeastern part of Austria 

and covering approx. 16.400 km² (Fig. 1). The area is part of the Eastern Alps with peaks 

reaching approx. 3.000 m asl. 

Forests cannot stop the devastating effect of large magnitude rockfall events, but for low 

magnitude rockfall events, forests provide effective protection (Hétu and Gray, 2000). The 

maintenance of forest stands with a protection function is often more cost-effective and more 

sustainable than technical measures (Kienholz and Mani, 1994; Motta and Haudemand, 

2000). However, in many mountainous regions it is not known whether active forest 

management ensures effective protection against rockfall. Therefore, the main aim was to 

1 Joanneum Research Forschungsgesellschaft mbH, Graz, AUSTRIA, herwig.proske@joanneum.at 

2 Institute for Geography and Regional Science, University of Graz, AUSTRIA 



assess the protective functions of the forest cover at a regional scale, as more than 60 % of 

the Province of Styria is covered by forests. 

The results are visualized using the example of the Triebenstein region in Upper Styria (cf. 

Fig. 1). 

Figure 1: Study area. The red dot marks the location of the Triebenstein region from which results are shown in Figures 2, 3 and 4. 

METHODS 

Two main aspects had to be taken into account: (a) the identification of potential source 

zones, and (b) the estimation of rockfall propagation zones. 

(a) Identification of potential rockfall source zones 

An empirical approach was used for the identification of potential source areas. Due to the 

availability of a 1 m - Digital Terrain Model (DTM) based on Airborne Laserscanning (ALS) 

escarpments could be identified with high spatial resolution. With regard to data processing 

capacities, the spatial resolution was reduced to 2 m. 

All lithological units in the study area were assigned to 16 geotechnical units (GTU) according 

to geotechnical properties (e.g. stability of escarpments, bedding and schistosity, weathering 

resistance). Threshold slope values for potential rockfall release areas were attributed to these 

geotechnical units based on field observations and inclinometer measurements as well as 

analysis of remote sensed data (Orthophotos, ALS-DTM). Morphometric parameters (e.g. 

profile curvature) and the tectonic situation (large faults and boundaries of nappes) were 

taken into account by introducing correction factors with a maximum value of 2°. The 


defined threshold slope angles cover values from 43° in tectonized carbonates up to 55° in 

massive granitic and gneissic rocks of the crystalline Central Alps. 

For the simulation, the mean block volumes of each GTU were classified in three volume 

classes representing the most probable event, according to field observations: (1) ≤0,008 m³; 

(2) >0,008 – ≤0,125 m³; (3) >0,125 – 3,375 m³. As the rockfall propagation zones were 

modelled area-wide, one average value for rock density had to be defined and was fixed with 

2,4 g/cm³ according to literature (e.g. Kobranova, 1989; Carmichael, 1984). 

Figure 2 displays the GTU and potential rockfall source areas of the Triebenstein region. 

The geological situation is characterized by rocks belonging to the paleozoic greywacke zone 

(Fig. 2a). The main rockfall source areas (Fig. 2b) are situated within carbonatic rocks, 

building steep cliffs up to 200 m height (cf. Fig. 3). The forested slopes below the cliffs 

(Fig. 3c) are mainly built by phyllites where large areas are covered by talus material. 

Figure 2: GTU (a) and modelled potential rockfall source areas (b) in the Triebenstein region. 

(b) Estimation of rockfall propagation zones 

The runout distances were estimated by velocity calculation based on a one parameter friction 

model (Scheidegger, 1975). With this method, the velocity of a rock particle is calculated 

along a profile line that is divided into a number of triangles. 

Depending on whether or not a rock breaks, 75–86% of the energy gained in the initial fall is 

lost in the first impact on the talus slope (Broilli, 1974; Dorren, 2003). The initial velocity on 

the talus slope is calculated by taking into account a factor K and in this way reducing the 

energy: 


vv ii = √2. gg. h ff . KK 

Formula 1 

where = initial velocity on talus slope [m/s]; = gravitational acceleration (9,81 m/s²); = 

height difference between start point and element I [m]; K = portion of energy which is 

reduced by impact. 

l r o oloa Formula Formula2o 

The further propagation velocity is calculated on the basis of energy conservation of a mass 

that is considered to move over a slope surface as defined by Scheidegger (1975). In this 

energy conservation approach, a friction coefficient is responsible for energy loss. For each 

cell in the falltrack, the velocity of the falling rock is calculated as follows: 

VV oooooo = √VV iiii 2 + 2gg (h − μμ ff + XX) 

Formula 2 

where: = outgoing velocity [m/s]; = incoming velocity [m/s]; = fall height [m]; = friction 

coefficient [–]; and = horizontal fall distance [m]. This velocity is calculated for each cell 

within the falltrack, starting at the rockfall source cell or at the impact cell on the talus slope, 

respectively (Dorren and Seijmonsbergen, 2002). 

Stopping occurs because energy is lost through friction forces and thus, velocity becomes 

zero. The value of the friction coefficient depends on the surface cover characteristics, 

including obstacles like vegetation and rocks. 

Basic friction coefficient values were defined according to literature (e.g. Meißl, 1998; Dorren 

and Seijmonsbergen, 2002; Wichmann, 2006). Small rocks retard more easily than bigger 

rocks; mainly, because during a rockfall the total kinetic energy of small rocks is lower than 

that of bigger rocks (Dorren, 2003). This was taken into account by the definition of different 

friction coefficients for each volume class. Forest cover characteristics were taken into account 

for the refined estimation of the friction coefficient. This step was based on a detailed 

characterisation of forest parameters on basis of ALS-data and satellite data which was 

performed within the frame of the same project (Schardt et al., 2015). According to the 

derived tree heights and their homogeneity, an automatic segmentation to forestal units was 

performed resulting in a GIS database with a total of 6.9 million polygons. The following 

forest parameters were selected as being relevant indicators of the surface roughness: (a) 

treetop number per unit area (4 classes), (b) crown coverage (3 classes), (c) height of upper 

layer (3 classes) and (d) vertical forest structure (2 classes). This resulted in the definition of 

24 forest types. The final refined friction coefficients are between 0,62 and 1,24 and were 

assigned to each of the forestal unit polygons. The friction coefficient values were calibrated 

by adjusting the modelled maximum runout distances to measured maximum runout 


distances in selected test sites (field measurements and analysis of remote sensed data). This 

was done for each volume class and each forest type in an iterative process. Subsequently for 

each volume class a friction coefficient file has been generated (cf. Fig. 3b). 

Figure 3: Detail of the Triebenstein region. (a) Hillshade visualization (b) Friction coefficients for Volume class 1 to forestal units (c) 

photograph of the most relevant rockfall area 

Potential transition cells for velocity calculation are assigned by using a multiple flow 

direction algorithm, based on the work of Gamma (2000). This algorithm was adapted for 

rockfall and compiled in the SAGA module Rock HazardZone by Wichmann (2006), not 

provided free of charge. 

The algorithm takes the local relief into account in order to calculate the magnitude of 

divergence yielding more realistic results. The method is implemented as a random walk in 


conjunction with a Monte Carlo approach. Modelled divergence is calibrated by three 

parameters: (1) a slope threshold, above which no divergence is modelled. (2) a parameter for 

the magnitude of divergence controlling whether a neighbouring cell exhibits a sufficiently 

high gradient to be selected as potential pathway and (3) a persistence factor that allows to 

increase the probability of that neighbouring cell, which features the same direction like the 

centre cell (Wichmann and Becht, 2006). 

The number of iterations was fixed to 1000 after tests with a higher number of iterations had 

not resulted in significant improvements of the outcomes. 

Modelling outputs consist of: (1) the number of transit frequencies. Due to high number of 

iterations (1000), the transition frequencies can be interpreted as transit probabilities of 

blocks approximatively. (2) The maximum velocities for each cell of the simulated blocks. 

Considering the three volume classes of blocks, approximate kinetic energies were calculated. 

Table 1 

The process area was finally classified with three qualitative susceptibility l categories ee la (low, Table Tabl 

moderate, high) based on the transition frequency and the estimated kinetic energy (cf. Table 

1). The Swiss codes are applied with regard to the limits of the kinetic energy (cf. Jaboyedoff, 

2005). 

Table 1: definition of susceptibility categories 

transition 

frequencies 

energies [kj] 

> 0 ≤ 30 > 30 ≤ 300 > 300 

1-5000 low moderate high 

> 5000 moderate high high 

By choosing appropriate friction coefficients, two scenarios were modelled: (1) a “bare earth” 

scenario simulating a complete absence of forest and (2) taking into consideration the forest 

cover at the time of the ALS flight mission. 

RESULTS 

The area wide modelling resulted in a total of approx. 603 km² of potential rockfall source 

areas (3,67 % of the study area), most of it situated in the Calcareous Alps and in crystalline 

rocks of the Niedere Tauern range. Based on field verifications in representative rockfall 

prone areas there is evidence that the used method rather over-estimates the extension of 

source areas. 

In the bare earth scenario an overall of 2.171 km² is shown as endangered by rockfalls which 

is 13,2 % of the area of Styria. This number is composed of 1,3 % classified as low, 5,4 % 

classified as moderate, and 6,5 % classified as high susceptible. 

Considering the forest scenario an overall of 1.541 km² is shown as endangered by rockfalls 

which is 9,4 % of the area of Styria. The portion of low susceptible areas rises to 1,7 % 


whereas the percentage of moderate susceptible areas decreases to 3,0 % and that of high 

susceptible areas decreases to 4,7 %. Field verifications of maximum runout distances in 

selected regions indicated that the results provided a high degree of accuracy. 

Due to different friction coefficients, the maximum runout distances between the bare earth 

scenario (low friction) and the forest scenario (higher friction) vary significantly. Especially 

multi-layered forest types with a high treetop density and with high (and therefore strong) 

trees reduce the propagation area. This can be interpreted as protective effect of forests 

(Fig. 4). 

Figure 4: Results of the rockfall modelling in the Triebenstein region: (a) “bare earth”-scenario; (b) scenario considering forest cover; (c) 

difference map; (d) and CIR orthophoto showing current forest cover and infrastructure). 

The protective effect of forests can be further visualized by calculating the difference of 

propagation areas between the forest scenario and the bare earth scenario (Fig. 4c). Thus the 

maximum reductions (three classes: from “high susceptible” to “not susceptible”) mainly can 

be found in the lowest parts of the rockfall runout zones and hence, in areas which are often 

used for infrastructure and settlements. 


CONCLUSIONS 

The presented method describes a generally applicable approach to model rockfall propagation 

areas. The approach is also suitable for large areas, which was done for the first time 

within the frame of the presented project. However, computing times up to several months 

have to be expected in large areas. 

High importance has to be given to the detailed identification of potential rockfall source 

areas, as the whole propagation modelling procedure is based on the results of this work step. 

Input parameters have to be calibrated in the field taking into account representative 

geological conditions that can be transferred over a wide area. In this context, additional 

efforts have to be made with regard to the generation of more detailed and reliable geological 

base maps. 

The modelling results of the presented method for the estimation of runout distances allow 

the classification into three susceptibility classes which can be used as indicative hazard map 

at a regional scale. 

Furthermore, different scenarios with regard to forest cover (e.g. reduced or missing forest 

cover) can be calculated, thus demonstrating the protective functions of forests. It is however 

required to have detailed information about relevant forest parameters and its effects on 

coefficients of friction. Here, again, extensive fieldwork is a fundamental precondition for 

receiving plausible results. 


The project was funded by the European Agricultural Fund for Rural Development and 

supported by the Austrian Federal Ministry of Agriculture, Forestry, Environment and Water 

Management, the Styrian Forestry Board and the Forest Association Styria. 

REFERENCES 

- Broilli L. (1974). Ein Felssturz im Großversuch. Rock Mechanics 3: 69–78. 

Carmichael R.S. (1984). Handbook of physical properties of rocks, Vol. III. CRC Press, Florida. 

- Dorren L.K.A. (2003). A review of rockfall mechanics and modelling approaches. Progress 

in Physical Geography 27 (1): 69–87. 

- Dorren L.K.A., Seijmonsbergen A.C. (2002). Comparison of three GIS-based models for 

predicting rockfall runout zones at a regional scale. Geomorphology 56: 49-64. 

- Gamma P. (2000). dfwalk - Ein Murgang-Simulationsprogramm zur Gefahrenzonierung. 

Geographica Bernensia G66. 144 p. 

- Hétu B., Gray J.T. (2000). Effects of environmental change on scree slope development 

throughout the postglacial period in the Chic-Choc Mountains in the northern Gaspé 

Peninsula, Québec. Geomorphology 32 (3-4): 335-355. 


- Jaboyedoff M., Dudt J. P., Labiouse V. (2005). An attempt to refine rockfall hazard zoning 

based on the kinetic energy, frequency and fragmentation degree. Natural Hazards and Earth 

System Sciences 5: 621–632. 

- Kienholz H., Mani P. (1994). Assessment of geomorphic hazards and priorities for forest 

management on the Rigi north face, Switzerland. Mountain Res. Dev. 14 (4): 321–328. 

Kobranova V.N. (1989). Petrophysics. MIR Publishers, Springer Verlag, 375 p. 

- Meißl G. (1998). Modellierung der Reichweite von Felsstürzen. Innsbrucker Geographische 

Studien 28. 249 p. 

- Motta R., Haudemand J.C. (2000). Protective forests and silvicultural stability. An example 

of planning in the Aosta Valley. Mountain Res. Dev. 20: 180–187. 

- Schardt M., Granica K., Hirschmugl M., Deutscher J., Mollatz M., Steinegger M., Gallaun H., 

Wimmer A., Linser S. (2015). The assessment of forest parameters by combined LiDAR and - 

Satellite data over Alpine regions - EUFODOS Implementation in Austria, Forestry Journal, 

The Journal of National Forest Centre – Forest Research Institute Zvolen, Lesn. Cas. For. J. 

61: 3-11. 

- Scheidegger A.E. (1975). Physical aspects of natural catastrophes. 289 p. 

- Wichmann V. (2006). Modellierung geomorphologischer Prozesse in einem alpinen 

Einzugsgebiet – Abgrenzung und Klassifizierung der Wirkungsräume von Sturzprozessen und 

Muren mit einem GIS. Eichstätter Geographische Arbeiten 15. Profil Verlag München/Wien. 

231 p. 

- Wichmann V., Becht M. (2006). Rockfall modelling: methods and model application in an 

alpine basin (Reintal, Germany). In: Böhner J., McCloy K.R., Strobl J. (eds): SAGA – Analysis 

and Modelling Applications. Göttinger Geographische Abhandlungen 115: 105-116. 



Sediment input from debris flows into mountain rivers: 

an event-based perspective 

Dieter Rickenmann, Dr. 1 ; Markus Gerber 2 ; Martin Böckli 3 

ABSTRACT 

Sediment transport in mountain streams is a function not only of upstream sediment input 

but also of sediment delivery from the steep lateral tributaries. In this study we analyzed a 

total of 32 debris-flow events which had occurred in the Swiss Alps. For 19 of the 32 studied 

events a sediment delivery occurred into the receiving stream. For the other 13 events no 

substantial part of the total event volume was transported into the receiving stream. We 

performed a statistical analysis to find a quantitative criterion for the question whether or not 

a debris flow with a given event magnitude will reach the confluence with the receiving 

stream. We developed a dimensionless index to respond to the above question. The index is 

based on the following three parameters: channel fan slope, length of the channel on the fan, 

and event volume. 

KEYWORDS 

debris flow; sediment input; mountain river; sediment connectivity; torrent fan 

INTRODUCTION 

Sediment transport in mountain streams is a function not only of upstream sediment delivery 

but also of sediment input from steep lateral tributaries. The amount of sediment potentially 

entering the main valley river from a tributary is likely to depend on the event magnitude of 

a debris flow in the torrent catchment as well as on the topographic conditions on the fan. 

These latter conditions may also be modified by man-made constructions such as bridges, 

houses, and protection structures. 

During a debris-flow event in a torrent catchment, the sediment transfer and deposition in 

the fan area depends primarily on the presence of sediment retention basins, the channel 

conveyance capacity in relation to the size of the debris flow, and potential clogging locations 

on the fan (e.g. narrow cross-sections under bridges crossing the channel). The sediment load 

transported to the fan apex can is often deposited on the fan outside of the channel, if the 

transport capacity in the channel is insufficient, for example in a flat reach upstream of the 

confluence. Streams with a catchment area of less than 25 km 2 and an average longitudinal 

slope greater than 5-10% can be classified as torrents (Rickenmann and Koschni, 2010; 

Rickenmann, 2014). 

1 Swiss Federal Research Institute WSL, Birmensdorf, SWITZERLAND, dieter.rickenmann@wsl.ch 

2 Züricher Hochschule für Angewandte Wissenschaften, Wädenswil and Swiss Federal Research Institute WSL, SWITZERLAND 

3 Swiss Federal Research Institute WSL and Holinger AG, Winterthur, SWITZERLAND 



For the quantification of solid transport in a mountain river, the potential sediment delivery 

from individual tributaries (torrents) must be known. For the estimation of the sediment 

delivery in debris-flow prone torrents there are several field-based methods (Marchi and 

D’Agostino, 2004; Jakob, 2005; Gertsch, 2009; Frick et al., 2008, 2011; Kienholz et al., 2010). 

Typically, with these methods an expected sediment load is estimated for the fan apex. 

There are many studies on the sediment connectivity of debris-flow fans with fluvial systems 

over longer time scales (Harvey, 2012). However, with regard to the important role of fans on 

the coupling or buffering of sediment cascades in the fluvial system, there are few quantitative 

investigations on the controlling factors of sediment delivery to the main valley river on 

an event-based time scale from debris flows in steep tributaries. 

For this study we analyzed the sediment input to mountain rivers from steep torrents based 

on 32 debris-flow events which had occurred from 1987 to 2009 in the Swiss Alps. Based on 

a statistical analysis we identified several parameters which can be used to estimate whether 

or not a debris flow (of a given event magnitude) delivers sediment into the receiving stream. 

Using some of the identified parameters, we developed a dimensionless index that can help to 

distinguish between the two cases. 

DATA AND METHOD 

The majority of the 32 case study events were taken from the PhD dissertation of Gertsch 

(2009) in steep torrents. The study catchments have surface areas mostly smaller than 20 km 2 

and mean channel gradients that are steeper than 10%. About a third of the study catchments 

have a sediment retention basin, mostly in the upper part of the fan near the apex. 

For the analysis, the 32 events were divided into the following two groups: (i) events and fan 

situations for which some of the total debris-flow material reached the receiving stream, 

labeled here as yes cases (YC), and (ii) events during which no sediment entered the receiving 

stream, labeled here as no cases (NC). For the first group, the available dataset did not, 

unfortunately, include sufficient information to quantify the amount of sediment which had 

entered the main stream. However, based on photo documents and on experience from other 

debris-flow events it is likely that for most events in group (i) only a relatively small fraction 

of the total sediment volume entered the receiving stream. 

In a first step, 18 parameters were determined for each event, characterizing the study catchment, 

the fan topography and the debris-flow event. Parameters characterizing the catchment 

and the fan were determined from the analysis of Digital Elevation Models (DEM 10m 

and DTM AV 2m of Swisstopo), the assessment of aerial photographs, and by field investigation. 

Event-related parameters were determined from the available event documentation and 

in some cases estimated with empirical formulas. The 18 parameters were: catchment area 

above fan apex, channel length from the apex to the confluence, mean fan slope, mean 


channel slope, mean channel slope according to Prochaska et al. (2008), mean channel slope 

until 25 m upstream of the confluence, mean channel slope for 100 m reach upstream of fan 

apex, Melton number (fan apex), Melton number (confluence), event volume, event volume 

without retention basin volume, maximum discharge (estimated from event volume), flow 

velocity (at maximum discharge), minimum cross-sectional area of channel on fan, estimated 

flow cross-sectional area, fan surface area, angle of fan channel to receiving stream (plan 

view), number of potential clogging sites. 

We then used statistical methods to search for a quantitative criterion or rule to distinguish 

between the two groups (i) and (ii). Some parameters were found to provide a reasonable 

differentiation between the YC and NC. Finally, using a subset of the most useful parameters, 

a dimensionless index was developed that can help to distinguish between the two cases. 

RESULTS 

Main processes on the fan during large debris-flow events 

For our dataset we compiled the main processes and fan situations which existed or occurred 

for large debris-flow events between the fan apex and the confluence with the main river 

(Fig. 1). 

Figure 1: Flowchart of the main processes on a fan during the debris-flow runout process. The line width is proportional to the number of 

observed cases. Out of 32 investigated events sediment entered the receiving stream in 19 cases (bottom line). 


Some important results can be described as follows: 

– For 19 of the 32 studied events a sediment delivery occurred into the receiving stream 

(YC). For the other 13 events no substantial part of the total event volume was transported 

into the receiving stream (NC). 

– For 21 events, an overtopping occurred out of the existing channel on the fan. Among 

these events there are 10 NC (not reaching the confluence) and 11 YC (reaching the 

confluence). 

– Although for 10 study catchments a sediment retention basin (SRB) exists on the fan, 

debris-flow material reached the confluence in some of these cases. 

A flow chart with the main processes occurring on the fan (see Fig. 1) could be used as a basis 

for the development of a decision tree that may eventually allow for a semi-quantitative 

estimation of the sediment delivery to the main stream. However, the number of possible 

sub-processes (many branching possibilities in the flow chart of Fig. 1) is very large in relation 

to the limited number of investigated events. Therefore, in the following analysis we mainly 

searched for criteria to distinguish between the two groups of events (i.e. YC and NC). 

Analysis of selected parameters 

Among the 19 pre-selected parameters we could identify five whose values allow a rough 

distinction between the NC and YC, in that there is no overlap of the quartiles of the value 

ranges (Fig. 2): (1) fan gradient, (2) channel gradient over 25 m horizontal distance upstream 

of the confluence, (3) mean flow velocity at maximum debris-flow discharge, (4) number of 

potential clogging locations, and (5) average channel gradient over 100 m horizontal distance 

upstream of the fan apex. Three of these parameters (2, 3, 5) are strongly correlated to the 

fan slope, and were therefore not used in the further analysis. 

Regarding the parameter length of the fan, Lf, we found that there were 5 NC with Lf > 1230 

m and 10 YC with Lf < 455 m, whereas the remaining 17 cases had intermediate LF values. 

This finding suggest that the longer is a torrential fan, the smaller is the probability that a 

debris flow reaches the receiving stream. 

All debris-flow events on a fan with a mean gradient, Sf, larger than 18 % or with a 

magnitude greater than 40,000 m 3 reached the receiving stream (Fig. 3). The more potential 

clogging locations are present on the fan, the lower is the probability that a debris flow 

reaches the receiving stream (Fig. 2). On fans with six or more potential clogging locations, 

only NC occurred. In contrast, if there were zero or one potential clogging location, there 

were only YC. 

We identified further parameters that may be suitable for a distinction between YC and NC: 

catchment area, mean channel gradient upstream of the fan, Melton number, ratio of 

calculated flow cross-section to minimum available flow cross-section, fan area, and angle of 


debris-flow channel to the receiving water. However, there is overlap of the quartiles of the 

value ranges between the two main groups (YC and NC). 

Figure 2: Range of values of five important parameters which may be related to the runout charac¬teristics. A = all events, B = no 

runout into the receiving stream (NC), C = runout into the receiving stream (YC), E =channel overtopping, F = no channel overtopping. 

Qmax is maximum debris flow discharge. A box includes the median value (thick line) and 50% of the data. It is limited by the upper 

and lower quartiles. Whiskers refer to the 5% and 95% values. Circles represent extreme values (not fully shown). 


Regarding the sensitivity to channel overtopping on the fan, we identified several parameters 

that may be useful for a delineation of this process: channel length (from apex to confluence), 

fan gradient, channel gradient over 25 m horizontal distance upstream of the confluence, and 

number of potential clogging locations. 

Dimensionless index 

Based on the above analysis, we searched for a quantitative criterion to distinguish between 

the two groups of events (YC and NC). Considering all the investigated parameters, some 

important correlation was identified among several of them, as for example between fan area 

and length of the fan channel or between the upstream channel gradient and the gradient of 

the channel on the fan. Among the following parameter pairs, however, no clear correlation 

could be identified: (i) fan gradient and fan length; (ii) fan gradient and event volume; (iii) 

event volume and fan length. 

Figure 3: Separation of YC and NC cases as a function of mean fan gradient (left panel) and event volume (right panel). 

As a result, we finally retained the three parameters fan gradient Sf, fan length Lf, and event 

volume Mexc for the development of a quantitative criterion. Here Mexc refers to the 

debris-flow event volume excluding the volume of the SRB if existing. By normalizing all 

1 i to onoa oua Eq1o 

three parameter values with the arithmetic mean of the parameter group (denoted by an 

overbar), a dimensionless index value was determined. By introducing a pre-factor of 0.42, 

the resulting range of index values lies between 0 and 1. 

IIIIIIIIII = 0.42 ∗ ( MM 0.33 

eeeeee 

∗ ( 

MM̅̅̅̅̅̅) 

LL 0.5 

̅̅̅ 

ff 

) ∗ SS 

eeeeee LL ff 

ff 

Equation 1 

The index permits a fairly good distinction between these events not reaching the confluence 

(NC) and those reaching the confluence (YC) (Fig. 4). Considering a fixed threshold of 0.06 

for the critical index value, only three YC events are not correctly classified (i.e. they would 

be incorrectly classified as NC events). The exponents of the different factors constituting 


Eq. 1 were determined by trial and error to obtain the best distinction between YC and NC 

events. 

DISCUSSION 

The torrent catchments included in our study are all prone to debris-flow occurrence. A 

limitation of our dataset is that there were few YC catchments with a natural fan, i.e. with 

only few anthropogenic structures. Among the catchments with NC, there was a relatively 

large proportion (69%) with a sediment retention basin on the fan. These two limitations, in 

combination with the relatively small number of 32 cases in total, may have influenced and 

biased our findings. 

Figure 4: Dimensionless index to separate YC and NC cases. The proposed critical limit of 0.06 (dashed line) approach correctly 

identifies 16 out of 19 events of the YC cases, and all 13 events of the NC cases are correctly identified with this limit. 

The proposed index allows for a reasonably good separation for debris-flow events and fan 

situations with or without sediment entering the receiving stream. It is possible that a more 

robust method could be developed by including more parameters such as the number of 

potential clogging locations in a detailed quantitative analysis. The index should be reviewed 

on the basis of further independent data. 

An alternative approach could consist in developing a decision tree, which could eventually 

also allow to estimate the fraction of the total event volume that reaches the receiving stream. 

Such a decision tree could be largely based on the flow chart of the main processes as 

depicted in Figure 2. The decision tree would start with a given debris-flow volume at the fan 

apex, and this volume would be reduced due to partial deposition along the flow path down 

to the confluence with the receiving stream or it may be enlarged due to sediment entrainment 

by erosion on the fan. To develop and test such an approach, a more detailed data base 


on debris-flow volumes entering receiving streams would be necessary to determine quantitative 

criteria at the branching points of the flow chart. A similar approach was developed in 

the context of a practice guide on "Estimation of the average annual sediment delivery into 

receiving streams" (BAFU, 2014). 


In this study we analyzed a total of 32 debris-flow events which had occurred in the Swiss 

Alps. For 19 of the 32 studied events a sediment delivery occurred into the receiving stream. 

For the other 13 events no substantial part of the total event volume was transported into the 

receiving stream. We performed a statistical analysis to find a quantitative criterion for the 

question whether or not a debris flow with a given event magnitude may reach the confluence 

with the receiving stream. We considered several topographic parameters which 

characterize the fan and channel properties, and we also included the event-based parameters 

in our analysis. 

We found that several of the pre-selected parameters were useful to discriminate between the 

two types of fan situations and events. We developed a dimensionless index, which is based 

on the following three parameters: channel fan slope, length of the channel on the fan, and 

debris-flow event volume. Selecting an optimal critical index value, this approach correctly 

identifies 16 out of 19 events in the first group (events with sediment entering the receiving 

stream); at the same time, all the 13 events in the second group (events without sediment 

reaching the receiving stream) are also correctly identified. 

The two topographic parameters of our index can be determined easily from digital elevation 

models. Additional compilations of comprehensive documentations of past debris-flow events 

including an estimate of the event volume and runout characteristics are important prerequisites 

for any further testing of the proposed index and for an improved development of a 

similar method. Other catchment or fan parameters could also be considered to possibly 

improve the discrimination method with an enlarged dataset. 


This study was performed in the context of the project «Feststofftransport in Gebirgseinzugs-gebieten» 

(Contract no. 11.0026.PJ / K154-7241) of the Swiss Federal Office for the 

Environment. The main part of the analysis was performed for the Bachelor thesis of Markus 

Gerber (Gerber, 2014). 

REFERENCES 

- BAFU (2014). Abschätzung der mittleren jährlichen Geschiebelieferung in Vorfluter – Praxishilfe. 

Hunziker, Zarn & Partner, Aarau; Lehmann, Hydrologie-Wasserbau, Urtenen – Schönbühl; 

belop gmbh, Sarnen; im Auftrag des Bundesamtes für Umwelt (BAFU), Bern, 75S. 


- Frick, E., Kienholz, H., Roth, H. (2008). SEDEX - eine Methodik zur gut dokumentierten 

Abschätzung der Feststofflieferung in Wildbächen. Wasser Energie Luft 100 (2): 131-136. 

- Frick, E., Kienholz, H., Romang, H. (2011). SEDEX (SEDiments and EXperts), Anwender-Handbuch. 

Geographica Bernensia P42, Geographisches Institut der Universität Bern, 

128p., ISBN 978-3-905835-27-4. 

- Gartner, J.E., Cannon, S.H., Santi, P.M., Dewolfe, V.G. (2008): Models to predict debris flow 

volumes generated by recently burned basins. Geomorphology 96: 339–354. 

- Gertsch, E. (2009). Geschiebelieferung alpiner Wildbachsysteme bei Grossereignissen - Ereignisanalysen 

und Entwicklung eines Abschätzverfahrens. Dissertation ( unveröffentlicht), 

Geographisches Institut der Universität Bern, Bern. 

- Gerber, M. (2014). Entwicklung eines Klassifizierungsverfahrens für den Feststoffeintrag in 

den Vorfluter durch Murgänge. Bachelorarbeit (unveröffentlicht), Zürcher Hochschule für 

Angewandte Wissenschaften, Wädenswil und Eidg. Forschungsanstalt WSL, Birmensdorf. 

- Harvey, A.M. (2012). The coupling status of alluvial fans and debris cones: a review and 

synthesis. Earth Surface Processes and Landforms 37: 64–76. 

- Jakob, M. (2005). Debris flow hazard analysis. In Debris-flow Hazards and Related Phenomena, 

ed. M Jakob, O Hungr (Chichester: Springer Praxis Publishing, 2005), pp. 411–443. 

- Kienholz, H., Frick, E., Gertsch, E. (2010). Assessment tools for mountain torrents: SEDEX© 

and bed load assessment matrix. Proc. International Interpraevent Symposium, Taipei, 

Taiwan, pp. 245–256. 

- Marchi, L., D’Agostino, V . (2004). Estimation of debris-flow magnitude in the Eastern 

Italian Alps. Earth Surface Processes and Landforms 29: 207–220. 

- Rickenmann, D. (2014). Methoden zur quantitativen Beurteilung von Gerinneprozessen in 

Wildbächen. WSL Berichte, Nr. 9, 105p. (www.wsl.ch/publikationen/pdf/13549.pdf) 

- Rickenmann, D., Koschni, A. (2010). Sediment loads due to fluvial transport and debris 

flows during the 2005 flood events in Switzerland. Hydrological Processes 24: 993–1007. 



Potential large wood-related hazards at bridges: the 

Długopole bridge in the Czarny Dunajec River, Polish 

Carpathians 

Virginia Ruiz-Villanueva 1,2 ; Bartłomiej Wyżga 3 ; Pawel Mikuś 3 ; Maciej Hajdukiewiczd 4 ; Markus Stoffel 1,2 

ABSTRACT 

Interaction between riparian vegetation and geomorphic processes in mountain streams can 

be amplified by abundant wood delivery to channels, high stream power and high sediment 

transport rates. At critical sections such as bridges, a quick succession of backwater can occur 

as a result of the reduction of the cross-sectional area. The aim of this work is to analyze 

wood transport and deposition during floods in the Czarny Dunajec River (Polish Carpathians) 

where it flows through the village of Długopole with a very narrow bridge cross-section. 

A numerical model which simulates the transport of large wood together with flow dynamics 

is applied and inlet and boundary conditions are determined based on field observations. 

Preliminary results showed that the bridge is easily clogged with wood under certain 

circumstances, although the blocking ratio and the probability of a log to be blocked depend 

on several factors. The final results will provide data to compute bridge clogging probability 

under the designed scenarios, and the potential impacts of the clogging on flood hazard. 

KEYWORDS 

driftwood, woody debris, clogging, flood, hazard 

INTRODUCTION 

The high potential risk associated with (flash) floods in mountain watercourses is a result of a 

rapid and complex hydrological catchment response (Borga et al., 2014). Besides high water 

levels in the drainage network, sediment transport and important morphological changes, the 

transport of large quantities of woody material must be considered an additional factor of 

flood risk in forested catchments (Mazzorana et al., 2011; Mao et al., 2013; Ruiz-Villanueva 

et al., 2014a). Interaction between riparian vegetation and geomorphic processes in mountain 

streams is amplified by abundant wood delivery to the channels, high stream power and high 

sediment transport rates. Recent floods across Europe highlighted some effects caused by 

large wood. An example can be the flood of August 2005 in Switzerland when, apart from 

flooding and enormous morphological changes in the streams, considerable amounts of wood 

1 Dendrolab.ch, Institute of Geological Sciences, University of Bern, SWITZERLAND, virginia.ruiz@dendrolab.ch 

2 Institute for Environmental Sciences, University of Geneva, SWITZERLAND 

3 Institute of Nature Conservation, Polish Academy of Sciences, Cracow, Poland; Faculty of Earth Sciences, University of Silesia, 

Sosnowiec, Poland 

4 Department of Environmental Engineering, Geomatics and Energetics, Kielce University of Technology, Poland 



were mobilized and deposited, resulting in high damage costs (Badoux et al., 2014; Rickenmann 

et al., 2015). 

Large wood (LW) in rivers has been described in the scientific literature since the 1970s 

(Swanson et al., 1976; Harmon et al., 1986; Montgomery et al., 2003). The presence of LW in 

river systems has been demonstrated to have very positive effects, for instance by enhancing 

the hydromorphological diversity of riverine habitats and representing a source of organic 

matter in channels (Gregory et al., 2003; Wohl, 2013; Beckman and Wohl, 2014). Therefore, 

management of LW has evolved in many regions and some forest management plans include 

guidelines to maintain riparian forest density (i.e. number of trees per unit area),buffer width 

and instream LW abundance (Spence et al., 1996). 

However, conflicts are usually greater in urban environments (Piégay and Landon, 1997), 

where the threat to infrastructure and public safety drives the most common management 

responses: to remove LW from the channel or to install debris racks to intercept LW transport 

(Bradley et al., 2005). Thus, LW management in urban areas implicitly assumes that wood is 

a problem and needs to be removed, with the practice resulting in degradation of aquatic 

habitats. It has been proved that wood removal often fails to prevent flooding, in part because 

of new input of LW during floods (Young, 1991; Gippel, 1995; Dudley et al., 1998). A more 

sustainable approach to manage LW in urban catchments is to redefine the problem as the 

inability of infrastructure to pass LW through the system (Lassettre and Kondolf, 2012). An 

alternative might be to modify the infrastructures. In addition the processes of LW input, 

storage and transport through the channel network might be accommodated by preserving 

zones of LW recruitment (forested hillslopes and riparian corridors) and areas of LW storage 

(gravel bars and floodplains) and maintaining pathways of LW transport, retaining the 

important ecological functions of LW (Lassettre and Kondolf, 2012). 

The first step in the improved LW management is therefore to identify the critical structures 

and the potential consequences in case of clogging. The partial clogging of bridges can cause a 

quick succession of backwater effects due to the reduction of cross-sectional area, which is 

accompanied by bed aggradation, channel avulsion and local scouring processes, which can 

ultimately lead to embankment/bridge collapse and floodplain inundation (Diehl, 1997; 

Comiti et al., 2007; Lyn et al., 2007; Mao and Comiti, 2010; Comiti et al., 2012; Badoux et al., 

2015; Lucía et al., 2015). As a result, flooded areas are likely to be different from those 

predicted from models where the presence of wood is not considered (Ruiz-Villanueva et al. 

2013), and therefore, this may result in the incorrect/uncertain estimation of flood risk. 

The ongoing research project FLORIST (Flood risk on the northern foothills of the Tatra 

Mountains), funded by the Polish-Swiss Joint Research Programme, aims at improving flood 

risk analysis in the region, including the analysis of large wood (Kundzewicz et al., 2014). 

During recent floods in Poland, such as those from 2001, 2010 and 2014, large quantities of 

wood were transported by mountain rivers, and large deposits of wood accumulated at some 


idge cross-sections, with adverse consequences (Hajdukiewicz et al, 2015). Wood deposits 

increased the hazardousness of the floods, with some of the above-mentioned effects. 

Therefore, the aim of this work is to analyse potential hazards related to wood transport and 

deposition at bridges during floods. We focused our investigations on the Czarny Dunajec 

River in the Tatra Mountains foreland (Polish Carpathians), where the river flows through 

the village of Długopole. Buildings in the village are located very close to the river and the 

bridge has a narrow cross-section (27 m) and is thus threatened by wood-related phenomena. 

This paper describes ongoing work within the FLORIST project and only preliminary results 

can be shown here. 

STUDY SITE 

The Czarny Dunajec (Figure 1) drains the Inner Western Carpathians in southern Poland. The 

river rises at about 1500 m above sea level (a.s.l.) in the high-mountain Tatra massif, with the 

highest peak in the catchment at 2176 m a.s.l. In the Tatra Mountains foreland, the river 

formed a non-cohesive alluvial plain consisting of resistant granitic and quartzitic particles 

Figure 1: (A) Location of the study area in the Polish Carpathians and location of the Czarny Dunajec River in relation to physiogeographic 

regions of southern Poland. Red rectangle indicates the study reach shown in panel B. 1 – high mountains; 2 – mountains of 

intermediate and low height; 3 – foothills; 4 – intramontane and submontane depressions; (B) Czarny Dunajec study reach upstream of 

the Długopole bridge; (C) view of the bridge from the left, upstream bank; (D) bank erosion and wood recruitment on the right bank a few 

tens of metres upstream of the bridge; (E) trees fallen to the river channel as a result of bank erosion (flow is from right to left). 

transported from the Tatras and sandstone gravel delivered to the Czarny Dunajec in the 

upper part of the foreland reach (Wyżga and Zawiejska, 2005). 

Characteristic features of the hydrological regime of the river are low winter flows and floods 

occurring between May and August due to heavy rainfall, sometimes superimposed on 


snow-melt runoff. Mean annual discharge of the river amounts to 4·4 m 3 s−1 at Koniówka, 

where the catchment area is 134 km 2 . The riparian forest is composed of alder and willow 

species with predominating young, shrubby forms of Alnus incana, Salix eleagnos, S. purpurea 

and S. fragilis, less frequent stands of older A. incana trees and occasional S. alba trees. 

The study reach is 1300 m long, with a channel width varying between 70 m and 20 m and 

amounting to 35 m on average, longitudinal slope of 0.006, and a drainage area of 145.7 km 2 . 

METHODS 

The numerical model developed by Ruiz-Villanueva et al. (2014a), called Iber-wood, is 

applied and inlet and boundary conditions are determined based on field observations. This 

model is fully coupled with a two-dimensional hydrodynamic model based on the finite 

volume method with a second-order Roe Scheme. Some of the parameters involved in the 

governing equations are wood density, angle of the log relative to flow, log length, log 

diameter, friction coefficient between the wood and the river bed, and the drag coefficient of 

the wood in water. The model includes two possible transport mechanisms for the movement 

of wood logs (floating or sliding/rolling) depending on wood density and water depth. In 

addition, translation and rotation (if one end of the wood piece is moving faster than the 

other) of logs are also included, based on flow velocity field. Interactions between logs and 

the channel boundaries and among logs themselves are also taken into account in the model. 

Therefore, log velocity and trajectory may change as a result of contacts with the banks or 

with other logs. The hydrodynamics and wood transport are coupled; thus, the hydrodynamics 

influence wood transport, but the presence of wood also influences hydrodynamics, 

adding a term to the bi-dimensional De Saint Venant equations (Ruiz-Villanueva et al., 

2014a). The model reproduces interactions between wood and infrastructures, computing 

whether a log can pass under or above the bridge deck, or become trapped by the structure, 

depending on the gate opening and width, or the weir length, water depth and wood 

diameter and length. When logs are trapped at the bridge, the drag force represents an 

opposite action to water flow, producing a rise in water level and a decrease in velocity 

(Ruiz-Villanueva et al., 2014b). The model has been already applied in the same river for 

other purposes related to the analysis of large wood dynamics (Ruiz-Villanueva et al., 2015a 

and 2015b). 

Topography of the studied reach is available from a LIDAR data from 2012; in addition, 23 

channel cross-sections in the vicinity of the Długopole bridge (18 cross-sections upstream of 

the bridge and 5 downstream) were used to update and improve the geometry. As a result, 

we obtained a DEM with 0.5 m pixel size resolution. Data from the nearest stream gauge 

station Koniówka is used for the calculation of flood discharges of a given probability/ 

recurrence interval, whereas the available rating curve was used for roughness (Manning’s n) 

calibration. In addition, water level observed during the flood of May 2014 was also used to 

validate model results. 


To characterize each piece of wood entering the reach, we established ranges of maximum 

and minimum lengths, diameters, and wood density based on the main types of trees re - 

cruited to this river. Stochastic variations of these parameters together with the position and 

angle with respect to the flow were used as well. We define a number of logs per minute to 

enter the simulation, assuming that wood recruitment is only occurring upstream of the 

study reach. 

The potential bridge clogging depends on many factors (Figure 2), such as (1) the approaching 

wood pieces (i.e. size and amount of wood transported in uncongested or congested 

manner, as defined by Braudrick et al., 1996), (2) the flow conditions (i.e. water level and 

flow velocity), (3) the geometry of the channel upstream the bridge (width, slope, curvature) 

and (4) the geometry of the bridge. 

Figure 2: Schematic sketch showing the three main factors influencing bridge clogging: (1) approaching wood (different size and 

different amount of wood); (2) flow conditions (in terms of water level and velocity field); (3) the geometry of the bridge (the one at 

Długopole has a single pier in the middle of the channel which is formed by a few steel columns linked with thinner steel elements) 

and the geometry of the channel. 

Running the model in a multiple scenario approach (to include the natural variability inherent 

to this process), we are analysing the three main factors affecting bridge clogging. We change 

the size and amount of wood pieces entering the river reach to determine the critical conditions 

in terms of wood. We also change the discharge to change the flow conditions. 


In further steps we will include a potential increase in bank erosion (see pictures D and E in 

Figure 1) which can also modify the flow conditions (by changing the flow velocity field). 

A probability of a bridge clogging can be computed with the ensemble from all the scenarios. 

PRELIMINARY RESULTS 

First run scenarios revealed the bridge as a critical section in the river if clogged. While for 

relatively high-magnitude floods (Q = 105 m 3·s-1, 10-year return period), the bridge capacity 

is enough to convey the flow, a reduction in the bridge cross-section caused by wood clogging 

significantly increases the flooding hazard for the nearby areas (Figure 3). 

Figure 3: Model results: flooded area and water depths (in metres) for a discharge of 105 m 3·s-1 in the study reach of the Czarny 

Dunajec River for the bridge cross-section at full capacity (A) and reduced by 90% (B). Flow is from left to right 

Preliminary results of the modelling of wood transport show that the bridge is easily clogged 

under certain circumstances, although the blocking ratio and the probability of a log to be 

blocked at the bridge depend on several factors. First of all, the size of the wood pieces, 

namely log length and diameter have a strong influence on the bridge clogging. Preliminary 

results show that the number of logs deposited at the bridge increases exponentially as log 

length and diameter increase (Figure 4). Log length seems to have a stronger effect on the 

entrapment than diameter. 


Figure 4: Model results: Percentage of deposited logs at the cross section upstream the bridge. Left: Simulated logs with fixed diameter 

of 0.2 m and different length values; right: Simulated logs with fixed length (equal to 10 m) and different diameters. In both graphs 

discharge is equal to 105 m 3·s-1. 

Flow conditions and the magnitude of the flood play an important role as well as the wood 

supply to the river reach and transport mechanism, and further scenarios will analyse these 

factors. Further results will provide data to compute the probability of bridge clogging under 

the designed scenarios, and the potential impacts of the clogging on hydrodynamics, flooded 

area and effects on the bridge stability. This information will be very useful for flood risk 

assessment and management of the river. 


This work was supported by the project Polish-Swiss FLORIST (Flood risk on the northern 

foothills of the Tatra Mountains; PSPB no.153/2010). The lead author acknowledges Dr. 

Ernest Bladé, Dr. Georgina Corestein and Marcos Sanz Ramos (FLUMEN Institute) for their 

support. 

REFERENCES 

- Andreoli, A., Comiti, F. Lenzi, M.A. 2007. Characteristics, distribution and geomorphic role 

of large woody debris in a mountain stream of the Chilean Andes. 1692, 1675–1692. 

- Badoux, A., Böckli, M., Rickenmann, D., Rickli, C., Ruiz-Villanueva, V., Zurbrügg, S., 

Stoffel, M. 2015. Large wood transported during the exceptional flood event of 24 July 2014 

in the Emme catchment (Switzerland). Proceedings of the Wood in World Rivers Conference, 

Padova, Italy, July 2015. 

- Badoux, A., Andres, N., Turowski, J.M. 2014. Damage costs due to bedload transport 

processes in Switzerland, Nat. Hazards Earth Syst. Sci., 14, 279–294. 

- Beckman, N.D. Wohl, E. 2014. Effects of forest stand age on the characteristics of logjams in 

mountainous forest streams. Earth Surface Processes and Landforms, 39, 1421–1431. 

- Borga, M., Stoffel, M., Marchi, L., Marra, F., Jakob, M. 2014. Hydrogeomorphic response to 

extreme rainfall in headwater systems: Flash floods and debris flows. Journal of Hydrology, 

518, 194–205. 


- Bradley J.B., Richards D.L., Bahner C.D. 2005. Debris Control Structures – Evaluation and 

Countermeasures. Report No. FHWA-IF-04-016, Hydraulic Engineering No. 9 (3rd edition), 

Federal Highway administration, Washington, D.C. 

- Diehl, T. 1997. Potential Drift Accumulation at Bridges. U.S. Department of Transportation 

Federal Highway Administration Research and Development Turner-Fairbank Highway 

Research Center, Georgetown Pike McLean, Virginia 22101-2296. Available: http://tn.water. 

usgs.gov/publications/FHWA-RD-97-028/FHWA-RD-97-028.pdf 

- Dudley, S., Fischenich, J. Abt, S. 1998. Effect of Woody Debris Entrapment on Flow 

Resistance. JAWRA Journal of the American Water Association 34, 1189–1197. 

- Gippel, C.J. 1996. Hydraulic Guidelines for the Re-Introduction and Management of Large 

Woody Debris in Lowland Rivers. Regulated Rivers: Research & Management, 12, 223–236. 

Gregory S.V., Boyer K.L., Gurnell A.M. (eds). 2003. The Ecology and Management of Wood 

in World Rivers, American Fisheries Society Symposium 37, Bethesda, MD. 

- Hajdukiewicz, H., Wyżga, B., Mikuś, P., Zawiejska J., Radecki-Pawlik, A. 2015. Impact of a 

large flood on mountain river habitats, channel morphology, and valley infrastructure. 

Geomorphology. In press. doi:10.1016/j.geomorph.2015.09.003 

- Harmon, M.E., Franklin, J.F., Swanson, F.J., Sollins, P., Gregory, S.V., Lattin, J.D., Anderson, 

N.H., Cline, S.P., Aumen, N.G., Sedell, J.R., Lienkaemper, G.W., Cromack Jr., K., Cummins, 

K.W. 1986. Ecology of coarse woody debris in temperate ecosystems. Adv. Ecol. Res. 15, 

133-302. 

- Kundzewicz, Z. Stoffel, M., Kaczka, R., Wyżga, W., Niedźwiedź, T., Pińskwar, I., Ruiz-Villanueva, 

V., Łupikasza, E., Czajka, B. 2014. Floods at the northern foothills of the Tatra 

Mountains — A Polish-Swiss research project. Acta Geophysica, 62, 620–641. 

- Lassettre, N.S., Kondolf, G.M. 2012. Large woody debris in urban stream channels : 

redefining the problem, River Res. Applic., 28, 1477–1487. 

- Lucía, A., F. Comiti, M. Borga, M. Cavalli, L. Marchi. 2015. Dynamics of large wood during 

a flash flood in two mountain catchments. Natural Hazards and Earth System Sciences 

Discussions, 3,1643–1680. 

- Lyn, D., Cooper, T. Yi, Y.K. 2003. Debris Accumulation at Bridge Crossings: Laboratory and 

Field Studies. Available at: http://docs.lib.purdue.edu/jtrp/48. 

- Mao, L., Comiti, F. 2010. The effects of large wood elements during an extreme flood in a 

small tropical basin of Costa Rica. In: De Wrachien D, Brebbia CA (eds) Debris flow III. WIT 

Press, UK, pp 225–236. 

- Mao, L. Andreoli, A., Iroumé, A.,Comiti, F.,Lenzi, F.A. 2013. Dynamics and management 

alternatives of in-channel large wood in mountain basins of the southern Andes. Bosque 

(Valdivia), 3415–16. 

- Mazzorana, B., Comiti, F. Fuchs, S. 2011. A structured approach to enhance flood hazard 

assessment in mountain streams. Natural Hazards, 67, 991–1009. 

- Montgomery, D.R., Collins, B.D., Buffington, J.M., Abbe, T.B. 2003. Geomorphic effects of 

wood in rivers. In: Gregory, S.V., Boyer, K.L., Gurnell, A.M. (Eds.), The Ecology and Management 

of Wood in World Rivers. American Fisheries Society, Bethesda, MD, pp. 21–47. 


- Piégay H, Landon N. 1997. Promoting ecological management of riparian forests on the 

Drome River, France. Aquatic Conservation: Marine and Freshwater Ecosystems 7, 287–304. 

- Rickenmann, D., Waldner, P., Usbeck, T., Köchil, D., Sutter, F., Rickli, C., Badeaus, A. 2015. 

Large wood transported during the 2005 flood events in Switzerland. Proceedings of the 

Wood in world rivers Conference, Padova, Italy, July 2015. 

- Rigon, E., Comiti, F. Lenzi, M.A. 2012. Large wood storage in streams of the Eastern Italian 

Alps and the relevance of hillslope processes. Water Resources Research, 48, 1–18. 

- Ruiz Villanueva, V. Ernest Bladé Castellet, Andrés Díez-Herrero, José M. Bodoque, Martí 

Sánchez-Juny. 2014b. Two-dimensional modelling of large wood transport during flash 

floods. Earth Surface Processes and Landforms, 39,438–449. 

- Ruiz-Villanueva, V. Ernest Bladé, Martí Sánchez-Juny, Belén Marti-Cardona, Andrés 

Díez-Herrero, José María Bodoque. 2014a. Two-dimensional numerical modeling of wood 

transport. Journal of Hydroinformatics, 16,1077. 

- Ruiz-Villanueva, V., Bartłomiej Wyżga, Joanna Zawiejska, Maciej Hajdukiewicz, Markus 

Stoffel. 2015a. Factors controlling large-wood transport in a mountain river. Geomorphology, 

in press. doi:10.1016/j.geomorph.2015.04.004 

- Ruiz-Villanueva, V., Bodoque, J. M., Díez-Herrero, a., Eguibar, M. a., & Pardo-Igúzquiza, E. 

2013. Reconstruction of a flash flood with large wood transport and its influence on hazard 

patterns in an ungauged mountain basin. Hydrological Processes, 27, 3424–3437. 

- Ruiz-Villanueva, V., Wyżga, B., Hajdukiewicz, H., & Stoffel, M. 2015b. Exploring large wood 

retention and deposition in contrasting river morphologies linking numerical modelling and 

field observations. Earth Surface Processes and Landforms, in press. doi:10.1002/esp.3832 

- Spence, B.C., Lomnicky, G.A., Hughes, R.M., Novitzki, R.P. 1996. An Ecosystem Approach 

to Salmonid Conservation. Draft Report No. TR-450-96-6057. ManTech Environmental 

Research Services Corporation: Corvallis, Oregon. 

- Swanson, F.J., Lienkaemper, G.W. Sedell, J.R.. 1976. History, physical effects, and management 

implications of large organic debris in western Oregon streams. U.S. For. Serv. Gen. 

Tech. Rep. PNW-56. 

- Wohl, E. 2013. Floodplains and wood. Earth-Science Reviews 123, 194–212. 

- Wyżga, B. Zawiejska, J. 2005. Wood storage in a wide mountain river: case study of the 

Czarny Dunajec, Polish Carpathians. Earth Surface Processes and Landforms, 30, 1475–1494. 

- Young, W.J. 1991. Flume study of the hydraulic effects of large woody debris in lowland 

rivers. Regulated Rivers: Research and Management, 6, 203–211. 



ProtectBio - Evaluation of the effects of protection 

forests on natural hazards due to gravity 

Arthur Sandri, dipl. Ing. ETH 1 ; Benjamin Lange, Dr. phil. nat. 1 ; Stéphane Losey, dipl. Ing. ETH 1 ; Bernhard Perren, dipl. Ing. ETH 2 

ABSTRACT 

Assessing the impact of biological protective measures is a major challenge. The project 

"PROTECT" of the National Platform for Natural Hazards (PLANAT) defines principles and 

processes to evaluate the impact of technical protection works. Within the framework of 

"ProtectBio", the methodology of PROTECT was adapted to biological measures, notably 

protection forests. This article shows that the principles for technical measures according to 

PROTECT are applicable to protection forests if some peculiarities of biological systems are 

considered. Thus, the impact of protection forests is comparable with the effects of technical 

protection works with the same objectives. The assessment of biological protection measures 

include several steps: the preparation, the rapid assessment, the evaluation of measures and 

the evaluation of effects. Some open questions remain concerning the quantification of the 

impact of the forest on medium and deep landslides as well as on floods. Notwithstanding 

these limitations, the here presented approach enables the quantitative assessment of the 

protection forests' effect and, thus, allows including biological measures in the integrated risk 

management. 

KEYWORDS 

Protection forest; Evaluation of Protection Measures; Natural Hazard Management; 

Risk Management 

INTRODUCTION 

The Federal Council mandated the national platform for natural hazard (PLANAT) to 

establish the strategy “Protection against Natural Hazards” (PLANAT 2004) which required as 

a core element the application of an integrated risk management. 

Integrated means that 

– all natural hazards are considered 

– all stakeholders are involved in the planning and implementation 

– all types of measures are included 

– all aspects of sustainability are taken into account 

Including all types of measures means that the protection against natural hazards is assured 

with both measures of prevention (spatial planning, technical and biological measures, 

1 Federal Office for the Environment FOEN, Ittigen, SWITZERLAND, arthur.sandri@bafu.admin.ch 

2 Impuls AG, SWITZERLAND 



preparedness measures) as well as measures of response (warning, rescue, damage protection, 

emergency measures) and measures of recovery. Integrated risk management is based on a 

comprehensive hazard and risk assessment and on an active communication about risk, 

which enables the community to correctly assess the risks and to act accordingly. An 

important bases of this comprehensive risk assessment is an in depth hazard and risk analysis 

which enables to quantify natural hazards and determines its process area. Moreover, the 

reliability and the effect of existing protection measures must be assessed. 

At a workshop of experts in natural hazards (FAN) in 2002 it was found that – depending on 

the hazard process – protection measures are considered very different (Romang et al. 2003). 

Therefore, PLANAT launched the project “Assessment of the effect of protection measures 

against natural hazards as a basis for their integration in spatial planning” called “PROTECT” 

(Romang et al. 2008). In the framework of this project, an evaluation method was developed, 

which includes nine principles to be tested in a four step procedure (Figure 1). 

Figure 1: Procedure to assess technical protective measures according to PROTECT. ¹: see Table 1. ²: Recommendations for spatial 

planning are not considered in ProtectBio 


However, PROTECT uses test criteria derived from civil engineering and designed for technical 

protection measures (as e.g. structural safety, serviceability, durability). Thus, a common basis 

for a comprehensible assessment of protective measures was created. It must therefore be 

examined whether and how these criteria can be applied to other protective measures such 

as protection forests. 

For this reason, the Federal Office for the Environment FOEN has initiated the project 

“ProtectBio” which is presented in this paper. The main objective of this paper is to show 

whether and how the principles and procedures of PROTECT are adaptable to biological 

protective measures in order to assess and compare their reliability and effects to technical 

protective measures. 

METHODOLOGY 

According to PROTECT, the assessment of technical protection structures includes the 

preparation and four consecutive steps: the rapid assessment to evaluate if there is an impact 

of the measure on the hazard process, the evaluation of measures to assess its reliability, the 

evaluation of the effects of the measures to quantify its impact and recommendations for 

spatial planning (Figure 1). Opposed to PROTECT, recommendations for spatial planning 

were not discussed within the framework of ProtectBio. Thus, step 4 is not elaborated. 

PROTECT defines nine principles for the assessment according to Figure 1. Table 1 explains 

these principles for technical measures. 

Table 1: Nine principles to prove the general suitability of technical protective measures according to PROTECT 

Principles 

1 Quantifiable Effect: the effect on the hazard process can be 

quantified or is at least discernible 

2 Uncertainties: the quantified impact of the measure on the hazard 

process exceeds the uncertainties in the assessment 

3 Scenarios: the assessment considers usual scenarios 

4 System Delimitation: the measure is assessed both with regard to the 

individual system and to the entire process area 

5 Permanent Availability: the protective effect is ensured for at least 50 

years 

6 Monitoring and Surveillance: the monitoring, the surveillance and in 

case of defects, the replacing is ensured 

7 Temporary Measures: temporary measures are not considered. 

8 Planned Measures: after a measure is realized, it is verified if the 

realization is according to the planning 

9 Time: since both protective measures and processes change in the 

course of time, the hazardous situation is verified periodically 


The preparation serves as a rough assessment to clarify whether these principles may be 

fulfilled and detailed analyses are appropriate. In-depth assessments of the nine principles are 

subsequently conducted in the framework of steps 1-3. One of the main goals of ProtectBio 

was to verify whether these principles and the procedure defined for technical protection 

structures are also usable for biological measures. In the context of this verification, some 

peculiarities of biological systems should be considered: 

1. Protection is a free service provided by the nature but only one of the functions of forests. 

2. Humans can influence the forest, but natural cycles and specific forest stand characteristics 

have to be considered. Forests cannot be planned to resist a specific impact of a certain 

event. 

3. The impact of the protection measure depends on natural cycles. 

4. The effect is area-wide. 

5. Forests and hazard processes interact with each other. 

Thus, due to these peculiarities, some deviations with regard to the interpretation of the 

principles are necessary when protection forests are assessed according to the procedure 

defined in PROTECT: 

Principles 1 and 2 (Quantifiable Effect and Uncertainties) 

It depends on the process to what extend the impacts of protection forests are quantifiable 

and whether the quantified impact exceeds the uncertainties. For avalanches, fall processes 

and shallow landslides, these principles are normally fulfilled. 

Principle 3 (Scenarios) 

Scenarios refer to events with differing probabilities of occurrence. The risk for an overload in 

forests may exceed that of technical measures since a forest is not planned to resist an event 

of a certain return period. 

Principle 4 (System Delimitation) 

The entire system includes the whole process area including forest and technical measures. 

Single systems describe a certain forest stand in a subterritory, e.g. the forest in the starting 

zone of an avalanche. 

Principle 5 (Permanent Availability) 

According to PROTECT, the protective effect should be ensured for at least 50 years (on 

condition that the measure is maintained). The return period of damages relevant to 

protection as storms, harmful organisms or forest fires at a specific site exceed 50 years even 

on pessimistic assumptions. Thus, permanent availability can be assumed. This does not mean 

that relevant damages never occur, but relevant events are rare and the protective functions 

do not disappear completely after a damage since laying wood and resilience (ability for 


egeneration of protection forests in case of damaging events) is often able to compensate the 

damage partially. 

Principle 6 (Monitoring and Surveillance) 

The guideline “Sustainability and success monitoring in protection forests” (Frehner et al. 

2005) is a tool that enables the monitoring and surveillance of protection forests at minimal 

costs. The need for action is derived from comparing the current state of the forest with the 

target profile, taking into account natural forest development. Local experts are responsible 

for both monitoring and surveillance which are a part of the forest planning. 

Principle 7 (Temporary Measures) 

Temporary measures are not considered when assessing the impact of protection forests. In 

the event of damages relevant to protection, temporary measures as laying wood may gain in 

importance. 

Principle 8 (Planned Measures) 

With the exception of reforestations, planned measures are of no significance. 

Principle 9 (Time) 

Changes in both the protective measures and hazard processes are considered by updating the 

target profiles for hazard processes in the guideline NaiS (Frehner et al. 2005) periodically. 

Table 2: Relevance of the protection measure „forest“ and its assessability. Green: relevance is possible and assessable. Yellow: 

relevance is only partially possible or relevance is only partially assessable. Red: no relevance or relevant, but not assessable. 

Natural hazard process 

Relevance of protection measure “forest” 

possible 

assessable 

Avalanches ● Flowing avalanche yes yes 

● Powder avalanche yes yes 

● Snow glide yes yes 

● Ice avalanche no - 

Fall ● Rock fall yes yes 

● Rock slide yes (only small events) yes 

● Rock avalanche no - 

● Ice fall yes yes 

Water ● Flooding yes rarely 

● Debris flow yes yes 

● Erosion yes yes 

Spontaneous 

● Shallow yes yes 

landslides 

● Intermediate yes no 

● Deep yes no 

Permanent landslides ● Intermediate yes no 

● Deep yes no 

Sinkhole / Subsidence ● yes no 


To sum up, the principles of PROTECT are applicable to protection forests when some 

pe culiarities of biological measures are taken into account. Thus, the procedure for technical 

measures according to Figure 1 (steps 1-3) is in principle transferable to protection forests. 

Furthermore, it was evaluated for which gravitational natural hazards a relevance of forests 

can be expected and, additionally, if this relevance is assessable. Table 2 gives an overview of 

this evaluation. 

RAPID ASSESSMENT 

The rapid assessment, as the first step after the preparation according to Figure 1, aims at 

giving a first overview of the situation and includes an estimation of the relevance of the 

protection measure. This evaluation reveals whether the measure has a relevant impact on 

the hazard and whether a more detailed investigation is appropriate. 

Thus, the rapid assessment should be practicable with minimal effort, i.e. as far as possible 

without fieldwork and complex calculations. Criteria used for the rapid assessment of a forest 

are, for example, the percentage of forest cover and its location in the process perimeter, 

stand characteristics like the canopy cover and the differentiation between stands that 

effectively have a protective effect and such which are presumably ineffective (e.g. young 

forests). 

Figure 2: Approach for the rapid assessment of protection forest in the starting zone of avalanches 


Within the framework of ProtectBio, decision schemes for frequent natural hazard processes 

were developed to simplify the rapid assessment. An example of such a scheme, concerning 

the relevance of the forest in starting zones of avalanches, is shown in Figure 2. 

EVALUATION OF MEASURES 

In this step, the reliability of protective measures is assessed with regard to their impact on 

the hazard process. According to PROTECT, the reliability depends on the structural safety of 

the measure, its serviceability and durability. The structural safety is the ability to resist an 

impact. As an example, a forest in an avalanche starting zone should resist a certain snow 

load. Thus, tree species and the coefficient of slenderness determine the structural safety in 

this case. 

The serviceability is the capacity of a forest to guarantee its protective function and depends 

on both the protection aim for the damage potential and the resulting state of the forest. It is 

assumed that serviceability is given if the condition of the forest fulfills the minimum target 

profile according to the guideline “Sustainability and success monitoring in protection forests 

(NaiS)” (Frehner et al. 2005). 

Durability means that the structural safety and the serviceability of a protective measure are 

ensured for at least 50 years (on condition that the measure is monitored and maintained). 

Crucial is the current condition of the protection forest and the probable development over 

50 years according to NaiS (Frehner et al. 2005) as well as the probability of relevant 

Figure 3: Evaluation of measures: determination of the reliability based on structural safety, serviceability and durability (according to 

Wasser & Perren 2014) 


damages. Moreover, regeneration is of large importance and, thus, ungulate browsing as a 

factor that affects tree establishment, growth and mortality. 

To sum up, the reliability of a protection forest is determined according to Figure 3 based on 

the assessment of its structural safety, its serviceability and durability. Currently, the reliability 

is determinable for hazard processes whose relevance is assessable according to Table 2. 

EVALUATION OF EFFECTS 

The evaluation of effects quantifies the significance of the protection measure to the hazard 

process and therefore to the risk for the damage potential. Stand characteristics that cause 

protective function are natural hazard specific. Consequently, parameters allowing the 

quantification of the protective function of a forest such as canopy cover (avalanche) or basal 

area (rockfall) have to be known and methods to quantify this influence must be available. 

These methods range from physically or probabilistic based models to expert evaluations and 

calculations and indirect evaluations. 

Existing approaches evolve and new methods are developed. Thus, the here presented 

statements about the quantifiability of the forests’ protective function represent the current 

state of the art. Usable to good quality quantifications are possible for hazard processes whose 

relevance is assessable according to Table 2. 

Open questions remain especially regarding landslides and floods. Even though there is an 

impact of forests to the water cycle and therefore to flood protection, it is rarely possible to 

quantify this connection. In hydrological models the forest is sometimes incorporated by 

adjusting the discharge coefficient or discharge measurements in forested watersheds include 

the impact of the forest implicitly. For shallow landslides, there are some efforts to include the 

influence of roots on the stability of slopes into models (e.g. Schwarz et al. 2012). Thus, 

significant improvements to validate the impact of the forest on soil stability are expected in 

the near future. 


ProtectBio shows that the procedure for assessing the effect of technical protection measures 

is in principle applicable to biological measures such as protection forests. 

Wherever hazard and risk assessments are made, the protective effects of forests should be 

adequately addressed in the future. As part of a comprehensive variant study of different 

protection measures, biological systems should be considered as possible alternatives to 

technical measures. Often, the variant studies do not lead to a clear preference of either a 

technical or biological solution but rather an optimal combination of both. For example, the 

better the effects of rock fall protection forests are assessable, the more accurate additional 

rock fall protection nets can be planned and designed. This combination suits therefore the 

demands of the integrated risk management. 


ProtectBio is not suitable for an area-wide application. The associated effort would be 

disproportional high. However, it makes sense to use it wherever detailed risk assessments 

and planning of protection measures are made. So far, ProtectBio has not been applicable as 

an automated processes, e.g. in a GIS, since georeferenced data of forest characteristics are 

rarely available. Moreover, the effect of protection forests on intermediate and deep landslides 

as well as on flooding is hardly quantifiable with existing approaches. Consequently, more 

research is needed in these areas. Furthermore, there are still open questions regarding the 

influence of gaps, openings and gullies in the forest cover. Not only size and number of 

openings and gullies are crucial for the resulting risk, but also their location in the transit 

area. Again, further work is necessary to gain simply applicable criteria. But overall, Protect- 

Bio enables the evaluation of the reliability and effectiveness of protection forests and their 

inclusion in the integrated risk management. 

LITERATURE: 

- Frehner M., Wasser B., Schwitter R. (2005). Nachhaltigkeit und Erfolgskontrolle im 

Schutzwald. Wegleitung für Pflegemassnahmen in Wäldern mit Schutzfunktion, Vollzug 

Umwelt. Bundesamt für Umwelt, Wald und Landschaft, Bern, 564 p. 

- PLANAT (2004): Sicherheit vor Naturgefahren. Vision und Strategie. Nationale Plattform für 

Naturgefahren, Biel, 41 S. 

- Romang H., Margreth S., Kienholz H., Böll A. (2003). Berücksichtigung von Schutzmassnahmen 

bei der Gefahrenbeurteilung. Workshop der Forstlichen Arbeitsgruppe 

Naturgefahren (FAN) 29./30.10.2002, St. Gallen, FAN-Sekretariat, 53 S. 

- Romang H. (Ed.) (2008). Wirkung von Schutzmassnahmen. Nationale Plattform für 

Naturgefahren PLANAT, Bern, 289 S. 

- Schwarz M., Cohen D., Or D. (2012). Spatial characterization of root reinforcement at stand 

scale: Theory and case study. Geomorphology 171 – 172: 190 – 200. 

- Wasser B., Perren B. (2014). Wirkung von Schutzwald gegen gravitative Naturgefahren – 

Protect-Bio. Schweizerische Zeitschrift für Forstwesen 165: 275-283. 



Backwater rise due to driftwood accumulation 

Isabella Schalko 1 ; Dieter Brändli 2 ; Lukas Schmocker, Dr. 1 ; Volker Weitbrecht, Dr. 1 ; Robert Michael Boes, Prof. Dr. 1 

ABSTRACT 

Transported driftwood during flood events can lead to accumulation at river infrastructures or 

intentionally be retained at driftwood retention structures. In both cases, the driftwood 

accumulation results in backwater rise upstream of the cross section and may consequently 

overtop the adjacent river embankments. During previous investigations various governing 

parameters for the estimation of the backwater rise were detected, but some findings are still 

contradictory. Within this study a series of hydraulic flume experiments were conducted to 

identify the decisive parameters on the backwater rise testing predefined driftwood accumulations. 

The effects of the approach flow condition (inflow flow depth and Froude number) as 

well as the driftwood accumulation characteristics (accumulation length, bulk factor and 

driftwood characteristics) were considered to enable the prediction of the expected backwater 

rise. The results of the experiments show that the backwater rise depends mainly on the 

Froude number and the driftwood accumulation characteristics (e.g. log diameter and 

driftwood accumulation compactness). 

KEYWORDS 

Backwater rise; bridge clogging; driftwood retention; flooding; large woody debris 

INTRODUCTION 

During the 2005 flood event in Switzerland approximately 30’000 t of driftwood were transported 

(Bezzola and Hegg 2007; VAW 2008; Waldner et al. 2010). The transported driftwood 

led to numerous accumulations and clogging of the flow cross section at river infrastructures 

such as bridges or weirs. This resulted in severe problems due to the backwater rise upstream 

of the blocked cross section and consequently flooding of the surrounding area. 

Engineering measures are necessary to reduce the destructive power of the interaction 

between transported driftwood and river infrastructures during flood events. They can be 

divided in (1) maintenance of the catchment area (e.g. removal of deadwood, erosion or 

landslide prevention and forest maintenance), (2) safe downstream conveyance of driftwood 

and (3) retention structures (e.g. racks or nets). The construction of retention structures is 

an essential measure to retain large driftwood volumes during flood events (Perham 1987; 

Wallerstein et al. 1996; Bradley et al. 2005; Hartlieb and Bezzola 2000). These retention 

structures also lead to a backwater rise and hence to decreasing flow velocity and increased 

sediment deposition (Bradley et al. 2005; Lange and Bezzola 2006), thereby intensifying the 

backwater rise. 

1 Laboratory of Hydraulics, Hydrology and Glaciology (VAW), ETH Zürich, SWITZERLAND, schalko@vaw.baug.ethz.ch 

2 Formerly VAW, ETH Zürich, SWITZERLAND 



When planning a driftwood retention structure, the backwater rise is a relevant design parameter 

for the rack height and the adjacent river embankments. The backwater rise directly 

determines the required rack height to prevent the retained driftwood from getting flushed 

over the rack. Furthermore, the resulting backwater rise at a bridge or weir due to driftwood 

blocking is an important parameter to conduct flood hazard assessments. 

STATE OF THE ART 

In order to investigate the backwater rise due to driftwood accumulation various studies were 

conducted. Knauss (1995) tested four different driftwood rack configurations (diagonal, 

V-shaped in and against flow direction and straight rack) and investigated the impact of coarse 

and fine driftwood material on the backwater rise. He defined the backwater rise due to driftwood 

accumulation at a rack with the backwater parameter α: 

Equation 1: Backwater parameter α (Knauss 1995). 

With h 2 

= flow depth with driftwood accumulation [m], h 1 

= initial flow depth without 

driftwood accumulation [m], v 1 

= initial flow velocity without driftwood accumulation [m/s], 

and g = gravitational acceleration [m/s 2 ]. 

The backwater parameter α depends mainly on the flow velocity and consequently on the 

specific discharge q, the driftwood characteristics as well as the rack configuration. For coarse 

driftwood α equals 1.5 and can increase up to α ≈ 2.3 for finer material. Compared to a 

straight rack, the V-shaped rack (V pointing in flow direction) results in a smaller backwater 

rise due to its larger rack length. The driftwood pile up at a V-shaped rack is reduced which 

favors the development of a driftwood carpet. Thus the driftwood placement is rather loose 

with a smaller backwater rise. In addition Knauss (1995) observed an increase of the 

backwater rise with increasing approach flow Froude number for the V-shaped rack. 

Rimböck (2003) studied the design of rope net constructions for driftwood retention. His 

experiments showed that the governing parameters affecting the backwater rise are the 

driftwood characteristics (mixture and wood type), discharge Q, bed slope J, as well as the 

channel roughness k st 

. The flow velocity at the upper end of the driftwood carpet should not 

exceed 0.8-1 m/s. Hence the driftwood is less compact and the risk of overtopping the 

retention structure can be reduced. 

A different rack configuration with the objective to reduce the backwater rise was introduced 

by Schmocker and Weitbrecht (2013). They presented the so-called bypass retention where 

the driftwood is retained parallel to the main stream in a bypass channel. The bypass channel 


is located at the outer bend of a river and the rack is placed parallel to the river axis. Due to 

the secondary currents in the river bend the driftwood is transported at the outer bend and 

into the bypass channel. The approach flow is parallel to the driftwood rack, which leads to a 

smaller backwater rise. Since the bed load remains in the main channel, the bypass retention 

has only a reduced influence on the sediment transport capacity. 

Further experiments on the backwater rise were conducted by Schmocker and Hager (2013). 

To simplify small scale model tests, they identified F o 

and the loosely placed driftwood volume 

V L 

as the key parameters for the accumulation process. 

Hartlieb (2014) conducted a dimensional analysis to parameterize the backwater rise due to 

driftwood accumulation. He also identified F o 

and the bulk factor a as the relevant parameters. 

The bulk factor a = V L 

/V S 

can be described by the loosely placed driftwood volume V L 

and 

the solid driftwood volume V S 

. Therefore, it defines whether a driftwood accumulation is 

compact (low bulk factor) or loose (high bulk factor). He evaluated several test results of 

driftwood accumulation at hydraulic structures and concluded that an increase of F o 

and a 

decrease of a lead to a higher backwater rise. 

OBJECTIVE 

Despite recent research, the knowledge on the backwater rise due to driftwood accumulation 

is still limited and some results are contradictory. Tab. 1 summarizes the relevant parameters 

for the backwater rise due to driftwood accumulation identified in previous studies. 

Tab. 1: Summary of relevant parameters for the backwater rise due to driftwood accumulation. 

The few available formulae for the backwater rise were established for a limited number of 

tests and apply mostly to a specific rack placement. Hence, the objective of the present study 

was to systematically investigate the resulting backwater rise due to a driftwood accumulation 


and to expand the existing parameter range. The effects of the approach flow condition 

(inflow flow depth and Froude number) as well as the driftwood accumulation characteristics 

(accumulation length and bulk factor, driftwood characteristics) were considered. The 

findings of this study should enable the estimation of the expected backwater at a driftwood 

rack. This would allow for a more precise design of driftwood retention structures and 

consequently an improvement of flood hazard assessments. 

SCALE MODEL 

Hydraulic Flume 

The experiments were conducted in a glass-sided flume at the Laboratory of Hydraulics, 

Hydrology and Glaciology (VAW). The flume is 8 m long, 0.4 m wide, 0.7 m high and its slope 

can be manually adjusted. A flow straightener at the inlet generated undisturbed inflow. The 

inflow discharge can be automatically regulated with a valve. The approach flow hydraulics 

(subscript o) are characterized by the flow depth h o 

and the flow velocity v o 

= Q o 

/(Bh o 

), or the 

Froude number F o 

= v o 

/(gh o 

) 1/2 , with Q o 

= discharge, B = channel width and g = gravitational 

acceleration. Fig. 1 shows the test setup and notation. 

Figure 1: (a) Test setup and notation. Initial approach flow conditions (F o 

, h o 

) without driftwood accumulation. UDS = Ultrasonic 

Distance Sensor. (b) Picture from the model test. 

The backwater rise is mainly governed by the initial driftwood accumulation, whereas the 

subsequently developing driftwood carpet has only a minor effect (Schmocker and Hager 

2013). The accumulation was therefore simplified as a pre-installed driftwood volume V D 

placed in between two racks. These racks, placed 3 m downstream of the intake, consisted of 

seven steel poles with a diameter of d = 0.008 m and therefore had a negligible effect on the 

overall backwater rise. The driftwood contained logs of length L L 

and diameter d L 

. The 


driftwood accumulation had a width of L A 

, a constant height of 0.4 m and a bulk factor of a = 

V L 

/ V S 

. The range of a = 2 - 5 was selected based on recent observations from WSL on 

driftwood accumulations in nature (Waldner et al. 2010). The approach flow depth h o 

and the 

resulting flow depth h = h o 

+ Δh (with Δh = backwater rise) were measured using two 

ultrasonic distance sensors, UDS 0 

placed 1 m upstream and UDS 1 

1 m downstream of the 

driftwood accumulation. Additional flow depths were measured using a manual point gauge. 

Model driftwood 

During the experiments the model driftwood without branches was divided in seven wood 

classes to represent a broad range of L L 

and d L 

(Fig. 2). The smallest class A consisted of 

matchsticks. While d L 

is rather constant for the classes A-D, it shows a wide range for classes 

E-G. Class G is a mixture of the smallest class A and class E. The model driftwood dimensions 

were scaled according to natural driftwood observed during the flood event 2005 (Waldner et 

al. 2010). Furthermore, the effect of fine material as small branches and leaves on the 

backwater rise was neglected. 

Figure 2: Wood classes used in the experiments with d L 

= log diameter, L L 

= log length and d Lc 

= characteristic log diameter. 

Another relevant aspect of the wood classes is the characteristic log diameter d Lc 

. It was 

calculated as a function of the mean circumference U m 

. In case of driftwood accumulation at a 

rack, water has to pass around the logs. Therefore the characteristic log diameter d Lc 

of the 

wood class can be calculated with Equation 2. 


Equation 2: Characteristic log diameter d Lc 

. 

With d L 

= log diameter [m], N(d L 

) = number of logs with d L 

in wood class and N = total 

number of logs. 

Test program and procedure 

The experiments were conducted within four test series to identify the impact of various 

parameters on the backwater rise. In series I the effect of F o 

was tested for three different h o 

with all other parameters kept constant. Series II to IV investigated the effect of the parameters 

a, L A 

and d Lc 

for different F o 

with all other parameters kept constant. The test program 

with the investigated parameters is listed in Tab. 2. 

Tab. 2: Test program 

Previous studies at VAW investigated the effect of driftwood mixture on the backwater rise 

and showed that the driftwood mixture could be represented by the mean diameter of the 

mixture (Schmocker and Hager 2013). Therefore, the effect of a mixture of various driftwood 

classes was not tested herein. Furthermore, the log length and shape were not investigated as 

separate parameters, as this is accounted for with the bulk factor. 

The experimental procedure can be described by the following steps: 

1. Measurement of h o 

and F o 

without driftwood accumulation. 

2. Inserting the two bar racks in the flume and adding the respective driftwood class and 

volume for the driftwood accumulation. 

3. Measurement of Δh in front of the driftwood accumulation for the respective ho and F o 

. 


RESULTS 

The Test series I has been conducted to identify the effect of F o 

on the backwater rise for 

various h o 

. Fig. 3a shows Δh for h o 

= 50, 100 and 150 mm, wood class G, a = 3.6 and L A 

= 

100 mm. F o 

was varied from 0.2 to 1.4. For h o 

= 150 mm only F o 

from 0.2 to 0.6 were tested, 

since the resulting backwater rise for F o 

≥ 0.8 overtopped the maximum driftwood accumulation 

height of 0.4 m. The results show that the backwater rise Δh increases linearly with 

increasing F o 

(Fig. 3a). The backwater rise Δh for h o 

= 100 mm and F o 

= 1.2 is 6.5-times higher 

than for F o 

= 0.2. In order to identify the effect of the approach flow depth on the backwater 

rise, the relative backwater rise Δh/h o 

was plotted on the ordinate (Fig. 3b). The data for all 

three approach flow depths h o 

collapse, hence the relative backwater rise is independent from 

h o 

. 

Figure 3: (a) Δh and (b) Δh/h o 

for three different h o 

various F o 

, wood class G, L A 

= 100 mm and a = 3.6. (c) Δh for h o 

= 100 mm, 

various F o 

, three different a, L A 

= 100 mm and wood class C. (d) Δh for h o 

= 100 mm, various F o 

, three different L A 

, wood class F and 

a = 3.8. 

The bulk factor a describes the compactness of the driftwood accumulation. During the 

experiments a certain bulk factor was established by placing the respective driftwood volume 

V D 

between the two racks. Fig. 3c shows the backwater rise for test series II with a = 2.4, 3.2, 

3.6, F o 

= 0.2-1.4, L A 

= 100 mm, h o 

= 100 mm and the wood class C. The experiments for a = 

2.4 were limited to F o 

= 0.2-0.8 to avoid an overtopping of the maximum driftwood accumulation 

height. The backwater rise Δh for F o 

= 0.8 equals 317 mm for a = 2.4 compared to 


Δh = 150 mm for a = 3.6. Therefore, the backwater rise is decreasing with increasing bulk 

factor (i.e. loose driftwood accumulation). A compact driftwood accumulation represents a 

higher flow resistance compared to a loose driftwood accumulation, leading to a higher 

backwater rise. These findings were observed for all wood classes. As expected the bulk factor 

acts like the porosity factor in groundwater flows. 

The impact of the driftwood accumulation length on the backwater rise (Test series III) is 

plotted in Fig. 3d. The experiments were conducted for F o 

= 0.2-1.4, h o 

= 100 mm, a = 3.8 and 

wood class F. The driftwood accumulation length was varied between L A 

= 50, 100 and 200 

mm. As plotted in Fig. 3d, the backwater rise increases with increasing L A 

. A higher accumulation 

length represents a greater flow resistance and consequently leads to an increased 

backwater rise. 

The geometric driftwood characteristics can be described by the characteristic log diameter d Lc 

of the respective wood class. A small d Lc 

allows for a more dense driftwood accumulation. 

Figure 4: (a) Δh for h o 

= 100 mm, L A 

= 100 mm and a = 3.6, (b) Δh for d Lc 

= 14 mm and F o 

= 1.2, (c) ∆h for d Lc 

= 2.3 mm 

and F o 

= 1.2. 


Therefore the driftwood accumulation contains a higher number of logs, which leads to a 

higher flow diversion when water is passing through. Test Series IV for the impact of d Lc 

on 

Δh were conducted for F o 

= 0.2-1.4 and wood classes A, B and D (Fig. 2). The backwater rise 

for d Lc 

= 2.3 mm, 8.5 mm and 14 mm (i.e. wood classes A, B, D) with a = 3.6, L A 

= 100 mm 

and h o 

= 100 mm is plotted in Fig. 4a. Two pictures from the model tests for class A, F o 

= 1.2, 

a = 3.6, L A 

= 100 mm, d Lc 

= 14 mm and d Lc 

= 2.3 mm are shown in Fig. 4b and Fig. 4c. The 

backwater rise Δh for F o 

= 0.8 equals 236 mm for d Lc 

= 2.3 mm compared to Δh = 153 mm for 

d Lc 

= 14 mm. Consequently, Δh increases with decreasing d Lc 

. 

CONCLUSIONS 

Hydraulic flume experiments were conducted to identify the relevant parameters affecting 

the backwater rise due to driftwood accumulation. During the experiments the approach flow 

conditions as well as the driftwood accumulation characteristics were systematically varied. 

The results show that the backwater rise increases with increasing Froude number F o 

and 

driftwood accumulation length L A 

as well as with decreasing characteristic log diameter d Lc 

and bulk factor a. Recently, numerical models have been developed to simulate driftwood 

transport, the accumulation processes and the resulting backwater rise. The results of this 

study may be used to validate these numerical models. 

The fine material like branches and leaves in the driftwood accumulation was neglected 

within this study. Further work at VAW aims to model the effect of fine material on the 

backwater rise and to combine all results in a design diagram for the expected backwater rise. 

This allows to estimate the backwater rise and consequently to design the height of a planned 

driftwood retention rack. Regarding bridge clogging, the estimated backwater rise helps to 

assess the hazard potential of existing bridges or river crossing structures during flood events 

with high driftwood transport. 

REFERENCES 

- Bezzola G.R. and Hegg C. (2007). Ereignisanalyse Hochwasser 2005 Teil 1: Prozesse, 

Schäden und erste Einordnung (Analysis of 2005 flood event, Part 1: Processes, damages and 

classification). Federal Office for the Environment FOEN, Swiss Federal Institute for Forest, 

Snow, and Landscape Research WSL. Umwelt-Wissen: 0825 (in German). 

- Bradley J.B., Richards D.L., Bahner C.D. (2005). Debris control structures: Evaluation and 

countermeasures. Report No.: FHWA-IF-04-016. U.S. Dept, Transportation, Federal Highway 

Administration, Washington, D.C. 

- Hartlieb A. and Bezzola G.R. (2000). Ein Überblick zur Schwemmholzproblematik (Overview 

of problems associated with driftwood). Wasser, Energie, Luft 92 (1/2): 1-5 (in German). 

- Hartlieb A. (2014). Maßgebende Parameter für den Aufstau durch Schwemmholz-verklausungen 

(Decisive parameters for backwater effects caused by floating debris jams). Proc. Intl. 

Symposium "Wasser- und Flussbau im Alpenraum", VAW-Mitteilung 228 (R. Boes, ed.), ETH 

Zurich: 485-493 (in German). 


- Knauss J. (1995). Von der oberen zur unteren Isar (From the upper to the lower Isar). 

Report 76: 23-66. TU Munich, Germany (in German). 

- Lange D. and Bezzola G.R. (2006). Schwemmholz: Probleme und Lösungsansätze (Driftwood: 

problems and countermeasures). VAW-Mitteilung 188 (H.-E. Minor, ed.), ETH Zurich 

(in German). 

- Perham R. E. (1987). Floating debris control: A literature review. Rep. REMR-HY-3, U.S. 

Army Corps of Engineers. Washington, DC. 

- Rimböck A. (2003). Schwemmholzrückhalt in Wildbächen (Driftwood retention in 

mountain torrents). Report 94. TU Munich, Germany (in German). 

- Schmocker L. and Weitbrecht V. (2013). Driftwood: Risk analysis and engineering measures. 

Journal of Hydraulic Engineering 139 (7): 683-695. 

- Schmocker L. and Hager W.H. (2013). Scale modelling of wooden debris accumulation at a 

debris rack. Journal of Hydraulic Engineering 139 (8): 827–836. 

- Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie (VAW). (2008). Ereignisanalyse 

Hochwasser 2005: Teilprojekt Schwemmholz (Analysis of the 2005 flood event: Sub-project 

drift). Bericht 4240. ETH Zurich, Switzerland (in German). 

- Waldner P., et al. (2010). Schwemmholz des Hochwassers 2005 (Driftwood during the 2005 

flood event). Final Rep., Federal Office for the Environment FOEN, Swiss Federal Institute for 

Forest, Snow and Landscape Research WSL. (in German). 

- Wallerstein N.P., Thorne C.R., Abt S.R. (1996). Debris control at hydraulic structures - management 

of woody debris in natural channels and at hydraulic structures. Report prepared for 

the U.S. Army Corps of Engineers, Waterways Experiment Station, Vicksburg, MS. 



Natural hazard induced risk: a dynamic individualised 

approach for calculating hit probability on networks 

Esther Schönthal, MSc 1 ; Margreth Keiler, PD Dr. 2 

ABSTRACT 

In the context of natural hazard and risk assessment in Switzerland, it is common to identify 

the collective and individual fatality risk (Bründl 2009). However, this approach is a static 

perspective. Consequently, the deduced results for the individual mortality risks on transport 

networks (street, railway), only describe the additive independent risk for any person on the 

endangered network section. Being aware of this limitation, it is the aim of this study to 

calculate the total hit probability for a specific single person, who is driving on a network 

section from A to B. Therefore, the calculation of individual risk is adapted and integrated in 

a model, whereby dynamic calculations for different scenarios are possible. The results shows 

that despite the changing of input parameters (speed, simulation time frame, …) the 

differences between the static and the dynamic approach mainly varies related to the hit 

probability. 

KEYWORDS 

natural hazards; risk analysis; hit probability; network analysis; GIS 

INTRODUCTION 

In the context of natural hazard and risk assessment in Switzerland, it is common to identify 

the collective and individual fatality risk (Bründl 2009). This approach, however, is a static 

perspective of the exposure to natural hazards (Bründl et al. 2010, Fuchs et al. 2013). The 

results of this approach, particularly for the fatality risks on transport networks (street, 

railway), only describe the additive independent risk for any person in general on the 

endangered network section. Being aware of this limitation, it is the aim of this study to 

quantify the consequences of a change in perspective from a static to a dynamic individualised 

perspective. Consequently, the objective is to calculate the total hit probability for a 

specific person who is driving on a network section from A to B (Fig. 1). 

The following main aspects are used as a framework for the study: The first step, starting from 

the static approach, is the analysis of how the calculations of hit probability have to be 

adapted to simulate the dynamic individualised perspective, and the subsequent implementation 

of this adapted modelling approach. Accordingly, the second focus lies on the assumed 

difference between the two approaches (static vs. dynamic individualised) by comparing the 

results providing the same baseline. In addition, the influence of the driving direction to the 

1 geo7 AG, Bern, SWITZERLAND, esther.schoenthal@geo7.ch 




esults of modelling the hit probability for the dynamic individualised perspective will be 

clarified. Finally, the system dynamic and sensitivity of the results according to the variability 

of the input parameters is tested. In this paper, only a selection of the overall results is 

presented. 

Fig. 1: Scheme of the static and dynamical individualised approach. 

METHODS 

In a first step, the mathematical formulas will be adapted with the concept of conditional 

probability. Secondly, a network data model system will be constructed for the modelling of 

the different aspects as outlined above. Finally, the planned analysis of sensitivity will be 

realised by variation of the different input parameters. 

1) Adaption of hit probability calculation 

The common calculation of hit probability related to natural hazards for individuals is a 

function of the following input parameters: the spatial appearance probability of the process 

(p(sa) j 

) , the probability of an event (p j 

), the length of the endangered section (g j 

), the speed 

of the individual (v i 

), and the probability of rear-end collision accident p(reca). Furthermore, 

in the case of a risk analysis on a network, two scenarios are defined, namely the direct hit 

(DH) and the ride-into-accident (RA) (Bründl 2009). Because the focus is on the aspects of 


the individual driver on the network, the influence of other decision-makers, who are 

causing closures or early-warning etc., will be ignored. Based on this aspect, the two 

probabilities are calculated per endangered section g j 

and per year (Formula 1). 

Formula 1: Calculation of the probability of direct hit and ride-into-accident out of Bründl (2009). 

All these parameters are also integrated into the adapted calculation for the dynamic 

individualised approach. In order to take the dynamic aspect better into consideration within 

the constructed calculation model, the probability of the ride-into-accident (RA) is completed 

with two elements. First, the spatial appearance probability of the process (p(sa) j 

) is integrated 

into the formula. Second, the probability of rear-end collision accident p(reca) is calculated 

depending on the speed of the individual (vi) deduced from the physical relation between the 

distance of braking deceleration and the speed, and by following the recommended values in 

ASTRA (2012) (Fig. 2). The blue function represents the physical connection of the braking 

distance and the speed of a car with a braking deceleration of 8 m/s 2 . 

Fig. 2: Derivation of the connection between the speed of the individual (v i 

) and the probability of rear-end collision accident p(reca). 

In addition, the assumption was taken; that a probability in the class “intermediate” of the 

classification after ASTRA (2012) nearly corresponds to a speed between 40 km/h and 60 

km/h. Based on these considerations, the relation between speed and the probability of 

rear-end collision accident is deduced. 


Subsequently, based on these formulas for the dynamic calculation, a norm probability for 

each scenario for one meter per hour is formed (Formula 2). The general determination is 

that for this unit the static and dynamic hit probability is identical. 

Formula 2: The deduced norm probabilities of the two scenarios direct hit and ride-into-accident. 

Afterwards the hit probability is calculated for each meter along the simulated path from A to 

B with the integration of the converse probability (the odds that the individual has travelled 

the path before without being hit) (Formula 3). 

Formula 3: The calculation of the hit probability of the dynamic approach. 

The reason for integrating the converse probability into the calculation is to consider the 

aspect that a specific individual moving on the section cannot exceed a hit probability above 

1. Thereby the variable X j 

represents the amount of the different natural hazard processes (z), 

which endangered the specific metre, and the variable l j 

is the length of the part in an 

endangered section with the same combination of natural hazard processes. 

2) Network data model system 

The main aim of the constructed network data model is the possibility to simulate different 

scenarios. In this context, the main requirement is that the data model is adaptable to any 

combinations of the above-mentioned input parameters and usable for any section in reality. 

The simulation model is built in Python, a programming language for geoprocessing (Python 

2015). Basically, the model calculates the hit probability for each metre by passing the path 

between A and B. Thereby, the smallest distance unit in the simulation model is one metre. 

First of all, the real section from A to B is imported in the calculation system. During this 

process a list with the endangered sections and their specific parameters is created. Afterwards, 

a simulation path as a sorted list of metres from A to B is constructed. Thereby each 

metre has the information of which processes endangered it. Finally the hit probability of the 

complete path is simulated through. 

The specific baseline for this study is represented by the section between Interlaken and 

Brienz in Switzerland. This course has a length of 18 kilometres and, overall, is endangered 


y 75 natural hazard processes. Table 1 shows an overview of the situation of exposure by 

natural hazards. 

Table 1: Overview of the situation of exposure by natural hazards for the simulation baseline between Interlaken and Brienz (CH). 

Thereby the endangered sections have been deduced from the natural hazard information 

maps of the canton of Berne (AGI 2015). The verification and further description and 

assessment of these sections are not subject of this study. 

3) Analysis of the results and their sensitivity to input parameters 

Based on the created simulation model, the results have been compared with the static 

approach and the influence of the direction of travel has been quantified. Furthermore, the 

sensitivity of the results dependent on the different parameters has been tested. For this 

purpose, a simple testing workflow has been defined. In a first series of simulations, some of 

the parameters that are dependent on the natural hazard have been varied, while, in a second 

series, the same was done with those parameters that are dependent on the individual. As a 

result, the dependency and the impact of the input parameters on the results have been 

evaluated. 

RESULTS 

For the initial situation S 0 

to compare the two approaches and testing the sensitivity of results, 

the parameters are defined as mentioned in Table 2. 

Table 2: Parameters setting of the initial situation S0 (adapted from Bründl (2009)). 

For the spatial appearance probability of the process, the standard values recommended in 

Bründl (2009) are applied. The probability of an event and the speed of the individual are 

fixed. Furthermore, the individual is passing the course of Interlaken to Brienz twice a day 

(once in each direction) and every day per year (one year = 365 days). That means that the 

simulation is running over a period of one year. 


Fig. 3: The growth of hit probability over one year for the static and dynamic perspectives. 

1) Comparison of the two approaches 

If we compare the results of the two approaches for one year (Fig. 3), it appears that the 

longer the simulation is running the range of the difference increase. 

The hit probability of the static approach shows a continuous growth, which means the rate 

of increase is constant and could theoretically exceed the value 1. In contrast, applying the 

dynamic approach, the rate of increase of hit probability decreases and, over time, it 

approximates the value 1. This behaviour is explained by the increasing converse probability 

factor, which is fed into the calculation of the dynamic approach and prevents the value from 

exceeding the value 1. If we compare the hit probabilities for the two approaches for one day 

and for one year (Table 3), the results indicate that the difference for one day is insignificant. 

However, by simulating over one year the difference is about 17 percent. 

Table 3: Comparison of the results of hit probability for the two approaches for the periods of one day and one year 


Correspondingly, the value of the norm probability plays an important role: The smaller the 

value of the norm probability, the smaller the difference will be after one year. This implies 

that the converse probability factor is growing slower. Furthermore, this connection is also 

reflected by the two different scenarios “direct hit” and “ride-into-accident”, whose values of 

hit probability per day differ by a factor of about 720. 

2) Influence of the direction of travel 

Figure 4 schematically visualise the results of the analysis regarding the influence of the 

direction of travel. Thereby the parameter V x 

represents the incidental costs if something 

happens (red cross). The coloured boxes stand for two different hazardous sections with a 

probability that something happens (p x 

). In the case, that we have no valuation of the event 

(V x 

= 1), only the probabilities are relevant. In this situation the results for driving from A to B 

or from B to A will be the same. Also in the two special cases if the valuation or the probability 

that something happens for the different sections are the same, the driving direction has no 

influence (as mentioned in the calculation in Fig. 4).Yet, if the various hazardous sections are 

evaluated differently (for example in the calculation of risks), then the different values will 

emerge for the two directions of travel. 

Fig. 4: Scheme of the influence of the driving direction for the results. 

3) Sensitivity 

Following the results for the two parameters, the probability of an event (p j 

) and the speed of 

the individual (v i 

) are presented. The speed is varied between 30 and 100 km/h. Figure 5 A 

shows the plotted results from a simulation period over one year. The lower the speed, the 

flatter is the graph of the result. The higher the speed, the more the curvature of the graph 

increases. 


The parameter probability of an event (p j 

) is varied with the common values in Switzerland 

between one event per year and one event per 300 years. Figure 5 B, presents the plotted 

results about a simulation period over one year. The more probable the event is, the steeper 

the graph of hit probability is growing before it is flattening by reaching the value 1. The rarer 

an event is, the flatter the hit probability is growing. Clearly apparent is that the hit probability 

of one individual cannot exceed the value of 1. In contrast, calculating the static individual 

hit probability for the different values of p j 

, the results may exceed 1. 

Fig. 5: A: The growth of hit probability over one year for different individual speed values. B: The growth of hit probability over one year 

for different event probabilities. Blue value: initial situation S 0 

. 

CONCLUSIONS 

Generally, the hit probability for an individual with the dynamic approach is limited to the 

value of 1. A value greater than 1, does not make sense for a probability. This means that the 

dynamic approach generates more realistic results for an individual person. The benefit of this 

approach is that it allows making a statement for the hit probability for a single specific 

commuter and not only for a group of individuals moving on the section. 

The larger the norm probability and the longer the simulation period, the differences increase 

between the static and dynamic approaches. At the level of probabilities, the driving direction 

has no influence. However in situations, where a valuation of the different probabilities is 

condacted, different values result for both directions of travel. 

This study proposes a change to a more dynamic perspective in risk analysis, starting with the 

hit probability on networks. The results according to the dynamic individualised perspective 


will also stimulate a new dimension for the discussion of risk management. Furthermore, the 

developed method features applications for further routing services and other mobile devices, 

and could additionally be applied to questions concerning the insurance sector. Finally, the 

benefit for road operators is the possibility to make more precise statements for single 

commuter with different usage behaviours on their roads. This paper presents an overview 

and some results from an ongoing study. More information and details were published in 

early 2016 (Schönthal, 2016). 

REFERENCES 

- AGI (2015): GK5, Naturgefahrenkarte des Kantons Bern 1:5‘000. Amt für Geoinformation 

(AGI) des Kantons Bern, Bern. 

- ASTRA (Ed.) (2012): Naturgefahren auf den Nationalstrassen: Risikokonzept. Methodik für 

eine risikobasierte Beurteilung, Prävention und Bewältigung von gravitativen Naturgefahren 

auf Nationalstrassen. Ausgabe 2012 V2.10. Bundesamt für Strassen ASTRA, Bern. 

- Bründl M., Bartelt P., Schweizer J., Keiler M., Glade T. (2010): Snow avalanche risk analysis 

- Review and future challenges. In: Alcántara-Ayala I., Goudie A. (eds.): Geomorphological 

hazards and disaster prevention. Cambridge University Press: 49-61. 

- Bründl M. (Ed.) (2009): Risikokonzept für Naturgefahren - Leitfaden. Nationale Plattform 

für Naturgefahren PLANAT, Bern. 

- Fuchs S., Keiler M., Sokratov S., Shnyparkov A. (2013): Spatiotemporal dynamics: the need 

for an innovative approach in mountain hazard risk management. Natural Hazards 64 (3): 

2083-2105. 

- Python (2015): Welcome to Python.org. www.python.org. last accessed 23. November 2015 

- Schönthal E. (2016): Naturgefahrenbedingte Risiken: ein individualisierter dynamischer 

Berechnungsansatz der Trefferwahrscheinlichkeit auf Netzwerken. Masterarbeit UNIGIS, 

Universität Salzburg. 



Flood Risk Map for the Canton of Zurich 

Hochwasser Risikokarte für den Kanton Zürich 

Christian Schuler 1 ; Thomas Egli²; Mirco Heidemann³; Manuela Häni 4 

ABSTRACT 

With the flood risk map, the Office of Waste, Water, Energy and Air (WWEA) and the GVZ 

(Buildings Insurance of the Canton of Zurich) present a tool which enables the identification 

and prioritisation of areas for risk-reducing measures. The risk map closes the gap between 

hazard mapping and the action plan. With increasing risks and ever tighter budgets, it offers 

a fundamental basis for all players working in the field of natural hazards. The tool combines 

various risk types and, for the first time, provides an overview of the effective flood risks in 

the canton of Zurich. 


Mit der Risikokarte Hochwasser legen das Amt für Abfall, Wasser, Energie und Luft AWEL 

und die GVZ Gebäudeversicherung Kanton Zürich ein Instrument vor, mit dem sich der 

Handlungsbedarf zur Reduktion von Risiken erkennen und priorisieren lässt. Die Risikokarte 

schliesst die Lücke zwischen der Gefahrenkartierung und der Planung möglichst effektiver 

Schutzmassnahmen – angesichts steigender Risiken und immer knapperer Budgets eine 

elementare Grundlage für alle Akteure im Naturgefahrenbereich. Das Instrument verknüpft 

verschiedene Risiko-Arten und verschafft erstmals den Überblick über die Hochwasser- 

Risiken im Kanton Zürich. 

KEYWORDS 

risk analysis; flood risk; Risk Map 

AUSGANGSLAGE 

Der Kanton Zürich hat ab 1998 die gemeindeweise Gefahrenkartierung in Angriff genommen, 

ab 2006 nach einem erweiterten Konzept. Die Gefahrenkarten sind mehr und mehr als 

wichtige Arbeitsgrundlage für Kantone und Gemeinden etabliert. Sie zeigen auf, wo eine 

Gefährdung besteht und wo mit welcher Wahrscheinlichkeit und welcher Stärke ein 

Hochwasser oder eine Massenbewegung (Hangmure, Rutschung, Steinschlag etc.) auftreten 

kann. 

Unmittelbar nach Abschluss der Gefahrenkartierung folgt die Umsetzung durch die Gemeinde. 

Einer der ersten Umsetzungsschritte ist die Erstellung einer Maßnahmenplanung. Ist das 

1 Office of Waste, Water, Energy and Air (WWEA), Zurich, SWITZERLAND, christian.schuler@bd.zh.ch 

2 Egli Engineering AG, St. Gallen, SWITZERLAND 

3 GVZ Gebäudeversicherung Kanton Zürich, Zürich, SWITZERLAND 

4 AWEL Amt für Abfall, Wasser, Energie und Luft, Zürich, SWITZERLAND 



Risiko für eine Gemeinde nicht tragbar, müssen Maßnahmen getroffen werden. Die Gefahrenkarten 

liefern keinen Hinweis, wo mit welchen Schäden zu rechnen ist. Daher werden für 

die Maßnahmenplanung und die damit verbundene Priorisierung der Maßnahmen als 

ergänzende Beurteilungsgrundlage Risikoanalysen benötigt. 

Bislang fehlte eine systematische Grundlage für die Beurteilung des Handlungsbedarfs für den 

Hochwasserschutz bei den zuständigen Behörden (Kanton für kantonale und Gemeinden für 

kommunale öffentliche Gewässer). 

METHODIK 

Für die Erarbeitung einer umfassenden Risikoübersicht haben das AWEL und die GVZ 

zusammen mit Experten aus verschiedenen Fachbereichen georeferenzierte Parameter resp. 

Themen nach ihrer vermuteten Auswirkung im Schadenfall eingeordnet. Auf diese Weise 

entstand eine quantitative nicht monetäre Risikoanalyse für den Kanton Zürich, die konkrete 

Hinweise für die Priorisierung und die Maßnahmenplanung liefert. Der Risikoanalyse zu 

Grunde gelegt sind die erarbeiteten Gefahren- und Intensitätskarten für die Hochwassergefährdung 

im Kanton Zürich. 

Fünf übergeordnete Risiko-Arten wurden betrachtet: Versorgungsrisiko, Personenrisiko, 

Kulturgutrisiko, Umweltrisiko und Sachrisiko. Untergeordnet wurden 56 Themenbereiche 

(Schutzgüter) erfasst und bewertet. Dazu gehören Denkmalschutz, Energie, Verkehr, 

Versorgung (Spitäler, Werkhöfe usw.), Kommunikationsinfrastruktur, Bevölkerungsdichte, 

Fruchtfolgeflächen, Schulhäuser, Gebäudeversicherungswert und weitere. Berücksichtigt 

wurden Themen, für welche georeferenzierte Daten vorliegen. 14 Themen konnten aufgrund 

ungenügender Datenqualität nicht berücksichtigt werden. 

Um die Vergleichbarkeit der verschiedenen Risiko-Arten zu gewährleisten, wurden die 

Themenbereiche nach ihrer Bedeutung klassiert. Damit wird das Risiko nicht direkt in 

Franken ausgedrückt. Dieser Ansatz ermöglicht es, alle einbezogenen Risiken miteinander zu 

verknüpfen. Das Berechnungsmodell wurde modulartig aufgebaut. Es kann flexibel erweitert 

und angepasst werden. 

Die Risikokarte zeigt als Resultat pro Hektar, wie groß das Hochwasserrisiko über alle 

einbezogenen Themen ist: groß, mittel, klein oder vernachlässigbar. Wo viele Rasterzellen mit 

mittlerem und großem Risiko beieinander sind, liegt ein so genannter Hotspot vor. Hier 

besteht Handlungsbedarf. 

GEFAHRENGRUNDLAGEN 

Die Gefährdung durch Hochwasser wurde anhand der Daten der Gefahrenkartierung im Kanton 

Zürich (Intensitätskarten HQ 30 

, HQ 100 

, HQ 300 

) sowie der Flächen der Restgefährdung 

(EHQ) gemäß Gefahrenkartierung, Datenstand 31. Juli 2015 ermittelt. 


Risiko-Arten 

Den Risikoarten untergeordnet sind folgende Parameter (Auszug): 

Tabelle 1: Zuordnung einiger Parameter zu den übergeordneten Risiko-Arten. 

Risiko-Art 

Versorgungsrisiko 

Personenrisiko 

Kulturgutrisiko 

Umweltrisiko 

Sachrisiko 

Parameter 

Leitungen und Netze (Strom, Gas, Kommunikation etc.) 

Wasserfassungen, Gewässerschutzbereiche 

Infrastruktur der Feuerwehr, Polizei, Zivilschutz 

Spitäler, Gefängnisse 

Straßen, Bahnen, Flughafen 

Sendeanlagen 

Schulgebäude, Universitäten 

Campingplätze, Familiengärten 

Sportanlagen, Freizeitanlagen 

Einkaufszentren, Multifunktionskomplexe 

Beschäftigte und Bevölkerungsdichte 

Denkmalschutz 

Chemie- und biologische Risiken 

Tankanlagen 

Gebäudewerte 

Fruchtfolgeflächen 

Für eine synoptische Risikokarte müssen die verschiedenen georeferenzierten Daten 

vergleichbar gemacht werden. Die Einordnung der Parameter erfolgte auf einer linearen, 

10-stufigen Skala, die je nach Thema eine unterschiedliche Bedeutung hat. Die zugeordnete 

Klassierung ersetzt in der Risikoberechnung den monetären Wert des Schutzgutes. 

Verletzlichkeit 

Nicht alle Themen weisen bezüglich Hochwasser die gleiche Verletzlichkeit auf. Während zum 

Beispiel bei einer schwachen Intensität Gasleitungen nicht beeinträchtigt werden, ist beim 

Verkehr bereits mit erheblichen Versorgungsengpässen zu rechnen. Den Themen wurden je 

nach Hochwasser-Intensitätsstufe (gering, mittel, stark) Verletzlichkeiten mit einem Wert von 

0 – 1 zugeordnet. Dabei bedeutet 1, dass Hochwasser das Thema komplett beeinträchtigt und 

0, dass für dieses Thema Hochwasser keine Wirkung hat. Wo keine Information zur Hochwasserintensität 

vorhanden war (EHQ), wurde eine pauschale Verletzlichkeit gewählt, welche 

normalerweise derjenigen der mittleren Intensität entspricht. Wo vorhanden, wurden die 

Werte für die Verletzlichkeit aus dem Risikotool des Bundes "EconoMe" verwendet (Quelle: 

BAFU 2013: EconoMe 1.0 Objektparameter, Stand 1.10.2013) 

Schadengrenze 

Durch die Einteilung in die Wiederkehrperioden 30, 100 und 300 Jahre entstehen große 

Sprünge. Ist bekannt, dass ein Fluss bereits bei einem 15-jährlichen Ereignis über seine Ufer 

tritt, erscheint diese Überflutungsfläche trotzdem erst in der Intensitätskarte der Wiederkehrperiode 

‚30 Jahre‘. Mittels einer manuell definierten Schadengrenze wird präzisiert, wann der 

erste Schaden auftritt, also beispielsweise bei 15 Jahren. Somit können für einzelne geografi- 


sche Gebiete unterschiedliche Schadengrenzen definiert werden. Die einheitlichen Schadengrenzen 

wurden wie folgt definiert: 

Tabelle 2: Definition der einheitlichen Schadengrenzen. 

Wiederkehrperiode Schadengrenze 

30 Jahre 15 Jahre 



EHQ (500 Jahre) 300 Jahre 

Damit wird gegenüber dem Risikokonzept der PLANAT (Nationale Plattform Naturgefahren 

der Schweiz) berücksichtigt, dass Schäden ab einer sogenannten Schadengrenze bis zum 

ersten in der Intensitätskarte dargestellten Ereignis ansteigen. 

Abbildung 1: Schadenkurve (blau) als Funktion der Schadensumme und der Wiederkehrperiode sowie Definition der unteren 

Schadengrenze. 

Die blaue Kurve ist die angenommene natürliche Schadenkurve. Das Risiko entspricht dem 

Integral unter der blauen Kurve. In der Berechnung wird diese Fläche mit der Treppenfunktion 

angenähert (grüne Balken). Jede Treppenstufenfläche entspricht dabei einem Risikobeitrag 

und berechnet sich aus dem gemittelten Schaden und der Differenz der Überschreitungswahrscheinlichkeiten. 

Mobiler Hochwasserschutz 

Es gibt Gebiete mit ortsfesten, mobilen Hochwasserschutzmaßnahmen (z.B. Dammerhöhungen 

mit Dammbalken). Gemäß Bundesempfehlung werden mobile Schutzmaßnahmen in der 

Gefahrenkartierung nicht berücksichtigt. Es macht aber Sinn, diese Schutzmaßnahmen bei 

einer Risikoanalyse zu berücksichtigen. Ähnlich wie bei der Schadengrenze wird der 

Einflussbereich dieser Schutzmaßnahmen geografisch definiert. Dazu wird die Wahrscheinlichkeit 

pro Wiederkehrperiode festgehalten, bei welcher das System korrekt funktioniert. Bei 


Faktor 0 resultiert dasselbe Risiko, wie wenn keine Maßnahme eingesetzt würde, bei Faktor 1 

wird kein Risiko mehr ausgegeben. 

PROZESSING 

Die gesamte Datenverarbeitung wurde in ArcGIS 10.2 for Desktop Basic durchgeführt. Für 

sämtliche Prozessierungsschritte wurden im ModelBuilder Modelle erstellt, damit die 

Arbeitsschritte einfach reproduzierbar sind. Für einzelne Verarbeitungsschritte ist die 

Erweiterung Spatial Analyst notwendig. Ergänzend wurden, insbesondere für die Risikoberechnung, 

Scripte in Python erstellt. Für die Erstellung und das Debugging wurde PyScripter, 

Version 2.7 verwendet. Die Aufbereitung von Linien- und Flächengeometrien in Punkteobjekte 

wurde mittels XTools Pro, Version 10.0 durchgeführt. 

Aufbereiten der Gefahrengrundlagen 

Dieser Prozessschritt muss nur einmal für alle Themen durchlaufen werden. Hierbei werden 

die Daten der Intensitätskarten mit der pro Jährlichkeit unteren Schadengrenze ergänzt. 

Damit wird berücksichtigt, dass ein Prozess häufiger auftreten kann als die entsprechende 

Intensitätskarte angibt (Beispiel: 150 anstatt 300 Jahre). Weiter kann der Einfluss mobiler 

Hochwasserschutzmaßnahmen einbezogen werden (siehe Formel 1). 

Umwandlung der Geodaten in Hektarraster 

Ziel ist es, alle Daten in Punktinformationen zu überführen. Punktdaten (z.B. Sendemasten) 

können direkt berechnet werden. Linienobjekte (z.B. Versorgungsleistungen) werden in 

Segmente unterteilt, die durch einen Schwerpunkt repräsentiert werden. Flächendaten (z.B. 

Flughafen) werden in Hektarflächen zerschnitten, die ebenfalls mit ihrem Schwerpunkt in die 

Berechnung einfließen können. 

Risikoberechnung 

Alle Parameter werden mit den aufbereiteten Grundlagendaten verknüpft und erhalten ihre 

Attribute wie Gefährdung, Auftretenswahrscheinlichkeit undSchadengrenze Danach werden 

die Verletzlichkeiten zugeordnet. Diese Schritte werden einzeln pro Thema und Wiederkehrperiode 

durchlaufen. Die Berechnung des Gesamtrisikos ist die Summe der pro Hektarzelle 

berechneten Einzelrisiken. 

Algorithmus 

Ein Risiko berechnet sich aus einem Schaden und der Wahrscheinlichkeit, dass dieser 

Schaden eintrifft. Die Wahrscheinlichkeiten eines (Hochwasser-) Ereignisses ist der Kehrwert 

der Jährlichkeit und bezeichnet, wie häufig im statistischen Mittel ein Ereignis auftritt. Das 

gesamt Risiko setzt sich aus den Risikobeiträgen der einzelnen Ereignissen zusammen. 


Berechnungsansatz: 

Formel 1: Ansatz für die Berechnung des Risikos. 

wobei: 

SE 

W 

Aw ij 

i 

j 

p i 

S ij 

R ij 

Verletzlichkeit 

Wert (Klassierung) 

Faktor zur Anpassung der Wahrscheinlichkeit 

Wiederkehrperiode 

Objekt 

Überschreitungswahrscheinlichkeit (Häufigkeit) 

Schaden für Objekt j bei Wiederkehrperiode i 

Risiko 

Aufbereitung und Darstellung 

Die Datenaufbereitung erfolgt für jede Hektarzelle mit dem Gesamtrisiko. Je nach Höhe des 

Wertes wird die Hektarzelle einer Farbe zugeordnet (Ziel war es, dass aus der Risikokarte 

Hochwasser für den gesamten Kanton 10 - 20 Hotspots herausstechen. Daraus folgt, dass die 

obersten 15 % der Hektarflächen dunkelrot sind). 

Tabelle 3: Festgelegte Definition der Risikoklassen (klein, mittel, groß). 

Wert / Quantil Aussage Farbe 

Zellen ohne Nach aktuellem Kenntnisstand keine weiß 

berechnetes Risiko Gefährdung vorhanden oder keine Werte 

analysiert. Kein oder vernachlässigbares 

Risiko. 

60 % - 85 % Großes Risiko vorhanden dunkelrot 

ERGEBNISSE 

Die „Risikokarte Hochwasser Kanton Zürich“ verschafft einen Überblick des Hochwasserrisikos 

auf kantonaler Ebene. Wo sich von Hochwasser gefährdete Gebiete mit einem untersuchten 

Schutzgut überschneiden, wird die Rasterzelle in hellrot, rot oder dunkelrot dargestellt. 

Erkennbar werden neben dem Schadenpotenzial an Sachwerten, wie viele Menschen 

betroffen sind, ob wichtige Verkehrsverbindungen beeinträchtigt sind, ob und in welchem 


Ausmaß mit Versorgungsunterbrechungen zu rechnen ist oder inwiefern wichtige Einrichtungen 

der öffentlichen Infrastruktur geschädigt werden könnten. Damit zeigt die Risikokarte 

nebst monetären Risiken auch schwer quantifizierbare Risiken, auch für die Umwelt und 

Kulturgüter. 

Die Berechnungen zur Risikoanalyse sind in einem Modell für ArcGIS von ESRI verfügbar. 

Ausgegeben werden berechnete Risiko-Werte zwischen 0 und 1.5. Die Resultate können als 

Datentabelle (Ranking von Themen in den Gemeinden) oder als Karte (strategisches 

Hilfsmittel) genutzt werden. 

Bis zu einem Wert von 0.00081742 (entspricht 60 % der Anzahl Rasterzellen) werden die 

Rasterzellen hellrot dargestellt. 

Tabelle 4: Abgrenzung der Risikoklassen. 

Wert Farbe Aussage 

0 weiß vernachlässigbares oder kein Risiko 

0 - 0.0081742 hellrot geringes Risiko 

0.0081742 - 0.00708 rot mittleres Risiko 

0.00708 - 1.5 dunkelrot großes Risiko 

Die Risikokarte zeigt, dass in einem Großteil der gefährdeten Gebiete gemäß Gefahrenkarte 

auch ein Schadenpotenzial vorhanden ist und somit ein Risiko resultiert. Die Ausprägung 

unterscheidet sich jedoch an vielen Stellen von der Gefahrenkarte. 

Trotz geringer Auflösung können die Gemeinden von der Risikokarte profitieren. Die 

Hotspots bilden die Basis um Prioritäten zu setzen und nach einer genaueren Betrachtung 

resp. detaillierteren Analyse Schutzmaßnahmen zu planen. 

Für die Maßnahmenplanung ist es notwendig sich ein genaues Bild der Situation zu machen 

und folgende Fragen zu klären: Welche Risiken bestehen in diesem Gebiet? Welches sind die 

Ursachen? Erst die detaillierte Analyse macht deutlich, wo tatsächlich Bedarf für einen 

besseren Schutz besteht. 

Eine wichtige Erkenntnis der Risikokarte ist, dass die Höhe des Risikos vor allem von der 

Nutzung und weniger von der Gefahr bestimmt wird. Hohe Risiken liegen nicht nur in den 

Bereichen, die auf der Gefahrenkarte mit einer mittleren oder hohen Gefahrenstufe ausgewiesen 

werden. Häufig treten hohe Risiken in Gebieten geringer Gefährdung oder sogar in 

Flächen der Restgefährdung auf. 

Diskussion 

Mit der vorliegenden Analyse ist es gelungen, für den Prozess Hochwasser eine Risikoübersicht 

über den ganzen Kanton Zürich zu erstellen. Sie steht der Allgemeinheit als Layer im 

Kantonalen Web-GIS und als Webdienst (wfs und wms) zur Verfügung. 


Abbildung 2: Der Hauptbahnhof der Stadt Zürich (grüner Punkt) ist ein bedeutender Verkehrsknotenpunkt. Auf der Gefahrenkarte 

(Abbildung links) weist er eine geringe Gefährdung auf (gelbe Flächen). Gemäß Risikokarte (Abbildung rechts) besteht hingegen eine 

Häufung großer Risiken (Quellen: www.maps.zh.ch/naturgefahren; www.maps.zh.ch/risikokarte). 

Die Risikokarte Hochwasser schließt für den Kanton Zürich die Lücke zwischen der Gefahrenkarte 

und der Maßnahmenplanung und verschafft eine systematische Übersicht über die 

Hotspots. 

Die Höhe des Risikos in jeder Rasterzelle ist die Summe von 42 berechneten Einzelrisiken 

(siehe Themen). Werden die Grenzen der qualitativen Einteilung in der Risikokarte geändert, 

kann das Gesamtbild stark beeinflusst werden. 


Die Risikokarte zeigt mit den Hotspots auf, wo große Risiken resultieren und somit besonders 

hoher Handlungsbedarf für den Hochwasserschutz besteht. Eine konkrete Kosten-Nutzen-Analyse 

kann mit den Daten der Risikoanalyse jedoch nicht durchgeführt werden, weil 

die Resultate nicht in CHF/Jahr sondern in qualitativer Form vorliegen. 

Verantwortungsträger haben konkret folgenden Nutzen: 

– Visualisierung, wo der größte Handlungsbedarf besteht 

– Möglichkeit zur systematischen Priorisierung über die Ortung von Hotspots 

– Transparente, reproduzierbare und objektive Kriterien für Entscheidungen 

Um genaue Aussagen machen zu können (z.B. auf Gemeindeebene), muss die Analyse auf 

lokaler Ebene entsprechend verfeinert werden. Lokale Kenntnisse müssen einfließen und 

insbesondere die effektive Verletzlichkeit von einzelnen Objekten mit hohem Wert / 

Bedeutung sollte verifiziert werden. Hinzu kommen eventuell lokal wichtige Risiken, welche 

in der vorliegenden Analyse nicht berücksichtigt wurden. 


Abbildung 3: Der Hauptbahnhof der Stadt Zürich (grüner Punkt) ist ein bedeutender Verkehrsknotenpunkt. Auf der Gefahrenkarte 

(Abbildung links) weist er eine geringe Gefährdung auf (gelbe Flächen). Gemäß Risikokarte (Abbildung rechts) besteht hingegen eine 

Häufung großer Risiken (Quellen: www.maps.zh.ch/naturgefahren; www.maps.zh.ch/risikokarte). 


HERAUSFORDERUNGEN 

Um dem Anspruch einer gesamtheitlichen Risikoanalyse gerecht zu werden, wurden 

verschiedene Risiko-Arten in die Berechnung integriert und auf einer einzigen Karte 

dargestellt. Da einige Themen nur schwer monetarisierbar sind, wurde zusammen mit den 

Fachverantwortlichen eine Klassierung der Themen vorgenommen. Diese Einteilung erfolgte 

nach subjektiven Kriterien, die vermutlich schwer reproduzierbar ist. 

Bei der Beschaffung und Sichtung der Daten hat sich gezeigt, dass deren Qualität sehr 

unterschiedlich ist. 14 interessante Datensätze konnten deswegen nicht berücksichtigt 

werden (z.B. Bodenbedeckung und Haltestellen). 

Die jetzt vorliegende Risikoübersicht erhebt keinen Anspruch auf Vollständigkeit. Weitere 

Themen können in eine spätere Version der Risikoanalyse integriert werden. 

Mit jeder neuen Gefahrenkarte kann auch die Risikokarte nachgeführt werden. Andere 

Risikoanalysen können mit der Risikokarte Hochwasser Kanton Zürich verglichen werden. 

LITERATUR 

- Egli Engineering AG (2014). Risikoanalyse Hochwasser Kanton Zürich, Schlussbericht. 

«Zürcher UmweltPraxis» ZUP (2014): Risikokarte Hochwasser Kanton Zürich, zur Naturgefahrenprävention. 

- PLANAT (2009): Risikokonzept für Naturgefahren, Leitfaden, www.planat.ch 



A probabilistic approach to flood hazard assessment 

and risk management in floodplains considering levee 

failures 

Silvia Simoni, Ph.D. 1 ; Gianluca Vignoli, PhD 2 ; Bruno Mazzorana, PhD 3 ; Claudio Volcan, M.S. 4 ; Francesco Maria Cesari, M.S. 5 

ABSTRACT 

Levees were originally built to confine rivers to a narrow straight path and to protect the 

surrounding territories from floods. Those territories which originally consisted mainly of 

rural areas, now have become productive areas. Levees can be a defence measure against 

flood as well as a threat when they fail. For this reason a comprehensive flood hazard 

assessment in floodplains can not proceed ignoring levee failures. 

Failures can occur through several mechanisms; a deterministic approach limits the result 

robustness, because the number of variables is high compared to available data. A probabilistic 

approach allows for a better handling of uncertainty, providing a robust frame to build 

flood scenarios. 

A methodology to tackle levee failures in flood hazard assessment is proposed within a 

semi-probabilistic framework and applied to a 53km-section of the Adige river in South Tyrol, 

Italy. This methodology encompasses a probabilistic levee stability analysis, a deterministic 

propagation of the flood and the probabilistic combination of possible scenarios. The results 

are a useful tool for hazard assessment and flood risk management. 

KEYWORDS 

Levee failure; flood hazard; flood mitigation; semi-probabilistic approach 

INTRODUCTION 

Floods have recently and historically occurred in European floodplains causing damaging and 

casualties (Foster, 2000; Barredo, 2009). Furthermore, recently, in the Po plain, in Italy, 

levees have failed with serious social and economical consequences (Govi and Turitto, 2000; 

Moasero et al 2012; Domeneghetti et al., 2013). Embankments and levees were originally 

built to confine rivers to a narrow straight path and to protect the surrounding territories 

from floods (Aschbacher et al., 2014). Those lands which originally consisted mainly of rural 

areas, now have become productive and industrialized areas, where cities and infrastructures 

have been built. From an economical and social perspective it is clear that those areas carry a 

1 Mountain-eering srl, Bolzano, ITALY, silvia@mountain-eering.com 

2 CISMA srl, Bolzano, ITALY 

3 Universidad Austral de Chile - Faculdad de Ciencias - Instituto de Ciencias Ambientales y Evolutivas, CHILE 

4 Hydraulic Engineering Autonomous Province of Bozen, ITALY 

5 hydro's - ingegneri associati, Bolzano, ITALY 



significant wealth which must be protected. Unfortunately experience shows that levees can 

be a defense 

measure against flood as well as a threat when they fail (Apel et al., 2009). For this reason a 

comprehensive flood hazard assessment in floodplains can not proceed ignoring levee failures. 

This involves the development of a methodology able to detect levee weaknesses and accounting 

for them. The physical processes to investigate are twofold: geotechnical and hydraulic. 

The breach development and formation within the levee body is determined by two factors: 

the hydraulic forcing and the geotechnical structure of the levees themselves, in terms of soil 

types and characteristics and layering conditions. (Morris et al. 2007; Morris, 2009) The first 

can be acquired by field investigations, such as boreholes through which soil samples can be 

collected and analyzed. Grain size distribution and hydraulic permeability are important 

parameters to assess levee stability. The latter can be investigated using in-situ penetration 

tests. Failures can occur through several mechanisms, such as piping, overtopping, erosion, 

slide of a portion of the slope, etc (Nagy and Tòth, 2005; Flor et al., 2010). In modeling the 

breaching process a challenge is related to understand the mechanisms with which the breach 

originates and how this relates to the discharge rate in the floodplain. In nature the breach 

failure is not an instant process, however in the modeling process this can be an assumption 

when the focus is on its effect on the floodplain. Given the complex reasnobable nature of the 

process, a deterministic approach strongly limits the robustness of the results, simply because 

the number of variables is too high compared to to the knowledge normally available for 

practical purposes (Mazzorana et al., 2009). On the contrary, a probabilistic approach would 

allow for a better handling of uncertainty, providing a robust frame for building scenarios. 

Here a methodology is proposed to tackle levee failures in flood hazard assessment within a 

probabilistic framework; the methodology is then applied to a 53km-long reach of the Adige 

river between Terlano, Bolzano and Salorno in South Tyrol. The results proved to be a useful 

tool not only for hazard assessment, but also for flood risk management. 

METHODOLOGY 

In this section we outline the theoretical framework for the proposed methodology. In 

general, breaching phenomena may be triggered by several physical processes which are 

highly unpredictable; for this reason deterministic approaches fail to describe them appropriately. 

This is due to several factors, among which: the difficult assessment of the geotechnical 

properties of the embankment and of the subjacent soil layers, the saturation condition of the 

embankment, the duration of the flood event and hydrograph’s shape. A probabilistic 

approach allows for uncertainty to be accounted for. The goal of the outlined methodology is 

the production of a reliable semi-probabilistic flood intensity map in a flood plain protected 

by earthen levees, given the hydrologic forcings and the levee characteristics (geometry and 

geotechnical properties), to be used as a tool for planning purposes. The semi-probabilistic 

flood intensity is given in terms of the maximum possible values of water depth and water 


velocities on the flood plain, obtained taking into account the occurrence probability of the 

flood, the occurrence probability of the levee failures and deterministic informations, such as 

the digital elevation model, levee geometry etc. The methodology is based on some reasonable 

assumptions: 1) breaches are statistically independent of each other; this implies that the 

occurrence of a breach will not influence the occurrence probability of breaches in other 

places along the river. This hypothesis accounts for the difficulty predicting a priori where the 

first breach occurs, in the context of scenarios made up of more breaches. 2) The width of the 

breach is assumed to be known. This parameter is estimated from statistical analyses of real 

breaches occurred during historical flood events (Nagy, 2006) 3) The breach is assumed to be 

instantaneous since for long-term maps (maximum flooding surface) short-time scale breach 

dynamics are not relevant (Fujita, 1987). 4)The flood hydrograph is assumed to be deterministically 

known in its essential characteristics, i.e. shape and peak which are estimated from 

data for a given return period. (Autonome Provinz Bozen, 2008 ) 

The methodology is applied to the river of interest and to its surrounding floodplain; it 

requires several input data, such as the hydrological forcing (i.e. 1-in-K-flood hydrograph), 

the digital elevation model of the flood plain, the levee geometry and geotechnical characterization. 

Levees can fail as a result of structural damage to levee itself (or to its foundation) and/or to 

hydraulic forces. Among all failure mechanisms two are particularly likely to occur in rivers 

running in floodplains and confined by earthen levees: overtopping and underseepage. 

Failures due to overtopping can occur when the overspilling flow erodes the levee's crest; 

PROBABILISTIC + LSD APPROACH 

Resistance 

(R) 

Forcing (F) = 

applied force 

likely failure 

Resistance and applied force [kN] 

Figure 1: Probabilistic approach to compute levee failure using the LSD method. 


failures due to underseepage can occur when the difference in hydraulic head between the 

waterside and the landside of the levee increases; as a consequence water starts to flow 

through the permeable foundation materials forming internal channels through erosion 

processes. These two mechanisms have been investigated in details and several approaches 

have been developed to compute the critical conditions for levees failure. In this work we 

chose the Vrouwenvelder (2001a) model to describe the critical overtopping discharge, the 

Bligh-Lane (1935) approach to identify the most critical cross sections along the river, in 

terms of proneness to underseepage processes, and the more detailed USACE (2000) and 

Sellmeijer (2006) (see Figure 3) methods to investigate those sections in details. Since 

geotechnical and hydraulic parameters necessary to apply the aforementioned methods are 

affected by uncertainty, the methodological setting was modified accordingly. 

The methodology requires the computation of fragility curves for the two described mechanisms. 

These curves define the breaching probability for a certain cross section, given the 

hydraulic forcings. Fragility curves for discrete levee cross sections and for the most likely 

failure mechanisms are computed based on a geotechnical model built for the levee body and 

its foundations from available data. Fragility curve calculations are based on a resistance 

analysis for the earthen levee identifying a critic state beyond which the levee loses its 

functionality. According to the limit state design approach (LSD), the critical state is a 

function of the resistance force R and the applied load F. These variables (R, F) are treated as 

stochastic variables described by probability distribution functions (Gaussian), characterized 

by a mean value and by a given variance. According to this approach the critical state, i.e. the 

levee failure, is more likely to occur when the curves describing the probability distribution of 

the two variables F and R partially overlap (Figure 1). The failure probability P is thus 

computed as P(Z

3 

h c = 3 √ 

2.78 · q 2 c 

g 

equation 2 

The meaning of the parameters is the following: c g 

(ms) is a coefficient representing the 

erosion endurance of grass, P t 

is a percentage indicating the ratio overtopping time to flood 

duration, t s 

(hours) is the flood duration, k is a roughness Strickler-type-factor of the riverside 

slope and α (degree) is the riverside slope angle. hc in equation 3 is the critical head corresponding 

to the critical overtopping flow q c 

. The model is built on a classification of the 

levee's mechanical features (e.g. presence and quality of grass coating, root depth, thickness 

of the fine material layer) and geometry. Assigning each parameter a probability distribution 

function (in terms of mean and variance) instead of a deterministic value and applying the 

propagation error theory, a failure probability due to overtopping was calculated for each 

section of the levees. 

For the underseepage case Z=i c 

-i A 

, Z[-] is the difference between the critical gradient i c 

[-] and 

i A 

which is the hydraulic forcing gradient evaluated applying the USACE (2000) or Sellmeijer 

(2006) formulation. Uncertainties associated to the aforementioned hydraulic variables, i.e. 

h c 

, h A 

, i c 

and i A 

, originate from the simplifications adopted in the modeling approach, such as: 

P3 

R3 

scenario S1: P(S1)=P 1 *P 2 *P 3 *P Q 

1-P3 

P2 

R2 

no R3 

scenario S2: P(S2)=P 1 *P 2 *(1-P 3 )*P Q 

P3 

1-P2 

R3 

scenario S3: P(S3)=P 1 *(1-P 2 )*P 3 *P Q 

P1 

no R2 

1-P3 

R1 

no R3 

scenario S4: P(S4)=P 1 *(1.P 2 )*(1-P 3 )*P Q 

PQ 

HQ Tr 

1-P1 

P3 

no R1 

P2 

R3 

scenario S5: P(S5)=(1-P 1 )*P 2 *P 3 *P Q 

R2 

1-P3 

no R3 

scenario S6: P(S6)=(1-P 1 )*P 2 *(1-P 3 )*P Q 

1-P2 

no R2 

P3 

R3 

scenario S7: P(S7)=(1-P 1 )*(1-P 2 )*P 3 *P Q 

1-P3 

no R3 

scenario S8: P(S8)=(1-P 1 )*(1-P 2 )*(1-P 3 )*P Q 

Figure 2: Sketch for the calculation of scenarios probability. 


estimation of the channel roughness, approximations in the topography, estimation of the 

grass root depth, and spatial variability of soil parameters . Uncertainties associated to 

dependent variables have been computed through the propagation of independent variable 

errors. 

For the computed N breaches, M failure scenarios are built by combining the breach 

formation in all possible manners (i.e. M = 2N). Given the N possible breaches, the scenario Si 

is characterized by the occurrence of Bi breaches and the nonoccurrence of (N-Bi) breaches. 

The probability of each scenario is then computed by multiplying the corresponding breach 

failure/non-failure probabilities, as shown in figure 2. The non-failure probability is the 

complement to 1 of the relative failure probability. 

To compute the corresponding flow depth and the velocity components within the floodplain, 

for each scenario hydrodynamic simulations carried out. From a practical and operational 

standpoint N+1 numerical simulations have to be performed considering the N levee failures 

separately, the “+1” simulation refer to the case of no failure. 

Since each point of the floodplain can be flooded by water coming from different breaches, 

the flood intensity (flow depth and velocity) is computed as the maximum water depth (and 

velocity) value among all simulation results which can cause a flood in that point. For 

example, if in the s-th scenario (out of M) a point in the floodplain (Point A) is flooded by 

water coming from two breaches (B1 and B2 out of N) the resulting flood map in point A is 

calculated as the maximum between the water depth values determined by each of the 

breach considered singularly (the calculation is repeated twice, first for the water depth and 

then for the flow velocities). 

h (s) 

(i,j) = max[h(b) (i,j) ] 

equation 4 

where h {s} is the water depth evaluated at point (i,j) of the floodplain, s, refers to the s-th 

(i,j) 

scenario, h {b} is the water depth at the point (i,j) due to the bth breach, and YsBreaches is 

(i,j) 

the number of breaches, which characterizes the s-th scenario, The resulting maps displaying 

flow depths and water velocities, obtained by taking into account the whole set of possible 

scenarios, are computed as a weighted average of the local water depths (and velocity) and 

the associated scenario probabilities. 

b =1,...Y(s) Breaches 

where H{i,j} is the water depth in the (i,j) point of the flood plain and Ps is the sth scenario 

probability. The same procedure is used for local velocity computations. 

∑ s=M 

s=1 

H (i,j) = 

h(s) (i,j) · T (s) 

(i,j) · P (s) 

T (s) 

equation 5 

∑ s=M 

s=1 T (s) 

(i,j) · P (s) 

(i,j) = { 

0 

if h 

(s) 

(i,j) =0 

1 if h (s) 

(i,j) > 0 


A CASE STUDY 

The methodology was applied to a 53-km-long reach of the Adige river between Terlano, 

Bolzano and Salorno in South Tyrol, Italy. The Adige river was artificially altered in the XIX 

century and now it runs almost straight along the Adige valley. Artificial earthen levees have 

been built and reinforced several times during the last century; in some parts they are 7m 

height. In the seventies along the Adige valley a highway was built partially on embankments 

that sometimes run parallel to river levees. At the gauging station of Bronzolo (15 km south 

of Bolzano) the watershed area is 6923 km², its elevation ranges from 3905 m a.s.l to 220 m 

a.s.l.. The peak discharge ranges from 1400 m³/s (1-in-30-year-flood) to 1832 m³/s (1-in-200- 

year-flood). The flood waves typically last for 90 hours. The hydrodynamic model was 

calibrated for roughness using historical series of flow data. 

The geometry of the Adige river had been previously surveyed and cross sections every 250 m 

were available. The topography of the floodplain was accurately described in order to 

reproduce topographical elements that can interfere with the flood, such as road and railway 

embankments, underpasses, etc. A good deal of data is available for the geotechnical 

characterization of the levees since in the last 15 years the Hydraulic Department of the 

Province of Bolzano (HDPB) carried out hundreds of boreholes, DPH, SPT tests and in-situ 

permeability tests. Furthermore laboratory analyses were carried out to obtain grain size 

distributions from several soil samples. The 23-km-long stretch between Terlano and Bolzano 

(North reach) is characterized by earthen levees with a maximum height of 3 m, whereas the 

30-km-long stretch between Bolzano and Salorno (South reach) is characterized by taller 

levees, with a maximum elevation of nearly 7m. Observed levee failure data, collected during 

the last two centuries along the Adige river, suggest that in the North reach overtopping is the 

major cause of levee failure, on the contrary underseepage processes are predominant along 

the South reach. 

From the available data a resistance model for both left and right levees was applied to 

calculate the critical overtopping flow triggering the erosion of levee crest, q c 

, using equation 

2 and 3 (Vrouwenvelder 2001a-b). Assigning each parameter a probability distribution 

function, in terms of mean and standard deviation) instead of a deterministic value and 

applying the propagation error theory, a failure probability due to overtopping was calculated 

for each section of the levees. A total of 11 levees segments were identified subjected to 

overtopping. Similarly, a geotechnical profile for the levees was built using data from DPH, 

SPT, boreholes and grain size distribution curves (Figures 4 and 5) to assign appropriate 

parameters to the the Bligh-Lane equation, translated into a probabilistic approach, with the 

aim of identifying the reaches more prone to underseepage. To those reaches the models of 

Sellmeijer (2006) and USACE (2000) (Figure 3 and 7) were applied, according to the 

sequence of subsoil layering. Hydraulic permeability values were derived from Lefranc tests. 


USACE (2000) 

L x 3 

H 

A 

1 

M 

B 

z 

d 

k 

k f 

b 

S 

L 2 

SELLMEIJER 

(2006) 

Figure 3: Sketch for the application of USACE 2000 and Sellmeijer 2006 models. 

Figure 4: Geotechnical profile built on borehole and DPH data provided by HDPB for a longitudinal section along the river (failure SX2). 


● 

● 

P_rotta per piping 

P_piping 

● 

●● ● ● ●● ● ● 

●●● 

●● ● 

●● 

● ●●● ● 

● ● ● ● ●● 

● ● 

● ● ● ● ● ● ● 

● ● ● ●●● ● ● 

●● ● ●● ● ● ● 

● 

● 

● 

● ● 

● 

Curva di fragilità − sifonamento Q200 

● ● 

● ● 

● 

● 

● 

● 

●● 

● 

● ● ● 

● 

● 

● 

● 

● 

●● 

● ● ● 

● 

● 

● 

●● 

● 

● 

● ● 

● 

● 

● 

● 

● 

● 

●● 

● 

● 

● 

● ● 

● ● 

●● 

● 

● 

● 

●● 

● 

● 

● 

● 

● 

● 

●● 

● ● 

● 

● 

● 

●● 

● 

● 

● 

● 

●● 

●● 

● 

● 

● 

● 

●●●● 

● ● 

●● 

● 

● 

● 

● 

● ● ● ●● 

●● ● 

● 

● 

● 

● ●● 

● 

● ● 

● 

● 

● 

● 

● 

●●● 

● ● 

● 

● 

● 

●● 

● 

● 

● 

●●● 

● ●● 

● 

● 

● 

●● 

● 

● 

● 

● 

●● ● 

● ● ●● ● ● 

● ●● 

● ● 

● 

● 

● 

● 

● 

● 

● 

● 

● 

●● 

● 

● 

● 

● 

● 

● 

● 

Figure 5: Geotechnical profile built on borehole and DPH data provided by HDPB for a cross section within the transect of depicted in 

figure 4 (failure SX2). 

failure probability [-] 

Probabilita di rottura [−] 

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 

cantilever + road 

embankment 

● ● 

● ●● ●●● ● 

● ● 

● ● 

no reinforcement 

● 

● ● ● ● 

● ● 

● ● ● 

cantilever 

0.0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3.0 3.3 3.6 3.9 4.2 4.5 4.8 5.1 5.4 5.7 6.0 6.3 6.6 6.9 

dislivello di carico idraulico [m] 

hydraulic head [m] 

TR30 TR100 TR200 

DX3 16% DX2 58% DX2 75% 

SX1 7% DX3 25% DX3 26% 

SX2 15% DX4 75% DX4 80% 

DX5 80% DX5 81% 

SX1 17% SX1 18% 

SX2 45% SX2 46% 

 

 

 

 

 

 

Figure 6: Fragility curves for all the cross sections along the left reach of the studied river Adige portion. The three curves indicate 

groups of sections where the levees was reinforced with cantilevers, other where a road embankment was also present in addition to the 

cantilever and other which were no reinforced. Overview of the weakest points along the Adige river, South of Bolzano. Failure probability 

for return period of the flood and for each weak point along the South reach of the Adige river. 


x WSE TR30 WSE TR100 WSE TR200 Z pc L D p [%] p [%] p [%] 

[km] [m s.l.m.] [m s.l.m.] [m s.l.m.] [m s.l.m.] [m] [m] TR30 TR100 TR200 

118.423 219.90 220.67 220.75 214.82 33.4 12.2 7 17 18 

118.473 219.85 220.62 220.70 214.71 35.7 12.2 5 12 13 

118.519 219.81 220.58 220.66 214.63 35.7 12.2 5 13 14 

118.564 219.78 220.55 220.62 214.50 37.2 12.1 4 11 12 

118.609 219.74 220.50 220.58 214.65 37.0 12.0 3 9 10 

118.655 219.70 220.45 220.52 214.66 36.9 12.0 3 8 9 

118.711 219.66 220.41 220.48 214.53 35.2 11.9 5 12 13 

118.721 219.61 220.35 220.42 214.52 36.1 11.8 4 10 11 

118.731 219.61 220.35 220.41 214.63 36.4 11.8 3 8 9 

118.756 219.61 220.35 220.42 214.64 36.3 12.3 3 9 9 

118.791 219.61 220.35 220.41 214.45 36.0 12.8 5 12 13 

118.837 219.55 220.29 220.36 214.52 35.2 13.3 5 13 14 

118.882 219.54 220.27 220.33 214.54 34.1 13.4 6 15 16 

.13.: Probabilità Figure 7: di Summary rotta arginale of the parameters in corrispondenza and the probability della of failure rotta computed SX1-Egna. for the Il failure significato SX1 for dei three simboli return period è il seguente: (RP 30, 100, 200 x chilometrica, WSE 

forzante 

year). 

idraulica 

x: position 

(livello 

of the cross 

della 

section, 

superficie 

WSE: water 

libera 

surface 

in alveo), 

elevation, 

Z 

Zpc: elevation of the floodplain, L: width of the left levee, D: 

pc quota del piano campagna (al piede dell’argine), L larghezza dell’argine, 

thickness of the foundation beneath the levee, p: failure probability. 

D spessore del terreno di fondazione e p probabilità di rotta. 

An example of longitudinal section of the geotechnical model is shown in Figure 4 for a reach 

along the left levee in southern part of the study area (named SX2, see Figure 6). An example 

of cross section along the same reach is shown in Figure 5. Applying the limit state design and 

a probabilistic approach, i.e. considering the limit state variables described by a pdf, fragility 

curves where computed for each cross section of the study stretch (Figure 6). In this analysis 

reinforcements of the levees were accounted for, such as cantilevers and lateral road 

embankment. Figure 7 displays a summary of the parameters and the probability of failure 

computed for the reach illustrated in Figure 5 for three return period (1 in 30, 100, 200 year 

flood). Figure 6 provides also an overview of the weakest points along the Adige river, South 

of Bolzano assigning each probable failure its probability (DX stands for right levee, SX stands 

for left levee). For 1-in-30-year flood failures due to underseepage can occur at 3 locations, 

for 1-in-100-year flood and 1-in-200-year flood the weak points are 6. The corresponding 

scenarios are 8 for 1-in-30-year flood, 64 for 1-in-100 and 1-in-200-year flood. 

Flood map calculations have been performed applying the described methodology, considering 

the hydrologic forcing for each return period of the flood, the levee failure characteristics 

and probability. Each numerical simulation refers to a state characterized by the presence of 

only one levee failure at a time (b in equation 4); numerical results were then combined 

using equation 4 in order to evaluate flow depth and flow velocity in the floodplain for each 

scenario. Finally the 1-in-K-year depth and velocity maps in the floodplain were computed 

using equation 5 over the M scenarios identified for the 1-in-K-year-flood. These maps 

summarize information related to the combination of M failure scenarios. Deterministic 

information comes from flooding patterns, which have been analyzed in details applying a 

hydrodynamic model, while probabilistic information comes from the levee failures probability. 

The final maps represent a synthetic set of information related to the expected effects on 

the flood plain for a given return period of the flood. 


± 

Water depth [m] 

0.04 - 0.25 

0.26 - 0.50 

0.51 - 0.75 

0.76 - 1.00 

1.01 - 1.50 

1.51 - 2.00 

2.01 - 2.50 

2.51 - 3.00 

3.01 - 3.50 

3.51 - 4.00 

4.01 - 4.50 

4.51 - 5.00 

Figure 8: Semi-probabilistic water depth [m] for the Adige floodplain due to 1-in-200-year flood. 


CONCLUSIONS 

The evaluation of flood hazard in a urbanized floodplain is a challenge due to both the 

complexity of the processes involved and a shortage of detailed data often experienced in 

applications. Firstly, the determination of hydrological forcing relies on the availability and on 

the completeness of historical series of data; secondly, the evaluation of the resistance of earth 

embankments to high flow conditions is difficult when only few data are available to 

characterize geomechanical parameters; thirdly, the propagation of flooding waves due to 

overtopping and breach formation requires highly detailed hydraulic models. Finally, the 

identification of a set of scenarios able to describe a good amount of possible flood configurations 

is unfortunately only a simplification of the reality. The semi-probabilistic approach 

described in this work attempts to tackle this challenge by taking into account uncertainties 

hidden in the aforementioned complexity. It provides a tool to compute the flood hazard in a 

floodplain by synthesizing the available data and yielding maps useful for risk management 

and urban planning. 

Results (water depth and velocity) relative to each return period of the flood, have been 

computed accounting for the joined occurrence of the flood and the levee failures. In contrast 

to a purely deterministic approach, which links effects to their own cause, these results 

provide a more reliable information on the actual hazard of the floodplain due to causes 

affected by non negligible uncertainties. The obtained flood intensity maps allow for the 

investigation of several levee failure scenarios in a robust manner; this provides a useful tool 

for flood hazard mapping and allows for appropriate risk mitigation strategies to be undertaken. 

It can also support decision makers and stakeholders to carry out cost benefit analysis. 

REFERENCES 

- Apel, H., Merz, B., Thieken, A.H., Influence of dike breaches on flood frequency estimation. 

Computers & Geosciences, 35, 5, 907-923, 2009. 

- Aschbacher, M., Hecher, P., Mazzorana, B. ,Development plan for rivers in South Tyrol. 6th 

Errc Conference, Vienna, 2014. 

- Autonome Provinz Bozen-Südtirol IHR—informationssystem zu hydrogeologischen Risiken. 

Methodischer Endbericht, Bozen, 2008 

- Barredo, J.I., Normalised flood losses in Europe: 1970–2006. Nat. Hazards Earth Syst. Sci., 9, 

97–104, 2009. 

- Domeneghetti, A., Vorogushyn, S., Castellarin, A., Merz, B., and Brath, A.: Probabilistic 

flood hazard mapping: effects of uncertain boundary conditions, Hydrol. Earth Syst. Sci., 17, 

3127-3140, doi:10.5194/hess-17-3127-2013, 2013. 

- Faeh R.; Mueller R.; Rousselot P.; Vetsch D.; Volz C.; Vonwiller L.; Veprek R.; Farshi D., 

BASEMENT- Basic Simulation Environment for Computation of Environmental Flow and 

Natural Hazard Simulation. Program manuals v 2.3. ETH, Zurich, 2012. 

- Fujita, Y., and T. Tamura, Enlargement of breaches in flood levees on alluvial plains. Journal 

of Natural Disaster Science, 9(1), pp. 37-60, 1987. 


- Flor, A., Pinter, N., Remo, J W.F, Evaluating levee failure susceptibility on the Mississippi 

River using logistic regression analysis, Engineering Geology, 116, (1–2), 139-148, 2010. 

- Foster, M.A., Fell, R., and Spannangle, M., The statistics of embankment dam failures and 

accidents. Canadian Geotechnical Journal, 37(5), pp. 1000-1024, doi: 10.1139/cgj-37-5-1000, 

2000. 

- Govi M., and Turitto O., Casistica storica sui processi d’interazione delle correnti di piena del 

Po con arginature e con elementi morfotopografici del territorio adiacente. Istituto Lombardo 

Accademia di Scienze e Lettere, 2000. 

- Masoero, A., Claps P., Asselman N. E. M,Mosselman E. and Di Baldassarre G, Reconstruction 

and analysis of the Po River inundation of 1951 Hydrological Processes, Wiley Online Library 

DOI: 10.1002/hyp.9558, 2012. 

- Mazzorana, B., Hübl, J., and Fuchs, S.: Improving risk assessment by defining consistent and 

reliable system scenarios. Natural Hazards and Earth System Sciences, 9, 145–159, 2009. 

- Morris, M., M. Dyer, and P. Smith, Management of flood embankments. A good practice 

review, Defra/Environment Agency, R&D Technical Report FD2411/TR1, 2007. 

- Morris, M., Breaching processes: a state of the art review, FLOODsite Project Report 

T06-06-03, FLOODsite, www.floodsite.net, 2009. 

- Lane E. W., Security from under-seepage: Masonery Dams on earth foundation. Trans. 

A.S.C.E., 100, 1935. 

- Nagy, L., Estimating dike breach length from historical data, Periodica Polytechnica, Serial 

Civil Engineering, 90(2), pp. 125-139, 2006. 

- Nagy, L., and S. Tòth, Detailed Technical Report on the collation and analysis of dike breach 

data with regards to formation process and location factors, Tech. rep., H-EURAqua Ltd., 

Hungary, 2005 

- Sellmeijer J., Numerical computation of seepage erosion below dams (piping). InProceedings 

of Third International Conference on Scour and Erosion, pp. 596–601, 2006. 

USACE, Design and Construction of Levees. US Army Corps of Engineers. Manual No. 

1110-2-1913, 2000. 

- Vrouwenvelder A. C. W. M., Steenbergen H. M. G. M., Slijkhuis K. A. H. (a). Theoretical 

manual of PC-Ring, part A: descriptions of failure models. Tech-report Nr. 98-CON- R1430, 

TUDelft, Delft. in Dutch, 2001. 

- Vrouwenvelder A. C. W. M.; Steenbergen H. M. G. M.; Slijkhuis K. A. H. (b). Theoretical 

manual of PC-Ring, part B: descriptions of failure models. Tech-report Nr. 98-CON- R1431, 

TUDelft, Delft. in Dutch, 2001. 



Landslide, flood and snow avalanche risk assessment 

for the safety management system of the railway 

Trento - Malè - Marilleva 

Gherardo Sonzio 1 ; Fulvio Bassetti 1 ; Ezio Facchin 1 ; Ettore Salgemma 1 ; Lucia Simeoni, Ph.D. .2 

ABSTRACT 

This paper defines the Criticality to identify sites that need further assessment or mitigation 

works for the safety of the Trento-Malè - Marilleva railway (Northern Italy). The railway is a 

typical example of a mountain railway exposed to natural hazards due to landslides, flooding 

and snow avalanches. In 2011 Trentino trasporti S.p.A., that is the quango responsible for the 

overall management of the railway, recognized the need of defining a systematic approach in 

the planning of the inspection activities, site-specific studies and mitigation works of natural 

hazards, and for this purpose has developed a classification method based on the concept of 

risk. The method defines five classes of "Criticality" C (none, mild, moderate, high or very 

high), to which the whole rail has been categorized. Criticality C is the product of the hazard 

H and the "works" O (C = H x O), where the factor O is greater than one in case of obsolete 

or even of absent substructures. 

KEYWORDS 

railway; landslide; flood and snow avalanche; safety management; risk assessment 

INTRODUCTION 

The railway line Trento - Malè - Marilleva in Trentino (Northern Italy) is 65 km long, 1,0 

meter gauge line, connecting the town of Trento to Marilleva (Figure 1). The railway line has 

generally a single track and it is equipped with the safety system ATP (Automatic Train 

Protection), which automatically stops the trains in case of they exceed the speed limit, or do 

not respect a signal. The average rail traffic is of 49 trains/day, the commercial speed is 

approximately of 35,0 km/h and the maximum speed is of 90 km/h. 

The line is passing within mountainous terrain through the Rotaliana plain, the Non Valley 

and Sole Valley. It was opened to the public in 1909, completely reconstructed in its own 

place at the end of the 1950s, and extended by 10 km to the current configuration at the 

beginning of the present century. It is a typical example of a mountain railway: it has a total 

rise of 700 m (from elevation 220 m a.s.l. in Trento to 900 m a.s.l. in Marilleva), a path made 

partly along hillsides, hairpin bends, 5783 m of tunnels (of which 2670 m in one single 

tunnel), 2580 m of bridges and viaducts, curve radius of 80 m and track gradients up to 50 ‰. 

1 Trentino trasporti S.p.a. Trento, ITALY, gherardo.sonzio@ttspa.it 

2 University of L’Aquila, ITALY 



Since 2009, Trentino trasporti S.p.A. (Tt) is the quango to which the Autonomous Province of 

Trento (PAT) has delegated the responsibility for the overall management of the railway 

including the operation and maintenance of the railway and its fixed infrastructures. The 

environmental context of the railway demands Tt to carry out the monitoring of risk 

conditions that can interfere with the line, among its various functions to ensure safe 

operation of the railway. Such risk conditions are those typical of an alpine environment, i.e. 

the risk of landslide, avalanche and flooding; furthermore, Tt is responsible for the construction 

and management of the identified risk mitigation works. 

From an operational point of view, Tt uses a group of dedicated employees (about 18 people), 

which are currently divided into three teams based locally and each coordinated by the 

Technical Head. Every day, each team carries out maintenance and surveillance of the track, 

its structures and of the lineside neighbourhood. The constant presence of the workers on the 

line, supported by the drivers on trains, allows the immediate identification of potential 

hazards. In particular, just before, during or after exceptional events (such as heavy rain, 

floods, heavy snowfalls, earthquakes, etc.) the workers are required to intensify the inspec- 

Figure 1: Location of the railway line Trento-Malè-Marilleva (Northern Italy). 

tions on the line. On some sites identified by Tt as "sensitive" due to greater hazard potential 

(either natural or related to the structures), the workers are asked to intervene before the 

passage of the first morning train (no night trains are operated on the line). If necessary, they 

have to adopt appropriate measures, such as the enforcement of slowdowns or interruption 

of train rides and the subsequent implementation of works aimed at risk mitigation, always in 

order to ensure the safety of the railway. In relation to the causes and severity of events that 

have originated the hazard, the workers are supported by the service personnel of Tt and 

consulting technical experts. 

In 2011 Tt recognized the need of defining a systematic approach in the planning of the 

inspection activities, site-specific studies and mitigation works of natural hazards, and for this 

purpose in the biennium 2011-2012, with its internal Technical Service resources, it has 

developed a classification method based on the concept of risk. The method defines five 

classes of "Criticality" C (none, mild, moderate, high or very high), to which the whole rail 

has been categorized. At that time, the Alpine Space project PARAmount (2015) dealing with 

risk management strategies for infrastructures was being carried out, and Tt had a few 

meetings with some partners, in particular with the Austrian Federal Railways, to exchange 

experiences in the management of natural hazards. This exchange would be more effective if 

Tt had been involved as an Observer and could participate to more meetings. 

Even though quantitative risk assessment (QRA) is increasingly encouraged in the geotechnical 

engineering community (see for example Lacasse S., 2013; Maciotta et al., 2015), 

qualitative procedures providing risk levels are still adopted with the purpose to identify sites 


ula 1 

that need further assessment or mitigation works, and do prioritize the activities (Bidwell 

et al, 2010; Winter et al, 2013). 

This paper describes the method, how it was implemented by reviewing the data already 

available to the company (from previous studies or maps carried out by PAT at large scales 

– accordingly the classification by Fell et al., 2008 large scales are intended from 1:5000 to 

1:10000) and how in some cases the Criticality values were updated after site-specific inspections 

and studies carried out at a detailed scale (1:500 to 1:1000). l r o oloa Formula 

1 oo alF 1r 

METHODS 

The method defines the criticality C as the product of an “hazard factor” H and a "work factor” 

O. The hazard factor H derives from the well-known formula to evaluate the total risk 

(Varnes D.J., 1984): 

Formula 1 

R=HEV 

Formula 1 

where R is the total risk, H the hazard, E the elements at risk and V the vulnerability. 

Only natural (geogenic) events were taken into account and specifically they were landslides 

(including: rock falls, slides, Deep Seated Gravitational Slope Deformations-DSGSDs), floods 

(distinguishing in: debris flows in the stream channels on the slopes of the valley, quick or 

slow flooding at the bottom) and snow avalanches. Moreover, an unique level of risk 

exposure equal to one along the whole line was assumed, in the sense that for example no 

distinction was made for the presence of one or two tracks (generally at the stations), rail 

station or stop, with the consequence that the vulnerability, referred to as the degree of loss 

ula 2 caused by the occurrence of an event of given intensity, was assumed l equal r to one o independently 

on the characteristics of the exposed element and of the intensity 2 oo I of the event. alF 

oloa Formula 2 

2r 

These two assumptions on the elements at risk and on the vulnerability actually prevent the 

estimation of risk and accordingly the definition by Fell et al. (2008) would reduce this 

method to an evaluation of the hazard. The hazard H was assumed as the product of the 

frequency F of occurrence of the natural event and the intensity I: 

Formula 2: 

H=FI 

formula 2 

Frequency F and intensity I were classified independently of the presence of mitigation works 

(e.g. rockfall protection fence, slides stabilized with retaining walls and drainage, etc) that 

may reduce the intensity or even avoid the occurrence of the event. In fact, since the 

planning and scheduling of construction and maintenance of mitigation works is one of the 

specific tasks that Tt has been asked to pursue for the safety of the railway line, it was decided 

not to reduce the hazard H due to the presence of mitigation works, but to amplify it with a 


Formula 3 

l r o oloa Form 

oo alF 

work factor O every time the structures were evaluated not to be either effective or functional 

to mitigate the risk. The product C: 

Formula 3: 

C=PO 

formula 3 

was defined Criticality. The purpose of the method was to obtain a ranking of criticality in 

order to appropriately manage, mitigate or prevent the consequences of an hazardous event. 

Scores were then given to frequency F, intensity I and works O, with the general rule that the 

larger the score the more critical the event to be prevented or mitigated with proper activities 

and works. More specifically, the absolute values of the scores, especially for the work factor 

O, were assigned by “trial and error” in order to be able to identify classes of criticality that 

were coherent with the planning based on the engineering judgment and successfully 

experienced in Tt for past events. 

Table 1: Intensity The frequency I scores. F scores were 0, 1 or 2. Based on the occurrence evaluated lic ere by analyzing to onloa the Table In 

databases published by national or local geological services and by alT reviewing the 1.oc data stored 

in the archive of the company, score 0 means that the event is impossible (no slope, cliff or 

stream are present) or not likely to occur, score 1 was for an event that had occurred rarely 

(once) or supposed possible, score 2 for an event that had already occurred more often than 

once. 

Table 1 

Table 1: Intensity I scores. 

Landslides 

Floods 

A 

B 

Rock Fall 

Slide 

Event 

Intensity 

A1 Boulders and large boulders 2 

A2 Cobbles 1 

B1 First-time failure 2 

B2 Active 1.5 

C Deep Seated Gravitational Slope Deformation - DSGSD 2 

D Debris flow in stream channels 2 

E Quick flooding 1.5 

F Slow flooding 1 

Snow Avalanche G Snow Avalanche 2 

Intensity I scores varied between 1 and 2 (Table 1). Score 2 was given to rapid and extremely 

rapid (Cruden et al., 1996) landslides, such as rock falls and first-time slides, and to DSGSDs. 

Even though these landslides may be inactive, quiescent or moving very slowly, they were 

scored 2 because of their large volumes and of a generally poor knowledge of the failure 

mechanisms that were acting. The score 2 to DSGSDs could be then reduced after specific 

studies. Snow avalanches and debris flow in stream channels were also scored by 2, given 


their rapid evolution and the tremendous consequences that would result from an impact 

with a passing train or simply with the tracks. Reactivated slides and quick flooding were 

given score 1.5, while the smallest score 1 was for slow flooding and falls of small rocks that 

can cause only minor damages to the trains without altering their normal cruise speed. 

e 2: Criticality The hazard C factor scores. H was then amplified by the work factor O, which Clic was ere equal to to oloa 3 in the Table terrae 

case of obsolete structures (i.e. not effective and/or not-functional alT for the purpose 2.oc of risk 

mitigation) and even to 4.5 if structures were absent. 

The resulting criticality C was then classified in 5 classes (Table 2) from nil (C0, C=0), which 

corresponds to the absence of dangerous phenomena, to very high (C4, C>18), which 

corresponds to very intense phenomena without mitigation works. 

Table 2 

Table 2: Criticality C scores. 

Criticality C0-Nil C1-Mild C2-Moderate C3-High C4-Very high 

Score 0 1-4 4-8 8-18 >18 

The high weight given to the work factor O in determining the criticality C level invites Tt 

to assume the role of controller of the safety of the railway line. Namely, Tt cannot cancel the 

natural hazard, but can mitigate its consequences with works, monitoring, inspections or 

maintenance activities, depending on the budgets. The ranking of C makes Tt constantly 

aware of the hazard conditions present along the line, as well as of the presence, the effectiveness 

and functionality of the remedial works. According to the C classification, Tt can 

manage and mitigate the consequences as budgets permit. 

The mild criticality C1 includes C scores ranging between 1 and 4, and then includes also 

the events of high intensity (I=2) and high frequency (F=2) with effective and functional 

mitigation (O=1). On the other hand, as soon as the aging of the works reduces their 

effectiveness (obsolete works such as old wood crib walls) C increases by a factor of three. 

For each C class, Tt has defined the actions to be carried out by the Transport Service, Work 

Service and Teams of Workers. For example, when criticality C4 occurs, the Transport Service 

Director, in agreement with the Technical Head, orders a temporary rail service interruption 

until the Team Workers have adopted short-term countermeasures. The rail service will then 

be reactivated with a continual monitoring by the workers, and meanwhile the Work Service 

will involve geotechnical engineers and engineering geologist for planning and designing the 

long-term mitigation works. In the case of criticality C0, the rail service is regular, no actions 

are required by the Work Service and Team Workers carry out the regular inspection of the 

line with a 15-day interval. 

RESULTS 

To date, along the entire railway line, 34 critical situations have been recognized, from mild 

to high, for a total length of about 7 km (i.e. more than 10% of the track). Figure 2 classifies 

the 34 critical situations in terms of type and criticality C score. Thirteen of the 34 critical situations 

are due to first-time slides, and ten of them are classified highly critical (C=C3, 


etween 8 and 18). The other ten C3 critical situations includes rock falls (3 C3 out of a total 

of 4 critical situations due to rock fall), one DSGSD, debris flows (5 out of 6) and snow 

avalanches (1 out of 2). 

Figure 2: Classification of the Criticalities. A1: large dimension-rock falls; A2: small dimension-rock falls, B1: first-time slides, B2: 

active slides, C: DSGSDs, D: Debris flows, E: Quick flooding, F: Slow flooding, G: snow avalanches. 

Table 3: Work It is worth factor noting values in for Table the 3 that C3 criticalities. 

the absence of mitigating works Click (O=4.5) here is the to major oloa cause Table 

alT 3.oc 

of the high criticality, and it should be borne in mind that in presence of DSGSD, due to its 

general large volume of soil, it is usually impossible to plan stabilization works, but monitoring 

or minor and local mitigating works, such as to align, restore or replace tracks, are the 

more effective actions. 

Table 3 

Table 3: Work factor values for the C3 criticalities. 

A1 B1 C D G 

O=3 1 2 0 1 0 

O=4.5 2 8 1 4 1 

In 2014, for 9 of the 34 critical situations, Tt charged geotechnical engineers and geologists to 

carry out site-specific analyses with the double aim of studying the most critical sites and of 

validating the method. For this purpose 5 high, 1 moderate and 3 mild critical sites were 

selected. Results are shown in Table 4. 


It is clear that the site-specific studies investigated the hazards at a larger scale (1 to 500, 

1000) and then many sites could be divided in sub-sites with different hazard scores. In three 

ble 4: Method cases Validation. the moderate or high criticality classes were confirmed, li hee to in donload one case Table it reduced nteaeent due to onio et 

the presence of effective works, in two cases it reduced alT only 4.do partially. In four cases the C 

classes increased because the works were actually recognized to be obsolete or absent, but 

fortunately for only a part of the investigated area. 

Table Table 4: Method 4 Validation. C class: “” increased, or “=” unchanged after the site-specific study. 

Before site-specific study 

After site-specific study 

Case Type I F O C Type I F O C C class change 

1 A1 2 2 4.5 18 A1 2 2 1 4 < 

2 A1 2 2 3 12 B2 1.5 2 4.5 13.5 = 

A2 1 2 3 6 < 

A1 2 2 4.5 18 = 

B2 1.5 2 4.5 13.5 = 

3 A2 1 2 1 2 A2 1 1 4.5 4.5 > 

B2 1.5 2 1 3 = 

A2 1 1 4.5 4.5 > 

4 B1 2 2 3 12 B1 2 2 3 12 = 

5 B2 1.5 1 4.5 6.75 B2 1.5 1 4.5 6.75 = 

6 B1 2 1 4.5 9 B1 2 2 4.5 18 = 

E 1.5 2 4.5 13.5 = 

7 B1 2 2 3 12 A2 1 1 1 1 < 

B1 2 2 3 12 = 

8 B1 2 2 1 4 A2 1 1 1 1 = 

B1 2 1 4.5 9 > 

B2 1.5 1 1 1.5 = 

B1 2 2 3 12 > 

9 A1 2 2 1 4 A1 2 2 3 12 > 

CONCLUSIONS 

The proposed method provides Tt with a systematical tool to identify and manage the critical 

situations due to natural hazards. In particular, it gives a major role to the work factor, 

because it represents the factor which Tt may control by planning the monitoring, inspections 

and mitigation works. The validation of the method by means of site-specific studies revealed 

that the work factor must be supported by a design of the structures in order to assess their 

effectiveness and that the type of the natural event must be identified by geomorphological 

surveys carried out at a detailed scale, according to Fell et al (2008). With these improvements 

the criticality could be computed by means of deterministic analyses (for example by 

slope stability analyses) instead of using scores based on engineering judgment and experience. 

So far the results have been obtained by using a spreadsheet that calculates the C-score 


to each interval of track, but it would be desirable to implement the criticality assessment in a 

GIS in order to facilitate the visualization and the real-time update. 

REFERENCES 

- Bidwell A., Cruden D., Skirrow R. (2010). Geohazard Reviews of Highway Corridors 

Through Mountainous and Foothills Terrain, Southwestern Alberta. Proc. of the 63 Canadian 

Geotechnical Conference, Calgary, Volume 1: 27-34. 

- Cruden D.M., Varnes D.J. (1996). Landslides Types and Processes. In: Turner AK, Schuster 

RL (eds). Landslides: investigation and mitigation (Special Report). Washington, DC, USA: 

National Research Council, Transportation and Research Board Special Report 247: 36–75. 

Fell R., Corominas J., Bonnard C., Cascini L., Leroi E. and Savage W.Z. (2008). Guidelines for 

landslide susceptibility, hazard and risk zoning for land use planning. Engineering Geology, 

102: 85-98 

- Lacasse S. (2013). Protecting society from landslides – the role of the geotechnical engineer. 

Proc. of the 18th International Conference on Soil Mechanics and Geotechnical Engineering, 

Paris, September 2-6 2013, 8th Terzaghi Oration: 15-34. 

- Maciotta R., Martin C.D., Morgenstern N.R., Cruden D.M. (2015). Quantitative risk 

assessment of slope hazards along a section of railway in the Canadian Cordillera—a 

methodology considering the uncertainty in the results. Landslides 1/2015. DOI:10.1007/ 

s10346-014-0551-4 

- PARAmount (2015). http://www.paramount-project.eu (accessed 02/09/2015). 

Varnes D.J. (1984). Landslide hazard zonation: a review of principles and practice. Natural 

Hazard Series, UNESCO, Volume 3: 

- Winter M.G., Harrison M., Macgregor F., Shackman L. (2013). Landslide hazard and risk 

assessment on the Scottish road network. Proceedings of the Institution of Civil Engineers 

Geotechnical Engineering 166 (6): 522–539. 



Efficient risk assessment of Norwegian railways 

combining GIS and field studies 

Heidi Hefre, M.Sc. 1 ; Kjetil Sverdrup-Thygeson, M.Sc. 1 ; Unni Eidsvig, M.Sc. 1 ; Øyvind Armand Høydal, M.Sc. 1 

ABSTRACT 

A GIS-based methodology for regional scale assessment of hazard and risk along railway 

corridors has been developed at NGI for the Norwegian National Rail Administration. 

Field investigation of hundreds of kilometres of railway is time-consuming and thus expensive 

to conduct. An assessment of risk along railway corridors is accomplished through 

substantial use of Geographical Information System (GIS), combining detailed Digital Elevation 

Models (DEM) and relevant railway data. The mapping method use advanced GIS analyses to 

identify potential high-hazard areas to closer inspect in field. 

A relative quantification of hazard and consequence level is carried out for the complete 

stretch of railway and combined to identify low, medium and high-risk areas. The results are 

presented in a series of detailed maps, showing the decision-makers where further investigations 

and mitigation measures are most needed. 

The GIS-based methodology is a time- and cost- efficient way to conduct continuous assessment 

of railway corridors, allowing fieldwork to be focused on inspection and evaluation of 

problem areas. 

KEYWORDS 

GIS; risk assessment; hazard mapping; railway; Norway 

INTRODUCTION 

Regional scale risk assessments along railway corridors using conventional mapping methods 

is time-consuming and thus expensive to conduct. Inspection of tens to hundreds of kilometres, 

obstructed by train traffic and often steep and densely vegetated terrain, make hazard 

assessment in the field challenging. 

This paper presents a new time- and cost-effective methodology for landslide hazard and risk 

assessment along railway corridors at a regional scale. The methodology combines advanced 

analyses using Geographical Information System (GIS) with field studies. The hazards 

included in this study are soil slides, embankment failure caused by insufficient drainage 

capacity and slope failures caused by river erosion. 

The presented methodology is developed by NGI through risk mapping of a total of 935 km 

of railway (nearly 23 % of all railway lines in Norway) for the Norwegian National Rail 

Administration. The purpose of the methodology is to provide a screening tool at the regional 

scale for problems involving linear infrastructures, to identify potential high-risk areas for 

detailed field investigations and future mitigation measures. 

1 Norwegian Geotechnical Institute, Oslo, NORWAY, kst@ngi.no 



METHODS 

The methodology applies GIS for estimating the natural hazard, consequence and risk level 

along the railway line, with the following aims: 

– Initial assessments to identify potential high-risk areas for soil slides, embankment failures 

and river erosion to be closer inspected in field 

– Classification of low and medium risk areas not to be inspected in field 

– Calculation of severity of consequence in case of a derailment along the railway corridor 

– Combining hazard assessment (as calculated using GIS or evaluated in field) and calculated 

consequence for a continuous risk assessment along the railway line 

– Present the results by producing risk maps for every kilometre of the mapped railway line 

The method combines a number of relevant information in order to perform an integrated 

analysis in GIS. The input data to the analyses are listed below. The availability and quality of 

these data will affect the results. 

– Digital elevation model (DEM) 

– Metered railway line 

– National landslide inventory 

– Culvert database with information on the drainage system 

– Quaternary geology maps 

– Rivers, streams and lakes 

– Existing river embankments 

– Train speed limits 

– Roads, train stations, crossings and bridges 

A high resolution DEM is of key importance, as the main purpose of the GIS-analyses is to 

analyse the terrain along the railway. The DEM is used for calculating distances, slope angles, 

slope directions, embankment- and slope-heights, and drainage paths for water and watershed 

areas. LiDAR data are preferred, but also elevation contours with 1 m interval are 

acceptable for producing a high resolution DEM of 1 or 2 m grid size. The railway line is 

divided into segments of 10 m unit lengths for analysis purposes. 

HAZARD ANALYSIS USING GIS 

The hazard analyses are performed to produce relative classifications of hazard level to 

identify potential high-hazard areas. Different parameters believed to be critical for each of 

the hazard types are analysed through GIS. In the analyses each railway segment receives a 

hazard score between 0 and 1 for each hazard type, where 0 represents no hazard and 1 

represents a high hazard level. The analyses for each hazard type are presented below. 


SOIL SLIDE AGAINST RAILWAY 

Soil slides in Norway are commonly triggered in slopes varying between 25° and 40°, 

depending on soil type. Exposed slopes are analysed in the soil slide hazard assessment using 

the following parameters: 

– The average slope angle within the exposed slope, H_soil1 

– Slope fall direction relative to railway, H_soil2 

– Soil type, H_soil3 

– Area of exposed slope, H_soil4 

GIS is used to identify all slopes with a slope angle between 25° and 40° and fall direction 

towards the railway, within 30 m to each side of the railway centre line (Figure 1). The 

analysis verifies that the slopes are soil slopes, by checking against quaternary geology maps, 

and ensures that the slope is of a significant extent. 

Figure 1: Soil slide hazard analysis using GIS: The terrain 30 m to each side of the railway is analysed to calculate average slope angle, 

fall direction and slope area. 


mula 1 

l r o o 

For each parameter a score between 0 and 1 is given according to predefined classes, and they 

together provide the soil slide hazard score according to Formula 1 

SSSSSSSS ssSSSSssss haaaaaaaass ssssSSaass = HH_ssSSSSSS 1 xx HH_ssSSSSSS 2 xx HH_ssSSSSSS 3 xx HH_ssSSSSSS 4 

Formula 1: Calculation of final hazard score for soil slide towards railway 

The soil slide hazard score is calculated for both sides of the railway, and the railway segment 

receives the highest of the two scores. 

INSUFFICIENT DRAINAGE CAPACITY CAUSING EMBANKMENT FAILURES 

Embankment failures along Norwegian railways are most often related to poor drainage and 

insufficient culvert capacity. An assessment of culverts along the railway is therefore 

conducted, by analysing the following parameters using GIS: 

– Expected discharge HQ 50 

, H_drain1 

– Culvert capacity, H_drain2 

– Upstream slope angle, H_drain3 

The expected discharge (H_drain1) is analysed by generating a hydrological correct DEM, and 

estimating the catchment area for all rivers and streams crossing the railway using the 

GIS-tool flow accumulation. The potential volume of discharge is then found by using a 

regression formula combining 50-years discharge and the associated catchment area. 

Each culvert is given the greatest calculated discharge within a radius of 30 m of the culvert. 

The expected discharge found for each culvert is then related to the culvert capacity (H_ 

drain2), which is calculated in m 3 /s from the given culvert dimensions based on capacity 

charts assuming inlet control. Possible blockage of the culvert is not accounted for. 

rmula 2 

The upstream slope angle (H_drain3) is used to express the sediment transport capacity. 

Sediment transport towards the culvert is associated with a higher probability of insufficient 

capacity. The upstream slope angle is found for every culvert by calculating the average Click slope here to do 

of the upstream waterway within the 10x30 m buffer. Slopes greater than 12° receives the 

highest score of 1. 

The drainage capacity hazard score for all culverts are calculated as shown in Formula 2: 

DDDDDDDDDDDDDDDD ccDDccDDccDDcccc hDDaaDDDDaa ssccssDDDD = HH_aaDDDDDDDD 1 

HH_aaDDDDDDDD 2 

xx HH_aaDDDDDDDD 3 

Formula 2: Calculation of final hazard score for embankment failures caused by insufficient drainage capacityy 

The relationship between the given culvert capacity and the estimated potential discharge will 

express whether the culvert capacity is sufficient. Values higher than 1 is set equal to 1, and 

implies that the culvert capacity is insufficient. 


SLOPE FAILURES CAUSED BY RIVER EROSION 

The critical parameters with respect to river erosion and associated slope failures included in 

the GIS-analyses are: 

– Distance between toe of railway embankment and river, H_erosion1 

– Height difference between toe of railway embankment and river, H_eroison2 

The first step is to locate the toe of the railway embankment and find the elevation using 

terrain analyses in GIS. This information is used for finding the distance and height difference 

between the toe of the embankment and the river. A distance of 30 m or more, or a height 

difference of 4 m or more, is regarded as a neglectable hazard level. Whereas a distance of 15 

m or less, and a height difference of 2 m or less, result in the highest hazard level. The values 

3 in between will result in an intermediate hazard value. 

l r o oloa 

The parameters are multiplied as shown in Formula 3, giving every railway segment a hazard 

score for slope failures caused by river erosion. The hazard score is compared to a national 

dataset with existing riverbank-stabilisation measures 

EEEEEEEEEEEEEE haaaaaaEEaa EEssEEEEss = HH_ssEEEEEEEEEEEE 1 xx HH_ssEEEEEEEEEEEE 2 

Formula 3: Calculation of final hazard score for slope failures caused by river erosion 

CONSEQUENCE ANALYSIS USING GIS 

For a continuous assessment for severity of consequence in the case of a derailment, a fully 

GIS-based methodology is applied. The consequence methodology is based on a methodology 

developed by Sweco (Sweco, 2012). This method was developed for risk assessments of rock 

cuts along the railway, based on traditional registration in the field. It was thus not well 

suited for regional scale risk mapping of several geohazards. 

After experiences gained in the first few years with these projects, NGI modified the existing 

consequence methodology to be fully based on GIS-analyses and with new classifications of 

the parameters. Three parameters are included in the consequence analyses and explained in 

more detail in the next paragraphs. 

ACCESSIBILITY FOR RESCUE 

The accessibility for rescue (C_access) is calculated by first finding the distance to the nearest 

railway station, level crossing and platform for every railway segment. Secondly, the 

accessibility from the nearest road is calculated using a "cost-analysis" in GIS. In this analysis, 

distance, slope and barriers, such as rivers and cliffs, are taken into account. The results of the 

cost-analysis are evaluated in order to choose suitable limits for defining accessibility into the 

three categories easy, intermediate and difficult (figure 2). 


Figure 2: Example of cost-analysis in GIS, used to define accessibility from road into “easy”, “intermediate” and difficult”. The green 

colour represents “easy access” to the railway, and surrounds the road in different width, depending on terrain slope and natural 

barriers. 

At stations, level crossings and platforms, it is assumed that rescue vehicles can access the 

railway track fast and easily. A short distance to these locations is thus regarded as a better 

alternative than reaching the track from a nearby road. The calculated distance is combined 

with the accessibility from road to express the Accessibility for rescue. The parameter is 

divided into seven classes with scores ranging from 1 to 8. 

TERRAIN AT PLACE OF DERAILMENT 

The shape of the terrain (C_terrain) within 30 m to each side of the railway is calculated 

using GIS through analyses of slope height, slope fall direction and distance to body of water. 

The higher the slope, the higher consequence score is achieved. If the slope ends in a body of 

water, the score is increased. If there is no slope (the railway goes through a two-sided 

cutting), the lowest score is achieved. The parameter is classified into six different classes with 

the associated scores ranging between 1 and 12. 

IMPACT SPEED 

Potential impact speed (C_speed) is calculated by using the sight distance together with the 

train speed limit. Sight distance is found using GIS based on the curvature of the railway line 

assuming a limited field of view, simulating dense vegetation 5m from the railway line. The 


train velocity is assumed to be constant during the reaction time and the duration of break 

application, and to decrease linearly during the retardation time. This means that in practice 

in curves the impact speed equals the speed limit. 

The speed limit and the sight distance will differ according to which direction the train is 

travelling. The potential impact speed is thus calculated in both directions, and the higher of 

the two results is chosen. The classification of impact speed is set to five different classes, with 

an impact speed less than 10 km/h resulting in the lowest score of 0, and an impact speed 

l r o oloa Formula ormula4o 

more than 80 km/h achieve the highest score of 10. 

The achieved scores for each of the three consequence parameters are finally summarised 

according to Formula 4, giving every railway segment a consequence score. 

CCCCCCCCCCCCCCCCCCCCCC CCCCCCssCC = CC aaaaaaaaaaaa + CC ttaattttaatttt + CC aassaaaass 

Formula 4: Calculation of final consequence score 

The consequence score is normalized to values between 0 and 1 for purposes of the risk 

analysis. 

FIELD INVESTIGATIONS 

Field investigations were carried out for potential problem areas, identified in the initial stages 

of the project. The railway sections to include for field investigations are identified from the 

following sources: 

– All sections identified with high-risk potential from GIS-analyses 

– Interviews with railway supervisors 

– One day overview inspection of the entire mapping area using designated slow-moving 

train 

– Information about historical slides on or near the railway 

– Documents describing problem areas 

The focus of the field investigations is to evaluate the condition and characteristics of steep 

side terrain, culverts, railway embankments and river embankments identified as potential 

hazards. 

All segments achieving the highest hazard score of 1 and segments classified with high-risk 

from GIS-analyses are subjected to a closer inspection in field. The results of the hazard 

analyses are updated after field investigations, with the hazard scores evaluated in field 

replacing those calculated by GIS-analyses. In potential risk areas, the GIS-analyses are thus 

always complemented and adjusted by expert opinion. 


RESULTING RISK ASSESSMENT 

Risk scores are calculated for all railway segments by multiplying the hazard and consequence 

score, and categorised into "low", "medium" and "high" risk level, according to Figure 3. This 

method provides a relative value of risk level for the analysed stretch of railway, allowing for 

comparison between the railway segments. 

Figure 3: The final hazard and consequence score for every railway segment are plotted together and categorised into “low”, “medium” 

and “high” risk level 

By counting the number of high-risk segments per kilometre railway, one can clearly 

illustrate where along the railway mitigation measures should be focused (Figure 4). 

Figure 4: The distribution of high-risk areas per kilometre of railway clearly shows where to prioritise detailed investigations and 

mitigation measures 

All final results are presented in a series of risk maps, covering the complete stretch of 

railway, each map displaying 1 km of the railway (figure 5). High-risk areas are highlighted 

in red, showing the decision makers where further investigations are most needed. This is a 

clear way to communicate information to the end user. 


Figure 5: Example of final risk map produced for every 1 km of the analysed railway. The hazard, consequence and risk level along the 

railway are shown continuously, in addition to risk level of culverts and location of historical slides. 


CONCLUSIONS 

Advanced analyses through GIS make it possible to conduct a continuous assessment of 

railway corridors, while fieldwork is focused on closer inspection and evaluation of problem 

areas. The selection process of these areas is supplemented by information from railway 

supervisors and other relevant sources. Regional scale mapping of landslide hazard and risk 

along railway corridors using a GIS based methodology have proven to be a time- and 

cost-efficient method. The quality of the GIS-analyses is highly dependent on a detailed 

terrain model of the proximity of the railway. The developed GIS-analyses provide a good 

indication of the location of potential problem areas, both with regard to the investigated 

hazards and the consequences of a derailment. It should however be underlined that the GIS 

analyses serve as a supplement to and not a substitution for field studies. 

FURTHER WORK 

There are parts of the GIS analysis that could be improved. First and foremost, we see a need 

to improve the calculation of the expected discharge in connection to culverts by a refined 

method for estimating the catchment area and extraction of associated parameters. This 

improvement is planned to be implemented in 2016. Furthermore, the method for estimating 

collision speed could be refined by taking into account the gradient of the railway line. The 

accessibility for rescue could take into consideration other factors, especially the type of 

ground (forest, farmland, bog, etc.). 

It is our opinion that the combination of methods presented in this paper identifies possible 

high-risk areas in a satisfactory manner, with the exception of the hazard relating to 

insufficient culvert capacity. This hazard type is particularly challenging, as it cannot be 

properly evaluated using detailed terrain models, overview mapping from inspection train or 

other available data sources. Performing capacity can only be revealed by a detailed inspection 

of culvert inlet and outlet. In order to verify the methods regarding culverts, some lowand 

medium-risk culverts could be randomly selected and examined in field, in order to 

detect possible underestimations of risk. 

REFERENCES 

Sweco (2012). Jernbaneverket, Veileder til verktøy for skredfarekartlegging. Report no. 

575012/01. 



Periglazial Hazard Indication Map: A Basic Instrument 

in Prospective Hazard Management 

Gefahrenhinweiskarte Periglazial: Instrument in der 

prospektiven Gefahrenbeurteilung 

Daniel Tobler, dipl. natw. ETH, Geologe 1 ; Peter Mani, lic. phil. nat., Geograph 2 ; Rachel Riner, lic. phil. nat, Geologin 1 ; Serena 

Liener, dr. phil. nat. 2 , Geographin; Nils Hählen, dipl. natw. ETH, Forstingenieur 3 ; Ricarda Bender-Gàl, Wasserbauingenieurin 4 

ABSTRACT 

Based on glacier retreat and degrading permafrost in high alpine regions across the Bernese 

Oberland (central to western Switzerland), several new source areas for natural hazard 

processes became evident within the last years. The evaluation of susceptible periglacial areas 

and the assignment of resulting processes by modelling will be one of the major tasks to be 

solved in near future. The prediction of those processes and their consequences is an 

interdisciplinary question. Meteorological scenarios for the next 50 years derived from 

climate change scenarios stand at the beginning of the decision chain. The susceptible 

periglacial areas which act as starting zones for mass movements (rockfall, landslides, debris 

flows) or new sediment sources can be calculated through sophisticated permafrost and 

glacier retreat models. 

Well established simulation tools have been used within the project. The result is a so-called 

periglacial hazard indication map visualizing endangered areas in the year 2060 for mass 

movement processes as well as other natural hazards like floods or subglacial lake outburst. 


Anhand von Klimaszenarien wurde die zukünftige Entwicklung der Einzugsgebiete sowie das 

Zusammenspiel der Veränderung der Gletscherausdehnung und des Permafrostes mit den 

Sturz- und Rutschprozessen, der Murgangauslösung, dem Geschiebetrieb sowie möglichen 

Folgeprozessen (z.B. Flutwelle) analysiert und modelliert. Die Resultate basieren auf einer 

neu entwickelten Methodik. Generell sind die einzelnen Prozessmodellierungen umfassend 

auf aktuellen, wissenschaftlichen Arbeiten abgestützt und mittels einer Sensitivitätsanalyse 

(vgl. separater Bericht) geprüft. Die Ergebnisse präsentieren die mögliche Gefahrensituation 

im Berner Oberland im Jahre 2060 unter Berücksichtigung der klimatischen Veränderungen. 

Die Modellresultate der einzelnen Prozesse verdeutlichen die mögliche Problematik im 

Zusammenhang mit degradierendem Permafrost, sich zurückziehenden Gletschern und damit 

1 GEOTEST AG, Zollikofen, SWITZERLAND, daniel.tobler@geotest.ch 

2 Geo7 AG 3012 Bern, SWITZERLAND 

3 Amt für Wald des Kantons Bern, Abteilung Naturgefahren, Interlaken, SWITZERLAND 

4 Tiefbauamt des Kantons Bern, Oberingenieurkreis 2, SWITZERLAND 



einhergehenden Destabilisierungen der Talflanken. Die am stärksten betroffenen Regionen 

sind Kandersteg, das Haslital sowie die hinteren Lütschinentäler. 

KEYWORDS 

process chains, permafrost, mass movements, climate change, process modelling 

EINFÜHRUNG 

In den hochalpinen Gebieten des Berner Oberlandes treten seit Jahren immer wieder neue 

Naturgefahrenherde auf, die durch den Rückgang von Gletschern und Permafrost entstehen. 

Sie bedrohen teilweise die Sicherheit von Siedlungen und Infrastrukturen. Die kantonalen 

Fachstellen des Kantons Bern haben aus diesem Grund die Studie Gefahrenhinweiskarte 

periglazial (kurz GHKperiGlazial) initiiert. Die Studie soll aufzeigen, wo sich infolge degradierenden 

Permafrosts und Gletscherrückzugs eine Veränderung der Naturgefahrenprozesse 

ergibt. 

BEARBEITUNGSGEBIET 

Der Untersuchungsperimeter umfasst sämtliche hydrologischen Einzugsgebiete im Berner 

Oberland, welche im Einflussbereich periglazialer Prozesse liegen (Figur 1). Der Perimeter 

reicht vom Simmental im Westen bis zum Gadmen- und Gental im Osten. Südlich verläuft 

der Perimeter entlang der hohen Alpengipfel an der Grenze zum Wallis; nördlich bilden 

Thuner- und Brienzersee sowie das Gental die Begrenzung. 

Figur 1: Der Untersuchungsperimeter Berner Oberland wurde für die Bearbeitung in vier Teilgebiete unterteilt. TG1 (Hasliaare), TG2 

(Lütschine), TG3 (Kander), TG4 (Simme) (Kartengrundlage: Swisstopo). 


VORGEHEN 

In einem ersten Schritt wurden die Klimaszenarien sowie Kenngrössen künftiger prozessauslösender 

Wetterereignisse, die bis 2060 im Berner Oberland zu erwarten sind, bestimmt. 

Basierend auf den prognostizierten Temperaturveränderungen wurde die künftige Ausdehnung 

von Gletscher- und Permafrostgebieten hergeleitet. 

Die Klimaveränderungen haben Auswirkung auf die Häufigkeit und die Magnitude der 

Naturgefahrenprozesse. Für die Erstellung einer Hinweiskarte ist nur die Veränderung der 

Magnitude entscheidend. Basierend auf diesen Rahmenbedingungen wurden die Wirkungsräume 

sowie die potenziellen Ablagerungskubaturen von Sturz-, Rutsch- und Murgangprozessen 

modelliert. Weiter wurden glaziale Prozesse wie Gletscherhochwasser, Eisabbrüche 

sowie Sekundärprozesse in Form von Impulswellen und Dammbrüchen analysiert (Figur 2). 

Die Bearbeitungstiefe der Prozessmodellierungen erfolgte auf Stufe „Gefahrenhinweis“, 

Massstab 1:25‘000. D.h. von den modellierten Prozessräumen wurden deren Umhüllende 

ausgeschieden. 

Figur 2: In der GHKperiGlazial modellierte Prozessketten. 


KLIMASZENARIEN 

Wichtigste Grundlage für die Herleitung der Klimaszenarien sind die Szenariendaten CH2011 

(2011), wobei die Daten des Emissionsszenarios A1B für den Szenarien-Zeitpunkt 2060 

verwendet wurden. Zwar entspricht die aktuelle Emissionsentwicklung eher dem pessimistischen 

Szenario A2. Da für das Szenario A1B detailliertere Daten vorliegen und sich die 

Auswirkungen der beiden Szenarien bis zum Zeitpunkt 2060 noch nicht wesentlich unterscheiden, 

wurde dieses verwendet. Für den Szenarienzeitpunkt 2085 zeigt das Szenario A2 

dann jedoch deutlich stärkere Auswirkungen als das Szenario A1B. 

Basierend auf diesen Daten wurden Szenarien für die Temperatur und den Niederschlag 

definiert. Bei den Temperaturen wurden neben den saisonalen Mitteltemperaturen auch 

Hitzeperioden und die Anzahl von Tautagen ausgewertet. Für den Niederschlag wurden 

neben den saisonalen Niederschlagssummen die Veränderungen der Intensivniederschläge 

betrachtet (Rajczak et al. ,2013). 

GLETSCHER UND PERMAFROST 

Das Szenario für den Gletscherrückzug wurde entsprechend dem Temperaturszenario aus den 

Daten von Linsbauer et al. (2013) abgeleitet. Zusätzlich zur zukünftigen Gletscherausdehnung 

wurden auch die Topographie des freigelegten Gletscherbetts und dessen Charakteristik 

hergeleitet. 

Grundlage für die Beurteilung der Veränderungen im Permafrost bildet die Karte der 

potenziellen Permafrostverbreitung (APIM) (Boeckli et al., 2012). Der darin abgebildete Index 

kann vereinfacht als anteilmässige Verbreitung von Permafrost interpretiert werden. Der 

Index liefert jedoch keine Angaben zur Mächtigkeit und Temperatur des Permafrosts. Deshalb 

wurde für die Abschätzung der Permafrostdegradation und deren Auswirkungen ein 

vereinfachter Ansatz entwickelt. 

Ausgehend von der heutigen mittleren jährlichen Lufttemperatur wurde die heutige 

Oberflächentemperatur abschätzt. Die Temperaturoffsets konnten aus der Oberflächencharakteristik, 

der Hangneigung und der Exposition abgeleitet wurden. In einem weiteren Schritt 

wurden für das Temperaturszenario die Offsets berechnet, wobei wiederum eine materialabhängige 

Sensitivität berücksichtigt wurde. Aus diesem Offset und dem Permafrostindex 

wurde ein Veränderungsfaktor abgeleitet, der in den Prozesssimulationen als Parameter für 

die Materialverfügbarkeit verwendet wird. 

MASSENBEWEGUNGSPROZESSE 

Die Simulation der betrachteten Massenbewegungsprozesse erfolgte für das Jahr 2060 in zwei 

verschiedenen Wahrscheinlichkeiten (gross, klein). Dabei wurden für sämtliche Prozessräume 

die möglichen Ablagerungskubaturen (Rutsch- und Sturzprozesse) respektive die maximalen 

Abflüsse (Flutwellen) und Geschiebefrachten (Murgänge) in Grössenordnungen von 10er 

Potenzen abgeschätzt. 


STURZPROZESSE 

Die Prozessräume potentieller Blockschläge und kleinerer Felsstürze wurden ausgehend von 

der Felsmaske (Swiss-TLM) sowie den ausgeaperten Felsbereichen in Gletscherrückzugsgebieten 

mit dem 3D-Sturzprogramm GEOTEST+Zinggeler (Tobler et al. 2009) modelliert. 

Basierend auf der Geologie und dem Durchtrennungsgrad des Gesteins (Kluftanalyse mit 

ColTop 3D (Bouissou 2011)) wurden die Ausbruchgebiete klassiert und Blockgrössen für die 

Modellierung festgelegt. Nach der Simulation wurden die Endpunkte der Sturztrajektorien 

mit den Informationen des Ausbruchgebietes attributiert (Permafrostverhältnisse, Klüftungsgrad 

des Gesteines, Blockgrösse), um die gesamte Ablagerungskubatur im Ablagerungsraum 

zu berechnen. Diese ergab sich aus der Summe in einer Rasterzelle abgelagerter Sturzblöcken 

resp. deren Volumina. 

RUTSCHPROZESSE 

Die Anrissgebiete spontaner Rutschungen wurden mit dem Modell SliDisp+ (Tobler et al., 

2013) modelliert, deren Auslaufbereiche mit dem GIS-Modell SliDepot (Tobler et al, 2013). 

Für die resultierenden Prozessflächen wurde pro Zelle die potenzielle Ablagerungskubatur 

kalkuliert, indem die mobilisierbare Mächtigkeit (Salciarini et al., 2006) mit der Rasterzellengrösse 

multipliziert wurde. Die mobilisierbare Mächtigkeit wird in den periglazialen Bereichen 

im Wesentlichen von der Permafrostveränderung dominiert. Für das Jahr 2060 wurden 

zusätzlich neue Anrissflächen in den Gletscherrückzugsgebieten berücksichtigt. 

Die Auswirkung der Klimaszenarien auf die permanenten Rutschungen ist nur schwer zu 

erfassen. Die Rutschungen liefern einen kontinuierlichen Geschiebeeintrag in die Gerinne. 

Diese Eigenschaft wurde über die Erhöhung der materialspezifischen Erodierbarkeit innerhalb 

der Rutschungen berücksichtigt, wobei die Parameter Rutschaktivität, Hangneigung, 

Permafrostveränderungen nach einer empirischen Klassierung in die Berechnung einflossen. 

Für die Erhöhung wurden die Parameter Rutschaktivität, Hangneigung und Permafrostveränderung 

nach einer empirischen Klassierung berücksichtigt. Die Abgrenzungen des permanenten 

Prozesses basierte auf bestehenden Grundlagen sowie InSAR-Daten. Lagen in zukünftigen 

Gletscherrückzugsgebieten deutliche Indikatoren vor, welche auf eine Aktivierung einer 

permanenten Rutschung hinweisen könnten, wurden diese ebenfalls erfasst (z.B. Hangfussentlastung 

im Bereich einer übersteilten Seitenmoräne, Übersteilte Rutschfront gekoppelt 

mit Murganggerinne, dokumentierte Ereignisse spontaner Rutschungen). 

MURGANGPROZESSE 

In einem ersten Schritt wurden die potenziellen Murganganrissgebiete ausgeschieden. Dabei 

wurden auch diejenigen Gebiete berücksichtigt, in welchen durch Gletscherrückzug oder als 

Folge von verstärkten Bergsturzablagerungen neue Murganganrisse entstehen können. 

Zur Bestimmung der potenziellen Murgangfracht wurde für jeden Punkt entlang der 

Trajektorie definiert, wieviel Material durch einen Murgang erodiert werden kann. Dies 

entspricht der Erosionsleistung eines Murganges auf den entsprechenden Laufmetern an der 


entsprechenden Stelle. Diese Erosionsleistung hängt im Wesentlichen von den geschieberelevanten 

Grundlagen sowie sämtlichen geschiebeliefernden Prozessen ab. Bei Murgängen aus 

Gletscherrückzugsgebieten oder aus auftauendem Permafrost wurde dabei von einer 

erhöhten Erosionsleistung ausgegangen. In den Permafrostdegradationsgebieten wurde dazu 

der Veränderungsfaktor verwendet. 

Ausgehend von den Anrissgebieten wurden die Murgangtrajektorien mit dem Modell 

MGSIM/dfwalk (Gamma 2000) modelliert. Entlang des Fliessweges jeder Trajektorie konnte 

die Erosion pro Gerinneabschnitt entsprechend der Erosionsleistung sowie in Abhängigkeit 

der Murganggeschwindigkeit bestimmt und fortlaufend zur Geschiebefracht aufsummiert 

werden. Beim Kegelhals wurden die Geschiebekubaturen durch das Modell RAMMS 

(Christen et al., 2012) abgenommen, um die Ausbreitung auf dem Kegel zu simulieren. 

Sehr seltene Ereignisse mit grossem Impakt 

Zu den sehr seltenen Ereignissen mit grossem Impakt zählen: Bergstürze, Gletscherabbrüche, 

Gletscherseeausbrüche. Gebiete mit einer Disposition für solche Grossereignisses wurden 

lediglich im periglazialen Raum für die kleine Wahrscheinlichkeit ausgeschieden und in der 

weiterführenden Prozesskette nicht weiter berücksichtigt. 

RESULTATE 

In den folgenden Abschnitten werden die Resultate der Modellierungen sowie die Veränderungen 

der Prozesse erläutert. Ein Darstellungsbeispiel findet sich Figur 3. 

Periglazial: Die Veränderungen im Periglazialgebiet werden im Wesentlichen durch den 

Rückgang der Gletscher sowie degradierendem resp. verschwindendem Permafrost verursacht. 

Der Anteil an Permafrostflächen und entsprechenden Veränderungen ist in den 

Teilgebieten Hasliaare und Lütschine am grössten. Westlich des Kandertals nehmen die 

Permafrostflächen und somit deren Relevanz deutlich ab. Insgesamt sind rund 12 % des 

Berner Oberlandes von Permafrostflächen bedeckt. 

Gletscherseen: Insbesondere im östlichen Berner Oberland kann durch den Rückzug von 

grossen und flachen Talgletschern eine hohe Anzahl neuer Gletscherseen entstehen, deren 

Volumina über 10 Mio. m 3 liegen können. In den zentralen und westlichen Gebieten des 

Berner Oberlandes ist mit geringerer Gletscherseebildung zu rechnen.Viele dieser potenziellen 

Gletscherseen entwickeln sich in sehr abgelegenen Lagen und tangieren Siedlungsraum 

kaum. Wenn jedoch Massenbewegungsprozesse in einer Prozesskette in einen Gletschersee 

stossen und so zu einem Seeausbruch führen, können weit entfernte Schadenpotenziale 

plötzlich bedroht werden. 

Gletscherabbrüche: Im östlichen Berner Oberland gibt es viele Gebiete mit steiler Vergletscherung, 

was beim Gletscherrückzug grundsätzlich zu einer hohen Anzahl von potentiellen 

Abbruchgebieten führt, die generell weit entfernt vom Schadenpotential liegen. Im zentralen 


Teil des Untersuchungsgebietes sind zwar weniger Gebiete gletscherbedeckt, der Siedlungsraum 

befindet sich jedoch vielerorts näher an steilen Gletschergebieten. Daher existiert in 

diesen Gebieten eine grössere Anzahl von Abbruchgebieten, deren Eisabbrüche direkt 

Siedlungsraum und wichtige Infrastrukturen treffen können. Im westlichen Berner Oberland 

gibt es aufgrund der geringen und wenig steilen Vergletscherung ein geringeres Ereignispotenzial. 

Die Entwicklung von potenziellen Eisabbruchgebieten ist stark mit Gletscherveränderungen 

verbunden. Gletscher, welche heute kein Eisabbruchpotential aufweisen, können in 

Zukunft jedoch durch Rückzug und veränderte Gletschergeometrie ein Gefährdungspotenzial 

entwickeln. 

Sturzprozesse: Die Ablagerungskubaturen aus Sturzprozessen sind im östlichen Berner 

Oberland aufgrund des grösseren Anteils von Felsgebieten höher als im Westen. Die 

angewandte Methodik, rohe Modellierung als Basis, Erhöhung der Ablagerungskubaturen 

aufgrund vermehrter Ausbrüche aus tektonischen Störzonen (Kluftsystemen) sowie 

Permafrostzonen führt insgesamt etwa zu einer Verdoppelung der modellierten Ablagerungskubaturen 

im Jahr 2060. Nebst der regionalen Variabilität zeigt sich auch im Kleinräumlichen 

ein eindeutiges Muster der Ablagerungen: Die Trajektorien aus den höherliegenden Ausbruchgebieten 

sammeln sich in den Runsen und Rinnen. Nur ein kleiner Teil der Blöcke 

reicht bis in den Talboden und damit in den Bereich des Schadenpotenzials. Die beträchtlichen 

Ablagerungen in den Rinnen bilden ein wichtiges Geschiebepotential für Murgänge. 

Rutschprozesse: Grundsätzlich kann festgehalten werden, dass alle Typen von Rutschungen 

ein massgebliches Geschiebepotential für Murgänge bilden. Spontane Rutschungen in 

Gerinnen können teilweise direkt Murgangprozesse auslösen. Die grossen permanenten 

Rutschungen in Gerinnenähe bilden bei Reaktivierungen enorme Geschiebeherde. Bei den 

permanenten Rutschungen werden stärkere resp. häufigere Reaktivierungen gegenüber der 

heutigen Situation in jenen Bereichen erwartet, wo sie bereits heute zu Reaktivierungen 

neigen und/oder im Einflussbereich des Periglazials liegen. 

Murgangprozesse: Veränderungen gegenüber der heutigen Situation sind v.a. dort zu 

erwarten, wo Murganganrisse im Permafrost oder in den Gletscherrückzugsgebieten liegen. 

Hier können einerseits neue Anrisse entstehen und andererseits deutlich grössere Geschiebefrachten 

verlagert werden. Dies wiederum führt zu grösseren Fliesshöhen (und damit 

Prozessintensitäten) sowie zu ausgedehnteren Ablagerungsflächen. Besonders betroffen sind 

das Haslital, die Lütschinentäler und das Kandertal. Im Engstligental sowie im Simmental sind 

grössere Veränderungen eher in den siedlungsfernen Gebieten zu erwarten. 


Figur 3: Detailausschnitt der modellierten Murgangprozessräume (violett) und Sturzablagerungen (blau und gelb). 167 

FAZIT 

Die GHKperiGlazial gibt für das Berner Oberland einen Überblick, wo als Folge der Klimaänderung 

am ehesten mit Veränderungen der Naturgefahren zu rechnen ist und welche 

Prozessarten davon betroffen sind. Die am stärksten betroffenen Regionen sind Kandersteg, 

das Haslital sowie die hinteren Lütschinentäler (Figur ). 

Die Studie beschreibt die massgeblichen Prozessketten und das Ineinandergreifen der 

einzelnen Teilprozesse. Am Anfang der Prozesskette liegen die Permafrostdegradation und der 

Gletscherrückzug. Je grösser die Veränderung, desto stärker die Reaktion des gesamten 

Systems. Der Norsektor umfasst rund ein Drittel der gesamten Permafrostflächen. Diese 

reichen hier in tiefere Höhenlagen. Mit dem Wissen, dass tieferliegende Gebiete von der 

Erwärmung generell stärker betroffen sind als höhere Lagen, ist es kaum erstaunlich, dass die 

Prozesse in den nordexponierten Gebieten im Jahre 2060 im Allgemeinen besonders verstärkt 

werden. 

Sehr viele Veränderungen laufen weit entfernt von den für die vorliegende Studie relevanten 

Siedlungsräumen ab und tangieren daher kaum Schadenpotenzial. Bergstrassen, Wanderwege, 

Alphütten und Seilbahnmasten stellen in diesem Zusammenhang kein relevantes 

Schadenpotential dar. Durch die Klimaänderung sind im Siedlungsgebiet keine neuen 


Figur 4: Zusammenfassende Übersicht zu den mutmasslich betroffenen Regionen im Berner Oberland. 173 

Gefahrenflächen zu erwarten. Hingegen kann die Häufigkeit oder die Intensität von einzelnen 

Ereignissen gegenüber der Vergangenheit zunehmen. 

Die GHKperiGlazial kann keine räumlich konkrete Aussage dazu machen, in welchen 

Gebieten effektiv Konflikte durch veränderte Gefahrenprozesse auftreten werden. Sie zeigt 

auf, wo grundsätzlich ein Potential dazu besteht. Auch Prozesse, die als Einzelereignis kein 

Problem darstellen, sondern erst in einer Häufung Schäden verursachen, lassen sich mit der 

gewählten Methode nicht bestimmen. 

Der Hauptnutzen der GHKperiGlazial liegt darin, die potentiell heiklen Gebiete zu identifizieren 

und zu priorisieren. Darauf basierend wird in den nächsten Jahren ein systematisches 

Monitoring aufgebaut, welches sicherstellen soll, dass ungünstige Entwicklungen frühzeitig 

erkannt werden und somit mehr Reaktionszeit für präventive Massnahmen zur Verfügung 

steht. 


LITERATUR 

- AG NAGEF Arbeitsgruppe Naturgefahren des Kantons Bern (2015): Klimawandel und 

Naturgefahren – Veränderungen im Hochgebirge des Berner Oberlandes und ihre Folgen. 

Linsbauer A., Paul F., Machguth H., Haeberli, W. (2013). Comparing three different methods 

to model scenarios of future glacier change for the entire Swiss Alps. Annals of Glaciology, 54 

(63), 241-253. 

- Rajczak J., Pall P., Schär C. (2013). Projections of extreme precipitation events in regional 

climate simulations for Europe and the Alpine Region. J. Geophys. Res.-Atmos., 118, 

3610-3626, doi: 10.1002/jgrd.50297. 

- Tobler D., Riner R., Pfeifer R. (2013). Modelling potential shallow landslides over large areas 

with SliDisp+. C. Margottini et al. (eds.), Landslide Science and Practice, Vol. 3, DOI 

10.1007/978-3-642-31310-3_6, Springer-Verlag Berlin Heidelberg 2013 

- Tobler D. Riner R., Pfeifer R. (2013). Runout modelling of shallow landslides over large 

areas with SliDepot. C. Margottini et al. (eds.), Landslide Science and Practice, Vol. 3, DOI 

10.1007/978-3-642-31310-3_32, Springer-Verlag Berlin Heidelberg 2013 

- Boeckli L., Brenning A., Gruber S., Noetzli J. (2012). Permafrost distribution in the 

European Alps: calculation and evaluation of an index map and summary statistics. The 

Cryosphere, 6(4), S. 807-820. 

- Christen, M.; Gerber, W.; Graf, Ch.; Bühler Y.; Bartelt, P.; Glover, J.; McArdell, B.; Feistl, T.; 

Steinkogler, W. 2012: Numerische Simulation von gravitativen Naturgefahren mit "RAMMS". 

Wildbach-, Lawinen-, Erosions- und Steinschlagschutz. 169, 282 - 293. 

- CH2011 (2011). Swiss Climate Change Scenarios CH2011. Published by C2SM, MeteoSwiss, 

ETH, NCCR Climate, and OcCC. Zurich, Switzerland. 

- Bouissou S., B. T. (2011). Influence of structural heterogeneities and of large scale topography 

on imbricate gravitational rock slope failures. Tectonophysics 526-529, S. 147-156. 

- Tobler, D., Graf, K., Krummenacher, B. (2009). Rockfall assessment of natural hazards by 

3D-simulation potential. Geophysical Research Abstracts, Vol. 11, EGU2009-6666-1, 2009, 

EGU General Assembly 2009 

- Gamma, P. (2000). dfwalk - Ein Murgang-Simulationsprogramm zur Gefahrenzonierung. 

Bern: Geographica Bernensia G66, Verlag des Geogr. Inst. Univ. Bern. 



The case study of Badouzih rockfall in northern Taiwan: 

mechanism, numerical simulation and hazard 

assessment 

Ching-Fang Lee, Ph.D 1 ; Ting-Chi Tsao, M.E. 1 ; Lun-Wei Wei, M.S. 1 ; Wei-Kai Huang, M.S. 1 

ABSTRACT 

The aim of this study is to reconstruct the rockfall movement trajectory and runout distance 

for understanding the falling dynamics in northern Taiwan. The Badouzih rockfall was 

triggered by rainfall on August 31st, 2013 in Keelung City in northern Taiwan. A 43 m 3 

volume rock was detached from the summit of the hillslope by toppling failure near 68K, 

Highway Route 2. The rockfall hazard caused road closure and car crush with the giant falling 

boulder. The field survey showed that the weathering process led to the extension of joint, 

during the heavy rainfall, the infiltration and the surface runoff washed the weathered 

material away, lead to the unstable and the fallen of the rock. The process can be divided into 

roll, fall and bounce, taking about 23 seconds in total. With the video record and the field 

investigation, a three dimensional numerical model (RAMMS::ROCKFALL) characterized by 

irregular block shape was adopted and verified in this study. 

KEYWORDS 

Rockfall; Badouzih; RAMMS; rock shape; trajectory 

INTRODUCTION 

Rockfall and debris slide are geologic hazard which occur in the costal mountainous area and 

usually been triggered by intense weathering and heavy rainfall. Rockfall can be categorized 

as a failure behavior of rock mass at a steep slopeland and weak joint plane (slope >55 [deg.]; 

Central Geological Survey). The moving type of rockfall is consist of fast free falling, toppling, 

rolling (with an obvious trajectory), and bouncing based on the landform, fragmentation 

condition, and strength of rock mass. The literature reveals the number of rockfall events 

between 82K-139K of Highway Route 2 (or called Northern Coastal Highway) reached 84 

during the period between 1994-1996 (total volume: 25,500 m 3 ). The spatial distribution of 

historical rockfall (1994-2003) on Highway Route 2 demonstrates it has a highly dependence 

on the seasonal climate, especially in autumn and winter. 

So far, a large number of computer numerical approaches have been developed for quantitative 

and modelling rockfall behavior based on different mechanical frameworks. The common 

numerical program for rockfall hazard analysis includes the following four methods: (1) 

Lumped mass approach (LMA); (2) Rigid body approach (RBA); (3) Discrete element method 

1 Disaster Prevention Technology Research Center, Sinotech Engineering Consultants, INC. Taipei, TAIWAN, tctsao@sinotech.org.tw 



(DEM); (4) Discontinuous deformation analysis (DDA). The approaches mentioned above 

provide the user the information of energy, trajectory, moving velocity, and jumping height 

of dangerous rockfalls, so the corresponding analysis of rockfall hazard can be performed for 

both advanced engineering practice and protective measures (Agliardi et al.,(2009); Cottaz 

et al.,(2010); Chen et al.,(2013); Leine et al., (2014); Giovanni et al.,(2015). We therefore use 

the RAMMS (RApid Mass Movement System) for reconstructing the 3D rockfall behavior and 

runout distance for understanding the characteristic of Badouzih rockfall event on August 

31st, 2013 (Figure 1, inset). In the paper, the available video record from dashboard camera 

(source: YouTube), remote sensing interpretation, geologic parameter collection, and field 

survey are summarized to reconstruct the failure mechanism of the rockfall. The approach 

allows us to build an irregular rock shape similar to the one in question instead of the typical 

spherical particle shape. The result of the 3D numerical simulation was compared with 2D 

Rocfall studies (Wei et al., (2014)). As a result a rockfall hazard map for this study site is 

produced and validated. 

Figure 1: Study area and geological setting of Badouzih, Keelung (Central Geological Survey, 1988, scale: 1/50,000). 

STUDY AREA AND METHODOLOGY 

The study area is located near Badouzih Harbor, about 10 km northeast of downtown 

Keelung City, northern Taiwan. The annual mean rainfall and averaged rainy days are 

3,772 mm and 197.6 days, respectively (Central Weather Bureau, from 1981-2010). The 


toe of the hillslope is encircled by Highway Route 2 (Figures 1, red rectangular area). The 

study area consists mainly of calcareous massive sandstone belonging to the Taliao Formation 

which features ridges and escarpments along the northern rocky coast. Apparently, as shown 

in Figure 1, the Taliao formation of rockfall area includes two types: one is sandstone (S.S), 

and another one is sandstone and shale interbedded (S.S & Sh. Interbedded). The attitude of 

the bedding plane is approximately N81°E/8°S and the attitude of the slope surface at the 

source area is approximately N70°W/35°N, forming an anaclinal slope. Regarding to the 

regional landforms, both intense erosion and weathering processes along the northern 

coastline forms a landform which characterized by steep cliffs (slope: 35°-80°). In addition, 

the environmental geological map published by the Central Geology Survey also indicated 

that the study area is situated in a high rockfall susceptibility zone. The highest rainfall 

intensity of 94.5 mm/h (August 31, 2013 15:00-16:00) occurred in the Keelung area during 

the influencing period of the Typhoon Kongrey. The rockfall disaster was triggered at 16:19 

after orographically rainfall. The intensity-duration-frequency analysis demonstrated that the 

short-term rainfall of approximately a 100-yr recurrence period plays a dominant role for 

triggering the rockfall event. 

To analysis the Badouzih rockfall event, aerial image interpretation (produced in 2004), 

digital terrain model (resolution: 5 m), field investigation, and rainfall data in the region were 

prepared and analyzed to understand the triggering mechanism and dynamic process firstly. 

Secondly, the parameter of environmental geology and dimension of rockfall were extracted 

to input in the numerical simulation. Three dimensional numerical simulation RAMMS was 

employed to study the dynamic characteristics of the rockfall event. RAMMS is a numerical 

modelling tool which developed by the WSL Institute for Snow and Avalanche Research SLF, 

and is used to predict and assess the natural hazard such as snow avalanche, debris flow, and 

rockfall. The RAMMS:: ROCKFALL module suggests a hard rigid-body approach and involves 

many types of contact drag force on complex terrain (for detailed theoretical background 

please see Leine et al., (2014) and Glover et al., (2015)). RAMMS can compute runout 

trajectory over terrain, including jump heights, velocity, rotational velocity, total kinetic 

energy and contact-impact force with 3D visualization. We defined a similar rock by using 

rock builder while measuring the size of the rock deposited on the highway (Table 1). To 

ensure the appropriate simulation for Badouzih rockfall event, the different geological setting 

and forest roughness were considered to describe the rockfall dynamic characteristics (Table 

2). The initial condition of release rock was set by trial and error method to trigger the rock 

sliding based on the interpretation of video record. Furthermore, the batch of numerical runs 

for rockfall simulation was performed for the purpose of assessing the influence area in this 

study. 


l ee la Table Tab 

Table 1: Parameter used for reconstructing the rockfall shape in the study. 

le 2 

Parameter value reference 

size of rockfall [m] 4.54 * 4.09 * 3.84 m in-situ measuring 

elevation of source area [m] 119 in-situ measuring 

density [kg/m 3 ] 2,650 experimental result 

release point (x, y) (329028, 2781980) aerial image(TWD97) 

initial rotational velocity (X, Y, Z) 

(rad/s) 

(0.3, 0.5, 0) video record 

rock type 

Equant_1.3 

rock mass [ton] 115.44 

rock volume [m 3 ] 42.75 

l calculated ee la by Table Table 

RAMMS 

Table 2: Geologic parameters used in the RAMMS simulation 

No. terrain material friction reference 

1 TI (S.S) colluvium 0.25 (soft) in-situ measuring 

2 TI(S.S and Sh. interbedded) colluvium 0.25 (soft) in-situ measuring 

3 Road and Fishery Harbor concrete 0.40 in-situ measuring 

4 

Forest type Height [m] Drug [kg/s] 

medium 1.5-3.0 1500 in-situ measuring 

RESULTS AND DISCUSSIONS 

With respect to the failure mechanism of rockfall, field investigation shows the event is 

composed of four successive movement behavior from source area to the toe, including 

rolling, falling, bouncing, and rolling (Figure 2). Evidently, one can observe the collapsed rock 

leaves a long trace on the gentle slope, and falls at the breakpoint on the cross-sectional 

profile (point a in Figure 2(a)). A falling vertical height of 40 m impacted and made a deep pit 

in the soft talus material (point b in Figure 2(b)). The following bouncing down the slope 

Figure 2: (a) Analysis of rockfall trajectoryon the cross-sectionalprofile (revised form Wei et al., (2014)) and (b) field survey verification 

displayed on aerial image taken by UAV (source: GIS Center, FCU, Taiwan). 


crashed into the houses situated along the rockfall path. The rock was rolling in the lower 

part of the talus before it stopped on the highway. 

Figure 3(a) shows 3D rockfall trajectory, contact points, and velocity as a function of horizontal 

distance. In comparison with the observed rockfall path (erosion marks) on the aerial image, 

both rockfall trajectory and stopping point in RAMMS simulation is much similar to the 

mapping result (Figure 3(c), (d)). The post-disaster mapping form the aerial image confirmed 

the exactly trajectory of rockfall and was compared with the simulation. The simulated result 

indicates the rolling motions of rock occurs on the upper part of the the gentle slope (H=55 to 

120 m), then it increases the moving velocity at the steep slope with bouncing until it impact 

the talus (H=25 to 55 m). As shown in Figure 3(b), the RAMMS::ROCKFALL model predicts a 

maximum rockfall speed of 20.5 m/s at the breakpoint of the slope during the bouncing stage 

of the rockfall trajectory. The contact points and movement patterns predicted by the model 

match with field observations and preliminarily demonstrate that the RAMMS::ROCKFALL 

model accurately depicts the rockfall disasters. 

Figure 3: The results of numerical simulation RAMMS-Rockfall: (a) 3D rockfalltrajectory and (b) velocity and contact points for Badouzih 

rockfall event; (c) the real rockfalltrajectory on the aerial image and (d) simulation trajectory in RAMMS. 


Figure 4 shows rockfall speed and trajectory predicted by RAMMS compared to 2D model 

results reported by Wei et al. (2014). In general, both speed and trajectory predicted by the 

3D model are less than 2D model predictions. However the distribution of speed and kinetic 

energy is similar. Differences between the models may be due to (1) Rock shape: in the 2D 

rockfall model, rocks are represented as particles and have few contact points with the slope; 

therefore the resistance to movement is lower. (2) Topographic model accuracy: a DTM of 

the actual landslide terrain is used in the RAMMS model and rockfall is affected by the three 

dimensional characteristics of the terrain. Consequently, the travel path of the particles in 

the three dimensional model differ from the travel path affected by a single two dimensional 

profile and resistance to movement resulting from topography is more evident. (3) Forest 

drag: using parameters that describe the vegetative cover of the terrain, RAMMS applies 

a corresponding drag and rock fall is subjected to an additional energy dissipation effect. 

Figure 4: Comparison of velocity and kinetic energy for the simulation results between 3D RAMMS and 2D Rockfall. 

The accuracy of the DTM affects the surface roughness applied to the rockfall. As the accuracy 

of the DTM is increased to levels typical of LiDAR measurements, the detailed representation 

of the topography causes surface roughness to increase. Crosta et al., (2015) modeled rockfall 

using a 2, 6 and 20 m resolution DTM. The high resolution DTM caused dispersion of the 

particle trajectories. To better understand the distribution and “hotspot” of rockfall predicted 

by RAMMS::ROCKFALL model, applying the same initial conditions, the calculations were 

repeated 100 times. Results are shown in Figure 5 for a 5 meter resolution DTM. Within the 


study area, there are two zones which have a high potential for rock fall. The gully at the left 

of the source area has a 55% probability of rockfall and is near the travel path of rockfall that 

has historically caused rock fall disasters. Additionally, 30% of the rockfall travels along the 

right gully and the last 15% of the rockfall volume remains suspended on the slope. In this 

study, the rockfall energy classification methodology described in Wei et al., (2014) was 

applied to categorize the hazard level of the modeled rockfall trajectories. Results demonstrate 

that Highway Route 2 and parts of the fishing port are within the rockfall hazard area. Mean 

velocity and kinetic rockfall energy are predicted to be 6.9 m/s and 5,968 KJ and deposition 

occurs 40 to 80 m (52%) from the actual rockfall location associated with the rockfall disaster. 

These results demonstrate that the RAMMS::ROCKFALL model is more effective than the 2D 

model for identifying rockfall disaster extent. Therefore, RAMMS may be suitable for warning 

and evacuation planning analysis. 

Figure 5: The kinetic rock energy map generated from batch runs in RAMMS: Rockfall (grid cell: 5x5 m). 

CONCLUSIONS 

The well-known Badouzih rockfall hazard triggered by rainfall on August 31st, 2013 in 

Taiwan was explored by 3D numerical approach in the study. RAMMS::ROCKFALL is able to 

integrate the detailed block shape, terrain material, and initial condition to modelling 

three-dimensional rockfall event. With comparing to the 2D numerical simulation in 

Badouzih rockfall event, the 3D simulation reveals the resolution of input DTM (surface 

roughness) and rock shape are main influenced factors to control the moving trajectory under 


the same initial and boundary conditions. However, the rolling, slipping and bouncing 

trajectory in Badouzih rockfall simulation presents a similar result in comparison with field 

survey. For the demand of reducing disaster consequences, the hazard map which associated 

with rock mass strength assessment (frequency) and numerical simulation (intensity) around 

alpine region can help predicting future rockfall occurrence. 

REFERENCES 

- Agliardi F., Crosta G.B., Frattini P. (2009). Integrating rockfall risk assessment and counter 

measure design by 3D modeling techniques. Nat Hazards Earth Syst Sci 9:1059-1073. 

- Chen G.Q., Zhen L., Zhang Y.B., Wu J. (2013). Numerical simulation in rockfall analysis: a 

close comparison of 2D and 3D DDA. Rock Mech Rock Eng 46:527-541. 

- Cottaz Y., Barnichon J.D., Badertscher N., Gainon F. (2010). PiR3D, an effective and 

userfriendly 3D rockfall simulation software: formulation and case-study application. Rock 

Slope Stability Symposium, Paris. 

- Crosta G.B., Agliardi F., Frattini P., Lari S. (2105). Key issues in rock fall modeling, hazard 

and risk assessment for rockfall protection. G. Lollino et al. (eds.), Engineering Geology for 

Society and Territory 2: 43-57. 

- Glover J., Bartelt P., Christen M., Gerber W. (2015). Rockfall-simulation with irregular rock 

blocks. G. Lollino et al. (eds.), Engineering Geology for Society and Territory 2: 1729-1733. 

- Leine R.I., Schweizer A., Christen M., Glover J., Bartelt P., Gerber W. (2014). Simulation of 

rockfall trajectories with consideration of rock shape. Multibody Syst Dyn 32(2): 241-271. 

- Wei L.W., Chen H., Lee C.F., Huang W.K., Lin M.L., Chi C.C., Lin H.H. (2014). The mechanism 

of rockfall disaster: A case study from Badouzih, Keelung, in northern Taiwan. Engineering 

Geology 183: 116-126. 



Leben 

mit 

Naturrisiken

Hazard and Risk Mitigation 

(structural, nonstructural measures, insurance) 


HAZARD AND RISK MITIGATION (STRUCTURAL, NONSTRUCTURAL MEASURES, INSURANCE) 

Integrated bed-load and driftwood retention in Kien - 

Findings from model-based testing and the 2011 flood 

Geschiebe- und Schwemmholzrückhalt Kien - 

Erkenntnisse aus den Modellversuchen und dem 

Hochwasser 2011 

Guido Lauber, Dr. 1 ; Jürg Speerli, Prof. Dr. 2 ; Warin Bertschi, M.D. 1 ; Armin Hemmi 1 

ABSTRACT 

During the 2005 flood,prolonged rainfall and high peak flows led to large quantities of bedload 

and driftwood causing overtopping at several locations. Following the events a protection 

project was launched which included a driftwood rack as well as a sediment retention 

basin. These solutions were designed and optimised using a hydraulic model. Between 

2007-2010 the various phases of the project were executed,including the expansion of an 

existing channel in the village. Throughout the flood of 2011, with peak flows matching a 

1/100-year event, hardly any driftwood was collected when compared to the flood in 2005. 

Such a scenario was considered unlikely and would have led to a large sediment discharge 

(lower efficiency of the sediment retention system). During the latter stages of the flood, 

the sediment discharge remain low as sufficient small-sized deadwood blocked the openings 

of the retention system. The design of these openings as well as the prediction and significance 

of driftwood are linked to the interaction of the floods as well as the bed-load and 

driftwood movement, remaining very complex and limited by uncertainties. 


Das Hochwasser vom 20.-22. August 2005 führte im Alpenraum zu grossräumigen Überschwemmungen. 

Im Kanton Bern war vor allem das Berner Oberland stark betroffen. 

Die Chiene beförderte sehr grosse Geschiebe- und Schwemmholzmengen zu Tal, was in den 

Dörfern Kien und Reichenbach im Kandertal zu massiven Ausuferungen führte. Auf der 

Grundlage einer detaillierten Ereignisanalyse wurden die Schutzprojekte entlang der Chiene 

in Kien und weiter oben im Kiental Boden erarbeitet. In Kien wurden u.a. eine Aufweitung 

des Gerinnes und ein Geschiebe- und Schwemmholzrückhalt projektiert. Das komplizierte 

Zusammenspiel von Geschiebe und Holz im Rückhaltebecken wurde in Modellversuchen an 

der HSR Hochschule für Technik Rapperswil untersucht und die daraus gewonnenen 

Erkenntnisse in der Projektierung berücksichtigt. Die verschiedenen Schutzbauwerke wurden 

in Etappen zwischen 2007 und 2010 realisiert. So war das Gesamtprojekt zum Zeitpunkt des 

Hochwassers im Oktober 2011 fertiggestellt und wurde einem ersten 1:1 Stresstest unterzo- 

1 Emch+Berger AG, Spiez, SWITZERLAND, spiez@emchberger.ch 

2 HSR Hochschule für Technik Rapperswil, Switzerland 



gen. In diesem Bericht werden kurz die Kernaussagen aus der Projektierung und Modellierung 

zusammengefasst, um daraus die Erkenntnisse aus dem Hochwasser 2011 aufzuzeigen. 

KEYWORDS 

flood protection; hydraulic modeling; event documentation; driftwood; bedload retention 

GEFAHRENSITUATION UND SCHUTZKONZEPT 

Während dem Sommerhochwasser 2005 hat die Chiene in Kien eine Abflussspitze von 

120-140 m 3 /s (HQ 100 

: 120 m 3 /s) erreicht. Insgesamt wurden während dem Ereignis 80‘000- 

100‘000 m 3 Geschiebe und 4‘000-6‘000 m 3 Schwemmholz mobilisiert und ein Grossteil davon 

auf dem Schwemmkegel in Kien abgelagert (s. Abbildung 1). 

Im Rahmen der Lokalen Lösungsorientierten Ereignisanalyse (Emch+Berger et al., 2006) 

wurden zum Schutz der Siedlungsgebiete Kien und Reichenbach eine Vielzahl unterschiedli- 

Abbildung 1: Aufnahmen vom Chiene-Hochwasser 2005 im Dorf Kien. Linkes Bild: Luftaufnahme vom Dorf Kien mit dem Standort der 

Dorfbrücke (roter Kreis). Rechtes Bild: Holzablagerungen bei der Dorfbrücke nach dem Hochwasser. 

cher Schutzmassnahmen geprüft. Die Bestvariante wurde im Rahmen eines Wasserbauplans 

(Emch+Berger et al., 2006) weiterbearbeitet und genehmigt. Sie sieht oberhalb von Kien 

auf dem Kegelhals die Realisierung einer Geschiebesperre mit vorgelagertem Schwemmholzrechen 

(s. Abbildung 2) sowie im Unterlauf (Schwemmkegel) eine Aufweitung des Gerinnes 

mit Erosionsschutz mittels Blocksteinschwellen und Blocksteinrampen bis zur Mündung in 

die Kander vor (Ausbauwassermenge plus Freibord). Der Dimensionierung wurde das 

Hochwasserereignis vom August 2005 zugrunde gelegt mit 70‘000 m 3 Geschiebe, 5‘000 m 3 

Schwemmholz und einer Abflussspitze von 120 m 3 /s. 


Abbildung 2: Linkes Bild: Hochwasserschutzmassnahmen Chiene in Kien - Übersicht Gesamtkonzept mit Rückhaltebecken (grün), Gerinneaufweitung 

und –stabilisierung mittels Blocksteinschwellen (orange), Blocksteinrampen (gelb), neue Dorfbrücke (pink), Damm 

Überlastkorridor (orange Linie). Rechtes Bild: Bau des vorgelagerten Schwemmholzrechens (2010) mit dahinterliegender Geschiebesperre. 

FUNKTION VON GESCHIEBESAMMLERN 

Die wichtigste Funktion eines Geschiebesammlers ist der Rückhalt von Geschiebe. 

Die Geometrie des Abschlussbauwerkes beeinflusst den Geschieberückhalt in grossem Masse 

(Zollinger 1983). Folgende Mechanismen führen zu einer Geschiebeablagerung im Sammlerraum 

(s. Abbildung 3): 

– (1) Reduktion der Abflusstiefe durch Gerinneaufweitung 

– (2) Reduktion des Sohlengefälles 

(3) Wasseraufstau und Seebildung im Geschiebesammler 

– (4) direkte Behinderung des Geschiebetriebes durch das Abschlussbauwerk 

Abbildung 3: Ablagerungsmechanismen in einem Geschiebesammler im Grundriss a) und Schnitt b): Gerinneaufweitung (1), 

Gefällsknick (2), Wasseraufstau (3), direkte Behinderung des Geschiebetriebs durch Abschlussbauwerk (4) (Zollinger 1983). 

Mit einer dem Abschlussbauwerk vorgelagerten Rechenkonstruktion wird das Schwemmholz 

im Zustand der Seebildung durch den Rechen zurückgehalten (Lange und Bezzola 2006). 


MODELLVERSUCHE GESCHIEBE- UND SCHWEMMHOLZRÜCKHALT 

Zur Überprüfung der Interaktion von Schwemmholz- und Geschieberückhalt wurden vorab 

an der HSR Versuche am hydraulischen Modell (Mstb. 1:35) durchgeführt (Speerli und 

Stucki, 2009). Abflüsse mit freier Oberfläche stehen unter dem Einfluss der Trägheits- und 

Schwerkraft, womit die Froude-Zahl massgebend wird. So können mit dem Massstabsfaktor 

= 35 und dem Ähnlichkeitsgesetz nach Froude die Werte aus dem Modell in Naturwerte 

umgerechnet werden. 

Als Eingangsgrössen dienten einerseits das Hochwasser 2005 mit unterschiedlichen 

Schwemmholzmengen sowie das EHQ Ereignis. Von einem Schwemmholzaufkommen in 

der Chiene kann aufgrund der Einzugsgebietsverhältnisse grundsätzlich ausgegangen werden. 

Deshalb stellt ein grosses geschiebeführendes Hochwasser ohne Holz ein eher unwahrscheinliches 

Extremereignis dar. 

Der Geschieberückhalt ist so konzipiert, dass sich ab einem Abfluss von 45 m 3 /s bzw. einer 

Verklausung der Durchlässe im Geschiebesammler ein See bildet und Schwemmholz an der 

Wasseroberfläche in Richtung Schwemmholzrechen schwimmt. 

LAGE UND DIMENSION DES RECHENS 

Bei grossen Holzmengen, wie sie vor allem im Überlastfall zu erwarten sind, bildet der vorgelagerte 

Schwemmholzrechen quasi einen zweiten, vorgelagerten höheren Geschieberückhalt. 

Einerseits können dadurch im Überlastfall auch beträchtlich grössere Geschiebemengen 

als die Dimensionierungsmengen zurückgehalten werden, andererseits steigen aber auch 

die Sohle und damit der Wasserspiegel im Rückhaltebecken entsprechend an (s. Abbildung 

4). Das rechnerische Bestimmen dieser Effekte ist mit einigen Unsicherheiten behaftet. 

Im hydraulischen Modell konnten unterschiedliche Szenarien untersucht und entsprechende 

Wasserspiegelkoten bestimmt werden (Sensitivitätsanalyse). Es sollte ein möglichst robustes 

Rückhaltesystem gefunden werden, welches unabhängig von diesen Unsicherheiten für 

grosse und kleine Geschiebe- und Schwemmholzmengen zuverlässig funktioniert. 

Abbildung 4: Linkes Bild: Schwemmholzrechen Varianten mit kurzem und langem Rechen. Rechtes Bild: Prinzipskizze: Einfluss Standort 

des Rechens auf Dammhöhe Geschiebesammler. 


Bei der Variante „kurzer Rechen“, welcher an den linksseitigen Erddamm des Sperrenbauwerks 

anschliesst, führt der Auflandungsprozess im Geschiebesammler zu einem sehr hohen 

seitlichen Damm (s. Abbildung 5): Eine der Hauptströmungen im Geschiebesammler erfolgte 

mehrheitlich entlang der linksseitigen natürlichen Berandung und des anschliessenden 

Erdammes. Dies führte dazu, dass sich das Geschiebe oberhalb des Rechens in hohem Masse 

in diesem Bereich ablagerte. Dies ist einerseits aus ästhetischen Gründen nicht erwünscht 

und andererseits bezüglich der Befahrbarkeit ungünstig, weil die Zufahrtstrasse über den 

Damm geführt werden muss (Platzbedarf, Steigungen). 

Abbildung 5: Linkes Bild: Prinzipskizze Variante kurzer Rechen, rechtes Bild: Modellversuch. 

Deshalb wurde eine etwa 15 m stromaufwärts verschobene Variante mit „langem Rechen“ 

geprüft, welche nicht an den Damm, sondern direkt an das neben der Strasse steil ansteigende 

natürliche Terrain anschliesst (s. Abbildung 6). Dadurch verteilt sich der enorme Holzteppich 

besser über den Rechen und vor allem bleibt der linksseitige Damm des Sperrenbauwerks 

praktisch unbeeinflusst von der Höhe des Schwemmholz-Rückstaus, was sich auch 

sehr positiv auf die Überlastsicherheit desselben auswirkt. 

Der lange Schwemmholzrechen ist 80 m lang, die Rechenstäbe 10 m hoch über der Flusssohle, 

die Rechenstäbe 1 m dick mit einem lichten Stababstand von 2.7 m dazwischen. 

Abbildung 6: Linkes Bild: Prinzipskizze Variante langer Rechen, rechtes Bild: Modellversuch am langen Rechen. 


GESCHIEBEAUSTRAG MIT UND OHNE SCHWEMMHOLZ 

Wie erwähnt, wird der Geschieberückhalt durch grosse Mengen Schwemmholz am Rechen 

erhöht. Wenn genügend Holz im Rechen verklaust, passiert praktisch kein Geschiebe den 

Rechen, dies gilt aber nur bis zum Erreichen der Hochwasserspitze (HQ 100 

). Beim abklingenden 

Hochwasser wird auch bei einem noch so mächtigen Holzteppich am Rechen eine nicht 

zu vernachlässigende Menge an Geschiebe unter dem Holzteppich hindurch ausgetragen, 

natürlich vorausgesetzt, der Sperrendurchlass ist nicht verklaust. Dies ist vor allem damit zu 

erklären, dass (sobald keine Seebildung mehr vorhanden ist) die Strömung im Geschiebesammler 

das abgelagerte Lockermaterial ohne Deckschicht und Verfestigung sehr einfach 

mobilisieren kann und der Gradient am Ende der Ablagerungen zur Sperre hin sehr steil ist 

(s. Abbildung 7). 

Abbildung 7: Linkes Bild: Geschiebeaustrag beim abklingenden Hochwasser ohne See, rechtes Bild Situation bei Seebildung. 

Im Versuch HQ 100 

, lang 

mit Schwemmholz betrug die Austragsrate von Geschiebe immerhin 

35‘000 m 3 von insgesamt 60‘000 m 3 (s. Abbildung 8 links). 

Wenn gar kein Schwemmholz vorhanden ist, breiten sich die Geschiebeablagerungen aufgrund 

der Seebildung deltaförmig von der Stauwurzel bis zur Geschiebesperre aus und der 

Geschiebeaustrag beginnt, sobald das Geschiebe die Sperre erreicht hat. Im absteigenden Ast 

der Hochwasserganglinie verschwindet die Seebildung und die Geschiebeaustragungsrate 

Abbildung 8: Linkes Bild: Geschiebeaustrag im hydraulischen Modell H Q100 lang 

mit Schwemmholz, rechtes Bild: HQ 100 lang 

ohne 

Schwemmholz. Der Geschiebeaustrag wurde während dem Ganglinienversuch insgesamt über 8 Zeitintervalle gemessen. 


steigt massiv an, falls kein Schwemmholz vorhanden ist. Dies führte zu einer Austragsrate 

von 50‘000 m 3 der insgesamt 60‘000 m 3 . Der Geschiebesammler wird also praktisch wieder 

entleert (s. Abbildung 8 rechts). 

Der Geschiebeaustrag kann durch eine Verkleinerung der Durchlassöffnungen reduziert 

werden. Dadurch werden aber auch kleinere Hochwasser zu Geschiebeablagerungen im 

Becken führen, was wiederum aufgrund der steigenden Unterhaltskosten (häufigere Ausbaggerungen) 

nicht erwünscht ist. Hier musste ein Optimum gefunden werden zwischen 

Sicherheit und Unterhalt. 

Wesentliche Erkenntnisse aus den Modellversuchen: 

– Die hydraulischen Modellversuche zeigen, dass ohne Schwemmholz und bei der Realisierung 

von zu grossen Durchlassöffnungen der Wirkungsgrad des Geschiebesammlers stark 

reduziert wird. 

– Mit kleinen Durchlassöffnungen und einem minimalen Holzeintrag von 1'000 m 3 konnte 

ein deutlich höherer Wirkungsgrad erreicht werden als ohne Schwemmholzbeigabe 

(Schwemmholzteppich am Rechen verzögert bzw. reduziert den Geschiebeaustrag). 

– Die Abmessungen der Durchlassöffnungen, die Lage und Dimension des Schwemmholzrechens 

sowie die Höhe des linksseitigen Dammes konnten ermittelt werden. Die beiden 

Durchlassöffnungen weisen eine Höhe von 0.75 m auf. Die ursprüngliche Öffnungsbreite 

wurde von je 7.5 m auf je 5.0 m reduziert. Die Grösse der Durchlassöffnungen kann durch 

Entfernen oder Hinzufügen von Stahlbalken verändert werden, so dass diese später nach 

Bedarf an die Erkenntnisse und Erfahrungen angepasst werden können. 

– Im Überlastfall (EHQ = 200 m 3 /s) wird die Strasse über den Rückhaltedamm mit mobilen 

Dammbalken versperrt. So kann bei einem Holzeintrag von 10‘000 m 3 knapp 137‘000 m 3 

Geschiebe zurück gehalten werden. Das entspricht mehr als dem doppelten potentiellen 

Rückhaltevolumen. 

HOCHWASSER 2011 

Am 10. Oktober 2011 hat das Zusammentreffen von intensiven Niederschlägen mit der 

Schmelze von Neuschnee ein starkes Hochwasser im Kandertal und insbesondere im Teileinzugsgebiet 

der Chiene verursacht. Für die Chiene in Kien wurde eine Abflussspitze von 

ca. 80-110 m 3 /s rekonstruiert. Insgesamt wurden im Geschiebesammler von Kien 25‘000- 

30‘000 m 3 Geschiebe aber lediglich 100-150 m 3 loses Schwemmholz abgelagert (s. Abbildung 

9). Das Ereignis vom Oktober 2011 war im Vergleich zum Hochwasser 2005 kleiner bzw. es 

handelte sich um ein kurzes Ereignis mit hoher Abflussspitze. Dadurch wurden eine deutlich 

kleinere Geschiebemenge und praktisch kein Schwemmholz (ausser geringe Mengen Totholz) 

mobilisiert. Die kleinen Mengen sind auch dadurch zu erklären, dass beim Ereignis 2005 

bereits sehr viel Material ausgeräumt wurde, bzw. die Bacheinhänge im Nachgang an das 

Ereignis ausgeholzt wurden. 


Abbildung 9: Fotos vom Hochwasser vom 10. Okt. 2011 (oben links) Schwemmholzrechen (oben rechts) Überfall Geschiebesperre (unten 

links) aufgeweitete Chiene oberhalb der Dorfbrücke (unten rechts) Geschiebe- und Schwemmholzablagerungen oberhalb des Rechens. 

ERKENNTNISSE UND DISKUSSION 

Nachfolgend sind die wesentlichen Erkenntnisse der hydraulischen Modellversuche und 

der während dem Hochwasser 2011 gemachten Beobachtungen aufgeführt: 

– Ein zuverlässiger und robuster Geschieberückhalt wird durch die Seebildung und eine 

entsprechend dimensionierte Durchlassöffnung gewährleistet. 

– Aber auch bei Seebildung beginnt ab dem Zeitpunkt, ab welchem der Sammler mit 

Geschiebe gefüllt ist, ein nicht unbedeutender Austrag von Geschiebe. 

– Dieser Austrag wird durch allfälliges Schwemmholz im Rechen reduziert (wie die Modellversuche 

zeigten). 

– Die Bandbreite der bei einem Ereignis angeschwemmten Schwemmholzmenge kann sehr 

gross sein, wie die Ereignisse 2005 (4‘000-6‘000 m 3 ) und 2011 (100-150 m 3 ) gezeigt haben. 

– In Kien wurden die Durchlassöffnungen so dimensioniert, dass bei einem Abfluss von 

45 m 3 /s die Seebildung einsetzt. Mit diesen Öffnungen wird im Modell ohne Schwemmholz 

ein Geschiebeaustrag erreicht, welcher an der obersten akzeptierbaren Grenze liegt 

(Optimierung zwischen Unterhaltskosten und Sicherheit). 

– Beim Ereignis 2011 war der Schwemmholzanfall sehr gering, so dass das wenige Schwemmholz 

am Rechen für den Geschiebeaustrag ohne Bedeutung blieb. 


– Weil aber bereits geringe Mengen an Totholz genügten, um die kleinen Durchlassöffnungen 

an der Sperre zu verklausen, wurde der Geschiebeaustrag beim abklingenden 

Hochwasser reduziert. 

– Der Rechen ist nun so konzipiert, dass grössere Baumstämme zurückgehalten werden, 

gröbere Äste und kleinere Stämme aber bei einem ansteigenden Hochwasser den unteren 

Bereich des Rechens möglichst passieren können, bis schliesslich bei vollständiger 

Seebildung im oberen Bereich (durch ein grobes Netz) sämtliches Holz zurückgehalten 

werden soll. Damit soll ein Verklausen der Durchlässe durch kleinere Holzteile beim 

ansteigenden Hochwasser bewusst ermöglicht werden (Reduktion des Geschiebeaustrags 

beim abklingenden Hochwasser). 


Beim Hochwasser 2005 führten lang anhaltend starke Niederschläge neben hohen Abflussspitzen 

vor allem zu sehr grossen Geschiebe- und Schwemmholzmengen, welche schliesslich 

grossflächige Ausuferungen verursachten. Der daraufhin projektierte Geschiebesammler 

mit Schwemmholzrechen wurde im hydraulischen Modell überprüft und optimiert. 

Zwischen 2007 - 2010 wurden die Schutzmassnahmen umgesetzt, d.h. in Etappen wurden 

der Geschiebesammler und der Schwemmholzrückhalt gebaut sowie das Gerinne im Dorf 

aufgeweitet. Während dem Hochwasser 2011, mit der Abflussspitze eines knapp 

100-jährlichen Ereignisses, wurde im Gegensatz zu 2005 kaum Schwemmholz mobilisiert. 

Ein derartiges Ereignis wurde während der Konzeption des Schutzprojektes als eher unwahrscheinliches 

Extremereignis betrachtet, welches zu einem erhöhten Geschiebeaustrag führen 

könnte (tieferer Wirkungsgrad Geschieberückhalt). 

Der Geschiebeaustrag beim abklingenden Hochwasser 2011 war trotzdem klein, weil klein - 

formatiges Totholz genügte, um die relativ schmalen Durchlassöffnungen zu verklausen. 

Die Dimensionierung der Durchlassöffnungen von Geschiebesammlern sowie die Prognose 

und Berücksichtigung von Schwemmholz bei der Auslegung einer Schutzmassnahme sind 

durch die Interaktion der Prozesse Hochwasser, Geschiebe- und Schwemmholztransport 

komplex und mit Unsicherheiten behaftet. Hydraulische Modellversuche können dazu 

beitragen, das Prozessverständnis massgeblich zu erhöhen. 


LITERATUR 

- Emch+Berger AG Bern, Hunziker, Zarn & Partner AG und Geotest AG (2006). 

Lokale Lösungs-orientierte Ereignisanalyse (LLE) Reichenbach. 

- Emch+Berger AG Bern und Hunziker, Zarn & Partner AG (2006). Wasserbauplan 

Hochwasserschutz Chiene in Kien. 

- Speerli J., Stucki A., HSR Hochschule für Technik Rapperswil, Institut für Bau und Umwelt 

(2009). Modellversuche Chiene. 

- Lange D., Bezzola G.R., VAW ETH Zürich (2006), VAW Mitteilung Nr. 188, Schwemmholz - 

Probleme und Lösungsansätze. 

- Zollinger, F., ETH Zürich (1983), Die Vorgänge in einem Geschiebeablagerungsplatz. 

Ihre Morphologie und die Möglichkeit einer Steuerung. Dissertation an der Eidgenössischen 

Technischen Hochschule Zürich, Diss. ETH Nr. 7419. 



Freeboard calculation of near critical flow 

Calcul de la revanche pour les écoulements proche 

d'un écoulement critique 

Niki Antonina Beyer Portner, Dr 1 ; Patrick Fellay, Bachelor of science in Civil Engineering 2 ; Karim Laribi, Master of science in 

Civil Engineering 3 ; Jean-Louis Boillat, Dr 4 

ABSTRACT 

The hazard assessment of the Vièze River crossing the city of Monthey showed that larges 

zones in the city could be flooded due to hydraulic capacity limitation of bridges. 

The authorities of Monthey thus decided to elaborate a flood protection project. In this 

context, a group of specialists was appointed to evaluate possible protection measures. 

In 2013, when the project was practically achieved, the Commission for Flood Protection 

of Switzerland published new recommendations concerning the freeboard of rivers. 

The required freeboard takes into account the uncertainty related to hydraulic and sediment 

transport computations as well as to drift blockage at bridges. The proposed conceptual 

approach is commonly applicable but requires to be completed in particular cases. For the 

Vièze River, the strict application of the guidelines leads to an unexpected over sizing, 

justifying the implementation of an adequate optimization process. The main focus was put 

on the water surface oscillations due to near-critical flow conditions. 

RESUME 

La carte des dangers de la Vièze qui traverse Monthey montre que de larges zones de la ville 

peuvent être inondées suite à une sous-capacité hydraulique de certains ponts. 

Les autorités de la ville ont par conséquent décidé de faire élaborer un projet de protection 

contre les crues. Un groupement de spécialistes a été mandaté pour développer des mesures 

de protections. 

En 2013, alors que le projet était pratiquement terminé, la Commission pour la Protection 

contre les Crues (CIPC) a publié des recommandations pour la détermination de revanche 

applicable aux aménagements de rivières. La revanche doit tenir compte des incertitudes dans 

le calcul de la ligne d’eau et des sédiments et du blocage de flottants aux ponts. Cette méthode 

permet de traiter la plupart des cas pratiques. Dans des cas particuliers elle doit être complétée 

ou étendue. Dans le cas de la Vièze, une application stricte conduit à un surdimensionnement, 

ce qui justifie une optimisation. 

L’accent est mis sur les ondulations de surface pour les écoulements proche d’un écoulement 

critique. 

1 HydroCosmos SA, Billens, SWITZERLAND, niki.beyer@hotmail.ch 

2 Ville de Monthey 

3 CERT ingénierie SA 

4 Hydro Boillat 



KEYWORDS 

Flood protection, freeboard, surface undulations 

INTRODUCTION 

La Vièze traverse la ville de Monthey, située dans la plaine du Rhône dans le Canton du 

Valais. Elle a donc une pente faible d’environ 1 %. La carte des dangers de 2002 (GILAT – 

ETEC, 2002) montre que la ville est menacée d’inondations, souvent d’intensité moyenne 

avec des profondeurs d’écoulement entre 0.5 et 2.0 m, dues principalement à la sous-capacité 

hydraulique de certains ponts. La Ville de Monthey a alors décidé de lancer un projet de 

protection contre les crues. 

Un groupement d’ingénieurs, d’urbanistes et de spécialistes en environnement a été mandaté 

pour évaluer les mesures possibles de protection contre les crues. Les objectifs concernent la 

protection contre les crues de la ville et l’intégration de la Vièze dans le réseau écologique de 

la plaine du Rhône. Les mesures de protection en ville requièrent une intégration harmonieuse 

dans le paysage urbain. 

L’analyse économique du projet (optimum des dommages potentiels par rapport aux coûts 

de construction) a confirmé que la condition de dimensionnement correspond à une crue 

centennale de 200 m 3 /s. Le projet élaboré consiste en une combinaison de différentes mesures : 

surélévation des digues en ville, rehaussement de certains ponts, abaissement du lit sur le 

Figure 1 : Situation de la Vièze et aménagements proposés. 


dernier kilomètre avant la confluence avec le Rhône et un élargissement de 500 m de long 

sur le secteur aval (cf. figure 1). Cet élargissement est dédié à la revitalisation du cours d’eau, 

au développement de la dynamique morphologique du lit et au rétablissement d’une 

connectivité pour la faune terrestre et aquatique. Les coûts du projet de construction sont 

évalués à 12 millions d’Euro. 

RECOMMANDATIONS CIPC 

En 2013, alors que le projet était pratiquement terminé, la Commission pour la Protection 

contre les Crues (CIPC) de l’Association Suisse pour l’Economie des Eaux a publié des 

recommandations pour la détermination de revanche applicable aux aménagements de 

rivières (CIPC, 2013). Ces recommandations visaient à standardiser la détermination de la 

revanche en Suisse en lien avec des paramètres spécifiques au projet, tenant compte du 

comportement hydraulique, du transport sédimentaire et du blocage de flottants contre les 

tabliers des ponts. Selon les recommandations, la revanche est ainsi composée de plusieurs 

éléments (cf. tableau 1), combinés par addition géométrique. En outre, la revanche doit être 

de 0.3 m au minimum et de 1.5 m au maximum pour les rivières du type de la Vièze. 

Pour la CIPC et les autorités de surveillance des aménagements de rivières, il est clair que ces 

recommandations doivent être appliquées aux projets cours d’eau, mais qu’elles peuvent être 

complétées ou étendues de cas en cas. Des dérogations sont ainsi possibles dans des situations 

particulières, si elles sont justifiées et étayées. 

Tableau 1 : Eléments composant la revanche selon les recommandations CIPC (CIPC, 2013). 

Dans le cas de la Vièze, l’écoulement de la rivière est proche de la condition critique (nombre 

de Froude, F=1) et la prise en compte de la hauteur d’énergie équivaut à systématiquement 

fixer la revanche à sa valeur maximale. Pour la géométrie actuelle, en considérant 1.5 m de 

revanche, la capacité actuelle de la Vièze se réduit à 80 m 3 /s, alors qu’elle a été dimensionnée 

en 1941 à ras-bord pour 200 m 3 /s environ (cf. figure 2). Le rehaussement des berges 


permettant de satisfaire une revanche de 1.5 m transformerait le cours d’eau en canal, 

confiné de part et d’autre par des hauts murs perchés au-dessus de la plaine habitée. Ceci 

impliquerait un impact important sur le paysage et induirait un risque potentiel supplémentaire 

en cas de rupture de digue. La stricte application des recommandations CIPC conduit 

ainsi à un surdimensionnement inattendu du profil en travers. 

Figure 2 : Capacité de la Vièze avec un écoulement à ras-bord et avec une revanche selon CIPC. 

MÉTHODOLOGIE 

Un modèle unidimensionnel avec transport solide a été utilisé pour analyser les écoulements 

de la Vièze à l’état actuel et à l’état de projet. Le modèle a été établi à l’aide du logiciel Dupiro, 

développé par HydroCosmos SA (www.hydrocosmos.ch) et la rugosité a été calée à l’aide 

d’observations lors de crues récentes. Le diamètre moyen a été déterminé sur la base d’un 

levé en ligne. Deux crues ont été analysées pour l’état de projet : une crue centennale avec 

200 m 3 /s de débit de pointe et une crue extrême avec 300 m 3 /s de débit de pointe. Le volume 

des sédiments transportés correspond à la capacité de transport à la limite amont du modèle. 

Pour la crue centennale, le volume correspond au volume annuel de sédiments estimé pour 

la Vièze. Il est encore supérieur à ce dernier pour la crue extrême. 

La Vièze est un cours d’eau canalisé entre l’entrée en ville et la confluence au Rhône. Elle 

est caractérisée par un tracé pratiquement rectiligne, une section quasi prismatique et par 

l’absence d’obstacles susceptibles de perturber l’écoulement. 


Toutefois, avec un nombre de Froude F variant entre 0.9 et 1.9 avec une moyenne de 1.2, 

la Vièze présente un écoulement systématiquement proche de la condition critique, F=1. 

La nature des écoulements qui en résulte est particulièrement instable car soumise à une 

alternance répétée de régimes fluviaux et torrentiels. Cette condition favorise la formation 

de ressauts hydrauliques, caractérisés par le passage d’un écoulement torrentiel (F>1) à un 

écoulement fluvial (F

où F1 est le nombre de Froude de l’écoulement amont et h1 et h2 les hauteurs d’eau aux 

limites amont, respectivement aval, du ressaut. 

RÉSULTATS 

Il est ainsi proposé de calculer l’amplitude des vagues de surface en application de la formule 

de Bélanger (équation 1). Les calculs effectués montrent que la revanche nécessaire associée 

à la formation de vagues est de l’ordre de fv = 0.5 m pour la crue de dimensionnement, 

respectivement fv = 0.6 m pour la crue extrême de 300 m 3 /s. 

En intégrant les autres éléments de la revanche selon les recommandations CIPC (cf. tableau 

2), la revanche nécessaire varie entre le minimum de 0.3 m et 1.1 m (cf. figure 4). Sur la 

figure 4, les endroits où l’écoulement est proche de l’écoulement critique sont visibles car la 

revanche est supérieure au minimum. Le pic vers le kilomètre 1400 est dû à un changement 

de section. La figure 4 montre également que la revanche peut être garantie sur de longs 

tronçons pour la crue de dimensionnement. Il reste un tronçon d’un kilomètre où les berges 

doivent être rehaussées. Ces interventions restent cependant modestes et peuvent être 

intégrées dans le paysage urbain. Cette réduction de la revanche est complétée par une 

analyse des cas de surcharge afin d’adapter les aménagements de protection pour garantir 

une gestion des risques résiduels pour les crues supérieures à la crue de dimensionnement 

(probabilité de blocage aux ponts, débordements favorisés dans des zone précises afin d’en 

protéger d’autres, résistance des digues à la submersion). 

Tableau 2 : Eléments composant la revanche selon les recommandations CIPC (CIPC, 2013) – Application à la Vièze. 

CONCLUSIONS 

Les nouvelles recommandations CIPC pour le calcul de la revanche conduisent, dans le cas 

de la Vièze, à un surdimensionnement des profils en travers. Tout en conservant l’esprit des 

recommandations, il est proposé de remplacer la revanche partielle correspondant à la 

hauteur d’énergie par la différence des hauteurs conjuguées d’un ressaut associé au nombre 

de Froude de l’écoulement. Les résultats issus de la littérature montrent que les ondulations 


Figure 4 : Revanche nécessaire calculée sur le cours d’eau de la Vièze et revanche minimale imposée au projet pour la crue de 

dimensionnement. L’origine du kilométrage se trouve à la confluence avec le Rhône. 

de surface, à considérer pour la revanche, atteignent au maximum la valeur proposée et que 

la hauteur d’énergie surestime largement le phénomène. 

Selon cette méthodologie, une revanche variant entre 0.3 et 1.1 m donne un ratio raisonnable 

avec la profondeur d’écoulement de la Vièze. Elle permet un rehaussement des berges 

intégrable dans le paysage urbain de la ville de Monthey. 


REFERENCES 

- Castro-Orgaz O. and Hager W. H. (2011). Turbulent near-critical open channel flow: 

Serre's similarity theory. Journal of Hydraulic Engineering ASCE, May 2011, p. 497-503. 

- CIPC (2013).La revanche dans les projets de protection contre les crues et de l’analyse de 

dangers. Recommandations de la Commission pour la protection contre les crues. Revue 

«Eau énergie air» – Cahier 2, Baden, Suisse. 

- Chanson H. (1996). Free-surface flow with near-critical flow conditions. Canadian Journal 

of Civil Engineering, Vol. 23, No 6, p. 1272-7284. 

- GILAT – ETEC (2002). Concept de protection contre les crues et concept de renaturation 

des cours d’eau sur la commune de Monthey – Rapport technique. 

- Keulegan G. H. and Patterson G. W. (1940). Mathematical theory of irrotational translation 

waves. Journal of Research of the National Bureau of Standards, RP 1273, US Department 

of Commerce, 24 (1), p. 47-101. 

- Mandrup-Anderson V. (1978). Undular hydraulic jump. Journal of Hydraulics Divison, 

104 (8), P. 1185-1188. 

- Sinniger R. O. et Hager W. H. (1989). Constructions hydrauliques. Ecoulements stationnaires. 

Traité de Génie Civil, Vol. 15, PPUR, Lausanne, Suisse. 



Planning of risk-based rockfall mitigation measures 

using 3D rockfall simulations 

Risiko-basierte Massnahmenplanung gegen 

Steinschlag mittels 3D Steinschlagsimulationen 

Thomas Bickel, Dipl.- Ing. 1 ; Markus Hodel, Dipl. Geograph, Akademischer Geoinformatiker 1 

ABSTRACT 

In Switzerland, in the past decades, natural hazard events caused repeatedly substantial 

damages to society. But limited financial resources necessitate risk-based mitigation strategies 

in order to fulfil the economic criteria of positive benefit-cost ratio. 

Meanwhile, official software tools provided by the federal authority are available for risk 

assessment (e.g. EconoMe, RoadRisk). Since these online-tools are initially developed for the 

evaluation of project prioritization and subsidization, they use standardized routines to 

sustain comparability. 

This paper presents a new approach to extend the established methodology of risk management 

that, however, allows a more realistic risk analysis for cost-effective rockfall mitigation 

measures on roads. After subdividing the investigated road section into appropriate subsections, 

the risk analysis is carried out for each subsection individually based on three-dimensional 

rockfall simulations. 

The implementation and results of this approach and its potential for risk analysis is illustrated 

on a virtual case study and compared with EconoMe 3.0. 

Finally, the feasibility, challenges and benefits of the extended approach are discussed with 

respect to mitigation project planning. 


In den vergangenen Jahrzehnten war die Schweiz wiederholt von Naturgefahrenereignissen 

betroffen, die erhebliche Schäden verursachten. Um die potentiell gefährdeten Personen- und 

Sachwerte nachhaltig zu schützen, müssen die knappen finanziellen Ressourcen priorisiert 

und in nachweislich Nutzen-Kosten-effiziente, risikoreduzierende Schutzmassnahmenkonzepte 

investiert werden. 

Zur Beurteilung der Projektpriorisierung und Subventionierung durch die öffentliche Hand, 

wurden vom Bund Berechnungsprogramme entwickelt (bspw. EconoMe, RoadRisk), die auf 

dem Ansatz der Nutzen-Kosten-Analyse basieren und ein wirtschaftliches Optimum 

1 Louis Ingenieurgeologie GmbH, Weggis, SWITZERLAND, thomas.bickel@louis-weggis.ch 



anstreben. Um die Vergleichbarkeit von Massnahmenprojekten zu gewährleisten, verwenden 

die Programme standardisierte und verallgemeinernde Routinen. 

Basierend auf der etablierten Risikostrategie präsentiert dieser Artikel eine Methodenerweiterung, 

die eine realitätsnahe Risikoanalyse für die Nutzen-Kosten-optimierte Massnahmenplanung 

gegen Steinschlag auf Linienobjekten vorschlägt. Dabei wird das zu untersuchende 

Linienobjekt in zweckmässige Abschnitte unterteilt und die Risikoanalyse unter Berücksichtigung 

der Möglichkeiten von 3D-Steinschlagsimulationen separat für jeden Teilabschnitt 

durchgeführt. 

Die Umsetzung und die Resultate dieses Ansatzes sowie das Potential für die Risikoanalyse 

werden an einem virtuellen Fallbeispiel vorgestellt und mit EconoMe 3.0 verglichen. 

Schlussendlich werden die Praxistauglichkeit und Vorteile mit Bezug auf die Schutzmassnahmenplanung, 

sowie zukünftige Herausforderungen diskutiert. 

KEYWORDS 

risk assessment; Rockfall; EconoMe; Risk Mitigation; Measures 

AUSGANGSLAGE UND ZIELSETZUNG 

Die schweizerische Kommission «Nationale Plattform Naturgefahren» (PLANAT) definiert 

Naturgefahren als hydrologische, meteorologische, geologische oder biologische Prozesse, die 

schädlich auf den Menschen sowie auf Sachwerte einwirken können. Somit rücken die von 

der Gesellschaft als Siedlungs- oder Industriegebiete genutzten oder mit Infrastrukturanlagen 

bebauten Räume in den Fokus der Naturgefahrenbearbeitung. 

Neben der ausschliesslichen Gefahrenbetrachtung (bspw. in Gefahrenkarten) hat unter dem 

Eindruck der Unwetterereignisse seit Ende der 1980er Jahre, die Analyse der aus Naturgefahren 

resultierenden Risiken an Bedeutung gewonnen (vgl. Bründl, 2015; Bründl et al., 2009). 

So wird das Risiko, definiert durch das Produkt der Faktoren «Schadenwahrscheinlichkeit» 

und «Schadenausmass», als Mass für die Gefährdung herangezogen und für die Abklärung 

eines Handlungsbedarfes eingesetzt. 

Für die Planung von Schutzmassnahmen gegen Naturgefahren wurden nach 2005 unter 

Federführung des Bundes Berechnungsprogramme zur Bewertung der Massnahmeneffektivität 

und Wirtschaftlichkeit auf der Basis der Risikooptimierung entwickelt. Die heute gebräuchlichen 

Programme «EconoMe» und «RoadRisk» weisen aber zugunsten der Projektvergleichbarkeit 

einen hohen Standardisierungsgrad auf, da sie als Entscheidungsgrundlage 

für die Projektpriorisierung und Mittelzuteilung durch den Bund gedacht sind (BAFU, 2015; 

ASTRA, 2015). Dem gestiegenen Anspruch in der Massnahmenplanung an eine realitätsnahe 

und robuste Ermittlung sowie Abbildung der Risikoverteilung im Schadenperimeter kann die 

einfache Standardanwendung der genannten Programme nicht genügen. 


Der hier vorgestellte Ansatz basiert auf dem Risikokonzept gemäss Strategie Naturgefahren 

Schweiz der PLANAT (Bründl, 2009) und zielt auf die detailliertere Analyse der Sturzrisiken 

auf beliebigen Teilabschnitten (TA) entlang linienförmiger Schadenobjekte (z.B. Verkehrsachsen). 

Die Auswirkungen dieser Methodenerweiterung auf die modellierte Risikoverteilung 

werden anhand eines fiktiven Beispiels vorgestellt und mit „EconoMe 3.0“ verglichen. Auf 

den Vergleich mit „RoadRisk“ wird dagegen verzichtet, da dieses Programm auch betriebswirtschaftliche 

Aspekte einbezieht. 

METHODENERWEITERUNG UNTER BERÜCKSICHTIGUNG VON 3D-STURZMODELLIERUNGEN 

Die Umsetzung des Risikokonzepts Naturgefahren in „EconoMe“ (Bründl et al., 2015) bildet 

die Ausgangslage für die hier vorgestellte punktuelle Methodenerweiterung. Das grundsätzliche 

methodische Vorgehen bleibt dabei unbeeinflusst. 

Grundgedanke und Motivation für die Erweiterung der bestehenden Methodik ist die 

stärkere Gewichtung der 3D-Sturzmodellierungen zur detaillierten Beschreibung potentieller 

Sturzprozesse in der Wirkungsanalyse (Analyse von Art, Ausdehnung und Intensität einer 

Gefährdung durch die zugrundeliegenden Szenarien). Gleichzeitig wird die Risikoanalyse 

um einen Faktor N(Fr) zur Berücksichtigung einer möglichen Fragmentierung des Ausbruchvolumens 

erweitert (in Anlehnung an den in „RoadRisk“ verwendeten „Ereignistyp ET“). 

Die Anzahl fragmentierter Sturzkörper N(Fr) ist abhängig vom Verhältnis der Volumina der 

Ausbruchs- und Wirkungsszenarien und ist im Rahmen der Feldaufnahmen gutachterlich 

zu bestimmen. 

Nach der Wirkungsanalyse ist in der Expositionsanalyse (Analyse von Lage, Anzahl, Art, 

Präsenz und Wert potentiell gefährdeter Objekte im Beurteilungsperimeter) eine Unterteilung 

des zu beurteilenden Strassenperimeters in eine Anzahl gleich langer Teilabschnitte N(TA) 

notwendig, wobei deren Länge zweckmässig zu wählen ist. Die Sturzintensitäten auf Höhe 

der Verkehrsachse werden nicht über den ganzen Projektperimeter gemittelt, sondern über 

die einzelnen Teilabschnitte. Folglich müssen auch die mit der Sturzintensität verknüpften 

Parameter der Konsequenzenanalyse (Analyse des Schadenausmasses durch Verschnitt der 

Erkenntnisse aus Wirkungs- und Expositionsanalyse) pro Teilabschnitt abgeleitet werden. 

Gemäss dem Leitfaden zum Risikokonzept kann die räumliche Auftretenswahrscheinlichkeit 

p(rA) verschieden abgeschätzt werden (vgl. Bründl, 2009, Kap. 3.4.1.3): 

– 1) Es kann für jede Prozessart und jedes Szenario ein Faktor bestimmt werden, der festlegt, 

welcher Flächenanteil im Mittel bestrichen wird. Diese Vorgehensweise ist in den bestehenden 

Berechnungsprogrammen umgesetzt, wobei dem Anwender Richtwerte zur standardisierten 

Verwendung vorgeschlagen werden. 

– 2) Die p(rA) kann mit Hilfe eines Ereignisbaumes abgeschätzt werden. Diese Variante wird 

im vorgestellten Ansatz verfolgt, wobei die drei verwendeten Kriterien A bis C in Abbildung 

1 dargestellt sind. 


Kriterium A im Ereignisbaum ist die gutachterlich festzulegende Anzahl an Sturzkörpern je 

Ereignis N(Fr). Das Kriterium B stellt die Erreichenswahrscheinlichkeit der Teilabschnitte p(E) 

TA (Wahrscheinlichkeit, dass ein betrachteter Teilabschnitt von einem Sturzereignis erreicht 

wird) dar. Deren Höhe wird aus 3D-Sturzmodellierungen abgeleitet und errechnet sich zu 

p(E)TA = n(Sk)TA/N(Sk), wobei n(Sk)TA die Anzahl Sturzkörper, welche den jeweiligen 

Teilabschnitt erreichen, und N(Sk) die Summe aller modellierten Sturzkörper der betrachteten 

Prozessquelle ist. Das Kriterium C des Ereignisbaumes berücksichtigt den effektiv vom 

Ereignis bestrichenen Anteil des Teilabschnittes und wird durch die räumliche Auftretenswahrscheinlichkeit 

innerhalb eines einzelnen Teilabschnittes p(rA)TA = d1/lTA beschrieben, 

wobei d1 die grösste Kantenlänge des fragmentierten Sturzkörpers und lTA die Länge eines 

Teilabschnittes ist. 

Abbildung 1: Ereignisbaum zur Herleitung der räumlichen Auftretenswahrscheinlichkeit p(rA)*TA für Sturzprozesse im Teilabschnitt TA 

Unter Berücksichtigung der oben aufgeführten Kriterien errechnet sich die erweiterte räumliche 

Auftretenswahrscheinlichkeit für Sturzprozesse pro Teilabschnitt p(rA)*TA folglich zu: 

Formel 1: Formel zur Abschätzung der erweiterten räumlichen Auftretenswahrscheinlichkeit p(rA)*TA für Sturzprozesse im 

Teilabschnitt TA. 

Formel 2: Formel zur Abschätzung der erweiterten räumlichen Auftretenswahrscheinlichkeit p(rA)*TA für Sturzprozesse im 

Teilabschnitt TA unter Berücksichtigung der Teilfaktoren. 

Unter Einbezug der Teilfaktoren ergibt sich: 

Mit der Verwendung von p(rA)*TA können Risiken in Anlehnung an die bestehende 

Methodik differenziert nach Teilabschnitten errechnet werden. Das wahrscheinliche 


Schadenausmass für Personen auf Strassen Aw(PS)TA (vgl. Bründl, 2009, Formel 3.14) 

errechnet sich neu pro Teilabschnitt und Szenario j zu: 

Formel 3: Formel zur Abschätzung des wahrscheinlichen Schadenausmasses für Personen auf Strassen und mechanischen 

Aufstiegshilfen im Teilabschnitt TA. 

wobei (S)TA die gemittelte Letalität von Personen im Teilabschnitt in Abhängigkeit des 

Prozesses und der Intensität, β der durchschnittliche Besetzungsgrad der Strassenfahrzeuge 

und DTV der durchschnittliche tägliche Verkehr sind. Der Faktor gj entspricht der Länge des 

Teilabschnittes lTA und v der durchschnittlichen Geschwindigkeit von Fahrzeugen. 

Der beschriebene Ansatz kann sinngemäss in allen Arbeitsschritten der Risikoberechnung 

(inkl. Berechnung der Sachrisiken) entsprechend umgesetzt werden. 

RANDBEDINGUNGEN DES FALLBEISPIELS 

Als Fallbeispiel wurde ein 750 m langer Strassenabschnitt gewählt, und in 30 Teilabschnitte 

von 25 Metern Länge unterteilt (vgl. Abbildung 2). Die Sturzgefährdung der Strasse sowie der 

Verkehrsteilnehmer resultiert aus drei Prozessquellen. PQ1 grenzt unmittelbar bergseits an die 

Strasse an, wohingegen PQ2 und PQ3 durch einen weitläufigen Transitbereich von der Strasse 

getrennt sind. Im Fallbeispiel wurden die Blockgrössen gutachterlich und zur Vereinfachung 

ohne Berücksichtigung einer Fragmentierung (d.h. N(Fr) = 1) im Ausbruchsbereich bestimmt 

(Tabelle 1) und bilden die Grundlage für die 3D-Sturzmodellierungen mit Rockyfor3D (V5.1). 

Betrachtet werden lediglich die Risiken infolge seltener Ereignisse (WP 30–100 Jahre). 

Tabelle 1: Prozessquellen PQ1–3. Die aufgelisteten Werte beziehen sich auf das seltene Ereignis (WP 30–100 Jahre). d1 = grösste 

Kantenlänge des fragmentierten Sturzkörpers; L = Länge des potentiell gefährdeten Strassenabschnittes; p(rA) = räumliche 

Auftretenswahrscheinlichkeit. 

RESULTATE UND DISKUSSION 

In Abbildung 2a sind die drei Prozessquellen, die modellierten Sturzbahnen, der betrachtete 

Strassenabschnitt sowie dessen potentiell gefährdete Bereiche inklusive der einwirkenden 

Sturzintensitäten abgebildet. Es zeigt sich die Abhängigkeit der errechneten Sturzverläufe von 

der jeweiligen Hangmorphologie im Transitbereich (besonders von Mulden, Runsen, Kuppen, 


Abbildung 2: Situation im Beispielperimeter für die seltenen Sturzereignisse (WP 30–100 Jahre) inkl. Lage der Prozessquellen, der 

Sturzbahnen, qualitativer Darstellung der errechneten Risiken je Teilabschnitt und der relativen Prozessquellenanteile je Teilabschnitt. 


etc.). Erwartungsgemäss werden potentielle Sturzkörper in Runsen kanalisiert und es ergeben 

sich dort die Blockdurchgangsmaxima, wo Kanalisierungs- und Überlagerungseffekte verschiedener 

Prozessquellen zusammen kommen. Des Weiteren ist ersichtlich, dass in allen 

potentiell gefährdeten Teilabschnitten mit starken Sturzintensitäten (Ekin > 300 Kilojoule) 

infolge seltener Sturzereignisse gerechnet werden muss. 

Nicht dargestellt sind die prozentualen Anteile aller Sturzkörper, welche die Strasse gemäss 

Modellierung erreichen. Diese wurden für die Prozessquellen PQ1, PQ2 bzw. PQ3 zu 66%, 

44% bzw. 20% errechnet, wobei gesamthaft 37% aller Sturzkörper bis auf die Strasse 

stürzen. 

Abbildung 2b stellt die einzelnen 25 m-Abschnitte der Strasse inkl. der mittels erweiterter 

Methodik errechneten kollektiven Risiken (Personen und Sachrisiken) dar. In den Kreisdiagrammen 

sind sowohl die Summen je Teilabschnitt als auch die relativen Prozessquellenanteile 

je Teilabschnitt dargestellt. In den Teilabschnitten mit hohen Blockdurchgängen werden 

die grössten Risiken errechnet. Sturzkörper aus der Prozessquelle PQ1 wirken flächig auf eine 

Vielzahl von Teilabschnitten. Demgegenüber wirken die Ereignisse aus den Prozessquellen 

PQ2 und PQ3 nur auf wenige Teilabschnitte, tragen dort jedoch massgeblich zum hohen 

Risiko bei und eine Überlagerung der Wirkungsbereiche führt zu deutlichen Risikospitzen. 

Die Teilabschnitte TA01–TA03, TA26, TA27 und TA30 sind nicht gefährdet. Die einzelnen 

Prozessquellen tragen 61%, 35% bzw. 4% zum gesamten kollektiven Risiko im Perimeter bei. 

Die errechneten Risiken der erweiterten Methodik (Variante 3) werden den Resultate des 

Standardvorgehens nach EconoMe3.0 gegenübergestellt, wobei für dieses zwei Varianten 

unter Verwendung der standardisierten p(rA)-Richtwerte (Variante 1) bzw. mit angepassten 

p(rA)-Werten (Variante 2) betrachtet werden (vgl. Tabelle 1). 

Die Resultate sind in Abbildung 3 zusammenfassend dargestellt. Die errechneten Risiken der 

Varianten unterscheiden sich sowohl ihrem Betrag nach als auch bzgl. ihrer Verteilung auf 

die Teilabschnitte deutlich, wobei die Variante 3 in der Summe die niedrigsten Werte liefert. 

Hinzu kommt, dass nur die Variante 3 eine realitätsnahe Verteilung auf die Teilabschnitte 

errechnen kann. In den Varianten 1 und 2 werden die Risiken als Summe über den gesamten 

Strassenabschnitt errechnet und auf die potentiell gefährdeten Teilabschnitte verteilt. 

Die Variante 1 errechnet für die PQ1 im Ereignisfall einen bestrichenen Strassenabschnitt 

von 17 m (=560 m*0.03), was der massgebenden Kantenlänge von 2 m widerspricht. 

Eine standardmässige Verwendung der Richtwerte führt, insbesondere bei langen bestrichenen 

Strassenabschnitten, zu nicht nachvollziehbar hohen Risikobeiträgen. Dem trägt die 

Variante 2 Rechnung, indem der p(rA)-Wert auf Basis der effektiven Kantenlängen hergeleitet 

wird. In der Folge reduzieren sich die errechneten Risiken um den Faktor 4 (Abbildung 3). 

Die Variante 3 berücksichtigt mittels der Modellierungen als einzige die morphologischen 

Charakteristika des Transitbereiches umfassend. Dies führt zu einer differenzierten Risikoverteilung 

innerhalb des gesamten betrachteten Strassenabschnittes. Ausserdem wird der Anteil 


der oberhalb der Strasse abgelagerten Blöcke quantitativ abgeschätzt. Es fällt auf, dass die 

Risikodifferenz zwischen Varianten 2 und 3 (57%) in etwa dem in den Modellierungen 

errechneten Anteil an abgelagerten Sturzkörpern (63%) entspricht. 

Abbildung 3: Kollektives Risiko in den Teilabschnitten infolge seltener Sturzereignisse (WP 30–100 Jahre) abhängig von der 

Berechnungsmethodik bzw. den jeweiligen Eingabeparametern. 


Die hier vorgestellte Methodenerweiterung eignet sich besonders zur Analyse einer Sturzgefährdung 

auf ein linienförmiges Schadenobjekt. Die bestehende Methodik nach EconoMe zur 

Abschätzung und Darstellung von Risiken kann im Grundsatz beibehalten werden. Vorsicht 

ist bei der standardisierten Verwendung von p(rA)-Richtwerten geboten (Variante 1), deren 

Plausibilität in jedem Fall zu prüfen ist. 

Über den gesamten Perimeter liefert die Risikoabschätzung mit angepassten p(rA)-Werten 

(Variante 2) mit nur geringem Zusatzaufwand plausible Risikowerte. Hinsichtlich der Massnahmenplanung 

auf dieser Grundlage ist eine gutachterliche Priorisierung der zu schützenden 

Teilabschnitte notwendig. Die erweiterte Methodik (Variante 3) liefert teilabschnittsscharfe 

Risikowerte und nutzt unter Umständen bereits vorhandene Daten aus 3D-Modellierungen. 

Die Resultate liefern eine gute Datengrundlage zur Priorisierung und Optimierung 

allfälliger Schutzmassnahmen innerhalb eines Projektperimeters nach den Kriterien vorgängig 

definierter Projektziele. Unter Voraussetzung robuster Grundlagen können mit dem 

vorgestellten Ansatz Kanalisierungs-, Überlagerungs- und Ablagerungseffekte ortsspezifischer 

Eigenschaften des Geländes quantitativ erfasst und daraus resultierende Risikospitzen im 

Hinblick auf eine risikobasierte Massnahmenplanung aufgezeigt werden. Dies führt idealer- 


weise zu einer verbesserten Nutzen-Kosten-Effizienz, wodurch der entstandene Mehraufwand 

während der Projektbearbeitung gerechtfertigt werden kann. 

Unsicherheiten der notwendigen Grundlagendaten (z.B. bei der Beurteilung der relevanten 

Naturgefahrenszenarien) reduzieren die Qualität der Resultate und können durch eine 

gewissenhafte Ausführung / Dokumentation sowie durch eine Plausibilisierung der Feldaufnahmen 

minimiert werden. Dies betrifft insbesondere die Festlegung von Anzahl und Grösse 

der Prozessquellen in Kombination mit den jeweiligen Szenarien, wodurch der Projektperimeter 

hinsichtlich der Steinschlagaktivität beschrieben wird. 

DANKSAGUNG 

Vielen Dank an Klaus Louis, der die Ausarbeitung des vorliegenden Artikels ermöglichte 

sowie an Conradin Zahno für seine Unterstützung. Ausserdem an Patrizia Köpfli und Raphael 

Zurfluh für ihre fachlichen Beiträge und an Michael Bründl für die aufschlussreiche Diskussion 

zum bestehenden Risikokonzept. 

LITERATUR 

- Bundesamt für Umwelt BAFU, 2015. Wirksamkeit und Wirtschaftlichkeit von Schutzmassnahmen 

gegen Naturgefahren. Online Tool. www. econome.admin.ch. 

- Bundesamt für Strassen ASTRA (Ed.), 2012. Naturgefahren auf den Nationalstrassen: 

Risikokonzept. Version V2.10 vom 19.12.2013. Dokumentation. 95 S. 

- Bundesamt für Strassen ASTRA, 2015. Wirksamkeit und Wirtschaftlichkeit von Schutzmassnahmen 

gegen Naturgefahren. Online Tool. www.roadrisk.admin.ch. 

- Bründl, M. (Ed.), 2009. Projekt A1.1 Risikokonzept für Naturgefahren: Leitfaden. 420 S. 

- Bründl, M., 2015. Umsetzung des Risikokonzepts in die Praxis. FAN Agenda 2015(2): 

15–16. 

- Bründl, M., Romang, H.E., Bischof, N., Rheinberger, C.M., 2009. The risk concept and its 

application in natural hazard risk management in Switzerland. Natural Hazards and Earth 

System Sciences 9: 801–813. 

- Bründl, M., Ettlin, L., Burkard, A., Oggier, N., Dolf, F. und Gutwein, P., 2015. EconoMe – 

Wirksamkeit und Wirtschaftlichkeit von Schutzmassnahmen gegen Naturgefahren: Formelsammlung. 

56 S. 



Protection structures against natural hazards: from 

failure analysis to effectiveness assessment 

Simon Carladous, PhD Student, MSc. Eng. 1 ; Jean-Marc Tacnet, Dr, MSc. Eng. 1 ; Mireille Batton-Hubert, Prof. 2 

ABSTRACT 

Protection structures aim to protect areas exposed to natural hazards. For instance, several 

clusters of check dams are located in the headwaters of a watershed, each having specific 

functions. Their structural, functional, and economic effectiveness must be assessed to assist 

decision makers in deciding on maintenance actions. Nevertheless, expert assessment 

generally focuses on a single check dam. It is based on the evaluation of several field 

indicators that are aggregated by experts. This paper aims to show how methods extracted 

from industry, such as dependability analysis and decision-making approaches, can be used 

to help formalize expert assessment with expert knowledge taken into account. For this 

purpose the formal methods are reviewed and their potential application introduced. 

This paper focuses on the use of indicators and criteria in various objective decision-making 

tools. For example, it compares the Analytic Hierarchic Process and the ELECTRE TRI 

methods to assess the functional effectiveness of a cluster of check dams. 

KEYWORDS 

check dams; effectiveness; functional analysis; Multi-Criteria Decision Methods 

INTRODUCTION 

In mountainous areas, torrential floods are sudden and destructive. Protection systems have 

specific functions to protect exposed elements from them. For instance, check dams control 

material volume and flow through stabilization of bed scouring and banks. More than 21,000 

old protection works, including 14,000 check dams, are registered in French public forests 

(Piton et al. 2015). Assessing their actual effectiveness is a key issue in maintenance decisionmaking. 

The effectiveness of a system may be defined as the level of objective achievement and takes 

into account three features: structural, functional, and economic. Its assessment is based on 

its capacity in relation to the objectives that were set. For instance, a check dam structure has 

to be stable under a debris-flow with a given intensity discharge. Nevertheless, while nominal 

capacity may be defined through functional and structural design, the real capacity of a 

structure may be reduced, depending on its condition. When analyzing structural pathology, 

field practitioners focus on structural and functional features. Degradation criteria that can 

1 Irstea - ETGR, Univ. Grenoble Alpes, Ecole Nationale des Mines de Saint-Etienne, AgroParisTech Saint-Martin-d'Hères, FRANCE, 

simon.carladous@irstea.fr 

2 Ecole des Mines de Saint-Etienne 

P_2016_FP078 


affect structural stability and the functional service of each structure or cluster of structures, 

are identified and qualitatively assessed. Following this, structural degradation, overall 

condition and emergency actions required on each structure are established through predefined 

qualitative classes. Pathology analysis is used by experts to propose adapted actions in 

order to maintain a structure in good working condition (Suda 2013). 

Assessing effectiveness is a multicriteria, multi-feature (structural, functional, and economic) 

and multiscale (check dam, device, and watershed) analysis. In practice, practitioners 

qualitatively assess the effectiveness of individual structures; there is no integrative method 

for this assessment. 

Current developments based on decision-aid methods (DAMs) aim to take into account 

expert knowledge in a formal integrative framework. This paper initially introduces the main 

principles from functional analysis to multi-criteria decision methods (MCDMs). We then 

assess the effectiveness of three clusters of check dams, as regards their given objectives and 

criteria, using two MCDMs. Finally, we review the main steps in assessing the effectiveness 

of a protection system and discuss remaining gaps. 

Figure 1: Multiscale and multicriteria aggregation to assess effectiveness of a cluster of check dams. 


METHODS 

This paper aims to show how DAMs can help to assess the effectiveness of protection systems. 

Accordingly, we do not go into great detail but provide the main principles. 

First, we use the safety and reliability analysis framework (Tacnet et al. 2012). This helps to 

specify the system components, their objective functions, the failure mechanisms, the criteria 

gj and the indicators Im that explain them (Fig. 1). Each Im assessment is related to the 

achievement of a function and is specified through a formal evaluation scale (Curt et al. 

2010). A list of Im indicators is an example of the results of a check dam pathology analysis 

(Suda 2013). In this paper we propose to use examples of these Im to directly show how they 

can be aggregated. 

Whatever the feature and system scale (check dam, a cluster of check dams), assessing the 

level of effectiveness is a decision-making problem (DMP). From a range of alternatives Ai, 

it may be necessary to choose the most effective one, sort them into predefined effectiveness 

classes, or rank them from the most to the least effective (Carladous et al. 2015; Roy 1985). 

As for any other DMP, the evaluation is based on the aggregation of the evaluations of several 

gj, assessed through Im (Fig. 1). This process is generally called the “expert judgement” and 

in our example remains focused on a single check dam as Ai. Nevertheless, several DAMs 

exist to perform this type of task. 

Figure 2: Adaptation of rule-based systems to the assessment of check dam effectiveness. a: An example of the structure of a 

hierarchical model used to assess the level of structural performance of a hydraulic dam, detailing the internal erosion failure mode 

(Curt 2010). b: Potential structure of a similar model to assess the level of performance of a torrential check dam. 


Rule-based systems are one such DAM. They are used to assess the structural performance 

level of flood dikes or hydraulic dams. Several failure mechanisms, MRj,, can affect them, e.g. 

internal erosion. Failure depends on the achievement of technical functions, Fk, such as 

drainage. The notations μMRj and μFk are the respective performance levels for MRj and Fk 

(Fig. 2a). Each μFk is assessed through several indirect (e.g. slab cracking) or direct (e.g. clear 

water leakage) indicators Im.(Fig. 2a). For hydraulic dams, a common discrete scale with a 

decreasing preference from 0 (excellent) to 10 (unacceptable) has been defined for all values 

of Im. In order to assess μFk, μMRj and the global performance level, several evaluations are 

aggregated according to rules such as “MAX” or “IF-ELSE” (Curt et al. 2010). For example, an 

expert has to assess the structural performance level of two dams, A1 and A2. For A1 he 

evaluates all values of Im: I2, I3, I5, I6, I8, I12, I13 and I14 = 0; I1 = 5; I7, I9, I10 and I11 = 7; 

I4 = 10. For A2 he produces another set of evaluations: I1, I4, I5, I8, I11, I12 and I13 = 0; I6, 

I9 and I10 = 2; I14 = 3; I2 = 5; I3 and I7 = 7. He obtains μMR4(A1) = 7 and μMR4(A2) = 2 

(Fig. 2a). The dam A2, which is in good condition, is more structurally sound regarding 

internal erosion than dam A1, which is in bad condition. Even if rule-based systems are easily 

understood by practitioners, establishing such a system needs expert elicitation and validation, 

which can be very time consuming. 

MCDMs can also help aggregate several gj criteria. Total aggregating methods are single 

synthesizing criterion approaches, based on the principle of preference transitivity. Outranking 

methods do not aim to give a single synthetizing criterion to help in decision making 

(Schärlig 1985). 

The Analytical Hierarchic Process (AHP) is an aggregation-based method (Saaty 1980; Tacnet 

2009). It consists of evaluating possible Ai according to preferences (represented by weights 

ula arlaou1rraoFormula1o 

ωj) expressed by the decision makers (DMs) on the different gj. The preference . The problem 

is first broken down from an overall goal to criteria and sub-criteria. At the lowest sub-criteria 

level (e.g. gj), alternatives Ai are compared in pairs providing their . Finally, for each Ai, the 

synthesizing criterion xi is given by Formula 1. 

xx ii = ∑ ωω jj . ωω iijj 

jj 

Formula 1: Calculation of the synthesizing criterion xi for the AHP method. 

ELECTRE TRI is a progression of the outranking ELECTRE methods introduced in the 1960s 

(Yu 1992). It aims to sort a number of Ai into predefined categories Ch, according to several 

gj. The lower and upper limits, respectively bh-1 and bh, of each Ch, have to be specified 

previously by the DMs, through corresponding evaluations for each gj. A fuzzy preference 

scale is defined through three thresholds: indifference (qj), strict preference (pj) and veto (vj). 

Following this, the set of evaluations xij is used to define the outranking relation of each Ai 

with limits bh. It is based initially on the calculation of partial concordance and credibility 

indices. Global concordance and credibility indices are then derived based on an arbitrary 

-cut strategy ( in [0,1]). The final binary assignment of each Ai to a given category Ch, is 

based on an arbitrary selected attitude choice (optimistic or pessimistic). 


Figure 3: Three clusters of check dams (cd) considered as alternatives and the criteria taken into account to formalize the DMP example. 

RESULTS 

In this part, we compare the AHP method with the ELECTRE TRI method on the same DMP. 

THE ACTUAL DMP 

The functional effectiveness of three clusters of check dams is assessed: A1 is a cluster of 

28 check dams, A2 contains 25 check dams and A3 contains 3 check dams (Fig. 3). Their 

purpose is to stabilize the longitudinal profile and limit lateral erosions. Debris flows are 

considered as the scenario of reference. We propose seven gj (Fig. 3). For each of them, the 


DMs have to give the evaluation xij for each Ai in addition to its weight ωj (Table 1). 

The preference order of the evaluation scale has to be specified. For example, the structural 

effectiveness assessment introduced in Fig. 2 is assessed through decreasing preference; 

indicator evaluation has lower preference while structural evaluation has higher preference. 

– g1 (increasing preference): free spillway dimensions, i.e., the rates (0–100%) between the 

smallest dimensions of all check dam free spillways and the reference scenario discharge. 

– g2 (decreasing preference): the orientation of the check dams, i.e., the mean of absolute 

angle of deviance (0–90 degrees) for all check dams. For each check dam, the angle of 

deviance is the absolute angle between the actual and the optimal orientations. 

– g3 (decreasing preference): the longitudinal implantation of check dams, i.e., the mean 

difference (in meters) between the optimal and actual implantation of each check dam. 

The difference can be technically assessed as the height difference between the calculated 

elevation of the next upstream check dam abscissa and its actual elevation for the lowest 

compensation slope. 

– g4 and g5 (decreasing preference): the mean structural effectiveness of the most significant 

check dams (g4) and of the others (g5). For each of them, structural effectiveness can be 

defined as the resistance to stresses due to the reference scenario. We propose an integer 

rating from 0 (stable) to 10 (unstable). In general, downstream check dams are the most 

significant. 

– g6 (decreasing preference): active longitudinal erosion, i.e., the rate (%) between the 

length of longitudinal erosion and the objective length of longitudinal stabilization. 

– g7 (decreasing preference): active lateral erosion, i.e., the rate (%) between the volume of 

active lateral erosion and the objective volume of lateral volume stabilization. 

Table 1: Data needed to apply the AHP and the ELECTRE TRI to the DMP. 

in general for AHP for ELECTRE TRI 

g j ω j x 1j x 2j x 3j ω 1j ω 2j ω 3j b 0 b 1 b 2 b 3 b 4 q j p j v j 

g 1 0.1 80 100 100 0.14 0.43 0.43 0 25 50 90 100 10 20 40 

g 2 0.2 0 -10 -20 0.54 0.30 0.16 -90 -25 -10 -5 0 5 10 30 

g 3 0.1 -1 

- 

0.5 

0 0.11 0.26 0.63 -10 -3 -1.5 0 10 0.5 1 2 

g 4 0.15 9 1 7 0.08 0.79 0.13 10 8 5 1 0 1 3 5 

g 5 

0.0 

5 

7 3 9 0.17 0.74 0.09 10 8 5 1 0 1 3 5 

g 6 0.2 -30 0 0 0.10 0.45 0.45 -100 -80 -50 -10 0 10 20 40 

g 7 0.2 -10 -20 -50 0.65 0.29 0.06 -100 -80 -50 -10 0 10 20 40 

1 


PROPOSITION OF A RULE-BASED SYSTEM 

For the evaluation criteria g4 and g5, the structural effectiveness of each check dam has to 

be assessed. A rule-based system based on expert knowledge is proposed, in order to allow 

assessment according to the foundation scouring failure mechanism (Fig. 2b). Foundations 

may be endangered by a regressive scouring (2), which depends on the geology (I1) and 

hydraulic conditions (I2, I3). Check dams with anti-scouring components (I4) are protected, 

but scouring can occur under foundations (I7) due to design (I5) or condition (I6) problems 

(μF1). If it occurs, it can affect the external stability of check dams (μF2). Its general 

movement (I8) is a direct indicator. Even if the unsupported foundation area is significant 

(I9), check dams with an external anchorage (I10) or with beam-support (I11) are externally 

stable. Specific visible cracks are a direct indicator of stability failure (I13). If no cracks are 

visible, structures are more or less vulnerable to scouring, e.g., gravity check dams in 

dry-stone are more vulnerable than gravity check dams in masonry, which in turn are more 

vulnerable than self-supporting check dams in concrete (I12). 

For the example under consideration, all check dams of A1 are in masonry without any 

scouring protection. A regressive scouring affects those downstream (Fig. 3). The structural 

performance level is consequently bad. Evaluation of g4 (for the most important check dams) 

is worse than for g5 (x14 = 9 and x15 = 7). By making assessments for each Ai, we obtain 

evaluations for g4 and g5 (Table 1). 

RESULTS FOR THE AHP METHOD 

Comparing alternatives in pairs provides ωij for each gj, e.g. g4, according to x14, x24,and 

x34 (Table 1), one prefers A2 to A1 with a level of 9, one prefers A3 to A1 with a level of 5, 

and one prefers A2 to A3 with a level of 7. It gives ω14 = 0.08, ω24 = 0.79, and ω34 = 0.13 

through the preferences matrix. Doing this for each gj, one fills in the Table 1. For each Ai, 

Formula 1 gives x1 = 0.3035, x2 = 0.4325, and x3 = 0.2640. It is possible to assess the 

functional effectiveness of each cluster of check dams on a continuous evaluation scale from 

0 to 1. Moreover, they can also be compared: A2 is strictly more effective than A1, which is 

equivalent/more effective than A3. 

RESULTS FOR THE ELECTRE TRI METHOD 

The DM sorts each Ai into one of the four effectiveness classes: C1 = ‘Not effective’; 

C2 = ‘Slightly effective’; C3 = ‘Moderately effective’; C4 = ‘Highly effective’. For each gj, 

the DMs must give the lower and upper limits (bh-1 and bh respectively) of each class, and 

also the thresholds pj, qj, and vj representing the fuzzy preferences on each gj evaluation 

scale (Table 1). In this example, the same thresholds are considered for each bh limit. 

Taking into account = 0.7 and the elements from Table 1, whatever the pessimistic or 

optimistic attitude of the DM, A1 is in the moderately effective class, whereas A2 is in the 

highly effective class. The A3 assignment depends on the attitude of the DM: in the pessimis- 


tic view it is in the slightly effective class, whereas in the optimistic view it is in the moderately 

effective class. 

CONCLUSIONS 

This paper shows how DAMs can help DMs assess the effectiveness of a protection system, 

such as a cluster of check dams. Several analysis steps are proposed: 1) specification of the 

DMP, 2) analysis of function, failure mode and pathology in order to extract effectiveness 

indicators (Suda 2013; Tacnet et al. 2012), 3) specification of the assessment scale and 

preference order of each indicator and criteria (Curt et al. 2010; Suda 2013), 4) elicitation 

of aggregation modes (rule-based system, criteria weights) (Curt et al. 2010; Schärlig 1985). 

This paper integrates results from a hypothetical rule-based system into MCDMs. Two 

MCDMs are compared demonstrating the data needed (matrices of preferences, thresholds 

to describe the limitations of classes, , etc.) which can actually be difficult to specify. 

This paper does not take into account imperfections in the evaluations of each indicator and 

criterion. Extended decision-support methods have recently been developed to take such 

features into account (Carladous et al. 2015; Tacnet 2009). 

This paper shows how safety and reliability analysis can be complementary to decision-support 

methods such as MCDMs, allowing expert knowledge to be taken into account in 

assessments of effectiveness. Actual derivation of expert knowledge for each DMP (systems, 

scales, effectiveness features), the combination of DAMs, imperfect information, and the 

validation of results remain challenging. 


REFERENCES 

- Carladous S. et al. (2015). Belief function theory based decision support methos: Application 

to torrent protection work effectiveness and reliability assessment. In Safety and Reliability of 

Complex Engineered Systems: 3643–3652. 

- Curt C. et al. (2010). A Knowledge Formalization and Aggregation-Based Method for the 

Assessment of Dam Performance. Computer-Aided Civil and Infrastructure Engineering, 25: 

171–184. 

- Piton G. et al. (sub.). Why do we build check dams? An historical perspective from the 

French experience. (Sub. to) Earth Surface Processes and Landforms. 

- Roy B. (1985). Méthodologie Multicritère d’Aide à la Décision, Paris, France. 

- Saaty T.L. (1980). Multicriteria Decision Making - The Analytic Hierarchy Process: Planning, 

Priority Setting, resource Allocation. 

- Schärlig A. (1985). Décider sur plusieurs critères. Panorama de l’aide à la décision multicritère, 

Presses polytechniques et universitaires romandes. 

- Suda J. (2013). Schutzbauwerke der Wildbachverbauung - Handbuch zur Durchführung 

einer Bauwerkskontrolle, Vienna, Austria. 

- Tacnet J.-M. et al. (2012). Efficiency assessment for torrent protection works: an approach 

based on safety and reliability analysis. In 12th international conference Interpraevent. 

Grenoble, France. 

- Tacnet J.-M. (2009). Prise en compte de l’incertitude dans l’expertise des risques naturels 

en montagne par analyse multicritères et fusion d’information. PhD Thesis,Ecole Nationale 

supérieure des Mines de Saint-Etienne. 

- Yu W. (1992). Aide multicitère à la décision dans le cadre de la problématique du tri: 

Concepts, méthodes et applications. Paris, France. 



Study of landslide run-out and impact on protection 

structures with the Material Point Method 

Francesca Ceccato 1 ; Paolo Simonini, Prof. 1 

ABSTRACT 

Reliable estimates of the impact forces induced by flow-like landslides on existing structures is 

of great importance for hazard assessment. This is customarily obtained by means of 

empirically based relationships but, significant differences may be encountered using the 

existing approaches. Large scale experimental studies are expensive and difficult to set up; for 

this reason a numerical technique able to simulate the material flow and its interaction with 

structures would be helpful for the hazard assessment as well as for the design of mitigation 

measures. This paper shows the applicability of the Material Point Method (MPM), a meshless 

method specifically developed to describe large deformations of bodies, to the study of 

granular flow propagation and impact forces on rigid structures. Complex shapes of the 

structure as well as different soil-structure interface properties are considered. It is shown that 

the MPM may represent a suitable tool to support the design of landslide mitigation measures. 

KEYWORDS 

MPM; granular flow; impact forces 

INTRODUCTION 

Several studies carried out in the past focused on the determination of the impulsive force 

produced by debris flows on vertical rigid structures, e.g. Armanini & Scotton (1993), Hübl et 

al. (2003), Moriguchi et al. (2009), but only a few have considered the shape of the structure 

(e.g. Shieh et al. 2008; Zanuttigh & Lamberti, 2006) or its flexibility (e.g. Canelli et al. 2012; 

Leonardi et al. 2014). Moreover, there are significant differences between the proposed 

approaches which render their practical use rather difficult. 

A numerical technique able to capture the key features of the debris flow and its interaction 

with a complex structure can contribute to assess the damage to existing structures and guide 

the design of protection measures. To this end, this paper investigates the potentiality of the 

Material Point Method (MPM) for these applications. 

The MPM has been specifically developed for large deformations of history dependent 

materials. It simulates large displacements by Lagrangian points moving through an Eulerian 

grid as shortly described in the next section. 

1 DICEA - University of Padua, ITALY, francesca.ceccato@dicea.unipd.it 



A 3D MPM code is applied in this study, thus complex geometries can be handled with ease. 

It features a specific algorithm to model soil-structure interaction and frictional sliding. 

The non-linear behaviour of soil is simulated by an elastoplastic model with Mohr-Coulomb 

failure criterion. The description of the flow propagation with a geomechanical model is quite 

unusual. Its use is customarily limited to small deformations such as the study of the landslide 

trigger mechanism, when the failure surface develops. Most of the numerical studies of flow 

propagation uses rheological models but it is important to note that it is very difficult to link 

geomechanical and rheological parameters under both small and large deformations. The 

MPM might close the gap between these approaches. 

In the following sections, it is shown that the MPM can reproduce the run out of a dry 

granular flow with good agreement with experimental results leading to the estimate of the 

impact forces on a vertical rigid wall, that are shown to be in good agreement with empirical 

observations. In addition to this, alternative shapes and surface characteristics of the structure 

are considered. 

THE MATERIAL POINT METHOD 

The MPM is a particle-based method developed since the 90’s for large deformations of history 

dependent materials (Sulsky et al. 1994). Recently, the method has been extended to coupled 

problems in order to simulate the soil-water interaction (Bandara & Soga, 2015; Jassim et al. 

2013) and unsaturated conditions (Yerro et al. 2015). The MPM has been successfully applied 

to the simulation of a number of geotechnical problems such as landslides (Andersen & 

Andersen, 2010), collapse of dams (Alonso & Zabala, 2011) and river-banks (Bandara & Soga, 

2015). 

The continuum body is discretized by a set of Lagrangian points, called material points (MP). 

They carry all the information of the continuum such as density, velocity, acceleration, stress, 

strain, material parameter as well as external loads. The MP do not represent single soil 

grains, as in Discrete Element Methods (DEM), but a portion of the continuum body. Large 

deformations are simulated by MP moving through a fix computational finite element mesh 

which covers the entire region of space into which the solid is expected to move. This grid is 

used to solve the system of equilibrium equations, but does not deform with the body like in 

Lagrangian Finite Element Method. 

The MPM code used in this study is being developed to solve 3D dynamic large deformation 

problems in geotechnical and hydromechanical engineering (Vermeer et al. 2013). The code 

features a contact formulation as presented by Bardenhagen et al. (2001) to model soil-structure 

interaction and frictional sliding. 


SIMULATION OF A GRANULAR FLOW 

The capability of the MPM to simulate the propagation of a dry granular flow is evaluated 

considering the result of a physical model test by Denlinger and Iverson (2001). The ex periment 

used a small flume with a bed surface inclined 31.4° adjoined to a horizontal runout 

plane. 

The non-linear soil behavior of loose sand is described with an isotopic elastic-perfectly plastic 

model with Mohr-Coulomb failure criterion. The input parameters are summarized in Table 

1. A static friction angle =40° and a static basal friction angle b 

=29° were measured by 

Denlinger and Iverson. However, the dynamic values of these parameters can be significantly 

lower. Sensitivity analyses showed that the displacements decrease by increasing and b, 

but the latter has a greater effect on the final run-out (Fig. 1). The best agreement with the 

experimental results is obtained with a basal friction angle of 26.6°. 

Figure 1: Run-out as function of the basal friction angle (left) and sand friction angle (right) for different time 

After material release, the flow accelerates gradually spreading and reaching the end of the 

inclined plane with a maximum velocity of 0.88 m/s, then the sand mass decelerates and 

stops. Figure 2 compares the propagation of the granular flow predicted by the MPM with 

that observed in the experiment. The maximum runout is observed after 1.50 s and the 

numerical prediction is in good agreement with the experiment. 

Table 1: Material properties of sand 

Bulk density [kg/m 3 ] ρ 1600 Dilatancy [°] ψ 0 

Initial porosity [-] n 0.4 Young’s modulus [kPa] E 1 

Friction angle [°] φ 40 Poisson’s ratio [-] ν 0.3 


Figure 2: Granular flow at t=0.32s (a), t=0.53s (b), t=0.93s (c) and t=1.50s (d): front view, top view and comparison with experimental 

results 

SIMULATION OF THE IMPACT ON A RIGID STRUCTURE 

This section discusses the impact of a granular flow on a structure in terms of forces and 

amount of material passing beyond the structure. The slope geometry and soil material 

parameters coincide with the previous section. The structure is placed at the end of the 

inclined plane and the basal friction angle is zero, thus maximizing the impact velocity. 

The discretization of the numerical model has been determined through preliminary analyses 

as a compromise between accuracy and computational cost. The maximum size of the 

elements is 8 mm along the sliding surface and 3 mm on the structure surface, 6880 MP 

are used to discretize the sand. 

Three different shapes of the structure are investigated: vertical dam, slanted dam and curved 

dam. For each shape, three values of the structure friction coefficient are considered to 

simulate various revetment materials: μ s 

=0, 0.3 and 0.6. 


Considering the free granular flow without any obstruction, at a cross section at the location 

of the structure, a maximum velocity of 1.96 m/s and a maximum flow height of 8.7 mm are 

observed, thus a Froude number Fr = v/(gh) 1 / 2 = 6.7 characterizes the flow. Fr>2 characterizes 

most of the small-scale tests, while real debris-flows usually show Fr

Figure 4: Fraction of san mass overtopping (not retained by) the wall (a), horizontal forces (b) and vertical forces (c) for different shapes 

of the dam 

Figure 5: Maximum impact pressure as function of Froude-number; comparison between MPM results and other experimental results 

collected by Hübl et al. (2009) 

reported by Hübl et al. (2009). The obtained p max 

=Fx, max 

/h (h=height of the dam) for the 

vertical wall is in good agreement with reference values for similar Froude numbers. 

The vertical and curved dam show similar values of the horizontal force, but the latter with 

a significant vertical force whereas the slanted dam shows similar values of the horizontal 

and vertical forces. The vertical component improves the stability of the dam against sliding 

and toppling. 


Increasing the structure friction coefficient the impact forces reduce. This effect is small for 

the vertical wall, but very important in case of slanted dam. In particular, F y 

decreases, thus 

reducing its stabilizing effect (Fig. 6). Because of the friction between the soil and the dam, 

the flow loses energy and the fraction of mass retained by the dam increases with μ s 

. 

Figure 6: Effect of structure friction coefficient on overtopping mass (left column) and forces (central and right colums) for different 

shapes 

CONCLUSIONS 

The MPM is a powerful tool to study landslides of the flow type and their interaction with 

structures. The method reproduces with good agreement the runout of a granular flow 

observed in a small scale experiment. 


The impact forces on structures of different shape can be calculated; moreover, the roughness 

of the surface can be included. This makes the MPM a valuable tool to support the design of 

mitigation measures. Indeed, simplified methods cannot evaluate the impact forces taking 

into account the 3D structure geometry and surface properties. 

For the considered example, it is shown that the vertical wall is the most effective shape in 

limiting soil overpassing, but the high horizontal impact forces might compromise its stability. 

In contrast, the curved and the slanted dam benefit of a stabilizing vertical component of the 

impact force. 

This study considers dry granular flows interacting with rigid structures, further developments 

of the research will include the soil-water interaction in saturated flow as well as 

flexible and permeable structures. 

REFERENCES 

- Alonso, E. E., & Zabala, F. (2011). Progressive failure of Aznalcóllar dam using the material 

point method. Géotechnique, 61(9), 795–808. 

- Andersen, S., & Andersen, L. (2010). Modelling of landslides with the material-point 

method. Computational Geosciences, 14(1), 137–147. 

- Armanini, A., & Scotton, P. (1993). On the dynamic impact of a debris flow on structures. 

In Proceedings Of The Congress-international Association For Hydraulic Research. 

- Bandara, S., & Soga, K. (2015). Coupling of soil deformation and pore fluid flow using 

material point method. Computers and Geotechnics, 63, 199–214. 

- Bardenhagen, S. G., Guilkey, J. E., Roessig, K. M., Brackbill, J. U., & Witzel, W. M. (2001). 

An Improved Contact Algorithm for the Material Point Method and Application to Stress 

Propagation in Granular Material, 2(4), 509–522. 

- Canelli, L., Ferrero, a. M., Migliazza, M., & Segalini, a. (2012). Debris flow risk mitigation 

by the means of rigid and flexible barriers - Experimental tests and impact analysis. Natural 

Hazards and Earth System Science, 12(5), 1693–1699. 

- Denlinger, R. P., & Iverson, R. M. (2001). Flow of variably fluidized granular masses across 

three-dimensional terrain: 2. Numerical predictions and experimental tests. Journal of 

Geophysical Research, 106(B1), 553. 

- Hübl, J., Holzinger, G., & Wehrmann, H. (2003). Entwicklung von Grundlagen zur 

Dimensionierung kronenoffener Bauwerke für die Geschiebebewirtschaftung in Wildbächen: 

Kleinmaßstäbliche Modellversuche zur Wirkung von Murbrechern. WLS Report 50 Band 3. 

Im Auftrag Des BMLFUW VC 7a. 

- Hübl, J., Suda, J., Proske, D., Kaitna, R., & Scheidl, C. (2009). Debris Flow Impact Estimation. 

International Symposium on Water Management and Hydraulic Engineering, 137–148. 

- Jassim, I., Stolle, D., & Vermeer, P. (2013). Two-phase dynamic analysis by material point 

method. International Journal for Numerical and Analytical Methods in Geomechanicsnumerical 

and Analytical Methods in Geomechanics, 37, 2502–2522. 


- Leonardi, A., Wittel, F. K., Mendoza, M., Vetter, R., & Herrmann, H. J. (2014). Particlefluid-structure 

interaction for debris flow impact on flexible barriers. Computer-Aided Civil 

and Infrastructure Engineering, in press. 

- Moriguchi, S., Borja, R. I., Yashima, A., & Sawada, K. (2009). Estimating the impact force 

generated by granular flow on a rigid obstruction. Acta Geotechnica, 4(1), 57–71. 

- Shieh, C. L., Ting, C. H., & Pan, H. W. (2008). Impulsive force of debris flow on a curved 

dam. International Journal of Sediment Research, 23(2), 149–158. 

- Sulsky, D., Chen, Z., & Schreyer, H. L. (1994). A particle method for hystory-dependent 

materials. Computer Methods in Applied Mechanics and Engineering, 118(1-2), 179–196. 

- Vermeer, P. A., Sittoni, L., Beuth, L., & Wieckowski, Z. (2013). Modeling soil-fluid and 

fluid-soil transitions with applications to tailings. In Taylings and Mine waste (pp. 305–315). 

Banff, Alberta, Canada. 

- Yerro, A., Alonso, E. E., & Pinyol, N. M. (2015). The material point method for unsaturated 

soils. Geotechnique, 65(3), 201–217. 

- Zanuttigh, B., & Lamberti, A. (2006). Experimental analysis of the impact of dry avalanches 

on structures and implication for debris flows. Journal of Hydraulic Research, 44(4), 522–534. 



Property protection against flooding shown on a 

complex practical example 

Objektschutz Hochwasser dargestellt an einem 

komplexen Praxisbeispiel 

Thomas Egli 1 ; Daniel Sturzenegger 1 ; Pierre Vanomsen 1 

ABSTRACT 

The canton of St. Gallen runs an integrated rescue centre. It receives all emergency calls to 

the police, fire brigade and ambulance and coordinates their operations. In reference to the 

hazard mapping the centre is subject to frequent flooding (Annuality of 30). Expected 

damages will concern the operational capability, the working staff and massive property 

damage. 

To reduce the high risk effectively, different protecting measures are proposed and evaluate 

due to feasibility, suitability for daily use, architecture, investment costs and maintenance 

costs. The proposed protecting measures may reduce the initial risk from 710'715 CHF/year 

to 66'000 CHF/ year. 


Die Kantonale Notrufzentrale (KNZ) des Kantons St. Gallen ist eine integrierte Notrufzentrale. 

Sie empfängt sämtliche Notrufe der Bevölkerung an die Polizei, Feuerwehr und Sanität 

und koordiniert die Einsätze. Aufgrund der Gefahrenkartierung ist davon auszugehen, dass 

bereits bei einem häufigen Hochwasser (Jährlichkeit 30) das Gebäude überschwemmt wird 

und neben den betrieblichen Störungen und Personenschäden auch massive Sachschaden im 

Rahmen von CHF 20 Mio. entstehen können. 

Damit das grosse Risiko wirksam gesenkt werden kann, werden verschiedene Schutzmassnahmen 

vorgestellt und aufgrund der technischen Machbarkeit, Praxistauglichkeit, Beeinflussung 

der gegebenen Architektur, Investitions- und Unterhaltskosten evaluiert. Mit den 

gewählten Schutzmassnahmen kann das Ausgangsrisiko von ursprünglich 710‘715 CHF/Jahr 

auf 66‘000 CHF/Jahr gesenkt werden. 

Mit Hilfe des risikobasierten Ansatzes lassen sich die effizienten Massnahmen für den 

Objektschutz ermitteln. Schliesslich sind neben den rein wirtschaftlichen Faktoren auch 

weitere Aspekte von gleichbedeutender oder sogar höherer Wichtigkeit und somit für 

den erfolgreichen Objektschutz gegen Hochwasser zu berücksichtigen. 

KEYWORDS 

property protection; risk; risk-based planning of measures; flooding 

1 Egli Engineering AG, St. Gallen, SWITZERLAND, egli@naturgefahr.ch 



EINFÜHRUNG 

Die Kantonale Notrufzentrale (KNZ) des Kantons St. Gallen liegt an den alten Stadtmauern 

von St. Gallen und ist ein Werk des Architekten Santiago Calatrava. Die KNZ steht im 

potentiellen Überflutungsgebiet der Steinach, siehe dazu auch Abbildung 3. 

Abbildung 1: Die Kantonale Notrufzentrale (KNZ) in St. Gallen. Der muschelförmige Bau ist umgeben von den historischen Bauten der 

Klosteranlage (Quelle: Egli Engineering AG). 

Die KNZ ist eine integrierte Notrufzentrale; hier werden sämtliche Notrufe der Bevölkerung 

an die Polizei, Feuerwehr und Sanität aus dem Kanton St. Gallen sowie teilweise aus 

Appenzell Innerrhoden und Ausserrhoden empfangen. Die KNZ alarmiert die Einsatzkräfte 

der Blaulichtorganisationen. Zusätzlich ist die KNZ die Alarmierungsstelle für den kantonalen 

Führungsstab und dient als Verkehrsleitstelle. 

Die für das kantonale Sicherheitsdispositiv ausserordentlich wichtige Funktion der KNZ 

verlangt auch den Betrieb während eines Elementarereignisses. Dies erfordert die Kenntnis 

der Eindringstellen des Wassers und den darauf basierenden Schutz der Infrastruktur mittels 

Objektschutzmassnahmen. 

METHODEN 

Die Methodik orientiert sich an bestehenden Grundlagen der Risikoanalyse, Risikobewertung 

und der Massnahmenevaluation und folgt grundsätzlich dem Risikokonzept für Naturgefahren 

(siehe Abbildung 2, Bründl M. (Ed.) (2009)). 


Abbildung 2: Methodisches Vorgehen für die risikobasierte Planung von Schutzmassnahmen gemäss Bründl M. (Ed.) (2009) 

ERGEBNISSE 

Risikoanalyse 

Die Risikoanalyse hat zum Ziel die Sach- und Personenrisiken quantitativ zu bestimmen 

und eine qualitative Einschätzung des Betriebsrisikos zu geben. 

Risikoanalyse: Gefahrenanalyse 

Wie in Abbildung 3 ersichtlich, zeigt die Gefahrenbeurteilung bereits bei einem Hochwasser 

mit einer 30 – jährlichen Wiederkehrdauer eine Überflutung an. Das Wasser fliesst über den 

ehemaligen Bachlauf längs der Strassen ab. Auch bei einem seltenen (HQ 100 

) und sehr 

seltenen Hochwasser (HQ 300 

und EHQ) entlastet der Vorfluter (die Steinach) auf dieselbe 

Weise. 

Aufgrund der Multiplikation von Fliesstiefe [m] und Fliessgeschwindigkeit [m/s] des Wassers 

wird die Intensität definiert. Die starke Intensität ist grösser 2 m 2 /s, die mittlere Intensität 

bewegt sich zwischen 0.5 und 2 m 2 /s und die schwache Intensität ist kleiner 0.5 m 2 /s. 

Bei Fliessgeschwindigkeiten kleiner 1 m/s ist nur die Fliesstiefe massgebend. 

Risikoanalyse: Expositionsanalyse 

Die möglichen Eintrittstellen des Wassers in das Gebäude sind in Abbildung 4 dargestellt: 

Die Eintrittstelle 1 umfasst ein Tor und die beiden Personaleingänge. Die Stelle 2 sind zwei 


Abbildung 3: Fliesstiefe und Fliessgeschwindigkeit bei der Kantonale Notrufzentrale (KNZ) (Naturgefahrenkommission des Kantons 

St. Gallen (2008). 

Aussentüren des Nachbargebäudes, welches mit der KNZ verbunden ist. Die Eintrittstelle 3 

sind Kanalisationsleitungen, welche im Hochwasserfall rückgestaut werden. Bei hohen 

Fliesstiefen kann Wasser über das Fenster in das Gebäude eindringen (Eintrittstelle 4) oder 

über die Leitungsdurchbrüche in das Gebäude gelangen (Eintrittsstelle 5). 


Abbildung 4: Eintrittstellen an der Kantonale Notrufzentrale (KNZ) 

Im Falle einer Überschwemmung der KNZ besteht für die Personen im Untergeschoss eine 

Gefährdung, im Obergeschoss sind die Personen sicher. Die durchschnittliche Aufenthaltsdauer 

der Einsatzdisponenten im Untergeschoss sind während eines Tages 166 Personenstunden; 

dies entspricht einer durchschnittlichen Belegung von 6.9 Personen während 24 Stunden. 

Risikoanalyse: Konsequenzensanalyse 

Sind die Eintrittstellen eingestaut, kommt es zur vollständigen Flutung des Untergeschosses 

(Betriebsraum, Apparateraum, Technikraum, Lüftung/Sänitär und Leitstelle) und zu Schäden 

im Erdgeschoss (Garageraum, Garderobe Uniform, Liftanlage, Korridor, Lager) und der 

Galerie (Garderobe, Dusche/WC, Aufenthaltsraum, Schleuse). Dabei ist von einem Totalschaden 

der Gebäudeinfrastruktur und -technik auszugehen, was einem gesamten Sachschaden 

von schätzungsweise CHF 20 Mio. entspricht. Die Kosten setzen sich zusammen aus 

den getätigten Renovationen im Jahre 2008 mit Investitionen in die Elektrotechnik, Lüftung 

und Stromverteilung im Umfang von CHF 12 Mio. Weiter werden bei einer Überflutung 

Heizungsanlage, Funk- und Telefonanlage, sowie Mobiliar beschädigt und im geringen 

Umfang werden auch Gebäudeschäden auftreten. Diese Schäden werden auf CHF 8 Mio. 

geschätzt. Der gesamte Sachschaden von CHF 20 Mio. beinhaltet keine weiteren Schäden, 

z. B. durch Verformungen infolge Gebäudeauftriebs. 

Die Wahrscheinlichkeit eines Todesfalls (Letalität) bei einer Überschwemmung in der KNZ 

wird wie folgt angenommen: Schwache Intensität: 0, mittlere Intensität 0.025, starke 

Intensität 0.24. Die Letalität ist hoch, da der Fluchtweg auf den Brandschutz ausgerichtet 

ist und unglücklicherweise dem Fliessweg des Wassers folgt. 


Risikoermittlung: Sachrisiko 

Es ist davon auszugehen, dass bereits bei einem 30 – jährlichen Hochwasser der Steinach 

der gesamte Sachschaden entstehen kann. Bei einem extremeren Ereignis (HQ 100 

, HQ 300 

und 

EHQ) steigt der Sachschaden kaum mehr an. 

Tabelle 1: Sachrisiko der Kantonale Notrufzentrale (KNZ) 

Wiederkehrperiode [Jahre] Sachschaden [CHF] Häufigkeit Sachrisiko [CHF/Jahr] 

30 20'000’000 0.023 460’000 

100 20'000’000 0.007 140’000 

300 20'000’000 0.002 40’000 

1’000 20'000’000 0.001 20’000 

Total 660’000 

Risikoermittlung: Personenrisiko 

Das Schadenausmass wird in Anzahl Todesfälle ausgedrückt und ist das Produkt aus der 

Anzahl anwesender Personen in der KNZ (6.9 Personen) und der Letalität. Der Todesfall 

kann als Geldwert ausgedrückt werden, dazu wird er monetarisiert. Wie hoch ein Todesfall 

monetarisiert wird, hängt ab von der gesellschaftlichen Zahlungsbereitschaft zur Verhinderung 

eines Todesfalles; in der Schweiz werden in der Regel CHF 5 Mio. verwendet 

(Bründl M. (Ed.) (2009)). 

Tabelle 2: Schadenausmass Personen 

Intensität Letalität Personen Monetarisierung [CHF] 

schwach 0 6.9 0 

mittel 0.025 6.9 862’500 

stark 0.24 6.9 828’0000 

Es wird angenommen dass bei der Wiederkehrperiode 30 und 100 Jahre eine mittlere Intensität 

und bei einem 300- und 1‘000-jährlichen Hochwasser starke Intensität im Gebäude 

vorliegt. Das Produkt der Jährlichkeit und des Schadenausmasses ergibt das Personenrisiko 

und beträgt 50'716 CHF/Jahr: 

Risikoermittlung: Betriebsrisiko 

Der technische Ausfall der KNZ lässt sich dank den redundanten Systemen (Einsatzleit- und 

Informationssystem der Kantons- und Stadtpolizei St. Gallen, ELIS) mit Einschränkungen 

bewältigen. Die Kosten für eine Betriebsverlegung werden nicht beziffert und fliessen nicht 

in die Risikoanalyse ein. 


Tabelle 3: Personenrisiko der Kantonale Notrufzentrale (KNZ) 

Wiederkehrperiode [Jahre] Personenschaden [CHF] Häufigkeit Personenrisiko [CHF/Jahr] 

30 862’500 0.023 19‘838 

100 862’500 0.007 6’038 

300 828’0000 0.002 16’560 

1’000 828’0000 0.001 8’280 

Total 50’716 

Risikobewertung 

Das Massnahmenziel von Schutzvorkehrungen an bestehenden Bauten wird so festgelegt, 

dass der jährliche Nutzen (Risikominderung) die jährlichen Kosten übersteigt. Im vorliegenden 

Fall führt dies zur Konsequenz, dass für Schutzmassnahmen Kosten von CHF 644‘715 

investiert werden können. Das Massnahmenziel entspricht im vorliegenden Fall einem 

300-jährlichen Ereignis. 

Massnahmenplanung 

Das hohe Risiko soll mit geeigneten Massnahmen wirksam reduziert werden (Egli T. (2005)). 

In der Folge werden einige Massnahmen genauer vorgestellt. Die Bemessung der Massnahme 

richtet sich nach der Wiederkehrperiode des Hochwassers (Schutzziel), der Fliesshöhe aus der 

Intensitätskarte und einer Reserve (Freibord) von 0.25 m. Daraus ergibt sich eine Höhe der 

Objektschutzmassnahme von 1.75 m. 

Massnahmenplanung: Vorschlag ‘Sperre’ bei der Eintrittstelle 1 

Auf der Länge von 6.1 m schützt ein Dammbalkensystem vor Hochwasser. Fix eingebaut und 

permanent sichtbar sind die seitlichen Führungsschienen und die im Boden versenkten 

Ankerpunkte. Falls aus architektonischer Sicht die Führungsschienen nicht akzeptabel sind, 

können diese auch ausgehängt werden, so verbleiben nur die Fixationspunkte im Mauerwerk. 

Entscheidend für die Funktionssicherheit ist ein rechtzeitiger und fachgerechter Aufbau 

des Systems. Dazu braucht es Personal, das permanent zur Verfügung steht und ein Warnsystem, 

das die Aktivitäten auslöst. 

Tabelle 4: Zusammenstellung der Kriterien zum Dammbalkensystem 

Umsetzung 

Praxistauglichkeit 

Architektur 

Unterhalt 

Investition 

Grundsätzlich einfach. Aufwendig im Fall der Datenübermittung mit 

Kabel. 

Keine Einschränkung zum heutigen Betrieb im Normalzustand. Doch 

unter Umständen ist mit häufigem Aktivierungsalarm zu rechnen 

und der Zutritt in die KNZ ist verwehrt. 

Kleiner Eingriff (seitliche Führungsschienen) 

Jahreskontrolle und periodische Reinigung 

- Dammbalken CHF 30'000.- inkl. bauliche Umsetzung 

- Warnsystem CHF 30'000.- inkl. bauliche Umsetzung 


Massnahmenplanung: Vorschlag ‘Barriere und abdichten Türen‘ bei der 

Eintrittstelle 1 

Hinter dem bestehenden Tor soll eine automatische Barriere den Wasserzutritt verhindern. 

Die Barriere ist innerhalb der Garage und damit äusserlich nicht sichtbar. Die Barriere wird 

nur im Ernstfall geschlossen und ist im Ruhezustand hochgefahren. 

Die beiden Personaleingänge müssen mit je einer zusätzlichen Drucktüre vor dem Wasser 

geschützt werden. Die Drucktüren müssen auf der Aussenseite angebracht werden, damit 

erhöht die Wassermasse den Anpressdruck auf den Türrahmen und die Dichtigkeit steigt. 

Massnahmenplanung: Schutzmassnahmen bei den restlichen Eintrittstellen 

Die restlichen Eintrittsstellen werden mit Hochwasserschutztüren (Stelle 2, siehe Abbildung 

4), Rückstauklappen (Stelle 3), Hochwasserschutzfenster (Stelle 4) und Leitungsabdichtungen 

(Stelle 5) abgedichtet. 

Massnahmenwahl 

Um die Eintrittstelle 1 zu schützen, wurden neben den oben genannten Vorschlägen auch 

ein Klappschott und ein Schlauchwehr vorgeschlagen. Beide Massnahmen sind jedoch bei 

der Detailplanung nicht weiter verfolgt worden, da die beengenden Verhältnisse im Torbereich 

nicht genügend Platz für diese beiden Varianten boten. 

Der Auftragsgeber entschied sich für den Einsatz eines Dammbalkensystems mit Alarmierung, 

wasserdichte Türen, Fenster, Rückstauklappen und Leitungsabdichtungen. 

Risikoanalyse nach Massnahmen 

Mit den ausgeführten Objektschutzmassnahmen ist die KNZ bei einem 300-jährlichen 

Hochwasser der Steinach geschützt mit einem Risiko nach Massnahmen von 66‘000 CHF / 

Jahr. Die Risikoanalyse nach Massnahmen geht von einer Wirksamkeit der Objektschutzmassnahmen 

von 90% aus. Dieser Wert ist hoch, besonders für mobile Massnahmen wie der 

Einsatz des Dammbalkensystems. Doch mit dem Alarmierungssystem und der permanenten 

Anwesenheit der Disponenten kann eine hohe Zuverlässigkeit erreicht werden. Entsprechend 

reduzieren sich die Sachschäden auf 10% zum ursprünglichen Wert. Personenschäden sind 

keine mehr zu erwarten da die Disponenten gewarnt sind. 

Es ergibt sich somit eine Risikoreduktion von ursprünglich 710‘715 CHF/Jahr auf 66‘000 CHF 

/Jahr, der Nutzen ist entsprechend 644‘715 CHF/Jahr 

FAZIT 

Die Kantonale Notrufzentrale (KNZ) liegt im Überflutungsbereich bei einem Hochwasser 

der Steinach. Die durchgeführte Risikoanalyse zeigt ein sehr hohes Risiko für die KNZ von 

710‘715 CHF/Jahr. Mit den aufgezeigten Objektschutzmassnahmen kann dieses Risiko 

reduziert werden auf 66‘000 CHF/Jahr. Die Schutzmassnahmen können somit das Risiko 

sehr effizient verringern. 


Abbildung 5: Das eingebaute Dammbalkensystem vor dem Tor der KNZ 


LITERATUR 

- Bründl M. (Ed.) (2009): Risikokonzept für Naturgefahren – Leitfaden. Nationale Plattform 

für Naturgefahren PLANAT, Bern. 

- Egli T. (2005): Wegleitung Objektschutz gegen gravitative Naturgefahren, Vereinigung 

Kantonaler Feuerversicherungen VKF, Bern. 

- Naturgefahrenkommission des Kantons St. Gallen (2008): Naturgefahrenanalyse Steinach. 

Projektverfasser Bänzinger Partner AG, Oberriet; Basler und Hofmann AG, Esslingen. 



Flood corridors to handle residual risk 

Markus W. Klauser, Dr. sc. ETH, Dipl.Ing. 1 ; Christian Willi, Dipl. Forsting. ETH 2 ; Werner Fessler, Dipl. Kult. Ing. ETH 3 ; 

Josef L.A. Eberli, Dipl. Kult. Ing. ETH 3 

ABSTRACT 

Settlements are spreading in the floodplains of rivers as a result of economic development 

and the rising demand for land. By means of structural and organizational measures, it is 

possible to reduce the integrated risk (in terms of expected loss per year, aggregated over all 

investigated scenarios). However, it should be noted that only spatial planning measures like 

the elimination of flood corridors (i.e. areas with reduced or no damage potential) as part of 

the urban planning measures of municipalities can guarantee a forward-looking, controlling 

effect. 

In the «Stanser Talboden, Switzerland», structural and organizational measures have already 

been implemented to protect settlements. These measures were part of the flood protection 

project «Engelberger Aa» and have shown a positive efficiency in the past. This study focuses 

on additional spatial planning instruments and shows that the flood protection can be further 

developed by creating a flood corridor, in a cost-effective and trend-setting way. 

KEYWORDS 

floods, hydraulic engineering - organizational - spatial planning measures, emergency 

planning, flood corridors, climate change, damage potential, effect-cost analysis 

INTRODUCTION 

Sustainable spatial planning is a key component in the integrated risk management of natural 

hazards. It is implemented by taking into account the hazard zones in the land-use plans of 

municipalities. These measures, such as building bans or the set-up of physical protection 

measures help to mitigate the negative effects of flood hazards, but do not provide an 

adequate solution to control settlement evolution. 

Flood corridors allow the reduction of settlement evolution in certain areas. In addition, the 

instrument is very flexible, allowing a certain worsening of the natural hazard situation due 

to e.g., climatic changes. The present study examines the instrument of flood corridors at the 

economic level using the example of the flood protection system «Engelberger Aa». We show 

that the implementation of such spatial instruments can be cost-effective measures complementing 

both existing hard (hydraulically engineered) and soft (organizational) measures. 

1 Tiefbauamt Kanton Nidwalden, Stans, SWITZERLAND, markus.klauser@nw.ch 

2 Ernst Basler + Partner AG 

3 Tiefbauamt Kanton Nidwalden 



The planning and design of the «Engelberger Aa» flood protection project started in 1987. 

The first four stages were completed at the end of 2007. The main elements of the project 

included four flood dikes to control excess load (flood dikes A-D, see Figure 1). These were 

created in locations where, in the case of excess load, the excess water with a low hazard 

potential can be discharged outside the river channel but in a controlled way. In traditional 

flood protection management, «excess load» would typically result in the uncontrolled 

breaching of structures (such as banks) and, in most cases, extensive damage as observed 

during flood events of the recent past. 

The flow path used in our case for the discharged water is called a flood corridor. Buildings 

or settlements located in the flow path are protected to the level of their protection objective 

and through the implementation of local measures. Thus it is guaranteed that the maximum 

volume of water that remains in the channel at each discharge point corresponds to the 

capacity of the next river section. As soon as the flood hydrograph returns to a level below 

the dimensioning-based water volume, flood water will again be discharged through the river 

channel. After the raising of the dikes, the maximum flow rate in the main settlement area 

between Dallenwil and Ennerberg has been 300m 3 /s. After the construction of the fourth 

flood dike, a maximum flow rate of 150 m 3 /s will remain in the river bed; this is the 

maximum discharge which can be channeled into the lake without causing any damage to 

the settlement of Buochs (Kanton Nidwalden, 2009) 

This study focuses on the flood corridor of the «Stanser Talboden» which affects the municipalities 

of Dallenwil, Oberdorf, Stans, and Stansstad (Figure 1). The corresponding excess load 

dike D deviates excess waters of the Engelberger Aa into the «Stanser Talboden» in the case 

of very rare event (extreme flood, EHQ), which in turn will result in a residual flooding risk 

(noteworthy, parallel acting torrent processes are not covered in this study and therefore not 

considered further). A weir edge secures flood dike D (EHQ excess overload) from failure. 

In organizational terms, the functionality of the flood corridor is guaranteed by a documented 

and practiced emergency planning. 

Settlement pressure at the «Stanser Talboden» is high and growth will likely continue in the 

future. Due to the current flood situation, it is therefore essential to guide settlement pressure 

and to align it within the flood corridors in a sustainable manner. Therefore, it is important to 

incorporate flood corridors in zoning plans of municipalities and to be able to control 

construction activity within this corridor through suitable provisions in zoning regulations. 

METHODS 

As part of the efforts to improve flood protection on the Engelberger Aa, various measures 

have been implemented in recent years or are currently under investigation. Based on the 

reference condition of the Engelberger Aa (before 2003), the following three measures have 

been realized or are planned: 


Lake 

Ennetbürgen 

Lake 

Stansstad 

Flood Corridor 

Buochs/Ennetbürgen 

60m 3 /s 

Flood dikes A and B 

45 m 3 /s 

45m 3 /s 

Buochs 

Flood Corridor 

Stanser Talboden 

Stans 

Flood dike C 

Oberdorf 

Engelberger Aa 

< 100 m 3 /s 

Flood dike D 

(EHQ-excess load) 

Dallenwil 

Figure 1: Schematic layout of the flood corridor within the „Stanser Talboden“ 

– Hydraulic Engineering: all hydraulic engineering measures within the flood protection 

project of the Engelberger Aa, such as spillways, dam remediation, bank protection, or 

flood proofing have been realized between 2003 and 2008. The associated calculations 

describe the risk-reduction achieved by hydraulic engineering measures. 

– Organization: the existing emergency planning complements the hydraulic engineering 

measures and is used by the emergency services to reduce flood impacts. The associated 

calculations describe the risk reduction achieved by organizational measures. 

– Spatial Planning: it is assumed on a notional basis that the flood corridor in the «Stanser 

Talboden» is spatially established in 2015 and able to control settlement evolution over the 

coming 70 years. It is important to mention that the flood corridor (which was built as part 

of the hydraulic engineering measures for very rare events) is already operational and 

functional. However, future settlement evolution is not yet guided by a forward-looking 

management like a flood corridor, which is simulated in this study. 


During the investigation, we started by defining the study area within the «Stanser Talboden». 

The actual risk situation was then determined based on existing risk assessments (Kanton 

Nidwalden, 2012). As a result, settlement evolution in the municipalities of the «Stanser 

Talboden» was then estimated based on data from the Canton’s agglomeration program 

(Kanton Nidwalden, 2011) for the next 70 years (i.e. until 2085). This was done as follows: 

– Assumption of an average increase in population by 0.55% per year over the next 70 years. 

– Definition and spatial localization of the necessary building zones. 

– Combination of the investigated settlement situation in 2085 with the flood hazard 

situation (i.e. intensity map). Determination of process intensities. 

– Conversion of potentially constructed areas into property categories. 

The assumed growth scenario results in a total settlement area requirement of about 200 ha 

within the four municipalities. Two-thirds of this land would be required as residential land; 

whereas one third was defined as land for public, commercial, or industrial buildings. Within 

the flood corridor about 70 ha of land are defined as construction land. Noteworthy, only 

one-third of this surface will not be flooded by the extreme flood of the Engelberger Aa. 

Residential development can be controlled through the introduction of a flood corridor zone. 

By refining the flood corridor zone through a definition of different sectors, the guiding effect 

of land-use development can be optimized as follows: 

– Sector A: Absolute construction ban. 

– Sector B: Designation and development of building zones is not possible – especially not yet 

constructed areas between buildings have to be removed. No possibility to add new 

buildings (except for replacement). 

– Sector C: Consolidation and realignment of existing building zones is possible (non-constructed 

areas between buildings, etc.). Existing building zones remain intact. 

Commercial and industrial as well as public areas will be planned and realized outside the 

flood corridor – but not inside. New residential buildings will be restricted to sector C. These 

measures result in a reduction of the damage potential in the flood corridor, but also in a 

reduction of land surface suitable for construction from 70ha to about 2ha which corresponds 

to roughly 90 housing units which can still be realized within sector C. 

We also considered property protection measures within the flood corridor, to be undertaken 

in the context of building renovations. 

The costs for the implementation of the flood corridor can be categorized as follows: 

– Compensation: costs related to the removal of building zones from the flood corridor were 

estimated at ca. 1,500,000.— CHF. 

– Property protection measures: a contribution of 5,000.— CHF was taken into account per 

building which needs to be renovated. 

– Procedural and administrative costs: it was assumed that administrative costs will be in the 

order of CHF 75’000.— CHF. 


– The costs for flood dikes, protection gates, terrain adjustments or excess load protection 

have been attributed to the hydraulic engineering measures for the flood protection project 

«Engelberger Aa». 

RESULTS 

The hydraulic engineering, organizational and planning measures can be evaluated based on 

two parameters, namely effect and efficiency. 

The effect can be investigated by comparing the integrated risk before and after the implementation 

of measures. Previous studies (Kanton Nidwalden, 2012) provide the results of the 

hydraulic engineering and organizational measures (Table 1). They show an overall reduction 

in integrated risk from ca. 4,261,000. to 1,292,000.— CHF/yr (-77%) as a result of the 

structural measures (Figure 1). Through the addition of organizational measures the 

integrated risk was reduced by an additional 400,000.— CHF/yr (or -16%). 

Table 1: Effect-cost ratios of the structural and organizational measures (Kanton Nidwalden, 2012). 

Risk 

Integrated risk 

Effect: 

Reduction of integrated risk 

Cost 

Costs of measures 

Efficiency 

Effect/Costs 

none 

Measure 

Hydraulic 

Engineering 

Hydraulic 

Engineering + 

Organisation 

[CHF/yr] 5,553,000.-- 1,292,000.-- 400,000.-- 

[CHF/yr] -- 4,261,000.-- 5’153,000.-- 

[CHF/yr] -- 332,000.-- 438,000.-- 

[-] 13 12 

Efficiency is achieved by comparing the effects of measures with their annual cost. The annual 

costs include investment costs as well as operation and maintenance costs. It can be shown 

that the hydraulic engineering measures account for ca. 332,000— CHF/yr. Additional cost 

for the organizational measures (ca. 106,000.— CHF/yr) increase the total coast to ca. 

438,000.– CHF/yr. 

Table 2 shows the effect of the spatial planning measures studied in this paper (Kanton 

Nidwalden, in preparation). First it can be seen that settlement activity in the following years 

will increase risk significantly – followed by an increase in integrated risk from ca. 400,000.– 

CHF/yr today to ca. 478,000.– CHF/yr in 2085. The implementation of the flood corridor can 

not only compensate this increase, but also reduce it to ca. 366,000.– CHF/yr, which means a 

reduction of 23% by 2085. 

The above effect and efficiency calculations for the spatial measures were derived with 

EconoMe (Bundesamt für Umwel BAFU, 2015), a tool especially designed for risk calcula- 


tions, and show that the implementation of a flood corridor will result in a effect-cost ratio 

of approximately 3 by 2085 (Table 2). Noteworthy, annual cost of 40,000.– CHF/yr for spatial 

planning measures have been divided into compensation of land being removed from 

building zones (and thus reclassified, also in monetary terms), property protection as well as 

administrative costs. By far the largest proportion of costs (ca. 90%) is attributed to the 

compensation for land being removed from the building zone. Ignoring these costs leads to 

a effect-cost ratio of about 26. The efficiency (effect-cost) ratio is also shown in Figure 2. 

Table 2: Effect-cost ratios of the structural-organizational and structural-organizational+spatial planning measures (Kanton Nidwalden, 

in preparation) for the reference year 2085. 

Risk 

Integrated risk 

Effect: 

Reduction in integrated risk 

Cost 

Cost of measures 

Efficiency 

Effect/Costs 

Hydraulic 

Engineering + 

Organization 

(2085) 

Measure 

Hydraulic 

Engineering + 

Organization + 

Spatial planning 

(2085) 

[CHF/yr] 478,000.-- 366,000.-- 

[CHF/yr] -- 112,000.-- 

[CHF/yr] -- 40,000.-- 

[--] -- 3 

6'000'000 

5'000'000 

no measures 

(before 2003) 

500'000 

400'000 

Integrated Risk [CHF/yr] 

4'000'000 

3'000'000 

2'000'000 

300'000 

400'000 450'000 500'000 

1'000'000 

Hydraulic Engineering 

(since 2008) 

Hydraulic 

Engineering+Organization 

(since 2008) 

Hydraulic 

Engineering+Organization 

(2085) 

Hydr.Engineering+ 

Organization+ 

Spatial Planning 

(2085) 

0 

0 100'000 200'000 300'000 400'000 500'000 600'000 

Annual Cost [CHF/yr] 

Figure 2: Effect-cost diagram. 


CONCLUSIONS 

The results of the structural and organizational risk analysis show a clearly positive effect-cost 

ratio of 13 for the hydraulic engineering and 12 for hydraulic engineering + organizational 

measures, respectively. The calculations are based on existing databases and are considered to 

be reliable. 

To determine the efficiency with the spatial planning measures, we used data that we 

predicted for the year 2085. However, such a long-term prediction must be interpreted with 

caution. Furthermore, different methodological approaches of risk analysis in studies (Kanton 

Nidwalden, 2012 on one hand side and study Kanton Nidwalden, in preparation on the other 

hand side) may have led to certain inaccuracies. 

Sensitivity considerations made clear that the effect-cost ratio of the definition of a flood 

corridor is significantly influenced by the expected construction activity and the compensation 

that potentially needs to be paid for land being removed from building zones. The quality 

of the calculated effect-cost ratio of 3 can therefore be classified as «indicative» under the 

above criteria. 

As a conclusion, the following principles can be described: 

– The designation of flood corridors form a meaningful and cost-effective complement to 

hydraulic engineering and organizational measures. 

– The effectiveness of structural measures is continually reduced, due to the progressive 

housing development, thereby leading to a continuous increase in damage potential. 

The designation of flood corridors counteracts this effect in an economic way. 

– A flood corridor is a very flexible tool which can be used to deal with future problems 

including climate change. 

REFERENCES 

- Bundesamt für Umwelt BAFU, Bundesamt für Bevölkerungsschutz BABS (2009). 

Adapt Alp – Case Study Nidwalden. Kurzdokumentation, erarbeitet durch die Ernst 

Basler+Partner AG. 13. Januar 2009 

- Bundesamt für Umwelt BAFU, Bundesamt für Bevölkerungsschutz BABS (2011). 

RiskPlan – IT Tool fürs Risikomanagement. http://www.riskplan.admin.ch. Zugriff Mai 2012. 

- Bundesamt für Umwelt BAFU (2015). EconoMe 3.0. Wirtschaftlichkeit von Schutzmassnahmen 

gegen Naturgefahren. www.econome.admin.ch. Zugriff 9.6.2015. 

- Bundesamt für Wasser und Geologie BWG (2002). Excel Tool für die Abschätzung des 

Schadenpotentials Überschwemmung und Übermurung, SchaPo 2.1. 

- Kanton Nidwalden (2009): Integrales Risikomanagement am Beispiel der Engelberger Aa. 

http//www. nw.ch. Zugriff August 2015. 

- Kanton Nidwalden (2011). Agglomerationsprogramm Nidwalden 2011. Bericht. 

www. nw.ch. 


- Kanton Nidwalden (2012). Wirksamkeit der Notfallplanung als Beitrag zum integralen 

Risikomanagement. Schlussbericht, erarbeitet durch die Ernst Basler+Partner AG. 

18.Oktober 2012 

- Kanton Nidwalden (2012). Gefahrenbeurteilung Überflutung durch Engelberger Aa. 

Gemeinden Dallenwil, Oberdorf, Stans, Stansstad. Technischer Bericht und Karten, erarbeitet 

durch die Niederer und Pozzi AG, 1. September 2012. 

- Kanton Nidwalden (in Erarbeitung). Risikoanalyse. Prozess Engelberger Aa im Perimeter 

Stanser Talboden. Technischer Bericht. 



Flood protection concept against enourmous debris 

flow scenarios 

Hochwasserschutzkonzept für Wildbach mit sehr 

grossen Murgangszenarien 

Rolf Künzi 1 ; Martin Amacher 1 ; Oliver Markus Hitz 3 ; Serena Liener 4 ; Christian Tognacca 5 ; Markus Zimmermann 6 

ABSTRACT 

In the last 500 years several large debris flow events of the Lammbach torrent in the Bernese 

Oberland with casualties were recorded. Since 22 check dams have been erected in 1896 

a huge amount of debris has been accumulating behind the check dams. A recent hazard 

analysis takes the failure of the old check dams into consideration. Flood scenarios vary 

from probable events with a total load of 40'000 m 3 to extreme events with a total load of 

750'000 m 3 . The consequent hazard situation on the alluvial fan demands for protective 

measures. 

In the frame of a flood protection concept the efficiency of four alternatives „Retain", 

„Convey", „Divert" and "Corridor" was examined. The analysis revealed that large events 

could be handled best with a combination of „Retain" and „Corridor". As a consequence of 

a participatory process an additional alternative „Decentralised Retention" was elaborated. 

This alternative is most effective during small and medium events. It was finally chosen 

as the best alternative although its measures are less effective during extreme events. 


Beim Lammbach im Berner Oberland wurden in den letzten 500 Jahren mehrere grosse 

Murgangereignisse mit Todesfällen dokumentiert. Seit dem Bau von 22 Sperren nach dem 

Ereignis 1896 wird hinter den Sperren des Lammbachs sehr viel Geschiebe zurückgehalten. 

In einer Gefahrenbeurteilung, welche einen möglichen Bruch der alten Sperren berücksichtigt, 

werden Hochwasserszenarien mit Geschiebefrachten auf dem Schwemmkegel zwischen 

40'000 m 3 bei einem häufigen Ereignis und 750’000 m 3 bei einem Extremereignis definiert. 

Daraus resultiert eine Gefährdung auf dem Schwemmkegel, welche Schutzmassnahmen 

erfordern. 

1 Flussbau AG SAH, Bern, SWITZERLAND, rolf.kuenzi@flussbau.ch 

2 Mätzener & Wyss Bauingenieure AG, Hauptstrasse 21, CH-3800 Unterseen 

3 Tiefbauamt des Kantons Bern, Oberingenieurkreis I, Schlossberg 20, CH-3601 Thun 

4 Büro geo7 AG, Neufeldstrasse 5 - 7, CH-3012 Bern 

5 beffa togniacca sagl, A San Rocch, CH-6702 Claro 

6 NDR Consulting GmbH, Riedstrasse 5, CH-3600 Thun 



Im Rahmen der Massnahmenplanung wurde der Nutzwert von vier Varianten “Rückhalten”, 

“Durchleiten”, “Umleiten” und “Korridor” analysiert. Es zeigte sich, dass mit einer Kombination 

der Varianten “Rückhalten” und “Korridor” eine Überlast am besten bewältigt werden 

kann. Aufgrund von Rückmeldungen aus der Mitwirkung der Bevölkerung wurde eine 

zusätzliche Variante “dezentraler Rückhalt” erarbeitet. Diese Variante erzielt die grösste 

Wirkung bei häufigen und mittleren Hochwasserereignissen und wurde schliesslich als 

Bestvariante gewählt, obwohl sich die Massnahmen der ursprünglich favorisierten Variante 

bei Überlast stabiler verhalten würden. 

KEYWORDS 

case study; torrent; debris flow; flood protection; high bed load potential 

AUSGANGSSITUATION 

Beim Lammbach im Berner Oberland auf dem Gemeindegebiet von Brienz, Schwanden und 

Hofstetten, wurden in den letzten 500 Jahren mehrere grosse Murgangereignisse mit 

Todesfällen dokumentiert. Im Mai 1499 wurde der Ortsteil Kienholz vollständig zerstört. 

Dabei waren 400 Todesopfer zu beklagen. 1824 führten Murgänge im Lamm-, Schwanderund 

Glyssibach zu 6 Todesopfern. 1896 verursachten zwei grosse Ereignisse mit einem mobilisierten 

Volumen von mindestens 250’000 m 3 grosse Schäden. Nach dem Ereignis von 1896 

wurden im Lammbach 22 Sperren errichtet. Davon wurden im Laufe der Zeit drei zerstört. 

Mit dem Bau der Sperren wurde die Bachsohle je nach Abschnitt um bis zu 20 m erhöht. 

Im Weiteren wurden im oberen Einzugsgebiet Hangstabilisierungsmassnahmen in Form von 

Aufforstungen mit über 8 Mio. Pflanzen und rund 80'000 m 3 Trockensteinmauern erstellt. 

Beim Zusammenfluss von Lamm- und Schwanderbach wurde zudem ein Ablagerungsraum 

mit einem Rückhaltevolumen von bis zu 70’000 m 3 geschaffen. 

Im 20. Jahrhundert sind keine ausserordentlichen Murgangereignisse mehr aufgetreten. 

Auch im Jahr 2005, als sich in den benachbarten Bächen Tracht- und Glyssibach zwei 

verheerende Murgangereignisse ereigneten, blieb der Lammbach ruhig. Dies dürfte unter 

anderem auch auf die Massnahmen im Einzugsgebiet zurück zu führen sein. Die Lammbachsperren 

wurden gebaut um eine Sohleneintiefung zu verhindern. Diese Funktion haben sie 

erfüllt. Hinter den Sperren des Lammbachs wurde in den vergangen rund 100 Jahren sehr 

viel Geschiebe zurückgehalten (> 1 Mio. m 3 ). Das Rückhaltevolumen der Sperren ist mittlerweile 

ausgeschöpft. Da die Sperren bereits über 100 Jahre alt sind, stellte sich bei der Überarbeitung 

der Gefahrenkarte die Frage, wie sich die Sperrentreppe bei zukünftigen Ereignissen 

verhalten würde. Aus diesem Grund wurden die Standfestigkeit und die Gebrauchstauglichkeit 

der Sperren untersucht und die Entwicklung ihres Zustandes unter der Annahme 

prognostiziert, dass die Bauwerke nicht mehr unterhalten und Instand gestellt würden 

(Kister et al. 2012). 


Unter Berücksichtigung all dieser Faktoren wurden für die Gefahrenbeurteilung Szenarien 

mit grossen Geschiebefrachten definiert. Die daraus resultierenden Intensitäten der Murgänge 

in Kombination mit der Nutzung des Wildbachkegels führen zu einer Gefährdung auf dem 

Schwemmkegel, welche Schutzmassnahmen erfordern. 

EINZUGSGEBIET VON LAMM- UND SCHWANDERBACH 

Das Einzugsgebiet vom Lamm- (3.6 km 2 ) und Schwanderbach (2.6 km 2 ) ist in nördlicher 

Richtung durch den Brienzergrat (höchster Punkt Brienzer Rothohrn mit 2'350 m ü. M.) 

und gegen Süden durch den Brienzer See (558 m ü. M.) begrenzt. 

Das obere Einzugsgebiet des Lammbachs ist geprägt durch den steilen, von Verwitterung 

geprägten Gipfelaufbau des Arnihaaggen mit besonders ausgeprägten Rutschhängen auf der 

Ostseite. Ganz allgemein muss der im Lammbachgraben anstehende Fels aufgrund von 

Sackungsbewegungen als nicht sehr verwitterungsresistent und ausgesprochen erosionsfähig 

bezeichnet werden. Im Projektgebiet wird das Festgestein nicht nur auf der Talsohle, sondern 

auch über grosse Strecken an den Talflanken von Lockergestein bedeckt. Der Lammbach hat 

seit dem Rückzug des würmeiszeitlichen Aaregletschers sein Bett mehrere 100 m tief in die 

anstehenden Felsschichten eingeschnitten und dabei im Aaretal einen mächtigen Schuttkegel 

aufgeworfen, welcher bis zum Brienzersee reicht. 

Der Schwanderbach ist in seinem oberen Einzugsgebiet ein typischer Jungschuttbach mit 

relativ bescheidenen Geschiebeherden. Er weist steile, aber grösstenteils grasbewachsene 

Hänge in der Südflanke des Brienzer Rothorns auf. Der einzig grosse Geschiebeherd im 

Schwanderbach stellt der Rutschkomplex von Ägerdi dar. Der Schwemmkegel des Schwanderbachs 

ist nur schwach ausgeprägt. 

Abbildung 1: Sperre V im Lammbach 


SZENARIEN 

Die Hochwasserabflüsse sind in der Tabelle 1 ersichtlich. 

Tabelle 1: Hochwasserabflüsse Lamm- und Schwanderbach 

Ereignis 

Hochwasserspitzen 

Schwanderbach Lammbach Lammbach inkl. 

Schwanderbach 

HQ 30 13 m³/s 8.3 m³/s 21 m³/s 

HQ 100 24 m³/s 15 m³/s 40 m³/s 

HQ 300 36 m³/s 23 m³/s 60 m³/s 

EHQ 46 m³/s 30 m³/s 77 m³/s 

Das enorme Geschiebepotential und die Interaktion von Witterung, Erosion, Rutschungen 

und bestehenden Schutzbauwerken machen den Geschiebehaushalt des Lammbachs zu 

einem komplexen System. Die Schätzungen für das Geschiebeaufkommen sind in der Tabelle 

2 dargestellt. Als Extremereignis wird im Lammbach das Versagen der Sperrentreppe 

angesehen. Sollten in Folge einer sehr grossen Rutschung und eines entsprechend grossen 

Murgangs ein Grossteil der Sperren versagen, würden die dahinter zurückgehaltenen 

Schuttmengen erodiert. Die Murgänge im Lammbach treten sowohl bei starken Gewitterniederschlägen 

als auch bei längeren Nässeperioden auf. Wegen der grossen Speicherkapazität 

der Sackungsmasse braucht es allerdings ein jeweils grosses Ereignis. Die Mergelschichten 

liefern viel Feinmaterial (siltiges und toniges Material). Dadurch können die Murgänge bei 

bescheidenem Wasserangebot langsam und bei hohem Wasserangebot aber auch schnell 

fliessen. Grosse Blöcke fehlen im Einzugsgebiet fast vollständig. Das abgelagerte Murgangmaterial 

besteht deshalb primär aus Steinen, Sand und Feinmaterial. Beim Schwanderbach 

können insbesondere nach längeren Nässephasen durch eine Aktivierung der Rutschung 

Ägerdi grössere Geschiebemengen ins Gerinne und bis zum Kegelhals gelangen (vgl. Tabelle 2). 

Tabelle 2: Geschiebeszenarien Lamm- und Schwanderbach 

Schwanderbach 

Lammbach 

Jährlichkeit Geschiebepotential Sperren Rutschung 

30-jährlich 15'000 m 3 40'000 m 3 alle intakt 

100- 

jährlich 

300- 

jährlich 

Extremereignis 

30'000 m 3 110'000 m 3 alle intakt 

65'000 m 3 210'000 m 3 einzelne 

versagen 

750'000 m 3 Grossteil 

versagt 

kleinere Rutschung(en) 

(5'000 - 10'000 m 3 ) 

Rutschung 30'000 m 3 bis zu 

50'000 m 3 ; davon fliessen ca. 

10'000 m 3 als Murgang ab. 

Grossrutschung bis zu 100'000 m 3 

(evtl. sogar mehr); davon fliessen 

ca. 50'000 m 3 als Murgang ab 

Grossrutschung deutlich über 

100'000 m 3 


Bei beiden Wildbächen führen ähnliche Niederschlagsereignisse zur Mobilisierung von 

grösseren Murgangereignissen . Da die Einzugsgebiete der beiden Bäche direkt nebeneinander 

liegen, muss mit dem gleichzeitigen Auftreten von Ereignissen in beiden Bächen 

ge rechnet werden. Diesem Umstand wurde bei der Szenarienbildung Rechnung getragen. 

VARIANTENSTUDIUM 

In einem ersten Schritt wurde eine grosse Auswahl an möglichen Einzelmassnahmen in 

Bezug auf die technische Machbarkeit und Schutzwirkung überprüft. 

Die Wirkung der Massnahmen wurde auf dem Schwemmkegel mit einer 2d-Modellierung 

mit dem Programm Flumen und zusätzlich einer unabhängigen gutachterlichen Beurteilung 

im Feld abgeschätzt. Die Murgangsimulationen mit FLUMEN (vgl. www.fluvial.ch, Beffa 

2003) löst die instationären, tiefengemittelten Flachwassergleichung mit der Methode der 

finiten Volumina und ist geeignet für die Untersuchung von Einphasenströmungen. Das 

Programm ermöglicht die numerische Nachbildung von Fliess- und Ablagerungsprozessen 

von Murgängen unterschiedlicher Rheologien. Für die Murgangsimulationen im Lammbach 

wurde ein "turbulent-yield" Ansatz (vgl. Naef et al. 2006) angewendet. 

Aufgrund der grossen mobilisierbaren Geschiebemengen im Einzugsgebiet und den durch 

das Siedlungsgebiet eingeschränkten Platzverhältnissen auf dem Schwemmkegel wurde rasch 

erkannt, dass zumindest einer teilweisen Erhaltung/Erneuerung der Sperrentreppe eine 

grosse Bedeutung zukommt. Um die Geschiebeeinträge im oberen Einzugsgebiet nicht zu 

vergrössern, sollen die forstlichen Massnahmen auch in Zukunft fortgesetzt werden, auch 

wenn deren Wirkung nicht eindeutig quantifizierbar ist. Aus den untersuchten Einzelmassnahmen 

wurden schliesslich folgende vier Grundvarianten gebildet: 

Variante 1 „Rückhalten“ 

Mit neuen Leitdämmen auf dem Schwemmkegel wird sichergestellt, dass sämtliches Geschiebe 

bis in den bestehenden Rückhalteraum transportiert wird. Das Rückhaltevolumen des 

Sammlers wird mit einer Sohlenabsenkung und durch die Erhöhung der bestehenden 

Abbildung 2: Variante 1 „Rückhalten“ 


Begrenzungsdämme des Geschiebesammlers vergrössert. Die Wirksamkeit der Massnahmen 

konnte durch die numerische Modellierungen bestätigt werden. 

Für den Bereich unterhalb des Rückhalteraums verblieben jedoch unzulässige Gefährdungen 

und das Verhalten bei Überlast ist konzeptbedingt problematisch. 

Variante 2 „Durchleiten“ 

Oberhalb des bestehenden Rückhalteraums sind analog zur Variante 1 neue Leitdämme 

vorgesehen. Zusätzlich soll das bestehende Rückhaltevolumen weniger stark vergrössert und 

mit Leitmauern unterhalb des bestehenden Rückhalteraums möglichst viel Material in den 

See geleitet werden. 

Aufgrund des relativ flachen Gefälles bis zum See muss mit einem Rückstau in den Korridor 

ausgehend vom Delta gerechnet werden. Die technische Machbarkeit des „Durchleitens“ ist 

somit nicht gegeben, weshalb diese Variante nicht weiter in die Überlegungen mit einbezogen 

wurde. 

Variante 3 „Umleiten“ 

Variante 3 sieht auf dem Schwemmkegel eine vollständige Umleitung des Lammbachs vor. 

Das neue Gerinne führt über heute landwirtschaftlich genutzte Flächen in einen 150'000 – 

200'000 m 3 fassenden Sammler und weiter in den Brienzersee. 

Abbildung 3: Variante 3 „Umleiten“ 

Die Kosten für die Variante 3 „Umleiten“ sind doppelt so hoch wie für die Variante 4 „Korridor“. 

Ein Umleitgerinne erfordert grosse Anpassungen an bestehenden Infrastrukturanlagen. 

Trotz der aufwändigen Massnahmen blieben bewohnte Gebiete entlang des neuen Gerinnes 

immer noch durch Murgänge mit starker Intensität gefährdet. Die Kostenwirksamkeit ist 

deutlich kleiner als 1 und somit die Subventionierbarkeit des Projektes durch Kanton und 

Bund nicht gegeben. 


Variante 4 „Korridor“ 

Die Variante „Korridor“ sieht eine Entlastung des bestehenden, aber vergrösserten Rückhalteraumes 

über einen durch seitliche Begrenzungsdämme definierten Korridor vor. Leitdämme 

am Anfang des Schwemmkegels begrenzen die Ausuferungen und Ablagerungen so dass 

die Murschübe bis in den Rückhalteraum oder Korridor geleitet werden können. 

Abbildung 4: Variante 4 „Korridor“ 

Die numerischen Modellierungen zeigten eine gute Wirksamkeit der Massnahmen. Es konnte 

aber nicht ausgeschlossen werden, dass unerwünschte Ablagerungen das Anspringen des 

Korridors erschweren und zu viel Material in den Rückhalteraum abfliesst und bewohnte 

Gebiete weiterhin mit starken Intensitäten gefährdet sind. 

Bei den verbleibenden drei Varianten 1, 3 und 4 wurden die Schutzwirkungen untersucht 

und die Varianten im Rahmen einer Nutzwertanalyse mit einander verglichen. Bei der 

Variante „Umleiten“ wirkten sich die schlechte Kostenwirksamkeit und die in Frage gestellte 

Subventionierbarkeit negativ auf die Bewertung und Rangierung aus. Die besser platzierten 

Varianten 1 „Rückhalten“ und 4 „Korridor“ wiesen praktisch identische Summen bei der 

quantitativen Bewertung auf. Beide Varianten hatten jede für sich relevante Nachteile. Bei 

der Variante „Rückhalten“ war die Restgefährdung unterhalb des Sammlers noch zu gross, 

eine weitere Vergrösserung der Rückhaltemassnahmen auf dem Schwemmkegel hätte 

wiederum Probleme mit der Landschaftsverträglichkeit verursacht. Die Variante „Korridor“ 

wurde in Bezug auf die Kriterien „Eingriff in die Landschaft“ und „Akzeptanz“ schlecht 

beurteilt. Aus dem Prozess des Variantenstudiums hat man am Schluss eine Kombination 

der Varianten „Rückhalten“ und reduzierten „Korridor“ als neue Variante „Rückhalt mit 

reduziertem Korridor“ entwickelt, welche sich als Bestvariante erwies. 


Abbildung 5: Variante „Rückhalten mit reduziertem Korridor“ 

PARTIZIPATION MIT PROJEKTAUSWIRKUNGEN 

Die Variante „Rückhalt mit reduziertem Korridor“ wurde im Rahmen des gesetzlich vorgegebenen 

Mitwirkungsverfahrens öffentlich aufgelegt. Während der Auflage konnten sich alle 

interessierten Personen zum Projekt äussern. Die Mitwirkung hat gezeigt, dass bei der 

Bevölkerung eine grosse Skepsis gegenüber einem Überlastkorridor besteht. Man äusserte 

Befürchtungen, dass ein Überlastkorridor in der Gemeinde Brienz zukünftige Siedlungsentwicklungen 

einschränken könne. Zudem wurde bemängelt, dass bei der Variante „Rückhalten“ 

keine Rückhaltemassnahmen im Lammbach oberhalb des Schwemmkegels untersucht 

worden sind. 

Die Überprüfung eines ergänzenden Geschieberückhaltes oberhalb des Schwemmkegels 

mit einem Abschlussbauwerk von 15 m Höhe ergab ein zusätzliches Rückhaltevolumen von 

70'000 m 3 . Aufgrund der grossen Wirkung wurde dieser Vorschlag als zusätzliche Variante 

„dezentraler Rückhalt“ in die weitere Planung aufgenommen. Dabei wurde auf die Ausbildung 

eines Überlastkorridors mit Leitdämmen verzichtet. 

Ein Vergleich der Varianten „Rückhalt mit reduziertem Korridor“ und „dezentraler Rückhalt“ 

zeigt, dass die Kosten je CHF 20 Mio. betragen. Die Wirkung auf die Gefahrenkarte nach 


Abbildung 6: Variante „dezentraler Rückhalt“ 

Massnahmen ist durch den zusätzlichen Rückhalt bis zu sehr seltenen Ereignissen (HQ 300 

) 

bedeutend (vgl. Abbildung 7). 

Wegen der riesigen mobilisierbaren Geschiebemengen beim Extremereignis (EHQ) unterscheiden 

sich die Intensitätskarten der beiden Varianten nach einem EHQ nicht. Bei der 

Variante „Rückhalt mit reduziertem Korridor“ wird ein Teil des anfallenden Materials via 

Korridor in Richtung See geleitet, bis auch dieser überlastet und schliesslich der gesamte 

Schwemmkegel betroffen ist. Bei der Variante „dezentraler Rückhalt“ bestehen zwei 

unabhängig voneinander funktionierende Rückhaltemassnahmen (oberer und unterer 

Abbildung 7: Gefahrenkarte nach Massnahmen Variante „Rückhalt mit reduziertem Korridor“ links, Variante „dezentraler Rückhalt“ 

rechts 


Rückhalt) bei welchen bei Extremereignissen am Ende ebenfalls der gesamte Schwemmkegel 

übermurt wird. Zusammengefasst kann gesagt werden, dass die Schutzwirkung der Variante 

„dezentraler Rückhalt“ bis zu sehr seltenen Ereignissen (HQ 300 

) besser beurteilt wird, die 

Variante „Rückhalt mit reduziertem Korridor“ bei extremen Ereignissen (HQ 300 

) gewisse 

Vorteile aufweist, welche sehr schwer zu quantifizieren sind. Ob sich z.B. eine Siedlungsentwicklung 

in den Bereich des Korridors mehr auf die Risiken auswirkt als eine Verdichtung 

derbestehenden Siedlungsstruktur ist nur sehr schwer zu beantworten und wurde nicht 

detaillierter untersucht. 

In Anbetracht all dieser Überlegungen und unter Berücksichtigung der Tatsache, dass der 

für den ursprünglich vorgesehenen Überlastkorridor beanspruchte Raum für die nächsten 

10 – 15 Jahre auch ohne Projekt durch Siedlungsbegrenzungslinien behördenverbindlich 

freigehalten wird, hat man sich auf der Basis einer Nutzwertanalyse für die Variante 

„dezentraler Rückhalt“ entschieden. 

Die Variante soll nun zum Bauprojekt weiter bearbeitet und bis 2017 planrechtlich sichergestellt 

werden. 

FAZIT 

Der Planungsprozess hat gezeigt, dass das Definieren von raumplanerischen und baulichen 

Schutzmassnahmen, welche bei extrem seltenen Ereignissen ausserhalb unserer Erfahrungswelt 

und trotz grossen Dimensionen immer noch eine Schutzwirkung haben, ausserordentlich 

schwierig und anspruchsvoll ist. Dies hauptsächlich in Bezug auf die Akzeptanz für 

Massnahmen, welche in Anbetracht des gewaltigen Ausmasses eines Extremereignisses nur 

eine begrenzte Wirkung aufweisen, sich aber auf das Landschaftsbild und die Siedlungsentwicklung 

auswirken. Diese Problematik verschärft sich im alpinen Raum, wo die Siedlungsflächen 

in ihrer Ausdehnung durch die Topografie beschränkt sind. Durch den partizipativen 

Prozess ist schlussendlich ein Projekt entstanden, das zwar nicht alle Probleme gänzlich zu 

lösen vermag, unter Berücksichtigung aller Bedürfnisse aber eine optimale Lösung darstellt. 


LITERATUR 

- Amacher, Martin, Christian Kaufmann, David Hodel (alle Mätzener&Wyss Bauing. AG), 

Markus Zimmermann (NDR Consulting GmbH), Serena Liener (geo7) und Christian Tognacca 

(beffa tognacca gmbh) 2014: Wasserbauplan Lamm- und Schwanderbach Mitwirkungsdossier 

- Zimmermann, Markus (NDR Consulting GmbH) 2015: Szenarien und Ereignisabläufe bei 

der Variante „dezentraler Geschieberückhalt“ 

- Kister, Bernd, Markus Zimmermann, Gabi Hunziker, Bruno Zimmerli und Walter Fellmann 

2012: Analysis of torrent protective structures as a basic element of the hazard mapping 

process. Determining the as-is-state of 100 years old check dams made of natural stone 

masonry. Grenoble: 12th Congress Interpraevent. 729 – 740 

- Beffa, Cornel: 2D-Strömungssimulation mit FLUMEN, Wiener Mitteilungen (2003) Band 

18, BOKU-Wien 

- Naef, D., D. Rickenmann, P. Rutschmann, B.W. McArdell, 2006: Comparison of flow 

resistance relations for debris flows using a one-dimensional finite element simulation model, 

Nat. Hazards Earth Syst. Sci., 6, 155–165 



The impact of debris flows on structures: practice 

revisited in light of new scientific results 

Dominique Laigle 1 ; Mathieu Labbé, PhD 2 

ABSTRACT 

Our study aims at determining values of the pressure in the vicinity of a structure subjected 

to a debris flow impact, and their evolution in time and space, for a given incident flow. We 

used a numerical model based upon the SPH (smoothed particles hydrodynamics) numerical 

method and simulated the flow features upstream of an obstacle impacted by a viscoplastic 

fluid wave. We focus interest on the pressures developed on the structure during the impact. 

The results are analyzed in view of the incident flow features, characterized by its Froude 

number Fr. We evidence a transition between what we name the dead-zone impact regime 

and the jet impact regime. We establish that this transition occurs for values of the Froude 

number between Fr ≈ 1.3 and 1.5. We evidence the respective roles of the gravitational and 

kinetic components of the pressure. Finally, we study the spatial distribution of the pressure 

applied to the obstacle and its evolution with time depending on the impact regime. 

KEYWORDS 

debris-flow; SPH; Protection structures; impact pressure 

INTRODUCTION 

The understanding and quantification of mudflow and debris flow – structure interactions in 

terms of modification of the incident flow and impact force applied to the obstacle are of 

paramount importance for the conception and design of structural countermeasures. In this 

context, the present study aims at determining local values of the flow velocity and pressure 

in the vicinity of a structure subjected to a debris-flow impact, as well as the changes in these 

variables over time and space, for a given incident flow. We use a 2D vertical numerical 

model based upon the SPH (smoothed particles hydrodynamics) numerical method. SPH is a 

particular method of treatment of fluid mechanics equations which is suitable for computing 

highly transitory free surface flows of complex fluids (viscoplastic, granular, etc.) in complex 

geometries. Thus, it is suitable for the treatment of debris-flow waves – structure interactions. 

We present our numerical experiments setup aiming at simulating a viscoplastic fluid wave 

impacting a simple obstacle. We notably focus interest on the pressures developed on the 

structure during the impact, on their evolution with the incident flow features, and on their 

spatial distribution over the obstacle. We also analyze our results in terms of physical 

phenomena explaining the observed pressures. 

1 Irstea, Grenoble, FRANCE, Dominique, dominique.laigle@irstea.fr 

2 Irstea, UR ETGR Snow avalanche and torrent control, Grenoble center, F-38402 St-Martin-d’Hères, France Univ. F-38041 Grenoble, 

France 



METHOD 

The SPH method 

The SPH (smoothed particle hydrodynamics) method is a mesh-free numerical method 

initially introduced by Gingold and Monaghan [1977]. It has since found many applications 

in fluid dynamics. Because of its mesh-free nature, it can handle large deformations of the 

simulated fluid. It also handles free surfaces naturally. This makes this method particularly 

well-suited for the simulation of the propagation of mudflows [Rodriguez-Paz and Bonet, 

2004], granular materials [Chambon et al., 2011] and their impact on a wall [Laigle et al., 

2007; Huang et al., 2011b]. 

Virtual experiments layout 

Our simulation domain (Figure 1) is derived from the experiments described in Tiberghien et 

al. [2007]. The experimental setup was designed to generate unsteady flows constituted of a 

steep front followed by a steady flow of constant depth h. It consisted of a 4 m long 30 cm 

wide flume with an inclination angle ranging from 0° to 10°. A tank was located upstream of 

the flume and stored the viscoplastic sample material. This tank was closed by a gate which 

could be instantaneously removed to release the fluid. A rectangular obstacle of height H was 

located downstream of the flume. The height H was adjusted from one experiment to another 

so as to maintain a constant aspect ratio h/H = 0.86. This value was arbitrarily chosen but is 

the order of magnitude of the ratio encountered in the field when considering real debris 

flows and protection structures. Simulations were carried out using an improved version of 

the SPH code presented in Laigle et al. [2007]. The fluid was modelled using about 60000 to 

80000 particles, resulting in about 20 particles along the vertical axis in the sheared region of 

the fluid. Each simulation was run on four cores on dual Xeon computers. The computation 

time for a single simulation ranged from one to two weeks depending on the number of SPH 

particles. 

Figure 1: Snapshot of a simulation showing the main features and dimensions of the virtual experiment. 


Our objective is to study the evolution of the local flow features in the vicinity of the obstacle 

for several values of the inclination angle. All other parameters of the virtual experiments 

(rheological parameters, height of the gate, and fluid level in the tank) are kept constant. This 

way, each inclination angle is related to a unique incident flow. In order to get a comparative 

criterion between numerical experiments and field flows, the following results will be 

presented not only on the basis of the inclination angle. We’ll indicate also the value of the 

Froude number of the incident flow. We adopt here the non-classical definition proposed by 

Tiberghien et al. [2007] in the framework of their experiments: 

Formula 1: definition of the Froude number adopted in the framework of the present study 

where ufront is not the fluid velocity but an average velocity of the front between the gate 

and the obstacle. h is the depth of the steady flow following the front propagation and θ is 

the inclination angle. The studied inclination angles in the simulations range from 3° to 16°, 

which corresponds to Froude number values ranging from 0.52 to 2.88, thus of the same 

order of magnitude as field debris flows. 

Virtual pressure sensors 

One of the goals of this work is to compute the pressure of the fluid in the vicinity of an 

obstacle. To this effect, we use virtual sensors [Laigle et al., 2007]. These sensors (Figure 2) 

are constituted of rectangular areas of the simulation domain over which various properties 

of the fluid are recorded. These properties include the average pressure, velocity, density and 

depth of the fluid. This is the method we use to compute the pressure on the wall, either at a 

given location (in which case the height of the sensor is equal to 1 cm, according to the flume 

experiments by Tiberghien et al. [2007]), or on the whole obstacle (the vertical extent of the 

sensor matches the height of the wall). 

Figure 2: location of the virtual pressure sensors off the upstream face of the obstacle 


RESULTS 

Preliminary results 

After Labbé [2015], the observation of general flow features during the impact shows that for 

gentler slopes, between 3° and 7°, the fluid level goes up gradually along the upstream wall of 

the obstacle and finally overflows with permanent contact with the upper face of the obstacle. 

A surface wave looking like a hydraulic bore propagates upstream and a roughly triangular 

zone of fluid at rest, we name the dead-zone, forms upstream of the obstacle. These observations 

are coherent with previous experimental results obtained by Tiberghien et al. [2007] 

among others. For steeper slopes, above 8°, a vertical jet forms immediately when the fluid 

impacts the obstacle. A dead-zone forms upstream of the obstacle but it is much smaller than 

at gentle slopes. After the first impact a jet is still present over the obstacle and its angle with 

the horizontal direction diminishes with time and stabilizes when the steady regime is 

reached. The jet presents no contact with the upper face of the obstacle. This kind of impact 

will be referred to as the jet impact regime. At intermediate inclination angles, between 7° 

and 8°, we observe a transition. While at gentler slopes the size of the dead-zone gets bigger 

when the inclination reduces, its size is almost constant at steeper slopes. 

Time evolution of the pressure on the obstacle 

In this section, we aim at examining the pressure applied by the fluid to the upstream wall of 

the obstacle. The height of the sensor is identical to the obstacle height. Figure 3 shows the 

evolution of the pressure versus time for a series of simulations carried out with an increasing 

flume inclination. We note that a pressure peak is systematically present when the fluid 

impacts the obstacle as well as a pressure plateau when the steady regime is established. The 

respective values of these pressures strongly evolve with the flume inclination. On gentle 

slopes, in the dead-zone regime, a small peak with duration about 0.01 s is observed at the 

impact followed by a slow increase of the pressure lasting a few seconds. Between 3° and 6°, 

the first pressure peak, whose intensity is lower than the pressure of the plateau, grows up 

with increasing inclination. In parallel, the pressure of the plateau diminishes with increasing 

inclination. From 7° onwards, the peak intensity overcomes the pressure of the plateau. The 

local minimum in the curve connecting the peak to the plateau disappears at 8°. The duration 

of the peak increases with slope to reach values about 0.1 s. 

Figure 3: Evolution of the simulated pressure applied to the upstream wall of the obstacle versus time for several values of the virtual 

flume inclination angle 


These results, combined with those of previous section, evidence two different impact regimes 

(dead-zone regime / jet regime). For the first one the maximum pressure is reached during 

the plateau when the steady regime is established. For the second one the maximum pressure 

is reached at the peak which immediately follows the impact. The transition between these 

regimes is observed between 6° (Fr = 1.12) and 7° (Fr = 1.31). 

Maximum pressures versus Froude number of the incident flow 

In this section we present the maximum pressures recorded by 4 sensors: 3 sensors 1 cm high 

located at the bottom, centre and top of the upstream face of the obstacle, and 1 global sensor 

covering all the obstacle height. These data are analyzed in reference to the Froude number 

(Figure 4). Between Fr = 0.5 en Fr = 1.2 pressures do not vary much for each sensor taken 

individually and curves are almost parallel. Higher pressures are observed at the bottom and 

pressures observed at the centre and global sensors are very similar. Between Fr = 1.2 and Fr 

= 1.6, the global sensor, bottom sensor, centre sensor and upper sensor respectively show a 

transition. Beyond a Froude value Fr = 1.6, the pressures increase almost linearly with the 

Froude number and the curves diverge. The pressure is systematically higher at the bottom 

and lower at the top of the obstacle. Once again, this evidences a transition in the impact 

regime with some additional information: the transition slightly depends on the position on 

the obstacle. 

Figure 4: Maximum pressure Pmax recorded by the sensors versus Froude number 

Analysis in terms of gravity and kinetic pressures 

In order to better understand the physics of the impact and to try to deduce rules for a better 

estimation of impact pressures in the field, we analyse here the respective contributions of 

the gravity pressure Pg and kinetic pressure Pkin. The kinetic pressure is given by: 

Formula 2 


Where u is the flow velocity, ρ the material density and β is a correcting coefficient which 

value is close to unity. The gravity pressure is given by: 

Formula 3 

These formulas constitute the basis of the models of evaluation of debris-flow impact pressure 

which can be found in the literature according to a synthesis proposed by Proske et al. (2011). 

In this section, we consider the maximum mean pressure applied to the whole obstacle 

Pmax,global (computed using the global sensor) but also the plateau pressure Pplateau 

(corresponds to the maximum pressure only in the dead-zone regime) and the peak pressure 

Pp (corresponds to the maximum pressure only in the jet regime). 

One can see in Figure 5 that the ratio Pplateau/Pg varies gently and almost linearly with the 

Froude number. For values of the Froude number higher than 1.2, the Pplateau and the 

Pmax,global curves diverge and this latter no longer varies linearly with the gravity pressure 

Pg. For Froude number values lower than 1.2, the ratio Pmax,global/Pg takes values from 2.8 

to 3.5. 

Figure 6 shows that the ratio Pp/Pkin varies quite quickly with the Froude number when the 

latter is lower than 1.0 and that the maximum pressure value Pmax,global is higher than the 

kinetic pressure Pkin. For values of the Froude number higher than 1.2, the Pp and Pmax- 

,global curves superpose and the ratio Pmax,global /Pkin is almost constant and equals 2. This 

value is coherent with the theoretical case of a perfect fluid impacting a plate perpendicular to 

the flow direction. In their synthesis, Proske et al. [2011] note that authors who adopt the 

kinetic model also propose values of the correcting coefficient from 2.0 for fine slurries to 5.0 

for coarse materials. 

Figure 5: Maximum pressure recorded by the global sensor to gravity pressure ratio Pmax, global/Pg and plateau pressure to gravity 

pressure ratio Pplateau/Pg versus Froude number 


Figure 6: Maximum pressure recorded by the global sensor to kinetic pressure ratio Pmax, global/Pkin and peak pressure to kinetic 

pressure ratio Pp/Pkin versus Froude number 

Spatial distribution of the pressure 

In this section we investigate the spatial distribution of the pressure applied to the upstream 

face of the obstacle and its evolution with time considering the impact regime. Whatever the 

regime, we can observe in Figures 7 and 8 that the first impact takes place at the bottom of 

the obstacle and the pressure spreads afterwards on the whole obstacle. After a sufficient 

time, the pressure profile is linear meaning that the plateau is reached and that the gravity 

pressure is the only contribution at work. One can see in Figure 7 that in the dead-zone 

regime, the first impact is not very strong and the maximum pressure at any point of the 

obstacle is clearly observed when the plateau pressure is reached. On the contrary, one can 

see in Figure 8 that in the jet regime, the first impact is very strong and the maximum 

pressure, at least at the bottom of the obstacle, is observed during the first impact phase. 

Figure 7: Distribution of the pressure applied to the upstream face of the obstacle at different times of the simulation carried out with Fr 

= 0.71, θ = 4° (first time of contact is at 3.187 s) 


Figure 8: Distribution of the pressure applied to the upstream face of the obstacle at different times of the simulation carried out with Fr 

= 2.22, θ = 12° (first time of contact is at 1.354 s) 

CONCLUSIONS 

The present study aimed at determining local values of the pressure in the vicinity of a 

structure subjected to a debris-flow impact, as well as its evolution in time and space, for a 

given incident flow. To carry out this study, we used a numerical model based upon the SPH 

numerical method. After briefly presenting SPH, we introduced our numerical experiments 

setup. In the presence of an obstacle, we simulated the local flows upstream of the obstacle at 

several times of the impact of a viscoplastic fluid wave. We focused interest on the time 

evolution and spatial distribution of pressures developed on the structure during the impact. 

The results were analyzed in view of the features of the incident flow characterized by its 

Froude number. We evidenced the existence of 2 impact regimes respectively named 

dead-zone and impact regime with a transition for Froude number values around 1.3 to 1.5. 

We evidenced the respective contributions of the gravity and kinetic pressures during the 

impact phase, depending upon the impact regime. We finally analyzed the spatial distribution 

of the pressure on the obstacle depending on the impact regime. 

REFERENCES 

- Chambon, G., Bouvarel, R., Laigle, D. and Naaim, M. (2011): Numerical simulations of 

granular free-surface flows using smoothed particle hydrodynamics. Journal of Non-Newtonian 

Fluid Mechanics, 166(12-13): 698-712. 

- Gingold, R. and Monaghan, J. (1977): Smoothed particle hydrodynamics-theory and 

application to non-spherical stars. Monthly Notices of the Royal Astronomical Society, 181: 

375-389. ISSN 0035-8711. 

- Huang, Y., Dai, Z., Zhang, W. and Chen, Z. (2011): Visual simulation of landslide fluidized 

movement based on smoothed particle hydrodynamics. Natural Hazards, 59(3): 1225-1238. 

ISSN 0921-030X. 

- Labbé, M. (2015) : Modélisation numérique de l’interaction d’un écoulement de fluide 

viscoplastique avec un obstacle rigide par la méthode SPH. Application aux laves torrentielles. 

PhD thesis. Université Grenoble Alpes. (in French) 


- Laigle, D., Lachamp, P. and Naaim, M. (2007): SPH-based numerical investigation of 

mudflow and other complex fluid flow interactions with structures. Computational Geosciences, 

11(4): 297-306. ISSN 1420-0597. 

- Proske, D., Suda, J., and Hübl, J., (2011): Debris flow impact estimation for breakers. 

Georisk, 5(2): 143-155. 

- Rodriguez-Paz, M. and Bonet, J. (2004): A corrected smooth particle hydrodynamics 

method for the simulation of debris flows. Numerical Methods for Partial Differential 

Equations, 20(1): 140-163. ISSN 1098-2426. 

- Tiberghien, D., Laigle, D., Naaim, M., Thibert, E. and Ousset, F. (2007): Experimental 

investigations of interaction between mudflow and an obstacle. In C. Chen and J. J. Major 

Eds., Proceedings of the International Conference on Debris-Flow Hazards Mitigation: 

Mechanics, Prediction, and Assessment, Chengdu, China. 281-292. 



People and buildings vulnerability to floods in 

mountain areas 

Marco Pilotti, Ph.D. 1 ; Luca Milanesi, Ph.D. 1 ; Roberto Ranzi, Ph.D. 1 

ABSTRACT 

Risk assessment needs rational and quantitative methodologies to describe hazard, vulnerability 

and exposure. This paper focuses on physical vulnerability functions based on mechanical 

schemes representing human stability and building structural resistance in a flow. These 

instruments were used to model the Gleno dam break event that occurred in 1923 in Valle di 

Scalve (Northern Italy). An innovative modelling approach adapting the domain bathymetry 

to the potential building collapse is proposed and positively tested. 

KEYWORDS 

2D model; Buildings vulnerability; Dam break; Flood mapping; People's vulnerability 

INTRODUCTION 

Hydraulic risk is a combination of hazard, exposure and vulnerability. Accordingly, quantitative 

procedures are needed to calculate the flow features, to identify the exposed targets 

(e.g., people, buildings, cars, etc.) and define the potential damage associated to each flow 

condition. Only this rational roadmap provides objective results so that the related legislative 

planning instruments can be understood, accepted and shared among stakeholders. The 

criteria used to assess exposure and vulnerability are still poorly discussed and this study 

focuses on people’s and buildings vulnerability to floods. 

Several models of people’s vulnerability to floods are available in the literature from both 

conceptual and experimental studies (e.g., Xia et al., 2014). They are useful to estimate the 

potential loss of life due to floods (Jonkman et al., 2008). Anyway, these models are mostly 

empirical and a satisfactory treatment of physical aspects such as the role of local slope and 

fluid density is still missing. Milanesi et al. (2015) proposed a weakly parametric conceptual 

model that includes a detailed treatment of these issues. The model, briefly summarized in 

the following, best matches the available experimental data. 

Although the most common method for the estimation of direct damage to buildings is still 

the application of stage-damage functions (Jongman et al., 2012), structural vulnerability 

models are available (e.g., Clausen & Clark, 1990) and worth to be investigated . The current 

study provides a conceptual scheme comparing the actions exerted by the flow on a simplified 

masonry building with the resistance of the structure itself, considering the potential 

failure mechanisms of a partly confined wall. 

1 Università degli Studi di Brescia, ITALY, luca.milanesi@unibs.it 



These criteria have been applied to the Gleno dam break event which occurred in 1923 with 

catastrophic consequences in Valle di Scalve, a right tributary of Valle Camonica, drained 

respectively by the Dezzo and Oglio rivers. The flood took about 45 min to flush the 21 

km-long stretch from the dam to the confluence of the rivers at Corna (Darfo). Pilotti et al. 

(2011) studied the event by a 1D model of the Dezzo river as far as the alluvial fan of Corna. 

Here a 2D code has been applied to model the flood on the fan, including a dynamic 

procedure accounting for building collapse. The preliminary results of the model are 

compared to observed data derived from historical documents. 

PEOPLE’S VULNERABILITY TO FLOODS 

As proposed by Lind et al. (2004), the human body is represented by a set of cylinders in 

vertical position on an inclined slope (Fig. 1) and impacted frontally by a steady uniform flow. 

The body weight W is decomposed in direction normal W N 

and parallel W P 

to the slope. 

Considering the pressure distribution of a parallel flow, the buoyancy force B N 

is normal to 

the bed. Since the body stands in vertical position on the slope, it is not normal to the flow 

field. Neglecting skin friction, the fluid dynamic force R is normal to the body frontal area and 

it can be decomposed in direction parallel and normal to the slope, giving drag D and lift L 

forces respectively. Friction T between the soles and the bed is the product of the coefficient μ 

(0.46, after calibration) and the effective weight w that is the algebraic sum of the forces 

normal to the slope: W N 

, B N 

and L 

Figure 1: Acting forces and their application points: (a) weight, (b) dynamic actions, and (c) buoyant and friction forces. Partial lateral 

view of the body. See Milanesi et al. (2015) for details on the symbols. 

Slipping instability occurs if the sum of the drag force and of the component of weight W P 

overcomes friction. Toppling instability occurs if the moment calculated with respect to the 

pivot point P, in this case the heel, of the component of the weight W N 

is exceeded by the 

destabilizing moments due to lift, drag, buoyancy and the component of the weight W P 

. A 

third condition is introduced to consider drowning by imposing a maximum admissible water 

depth as a function of the height of the neck. The depth safety limit, as a function of the flow 

velocity U, is given by the minimum among slipping, toppling and drowning depths (Fig. 2). 


Figure 2; Stability curves for adults (m=71 kg, Y a 

=1.71 m, thin line) and children (m=22.4 kg, Y c 

=1.21 m, thick line): (a) slipping (b) 

toppling (c) drowning; combined in Fig. (d) (ρ=1000 kg/m 3 and =0°). A, B, and C influence areas in (d) represent respectively the 

drowning, toppling and slipping controlled areas for children. Fig. (e) shows literature experimental data and the curves by Xia et al. 

(2014) (dashed). High, medium and low vulnerability are respectively identified by red, orange and yellow. See Milanesi et. al. (2015) 

for details on the cited studies. 

BUILDINGS STRUCTURAL VULNERABILITY TO FLOODS 

Alpine traditional masonry buildings are modelled using a simple structure with weight-bearing 

walls. Accordingly, it is subdivided in elementary modules independent from each other, 

whose stability is assessed with respect to the wall impacted by the flow. Under the action of 

a horizontal acceleration normal to the wall, the wall flexes out-of-plane, rotating around a 

horizontal or vertical joint, depending on the constraint exerted by the surrounding walls. 

The loading force F per unit width is made of a hydrostatic F st 

and a dynamic F d 

term, the 

latter amplified to represent the impulsive behaviour of the impact (Cross, 1967). 


Figure 3: Scheme of the building (a) and detail of the impacted wall with tracks of the failure joints (b). Horizontal (c) and vertical (d) 

joint failure mechanisms. 

The first failure mechanism is modelled considering a vertical strip of the wall of unit width, 

thickness t, and height Y as a simply supported beam loaded by a trapezoidal pressure 

distribution. It fails along a horizontal joint of thickness a (x-x’, Fig. 3b) if the maximum 

moment M max-h 

caused by the flow pressure is greater than the stabilizing moment M u-h 

due 

to the vertical loads N. 

MM mmmmmm−h ≥ MM uu−h → MM mmmmmm−h ≥ NN ( tt 2 − mm 2 ) 

Formula 1 

The second failure mechanism is modelled by considering the portion of the wall of height h 

and width L as a laterally simply supported beam. The pressure distribution is uniform and 

the stability is provided by a resisting arch that transfers the load F 0 

on the sidewalls and by 

the friction force per unit width F μ 

at the interface at height h. The vertical joint (y-y’, Fig. 3b) 

triggers if the maximum moment on the beam M max-v 

is greater than the resisting moment 

M u-v 

. 

MM mmmmmm−vv ≥ MM uu−vv → MM mmmmmm−vv ≥ FF oo (tt − mm ) + FF LL 2 

2 

μμ 

8 

Formula 2 


The limiting stability condition is represented by the horizontal joint mechanism. 

Figure 4: Stability thresholds associated to the described mechanisms calculated for a 2.5 storey building with walls 0.45 m thick. 

THE GLENO DAM BREAK TEST CASE 

The Digital Terrain Model (DTM) of the domain was derived from LIDAR data, filtered from 

buildings and vegetation, with 1 m resolution (Source: Italian Ministry of the Environment 

and Protection of Land and Sea). It was averaged at uniform sized cells of 7.5 m in order to 

represent buildings and limit the computational effort. The urban area and the path of the 

Dezzo river in 1923 were reconstructed through historical maps. Because of the lack of 

quantitative and reliable information regarding the original geometry of the Dezzo river, a 

rectangular approximation was assumed. The original path was derived from historic maps. 

Figure 5: a) Location of the study area. b) Input hydrograph (Pilotti et al., 2011). c) Discretization of the hydrograph for the dynamic 

approach. 


The code used to compute local flow depth and velocity (FLO-2D) solves the full two 

dimensional (2D) shallow water equations (SWE) on a Cartesian grid using an explicit, 

central, finite difference numerical scheme along eight flow directions. This software, already 

applied successfully to the dam break study of Cancano (Pilotti el al., 2014), was chosen also 

for its compliance with the requirements of the Federal Emergency Management Agency 

(FEMA) in USA about hydraulic modelling. 

Urban flooded domains have uneven roughness and traditional a priori estimates of roughness 

are not feasible in case of impulsive flood waves affecting significantly the domain 

morphology and roughness. Accordingly, roughness was assumed constant and calibrated 

(k s 

=30 m 1 /3 /s) with reference to the depth observed in several points of the domain. 

It was shown (e.g., Soares-Frazão & Zech, 2008) that the presence of buildings as well as their 

alignment influence the flow field and the flood extent. This is especially true as far as 

impulsive floods are concerned since their destructive impact on buildings can dynamically 

alter the bathymetry. Accordingly, in the following two different modelling approaches will 

be compared: a static one, keeping the buildings in their initial position during the entire 

simulation based on the hydrograph in Fig. 5b, and a dynamic one, provided by a series of 

Figure 6: Differences in maximum flow depth (a) and velocity (b). Blue indicates that the dynamic simulation provides values greater 

than the ones from the static model; red colour indicates the opposite situation. White indicates substantial equality between the results 

of the two approaches. GREEN: flooded buildings; YELLOW: submerged buildings; ORANGE: partly destroyed buildings; RED: destroyed 

buildings. 


Figure 7: (a) Validation of the structural vulnerability model; the red dots indicate where, according to the presented model (section 2), 

buildings would be destroyed by the flow in the dynamic approach. Risk to people maps from the static (b) and dynamic (c) approaches. 

simulations of partial hydrographs (Fig. 5c). At the end of each partial simulation, the 

buildings destroyed according to the structural vulnerability model are removed from the 

domain and a roughness value k s 

=10 m 1/3 /s is assigned to the area previously covered by the 

structure to represent the presence of debris. The following simulation is then run from the 

beginning of the hydrograph with the updated bathymetry. 

RESULTS AND CONCLUSIONS 

The main difference between the two simulations is related to the lateral spread of the flow 

due to the interaction with the buildings (Fig. 6). The static simulation estimates greater 

depths and velocities in the lateral areas of the fan because of the diffusion caused by 

buildings. On the contrary, with the dynamic approach the flow is more concentrated along 

the centre of the domain because of the gradual removal of the destroyed structures based on 

the vulnerability model, whose reliability is demonstrated in Fig. 7a. The comparison of the 

estimated flow depths with data from historical images proves that the dynamic approach is 

more accurate than the static one (RMSE=0.69 m and 0.78 m respectively). Risk was 

classified by the vulnerability criterion for people, whose exposition was considered unitary. 

As usual in extreme events, Figs. 7b and 7c show a strong polarization toward high risk areas. 

The main advantages of the model of people vulnerability is the direct dependency of the 

thresholds from fluid density and local slope that are key parameters especially in mountain 

areas. Similarly, on the contrary of empirical formulae (e.g., Clausen & Clark, 1990), the 

physically based model of buildings structural vulnerability can be adapted to different 

structural typologies. Finally, the dynamic modelling of extreme events allows a more realistic 

hazard assessment and an accurate estimate of the consequences on structures. The simple 

physically based procedures presented in this study allow a reliable reproduction of flood 


events and a more objective representation of risk that might guarantee comprehensibility 

and acceptability of the related plans and constraints to the stakeholders. 

REFERENCES 

- Clausen L., Clark P. (1990). The development of criteria for predicting dam break flood 

damages using modeling of historical dam failures. Proc. Int. Conf. on River Flood Hydraulics, 

W. White (ed.), Wiley, U.K, 369–380. 

- Cross, R.H. (1967). Tsunami Surge Forces. Journal of the Waterways and Harbors Division 

93(4): 201-231. 

- Jongman B., Kreibich H., Apel H., Barredo J. I., Bates P. D., Feyen L., Gericke A., Neal J., 

Aerts J.C.J.H., Ward P.J. (2012). Comparative flood damage model assessment: towards 

a European Approach. Nat. Hazards Earth Sys. 12: 3733–3752. 

- Jonkman S.N., Vrijling J.K., Vrouwenvelder A.C.W.M. (2008). Methods for the estimation 

of loss of life due to floods: a literature review and a proposal for a new method. Nat Hazards 

46: 353-389. 

- Lind N., Harftord D., Assaf H. (2004). Hydrodynamic models of human stability in a flood, 

J. Am. Water Resour. As. 40(1): 89–96. 

- Milanesi L., Pilotti M., Ranzi R. (2015). A conceptual model of people’s vulnerability to 

floods. Water Resour. Res. 51(1):182-197. 

- Pilotti M., Maranzoni A., Tomirotti M., Valerio G. (2011). The 1923 Gleno dam-break: 

case study and numerical modelling. J. Hydraul. Eng. 137(4): 480-492. 

- Pilotti M., Maranzoni A., Milanesi L., Tomirotti M., Valerio G. (2014). Dam-break modeling 

in alpine valleys. J. Mt. Sci. 11(6): 1429-1441. 

- Soares-Frazão S., Zech Y. (2008). Dam-break flow through an idealized city. J. Hydraul. 

Res. 46(5): 648-658. 

- Xia J., Falconer R.A., Wang Y., Xiao X. (2014). New criterion for the stability of a human 

body in floodwaters. J. Hydraul. Res. 52(1): 93-104. 



Modern flood protection and rehabilitation concepts 

at pre-alpine alluvial rivers 

Michael Müller, Ph.D. 1 ; Peter Billeter, Ph.D. 1 ; Matthias Mende, Ph.D. 1 ; Manuel Zahno, MSc. 1 ; Adrian Fahrni 2 

ABSTRACT 

In Switzerland, the big pre-alpine alluvial river corridors are embedded in urban, industrial, 

agricultural or recreational zones. Planning and construction of flood protection and 

restoration measures on these rivers require consideration of various interests and is strongly 

influenced by restrictive constraints in time and space. 

Three examples of flood protection projects reveal challenges met during planning and 

construction stages. All measures presented were subject to various actors' opinion and 

implemented under particular site conditions. A large flood plain was exploited for widening 

the Aare River upstream of Berne, enhancing wetlands of national interest. Residential and 

industrial zones at the Kleine Emme River at Littau only allowed increasing flood capacity 

by heightening the cross section. However, the ecological and recreational environment was 

enhanced by instream restoration measures such as flow structures, low flow channels as 

well as micro groins. The Linth Channel project included both river widening and instream 

restoration. In each case, excess flood evacuation concepts guarantee the functionality of 

implemented protection structures under even more severe flow conditions. 

KEYWORDS 

Flood control; river restoration; river widening; instream river training; excess flood evacuation 

INTRODUCTION 

Today's flood protection measures aim at controlled and safe run-off of a specific design flood 

on one hand, and an enhancement of ecological conditions and fish habitats on the other 

hand. Thus, if on a given river reach civil engineering works for flood control are planned a 

detailed analysis of possible restoration measures has to be carried out simultaneously. If local 

conditions allow, river widenings present a very suitable solution for both flood protection 

and ecological diversification. If widening cannot be considered due to densely built or 

sensitive surroundings, specific local measures in the river section have to be considered such 

as flow structures created by instream restoration measures, low flow channels and micro 

groins. This is in accordance with the Federal Act on the Protection of Waters which defines 

the legal context for planning and construction on Swiss rivers. 

1 IUB Engineering AG, Belpstrasse 48, CH-3007 Bern, SWITZERLAND, michael.mueller@iub-ag.ch 

2 Tiefbauamt Kanton Bern, OIK II, Schermenweg 11, CH-3001 Bern, SWITZERLAND 



The environment in the Swiss Midland is dominated by urban areas and most of the big prealpine 

alluvial river corridors are embedded in urban, industrial, agricultural or recreational 

zones. Consequently, if a certain river reach is planned to be enhanced by flood protection 

and restoration measures, all these different parties become somehow involved in the project. 

Thus, during planning and construction phases of flood control measures for Swiss pre-alpine 

alluvial rivers, a big variety of actors, different and sometimes contradictory interests as well 

as multiple constraints in time and space have to be considered. Solutions and compromises 

have to be found to reduce flood damages, improve river habitats and, after all, achieve 

public acceptance and satisfaction of the involved partners. 

According to Swiss standards the design flood discharge Q dim 

depends on the potential of 

damage. For urban areas or relevant infrastructure the discharge Q dim 

often has annual 

probability of P = 0.01 and the dam heights account for a safety margin of typically 0.5 m 

to 1.0 m with regard to the water level at design discharge. In any case and in addition to 

a solution for the design flood, an evacuation concept has to be proposed for the case of an 

excess flood event. For such events either extreme discharge with an annual probability of 

occurrence of P < 2‰ (corresponding to a 500-years return period), extreme bedload or 

extreme driftwood transport are assumed. The excess flood evacuation concept includes the 

definition of an emergency and alerting concept as well as constructive measures to assure 

the functioning of the flood protection measure during the extreme event, for example by 

a controlled flooding of less populated river corridors. 

The paper focuses on two basically different approaches that can be adopted regarding flood 

protection measures and river restoration - solutions with and without river widening - based 

on three practical examples of executed or planned flood protection works on the Swiss Rivers 

Aare, Kleine Emme and Linth which were or are to be carried out under challenging site 

conditions and particular constraints. Furthermore, concepts and specific solutions to handle 

the subject of excess flood control are discussed. 

FLOOD CONTROL INVOLVING RIVER WIDENING AND RESTORATION 

In the beginning of the 20th century, flood protection concepts often included straightening 

and channelizing of river reaches. Nowadays and based on new laws on water protection and 

hydraulic engineering the idea consists in giving back terrain to the rivers, allowing them 

to regain their meandering or braided character and to flood specific areas where a certain 

degree of damage can be accepted. The challenge consists in controlling the effects of the 

morphological processes to reach a sustainable situation for both nature and civilization. 

Constructive measures include protection of eroding banks, stabilization and structuring 

of the cross section invert as well as increasing flow diversity to reach optimum ecological 

conditions. 


Over the last two decades, several flood events of the Aare River affected the Selhofen- 

Zopfe region (Switzerland, Figure 1a), a zone involving agriculture, drinking water supply, 

protected wetlands of national interest as well as a small international airport. The flood 

event in 1999, when the discharge of the Aare river reached about Q = 610 m 3 /s corresponding 

to a 500-years-flood, caused considerable damages to the local communities and 

infrastructure. In addition, the existing levee as well as the concrete groin structures built in 

the early 20th century were in need of reconstruction. Therefore, a flood mitigation project 

was launched by the Canton of Berne. Various actors and interests had to be considered 

during planning phase; on the left river bank, flood protection measures had to be compatible 

with zones of important drinking water wells for the city of Berne as well as a recreational 

zone including the protected wetlands. The right river bank also hosts two drinking water 

wells, a very popular walking path along the river as well as a public swimming pool. 

The flood mitigation measure thus had to satisfy multiple technical, environmental and 

also public requirements. 

On the outer bank, the existing damaged longitudinal concrete groins (Figure 1b) were 

dismantled and replaced by riprap to increase the river cross section. Three new micro groins 

were built to guide the main flow back into the center of the cross section to increase flow 

diversity and to prevent outer bank erosion (Figure 1c). This rough and permeable waterfront 

protection allows both stabilizing the outer bend and protecting the drinking water supply 

zones. The excavation of a new secondary channel (Figure 1d) and the implementation of 

new longitudinal gravel bars enhance the natural fluvial environment and assure the 

conservation of the attractive walking path with direct access to the river for the local 

population. Punctually, flood-proofing measures were taken to protect existing infrastructure 

Figure 1. Overview of flood mitigation and restoration measures on the Aare River upstream of Berne, b) old longitudinal concrete groins 

and c) new enhanced erosion bank protection with micro groins and gravel bank, d) new natural secondary channel, inner bank widening 

during normal (e) and flood conditions (f), and g) enhanced Giesse Channel. 


Figure 1b: Flood mitigation and restoration project Aare River: 

right bank initial state 

Figure 1c: Flood mitigation and restoration project Aare River: 

right bank enhanced state 

Figure 1d: Flood mitigation and restoration project Aare River: 

right bank natural secondary channel 

Figure 1e: Flood mitigation and restoration project Aare River: left 

bank widening normal discharge 

Figure 1f: Flood mitigation and restoration project Aare River: left 

bank widening flood discharge 

Figure 1g: Flood mitigation and restoration project Aare River: left 

bank natural Giesse Channel 

such as a small restaurant on the river side or the already enhanced confluence between the 

Gürbe and the Aare River. 

The existing flood protection levees on the inner bank were removed and rebuilt at the very 

possible limits of the most sensitive drinking water zone. Such, the river could be widened 

by up to 200 m (Figure 1e and f). Stream barbs along the new levee as well as small stream 

islands form a rough and permeable element against erosion. Within the widened zone the 

so called Giesse Channel, a small side channel fed by groundwater, was also partially moved 

and given a new meandering course with instream structures for fish shelter and wooden 

protection at the outer bank (Figure 1g). Most of the construction works, clearing and truck 


transports took place in the nature conservation area and were therefore restricted to a 

very minimum. 

The flood mitigation works were completed in June 2015. One month earlier, in beginning of 

May 2015, the passage of a flood with Q max 

= 510 m 3 /s, corresponding to a 50-years return 

period, confirmed the adequate functioning of the almost terminated flood protection system. 

The Linth Channel between Lake Walen and Lake Zürich presented insufficient flood 

capacity, deficits regarding excess flood evacuation possibilities as well as damaged bank 

protection. Therefore, the Linth Channel flood protection project including a series of flood 

mitigation and restoration measures was implemented (Billeter and Keller 2013). The 

historical Glaner Linth was meandering directly to Lake Zurich before being deviated to Lake 

Walen by Conrad Escher at the beginning of the 19th century with the purpose of flood 

protection. This major river correction resulted in a deficit of natural alluvial banks in the 

so-called Escher Channel. At Klein Gäsischachen, where topography and situation allowed 

for some flexibility regarding the flood protection along the right river bank, it was decided to 

remove the right levee partially in order to achieve a local widening that re-substitutes the 

alluvial bank structures lost by the 19th century river correction. During construction works, 

the water was led through a previously excavated diversion channel which served as initial 

secondary channel of the widening section once the existing right flood protection levee was 

removed and initial breaches allowed flooding of the widening area. 

Figure 2: River widening at Klein Gäsischachen (Lindth Channel), a) picture taken during physical modelling tests, b) result from 

physical modelling tests: river morphology after passage of a flood event (river bed erosion = positive values, redish colours, sediment 

deposition = negative values, blueish colours), and c) protoype behavior. 


The development of river morphology was studied in hydraulic modelling tests at VAW, ETHZ 

(Figures 2a and b) and the same behavior could be observed under prototype conditions 

during flood events occurring relatively shortly after the completion of construction works 

(Figure 2c). 

At Hänggelgiessen, a local widening was realized aiming the creation of a new regularly 

flooded wetland, including a natural secondary channel on one hand and offering a small 

retention volume of 1.0 Mio. m 3 on the other hand (Figure 4a). In this particular case, the 

river widening was thus combined with an excess flood evacuation system, presented later 

on in this paper. 

FLOOD CONTROL AND RESTORATION WITHIN A GIVEN RIVER WIDTH 

In cases where the local environment is either densely built or exploited in a way that no 

river widening can be realized, other constructive measures have to be considered to increase 

flood capacity and flow diversity in the given reach. In August 2005, the Kleine Emme River 

which is characterized by a mean annual discharge of 16 m 3 /s carried some Q = 650 m 3 /s in 

combination with a lot of sediment and driftwood. This flood event corresponding to a 

70-years return period caused inundations in the adjacent zones with damages estimated 

to about 65 Mio. CHF to hydraulic structures and about 200 Mio. CHF to buildings in the 

vicinity of the river. Therefore, planning of flood mitigation measures was started by the 

Canton of Lucerne with the objective of both sustainable flood protection and river restoration. 

Today, the flood protection project is in authorization phase and includes various 

measures along some 23 km of the Kleine Emme River. River widening as presented in the 

preceding paragraph is also planned in less populated areas, but upstream of the confluence 

with the Reuss River at Littau-Emmen (Switzerland) residential and industrial zones close 

to the river restrict widening possibilities to a very minimum. 

Consequently, flood capacity in this river reach will be increased by the implementation of 

new dams, side walls and local river bed excavation to heighten the existing trapezoidal 

channel-like cross section. The design flood at the corresponding reach is Q dim 

= 700 m 3 /s 

which equals to a 100-years flood. The width and bank height of the enlarged cross section 

accounts for a safety margin added to water level at design flood allowing a flood up to 1.3 to 

1.5×Q dim 

to discharge without overtopping (Billeter et al. 2014). However, for environmental 

purpose and to meet the requirements of the Federal Act on the Protection of Waters, the 

river bed will be enhanced to improve morphological variability. Instream restoration 

measures such as flow structures, induced low flow channels as well as micro groins aim at 

an increased variability of flow characteristics (flow velocity and water height) and a stabilization 

of the river bed (Mende 2012, Figure 3). At the same time, they allow to enhance fish 

habitat conditions and the overall ecological and recreational environment. 


Figure 3: Instream restoration measures at the Kleine Emme in the urban area of Littau-Emmen, a) schematic drawing, b) example 

of wooden groins at the Kander River (Märkt, Germany, photo courtesy of Erich Linsin), c) example of rootstock structures at the Emme 

River (Switzerland). 

The confluence with the Reuss River, the so called "Reusszopf" is designed with two river 

branches to dissolve the strongly channelized character of the flow. Thanks to the future flow 

diversity and the possibility to bypass the present confluence sill, the two rivers will be 

connected and allow fish migration in both upstream and downstream direction. 

Where the existing width of the Linth Channel had to be maintained two different concepts 

were applied to reach an ecological enhancement of the river banks. Initially, the channel was 

protected by steep and partially solidified riprap (slope 1:1). The restoration project included 

channel reaches with stream barbs on one hand and reaches with shallow banks on the other 

hand, allowing a diversification of the strongly channelized flow conditions (Figure 4a). 

EVACUATION OF EXCESS FLOOD EVENTS 

The planning of flood control is effectuated for a given design discharge. However, some of 

the implemented protection structures must resist even more severe flow conditions, the so 

called excess flood. To guarantee the functionality of the overall flood protection concept at 

a long term, solutions for a secure evacuation of this extreme event have to be planned and 

built. For the examples described above, excess discharge is released into extended flood 

corridors. 


Upstream of the Selhofen-Zopfe area, the flood protection levees are built to a level which 

allows lateral overtopping along a very specific bank reach if the discharge in the Aare River 

is higher than the design flood for the downstream protection structures. The additional 

discharge is evacuated through a specific excess flood corridor protecting the local airport 

from flooding. 

The same method has been applied at various other rivers e.g. the Kleine Emme River and 

the Linth Channel, where excess flood discharge is deviated into a specific corridor by an 

artificially controlled weir structure. Upstream of the widening Hänggelgiessen on the Linth 

Channel (Figure 4a), the enhanced channel presents a capacity of HQ 500years 

= 500 m 3 /s while 

the capacity of the downstream reach is limited to about HQ 300years 

= 420 m 3 /s. The remaining 

80 m 3 /s are discharged over a weir structure into the secondary channel which is normally 

draining the so called Schänner plain. If incidentally the extreme flood event coincides with 

a flood in this area, the proper Schänner plain still presents some topographical depressions 

with low damage potential assuring an additional accumulated retention volume of 

1.0 Mio. m 3 for peak flood routing. 

Figure 4: a) Local widening and excess flood evacuation system at Hänggelgiessen on the Linth Channel, and b) comparison between 

Linth Channel before (above, original cross section with steep bank slopes) and after the realization of the flood control and river 

restoration project (below, shallow banks). 

CONCLUSIONS 

Today, hydraulic and river engineers face two main controversial tendencies when planning 

and realizing flood control and river restoration on big rivers in the Swiss Midland. On one 

hand, the surroundings are often densely built and/or used for residential, industrial or 

agricultural purpose which results in a high damage potential during flood events and thus 

requires intervention. However, possibilities for flood mitigation are restricted due to the little 

space available. On the other hand, the law requires an enhancement of morphological and 

habitat conditions and local people often claim access to the river for recreational use, 

preferring natural river conditions instead of channelized streams. 

Whenever local conditions allow, river widening has proven to be very effective on flood 

mitigation and in connection with the implementation of secondary channels, stream islands 

and micro groins it allows diversifying flow characteristics to a very high level. This results in 


oth habitat enhancement for fauna and public attraction due to a restored natural river 

environment. Flood mitigation measures respecting the ecological aspects of river restoration 

must also be proposed for very confined river reaches. Instream restoration measures present 

very valuable tools to improve general ecological habitat and fish migration conditions and 

can be applied in both small and large channel-like cross sections. However, they will not 

fully restore natural braiding or meandering processes or wetland dynamics. In any of the 

afore given practical examples, specific solutions were found to guarantee a secure design 

flood passage on one hand and a habitat enhancement due to instream structures and low 

water or secondary channels on the other hand. 

REFERENCES 

- Billeter P., Keller Y. (2013). Das Projekt Linthkanal, Proceedings of the Symposium 

"Projekt Hochwasserschutz Linth 2000". June 2013, Linthverwaltung, Lachen, Switzerland. 

pp. 111-122 

- Billeter P., Mende M., Jenzer J. (2014). Flood Characteristics and Flood Protection Concepts 

in the Reuss Catchment Basin. Proceedings of the International Conference on Fluvial 

Hydraulics "River Flow", September 3-5, 2014, Lausanne, Switzerland. 

- Mende M. (2012). Instream River Training - Naturnaher Flussbau mit minimalem 

Materialeinsatz. 

- Korrespondenz Wasserwirtschaft, 5(10): 537–543. 



Bedload trapping in open check dam basins - measurements 

of flow velocities and depositions patterns 

Guillaume Piton, Eng. 1 ; Ségolène Mejean, Eng. 2 ; Costanza Carbonari, Msc. 3 ; Jules le Guern, Msc. 2 ; Hervé Bellot, Msc. 2 ; 

Alain Rrcking, PhD 2 

ABSTRACT 

In steep slope streams, torrential-hazards mainly result from abrupt and massive sediment 

deposits. Open check dams are regularly used in natural hazard mitigation to trap sediment 

and driftwood. A good comprehension of the phenomena that occur in these structures is 

needed to optimize their design. In this paper, we present new results from small scale 

experiments addressing (i) a validation of water stage - discharge formula proposed in the 

literature for slit and slot dams; (ii) recommendations in the use of formula dedicated to 

deposition-thickness-estimation; (iii) geomorphic and hydraulics descriptions that seek to 

help field practitioners and numerical modelers to better understand what can be observed 

in labs and in the field and what kind of phenomena should be modeled. A special attention 

has been paid to highlight the implication of our results in the use of formula and in structure 

design and maintenance. 

KEYWORDS 

slot and slit dams; sediment trap; photogrammetry; Large Scale PIV; small scale model 

INTRODUCTION 

In steep slope streams and especially on their fan part, torrential–hazards mainly result from 

abrupt and massive sediment deposits. To curtail such phenomenon, soil conservation 

measures as well as torrent control works have been undertaken for decades. Since the 1950s, 

open check dams complement other structural and non-structural measures in watershedscale 

mitigation plans [Armanini et al., 1991; VanDine, 1996]. Hundreds of these structures 

have thus been built for about 60 years. Their design evolved with the improving comprehension 

of torrential hydraulics and sediment transport processes; however numerous open 

check dams have a general tendency to trap most of the sediments supplied by the headwaters 

and to weakly self-clean. Secondary effects such as channel incision downstream of 

the structures often occur after their creations. Sediment starvation trends tend to propagate 

to the main valley rivers and to disrupt past geomorphic and ecologic equilibriums. 

To minimize useless dredging operations and to promote sediment continuity, while maintaining 

the mitigation effect of open check dams, a better selectivity of sediment trapping 

must be sought in open check dams [Armanini et al., 1991; SedAlp, 2015]. To approach 

1 IRSTEA Saint-Martin-d’Hères, FRANCE, guillaume.piton@irstea.fr 

2 Irstea, centre de Grenoble, UR ETGR, St Martin d’Hères, France and Univ. Grenoble Alpes, F-38041 Grenoble, France 

3 Univ. Firenzi, Italy and Irstea, centre de Grenoble, UR ETGR, St Martin d’Hères, France and Univ. Grenoble Alpes, F-38041 Grenoble, 

France 



optimal structures that would trap sediments during dangerous floods and flush them 

partially during small floods, we must improve the scientific knowledge on hydraulic and 

deposition processes that occur in sediment traps during floods. 

Four trapping processes (TP) eventually act in sediment trap basins (Fig. 1 a): a decrease in 

transport capacity due to a milder energy slope in the basin (TP1); a decrease in transport 

efficiency due to flow spreading in a basin wider than the upstream channel (TP2); a drop 

in the shear stresses in the calm water area upstream of the dam (hydraulic control - TP3); 

and mechanical blockages against the dam openings (mechanical control - TP4). 

Figure 1: a) Trapping Processes (TP) resulting in sediment deposition in sediment trap basins: (TP1) slope decrease; (TP2) width 

increase; (TP3) delta-type hydraulic control; and (TP4) direct mechanical blockage and; Longitudinal and transverse view of flows 

through: b) a slot dam and c) a slit dam. 

The mechanical blockage process [Fig. 1 (TP4)] is relatively well understood [Piton and 

Recking, 2015a, 2015b]. On the contrary, case studies testing the existing criteria on (TPs 1-3) 

remain scarce. This paper addresses three scientific issues, two deal with (TP3): (i) How 

accurate are the current hydraulic approaches describing the water stage–discharge Eqs.? and, 

(ii) Do the approaches proposed for deposit thickness estimation give satisfying results? The 

last issue concerns (TPs 1-2): (iii) what are the flow conditions and the resulting geomorphic 

patterns that occur during sediment trap basin filling in a relatively wide and mild basin? We 

gather new elements on these questions, based on an experimental approach using a Froude 

scale model. The first runs were performed in pure water hydraulics to validate water 

stage-discharge Eqs.. Sediment was added in the subsequent runs to look at deposition 

processes. 

MATERIAL AND METHODS 

The small-scale sediment trap was built in a 6-m-long, 1.2-m-wide, 0.4-m-deep, 10%-steep 

tilting flume. The water was recirculated and measured by a flowmeter with a maximum 

discharge of 4 l/s. The sediment feeder was composed of a hopper, associated with a conveyor 

belt, with a maximum solid discharge capacity of 200 to 300 g/s depending on the grain size 

distribution of the sediment mixture. Two sediments mixtures were used, hereafter refer to as 


GSD1 & GSD2, consisting in natural poorly sorted sediments with diameter from 0.2 to 

20 mm. The median grain size D 50 

of GSD1 and GSD2 are 3.8 and 2.4 mm, respectively; and 

the mean arithmetic diameters are of 6.4 and 4.9 mm, respectively. 

High quality pictures of the flume were taken with two CANON 100D cameras fixed on a 

trolley to the ceiling of the laboratory. Digital elevation model (DEM) of the deposits were 

reconstituted, with a 1-mm accuracy [Le Guern, 2014], using the Agisfost Photoscan 

software. Before each DEM measurement, a high speed camera Phototron FASTCAM took 

videos of the flow at 125 frames/s. The Fudaa software was used to obtain surface flow 

velocity fields by large scale Particle Image Velocimetry (PIV) [Carbonari, 2015]. A point 

gauge fixed on three graduated perpendicular rails allowed to measure flow surface and 

bed altitudes with a 2-mm-accuracy. 

RESULTS 

Open check dam pure water hydraulics 

The hydraulic control of the deposit [Fig. 1 (TP3)] is a characteristic delta dynamic, i.e. is 

controlled by the water level of the tranquil water area formed by the open check dam 

backwater effect. A large number of open check dam are designed as slit or slot types 

(Eq. 1-5) (Fig. 1 b & c). Table 1 gathers the existing water stage–discharge l Eqs. ee describing la the Table flow TE1-. 

conditions that may occur in these dams: free surface or pressure flow (Fig. 1 b & c). 

Table 1: Water stage – discharge formula for slot ant slit dams 

Flow type Source Equation Eqs. number 

Free surface Zollinger [1983] Q = 2⁄ 3 

μw 0√2g H 3 (1) 

Free surface Armanini and Larcher [2001] Q = w 0 

√g( 2HH (2)* 

⁄ )3 

3 

Free surface D’Agostino [2013] Q = w 0 

√g( 2HH ⁄ 

3 ∗ 1.2 

) 3 (3)* 

Pressure flow Torricelli [1644] Q = μw 0 h 0 √2g(H − h 0 ⁄ 2) (4) 

Pressure flow Zollinger [1983] Q = 2⁄ 3 

μw 0 [√2gH 3 − √2g(H − h 0 ) 3 ] (5) 

Note: see Fig. 1 for parameter definition, with Q the discharge [m 3 /s], w0 the width of the opening [m], h0 the height of 

the opening [m], µ the contraction coefficients [-], g the gravitational acceleration [m/s²] and H, the hydraulic head [m] 

with H = h + V²/2g with h the water depth [m] and V the flow velocity in the upstream section [m/s] approximated by 

Q/Wh, with W the basin width [m]. 

* Eqs. 2 and 3 are equivalent to Eq. 1 with values of µ = 0.577 and µ = 0.439, respectively. 

Eq. 2 and 3 are based on theoretical considerations. Zollinger [1983] proposed using a value 

of 0.65 for µ but did not provide the calibration data. We propose to use our measurements to 

discriminate which equation and which µ-value are the most relevant in typical torrential 

flows. 


Height dam-configurations were tested, without sediment transport, to answer this question: 

3 slits w 0 

= [6;10;14cm] and 5 slots w 0 

xh 0 

= 10x6cm, 10x4cm, 10x2cm, 14x6cm, 14x4cm, 

14x2cm. The dam was made of a 7.5mm-thick PVC plate. The basin immediately upstream 

of the dam was 20cm-wide and its floor was covered with pebbles 14-18mm in diameter. 

64 measurements with Q [0.5;3.8 l/s] and h [16;140 mm] were taken at different stable 

states. The results are synthetized in Fig. 2 and more details on the experimental set-up, the 

data and an error analysis can be found in Mejean [2015]. 

Figure 2: Water stage–discharge relationship analyses: calibration of µ, contraction coefficient for a) free surface flows in slits and slots 

using Eq. 1 and b) pressure flows in slot using Eq. 5 and; c) comparison between experimental data and computed water-stage 

discharge. Note: Eqs. 1, 4 and 5 were used taking into account the inertia term V²/2g in H the hydraulic head term of the formula 

The linear best fit (Fig. 2 a) confirms the µ-value of 0.65 in free surface flow. Eq. 1 is 

thus recommended, rather than Eqs. 2 and 3, for equivalent thin-dam-configurations, i.e. for 

h/e [0.05;0.47], with e the dam thickness in the flow direction (Fig. 1 b). The results still 

need to be confirmed for thicker dam configurations were the Eq. 2 may be more adapted. 

For pressure flows in slots (Fig. 2 b), the linear best fit shows a µ-value of 0.68, close 

to the 0.65 value also retained by Zollinger [1983]. Eqs. 4 and 5 are almost indistinguishable 

in our results (Fig.2 c) if the hydraulic head, in the Torricelli [1644] formula, is corrected by 

half the slot height as written in Eq. 4. 

For high H/h 0 

ratio, Eqs. 4 and 5 underestimate slightly h, and conversely they overestimate 

h for low H/h 0 

ratio (Fig. 2 c). It results from the use of a constant value of µ whereas the 


contraction effects increase with H/h 0 

. Complementary experiments should be performed to 

calibrate a variation of µ with the contraction, i.e. varying with H/h 0 

and W/w 0 

, this in 

torrential context, i.e. with steep slope, rough beds and sediment transport. In addition our 

results demonstrate that taking into account the inertia term V²/2g in H the hydraulic 

head is important: Using the sole h term in the Eqs. would result in an underestimation of 

the structure discharge capacity and consequently in overestimation of the deposition and 

trapping performance (see later), which is not conservative in hazard mitigation. 

Delta thickness estimation 

During a trap filling, when sediments reach the open check dam backwater area, where flow 

velocities are low, they deposit as a delta. ∆Z, the delta thickness at the front (Fig. 1 a), 

directly controls the trapped sediment volume, e.g., a lower ∆Z means a lower trapped 

l r o oloa Formula 6 

volume. ∆Z estimation is thus a key step in the trap design. Armanini and Larcher [2001] and 

Jordan et al. [2003] proposed formulas to estimate ∆Z that can be rearranged as follow 

[Mejean, 2015]: 

92 

[Armanini and Larcher, 2001] 

l r o oloa Formula 7 

∆Z(T) = h(TT) ( W−w 0 

) [1 + ( W−w 0 

⁄ )] 

(6) 

w 0 w 0 

Equation 6 

[Jordan et al., 2003] 

[Jordan et al., 2003] T 

EQUATION 7 

0 

∆Z(T) = 1 T ∫ h(tt)dt 

93 Equation With 7W, the basin-width upstream of the open check dam; w 0, the slit-width, T the duration since the 

94 beginning of the flood and h(T) the flow-depth upstream of the structure at time T (Fig. 1 b & c), 

With W, the basin-width upstream of the open check dam; w 0 

, the slit-width, T the duration 

95 varying in time during the flood and computed using Eq. 1 and µ=0.65. 

since the beginning of the flood and h(T) the flow-depth upstream of the structure at time T 

(Fig. 1 b & c), varying in time during the flood and computed using Eq. 1 and µ=0.65. 

96 It is worth stressing that Eq. 6 has been calibrated in laterally confined flows, i.e. the flow covered all 

97 It is the worth deposit stressing due to that the Eq. flume 6 has relatively been calibrated narrow width in laterally (0.4 m). confined On the contrary, flows, i.e. Eq. the 7 has flow been 

98 covered calibrated all the in laterally deposit unconfined due to the flows, flume i.e. relatively the deposit narrow showed width a more (0.4 classical m). On delta the contrary, pattern with Eq. 

99 7 has few been mobile calibrated active channels in laterally narrower unconfined than the flows, total basin i.e. the width. deposit In addition, showed the a more authors classical of Eq. 6 

delta pattern with few mobile active channels narrower than the total basin width. In addition, 

100 assumed that the delta thickness is always equilibrated with the flow constraints and thus only 

the authors of Eq. 6 assumed that the delta thickness is always equilibrated with the flow 

101 depends on the instantaneous water depth h(T). On the contrary, Eq. 7 takes into account all the 

constraints and thus only depends on the instantaneous water depth h(T). On the contrary, 

102 

Eq. 

water 

7 takes 

depth 

into 

evolution 

account 

during 

all the 

the 

water 

flood 

depth 

through 

evolution 

the integral. 

during the flood through the integral. 

103 

To test 

To test 

these 

these 

Eqs., 

Eqs., 

five 

five 

flood 

flood 

experiments 

experiments 

were 

were 

performed 

performed 

in 

in 

the 

the 

aforementioned 

aforementioned 

flume, 

flume, 

under 

with 

104 constant solid solid concentration, C = 

Qs 

, varying from 1 1 to to 5%, 5%, with with Qs, the Qs, solid the discharge. solid discharge. Triangular 

Qs+Q 

Triangular hydrographs hydrographs were used were with water used discharge with water reaching discharge 2.75 l/s reaching at the peak 2.75 for l/s all at runs. the The peak cumulated for all 

105 

106 

sediment supply was the same in the all runs (500 kg). The hydrograph duration was thus inversely 


107 proportional to the concentration (see Mejean [2015] for more details). 

(7)

uns. The cumulated sediment supply was the same in the all runs (500 kg). The hydrograph 

duration was thus inversely proportional to the concentration (see Mejean [2015] for more 

details). 

In our experimental conditions, the flows were laterally unconfined, i.e. sediment entering 

the open check dam-backwater-area was transported in an active channel narrower than the 

basin width and flowing between deposit terraces. Several photogrammetric measurements 

were taken on each experiment at different times on the hydrographs, thus for various 

instantaneous discharges and water depths. Taking it into account and, as could be expected 

from Eqs. 6 and 7, ∆Z evolved between measurements. Width-averaged-deposit-longitudinal 

profiles were extracted from the DEMs [Mejean, 2015]. The delta-thickness was identified at 

the break in the slope and measured from the bottom of the open check dam (Fig. 1 a). Fig. 3 

shows the comparison between measured and calculated values of ∆Z, using Eq. 6 and 7. 

Figure 3: Comparisons between measured values of delta thickness, ∆Z, and predicted values by a) Eq. 6 and b) Eq. 7, in pure water 

hydraulics and with ∆Hsed. Taking into account ∆Hsed in the h estimation is necessary to achieve a satisfying estimation of ∆Z. 

One can notice in Fig. 3 that, all other parameters being geometrically fixed, Eqs. 6 and 7 

underestimate ∆Z when using the pure water hydraulics described by the unique 

Eq. 1 to estimate h(t). To estimate the total water depth l of sediment-laden r o oloa flows in Formula slits and 

slots, Piton and Recking [2015a] recommend to take into account an additional head loss 

related to the sediment transport, Δ sed 

= 1.5D MAX, 

, with D MAX 

the maximum transported 

sediment diameter [m]. Using this adaptation to the clear water hydraulics, the water depth is 

computed using: 

h(tt) = h pppppppp wwwwwwpppp (tt) + 1.5D MAX (8) 

Equation 8 


Taking into account this additional head loss, Eqs. 6 and 7 give a rather good 

approximation of ∆Z, though some scattering remains (Fig. 3 a & b, grey dots). This 

scattering is probably due to ∆Z-measurement uncertainties (the transversal profiles at the 

delta fronts were not flat, thus their representative height, here taken as the median altitude 

on the profile, are subject to interpretation) combined to a natural variability of the phenomenon, 

e.g., high deposition or incision preceding the measurement. In addition, the hypothesis 

that the delta-front-shape is always equilibrated with the flow constrains rapidly fall in 

defect during the hydrograph recession, as supposed initially by Armanini and Larcher 

[2001], explaining the low ∆Zs computed in Fig. 3 a. Eq. 6 is thus not adapted to compute ∆Z 

in wide basins during the hydrograph recession. Eqs. 7 give thus a good approximation 

of ∆Z if the water depth is correctly computed, i.e. taking into account all additional 

head losses related to torrential hydraulics (strong sediment transport, driftwood 

accumulation). 

GEOMORPHIC DESCRIPTION OF A SEDIMENT TRAP FILLING 

When entering the basin, flows and sediment pass from a steep-laterally-confined to a 

milder-laterally-unconfined situation. In this situation, Zollinger [1983] observed both 

mono-channelized and braided fan-shape deposits. The transition from confined to unconfined 

flows raises complex issues in field observations and in numerical modeling [Piton and 

Recking, 2015a]; e.g. does channelized and braided patterns come from different flood-types? 

or how to compute the deposition slope of an unconfined massive bed-load supply? 

13-experiments, 5 with a slit-dam (analyzed before) and 8 with a basin without open-checkdam 

(same concentration and hydrographs feedings, two sediment mixtures [Mejean, 2015]), 

were performed to observed deposition-processes and hydraulics’ conditions of bed-load 

trapping. Cycles of channelized and braided-like-patterns were systematically observed (e.g. 

Fig. 4). These patterns are simply different phases of a basin-filling, which is not a deposits-continuous-progression 

but rather jerky-sediment-propagations occurring after reconstitution 

of sufficient sediment stocks in the inlet-vicinity. Grain size sorting and deposit armoring 

play key roles in these cycles: braided patterns were observed to be steep and paved while 

channelized pattern to be milder, with a bed smoothed by the finer subsurface materials 

released during the channelization. 

There is thus not a unique value of deposit-slope but rather a range of slope in which a 

dynamic-equilibrium fluctuates. A method to estimate the slope range should be developed in 

further analysis. In addition, laterally-confined complementary experiments demonstrated 

that the deposition-slope increases in unconfined configurations (see Mejean [2015]). 

As a method to estimate deposition slope is still lacking, it is generally recommended in new 

sediment trap design to measure deposition slopes in the field, for example upstream of 

existing check dams, to estimate the deposition slope in the future trap. Our results demonstrate 

that deposition slopes measured above check dams, i.e. in quite confined configura- 


tions, must be considered as minimum values of the possible bedload deposit slopes in 

sediment trap basins. One must note that it is not the case for mud flows which may deposit 

with very gentle slopes [Piton and Recking, 2015a]. 

Partial self-cleaning was also observed systematically: (i) during the hydrograph recession in 

slit-dam experiments and conversely (ii) during the peak-flows in dam-less experiments. 

Slit dams prevent peak-flow-releases due to the delta-like dynamics [Fig. 1 (TP3)], which is 

maximum at the peak flows. The volumes that were stored in the delta-front were subsequently 

re-eroded and partially flushed during the flow-recession, leaving terraces’-like 

patterns. The observation of a clear-incised-channel in the downstream part of the deposits is 

thus an evidence of partial self-cleaning. Adding a slit-grill or driftwood would have created 

a jam on the slit and prevented this self-cleaning phenomenon [Piton and Recking, 2015a, 

2015b]. 

Figure 4: Photos of flows, DEM representation of deposit thickness and surface velocity fields of a) braided like-patterns during the 

initial filling of the inlet-vicinity; b) armor breaking leading to a channelization transferring sediment to the basin central part; c) new 

braided like patterns followed by d) another channelization, whose channel connect the inlet to the outlet, leading to a partial 

self-cleaning. Flow from the top to the bottom of the pictures. 

HYDRAULICS ON MASSIVE DEPOSITS 

PIV and DEM measurements were analyzed to deduce the slope, Froude and Shields numbers 

on massive-bed-load deposits (Fig. 5). Before each DEM and LSPIV measurements, the water 

depth, h gauge 

, has been measured using the point gauge. At the same coordinates, the 


171 

Hydraulics on massive deposits 

ts s zed to deduce the slope, Froude and Shields numbers on 

172 PIV and DEM measurements were analyzed to deduce the slope, Froude and S 

re d to each deduce DEM the and slope, LSPIV Froude measurements, and Shields numbers the 

173 

water on massive-bed-load 

depth, 

were analyzed to deduce the slope, Froude and Shields numbers on 

deposits (Fig. 5). Before each DEM and LSPIV measurements 

t ig. each gauge. 5). Before DEM At and each the LSPIV same DEM and measurements, coordinates, LSPIV measurements, the the topographical water the 174 depth, water hdepth, 

gauge, 

profile, has been measured using the point gauge. At the same coordinates, the to 

ing o auge. measured. the point At the gauge. same The At coordinates, surface the same velocity coordinates, the topographical was the interpolated topographical 175 profile, transverse profile, 

on a 1- to the flow direction, was also measured. The surface velocity was i 

topographical profile, transverse to the flow direction, was also measured. The surface 

ion, measured. was also The measured. surface The velocity surface was velocity interpolated was interpolated velocity was interpolated on a 1-mm-transversal 176 on a 1- on a 1- 

alues/profile). The mean value of the interpolated mm-transversal surface step on step the on profiles the profiles (N (N values/profile). The The mean value of the i 

profiles (N values/profile). The lues/profile). mean The value mean of value the of interpolated the interpolated surface 177 surface 

velocities, = 

 

ough estimation of = mean value of the interpolated surface 

, the profile mean velocity. The 

, gives a rough estimation of = , the profile 

 

, gives a rough estimation gh estimation of , the of = of = , the profile mean velocity. The 

, profile the profile 

mean mean velocity. The deposition slope S has been measured on the DEM 

178 

The 

deposition slope S has been measured on the DEM along a longitudinal pro 

on measured the 

along 

DEM the a 

along DEM along longitudinal 

a a longitudinal profile profile passing passing by by the transversal 

by the 

the 

profile. Rough estimation of, Fr and 

n the DEM along a longitudinal profile passing 179 by the 

timation transversal profile. Rough estimation of Fr and , the Froude and the Shield 

Fr and of Fr with and the Shield stress for the D 84 

,respectively, were computed using: 

r and , and the 

Froude , the Froude and and the the Shield stress for the the D 84 D, 

84 , 

, the Froude and the Shield stress for the D 84 , 

respectively, sediment were computed density taken using: as 1.65. 

 

 

All and 

180 

, with 

 

 

 

 

 

 

 

and 

dimensionless and 

 

d using: 

 

and 

, with ∆ the submerged 

 

, with ∆ the submerged 

number , with analyzed ∆ the submerged hereafter are representative of local values of flow features in 

 

65. All dimensionless number analyzed hereafter are representative 181 sediment of 

density taken as 1.65. All dimensionless number analyzed hereafter ar 

active channels. They are not averaged on the basin width. 

n nsionless active channels. number number They analyzed are analyzed not hereafter averaged hereafter on are the representative basin are representative width. 

182 of of 

local values of flow features in active channels. They are not averaged on the basi 

nnels. They They are are not not averaged averaged on the on basin the width. basin width. 

Figure 5: Rough estimation of main flow dimensionless numbers: a) deposition slope vs inlet solid concentration; b) Froude number vs 

inlet solid concentration and; c) Froude number vs Shield number 

The absolute values illustrated in Fig. 5 are rough estimations because of the high uncertainties 

on h gauge 

, however some general trends of interactions between geomorphology and 

hydraulics can be observed: (i) the deposition slope strongly fluctuates for a given solid 

concentration (Fig. 5 a) highlighting the geomorphic cycles' magnitude and that armoring 

process lead to various equilibrium slope despite a constant solid concentration at the inlet. 

(ii) The Froude number also strongly fluctuates (Fig. 5 b) which is likely to be mainly related 

to the varying velocities related to varying bed roughness within the geomorphic and 

armoring cycles. (iii) Interestingly an unintuitive inverse correlation seems to appear between 

the Froude number and the Shields number (Fig. 5 c): Low Shields numbers were computed 

on the milder slopes observed during chenalisations but where the smooth bed allowed high 

velocities and Froude numbers. Conversely, steep paved braided fan-patterns showed high 

slopes (and thus Shield numbers) and low velocities (and thus Froude numbers) due to their 

rough and paved beds. Local measurements of sediment diameter should be done to compute 

more accurate value of Shields numbers and to confirm this inverse correlation. These 

preliminary results need to be more deeply analyzed to specify the autogenic fluctuating 

hydraulics of massive bed-load deposits related to grain size sorting and to define which 

friction law and transport formula are the most relevant to compute such phenomena. 


CONCLUSIONS 

The new elements listed in this paper will help designers to more accurately design and 

numerically model the structures, specifically slot and slit dams, so to better adapt them to 

each site and their natural-hazard-specificities and mitigation-objectives. They also highlight 

the varying nature of sediment transport of poorly sorted mixtures and the necessity to push 

further the research on this subject, which, so far, is not enough understood to provide 

accurate design methods to practitioners. 

ACKNOWLEDGMENTS 

This study was funded by Irstea, the INTEREG-ALCOTRA European RISBA project, and the 

ALPINE SPACE European SEDALP project. The authors would like to thank Matjaž Mikoš for 

his editorial work and two reviewers who help us to improve this paper. 

REFERENCES 

- Armanini et al., 1991. From the check dam to the development of functional check dams. 

Fluvial Hydraulics of Mountain Regions 37, 331–344. 

- Armanini & Larcher, 2001. Rational criterion for designing opening of slit-check dam. 

J. of Hydr. Eng. 127, 94–104. 

- Carbonari, C., 2015 Experimental observations on the functioning of sediment trap basins: 

LSPIV measurements of low submersion flows, Msc. Thesis. 

- D’Agostino, V., 2013. Filtering-retention check dam design in mountain torrents, in: 

Check Dams, Morphological Adjustments Erosion Control Torrential Streams., pp. 185–210. 

- Jordan et al., 2003. Modélisation physique d’un piège à graviers, le cas du Baltschiederbach. 

Wasser Energie Luft 95, 283–290. 

- Le Guern, J., 2014. Modélisation physique des plages de dépôt : analyse de la dynamique 

de remplissage, Msc. Thesis. 

- Mejean, S., 2015. Caractérisation des conditions hydrauliques du piégeage de la charge 

sédimentaire grossière des torrents, Msc. Thesis. 

- Piton & Recking, 2015a. Design of sediment traps with open check dams I: hydraulic and 

deposition processes. J. of Hydr. Eng. In press. 

- Piton, G., Recking, A., 2015b. Design of sediment traps with open check dams II: 

woody debris. J. of Hydr. Eng. 142(2). 

- SedAlp, 2015. Work Package 6 Final report - Interactions with structures. Alpine Space 

European project 

- Torricelli, E., 1644. Opera geometrica, Firenze. 

- VanDine, D.F., 1996. Debris Flow Control Structures for Forest Engineering. Res. Br., B.C. 

Min. For., Victoria, BC. 

- Zollinger, F., 1983. Die Vorgänge in einem Geschiebeablagerungsplatz (ihre Morphologie 

und die Möglichkeiten einer Steuerung), ETH Zurich PhD Thesis. 



Sediment Management in Alpine Catchments: Area of 

conflicts between protection needs, complexity of 

waste law and quality goals for water bodies 

Sedimentmanagement in Alpinen Einzugsgebieten: 

Konfliktfeld zwischen Schutzbedarf, abfallrechtlicher 

Komplexität und Qualitätszielen für Fließgewässer 

Florian Rudolf-Miklau, Priv. Doz. Dr. 1 ; Susanne Mehlhorn, Dipl.-Geogr. 2 

ABSTRACT 

Flood related sediment management has to take into account as well the protection function 

as the morphologic, ecological and chemical quality goals according to the EU WFD. Sediment 

materials, for which disposal is intended (in particular after floods or debris flow) or contamination 

is proved, are subject to the regime of waste law. In this paper a 5-phase-model for the 

sediment management in torrential watersheds is presented, which primarily aims at the goal 

to keep sediments as fare as possible in the water course (relocation, dotation into water 

body) or use it for the purpose of flood (torrent) control. Only in evitable cases sediment shall 

be used for other purposes (raw or building material, agricultural use) or deposited as these 

options are combined with a complex legal framework and higher expenses. A guideline is 

the state of elaboration, which will provide concepts for a structured and efficient sediment 

management in torrent watersheds. 


Das schutzwasserwirtschaftliche Sedimentmanagement hat neben der Schutzwirkung auch 

den morphologischen, ökologischen und chemischen Qualitätszielen der EU WRRL zu 

entsprechen. Feststoffe, für die eine Entledigungsabsicht besteht (insbesondere nach 

Hochwasser oder Muren) oder deren Kontamination nachgewiesen wird, fallen unter das 

Regime des Abfallrechts. Der Beitrag behandelt ein 5-Phasen-Modell des Sedimentmanagements 

in Wildbacheinzugsgebieten, dessen primäres Ziel die Belassung der Sedimente im 

Gewässeregime (einschließlich Umlagerung und Wiedereinbringung) oder in schutzwasserwirtschaftlicher 

Verwendung ist. Nur in unvermeidlichen Fällen ist eine sonstige Verwendung 

(Rohstoff, Baustoff, landwirtschaftliche Verwertung) oder Deponierung des Materials 

angezeigt, da diese Optionen in einem komplexen Rechtrahmen erfolgen und mit höheren 

wirtschaftlichen Aufwendungen verbunden sind. Ein Leitfaden des Österreichischen 

Wasser- und Abfallwirtschaftsverbandes (ÖWAV) ist im Stadium der Ausarbeitung, der die 

1 BMLFUW Vienna, AUSTRIA, florian.rudolf-miklau@die-wildbach.at 

2 Austrian Torrent and Avalanche Control Service 



Grundlage für ein strukturiertes und effizientes Sedimentmanagements in Wildbacheinzugsgebieten 

bilden soll. 

KEYWORDS 

sediment management; torrent control; waste law; sediment continuum; contamination. 

VORBEMERKUNG 

Der Beitrag gibt den Stand eines interdisziplinären Expertendiskurses wieder, der 2016 in 

der Erstellung einer Richtlinie des Österreichischen Wasser- und Abfallwirtschaftsverbandes 

(ÖWAV) münden soll. Es wird auf europäisches und nationales österreichisches Recht Bezug 

genommen. 

BEGRIFFSBESTIMMUNG UND BEDEUTUNG FLUVIATILER SEDIMENTE 

Sedimente bilden einen elementaren Bestandteil von Fließgewässern und umfassen gemäß 

DIN 4049-1 klastische, chemische und biogene Komponenten. Der Begriff „Feststoff“ iSd 

ÖNORM 2400 beschreibt die Gesamtheit der vom Fließgewässer transportierten festen Teile, 

also Geschiebe (gleitend, rollend oder springend fortbewegte Gesteinsteile), Schwebstoffe 

(durch Turbulenz in Schwebe gehaltenen Partikel) und Schwimmstoffe (insbesondere Holz). 

Feststoffe gelangen überwiegend durch Erosionsprozesse ins Gewässer, werden je nach 

Wassermenge, Fließgeschwindigkeit und Kornspektrum unterschiedlich weit transportiert, 

zerkleinert und schließlich zur Ablagerung gebracht (sedimentiert). Das Grundmodell der 

Feststoffbilanzierung eines Fließgewässerabschnitts (Abbildung 1) ist im Wesentlichen eine 

Funktion des Ein- und Austrags sowie der räumlichen und zeitlichen Veränderung der 

Verlagerungsprozesse und der Morphologie in der Gerinnesohle. Das Ausmaß des Eintrags 

aus dem Oberlauf oder aus Zubringern kann sich durch natürliche oder anthropogene 

Einflüsse verändern. Sedimentation und damit verbundene Sohlauflandung tritt auf, wenn 

die Transportkapazität der Wasserwelle nicht ausreicht, um das eingetragene Sediment weiter 

zu transportieren. Ein die Transportkapazität nicht auslastender Feststoffeintrag führt meist 

zur Erosion im Fließgewässerabschnitt, wodurch die Sohlneigung sowie mittelbar die 

Schleppspannung und Transportkapazität reduziert werden. (DWA 2012, 23) Weiters tritt 

natürliche Umlagerung auf, wenn in einem Abschnitt Feststoffe wiederholt erodiert und 

deponiert werden (ONR 24800, 4.1.9). Die Sedimententnahme durch Baggerung führt meist 

zu einem Austrag, eine Rückgabe des Materials an anderer Stelle ins Gewässer (Wiedereinbringung) 

bewirkt einen Sedimentinput (künstliche Umlagerung). (DWA 2012, 23) 

Aus der Perspektive des Hochwasserschutzes führt der Sedimenthaushalt in Fließgewässern 

(Überschuss, Gleichgewicht oder Mangel) zu bedeutenden morphologischen Effekten, 

insbesondere zur Tiefenerosion, zur Auflandung, zur Verklausung oder zur Gerinneverlagerung, 

denen durch „klassische“ Maßnahmen der Wildbachverbauung wie die Konsolidierung 

der Gewässersohle, die Regulierung des Gewässerlaufs, den Geschieberückhalt oder die 

Räumung begegnet wird (Rudolf-Miklau 2009, 148). Andere Funktionen von Sedimenten, 


Abbildung 1: Grundmodell der Fließgewässerbilanzierung in Gewässerabschnitten 

vor allem die ökologische Bedeutung als Lebensgrundlage von Fischen und Makrozoobenthos, 

die morphologische Wirkung für die Formung und Rauigkeit des Gewässerbetts sowie 

die Verwendung als Rohstoff und Baumaterial, müssen in einem gesamtheitlichen Sedimentmanagement 

ebenfalls berücksichtigt werden. Aus rechtlicher Sicht können Sedimente dem 

Charakter nach Ressource, Schutzgut oder Schadensfaktor sein (Bergthaler et al. 2015, 5). 

Sediment besitzt außerdem die Eigenschaft, „eine Reihe von schwer löslichen und schwer 

abbaubaren Umweltchemikalien und Abfallstoffen anzureichern“ (Geoakkumulation), wobei 

die Konzentration bis zu tausendfach höher als in der Wasserwelle sein kann (Müller 1986, 

110). Auf die Eigenschaft von Sediment als Schadstoffträger zielen insbesondere die Umweltqualitätsnormen 

des Art 2 Z 35 der EU-WRRL (2000/60/EG) ab, wobei die entsprechenden 

nationalen Qualitätszielverordnungen (AT: QZV Chemie OG 2006; D: OGewV 2011) bisher 

konkret nur wenige Schadstoffe ansprechen. Neben der Geoakkumulation kommt auch ein 

Eintrag von anthropogen abgelagerten Schadstoffen – etwa aus Deponien, Bergbauhalden 

oder der Landwirtschaft – durch Abschwemmung, Lösung oder Erosion als Kontaminationsquelle 

in Frage. 

Schutzwasserwirtschaftliches Sedimentmanagement umfasst Maßnahmen zur Aufrechterhaltung 

des Sedimentkontinuums oder der Wiedereinbringung von Sedimenten ins Gewässer, 

die Schaffung von Sedimentationsräumen (-becken), die Verbesserung der Transportkapazität 

durch bauliche Maßnahmen, die Räumung von Sedimenten (nach Ereignissen), die künstliche 

Umlagerung und Zwischenlagerung, die wirtschaftliche Verwertung der Räumgutes, die 

Schaffung von Deponieflächen sowie die Entsorgung. 


SEDIMENTMANAGEMENT IN WILDBACHEINZUGSGEBIETEN: AUFGABEN UND FUNKTIONEN 

Wildbacheinzugsgebiete (in AT: ca. 12.500) sind in besonderem Maße durch fluviatile oder 

murartige Verlagerungsprozesse von Feststoffen geprägt (vgl. Legaldefinition „Wildbach“ nach 

§ 99 Abs 1 ForstG). Meist handelt es sich um Einzugsgebiete außerhalb der intensiv von 

Menschen genutzten Gebiete, in denen natürlich anstehende Fest- und Lockergesteine 

abgetragen werden. Die primären Sedimentationsflächen der Wildbäche sind Schwemm- oder 

Murkegel, in denen große Teile der abgetragenen Massen aus den Einzugsgebieten über 

Jahrhunderte gespeichert werden. (Bergmeister et al. 2008, 5). Durch die Errichtung von 

Siedlungen und entsprechende Schutzbauwerke sind in den in den Alpen zahlreiche 

Schwemmkegelflächen als natürliche Sedimentationsräume ausgeschaltet und das Feststoffregime 

alpiner Wildbäche nachhaltig verändert worden. 

In den letzten Jahrzehnten hat sich die Technologie der Feststoffbewirtschaftung in Wildbacheinzugsgebieten 

durch Retention, Dosierung oder Filterung (ONR 24800, Tabelle 5, 40) mit 

großen Sperrenbauwerken rasch entwickelt. Gleichzeitig wurde dadurch der Erhaltungsaufwand 

für die Räumung der Rückhaltebecken stark erhöht. In Österreich existiert in Wildbacheinzugsgebieten 

nach groben Schätzungen ein räumbares Feststoffretentionsvolumen von 

60 Mio. m³. Diese Retentionsbecken müssen zur Aufrechterhaltung der Schutzfunktion 

regelmäßig, jedenfalls aber nach Hochwasser und Muren geräumt werden. In „normalen“ 

Jahren ist ein Räumvolumen von 0,5 – 1,5 Mio. m³ zu bewältigen, welches in Katastrophenjahren 

noch wesentlich höher liegen kann. Alleine im Bundesland Salzburg waren im 

Hochwasserjahr 2013 345.000 m³ aus Rückhaltebecken zu räumen. Die durchschnittlichen 

Räumungskosten einer Baggerung einschließlich Deponie sind in den letzten Jahren 

dramatisch gestiegen und erreichen ca. € 10,-/m³ (für dislozierte kommerzielle Bodenaushubdeponien 

sogar bis zu € 35,-/m³). Die jährlichen Investitionen in Räumung von Rückhaltebecken 

liegt zwischen € 5 und 10 Mio. 

Sowohl die stark angestiegenen Räumvolumina als auch die mit der Europäischen Abfallrichtlinie 

(EU-ARL, 2008/98/EG) verbundenen Anforderungen an die Deponierung oder 

Behandlung kontaminierter Sedimente stellen die Wildbachverbauung vor neue Herausforderungen 

und haben das Potenzial, bewährte Schutz- und Managementkonzepte in Frage zu 

stellen. Es ist daher erforderlich, das Sedimentmanagement in Wildbacheinzugsgebieten unter 

Berücksichtigung der morphologischen, ökologischen, rechtlichen und wirtschaftlichen 

Perspektiven zu einem umfassenden System weiterzuentwickeln. Grundlage dafür bildet eine 

aus der Nomenklatur der feststoffwirksamen Schutzmaßnahmen (ONR 24800, 4.6.3) sowie 

den Rechtsbegriffen des Forst-, Wasser- und Abfallrechts entwickelte Systematik der 

Managementfunktionen. 

PROBLEMSTELLUNG: SEDIMENTMANAGEMENT UNTER DEM REGIME DES ABFALLRECHTS 

Feststoffe aus Wildbacheinzugsgebieten stammen fast ausschließlich aus natürlich anstehenden 

Lockergesteinsherden ohne erhebliche anthropogene Beeinflussung, sodass grundsätzlich 


von der Eigenschaft des „nicht kontaminierten Sediments“ gemäß § 3 Abs 1 Z 7 AWG 

ausgegangen werden kann. Allerding kann ohne Nachweis der Kontaminationsfreiheit nicht 

abschließend angenommen werden, dass keine Verunreinigung der Sedimente aus anthropogenen 

oder geogenen Ursachen gegeben ist. In diesen Fällen wäre nämlich die Abfalleigenschaft 

des Sediments anzunehmen. 

Die EU-ARL schließt gemäß Art 2 Abs 3 „Sedimente, die ua. zum Zweck der Bewirtschaftung 

von Gewässern oder der Vorbeugung gegen Überschwemmungen oder der Abschwächung 

der Auswirkung von Überschwemmungen umgelagert wurden und erwiesenermaßen nicht 

gefährlich sind“, aus ihrem Anwendungsbereich explizit aus. Eine ähnliche Ausnahmebestimmung 

enthält § 3 Abs 1 AWG für „nicht kontaminierte Sedimente“. Der Begriff „Kontamination“ 

bezieht sich ausschließlich auf anthropogen verursachte Verunreinigungen. Geogene 

Hintergrundbelastungen gelten nicht als Kontamination im Sinne der zitierten Rechtsnormen. 

Die Verpflichtung des Nachweises der Kontaminationsfreiheit ist jener Person auferlegt, 

die sich auf die betreffende Ausnahme berufen will. (Bergthaler et al. 2015, 14) Das Ausmaß 

und die Methode der Nachweisführung (Augenschein oder analytische Laborprüfung) 

ergeben sich aus den nationalen Beweisregeln der ÖNORM S 2126. Bei Sedimenten aus 

naturnahen Wildbächen, bei denen im Oberlauf keine maßgeblichen Direkteinleiter oder 

sonstige potenziell kontaminierenden Einwirkungen auf das Gewässer bekannt sind, wird 

man mit einem Augenschein das Auslangen finden. Bei bekannter gewerblicher oder 

bergbaulicher Nutzungsgeschichte im Einzugsgebiet wird hingegen hinsichtlich der in Frage 

kommenden Belastungen eine ergänzende analytische Untersuchung indiziert sein. 

Die Ausnahme des Sediments von der Abfalleigenschaft nach Art 2 Abs 3 EU-ARL ist an den 

Begriff „Umlagerung innerhalb von Oberflächengewässern“ gekoppelt, konstatiert also einen 

unabdingbaren Gewässerbezug. Die Auflistung der möglichen Maßnahmen in den Leitlinien 

der Kommission zeigt aber, dass sich diese nicht auf das Bett des Fließgewässers beschränken 

müssen, sondern dass auch im Landbereich durch gewässerbezogene Maßnahmen noch 

zulässige Verwendungszwecke (insbesondere im Bereich des Hochwasserschutzes und der 

Wildbachverbauung) bestehen (Bergthaler 2015, 20). Eine weitere anwendbare Ausnahmemöglichkeit 

besteht gemäß § 3 Abs 1 Z 8 AWG, wenn geräumte Sedimente den „nicht 

kontaminierten Böden und anderen natürlich vorkommenden Materialien“ zugeordnet 

werden. Hier ist zwar eine räumliche Einschränkung des Materials auf den "Ort, an dem es 

ausgehoben wurde" (Bodenaushub) formuliert. Als Abgrenzungskriterium gilt der Bereich, 

der von der jeweiligen Baugenehmigung umfasst ist, und daher nach der jeweiligen Umschreibung 

des Bauprojekts durchaus weitreichend auszulegen ist. (Bergthaler 2015, 20) Dies 

bedeutet, dass bei schutzwasserwirtschaftlichen Baumaßnahmen die ausgehobenen Sedimente 

an allen bautechnischen Einsatzbereichen entlang der projektgegenständlichen Fließstrecke 

verwendet werden können und immer noch vom Abfallbegriff ausgenommen bleiben. 

In diesem Kontext ist in Österreich auch an eine Reaktivierung des Rechtsinstituts des 


Perimeters (Arbeitsfeldes) nach § 1 WLV-G zu denken, welches dem hier verwendeten 

Baustellenbegriff entsprechen dürfte. 

Liegen die Voraussetzungen des § 3 Abs 1 Z 7 und 8 AWG nicht vor, wird davon auszugehen 

sein, dass es sich bei Sedimenten im Falle der Räumung um Abfall handelt. Dies gilt in jedem 

Fall, wenn das Material entsorgt und außerhalb des Gewässers endgültig deponiert werden 

soll. In diesem Fall sind in Österreich die Maßstäbe des Bundesabfallwirtschaftsplanes 

(BMLFUW 2006) bzw. bei einer Beseitigung die Bestimmungen der Deponieverordnung 

(DVO 2008) anzuwenden. Auch wenn hier nur Auszüge aus der viel umfangreicheren 

rechtlichen Analyse des Sedimentmanagements (Bergthaler et al. 2015) wiedergegeben 

wurde, so kann doch der Schluss gezogen werden, dass bei der Behandlung von geräumten 

Sedimenten ein planmäßiges und strukturiertes Vorgehen obligatorisch ist, um nicht 

unbeabsichtigte Rechtsfolgen und wirtschaftliche Mehrbelastungen zu generieren. 

PHASENMODELL DES SEDIMENTMANAGEMENTS UND VERWERTUNGSREGIME 

Als Basis für das Ziel der Entwicklung eines strukturierten und rechtlich fundierten Konzepts 

für das Sedimentmanagement in Wildbacheinzugsgebieten wurde – aufbauend auf dem 

Grundmodell nach Abbildung 1 – ein 5-Phasen-Modell entwickelt, welches mit den für die 

Sedimente in Frage kommenden Behandlungs- und Verwertungsregimen (Schutzwasserwirtschaftliches 

Regime, Land- und forstwirtschaftliches Verwertungsregime, Rohstoffverwertungsregime, 

Abfallregime) gekoppelt wurde. Das Modell umfasst folgende Phasen: 

– Phase 0: Sedimentkontinuum oder schutzwasserwirtschaftliche Maßnahmen im Gewässer 

– Phase 1: Anfall und erste Disposition 

– Phase 2: Zwischenlagerung 

– Phase 3: Weitere Verwendung (Verwertung) und Entsorgung (Deponierung) 

– Phase 4: Wiedereinbringung in Gewässer 

Abbildung 2 zeigt den Materialfluss bezogen auf die möglichen Arten der Behandlung der 

Sedimente und der abfallrechtlichen Qualifikation des Materials. Als Grundprinzip des 

Modells gilt, dass aus ökologischen und morphologischen Gründen nach Möglichkeit die 

Feststoffe im Sedimentkontinuum des Wildbaches belassen oder im Gewässer umgelagert 

werden sollen. Wenn jedoch eine Räumung erforderlich ist, soll die Verwertung für schutzwasserbauliche 

Zwecke angestrebt werden. Aufgrund der rechtlichen und wirtschaftlichen 

Lasten ist eine sonstige Verwertung außerhalb des Gewässers oder eine Deponierung nur in 

unvermeidbaren Fällen anzustreben. 

Ein durchgängiges Feststoffregime oder Schutzkonzepte, die den Sedimentfluss aufrecht 

erhalten, stellen im Sinne des “River Continuum Concepts“ (Vanote et al. 1980, 130f) sowie 

der Qualitätsziele der EU-WRRL (2000/60/EG) den Optimalzustand dar (Phase 0) und sind 

auch aus schutzwasserbaulicher Sicht am kosteneffizientesten. Allerdings kann in Wildbächen 

ein natürliches Feststoffregime nur in den Grenzen der Sicherheit des menschlichen 

Lebens- und Wirtschaftsraums aufrechterhalten werden. Einige Funktionstypen der Wild- 


Abbildung 2: Flowchart des Sedimentmanagements in Wildbacheinzugsgebieten 

bachverbauung – insbesondere die Konsolidierung, Dosierung, Filterung oder Ablenkung 

(ONR 24800, 4.6.3) – zielen auf den Feststofftransport bei Hochwasser und begünstigen das 

Sedimentkontiunuum bei Normalwasserführung sowie die morphologische Balance des 

Wildbaches. Die kleinräumige Umlagerung von Sedimenten (im engeren schutzwasserwirtschaftlichen 

Sinne) wird zur Überbrückung einer abschnittweise nicht ausreichenden 

Transportkapazität eingesetzt. 

Kommt es zu massiven Feststoffablagerung und stellen diese ein Abflusshindernis der freien 

Fließstrecke dar, ist eine Räumung (Sedimententnahme) durchzuführen (Phase 1). Gesetzliche 

Räumpflichten bestehen nach §§ 39 und 47 WRG sowie in Form der Wildbachräumung 

nach § 101 Abs 6 ForstG. Eine verwaltungsrechtliche Bewilligungspflicht für Räumungen 

besteht nur bei nutzungs- oder gewinnorientierter Sedimentgewinnung (z.B. Schotterabbau 

iSd § 1 Z 8 MinroG; Entnahme als Wassernutzung iSd § 9 WRG), nicht jedoch in Erfüllung 

gesetzlich normierter Räumungs- und Instandhaltungspflichten. In dieser Phase besteht 

weiterhin ein räumlicher Gewässerbezug, da das geräumte Sediment unter der Annahme 

einer schutzwasserwirtschaftlichen Verwertung im Nahbereich zwischengelagert und sortiert 

wird. 

Von einer Zwischenlagerung (Phase 2) ist zu sprechen, wenn das Sedimentmaterial iSd § 2 

Abs 7 Z 4 AWG für den Weitertransport nicht länger als 1 Jahr bereitgehalten oder vor einer 


Verwertung nicht länger als 3 Jahre gelagert wird und daher keine Deponie (endgültige 

Ablagerung) vorliegt. Die Zwischenlagerung verschafft zeitlichen Spielraum, um den 

Sedimentanfall zu puffern und eine schutzwasserwirtschaftliche Verwendung, Wiedereinbringung 

ins Gewässer oder sonstige Verwertung des Materials sicher zu stellen. Allerdings sind 

dafür geeignete Lagerflächen im ausreichenden Umfang vorzuhalten, für die verwaltungsrechtliche 

Bewilligungstatbestände bestehen (z.B. HQ-30 nach § 38 WRG, befristete forstliche 

Rodung nach § 18 ForstG, naturschutzrechtliche Bewilligungs- und Anzeigepflicht). Abfalleigenschaft 

ist für nicht kontaminierte Sedimente mangels Entledigungsabsicht in der Phase 1 

und 2 nicht anzunehmen. 

Phase 3 umfasst alle Arten der Verwertung außerhalb des Gewässers, als Baumaterial (z.B. 

Straßenbau), als produktgleicher Rohstoff (Schotter- und Kiesgewinnung) oder für landwirtschaftliche 

Zwecke (Geländegestaltungen, Rekultivierung, Einarbeitung in den Boden). Bei 

der Weitergabe (Verkauf) der Sedimente sind insbesondere eigentumsrechtliche und 

gewerberechtliche Aspekte zu beachten. Besteht kein neuer Verwendungszweck für das 

zwischengelagerte Sedimentmaterial und liegen die Ausnahmetatbestände nach § 3 Abs 1 

AWG 2002 nicht vor, so ist durch die Weitergabe des Sediments wohl Entledigungsabsicht 

indiziert und ist mit der Übergabe an einen Abfallbehandler dieser explizit mit der umweltgerechten 

Verwertung und Beseitigung zu beauftragen. Als Deponietypen für nicht kontaminierte 

Sedimente kommen definitionsgemäß Bodenaushubdeponien in Frage, für Sedimente, 

die mit Fremdbestandteilen verunreinigt sind, ist die Kategorie der Baurestmassendeponie 

einschlägig. (Bergthaler 2015, 142) Das Einbringen von Abfällen in einen Deponiekörper 

unterliegt grundsätzlich der Altlastenbeitragspflicht, im Falle von Katastrophenereignissen ist 

eine Deponierung aber nach § 3 Abs 4 AlSAG nach Bestätigung durch die Gemeinde 

beitragsfrei (Altlastensanierungsabgabe) möglich. 

Als Phase 4 wurde die Wiedereinbringung von Sedimentmaterial in ein Gewässer definiert, 

die grundsätzlich nur für nicht kontaminiertes Material möglich ist. Eine wasserrechtliche 

Bewilligungsfähigkeit nach § 32 Abs 2 lit a WRG ist nur unter bestimmten Rahmenbedingungen 

gegeben, beispielsweise nur bei größeren Gewässern mit ausreichender Transportkapazität 

an geeigneten Orten und zum passenden Zeitpunkt. 

SCHLUSSFOLGERUNGEN UND AUSBLICK 

Schon aufgrund der ökologischen und morphologischen Bedeutung des Sediments in Fließgewässern, 

ebenso jedoch wegen der rechtlichen Komplexität und wirtschaftlichen Aufwendungen 

ist das in diesem Beitrag dargestellte 5-Phasen-Modell auf die weitgehende Belassung 

oder kleinräumige Umlagerung der Sedimente im Feststoffregime der Wildbäche ausgerichtet. 

Wenn eine Räumung aus technischen oder rechtlichen Gründen geboten ist, soll das Sedimentmanagement 

eine möglichst wirtschaftliche Verwendung des Materials sicherstellen. 

Dabei ist der schutzwasserbaulichen Verwertung oder der Wiedereinbringung ins Gewässer 

der Vorzug einzuräumen. Das zeitliche und mengenmäßige Auseinanderfallen des Sediment- 


anfalls und Materialbedarfs kann in bestimmten Fristen (3 Jahre) oder räumlich begrenzten 

Bereichen (Baustellen, Perimeter) durch Zwischenlagerung gepuffert werden. In Wildbacheinzugsgebieten 

sind daher entsprechende Zwischenlagerflächen sicherzustellen und ein 

Konzept für die mittelfristige Materialverwendung zu erstellen. Dieses Konzept soll auch 

konkrete Optionen für andere Verwertungsmöglichkeiten (landwirtschaftliche Verwertung, 

Schottergewinnung, Baumaterial) mit einschließen. Ist die Deponierung des Materials wegen 

des extremen Überschusses (nach Hochwasserereignissen) unausweichlich, so wird die 

vorausschauende Schaffung von bewilligten Bodenaushubdeponien als Teil von Schutzprojekten 

oder im Nahbereich der Wildbacheinzugsgebiete empfohlen. 

Das vorgestellte Modell versucht, die komplexen rechtlichen und technischen Zusammenhänge 

des Sedimentmanagements für die Praxis im Sinne einer effizienten und strukturierten 

Vorgehensweise umsetzbar zu machen. Es bestehen allerdings weiterhin erhebliche Rechtsunsicherheiten 

im Kontext des Abfallrechts, die das etablierte System der Wildbachverbauung 

in der Behandlung von Feststoffen erheblich hemmen oder gar in Frage stellen. Der in 

Ausarbeitung befindliche Leitfaden des Österreichischen Wasser- und Abfallwirtschaftsverbande 

(ÖWAV) soll hier eine rechtliche Klärung und Handlungssicherheit für die Praxis 

herbeiführen. 

LITERATUR 

- Bergmeister K., Suda J., Hübl J., Rudolf-Miklau, F. (2008). Schutzbauwerke der Wildbachverbauung. 

Ernst und Sohn Berlin. 

- Bergthaler W., Wagner E., Jandl C. (2015). Sedimentmanagement – Rechtliche Aspekte. 

Johannes Kepler Universität Linz. Studie iA des BMLFUW, unveröffentlicht. 

- BMLFUW (2006). Bundesabfallwirtschaftsplan. Bundesministerium für Land- und Forstwirtschaft, 

Umwelt und Wasserwirtschaft Wien. 

- Deutsche Vereinigung für Wasserwirtschaft, Abwasser und Abfall (DWA) (2012). Sedimentmanagement 

in Fließgewässern: Grundlagen, Methoden, Fallbeispiele. Merkblatt DWA 

M-526. 

- Müller G. (1986). Schadstoffe in Sedimenten – Sedimente als Schadstoffe. Mitteilungen der 

österreichischen geologischen Gesellschaft (Umweltgeologie Bd. 79). 

- Rudolf-Miklau F. (2009). Naturgefahren-Management in Österreich. Lexis Nexis Orac. 

- Vannote R. L., Minshall G. W., Cummins K. W., Sedell J. R., Cushing, C. E. (1980). The river 

continuum concept. Canadian journal of fisheries and aquatic sciences 37(1): 130-137. 

RECHTSQUELLEN 

Abfallwirtschaftsgesetz 2002, idF österr BGBl. I Nr. 193/2013 (AWG). 

Altlastensanierungsgesetz 1989, idF österr BGBl. I Nr. 103/2013 (AlSAG). 

Forstgesetz 1975, idF österr BGBl. I Nr. 102/2015 (ForstG). 

Mineralrohstoffgesetz 1999, idF österr BGBl. I Nr. 80/2015 (MinroG). 

Deponieverordnung 2008, idF österr BGBl. II Nr. 104/2014 (DVO). 


Wildbachverbauungsgesetz, RGBl. Nr. 117/1884 (WLV-G). 

Qualitätszielverordnung Chemie Oberflächengewässer, österr BGBl. II Nr. 96/2006 (QZV 

Chemie OG). 

Verordnung zum Schutz der Oberflächengewässer (Oberflächengewässerverordnung vom 20. 

Juli 2011, dt BGBl. I S. 1429 (OGewV). 

Richtlinie 2008/98/EG des Europäischen Parlamentes und des Rates vom 19. November 2008 

über Abfälle (EU-ARL). 

Richtlinie 2000/60/EG des Europäischen Parlamentes und des Rates vom 23. Oktober 2000 

zur Schaffung eines Ordnungsrahmens für Maßnahmen der Gemeinschaft im Bereich der 

Wasserpolitik (EU-WRRL). 

Wasserrechtsgesetz 1959, idF österr BGBl. I Nr. 98/2013 (WRG). 

NORMATIVE VERWEISUNG 

DIN 4049-1 – Hydrologie, Grundbegriffe. Ausgabe: 1992-12. 

ÖNORM B 2400 - Hydrologie - Hydrographische Fachausdrücke und Zeichen - Ergänzende 

Bestimmungen zur ÖNORM EN ISO 772. Ausgabe: 2003-01-01. 

ÖNORM S 2126. Grundlegende Charakterisierung von Aushubmaterial vor Beginn der 

Aushub- oder Abräumtätigkeit. Ausgabe: 2010-08-15. 

ONR 24800. Schutzbauwerke der Wildbachverbauung - Begriffe und ihre Definitionen 

sowie Klassifizierung. Ausgabe: 2009-02-15. 



Physical modelling optimization of a filter check dam 

in Switzerland 

Sebastian Schwindt, MSc 1 ; Giovanni De Cesare, Dr. 2 ; Jean-Louis Boillat, Dr. 3 ; Anton J. Schleiss, Prof. Dr. 1 

ABSTRACT 

After a major flood with a peak discharge of 177 m³/s in the city of Martigny (Valais, Switzerland) 

by the alpine river Drance in October 2000, the local authorities are strengthening 

efforts for the protection of the urban area. The Drance has an annual average discharge of 10 

m³/s and the most important floods occur in the late summer. Two lateral torrents potentially 

supply the Drance with debris flow directly upstream of the urban area, where the installation 

of a filter check dam is foreseen for the retention of sediments in case of flood events. 

The functionality of the structure was analyzed on a physical model at a geometric scale of 

1:42 for different scenarios, i.e. clear water conditions, regular bedload and driftwood 

transport. The flow pattern and the sediment retention upstream of the filter check dam were 

analyzed and the energy dissipation downstream was optimized by modifications of the 

stilling basin. 

KEYWORDS 

filter check dam; physical model; retention of bedload; flood protection; driftwood. 

INTRODUCTION 

After a major flood event in the city of Martigny (Lower Rhone Valley, Switzerland) in 

October 2000, the local authorities are strengthening efforts for the protection of the urban 

area, like recently in the neighboring municipality of Riddes (Bianco et al. 2014). Martigny is 

crossed by the alpine river Drance, whose catchment area is under the influence of several 

hydropower reservoirs and which suffer from floods mainly in late summer, due to superimposed 

glacier melting and rainfall runoff. 

As indicated in Figure 1, the upper branch of the Drance River is flowing from south-east to 

north-west. Downstream of the village of Les Valettes, a canyon stretch with steep hill slopes 

receives five lateral torrents just upstream of the urbanized area of Martigny-Croix and 

Martigny. 

In the framework of the flood protection program, the trapping of sediment is foreseen at 

the downstream end of the canyon stretch by means of a filter check dam. The project site is 

confined laterally by steep hill slopes which impede the excavation of a retention basin. 

Therefore the height of the filter check dam is designed to retain the sediments in the canyon 

1 Ecole Polytechnique Federale de Lausanne, SWITZERLAND, sebastian.schwindt@epfl.ch 

2 Laboratoire de Constructions Hydrauliques - École Polytechnique Fédérale de Lausanne, Switzerland 

3 formerly Laboratoire de Constructions Hydrauliques - École Polytechnique Fédérale de Lausanne, Switzerland 



for a 100-year flood. In this study, the hydraulic behavior of the pre-designed filter check 

dam was analyzed and optimized by means of a physical model. 

Figure 1: Map of the region of interest, indicating the modelled river reach. Reproduced with the authorization of Swisstopo JA100120. 

METHODS 

Hydrological, geomorphological framework 

The river has been reproduced on an 850 m long reach with bed slopes varying between 

2.4 % and 3.1 %. The part upstream of the sediment control dam is about 600 m long and 

includes the inflows of the Lavanchy and St. Jean Torrents. The downstream part serves for 

the analysis of energy dissipation. The range of discharges tested in the experiments is shown 

in Table 1, where Qd denotes the design discharge for clogging of the orifice. 

Table 1: Range of discharges analyzed in the experiments in m³/s at prototype dimensions. Qd denotes the design discharge of opening 

clogging. 

Q mean Q d 

HQ 2.33 HQ 5 HQ 10 HQ 20 HQ 50 HQ 100 HQ 300 HQ 500 (EHQ) 

10 30 86 105 119 130 175 230 280 345 

Two different grain size distributions are applied in order to account for the differences 

between the coarser channel bed and the finer sediment supply of debris flow from the lateral 

torrents. The bed mixture is characterized by D30 = 29 mm, D50 = 92 mm and D90 = 435 mm, 


where the mean diameter is Dm = 165 mm. The supply mixture is characterized by 

D30 = 25 mm, D50 = 43 mm and D90 = 156 mm, where the mean diameter is Dm = 75 mm. 

Driftwood 

Since the catchment area of the Drance is partially covered by alpine forests, driftwood is 

likely to occur during floods. The blockage of hydraulic structures like the filter check dam 

and the bridges through the town of Martigny during floods have to be avoided. The behavior 

of driftwood and retention measures is studied by injecting piles of wood at the model 

entrance with different structural setups. The driftwood samples consist of 15 trunks of 10 m 

of length, 10 trunks of 15 m of length and 2 rootstocks. 

Scaling and model setup 

The model has a geometric scale of 1:42 and is about 21 m long and 2.8 m wide. In order to 

promote the participation of local stakeholders and the inhabitants of Martigny, the model 

was built in the neighborhood of the future establishment site. The geometric scale was 

chosen due to spatial limitations (upper size limit) and the respect of Froude and sediment 

transport similarities (lower size limit). 

The model is shown on Figure 2 with the upstream and downstream basins for water storage 

of 0.75 m³ and 4 m³ respectively, the two pumps and the chutes for the lateral torrents. 

Figure 2: View of the model with the lateral torrents, the supply pumps as well as the upstream and downstream water storage basins. 

The transported sediments, supplied at the upstream end of the channel, are captured in 

industrial filter bags which are suspended in the downstream basin. 


Layout of the filter check dam 

The initial configuration of the filter check dam implies two 10 m wide standard crest spillways, 

a stilling basin furnished with 1.5 m large blocks and a 4 m large central pile for the 

placement of a sluice gate covering the 4 m wide orifice of the filter check dam (prototype 

dimensions). The shape of the stilling basin is orientated at the standard design of a stilling 

basin. As the structure will be located in a right hand curve of about 24°, the crest axis of 

the spillways is inclined by 12° with respect to the upstream river axis for guiding the flow 

towards the downstream river axis. 

The geometry of the orifice is crucial because a too large opening will lead to excessive 

sediment transit during floods, i.e. self-flushing, offering therefore insufficient protection 

downstream. A too small opening on the contrary will retain too much sediments, leading to 

early or even permanent backfilling of the retention volume. The elements of the structure, 

described in Figure 3, were designed for a 100-year flood. 

Figure 3: Elements of the filter check dam: opening, central pile, crest spillways and stilling basin with 1.5 m large blocks (prototype 

dimension): a) dry model from upstream and b) lateral view during a 100-year flood. 

The effects of the structure on energy dissipation were analyzed with respect to the central 

pile (placed / suppressed), the implementation of a sluice gate in front of the orifice for the 

flushing control (opened / closed), the shape and length of the stilling basin (standard stilling 

basin / pear-shaped) and sediment transit through the filter check dam. 


Measurements 

The water level was recorded by means of ten ultrasonic probes and the water discharges 

were controlled individually for each pump by two flow meters. The sediment output was 

measured at the end of each experiment by means of an industrial scale. Video and picture 

documentation was constituted to highlight dynamic processes like the behavior of driftwood, 

the development of flow patterns and the sediment transport. 

Scenarios 

Combined with modifications of the layout, different discharge scenarios and solid transport 

events were studied. The whole range of discharges according to Table 1 was tested for the 

initial layout configuration with large central pile and standard-shaped stilling basin, under 

steady flow conditions without bedload. The focus of interest is the behavior of the structure 

in case of ordinary flow conditions (about 10 m³/s), frequent floods i.e. HQ 2.33 with 86 

m³/s, 100-year flood with 230 m³/s and extreme flood event with 345 m³/s which comply 

with 0.87 l/s, 7.52 l/s, 20.12 l/s and 30.18 l/s at model scale, respectively. 

For the 100-year and the extreme flood scenarios, the functioning of the structure must be 

guaranteed for ordinary bedload transport as well as with debris flow from the lateral 

torrents. The hydrograph and the corresponding sediment curve which apply for a 100-year 

flood event are shown in Figure 4. 

Figure 4: Showcase hydrograph and sediment curve of a 100-year flood, based on the flood from the year 2000. 

The sediment curve is based on the evaluation of bedload from the results of a 1D numerical 

simulation with Gesmat code at the model entrance. The peak of the solid discharge is about 

3.5 t/s, which correspond to about 18.5 kg/min at model scale. 


RESULTS 

Hydraulic control and sluice gate 

The hydraulic effects of the structure for the initial configuration with central pile and active 

orifice were studied experimentally and theoretically by means of a stage-discharge relation. 

The small width of the orifice (less than 50 % of the river width) imposes backwater for 

nearby all discharge conditions. 

Central Pile and Driftwood 

The central pile influences the backwater and induces driftwood clogging on the right bank 

spillway in 10 % of the tests. The presence of the pile has only minor effects on energy 

dissipation in the stilling basin. Initially, flow deflectors were foreseen for deviating the flow 

towards the left river bank directly upstream of the filter check dam, where the driftwood 

was intended to be retained in case of floods, according to Schmocker and Weitbrecht (2013). 

The authors propose to install vertical piles between the main channel bed and the river 

banks longitudinally to the river axis in an outer river bend, where vertically oriented 

vortices occur during floods and push the driftwood towards the outer bank. Thus, the 

driftwood can be trapped behind the barrier of piles. In the present case, due to the influence 

of backwater, the solution with flow deflectors cannot establish the suggested flow conditions. 

Another promising position for a pile structure was identified in an outer right hand bend 

about 350 to 450 meters upstream of the filter dam. Here, driftwood can be retained, but only 

if the pile structure also crosses the river axis. 

Stilling basin 

The key parameter for the evaluation of the efficiency of the stilling basin is the energy 

dissipation by means of water surface level and flow velocity measurements upstream and 

downstream of the stilling basin. The initial shape according to a “standard” stilling basin had 

the poorest performance in terms of energy losses. The energy dissipation was improved by 

adapting the geometry of the stilling basin to a pear-shape with the narrowing towards the 

downstream direction. Further improvements were achieved by increasing the depth of the 

basin by about 1.5 m and elongating the stilling basin by about 10 m in prototype dimensions. 

As the structure is situated in a right hand bend, the energy dissipation was even more 

pronounced by applying an asymmetric deepening of about 1 m more along the right river 

bank, as this measure makes the velocity profile more uniform at the downstream end of the 

stilling basin. The stages of improvement of the energy dissipation are illustrated in Figure 5, 

based on the energy head of a 100-year flood. 

The experiments have shown that the 1.5 m large boulders are mobilized during floods, thus 

requiring stabilization measures. 


Figure 5: Improvement of the stilling basin by adapting the shape, pool depth, length and 183 symmetry with respect to the longitudinal 

river axis based on the energy head during a 100- 184 year flood (HQ100). 

Sediment Transfer 

The sediment transfer through the filter check dam, notably the opening, is strongly linked 

to the flow pattern along the structure which was analyzed by injecting tracer color. The observations 

are documented in Figure 6 for a discharge of 30 m³/s. 

Figure 6: Flow pattern upstream of the filter check dam for a discharge of about 30 m³/s. 


Within the range of the tested discharges (minimum 10 m³/s), a quiet water zone established 

at all discharges upstream of the dam towards the left river bank. As shown in Figure 6, 

sediment deposits at the outer bound of the recirculation zone close the main current. 

The behavior of the structure during floods was studied by reproducing the hydrograph of 

Figure 4 by incremental steps and without closing the weir. The experiment was stopped at 

the flood peak for the analysis of the depositions. A large part of the sediment was retained 

upstream of the filter dam with a deposition slope of about 0.7 %. However, about 1/3 of the 

injected sediment passed the structure and was captured in the sediment bag. In prototype 

dimensions, about 20’000 m³ of sediments would have been directed towards the town of 

Martigny. 

CONCLUSIONS 

The initial layout of the filter check dam included standard spillways and a stilling basin 

designed for a 100-year flood (HQ 100 

) with a 4 m large central pile and orifice. The design of 

the basin was experimentally improved stepwise by adaptations of its shape, the length and 

the depth. An additional improvement was achieved by increasing the pool depth asymmetrically 

along the right bank, as the structure is situated in a right hand bend. 

The experiments have shown that about 2/3 of the sediments are retained by the structure 

during a 100-year flood. The rest can be safely transported through the channelized section in 

the town of Martigny. 

The actual design of the orifice probably requires a control measure, i.e. a sluice gate to avoid 

self-flushing in the falling limb of a flood event. Alternative remedy measures are conceivable, 

like the installation of vertical poles or horizontal beams in shape of a mechanical barrier 

upstream of the orifice. This alternative will be examined on the model too. 

Driftwood is partially blocked when the central pile applies. However, a retention zone about 

350 m to 450 m upstream of the filter check dam was identified for the installation of a 

driftwood retention device. 

Many of the existing open check dams were designed based on the own experience of the 

responsible engineer (Armanini and Larcher, 2001). For safety reasons, it was considered that 

it is more convenient for the filter check dam at the Drance River to design a narrow opening 

with a gate. However, the installation and the handling of a sluice gate during floods are 

critical issues. As an alternative, the enlargement of the opening with the implementation of 

a grid rack has to be considered but was not yet implemented in the framework of this study. 


REFERENCES 

- Armanini A., Larcher M. (2011). Rational criterion for designing opening of slit-check dam. 

Journal of hydraulic engineering 127: 94-104. 

- Bianco M., Bianco P., De Cesare G. (2014). Design of a bed load trap and driftwood filtering 

dam, analysis of the phenomena and hydraulic design. Proceedings of the Symposium CIPC 

KOHS 2014, Swiss Competences in River Engineering and Restoration: 129-139. 

- Piton G., Recking A. (2015). Design of Sediment Traps with Open Check Dams. I: 

Hydraulic and Deposition Processes. Journal of Hydraulic Engineering: 1-16. 

- Rickenmann D. (1997). Schwemmholz und Hochwasser. Wasser Energie Luft 89(5/6): 

115-119. 

- Schmocker, L., Weitbrecht, V. (2013). Driftwood: Risk Analysis and Engineering Measures. 

Journal of Hydraulic Engineering 139: 683-695. 

- Zollinger F. (1983). Die Vorgänge in einem Geschiebeablagerungsplatz: ihre Morphologie 

und ihre Möglichkeiten einer Steuerung. Dissertation Eidgenössische Technische Hochschule 

Zürich 7419: 264. 



Settlement dynamics in floodplains: from assessing 

future flood hazard exposure to developing spatial 

adaptation measures 

Walter Seher, Ph.D. 1 ; Lukas Löschner 1 

ABSTRACT 

Based on empirical research from three Austrian flood-prone municipalities, this contribution 

discusses the impact of settlement dynamics in floodplains on future changes in flood hazard 

exposure. The authors present a methodological framework for a GIS-based assessment of 

settlement dynamics in 300-year flooding areas and reflect options of implementing a 

risk-oriented approach in flood-related spatial planning. Findings indicate the following: i) 

increases in flood hazard exposure are to a large extent determined by the kind of building 

land (residential, commercial etc.) displayed in local land use plans; ii) the supply of building 

land (i.e. vacant lots and development areas) in floodplains is a key driver of future flood 

hazard exposure; iii) "context matters", i.e. flood-related options in spatial planning are 

strongly influenced by topographic conditions (e.g. the amount of land available for permanent 

settlement) and iv) there is further scope to specifically use zoning or local development 

plans for implementing a risk-oriented approach in flood-related planning. 

KEYWORDS 

Flood hazard exposure; settlement development; exposure assessment; risk-based spatial 

planning. 

INTRODUCTION 

In the past years flood policy in Austria gradually shifted from a structural, security-based 

approach of flood protection towards an integrated, risk-based approach of flood risk management. 

Whereas the 'traditional' approach was informed by a firm belief in controlling rivers 

via engineering solutions, Austrian flood policy increasingly aims at reducing the vulnerability 

to flooding (Nordbeck, 2014). Due to its pivotal importance for flood risk prevention, 

spatial planning is assuming a more central role within the nascent paradigm of an integrated 

management of flood risks. Spatial planning is operative in reducing vulnerability by 

minimizing the exposure to flood hazards. Hazard exposure can be effectively reduced by 

allocating demands for future land uses according to the suitability of locations. Evaluating 

suitability focuses on the question whether the flood risk for the location concerned is 

tolerable and acceptable. Existing zoning regulations in state spatial planning laws provide the 

necessary regulatory framework (Kanonier, 2005). Flood risk evaluation and the subsequent 

1 University of Natural Resources and Life Sciences, Vienna Institut of Spatial Planning and Rural Development, Vienna, AUSTRIA 

walter.seher@boku.ac.at 



land use decisions are usually based on the spatial extension of flood events with a defined 

level of occurrence (e.g. 100-year flood), thus regarding hazards rather than vulnerabilities. 

Despite significant progress to halt the continuous upward trend in the exposure of people 

and assets in flood-prone areas, settlement development, however, remains the strongest 

driver of flood-related economic damage and losses (Elmer et al., 2012). In particular the 

unchecked increase in damage potential in extreme hazard areas beyond the regulated 

flooding zones (e.g. 300-year flooding areas where land development is not regulated by 

statutory provisions in spatial planning laws) provides a fundamental challenge for policy 

makers and illustrates the need to anticipate settlement development in floodplains and to 

develop risk-based spatial planning approaches (see among others Hess, 2011; Greiving, 

2002). Risk-based spatial planning not only refers to hazards but also to vulnerability and 

potential damages when considering future land uses. A risk-based approach to spatial 

planning is calling for additional information, not only about hazards but also concerning 

hazard exposure, sensitivity and adaptive capacity. 

Against this background the authors present an analytical framework for a GIS-based 

assessment of settlement dynamics in 300-year flooding areas. Based on empirical findings 

from three Austrian case studies the paper discusses options of implementing a risk-oriented 

approach in flood-related spatial planning. The contribution addresses the following research 

questions: 

– What is the influence of settlement change in floodplains on future flood hazard exposure? 

– What are the main driving factors of settlement growth in floodplains? 

– What are (spatial planning) options to coordinate settlement development in floodplains 

and minimize the increase in damage potentials? 

METHODS 

This paper is based on empirical findings which were developed in the research project 

“RiskAdapt – Anticipatory Flood Risk Management under Climate Change Scenarios: From 

Assessment to Adaptation”. In the course of this project the authors conceptualised and 

applied a framework for tracing settlement trajectories in floodplains and assessing future 

flood hazard exposure (for the year 2030). Combining elements of scenario-planning, zoning 

and hazard exposure mapping and integrating different forms of knowledge (scientific, local, 

practical) on floodplain development, the framework was empirically tested in the following 

three Austrian case study municipalities to assess future changes in hazard exposure: 

Gleisdorf (Styria), Altenmarkt im Pongau (Salzburg) and Perg (Upper Austria). The results 

were presented in stakeholder workshops (involving academics and practitioners from spatial 

planning and water management as well as local and regional decision makers) and used as 

inputs for the discussion of anticipatory adaptation measures. 


To assess future changes in flood hazard exposure, the current exposure of risk elements 

needs to be known. Therefore, in a first step, the calculated flooding areas and inundation 

depths for 100-year flood events (HQ 100 

) and 300-year flood events (HQ 300 

) were intersected 

in ArcGIS with the (geo-referenced) federal building and housing register (2013). Based on 

the attributes “type of building” and “built-up area” it was determined to which extent e.g. 

residential, commercial, industrial or public buildings are exposed to the respective flood 

hazard scenarios. 

In a second step a scenario of settlement development (in potentially flood-prone areas) was 

generated for each case study until the year 2030. Based on the analysis of census and economic 

data as well as spatial planning instruments (i.e. municipal zoning plans and regional 

development plans) we calculated population and household trends, the availability of 

building land reserves and the expected demand for housing and commercial or industrial 

land uses. The settlement scenarios were complemented by in situ knowledge of local decision 

makers (mayors, chief officers and heads of the municipal building authorities) and the 

municipalities’ spatial planning consultants. In the course of in-depth interviews they were 

asked i) to comment on expected trajectories of land use change in their municipality, ii) to 

identify priority areas of settlement development and iii) to specify the development of vacant 

building plots or the expected demolition of buildings until the year 2030. Their input was 

used to check the plausibility of the scientific assumptions and to increase the robustness of 

the exposure scenarios. 

In a third step, the settlement scenarios were mapped in ArcGIS (see Figure 1) based on the 

nomenclature of the building and housing register (i.e. each new point feature was assigned 

attributes regarding the expected building type and built-up area). The digital cadastral map 

was used as a reference frame to ensure the best possible localisation of the new buildings 

within the building lot. 

The above-outlined assessment of future exposure was presented as part of an integrated 

flood risk assessment in each case study area in the form of scientist-stakeholder workshops 

(see Löschner et al., 2015). The aim of the workshops was to discuss the assessment and 

management of current and future flood risks, in particular by i) reflecting determinants of 

risk based on different scenarios (maps), ii) identifying/verifying local context conditions and 

pre-existing policy processes, and iii) developing and prioritizing adaptive measures for 

extreme flooding scenarios. 

To stimulate discussion, the workshops began with a presentation of the quantifiable results 

of the flood risk assessment. This scientific input was delivered in an interactive setting via 

plotted maps and aimed at providing an impetus for the discussion and development of 

adaptation measures in a World Café setting. Accordingly, the discussion was grouped into 

three roundtables (with three to five representatives) each having a different focal point 


Figure 1: GIS-based land development scenario in potentially flood-prone areas. 

(based on the cycle of flood risk management): structural measures of flood protection; 

planning measures of flood prevention and coping measures to increase flood preparedness. 

At the end of the World Café, the moderators presented all adaptive measures from the 

roundtable discussions in the plenum. In a next step, the participants prioritized the different 

measures according to their own preferences. 

RESULTS 

This section presents key results of the exposure assessment as well as findings from the three 

stakeholder workshops concerning proposed planning measures of flood prevention. Based 

on the respective land development scenarios, the GIS-based assessment of future changes in 

flood hazard exposure indicates that all three case studies can expect an increase in potentially 

affected buildings until the year 2030 (see figure 2). 

Specifically, the following findings can be derived from the analysis: 

– Current protection against 100-year flood events: The current (2015) number of buildings 

in 100-year flooding areas reflects the effectiveness of the local flood-protection scheme. 

In the case of Gleisdorf current building stock is only marginally affected (due to a 

retention basin and linear flood protection measures). In Perg, on the other hand, where 

flood protection infrastructure is largely missing, more than 175 buildings with a total area 

of around 4 ha are flooded in HQ 100 

. 


Figure 2: Buildings located currently (2015) and in the year 2030 in 100- and 300-year flooding areas 

– Current exposure to 300-year flood events: All three cases have large building stock (2015) 

in 300-year flooding areas. In Gleisdorf, the exposed buildings consist mainly of large-scale 

commercial and industrial buildings which are located adjacent to the flood-protected 

areas. In the other two case studies the potentially affected buildings are overwhelmingly 

residential buildings (i.e. single-family homes or apartment buildings). 

– Future settlement development in 100-year flooding areas: Depending on state planning 

law and local zoning plans, the three municipalities can expect varying land development 

in 100-year flooding areas until the year 2030. According to the scenario in Gleisdorf no 

new buildings will be erected in these areas. In Perg around fifteen new (apartment) 

buildings shall be built until 2030, in Altenmarkt around eighteen (residential, commercial 

and industrial buildings) shall be developed, leading to a sharp future increase in flood 

hazard exposure in 100-year flooding areas. 

– Future settlement development in 300-year flooding areas: In all municipalities high 

settlement dynamics can be expected in 300-year flooding areas until 2030. In Perg, future 

land development in these areas (50 buildings/ca. 9.400 m 2 ) is mainly due to the construction 

of small-scale residential buildings. In Gleisdorf, on the other hand, a total of nine new 

large-scale manufacturing and commercial buildings (total area of around 10.400 m 2 ) can 

be expected; whereas in Altenmarkt future land development (30 buildings/ca. 11.300 m 2 ) 

reflects a mix of both residential and commercial/industrial uses. 

Despite differences in topography, land use and hazard potential, two general conclusions 

regarding the driving factors of future flood hazard exposure can be drawn from the comparison 

of the three case studies. First, the different type and dynamic in future land development 

in flood hazard areas indicates that increases in flood hazard exposure are to a large 

extent determined by the kind of building land (residential, commercial etc.) displayed in the 

zoning plans. This becomes evident from the average area affected per building (see figure 2), 

as it shows that the three municipalities set different priorities (e.g. commercial/industrial 

vs. residential land uses) in land development. Secondly, findings indicate that future flood 


hazard exposure depends on the supply of suitable building land (i.e. the amount of vacant 

lots and development areas). While some municipalities have the possibility to provide 

building land in areas outside of flood hazard zones, others are constrained i.a by alpine 

topography and thus have no alternative but to develop valley basins all the while keeping 

the hazard exposure and the flood risk as low as possible. 

These constraints were also partly reflected in the stakeholder workshops. Along with 

structural measures of flood protection as well as coping measures to increase flood preparedness, 

the stakeholders proposed a range of planning measures to mitigate the likely future 

increase in flood hazard exposure. In sum there was a strong consensus to display 300-year 

flooding zones in local land use plans in order to improve the information base for extreme 

flood events. To reduce an increase in damage potential in hazard zones many participants 

also demanded that extreme flood events should be considered in issuing building permits, in 

order to foster flood-proof building adaptations in these areas. Finally, workshop participants 

advocated a harmonization of hazard data available, more widespread application of local 

land use plans to better allocate land for retention basins or flooding corridors and to allow 

for a gradual land development, from low to high flood hazard areas. Generally the proposed 

planning measures can complement existing flood-related statutory provisions in spatial 

planning laws. 

CONCLUSIONS 

The findings presented were generated on the basis of case study analyses. As all case studies 

can expect high population, economic and settlement growth until the year 2030 they each 

face an increase in flood hazard exposure. However, the different levels in increase indicate 

that “context matters”. Development options are strongly influenced by topographic 

conditions. The amount of land suitable for permanent settlement is a limiting factor for 

spatial planning in general and flood-related planning in particular. Besides topography the 

availability of suitable building land, i.e. vacant lots and development areas, inside and 

outside of potential floodplains is a key driver of future flood hazard exposure. Where 

alternatives to floodplain development are rare an increase in flood hazard exposure can be 

expected, especially in areas where building bans are not in force due to low probability and 

intensity of flooding. Finally, increases in flood hazard exposure are to a large extent 

determined by the kind of development intended. As the three cases were chosen to reflect 

different spatial types (i.e. alpine, peri-urban, urban) the results concerning the drivers of 

settlement growth in floodplains seem to be generally applicable to other floodplain situations. 

The respective characteristics of flood hazard, however, are unique to each case. 

Despite fundamentally different spatial contexts, the results of the stakeholder workshops 

show that preventing an increase in damage potential is a challenge common to flood-prone 

municipalities. The proposed planning options, however, only partially reflect the increase of 

flood hazard exposure caused by settlement growth both in areas with low flooding probabili- 


ty (i.e. 300-year flooding areas) and in areas of residual risk (i.e. areas protected up to a 

certain degree of flooding). Proposed measures, like harmonizing hazard data or allocating 

land for retention basins or flooding corridors focus on flood hazards rather than on vulnerabilities. 

A risk-based approach in flood-related spatial planning calls for an adjustment of land 

use intensities (influencing flood hazard exposure) according to the probability and intensity 

of flood hazards. Land use adjustment in floodplains is reflected in proposed measures like 

considering extreme flood events in issuing building permits or zoning land use intensities 

according to the level of flood hazards. Furthermore, local development planning fits into the 

requirements of a risk-based planning approach. Local development plans are implemented 

to specify the layout of building land in areas to be developed, thus enabling regulations to 

ensure flood-adapted buildings and infrastructure. Participatory development planning is able 

to spread information about existing flood risks and thus contributes towards increasing risk 

awareness of people involved. Collaborative approaches integrating flood protection and 

development planning offer possibilities to reduce an increase in damage potential by 

reducing hazard exposure and raising adaptive capacity. 

In order to cope with settlement dynamics in floodplains flood managers will have to broaden 

their portfolio and have to move beyond flood protection (via technical measures) and risk 

prevention (via the allocation of land uses) and strengthen (local) adaptive capacities, including 

the leverage of spatial planning instruments to develop flood-adapted buildings and 

land uses. The EU Flood Directive (2007/60/EC) provides a strong impetus for the shift towards 

an integrated flood risk management. The directive is currently in the final stage of 

implementation with flood risk management plans being completed. As flood risk management 

plans aim to reduce the likelihood and the adverse consequences of flooding measures 

of risk-based spatial planning should be considered in order to correspond to the demand of 

White (1942) to “readjust land occupance and floodplain phenomena in a harmonious 

relationship”. 

REFERENCES 

- Directive 2007/60/EC of the European parliament and of the council of 23 October 2007 

on the assessment and management of flood risks. 

- Elmer, E., Hoymann, J., Düthmann, D., Vorogushyn, S., Kreibich, H. (2012): Drivers of 

flood risk in residential areas. Natural Hazards and Earth System Sciences 12: 1641-1657. 

Greiving, S. (2002): Räumliche Planung und Risiko. Gerling Akademie Verlag, München. 

- Hess, J.T. (2011): Schutzziele im Umgang mit Naturrisiken in der Schweiz. vdf Hochschulverlag 

AG, Zürich. 

- Kanonier, A. (2005): Naturgefahren im österreichischen Raumordnungsrecht. In: 

Österreichische Raumordnungskonferenz (ÖROK) (Hrsg.): Präventiver Umgang mit Naturgefahren 

in der Raumordnung. Materialienband, Schriftenreihe der Österreichischen 

Raumordnungs- konferenz Nr. 168, Wien. 


- Löschner, L., Nordbeck, R., Scherhaufer, P., Seher, W. (2015): Scientist-stakeholder 

workshops: a collaborative approach for integrating science and decision-making in Austrian 

flood-prone municipalities. Environmental Science and Policy. Available online at: 

http://www.sciencedirect.com/science/article/pii/S1462901115300575 

(accessed 10 September 2015). 

- Nordbeck, R. (2014): Klimawandel und vorsorgender Hochwasserschutz in Österreich: eine 

entwicklungsdynamische Analyse der Anpassungskapazitäten (2002-2012). InFER Discussion 

Paper 2-2014, Vienna. 

- White, G.F. (1942): Human adjustment to floods. Research Paper No. 29, University of 

Chicago. 


The findings presented in this paper were developed in the project RiskAdapt (Anticipatory 

Flood Risk Management under Climate Change Scenarios: From Assessment to Adaptation; 

grant number: KR11AC0K00275). The project was funded within the Austrian Climate 

Research Program (ARCP) by the Austrian Climate and Energy Fund. For more information, 

see https://riskadapt.boku.ac.at/. 



Flood losses in Switzerland compared with 

hazard maps 

Überschwemmungsschäden in der Schweiz im 

Vergleich mit Gefahrenkarten 

Luzius Thomi, PhD, MA 1 ; Matthias Künzler, MSc 1 ; Raoul Kern, MSc 1 

ABSTRACT 

This paper compares spatial claim data of the Swiss Mobiliar Insurance Company with flood 

hazard maps. We analyse the locations of inundation claims as a function of flood hazard 

zones. Moreover, we quantify the ratios of claims caused by floods, backwater in the 

canalisation, and groundwater. 

About 40 % of the analysed claims are due to floods (including surface runoff). Groundwater 

and, to a much higher extent, backwater are responsible for the remaining claims. One of two 

flood claims occurs in the so called white or "no danger" zone. These claims are supposed to 

be caused by surface runoff. 

To improve flood risk management, more knowledge is needed about the processes of surface 

runoff and backwater. Additionally, the affected areas have to be identified and mapped. 

Moreover, raising the awareness of surface runoff and backwater is necessary. 


Dieser Beitrag vergleicht Schadendaten der Schweizerischen Mobiliar Versicherungsgesellschaft 

mit Hochwassergefahrenkarten. Er geht der Frage nach, in welchen Gefahrenzonen 

sich die Überschwemmungsschäden befinden und in welchem Verhältnis Schäden durch 

Hochwasser, Rückstau aus der Kanalisation und Grundwasser stehen. 

Nur rund 40 % aller Überschwemmungsschäden sind durch Hochwasser (inkl. Oberflächenabfluss) 

bedingt. Grundwasser und hauptsächlich Rückstau sind für die restlichen Schäden 

verantwortlich. Von den eigentlichen Hochwasserschäden wiederum befindet sich die Hälfte 

in der weissen, gefahrenfreien Zone. Hierbei dürfte es sich um Schäden durch Oberflächenabfluss 

handeln. 

Für einen umfassenden Schutz vor Überschwemmungen müssen die Gefährdungen durch 

Oberflächenabfluss und Rückstau besser verstanden und z. B. in Form einer Karte bezeichnet 

werden. Zudem muss das Bewusstsein für Oberflächenabfluss und Rückstau gesteigert 

werden. 

KEYWORDS 

floods; losses; hazard maps; surface runoff; insurance 

1 Swiss Mobiliar Insurance Company, Bern, SWITZERLAND, luzius.thomi@mobiliar.ch 



EINFÜHRUNG 

In den letzten zehn Jahren haben Hochwasser und Murgänge in der Schweiz Sachschäden 

von mehr als CHF 4 Milliarden (WSL 2014) verursacht. Die Schweizerische Mobiliar Versicherungsgesellschaft 

(kurz Mobiliar) versichert rund jeden dritten Haushalt und ist damit 

grösster Sachversicherer der Schweiz. Sie war deshalb von diesen Unwettern besonders 

stark betroffen. 

Überschwemmungsschäden werden nicht nur durch hochwasserführende Gewässer 

verursacht, sondern auch durch Rückstau in der Kanalisation und hohe Grundwasserspiegel. 

Nachfolgend werden die Resultate einer schweizweiten Auswertung von Überschwemmungsschäden 

dargestellt und diskutiert. Durch die Verknüpfung mit Gefahrenkarten können 

Aussagen zur Verteilung der Überschwemmungsschäden gemacht werden. 

BEGRIFFE UND FRAGESTELLUNG 

Überschwemmungen können direkte Schäden (z. B. an Gebäuden, Fahrzeugen, landwirtschaftlichen 

Kulturen) und indirekte Schäden (z. B. Betriebsausfälle, temporäre Schliessung 

wichtiger Verkehrsachsen) verursachen. Der vorliegende Artikel beschränkt sich auf direkte 

versicherte Überschwemmungsschäden an Gebäuden und Gebäudeinhalt (Fahrhabe), wobei 

folgende Definitionen verwendet werden (Abweichungen zu anderen Definitionen in den 

Bereichen Naturgefahren oder Versicherungswirtschaft sind möglich): 

– Überschwemmungsschäden: 

Oberbegriff für alle Schäden durch hochwasserführende Gewässer, Oberflächenabfluss, 

Rückstau aus der Kanalisation und Grundwasser. 

– Hochwasserschäden: 

Dabei handelt es sich um Schäden, die durch oberirdisch eintretendes Wasser verursacht 

werden, d. h. durch ausufernde Gewässer und Oberflächenabfluss. Die beiden Prozesse 

werden in der Sachversicherung nicht unterschieden. Die Verordnung über die Beaufsichtigung 

von privaten Versicherungsunternehmen (Aufsichtsverordnung, AVO; SR 961.011) 

vom 9.11.2005 zählt die Hochwasserschäden zu den so genannten Elementarschäden (Art. 

173 Abs. 1). 

– Rückstauschäden: 

Schäden durch Wasser, das aufgrund eines Rückstaus aus der Kanalisation ins Gebäude 

eintritt. Gemäss der AVO sind Schäden durch Rückstau von Wasser aus der Kanalisation 

keine Elementarschäden, egal welche Ursache sie haben (Art. 173 Abs. 3 lit. b). 

– Grundwasserschäden: 

Schäden, die durch einen hohen Grundwasserspiegel entstehen. Sie gelten gemäss AVO 

ebenfalls nicht als Elementarschäden (Art. 173 Abs. 3 lit. a). 

Während in 19 von 26 Kantonen Elementarschäden an Gebäuden – und damit auch 

Hochwasserschäden – durch kantonale Gebäudeversicherungen abgegolten werden, sind 

Schäden an der Fahrhabe (ausser in den Kantonen Waadt und Nidwalden, in denen ebenfalls 


ein staatliches Monopol gilt) wie auch die Gebäudeschäden in den restlichen sieben Kantonen 

und im Fürstentum Liechtenstein durch private Versicherungen gedeckt. Bei Rückstau 

und Grundwassersaufstoss besteht kein staatliches Monopol. Entsprechende Schäden werden 

somit mehrheitlich durch die Privatassekuranz versichert. 

Der vorliegende Artikel geht der Frage nach, wie Überschwemmungsschäden in Bezug auf 

Gefahrenzonen verteilt sind und wie sie sich zusammensetzen: 

– Wie sind die Überschwemmungsschäden bezüglich der Gefahrenzonen verteilt? 

– Wie lässt sich diese Verteilung erklären? 

– Welches Verhältnis besteht zwischen Schäden durch Hochwasser, Rückstau und Grundwasser? 

METHODEN 

Für die Auswertung der Überschwemmungsschäden wurden folgende Datensätze verwendet: 

– Schadendaten der Mobiliar: 

– Ausgewertet wurden alle Überschwemmungsschäden an Gebäuden und Fahrhabe 

zwischen 2003 und Ende Juni 2015 in der Schweiz und im Fürstentum Liechtenstein, die 

adressgenau vorliegen und geocodiert werden können (insgesamt rund 68‘000). Schäden 

an Fahrzeugen wurden nicht berücksichtigt. 

– Gefahrenkarten für Hochwasser: 

Die Karten wurden bei den Kantonen bezogen (Stand: Juni 2015) und zusammengeführt. 

Ausser für den Kanton Waadt standen für jeden Kanton Gefahrenkarten für die Auswertung 

zur Verfügung, zumindest für einen Teil des besiedelten Gebiets. Die Gefahrenkarten 

bezeichnen die durch Ausuferung von Gewässern überschwemmbaren Flächen in fünf 

Gefahrenstufen (siehe Tabelle 1; zur Machart der Schweizer Gefahrenkarten, siehe 

PLANAT 2012). Überschwemmungsflächen durch Oberflächenabfluss sind in der Regel 

nicht abgebildet. 

– GeoPost Coordinates der Schweizerischen Post AG: 

Das koordinatengenaue Adressverzeichnis umfasst alle postalisch bedienten Gebäude der 

Schweiz und beinhaltet u. a. Angaben zur Anzahl Haushalte pro Adresse (Stand des 

Datensatzes: Juli 2015). 

– Landschaftsmodell „swiss TLM3D“ des Bundesamts für Landestopografie swisstopo (Version 

1.3), Topic Bauten. Es umfasst u. a. sämtliche Gebäude der Schweiz und des Fürstentums 

Liechtenstein in vektorieller Form. 

– Mithilfe einer GIS-Applikation wurden die Schaden- und Adressdaten (Punktdatensätze) 

sowie die Gebäude (Polygondatensatz) mit den Gefahrenkarten räumlich verknüpft. 

Dadurch konnten Erkenntnisse gewonnen werden, wie sich die Schäden, Adressen und 

Gebäude auf die Gefahrenzonen verteilen. Gebäude, die in mehr als einer Gefahrenzone 

liegen, wurden der höchsten Gefahrenzone zugeordnet. 

Die Schadenorte wurden mit den Adressdaten der Post geocodiert. Der Punkt liegt in der 

Regel beim Gebäudeeingang. Für die Beurteilung der Gefährdung ist aber nicht der Gebäude- 


eingang relevant, sondern der Gebäudegrundriss. Um Verzerrungen zu verhindern, wurde 

geprüft, wo die Punkte in den jeweiligen Gefahrenzonen liegen. Dabei zeigte sich, dass die 

Verteilung der Schadenorte innerhalb der Gefahrenzonen vergleichbar ist. Einzig in der 

weissen Zone liegen die Punkte im Schnitt etwas weiter weg von der nächsten Gefahrenzone, 

was sich durch die grosse räumliche Ausdehnung der weissen Zone erklärt. Die Auswertung 

von Punktdaten wird somit als aussagekräftig beurteilt. Zwar gibt es Fälle, in denen ein Teil 

eines Gebäudes nicht in derselben Gefahrenzone liegt wie der Adresspunkt. Über die grosse 

Anzahl an Schadendaten und die ähnliche Verteilung der Punkte in den Gefahrenzonen 

dürfte sich dies aber ausgleichen. Ganz abgesehen davon weisen die Gefahrenkarten in der 

Regel keine metergenaue Präzision auf. 

ERGEBNISSE 

Die verfügbaren Hochwassergefahrenkarten decken 64 % aller Gebäude und 68 % aller 

Haushalte der Schweiz ab (siehe Tabelle 1). Die Verteilung nach Gefahrenzone der bei der 

Mobiliar versicherten Haushalte und Firmen ist im Wesentlichen mit jener von Gebäuden 

und Haushalten vergleichbar. 

Tabelle 1: Prozentualer Anteil der Gebäude (n = 2.56 Mio.) und Haushalte (n = 4.27 Mio.) nach Gefahrenzone (zur Definition der 

Gefahrenzonen, siehe PLANAT 2012). 

Gefahrenzone Bedeutung Gebäude Haushalte 

Nicht untersucht Keine Gefahrenkarte vorhanden 36 % 32 % 

Weisse Zone Keine oder vernachlässigbare Gefährdung 48 % 53 % 

Gelbweisse Zone Restgefährdung: Gefährdung bei sehr seltenen Ereignissen 4 % 5 % 

Gelbe Zone 

Blaue Zone 

Geringe Gefährdung: mittelhäufige Gefährdung mit geringer 

Intensität 

Mittlere Gefährdung: häufige Gefährdung mit geringer 

Intensität bzw. mittelhäufige Gefährdung mit mittlerer Intensität 

7 % 6 % 

4 % 3 % 

Rote Zone Erhebliche Gefährdung: generell starke Intensität 1 % 1 % 

Die Hälfte aller Schäden und etwa ein Viertel der verursachten Kosten liegen in der weissen 

Zone. Zudem sticht die gelbe Zone heraus, in der die Schäden punkto Anzahl und Kosten 

deutlich über jenen der roten und blauen Zone zusammen liegen (siehe Tabelle 2). 

Tabelle 2. Prozentualer Anteil der Hochwasserschäden nach Gefahrenzone. Schäden in nicht untersuchten Gebieten (ohne Gefahrenkarte) 

sind nicht berücksichtigt. 

Gefahrenzone Anzahl Schäden Kosten 

Weisse Zone 50 % 24 % 

Gelbweisse Zone 10 % 16 % 

Gelbe Zone 23 % 34 % 

Blaue Zone 15 % 22 % 

Rote Zone 2 % 4 % 


Unter der Annahme, dass die Gefahrenkarten korrekt erstellt worden sind, heisst das, dass die 

Schäden in der weissen Zone vermutlich durch Oberflächenabfluss (z. B. aus landwirtschaftlichen 

Flächen oder aufgrund unzureichender Siedlungsentwässerung) entstanden sind. Auch 

wenn man bedenkt, dass rund ein Fünftel der Schäden in der weissen Zone weniger als 10 m 

von einer Gefahrenzone entfernt liegt, dürfte die Grössenordnung von 50 % Schäden durch 

Oberflächenabfluss stimmen: Auch in gelben, blauen und roten Zonen kann es zu Schäden 

durch Oberflächenabfluss kommen. Diese können aber anhand der zur Verfügung stehenden 

Daten nicht quantifiziert werden. 

Hochwasserschäden kosten zwar am meisten, Rückstauschäden treten aber am häufigsten 

(47 %) auf, vor Hochwasserschäden (40 %) und Grundwasserschäden (13 %). Es fällt auf, 

dass Hochwasser- und Rückstauschäden oft gleichzeitig entstehen. Dies ist insofern nicht 

erstaunlich, als Rückstau und Hochwasser häufig dieselbe Ursache haben, nämlich langanhaltende 

und/oder intensive Niederschläge. Trotz Monopol der kantonalen Gebäudeversicherungen 

für die Deckung von Hochwasserschäden an Gebäuden in 19 von 26 Kantonen dürfte die 

Grössenordnung der oben genannten prozentualen Anteile der Überschwemmungsschäden 

stimmen. Denn wird ein Gebäude durch Hochwasser beschädigt, ist in vielen Fällen auch die 

Fahrhabe betroffen. Damit ist der Schadenfall in der vorliegenden Auswertung berücksichtigt. 

FALLBEISPIELE 

Die schweizweiten Ergebnisse werden nachfolgend anhand von drei konkreten Fallbeispielen 

illustriert. 

Altstätten (Kanton St. Gallen): 

Am 28. Juli 2014 führten in der St. Galler Gemeinde Altstätten Starkniederschläge zu hohen 

Abflüssen und in der Folge zu Ausuferungen. Weit über 90 % aller durch die Mobiliar 

getragenen Überschwemmungsschäden sind auf Hochwasser zurückzuführen. Schäden durch 

Rückstau oder Grundwasser gab es kaum (siehe Abbildung 1). Die Hochwasserschäden 

wurden grösstenteils durch die Ausuferung des Stadtbachs verursacht. 92 % davon befinden 

sich in der roten, blauen, gelben oder gelbweissen Zone, wobei knapp die Hälfte der Schäden 

in der blauen Zone liegt. Die Schadenorte stimmen also gut mit der Hochwassergefahrenkarte 

überein. 

Marly (Kanton Freiburg): 

Die Freiburger Gemeinde Marly war in den letzten Jahren mehrmals von Unwettern 

betroffen. Zwar ist das Siedlungsgebiet durch mehrere Bäche bedroht. Dennoch machen die 

Schäden in den Gefahrenzonen – die mit grosser Wahrscheinlichkeit auf die Bäche zurückzuführen 

sind – nur etwa einen Drittel aller Hochwasserschäden aus. Zwei Drittel befinden sich 

abseits von Gewässern und sind somit vermutlich durch Oberflächenabfluss verursacht 

worden. Die Hochwasserschäden wiederum machen nur etwa die Hälfte aller Überschwemmungsschäden 

aus (siehe Abbildung 1). Der Fall der Gemeinde veranschaulicht die Proble- 


matik also sehr deutlich: Mit den heutigen Gefahrenkarten lässt sich nur ein Teil möglicher 

Überschwemmungsgebiete erkennen. 

Rubigen (Kanton Bern): 

Am 20. Juli 2007 haben intensive Niederschläge in der Gemeinde Rubigen zu starkem Oberflächenabfluss 

geführt. 90 % aller Hochwasserschäden liegen in der weissen Zone. Nur eine 

kleine Anzahl an Schäden ist durch einen hochwasserführenden Bach verursacht worden. 

Abbildung 1: Aufteilung der Überschwemmungsschäden in Hochwasser- bzw. Rückstau- und Grundwasserschäden. 

DISKUSSION 

Die Gefahrenkarte gilt heute als die wichtigste Grundlage, was die Erkennung von Überschwemmungsgebieten 

angeht. Dass sie sich auf Ausuferungen von Gewässern beschränkt, 

dürfte ausserhalb von Fachkreisen kaum bekannt sein. Daraus ergibt sich ein lückenhafter 

Umgang mit Überschwemmungsgefahren und somit ein ungenügender Schutz vor Schäden. 

Problem gelbe Zone: Zwar sind die Überschwemmungsintensitäten in der gelben Zone 

meist gering, trotzdem gibt es hier mehr Schäden als in der blauen und roten Zone zusammen. 

Heute sind raumplanerische (z. B. Ausscheidung neuer Bauzonen) und baurechtliche 

Massnahmen (z. B. Bauauflagen) – zum Teil sogar ganze Hochwasserschutzprojekte – in der 

Regel auf die Beseitigung des Risikos in der blauen und roten Zone ausgerichtet. Dies ist 

insofern sinnvoll, als damit Schäden durch Prozesse mit hohen Intensitäten (d. h. hohe 

Fliessgeschwindigkeiten oder grosse Überschwemmungstiefen) vermindert werden, insbeson- 


dere auch Personenschäden. Im Sinne einer wirksamen Schadenminderung reicht es aber 

nicht aus, gelbe Zonen als reine Hinweiszonen (vgl. ARE et al. 2005) zu bezeichnen, also als 

Zonen, in denen etwa Grundeigentümer lediglich auf die Gefährdung aufmerksam gemacht 

werden, ohne konkrete Einschränkungen zu erlassen. Es braucht adäquate planerische, 

organisatorische und/oder bauliche Massnahmen, um das Risiko zu senken, z. B. durch 

bauliche Auflagen, freihalten von Abflusskorridoren, risikobasierte Raumplanung (vgl. ARE 

et al. 2005, Camenzind & Loat 2014) usw. 

Problem Oberflächenabfluss: Der Oberflächenabfluss macht rund die Hälfte aller Hochwasserschäden 

aus, wird jedoch in den meisten Gefahrenkarten nicht ausgewiesen. Als 

Konsequenz daraus wird die weisse Gefahrenzone als gefahrenfrei verstanden, was de facto 

vielerorts nicht der Fall ist. Um dem Problem Herr zu werden, braucht es ein besseres 

Verständnis des Prozesses Oberflächenabfluss und eine Übersichtskarte, welche die betroffenen 

Gebiete ausweist. Zudem müssen Gemeinden, Architekten und Grundeigentümer stärker 

sensibilisiert und geeignete Anreize geschaffen werden, damit Massnahmen gegen Oberflächenabfluss 

auch umgesetzt werden. 

Problem Rückstau: Schäden durch Rückstau in der Kanalisation übersteigen anzahlmässig 

die Hochwasserschäden. Sie sind nicht ein rein technisches Problem der Siedlungsentwässerung, 

sondern stehen oft in direktem Zusammenhang mit hohen Niederschlagsintensitäten 

und grossen Abflüssen. Um Massnahmen ergreifen zu können, müssen die Gebiete identifiziert 

werden, in denen es zu Rückstauschäden kommen kann. Denkbar sind etwa Übersichtskarten, 

wie sie z. B. der Kanton Zürich erstellt hat (vgl. Schuler 2014). 

Problem Kommunikation: Gefahrenkarten berücksichtigen in der Regel weder Oberflächenabfluss 

noch Rückstau in der Kanalisation. Weisse Zonen werden gleichgesetzt mit 

„keine oder vernachlässigbare Gefährdung“ (vgl. Bundesamts für Umwelt BAFU, 

http://www.bafu.admin.ch, Thema Naturgefahren, konsultiert am 21.8.2015). Auch wenn 

dies bei einer eingeschränkten Betrachtungsweise auf die untersuchten Prozesse – nämlich 

die hochwasserführenden Gewässer – korrekt ist, ist es irreführend. Es braucht eine umfassendere 

Kommunikation, insbesondere für Gemeinden und die betroffene Bevölkerung, 

damit Überschwemmungsgefahren in ihrer Ganzheit erkannt und Schäden frühzeitig 

verhindert werden. 

FAZIT 

Die Auswertung von rund 68‘000 Schäden durch Hochwasser, Rückstau und Grundwasser 

zeigt, dass nur 40 % aller Schäden durch Hochwasser bedingt sind und davon wiederum nur 

etwa die Hälfte durch die Ausuferung von Gewässern. Hochwasserschäden in der gelben 

Gefahrenzone sind anzahl- und kostenmässig bedeutender als in der blauen und roten Zone 

zusammen. Am meisten Schäden entstehen durch Rückstau in der Kanalisation. Allerdings 


ist der Durchschnittsschaden durch Rückstau und Oberflächenabfluss deutlich geringer als 

jener durch Ausuferung von Gewässern. 

Für einen nachhaltigen Umgang mit Überschwemmungsrisiken leiten sich folgende Schlussfolgerungen 

ab: 

– Die Gefahrenkarten müssen auch das Gefahrenpotenzial von Oberflächenabfluss sowie von 

Rückstau und Grundwasseraufstoss aufzeigen. Zudem muss das Prozessverständnis von 

Oberflächenabfluss verbessert werden. 

– Massnahmen raumplanerischer und baurechtlicher Art dürfen sich nicht auf die blaue und 

rote Zone beschränken. Es braucht Anreize zur Reduktion der Verletzlichkeit von 

Gebäuden gegenüber Überschwemmungen in der gelben, gelbweissen und weissen Zone. 

– Die Kommunikation, insbesondere gegenüber Nicht-Fachleuten, muss angepasst werden. 

Gefahrenkarten zeigen nicht abschliessend, ob eine Überschwemmung möglich ist oder 

nicht. Ohne eine entsprechende Sensibilisierung wird insbesondere in der weissen Zone 

ein Gefühl falscher Sicherheit vermittelt. 

LITERATUR 

- ARE, BWG, BUWAL (2005). Raumplanung und Naturgefahren. Empfehlung. Bern, 

Bundesamt für Raumentwicklung (ARE), Bundesamt für Wasser und Geologie (BWG), 

Bundesamt für Umwelt, Wald und Landschaft (BUWAL). 

- Camenzind R., Loat R. (2014). Risikobasierte Raumplanung – Synthesebericht zu zwei 

Testplanungen auf Stufe kommunaler Nutzungsplanung. Bern, Nationale Plattform Naturgefahren 

(PLANAT), Bundesamt für Raumentwicklung (ARE), Bundesamt für Umwelt (BAFU). 

- PLANAT (2012). Lesehilfe für gravitative Gefahrenkarten. Bern, Nationale Plattform 

Naturgefahren (PLANAT). 

- Schuler Ch. (2014): Gefahrenkarten Kanton Zürich, Lesehilfe. Zürich, Amt für Abfall, 

Wasser, Energie und Luft (AWEL). 

- Verordnung über die Beaufsichtigung von privaten Versicherungsunternehmen (Aufsichtsverordnung, 

AVO) vom 9. November 2005 (Stand am 1. Januar 2013), SR 961.011. 

- WSL (2014): Development of damage 1972 2013 (taking inflation into account). Website: 

http://www.wsl.ch/fe/gebirgshydrologie/HEX/projekte/schadendatenbank/index_EN, last 

update: 13.03.2014, accessed 03.03.2015. Swiss Federal Institute for Forest, Snow and - 

Landscape Research (WSL) 



Controlled and efficient bed load management 

by means of variable drain locks embedded in a 

crownclosed large drain sediment control dam 

Aktive und effiziente Geschiebebewirtschaftung 

mit Hilfe variabler Dolenverschlüsse an einer 

kronengeschlossenen großdoligen Bogensperre 

Arthur Vogl, DI 1 ; Mathias H. Luxner, DI Dr.techn. 2 ; Hubert Agerer, DI 1 

ABSTRACT 

In 1991 - 1992, a crownclosed large drain arch dam, having a height of 17 meters, was built 

in the middle course of "Schnannerbach". Three weeks upon completion the sediment 

control dam already proved its functionality when a debris flow event of approximately 

30000m³ of bed load was retained by the dam so that damage to the village in the lower 

course was prevented. However, in 2005, during a steady rain and flood event, which 

resulted in huge bed load transport, the sediment control dam did not operate as intended. 

It failed because the drains blocked only partially. The result was massive damage to a village. 

As a makeshift, the flow through the drain penning was limited by means of rakes. The 

consequence of this modification was that even during snowmelt flood, bed load was retained 

by the sediment control dam. This was not the original intention and led to increased cost in 

terms of excavating the sedimentation basin. Since 2005 the basin had to be excavated six 

times. 


In den Jahren 1991 und 1992 wurde eine kronengeschlossene großdolige Bogensperre im 

Mittellauf des Schnannerbaches errichtet. Drei Wochen nach deren Fertigstellung wurde 

diese Sperre durch einen Murgang beaufschlagt. Bei diesem Ereignis konnten mit Hilfe der 

Geschiebedosiersperre rund 30.000m³ Material schadlos abgelagert werden. Im August 2005 

fand ein weiteres Schadereignis statt. Im Gegensatz zu 1992 gab es keinen murartigen 

Schwall, bei dem die Großdolen verklausten, sondern über mehrere Stunden war ein 

Hochwasser mit sehr viel Geschiebetrieb zu beobachten. Prozessbedingt funktionierte die 

großdolige Sperre nicht. Es kam nur kurzzeitig zu einem Rückstau. Die Folge waren große 

Schäden am dicht besiedelten Schwemmkegel. 

Als Sofortmaßnahme wurde das vorhandene Durchflussprofil der Großdolen verkleinert. 

Infolge dessen lagerte sich in den Jahren nachher bereits während der Schmelzhochwässer 

1 Austrian Service for Torrent and Avalanche Control, Imst, AUSTRIA, arthur.vogl@die-wildbach.at 

2 Luxner Engineering ZT GmbH 



egelmäßig Geschiebe im Becken hinter der Bogensperre ab. Diese Anlandungen verursachten 

hohe jährliche Räumungskosten. Mit Hilfe eines variablen Dolenverschlusses, der sowohl 

automatisch als auch manuell bedienbar ist, soll zukünftig sichergestellt werden, dass die 

Sperre bei Murgängen sowie bei Hochwasser mit Geschiebetrieb funktioniert und damit die 

Räumungskosten massiv gesenkt werden können. 

KEYWORDS 

bed load management; rubber dam; sediment control dam 

EINLEITUNG 

Seit dem Jahr 1852 wurden am Schnannerbach in der Gemeinde Pettneu am Arlberg im Tiroler 

Stanzertal insgesamt elf Schadereignisse dokumentiert, siehe Abb.1. Mit den ersten 

technischen Verbauungsmaßnahmen begann die Wildbach- und Lawinenverbauung im Jahr 

1914. Ende der 70-iger und anfangs der 80-iger Jahre des letzten Jahrhunderts wurde der 

Schnannerbach im gesamten Schwemmkegelbereich mit einer Grundschwellenstaffelung und 

beidseitigen Leitwerken in Zementmörtelmauerwerk verbaut. Die Errichtung eines Geschieberückhaltes 

war aus Platzgründen bzw. wegen des damals nicht erschließbaren Mittellaufs 

nicht möglich. 

Abbildung 1: Geographische Übersicht mit Gefahrenzonen am Schwemmkegel. 


Am 29. Juli 1990 fand in diesem Einzugsgebiet das bis dato größte Ereignis statt. Ein Mur - 

stoß - ausgelöst durch ein Starkniederschlagsereignis mit Hagelschlag - verursachte am dicht 

besiedelten Schwemmkegel Schäden an insgesamt 12 Wohn- und 2 Wirtschaftsgebäuden 

sowie Straßen und landwirtschaftliche Flächen. Das Unterlaufgerinne war infolge massiver 

Anlandungen nicht in der Lage, das anfallende Wasser und Geschiebe abzuführen. Weitere 

Schadereignisse folgten in den Jahren 1991, 1992 und 2005, siehe Abb. 2. 

Abbildung 2: Die linke Abbildung zeigt das Ereignis vom 23. August 2005. Rechts das Ereignis vom 29. Juli 1990. 

Nach dem folgenschweren Ereignis von 1990 wurde in den Jahren 1991 und 1992 im unteren 

Mittellauf eine rd. 17 Meter hohe kronengeschlossene großdolige Bogensperre mit 

einem Geschieberückhaltebecken, das ein Fassungsvermögen von bis zu 35 000m³ aufweist, 

errichtet, siehe Abb. 3. Diese großdolige Bautype ergab sich aus der Tatsache, dass in diesem 

Einzugsgebiet bis damals nur Murereignisse stattfanden. Um den Sperrenstandort bzw. das 

Geschieberückhaltebecken oberhalb einer rd. 100 m langen und bis zu ebenso hohen, nur 

wenige Meter breiten Schluchtstrecke zu erreichen, war die Errichtung von zwei, insgesamt 

265 m langen Zufahrtstunneln notwendig. 

Bereits drei Wochen nach Fertigstellung der Bogensperre fand ein weiteres Ereignis statt; 

auch dieses Mal infolge eines konvektiven Starkniederschlages. Bei diesem Murgang konnte 

die gesamte Geschiebemenge aus dem Einzugsgebiet oberhalb der Bogensperre schadlos 

hinter dem Sperrenbauwerk zurückgehalten werden. 

Während eines advektiven Niederschlagereignisses im August 2005 gab es in Westösterreich 

große Sachschäden im Bereich von Siedlungen sowie Infrastruktureinrichtungen und landwirtschaftlichem 

Kulturland. Überwiegend waren davon der Inn und seine großen Zubringer, 

wie beispielsweise die Trisanna und die Rosanna betroffen. Aber auch in kleineren 

Einzugsgebieten verursachte dieses Ereignis Schäden. Am Schnannerbach waren insgesamt 

10 Wohnhäuser, drei Fremdenverkehrsbetriebe, zwei Gewerbebetriebe, zwei öffentliche 

Gebäude sowie landwirtschaftliches Kulturland betroffen. 


Abbildung 3: Die 17 Meter hohe kronengeschlossene großdolige Bogensperre. 

Bis zum August 2005 wurden im Schnannerbach nur Schadereignisse durch Murgänge 

infolge konvektiver Starkniederschläge dokumentiert. Während des lang anhaltenden 

Landregens am 22. und 23. August 2005 wurden sehr große Mengen von Kalk/Dolomitgrus 

sowie Mergel mobilisiert und im Gerinne abtransportiert. Das 1991 bis 1992 errichtete 

Geschieberückhaltebecken war nur in sehr geringem Ausmaß verlandet bzw. in der Lage 

das Schadgeschiebe zurückzuhalten. Nach Analyse des Ereignisses wurde ersichtlich, dass die 

Dolen der Bogensperre, welche auf einen Murgang ausgelegt waren, nur über einen kurzen 

Zeitraum des mehrtägigen Ereignisses verschlossen waren. Gründe dafür waren vor allem die 

fehlenden Grobgeschiebekomponenten sowie das fehlende Wildholz. 


Nach diesem Ereignis forderte die betroffene Bevölkerung vehement die Verkleinerung der 

Dolen, damit Geschiebe unabhängig vom Prozesstyp Murgang oder Hochwasser mit Geschiebetransport 

immer im Rückhaltebecken zurückgehalten werden kann. Dies wäre jedoch mit 

den gravierenden Nachteil verbunden, dass auch bei kleineren Ereignissen, wie beispielsweise 

den jährlichen Schmelzhochwässern nicht wenig Geschiebe im Becken abgelagert wird und 

man damit unweigerlich mit höheren Räumungskosten rechnen müsste. 

Die gewünschten Anforderungen an das Sperrenbauwerk sind daher sehr konträr: einerseits 

soll der Stauraum der Sperre so lange wie möglich leer bleiben, um Rückhaltekapazitäten so 

lang wie möglich verfügbar zu haben sowie um Kosten (Räumung) zu sparen und anderer- 


seits muss der Geschieberückhalt sowohl bei einem Murgang als auch bei einem Hochwasser 

mit Geschiebetransport funktionieren, sodass im Schwemmkegelbereich bzw. im Unterlauf 

keine Schäden entstehen. 

Um das vorhandene Geschiebeablagerungsbecken bestmöglich ausnutzen zu können, 

müssten also, je nachdem ob es sich um einen Murgang oder um ein langanhaltendes 

Hochwasserereignis mit Geschiebetransport handelt, unterschiedliche Sperrenöffnungstypen 

vorhanden sein. 

Im Rahmen einer Vorstudie wurde nach Möglichkeiten gesucht, die bestehenden Dolenöffnungen 

derart zu adaptieren, dass der Spagat zwischen Sicherheit und Minimierung der 

Räumungskosten bzw. der zu gewährleistenden Schutzfunktionalität bei den beiden 

vorerwähnten Prozesstypen geschaffen wird. 

Aufgrund der Tatsache, dass die Bogensperre schon seit 1992 besteht, ergaben sich neben den 

vorerwähnten Hauptanforderungen noch Rahmenbedingungen für die Umsetzung der 

Maßnahmen. So war zum einen der für die Adaptierungsarbeiten verfügbare Raum durch die 

Dolengröße sowie -lage bereits vorgegeben und zum anderen sollte in den Sperrenbestand - 

nicht zuletzt aus statischen Gründen - möglichst wenig eingegriffen werden. Eine weitere 

Anforderung war, dass die neue Anlage im Extremfall die wichtigsten Funktionen auch ohne 

externe Energiequelle, nur durch menschliche Kraft, ausführen kann. Zudem ist der Bereich 

um die Bogensperre inklusive Zugang/fahrt im Ereignisfall extrem steinschlaggefährdet. 

Aus diesem Grund muss die neue Anlage auch aus der Ferne bedienbar sein. 

Für die Überwachung bzw. die Analyse von Ereignissen sowie um für die Abwicklung 

künftiger Ereignisse zu lernen bzw. Erfahrungen zu sammeln soll die Anlage zusätzlich 

umfangreiche Datenaufzeichnungen ermöglichen. 

Abbildung 4: Entscheidungsmatrix der Vorstudie (grün- gut, orange - neutral, rot - schlecht). 


PROBLEMLÖSUNG 

Im Rahmen der Problemlösungsphase wurde eine umfangreiche Variantenstudie durchgeführt. 

Dabei wurden vier Möglichkeiten eines variablen Dolenverschlusses entwickelt und 

bewertet. Anhand verschiedener Kriterien stellte sich im vorliegenden Fall schließlich ein 

Schlauchwehr als variabler Dolenverschluss als am besten geeignet dar. Abb. 4 zeigt die 

Gegenüberstellung anhand der gewählten Bewertungskriterien im direkten Vergleich zwischen 

einem klassischen Schütz- und einem Schlauchwehr. 

Abbildung 5: Fotomontage mit den wichtigsten Anlagenteilen im Bereich der Bogensperre. 

Das Konzept der Anlage besteht darin, dass in den obersten drei der vier derzeit bestehenden 

und im Zuge des Umbaus geringfügig adaptierten Großdolen der Bogensperre jeweils ein 

Schlauchwehr integriert wird, welches im Ausgangszustand offen ist. Jedes derartige Wehr 

besteht im Wesentlichen aus einem flexiblen Schlauch, ähnlich dem häufig bei Wasserfassungen 

für Kleinwasserkraftwerke verwendeten, welcher durch das Befüllen/Ablassen von 

Wasser und der damit verbundenen Volumenzu- und abnahme die Querschnittsfläche der 

Dole variieren kann. Jedes Wehr fasst im gefüllten Zustand ca. 6 m³ Wasser. 

Die zum Befüllen der Schlauchmembranen erforderliche Wassermenge befindet sich in 

einem betonierten Hochbehälter der 30 m³ fassen kann und ca. 30 Höhenmeter über dem 


untersten Wehr liegt. Durch die Höhendifferenz wird der im Betrieb erforderliche Innendruck 

von 275 kPa in den Membranen sichergestellt. Weiters ist dadurch ein Befüllen aller Wehre 

nahezu ohne externe Energiezufuhr, nur durch manuelles Öffnen eines Schiebers, möglich, 

was eine Anforderung für Notsituationen darstellt. 

Der Hochbehälter und die Wehre bilden zusammen mit dem Tiefbehälter, welcher als 

Zwischenspeicher bei der Entleerung dient, der Pumpe und dem Rohrleitungssystem einen 

geschlossenen Wasserkreislauf, siehe Abb. 5. D.h. das bei der Entleerung eines Wehres 

anfallende Wasser gelangt zuerst in den Tiefbehälter und wird von dort mittels der Pumpe 

wieder in den Hochbehälter gefördert. Um mögliche Frostschäden zu verhindern, wird der 

Wasserkreislauf über den Winter entleert. 

Die Anlage wird über eine elektronische Steuerung bedient und überwacht. Diese befindet 

sich in einem Gebäude im luftseitigen Bereich der Bogensperre. Das Öffnen und Schließen 

der Dolen sowie die Interpretation der Messwerte erfolgt durch eine örtliche Kommission 

unter dem Vorsitz des Bürgermeisters. 

Im Bereich der Mündung des Schnannerbaches in die Rosanna ist ein elektronischer 

Messpegel installiert, welcher mit der Steuerung der Anlage verbunden ist. Zusätzlich 

befindet sich im gleichen Bereich eine färbige Pegelmarkierung welche die visuelle Einschätzung 

des Wasser-/Geschiebepegels in Form von Ampelfarben ermöglicht. Beides dient als 

Entscheidungshilfe für die Kommission. 

Wie bereits erwähnt, soll die bestehende Bogensperre baulich möglichst wenig verändert 

werden. Diese Vorgabe schließt eine in die Mauer integrierte Zu- und Ableitung zum Befüllen 

und Entleeren der Schlauchmembrane aus. Aus diesem Grund ist jedes Wehr in einen 

Stahlrahmen integriert, welcher als Ganzes in die Dole eingebaut wird. Die Zu- und Ableitung 

ist ebenfalls in diesen Rahmen integriert. Diese Konstruktionsweise bietet zudem den Vorteil, 

dass jedes Wehr in der Werkstätte als Ganzes vorgefertigt werden kann und somit die 

Montagezeit auf der Baustelle wesentlich reduziert sowie den Montageablauf dort deutlich 

vereinfacht und auch verbilligt wird. An der Bogensperre sind durch diese Konstruktionsweise 

somit nur Bohrungen für die Verankerung der Wehre sowie kleinere Ausbrüche 

notwendig. 

Um die hohen Anforderungen an die Anlage sowie deren ständige Verfügbarkeit zu gewährleistet, 

ist die Steuerung im Industriestandard ausgeführt. Sie besteht im Wesentlichen aus 

Messtechnik, Automatisierungsebene und Prozessebene. Die Prozessabläufe werden hier in 

der Feldebene programmiert und dann ans Leitsystem übermittelt. Ein zentrales Leitsystem 

steuert und überwacht die gesamten Anlagenkomponenten und ist die zentrale Anlaufstelle 

der Prozessdaten. Die Visualisierungen, die Archivierung, das Reporting sowie das Alarm- 

Management werden vom Prozessleitsystem verwaltet und geregelt. 


Zum Einsatz kommt, aufgrund der Priorität der Anlage, ein Alarmsystem mittels Sprachprozessor 

(Sprachalarmierung); dieses bietet ein hohes Maß an Sicherheit im Störfall. 

Mittels PC, iPad oder Laptop kann die Anlage vollwertig aus der Ferne bedient und überwacht 

werden. Die hohe Verfügbarkeit der Anlage ist auch bei Stromausfall (Notstromaggregat) 

oder einer defekten Internetverbindung (W-Lan-Access-Point) garantiert. 

Die Eigenverantwortung der Entscheidungsträger ist sehr groß, da über die Steuerung des 

mobilen Dolenverschlusses (vergl. Abb. 6) direkt in den Ablauf des Katastrophenszenarios 

eingegriffen wird. 

Abbildung 6: Organisationsorganigramm während eines Hochwasserereignisses mit Geschiebetransport. 


FAZIT 

Der mobile Dolenverschluss soll die intelligente Antwort auf die unterschiedlichen Anforderungen 

aufgrund der Prozesstypen Murstoß und Hochwasser mit Geschiebetransport 

darstellen. Die Dolen sollen so spät wie möglich geschlossen werden, um den schadlosen 

Geschiebetrieb so lang wie möglich zu gewährleisten bzw. so bald wie möglich wiederum zu 

ermöglichen und dadurch das Geschiebeablagerungsbecken so gut und so lang wie möglich 

zu nutzen. Im Idealfall soll die abklingende geschiebeentlastete Hochwasserwelle genutzt 

werden, um allfällige Anlandungen im Unterlauf des Schnannerbaches wieder auf natürlichem 

Weg zu räumen. Diesen geschiebeentlasteten Hochwasserabfluss soll auch durch das 

Schließen einer zusätzlichen Dole künstlich initiiert werden. 

Der mobile Dolenverschluss soll eine umfassende Möglichkeit darstellen, in die Prozessdynamik 

eines Wildbaches, im konkreten Falls des Schnannerbaches variabel sowie situationsbezogen 

einzugreifen. Er bietet damit die Chance, den Spagat zwischen Sicherheit und 

Ökonomie, dh. Minimierung der Räumungskosten zu schaffen. 

Im Gegensatz zu herkömmlichen Schutzmaßnahmen in der Wildbach-, Lawinen- und 

Steinschlagverbauung, deren Funktionalität während eines Ereignisses im Zusammenhang 

mit dem Prozesstyp in der Regel nicht aktiv beeinflussbar ist, ist dies bei der, in diesem 

Aufsatz vorgestellten Anlage möglich, ja sogar erforderlich. Dies setzt jedoch eine hohe 

Eigenverantwortung und auch Einsatzbereitschaft jener Personen voraus, die diese Anlage 

warten, betreuen und im Ereignisfall auch bedienen. Diese hohe Eigenverantwortung und 

Einsatzbereitschaft müssen durch entsprechende Schulungen und Katastrophenübungen 

sowie den Lehren und Erfahrungen aus abgelaufenen Ereignissen permanent unterstützt 

werden. 

Wie auch bei anderen aktiven Schutzmaßnahmen mit temporärer Wirkung (z.B. künstlicher 

Lawinenauslösung) auch, werden die variablen Dolenverschlüsse in der Gefahrenzonenabgrenzung 

nicht berücksichtigt. 



Actions for the Maintenance and Lifespan prolongation 

of SABO Facilities 

Hisashi Watanabe, Professional engineer 1 ; Toshio Mori 

ABSTRACT 

Japan is one of the countries most prone to sediment-related disasters in the world. For 

centuries, various SABO works, which are measures to counter erosion and control sediment, 

have been constructed in Japan. In the modern era, countermeasures that protect people 

from localized sediment-related disasters, such as debris flows or cliff failures, have been 

emphasized, considering how such disasters actually occur. 

Many locations in Japan are prone to sediment-related disasters. Therefore, It is very 

important to maximize the functionality and extend the lifespan of existing SABO facilities. 

Our organization, The Sabo Frontier Foundation, evaluates the soundness of SABO facilities 

and reviews measures to counter sediment-related disasters based on our research. 

In fiscal 2015, the Erosion and Sediment Control Department of the Ministry of Japan issued 

the Manuals, to maintain the functions of existing structures and ensure the future use of 

those structures. Nation-wide efforts are now being taken in this respect. This paper outlines 

the two manuals and introduces the actions taken. 

KEYWORDS 

National Resilience; Maintenance and Lifespan prolongation of SABO facilities; effective use 

of existing SABO facilities; inspection; soundness evaluation 

INTRODUCTION 

Japan is one of the countries most prone to sediment-related disasters in the world. For 

centuries, various SABO works, which are measures to counter erosion and control sediment, 

have been constructed in Japan. In the modern era, countermeasures that protect people 

from localized sediment-related disasters, such as debris flows or cliff failures, have been 

emphasized, based on how such disasters occur. 

Many locations in Japan are prone to sediment-related disasters and little progress has been 

made in the countermeasures taken using SABO facilities, which are at about 20% capacity. 

It is very important to maximize the functionality and extend the lifespan of existing SABO 

facilities. Our organization, The Sabo Frontier Foundation, evaluates the soundness of SABO 

facilities and reviews measures to counter sediment-related disasters based on our research. 

The issues of maintenance and prolongation of the lifespan of SABO facilities are important 

globally; some countries have already implemented countermeasures to the problem. In fiscal 

1 SABO FRONTIER FOUNDATION (SFF), TOKYO, JAPAN, kenkyu2@sff.or.jp 



2015, the Erosion and Sediment Control Department of the Ministry of Land, Infrastructure, 

Transport and Tourism (MLIT), Japan, issued the “Planning Manual for Maintaining and 

Prolonging the Lifespan of SABO Facilities” and the “Inspection Procedure Manual for SABO 

Facilities”, to maintain the functions of existing structures and ensure the future use of those 

structures (these manuals are currently available only in Japanese). Nationwide efforts are 

now underway in this respect. This paper outlines the two manuals and introduces the 

actions taken. 

PROBLEMS FACING JAPAN 

1) Climate of Japan and aggravation of natural phenomena 

Japan is part of the Pacific Ring of Fire and much of the terrain is steep and complex. 

Geologically, volcanic and sedimentary rocks are distributed in a diverse mosaic with many 

faults. Since the beginning of the Holocene (from approximately 10,000 years ago until the 

present), Japan has experienced the fourth greatest number of eruptions globally, accounting 

for 7.3% of all volcanic eruptions. In addition, about 20.5% of all earthquakes with a 

magnitude of 6.0 or higher have occurred in Japan. 

Analysis of the source waters and routes of typhoons, including hurricanes, in the Northern 

Hemisphere shows that their routes are concentrated in three tracks, one of which leads from 

Southeast Asia towards the eastern part of the Sea of Okhotsk via the waters near Japan. 

In Japan, mountains comprise 61.0% of the total land area, while less than 40% of Japan is 

habitable. As a result, the population density in the habitable area exceeds 900 people per 

km 2 . Considering the topography, geology, climate, and lack of inhabitable land, it is 

understandable that Japan is prone to sediment-related disasters. 

Recently, typhoons, earthquakes, and volcanic eruptions have increased in scale. In addition, 

the frequency of localized torrential rain from continuously generated clouds, the so-called 

“back-building phenomenon”, has increased. This has led to large-scale sediment-related 

disasters, such as those observed in the Niigata Prefecture Chuetsu Earthquake (2004; Fig. 1) 

Figure 1: Overview of the damage caused The Niigata Prefecture Chuetsu Earthquake 


and extensive sediment-related disasters in the Kii Peninsula (2011, Fig. 2) and Hiroshima 

(2014; Fig. 3). Consequently, appropriate countermeasures are required to ensure regional 

safety. 

Figure 2: Overview of sediment-related disaster in Kii Peninsula, and Period rainfall distribution (August 30 ~ September 5) 

2) Decrease in investment capacity 

In Japan, SABO works have been built for a long time. Since 1875, some 90,000 check dams 

and 10,000 km of channel works have been constructed. Nevertheless, more than 1,000 

sediment-related disasters occur annually. Therefore, more structural and non-structural 

measures need to be taken. Structural measures to deal with natural phenomena include 

maintaining the “function and performance" of SABO facilities, reinforcing and improving 

them, and developing new facilities. (here, the “performance of a SABO facility” is defined as 

the structural safety of the SABO facility, while the “function of a SABO facility” is defined 

as its ability to prevent sediment-related disasters.) Of those existing, increasing numbers are 

over 50 years old and it is essential to counteract damage or deterioration of these facilities. 


Figure 3: Overview of sediment-related disaster in Hiroshima, and 60 minutes accumulated rainfall (13:00 August 19–13:00 August 20) 

There is also an urgent need to update the technical standards of any SABO facility that is 

“incompatible with the latest technical standards”. 

However, the increasing costs of social security in Japan make it difficult to increase capital 

maintenance costs. Therefore, to reduce the total cost of SABO facilities, i.e., the cost of 

maintaining the installations in an era of limited budgets, SABO facilities must be developed 

and maintained efficiently and effectively. Their maintenance and lifespan are very important, 

together with evaluations of their function and performance (see chapter 2 ”Evaluating 

the soundness of a SABO facility” below) and making appropriate repairs and reinforcement 

when needed. 

INSPECTION METHODS 

1) The contents of manuals 

The Erosion and Sediment Control Department of the MLIT issued the “Planning Manual for 

Maintaining and Prolonging the Lifespan of SABO Facilities”, which outlines how to maintain 

and preserve the functionality of existing SABO facilities. In particular, the sections on the 

“Inspection of a SABO facility” and “Evaluation of the soundness of a SABO facility” are 


important considerations in the “Inspection Procedure Manual for SABO Facilities”, which 

outlines inspection methods. 

2) Evaluating the soundness of a SABO facility 

First, the soundness and fitness (here, “fitness” is defined as the soundness against aging by 

weathering) of the SABO facility of interest is evaluated visually. This inspection determines 

whether any damage or deterioration has occurred in the SABO facility. The effect that the 

observed damage or deterioration has had on the function of the SABO facilitiy is categorized 

into one of three grades (e.g., Fig. 4); SABO facilities in the “Anomaly level a” category are 

excluded from further measures at this stage. 

Figure 4: Example of an evaluation of the level of anomaly by part (source: Inspection Procedure manual for SABO Facilities) 

In addition to quantitative information on anomalies detected in the evaluation, factors that 

may cause damage or deterioration to the facility (such as the structure, material, model or 

characteristics of the catchment area) are observed, evaluated, or analyzed (Fig. 5). The 

overall soundness of the facility is evaluated comprehensively based on all this information 

(Fig. 6). 

For example, if the same kind of abrasion of the levee crown has occurred at multiple 

facilities, an evaluation is made based on the types of internal material (e.g., concrete or 

rubble concrete). If the same kind of cracking has occurred at multiple facilities, an evaluation 

is made based on the direction of cracking (e.g., vertical, horizontal, diagonal, etc.). The 

immediate surroundings of the SABO facility, for example the location and volume of 

deposited sediment, the stability of the facility site, and the stability of the substrate near the 

wing of the facility are also evaluated. 


Figure 5: Example of the evaluation corresponding to the facility structures, materials, and types (Abrasion or Corruption of levee crown, 

in Concrete dam and Masonry dam) 

Figure 6: Comprehensive evaluation of SABO facilities 


"The Performance of SABO Facilities" will be used to evaluate the suitability of technical 

standards (safety standards) and as a reference when deciding a measure’s priority, such as 

the placement of conservation targets and SABO facilities, the importance of SABO facilities, 

the recommended size of facilities, the cost of the countermeasures, and the ease of construction 

of countermeasures. This will help to prioritize the measures to be taken. 

3) In-need-of-repair SABO facilities and follow-up observations 

SABO facilities are being classified as those that require urgent repairs (hereinafter “in-needof-repair” 

SABO facilities) and those that need follow-up observation (hereinafter “follow-up 

observation” SABO facilities), based on the soundness evaluation. This classification is used to 

identify factors that lead to a loss of function and performance of SABO facilities. The 

personnel undertaking this classification work must have relevant experience or qualifications 

related to SABO. 

For in-need-of-repair SABO facilities, effective repair measures need to be developed based 

on the characteristics of any damage or deterioration (structure, model, material, catchment 

area characteristics, etc.). Follow-up observations of SABO facilities will include the implementation 

of an inspection that can monitor the progress of damage or deterioration (e.g., 

abrasion of the levee crown, increase or decrease in crack width) and recording of the 

inspection results. In this way, SABO facilities will be classified into those requiring observation 

and those requiring immediate measures, so that the limited budget can be allocated 

efficiently to protect the safety of the region. 

Figure 7: Result of analyzed and evaluation about the soundness of the SABO facilities in five under the direct control SABO works office 


In 2014, we evaluated the soundness of SABO facilities under the direct control of five SABO 

works offices. These investigations showed “abrasion of the levee crown” and “scouring of the 

foundation of the dam”, typical damage or deterioration that leads to loss of function and 

performance of the facility. On average, approximately 12% of SABO facilities were “in-needof-repair” 

and 58% required “follow-up observations” (Fig 7). 

4) A case example 

SABO facilities are being repaired to maintain their functions. In addition, their structures are 

being reinforced and function improved to achieve effective utilization of facilities from the 

viewpoint of maintenance and prolonging the lifespan of SABO facilities (a reduction of total 

cost) (Fig. 8). 

For example, when a SABO facility suffers from repeated abrasion of the levee crown, it is 

repaired with granolithic concrete to reduce the repair frequency (L). When an apron and 

counter dam are constructed to prevent scouring at the foundation (S), appropriate measures 

are taken using a variety of actions, including widening the dam to improve stability (s1), 

increasing the crown and wing widths to confer greater protection against debris flow (s2), 

and raising the dam height to increase the sediment trap capacity of the SABO facility (s3). 

Figure 8: Example of measures in light of Maintenance and Lifespan-keeping of SABO facilities (reduction of total cost) 


BUDGET AND SYSTEM 

The MLIT established a system that provides for a SABO facility manager who implements 

a “Maintenance and Lifespan-prolongation Plan” according to the manuals, with a national 

subsidy to cover part of the work costs in 2015. To take advantage of this system, the 

prefecture began to create a “Maintenance and Lifespan-prolongation Plan”, with a deadline 

of fiscal 2018. In addition, facilities under the direct control of SABO works undergoing 

construction and maintenance must be investigated by the end of fiscal 2016, and measures 

should then be implemented. To contribute to these plans, our organization will study the 

soundness of SABO facilities and countermeasures for various types of SABO facility. 

REFERENCES 

- Planning Manual for Maintaining and Prolonging the Lifespan of SABO Facilities: 

The Erosion and Sediment Control Department of the Ministry of Land, Infrastructure, 

Transport, and Tourism of Japan (Currently available only in Japanese.) 

- Inspection Procedure Manual for SABO Facilities: The Erosion and Sediment Control 

Department of the Ministry of Land, Infrastructure, Transport and Tourism of Japan 

(Currently available only in Japanese.) 

- Analysis of the as-is-state of old natural stone barriers in the Bernese Alps: Bernd Kister, 

Lucerne University of Applied Science and Arts, Horw, Switzerland 


Emergency Management 

(emergency planning, early warning, intervention, recovery) 


EMERGENCY MANAGEMENT (EMERGENCY PLANNING, EARLY WARNING, INTERVENTION, RECOVERY) 

Outflow forecast on a mountain river! 

(Emergency Management in the Canton of Nidwalden) 

Werner Fessler, Dipl. Kult. Ing. ETH 1 ; Markus Klauser, Dr. sc. ETH 1 , Dipl.Ing.; Peter Seitz, Dipl. Ing. 1 ; 

Josef Eberli, Dipl. Kult. Ing. ETH 1 

ABSTRACT 

Emergency management in the Canton of Nidwalden ensures timely and targeted intervention 

in case of a flood event on the Canton's main river, the „EngelbergerAa". In case of an 

event, the emergency planning includes the definition of processes and responsibilities as 

well as monitoring. It provides necessary resources and basis for decision making. In addition, 

a reliable forecast of the expected development is fundamental for an effective intervention. 

However, customary rainfall-outflow models do not represent the EngelbergerAa, a river 

characterized by steeply rising hydrographs with pronounced peaks. The Canton Nidwalden 

developed for that reason a customized rainfall-outflow model, based on a simple three-storage-reservoir-system 

as a function of time. The developed model for outflow-prediction in the 

EngelbergerAa has proven its steadiness during various incidents and generated valuable 

services. Despite the simple architecture of the applied algorithms, the tool allows to trigger 

and control the necessary measures at the right time. Though, the experience and assessment 

of management staff as well as task forces have to complete the resulting forecasts. 

Durch die Notfallorganisation wird im Kanton Nidwalden sichergestellt, dass bei Hochwasser 

an der Engelberger Aa gezielt und rechtzeitig interveniert wird. In der Notfallplanung 

werden, in Berücksichtigung relevanter Kennwerte, für den Ereignisfall Abläufe, Überwachungen 

und Zuständigkeiten festgelegt, sowie die erforderlichen Hilfsmittel und Entscheidungsgrundlagen 

bereitgestellt. Entscheidend für eine wirkungsvolle Intervention ist aber 

auch eine zuverlässige Vorhersage der zu erwartenden Entwicklung. 

Gebräuchliche Niederschlag-Abfluss-Modelle werden dem Gebirgsbach-Charakter der 

Engelberger Aa mit einer steil ansteigenden Abflussganglinie und einer ausgeprägten Spitze 

jedoch kaum gerecht. Für die Vorhersage des erwarteten Abflusses wird deshalb ein - beim 

Tiefbauamt des Kantons Nidwalden entwickeltes - Niederschlags- Abflussmodel eingesetzt. 

Das "N-A-Modell Engelberger Aa" simuliert grundsätzlich ein integrales Drei-Speicher-System 

in Funktion der Zeit. Das für die Vorhersage an der Engelberger Aa entwickelte Niederschlag-Abfluss-Modell 

hat sich während verschiedenen Ereignissen bewährt und wertvolle 

Dienste geleistet. Trotz der einfachen Architektur der Simulationsalgorithmen erlaubt die 

Abflussvorhersage - ergänzt durch Erfahrungen und Beurteilungen von Führung und 

1 Tiefbauamt Kanton Nidwalden, Stans, SWITZERLAND, werner.fessler@nw.ch 



Einsatzkräften - die notwendigen Massnahmen der Notfallplanung rechtzeitig auszulösen 

und zu steuern. 

KEYWORDS 

Emergency management, Krisen-Management; Emergency planning, Notfallplanung; 

outflow forecast, Abfluss-Vorhersage; integral rainfall-outflow model with three storage 

reservoirs, integrales Niederschlag-Abfluss-Modell mit drei Speichern; Excel routine 

mountain stream, Excel-Routine Gebirgsbach 

EINFÜHRUNG 

Hochwasserschutz ist dann wirksam, wenn vorausschauende, organisatorische Massnahmen 

der Notfallplanungen optimal mit raumplanerischen und baulichen Massnahmen korrespondieren. 

Dabei wird das Risiko durch die kombinierte Umsetzung der Massnahmenelemente 

deutlich reduziert und Schäden können verhindert bzw. vermindert werden. Entscheidend 

für eine rechtzeitige und wirkungsvolle Intervention ist aber auch eine frühzeitige und 

zuverlässige Alarmierung. 

Im Kanton Nidwalden obliegt die Verantwortung für die Engelberger Aa dem Kanton. In den 

Jahren 1998 bis 2007 wurden von Dallenwil bis Buochs - als Schlüsselelemente des integralen 

Risikomanagements - verschiedene bauliche Massnahmen am Gewässer realisiert. Die 

Anforderungen an ein robustes Systemverhalten und vorliegende Platzverhältnisse begründen 

eine sukzessive Entlastung. Hierbei werden Wassermassen, welche die Gerinnekapazität 

übersteigen, an festgelegten Entlastungsstellen ausgeleitet. Dadurch kann eine Überlastung 

des Gewässer-Systems ausgeschlossen werden und dessen Funktionalität bleibt jederzeit 

gewährleistet. Die resultierenden Abflüsse ausserhalb des Gerinnes werden in Abflusskorridoren 

mit möglichst geringem Schadenpotential flächig zum See geleitet. 

Mit den baulichen Massnahmen wurde gleichzeitig eine Notfallplanung erarbeitet und 

umgesetzt. Durch die Notfallplanung wird sichergestellt, dass bei Hochwasser gezielt und 

rechtzeitig interveniert wird. In Berücksichtigung relevanter Kennwerte werden für den 

Ereignisfall Abläufe, Überwachungen und Zuständigkeiten festgelegt sowie die erforderlichen 

Hilfsmittel und Entscheidungsgrundlagen für eine zielgerichtete, speditive und zeitnahe 

Reaktion bereitgestellt. 

Das resultierende Krisenmanagement stellt - in Zusammenarbeit mit den örtlichen Interventionskräften 

- eine Verbundaufgabe der relevanten kantonalen und kommunalen Kompetenzen 

dar. Ein Stab aus Fachleuten beurteilt kritische Situationen laufend, ergreift Massnahmen 

und gewährleistet eine rechtzeitige Alarmierung der Einsatzkräfte und der Bevölkerung. 

Die Alarmierung der Führungskräfte bezüglich der Engelberger Aa erfolgt basierend auf den 

automatisch übermittelten Abflussmessungen der Messstelle Buochs. Ergänzend werden bei 

kritischen Wetterentwicklungen entsprechende Warnungen abgesetzt. Die Abläufe orientie- 


en sich an den „Pegel-Marken" 80m 3 /s (=> Beobachtung), 100m 3 /s (=> Vorbereitung) und 

125m 3 /s (=> Intervention). 

Damit fundierte Entscheide gefällt und zweckmässige Massnahmen ergriffen werden können, 

ist auch eine zuverlässige Vorhersage erforderlich. Die „Zentrale Wasserbau" - eine dem 

kantonalen Führungsstab unterstellte Fachinstanz - ist für die hydraulische Vorhersage der 

erwarteten Entwicklung zuständig. Mit dem Ziel, den Entscheidungsträgern einen Vorsprung 

zu verschaffen, wird die Vorhersage der Abflussentwicklung fortlaufend aufbereitet. Hierfür 

wurden einfache Algorithmen entwickelt und als Berechnungsroutine in Excel bereitgestellt. 

In Berücksichtigung der Abflusssituation und in Anwendung von Niederschlagsmessungen / 

Niederschlagsprognosen werden mit Hilfe der Berechnungsroutine insbesondere nachfolgende 

Fragestellungen fokussiert: 

– Wann werden voraussichtlich die Kennwerte der Notfallplanung erreicht? 

– Wie wird sich die Abflussdynamik der Engelberger Aa entwickeln? 

– Welche mutmasslichen Abflussspitzen sind zu erwarten? 

– Wann muss voraussichtlich mit der Abflussspitze gerechnet werden? 

METHODEN 

Gebräuchliche, grossräumige Niederschlag-Abfluss-Modelle werden dem Gebirgsbach-Charakter 

der Engelberger Aa mit einer steil ansteigenden Abflussganglinie und einer kurzen, 

ausgeprägten Spitze kaum gerecht. Für die Vorhersage des erwarteten Abflusses wird deshalb 

ein beim Tiefbauamt des Kantons Nidwalden entwickeltes Niederschlags-Abflussmodell 

eingesetzt. Als Input berücksichtigt das "N-A-Modell Engelberger Aa" erhältliche Messungen 

von Abfluss und Niederschlag ergänzt mit verschiedenen Niederschlags-Prognosen. 

Die verfügbaren Daten werden fortlaufend, manuell in eine „Excel-Routine" eingepflegt. 

Das "N-A-Modell Engelberger Aa" simuliert grundsätzlich ein integrales Drei-Speicher-System 

in Funktion der Zeit (vgl. Abbildung 1). Jeder der drei Speicher nimmt das zugehörige 

Niederschlagsvolumen auf (Speicherfüllung), speichert es (Retention) und gibt die Wassermassen 

sukzessive an den Abfluss weiter (Speicherentleerung). In Anlehnung an das Reaktionsverhalten 

werden die virtuellen Speicher mit Fern (=> langsam); Mittel und Nah (=> schnell) 

bezeichnet. 

Die einzelnen Speicher bzw. deren Reaktion werden durch „Speicherparameter" (vgl. Tabelle 

1) charakterisiert, welche das zugehörige Einzugsgebiet repräsentieren. Aufgrund der gegenseitigen 

Interaktion ist eine Anpassung der "Speicherparameter" während einem Ereignis 

kaum angezeigt. Die „Speicherparameter" wurden anhand bekannter Messreihen für die 

Engelberger Aa kalibriert. Da Extremereignisse für die Kalibrierung nur beschränkt zur 

Verfügung stehen, stellt das Ereignis 2005 bis dato das massgebende Extrem-Hochwasser- 

Referenz- Ereignis dar. 


Abbildung 1: Schematische Funktionsweise N-A-Modell Engelberger Aa (Drei-Speicher-Modell) 

Tabelle 1: Modellparameter N-A-Modell Engelberger Aa (Drei-Speicher-Modell) 

Die "Grundparameter" (vgl. Tabelle 1) dienen der Steuerung der berechneten Abflussganglinie. 

Durch Änderung der kalibrierten "Grundparameter" besteht die Möglichkeiten die 

Abflussvorhersage während einem Ereignis visuell zu justieren, gleichzeitig ist diese 

Manipulation der Grundparameter jedoch kritisch zu hinterfragen bzw. zu begründen: 

– QB: vertikale Verschiebung der Abflussganglinie (+ grösserer bzw. - kleinerer Abfluss) 

– ∆t': horizontale Verschiebung der Abflussganglinie (+ späterer bzw. - früherer Abfluss) 

– nA: Geografische Niederschlagsverteilung im Einzugsgebiet (ΣN < oder > ΣMessung/ 

Prognose) 

– FEZG: Abflusswirksames Einzugsgebiet (z.B. < FEZG - Schneefall; > FEZG + Schneeschmelze) 

Das "N-A-Modell Engelberger Aa" berücksichtigt das gespeicherte Volumen als Grundlage 

für die abflussrelevanten Anteile der einzelnen Speicher. Die Berechnungsalgorithmen der 

„Excel-Routine" ermitteln pro Speicher zeitbezogen die integrale bzw. summarische Bilanz 

der anteilsmässigen Niederschlags- und Abflussvolumen (vgl. Tabelle 2). 

Der Vorgang der Speicherfüllung ist geprägt durch den im Einzugsgebiet gemessenen bzw. 

vorhergesagten Niederschlag. Die Berechnungsalgorithmen der „Excel-Routine" berücksichtigen 

eine anteilsmässige Verteilung des Niederschlags auf die einzelnen Speicher (vgl. Tabelle 

3). Die räumlichen Unterschiede werden durch die gewichtete Mittelwertbildung verschiede- 


Tabelle 2: Speichervolumen zum Zeitpunkt t als Bilanz im zugehörigen Berechnungszeitraum 

Tabelle 3: Berechnung der Speicherfüllung (aufgefangenes Niederschlagsvolumen) pro Speicher 


ner Regenmessstationen abgebildet. Ergänzend oder alternativ kann der abflusswirksame 

Niederschlag erhöht bzw. reduziert werden. Diesbezüglich ist die gutachterliche Interpretation 

von verfügbaren Summendarstellungen (z.B. Niederschlagsradar kumuliert über 6h) im 

relevanten Einzugsgebiet erforderlich. 

Die aktuelle Speicherentleerung wird durch die im virtuellen Speicher zwischengelagerten 

Wassermassen gesteuert. Die Berechnungsalgorithmen der „Excel-Routine" berücksichtigen 

hierfür die relativen Füllstände (vgl. Tabelle 4). Bei der Engelberger Aa wird die resultierende 

Entleerungsrate mit einer linearen Abhängigkeit (f=1) von der relativen Speicherfüllung 

abgeleitet. Alternativ kann mit einer Funktionalitätspotenz (f 1) die Entleerungscharakteristik 

der Speicher beeinflusst werden (f>1 => grössere Retention => geringere Abflüsse; f< 1 

=> geringere Retention => grössere Abflüsse). 

Die Entleerung der virtuellen Speicher hängt vom aktuell gespeicherten Volumen ab, weshalb 

im Grundsatz eine iterative Berechnung erforderlich ist. Durch kleine Berechnungsintervalle 

lassen sich das abgeleitete Abflussvolumen bzw. das aktuell gespeicherte Volumen durch 

Tabelle 4: Berechnung der aktuellen Entleerung der Speicher zum Zeitpunkt t 


Anwendung des vorgängigen Berechnungsschritts (t=t-∆t) annähern, womit auf die Iteration 

verzichtet werden kann. 

Mit den Rechenalgorithmen gemäss Tabellen 2 bis 4 lassen sich im zeitlichen Verlauf für jeden 

Berechnungspunkt sowie für die einzelnen Speicher die jeweiligen Speichervolumen sowie 

die resultierenden Entleerungen ermitteln. Die Aufsummierung der Entleerungen ergibt 

schlussendlich den voraussichtlichen Abfluss in der Engelberger Aa (vgl. Tabelle 5) und in 

der Zeitachse die mutmasslich zu erwartende Abflussganglinie. 

ERGEBNISSE 

Das für die Vorhersage des zu erwartenden Abflusses an der Engelberger Aa entwickelte 

Niederschlag-Abfluss-Modell wurde massgeblich am Ereignis 2005 entwickelt und kalibriert. 

Anwendungen in den Folgejahren führten vor allem zur Optimierung der Programmierung 

der Berechnungsroutinen. Gleichzeitig wurden auch die Eingabemöglichkeiten betreffend der 

verfügbaren Datengrundlagen (Messungen und Prognosen) angepasst. An den Kerninhalten 

des Modells waren jedoch aufgrund der Erfahrungen keine massgeblichen Anpassungen 

angezeigt. Ergänzend wird momentan die Anwendung für weitere Gewässer-Systeme abgeklärt, 

wobei erste Resultate vielversprechend sind. 

Tabelle 5: Berechnung des Abflusses in der Engelberger Aa als Summe der Speicherentleerungen 

Die Kalibrierungscharakteristik des Hochwasserereignisses vom August 2005 ist in Abbildung 

2 dargestellt (Berechnungsintervall 30 Min). Das Abflussdiagramm zeigt die gemessenen und 

die mit dem "N-A-Modell Engelberger Aa" berechneten Abflüsse, sowie die zugehörige 

Auf teilung auf die einzelnen Speicherentleerungen. Als Berechnungsgrundlage werden die 

regional gemittelten Niederschlagsmessungen berücksichtigt. Visuell kann festgestellt werden, 

dass die Abflussganglinien harmonieren, auch wenn einzelne Ausschläge nicht exakt 

abgebildet werden. Sowohl die Grössenordnung als auch der Zeitpunkt der Abflussspitze 

stimmen recht gut überein. Hingegen setzt der gemessene Anstieg in der zweiten Phase 


Abbildung 2: Abfluss und Kalibrierung an der Engelberger Aa für das Ereignis 2005 

(22.08) etwas früher ein als der Berechnete, was zweiter jedoch durch eine steilere Reaktion 

kompensiert. 

Die Abweichungen in der Differenzbetrachtung erscheinen auf den ersten Blick etwas gross. 

Allerdings ist zu berücksichtigen, dass die räumliche/zeitliche Variation des Niederschlags 

nur rudimentär berücksichtigt wird und die Auswertung keine zeitliche Variation beinhaltet. 

Entsprechend muss die Differenz in der Aussagekraft relativiert werden. Anderseits zeigt der 

Trend der Differenzen, dass die Kalibrierung den Abfluss in ihrer Gesamtheit recht gut 

wiederspiegelt. Auch die Korrelation der Berechnung mit der Messung berücksichtigt keine 

zeitlichen Varianzen. Dennoch kann festgehalten werden, dass der Abfluss tendenziell etwas 

unterschätzt wird, während die Abflussspitze eher überschätzt wird. 

Die Grafik zeigt auch die Problematik möglicher Modellanpassungen während einem 

Ereignis. Wird beispielsweise aufgrund der festgestellten Abweichung am 22.8 vormittags 

(ca. 30m 3 /s) durch Anpassung der "Grundparameter" (vgl. Tabelle 1) der berechnete Abfluss 

erhöht hat dies zur Folge, dass die Abflussspitze am Nachmittag und Abend noch stärker 

überschätzt wird. 

Die Abbildung 3 zeigt ein Beispiel einer Prognose für das Hochwasserereignis Anfang Juni 

2013 mit dem aktuell angewendeten Layout (Ausschnitt) des "N-A-Modells Engelberger Aa". 

Zur Illustration wurde der effektiv gemessene Abfluss ergänzt. Das Beispiel zeigt, dass der 

Anstieg der Abflussganglinie in Anwendung der Cosmo2-Prognosen sehr gut abgebildet wird. 

Mit der Cosmo7 Prognose wird der Abfluss etwas überschätzt. Allerdings ist der Zeitpunkt der 

vorhergesagten Abflussspitzen nahezu identisch, was darauf hindeutet, dass die Niederschlagsvorhersagen 

gleichartig verlaufen und entsprechend zuverlässig erscheinen. Aufgrund 


Abbildung 3: Abflussprognose an der Engelberger Aa vom 31. Mai 2013 um 18:00Uhr 

der Cosmo7-Prognose ist damit zu rechnen, dass in der Nacht vom 2. auf den 3. Juni eine 

zweite Abflusswelle zu erwarten sein wird, welche jedoch aufgrund der effektiven Niederschläge 

nicht eingetreten ist. Auch diese Grafik lässt tendenziell eine Überschätzung des 

Abflusses vermuten, allerdings ist zu berücksichtigen, dass die effektiven Niederschläge etwas 

geringer ausfielen als prognostiziert. 

FAZIT 

Das für die Vorhersage an der Engelberger Aa entwickelte Niederschlag-Abfluss-Modell hat 

sich während verschiedenen Ereignissen bewährt und wertvolle Dienste geleistet. Trotz der 

einfachen Architektur der Simulationsalgorithmen erlaubt die Abflussvorhersage - ergänzt 

durch Erfahrungen und Lagebeurteilungen von Führung und Einsatzkräften - die notwendigen 

Massnahmen der Notfallplanung rechtzeitig auszulösen und zu steuern. Bei der Interpretation 

sind allerdings auch die Systemgrenzen zu berücksichtigen. Die verfügbaren Niederschlags-Prognosen 

unterliegen im Gebirge bezüglich zeitlicher und räumlicher Verteilung 

grossen Unsicherheiten. Insbesondere lokale Gewitter sind nach wie vor nur kurzfristig vor - 

hersehbar. Ein grosser Vorteil der manuellen Datenpflege ist der Umstand, dass man sich 

zeitnah und fortlaufend mit der aktuellen Wettersituation und der Entwicklung auseinander 

setzen muss, was die Interpretation der Ereignisse massgeblich begünstigt. 


Im Kanton Nidwalden gewährleistet ein engagiertes Team rund um den kantonalen Führungsstab 

in Zusammenarbeit mit den örtlichen Interventionskräften ein risikobasiertes 

Krisenmanagement. Durch Erfahrungen aus vergangenen Ereignissen konnte das Krisenmanagement 

kontinuierlich verfeinert werden. Die Vorhersagen an der Engelberger Aa 

als kleines Puzzleteil werden vorliegend vorgestellt. Im Dienste der Bevölkerung soll eine 

zeitgemässe Risikokultur gelebt, vorgelebt und das diesbezügliche Bewusstsein sukzessive 

erweitert werden. Denn die absolute Sicherheit im Umgang mit Naturgefahren wird es 

nie geben. Die stetige Auseinandersetzung mit dem Risiko (Restrisiko) und den resultierenden 

Wirkungen, aber auch die zeitnahe Interpretation der mutmasslichen Entwicklung 

eines Ereignisses, sind entscheidend für eine wirkungsvolle Schadensminderung und 

bilden entsprechend eine Kernkompetenz des Krisenmanagements, nicht nur im Kanton 

Nid walden. 

LITERATUR 

- Tiefbauamt Kanton Nidwalden (2009), Integrales Risikomanagement am Beispiel 

Engelberger Aa 

- Tiefbauamt Kanton Nidwalden (2007), Notfallplanung Engelberger Aa, Revisionsstand 2014 

- Tiefbauamt Kanton Nidwalden (2015), NG_Berechnung_ProTool_E_Aa_NOTFALLVERSION, 

V12 



Intervention planning as a preventive tool for integral 

natural hazard management in South Tyrol/Italy 

Willigis Gallmetzer, Mag. 1 ; Martin Eschgfäller 2 ; Roland Fasolo 1 ; Peter Egger 1 

ABSTRACT 

Intervention planning plays an important and central role in an integrated natural hazard and 

risk management. Whereas previously active measures where realised to protect settlements 

and infrastructures from natural disasters nowadays urban planning instruments for natural 

risk prevention and intervention planning as an instrument for preparedness became more 

and more important and are increasingly applied. 

The Department for Hydraulic Engineering and the Department for Civil protection of the 

Autonomous Province of Bolzano-South Tyrol/Italy elaborated three instruments to improve 

the emergency preparedness. 

The first one a water hazard intervention handbook as a "roadmap" with responsibilities, 

activities and procedures in case of floods or debris flow. 

The second one is the so called intervention map for fire brigades to better manage the first 

emergency phase in the immediate aftermath of debris flow or avalanche events and the third 

on is the so called intervention plan. 

The following article describes the three preventive tools, their content and the first experiences 

of the task forces and involved person applying these two instruments. 

EINSATZPLANUNG ALS VORSORGEINSTRUMENT IM INTEGRALEN NATURGEFAHRENMANAGEMENT IN SÜDTIROL/ 

ITALIEN 

Die Interventionsplanung spielt im integralen Naturgefahren und –risikomanagement eine 

wichtige und bedeutende Rolle. Während früher aktive Verbauungsmaßnahmen zum Schutz 

vor Naturgefahren errichtet wurden, werden heutzutage raumplanerische Elemente zur 

Prävention naturbedingter Risiken und die Einsatzplanung als Instrument zur Vorbereitung 

immer wichtiger und finden vermehrt Anwendung. 

Die Abteilung Wasserschutzbauten und die Abteilung Zivilschutz der Autonomen Provinz 

Bozen-Südtirol/Italien haben drei Hilfsinstrumente zur Verbesserung der Notfallplanung 

erarbeitet: das Einsatzhandbuch Wildbach als Anleitung im Ereignisfall bei Hochwasser und 

Muren mit Beschreibung von Verantwortlichkeiten, Tätigkeiten und Abläufen und als zweites 

Instrument die so genannte Einsatzkarte für die Feuerwehren zur besseren Bewältigung der 

Erstphase bei Einsätzen unmittelbar nach Murereignissen oder Lawinen und als drittes den so 

genannten Einsatzplan. Der schriftliche Beitrag beschreibt die drei Vorsorgeinstrumente, 

1 Autonomous Province of Bolzano South Tyrol, Civil Protection Agency, Bolzano South Tyrol ITALY, willigis.gallmetzer@provinz.bz.it 

2 ambio-alp forstingenieure, Expert in risk prevention planning 



deren Inhalt und die ersten Erfahrungen der Einsatzkräfte und Beteiligten bei deren 

Anwendung. 

KEYWORDS 

intervention planning; intervention map; hazard intervention handbook; natural hazard 

and risk management. 

1. EINFÜHRUNG 

Der Alpenraum ist immer wieder von Naturkatastrophen wie Hochwasserereignissen, 

Muren oder Lawinen betroffen. So haben in den Jahren 2008 bis 2014 allein in Südtirol 

666 Unwetterereignisse große Schäden angerichtet. Am häufigsten waren Verkehrswege 

und Landwirtschaftsflächen betroffen, aber auch an den Schutzbauten selbst waren in rund 

1/3 der Ereignisfälle größere Schäden zu verzeichnen (Autonome Provinz Bozen-Südtirol, 

Abteilung Wasserschutzbauten, 2015a). Zwar sind die Schäden an Gebäuden und Infrastrukturen 

von der Anzahl her geringer, dafür sind aber die monetären Schäden höher. Leider 

waren bei den Unwetterereignissen auch Todesopfer zu beklagen. Neben und auch in diesen 

Gefahrenzonen entwickeln sich Siedlungen, Landwirtschaft, Industrie, Tourismus, Handel 

und Infrastrukturen immer weiter. Mit der Veränderung des wirtschaftlichen Umfeldes verändert 

sich auch das Schadenspotential; dieses steigt stetig an. Dem gegenüber wächst gleichermaßen 

auch das Schutzbedürfnis der Bevölkerung. In der Vergangenheit waren die Schutzstrategien 

vor allem auf aktive Verbauungsmaßnahmen ausgerichtet, mit dem Ziel die 

Naturgefahrenprozesse direkt zu beeinflussen. So existieren derzeit in Südtirol rund 28.000 

Querbauwerke der Wildbachverbauung, das sind 3,8 Querbauwerke pro km² und auf die 

Bevölkerung Südtirols umgerechnet 1 Querbauwerk pro 18 Einwohner (Autonome Provinz 

Bozen-Südtirol, Abteilung Wasserschutzbauten, 2015b). Heute begegnen die zuständigen 

Behörden den Naturgefahren auf mehreren Ebenen. Neben den nach wie vor erforderlichen 

Verbauungsmaßnahmen zum Schutz von Siedlungen und Infrastrukturen spielen territoriale 

Planungsinstrumente wie Raumplanung unter Berücksichtigung der Gefahrenzonen, die 

Sensibilisierung der betroffenen Anrainer, Selbstschutzmaßnahmen und die Schulung der 

beteiligten Einsatzkräfte als vorbeugende Notfallmaßnahmen eine zentrale Rolle im integralen 

Naturgefahrenmanagement. Die Auswahl der Schutzstrategien für Risikozonen wird von 

den beteiligten Entscheidungsträgern in Abhängigkeit von einer Gefahrenstudie, Risikoanalyse, 

Kosten-Nutzen-Rechnung und zeitlichen und finanziellen Ressourcen getroffen. 

Drei Hilfsinstrumente, welche die präventive Notfallplanung der Einsatzkräfte in Südtirol 

unterstützen sind die Einsatzkarten und -pläne für Wasser- und Lawinengefahren für die 

örtlichen Feuerwehren und das Einsatzhandbuch Wildbach mit dem Sonderplan Hochwasserdienst 

für die Flüsse Etsch und Eisack. 


2. INSTRUMENTE DER INTERVENTIONSPLANUNG BEI NATUREREIGNISSEN 

In Südtirol werden zusätzlich zu den Zivilschutzplänen drei Instrumente zur Einsatzplanung 

im Bereich der Naturgefahren unterschieden: die Einsatzkarten, die Einsatzpläne und die 

Einsatzhandbücher. Für die Erstellung von Einsatzkarten haben die Landesabteilungen Zivil - 

schutz und Wasserschutzbauten mit einem Zivilschutzsachverständigen ein Handbuch erstellt. 

Die Einsatzkarten dienen den operativen Kräften als Einsatzstütze bei seltenen Er eignissen 

mit mäßig hohem Komplexitätsgrad sowie räumlich eng begrenzten Schadenswirkungsbereichen. 

Für räumlich ausgedehntere Ereignisse, sowie für Ereignisse mit erhöhtem Gefahrenpotential 

und erhöhtem Komplexitätsgrad im Einsatz ist das ausführlichere und umfangreichere 

Instrument des Einsatzplanes vorzusehen. Das Einsatzhandbuch dient zur Bewältigung 

von Naturereignissen von sehr hohem Komplexitätsgrad, einer weiträumigen Ausdehnung 

und mit sehr großen Schadenspotentialen. 

EINSATZKARTE 

Mit der Ausarbeitung von Einsatzkarten werden folgende Ziele verfolgt (Ernst Basler + 

Partner, 2008): 

– Verringerung der durch Naturereignissen bedingten Schäden an Menschen, Gütern, 

Umwelt sowie der wirtschaftlichen Tätigkeit 

– Optimierung des Personal- und Mitteleinsatzes 

– Sicherstellung des Informationstransfers während eines Ereignisses 

– Bereitstellung einer Entscheidungsgrundlage für situatives Handeln 

– Verbesserung der Sicherheit der Einsatzkräfte 

– Die Einsatzkarte dient stets dem aktiven Einsatzmanagement. Mit ihr wird ein Instrument 

zur Überbrückung des Zeitraumes vom Erkennen der Gefahr bis zur Umsetzung von 

technischen oder raumplanerischen Maßnahmen geschaffen. Darüber hinaus wird anhand 

der Einsatzplanung ein Management von Restrisiken möglich. Die Einsatzkarte ist ein 

Hilfsmittel und soll auf der Ebene der Einsatzleitung zur Unterstützung bei der Entscheidung 

über zu treffende Erstmaßnahmen verwendet werden. Durch die Vorwegnahme der 

Risikoanalyse und durch die präventive gedankliche Ausarbeitung möglicher Erstmaßnahmen 

verschafft die Einsatzkarte den Entscheidungsträgern einen Zeitgewinn und einen 

Wissensvorsprung. Einsatzkarten können prinzipiell für alle operativen Kräfte und für jedes 

Gefahrenszenario erstellt werden (Eschgfäller M., 2012). 

MÖGLICHKEIT UND GRENZEN BEI EINSÄTZEN 

Die Bandbreite der möglichen Szenarien, denen man bei Einsätzen in Folge von Naturereignissen 

wie Hochwasser, Muren, Lawinen begegnen kann, ist groß. Die Ereignisdokumentationen, 

Gefahrenhinweiskarten, Gefahrenstudien oder Gefahrenzonenpläne liefern realistische 

und wissenschaftlich fundierte Hinweise über die Entwicklungsmöglichkeit dynamischer 

Naturgefahrenprozesse und bilden die Basis der Einsatzkarten. Die darin festgehaltenen 

kartografischen Informationen sind wichtige Hinweise zu Gefahrenbereichen, die nur unter 

besonderen Vorsichtsmaßnahmen oder überhaupt nicht betreten werden dürfen und zu 


sicheren Räumen, welche z.B. für die Einrichtung von Bereitstellungsflächen verwendet 

werden können. Über die Vorabgrenzung von Gefahrenzonen ist eine erste Abschätzung der 

Handlungsmöglichkeiten im Raum möglich. Durch die Lageerkundung vor Ort werden die 

kartographischen Erstinformationen zur vermuteten Lage ergänzt, bestätigt oder korrigiert. 

Die meisten der Kleinst- und kleinen Einzugsgebiete (

EINSATZKARTE: Fartleisbach ID - G.230 

Ebenwieserhof 

Haselstauder Brücke 

Obereishof 

E-Werk Schwarz 

Fartleisbach 

Uhlen 

Untereishof 

E5 

Haselstauderhof 

Pfeiftalerhof 

Brücke Sportanlagen 

Sehr hohe Gefahr 

Hohe Gefahr 

Mittlere Gefahr 

 

Abbildung 1: Ausschnitt aus einer Einsatzkarte (Eschgfäller M., 2012) 


Lawine) beschreibt. In der Karte selbst sind die Gefahrenbereiche, Risiken und Erstmaßnahmen 

anhand von taktischen Zeichen dargestellt (siehe Abbildung 1). 

Auf der Rückseite der Einsatzkarte findet sich eine Auflistung der Risiken. Die Klassifizierung 

der Risiken wurde von den Richtlinien zur Erstellung der Gemeindezivilschutzpläne der 

Autonomen Provinz Bozen in leicht abgeänderter Form übernommen. In der Beschreibung 

werden die erkannten Risiken analysiert, aufgelistet und in Risikokategorien gegliedert. 

Darauf folgt ein Abschnitt über Entscheidungsregeln im Einsatz und über Erstmaßnahmen. 

Die Entscheidungsregeln und Maßnahmen werden hierbei ebenfalls von den Feuerwehren 

ausgearbeitet und nach Prioritäten gereiht. Am Ende des textlichen Teils sind die zu benachrichtigenden 

Stellen und Personen aufgelistet. Ein Register mit den dazugehörigen Kontaktdaten 

wird von der Feuerwehr gesondert in der Einsatzzentrale aufbewahrt, oder ist Teil des 

Gemeindezivilschutzplans (Eschgfäller M., 2012). 

EINSATZPLAN 

Ist das Ereignis zu komplex, der Schadenswirkungsbereich zu groß oder nimmt das Gefahrenpotential 

zu, so tritt an die Stelle der Einsatzkarte, der Einsatzplan. Dieser besteht aus 

mehreren thematischen Kartenblättern und aus einer ausführlichen Risikoanalyse und aus 

einem detaillierten Handlungsschema. Hier können auch im Zuge eines Naturereignisses 

auftretende Dominoeffekte wie Ausfälle in der Strom- oder Wasserversorgung berücksichtigt 

und behandelt werden. Zudem wird das koordinierte Zusammenspiel mehrerer Einsatzorganisationen 

festgeschrieben. 

EINSATZHANDBUCH WILDBACH MIT SONDERPLAN HOCHWASSERDIENST 

Die Landesabteilung Wasserschutzbauten hat für Mur- und Hochwasserereignisse eine Art 

„roadmap“ ausgearbeitet, das so genannte „Einsatzhandbuch Wildbach“. Dieses legt die 

Zuständigkeiten, Tätigkeiten und Abläufe vor und während eines Ereignisses detailliert fest. 

In diesem Einsatzhandbuch ist auch der „Sonderplan Hochwasserdienst“ enthalten, welcher 

die Maßnahmen im Falle eines Hochwassers an Etsch und Eisack regelt. Es wurden Aufgaben 

formuliert, Abläufe durchgespielt und festgelegt und Kompetenzen zugewiesen – dies alles 

mit dem Ziel, klare Regeln für Katastrophenlagen festzuschreiben. Das Einsatzhandbuch 

Wildbach umfasst Kapitel zu Übersichten über Strukturen, Führungsorganisation, Aufgaben 

von Bereitschaftsdiensten, Aufgaben der Einsatzleitung und Aufgaben der Stabsarbeit. Als 

Führungssystem für die Bewältigung eines Einsatzes im Ereignisfall wird die „Dienstvorschrift 

100 (DV 100) zur Führung und Leitung im Einsatz“ angewandt, d.h. der Einsatz wird mit 

Hilfe der Stabsarbeit (S1-S6) abgewickelt: S1 Personal, S2 Lage, S3 Einsatz, S4 Versorgung, 

S5 Information und S6 Kommunikation. Das Einsatzhandbuch begleitet die Verantwortlichen 

Schritt für Schritt in ihrer Einsatztätigkeit. Alle wichtigen Informationen sind in knapper 

Form in einer Art Checkliste dargestellt. Für den speziellen Fall von Hochwasser an der Etsch 

greift der „Sonderplan Hochwasserdienst“. Dieser legt anhand einer Checkliste alle Abläufe 

der Bereitschaftsdienste der verschiedenen Landesabteilungen, der Einsatzleitung und der 


Stabsarbeit im Detail fest. Die Einsatztätigkeit wird geografisch in neun „technische Wildbachbezirke“ 

unterteilt, die den bestehenden Feuerwehrbezirken entsprechen. Die gesetzlich zugewiesenen 

Aufgaben der Abteilung Wasserschutzbauten gliedern sich in Vorhersage, Vorbeugung, 

Bereitschaftsdienst, Einsatz und Wiederinstandsetzung. 

Vorhersage: Durch die modernen Mittel der Meteorologie ist die Vorhersage von extremen 

Wetterlagen heute relativ zuverlässig. Im Ernstfall erörtern das Hydrographische Amt, die 

Abteilung Wasserschutzbauten und alle weiteren betroffenen Behörden und Organisationen 

im Rahmen von Bewertungskonferenzen die Lage. 

Vorbeugung: Das Hauptaugenmerk gilt der Erkennung von Gefahren, Vorbeugung von 

Schäden, Planungs- und Absicherungsmaßnahmen. 

Bereitschaftsdienst: Der rund um die Uhr erreichbare Bereitschaftsdienst aktiviert im 

Ereignisfall alle erforderlichen Dienste der Abteilung. 

Einsatz: Im Einsatzfall sind die Aufgaben der Abteilung: 

– die Vorhersage von Ereignissen 

– die Beratung der Zivilschutzbehörden der Landes- und Gemeindeverwaltungen 

– die Überwachung der Wasserläufe und Stauanlagen 

– der Hochwasserdienst an Etsch und Eisack 

– die direkte Intervention mit Mannschaften und technischem Gerät an Schadensorten 

– Erstellung einer Dokumentation all dieser Tätigkeiten 

Wiederinstandsetzung: Nach dem Ende der akuten Phase von Ereignissen, die Schäden an 

Wasserläufen oder anderen Einrichtungen verursacht haben, behebt die Abteilung in erster 

Linie die Schäden im eigenen Zuständigkeitsbereich. 

Für den Ernstfall sind die Abläufe in Flussdiagrammen detailliert festgehalten. Erreichen 

Etsch und/oder Eisack Pegelstände, die auf eine erhöhte Gefahr hinweisen, informiert das 

Hydrografische Amt den Bereitschaftsdienst der Abteilung Wasserschutzbauten. Steigen die 

Pegel weiter auf das Niveau der „Vorwarnstufe“, werden die Techniker des Hochwasserdienstes 

alarmiert und die Hochwasserzentrale besetzt. Eine Bewertungsgruppe bestehend aus 

Fachleuten der Abteilung Wasserschutzbauten und des Hydrographischen Amtes berät über 

die Lage. Gleichzeitig werden über die Landesnotrufzentrale die Bezirkseinsatzzentralen der 

Feuerwehren alarmiert, die ihrerseits die zuständigen lokalen Freiwilligen Feuerwehren zum 

Einsatz rufen. Bei Erreichen der „Warnstufe“ kontrollieren Deichwachen (Personal der 

Freiwilligen Feuerwehr) die Flussdämme. Nach weiterem Ansteigen des Wasserstandes 

müssen Pumpwerke der Abzugsgräben und eines E-Werk abgeschaltet werden. Für die 

Szenarien „drohender Dammbruch“ und „erfolgter Dammbruch“ sind weitere spezifische 

Mitteilungen und Maßnahmen vorgesehen. Sobald die Pegelstände wieder unter die 

„Warnstufe“ sinken, werden die Deichwachen an den Dämmen abgezogen und bei weiterem 


Sinken des Pegels unter die „Vorwarnstufe“ wird auch die Besetzung der Pegelmessstellen 

aufgehoben und sodann der Hochwasserdienst beendet. Sämtliche Aktionen werden mit 

genau vordefinierten Formularen von der Hochwasserzentrale aus angeordnet. Nach 

Übungen und nach Einsätzen im Ernstfall wird in Nachbesprechungen der Ablauf noch 

einmal in allen Details besprochen und eventuelles Verbesserungspotential ermittelt 

(Egger P., 2014). 

6. ERGEBNISSE 

Durch das Verfassen eines Handbuches zum Erstellen von Einsatzkarten für die Bereiche 

Wasser- und Lawinengefahren ist dieses Einsatzinstrument landesweit in Südtirol/Italien auf 

eine einheitliche Basis gestellt. Somit wird eine uniforme Lesart und Interpretation der 

Karten erreicht. Darüber hinaus fügen sich künftig diese Einsatzinstrumente nahtlos in die 

übrigen planerischen Zivilschutzinstrumente, wie den Gemeindezivilschutzplänen und 

anderen Sonderplänen ein und stehen mit diesen nicht im Widerspruch. 

In Zusammenarbeit mit den örtlichen Feuerwehren wurden bereits mehrere Einsatzkarten 

und Einsatzpläne ausgearbeitet. Durch die Mitwirkung der Einsatzkräfte bei der Erstellung 

der Unterlagen erhalten sie einen professionellen Zugang zur Thematik und werden für 

besondere Gefahrensituationen und Risiken bei Hochwasser-, Muren- und Lawinenereignissen 

sensibilisiert. Kausale Zusammenhänge und mögliche Dominoeffekte in der Ereigniskette 

werden von ihnen erkannt und die Führungskräfte können sich auf die zu treffenden 

Entscheidungen und Maßnahmen vorbereiten. Zudem stellen die Einsatzkarten ein praktikables 

Übungsinstrument dar. 

Seit Bestand des Sonderplanes Etsch im Jahr 2000 wurden schon mehrere Male Hochwassereinsätze 

an der Etsch gemäß diesem Plan mit Erfolg abgewickelt, so z.B. im August 2014. 

Darüber hinaus werden sämtliche Abläufe einmal im Jahr im Rahmen einer groß angelegten 

Hochwasserübung an Etsch und Eisack beübt. Dabei werden in der Regel von einem 

Feuerwehrbezirk mehrere Szenarien durchgespielt, während in den übrigen Bezirken 

lediglich die Kommunikationswege getestet werden. Bei der Hochwasserübung im Herbst 

2014 wurden z.B. im Feuerwehrbezirk Brixen mehrere Hochwassersituationen simuliert. 

Insgesamt waren dabei an die 400 Personen im Einsatz. Das Einsatzhandbuch Wildbach 

wurde im Jahre 2007 eingeführt und seitdem wurden schon mehrere Einsätze bei Mur- und 

Hochwasserereignissen an Wildbächen damit abgearbeitet. So wurde bei den Ereignissen im 

Pfitschertal und im Sterzinger Raum im Sommer 2012 die Hochwasserzentrale besetzt und 

die Einsätze wurden von dort aus gemäß Handbuch koordiniert. 

7. FAZIT 

Für die Notfallplanung bei Wassergefahren liegen nun in Südtirol mit den Einsatzkarten, den 

Einsatzplänen und dem Einsatzhandbuch, den Zivilschutzdiensten drei Instrumente vor, die 

eine wesentliche Verbesserung der Notfallplanung ermöglichen und damit einen wichtigen 

Beitrag zum integralen Risikomanagement leisten. 


Mit den Einsatzkarten und -plänen werden den Feuerwehren für Mur- und Lawinenereignisse 

Planungsinstrumente zur Bewältigung der Erstphase bei Einsätzen an ausgewählten 

Wildbächen und Lawinenstrichen zur Verfügung gestellt. Es handelt sich hierbei um 

Siedlungsbereiche mit einem gewissen Risikopotential und/oder einem hohen Komplexitätsgrad 

im Einsatz. Mit diesen Einsatzhilfen wird eine Verbesserung der Sicherheit der Einsatzkräfte, 

eine Optimierung des Personal- und Mitteleinsatzes, die Sicherstellung der Informationsflüsse 

und eine Entscheidungsgrundlage für situatives Handeln zur Verfügung gestellt. 

Das dadurch verbesserte aktive Einsatzmanagement hat zum Ziel durch gezielte einsatztaktische 

und operative Vorbereitungen dem nahenden Naturereignis bestmöglich zu begegnen 

sowie ereignisbedingte Schäden an Menschen, Tieren, Gütern und Umwelt zu verringern und 

Wiederinstandsetzungszeiten so kurz wie nötig zu halten. 

7. LITERATUR 

- Autonome Provinz Bozen-Südtirol, Abteilung Wasserschutzbauten (2015a). Ereigniskataster 

ED30. Unveröffentlicht. 

- Autonome Provinz Bozen-Südtirol, Abteilung Wasserschutzbauten (2015b). Schutzbautenkataster 

Baukat30. Unveröffentlicht. 

- Egger P. (2014). Der Ernstfall – In: Geschichte der Etsch von Kurt Werth. Athesia Verlag: 

291-294. 

- Ernst Basler + Partner AG. (2008). TIMIS flood – Interventionskarte. Konzept 2008. 

Zollikon. Unveröffentlicht: 1 

- Eschgfäller M. (2012). Handbuch zum Erstellen von Einsatzkarten für die Bereiche 

Wasser- und Lawinengefahren. IREK – Integrales Raumentwicklungskonzept für ausgewählte 

Lebensräume des Wipptales, Modul 4 Schutz- und Raumentwicklungskonzepte. Im Auftrag 

der Abteilung Brand- und Zivilschutz – Bozen. unveröffentlicht. 

Gebäudeversicherung Graubünden; Amt für Wald Graubünden (2009): Kurzanleitung 

Einsatzkarte. Vom Wissen zum Handeln. With assistance of Markus Fischer, Reto Hefti. 

Davos: 3. 

- Romang H. (2006). Einsatzkarte – vom Wissen zum Handeln. Pilotprojekt Einsatzkarte 

Schlussbericht. Davos: 3, 10. 

- Romang H., Wilhelm Ch. (2008). Kurzanleitung Einsatzkarte – vom Wissen zum Handeln. 

Davos: 4. 

- Staffler H. et al. (2008). Organisation der Einsatzleitung in der Abteilung Wasserschutzbauten 

der Autonomen Provinz Bozen – Südtirol bei Hochwasserereignissen. Interpraevent 2008 

– Conference Proceedings, Vol. 2: 87-97. 



Multiple early warning system on rock walls above 

a railway line in the Bernese Oberland (Switzerland) 

Multiples Früherkennungssystem an Felswänden 

oberhalb einer Eisenbahnstrecke im Berner Oberland 

(Schweiz) 

Ueli Gruner, Dr. phil. nat. 1 ; Hans-Heini Utelli 2 

ABSTRACT 

From the rock walls of a mountain section of the „Lötschberg" railway line between the 

Bernese Oberland and Valais (Switzerland), rockfall events have occurred repea-tedly, that 

are drastic for railway operations. That's why numerous and costly protective structures have 

been previously built and compemented steadily until today. However, these can not protect 

from large-volume falls. In order to further reduce the remaining death risk, a multiple early 

warning system was installed. The monitoring methods include simple manual measurements 

(distance detection between two pins) as well as best practices (such as tachymetry and Tele 

Joint meters) and most modern technology (laser and radar scanning). For the implementation 

of the results of this monitoring program, a detailed early warning dispositif was 

established, that indicates, which measures have to be taken in the case of relevant rock 

movements. 


Aus den Felswänden einer Bergstrecke der Lötschberg-Eisenbahnlinie zwischen dem Berner 

Oberland und Wallis (Schweiz) haben sich immer wieder für den Bahnbetrieb einschneidende 

Sturzereignisse ereignet. Deswegen wurden schon früher zahlreiche und aufwändige 

Schutzbauten erstellt und bis heute immer wieder ergänzt. Diese können jedoch grossvolumige 

Stürze nicht zurückhalten. Um das verbleibende Todesfallrisiko weiter zu reduzieren, 

wurde ein multiples Früherkennungssystem eingerichtet. Die Überwachungsmethoden 

reichen von einfachen Handmessungen (Abstandserfassung zwischen zwei Bolzen) über 

bewährte Verfahren wie Tachymetrie und Telejointmeter bis zur modernsten Technologie 

(Laser- und Radar-Scanning). Für die Umsetzung der Ergebnisse aus dieser Messüberwachung 

besteht ein detailliertes Frühwarndispositiv, das die einzelnen Handlungsschritte bei 

relevanten Felsbewegungen aufzeigt. 

KEYWORDS 

rock fall; rock monitoring; early detection; early warning; alert dispositif 

1 Kellerhals+ Haefeli AG Bern, SWITZERLAND, ueli.gruner@k-h.ch 

2 Impuls AG 



EINLEITUNG 

Die BLS Netz AG betreibt auf der Strecke zwischen Bern und Brig (Berner Oberland/Wallis, 

Schweiz; Fig. 1) seit über 100 Jahren die Lötschberglinie, welche Gebiete nördlich der Alpen 

mit Italien verbindet. Auch nach der Eröffnung des Lötschberg-Basistunnels im Jahr 2007 

wird die doppelspurige Bergstrecke zwischen Frutigen und Brig zum Transport von Personen 

und Gütern verwendet. Auf dem rund 17 km langen Abschnitt zwischen Frutigen und 

Kandersteg im Berner Oberland verläuft die Bahnlinie unter teilweise hohen Felswänden. 

Für die Bahnlinie besteht auf vielen Abschnitten ein grosses Gefahrenpotenzial. So kam es 

in der Vergangenheit immer wieder zu Felsstürzen, welche neben Sachschäden z.T. auch 

mehrtägige Bahnunterbrüche verursachten. 

Im Rahmen einer Gefahren- und Risikoanalyse wurden für die gravitativen Naturgefahrenprozesse 

auf der Bahnstrecke Frutigen - Kandersteg diejenigen Abschnitte ausgeschieden, 

bei welchen für Mensch und Anlage ein erhöhtes Risiko besteht (Utelli et al., 2016). 

Figur 1: Situation des Überwachungsgebiets der Eisenbahnstrecke Frutigen – Kandersteg (Lötschberglinie, rot markiert) mit den rund 30 

Liefergebieten (Flächen) und den Gefahrengebieten (eingekreiste Flächen) 

Als Schutzziel wurde festgelegt, dass das durch Naturgefahren verursachte, zusätzliche 

individuelle Todesfallrisiko für Bahnreisende nicht mehr als 1 x 10-5/Jahr betragen soll. Um 

dies zu erreichen, wurde ein Massnahmenkonzept bzw. -plan erarbeitet. Dieser sieht neben 

den bereits weitgehend umgesetzten baulichen Massnahmen auch organisatorische Massnahmen 

vor. Dazu zählt nebst einer elektronischen Steinschlagüberwachung in den bestehenden 


Schutznetzen (mit vorübergehendem Fahrleitungsunterbruch bei einem Input) ein in den 

letzten Jahren eingerichtetes, multiples Früherkennungssystem für Felsbewegungen. Mit 

diesem System (bzw. mit den bei signifikanten Bewegungen vorzunehmenden Massnahmen) 

kann das nach den baulichen Massnahmen verbleibende Personenrisiko nochmals um rund 

2/3 reduziert werden. 

Die Früherkennung beruht auf sieben verschiedenen Messmethoden (vgl. Tab. 1), die weiter 

unten näher beschrieben werden. Für das System gelten gewisse, vor allem geologisch 

bedingte Voraussetzungen, auf die im nachfolgenden Kapitel näher eingegangen wird. 

Tabelle 1: Zusammenstellung der Messmethoden für die Erfassung von Felsbewegungen oberhalb der Eisenbahnstrecke 

Frutigen – Kandersteg 

Messmethoden Anzahl Messpunkte Messrhythmus/Jahr Messgenauigkeit 

Tachymetrie 41 2 x ca. 1 mm 

Handmessstellen 70 2 x ca. 1 mm 

Siegel 32 2 x ca. 0.5 mm (Risse) 

Radar-Scanning 3 Gefahrengebiete 1 x ca. 1 mm 

Laser-Scanning 2 Gefahrengebiete 1 x ca. 2 cm bis 4 cm 

Telejointmeter 3 x 3 („Gstryfet Birg“) permanent 0.1 mm 

Inklino-/Extensometer 6 („Gstryfet Birg“) alle 3 Jahre 1 mm auf 10 m/0.1 mm 

VORAUSSETZUNGEN FÜR DAS FRÜHERKENNUNGSSYSTEM 

Bei einer Früherkennung von grösseren Felsbewegungen im Überwachungsgebiet gehen die 

Bearbeiter von folgenden Voraussetzungen aus: 

Grössere Felspartien brechen i.d.R. nicht spontan ab. Felsinstabilitäten entstehen nämlich 

durch einen häufig Monate oder sogar Jahre dauernden Entfestigungs- und Ablösungsprozess 

in Form eines Kohäsionsabbaus. Sie kündigen sich somit in erster Linie durch eine Beschleunigung 

der Felsbewegungen an. Diese fortschreitenden Ablösungsprozesse können durch eine 

gut instrumentierte und an den massgeblichen Stellen eingerichtete Felsüberwachung meist 

rechtzeitig erkannt werden. 

An die Messgenauigkeit sind unterschiedliche Anforderungen zu stellen. Bei grösseren Fels - 

massen sind Bewegungen in der Grössenordnung von cm bis dm möglich, ohne dass ein 

Absturz erfolgt. Eine Messgenauigkeit im mm-Bereich ist somit absolut genügend. Dies umso 

mehr, als sich infolge von Temperaturschwankungen Felsbewegungen ergeben können, die 

allein im Tagesverlauf Unterschiede von bis zu 1 mm und im Jahresverlauf solche von bis zu 

4 mm zeigen, ohne dass sich dabei eine generelle Veränderung der Felsstabilität ergibt 

(Gruner 2008, Gruner 2012). 


Bei einer frühzeitigen Erkennung von relevanten Bewegungen bleibt i.d.R. genügend Zeit, 

um entsprechende Massnahmen vor einem definitiven Absturz zu ergreifen (z.B. Sprengungen, 

bauliche Massnahmen etc.). 

GEOLOGIE UND LIEFERGEBIETE DER STURZPROZESSE 

Die Felswände oberhalb der Bahnlinie bestehen aus z.T. massigen Kalken und Kieselkalken 

der Kreide (Wildhorn-Decke des Helvetikums). Die ausgeprägten Verfaltungen und Verschuppungen 

der Schichtabfolgen führen dazu, dass das Gestein ein z.T. enges Trennflächengefüge 

aufweist. Diese Ausgangslage hat zur Folge, dass es im Untersuchungsgebiet immer wieder 

zu Stein-und Blockschlag sowie auch zu Felsstürzen (>100 m 3 ) gekommen ist. 

Im Rahmen der eingangs erwähnten Gefahren- und Risikoanalyse wurden im Untersuchungsgebiet 

rund 30 Liefergebiete für Sturzprozesse ausgeschieden. Einige davon stellen für die 

Bahnlinie jedoch keine Gefährdung dar (Tunnels, ausreichend dimensionierte Schutzbauten, 

geringe Reichweite der Sturzkörper etc.). Es verbleiben neun so genannte Gefahrengebiete, 

aus denen grobblockige Felsteile oder sogar grossvolumige Felspartien ausbrechen und trotz 

den errichteten Schutzwerken bis auf die Bahnlinie gelangen können (vgl. Fig. 1). Das Ziel 

des Früherkennungssystems ist es, diese potenziellen grossen Felsinstabilitäten messtechnisch 

möglichst vollständig und frühzeitig zu erfassen, um allfällig notwendige Massnahmen rechtzeitig 

ergreifen zu können. 

MESSMETHODEN 

Tachymetrie 

Drei Gefahrengebiete werden tachymetrisch überwacht (total 41 Reflektoren, Messdistanz 

ca. 600 bis 800 m). Die Reflektoren wurden an grösseren, potenziell instabilen Felspartien 

(meist > mehrere 10 m 3 ) montiert, wo klare Ablösungserscheinungen vom gesunden Felsverband 

festgestellt werden konnten. Vorteilhaft ist, dass die tachymetrische Fernüberwachung 

nebst der genügenden Genauigkeit von rund 1 mm auch kostengünstig ist und bei 

Bedarf problemlos um weitere Messpunkte ergänzt werden kann. Zudem können die 

Messpunkte, falls notwendig, auch permanent gemessen und mit Schwellenwerten (z.B. für 

eine Intervention) versehen werden. Nachteilig ist bei dieser Methode, dass sie bei schlechter 

Sicht (Nebel) keine brauchbaren Resultate liefert und zudem nur eine bedingte räumliche 

Aussagekraft hat. 

Handmessungen 

Die insgesamt 70 Handmessstellen (total 130 Messpunkte) und die 32 Siegel liegen verteilt 

auf alle neun Gefahrengebiete. Die Messstellen befinden sich an Felsklüften, d.h. es werden 

i.d.R. nur Teilbereiche einer Felsinstabilität gemessen. Dank der speziell konstruierten 

Messspitzen (aufgeschraubt auf in den Fels eingebohrten Bolzen) liegt die Messgenauigkeit 

im mm-Bereich. Die Messstellen wurden an mittelgrossen, potenziell instabilen Felspaketen 

eingerichtet (ca. 5 bis 50 m 3 ), bei welchen infolge der Vegetation kaum eine andere sinnvolle 


Überwachungsmöglichkeit bestand. Vorteilhaft ist, dass die Handmessungen nebst der 

genügenden Genauigkeit auch kostengünstig sind und bei Bedarf problemlos um weitere 

Messstellen ergänzt werden können. Nachteilig ist bei dieser Methode, dass das Gefahrengebiet 

jeweils begangen werden muss; zudem besteht nur eine bedingte räumliche Aussagekraft. 

Siegel 

Die 32 eingerichteten Siegel aus Zement wurden meist als Ergänzung zu den Handmessstellen 

an den gleichen Gefahrenstellen angebracht. Es gelten die gleichen Merkmale wie bei der 

Handmessmethode. 

Radar-Scanning 

Mit einem Radar-Scanning (Radar Interferometrie) können von einem Geländestandort 

aus mittels eines mobilen Radargerätes ganze Felswände abgetastet werden (vgl. Wiesmann, 

Gruner 2011). Bei einer Messdistanz von rund 700 m bis 900 m beträgt die Messgenauigkeit 

rund 1 mm und die minimal detektierbare Felsfläche ca. 6 m 2 . Mit dem Radar-Scanning 

werden von einem Standort aus drei Gefahrengebiete überwacht. Vorteilhaft ist, dass das 

Radar-Scanning nebst der genügenden Messgenauigkeit eine grosse Felsfläche mit einer 

einzigen Messung erfasst und dass das Gerät auch permanent eingesetzt und mit einer 

Alarmierungsanlage kombiniert werden kann. Nachteilig ist bei dieser Methode, dass sie 

relativ teuer ist und dass die Vegetation in der Felswand falsche Signale übermitteln und 

so eine vermeintliche Bewegung darstellen kann. 

Laser-Scanning 

Mit dem mobilen Gerät für das Laser-Scanning kann ebenfalls eine Felswand als Ganzes 

erfasst werden. Die Messgenauigkeit ist bei dieser Methode jedoch mit 2 cm bis 4 cm 

bedeutend niedriger als bei allen übrigen Messverfahren. Diese Messmethode wird bei zwei 

Gefahrengebieten eingesetzt, wo die Messdistanzen relativ gering sind (ca. 300 m bis 500 m) 

und die Messgenauigkeit somit etwas besser ist. Die minimal detektierbare Felsfläche liegt bei 

wenigen m 2 . Die gegenüber dem Radar-Scanning geringere Messgenauigkeit wird durch die 

geringeren Kosten etwas aufgewogen. Auch beim Laser-Scanning kann das Gerät für eine 

permanente Überwachung eingesetzt werden. Nachteilig ist, dass auch hier die Vegetation 

in der Felswand falsche Signale bzw. eine vermeintliche Bewegung suggeriert. 

Telejointmeter 

Im zentralen Abschnitt der Strecke Frutigen – Kandersteg befindet sich, etwa 300 Höhenmeter 

über der Bahnlinie, eine grosse instabile Partie mit einer Kubatur von rund 100‘000 m 3 

(„Gstryfet Birg“). Die rund 100 m hohe Felswand ist bergseits zum grossen Teil durch eine 

weit offene Hauptkluft vom stabilen Wandbereich abgetrennt. Die felsmechanisch mögliche 

Kippbewegung der labilen Felsmasse wird – nebst tachymetrischen Messungen – mittels an 


drei Standorten der Hauptkluft angeordneten Telejointmeter permanent, d.h. jede Minute, 

erfasst. 

Inklino-/Extensometer 

Gemäss dem geologischen Modell bewegt sich die instabile Gebirgsmasse des „Gstryfet Birg“ 

entlang von mehreren Gleitebenen talwärts. Zur Früherkennung solcher Gleitbewegungen 

werden ergänzend in sechs Bohrlöchern periodisch Inklino- und Extensometermessungen 

gemacht. 

FRÜHWARN- UND ALARMDISPOSITIV 

Durch das grosse Spektrum von Messmethoden mit regelmässigen Messungen wird eine 

grosse Anzahl von Messwerten generiert. Um diese Resultate sinnvoll und praxistauglich zu 

verarbeiten, wurde im Jahr 2012 zusammen mit dem Früherkennungssystem ein Frühwarndispositiv 

mit einem detaillierten Verfahrensablauf entwickelt (vgl. Fig. 2). 

Figure 2: schematische Darstellung des Verfahrensablaufes beim Frühwarndispositiv der Eisenbahnstrecke Frutigen – Kandersteg 

Grundsätzlich wird mit der Festlegung von Schwellenwerten zurückhaltend umgegangen, 

dies v.a. auch, um vorsorgliche Sperrungen der Bahnlinie wegen unbegründeter Alarme 

möglichst zu vermeiden. Die Beurteilung erfolgt deshalb schrittweise in Bezug auf die relativen 

Bewegungen (Trendbeurteilung). 

Gemäss dem Dispositiv ist es die Funktion und Aufgabe des Gelogen, die Messergebnisse 

regelmässig zu beurteilen und zu interpretieren. Zeigen einzelne Messwerte signifikante 

Felsbewegungen, findet eine Feldbeurteilung statt, und das Messdispositiv wird gegebenenfalls 

ergänzt und/oder der Messrhythmus erhöht. Die Bahnbetreiberin – die BLS Netz AG – 

wird bei diesen Schritten jeweils informiert. Werden Schwellenwerte relevant überschritten, 

kann eine permanente Felsüberwachung eingerichtet werden (mit Alarmdispositiv). 


Denkbar sind zudem sofortige bauliche Massnahmen wie z.B. Felssicherungen oder auch 

Sprengungen. Im Notfall kann die Strecke gesperrt werden. Bei diesen Schritten – als 

Intervention bezeichnet – wird die Bahnbetreiberin in die Entscheidung und auch in die 

Verantwortung eingebunden. 

Bei der grossen, rund 100‘000 m 3 umfassenden, labilen Felspartie „Gstryfet Birg“ wurde 

das Zeitfenster zwischen periodischen Einzelmessungen als (zu) lang betrachtet, d.h. die 

Felsbewegungen werden dort bereits seit längerer Zeit permanent mit Telejointmeter erfasst. 

Es besteht ein separates Alarmdispositiv mit definierten Schwellenwerten für die einzelnen 

Gefahrenstufen. Für jede der vier ausgeschiedenen Gefahrenstufe (Normalfall, Warn-, 

Alarm- und Interventionsstufe) gibt es einen Vorgehensplan mit jeweils einzelnen Vorgehensschritten 

und Meldeabläufen. 

ERGEBNISSE 

Seit der Einführung des integralen Früherkennungssystems im Jahr 2012 haben sich keine 

relevanten Felsbewegungen ergeben, welche eine wichtige Handlung im Sinne des Frühwarndispositivs 

ausgelöst hätten. Handmessungen, Siegelüberwachung, tachymetrische wie 

auch Inklino- und Extensometermessungen zeigten bisher keine beschleunigenden Felsbewe- 

Figure 3: Aufnahme eines Laser-Scanning mit festgestelltem Felsausbruch von rund 2 m 3 (Kreis; oben), Darstellung des Felsausbruchs 

als Punktewolke (Fläche, Profile; unten) 


gungen. Allerdings ist bei den bereits seit über 10 Jahren laufenden Messungen am „Gstryfet 

Birg“ ein klarer, talwärts gerichteter Trend der Bewegung erkennbar (jährlich 

jeweils ca. 1-2 mm); dies wird jedoch gemäss Frühwarndispositiv als „Normalfall“ betrachtet. 

Die permanenten Bewegungsmessungen mittels Telejointmeter am „Gstryfet Birg“ zeigen den 

erwähnten Trend der Bewegung ebenfalls. Allerdings sind hier die Temperatur bedingten bzw. 

jahreszeitlichen Einflüsse auf die Felsbewegungen erkennbar: Im Winter bei kalter Witterung 

erfolgt eine Kontraktion des Gesteins und somit eine leichte Öffnung der Kluft. Im Sommerhalbjahr 

weitet sich die Gesteinsmasse etwas aus und die Kluft schliesst sich leicht (vgl. 

Gruner 2008). Diese felsmechanische Erkenntnis dient auch zur Interpretation der Einzelmessungen 

bei den anderen Messmethoden. 

Das Radar-Scanning im Abschnitt Frutigen – Kandersteg hat bisher keine relevanten Felsbewegungen 

aufgezeichnet. Die Erfahrung aus einem benachbarten Überwachungsgebiet für 

eine Nationalstrasse hat jedoch gezeigt, dass dort auch bei kleinen Felsvolumina von wenigen 

10 m 3 Bewegungen im mm-Bereich erkannt und anschliessend im Feld eindeutig verifiziert 

werden konnten. 

Auch mittels Laser-Scanning konnten auf der Strecke Frutigen – Kandersteg bisher keine 

Felsbewegungen frühzeitig detektiert werden. Hingegen wurden kleinere Felsabbrüche 

festgestellt (Volumen bis ca. 5 m 3 ; vgl. Fig. 3), wo allerdings vorgängig keine entsprechenden 

Signale festgestellt wurden. Dies dürfte einerseits auf die eher schlechte Messgenauigkeit 

zurückzuführen sein, anderseits aber auch auf das Messintervall (jährliche Messung), 

innerhalb welchem es zu einer schnellen Ablösung aus dem Felsverband gekommen ist. 

Dank dem Hinweis aus dem Laser-Scanning wurden die einzelnen Stellen nach dem Ereignis 

jeweils besichtigt, um nach allfällig verbleibenden Felsinstabilitäten zu suchen (was aber 

nicht der Fall war). 


Ein integrales Früherkennungssystem kann, falls es stufengerecht und massgeschneidert auf 

die Lokalverhältnisse eingerichtet wird, für die Sicherheit der Bahn einen wichtigen Beitrag 

leisten. Dabei ist eine Kombination von bewährten Messmethoden (z.B. Handmessungen) 

und moderner Technologie (z.B. Radar-Scanning) anzustreben. Entscheidend ist jedoch, dass 

die Messergebnisse durch erfahrene, mit den lokalen Verhältnissen vertraute Geologen 

ausgewertet und interpretiert werden. Dazu braucht es nicht nur Fachwissen, sondern auch 

eine auf Erfahrung aufgebaute Intuition, um die Signale eines kommenden Ereignisses 

rechtzeitig zu erkennen und entsprechend zu interpretieren. Dies bedeutet, dass der Geologe 

ein gewisses Mass an Gelassenheit aufweisen sollte, damit die Bahnstrecke nicht bereits bei 

kleinen Felsbewegungen gesperrt wird. Es bedeutet aber auch, dass er die Messresultate selbst 

in ruhigen Zeiten mit Aufmerksamkeit verfolgen sollte, damit er relevante Bewegungssignale 

– im Sinne der Frühwarnung – rechtzeitig erkennt. 


REFERENZEN 

- Gruner U. (2008). Klimatische und meteorologische Einflüsse auf Sturzprozesse. 

Proceed. Conf. INTERPRAEVENT 2008/2: 147-158. 

- Gruner U. (2012). Sturzereignisse in der Schweiz – eine statistische Auswertung. 

Swiss Bull. angew. Geologie 17/2: 63-71. 

- Utelli H.-H., Kuster F., Pfammatter Ch. (2016). Integrales Naturgefahrenmanagement 

der BLS AG. Proceed. Conf. INTERPRAEVENT 2016. 

- Wiesmann A., Gruner U. (2011). Radar-Interferometrie im Einsatz für die Stabilitätsüberwachung 

von grossflächigen Felswänden. Swiss Bull. angew. Geologie 16/1: 51-55. 



Applied flood-risk-management in the Machland-Nord, 

Upper Austria 

Angewandtes Hochwasser-Risikomanagement im 

Machland-Nord, Oberösterreich 

Raimund Heidrich, DI 1 

ABSTRACT 

Due to previous flood-events 7 communities in the Machland Nord have found a flood- protection-association 

in 1993. The association ordered a flood-protection-study, which followed 

an at this time innovative and integral approach. The result was a flood-protection-concept 

combining resettlement, building of linear flood-protection-systems and object-protection 

measures. 

During the construction of the flood-protection-system extensive preventive measures on 

identification and reduction of flood-risk (including residual risk) were implemented for the 

first time. Simultaneously the elaboration of organizational measures for a better coping of 

floods considering the new flood-protection-system were dictated. During all processing 

phases an intense networking process of all responsible authorities over different administration 

levels with the operator of the flood-protection-system was reached. Additional 

important players were integrated regularly. According to an integrative approach information 

of the affected public was also considered. 

The core components of the flood-risk-management in Machland Nord were hazard analysis 

using numerical 2d-modelling, preparation of operating regulations for the operator, 

preparation of emergency plans for the authorities and the preparation of evacuation plans 

for the communities. 


Aufgrund vergangener Hochwasserereignisse haben sich 7 Gemeinden im Machland Nord im 

Jahr 1993 zum Hochwasserschutzverband Machland-Nord zusammengeschlossen und eine 

Hochwasserschutzstudie in Auftrag gegeben. Diese verfolgte einen zum damaligen Zeitpunkt 

innovativen in Österreich einzigartigen integralen Ansatz. Es wurde ein Hochwasserschutzkonzept 

entwickelt, das eine Kombination aus Absiedlungsmaßnahmen, linearen Hochwasserschutzanlagen 

und Objektschutzmaßnahmen beinhaltet. 

Im Zuge der Errichtung der Hochwasserschutzanlage wurden erstmals umfangreiche vorbeugende 

Maßnahmen hinsichtlich der Erkennung und Verringerung des Hochwasser-Risikos 

(inkl. Restrisikos) gesetzt. Gleichzeitig wurde unter Berücksichtigung der neuen Schutzanlage 

die Ausarbeitung organisatorischer Maßnahmen zur besseren Bewältigung von Hochwässern 

1 riocom - consulting engineers for water management and environmental engineering, Vienna AUSTRIA, 

raimund.heidrich@riocom.at 



vorgeschrieben. In allen Phasen der Bearbeitung wurde eine intensive Vernetzung der zuständigen 

Behörden über die unterschiedlichen Verwaltungsebenen hinweg mit dem Betreiber 

der Hochwasserschutzanlage erreicht, unter regelmäßiger Einbindung weiterer wichtiger 

Akteure. Im Sinne eines integralen Ansatzes wurde auch die Information der betroffenen 

Bevölkerung berücksichtigt. 

Die Kernelemente des Hochwasser-Risikomanagements im Machland-Nord sind die Gefahrenanalyse 

durch numerische 2d-Modellierungen, Erstellung von Betriebsvorschriften für 

den Betreiber der Schutzanlage, die Erstellung von Notfallplänen für die Behörden sowie die 

Erstellung von Evakuierungspläne für die Gemeinden. 

KEYWORDS 

2D model; hazard-analysis; action plan; contingency plan; evacuation plan 

EINFÜHRUNG 

Als Reaktion auf vergangene Hochwasserereignisse haben sich 7 Gemeinden im Machland 

Nord (Bezirk Perg, Oberösterreich) Mauthausen, Naarn, Mitterkirchen, Baumgartenberg, 

Saxen, Grein und St. Nikola im Jahr 1993 zum Hochwasserschutzverband Machland-Nord 

zusammengeschlossen, mit dem Ziel, weite Teile der Region gegen die immer wieder 

auftretenden Donauhochwässer zu schützen. (Schwingshandl A., Liehr C., Heidrich R. 

(2013)) 

Der Bezirk Perg liegt am linken Donauufer an der Grenze zu Niederösterreich (siehe 

Abbildung 1). Die geschützten Siedlungsbereiche von drei der sieben Gemeinden liegen 

direkt an der Donau, in den restlichen Gemeinden sind diese im Hinterland des linksufrigen 

Rückstaudammes des Donaukraftwerkes Wallsee-Mitterkirchen (siehe Abbildung 2) situiert. 

In einer Hochwasserschutzstudie und nachfolgenden Planungen wurde ein zum damaligen 

Zeitpunkt innovatives, in Österreich einzigartiges Hochwasserschutzkonzept entwickelt, das 

eine Kombination aus Absiedlungen, etwa 34km linearen Hochwasserschutzanlagen sowie 

36 Objektschutzmaßnahmen beinhaltet. Die aus diesen Maßnahmen bestehende Hochwasserschutzanlage 

„Machland-Nord“ wurde in den Jahren 2009 bis 2012 errichtet. Für den 

laufenden Betrieb wurde eine Betreibergesellschaft eingerichtet, die Machland-Damm 

Betriebs GmbH. (Machland-Damm Betriebs GmbH) 

Im Zuge der Errichtung der Hochwasserschutzanlage wurden erstmals umfangreiche 

vorbeugende Maßnahmen hinsichtlich der Erkennung und Verringerung des Hochwasser- 

Risikos (inkl. Restrisikos) in den Machlandgemeinden gesetzt. Gleichzeitig wurde in Abstimmung 

auf die neue Schutzanlage die Ausarbeitung organisatorischer Maßnahmen zur 

besseren Bewältigung von Hochwässern vorgeschrieben. In allen Phasen der Bearbeitung 

wurde dabei eine intensive Vernetzung der zuständigen Behörden (Gemeinden und Bezirkshauptmannschaft) 

über die unterschiedlichen Verwaltungsebenen hinweg mit dem Betreiber 


der Hochwasserschutzanlage und einer regelmäßigen Einbindung weiterer wichtiger Akteure 

erzielt. 

Die Kernelemente des Hochwasser-Risikomanagements im Machland-Nord sind: 

– Gefahrenanalyse durch numerische 2d-Modellierungen von Überlastereignissen 

– Erstellung von Betriebsvorschriften für den Betreiber der Schutzanlage 

– Erstellung von Notfallplänen für die Behörden (auf Gemeinde- und Bezirksebene) 

– Erstellung von Evakuierungsplänen für die Gemeinden 

Die Bearbeitungen wurden durch zahlreiche Workshops und Informationsveranstaltungen 

auf Gemeinde- und Bezirksebene begleitet, um die unterschiedlichen Zielgruppen über die 

Bearbeitung dieser Themen und die für sie wesentlichen Ergebnisse zu informieren. Moderne 

Hochwasserschutzanlagen haben Auswirkungen auf verschiedenen Ebenen. Sie verhindern 

hochwasserbedingte Schäden in Siedlungsgebieten, an Infrastruktur, Gewerbe und Landwirtschaft. 

Sie tragen wesentlich dazu bei, die Gefährdung von Personen zu reduzieren. Gleichzeitig 

entstehen jedoch neue Gefährdungsszenarien – vor allem in Versagensfällen oder aufgrund 

der für die Betroffenen unbekannten Situation. 

Im Sinne eines integralen Risikomanagements ist bei Errichtung von Hochwasserschutzanlagen 

das Thema Restrisiko (z.B. durch ein Versagen der Schutzanlage) zu behandeln. 

In solche Risikoanalysen sind sämtliche handelnden Akteure einzubinden, um deren 

Fachwissen einzubringen und deren Akzeptanz der Ergebnisse und Produkte zu gewährleisten. 

Dann werden Betriebsvorschriften, Notfallpläne und Evakuierungspläne im Hochwassereinsatz 

auch verwendet. 

DIE HOCHWASSERSCHUTZANLAGE “MACHLAND-NORD” 

Die Hochwasserschutzanlage “Machland-Nord” besteht aus etwa 30,1km Erddämmen, 3,9km 

mobilen Hochwasserschutzelementen und einigen hundert Meter Hochwasserschutzmauern. 

Diese und die 24 Trafostationen, 76 Pumpwerke und 28 Absperrbauwerke bzw. Schieberschächte 

machen den Machlanddamm, so der in der Bevölkerung gebräuchliche Name für 

das Schutzbauwerk, zu einer der komplexesten und größten Hochwasserschutzanlagen in 

Österreich. (Machland-Damm Betriebs GmbH) 

Entlang der Hochwasserschutzanlage wurden insgesamt 13 Pegelmessstellen zur Beobachtung 

der Wasserspiegellagen errichtet. Die Messung erfolgt mittels redundant ausgeführter 

Drucksonden. Die Beobachtungen werden an die Leittechnikzentrale fernübertragen. 

Die Steuerung der Pumpstationen und Verschlussobjekte erfolgt autark an den einzelnen 

Objekten, die Überwachung ist zentral über die Leittechnikzentrale des Betreibers möglich. 

Der umfangreiche Einsatz von mobilen Hochwasserschutzsystemen stellt für alle Beteiligten 

eine große Herausforderung vor allem beim Aufbau und der Überwachung dieser Anlagenteile 

dar. Beispielhaft seien hier der etwa 1,6km lange mobile Hochwasserschutz der Gemeinde 

Mauthausen sowie der 750m lange Mobilschutz der Gemeinde Grein erwähnt. Ersterer weist 


eine maximale Höhe von 2,6m auf. In drei Aufbauphasen werden 218, 189 und 132 

Steher aufgestellt, zwischen denen 658, 2050 und 2698 Dammbalken einzulegen sind. 

In der Gemeinde Grein beträgt die maximale Höhe der Mobilschutzwand entlang der 

Donaulände sogar 4,4m. Diese wird in 5 Aufbauphasen von bis zu 4 parallel arbeitenden 

Trupps errichtet. Am aufbauintensivsten erweist sich hierbei die Phase 2 mit rund 

200 Stützen und gut 700 Dammbalken. 

GEFAHRENANALYSE MITTELS NUMERISCHER 2D-MODELLIERUNGEN 

Aufgrund unterschiedlicher Exposition der Gemeinden im Machland-Nord unterscheidet sich 

die Überflutungssituation bei Donauhochwasser in den 7 Gemeinden. Historisches Wissen 

über den Ablauf und das Auftreten von hochwasserbedingten Überflutungen wurde durch 

die Errichtung der Hochwasserschutzanlage „Machland-Nord“ obsolet. Daher war es notwendig, 

das Überflutungsgeschehen für den Bemessungsfall sowie den Verlauf von Überflutungen 

bei Überlastereignissen mittels numerischer hydrodynamischer 2d-Modellierungen 

abzubilden. Die untersuchten Überlastfälle umfassen die Überströmung der Schutzanlage 

aufgrund eines Ereignisses größer dem Bemessungsfall, Dammbrüche, Versagen mobiler 

Hochwasserschutzanlagen sowie Systemversagen (hier im speziellen den Ausfall von Pumpstationen). 

Je nach untersuchtem Szenario erfolgten die Modellierungen stationär oder 

instationär. Die Welle des Hochwassers vom August 2002 diente als Vorlage für die instationären 

Simulationen. 

Die Modellierungsergebnisse wurden für unterschiedliche Zwecke ausgewertet. Der Betreiber 

der Hochwasserschutzanlage benötigte vor allem Informationen über den Anstieg der 

Wasserspiegel zu Beginn des Hochwassers, zur Ausarbeitung von Zeit-Phasen-Plänen für den 

gestaffelten Aufbau der mobilen Hochwasserschutzanlagen. Für die behördlichen Einsatzleitungen 

der Gemeinden und des Bezirks waren die potentiellen Überflutungen, deren 

zeitlicher Verlauf sowie die zu erwartenden Wasserstände im geschützten Hinterland in 

Überlastfällen von Bedeutung, um eventuelle Evakuierungen bestmöglich vorbereiten zu 

können. 

Als Ergebnis der numerischen 2d-Modellierungen wurden Gefahrenkarten erstellt, die das 

potentielle Überflutungsgebiet und die zu erwartenden Wassertiefen sowie das Ausbreitungsverhalten 

zeigen. Die erstellten Planunterlagen unterstützen die Entscheidungsträger bei 

der Ermittlung potentieller Gefahren infolge von Versagensszenarien oder Überlastfällen, 

sie ermöglichen die Identifizierung betroffener Siedlungsgebiete und können dadurch als 

Planungsgrundlage für die Steuerung einer gezielten Information Betroffener dienen. 

BETRIEBSVORSCHRIFTEN FÜR DIE HOCHWASSERSCHUTZANLAGE 

Die Machland-Damm Betriebs GmbH hat als Betreiber der Hochwasserschutzanlage „Machland-Nord“ 

den behördlichen Auftrag, die Schutzanlage jederzeit in einem voll funktions- 


tüchtigen Zustand zu erhalten. Als Hilfestellung für den Betreiber wurden Betriebsvorschriften 

erstellt. Diese umfassen Betriebspläne für den Trockenwetterbetrieb, den Hochwasser - 

betrieb sowie den Überlastfall. 

Zur Erfüllung des behördlichen Auftrags ist es notwendig, regelmäßige Wartungs- und In - 

standhaltungsmaßnahmen im Trockenwetterbetrieb durchzuführen. Dazu geben die 

Betriebsvorschriften Art, Umfang und Intervall dieser Maßnahmen an. Beigelegte Formblätter 

dienen der lückenlosen Dokumentation der Arbeiten. Im Hochwasserbetrieb sind Kontrollund 

Überwachungsmaßnahmen zu setzen. Es sind Wasserstände und Freibordwerte an 

Lattenpegeln abzulesen. Entdeckte Schäden an der Schutzanlage sind umgehend zu melden. 

Für diese Tätigkeiten liegen den Betriebsvorschriften Formblätter zur Dokumentation und 

Meldung bei. Die Schutzanlage ist auch in Überlastfällen bestmöglich zu beobachten und – 

in Zusammenarbeit mit den Katastrophenschutzbehörden und Einsatzorganisationen - zu 

verteidigen. 

NOTFALLPLÄNE DONAUHOCHWASSER DER BEHÖRDEN 

Trotz der Errichtung der Hochwasserschutzanlage „Machland-Nord“ haben die behördlichen 

und technischen Einsatzleitungen der 7 Gemeinden im Machland Nord sowie des Bezirkes 

Perg im Hochwasserfall zahlreiche Maßnahmen zur Vorbereitung und Bewältigung eines 

Donauhochwassers in ihrem Zuständigkeitsbereich zu treffen. Eine Kernaufgabe dabei ist 

die Information der Betroffenen. Die Errichtung der Hochwasserschutzanlage hat für die 

zuständigen Katastrophenschutzbehörden sogar zu zusätzlichen Aufgaben geführt. Neben 

den Notfallplänen für den Hochwasserfall wurden auch Notfallpläne für die vorab analysierten 

Überlastfälle erstellt. Dabei werden zusätzlich Evakuierungsmaßnahmen berücksichtigt. 

Die Notfallpläne enthalten dazu Maßnahmen der Warnung / Alarmierung, Information und 

Verständigung sowie der Koordination mit den wesentlichen Akteuren. 

Die Notfallpläne Donauhochwasser beinhalten eine Sammlung wesentlicher Tätigkeiten, 

zusammengestellt in einem Maßnahmenkatalog und ergänzt um Maßnahmenblätter. Die 

Maßnahmen werden dabei konkreten Kriteriums-Wasserständen zugeordnet, bei deren 

Erreichen sie erstmals bzw. einmalig gesetzt werden. Der Maßnahmenkatalog stellt dadurch 

eine Liste zu setzender Maßnahmen und Tätigkeiten dar, die mit steigendem Wasserstand 

nach und nach abzuarbeiten sind. Ziel der Notfallpläne Donauhochwasser ist es, den 

handelnden Akteuren in den zuständigen Katastrophenschutzbehörden Leitlinien für deren 

Handeln im Hochwasserfall zu geben. Der Maßnahmenkatalog ist daher nicht als starre 

Vorgabe zu sehen, da Hochwässer stark unterschiedlich ablaufen können, was unterschiedliche 

Vorgehensweisen bezüglich der Vorbereitung, Bewältigung und Verteidigung notwendig 

macht. Die Maßnahmenblätter enthalten detaillierte Informationen zur Durchführung der 

geplanten Tätigkeiten, Lageskizzen und Verweise auf die jeweils gültigen Gesetzestexte. 

Die Kriteriums-Wasserstände beziehen sich primär auf lokale Pegel, die von den Gemeinden 

selbst beobachtet werden können bzw. über deren Pegelstände sie in regelmäßigen Abstän- 


den informiert werden. Ist dies nicht möglich kann auf allgemein zugängliche Beobachtungsdaten 

von 2 Pegelstationen des Landes Oberösterreich zurückgegriffen werden. Für diese 

stehen im Hochwasserfall zusätzlich Prognosewerte der Wasserstände zur Verfügung. Die 

Verwendung lokaler Pegel hat den Vorteil, dass Abweichungen infolge unsicherer Pegelrelationen 

ausgeschaltet werden. Dies ist vor allem in jenen Gemeinden wichtig, die im Hinterland 

liegen und für die eine direkte Umlegung eines Donauwasserstandes auf lokale Wasserstände 

meist nur ungenau möglich ist. 

EVAKUIERUNGSPLÄNE DER GEMEINDEN 

Die Evakuierungspläne der 7 Gemeinden im Machland Nord schließen thematisch direkt an 

die Notfallpläne der Gemeinden an. Wird gemäß Notfallplan Donauhochwasser eine 

Evakuierung eingeleitet, können die Behörden auf die Inhalte und Produkte der Evakuierungspläne 

als Informationsquelle zurückgreifen. Diese zeigen in Textdokumenten, Datenblättern 

und Lageplänen wer oder was, wann, von wem und womit, wohin evakuiert 

wird und von wem er/sie/es dort versorgt wird. Die Unterlagen sind dabei so gestaltet, dass 

sie leicht ins Feld mitgenommen werden können und beispielsweise ortsfremden Einsatzkräften 

als Arbeitsgrundlage und Informationsquelle dienen können. 

Ein wesentlicher Schritt im Zuge der Erstellung der Evakuierungspläne ist die Abgrenzung 

von Evakuierungszonen. Das sind Bereiche die hinsichtlich des Gefahrenpotentials als 

vergleichbar angesehen werden können. Für diese Evakuierungszonen wurde die Anzahl 

betroffener Personen und zu evakuierender Nutztiere zusammengestellt. Weiters wurden 

Evakuierungsrouten und Sammelpunkte für Personen ohne eigenem Fahrzeug definiert. 

Ergänzt werden die Evakuierungspläne durch Analysen und Erhebungen zu den Themen 

Transportlogistik (z.B.: wo kann ich Transportfahrzeuge anmieten) und Unterkunftsmanagement 

(Kriterien für Notunterkünfte; geeignete Standorte). 

FAZIT – ERFAHRUNGEN IM HOCHWASSER VON JUNI 2013 

Eine effektive Maßnahme zur Verringerung von Hochwasser-Risiko ist die Verringerung des 

Schadenspotentials, beispielsweise durch Absiedlung in hochwassersichere Bereiche. Dieser 

Ansatz wurde in der Planungsphase des Hochwasserschutzes im Machland Nord als Teil eines 

integrierten Hochwassermanagements in Österreich erstmals großflächig angewandt. Dabei 

wurden über 230 Wohnobjekte abgesiedelt. Die Bewohner erhielten eine entsprechende 

Ablöse für ihr Wohnobjekt und wurden von Seite der Gemeinden bei der Suche nach 

Ersatzstandorten unterstützt. 

Das Hochwasser von Juni 2013, das das Bemessungsziel der Hochwasserschutzanlage erreicht 

bzw. mancherorts leicht übertraf hat gezeigt, dass die frühzeitige Einbindung der zuständigen 

Akteure des Hochwasser- und Katastrophenschutzes dazu führt, Akzeptanz und einen hohen 

Informationsgrad die erstellten Unterlagen und Operate betreffend zu schaffen. 


Abbildung 1: Lage des Projektgebietes in Ost-Österreich 

Die Rollenverteilung der handelnden Akteure im Machland Nord ist durchaus komplex. Als 

Beispiel sind die örtlichen Feuerwehren zu nennen. Diese erhalten Anfragen und Aufträge 

von mehreren Stellen. Sie errichten den mobilen Hochwasserschutz und führen Dammwachen 

im Auftrag des Dammbetreibers durch, gleichzeitig setzen sie Verteidigungsmaßnahmen 

oder führen Evakuierungen im Auftrag der behördlichen Einsatzleitung durch. Umso 

wichtiger ist es gut koordiniert vorzugehen, um potentiellen Schwachstellen wie Personalengpässen 

frühzeitig entgegen zu wirken. 

Die notwendige enge Kooperation und Koordination zwischen der Machland-Damm Betriebs 

GmbH als Betreiber der Hochwasserschutzanlage, den behördlichen Einsatzleitungen auf 

Bezirks- und Gemeindeebene, der technischen Einsatzleitungen auf Bezirks- und Gemeindeebene, 

dem Bezirkspolizeikommando und den Polizeiinspektionen, dem Rotem Kreuz sowie 


Abbildung 2: Übersichtslagedarstellung der Hochwasserschutzanlage „Machland-Nord“ 

den Straßenmeistereien hat während des Hochwassers von Juni 2013 vorbildlich und effektiv 

funktioniert. Die stets verfügbare Fachberatung und offene Kommunikation durch die vom 

Betreiber eingesetzten externen Experten hatten großen Mehrwert für die behördlichen 

Einsatzleitungen, vor allem zu fortgeschrittenem Zeitpunkt als der Überlastfall drohte. 

(Schwingshandl A., Liehr C., Heidrich R. (2013)) 

Dank der Umsetzung der Maßnahmen des Trockenwetterbetriebs und Hochwasserbetriebs 

der Betriebsvorschriften wurden die Vorbereitungs- und Kontrollmaßnahmen des Betreibers 

zeitgerecht und korrekt umgesetzt. Dadurch bedingt waren die mobilen Hochwasserschutzanlagen 

rechtzeitig einsatzbereit, die erforderlichen Straßensperren und Umleitungsstrecken 

ordentlich eingerichtet sowie die elektromaschinellen Anlagenteile wie Pumpstationen, 

Schieberbauwerke und Energieversorgungspunkte betriebsbereit. 

Die vom Betreiber organisierten Dammwachen (Durchführung durch die örtlichen Feuerwehren) 

leisteten bei teilweise widrigen Wetterbedingungen ausgezeichnete Arbeit. Die 

aufmerksamen Dammwachen hatten durch ihre raschen Meldungen erheblichen Anteil 

daran, dass die aufgetretenen Problemstellen an den Erddämmen erfolgreich verteidigt 

werden konnten. (Schwingshandl A., Liehr C., Heidrich R. (2013)) Gleichzeitig standen die 

örtlichen Feuerwehren der betroffenen Bevölkerung im Zuge des Katastrophenhilfsdienstes 

im Auftrag der behördlichen Einsatzleitung zur Seite. Ein Hauptgrund für die ausgezeichnete 

Arbeit der örtlichen Feuerwehren waren die gute Schulung der Einsatzkräfte und die 

Teilnahme an den regelmäßig durchgeführten Montage- und Alarmübungen durch den 

Betreiber im Trockenwetterfall. 


Während des Hochwassers von Juni 2013 wurden einzelne Siedlungsgebiete aufgrund hoher 

Beaufschlagung der Schutzanlage oder beobachteter Schadstellen an den Erddämmen 

vorsorglich evakuiert. Die dabei gewonnenen Erfahrungen wurden bei der Erstellung der 

Evakuierungspläne berücksichtigt. 

Hochwasser-Risikomanagement ist ein andauernder Prozess mit zahlreichen Beteiligten. 

Neben der Schaffung von Fachgrundlagen ist es wichtig eine gute Vernetzung der handelnden 

Akteure und Beteiligten in diesem Prozess zu schaffen. Der Wissensstand um die 

Fachgrundlagen, Abläufe und erforderlichen Maßnahmen ist speziell in längeren Phasen 

ohne Hochwasser durch regelmäßige Stabs-, Alarm- und Einsatzübungen hoch zu halten. 

Dies erleichtert es auch neuen Akteuren z.B. nach Personalwechsel leicht in das bestehende 

System einzusteigen. 

Eine wesentliche Aufgabe im Hochwasser-Risikomanagement besteht in der Information der 

Betroffenen. Je höher deren Wissenstand über die eigene Gefährdung, mögliche Vorsorge-, 

Verteidigungs- und Evakuierungsmaßnahmen, desto eher werden diese Gruppen im 

Ereignisfall Verständnis für behördliche Anordnungen aufbringen und entsprechend positiv 

darauf reagieren. Vor Errichtung der Hochwasserschutzanlage „Machland-Nord“ waren große 

Gebiete des nun geschützten Raumes auch bei kleineren häufigeren Hochwässern von 

Überflutungen betroffen. In diesen Gebieten hat die Bevölkerung in den letzten Jahrzehnten 

gelernt mit dem Hochwasser zu leben und entsprechend darauf zu reagieren. Diese Situationen 

fallen zukünftig weg. Hochwasser wird nur mehr selten zu Überflutungen hinter der 

Schutzanlage führen. Vor allem ortsfremde, neu zugezogene Personen können die Gefährdung 

durch Hochwasser aufgrund der fehlenden kleineren häufigeren Überflutungen nicht 

erlernen. Daher ist es von besonderer Wichtigkeit, dass die zuständigen Katastrophenschutzbehörden 

hier in regelmäßigen Abständen informieren. 

Gut geschulte Entscheidungsträger und eine gut informierte Bevölkerung erleichtern die 

Arbeit im Hochwasserfall für alle Beteiligten und erhöhen die Chance Leben zu schützen und 

Schäden zu verringern. 

LITERATUR 

- Schwingshandl A., Liehr C., Heidrich R. (2013). „Hochwasserschutz Machland-Nord – Bewährungsprobe 

im Hochwasser Juni 2013“. Österreichische Wasser- und Abfallwirtschaft 

7-8/13: 273-279. 

- Machland-Damm Betriebs GmbH; www.machlanddamm.at 



Flood forecasting system for the Tyrolean Inn River 

(Austria): current state and furtherenhancements of 

a modular forecasting system for alpine catchments 

Matthias Huttenlau, Dr. 1 ; Johannes Bellinger 1 ; Paul Schattan 1 ; Kristian Förster 1 ; Felix Oesterle 1 ; Katrin Schneider 1 ; 

Stefan Achleitner 2 ; Johannes Schöber 3 ; Georg Raffeiner 4 ; Robert Kirnbauer 5 

ABSTRACT 

Several severe flood events in Central Europe show the vulnerability of society living along 

water courses. On the one side, operational flood forecasting systems contribute to reduce 

vulnerability and increase resilience in flood risk management. On the other side, a runoff 

forecasting system used by energy providers also has an economic dimension in terms of their 

day-to-day business providing advanced planning capabilities used in the hydro-power plant 

operation. Combining these two aspects, the best way for its realization is a coordinated 

development, which is able to (i) fulfill its responsibility in the context of flood risk management 

and (ii) support the advanced hydro-power plant management of the energy provider. 

In this context, the following paper presents the current state and further enhancements of 

the flood forecasting system of the Tyrolean Inn River. Especially in its alpine catchments the 

operation of the forecasting system is challenging. Short times of runoff concentration times 

paired with a terrain of complex processes for runoff generation and flood wave propagation. 

KEYWORDS 

operational flood forecasting; hybrid hydrological/hydraulic system; Inn River; predictive 

accuracy; alpine hydrology 

INTRODUCTION 

The Austrian Province of Tyrol is a typical mountain region, where process dynamics are 

generally higher than in the low lands due to strong altitudinal gradients, which are relevant 

for flood forecasting at small spatial and temporal scales. Only a small proportion of the 

total area, which is concentrated in the valley floors in close vicinity to streams, is available 

for permanent human settlement. Tyrol is part of the Eastern Alps and covers an area of 

12,640 km 2 with an elevation range from 462 m up to 3,798 m a. s. l. Due to topographically 

controlled limits, only 12.8 % of the area is available for permanent settlement of which only 

0.4 % is declared as residential area populated by approximately 700,000 inhabitants (Amt 

der Tiroler Landesregierung, 2014). The circumstance that a very high proportion of the 

1 alpS Centre for Climate Change Adaptation, Innsbruck, AUSTRIA, huttenlau@alps-gmbh.com 

2 Unit of Hydraulic Engineering, University of Innsbruck (Austria) 

3 TIWAG - Tiroler Wasserkraft AG, Innsbruck (Austria) 

4 Hydrographic Service, Regional Government of Tyrol, Innsbruck (Austria) 

5 Institute for Hydraulic and Water Resources Engineering, Vienna University of Technology, (Austria) 



esidential areas are located in valley floors leads consequently to a high exposure to flooding. 

In Tyrol, 105,330 people and a cumulative asset exposure of buildings and their contents 

(e.g. residential, public, trade and industry) of approximately EUR 7.6 billion are exposed 

within floodplains with a recurrence interval of 300 years (Huttenlau et al. 2010). 

The vulnerability of society living along watercourses in general was clearly shown by the 

severe European flood events in the years 1999, 2002, 2005 and 2013. Currently no reliable 

information about changes in the characteristics or the return periods of future flood events 

in Austria is available, especially under climate change conditions (Nachtnebel et al., 2014). 

Irrespective of different future damage potentials and damage symptoms, a reliable increase 

of the flood damages, mainly caused by the population growth and socioeconomic development, 

can be expected (König et al., 2014). Flood forecasting as a central component of early 

warning and decision support systems is an efficient instrument to minimize false alarms and 

enables effective emergency response for decision making. Beside other preliminary measures 

in the framework of integrated flood risk management (IFRM) like flood protection and land 

use planning, emergency management is one of the major pillars to reduce flood risk and 

a matter of public policy. Energy suppliers also apply runoff forecast systems to advance 

decision making in their day-to-day business and to support management decisions with 

respect to flood control through hydropower plants (often a requirement of official decisions). 

Combining these two aspects, the best way for its realization is a coordinated development, 

which is able to fulfill its responsibility in the context of flood risk management and also 

support the advanced hydro-power plant management of energy suppliers. Such a development 

of an operational forecasting system for the Tyrolean part of the Inn River was realized 

in a joint public-private research initiative of the affiliated institutions. In detail, the development 

of the system was mainly located at the research center “alpS – Centre for Climate 

Change Adaptation”, which participates on the COMET-program of the Austrian Research 

Promotion Agency (FFG). The program builds a platform for jointly defined research 

programs in the field of science-industry cooperation. Besides the partitioning of the costs the 

close cooperation includes furthermore regular communications about practical experiences 

as well as the latest scientific knowledge to fulfil the requirements. 

Especially the operationalization and advancement of such forecasting systems, with respect 

to complexity of runoff generation and flood wave propagation, is challenging. This is 

especially true considering alpine catchments, which are characterized by very short times of 

runoff concentration, which in turn limits the response time. This contribution presents the 

current state of the operational forecasting system HoPI (“Hochwasserprognose für den 

Tiroler Inn”), describes some operational applications, and shows ongoing research activities. 

HISTORY AND SETUP OF THE FLOOD FORECASTING SYSTEM 

Following the flood event in 2002, the development of the presented flood forecasting system 

started with the prototype of the modular-based, hybrid hydrological/hydraulic system 

(Senfter et al., 2009) from 2003 to 2006. Unfortunately, the system was not yet available for 

the operational flood forecasting service at the Hydrographic Office of the Province of Tyrol, 


when a Vb (5b) track cyclone struck against the north side of the Alps in August 2005. 

This heavy precipitation event led to massive floods within the Alpine Ridge. In the Paznaun/ 

Stanzer valley and the district Außerfern (see figure 1), floods were measured with return 

periods greater than 100 years. At the gauge Innsbruck, a return period of 200 years was 

observed (Godina, 2006). A verifiable reduction of the maximum instantaneous discharge of 

about 15 cm was attributable to the retention in several water reservoirs of the TIWAG and 

also to inter-basin diversion from the catchments of the Paznaun/Stanzer valley into water 

reservoirs of the Vorarlberger Illwerke. Thus, hydro-power plant operation helped to prevent 

the imminent overbank flooding in the city of Innsbruck (Schönlaub und Hofer 2009). 

Figure 1: Spatial extent of the HoPI system including the hydrologic/hydraulic model components. The models for the Inn headwaters in 

Switzerland are currently under preparation. 

After the successful integration into the operational flood forecasting service at the Hydrographic 

Office of the Province of Tyrol the developments of the subsequent project phase 

(2006-2009) were mainly characterized by further adaptations to requirements for practical 

use. This improvement ensured the reliability of the whole system and formed the basis for 

the current model optimization as well as the identification and quantification of system 

uncertainties. Achleitner et al. (2012) analyze the operational performance of the hydrological 

models and result in a strong dependency to the meteorological observations and 


meteorological forecasts, respectively. Furthermore they propose the implementation of 

model state updating routines for further improving of the forecasting quality. The current 

operational forecasting system is set up as a permanently operating tool, running at hourly 

time steps, for the 200 km-long segment of the Austrian part of the River Inn between the 

Swiss-Austrian border and the Austrian-German border. The river catchment at the Austrian-German 

border covers an area of 9,700 km 2 in total. Forecasted discharges at the 

Austrian-German border are provided as input to the forecast model for Bavaria (Germany). 

The forecasting system HoPI comprises different types of deterministic models and data 

management software components. Its current implementation has a modular hybrid 

hydrological / hydraulic character (Achleitner et al., 2009) and is structured in (i) the data 

management and preprocessing, (ii) the hydrological simulation of tributary catchments and 

(iii) the hydraulic simulation of the Inn River. The procedure starts with the import of 

observed data since the last system run and the prognosis data for the upcoming forecast. 

Measured meteorological station data is interpolated and processed as input data for the 

models. For the forecasting period, the observed time series are extended with meteorological 

forecast data. Currently, the runoff generation from 10 glaciated sub-catchments, with an 

overall area of 620 km², is modelled with the fully distributed water-balance model SES 

(Snow and Icemelt Model) (Asztalos et al., 2007, Schöber et al., 2014). The accumulation and 

melt processes of snow, firn and ice are considered on basis of an energy-balance-equation on 

a 50 m grid. The simulated outflow of snow, firn, glacier ice, rock and subsurface flow to the 

catchments outlet is transformed via five Nash-Cascades. To run this model, data of precipitation, 

temperature, humidity, wind speed, and global radiation are required. Subsequently, the 

runoff modelled by SES is used as additional boundary conditions for the downstream 

hydrological model. By contrast, runoff generation from non-glaciated sub-catchments is 

modelled with the semi-distributed water balance model HQsim (Kleindienst, 1996) based on 

the HRU concept, which aggregate areas with similar soil type, aspect and elevation. At each 

HRU, processes of snow accumulation and melt (day-degree approach), evapotranspiration, 

interception, infiltration into the unsaturated zone as well as the saturated zone are included. 

Thereby the runoff is partitioned into surface flow, interflow and base flow. After calculation 

of runoff at each HRU, it is concentrated at the nearest point of the channel network and 

routed downstream by a non-linear storage cascade towards the catchment’s outlet. In total, 

50 tributary catchments, including 13 gauged catchments, are independently simulated. 

As before, the modelled runoff provides the boundary conditions for the commercial 

1D-hydraulic model FluxDSS/DESIGNER/FLORIS2000, which represents the Tyrolean Inn 

segment from the gauge Martinsbruck at the Swiss border to the gauge Kufstein at the 

German border (fig. 1). 

APPLICATION 

The briefly described hybrid model has a temporal forecast period up to 48 h and is driven by 

hydro-meteorological parameters of the Integrated Nowcasting through Comprehensive 

Analysis (INCA) System provided by the Austrian Meteorological Office (ZAMG) on hourly 


asis. At present, the models are initialized using online transmitted meteorological and 

hydrological data prior to the forecast, for which INCA provides all needed boundary 

conditions. Currently, HoPI is run every hour at the Hydrographic Office of the Province of 

Tyrol mainly for flood forecasts and also at TIWAG where it is used as well as decision support 

system for water management tasks. An additional instance is operated once a day at the 

research centre alpS, but to ensure the system safety the results are shared among the three 

stakeholders. However, further actions require decisions by a hydrologist on duty. For 

example, supported by HoPI predictions, the water level in the reservoirs of storage power 

plants can be lowered as early as possible by increased power generation, if an extreme event 

is expected to occur. In the case of a pumped-storage power station, water can be pumped 

from lower to upper stage reservoirs (Hofer et. al, 2013). In case of emergency the collaboration 

of relevant government departments (e. g. Tyrolean Regional Hazard Warning Centre, 

Tyrolean Fire Service Association) is legally predefined. The Hydrographic Office of the 

Province Tyrol informs, based on the runoff predictions of HoPI as well as expert assessments 

of the meteorological forecast, about the current flood risk and updates the situation by 

current measurements and predictions within the event. The hydro power plant management 

remains by the operators, who usually adapt the operating schedule to relieve the situation. 

If a danger to life and limb is foreseeable, the government agency could give eventual 

instructions to protect the general public interest. Since HoPI was implemented into the 

operational flood forecasting, no major flood event occurred at the River Inn in Tyrol. Within 

the extreme flooding in Central Europe in June 2013, an occluded front system reached the 

north side of the Alps (Hydrographischer Dienst Tirol, 2013). The heavy rainfall passed the 

HoPI catchments and hit the adjacent catchments. Nevertheless, floods with a return period 

of around 20/25 years were observed at the Inn tributary catchments “Brandenberger Ache” 

and “Brixentaler Ache”. Figure 2 (a) features the hydrograph of the gauge “Bruckhäusl” at 

the Brixental tributary. In May 2013 the runoffs were slightly overestimated using the 

measured data, but with the onset of the rainfall the model simulated the flood event very 

well. Solely the simulated peak runoff (173 m³/s) underestimates the observed value (190 

m³/s) by around 9 %. Figure 2 (c) shows the HoPI-prognoses obtained several hours before 

the event. Due to their strong relation to the meteorological forecasts, the prognoses 

estimated the observed runoff not earlier than 18 hours before the event. Besides its initially 

defined target the flood forecasting, the HoPI system is currently used for several water 

management tasks, such as the prognosis of appropriate runoffs for flushing sediments from 

river power plants. Figure 2(b) depicts the HoPI-simulation of such a situation: within the 

second and third week in July 2014 a cyclone lead to heavy thunderstorms in the north of 

Tyrol with an increasing runoff exceeding the flood mark of the yearly return period at the 

gauge “Bruckhäusl”. Due to satisfactory hydrological forecasts of the HoPI system (Fig. 2 (d)), 

this event was used to flush the local hydro power plant by the TIWAG. 


Figure 2: The figures (a) and (b) show observed and simulated hydrographs (with observed forcing data) of two exemplarily flood events 

at the “Bruckhäusl” gauge (Brixentaler Ache). Hydrographs in figures (c) and (d) show the accuracy of the HoPI-prognoses obtained 

several hours before the event. 

CONCLUSION AND FURTHER ADVANCEMENTS 

Overall, the presented forecasting system is a reliable and powerful tool for flood forecasting. 

Since nature is a very complex system, there will always be poorly predictable events, even if 

the presented applications deliver satisfactory results. To further increase the reliability and 

flexible use of the HoPI system for future tasks within the water resources management of 

Tyrol, different enhancements are planned such as: 

– An alternative online coupling of both applied hydrological models to benefit from their 

specific advantages: the power of the SES model is doubtless the snow simulation of 


unforested areas, whereas a more sophisticated concept of soil processes and runoff routing 

is implemented in HQsim. 

– Since the accuracy of the simulated runoff is mainly related to the input forcing data, the 

reliability of the predicted runoff up to 48 h depends as well as on the applied INCA data. 

Ensemble datasets for INCA, which became available recently, are to be used besides the 

main INCA run to introduce an uncertainty bandwidth in the forecast. 

– To provide an additionally medium-range forecast (up to 10 days), the current forecast 

period will be extended using statistically corrected Global Forecast System (GFS) data. 

– In alpine headwater catchments, both total solid precipitation and snow redistribution are 

major sources of uncertainty. Terrestrial laser scan campaigns and new methods for 

measuring snow water equivalent (SWE) at different spatial scales are applied to further 

develop snow-hydrological modelling. 

– To compensate for poor simulation of snow melt in spring periods (resulting from incorrect 

snow accumulation during winter), data assimilation from different snow data sets is 

planned. Remote sensing data from optical sensors (e.g. MODIS, Sentinell-2) will be used 

to provide information of the snow-covered area in daily or near daily temporal resolution. 

Synthetic Aperture Radar (SAR) based products will be used to assimilate areas where 

snow is melting. Despite few stations measuring SWE directly, an overall number of 100 

snow gauges, distributed over the whole tributary catchments, will be used as additional 

data to estimate SWE from snow depth (Schöber 2015), serving as an additional calibration 

dataset besides runoff and snow cover. 

LITERATURE 

- Achleitner, S., Rinderer, M., and Kirnbauer, R. (2009): Hydrological modeling in alpine 

catchments: sensing the critical parameters towards an efficient model calibration, in: Water 

Sci. Technol. 60(6), S. 1507–1514. 

- Achleitner, S., Schöber, J., Rinderer, M.,Leonhardt, G., Schöberl, F., Kirnbauer, R., and 

Schönlaub, H. (2012): Analyzing the operational performance of the hydrological models in 

an alpine flood forecasting system, in: J. Hydrol., 412-413, S. 90–100. 

- Amt der Tiroler Landesregierung (2014): Statistisches Handbuch Bundesland Tirol 2014, 

Innsbruck. 

- Asztalos, J., Kirnbauer, R., Escher-Vetter, H., and Braun, L. (2007): A distributed energy 

balance snow and glacier melt model as a component of a flood forecasting system for the Inn 

River, in: Proceedings of the Alpine Snow Workshop, 5-6. Oktober 2006, Berchtesgaden 

National Park research 53, S. 9–17. 

- Godina, R., Lalk, P., Lorenz, P., Müller, G., and Weilguni, V. (2006): Hochwasser 2005 – 

Ereignisdokumentation, in: Teilbericht des Hydrographischen Dienstes, Bundesministerium 

für Land- und Forstwirtschaft, Umwelt und Wasserwirtschaft, Wien. 

- Hofer, B., Schöber, J. and Perzlmaier, S. (2013): Flood control – Principles for the operation 

of existing and the planning of new storage power plants, in: International Conference and 


Exhibition (Hydro 2013). Promoting the Versatile Role of Hydro. Sutton: Aqua-Media 

International, 18.06. Innsbruck 

- Huttenlau, M., Stötter, J., and Stiefelmeyer, H. (2010): Risk-based damage potential and loss 

estimation of extreme flooding scenarios in the Austrian Federal Province of Tyrol, in: Nat. 

Hazards and Earth Syst. Sci. 10(12), S. 2451–2473. 

- Hydrographischer Dienst Tirol (2013): Hochwasserereignis 31.5. bis 2.6.2013 im Tiroler 

Unterland. Hydrologischer Kurzbericht. URL https://www.tirol.gv.at/fileadmin/themen/ 

umwelt/wasserkreislauf/downloads/Kurzbericht_Hochwasser_Juni_2013.pdf. 

Kleindienst (1996): H. Erweiterung und Erprobung eines anwendungsorientierten hydrologischen 

Modells zur Gangliniensimulation in kleinen Wildbacheinzugsgebieten. Master 

Thesis, Ludwig Maximilians Universität München. 

- König, M., Loibl,W., Steiger, R., Aspöck, H., Bednar-Friedl, B., Brunner, K. M., Höferl, K. 

M., Huttenlau, M., Walochnik, J., and Weisz, U. (2014): Der Einfluss des Klimawandels auf 

die Anthroposphäre, in: Österreichischer Sachstandsbericht Klimawandel 2014, APCC, S. 

641–704. Verlag der Österreichischen Akademie der Wissenschaften, Wien, Österreich. 

- Nachtnebel, H. P., Dokulil, M., Kuhn, M., Loiskandl, W., Sailer, R., and Schöner, W. (2014): 

Der Einfluss des Klimawandels auf die Hydrosphäre, in: Österreichischer Sachstandsbericht 

Klimawandel 2014, APCC, S. 411–466. Verlag der Österreichischen Akademie der Wissenschaften, 

Wien, Österreich. ISBN 978-3-7001-7699-2. 

- Senfter, S., Leonhardt, G., Oberparleiter, C., Asztalos, J., Kirnbauer, R., Schöberl, F., and 

Schönlaub, H. (2009): Flood Forecasting for the River Inn, in: Veulliet, E., Stötter, J., and 

Weck-Hannemann, H.: Sustainable Natural Hazard Management in Alpine Environments, S. 

35–67, Springer-Verlag, Berlin Heidelberg. 

- Schöber, J., Schneider, K., Helfricht, K., Schattan, P., Achleitner, S., Schöberl, F., Kirnbauer, 

R. (2014): Snow cover characteristics in a glacierized catchment in the Tyrolean Alps – 

Improved spatially distributed modelling by usage of Lidar data, J. Hydrol. 519, S. 3492–3510. 

- Schöber, J., Achleitner, S., Bellinger, J., Kirnbauer, R., Schöberl, F. (2015): Analysis and 

modelling of snow bulk density in the Tyrolean Alps, in: Hydrol. Res. (in press). 

- Schönlaub, H., Hofer, B. (2009): Die Hochwassersituation bei abgeleiteten Bächen, in: 

WasserWirtschaft 09, S. 23-29. 



Advanced Flood Forecasting for Switzerland 

Erweiterte Hochwasservorhersagen für die Schweiz 

Karsten Jasper 1 ; Martin Ebel 1 

ABSTRACT 

Over the last few years the Federal Office for the Environment FOEN has put considerable 

effort into improving the quality and the range of their operational flood forecasts. Today 

the FOEN not only provides hydrological forecasts for the rivers and lakes of the Swiss Rhine 

catchment. It also produces and disseminates forecasts for the river systems of Rhone and 

Ticino. The underlying data basis has been significantly improved. This allows the implementation 

of detailed water balance models, yielding a higher modelling quality especially when 

modelling smaller and hydrologically complex catchments. Moreover different hydrological 

models are implemented in FOEN's forecasting system for a growing number of river basins. 

Hydrological forecasts thus can not only be compared on the basis of different numerical 

weather forecasts, but it is also possible to evaluate forecasts of up to 3 different hydrological 

models for several catchments. The use of these hydrological model ensembles will allow 

the comparison of the different modelling approaches and a calculation of the predictive 

uncertainty of a current forecast. 


In den letzten Jahren unternahm das Bundesamt für Umfeld (BAFU) erhebliche Anstrengungen, 

um den Umfang und die Qualität seiner operationellen Hochwasservorhersagen zu 

verbessern. Heute stellt das BAFU nicht nur für die Gewässer des schweizerischen Rhein- 

Einzugsgebietes hydrologische Vorhersagen zur Verfügung, sondern auch für die Gewässersysteme 

von Rhone und Ticino. Die Nutzung von verbesserten Datengrundlagen und der 

zunehmende Einsatz von detaillierten Flussgebietsmodellen führten zu einer allgemein 

höheren Qualität der hydrologischen Modellierung, insbesondere in komplexen oder kleineren 

Einzugsgebieten. Für immer mehr Flussgebiete werden am BAFU inzwischen nicht 

mehr nur meteorologische sondern auch hydrologische Modell-Ensembles gerechnet. Damit 

können im Einzelfall die Ergebnisse von bis zu drei verschiedenen Abflussmodellen miteinander 

verglichen und daraus ein Mass für die hydrologische Modellunsicherheit gewonnen 

werden. Aktuell werden regelmässig für mehr als 150 Flussabschnitte und Seen hydrologische 

Vorhersagen mit einem Vorhersagehorizont von bis zu 10 Tagen verbreitet. 

KEYWORDS 

operational flood forecasting; hydrological models 

1 Federal Office for the Environment FOEN, Ittigen, SWITZERLAND, karsten.jasper@bafu.admin.ch 



EINFÜHRUNG 

Operationelle Abfluss- und Wasserstandvorhersagen werden vom Bund seit Mitte der 80er 

Jahre durchgeführt. Bis vor wenigen Jahren waren diese Vorhersagen auf die Hauptgewässer 

im Einzugsgebiet des Rheins bis Rheinfelden beschränkt. Mit der operationellen Nutzbarmachung 

des Vorhersagesystems Delft-FEWS (Werner et al., 2013) und dem darin integrierten 

semi-distributiven Modell HBV-96 (Lindström et al., 1997) erweiterten sich jedoch in 2007 

die Möglichkeiten der Vorhersage am Bundesamt für Umwelt (BAFU). Von nun an waren am 

BAFU, dem hydrologischen Vorhersagezentrum der Schweiz, operationelle modellgestützte 

Vorhersagen für das gesamte Rhein-Gebiet bis Basel (etwa 36‘000 km 2 ) möglich. Die im 

HBV-96 verwendete Gebietsstruktur unterteilt das Vorhersagegebiet in mehr als 60 Teilgebiete 

mit Flächengrössen zwischen 100 und 3‘300 km 2 (Bürgi, 2008). Abflussvorhersagen sind 

jeweils für den Auslass eines jeden Teilgebietes möglich. Die Erfahrungen im operationellen 

Vorhersagebetrieb zeigten, dass mit dem eingesetzten Modell vor allem für grössere Flussgebiete 

qualitativ gute Abflussvorhersagen erreicht werden können. Für viele der mittleren und 

kleineren Zuflüsse waren hingegen oft keine zufriedenstellenden Vorhersagen möglich. 

Grund dafür sind vor allem modelltechnische Vereinfachungen in der Beschreibung von 

komplexen Einzugsgebieten. Hierzu gehören insbesondere eine zu geringe räumliche 

Auflösung der alpinen Gebiete und die Verwendung von stark vereinfachten Modellansätzen, 

z.B. zur Berechnung von Bodenfeuchte sowie von Schnee- und Gletscherschmelze. 

Im Rahmen des schweizerischen Aktionsplanes zur Optimierung von Warnung und Alarmierung 

vor Naturgefahren (OWARNA) wurden daher mit Sicht auf den hydrologischen 

Vorhersagebetrieb zwei Hauptziele formuliert: (a) Steigerung der Vorhersagequalität durch 

Berücksichtigung von verbesserten Datengrundlagen und dem Einsatz von detaillierten 

Flussgebietsmodellen sowie (b) Ausdehnung der hydrologischen Vorhersagen auf die 

mittleren und grösseren Gewässer der gesamten Schweiz (Hess, 2010). Mit der Inkraftsetzung 

der Alarmierungsverordnung (AV) in 2011 wurde zudem der gesetzliche Rahmen für die 

Arbeiten zur Zielerreichung gesetzt (Amiguet et al., 2016). 

RÄUMLICHE AUSDEHNUNG DER HYDROLOGISCHEN VORHERSAGEN 

Die «hydrologische Schweiz» umfasst unter Berücksichtigung ihrer ausländischen Gebietsanteile 

eine Gesamtfläche von knapp 58‘000 km 2 . Sie übertrifft damit die Schweizer Landesfläche 

um mehr als 40%. Das Rhein-Gebiet stellt mit etwa 63% den grössten Flächenanteil 

dar. Ihm folgen die Flussgebiete Rhone (18%) und Ticino (11%). Weniger als 8% der Gesamt - 

fläche entfallen auf die restlichen Flussgebiete (Inn, Doubs, Adda, Etsch). 

Das Vorgehen für den weiteren modellgestützten Ausbau der Hochwasservorhersage orientierte 

sich an der Grösse des bisher noch nicht erfassten Gebietes, aber auch an dessen 

Gefahrenpotenzial bei Hochwasser. Folgerichtig wurden nach der operationellen Inbetriebnahme 

des HBV-96 Rhein-Modells im Jahr 2007 vorrangig Vorhersagelösungen für die 

Flussgebiete von Rhone und Ticino erarbeitet. 


Das entwickelte Vorhersagemodell für die schweizerische Rhone deckt nicht dessen gesamtes 

Einzugsgebiet ab (bis zum Pegel Chancy), sondern zunächst nur den Bereich bis zum 

Genfersee (Fläche ca. 5‘500 km 2 ). Es basiert auf dem deterministischen, flächendifferenzierten 

Wasserhaushaltsmodell WaSiM (Schulla, 2015) und liefert seit 2012 operationelle Abflussvorhersagen 

für 13 Pegelstandorte. Das Modell berücksichtigt alle wesentlichen Einflussfaktoren, 

wie z.B. den Betrieb der zahlreichen Stauanlagen oder eine flächendetaillierte Schnee- und 

Gletscherschmelze. Neben dem am BAFU verwendeten Rhone-Modell existiert auch ein 

modellgestütztes regionales Vorhersagesystem, welches vom Kanton Wallis betrieben wird 

(Garcia Hernandez et al., 2014). 

Das für die Vorhersagen berücksichtigte Einzugsgebiet für den Ticino erstreckt sich bis zum 

Ausflussbereich des Lago Maggiore (Ticino bei Sesto Calende). Es umfasst eine Einzugsgebietsfläche 

von insgesamt 6‘600 km 2 und berücksichtigt auch die italienischen Zuflüsse zum 

Lago Maggiore, wie z.B. den Toce. Die hydrologischen Vorhersagen werden im Ticino-Gebiet 

durch eine kombinierte Anwendung zweier Modelle erhalten: Zunächst wird das hochauflösende, 

hydrotop-basierte Wasserhaushaltsmodell PREVAH (Viviroli et al., 2009) für die 

Berechnung der Abflussbildung verwendet. Anschliessend werden die von PREVAH simulierten 

Teilgebietsabflüsse an das Modell RS3.0 übergeben und mit diesem Gerinneabflüsse und 

Seewasserstände berechnet. RS3.0, dessen grundlegender Aufbau in García Hernández et al. 

(2007) beschrieben ist, berücksichtigt alle relevanten Daten und Strukturen zur Steuerung 

der Tessiner Speicheranlagen. Im Zusammenspiel mit Delft-FEWS bietet es darüber hinaus 

die Möglichkeit der Berechnung von Szenarien für die Seensteuerung im Hochwasserfall. 

Das gekoppelte Modellsystem wurde in enger Zusammenarbeit mit dem Kanton Tessin erstellt 

und löst das bisherige kantonal genutzte Modellsystem ab. Es wird in den operationellen 

Vorhersagebetrieb am BAFU überführt und dann regelmässig Wasserstands- und Abflussvorhersagen 

für mehr als 20 verschiedene Pegelstandorte liefern. 

EINSATZ VON DETAILLIERTEN FLUSSGEBIETSMODELLEN 

Parallel zur räumlichen Ausdehnung der hydrologischen Vorhersagen wird am BAFU 

fortlaufend in die Verbesserung der Vorhersagequalität investiert. Entsprechend wurde in 

den letzten Jahren die Abflussvorhersage zunehmend mit Modellen erweitert, welche das 

hydrologische Prozessgeschehen im Schweizer Alpenraum adäquat abbilden können. 

Die Anforderungen an die eingesetzten Modelle für den operationellen Vorhersagebetrieb 

sind hoch: Die Modelle müssen unterschiedlichste Gebietseigenschaften (z.B. mit und ohne 

Gletschereinfluss) und relevante Eingriffe auf den Abfluss- und Wasserhaushalt in geeigneter 

Weise berücksichtigen können (z.B. Seeregulierung, Speicherbetrieb durch Kraftwerke, 

Zu- oder Ableitungen). Im Weiteren müssen sie auch in der Lage sein, die aktuell verfügbaren 

Datengrundlagen optimal im Vorhersagebetrieb einzubinden. Die Modelle müssen für 

operationelle Anwendungen optimiert sein, aber dennoch die komplexen hydrologischen 

Prozesse in der Schweiz abbilden können. 


Vor diesem Hintergrund wurde am BAFU in 2010 damit begonnen, das bisher genutzte 

konzeptionelle Vorhersagemodell für den Rhein (HBV-96) schrittweise durch detailliertere 

Modelle zu ergänzen. Hierzu wurden einerseits bestehende operationelle Modellanwendungen 

in das BAFU-Vorhersagesystem integriert, wie z.B. das im Kanton Zürich betriebene 

Modellsystem PREVAH-Sihl (Zappa et al., 2010) oder das in der Hochwasservorhersagezentrale 

von Baden-Württemberg (LUBW) eingesetzte Modell LARSIM (Bremicker, 2010). 

Andererseits wurden auch neue regionale Vorhersagemodelle entwickelt, wie z.B. WaSiM- 

Alpenrhein (Schulla, 2013). Einen Überblick über die flächendetaillierten Regionalmodelle, 

welche derzeit im BAFU-Vorhersagesystem Verwendung finden, geben Tabelle 1 und 

Abbildung 1. Die Operationalisierung von weiteren prozessorientierten Flussgebietsmodellen 

ist vorgesehen, um vorhandene Lücken in der hydrologischen Vorhersagelandschaft zu 

schliessen und die Güte der Vorhersagen, insbesondere in kleineren und mittleren Flussgebieten, 

zu steigern. 

Abbildung 1: Aktuelle Abdeckung der Schweiz mit flächendetaillierten Vorhersagemodellen 

MODELLSTRATEGIE FÜR DIE HYDROLOGISCHE VORHERSAGE 

Die Vorhersagestrategie am BAFU sieht mittelfristig die Gesamtabdeckung der Schweiz mit 

regionalen WaSiM-Modellen vor. Damit läge praktisch ein Vorhersagemodell gleichen Typs 

flächendeckend in 500 x 500 m Auflösung für alle Flussgebiete der hydrologischen Schweiz 

vor. Zusätzlich werden im Sinne eines hydrologischen Multi-Modell-Ansatzes für mehrere 

Einzugsgebiete (insbesondere für Hot-Spot-Regionen) weitere Flussgebietsmodelle im 


Tabelle 1: Aktuelle flächendetaillierte Modellanwendungen in der hydrologischen Vorhersage am BAFU 

Modell Pegeleinzugsgebiet (Fläche) Jahr der Inbetriebnahme 

PREVAH Sihl bis Sihlhölzli (336 km 2 ) 2010 

Linth bis Mollis (600 km 2 ) 2010 

WaSiM Emme bis Wiler (940 km 2 ) 2010 

Rhone bis Genfersee (5‘350 km 2 , geschätzt) 2012 

Alpenrhein bis Bodensee (6‘700 km 2 , 

2014 

geschätzt) 

Aare bis Bern (2‘945 km 2 ) 2015 

Thur bis Andelfingen (1‘696 km 2 ) 2015 

Töss bis zum Rhein (425 km 2 , geschätzt) 2016 

Glatt bis Rheinsfelden (416 km 2 ) 2016 

PREVAH + RS3.0 Ticino bis Sesto Calende (6‘600 km 2 ) 2016 

LARSIM Rhein bis Basel (35‘897 km 2 ) 2016 

Vorhersagesystem am BAFU nutzbar sein. Die Abflussvorhersagen der verschiedenen Modelle 

helfen dem Prognostiker bei der Abschätzung der aktuellen hydrologischen Modellunsicherheit. 

Im Bedarfsfall kann für ausgewählte Einzugsgebiete von den Ergebnissen des Hauptmodells 

abgewichen und auf die Vorhersagen eines Alternativmodells zurückgegriffen werden. 

Das Vorhersagesystem Delft-FEWS wurde und wird im Auftrag des BAFU so weiterentwickelt, 

dass es den Prognostiker bei diesem Entscheidungsprozess sowie bei allfälligen 

Szenario-Rechnungen durch dafür massgeschneiderte Tools optimal unterstützt. 

Das beschriebene Vorgehen zur parallelen Anwendung von mehreren Modellen (hydrologisches 

Modell-Ensemble) steht im Einklang mit dem seit vielen Jahren verwendeten meteorologischen 

Multi-Modell-Ansatz. Seit dem Jahr 2007 bilden die Resultate von vier verschiedenen 

Wettervorhersagemodellen die Grundlage für die operationellen Abflussvorhersagen am 

BAFU. Die berücksichtigten numerischen Wettermodelle haben räumliche Auflösungen 

zwischen 2 und 16 km und Vorhersagelängen von 33 bis 240 Stunden. Sie liefern entweder 

deterministische Vorhersagen (COSMO-2, COSMO-7, ECMWF) oder stellen Ensemble-Vorhersagen 

bereit (COSMO-LEPS). Ab 2016 werden Daten von verfeinerten Wettermodellen 

für die Wasserstands- und Abflussvorhersagen verfügbar sein, und zwar von einem deterministischen 

1.1km-Modell (COSMO-1) und von einem 2.2km-Ensemble-Modell (COSMO-E). 

VERBESSERTE DATENLAGE 

In den letzten Jahren wurden grosse Fortschritte in der zeitnahen Bereitstellung von 

meteorologischen Daten der Bodenmessnetze gemacht. Die Automatisierung von Messstationen 

der MeteoSchweiz wie auch die zunehmende Nutzbarmachung von Daten anderer 

Messnetzbetreiber (kantonale Fachstellen, Institutionen, Private, ausländische Dienste und 


Behörden, etc.) führte zu einer markanten Erhöhung der operationell verfügbaren Datendichte, 

insbesondere im Bereich der Niederschlagsmessungen. Insgesamt werden derzeit die 

Messdaten von mehr als 800 meteorologischen Stationen in das Vorhersagesystem am BAFU 

eingespeist, darunter von etwa 260 Stationen, welche von der MeteoSchweiz betrieben 

werden (automatische Stationen vom SwissMetNet sowie automatisierte Niederschlagsstationen). 

Im Vergleich zu 2007 vervielfachte sich damit die Anzahl der für die hydrologische 

Vorhersage nutzbaren Wetterstationen, und zwar über alle Höhenbänder. Eine erhöhte 

meteorologische Stationsdichte, insbesondere bei den Niederschlagsmessungen, ist aus Sicht 

der hydrologischen Vorhersage extrem wertvoll, da sie die Interpolationen der punktuellen 

Messwerte in die Fläche zu verbessern hilft. Werden die meteorologischen Messdaten 

realistischer interpoliert, dann sind auch genauere hydrologische Modellnachführungen 

möglich. Die Initialbedingungen für die anschliessenden Vorhersagesimulationen verbessern 

sich entsprechend (genauere Systemzustände). 

Aber nicht nur die Anzahl der verfügbaren Messstationen hat Einfluss auf die Interpolationsgüte 

und damit auf die Wertigkeit einer meteorologischen Station im Vorhersagesystem. 

Weitere wichtige Einflussfaktoren sind deren Datenqualität, deren geographische Lage, deren 

Intervall der Datenlieferung oder deren Parameterauswahl. Die Anforderungen der am BAFU 

verwendeten Vorhersagemodelle an die meteorologischen Eingangsdaten sind in Tabelle 2 

zusammengefasst. 

Tabelle 2: Benötigte meteorologische Eingangsgrössen für die am BAFU verwendeten Vorhersagemodelle; Niederschlag (N), 

Lufttemperatur (LT), Windgeschwindigkeit (W), relative Luftfeuchtigkeit (RF), Taupunkttemperatur (TD), Globalstrahlung (GS), relative 

Sonnenscheindauer (SSD), Luftdruck (LD) 

Modell P T W RF TD GS SSD LD 

HBV x x x - x - - - 

PREVAH x x x x - x x - 

WaSiM x x x x - x x - 

LARSIM x x x x - x - x 

In Anbetracht der Bedeutung des Niederschlages für die Abflussvorhersage werden im 

BAFU-Vorhersagesystem nicht nur stationsbezogene Niederschlagsmessungen verarbeitet, 

sondern auch vorprozessierte Kombinationsprodukte, welche aus Echtzeit-Messungen vom 

Niederschlagsradar und Bodenstationen bestehen (Sideris et al., 2013). Diese als CombiPrecip 

bezeichneten Flächenprodukte des Niederschlags stellen attraktive Ergänzungen zur 

klassischen stationsbezogenen Niederschlagsinterpolation dar. Aktuelle Untersuchungen 

zeigen, dass mit Hilfe von CombiPrecip die Simulation von Hochwassern in Einzelfällen 

deutlich verbessert werden kann. Die Verbesserung gegenüber der klassischen Interpolation 

von Punktmessungen ist bisher jedoch nicht systematisch, sondern offenbar von der 


etrachteten Region und vom Ereignistyp abhängig. Insofern kann CombiPrecip bis auf 

weiteres nur eine Ergänzungsrolle in der hydrologischen Vorhersage am BAFU einnehmen. 

Die Aufnahme von neuen flächendetaillierten Flussgebietsmodellen ging einher mit der 

Berücksichtigung einer erhöhten Anzahl an hydrologischen Messstationen, vor allem in den 

komplexen und oft schwierig zu modellierenden Zuflussgebieten. Allein für das Rhein-Gebiet 

werden heute die Abflüsse und Wasserstände von etwa 120 Stationen für die Nachführung 

der hydrologischen Vorhersagemodelle genutzt. Diese Stationsanzahl liegt deutlich über 

derjenigen von 2007 (HBV-Rhein mit 70 Stationen). Die verwendeten Stationen gehören 

überwiegend zum Messnetz des BAFU. Inzwischen werden aber auch zunehmend kantonale 

und ausländische Stationen sowie Stationen von Kraftwerksbetreibern in die Vorhersagen 

einbezogen. Ein dichteres Messnetz an Abfluss- und Wasserstandsmessungen bedeutet in 

erster Linie mehr Kontrollpunkte für die Modellrechnungen, was sich insbesondere im 

Hochwasserfall positiv auf die Güte der Abflusssimulation und damit der Vorhersage 

auswirkt. 

Neben modellrelevanten Verbesserungen bei den zeitbezogenen Daten gab es im Rahmen 

des Modellausbaus auch Fortschritte in der Nutzung von raumbezogenen Daten. Prozessorientierte 

Vorhersagemodelle haben hier naturgemäss grosse Vorteile gegenüber vereinfachten 

Modellen. Sie sind wesentlich besser in der Lage, die für die hydrologische Modellierung 

allgemein verfügbaren Datengrundlagen auszunutzen. Die neuen Flussgebietsmodelle 

am BAFU verwenden jeweils hochauflösende Flächendaten zur Orographie, zur Landnutzung, 

zum Boden und zur Hydrogeologie, um daraus modellrelevante Parameter, wie z.B. 

die Speicher- und Leitfähigkeiten im Untergrund, abzuleiten. Ausserdem berücksichtigen 

sie möglichst detailliert Gerinnedaten (Geometrie, Rauhigkeiten) sowie Betriebsregeln zur 

Seen- und Speichersteuerung. 

ERGEBNISSE 

Die Vorteile von detaillierten gegenüber einfachen Modellen zeigen sich oft bei Abflusssimulationen 

in sogenannten Kopfgebieten, d.h. in Gebieten ohne gemessene Zuflüsse. 

Im vorliegenden Beitrag ist dies beispielhaft für ein Hochwasserereignis im Einzugsgebiet 

der oberen Emme (124 km 2 ) illustriert, welches im Mai 2015 durch Starkregen ausgelöst 

wurde. Die im ersten Vorhersageabschnitt (bis +30h) weitgehend zutreffenden Niederschlagsvorhersagen 

der Wettermodelle COSMO-2 und COSMO-7 resultieren hier nur beim 

detaillierten Modell (WaSiM) in qualitativ hochstehende Abflussvorhersagen (vgl. Abb. 2 

und 3). Weniger gute Vorhersagen werden im aktuellen Beispiel durch die Daten des 

ECMWF-Wettermodells erzielt. Die Ergebnisse zeigen erwartungsgemäss, dass die Qualität 

einer Abflussvorhersage, insbesondere in kleineren Einzugsgebieten, entscheidend von 

der Güte der Niederschlagsvorhersage abhängt. 


Abbildung 2: Operationelle Hochwasservorhersagen mit dem Modell HBV-96 für den Pegel Emme-Eggiwil am 01.05.2015 (06 Uhr) 

Abbildung 3: Operationelle Hochwasservorhersagen mit dem Modell WaSiM für den Pegel Emme-Eggiwil am 01.05.2015 (06 Uhr) 


FAZIT 

In den letzten Jahren hat die Hochwasservorhersage am BAFU die Qualität und den Umfang 

ihrer bereitgestellten Abflussvorhersagen deutlich gesteigert. Dazu trugen neben einer verbesserten 

Daten- und Modellsituation auch Weiterentwicklungen im Bereich der Vorhersageplattform 

Delft-FEWS bei. Insgesamt ist das BAFU seinem Ziel der Gesamt abdeckung 

der Schweiz mit modellgestützten hydrologischen Vorhersagen einen grossen Schritt näher 

gekommen. Der zunehmende Einsatz von detaillierten Flussgebietsmodellen ermöglicht 

verbesserte Abflusssimulationen und trägt damit zu einer Erhöhung der Vorhersagegenauigkeit 

bei, insbesondere in komplexen Einzugsgebieten. Die Nutzung von meteo rologischen 

und hydrologischen Modell-Ensembles hilft, die Vorhersageunsicherheit besser abzuschätzen. 

LITERATUR 

- Amiguet C., Bürgi T., Murer D., Schmutz C., Volken D. (2016): Warnungen der Bundesfachstellen 

der Schweiz vor Unwetter und Hochwasser. Interpraevent 2016 

- Bremicker M. (2010): Das Wasserhaushaltsmodell LARSIM – Modellgrundlagen und 

Anwendungsbeispiele. Freiburger Schriften zur Hydrologie, Band 11, 119 S. 

Bürgi T. (2002): Operational flood forecasting in mountainous areas - an interdisciplinary 

challenge. In: Spreafico M. and Weingartner R. (eds.). International Conference in Flood 

Estimation. CHR Report II-17: 397–406. Bern 

- Bürgi T. (2008): Operationelle Hochwasservorhersagen für das Einzugsgebiet des Rheins 

in der Schweiz. Interpraevent 2008. Conference Proceedings, Vol. 1, 51-62 

- García Hernández J., Jordan F., Dubois J., Boillat J.-L. (2007). Routing System II: 

Flow modelling in hydraulic systems. In A. Schleiss (ed.), Communication 32 du Laboratoire 

de Constructions Hydrauliqes. Lausanne: EPFL 

- Garcia Hernández J., Claude A., Paredes Arquiola J., Roquier, B., Boillat J.-L. (2014): 

Integrated flood forecasting and management system in a complex catchment area in the 

Alps: Implementation of the MINERVE project in the Canton of Valais. In A. Schleiss, J. 

Speerli, R. Pfammatter (eds.), Swiss Competences in River Engineering and Restoration. 

London. ISBN 978-1-138-02676-6 

- Hess J. (2010): OWARNA Folgebericht. 53 S. 

- Lindström G., Johansson B., Persson M., Gardelin M., Bergström S. (1997): Development 

and test of the distributed HBV-96 hydrological Model. Journal of Hydrology 201: 272-288 

- Schulla J. (2013): Modellaufbau WaSiM-Alpenrhein. Interner Bericht. BAFU. 216 S. 

- Schulla J. (2015): Model Description WaSiM. 335 S. (verfügbar auf www.wasim.ch) 

- Sideris IV., Gabella M., Erdin R., Germann U. (2013): Real-time radar-rain-gauge merging 

using spatio-temporal co-kriging with external drift in the alpine terrain of Switzerland. 

O.J.R. Meteorol. Soc.. doi: 10.1002/qj.2188 


- Viviroli D., Zappa M., Gurtz J., Weingartner R. (2009): An introduction to the hydrological 

modelling system PREVAH and its pre- and post-processing tools. Environmental Modelling 

& Software 24: 1209–1222 

- Werner M., Schellekens J., Gijbers P., van Dijk M., van den Akker O., Heynert K. (2013): 

The Delft-FEWS flow forecasting system. Environmental Modelling & Software 40: 65-77 

- Zappa M., Jaun S., Badoux A., Schwanbeck J. und weitere Autoren (2010): IFKIS-Hydro 

Sihl: Ein operationelles Hochwasservorhersagesystem für die Stadt Zürich und das Sihltal. 

Wasser Energie Luft, Heft 3: 238-248 



Can Twitter catch precursory phenomena before 

sediment disasters? 

Masaru KUNITOMO, master 1 ; Joko KAMIYAMA, master 1 ; Kazuki MATSUSHITA, master 1 ; Yuzuru YAMAKAGE, doctor 2 ; 

Kunihiro TAKEDA, master 2 ; Cheng QIU, doctor 3 ; Akiko ITO, bachelor 3 ; Takeru ARAKI, master 3 

ABSTRACT 

In Japan, the Cabinet Office recommends that municipalities immediately judge issuance of 

evacuation instructions when they grasp the signs of sediment-related disasters. However, 

this early warning mechanism has not necessarily functioned effectively, because it is difficult 

for local governments to get such information from local residents. In such situation, "twitter" 

has been sharply growing in the number of users and is being incorporated into social system. 

Then, with focus on the real-time property of "twitter", the authors have started a research to 

identify information on the precursory phenomena of sediment-related disasters contained in 

users' "tweets". Then, we confirmed that tweets are fairly effective in detecting precursory 

phenomena etc. particularly in areas with a large population size. And that application to 

early warning systems can be expected. 

KEYWORDS 

Sediment Disaster; early warning; SNS; Twitter 

INTRODUCTION 

It has been said that signs of sediment-related disasters, such as "earth rumbling" and "earth 

smelling," are important early warning signal, and the Cabinet Office of Japan recommends 

on its guideline that municipalities immediately judge issuance of evacuation instructions and 

local residents quickly take actions to protect their safety when they grasp the signs. In fact, 

there are many cases reported where local residents found precursory phenomena of 

sediment-related disasters and evacuated together with family members and/or neighboring 

residents to avoid personal injuries. However, there are also reports that even if local residents 

find precursory phenomena of sediment-related disasters, such information is transferred 

only to family members and/or neighboring residents, and is rarely reported to local 

governments (Miyase et al., 2009). Thus, information on the precursory phenomena of 

sediment-related disasters is effective in judging evacuation, but there are some issues to be 

solved concerning quick collection and sharing of such information to use such phenomena 

information for early warning system. 

1 National Institute for Land and Infrastructure Management MLIT, Tsukuba, JAPAN, kamiyama-j253@nilim.go.jp 

2 Fujitsu Laboratories Ltd. 

3 Nippon Koei Co. Ltd. 

IP_2016_vc 


In such situation, "twitter" has been sharply growing in the number of users with the 

advantages of real time and spreading properties and is being incorporated into social system. 

In disaster prevention as well, cases of using twitter as a risk communication tool has been 

increasing. 

Then, with focus on the real-time property of "twitter" information, the authors have started 

a research to identify information on the precursory phenomena or occurrences of sediment-related 

disasters contained in users' "tweets" stating uncertainties or fears about heavy 

rain etc. and incorporate them into early warning systems. 

In this study, we analyzed the tweets posted in the events of the July 2012 Northern Kyushu 

Heavy Rain and the August 2014 Hiroshima Heavy Rain and studied the possibility of 

detecting the situation of the site where a sediment-related disaster occurred using twitter 

information. 

CHARACTERISTICS OF TWITTER 

Twitter is a kind of social media called "mini blog" or "micro blog." As characteristics of 

twitter, Kazama (2012) states that twitter information quickly and widely spreads due to 

the high real-time property despite the limitation to a maximum of 140 characters per tweet, 

easier information exchange with or transfer to other users, and more casual and easier 

communication of information among twitter users than general SNS media. 

There are many cases reported where twitter was very effective due to its real time and 

spreading properties in risk communication immediately after the occurrence of disasters 

(e.g., Ishikawa et al, 2012). On the other hand, it is also fact that some are concerned about 

spread of false rumors. 

Thus, there seems to be pros and cons on using twitter as a risk communication tool. 

However, Taniguchi (2012) points out that twitter is effective as an information sharing tool 

in the initial phase of a disaster where it is difficult to grasp the overview even if negative 

effects such as spread of false rumors are disseminated. 

Furthermore, in recent years, studies to probe the possibility of using twitter as a social sensor 

have been also going on, including estimation of the epicenter of an earthquake by analyzing 

tweets and detection of the spread of influenza (e.g., Sakaki et al., 2010). 

As we reviewed the findings of such prior studies comprehensively, it is considered that use of 

twitter as a social sensor enables detection of natural phenomena that are hardly detected by 

physical sensor, although the reliability and stability of twitter are much lower than those of 

physical sensors, and that twitter can be a useful tool to visualize what cannot be detected by 

conventional means by supplementing physical sensor information. 


EXAMINATION OF THE POSSIBILITY OF GRASPING THE CONDITION OF THE DISASTER SITE 

ANALYTICAL METHOD 

After screening the twitter information on the Web to identify the municipalities where the 

occurrence of a disaster is referred to in posted tweets ("municipality estimation"), we 

extracted tweets related to heavy rain and classified them into the situation categories such as 

sediment-related disasters and weather condition etc. in order to grasp the situation of the 

site with tweet content of such extracted information and set key words that represent each 

situation category. 

Then, we extracted tweets that include a key word we set up and estimated the situation 

based on changes in the number of tweets in each situation category. Further, we examined 

the effectiveness of grasping the situation from the tweets posted with critical content 

suggesting an emergent situation. To carry out municipality estimation, we adopted the 

procedure taken by Takeda et al (2014) through partial simplification. 

SUBJECT DISASTER AND OUTLINE OF TWITTER INFORMATION 

This study examined the sediment-related disasters caused by the Northern Kyushu Heavy 

Rain and the Hiroshima Heavy Rain, which hit Minami-Aso village, Aso city, Kumamoto 

prefecture ("Aso") and Hiroshima city, Hiroshima prefecture ("Hiroshima"), respectively, and 

in both of which it is known that local residents found precursory phenomena of sedimentrelated 

disasters (e.g., Sakai et al., 2013) (Figure 1). Time zones selected to analyze data were 

Figure 1: Locations and outline of subject disaster sites 


from 16:00 on July 11 to 12:00 on July 12, 2012 for the Northern Kyushu Heavy Rain and 

from 1:00 to 5:30 on Aug. 20, 2014 for the Hiroshima Heavy Rain. In both events, rainfall 

rapidly increased at midnight and sediment-related disasters occurred at dawn and caused 

damage etc. to many people. 

Of the tweets posted during the duration for analysis above, 2,207 tweets were estimated 

to have been posted from Aso, and 5,813 tweets, from Hiroshima. 

EXTRACTION AND CLASSIFICATION OF TWEETS RELATED TO DISASTERS 

Of the tweets of which locations were estimated, we extracted 968 tweets related to the Aso 

and 1,617 tweets for the Hiroshima, based on the contents of tweets. Further, the extracted 

tweets were classified into the "situation categories" of Table 1. 

Table 1: Situation category. 

KEY WORD SETTING BY TEXT ANALYSIS ACCORDING TO SITUATION CATEGORIES 

Using "KH Coder," software for text analysis, we automatically extracted frequent words 

according to situation categories by analyzing the tweets after decomposing them into part of 

Table 2: Examples of key words (Category related to sediment-related disasters). 


speech. From among the words extracted, we adopted as key words representing the content 

of each situation category (e.g. "land slide", "disaster", etc. for "Sediment-related disasters" 

and "Dangerous", "Scary", etc. for "Feeling and state of mind."). Table 2 provides examples 

for the category of "Sediment-related disasters." 

Figure 2: Time series changes in the number of tweets in the category of sediment-related disasters (Hiroshima) 

Figure 2 shows in graph form time series changes in the number of tweets of which contents 

were classified and extracted after text analysis for the category of sediment-related disasters 

in the Hiroshima Heavy Rain (, which do not necessarily contain key word(s)) and the 

numbers of tweets extracted mechanically using the set key words, together with changes in 

intensity of rainfall. Since the trends of both types of numbers were generally consistent, it 

was verified that the key words set up can be used to capture changes in the number of 

tweets in each situation category. 

GRASP OF THE SITUATION USING KEY WORDS 

Figure 3 shows the timing of response actions etc. to the disaster in Aso and rainfall, changes 

in the number of tweets extracted using key words for each situation category, and the timing 

of appearance of tweets showing critical content. 

The number of tweets in the weather related category is increasing continuously, which 

suggests the situation where heavy rainfall is continuing. Meanwhile, tweets in the categories 

of flood and sediment-related disasters begin to increase about 7:00, which is far behind the 

time of the first reporting of the disaster. However, when we checked the contents of tweets 

individually, we found that the occurrence of a sediment-related disaster could be detected a 

little earlier (6:22). Additionally, when tweets are not limited to those for which the location 


Figure 3: Grasped disaster situation by analyzing tweets (Aso) 

is estimated, relevant tweets appeared in a range of heavy rain areas including Aso in much 

earlier time zones, e.g., "Landslide, now" at the time between 2:00 and 3:00. 

Figure 4 provides the results of analysis conducted on the Hiroshima Heavy Rain. For 

Hiroshima, changes in the number of tweets and contents of tweets also suggest the situation 

where heavy rainfall is continuing. This is similar to Aso, but the number of tweets in the 

category of sediment-related disasters increased in time zones much earlier than the 

occurrence of the sediment-related disaster, which is different from the situation in Aso. 

Further, in Hiroshima, when tweets are not limited to those for which the location is 

estimated, tweets stating the detection of precursory phenomena appeared at 2:50, earlier 

than the time zone during which debris flow is said to have occurred intensively, such as 

"I've been hearing a sound like the one heard when stones are rolling down the river" (this 

was estimated since heavy rain was falling only in Hiroshima throughout the Japan when 

the tweet was posted). Difference of the size of population (the number of twitter users) is 


Figure 4: Grasped disaster situation by analyzing tweets (Hiroshima) 

clearly the main factor of the early increase in the number of tweets in Hiroshima compared 

to Aso. Therefore, in order to detect precursory phenomena etc. at an earlier phase, it is 

important to extract tweets including critical content individually, as well as changes in the 

number of tweets. At present, when using twitter as an early warning system, it is necessary 

to provide measures to prevent omission of critical tweets, such as joint use of tweets of 

which locations are limited to prefectural level. 

CONCLUSIONS 

This study examined the possibility of grasping precursory phenomena from changes in the 

number of tweets extracted using key words and from the timing of appearance of tweets 

including critical content by setting key words after classifying tweets related to the heavy 

rain into situation categories. As a result, we confirmed that tweets are fairly effective in 

detecting precursory phenomena etc. in areas with a large population size, such as Hiroshima. 

It was also found that information on the feelings, state of mind, etc. of local residents, which 


can be of any help in determining directions for evacuation, can be fairly grasped and that 

application to warning / evacuation systems can be expected. 

On the other hand, in areas with a small population, it is difficult to detect precursory 

phenomena etc. only with the increase in the appearance of tweets since the number of 

tweets is very small. Therefore, in order to grasp the precursory phenomena of sedimentrelated 

disasters, etc. using twitter information, including areas with a small population, we 

intend to grasp situation changes sensitively by introducing the technology for evaluating the 

value of each tweets according to areas and to reduce omissions of tweets that include critical 

content by improving the accuracy of location estimation. 

We also intend to develop application software to improve the reliability as disaster prevention 

information, e.g. by overlapping display of the heavy rainfall areas obtained from the 

precipitation radar. 

REFERENCES 

- Masayuki M., Toshihiro K., Shota K. (2010) Residents' Attitude Survey in the July 2009 

Yamaguchi Heavy Rain Disaster, FY 2010 Japan Society of Erosion Control Engineering 

Presentation Summary 

- Kazuhiro K. (2012) Information Dissemination in Twitter, Journal of the Japanese Society 

for Artificial Intelligence, Vol. 27, No. 1: 35-42. 

- Ikki O., Yutaka M. (2012) Twitter and social media, Journal of the Japanese Society for 

Artificial Intelligence, Vol. 27, No. 1: 34. 

- Tetsuya I., Akiyuki K., Kimiro M. (2012) Investigation of information sharing by Twitter 

users during the 2011 heavy snow disaster in San-in region of western Japan, A Collection of 

Papers of the Institute of Social Safety Science No. 17. http://isss.jp.net/isss-site/wp-content/ 

uploads/2013/08/2012-014_cd.pdf 

- Shinichiro T. (2012) Effectiveness of twitter at disasters --- Taking as an example the heavy 

rain disaster by Typhoon No. 12 in September 2011, Disaster Information No. 10: 56-67. 

Sakaki, T., Okazaki, M., Matsuo, Y. (2010) Earthquake shakes twitter users: Real-time event 

detection by social sensors, Proc. 19th Conf. on World Wide Web, 851-860. 

- Kunitaka T., Tadanobu F., Shigetaka T., Yuzuru Y., Ken A., Akiko I., Ken M., Junichi K., 

(2014) Estimation of Occurrence of Disaster Using Twitter Data --- Report on the cases 

investigated concerning sediment-related disasters in the 2012 Northern Kyushu Heavy Rain, 

FY2014 Japan Society of Erosion Control Engineering Research Presentation Summary B: 

172-173. 

- Nobuaki S., Ryoichi M., and Toshihiro K. (2013) Behaviors of residents in case of sediment-related 

disaster caused by the July 2012 Northern Kyushu Heavy Rain, Journal of the 

Japan Society of Erosion Control Engineering (Shin-Sabo), vol. 66, No. 2: 57-63. 



The flood warning service of the Austrian Federal 

Railways 

Günther Kundela, DI 1 ; Ines Fordinal, DI 2 ; Florian Mühlböck, DI 2 ; Christian Rachoy, DI 1 

ABSTRACT 

According to Austrian legislation, the Austrian Federal Railway Infrastructure Company 

(OEBB Infra) is responsible, on one hand, for the safety and on the other hand for the high 

availability of the national and transnational railway tracks in Austria. To fulfil these high 

demands different countermeasures against natural hazards are necessary. Beside a high 

range of technical protection measures also organisational measures such as flood warning 

services are very important. On the basis of experiences a 3-phase warning service was 

developed. With the incoming weather forecasts the Early Warning Phase of the OEBB Flood 

Warning service starts and preliminary analyses of the expected development of the flood 

event are initiated. The most important step during the event phase is to initiate the required 

emergency measures. The end of a flood event is the beginning of the next event. An 

accurate event documentation is very helpful to improve the warning processes and to plan 

optimal countermeasures. 

KEYWORDS 

Railway infrastructure; flood warning service 

INTRODUCTION 

Since natural hazards have impact on the safety, availability and economy of railway 

infrastructures, therefore the protection of transported people, the railway staff and goods 

must be the top priority of natural hazard management. Regarding the fact that damage on 

railway infrastructure can be substantial and hindering, the Austrian Federal Railway 

Infrastructure company (OEBB Infra) is continuously improving its services and fully liable 

for safety issues. 

Floods, especially those with a large spatial extent, cause enormous costs for recovery of the 

railway infrastructure such as train stations, tracks, bridges or catenaries. The flood events on 

the Arlberg in 2005 and the flood event on the March River in 2006, as examples, generated 

total costs of 80 Million Euro for infrastructure recovery. Following the risk circle (Figure 1) 

there are different possibilities to reduce the risk of natural hazards. Risk maps, as a valuable 

support for planning preventive measures, help to identify flood-prone areas. Technical 

countermeasures such as flood protection dams in Austria are designed by using flood events 

with a 100-year return period and neglecting larger events. Warning services, contingency 

1 ÖBB-Infrastruktur AG, Vienna, AUSTRIA, Christian.Rachoy@oebb.at 

2 Riocom Ingenieurbüro A. Schwingshandl Handelskai 92 1200 Wien 



planning and –training as organisational measures support the handling of disasters. As 

pointed out by Rudolf-Miklau (2009), measures, if applied in optimal chronology and 

functional order, are likely to achieve their intended effect. 

However, the hazard itself cannot be avoided, thus the main objective of organisational 

counter measures is to protect lives. The early observation of natural hazard events through 

the use of warning systems is crucial, especially in the case of failure of technical countermeasures. 

Therefore, OEBB Infra is intensively working on the development of comprehensible 

and robust systems to predict meteorological and hydraulic events. 

Figure 1: OEBB Infra - Risk Cycle 

STRATEGIES AND TOOLS 

Natural Hazard Management aims at reducing risk specifically by following the phases of the 

Risk Cycle. When it comes to floods, infrastructure operators have various means for the 

protection of technical facilities as well as for the absolute safety of their customers and 

employees. Primarily, there are technical safety measures such as flood protection dams or 

mobile flood protection walls alongside the endangered track. By creating flood risk maps, 

endangered hot spots of the railway can easily be identified and thus contribute to preventive 

measures as best as possible. In the phases of provision and preparation, organisational 

measures are included, such as emergency trainings which are executed on a regular basis to 

ensure a well prepared and skilled staff. Furthermore, alarm plans are elaborated for specific 

railway sections and warning levels. The OEBB Infra operates its own weather informationand 

warning system called “infra:wetter”. This includes Austrian-wide weather forecasts, 

real time weather radar data as well as customised warning levels for defined rail sections. 

Additionally, the web page supplies all users with specific information about fire- and 


avalanche risk and gale warnings. It provides those responsible for the affected area standardised 

and replicable weather warnings for the assigned parts of the railway network via text 

message and e-mail. These customised warnings are crucial to the preparation of organisational 

countermeasures against extreme weather conditions. Nevertheless, it does not offer 

hydrological and hydraulic information about rivers relevant to the railway network. 

Due to gathered experiences at the last events it became clear that the earlier one knows 

about possible flood events the better organisational countermeasures can be taken, such as: 

– Constant observation of the flood-prone hot spots 

– Closing of the affected tracks in advance 

– Evacuation of flooded areas 

METHODS AND IMPROVEMENTS 

In extension to the former approach, which was exclusively based on weather and precipitation 

forecasts, the flood risk warnings also incorporate the analysis of historic flood event data 

as well as flood forecasts and the expertise of hydrological warning services. The utilised flood 

risk maps also include three-dimensional spatial analyses which take into account railway 

structures such as bridges and culverts instead of the flooded areas along the track only. 

A fundamental value of information lies in the height of the railway facilities referred to the 

considered water level (so called ‘warning sections’). The inclusion of as much existing data 

and information as possible (for example administrative hydrological warning systems and 

the expertise of Federal Hydrographic Services) is just one crucial improvement of the flood 

warning service. –. Through well-established hydrological models of the Hydrographic 

Services, flood level forecasts (depending on the size of the observed river catchment) can be 

provided several hours in advance. In addition, the Natural Hazard Management of the OEBB 

Infra together with the hydrologists of the federal services determine on certain water gauges 

which are of particular interest for the flood warnings for the railway tracks. In case of an 

announced flood event, the water levels of the selected water gauges will be included in the 

flood report of the Federal Hydrographic Services and thus help to further improve the local 

flood warnings. 

ASSESSMENT OF FLOOD POTENTIAL FOR RAILWAY INFRASTRUCTURE IN AUSTRIA 

The Austrian Federal Ministry of Agriculture, Forestry, Environment and Water Management 

(BMFLUW) holds a complete dataset of flood-prone areas, water depths and flow velocities 

in Austria. This dataset is a collection of all available data from the technical departments of 

Austria’s nine provinces, the Austrian service for torrent- and avalanche control as well as 

the Federal Ministry. It covers three scenarios in form of the return periods of 30, 100 and 

300 years and serves as a valuable, nationwide overview of the flood situation and offers a 

good basis for the assessment of the flood potential for the railway infrastructure in Austria. 


FLOOD WARNING LEVELS 

A first step for the detection of flood affected areas was a spatial intersection between the 

railway network and the available flood scenarios. To emphasise the relevance of the data 

basis, a distinction between the various classes of data origin was displayed as separate maps. 

In a later stage, the accomplishment of a delineation of relevant areas for the warning process 

which would take into account three-dimensional flood height information from a digital 

elevation model was a top priority. The combination of distance to the rails and the height 

difference between the water level and the top of the railway embankment is classified in 

three warning levels: 

– Orange warning section: marks flooded areas close to the railway track with significant 

height differences and thus rails are unlikely to be affected by a flood. 

– Red warning section: identifies flooded areas close to the railway track with little height 

differences and thus rails are likely to be affected by a flood. 

– Purple warning section: reveals intersections between flood areas and railway tracks. 

A detailed explanation is given in Figure 2. The warning sections visualise the potential 

extent of flood scenarios for three different return periods. This information represents 

different cases in a static way, whereas the dynamic part is included by incorporating 

incoming (hydrographic) warnings. Depending on the forecasted warning, the most appropriate 

case is used to identify areas where measures have to be taken. The spatial classification 

Figure 2: Classification of Warning Sections 


of warning sections for flood events is designed in a way that is consistent with the warning 

levels of the weather warning system already in use. This enables an implementation into the 

existing internal web GIS and provides the users with an already familiar classification 

scheme. 

PROCEDURE OF THE FLOOD WARNING SERVICE 

The Flood Warning Service is divided into three different phases, using the available data 

basis to achieve the best possible result (Figure 3). Furthermore, the on-call availability of a 

service operator providing this service, as well as the involvement of several key figures 

guarantee a successful, smooth and especially continuous performance of the Flood Warning 

Service. 

Figure 3: Procedure of the Flood Warning Service of the OEBB Infra 

PHASE 1: EARLY WARNING 

With the incoming weather warning from the above mentioned internal weather warning 

system(usually 3 to 5 days prior to the forecasted event), the ‘Early Warning Phase’ starts 

together with the initiated preliminary analyses of the expected development of the flood 

event. These analyses use characteristic values and summarised flood profiles of known 

historic flood events in the affected regions and river basins. The flood event profile aims at 


giving a clear overview of each individual historic flood event. It contains detailed information 

about the hydro-meteorological system state, the precipitation, the runoff and a textual 

description of the event. 

By comparing weather forecasts to historic flood events, a rough estimation of the scale of the 

expected flood (return period) is attempted. The responsible federal Hydrographic Services 

are consulted in this early phase of the warning system for the characteristics of the catchment. 

The next step is to approximately identify the affected railway sectors according to the 

warning level classification. The periodic flood situation report summarises all crucial 

information, which was collected, analysed and prepared by the Flood Warning Service 

during the Early Warning Phase. It supports the Natural Hazard Management of OEBB Infra 

in taking decisions and initiating measures in the emergency planning. 

PHASE 2: WARNING 

The Warning Phase is initiated as soon as specific flood forecasts are provided by the Provincial 

Hydrographic Services. The flood forecasts are the result of hydrological models and are interpreted 

by experts with comprehensive regional knowledge. Thus, the preliminary analysis 

can be updated due to this additional data. On the basis of hydrological modelling, the estimation 

of the expected flood (return period) ensures a rather accurate identification of 

affected railway sectors. 

Furthermore current water gauge levels are published by the Hydrographic Central Office. 

Subsequently, predefined threshold values are monitored for each identified railway sector. 

As before, all crucial information for the support of the emergency planning is summarised in 

the flood situation report. The warning phase lasts from when flood forecasts are available 

until certain water level thresholds are exceeded. 

PHASE 3: EVENT 

With the observation of the water gauges, once again, the expected event scale can be updated 

and the identification of affected railway sectors can be derived. The most important 

measure during the event phase is the initiation of the required emergency measures, which 

means that an incident command is installed and the general management decides about 

temporary mitigation measures such as constant observation, mobile flood protection or 

railway closures. Additionally, constant communication between the involved authorities 

(OEBB, the Flood Warning Service and the Hydrographic Services) is highly important. 

The flood situation reports are generated in shorter intervals and focus on the temporal 

course of the event. As soon as the water gauges show a decreasing in water level, the 

general management decides on the end of the alarm. 

Shortly after the flood event, analysis is carried out and the possibility for reflection among 

the involved experts to gain fundamental experience and to improve the Flood Warning 

Service with every event is given. 


FLOOD SITUATION REPORT 

The Flood Situation Report (Figure 4) is the core element of communication between the 

Flood Warning Service and the OEBB. It summarises current information on potential flood 

risk in a catchment. It supports the Natural Hazard Management of the OEBB Infra by providing 

important professional advice for the decision making and initiate mitigation measures 

during the emergency planning. More specifically, it contains the following parts: 

– River catchment 

– Description of meteorological and the hydrological situation 

– Characterisation of the system’s state (soil moisture, snow fall, snowmelt, zero degree line) 

– Estimation of expected runoff (scale of flood event as return periods) 

– Identification of affected railway sections 

– Summarised experts’ opinion 

Figure 4: Structure of the Flood Situation Report 

Including the expertise of Provincial Hydrographic Services, the system’s state is characterised 

by a qualitative assessment on whether soil moisture or snowmelt are likely to influence the 

flood development. The approach to estimate the expected runoff depends on the warning 

phase and available data. The pre-assessment is based on historic flood events, later on 

specific flood forecasts which are published by the Hydrographic Services and can be used to 


update the estimation. To supply those responsible for the emergency planning with crucial 

information, the identification and display of the affected railway sections for the return 

periods of 30, 100 and 300 years (orange, red and purple sections in Figure 2) are highly 

important and count as final steps. 

CONCLUSIONS 

According to Austrian legislation, the Austrian Federal Railway Infrastructure Company 

(OEBB Infra) is, on one hand, responsible for the safety and, on the other hand, for the high 

availability of national and transnational railway tracks in Austria. 

To fulfil these high demands, different countermeasures against natural hazards are necessary. 

Beside a high range of technical protection measures like dams, barriers or fences, organisational 

measures such as warning systems, training, and contingency plans are equally 

important. 

There are three major reasons for the utilisation of organisational measures. First of all, 

technical countermeasures are not always possible due to natural, technical and legal reasons. 

Secondly, the planning, approval and construction of technical measures take very long – often 

years. Thirdly, technical countermeasures are planned by using design events: if a natural 

hazard event exceeds the design event, organisational measures help to protect human life. 

Warning systems are inexpensive and can be well adapted to resist influences from climate 

change. 

The accurate discharge models of the different catchment areas build the basis for planning 

countermeasures against flood events. In Austria, these discharge models are calculated by 

nine different federal countries. For a nationwide infrastructure operator, it is challenging to 

combine and interpret the different data. 

The quality of weather forecasts is very high, since the prognosis of the amount and the 

spatial position of precipitation concerning larger catchment areas, for example, are very 

accurate three days prior to an expected event. Because of the limited local resolution of 

prognosis models, it is very difficult to make forecasts for small catchment areas. 

Before, during and after a flood event, information and communication are crucial for a 

well-functioning warning system. The major principle of the flood warning system of the 

OEBB is the utilisation of existing information. The delivery of expert knowledge by the 

warning centres of the federal countries in Austria and the punctual contact with the 

Hydrographic Services serve as a fundament for taking decisions. Large infrastructure 

operators (18.000 employees at the OEBB Infra) need a well-functioning communication 

system, which includes an early conveyance of information to boot up the emergency 

organisation. Also, it is very important to communicate the timely all-clear. 


The end of a flood event is the beginning of the next event, thus, an accurate event documentation 

is very helpful to improve the warning processes and to optimally plan countermeasures. 

Even though warning systems, as organisational countermeasures, help to protect human 

life and to optimise the disposability of traffic, they are not able to reduce damage on infra- 

structure. 

REFERENCES 

- Rudolf-Miklau F., (2009). Naturgefahren-Management in Österreich 

Vorsorge-Bewältigung – Information, Wien. 

- BMLFUW, (2009). Flood Risk II, Vertiefung und Vernetzung zukunftsweisender 

Umsetzungsstrategien zum integrierten Hochwassermanagement, Wien. 

- TUNSTALL, S. (2006): Damage reducing effects of flood warnings. In: FLOODsite (Hrsg.): 

Guidelines for Socio-economics flood damage evaluation. Wallingford, UK. 



Early Flood Warning for the City of Zurich: Evaluation 

of real-time Operations since 2010 

Katharina Liechti, Dr. 1 ; Matthias Oplatka, Dr. 2 ; Natascha Eisenhut, M. Sc. 2 ; Massimiliano Zappa, Dr. 1 

ABSTRACT 

The flood forecast system for the river Sihl in Zurich (Switzerland) is operated by the Federal 

Research Institute WSL on behalf of the Canton of Zurich since 2007. The close collaboration 

between forecasters and decision makers fosters continuous development of the system. 

A management tool was implemented in the system that allows computing scenarios for 

predicted critical events. The forecast system comprises three forecast chains differing in lead 

time, spatial resolution and update cycle amongst others. These are the two deterministic 

chains driven by the numerical weather predictions COSMO-2 and COSMO-7 and the 

ensemble forecast chain driven by the COSMO-LEPS ensemble. The hydrological forecasts 

resulting from the three model chains and additional helpful information about the conditions 

in the catchment are made available on an online platform. The statistical evaluation 

of the three forecast chains showed that the ensemble forecasts clearly outperform the 

deterministic forecasts and are more reliable especially for taking decisions that need a lead 

time of more than just a few hours. 

KEYWORDS 

flood forecasting; early warning; hydro-meteorological forecast system 

INTRODUCTION 

This contribution presents the flood forecasting system for the river Sihl for the City of Zurich 

and its evaluation over more than five years of operational use. 

In summer 2005 big parts of Switzerland were flooded after 3 days of heavy rain. The event 

caused damages of more than three billion Swiss Francs. While many places experienced the 

worst flood damages recorded, Zurich City stayed relatively dry. However, the city centre of 

Zurich has a high damage potential, which is estimated to about five billion Swiss Francs. 

A lot of infrastructure had been constructed on the alluvial fan of the river Sihl during the 

last century. A closer look at the event of 2005 showed that if the centre of precipitation 

would have been over the Sihl catchment, the city centre, including Zurich central railway 

station would have been flooded. One of the main problems in Zurich is that the river Sihl 

crosses the central railway station. The riverbed of the Sihl is embraced by the underground 

tracks and the ground level tracks of the railway station, limiting its capacity to estimated 

350m 3 /s, corresponding roughly to a hundred year flood (FOEN, 2014). So a forecast system 

for the river Sihl is needed to be able to take prevention measures in case of expected flood 

1 Eidg. Forschungsanstalt WSL, Birmensdorf, SWITZERLAND, kaethi.liechti@wsl.ch 

2 Kanton Zürich Baudirektion Amt für Abfall, Wasser, Energie und Luft 



events. The most important measure that could be taken to prevent a flood event in Zurich is 

to drawdown the Sihl lake, which is a reservoir lake for hydropower production and can act 

as a retention basin for about 46% of the catchment area (Figure 1). This action needs a lead 

time of about 1 to 3 days. 

The forecasting system developed for this purpose is run operationally since 2007 by the 

Swiss Federal Institute for Forest, Snow and Landscape Research WSL on behalf of the 

Canton of Zurich (Office for Waste, Water, Energy and Air) (Addor et al. 2011; Zappa et al. 

2010). In this contribution the system set up, a statistical evaluation and experiences with 

the system in operational use are presented. 

Figure 1: Sihl catchment. The upper part of the catchment coloured in orange (46 %) belong to the accumulation area of the reservoir 

lake Sihlsee. The remaining catchment area consists of the two tributaries Alp and Biber and the narrow Sihl valley between the lake 

and Zurich City (source Addor et al. 2011). 

METHODS 

The presented flood forecasting system consists of a meteorological and a hydrological part 

(Zappa et al. 2010). Three different meteorological model forecasts are used to drive the 

discharge forecasts. They differ in their spatial resolution, lead time and update cycle (Table 

1). The models COSMO-2 and COSMO-7 are deterministic models, which forecast for each 

time step one single forecast value. COSMO-LEPS on the other hand is a probabilistic model, 

which consists of 16 ensemble members. So for each time step COSMO-LEPS forecasts 16 

equally likely forecast values (Montani et al. 2011). 


Table 1: Numerical weather prediction models used for driving the hydrological model PREVAH. (* After computation and dissemination 

of COSMO-LEPS 120 h lead time are left in the hydrological forecast). 

Model Horizontal resolution Initialisation Lead time Member 

COSMO-2 2.2 km 00, 03, 06, ... UTC 24h 1 

COSMO-7 6.6 km 00, 06, 12 UTC 72h 1 

COSMO-LEPS 7 km 12 UTC 132h* 16 

These meteorological forecasts are fed into the semi-distributed hydrological model PREVAH 

(Viviroli et al. 2009) and result together with the current state of the catchment in discharge 

forecasts for the river Sihl. The discharge forecasts have a temporal resolution of one hour 

and a lead time and update cycle according to the driving meteorological models. 

These forecasts are made available on an online platform to the decision makers of the 

Canton of Zurich (Badoux et al. 2010). The platform includes not only the discharge forecasts 

for the river Sihl (Figure 2), but also forecasts of the level of the Sihl lake and other meteorological 

and hydrological parameters that help judging the current and expected situation. 

Furthermore a management tool allows the realization of discharge scenarios in a pre-event 

phase. The most valuable element of this flood forecasting system is however the good and 

close collaboration between researchers and decision makers. Only good communication 

Figure 2: Screenshot of the online platform showing the COSMO-LEPS forecast for the river Sihl in Zurich for the period of May 31st to 

June 4th 2013. The coloured background of the available forecast locations correspond to the highest warning level reached in the 

forecast period. The gray panel on the left hand side offers the opportunity to easily switch between the different forecast products and 

other additional information like measurements, radar data, meteograms amongst others. 


makes it possible to operate and further develop a system that meets the user need and can 

unfold its full potential. 

Statistical Evaluation 

Since the last model update in late 2009, forced by the increase of horizontal resolution of the 

meteorological models, more than five years of continuous forecasts from the three different 

forecast chains are available for the evaluation of the hydrological forecast system. 

The evaluations include data from the months March to October of the years 2010 to 2014. 

The reason for this restriction is, that most flood events in the river Sihl occur during these 

months due to snowmelt, thunderstorms and long lasting precipitation events. Also during 

these months forecasters and end users are on standby duty. 

The performance of the three model chains were compared to each other by means of several 

statistical scores. Here the evaluation with the coefficient of determination (R 2 ) and the Brier 

Skill Score (BSS) is presented. These scores are a mix of easy understandable measures of 

agreement (R 2 ) and well-tailored advanced metrics able to provide hints on the quality of 

both deterministic and probabilistic forecast systems (Addor et al. 2011; Liechti et al. 2013). 

The coefficient of determination R 2 is calculated using the observed and forecast daily mean 

runoff values. It is the squared Pearson correlation coefficient, which for its part is a measure 

for the linear correlation between forecast and observation. R 2 is a deterministic measure, 

therefore for the ensemble forecasts the daily mean value of the ensemble mean is used. R 2 

indicates the percentage of observed variance that can be explained by the forecast. So the 

closer R 2 is to one, the better the forecast explains the observation. In the presented case the 

evaluation is done by lead time, days 1 to 5 for COSMO-LEPS mean, day 1 to 3 for COSMO-7 

and day 1 for COSMO-2. 

The Brier Skill Score (BSS) is a measure which allows direct comparison of deterministic and 

probabilistic forecasts without the previous reduction of the ensemble to its median or mean. 

The BSS is based on the Brier Score (BS) which is the mean squared error of the probability 

forecast to exceed a predefined threshold given the observed outcome (exceeding/not exceeding 

the threshold). The BSS then describes, for the predefined threshold, how much better or 

worse the BS of the forecast is compared to the BS of the climatological forecast. This makes it 

a good measure to show the added value of the forecast system compared to the climatological 

‘guess’. For the evaluation with the BSS the hourly time series of forecasts and observations 

were aggregated with a centred running maxima of 13 hours. This gives the forecast 

system a tolerance of plus/minus 6 hours. Equations and closer descriptions of the presented 

scores can be found in Wilks (2006). 

Management tool 

During the operational use of the presented forecast system it came clear that for the management 

of individual events a tool was desirable that allows to simulate scenarios. Therefore a 


management tool was built, which allows computing scenarios by varying the inflow into the 

Sihl lake, draw down of the Sihl lake, used turbine capacity of the hydro power plant and the 

contribution from the tributaries not flowing into the reservoir lake first. The time steps over 

which these parameters should be applied in the calculation of the scenario can be freely 

chosen by the user of the tool. The tool is accessible online to all eligible users. 

RESULTS 

Coefficient of determination R 2 

As to be expected the coefficient of determination R 2 between forecast and observation 

decrease with increasing lead time. Table 2 lists the R 2 values for the three forecast chains by 

lead time for the individual years and the entire period. For the first forecast day both 

deterministic forecast chains reach mainly higher R 2 than the ensemble forecast. On forecast 

days two and three the ensemble forecast driven by COSMO-LEPS reach higher R 2 than the 

deterministic COSMO-7 (except day two 2012). Even the fourth forecast day of the ensemble 

forecast still reaches higher R 2 than the third forecast day of the deterministic COSMO-7. 

Table 2: R 2 of forecast observation pairs of daily mean runoff, listed by lead time for each year from 2010 to 2014 and for all years 

together. Only data from the months March to October are used. 

2010 2011 2012 2013 2014 2010-2014 

COSMO-LEPS Day 1 0.79 0.87 0.81 0.86 0.84 0.83 

Day 2 0.65 0.49 0.58 0.75 0.52 0.62 

Day 3 0.38 0.43 0.54 0.70 0.52 0.52 

Day 4 0.28 0.41 0.49 0.64 0.40 0.44 

Day 5 0.22 0.15 0.23 0.53 0.30 0.29 

COSMO -7 Day 1 0.85 0.87 0.86 0.88 0.83 0.86 

Day 2 0.62 0.39 0.66 0.69 0.45 0.58 

Day 3 0.48 0.25 0.46 0.58 0.25 0.42 

COSMO -2 Day 1 0.83 0.83 0.87 0.91 0.81 0.86 

Brier Skill Score 

The BSS was calculated for several thresholds. Shown here are the results for the thresholds 

corresponding to the 90-% and 95-% quantile of the discharge climatology (March to 

October 2007 to 2014, centred running maxima over 13 hours). In addition to the three 

forecast chains also the ensemble median is evaluated and treated as a deterministic forecast. 


Generally it can be seen that the BSS decreases with increasing threshold and with increasing 

lead time (Figure 3). COSMO-2 is better or equally good as COSMO-7 over its entire runtime. 

For the lower threshold tested COSMO-LEPS median and COSMO-7 are in the same range. 

For the higher threshold tested COSMO-LEPS median is performing a bit better than 

COSMO-7. The COSMO-LEPS ensemble forecast is clearly the best performing forecast chain 

and reaches positive BSS values over its entire forecast period of five days. 

Figure 3: Brier Skill Score. Thresholds 19.5 m 3 /s and 34.75 m 3 /s correspond to the 90% and 95% quantiles of the discharge climatology 

(centerd 13h running maxima, March to October 2007 to 2014). 

Management tool 

The Management tool is illustrated using an event from 2013. The COSMO-LEPS forecast of 

May 31st predicted high discharge, i.e. four ensemble members exceeding the third warning 

level of 200 m 3 /s, for June 1st and 2nd which would fill the reservoir lake (Figure 4). 

The end-users then informed the forecasters that they plan to draw down the reservoir lake 

by 80 m 3 /s over the next 36 hours. 

According to the available forecasts at that time this drawdown would have overlapped with 

discharge peaks from the tributaries Alp and Biber. The forecasters therefore used the 

management tool to calculate the expected scenario according to the end-user’s plan and 

according to a counter proposal from the forecasters (Figure 4). The resulting scenarios made 

clear that with the planned drawdown of 80 m 3 /s over the next 36 hours the probability to 

exceed the third warning level of 200 m 3 /s was quite high (Figure 4, scenario 1). Thus, after a 


short teleconference with the forecasters to discuss the scenarios and their implications, the 

end-users decided to shorten the drawdown to a few hours only reducing the probability of 

exceeding 200 m 3 /s in Zurich significantly (Figure 4, scenario 2). 

The observed event then consisted of four peaks distributed over three days. A first artificial 

peak from the ordered draw down reached 130 m 3 /s, the second peak originating from the 

tributaries reached 190 m 3 /s, more intensive rain lead to a peak of 140 m 3 /s, and the last peak 

originated from the combination of a rule based draw down and more intensive precipitation 

reached 160 m 3 /s. 

Sisee ee init 21 21 

Si ric init 21 21 

Forecast 

WL3 

WL2 

WL1 

discharge [m 3/s] 

WL3 

WL2 

WL1 

Fr 31. May Sat 1. Jun Sun 2. Jun Mon 3. Jun Tue 4. Jun 


hours 

hours 

Sisee otet 

discharge [m 3/s] lake level [m a.s.l.] 

Si ric 

Scenario 1 

WL2 

WL1 

an eers 

ecee 2 s 

WL3 

WL2 

WL1 



hours 

hours 

Sisee otet 

discharge [m 3/s] discharge [m 3/s] 

Si ric 

Scenario 2 

discharge [m 3/s] 

WL3 

WL2 

WL1 

n to eers 

ecee 2 s 

WL3 

WL2 

WL1 


hours 


Figure 4: Example of management tool application. Top row: COSMO-LEPS forecasts for the lake level (left) and the Sihl in Zurich (right). 

Middle row: Scenario 1 releasing 80 m 3 /s over 36 hours from the lake and corresponding expected peak flow in Zurich. Bottom row: 

Scenario 2 releasing 80 m 3 /s over 9 hours and corresponding expected peak flow in Zurich. 

hours 

CONCLUSIONS 

The statistical evaluation shows that for a system like the river Sihl upstream of Zurich, which 

needs a lead time of one to three days to take preventive measures in case of a coming event, 

the ensemble forecasts are more reliable than the deterministic forecasts. 


The management tool proved to be useful to handle individual critical events. For the event 

presented here, the tool helped to plan the drawdown of the reservoir lake such that a 

coincidence of peak flows from tributaries and from the lake outlet were avoided. The tool 

is not perfect yet and needs further improvement to ease its handling. 

The near future will bring significant changes to the system. MeteoSwiss will produce new 

numerical weather predictions and replace the three forecast products with only one 

deterministic forecast product called COSMO-1, and the ensemble forecast product 

COSMO-E. The spatial resolution of these products will be 1.1 km and 2.2 km and the 

number of ensemble members will increase to 21. Furthermore it is planned to implement 

post processing into the operational forecast system. Both developments are expected to 

further improve the discharge forecasts for the river Sihl. 

REFERENCES 

- Addor N., Jaun S., Zappa M. (2011). An operational hydrological ensemble prediction 

system for the city of Zurich (Switzerland): skill, case studies and scenarios. Hydrology and 

Earth System Sciences 15(7): 2327-2347. 

- Badoux A., Zappa M., Oplatka M., Bösch M., Jaun S., Steiner P., Hegg C., Rhyner J. (2010). 

IFKIS-Hydro Sihl: Beratung und Alarmorganisation während des Baus der Durchmesserlinie 

beim Hauptbahnhof Zürich. Wasser Energ. Luft 102(4): 309-320. 

- FOEN, Federal Office for the Environment. (2014). http://www.hydrodaten.admin.ch/lhg/ 

sdi/hq _studien/hq_statistics/2176hq.pdf. Access: 06.01.2016. 

- Liechti K., Zappa M., Fundel F., Germann U. (2013). Probabilistic evaluation of ensemble 

discharge nowcasts in two nested Alpine basins prone to flash floods. Hydrological Processes 

27(1): 5-17. 

- Montani A., Cesari D., Marsigli C., Paccagnella T. (2011). Seven years of activity in the field 

of mesoscale ensemble forecasting by the COSMO-LEPS system: main achievements and open 

challenges. Tellus Series a-Dynamic Meteorology and Oceanography 63(3): 605-624. 

- Viviroli D., Zappa M., Gurtz J., Weingartner R. (2009). An introduction to the hydrological 

modelling system PREVAH and its pre- and post-processing-tools. Environmental Modelling & 

Software 24(10): 1209-1222. 

- Wilks D. S. (2006). Statistical methods in the atmospheric sciences. Amsterdam, Elsevier. 

- Zappa M., Jaun S., Badoux A., Schwanbeck J., Addor N., Liechti K., Roeser I., Walser A., 

Viviroli D., Vogt S., Gerber M., Trösch J., Weingartner R., Oplatka M., Bezzola G.R., Rhyner 

J.(2010). IFKIS-Hydro Sihl: Ein operationelles Hochwasservorhersagesystem für die Stadt 

Zürich und das Sihltal." Wasser Energ. Luft 102(3): 238-248. 



A new data management infrastructure for improved 

analysis and real-time publication of flood events in 

the Canton of Aargau 

Christophe Lienert 1 

ABSTRACT 

In order to take appropriate countermeasures before and during flood events, decision 

makers, crisis committees, and the public need fast, secure and easy access to real-time hydrological 

data. In order to meet the increased needs and expectations of internal and external 

parties, new Desktop and Web-tools for real-time data analysis, data collation, interoperability, 

and interactive visualization have been introduced, and further developed by the Swiss 

Canton of Aargau. Customizable tools for hydrological data analysis and cartographic 

publication are now available, tailored to early warning and flood management. Various 

features of the newly launched data infrastructure, and the real-time map HydroWeb, are 

highlighted. 

KEYWORDS 

early warning; flood management; hydrology; real-time cartography; data interoperability 

INTRODUCTION 

In the Swiss Canton of Aargau, flood hydrology, emergency management and early warning 

is a high priority as four of the five largest rivers of Switzerland join on its territory. Superficial 

waters of about two thirds of Switzerland's area, plus some parts of Germany and Austria, 

flow through the Canton and are discharged by the Rhine. Floods of different severity have 

occurred in the past, and will occur in the future. Timely and precise information about the 

state of the waters, as well as forecasts and runoff predictions provided by Swiss Federal 

agencies, and coordination with upstream Cantons are vital to prepare for, counteract and 

coping with floods. 

In 2012, the Swiss Canton of Aargau has launched the two-years project "Introduction 

WISKI", during which a commercial water information system software was evaluated, 

tested, purchased, and implemented. The goal was to replace the previous software for better 

management of large archive and real-time hydrological datasets, for enhanced hydrological 

and GIS-based analysis, and for better interface management with Web-based visualization 

and publication products. The project contained the introduction of a WISKI-Desktop 

component and a WISKI-Web component (called HydroWeb). The overall project consisted of 

the following working packages and deliverables: 

1 Canton of Aargau, Aarau, SWITZERLAND, christophe.lienert@ag.ch 



– Securing current, parallel data management operations 

– Development of a comprehensive, extendable data model to grant role-based access. 

– Conversion of telephone-based and FTP-based data retrieval. 

– Processing and migration of legacy data, stage-discharge relationships, authoring of new 

quality concepts and time series models. 

– Realization of new manual and automated publications routines, particularly the generation 

of approved and provisional hydrological yearbook sheets and special reports 

– Linking to desktop and Web-based geo-information systems (GIS) for spatio-temporal 

analysis and cartographic visualization. 

– Adjustable interface management for cloud-based data exchange by means of standardized 

data interoperability tools. 

– Complete re-design and implementation of the Web-based hydrological map publication 

(browser and mobile enabled) within the existing Cantonal IT-environment. 

– Internal training of staff after rollout. 

– Provision of a location-independent Intranet-based access to real-time data and analysis 

tools for cantonal flood management staff (flood on-call service). 

The decision of the Canton of Aargau fell on the software WISKI which is marketed by the 

German company Kisters Inc., Aachen, Germany. One of the world's market leaders in the 

domain of information systems for water resources management, the company now also 

specializes in Web-based services and application programmable interfaces (API), reflected by 

the entrance of the product Kisters Web Interoperability Solution (KiWIS) in the company's 

portfolio. The company also actively participates in the development of Open Geospatial 

Consortium's WaterML and TimeSeriesML standard, ensuring that hydrological data comply 

with international standards (Horsburgh et al. 2009, Yu et al. 2015). 

KiWIS allows the Canton of Aargau to tap the full potential of their newly set-up large, 

hydrological database WISKI. It offers tailored data in real-time to either end-users (e.g., flood 

committees, public users), or third party applications (e.g., HydroWeb). WISKI acts as a 

comprehensive (thick) internal Desktop application for complex hydrological data analysis. 

Products based on KiWIS are Web-based (thin) and therefore rather trimmed to the relevant 

information, ideal for public users or experts who need to monitor hydrological situations at a 

glance, before initiating more detailed analysis. 

METHODS 

Linking real-time measurement data with GIS-data 

WISKI-Desktop is equipped with GIS interfaces and an additional extension ensures that 

WISKI may be used in GIS or GIS may be used in WISKI. The goal of the project component 

HydroWeb, however, was to leverage KiWIS for (carto)graphic visualization of hydrological 

real-time data on the Web. The main question was how to handle, and combine huge 

amounts of real-time data with the cantonal spatial data infrastructure and its Web-based 


publication utility AGISviewer (Lienert & Meier 2014, Vitolo et al. 2015), since this was one 

of the main requirements of HydroWeb. The AGISviewer framework had, for the first time, to 

meet the requirement of handling real-time data. So far, predominantly static, pre-processed 

Cantonal geo-information products are published in this framework. The choice to realize 

HydroWeb in the AGISviewer framework, as opposed to an entirely external framework, was 

based on the arguments that measurement data stay in-house and many existing functionalities 

and services are reusable. 

Integrating various data sources 

In HydroWeb, two main distinctions are made as to data sources and measurement network: 

1) data sources owned by Canton of Aargau 2) third-party data from other public authorities, 

such as several Federal Offices, other neighboring Cantons, German State Baden-Wuertemberg 

(neighboring the Canton of Aargau across the river Rhine), research institutes, and 

private measurement network operators. Further distinctions are made between a) real-time 

versus archive, b) enduring versus temporary time series, c) original (non-editable) versus 

productive (editable) time series, d) persistent versus on-the-fly calculated time series. 

The following measuring parameters are stored in the Cantonal WISKI database: precipitation, 

soil moisture, soil temperature, water temperature, air temperature, groundwater 

height, pumping volumes, river discharge, various drinking water quality parameters, and 

various bathwater quality parameters. Further environmental parameters will be integrated 

in the near future (Lienert & Meier, 2014). 

Basic data model 

On top of the vendor data model, a cascade of several steps describes and determines the 

basics of the overall WISKI data model operated by the Canton of Aargau: 

– data owner (e.g., Canton of Aargau = AG) 

– organization (e.g. Division Landscape and Waters = ALG) 

– gauging site (e.g., City of Rheinfelden, and gauge number 0374) 

– gauge type (e.g. flowing waters = FG, or meteorology = METEO, etc.) 

– measuring parameter (e.g., water discharge = Q) 

– time series (e.g., original series with raw data = Cmd.O) 

– agents (e.g., agents to import raw, unverified data from some source) 

Each of the above mentioned steps has up to a dozen further attributes attached to them, and 

further interrelations. For example, on each gauging site level, information is stored regarding 

identification, location (geo-coordinates), and flood statistics. 

The Web-Service interface KiWIS 

Technically speaking, KiWIS is Web-based application programming interface (API) of WISKI, 

based on the Representational State Transfer (REST). Using KiWIS, the Canton of Aargau has 

established a basis to considerably facilitate online collation and publication of hydrological 

data in real time, to support flood management and other activities. In summary, KiWIS: 


– allows for the publication of large amounts of time series data 

– supports authorities to perform their legal data publication tasks 

– helps foster cooperation with third-parties 

– distributes specific data and information to assigned user groups 

– enables customers build customized Web-applications and data-hubs, in order to use data 

across applications 

Data requests from external customers may now be conveniently answered by URLs contain 

all desired information. Other data services were deployed to deliver real-time data to other 

program components. One component is the real-time list; another component is the 

real-time map (HydroWeb). Both components are discussed in the next chapter. 

RESULTS 

Data analysis and publication using real-time lists 

The most current data are accessible through either maps or list. The list shown in Fig. 1 

is optimized for mobile devices and complies with cantonal corporate design guidelines. 

As maps are not yet accessible via mobile devices, mobile users are automatically redirected 

to the list. 

Figure 1: Real-time data list optimized for mobile devices. River discharge, water levels, water temperatures or precipitation are 

available. Colored data rows indicate the exceedance of pre-defined warning levels. 


The list is derived from a tailored KiWIS service and is further processed on the server for 

better readability. The user may choose one of the measured parameters (river discharge, 

water level, water temperatures, precipitation) using drop down functionality. The list is 

immediately reloaded when the parameter is changed, or when the browser is refreshed. 

For river discharge, the current values are related to different warning levels. Depending on 

whether and what warning levels are reached, data rows are automatically colored. Warning 

levels and its coloring comply with Swiss national flood management standards and ensure 

quick looks. The real-time data list is accessible at www.ag.ch/hydrometrie/liste. 

Figure 2: Real-time map HydroWeb on the national scale. Real-time data from neighboring areas and upstream Cantons are also 

integrated. Base map: Topographic landscape model TLM by Swiss Federal Office of Topography, swisstopo (provided as Webmap-service). 

Data analysis and publication using real-time maps 

In Fig. 2, the real-time hydrological map with the territory of the Canton of Aargau (dashed 

border) is shown. Tripartite circle map symbols depict the measured parameters river 

discharge (dark blue), water levels (green), water temperature (red), and full blue circle 

symbols represent precipitation gauges. 

Depending on the chosen map scale, specific interactivity and responses are provided on 

HydroWeb. Clicking on the map symbols in the legend window on the right either activates 

or deactivates onMouseOver-functionality. Depending on the scale, gauge information is 

either provided when the mouse is moved over the symbol, or attached to the symbol and 

therefore visible as a flag. 

As shown in the map in Fig. 2, several other gauges are visualized outside the cantonal 

territory. Joint measuring networks (e.g., the precipitation network operations between 


Cantons Aargau and Lucerne), as well as real-time data exchange agreements with partner 

networks allow to publish more than just cantonal data. Key gauges of upstream Cantons are 

therefore visualized jointly on HydroWeb, helping to get the bigger picture when assessing 

and analyzing floods. 

Space conflicts and cluttering of map symbols on larger overview map scales have been 

addressed by a scale-dependent filter, giving one gauge location priority over the other. The 

priority of a gauge was determined expert-based, depending on its overall significance for 

flood management, measuring network affiliation, measuring quality, and representativity. 

The prioritization is controllable using a configuration file. 

Data widgets, as well as interactive search, and filter functionality are an integral part of 

HydroWeb. The functionalities are shown in Fig. 3 on the right side. Users may search for 

gauge locations or numbers, and names of municipalities. Likewise, there are filter options 

that automatically remove gauges on the map. The following filter criteria are available: 

– data ownership 

– catchment area 

– measuring parameter 

– date of measurement 

Figure 3: Real-time map HydroWeb on the catchment scale. The data widget contains hourly precipitation data. Time series may be 

viewed on different temporal aggregation levels (10min, 1h, 1d), and different retrospect (2d, 7d, and 31d). 

The map in Fig. 4 contains small flags next to the gauge, some of which with a yellow color, 

as these have reached a specific warning level. When clicking on such a gauge, the widget 

shows the time series graph, plus the yellow warning level. The widgets are movable and 

resizable, and each displays uniform metadata on their left side, and time series information 


on the right side. On the right side, information is organized by tabs and contains the 

following items: 

– discharge with time series graphs 

– water level with time series graph 

– water temperature with time series graph 

– precipitation with bar graphs, cumulative curves 

– statistics with extreme values (river discharge) 

– statistics with intensity, sum and event analysis (for precipitation) 

– annuals with verified, authorized data of the past year(s) 

– gauge images 

As shown in Fig. 1, the entire list of available gauges may also be retrieved in HydroWeb, 

displayed in Fig. 4 on the right side. The list may be sorted as to station name, parameter 

name, measuring value, or time of last measurement. Clicking on a list item releases an 

automatic zoom to the chosen gauge. HydroWeb is accessible at www.ag.ch/hydrometrie/ 

karte. 

Figure 4: Real-time map HydroWeb on a regional scale. Yellow value flags next gauge symbols indicate moderate danger. In the widget, 

the graph is visualized jointly with the warning threshold. 

FURTHER DELIVERABLES 

HydroWeb is supposed to be an "anytime-anywhere" tool for flood management and early 

warning. It is available in all major browsers and does not require any plug-ins. comprehensive 

load tests were carried out before rollout and findings showed that the system is capable 

of handling 5000+ concurrent visitors without noticeable compromises. In order to calculate 

and analyze visitor statistics, especially during flood events, the well-known Google Analytics 

tool has been integrated. 


Each system component (servers, services, databases, middleware) of the entire WISKI- 

Architecture is regularly checked by another server for its availability and functionality. 

A dash board specifically built for this purpose allows for quick views and performance checks. 

In case some component is (repeatedly) underperforming, emails are dispatched to cantonal 

IT-staff. 

CONCLUSION AND OUTLOOK 

Due to extended technical and organizational requirements, the Swiss Canton of Aargau 

has introduced new measurement data software, in order to meet the increased needs and 

expectations of internal and external parties during floods. In order to take appropriate 

countermeasures before and during flood events, decision makers, crisis committees, and 

specialists need rapid, secure and easy access to real-time discharge and precipitation data. 

In September 2014, the Canton of Aargau has introduced the new Desktop-Software WISKI 

and its Web-based module KiWIS, facilitating data exchange in standardized format, and 

timely processing and publication of hydrological data, relevant to flood management and 

early warning. The discussed hydrological real-time map HydroWeb constitutes a "front-end" 

of these real-time data workflows. It allows for cartographic access to multi-source, multiparameter, 

and multi-format data in a harmonized, interactive, and attractive way. 

It may be assumed that WISKI will be established at several Cantons and Federal agencies in 

Switzerland as a tool for hydrological (and other environmental) data management. With this 

present contribution, experiences and deliverables of the successful introductory project 

WISKI are shared. Further developments comprise the integration of real-time raster data 

sets. Presently, a project is carried out that deals with the integration and visualization of 

raster-based precipitation distribution data in WISKI, and its interface through KiWIS for 

interactive visualization in HydroWeb. 

LITERATURE 

- Horsburgh J., Tarboton D., Piasecki M., Maidment D., Zaslavsky I. (2009). An integrated 

system for publishing environmental data. Environmental Modelling & Software 24(8): 

879-888. 

- Lienert C., Meier S. (2014). Environmental Data Visualization EnVIS – Linking real-time 

sensor data with spatial data Infrastructures for Web-based visualization. In: Bandarova T., 

Konecny M., Zlatanova S. (Eds.). Thematic Cartography for the Society. Berlin: Springer, 

293-304. 

- Vitolo C., Elkhatib Y., Reusser D., Macleod C., Buytaert, W. (2015). Web Technologies for 

Environmental Big Data. Environmental Modelling & Software 63: 185–198. 

- Yu J., Taylor P., Cox S., Walker G. (2015). Validating observation data in WaterML 2.0. 

Computers & Geosciences 82: 98-110. 



Radar-based Warning and Alarm Systems for Alpine 

Mass Movements 

Lorenz Meier, Dr. 1 ; Mylène Jacquemart 1 ; Bernhard Blattmann 1 ; Sam Wyssen 2 ; Bernhard Arnold 3 ; Martin Funk, Prof. Dr. 4 

ABSTRACT 

We present two applications of Doppler radar and Ground-Based Interferometric Synthetic 

Aperture Radar (GBInSAR) in warning and alarm systems for alpine mass movements. 

Near Zermatt, two long-range Doppler radars are used to detect snow avalanches and close 

the road below in real time. Local authorities have access to the data at all times, and can 

reopen the road from any computer or mobile device. 

In the Saas Valley, ice avalanches from the glaciated northwest face of Weissmies Mountain 

pose a threat to infrastructure and people below. We use GBInSAR to continuously monitor 

ice surface velocities on this face to detect and warn of potential ice fall. 

Radar technologies have proven reliable and versatile in warning and alarm systems because 

they can monitor large areas without the need for in-situ installations and are largely 

unaffected by weather. 

KEYWORDS 

radar; mass movements; early warning; alarm system; road safety 

INTRODUCTION 

Alpine mass movements like snow and ice avalanches, rockfall and landslides pose an inherent 

threat to villages, roads or ski resorts. In many cases, structural measures can provide 

long-term safety. In some cases, however, these measures are too costly, negatively impact 

scenery or cannot be constructed where needed. In these situations, an electronic warning or 

alarm system, paired with organizational measures, can significantly reduce risk of injury, 

fatality and damage to property. Electronic alarm systems measure an event in real-time 

(e.g. the avalanche velocity or flow height) and trigger an automatic response like closing a 

road or stopping a train before the event reaches the threatened area. Warning systems 

measure precursory events (e.g. the acceleration of surface movements of a rock wall that 

announces possible rockfall) and provide the data to local authorities and experts as a basis 

for decision-making (Sättele & Bründl, 2015). 

Where roads, railways or ski resorts are threatened by avalanches, preventive avalanche 

release using explosives is a common mitigation method (Gubler & Wyssen, 2002). However, 

the result of blastings must be verified before reopening the threatened area. Since this can 

1 GEOPRAEVENT AG, Zürich, SWITZERLAND, lorenz.meier@geopraevent.ch 

2 Wyssen Avalanche Control AG 

3 Gemeinde Zermatt 

4 ETH Zürich 



e difficult under darkness, fog or heavy snow fall, sensors without visibility requirement are 

a strong asset (Meier et al. 2010, Scott et al. 2006). 

Contrary to snow avalanches, ice avalanches are predictable in some circumstances. A sharp 

increase in surface velocities at the unstable margin of a glacier is a clear sign of an impending 

icefall. The rupture time can be predicted based on a time series of surface velocity data. 

(Faillettaz et al., 2015). The most recent glacier catastrophe in Switzerland occurred in 1965, 

when two million m³ of ice broke off from the tongue of Allalingletscher and buried the 

construction site of the Mattmark dam, killing 88 people (Vivian, 1966). 

Doppler radars, while best known as speed traps used by the police, have been applied to 

measure fast processes like avalanches (Lussi et al., 2012), and are widely used to monitor 

eruption dynamics in volcanology (e.g. Seyfried & Hort, 1999). Ground Based Interferometric 

Synthetic Aperture Radar (GBInSAR) has been used to study a variety of slower processes, 

from landslides, (rock) glacier motion, and tidewater glacier dynamics to displacements in 

man-made structures (Luzi et al., 2007; Monserrat et al., 2014; Nolesini et al., 2013; 

Rödelsperger 2011). 

We report on two radar applications where these technologies have been integrated into 

operational alarm and warning systems: The first detects avalanches within an area of 

roughly two square kilometer using Doppler radars. We show that they can significantly 

improve avalanche release verification and be used in an automated way to close and reopen 

roads to traffic in real time. The second uses GBInSAR to continuously monitor the surface 

flow field of a glacier, data that can be analyzed to forecast icefall events. 

METHOD 

Radar provides several advantages over other surveying techniques. First, it does not require 

sensors to be installed within the observed area, keeping both people and instrumentation 

out of harm's way. Second, a single radar is sufficient to monitor areas spanning several 

square kilometers. Third, in contrast to optical methods, radar measurements are largely 

unaffected by low visibility. The atmosphere is mostly transparent to radar radiation and 

measurements are even possible during rain, snowfall or fog. 

For both types of radar systems described here, data quality is crucial for reliable operation. 

If an alarm system is not sufficiently reliable or accurate, it may miss an event. Requirements 

for warning systems are slightly less strict, depending on the expected delay between 

precursor and event. 

DOPPLER RADAR 

Radar transmits electromagnetic radiation, usually in the range of a few to a few tens of GHz. 

This radiation is reflected by objects in the target area. If such an object is moving towards 

or away from the radar, the frequency of the reflected radiation is offset from the original 

radar frequency. Known as the Doppler Effect, objects moving toward the radar reflect a 

higher frequency, while objects moving away from the radar reflect a lower frequency. 


This phenomenon is commonly experienced in daily life: the sound of an approaching 

ambulance changes its pitch abruptly when it passes the observer. An object’s velocity and 

direction relative to the radar can thus be calculated by comparing the transmitted and 

reflected signal frequencies (Figure 1A). 

Gravitational mass movements like avalanches, rockfall and debris flows all move with 

velocities of a few meters to a few tens of meters per second and can easily be measured using 

Doppler radars. However, Doppler radars are sensitive to all movements within their target 

area. Thus, additional criteria are needed to distinguish dangerous movements from “harmless” 

objects in motion, like helicopters or wildlife. We have conducted an extensive field 

study at 8 locations in Switzerland with several Doppler radars that measure velocity, range 

and direction to moving targets within a field of view of 90° (horizontal) × 10° (vertical) and 

a range of up to 2’000 m. We have acquired more than 10 terabytes (TB) of radar raw data 

that contain radar signals of approximately 200 avalanches, 100 rockfall events and 1 debris 

flow. These data were used to develop an algorithm that reliably detects snow avalanches and 

rockfall events in real time. The algorithm determines avalanche location and speed, and 

estimates its size. It can be configured to meet local needs. For example, if there are several 

gullies in the radar field of view, the software can be configured to only trigger an alarm if an 

avalanche of a certain minimum size is detected in one or more specific gullies. 

So far, four Doppler radars of this kind have been deployed in operational projects: two in 

Zermatt (see below), one for avalanche safety of a mountain railway, and one mounted on 

a Wyssen avalanche tower for the verification of artificially released avalanches. 

Figure 1A: Doppler radars calculate velocity and direction based on the shift in received frequency. 1B: GBInSAR systems measure 

deformation based on the phase shift of the received signal. 


GBInSAR 

Ground-based SAR interferometry is widely used to monitor slope movements (e.g. Monserrat 

et al., 2014; Nolesini et al., 2013), but its application to glacier motion in an operational 

warning system is a novelty. Similar to Doppler radar, it transmits radar radiation and 

analyzes the signal reflected by the target area. In contrast however, GBInSAR systems 

analyze changes in the phase of the reflected radiation instead of changes to the frequency. 

Therefore, they can measure displacements as small as 0.1 mm between two acquisitions. 

The IBIS-FL radar system used in the present study works with a wavelength of 17 mm. The 

distance to a target is an arbitrary number of full wavelengths, plus a remaining fraction of a 

wavelength of at most 17 mm. If a target moves between one measurement and the next, this 

fraction will change, recorded as the phase shift between two measurements (Figure 1B). 

An important limitation of this principle is that displacements larger than a quarter of the 

system’s wavelength, i.e. ± 4 mm, cannot be resolved unambiguously. Furthermore, only the 

line-of-sight component of the displacement can be measured, and the atmosphere can cause 

delays that need to be corrected later. For an acquisition interval of 1 minute, the maximum 

velocity that can be detected is about 6 m per day. Mathematical methods (spatial phase 

unwrapping) can extend this limit further. 

The area of coverage depends on the local terrain and the type of antenna used. The field 

of view of the IBIS-FL radar is roughly 60° (horizontal) × 30° (vertical) with a range up to 

4’000 m. The resolution, or ‘pixel size’, is 0.75 m in the along-range direction and 4.4 mrad 

in the cross-range direction (i.e. 4.4 m at a range of 1’000 m). Details regarding the GBInSAR 

measurements can be found in Rödelsperger 2011. 

STUDY SITES 

Both study sites are situated in the southern Swiss Alps in the Canton of Valais, and were 

established upon request of the local authorities. 

SITE 1: ZERMATT 

Zermatt is one of the most famous mountaineering and skiing destinations in Switzerland, 

with a year-round population of roughly 6’000 and a yearly overnight stay count of about 

2 million. The cantonal road between Täsch and Zermatt is the only access road to Zermatt. 

While most tourists travel to and from Zermatt by train, residents and commercial vehicles 

rely heavily on the avalanche-threatened road. Avalanches are released artificially by 

helicopter blasting when conditions permit. As an additional safety measure, we installed 

Doppler radar alarm systems on the opposite side of the valley to monitor three avalanche 

channels from a distance of 800-1800 m (Figure 2A). A first Doppler radar was installed in 

January 2015, while a second was added in December 2015 along with five traffic lights 

and four road barriers 

The data from the Doppler radars are analyzed in real time. If our algorithm determines an 

avalanche is occurring, the traffic lights are immediately turned to red and the barriers are 


lowered to block access to the threatened sections of the road. In addition, the alarm system 

turns a traffic light 3 km north in Täsch to red to prevent any additional traffic from traveling 

up the road. 

The road lies at approximately 1’600 m altitude (relative to sea level). Radar 1 watches for 

avalanches at altitudes of 2’100 to 2’400 m. Depending on the altitude of the release zone, 

and the size and velocity of the avalanches, it takes them 30-75 s to reach the road (or 

100-150 s for wet snow avalanches). However, when initially detected by Radar 1, it is not 

yet known whether or not an avalanche will actually reach the road. This depends on a 

number of factors, amongst them the distribution of snow in the lower part of the avalanche 

channel. Because of the limited warning time, the road has to be closed temporarily. To help 

the local authorities reopen the road more quickly, three webcams with infrared lights were 

installed for night and daytime observation. After receiving an alarm as automated texts and 

calls to their mobile phones, authorized users can access an online platform to check the 

cameras and, if no debris is present on the road, remotely reopen the road. Since an on-site 

visit was previously required, the webcams and the remotely controlled traffic lights and 

Figure 2A: Doppler radar installation near Zermatt. Radars 1 and 2 detect avalanches across the valley. Yellow/orange areas roughly 

indicate the radar target area. Radar 2 permits an automatic reopening of the main road. 2B: The IDS IBIS-FL ground based radar 

interferometer is mounted on top of the cable-car station, measuring surface flow velocities of Triftglacier at Weissmies roughly 1.5 km 

away. 


arriers significantly shorten the time needed to reopen the road from about 45 minutes to 

10 minutes. 

Radar 2 was installed with the aim to simplify the reopening process even further: this radar 

targets an area below the field of view of Radar 1, at an altitude of 1’800 to 1’900 m. 

This data can be used to automatically reopen the road, since it can verify whether or not 

an avalanche detected by Radar 1 reached lower altitudes. 

A short documentary about the system by the Swiss National TV in English can be found 

here: http://gpr.vn/swi 

SITE 2: WEISSMIES 

A large part of the glaciated northwest face of Weissmies Mountain in the Saas Valley of 

Switzerland recently became unstable. The likely causes of this instability are climate-induced 

glacier thinning of the supporting Triftgletscher and a progressive warming from freezing to 

melting at the ice-bed interface. Consequently, in summer 2014, a portion of the glaciated 

face with roughly 800‘000 m 3 of ice switched into an “active phase” with high flow velocities 

(Preiswerk et al., in press). Such fast flow increases the likelihood of a major icefall, which 

could endanger human populations and infrastructure in the Saas Valley. A monitoring 

campaign was initiated to detect any early warnings of dangerous break-off events to allow 

a timely evacuation of threatened areas. In October 2014, a GBInSAR system and a webcam 

were installed on the roof of the cable-car station at 3’142 m.a.s.l. (Figure 2B). The distance 

from the radar to the target area is roughly 1.5 km, resulting in a mean pixel width of about 

6 m. The radar system operates continuously, scanning at intervals of 2 minutes. The focused 

raw data is uploaded to an offsite server for further analysis. While GBInSAR is not impacted 

by darkness or low visibility, the measurements are influenced by air temperature, pressure 

and humidity. Other influences include surface ablation and snow accumulation, both of 

which can be misinterpreted as glacier movement. Solar radiation and rain change the way 

the radar signal is reflected within the snowpack due to the different refractive indices of ice 

and water. These effects are corrected using different mathematical algorithms that consider 

stable reference areas. The final velocity solutions are then uploaded to a password-protected 

website once every hour, where they are available to glaciologists for interpretation. In addition, 

photogrammetric, seismic and GPS measurements have also been performed (Preiswerk 

et al., in press). 


During the winter of 2014-2015, Doppler Radar 1 recorded 32 avalanches in Zermatt, all 

of which were confirmed by the local authorities. Avalanches could be localized with an 

accuracy of 2-3° in azimuth and within a few ten meters in range (Figure 3, top left, example 

from Weissmies). The road was only buried twice by avalanche debris, as most avalanches 

stopped above the road. We have no indications or reports that the alarm system missed any 

avalanches. Although the system has not yet experienced extreme weather such as sleet or 


Figure 3: Top left: Ice avalanche recorded on camera and measured with Doppler radar. Top right: Example of a surface flow field on the 

main arm of the glacier and the unstable section. Bottom: Time series of surface velocities (averaged over the surface area) from 

October 2014 to January 2016. 

heavy rain, it is designed to notify users once the absorption or reflection of the atmosphere 

rises such that the radar pulses no longer reach across the valley and back. 

In 2016, between January 3 and February 15, Radar 1 detected 7 avalanches. Two of these 

reached the detection area of Radar 2 and triggered the alarm. All detected avalanches were 

confirmed by the local authorities (four of which were released artificially by blasting). At the 

time of writing, no false alarms have yet occurred. Avalanche sizes are detected by the radar 

system and can be used as an alarm criterion: during the winter, algorithms were slightly 

modified to avoid closing the road when only small avalanches were detected by Radar 1. 

At Weissmies, glacier surface velocities were reliably estimated from the GBInSAR surface 

displacements over a period of over 15 months (Figure 3) and are consistent with GPS 

measurements (Preiswerk et al.,). The system has operated continuously since October 11, 

2014, and the resulting velocity data were uploaded at least once per day on 434 of the 

452 days. On the other 18 days (4% of the days), no reliable measurements were possible 

due to snowfall, snow drift or other difficult meteorological conditions. During the warm 

summer months, measurements were most reliable in the early morning, while wet snow 

inhibited measurements in the afternoon. The reflected radar signal remained strong enough 

at all times, but often decorrelated quickly under these wet-snow conditions. 

The main glacier velocity showed some seasonal changes, and the unstable area slowed down 

from roughly 20 cm/day in October 2014 to 5 cm/day by early spring 2015. Small ice fall 

events of a few 1’000 m 3 occurred about once per month. Except for a few very small failures, 

they were preceded by a 50-200% increase in local surface velocity. The five largest events 


had volumes of around 10’000 m 3 . We could successfully predict their time of failure one to 

a few days in advance by manually analyzing the velocity time series. We did not observe 

locally increasing surface velocities that did not lead to ice avalanches. 

SUMMARY 

Radar is a very versatile and robust technology for monitoring mass movements and early 

warnings of rupture phenomena like landslides and rock or ice avalanches. It can precisely 

locate the moving areas within a large field of view and measure velocities that are orders 

of magnitudes apart: from millimeters to meters a day with GBInSAR to many meters per 

second with Doppler radar. The measurements can be performed remotely, without accessing 

the area under observation, and are largely unaffected by weather conditions. 

REFERENCES 

Gubler H., and Wyssen S. (2002). Artificial release of avalanches using the remote controlled 

Wyssen Avalanche Tower. International Snow Science Workshop, Pentiction, B.C. 

Faillettaz J., Funk M., Vincent C. (2015). Avalanching glacier instabilities: Review on 

processes and early warning perspectives. Rev. Geophys., 53. 

Lussi D., Schoch M., Meier L., Ruesch M. (2012). Projekt Lawinendetektion Schlussbericht. 

WSL Institut für Schnee und Lawinenforschung SLF, 43p. 

Luzi G., Pieraccini M., Mecatti D., Noferini L., Macaluso G., Tamburini, A., Atzeni C. (2007), 

Monitoring of an Alpine Glacier by Means of Ground-Based SAR Interferometry, IEEE 

Geoscience and Remote Sensing Letters, vol. 4, no. 3, pp. 495-499. 

Meier L., and Lussi D. (2010). Remote detection of snow avalanches in Switzerland using 

infrasound, doppler radars and geophones. International Snow Science Workshop, Squaw 

Valley, CA. 

Monserrat O., Corsetto M., Luzzi G. (2014). A review of ground-based SAR interferometry 

for deformation measurement. ISPRS Journal of Photogrammetry and Remote Sensing, 

Vol. 93, pp. 40-48. 

Nolesini T., Di Traglia F., Del Ventisette C., Morett S., Casagli N. (2013). Deformations and 

slope instability on Stromboli volcano: Integration of GBInSAR data and analog modeling, 

Geomorphology, Vol. 180–181, pp. 242-254. 

Preiswerk L.E., Walter F., Anandakrishnan S., Barfucci G., Beutel J., Burkett P.G., Dalban 

Canassy P., Funk M., Limpach P., Marchetti E., Meier L., Neyer F., in press. Monitoring 


unstable parts in the ice-covered Weissmies northwest face. Interpaevent 2016 - Conference 

Proceedings. 

Rödelsperger S. (2011). Real-time Processing of Ground Based Synthetic Aperture Radar 

(GB-SAR) Measurements. Verlag der Bayerischen Akademie der Wissenschaften in Kommission 

beim Verlag C.H. Beck, Münschen. 

Sättele M. and Bründl M. (2015). Praxishilfe für den Einsatz von Frühwarnsystemen für 

gravitative Naturgefahren, WSL- Institut für Schnee- und Lawinenforschung SLF, Bundesamt 

für Bevölkerungsschutz/BABS, Bern. 

Scott E.D., Hayward C.T., Colgan T. J., Hamann J.C., Kubichek R.F., Pierre J.W., and Yount J. 

(2006). Practical implementation of avalanche infrasound monitoring technology for 

operational utilization near Teton Pass Wyoming. In Proceedings ISSW, pp. 714 - 723. 

Seyfried R. and Hort M. (1999). Continuous monitoring of volcanic eruptions dynamics: 

a review of various techniques and new results from a frequency-modulated radar Doppler 

system. Bulletin of Volcanology, Vol. 60, Issue 8, pp. 627 – 639. 

Vivian R. (1966). La catastrophe du glacier d'Allalin. Revue de Géographie Alpine, Vol. 54, 

Issue 1, pp. 97 - 112. 



Practical experience with the Flood scenario 

catalogue Carinthia, a handbook for flood forecast 

and warning. 

Praktische Erfahrungen mit dem Hochwasserszenarienkatalog 

Kärnten, ein Handbuch zur 

Hoch wasser prognose und –warnung 

Johannes Moser, DI 1 ; Christian Kopeinig, DI 1 ; Kurt Rohner, DI 1 

ABSTRACT 

The flood scenario catalogue Carinthia is an expert system that helps to improve the quality 

and stability of flood forecasts during a flood event achieved by profound analyzes and 

numerous calculations beforehand. Results of thousands of simulations are collected in a 

database and displayed in a handbook by diagrams in a user friendly way. Thus, for experts 

it is possible to make well-founded statements about the future development of floods for 

a variety of forecasting problems. This is possible even without a computer and online model 

calculations, for example, when computational models are not available. 

Typical questions during flood warning situations as: "What is the maximum discharge we 

have to expect? When and where is the peak of the flood? Where do we have to expect 

dangers due to the flood?" can be answered in good quality with support of this system. 

KEYWORDS 

Flood, flood warning, flood forecast, scenario catalogue, civil protection 


Der Hydrografische Dienst Kärnten betreibt für die Warnung der Bevölkerung, die Information 

von Behörden, Einsatzkräften und Mitarbeitern im Hochwassermanagement ein Hochwasserwarnservice 

mit Internetdienst und SMS Versand. Während eines Hochwassers müssen 

Entscheidungen dabei rasch und oft auch unter stressigen Bedingungen getroffen werden. 

Die Zuverlässigkeit der Einschätzung des weiteren Verlaufs des Hochwassers ist deshalb von 

großer Bedeutung. 

Es wurde deshalb in Zusammenarbeit mit der Technischen Universität Wien, zusätzlich zum 

kontinuierlich betriebenen Niederschlags-Abflussmodell, ein Expertensystem entwickelt, das 

den Hydrologen im Hochwasserfall dabei unterstützt Prognosen in hoher Qualität und 

Stabilität zu erstellen (Moser, Kopeinig, 2006). 

1 Regional Government of Carinthia, Klagenfurt, AUSTRIA, johannes.moser@ktn.gv.at 



Modellgestaltung, Kalibrierung, Auseinandersetzung mit beobachteten Hochwasserereignissen, 

Ermittlung von wesentlichen Hochwasserereignistypen und Szenarien nahmen zwar 

etliche Jahre in Anspruch, die dabei gewonnenen Erfahrungen und Kenntnisse sind für den 

Hydrologen aber von immensem Nutzen. 

DATENGRUNDLAGE 

Die Entwicklung des Hochwasserszenarienkataloges erforderte als ersten Schritt eine eingehende 

Analyse historischer Hochwasserereignisse. Dazu wurden ca. 10 Ereignisse pro 

Pegel, Großteils im Zeitraum 1980 – 2010 (einzelne große Ereignisse auch davor) ausgewählt, 

um typische Muster zu erkennen und mit den vorbereiteten Szenarien möglichst viele real 

ablaufende Hochwasserereignisse einschätzen zu können. Die Größe der Einzugsgebiete 

beträgt 75 bis ca. 5000 km². 

In einem ersten Schritt wurden dafür aus den vorhandenen Daten von ca. 120 Nieder - 

schlags- und Temperaturmessstationen flächig interpolierte Niederschlags- und Temperaturverteilungen 

für ein Raster von 90 mal 228 km und einer Zellgröße von 1 km² berechnet. 

Die Interpolation erfolgte mittels Kriging Verfahren. Dies erfolgte von den 50er-Jahren des 

vorigen Jahrhunderts bis 1990 mit einer zeitlichen Auflösung von einem Tag, ab 1990 mit 

einem Zeitschritt von einer Stunde. Die Höhenabhängigkeit des Niederschlages wurde mit 

einem räumlich variablen, aber zeitlich fixen Höhengradienten berücksichtigt. Der Höhengradient 

der Temperatur wurde für jeden Zeitschritt aus den aktuell gemessenen Stationsdaten 

berechnet. 

Aus den Gebietsverteilungen wurden für alle im Katalog erfassten Pegelstellen Ganglinien 

des mittleren Gebietsniederschlages, der Gebietstemperatur und des Gebietsregens berechnet. 

Für die Ermittlung des Gebietsregens wurde für jede Rasterzelle aufgrund der Temperatur 

eine Unterscheidung in Regen oder Schnee durchgeführt. 

ANALYSE HISTORISCHER HOCHWÄSSER 

Auf der Basis der berechneten Gebietsganglinien wurden für die wichtigsten Pegelmessstellen 

der Vorregen während der 10 Tage vor dem Hochwasserereignis, der Ereignisregen sowie das 

Volumen der Hochwasserwelle ausgewertet. Daraus konnte ein fiktiver Abflussbeiwert des 

Direktabflusses berechnet werden. Die Hochwasserereignisse wurden in einem „flow ratio“ 

– „Wasserdargebot“ Diagramm aufgetragen. Dabei ergab sich eine Clusterung in Abhängigkeit 

von der Vorfeuchte (Vorregen). Auf der Basis dieser Auswertungen und zusätzlicher 

Plausibilitätsüberlegungen wurden in den Diagrammen dann auch die für die Katalogsimulationen 

verwendeten „flow ratio“ Werte des Direktabflusses festgelegt. (siehe Abb. 1 auf 

Folgeseite) 


Abbildung 1: Pegel Oberdrauburg, Drau: Aus Eichereignissen rückgerechnete und für die Katalogsimulationen festgelegte flow-ratio 

Werte für den Direktabfluss und für den trockenen, mittleren und feuchten Ausgangszustand. 

Abbildung 2: Gewählte zeitliche Muster der Niederschlagsverteilungen. 

Abbildung 3: Festgelegte räumliche Muster der Niederschlagsverteilungen für das Flussgebiet obere Drau. 


Zusätzliche Betrachtungen ergaben, dass die Niederschlagsintensität (Regendauer), sowie die 

zeitliche und räumliche Niederschlagsverteilung Einfluss auf die Hochwasserspitze ausüben, 

wobei bei der zeitlichen Verteilung des Niederschlags häufig folgende Muster beobachtet 

wurden: (siehe Abb. 2 auf Folgeseite) 

Bei der räumlichen Verteilung der Niederschläge zeigte sich, dass kleinere Einzugsgebiete 

nicht unterteilt werden müssen, beim Haupteinzugsgebiet der Drau wurden folgende Muster 

beobachtet: (siehe Abb. 3 auf Folgeseite) 

MODELLKONZEPT UND MODELLWAHL 

Für die Simulation der Katalogszenarien wurde das Modell HEC-HMS des U.S.A.C.E 

verwendet. Die Berechnung erfolgte Ereignisbasiert mit einem 3-Speicher Ansatz (surfaceflow/schnell, 

interflow/mittel, baseflow/langsam). Die Berechnung der Abflusskonzentration 

geschah mittels Clark – Verfahrens (Clark, 1943). Die Modellparameter wurden an die bei der 

Analyse historischer Hochwässer gewonnenen Werte angeeicht (Schindler, 2006). 

Der Wellenablauf im Gerinne wurde mit den Methoden der kinematischen Welle und des 

Muskingum – Cunge Verfahrens (Dyck, 1995) berechnet. 

Abbildung 4: Pegel Oberdrauburg, Drau: Für die Katalogsimulationen festgelegte flow-ratio Werte für den schnellen und mittleren 

Abflussanteil und den trockenen, mittleren und feuchten Vorfeuchtezustand. 

MODELLKALIBRIERUNG 

Bei der Modellkalibrierung wurden, für eine begrenzte Zahl von Niederschlagsmustern 

(siehe Abb. 2 und Abb. 3), die in der Analyse der historischen Ereignisse gewonnenen 

flow-ratio Werte des Direktabflusses auf die Modellspeicher „schnell“ und „mittel“ aufgeteilt. 

Es wurden dabei für jeden Pegel, und für jede Abflusskomponente, Diagramme erzeugt, 


aus denen der jeweilige „speicherspezifische flow ratio - Wert“, in Abhängigkeit von der 

Niederschlagsmenge und der Vorfeuchte im Einzugsgebiet, ablesbar ist. 

HOCHWASSERSZENARIENKATALOG 

Der Hochwasserszenarienkatalog Kärnten stellt auf Basis der zuvor beschriebenen Modellierungen 

eine umfangreiche Zusammenschau der Hochwasserscheitel von möglichen Hochwasserszenarien 

für Kärnten dar. 

Dabei wurden die Darstellungen in Diagrammen so gewählt, dass durch den Benutzer auch 

eine „visuelle“ Interpolation bzw. Extrapolation zwischen den Werten möglich ist. 

Bei der grundlegenden Annahme einer räumlich gleichmäßigen Niederschlagsverteilung 

wurden 336 Szenarien pro Pegelmessstelle (7 Stufen der Niederschlagssumme, 4 Dauer stufen, 

4 unterschiedliche zeitliche Verteilungen, 3 Zustände der Vorfeuchte) berechnet. Bei den 

Pegelstellen mit großem Einzugsgebiet wurden zusätzlich drei räumlich variierende Regenverteilungen 

untersucht (siehe Abb. 3), dadurch ergaben sich dann 1.344 Szenarien pro 

Messstelle. Insgesamt ergaben sich für 41 Pegel ca. 20.000 Simulationen bzw. Ergebnisse. 

Sämtliche Modellergebnisse wurden in einer Datenbank gespeichert und in Ergebnis- bzw. 

Katalogblättern dargestellt. 

Bei dieser großen Anzahl an vorliegenden Berechnungen liegt es auf der Hand, dass für die 

erfolgreiche Anwendung des Szenarienkataloges die praxisgerechte Aufbereitung der Daten 

von entscheidender Bedeutung ist. 

Die Katalogszenarien wurden deshalb in den für den praktischen Einsatz vorgesehenen 

Diagrammen nach folgenden Merkmalen gegliedert: 

– Niederschlagshöhe: Die Summe aus Niederschlag und Schneeschmelze im Laufe eines 

Ereignisses [mm] in Stufen von 25 bzw. 50 mm. Die Schneeschmelze wird dem parallel 

betriebenen Niederschlag-Abflussmodell entnommen oder vereinfacht abgeschätzt. 

– Niederschlagsdauer: Dauer des Ereignisniederschlags in Stunden [h]. Für alle Einzugsgebiete 

wurden folgende Dauerstufen gewählt: 12 h, 24 h, 48 h und 72 h 

– Vorbefeuchtung: Beschreibung des Ausgangszustandes bei Beginn eines Ereignisses 

bezogen auf die Bodenfeuchte. 

– Zeitliches Muster der Niederschlagsverteilung (Blockregen, Anfangsbetont, Doppelwelle, 

Endbetont), jeweils ein eigenes Blatt. 

– Räumliche Muster der Niederschlagsverteilung (bei großen Einzugsgebieten), jeweils ein 

eigenes Blatt. 

ANWENDUNG DES SZENARIENKATALOGES 

Wie in den Beispieldiagramm für den Pegel Krottendorf/Lavant zu sehen ist (Abbildung 5), 

sind die Abflussspitzen QMax der Hochwasserszenarien als kleine Rechtecke für bestimmte 

Niederschlagsmengen, Niederschlagsdauern und Bodenfeuchtezustände dargestellt. Sie sind 


Abbildung 5: Katalogdiagram für den Pegel Krottendorf / Lavant für das Szenario Doppelwelle. 

entsprechend der Modellsimulation als Scheitelwerte von zu erwartenden Hochwasserabflüssen 

zu verstehen. 

Wichtigste Eingangsgröße ist die prognostizierte Niederschlagsmenge. Diese wird in der Regel 

vom Wetterdienst als Gebietsniederschlagshöhe für die einzelnen Regionen bekanntgegeben. 

Für jede Niederschlagsmenge wird im Diagramm horizontal nach der Vorbefeuchtung (von 

links nach rechts: trocken, mittel, feucht) und vertikal nach der Niederschlagsdauer (von 

oben nach unten: 12, 24, 48, 72 Stunden) unterschieden. Für die verschiedenen zeitlichen 

Niederschlagsverteilungen (Blockregen, Anfangsbetont, Endbetont, Doppelwelle) gibt es 

jeweils ein eigenes Blatt, bei Stationen mit räumlicher Unterteilung gibt es zusätzlich eigene 

Blätter pro Verteilung. Das Diagramm lässt auch eine visuelle Interpolation zwischen den 

Katalogwerten zu. Unterschiedliche Szenarien mit veränderten Eingangsgrößen können 

rasch abgeschätzt werden (Moser, Kopeinig, 2009). 

Zur Einordnung der Hochwasserscheitel nach der Auftrittswahrscheinlichkeit bzw. Jährlichkeit 

wurde der Diagrammhintergrund entsprechend den Hochwasserkennwerten eingefärbt. 

Wie man dem Diagramm entnehmen kann, kann ein und dieselbe Niederschlagsmenge durch 

unterschiedliche Kombinationen von Niederschlagsdauer, zeitlichen Verlauf und Bodenfeuchte 

stark variierende Spitzenabflüsse bewirken. Allgemein ist erkennbar, dass mit zunehmender 

Niederschlagssumme der Einfluss der Vorfeuchte abnimmt. 

Zusätzlich zur analogen Form des Kataloges wurde ein Softwareprogramm entwickelt 

welches es erlaubt anhand von Filterkriterien wie Niederschlagssumme, Vorfeuchte, Dauer, 

räumliche und zeitliche Verteilung für eine Station schnell die passenden Katalogszenarien zu 


Tabelle 1: Ausschnitt aus der generellen Gefährdungsliste für die Lavant. 

finden. Innerhalb dieses „Digitalen Szenarienkataloges“ können gemessene und simulierte 

Ganglinien des Gebietsniederschlages und des Abflusses auch visualisiert werden, was eine 

gute Einschätzung des Weiteren zeitlichen Verlaufes eines Hochwasserereignisses erlaubt. 

GEFÄHRDUNGSLISTEN 

Um auch Auskunft über Auswirkungen und mögliche Gefährdungsbereiche geben zu 

können, ergänzen sogenannte Gefährdungslisten den Szenarienkatalog (siehe Tab. 1 auf 

Folgeseite). Darin sind, geordnet nach der Jährlichkeit des gefährdenden Abflusses, tabella- 

Abbildung 6: Verlauf von Abfluss und Wasserdargebot für den Pegel Krottendorf / Lavant für das Hochwasserereignis am 20.6.2004. 


Tabelle 2: Kennwerte des Hochwasserereignisses am Pegel Krottendorf / Lavant am 20.6.2004. 

risch kritische Bereiche aufgelistet. Sind also die Spitzenabflüsse mit Hilfe des Szenarienkataloges 

abgeschätzt, können die damit verbundenen Gefährdungsbereiche schnell genannt 

werden. 

EREIGNISANALYSE UND DOKUMENTATION 

Der Katalog wurde in den Jahren 2003 bis 2010 erstellt und kommt seither im Rahmen des 

Hochwasserwarnservices mehrmals pro Jahr zum Einsatz. Die fortlaufende Analyse und 

Dokumentation von aktuellen Hochwasserereignissen ist dabei ein wesentlicher Bestandteil 

zur Validierung des Kataloges. Auswirkungen von Unsicherheiten in der Datenlage und die 

Qualität der Simulationsberechnungen sollen durch die Ereignisanalysen aufgezeigt werden. 

Ziel ist eine laufende Verbesserung der Abschätzgenauigkeit der Hochwasserwelle. Zur 

Erlangung von Routine und Erfahrung im Umgang mit dem System ist die Ereignisanalyse 

ebenfalls hervorragend geeignet. 

Allgemein kann nach mehr als 5 Jahren Anwendungserfahrung gesagt werden, dass die Güte 

der Prognosen mit jenen des parallel betriebenen numerischen Modells vergleichbar ist. 

BEISPIEL: HOCHWASSER AN DER LAVANT IM JUNI 2004 

– Zeitliches Niederschlagsmuster: schwach ausgeprägte Doppelwelle (auch Blockszenario 

möglich), Niederschlagsfront mit eingelagerten Gewittern. 

– Wasserdargebot gleich Niederschlag (kein Schneeinfluss): ca. 80 mm. 

– Niederschlagsdauer: ca. 24 Stunden. 

– Ausgangszustand / Bodenfeuchte: feucht (180 mm Regen im Vormonat, Abfluss ~ 2 x MQ) 

Katalogabschätzung laut Hochwasserkatalogblatt für den Pegel Krottendorf (siehe Abbildung 

5) ergibt eine Abflussspitze von ca. 220 m³/s oder ca. HQ 40 

. 

Die gemessenen Abflüsse waren mit 235 m³/s etwas höher als im Katalogszenario „feucht“, 

wahrscheinlich aufgrund eingelagerter Gewitterzellen und ungleichmäßiger Regenverteilung. 


ZUSAMMENFASSUNG: 

Die Anwendung des Hochwasserszenarienkataloges ist für den Hydrographischen Dienst 

Kärnten mittlerweile zur Routine geworden. Der Katalog ist eine äußert wertvolle Expertise, 

die viele Informationen und Ergebnisse auf Basis von Modellrechnungen beinhaltet und den 

Experten in „bedrängten“ Zeiten fachlich in seiner Bewertung und Entscheidung unterstützten 

kann. Der Katalog schafft einen raschen und vielfältigen Überblick über mögliche 

Hochwasserszenarien auf Basis von Gebietsniederschlagsprognosen und ist in seiner 

Anwendung doch einfach gehalten. 

Eine fortlaufende Evaluierung von Hochwasserereignissen und Katalogergebnissen ist 

gefordert um den Umgang mit dem Katalog und den damit verbundenen Unsicherheiten zu 

lernen, und falls erforderlich den Katalog zu adaptieren bzw. zu ergänzen. 

REFERENZEN 

- Moser, Kopeinig (2006): Hochwasserwarnung in Kärnten, Wiener Mitteilungen, Band 199, 

S 23-38. 

- Moser, Kopeinig (2009): Hochwasserwarnung in Kärnten – ein Praxisbeispiel, Wiener 

Mitteilungen, Band 216, S213-230. 

- Clark (1943): Storage and the unit hydrograph, Proc. ASCE Vol. 9. 

- Dyck (1995): Grundlagen der Hydrologie. 

- Schindler (2006): Modifikation des HEC-HMS Modells im Rahmen der Entwicklung eines 

Hochwasserprognosemodells. Diplomarbeit, TU Wien. 



Avalanche detection systems: A state-of-the art 

overview on selected operational radar and infrasound 

systems 

Walter Steinkogler, Ph.D. 1 ; Lorenz Meier, Ph.D. 2 ; Stian Langeland 1 ; Sam Wyssen 1 

ABSTRACT 

The application of detection systems allow a reduction of closure time of roads in combination 

with a reduction of residual risk. Over the last years, the developments and advances of 

radar (LARA) and infrasound (IDA) avalanche detection systems and especially the integrated 

visualization (PIA) significantly improved the operational applicability and showed their 

capability to support the avalanche control work. In this work, we present results from 

operational experience and recent developments of these systems. The results from verification 

campaigns, such as minimum detectable size, longest distance of detection and detection 

of wet snow avalanches are discussed. Furthermore, we indicate current shortcomings and 

future development directions. 

KEYWORDS 

Avalanche Detection; Radar LARA/SARA/PETRA; Infrasound IDA; Integrated Visualisation 

PIA 

INTRODUCTION 

Snow avalanches pose a direct threat for people and infrastructure during winter time. 

Governmental agencies protect settlements and traffic routes using permanent measures 

(tunnels, steel structures, etc.) and/or active and passive temporary measures (e.g. road 

closures, evacuations, preventive avalanche release, avalanche forecasting, etc.). The 

preventive release of snow avalanches along traffic routes has been applied since many 

years as permanent measures are too expensive or not feasible to construct for certain areas. 

Furthermore, due to the increased mobility of people, long-lasting closures of roads and 

railway lines are less and less accepted. The preventive release methods are much more 

effective when the success of preventive releases can be verified reliably. The application of 

detection systems allows a reduction of closure time of roads in combination with a reduction 

of residual risk and aid the avalanche control team in their decision making. Site-specific 

alarm thresholds can be set for automatic closure of traffic lines. In addition, the knowledge 

of the occurrence, frequency and size of avalanche events can assist regional or local authorities 

who are responsible for the control and forecasting of avalanche hazard. 

1 Wyssen Avalanche Control, Reichenbach, SWITZERLAND, walter@wyssen.com 

2 GEOPRAEVENT AG 



A variety of systems for the detection of avalanches was tested in recent years and partly 

transferred into operational use at traffic route operations and ski resorts. Depending on the 

aim of the operation and the object to protect, the most suitable system should be selected 

(Table 1). 

Table 1: Overview of different avalanche detection systems and their suitability for different operations. 

In this study we focus on gathered operational experiences and recent developments of radar 

(LARA, SARA, PETRA) and infrasound systems (IDA) (Figure 1). Furthermore, the inevitable 

incorporation of these systems in a practitioner-friendly and easy to operate platform (PIA) 

is described. Other methods for the detection of avalanches exist, such as geophones in the 

release area or along the path, or trigger lines in the path, but are not discussed in this paper 

as here we focus on remote detection methods. 

Figure 1: Schematic overview of operational radar (LARA, SARA) and infrasound (IDA) detection systems. 


RADAR SYSTEMS LARA AND SARA: 

Technical description: 

Radars have been applied for the detection of avalanches for many years. In most cases 

(pulsed or frequency modulated) Doppler radars are used (Gubler and Hiller, 1984, Gauer et 

al. 2007, Fischer et al., 2014), emitting electromagnetic waves at a certain frequency, which 

are then reflected and travelling back to the radar. To detect an avalanche by radar, the 

avalanche movement must at least partly be directed towards the radar, as the line-of-sight 

component of the velocity is measured. The avalanche velocity leads to a Doppler-shifted 

signal in frequency space, allowing the radar to discriminate between moving and static 

targets. Hence, avalanches can only be detected by radar once they are in motion. 

Experience with LARA/SARA: 

In 2011 Wyssen Avalanche Control AG installed the first version of the Long range Avalanche 

Radar, LARA, in Ischgl, Austria (Figure 2, left). The purpose of the radar installation was i) 

Verification of the controlled release and ii) Gathering information about spontaneous 

avalanche activity. Over the last years, the radar has been working very reliably and 

satisfactorily and it became a standard operational tool of the safety staff. Consequently, three 

more radars of the same type were installed in Austria and Switzerland. The big advantage of 

the radar is the accurate detection of even small avalanche events, e.g. preventively released 

ones. The shorter the distance to the radar antenna and the better the weather conditions 

(i.e. no rain, no snowfall), the smaller the detectable avalanches are (events of a few 100 m³ 

in a distance of 1.5 km were detected with radars of the newest generation). On the contrary, 

the monitored area is limited to one single avalanche path. Multiplexing with multiple 

antennas is possible and applied in some locations. However, it has some limitations as 

multiplexing more than two or three antennas can become difficult. Typically, measurements 

are taken once per second per antenna, i.e. with three antennas each antenna would be 

‘blind’ during 2 seconds. Still, even three 5 degrees antennas only cover a limited area 

compared with the new 90°x10° radars. 

Figure 2: A LARA (left) and a SARA (right) installation used for avalanche detection for operational road protection. 


To account for the limited size of the monitoring area, a new radar system with a much wider 

opening angle (up to 90°) was tested to account for this shortcoming. The newest avalanche 

radar version is a Range-Doppler Radar operating at the X band, with the ability to measure 

azimuth angle, range and velocity. It has an opening angle of 90° horizontally and 10° 

vertically and captures all movements within this area (Figure 3). It can detect a large 

avalanche up to a range of 2000 m (LARA) or 500 m (SARA), covering an area of up to 1 km 2 

(LARA). Power can be provided by fuel cells or by permanent power supply if available. 

To ensure the system reliability of the radars they are constantly monitored remotely and 

maintained on-site every year. 

Since radar systems provide data in real-time, alarm thresholds can be defined which allows 

to also use the system for the automatic closure of traffic lines. Up to five independent 

algorithms look for patterns in the radar data that are typical for avalanches. Depending on 

the local requirements for detection probability and the tolerance for false alarms, between 

one and five algorithms are necessary to trigger an alarm. 

Figure 3: Example of LARA of newest generation monitoring an ice avalanche. Camera view of moving avalanche (left) and 

corresponding radar image (right, colors showing signal strength). 


Based on the success of the avalanche radar, the Short distance Avalanche Radar SARA with 

less energy consumption was developed. It was mounted directly on the Wyssen avalanche 

towers (Figure 2, right) to get immediate information about the success of the avalanche 

release within the effective range of the system. This is a much needed feature for verification 

of preventively released avalanches. Furthermore, other uses of SARA, such as the detection 

of persons moving in the area endangered by avalanches, were successfully tested (Video: 

http://gpr.vn/PETRA ). 

Between November 2014 and now, SARA/LARA systems of the newest version have been 

installed at six different locations. Three SARAs were mounted on Wyssen Avalanche Towers, 

and three LARAs were installed at locations where frequent spontaneous and blasted 

avalanche events occur. One of these locations is the Triftglacier at Weissmies (Figure 3), 

where a variety of other detection systems are deployed (see L. Preiswerk et al, INTER- 

PRAEVENT paper). In total, about 10 terabytes of data were gathered and more than 

200 avalanches identified (Video: http://gpr.vn/LARA ). 

INFRASOUND SYSTEM IDA 

Technical description: 

Infrasound waves are low frequency (

could be even enhanced. Based on this success additional systems were installed and 

currently four systems are used in Switzerland and two in Norway. 

In Switzerland an extensive verification campaign was conducted in Winter 2014/15. IDA 

was used to monitor certain avalanche paths which endanger a local road and to define the 

smallest avalanche size which can be detected by the system. Although the system detected 

many of the smaller slides (size 1 – 2), they were not automatically visualized and identified 

as avalanches as they were below the defined thresholds. Mid-sized and large dry slab 

avalanches were correctly detected. In azimut direction the detected avalanches fit the 

observations with an accuracy of ± 3°. Additionally, large dry avalanches could be detected 

up to 14 km away from the IDA system. 

In the case of wet snow avalanches some of the larger events were not detected by IDA 

automatically. Yet, post-processing of the data revealed that the system recorded the 

avalanches but the signal did not fulfill the detection criteria, which were defined for dry slab 

avalanches. Even though the main source of infrasound from avalanches is produced by the 

powder cloud, these results indicate that larger wet snow avalanches do produce enough 

infrasound to be detected by IDA. Strong ambient noise, such as wind, has shown to 

complicate the identification of the avalanche signal. 

Furthermore, the IDA system was also tested if it is also suitable for use in climatic conditions 

of high alpine areas in Norway. Multiple avalanches were automatically detected and several 

others by post-processing of the data after the season. At one of the locations, more than two 

metres of dense (250-300 m 3 ) snow with several ice layers covered the sensors which 

influenced the quality of the signals. Therefore, the overall impact on the system due to large 

snow depths will be further investigated in the next winter season. 

IDA proved to be a very valuable tool for gathering information about avalanche activity of 

multiple avalanche paths in a larger area. Since IDA is continuously monitoring it also 

provides data on spontaneous avalanche activity, which can be very useful information for 

the local avalanche control team (Figure 4, green arrows). 

PLATFORM OF INTEGRATED AVALANCHE INFORMATION PIA: 

For road authorities operating several avalanche release and detection systems, simplicity is 

one of the key demands. In order to satisfy this need, an interdisciplinary project was 

launched in 2013 to develop an online information platform, PIA, Platform of Integrated 

Avalanche information. The goal was to gather relevant information from avalanche release 

systems (e.g. avalanche towers) and detection systems (e.g. geophones, LARA, IDA) and to 

visualize the results in a clear and simple way, making it possible to get a good overview at a 

glance using a mobile phone or laptop. PIA has been successfully applied in operational use 

for road protection in Gonda, Switzerland since winter 2013/14 (Figure 5). 


Figure 4: Online map visualization of the IDA system showing preventively released avalanches (red arrows), explosions (orange arrows) 

and natural avalanche activity (green arrows). Blue circles indicate positions of preventive avalanche release systems. 

Figure 5: PIA platform showing an operational example of artificial avalanche release and verification for the Gonda avalanche 

(Switzerland). Black circles indicate the positions of preventive avalanche release systems along the ridge with the red circle and text, 

e.g. GONDAC, showing the last detonated avalanche tower in the current list of events (left). Yellow circles are geophone locations and 

red and blue filled areas indicate areas where geophones detected motion. Black rectangles along the path indicate the area monitored 

by the two radar antennas with detected avalanches in red ellipsoids 


RESULTS AND OPERATIONAL EXPERIENCE 

From an operational point of view both systems have proven to have reached a technological 

level at which they work reliable, both in terms of system stability and avalanche detection 

performance, and can significantly assist local avalanche control teams (Table 2). Both 

systems need a calibration period (a few avalanches of typical size for the avalanche path) to 

optimize the alarm parameters and fine-tune to the local conditions and thus minimize false 

alarms. Generally, an intensive and well-prepared planning phase is essential to achieve the 

desired functionality and accuracy of the systems. 

Systems as LARA are very suitable to monitor a single avalanche path, such as the Gonda 

avalanche path in Figure 5. Large avalanches were reliably detected and smaller avalanches, 

i.e. a few 100 m³, were detected up to a distance of 1.5 km in good weather conditions. 

Depending on the terrain, the new LARA generation with a horizontal opening angle of 90° 

allows to monitor multiple avalanche paths (Figure 3). Additionally, the automatic alarm 

messages reliably inform the local authorities about spontaneous avalanche activity in the 

corresponding path. 

The infrasound system IDA proved to be able to successfully monitor the avalanche activity 

of medium-sized to large avalanches in an area up to 5 km radius. IDA is also able to detect 

smaller avalanches although they are often not automatically displayed as they do not fulfill 

all the criteria by the processing algorithms. Furthermore, the accuracy of the system 

decreases for small avalanches. Small wet avalanches were not detected but larger ones are 

recorded by the IDA system. 

Table 2: Summary and technical details of radar and infrasound systems. 



Over the last years, the developments and advances of (radar and infrasound) avalanche 

detection systems and especially the integrated visualization (PIA) worked very reliably and 

showed their capability to support the avalanche control work. 

Recent verification campaigns and an increasing number of installations in different climatic 

regions have allowed to better define the technical capabilities of the systems. For the 

radar-based detection systems (LARA, SARA, PETRA), future data needs to be gathered on its 

operation during varying meteorological conditions (e.g. during wet snowfall and heavy 

rain). 

For the infrasound system (IDA), the influence of wind is challenging. A new sensor 

generation in combination with optical fiber cables will allow for a much better signal-tonoise 

ratio and looks promising to improve the system performance. As large wet snow 

avalanches were recorded by the system but not automatically classified them as avalanches, 

an adaptation of the processing algorithms should allow to also detect these avalanches. 

A further improvement of the IDA system will be an index for the general avalanche activity 

in the area with much less strict filters in the algorithms. Even though this will possibly result 

in a higher false alarm rate, this should allow measuring a general increase in avalanche 

activity in the observation area (also of smaller avalanches). 

REFERENCES 

Bedard, A., (1994). An evaluation of atmospheric infrasound for monitoring avalanches. 

Proceedings, 7th International Symposium on Acoustic Remote Sensing and Associated 

Techniques of the Atmosphere and Oceans, 3–5 October, Boulder, CO. 

Fischer J. T., Fromm R., Gauer P., Sovilla B (2014) Evaluation of probabilistic snow avalanche 

simulation ensembles with Doppler radar observations. Cold Regions Science and Technology, 

Volume 97, 151 - 158 

Gauer, P., Kern, M., Kristensen, K., Lied, K., Rammer, L.,Schreiber, H. (2007). On pulsed 

Doppler radar measurements of avalanches and their implication to avalanche dynamics, 

Cold Regions Science and Technology, 50, 55–71 

Gubler H., Hiller M. (1984). The use of microwave FMCW radar in snow and avalanche 

research. Cold Regions Science and Technology, Volume 9, Issue 2, 109-119 

Kogelnig, A., Suriñach, E., Vilajosana, I., Hübl, J., Sovilla, B., Hiller, M. and Dufour, F. (2011) 

On the complementariness of infrasound and seismic sensors for monitoring snow avalanches, 

Natural Hazards and Earth System Sciences, 11(8), 2355-2370 


Marchetti E., Ripepe M., Ulivieri G., Kogelnig A. (2015) Infrasound array criteria for 

auto matic detection and front velocity estimation of snow avalanches: towards a real-time 

early-warning system. Natural Hazards and Earth System Sciences 3(4): 2709-2737 

Ulivieri, G., Marchetti, E., Ripepe, M., Chiambretti, I., De Rosa, G. and Segor, V. (2011) 

Monitoring snow avalanches in Northwestern Italian Alps using an infrasound array, Cold 

Regions Science and Technology, Volume 69, Issues 2–3, December 2011, Pages 177–183P 



Integrated risk management of natural hazards by the 

railway Company BLS Netz AG 

Hans-Heini Utelli, dipl. Natw. ETH 1 , Geologe; Christian Pfammatter, dipl. Forst. Ing. ETH 2 ; Franz Kuster, dipl. Geograph 3 

ABSTRACT 

The routes of the BLS railway cross the alps and are exposed to gravitational natural hazards. 

In order not to injure the protection goal for travelers, the railway company BLS Netz AG has 

implemented numerous structural, silvicultural and organizational measures. Thanks to a risk 

analysis and the subsequent integral planning of measures those sections are protected where 

the risk and the cost-benefit difference of measures or combination thereof are high. 

The so reached security level is as long as possible held with a long-term conservation plan 

for all protective structures, the maintenance of protective forests and the monitoring of rock 

instability. The security level should be periodically checked with a renewed risk analysis. 

KEYWORDS 

railway; gravitational hazards; risk analysis; integrated planning of measures; long-term 

conservation plan 

EINFÜHRUNG 

Die Bahnlinie der BLS Netz AG in den Berner Alpen zwischen Frutigen und Kandersteg 

(Schweiz) ist eine für den nationalen und internationalen Verkehr wichtige Eisenbahnverbindung. 

Sie ist - wie andere Bahnlinien in den Alpen auch - auf verschiedenen Abschnitten 

durch gravitative Naturgefahren betroffen. In der Vergangenheit wurden bereits umfangreiche 

Massnahmen zum Schutz gegen Naturgefahren ergriffen. Dennoch bleibt die Strecke vor 

allem durch Stein- und Blockschlag, Felssturz, Überschwemmung oder Übersarung, Lawinen 

und Übermurung gefährdet. (siehe Abb. 1 auf Folgeseite) 

Die BLS Netz AG beabsichtigt eine ausreichende Sicherheit auf dieser Strecke dauerhaft zu 

gewährleisten. Die Strategie dazu basiert auf den folgenden Elementen, die in den nächsten 

Kapiteln näher erläutert werden: 

– Periodische Analyse und Bewertung des IST-Zustandes hinsichtlich Gefährdung und 

Risiken. 

– Konsequente Umsetzung der integralen Massnahmenplanung mit: 

– a) Evaluation und Planung von zusätzlichen baulichen , waldbaulichen und / oder 

organisatorischen Schutzmassnahmen auf der Basis von Nutzen-Kosten-Überlegungen, 

– b) Erhalt der bestehenden wie auch der in Zukunft zu erstellenden Schutzbauten sowie 

Pflege der Schutzwälder. 

1 IMPULS AG Wald Landschaft Naturgefahren, Thun, SWITZERLAND, hans-heini.utelli@impulsthun.ch 

2 Amt für Wald Kanton Bern, Abteilung Naturgefahren 

3 BLS Netz AG 



Abbildung 1: Übersicht über die Bahnlinie der BLS Netz AG zwischen Frutigen und Kandersteg im Berner Oberland, Schweiz. 

PROBLEMATIK 

Risikoentwicklung ohne Massnahmen 

Auf der BLS -Linie Frutigen – Kandersteg wird das durch gravitative Naturgefahren ausgelöste 

kollektive Personen- und Sachrisiko unabhängig von den Veränderungen in den Gefahrenprozessen 

zunehmen 

– a) weil Zugsfrequenzen und Passagierzahlen sowie die Ansprüche an die Verfügbarkeit 

auch im öffentlichen Verkehr in den nächsten Jahren zunehmen und 

– b) weil sich der Zustand der bestehenden Schutzbauten ohne gezieltes Unterhaltsmanagement 

verschlechtern wird und somit deren Zuverlässigkeit abnimmt. Wenn dadurch 

Ereignisse nicht mehr von den Schutzbauten aufgehalten werden, werden Schäden durch 

Naturereignisse an Personen und Infrastruktur wieder wahrscheinlicher 

(siehe Abbildung 2). 

Verletztes Schutzziel? 

Die BLS Netz AG will sicherstellen, dass die Passagiere auf der Bahnstrecke Frutigen-Kandersteg 

keinem zu hohen Risiko ausgesetzt sind: das vom Kanton Bern definierte und von der 

BLS mitgetragene Schutzziel für einen typischer Pendler, welcher diese Bahnstrecke zwei Mal 

pro Tag benützt, soll nicht verletzt werden. Deshalb will die Bahn diese Risiken kennen und 

falls das Schutzziel verletzt ist, wissen wo Massnahmen am effizientesten umzusetzen sind. 


Abbildung 2: Schematische Darstellung der Risikoentwicklung entlang einer Verkehrsachse aufgrund unterschiedlicher Handlungsoptionen. 

METHODEN 

Risikoanalyse und -bewertung 

Um feststellen zu können, ob die Sicherheit auf dem Schienennetz der BLS Netz AG ausreichend 

ist, müssen zuerst die Gefährdung und das daraus resultierende Risiko im IST-Zustand 

wie auch der Soll-Zustand bekannt sein. Um den IST-Zustand zu erfassen liess die BLS 

Netz AG eine Vorstudie ausarbeiten, um eine Übersicht über die bestehende Gefährdung 

und die daraus resultierenden Risiken für Menschen und Sachwerte zu erlangen. Für die 

Gefahrenbeurteilung wurden sämtliche Einzugs- oder Liefergebiete von Naturgefahrenprozessen 

berücksichtigt, welche die Bahnlinie betreffen können. Die Ausarbeitung der 

Gefahren- und Risikoanalyse erfolgte nach dem Risikokonzept der nationalen Plattform 

Naturgefahren (PLANAT) (Bründl, 2009) und lässt sich wie folgt zusammenfassen: 

– Bestimmung der Liefergebiete und Definition der Szenarien für die Prozesse Sturz, Wasser, 

Lawinen und Rutschungen 

– Bestimmung der massgebenden Einwirkung entlang des Gleises in Form von Intensitätskarten 

für die spezifischen Szenarien 

– Erfassung von Details des Schadenpotenzials durch BLS Netz AG (Zugsfrequenz, Fahrgeschwindigkeit, 

Besetzungsgrad, Fahrgeschwindigkeit, usw.) 


– Bestimmung des Risikos streckenbezogen und für jede Prozessquelle 

Die Wirkung der bestehenden Schutzbauten, der bestehenden organisatorischen Überwachungsmassnahmen 

und des Schutzwaldes sind in die Beurteilung der Gefährdung und der 

Risiken eingeflossen. In der Risikoanalyse wurden folgende Schadenbilder berücksichtigt: 

– Schäden an Personen (Passagiere und Zugspersonal) Personenrisiken 

– Schäden infolge Räumung und Wiederherstellung der Gleisanlagen und dazugehörender 

Infrastruktur Sachrisiko Räumung und Wiederherstellung 

– Schäden am Rollmaterial Sachrisiko Rollmaterial 

– Kosten für Ersatzbusbetrieb Risiko Verfügbarkeit 

Mögliche Kollisionen mit einem Gegenzug nach einer Entgleisung wurden zwar evaluiert, 

aber in der Risikoanalyse nicht berücksichtigt. Zur Aufsummierung von Personen- und 

Sachrisiken wurde das kollektive Personenrisiko mit dem Grenzkostenansatz von 5 Mio. CHF 

pro statistischen Todesfall monetarisiert (Bründl, 2009). 

Mit der Risiko- und Massnahmenbewertung wird der SOLL-Zustand definiert. Die Risikostrategie 

der BLS Netz AG sieht - gestützt auf die Vorgaben des Kantons Bern - vor, dass das 

durch Naturgefahren verursachte zusätzliche individuelle Todesfallrisiko für Bahnreisende 

nicht mehr als 1 x 10-5 / Jahr betragen soll (=Schutzziel). 

Integrale Massnahmenplanung 

Massnahmen sollen nur dann geplant werden, wenn das Schutzziel verletzt ist. Im Rahmen 

der oben erwähnten Vorstudie wurden deswegen Massnahmenkonzepte gestützt auf die 

obigen Vorgaben für diejenigen Stellen ausgearbeitet, wo das Risiko am grössten ist. Gibt es 

für eine Gefahrenstelle mehrere Handlungsoptionen, so soll diejenige Massnahme oder 

Massnahmenkombination gewählt werden, die die grösste Nutzen-Kosten-Differenz aufweist 

und die auch mit den weiteren Kriterien im Sinne der Nachhaltigkeit im Einklang steht. Die 

Berechnung des Nutzens wie auch die Aufrechnung der jährlichen Kosten erfolgt nach dem 

Ansatz des Risikokonzeptes der PLANAT (Bründl, 2009). Für die Wahl der richtigen Schutzmassnahme 

sind vertiefte Prozesskenntnisse wie auch Kenntnis der lokalen Gegebenheiten 

vor Ort einerseits und Kenntnisse über die Wirkungsweise und Kosten der Massnahmentypen 

andererseits notwendig. 

ERGEBNISSE 

Risikoanalyse und -bewertung 

Der im Detail untersuchte Streckenabschnitt (siehe Abbildung 1) verläuft zwischen Frutigen 

und Kandersteg (km 13.5 bis km 33.7). Auf der rund 20 km langen untersuchten Strecke 

sind 15 km von Naturgefahren aus total 77 verschiedenen Prozessquellen betroffen. 

Die mit dem Grenzkostenansatz monetarisierten, kollektiven Personenrisiken machen rund 

2/3 und die Sachrisiken rund 1/3 des Gesamtrisikos aus. Die Aufteilung der Risiken auf die 

verschiedenen Schadenarten zeigt, dass die Personenrisiken, im Zusammenhang mit 

Sturzereignissen dominieren (rund 55% des Gesamtrisikos). Demgegenüber verursachen die 


Wasser- und Lawinenprozesse in erster Linie Sachrisiken infolge Räumung und Wiederherstellung 

der Infrastruktur. Die Schäden am Rollmaterial sind über alles gesehen untergeordnet 

(rund 15%); sie entstehen vor allem bei der Kollision mit abgelagertem Sturzmaterial. 

Die Risiken infolge Verfügbarkeit machen insgesamt nur rund 6% aus; sie entstehen in erster 

Linie nach Sperrungen infolge eines Ereignisses. Die Ergebnisse der Risikoanalyse zeigten 

hinsichtlich Schadenhäufigkeit eine gute Übereinstimmung mit dem Unfallgeschehen infolge 

Naturgefahren auf dieser Strecke. Die Zahlen zum Schadenausmass liessen sich nicht 

vergleichen, da die entsprechenden Zahlen aus der Vergangenheit fehlten. Aufgrund der 

bisherigen Unfallbilder wird das Schadenausmass in der Risikoanalyse vermutlich eher 

überschätzt. Das Schutzziel des individuellen Todesfallrisikos wird im IST-Zustand nicht 

erreicht, der Wert ist um einen Faktor 6 zu hoch. 

Dank der Analyse konnten besonders risikoreiche Streckenabschnitte und Liefergebiete 

identifiziert und für die Massnahmenplanung priorisiert werden. 

ERREICHEN DES ANGESTREBTEN SICHERHEITSNIVEAU MIT INTEGRALER KOMBINATION NEUER 

SCHUTZMASSNAHMEN 

Da das angestrebte Schutzniveau im IST-Zustand nicht erreicht wird, muss das Risiko 

reduziert werden. Dies kann durch neue Schutzmassnahmen geschehen. In der Planung 

stellte sich an verschiedenen Stellen heraus, dass bauliche Systeme in diesem steilen Gelände 

entweder nicht realisierbar (z. B. Steinschlagschutzdämme), in ihrer Wirkung begrenzt 

(z. B. Steinschlagschutznetze) oder finanziell sehr aufwändig (z. B. verstärkte Galerien, 

Tunnels) sind und deswegen nicht zur Umsetzung empfohlen werden können. An diesen 

Stellen sind Überwachungsmassnahmen, z. T in Kombination mit den bestehenden baulichen 

Massnahmen die kostenwirksamste Lösung. Das angestrebte Schutzziel kann nur erreicht 

werden, wenn alle möglichen Massnahmentypen sowie deren Kombination berücksichtig 

werden (siehe auch Abbildung 5): 

– Bauliche Massnahmen: Neubau von Schutzbauwerken im Umfang von 3 Mio CHF für den 

Zeitraum 2013 - 2017 

– Organisatorische Massnahmen: Messtechnische wie auch manuelle Überwachung von 

Felsinstabilitäten (vgl. Gruner & Utelli 2016) und Reinigung von gleisnahen Felswänden 

– Organisatorische Massnahmen: Elektrische Überwachung der linearen Schutzbauwerke 

entlang des Trassees 

Im Folgenden soll der letzte Massnahmentyp erläutert werden, da damit das Personenrisiko, 

welches den grössten Anteil des Risikos ausmacht, sehr effizient reduziert werden kann. 

Elektrische Reissdraht-Überwachung 

Auf vielen Streckenabschnitten ist beim Personenrisiko das Schadenbild "Kollision mit 

abgelagertem Sturzmaterial" dominierend. Deswegen erwies sich dort die Überwachung von 

bestehenden oder neuen Schutzbauwerken als sehr effiziente und auch effektive Massnahme. 

Die Überwachung detektiert Ereignisse, die die Kapazität der Schutzmassnahme überschrit- 


ten, diese durchschlagen und sich auf dem Gleis abgelagert haben (siehe Abbildung 3). 

Der nächste heranfahrende Zug wird gestoppt und mit Fahrt auf Sicht durch den betroffenen 

Streckenabschnitt geleitet, so dass jederzeit einen Bremsung möglich ist und somit eine 

Kollision verhindert werden kann. Das Risiko für Zugspendler wird damit signifikant gesenkt. 

Parallel dazu kontrolliert eine Equipe der BLS Netz AG die Schadenstelle und kann gegebenenfalls 

weitere Massnahmen einleiten. 

Abbildung 3: Elektrische Überwachung von Schutzbauwerken. Die in den Bremsen oder im Bauwerk eingebauten Drähte reissen im 

Überlastfall und leiten ein Signal an die Zugsleitzentrale. 

FAZIT INTEGRALE KOMBINATION NEUER SCHUTZMASSNAHMEN 

Mit den aktuell sich in der Realisierung befindenden zusätzlichen baulichen und organisatorischen 

Schutzmassnahmen wird das kollektive Risiko um einen Faktor 2, das individuelle 

Todesfallrisiko um mehr als einen Faktor 4 gesenkt. Rechnerisch wird das individuelle 

Todesfallrisiko noch bei rund 1.5 x 10-5 / Jahr liegen. Aufgrund verschiedener Unsicherheiten 

(Gefahren-Szenarien, Wiederkehrperioden von Ereignissen, Unfall-Szenarien, Berücksichtigung 

organisatorischer Massnahmen etc.) darf die Genauigkeit dieser Berechnung nicht 

überbewertet werden. Die rechnerisch präzise, abschliessende Antwort auf die Frage nach der 


ausreichenden Sicherheit ist darum nicht immer sinnvoll. Weitere bauliche Massnahmen sind 

hier nicht mehr kostenwirksam. Die Sicherheit wird aber auf alle Fälle erhöht und es darf 

trotz aller rechnerischen Unsicherheiten angenommen werden, dass die BLS Netz AG auf 

dieser Strecke heute das angestrebte Sicherheitsniveau wirksam erreicht. 

Halten des angestrebten Sicherheitsniveaus mit dem Erhalt von Schutzbauwerken 

Um das mit den bisherigen Massnahmen erreichte Schutzniveau dauerhaft sichern zu 

können, bedarf es eines gezielten Schutzbautenmanagements, bestehend aus einem vollständigen 

Schutzbautenkataster und der fortwährenden Erhaltungsplanung. 

Die BLS hat in ihren beiden Forst- und Felssicherungsequipen seit Jahrzehnten erfahrene 

Mitarbeitende für die Bauwerksüberwachung und die Unterhaltsarbeiten. In den letzten 

Jahren konnte mit der einheitlichen Ersterfassung der bestehenden Schutzbauwerke im 

Gelände und der Ablage in einem Schutzbautenmanagementsystem (geobasierte Datenbank) 

das Wissen über Lage und Zustand der Bauwerke dauerhaft gesichert und für andere 

Involvierte zugänglich gemacht werden (siehe Abbildung 4 links): die BLS Netz AG ist auf 

dem Streckennetz im Kanton Bern für aktuell 750 Schutzbauwerke zuständig. 

Die einheitliche Beurteilung des Zustands und des Handlungsbedarfs vor Ort sowie die 

Abschätzung der geeigneten Unterhaltsmassnahmen ermöglichten es erstmals, ein mehrjähriges 

integrales Erhaltungsprojekt (vgl. Pfammatter 2016) für das gesamte Berner Streckennetz 

der BLS Netz AG fundiert auszuarbeiten: im laufenden Erhaltungsprojekt 2012-2016 sind für 

Instandhaltungen und Instandsetzungen gut CHF 450‘000.- budgetiert. Pro Jahr ist die 

BLS-Forstequipe mehrere Wochen mit Unterhaltsarbeiten in verschiedenen Verbauungsgebieten 

entlang des Streckennetzes tätig. Aktuell handelt es sich um Instandsetzungen oder 

Rückbauten von älteren Trockensteinmauern, Ertüchtigungen von Stahl- und Holzbarragen, 

Instandhaltungen von Netzabdeckungen und Instandsetzungen aller Werktypen nach 

Ereignissen (siehe Abbildung 4 rechts). Die ausgeführten Arbeiten und die neue Zustandsbeurteilung 

werden im Schutzbautenmanagementsystem nachgeführt. Nach Abschluss des oben 

erwähnten Instandsetzungsprojektes sollten sich darin keine Schutzbauwerke mehr in 

alarmierendem oder in mittelfristig nicht tolerierbarem Zustand befinden. Die Zuverlässigkeit 

der Verbauungen ist dann bis zum jeweiligen Dimensionierungsszenario hoch. So können die 

Schutzbauwerke in der Gefahren- und Risikobeurteilung angemessen berücksichtigt werden. 

Mit dieser umfassenden, bei der BLS institutionalisierten und von Bund und Kanton mit 

Beiträgen unterstützen Erhaltungsplanung wird sichergestellt, dass die Schutzbauten 

möglichst lange eine hohe Zuverlässigkeit behalten. (siehe Abbildung 4) 

HALTEN DES ANGESTREBTEN SICHERHEITSNIVEAUS MIT DER PFLEGE DER SCHUTZWÄLDER 

Ein stabiler Schutzwald ist für den Schutz vor Naturgefahren eine wichtige und oft auch 

effizienteste Massnahme. Dem war und ist sich die BLS seit dem Bau der Lötschberg-Bergstrecke 

vor mehr als 100 Jahren bewusst. Deshalb hat sie schon sehr früh grossflächig 

das Land oberhalb der Gleise erworben und aktiv unbewaldete Gebiete aufgeforstet 


Abbildung 4: Die Lage der bestehenden Schutzbauten wird im Gelände mittels GPS erfasst, der Zustand des Bauwerks wird mittels 

visueller Kontrolle durch Fachleute beurteilt. Das Fundament dieser Betonmauer muss gesichert werden. (links). Diese Blocksteinmauer 

wurde mit armiertem Ortsbetonfundament und Drahtsteinkorb aufgrund der Zustandsbeurteilung instand gestellt(rechts). 

(Schwarz, 1996). Hier wird bis heute konsequent flächig Schutzwaldpflege betrieben. Diese 

Pflege wird nach der Methode Nachhaltige Pflege im Schutzwald NaiS (Frehner et al, 2005) 

in Waldbauprojekten geplant und mit jährlicher Festlegung des Arbeitsprogramms durch die 

eigene Forstequipe ausgeführt. 

FAZIT 

Dank der Gefahren- und Risikoanalyse beantwortet die BLS Netz AG periodisch die Frage, 

ob ein ausreichendes Sicherheitsniveau bezüglich Naturgefahren erreicht ist. Die Risikoanalyse 

liefert die Grundlage, um Abschnitte und Prozessquellen mit grossen Risiken zu identifizieren, 

für die integrale Massnahmenplanung zu priorisieren und um Massnahmen hinsichtlich 

Nutzen-Kosten-Verhältnis zu beurteilen. Die Einhaltung des Schutzziels muss periodisch 

überprüft werden. Die Überprüfung soll möglichst auf das ganze Streckennetz ausgedehnt 

werden. 

Erst dank der Kombination von baulichen, waldbaulichen und organisatorischen Massnahmen 

wird das angestrebte Schutzziel erreicht sowie die Verfügbarkeit der Strecke erhöht 

(siehe Abbildung 5). Um das erreichte Sicherheitsniveau langfristig zu erhalten, müssen der 

Unterhalt der Schutzbauwerke und die Pflege des Schutzwaldes sichergestellt sein. 


Abbildung 5: Das integrale Naturgefahrenmanagement der BLS Netz AG beinhaltet (von oben links im Uhrzeigersinn): Instandhaltung 

und Instandsetzung von bestehenden Bauwerken,Überwachung von instabilen Felspartien, Überwachung von bestehenden 

Schutzbauwerken, waldbauliche Massnahmen sowie die punktuelle Ergänzung mit neuen Schutzmassnahmen. 

Dank des integralen Lösungsansatzes waren für Bund und Kanton die Voraussetzungen 

gegeben, um all diese Massnahmen auch finanziell unterstützen zu können. 

LITERATUR 

- Bründl, M. (Ed.) (2009): Risikokonzept für Naturgefahren (Leitfaden). Nationale Plattform 

Naturgefahren PLANAT: 420 S. 

- Frehner, M.; Wasser, B.; Schwitter, R. (2005): Nachhaltigkeit und Erfolgskontrolle im 

Schutzwald. BAFU, Bern: 564 S. 

- Gruner, U., Utelli, H.-H. (2016) (in Review): Multiples Früherkennungssystem an Felswänden 

oberhalb einer Eisenbahnstrecke im Berner Oberland (Schweiz) 

- Pfammatter, Chr. (2016): Successful management of protective structures in long-term 

conservation projects. Interpraevent 2016 Extended Abstract: 3 S. 

- Schwarz, W. (1996): Schutz vor Naturgefahren auf der Nordrampe. BLS Lötschbergbahn, 

Bern: 126 S. 



Storm and Flood Warnings issued by Switzerland's 

Specialist Federal Agencies 

David Volken, PhD 1 , Coralie Amiguet 1 ; Therese Buergi 1 ; Daniel Murer 2 ; Christoph Schmutz, PhD 2 ; 

ABSTRACT 

Protecting the population against natural hazards is a key task of the state. To improve storm, 

flood and avalanche warnings and alerts, the NBCN Intervention Ordinance came into force 

at the federal level in 2011, the Alarm Ordinance (AlarmO) was amended and the Radio and 

Television Ordinance (RTVO) was expanded. These ordinances introduced comprehensive 

provisions governing the responsibilities for natural hazard warnings and their communication 

to the authorities and the population. As a result, MeteoSwiss is responsible for weather 

hazards, the FOEN is responsible for floods and forest fires, and the WSL is responsible for 

avalanches. When hazardous events occur, these specialist federal agencies work closely 

together and distribute their warning products to cantonal authorities. The federal Special 

Natural Hazards Committee may become active in the event of large-scale level 4 or 5 natural 

hazards. This committee was first created when the flooding occurred in early June 2013. 

KEYWORDS 

Warning; Storm; flood; natural hazard 

EINFÜHRUNG 

Der Schutz der Bevölkerung und ihrer Lebensgrundlagen vor Naturgefahren ist eine zentrale 

Aufgabe des Staates. Mit dem integralen Risikomanagement, welches alle Handlungsoptionen 

zur Verbesserung der Sicherheit nutzt, soll den Naturgefahren mit ihren grossen Unsicherheiten 

begegnet werden. Öffentlichkeit und Politik fordern einen besseren Schutz vor Naturgefahren 

und insbesondere eine frühzeitige Warnung und Alarmierung (Hess et al., 2012). 

Durch eine gesamtheitliche und vernetzte Beurteilung der aktuellen Gefahrensituation und 

einer optimierten zeitlich-räumlichen Auflösung der Meteo- und Abflussvorhersagen und 

deren Beurteilung lassen sich die Qualität der Warnungen und Orientierungen der Behörden 

und der Bevölkerung verbessern (Hess et al., 2012). 

METHODEN 

Die Inkraftsetzung der ABCN-Einsatzverordnung und der Alarmierungsverordnung (AV) 

sowie die Ergänzung der Radio- und Fernsehverordnung (RTVV) am 1. Januar 2011 hat 

die Zuständigkeit für die Warnungen vor Naturgefahren sowie deren Vermittlung an die 

Behörden und die Bevölkerung umfassend geregelt (Wernli-Schärer et al., submitted). 

Die Fachstellen des Bundes für Naturgefahren wurden für alle relevanten Gefahrenarten 

1 Bundesamt für Umwelt, Ittigen, SWITZERLAND, david.volken@bafu.admin.ch 

2 Bundesamt für Meteorologie und Klimatologie, MeteoSchweiz 



estimmt, die Warnskala wurde für alle Stellen auf fünf Gefahrenstufen (keine oder geringe 

(grün), mässige (gelb), erhebliche (orange), grosse (rot), sehr grosse Gefahr (dunkel rot)) 

festgelegt und die Zusammenarbeit zwischen Bund und Kantonen wurde optimiert. 

Im Falle von grosser und sehr grosser Gefahr können zudem alle konzessionierten Radio- und 

Fernsehanbieter zur Verbreitung der Warnungen verpflichtet werden. Die Lehren aus 

früheren Ereignissen (z.B. Lawinenwinter und Frühjahrs-Hochwasser 1999 (BWG, 2000), 

extreme Hochwasserereignisse 2005 (Bezzola et al., 2007 und 2008) und 2007 (Bezzola et al., 

2009)) führten im Rahmen des Lenkungsausschusses Naturgefahren (LAINAT) zum Ausbau 

und zur Institutionalisierung einer verbesserten und engeren Zusammenarbeit zwischen den 

Fachstellen des Bundesamts für Umwelt (BAFU), MeteoSchweiz, WSL-Institut für Schneeund 

Lawinenforschung SLF, dem Schweizerischen Erdbebendienst sowie dem Bundesamt für 

Bevölkerungsschutz (BABS) (Wernli-Schärer et al., submitted). 

WETTERWARNUNGEN 

Warnungen vor den Gefahren des Wetters haben in der Schweiz Tradition. Die Ausgabe von 

Sturmwarnungen für die ersten Schweizer Seen wird bereits seit etwa 80 Jahren praktiziert. 

Im Jahr 2000 wurde mit dem Bundesgesetz über die Meteorologie und Klimatologie (MetG) 

der Warnauftrag umfassender definiert und dem Bundesamt für Meteorologie und Klimatologie 

MeteoSchweiz zugewiesen. In den Folgejahren wurden die Wetterwarnungen inhaltlich, 

räumlich und zeitlich weiterentwickelt und verfeinert. Darin eingeschlossen ist die Absprache 

mit den Empfängern, insbesondere den Kantonsbehörden. Heute gibt es Warnungen vor 

Wind, Gewitter, Regen, Schnee, Strassenglätte, Hitze und Frost. Die Warnungen werden in 

einer hohen Auflösung für 159 Warnregionen (vgl. Abb. 1) und 51 Warnobjekte (Seen und 

Flughäfen) erstellt. Die Behörden erhalten in Fällen, wo ein signifikantes Ereignis ab der 

Stufe 3 mit genügend Sicherheit (40%-70% Wahrscheinlichkeit) und einem Vorlauf von bis 

zu drei Tagen erwartet wird, eine sogenannte Vorwarnung. Für die Bevölkerung gibt es ein 

ähnliches Produkt: Der Warnausblick ist eine Vorschau auf die erwarteten Unwetterwarnungen 

der nächsten Tage. Er wird auf den Warnseiten des Internetauftritts und der mobilen App 

von MeteoSchweiz schraffiert dargestellt, um ihn von den effektiv ausgegebenen Warnungen 

zu unterscheiden (vgl. Abb. 1). 

HOCHWASSERWARNUNGEN 

Das Bundesamt für Umwelt (BAFU) hatte vor 2011 keinen gesetzlichen Auftrag, vor 

Hochwasser zu warnen. Statt offizieller Warnungen wurden im Bedarfsfall Hochwasserinfos 

für genau definierte Abflussmessstellen an nationale, kantonale und private Kunden, wie 

zum Beispiel die Behörden des Kantons Aargau oder die Rheinschifffahrt in Basel, übermittelt. 

Mit der Revision der Alarmierungsverordnung wurde das BAFU verpflichtet, für 

sämtliche grössere Gewässer in der Schweiz hydrologische Vorhersagen zu erstellen und im 

Hochwasserfall die Behörden zu warnen. In Absprache mit den Kantonen wurden die 

Gewässer (Flüsse und Seen) definiert, für welche der Bund Warnungen ausgibt (vgl. Abb. 2). 

Es handelt sich dabei um nationale Gewässer, entlang welchen ein grosses Schadenpotential 


Abbildung 1: Darstellung des Warnausblickes (schraffierte Flächen) und der Warnungen (ausgefüllte Flächen) in der mobilen App von 

MeteoSchweiz. Die feinen, weissen Linien trennen die einzelnen Warnregionen. Hier eine Gewittersituation. 

Abbildung 2: Karte mit den Gewässern von gesamtschweizerischem Interesse. 


vorhanden ist. Es sind auch Flüsse, die durch mehrere Kantone fliessen oder internationale 

Fliessgewässer sowie Seen, an welche mehrere Kantone grenzen oder welche die Grenze zu 

Nachbarländern bilden. 

Ein regionales Hochwasserereignis im Kander- und Lötschental hat am 11. Oktober 2011 

Millionenschäden verursacht. Da diese Regionen keine Flüsse von nationaler Bedeutung 

aufweisen, konnte beim Hochwasser vom 11. Oktober 2011 keine Warnung herausgegeben 

werden. Als Reaktion auf dieses Ereignis dehnte das BAFU das Konzept der Warnungen für 

nationale Gewässer auf kleinere und mittlere Fliessgewässer (regionale Hochwasserwarnungen) 

aus. Seit dem erstellt das BAFU regionale Hochwasser-Warnungen für 26 Regionen. 

Die Gefahrenstufen geben Auskunft über die Intensität des Ereignisses, die möglichen 

Auswirkungen und Verhaltensempfehlungen. Die Schwellenwerte, die die Gefahrenstufen 

abgrenzen, werden ausgehend vom vorhandenen Wissen über das Verhalten des jeweiligen 

Fliessgewässers festgelegt (Pegel ab dem das Gewässer über die Ufer tritt, ab den ersten 

Schäden eintreten usw.). Diese Schwellenwerte entsprechen in etwa der Jährlichkeit von 

Hochwasserereignissen, also einer Wiederkehrperiode von durchschnittlich 2, 10, 30 oder 

100 Jahren. 

Bei den regionalen Hochwasserwarnungen, die für kleine und mittlere Flüsse mit Einzugsgebieten 

bis 300 km 2 ausgegeben werden, gibt es zurzeit 2 Stufen (grün = keine oder geringe 

Gefahr, gelb = Hochwasserdisposition an kleinen und mittleren Gewässern). Diese werden 

voraussichtlich ab Sommer 2016 auf 4 Stufen ausgedehnt. 

ZUSAMMENARBEIT DER FACHSTELLEN 

Zeichnet sich heute beispielsweise ein markantes Niederschlags-Hochwasser-Ereignis ab, 

finden zwischen den Prognostikern des BAFU und der MeteoSchweiz regelmässige Telefonkonferenzen 

zur Abstimmung der Lage und der Warntätigkeit statt. Während MeteoSchweiz 

ohnehin einen 24-Stundenservice pflegt, intensiviert das BAFU in diesen Fällen den 

Vorhersagedienst. Spielt beim Ereignis auch die Schneeschmelze eine wichtige Rolle, wird der 

schneehydrologische Dienst des SLF zugeschaltet. Darüber hinaus kommen bei grossen 

Ereignissen die Führungsorganisationen der jeweiligen Bundesämter zum Einsatz. Bei Bedarf 

werden diese ämterübergreifend im Fachstab Naturgefahren zusammengeschlossen. 

Warnungen werden verfasst und via die Nationale Alarmzentrale (NAZ) über gesicherte 

Vermittlungskanäle an die Einsatzzentralen der betroffenen Kantone verschickt. Zudem wird 

seit 2014 ein gemeinsames Naturgefahrbulletin des Bundes erstellt, welches die gesamte 

Naturgefahrensituation aller Fachstellen beschreibt. 

Die Naturgefahrenfachstellen des Bundes stellen Messungen, Beobachtungen sowie Vorhersagen 

und Warnprodukte auf der Gemeinsamen Informationsplattform Naturgefahren (GIN) 

primär den kantonalen Behörden zur Verfügung. Das Naturgefahrenportal (www.naturgefah- 


en.ch) fasst seit 2014 die aktuellen Warnungen des Bundes, Medienmitteilungen und die 

Naturgefahrenbulletins für jedermann öffentlich zugänglich zusammen (Neversil et al., 

submitted). Das Naturgefahren-Portal bietet zudem mobile Seiten an. Damit können die 

wichtigen Informationen auch unterwegs abgerufen werden. 

MeteoSchweiz publiziert seit dem Herbst 2015 sämtliche Naturgefahrenwarnungen des 

Bundes auf der mobilen App. Die Benutzer können sich mit der App nicht nur die aktuellen 

Unwetterwarnungen sondern sämtliche Naturgefahrenwarnungen als Push-Nachricht 

schicken lassen. Damit sind sie laufend über die aktuelle Naturgefahrensituation im Bild. 

Die App von MeteoSchweiz wurde 3.6 Mio. Mal heruntergeladen (Stand August 2015) 

(Neversil et al., submitted). Damit ist die Reichweite der Naturgefahrenwarnungen gross. 

Muss die Bevölkerung rasch vor einer sehr grossen Gefahr gewarnt werden, so steht den 

Fachstellen seit dem Jahr 2011 das Mittel der verbreitungspflichtigen Bevölkerungswarnung 

zur Verfügung (AV und RTVV). Diese Warnungen müssen dann von den verpflichteten 

Rundfunksendern über Fernsehen und/oder Radio rasch verbreitet werden. Dieses Instrument 

gilt es verantwortungsvoll einzusetzen. Bisher wurde erst einmal im August 2011 eine 

verbeitungspflichtige Warnung ausgegeben (Gewitter). 

Grosse Naturgefahrenereignisse werden heute von den Fachstellen des Bundes eng geführt 

begleitet. In den Naturgefahrenfachstellen des Bundes bestehen Führungs- oder Einsatzorganisationen, 

welche zeitnah ein konsistentes Handeln der Fachstellen gewährleisten. 

Bei MeteoSchweiz ist die EO Met (Einsatzorganisation MeteoSchweiz) und beim BAFU die 

FO BAFU (Führungsorganisation) erreichbar und wird seit 2012 bei Warnungen der Stufen 

4 und 5 beigezogen, resp. hochgefahren. Bei sehr grossen und/oder kombinierten Naturgefahrenereignissen 

übernimmt der übergeordnete Fachstab Naturgefahren die Führung in 

der Ereignisbewältigung. Die Prozesse des Fachstabs Naturgefahren sind sowohl mit den 

Prozessen der Fachstellen als auch mit dem übergeordneten Bundesstab ABCN abgestimmt. 

Der Bundesstab ABCN kommt unter anderem bei grossen Naturkatastrophen zum Einsatz 

und erhält die Fachinformation von den Naturgefahrenfachstellen koordiniert durch den 

Fachstab Naturgefahren. 

ZUSAMMENARBEIT MIT DEN KANTONEN 

Die Grundlagen der Hochwasser-Warnungen wurden beim Bundesamt für Umwelt vor 

Inkraftsetzung der revidierten Alarmierungsverordnung erarbeitet. Dabei wurden das 

Gewässernetz und die jeweiligen Gefahrenstufen definiert. In regelmässig stattfindenden 

Absprachen zwischen BAFU und den Kantonen werden die Gefahrenstufen und das 

Gewässernetz überprüft und punktuell angepasst. Im Hochwasserfall finden regelmässig 

telefonische Absprachen zwischen der Hochwasser-Vorhersagezentrale des BAFU und der 

Kantone statt. Zusätzlich zu den Warnungen und Naturgefahrenbulletins des BAFU, ist es für 

die Kantone wichtig, dass sie mit den Prognostikern des BAFU im telefonischen Kontakt die 


kurzfristige Entwicklung der Gewässer diskutieren können, um eine fundierte Einschätzung 

der Lage in ihrem Kanton zu erhalten. Gerade mit den Kantonen Bern, Luzern und Zürich, 

die für die Seeregulierungen an Bieler-, Vierwaldstätter- und Zürichsee zuständig sind, ist der 

gegenseitige Infofluss betreffend Seeregulierung und Steuerung der Schleusen sehr zentral. 

Weiter stellen das BAFU und die Kantone während eines Ereignisses mit gegenseitigen 

Absprachen eine widerspruchsfreie Kommunikation zur Bevölkerung und den Medien sicher. 

Für den gegenseitigen Austausch, finden zwischen den Bundesämtern MeteoSchweiz, BAFU 

und dem SLF sowie den Kantonen jährlich im Herbst eine Warnkonferenz statt. Die Warnkonferenz 

richtet sich vornehmlich an die Naturgefahrenfachstellen und die Führungsorganisationen 

der Kantone. Das Ziel dieser Warnkonferenzen ist der Rückblick auf das vergangene 

Warnjahr, die Beurteilung der Warnungen der Bundesfachstellen sowie der Gedankenaustausch 

zwischen Bund und Kantonen. 

HOCHWASSEREREIGNIS VOM JUNI 2013 

Das Hochwasserereignis von anfangs Juni 2013 lässt sich hydrologisch und meteorologisch 

mit demjenigen von 1999 vergleichen, auch was die Vorgeschichte betrifft: Zwischen dem 

31. Mai und dem 2. Juni 2013 fielen im Mittelland in 48 Stunden zwischen 80 und 150 mm 

Niederschlag. In einem Streifen von den Schwyzer Alpen bis ins Appenzell wurden 150 bis 

200 mm gemessen, rund um den Säntis sogar bis 250 mm. Lokal wurden Wiederkehrperioden 

von 100 bis 200 Jahren beobachtet. Die Fliessgewässer und Seen sind in den betroffenen 

Regionen markant angestiegen. Verbreitet wurden Abflüsse mit Wiederkehrperioden von 

2 bis 10 Jahren, lokal bis 50 Jahren beobachtet. An der Reuss, der Thur sowie am Hochrhein 

wurde die Gefahrenstufe 3 (HQ 10 

bis HQ 30 

) erreicht. In der Ostschweiz traten einige Gewässer 

über die Ufer und überschwemmten Agrarland und teilweise Siedlungsgebiete. 

Auf dem Hochrhein und der Aare musste die Schifffahrt eingestellt werden. 

Die Beurteilung der Wetter- und Hochwasserentwicklung durch die Fachstellen des Bundes 

(MeteoSchweiz, BAFU und SLF) fand ab dem 29. Mai jeden Morgen in Form einer Telefonkonferenz 

statt. MeteoSchweiz warnte ab dem 30. Mai vor Starkregen der Gefahrenstufe 4, 

das BAFU ab dem 31. Mai vor Hochwasser der Gefahrenstufe 2 bzw. 3. Die Warnstufen 

wurden während des Ereignisses laufend den neusten Entwicklungen angepasst (vgl. Abb. 3) 

Die Warnprodukte wurden von MeteoSchweiz und dem BAFU via dem offiziellen Verbreitungsweg 

über die Nationale Alarmzentrale NAZ und den Kantonspolizeien direkt an die 

Behörden der Kantone verbreitet. Das Naturgefahrenbulletin mit den umfassenden Informationen 

zum Ereignis wurde vom 31.5 bis 2.6. täglich oder sogar 2-mal täglich aktualisiert. 

Am 31. Mai wurde zwischen BAFU und MeteoSchweiz entschieden, dass der Fachstab für 

Naturgefahren des Bundes aufgeboten werden soll. Gleichentags fand noch der erste 

gemeinsame Rapport statt. Es wurde entschieden, ein Pressecommuniqué auf die Abendnachrichten 

zu publizieren und auf eine verbreitungspflichtige Bevölkerungswarnung zu 


Abbildung 3: Übersicht über die Produkte der Fachstellen für Naturgefahren bzw. des Fachstabs Naturgefahren während des 

Hochwassers vom 31.5-3.6.13. 

verzichten. Zur laufenden Beobachtung und Beurteilung der Hochwassersituation fanden in 

der Folge die Telefonkonferenzen jeweils zweimal täglich statt. Parallel zu den Mess-, 

Überwachungs-, Beurteilungs- und Warntätigkeiten setzten die Naturgefahrenfachstellen des 

Bundes auf eine kontinuierliche, aktive und transparente Medienkommunikation. Mit diesen 

Massnahmen konnten nachweislich Schäden verhindert oder vermindert werden (Buser et 

al., 2013). 


Mit der Revision der Alarmierungsverordnung wurden die rechtlichen Grundlagen für den 

Warnprozess vor Naturgefahren geschaffen. Die Bundesämter BAFU, MeteoSchweiz und SLF 

arbeiten seit dem Jahr 2011 in Unwetterlagen intensiv zusammen und setzten ihre Warnprodukte 

an die Kantonsbehörden ab. Beim Hochwasserereignis im Jahr 2013 konnte die 

Zusammenarbeit unter den verschiedenen Bundesämtern in Echtzeit erfolgreich überprüft 

werden. Dabei konnten die erarbeiteten Konzepte und organisatorischen Massnahmen 

erfolgreich in der Praxis getestet werden. 


LITERATUR 

- Bezzola G.R., Hegg Ch. (Ed.) (2007): Ereignisanalyse Hochwasser 2005, Teil 1 – Prozesse, 

Schäden und erste Einordnung. Bundesamt für Umwelt BAFU, Eidgenössische Forschungsanstalt 

WSL. Umwelt-Wissen Nr. 0707. 

- Bezzola G.R., Hegg Ch. (Ed.) (2008): Ereignisanalyse Hochwasser 2005, Teil 2 – Analyse von 

Prozessen, Massnahmen und Gefahrengrundlagen. Bundesamt für Umwelt BAFU, Eidgenössische 

Forschungsanstalt WSL. Umwelt-Wissen Nr. 0825. 

- Bezzola G.R., Ruf W., Jakob A. (Ed.) (2009): Ereignisanalyse Hochwasser August 2007 – 

Analyse der Meteo- und Abflussvorhersagen; vertiefte Analyse der Hochwasserregulierung 

der Jurarandgewässer. Bundesamt für Umwelt. Umwelt-Wissen Nr. 0927. 

- Buser M. et al. (2013): Hochwasserereignis an der Alpennordseite vom 31. Mai bis 3. Juni 

2013. 

- BWG (Bundesamt für Wasser und Geologie), (2000): Hochwasser 1999, Analyse der 

Ereignisse. Studienbericht Nr. 10/2000. Bundesamt für Wasser und Geologie, Bern. 

- Hess J., Schmid F. (2012): Towards optimised early warning. Developments in Switzerland. 

12th Congress Interpraevent 2012, Grenoble France. 

- Neversil B. et al. (submitted): A better informed public through the Swiss Confederation’s 

joint natural hazards portal. Interpraevent 2016 (Poster) 

- Wernli-Schärer L., Bialek R., Buser M., Flury Ch., Gerber B., Haslinger F., Hegg Ch., Ottmer 

B., Overney O., Romang H., Schmutz Christoph, Schweizer J. (submitted): Strategien zur 

Reduktion der Naturgefahrenschäden durch optimierte Warnung, Alarmierung und Intervention 

in der Schweiz. Interpraevent 2016 

- SR 520.12, Alarmierungsverordnung (AV) 

- SR 784.401 Radio- und Fernsehverordnung (RTVV) 

- SR 429.1 Bundesgesetz über die Meteorologie und Klimatologie (MetG) 



Strategies for the reduction of natural hazards 

damages by optimized warning, alarming and 

intervention in Switzerland 

Lilith Wernli-Schärer, MSc 1 ; Roland Bialek 2 ; Martin Buser 1 ; Christoph Flury 2 ; Bruno Gerber 3 ; Florian Haslinger 4 ; Christoph 

Hegg 5 ; Birgit Ottmer 7 ; Olivier Overney 1 ; Hans Romang 1 ; Christoph Schmutz 6 ; Jürg Schweizer 7 

ABSTRACT 

With the goal of protecting the population more effectively against natural hazards, the 

Federal Council initiated a project to optimise warnings and alerts in the event of natural 

hazards, known as the OWARNA project (Optimierung von Warnung und Alarmierung bei 

Naturgefahren). OWARNA has made it possible to implement measures to improve flood 

forecasts, such as their quality and availability, strengthen and standardise cooperation at the 

federal level, provide better information to local authorities and the public, and train local 

natural hazard advisors. The significant progress achieved through this project contributes 

to a functioning warning system. Future challenges are to establish crisis-proof forecast and 

warning systems as well as increase the readiness of the population to handle warnings 

properly. To meet these challenges, the authorities and the population will essentially need 

to better understand the potential impacts of natural hazards. In addition, there is a need for 

standardisation among suppliers of storm warnings. 

KEYWORDS 

early warnings and alerting, preparedness, response, national coordination 

AUSGANGSLAGE 

Der Schutz der Bevölkerung und ihrer Lebensgrundlagen vor Naturgefahren ist eine zentrale 

Aufgabe des Staates. Zum Schutz vor Naturgefahren werden in der Schweiz pro Jahr 

insgesamt rund 2.9 Milliarden Franken aufgewendet, was rund 0.6 Prozent des Bruttoinlandproduktes 

entspricht (BAFU, 2015). Die Hochwasserereignisse vom Sommer 2005 führten in 

der Schweiz zu Schäden von rund 3 Milliarden Franken und forderten sechs Todesopfer 

(Bezzola & Hegg, 2008). Mit diesem Ereignis wurden Probleme und Grenzen der Vorhersage 

von seltenen Naturereignissen offengelegt (Bezzola & Hegg, 2007). Die Erkenntnisse aus dem 

Jahr 2005 wurden durch die Hochwasserereignisse vom Sommer 2007 bestätigt. 

1 Geschäftsstelle LAINAT, Ittigen, SWITZERLAND, lilith.wernli-schaerer@bafu.admin.ch 

2 BABS 

3 Kanton Bern 

4 SED 

5 WSL 

6 MeteoSchweiz 

7 SLF 



Diese Ereignisse führten zu Unwetterschäden von insgesamt 710 Millionen Franken und 

forderten vier Todesopfer (Bezzola & Ruf, 2009). Die Schäden wurden nicht alleine vom 

Ausmass des Ereignisses beeinflusst. Eine effiziente Prävention, Intervention sowie Schadenbewältigung 

wirken sich ebenso erheblich auf das Schadensausmass aus (Bezzola & Hegg, 

2008). Organisatorische Massnahmen wie Sperrungen gefährdeter Gebiete, der Einsatz 

mobiler Hochwasserschutzelemente oder die Evakuation gefährdeter Personen können im 

Vergleich zu baulichen Massnahmen schneller umgesetzt werden. Sie sind bei selten 

auftretenden Gefahrenprozessen oft auch kostengünstiger, da ein grosser Teil der Kosten nur 

dann anfällt, wenn sie zum Einsatz kommen (Baumgartner, 2015). Darin besteht aber auch 

ihr grösster Nachteil: Sie erfordern eine Intervention vor dem Ereignis, um ihre Wirkung zu 

entfalten. Voraussetzung ist deshalb eine rechtzeitig vorliegende, zuverlässige und verständliche 

Information über die zukünftige Entwicklung der Naturgefahrenprozesse, damit die 

lokalen Interventionskräfte einen Einsatz verhältnismässig führen können. Es stellte sich 

deshalb die Frage, mit welchen Massnahmen die Warnungen vor Naturgefahren optimiert 

werden können, um durch vermehrte vorsorgliche Intervention Personen zu schützen und 

Sachschäden zu vermeiden. 

ORGANISATIONALER RAHMEN 

Naturgefahren werden in der Schweiz mittels einer zwischen den drei Staatsebenen (Bund, 

Kantone und Gemeinden) geteilten Verantwortung angegangen. Die Naturgefahrenfachstellen 

des Bundes informieren und warnen Bevölkerung, Medien sowie kantonale und 

kommunale Behörden vor drohenden Naturgefahren. Die Fachstellen des Bundes behandeln 

dabei folgende Gefahrenbereiche (Alarmierungsverordnung, SR 520.12, Art. 9): 

– Gefährliche Wetterereignisse: das Bundesamt für Meteorologie und Klimatologie 

– Hochwasser und damit verbundene Rutschungen sowie Waldbrände: das Bundesamt für 

Umwelt 

– Lawinengefahren: das Institut für Schnee- und Lawinenforschung der Eidgenössischen 

Forschungsanstalt für Wald, Schnee und Landschaft 

– Erdbeben: der Schweizerische Erdbebendienst 

Das Bundesamt für Bevölkerungsschutz unterstützt jene Stellen, die in der Vorbeugung 

kollektiver Risiken und in der Ereignisbewältigung tätig sind, insbesondere betroffene 

Bundesstellen, die Kantone und die Partnerorganisationen des Verbundsystems Bevölkerungsschutz. 

Die zuständigen Bundesbehörden haben fünf einheitliche Warnstufen eingeführt. Diese sind 

in der Alarmierungsverordnung verankert und basieren auf der fünfteiligen europäischen 

Lawinengefahrenskala, auf welche sich die Lawinenwarndienste der Alpenländer 1993 

geeinigt haben (SLF, 2015). Sind in einer Gefahrensituation mehrere Fachstellen zuständig, 

erlassen sie gemeinsame Warnungen. Die Federführung wird dabei von der Fachstelle am 

Ende der Prozesskette übernommen. Die zuständige Fachstelle übermittelt Warnungen der 


Nationalen Alarmzentrale, welche diese an die kantonalen Behörden weiter leitet (Alarmierungsverordnung, 

SR 520.12, Art. 9). Die Kantone verantworten die Notfallplanung und die 

Gemeinden planen und implementieren präventive Massnahmen. 

ZIELSETZUNGEN DES PROJEKTES OWARNA UND DIE STRATEGIE DES BUNDES 

Wie die Ereignisanalyse der Hochwasser 2005 gezeigt hat, kann das Schadensausmass von 

Naturgefahren mit einer optimierten Warnung, Alarmierung und Intervention um 20 Prozent 

reduziert wer-den, wenn Warnung, Alarmierung und Information rechtzeitig erfolgen und 

damit Menschen und Sachwerte rechtzeitig in Sicherheit gebracht werden können (Hess & 

Schmid, 2012). Zudem hat die Analyse der Hochwasser 2005 und 2007 aufgezeigt, dass 

personelle Defizite zur Aufrechterhaltung der kritischen Geschäftsprozesse in Notfallsituationen 

behoben werden müssen. So hat der Bundesrat in Folge der Ereignisse 2005 das Projekt 

OWARNA initiiert. Konkret hat der Bundesrat im Jahr 2007 Massnahmen zur Verbesserung 

der Hochwasservorhersage sowie zur besseren Information der kantonalen und kommunalen 

Behörden und Bevölkerung beschlossen. Ebenso wurde die Notwendigkeit einer umfassenden 

Notfallplanung bzw. eines Business Continuity Management (BCM) bei allen am 

Warnprozess beteiligten Fachstellen des Bundes erkannt. Ziel ist es, im Ereignisfall einen 

24-Stunden Vorhersagebetrieb sowie die fachliche Vernetzung – und damit die fachliche 

Beratung im Ereignisfall – sicherzustellen. In umfangreichen Analysen und Workshops 

zwischen den zuständigen Bundesfachstellen und den Fachstellen und Führungsorganisationen 

der Kantone wurde die Aufgabenteilung zwischen Bund und Kantonen geklärt. Es zeigte 

sich, dass bei zahlreichen Fragestellungen die Verantwortung auf Stufe Bund auf mehrere 

Organisationen verteilt ist. Um die nötige Koordination zwischen den Bundesstellen 

durchführen zu können, haben sich die Bundesfachstellen für Naturgefahren 2008 zum 

Lenkungsausschuss Intervention Naturgefahren (LAINAT) zusammengeschlossen. Der 

LAINAT ist seit dem 1. April 2009 operativ und ermöglicht es, politische, strategische und 

fachliche Fragestellungen innerhalb institutionalisierter Abläufe nachhaltig zu klären sowie 

die Vorhersage- und Warntätigkeit der Bundesstellen zu koordinieren und kontinuierlich 

weiterzuentwickeln. 

Kurz zusammengefasst verfolgt der Bund damit die Strategie, einen effektiv funktionierenden 

Warnprozess zu etablieren und zu erhalten. Im Folgenden wird der aktuelle Stand der 

Umsetzung der wichtigsten OWARNA-Massnahmen beschrieben. 

OPTIMIERTE WARNUNG, ALARMIERUNG UND INTERVENTION 

Um die Ziele von OWARNA zu erreichen, braucht es einen funktionierenden Warnprozess. 

Jedes einzelne Element trägt wesentlich zum Funktionieren des Gesamtsystems bei (Hess & 

Schmid, 2012). Im Folgenden werden die entsprechenden Elemente sowie der aktuelle Stand 

der Umsetzung der OWARNA-Massnahmen beschrieben: 


Verbesserung der Vorhersagesysteme: Eine verbesserte räumliche und zeitliche Auflösung 

der Vorhersagemodelle ermöglicht einerseits die Qualität und Aussagekraft der 

Wetter- und Abflussvorhersagen zu verbessern und andererseits Warnungen räumlich und 

zeitlich präziser zu verfassen. MeteoSchweiz hat das gesamte Netzwerk der Niederschlagsradars 

erneuert und arbeitet gegenwärtig daran, die radartechnisch bisher schlecht abgedeckten 

inneralpinen Gebiete (Wallis und Graubünden) mit zwei zusätzlichen, hochalpinen Radarstationen 

zu versorgen. Eine dieser Radarstationen (Pointe de la Plaine Morte) konnte bereits im 

Frühjahr 2014 dem Betrieb übergeben werden. Die fünfte Radarstation auf dem Weissfluhgipfel 

steht seit Oktober 2015 im Testbetrieb. Gleichzeitig wird das Niederschlags-Bodenmessnetz 

von MeteoSchweiz ausgebaut und automatisiert. Bei der Verbesserung der Vorhersagesysteme 

wurden massgebende Fortschritte erzielt. So können nun für nationale 

Fliessgewässer von gesamtschweizerischem Interesse sowie auch für kleine und mittlere 

Gewässer hydrologische Vorhersagen gerechnet werden (vgl. Amiguet et al. 2016). Eine 

verbesserte räumliche und zeitliche Auflösung der Vorhersagemodelle ermöglicht eine 

Optimierung der Qualität und Aussagekraft der Wetter- und Abflussvorhersagen sowie 

räumlich und zeitlich präzisere Warnungen. Im Projekt COSMO-NeXt erhöht MeteoSchweiz 

die Auflösung des Wettervorhersagemodells auf 1 Kilometer Maschenweite (COSMO-1) und 

stellt Ensemble-Rechnungen mit 2 Kilometer Maschenweite mit 21 Membern zur Verfügung 

(COSMO-E). Für die aktuelle Gefahreneinschätzung und zeitnahe Warnungen ist das 

sogenannte Nowcasting wichtig. Neue Systeme auf der Basis der Messinfrastruktur (insbesondere 

Niederschlagsradar) erlauben die rasche und räumlich hochaufgelöste Erstellung und 

Verbreitung von Gewitterwarnungen (vgl. Gaia et al.). Damit werden die erfolgreichen 

Entwicklungen fortgesetzt, welche 2003 im Thunderstrom Radar Tracking (TRT) ihren 

Ursprung haben. 

Intensivierung und Standardisierung der Zusammenarbeit auf Bundesebene: Oft 

handelt es sich um kombinierte Ereignisse, beispielsweise wenn langanhaltende Niederschläge 

mit einer intensiven Schneeschmelze zu Hochwasser führen. Um solche Ereignisse in ihrer 

Gesamtheit zu erfassen, wurde die Zusammenarbeit zwischen den Naturgefahrenfachstellen 

des Bundes intensiviert und standardisiert. So wurden 2014 auf technischer Ebene gemeinsame 

Standards zum Austausch von Warnungen zwischen den Fachstellen und Dritten 

vereinbart. Auf inhaltlicher Ebene vermittelt das sogenannte Naturgefahrenbulletin seit 2014 

bei besonders kritischen Lagen weiterführende Informationen wie eine gemeinsame 

Lagedarstellung. Es wird entweder einzeln durch die betroffene Fachstelle oder im Falle eines 

kombinierten Ereignisses gemeinsam durch die Fachstellen bzw. den Fachstab Naturgefahren 

erstellt. Die fachlichen Absprachen zwischen den Naturgefahren-Fachstellen des Bundes 

finden im „Fachstab Naturgefahren“ statt. So kann sichergestellt werden, dass die vorhandene 

Expertise und Erfahrung in die Lagebeurteilung einfliesst und widersprüchliche Aussagen 

vermieden werden. Der Fachstab Naturgefahren wurde 2010 geschaffen, ist rechtlich 

verankert (ABCN-Einsatzverordnung) und nimmt folgende Aufgaben wahr: 


– Er sammelt und interpretiert die Erkenntnisse der Partner und fasst sie zu einer Gesamtbeurteilung 

zusammen, die unter allen Fachstellen abgestimmt ist. 

– Er verfasst zuhanden der kantonalen Behörden und der Bevölkerung gemeinsame 

Bulletins, Warnungen und Verhaltensempfehlungen, welche sich auf eine Gesamtbeurteilung 

der Gefahrensituation abstützen (Prinzip der „Single Official Voice“). 

– Bei einem Ereignis des Bundesstabs ist der Fachstab Naturgefahren für das Erstellen und 

Nachführen der Fachlage Naturgefahren verantwortlich. 

Der Fachstab Naturgefahren kam beim Hochwasser vom Mai/Juni 2013 erfolgreich zum 

Einsatz. So hatte der Fachstab Naturgefahren des Bundes während der gesamten Zeit des 

Hochwassers den Kommunikationslead. Der Informationsfluss funktionierte reibungslos und 

die Organisationen des Bundes arbeiteten effizient zusammen. Die Informationen gelangten 

schnell und präzise zu den Kantonen, die sie wiederum den Einsatzstäben weiterleiteten. 

Im Vergleich zu den Hochwasserereignissen von 2005 und 2007 wurde damit eine klare 

Verbesserung erzielt. 

Verbesserung der Kommunikation und Informationsprodukte: Damit die erarbeiteten 

Informationsprodukte einen Beitrag zur Bewältigung des Ereignisses leisten können, müssen 

diese den Empfängern bekannt sein. Sie müssen zudem verfügbar und verständlich formuliert 

sein. Es wurde deshalb sowohl für die kantonalen Behörden als auch für die Öffentlichkeit 

je ein hochverfügbarer Informationskanal aufgebaut: 

– Als Kanal für die kantonalen Behörden wurde 2010 die Gemeinsame Informationsplattform 

Naturgefahren (GIN) in Betrieb genommen (Heil et al., 2014). Auf dieser Plattform 

stellen die Fachstellen des Bundes den Naturgefahren-Fachleuten in Bund, Kantonen und 

Gemeinden gemeinsam ihre Produkte zu den verschiedenen Naturgefahren zur Verfügung. 

Diese umfassen Mess- und Beobachtungsdaten, Vorhersagen, Warnungen, Modelle und 

Bulletins. Damit verfügen die Sicherheitsverantwortlichen rasch und in übersichtlicher 

Form über wichtige Informationen. Im Speziellen wurden 2014 erweiterte Informationen 

für Lawinenfachleute integriert. Heute nutzen rund 1‘500 Personen GIN. Während des 

Hochwasser-Ereignisses von Anfang Mai 2015 haben sich insgesamt 430 Naturgefahrenfachleute 

mindestens einmal bei GIN eingeloggt. GIN wurde durchschnittlich von 

160 Naturgefahrenfachleuten pro Tag besucht. 

– Als Kanal für die Öffentlichkeit haben die Naturgefahrenfachstellen des Bundes gemeinsam 

ein Internetportal entwickelt, auf welchem die aktuelle Naturgefahrenlage in der Schweiz 

seit 2014 auf einen Blick erfasst wird (www.naturgefahren.ch). Die Naturgefahrenfachstellen 

des Bundes haben das Portal gemeinsam im Auftrag des Bundesrates entwickelt und 

damit ein weiteres Element zur Verbesserung der Warnung bei Naturgefahren realisiert. 

Es beinhaltet Verhaltensempfehlungen und Hintergrundinformationen zu Naturgefahrenprozessen 

und den dafür zuständigen Bundesfachstellen. Damit wurde ein gemeinsamer 

Verbreitungskanal für Warnungen der Fachstellen des Bundes geschaffen. Die Besucherzahlen 

des Naturgefahrenportals korrelieren jeweils mit der Ereignislage. So verzeichnete 


das Portal im Juli 2014 (intensiver Dauerregen sowie Hochwassergefahr an Flüssen und 

Seen) bis zu 18‘000 Besucher/-innen pro Tag bzw. im Mai 2015 (Starkniederschläge und 

Hochwasser auf der Alpennordseite) bis über 8‘000 Besucher/-innen pro Tag. 

– Für die Information und Warnung der Öffentlichkeit kann neben Medienmitteilungen und 

Naturgefahrenbulletin auch auf das Instrument der verbreitungspflichtigen Warnung 

zurückgegriffen werden. Konkret können bei grosser und sehr grosser Gefahr gemäss Art. 2 

in Verbindung mit 

– Art. 9 der Alarmierungsverordnung (AV) bestimmte Medien zu einer Verbreitung der 

Warnungen an die Bevölkerung verpflichtet werden. 

Ergänzend zu den oben beschriebenen Kanälen bieten die Webseiten der einzelnen Fachstellen 

weiterhin umfassende Informationen. 

Die bisherigen Naturgefahrenereignisse zeigten, dass die Erstellung gemeinsamer Informationsprodukte 

und deren Verbreitung gut funktionieren. Der Zeitplan dazu ist sehr eng und 

die Erstellung bindet viele Personalressourcen. Die Rückmeldungen aus den Kantonen zeigen 

jedoch, dass die von den Fachstellen gemeinsam erstellten Naturgefahrenbulletins als 

qualitativ hochwertig und vertrauenswürdig wahrgenommen werden. 

Im Falle eines Ereignisses ist eine begleitende Kommunikation zentral. So hat sich beim 

Hochwasser Mai/Juni 2013 das schon früh einsetzende mediale Begleiten des Warnprozesses 

sehr gut bewährt. In diesem Fall hat es dazu geführt, dass auf eine verbreitungspflichtige 

Bevölkerungswarnung verzichtet werden konnte. Indem laufend über die neuesten Entwicklungen 

informiert wurde, konnte ein Erwartungsmanagement betrieben werden. Zudem 

wurde mittels Sprachregelungen für Medienanfragen eine zwischen den Bundesfachstellen 

gut abgestimmte Kommunikation erwirkt. 

Ausbildung von lokalen Naturgefahrenberatern: Am Schluss der Warnkette muss 

sichergestellt sein, dass die erhaltenen Informationen lokal richtig interpretiert und angemessene 

Massnahmen ergriffen werden. Die Ereignisanalyse der Hochwasser 2005 zeigte 

deutlich, dass vor allem auf lokaler Ebene bei der Bewältigung von Naturgefahren in 

fachlicher Hinsicht Lücken bestehen. Im Ernstfall müssen sich die Führungsgremien und 

Interventionskräfte auf Fachwissen vor Ort abstützen können, um die Lage umfassend zu 

beurteilen und die richtigen Entscheidungen zu treffen. Es braucht lokale Naturgefahrenberaterinnen 

und -berater (LNGB), die sowohl über das nötige Fachwissen zu Gefahrenprozessen 

als auch über Kenntnisse der örtlichen Besonderheiten verfügen, analog zu den bei 

Lawinendiensten bewährten Strukturen (Bründl et al., 2014). Dies geschieht durch Ausbildung 

von lokalen Naturgefahrenberatern und die Erarbeitung einer Notfallplanung. 

Die lokalen Naturgefahrenberater beraten den Einsatzleiter vor Ort zu möglichen Entwicklungen 

der Situation. Die Notfallplanung bietet ihm bereits vorbereitete Massnahmen, die bei 


Bedarf in kurzer Zeit umgesetzt werden können. Das Prinzip des LNGB wurde bereits von der 

Mehrzahl der Kantone übernommen. Damit wird das lokale Wissen gestärkt. 

Ferner wurde im Zuge der OWARNA-Massnahmen für bevölkerungsschutzrelevante Lagen 

ein Melde- und Lagezentrum an der Nationalen Alarmzentrale (NAZ) beim Bundesamt für 

Bevölkerungsschutz aufgebaut. Das Melde- und Lagezentrum gewährleistet im Ereignisfall 

eine schweizweite Lageübersicht sowie eine sichere und zeitgerechte Verbreitung von 

Warnungen. Gleichzeitig werden die betroffenen Kantone durch das Bereitstellen von 

Informationen in der Beurteilung der Lage und damit in der Ereignisbewältigung unterstützt. 

ZUKÜNFTIGE HERAUSFORDERUNGEN 

Auch in Zukunft ist mit Ereignissen, welche mit den Jahren 2005 und 2007 vergleichbar sind, 

zu rechnen (Beniston, 2007). Gemäss der Zusammenfassung des Berichts CH2011 sind 

„Projektionen der Häufigkeit und Intensität von Niederschlagsereignissen mit grösseren 

Unsicherheiten behaftet, markante Änderungen können jedoch nicht ausgeschlossen werden. 

Zusätzlich wird eine Verschiebung von festem Niederschlag (Schnee) hin zu flüssigem 

Niederschlag (Regen) erwartet, was das Überschwemmungsrisiko speziell in niedrigen Lagen 

vergrössern würde“ (CH2011). Bisher wurden bereits bedeutende Massnahmen zur Reduktion 

der Naturgefahrenschäden verwirklicht, welche sich in den letzten Jahren bewährt haben. 

Die grösste zukünftige Herausforderung liegt in der Aufrechterhaltung und Weiterentwicklung 

dieser Errungenschaften. Gleichzeitig gibt es verschiedene Handlungsfelder, welche 

angegangen werden müssen: 

– Krisensichere Vorhersage und Warnsysteme sowie Krisenkommunikation: 

Die Qualität der Information konnte insgesamt durch verbesserte und neu geschaffene 

Mess- und Prognosesysteme gesteigert werden. Ebenso sind der Aufbau und die Aufrechterhaltung 

des BCM durch die einzelnen Fachstellen gewährleistet. Die Verfügbarkeit der 

für Prognosen, Warnungen und Alarmierungen benötigten Systeme und Kommunikationswege 

müssen jedoch auch bei einem Ausfall der Stromversorgung, der Telefonie und des 

Internets sichergestellt sein. Bei schwerwiegenden Störungen wie einer anhaltenden 

Unterversorgung mit Strom kann dies derzeit auf nationaler Ebene noch nicht vorausgesetzt 

werden. Der Aufbau und Betrieb eines redundanten, stromsicheren Kommunikationsnetzes 

kann nur im Verbund mit den anderen Partnern im Bereich Sicherheit sinnvoll 

umgesetzt werden. Entsprechende Projekte werden derzeit ausgearbeitet. 

– Vereinheitlichung bei den Anbietern von Unwetterwarnungen: Private Warnungsanbieter 

operieren zurzeit vor allem im meteorologischen Bereich, wobei auch Aussagen 

zu wetterbedingten Gefahren wie Hochwasser und Waldbrand gemacht werden. Problematisch 

ist zurzeit, dass private Anbieter mit unterschiedlichen Gefahrenstufen, Schwellenwerten, 

Gefahrenstufenbezeichnungen, Warnregionen und Farbkodierung operieren. Dies 

führt zu Verwirrung bei den Interventionskräften und der Bevölkerung und schadet 

letztlich der Glaubwürdigkeit und dem Ziel der Warnungen insgesamt, Sicherheit zu 

schaffen. Deshalb wurde Ende 2014 ein parlamentarischer Vorstoss (Postulat Vogler, 2014) 


an den Bundesrat überwiesen, der eine Vereinheitlichung bei den Anbietern von Unwetterwarnungen 

fordert. 

– Bereitschaft der Bevölkerung, bei Naturgefahrenereignissen adäquat zu handeln: 

Diverse Beispiele gelungener Interventionen und das grosse Interesse an Notfallplanungen 

zeigen, dass bei den kantonalen und kommunalen Interventionskräften ein Umdenken 

stattgefunden hat. Bei der Bevölkerung eigenverantwortliches Handeln auszulösen ist 

ungleich schwieriger. Das Wissen, in einem Gefahrengebiet zu leben, alleine reicht nicht 

aus. Notwendig ist ein individuelles Verantwortungsbewusstsein, kombiniert mit der 

Kenntnis um sinnvolle Verhaltensweisen. Gerade angesichts der zunehmenden Mobilität 

der Bevölkerung ist die Erhöhung der Bereitschaft der Bevölkerung zum adäquaten 

Umgang mit Warnungen durch eigenverantwortliches Handeln zentral. Um dies zu 

erreichen ist der gezielte Einbezug externer Stellen notwendig: Volksschulen, Bildungsinstitute 

für Architekten, Planer und Ingenieure, Medien, private Warnanbieter und Gebäudeversicherungen 

sind einige von ihnen. Im Rahmen des LAINAT sollen weitere Anstrengungen 

unternommen werden, um das bestehende Ausbildungs- und Bildungsangebot optimal 

einzusetzen und bei Bedarf zielgruppengerecht zu erweitern. Dabei sollen bestehende 

Synergien genutzt und Doppelspurigkeiten vermieden werden. 


Zehn Jahre nach dem Hochwasser 2005 konnten mit dem Projekt OWARNA in den letzten 

Jahren verschiedene Optimierungsmassnahmen, welche zu einem funktionierenden 

Warnprozess und damit zur Reduktion der Naturgefahrenschäden beitragen, umgesetzt 

werden. 

Mit der Schaffung des LAINAT konnte die Zusammenarbeit der Fachstellen des Bundes 

konsolidiert werden. Die Koordination der Vorhersage- und Warntätigkeit der Bundesstellen 

durch diese Organisation hat sich gut etabliert und ist für die Weiterentwicklung und 

Aufrechterhaltung des Bestehenden sowie zur Abdeckung des zukünftigen Handlungsbedarfs 

weiterhin erforderlich. 

LITERATUR 

- Amiguet, C.; Bürgi, T.; Murer, D.; Schmutz, Ch.; Volken, D. (submitted): Warnungen 

der Bundes-Fachstellen der Schweiz vor Unwetter und Hochwasser. Interpraevent 2016. 

BAFU (2014): Zustandsbericht Naturgefahren, 5. Gefahrengrundlagen und Prävention 

(Massnahmen). http://www.bafu.admin.ch/umwelt/status/03995/index.html?lang=de. 

Zugriff: August 2015. 

- Baumgartner, C. (2015): Mobiler Hochwasserschutz in urbanen Gebieten: Ein Überblick 

und Anwendungsmöglichkeiten einzelner mobiler Hochwasserschutzsysteme. disserta Verlag. 

- Beniston, M.; Stephenson, D.; Christensen, O.; et.al (2007): Future extreme events in 

European climate: an exploration of regional climate model projections. Climate change, 

81:71–95. Springer Science+Business Media B.V. 


- Bezzola, Gian Reto; Hegg, Christoph (Ed.) 2007: Ereignisanalyse Hochwasser 2005, Teil 1 

– Prozesse, Schäden und erste Einordnung. Bundesamt für Umwelt BAFU, Eidgenössische - 

Forschungsanstalt WSL. Umwelt-Wissen Nr. 0707. 

- Bezzola, Gian Reto; Hegg, Christoph (Ed.) 2008: Ereignisanalyse Hochwasser 2005, Teil 2 

– Analyse von Prozessen, Massnahmen und Gefahrengrundlagen. Bundesamt für Umwelt 

BAFU, Eidgenössische Forschungsanstalt WSL. Umwelt-Wissen Nr. 0825. 

Bezzola, Gian Reto; Ruf, Wolfgang; Jakob, Adrian (Ed.) 2009: Ereignisanalyse Hochwasser 

August 2007 – Analyse der Meteo- und Abflussvorhersagen; vertiefte Analyse der Hochwasserregulierung 

der Jurarandgewässer. Bundesamt für Umwelt. Umwelt-Wissen Nr. 0927. 

- Bründl, M.; Etter, H. J.; Steiniger, M.; Klingler, C.; Rhyner, J.; Ammann, W. (2004): IFKIS - 

a basis for managing avalanche risk in settlements and on roads in Switzerland, Nat. Hazards 

Earth Syst. Sci., 4(2), 257-262. 

- CH2011: Zusammenfassung Szenarien zur Klimaänderung in der Schweiz CH2011. 

Published by C2SM, MeteoSwiss, ETH Zurich, NCCR Climate and OcCC. http://www.ch2011. 

ch/pdf/CH2011summaryDE.pdf. Zugriff: November 2015. 

- Hess, Josef; Schmid, Franziska (2012): Towards optimised early warning. Developments 

in Switzerland. 12th Kongress Intrepraevent 2012, Grenoble France. 

- Heil, B.; Petzold, I.; Romang, H.; Hess, J. (2014): The common information platform 

for natural hazards in Switzerland, Nat. Hazards, 70(3), 1673-1687. 

- Gaia et al. (submitted): Thunderstorm warnings: design and development of a realtime 

automatic warning system for the authorities and the population. Interpraevent 2016. 

- Postulat Vogler 2014 (14.3694): Notwendige Vereinheitlichungen bei den Anbietern 

von Unwetter-warnungen. 

- SLF (2015): Die Europäische Lawinengefahrenskala. http://www.slf.ch/schneeinfo/ 

zusatzinfos/interpretationshilfe/lawinengefahrenbegriffe/europaeische_skala/index_DE. 

Zugriff: September 2015. 

- SR 520.12, Alarmierungsverordnung (AV) 

- SR 520.17, ABCN-Einsatzverordnung. 


ISBN 978-3-901164-24-8 

WWW.INTERPRAEVENT.AT

INTERPRAEVENT 2016

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?