D3.1 – Technical Evaluation Report of Current Methods of Hazard Mapping of Debris Flows, Rock Avalanches, and Snow Avalanches

SIXTH FRAMEWORK PROGRAMME

Project no. 018412

IRASMOS

Integral Risk Management of Extremely Rapid Mass Movements

Specific Targeted Research Project

Priority VI: Sustainable Development, Global Change and Ecosystems

D3.1 – Technical evaluation report of current methods of hazard mapping of debris flows, rock avalanches, and snow avalanches

Due date of deliverable: 31/05/2008

Actual submission date: 15/07/2008

Start date of project: 01/09/2005 Duration: 33 months

WSL Swiss Federal Institute for Snow and Avalanche Research

Revision [2]

Project co-funded by the European Commission within the Sixth Framework Programme (2002-2006) Dissemination Level PU Public PP Restricted to other programme participants (including the Commission Services) RE Restricted to a group specified by the consortium (including the Commission Services) CO Confidential, only for members of the consortium (including the Commission Services) SIXTH FRAMEWORK PROGRAMME PRIORITY VI Sustainable Development, Global Change and Ecosystems

SPECIFIC TARGETED RESEARCH PROJECT

INTEGRAL RISK MANAGEMENT OF EXTREMELY RAPID MASS MOVEMENTS

WORK PACKAGE 3: HAZARD ASSESSMENT AND MAPPING OF RAPID MASS MOVEMENTS

DELIVERABLE D3.1 Hazard mapping of extremely rapid mass movements in Europe State of the art methods in practice

Edited by: MASSIMILIANO BARBOLINI UNIVERSITY OF PAVIA

OCTOBER 2007 SIXTH FRAMEWORK PROGRAMME

PRIORITY VI - Sustainable Development, Global Change and Ecosystems

SPECIFIC TARGETED RESEARCH PROJECT “IRASMOS” Integral Risk Management of Extremely Rapid Mass Movements (contract n. 018412)

DELIVERABLE D3.1 Hazard mapping of extremely rapid mass movements in Europe State of the art methods in practice

Edited by: MASSIMILIANO BARBOLINI UNIVERSITY OF PAVIA, ITALY, WP3 LEADER

Contributions by: MASSIMILIANO BARBOLINI UNIVERSITY OF PAVIA, ITALY, TASK 3.4 LEADER FEDERICA CAPPABIANCA UNIVERSITY OF PAVIA, ITALY, TASK 3.1 LEADER OLIVER KORUP SWISS FEDERAL INSTITUTES FOR SNOW AND AVALANCHE RESEARCH, SWITZERLAND, TASK 3.3 LEADER DIETER RICKENMANN UNIVERSITY OF NATURAL RESOURCES AND APPLIED LIFE SCIENCES, AUSTRIA, TASK 3.2 LEADER

TABLE OF CONTENTS

LIST OF FIGURES …………………………………………….. pag. 4

LIST OF TABLES ……………………………………………… pag. 6

FOREWORD …………………………………………………… pag. 7

SUMMARY ……………………………………………………. pag. 8

Chapter 1 DEFINITIONS …………………………………………………. pag. 9 1.1 General definitions ……………………………………... pag. 9 1.2 Process-related definition ……………………………… pag. 11 1.2.1 Debris flow …………………………………………………. pag. 11 1.2.2 Rock avalanches ……………………………………………. pag. 13 1.2.3 Snow avalanches ……………………………………………. pag. 16

Chapter 2 DEBRIS FLOW ………………………………………………… pag. 21 2.1 Hazard mapping criteria ………………………………... pag. 21 2.2 Hazard mapping tools ………………………………….. pag. 26 2.2.1 Identification of debris flow hazard ………………………… pag. 26 2.2.2 Evaluation of debris flow occurrence ………………………. pag. 29 2.2.2.1 Triggering conditions and relevant factors ………… pag. 2.2.2.2 Initiation mechanisms ………………………………… pag. 2.2.2.3 Magnitude and frequency …………………………….. pag. 2.2.3 Dynamical behaviour and runout distance prediction of debris flow ………………………………………………….. pag. 35 2.2.3.1 Empirical-sttistical approaches ……………………… pag. 35 2.2.3.2 Analytical-deterministic approaches ……………….. pag. 42 2.2.3.3 Simulation models …………………………………….. pag. 43 2.2.3.4 Physical models ……………………………………….. pag. 48 2.2.3.5 Conclusions …………………………………………….. pag. 48 2.3 Hazard map types ………………………………………. pag. 49 2.3.1 Hazard index maps (overview scale) ……………………….. pag. 49 2.3.2 Hazard maps (detailed scale) ……………………………….. pag. 50 2.3.3 Overview on methods for hazard mapping …………………. pag. 51 2.3.4 Scenarios and uncertaintiy ………………………………….. pag. 52

Chapter 3 ROCK AVALANCHES …………………………………………. pag. 54 3.1 Hazard mapping criteria ……………………………….. pag. 54 3.2 Hazard mapping tools ………………………………….. pag. 56

2 3.2.1 Identification of of rock avalanches hazard ………………… pag. 56 3.2.1.1 Rock avalanche morphology and sedimentology …. pag. 56 3.2.1.2 Causes of rock avalanches …………………………… pag. 59 3.2.2 Evaluation of rock avalanches occurrence …………………. pag. 60 3.2.2.1 Trigger mechanisms and possible precursors ……... pag. 60 3.2.2.2 Engineering, Geological and geotechnical approaches ……………………………………………... pag. 62 3.2.2.3 Magnitude and frequency …………………………….. pag. 63 3.2.2.4 On- and Off-site hazards ……………………………... pag. 65 3.2.3 Dynamical behaviour and runout distance prediction of rock avalanches ………………………………………………….. pag. 70 3.2.3.1 Modelling approaches ………………………………… pag. 70 3.2.3.2 Data input requirements ……………………………… pag. 71 3.2.3.3 Empirical approaches ………………………………… pag. 72 3.2.3.4 Analytical approaches ………………………………… pag. 74 3.2.3.5 Simulation models …………………………………….. pag. 77 3.2.3.6 Discussion on model result …………………………... pag. 79 3.3 Hazard map types ……………………………………… pag. 79

Chapter 4 SNOW AVALANCHES …………………………………………. pag. 81 4.1 Hazard mapping criteria ……………………………….. pag. 81 4.2 Hazard mapping tools ………………………………….. pag. 83 4.2.1 Identification of snow avalanches hazard ………………….. pag. 83 4.2.1.1 Using historical data …………………………………. pag. 83 4.2.1.2 Naturalist approach …………………………………… pag. 91 4.2.2 Evaluation of snow avalanches occurrence ………………… pag. 94 4.2.3 Dynamical behaviour and runout distance prediction of snow avalanches ……………………………………………. Pag. 98 4.2.3.1 Statistical approaches ………………………………. pag. 98 4.2.3.2 Analytical approaches ………………………………… pag. 101 4.2.3.3 Simulation models ……………………………………... pag. 101 4.3 Hazard map types ……………………………………… pag. 108

ANNEXES ……………………………………………………... pag. 114 Annex A – Legislation ……………………………………….. pag. 114

REFERENCES …………………………………………………. pag. 136

LIST OF FIGURES

Figure 1-1: Oblique aerial view of 1987 Val Pola rock avalanche (35 × 106 m3) and dammed lake, near Bormio, Italy pag. 14 (Crosta et al., 2004) Figure 1-2: Mount Munday rock avalanche, Ice Valley Glacier, Coast Mountains, Canada. Length of rock avalanche is pag. 15 4.7 km (Evans et al., 2002) Figure 1-3: Schematic representation of the components of an avalanche path (Image from "Advanced Avalanche pag. 17 Safety Course Manual" Copyright © 1998 Canadian Avalanche Association) Figure 1-4: Main morphological criteria of avalanche classification pag. 18 Figure 2-1: Main elements which need to be considered in assessing the hazard of debris flows. Elements mainly pag. 22 relevant for debris-flow frequency and magnitude have a green background colour, and those mainly relevant to determine flow intensity and affected areas have a blue background colour Figure 2-2: Hazard rating system used in Switzerland: red = high, blue = medium, yellow = low, white = no hazard. pag. 23 The hazard degree (three colors) is a function of the intensity and probability of an event. Figure 2-3: Relation between critical rainfall intensity and duration for debris-flow occurrence: data from the Swiss pag. 29 Alps (from Zimmermann et al., 1997a), and comparison with other threshold lines. Figure 2-4: Three typical debris-flow initiation mechanism (from VAW, 1992): bed erosion and fluidisation, slope pag. 31 instability, temporary blocking of water and solid discharge Figure 2-5: Data on event magnitude of debris flows. A few data points from Italian events refer to torrential floods pag. 33 with sediment transport. Figure 2-6: Relationships between debris-flow peak discharge and event magnitude (from Rickenmann, 1999). pag. 36 Figure 2-7: Application of Manning equation (2) to debris flows, using a pseudo Manning n value back-calculated for pag. 37 each sub-dataset (from Rickenmann 1999, Rickenmann & Weber 2000). Figure 2-8: Application of empirical velocity equation (3) to debris flows (of data sets in Table 222-4); equation was pag. 38 first developed for water flows (from Rickenmann 1999). Figure 2-9: Travel angle vs. volume of mass movement. Data include 140 debris flows from the Swiss Alps and 51 pag. 39 large landslides/rock avalanches (for data sources see Rickenmann 1999). Also shown is the approximate range for 101 data points of unobstructed landslides plus unobstructed and channelized debris flows discussed in Corominas (1996). Figure 2-10: Observed runout distance of debris flows and rock avalanches compared to predictions with equation (7), pag. 40 which was derived for debris flows. Data include 140 debris flows from the Swiss Alps and 51 large landslides/ rock avalanches (for data sources see Rickenmann, 1999). Figure 2-11: Travel angle of debris flows vs. watershed area. Data include 144 debris flows from the Swiss Alps, where pag. 41 the deposits are characterized as “cs” = clast supported, “sg” = sand and gravel, and “ms” = matrix supported (modified from Zimmermann et al., 1997). Figure 2-12: Typology of the examined flows: (a) loose bed, immature flow (sheet-flow); (b) loose bed, mature flow; (c) pag. 45 loose bed, plug flow; (d) solid bed flow. Figure 2-13: (a) Calculated deposition patterns for Val Varuna debris-flow cone with simulation model of Zimmermann pag. 50 et al. (1997a) [source: Gamma (2000)]; (b) Observed deposition of debris flow in 1987 on the fan of Val Varuna [photo Swiss Federal Office of Torpography of 21.9.1987, archive KSL, flight line 080003, image no. 9871] Figure 2-14: Example of simulation with model of Zimmermann et al. (1997a) used for hazard index mapping; debris- pag. 50 flow paths are marked in red [source: M. Zimmermann, Geo7] Figure 2-15: Example of hazard map at detailed scale (scale of the single path). Case study: Sauris (IT) debris flow. pag. 51 Dynamical model: TRENT-2D. Mapping criteria: (a) Rickenmann (2005b) criteria; (b) PAT (Trento Province) criteria. Figure 3-1: Aerial view of the rainfall-triggered 2006 Guinsaugon rockslide-debris avalanche, Philippines, that killed pag. 55 more than 1000 people (Evans et al., 2006). Figure 3-2: Hummocky surface of volcanic debris-avalanche deposit along north side of Chakachatna River, Alaska. pag. 57 View is toward south. Photograph by C.F. Waythomas, July 1998 (Waythomas and Nye, 2002). Figure 3-3: Oblique air photo draped on DEM of the Cheam 5000-year old rock avalanche (175 × 106 m3), Fraser pag. 58 Valley, British Columbia, Canada. Conspicuous hummocky terrain together with a number of geological boreholes led to the identification of this major event (Orwin et al., 2004). Figure 3-4: Boulder carapace of rock avalanche that ran out on West Fork Glacier, Alaska, during the M 7.9 Denali pag. 59 th k f 3 N b 2002 (Jib t l 2006) N t f l

4 earthquake of 3 November 2002 (Jibson et al., 2006). Note person for scale. Figure 3-5: High-resolution digital photogrammetry of successive air photos allows computation of displacement pag. 63 vectors on rock slope adjacent to Aletsch Glacier, Swiss Alps. Detailed knowledge of such processes as potential precursors of catastrophic rock-slope failure are essential for site-specific hazard and risk management (Kääb, 2002). Figure 3-6: Suggested steps in compiling a regional-scale inventory on rock avalanches (Korup, 2005b). pag. 64 Figure 3-7: Cumulative volume distribution for a worldwide dataset of 142 rockfalls derived from cliffs. Note power- pag. 65 law trend of rockfalls with volumes between 107 and 1010 m3 (Dussauge et al., 2003). Note that events in this size domain also contain rock avalanches. Figure 3-8: Changes to river-channel long profile following catastrophic sediment input from 1999 Mt Adams rock pag. 66 avalanche, Southern Alps, New Zealand (Korup et al., 2004). Figure 3-9. Downstream effects of confined river morphology following failure of the 1999 Mt Adams rock-avalanche pag. 67 dam, Southern Alps, New Zealand. A: False colour satellite image showing rock avalanche (ra) and landslide dam (ld). B: Channel changes measured between locations 1 and 2 in (A). (Korup et al., 2006) Figure 3-10: Example of the persisting problems associated with large rock avalanches, State Highway 93, Southern pag. 68 Alps, New Zealand. This important traffic route was for years routed across heavily fragmented rock- avalanche debris that was emplaced ~2000 years ago. Continued rock fall and debris flows from the deposit necessitated the construction of the Southern Alps’ first viaduct. Note that pylons are grounded in >30-m thick rock-avalanche debris (Korup, 2004a). Figure 3-11: Possible chains of events relating to the formation and failure of a rock-avalanche dam (Korup, 2005a). pag. 69 Figure 3-12: Sketch highlighting several geomorphic hazards expressed as dependent probabilities resulting from the pag. 69 formation and failure of a rock-avalanche dam in a mountain-foreland system. P(D|O), for instance, denotes the conditional probability of spatial impact by a dam-break flood, given the failure of the dam. Figure 3-13: The relationship between the ratio of rock-avalanche height over length (H/L) and volume is an often cited pag. 70 measure of the mobility of such extremely rapid mass movements (Collins and Melosh, 2003). Figure 3-14: Types of landslides responsible for damming rivers from a sample of 353 worldwide examples (Ermini and pag. 73 Casagli, 2003). Figure 3-15: Relationship between total area A affected by rock avalanches and the potential energy of the mass prior pag. 75 to failure (Dade and Huppert, 1998). Figure 3-16: Time steps in a numerical model of the 1987 Val Pola rock avalanche, Italy. This model is based on a pag. 78 Voellmy rheology also used for simulating the runout of snow avalanches (Crosta et al., 2004). Figure 4-1: Avalanche hazard zones defined according to France criteria: in violet high hazard (A3), in orange pag. 82 moderate hazard (A2) and in rose low hazard (A1). France mapping criteria also include the largest expected avalanche (yellow, AMV) and the protective forests (green, V) Figure 4-2: Example of site photograph. It reports from top to bottom, the name of the village, the zip code and the site pag. 85 number, as well as some red lines for the avalanche main flows Figure 4-3: Example of the observation map for the avalanche permanent survey. It reports for site # 202 from uphill pag. 86 to downhill: the reference number of the site, the edge of the avalanche site open to downhill, the avalanche main flows, the observation and alert thresholds, and the observation point. Moreover, the grey areas report witness testimonies from the past avalanche map. Figure 4-4: Cartography obtained from the testimony by father Jean-Baptiste Dédevens –1720 pag. 91 Figure 4-5: Disc of a tree impacted by a snow avalanche. pag. 93 Figure 4-6: Operator scrutinazing aerial pictures through a stereoscope pag. 94

Figure 4-7: Preliminary analysis for 1-D simulation of avalanche dynamics. Case: Vallecetta avalanche, Bormio (see pag. 95 Figure 4 below). a) Aerial view (Picture regione Lombardia, adapted in Riboni et al., 2005). b ) Definition of the release zone (green area) and of the main avalanche profile (green dots)

Figure 4-8: Conceptual approach for avalanche hazard mapping a) Snowfall input. Adapted from Ancey et al., 2004. pag. 95 b) Scheme of the avalanche dynamics (by courtesy of SLF Davos)

Figure 4-9: Avalanche hazard map, according to the Swiss approach (yellow, blue and red zone pag. 96 Figure 4-10: Overview of dymnamical models pag. 101

Figure 4-11: Example of the past avalanches map. It reports witness testimonies in dark gray and photo-interpretation pag. 110 in lighter gray (respectively magenta and orange in original colored map)

Figure 4-12: Example of CLPV. The maximum avalanche expansion on each site is drawn, in violet, from the analysis pag. 110 of historical documentation (§ 4.2.1.1) and, in orange, from photo interpretation (§ 4.2.1.2).

Figure 4-13: Scheme of the different phase in the creation of an hazard map. pag. 111 Figure A1-1: Derivation of hazard zones from probability and intensity of the hazardous process (BUWAL, BRP & pag. 118 BWW 1997)

5 BWW, 1997) Figure A2-1: Example of a susceptibility/hazard map for volcanic debris avalanches, Iliamna Volcano, Alaska. pag. 120 Figure A2-2: Extract from the West Coast Emergency Plan issued by the West Coast Regional Council, i.e. the major pag. 121 administrative authority in charge for natural hazards mitigation in the western Southern Alps, New Zealand. Note the explicit treatment of the term “rock avalanche”, which is a frequent process in this tectonically active mountain belt. Figure A3-1: Classification of the hazard zones for snow avalanches according to the Swiss guidelines (BFF and SLF, pag. 132 1984).

LIST OF TABLES

Table 1-1: Expressions for debris flows and related processes in some European languages pag. 12 Table 1-2: Expressions for rock avalanches and related processes in some European languages pag. 16 Table 1-3: Expressions for snow avalanches and related processes in some European languages pag. 19 Table 1-4: The Canadian avalanche size classification (McClung and Schaerer, 1993) pag. 20 Table 2-1: Criteria to delimit intensity classes for debris flows and related processes (Switzerland). pag. 23 Table 2-2: Definition of hazard zones or degrees of hazard, as applied in Switzerland (Petrascheck & Kienholz, pag. 24 2003). Table 2-3: Empirical relationships to estimate the event magnitude of debris flows and/or torrential floods with pag. 32 sediment transport. Table 2-4: Tentative estimated channel debris yield rates in western Canada (Hungr et al., 1984). pag. 33 Table 2-5: Empirical formulae relating debris-flow peak discharge and event magnitude (from Rickenmann, 1999; pag. 36 Jakob, 2005). Table 2-6: Data on mean flow velocity of debris flows (from Rickenmann, 1999). pag. 37 Table 2-7: Methods used for the assessment of debris flow parameters (Rickenmann, 2001, 2001b, 2003). pag. 52 Table 3-1: Key characteristics of hot and cold avalanche deposits in volcanic terrain (Stewart et al., 2003). pag. 59 Table 4-1: Reference values fo avalanche intensity and frequency adopted in Italy and Switzerland for avalanche pag. 82 hazard mapping. High hazard zone are coloured in red; blue corresponds to moderate hazard and yellow to low hazard. Table 4-2: Swiss reference values of P*72 according to different return periods (Salm, 1990) pag. 96 Table 4-3: Topographic parameters governing maximum runout distance pag. 99 Table 4-4: Regression models proposed by for different mountainous regions. pag. 99 Table A1-1: Specifications for the design event (150 years)and a frequent event (10 years)with respect to torrent pag. 115 related processes Table A1-2: Definition of Swiss hazard zones (BUWAL, BRP & BWW, 1997) pag. 118 Table A1-3: Classification of intensity for damaging processes (BUWAL, BRP & BWW, 1997) pag. 118 Table A1-4: Criteria for the intensity of a process for settlement areas (BUWAL, BRP & BWW, 1997) pag. 119 Table A1-5: Definition of the probability classes (BUWAL, BRP & BWW, 1997) pag. 119 Table A3-1: Specifications for the design event with respect to snow avalanches pag. 123 Table A3-2: Specifications for the design event with respect to snow avalanches (Barbolini et al., 2005). T is the return pag. 125 period(years) and P the avalanche impact pressure (kPa) Table A3-3: Safety requirements for the location of buildings pag. 128

EDITOR'S FOREWORD

The present report is a truly European product in that researchers in the field of natural hazard from polar latitudes (Norway) to Mediterranean coast (Italy and France) have contributed to it, sharing knowledges from their own country with colleagues from other countries. Nevertheless, it combines knowledges coming from process-based researches and experiences, contributing to a valuable “integrated” wiev to natural hazard prevention in mountain communities.

The spirit of collaboration on this report as well as on many other subjects within the IRASMOS project definitely proves that frontiers between countries have a weak meaning in natural hazard science, for the benefit not only of the research community, but also of the citizens in all countries threatened by natural hazards.

I wish to sincerely thank all my co-authors of this report, listed on the frontispiece, for their willingness to ease my editorial work, for their speedy responses to my many wishes, for additional information and pictures and for proof-reading.

Other persons from the IRASMOS consortium that have made significant contributions to the report that I wish to gratefully acknowledge at this point are the following: Sven Fuchs and Markus Holub (BOKU), Florance Naaim, Domenique Laigle and Nicolas Eckert (CEMAGREF), Luigi Fraccarollo (CUDAM), Daniele Bocchiola (MIL), Erik Hestnes (NGI), Michael Bruendl (SLF).

A special thanks goes to Gloria Furdada and Ema Suriñach, from University of Barcelona, for checking my “naïf” Spanish translations along the text.

Pavia, October 30, 2007

Massimiliano Barbolini

SUMMARY

For easy reference, in Chapter 1 the terminology of main concern for the report is defined, including most commonly used “process-specific” expressions.

Chapter 2, 3 and 4 pose the attention on different extremely rapid mass movements, that is debris flow, rock avalanches and snow avalanches, respectively. Hazard mapping criteria, hazard mapping tools and types of maps are described in detail for each of the considered process.

In Chapter 5 some tentative conclusions arising from a critical analysis of the state-of- the-art are proposed.

Annex A closes the report, with a summary of the current legislations at European scale for mapping the areas endangered by the three considered natural processes.

Chapter 1

DEFINITIONS

1.1 General definitions

Introduction Natural hazard research began as a study of cultural perceptions to extreme naturally occurring events. Since the advent of the study, natural hazard literature has become steeped in terminology that is sometimes confusing to the reader. The concepts of hazard and risk are often confused with one another. Because these terms are casually used in everyday vernacular, their meaning lack precision; not to mention that translated into different languages the terms take on an even broader range of definition. Authors of various hazard-related textbooks attempt to define the terminology, but similar to the human fingerprint each author's definitions are unique: one author may provide a definition for risk, where another author uses the same difinition for hazard. Furthermore, the concepts of hazard and risk are a function of individual and/or cultural perceptions and what a particular individual and/or culture values. Because extreme naturally occurring events are ubiquitous, the terminology is also variable on differing perceptions of the event.

Maybe it is too arduous to apply universal meanings to the terminology and concepts of natural hazard studies, and it is probably up to the individual to assing his/her own definitions based on his/her own perceptions. However, in order to avoid ambiguities in the frame of the present report, below we illustate documented definitions of the main terms used in natural hazard studies to which the reader should refer.

Hazard Most authors agree that the definition of a hazard has to include the potential for an harmful interaction between humans and an event:

10 • a condition or situation which has the potential to create harm to people, property, or the environment (IPENZ, 1983);

• a naturally-occurring or human-induced process or event with the potential to create loss (Smith, 1996).

Some authors also include in the definition of hazard a sense of cultural perception, or explicitly point out the variability of the hazard concept based on cultural perception of what peolple value:

• an interaction between a system of human resource managment and an extreme or rare natural phenomenon, which may be goephysical, atmospheric or biological in origin, greatly exceeding normal human expectations in terms of its magnitude or frequency, and causing a major human hardship with significant material damage to infrastructure and/or loss of life or disease (Chapman, 1994);

• An interaction between humans and extreme natural events with respect to cultural perceptions and value systems (Gravley, 2001).

Risk Either directly or indirectly the definition for risk are unanimously a combination of probability and loss:

• the probability that a hazard will be realised and the probability of the harm itself (IPENZ, 1983);

• the actual exposure of something of human value to a hazard, often regarded as the combination of probability and loss (Smith, 1996),

even if some authors push the concept of risk further by taking into account the vulnerability of a particular population:

• a function of the probability of the specified natural hazard event and vulnerability of cultural entities (Chapman, 1994);

• an assessment of what a particular group of people value and the probability of losing what they value as the result of a hazard (Gravley, 2001).

Quantitative definitions The United Nations Disaster Relief Organization (UNDRO, 1982) provides a of risk quantitative definition for risk (R) by multiplying the probability of hazard occurrence (H), the vulnerability (V) and the social elements at risk (E):

R = E × H × V.

Similarly, according to the Committee on Risk Assessment of the Working Gorup on Landslides of the International Union of Geological Sciences (IUGS, 1997), quantitative risk assessment involves expressing the risk as a function of hazard, element at risk, and vulnerability, where these terms are defined as follows:

11 Risk: A measure of the probability and severity of an adverse effect to health, property or the environment form ERMM.

Hazard: The probability a particular ERMM(*) occurs within a given time.

Element at risk: The population, building and engineering works, economic activities, public service utilities and infrastructures in the area potentially affected by ERMM.

Vulnerability: The degree of loss to a given element or set of elements within the area affected by ERMM(s). It is expressed on a scale of 0 (no loss) to 1 (total loss). For property, the loss will be the value of the property, and for persons it will be the probability that a particular life (the element at risk) will be lost, given the person(s) affected by the ERMM.

Note: In the previous definitions the world landslide has been substituted by ERMM, acronym for: Extremely Rapid Mass Movements (i.e. Debris Flows, Roch Avalanches, Snow Avalanches).

1.2 Process-related definitions

1.2.1 Debris flow

Defining debris flow A debris flow is a mixture of water, poorly sorted sediment and other debris, typically flowing rapidly, with one or more surges and a coarse-grained front, down steep mountain channels to a fan. Both solid and fluid forces strongly influence the motion, distinguishing debris flows from related phenomena such as rock avalanches and sediment-laden floods. A debris flow has a higher solid concentration than normal or hyperconcentrated streamflow on the one end, and a higher water content than a landslide or rock avalanche on the other end of the spectrum. The material involved in debris flows usually includes sediment from the clay size up to boulders, and organic components such as woody debris may also be present. Debris flows in the more restrictive sense have a significant amount of coarse particles particularly near the front region.

Debris and mud flows A mud flow is a mixture of water and predominantly clay-, silt- and sand-sized particles, making up a viscous slurry and typically flowing rapidly, with one or more surges and possibly including organic debris, down steep and also more gentle mountain channels. The term mud flow usually refers to fine-grained debris flows. The presence of fine cohesive particles such as clay and a high solid concentration imply that the mixture is viscous and the flow behaviour is often laminar. The viscous slurry prevents rapid drainage of water. Mud flows are generally associated with longer runout distances than coarse-grained debris flows.

12 However, debris and mud flows show similar general characteristics. The bulk behaviour of both debris and mud flows is characterised as a non-Newtonian fluid. At low shear stresses or shear rates the mixture may stop flowing at non-zero slope gradients (e.g. due to a yield stress). Debris flows including a substantial amount of coarse particles are probably more frequent in torrential catchments of the European Alps than mud flows. However, depending probably upon the local geology, and lithology, large areas are recognized to produce essentially mud flows or so-called muddy or viscous debris flows.

Hillslope debris flows Hillslope debris flows have a similar material composition as debris or mud flows. In contrast to debris or mud flows, hillslope debris flows do not occur in an established or predefined channel. A hillsope debris flow usually does not occur repeatedly within a historic time scale at a given location. In its initial stage, it may be similar to a landslide (debris slide), but after internal distortion and development of higher velocities it has a flow character similar to debris flows. Multiple surges are uncommon. There is no unique terminology to distinguish between different flow types which reflects to some degree the complexity of these phenomena. For some terms, even contradictory definitions have been proposed. Here, we adopt the most common use in the scientific English literature. Debris flow is often used as general expression referring to all the three flow types discussed above. In the following, we also use it in this general sense and only refer to mud flows and hillsope debris flows where an explicite distinction is necessary.

In Table (1-1) an overview is given on the most commonly used terms for the different flow types in several languages.

Table 1-1: English Englisch German French Italian Spanish terminology alternate Expressions for debris flows adopted here expressions and related processes in some European languages debris flow debris torrent, (granularer) lave torrentielle colata di detrito, corriente de debris avalache Murgang (granulaire) lava torrentizia derrubios,

mud flow mudflow Schlammstrom, coulée de boue, colata di fango Colada de barro feinkörniger lave torrentielle Murgang boueuse

hillslope debris debris Hangmure coulée boueuse scivolamento di Colada de flow avalanche, detrito derrubios mudslide

Hyper- debris flood, Intensiver charriage hyper- - Flujo concentrated mud flood, Geschiebe- concentré * hiperconcentrad flow immature debris transport * o flow

normal - Hochwasser crue avec piena con Flujo con carga streamflow, mit Geschiebe- charriage trasporto solido de fondo flood including transport fluvial sediment transport

* these expressions refer to the sediment transport conditions (immature debris flow) which occur in hyperconcentrated flow or in normal streamflow at steep slopes

Earth flows Flowing sediment-water mixtures with solid concentrations higher than debris or mud flows are generally called landslides or earth flows but there is a numberofother

13 flows are generally called landslides or earth flows, but there is a number of other expressions among which also debris avalanche is used (thus being in contradiction to the use of this term for hillslope debris flows).

Hyperconcentrated Hyperconcentrated flows have sediment concentrations higher than for fluvial sediment flows transport but lower than for debris and mud flows. Hyperconcentrated flows are not associated with a surging flow behaviour (as debris and mud flows), and the fluid mixture is characterised as a non-Newtonian fluid with a yield stress (contrary to normal streamflow), implying low settling velocities of sand-sized paricles. Hyperconcentrated flows occur through dilution of debris or mud flows, or by further entrainment of solid material through sediment transporting flows.

The volumetric solid concentration in the frontal and main part of a debris-flow surge is of the order of 50 % to 85 %. In the case of channelised debris and mud flows there is a close relationship to hyperconcentrated and normal streamflow, which are characterised by decreasing sediment concentrations. In hyperconcentrated flows, the volumetric solid concentration is of the order of 25 % to 50 %. The exact limits between different flow types are not generally known, and they depend also on grain size distribution and on grain composition.

The transition between different process types may occur towards the rear part of a debris-flow surge or through dilution by normal streamflow when entering higher order channels. A transition between debris flows, hyperconcentrated and normal streamflow is also possible in a given torrent channel during the same rainstorm event. With normal streamflow conditions prevailing over a longer time period, fluvial sediment transport may also result in a considerable throughput of solid material to the fan area.

For a more detailed discussion of the characteristics and proposed classifications of solid-water flows, see for example Hungr et al. (2001).

1.2.2 Rock avalanches

Defining rock Rock avalanching is the extremely rapid (>30 m/s), flow-like movement of large avalanches volumes (>106 m3) of initially intact rock masses that are increasingly fragmented during the runout process. A key characteristic of rock avalanches is their unusual long (or “excess”) runout (Heim, 1932; Hsü, 1975; Nicoletti and Sorriso-Valvo, 1991; Legros, 2002), which cannot be explained by conventional friction physics; a number of mechanistic theories have attempted to explain the excess travel distance. The phenomenon of this extreme mobility is limited to larger failure volumes (which are alternatively termed “sturzstrom” after Heim, 1932), although discrete values are difficult to establish. The German term “Bergsturz” addresses slope failures with a minimum volume of 106 m3. Nearly two third of the largest catastrophic landslides on Earth that have occurred in a terrestrial setting can be classified as rock or debris avalanches (Korup et al., in prep.).

14 Figure 1-1:

Oblique aerial view of 1987 Val Pola rock avalanche (35 × 106 m3) and dammed lake, near Bormio, Italy (Crosta et al., 2004).

Rock vs. debris From a physical perspective, rock avalanches are extremely rapid granular mass flows, avalanches often involving solid, liquid, and gaseous phases. Rock avalanches sensu stricto are dry phenomena that travel without significant aid of water. The term “debris avalanche” is generally used in the context of catastrophic failures on the flanks of volcanic edifices, although it is in some countries (e.g. Canada) also used to describe rapid debris-flow like phenomena (see Table 1-1 for terminology on debris flows). In few cases, failures of tectonically weakened rocks with material properties similar to those of debris are also covered by this term (Wright, 1999).

Both morphology and dynamic characteristics of volcanic debris avalanches strongly resemble those of rock avalanches. The main difference is that volcanic debris avalanches usually have higher water and clay content, which partly leads to substantially greater travel distances and affected areas, while their movement does not imply catastrophic comminution of rock particles down to sand and silt size. A special feature of volcanic debris avalanches are toreva blocks, which are exceptionally large fragments of deposits that remained largely intact during catastrophic failure (Ponomareva et al., 2006). Evidence of comparably sized blocks in rock-avalanche deposits has so far not been documented. Quite conversely, the characteristic

15 fragmentation of initially intact rock masses during flow motion is what also distinguishes rock avalanches from volcanic debris avalanches and other large and coherent rock(-block) slides alike (Davies and McSaveney, 2002; Davies et al., 2006). The resulting deposits are commonly referred to as chaotic brecciae or megabrecciae (Schramm et al., 1998), mainly consisting of poorly sorted and very angular debris.

Figure 1-2:

Mount Munday rock avalanche, Ice Valley Glacier, Coast Mountains, Canada. Length of rock avalanche is 4.7 km (Evans et al., 2002).

Composite rock Like many landsliding phenomena, rock avalanches may be part of multi-phase mass avalanches movements. For instance, they may evolve from initial sliding or falling of coherent rock masses, and conversely transform into debris flows, or in the case of volcanic debris avalanches, lahars. In many cases, entrainment of valley-fill sediments or glacier ice, or catastrophic displacement of water bodies, may often modulate or even dominate their runout dynamics (Abele, 1989; Hungr and Evans, 2004; Huggel et al., 2005). Consequently, the dominant types of motion are combined for use in technical jargon, i.e. rockslide/rock avalanche or debris avalanche/lahar (Cruden and Varnes, 1996).

16 Table 1-2: English Englisch German French Italian Spanish terminology alternate Expressions for rock adopted here expressions avalanches and related processes in some European rock avalanche sturzstrom Bergsturz avalanche de marocche avalancha de languages pierres rocas

volcanic debris (hot/cold) Vulkanischer Lahar avalanche volcanic Bergsturz landslide

toreva blocks Toreva Block

rock/ice Fels-/Eislawine debâcle Avalanche de avalanche rocas y hielo

Rock- Bergsturz/Mur debâcle avalanche/debri gang s flow

1.2.3 Snow avalanches

Defining snow Avalanches may be a harmless trickle of loose snow descending to a new angle of avalanches repose, or a large devasting mass of snow, ice and earth, moving at high speed down a lengthy slope with enough energy to destroy anything in its path. They may occur under widely different conditions and take various forms.

Avalanche path An avalanche paths can generally be divided in three zones, see Figure 1-3:

• starting zone (or zone of origin);

• track (or zone of transition);

• runout zone (or zone of deposition).

The starting zone of an avalanche is where the unstable sheet of snow breaks loose from the rest of the snowpack and starts to slide down the slope. Some avalanche paths have multiple starting zones which led into a common track and runout zone. The avalanche track is the part of path below the starting zone that connects the starting zone with runout zone; it can be represented by a gully or by an open-slope. Runout zone or zone of deposition is the area where deceleration of avalanche occur, the snow masses are deposited, and the avalanche come to a rest.

17 Figure 1-3:

Morphological A comprehensive classification of snow avalanches, including related teminology, is classification given in the International Avalanche Classification published by the Unesco (1981). This classification is essentially based on the following morphological criteria (see Figure 1-4):

• the manner of starting (slab vs loose avalanche)

• the type of motion (flowing vs airborne avalanche)

• the position of the bed surface (surface vs ground avalanche)

• the form of the path (channelled vs open-slope avalanche)

• the free water content of snow (dry vs wet avalanche)

Depending on the type of rupture which originates the snow avalanche, avalanches are classified as loose snow or slab. Loose avalanches initiate on or near the snow surface at a particular point where internal cohesion is surpassed by slope angle. Loose avalanches spread in a triangular pattern as they move down slope. Slab avalanches occur when an entire cohesive block of snow releases due to a subsurface failure (typically occurring in the so called "weak layer"). Both types of avalanche can occur in wet and dry snow conditions. Dry slab avalanches are generally the most significant in terms of hazard mapping and land use planning. However wet loose avalanches, that flow more slowly and close to the ground, may have significant destructive impact on vegetation.

Two distinct limiting forms of motions exist: (a) motion close to the ground and following the ground contours of a snow mass of large bulk density (of the order of 10-

18 1 g/cm3), referred to as flowing (or dense-snow) avalanche; (b) motion of snow cloud of low bulk density (of the order of 10-2 g/cm3), consisting of snow particles suspended in air, called airborne (or powder-snow) avalanche. Airborne avalanches requires dry snow and their formation is favoured by a sudden increase of slope angle of the avalanche track or when the avalanche jumps over a precipice. Flowing avalanche can occur under wet or dry conditions.

Some wet snow avalanches extend down to the ground, whereas others, and generally dry-snow gravity flow, have the lower boundary within the snow cover. This introduces the further distinction between ground and surface avalanche. When the avalanche track is represented by a gully or a ravine we refer to a channeled avalanche, in opposition to an open-slope avalanche.

Figure 1-4: (A) Manner of starting

Main morphological criteria of avalanche classification

SLAB AVALANCHE LOOSE-SNOW AVALANCHE

(B) Type of motion

FLOWING AVALANCHE AIRBORNE AVALANCHE

SURFACE AVALANCHE GROUND AVALANCHE

(D) Form of the path

CHANNELLED AVALANCHE OPEN-SLOPE AVALANCHE

19 In Table 1-3 an overview is given on the most commonly used terms for the different avalanche types in several languages.

Table 1-3: English French German Italian Spanish

Expressions for snow Slab avalanche Avalanche de Schneebrettlawine Valanga a lastroni Alud de placa avalanches in some plaque de neige European languages Loose avalanche Avalanche de neige Lockerschneelawine Valanga a debole Alud de nieve meuble coesione suelta

Flowing avalanche Avalanche de neige Fliesslawine Valanga radente Alud de nieve (or dense-snow coulante (o valanga di neve densa avalanche) densa)

Airborne avalanche Avalanche de neige Staublawine Valanga Alud de nieve (or powder-snow poudreuse nubiforme (o polvo avalanche) valanga di neve polverosa) Alud de aerosol

Ground avalanche Avalanche de fond Grundlawine Valanga di fondo Alud de fondo

Surface avalanche Avalanche de Oberlawine Valanga di Alud de superficie surface superficie

Dry avalanche Avalanche de neige Trockenschneelawine Valanga di neve Alud de nieve seca sèche asciutta

Wet avalanche Avalanche de neige Nassschneelawine Valanga di neve Alud de nieve mouillée bagnata húmeda

Channelled Avalanche de Runsenlawine Valanga in Alud canalizado avalanche couloir canalata

Open-slope Avalanche de Flächenlawine Valanga di Alud sin confinar avalanche versant versante

Size classification Snow avalanche are classified according to size as well. Canadian system, which is widely used also in Europe, classify avalanche size according to its destructive potential (see Table 1-4), using an approach similar to the Mercalli scale used for ranking the shaking intensity of earthquakes. Different sizing systems are employed elsewere in the world. The reader could refer to Appendix D of McClung and Schaerer (1993) for a detailed discussion on the other avalanche size classification system proposed so far.

20 Table 1-4: Size Description Typical Mass [×103 Typical Path Typical Impact kg] Length [m] Pressures [kPa] The Canadian avalanche size classification (McClung and 1 Relatively <10 10 1 Schaerer, 1993) harmless to people

2 Could bury, injure 100 100 10 or kill a person

3 Could bury a car, 1'000 1'000 100 destroy a small building, or break a few trees

4 Could destroy a 10'000 2'000 500 railway car, large truck, several buildings, or a forest with an area up to 4 hectares

5 Largest snow 100'000 3'000 1'000 avalanches known; could destroy a village or a forest of 40 hectares

Chapter 2

DEBRIS FLOWS

2.1 Hazard Mapping Criteria

Introduction A recent summary of hazard assessment of debris flows is given by Huerlimann et al. (2006). First, the probability of debris-flow occurrence is evaluated, and second the runout behaviour is analysed. Generally, the results of these two steps should be summarized in a zonation map representing the hazard degrees, while the assessment typically ends with a conclusion on hazard mitigation and reduction.

The evaluation of the temporal and spatial probability of debris flows is generally a difficult and time-consuming task, which includes field surveys, interpretation of aerial photos, enquiries, studies of archives, etc. Temporal probability can be obtained from historic data, and the final result should be a magnitude–frequency relationship. After determining the potential debris flow initiation zones, the runout behaviour is analysed in order to delimit the extension of the endangered zones. The results of the two previously mentioned steps are summarized in a hazard map. A hazard map is divided into different hazard degrees, which may be defined based on a hazard matrix. The fundamental components of such a matrix should be the intensity and the probability of occurrence of the process (Hungr, 1997; Raetzo et al., 2002; Petrascheck and Kienholz, 2003). Sometimes magnitude is taken as an index for intensity. However, the term “intensity” is preferable as it better symbolises the destructiveness of the process. In the Swiss procedure, for example, the intensity is defined by the impact pressure of a debris flow, which is given by velocity and flow depth. The probability of occurrence can also be expressed by the frequency or the recurrence period of the process.

Instead of simply using the probability of occurrence, in France it is preferred to use now the notion of “probability of a given point to be reached by the flow”. This is the combination of the probability of occurrence (volume for example) with some

22 probability of evolution of the flow on the alluvial fan. This latter greatly depends upon the local topography but also upon the occurrence of disturbing phenomena (local overflow, clogging by boulders, woody debris…) that are likely to modify the flow. This aspect is further discussed in paragraph 2.34 "Scenarios and uncertainty".

The final step of each hazard assessment refers to the mitigation and the reduction of the existing hazard. Different approaches have been proposed to reduce loss from landslides, including restriction of development or protection against hazard (Huebl and Fiebiger, 2005).

In assessing the potential hazard of debris flows, basically two aspects have to be analysed: (a) debris-flow occurrence, and (b) characteristics of the flow process. Figure (2-1) shows the main elements to be considered and their interrelationship (Rickenmann, 2001). An essential element of the hazard mapping are the criteria to distinguish between different classes of intensities and probabilities. Basically, a number of different approaches may be used to determine the process intensity in endangered areas, depending on the required level of detail for the hazard map.

Figure 2-1:

Main elements which need to be considered in assessing the hazard of debris flows. Elements mainly relevant for debris-flow frequency and magnitude have a green background colour, and those mainly relevant to determine flow intensity and affected areas have a blue background colour.

Criteria In Switzerland, a hazard rating system is used (Petrascheck & Kienholz, 2003) which defines different danger levels according to a combination of the hazard probability (defined with a return period or frequency of occurrence) and the hazard intensity (Figure 2-2). This is scheme is used for different natural haazrd types. Generally, the hazard intensity is strongly linked to the magnitude of an event. Therefore the assessment of magnitude-frequency relations for each hazard type is of primary importance. The probability and intensity of an event has to be assessed for all areas which are potentially endangered, for example on a torrential fan. Independent of the type of hazard, each class should have about the same effect on man and buildings. The concept of combining hazard intensities and probabilities to arrive at danger levels for hazard maps is described in (BUWAL/BWW/BRP 1997, BWW/BRP/BUWAL 1997).

23 Figure 2-2:

Hazard rating system used in Switzerland: red = high, blue = medium, yellow = low, white = no hazard. The hazard degree (three colors) is a function of the intensity and probability of an event.

In the Swiss recommendations, the intensity for debris flows and related processes is expressed as a function of impact pressure, flow velocity and/or flow depth. As flow depth is difficult to determine in some cases, an estimated depth of deposition may be taken as an approximation. Table 2-1 shows the critical values to delimit hazard intensity classes proposed in the Swiss recommendations (BWW/BRP/BUWAL 1997). One major objective of the hazard rating system is to define intensity classes which are comparable between different hazard processes. Therefore the critical values for debris flows and hillslope debris flows have been slightly modified here (Rickenmann, 2005b).

Table 2-1: Process low intensity medium intensity high intensity Snow avalanche (*) P < 3 kN/m2 3 kN/m2 < P 3 kN/m2 P > 30 kN/m2 Criteria to delimit intensity Debris flow (*) -- D < 1 m D > 1 m classes for debris flows and related processes and and (Switzerland). v < 1 m/s v > 1 m/s Dynamic flooding (*) v h < 0.5 m2/s 0.5 < v h < 2 m2/s v h > 2 m2/s

Debris flows, hillslope debris hd < 0.4 m hd < 1.0 m hd > 1.0 m flows (**) and (alternative criteria) and or v < 0.4 m/s

0.4 m/s < v < 1.5 m/s v > 1.5 m/s

P = avalanche impact pressure on an obstacle D = thickness of debris front v = flow velocity (flood or debris flow) h = flow depth

hd = depositional depth (*) Swiss recommendations 1997 (**)Values proposed by Rickenmann (2005b). In the Swiss recommendations, the criteria for the impact intensity are defined from the effect the impact forces have on the safety of human life and structures.

24 High intensity is defined as impact when human life even in a house is not safe, because either the house is deeply flooded or may collapse suddenly owing to the acting forces

Medium intensity is defined as danger to life outside the building. The intensity of impact is not sufficient to destroy a solid building but may damage parts of the structure. It is high enough to endanger the life of an unprotected person outside.

Low intensity is when danger to life is not significantly raised and no great structural damage is to be expected. However damage may occur inside the house. The typical case is shallow flooding with depths below 0.5 m.

In France, the current legislation also consists of recommandations based upon very general laws. This means that the procedure in itself is not compulsory. Regarding Table 2-1, inside the methodolical guide under preparation in France (not yet validated by the Ministry of Ecology and Sustainable Development), we have considered that the intensity is always high in zones reached by debris flows. Therefore, criteria on flow depth and velocity are only used for other kinds of torrential floods.

Degree of hazard or The classification follows the principle that the lower the probability, the higher the hazard zones admissible impact intensity for the same classification. Only for high impact intensities with danger to life, even within houses, is the hazard always considered high, as shown by the ‘red’ areas in Figure 2-2. The classification into different zones has direct consequences for land use planning as shown in Table 2-2 below.

Table 2-2: RED: high hazard • People are at risk of injury both inside and outside buildings. Definition of hazard zones or • A rapid destruction of buildings is possible. degrees of hazard, as applied in Switzerland (Petrascheck or: & Kienholz, 2003). • Events occurring with a medium intensity, but with a high probability of occurrence. In this case, people are at risk outside buildings, or buildings can no longer house people. The red zone mainly designates a prohibition domain (area where development is prohibited). BLUE: moderate hazard • People are at risk of injury outside buildings. Risk is considerably lower inside buildings. • Damage to building structure should be expected, but not a rapid destruction, as long as the construction type has been adapted to the present conditions. The blue zone is mainly a regulation domain, in which severe damage can be reduced by means of appropriate protective measures (area with restrictive regulations). YELLOW: low hazard • People are at low risk of injury. • Slight damage to buildings is possible. Damage might occur inside the building but not at the structure. The yellow zone is mainly an alerting domain (area where people are notified at possible hazard). YELLOW-WHITE HATCHING: residual danger Low probability of a high intensity event that can be designated by yellow-white hatching. The yellow-white hatched zone is mainly an alerting domain, highlighting a residual danger. WHITE: According to currently available information no or negligible danger. In France: we now propose (not yet validated by the Ministry of Ecology and Sustainable Development) to have only two classes of intensity : high and medium and a probability of reachness (of a given point by the flow) in 4 categories. The hazard is

25 defined by crossing these two scales and also by an index giving information on the considered phenomenon (debris flows, hyperconcentrated flow…), the intensity and the probability.

The matrix of Figure (2-2) was thoroughly discussed among different expert groups in Switzerland concerned with natural hazards. One group wanted to give the maximum of information, thus distinguishing between the 9 fields of the matrix. For another group two classes (interdiction and prescription) appeared sufficient. The divergence was mainly due to the experience obtained so far with hazard zones from avalanches, which have existed in Switzerland since 1984. The words ‘hazard zones’ were linked to high danger to life, which is true for avalanches. For this type of hazard the extent of low impact zones is small. This is in contrast to flood hazards, where shallow flooding is widespread and causes large material damages. This may be illustrated by comparison of the average annual damage in Switzerland. Floods cause two to three casualties and material losses of about 250 Mio. CHF. With avalanches, about 20 persons are killed annually but average damage remains in the order of magnitude of 20 to 30 Mio. CHF. This underlines the importance of the low hazard zone (yellow) for flood hazards. Since damages may be high the question is raised whether a prescription should be made also for the yellow (low hazard) zone. It was finally decided to include it as recommendation only. Thus, the cantons or the municipalities can set up regulations, but are not obliged to do so. This improved the acceptance and gave more flexibility to the municipalities when establishing building codes. Finally, it is in the interest and the responsibility of the land owner to protect himself and his goods against damage. He should not be freed from his obligation by numerous regulations which are difficult to control.

The matrix proposed in France has a similar shape to the Swiss one in Figure (2-2). However, the intensity being always considered high for debris-flows, only the probability of reachness (as defined previously) determines the hazard level. This probability is devided in categories: high, medium, low, potential (extremely low) and we do not directly refer to return periods of 30, 100 and 300 years as you do in Switzerland.

A particularity is the zone of residual risk. The idea may be illustrated with two examples. For rock avalanches the intensity is always high, thus it must be classified as high hazard (red). However, probability is extremely difficult to asses and in many cases very low, which would result in a zone with negligible hazard (white). To classify an area under this condition as either high or negligible hazard is unsatisfactory, resulting in either too stringent or too lax prevention requirements. Probability can be predicted better, if movements are observed, since rock avalanches announce themselves by an acceleration of movements. By classifying the area as ‘residual risk’, the local authorities can be required to establish an observation scheme, by which the potential danger is further assessed. Existing settlements can remain untouched until the observations show that it is not only a suspicion but a real danger. Another typical example are areas protected by dikes. The failure probability of a well maintained dike is very low. However, the safety of the areas protected by dikes is lower compared to high lying areas. The recent flood events of 1997 on the Oder

26 (Germany), 1999 in the canton Berne (Switzerland) and in Bavaria (Germany), 2000 in the canton Valais (Switzerland) and the Aosta region (Italy), and finally 2002 in Austria, Germany an the Czech Republic showed that an extreme event, larger then a usual design event can always occur. Therefore emergency organizations must be prepared for such events and very sensitive structures should be avoided in such areas. The identification of areas with such low hazard is therefore necessary, although it might not influence normal construction activity.

2.2 Hazard mapping tools

2.2.1 Identification of Debris Flow Hazard

With respect to the hazard assessment of torrent catchments, it is important to determine whether debris flows are likely to occur or not. The basic requirements for the occurrence of debris flows are steep slopes, sufficient volumes of debris material relatively easy to mobilize, and sufficient water to trigger the flow. Based on a geomorphologic assessment of the torrent channel and the fan, further characteristics may also be used to estimate a relative probability of debris-flow occurrence (Aulitzky, 1980; Rickenmann, 1995).

Debris flow hazard recongnition is a first important step and may be based on the following elements (Jakob, 2005):

- geomorphic evidence, including also the appearance of the alluvial or colluvial fan

- aerial photographs, satellite imagery, topography

- geomorphology and topography

- historical accounts and records

These elements give information on past events and for example on ancient channels likely to influence the flow direction. One main objective of this part is to understand the behaviour of the catchment in the past and to record areas historically threatened by torrential phenomena. This study sometimes allow a preliminary localisation and caracterisation of ancient catastrophic events and ancient countermeasures. It can provide information on the type of flood (debris flow, hyperconcentrated flow?) most often producing catastrophic events. A synthesis can take the form of a map of historical phenomena.

Recording of natural phenomena and countermeasures:

This study is carried out at the catchment scale. Its general objective is to compare erosive processes inside the catchment to technical countermeasures already set up to fight against these processes and assess the capacity of these latter to face all events likely to occur.

27 In a first time, a recording and localisation of processes likely to provide some material or woody debris to the floods is carried out and possibly mapped and qualified according to the intensity of its activity. This is carried out using aerial photographs, satelitte imagery, geological maps and specific field surveys whose objective is to precise the extension of main phenomena.

In the same time, existing technical countermeasures are recorded and localised. For each of them, it is necessary to control its general state, its design hypothesis, the frequency of keeping it in order actions, its behaviour during past events and the existence of some indentified person or service in charge of keeping it in order, when such information is known.

In this phase, it is also necessary to imagine the behaviour of the catchment and existing countermeasures in case of exceptional event and especially the threat generated by the protection structure when its capacity is overcome.

Land use:

The presence of infrastructures is likely to influence the flow direction.

Debris-flow occurrence Debris flows typically occur in a torrent catchment. The term torrent refers to a steep vs. fluvial sediment water course in mountainous terrain which is characterised by sporadic and sudden transport events high discharges of both water and sediments. A torrent catchment in the Alps usually comprises a catchment area less than one up to several ten square kilometres. The channel gradient may vary from more than 60 % to a few % in the fan area. Similarly as with snow avalanches and other geophysical mass movements, the torrent system and the debris-flow process can be characterised by three main zones: The headwater area or initiation zone where the flow is triggered, the transit zone (gully and channels) where entrainment of more solid material may occur, and the debris fan area where often major deposition takes place.

It is particularly important to identify the possible occurrence of debris flows on a given catchment, the subsequent threat being most of the time higher, debris flow is often the “reference flood”. The existence of other types of floods with bedload transport should not be neglected though.

The occurrence of debris flows requires that the following three parameters are present in a critical combination:

(i) sediment availability

(ii) steep slopes

(iii) water input

In torrent catchments, the conditions (i) and (ii) are often satisfied in many locations. In the European Alps and many other regions, the triggering factor (sufficient water input) is precipitation in most cases, and sometimes also snow melt.

28 A generally high availability of easily erodible solid material or increased sediment input to channels by temporarily activated slope movements are a preriquisite that one or more debris-flow events per year may occur over a longer time period. In a given torrent catchment, small to medium size events (up to a few ten thousand cubic metres) may typically occur with time intervals of years to decades, whereas very large events (more than about hundred thousand cubic metres) may happen only once or twice during a one-hundred year period.

Debris flows in the Alps may involve total sediment volumes of a few hundred to a some hundred thousand cubic metres. The sediment supply may either stem from point sources such as landslides or from incision of the torrent bed by vertical erosion. The total event magnitude is often used as a rough indication to characterise the intensity of a debris flow. This parameter largely influences the flow behaviour in the channel and – in the case of overtopping - the extent of the affected areas on the fan. Peak discharges, flow depths and flow velocities are essential parameters to determine possible impact forces and to estimate whether a debris flow can be conveyed through the existing channel on the fan or whether it will overtopp and possibly cause damage in the deposition area.

A distinction between debris flows and fluvial sediment transport based primarily on the Melton number has been proposed in several studies (Marchi & Brochot 2000; Bardou 2002; Wilford et al. 2004). Such a distinction may be helpful for a preliminary hazard assessment.

Disposition of a The occurrence of debris flows depends on several factors such as the hydrogeologic, catchment for debris- geologic and topographic conditions. Put in a simplified way, the following criteria flow occurrence must be satisfied: (i) availability of sediment, (ii) steep slopes, and (iii) sufficient water input. The parameters (i) and (ii) define the disposition or susceptibility of a catchment to produce debris flows. The parameter (iii) points to triggering conditions: If the stress by water infiltration and runoff exceeds a threshold value, slope instabilities and debris flows may occur.

Long-lasting rainfalls may increase the susceptibility to slope instabilities. In a catchment with limited sediment availability, weathering may be important to replenish a certain sediment potential necessary for debris-flow occurrence. If a stress is applied to the the sediment sources and the channels in the catchment, for example by subsurface and surface runoff following intense rainfall, a debris flow may be initiated.

The sediment sources may be concentrated in the headwater area at distinct locations such as moraines, talus slope or other glacio-fluvial deposits. Alternatively or additionally, the sediment material is entrained along the flow path.In this case there may be regular sediment input from small tributaries, or the bulk of the material may result from incision of the main torrent bed in easily erodable rock formations (e.g. shist formations).

29 2.2.2 Evaluation of debris-flow occurrence

Triggering conditions The most common trigger of debris flows is sufficient water input during precipitation and relevant factors events. Other sources of important water input into debris or soil material at steep slopes are snowmelt (during spring and early summer), melting ice during a volcanic eruption, glacial lake outburst floods, and sudden release of subsurface water stored near the outlet of a glacier drainage system.

The three main initiation mechanisms are landslide-type failures (which may also result in hillslope debris flows), channel-bed failure and temporary blockage of sediment and water flows in the channel, enhancing the surge-like flow behavior (e.g. Rickenmann & Zimmermann, 1993). Berti and Simoni (2003) mention two broad approaches which may be used to quantitatively assess the first two initiation mechanisms: Assessment of landslide derived debris flows typically relies on Coulomb failure coupled with critical state soil mechanics concepts for the explanation of failure-induced liquefaction of material. For the channel-bed initiation a hydraulic approach may be used to define critical conditions for debris flow formation (Tognacca et al., 2000; Berti & Simoni, 2005).

The mobilisation of debris material during debris flow events is related either to the onset of sediment transport due to water runoff or to slope failures caused by an increase in pore-water pressures. Both the runoff formation and slope instabilities are a function of rainfall intensity and cumulative precipitation (or water input in another form). Therefore, a critical combination of rainfall intensity and duration has been proposed as criterion to define the conditions above which debris flows can occur. Some threshold lines for debris-flow occurrence are compared with data from Switzerland in Figure (2-3).

Figure 2-3:

Relation between critical rainfall intensity and duration for debris-flow occurrence: data from the Swiss Alps (from Zimmermann et al., 1997a), and comparison with other threshold lines.

Two kinds of meteorological events can be distinguished for the triggering of debris

flows (they are also represented by the two clusters of data points in Figure (2-3)). Type A events: Short thunderstorms or high-intensity downpours are likely to generate flood events in torrents which promote debris-flow formation by material entrainment

30 along the main channel. In Switzerland, a minimal rainfall intensity of about 30 mm/h and a minimal cumulative rainfall of about 40 mm are necessary to trigger type A events. The precipitation may also partly occur in form of hail, a fact often emphasised by local observers. In Switzerland, most thunderstorm related debris flows occur in the months from June through September. Type B events: Sustained regional rainstorms result in substantial water infiltration, subsurface runof and partial or complete saturation of the soil. These soil conditions favour increased surface runoff, and the soils are susceptible to slope failures which may trigger the formation of debris flows. Another triggering situation is a combined effect of rainstorm and snowmelt. In Figure (2-3), the two Swiss debris-flow events with an average rainfall intensity below 1 mm/h occurred during snowmelt periods.

Note that there is a considerable uncertainty in defining threshold conditions as given in Figure (2-3). The rainfall data may not necessarily be representative for the debris- flow initiation site, have a limited temporal resolution, and the relevant rainfall duration is generally unknown. Furthermore, differences in geological and hydrogeological conditions are not accounted for in the above threshold relationships which are based on data from many different areas.

Initiation mechanisms In general, three typical initiation mechanisms can be distinguished (Figure (2-4)): (a) gradual or sudden entrainment of solid material by bed and bank erosion, possibly including a partial fluidisation of the bed; (b) slope instability, initiating a landslide which transforms into a debris or mud flow;and (c) a temporary blocking of the solid discharge in a stream channel. The blocking may be associated with pulsing bed load movement including coarse particles, and with channel constrictions, possibly favoured by woody debris. Alternatively, it may be caused by a sudden input of a large mass of solid material, either through a slope failure or by a debris flow from a tributary channel.

31 Figure 2-4:

Three typical debris-flow initiation mechanism (from VAW, 1992): bed erosion and fluidisation, slope instability, temporary blocking of water and solid discharge

Mountain torrents are often characterised by a very irregular channel morphology. Channel constrictions may be due to large boulders or bedrock outcrops. At such locations a temporary blockage of water and sediment flow may occur, possibly favoured by the presence of woody debris. The combination of woody debris and coarse particles is very effective in blocking a flow path, as is known from slit type outlet structures of sediment retention basins. Upstream of the temporary blockage, more sediment is accumulated until the pressure is sufficient to cause a collapse, suddenly releasing large quantities of sediment and water. Such a “dam-break” flow can transform into a debris flow.

Glacial lake outburst floods (GLOF’s) are characterised by the release of large quantitites of water and much greater peak flows as compared to a rainfall-induced flood in the same catchment. Below moraine-dammed lakes there are often ideal conditions for the intitation of debris flows: steep gradients and and easily erodable material with a wide grain size distribution. Therefore, many of the largest debris flows observed in alpine environments are the result of glacial lake outburst floods (Haeberli, 1983).

Magnitude and At present, there are no rigorous methods which would allow a strict assessment to frequency determine an exact probability of debris-flow occurrence, be it based either on physically measured characteristics of a catchment or on a statistical analysis. If there

32 is information available on past debris flow events, this is often the most reliable indication. It is noted that the concept of „recurrence intervals“, as borrowed from flood frequency analysis, may be problematic when applied to debris flow events (Davies, 1997). For example, temporary storage of sediments may be of importance, and thus an event may depend on previous ones. If information on historic events is available, it may be possible to determine a characteristic pattern between debris flow frequency and debris-flow volume for a particular catchment (Zimmermann et al., 1997a,b).

In most torrents, no measurements are available on past events. Nevertheless, a semi- quantitative or qualitative assessment of past events should be made as far as possible. Such an assessment may be based on historic documents (e.g. written documents from forest, torrent control, or river services), or on interviews with the local population. A terrain analysis, possibly supported by maps and aerial photographs and including the interpretation of geomorphic features (“silent witnesses”), can also give information on past events. These assessments are essential for determining a magnitude-frequency pattern of a given torrent and may also help to delimit endangered areas.

A very rough estimate of a potential (extreme) event magnitude may be made with empirical relationships which are generally based on morphometric parameters of the catchment. Several relationships are listed in Table (2-3). However, these relations can only give either a "maximum" or a mean value of a potential event magnitude, since its variability for a given catchment area is very large, typically covering several order of magnitudes. Figure (2-5) shows data on event magnitude of debris flows and/or torrential floods with sediment transport versus catchment area.

Table 2-3: Formula N Source

Kronfellner-Kraus (1984); 3 M = K Ac 100 Sc 1420 M Maximum event magnitude [m ] Kronfellner-Kraus (1987) Empirical relationships to 3 Ma Average event magnitude [m ] estimate the event magnitude M = 37000 A 0.78 ~ 65 Zeller (1985; “extreme c A Catchment area [km2] of debris flows and/or conditions for sediment c yield”) torrential floods with Sc Mean channel slope [-] sediment transport. 2.3 15 (*) Hampel (1977) Ma = 150 Ac (100 Sf -3) Sf Mean fan slope [-] Rickenmann and L Length of active channel [m] M = Lc (110 - 250 Sf) 82 c Zimmermann (1993) K Torrentiality factor [-] M = 13600 A 0.61 551 Takei (1980) a c IG Geologic index [-] 0.67 D’Agostino (1996) N Number of events [-] Ma = 29100 Ac 64

1.28 D’Agostino & Marchi (2001) Ma = 70 Ac Sc IG 64

33 Figure 2-5:

Data on event magnitude of debris flows. A few data points from Italian events refer to torrential floods with sediment transport.

In order to improve the relationships, a more detailed assessment of the geologic and lithologic factors is necessary, similarly to the approach of D’Agostino et al. (1996) in Table (2-3). Data handling with a GIS will help to process such spatially variable information. In smaller catchments with a high sediment potential, the surface runoff during a rainstorm may be the limiting factor for the debris-flow event magnitude. In this case, the estimation of a flood hydrograph can serve as a basis to determine a debris-flow event magnitude, for example by assuming an average solid concentration of 70 % by volume.

Typically, most of the sediments feeding a debris flow originate from locations close to the channel or from channel erosion. Therefore, another simple method consists in estimating maximum channel debris yield rates along characteristic reaches, and summing them up over the expected flow length. Tentative channel debris yield rates are given by Hungr et al. (1984) based observations in western Canada (Table (2-4)). For some larger debris flow events in Switzerland, mean sediment yield per length of active channel were from 40 m3/m to 90 m3/m, and locally eroded cross-sections of 500 m2 to 650 m2 were observed (Rickenmann & Zimmermann, 1993). Similar values resulted after the passage of glacial lake outburst floods in Switzerland (Haeberli, 1983).

Channel Erodibility Channel Gradient Bed Table 2-4: Side slopes Stability condition1 debris yield coefficient type [deg] material rate2 [m³/m] [m³/(m.km)] Tentative estimated channel Stable, practically A 20 – 35 Bedrock Nonerodible 0 – 5 0 – 5 debris yield rates in western bare of soil cover Canada (Hungr et al., 1984). Thin debris Nonerodible B 10 – 20 or loose soil Stable 5 – 10 5 – 10 (bedrock) over bedrock

Deep talus Less than C 10 – 20 Stable 10 – 15 10 – 15 or moraine 5m high

Deep talus Talus, over Side slopes at D 10 – 20 15 – 20 15 – 30 or moraine 5m high repose

Side slopes Up to 200 Deep talus Talus, over Not E 10 – 20 potentially unstable (consider as or moraine 20 m high applicable (landslide area) point source)

1 Prior to the expected debris torrent event 2 For drainage areas of 1 – 3 km²

34 A more detailed method consist in classifying homogeneous subunits of a catchment according to their role in sediment production and potential sediment delivery to a channel during a flood or debris-flow event. Such an assessment can be made with the help of a GIS. A preliminary assessment may include only the major relevant factors such as slope gradient and exposition, proximity to a channel, type of geology and lithology (Zimmermann & Lehmann, 1999). A more reliable estimate of a potential event magnitude requires a detailed geomorphologic field assessment of potential sediment sources in the whole catchment and in particular along the channel network (Spreafico et al, 1996). In case of larger slope instabilities it may also be necessary to make a geotechnical assessment, possibly analyse soil samples or even set up a monitoring program.

As can be inferred from the above discussion, it is often difficult to make reliable predictions also about the frequency of future events. In the absence of any information on past events, a rough estimate of the occurrence of future events may be made via the probability of triggering (rainfall) events. It is obvious that the main limitations to applying statistical approaches are the lack of data and the complexity of the initiation process.

Debris flows and torrential floods tend to entrain material in the headwater area and along the steeper reaches of the flow path (transit zone). Massive erosion of the channel bed and the adjacent side-slopes may occur, involving dowcutting of several metres. The frontal part of the flow acts to destabilize both the bed and the banks while material is entrained by the more fluid rear part of a surge or by torrential flood waters. Woody debris lying in the channel is often carried along by debris flows. If the entire debris flow surge is already heavily loaded with debris, the risk of substantial erosion in the fan area is smaller than for water floods with low sediment concentrations.

A detailed assessment of a potential future event magnitude often requires field investigations. On the one hand, reconstruction of past events may be improved by an interpretation of geomorphic features and debris flow deposits. On the other hand, the potential magnitude of a future event is estimated most reliably by mapping and assessing all possible sediment sources which may trigger or feed a debris flow. The interdisciplinary character of the phenomenon and of the triggering processes ideally requires to study the geomorphology, geology and lithology, channel network and hydrology, and vegetation of a torrent catchment. The interpretation of related investigations helps to estimate the thickness and areal extent of sediment sources, and their likelihood to contribute to a debris flow event. Precise quantitative methods are generally lacking because the mobilization of debris is a complex process depending on a number of factors, which show a large spatial variability, and in addition the water conditions may strongly vary during an event. Data acquisition by remote sensing techniques or by laboratory analysis of soil samples can improve the interpretation based on field observations.

2.2.3 Dynamical Behaviour and Runout Distance Prediction of Debris Flow

A major step in a debris-flow hazard assessment is the delineation of endangered areas. A discussion of methods for runout assessment of debris flows is given for example in Hungr et al. (1984), Cannon (1989), Bathurst et al. (1997), Fannin and Wise (2001), McDougall and Hungr (2003), and a recent summary is provided by Rickenmann (2005). Prediction of runout distance may be divided into empirical-statistical and dynamic methods. Though empirical-statistical approaches are often easy to use, they should only be applied to conditions similar to those on which their development is based. Apart from total travel distance or runout distance on the fan, there are also empirical approaches to estimate peak discharge and flow velocity. Dynamic models are physically based and consider the momentum or energy conservation of the flow. Such approaches for predicting velocity and runout of mass movements have been developed either for mass point or lumped mass models, or for continuum based models which also simulate the deformation of the moving mass along the flow path. A major difficulty in developing dynamic models for debris-flow runout prediction is the choice of appropriate friction parameters or material rheologies (e.g. Hungr, 1995; Iverson, 1997, Rickenmann et al. 2005; Naef et al., 2006). Partly similar approaches are being used for snow avalanches (Barbolini et al., 2000) and for landslides or rock avalanches (Crosta et al., 2004). A recent summary of runout prediction methods for debris flows is given in Rickenmann (2005).

2.2.3.1 Empirical-statistical approaches

Peak discharge and In Figure (2-6), data of peak discharge and debris-flow volume are shown. Also plotted peak flow or front is a semi-theoretical line which is derived from the assumption that Froude scaling velocity must be satisfied for flows of different size but having essentially the similar physical properties. The corresponding equation has the following form:

Qp = 0.1 M5/6 = 0.1 M0.833 (1a)

This equation is valid for granular or bouldery type debris flows. Similar equations are given in Table (2-5). In these equations M is given in [m3] and Qp, in [m3/s]. Note that the exponents are close to the exponent required from Froude-scaling as given in eq. (1a). The data presented in Figure (2-6) pertain predominantly to debris flows with a granular front, except for the Jiangjia ravine. The data representing mud flows and those representing volcanic debris flows with predominantly fine material have smaller peak discharges for a given debris-flow volume than the granular type debris flows.

36 Figure 2-6:

Relationships between debris-flow peak discharge and event magnitude (from Rickenmann, 1999).

Table 2-5: Data basis Formula Eq. Source

0.780 Empirical formulae relating granular debris flows (Japan) Qp = 0.135 M (1b) Mizuyama et al.(1992) debris-flow peak discharge and event magnitude (from 0.90 bouldery debris flows Qp = 0.04 Mw (1c) Jakob (2005) Rickenmann, 1999; Jakob,

2005). 0.790 muddy debris flows Qp = 0.0188 M (1d) Mizuyama et al.(1992)

1.01 volcanic debris flows Qp = 0.003 Mw (1e) Jakob (2005)

0.831 Merapi volcano (Indonesia) Qp = 0.00558 M (1f) Jitousono et al. (1996)

0.870 Sakurajima volcano (Japan) Qp = 0.00135 M (1g) Jitousono et al. (1996)

To describe the flow behavior of debris flows, a number of approaches and flow resistance equations have been proposed. To evaluate constitutive equations for the shear behavior of different debris-flow materials, detailed measurements under natural conditions would be required but are difficult to obtain. Therefore many of the proposed flow-resistance equations for the mean velocity are based either on empirical data of mean flow parameters of prototype flows or measured velocity profiles in laboratory flows, in which simplified material mixtures have been used to simulate debris flows.

The peak or front flow velocity V of debris flows may be estimated using a Manning- Strickler type equation (Rickenmann, 1999):

V = (1/n) h0.67 S0.5 (2)

where h is the flow depth, S is the channel slope, and pseudo-Manning n values are around 0.1 s/m1/3. Similar empirical equations were developed for the mean velocity of debris flows in Russian, Chinese and Japanese studies, as reported in Costa (1984).

Figure (2-7) compares the application of the Manning equation to debris flows, where different pseudo-Manning n values have been selected for each of the datasets A, B, C, and D listed in Table (2-6). For each dataset, a best fit mean n value was determined (Rickenmann and Weber, 2000); these values appear to reflect a difference in scale and

37 material composition. The coarser grained alpine type debris flows (set A and C) tend to require a higher value for n than the finer grained mudflows and lahars (set B and C). This trend is in general agreement with the grouping of the relationships between Qp and M in Figure (2-6). An analogous conclusion can be made from the analysis of surface velocity measurements of debris flow surges in the Kamikamihori valley in Japan where the friction coefficient (similar to a Chezy value) decreases with increasing content of coarse gravel in the flow (Suwa et al., 1993).

Figure 2-7:

Application of Manning equation (2) to debris flows, using a pseudo Manning n value back-calculated for each sub-dataset (from Rickenmann 1999, Rickenmann & Weber 2000).

Table 2-6: Data- Flow type Country/Region N Source set Arattano et al. (1996) Data on mean flow velocity (A) granular debris Italy, T. Moscardo 7 flows of debris flows (from Rickenmann, 1999). Japan, Kamikamihori valley 12 Okuda & Suwa (1981); H.Suwa, written comm. (1997) USA, Mt. St. Helens 6 Pierson (1986) (Shoestring Site) (B) mud flows China, Jiangia gully 33 M. Jakob, written comm. (1995); Z. Whang, written comm. (1997) (C) mostly granular Swiss Alps 29 VAW (1992); M. Zimmer- debris flows mann, written comm. (1996) (D) lahars USA, Mt. St. Helens (Pine 20 Pierson (1985) Creek and Muddy River) Columbia, Nevado del Ruiz 17 Pierson et al. (1990)

Analyzing clear water flows in torrents and gravel-bed rivers, an empirical equation has been developed where the mean flow velocity is expressed as a function of the discharge, Q, the slope, S, and the characteristic grain size d90, for which 90% of the bed material is finer in diameter (Rickenmann, 1994, 1996). For the debris-flow data of Table 3, there is not sufficient information on grain size distribution, therefore a simplified version of the equation has been applied (Rickenmann 1999):

V = 2.1 Q0.33 S0.33 (3)

where V is in [m/s], Q in [m3/s] and S is a fraction (sin of the bedslope angle). Application of eq. (3) results in a reasonable agreement between calculated and observed velocities for debris flows, with a similar scatter as for clear water flows, as shown in Figure 2-8. However, for the debris-flow dataset D (lahars) eq. (3) with the

38 constant 2.1 systematically overpredicts observed velocities.

Figure 2-8:

Application of empirical velocity equation (3) to debris flows (of data sets in Table 222-4); equation was first developed for water flows (from Rickenmann 1999).

Velocity estimates of the peak flow part of a debris or mud flow can also be obtained

from geomorphic evidence in channels regarding superlevation of the flow in bends or run-up heights against obstacles such as a bedrock ridge. The superelevation equation is:

V = (k1 rc g cos β tanδ)0.5 (4)

where k1 = correction coefficient, rc = radius of curvature of the bend, g = acceleration due to gravity, β = superelevation angle, δ = channel slope. In many debris-flow studies k1 = 1 is assumed (e.g. Costa, 1984). Hungr et al. (1984) and Bulmer et al. (2002) suggest to use k1 = 0.1 to 0.5. The equation with the run-up height is:

V = (2 g Δh)0.5 (5)

where Δh = run-up height.

Total travel distance Several empirical approaches to estimate debris-flow and landslide runout distance are and and runout in discussed by Bathurst et al. (1997). It is often assumed that debris-flow deposition depositional zone begins at slopes of 10°. These approaches estimate runout distance using one or a combination of the following factors: flow volume, mean slope of the transportation zone, elevation difference from the starting point to the point where deposition begins, or travel angle. Bathurst et al. (1997) tested various approaches with landslide and hillslope debris-flow data from Idaho and found that none of the approaches are accurate over the full range their field data. To estimate sediment delivery into streams at a watershed scale, they recommended using at least two approaches in parallel, thus providing information on the uncertainty of the prediction.

For a given volume, debris flows usually show a larger mobility, or lower travel angles, than landslides and rock falls (Figure (322-4)); see also Corominas, 1996; Iverson, 1997; Rickenmann, 1999; Legros, 2002). In a study of the mobility of 204 landslides, Corominas (1996) concluded that the scattering in the plots of travel angle versus event

39 volume is partly due to effects of obstacles and topographic constraints, which generally tend to reduce the landslide mobility. For a subset of 71 debris flows, including debris slides and debris avalanches but excluding mudflows and mudslides, he developed an empirical relationship, which may be given as:

L = 1.03 V 0.105 H (6)

where H is in m, L in m, V in m3, and V varies between 102 and 1010 m3. Equation (6) has a correlation coefficient r2 = 0.76 for this dataset. Both unobstructed and channelized debris flows show a similar mobility, although the latter ones are reported to be slightly more mobile. Unobstructed flows refer to mass displacements over a regular topographic surface without obstacles or lateral restrictions. For volumes smaller than 104 m3, debris flows occurring in dense forests appear to be clearly less mobile than channelized or unobstructed flows. A regression analysis using mainly debris-flow data from the Swiss Alps produced the following equation (Rickenmann, 1999):

L = 1.9 V 0.16 H0.83 (7)

where L is in m, V in m3 and H in m. Equation (7) has a correlation coefficient r2 = 0.75 for the dataset which includes 160 debris flow events with L ranging from 300 to 12,600 m, M from 7x102 to 106 m3, and H from 110 to 1820 m.

Figure 2-9:

Travel angle vs. volume of mass movement. Data include 140 debris flows from the Swiss Alps and 51 large landslides/rock avalanches (for data sources see Rickenmann 1999). Also shown is the approximate range for 101 data points of unobstructed landslides plus unobstructed and channelized debris flows discussed in Corominas (1996).

A comparison of equation (6) and (7) with debris flows data show similar scatter of the

data in relation to the two equations. An example of applying (7) is shown in Figure (322-5), where additional data on rock avalanches follow a similar trend but have comparatively shorter travel distances than the debris flows, which may be explained by lower water concentrations in the flowing masses. The Mount St. Helens lahars and the Nevado del Ruiz mud flow, on the other hand, have comparatively high L values, possibly reflecting either larger water concentrations in the flowing mixtures or higher clay and silt contents compared to the majority of the mostly alpine-type debris flow data.

40 Figure 2-10:

Observed runout distance of debris flows and rock avalanches compared to predictions with equation (7), which was derived for debris flows. Data include 140 debris flows from the Swiss Alps and 51 large landslides/ rock avalanches (for data sources see Rickenmann, 1999).

When applying equation (6) or (7) for predictive purposes, a debris-flow starting point

and the longitudinal profile of the expected flow path must also be defined. For a given estimate of V, a solution for L can then be determined either graphically or mathematically. A similar procedure has been proposed by Iverson et al. (1998) to predict the depositional extent of lahars with the LAHARZ model.

Based on almost the same data set as used by Rickenmann (1999), including 144 debris flows from the Swiss Alps, Zimmermann et al. (1997a) defined a lower envelope of tan β as a function of the catchment area Ac [km2]:

tanβmin = 0.20 Ac-0.20 (8)

Equation (8) may be used to determine a probable maximum travel distance (Figure (322-6)). Including eight additional observations from the Canadian Cordillera, the lowest travel angles (lowest value tanβmin = 0.07) observed for the Swiss data are associated with debris flows with a high proportion of fine material (data points labeled “matrix supported”). For coarser-grained debris flows (data points labeled “clast supported” or “sand and gravel”), the minimum observed travel angle is tanβmin = 0.19 (Rickenmann and Zimmermann, 1993).

41 Figure 2-11:

Travel angle of debris flows vs. watershed area. Data include 144 debris flows from the Swiss Alps, where the deposits are characterized as “cs” = clast supported, “sg” = sand and gravel, and “ms” = matrix supported (modified from Zimmermann et al., 1997).

Cannon (1993) developed an empirical model for the volume-change behavior of

debris flows, based on 26 debris flows in Hawaii. The initial debris-flow volume must be known or assumed, and no material entrainment is considered. The average volume change is assumed to be a function of the slope angle along the flow path and the degree of flow confinement by the channel. Consideration of vegetation type through which the flow traveled did not significantly improve the model. A similar but more comprehensive approach used 449 landslide/debris flow events from Queen Charlotte Island, Canada, on clear-cut, glaciated hillslopes (Fannin and Wise, 2001). Both entrainment and deposition are considered for the volume balance. Most initial failures volumes are less than 500 m3, with peak volumes less than 4000 m3, and travel distances less than 600 m (Wise, 1997). Model development is based on 131 debris- flow events comprising morphological information on 533 reaches. Five different regression equations describing the volume change were developed, considering separately three typical flow modes (unconfined, confined, and transition flow). Entrainment and/or deposition functions are specified for different slope ranges, and depend on several of the following geometric and derived parameters for each reach: slope angle , reach length, width of entrainment or deposition, entering flow volume, and a bend-angle function (which depends on vertical and horizontal change in flow direction and on the entering flow volume).

The empirical-statistical LAHARZ model (Iverson et al., 1998) predicts runout lengths and areas affected by lahars, and is based on a scaling analysis of key parameters and a statistical analysis using data from 27 lahars at nine volcanoes. Two semi-empirical relationships are used in the model: one between the cross-sectional flow area A and the lahar volume V, and the other between the planimetric area B of the inundated valley and the lahar volume V; the latter relationships is of the form B ∼ V2/3. The method has a high potential to be also applicable to debris flows, as suggested by limited data included by Iverson et al. (1998). These data show a slight systematic deviation from the trends defined by the lahar data, indicating that non-volcanic debris flows move less fluidly and form appropriately thicker deposits than most lahars. A

42 similar conclusion is made by Crosta et al. (2003) who derived a similar relationship between the planimetric area B the debris-flow volume V based on 116 granular debris flows in the Italian Alps. The approach of Iverson et al. (1998) has been implemented by Hofmeister and Miller (2003) in a GIS-based model of debris-flow deposition zones for regional hazard assessments in Oregon.

2.2.3.2 Analytical-deterministic approaches

For a given torrent channel and fan topography, larger flows tend to result in longer

runout lengths than smaller flows having similar material properties. Empirical evidence shows that larger debris-flow volumes also result in higher peak discharges (Mizuyama et al., 1992; Jakob and Bovis, 1996; Rickenmann, 1999; Jakob, 2005) which in turn are associated with higher flow velocities and/or larger flow cross- sections and flow depths.

Based on a momentum consideration for a flow traveling over a surface with a constant slope, the runout length can be described by a theoretical equation (Takahashi, 1991; Hungr et al. 1984), which requires the following input parameters: runout slope angle, entry channel slope angle, entry velocity, entry flow depth, and the friction slope which is assumed to be constant along the runout path and accounts only for sliding friction. The model assumes a constant discharge from upstream and that there is no change in flow width after the break in slope. It is shown in Rickenmann (2005) that for given values of slopes and entry flow depth for typical runout conditions on alpine torrential fans, the runout length calculated with this approach assumes a minimum value for entry flow velocities ranging between about 2 m/s and 4 m/s. This observation may have some practical relevance in that uncertainties on the entry velocity in this range are not so sensitive for predicting runout with the given approach. The key parameter in the approach is the friction slope along the runout path which can only be determined by back-calculation with observational data. It may be noted that a similar runout model as that of Takahashi-Hungr for debris flows had earlier been developed for dynamically similar snow avalanches (Salm, 1966).

Rickenmann (2005) compared analytical expressions proposed for the estimation of the runout length of avalanches (Voellmy,1955), landslides (Van Gassen and Cruden, 1989) and debris flows (Perla et al.,1980; Körner, 1980). He concluded that these models will result in different values for the friction slope if applied to the same case. The mass point model of Perla et al. (1980), was applied by Zimmermann et al. (1997a), Gamma (2000) and Genolet (2002) to a total of 75 Swiss debris flow events. In these studies, back-calibrations of the two model parameters were made based on travel distance information and in some cases also on flow velocity (estimated from superelevation). However, in particular the results of Genolet (2002) showed that the friction slope is strongly correlated with the terrain slope at terminal deposition of the debris flow, a situation where the calculated runout distance is very sensitive to small changes in the friction slope. This two parameter mass point model incorporating the Voellmy fluid approach has been used for preliminary hazard assessments in Switzerland.

43 Iverson (1997) developed a two-phase mixture model that generalized the Savage- Hutter (1989) granular avalanche model, and Iverson and Denlinger (2001) extended the one-dimensional Iverson (1997) model to multidimensional flow. In the context of assessing the effect of organic debris on debris flow runout at the watershed scale, Lancaster et al. (2003) use a simplified form of the Iverson and Denlinger (2001) equation, where the debris flow is treated as a one-dimensional point process, and the velocity and depth are functions only of time. Changes in density and depth are prescribed by entrainment of sediment, wood and water and by changes in channel and valley geometry. The effect of wood on debris flow runout is considered by accounting for the momentum needed to accelerate entrained wood and its effect on bulk density and flow depth. The Lancaster model was calibrated by adjusting the pore pressure term such that the simulated debris-flow travel distance distribution best matched the observed distribution. A pore pressure of 1.8 times the hydrostatic pressure resulted in a satisfactory calibration. The calibration procedure included 29 debris flows mapped in a study area in the Oregon Coast Range, which have travel distances from 30 to 680 m and known deposit volumes from 35 to 2500 m3. The average wood fraction in the deposits was 60% by volume. Disregarding wood effects substantially increased modeled runout distances for flow conditions associated with larger travel distances including wood effects (Lancaster et al., 2003).

2.2.3.3 Simulation models for the propagation and arrest of flows of non-Newtonian fluid

Predicting tools aiming to consider, to some extent, the dynamics of debris/mud flows under unsteady conditions, have to rely over momentum and mass conservations of each phase devised in the flow mixture. As a result, these equations allow the determination of the flow parameters and deformation of the mass along the entire track, including deposition. In this type of approach the first thing is therefore to develop an appropriate mathematical model, as herein pointed out; the subsequent use of a numerical method allows to obtain approximate solutions. The mathematical model may describe the motion of an assembly of particles elements, in a Lagrangian view, or of a continuum flowing domain, in an essentially Eulerian view. Methods belonging to the Lagrangian family are boosting up their potentiality more and more. Particle dynamics, cellular automata, SPH are some of these approaches. Nevertheless, their use is mostly limited to fundamental issues in the research field, in test or laboratory cases, and seem yet to encounter problems (e.g., because of exceedingly long computing time) in real-event applications. Continuum based simulation models, on the contrary, have probably come further in exploiting their skills, and have been extensively applied to real cases. For this reason they only are considered in the following. One more feature in the continuum approach, in practical applications, is the employment of vertically (or section) integrated flow models, where the physical domain is therefore reduced to 1D or 2D. The description of debris flow is in this way entrusted a shallow-water type model, with clear advantages: the relevant computing codes may exploit the huge numerical machinery already developed for riverine flows with/without sediment transport. 3D models for non-Newtonian fluid are anyway available (Frenette et al, 2002).

44 The mathematical model consists of a system of partial differential equations (PDEs) which, to be handled, may be supplemented with phenomenological closures. These closures typically are algebraic relations expressing the averaged solid concentration (or the solid discharge, in alternative), the averaged (on the flow section boundary) shear stress, the rate of erosion/deposition and so on, depending on the specific modelling assumptions.

Among the many physical issues which take place in flows with high solid concentration, there are two major distinctions which are here deployed as priority. The first one (choice A) is the distinction between flows exchanging mass (liquid and solid) and attendant momentum through the instantaneous bed-profile, and flows running over underformable walls. Mobile (case A1) or rigid (case A2) bed are referred to point out either possibility. The second one (choice B) has strong links with rheological issues, and concerns the type of fluid, which may be a mixture of distinguished coarse grains and water (case B1), or a suspension of fine solids in water (mudflows, case B2). In B1 option, different rheological laws may be assumed for the liquid and the solid phases, whereas choice B2 means that the inners stresses refer to the bulk-density fluid. Choice A and B are somehow correlated: mixture with coarse grains are likely to produce depositions and erosions, whereas suspension are unlikely.

As far as choice A is concerned, Figure 223-7, concerning a mixture with well sorted, coarse grains, helps the reader to realize as it significantly affect the phenomenon. In this Figure, yw represents the free-surface of the liquid phase, β is the slope angle, the arrows mean direction of increasing/decreasing of the indicated quantities, Cs is the depth-averaged concentration. They show that, for flows in uniform conditions over a loose-bed (case A1-B1), different flow regimes take place as the bed slope changes, and that the Froude number reduces as the slope increases (Armanini et al., 2005). The regimes ar referred to as sheet, mature and plug flow. The mature flow presents a unique free-surface for the solid and liquid phase, and takes place at a slope value which may be determined (Armanini et al., 2005). At slope values smaller or greater than the mature limit, sheet flows, having a liquid layer on the top, and plug flows, having a dry-solid layer on the top, show up. Accordingly, as the slope increases, there is a solid-concentration increment and a progressive dominance of frictional stress- mechanism layers over collisional ones. These findings contrast with the case of flows over a rigid wall (case d in the Figure), where different trends are observed. A theoretical description of this matter, despite its relevance, is not yet available, because of the difficulties arising in dealing with the involved rheological matter (Armanini et al, 2005). Therefore, the exact phenomenological closures, such as that one expressing the depth-averaged concentration, may not be known. In the following section “Examples”, approximated phenomenological closures will be provided. The case of a rigid bed (A2) is generally less difficult to be modeled, since the sediment concentration is not constrained to reach a loose-bed on the wall, and, as a consequence, only collisional stresses (and no frictional ones) may be present throughout the flow depth. Furthermore, only the shear stress on the boundary has to be assigned.

45 Figure 2-12:

Typology of the examined flows: (a) loose bed, immature flow (sheet-flow); (b) loose bed, mature flow; (c) loose bed, plug flow; (d) solid bed flow.

Cases with spatial or temporal transitions from rigid to movable bed should also be

considered (i.e., when the flow is running over a recent deposition or when protection works as sills or check-dams are encountered), making the whole analysis more difficult.

As far as choice B is concerned, the interpretation of the rheological inner stresses depends strongly on it. In debris flows (B1), separate consideration of stresses in the fluid and in the liquid phases may be advised. With regard to the solid phase only, collisional and frictional regimes may be considered. The former may be represented by the seminal relationships provided by Bagnold (1956). Recent relevant interpretations (Jenkins and Hanes, 1998, Armanini et al., 2005) are based on the kinetic theory for gasses. The latter is considered in different ways; among others, it is here reminded the interpretation by Savage (1998) and that one by Jop et al. (2006), who, in different ways, propose a viscous-plastic type consitutive law. In this two- phase view of the mixture, the shear stresses inside the liquid-phase are usually neglected in comparison with those ones in the solid-phase. On the contrary, the water is usually assumed to reduce the normal stresses of the solid phase by the Archimedean factor. This point is effective in models which exploit a Coulomb-type stress for the frictional stresses. For suspended, or mud flows (B2), the flow may be seen as formed by a non-homogeneous, single-phase, fluid. The shear stresses are non-Newtonian viscous, with strong dependance of viscosity over the solid concentration and shear rate (Phillips et al., 1992; Quemada 1998). The phenomenological closures require, to be obtained, the vertical integration throughout the flow depth, of the solid concentration and of velocity. It may be performed only numerically; final results cannot be therefore expressed, at the moment, in some closed form. In the field of suspension flows, significant contributions have not yet appeared, able to manage the routing of real floods down gulleys or slopes. Indeed, the routing of real floods down gulleys or slopes are usually dealt with by some simplified models for homogeneous fluid (although it is not), such as Bingham, Herschel-Bulkley, Voellmy or others, in which the rheological parameters are assumed constant in time and space.

In particular, no published model (to best of writer’s knowledge) has till now considered the case (A1-B2), that is the case of a suspension flowing over a mobile bed, with erosion and deposition stages of the suspended solid. Given the above major classification, it is now possible to proceed with a short resume of works dealing with real applications.

In many numerical debris-flow simulation models, the solid-fluid mixture is considered as a quasi-homogeneous fluid (A2-B2). These models may consider only deposition phenomena, which take locally place when the flow stops. Situations concerning the

46 remobilization of some underneath layer of mud is physically possible but rarely considered (Zanuttigh and Lamberti, 2004). A number of these models are partly or fully based on a rheological formulation for a Bingham or viscoplastic fluid (e.g., Fraccarollo and Papa, 2000; Imran et al., 2001; Laigle et al., 2003, Liu and Huang, 2006) and employed in 1D or 2D applications. Several applications to natural debris flows modified the pure Bingham model by adding a friction term accounting for channel roughness and turbulence (O’Brien et al., 1993; Han and Wang, 1996; Jin and Fread, 1999). More recently, a comparison of different modeling approaches with field observations in the European Alps (McArdell et al., 2003; Rickenmann et al., 2005) and at the Kamikamihori valley in Japan (Rickenmann and Koch (1997) used the following flow laws (or constitutive equations) in the simulation models: turbulent laws (e.g. Manning, Chézy), some having also a stop term (e.g. Voellmy), laminar (Bingham, Newtonian laminar, Generalized Viscoplatic Fluid), and inertial formulations (dilatant/grain shearing) as well as combinations of flow laws when appropriate. If the correct stopping location was simulated in the one-dimensional (1D) applications, it was possible to approximately match the velocity observations in most cases. However, in this case there is a general trend of under-prediction of flow depth or flow area. A general conclusion from the two-dimensional (2D) simulations is that the topography (for given input conditions) plays a key role with regard to the deposition pattern. A more detailed spatial resolution of the channel and fan topography strongly improves the model results in many cases. In comparison, the choice of a particular constitutive equation appears to be relatively less important.

Iverson and Denlinger (2001) extended the one-dimensional Iverson (1997) model to multidimensional flow (see also section 13.2.3). Mixture theory is used to develop the two-phase flow equations. The approach assumes that the apparent debris rheology can evolve as debris-flow motion evolves. Constitutive equations (or rheologic "laws") are specified for the solid and fluid phases independently of one another, and the solid and fluid phases interact through coupling stresses. Iverson and Denlinger (2001) assume that the coupling stresses result from Darcian flow of fluid relative to the solid grains, and that this flow evolves with time and position. In their approach pore pressure may vary within the solid-fluid mixture. The model is reported to require little or no calibration because model parameters like basal and internal friction angles of solid grains, fluid viscosity, and mixture diffusivity can be constrained by independent measurements. The model appears to perform quite well in predicting runout distance (R. M. Iverson, written comm. 2003). Being the bed considered undeformable, this model is to be considered a (A2-B1) type. In a similar modeling framework, Hungr (1995) proposed a one-phase model for rapid flow slides and debris flows that better represents some soil mechanics properties of the fluid-solid mixture. Different rheological laws can be selected. As in Iverson’s works, the friction stress is ruled out by pore-pressure conditions that may be intermediate between fully drained and liquefied. The model has been applied to several field cases (Ayotte and Hungr, 2000), and an extended version is under development (McDougall and Hungr, 2003). Interesting enough, Hungr’s (1995) model incorporated material entrainment and deposition along the flow path, at a rate which is prescribed (as input variables) instead being linked, by closure assumption, to the flow dynamics. All the models reported in this paragraph are also interesting for the analysis of lateral spreading of the debris

47 flow over a fan.

Stepping forward towards two-phase models, it is often assumed that inertial grain flow determines the flow behaviour. Most of these models closely follow the approach of Takahashi et al. (Takahashi, 1991; Takahashi et al., 1992; Nakagawa et al., 2000; Ghilardi et al., 2003). In general, these two-phase models allow for a variation in sediment concentration and can also calculate bed level changes (A1-B1). In these models it is anyway generally different the way they deal with phenomenological closures and with coupling between flow dynamics and bed changes. Sometime the density change of the bulk density is neglected. Valiani and Caleffi (2003) and Fraccarollo,et al.(2003), were careful in dealing with the coupling effects under very high transient conditions. IMPACT works (European Project of the Fifth (EC) RTD Framework Programme (1998-2002, Contract No. EVG1-CT-2001-00037) were particularly focused on extreme flood processes including flood propagation with sediment movement. In July 1996, the collapse of a secondary dam along the Lake Ha!Ha! reservoir resulted in a severe dam-break wave with spectacular morphological changes in the 30 km long river, till the confluence with the Saguenay River in the Ha!Ha! Bay. Many models, from different Institutions, have been applied to this case, resulting that, although quite sophisticated, models essentially did not fit very well with the huge amount of observed data.

Some words are, in the end of this Section, devoted to the use of the FLO-2D (O’Brien et al. (1993), a commercially available code, quite often employed for debris/mud-flow propagation over a rigid topography (type A2-B2), also with research purposes. For several 1D and 2D simulations, a comparison of the relative importance of the three friction terms of the hybrid FLO-2D flow law approach was made. In this approach, the total friction is determined as a combination of yield, viscous, collisional and turbulent stress components where the latter two are represented by one single coefficient. The comparison indicates that in many cases the turbulent-collisional friction term dominates for faster moving flow in channelized reaches; at slower velocities and in particular for depositional flows on the fan outside the channel, a yield strength or stopping term becomes more important as well (Rickenmann et al., 2005). As a relatively simple model, the Voellmy fluid approach was found to be rather robust in terms of numerical stability of the simulations, and the back-calculated Voellmy parameters are consistent between the 1D and the 2D simulations.

In conclusion, for many of these continuum based models, application to a real debris- flow event produced reasonable agreement with the observed deposition pattern. However, for reliable run out prediction, the models should be first tested rigorously against several field events. The simple friction law approaches in many of these models require either estimation of the magnitudes of the friction parameters or calibration of the models to match previous events. Being data on past event available or not, it is anyway suggested to carry out laboratory experiments (accepting to reduce the coarse grain-size fractions if too big) devoted to explore the rheological behavior of the mixture involved in the flow. By doing so, with the parameters of the phenomenological closures determined in an independent way, no (or minimum) calibration would be necessary.

48 At the end it is here reminded the importance to preliminary choose the appropriate model for the specific event to be simulated. In particular, options of type A and B above discussed are essential for a correct set-up of the analysis.

2.2.3.4 Physical models

Even though much more used for the design of countermeasures than for hazard

mapping, physical modelling can be of some help to this latter purpose. Physical modelling consists of representing phenomena at a small scale in similarity with real phenomena. For debris-flows, this requires to use specific model fluids (water-clay mixtures for instance) whose adequate composition remains an important scientific issue. The similarity criteria are generally based upon the Froude number and other dimensionless numbers chosen according to the assessed rheology of the debris-flow material.

2.2.3.5 Conclusions

Topography and volume of mass movement control runout distance to a considerable

extent. If an analysis similar to the LAHARZ method by Iverson et al. (1988) can be made for debris flows based on detailed data from the runout zone topography, this method can potentially improve the success of empirical-statistical methods for the prediction of the runout length. For the empirical-statistical methods, experience applying one method to different settings is still limited; however, potential exists for applying them in preliminary hazard assessments.

Among the analytical approaches, there are several approches such as the Takahashi- Hungr model, the sliding friction model and the Körner-Perla model. These approches need to be more systematically compared with field observation for a more reliable determination of appropriate values for the friction slope.

The most complete description of debris-flow behavior can be provided by continuum based dynamic simulation models. Only for some model applications were exact values of model parameters known or determined a priori. Typically, appropriate values for the rheologic or friction parameters were assumed or back-estimated from field observations. However, most of these models have not yet undergone rigorous testing with field events, which is necessary before they can be reliably used for predictive purposes. When comparing model simulation results with observations of natural debris flows, some general debris-flow characteristics needed for hazard assessment can be reasonably well simulated with simple dynamic models if prior calibration of the model parameters is possible (Rickenmann et al. 2005; Naef et al., 2006).

2.3 Hazard maps types

Introduction Hazard mapping of areas directly exposed to torrential floods requires: (a)some knowledge of points where overflow (local to generalised) outside the torrent channel is likely to occur; (b) analysis of the conditions of overflow by floods with large solid

49 discharge.

During flooding, the overflow can result as a consequence of: (a) reduction of the channel cross-section by previous material deposition; (b) channel cross-section too small compared to a large event, as for example a debris flow of a given probability (e.g. "reference" debris flow); (c) presence of human infrastructures in the channel (bridges, etc.); (d) failure of some inadequate protection structure; (e) formation of temporary debris dams plugging the channel; (f) strong erosion of lateral banks of the channel, likely to generate a strong deviation of the flow.

For all scenarios (inclduing e.g. "reference" debris flow) subsequent spreading of debris flows is analysed on the basis of identified overflow points and taking into account possible disturbing factors likely to generate more critical situations.

2.3.1 Hazard index maps (overview scale)

Overview hazard maps are used in regional planning (Petrascheck & Kienholz, 2003).

Regional planning, as done for example in Switzerland, establishes rules for the planning on a municipality level. Therefore only the regions where natural dangers must be considered are included. This requires identification of the types of hazards and the potentially endangered areas where limits may be set generously, since the later detailed studies for local planning seldom exceeds the study area as indicated in the overview map. This map should cover the whole territory. The accuracy is limited by the available financial and time resources. The stepwise approach limits the detailed studies only for the areas where it is necessary.

Hazard index maps are established at a general scale (e.g. 1:50'00 or 1:25'000) to identify potentially endangered areas and those not endangered. At this level, events with a relatively long return period (e.g. 300 years) may be considered to arrive at a “generous” delineation of endangered areas. These areas basically depend on the runout distance and the lateral deposition pattern of debris flows on the fan. Empirical methods or simple models may be used to determine the potentially endangered areas.

As an example for hazard index mapping of debris flows, a GIS based simulation model has been developed which is described in Zimmermann et al. (1997a) and Gamma (2000). The basic part is a mass point model which allows to estimate total travel distance of debris flows. This model incorporates a Voellmy fluid flow resistance approach, and two parameters need to be calibrated (Rickenmann, 1990). An empirical relation is used to determine the starting locations (using a critical slope angle) of potential debris-flow events. To estimate potentially affected areas on the fan, a stochastic simulation of the lateral flow angle on the fan is introduced (requiring prior calibration). Relating the number of simulation runs to the event magnitude results in larger affected areas for bigger debris-flow events. Figure (2-13) shows and an example illustrating the calibration of the lateral spreading on the fan, and Figure (2- 14) gives an example of a hazard index map produced with this model.

50 Figure 2-13:

(a) Calculated deposition patterns for Val Varuna debris-flow cone with simulation model of Zimmermann et al. (1997a) [source: Gamma (2000)]

(b) Observed deposition of debris flow in 1987 on the fan of Val Varuna [photo Swiss Federal Office of Torpography of 21.9.1987, archive KSL, flight line 080003, image no. 9871]

Figure 2-14:

Example of simulation with model of Zimmermann et al. (1997a) used for hazard index mapping; debris-flow paths are marked in red [source: M. Zimmermann, Geo7]

2.3.2 Hazard maps (detailed scale)

Detailed hazard maps are the basis for spatial planning on a municipality level

(Petrascheck & Kienholz, 2003). If hazard mapping procedures have been established, ill f l h ldb ll di hihh d( d) d

51 typically no future settlements should be allowed in high hazard (red) zones, and building within the moderate (blue) hazard zones should be avoided as much as possible. If existing built up areas are classified as high or moderate hazard zones, technical protection measures must be studied to reduce the risk. But risk mitigation is not only done by spatial planning and protection works.

Figure 2-15:

Example of hazard map at detailed scale (scale of the single path). Case study: Sauris (IT) debris flow. Dynamical model: TRENT- 2D. Mapping criteria: (a) Rickenmann (2005b) criteria; (b) PAT (Trento Province) criteria.

The hazard map is also an important basis to:

• establish construction rules for zones with moderate or low hazard degrees

• provide information for the land owner on the hazard situation and associated construction standards

• prepare emergency measures for the whole area including those with residual risk.

• discuss and implement measures aiming at hazard and risk reduction

It is obvious that the information provided by the hazard maps is not sufficient for detailed planning of all above measures. But it highlights an area saying “in case you want to do something you have to consider this hazard – either obligatory or deliberately for your own safety”. The necessary detailed information should be found in technical reports accompanying the hazard map. This background information is compulsory for each map, since it must be documented why a certain area is classified in a specific way.

2.3.3 Overview on methods for hazard mapping

Table (233-1) lists a number of methods which may be used for debris-flow hazard

assessment. This list also gives information on required input parameters, on some limitations of methods, and on the the suitability of the methods for producing hazard index maps at a less detailed scale or proper hazard maps at a more detailed scale.

Ideally, the elaboration of hazard maps may include the interpretation of past events, air photographs and thematic maps, field investigations, use of empirical methods, and

52 possibly the application of numerical simulation models.

Method Basis, Input Restrictions, Remarks Table 2-7:

Event magnitude Methods used for the assessment of debris flow empirical relationships catchment morphology insufficient prediction accuracy parameters (Rickenmann, 2001, 2001b, 2003). geology/lithology further investigations desirable erodability (comb. with GIS) used for producing hazard index maps sediment yield past observations, geomorpho- time consuming, involves subjective assessment logic assessment (input from judgement tributaries and slope failures)

Frequency of events

past events dates of past events (with good indicator also for deposition estimate of magnitude) behaviour threshold rainfall frequency of threshold rainfall rough assumption, no or limited conditions conditions correlation with event magnitude dating methods dendrochronology, linchenometry used in research, often not feasible for practical application

Initiation mechanisms

slope instability analysis rainfall, soil + hydrogeological soil mechanics, transition to debris conditions flows; often heterogeneous conditions; lacking field observations

threshold discharge in peak runoff, bedload initiation analogy to sediment transport channel other triggering e.g. glacier lake outburst, snow melt mechanisms

Flow propagation

flow resistance laws event magnitude, discharge, flow empirical, resistance coefficients (uniform flow) depth, channel slope depend on material composition and water content "hydraulic" simulation with constitutive equations for different limited verification with field events so numerical models "fluids", channel geometry, input far; criteria for applicability of flow types hydrograph still lacking

Deposition on fan (runout distance)

Travel angle longitudinal profile, catchment very rough estimate, (general slope) area problems with convex profiles empirical relationship Event magnitude, elevation very rough estimate difference (relief height) GIS based models topography, combined with mostly precalibration of parameters empirical runout prediction models "hydraulic" simulation with constitutive equations for different limited verification with field events so numerical models "fluids", fan geometry, input far; criteria for applicability of different hydrograph models not well known

2.3.4 Scenarios and uncertainty

As discussed elsewhere, there are often major uncertainties both in establishing an

appropriate magnitude-frequency relationship and in the assessment of the flow and deposition behaviour. In addition, there may be a strong interaction with other processes or elements such as bedload transport and woody debris, further complicating the assessment. It can therefore be necessary to work with scenarios. For example, the presence of woody debris may lead to a blockage of the flow in the torrent channel underneath a bridge, resulting in overtopping of the flow on the fan. On the contrary, without woody debris the channel conveyance capacity may be sufficient to carry the estimated debris-flow discharge for the same event. Such scenarios also

53 have to be associated with a probability of occurrence.

It is important to take into account this uncertainty related to different scenarios for some of which it can be very difficult to assign a probability. For example, in France it is now proposed that uncertainty be considered very soon in the hazard mapping procedure by the responsible experts. It is recommended that several scenarios be considered (if pertinent) when defining the reference flood.

Taking into account these points in practice essentially requires some expertise from people in charge of hazard mapping. This expertise has to use as much as possible the hazard mapping tools presented here-before. However, human reasonning still has a paramount importance in hazard mapping procedures. For this reason it is important to include information about the assumptions regarding different scenarios which form the basis of a precise delineation of hazard zone, in order to make the hazard mapping process transparent, documenting also somewhat "subjective" decisions which are based on the experience of the expert(s) responsible for the haazrd mapping. This is strongly recommended for example both in the Swiss hazard mapping procedure and in France related to the preparation of the PPR maps.

If several hazardous processes can occur in the same area, the final hazard map will result from overlaying the danger levels of all processes, and the most unfavourable danger level will be retained on the map. However, it is important that also a technical report is produced which describes all steps and methods used to construct the hazard map. This documentation helps to understand the assumptions, arguments and calculations which form the basis of the hazard assessment. Furthermore, it is also important to clearly show uncertainties of the assessment. Both the establishment of magnitude-frequency relations and the assessment of the flow behaviour are associated with uncertainties. Therefore, the resulting hazard zones should ideally be related to probability distributions which reflect for example the uncertainties of flow magnitude and of the occurrence of scenarios.

Chapter 3

ROCK AVALANCHES

3.1 Hazard Mapping Criteria

Introduction For a long time, rock avalanches were perceived as large-scale manifestations of catastrophic hillslope adjustment in response to debuttressing of alpine valley walls following deglaciation. With removal of large valley glaciers, oversteepened rock slopes in U-shaped valley troughs gradually become unable to support the new stress regime, and thus give rise to large-scale rock-slope failure. Such processes can be observed in the vincitiy of downwasting glaciers in many mountain belts throughout the world (e.g. Kääb, 2002; Holm et al., 2004). With such exceptionally large events occurring only once per many years on average in present times, the view developed that most rock avalanches had occurred during the stages subsequent to the Last Glacial Maximum, and Early Holocene.

Rock avalanches past Thanks to detailed research on historic occurrences (Heim, 1932; Abele, 1974; and present Eisbacher and Clague, 1984) and recent advances in Quaternary geochronology there is now, however, ample evidence that rock avalanches have continued to occur throughout the Holocene (von Poschinger and Haas, 1997; Kubik et al., 1998; Ballantyne and Stone, 2004). What is more, historic accounts clearly demonstrate that rock avalanches also have incurred a high number of fatalities and damage in many mountain belts even in historic times (Heim, 1932; Abele, 1974; Raetzo et al., 2002; Segalini and Giani, 2004; Blikra et al., 2005).

In tectonically active mountain belts and fault-bounded mountain fronts especially, rock avalanches may occur rather frequently, i.e. more than once in a decade for a given region (e.g. McSaveney, 2002), mainly because seismic activity together with high rates of erosion promote causes and triggers of such catastrophic slope failures. The same applies to volcanic arcs, where large debris avalanches together with sector

55 collapses of volcanic edifices are important elements of erosion and maintenance of slope angles (Ponomareva et al., 2006). Large catastrophic rockslides and rock avalanches are also common in mountainous regions on passive continental margins, where rapid surface uplift due to postglacial isostatic rebound of the order of 101 mm/yr causes stress release and rock-slope dilatation (Sandersen et al. 1996). The historic occurrence of such failures together with the fact that their deposits are readily discernible on sonar images of fjord bottoms indicates their epsiodic recurrence during postglacial times.

The largest rock and debris avalanches have rapidly mobilise rock-mass volumes of up to 109 m3, and five events exceeding this size have occurred in the 20th century alone (Korup and Glade, in prep.). One of the latest such giant events is the earthquake- triggered Usoi rockslide-rock avalanche (2.2 × 109 m3) in the Tajik Pamirs, which formed a 600-m high natural dam across the Murgab River in 1911 at the former site of the Usoi village. The impounded Lake Sarez is now >60 km long, and contains some 17 km3 of water (Schuster and Alford, 2004). Similarly, another earthquake-triggered rock avalanche (1.2 × 108 m3), Northern Tajikistan, obliterated the town of Khait, allegedly causing thousands of fatalities in 1949 (Leonov, 1960).

Recent examples One of the most recent rock avalanches occurred at Hattian Pir near Dandbeh, Pakistan, killing some 1000 people, in the wake of the 2005 M 7.2 Kashmir earthquake (Harp and Crone, 2006). The destructive 2002 Karmadon ice-rock avalanche in the Caucasus demonstrated the potential of extreme travel distance (i.e. ~19 km) due to exceptionally high ice content (Huggel et al., 2005). In contrast, the 2006 Guinsaugon rockslide- debris avalanche, Philippines, which killed more than 1000 people (Lagmay et al. 2006), highlighted the potential for such extremely rapid mass movements in the vicinity of active faults and in tropical environments alike. In the European Alps, the 2003 Thurwieser rock avalanche, Italy, resulted from collapse of a rock pillar, and mobilised some 2 × 106 m3, but fortunately ran out in uninhabited alpine terrain.

Figure 3-1:

Aerial view of the rainfall- triggered 2006 Guinsaugon rockslide-debris avalanche, Philippines, that killed more than 1000 people (Evans et al., 2006).

Practical hazard and Although rock avalanches continue to cause substantial damage and claim lives in

56 risk management many mountain regions throughout the world, in European countries very few strategies exist in the way of detecting, quantifying, or managing any hazards or risks from this process. Most practical approaches have so far implictly or explicitly relied on the notion that rock avalanches are very rare events in comparison to other extremely rapid mass movements such as debris flows or snow avalanches. Hence, their management has effectively remained at levels of residual or acceptable risk. This is also partly based on that reasoning that the high physical destruction potential of large rock avalanches renders the range of structural countermeasures to modulate of reduce physical impact forces largely inefficient at most scales. However, in countries where rock avalanches occur more frequently, e.g. Canada, New Zealand, Japan, Taiwan or Papua New Guinea, first steps toward a more explicit treatment in hazard and risk management have been taken (see Appendix). The assessment of hazards and risks from volcanic debris avalanches has also earned substantial interest in Alaska and the Pacific Northwest of the USA, and Japan.

Nevertheless, little attention has been given as to attempt any quantification of the hazard and risk from rock avalanches, and most notions in this regard remain qualitative. There are some notable exceptions (e.g. Whitehouse and Griffiths, 1983), and a growing number of regional studies and inventories on large landslides including rock avalanches helps constraining their past probabilities of occurrence as a preliminary means towards a quantitative hazard assessments.

The following sections intend to give an overview on the currently used methods and limitations in quantifying the hazard from rock avalanches.

3.2 Hazard Mapping Tools

3.2.1 Identification of rock avalanches hazard

3.2.1.1 Rock-avalanche morphology and sedimentology

From detailed studies and observational evidence it is generally accepted that rrock avalanches have on average a lower recurrence than debris flows or snow avalanches within a given area. Therefore, any quantitative estimate of rock-avalanche frequency per unit time and unit area (= hazard) is dependent on the completeness of all available historic records. Moreover, this effectively requires the timescale of interest to be extended to prehistoric times, and relies heavily on the correct detection and interpretation of former geomorphic evidence of rock avalanches. This in turn is a function of the preservation potential of rock-avalanche morphology in a given landscape, and largely determined by, among others, post-emplacement erosion potential, catchment position, and sedimentology of the debris deposit. A common problem with geological evidence is, for instance, that younger and larger events may have obliterated all evidence of former events. Thus, inadvertent undersampling of past rock-avalanche occurrences will lead to underestimates of their probability of occurrence, and hence, levels of hazard.

57 Figure 3-2:

Hummocky surface of volcanic debris-avalanche deposit along north side of Chakachatna River, Alaska. View is toward south. Photograph by C.F. Waythomas, July 1998 (Waythomas and Nye, 2002).

Morphology The morphology of rock-avalanche detachment areas is perhaps the least conspicuous and highly variable feature, and there seem to be no unique features associated. Although in many cases amphitheatre-, bowl-, or even cirque-shaped (e.g. Smith et al., 2006), planar detachment areas are equally possible. Frequently such detachment areas resemble glacial cirques (Turnbull and Davies, 2006), while post-failure rock fall processes commonly have built extensive scree slopes, thus further restricting access to the detachment plane and masking any evidence of prior large-scale instability. It is unclear whether rock-avalanche detachment areas would differ significantly from those of other large-scale rock-slope failures in a morphological way.

Rock-avalanche deposits are typically elongate and lobe- or tongue-shaped debris sheets indicative of rapid flow movement. Their surface commonly features sets of longitudinal and/or transverse furrows, ridges, and swales, whereas their distal margins are sharply defined and often lip- or finger-shaped. Distal portions may run up (swash) against obstacles, opposite hillslopes, or even overtop small interfluves (McSaveney, 2002). Elongate depressions between such swash zones and the underlying hillslopes were termed “Brandungstälchen” (swash valleys) by Heim (1932). Surface depressions may contain perched ponds, especially if reworking of debris has helped to seal the deposit surface.

58 Figure 3-3:

Oblique air photo draped on DEM of the Cheam 5000- year old rock avalanche (175 × 106 m3), Fraser Valley, British Columbia, Canada. Conspicuous hummocky terrain together with a number of geological boreholes led to the identification of this major event (Orwin et al., 2004).

Sedimentology There have been several recent attempts to characterise and quantify rock-avalanche sedimentology both in terms of individual grain size-distributions as well as internal structure of whole deposits. The key morphological and depositional features of rock avalanches are summarised by Hewitt (1999), who also pointed out several apparent similarities with glacigenic sediments. One distinctive feature of all rock-avalanche deposits is the angular to very angular shape of clasts, which in an ideal stratigraphic column are inversely graded, and topped by a bouldery carapace. Below this surface armour, the particle size distribution of rock-avalanche debris follows a fractal distribution. Although morphologically intact, larger clasts show distinctive fracturing, giving rise to a characteristic jigsaw structure, in which angular particles are interlocked, creating little visible void space.

Stewart et al. (2003) have underlined the importance of differentiating between hot and cold rock avalanches in volcanic terrains, i.e. those related to a volcanic eruption and non-volcanic slope instability, respectively. They argued that the differing runout characteristics of these extremely rapid mass movement, incorrect interpretation of sediments could lead to significant underestimates in hazard apraisals.

59 Table 3-1:

Key characteristics of hot and cold avalanche deposits in volcanic terrain (Stewart et al., 2003).

A rare occurrence at the base of large rockslide/rock avalanche deposits is “frictionite”, which has also been termed hyalomylonite (Legros et al., 2000; Weidinger and Korup, in press). This glassy or pumice-like material resembles fault gouge and attests to high frictional heat at the sliding base, which has contributed to partial melting of minerals. In some cases, the correct identification of this material together with other characteristics of rock-avalanche deposits may be the only evidence left in areas og high erosion, where other morphological evidence is rapdily obliterated (Weidinger and Korup, in press).

Finally, The sedimentology of river-blocking rock avalanches is of utmost importance, as it largely controls the stability of the dam against seepage erosion, and lateral undercutting, or other types of catastrophic failure. Based on a range of case examples, Weidinger et al. (2002) have begun to explore first-order relationships between clast size and dam stability, which remain to be tested by numerical stability models.

Figure 3-4:

Boulder carapace of rock avalanche that ran out on West Fork Glacier, Alaska, during the M 7.9 Denali earthquake of 3 November 2002 (Jibson et al., 2006). Note person for scale.

3.2.1.2 Causes of rock avalanches

A growing number of case studies and inventories of rock avalanches throughout the world has helped to empirically constrain their underlying causes and triggers with respect to boundary conditions such as tectonics, climate, and rock type (e.g. Heim, 1932; Abele, 1974; Whitehouse, 1983; Eisbacher and Clague, 1984; Hewitt, 1998; Hermanns and Strecker, 1999; Legros, 2002; Geertsema et al., 2006; Ponomareva et

60 al., 2006; Korup et al., in prep.).

Causes Typical causes of rock avalanching are not significantly different from those leading to other types of large-scale rock-slope failure. They include

• gravitational and topographically enhanced stress along (but not limited to) major discontinuities;

• gradual reduction of rock-mass strength through repeated earthquake shaking;

• gradual weakening of rock-mass strength in fault-zones by hydrothermal alteration and hanging-wall shattering;

• slope dilatation and/or loading following precursory landsliding; and

• incision-driven loss of internal cohesion or lateral support.

The latter preparatory factor also includes persistent and gradual effects of weathering, fluvial or artificial slope undercutting, and also melting of alpine permafrost ice in jointed rock slopes (Bottino et al., 2002). Given the depths of failure surfaces of most rock avalanches, the role of vegetation and root strength is in most cases negligible.

These conditions are typically provided in tectonically active orogens, fault-bounded escarpments, and volcanic arcs. Notably, high relief or slope steepness are not necessary preconditions for large rock avalanches, as several occurrences have also been documented in moderate mountain topography. For instance, one of the world’s largest rockslide/rock avalanches at Seidmarreh, Iran (2.4 × 1010 m3) has failed in rather moderate relief. Moreover, rock-mass defects on the hillslope scale and below may play an important role in slope stability, hence complicating approaches to derive a regional pattern of rock avalanche occurrence as a function of large-scale structural geological characteristics (e.g. Hermanns and Strecker, 1999).

3.2.2 Evaluation of of rock avalanches occurrence

3.2.2.1 Trigger mechanisms and possible precursors

Triggers The occurrence of rock avalanches is frequently bound to a trigger mechanism that causes a transient change to the ratio of shear strength versus shear stresses rock slope at the expense of stability.

The most common trigger mechanisms of catastrophic rock avalanches include large earthquakes, high-intensity rainstorms (sometimes in conjunction with excessive snowmelt), and fluvial undercutting. Importantly, several historic rock avalanches occurred without any observed triggers (Eisbacher and Clague, 1984; McSaveney, 2002). In these cases failure is likely to result from gradually exceeded intrinsic stress thresholds, and may represent the catastrophic culmination of preceding slope movements, as several historic source areas of catastrophic rock avalanches were sited on large-scale and slow-moving sackung-type landslides (Radbruch-Hall, 1978;

61 Eisbacher and Clague, 1984).

General predictability Determing the probability of occurrence for rock avalanches in a given area is not a straightforward task. A standard approach is to perform statistical analyses on inventories that are believed to contain a significant and statistically representative sample. An alternative option is to infer the probability of occurrence from that of the triggering mechanisms. Thus, the predictability of earthquake- or rainfall-triggered rock avalanches depends on the knowledge of the spatio-temporal distribution of the trigger events, i.e. their magnitude and frequency, temporal duration, and any underlying critical thresholds that need to be exceeded.

Earthquake-triggered rock avalanches are as difficult to predict as are earthquakes in general, i.e. the chance of successfully forecasting coseismic events is very low. The same applies to those triggered by fluvial undercutting or gradual loss of rock-mass strength beyond a critical stress threshold. Quantification of the dynamics of local geotechnical parameters is scarce, and usually requires a well-defined and cost- intensive monitoring programme, usually reserved for case studies of imminent threat to population or infrastructure.

The potential of predictability increases where rock avalanches are preceded by slow- moving deep-seated slope deformation, although such low-activity phases may extend from decades to millennia, before catastrophic “creep rupture” (Radbruch-Hall, 1978). Although several cases of such catastrophic culmination of gravitational slope deformation (“Sackung”, “Talzuschub”) are known, it remains highly challenging to approximately predict the timing, magnitude, and type of resulting rock-slope failure from preceding slope deformation. Typical warning signs of pending catastrophic failure reported include rapid opening of tension cracks, gradual increases in rock fall activity, or changes to spring discharge. These warning signs however do not indicate whether any pending rock-mass failure will eventually transform into a rock avalanche, or whether the failure will be of a more coherent type.

Precursory processes This highlights a general difficulty of research on and mitigation of rock avalanches, i.e. the exact delimitation of any preceding processes in space and time. Most studies regard large rock-slope failures as discrete short-lived events, but rarely address the potential for rock avalanches to develop as a catastrophic culmination from continuous rock-slope deformation (Eisbacher and Clague, 1984; Chigira et al., 2003). Nevertheless, the careful detection and documentation of any such precursory processes and/or their associated diagnostic landforms are of high relevance to site- specific hazard zonation. Not surprisingly, many case studies were instigated by such phenomena, and allowed timely detection, monitoring, and—in some cases—early warning.

Precursors and Conversely, early warning is rarely feasible without any reliable detection and prediction successful monitoring of suspected failure sites in the first place (Crosta and Agliardi, 2002). As a consequence, any sophisticated geotechnical site investigations augmented by numerical (process) modelling and real-time monitoring (Willenberg et al., 2002; Tarchi et al., 2003) are limited to sites that had either already failed or were about to in

62 short periods of time (Sartori et al., 2003; Segalini and Giani, 2004). It follows that the detection and use of precursors for rock-avalanche hazard assessments means is less applicable to larger regions.

3.2.2.2 Engineering geological and geotechnical approaches

Rock-slope stability Detailed quantitative assessment of rock-slope stability at a given site provides deterministic means to address the probability of failure, but it also demands very costly resources.

One reason for this is that, although standard limit equilibrium (e.g. Factor of Safety) analyses provide a physical basis for (rock-)slope stability, most geotechnical parameters vary considerable along a given slope portion (Wilson et al., 2003), and may not be considered representative for critical slope sections. The shear size of many potential failure-prone investigation areas, which commonly extend over several km2, restricts the number of borings and drillholes for geotechnical investigation. Many innovative geodetic methods such as airborne high-resolution photogrammetry, LIDAR, ground-based InSAR, or differential GPS measurements have helped quantify discontinuities, deformation rates, and hydrogeological conditions of the creep phase of rock slopes at increasingly high levels of detail.

These methods provide detailed insights into the slope dynamics before commencement of catastrophic failure. They have, however, added little in terms of constraining the probability or timing of catastrophic failure. In some cases, the use of the so-called inverse velocity relationship has been shown to be somewhat successful. This empirical approach is based on the observation that the reciprocal velocity plotted against a time axis of several gradually accelerating rock masses has helped predict the time of failure within reasonable constraints (Sornette et al., 2004).

63 Figure 3-5:

High-resolution digital photogrammetry of successive air photos allows computation of displacement vectors on rock slope adjacent to Aletsch Glacier, Swiss Alps. Detailed knowledge of such processes as potential precursors of catastrophic rock-slope failure are essential for site- specific hazard and risk management (Kääb, 2002).

3.2.2.3 Magnitude and frequency

Regional-scale Most hazard and risk assessments pertaining to debris flows and snow avalanches rely approach to rock- on expert or inventory knowledge of channels or slopes that are particularly prone to avalanche hazard these processes. As outlined above, the main difference to rock avalanches is that, very often, rock-slopes prone to catastrophic failure are very difficult to identify. Therefore, a typical first-order and more practical approach to hazard zonation starts at the regional scale rather than at the local scale of individual slopes.

Ideally, this should complement the higher degree of detail, but lower coverage of space, from studies of well-studied and highly instrumented rock slopes. However, such links between regional- and hillslope-scale studies have not been fully exploited as yet, mainly because of the difficulties in up- or downscaling of the relevant parameters. Despite there being a high number of known unstable rock slopes in the European Alps, no generally accepted set of criteria exists to objectively quantify their potential for issuing large rock avalanches.

Magnitude and Given these constraints, a regional magnitude-frequency approach appears to be frequency from adequate as a first-order estimate towards quantifying the probability of rock inventory data avalanching per unit time and/or unit area. Importantly, this approach also fully meets the requirements of the definition of landslide hazard in general (Cruden and Varnes, 1996). The necessity of an inventory as an essential step in hazard and risk assessment is thus also called for when deadling with rock avalanches. Practical scales for such an excercise are 1:50,000 to 1:200,000, depending on the level of detail and minimum volume or area of the rock-avalanche events to be included.

64 Figure 3-6:

Suggested steps in compiling a regional-scale inventory on rock avalanches (Korup, 2005b).

Existing rock-avalanche inventories often list a number of volumes and dates, which may be readily used for such an analysis (e.g. Abele, 1974; Whitehouse, 1983; Hewitt, 1998). Within the field of Quaternary geochronology, there is now a wide range of absolute dating methods (e.g. 14C, 10Be, 3He, dendrochronology, lichenometry) available to constrain the age of a number of deposits with unknown ages to fill in the gaps in the magnitude-frequency distribution.

Several studies have found that, within a specified area, the frequency f(r) of a landslide (expressed e.g. per unit area or per unit area and time) is a function of its magnitude (e.g. measured in terms of volume, area affected, or runout length), and therefore it should be possible to approximate f(r) from a best-fit model (Hungr et al., 1999; Dussauge et al., 2003; Malamud et al., 2004). Hence, only frequency distributions measured in space and time are useful for hazard estimates.

What is more, most of these studies have relied on extensive datasets containing landslides below a certain size threshold, usually << 107 m3 (Dussauge et al., 2003). Compared to the observed upper size limit of rock avalanches, i.e. ~1010 m3, most of these inventories contain small events. Considerably less is known about the recurrence of larger landslides, and it remains largely untested whether the statistical relationships remain valid beyond the size spectrum and observation period they were derived from. Few studies explicitly attempted to quantify the probability of occurrence of larger catastrophic landslides from inventories that span several millennia (Whitehouse and Griffiths, 1983).

One obvious limitation to this approach is the preservation potential of rock-avalanche deposits in actively eroding landscapes. Smaller deposits are theoretically much more susceptible to being eroded or buried by subsequent geomorphic processes. Another drawback of the magnitude-frequency approach is that it does not provide any spatially

65 explicit zonation of rock-avalanche hazard in general.

Figure 3-7:

Cumulative volume distribution for a worldwide dataset of 142 rockfalls derived from cliffs. Note power-law trend of rockfalls with volumes between 107 and 1010 m3 (Dussauge et al., 2003). Note that events in this size domain also contain rock avalanches.

Magnitude and Alternatively, first-order hazard estimates of large landslides use magnitude-frequency frequency of trigger distributions of their triggers as a surrogate (Keefer, 1999). This re-quires sufficient mechanisms knowledge on the underlying intensity thresholds. Many of the large catastrophic landslides documented have been triggered by strong earthquakes and volcanic eruptions, and there are empirical relationships between the earthquake magnitude and that of coseismic landsliding, quantified e.g. in terms of total affected area or volume (Keefer, 1999). The limits of this approach were revealed by several recent earthquakes that triggered significantly fewer landslides than expected (Jibson et al., 2006), whereas conversely many catastrophic landslides have occurred without any trigger observed.

3.2.2.4 On- and off-site hazards

Because of their size range (106–1010 m3), rock avalanches are usually produce kilometre-scale geomorphic effects. What is more, with increasing volume, the probability of significant geomorphic interaction with the drainage network, and thus, the possibility for spatial and temporal translation of geomorphic impact, also increases. This is a crucial point of neglect in many state-of-the-art landslide hazard assessments is the potential for any adverse off-site or long-term effects. What is more, current reviews do not address this important issue (Aleotti and Chowdbury, 1999; Dai et al., 2002).

Sediment discharge Generally, these effects on the drainage network derive from response to instantaneous and channel instability and excessive sediment input. Large rock avalanches in particular deliver substantial amounts of highly fragmented debris to river channels, with local thicknesses of up 1000 m (Hewitt, 1998; Harden, 2001). This overwhelms the transport capacity of even large river systems, and temporary blockage and aggradation ensue. Depending, among others, on emplacement morphology, degree of rock fragmentation, and fluvial erosion potential, short-term peak sediment yields from actively eroding rock avalanches may attain 105 m3 km–2 yr–1 when specified for the upstream contributing catchment area (e.g. Chen et al., 2005). The highest such sediment yields from rock-avalanche deposits

66 are documented from tecontically active mountain ranges with high amounts of precipitation, such as the Himalayas, Taiwan, Papua New Guinea, or the New Zealand Southern Alps (Korup et al., 2006).

Figure 3-8:

Changes to river-channel long profile following catastrophic sediment input from 1999 Mt Adams rock avalanche, Southern Alps, New Zealand (Korup et al., 2004).

River blockage The extreme case of sediment input to river sytems often leads to the formation and failure of rock-avalanche dams. For a given catchment position, the probability for occurrence of such dams increases with the size of the rock avalanche, although narrow bedrock gorges are espeially prone to formation of high blockages. These give rise initially to backwater inundation, wheras sudden failure of the dam may lead to catastrophic outburst flows (Costa and Schuster, 1988; Korup, 2002; Schneider et al., 2004; Dunning et al., 2006; Korup and Tweed, in prep.).

Typical effects following dam failure include the slow downstream conveyance of sediment leading to in-channel and coarse floodplain aggradation (Korup et al., 2004); ensuing river channel instability such as sudden avulsions (Clague et al., 2003; Korup, 2004b); secondary slope instability along aggraded sections of valley floor; and enhanced flood frequency.

Another type of off-site effect is the formation of displacement waves by rock- avalanche impact into water bodies, such as alpine lakes, fjords, but also artificial reservoirs. In the Norwegian fjords, for instance, rockslide-derived displacement waves have accounted for the highest number of historic fatalities from a single natural hazard (Sandersen et al., 1996).

67 Figure 3-9:

Downstream effects of confined river morphology following failure of the 1999 Mt Adams rock-avalanche dam, Southern Alps, New Zealand. A: False colour satellite image showing rock avalanche (ra) and landslide dam (ld). B: Channel changes measured between locations 1 and 2 in (A). (Korup et al., 2006)

Persistence of In populated alpine areas, any impacts related to landslide-sediment reworking and geomorphic effects of transport are also particularly relevant to artificial reservoirs and hydropower schemes rock-avalanche debris (Poisel et al., 2003). Excessive sediment yields from rock-avalanche dams may contribute to accelerated infill of artificial reservoirs. Moreover, large rock-avalanche deposits may act as long-term debris reservoirs for recurring rock falls, shallow debris slides, and debris flows (Liao and Chou, 2003). There are also several reported cases where major portions rock-avalanche deposits may fail catastrophically (Eisbacher and Clague, 1984). Moreover, many of the (pre-)historic rock avalanches in now populated mountain areas have completely re-shaped valley floors over several tens of square kilometres, and their hummocky deposits are persistent obstacles to modern infrastructure and land use.

68 Figure 3-10:

Example of the persisting problems associated with large rock avalanches, State Highway 93, Southern Alps, New Zealand. This important traffic route was for years routed across heavily fragmented rock-avalanche debris that was emplaced ~2000 years ago. Continued rock fall and debris flows from the deposit necessitated the construction of the Southern Alps’ first viaduct. Note that pylons are grounded in >30-m thick rock-avalanche debris (Korup, 2004a).

Range of off-site A good example are the extensive deposits related to the 1.2 × 1010 m3 Flims rockslide- impacts rock avalanche in the Swiss Alps. this giant Early Holocene slope failure had dammed the Rhine river up to a height of ~400 m, creating extensive lake and subsequent outburst deposits, which are traceable as far as Lake Constance (Schneider et al. 2004).

The physical impact range of such off-site effects is dependent on the overall valley geometry, but in the case of catastrophic outburst flows may in exceptional cases attain several hundreds of kilometres or more. Hermanns et al. (2004) describe the Barrancas rock-avalanche dam failure on Rio Colorado, Patagonia, which in 1914 caused a debris flow mobilising 120 × 106 m3 of sediment, which was mainly eroded from the dam. The resulting depositional terraces attained heights of up to 20 m above the channel floor over ~60 km, and led to the damming of several tributaries, although the flow affected ~1250 km as far as the Atlantic coast. Another example that has attracted worldwide attention is the world’s highest dam, formed by the Usoi rockslide that dammed the Murgab River to form 60-km long Lake Sarez, Tajikistan (Schuster and Alford, 2004).

Despite advances in numerical dam-break modelling, the effects of catastrophic outburst floods and debris flows from such lakes cannot be treated in a fully probabilistic manner, and it is standard practice to formulate a range of scenarios (Risley et al., 2006). The consequences of what would happen following a catastrophic drainage of the 1.7 × 1010 m3 Lake Sarez was exemplified, albeit at a smaller scale, by the failure of the 2001 Zhamulongba rock/ice avalanche dam, Tibet (Shang et al.,

69 2003). After breaching of the dam, a huge flood wave caused widespread destruction and fatalities in Indian provinces downstream. Further and more persistent consequences of landslide-dam failures include pulses and delays superimposed on the long-term sediment transport in formerly blocked rivers.

These and many other case histories have demonstrated the potential of large rock avalanches to trigger a sequence of cascading processes. Such potentially adverse chains of events have so far been poorly recognised in hazard and risk assessments in general. Their importance lies in that even catastrophic rock avalanches that occur in unihabited parts of mountain belts may trigger hazardous consequences for communities up- or downstream.

Figure 3-11:

Possible chains of events relating to the formation and failure of a rock-avalanche dam (Korup, 2005a).

Figure 3-12:

Sketch highlighting several geomorphic hazards expressed as dependent probabilities resulting from the formation and failure of a rock-avalanche dam in a mountain-foreland system. P(D|O), for instance, denotes the conditional probability of spatial impact by a dam- break flood, given the failure of the dam.

70 3.2.3 Dynamical Behaviour and Runout Distance Prediction of rock avalanches

3.2.3.1 Modelling approaches

Why model? The modelling of large catastrophic landslides in general, and long-runout rock avalanches in particular, is of substantial interest for several reasons. First, the encapsulation of what appears to be a highly energetic process of relief destruction is of intrinsic interest to geosciences in terms of its mass-transport capability and resulting landform changes, particularly disciplines such as process geomorphology, Quaternary geo(morpho)logy, or long-term landscape evolution. The dynamics of rock avalanching itself, however, poses many interesting questions and challenges to analytical and fluid mechanics, the physics of Coulomb friction and granular mass flows, as well as conservation of mass and energy.

From an applied perspective, modelling of rock-avalanche dynamics and their runout is of utmost interest to natural hazard and risk management. In the case of debris flows and snow avalanches, dynamic models are needed to spatially delineate the area of direct physical impact from a spreading mass movement as a function of well-defined initial and boundary conditions, such as, among others, the initial volume the surrounding valley geometry. These simulated runout dimensions are then coupled with estimates of the average recurrence interval for a defined event volume. This practice is more difficult for rock avalanches, mainly because of the problems of reliably modelling the excess runout, which may lie between 102 and 103 m. In many mountain valleys, underestimating travel distances at these scales may be crucial, especially as direct physical impact by rock avalanches can be assumed to cause full destruction.

Figure 3-13:

The relationship between the ratio of rock-avalanche height over length (H/L) and volume is an often cited measure of the mobility of such extremely rapid mass movements (Collins and Melosh, 2003).

Types of models Three main types of models have governed research on dynamics and runout of rock avalanches:

• empirical models;

71 • analytical models;

• numerical models.

Though each of these model types have their advantages and disadvantages, they have served to promote our knowledge on the phenomenon of rock avalanches and related granular mass flows and rapid mass movements, in general. The following overview gives a brief summary of the input requirements, underlying assumptions, applicability, and limitations of various models used to explain the physics of rock avalanches.

3.2.3.2 Data input requirements

Initial topography The success of any modelling effort in terms of reliably predicting rock-avalanche runout is strongly dependent on the quality and accuracy of input data for initial and boundary conditions. Generally speaking, for any model, the quality of the output is determined by the quality of its input.

For large rock avalanches, this poses an interesting and far from trivial problem, since the initial conditions for many events remain poorly defined or mostly unknown. Only in a number of historic occurrences is it for instance possible to reconstruct the initial, i.e. pre-failure, topography from photography, topographic maps, or ideally, even digital topographic data. Geotechnical back analyses only address the limit force equilibria prior to failure, while only indirectly indicating any requirements of the pre- failure topography. Likewise, volume estimates of rock-avalanche deposits may differ quite substantially from those of source areas, even when they are seemingly obvious and well-defined by topographic break lines, such as amphitheatre-shaped detachment areas (e.g. Abele, 1974). Likewise, the initial internal rock-mass structure of the original source area, which may have exerted considerable control on the shape and size of the detaching rock mass, is usually intractable to reconstruct.

The problem of initial conditions is much easier dealt with in terms of the causes and trigger mechanisms of rock avalanching, since most models are designed to commence at the onset of the failure process without prior knowledge on the underlying process history. For event-based triggers such as earthquakes or rainstorms, this instance may be applicable, although there remains in some cases the problem of potential initial motion prior to catastrophic failure (see below).

Runout behaviour and The numerical prediction of the runout of catastrophic rock avalanches is one of the valley geometry greatest problems posed to hazard assessement. Excess runout diatances may range between 101 and 103 m. The topography of the failure site and its immediate surroundings are important boundary conditions for modelling the runout characteristics of rock avalanches. Nicoletti and Sorriso-Valvo (1991) demonstrated that the runout behaviour of rock avalanches may be a function of major landform elements in the runout zone. However, several studies have also shown that large rock avalanches may obliterate minor valley-floor landforms, and, in exceptional cases, contain phases of pure falling motion without significant bedrock contact (Heim, 1932), or even override minor interfluves (e.g. Hewitt, 1998; McSaveney, 2002),

72 representing extreme cases of the typical phenomenon of swash or run-up on opposite valley flanks.

Typically, sudden impact of the moving rock-avalanche mass against major obstacles such as the juxtaposed valley flank may also cause substantial diversion of the bulk rock mass of directions normal to that of initial movement, usually along the major valley axis, causing a characteristic L-shaped deposit planform to develop. Hence, it is critical to capture the three-dimensional valley topography below the detachment area on a 101-km scale as accurately as possible. This should also include any water bodies in the runout path, since the fragmentation mechanics and runout behaviour of rock avalanches will significantly differ when in contact with underlying water bodies. The corresponding and newly emergent field of research on landslide tsunami (Pedersen et al., 2002) deals predominantly with the dimensions and propagation of displacement waves as the main issue of concern (e.g. Lynett and Liu, 2005; Panizzo et al., 2005), whereas the actual runout underwater is only of secondary concern.

Flow conditions Other critical parameters to calibrate models of rock-avalanche dynamics and runout include estimates of flow conditions, which usually address a mean or maximum velocity of the moving mass. Similarly to debris flows, several rock avalanches have been documented to show effects of superelevation, especially where they have been subject to channelization along major valley axes. This superelevation effect can be exploited to infer mean flow velocities (see section on debris flows). Likewise, the runup or swash height of rock avalanche debris over obstacles or opposing hillslopes may be used to infer the necessary velocity of motion. Clavero et al. (2002) used impact marks of projectiles from volcanic debris avalanches on exposed rock surfaces to infer the flow velocity required.

3.2.3.3 Empirical approaches

Morphometry of rock A range of studies has conducted empirical analyses of various morphometric avalanches properties of rock avalanches. Amongst the most popular approaches are bivariate plots of drop height H or runout L, versus volume V, which highlight the excess travel distance of large rock avalanches (e.g. Nicoletti and Sorriso-Valvo, 1991; Corominas, 1996; Legros, 2002). The ratio of H/L has often been used as a key measure of characterising the runout of rock avalanches, and has been earlier (and erroneously) interpreted as the internal angle of friction of the moving rock mass. Nicoletti and Sorriso-Valvo (1991) further proposed an excess runout length, which quantifies the additional travel distance of rock avalanches in excess of those of rock-slope failures having a predefined coefficient of friction.

Other useful morphometric measures include the maximum swash height, i.e. the vertical extent of run-up of rock-avalanche debris on valley flanks opposite of the detachment slope, and any preserved super-elevation trimlines, where the moving rock mass has taken sharp bends predefined by valley topography. Both swash and superelevation help to estimate the maximum velocity of the rock avalanche during motion.

However, most of these scatter plots show considerable scatter even on log-log plots. This is due to a number of uncertainties regarding the exact determination of size

73 parameters for rock avalanches, especially where deposits have been partly eroded or buried by subsequent geomorphic processes, but possibly also due to other poorly constrained controls such as lithology, climate, or valley geometry. Importantly, this scatter also significantly reduces the prediction potential of such empirical relationships.

Nonetheless, the use of morphometric parameters, such as fahrböschung and shadow angles remains a popular tool for estimating the approximate runout from rock-slope failures of a magnitude well below that of rock avalanches (< 106 m3).

Morphometry of Although not developed exclusively for rock avalanches, several studies have landslide dams attempted to characterise empirically form and process relationships for naturally- formed landslide dams. These are first-order estimates methods, which give a good overview on the types of key parameters and their most likely ranges of value (Korup, 2002). Several studies have shown functional links between morphometric characteristics of landslide dams, such as their height, volume, or contributing catchment area, to their projected life span (Ermini and Casagli, 2003; Korup, 2004a). A key problem in this regard, however, is that rock-avalanche dams are dynamic features, and that the dimensions of both dams and lakes are amenable to change at rates dictated by geomorphic processes in the studied area. In other words, a dam that appears to be stable at the time of observation may fail shortly after: there are several such accounts where rock-avalanche dams had remained stable for hundreds, if not thousands, of years, and then suddenly failed catastrophically (Eisbacher and Clague, 1984; Costa and Schuster, 1991; Hermanns et al., 2004). Since such methods are fully empirical and based on data from various regions within differing environmental boundary conditions, they cannot be used as adequate substitutes for detailed and site- specific analyses of dam stability.

Figure 3-14:

Types of landslides responsible for damming rivers from a sample of 353 worldwide examples (Ermini and Casagli, 2003).

Catastrophic outburst Among the most severe hazards from unstable rock-avalanche dams is the downstream of rock avalanche- impact of outburst floods, which very often transform into debris flows by entraining dammed lakes channel sediments along their flow path (Schuster, 2000). There are several empirical

equations that relate observed peak discharge Qp to effective dam height, and volume or potential energy of water released during the event (e.g. Costa and Schuster, 1991). Notably, some of the largest floods in documented history have been caused by sudden 5 failure of rockslide or rock-avalanche dams, with estimated Qp up to the order of 10 m3 s–1.

74 Walder and O’Connor (1997) proposed a more physical-based means to determine peak discharge from earthen dams, while noting that there were still many uncertainties regarding the actual physical processes of dam failure. They also noted that the internal material properties of the dam are of great importance, especially for determining an adequate value for the breach rate, which plays a leading role in predicting peak discharge (Wassmer et al., 2004).

3.2.3.4 Analytical models

Iverson (2003) recently summarised the basic approaches and shortcomings to the modelling of runout of large landslides. Many of his findings also apply to rock avalanches. On the process side, however, a considerable range of sometimes diverging theories have addressed the unusual runout of rock avalanches with regard to conventional frictional theory.

Coulomb friction – One of the earliest approaches to model the runout of rock avalanches was undertaken applications and by Heim (1932), who used a Coulomb slide-block model, based on Newton’s second limitations law:

dv ρh = ρghsinθ − ρghcosθ tanφ (3.1) dt

where ρ is density of a rigid rock mass; h is its (uniform) height; v is velocity; t is time; g is acceleration due to gravity; θ is hillslope angle; and φ is the Coulomb friction angle, i.e. the ratio of shear to normal forces at the sliding plane. This important rock property is readily obtainable from laboratory experiments (30° < φ < 40°), and has therefore greatly contributed to the popularity of this approach. Integration of equation (3.1) over travel distance x, while implicitly assuming a point mass with zero volume, yields the well-known morphometric characterisation (see § 3.2.3.3):

H = tanφ (3.2) L

where H is the maximum vertical drop height of the rock avalanche, i.e. the vertical elevation difference between its crown scarp and deposit toe; and L is the maximum runout length. Scheidegger (1973) interpreted the ratio in equation (3.2) as the coefficient of internal friction of a given rock avalanche.

The most important limitation to this simple model comes from empirical observations of rock-avalanche runout L, which beyond a somewhat diffusely defined “threshold” volume V ~106 m3, is higher than predicted by the range of typical values of φ in equation (3.2). In other words, larger rockslide and rock avalanches do travel for longer distances than one would expect from conventional frictional theory. This is the reason, why simple sand-pile laboratory experiments do not represent the mechanics of large rock avalanches in the field (Davies and McSaveney, 1999). A measure of this “excess runout” was proposed by Hsü (1975):

75 H L = L − (3.3) e tan32°

assuming that φ = 32° would represent a typical angle of friction. Le however scales somewhat systematically with V over several orders of magnitude, and hence remains unaccounted for when using both point-mass assumptions and a Coulomb friction term as in equation (3.1).

Dade and Huppert (1998) proposed a formulation based on the conservation of energy:

gMH A = λ1/3 ( )2/3 (3.4) τ

where A is the total area covered by the rock-avalanche deposit; λ is the mean width- length ratio of the rock-avalanche deposit; M is deposit mass; and τ is resisting shear stress. This states a relationship between the total area affected (i.e. the range of spatial impact in hazard assessment) and the potential energy of the rock mass prior failure raised to the power of two-thirds. In many cases, however, both M and H as well as λ in particular are unknown, hence limiting the use of this formula for predicting hazard zones.

Figure 3-15:

Relationship between total area A affected by rock avalanches and the potential energy of the mass prior to failure (Dade and Huppert, 1998).

The case for low basal As Iverson (2003) notes, there have been several modifications of equation (3.1) to friction include e.g. changes in mass resulting from entrainment or deposition during runout (Hungr and Evans, 2004). However, in some way or the other they neglect to satisfyingly address basic physical constraints, such as conservation of mass or momentum. Conservation of volume, which is employed for studying extraterrestrial long-runout rock-slope failures (e.g. Barnouin-Jha et al., 2005):

∂h ∂ = (hu) = 0 (3.5) ∂t ∂x

where u is mean rock-avalanche velocity; is a premise not directly applicable to rock avalanches (or other large landslides, for that matter), given the importance of dilatance of the moving mass, and any processes of entrainment or deposition during motion (Hungr and Evans, 2004).

76 Nevertheless, the observed excess runout of rock avalanches has invoked various a number of physically-based theories that advocate modes of reducing internal and/or basal friction in a way that would allow for the extreme travel distances observed.

Lubrication due to gas Early avenues of explanation postulated that trapped air cushions below the moving and fluid pressures rock-avalanche mass or entrainment of groundwater-saturated valley fills would produce high enough fluid pressures to sufficiently reduce basal friction during motion (Abele, 1997). The generally higher pore-water water and clay-mineral content in “cold” volcanic debris avalanches or mixed ice/rock avalanches indeed appears to produce much higher runout (>10 km; Capra and Macías, 2002; Shang et al., 2003; Huggel et al., 2005), and lower values of mean deposit thickness, than for non-volcanic or dry rock avalanches.

The detection of giant rock avalanches on extraterrestrial bodies such as the Moon or Mars (e.g. Barnouin-Jha et al., 2005; Bulmer and Zimmerman, 2005), however, casts serious doubts on the need of additional lubricant media to produce excess runout (although there are circular arguments that spell out the necessity of extraterrestrial liquid water for producing such phenomena). Moreover, Brodsky et al. (2003) argued on the basis of teleseismic data of three large volcanic debris avalanches that volcanic gas would not affect the apparent friction during motion.

Acoustic fluidisation Another attempt to explain Le is that of acoustic fluidisation, during which “transient, high-frequency pressure fluctuations, generated during the initial collapse and subsequent flow of a mass of rock debris, may locally relieve overburden stresses in the rock mass and thus reduce the frictional resistance to slip between fragments.“ (Collins and Melosh, 2003).

Dynamic fragmentation A recently developed theory introduces the concept of dynamic fragmentation associated with high dispersive pressures due to close grain-to-grain interactions, which lead to break up of clasts along failure surfaces (Davies et al., 1999). This

fragmentation-driven dispersive force Pf can be expressed in one-dimensional terms as

∂d P = −c (3.6) f ∂x

where c is a dimensional constant, d is the depth of the moving rock avalanche at longitudinal position x (Davies and McSaveney, 2002). Importantly this theoretical approach does not require the initial rock mass to be jointed in any way at initial

detachment time t = t0, nor does it require the moving rock mass to contain and/or entrain any potential “lubricants” en route. Moreover, it explicitly accounts for the often observed comminution of rock-avalanche deposits with a grain size distributions that have a fractal dimension similar to that recently produced by a 3D granular shear model of a lattice solid (Abe and Mair, 2005). Hence, it is one of the few physical- based theories that addresses the important effect of variable grain size distributions during motion.

77 3.2.3.5 Simulation models

Rock-avalanche There exists a number of numerical models that were developed to simulate the runout dynamics and runout of granular flows, and which have been applied to rapid mass movements of the flow type.

The practical application of models of granular flows were also the subject of the previously EU-funded project DAMOCLES (“Debrisfall Assessment in Mountain Catchments for Local End-Users”; Contract No. EVG1-CT-1999-00007), and Crosta et al. (2001; 2003) provide a good overview on available models and underlying assumptions.

McDougall and Hungr (2004) further developed the continuum mechanics process model of Hungr (1995) for the analysis of motion of rapid flow slides, debris flows, and avalanches across three-dimensional (3D) terrain, based on the physical principles of mass and momentum conservation. Importantly, their model takes into account the effects of centripetal acceleration on rock-avalanche motion due to curvature of the underlying slopes:

2 σ z = ρh(g cosα + v κ z ) (3.7)

where σz is bed-normal stress; hz is the bed-normal flow depth; α is the inclination of the x-y plane from the vertical; v is depth-averaged flow velocity in the direction of

motion; and κz is the bed-normal curvature of the flow path in the direction of motion. They defined the frictional rheology at the bed by:

v τ = x σ tanφ (1− r ) (3.8) zx v z d u

v τ = y σ tanφ (1− r ) (3.9) zy v z d u

where τzx and τzy is bed shear stress opposing the direction of motion; vx and vy are the

components of velocity v in the x- and y-direction, respectively; ru is the ratio of pore

pressure to total bed-normal stress; and φd is the dynamic basal friction angle. Pirulli (2004) has adapted this model for simulating the runout of three large rock-slope failures in the European Alps. Crosta et al. (2004) used a similar 3D runout model based on Voellmy rheology within a Lagrangian frame of reference to model the travel distance and deposit shape of the 1987 Val Pola rock avalanche, Italy,

Denlinger and Iverson (2004) developed model of granular avalanche-type landslides across 3D terrain solely on the basis of continuum mass and momentum conservation, and stress generation between moving grains controlled by Coulomb friction. The obvious advantage of this model and the one of McDougall and Hungr (2004) is their applicability to rock avalanches, snow avalanches, debris flows, and pyroclastic flows alike, highlighting their general physical principles and interrelationships. What is more, the prediction potential of the model was directly tested and verified by physical

78 laboratory experiments involving the avalanche-like movement of dry sand over 3D terrain (Iverson et al., 2004).

The majority of these models are based on physical laws from fluid dynamics and based on conservation of mass or energy. However, Chemenda et al. (2005) note that, although numerical modelling appears to be the most powerful simulation tool for large landslides at present, physical lab experiments may be more appropriate to compensate for difficulties experienced when numerically modelling 3D brittle failures and large inelastic strain.

Based on the same physical principles, Kelfoun and Druitt (2005) presented a model for the catastrophic dry volcanic debris avalanche derived from a sector collapse of Socompa volcano, Chile, by assuming a constant retarding stress during runout, while using the surficial deposit morphology and runout as further constraints for calibration.

Thus it is not surprising that very few models explicitly address the effect of rock-mass fragmentation and comminution during rock-avalanche motion, especially given that many rock avalanches initiated as more or less intact rock blocks.

Smith et al. (2006) employed such a fragmentation term in the DAN model, and produced a runout very similar to that of a medium-sized rock avalanche. Gray et al. (2003) provided model expressions of describing the flow of granular avalanches around obstacles in their runout, using hydraulic theory, and thus contributed to the little studied process interactions of granular avalanches with local topography or even structural countermeasures. Other recent approaches such as the RASH3D model of Pirulli et al. (2006) have employed the St. Venant’s Equation which is used in fluid dynamics.

Figure 3-16:

Time steps in a numerical model of the 1987 Val Pola rock avalanche, Italy. This model is based on a Voellmy rheology also used for simulating the runout of snow avalanches (Crosta et al., 2004).

79 3.2.3.6 Discussion on Model Results

In summary, there exists a range of models for describing and explaining the physical behaviour of rock avalanches as prime examples of fragmenting granular flows. While empirical approaches are useful first-order approximations to encapsulate the phenomenon with few readily obtainable parameters, their limitation lies in their lack of spatial explicit predictions.

A range of one- to three-dimensional numerical models overcome this problem, hence paving the way for providing spatial information on the expected runout of a rock avalanche, given adequately specified initial volume and valley topography. The majority of these models are based on physical laws from fluid dynamics and conservation of mass or energy, although very few explicitly address the effect of rock- mass particle comminution during rock-avalanche motion (e.g. Smith et al., 2006). Most of these models are readily calibrated for single events, and produce satisfying results for simulating the runout. However, these models remain hardly applicable to other site with differing initial and boundary conditions, and most are used for back- analysing a single event rather than predicting possible future ones. The recent occurrences of ice/rock avalanches with runout >10 km demonstrate that rock avalanches are element within a continuum of complex multi-phase granular flows, which is presently intractable to formulate within a single numerical model.

Hence, there is a clear research gap in providing a transferable, let alone universally applicable, numerical runout model to rock-avalanche motion.

3.3 Hazard Map Types

Hazard maps for rock avalanches in alpine terrain comparable to those derived for snow avalanches and debris flows are very scarce, and the few exceptions do not follow any specific standard. Importantly, there are even few attempts of explicitly mapping the susceptibility to rock avalanches as a first-order substitute, as is common practice in the majority of studies on landslide hazard assessments. The main reason for this, of course, is the lack of detailed knowledge on the threedimensional rock-mass structure on a regional scale. Hence the regional-scale approach to hazard assessment is limited to a statistical magnitude-frequency estimate independent of any lithological or structural conditions, as well as the spatio-temporal occurrence of trigger mechanisms.

There are ways to approximate the probabilistic occurrence of rock avalanches through combined modelling of their triggering mechanisms with topographic site characteristics. For instance, seismic hazard zonation maps may, together with critical topographic parameters for rock-avalanche triggering, provide a first-order approximation of this problem (Jibson et al., 2000).

More advances in hazard and risk assessments of rock avalanches have been made in countries outside of Europe, where they are a more frequent phenomenon. In several cases, for instances, there are site-specific suceptibility and hazard maps for a number

80 of volcanoes, which have repeatedly produced large debris avalanches. Common approaches to delineate hazard zones for such volcanic landslides include the use of scenario-based values of H/L = 0.2, H/L = 0.1, etc, which are then computed from topographic base data. The flanks of many volcanic edifices host evidence of former such events, so that cross-checks with field data allow a more realistic assessment of the hazard zonation thus derived.

Ideally, hazard maps pertaining to both volcanic and non-volcanic rock avalanches should contain both the immediate hazard of physical impact from the moving rock- avalanche debris, and any subsequent hazards arising after the deposit has come to a halt. The latter will in most cases require a more dynamical approach to hazard mapping, as boundary conditions are likely to change with possible formation and failure of rock-avalanche dams. Given the various timescales that off-site impacts may operate on, hazard mapping efforts will certainly need to be adjusted to the specific on- site needs following a given rock-avalanche event.

Chapter 4

SNOW AVALANCHES

4.1 Hazard Mapping Criteria

Introduction The aim of hazard mapping is to present the spatial variation of hazard on geographical maps. In order to avoid damages and fatalities, the simplest strategy is to avoid the presence of any human being or constructions in an endangered area, that can be characterized by the maximum run-out distance of an avalanche. However, it is often impossible to keep the infrastructures and human activities out of endangered areas. Thus the necessity to characterize snow avalanche intensity as a function of return period, in order to appropriately dimension protective measures and regulate land use activities according to the possible avalanches destructive power and related consequences on the exposed goods.

In Europe avalanche hazard mapping differs from country to country. Hazard mapping criteria reflects the definition of hazard levels as functions of different parameters: event frequency, event magnitude and individual risk.

Hazard mapping In Norway, hazard mapping is based on event frequency, mainly estimated on the base of based on event historical data analysis and statistical prediction models. frequency Three security classes are defined according to the maximum nominal avalanche frequency per year, or in an analogous way on the return period (years). Security Class 1, corresponding to the highest hazard level, is defined where T < 100 years, security class 2 where 100 < T < 1000 years and security class 3 (safe areas) where T > 1000 years.

Hazard mapping Other countries distinguishes hazardous levels on the base of event intensity. For mapping based on event purposes avalanche intensity is usually identified by the avalanche impact pressure, defined as magnitude P=k ρ v2, where v is the avalanche flow velocity (ms-1), ρ is the snow avalanche density (kgm-3) and k is a coefficient with values that are function of the avalanche type (dense vs powder) and

82 of the obstacles shape and size.

In France, adopting a design avalanche event characterized by a return period T=100 years (century hazard), three levels of increasing hazard are defined according to the following intensity thresholds: low hazard where P<1 kPa, moderate hazard where 1 kPa≤P<30 kPa and high hazard where P≥30 kPa (Figure 4-1). In Austria, the design avalanche event is characterized by a return period T=150 years, and only two zones of increasing hazard are defined according to the following intensity thresholds: yellow zone where 110 kPa.

Figure 4-1: Avalanche hazard zones defined according to France criteria: in violet high hazard (A3), in orange moderate hazard (A2) and in rose low hazard (A1). France mapping criteria also include the largest expected avalanche (yellow, AMV) and the protective forests (green, V)

Hazard mapping The third mapping criteria uses both event frequency and event intensity to define hazard levels based on event in avalanche prone areas. Although the different countries define an increasing hazard as frequencyand avalanche intensity and frequency increase (i.e. as avalanche return period decrease), the magnitude thresholds between hazard levels changes from country to country (Table 4-1).

This reflects the scarce knowledge on the potential effects of avalanches on people, structures and infrastructures and the need of explicitly evaluate risk to clearly understand what “high”, “moderate” and “low” hazard levels concretely mean in terms of human life or economic losses.

Table 4-1: Hazard zone Italy Switzerland Reference values fo avalanche intensity and frequency adopted in T=30 yrs and P>3 kPa T= 30 yrs Italy and Switzerland High or or for avalanche hazard T=100 yrs and P>15 kPa T=300 yrs and P>30kPa mapping. High hazard zone are coloured in T=30 yrs and P<3 kPa T=300 yrs and P<30 kPa red; blue corresponds Moderate or (for powder avalanches to moderate hazard T=100 yrs and 3300 yrs Low or (for powder avalanches T=300 yrs T=300 yrs and P<3 kPa)

Hazard mapping In Iceland hazard zoning is based on the estimation of the probability of being killed by an based on avalanche if one lives or works in a building under a hazardous hillside, that is on individual risk individual risk (Jonasson et al., 1999). Hazard zones are delineated using risk levels of 0.2×10-4, 0.7×10-4 and 2×10-4, with a risk lower than 0.2×10-4 considered acceptable (Arnalds et al., 2004).

83 4.2 Hazard Mapping Tools

4.2.1 Identification of snow avalanches hazard

4.2.1.1 Using historical data

Introduction A general sense for historical: by historical we shall mean both the strictly documented history and the broadest sense covering also oral testimony and equally all the traces that can be used as signs of past facts or hints to infer them: geological or geomorphological observations, tree rings for dendrochronological analysis, vegetation structure. All these sources must be exploited together to reconstruct an image of a past situation, event, or chronology. This presentation concerning historical data is divided into two parts:

- from 0 to 100 years

- from 100 years to 200 years

and doesn’t deal with visual observation of signs of past facts. This subject is dealt within the part “Naturalist approach”, treated afterwards.

These boundaries are arbitrary and questionable. First, the boundaries between the periods (from 0 to 100 years and from 100 years to 200 years) are driven by situation observed in various European countries, such as France and Italy: the avalanche survey began around the beginning of the 20th century. In other European countries, the boundaries may be different even if methodology remains the same.

Secondly, once a visual observation of past event has been done, the record has become a document, an archive and is therefore reflected from that time on in the realm of history. For example it is the case for photo-interpretation used to draw up hazard registration maps, such as CLPA in France, CLPV in Italy, MZA is Spain and avalanche cadastre map in Switzerland. The distinction between historical data and signs of past facts is not obvious.

4.2.1.1.1 From 0 to 100 years

Avalanche The chronicle of past avalanches begun, in most countries, around the beginning of the 20th permanent survey: century (Mougin, 1922) and, since then, hundreds of forest rangers registered on a template each the example of the event that occurred on the surveyed sites. Avalanche counts are registered, as well as French avalanche quantitative and qualitative information such as meteorological conditions, release and runout database altitude, deposit volumes, etc (Jamard et al., 2002). All the past observations are collected by the preposed institutions or reseach centres and stored in databases.

For such a large chronicle, data quality is of course subject to serious cautions (Garcia, 2002). For instance, some paths were not very well located in the past, whereas some others were not perfectly surveyed during the whole time period.

84 The collection of these avalanche data constitute the first step in the creation of historical registration maps (§ 4.3), together with the aerial photographs of the sites (which allows to check the localization of the site on a map finding avalanche marks on vegetation and relief (§ 4.2.1.2)) and the analysis of past written documents (§ 4.2.1.1.2) as well as witness testimonies.

Here follows the procedure adopted in France to analyze and map these historical data, since the French procedure is well structured and is representative of the methodology used in different countries in the updating of the avalanche historical cartography.

The “Enquête Permanente sur les Avalanches-EPA” is a chronicle, describing the avalanche activity on 5.000 determined French sites in the Alps and the Pyrenees. Its aim is not to register all the avalanches in France, but to be as exhaustive as possible on the considered sites.

In France, on February 12th, 12 persons were killed and 14 mountains chalets destroyed at Montroc (town of Chamonix). People were mostly killed in or just around their house. The French Secretary of Environment charged inspectors with establishing the conditions of the accident and the efficiency of avalanche mitigation policies. The inspectors concluded that the Department of Environment should improve some existing programs such as avalanche permanent survey or the past avalanche map. The Department of Environment assigned the Cemagref and the National Forest Service (Office National des Fôrets / ONF) to develop these projects.

Therefore, since 2001, the chronicle has been improved (Belanger and Cassayre, 2004): the objective was to focus on the sites with maximal exposure and to reduce uncertainties concerning the values stored in the database. The survey was improved by restructuring the network of forest rangers: for each site, the Cemagref and ONF identified a primary observer and a substitute one, so that no observation event could be missed. However, the survey has remained a delayed time observation (and not a real-time one). Then the Cemagref and ONF have been updating the list of sites, to take into account the evolution of constructions and roads. Three agents are in charge of deciding this update on the field: one is the main observer of the site, another one is a technician from the mountain risk unit, and the last one is a full-time specialized technician from the Cemagref. They decide to stop, to continue or to begin event observation on the site, and they write down their argumentation. The two main factors considered are the importance of potential and present risks, and the quality of data collected since the beginning of observation. These three persons report on several complementary supports: a site photograph, an observation map, and a site template.

- Photographs allow observers (and especially substitute ones) to check the localization of the site on the map (see Figure 4-2).

- The map is drawn on a black and white topographic background, with a scale of 1/25000. The team reports the edge of the avalanche site, open to downhill, and the avalanche main flows (see Figure 4-3).

The 3 persons decide some points of protocol, which are drawn on the map:

- an observation threshold: observers systematically report events which flow under the

85 threshold; they report an event that remain over the threshold only if they estimate it has some special characteristics.

- An alert threshold: see selected event investigations described hereafter.

- A point of observation: it is a place where you can easily go most of the time and which gives you a good view of the extent of avalanche. This information is especially useful for substitute observer. Observers shall examine events at least from this place. We normally take the photograph from this place.

The Cemagref registers all these data, and then publishes them in binders. They include maps, photographs and site bills, which are distributed to the observers and to the technicians from the mountain risk units. All the sites will be updated at the end of 2006. The Cemagref has re-edited the full protocol of observation much more precisely. However, the bases of reports still rely on an adaptation of the International avalanche atlas (De Quervain, 1981). This protocol is described and illustrated in an observation handbook, given to all observers (Garcia, 2003).

The Cemagref registered data of all events in a database since 1970. However there were some intermediate supports (both on paper and computers), and some transcription or migration mistakes are still possible. Therefore, the Cemagref has decided to scan all the observation notebooks that were completed. After scanning the notebook, Cemagref provides the users with the databases together with the scans, as complementary information and crosschecking tool for the older events.

The usual valorization of this unique database is local predetermination using naturalistic approaches (Cemagref ETNA 2000), advanced physical modelling (Naaim et al., 2004) or coupled statistical-dynamical approaches (Meunier and Ancey, 2004), (Eckert et al. 2006). But large scale studies relying on the data collected in an entire district are now also investigated.

Figure 4-2:

Example of site photograph. It reports from top to bottom, the name of the village, the zip code and the site number, as well as some red lines for the avalanche main flows.

86 Figure 4-3:

Example of the observation map for the avalanche permanent survey. It reports for site # 202 from uphill to downhill: the reference number of the site, the edge of the avalanche site open to downhill, the avalanche main flows, the observation and alert thresholds, and the observation point. Moreover, the grey areas report witness testimonies from the past avalanche map.

The past and The first CLPA maps date back to 1970, the year an avalanche killed 39 young people in Val geomorphologic d’Isère (Savoie, France) (§ 4.3). A few months after the disaster, the French Government made map: the example the Cemagref responsible for mapping all the major threatened areas: more than 6000 sq.km. of of the French the French Alps and Pyrénées were mapped between 1970 and 1978 including the most tourist- CLPA crowed areas during winter time and the main mountain villages.

At the beginning of the nineties, this first issue has been updated and a reprint produced. Then, as said previously, after the Montroc avalanche in 1999, Department of Environment assigned the Cemagref of improving some points of CLPA execution. Since 2002, this program is called in France “Carte de Localisation des Phénomènes d’Avalanches” but it is still rather better known under its older name “Carte de Localisation Probable des Avalanches”, the abbreviation remaining CLPA.

The Cemagref improved the map by adding two documents which are now systematically distributed with the map.

- Since the beginning, the Cemagref has kept tracks of testimony references and main facts on a paper file for each avalanche. The Cemagref has digitized these data in a database, and published them, to the exception of witnesses’ names.

- For each mountain massif, the Cemagref has edited a synthesis of the main avalanches that happened. It also includes descriptions of mountains, geomorphology, environment, and it integrates climate and precipitation data in relation with snow and avalanche. French Meteorological departments writes these last two parts.

The Cemagref has also implemented regular update processes:

- Since the last edition, ten to thirty years ago, new events have occurred, and some of them may have been bigger than the ones known before: the envelope of past events

87 shall be extended. Each year, the Cemagref makes a full investigation on ten percent of the studied valleys, searching for such big events. It is the decennial update.

- In addition, there is an annual update. It is limited to places where professionals send the Cemagref some information about events exceeding the drawing of avalanche on the map. The Cemagref collects part of these information thanks to the selected event investigations described hereafter.

The Cemagref complements now the photo interpretation with a field analysis, to find evidences of avalanche on the vegetation and relief, because some of them cannot be seen on aerial photographs: for example, a narrow corridor where avalanche has destroyed trees, but which is still hidden by neighboring trees.

In addition, news valleys or parts of valleys are studied: at the present time the Cemagref has studied 7000 sq.km of valleys and their neighborhoods.

The Cemagref registers all these data, and then publishes them in binders. They gather maps, testimony references and main facts, as well as massif synthesis. We distribute them to leaders, professionals, forest rangers and witnesses. All professionals appreciate this map for its precision and its exhaustiveness: It is he reference map for avalanche in France. However, nobody shall forget that this is not a full prospective or hazard map (even if it is an essential part of it).

Complementaritie The survey and the map are fully complementary. They share some common characteristics: s between CLPA and EPA Both focus on areas where there are constructions or roads. They are not adapted to mountain sport activity, which need short-time analysis of the snow stability.

All technicians use both of them as references about past avalanches, as opposed to broader public. Those ones do not know avalanche observations well, neither do they understand their precise meaning. To inform people about natural risks, the Department of Environment asked us to widely spread data. It requires that we vulgarize the above notions about avalanche observation, such as differences between past avalanche and hazard identification, or avalanche forecasting.

From this purpose, the Cemagref shall put the data on an Internet site, with a geographical interface. A temporary web site already allows consulting the past avalanche map. The address is http://www.avalanches.fr/

To make it possible, we have put all the data about events, sites, and observation network on databases. They are on the software Microsoft Access for textual parts, and ESRI ArcGis station for geographical data. They will be migrated on Oracle, with the connection Arc SDE for geographical data.

The Cemagref publishes all maps on A3 sheet (approximately 420 x 280 mm). The cutting grid is regular and it uses the same base for the map survey and the past avalanche map. There are approximately 300 tiles for the survey and 500 tiles for the past avalanche map.

88 The Cemagref distributes the products (maps, cards, photographs) in formal binders. It allows easier fillings and updates.

The event Another program concerns investigations about selected events, wherever they happened. This investigation: the program covers all mountain risks: avalanches, torrents, rock falls, landslides, etc. The criteria to example of the select these events are: special size or characteristics, damages or proximity to constructions or French avalanche open roads. Mountain risk technicians have reported the descriptions of the context, the event database and the damages. Then, these event investigations are registered in a database. This database is registererd by the National Forest Service (Office National des Fôrets / ONF)

For avalanches, the investigations are connected to the avalanche permanent survey and the past avalanche map:

- All observers of the permanent survey have to alert mountain risk units if the event goes over the alert threshold predefined for each site of the avalanche permanent survey; or, for all avalanche sites, if the event exceeds the drawing on the past avalanche map.

- A technician investigates the event. Besides his report, he draws a map of the event, and sends these documents to the Cemagref.

The Cemagref integrates these information to the past avalanche map, in the annual update.

4.2.1.1.2 From 100 to 200 years: the place of historian

Many decision processes make it a rule not to consider more history than is immediately present in the mind of the attendees. The overwhelming majority run on such financial and schedule bases that only superficial investigations can be envisioned. Nevertheless natural risk management in general, and avalanche risk in particular, cannot ignore the far-off past. It is particularly true for post-crisis inquiries and court requisition (court seek to find similar cases to show that an accident or a catastrophe was foreseeable) in which case much financial limitations are released. Therefore, the search for historical, oral or written information, of past events is of paramount importance, and therefore can not be ignored. Three key question appears in order to determine the best way of taking into account old information:

- how to collect historical information systematically ?

- how to analyse his content taking into account the historical context ?

- how to archive this historical information ?

In France, historians and natural hazard professional finalized a methodology (Coeur et al., 1998) related to collection and storage of historical sources and the interpretation of historical data.

Collection of As information can be in oral or written form, two ways of investigations were developed: information - Collection of oral information: it consist in interviewing witnesses using a method which

89 allows : (i) to evaluate the credibility of information thanks to questions which are indirectly crosschecked between them, and (ii) to know the connection between the witness and the facts he reports (property owner, technician which intervenes on the site, simple visual witness...). Each interview is devoted to get other witnesses, this cycle of "connected" testimonies stops when the first witnesses are again quoted.

- Collection of written information: Archive exploration is rather complex. For example French public archives are periodically vetted, classified and moved to the public record archives for each department. They are then classified according to quite precise rules which must be known. Moreover, there are also private archives to which it is difficult to have access. And there can be a huge amount of locations where to check documents. They are often located by the means of oral testimonies; in this case the historian training is very invaluable.

Historical information retrieving is thus an expert work, and it is very delicate for a neophyte to find something. In addition, this work is long and tiresome, and it is not appropriate for searching for information related to a given avalanche path. It is more suitable to obtain historical information on a whole territory which corresponds to the archive (communal or departmental).

The interpretation One of the main concerns of this work consists in obtaining information on the localization of of historical the described event. Indeed, if it is impossible to position the avalanche layout on a map, even information partially the information loses all interest in the framework of risk management and risks zoning . As far as he focus on phenomena the natural hazards professional will try to extract facts. However judging facts requires catching the impression made on people of possibly several centuries ago, which implies an understanding of their representation of the phenomena. Moreover it is imperative to cross some information so as to restore the reality of the facts through the documents available. Some requests for a subsidy or an exemption of taxes is likely to magnify damages to obtain as much as possible. This phase of analysis is very significant. The historian and natural hazards professional will intertwine their knowledge in close cooperation trying to see through the eyes of their ancestors but in the light of today : the historians can, starting from the nature of the document (personal notebook, deliberation, report...), of its author (journalist, clergyman, mayor, engineer, victim…), express an opinion about its contents by putting events into their historical, political, religious and administrative context. The knowledge of the customs and habits can also throw light on the reality of the reported facts. The technician, on the basis of his knowledge of the phenomena, can evaluate the physical reality which can be described in a misrepresented way. Sometimes the terrain has undergone such transformations that the place cannot be considered the same. Toponymy must also be analyzed with a very great circumspection : the place names can be sometimes common or has sudden changes of orthography. Finally the place could vary in time; the localisation of all the natural or artificial reference marks must be checked, such as for example the location of a road between two villages which could vary in time. As an example, following the dramatic avalanche of Montroc, in Chamonix, France, 1999, an old writing wasdug out that reported the avalanche passing the road. The accusation was to have neglected this knowledge. Anyway, it turned out that, at that time, the road was on the other side of the river, and that was changed probably because of the menace.

The data storage Collecting and archiving historical sources are indissociable.

90 For any indexed document, it is essential to indicate:

- who found this document;

- how was it found (chance, systematic research, …);

- where was it found, and where is it possible to consult the documents;

- the author of the document;

- its nature;

- its purpose;

- the interpretation of the historian and the technician. It must be kept a traceability of this interpretation. Not only the conclusions must be archived.

- The date of discovered document;

- the date of communication of its knowledge;

For a given territory, it is essential to indicate:

- who made the document retrievals?

- With which aim was made the research ?

- Which were the decision-making process, consulted archives?

- Which archives are not consulted , were not found are considered disappeared or destroyed, or gave no result ?

In such way incremental research should be possible. If the condition of a fulfilling referencing is not fulfilled, further research will most likely go over and go through the same steps, because a tedious crosschecking will have to be made by hand to establish what documents have been actually included in the first exploration. The case is even worse for witnesses.

History could bring to light facts more or less well described, hard to get to, and allow to form an opinion on their veracity and which credence one could give to such events in a contemporary process of risk management. Lastly, in front of the difficulties encountered to analyze historical document, one could make improvements in our practice of contemporary information storage so that our successors are not confronted with the same difficulties. However, it should be recognized that archive search is long and tiresome, and today’s policies of risk management are often incompatible with careful historical investigations, due to the financial aspect but also due to the duration of such studies: It is not always possible to differ certain decisions waiting for the completion of the project.

Example of historical « (Source: Fonds M. Burnet, bulletin paroissial, janvier 1935) testimony (father Jean- Baptiste Dédevens – 1720) and related Du 19 au 20 février 1720, environ minuit, il est venu une avalanche de la Golèze, qui, se joignant avec celle des Poupes,

91 cartography obtained sont venues toutes deux fondre vers la digue située au-dessus de l’église, lesquelles avalanches, heureusement, se from testimony partagèrent par le moyen de cette mauvaise et méchante digue ou tourne, qui tourna si à propos, qu’une partie passa de çà et de là de l’église et de la cure, sans l’endommager en aucune manière ; toutefois, la partie du vent mis en bas, la maison de la confrérie du Saint-Esprit, sise pour lors au milieu de la tourne, au-dessus de l’église, brisa la croix du cimetière, où la neige était élevée plus de trois pieds au-dessus de la moitié du clocher, cassa le toit d’une boutique inutile, sise au bout du jardin, passa par le milieu du jardin tenant de là jusqu’à la maison de la Gervaise Semblanet, et s’alla terminer à la rivière ou eau de Bérard. Du côté de bise, elle ne passa pas le grand chemin, mais elle s’étendit bien loin du côté du Molard, jusqu’à la fenêtre de la chambre, de la maison de Jean Ancey ce qui causa un si grand effroi que je résolus de quitter la cure à cause du grand danger.

Dédevens, curé ».

Figure 4-4:

Cartography obtained from the testimony by father Jean-Baptiste Dédevens –1720

4.2.1.2 Naturalist Approach

Topographic and Whatever the extent of the study, the nature of the objective of protection or the precision of the geomorphologic required opinion, the field knowledge is essential: it gives the expert a lot of information. It is approach necessary to visit the site concerned, generally several times and if possible at various seasons. These recognitions are supplemented by a fine analysis of local topography and all archives and testimonies related to the past avalanches in the path. This first stage is often long and requires much care. Only this step really enables to integrate all the information about the question.

A work only based on photographs, maps, or digital elevation model is destined for coarse error, even for failure. Only the direct observation of the site makes it possible to synthesize its

92 principal characteristics.

The examination of the specific geographical characters, with all their distinctive features, and the research into all the traces done by the past avalanches, is carried out effectively through 2 steps:

- the geographical study of the avalanche site;

- the historical study of the past avalanches (see § 4.2.1.1).

These two studies strongly deserve to be simultaneously started. The geographical study examines the site regarding it as the “theatre” of the avalanche. The determining elements for release and evolution of the phenomenon must be observed at two levels: in a global way (regional situation, characteristic altitudes, general exposure, dimensions, etc.) and in a detailed way.

For each starting flowing and runout zone, it is necessary to define different characteristics:

– topographic and geographic (surfaces, slopes, differences in height, profiles, altitudes, exposures,), with, for example, a table of values and a slopes map;

– geomorphologic: singularities, type of path (unconfined or gully), roughness (including roughness due to the vegetation).

According to the cases, it is necessary to go further :

– in the starting zone: mainly to evaluate the potential of entrained snow including the research of the homogeneous units of area, their relative position and their interdependence;

– in the flowing zone: to discover the site’s singularities which may influence the avalanche dynamics (slope’s change, possibility of channelling or spreading out for the avalanche, sinuosity);

– in the runout zone: to understand the conditions of deceleration and of snow deposit with the slope’s change.

Nevertheless it is necessary to appreciate the real influence, the conditions of action, the limiting factors, the effects of combination of each of these elements. In such way, it will be possible to determine the relative influence of each parameter regarding the avalanche dynamics. References to other similar sites are then advisable.

Dendrochronolog In the timbered corridors, the dendrogeomorphology can bring a response while making it y possible to reconstitute a chronology of the major avalanches (Burrows et al., 1976; Schweingruber, 1996). Indeed, when a tree undergoes a shock violent enough (falls of block, avalanche, close tree...), the ring of the year will develop a wood of reaction in its opposite part i.e. a wood of compression, yellow or brown-red, short, with dense cells and thick partitions which is formed on the opposite side of the impact (Figure 4-5). This wood of reaction and the scars of abrasion are the indicators retained like witnesses of the passage of an avalanche in

93 many studies undertaken since about thirty years. Concretely, the method consists in using a drill to carry out samples in the trees. One can also use a slicer to take a disc if the tree has already died or is very damaged. These samples are carried out in zones not subjected to the avalanches (series of reference) and in suspect zones (samples). The comparison of the owners of growth of the samples with the series of reference makes it possible to highlight avalanche years which are then validated by a visual checking of the discs (presence of wood of reaction). Reconstitutions were thus carried out in the Rocky Mountains (Potter, 1969; Carrara, 1979; Butler and Malanson, 1985), in the Italian Alps (Comunello et al., 2001; Bezzi et al., 2003), in the mounts Chic-Chocs (Boucher et al., 2004), in the Spanish Pyrenees (Muntan et al., 2004) and in France in the valley of High Romanche (the Hautes Alpes) in 2005 (Perfettini and Corona, 2006).

Figure 4-5:

Disc of a tree impacted by a snow avalanche.

Photo- This method, developed by the French Institut Géographie National (IGN) in 1969, requires the interpretation use of pairs of panchromatic monochrome aerial pictures taken at the end of summertime in the avalanche path. The operator scrutinizes these pictures through a stereoscope (Figure 4-6). Thus, he can get information, sometimes tiny details, such as avalanche tracks in the vegetation or differences in height (obvious in the mature forest areas), or marks on human activities as well as landscape features which could make easier a snow slide release, flow, braking and stopping. Because of these requirements, the use of winter pictures taken when snow covers most of those details is unsuitable. For example all this obtained information is printed in orange on CLPA. Recently a field investigation complements this process to improve it.

94 Figure 4-6:

Operator scrutinazing aerial pictures through a stereoscope

4.2.2 Evaluation of snow avalanches occurrence

Introducton A good knowledge of avalanche frequency and dimension is crucial for avalanche hazard mapping procedure (§ 4.1). Historical avalanche events give an idea of the size and type of avalanche that may occur in the future; unfortunately, for many avalanche tracks, information about historical avalanche events is incomplete and, subsequently, uncertainty concerning avalanche frequencies and dimensions exists. Statistical models are used to determine run-out distances for a given probability from a basis of recorded known avalanche run-outs (McClung and Lied, 1987; McClung et al., 1989; McClung and Mears, 1991; Barbolini and Cappabianca, 2002). The major limitation of this approach is the usual lack of a sufficient number of recorded events upon a given avalanche path. Some author tried to circumvent the problem by standardising records from several different avalanche paths to produce a combined artificial history for an average avalanche path within a mountain region. Transferring avalanche between paths has been done either by topographical criteria (Keylock et al., 1999) or by using physical models (Jonasson et al., 1999). However the main weaknesses of statistical approaches, that is extrapolate the empirical probability distribution fitted to field data in order to estimate the avalanche frequency in area not covered by historical information, still remains.

The run-out distance is not sufficient alone to characterize the dimension and the magnitude of an event, required for hazard mapping in many countries (§ 4.1) and for the reliable dimensioning of protective measures. More detailed information about impact pressures, their duration, flow height and run-up height are required and consequently the use of analytical and simulation models for snow avalanches (§ 4.2.3.2-3).

Magnitude Given an avalanche site the avalanche run out and magnitude for assigned frequency of frequency occurrence, or return period, in year, are usually derived from the statistics of the input mostly relationship related to avalanche event magnitudes, i.e. the release depth (Barbolini et al., 2003; Bocchiola et al., in press). This implies the assumption of a release area constant over each avalanche site. Maggioni and Gruber (2003) studied the influence of topographic parameters on the frequency distribution of the release area, giving a first contribution to directly use it as input for uncertainty modelling of avalanche run-out distances and impact pressures using Monte Carlo methods.

The release depth is often assumed to coincide with the snow depth precipitation in the three

days before the event, or three days snow fall depth, P72 (Burkard and Salm, 1992; Ancey and

95 Meunier, 1999). The value of P72 is calculated by evaluating the “increase in snow pack depth during a period of three consecutive days of snow fall” (Sovilla, 2002; Barbolini et al., 2004).

The value of P72 for every year is taken as the greatest observed value of the positive difference in snow depth calculated using a three days wide window, moving by one day steps (Barbolini et al., 2004). This is evaluated with respect to a flat area and then properly modified for local slope conditions and snow drift overloads (Salm et al., 1990; Barbolini et al., 2002; Barbolini et al., 2003).

The T-years avalanche is therefore simulated as the avalanche with depth at release equating the

(estimated) T-years value of P72, P72(T)

Notice that in some cases, more complex schemes can be adopted, e.g. by deriving the frequency distribution of snow avalanches magnitude by synthetic simulation of the occurrence

of values P72 above given thresholds (as in Ancey et al., 2004), therefore profiting of a greater amount of information.

However, such approaches provide are non standard tools and require relevant expertising, considerable data handling, besides being time consuming.

The typical conceptual scheme adopted is as reported in Figures 4-7 to 4-9. First, a reliable release zone is identified, based on historical findings and some critical avalanche profile(s) is chosen, as in Figure 4-7. In case, sensitivity analysis can be carried to the choice of the proper release zone and profile(s).

Then, the input of P72(T) is fed to a properly calibrated dynamic model (Figure 4-8), to calculate the run out zone and snow depth, velocity and pressure for each point along the avalanche track.

This eventually result into hazard mapping, exemplified in Figure 4-9 according to the Swiss guidelines (red, blue and yellow zone).

Figure 4-7:

Preliminary analysis for 1-D simulation of avalanche dynamics. Case: Vallecetta avalanche, Bormio (see Figure 4 below). a) Aerial view (Picture regione Lombardia, adapted in Riboni et al., 2005). b ) Definition of the release zone (green area) and of the main (a) avalanche profile (b) (green dots)

Figure 4-8:

Conceptual approach for avalanche hazard H72 : Maximum annual Three days mapping a) Snowfall cumulated snow fall input. Adapted from Ancey et al., 2004. b) Scheme of the avalanche dynamics (by courtesy of SLF Davos)

96 Davos) (a) (b)

Figure 4-9:

Avalanche hazard map, according to the Swiss approach (yellow, blue and red zone).

According to Salm et al. (1990) the average release depth, measured perpendicularly to the soil)

can be calculated as follows:

0.05(z − 2000) P ()T = P* ()T ⋅ f ()ψ + (4.1) 72 72 100

* Where P72()T is the maximum snow precipitation over three consecutive days, with return period T; f ()ψ is a slope factor that takes into account the decreasing of the release depth as the terrain slope increase and z is the average altitude of the release area above the sea level.

In Switzerland the statistical analysis of snowfall data from the archives of the Federal Institute * for Snow and Avalanche Research (SLF) led to the definition of reference values of P72 different values of return period (see Table 4-2).

Return period T * Table 4-2: Maximum three day snowfall depth - Reference value of P72 , ψ =28° (years) Swiss reference values Small (m) High (m) of P*72 according to 30 0.88 1.50 different return periods (Salm, 1990) 100 1.06 1.81 300 1.23 2.10

Regional These reference values have sense if used inside the Swiss territory, but loose their validity if approach to applied to different mountain regions. Thus, in the other countries the statistical estimation of T- frequency years quantiles of P72 can be carried out by distribution fitting of the single site observed data, assessment of H72 i.e. of the maximum annual observed values of P72 for a range of years of observation at a given snow gauging station. However, according to the theory of extreme values (Darlymple, 1960; Hosking et al., 1985; Hosking and Wallis, 1993; Kottegoda and Rosso, 1997; De Michele and Rosso, 2001; Katz et al., 2002) in order to provide reliable estimates of the T-years quantiles using empirical distribution fitting, a least number of observations is required, in the order of

nobs = T/2. For T = 300 years, this amounts to about 150 years of sampled data. As an example, in the Italian alps, except for a very few cases, only short series of observed snow depth are available, covering a period of 20 years or so (Bocchiola and Rosso, in press). This is similarly true for many countries; for instance, in the Swiss Alps, daily snow data series are generally available for periods of about 60 to 70 years (Laternser and Schneebeli, 2003), resulting in about T = 140 years, still unfit, albeit surely more reliable.

97 The lack of observed data for distribution fitting of extreme values in hydrological sciences can be overcome by using regional approaches, including index value approach (Darlymple, 1960; Sveinsson et al., 2001; Bocchiola and Rosso, in press; Bocchiola et al., in press).

This implies that values of a hydrological variable that are scaled, i.e. divided by an index value (Bocchiola et al., 2003) have identical frequency distributions across all sites within a given homogenous area, or region (Hosking and Wallis, 1993; Burn, 1997).

This allows distribution fitting by using only one sample, obtained by grouping the scaled values of the hydrological variable observed at the gauging stations inside the homogeneous area, thus increasing the available number of samples and reducing estimation uncertainty (De Michele and Rosso, 2001). The so obtained distribution, valid for the whole considered region, is usually named growth curve. The use of this growth curve results in more reliable estimates as compared to single site analysis (Madsen et al., 1997; Katz et al., 2002). The regional approach has been used also for avalanche mapping procedure (Barbolini et al., 2002; Barbolini et al., 2003).

The index value can be evaluated according to different approaches (Bocchiola et al., 2003; Robson and Reed, 1999; Sveinsson et al., 2001). The expected value, or the mean of the single site distribution, estimated by the sample average of the observed values at the specific site i is used by Barbolini et al. (2003) and Bocchiola et al. (in press):

1 Yi μH 72 i = ∑ H72i,y (4.2) Yi y=1

th where Yi is the number of years of observation and the suffix y indicates the year y . The related standard error of estimation is:

σ H 72i σ μH 72i = (4.3) Yi

with σH72i sample standard deviation of H72 at the specific site i. The scaled value of H72 at each specific site i is therefore defined as:

* H 72i H 72i = ≈ Fi ()1; .. (4.4) μH 72i

* The symbol Fi indicates the cumulated probability distribution of H72 at site i. The average of * H72i is obviously 1 and the remaining moments need to be estimated from data. The key hypothesis for the regional approach is that the distribution Fi is the same at each site i. The first step in regional analysis is the definition of the homogenous regions (Burn, 1997), i.e. those regions where homogeneity of the function Fi holds for each site. This requires first partitioning the group of the considered stations into sub-groups, or regions. Partitioning can be based on grouping techniques, including for instance Principal Components Analysis (Rohrer et al., 1994; Baeriswyl and Rebetez, 1997; Bohr and Aguado, 2001), region of influence assessment (ROI, e.g. Castellarin et al., 2001), or seasonality measures (Burn, 1997; De Michele and Rosso,

98 2002). Then, appropriate tests are carried out to verify the homogeneity of the frequency

distribution Fi into each region. The typically adopted approaches assess either the homogeneity of the L-moments of the observed data sample (Hosking and Wallis, 1993; Burn, 1997), or the homogeneity of the single site distributions (Wiltshire, 1986).

* Accordingly, one can hypothesize homogeneity of the distribution of H72i in the considered region and evaluate a proper distribution fit.

General Extreme Value (GEV) distribution provides acceptable description of extreme values for a wide range of variables, including precipitation (Cong et al., 1993), floods (De Michele

and Rosso, 2001; Katz et al., 2002) and three day snow fall depth, H72 (Ancey and Meunier, 1999; Barbolini et al., 2002; Sovilla, 2002; Barbolini et al., 2003; Bocchiola et al., in press).

The GEV quantiles featuring T-years return period are evaluated as:

* α p H 72(T ) = ε p + ()1− exp()− k p yT (4.5) k p

with yT Gumbel variable, yT = -ln(-ln((T-1)/T)) and εp, αp, kp, location, scale and shape parameters of the distribution.

4.2.3 Dynamical behaviour and run-out distance prediction of snow avalanches

Introduction In 1998, during the SAME project (Snow Avalanche Modelling, Mapping and Warning in Europe – 4ième PCRD), the deliverable 4 was dedicated to a survey of computational models for snow avalanche motion: various models for computation of avalanche motion, both empirical procedures including statistical and comparative models for run-out distance computations as well as dynamics models describing the physics of dense and powder snow avalanches, the coupled combination of these, and slush flows were presented. Ten years later, some of them have been discarded or have been improved and presentations need to be updated. This part focus on and is limited to models which are currently used in European countries for hazard mapping and for expertises.

4.2.3.1 Statistical approaches for determining avalanche run-out

Empirical models for snow avalanches are based on statistical/topographical models or comparative models for estimation of avalanche run-out distance.

The first empirical models developed were the regressive models (Bovis and Mears, 1976; Lied and Bakkeøi, 1980). In these models the longest run-out distances from well documented avalanche paths are related, through statistical regression analysis, to a selection of topographical parameters (§ 4.2.3.1.1).

In inferential models probability distribution functions are fitted to a set of run-out distances, properly expressed (see § 4.2.3.1.2). These models were developed at a regional scale (McClung and Lied, 1987), that is run-out distances from different avalanche sites are adequately scaled and grouped into an unique data sample and an appropriate distribution function is fitted. Recently Barbolini and Cappabianca (2002) proposed an inferential analysis of historical run-

99 out distances at a local scale, using all the data systematically recorded, in a given time period, on an avalanche path and not only the maximum events. Thus it is possible to obtain the relation between run-out distance and return period for the analysed site.

At least, comparative models are based on methods for evaluating the similarity between path profiles (Butler and Malanson, 1992).

Empirical procedures are normally applied to dense snow avalanches. However, in principle, there is no reason why they could not be applied to slush flows and powder snow avalanches if a sufficient number of precise observations are available.

4.2.3.1.1 The Lied and Bakkehøi statistical α/β-model

The statistical α/β-model (Lied and Bakkehøi, 1980, Bakkehøi et al., 1983, Lied and Toppe, 1988, Bakkehøi and Norem, 1994) was developed at NGI and governs maximum run-out distance (defined by the α angle, average inclination of the avalanche path between the starting point and the run-out position along the profile) solely as a function of topography. The run-out distance equations were found by regression analysis, correlating the longest registered run-out distance from 206 avalanche paths to a selection of topographic parameters. The parameters that had proved to be most significant are presented in Table 4-3.

Table 4-3: Symbol of Parameter description parameter Topographic parameters governing β (°) Average inclination of avalanche path between starting point and point of 10o maximum runout inclination along terrain profile. distance θ (°) Inclination of top 100 vertical meters of starting zone. H (m) Total height difference between starting point and lowest point of best-fit parabola y=ax2+bx+c. y″ (m-1) Related to curvature of avalanche path.

The β-angle is empirically found to be the best characterisation of the track inclination, and the regression analysis revealed that the β-angle is also the most important topographic parameter. In fact, in general it would appear that β is the only statistically significant terrain parameter. The usual form of the model is that of a simple linear regression relation: α =Aβ+ B, thus the name of “α/ β-model”.

Studies carried on in different mountainous regions confirmed both the advisability to use topographical parameters in order to predict avalanche run-out distances of extreme events and the importance of the β angle for α estimation. Table 4-4 gives some predictive relations obtained for different European countries.

Accuracy of Table 4-4: Mountain area No. of Hypothesis Equation the model (References) events 2 Regression models σ(°) R (-) proposed by for α = 0.96β −1.4 2.3 0.85 different mountainous −4 2.28 0.85 regions. 206 All avalanches α = 0.92β − 7.910 H Norway (Bakkeoi et al., + 0.024 Hy''θ + 0.04 1983) H>900 (m) α = 0.94β 0.035θ − 2.6 1.02 0.81 20 β ≤ 30°

100 α = 0.89β + 0.66 1.62 0.73 Italy – '' 1.31 0.82 Cordevole basin 53 All avalanches α = 0.89β − 0.62Hβy (Barsanti, 1990) + 27.36 Hy''θ − 2.6 Canada α = 0.93β - 0.75 (McClung and 126 All avalanches Mears, 1991) Alaska α = 0.74β + 3.67 - 0.58 (McClung and 52 All avalanches (α = 0.86β ) Mears, 1991) Sierra Nevada α = 0.67β + 2.5 - 0.6 (McClung and 90 All avalanches (α = 0.76β ) Mears, 1991) Colorado α = 0.63β + 4.68 - 0.5 (McClung and 130 All avalanches (α = 0.8β ) Mears, 1991) 88 All avalanches α = 0.5β + 0.35θ −1.01 3.8 0.63 Japan (Fujisawa et al., No buildings α = 0.92β + 490.4y'' −1.49 - 0.98 1993) 18 interaction L>100 m Italy –Rabbi α = 0.87β +1.71 1.86 0.85 Valley '' 1.45 0.91 54 All avalanches α = 0.94β − 0.36Hβy (Castaldini, 1994) + 4.36θ βy'' + 6.8910−5 Hβ −1.77 α = 0.946β − 0.83 1.5 0.92 −6 '' 1.3 0.94 Austria 80 All avalanches α = 0.97β − 0.610 Hy θ (Lied et al., '' 1995) − 0.032y − 0.07θ + 1.54 β ≤ 25° α = 1.14β − 4.66 0.87 0.77 France 2.69 0.66 168 All avalanches α = 0.82β + 2.82 (Adjel, 1995) Iceland α = 0.85β 2.3 0.71 (Johannesson, 45 All avalanches 1998) Catalan 216 All avalanches α = 0.97β −1.2 1.74 0.87 Pyrenees α = 0.86β +1.* 5 1.98 0.75 (Furdada et al., 64 All avalanches 1998) Italy – Valtellina 94 All avalanches α = 0.943β 1.4 0.92 (Fellini, 1999) 17 Livigno Valley α = 0.882β 1.93 0.88

4.2.3.1.1 The McClung and Lied Runout Ratio Model

The Runout Ratio Model was proposed by McClung and Lied (1987) and then revisited by Mclung et al. (1989) and Mclung and Mears (1991). It is based on the same parameters as the α/β model, but a new topographic parameter is introduced in order to describe the run-out distances of extreme events: the runour ratio, RR:

ΔX tan β − tanα RR = = (4.6) X β tanα − tanδ

where Δx is the horizontal distance from the α-point to the β-point, Xβ is the horizontal distance from the starting zone to the β-point and δ, the angle from the α-point to the β-point.

The analysis of data from several mountain ranges shows that the Runout Ratio is a random variable, statistically independent from the topographic variables that describe the avalanche site, and that the Extreme Value Type I or Gumbel distribution usually shows a good fit to

101 Runout Ratio datasets.

The cumulative distribution function for this distribution is given by:

P( RR ) = exp{}− exp[]− (RR − u)/ b (4.7)

where u is a location parameter and b is a scale parameter of the distribution.

The model parameters, u and b, varies from the mountain ranges. Examples of applications of this model and a comparative analysis with the regressive model α/β are given by Jóhannesson (1998) and McClung (2001).

4.2.3.2 Dynamical models

Dynamical models considers the physical processes underlying the avalanche motion and make use of sets of equation able to describe, at least in principle, the motion of a released snow mass from initiation to run-out. While empirical procedures may permit an assessment of run-out distance, dynamical models give much additional information concerning the nature of the sliding event (flow heights, velocities, etc.). These information are crucial for hazard mapping (§ 4.1) and risk calculation, where avalanche intensity determines the degree of damages to exposed elements. An overview of avalanche dynamical models is given in Figure 4-10.

Figure 4-10: Dynamical models Overview of dymnamical models Analytical models Simulation models

Flowing Flowing Powder Flowing and avalanches avalanches avalanches powder

“Sliding block Deformable body “Binary mixture” “Density current” “Coupled” “models models models models models

4.2.3.2.1 Analytical approaches

These models, called “sliding block” models, apply to dense snow avalanches and describe the flowing snow mass as a rigid body descending on a simplified topography.

If avalanche dimensions (length, width and height) are negligible respect to the length of the avalanche track, one can represent the avalanche motion trough the motion of a massive point along a trajectory. The second Newton law gives:

d()Mv dv dM = M + v = F − F (4.8) dt dt dt g r

That is

1 ⎛ dM ⎞ a = ⎜ Fg − Fr − v ⎟ (4.9) M ⎝ dt ⎠

Where M, v and a are the instantaneous avalanche mass, velocity and acceleration respectively.

102 Fg is the slope parallel component of the avalanche weight and Fr the resultant of resistive forces. Fr is the result of numerous phenomena which dissipate avalanche energy and different are the relation used to describe it (Kozik, 1962; Eglit, 1984; Perla, 1980).

If the avalanche can increase or diminish its mass along its motion through snow erosion or deposition along the track a further term is to be considered in (4.8):

dm F = v (4.10) e dt

Where m is the eroded/deposited mass.

The two analytical models mostly used are the PCM model (Perla et al., 1980) and the “Voellmy” block model (Voellmy, 1955), this latter has been adopted for many years in Switzerland as a technical base for hazard mapping purposes (Salm et al., 1990) and widely diffused in Italy from practitioners who operate in avalanche field.

4.2.3.2.1.2 The PCM block model

The 2-parameter PCM-model (Perla et al., 1980) describe the avalanche as a one-dimensional block of finite mass moving on a path of varying curvature. The reference point is the initial rest position of the block’s centre of mass. It is based on eq. (4.8) where the temporal derivatives are transformed into spatial derivatives. The resistance force is a function of a Coulomb friction term and a square velocity dependent term:

2 Fr = μ Mg cosθ + kv (4.11)

Where q is the local slope of the avalanche trajectory, μ and k the Coulomb friction and turbulent friction coefficients respectively.

The equation of momentum (4.8) becomes:

1 dv2 D = g()sinθ − μ cosθ − v2 (4.12) 2 ds M where s is the slope position and D=(k+dM/ds). The ratio M/D (m) is a parameter characteristic to the model, called mass-to-drag ratio. The inclination and perhaps the adjustable parameters are not constant along the path. For the model application an iterative solution procedure is described. The slope is divided into small segments of constant inclination and parameter values and the avalanche velocity at the end of each segment is calculated using the following relation:

0.5 B ⎡ ⎛ M ⎞ A ⎤ vi = ⎢αi ⎜ ⎟ ()1− exp βi + ()vi exp βi ⎥ (4.13) ⎣ ⎝ D ⎠i ⎦

103 A Where vi is the flow velocity at the beginning of the segment;

αi = g()sinθi − μi cosθi

2Li βi = − ()M D i A B The initial velocity of the successive segment, vi+1 is then assumed equal to vi . When A θi > θi+1 the following correction in vi+1 is introduced:

A B vi+1 = vi cos()θ i −θ i+1 (4.14)

This procedure is iteratively applied until the block centre of mass stops before the end of the ith segment. The runout distance, from the beginning of this segment is:

⎡ A 2 ⎤ ⎡()M D i ⎤ (vi ) S = ⎢ ⎥ ln⎢1− ⎥ (4.15) 2 ⎢ α ()M D ⎥ ⎣ ⎦ ⎣ i i ⎦

For constant inclination and parameter values along an infinitely long slope, the result is analogous to that of Voellmy. Flow height and lateral spreading of the flow are not calculated by the model and are estimated basing on field observations and available historical data. The authors of the model (Perla et al., 1980) proposed the following ranges of values to use for resistance parameters: 0.1 ≤ μ ≤ 0.5 and 102 ≤ M D ≤ 104 .

Anyway, because of their considerable variability, a site specific calibration of the model is hopeful before its application for mapping purposes. Examples of model calibration in different mountain regions are given by McClung and Schaerer (1983), Bakkeøi et al.(1981) and Jónasson et al. (1999).

4.2.3.2.1.1 The Voellmy block model

Voellmy’s (1955) model is a one-dimensional block model for the calculation of avalanche runout distance. The sliding mass is considered as an endless fluid of height H reaching a terminal velocity by equilibrium of gravitational forces and shear forces on an infinitely long slope of constant inclination θ1. Based on hydraulic theory, the shear forces are represented by a dynamic drag proportional to the terminal velocity squared on the free, upper surface and a combination of a similar dynamic drag and a Coulomb friction proportional to the normal forces along the bed.

Hence the terminal velocity is expressed by the two-parameter formula:

1 2 Vt = [ξ H ()sinθ1 − μ cosθ1 ] (4.16) where density and drag coefficients are lumped together into the “coefficient of turbulent friction”, ξ [m/s2], and μ is the Coulomb friction coefficient. To account for lateral

104 confinement, H is replaced by the hydraulic radius (flow cross-sectional area divided by wetted perimeter).

The deceleration starts at a certain reference point, normally located where the actual slope inclination equals tan-1μ. From this point the run-out distance on a slope of constant inclination θ2 is computed by energy considerations:

2 2 −1 S = Vt []2g()()μ cosθ2 − sinθ2 +Vt g / ξ H D (4.17)

HD is the mean depositional depth accounting for the energy loss due to pile-up of debris and g is the acceleration of gravity.

The computed run-out distance is based on the assumption that terminal velocity is reached, and depends strongly on the selected location of the reference point, as well as on the values of the input parameters.

The values of μ and ξ are discussed by Buser and Frutiger (1980) and by Martinelli et al. (1980).

4.2.3.2.2 Simulation models

Simulation models can be distinguished depending on the type of avalanche they intend to simulate.

Referring to dense avalanches these models are based on the hypothesis that the avalanche can be described as a continuum deformable body descending on a real topography. Respect to analytical approaches these models allow a more detailed description of avalanche dynamical features. Depending on the mathematical formulation adopted for mass and momentum balance equations this class of models can be splitted in two sub-classes: (a) Navier Stokes Models and (b) Hydraulic models. The first make use of the complete form of the mass and momentum balance equations. Depending on the rheological formulation for the mathematical behaviour of the moving snow they can be subdivided in Newtonian models (Lang et al., 1979a-b; Dent and Lang, 1980; Hunt, 1995) and biviscous models (Dent and Lang, 1983). The latter use instead depth-averaged balance equations similar to those commonly used in hydraulic problems (§ 4.3.2.2.1); in this case the dissipative processes can be synthesized in a basal friction term, for which many different expressions have been proposed in literature (Grigorian, 1979; Savage and Hutter, 1989; 1991; Sampl, 1993; Naaim and Ancey, 1992; Natale et al., 1994; Bartelt et al., 1997 a-b, Eglit, 1998; Norem et al., 1987; 1989).

The approached adopted to describe powder snow avalanches are different from that of dense avalanches (Savage and Hutter, 1990). “Density current” models are based on local balances of total mass and linear momentum, often integrated over the current height. These models are restricted to the steep part of the track, where phase-separation and mass-change effects may be of minor importance under certain conditions. “Binary mixture” models describe the airborne turbulent particle flow as a two-phase flow (snow grains and air). In this binary description, mass and momentum balances are formulated for each of the phases and their interaction is accounted for by the mutual interaction force. Since powder snow avalanche are highly

105 turbulent, the equations have to be time averaged and the set must be closed by a turbulence closure model. These kind of models usually disregarded the interaction with the dense flow and thus are not able to model the early stages of a powder snow avalanche formation. These models disregard the interaction with the dense flow and thus are not able to model the early stages of powder snow avalanche formation.

“Coupled models” tempt to describe the dense flow as well as the formation and movement of the powder snow part, together with the coupled flow of both parts (Bartelt et al., 2000; Naaim, 1999). Coupled models are built upon separate sub-models for the dense and the powder layer, with an additional sub-model to describe the exchange of mass and momentum between these layers.

An extensive review of these models up to 1998 can be found in Harbitz (1998). The main features of the models currently used for hazard mapping, and expertises, both one-dimensional and two-dimensional are given below.

4.2.3.2.2.1 One-dimensional hydraulic models

In order to describe the longitudinal spreading of a dense avalanche, several one-dimensional time-dependent models have been proposed (Grigorian, 1979; Sampl, 1983; Eglit, 1968, 1998; Naaim and Ancey, 1992; Natale et al., 1994 ). These models are based on the fact that usually the length of an avalanche is much larger than its thickness, so an approach similar to that applied in shallow water theory or in hydraulics can be used.

The coupled partial differential equations for an avalanche flow has the form (Eglit, 1968; Harbitz, 1998):

∂h ∂hu + = q (4.18) ∂t ∂x e / d

2 ∂hu ∂(hu ) ∂ 2 + = ghsinθ − ()K gh cosθ − f − f (4.19) ∂t ∂x ∂x a / p 1 2

In these relations t and x are the time and the coordinate along the slope, u(x, t) and h(x, t) are the velocity and depth of the flow, qe/d is the erosion/deposition rate, θ(x) is the slope angle, Ka/p the active/passive pressure coefficient that takes into account the anisotropy of normal stresses, f1 and f2 the flowing and the entrainment friction respectively.

The main differences between the models of this type consist in the rheological description of the snow from which the formulation of friction f1 depends, and the inclusion of erosion/deposition processes.

VARA-1D model (Natale et al., 1994; Nettuno, 1996) neglects entrainment (qe/d=0), snow is considered a perfect fluid (Ka/p is set equal to 1) and resistance coefficient denote Coulomb dry friction and an hydraulic type friction independent from the flow height:

106 ⎛ n2 ⎞ ⎜ 2 ⎟ f1 = sign()u h⎜ μ g cosθ + u ⎟ (4.20) ⎝ h ⎠

The NIS model (Norem et al., 1987, 1989; Sovilla, 2004; Issler et al., 2005) is based on the comprehensive Criminale-Ericksen-Filbey rheology comprising cohesion and dispersive pressures, thus comprising cohesion and dispersive pressures, thus potentially encompassing visco-plastic as well as granular material behaviour. However the implementation currently used

neglects cohesion and the friction f1 is given by:

s u 2 9ν (u − u ) ∂(u − u ) f = g b cosθ + 0 − g 1 h 0 h 0 (4.21) 1 ρ h 4h 2 ∂x

Where s is the velocity squared friction coefficient, b is the dry friction coefficient, υ1 is the

normal stress viscosity and uh and u0 are the velocities at the top surface and at the base of the avalanche respectively. Another difference relative to VARA-1D is an earth pressure coefficient depending on the slope angle:

9ν ()u − u 2 K = 1+ 1 h 0 (4.22) a / p 8gh3cosθ

1D models with Some one-dimensional models also allow to take into account changes in the path width by variable path means of integration over the width. Integration of the balance equations for mass and width momentum over the flow height, h, and user-prescribed width, W, leads to evolution equations for h and the average velocity, u:

∂hW ∂hWu + = Wq (4.23) ∂t ∂x e / d

2 ∂hWu ∂(hWu ) ∂ 2 + = ghWsinθ − ()K gWh cosθ − A()f − f (4.24) ∂t ∂x ∂x a / p 1 2

Where A(x, t) is the cross-sectional flow area, given by A(x, t)=W(x) h(x, t).

An example of this model category is the Swiss model FL-1D (Bartelt et al., 1999; Sartoris and Bartelt, 2000), the AVAL-1D module that reproduce dense avalanches. The flowing friction is described, as in the VARA-1D model, using a Voellmy-fluid law (Voellmy, 1955):

107 g u 2 f = μ g cosθ + (4.25) 1 ξ h

The active /passive pressure coefficient Ka/p is governed by the relation:

Ka ⎫ 2 ⎛ θ ⎞ ⎬ = tan ⎜45° ± ⎟ (4.26) Kb ⎭ ⎝ 2 ⎠

Respect to analytical models (§ 4.2.3.2.1) one-dimensional hydraulic models give more information on avalanche characteristics such as impact pressures, velocities and snow depth as functions of space and time, and run out distances. The input preparation for simulations is very simple, as for analytical models one needs to define the avalanche track, the release snow depth and length, the friction parameters and, in the second type of 1D model described, the avalanche width in different sections along the track. A limitation typical of the use of 1D models is the lacking of information on the spreading of the avalanche in the run out area, that can be overcome simulating the avalanche along different trajectories in the stopping area with the support of historical registration of past events.

4.2.3.2.2.1 Two-dimensional hydraulic models

The typical formulation of a two-dimensional hydraulic type model for the simulation of dense avalanches with a Voellmy-fluid friction law is the following:

∂h ∂hu ∂hv + + = q (4.27) ∂t ∂x ∂x e / d

∂hu ∂ hu 2 ∂ huv ∂ ⎛ h 2 ⎞ ()() ⎜ ⎟ + + = − ⎜ K a / p cosθ x ⎟ + ∂t ∂x ∂y ∂x ⎝ 2 ⎠ ⎛ ⎞ ⎜ u ⎟ g 2 2 + gh sinθ x cosθ y − μ cosθ x − u u + v (4.28) ⎜ 2 2 ⎟ ⎝ u − v ⎠ ξ

∂hu ∂ hv 2 ∂ huv ∂ ⎛ h 2 ⎞ () () ⎜ ⎟ + + = − ⎜ K a / p cosθ x ⎟ + ∂t ∂y ∂x ∂y ⎝ 2 ⎠ ⎛ ⎞ ⎜ v ⎟ g 2 2 + gh sinθ x sinθ y − μ cosθ x − v u + v (4.29) ⎜ 2 2 ⎟ ⎝ u − v ⎠ ξ

Where u and v are the components of avalanche velocity in x and y directions.

The two 2D models for snow avalanche simulations currently used for hazard mapping in Europe are MN2L (Naaim et al., 2003) and SAMOS (Sampl and Zwinger, 2004) models. Even

108 if only the general formulation of the equation describing the dense flow are given here, these two models try to describe all the possible component of snow avalanches, dense and powder.

MN2L model comprises a depth-averaged, two-dimensional dense flow model and a three dimensional two-phase (air and particles) model for the powder layer, also with the transition layer collapsed to an interface. The lowest layer (dense part), most concentrated, is considered as a granular flow. According to the shear rate and the Froude number, the flow is described either by the Mohr-Coulomb frictional model, or by an inertial model resulting from the kinetic theory applied to granular material (Bonamy et al., 2002; Bouchet et al., 2004). The second layer (powder part), of weak concentration, accounts for the development of the aerosol; the flow is turbulent diphasic there and the energy dissipation is dominated by the turbulence of the interstitial fluid (Chen and Wood, 1985). The initiation and the development of the aerosol result from the erosion at the top of the dense layer. This exchange happens across a transition layer and the interface between the two layers is treated by a model resulting from the saltation theory.

In SAMOS the dense flow layer of the avalanche is assumed to behave like a quasi-statically deforming pile of granular material, sliding on a thin shear-layer at its bottom (Savage and Hutter, 1989). Depending on the predominant flow regime (“quasi-static” regime described by a dry Coulombian friction law, or “grain inertia” regime described by Bagnold, 1954) in the boundary condition at the bottom friction can depend upon the square of the strain rate or on the square of the velocity (Sailer, 2003). In the powder-snow or suspension layer, The (Favre-averaged) Navier-Stokes equations are solved with a particle transport equation together with a transport equation for particle volume concentration c, accounting for turbulent diffusion and gravitational settling. Closure of the equations is achieved with the help of the k-ε turbulence model. The two layers are coupled through a “transition model”, which prescribes the transfer of snow mass between the two parts. It is assumed that the logarithmic “law of the wall” is equally applicable to particle diffusion across the interface as to the turbulent momentum flux (Zwinger, 2000).

These models have the potentiality to give a detailed description of avalanche flow behaviour, but their complexity and the scarce available field observations are not enough to calibrate all the parameters required by them.

4.3 Hazard Map Types

Depending on map content and methods used in data collection and data processing it is possible

to distinguish between two types of hazard maps:

Hazard registration maps. Maps containing historically known avalanches, compiled from literature, documents, interviews and by field investigations and photo interpretation.

Hazard zoning maps. Maps which define risk areas compiled on the basis of known historic events, geomorphological investigations and the use of frequency/runout calculation models. The hazard zones should correspond to the safety requirements in the national building regulations, or specify other frequency/magnitude conditions of the hazard zone.

Hazard These maps, usually at a scale of 1:25000 or 1:10000, are informative documents that sum up

109 registration maps synthetically the information concerning an avalanche path, recording the maximum extension of events occurred on the site without any reference to event frequency or intensity and related risk. The defence works of each site are drawn up but only to describe the actual situation of the site, without indications on their efficiency or maintenance.

The main objective of these maps is to create an inventory of historical events, and preserve the memory of past avalanches synthetically. They don’t give any indication concerning avalanche intensity or frequency and can’t be directly used for planning purposes since they don’t allow to distinguish between areas subject to different degrees of hazard.

Examples of these maps are EPA and CLPA in France (Figure 4-11), CLPV in Italy (Figure 4-12) and the map of the cadastre in Switzerland (§ 4.4).

These maps reported the envelope of event areas, by two independent methods:

- an interpretation of stereo-aerial photographs, to find avalanche marks on vegetation and relief (see § 4.2.1.2). All the obtained information are printed in orange on the French CLPA and Italian CLPV.

- An analysis of archived newspapers, photographs and technical reports as well as witness testimonies: a local survey is carried out in the field with the help of witnesses (forest rangers, ski patrollers, road workers, shepherds,…) who indicate the maximum boundaries of avalanches they saw or heard about, and the affected area. This survey takes place in summertime so that people responsible of these maps can link more easily their statements with the terrain features observed on the aerial pictures and printed on the contour maps. The definitive information put on the maps is the envelope of the different reported phenomena concerning the same site. This information is printed in magenta (Figures 4-11, 4-12), all the permanent control works are printed black.

Both in France and Italy the boundaries of the different events known on an avalanche site are not distinguished: only the global envelope of all the events is drawn. These maps do not give indication of the potential risk evaluated considering the return period and the power of avalanche.

110 Figure 4-11:

Example of the past avalanches map. It reports witness testimonies in dark gray and photo- interpretation in lighter gray (respectively magenta and orange in original colored map)

Figure 4-12:

Example of CLPV. The maximum avalanche expansion on each site is drawn, in violet, from the analysis of historical documentation (§ 4.2.1.1) and, in orange, from photo interpretation (§ 4.2.1.2).

Hazard zoning In hazard maps the areas endangered by avalanches are divided into different hazard degrees, maps usually “high”, “moderate” and “low”, depending on avalanche frequency and/or intensity (§ 4.1). Since they are essentially used for planning purposes, these maps are at a more detailed scale (1:5000, 1:2000, 1:1000) rather than the inventory maps. According to the hazard zoning, regulations can be derived to establish construction criteria for new buildings, decide whether allow new settlements, plan countermeasures for hazard reduction and emergency measures in the endangered areas.

The creation of these maps implies the analysis of all the available data on site features and history (§ 4.2.1), field investigations and often the use of simulation models for the calculation of avalanche intensity in the run-out area (§ 4.4.3) for events of assigned return period (Figure 4- 13).

Depending on the local legislation hazard maps differs from country to country, both in the

111 definition of what is to be considered as an high, moderate or low hazard, and in the criteria adopted to define it. In Switzerland these maps are created following the “Directives pour la prise en considération du danger d’avalanches lors de l’exercise d’activités touchant l’organisation du territorie” (Office Fédéral des Forets, 1984) which report how to define different hazard levels using dynamical simulation models. In France the current procedure is called PPR (Plans de prévention des risques naturels prévisibles) while in Italy there is not a national regulation concerning the criteria for hazard maps and the “Linee guida metodologiche per la perimetrazione delle aree esposte al pericolo di valanghe” (AINEVA, 2006) constitute a guide for the creation of these maps following the Swiss example.

Source of uncertainties in hazard mapping are strictly connected to the ones of the data used to create them. Particular attention is to be paid to the analisys of nivometereological data as an input for dynamical models, for the determination of the “frequency-magnitude” relationship (§ 4.2.2) and to the choice of the model parameters. In current practice one often has to face the problem of scarce historical documentation of past events and, despite of calibration tables or laws for the dynamical models (Salm et al. 1990; Sovilla, 2002; Barbolini et al., 2004), the degree of subjectivity in the choice of the parameters to adopt in the simulation still remains.

Figure 4-13:

Scheme of the different phase in the creation of an hazard map.

Traceability Risk management processes implies several steps such as hazard assessment and mapping, exposure analysis for inhabited zones, and risk zoning. All these steps require expert analysis based on available methods and data. From one area to another, data availability and quality can be very different.

It is therefore quite difficult to take into account all the context that led to those specific studies. All the people who have to use the produced documents are interested in knowing more about

112 their implementation conditions. Traceability is a way to share and trace the information related to natural hazards. Two kinds of information have to be saved: what kind of data was used to produce a document ? What are the hypothesis, methods and reasoning process who led to the final result (an hazard zoning map, a risk map...)?

In France, for instance, a study was carried out to analyze the possibilities of improving the traceability of data and reasoning processes through the French risk prevention plans. On the 2nd January 1995, a law was created that outlined the Risk Prevention Plans. This still remains the main non-structural instrument for the State to protect from any natural hazards. Most of the time, it is based on existing information and studies. No specific hazard modelling should be done except in very critical situations where the nature of the stakes requires a high level of protection.

The zoning map and the reglementary rules are the two main documents, which compose the risk prevention plan. The zoning map is made by the technical service responsible for managing the administrative procedure. The zoning map shows the limits of the zones where can be applied:

• prohibitions;

• homogenous reglementary rules; and

• protection and prevention measures.

These different kinds of measures are based on land-use and town and country decisions. Two types of zones are identified:

• red zones where any buildings are prohibited;

• blue zones where building is possible only through specific instructions;

In addition to these maps, the Risk Prevention Plans should include and explicit clearly the method used to characterize the phenomenon and to establish the zoning map.

This determination is often the result and a combination of some expert judgments. It is therefore quite difficult to identify clearly the criterions and the process used to fix the different limits of the zoning map. Two kinds of weaknesses may therefore be pointed out: it is difficult to update the existing prevention plans and the population do not easily understand and accept the zoning maps. Traceability implies to be able to describe both data and reasoning processes. What are the primary data? In which reasoning processes are they used for? How can we build and share a common terminology? Metadata (according to ISO 19115 extended to some genealogy point of view) appears as an interesting possibility to establish the linkage between data and results and compare global processes from one case to another.

113

Annex A

LEGISLATION

A.1 - Debris flows

Introduction Serving as a prerequisite for planning and implementing protection measures as well as a steering instrument in safety planning, hazard zone maps have to be based on clear defined legal acts. Below you can find an overview of legislation of selected member countries of the IRASMOS project.

A.1.1 Austria

In Austria, among others,

• the Forest Act1 and

• the Water Act2

are the most important laws acting as legal framework for hazard zone mapping.

The implementation of these laws is assigned to the Federal Ministry of Agriculture, Forestry, Environment and Water Management (BMLFUW) and administrated by the government departments of the Austrian Service for Torrent and Avalanche Control (WLV)3 and the Federal Water Engineering Administration.

Legal framework The Forest Act (§ 8b) of 1975 prescribes, that Austria’s catchment areas have to be delimited according to their endangerment against natural hazards like torrential floods or avalanches Forest Act (§ 99), and, as well as zones reserved for further

1 BGBl 1975/440 idF BGBl I 2005/87 2 WBFG (1985) 3 ForstG § 11 Abs. 1

114 measures have to be distinguished in hazard zone maps. Forest Act (§ 11) governs the generation of hazard maps and the involvement of communes and population. The 4 content and design of those maps are specified by a Decree of the Federal Minister .

According to § 5 (2) of the Ordinance on Hazard Zoning, not only all available data, information and interactions in between natural hazards have to be considered for developing hazard zone maps, but also man-made influences like infrastructure devices and settlement activities have to be taken into account.

Hazard maps usually cover the area of one community or only certain parts, if these are needed for the implementation of mitigation measures. Hazard maps should be comparable due to a preliminary survey of the Federal Ministry during the approval process.

Cartographic part The cartographic part consists of a map showing the relevant hazardous processes and another map showing the delimited hazard zones. The map with the hazardous processes can be a topographical map or an aerial view (scale is 1:50.000 or 1:25.000). The delimited hazard zones are visualized by a cadastral map with a scale of 1:5.000.

At the base of a design event with a return period of 150 years and a frequent event wit a return period of 10 years, § 6 of the Ordinance of Hazard Zoning5 prescribes the criteria for delimitation of hazard zones. According to these regulations, red (a permanent use for settlement and traffic purposes is not possible or only possible with unproportionally high effort) and yellow (a permanent use is impaired) hazard zones and specific reserved areas have to be displayed in the hazard maps. Blue reservation areas are reserved for future mitigation measures, brown reference areas inform about other natural hazards like landslides and rock fall (magnitude and frequency are not assessed) and purple reference areas have to be preserved as areas which act already as a natural protection due to their character.

Criteria for delimitating hazard zones according to torrential hazards are shown in Table A.1-1.

Criteria Zones Design event Frequent event Table A1.1:

Stagnant water TR waterdepth ≥ 1,5 m water level mark HQ10 > 0,5 Specifications for the design HQ 02 TY waterdepth < 1,5 m water level mark HQ10 < 0,5 event (150 years)and a mHQ<02m frequent event (10 Running water TR height of energyline ≥ 1,5 m HQ10: height of energyline ≥ years)with respect to torrent 025m TY height of energyline < 1,5 m HQ : height of energyline < related processes 10 025m Erosion gullies TR depth ≥ 1,5 m erosion gullies possible TY depth < 1,5 m runoff without erosion gullies see running water“ Bedload deposits TR height of deposits ≥ 0,7 m bedload deposits possible

4 Decree of the Federal Minister; Verordnung des Bundesministers für Land- und Forstwirtschaft vom 30. Juli 1976 über die Gefahrenzonenpläne (GefahrenzonenplanVO), BGBl 1976/436 5 GefahrenzonenplanVO § 6

115 Bedload deposits TR height of deposits ≥ 0,7 m bedload deposits possible TY height of deposits < 0,7 m no bedload deposits, see „running water“ Post failure slope TR top edge of the post-failure - movement due to slope movement epth/lateral erosion TY safety area Debris and earth flows TR boundary of distinct debris - flow deposits TY Retrogressive erosion TR possible extent no assessment TY Consider criteria of „erosion gullies“ and „post failure slope movement” Written part Base of the written part is a detailed listing of the catchment’s properties like, among others, catchment area, geology, vegetation, hydrology, discharge and potential bedload capacity.

The written part comments on the cartographic part of the hazard map. The process of delimiting hazard zones and the justification of the assessment is included as well as all data (maps, aerial photographs, references, reports, chronicles …) used for the hazard zoning. Further recommendations are given considering spatial planning activities, construction requirements and safety purposes. This can lead to restrictions for the yellow hazard zone.

Hazard zone maps, developed by the Austrian Service for Torrent and Avalanche Control, have to be reviewed by the Federal Ministry of Agriculture, Forestry, Environment and Water Management (BMLFUW). After this review, the hazard zone maps are open for the public for at least four weeks. Persons concerned can hand in objections by a written statement. After expiration of this period, a committee, respectively consisting of a representative of the Federal Ministry of Agriculture, Forestry, Environment and Water Management, of the province, of the municipality and of the Austrian Service for Torrent and Avalanche Control will proof the draft. During an on-site inspection the objections of persons concerned are considered. After agreement, the approved hazard zone maps are to be open for public (Forest Act§ 102) at the Austrian Service for Torrent and Avalanche Control, the district administration and the municipality.

Interpreting the Austrian law, hazard zone maps don’t have any ordinance character (KHAKZADEH 2004). Due to this fact, no commandment, prohibition or permission for the citizens can be deduced. Summarized, Austria’s hazard maps can be described as an expert opinion with a prognostic character (ZOPP 2004).

A.1.2 Switzerland

Legal acts Due to the Swiss federal system, the implementation of hazard zoning differs considerable at cantonal level according to the recommendations of the federal government.

The federal law considering natural hazards are the base for cantonal law. At cantonal level the federal law is modified taking local circumstances into account. This

116 procedure assures the legislative autonomy of the Swiss cantons.

Natural hazards are respected in the spatial planning act, building act, hydraulic engineering act and the forest act. Further building insurances deal with natural hazards.

In former times the „Bundesgesetz betreffend die eidgenössische Oberaufsicht über die Forstpolizei (FpolG)“ has been the exclusive act for hazard zone mapping. Meanwhile the following acts are decisive for hazard zoning in Switzerland:

• Bundesgesetz über den Wald6 (WAG)

• Bundesgesetz über den Wasserbau7

• Bundesgesetz über die Raumplanung8 (RPG)

• Verordnung über den Wald9 (WaV)

In Switzerland two types of natural hazard related maps are differentiated:

• Hazard indication map (1:10.000 to 1:50.000)

• Hazard map (1:2.000 to 1:10.000)

The hazard indication map serves as a base for planning and recognition of conflict of interests. It shows an approximate overview over the situation and the type of hazards without quantification for single regions or cantons.

The hazard map is the base for land use planning and the planning of mitigation measures against natural hazards. It lists detailed information to the hazard type, the spatial extent and the intensity of the hazard with regard to already settled or future settled areas as well as for infrastructure facilities. The delimitation occurs at the level of parcels and is done by cantons and communities.

Hazard zones The Swiss system distinguishes four hazard zones (Table A1-2-1):

6 http://www.admin.ch/ch/d/sr/9/921.0.de.pdf 7 http://www.admin.ch/ch/d/sr/7/721.100.de.pdf 8 http://www.admin.ch/ch/d/sr/7/700.de.pdf 9 http://www.admin.ch/ch/d/sr/9/921.01.de.pdf

117 Table A1-2: Zone Description People are endangered inside and outside of buildings. Sudden failure of buildings is Definition of Swiss hazard Area of possible. zones (BUWAL, BRP & prohibition Frequent events occur with lower intensity. People are endangered especially outside of BWW, 1997) buildings. People are endangered especially outside of buildings, inside they are quite save. Area of Damage of buildings is possible; failure of the building will be improbably if legal precept structural restraints have been considered. Area of People are rarely endangered. Marginal damage to buildings and interferences are instruction possible. Inside of the buildings considerable tangible damages can occur. Area of Occurrence of tangible endangerment with high potential of damage (very low instruction probability, high intensity) or compared to nowadays an obvious increase of probability.

The degree of exposure is defined by the intensity and the return period of a hazardous process. These two parameters are combined to different hazard levels by the aim of a matrix (Figure A1-1)

Figure A1-1:

Derivation of hazard zones from probability and intensity of the hazardous process (BUWAL, BRP & BWW, 1997)

Delimitation of hazard The delimiting conditions for the discrimination of the different intensity levels have zones been chosen in such a manner, that comparable transport processes (avalanche, flood and debris flow) will result in comparable stress and endangerment (RICKENMANN, 2001). The intensity of a potentially damaging process can be classified into three categories (Table A1-2-3):

Table A1-3:

Classification of intensity for damaging processes (BUWAL, BRP & BWW, 1997)

Critical values have been derived for settlement areas but they can be derived analogous for other land use (Table A1-2-4).

118 Table A1-4:

Criteria for the intensity of a process for settlement areas (BUWAL, BRP & BWW, 1997)

Using different return periods, the semi-quantitative expressed probability is classified into three classes (Table A1-2-5):

Table A1-5:

Definition of the probability classes (BUWAL, BRP & BWW, 1997)

Defining a certain period (expected useful life, wear lifespan), the probability of occurrence can be estimated by use of this equation:

n ⎛ 1 ⎞ p = 1− ⎜1− ⎟ ⎝ T ⎠

p = probability

T = annuality, recurrence period

n = period under review (e.g. wear lifespan)

Implementation and Swiss tools for spatial planning are the “master plan” and derived from this the “land legal obligation use plan”.

According to the act of spatial planning10, in this “master plan” the Swiss cantons have to define areas, which are endangered by natural hazards or other destructive impacts. It is the canton’s jurisdiction to derive the “master plan”, which is legal obligate. First of all, according to RPG 1979 Art 14, areas for building, agriculture and protection are delimited and described. Further, areas of different use can be defined (RPG 1979 Art 18), which can interfere with hazard zones (“positive planning”). Deriving the areas of use from the hazard zones is the so called “negative planning”. Thus, the integration and the level of legal obligation of the hazard (indication) maps are determined by the

10 Bundesgesetz über die Raumplanung (RPG 1979 Art 6)

119 cantons.

No standardized critical values are prescribed for the federal territory, such announcements have the character of recommendations.

At the level of land use planning the grade of ascertainment and legal obligation has to be defined in such a way, that an adequate consideration of natural hazards is guaranteed. The land use plan is generally binding.

Basically an appreciation of interests has to be performed, whereas it is implausible, that other interests are more grave than objective delimitated hazard zones. The hazard map serves as a base for delimitating the hazard zones.

A.2 - Rock avalanches

Experiences in Europe In Europe, catastrophic rock avalanches are generally perceived as rare events. The and overseas estimated probabilities of occurrence often are within the range of residual and acceptable risk, and at the European level little mention exists in legislatory frameworks about mitigating these perceived risk levels.

However, in mountainous countries affected more frequently by rock avalanches, such as e.g. Canada, Taiwan, or New Zealand, awareness of these phenomena has led to their explicit mentioning, if not consideration, in regulatory aspects of hazard and risk mitigation, moreso in areas of active volcanism, where catastrophic debris avalanches are a reccurring hazard.

Figure A2-1:

Example of a susceptibility/hazard map for volcanic debris avalanches, Iliamna Volcano, Alaska.

Southern Alps, New The tectonically active Southern Alps of New Zealand is a region, where numerous

120 Zealand rock avalanches have been found to occur throughout at least the last 9000 years (Whitehouse and Griffiths, 1983). The occurrence of four rock avalanches in these mountains in the 1990s near areas populated or frequented by international tourism alone has led to a much better appreciation of the adverse consequences involved. For instance, the West Coast Emergency Management Plan issued by the local administrative authority, explicitly recognises rock avalanches as well as landslide dams as their possible and highly likely consequences.

Figure A2.2:

Extract from the West Coast Emergency Plan issued by the West Coast Regional Council, i.e. the major administrative authority in charge for natural hazards mitigation in the western Southern Alps, New Zealand. Note the explicit treatment of the term “rock avalanche”, which is a frequent process in this tectonically active mountain belt.

Swiss Alps Although hazard zonation and mapping guidelines in Switzerland do recognise rock avalanches as a potentially occurring process, the existing classification subsumes rock avalanches as (Gross-)Rutschungen or “large and deep-seated landslides”. According to the methodology for zoning hazards from landslides, rock avalanches need to be treated as “high intensity” processes, given that their detachment depth would in most cases easily exceed 2 m as the critical threshold. The same applies for

121 rock avalanches derived from a precursory rock fall phase, where impact energies are mostly > 300 kJ as the critical value for “high intensity” processes. Notably, the Swiss approach differentiates between rock falls and landslides, and hence rock fall/rock avalanches might fall into both categories.

A.3 - Snow avalanches

Introduction In the following the current legislation in European countries affected by snow avalanches is reported. Legislation is very important if one wants to understand the differences between the tools used in the different countries for hazard mapping, since hazard mapping procedures, hazard tools and hazard map types are used and done in legislative surroundings which varies as function of countries.

A.3.1 Austria

Legal framework The legal basis of dealing with natural hazards at a federal level is provided by the federal law related to forests (ForstG1), where article II related to forestal spatial planning regulates the hazard mapping in principal. The implementation of this law is assigned to the Federal Ministry of Agriculture, Forestry, Environment and Water Management (BMLFUW) and administrated by the government departments of the Torrent and Avalanche Control Service (WLV2) . In § 8 of the ForstG, the hazard map is defined as forestal development plan; in § 11, the administrative procedure of generating the hazard map and the involvement of communes and population is specified. Further provisions related to the map content and design are determined by the ordinance on the hazard maps3, as well as regulations on which spatial entities hazard map have to be compiled, and which content is needed for the cartographic part and text component of the hazard map. According to the ordinance, hazard maps form the basis for: (i) the development and implementation of mitigation measures and the prioritisation of those measures according to their urgency and for (ii) planning issues related to spatial planning activities, civil engineering and national safety purpose4. The overall aim of hazard mapping is the delimitation of areas endangered by torrent processes and avalanches, the degree of hazard and the delimitation of those areas to be reserved for mitigation purpose.

On the state level, associated statutory sources can be found in all federal states of Austria with the exception of Vienna. The hazard map is generally based on the community area, or on certain parts of a community if appropriable due to the size of the community or if this is a particular need for the implementation of protection measures5. Due to a preliminary survey of the Federal Ministry during the approval process, all hazard maps should become comparable.

Cartographic part The zoning is prescribed within the ordinance on the hazard maps6. According to those regulations, red and yellow hazard zones and specific reserved areas and areas for information have to be delimitated in the hazard maps. If existing, red and yellow hazard zones and blue reserved areas have to be displayed, whereas the delineation of brown and purple areas for information is optional. Thereby the outline of those areas is only based on technical criterions and not on the outline of the land plots.

122 The red hazard zone includes all areas endangered by torrents and avalanches in a degree that a permanent use for settlement and traffic purpose is not possible due to the estimated damage resulting from the design event; or where the use would only be possible with disproportional subsidies by the public sector in protection measures7.

The yellow hazard zone includes all areas influenced by the effects of torrents and avalanches if used for settlement or traffic purpose8.

Blue areas to be reserved include all those areas (i) needed for the implementation of technical or forestall protective measures or the maintenance of such areas and (ii) needed for a certain form of land management due to an implemented mitigation measure9.

Brown areas include all areas affected by natural hazards apart from torrent processes and avalanches, such as rock fall or landslides10. Purple indicates those areas where the protective function is dependent on the conservation of soil or terrain properties11.

The basis of the red and yellow hazard zones is given by the design event, which is an event with a recurrence interval of approximately 150 years12. However, with respect to avalanches, this design event is specified further within the instructions of the Ministry to the subordinated government departments of the Torrent and Avalanche Control Service (Table A3-1).

Table A3-1: Criterion Hazard zone Design event T=150 yrs Pressure (P) [KN/m2] red P > 10

Specifications for the design yellow 1 < P < 10 event with respect to snow Deposition height (H) 13 red H > 1.5 avalanches [m] yellow 0.2 < H < 1.5 According to the legal regulations, hazard maps are evaluated and refereed reports of the summary outline of all possible design events14 and therefore have been judged by the Austrian administrative tribunal as qualified report with a prognostic character15. Thus, the consideration of hazard maps in all planning activities by communes and citizens cannot be obligatorily enforced. Therefore, the BMLFUW introduced the so- called estoppel, in which a non-consideration of hazard maps will exclude statutory corporations from governmental subsidies for protective measures against torrent and avalanche hazards16.

Written part The text component completes and comments on the cartographic part of the hazard map and therefore ensures the comprehensibility of the hazard plan. Correspondingly, all map bases and data used are included, as well as a justification of the assessment and depiction of endangered areas, and special notes for spatial planning activities, construction and safety purposes. The data to be considered include maps, outlines, aerial photographs, references used and reports taken into account during the procedure of hazard mapping. Furthermore, descriptions of all areas with respect to avalanches and torrent processes, and the related catchment characteristics such as area, discharge and potential bed load during a design event have to be described in the written part of the hazard plan. This description has to be carried out using a

123 standardised official form. Additionally, for every catchment area an inspection sheet has to be compiled, including all observations relevant with respect to an assessment of those watersheds, such as possible sources of bed load. The written part of the hazard assessment is completed with suggestions of how to incorporate the mapping results in the local development planning and with potential construction requirements or restrictions for the yellow hazard zone.

1 BGBl 1975/440 idF BGBl I 2005/87 2 ForstG § 11 Abs. 1 3 Decree of the Federal Minister; Verordnung des Bundesministers für Land- und Forstwirtschaft vom 30. Juli 1976 über die Gefahrenzonenpläne (GefahrenzonenplanVO), BGBl 1976/436 4 GefahrenzonenplanVO § 1 5 GefahrenzonenplanVO § 3 6 GefahrenzonenplanVO § 6 7 GefahrenzonenplanVO § 6 lit a 8 GefahrenzonenplanVO § 6 lit b 9 GefahrenzonenplanVO § 6 lit c 10 GefahrenzonenplanVO § 7 lit a 11 GefahrenzonenplanVO § 7 lit b 12 GefahrenzonenplanVO § 6 13 BMLF, Decret 52.240/10-VC6a/99 idF BMLFUW 03/2001 14 BMLF, Richtlinien für die Gefahrenzonenabgrenzung (1989) 15 VwGH 27.03.1995, 91/10/0090 16 Länger, Geschichtliche Entwicklung der Gefahrenzonenplanung in Österreich, Wildbach- und Lawinenverbau 152/2005, 13 A.3.2 France In 1995, a law was created that outlined the Risk Prevention Map (PPR in French), which is a main non-structural instrument for the state to protect from natural hazards. This mapping at the local scale shows the exposed zones (hazard and risk map) and prescribes the prevention actions (report) to be implemented both by private owners and/or public organizations. This is a basic risks map at the local scale, to be shifted into the statutory land-use map (POS). To carry out the PPR documents, several methodological guides have been elaborated under the responsibility of the Public Authority. The first one is a general guide, giving a methodology to the operator, in order to draw up the mapping and to establish the attached report including prescriptions. Next ones are more specific to each natural hazards. For avalanche, this guide (http://www.prim.net/professionnel/documentation/guide_avalanche/page01.ht ml), has been drawn up by RTM (office for the conservation and protection of the mountain-areas : Services de Restauration des Terrains en Montagne in French). And a part of this guide is dedicated to the hazard zoning. The reference avalanche event is characterized by a return period T=100 years (century hazard), three levels of increasing hazard are defined: low hazard (rose zone on the map) where P<1 kPa, moderate hazard (orange zone on the map) where 1 kPa< P< 30 kPa and high hazard where P≥30 kPa (violet zone on the map). Hazards map doesn’t take into account vulnerability of buildings or people. Nevertheless it is advisable to differentiate between building or infrastructure safety and people safety. If it is acceptable from an economic point of view that damage may occur to the buildings for extreme events (greater than centennial) it is no more the case if we consider people in mortal danger outside the buildings. That ‘s why the methodological guide introduces for the same avalanche path a second reference event which is called maximum likely event (AMV for avalanche maximale vraisemblable in French) (yellow zone on the map). It is the maximum spatial variation which can be imagined. This second reference avalanche differs from the maximum historical

124 event. If it is calculated thanks to a numerical model, the chosen return period is three centuries. Such reference event allows to determine an alerting domain highlighting a residual danger. At the present time, there is no general agreed opinion on AMV : mayors, which are responsible of emergency measures in the endangered areas, are strongly affected by this zoning. They consider that national government must make clear the concept and his mode of enforcement. Another zone, the green zone, appears on the hazard maps : it concerns forests which have protective effects in a potential avalanche release area. The release area could be detected thanks to topographic criteria and using a DTM (digital terrain model). There is no standardized method to establish hazard maps for avalanche. But general rules are described in the methodological guide. Assessment of reference events is mainly based on data collection (EPA, CLPA, surveys,…) and analysis, field investigations and analogy with well-known similar paths. In France simulation models for the calculation of run-out distance are not used apart from complex avalanche paths with high vulnerability. In this case the nivometeorological input parameter of the dynamic model is the snow depth with a return period of one century. An other important point is that effects of permanent protective measures (such as dam, mounds, snow rakes…), which are drawn on hazard maps, are not taken into account in hazard assessment.

A.3.3 Italy

Legal Framework In Italy a national regulation concerning avalanche hazard mapping is missing and Alpine regions produce their own laws to take snow avalanches into account in land use planning procedures. Recently, as a result of a collaboration project between AINEVA (Italian Association for Snow and Avalanche) and the University of Pavia, new “Guidelines for Avalanche Hazard Mapping” (Barbolini et al., 2005b) have been published. This document proposes criteria and methods to proceed in the different phases of territory planning in avalanche prone areas and it represents a reference for the Italian public administrations which operate in mountainous regions. The mapping criteria contained in this document regard exclusively urban settlements; traffic roads, electric systems, ski and chair lifts, etc. are excluded. In the following the mapping criteria and some relevant aspects of these "Guidelines" are briefly presented.

Cartographic part Three zones with decreasing degree of hazard are defined on the base of avalanche return period and avalanche impact pressure (Table A3-2).

Red hazard zones include those areas that can be reached relatively frequently by avalanches, even with moderate destructive power, or by rare events with high destructive potential.

Blue hazard zones include areas reached by residuals effects of relatively frequent avalanches, or rarely by avalanches with moderate destructive power.

Yellow hazard areas are those areas that can be reached by the residual effects of rare events. Also areas reached by extreme events belong to yellow hazard zone.

Table A3-2: Hazard Hazard zone Specifications for the design level

125 event with respect to snow High Red T=30 yrs and P>3 kPa or T =100 yrs and P>15 kPa avalanches (Barbolini et al., 2005). T is the return period(years) and P the Moderate Blue T=30 yrs and P<3 kPa or T =100 yrs and 3

Some important aspects of these new guidelines that should be highlighted are the following:

In the calculation of avalanche impact pressure also the static component should be accounted for; that means that mapping is also based on the deposition depth.

Monitoring and evacuation plans must be prepared to increase safety of people living in red, blue and yellow zones.

Modification of hazard maps after realisation of defence works (or reforestation) is allowed, under certain conditions (no new white areas, that means yellow boundary unmodified, but only modification of the boundary between red/blue and blue/yellow zones; maintenance plan for the defence works must be operative; etc.)

Maps upgrades and Some general criteria for the updating of hazard maps is also suggested. On the base modification of new information hazard maps have to be checked and, if necessary, modified. Particular attention has to be paid to:

1. historical documentation of past events, not already used in the creation of the hazard map and that can modify it;

2. new avalanche events (that is avalanches in areas not registered in the avalanche cadastre) or know events with unexpected extraordinary magnitude;

3. natural or artificial territory modifications that increase the exposition factor (e.g. deforestation of the release area). Hazard map updating also has to take into account for up-to-date nivometereological data and the development of new calculation tools.

Also factors that can decrease hazard levels have to be considered, for example natural reforestation of the release area or the construction of protection works in the release and/or runout areas. In that case, in order to maintain a control on the areas potentially exposed to "extreme" avalanche action, the up-to-date hazard map should consist only in a re-classification of hazard levels, maintaining the total surface of the mapped territory (that is the yellow area boundary) unchanged. The updating process should be done making use of expert evaluations which:

1. evaluate the effect of reforestation on the release and motion of snow mass depending on the type, density and age of the vegetation and its exposure to risk factors that can reduce its efficacy;

2. evaluate how the protection works modify release conditions and the avalanche dynamics. The possibility of hazard re-classification in the valley bottom has to be evaluated considering the technical life of defence works.

126 Land use restriction According to the hazard level classifications land use limitations and specifics safety requirements are fixed for the different hazard zones.

Due to the high level of hazard associated to the red zone, in these areas the construction of new buildings (both residential and commercial) is not allowed. In the blue (moderate hazard) zones, new constructions are allowed, but with "strong" restrictions (low building indexes, properly reinforced structures, etc.). In the yellow (low hazard) zones new constructions are allowed, with "minor" restrictions (public facilities, like schools and hotels, must be avoided).

A.3.4 Norway

Legal framework Legal demands on avalanche safety exist in pursuance of the Building and Planning Act, the Working Environment Act and the Act relating to the Regulation of Water Resources. The safety requirement for buildings varies according to the probable risk of personal injury. The Municipal Council is the competent authority. In construction employers are obliged to take precautions to prevent avalanche accidents.

The Norwegian Geotechnical Institute is the only institution which undertakes avalanche research and consulting in Norway. Our quality assurance system is intended to satisfy the demands of the International Standard ISO 9001.

Neither legal regulations nor other written documents can prevent liable persons from making mistakes in this field of work. However, damage caused by avalanches will normally be compensated in full. If a party has shown gross negligence, judicial practice shows that the defendant normally is found guilty of compensatory negligence. All acquired knowledge demonstrates that both human and economic consequences may be considerable if legal demands and quality policy are ignored.

The Building and Planning Act and the Working Environment Act both have demands concerning avalanche safety. Legal requirements were first established in the Building Act of 1924. The Act was put into force for the whole country in January 1966.

Instructions on avalanche safety are also established in pursuance of the Act Relating to the Regulation of Water Resources. On the other hand, the Road Act has no such demands concerning avalanche safety. However, this does not mean that avalanche problems are neglected in highway planning.

In accordance with the Building and Planning Act, development in hazardous areas can be avoided or prohibited at three different stages: At the municipal area planning level by tying up potential hazard areas, at the area regulation planning level by marking off hazardous areas and thirdly due to the Building Site Regulation by the specific safety standards.

The general clause of the Building Regulation states that:

"Buildings and directly adjacent external areas in use shall be situated, dimensioned and instructed so that there is reasonable safety against personal injury occurring because of such loads which may be foreseen.

127 ...... Buildings where a total collapse would cause serious or extremely serious risk of personal injury shall be constructed or situated so that accidental action in a small part of the building will not lead to an extensive collapse."

The safety requirement of the different types of buildings and their outside areas varies according to the probable risk of personal injury. There are three safety classes (Table A3-3).

Table A3-3: Safety Consequences of Highest nominal Safety requirements for the Categories of buildings class structural failure annual probability location of buildings of natural hazards 1 Less serious 10-2 - Garage for max. 2 cars, boat houses etc. - Storage sheds occasionally in use - Halls of plastic-based fabrics - Agricultural buildings etc., if frequently used class 2 or 3 2 Serious 10-3 - Buildings not exceeding two storeys of moderate span and in normal use - Industrial and storage buildings of one storey not accessible to general public, with ≤ 5 persons per 100 m2. Distance to other buildings, roads etc. ≥ height of façade - Tall masts, independent towers, silos and chimneys outside of built up areas 3 Extremely serious < 103 - Buildings not included in class 1 and 2 The tolerated nominal annual probability for hazards to buildings in safety class 3 (< 10-3) should be decided on according to the stipulated total risk due to natural hazard. The higher the consequences the lower the probability of hazard should be allowed. The Municipal Building Committees shall approve the highest nominal annual probability of hazard in these cases.

Buildings of safety class 2 and 3, which already exist within hazardous areas, may be rehabilitated or rebuilt. However, the highest nominal annual probability of hazards should not exceed 3 x 10-3 in class 2 and 10-3 in class 3.

As indicated in the general clause, buildings and their outside areas may also be dimensioned or otherwise secured so that the specific safety standard is fulfilled.

128 The competent authority in the case of the Building and Planning Act is the Municipal Council. The Municipal Building Committee recommends the plans. The County Administration is obliged to give guidance.

In the official guideline to the Building Regulation, the local authorities are advised to cooperate with the Norwegian Geotechnical Institute to elucidate natural hazards in all area planning.

In case of application for a concession in accordance with the Water Resources Act there is a premise that the potential avalanche hazard of the actual areas shall be evaluated by an expert during the planning stage. At this stage the builder or his consultant is responsible for ensuring that the appropriate steps are taken. The evaluation document should be enclosed with the application.

In accordance with the Regulation relating to the Working Environment Act employers are obliged to take precautions to prevent avalanche accidents at all exposed locations in potentially hazardous areas. The avalanche risk to access roads, camp locations and construction sites has to be evaluated by an avalanche expert. The expert shall prescribe the necessary safety and preparedness measures, and work out an action plan and/or appropriate precautions to be followed in hazardous situations. The builder or main contractor is responsible for directing and coordinating the safety precautions.

Quality policy and The Norwegian Geotechnical Institute (NGI) is the only institution which undertakes liability avalanche research and consulting in Norway. NGI is a non-profit based foundation with approx 10% governmental funding for research.

NGI undertakes work according to:

- NGIs General Conditions (which specifies validity, man-hour compensation, rental of equipment, refundable expenses, payment, liability, adjustment of rates and possibility of separate contract.)

- NS 3480 "GEOTECHNICAL DESIGN Foundation, Earth and Rock Engineering"

- NS 8402 "General conditions of contract for consultancy commissions with remuneration on the basis of actual time taken”.

There is no specific Norwegian Standards, or proposals for such standards, in the field of avalanche zoning and protection. We therefore have to adjust principles from related fields of activity.

The intent of our quality assurance system is to satisfy the demands of the International Standard ISO 9001.

Particularly important in the process are:

- there should be no doubt (at any given time) as to who is responsible for what

- the professional work should be based on updated professional knowledge and

129 methods

- documentation of data, calculations and techniques should be accessible

- the written documents have to be concise and unambiguous

- the quality assurance system should be as comprehensive as needed to meet the quality objectives

Consultants are obliged to have liability insurance which, if possible, covers full responsibility according to the contract. Liability towards the client is normally limited for individual claims and for all claims in one and the same policy year. NGIs policy comprises liability under the laws in force which is incurred for damage anywhere in the world.

Insurance and disaster In January 1980 a new act became operative in Norway which states that all objects assistance with Fire Insurance are also obliged to take out Natural Hazard Insurance. Damages caused by avalanches will normally be compensated in full unless the client has shown gross negligence. However, insurance companies will neither initiate any hazard evaluation nor safety measures. They may - on the other hand - increase the insurance premium or refuse rebuilding.

Unfortunately, damage caused by avalanches does not automatically provoke evaluation of hazard and safety measures. The injured party will normally have to initiate settlement of the problem with the local authorities and/or the National Fund for Natural Disaster Assistance.

The local authorities or the National Fund for Natural Disaster Assistance will then contract NGI, the distinction of being the only institution undertaking avalanche consulting in Norway. Government and private enterprises, builders and contractors will contact NGI directly.

A.3.5 Switzerland

Legal framework In Switzerland there are three laws at the federal level which regulate land use and hazard mapping in the field of natural hazards.

- The federal land planning law (“Raumplanungsgesetz”, RPG, 1979) and the land planning regulation (“Raumplanungsverordnung”, RPV, 2000).

- The federal forest law (“Waldgesetz”, WaG, 1991) and the federal forest regulation (“Waldverordnung “, WaV, 1992).

- The federal water construction law (“Wasserbaugesetz”, WBG, 1991) and the federal water construction regulation (“Wasserbauverordnung”, WBV, 1994).

For land use planning in the field of avalanches only the land planning law and the forest law are of particular interest.

The regulation of land use in Switzerland is organized at three levels. At the federal

130 level concepts, object plans (“Sachplan”) and recommendations are developed which act as a leading guideline for planning at the cantonal level and the communal level (Art. 13, RPG and Art. 14ff., RPG). For considering the interaction between land use planning and natural hazards a recommendation was issued in 2005 (“Raumplanung und Naturgefahren”; ARE, BWG and BUWAL, 2005). For dealing with avalanche danger the “Richtlinien zur Berücksichtigung der Lawinengefahr bei raumwirksamen Tätigkeiten“ (BFF and SLF, 1984) is the most important guideline or recommendation for the consideration of avalanche danger at the federal and cantonal level.

At the cantonal level general guiding plans (“Richtpläne”) define those areas which are endangered by natural hazards (Art. 6, RPG). The guidelines are compulsory for authorities, they describe the state and the planned development of the landscape. They indicate those areas which are either endangered or not without specifying the level of danger in detail. During the development of these plans the cantonal authorities are supported by federal authorities. The cantonal guiding plan consists of a map in the scale of 1:25’000 to 1:50’000 and an accompanying technical report. It is revised approximately every 10 years.

At the communal level the land use plans (“Nutzungsplan”, Art. 14ff, RPG) are the basis for land use regulation. In a land use plan building areas, agricultural areas and protection areas are distinguished. Other areas like e.g. hazard areas can be separated and can overlay with the other three areas. The recommendations published in 1984 (BFF and SLF, 1984) are the basis for the development of hazard maps which indicate areas endangered by avalanches at different levels (the classification is presented in the next section). Hazard maps are the basis for land use plans at the communal level. They are mandatory for land owners and are approved by the cantonal authorities. Land use plans are revised every ten to fifteen years.

The building permission (“Baubewilligung”) at the communal level is the third legislation. The permission for building in the building zone is issued by the commune, for building outside the building zones the cantonal authorities have to agree. In an application for building it is checked whether the federal, the cantonal and the communal legislation is fulfilled.

In the federal forest law it is regulated that the cantons have to protect humans and high assets against natural hazards (Art, 19, WaG, 1991). In Art. 15 (1) of the federal forest regulation (WaV, 1992) it is prescribed that the cantons have to generate hazard maps and in Art. 15 (2) it is stated that they have to regard the recommendations issued by the federal authorities and institutions.

Classification and In a hazard map four levels of hazards are distinguished and displayed as red, blue, cartographic display yellow and white areas in the map. The parameters for classification are defined by the of hazard level impact pressure and return period of the expected avalanches (BFF and SLF, 1984).

An area is attributed to the red zone if:

Condition a) Avalanches with an impact pressure of ≥ 30 kPa can occur within a

131 return period of 300 years.

Condition b) Avalanches within an impact pressure of ≤ 30 kPa can occur within a return period of 30 years.

Only one of both conditions must be fulfilled in order an area is being attributed to the red zone.

An area is attributed to the blue zone if avalanches with an impact pressure ≤ 30 kPa can be expected within 30 to 300 years. Areas affected by powder avalanches with an impact pressure ≤ 3 kPa and a return period within 30 years can also be attributed to the blue zone.

An area can be attributed to the yellow zone (not mandatory) if powder avalanches within an impact pressure of ≤ 3 kPa can be expected with a return period of ≥ 30 years or for areas where very rare avalanches (return period > 300 years) or undocumented avalanches can be expected.

In a white area no avalanches have to be expected.

Existing hazard zones should be displayed in the object plans (“Sachpläne”), general guiding plans (“Richtpläne”) and land use plans (“Nutzungspläne”) as a basis for further planning at the respective level.

Figure A3-1: Classification of the hazard (red) zones for snow avalanches Condition (a) according to the Swiss guidelines (BFF and SLF, (red) 1984).

red blue yellow Condition (b)

Impact pressure avalanche kPa avalanche pressure Impact Red or blue or yellow blue

Return period in years

Practical application The hazard maps are the basis for land use planning at the communal level. The in building regulation cantonal authorities take care that hazard maps developed by a commune or by private consultants commissioned by a commune do regard the existing guidelines and that they are applied in practice according to legislation. In many this task is fulfilled by a cantonal hazard commission.

Putted in practice the following restrictions are formulated in the guidelines (BFF and SLF, 1984):

132 1. In red zones no building zones are allowed. Existing buildings (erected before a hazard map) may remain in the endangered area but the best possible protection have to be established. New buildings or reconstruction of buildings are only allowed in extraordinary cases if they are strong economic constraints. The buildings have to be equipped with direct object protection measures which can withstand the extreme forces as far as possible. Alarm and evacuation plans have to be established. In every case experts have to be consulted in this process. Experiences from the cantons after the avalanche winter in 1999 showed that allowance of reconstruction in the red zone is very low and handled very restrictive.

2. In the blue zones the creation of building zones should be promoted very carefully. In any case if there is land available in safer areas these areas should be considered first. Buildings must be equipped with protection measures like reinforced walls, no windows at the hillside etc. The relation of building area within the blue zone should be low, i.e. in the range of 20 %. Special attention should be spend to large groups of people which stay in the endangered area for a longer duration. Also for these areas evacuation plans and an alarm organisation has to be established. Safe access ways must be available, especially for evacuation purposes.

3. In cases of high avalanche danger residents in yellow zones should be warned by an organisation that staying outside buildings is dangerous. Windows and other opening of houses should be closed. Doors and windows should dimensioned in a way that they withstand forces induced by powder avalanches.

4. Smaller isolated areas which can only be accessed via areas exposed to higher hazard (e.g. a house in a white zone area can only be accessed via a track in a red zone) should be treated the same way like an area in the more endangered zone.

133

References

Definitions Chapman, D. 1994. Natural Hazards, Oxford University Press, 174 pp. Gravley, 1991. Risk, Hazard and Disaster. (http://homepages.uc.edu/~huffwd/Volcanic_HazardRisk/Gravley.pdf). Hungr, O., Evans, S.G., Bovis, M.J., Hutchinson, J.N. (2001): A review of the classification of landslides of the flow type. Environmental & Engineering Geoscience, VII (3), pp. 221-238. IPENZ. 1983. Engineering Risk, report of the President’s task committee on Professional Practice and Risk 1983, Institution of Professional Engineers New Zealand, 95 pp. IUGS, 1997. Quantitative risk assessment for slopes and landslides – The state of the art. In Landslide risk assessment: Proceedings of the International Workshop on Landslide Risk Assessment, 19-21 February 1997, Honolulu, Hawaii, USA. Edited by D.M. Cruden, and R. Fell, A.A. Balkema, Rotterdam, The Netherlands, pp. 3-12. Smith, K. 1996. Environmental Hazards- Assessing Risk and Reducing Disaster, Routledge, 2nd ed., 389 pp. UNDRO (United Nations Disaster Relief Organization). 1982. Natural Disasters and Vulnerability Analysis, Geneva, Office of the United Nations Disasters Relief Coordinator. UNESCO. 1981. Avalanche Atlas. Illustrated International Avalanche Classification. International Commission on Snow and Ice. Ed.UNESCO, Paris, 265 pp. Varnes, D.J. 1984. Landslide hazard zonation: a review of principles and practice. Natural hazard, 3, Ed. Unesco, 63 pp.

Debris Flows Armanini A, Capart H, Fraccarollo L, and Larcher, M.,(2005): Rheological stratification in experimental free-surface flows of granular-liquid mixtures , Journal of Fluid Mechanics 532, 269-319. Aulitzky, H.: 1980, Preliminary two-fold classification of torrents, Proc. Int. Symp. Interpraevent, Bad Ischl, Austria, Bd. 4, pp. 285–310. Ashida, K., and Michiue, (1972), M. Study on hydraulic resistance and bedload rate in alluvial streams. Trans. Jpn. Soc. Civil Eng., 206, pp. 59-69. Ayotte, D., Hungr, O. (2000): Calibration of a runout prediction model for debris flows and avalanches. In: Wieczorek, G.F. and Naeser, N.D. (eds), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Proceedings 2nd International Conference, Taipei, Taiwan, pp. 505-514. Balkema, Rotterdam. Bagnold, R.A. (1954). Experiments on a gravity-free dispersion of large solid spheres in a Newtonian fluid under shear. Proc. Roy. Soc. London A 225, 49-63. Bathurst, J. C., Burton A., and Ward, T. J. (1997) Debris flow run-out and landslide sediment delivery model tests. J. Hydr. Eng. 123, 5, 410-419. Berti, M. and A. Simoni (2005). "Experimental evidences and numerical modelling of debris flow initiated by channel runoff." Landslides 2(3): 171-182. Berti, M. and A. Simoni (2003). Debris flow initiation in a highly-conductive soil. International Conference "Fast Slope Movements. Prediction and prevention for risk mitigation", Naples, Patron, Bologna. Bulmer, M.H., Barnouin-Jha, O.S., Peitersen, M.N., Bourke, M. (2002): An empirical approach to studying debris flows: Implications for planetary modeling studies. Journal of Geophysical Research 107 (E5): 10.1029/2001JE001531. BUWAL/BWW/BRP (1997): Empfehlungen: Berücksichtigung der Massenbewegungsgefahren bei raumwirksamen Tätigkeiten. Herausgegeber: Bundesamt für Umwelt, Wald und Landschaft (BUWAL), Bundesamt für Wasserwirtschaft (BWW), Bundesamt für Raumplanung (BRP), EDMZ Nr. 310.023d, Bern, 42p. (in German; French version also available) BWW/BRP/BUWAL (1997): Berücksichtigung der Hochwassergefahren bei raumwirksamen Tätigkeiten. Empfehlungen. Herausgegeber: Bundesamt für Wasserwirtschaft (BWW), Bundesamt für Raumplanung (BRP), Bundesamt für Umwelt, Wald und Landschaft (BUWAL), EDMZ Nr. 804.201d, Bern, Switzerland. 32p. (in German; French version also available) Cannon, S. H. (1989) An approach for estimating debris flow runout distances. In: Proceedings Conference XX, International Erosion Control Association, Vancouver, B.C., pp. 457-468. Cannon, S. H. (1993) An empirical model for the volume-change behavior of debris flows. In: Shen, H. W., Su, S. T. and Wen, F. (eds), Hydraulic Engineering ’93. Vol. 2, pp. 1768-1773. ASCE, New York. Corominas, J. (1996) The angle of reach as a mobility index for small and large landslides. Canadian Geotechnical Journal 33, 260-271. Costa, J. E. (1984): Physical geomorphology of debris flows. In J.E. Costa & P.J. Fleischer (eds.), Developments and applications of geomorphology: pp. 268-317. Springer: Berlin. Crosta, G. B., Cucchiaro, S. and Frattini, P. (2003) Validation of semi-empirical relationships for the definition of debris-flow behavior in granular materials. In: Rickenmann, D. and Chen, C.L. (eds), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Proceedings 3rd International DFHM Conference, Davos, Switzerland, pp. 821-831. Millpress, Rotterdam. Crosta, G.B. Chen, H., Lee, C.F. (2004): Replay of the 1987 Val Pola Landslide, Italian Alps. Geomorphology 60 (1-2): 127-146. D'Agostino, V., Marchi, L. (2001). Debris flow magnitude in the Eastern Italian Alps: data collection and analysis. Physics and Chemistry of the Earth (C), 26(9), 657-663. D'Agostino, V., Cerato, M., Coali, R. (1996): Il trasporto solido di eventi estremi nei torrenti del Trentino Orientale. Internationales Symposium Interpraevent, Garmisch-Partenkirchen, Tagungspublikation, Band 1, 377-386. Davies, T. R. H. (1997). “Using hydroscience and hydrotechnical engineering to reduce debris flow

134 hazards.” Proc. First Int. Conf. on Debris Flow Hazards Mitigation, San Francisco, USA, ASCE, 787- 810. Fannin, R. J. and Wise, M. P. (2001) An empirical-statistical model for debris flow travel distance. Canadian Geotechnical Journal 38, 982-994. Frenette, R, Zimmermann, T, and Eyheramendy, D, (2002): Unified modeling of fluid or granular flows on dam-break case. Journal of Hydraulic engineering 128 (3): 299-305. Fraccarollo, L. and Papa, M. (2000) Numerical simulation of real debris-flow events. Physics and Chemistry of the Earth, B, 25, 9, 757-763. Fraccarollo, L., Capart, H. and Zech, Y. (2003): A Godunov method for the computation of erosional shallow water transient. International Journal for Numerical Methods in Fluids, 41(9):951-976. Gamma, P. (2000) dfwalk – Ein Murgangsimulationprogramm zur Gefahrenzonierung. Geographica Bernensia, G66, Geographisches Institut der Universität Bern, 144p. [in German] Genolet, F. (2002) Modélisation de laves torrentielles – Contribution à la paramétrisation du modèle Voellmy-Perla. Postgraduate thesis, Ecole Polytechnique Fédérale de Lausanne, Suisse, 70p. + annexes. [in French] Ghilardi, P., Natale, L. and Savi, F. (2003) Experimental investigation and mathematical simulation of debris-flow runout distance and deposition area. In: Rickenmann, D. and Chen, C.L. (eds), Debris- Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Proceedings 3rd International DFHM Conference, Davos, Switzerland, pp. 601-610. Millpress, Rotterdam. Haeberli, W. (1983): Frequency and characteristics of glacier floods in the Swiss Alps. Annals of Glaciology, 4, 85-90. Han, G. and Wang, D. (1996) Numerical modeling of Anhui debris flow. Journal of Hydraulic Engineering 122, 5, 262-265. Hampel, R. (1977): Geschiebewirtschaft in Wildbächen. Wildbach- und Lawinenverbau, 41(1), 3-34. Hofmeister, R. J. and Miller, D. J. (2003) GIS-based modeling of debris-flow initiation, transport and deposition zones for regional hazard assessments in western Oregon, USA. In: Rickenmann, D. and Chen, C.L. (eds), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Proceedings 3rd International DFHM Conference, Davos, Switzerland, pp. 1141-1149. Millpress, Rotterdam. Huebl, J., Fiebiger, G., 2005. Debris-flow mitigation measures. In: Jakob, M., Hungr, O. (Eds.), Debris-flow Hazards and Related Phenomena. Springer, Berlin, pp. 445–487. Hungr, O. (1995): A model for the runout analysis of rapid flow slides, debris flows, and avalanches. Canadian Geotechnical Journal, 32, 610-623. Hungr, O., (1997). Some methods of landslide hazard intensity mapping. In: Cruden, D., Fell, R. (Eds.), Int. Workshop on Landslide Risk Assessment. Balkema, pp. 215–226. Hungr, O., Morgan, G. C. and Kellerhals, R. (1984) Quantitative analysis of debris torrent hazards for design of remedial measures. Canadian Geotechnical Journal 21, 663-677. Hungr, O., Evans, S.G., Bovis, M.J., Hutchinson, J.N. (2001): A review of the classification of landslides of the flow type. Environmental & Engineering Geoscience, VII (3), 221-238. Iverson, R. M. (1997) The physics of debris flows. Review of Geophysics 35, 3, 245-296. Iverson, R. M. and Denlinger, R. P. (2001) Flow of variably fluidized granular masses across three- dimensional terrain, 1. Coulomb mixture theory. Journal of Geophysical Research 106, B1, 537-552. Iverson, R. M., Schilling, S. P. and Vallance, J.W. (1998) Objective delineation of lahar-inundation zones. Geological Society of America Bulletin 110, 8, 972-984. Jakob, M., 2005. Debris-flow hazard analysis. In: Jakob, M., Hungr, O. (Eds.), Debris-flow Hazards and Related Phenomena. Springer, Berlin, pp. 411–443. Jakob, M. and Bovis, M. J. (1996) Morphometric and geotechnical controls of debris flow activity, southern Coast Mountains, British Columbia. Zeitschrift für Geomorphologie, Supplement Band 104, 13-26. Jenkins, J.T. and Hanes, D.M. (1998). Collisional sheet-flow of sediment driven by a turbulent fluid. J. Fluid Mech. 370, 29-52. Jin, M. and Fread, D. L. (1999) 1D modeling of mud/debris unsteady flows. Journal of Hydraulic Engineering 125, 8, 827-834. Körner, H.J. (1980): Modelle zur Berechnung der Bergsturz- und Lawinenberechnung. Internationales Symposium Interpraevent, Bad Ischl, Tagungspublikation, Band 2, pp. 15-55. Kronfellner-Kraus, G. (1984): Extreme Feststofffrachten und Grabenbildungen von Wildbächen. Internationales Symposium Interpraevent, Villach, Tagungspublikation, Band 2, 109-118. Kronfellner-Kraus, G. (1987). Zur Anwendung der Schätzformel fuer extreme Wildbach-Feststofffrachten im Süden und Osten Oesterreichs.Wildbach- und Lawinenverbau, 51(106), 187-200. Laigle, D. Hector, A. F., Hübl, J. and Rickenmann, D. (2003) Comparison of numerical simulation of muddy debris flow spreading to records of real events. In: Rickenmann, D. and Chen, C.L. (eds), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Proceedings 3rd International DFHM Conference, Davos, Switzerland, pp. 635-646. Millpress, Rotterdam. Lancaster, S. T., Hayes, S. K. and Grant, G.E. (2003) Effects of wood on debris flow runout in small mountain watersheds. Water Resources Research, 39, 6, 1168, doi:10.1029/2001WR001227, 21p. Legros, F. (2002) The mobility of long-runout landslides. Engineering Geology 63, 301-331. Liu, K.F. and Huang, M. C. (2006): Numerical simulation of debris flow with application on hazard area mapping. Computational Geosciences (2006). McArdell, B.W., Zanuttigh, B., Lamberti, A. and Rickenmann, D. (2003) Systematic comparison of debris flow laws at the Illgraben torrent, Switzerland. In: Rickenmann, D. and Chen, C.L. (eds), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Proceedings 3rd International DFHM Conference, Davos, Switzerland, pp. 647-657. Millpress, Rotterdam. McDougall, S. D. and Hungr, O. (2003) Objectives for the development of an integrated three-dimensional continuum model for the analysis of landslide runout. In: Rickenmann, D. and Chen, C.L. (eds), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Proceedings 3rd International DFHM Conference, Davos, Switzerland, pp. 481-490. Millpress, Rotterdam.

135 Mizuyama, T., Kobashi, S. and Ou, G. (1992) Prediction of debris flow peak discharge. Internationales Symposium Interpraevent, Bern, Switzerland, Tagungspublikation, Band 4, pp. 99-108. Naef, D., Rickenmann, D., Rutschmann, P., McArdell, B.W. (2006): Comparison of flow resistance relations for debris flows using a one-dimensional finite element simulation model. Natural Hazards and Earth System Sciences, 6:155-165. Nakagawa, H., Takahashi, T. and Satofuka, Y. (2000) A debris-flow disaster on the fan of the Harihara River, Japan. In: Wieczorek, G.F. and Naeser, N.D. (eds), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Proceedings 2nd International Conference, Taipei, Taiwan, pp. 193-201. Balkema, Rotterdam. O'Brien, J. S., Julien, P. Y. and Fullerton, W. T. (1993) Two-dimensional water flood and mudflow simulation. Journal of Hydraulic Engineering 119, 2, 244-261. Perla, R., Cheng, T. T. and McClung, D. M. (1980) A two parameter model of snow avalanche motion. Journal of Gaciology 26, 94, 197-208. Petrascheck, A. and Kienholz, H. (2003) Hazard assessment and mapping of mountain risks in Switzerland. In: Rickenmann, D. and Chen, C.L. (eds), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Proceedings 3rd International DFHM Conference, Davos, Switzerland, pp. 25-38. Millpress, Rotterdam. Phillips, R. J., Armstrong, R. C., Brown, R. A. (1992): A constitutive equation for concentrated suspensions that account for shear-induced particle migration, Phys. Fluids A, 4, 30-41. Quemada, D., (1998): Rheological modelling of complex fluids. I. The concept of effective volume fraction. European Physical Journal (Applied Physics), 1 119-127. Raetzo, H., Lateltin, O., Bollinger, D., Tripet, J.P., 2002. Hazard assessment in Switzerland – codes of practice for mass movements. Bulletin of Engineering Geology and the Environment 61, 263–268. Rickenmann, D. (1990) Debris flows 1987 in Switzerland: modelling and sediment transport. IAHS Publ. no. 194, pp. 371-378. Rickenmann, D. (1994): An alternative equation for the mean velocity in gravel-bed rivers and mountain torrents. In: G.V. Cotroneo & R.R. Rumer (eds), Proceedings ASCE 1994 National Conference on Hydraulic Engineering, Buffalo N.Y., USA, Vol. 1, 672-676. Rickenmann, D. (1995): Beurteilung von Murgängen. Schweizer Ingenieur und Architekt, 48, 1104-1108. Rickenmann, D. (1996): Fliessgeschwindigkeit in Wildbächen und Gebirgsflüssen. Wasser, energie, luft, 88 (11/12), 298-304. Rickenmann, D. (1999) Empirical relationships for debris flows. Natural Hazards 19, 47-77. Rickenmann, D. (2001): Murgänge in den Alpen und Methoden zur Gefahrenbeurteilung. In: Mitteilungen des Lehrstuhls und Instituts für Wasserbau und Wasserwirtschaft, Rheinisch-Westfälische Technische Hochschule Aachen, Band 124, 51-77. Rickenmann, D. (2003): Methoden zur Beurteilung von Murgängen. Teil-Bericht für das Projekt ETALP des BMLFUW, Wien, Oesterreich. Rickenmann, D. (2005a): Runout prediction methods. In: M. Jakob & O. Hungr (eds), Debris-Fow Hazards and Related Phenomena, Praxis-Springer, pp. 263-282. Rickenmann, D. (2005b): Hangmuren und Gefahrenbeurteilung. Kurzbericht für das Bundesamt für Wasser und Geologie. Universität für Bodenkultur, Wien, und Eidg. Forschungsanstalt WSL, Birmensdorf. Rickenmann, D. and Koch, T. (1997) Comparison of debris flow modelling approaches. In: Chen, C.L. (ed.), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Proceedings 1st International DFHM Conference, San Francisco, USA, pp. 576-585. American Society of Civil Engineers, New York. Rickenmann, D., Weber, D. (2000): Flow resistance of natural and experimental debris flows in torrent channels. In: G.F. Wieczorek & N.D. Naeser (eds.), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment; Proceedings 2nd International DFHM Conference, Taipei, Taiwan, August 16-18, 2000, Rotterdam, Balkema, 245-254. Rickenmann, D. and Zimmermann, M. (1993) The 1987 debris flows in Switzerland: documentation and analysis. Geomorphology 8, 175-189. Rickenmann, D., Laigle, D., McArdell, B.W., Hübl. J. (2005): Comparison of 2D debris-flow simulation models with field events. Computational Geosciences, DOI: 10.1007/s10596-005-9021-3 (online). Rosatti, G., Fraccarollo, L. (2006): Simulazioni bidimensionali di colate detritiche in canaloni erodibili. 30° Convegno di Idraulica e Costruzioni Idrauliche", Roma (in press). Salm, B. (1966) Contribution to avalanche dynamics. Proceedings International Symposium on Scientific Aspects of Snow and Ice Avalanches, Christchurch, N.Z., IAHS Publ. No. 69, pp. 199-214. Savage, S.B. (1998). Analyses of slow high-concentration flows of granular materials. J. Fluid Mech. 377, 1- 26. Savage, S. B. and Hutter, K. (1989) The motion of a finite mass of granular material down a rough incline. Journal of Fluid Mechanics 199, 177-215. Spreafico, M., Lehmann, C., Naef, O. (1996): Empfehlung zur Abschätzung von Feststofffrachten in Wildbächen. Teil I: Handbuch, 46p. + Anhang, und Teil II: Fachliche Grundlagen, 113p., Groupe de travail pour l'hydrologie operationelle (GHO), Mitt. Nr. 4, Landeshydrologie und –geologie, Bern. Takahashi, T. (1991) Debris Flow. IAHR Monograph Series, Balkema, Rotterdam, 165p. Takahashi, T., Nakagawa, H., Harada, T. and Yamashiki, Y. (1992) Routing debris flows with particle segregation. Journal of Hydraulic Engineering 118, 11, 1490-1507. Takei, A. (1984): Interdependence of sediment budget between individual torrents and a river-system. Internationales Symposium Interpraevent, Villach, Tagungspublikation, Band 2, 35-48. Tognacca, C., Bezzola, G.R., Minor, H.-E. (2000): Threshold criterion for debris-flow initiation due to channel-bed failure. In G.F. Wieczorek & N.D. Naeser (eds), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment; Proceedings 2nd International DFHM Conference, Taipei, Taiwan, August 16-18, 2000, Balkema, Rotterdam, 89-97. Valiani, A. and Caleffi, V. (2003) Numerical simulation of dam-dreak flow on granular bed: intense sediment transport vs. debris flow modeling. In: Rickenmann, D. and Chen, C.L. (eds), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Proceedings 3rd International DFHM

136 Conference, Davos, Switzerland, pp. 601-610. Millpress, Rotterdam. Van Gassen, W. and Cruden, D. M. (1989) Momentum transfer and friction in the debris of rock avalanches. Canadian Geotechnical Journal 26, 623-628. VAW (1992) Murgänge 1987: Dokumentation und Analyse. Unpublished Report, No. 97.6, Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie (VAW), ETH Zürich, 620p. Voellmy, A. (1955) Über die Zerstörungskraft von Lawinen. Schweizerische Bauzeitung, Jahrgang 73, Hefte 12:159-162, 15:212-217, 17:246-249, 19:280-285. [in German] Wilford, D., M. Sakals, et al. (2004). Recognition of debris flow, debris flood and flood hazard through watershed morphometrics. Landslides 1: 61-66. Wise, M. P. (1997) Probabilistic modelling of debris flow travel distance using empirical volumetric relationships. M.A.Sc. thesis, University of British Columbia, Vancouver, B.C., Canada. Zanuttigh, B. and Lamberti, A. (2004) Numerical modelling of debris surges based on shallow-water and homogeneous material approximations. Journal of Hydraulic Research 42, 4, 376-389. Zeller, J. (1985): Feststoffmessung in kleinen Gebirgseinzugsgebieten. Wasser, energie, luft, 77(7/8), 246- 251. Zimmermann, M., Lehmann, C. (1999). Geschiebefracht in Wildbächen: Grundlagen und Schätzverfahren. Wasser, energie, luft, 91 (7/8), 189-194. Zimmermann, M., Mani, P., Gamma, P., Gsteiger, P., Heiniger, O., and Hunziker, G. (1997a) Murganggefahr und Klimaänderung - ein GIS-basierter Ansatz. Schlussbericht NFP 31, vdf-ETH, Zürich, Switzerland, 161p. Zimmermann, M., Mani, P., Romang, H. (1997b). Magnitude-frequency aspects of Alpine debris flows. Eclogae Geologicae Helvetiae, 90(3), 415-420.

Rock Avalanches Abe, S., Mair, K., 2005. Grain fracture in 3D numerical simulations of granular shear. Geophysical Research Letters 32, L05305, doi:10.1029/2004GL022123. Abele, G., 1974. Bergstürze in den Alpen – ihre Verbreitung, Morphologie und Folgeerscheinungen. Wissenschaftliche Alpenvereinshefte 25, 230pp. Abele, G., 1997. Rockslide movement supported by the mobilization of groundwater-saturated valley floor sediments. Zeitschrift für Geomorphologie Supplementband N.F. 41, 1-20. Aleotti, P., Chowdbury, R., 1999. Landslide hazard assessment: summary review and new perspectives. Bulletin of Engineering Geology and the Environment 58, 21-44. Ambrosi, C., Crosta, G.B., 2006. Large sackung along major tectonic features in the Central Italian Alps. Engineering Geology 83, 183-200. Bachmann, D., Bousissou, S., Chemenda, A., 2004. Influence of weathering and pre-existing large scale fractures on gravitational slope failure: insights from 3-D physical modelling. Natural Hazards and Earth System Sciences 4, 711-717. Ballantyne, C.K., Stone, J.O., 2004. The Beinn Alligin rock avalanche, NW Scotland: cosmogenic Be-10 dating, interpretation amd significance. Holocene 14, 448-453. Barnouin-Jha, O.S., Baloga, S., Glaze, L., 2005. Comparing landslides to fluidized crater ejecta on Mars. Journal of Geophysical Research 110, E04010, doi:10.1029/2003JE002214. Blikra, L.H., Longva, O., Harbitz, C., Løvholt, F., 2005. Quantification of rock-avalanche and tsunami hazard in Storfjorden, western Norway. In: Senneset, K., Flaate, K., Larsen, J.O. [Eds.] Landslides and Avalanches. Proceedings 11th International Conference and Field Trip on Landslides, 1-10 September 2005. Balkema/Taylor & Francis, London, pp. 57-63. Bottino, G., Chiarle, M., Joly, A., Mortasa, G., 2002. Modelling rock avalanches and their relation to permafrost degradation in glacial environments. Permafrost and Periglacial Processes 13, 283-288. Boultbee, N., Stead, D., Schwab, J., Geertsema, M., 2006. The Zymoetz River rock avalanche, June 2002, British Columbia, Canada. Engineering Geology 83, 76-93. Brodsky, E.E., Gordeev, E., Kanamori, H., 2003. Landslide basal friction as measured by seismic waves. Geophysical Research Letters 30, 2236, doi:10.1029/2003GL018485. Bulmer, M.H., Zimmerman, B.A., 2005. Reassessing landslide deformation in Ganges Chasma, Mars. Geophysical Research Letters 32, L06201, doi:10.1029/2004GL022021. Capra, L., Macías, J.L., 2002. The cohesive Naranjo debris-flow deposit (10 km3): A dam breakout flow derived from the Pleistocene debris-avalanche deposit of Nevado de Colima Volcano (Mexico). Journal of Volcanology and Geothermal Research 117, 213-235. Chemenda, A., Bouissou, S., Bachmann, D., 2005. Three-dimensional physical modeling of deep-seated landslides: New technique and first results. Journal of Geophysical Research 110, F04004, doi:10.1029/2004JF000264. Chen, R.F., Chan., Y.C., Angelier, J., Hu, J.C., Huang, C., Chang, K.J., Shih, T.Y., 2005. Large earthquake- triggered landslides and mountain belt erosion: The Tsaoling case, Taiwan. Comptes Rendues Geoscience 337, 1164-1172. Chigira, M., Wang, W.N., Furuya, T., Kamai, T., 2003. Geological causes and geomorphological precursors of the Tsaoling landslide triggered by the 1999 Chi-Chi earthquake, Taiwan. Engineering Geology 68, 259-273. Clague, J.J., Turner, R.J.W., Reyes, A.V., 2003. Record of recent river channel instability, Cheakamus Valley, British Columbia. Geomorphology 53, 317-322. Clavero, J.E., Sparks, R.S.J., Huppert, H.E., Dade, W.B., 2002. Geological constraints on the emplacement mechanism of the Parinacota debris avalanche, northern Chile. Bulletin of Volcanology 64, 40-54. Collins, G.S., Melosh, H.J., Acoustic fluidization and the extraordinary moblility of sturzstroms. Journal of Geophysical Research 108, NO. B10, 2473, doi:10.1029/2003JB002465. Corominas, J., 1996. The angle of reach as a mobility index for small and large landslides. Canadian Geotechnical Journal 33, 260-271. Costa, J.E., Schuster, R.L., 1988. The formation and failure of natural dams. Geological Society of America Bulletin 100, 1054-1068. Crosta, G.B., Agliardi, F., 2002. How to obtain velocity thresholds for large rockslides. Physics and

137 Chemistry of the Earth 27, 1557-1565. Crosta, G.B., Calvetti, C., Imposimato, S., Roddeman, D., Frattini, P., Agliardi, F., 2001. Granular flows and numerical modelling of landslides. DAMOCLES Report, 71p. Crosta, G.B., Chen, H., Lee, C.F., 2004. Replay of the 1987 Val Pola Landslide, Italian Alps. Geomorphology 60, 127-146. Crosta, G.B., Imposimato, S., Roddeman, D., 2003. Numerical modelling of landslide stability and runout. Natural Hazards and Earth System Sciences 3, 523-538. Cruden, D.M., Varnes, D.J., 1996. Landslide types and processes. In: Turner, A.K., Schuster, R.L. (Eds.), Landslides, investigation and mitigation. Special Report 247, Transportation Research Board, National Research Council, pp. 36-75. Dade, W.B., Huppert, H.E., 1998. Long-runout rockfalls. Geology 26, 803-806. Dai, F.C., Lee, C.F., Ngai, Y.Y., Landslide risk assessment and management: an overview. Engineering Geology 64, 65-87. Davies, T.R.H., 2002. Landslide-dambreak floods at Franz Josef Glacier township, Westland, New Zealand: A risk assessment. Journal of Hydrology (New Zealand) 41, 1-17. Davies, T.R., McSaveney, M.J., 1999. Runout of dry granular avalanches. Canadian Geotechnical Journal 36, 313-320. Davies, T.R., McSaveney, M.J., 2002. Dynamic simulation of the motion of fragmenting rock avalanches. Canadian Geotechnical Journal 39, 789-798. Davies, T.R., McSaveney, M.J., Hodgson, K.A., 1999. A fragmentation-spreading model for long-runout rock avalanches. Canadian Geotechnical Journal 36, 1096-1110. Davies, T.R., McSaveney, M.J. and Beetham, R.D., 2006. Rapid block glides: slide-surface fragmentation in New Zealand's Waikaremoana landslide. Quarterly Journal of Engineering Geology and Hydrogeology 39, 115-129. Davies, T.R.H., Scott, B.K., 1997. Dambreak flood hazard from the Callery River, Westland, New Zealand. Journal of Hydrology New Zealand 36, 1-13. Denlinger, R.P., Iverson, R.M., 2004. Granular avalanches across irregular three-dimensional terrain: 1. Theory and computation. Journal of Geophysical Research 109, F01014, doi:10.1029/2003JF000085. Dunning, S.A., Rosser, N.J., Petley, D.N., Massey, C.R., 2006. Formation and failure of the Tsatichhu landslide dam, Bhutan. Landslides 3, 107-113. Dussauge, C., Grasso, J.R., Helmstetter, A.S., 2003. Statistical analysis of rockfall volume distributions: implications for rockfally dynamics. Journal of Geophysical Research-Solid Earth 108, doi: 10.1029/2001JB000650.. Eisbacher, G.H., Clague, J.J., 1984. Destructive mass movements in high mountains – hazard and management. Geological Survey of Canada Paper 84-16, 230p. Erismann, T.H., Heuberger, H., Preuss, E., 1977. Der Bimstein von Köfels (Tirol), ein Bergsturz-‘Friktionit’. Tschermaks Mineralogische und Petrographische Mitteilungen 24, 67-119. Ermini, L., Casagli, N., 2003. Prediction of the behaviour of landslide dams using a geomorphological dimensionless index. Earth Surface Processes and Landforms 28, 31-47. Evans, S.G., et al., 2002. Climate change and geomorphological hazards in the Canadian Cordillera; the anatomy of impact and some tools for adaptation. Scientific Report Climate Change Action Fund Project A099. Evans, S.G., Guthrie, R.H., Roberts, N.J., Bishop, N.F., 2006. The disastrous 17 February 2006 rockslide- debris avalanche on Leyte Island, Philippines: a catastrophic landslide in tropical mountain terrain. Natural Hazards and Earth System Sciences 7, 89-101. Forlati, F., Gioda, G., Scavia, C., 2001. Finite element analysis of a deep-seated slope deformation. Rock Mechanics and Rock Engineering 34, 135-159. Friedmann, S.J., Kwon, G., Losert, W., 2003. Granular memory and its effect on the triggering and distribution of rock avalanche events. Journal of Geophysical Research 108(B8), doi:10.1029/2002JB002174. Geertsema, M., Clague, J.J., Schwab, J.W. and Evans, S.G., 2006. An overview of recent large catastrophic landslides in northern British Columbia, Canada. Engineering Geology 83, 120-143. Gray, J.M.N.T., Tai, Y.C., Noelle, S., 2003. Shock waves, dead zones, and particle-free regions in rapid granular free-surface flows. Journal of Fluid Mechanics 491, 161-181. Harden, C., 2001. Sediment movement and catastrophic events: The 1993 rockslide at La Josefina, Ecuador. Physical Geography 22, 305-320. Harp, E.L., Crone, A.J., 2006. Landslides triggered by the October 8, 2005, Pakistan earthquake and associated landslide-dammed reservoirs. US Geological Survey Open-File Report 2006-1052, 13p. Heim, A., 1932. Bergsturz und Menschenleben. Beiblatt zur Vierteljahresschrift der Naturforschenden Gesellschaft Zürich 20, 217pp. Helmstetter, A., Sornette, D., Grasso, J.R., Andersen, J.V., Gluzman, S. Pisarenko, V., 2004. Slider block friction model for landslides: Application to Vaiont and La Clapière landslides. Journal of Geophysical Research 109, B02409, doi:10.1029/2002JB002160. Hermanns, R.L., Niedermann, S., Ivy-Ochs, S., Kubik, P.W., 2004. Rock avalanching into a landslide- dammed lake causing multiple dam failure in Las Conchas valley (NW Argentina) – evidence from surface exposure dating and stratigraphic analyses. Landslides 1, 113-122. Hermanns, R.L., Strecker, M.R., 1999. Structural and lithological controls on large Quaternary rock avalanches sturzstroms in arid northwestern Argentina. Geological Society of America Bulletin 111, 934-948. Hewitt, K., 1998. Catastrophic landslides and their effects on the Upper Indus streams, Karakoram Himalaya, northern Pakistan. Geomorphology 26, 47-80. Hewitt, K., 1999. Quaternary moraines vs catastrophic rock avalanches in the Karakoram Himalaya, Northern Pakistan. Quaternary Research 51, 220-237. Holm, K., Bovis, M. and Jakob, M., 2004. The landslide response of alpine basins to post-Little Ice Age glacial thinning and retreat in southwestern British Columbia. Geomorphology 57,: 201-216.

138 Hsü, K.J., 1975. Catastrophic debris streams (sturzstroms) generated by rockfalls. Geological Society of America Bulletin 86, 129-140. Huggel, C., Zgraggen-Oswald, S., Haeberli, W., Kääb, A., Polkvoj, A., Galushkin, I., Evans, S.G., 2005. The 2002 rock/ice avalanche at Kolka/Karmadon, Russian Caucasus: assessment of extraordinary avalanche formation and mobility, and application of QuickBird satellite imagery. Natural Hazards and Earth System Sciences 5, 173-187. Hungr, O., 1995. A model for the runout analysis for rapid flow slides, debris flows, and avalanches. Canadian Geotechnical Journal 32, 610-623. Hungr, O., Evans, S.G., 2004. Entrainment of debris in rock avalanches: An analysis of a long run-out mechanism. Geological Society of America Bulletin 116, 1240-1252. Hungr, O., Evans, S.G., Bovis, M., Hutchinson, J.N., 2001. Review of the classification of landslides of the flow type. Environmental and Engineering Geoscience 7, 221-238. Hungr, O., Evans, S.G. and Hazzard, J., 1999. Magnitude and frequency of rock falls and rock slides along the main transportation corridors of southwestern {British Columbia}. Canadian Geotechnical Journal 36, 224-238. Iverson, R.M., 2003. How should mathematical models of geomorphic processes be judged? In: Wilcock, P.R., Iverson, R.M. (Eds.), Prediction in geomorphology. American Geophysical Monograph 135, AGU Washington, DC, pp. 83-94. Iverson, R.M., Logan, M., Denlinger, R.P., 2004. Granular avalanches across irregular three-dimensional terrain: 2. Experimental tests. Journal of Geophysical Research 109, F01015, doi:10.1029/2003JF000084. Jibson, R.W., Harp, E.L., Schulz, W. and Keefer, D.K., 2006. Large rock avalanches triggered by the M 7.9 Denali Fault, Alaska, earthquake of 3 November 2002. Engineering Geology 83, 144-160. Kääb, A., 2002. Monitoring high-mountain terrain deformation from repeated air- and spaceborne optical data: examples using digital aerial imagery and ASTER data. ISPRS Journal of Photogrammetry & Remote Sensing 57, 39-52. Keefer, D.K., 1999. Earthquake-induced landslides and their effects on alluvial fans. Journal of Sedimentary Research 69, 84-104. Kelfoun, K., Druitt, T.H., 2005. Numerical modeling of the emplacement of Socompa rock avalanche, Chile. Journal of Geophysical Research 110, B12202, doi:10.1029/2005JB003758. Kilburn, C.R.J., Pasuto, A., 2003. Major risk from rapid, large-volume landslides in Europe (EU Project RUNOUT). Geomorphology 54, 3-9. Korup, O., 2002. Recent research on landslide dams – a literature review with special attention to New Zealand. Progress in Physical Geography 26, 206-235. Korup, O., 2004a. Geomorphometric characteristics of New Zealand landslide dams. Engineering Geology 73, 13-35. Korup, O., 2004b. Landslide-induced river channel avulsions in mountain catchments of southwest New Zealand. Geomorphology 63, 57-80. Korup, O., 2005a. Geomorphic hazard assessment of landslide dams in South Westland, New Zealand – fundamental problems and approaches. Geomorphology 66, 167-188. Korup, 2005b. Distribution of landslides in southwest New Zealand. Landslides 2, 43-51. Korup, O., McSaveney, M.J., Davies, T.R.H., 2004. Sediment generation and delivery from large historic landslides in the Southern Alps, New Zealand. Geomorphology 61, 189-207. Korup, O., Clague, J.J., Hermanns, R.L., Hewitt, K., Strom, A.L., Weidinger, J.T., in prep. Giant landslides, topography, and erosion. Korup, O., Glade, T., in prep. Hazard and risk from extreme events—The case of large and catastrophic landslides. Korup, O., Tweed, F., in prep. Formation and failure of ice, moraine, and landslide dams in mountainous terrain. Kubik, P.W., Ivy-Ochs, S., Masarik, J., Frank, M., Schlüchter, C., 1998. 10Be and 26Al production rates deduced from an instantaneous event within the dendro-calibration curve: the landslide of Köfels, Ötz Valley, Austria. Earth and Planetary Science Letters 161, 231-241. Lagmay, A.M.A., Ong, J.B.T., Frenandez, F.E.D., Lapus, M.R., Rodolfo, R.S., Tengociang, A.M.P., Soria, J.L.A., Baliatan, E,G., Quimba, Z.L., Uichanco, C.L., Paguican, E.M.R., Remedio, A.R.C., Lorenzo, G.R.H., Valdivia, W., Avila, F.B., 2006. Scientists investigate recent Philippine landslide. EOS Transactions 87, 121-124. Legros, F., 2002. The mobility of long-runout landslides. Engineering Geology 63, 301-331. Legros, F., Cantagrel, J. M., Devouard, B., 2000. Pseudotachylyte (Frictionite) at the base of the Arequipa volcanic landslide deposit (Peru): Implications for emplacement mechanisms. Journal of Geology 108, 601-611. Leonov, N.N., 1960. The Khait 1949 earthquake and geological conditions of its origin. Proceedings of Academy of Sciences of the USSR Geophysical Series 3, 409-424 (in Russian). Li, M.H., Hsu, M.H., Hsieh, L.S., Teng, W.H., 2002. Inundation potential analysis for Tsao-Ling landslide lake formed by Chi-Chi earthquake in Taiwan. Natural Hazards 25, 289-303. Liao, W.M., Chou, H.T., 2003. Debris flows generated by seepage from landslide dams. In: Rickenmann, D., Chen, C.L. (Eds.), Proceedings 3rd International Conference on Debris-Flow Hazard Mitigation, Millpress, Rotterdam, pp. 315-325. Lynett, P., Liu, P.L.F., 2005. A numerical study of the run-up generated by three-dimensional landslides. Journal of Geophysical Research 110, C03006, doi:10.1029/2004JC002443. Malamud, B.D., Turcotte, D.L. Guzzetti, F. Reichenbach, P., 2004. Landslide inventories and their statistical properties. Earth Surface Processes and Landforms 29, 687-711. McDougall, S., Hungr, O., 2004. A model for the analysis of rapid landslide motion across three-dimensional terrain. Canadian Geotechnical Journal 41, 1084-1097. McSaveney, M.J., 2002. Recent rockfalls and rock avalanches in Mount Cook National Park New Zealand. In: Evans, S.G., DeGraff, J.V. (Eds.), Catastrophic Landslides: Effects, Occurrence, and Mechanisms. Geological Society of America Reviews in Engineering Geology XV, pp. 35-70.

139 Nicoletti, P.G., Sorriso-Valvo, M., 1991. Geomorphic controls of the shape and mobility of rock avalanches. Geological Society of America Bulletin 103, 1365-1373. Orwin, J.F., Clague, J.J., Gerath, R.F., 2004. The Cheam rock avalanche, Fraser Valley, British Columbia, Canada. Landslides 1, 289-298. Panizzo, F., De Girolamo, P., Petaccia, A., 2005. Forecasting impulse waves generated by subaerial landslides. Journal of Geophysical Research 110, C12025, doi:10.1029/2004JC002778. Pedersen, S.A., Larsen, L.M., Dahl-Jensen, T., Jepsen, H.F., Pedersen, G.K., Nielsen, T., Pedersen, A.K., Platen-Hallermund, F., Weng, W., 2002. Tsunami-generating rock fall and landslide on the south coast of Nuussuaq, central West Greenland. Geology of Greenland Survey Bulletin 191, 73-83. Pirulli, M., 2004. Numerical analysis of three potential rock avalanches in the Alps. Felsbau 22, 32- 38. Pirulli, M., Scavia, C., Mortara, G., Tamburini, A., Deline, P., Noetzli, J., 2006. The Thurwieser rock avalanche, Italian Alps: description and dynamic analysis. Geophysical Research Abstracts 8, 08482. Plafker, G., Ericksen, G.E., 1978. Nevados Huascaran avalanches, Peru. In Voight, B. (Ed.), Rockslides and Avalanches I: Natural Phenomena Publications. Elsevier, Amsterdam, pp. 277-314. Poisel, R., Hermann, S., Kittl, H., Preh, A., 2003. Die Grosshangbewegung Larchberg-Galgenwald bei Murau. Felsbau 21, 110-119. Pollet, N., Cojean, R., Couture, R., Schneider, J.L., Strom, A.L., Voirin, C., Wassmer, P., 2005. A slab-on- slab model for the Flims rockslide (Swiss Alps): Canadian Geotechnical Journal 42, 587-600. Ponomareva, V.V., Melekestsev, I.V., Dirksen. O.V., 2006. Sector collapses and large landslides on Late Pleistocene-Holocene volcanoes, Kamchatka, Russia. Journal of Volcanology and Geothermal Research 158, 117-138. Poschinger, A. von, Haas, U., 1997. Der Flimser Bergsturz, doch ein warmzeitliches Ereignis? Bulletin für Angewandte Geologie 2, 35-46. Raetzo, H., Latelin, O., Tripet, J.P., Oswald, D., Dapples, F., 2002. Landslides and evaluation of triggering factors, hazard assessment practice in Switzerland. In: Rybar, J., Stemberk, J., Wagner, P. (Eds.), Landslides. Proceedings 1st European Conference on Landslides 24-26 June 2002, Prague. Balkema, Rotterdam, pp. 455-460. Radbruch-Hall, D.H., 1978. Gravitational creep of rock masses on slopes. In: Voight, B. (Ed.), Rockslides and avalanches 1. Elsevier, Amsterdam, pp. 607-657. Risley, J.C., Walder, J.S. and Denlinger, R.P., 2006. Usoi dam wave overtopping and flood routing in the Bartang and Panj Rivers, Tajikistan. Natural Hazards 38, 375-390. Sandersen, , F., Bakkehøi, S., Hestnes, E., Lied, K., 1996. The influence of meteorological factors on the initiation of debris flows, rockfalls, rockslides and rockmass stability. In: Senneset, K. (ed) Landslides. Proceedings of the 7th Symposium on landslides, Trondheim, 17-21 June 1996, pp. 97- 114. Sartori, M., Baillifard, F., Jaboyedoff, M., Rouiller, J.D., 2003. Kinematics of the 1991 Randa rockslides (Valais, Switzerland). Natural Hazards and Earth System Sciences 3, 423-433. Scheidegger, A.E., 1973. On the prediction of the reach and velocity of catastrophic landslides. Rock Mechanics 5, 231-236. Schneider, J.L., Pollet, N., Chapron, E., Wessels, M., Wassmer, P., 2004. Signature of Rhine Valley sturzstrom dam failures in Holocene sediments of Lake Constance, Germany. Sedimentary Geology 169, 75-91. Schramm, J.-M., Weidinger, J.T., Ibetsberger, H.J., 1998. Petrologic and structural control on geomorphology of prehistoric Tsergo Ri slope failure, Langtang Himal, Nepal. Geomorphology 26, 107-121. Schuster, R.L., Alford, D., 2004. Usoi Landslide Dam and Lake Sarez, Pamir Mountains, Tajikistan. Environmental and Engineering Geoscience 10, 151-168. Segalini, A., Giani, G.P., 2004. Numerical model for the analysis of the evolution mechanisms of the Grossgufer rock slide. Rock Mechanics and Rock Engineering 37, 151-168. Shang, Y., Yang, Z., Li, L., Liu, D., Liao, Q., Wang, Y., 2003. A superlarge landslide in Tibet in 2000: Background, occurrence, disaster, and origin, Geomorphology 54, 225-243. Sørensen, S.A., Bauer, B., 2003. On the dynamics of the Köfels sturzstrom. Geomorphology 52, 11-19. Smith, G.M., Davies, T.R., McSaveney, M.J., Bell, D.H., 2006. The Acheron rock avalanche, Canterbury, New Zealand—morphology and dynamics. Landslides 3, 62-72. Smyth, C.G., Royle, S.A., 2000. Urban landslide hazards: incidence and causative factors in Niterói, Rio de Janeiro State, Brazil. Applied Geography 20, 95-117. Sornette, D., Helmstetter, A., Andersen, J.V., Gluzman, S., Grasso, J.R., Pisarenko, V., 2004. Towards landslide predictions: two case studies. Physica A, 605-632. Stead, D., Eberhardt, E., Coggan, J.S., 2006. Developments in the characterization of complex rock slope deformation and failure using numerical modelling techniques. Engineering Geology 83, 217-235. Stewart, M.L., Russell, J.K., Hickson, C.J., 2003. Discrimination of hot versus cold avalanche deposits: implications for hazard assessment at Mount Meager, B.C. Natural Hazards and Earth System Sciences 3, 713-724. Tarchi, D., Casagli, N., Moretti, S., Leva, D., Sieber, A.J., 2003. Monitoring landslide displacements by using ground-based synthetic aperture radar interferometry: Application to the Ruinon landslide in the Italian Alps. Journal for Geophysical Research 108B, ETG 10-1- ETG 10-14. Turnbull, J.M., Davies, T.R.H., 2006. A mass movement origin for cirques. Earth Surface Processes and Landforms 31, 1129-1148. Walder, J.S., O’Connor, J.E., 1997. Methods for predicting peak discharge of floods caused by failure of natural and constructed earthen dams. Water Resources Research 33, 2337-2348. Wassmer, P., Schneider, J.L., Pollet, N., Schmitter-Voirin, C., 2004. Effects of the internal structure of a rock–avalanche dam on the drainage mechanism of its impoundment, Flims sturzstrom and Ilanz paleo-lake, Swiss Alps. Geomorphology 61, 3-17. Waythomas, C.F., Nye, C.J., 2002. Preliminary volcano-hazard assessment for Mount Spurr, Alaska. U.S. Geological Survey Oopen-File Report 01-482, 46p.

140 Weidinger, J.T., Korup, O., Petrologic evidence for a large Late Quaternary rockslide near Kanchenjunga, Sikkim Himalayas, India – implications for extreme events in mountain relief destruction. Geomorphology (in press). Weidinger, J.T., Schramm, J.M, Nuschej, F., 2002. Ore mineralization causing slope failure in a high-altitude mountain crest—on the collapse of an 8000 m peak in Nepal. Journal of Asian Earth Sciences 21, 295-306. Weidinger, J.T., Wang, J.D. and Ma, N.X., 2002. The earthquake-triggered rock avalanche of Cui Hua, Qin Ling Mountains, P. R. of China - the benefits of a lake-damming prehistoric natural disaster. Quaternary International 93-4, 207-214. Whitehouse, I.E., 1983. Distribution of large rock avalanche deposits in the central Southern Alps, New Zealand. New Zealand Journal of Geology and Geophysics 26, 272-279. Whitehouse, I.E., Griffiths, G.A., 1983. Frequency and hazard of large rock avalanches in the central Southern Alps. Geology 11, 331-334. Willenberg, H., Ladner, F., Spillmann, T., Low, S., Sambeth, U., Evans, K., Eberhardt, E., 2002. Aufbau eines Multi-Parameter-Überwachungssystems für Felsrutschungen. Felsbau 20, 44-51. Wilson, A.J., Petley, D.N., Murphy, W., 2003. Down-slope variation in geotechnical parameters and pore fluid control on a large-scale Alpine landslide. Geomorphology 54, 49-62. Wright C.A., 1999. The AD 930 long-runout Round Top debris avalanche, Westland, New Zealand. New Zealand Journal of Geology and Geophysics 41, 493-497.

Snow Avalanches Adjiel, G. (1995). Methodes statistiques pour la determination de la distance d’arret maximale des avalanches. La Houille Blanche, 7, 100-104. Ancey, C., Meunier, M. (1999). Utilizzo di tecniche statistiche per la determinazione di scenari di valanghe di riferimento e aiuto alla decisione [Use of statistical techniques for the definition of avalanche scenarios and as a support for decisions], Proc. of TRACE 99 (Table Ronde Avalanche Control in Europe), Breuil Cervinia, Italy, 19–20 April 1999, 47–58, In Italian. Ancey, C., Meunier, M., Richard, D. (2002). Inverse problem in avalanche dynamics models, Water Resour. Res., 39(4), 1099, doi:10.1029/2002WR001749, 2003. Ancey, C., Gervasoni, C. Meunier, M. (2004). Computing extreme avalanches, Cold Regions Science and Technology, 39, 161– 180. Baeriswyl, P. A., Rebetez, M. (1997). Regionalization of precipitation in Switzerland by means of principal components analysis, Theor. App. Climatol., 58, 31-41. Bakkehøi, S. and Norem, H. (1994). Sammenlikning av metoder for beregning av maksimal utløpsdistanse for snøskred (Comparison of methods for calculation of maximum avalanche runout distance). Norwegian Geotechnical Institute, report 581200-30 (in Norwegian). Bakkehøi, S., Domaas, U. and Lied, K. (1983). Calculation of Snow Avalanche Run-out Distance. Annals of Glaciology, Vol. 4, 24-29. Also in: Norwegian Geotechnical Institute, publication No. 151, 1984. Bakkehøi, S., Cheng, T., Domaas, U., Lied, K., Perla, R.I. and Schieldrop, B. (1981). On the computation of parameters that model snow avalanche motion. Canadian Geotechnical Journal 18(1), 121-130. Barbolini, M. (1998). “VARA one- and two-dimensional models”. In: A Survey of computational models for snow avalanche motion. Edited by C. Harbitz. NGI Report No. 581220-1, 59-63. Barbolini, M. (1999). Dense Snow Avalanches: Computational Models, Hazard Mapping and related Uncertainties. PhD Thesis, University of Pavia. Barbolini, M., Gruber, U., Keyloch, C., Naaim, M. and Savi, F. (2000). “Application and evaluation of statistical and hydraulic-continuum dense-snow avalanche models to five real european sites”. Cold Regions Science and Technology, 31(2), 133-149. Barbolini, M., Savi, F. (2001). Estimate of uncertainties in avalanche hazard mapping. Annals of Glaciology. 32. pp 299-305. Barbolini, M., Ceriani, E., Del Monte, G., Segor, V. & Savi, F. (2001). “The "Lavanchers" avalanche of February 23th 2000, Aosta valley, Italy”. In: Proceeding of the International Snow Science Workshop: a merging between theory and practice (ISSW 2000). Big Sky, Montana, USA, October 1st - 6th 2000, pp 519-527. Barbolini, M., Natale, L., Savi, F. (2002). Effect of release conditions uncertainty on avalanche hazard mapping, Nat. Hazards, 25, 225-244. Barbolini, M., Cappabianca, F. and Savi, F. (2003). “A new method for the estimation of avalanche distance exceeded probabilities”. Surveys in Geophysics, 24, 5-6, pp. 587 - 601. Barbolini, M., Cappabianca, F. & Savi, F. (2004). “Risk assessment in avalanche prone areas”. Annals of Glaciology, 38, pp.115-122. Barbolini M., Natale L., Cordola M., Tecilla, G. (2004). Linee Guida metodologiche per la perimetrazione delle aree esposte al pericolo di valanghe. Neve e Valanghe, 53, 6-13. (In Italian with abstract in English). Barbolini, M., Cappabianca, F., Frigo, B. & Sailer, R. (2006). “The vulnerability of buildings affected by powder avalanches”. RISK 21: Coping with Risks Due to Natural Hazards in the 21st Century. Balkema Publishers. pp. 227-236. Barsanti, M. (1990). Calcolo della distanza di arresto delle valanghe sulla base di parametri topografici del pendio. Neve e Valanghe, 9, 86-97. Bartelt, P., Gruber, U. and Salm, B. (1997a). Numerical Modelling of Dense Snow Avalanches Using a Voellmy-Fluid Flow Law Solved by a Finite Difference Method. SFISAR Internal Report No. 716, Davos, CH. Bartelt, P., Salm, B. and Gruber, U. (1997b). Modelling dense snow avalanche flow as a Criminale-Ericksen- Filby fluid without choesion. SFISAR Internal Report No. 717, Davos, CH. Bartelt, P., Kern, M. and Christen, M. (2000). A mixed flowing/powder snow avalanche model. In: Proceedings of the International Snow Science Workshop: a merging between theory and practice (ISSW 2000). Big Sky, Montana, USA, October 1st-6th 2000, 280-289. Bélanger, L., Cassayre, Y. (2004). Projets for past avalanche observation and zoning in France, after 1999 catastrophic avalanches. Proceedings of the International Snox Survey Workshop. 19-24 September

141 2004. Jackson Hole, Wyoming. Pp 416-422. Bezzi M., Cantiani M.G., Ciolli M., Comunello G., Cherubini P., 2003. Leggere gli anelli degli alberi per ricostruire la frequenza e l'estensione delle vlanghe nel passato. - In: De Angelis P., Macuz A., Bucci G., Mugnozza G.S. (eds) Atti del III CongressoNazionale: Alberi e foreste per il nuovo millennio. Viterbo, Società Italiana di Selvicoltura ed Ecologia Forestale, pp. 147-152. Blagovechshenskiy, V. (2001). “Estimation of avalanche friction coefficients at Zailiysliy Altau range (Kazakhstan)”. In: Proceedings of the II International Conference on Avalanche and Related Subjects, Kirovsk, Russia, 3–7 September 2001. Bocchiola, D., De Michele, C., Rosso, R. (2003). Review of recent advances in index flood estimation, Hydrology and Earth System Sciences, 7(3), 283-296. Bocchiola, D., Rosso, R., in press. On the distribution of the daily Snow Water Equivalent in central Italian Alps, Adv. Wat. Res. Bocchiola, D., Medagliani, M., Rosso, R., in press. Regional snow depth frequency curves for avalanche hazard mapping in central Italian Alps, Cold Regions Science and Technology. Bohr, G. S., Aguado, E. (2001). Use of April 1 SWE measurements as estimates of peak seasonal snowpack and total cold-season precipitation. Water Resour. Res., 37, 1, 51-60. Bonamy, D., Daviaud, F., Laurent, L., Bonetti, M., Bouchaud, J.P. (2002). Multiscale clustering in granular surface flows . Physical Review Letters, 89(3), 343011-343014. Boucher D., Filion, L., Hétu B. (2004). Reconstitution dendrochronologique et fréquence des grosses avalanches de neige dans un couloir subalpin du mont Hog's Back, Gaspésie centrale (Québec). Géographie Physique et Quaternaire, 58. Bouchet, A., Naaim, M., Bellot, H., Ousset, F. (2003). Experimental study of dense snow avalanche, velocity profiles in quasi-permanent and fully developed flows. Annals of Glaciology, 38, 30-34. Bozhinsky, A.N. and Losev, K.S. (1998). “The fundamentals of Avalanche Science”. Mitteilungen Nr. 55, SLF, Davos. Burkard, A., Salm, B. (1992), Die bestimmung der mittleren anrissmachtigkeit d0 zur berechnung von fliesslawinen [Etimate of the average release depth d0 for the calculation of flowing avalanches], Internal Report of the Swiss Federal Institute for Snow and Avalanche Research, No. 668, Davos, Switzerland. (In German, also available in French under the title: Estimation de l’epaisseur moyenne de declenchement d0 pour le calcul des avalanches coulantes, translated in French by C. Ancey, Cemagref, 1994). Burn, D. H. (1997). Catchment similarity for regional flood frequency analysis using seasonality measures, J. of Hydrol., 202, 212-230. Burrows C.J., Burrows V.L., (1976). Procedures for the study of snow avalanche chronology using growth layers of woody plants. Institute of Arctic and Alpine Research, Occasional Paper no. 23, University of Colorado, pp. 13-24. Buser, O. and Frutiger, H. (1980). Observed maximum run-out distance of snow avalanches and the determination of the friction coefficients μ and ξ. Journal of Glaciology, 26 (94), 121-130. Butler D.R., Malanson G.P., 1985. A reconstruction of snow-avalanche characteristics in Montana, U.S.A., using vegetative indicators. Journal of Glaciology 31(108), pp.185-187. Carrara P.E., 1979. The determination of snow avalanche frequency through tree-ring analysis and historical records at Ophir, Castaldini, R. (1994). Sul calcolo della distanza di arresto delle valanghe. Neve e Valanghe, 21, 50-61. Castellarin, A., Burn, D.H., Brath, A. (2001). Assessing the effectiveness of hydrological similarity measures for flood frequency analysis, J. of Hydrol., 241, 270-285. Cemagref ETNA (2000). Commune de Chamonix Mont Blanc, projet de Centre de secours principal près des Pélerins, étude du risque d’avalanche. Rapport d’expertise, F. Rapin coord. Chen, C.P., Wood, P.E. (1985). A turbulence closure model for dilute gas particle flows. Canadian Journal of Chemical Engineering, 63(3), 349-360. Christen, M., Bartelt, P. and Gruber, U. (2002). “AVAL-1D: An avalanche dynamics program fort he practice”. In: Proceedings of the International Congress Inerpraevent 2002 in the Pacific Rim, 14-18 October, Matsumoto, Japan. Vol. 2, pp. 715-725. Coeur, D., Lang, M., Naulet R., Burnet R., Strazzeri, D., “Histoire et connaissance des phénomènes naturels extrêmes », 1998, Ingénierie – EAT- Risques naturels, p.15-26 Colorado. Geological Society of America Bulletin, 90, pp. 773-780. Comunello G., Bezzi, M., Cherubini P., Ciolli M., Cantiani M.G., 2001. Tecniche GIS e di dendrocronologia applicata per lo studio di aree potenzialmente soggette al rischio di valanghe. - Linea Ecologia - Economia Montana 33, 4, pp. 58-62. Cong, S. Li, Y., Vogel, J., Shaake, J.C. (1993). Identification of the underlying distribution form of precipitation by using regional data, Water Resour. Res., 29, 1103-1111. Darlymple, T. (1960). Flood frequency analysis, U.S. Geological Sur. Water Supply Pap., 1543-A. De Michele, C., Rosso, R. (2001). Uncertainty assessment of regionalized flood frequency estimates, J. Hydrol. Eng., ASCE, 6(6), 453-459. De Michele, C., Rosso, R. (2002). A multi-level approach to flood frequency regionalization, Hydrology and Earth System Sciences, 6(2), 185-194. Dent, J.D. and Lang, T.E. (1980). Modelling of snow flow. Journal of Glaciology, 26 (94), 131-140. Dent, J.D. and Lang, T.E. (1983). A biviscous modified bingham model of snow avalanche motion. Annals of Glaciology 4, 42-46. Eckert, N., Parent, E., Richard, D. (2005). Revisiting statistical–topographical methods for avalanche predetermination: Bayesian modelling for runout distance predictive distribution. Submitted to Cold Regions Science and Technology. Eglit, M.E. (1998). Mathematical modelling of dense avalanche. In: Proceedings of the Anniversary Conference for the 25 Years of Snow Avalanche Research at NGI, Voss, Norway, 12-16 May, 1998, NGI Publications No. 203, 15-18. Fellini, M. (1999). Calcolo della distanza di arresto di valanghe estreme: applicazione di un modello statistico-topografico all’Alta Valtellina ed all’Alta Valmalenco. Tesi di Laurea. Università di Pavia.

142 Fujisawa, K., Tsunaki, R. and Kamiishi, I. (1993). Estimating snow avalanche runout distancefrom topographic data. Annals of Glaciology, 18, 239-244. Furdada,G and Vilaplana, J.M. (1998). Statistical prediction of maximum avalanche runout distances from topographic data in the western Catalan Pyrenees. Annals of Glaciology, 26, 285-288. Galbiati, G.L. and Savi, F. (1996). “Simulazione matematica del ruscellamento e dell’infiltrazione; applicazione alla scala della particella”. In: Scritti dedicati a Giovanni Tournon, presented at the Workshop: I problemi dei grandi comprensori irrigui, Novara I, Giugno 1996, 6-7 (In Italian). Garcia, S. (2002). Analyse de la régularité des observations de l’Enquête Permanente sur les Avalanches. DESS Ingéniérie Mathématique, Université Joseph Fourrier, Grenoble, France. 101 p Garcia S., Bélanger L., Bernard G., Borrel G., Burnet R., Fattori F., Strazzeri S. (2003). Manuel à l’attention des observateurs pour l’Enquête Permanente sur les Avalanches (EPA). Cemagref. Granet-Abisset A.M., Brugnot, G., Avalanches et risques. Regards croisés d’ingénieurs et d’historiens. Grenoble, publications de la MSH-Alpes, 2002, 182p. Grigoryan, S.S. (1979). Novii zakon trenija i mehanizm krupnomasshtabnih gornih obvalov i opolznei (A new law of friction and mechanism for large - scale slag heaps and landslides). Dokl. Akad. Nauk SSSR 244(4), 846-849; English transl. in Soviet Phys. Dokl. 24 (1979). Gubler, H. (1993). “Swiss Avalanche-Dynamics Procedures for Dense Flow Avalanches”. AlpuG, Dr. H. Gubler, Richtstattweg 2, CH-7270 Davos Platz. Harbitz, C. (1998). A survey of computational models for snow avalanche motion. NGI Report No. 581220-1 (Also Deliverable No. 4 of the 4th Framework EU Project SAME: Snow Avalanche Mapping, Model validation and Warning Systems in Europe, Published by the European Commission, EUR19069). Hopf J. (1998). “An Overview of Natural Hazard Zoning with Special Reference to Avalanches”. In: Proceedings of the Anniversary Conference for the 25 Years of Snow Avalanche Research at NGI, Voss, Norway, 12−16 May, 1998, NGI Publications No. 203 Hosking, J. R., Wallis, J. R., Wood, E. F. (1985). An appraisal of the regional flood frequency procedure in the UK flood studies report, Hydrol. Sci. J., 30, 85-109. Hosking, J .R. M., Wallis, J. R. (1993). Some statistics useful in regional frequency analysis, Water Resour. Res., 29(2), 271-281. Hunt, B. (1995). Newtonian fluid mechanics treatment of debris flows and avalanches. Journal of Hydraulic Engineering, 120 (12), 1350-1363. IUGS Working Group on Landslides, Committee on Risk Assessment. (1997). “Quantitative risk assessment for slopes and landslides – The state of the art”. In Landslide risk assessment: Proceedings of the International Workshop on Landslide Risk Assessment, Honolulu, Hawaii, USA, 19-21 February 1997. Edited by Cruden, D.M. and Fell, R. A.A. Balkema, Rotterdam, pp. 3-12. Jamard, A. L., Garcia, S., Bélanger, L. (2002). L’enquête permanente sur les Avalanches (EPA). Statistique descriptive générale des événements et des sites. DESS Ingéniérie Mathématique option Statistique, Université Joseph Fourrier, Grenoble, France. Available online at http://www.avalanches.fr/. 101 p. Jónasson, K., Sigurdson, S. and Arnalds, P. (1999). “Estimation of Avalanche risk”. Vedurstofu Islands n. R99001-UR01, 44 pp. Jónasson, K., Sigurðsson, S.P and Arnalds, P. (1999). Estimation of avalanche risk. Reykjavik, Veðurstofu Islands (Report No. R99001-URO). Jóhannesson, T. (1998). Icelandic avalanche runout models compared with topographical models used in other countries. NGI Publications 203. (25 Years of Snow Avalanche Research at NGI, Anniversary Conference, Voss, Norway, 12 16 May, 1998, Proceedings). Norwegian Geotechnical Institute, Oslo. Johannesson, T. and Arnalds, P. (2001). “Accidents and economic damage due to snow avalanche and landslide in Iceland”. Jokull, 50, 81-94. Katz, R. W., Parlange, M. B., Naveau, P. (2002). Statistics of extremes in hydrology, Adv. Wat. Res., 25, 1287–1304. Keylock, C.J., McClung, D.M. & Magnússon, M.M. (1999). “Avalanche risk mapping by simulation”. J. Glaciol., 45(150), pp. 303-314. Keylock, C. J. and Barbolini, M. (2001). “Snow avalanche impact pressure –vulnerability relations for use in risk assessment”. Canadian Geotechnical Journal, 38,227-238. Kottegoda, N., Rosso, R. (1997). Statistics, Probability and Reliability for Civil and Environmental Engineers, Mc Graw-Hill. Kozik, S.M. (1962). The calculation of snow avalanche motion. Leningrad (in Russian). 76 pp. Lang, T., Dawson and Martinelli, M. (1979a). Application of numerical transient fluid dynamics to snow avalanche flow. Part1. Development of the computer program Avalanch. Journal of Glaciology, 22, 107-115. Lang, T., Dawson and Martinelli, M. (1979b). Application of numerical transient fluid dynamics to snow avalanche flow. Part2. Avalanche modelling and parameter error evaluation. Journal of Glaciology, 22, 117-125. Laternser, M., Schneebeli, M. (2003). Long-term snow climate trends of the Swiss Alps (1931–99), Int. J. Climatol., 23, 733–750. Lied, K. and Bakkehøi, S. (1980). Empirical Calculations of Snow-Avalanche Run-Out Distance Based on Topographic Parametres. Journal of Glaciology, Vol. 26, No. 94, 165- 177. Also in: Norwegian Geotechnical Institute, publication No. 133, 1981. Lied, K. and Toppe, R. (1988). Calculation of maximum snow-avalanche run-out distance by use of digital terrain models. Annals of Glaciology, 13, 164-169. Also in: Norwegian Geotechnical Institute, publication no. 183, 1992. Lied., K, Weiler, C., Bakkeøi, S. and Hopf, J. (1995). Calculation methods for avalanche run-out distance for the Austrian Alps. Norwegian Geothecnical Institute, Report No. 581240-1. McClung, D.M. (2001). Extreme avalanche runout: a comparison of empirical models. Canadian Geotechnical Journal, 38, 1254-1265. McClung, D.M. and Mears, A.I. (1991). Extreme value prediction of snow avalanche runout. Cold Regions

143 Science and Technology, 19, 163-175. MacCormack, R. (1969). “The effect of viscosity in hypervelocity impact cratering”. AIAA Paper 69-354. Madsen, H., Pearson, C. P., Rosbjerg, D. (1997). Comparison of annual maximum series and partial duration series methods for modeling extreme hydrologic events 2. Regional modelling, Water Resour. Res., 33(4), 759–769. Maggioni, M., Gruber, U. (2003). The influence of topographic parameters on avalanche release dimension and frequency, Cold Regions Science and Technology, 37, 407– 419. Martinelli, M. jr. (1986). A test of the avalanche runout equations developed by the Norwegian Geotechnical Institute. Cold Reg. Sci. Technol. 13 (1), 19-33. Martinelli, M.jr., Lang, T.E., and Mears, A.I. (1980). Calculations of avalanche friction coefficients from field data. Journal of Glaciology 26 (94), 109-119. Mellor M. (1978): “Dynamics of snow avalanches”. In: Rockslides and Avalanches. Ed. B.Voight, Amsterdam, Elsevier. 753-792. Meunier, M., Ancey, C. (2004). Towards a conceptual approach to predetermining high-return-period avalanche run-out distances. Journal of Glaciology. 50-169. pp 268-278. McClung, D.M. and Lied, K. (1987). Statistical and geometrical definition of snow avalanche runout. Cold Reg. Sci. Technol., 13 (2), 107-119. McClung, D.M. and Mears, A. 1991. Extreme value prediction of snow avalanche runout. Cold Regions science and technology, 19, 163-175. McClung, D.M. and Schaerer, P. A. (1993). The Avalanche Handbook. Seattle, WA, The Mountaineers. McClung, D.M., Mears, A.I. and Schaerer, P. (1989). Extreme avalanche run-out: Data from four mountain ranges. Annals of Glaciology Vol. 13, 180-184. Mougin, P. (1922). Les avalanches en Savoie. Ministère de l'Agriculture, Direction Générale des Eaux et Forêts, Service des Grandes Forces Hydrauliques, Paris. pp 175–317. Muntan E., Andreu L., Gutierrez E., Martinez P., 2004. Dendrochronological study of the Canal del Roc Roig avalanche path: first results of the Aludex project in the Pyrenees. Annals of Glaciology, Vol. 38, No. 1., pp. 173-179. Naaim, M. (1999). Modelisation des Avalanches Mixtes de Niege Seche. La Houille Blanche, 5, 54-69. Naaim, M. and Ancey, C. (1992). Dense avalanche model. European Summer University,Chamonix, Cemagref Publications, 173-181. Naaim M., Faug T., Naaim-Bouvet F. (2003). Dry granular flow: erosion and deposition modelling. Surveys in Geophysics,24, 569-585. Naaim, M., Naaim-Bouvet, F., Faug, T., Bouchet, A. (2004). Dense snow avalanche modelling: flow, erosion, deposition and obstacle effects. Cold Regions Science and Technology 39. pp 193-204. Natale, L. and Savi, F. (1992). “Propagazione di un’onda di sommersione in un canale vuoto”. Jornadas de enquentro trilateral para el estudio de la idraulica de las ondas de submersion. Zaragoza, Sept. 1992 (In italian). Natale, L., Nettuno, L., and Savi, F. (1994). “Numerical simulation of snow dense avalanche: an hydraulic approach”. In Hamza, M.H. (eds), Proceedings of 24th International Conference on Modelling and Simulations, MS’94, 2-4 May, Pittsburgh, Pennsylvania. Anaheim, IASTED ACTA PRESS, 233-236. Nettuno, L. (1996). La modellazione delle valanghe di neve densa: aspetti modellistici e sperimentali. PhD Thesis, University of Pavia. Norem, H., Irgens, F. and Schieldrop, B. (1987). A continuum model for calculating snow avalanche velocities. Proceedings of Avalanche Formation, Movements and Effects, Davos 1986, IAHS publication 162, 363-378. Also in: Norwegian Geotechnical Institute, report 58120-9. Norem, H., Irgens, F. and Schieldrop, B. (1989). Simulation of Snow-Avalanche Flow in Run-Out Zones. Annals of Glaciology Vol. 13, 218-225. Nuijc, G.L. (1994). “Efficient implementation of non-oscillatory schemes for the computation of free surface flows”. Journal of Hydraulic Research, 33(1), 101-111. Perfettini P., Corona C., 2006. Reconstitution dendrogéomorphologique d’une chronologie spatialisée d’avalanches : l’exemple du dépôt Pierres Jean-Jeanne, Massif de l’oisans, Alpes du Nord, France. Journal of Glaciology (sous presse). Perla, R.I. (1980.) Avalanche Release, Motion and Impact. Dynamics of Snow and Ice Masses,Academic Press, Inc. Perla, R.I., Cheng, T.T. and McClung, D.M. (1980). A Two-Parameter Model of Snow-Avalanche Motion. Journal of Glaciology, 26, (94), 197-207. Potter Jr N., 1969. Tree-ring dating of snow avalanche tracks and the geomorphic activity of avalanches, Northern Absaroka Mountains, Wyoming. Geol. Soc. Amer. Special Paper, 123, pp. 141-165. Quervain de M.R., Crecy de L., Lachapelle, E.R., Losev, K. and Shoda M. (1973). “Avalanche Classification”. In: Proposal of the Working Group on Avalanche Classification of the International Commission on Snow and Ice. Hydrological Sciences Bullettin XVIII 4. Robson, A., Reed, D. (1999). Flood estimation handbook (FEH), Institute of Hydrology, Wallingford, U.K., Vol. 3. Rohrer, M.B., Braun, L.N., Lang, H. (1994). Long Term Records of Snow Cover Water Equivalent in The Swiss Alps: 1. Analysis, Nordic Hydrology, 25, 53-64. Sailer, R. (2003). “Case studies with SAMOS - Comparison with observed avalanches”. AVL - Advanced Simulation Technologies, International User Meeting 2003. Salm, B. (1993). “Flow, Flow transition and runout distances of Flowing Avalanches”. Annals of Glaciology 18, 221-226. Salm, B., Burkard, A. and Gubler, H. (1990). “Calcul des Avalanches: Une Metode pour le Praticien avec des Exemples », SFISAR Message No 47 (Translated in French by Ancey C., 1994). Sampl, P. (1993). Current status of the AVL avalanche simulation model. Numerical simulation of dry snow avalanches. In: Proceedings of Pierre Beghin International Workshop on rapid gravitational mass movements, 6-10 December 1993, Grenoble, France, CEMAGREF Ed., 269-275. Sampl, P. & Zwinger, T. (2004). “Avalanche Simulation with SAMOS”. Annals of Glaciology, 38, 393-398. Savage, S.B. and Hutter, K. (1989). The motion of a finite mass of granular material down a rough incline. J.

144 Fluid Mech., 199, 177-215. Savage, S.B. and Hutter, K. (1991). The dynamics of granular avalanche from initiation to runout. Part I: analysis. Acta Mechanica, 86, 201-223. Savage, S.B. and Hutter, K. (1990). The dynamics of avalanches of granular materials from initiation to runout, Part I: Analysis. Acta Mech. 86, 201-223 (1991). Schweingruber F.H. (1996). Influence of snow. In Tree Rings and Environment. Dendroecology. Swiss Federal Institute for Forest, Snow and Landscape Research, Birmensdorf, Paul Haupt Verlag, Berne. pp. 183-196. Sovilla, B. (2002). Input parameters for dense snow avalanches simulation, [Parametri di input per simulazioni di valanghe dense], In: Course of avalanche dynamics, Eidg. Inst. Schnee- und Lawinenforsch. SLF Davos, Held in Bormio, 4-5 december 2002 (In Italian). Sveinsson O., Boes, G. B., Duane C., Salas, J. D. (2001). Population index flood method for regional frequency analysis, Water Resour. Res., 37(11) , p. 2733-2748. Voellmy A. (1955). “Über die Zerstörungskraft von Lawinen”, Schweiz. Bauzeitung No. 73, 159-165, 212- 217, 246-249, 280-285. Wilhelm, C. (1998). “Quantitative risk analysis for evaluation of avalanche protection projects”. In: Proc. of 25 Years of Snow Avalanche Research at NGI, Voss 12-16 May 1998, Hestness, E. Ed., NGI Wiltshire, S.E. (1986). Regional floods frequency analysis I: Homogeneity statistics, Hydrological Sciences Journal, 31, 3-9.Publications 203, Oslo, pp. 288-293. Zwinger, T. (2000). Dynamik einer Trockenschneelawine auf beliebig geformten Berghängen. Diplomarbeit, Technische Universität Wien, Technische Hochschule Darmstadt.

145