Socio-Economic and Ecological Evaluation and Modelling Methodologies

Integrated Flood Risk Analysis and Management Methodologies

Socio-economic and ecological evaluation and modelling methodologies

Date 29th February 2008

Report Number T10-07-13 Revision Number 1_2_P10

Deliverable Number: D10.1 Due date for deliverable: November 2007 Actual submission date: November 2007 Sue Tapsell FHRC/MU FLOODsite is co-funded by the European Community Sixth Framework Programme for European Research and Technological Development (2002-2006) FLOODsite is an Integrated Project in the Global Change and Eco-systems Sub-Priority Start date March 2004, duration 5 Years Document Dissemination Level PU Public PU 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)

Co-ordinator: HR Wallingford, UK Project Contract No: GOCE-CT-2004-505420 Project website: www.floodsite.net

Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

DOCUMENT INFORMATION

Title Socio-economic and ecological evaluation methodologies Lead Author Sue Tapsell Sally Priest, Dennis Parker, Edmund Penning-Rowsell, Christophe Viavattene, Theresa Wilson, John Handmer - FHRC/MU Arjan Wijdeveld, Marjolein Haasnoot, Reinaldo Penailillo - WL | Delft Hydraulics Contributors Frank van den Ende, RIZA, Dutch Governmental Institute Paul van Noort, , RIZA, Dutch Governmental Institute Frank Messner, Volker Meyer, Dagmar Haase, Sebastian Scheuer, Anne Schildt - UFZ Celine Lutoff, Isabelle Ruin - INPG Distribution Public Document Reference T10-07-13

DOCUMENT HISTORY

Date Revision Prepared by Organisation Approved by Notes 30/11/07 1_0_P10 S. Tapsell FHRC/MU 10/12/07 1_1_P10 S. Tapsell FHRC/MU 29/02/08 1_2_P10 S. Tapsell FHRC/MU 1_3_P01 Paul Samuels HR Final Formatting Wallingford

ACKNOWLEDGEMENT

The work described in this publication was supported by the European Community’s Sixth Framework Programme through the grant to the budget of the Integrated Project FLOODsite, Contract GOCE-CT- 2004-505420.

DISCLAIMER

This document reflects only the authors’ views and not those of the European Community. This work may rely on data from sources external to the FLOODsite project Consortium. Members of the Consortium do not accept liability for loss or damage suffered by any third party as a result of errors or inaccuracies in such data. The information in this document is provided “as is” and no guarantee or warranty is given that the information is fit for any particular purpose. The user thereof uses the information at its sole risk and neither the European Community nor any member of the FLOODsite Consortium is liable for any use that may be made of the information.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 ii Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

SUMMARY

Theme 1 of the FLOODsite project provides new knowledge and understanding to derive risk analyses for flood prone areas. With the recognition that it is not possible to protect from all floods, innovative ways need to be found to manage the associated risks and likely consequences. The overall objective of Task 10 within this Theme is to focus research efforts on innovative methods to understand, model and evaluate socio-economic and ecological flood damages; this report presents the results of key research from four projects (Activities) conducted under Task 10 of the FLOODsite project. These reports are:

1. Building a model to estimate Risk to Life for European flood events (project document T10- 07-10) 2. Modelling the damage-reducing effects of flood warnings (project document T10-07-12) 3. Toxic Stress: the development and use of the OMEGA modelling framework in a case study (project document T10-07-14) 4. GIS-based Multicriteria Analysis as Decision Support in Flood Risk Mangement (project document T10-07-06).

Each of these Activities comprises a separate, and substantial, research project in its own right encompassing the use of different disciplines and approaches and different teams of researchers. Detailed Milestone reports have been produced for each of the Activities. These reports should be referred to for full accounts of the individual research projects, including full literature reviews, research methodologies, results and discussions.

The research covered in this Deliverable report encompasses research on socio-economic and ecological evaluation and modelling methodologies focusing on risk receptors, that is: people, buildings and the environment. Activity 1 focuses very much on the social impacts and damages – those to human life and health. The emphasis in Activity 2 is largely on assessing economic and financial damages and the benefits of flood warning systems. Activity 3 focuses specifically on ecological damages to the natural environment, including habitats and species. Finally, Activity 4 focuses on developing a multi-criteria approach to evaluation of various types of flood damages, which can include those addressed within the other Activities, thus drawing together the diverse elements of the research Task.

The research objectives also directly relate to the European Directive on the Assessment and Management of Flood Risks (EU 2007/60/EC of 23 October 2007), in particular with reference to Articles 1, 4, 6 and 7.

The overall combined results from the research should lead to a better understanding and quantification of flood impacts and therefore the provision of evaluation methodologies, techniques and approaches to guide end-users in decisions on levels of investment, preparedness planning and emergency response strategies in future flood risk management across Europe.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 iii Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 iv Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

CONTENTS

Document Information ...... ii Document History ...... ii Acknowledgement...... ii Disclaimer ...... ii Summary ...... iii Contents ...... v

1 Introduction ...... 13 1.1 Background...... 13 1.2 Purpose and objectives ...... 14 1.2.1 European Floods Directive...... 14 1.3 Structure of report...... 15

2 Activity 1: Building models to estimate loss of life for flood events...... 16 2.1 Introduction and Context ...... 16 2.2 Aims of Activity 1 ...... 16 2.3 Structure...... 17 2.4 Floods and risk to life or injury ...... 17 2.5 Factors affecting cause of death or injury...... 18 2.5.1 Examples of common flood scenarios in Europe...... 20 2.6 Methods to calculate flood risks to people ...... 20 2.6.1 The ‘Flood risks to people’ methodology ...... 21 2.6.2 Quantifying the risk to people for a single event ...... 24 2.6.3 Risk to People mapping ...... 24 2.7 Initial model adaptation of the Risk to People model to the European context...... 27 2.8 Methodology and data collection...... 28 2.8.1 Data collection methods...... 28 2.8.2 Limitations and problems arising from data collection...... 29 2.8.3 Usability of Continental European data for Risk to People methodology ...... 29 2.9 Analysis of the circumstances and causes of European flood-related deaths...... 32 2.9.1 Reporting of fatalities...... 32 2.9.2 Circumstances leading to death from flooding...... 32 2.9.3 Characteristics of datasets with no fatalities ...... 37 2.10 Additional case studies ...... 41 2.10.1 The Gard flood of 2002, France...... 41 2.10.2 The Boscastle flood of 2004, UK...... 41 2.11 Calibration of the UK Risk to People Model...... 44 2.11.1 Sensitivity analysis...... 44 2.11.2 Model application...... 45 2.11.3 Summary of limitations of the current Risk to People model for application in a European context ...... 46 2.12 Adaptations and revisions to the UK Risk to People model...... 47 2.13 Recommendations for refining the UK Risk to People methodology ...... 49 2.14 Proposed European Risk to Life model ...... 50 2.14.1 Conceptual model...... 50 2.14.2 Hazard factors ...... 52 2.14.3 People exposure ...... 54 2.14.4 Risk to Life from flooding ...... 58 2.14.5 Mitigating factors...... 60 2.14.6 A new approach to assessing Risk to Life from flooding in Europe...... 61 2.14.7 Application of the threshold European Risk to Life model to the

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 v Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

European flood event data...... 63 2.15 Mapping risk to life ...... 64 2.16 Conclusions and recommendations for further research...... 67 2.16.1 Summary of research...... 67 2.16.2 Remaining issues...... 68 2.16.3 Recommendations for further research ...... 68

3 Activity 2 Developing models to estimate the benefits from flood warnings...... 70 3.1 Introduction ...... 70 3.1.1 Floods and flood losses in Europe: the background to the development of flood warning systems ...... 70 3.1.2 A revolution in flood warning systems in Europe...... 70 3.2 Different disciplinary perspectives and their potential contributions to understanding and improving flood warnings ...... 73 3.2.1 The flood loss reducing effects of warnings ...... 75 3.2.2 Project objectives and deliverables ...... 76 3.3 Methodological approach ...... 77 3.3.1 Introduction...... 77 3.3.2 Data constraints on research and application of research results ...... 77 3.3.3 Rethinking and reconceptualisation ...... 77 3.3.4 Re-analysis of existing social survey data...... 77 3.3.5 Partner survey...... 79 3.3.6 The expert interview tool ...... 79 3.3.7 Secondary data ...... 80 3.3.8 Case Studies ...... 80 3.3.9 Methodological implications of data constraints...... 81 3.4 Flood Warning Response...... 81 3.4.1 Introduction...... 81 3.4.2 Different physical and social contexts are critical in understanding flood warning response ...... 81 3.4.3 Factors likely to encourage or inhibit flood warning response and features of flood warning systems which are likely to enable people to respond more effectively to warnings...... 82 3.4.4 Weighing the significance of the evidence...... 83 3.4.5 The complexities and similarities of flood warning response...... 84 3.4.6 Factors determining flood warning response effectiveness ...... 85 3.5 The UK FHRC model for estimating the flood losses avoided by flood warnings ...... 86 3.5.1 The 1991 model of damage saving generated by flood warnings...... 86 3.5.2 The Environment Agency (2003) variant...... 86 3.5.3 The limitations of the UK FHRC model ...... 87 3.5.4 The latest survey evidence ...... 88 3.5.5 Conclusions...... 89 3.6 European variants of the UK FHRC model...... 90 3.6.1 Re-conceptualising the UK FHRC model for continental Europe ...... 91 3.6.2 Application of the EU FHRC model ...... 93 3.6.3 Data availability and proxies...... 95 3.7 Strengths and shortcomings of the existing UK/EU FHRC model...... 95 3.8 Modelling the wider response to flood warnings: The Flood Warning Response and Benefits Pathways Model (FWRBP model) ...... 99 3.8.2 Opportunities and limitations associated with the ‘pathways’ model...... 105 3.8.3 How the model may inform flood warning system enhancement...... 106 3.9 Case Study: Modelling the flood damage reducing effects of flood warnings in Grimma, River Mulde, Germany ...... 106 3.9.1 Case study context and data ...... 106 3.9.2 Flood management in Germany ...... 109

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 vi Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

3.9.3 The current Saxon flood warning system...... 109 3.9.4 The Grimma Flood Warning System ...... 111 3.9.5 Flood damages in Grimma ...... 112 3.9.6 Post flood response and recovery in Grimma ...... 114 3.9.7 Application of the EU FHRC model to Grimma...... 115 3.9.8 Application of the Flood Warning Response Benefits Pathway (FWRBP) model to Grimma, River Mulde...... 121 3.9.9 Discussion and conclusion ...... 128 3.10 Principal findings and conclusions ...... 129 3.10.1 The purpose and methods of the research project ...... 129 3.10.2 The evolution of flood warning systems in Europe ...... 130 3.10.3 The estimation of flood warning benefits ...... 131 3.10.4 The application of flood loss reduction models to case studies ...... 132 3.10.5 Enabling a better response to flood warnings ...... 132 3.10.6 Methods for collecting data...... 136 3.10.7 The main scientific progress and advances made by this research project ...137 3.10.8 Gaps in research knowledge...... 137

4 Activity 3: The effects of floods and flood-induced pollution on ecosystem health ...... 139 4.1 Introduction to the OMEGA modelling framework ...... 139 4.2 Risk assessment based on species sensitivity distribution...... 139 4.3 Modelling framework ...... 140 4.4 New developments ...... 141 4.5 From effect on the total population to effect on specific groups ...... 141 4.6 Combination of toxic effects of different toxicants / Mode of Action ...... 142 4.7 Western Scheldt case study: toxic risk prediction in surface water...... 143 4.7.1 Introduction...... 143 4.7.2 Materials and methods ...... 143 4.7.3 Results and discussion...... 146 4.7.4 Conclusions on the Western Scheldt case study ...... 152 4.8 Middelburg Case Study: Spreading of polluted sediments due to flooding ...... 153 4.8.1 Introduction...... 153 4.8.2 Methods and Materials...... 154 4.8.3 Simulation results...... 165 4.8.4 Recommendations...... 170 5 Activity 4: Developing methodological foundations for GIS-based multicriteria evaluation of flood damage and risk (UFZ) ...... 171 5.1 Background...... 171 5.2 Research Results...... 172 5.2.1 Problem Definition...... 172 5.2.2 Selecting Evaluation Criteria ...... 173 5.2.3 Alternatives ...... 173 5.2.4 Criteria Evaluation: Risk Maps...... 174 5.2.5 Criterion Weights...... 179 5.2.6 Decision Rules ...... 179 5.2.7 Results & Sensitivity...... 179 5.2.8 Outlook: Multicriteria project appraisal...... 183 5.2.9 Conclusions...... 183

6 Overall conclusions...... 185

7 References ...... 187

8 Appendix A: Data collection template ...... 202

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 vii Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Tables Table 2.1: Area Vulnerability scores ...... 23 Table 2.2: Flood warning components applied to the Risk to People model for Continental European case studies ...... 27 Table 2.3: List of flood event data received (in chronological order)...... 31 Table 2.4: Causes of death in the UK calibration events ...... 35 Table 2.5: Causes of fatality in the European events...... 36 Table 2.6: Prominent factors affecting death in both the UK and wider European flood events (XX = Most important factors) ...... 38 Table 2.7: Data for those events where no deaths occurred - key characteristics are shown in bold ...40 Table 2.8: Sensitivity Analysis – Nz and Hazard Rating...... 44 Table 2.9: Sensitivity Analysis – Area Vulnerability ...... 44 Table 2.10: Sensitivity Analysis – People Vulnerability ...... 45 Table 2.11: Summary of Risks to People Model results compared to actual fatalities in Europe ...... 46 Table 2.12: The different variables included within the statistical analyses...... 48 Table 2.13: Flood hazard thresholds as a function of depth and velocity...... 53 Table 2.14: Flood hazard thresholds as a function of depth and velocity...... 54 Table 2.15: Critical parameters for damage to motor vehicles applied to dam break flooding ...... 54 Table 2.16: Flood conditions leading to the partial or total damage of buildings in Finland ...... 55 Table 2.17: Instances of building collapse within the European flood events...... 56 Table 2.18: Flood hazard thresholds as a function of depth and velocity...... 57 Table 2.19: Categories indicating an area’s vulnerability to flood waters...... 57 Table 2.20: Main factors leading to fatalities from flooding ...... 58 Table 2.21: Categories of mitigating actions ...... 61 Table 2.21: Application of the threshold model to European flood events...... 63 Table 3.1: Examples and case studies used for the testing and refinement of the models and the wider approach...... 80 Table 3.2: Flood warning equation parameters...... 87 Table 3.3: Flood warning damage reduction example...... 89 Table 3.4: Potential data sources and proxies for the components within the EU FHRC model...... 97 Table 3.5: Three categories of potential responses to flood warnings which have the potential to generate benefits of flood warnings (i.e. flood damages avoided) ...... 98 Table 3.6: Eight pathways of possible responses to flood warnings which have the potential to generate benefits of flood warnings (i.e. flood damages or human losses avoided)...... 101 Table 3.7: The four stages of the German flood warning system ...... 110 Table 3.8: Estimated flood damages experienced in Grimma during the 2002 flood event ...... 113 Table 3.9: Meso-scale damage estimation of average annual damages for Grimma town centre .....113 Table 3.10: Sources of recovery from damages sustained in the 2002 floods...... 114 Table 3.11: Values to apply to the EU FHRC model...... 118 Table 3.12: Recalculation of the damage savings possible through moving of contents in Grimma ..121 Table 3.13: The differential importance of the flood warning and benefit pathways in floods of different magnitude in Grimma, Germany ...... 122 Table 3.14: Information concerning precautionary measures undertaken by households located on the Elbe tributaries...... 126 Table 3.15: Application of the FWRBP model for Grimma...... 127 Table 3.16: Application of the FWRBP model to Grimma with the proposed flood protection scheme installed...... 128 Table 3.17: Prescription for improving flood warning response ...... 135 Table 4.1: General relation of the processes in the heavy metal model for copper...... 158 Table 4.2: Initial conditions in the water quality model...... 160 Table 4.3: Boundary conditions for the heavy metals during the January 2010 event...... 161 Table 4.4: Boundary conditions for suspended solids during the January 2010 event...... 161 Table 4.5: Total amount of sediment in the model...... 161 Table 4.6: Gross yearly emissions (kg) of Cadmium, Copper and Zinc in Middelburg...... 161

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 viii Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 4.7: Gross yearly emissions (kg) of Cadmium, Copper and Zinc in five built-up areas...... 162 Table 4.8: Gross yearly emissions (g) of heavy metals from the wear of copper rails and oil leaks.162 Table 4.9: Cadmium and copper emissions from artificial and pig manure ...... 162 Table 4.10: Load of metals coming from a pump station in a calamity...... 162 Table 4.11: Yearly loads (g) of heavy metals from industries...... 163 Table 4.12: Calculated volumes (m3) of the inundated areas. The water depth in the 2D plane is calculated with the hydrodynamic model. The area is estimated according to the grid...... 163 Table 4.13: Total concentrations of heavy metals (g/m3) at the different areas (100 times amplificatied)...... 163 Table 4.14: Kd coefficients of metals for sand, silt and clay...... 164 Table 4.15: Process coefficients in the sediment model, with default (non-calibrated) coefficient values...... 164 Table 4.16: Mass balance for the sediment in the 2D model for both stages (January 15th and January 29th)...... 167 Table 4.17: Distribution (percentage) of the three sediment fractions during both stages...... 167 Table 4.18: Mass balance for the heavy metals in the 2D model for both stages (January 15th and January 29th)...... 170 Table 4.19: Distribution (percentage) of the heavy metals during both stages...... 170 Table 4.20: Parameters in the sediment model ...... 170 Table 5.1: Criteria of environmental risk assessment ...... 176

Figures Figure 2.1: Risk to People model flood injuries formula ...... 22 Figure 2.2: Risk to People model flood fatalities formula ...... 22 Figure 2.3: Risk to People model hazard rating ...... 22 Figure 2.3: Risk to people model hazard rating formula...... 22 Figure 2.4: Risk to People model area vulnerability score...... 23 Figure 2.5: Risk to People model people vulnerability score...... 23 Figure 2.6: Formula for calculating the number of deaths and/or injuries...... 24 Figure 2.7: Example of the application of the Risk to People methodology...... 25 Figure 2.8: Steps for mapping, after HR Wallingford (2005a, p36-48) and HR Wallingford ...... 26 Figure 2.9: Map of locations where flood event data have been gathered ...... 30 Figure 2.10: Circumstances surrounding flood-related death broken down by age for the European flood events ...... 37 Figure 2.11: Photograph taken in the centre of the village during the Boscastle flood...... 42 Figure 2.12: Example of some of the debris moved by the floods in the Boscastle flood event...... 42 Figure 2.13: Members of the public rescuing others from the floodwaters in Boscastle ...... 43 Figure 2.14: Expression characterising the effects on people exposed to the flooding risk...... 50 Figure 2.15: Method for calculating flood risks to people ...... 51 Figure 2.16: Proposed conceptual model for assessing risk to life ...... 51 Figure 2.17: Loss of stability figures taken from Abt et al. (1989) and Karvonen et al. (2000)...... 52 Figure 2.18: Recalculation from the HR Wallingford (2005a) of the ‘danger’ thresholds for a range of different depths and velocities...... 53 Figure 2.19: First half of threshold model indicating the risk of life from flooding...... 59 Figure 2.20: Threshold approach to assessing Risk to Life from flooding in Europe...... 62 Figure 3.1: The principal elements of a flood forecasting, warning and response system...... 71 Figure 3.2: A perspective on the positioning of flood forecasting and warning systems within flood management measures in general...... 72 Figure 3.3: The versatility of structural and non-structural flood mitigation measures ...... 73 Figure 3.4: Floodplain user’s perceptions, attitudes and warning response behaviours ...... 74 Figure 3.5: The complex information and data flows between actors which create major challenges within flood forecasting, warning and response systems as science and technology advances ...... 74

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 ix Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 3.6: The losses and damages caused by floods which may be reduced by flood warnings ...... 75 Figure 3.7: Improving the effectiveness of a flood warning system over time...... 76 Figure 3.8: Diagram illustrating the research design...... 78 Figure 3.9: The variables which determine flood warning response either by inhibiting or enabling response by individuals occupying flood prone locations ...... 85 Figure 3.10: Potential evacuation scenarios affecting people’s ability to save losses due to flood warning, including damages to property contents...... 93 Figure 3.11: Potential scale of flood warning benefit components...... 99 Figure 3.12: Flood Warning Response and Benefit Pathways (FWRBP) model...... 100 Figure 3.13: Flood warning responses, benefits and costs ...... 103 Figure 3.14: Map showing Grimma within the Mulde catchment, and also the other Elbe tributaries.107 Figure 3.15: Graph highlighting the most serious flood events affecting Grimma and their maximum flood depth (water level in cm)...... 108 Figure 3.16: The reporting and information chain for flood related messages in Saxony ...... 109 Figure 3.17: Screenshot from webcam of the level of the River Mulde at the Pöppelmannbrücke in Grimma ...... 111 Figure 3.18: Images of flood damage in Grimma during and after the 2002 flood...... 112 Figure 3.19: Components of the EU FHRC model ...... 115 Figure 3.20: Histogram from Merz et al. (2004) illustrating the mean damage (total damage, damage to building structure, damaged to fixed inventory, damage to movable inventory) for different economic sectors...... 116 Figure 3.21: The emergency measures performed, in descending order, as a percentage of all interviewed people for the Elbe, Elbe tributaries and Danube (multiple answers possible)...... 118 Figure 3.22: The reasons why people did not undertake emergency measures during the 2002 flood (multiple answers are possible) ...... 119 Figure 3.23: Indicative depth-damage curve illustrating the proportion of the average annual damages that is more difficult to save in deep floods through moving ...... 120 Figure 3.24: Inundation model showing expected flood depths in the 1 in 200 year flood event in Grimma ...... 120 Figure 3.25: The planned location of proposed structural flood defences for Grimma ...... 123 Figure 3.26: Breakdown of estimated flood damages highlighting the damages caused by the 2002 flood in the Elbe catchment (after Kreibich et al. (2007; 12))...... 125 Figure 4.1: Combining individual NOEC levels ...... 140 Figure 4.2: From NOEC’s to Potential Affected Fraction (PAF) ...... 140 Figure 4.3: Modelling framework for determining critical concentrations (bioaccumulation) in aquatic biota (OMEGA)...... 141 Figure 4.4: Flow schema of OMEGA input and output, including QSAR relations and decision rules...... 142 Figure 4.5: Monitoring points water quantity and quality Western Scheldt ...... 144 Figure 4.6: Trends in heavy metal concentrations at Schaar van Ouden Doel (2000)...... 144 Figure 4.7: Trends in organic pollutants concentrations at Schaar van Ouden Doel (2000)...... 145 Figure 4.8: Modelled Western Scheldt area ...... 145 Figure 4.9: Measured (dots) and calculated concentrations (lines) at eastern calibration point (Schaar van Ouden Doel)...... 147 Figure 4.10: Example of zinc concentration in the middle of the Western Scheldt during one year (2000), including tidal fluctuation ...... 147 Figure 4.11: Example of zinc during the year and as function of the place ...... 148 Figure 4.12: Zink concentrations on 14-02-2000 (spring peak) Figure 4.13: Zink concentrations on 11-10-2000 (autumn low) ...... 149 Figure 4.14: Copper conc. on 14-02-2000 (spring peak) Figure 4.15: Copper conc. on 11-10-2000 (autumn low)...... 149 Figure 4.16: Verdronken Land van Saeftinghe’...... 150 Figure 4.17: Dissolved concentrations for reference point ‘Verdronken Land van Saeftinghe’...... 150 Figure 4.18: Comparison of chronic versus acute stress, August 2000...... 151

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 x Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 4.19: ms-PAF for chronic exposure for different groups of organisms and total-ms PAF, location Land van Saeftinghe ...... 151 Figure 4.20: Contribution of single toxicants to the ms-PAF for chronic exposure at different moments in time, location Land van Saeftinghe...... 152 Figure 4.21: Overview of the study area Middelburg ...... 153 Figure 4.22: Elevations in the study area (right). Inundated area (left)...... 154 Figure 4.23: Water level at the inflow boundary of the 1D model...... 155 Figure 4.24: Configuration of the Sobek modules that are used for the sediment transport calculations...... 156 Figure 4.25: Re-suspension –sedimentation in the sediment model...... 157 Figure 4.26: Model for copper in the water and in the sediment...... 158 Figure 4.27: Map of risks for Middelburg and surroundings (www.risikokaart.nl)...... 159 Figure 4.28: Location of sources of heavy metal emissions. D-locations are considered urban locations, R-locations are agricultural locations, A and B are industries...... 159 Figure 4.29: Tangential tension results on January 15th and January 29th...... 165 Figure 4.30: Amount of freshly deposited sand (g) on January 15th and January 29th. The sedimentation velocity of sand is 30 m/d...... 166 Figure 4.31: Amount of freshly deposited silt (g) on January 15th and January 29th. The sedimentation velocity of silt is 7 m/d...... 166 Figure 4.32: Amount of freshly deposited clay (g) on January 15th and January 29th. The sedimentation velocity of clay is 0.3 m/d...... 166 Figure 4.33: The thickness (m) of the freshly formed sediment layer on January 15th and January 29th...... 168 Figure 4.34: Amount of freshly deposited cadmium (g) on January 15th and January 29th...... 168 Figure 4.35: Amount of freshly deposited copper (g) on January 15th and January 29th...... 169 Figure 4.36: Amount of freshly deposited zinc (g) on January 15th and January 29th...... 169 Figure 5.1: Damage-probability curve ...... 174 Figure 5.2: Expected inundation depth for a 200-year flood event (City of Grimma)...... 175 Figure 5.3: Annual Average Damage (AAD) (City of Grimma): mean estimation ...... 176 Figure 5.4: Environmental risk (City of Grimma): standardised values (0-1) ...... 177 Figure 5.5: Annual affected population (City of Grimma): ...... 178 Figure 5.6: Social hot spots at risk and their probability of being flooded (City of Grimma): ...... 178 Figure 5.7: Example for selected “high risk areas” by the disjunctive approach...... 180 Figure 5.8: Standardised multicriteria risk: large weight on economic & population criteria (40% each)...... 181 Figure 5.9: Standardised multicriteria risk - criteria score sensitivity: minimum value of annual average damage (weights as Figure 5.8) ...... 181 Figure 5.10: Standardised multicriteria risk - criteria score sensitivity: maximum value of annual average damage (weights as fig. 4.8) ...... 182 Figure 5.11: Standardised multicriteria risk - weight sensitivity: large weight on environmental criterion (0.625) ...... 182 Figure 5.12: Change in standardised multicriteria risk due to HWSK-measures (weights as 4.8) ...... 183

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 xi Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 xii Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

1 Introduction

1.1 Background

The extent of flooding is expected to increase over the next 50 to 100 years owing to the effects of climate change and global warming (Kundzewicz et al., 2005; IPCC, 2007; Stern, 2007). As a result of this, regional changes to flood distribution may mean that areas not previously affected by flooding may become newly afflicted (Few, 2006). With the recognition that it is not possible to protect from all floods, we need to find innovative ways to manage the associated risks and likely consequences. Within this context there is a growing awareness that floods are not just about damage to buildings, the environment and economy but that they are also “people problems” as well as “water problems” and that human behaviour can be adapted to mitigate the effects that floods can bring. A clear understanding of the likely risks and impacts of future flooding, the potential benefits of mitigating these impacts and appropriate methodologies to evaluate the impacts are therefore necessary to inform future investment in flood risk management options.

The research covered in this report encompasses socio-economic and ecological evaluation and modelling methodologies. In particular, with reference to the Source-Pathway-Receptor risk approach developed within the Foresight Future Flooding project (Evans et al., 2004), this research focuses on methodologies to determine damages and losses to receptors, that is: people, buildings and the environment. Figure 1.1 outlines the four components or Activities included in this Task as outlined in the Research Implementation Plan.

Task 10: Socio-economic evaluation and modelling

Activity 1 Activity 2 Activity 3 Activity 4 Leader: MU/FHRC Leader: MU/FHRC Leader: WL/Delft Leader: UFZ Model for estimation Damage reducing GIS-based Flood induced pollution of loss of life effects of warnings multicriteria evaluation

Action 1 Phase I - factors Action 1 Data collection Action 1 State-of-the-art on Action 1 Literature leading to risk for floods and environmental vulnera- review of GIS- (MU/FHRC) flood warnings bility (WL/Delft) based MCA Action 2 Phase II - (MU/FHCR) Action 2 State-of-the-art approaches relation of risks Action 2 Application of ecotoxicological models (UFZ) to people and database to (WL/Delft) Action 2 Setup of hazards selected pilot Action 3 Expert meeting on methodology (MU/FHRC) studies. further steps (WL/Delft) (UFZ) Action 3 Phase III - Calibration of Action 4 Definition of relevant Action 3 Exemplary GIS calibration of model in a indicators for ecological land register model Europe-wide quality (WL/Delft) (UFZ) (MU/FHRC) context Action 5 Case study and Expert Action 4 Link Action 4 Phase IV - flood (MU/FHRC) meeting on further steps methodology to risk maps incl. (WL/Delft) the GIS Model database (UFZ) (MU/FHRC)

Task 10 will improve the evaluation of flood damage by (i) refining a loss-of-life model and developing a methodology for mapping life risks; (ii) advancing a model to estimate the economic benefits from flood forecasting and warning systems; (iii) modelling the effects of floods on ecological systems and species; and (iv) multi-criteria evaluation to evaluate flood risk and flood damage under uncertainty Figure 1.1 Structure of Task 10

Each of these Activities comprise a separate, and substantial, research project in their own right encompassing the use of different disciplines and approaches and different teams of researchers.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 13 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Detailed Milestone reports have been produced for each of the Activities (see Priest et al., 2007; Parker et al., 2007; Wijdeveld et al., 2007; Meyer, Scheuer & Haase, 2007). These reports contain full accounts of the individual research projects, including full literature reviews, research methodology, results and discussions. This Deliverable report comprises abridged versions of each of these reports.

Two additional reports are also related to Task 10. The first is a case study of the 2002 flooding in the Gard region of France (Lutoff and Ruin, 2007) and is based on research conducted for a PhD thesis (see Ruin 2007). The second report (Tapsell et al., 2008) focuses on understanding the complex health impacts of floods and presents a conceptual model of the various factors impacting human health and well-being with respect to flooding. This research was originally intended to be a part of the work for the Risk to Life model in Activity 1, but on reflection it was decided that it should be reported separately. This report is due to be completed in April 2008.

1.2 Purpose and objectives

Theme 1 of the FLOODsite project will provide new knowledge and understanding to derive risk analyses for flood prone areas. The overall objective of Task 10 within this Theme is to focus research efforts on innovative methods to understand, model and evaluate flood socio-economic and ecological damages. Four sub-objectives (outlined in Figure 1.1 above) are important in this context:

1. The first is to better understand the process of loss of life in flood events, to develop a model to estimate risk to life and to develop a methodology for mapping such risks.

2. The second objective is to produce a model to estimate the effectiveness of flood forecasting and warning systems in terms of damages saved, using case study data. The results will help to provide information on whether further investment should go to technical flood forecasting systems or into dissemination arrangements, or into efforts to strengthen social resilience of local communities.

3. Thirdly, the effects of floods on ecological systems and species will be examined and modelled while considering the effects of inundation and pollution loads.

4. Fourthly, since the evaluation of flood damages and flood risk must often be done with several criteria, and with a high degree of uncertainty involved, research will be conducted on the methodological foundation of multi-criteria evaluation to evaluate flood damage under uncertainty, referring in particular to GIS-based damage models.

Activity 1 focuses very much on the social impacts and damages – those to human life and health. The emphasis in Activity 2 is largely on assessing economic and financial damages and the benefits of flood warning systems. Activity 3 focuses specifically on ecological damages to the natural environment, including habitats and species. Finally, Activity 4 focuses on developing a multi-criteria approach to evaluation of various types of flood damages, which can include those addressed within the other Activities, thus drawing together the diverse elements of the research Task.

1.2.1 European Floods Directive The above objectives also directly relate to the European Directive on the Assessment and Management of Flood Risks (EU 2007/60/EC of 23 October 2007). In particular, the research addresses the Directive in a number of ways as follows:

• Article 1: by contributing innovative evaluation and modelling methodologies “aiming at reduction of the adverse consequences for human health, the environment, cultural heritage and economic activity associated with floods in the Community”.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 14 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

• Article 4 (2b): by providing “a description of the floods which have occurred in the past and which had significant adverse impacts on human health, the environment, cultural heritage and economic activity, and for which the likelihood of similar future events is still relevant, including their flood extent and conveyance routes and an assessment of the adverse impacts they have entailed”.

• Article 4 (2d): by providing “an assessment of the potential adverse consequences of future floods for human health, the environment, cultural heritage and economic activity”.

• Article 6 (2, 4, 5): by “the preparation of flood risk maps for at-risk areas showing such elements as: flood extent, depths and flow velocity, potential adverse consequences expressed in terms of indicative number of inhabitants potentially affected, type of economic activity, information on floods with a high content of transported sediments and other significant sources of pollution” and other factors.

• Article 7 (3): by providing methodologies for use in flood risk management plans which “take into account such relevant aspects as costs and benefits”.

1.3 Structure of report

The remainder of the report is therefore structured in key four parts, each focusing on one of the Task objectives outlined above. Section 2 reports on the development of a Risk to Life model to estimate loss of life from flood events within the European context. This is followed in Section 3 by the development of a new methodology to estimate the diverse benefits of flood warning systems. Section 4 reports on developing a model framework to predict the eco-toxicological stress levels during flooding and to determine the spreading of pollutants. Section 5 discusses GIS-based multi-criteria analysis as decision support in flood risk management. The final section briefly attempts to draw the research together and to offer some overall conclusions.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 15 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

2 Activity 1: Building models to estimate loss of life for flood events

2.1 Introduction and Context

Floods are some of the most frequently reported and costly natural disasters world-wide (Hewitt, 1997). They are also one of the most widespread climatic hazards that pose multiple risks to human lives, infrastructure, economies and the natural environment. With the frequency and intensity of flooding predicted to increase in the future, regional changes to flood distribution may mean that areas not previously considered to be at risk may become newly afflicted (Few, 2006). Floods may also result from a range of different causes, for example, intense or prolonged rainfall, snow and ice melt, tidal surges, dyke failure, incapacity of drainage systems, and rising ground water. Some flood events may even involve a combination of one or more of these types of flooding e.g. fluvial and tidal, further complicating the situation. In addition, floods may be fast onset e.g. within an hour or two in rapid response catchments, or slow onset over many hours or even days in the lower floodplains. These differences in the type of flood events can have serious implications for the risk to human life. For example, in England and Wales river floods have been typically small-scale, short- lived and shallow resulting in few deaths, while in other parts of continental Europe where the river systems are much larger, such as Germany, flooding may be deep and spread over large areas and last for many days or weeks posing greater risk to life. In northern Italy, floods in the mountainous areas may pose additional problems as they may contain large amounts of mud and other large debris.

Along with the increase in the frequency of flood events in recent years there has been a rise in the numbers of deaths reported and attributed to flooding. Yet, to date, we know very little about the likely loss of life in floods, and the various causes. We do not yet have appropriate techniques that predict the incidence of loss of life in floods, or the potential for flood mitigation measures to reduce this loss. Therefore, in order to reduce the risk to life it is necessary to understand the causes of loss of life in floods in order to pinpoint where, when and how loss of life is more likely to occur and what kind of intervention may be effective. In particular, with reference to the Source-Pathway-Receptor risk approach developed within the Foresight Future Flooding project (Evans et al., 2004), this research focuses on methodologies to determine damages and losses to human receptors.

2.2 Aims of Activity 1

The results reported here comprise an abridged version of Milestone 10.1 (Priest et al., 2007) which reports the full outputs from Activity 1 of Task 10. This Activity focused on further developing a methodology to estimate risk to human life and serious injury from flood events. The objectives of the research were therefore:

• to map, through the use of GIS and building partly on existing work, the outputs of the risk to life model(s) providing estimates of the potential loss of life in floods.

It was further aimed to produce risk to life models that are usable at different scales. This flexibility was thought to be essential as not all European countries have detailed flood data that is readily available. It needs to be noted that although Europe experiences many different types of flooding, this research only examines risk to life and health impacts related to fluvial flooding. Coastal or other types of flooding have not been included due to time and funding constraints.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 16 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

The above objectives also directly relate to the European Directive on the Assessment and Management of Flood Risks (EU 2007/60/EC of 23 October 2007). In particular, the research addresses the Directive in a number of ways as follows:

• Article 4 (2d): by providing “an assessment of the potential adverse consequences of future floods for human health, the environment, cultural heritage and economic activity”.

• Article 7 (3): by providing methodologies for use in flood risk management plans which “take into account such relevant aspects as costs and benefits”

2.3 Structure

This section of the report begins with a very brief review of some relevant literature on types of floods and factors affecting potential risk to human life. A full literature review is included in Priest et al. (2007). This research has taken as a starting point a methodology developed in the UK by HR Wallingford et al. (2003; 2005) which calculates flood risk to people from river and coastal flooding. The ‘Flood risks to people’ methodology is thus explained along with its application to date and adaptations of the model carried out for this research. The methodology used for this research is outlined along with the limitations and problems arising from data collection. The circumstances and causes of European flood-related deaths are analysed and recommendations are made for refining the UK model. The next section discusses the calibration of the Risk to People model with data from a number of European flood events. Adaptations and revisions to the current UK model are then discussed and the data analyses that were conducted for the research are explained. A proposed European Risk to Life model is then presented followed by a suggested approach for mapping the risk to life using the revised model. The final Section draws together the key research findings along with recommendations for further research.

2.4 Floods and risk to life or injury

The Emergencies Disaster Database (EM-DAT: The OFDA/CRED International Disaster Database (www.em-dat.net1, Université Catholique de Louvain, Brussels, Belgium) records a total of 2,516 flood disasters in the period 1980- 2006, accounting for 176,824 deaths and some 2,600 million people affected world-wide. In Europe, although the numbers of deaths from floods are not as high as in other parts of the world, flooding is the most common natural disaster, and deaths are not uncommon (WHO- Europe, 2002a); much flood risk management effort is therefore aimed at reducing these losses. Recent floods in Europe have resulted in a number of fatalities. For example, the 1997 Oder floods were the largest floods on record in Poland (Kundzewicz et al., 1999) and caused 50 deaths

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 17 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

(Kundzewicz and Kundzewicz, 2005). In August 2005 at least 50 people died, 33 of them in Romania, due to flooding caused by heavy rains in Austria, Bulgaria, Germany, Romania, and Switzerland (www.em-dat.net1). The August 2002 floods in Central Europe caused more than 100 fatalities in Austria, the Czech Republic, Germany, Hungary and the Russian Federation (WHO-Europe, 2002a). Guzzetti et al. (2005) highlight that floods and landslides kill people almost every year in Italy (www.em-dat.net1) and examinations of media reports suggest the summer 2007 flooding in England and Wales caused 10 deaths (both directly and indirectly).

Kundzewicz et al. (2005, p.167) cite the following changes as possible reasons for the increase in flood risk and vulnerability in Central Europe:

• Changes in terrestrial systems (hydrological systems and ecosystems), land cover and usage, river regulation: straightening of channels, embankment, changes in the conditions that transform precipitation into run-off causing a higher peak and shorter time-to-peak. • Changes in socio-economic systems, increasing exposure and potential damage due to floodplain development, higher wealth in flood prone areas; changing risk perception. • Changes in climate e.g. increased intensity of precipitation, seasonality, circulation patterns.

The vast majority of flood fatalities occur during the impact phase of flood events. Jonkman and Kelman (2005, p.76) define a flood-related fatality as ‘a fatality that would not have occurred without a specific flood event’ although they accept that this definition raises questions regarding the timing of the death. They also define ‘flood’ as ‘the presence of water on areas that are usually dry’. However, there are consistency problems when classifying flood deaths (Jonkman, 2003) and no one “standardised universally-accepted method exists for determining whether deaths are caused by a natural disaster” either directly or indirectly (Schlenger et al., 2006, p12). Global figures on flood- related fatalities do not in general include tsunamis or hurricane victims, even though according to Jonkman and Kelman’s (2005) definition of flood they should be included as ‘flood victims’. Drowning is not the only cause of death in a flood and many of the drownings are car related, as discussed below in later sections. Moreover, when people drown in a car the deaths are often considered to be traffic deaths (Kelman, 2004).

2.5 Factors affecting cause of death or injury

Mortality and morbidity can depend upon the type of flood event and various other factors. In flash floods and other situations where the impact is more immediate, most deaths are due to drowning while injury is usually a result of moving debris and high winds (Legome et al., 1995). Mortality associated with a flood will depend on the flood characteristics (e.g. depth, velocity and speed of onset) but the way people respond to floods is also a critical factor.

Jonkman and Kelman (2005) propose a framework for analysing flood deaths: they suggest that a combination of hazard factors and vulnerability factors result in a flood death due to a specific medical cause (e.g. physical trauma, drowning, heart attack). Flood hazard factors used to calculate how floods impact upon people include depth of water, rate of rise, velocity, wave characteristics and debris and pollutants load (Jonkman and Kelman, 2005). The meteorological conditions that accompany floods can also cause additional deaths, for example, in car accidents due to more collisions and falling trees by high winds (Jonkman, 2003). Injuries and deaths may also occur during clean up (Noji, 1993; MMRW, 1989).) or as a result of undertaking rescues (Jonkman and Kelman, 2005; MMRW, 2000). Rescues from fast flowing waters in particular present high hazards.

1 In order for an event to be recorded into the database, at least one of the following criteria must be fulfilled: 10 or more people killed, 100 or more people affected/injured/homeless, significant disaster, e.g. ‘worst disaster in the decade’, significant damage, e.g. ‘most costly disaster’. Source: www.em-dat.net Data Accessed 22/01/07

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 18 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Certain characteristics of people, their community or property can also increase the risk to life of those affected by flooding. These characteristics include (among others) the presence of elderly or ill people, particular types of property (e.g. single storey), no previous experience or awareness of flooding, poor community support, the need to evacuate and live in temporary accommodation (HR Wallingford, 2003). Vulnerabilities that can potentially lead to a flood death include age, gender, prior health and disability (physical and mental), swimming ability, previous flooding experience, clothing, activity and behaviour (e.g. sleeping, attempting a rescue, evacuation), impairment (e.g. due to alcohol or drugs) and knowledge of the area. The literature reviewed by Jonkman and Kelman (2005) rarely makes links between hazard factors, individual vulnerabilities and the medical cause of death.

Other circumstances may contribute to reduce the death toll caused by a flood. In 1988, a flash flood in the region of Nimes, France, damaged 45,000 homes and destroyed more than 1,100 vehicles. However, there was a relatively low number of deaths (9) and serious injuries (3) reported despite the fact that only 17% of the population were aware that they lived in a flood prone area. The low number of deaths can be attributed to the fact that the flood occurred early in the morning when people were still at home and that before the peak of the flood water was already blocking traffic on the roads. The limited death toll was also attributed to the mild temperature, official rescue operations and rescues by civilians (Duclos et al., 1991).

An additional factor that deserves attention is human behaviour. In European floods particularly, deaths are strongly related to risk-taking behaviour (Jonkman, 2003), and the World Health Organization (2002a) estimate that up to 40% of health impacts due to flooding result from such behaviour. For example, people attempting to walk or swim in flood waters can be swept away, the danger of this being higher in fast flowing waters (Jonkman, 2003). HR Wallingford (2003) argue that 0.3m of water is sufficient to cause instability to small, light or low motor vehicles while emergency vehicles may resist waters of up to 1 m in depth (HR Wallingford 2005a); safe evacuation by higher and larger vehicles is only possible up to the depth of 0.4m. People often underestimate the danger of flood waters and lack imagination about what can happen. Some people have been reported as dying whilst trying to save pets and belongings, and some were simply ‘flood tourists’ watching or photographing the flood waters. ‘Flood tourism’ has been reported in several recent European floods, including large groups of people gathering on river banks and pursuing recreational boating on flooded streams (Jonkman and Kelman, 2005; Wilson, 2006). Male fatalities frequently outnumber female with regards to risky behaviour, in one Australian example by as many as 4:1 (Coates, 1999). Risky behaviour is often caused by lack of knowledge of what is best to do in a flood situation. One of the main difficulties of flood management lies in educating the public to react in an appropriate way before or during a flood (WHO-Europe, 2002b). For instance, 95% of flash flood victims try to outrun the waters along their path rather than climb rocks or go uphill to higher grounds (Facts about flooding, no date, http://www.weather.com/safeside/flood/facts.html).

Deaths by fire, electrocution and carbon monoxide poisoning have also been reported as an indirect consequence of flooding, often during the clean up phase (e.g. Jonkman and Kelman; 2005; MMRW, 2005; MMRW, 2000; MMRW, 1989) and these deaths generally occur inside buildings (Jonkman and Kelman, 2005). Identifying the relationships between the different variables is therefore one of the problems with modelling loss of life (Jonkman et al., 2002).

The difference between surviving and not surviving a flood event may also be strongly influenced by random, unpredictable factors. For instance, the presence/absence of floating debris that can be clung to, and the availability of shelter, are largely a matter of luck (Aboelata et al., 2003). Therefore, although it might be possible to estimate and define the broader risks to society from a flood event, it is extremely difficult to estimate risk to an individual. With small numbers of fatalities being recorded, fortuitous or unfortunate circumstances leading to more or fewer fatalities may greatly impact upon the total recorded numbers of deaths.

Four broad sets of flood characteristics can be identified which are seen to influence the number of fatalities or injuries in the event of a flood:

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 19 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

• Area characteristics (exposure, type and structure of buildings, flood warnings etc.) • Flood characteristics (depth, velocity, debris, speed of onset, time of day/year etc.) • Population characteristics (age, prior health, disability, language constraints, presence of tourists/visitors, behaviour etc) • Institutional response (evacuation, rescue etc.)

Loss of life is thus caused by a combination of the above characteristics, for example a high depth and velocity, coupled with vulnerable housing and population, no flood warning and risky behaviour. These different combinations provide different potential scenarios and risk to life

2.5.1 Examples of common flood scenarios in Europe The most frequent types of flood event in Europe are flash floods and riverine floods (Jonkman, 2005). Flash floods have the highest mortality and usually affect smaller areas. River floods affect larger areas and consequently more people but generally cause lower mortality. A review of 220 flood events in Europe recorded by the OFDA/CRED International Disaster Database showed that the average mortality (number of fatalities/total affected) for flash flood events is 5.6% of affected populations compared to 0.47% for river floods (Jonkman, 2005). Flash floods are frequently caused by heavy rain in hilly or mountainous areas. They allow no warning or just a short warning, which increases area vulnerability, as it affects the position of people within the floodplain. Flash floods are not only associated with steep terrain but also with the flooding of flat areas where the slope is too small to allow the run-off of floodwaters. The water accumulates on the surface or in low-lying areas such as underground car parks or basements (Kron, 2002). During the June 2007 floods in Hull, UK, intense rainfall resulted in the urban drainage systems being overwhelmed thus flooding thousands of properties (Coultard et al., 2007; Crichton, et al., 2007).

Slow rising floods tend to affect larger areas than flash floods and usually provide longer lead times for warnings. Substantial riverine floods occurred in Central Europe in 1993, 1995 and 1997 (Rosenthal and Bezuyen, 2000) and more recently in 2002 and 2005. In recent years (e.g. 1997, 2001, 2002) major flooding has taken place in summer generated by intensive rainfall during a long wet spells that covered vast areas. As well as slow rising floods in the main rivers, these events also caused violent flash floods in the smaller catchments (Kundzewicz et al., 2005). The combination of snow melt and rainfall is another common cause of flooding in winter and spring. This type of flood is typical of Central Europe (Colombo and Vetere Arellano, 2002) and Italy (Guzzetti et al., 2005).

Campsites are particularly vulnerable areas as tents and caravans may not provide safe refuge in the event of heavy rainfall or a flash flood. People vulnerability in campsites may also be high due to the presence of families with children and also of people that do not necessarily know the area. Tourists are thus particularly vulnerable to flash floods (Lutoff and Ruin, 2007; Ruin, 2007). An example is Biescas, Spain where in 1996 86 people died as a consequence of the floodwaters and mud that covered a campsite during a flash flood (WHO-Europe, 2002a).

2.6 Methods to calculate flood risks to people

Several methods have been developed as a means to calculate the potential risk to life from flood events; Jonkman et al. (2002) reviewed a number of these methods. The number of fatalities caused in a flood depends on a large number of characteristics (as discussed above), however, most of the models reviewed by Jonkman et al. (2002) limit themselves by only taking into account some of these characteristics when modelling loss of life.

Many of the existing risk to life models are designed to predict fatalities for either large-scale floods caused by flood defence failure in low-lying areas (e.g. Jonkman, 2003, 2007) or dam or dike break scenarios (e.g. Waarts, 1992, Graham, 1999). However, flood events in many other parts of Europe are

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 20 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 quite different from these situations, thus these models may not be applicable in these different situations. These models also largely involve mortality fractions based on empirical observations, which are then applied to the exposed population according to flood severity and other parameters such as flood warning and/or awareness.

One project in the UK that developed a different model to predict loss of life was the Flood Risks to People Project (HR Wallingford, 2003; 2005a). The Flood Risk to People method is different in that fatalities for a particular event are calculated as a function of injuries, which in turn are estimated according to the flood, area, and population characteristics, rather than applying a uniform mortality fraction to the exposed population. The Risk to People model for estimating loss of life is fully described in the next Section.

2.6.1 The ‘Flood risks to people’ methodology The Flood Risks to People Project Phase 1 and 2 developed a methodology for assessing flood risks to people. The project covered deaths or serious harm that occur as a direct result of a flood either during or up to one week after the event. The methodology was developed to be applicable in England and Wales and can be used to assess and map flood risk to people at different scales (for example, flood event, regional or national scale). Phase 1 of the project developed a formula that combined information on flood hazard with criteria related to the vulnerability of areas and people to flooding for estimating the likelihood of serious injury or death from flooding. Phase 1 calibrated the formula using three historical UK flood case studies of Norwich 1912, Lynmouth 1952 and Gowdell 2000 (HR Wallingford, 2003). During Phase 2 (HR Wallingford, 2005a) a number of changes were made to the formula and the model was retested on the case study of the Carlisle flood in 2005. The basic method and the final model developed (henceforth known within this report as the Risk to People Model) will be explained below, however readers are advised to refer to the reports generated by the original two projects (HR Wallingford 2003; 2005a) for more detailed information concerning the model’s development and case study testing.

As the Risk to People model has been to some extent useful in the assessment of risk from flooding in the UK, this model was taken to form the basis for modelling of risk to life within Task 10 and was tested for its applicability within the wider European context. Following the results of this calibration with European flood event data, the plan was to refine the model if necessary to apply to flooding in the rest of Europe.

The model is based upon a series of different criteria and there are obviously many other factors that the model does not consider. This issue is raised again in later discussion when beginning to consider the appropriateness and/or adaptability of the Risk to People model to European floods more generally. The model works on the premise that the numbers of deaths can be calculated from a function of those who are within the ‘at risk population’: that is those who are exposed to the flood hazard.

Three broad sets of characteristics have been used to try to determine the number of fatalities and serious injuries from a flood event. These are indicative of the often held notion that flood risk is a product of the hazard and the vulnerability to a particular event (Raynor and Cantor, 1987; Dracup and Kendall, 1990; Blong, 1997; Lewis, 1999). The characteristics considered within the model are:

• Flood characteristics (depth, velocity, etc.) • Location characteristics (inside/outside, nature of housing) • Population characteristics (age, health, etc.) (HR Wallingford, 2003).

Based on these characteristics, the following formula (Figure 2.1) was developed to calculate the number of injuries produced by a single flood event. More specifically the Risk to People model functions by initially calculating the expected numbers of serious injuries from a flood event and then this figure is used to calculate the expected numbers of deaths from the event. Therefore the model

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 21 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

works by estimating the number of fatalities by taking this as a function of the number of those injured.

Number of Injuries N(I) = 2NZ x (Hazard Rating x Area Vulnerability) x People Vulnerability 100 where, • N(I) = number of injuries • NZ = population living in the floodplain • Hazard Rating = function of the flow characteristics of the flood, i.e. depth (m) and velocity (m/s) and the ‘debris factor’ (score). • Area Vulnerability = function of the effectiveness of flood warnings, speed of onset, type of buildings • People Vulnerability = function of the number of very old people (over 75) and long term sick/ disabled/infirm. This factor is expressed as a percentage

Figure 2.1: Risk to People model flood injuries formula

The number of fatalities from a flood event can then be calculated from the injuries formula as this is considered by the Risk to People methodology to be a function of the number of injuries and the hazard rating. Thus, the more severe the flood (in terms of depth, velocity and debris), the higher the proportion of fatalities among the injured. Figure 2.2 illustrates this formula and its components and these characteristics are now discussed in more detail in the following sections.

Fatalities = 2N(I) x HR 100 where, • N(I) = number of injuries • Hazard Rating = function of the flow characteristics of the flood, i.e. depth (m) and velocity (m/s) and the ‘debris factor’ (score).

Figure 2.2: Risk to People model flood fatalities formula

2.6.1.1 Flood Hazard Rating (HR) The flood hazard rating is used in the methodology as a variable that affects the proportion of exposed people which are injured or killed. The degree of hazard that flood waters present to people depends on the velocity and depth and the presence of debris. After considering a number of alternative equations with reference to experimental data, the final formula used to calculate the hazard rating to people is shown in Figure 2.3.

HR = d x (v + 0.5) + DF where, HR = (flood) hazard rating; d = depth of flooding (m); v = velocity of floodwaters (m/sec); and DF = debris factor (= 0, 0.5, 1 depending on probability that debris will lead to a significantly greater hazard) Figure 2.3: Risk to People model hazard rating

Figure 2.3: Risk to people model hazard rating formula

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 22 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Experimental work from Abt et al. (1989) and the Karvonen et al. (2000) was reviewed and the above formula was found reliable for determining the threshold for losing stability in flood conditions. However, it was also acknowledged that there are differences between the two experiments and that they do not reproduce real-life conditions. Therefore, it was recommended to be precautionary in setting and describing hazard classes.

2.6.1.2 Area Vulnerability (AV) The Area Vulnerability determines the number of people exposed to the flood and also provides a consideration about whether they are able to escape the flood waters. At any particular time people may be in a number of locations for instance; outdoors, indoors, inside a vehicle or in a basement. The Area Vulnerability therefore is a function of flood warning, speed of onset and the nature of these areas (Figure 2.4).

Area Vulnerability (AV) = Speed of onset + Nature of the area + Flood warning

Figure 2.4: Risk to People model area vulnerability score

In the methodology, each of the factors received a 1, 2, or 3 score illustrated in Table 2.1.

Table 2.1: Area Vulnerability scores Parameter 1 - Low risk area 2 - Medium risk area 3 - High risk area Speed of onset Onset of flooding is very Onset of flooding is Rapid flooding gradual (many hours) gradual (an hour or so) Nature of area Multi-storey apartments Typical residential area Bungalows, mobile homes, (2-storey homes); busy roads, parks, single commercial and industrial storey schools, campsites, etc. properties Flood warning Score for flood warning = 3 - (P1 x (P2 + P3)) where P1 = % of Warning Coverage Target Met2 P2 = % of Warning Time Target Met3 P3 = % of Effective Action Target Met4 Area Vulnerability (AV) = sum of scores for ‘speed of onset’, ‘nature of area’ and ‘flood warning’

2.6.1.3 People vulnerability (PV) The Risk to People methodology examined a number of factors that made people more (or less) vulnerable to flooding or the effects of flooding. These are explained in more detail in HR Wallingford (2005a p16-17). Those illustrated in the formula in Figure 2.5 are those that were considered to be the most important within UK floods. This variable is based on the idea that the very old and the infirm/disabled/long-term sick will be more at risk (young children would also theoretically be more at risk, but they are unlikely to be left alone in a flood).

People Vulnerability (PV) = % residents suffering any long-term illness + % aged 75 or over

Figure 2.5: Risk to People model people vulnerability score

2 This is based on the Environment Agency’s 80% target 3 Two-hour warning time 4 Target: 75% of people taking action

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 23 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Other socio-demographic factors such as income, education, employment status, car ownership and family status will be associated with long term physical and psychological effects but not with direct physical injuries/deaths.

The following formula (Figure 2.6) is used to calculate the numbers of deaths and/or injuries.

N(I) = N x X x Y

where;

N (I) = number of deaths/injuries N = population within the floodplain X = the proportion of the population exposed to a risk of death/injury for a given flood (based on the Area Vulnerability) Y = proportion of those at risk that will suffer death/injury, based on the People Vulnerability

Figure 2.6: Formula for calculating the number of deaths and/or injuries

2.6.2 Quantifying the risk to people for a single event With the Risk to People model the flood event needs to be properly zoned in order to be able to apply it to single events. This is so that the characteristics and therefore the values assigned to each of the variables are as homogenous as possible over that area. This is particularly important in relation to the hazard rating component as the degree of hazard (depth, velocity and debris) depends on the position of people within the floodplain. How this has been achieved in the UK situation is that the risk zones have been defined as being areas of different distances from the river or the coast (i.e. the source of the water in question). HR Wallingford (2005a, p20-23) present a hypothetical case study to illustrate the application of the Risk to People methodology to a flood event. This hypothetical case study (in shortened version) is presented in Figure 2.7. The example has been generated for a 100 year flood (1% probability). It is of course possible however to undertake the same calculation for events of other probabilities. Further examples including those that were used to calibrate the Risk to People methodology can be found in HR Wallingford (2005a, p25-35).

2.6.3 Risk to People mapping A further output of the Risk to People project was a mapping component whereby a Geographical Information System (GIS) was used to analyse, manage and communicate flood risks to people. The approach for mapping flood risks to people as reported in the Risks to People project (HR Wallingford, 2005a) is illustrated in Figure 2.8.

There are two elements to be considered in the mapping of flood risks to people:

• A Flood Hazard map, which represents the hazard rating component of the Risks to People methodology and describes the hydraulic conditions affecting the at-risk population. • A Flood Vulnerability map, which combines the Area Vulnerability and Population Vulnerability components of the methodology and describes the vulnerability of both the area, in terms of e.g. flood warnings and land use, and the population, in terms of age and health status.

The Risks to People map finally combines these to describe the individual risk of serious harm due to flooding.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 24

Hazard zones, Hazard Rating and Area Vulnerability

Distance from NZ Typical Typical Debris factor Hazard Flood warning Speed of onset Nature of area AV = FW + river/ coast Population depth d velocity v (DF) Rating (FW) (ONSET) (AREA) ONSET in risk zone (m) (m/s) = d (v+0.5) + +AREA DF 0-50 25 3 2 1-possible 8.5 2.15 3 2 7.15 50-100 50 2 1.8+ 1-possible 5.6 2.15 2 1 5.15 100-250 300 1 1.3 1-possible 2.8 2.15 2 3 7.15 250-500 1000 0.5 1.2 1-possible 1.85 2.15 1 2 5.15 500-1000 2500 0 0 0-unlikely 0 2.15 1 2 5.15

Determining the number of people exposed Calculating the People Vulnerability

Distance NZ Hazard Area X = HR x NZE Distance from Factor 1 = % Factor 2 = % Y (People from river/ Rating Vulnerabili AV river/ coast of very old disabled/ vulnerability) coast (HR) ty (AV) infirm = 1 + 2 0-50 25 8.5 7.15 61% 15 0-50 15% 10% 25% 50-100 50 5.6 5.15 29% 14 50-100 10% 14% 24% 100-250 300 2.8 7.15 20% 60 100-250 12% 10% 22% 250-500 1000 1.85 5.15 10% 95 250-500 10% 15% 25% 500-1000 2500 0 5.15 0% 0 500-1000 15% 20% 35% The area vulnerability score is then multiplied by the hazard rating to obtain X or In order to determine the numbers of deaths/injuries, the number of percentage of people exposed. The number of people affected (NZE) is obtained people exposed (NZE) is multiplied by a factor Y based on the People by multiplying the percentage of people exposed (X) by the number of people in Vulnerability. Y is a function of the percentage of over 75s and the percentage of long-term ill. the area (NZ).

Determining the number of deaths/injuries Distance from NZE Y = 1 + 2 Number of injuries Fatality Number The number of injuries is obtained by multiplying the river/ coast = 2 x Y x NZE rate = 2 x of deaths people at risk (NZE, see table 6.6.2) by their HR vulnerability. The number of deaths is a function of 0-50 15 25% 8 17% 1 the HR and is calculated by multiplying it by 2. In 50-100 14 24% 7 11% 1 this example, the estimated number of deaths is 5 and 100-250 60 22% 26 6% 1 the injuries 89: 250-500 95 25% 48 4% 2 500-1000 0 35% 0 0% 0 All 184 89 5

Example figures taken from HR Wallingford (2005a, p25-35). Figure 2.7: Example of the application of the Risk to People methodology

Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

1) Define the spatial extent of the study. The spatial scale will affect the type of data to be used and the level of risk to be assessed.

2) Decide on the severity (return period/annual probability) of the design flood event or events. The UK guidance recommends a minimum of five events and the choice of events is influenced by the presence and quality of flood defences. For areas that are protected by flood defences, more severe events should be modelled compared to undefended areas. For the UK, the recommendations (expressed in terms of the return period) are:

Undefended area with regular flooding – 20, 50, 100, 250, and 1000 year events Defended area (standard of defence 1:75 years) – 100, 200, 300, 500, and 1000 year events

3) Define flood hazard zones based on hydraulic model outputs and topography or define using existing flood risk maps and expert judgement. For large, low-lying floodplains it may be appropriate to define zones based on distance from the source of flooding.

4) Calculate the Hazard Rating for the flood zones:

HR = D * (V+0.5) + DF

The debris factor may be seen as optional; the Criteria for Identifying Rapid Response Catchments report (Environment Agency, forthcoming) did not apply a debris factor because they considered that there was not sufficient data available to determine an appropriate value. If a debris factor is to be included, Table 1 gives guidance on selecting a debris factor based on land cover and flood characteristics. Otherwise, the debris factor can be omitted from the calculation. Figure ~1 shows the Flood Hazard zones generated for the Thamesmead area in the UK.

Suggested debris factors for different flood depths, velocities and dominant land cover type

Depths (m) Pasture/Arable Woodland Urban

0 – 0.25 0.0 0.0 0.0 0.25 – 0.75 0.0 0.5 1.0 d> 0.75 and/or v> 2m/s 0.5 1.0 1.0

Further guidance on flood hazard mapping is given above (in Section 3.5.1)

5) For the flood hazard zones identified, use local maps (equivalent to the UK Ordnance Survey maps) and population data to develop an understanding of land use and population characteristics. Also review formal flood warning systems.

Assign scores for speed of onset, land use type, and flood warning quality to the hazard zones to give the area vulnerability score. The area vulnerability score is obtained by assigning values of 1 to 3 to these variables, where 1 represents low risk, and 3 represents high risk. These values are then summed to give the Area Vulnerability score. Because the scoring system is very basic, it is probably not worthwhile devoting a great deal of time and effort to these variables; a broad-brush assessment should suffice. Figure 3.10 shows the Area Vulnerability component of the flood zones identified by the Flood Hazard methodology.

6) Calculate or estimate the percentage of the population that is elderly (e.g. over 65) and the percentage of the population that suffers from long-term illness. The total of these two percentages is the People Vulnerability score.

Note that it may be necessary to divide a flood hazard zone into two or more sub-zones if the land use and/or population characteristics are substantially different within the zone.

Further guidance on flood vulnerability mapping is given in Section 3.5.2

7) Once the flood hazard zones are identified and flood vulnerability within the zones is defined, it is a simple process to calculate the risks to people for each zone. This can be done within a GIS (e.g. ArcView and ArcMap have the necessary functionality) or the data can be exported to a spreadsheet package such as MS Excel. Figure 3.11 shows the Risk to

Figure 2.8: Steps for mapping, after HR Wallingford (2005a, p36-48) and HR Wallingford (2005b, p33-46)

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 26 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

2.7 Initial model adaptation of the Risk to People model to the European context

Following a brief investigation concerning the data sources and information that are available in other European countries it was apparent at an early stage that it would be necessary to adapt the Risk to People methodology and the model produced if it was going to be possible to use it to estimate the risk to life from Continental European floods. In addition, it was important at the outset to examine and compare the characteristics of European flood events in order to more fully inform the data collection component of this study. This section briefly examines both the initial change to the model and a consideration of additional factors that may be required to assess other types of European flooding.

An initial change was required to the Flood Warning aspect of the Area Vulnerability component of the model due to the fact that the UK appears to hold more information than much of the rest of Europe regarding the reliability, effectiveness and people’s response to flood warnings. Therefore, when applying the model to the situation in Europe a much simpler scale has been adopted, although the same numbering of 1 to 3 has been applied. These alternative criteria are shown in Table 2.2.

Table 2.2: Flood warning components applied to the Risk to People model for Continental European case studies Figure assigned to 1 2 3 the model Explanation Good Fair None Where the majority of Warning received by No warnings received people at risk received some of the population warning with adequate with adequate lead time lead time

Thus, an issue identified that needs consideration in other European flooding relates to the availability of data. It is very problematic even when a flood is well-documented to assign values to many of the Risk to People model criteria as many of these would be required to be estimates. Specific issues of this nature were considered in detail within the main report (Priest et al., 2007), including a detailed case study of the 2004 Boscastle flood in the UK. However, one major issue that needs to be raised at this juncture is the impact of zoning on the results. Zoning has of course been attempted where it has been possible, however in a number of circumstances this has not been possible and therefore questions concerning the applicability of the results can be raised.

A further issue relates to the characteristics of floods in Continental Europe which can vary considerably from those that occur in England and Wales. Due to the variations in climate, land use, hydrological conditions and catchment scale (as well as many other variables) many different types of flooding are experienced (Penning-Rowsell and Peerbolte, 1994). In addition to this, the floods in the UK are often considered to be less severe than many of those experienced in Continental Europe, where events are often faster, deeper and more extensive, and therefore can be considered more dangerous to people. The current methodology does not take into account flood damage to buildings and the risk to people associated with building collapse nor does it consider the ‘vehicles factor’ highlighted above (HR Wallingford, 2005a). Moreover, the characteristics of many Continental European floods mean that aspects of people’s vulnerability come under question.

Additionally, some areas of Europe experience slow rising flooding which not only permits flood warning but also evacuation. This is a particularly important factor as it may directly affect the population at risk, which has a large impact upon the results of the model. Finally, information about additional factors such as whether there are any language constraints and awareness of flood risk has been gathered. The full set of factors included during data collection is illustrated in Appendix A. The

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 27 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 next Section discusses the methodology used in the research for this Task in developing a European Risk to Life model based upon the Risk to People model outlined above.

2.8 Methodology and data collection

A key factor when developing a model such as that proposed in this study is the availability of, and access to, high quality, reliable and detailed data (Jonkman and Kelman, 2005). Without such data the output from any model is likely to be at best crude and at worst extremely inaccurate. A good dataset was therefore vital for the calibration of the current Risk to People model, for any revisions to the model, and for successfully achieving the aims of Task 10 Activity 1. The original Risk to People model was developed using data from three flood events (Norwich 1912, Lynmouth, 1952 and Gowdell, 2000) (seven flood zones) in the UK and then further tested using five risk zones from the Carlisle 2005 flood event. In order to test the model more widely it was aimed to gather data for a larger number of European flood events. The methodology employed for collecting European data for use in calibrating the model is outlined in the following sections along with limitations and problems arising from the data collection.

2.8.1 Data collection methods Following a review of the relevant literature and identification of factors leading to risk to life, a template was produced outlining the relevant data required for the study (see Appendix A). Key flood events in Continental Europe, preferably within the last five years, were identified and searches were made on the internet. These searches resulted in little detailed data but were able to provide information on flood events, numbers of fatalities and people affected by country, date and type of disaster. In parallel with this, a number FLOODsite project partners were contacted in those countries where recent flood events had been identified to have fatalities e.g. Germany, Poland, France, Italy and Spain. These partners were sent the data template along with the background information on the aims of the research and guidance on the type of data required for completing the template. When supplying the data partners were also asked to indicate the source of the data along with the quality and levels of uncertainty or accuracy (e.g. sourced from government data, local authorities, residents observation, media reports etc. or if taken from a model, measured or estimated data).

Data was finally obtained from various locations and flood events in Italy, Spain, Germany and France. Other requests for data met with no success, either because the partners did not have the data, because there had been no deaths in the flood events in their areas, or because partners simply did not have the time to gather the data. Therefore additional data had to be requested from sources outside the FLOODsite project e.g. in Poland, Slovenia, the Czech Republic and Romania. In some cases this necessitated translation of information into other languages e.g. Czech and Slovenian. Some of these requests for data received no response, despite follow-up emails, and were eventually abandoned. However, successful contacts were made in Poland and the Czech Republic, both of which resulted in good data being obtained as well as very useful background information on the flood events and locations. As this data did not come from FLOODsite partners it had to be purchased from the suppliers on a commercial basis. Initially data was only requested for flood events with fatalities. Towards the end of the data collection period some data were also collected for events with no fatalities. These data were used for comparative purposes when testing calibration of the model. Figure 2.9 maps those locations from which data was received and included for the calibration of the model.

Overall, data were received for flood events between 1997 and 2005. In total the data appears to cover 11 different flood events at 25 locations across six European countries, providing 43 different data zones which have 82 deaths. Table 2.3 gives a breakdown of from where the data were collected, the highlighted rows grouping those variables from the same flood event. In the final analysis, not all the flood event data received from all locations were able to be used. Two zones (Botarell and Cagarel) were discarded because some of the key data (e.g. flood depth, population exposed) was missing, six (Stronie Slaskie C, Klodzko Town B, Duszniki Zdroj A, Ladek Zdroj C, Polanica Zdroj B and

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 28 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Dresden B) were not used due to the fact that the population exposed to flooding was zero (since the premise of the model is that deaths are a function of the population exposed) and one zone (Gard) because the data was not able to be disaggregated into different hazard zones. The data from those locations that were applied in the final calibration of the model are indicated in Table 2.3. This leaves 34 usable European datasets which have 52 deaths in total.

2.8.2 Limitations and problems arising from data collection Limitations in the availability of data have already been outlined above. This highlights the need for the establishment of reliable, systematic and consistent methods for collecting data following flood events across Europe, as well as for the need to make available data that is collected; a key constraint concerns the level at which data are collected. For example, in England and Wales the Environment Agency conducts regular post-event surveys and data collection at a national level. However, even here data is not easily accessible and there is no national database of such data. In Federal countries like Germany, where each individual State is responsible for the collection of relevant data, the problem is more complicated and each State would need to be contacted separately to obtain any available data. In other countries national-level data is simply not available.

The difficulty in obtaining accurate and reliable data had not been anticipated at the outset. In many cases data on certain variables were simply not available (e.g. numbers of serious injuries, numbers of people evacuated, census data) or it was crudely estimated (e.g. data on flood depths and velocity), often for understandable reasons. In many cases estimates were given, or indeed proxies. Moreover, the majority of data received was for flash flood events and more data on slow rising floods would have been useful. Data supplied by the FLOODsite partners and others originated from a wide range of sources, some being more reliable and robust than others. In several cases no information was given on data sources. Finally, the length of time taken to gather the necessary data for calibrating the model was in many cases extreme, taking many months from the initial request to receipt of the data, and leading to severe delays on the project. The next section explores the relevance of the European data collected for testing the Risk to People model and the potential problems associated with these data.

2.8.3 Usability of Continental European data for Risk to People methodology The Risk to People methodology appears very sensitive to the zoning of flood events (e.g. that the similar flood characteristics are experienced over the whole zone) and therefore any data that is input into the model is also required to be correctly zoned. This is very difficult as not all the characteristics will be homogenous over the same zones. From the description of the model by HR Wallingford (2005a), and from looking at the makeup of the model, the most important components to ensure that floods are correctly zoned are the hazard rating and the populations that are exposed to flooding. It would be extremely difficult to construct flood zones where each of the different variables within the model was entirely consistent, as this would necessitate the creation of a large number of very small zones for each flood event. Information at this level of detail just does not exist in many locations. Despite this, those providing this study with data were asked to define different risk zones within the same flood if this were indeed possible and many did manage to achieve this for some of the flood data provided. In one case (the Gard, France) it was not possible to disaggregate the data due to the size of the area and population covered, therefore the model could not be applied to this dataset.

Similar to the problems of defining risk zones and the flood characteristics across different areas, estimating both the population at risk from any particular flood event, and subsequently the people vulnerability characteristics (e.g. the percentage of the at risk population over 75 years of age and the percentage of the long-term sick) is equally problematic. In most cases these populations have been estimated from a number of different sources including post-flood reports, the numbers of properties affected and census data. These of course do not include visitors to the area or any other transient populations, and inevitably the census geography will not match the risk zones.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 29

Figure 2.9: Map of locations where flood event data have been gathered

Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 2.3: List of flood event data received (in chronological order)5 Date Location Country Flood type Deaths Applied In model 6-31 July 1997 Olomouc Czech Slow rise (around 3 days before danger 2 Yes Republic situation) 6-11 July 1997 Otrokovice Czech Slow rise (around 3 days before danger 1 Yes Republic situation) 6-31 July 1997 Troubky Czech Middle risk onset 9 Yes Republic 6/7 July 1997 Stronie Slaskie A Poland Flash flood/ Medium risk onset 1 Yes 6/7 July 1997 Stronie Slaskie B Poland Flash flood/ Medium risk onset 0 Yes 6/7 July 1997 Stronie Slaskie C Poland Flash flood/ Medium risk onset 0 No 7/8 July 1997 Klodzko Town A Poland Flash flood/ Medium risk onset 1 Yes 7/8 July 1997 Klodzko Town B Poland Flash flood/ Medium risk onset 0 No 7/8 July 1997 Klodzko Town C Poland Flash flood/ Medium risk onset 5 Yes 7/8 July 1997 Klodzko Town D Poland Flash flood/ Medium risk onset 0 Yes 7/8 July 1997 Klodzko Gmina A Poland Flash flood/ Medium risk onset 0 Yes 7/8 July 1997 Klodzko Gmina B Poland Flash flood/ Medium risk onset 0 Yes 7/8 July 1997 Klodzko Gmina C Poland Flash flood/ Medium risk onset 1 Yes 7 July 1997 Miedzylesie A Poland Flash flood/ Medium risk onset 0 Yes 7 July 1997 Miedzylesie B Poland Flash flood/ Medium risk onset 1 Yes 7 July 1997 Miedzylesie C Poland Flash flood/ Medium risk onset 0 Yes 6/7 July 1997 Bystrzyca Klodzka A Poland Flash Flood/ Medium onset 0 Yes 6/7 July 1997 Bystrzyca Klodzka B Poland Flash Flood/ Medium onset 0 Yes 7 July 1997 Ladek Zdroj A Poland Flash Flood/ Medium onset 0 Yes 7 July 1997 Ladek Zdroj B Poland Flash Flood/ Medium onset 0 Yes 7 July 1997 Ladek Zdroj C Poland Flash Flood/ Medium onset 0 No 7 July 1997 Ladek Zdroj D Poland Flash Flood/ Medium onset 0 Yes 22/23 July 1998 Polanica Zdroj A Poland Flash Flood/ Medium onset 0 Yes 22/23 July 1998 Polanica Zdroj B Poland Flash Flood/ Medium onset 0 No

22/23 July 1998 Duszniki Zdroj A Poland Flash flood/ Medium risk onset 0 No 22/23 July 1998 Duszniki Zdroj B Poland Flash flood/ Medium risk onset 0 Yes 22/23 July 1998 Duszniki Zdroj C Poland Flash flood/ Medium risk onset 7 Yes 14 August 1998 Fortezza Italy Flash flood/High onset 5 Yes 29 August 1998 Fella A Italy Flash flood/High onset 1 Yes 29 August 1998 Fella B Italy Flash flood/High onset 1 Yes 10th June 2000 La Farinera Spain Flash flood/High onset 1 Yes 10th June 2000 Magarola Spain Flash flood/Medium risk 4 Yes 22 October Botarell Spain Flash flood/ Medium risk onset 4 No as some data 2000 missing 8/9 June 2002 Gard France Medium to high risk Flash flooding but 23 No as data was the data was not disaggregated unzoned 12/13 August Dresden A Germany Slow to Medium rise flooding 1 Yes 2002 12/13 August Dresden B Germany Slow to Medium rise flooding 0 No 2002 13 August 2002 Erlln, Mulde Germany Slow rise flooding 0 Partly 13 August 2002 Grimma, Mulde Germany Slow rise flooding 0 Partly 13 August 2002 Eilenburg – Karl- Germany Slow rise flooding 0 Partly Marx-Siedlung, Mulde 7 September Cambrils Spain Flash flood/ Medium risk onset 3 Yes 2004 13th October Cagarel Spain Slow onset (was defined as low risk) 3 No as some data 2005 missing 12-15th October Calonge Spain Flash flood/High onset 1 Yes 2005 22 October Cassano Murge Italy Flash flood/High onset 7 Yes 2005

5 The shading in the table groups zones from the same flood event.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 31 29th February 2008 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

In addition to the need to ensure that the characteristics that are provided are as homogenous as possible across individual risk zones, a further potential problem with the data provided is that it is a snap-shot of the flooding situation and often does not reflect the changing risks within a single flood event. Therefore there is some doubt concerning which values should be provided for the physical characteristics. In many cases the maximum values have been provided, however this will have only occurred at the peak of the flooding. This is particularly significant in those floods whereby there was a period of less severe flooding prior to the floodwater’s peak which would not only serve as a warning to people, but also in the more severe cases allow them to escape from danger.

The above factors raise questions about both the purpose of modelling risk to life (for instance is it calculating a worst case scenario?) as it is unlikely to be able to forecast accurately the exact number of deaths. Or is it to be used as a guide to the identification of those areas which are most likely to suffer fatalities from flooding? These questions are returned to in the concluding Section.

2.9 Analysis of the circumstances and causes of European flood-related deaths

This Section provides a more in-depth analysis of European flood events and those factors leading to fatalities. It looks at whether these factors are included, or could be included, within the UK model. Those flood events (zones) where no deaths were reported are also analysed to identify possible reasons why no fatalities occurred, and therefore to lead to possible model refinements.

2.9.1 Reporting of fatalities As reported above, there is no systematic system for collecting flood data across Europe and this includes information about deaths. Therefore data on this issue is often difficult to find and the level of detail presented varies enormously. Much of this information needs to be pieced together from a number of sources including official flood reports, media reports and unofficial sources. The reliability of much of this data from whatever source is often also reliant upon eye witness reports, the quality of which may vary greatly. Coupled with issues relating to the availability of data surrounding known flood deaths, there is also the issue of “missing deaths” or deaths that may have been misattributed to a flood event. Jonkman and Kelman (2005) discuss the problems associated with the indirect deaths associated with flood events. These are often victims who died some days or weeks after the event and often whose death was not reported as being flood related. Therefore, it is necessary to be cautious when using information about flood-related fatalities.

The age of an event is another issue. The two main calibration events for the UK Risk to People model were Lynmouth which was 1952 and Norfolk in 1912. Much has been written about the Lynmouth event and there is much information about cause of death. Information about the 1912 event is much more difficult to find and the validity of the sources more difficult to assess.

The following sections examine in as much detail as possible the deaths that have occurred in both the UK calibration datasets as well as those from the newly collected European data. A general examination of those factors leading to a loss of life is made to identify overall patterns and trends before a more detailed analysis of case study examples.

2.9.2 Circumstances leading to death from flooding Analyses were undertaken to compare and contrast the ways in which people died in the UK events with the factors surrounding death in the wider European case studies. This was in order to determine whether there are any differences which might be leading to the over-prediction of deaths observed. Tables 2.4 and 2.5 illustrate, in as much detail as was available, how the victims died and other personal information; they also include details about the sources of the information.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 32 29th February 2008 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

2.9.2.1 UK flood-related deaths As can be seen in Table 2.4, there were a number of different reasons why people died in the UK events, and information from all of the events is adequate to make some conclusions about the causes of death. The Lynmouth case study is the most useful, due to the higher number of fatalities and the level of detail that has been recorded about the circumstances leading to casualties. Fatalities in this event were caused primarily by the severity of the hazard with reports of people being swept away by the flood waters or being trapped or plunged into the flood waters when the buildings failed. Indeed, 22 of the 34 overall deaths reported during this flood event were killed when the buildings they were in collapsed around them. The main factors in this event leading to the deaths were: fast onset of the floods, the fact that its full force occurred after dark, and the severity of the event. The ferocity of the flood is also illustrated by the difficulty in recovering some of the bodies of those who had died, some were not found until weeks later and miles away from where they had gone missing; four of the victim’s bodies were never found.

What should also be considered with the Lynmouth event are the large numbers of people who, despite terrible flood conditions, still managed to escape and survive. The information used to calibrate the Risk to People model for this event states that there were a total of 400 people living in the floodplain (Nz score). Prosser (2001) estimates that on the afternoon before the flooding there were 1,700 people in the village of Lynmouth, the majority of which were tourists. People managed to survive by sheltering on top-floors of properties (and were fortunate enough not to be in a property that collapsed) or by evacuating the zones immediately next to the river and moving to higher ground. In spite of the poor conditions and problems in communication, successful efforts were made by the local police, firefighters and ordinary citizens to rescue people trapped in buildings.

The physical characteristics of the Carlisle event contrast significantly with that of Lynmouth. Although 2,800 properties were flooded (Environment Agency, 2005c) and the numbers estimated to have been exposed to the flooding were more than 5 times the number exposed during the 1952 Lynmouth event, the rate of rise of the flood event was much slower, which afforded the opportunity for flood warning and emergency management procedures to be implemented and for people to be successfully evacuated. The two deaths in this flood event relate to two elderly women (aged 79 and 85) who despite being contacted on a number of previous occasions by the Environment Agency did not sign up to their flood warning system and therefore it appears that they were either not sufficiently warned about the event or did not (or were not able) to take action. In addition to the official warning, these women also did not respond to a direct warning from rescuers knocking on their doors (Environment Agency, pers. comm.) In this case it does appear that the age and vulnerability of these two flood victims contributed to their deaths. Despite this event contrasting greatly with that of the Lynmouth flood event, the Risk to People model appears to function effectively.

2.9.2.2 Continental European flood-related deaths Table 2.5 looks in more detail at the factors leading to deaths in the Continental European data sets and shows that there were different factors leading to flood-related deaths in the wider European case studies. Over the 18 flood events where fatalities occurred, there were 74 fatalities in total, with drowning and the collapse of buildings appearing to be the main causes of death, although the circumstances surrounding fatalities differ. Precise information about all of the deaths is not available; however there is sufficient data on enough of the deaths to look in more detail at the breakdown of these fatalities and to comment upon how this influences the performance of the Risk to People model. For instance, the model assumes (through the people vulnerability component) that those over the age of 75 and the long-term ill are more vulnerable to flood events.

Males make up the highest proportion of the fatalities, accounting for almost two-thirds of deaths during the flooding. For those deaths where both the circumstances surrounding the death and the gender is known (n = 40) 30% of the males who died were outside and drowned. 22% of the deaths occurred during rescue attempts of some kind, whether of humans or animals. A further 30% of the male fatalities were from car-related incidents. This is similar to the female component where 24% of

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 33 29th February 2008 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 the deaths were car-related. However, the main location for female deaths was inside the property where 47% of women died from downing after being trapped in contrast with 9% of males who died within buildings or from building collapse respectively. These results also confirm the psychological literature which suggests that males have a tendency to take greater risks (Byrnes et al., 1999),

For the 74 fatalities that occurred during the European flood events we have age data for 60% of the sample. Age compares relatively well with the general age curve for the region suggesting that the age at which people died is a relatively good fit with the general demographic. The main outlier appears as expected to be with the older age group. If it is assumed that the term “older resident” applies to those above the age of 65, then the graph would be even more heavily weighted towards the older age groups. This is of importance as there is an age factor included within the “people vulnerability” component of the Risk to People model.

Where known, Figure 2.10 illustrates the circumstances surrounding the deaths of those people within each age category. The main points of interest within this graph are the importance of car-related deaths to those in the 40-49 age category, and the increasing importance of drowning within homes and properties as the ages of people increase. There is also a rise and then fall in the numbers of people who were killed by flooding in the outdoors, peaking in the 50-59 age category.

The results reported here are not significantly different from other studies of this kind e.g. Jonkman, 2005; Jonkman and Kelman, 2005). They thus provide some insight into those factors that need to be modelled within any attempt to estimate the risk to life from flooding events. Those datasets where there were no fatalities are also important to see whether there are any clear factors that are preventing injuries and deaths from flooding and to identify whether there are any additional factors that need to be included within a method for estimating loss of life from flooding.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 34 29th February 2008

Table 2.4: Causes of death in the UK calibration events

Flood event Number of Cause of death (M = male, F=female) Sources of data deaths Gowdell (2000) No deaths N/A n/a Norfolk (1912) 4 • A baby drowned (M5 months). He was in a boat with his family, when it sank and he was lost when his mother became Eastern Daily Press (2004) unconscious • 1 man drowned after becoming exhausted after wading through the waters to rescue many people who were trapped in their houses • 1 post man drowned when driving horse and cart through deep water • 1 lady died of ‘fright’ while being carried into a boat Lynmouth 34 There were a range of deaths in Lynmouth and the surrounding area. Prosser (2001) and Delderfield (1952) Lynmouth deaths (28) (1981) A number of deaths were characterised by the collapse or partial collapse of six buildings and therefore perished from injuries sustained in the collapse or subsequent drowning. • A disabled mother (64), her son (27) and her daughter and her husband and children (F32, M37, M11 and M9) and 2 visitors (M56 and F52) were all killed in a cottage collapse (the father (63) survived) • A family of four (M30, F32, M3 and M3 months) died when their cottage collapsed. • An elderly couple (M80 and F72) and her brother (78) refused to leave their property and were killed later when that cottage collapsed. • Two elderly women (75 and 77) died when the cottage they were in collapsed. • Two seasonal workers (F48 and F40) from Surrey disappeared from the Lyn Valley hotel and are believed to have been swept away when a section of the hotel collapsed. • The cook of the West Lyn Hotel (F60) was swept away when a section of that hotel collapsed. • Four victims, a grandmother (54), her grandson (8) and two women visitors that they had taken in due to the weather (21 and 22) were swept away trying to escape the flood waters when they fell through a gap where the road had collapsed. • Another woman (56) lost her footing and was drowned when trying to escape the flood waters. • A man (53) was swept away and drowned; whilst trying to rescue others. • The deaths of two people (an unidentified lady and a man of 50) are unaccounted for

Parraccombe deaths (3) • Two holiday makers a mother (46) and her son (14) were killed when their chalet was washed away • A man (60) disappeared, believed to have been swept away and drowned, whilst trying to rescue others.

Filleigh deaths (3) • Three Manchester scouts (M11, M11, M14) were swept away from their campsite and drowned Carlisle (2005) 3 • Two elderly women (79 and 85) died in adjacent flooded homes in Carlisle (had not signed up to the flood warning service BBC News Website (2005); despite being written to by the Environment Agency) Environment Agency (2005c; • 1 man (63) died when a barn collapsed onto his caravan in Hethersgill, Cumbria 2006); Government Office for the North West (2005)

Table 2.5: Causes of fatality in the European events

Flood event No. of Cause of death Sources of data deaths Fella a, Italy (1998) 1 1 (51, woman) swept away by flood her body was found 10km downstream two weeks later Partner questionnaire Fella b, Italy (1998) 1 1 (57, man) swept away and buried by sediments and debris flow Partner questionnaire Cassano Murge, Italy 7 5 members of the same family (mother, 49, Father, 52; children 27, 23, 14) swept away in their car when Partner questionnaire (2005) crossing a bridge that collapsed; 1 driver (24, man) swept away in his car by a small torrent; 1 other drowned in a small river during the flood Fortezza, Italy (1998) 5 5 tourists (German speaking family) were swept away in car when it was hit by a debris flow Partner questionnaire Calonge, Spain (2005) 1 1 woman (75) swept away by the floodwaters Newspaper reports Cambrils, Spain (2004) 3 A couple (49 and 44) and the brother of the woman (41) all Andorran. Main roads were closed but they Newspaper reports continued their trip using a secondary road. The river dragged them when they tried to cross it in their car. La Farinera, Spain (2000) 1 1 woman (83) drowned in her home Newspaper reports Magarola, Spain (2000) 4 2 died when their car fell into the river through a destroyed bridge Newspaper reports 2 policemen died during the search for the first two victims Duszniki Zdroj c, Poland 7 Drowning (but no other details) Workers of the town Council (1998) Klodzko Gmina c, Poland 1 1 man (64) drowned whilst he was trying to catch some boards carried by the water – there was some Workers of the town Council (1997) speculation that he was drunk Klodzko Town a, Poland 1 6 people drowned (5 men 41, 58, 60, 61, 81) and 1 woman (57). One of the men (81) did not want to be Workers of the town Council (1997) evacuated and stayed in his flooded apartment; one was visiting a friend and was swept away when he was Klodzko Town c, Poland 5 leaving. (1997) Miedzylesie b, Poland 1 1 man (34) was sleeping when his bedroom was flooded and collapsed into the floodwaters Partner questionnaire (1997) Stronie Slaskie a, Poland 1 A man (45) drowned when washed from a bridge whilst he was rescuing children standing there. Partner questionnaire (1997) Troubky, Czech (1997) 9 7 older residents died from crush injuries as their buildings collapsed. A younger man was killed by a falling General report on the Floods in the Morava river basin beam A ninth victim (who initially survived the collapse of her home) died in hospital a few days later and Dyje River basin in July 1997, Povodí The homes were constructed unfired clay brick Olomouc, Czech (1997) 2 Drowned when trying to escape the floodwaters Povodí Moravy Water Management Control Centre records Otrokovice, Czech (1997) 1 Drowned when trying to escape on an inflatable bed General report on the Floods in the Morava river basin and Dyje River basin in July 1997, Povodí and General Report on the floods in the Zlín area in July 1997, Zlín District Authrorities. Dresden, Germany (2002) 1 Casualty due to collapsed building Personal communication with the Office for the Environment, City of Dresden Gard, France (2002) 23 5 deaths occurred due to the use of motor vehicles all in separate incidents (F46, F46, M52, M55 and M70) Review of newspapers, municipality services, post flood 7 people died inside their homes from drowning; the following died following a dike break on the Rhône reports from the rescue services (F84, F54, F67, F75 and F77) and the others died due to drowning from rising water levels in their homes Lutoff and Ruin (2007) (F52, F71 and F84) ; 2 died in their homes from heart attacks (F72 and M52) 3 people who were outside drowned (F46) and two of those were attempting to rescue animals (M35 and M?) 5 tourists were killed after they did not evacuate their campsites quickly enough (M42 and 2 children 2 and 6) and (M74 and F34) from a second campsite. Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Numbers of fatalities 4

0 0-9 10-19 20-29 30-39 40-49 50-59 60-69 70-79 80+ Drowned outside Car-relatedAge range (Years) Rescue-related Drowned in homes Collapsed building Campsite-related Indirect causes (heart attack)

Figure 2.10: Circumstances surrounding flood-related death broken down by age for the European flood events

2.9.2.3 Further factors leading to fatalities The complex nature of flooding events and the differences that exist between the circumstances surrounding deaths means that it is important to try to find some common elements that are present to enhance the modelling of losses. Data on the flood characteristics of each of the events has been analysed alongside the actual circumstances of the fatalities to try to identify the main factors that are affecting the number of injuries and deaths (Table 2.6). These highlight the importance of flood velocity, building collapse, people vulnerability and behaviour.

2.9.3 Characteristics of datasets with no fatalities In addition to examining the circumstances surrounding fatalities in the 21 flooding events where there have been deaths, it is also important to examine those 18 zones in Poland and 3 cases in Germany for which data are available where flooding was experienced yet no deaths occurred. These examples can be split into two separate groupings, those that are zones from a hazard location where deaths were experienced (in the cases of Stronie Śląskie, Klodzko Town, Klodzko Gmina, Duszniki Zdroj and Meidzylesie) and those from locations where there were no deaths at all were experienced (Ladek Zdroj, Bystrzyca Klodzka and Polanica Zdroj (Poland) and Erlln, Grimma and Eilenburg (Germany)).

The low numbers of people exposed to the flooding is one reason why no deaths occurred in a number of the flood events. The flood event in Klodzko Town zone B affected industrial areas where there is no residential population and the numbers of people in the area was also low due to the timing of the flood being at night and on a holiday. Similarly, the populations exposed in other zones is low, as the data shows that Stronie Slaskie C, Klodzko Town B, Duszniki Zdroj A, Ladek Zdroj C, Polanica Zdroj B and Dresden B are areas of industry, farmland, forest or meadows where no or few people were exposed.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 37

Table 2.6: Prominent factors affecting death in both the UK and wider European flood events (XX = Most important factors)

Speed Timing Flood Debris Structural Rescue- People Human Awareness/Language Flood event of of Cause of death Velocity Content Collapse related Vulnerability Behaviour difficulties onset flood Norfolk, UK (1912) XX X X A baby drowned (M5 months) after boat he was in capsized; a man drowned attempted to rescue others, a man drowned when driving horse and cart through deep water and a lady died of ‘fright’ while being carried into a boat Lynmouth, UK (1952) XX XX X X X Range of deaths but most were either swept away by the flood waters, hit by debris or trapped in collapsing buildings (see Table 7.1 for more details) Carlisle, UK (2005) XX X X Two elderly ladies (79 and 85) died in adjacent flooded homes in Carlisle a man (63) died when a barn collapsed onto his caravan in Hethersgill, Cumbria Fella a, Italy (1998) XX 1 (51, woman) swept away by flood her body was found 10km downstream two weeks later Fella b, Italy (1998) X XX 1 (57, man) swept away and buried by sediments and debris flow

X X XX 5 members of the same family (mother, 49, Father, 52; children 27, 23, 14) swept Cassano Murge, Italy away in their car when crossing a bridge that collapsed (2005) 1 driver (24, man) swept away in his car by a small torrent 1 other drowned in a small river during the flood Fortezza, Italy (1998) XX X XX X 5 tourists (German speaking family) were swept away in car when it was hit by a debris flow Calonge, Spain (2005) X XX 1 woman (75) swept away by the floodwaters

XX X A couple (49 and 44) and the brother of the woman (41) all Andorran. Main roads Cambrils, Spain (2004) were closed but they continued their trip using a secondary road. The river dragged them when they tried to cross it in their car. La Farinera, Spain X XX 1 woman (83) drowned in her home (2000) X X XX 2 died when their car fell into the river through a destroyed bridge Magarola, Spain (2000) 2 policemen died during the search for the first two victims Duszniki Zdroj c, X Drowning (but no other details) Poland (1998) Klodzko Gmina c, XX 1 man (64) drowned whilst he was trying to catch some boards carried by the water – Poland (1997) there was some speculation that he was drunk Klodzko Town a, c X X X 6 people drowned (5 men 41, 58, 60, 61, 81) and 1 woman (57). One of the men (81) did not want to be evacuated and stayed in his flooded apartment; one was visiting a Poland (1997) friend and was swept away when he was leaving. Miedzylesie b, Poland XX X X 1 man (34) was sleeping when his bedroom was flooded and collapsed into the (1997) floodwaters Stronie Slaskie a, Poland XX X A man (45) drowned when washed from a bridge whilst he was rescuing children (1997) standing there X XX 7 older residents died from crush injuries as their buildings collapsed. A younger man was killed by a falling beam, A ninth victim (who initially survived Troubky, Czech (1997) the collapse of her home) died in hospital a few days later, The homes were constructed of clay brick Olomouc, Czech (1997) X 1 drowned when trying to escape the floodwaters

Otrokovice, Czech X X 1 drowned when trying to escape on an inflatable bed (1997) Dresden, Germany XX 1 was killed in a collapsed building (2002) 5 motor vehicles deaths (F46, F46, M52, M55 and M70), 7 drowned in homes, (F84, F54, F67, F75 and F77) died in a dike break drowned after water in homes (F52, F71 Gard, France (2002) and F84), 2 died in their homes from heart attacks (F72 and M52) 3 people who were outside drowned (F46) and two of those were attempting to rescue animals (M35 and M?) 5 tourists were killed in campsites (M42 and 2 children 2 and 6, M74 and F34) Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

There are however a number of events occurring in areas that have relatively large populations exposed which also reported zero deaths. Table 2.7 shows the data collected from these events in order to investigate more fully possible intervening factors. In all of the zones it was reported that there were few or no language difficulties and the speed of onset was considered to be a medium risk (i.e. the rise of the flood water occurred over a few hours). Data was also collected on the presence or absence of debris in the flood waters, and for all of the data sets large debris was observed during the flood, including such objects as cars, trees, rocks and parts of damaged buildings. As can also be seen the hazard ratings that have been calculated for each of the zones are comparable to those flood events where fatalities did occur, therefore it can be argued that the flood events are severe enough and have the flood characteristics for deaths to have occurred. It is difficult, if not impossible, to state with certainty why the flooding in these zones did not lead to any deaths. However the analysis did attempt to identify common factors that may have contributed to the fact that the populations exposed managed to survive the flood waters. These factors can be summarised as follows:

• Rural properties: smaller populations at risk and a lower concentration of buildings and domestic properties; floodwaters more able to spread out over the floodplain, thereby dissipating energy and less likely to cause injury and death; more space for debris to move between properties. • Type of buildings: multi-storey housing allowing people to escape the floodwaters until either they subsided or people were rescued or evacuated. • Flood warning: nine out of the sixteen cases where no deaths were experienced reported some kind of flood warning prior to the onset of flooding, whether formally or informally delivered. This might be a significant factor about why people were able to take appropriate action and therefore reduce the loss of life. • Evacuation: seven out of the sixteen zones reported that there was some evacuation either prior to the flood or directly afterwards (thereby avoiding the possibility of subsequent flood- related deaths). The levels and timings of the evacuations vary considerably between hazard zones. Evacuation appears to be highly significant in at least three of the risk zones: Bystrzyca Klodzka zones A and B and Eilenburg in the Mulde region of Germany. However, in addition to the positive impacts of evacuation there is also the possibility that the act of evacuation might in fact increase a person’s likelihood of injury or death (Jonkman and Kelman, 2005; Drobot et al., forthcoming). It appears that the effectiveness of evacuation procedures in preventing fatalities is very closely linked to both the type and characteristics of the flooding as well as the timeliness of the evacuation and associated flood warning.

In addition to the presence of the variables described above (or a combination of these variables) there is also likely to be an element of chance. The case study of the flood in Duszniki Zdroj Zone B could be an example of where this played a part, where, on the face of it the circumstances of this flood suggest that loss of life and significant numbers of injuries would be experienced. For example, the area is one which has high degrees of tourism and recreation and it has been noted that there are large numbers of guest houses and hotels located within the risk zone. The flood experienced was also close to the height of the tourist season in July 1998 and began in the late evening/night. There were no official flood warnings and the time of day (the flood started in night time of a public holiday) may have meant that observations of the flood waters rising could have been hampered. In addition to the large numbers of visitors (which one would assume had low flood awareness) there were also large numbers of elderly people among the ‘at-risk’ population. It was estimated that a third of the population in resident in the area were over the age of 75. It can therefore only really be speculated as to why, with a flood of up to 4 metres depth with velocities up to 10 m/s, there were no recorded deaths.

Two other case studies completed for the research can be used to illustrate the role of chance: the Gard flooding in France in 2002 and the Boscastle flood in England in 2004; both of these events were flash floods.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 39 29th February 2008

Table 2.7: Data for those events where no deaths occurred - key characteristics are shown in bold

FLOOD CHARACTERISTICS AREA CHARACTERISTICS PEOPLE CHARACTERISTICS

Flood event Depth Velocity Hazard Time of Flood Duration Land use Flood warning system Type of Buildings Building Evacuation Population Age Risk (m) (m/s) Rating collapse 75+ Awareness

Bystrzyca 5 10 53.5 Night, holiday 1 day Urban None Multi-storey houses None 630 people 450 (est) 25 Low Klodzka A evacuated over the two zones Bystrzyca 2 5 12 Night, holiday 1 day Rural, fields, forest None Single-storey houses 2 1050 (est) 60 Low Klodzka B

Duszniki 4 10 43 Evening/ night 12 hours Tourist and None Multi-storey (including guesthouses, None None 450 (est) 150 Low Zdroj B recreation hotels, restaurants, parks stadium)

Klodzko Town 4 Up to 5 23 Night, holiday 2 days Industrial, fields Warned by the municipal police Two sports stadiums, single-storey None Some attempts, 30 (est) 5 Low D by megaphones houses, railway track, storehouse but few wanted to leave

Klodzko 2 5 12 Late evening/ 3 days Rural (fields None used Single, single-storey houses None People evacuated 120 (est) 10 Low Gmina A night holiday mainly) after the flood event

Klodzko 3.8 5 21.9 Late evening/ 2 days Rural; forests and None Single-storey houses, scattered 1 None 100 (est) 7 Low Gmina B night holiday fields (majority)

Ladek Zdroj 3 5 17.5 Noon/ 24 hours Rural, scattered Warned by the municipal police Two-storey houses 1 Evacuation was 180 (est) 14 Low A settlement by megaphones announced - 10 afternoon evacuated

Ladek Zdroj 4.6 5 26.3 Noon/ 24 hours Urban, concentrated Warned by the municipal police Multi-storey buildings None Evacuation was 250 (est) 20 Low B settlement and fire brigades by megaphones– announced - 60 afternoon also radio warnings Evacuated

Ladek Zdroj 2 5 12 Noon/ 24 hours Rural, scattered Warned by the municipal police Two-storey buildings None Evacuation 300 (est) 22 Low D settlement and fire brigades by megaphones– announced but no- afternoon also radio warnings one was evacuated

Meidzylesie A 2.5 10 27.25 Middle of day, 12 hours Concentrated None Multi-storey houses 1 None 370 (est) 35 Low holiday settlement

Meidzylesie B 1.5 10 16.75 Middle of day, 12 hours Fields, meadows, None Single, single-storey buildings None None 140 (est) 12 Low holiday forest single buildings

Polanica 5 10 53.5 Late evening, 12 hours Urban concentrated Warned by the municipal police Multi-storey houses (majority) 3 None 420(est) 30 Low Zdroj A night settlement by megaphones

Stronie 2 10 22 Night time, 3 days Rural, fields Warned by the municipal police Single-storey houses (ground floor None People evacuated 400 (est) 100 Low Śląskie B holiday and fire brigades by megaphones– plus an attic) after the flood also radio warnings event

Erlln 1.6 4 7.7 Morning 3 days Residential No official but 25% received an Detached houses (mainly two None None 100 (est) 19 Medium ‘unofficial’ warning storeys)

Grimma 3 7 23 Afternoon 3 days Mixed use (city People warned by the police at Mostly 2-3 storey houses, schools, 50 None organised, 1200(est) 100 Low centre with around midnight, the peak of the churches. In the main streets: shops but 100 people commercial, flood did not occur until 2pm the in the ground floor, residential in needed to be residential etc) following day the upper floors rescued post- flood

Eilenburg 2 4 10 5-6pm 12 days Residential Very good. Reported that 98% of Detached houses, mainly two storeys None Most people left 250-300 (est) 55 Medium people warned by police 5 hours 5 hours before before the flood the flood Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

2.10 Additional case studies

2.10.1 The Gard flood of 2002, France The Gard flood of 2002 is an example of severe flash flooding over a wide area. A separate report containing a more detailed discussion of the case study is available on the FLOODsite website (Lutoff and Ruin, 2007 - Project Report Number T10-07-03; also see Ruin, 2007). The Gard region is a flash flood prone département located at the foothills of the Cévennes mountains close to the Mediterranean sea, in Southern France. Due to the nature of the flooding (i.e. small catchments with very short response times) it is not possible to issue flood warnings. The disaster area covered 297 municipalities (80% of the Gard department), taking 23 lives.

Lutoff and Ruin (2007) report fatalities being mostly old and disabled people (9 of them died in their homes), and road users (5 people). During this event tourists also appeared to be vulnerable with a total of 5 victims, two of whom were children, who perished when they were not able to evacuate their campsites quickly enough. About 600 vehicles were involved in the operation rescuing 2,940 people. 40 of these vehicles were lost and 200 were damaged. 1,260 people were rescued by 20 helicopters. The flood event started on a Sunday night when less people were on the roads compared to weekdays. Considering that more than 200 school buses transporting 4,000 children circulate on week days in this sector, this gives an indication of the potential risk. Thus, if the flood had happened during the height of the tourist season and during a week day, many more fatalities are likely to have occurred. The authors conclude that mobility is one of the main circumstances of fatalities and that a way should be found to include this in the risk to life model.

2.10.2 The Boscastle flood of 2004, UK Boscastle (and neighbouring Crackington Haven) are examples of risk zones whereby the risks from flooding to loss of life were extremely severe, yet no deaths resulted. It is a useful application for the Risk to People methodology as not only does it highlight the significance of the data input to the model, and the fact that a range of values might be a more appropriate output rather than an absolute prediction of deaths, it also highlights factors that the current methodology does not really consider.

The Boscastle case study (discussed in detail in Priest et al., 2007) illustrates how the context of modern flood warning practices, incident management and search and rescue scenarios can be significant in reducing loss of life. It also highlights how a series of circumstances and the efforts of rescuers can greatly alter the chances of experiencing fatalities from flood events. The event was used as another UK case study from which to calibrate the Risk to People model. The flood shares similarities with many of the Continental European flood events as well as the Lynmouth flood of 1952. The Lynmouth flood formed part of the basis for the calibration of the original model and was therefore modelled by the Risk to People methodology with some success. Therefore it might be anticipated that the Risk to People model should perform well when applied to the Boscastle case study.

Boscastle is a small village located on the North Coast of Cornwall in the South West of England. The step-sided valleys funnelled the floodwaters resulting in severe flooding that was bad enough to seriously damage some residences and completely destroy others. In addition, cars were transported by the flood waters from the visitors’ car park down the main stream, out into the harbour and out to sea (Environment Agency, 2005c). One of the major characteristics of the flooding was the speed with which the waters rose. The fast and deep nature of the flood waters and the fact that there was little or no warning of the flood event meant that there were many people trapped in Boscastle, mainly within the residential and commercial buildings. An extensive search and rescue effort took place.

The low numbers of injuries and the fact that no one was killed in the flood, and in particular in the village of Boscastle where flood waters were at their most ferocious, has been described as a miracle (Rowe, 2004), see Figures 2.11 and 2.12. As well as some good fortune being evident during the flood

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 41 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 it is also the case that there were a number of different aspects which meant that no-one lost their lives in the flooding and injuries were minimised. The first aspect that was fortuitous was the timing of the flood event itself.

Figure 2.11: Photograph taken in the centre of the village during the Boscastle flood. (© David Flower6)

Figure 2.12: Example of some of the debris moved by the floods in the Boscastle flood event (© David Flower7)

6 Photograph have been taken from ICDDS (2007)

7 Photograph have been taken from ICDDS (2007)

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 42 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

The flooding occurred during daylight hours and therefore it was possible for a number of people in the area to see the rain and witness the waters rising. Rowe (2004) also describes how many members of the local community phoned friends, family and neighbours in the town and in neighbouring areas warning them of the danger. Despite no effective official flood warning prior to the flood event, the reaction of the authorities and members of the public, once it was realised that flooding was occurring and that lives were threatened, was extensive. Had the flood occurred at night it is possible that lives would have been lost, as rescue by helicopter would have been more difficult.

The presence of visitors and tourists in an area during flooding might be considered to be problematic due to the fact that they are likely to have a lower understanding of the flood risk and what actions to take at the time of an event. However, this was not the case with this event as among the local population awareness of the flood risk was generally low and the actions of many visitors was considered to be one of the contributing factors to successful rescue of people from the floodwaters. Many people (mainly young to middle-aged males) were reported to have returned to areas of the village where people required rescuing after ensuring that their own families were safely away from the danger areas (Figure 2.13) (Rowe, 2004). In addition, there is some anecdotal evidence to suggest that the presence of surfers (i.e. mainly young, able-bodied males) who were present in Boscastle that day were fundamental to the assistance of those less able members of the Boscastle community to escape from the flood waters (B.Watts, Environment Agency, pers. comm.). Finally, a number of military helicopters were fortuitously in the local area, and along with a Coastguard helicopter, were able to airlift 97 people from the village, many of whom were stranded on the roofs of buildings (South West Resilience Forum, 2006).

Figure 2.13: Members of the public rescuing others from the floodwaters in Boscastle (© David Flower8)

The Boscastle and Crackington Haven risk zones are examples whereby the risks from flooding to loss of life were extremely severe, yet no deaths resulted. The range of fatalities predicted by the model varied from 0.67 to 109.2 for Boscastle and from 0.13 to 4.69 for Crackington Haven, depending upon the values input. Boscastle illustrates that factors such as the timing of the flood, unofficial warning systems and evacuation are major components that are missing from the current approach. These will

8 Photographs have been taken from ICDDS (2007)

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 43 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 also be more important when considering the types of flood events which are more common in Continental Europe; that is floods that can be deep, have a rapid onset, high velocity and where evacuation is a more common and necessary response.

2.11 Calibration of the UK Risk to People Model

2.11.1 Sensitivity analysis In the following sections the general applicability of the Risk to People model for Continental European flood events is evaluated. A sensitivity analysis was carried out on the European data collected and the model results for the European case studies are described and discussed.

In order to show how much influence each component of the model equation has on the outcomes, the attributes of each baseline variable were changed in turn, while the other variables were held constant. Carlisle Zone B was chosen as the baseline for the analysis since it is arguably illustrative of a ‘typical’ UK flood. The summarised results are shown in Tables 2.8 to 2.10.

Table 2.8: Sensitivity Analysis – Nz and Hazard Rating

Nz Depth Velocity Debris HR Predicted Predicted % Factor score injuries fatalities change Carlisle zone B 420 1.0 0.5 0.0 1 12 0.24 Population x 2 840 1.0 0.5 0.0 1 24 0.48 100 Population x 3 1,260 1.0 0.5 0.0 1 36 0.72 200 Change depth to 2m 420 2.0 0.5 0.0 2 24 0.95 300 Change depth to 2.5m 420 2.5 0.5 0.0 2.5 30 1.49 525 Change depth to 3m 420 3.0 0.5 0.0 3 36 2.14 800

Change velocity to 1m/s 420 1.0 1.0 0.0 1.5 18 0.53 125

Change velocity to 1.5m/s 420 1.0 1.5 0.0 2 24 0.95 300

Change velocity to 2m/s 420 1.0 2.0 0.0 2.5 30 1.49 525

Change DF to 0.5 420 1.0 0.5 0.5 1.5 18 0.53 125 Change DF to 1 420 1.0 0.5 1.0 2 24 0.95 300

Table 2.9: Sensitivity Analysis – Area Vulnerability

Flood Speed of Nature of AV score Predicted Predicted % Warning onset area injuries fatalities change Carlisle zone B 2 1 2 5 12 0.24 Change warnings to 3 3 1 2 6 14 0.29 20 Change speed of onset to 3 2 3 2 7 17 0.33 40 Change nature of area to 3 2 1 3 6 14 0.29 20 Change all to 3 3 3 3 9 21 0.43 80

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 44 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 2.10: Sensitivity Analysis – People Vulnerability Predicted1 Predicted Nz HR AV PV (%) % change injuries fatalities Carlisle zone b 420 1.0 5 28 12 0.24 change PV to 100% 420 1.0 5 100 42 0.84 253 change PV to 50% 420 1.0 5 50 21 0.42 77 change PV to 0% 420 1.0 5 0 0 0.00 -100 change all high with PV 25% 420 14.5 9 25 274 79.47 33,332 change all low with PV 25% 420 0.5 3 25 3 0.03 -87 change all high with PV 50% 420 14.5 9 50 548 158.95 66,764 change all low with PV 50% 420 0.5 3 50 6 0.06 -73 change all high with PV 100% 420 14.5 9 100 1,096 317.90 133,628 change all low with PV 100% 420 0.5 3 100 13 0.13 -47

1 The cells shaded grey show where predicted injuries exceed the at-risk population

2.11.1.1 Summary Changes to the Hazard Rating component of the model have dramatic, yet plausible, results. Changes to the Area Vulnerability component have a much less pronounced effect on the model outputs, largely because of the limited range of values that can be selected. The model is hugely sensitive to People Vulnerability because of its function as a multiplier in the final calculation of injuries. The sensitivity analysis also revealed a structural flaw in the model in that when all variables are high (by UK standards) and PV is equal to or greater than 50 per cent, there are actually more people injured than are resident in the hazard zone.

2.11.2 Model application When applied to the Carlisle floods of January 2005 (the only non-calibration event) the Risk to People model generates a result that is consistent with reality. This correlation is obviously influenced by the inclusion of the calibration events, around which the model was designed, and the small size of the sample (see Priest et al. (2007) for full model results and discussion).

However, when applied to the European case studies, the Risk to People model results can only be described as erratic, as Table 2.11 shows (note that this table is a summary of the model inputs and results, the full array of data can be seen in Priest et al. (2007) Appendix D). In most cases the model overestimates the number of deaths, and while some of these overestimates are moderate, such as La Farinera, Spain, others are huge, such as the Klodzko Town zone c. It is also clear from the table that the HR values of continental floods are significantly higher than those so far obtained in the UK.

There is no significant correlation (parametric or nonparametric) between the predicted and actual fatalities in the European case studies. In two cases the model underestimates the number of deaths. The Cambrils case study result (two deaths predicted, three occurred) appears to be a function of the very low PV score of 1 per cent. It is not clear why the Troubky case underestimates the number of deaths (2 deaths predicted, 9 occurred) although this could be due to incorrect zoning (the importance of correct zoning information is discussed below). The correct prediction for Fortezza of five fatalities is probably due to the small number of people in the hazard zone (estimated at 60) which offsets the high HR value of 39.

The Klodzko town (c) case study has an extremely high Hazard Rating. This is responsible for a gross distortion of the predicted injuries and fatalities values, revealing another structural flaw in the model in that more deaths than injuries are predicted.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 45 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 2.11: Summary of Risks to People Model results compared to actual fatalities in Europe Predicted Predicted Actual Country Flood event Nz HR injuries deaths deaths Italy Fella a 400 9.8 47 9 1 Italy Fella b 500 11.5 58 13 1 Italy Cassano Murge 20,000 17.5 4,263 1,492 7 Italy Fortezza 60 39.0 7 5 5 Spain Calonge 1,300 3.0 137 8 1 Spain Cambrils 2,000 6.0 14 2 3 Spain La Farinera 200 6.0 33 4 1 Spain Magarola 300 47.5 160 152 4 Poland Duszniki zdroj 120 43.0 66 57 7 Poland Klodzko gmina 1,050 21.9 210 92 1 Poland Klodzko town zone a 200 23.0 58 27 1 Poland Klodzko town zone c 2,500 69.2 1,330 1,841 5 Poland Miedzylesie 876 12.0 134 32 1 Poland Stronie Slaskie 2,000 43.0 826 710 1 Czech R Troubky 2,010 2.6 37 2 9 Czech R Olomouc 28,200 2.8 384 22 2 Czech R Otrokovice 19,000 3.9 389 31 1 France Gard 230,510 49.0 121,986 119,546 23 Germany Dresden 300 9.1 44 8 1

2.11.3 Summary of limitations of the current Risk to People model for application in a European context The Risk to People model was developed for the UK under a number of constraints. The first is that the model was specifically designed for floods in England and Wales; this means that, because there are so few flood fatalities in the UK as a whole, there were very few suitable events available for model development, calibration and testing. In addition to the project’s time and resource constraints, the other condition on the product was that its results had to be mappable; this restricted the type of data that could be used in the model to those with a locational element. The Hazard Rating component of the formula was also not designed for the major rivers and mountainous catchments of Continental Europe. The extreme values for HR generated by the European data contribute to the dramatic over- predictions that have been described.

The model was thus found to contain two structural weaknesses: a Hazard Rating of greater than 50 yields the result that more fatalities are predicted than injuries; when HR and PV values are high the model becomes unstable and tends to predict more injured people than are in the hazard zone.

Insufficient account was also taken of institutional arrangements such as evacuation and rescue operations in the Area Vulnerability component of the model. In addition, the model is hugely sensitive to People Vulnerability which is arguably of less importance in the wider European context than it is in the UK.

Nevertheless, the model results for the UK case studies are reasonable estimates, despite the logical inconsistencies that have been identified in the formula. Unfortunately the model is less successful when applied to the Continental European case studies, and generally over-predicts the fatalities. This

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 46 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 is partly because the high hazard ratings associated with floods in Continental Europe and the logical flaw in the model that simply did not apply at the comparatively low HR values obtained in the UK.

Although redesigning the model would ideally require good quality data from many more flood events than were available for this research, some simple alterations have been made to the existing model in an attempt to improve its predictive capability. These modifications to the model are discussed in the following Section.

2.12 Adaptations and revisions to the UK Risk to People model

Potential refinements and revisions to the current UK model were explored and attempts made to develop a new product which more accurately models Continental European fatalities. The aim was to reduce the number of fatalities predicted by the model to more realistic levels whilst retaining the relationship between predicted fatalities and flood severity, as represented by the Hazard Rating.

The European data was also examined in detail in order to better understand which factors are most important to the assessment of risk to life within the wider European context. This was achieved by performing a number of statistical analyses (cluster analyses, multiple regression analyses and Principal Component Analyses) to explore the datasets in more detail and understand the relationships between the different variables and the numbers who died during flooding events (see Priest et al., 2007, for full details). These data variables can be seen in Table 2.12. It was necessary to include those additional factors which were not examined within the Risk to People methodology but which have been highlighted as being potentially important in assessing the risk flooding in Continental Europe. There are a number of ways in which each of these variables might therefore have been included. In a number of cases these factors were represented in more than one way to try to capture whether or not they are significant. The revised model is presented in the next section.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 47 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 2.12: The different variables included within the statistical analyses Variable Description Risk to People Factors included within the original Risk to People methodology Population at risk (Nz) The population who are in the risk zone and who are at risk from the flood Depth The depth of the flood (in metres) Velocity The Velocity of the flood (in metres per second) Debris factor (DF) A factor related to the amount of debris found in the flood waters (A factor of 0, 0.5 or 1 is assigned) Speed of onset Speed of onset is represented as an ordinal scale: 1. Slow onset 2. Medium onset 3. Rapid onset Nature of area An ordinal scale with variables representing the characteristics of the area: 1. Multi-storey apartments 2. Typical residential area (2-storey homes); commercial and industrial properties 3. Bungalows, mobile homes, busy roads, parks, single storey schools, campsites, etc. Flood warning Flood warning is represented as an ordinal scale: 1. Good - Where the majority of people at risk received warning with adequate lead time 2. Fair - Warning received by some of the population with adequate lead time 3. None - No warnings received Percentage of the long-term ill An estimation of the percentage of the ‘at risk’ population that are long-term unwell Percentage of over-75 years An estimation of the percentage of the ‘at risk’ population that are over the age of 75 years Alternative Risk to People These are where actual measures have been used rather than putting these factors variable representations on an ordinal scale Lead time of flood warning This factor is presented in minutes and constitutes the length of time before the flooding a warning was given. Actual time to flood onset This is the time from when the first signs of change were noticed in water levels or from signs of precipitation to when the floods threaten people and their property (in minutes). New variables Awareness of flood risk This has been developed on an ordinal scale of; 1. High awareness 2. Medium awareness 3. Low awareness Building collapse (1) This is a presence/absence factor which indicates whether buildings collapsed or not. Building collapse (2) This variable is the actual numbers of buildings that collapsed during the flooding Building collapse (3) A ratio of buildings collapsed to population at risk from flooding Evacuation (1) A presence/absence variable (represented as 1, 0) representing whether people were evacuated from the flooded zone Evacuation (2) The percentage of people at risk that were evacuated AFTER the flood event Evacuation (3) The percentage of people at risk that were evacuated BEFORE or DURING the flood event Evacuation (4) The percentage of people at risk that were evacuated EITHER before or after the flood event Flood Duration The length of time a zone has experienced flooding (in hours) Population with language This is a presence/absence value which is evaluating whether language constraints constraints (1) are problematic Population with language This is the percentage of the population at risk that have particular language constraints (2) constraints. Time of the flood This is a factor (1,0) which indicated whether the flooding occurred during the day or the night.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 48 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

2.13 Recommendations for refining the UK Risk to People methodology

The analysis of the circumstances leading to fatalities from flooding in the examples examined highlights potential differences between the significant factors causing deaths in the UK and those in other parts of Europe. Several recommendations for refinement of the Risk to People methodology were therefore made to make it more applicable within a wider European context.

• People vulnerability should be given less prominence: due to the more severe and different types of flooding that occur in Europe the vulnerability components may not be as relevant in the wider European context. This is especially the case with flash flooding where the flood waters have high velocity and depth, therefore anyone caught up in these waters will be vulnerable. The analyses on age indicates that those who appear to be most vulnerable (e.g. the elderly) are often no more vulnerable than others, therefore the degree to which this is useful for the model, particularly with regard to flash flooding, needs to be questioned. Indeed, there is some evidence to suggest that in certain circumstances more elderly people might be, by their behaviour, less vulnerable as they are more likely to recognise their own limitations and remain inside away from the flood waters, whereas other younger more able-bodied people might venture outside. In some flood events it might be wiser for people to stay within their properties if they are not able to evacuate from the area prior to the floodwaters arriving.

• Place more prominence on the effect of flood warnings: the analysis of those flood events where no deaths occurred illustrates the potential importance of both official and unofficial flood warning systems. Although official warning is included within the Risk to People methodology, this might be having a more important impact than the model is currently recognising.

• Place more prominence on type of buildings: from the analysis of the ‘no death’ events the presence or absence of two-storey buildings is recognised in having an impact on the loss of life. Again this is partly included within the current methodology, within the nature of the area component. However, the model could be altered slightly to strengthen this component.

• Include a population density factor: the density of properties in an area and the land use appear to have an impact upon whether flooding fatalities. Therefore, it would be interesting to investigate whether it is possible or significant to put in a measure of urban density to reflect whether it is a rural or urban landscape that is being affected.

• Place more prominence on the debris factor: within the original Risk to People model a factor for the presence or absence of debris is included. This factor may be having a greater influence on the likelihood of fatalities in Continental Europe and therefore the significance of this factor should be examined. In particular, whether the presence of debris is more likely to lead to building collapse needs to be explored.

Thus changes to the model were also guided by the sensitivity and statistical analyses. The sensitivity analysis determined how much influence the model components had on the model results and this, in turn, determined the type of amendment made to the model. For instance, the sensitivity analysis illustrated that Area Vulnerability had the smallest effect on the outcomes so the amendments described sought ways of increasing the effect of this parameter. The People Vulnerability component was shown to have a strong (and potentially destabilising) influence on model outcomes and so the amendments were geared towards reducing its effect, and thus it’s potential to trigger model instability. The statistical analysis determined which cases were included in the revised model. The Czech case studies, for instance, were not included because the flood characteristics were significantly different to the other case studies. In addition, we also attempted to account for the effects of evacuation by subtracting evacuation figures (where these were available) from the Nz value. However, this amendment had no significant effect on the model predictions and the results are not included here.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 49 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Some amendments were more successful than others. The most promising amendment to the formula was to convert the PV percentage to an absolute value and then use the reciprocal of this in the calculation. However, good quality data from many more case studies would be necessary to evaluate the usefulness of this, and other, formula modifications.

As mentioned from the outset, there are a number of variables of importance that the Risk to People model does not consider; such as building collapse and evacuation. However, in most UK events these appear of limited importance. One aspect that has not really been incorporated but was fundamental in the Boscastle situation is the impact of search and rescue on the outcome of an event and the number of fatalities. In addition to missing variables, some components of the model appear to function better than others, for instance the Environment Agency have little confidence in the function of the debris factor within the equation (Environment Agency, pers. comm.).

However, the model has only really been tested and calibrated on a small number of case studies, many of which occurred a number of years ago since when there have since been major advances in flood warning, communication technologies and changes to search and rescue practices. The case studies used to calibrate and test the approach are all quite severe flooding; the approach has also not really been tested on more minor flooding cases in the UK. Similarly, due to the nature of the data supplied it has not been possible to test this approach with Continental European flooding of a similar magnitude and type to that of the UK. If data could be found for these types of events, it may be found that this approach performs much better and could be applied in these circumstances.

The following Section will therefore introduce an alternative ‘threshold’ approach to modelling risk to life that simply focuses on the variables identified as the most significant; that is depth and velocity of the floodwaters, and the exposed population (including mitigating variables that might impact upon the numbers of people exposed to flooding).

2.14 Proposed European Risk to Life model This Section explores the development of a ‘threshold’ approach to the assessment of risk to life in Europe. The aim is to combine information on the factors considered to be the most significant when estimating risk to life with other empirical evidence and theoretical knowledge to develop a simple ‘banded’ or threshold approach to risk to life assessment.

2.14.1 Conceptual model At the highest level, the theory explored within the original Risk to People model is still applicable to the situation in Europe (Figure 2.14).

E = f (F, L, P)

Where E is the nature/extent of effects (on those exposed), F is the flood characteristics (depth, velocity), L is the location characteristics (inside/outside buildings, nature of housing etc) and P is the population characteristics (age, health). HR Wallingford (2003, p15).

Figure 2.14: Expression characterising the effects on people exposed to the flooding risk

Adding the numbers of people who are exposed to the hazard the Risk to People approach is illustrated in the following diagram (Figure 2.15).

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 50 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 2.15: Method for calculating flood risks to people HR Wallingford (2005c, p2)

However, according to the previous analyses of the situation within other parts of Europe this does not really fully explain the situation leading to risk to life from flooding. In addition, the flood hazard component in this model excludes the role of building collapses on the eventual numbers of people who are fatally injured by flood waters. As explained previously this was the main cause of fatalities in a number of case studies and if it occurs is a significant threat to life. The role of evacuation (either formally organised or informally undertaken) and its positive impact upon the numbers of people who are exposed to the hazard is also not really considered within the Risk to People methodology. This is likely to be a consequence of the fact that in the UK planned evacuation from flooding events is very rarely necessary. Although Figure 2.16 does address the broad issues involved in assessing the risk to life from flooding, when considering the situation in Europe it is possible to propose a more specific conceptual model.

Risk to Life in Europe = f(F, Ex, Pv, -M)

Where:

F is the flood hazard characteristics (e.g. depth, velocity), Ex is the exposure to the hazard (related to the nature of the area, whether people can avoid direct contact with the flood waters without being threatened by building collapse), Pv is people vulnerability (the importance of this variable will depend upon the severity; for instance in some circumstances, such as very severe floods, this variable is redundant) M are the mitigating actions (is there sufficient warning to enable people to evacuate the area entirely or seek appropriate shelter from the flood waters).

Figure 2.16: Proposed conceptual model for assessing risk to life

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 51 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

2.14.2 Hazard factors From the statistical analyses undertaken as part of this project and from other studies investigating the potential number of lives lost from flooding, the most important factors are flood depth and velocity and it is these hazard factors that will first be considered. A number of studies have explored the notion of human stability in flowing water which (Abt et al., 1989; Karvonen et al. 2000; Lind et al., 2004; Jonkman et al., 2005; HR Wallingford, 2005, 2005; Jonkman and Penning-Rowsell, forthcoming). Each of these studies has in their different way attempted to identify those speeds and depths where human safety is compromised. It must be noted from the outset that this is a very difficult undertaking as each individual may differ, as a person’s ability to withstand flood water is dependent on such factors as their height, weight, age and physical condition. Lind et al. (2004) have identified that the drag on a person in the water also contributes to their instability in flood waters. In their study they argue that this drag is determined by the clothing that they are wearing, as generally those who are wearing bulkier clothing will experience more drag in the water. Additionally, a person’s ability to remain stable in flood waters may also be affected by other environmental conditions such as whether there is good visibility, the temperature of the water and the presence or absence of debris. Despite there being obvious differences in people’s ability to maintain stability and safety during flooding, it is possible to identify broad thresholds at which flood waters will become dangerous to people. First it is necessary to consider how depth and velocity should be represented.

Many of the studies have used the product number of depth multiplied by velocity as a function to describe when human stability is compromised. Although as described above it is difficult to quantify the thresholds, Abt et al. (1989) argue that the product number can be used as a rough indicator or predictor of when a human would become unstable in flood waters. However, HR Wallingford (2003) argue that velocity is more important than depth and offer the alternative equation depth x (velocity + 0.5) as they argue that this offers a better estimation of the hazard to people. Using this formula, Figure 2.17 identifies thresholds where different individuals are in danger from the flood waters, based on their height and weight. The resulting thresholds are displayed in Table 2.13.

Figure 2.17: Loss of stability figures taken from Abt et al. (1989) and Karvonen et al. (2000)9 Source: HR Wallingford (2005, p7).

9 The estimates of the heights and weights illustrated in this graph were based on figures taken from UK Department of Health data with average figures.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 52 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 2.13: Flood hazard thresholds as a function of depth and velocity d x (v + 0.5) (m2/s) Degree of flood Description hazard <0.75 Low Caution “Flood zone with shallow flooding water or deep standing water” 0.75 – 1.25 Moderate Dangerous for some (i.e. children) “Danger: Flood zone with deep or fast flowing water”

1.25 – 2.50 Significant Dangerous for most people “Danger: Flood zone with deep fast flowing water”

>2.50 Extreme Dangerous for all “Extreme danger: Flood zone with deep fast flowing water” Source: HR Wallingford (2005, p8).

Since the model developed here will include other factors where people either might find themselves by chance, or use to provide shelter (such as motor vehicles or buildings), it is also necessary to identify similar thresholds for a depth x velocity function for these features. After re-plotting the results found in both Karvonen et al. (2000) and Abt et al. (1989), and adopting the approach undertaken in HR Wallingford (2005a) Figure 2.18 calculates the depth-velocity function for each variable and identifies potential new thresholds based on a depth-velocity product.

Depth 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5 Depth x Velocity 0.25 0.0625 0.125 0.1875 0.25 0.3125 0.375 0.4375 0.50 0.5625 0.625 0.5 0.125 0.25 0.375 0.5 0.625 0.75 0.875 1.00 1.125 1.25 1 0.25 0.5 0.75 1 1.25 1.5 1.75 2.00 2.25 2.5 Velocity 1.5 0.375 0.75 1.125 1.5 1.875 2.25 2.625 3.00 3.375 3.75 2 0.5 1 1.5 2 2.5 3 3.5 4.00 4.5 5 2.5 0.625 1.25 1.875 2.5 3.125 3.75 4.375 5.00 5.625 6.25 3 0.75 1.5 2.25 3 3.75 4.5 5.25 6.00 6.75 7.5 3.5 0.875 1.75 2.625 3.5 4.375 5.25 6.125 7.00 7.875 8.75 4 1 2 3 4 5 6 7 8.00 9 10 4.5 1.125 2.25 3.375 4.5 5.625 6.75 7.875 9.00 10.125 11.25 5 1.25 2.5 3.75 5 6.25 7.5 8.75 10.00 11.25 12.5

Figure 2.18: Recalculation from the HR Wallingford (2005a) of the ‘danger’ thresholds for a range of different depths and velocities. Adapted from HR Wallingford (2005a, p9).

It is acknowledged that an individual’s ability to remain stable in flood waters is not only a function of the depth and velocity of the water, and their height and weight, but may also be linked to other variables such as their physical condition, or other circumstances such as lighting, underfoot conditions, cold water, presence of debris in the water or if they are carrying a load or assisting other people; indeed different experiments and assessments of this variable have provided different estimates of these figures. Therefore based on the literature, a range of variables are provided in Table 2.14, presenting a low, mid, high and extreme hazard estimate to all of the thresholds.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 53 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 2.14: Flood hazard thresholds as a function of depth and velocity Depth x velocity (m2/s) Hazard from Description Low range Mid-range High Range flooding Caution <0.1 <0.25 <0.50 Low “Flood zone with shallow flood water or deep standing water” Dangerous for some (i.e. children and elderly) 0.10 to 0.30 0.25 to 0.50 0.25 to 0.70 Moderate “Danger: Flood zone with deep or fast flowing water” Dangerous for most people 0.40 to 0.70 0.5 to 1.10 0.90 to 1.25 High “Danger: Flood zone with deep fast flowing water” Dangerous for all 0.9 to 1.25 1.10 to 3.00 >3.00 Extreme “Extreme danger: Flood zone with deep fast flowing water” Source: Adapted from HR Wallingford (2005a, p8).

2.14.3 People exposure The analysis above indicates the potential risk to human stability from flood waters. However not everyone will be outside during a flood; some people will be inside or will seek shelter from direct contact with the floodwaters. For instance, people who are located in a well-constructed three storey building made of bricks are likely to be less exposed or vulnerable to the threats of flood waters, than those who are staying in mobile accommodation. Those people in vehicles may be exposed and thus vulnerable to flooding. Reiter (2000) (taken from the RESCDAM work) suggests ranges for the risk and damages that can occur to personal vehicles. Although these have been applied within a dam break analysis these are still useful and relevant when considering other types of flooding. These thresholds can be seen in Table 2.15.

Table 2.15: Critical parameters for damage to motor vehicles applied to dam break flooding Risk of damage Damage parameter (depth x velocity) m2/s Small damages, small Medium damages, Total damages, very danger Medium danger high danger Personal cars < 0.3 0.50 - 0.60 > 0.6

Source: Reiter (2000, p11).

2.14.3.1 Vulnerability of areas It is necessary to identify those locations which will have a higher degree of vulnerability than others, either by their character or by the presence of a large number of vulnerable people (e.g. children and/or elderly or sick people). For instance areas with campsites, locations of mobile properties or areas with large open recreational spaces will provide little shelter from direct contact with flood waters. Urban residential areas or other locations with buildings should in theory provide a higher level of shelter from flood waters, although the degree of this will vary according to the building type, the quality of construction and the number of storeys that a property has. However, in severe flooding the integrity of this shelter may be compromised by either structural damages or in some instances total collapse. It is therefore necessary to comment upon the resistance of buildings to flood waters.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 54 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

2.14.3.2 Building integrity and collapse There are a number of factors to consider when attempting to place thresholds on the ability of a building to withstand floodwaters. These include the materials that the building is made from (e.g. timber, brick, stone or a mix of materials), the quality of the original construction, construction methods, the age and condition of the property. These variables make it very difficult to put exact thresholds on building collapse and therefore those described below are only broad guidelines. It is recommended that if the model developed here is applied to a specific location, these thresholds be altered to reflect the local style and nature of the building fabric.

This study adopts the thresholds developed by Karvonen et al. (2000) which identified different depth velocity product thresholds (see Table 2.16) for partial and total damage of three types of properties; wooden (unanchored), wooden (anchored) and masonry concrete and brick. For masonry, concrete and brick buildings there is a velocity threshold (of >2m/s) which is also significant to the onset of structural damages.

Table 2.16: Flood conditions leading to the partial or total damage of buildings in Finland House type Partial damage Total damage vd = velocity x depth Wood framed- unanchored vd > 2 m2/s vd > 3 m2/s Wood framed-anchored vd > 3 m2/s vd > 7 m2/s Masonry, concrete and brick v > 2 m/s and v > 2 m/s and vd > 3 m2/s vd > 7 m2/s Source: Karvonen et al. (2000, p18).

It is interesting to look at these thresholds in the context of the European flood events that have been studied in this report. Building damage has not been recorded in all cases but the numbers of buildings or structures (such as bridges or roads) that suffered collapse has been identified. Of the 52 deaths recorded 11 were caused by properties either partially or fully collapsing, whereas a further 7 deaths were caused by the collapse of roads or bridges, see Table 2.17.

In most cases the buildings do appear to be collapsing under the situations explained above, although the relationship between the depth-velocity product and structural collapse appears to be very complex as there are a number of anomalous events. Half of the events studied had properties or bridges that totally collapsed, most of these did have values that were greater than the suggested depth-velocity product of greater than or equal to 7m2/s by Karvonen et al. (2000). Additionally, there were a number of events that also had a depth velocity product number greater than 7m2/s, yet no building collapses were reported. This does not necessary mean that the threshold levels chosen are incorrect; merely that buildings might not always be completely destroyed or even that total collapses are not accurately reported. A further reason for this might be related to the depths and velocities reported within the hazard zone. In some cases these values are potentially the highest values and therefore not all buildings will be equally exposed to these severe levels. Moreover, if it is an urban area that is affected not all of the buildings will be equally exposed to the same flood depths and velocities; some buildings may have a sheltering or shadowing affect on other buildings, reducing the severity of the depth- velocity value that they experience.

There are also a couple of anomalous events where the depths and velocities were relatively low yet there were many buildings that collapsed. These events, Troubky and Olomouc, are both categorised as slow rise floods and both zones are found within the Czech Republic. In both of these events hundreds of buildings collapsed (337 and 208 respectively). In this situation it appears to be building materials and construction that are to blame for the high level of building collapse in this region. The majority of the properties in this rural region were constructed cheaply some as long ago as 150 years, out of plaster over unfired brickwork (FLOODsite partner survey response).

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 55

Table 2.17: Instances of building collapse within the European flood events Flood event / Hazard N (Z) Depth Velocity Depth11 No. of collapsed buildings Other collapsed structures Total Collapse-related fatalities 10Zone x event velocity fatalities Fella a 400 2.5 3 7.5 0 – But 40 homes buried by sediment and 1 almost 100 damaged by the flood waters Fella b 500 3 3 9 0 – But 250 houses buried by sediments and 1 hundreds more damaged Cassano Murge 20000 3 5 15 0 Collapses of roads and railways. 7 5 members of the same family swept away in their car when crossing a bridge that collapsed; Fortezza 60 4 9 36 0 Collapses of roads and railways. 5 Calonge 1300 1 2.5 2.5 0 1 Cambrils 2000 3 1.5 4.5 0 3 La Farinera 200 2 2.5 5 1 Two small bridges were 1 reported to have collapsed Magarola 300 5 9 45 0 Bridge collapse 4 Two people died in their car when a bridge collapsed. In addition two policemen who went to look for them also perished Duszniki zdroj zone c 120 4 10 40 0 7 Klodzko gmina zone c 1050 3.8 5 19 4 1 Klodzko town zone a 200 4 5 20 0 1 Klodzko town zone c 2500 6.5 10 65 5 5 Miedzylesie zone b 876 2 5 10 4 1 1 man (34) was sleeping when his bedroom was flooded and collapsed into the floodwaters Stronie Slaskie zone a 2000 4 10 40 0 1 Troubky 2010 2 0.3 0.6 337 - 50% of the buildings were destroyed12 9 All nine died in collapsed buildings Olomouc 28200 2 0.42 0.84 208 - Most collapsed buildings were made 2 of unfired bricks or a mix of materials Otrokovice 19000 3.5 0.34 1.19 0 – but 1082 houses were damaged of which 1 562 were out of use for a long time Dresden 300 1.8 4 7.2 1 - Around a further 20 buildings that were 1 The loss of life occurred when the building collapsed being constructed were destroyed. Erlin 100 1.6 4 6.4 0 0 Grimma 1200 3 7 21 50 0 Eilenburg 275 2 4 8 0 0 Bystrzyca Klodzka zone a 450 5 10 50 0 0 Stronie Slaskie zone b 400 2 10 20 0 0 Polanica Zdroj zone a 420 5 10 50 3 0 Miedzylesie zone c 138 1.5 10 15 0 0 Miedzylesie zone a 369 2.5 10 25 1 0 Ladek Zdroj zone d 300 2 5 10 0 0 Ladek Zdroj zone b 250 4.6 5 23 0 0 Ladek Zdroj zone a 180 3 5 15 1 0 Duszniki zdroj zone b 450 4 10 40 0 0 Klodzko gmina zone a 120 2 5 10 0 0 Bystrzyca Klodzka zone b 1050 2 5 10 2 0 Klodzko town zone d 30 4 5 20 0 0 Klodzko gmina zone b 100 3.8 5 19 1 0

10 The shaded events are those which experienced some property or other structural collapse 11 These are the events with a depth velocity product score of greater than 7m2s. 12 The events highlighted in red have high numbers of buildings that collapsed, but with low depths and velocity

Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

It is also necessary to add an additional depth-velocity threshold to the model to reflect at which point the majority of buildings will be vulnerable from collapse; thereby making people directly vulnerable not only to the flood waters but to the effects of building collapse itself. The additional threshold of > 7m2/s where all buildings that are in direct contact with the flood waters are vulnerable (assuming that velocity ≥ 2m/s) is added in Table 2.18.

Table 2.18: Flood hazard thresholds as a function of depth and velocity Depth x velocity (m2/s) Hazard from Description Low range Mid-range High Range flooding Caution <0.1 <0.25 <0.50 Low “Flood zone with shallow flood water or deep standing water” Dangerous for some (i.e. children and elderly) 0.10 to 0.30 0.25 to 0.50 0.25 to 0.70 Moderate “Danger: Flood zone with deep or fast flowing water” Dangerous for most people 0.40 to 0.70 0.5 to 0.11 0.90 to 1.25 High “Danger: Flood zone with deep fast flowing water” Dangerous for all “Extreme danger: Flood zone with deep fast 0.9 to 1.25 1.1 to 7.00 >7.00 Extreme flowing water where properties will be prone to structural damage; poorly-constructed and wooden buildings may collapse.” Dangerous for all “Extreme danger: Flood zone with deep fast >7.00 >7.00 >7.00 Extreme flowing water where all properties are vulnerable to collapse or serious structural damage” Source: Adapted from HR Wallingford (2005a, p8).

Table 2.19 integrates all of the different components of the vulnerability of the area and identifies those areas that are best able to reduce the chances of an individual’s exposure to the floodwaters.

Table 2.19: Categories indicating an area’s vulnerability to flood waters 1. Low Vulnerability 2. Medium vulnerability 3. High vulnerability These areas will have multi-storey This category is a typical This category will include areas buildings that would provide safer residential area with mixed land which provide little protection to places for people to escape to. use (e.g. residential and industrial individuals from flood waters. The These areas will also have well- mixes) and mixed types of type of land use within this zone constructed properties made out of buildings (i.e. areas with single and would include mobile homes, solid materials such as masonry multi-storey properties) campsites. It also includes areas of concrete and brick poorly-constructed properties which would be more vulnerable to structural damage or collapse and single storey dwellings which would only offer limited protection in deep waters.

Similar to the approach adopted in the Risk to People methodology, three categories are proposed to indicate the different vulnerabilities for locations affected by flooding. These categories are based on four main factors: type of land use (i.e. whether there are proper buildings where shelter can be sought); number of floors of a property, indicating whether people are able to escape from flood waters; structural integrity of buildings, their building material and the integrity of construction; and the presence of particularly vulnerable groups or activities (e.g. schools, residential care homes).

T10-07-13_Deliverable_D10_1_ report_V_1_2_P10.doc 29th February 2008 57 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

2.14.4 Risk to Life from flooding From the information above it is possible to construct a threshold model highlighting the consequences of flooding at different depths and velocities using the depth-velocity product. Figure 2.19 combines the thresholds for people directly exposed to the flood waters and the information about whether particular areas are vulnerable and illustrates these thresholds and identifies the risks associated with flood waters at each of the different levels. The model provides four different risk levels each also illustrated by a different colour; Extreme risk (red), High risk (orange), Medium risk (yellow) and Low risk (green).

It is also possible with this model of Risk to Life to provide some indication of the dominating factors leading to injuries and fatalities from flooding of difference levels. This is illustrated in Table 2.20 and Figure 2.19, which comprise the first part of the new Risk to Life model. However, due to the complexity of the factors leading to death, and particularly in relation to those areas in the most vulnerable zones where physically vulnerable properties are found due to poor construction or unsuitable materials, this can only be a broad assessment.

Table 2.20: Main factors leading to fatalities from flooding Depth-velocity Nature of area Main factor leading Description thresholds (m2/S) categories to fatalities <0.25 All Low risk There is low risk to people from the flood waters.

0.25 – 0.50 All People Vulnerability The fatalities are likely to be concentrated dominated – some amongst the vulnerable people e.g. children Behaviour-related either playing in or near flood waters, or elderly people (often trapped in their properties) 0.50 – 1.10 Low and medium Behaviour dominated In most circumstances people will be able to vulnerability find shelter away from the floods, however, deaths and injuries may still occur if people undertake risky activities such as driving through the floodwaters or taking unnecessary risks in the waters 0.75-1.75 High vulnerability Hazard dominated In these situations, fatalities are likely to occur from direct contact with the flood 1.75-7.00 Low and medium waters vulnerability 1.75-7.00 High vulnerability Hazard and building Fatalities will occur if people are in direct collapse dominated contact with the flood waters or if caught in >7.00 All buildings that are structurally compromised by the flood waters.

It is also important to remember that at all levels of flood severity (i.e. those events with a higher depth-velocity component) people vulnerability will remain a factor as those in this category are potentially less able to take action on their own or evacuate from areas. Similarly, the behaviour of people during flooding is also important, particularly on the fringes of the very high hazard zones where depths and velocities will be lower but still will be dangerous. Therefore, undertaking risky or inappropriate activities at higher depth/velocity levels will still impact greatly upon an individual’s risk of injury or death from flooding.

T10-07-13_Deliverable_D10_1_ report_V_1_2_P10.doc 29th February 2008 58 DEPTH x OUTDOOR NATURE OF THE STRUCTURAL RISK TO LIFE FROM FLOODING FATALITY VELOCITY HAZARD AREA DAMAGE FACTOR MID-RANGE 3. High vulnerability Risk to life in this scenario is extreme as not only are those in (including mobile homes, the open very vulnerable to the effects of the flood waters but campsites, bungalows and Total collapse may those who have also sought shelter are also very vulnerable due poorly constructed occur. Structural to the fact that building collapse is a real possibility >7m2s-1 properties) damages 2. Medium vulnerability probable (Typical residential area in particular for mixed types of properties) properties with 1. Low vulnerability poor quality (Multi-storey apartments and building fabric masonry concrete and brick properties) 3. High vulnerability All those exposed to the hazard outside will be in direct danger dominated (including mobile homes, from the floodwaters. Those living in mobile homes will be at campsites, bungalows and risk from the high depths and velocities and those in single poorly constructed storey dwellings will be at risk from not being able to escape to Extreme properties) upper floors. Those in very poorly constructed properties will Hazard and building collapse also be vulnerable from structural damages and/or building 1.10 to Dangerous for all collapse. 7 m2s-1 2. Medium vulnerability All those exposed to the hazard outside will be in direct danger (Typical residential area from the floodwaters. Damages to structures are possible. mixed types of properties) Structural damages Those in unanchored wooden frames houses are particularly possible vulnerable. With very deep waters there is the risk of some not being able to escape. 1. Low vulnerability All those exposed to the hazard outside will be in direct danger (Multi-storey apartments and from the floodwaters. In this scenario those residing in these masonry concrete and brick properties have the lowest risk although structural damages are properties) still possible in wooden properties 3. High vulnerability Structural damages Those outside are vulnerable from the direct effects of the (including mobile homes, and collapse floodwaters. In addition, those in single storey dwellings will be campsites, bungalows and possible for vulnerable in deeper waters. People will also be afforded little

poorly constructed properties with protection in mobile homes and campsites. Those in very Hazard Dominated properties) poor quality poorly constructed properties will also be vulnerable from building fabric structural damages and/or building collapse. Vehicles are also likely to stall and lose stability. 2. Medium vulnerability Anyone outside in the floodwaters will be in direct danger. It is (Typical residential area at this point where behaviour becomes significant as structural 0.50 to 2 -1 mixed types of properties) Structural damages are less likely; those inside should mostly be 1.10 m s damages – less protected. Vehicles are likely to stall and lose stability. Are High likely and less people undertaking inappropriate actions such as going outside severe when is it not necessary? 1. Low vulnerability Anyone outside in the floodwaters will be in direct danger from

Dangerous for most (Multi-storey apartments and the floodwaters. It is here at this point where behaviour masonry concrete and brick becomes significant as structural damages are less likely so Behaviour properties) those inside should be on the most part protected. Vehicles are dominated likely to stall and lose stability. Are people undertaking inappropriate actions such as going outside when is it not necessary? 3. High vulnerability Structural damages Only the most vulnerable should be in direct danger from the (including mobile homes, possible for floodwaters. (e.g. children and the elderly); in this category the campsites, bungalows and properties with shelter may not protect them. Motor vehicles may become poorly constructed poor quality unstable at these depths and velocities. Those in very poorly properties) building fabric constructed properties may also be vulnerable from structural 0.25 to damages. 2 -1 0.50 m s 2. Medium vulnerability Only the most vulnerable should be in direct danger from the (Typical residential area floodwaters (e.g. children and the elderly). Motor vehicles may mixed types of properties) Unlikely become unstable at these depths and velocities. Those who

Moderate seek shelter should be safe. 1. Low vulnerability Only the most vulnerable should be in direct danger from the (Multi-storey apartments and floodwaters. (e.g. children and the elderly). Motor vehicles may People vulnerability People vulnerability Dangerous for some

masonry concrete and brick become unstable at these depths and velocities. Those who some dominated though

properties) seek shelter should be safe. fatalities behaviour-related

3. High vulnerability A very low risk to adults either out in the open or who is in a (including mobile homes, property. There may be a threat to the stability of some vehicles campsites, bungalows and even with these low depth-velocity factors. <0.25 m2s-1 poorly constructed Unlikely properties) 2. Medium vulnerability (Typical residential area Low mixed types of properties) Caution 1. Low vulnerability Low risk (Multi-storey apartments and masonry concrete and brick properties) Figure 2.19: First half of threshold model indicating the risk of life from flooding

59 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

2.14.5 Mitigating factors After defining the factors that contribute to the flood hazard it is important to realise that in most cases actions are taken to not only reduce the impacts of these flood hazards but also to reduce the public’s exposure to the hazard.

In many instances in Europe, evacuation is a real and important option that is used to mitigate the impacts of the worst flood events. Indeed, half of the flood events studied in this report (n=17) had some level of official evacuation taking place either before, during or following the onset of flooding, although the numbers of people that were evacuated varied from 2% of the population (in Olomouc, Czech Republic) to over 95% of the population (in Eilenburg, Germany). Official evacuation levels will no doubt be affected by the lead time before the warning, but may also be impacted by other variables such as previous experience of flooding, levels of trust not only in the forecast but also in that their property will not be looted, as well as the availability of a flood-free route out of the danger zone. Self-evacuation levels may also depend upon previous experience of the hazard, availability of transport and knowledge of a flood-free route. In this first iteration of the model two levels for evacuation are presented: full and partial.

Obviously, in the situation where all (or close to all) of the population at risk is moved from the situation the level of risk will be reduced as there will be no (or few) people left for the flooding to impact. When adapting and developing the model further or tailoring it for a specific region, the evacuation component might be split into more categories and rough percentages given for each (e.g. <25% evacuation, 25-50% evacuation, 50-75% evacuation or >75% evacuation) thus providing an extra level of detail.

In situations other than evacuation, the mitigating action undertaken may still mean that large numbers of people are still in the area of the hazard, but are not necessarily exposed. In other, less severe, circumstances an effective flood warning with a longer lead time will be sufficient to allow people to get out of direct contact with the flood waters and then potentially reduce their chances of being injured or killed. The effectiveness of this shelter however, is also related to the severity of the depths and velocities experienced and the structural integrity of the properties. For instance, in the situation whereby the most extreme hazard is experienced (i.e. where depth-velocity is >7m2/s) all those remaining in the area will be exposed to the possibility of partial or total collapse of buildings. However despite the threat of structural collapse people are still likely to be safer sheltering in buildings rather than exposing themselves to the flood waters. In other situations however, the speed of onset might be such that a flood warning is not possible or there may be areas where flood warnings are not available, not effective or are too late to provide people with the options of taking action. To represent this type of situation two more categories have been added to the model: no flood warning available or a short lead time and effective flood warning with adequate lead time.

Table 2.21 indicates the four broad categories of mitigating factors that have been added to the model. However it is acknowledged that these categories might be able to be more refined and detailed when applying the model to a specific region or town. In this instance these broad categories will indicate the principles or the methodology and can be used at a broad scale to highlight whether the risk category explained in Sections 1.6.3 and 1.6.4 is realistic to the area, or whether official flood response measures will reduce these categories. For instance, if an area is subject to flash flooding, broad-scale evacuation prior to the event is unlikely, as is flood warning with adequate lead time, therefore category 4 (No flood warning or short lead time) should be assigned. Similarly, this category would be assigned to regions that have no flood warning systems or regions where, in the case of past flooding, flood warnings have not been effective. Regions that have effective flood warning systems with good evacuation plans in place might chose to select categories 1 (Full evacuation) or 2 (Partial evacuation) depending upon past experiences.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 60 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 2.21: Categories of mitigating actions Mitigating factor Description Outcome

1. Full evacuation A flood warning and then evacuation order have been Most people have been able following a flood provided in sufficient time before the flooding. There are to evacuate the area and warning plans and resources in place to enable the majority of those therefore not exposed to the in the risk zone to evacuate (or self-evacuate) from the risk flooding zone. 2. Partial A flood warning and then evacuation order have been Some people have evacuation provided with sufficient time before the flooding. There evacuated the area following following a flood are plans and resources in place to enable some of those in receipt of a warning. The warning the risk zone to evacuate (or self-evacuate). Some of those rest of the population remain remaining in the area at risk would have received a flood in situ but have had the warning and will have had the opportunity to seek shelter. chance to receive a flood In some instances the partial evacuation might be targeted warning and have had the at vulnerable groups, such as children or the elderly. time to react. Some people may not receive the warning or advice to evacuate or may choose not to leave the area. 3. Flood warning A percentage of the population will have received a flood Most people remain in situ with adequate lead warning with enough time to react and get to safety. There and therefore may be time with mixed may however be mixed effectiveness of this warning exposed (or expose responses system and/or mixed responses to the warning. This may themselves). But the flood depend on the dissemination strategy, the experience of warning will permit people the warning agency and/or the people and awareness of the the time to react and seek most appropriate action to take. shelter. 4. No flood This may be a region that has no (or an ineffective) flood The majority of the warning or short warning service. It may also be a area of flash flooding, population are in situ when lead time where forecasting and warning with sufficient time for an flooding occurs and are not effective warning to be delivered is difficult. warned or warned very close to the flood occurring.

All of these categories do not take into consideration “unofficial” or unplanned action by individuals or communities (though the model could be refined to do this if a region has good information about how the public have reacted in past floods) nor are they able to account for the effectiveness of a person’s response to these factors. Thereby local experience of the flooding situation and how people react to flooding would need to be added to the model and the categories refined in order to improve on the assessment provided, although it is difficult to see how this could be done.

2.14.6 A new approach to assessing Risk to Life from flooding in Europe Figure 2.20 combines the hazard and exposure thresholds and the mitigating factors to provide a model from which the risk to life can be assessed at different scales. Although only a broad assessment, this approach can be applied at a range of scales (though as mentioned previously it might be developed and refined further for a local or regional context). The purpose of the model is to allow flood managers to make a general and comparative assessment of risk to life and also where to target resources before, during and after flooding. The final column in Figure 2.20 provides some high-level suggestions, although these again may be made more detailed depending on the scale of application and the purpose (e.g. for planning evacuation, the locating of emergency shelters or where enhancing flood risk awareness should be targeted). One advantage of this scaled approach over the methodology applied in the Risk to People methodology is that although it is still necessary to zone areas according to the hazard characteristics and vulnerability, it is not necessary to zone them homogenously for both features; this is best illustrated below when looking at risk mapping). Therefore, areas of differing hazard and areas of differing vulnerability can overlap and intersect and a risk level be assessed for each different combination.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 61

RISK TO LIFE WITHOUT ANY MITIGATING ACTIONS RISK TO LIFE WHEN MITIGATING ACTIONS ARE APPLIED D x V OUTDOOR NATURE OF THE STRUCTURAL RISK TO LIFE MAIN RISK TO 14 15 FACTOR HAZARD AREA DAMAGES CATEGORIES WITHOUT FACTOR MITIGATION FACTOR LIFE ACTIONS MID- (VULNERABILITY) MITIGATION LEADING TO WITH 13 RANGE FATALITIES ACTIONS 3. High vulnerability No flood warning or short lead time The emphasis in these situations should be on search and rescue if this is possible. Resources should be targeted on identifying the areas or groups of (including mobile homes, people who are in most immediate danger. Particular efforts should be made to ensure that the population are risk-aware and that they know how to campsites, bungalows and respond when flooding of this magnitude occurs. If possible, a flood warning service should be developed. Where possible ensure that people do not poorly constructed reside in areas that suffer such severe flooding where there is no flood warning. properties) Risk to life in this scenario is Flood warning with adequate lead time The emphasis in these situations should be on search and rescue if this is possible. Resources should be targeted on identifying the areas or groups of extreme as not only are those in and mixed response people who are in most immediate danger. Particular efforts should be made to ensure that the population are risk-aware and that they know how to the open very vulnerable to the respond when flooding of this magnitude occurs. >7m2s-1 Total collapse effects of the flood waters but 2. Medium vulnerability Focus should be on ensuring that as many people as possible are evacuated safely (i.e. with enough time not to be caught out in the flooding). Search may occur. those who have also sought shelter Partial evacuation following a flood (Typical residential area and rescue operations can then focus on the smaller number of people who remain in the risk area. Particular efforts should be made to ensure that the are also very vulnerable due to the warning mixed types of properties) Structural population are risk-aware and that they know how to respond when flooding of this magnitude occurs. damages fact that building collapse is a real 1. Low vulnerability probable possibility Full evacuation following a flood If a full evacuation occurs, most people will be out of immediate danger. Efforts therefore need to focus on ensuring the well-being of those who have (Multi-storey apartments and in particular for warning been evacuated from the area. This zone is not green because there is likely to be some risk during the evacuation period and also because of the fact masonry concrete and brick that these events may not occur in isolation (i.e. people might need to move through areas of more minor flooding.) properties) properties with poor quality No flood warning or short lead time The emphasis in these situations should be on search and rescue if this is possible. Resources should be targeted on identifying the areas or groups of building fabric people who are in most immediate danger. Particular efforts should be made to ensure that the population are risk-aware and that they know how to All those exposed to the hazard respond when flooding of this magnitude occurs. If possible, a flood warning service should be developed. Where possible ensure that these types of outside will be in direct danger land use are not located in areas that could suffer such severe flooding. from the floodwaters. Those living dominated The emphasis in these situations should be on ensuring that as many people as possible are warned and know what to do to protect themselves, 3. High vulnerability in mobile homes will be at risk Flood warning with adequate lead time following this search and rescue should be carried out if this is possible. Proactively, education should focus on flood risk awareness and preparation. (including mobile homes, from the high depths and and mixed response Where possible ensure that these types of land use are not located in areas that could suffer such severe flooding. campsites, bungalows and velocities and those in bungalows poorly constructed will be at risk from not being able Partial evacuation following a flood Focus should be on ensuring that as many people as possible are evacuated safely (i.e. with enough time not to be caught out in the flooding). Search and rescue operations can then focus on the smaller number of people who remain in the risk area. Particular efforts should be made to ensure that the properties) to escape to upper floors. Those in warning very poorly constructed properties population are risk-aware and that they know how to respond when flooding of this magnitude occurs. Where possible ensure that these types of land 1.1 to will also be vulnerable from use are not located in areas that could suffer such severe flooding. 7 m2s-1 structural damages and/or building Full evacuation following a flood If a full evacuation occurs, most people will be out of immediate danger. Efforts therefore need to focus on ensuring the well-being of those who have

Hazard and building collapse collapse. warning been evacuated from the area. This zone is not green because there is likely to be some risk during the evacuation period and also because of the fact that these events may not occur in isolation (i.e. people might need to move through areas that are flooding to a lesser degree.) Where possible ensure

that these types of land use are not located in areas that could suffer such severe flooding. Extreme No flood warning or short lead time The emphasis in these situations should be on search and rescue if people are exposed if this is possible. Resources should be targeted on identifying the areas or groups of people who are in most immediate danger (e.g. those on the ground floor). Efforts and resources should also be targeted at Dangerous for all All those exposed to the hazard ensuring that the population are risk-aware and that they know how to respond when flooding of this magnitude occurs. If possible, a flood warning system should be developed. outside will be in direct danger The emphasis in these situations should be on ensuring that as many people as possible are warned and know what to do to protect themselves and from the floodwaters. Damages to Flood warning with adequate lead time 2. Medium vulnerability structures are possible. Those in and mixed response where to go for safety, following this search and rescue should be carried out if this is possible. Proactively, education should focus on flood risk (Typical residential area unanchored wooden frames awareness and preparation. Where possible ensure that vulnerable land uses are not located in areas that could suffer such severe flooding. mixed types of properties) houses are particularly vulnerable. Partial evacuation following a flood Focus should be on ensuring that as many people as possible are evacuated safely (i.e. with enough time not to be caught out in the flooding). Search With very deep waters there is the warning and rescue operations can then focus on the smaller number of people who remain in the risk area. Efforts and resources should also be targeted at risk of some not being able to ensuring that the population are risk-aware and that they know how to respond when flooding of this magnitude occurs. Where possible ensure that escape. vulnerable land uses are not located in areas that could suffer such severe flooding. Full evacuation following a flood If a full evacuation occurs, most people will be out of immediate danger. Efforts therefore need to focus on ensuring the well-being of those who have Structural been evacuated from the area. damages warning Possible No flood warning or short lead time The emphasis in these situations should be on search and rescue if people are exposed if this is possible. Resources should be targeted on identifying the areas or groups of people who are in most immediate danger (e.g. those on the ground floor). Efforts and resources should also be targeted at ensuring that the population are risk-aware and that they know how to respond when flooding of this magnitude occurs. If possible, a flood warning All those exposed to the hazard system should be developed. outside will be in direct danger The emphasis in these situations should be on ensuring that as many people as possible are warned and know what to do to protect themselves and 1. Low vulnerability Flood warning with adequate lead time from the floodwaters. In this where to go for safety. Following this, search and rescue should be carried out if this is possible. Proactively, education should focus on flood risk (Multi-storey apartments and and mixed response scenario those residing in these awareness and preparation. masonry concrete and brick properties have the lowest risk properties) Partial evacuation following a flood Focus should be on ensuring that as many people as possible are evacuated safely (i.e. with enough time not to be caught in the flooding). Search and although structural damages are rescue operations can then focus on the smaller number of people who remain in the risk area. Efforts and resources should also be targeted at ensuring still possible in wooden properties warning that the population are risk-aware and that they know how to respond when flooding of this magnitude occurs. Full evacuation following a flood If a full evacuation occurs, most people will be out of immediate danger. Efforts therefore need to focus on ensuring the well-being of those who have warning been evacuated from the area. No flood warning or short lead time The emphasis in these situations should be on search and rescue if people are exposed if this is possible. Resources should be targeted on identifying Those outside are vulnerable from the areas or groups of people who are in most immediate danger. Efforts and resources should also be targeted at ensuring that the population are risk- Hazard dominated Structural the direct effects of from the aware and that they know how to respond when flooding of this magnitude occurs. If possible, a flood warning system should be developed. Where damages and floodwaters. In addition, those in possible ensure that these types of land use are not located in areas that could suffer such severe flooding. 3. High vulnerability collapse possible bungalows will be vulnerable in Flood warning with adequate lead time The emphasis in these situations should be on ensuring that as many people as possible are warned and know what to do to protect themselves and (including mobile homes, for properties deeper waters. People will also be where to go for safety. Following this, search and rescue should be carried out if this is possible. Proactively, education should focus on flood risk afforded little protection in mobile and mixed response campsites, bungalows and with poor quality awareness and preparation. Where possible ensure that these types of land use are not located in areas that could suffer such severe flooding. poorly constructed homes and campsites. Those in properties) building fabric very poorly constructed properties Partial evacuation following a flood Focus should be on ensuring that as many people as possible are evacuated safely (i.e. with enough time not to be caught out in the flooding). Search will also be vulnerable from warning and rescue operations can then focus on the smaller number of people who remain in the risk area. Particular efforts should be made to ensure that the structural damages and/or building population are risk-aware and that they know how to respond when flooding of this magnitude occurs. Where possible ensure that these types of land use are not located in areas that could suffer such severe flooding. collapse. Vehicles are likely to also stall and lose stability. Full evacuation following a flood If a full evacuation occurs, most people will be out of immediate danger. Efforts therefore need to focus on ensuring the well-being of those who have warning been evacuated from the area.

No flood warning or short lead time During flooding resources should be targeted at assisting the most vulnerable in the community and ensuring that they are safe both before and after Anyone outside in the floodwaters flooding and helping them to avoid the risk. Introduction of a flood warning service where possible. If flood warnings are really not possible, attention will be in direct danger from the should focus on ensuring the population is risk aware and know how to respond during flooding. floodwaters. It is here at this point Flood warning with adequate lead time Flood warnings should be provided as early as possible to warn as many people as possible. Focus should be on instructing the population on the best where behaviour becomes significant course of action to ensure that they act appropriately and get to, or remain in, a place of safety. Resources before flooding should be targeted at raising 0.5.to 2. Medium vulnerability as structural damages are less likely and mixed response 2 -1 public awareness of flood risk and how to respond to flood warnings. 1.1 m s (Typical residential area so those inside should be on the most High High mixed types of properties) part protected. Vehicles are likely to Partial evacuation following a flood Where possible and where there is time people should be encouraged to evacuate. Efforts should be targeted on assisting those who are unable to stall and lose stability. Are people warning evacuate themselves. Proactively, efforts should be made to ensure that the population is aware of the risk of flooding and know how to respond during undertaking inappropriate actions flooding. such as going outside where is it not

Dangerous for most for Dangerous Structural necessary? Full evacuation following a flood If a full evacuation occurs, most people will be out of immediate danger. Efforts therefore need to focus on ensuring the well-being of those who have damages – less warning been evacuated from the area likely and less No flood warning or short lead time During flooding resources should be targeted at assisting the most vulnerable in the community and ensuring that they are safe both before and after severe Anyone outside in the floodwaters flooding and helping them to avoid the risk. Introduction of a flood warning service where possible. If flood warnings are really not possible, attention will be in direct danger from the should focus on ensuring the population is risk- aware and know how to respond during flooding. floodwaters. It is here at this point Flood warnings should be provided as early as possible to warn as many people as possible. Focus should be on instructing the population on the best 1. Low vulnerability where behaviour becomes significant Flood warning with adequate lead time as structural damages are less likely course of action to ensure that they act appropriately and get to, or remain in, a place of safety. Resources before flooding should be targeted at raising (Multi-storey apartments and and mixed response so those inside should be on the most public awareness of flood risk and how to respond to flood warnings. masonry concrete and brick part protected. Vehicles are likely to properties) Partial evacuation following a flood Where possible and where there is time people should be encouraged to evacuate. Efforts should be targeted on assisting those who are unable to stall and lose stability. Are people evacuate themselves. Proactively, efforts should be made to ensure that the population is aware of the risk of flooding and know how to respond during undertaking inappropriate actions warning flooding. such as going outside where is it not necessary? Behaviour dominated Full evacuation following a flood If a full evacuation occurs, most people will be out of immediate danger. Efforts therefore need to focus on ensuring the well-being of those who have been evacuated from the area. warning Only the most vulnerable should be No flood warning or short lead time During flooding resources should be targeted at assisting the most vulnerable in the community and ensuring that they are safe both before and after Structural in direct danger from the flooding and helping them to avoid the risk. Introduction of a flood warning service where possible. If flood warnings are really not possible, attention floodwaters. (e.g. children and the should focus on ensuring the population is risk-aware and know how to respond during flooding. damages possible elderly). They are obviously most The warning should be concentrated on raising awareness of the potential for danger and in particular at those most likely to be exposed to the flood 3. High vulnerability for properties vulnerable as they are less able to Flood warning with adequate lead time (including mobile homes, waters (e.g. water-related recreational activities). Resources should be targeted to assist the most vulnerable groups. Education should also be focussed with very poor save themselves from the flood and mixed response campsites, bungalows and waters and in this category the on informing people about the dangers of certain activities (e.g. driving or swimming). quality building poorly constructed shelter may not protect them. Motor Partial evacuation following a flood Evacuation efforts should target those most vulnerable in these areas and ensure that they are assisted to leave the area or moved to more secure properties) fabric vehicles may become unstable at warning locations. Resources should be used to ensure the well-being of those who have been evacuated as well as those who remain. these depths and velocities. Those in very poorly constructed properties Full evacuation following a flood An unlikely scenario for this type of risk level. Resources should be concentrated on evacuating and assisting those most vulnerable, particularly in this may also be vulnerable from zone where there is limited shelter. structural damages. warning

No flood warning or short lead time During flooding resources should be targeted at assisting the most vulnerable in the community and ensuring that they are safe both before and after flooding, and helping them to avoid the risk. Introduction of a flood warning service where possible. If flood warnings are really not possible, attention 0.25 to Only the most vulnerable should should focus on ensuring the population is risk-aware and know how to respond during flooding. 0.5 m2s-1 be in direct danger from the 2. Medium vulnerability Flood warning with adequate lead time The warning should be concentrated on raising awareness of the potential for danger and in particular at those most likely to be exposed to the flood floodwaters (e.g. children and the (Typical residential area and mixed response waters (e.g. water-related recreational activities). Resources should be targeted to assist the most vulnerable groups. Education should also be focussed elderly). Motor vehicles may mixed types on informing people about the dangers of certain activities (e.g. driving or swimming). become unstable at these depths of properties) Evacuation is unlikely to be needed in this scenario as the risk to people is low. There may be the need to target specific groups who may be at risk and velocities. Those who seek Partial evacuation following a flood warning (e.g. old people’s homes or schools). Moving vulnerable people like the elderly may have adverse long-term impacts.

Moderate shelter should be safe. Full evacuation following a flood An unlikely scenario for this type of risk level. Resources should be concentrated on evacuating and assisting those most vulnerable. Attention should then be turned to assisting in the protection and clear-up operation of those whose properties have been flooded. warning Dangerous for some for Dangerous Unlikely No flood warning or short lead time During flooding resources should be targeted at assisting the most vulnerable in the community and ensuring that they are safe both before and after flooding, and helping them to avoid the risk. Introduction of a flood warning service where possible. If flood warnings are really not possible, attention should focus on ensuring the population is risk-aware and know how to respond during flooding. Only the most vulnerable should The warning should be concentrated on raising awareness of the potential for danger and in particular at those most likely to be exposed to the flood be in direct danger from the Flood warning with adequate lead time 1. Low vulnerability waters (e.g. water-related recreational activities). Resources should be targeted to assist the most vulnerable groups. Education should also be focussed floodwaters. (e.g. children and the and mixed response (Multi-storey apartments and on informing people about the dangers of certain activities (e.g. driving or swimming). elderly). Motor vehicles may masonry concrete and brick become unstable at these depths Partial evacuation following a flood Evacuation is unlikely to be needed in this scenario as the risk to people is low. There may be the need to target specific groups who may be at risk properties) and velocities. Those who seek warning (e.g. old people’s homes or schools.) Although it must be remembered where the risk is low, moving vulnerable people like the elderly may have an shelter should be safe. adverse impact upon long-term health. Resources should be used to ensure the well-being of those who have been evacuated as well as those who though some behaviour-related People vulnerability dominated remain.

Full evacuation following a flood An unlikely scenario for this type of risk level. Resources should be concentrated on evacuating and assisting those most vulnerable. warning 3. High vulnerability (including mobile homes, There may be the need to target specific groups that may be at risk (e.g. old people’s homes or schools) though most actions will be by warning people campsites, bungalows and to be cautious around the floodwaters. poorly constructed In terms of risk to life flood warnings and

A very low risk to adults either properties) evacuation will have little impact as there out in the open or who is in a <0.25 m2s-1 2. Medium vulnerability property. There may be a threat to is a low risk to life from the event itself, Unlikely (Typical residential area the stability of some vehicles even though a flood warning may be used to Low mixed types at these low depth-velocity

Caution ask people to exercise caution and take of properties) factors. 1. Low vulnerability action to reduce damages.

(Multi-storey apartments and Low risk masonry concrete and brick properties) Figure 2.20: Threshold approach to assessing Risk to Life from flooding in Europe Key Extreme Risk High Risk

13 The mid-range values are presented here. See Sections 9.2 and 9.3 for an explanation of the ranges of hazard variables. 14 Four different mitigation variables are presented in the model. See Section 9.5 for a description of how these might be tailored to specific location or scenario. Moderate Risk Low Risk 15 The actions provided here are indicative. If possible, these should be tailored to the local situation and circumstances.

Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

2.14.7 Application of the threshold European Risk to Life model to the European flood event data The new model has mainly been developed to be able to assess the risk to life from flooding in an area. The model (Figure 2.20) should be used by working from left to right, and users should firstly identify the depth and velocity characteristics of the area of interest to them and select the level which best matches the depth-velocity products estimated for their area. It is then necessary to assess an area’s vulnerability, by examining the land use, type and quality of buildings and whether there are any particularly vulnerable groups of people present. Additionally, it may be necessary to assess whether large numbers of people are likely to be vulnerable in motor vehicles, for instance if a major road crosses the zone of interest. By selecting the hazard and then the vulnerability, an initial assessment of the level of risk for an area is then presented in the column Risk to life categories without mitigation. A user can then select which flood warning or evacuation category is likely to be present within this area and therefore assess the resulting risk to life once this has been applied as mitigating actions may have the effect of reducing the risk.

Following this example, it is useful to apply the model to some of the flood events that have been explored within this study. It must be remembered however, that applied in this way on an event basis the approach will suffer from similar zoning problems to when the Risk to People methodology is applied as it will necessitate the averaging of different events. A range of different types of flooding events and events of different outcomes is illustrated in Table 2.21.

Table 2.21: Application of the threshold model to European flood events Flood event DxV Area Mitigating Risk Event Description m2/s Vulnerability factor Level Deaths Boscastle, 2 Medium No flood 0 The threat of risk to life is high, but deaths were High UK warning averted due to the efforts of search and rescue.

Carlisle Zone 2 Medium Flood 2 The two deaths in the Carlisle event were due to D, UK warning and People Vulnerability as they were two elderly Medium some women living alone who died in their own homes

evacuation who were not warned and were not assisted in evacuating from their homes. Troubky 0.6 High Flood 9 Although the risk in this category is medium the Czech warning main factor was the very poor building fabric. Republic Deaths caused due to collapsed buildings constructed from materials not resilient to flood High waters. When buildings do collapse the deaths

from flooding increase greatly. The majority of the people who died were elderly therefore this appears to reflect the people vulnerability-dominated deaths at this level Klodzko 20 High Flood 1 Victim was killed by direct contact with the flood Town A warning waters. It was argued that many people heeded the Extreme warning and avoided venturing outside. In

addition, there were no instances of building collapse. Calogne 1 Medium Partial 1 A 75 year-old woman drowned in the channel. evacuation Therefore the death in this case is related to both Low people vulnerability and behaviour-related as the

majority of the people close to the channel were evacuated. Cambrils, 3 Medium Flood 3 All three deaths in this incident were behaviour High Spain warning related as they unnecessarily drove their car onto a

flooded road. Eilenburg, 8 Low Full 0 Deaths avoided by action of effective warning and Germany evacuation Medium evacuation

Although the approach is highlighting the severity of flooding experienced, it has been difficult to test whether the approach is differentiating between events, as the data that has been collected as part of

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 63 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 this study are some of the worst floods experienced in Europe over the last two decades. The approach needs further testing on events of many different magnitudes. Similar to the limitations of the Risk to People methodology, this approach cannot take account of chance or all actions that people undertake during flooding. The following Section will demonstrate the use of this approach within a mapping methodology and provide case study illustrations of its capacity to map risk to life from flooding.

2.15 Mapping risk to life

The Risk to Life model developed in this project is based on a decision support tree and has a semi- qualitative approach. The qualitative outputs (Risk level) are assessed with the hazard and vulnerability values of a given area and different conditions. The GIS format used in this methodology is a vector format with polygon topology. The hazard and vulnerability maps are first created with existing data and are then combined through a “union” process, i.e. features of both layers are combined into one feature, while maintaining the original features and attributes.

The Risk to Life model defines different outdoor hazards according to depth-velocity factor thresholds. To produce the hazard map it is necessary to define the extent of the flood and the depth- velocity value for a defined event (associated probability). The best approach is to use output from 1D/2D models. The hazard map could contain a “fixed” flooding area where the depth-velocity would change with different flood scenarios. If the hydraulic model gives the spatial and temporal dynamic of the depth and velocity value (usually in a raster format), it is then useful to have “flexible” flooding areas. This means that for each time-series the hazard map has to be rebuilt according to the threshold values of the Risk to Life model and the output of the hydraulic model.

The vulnerability map contains three levels of information: the vulnerability level, the population component and the mitigation factor. The population component and the mitigation factor are considered as additional information and do not necessarily affect the spatial extent of the features. It is thus mainly the nature of the area that defines the features boundary and the vulnerability scale. The model is based on the type of buildings that could be found in an area. Spatial information usually available includes the type of land-use, such as residential area, industrial and commercial units, recreational area, open field, etc. The vulnerability could change depending on the activities within an area e.g. the presence of schools, hospitals or care homes.

The resulting layer, or more exactly resulting database, is used as an input of the decision tree model to produce the risk map. The decision tree model has been developed in Visual Basic Application on the ARCGIS9.1® platform to represent the decision tree process to facilitate the processing of mapping. Figure 2.21 shows the input interface of the model. The user can change, if necessary, the hazard threshold values and the risk impact factor to calibrate the model. Figure 2.22 presents the output interface of the model for one of the case studies. The results of the model could be mapped with different indicators. A general representation of the mapping methodology is shown in Figure 2.23 and is illustrated with one of the case studies from the Milestone report (Priest et al., 2007) (Thamesmead, UK). The risk level defines for each area the level of the risk to life based on a scale of five qualitative values (low to extreme). The exposure factor gives the potential number of people exposed to the risk multiplied by the risk factor associated with a risk class. It could be expressed per ha (as shown in Figure 2.23). From top to bottom, the hazard map and vulnerability map are spatially unified and are used as an input to the risk to life model. Risk level and exposure factor are the resulting map of the model.

Most of the potential of the model has not yet been tested e.g. comparisons of different events, of population scenarios and of warning and evacuation scenarios. Moreover, the results from the model have not been calibrated with real events. There is thus still much further research that can be undertaken: the model requires further testing and calibration and refinements in order for it to be effectively operationalised. We also strongly believe that participatory processes at case study sites

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 64 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 would be highly relevant to identify vulnerability areas, to build scenarios and to ground-truth the model.

Figure 2.21: Input interface – the risk to life model

Figure 2.22: Results in the Output interface – The risk to life model - UK case study

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 65 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 2.23: Mapping Process applied to the Thamesmead UK case study.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 66 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

2.16 Conclusions and recommendations for further research

2.16.1 Summary of research The overall objective of Task 10 of the FLOODsite project is to focus research efforts on innovative methods to understand, model and evaluate flood damages. One aspect of this damage relates to risk to human life and serious injury resulting from flooding. In order to reduce the risk to life it is necessary to understand the causes of loss of life in floods in order to pinpoint where, when and how loss of life is more likely to occur and what kind of intervention and flood risk management measures may be effective in eliminating or reducing serious injuries and fatalities. The objectives of this research were therefore as follows:

• to further develop a model, or models, that will provide insight into, and estimates of, the potential loss of life in floods, based on work already undertaken in the UK and new data collected on flood events in Continental Europe; • to map, through the use of GIS and building partly on existing work, the outputs of the risk to life model(s) providing estimates of the potential loss of life in floods.

The research took as a starting point the Risk to People model developed in the UK (HR Wallingford, 2005) and assessed the applicability of this model for flood events in Continental Europe, which tend to be more severe. Data on flood events were gathered from 25 locations across six European countries as well as data from an additional case study in the UK. A number of problems were identified with the current model when applied to the flood data from Continental Europe. Firstly, despite the model yielding reasonable estimates for UK case studies, the Hazard Rating component of the formula was not designed for the major rivers and mountainous catchments of Continental Europe and the extreme values for HR generated by the European data resulted in dramatic over-predictions of injuries and fatalities. The model was found to contain two structural weaknesses: a Hazard Rating of greater than 50 was seen to result in more fatalities being predicted than injuries, and when HR and PV values were high the model became unstable and tended to predict more injured people than were in the hazard zones. Moreover, research into the factors surrounding European fatalities also highlighted that more account needs to be taken of institutional arrangements and mitigating factors such as evacuation and rescue operations in the Area Vulnerability component of the model. In addition, the model was hugely sensitive to People Vulnerability, which is arguably of less importance in the wider European flooding context than it is in the UK.

Thus a new semi-qualitative ‘threshold’ model which combines hazard and exposure thresholds and mitigating factors has been developed to assess risk to life from flooding in a wider European context. The model has been designed to be flexible enough to be used and applied at a range of scales, from a broad assessment at a regional or national scale, to a more detailed local scale. This flexibility is essential as not all European countries have detailed flood data that is readily available. The model should be used as a tool to allow flood managers to make general and comparative assessments of risk to life and to consider the targeting of resources before, during and after flooding. The new model also permits simple mapping of risk to life which again can be applied at various scales.

It is recognised that flooding can occur on a wide scale stretching often limited resources and personnel. Therefore, different flood risk zones should not be examined in isolation, their risk needs to be assessed and the outcome integrated into a broader and comprehensive action plan. One advantage of this scaled approach over the methodology applied in the Risk to People model is that although it is still necessary to zone at-risk areas according to hazard characteristics and vulnerability, it is not necessary to zone them homogenously for both features. Therefore, areas of differing hazard and areas of differing vulnerability can overlap and intersect and a risk level can be assessed for each different combination

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 67 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

It is acknowledged that the new Risk to Life model and its thresholds need to be tested further on a range of different flood events to investigate the validity of the approach, and in particular the thresholds selected. Despite this, it is hoped that this approach permits an initial assessment and prediction of the risk to life which can subsequently be enhanced and refined with local knowledge. Thus when applying the model at a high resolution, it is also recommended that the approach is used iteratively, with users applying their own knowledge and experience to tailor the categories and what they contain to their local specific situations. The approach also permits some scope for uncertainty, in particular within the depth and velocity data. The model will of course be most sensitive to error at the thresholds of the different depth-velocity product classes.

Overall, the research has increased the understanding of the factors surrounding fatalities from flood events in the broader European context. It has also highlighted the potential roles of factors such as building collapse, human behaviour and the role of chance in effecting fatalities and injuries, as well as the benefits of mitigating measures such as evacuation and flood warnings.

2.16.2 Remaining issues A number of problems remain in refining the model further. Firstly, the results generated from the application of the new approach are, similar to other models of this type, hugely sensitive to the data input into the model and the different values attributed to the model components. This factor, along with general limitations in the availability of data, have highlighted the need for the establishment of reliable, systematic and consistent methods for collecting data following flood events across Europe, as well as for the need to make available such data that is collected. A key constraint relates to who is responsible for collecting such data, which at present varies from agencies at local, regional and national levels. The EU has recognised the need for greater European co-ordination on flood risk management, therefore it is suggested that protocols are needed to address the data issue. Moreover, any future research project that requires the collection of such data at a European scale needs to allow sufficient time and resources, and requests for data need to be made well in advance of when it is actually needed.

Several questions can also be raised at this point about the purpose of modelling risk to life. For instance, is it aimed at modelling a worse case scenario? It is unlikely that it will ever be possible to estimate accurately the number of deaths from a flood event. Therefore, should the modelling simply be used as a guide to the identification of those areas which are most likely to suffer fatalities from flooding? It is also possible to question the idea of trying to model risk to life for the whole of Europe due to the large differences in types of flood hazard, area and people vulnerability and institutional arrangements.

2.16.3 Recommendations for further research Several recommendations can be made to take this research forward.

• In order to refine the model further good quality data from many more flood events is required than were available for this research. To facilitate the data collection, this would need the cooperation of European governments in making data available rather than having to purchase it on a commercial basis.

• In particular, more data from slow-rising flood events needs to be collected and analysed to further identify factors impacting upon risk to life in these situations.

• A further suggestion could be to produce separate risk to life models for different types of flood events e.g. flash floods (for both urban and rural areas), slow rising floods, coastal floods, dam or dike break or breaching etc., although this would again require large amounts of data.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 68 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

• Exploratory research with flood risk managers, local authorities and other stakeholders across Europe could initially be conducted in order to assess the type of information and models that would be of most use in different situations. This could then form the basis for taking the research forward to produce practical and easy to use tools that are fit for purpose and which take account of available resources.

• In order for the mapping methodology to be effectively operationalised more work is needed. This will include calibrating the model with real events e.g. testing it using comparisons of different events, of population scenarios and of warning and evacuation scenarios. Participatory processes at case study sites would be highly relevant to identify vulnerability areas, to build scenarios and to ground-truth the model.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 69 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

3 Activity 2 Developing models to estimate the benefits from flood warnings

3.1 Introduction

3.1.1 Floods and flood losses in Europe: the background to the development of flood warning systems

During the past decade, a number of European countries have suffered tragic loss of life and enormous economic damage owing to flooding. Some, including north-west Romania, south-eastern France and southern and central Germany, have suffered repeated damaging flooding. It is clear that floods now present serious and growing risks to a significant number of Europeans and to Europe’s economy and society.

Floods are the most common, and also among the most widespread, natural disasters in Europe. Flood fatalities do not appear to be increasing, but flood damages are almost certainly increasingly significantly. The following evidence presents the overall picture. The number of flood disasters reported has been increasing significantly, particularly recently. In 2002 alone, 20 European countries experienced at least one disastrous flood. According to EMDAT, floods have caused 2,626 deaths in Europe over the thirty year period (1973-2002) (Hoyois and Guha-Sapir, 2003): an annual average of 87.5. According to EMDAT, during the past 30 years almost 9 million Europeans were affected by floods, with their growing number being linked with the increasing number of disastrous floods. Taking the period 1991-1995, the European Environmental Agency report the cost of flood damage being between 1991 and 1995 as estimated at €99 billion (European Environmental Agency, 2001; 7): an average annual damage toll of approximately €25 billion.

Flood damages, and trends in flood damages, are difficult to estimate. This is because flood damages are widely diffused, inaccurately reported, difficult to measure and of different types, some of which are more hidden and more difficult to assess than others. In the UK current annual average flood damage is estimated at £1,400 million (€2.1 billion), but under one growth scenario it could expand to around €40 billion by 2080, in the unlikely case that flood management policies and expenditures remain unchanged (Office of Science and Technology, 2004).

Population growth in urban corridors coincident with river valleys, rising standards of living, increased consumption, the location of people and economic assets in floodplains and in coastal zones, and climate change impacts upon flood flows and flood frequency, are all believed to be combining to significantly increase flood damage potential. A strong belief that flood damage potential is rising in Europe is one of the key rationales for the recent introduction of an EU ‘Floods Directive’ (European Parliament, 2007) which seeks to place flood assessment and flood management on a firmer footing in Europe.

3.1.2 A revolution in flood warning systems in Europe

Systems for detecting floods and forecasting them have been in place for many years along some of Europe’s major rivers and in some parts of Europe. Typically flood warning systems evolved from initiatives to collect and record meteorological and hydrological data, and from the foundation and development of meteorological and hydrological institutions. Often institutional missions and scientific interest centred much more in technical development of flood detection and of flood forecasting than upon flood warning communication and warning response (Figure 3.1). This sometimes meant that these ‘downstream’ elements of flood forecasting, warning and response systems received much less attention than was necessary.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 70 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Across Europe there are now many initiatives to improve flood forecasting and warning systems, suggesting that we may be on the verge of a revolution in flood warnings. Initiatives now focus much more upon all of the elements of the flood forecasting, warning and response systems with technical, social, economic and institutional advances being worked upon. The full potential of these initiatives is a significant increase in public safety and security in Europe’s flood zones, and a significant decrease in flood losses including lives lost and damages to property and infrastructure. This potential can be realised as long as the goal of these of initiatives is a holistic and fully integrated one which drives the improvement of the whole flood forecasting, warning and response process, rather than just isolated component parts.

pre-flood preparations of emergency response agencies interrogation of hydrometric networks training for readiness transmission and recording public sensitization and flood of hydrometric data education about floods and flood warnings preparation, flood weather radar data awareness detection severe weather forecasting & readiness and warning

receiving, integrating and processing met. and hydro. post-event debriefing data post-flood flood lessons learned warning running flood forecast forecasting models enhancements evaluation implemented forecasting flood levels and their timing for specific locations

flood flood warning warning response flood formulation interpreting forecasts & understanding likely pre-flood damage warning flood impacts reducing actions commun- issuing appropriate emergency response ication warnings agency engagement and loss mitigation updating warnings

public damage reducing warning dissemination/ actions media engagement

warning confirmation

Figure 3.1: The principal elements of a flood forecasting, warning and response system

A critical problem affecting flood warning response is flood warning lead time which is often too short to permit effective institutional and other responses. The European Flood Forecasting System (EFFS) project has been developing a European flood forecasting systems for 3 or 4-10 days in advance of an event. A European Flood Alert System (EFAS) has also been developed using ECMWF’s medium-range deterministic and ensemble forecasts, and has a spatial resolution of 5 km, and 1 km for selected river basins. Current initiatives also include extending the territorial coverage of flood forecasting and warning systems. Flood forecasting capabilities are being upgraded. Advances in meteorological, hydrological and engineering sciences are fast generating a range of further methodologies for forecasting weather and flood events, including ensemble prediction systems (EPS) and new hydrodynamic models. Each of these advances is presenting new challenges to scientists and those within the flood forecasting and warning chains, not least in working out how best to utilise and communicate the data they produce, including uncertainty data. Progress is being made in harnessing recently introduced and new information and communication technologies to communicate flood warnings in many parts of Europe. These communication techniques enable both the timeliness of, and public access to, warnings to be improved. A revolution is taking place in the public availability of flood information and flood warnings. People are increasingly able to rapidly obtain flood information from a variety of sources through personalised communication networks, although the information available may now be occasionally conflicting.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 71 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

The acid test for flood forecasting and warning systems is whether or not they lead to appropriate warning responses which save lives and reduce flood damages. Although there have been many positive experiences, unfortunately, this has not always been the case. Institutional responses to flood warnings are sometimes rendered ineffective by institutional complexity and organisational shortcomings. Individual responses may be hindered by lack of knowledge of the meaning of flood warnings, and how to respond appropriately to a flood warning.

In most parts of Europe, an engineered, constructional method of protecting communities and infrastructure against floods has been the most common approach, coupled with a technical approach towards forecasting floods based upon monitoring weather, runoff and related natural conditions. There is no doubt that this overall ‘structural’ approach to flood protection has been effective but change is taking place. Alternative and complementary approaches are required, especially those which focus upon changing human behaviour towards environmental risks. An array of non- structural flood measures exists and success is now viewed as being more likely using a ‘portfolio’ approach in which both structural and non-structural methods are adopted. Generally, flood warning systems are categorised as a non-structural, partly technical, flood management measure (Figure 3.2). The bases of flood warning systems are twofold: a) an ability to detect flood generating conditions and to forecast them in an accurate, reliable and timely manner (largely a set of technical processes and issues); and b) an ability to translate flood forecasts into warnings which can then be communicated to those at risk in a timely manner leading to appropriate behavioural responses to warnings which succeed in reducing threat to life and property (a combination of technical and behavioural processes and issues).

Dams, reservoirs, retarding basins

Channel modifications

Water control Levee banks measures Flood proofing Structural Catchment modifications measures

Schemes of drainage & Land use control flood protection measures Flood forecasting, flood warning and emergency planning

Planning controls Non-structural measures Financial relief and loss reduction Acquisition and relocation

Flood insurance

Public information and education

Figure 3.2: A perspective on the positioning of flood forecasting and warning systems within flood management measures in general Source: Penning-Rowsell and Peerbolte (1994, p6).

Flood warning systems are sometimes closely combined with structural strategies. Where there is no structural flood protection at all, flood warning systems may be used in conjunction with emergency planning and response systems as the primary flood management measure.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 72 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

The rationale for flood warning systems, which often means that they become prominent in flood management strategies, usually rest first and foremost upon a) providing public security and safety; b) preventing physical injury; and c)preventing loss of life. Reduction of economic and financial damage is also an important rationale, and it is usually considered to be an important ‘secondary’ function of flood warnings. It is this aspect of flood warnings which is the focus of this research report.

Flood warning systems are versatile in the sense that they may be employed to reduce flood risk (i.e. the probability of flooding), flood exposure (i.e. people and property located in floodplains) and flood vulnerability (i.e. the capacity of people to cope with, resist and recover from a flood (Figure 3.3).

Barriers, barrages,dams and river regulation

River channel improvements

Flood risk Dikes, levees, embankments reduction Flood protection and drainage Structural Flood abatement through forestation measures

Flood exposure Flood proofing reduction measures Flood forecasting and warning

Preparedness, planning, evacuation

Public awareness raising Non-structural measures Flood Land use regulation vulnerability reduction measures Property purchase and relocation

Flood insurance

Compensation

Figure 3.3: The versatility of structural and non-structural flood mitigation measures Source: Parker (2000, p15).

3.2 Different disciplinary perspectives and their potential contributions to understanding and improving flood warnings

Researchers and practitioners approach flood warnings systems from very different perspectives. For example, for natural scientists (e.g. meteorologists, hydrologists, engineers) the basis of flood warning systems principally lies in the technical domains of flood detection, flood forecasting and flood routing, but also increasingly in flood warning communication where the rapid pace of innovation in information and communication technologies has opened up exciting possibilities. On the other hand, for social scientists flood warnings are an opportunity to observe, understand and model human behaviour in extreme circumstances. Flood warnings viewed as a process of risk communication designed to influence of other factors (such as anxiety, or trust in authority) on behavioural response to flood warnings (Figure 3.4).

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 73 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Risk communicators (e.g. flood warning Risk communication recipients managers in a flood risk management agency) e.g. floodplain occupants or flood warning recipients

• Denial Communication •Message • Competing channels receipt priorities •Message and risk • comprehension messages

Design of risk Attitude/ Risk Flood warning communication Target Flood risk behavioural assessment message - flood warning audience Perception change (e.g level of (e.g. who, how to (e.g. flood (e.g. a defined (e.g. ranging (e.g. begin flood risk, reach with what zone map; floodplain from aware to make family consequenes of message, when and a flood population) to unaware) flood and flood flooding) how often; behaviour warning) warning plan) change desired))

Diversity or heterogeneity Experience • Iconic and • •Age • Knowledge • Uncertainty informational • Stress level • Gender of appropriate • Quality of elements • Social group • Language actions data • Behavioural advice •Trust • Special needs • Flood experience content • Heuristics • Cultural factors •Biases • Personal • Specificity benefit-cost • Anxiety level • Trust in authority

Figure 3.4: Floodplain user’s perceptions, attitudes and warning response behaviours

Figure 3.5 presents a further perspective on risk communication issues in the case of flood warnings and emphasises the importance of communication of flood risk between the different actors in the process including scientists (e.g. weather and flood forecasters), flood warners and flood defence operators, professional emergency managers and utility company managers and members of the public. The challenge is to combine these perspectives, and the insights which they bring, to design effective flood warning systems and to improve them over time. In this project we bring predominantly social science perspectives to bear upon the problems of modelling flood damage savings generated by flood warnings, and improving flood warning systems.

Figure 3.5: The complex information and data flows between actors which create major challenges within flood forecasting, warning and response systems as science and technology advances

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 74 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

3.2.1 The flood loss reducing effects of warnings

In theory a reasonably accurate, reliable and timely flood warning system is capable of reducing almost any flood loss - any of the flood losses which appear in Figure 3.6 can become a loss avoided by (or ‘benefit’ of) a flood warning. In practice, the flood losses avoided by any particular flood warning appear to be a function of a range of variables, including flood warning lead time (defined here as the time between the receipt of a flood warning and the onset of flooding at a location); and the speed and appropriateness of the response to the warning by those affected and those responsible for defending communities against floods.

Figure 3.6: The losses and damages caused by floods which may be reduced by flood warnings Source: Parker (1999, p39).

An important element of this research project is an examination of these and other variables which affect the amount of flood damages avoided once a flood warning is issued. A model, first published by the Flood Hazard Research Centre, Middlesex University in 1991 (Parker, 1991), currently exists for estimating benefits of flood warnings in terms of flood damages avoided. This model, which currently applies in the UK only (and hereafter referred to as the UK FHRC model), may be used to improve the effectiveness of flood warnings over time (Figure 3.7) and represents the starting point for the current research project.

This research report focuses almost entirely upon the benefits of flood warnings, and the warning responses and behaviours which lead to these benefits, but it should also be recognised that flood warnings have costs. To derive net benefit of flood warning systems, the costs of flood warning systems should be deducted from the benefits, but this is not part of this particular research project or report.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 75 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 3.7: Improving the effectiveness of a flood warning system over time

3.2.2 Project objectives and deliverables

The objectives and deliverables of this research project, as set out in the Research Implementation Plan (RIP) January 2006 and the Details of Work (DOW) April 2006, are as follows.

1. To collect new information and available data on the damage-reducing effects of flood warnings and on those factors affecting people’s propensity to act upon receipt of a warning, and to refine the existing UK FHRC model of the economic benefits of warnings.

2. To calibrate a simple model developed by MU/FHRC for a range of flood circumstances so that this model can become a standard method for evaluating flood damages and the effect that warnings can have on these values.

3. To produce a modified FHRC model and data on a) how people respond to flood warnings b) what actions people take to protect people and property c) what features of flood warning systems would be best to enable people to do more d) some insights into what factors encourage, limit of inhibit the ability of households and others to respond to flood warnings e) some estimates of the actual damages saved according to key variables (e.g. type of warning and warning lead time)

4. To apply the flood loss reduction model to some FLOODsite case studies to demonstrate the viability of flood warnings as a major non-structural flood mitigation option

5. To review and refine methods for collecting data on the damage reducing actions taken by households upon receipt of a warning

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 76 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

3.3 Methodological approach

3.3.1 Introduction To meet the objectives and deliverables of this study it was necessary to use a particular set complementary research strategies and methods. These were designed to overcome some of the data constraints faced in research of this nature (see 3.3.2 below) and include the following:

• the use and re-analysis of existing social survey data (both qualitative and quantitative data); • re-thinking of the existing conceptual model of flood damage reduction produced by flood warnings; and the conceptual development of this model into a broader one which is more appropriate for Europe; • data from FLOODsite partner surveys designed to gather up to date profile information and data relating to particular aspects of flood warning systems; • expert interviews in two selected countries; • secondary source data from a range of sources and the incorporation of these data into simulation models; • simulation modelling and model testing; and • case study demonstration of the application of models.

The main field data collection elements of the research involved the expert interviews in France and the Czech Republic, and the gathering of case study information from Grimma, Germany and Grenoble, France. Our research design, which incorporates the above elements, is shown in Figure 3.8.

3.3.2 Data constraints on research and application of research results With some exceptions, reliable flood damage data does not exist for most parts of Europe, and data on the socio-economic profile of floodplain populations only exists where particular surveys have been targeted to collect these data. In many cases no-one has even thought of collecting the data which our research approach might use. In some respects, therefore, Europe resembles an under-developed region in terms of the sophistication of socio-economic data which is available on flooding, and this limits the kind of research which can be undertaken to meet the research objectives of this project, and the application of models which we have produced in the course of this research.

3.3.3 Rethinking and reconceptualisation An important requirement in the objectives of this research project is to develop the UK FHRC model for a range of flood circumstances: to make the model more applicable in Europe. When we reconsidered the UK FHRC model we recognised its limitations: a) it applies more to shallow floods in which people can move valuables out of the reach of floodwaters; b) the model is not well suited to modelling damage savings in flash floods; c) a wider range of responses to flood warnings has become available and is increasingly being used and currently the model excludes; and d) the UK FHRC model does not apply to circumstances in which mobile structural flood defences (e.g. flood barriers of various kinds) work in conjunction with flood forecasting, warning and response systems, and we believe that it therefore significantly under-estimates flood damage savings. A new Flood Warning Response and Benefit Pathways (Section 3.8) is the outcome of this process of re- conceptualisation, and it provides a model which more completely captures flood warning benefits in Europe.

3.3.4 Re-analysis of existing social survey data Existing recent social survey data is re-analysed in order to a) allow the updating and refinement of the existing UK FHRC model, and b) produce and interpret evidence on how people respond to flood warnings and the protective measures which they take. Qualitative and quantitative data from a range

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 77 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

FHRC MODEL UPDATE DEVELOPMENT OF FLOOD WARNINGS

RESPONSE BENEFITS PATHWAY MODEL

FHRC Model updated for the UK situation using Development of the model to include additional further analysis survey results components into a benefits model

FLOODsite PARTNER SECONDARY RESEARCH

SURVEY

Conducted to provide initial Investigated published sources broad assessment of the flood to explore the flood warning warning schemes in Europe arrangements in Europe.

Identification of the countries France and Czech Republic for in- depth national case studies

EXPERT INTERVIEWS IN FRANCE AND THE CZECH REPUBLIC

Expert interviews conducted in both France and the Czech Republic exploring the flood warning systems and wider context of flood management strategy. Experts interviewed in different areas of responsibility and different organisations and scales of management

Identification of case studies and SECONDARY DATA Realisation that in some areas of data with which test the models Europe the severe data From a range of different constraints evident make the sources (e.g. from flood reports, application of the models simulated data etc.) impossible

MODEL TESTING GRENOBLE CASE STUDY The FHRC and FWRBPM are tested within the European context on a number of different case Local case study undertaken to studies. These case studies illustrate different examine the situation where available geographical scales and levels of data availability data is extremely limited (see Sections 7, 8 9 and 10)

REFINEMENT OF FHRC AND FWRBPM DEVELOPMENT OF A MORE QUALITATIVE APPROACH

The refinement and development of models which are Assessing the damage-saving benefits of flood applicable within the European context and guidelines warnings using a qualitative approach to overcome of their use. data shortage/gaps.

Figure 3.8: Diagram illustrating the research design

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 78 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 of flood locations in England and Wales, gathered in a UK Defra/Environment Agency sponsored research project, is extensively re-analysed and combined with analysis of the Floodsite Task 11 research projects on the vulnerability and response of communities to flooding. We have extracted and interpreted data from these Task 11 research reports which provide empirical evidence on, and insight into, flood warning response and protective behaviour.

3.3.5 Partner survey Early in the project our partners were identified as an initial source of information on aspects of flood warning systems which are rarely published and available in any secondary source. Also we wanted an up-to-date picture of flood warning systems in Europe, and we asked for experts and specialist reports to be identified. One, or sometimes several, persons from one partner research organisation in each of the thirteen countries participating in the FLOODsite project (i.e. Belgium, Czech Republic, France, Germany, Greece, Hungary, Italy, the Netherlands, Poland, Portugal, Spain, Sweden and the United Kingdom) were selected, and a partner questionnaire was sent to them with a request for a response. The number of full survey responses from the FLOODsite partners was limited to five, with such responses being received the UK, France, Germany, The Netherlands and Hungary. Our Belgian partner responded that he felt unable to provide answers to many of the questions in the questionnaire because of a lack of information and data. We interpret the lack of response by other partners partly to the same reason – that information and data are lacking. The partner survey allowed different flood warning systems and their strengths and weaknesses to be identified, and it provided sufficient information to be able to select those systems to be explored in more detail in the in-depth case studies. The partner survey proved useful in identifying key flood warning experts within each country.

Following the broad survey of countries in Europe, two countries were selected for in-depth expert to gain a depth of insight; these are France and the Czech Republic. The flood warning and flood management frameworks of these countries are presented in detail in the comprehensive milestone report (T10_07_12) which explores;

a) the current stage of development of flood forecasting and warning systems in two selected European countries – France and the Czech Republic; b) the extent to which flood damage reduction is a central rationale for the existence and development of these systems, and their extension where this is planned; c) the extent to which the French and Czechs are currently modelling flood damages, and the benefits of flood warnings; and d) the emphasis currently place in these countries on flood warning response including the measures being taken to enhance it, together with any barriers to this that might be perceived.

Although not fully developed here, these case studies provide a deeper understanding of the opportunities and barriers in these two countries to adopting the flood damage reduction or flood warning benefit estimation models and have informed the conceptualisation of the models to the wider European context.

3.3.6 The expert interview tool It became apparent early on in the project that the availability of data was limited, and therefore the availability of values for some of the components in the models was constrained, it was necessary to adopt alternative approaches to acquiring data. A comprehensive tool was developed to guide the interviews and to gain data through the expert interviewing process. This technique asked questions concerning the development, operationalisation and success of flood warning schemes and permitted detailed questions about the specific nuances of flood management strategies. Respondents were questioned about those variables that they felt were important in preventing flood damages and were able to provide specifics about how people respond during a flood event.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 79 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

It was also important within this interviewing process to establish the validity of the different models to identify the losses saved through different routes. The opinions of these key flood management experts were fundamental to the further development, refinement and application of the assessment approaches. The strengths, weaknesses, history and future plans for flood warnings were able to be investigated, alongside the wider context of the flood management system to see how improvements to flood risk management might lead to different implications for the model. The interview tool not only ensured that the background and effectiveness of the different aspects of the flood warning scheme were investigated, but also where specific data were known to be limited (for example the reliability of the warning scheme, the response of individuals to flood warnings and estimates of the numbers of people living in flood risk areas or who are protected by defences) experts could be asked to provide their judgemental estimates for these values. Participants were also asked to provide an estimate of their confidence about the judgements which they made. Even so some experts found it impossible to provide values in some cases.

3.3.7 Secondary data We used a wide range of secondary data in the research project, although as we have already noted above secondary data on socio-economic aspects of flooding and flood warning were often simply unavailable. We had hoped, for example, to find data on the number of flood prone properties in different countries and the proportion of these which are protected from flooding in some way, but such data do not exist in most cases. It is also very difficult to get data from secondary sources on expected annual flood losses at the country or region or catchment level.

3.3.8 Case Studies Within milestone report T10_07_12 a range of examples and case studies have been adopted to demonstrate the application and flexibility of the models. These include different geographical scales, different types of flooding and flood severity and the different levels and refinements of data that are available. These examples and case studies and their features are illustrated in Table 3.1.

Table 3.1: Examples and case studies used for the testing and refinement of the models and the wider approach. Example locations Geographical- Data availability Flooding type Flooding Use type scale severity FHRC MODEL (originally developed for the UK) England and Wales National-scale Data-rich Mixed season riverine Low to Average damages flooding moderate saved per residential property River Elbe, in the Large-scale Medium level of Mainly slow onset Moderate to Average annual Czech Republic catchment area data availability rainfall flooding severe damages Middle Loire, France Small-scale Medium level of Slow-onset rainfall Low to Event based catchment data availability flooding moderate approach Grimma16, Germany Small-scale Data-rich Slow-onset rainfall Moderate to Both 2002 and approach flooding severe average annual damages FLOOD WARNING RESPONSE BENEFITS AND PATHWAY MODEL England and Wales National-scale Data-rich Mixed season riverine Low to Average annual flooding moderate damages River Elbe in the Czech Large-scale Medium level of Mainly slow onset Moderate to Average annual Republic catchment area data availability rainfall flooding severe damages Middle Loire, France Small-scale Medium level of Slow-onset rainfall Low to Event based catchment data availability flooding moderate approach Grimma4, Germany Small-scale Data-rich Slow-onset rainfall Moderate to Both 2002 and approach flooding severe average annual damages QUALITATIVE ASSESSMENT OF THE LOSS-SAVING BENEFITS OF FLOOD WARNINGS Grenoble4, France Medium-scale Data-poor Mixed flooding Moderate to Qualitative area (including flash severe assessment flooding, dam burst)

16 These shaded locations are examined in more depth in T10_07-12 and form full case studies

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 80 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Not all of these examples and case studies are presented within this report. One case study Grimma in Germany is presented in Section 3.9 to highlight the application of both the modified EU FHRC model and the Flood Warning Response and Benefits model.

3.3.9 Methodological implications of data constraints Currently, the level of precision of data which is available varies considerably between and within countries. This has serious implications for the successful application of the models. For instance, unlike for the UK it was very problematic to gain estimates of figures such as estimated annual damages or the numbers of people living at risk from flooding at the national level for other European countries. Since these data have not been routinely collected or been estimated by the organisations involved in flood management, there was some reluctance for experts to estimate values such as these for the national level. However, a little more data of this kind were available at the scale of the catchment or the local and regional level. This of course has implications for the testing of the FHRC and FWRBP models and any future wider application of the models within the European context.

3.4 Flood Warning Response 3.4.1 Introduction Evidence about how people respond to flood warnings in different types of floods, and how they currently seek to protect themselves from flooding is important to how many damages can be saved. In addition, focus is turned to those factors which are likely to encourage or inhibit flood warning response, and the features of flood warning systems which are likely to best enable people to respond more effectively to flood warnings.

Evidence on how people currently respond to flood warnings is not particularly extensive. This is because reports of flooding often focus upon the conditions producing flooding, the flood damages sustained in the event and/or the performance of a flood defence system, rather than upon recording in any detail the actions taken by people when they receive a flood warning. There are, however, some important and insightful exceptions which are analysed and presented in the full milestone report. Only some of the key messages and their implications are presented here.

3.4.2 Different physical and social contexts are critical in understanding flood warning response The physical characteristics of flooding, and the social circumstances of those to whom flood warnings are issued are absolutely crucial to any understanding and interpretation of how people respond to flooding and flood warnings. In this sense ‘context is everything’ in understanding and interpreting flood warning response, and in ultimately in identifying the circumstances which are most conducive to an effective response to flood warnings. For this reason we have taken trouble to summarise what is known about the physical and social contexts of the flood warning responses referred to below (although the original sources should be consulted for a full picture).

Variables such as the severity of the flood threat (including the perception of flood risk severity), the rate of rise (or flood-to-peak interval) of floodwaters, and the time available between a flood warning being received and the onset of flooding, are usually considered to be critically important variables in understanding people’s response to a flood warning. Whether a flood is a slow onset, medium-onset or rapid-onset event can be expected to greatly affect warning response, and in the discussion below we seek to provide evidence from flood situations across the speed of onset spectrum. Also people’s experience of flooding, and experience of previous flood warnings (a surrogate of which may be the past frequency of flooding), is also very likely to have an important affect on their warning response, as is their knowledge of how to respond appropriately to a flood warning. But more fundamentally, factors such as people’s expectations and assumptions about the distribution of responsibility for responding to flooding and flood warnings, for saving damage and for compensating for flood damages, have a very important influence on their attitudes and behaviour in floods. Therefore, in

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 81 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 what follows we seek to identify, where possible and in some detail (i.e. where the information is available), the flooding circumstances and the social circumstances which relate to particular flood warning responses.

A further contextual issue which is highly pertinent to both flood warning response and questions about how people protect themselves from floods, is how people prepare for flooding or for a flood warning. Clearly, the extent to which people are prepared for a flood and a flood warning, and the kind of preparations they make, or precautionary measures they take, will influence their response to a flood warning. Therefore, in the evidence discussed below, we refer to research information on this subject.

The evidence assembled below is drawn from a wide variety of empirical surveys and research studies. The focus is evidence principally from Europe, although we draw upon research from outside of Europe where necessary. A variety of research designs have been in the research which we examine. This means that the variables, themes or questions or categories examined or used in one study are often not replicated in another study. Therefore, it is not always possible to draw together an entirely comparable or completely consistent picture of the variables underlying flood warning response from the empirical evidence which has been assembled.

3.4.3 Factors likely to encourage or inhibit flood warning response and features of flood warning systems which are likely to enable people to respond more effectively to warnings

Handmer and Ord (1986) identify the following range of variables which are likely to affect response to flood warnings: • community and situational context, including socio-political culture, preparedness and situational factors such as the time of day, of the day of the week; • disaster characteristics such as timing, environmental cues; • warning source and mode of communication, including credibility (number of false warnings), credibility of the warning source, availability and number of warning sources; • warning message delivery, including urgency in the announcement, whether or not the message is delivered personally, the number of warnings received; • message content, including specificity about location, timing and magnitude of impact); information on the consequences of ignoring warnings; details of the desirable response; use of non-technical language; unambiguity; and perceived certainty; • local context; including preparedness; disaster education; previous experience; the group context; community involvement; living alone; and the response of others; and • demographic factors, including age; gender; residency; ethnicity; income and education; and personality.

Situational factors can either inhibit or encourage levels and effectiveness of flood warning response. For example, people may find it difficult to respond to a flood warning if it is received after night-fall (i.e. when it is dark). Certainly in England farmers who have to move livestock out of the path of floods report that receiving a flood warning at dusk is problematic because they then have to move stock in the dark and conditions on floodplains can be dangerous. In Mauritius the tropical cyclone warning system has been constructed so that a warning is given out several hours in advance of nightfall rather than at nightfall to enhance response (Parker and Budgen, 1998). Floods may occur in holiday periods when a high proportion of flood-prone people may be away from their homes (availability to receive a warning therefore falls): the Easter 1998 floods in England occurred on an Easter Bank Holiday weekend when such factors became relevant. The behavioural content of warning messages is very important and influences warning response. For example, the contrast between those who knew what to do once they had received a flood warning in 2002 Danube floods

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 82 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 compared with those in the Elbe, where the flood warning messages contained no information on how to respond, is large.

Many studies have found that, where they have a high credibility, warning from government agencies stimulates a better response, but UNDRO 1984) found that this was the case only for higher socio- economic groups; those of lower socio-economic status may not be reached by hierarchical, powerful agencies (Schware and Lippoldt, 1982). Scanlon and Frizzell (1979) and Schware and Lippoldt (1982) found that ‘word of mouth’ communication between people was the most effective way of warning people and eliciting an adaptive response. Downing (1977) found that the more sources of confirmation there were for a warning the better the response, and he and UNDRO (1984) also found that a moderate to high degree of urgency stimulated adaptive responses. A considerable amount of research has gone into exploring the relative strengths of signal words that may be used in warnings (Edworthy and Adams, 1996; 31). Wolgater and Silver (1990) tested the arousal strength of 20 words (including ‘Note’, Prevent’, ‘Harmful’, ‘Warning’, ‘Urgent’, ‘Danger’, ‘Fatal’ and ‘Deadly’). The word ‘Note’ had the lowest arousal strength whereas the word ‘Deadly’ had the highest arousal strength, and in general this body of research has found that differences in signal strength of words are common and fairly robust. Handmer and Ord (1986) report that adaptive behaviour appears to be directly related to the number of warnings a person receives, with more warnings for the same event resulting in improved response.

Flood warning response will be inhibited, possibly seriously so, if flood warning communications are ineffective and delayed and untimely, and will be enhanced in conditions which are opposite to these. In the section below we make the assumption that flood warnings are communicated effectively and in a timely manner to members of at risk communities, and instead we focus upon warning response which is at the core of the purpose of this Section. It should be understood, however, that this assumption is a large one because in the past flood warning systems have frequently failed or under- performed because of insufficient warning dissemination arrangements. Also the empirical evidence above clearly demonstrates that the proportion of flooded respondents receiving a flood warning can be disappointingly low. Sometimes failure or under-performance is to do with late detection or forecasting of floods; inadequate knowledge of floodplain populations in the path of a flood and databases which do not hold accurate contact details of those in the path of a flood; delay and/or human error and/or technological failure; or problems which mean that flood warnings sent are not received, or are not received and understood by recipients (Parker and Neal, 1990; Parker, 2004). Flood warning communications are benefiting from the rapid rate of development of information and communication technologies, especially personalised means of receiving information, and in general this is enhancing the speed by which flood warnings can be communicated to those at risk (Parker and Haggett, 2001; Tapsell et al., 2004; Andryszewski et al., 2005).

3.4.4 Weighing the significance of the evidence The evidence analysed (and fully presented in T10-07-12) provides a great deal of insight into the factors which are likely to encourage or inhibit people’s response to flood warnings. It is useful that the evidence is of complementary nature, including empirical data from surveys of flood prone populations either after or in anticipation of flood events; high-level judgemental data from experts and FLOODsite partners; and data from the research literature. It is, of course, important to weigh this evidence carefully because it is of different types and because the empirical evidence is unevenly available across Europe. Currently much more evidence is available from parts of Europe with comparatively high per capita incomes and where flood warning systems may be more developed than perhaps elsewhere in Europe. It is impossible to know with any certainty, but the possibility exists that the empirical data assembled above may represent those locations and socio-economic conditions in which flood warning response is more developed and better than on average in Europe. It appears likely that flood warning response problems may be greater in some parts of Europe than indicated by the empirical evidence above, even though this evidence currently suggests limited success in flood warning response other than in saving life where the picture appears to be different.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 83 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

It is clear that in many of the flood locations from which empirical evidence has been assembled above, loss of life has largely been avoided through flood warnings of one kind or another. Yet floods continue to cause life of life in other locations in Europe. Referring to the EMDAT database maintained by the Catholic University of Leuven, Penning-Rowsell et al. (2004) observe that deaths caused in floods in Europe may be declining as a result of an expansion and improvement of flood warning systems (Parker et al., 2007).

3.4.5 The complexities and similarities of flood warning response It is clear from an examination of evidence on flood warning response and related protective behaviours that, although there is an evident degree of similarity about how people currently respond to flood warnings (or how they say they will respond to them), the factors influencing warning response and protective behaviours are complex and complexly inter-related. Apart from anything else, this suggests that it may be difficult, but by no means impossible, to create the conditions in which flood warning systems will work well – when working well includes inducing high levels of appropriate response and damage and loss saving. This kind of view prompted Handmer (2000) to ask whether flood warnings are futile. We examine the conditions which are likely to induce favourable outcomes below.

Figure 3.9 attempts to capture as much of the complexity as possible which the evidence reveals surrounds and influences (indeed determines) flood warning response. The current similarities of flood warning response, as they arise from the evidence presented above, appear to be as follows:

- a significant proportion of people are likely to do nothing to prepare for floods and flood warnings, and are likely to take no action following a flood warning; - a significant proportion of people will use personal observations and family and community networks (i.e. personal and unofficial warning systems), and may prefer these as more trustworthy than official flood warnings; - many people use a combination of personal, unofficial/community, and official warning systems when they perceive a flood threat; - staged warnings (where there are different levels of warning according to expected severity of event) appear to elicit greater response as the warning level is increased; - people have a strong need to gain confirmation of a flood warning, no matter what the source of this warning may be – in some circumstances the actions people take to confirm a flood warning may lead them into danger; - people have a higher incentive to save their lives than to save flood damage, and many will respond effectively to protect their life (e.g. by ‘running away’ or evacuating) once they assess the threat reasonably accurately; - many people seek to save possessions, especially valuables and memorabilia when they receive a flood warning; - many seek to move vehicles to flood-free ground on receipt of a flood warning; - many seek to prevent entry of floodwaters to their properties, but the inexperienced and un- knowledgeable resort to methods which are likely to be ineffective; - a minority are prepared for a flood and a flood warning with contingent flood proofing methods (e.g. flood boards, flood gates); - a minority adapt their homes to make them more resilient and resistant to flooding; - warning responses which reduce psychological trauma and stress are counted by flood victims in their definition of ‘effective’ response; - many will posses flood insurance where flood insurance is widely available; - the possession of flood insurance may be negatively associated with adoption of other warning responses (unless required by insurance companies); - those experienced in flooding will make more preparations and take more actions on receiving a flood warning than those who are inexperienced.; and - drivers may under-estimate flood risks, especially flash flood risks, and need to be incorporated into flood risk awareness raising and flood warning measures.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 84 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 3.9: The variables which determine flood warning response either by inhibiting or enabling response by individuals occupying flood prone locations

3.4.6 Factors determining flood warning response effectiveness In practice most of the variables in the centre and on the right hand side of the inner circle in Figure 5.3 affect flood warning effectiveness, but several further variables are identified which have not yet been discussed. They include the proportion of assets in properties which are moveable (see Section 3), the warning lead time (which has been discussed above), the characteristics of the properties and especially whether they have storage space on higher storeys to which assets can be moved, and the availability of assistance.

It is feasible to adapt properties so that contents and fixtures and fittings are moveable, and to make them more easily moveable. Arranging for assistance by, for example, local municipalities making a contract with local removal firms to be available to move assets in advance of a flood, and to return them after the event, can increase the amount of flood damage saved during a flood.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 85 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

3.5 The UK FHRC model for estimating the flood losses avoided by flood warnings

3.5.1 The 1991 model of damage saving generated by flood warnings Research on the damage reducing effects of flood warnings in the UK is sparse, mainly because a precondition for such research is that reasonably reliable and accurate flood damage data are available. As such damage became available from the late 1970s onwards in the UK, so research evidence began accumulating on the likely effects of flood warnings on flood damage (e.g. Penning- Rowsell et al., 1978). Parker (1991) brought together research evidence and the summary equation used by Parker (1991), developed by Green (CNS Scientific & Engineering Services, 1991), eventually became a basis for assessing the economic impact of warning arrangements for the residential sector in the UK.

FDA = PFA x R1 x PRA x PHR x PHE ...... Equation 3.1 (see Table 3.2)

Research on flood damages during the 1970s-1990s showed that the percentage of direct flood damage which could be saved following receipt of a flood warning (i.e. PFA) was limited to the total value of moveable contents and inventories of properties. Research on the performance of flood warning systems in England and Wales during the 1980 and 1990s revealed that the reliability (R1) of flood warnings is a critical factor in a) the effectiveness of response to flood warnings and b) the damage savings generated by flood warnings. In a series of research projects designed to reveal the performance of flood warnings, R1 was found to be much lower than 100% leaving much room for improvement. Green (CNS Scientific Engineering Services, 1991) argued that the availability (PRA) and ability (PHR) of residents to respond to a flood warning also reduced the potential damage saving of warnings. Those who were not at home to receive a flood warning would not subsequently act to reduce damage, and those who were physically or mentally unable or incapable of responding to a flood warning by removing damageable goods would also be unable to act. Furthermore, Green noted from research undertaken in the 1980s and 1990s that not all those receiving flood warnings knew what to do in response to a flood warning (PHE) i.e. responses were sometimes ineffective further eroding damage saving potential.

3.5.2 The Environment Agency (2003) variant The Environment Agency currently bases its national flood warning investment strategy upon the UK FHRC model explained above. However, the Agency adapted the model in various ways. First, they introduced a ‘coverage’ (C) factor since only a proportion of England and Wales is instrumented to the degree that a flood warning service can be provided, and they were interested in estimating the national benefits of investing in flood warning services including extending the service into areas currently not provided with a service. Secondly, the Agency modelled FDA slightly differently using annual average damages (AAD) and a damage reduction factor (DR). The result is Equation 3.2 (Environment Agency National Flood Warning Centre, 2003):

FDA = (AAD x DR x C) x (R2 x RA x PR x RE) ...... Equation 3.2 (see Table 3.2)

The Meteorological Office also employed the UK FHRC model in the context of severe weather warnings of pluvial floods (Meteorological Office/Posford Duvivier 2003). The model was developed mainly for households and Parker (1991) noted the limited research evidence available on non- residential property on which to base a model variant. Such data remain sparse.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 86 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 3.2: Flood warning equation parameters Element Equation 3.1 (Parker 1991) and Equation 3.2: Environment Agency NFWC Equation 3.3 (2003) variant Damage avoided FDA Estimated actual flood damage FDA Estimated actual flood damage avoided avoided owing to the flood owing to the flood warning warning Potential damage PFA Potential flood damages AAD Annual average damage to be avoided avoided (property plus road DR Damage reduction (the % amount of pre- vehicle damages avoided) flooding action that can be taken to reduce the cost of the flooding event) The warning R1 Reliability (%) of the flood C Coverage of flood warning service (the system’s warning process (i.e. the proportion/% of properties within the effectiveness proportion of the population at indicative flood plain that have been risk which is warned with offered an appropriate flood warning sufficient lead time to take service). action) R2 Service effectiveness (the proportion/% of flooded serviced properties that were sent a timely, accurate and reliable flood warning). Response PRA Proportion/% of householders RA Availability (the proportion/% of flooded availability available to respond to a services properties that received such a warning warning) Response PHR Proportion/% of householders PR Ability (proportion/% of householders capability able to respond to a warning able to understand and respond to such a warning) Response PHE Proportion/% of households RE Effective action (proportion/% of effectiveness who respond effectively. serviced properties either willing to take effective action or which have actually taken effective action following a flood warning to reduce flood damages). Equation 2.3 (see below) Combined RAS Proportion/% of householders measure of who receive a flood warning reliability (R1) message, based upon a) and PRA success in disseminating a warning and b) the availability of householders to receive it Total potential TPD The total potential monetary damages value of damages to structure and contents inventory Potential PID The potential monetary value inventory of damages to contents damages inventory items

Potential MID The potential monetary value moveable of damages to moveable inventory contents inventory items damages

3.5.3 The limitations of the UK FHRC model The model has limitations. First, it addresses damage saving achieved by individual property owner’s efforts, through moving and raising assets. The model does not cover damage reduction through householders preventing water entering their property, by effective sandbagging, pumping or closing openings. Second, the model deals only with damage reduction by individual householders, and not the potential for community response to a flood warning to reduce damages, for example with state- funded temporary flood barrier systems (Stokes and May, 2004). Clearly, the benefits here are potentially greater than those realised by individual householders moving their assets, and include the

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 87 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 avoidance of many of the intangible flooding impacts (Tapsell et al., 2002; Penning-Rowsell and Tapsell, 2002).

Secondly, the model is based upon an ‘event specific’ approach. Therefore it fails to take account of any wider benefit effects that longer-term publicity and education associated with flood warning systems has upon reducing damageability.

Finally, the model is not designed to take account of reductions in loss of life, injury and the other health impacts of flooding for which flood warnings are partly designed.

3.5.4 The latest survey evidence Research undertaken in 2004/05 by the Flood Hazard Research Centre for DEFRA and the Environment Agency allows the variables in Equation 3.1 above to be recalibrated for damage savings generated by flood warnings for households or residential properties. At the same time, a revised version of the 1991 model (see Equation 3.3 below) is postulated based upon the latest evidence. Full analysis of this survey data its implications and limitations are presented in Section 3 of the Milestone Report T10-07-12.

FDA = (TPD x PID x MID) x RAS x PHE ...... (Equation 3.3)

Where the variables are as defined in Table 3.2 TPD = Total potential damages PID = Potential inventory damage MID = Moveable inventory damage RAS = Reliability of the flood warning process combined with proportion of householders available to respond to a warning. PHR is now excluded on the grounds of the latest research findings from England and Wales but might warrant inclusion elsewhere if local social survey results indicate that ability (or disability) affects warning response (see Section 5 for a discussion of this.) PHE = Effective response – the proportion (%) of properties for which an appropriate flood warning service is provided where the occupants are either willing to take effective action or which have actually taken effective action following a flood warning to reduce flood damages

Table 3.3 provides a worked example of flood damage reduction to illustrate this using the revised model (Equation 3.3) calibrated using data from the survey research. In the 1991 model PFA is used to estimate flood damages at various flood depths for 4 flood warning lead times (2, 4, 6 and 8 hours). In the latest model (Table 3.3) damage savings are estimated for just two flood warning lead times i.e. < 8 hours and > 8 hours. This is a limitation of the latest model. Further modelling is required to reliably estimate flood warning damage reduction for a greater number of warning lead times, including very short lead times.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 88 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 3.3: Flood warning damage reduction example Item/ calculation Description % (Y) £ (X) Example Calculation factor TPD (A) Total Potential damages 100 30,000 (€43,478) PID (B) Potential Inventory damage ( as 52 15,600 BY*AX a % of TPD) (€22,609) MID (C) Moveable Inventory damage (as 41 6,396 CY*BX a % of Potential Inventory (€9,270) damage) RAS (D) Households in receipt of a 38 warning PHE Effectiveness of : (E) < 8 hour warning 55 (F) > 8 hour warning 71 Total Potential damage saved by: < 8 hour warning 4.46 1,337 AY*BY*CY*DY*EY (€1,938) > 8 hour warning 5.75 1,726 AY*BY*CY*DY*FY (€2,501)

Potential Inventory damage saved by: < 8 hour warning 8.57 1,337 CX*DY*EY (€1,938) > 8 hour warning 11.06 1,726 CX*DY*FY (€2,501) Note: <8 and >8 warning lead time data derived from sample of 110 interviews where residents could report their warning lead time.

3.5.5 Conclusions Flood warnings have become central tool of flood risk management strategy in Britain, as elsewhere. In general it is hoped that flood warning systems can contribute to making the occupation of floodplains more viable and sustainable. However, our research indicates some of the complexity and difficulty involved in significantly improving levels of warning response, thereby gaining greater economic benefits from warnings. In this regard the research has wider potential relevance for hazard and disaster managers seeking to improve warning response through engaging the public and seeking to influence their hazard response behaviours.

Important steps forward have been made by the Environment Agency in improving flood warning coverage and reliability, direct communication of flood warnings to those at risk, and in making flood risk, flood warning and warning response information more widely available, but the positive impacts of these advances appears slow to take effect. It could be that a time lag is involved before the relatively recent enhancements to flood warning systems fully impact in terms of improved warning response rates and higher damage savings. Currently, however, the value of flood warnings in terms of damages saved is modest, and lower than indicated by previous applications. This research demonstrates that the extent of flood warning damage savings by moving and raising household assets is limited by the value of what can be moved. Only 21% of the total flood damage potential for UK residences is avoidable by householder’s damage-saving responses to flood warnings, and only about a quarter (i.e. 24%) of this damage is currently avoided by their warning responses.

The model of economic benefits of flood warnings developed in 1991 represents a step forward in allowing flood warning benefits to be estimated for residences. However, the 1991 model now

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 89 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 requires simplification. The simpler and more transparent approach to estimating total potential damages (PFA) reported in this paper and contained in Equation 3.3 should now be used. The ability to respond parameter (i.e. PHR) should be removed from Equations 3.1 and 3.2: the survey results indicate that those who are disabled, who are in ill-health or who are elderly save as much damage proportionately as others. Secondly, downwardly revised proportions for other equation parameters should be used.

Many householders remain unwarned because of the limitations created by warning system reliability and householder availability problems. Those in higher social grades are more likely to receive warnings than those in lower social grades, and these factors appear to be important contributory factors to modest damage savings. The economic benefits model assumes that flood warning lead time is a significant factor in avoiding flood damages, and the survey results support this assumption. Having prior experience of flooding, receiving more informative warning messages, receiving help from outside of the home, and being connected to an AVM system are all factors associated with higher damage savings. Although the majority of householders whose homes are flooded take action to save assets, these actions are often ineffective.

The fact that our regression analyses only explain 12-13% of the variance in warning response behaviour indicates that other factors require further consideration. Findings from the socio- psychological and behavioural research are likely to be highly relevant here, and are explored elsewhere (e.g. Parker et al., 2007). Personality, personal experience of hazards and individual risk perception are areas that merit further research. Factors such as trust in institutions, and the level of community cohesion and institutional and social organisation that were given little attention in the study – and are difficult to quantify – also require further investigation in order to identify what motivates people to respond or not to respond to flood warnings.

Apart from harnessing lessons from socio-psychological and behavioural research, the research provides some important policy pointers. Improving the reliability of the flood warning process, and the proportion of householders available to respond to a warning, is crucial. By using mobile telephones to communicate flood warnings, rather than relying on landlines, the Environment Agency has been working to improve the availability factor. As in Britain, attention is needed to providing those in lower social grades, and those with special needs, with warnings. The Agency’s new Flood Warning Direct service is designed to be more accessible to such groups and should have a positive impact on warning response in future. Anything which can increase warning lead time, such as precipitation forecasting is very important, but there are difficulties in reliably forecasting floods to give, say, more than 8 hours warning lead time in Britain where floods are mostly rapid-onset events by global standards. Rapid, accessible and personalised information communication technology applied to flood warnings is also very important and being harnessed. Flood warning effectiveness can be enhanced to some extent by employing education as a substitute for flood experience. Therefore, publicity and information programmes about flood risk and flood warnings will be of continuing importance, aimed particularly at improving the effectiveness of householder’s damage-saving responses. Lastly, organising outside help for households to move their valuable belongings when a flood warning is issued is likely to increase warning benefits.

The following Section considers the application and re-conceptualisation of this approach to the rest of Europe.

3.6 European variants of the UK FHRC model

3.6.1 Re-conceptualising the UK FHRC model for continental Europe This section seeks to reconsider and re-conceptualise the UK FHRC and its applicability to other locations in mainland Europe. The UK FHRC model designed by Green (CNS Scientific and Engineering Services) and elaborated by Parker (1991) has been revisited and revised above following

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 90 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

new evidence about how people and organisations respond to flooding in England and Wales. This revised model (Equation 3.3) is presented above.

One particular refinement to the UK FHRC model from previous versions has been to simplify and remove the ability to respond variable PHR. This eliminates the differentiation between those households that contain people who might not be able to take effective action on the grounds of ability (such as disabled or the long term unwell) and those that do not. The latest research (Parker et al., 2007) indicates that there is no significant difference between the amounts that they are able to save. Whether or not to include this variable in a re-conceptualised European version of the model is difficult decide because of apparently contradictory empirical evidence. There may be some circumstances (e.g. areas with much higher than average proportions of people from vulnerable groups, or where post-flood social survey evidence indicates) where the ability to respond component may be relevant in affecting the amount of flood damage saved. However, in most circumstances where there is a population with diverse socio-economic and health characteristics it appears probable that the ability to respond variable is not influential as far as damage savings are concerned. This is because vulnerable people are likely to receive assistance.

The second, major revision of the existing UK FHRC model reported is the combination of the PRA (the proportion of households able to respond to a warning) and the R (the reliability of a flood warning) in a new RAS factor. Social survey responses to the question ‘Did you receive a flood warning?’ are affected by a combination of these two factors, and in order to use social survey evidence of this nature it is appropriate to use the combined RAS factor. This also recognises increased success in disseminating a flood warning (meaning that a person is not necessarily required to be at home) and increased information collected by the Environment Agency in England and Wales about the numbers of people who received a warning. Whether the RAS component is relevant and suitable within an EU variant of the model again depends on what data are available and how they are collected. For example, social survey data collected in, say Germany, yielding the same responses as social survey data acquired in the UK would mean using RAS. However, if there are separate measures of warning effectiveness and householder availability the factor may be separated into its component parts i.e. PRA (availability) and R1 (reliability) (see Table 3.1)

An initial re-conceptualisation of the UK FHRC model for use in Europe would need to include both of the two UK FHRC (Equations 3.1 and 3.3) variants;

FDA = (TPD x PID x MID) x RAS x PHE ...... (Equations 3.4) FDA = (TPD x PID x MID) x (PRA x R1) x PHE

Where the variables are as defined in Table 3.2 TPD = Total potential damages PID = Potential inventory damage MID = Moveable inventory damage RAS = Reliability of the flood warning process combined with proportion of householders available to respond to a warning PAR = Proportion of householders available to respond to a warning R 1 = Reliability % of the flood warning process (i.e. the proportion of the population at risk which is warned with a sufficient lead time to take action PHE = Effective response – the proportion (%) of properties for which an appropriate flood warning service is provided, where the occupants are either willing to take effective action or which have actually taken effective action PHR may be excluded if there is no information proving otherwise. It is recommended however that the population characteristics of the area of interest are considered before applying the model.

Evidence from the analysis of flood warning responses and from the expert interviews undertaken in both the Czech Republic and France suggests, clearly indicate that there is further important component impacting upon the amount of loss savings possible from flood warnings which is not taken into account in the UK FHRC model. This is the evacuation of people from homes and other properties. The UK FHRC model was developed from a large number of post-flood social surveys in

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 91 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 which the mean depth of flooding is low or shallow. Although deeper floods occur in the UK, cases of them were not included in the data sets used to develop the UK FHRC model and evacuation of properties was minimal. Many parts of Europe experience much deeper flooding than considered in the development of the UK FHRC model, and there are also cases such as the Mulde floods affecting Grimma in southern Germany (see Section 10) where floodwaters velocities and flood depth combined mean that large-scale evacuation is vital to save lives. In areas at risk from these more severe floods, evacuation of people is more common and is often large-scale. In some circumstances mass evacuation of people will occur. Van Duin and Bezuyen (2000) describe how 250,000 people in the Netherlands evacuated because of a fear of dyke breaks during 1995 flood event. Mass evacuation is a relatively rare event, but smaller scale evacuations occur more commonly.

The numbers of people evacuated prior to the flooding and the timing of this evacuation (i.e. the amount of time between people being aware or officially warned that a flood is coming and the time they evacuate or are evacuated) Both of these factors directly effect the time and manpower available to undertake actions to move and safeguard contents of properties. In turn this will affect the amount of damages saved. Evacuation is likely to affect both the availability component (RAS) and/or the effectiveness of component (PHE) of the model. To take this new set of factors into account, and to make the model more applicable to the European scene (as well as to cases of deeper and more severe flooding in the UK, we have developed an alternative equation (henceforth known as the EU FHRC model) (Equation 3.5).

FDA = (TPD x PID x MID) x RAS x PHE x EVAC ...... (Equations 3.5) FDA = (TPD x PID x MID) x (PRA x R) x PHE x EVAC

Where the variables are as defined in Table 3.2 TPD = Total potential damages PID = Potential inventory damage MID = Moveable inventory damage RAS = Reliability of the flood warning process combined with proportion of householders available to respond to a warning PHE = Effective response – the proportion (%) of properties for which an appropriate flood warning service is provided, where the occupants are either willing to take effective action or which have actually taken effective action EVAC = proportion of properties whose residents are evacuated and who therefore were not able to save property contents PHR is may be excluded if there is no information proving otherwise. It is recommended however that the population characteristics of the area of interest are considered before applying the model.

Although the introduction of the EVAC component adds an important new dimension to the model, and makes it more widely applicable in Europe, it must be recognised that it remains a simplified version or reality. Evacuation may have a much more complicated impact upon the amount of loss savings presented. There are a number of different scenarios with regards to evacuation and the affects these have upon residents’ ability to prevent damages to their property contents. These scenarios (illustrated in Figure 3.10) assumes that a flood warning has been provided and that the percentages (i.e. TPD, PID and MID) of the losses that can be saved are equal.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 92 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

No evacuation Evacuation No time to move contents No time to move contents

No damage savings and where No damage savings and where severe severe flooding is expected, human flooding is expected evacuation should losses might be experienced limit human losses

No evacuation Evacuation Time to move some or all contents Time to move some or all contents

This is the ideal situation. Losses As well as being dependent on RAS and PHE, the saved will be related directly to RAS loss savings will also be related to the proportion and PHE of the population evacuated and the time available to move contents prior to the evacuation

Figure 3.10: Potential evacuation scenarios affecting people’s ability to save losses due to flood warning, including damages to property contents

The scenarios in Figure 3.10 are not mutually exclusive. In most evacuation scenarios flooding characteristics (i.e. depth, velocity, debris load etc.) varies spatially and temporally. Average figures for say, the time available for moving damageable contents before evacuation becomes advisable or necessary, may be used. However, particularly where large areas are at risk from flooding, and the floodplain is relatively extensive and capable of producing varied flood depths etc., the model may be applied with a greater differentiation in the EVAC component.

3.6.2 Application of the EU FHRC model One such application is when the model is used in a segmented manner, particularly when the model is applied to calculate the potential loss savings from a specific flood event (e.g. for the 1 in 150 year flood event). For instance, consider the following hypothetical scenario.

Following a general flood warning of 8 hours it is noted that 70% of people received a warning. Of those 70% that received a warning 50% of the at risk population need to be evacuated during flooding; 10% of those in the most at risk region are evacuated almost immediately, 20% are asked to evacuate after 2 hours and the final 30% are asked to evacuate after 6 hours.

Since 70% of the population received a flood warning; The RAS component is equal to 70%. Similarly, the potential damages that could be saved component of the model remains the same (TPD x PID x MID).

The main factor that needs to be reflected in the model is to PHE. In the above scenario 50% of residents are warned and have the maximum of eight hours before flooding to take action to move the contents of their properties out of the reach of floodwaters. Assuming that these residents are safe to remain and remove their moveable property, they would be able to save a significant amount of damage. This assumption is confirmed by Thieken et al. (2007) who observed that during the 2002 flood, those who had more time were able to save higher proportion of the losses. Those residents evacuated immediately following a warning will not be able to save any damages. Those who have a reduced warning time due to the necessary evacuation will therefore have reduced damage savings. In our model these different levels of effectiveness are labelled as follows:

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 93 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

• for those not evaluated PHEmax, (50%) • residents immediately PHE0 (10%) • residents evacuated after 2 hours PHEmin (20%) • residents evacuated after 6 hours PHEmed (30%)

Using these as measures as guidance the following formula may be applied as follows (Equation 3.6) whereby the evacuation component (EVAC) is replaced by segmenting the effective response component).

Evacuation influence

FDA = (TPD x PID x MID) x RAS x {(PHEmax x 50%)+(PHE0 x 10%)+(PHEmin x 20%)+(PHEmed x 30%)}

...... (Equation 3.6) Where the variables are as defined in Table 3.2 TPD = Total potential damages PID = Potential inventory damage MID = Moveable inventory damage RAS = Reliability of the flood warning process combined with proportion of householders available to respond to a warning PHE = Effective response – the proportion (%) of properties for which an appropriate flood warning service is provided, where the occupants are either willing to take effective action or which have actually taken effective action

The potential difficulty with applying the model in this way is having sufficient knowledge about how evacuation will occur. For example, a judgement about the proportion of the population that are evacuated in different flood circumstances will need to be made. This usually requires quite detailed experience of the flood hazard. However, many areas where loss of life is threatened will have detailed evacuation procedures. These procedures will normally be based upon an assessment of the risk to people from flooding and an expectation of the numbers of people that will need to be evacuated in given circumstances.

Estimating the effectiveness of the loss saving action undertaken within each of the different scenarios may present problems. Some estimates may be made by using evidence from previous flood experiences. Evidence about how much people were able to save floods with different flood warning lead times may be available, and these data may provide a guide to the effectiveness of action with 2 hours, 4 hours, 6 hours etc. lead time. These figures might not be directly applicable in the situation whereby people know from the outset that they will need to evacuate; as opposed to an evacuation order that is given following the worsening of a flood situation. Understandably when evacuation is expected , residents may focus upon gathering the belongings which they will need if they are out of their home for a long period (e.g. clothes, documents, medicines) rather than spending the time saving damages. If a longer period of time is available a mixture of loss saving activities and the gathering of essential items is likely to occur. Other circumstances that reduce the effectiveness of damage saving include whether people need to evacuate just themselves and their families. A family may include a number of dependents (e.g. elderly family members, children, or other relatives) that they need to assist, thereby reducing the time available for damage reducing action.

The amount of damages that can be saved in these scenarios is something that needs to be researched further in order to refine the application of the model. Some of these issues are explored further when the model is applied to the case studies.

An associated issue relates to situations in mainland Europe where flooding experienced is much more severe and deeper than is typically the case in the UK. There are occasions when flooding is so deep that it reaches into the second storey of properties. These higher storeys are traditionally

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 94 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 employed to store contents that that have been moved from lower storeys to safeguard them from the floods. The 2002 floods on the River Elbe, in places such as Grimma, were as deep as this. In such cases damage savings can be expected to be reduced. In addition, in situations where buildings collapse into the water during flooding, contents which have been moved are likely to be lost. These circumstances and how they impact upon loss savings is investigated further in Section 3.9.7.6 when considering the case study of Grimma.

3.6.3 Data availability and proxies Data are scarce in some regions and concerning some aspects of flood warning and response. Ideally, empirical evidence is used to calibrate the model, either through the observations and opinions of flood warning operators or from post event surveys. In areas whereby either flooding has not occurred or a new flood warning system has been developed and is untested, other data sources (such as simulated data) or proxies for these data may well be required.

Prior to the recalibration of the UK FHRC model (Section 3.5) census data was used to provide values for some of the components of the model. Table 3.4 provides some examples of data which may be used to calibrate the EU FHRC model. Both the EUROSTAT and ESDS INTERNATIONAL datasets have country and some regional-level statistics which provide data on aspects including health, population and employment. All of these features may provide useful proxies. More specific country or regional-level statistics or Census data may also be available to provide a much more reliable estimation of some of the data proxies.

These data sources are not specific to the ‘at-risk’ population however, and will only be able to provide indicative estimates of the numbers of people likely to be available to receive a flood warning or those able to respond (if this is deemed to be relevant).

The effectiveness of the damage saving activity is one component that is much more difficult to estimate, and as it is very specific to the flood warning system. Some information may be available from previous flood events, where damages have been estimated. In addition anecdotal evidence of activity may be available, or it may be possible to use the expert judgements of flood managers. Some information regarding incentives for action (such as the percentage of residents with flood insurance) might also provide indications about the likelihood of people taking damage reducing action following a flood warning.

3.7 Strengths and shortcomings of the existing UK/EU FHRC model The principal strength of the existing UK FHRC model is that it provides a working method for estimating the flood damages avoided by flood warnings in a) households and b) arising through the moveable contents or inventories of houses being moved out of the reach of flood warnings. Calibration for England and Wales is also detailed and based upon large data sets derived from post- flood surveys of flooded households. To date, the flood defence agency for England and Wales, the Environment Agency (EA), has based its flood warning strategy for 2003/4 to 2012/13 on this model (Environment Agency National Flood Warning Centre, 2003). This involves investing £250 million over 10 years.

Some of the shortcomings of the existing UK FHRC model of flood damage reduction generated by flood warnings are discussed in 3.5 above. The principal limitation is that the model assumes that the only damage-reducing responses to flood warnings are in the form of the contents of properties being moved to higher levels or being evacuated from properties by their owners. The effects of flood fighting or contingent flood proofing, such as sandbagging or closing openings to prevent floodwater ingress which may be triggered by a flood warning, are not taken into account. Such ‘resilience measures’ have become more common in parts of Europe since the early 1990s and many more flood resilience products are available on the market than formerly.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 95

Table 3.4: Potential data sources and proxies for the components within the EU FHRC model

Variables within the model Potential data sources/proxies PFA Potential flood damage avoided – the total damage avoidable if all moveable terms of See below contents of a property are successfully moved out of the reach of floodwaters FDA = (TPD x PID x MID)

TPD Total potential damages (assuming both the property and contents are damaged, but not necessarily totally destroyed) Actual or simulated damage data PID Potential inventory damages (the proportion of TPD which is potential inventory (i.e.contents) damage MID Moveable inventory damages (the proportion of PID which is moveable inventory (i.e.contents) damage R Reliability or effectiveness of the flood warning service – the proportion (%) of Information on this nature may come from a range of sources. There may be post-flood event evidence flooded properties for which an appropriate flood warning service is provided, that was highlighting the number of people who received a warning during a flood. sent a timely, accurate and reliable flood warning If this is not available it is necessary to consider the percentage of the population who are covered by a particular flood warning scheme. In addition, information about the reliability of a flood warning system may be provided through consultation with the authorities providing and/or disseminating that warning. PRA Availability – the proportion (%) of flooded properties for which an appropriate flood These statistics are attempting to discover who is at home to receive a warning. Useful categories might warning service is provided, that actually received a flood warning. include: • population in jobless households (e.g. either unemployed or retired) (from EUROSTAT or other national statistics) As stated above this category is important if the flood warning system requires people • population that work from home (UK Flexibility statistics and similar European statistics ) to be at home in order to receive a warning • % of the country with internet access (EUROSTAT)

PHR Ability – the proportion (%) of property occupants able to understand and respond to a Those unable to move contents/undertake evasive action or do not understand the warning message flood warning) EUROSTAT statistics or other national statistics • People having a long-standing illness or health problem • Activity restriction in the past 6 months • Age (% of the population over the age of 70)

Other national or regional statistics might include • % of those from an ethnic minority with language difficulties • % registered blind or partially-sighted PHE Effective response – the proportion (%) of properties for which an appropriate flood • % of those who had experienced a flood warning at their current address warning service is provided, where the occupants are either willing to take effective • % of those who knew they were at flood risk action or which have actually taken effective action following a flood warning to • % of those who have received any flood warning/awareness education reduce flood damages • % of those households who do not have any recovery mechanism to recover from flooding EVAC The percentage of people evacuated (with more detail if possible about the timing of The main sources for this are likely to be evacuation) • flood event surveys • Evacuation plans • Expert judgement from flood managers

Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Also in England and Wales the relationship between those living in flood risk areas and flood risk management organisations has changed over the past decade – a change which is mirrored in a number of other European countries. In England and Wales this change began when in 1996 the Environment Agency, which is the flood defence and forecasting agency, was also given the lead responsibility for disseminating flood warnings. This change was subsequently reinforced by extensive flooding in 1998 which led to a major re-evaluation of the flood forecasting and warning service. Information previously hidden from public view, such as detailed flood risk maps, is now widely available on the internet. Through this, and in other ways, those living in flood risk areas are being encouraged to engage in managing flood risk and to adopt both temporary and permanent resilience measures, as well as to become more familiar with flood warning systems and possible damage-reducing responses to them. Secondly, when the model was first conceptualised in 1991 temporary demountable flood defences and related community based measures were not used in the UK or elsewhere in Europe and were not considered in the model. However, since 1991 such measures have also become more common and often rely upon flood warnings being available giving sufficient lead time for these measures to be implemented.

A critique of the existing UK FHRC model was published by the FHRC in 2005 (Tunstall et al., 2005) in which the need for a broader approach to capturing the benefits of flood warning is set out. Recent scoping research undertaken by the Scotland and Northern Ireland Forum for Environmental Research (SNIFFER, 2006) used this critique recognising that the potential for capturing benefits from flood warnings is greater than suggested in the methodology originally developed in 1991. The SNIFFER scoping study17 recommends a more holistic approach to modelling the benefits of flood warnings, and one which takes into account the potential of a) operational and b) resilience measures – both identified as in need of further attention as by Tunstall et al. (2005) – as well as c) the benefits of moving possessions. This scoping report identifies the measures contained in Table 3.5 as those which may potentially lead to benefits of flood warnings (i.e. flood damages avoided) as a result of a timely flood warning being issued, received and acted upon. Given sufficient flood warning lead time, maintenance of the efficiency of watercourses as flood conveyance channels can be effective in keeping flood levels to the minimum. Other operational responses, such as maintaining the integrity of flood defences can also be potentially very effective in reducing flood levels and will lead to lower flood damages than would otherwise be the case.

Table 3.5: Three categories of potential responses to flood warnings which have the potential to generate benefits of flood warnings (i.e. flood damages avoided) (SNIFFER, 2006; p30) Flood warning Examples response Moving possessions Moving contents of houses Moving livestock Moving finished goods in manufacturing premises Operational responses Remove blockages from watercourses Clear debris screens Weed and tree clearance from channels Emergency repair of failing flood defences Making breaches in secondary flood banks and informal defences to lower flood levels Implementation of Temporary flood proofing or resilience measures for individual properties** resilience measures* Demountable/temporary defences for individual properties** Pumping out water from basements * only temporary or contingent resilience measures contribute to reducing flood damages by flood warning and these measures must be distinguished from permanent resilience measures ** a distinction is made between ‘kite-marked’ flood resilience measures and products, and those which have not met a standards-test (see for example: www.abi.org.uk/Display/File/Child/553/Flood_Resilient_Homes.pdf)

17 The scoping study is currently being followed up by a further research project sponsored by SNIFFER entitled ‘Proposed methodology for assessing the benefits of flood warning’ which is due to report in January 2008. T10-07-13_Deliverable_D10_1_ report_V_1_2_P10.doc 29th February 2008 97

Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

In the research sponsored by SNIFFER (2006) the potential scale of the three flood warning benefit components was displayed diagrammatically (Figure 3.11), suggesting that the moving of possessions is likely to lead to much smaller benefits than operational responses and resilience measures (with resilience measures being identified as potentially providing by far the largest benefits).

Figure 3.11: Potential scale of flood warning benefit components Source: SNIFFER (2006, p85)

Taking into account developments since 1991, we have reconceptualised flood warning responses and benefits to generate a new, improved model called the ‘Flood Warning Response and Benefit Pathways Model’ (FWRBP).

3.8 Modelling the wider response to flood warnings: The Flood Warning Response and Benefits Pathways Model (FWRBP model)

We have developed a complementary approach to modelling the economic flood loss savings which flood warnings may generate. This approach seeks to address some of the shortcomings and limitations of the existing UK based model explained above, and takes further the thinking developed by Tunstall et al. (2005). The model developed in this Section is a much expanded (i.e. broader) model compared to the existing models discussed although it can incorporate these as one element.

Taking the kind of holistic perspective recommended in SNIFFER (2006) eight principal response pathways to flood warnings can be identified (Figure 3.12). These eight pathways represent the theoretical range of choice of damage and loss reducing responses to flood warnings. Each is potentially capable of reducing flood damages and producing flood warning benefits, although in many parts of Europe this potential is currently far from fully exploited for various reasons which we explore below. Examples of these potential responses are given in Table 3.6.

T10-07-13_Deliverable_D10_1_ report_V_1_2_P10.doc 29th February 2008 98

Monitoring, Flood Flood defence agency Flood Flood Public dissemination devices mobilises its resources detection, (e.g. internet, media) warning and forecasting warning

Floodplain occupants and Flood defences users are operationalised

e.g. flood barriers Official emergency response are closed, flood storage (e.g. fire service, police, local authorities, military, mayors etc.) areas operationalised

BUSINESS AND UTILITY USERS Householders Finance Hospitals/ Public Retail Transport institutions health care Building Outlet network and offices facilities owners managers managers

Farm Manufacturing Public Logistics Transport Integrity of flood enterprises companies Utility Sector Network defences is monitored Watercourse Evacuation of manager manager users And maintained capacity areas at risk, maintenance before and during flooding

Operation of Move house Activate formal or ad hoc business community- Search contents and/or Activate contingent flood proofing continuity plans e.g. reschedule based and rescue evacuate measures to buildings and other assets business to minimise loss measures belongings

1 2 3 4 5 6 7 8 FDO CBO WCM SAR EVAC CME CFP BCP

Flood damages Flood damages Flood damages Flood damages Flood damages Flood damages avoided avoided avoided avoided avoided avoided

If businesses losses are avoided, and job Measures will reduce water levels If damages are avoided human If damages are avoided human security guaranteed avoided human and therefore also lower the losses through reduction in losses through reduction in losses through reduction in stress/anxiety threat to people stress/anxiety may be observed stress/anxiety may be observed may be observed

Human Human Human Human Human losses losses losses losses avoided losses avoided avoided avoided avoided

Figure 3.12: Flood Warning Response and Benefit Pathways (FWRBP) model

Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 3.6: Eight pathways of possible responses to flood warnings which have the potential to generate benefits of flood warnings (i.e. flood damages or human losses avoided)

Flood warning response Example Flood defence operationalisation (FDO) Closure of a flood barrier Diversion of flood flows into a flood diversion channel Opening of flood detention of flood storage areas Use of flood storage capacity in flood dam River regulation Emergency repair of failing flood defences Making breaches in secondary flood banks and informal defences to lower flood levels Community-based options (CBO) Demountable flood defences provided for a community, neighbourhood or road Community pumping schemes Watercourse maintenance (WCM) Remove blockages from watercourses Clear debris screens Weed and tree clearance from channels Search and rescue (SAR) Rescue of people from flooded properties or areas Evacuation (EVAC) Pre-flood evacuation of people from flood- prone properties and areas Contents moved or evacuated (CME) Moving possessions within properties to a higher level, or moving possessions to another location Contingent flood proofing (CFP) Use of property temporary resilience measures Business continuity planning (BCP) Deployment of business continuity plans to reduce direct and indirect flood damages to businesses

From Figure 3.12 and Table 3.6 we can see that the movement of contents of properties (either their movement upstairs or their evacuation from the property) (CME in Figure 3.12) is only one of eight possible flood warning response and benefit ‘pathways’. This means that the existing UK/EU FHRC model is only likely to estimate a small proportion of total potential flood warning benefits.

Structural flood defences are sometimes used in conjunction with flood forecasting and warning systems. A major, large-scale example is the Thames tidal flood exclusion barrier which protects London from tidal flooding, and which has to be closed once a flood forecast and flood warning is received by the manager of the barrier. In fact, the river Thames tidal flood defence system as a whole relies upon the Thames barrier and several smaller barriers, all of which must be closed once a flood is forecast. In addition, the flood walls and embankments downstream of the Thames barrier contain over twenty openings (used for access to the river) which also have to be closed to make the defences ‘watertight’, and the integrity of these defences also depends upon a flood forecast and receipt of a flood warning by the floodgate operators. Similar structural flood defence systems exist in other parts of the UK and elsewhere in Europe, such as in The Netherlands. The flood damages saved by these systems are attributable to the combined effect of the structures and flood forecasting and warning systems, and indeed these kinds of structural defences depend entirely upon a fully functioning flood forecasting and warning system.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 100 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

In practice there are many other ways in which the operationalisation of structural flood defences is dependent or partly dependent upon flood forecasting and warning systems. Some of Europe’s major rivers are heavily regulated and the process of regulation which delivers different degrees of flood protection makes use of flood forecasting and warning data. An example is the upper Danube in Germany and Austria where there are numerous dams and man-made barriers, and downstream of Bratislava where hydropower plants diverts large quantities of water into a side channel and reservoir.

Community-based options are taken here to mean temporary demountable flood defences. These temporary defences are erected or positioned in the days and hours prior to a flood, and depend entirely upon a reliable flood forecast and warning for their deployment. They have become widely used in the River Severn valley in England and in other locations in Europe, including a high-profile scheme in Prague, Czech Republic.

Watercourse maintenance aims at maintaining the efficiency of channels to carry river and flood waters. In many parts of Europe watercourse maintenance is routinely undertaken, but may also be activated by a flood forecast and warning in the period just before a flood is anticipated in order to keep flood levels to a minimum thereby potentially reducing the extent of the area flooded and the depths of flooding.

Search and rescue and evacuation measures are common responses to flood warnings, and are aimed primarily at saving life and other adverse effects of flooding on human beings. They do not usually contribute much to financial or economic damage saving, and may lead to such savings being low in particular circumstances, e.g. where flood warning lead times are short. They are included on the FWRBP model because they can have an impact on lowering financial and economic damage saving.

The movement of possessions to higher levels or to locations beyond the floodplain is a time- honoured response to flood warnings, and is a well developed response pathway in the United Kingdom particularly in locations where there is reasonable flood warning lead time and the flooding conditions expected do not present a major threat to people in terms of loss of life because the floods are shallow.

Contingent flood proofing is a particular type of flood proofing originally identified by Sheaffer (1960, 1967) in the USA. It has taken about 40 years to fully recognise the significance of these flood proofing measures in Europe. Sheaffer distinguished three types of flood proofing: permanent, contingent and flood-fighting. Today in Europe these measures are usually referred to as flood resilience measures. Contingent flood proofing measures are flood resilience measures which are ‘contingent upon’ (i.e. they depend upon) a flood warning being received, after which they are deployed. They are planned measures put into operation in advance of a flood to avoid or reduce flood damage. They are distinguished from flood fighting measures which are last-minute, unplanned measures which are often used by those about to be flooded (these measures typically include the use of sandbags).

3.8.1.1 Modelling the net benefits of flood damage reducing responses to flood warnings Each of the eight flood warning response pathways in Figure 3.12 is capable of generating flood damage savings, but these benefits must be set against the costs associated with these responses to gain a measure of the net benefit of flood warnings. This net benefit analysis is set out in Figure 3.13 which identifies some of the principal costs associated with flood warning response.

This research project focuses mainly upon the flood damages and losses saved by flood warnings, and in the analyses which follow the costs of flood warning and related flood defence systems are not explored further. However, it should be noted that costs need to be deducted from benefits to derive net benefits of flood warnings.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 101

Warning Response Warning benefit Response cost Net monetary benefit/cost

Cost benefit analysis should ensure that the Flood defence Damages to properties and/or businesses Capital cost of scheme and maintenance scheme has a net benefit- although this is over the operationalisation will be reduced in the area protected. (in many countries work undertaken lifetime of the scheme. Benefits are to those (FDO) Potential wider benefit to local economy using taxpayers funds) protected and potentially to the wider society

Cost benefit analysis should ensure that the Community-based Damages should be reduced to the area Capital cost, maintenance of schemes, scheme has a net benefit- although this is over the options that is protected operationalisation (e.g. labour) and lifetime of the scheme. Benefits are to those (CBO) maintenance costs. Usually undertaken using protected and potentially to the wider society national or local taxpayers funds

Should have a net positive benefit, though it is Watercourse difficult to quantify the proportion of the Benefits are difficult to quantify – Implementation costs (e.g. equipment maintenance damages saved by this action, especially in a though should lower or prevent flooding and labour) – maybe undertaken by (WCM) large flood – the benefits may be to public or downstream of clearance. public or private individuals private organisations

Search and Rescue Cost of undertaking the search and (SAR) Human losses rescue operation most often paid for If human life is not given a monetary value there is likely to be a net cost to this action- although avoided from taxpayers funds. Flood there are obviously wider societal benefits.

Warning Cost of arranging and undertaking the evacuation also potential costs in Evacuation accommodating the evacuated. Main costs If human life is not given a monetary value there (EVAC) Human losses likely to be funded by local or national is likely to be a net cost to this action- although avoided taxpayers, but maybe some cost to individuals there are obviously wider societal benefits. evacuated (e.g. for accommodation and/or return)

Contents moved or Those items moved will potentially be evacuated saved from flood waters depending on May be minor costs to the individual The net benefits will be to the individual (CME) the water depth

There are likely to be net benefits, but these Contingent flood Cost of implementing measures and might only be evident following a number of proofing Prevented or reduced damage to potentially some operating costs. These costs events whereby the first savings cancel out the (CFP) property or business affected are likely to be paid either by the individual capital and implementation costs of measures (or potentially their insurer through retrofitting) or the business involved.

Direct damage and business interruption Costs of developing the plan, savings to the business – but also Business continuity Depending on the action – reduction in planning implementation, update and potentially potentially wider local economy benefits if damage and business disruption costs (BCP) equipment or storage if stock moved. business remains healthy and an employer.

Figure 3.13: Flood warning responses, benefits and costs

Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

3.8.1.2 Further development of the FWRBP model The analysis represented by Figure 3.12 leads to the following equations which may be applied at a range of scales or levels of resolution. The equations may, for example, be applied at the national level to assess the benefits of flood warning for national investment decision-making purposes, or at a community level. Finer grained quantitative data are generally required within the equations as the level of resolution is increased.

Firstly, we need to distinguish between those damages saved by flood warning systems working in combination with structural flood defences, from those which arise in other ways.

TEDS = EDS1 + EDS2 ...... Equation 3.7

where:

TEDS = total economic damage saving generated by flood warnings EDS1 = economic damage saving generated by flood warning systems working in combination with large scale structural flood defence systems EDS2 = economic damage saving generated by flood warning systems working without the support of large scale structural flood defence systems

The calculation or estimation of expected average annual flood damages (EAD) is now a commonly employed step in benefit-cost analyses of flood mitigation projects, and is derived by establishing a loss-probability relationship for any floodplain or area under investigation (see for example Penning- Rowsell et al., 2005a). To estimate EDS1 we need to take into account the proportion of unprotected floodplain properties (UFP), and estimate the proportion of EAD1 which will be saved in the unprotected properties by flood warning systems (FDO). Estimates are likely to be fairly crude, at least initially, and this model assumes that flood defences will not be breached or overtopped. The UFP value can be increased judgementally to take these effects into account if necessary.

EDS1 = EAD1 x PFP x FDO ……………………… ...... Equation 3.8 where:

EAD1 = expected annual average flood damage without any flood defences PFP = the protected floodplain: the proportion of properties at risk from flooding which are unprotected by structural flood defence systems FDO = the proportion of EAD1 which is likely to be saved by the operationalisation of flood defence measures which are dependent upon a flood being forecast and a warning be available to the operators of flood defences

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 103 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Estimates of EDS2 require values to be assigned to each of the warning response variables in Figure 3.12:

EDS2 = (EAD1 x UFP x CBO) + (EAD1 x UFP x WCM) + (EAD1 x UFP x SAR) + (EAD1 x UFP x EVAC) + (EAD1 x UFP x BCP) + (EAD1 x UFP x CFP) + (EAD1 x UFP x CME) ...... Equation 4.318 where:

UFP = unprotected floodplain: the proportion of properties at risk from flooding which are unprotected by structural flood defence systems and relies upon a flood forecasting and warning system for protection of life and property.

CBO = proportion of properties protected by small-scale demountable/moveable flood defence systems installed at the community or neighbourhood level following a flood forecast and warning.

WCM = proportion of potential flood damage saved by watercourse maintenance activities before and during a flood (e.g. removal of blockages, maintenance of efficient flood flows, protection against overtopping and breaching where feasible, deployment of contingency flood storage areas).

SAR = proportion of flood damage to property saved by search and rescue operations. This will normally be zero or close to zero as SAR is usually mainly aimed at saving life and limb.

EVAC = proportion of flood damage to property saved by human evacuation operations. This will normally bed zero or close to zero as EVAC is aimed at saving life and limb.

BCP = proportion of flood damage to property and business activities avoided by the use of business continuity plans.

CFP = proportion of flood damage to property avoided through contingent flood proofing measures, operated once a flood warning is received.

CME = proportion of flood damage to property avoided by occupants moving contents either upstairs out of the reach of the flood, or by evacuating property from path of flooding.

3.8.2 Opportunities and limitations associated with the ‘pathways’ model The FWRBP model provides an opportunity to trace a much wider range of economic benefits than the existing UK/EU FHRC model, even in its modified form, and to exclude much less. The model presents a much truer, and much less partial, picture of warning response than the first model. The analysis may also be taken deeper into economic effects. Although they remain very difficult to evaluate and measure (Parker et al., 1987), indirect flood losses are potentially captured by the pathways model (i.e. through analysis of the effects of business continuity plans), whereas they are not in the first model.

The pathways model is not dependent upon regular post-event social survey derived data for parameter calibration; although in some cases these data have been used to estimate the value and loss-saving potential of the pathways. Such data collection can be costly. The model could be used as a national or regional level to generate low resolution, general parameter values, and may be used at a catchment or settlement level with values being derived from survey evidence.

18 The application of this model can be seen in Section 8 of the milestone report T10_07_12 and in Section 3.9

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 104 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

3.8.3 How the model may inform flood warning system enhancement The FRWBP model suggests a range of ways in which flood warnings may be responded to in order to reduce flood damage potential. Consideration of this model in each situation in which flood warning systems are installed can lead to additional – and particularly valuable and earlier – damage- reducing measures being taken in future. For example, at the moment it would be rare to find that all of the damage reducing means indicated in Figure 3.12 have been systematically considered and evaluated in any particular flood and flood warning case. We have discovered cases where measures which are applicable are not currently being considered and employed.

Clearly in some flood locations and conditions not all of the flood damage reducing measures indicated in Figure 3.12 will be applicable. In such cases damage reduction will depend more heavily on one, two or a small number of measures, but the consideration of combinations of measures can lead to increased damaged reduction. It may be that, in particular cases, the full potential of watercourse maintenance on receipt of a flood warning has not yet been exploited, and similarly with other measures.

Rapid response catchments and flash flood locations pose a particular challenge where there are no structural flood defences. In these cases, focus upon WCM, SAR, EVAC, CFM and CME becomes critical, and BCP can also make contributions to damage savings by ensuring the most rapid return to normal business conditions possible. CBO in terms of erection of demountable defences appears to be largely infeasible in flash flood circumstances, although pumping systems might be feasible.

We return to these issues in Section 3.10 where we set out our principal findings.

The following section illustrates the application of both the EU FHRC model and the new Flood Warning Benefits and Response Pathways model to the case study of Grimma, Germany.

3.9 Case Study: Modelling the flood damage reducing effects of flood warnings in Grimma, River Mulde, Germany

3.9.1 Case study context and data The town of Grimma, located on the River Mulde in the Elbe catchment, suffered an extreme flood event in 2002. The wider flooding in 2002 in the Elbe catchment, in which Grimma is located, led to a considerable amount of research in the aftermath of the flood and to a wealth of information about the various settlements affected, including Grimma, the flood, the response of both the authorities and the public to the flood, and also the flooding consequences (DKKV, 2004; Saxony State Office for Environment and Geology, 2004; Thieken et al., 2005; 2007; Kreibich et al., 2005; 2007; Steinführer and Kuhlicke, 2007). We have extracted data relating to Grimma from these sources. In some cases the sources fail to disaggregate information to the town level, and it has been necessary in these cases to draw implications about Grimma and the flood response there.

In addition to these secondary sources, data has come from interviews held with ten key stakeholders involved in the flood management process. These interviews were undertaken by UFZ (Helmholtz Centre for Environmental Research), and include (among others) the mayor of Grimma and representatives from the town planning council, the water authority, local trade association, the Saxon dam authority, the regional planning authority, and the Saxon Ministry for the Environment and Agriculture.

3.9.1.1 Physical flood producing features of the Mulde catchment Grimma is a small town of approximately 17,000 people located in Saxony within the catchment of the River Mulde, which is a tributary of the River Elbe. The Mulde catchment covers an area of

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 105 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

7.345km² (Reichhoff and Refior, 1997; p3) and includes almost two thirds of the area of the Saxon Ore Mountains (Erzgebirge) – see Figure 3.14.

Figure 3.14: Map showing Grimma within the Mulde catchment, and also the other Elbe tributaries Source: http://mulde.hydrology.rub.de/images/mulde_gebiet_en.jpg (05.11.2007)

The Mulde is itself fed by three tributaries: the Freiberger Mulde (spring source: 855m above sea level) from the south east, the Zwickauer Mulde (spring source: 770m above sea level) from the south west, and the River Zschopau from the south. The two principal Mulde tributaries unify upstream of Grimma at an altitude of 133m above sea level. Here they form the United Mulde (Vereinigte Mulde), which flows through Grimma and other towns downstream.

The evolution of major floods within the Mulde catchment is fostered by its spatial features. These include a large share of the mountainous upstream catchment areas, high relief energy, a narrowing catchment shape, and a small share of downstream lowland areas, where river discharges might otherwise spread out instead of forming flood peaks.

3.9.1.2 Climatic factors A high proportion of the Mulde catchment is characterised by relatively high relief energy. Heavy precipitation can rapidly develop into floods such as the event which occurred in August 2002. In this event soil infiltration capacity was overwhelmed by the precipitation volume and intensity. The two main flood producing events are a) sudden snow melt events and b) heavy summer rainfall caused by special weather constellations - the so called Vb-situations. These bring warm and wet air masses northwards from the Mediterranean Sea. The precipitation prior to the August 2002 flood was a maximum of approximately 350 mm/m² within 24 hours19.

3.9.1.3 Characteristics of the town of Grimma Grimma’s has a total population of 17,000. The flood prone town centre is inhabited by approximately 2,000 people. Grimma developed from two early medieval settlements, one developed by merchants, and the other by craftsmen. Owing to two fords in the river Mulde, trading of goods

19 http://de.wikipedia.org/wiki/Elbehochwasser_2002

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 106 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 became important over time. Furthermore the town acquired the privilege of trading timber from the Ore Mountains. The timber was shipped from the south on the Mulde as caber rafts. Important trading routes (especially the Via Regia) crossed through Grimma and contributed to the creation of local economic wealth. Besides fisheries and mills, river water was of importance for tanning industries (glove manufacturing). The river became an important feature of local identity. Even though the river has now lost much of its economic importance to Grimma, today is attracts water sports and tourists.

3.9.1.4 Historical flood events in Grimma Floods are not a recent phenomenon in the Mulde. For Grimma the following flood events shown in Figure 3.15 are reported from different sources.

Figure 3.15: Graph highlighting the most serious flood events affecting Grimma and their maximum flood depth (water level in cm20) Source: Büttner (no date)

In former times the people of Grimma tried to develop strategies for addressing the flood risk. In the 15th century large scale dumping of soil and debris was carried out within the medieval town to raise the ground level and reduce flood risk (Pesenecker, pers. comm., 2005). Summer as well as winter floods pose problems for Grimma, and severe flood events have occurred in both seasons in the past.. While summer floods are caused mainly caused by Mediterranean cyclones, the winter floods tend to be a consequence of sudden snow melt events and ice jams which occur at the stone bridge to the north of, or just downstream of, Grimma. Dynamiting of ice shields from the front of the bridge is documented from the 18th until the early 20th century. In the past considerable flood damage has been experienced in the town. The town wall was destroyed in the past, and two churches also suffered severely.

3.9.1.5 Geological risk factors The town centre is located directly in the floodplain and is underlain by loam and quaternary gravels. These enable the water table (hydraulically connected to the river water body) to rise unhindered and to flood the town from beneath. Furthermore, the river bank opposite the town is naturally formed by granite walls: thus floodwater cannot spread to that river side. In short the town centre is surrounded by obstacles to floodwaters and these keep water from expanding in the plane.

3.9.1.6 Constructional risk factors The constructional configuration of the city centre contributes to the flood damages suffered. The direction of the major streets in the city centre is parallel to the river Mulde. Thus, the flood water can

20 Even though the absolute water levels have never reached the 2002 value, the relative water level in the city of Grimma was probably as high as that, since the ground level has increased through time (esp. due to fires, after which people rebuilt houses on the basis of old house debris).

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 107 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 reach a high velocity when flowing through the town. During the 2002 flood, the streaming water washed out huge holes from the surface with a depth of up to 4 metres in the places where it passed at highest velocity. The southern end of the town is completely open to the river and the stone bridge (Pöppelmannbrücke) can cause flood heightening due to jams by tree trunks or the masses of flowing water themselves.

3.9.2 Flood management in Germany Steinfuhrer and Kuhlicke (2007) provide a comprehensive examination of the flood management system in Germany and Saxony. However, briefly the structure of the German flood management system is as follows.

Country level: Flood forecasting is the responsibility of the German Weather Service21. It delivers general flood warnings. The majority of responsibilities for flood warning dissemination are anchored on the federal state level.

State level: Technical measures in flood protection are a task of special state agencies that deal with questions of water engineering, dams and reservoirs22. For Saxony it is the Landestalsperrenverwaltung (Saxon Dam Authority) which is responsible for raw water supply, watercourse maintenance and flood protection (State Reservoir Administration, 2007). This authority has its own network of river gauging stations within all river catchments and sub-catchments in Saxony. It provides detailed flood warnings for affected catchments. To secure an effective and efficient handling of flood information, the Saxon Flood Centre was founded.

3.9.3 The current Saxon flood warning system The flood information and alert service in the Free State of Saxony is headed by the Flood Control Centre based in the Saxon State Agency for Environment and Geology (Landesamt für Umwelt und Geologie –LfUG), Figure 3.16. The Saxon Flood Centre provides all relevant flood information directly to each authority with flood defence responsibilities, as well as to any third parties (private persons) at special risk of flooding (using an SMS network). All flood messages and warnings are constantly available on the information platform of the Saxon Flood Centre.

Figure 3.16: The reporting and information chain for flood related messages in Saxony Source: Saxon Flood Centre Website (2007)

21 http://www.dwd.de/en/en.htm 22 They are named differently in each of the Federal States and it is important to note that besides having responsibility for flood protection, their other responsibilities and remits sometimes differ.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 108 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

In case of rising water levels, all authorities within the potentially affected river catchment are informed, in particular the Mayors of all villages and towns. If the municipalities do not assign a different unit, the local fire brigades are also responsible for the local flood fighting measures.

In law, flood warning in Germany operates on a 4 tier system. These are briefly summarised in Table 3.7.

Table 3.7: The four stages of the German flood warning system Stage Description Stage 1 This is based on a constant analysis of the meteorological and hydrological situation. At the local level during this phase of the warning, the alarm plans are routinely checked and the utilisability of the equipment is controlled. Stage 2 At the local level, the dikes along the river as well as endangered buildings are now systemically observed. Additionally, the operational readiness of the responsible staff and the flood protection material is controlled. Stage 3 At the local level, the dikes are constantly monitored and possible mobile preventive safety measures are put into place. Additionally, a task force is established consisting of people who are given responsibility during the crisis. Furthermore, special communication channels are installed and further man power for a possible active flood defence is concentrated. Stage 4 At the local level, the organisations responsible now have to prepare everything for a possible evacuation of the population. Additionally, man power and material are concentrated for flood defence. If the final stage of warning is no longer sufficient to handle the situation, the regional district officially declares a disaster.

Important stages of the flood warning process follow the exceedance of level 3. At this level the flood fighting brigades begin with protection measures, such as erection of mobile defences and sandbags. After warning level 4 has been declared, people evacuation procedures being to be implemented.

3.9.3.1 The effectiveness of the Saxon Flood Warning System In the August 2002 flood, the Saxon flood warning system revealed shortcomings. The flood experience revealed that, at all levels, the communication and transfer of information was unsatisfactory. The clarity of the communication hierarchy required improvement and roles and responsibilities required clarification. The Kirchbach Report (2002) examined the system and recommended the flowing changes:

• concentration of responsibilities for information and alerting into one unit; • concrete guidance for authorities; • streamlining of information pathways (in a top-down system); and • safeguarding of contact availability in all communities and municipalities at any time.

The August 2002 flood is described time and time as an extraordinary and completely unexpected event. Thus, the Saxon flood warning system was subjected to a particularly severe test. Prior to this event, floods of slower onset and smaller magnitude had always been predicted and communicated satisfactorily to the local level. In the August 2002 flood, State authorities state that there was a general lack of information available about the likely flood onset velocity, and the magnitude and duration of the flood.

Following the reorganisation of the flood warning system in Saxony, the system is now believed to be much more efficient and effective, especially regarding clarity of event prediction and the speed of information communication and transfer. The “baptism of fire” for the new Saxon flood warning and information system was the 2006 spring flood which affected parts of the Upper Elbe catchment around Dresden and Saxon Switzerland. Due to early information and warning, people were

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 109 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 successfully informed about the flood and its probable magnitude one week in advance. They had sufficient time to remove their belongings from the flood prone area. As a consequence the material damage was low. To date, no event comparable to the exceptional 2002 event has occurred. Since experience with new flood extremes is still very limited, the effectiveness of the system is only proved for smaller and medium magnitude events.

3.9.4 The Grimma Flood Warning System Like many other towns along the Mulde, Grimma did not receive any proper prior warning or related information about the August 2002 flood. After the flood, and given the lack of a timely flood warning by the administrative district office, the town council decided to install an autonomous flood warning system using some of the financial donations provided following the event. This flood warning systems comprises the following five elements:

• central hooter sirens on town roofs and a central flood announcement system; • autonomous SMS – information network (after announcement of alarm level 1;) • a river gauge camera – live streaming on the internet (see Figure 3.17;) • 24hours information in situations of approaching flood conditions s on local TV Muldental; and • house threshold measuring: to assess how much time is left until flooding.

Despite the new central information platform and warning unit (in the Saxon Flood Centre) the Grimma flood warning system now exists and is intensively used by the town’s population. This is especially true of the river gauge web camera illustrated in Figure 3.17). Other settlements, including Dresden, have set up independent flood warning systems. These flood warning systems are currently additional to the legal requirements and are currently being tolerated by the State authorities. Tolerance may be reduced and may end if problems occur due to overlapping or conflicting or contradictory warnings being issued by two different flood warning organisations or if incorrect information is provided that leads to additional damages or threatens public safety.

Figure 3.17: Screenshot from webcam of the level of the River Mulde at the Pöppelmannbrücke in Grimma Source: Stadtverwaltung Grimma Website (2007)

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 110 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

3.9.4.1 Effectiveness of the Grimma Warning System In some of the interviews undertaken by UFZ (Helmholtz Centre for Environmental Research), people from the local authorities were asked how effective they believed the autonomous flood warning system was in Grimma. All respondents stressed the importance of the flood warning system in improving the flood preparation or lead time, especially after the very short lead time experienced in 2002. The respondents believe that a large proportion of material losses may be avoided by removing assets from the flood risk area before the onset of inundation. Thus, the system is considered by local people to be potentially highly effective. However, the people of Grimma are also interested in structural flood measures.

3.9.5 Flood damages in Grimma The 2002 floods in the Elbe region caused huge damages with inundation flood depths range between two and four metres. Figure 3.18 provides some photographic evidence of the devastation caused in Grimma. Damage estimates for the 2002 event in Germany were reported to be €9.2 billion (DKKV, 2004) in November 2002.

Figure 3.18: Images of flood damage in Grimma during and after the 2002 flood Sources: http://proheritage.info/saxony02/Grimma1Ffb.jpg (top left image) http://www.greenpeace.de/themen/klima/nachrichten/artikel/neue_klimastudie_verheisst_nichts_gutes/(Bottom left image) http://www.stern.de/wissenschaft/natur/511015.html?eid=511340&nv=ex_L3_ct (Right hand image)

The City Administration for Grimma has presented estimates of the flood damages caused by the 2002 floods, as well as a breakdown of these into damages types (Table 3.8) In the 2002 flood event just under €100 million damage was caused to private residential property, of which only 16% (€12 million) was caused to the contents inventory of the property. This percentage is lower than what might be expected in less severe flooding. Structural damage to buildings was particularly high in Grimma owing to the depth and velocity flooding.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 111 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 3.8: Estimated flood damages experienced in Grimma during the 2002 flood event Damage type Damages (in €) Percentage of the total Urban development measures (including 62 701 856 28% administrative buildings) Residential households Buildings 62 539 000 28% Interior 12 130 000 6% Non Residential Properties Buildings and Interior 41 894 000 19% Infrastructure 14 557 549 7% Sports and leisure facilities 7 660 883 3% Schools 15 202 186 7% Other social facilities 3 411 379 2% Total 220 096 853 100%

Source: City administration of Grimma (Stadtverwaltung Grimma, 2003)

Meyer et al. (2007) undertook meso-scale flood damage estimation23 for an area including the town of Grimma. This research presents a range of estimated annual average damages including the direct damages to both non-residential and residential structures and their contents. It is important to note that damages to vehicles or infrastructure such as roads and railways are excluded from these estimates. Indirect damages, particularly business disruption and the associated losses in revenue, are also excluded.

Table 3.9 presents three estimates of average annual damages for Grimma. The average annual damages have also been re-calculated taking into account the planned structural flood defences (discussed in more detail in Section 3.9.8.1). The implementation of these measures which are to be designed to the 1 in 100 year standard is calculated to save between 26% and 36% of the estimated average annual damages for Grimma.

Asset value data for flood risk areas of Grimma indicate that in some areas of the town, residential damages contribute to 40% to 55% of the total damages, whereas in others areas this percentage is very low (V. Meyer, pers. comm.) Therefore, in the subsequent analysis below an average of residential damages comprising between 30% and 41% of the total damage is used. These values are corroborated by the flood damages experienced in the 2002 event. As illustrated in the data in Table 3.8, residential damages accounted for 34% of the total damages sustained.

Table 3.9: Meso-scale damage estimation of average annual damages for Grimma town centre Average annual damages Minimum Mean Maximum Status quo 347765 559660 832868 With planned defences 89350 173658 298096 Difference -258415 -386002 -534771

Source: Meyer et al. (2007)

23 For full details about the method of estimating average annual damages please refer to Meyer, V., Scheuer, S. and Haase, D. (2007) GIS-based Multicriteria Analysis as Decision Support in Flood Risk Management. UFZ Discussion Papers 6/2007. Leipzig. http://www.ufz.de/data/DP_2007_06_Meyer7117.pdf

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 112 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

3.9.6 Post flood response and recovery in Grimma

3.9.6.1 Compensation of losses The 2002 flood caused about €6.2 billion damage in Saxony. In Grimma €220,096,853 damages were registered. These losses were compensated from different sources. Table 3.10 gives an overview of the contributions from different sources. The figures are 2002 values and reveal that some damages were not compensated for. The German Construction Bank (KfW) offered low interest credits for people who wanted to reconstruct their houses. In addition, further Governmental compensation was provided but there is no comprehensive data on the sums involved. Altogether, in terms of flood loss compensation, UFZ surveys (e.g. in Eilenburg on the Mulde) reveal a high degree of satisfaction among the affected population (see 3.9.6.3).

Table 3.10: Sources of recovery from damages sustained in the 2002 floods Purpose Source Sum (€) Recovery of private households Private donations 7 953 693 (Sum losses:€12 130 793) Insurance pay outs 3 313 836 (Sum assistance:€ 12 054 130) Emergency aid from the district council 457 600 Emergency aid from municipality 319 000

Recovery of damages on buildings Private donations 4 961 084 (Sum losses: €62 539 215) Insurance pay outs 8 970 321 (Sum help: €39 460 335) Emergency aid from the district council 1 000 Credits from Saxon Construction Bank 25 527 929 (SCB)

Recovery of small industry and trade Private donations 5 024 685 (Sum losses €41 894 156) Insurance pay outs 304 641 (Sum help: €9 955 430) Emergency aid from the district council 7 000 Emergency aid from municipality 348 000 Credits from SCB and German CB (KfW) 4 271 104 Source: http://grimma.de/11_infos/shownews.php

3.9.6.2 Rebuilding Works The rebuilding works in Grimma started very quickly after the flood and by the end of 2004 most damage had been repaired and the town had been reinstated to is former flood prone state or better. During interviews with the Grimma authorities in 2005, the majority of the private house owners or landlords were reported to have failed to install flood proofing measures during the reconstruction. In Grimma, 45 buildings had to be demolished because of heavy structural flood damage, but very little evidence of this remains today.

3.9.6.3 Impact of compensation and other measures on incentives for action The impact of compensation and structural flood protection measures on private incentives for action is of importance. The flood event of 2002 is perceived by local people to be extraordinary and unexpected. People expected Governmental support and in most cases received it. Most of the households were insured properly with elemental damage insurances (insurance contracts from GDR time encompassed all types of damage; the GDR state insurance was bought by the Alliance Insurance Group after 1990). After the flood, the insurance companies raised the cost of insurance for people who reside in flood risk areas. Thus, not all households can now afford “all-inclusive” insurance any more and now live without it. Only a minority of residents have adopted precautionary constructional resistance measures. In a study of Eilenburg people demonstrated their interest in a high degree of flood awareness and reported that they are much more aware now of what to do in the case of a flood warning, although most believe that a recurrence of the 2002 event is very improbable. In other

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 113 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 regions in Germany that experience more frequent flooding, the incentive for private action seems to be much higher than in Grimma, especially since the cost of insurance is increasing. In relatively infrequently flood affected areas, such as Grimma, people tend to be relying upon a new structural flood defence system being installed and upon a perceived low probability of flooding.

3.9.7 Application of the EU FHRC model to Grimma Data for the model application are from a range of sources, including: a) interviews with the local authorities in Grimma, and b) 2002 post-event reports and surveys of flood affected residents (including Steinfuhrer & Kuhlicke, 2007; Thieken et al., 2005, 2007; Kreibich et al., 2005; 2007). The 2002 flood event provided an excellent opportunity to better understand the impact of flooding and also provided an impetus for investigating flood responses. The EU FHRC model (equation 3.5) discussed in Section 3.6 is represented in Figure 3.19. In the situation in England and Wales the ability (PHR) variable was excluded in the revised model on the grounds of the latest research findings from England and Wales. However, for Grimma we have initially reintroduced this variable so that we can establish whether or not it is likely to be a significant factor in determining flood warning response.

In Grimma flooding only becomes significant in terms of flood damage potential in a 1 in 50 year event or greater. Floods with shorter return periods than this affect very few properties and where they do people are usually well prepared. In the application below we therefore only examine flooding at and above the 50 year event.

FDA = (TPD x PID x MID) x RAS x PHE x EVAC where:

FDA = Flood damage avoided - the estimated actual flood damage avoided owing to the flood warning TPD = Total potential damages PID = Potential inventory damage MID = Moveable inventory damage RAS = Reliability of the flood warning process combined with proportion of householders available to respond to a warning. PHR = Ability - the proportion (%) of property occupants able to understand and respond to a flood warning) PHE = Effective response – the proportion (%) of properties for which an appropriate flood warning service is provided, where the occupants are either willing to take effective action or which have actually taken effective action following a flood warning to reduce flood damages EVAC = proportion of properties whose residents are evacuated and who therefore were not able to save property contents

Figure 3.19: Components of the EU FHRC model

3.9.7.1 Potential flood damage avoided To model potential flood damage avoided in Grimma we need to consider TPD, PID and MID (Figure 3.19). For Grimma, values for these variables are based primarily upon calculations from the estimations of the average annual flood damages. Table 3.7 highlights that the mean estimated annual damage (above a 1 in 50 year level: below this level damages are assumed to be zero) is €559 660. On average it is thought that between 30 and 41% of total damages are attributable to domestic property, and therefore, taking a mid value of 35% from this range, the estimated annual average damages for residential property is considered to be €195 880.

Figure 3.20 is from Merz et al. (2004) and provides data collected from the HOWAS database. This database provides average flood damages for properties collected from a number of flood events in Germany. Although none of the events are located in our region of interest, the HOWAS values provide an approximate percentage for the components of contents inventory damages for properties; both total and moveable. From Figure 3.20 potential contents inventory damages (PID) are estimated as being 40% of total household damages. Of these potential damages, moveable inventory damages

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 114 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

(MID) are considered to make up the majority, and are therefore estimated to be 80% of these potential contents inventory damages.

Figure 3.20: Histogram from Merz et al. (2004) illustrating the mean damage (total damage, damage to building structure, damaged to fixed inventory, damage to movable inventory) for different economic sectors

3.9.7.2 Reliability and effectiveness of the flood warning and the householders ability to respond (RAS) RAS is a difficult value to estimate for Grimma. There are no relevant data relating directly and exclusively to Grimma. From expert interviews undertaken in Grimma it is estimated that in August 2002 90% of people would have been available for a flood warning, although shortcomings were clearly apparent in the flow and communication of information and the flood warning. Thieken et al. (2007) conducted a study in the German language investigating the 2002 flooding. One of their three interview groups is from the Elbe tributaries which includes Grimma. One of their research questions concerned how people found out about flooding. 42% of people reported that they did not receive a flood warning. By implication 58% of people received a flood warning. Since the 2002 flood a new flood warning system has been introduced in Saxony and in Grimma, and above we refer to this as working satisfactorily but this is in locations neighbouring Grimma (in small to medium scale events) and not in Grimma itself. Our assessment is that the value of 58% should be used for RAS in Grimma. Indeed, in comparison to England and Wales this is a high value.

3.9.7.3 Ability to take action Ability (PHR) to take action requires further examination. National, regional and local statistical sources allow estimates to be made of the proportions of the population who might be unable to take damage reducing measures (see Section 3.6.3).

Age For the Saxony region 23% of the population are over the age of 70 years, whereas 15% of the population are over 70 years (Gesundheitsberichterstattung des Bundes, 2007). These estimates are for 2006 and are higher than the averages for the rest of Germany which are 19.5% and 12.9% respectively.

Disability Various statistics are available because there are many different definitions of disability. The EUROSTAT database (Eurostat, 2007) presents one main statistics of interest. This is a percentage disabled for all of Germany in 2001. It is based upon an assessment of the fraction of the population who are ’Hampered in daily activities by any physical or mental illness or disability’. This source

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 115 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 identifies 10.2% of the German population as being severely hampered, and a further 29.5% of the population who are to some extent hampered.

Regional statistical sources suggest a lower value for disability. Data from the Federal Health Monitoring database (Gesundheitsberichterstattung des Bundes, 2007) of the ‘Number of severely handicapped persons with official certification’ provides a figure of 300,489 for Saxony which equates to 7% of the population. Information for Grimma’s municipality is also available from Statistiches Landesamt des Freistaates Sachsen (2007) which suggests that 65 out of 1000 people are disabled in Muldentalkreis, or 6.5%. During the August 2002 event it was reported that ten people needed to be evacuated from the flood zone by ambulance due to disability or illness.

Understanding the warning Finding information about the numbers of people who are unable to understand a flood warning is problematic General statistics identify those who are either not German nationals, or those who are now nationalised but who are originally from a different country or ethnic origin. However, these sources do not identify those who do not speak or understand the German language. Information from interviewing key flood stakeholders suggests that in Grimma in 2002 there were few communication issues owing to language difficulties.

In summary, the percentages of both the elderly and disabled are relatively comparable to the UK, and it is estimated that there are few people in Grimma who would not understand a flood warning. In addition, apart from the 10 people who were reported as being assisted to leave their homes in Grimma in August 2002, there is no evidence to suggest that the population are different to those on which they conclusions about PHR are made in Section 3.5. We therefore exclude ability (PHR) from the model in this application (Table 3.10).

3.9.7.4 Effective response Effective response (PHE) is the proportion (%) of properties for which an appropriate flood warning service is provided; where the occupants are either willing to take effective action or which have actually taken effective action following a flood warning to reduce flood damages. Evidence from Section 5 clearly indicates that the preparedness of people is higher where the experience of flooding is frequent. Recency of flooding can also increase preparedness. However, expert interviews undertaken in Grimma indicate that, owing to the relative infrequency of flooding there, few are properly prepared. There are, however, few quantitative data to support this assertion.

Steinführer and Kuhlicke (2007) undertook a survey of people in the catchment of the middle Mulde in 2005. When asked about whether they were prepared for flooding in the 2002 flood, 85% of those affected stated that they were not prepared at all for flooding. They were not, therefore, able to respond effectively to reduce flood damages. Thieken et al. (2007) also investigated effectiveness in their survey of the 2002 floods. Their results indicate that, of those who received flood warnings, many attempted to take damage reducing actions (Figure 3.21).

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 116 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 3.21: The emergency measures performed, in descending order, as a percentage of all interviewed people for the Elbe, Elbe tributaries and Danube (multiple answers possible) Source: Thieken et al. (2007, p1029).

Just over 50% of people took moveable contents upstairs. Steinführer and Kuhlicke (2007) also examined the types of actions which they categorise as ‘Ad hoc activities reducing material damage’. People were asked what they did first after hearing that a large flood was coming. 31% reported that they tried to secure something (21%) or many things (10%), whereas a further 22% left their homes but took their most important things with them. Often, however, the most important possessions are likely not to always be the most valuable when measured in monetary terms. Taking all of the available evidence into account our assessment is that a value of 31% should be applied to PHE in this example. In selecting this value we have weighted our assessment towards what is specifically known about Grimma.

3.9.7.5 Calculating flood damages avoided (FDA) From the above sections it has been possible to assign values to each of the components of the FHRC model. Table 3.11 summarises these figures. For this initial calculation it is assumed that the evacuation component is not relevant, primarily because of the fact that evacuation will only occur in the most severe events and therefore only affect a small proportion of the depth damage curve. Prior to 2002, the last flooding event that prompted evacuation was in 1974. In addition, to this the circumstances of the flooding mean that even when there is a flood warning and then an evacuation order there is still a relatively long lead time (6 to 8 hours which might increase with the new flood warning system) whereby residents are still able to take action to secure their possessions.

Table 3.11: Values to apply to the EU FHRC model Model component Estimated percentage Total potential residential damages TPD €195 880 Potential inventory damage (as a percentage of TPD) PID 40% Moveable inventory damage (as a percentage of PID) MID 80% Reliability of the flood warning process combined with RAS 58% proportion of householders available to respond to a warning. Effective response PHE 31% Evacuation EVAC n/a The calculation of flood damages avoided is as follows:

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 117 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

FDA = (TPD x PID x MID) x RAS x PHE = 195 880 x 0.40 x 0.80 x 0.58 x 0.31 = € 11 270

The total potential residential flood damages avoided annually by the timely activation of flood warnings in Grimma is €11 270 or 5.8% of the overall estimated average annual residential damages. This equates to 2.1% of the overall estimated average annual damages for Grimma.

3.9.7.6 Commentary on the results Flooding in Grimma is complicated by the large range of flood depths experienced. Many of the figures in Table 3.10 are selected from information gathered from the 2002 flood event. This was a very extreme and infrequent, and arguably untypical, event and may not provide a sound guide to people’s typical damage reducing actions. In addition, the severity of the flood meant that a level 4 flood warning was issued only six hours after the level 2 warning. This final flood warning level advises people to quickly evacuate the area. In an event such as this, where flooding onset is rapid, flooding is very deep, and the consequences of remaining in the flood risk area are likely to be fatal, much of the focus of individuals is to remove themselves from the danger as soon as possible. This may indicate that in less severe flood events the proportion of people likely to undertake effective action may be higher than the 31% value used above. In flood events in which the rate of rise is slower, and flood depths are lower, there may well be more time for damage saving actions as the priority given to damage saving may rise. This is another reason why applying the evacuation component of the model would be inappropriate in these circumstances as by doing so would further reduce the loss savings and potentially even further underestimate their value.

Thieken et al. (2007) asked those who took no emergency measures during the 2002 flood, why they did not take action (Figure 3.22). Although it is difficult to present absolute percentages as multiple answers were allowed, the data indicate that the overwhelmingly most important reason was that people believed that it was too late to do anything. People may not have received a flood warning with sufficient lead time to act.

Figure 3.22: The reasons why people did not undertake emergency measures during the 2002 flood (multiple answers are possible) Source: Thieken et al. (2007, p1028)

Those who undertook emergency measures in 2002 were also asked to comment upon how effective they felt these measures were Thieken et al. (2007). On a scale of 1 to 6 (1 being the most effective) the safeguarding of contents only received an effectiveness score of 3, whereas many of the other actions scored higher. The main factor influencing effectiveness appears to be flood depth. Many of those who secure contents are likely to move inventory to the upper floors of their properties. However, the mean depth of floodwaters in Grimma in August 2002 is estimated at between 2 and 4 metres). In many properties flooding therefore reached into the storey above ground floor level

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 118 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 limiting damage saving. In the 50 or so buildings which collapsed in flood in Grimma, many of their contents were lost.

Figure 3.23 illustrates an indicative depth damage curve and the point in the curve whereby the mean flood depth is greater than 2.5 metres24; the depth at which water will begin to flood the second storey of most properties. In instances where the average flood depth is 2.5m or higher very few contents (that is not removed from buildings) will be saved. In Figure 3.23 are shaded red identifies those damages in which the moving of contents to an upper storey is likely to produce few damage savings, as the moved contents will subsequently become inundated. A more refined estimate of flood damage savings is possible if the value of those damages in the red shaded area are removed from the possibility of damage saving. In Grimma floods ranging from a 1 in 75 year to a 1 in 200 year event are likely to require this treatment. Meyer et al. (2007) modelled the extent and depth of 200 year flood in Grimma, which indicates 2 to 3 metre flood depths (Figure 3.24).

Damage

2.5m Depth

Figure 3.23: Indicative depth-damage curve illustrating the proportion of the average annual damages that is more difficult to save in deep floods through moving

Figure 3.24: Inundation model showing expected flood depths in the 1 in 200 year flood event in Grimma Source: Image courtesy of Volker Meyer (Meyer et al., 2007)

24 Please note that this exact depth may vary between different regions and countries and may be altered according to the typical style of building found in an area.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 119 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

If the value of the 200 year flood is taken to be the threshold above which the savings of damages becomes more difficult because of inundation above ground floor ceiling levels, the flood damage savings calculation can be reworked as in Table 3.12.

Table 3.12: Recalculation of the damage savings possible through moving of contents in Grimma Model component Estimated percentage Total potential residential damages (approximately recalculated TPD €100000 for the proportion of the damages below a 1 in 200 year level) Potential inventory damage (as a percentage of TPD) PID 40% Moveable inventory damage (as a percentage of PID) MID 80% Reliability of the flood warning process combined with RAS 58% proportion of householders available to respond to a warning. Effective response PHE 31%

The recalculation for flood damages avoided is as follows:

FDA = (TPD x PID x MID) x RAS x PHE = 100 000 x 0.40 x 0.80 x 0.58 x 0.31 = € 5754

The total potential residential flood damages avoided annually by the timely activation of flood warnings in this case have been reduced to €5754. Therefore, when discounting the proportion of the damages when flooding averages above 2.5m, it suggests that only 1% can be effectively saved from the overall estimated average annual damages for Grimma.

3.9.8 Application of the Flood Warning Response Benefits Pathway (FWRBP) model to Grimma, River Mulde Table 3.13 presents three flooding scenarios for Grimma and highlights the circumstances in which different flood warning response and benefits pathways become more or less significant to the damages saved. The three scenarios are a) a flood that exhibits a mean depth of 0.6 metres and a maximum depth of 1 metre (approximately a 1 in 75 event), b) a flood that exhibits an mean depth of 1.2 metres and a maximum depth of around 2 metres (approximately a 1 in 100 year event), and c) the 2002 event where the flooding ranged from 2 to 4 metres in depth. The 2002 flood is estimated as being between a 1 in 200 and a 1 in 250 year event.

Table 3.12 suggests the importance of each of the different factors affecting flood warning benefits, but is a simplification of reality. The warning response pathways which we assess to be most important in Grimma are shaded in the table, and differ according to the flood event. This suggests that a different flood warning response strategy may be advisable for each flood in the case of Grimma. Also for some of warning response and benefit pathways, the significance of the damage saving strategy may vary spatially across the floodplain. For example, CFP is only feasible with floodwaters up to less than 1.5 metres, and so CFP may be feasible and useful in some parts of the floodplain but not in others.

In the August 2002 flood in Grimma, the loss savings consequent upon a flood warning were concentrated in the evacuation (EVAC) and search and rescue (SAR) components of the model, because nearly everyone managed to vacate the flood risk area prior to the most extreme flooding. Although in Grimma there was no official organised evacuation before the flooding, 100 people (around 8% of the flood risk population) were rescued from their flooded houses during the flooding (Priest et al., 2007). Other residents of Grimma were required to self-evacuate. Because people evacuation and search and rescue was the key response to a flood warning in the 2002 flood in Grimma, flood damage savings to contents was lower than might have been in a more modest flood event.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 120

Table 3.13: The differential importance of the flood warning and benefit pathways in floods of different magnitude in Grimma, Germany

Warning and benefit Description of pathway components 1 in 75 year event (approximately) 1 in 100 year event (approximately) 2002 flood event (c. 200-250 year event) response pathway Maximum depth 1m 2m 4 metres Mean depth 0.6m 1.2m 2-4 metres FDO Operationalisation The main flood defence features that need The regulation of the flows will have a large impact The regulation of the flows may have a large impact FDO is unlikely to be effective. The of flood defences operationalisation in Grimma are the use of dams to and may be sufficient to avoid the worst of the and may be sufficient to avoid the worst of the magnitude of this flood means that utilising regulate the flows (although new defence measures are flooding. flooding. This may make a smaller difference than in the dams to regulate the flow and pumping out planned). In addition, the new pumping scheme can be the case of a lesser event, but it still may be significant water will be relatively ineffective. used to reduce the amount of water damage seeping in reducing the worst flows or allowing people more behind the defences. time to move property out of the flood zones. CBO Operationalisation In Grimma, logs are used at the side of the river to These measures will be effective in preventing the These measures will be effective in preventing the These features will mostly be ineffective in h of community prevent flood damage. These require 2 to 3 hours to be damages to the structures of buildings through damages to the structures of buildings through debris. flood velocities and depths characteristic of based measures erected, and they are stored close to the town centre. debris. Demountable defences may be overtopped if the this event. They are not designed to prevent flooding but to maximum depths of 2 metres are experienced. prevent debris from causing damages. WCM Watercourse There is an obligation on the Federal authorities to In this situation, a timely flood warning may permit A timely flood warning may permit emergency Both insignificant and potentially dangerous maintenance maintain and clear watercourses as the Mulde is a level emergency clearance of the river to be carried out if clearance of the river if required. Although in the to the operatives in this magnitude of 1 river. This duty is required both pre-and during required, particularly at the bridge above Grimma. higher flood depths there may be less time to safely flooding. flooding. carry this out. SAR Search and rescue Search and rescue is a major aspect of the response to SAR may well be necessary for people trapped in properties or find themselves in areas where there is fast SAR was very important in saving 100 people flooding in Grimma, and is coordinated through the flowing water. Instances of rescue will be lower in less severe flood conditions and events with a sufficient during the 2002 event and will be significant and usually the local fire brigade. lead time to warning. in any event of this magnitude. EVAC Evacuation In the 2002 pre-event evacuation was not formally Evacuation might be appropriate where there is fast flowing water or vulnerable members of the community Although not necessarily significant in the undertaken, although neighbouring towns in the region are exposed to the danger of flood waters. Instances of evacuation are expected to be less, the lower the 2002 flooding, evacuation in these instances experienced pre-flood evacuation.. Evacuation is magnitude of the flooding. will help to avoid human losses. prepared at the fourth and highest level of flood warning process. After this level a disaster is announced. CME Moving/ This is an option open to all householders who are Given enough warning time contents inventories Given enough warning time contents inventories could This pathway was not particularly significant evacuation of physically able, have somewhere to move their could be moved out of the flood waters and damages be moved out of the flood waters and damages saved. in the 2002 event as the majority of effort was house contents belongings (i.e. have a property of two storeys or saved. in ensuring safety to people. In addition, for more), who have received a timely warning and who those who did manage to move some of their are available to react. Awareness that this is an belongings, the depth of the flooding would appropriate action to take is also a factor in the success mean that moving it upstairs would not have of this measure. been sufficient. CFP Contingent flood This might include activities such as erecting flood These features would be effective, according to their There will be some areas of the floodplain where Similar to above, in such deep flooding these proofing measures boards or private barriers to prevent the water entering design standard, in preventing flood water entry, flooding is greater than 1m and these measures will be measures are not significant as the flooding in the building. assuming that they are erect both promptly and rendered ineffective. most areas would be higher than the measures correctly. can cope with rendering them ineffective.. BCP Implementation of This may take the form of flood proofing or individual If implemented correctly, planning will be The success of the flood proofing of businesses will Any flood proofing activities or moving of business defences, or the moving of equipment or stock from significant in this event context provided a timely be dependent upon their design effectiveness, if 2 inventory or equipment (unless completely contingency the threat of the water, and the establishment of warning. metres of water is experienced some of these flood away from the site) would prove to be planning. alternative emergency flood free business locations. proofing measures will become ineffective. Similarly, ineffective due to the large amounts of movement of stock and equipment would need to be flooding experienced. either out of the flood zone or above 2m. Summary Damage savings generated by flood warnings are Damage savings generated by flood warning are more Damage savings generated by flood warnings feasible in such an event, particularly as the majority difficult and will be a mix of human and damage loss will be primarily concentrated on the human of the warning response and benefit pathways offer savings depending upon where people are located in loss savings Property/contents damage potential benefit. Human loss savings might be the flood risk zone. Damage savings are likely to savings are likely to be lower for events of necessary in some circumstances where individuals come primarily from flood defence operationalisation, this magnitude. or groups may need assistance. community measures, business contingency planning and the moving of contents

Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

3.9.8.1 Flood Defence Operationalisation (FDO) At present there is a no structural flood protection in Grimma. However, following the devastation caused by the 2002 floods, a flood wall is planned. The proposal is to design and construct protection up to a standard of the 1 in 100 year flood. Following implementation of this wall the estimated mean annual average damages for the town have been estimated to be reduced to €173 658 (reducing current potential average annual damages by 69%) (V. Meyer, pers. comm.). This proposal has been in the planning stages since 2003 and construction is planned for 2008. The proposed measures are shown in Figure 3.25 and comprise:

• construction of a new wall section combined with mobile elements (orange line in the image); • reconstruction of the old city wall and integration of a protection wall into the city wall (red line); • object protection (integration into existing walls), new protection wall and mobile flood defence components (green line); and • construction of a rampart with an integrated protection wall (yellow line).

Figure 3.25: The planned location of proposed structural flood defences for Grimma Source: www.umwelt.sachsen.de/.../staatsbetriebe/ltv/downloads/Hochwasserschutz_Grimma_Massnahmekonzept.pdf

Since the ground is made up of gravel and sand, water level changes occurring during floods must be controlled. Introducing an adequate pumping scheme is rated as one of the most challenging parts of providing flood protection for Grimma especially as the pumps will require a reliable source of energy during an event. The proposals clearly indicate that the flood defence measures will require operationalisation to be effective. It is believed that the gaps in the defences can be closed quickly; within 2 to 3 hours of this being considered necessary. A flood warning lead time of at least 3 hours is therefore required for the integrity of the proposals to be maintained.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 122 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Further complementary measures are also being considered, including the provision of further upstream flood storage by the use of sluices and dams. These can help moderate flows and reduce flood peaks. However, a warning lead time of 2 to 3 days is required to make these measures work effectively. The Mulde is one of the fastest rising rivers of its type. Flood detention and storage may only be effective over longer periods of higher than average flows rather than in peaky events. These measures are also only likely to be effective for short return period flood events. A nominal value of 1% damage saving (prior to the construction of the flood wall) is assigned to this category in the application below.

3.9.8.2 Watercourse maintenance (WCM) In floods affecting Grimma debris jams against the Pöppelmannbrücke causing heightened flood levels and the potential for increased flood damages. The responsibility for clearing and maintaining the watercourse rests with the federal authorities because the Mulde is a level 1 category river. These authorities have an obligation prior and during flood events to clear watercourses and to remove debris. Reconstruction of the Pöppelmannbrücke following the damage caused in 2002 is planned. However, the replacement bridge has been designed to reduce the amount of obstruction that takes place during flooding.

Undertaking routine watercourse maintenance before flooding events should be straightforward. More difficulty in undertaking these activities will be experienced during flood conditions or when the river is rising, owing to increased flood depths and velocities. WCM is therefore only likely during flood warning levels 1 and 2 (Table 3.7): beyond this the work becomes too dangerous. Watercourse maintenance will be possible and most effective in the more frequent, but less severe events in which the clearance of the channel will have a greater impact upon reducing flood levels and related flood damages. During severe flooding the impact of clearing the channel will be less pronounced.

WCM is likely to be most effective in the shallower floods. Our assessment is that a value 5% may reflect the effect of WCM on reducing flood damages in Grimma.

3.9.8.3 Community based Operationalisation (CBO) Stop logs which are deployed in Grimma just prior to or during a flood have little effect upon preventing water from entering properties. However, they make some contribution to preventing debris damage to properties. The use of stop logs as a temporary flood defence measure is most effective at reducing damages during flooding of shallower depths, although their effectiveness needs examining further. As floodwater is still able to inundate buildings with the stop logs in placed, only some of the building damages and contents will be saved. Our assessment is that a value of 1% reflects the damage saving potential of these measures in Grimma.

3.9.8.4 Business Continuity Planning (BCP) There is little information about the role of business continuity planning in Grimma, or about how businesses were affected in the 2002 flood event. Section 3.9.5 presents the estimated damages that occurred in Grimma during 2002: €41 894 000 is estimated as being the direct damages to non- residential properties (including businesses), both to the contents and building structures However, no estimates exist for the indirect cost of business interruption or for losses to the local and regional economy. Such losses are also excluded from the estimated average annual damage calculations for Grimma.

Kreibich et al. (2007) undertook a survey of 412 companies that were affected by the 2002 flood event in the Elbe region. The town of Grimma was included in the survey but results are not disaggregated for Grimma. As the survey also included a range of sectors, it is possible to make some tentative inferences about the behaviour of businesses in Grimma during the 2002 floods. For some of the results the data is also broken down between those businesses in flash flooding areas (which include Grimma) and therefore for these, more specific data can be applied. 48% of companies surveyed

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 123 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 reported not receiving a flood warning: thus the damage reducing actions they were able to take were very limited. The mean flood warning lead time for this area was calculated at 10 hours (median 6 hours).

Figure 3.26 provides some estimation of the breakdown in damages in the 2002 floods, and provides evidence about the savings which may be made through taking emergency measures. The International Commission for the Protection of the Rhine (2002) also estimates that as much as between 50-75% of damages may be saved by businesses undertaking effective emergency measures and through contingency planning. These values appear optimistic to us and probably reflect and indicative potential rather than current actuality.

As previously stated since the estimated above were included areas of both flash flooding and slower onset flooding, they are potentially presenting an average position for Grimma, (i.e. accounting for the extreme circumstances and more modest events). There are also no additional figures available for this pathway and so the estimation of 15% of businesses saving 40% of total damage (the average of the buildings and interior damages) will be applied. Further research on the situation is of course required to refine this estimate further.

Figure 3.26: Breakdown of estimated flood damages highlighting the damages caused by the 2002 flood in the Elbe catchment (after Kreibich et al. (2007; 12))

As indirect business interruption losses are excluded within the estimated average annual damages for Grimma, they are also excluded from the calculation of damage savings. From Figure 3.26 there is an indication that on average business interruption is estimated to have generated about 25% of the total flood damages in the 2002 event. Therefore, undertaking business continuity planning may well provide a means of increasing flood warning benefits in the future.

3.9.8.5 Individual level action including Contingent Flood proofing (CFP) and Contents moving and evacuation (CME) These two warning response and benefit pathways are considered together here as they both involve action by individual householders and may be linked. Section 3.9.7 discusses the impact that the raising or evacuation of contents may have upon the damages sustained during flooding. Using this procedure we estimate that, on average, 6% of total damages can be saved by CME.

There is not much information on CFP in Grimma, almost certainly indicating that it is not common. The flood managers in Grimma corroborated this view. Thieken et al. (2007) and Kreibich et al. (2005) provide some evidence. Table 3.14 summarises Thieken et al’s evidence and highlights those precautionary measures taken before the 2002 floods in the survey group containing Grimma.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 124 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 3.14: Information concerning precautionary measures undertaken by households located on the Elbe tributaries Precautionary measure Percentage of households in the Elbe tributaries that had undertaken this activity Adapted building use c. 15% Adapted furniture and equipment c. 15% Moved utilities upstairs c. 10% Sealed cellars c. 5% Purchased flood barriers c. 8%

(Data estimated from Figure 3 in Thieken et al. (2007, p1025).

Kreibich et al. (2005) argue that generally people were not well prepared for the 2002 flooding. As illustrated in the table above, few households had undertaken any precautionary measures despite living in an extreme flood risk area. Kreibich et al. also go on to suggest that significant savings of between 46 and 53 % could be made to both structural and contents damage by adapting buildings. However, the adaptations which they consider would help reduce flood damages but they are not examples of contingent flood proofing: they are permanent flood proofing measures not dependent upon a flood warning.

Only around 8% of those in the Elbe tributaries had purchased private flood barriers before the 2002 flood. A further 12% had done so following the 2002 flooding and the survey suggested that this value would rise to close to 30% of households because 10% reported planning purchasing flood barriers in the six months following the survey. Thus the importance of these measures may increase in future. However, the value of 30% might not be applicable to Grimma which suffers from more rapid onset flooding than the other areas included in regions surveyed by Kreibich et al. (2007). Also proposals to build structural flood defences in Grimma are likely to deter people from purchasing and using CFP measures. Overall, our assessment is that a value of 8% of households adopting CFP measures is realistic for the Grimma both now and in the future.

It is also important to consider the effectiveness of these flood proofing measures. Kreibich et al. (2005) argue that in the 2002 floods the implementation of private water barriers had no significant impact on the level of damage to contents. The severity of the event may be responsible for this as many these barriers would have been overtopped. The impact of barriers on saving building damages is reported as being slightly more significant, with a reduction in damages by 29% for those properties fitted with barriers. To summarise, in the 2002 flood on the Elbe, the 7% of the households affected managed to save no significant contents damages; but 29% of buildings damages were saved by using CFP measures.

However, it is again important to consider the influence of taking figures from an extreme event. The International Commission for the Protection of the Rhine (2002) argues that if the floods barriers remain intact and are not overtopped by floodwaters, these barriers are able to reduce damages by 60- 80%. An estimate of 8% of households (from those who already have or intend to purchase barriers) of average savings of 35% has been tentatively used for this value as for some events the maximum damage savings of around 70% will be made and for the very extreme events no savings will be made. On average, we assess the value of total flood damages to building and their contents to be reducible by 3% (of the residential property component) through employing CFP measures.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 125 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

3.9.8.6 Calculation of the current potential loss savings from flood warnings employing the FWRBP model Table 3.15 presents the values assigned to each warning response and benefit pathway for Grimma, and the calculation of potential flood damage savings attributable to flood warnings in this town.

Table 3.15: Application of the FWRBP model for Grimma Warning and benefit Percentage value Model formula Calculation for Grimma (€) response pathway assigned Flood defence 1% EDS1 = (EAD x 559660 x 0.01 5 597 operationalisation FDO) (FDO) EDS1 5 597 Watercourse 5% (EAD x WCM) + 559660 x 0.05 27 983 maintenance (WCM) Community based 1% (EAD x CBO) + 559660 x 0.01 5 597 operationalisation (CBO) Business Continuity 6% of the proportion (EAD x BCP) + 364000 x 0.06 21 840 Planning (BCP) of AAD related to non- residential Contingent flood 3% of the proportion (EAD x CFP) + 195 880 x 0.03 5 876 proofing (CFP) of AAD related to residential properties Moving and/or 7% of the proportion (EAD x CME) 195 880 x 0.07 13 712 evacuation of Contents of AAD related to (CME) residential properties EDS2 75 008 TEDS = EDS1 5 597 +75 008 + EDS2 TOTAL €80 605 DAMAGE SAVING

Therefore, as Table 3.15 illustrates, the total loss savings according to our application of the FWRBP model top Grimma is €80 605, or 14% of the total expected average annual damages.

3.9.8.7 Calculation of the future potential loss savings employing the FWRBP model for Grimma The construction of the proposed structural flood defences together with the proposed pumping scheme will impact significantly on the expected average annual damages (these are reduced to € 173 658 in Table 3.8). However, in order to be fully effective parts of the defences require operationalisation. Therefore, some of the damage savings are attributable to FDO measures. Since, it is virtually impossible) to model the effectiveness of the defences should the gaps in the defences not be closed, or when the pumping scheme fails to work, we make the assumption that the defences are 100% efficient; and therefore the maximum loss savings will accrue.

Since the floodwall will only be effective up to the 1 in 100 year standard,, residual flood damages will occur.. It is assumed that the floodwall will overtop (rather than breach) and, therefore, the losses on the floodplain might still be reduced. For the case shown in Table 3.16, we use the same values as in Table 3.15, but we apply them to the revised estimate of expected average annual damages. Table 3.16 provides the recalculation for Grimma in a post-structural protection era.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 126 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 3.16: Application of the FWRBP model to Grimma with the proposed flood protection scheme installed Warning response and Percentage value Model formula Calculation for Grimma (€) benefit assigned Flood defence In this scenario the FDO component has 559660 -173658 386 002 operationalisation been re-estimated with a revised version of (FDO) the average annual damages calculated with the presence of the flood wall (see Table 9.3) Watercourse 5% (EAD x WCM) + 173658 x 0.05 8 683 maintenance (WCM) Community based 1% (EAD x CBO) + 173658 x 0.01 1 737 operationalisation (CBO) Business Continuity 6% of the proportion (EAD x BCP) + 112878 x 0.06 6 773 Planning (BCP) of AAD related to non- residential25 Contingent flood 3% of the proportion (EAD x CFP) + 60 780x 0.03 1 823 proofing (CFP) of AAD related to residential properties26 Moving and/or 7% of the proportion (EAD x CME) 60 780x 0.07 4 255 evacuation of Contents of AAD related to (CME) residential Properties16 EDS2 23 241 TEDS = EDS1 + 386 002 + 23 241 EDS2 TOTAL €409 243 DAMAGE SAVING

Applying the FWRBP model, the potential total future loss savings increase significantly because of the installation of structural flood defences which requires operationalisation on receipt of a flood warning to be effective. Damage savings are estimated to be €409 243 which is 73% of the estimated total average annual damage potential (without any flood damage reduction measures). 69% of these damage savings are generated by the effective operationalisation of the flood defences.

3.9.9 Discussion and conclusion

The purpose of this section is to apply and calibrate both the EU FHRC and the FWRBP models to the town of Grimma on the River Mulde, and in so doing to provide a case study. We encountered data difficulties in the application of both models. Given this we assembled the data and qualitative evidence available to us and made assessments grounded upon this evidence to formulate values for use in the model calculations. The modelling may be improved in future by obtaining higher quality data, but this is likely to require further targeted research. It is clear that the outcomes of the models are much more sensitive to some values than others, and so value sensitivity should be used to guide data improvement effort in future.

It is instructive to compare the results of the modelling. It is clear that an application of the EU FHRC model misses many of the potential benefits of flood warnings in Grimma, and is therefore likely to be misleading on its own. For example, for the current situation in Grimma (i.e. prior to implementation of the proposed structural flood protection scheme), we estimate that 2.1% of total expected annual average flood damages are saved by flood warnings (or 6% of the residential flood damage potential).

25 Assuming that this figure still remain as 65% of the new estimated average annual damages 26 Assuming that this figure still remain as 35% of the new estimated average annual damages

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 127 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

These values are approximately halved if the effect of second storey flooding and deep flooding is taken into account. In monetary terms, therefore, the potential damage savings attributable to moving or evacuating property contents are very modest. Limited flood warning reliability, property occupier availability and warning response action effectiveness contribute to these damage savings being modest, and if these factors were less of a constraint in future, flood damage savings might be a little less modest. However, a key factor driving down flood damage savings in Grimma is the priority which has to be given to saving lives through evacuation which takes effort away from material flood damage savings.

When the FWRBP model is applied to Grimma, much larger flood damage savings are attributable to flood warnings. 14% of the total expected average annual flood damages are potentially avoided. Damage savings are still limited by the priority given to life saving, but the damages saved are now significant in monetary terms. The impact of installing the proposed flood mitigation scheme for Grimma is to increase the potential damage savings attributable to flood warnings to a much higher level – almost 5 times more than the savings without a scheme – and these savings are very appreciable ones.

Grimma is not a typical case because of the depth of the flooding which can occur. The rapidity of flooding onset also places Grimma towards the end of the flooding spectrum, although some of Europe’s most challenging flood risks are represented by rapid-onset flooding risk. Given the range of flood depths experienced in Grimma, the case highlights the fact that different flood warning response strategies may be needed for floods of different magnitude. This is well illustrated in Table 3.13 in which different combinations of response pathways are suitable for each of the flood magnitudes considered. Grimma’s flood characteristics probably magnify this finding somewhat and it may not be so applicable to cases where flooding characteristics are less extreme and the range of flooding conditions is less. But this finding suggests that, generally, flood warning response strategies might need to be different for short and long (particularly extreme) events in the same location.

3.10 Principal findings and conclusions 3.10.1 The purpose and methods of the research project

3.10.1.1 Purpose This purpose of this research project is to contribute ways of improving the benefits that society derives from flood forecasting and warning systems, so that such systems can play their fullest part in managing flood risks successfully in Europe.

Flood forecasting and warning systems have a significant role to play within integrated flood risk management. Adopting the ‘portfolio approach’ which is currently being advocated and which has started to be implemented in Europe, flood forecasting and warning systems have the potential to be combined with both structural and other non-structural measures to make the greatest positive impact in terms of reducing flooding risks to society. To some extent this integrated, portfolio approach is already adopted and in practical beneficial use in parts of Europe, but the potential to extend the approach – using flood forecasting and warning systems – appears to be large. This potential cannot be realised, however, until we more fully understand the benefits of flood warning systems and before we can enhance flood warning response which is woefully insufficient in many current cases.

The purpose of this research project is directly relevant to the implementation of the 2007 EU Directive on the Assessment and Management of Flood Risks (EU, 2007). Firstly, the Directive states that:

‘flood risk management plans shall take into account relevant aspects such as costs and benefits….’ of flood risk management strategies (EU Directive Section 4, paragraph 3).

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 128 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

The research reported here directly addresses this requirement because it focuses upon the development of several models which enable the benefits of flood warning systems to be estimated (these benefits being flood damages saved by flood warnings).

Secondly, the Directive states that:

‘Flood risk management plans shall address all aspects of flood risk management focusing on prevention, protection, preparedness, including flood forecasts and early warning systems and taking into account the characteristics of the particular river basin or sub-basin’ (EU Directive Section 4, paragraph 3).

The research reported here directly addresses early warning systems and related preparedness and protective measures, and it does so by distinguishing between different characteristics of floods.

3.10.1.2 Methods We have addressed the research objectives using a social science research methodology, but one which is adjusted to the particular purposes of the project. Our research is conceived in two theoretical frameworks. Firstly, in theoretical terms the research on the estimation of flood warning benefits sits within a welfare economics, benefit-cost framework in which the objective is to identify those improvements (e.g. the extension of flood warning arrangements, or the lengthening of flood warning lead time) which deliver net economic efficiency benefits to society, but a framework which we recognise might be more appropriately considered as a multi-criteria analytical framework. Whatever the analytical framework being considered, our research on flood warning benefits is designed to help decision-making make wise investment decisions about flood warning systems in the future. Secondly, in theoretical terms our research on improving the response to flood warnings employs a behavioural paradigm in which we do not view flood warning response as some kind of mechanical or pseudo-mechanical human reaction to flood risk information. Nor do we view flood warnings as only emanating from an official flood forecasting and warning source as is often assumed in the design of such systems. Behavioural theory suggests quite the opposite: that human beings perceive flood risks and respond to them according to a wide range of social, psychological and behavioural variables which themselves are likely to be influenced by cultural, political and institutional factors.

Our methods of data collection were heavily influenced by the availability of data on socio-economic dimensions of flood risk and floodplain use, flood damage, flood warning systems and warning response. These are all areas where there is currently a paucity of data in Europe, and where data collection methods need to be greatly improved (see 3.10.5 below).

We make the methodological assumption in this paper that our research models and results apply throughout Europe, but in so doing we incorporate socio-political and institutional factors which vary greatly across Europe. Where these differences are likely to be important we identify and/or discuss them.

3.10.2 The evolution of flood warning systems in Europe We introduce the idea of a revolution in flood warning systems in Europe. We conclude that this revolution has a) partly taken place and b) has partly not yet happened. The proportion of development/potential development assigned to a) and b) in Europe depends upon one’s viewpoint and vision, but in our view they might be distributed in a 25%/75% manner. In other words, advances have been and are being made, but a great deal of potential remains to be realised and capitalised upon.

A pattern exists in the way in which flood forecasting and warning systems have developed, and are being improved, in Europe. Investment and advances tend to be focused first and foremost at the flood detection and forecasting end of the flood forecasting and warning process, and as advances are

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 129 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 made in these areas eventually advances are also made in flood warning communications. This has been noticeable in recent years as flood warning systems have benefited from the rapid advancement in personalised information and communication technologies. It is only when it is recognised that responses to flood warnings may be inadequate, and that the monetary benefits of flood forecasting and warning systems are not being achieved, that interest begins to focus upon these benefits and their calculation. At the same time questions start to be asked about why flood warning responses are insufficient and why flood warning systems do not appear to be as effective as perhaps originally believed, except perhaps in reducing loss of life in floods. At this stage the spotlight turns to issues surrounding ways of enhancing flood warning response and the take-up of flood preparedness and self-protection measures by property occupants.

Although the pattern described above is a simplification of how flood warning systems may evolve, the general pattern of evolution is a common one. The current focus in Europe upon non-structural flood management measures, and national flood management policies which emphasise ‘finding room or space for rivers or water’ has helped to strengthen the focus upon flood receptors (e.g. people living in floodplains) and the extent to which their responses to floods are adequate and may be improved.

Based on the research evidence presented in this research report, below we identify the conditions which we believe are necessary to put in place in order to greatly enhance flood warning response and the associated benefits of flood warnings. We believe that, although technical advances in flood detection and forecasting are vitally important, much of the economic benefit of such advances will be ‘frittered away’ and lost if considerable attention is not paid to providing or creating the right conditions for high and successful levels of flood warning response.

3.10.3 The estimation of flood warning benefits One of the principal results from this research project is the provision of two methods for estimating, in quantitative monetary terms, the flood damage savings produced by flood warnings. Both methods are now suited to application and use in Europe, but these applications depend upon the availability of data. To counter data unavailability we suggest flexible ways in which these models may be used.

3.10.3.1 The models for estimating the benefits of flood warnings In this research project we estimate the effectiveness of flood warnings according to the flood damages they save where these flood damages are to property of one sort or another. Flood warnings also save lives and prevent physical and other injuries and trauma, but these benefits are excluded in this project except for mention of them where relevant.

The starting point for this research project was an existing model for estimating flood damages avoided by flood warnings which had been produced in the UK during the 1990s. This model (technically a variant closely based upon the model) had been employed by the Environment Agency in the UK to estimate potential flood damages avoided by flood warnings, in order to underpin the national flood warning investment strategy. This is the UK FHRC model.

At the beginning of the current research project it was apparent that the existing UK FHRC model required recalibration in the light of new survey evidence from England and Wales and that, more importantly, the model needed further refinement. Recalibration was achieved through a Defra/Environment Agency research project (Tunstall et al., 2005), but reanalysis of the data was required as part of the current research project in order to refine the model (Parker et al., 2007). The initial product of the current research project is a refined UK FHRC model, which is presented and discussed in Section 3.5.4.

The refined UK FHRC model was then revisited and critically reviewed in the context of the types of cases which we encountered in the course of this project across Europe. We found that the model was inadequate, particularly because evacuation of floodplain properties appears to have been much more significant in recent floods than in those analysed for the UK research (i.e. Tunstall et al., 2005). The

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 130 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 outcome of this critical review and conceptual ‘testing’ against European conditions is that the model has been further refined as part of this project. The re-conceptualisation is reported in Section 3.6.

The end-product from the research development of the UK FHRC model is a model which we believe is applicable in Europe and a wide range of flood conditions present in Europe, given the availability of data (the specificity of which is as flexible as is possible to help overcome potential data difficulties). The research on this model is specifically directed at meeting project objective 1 (see 3.2.2). This model allows the estimation of flood damage savings from flood warnings which are generated by property owners and occupants responding to the warnings by moving their property contents out of the reach of floodwaters. This kind of response is the main response which currently underpins the national flood warning investment strategy in England and Wales, and which has driven forward a range of flood warning improvements there.

During the re-conceptualisation of the UK FHRC model, we developed an entirely new model of flood warning response and benefits which is presented in Section 3.8. The Flood Warning Response and Benefits model (FWRBP model) is the second major product of this research project, and it is specifically directed at satisfying research objective 2 (see 3.2.2). It may be applied in a particularly wide range of flood circumstances (wider than the first model above). This model is fully applicable to Europe and has the benefit of capturing a much wider range of flood warning benefits than the UK and the Europeanised FHRC model. This is because it focuses upon a wider range of warning responses which are becoming more common in Europe, but which need to be developed more as part of a new flood warning response strategy (see below).

3.10.3.2 Estimates of actual flood damages saved according to key variables It has proved particularly difficult to provide estimates of actual flood damages save by flood warnings. This is because a pre-requisite for estimating flood damage savings from warnings is the availability of base-line flood damage data. Unfortunately, in most parts of Europe flood damage data do not exist in the same way as they do, for example, in the UK, and there has been no history of the development of such data sets. Through extensive networking, and through our partner surveys, we managed to identify a number of such data sets, none of which exist for the national level as in England and Wales. For example, we found data sets for the Czech Elbe and parts of the Loire valley as well as for parts of Germany. Estimates of flood damages saved draw upon these data sets

3.10.4 The application of flood loss reduction models to case studies As indicated above, we have incorporated a number of case studies into the research. This was part of the initial research design and was incorporated in a version of the objectives specification (i.e. the RIP for this project). Initially, it was hoped that we could use FLOODsite pilot studies, but this proved unfeasible and we identified other case studies for this project.

The rationale for the choice of case studies is explained in Section 3.3.8 and the case study of Grimma on the River Mulde is presented in this report. The other examples and case studies detailed are presented in the full milestone report (T10-07-12).

3.10.5 Enabling a better response to flood warnings

3.10.5.1 How people respond to flood warnings, and the protective actions which they take Objective 3a)-3b) is a major one in this project. Section 3.4 provides the implications of a comprehensive analysis of flood warning responses presented in the milestone report associated with this research. We are fortunate to have empirical research evidence relating to different types of floods, including rapid response, medium speed response and slow response floods. Although we discovered an evident degree of similarity about how people currently respond to flood warnings across Europe, we conclude that the factors affecting warning response and protective behaviours (including preparedness for floods) are complex and complexly inter-related.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 131 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

From examining the current similarities in the way in which people in Europe respond to flood warnings, it is clear that currently a significant proportion of people are likely to do nothing to prepare for floods and are likely to take no action, other than perhaps actions designed to save their lives, following a flood warning. Many people are found to use a combination of flood warning methods, including a) personal observation to self-warn, b) unofficial community warning systems (based upon kinship, neighbourhood and community linkages and observations), and c) official flood warnings which are communicated by local or central government or their agents. The need to get confirmation of a flood warning, before taking any damage or loss reducing action, is an established research finding which receives further strong support from the evidence available to us in this research report. People most readily respond to warning to save their own lives, or the lives of others (especially family members and pets), but in seeking to confirm a warning a significant proportion of people further endanger their lives through the information seeking and curiosity satisfying actions which they take. Of those who take action to reduce flood damage once they have received a flood warning, the most common responses appear to be to try to prevent floodwater entry and to remove possessions including cars, items deemed important (such as medicines) and other possessions (usually memorabilia first and foremost). A relatively small minority of people prepare for flooding by taking contingent flood proofing or contingent resilience measures (i.e. measures, the implementation of which, depends first upon receiving a flood warning. An even smaller proportion of people adapt their properties permanently to flooding in order to reduce flood damage potential. Some people underestimate the flood risk. French research reveals that among car drivers more underestimate flood risk than over-estimate it, which has an impact upon their risk behaviours including their behaviours after receiving a flood warning.

We found that ‘context is everything’ in understanding flood warning response. This is so in the case of both the characteristics of the flood event and the characteristics and flood experience of flood warning recipients. Speed of onset appears to be the most critical variable affecting flood warning response because it affects the flood warning lead time available to respond. Flood depth and flood velocity are also important influencing variables because they constrain the responses that can be taken effectively. Socio-economic characteristics of floodplain populations are found to be very complexly related to flood warning responses, except in the case of type of tenure which is consistently found to differentiate both preparedness levels and warning response. One of the most powerful predictors of the prevalence of preparatory measures and flood warning responses is experience of flooding. Flood experience encourages flood warning response.

We conclude with the following generalisations about how people are currently likely to respond to flood warnings, from the evidence which we have assembled in this research report:

• appropriate response to flood warning is currently close to non-existent or is often muted, and currently only in a minority of cases is it likely to be better than this or extensive; • there is an identifiable set of reasons why people do not currently respond to flood warnings appropriately (see below); • the monetary flood damage savings achieved by flood warnings – originating from property occupiers response to flood warnings only – is currently likely to be modest; and • the potential for improving appropriate response is high; and • the potential for improving warning benefits is reasonably high.

3.10.5.2 The FWRBP model as a template for flood warning response strategies The FWRBP represents a new and comprehensive flood warning response strategy. This strategy integrates particularly well with the new flood risk management strategies which are being pursued in

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 132 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Europe which emphasise a) a portfolio approach, b) non-structural solutions, c) individual responsibility and self-help, and d) resilience measures.

The FWRBP model is directly relevant for the practice of flood risk management, and should be moved into practical application as soon as possible in European member states. It is not currently being applied and it has the potential to enhance flood management benefits and to advance portfolio flood risk management.

During the course of this research project we discovered that those designing flood warning systems and those seeking to encourage an effective response to flood warnings, rarely considered the full range of flood warning response pathways contained in the FWRBP model (Figure 3.12). We believe that in future flood risk management, and flood warning response, strategies should be informed by the FWRBP model, using the model as a template to plan warning response within a portfolio approach.

Clearly, in some cases only one or a few warning response pathways may be relevant and useful, but in other cases we anticipate that the model template will act to suggest additional ways in which flood warning can be capitalised upon to save losses and damages.

3.10.5.3 Factors which encourage and inhibit response to flood warnings It is clear that the socio-political, institutional context and conditions in which people live and work, and in which flood warning systems are embedded, exerts significant influence upon their predispositions to prepare for flooding and to respond to flood warning with protective and damage- saving actions.

In most parts of Europe the state (central, state or local government or their agencies) has a statutory flood risk management role, and this is usually discharged by providing structural flood defences amongst several management measures. Because structural flood protection, and some other areas of flood protection (e.g. provision of flood storage areas, flood control dams etc.), have many of the attributes of being a public good, this common approach is generally considered to be a legitimate one for the engagement of the state. The state may also have extended its role into providing other measures which support flood protection, such as providing flood risk maps. However, unless there is a commonly shared understanding of the role of the state and the role of individual citizens in preventing and protecting against floods, those living and using floodplains may conclude that flood protection is entirely a state matter and not one for them. According to the evidence assembled in this research report, this position appears to have been reached in northern Italy, in Hungary and to some extent in Scotland. De Marchi et al. (2007) accurately refer to this syndrome as the ‘delegation of responsibility for safety’. A pre-requisite for an effective degree of flood preparedness and warning response is therefore the cultivation and constant reinforcement of understandings about shared responsibility for flood protection amongst the flood prone population (Table 3.17).

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 133 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 3.17: Prescription for improving flood warning response

1. Cultivation, reinforcement and constant communication of an understanding amongst the public of the boundaries of responsibility for flood protection of a) the state and b) the citizen (referring where appropriate to the history of flood defence).

2. Nurturing of public trust in flood risk management authorities and agencies.

3. Evolution of either a publicly funded compensation scheme in which compensation is made dependent upon adoption of permanent and contingent flood resilience measures (i.e. including ones which depend upon a flood warning for their operation) or private flood insurance (in which claim settlements are dependent upon the same) with very limited state compensation arrangements.

4. Cultivation of a high degree of understanding among protected citizens about the standards of the flood protection on which they depend, the limits of protection and the consequences of defences being breached or overtopped.

5. An approach to flood risk information which is open as far as all members of society are concerned, coupled with extensive measures to make flood risk information accessible, backed up by a continuous programme of location specific and general flood awareness campaigns. Public information must include the details of flood warning arrangements, and local community members should be enlisted as ‘flood wardens’ to help communicate these details and to raise levels of understanding of them.

6. Encouragement of the retention of local flood knowledge (including knowledge of past flood events, their characteristics and impacts) and knowledge about ways of preparing for floods, responding to flood warnings and taking appropriate and effective actions. Development of understanding and knowledge for the reasons for false warnings. Requires local knowledge to be valued and a forum to display and discuss it with local people.

7. Explicit recognition of different forms of flood warning (a) personal self-warning, (b) unofficial community based warnings, (c) official warning with recognition of the value of each, and the value of combining these forms. Design of official flood warning systems which seek to incorporate local knowledge and local social networks so that personal, unofficial and official warning systems are integrated rather than provision of a top- down, official warning system. Close working with flood prone populations where possible to ensure that warning systems accurately serve their needs, and are reviewed with local community input. This should include input on false warnings and the level of acceptability of them to the community.

8. Measures to ensure a high proportion of floodplain users are available to be contacted to receive a flood warning. Use of mobile telephone and personalised communication technologies to reach people wherever they may be. 9. Measures to ensure that vulnerable members of the community with low or non-existent preparedness and flood warning response capacity receive assistance (including from other community members) to help them prepare and respond. 10. Measures (such as local flood fairs) designed to inform members of flood prone communities of the nature of the flood risk and the permanent measures and products which are available to help them protect themselves from flooding.

11. Measures targeted at particular categories of floodplain user who may lack familiarity with local conditions and flood risk, such as car drivers and tourists.

12. Lengthen flood warning lead time and ensure that flood warnings are communicated appropriately (i.e. with factual and behavioural content of the kind required by flood prone community members) and in a timely manner to all who require them. This means that detailed and accurate records of flood prone addresses and contact details must be maintained: itself requiring high quality floodplain mapping as a pre-requisite.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 134 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

The measures contained in Table 3.17 need to be translated into detailed action plans attuned to national and local circumstances before they can be applied, but these measures provide an important prescription for the improvement of the warning response dimensions of flood forecasting and warning systems. National and local translations of these measures need to be seriously considered and applied as soon as possible in European Union member states.

3.10.6 Methods for collecting data

Objective 5 (see 3.2.2) is concerned with methods for collecting data on the damage reducing actions taken by households upon receipt of a warning. We specifically address this objective in 3.10.6.2 below but, before we do so, we briefly comment upon the collection of wider socio-economic data on floods.

3.10.6.1 Socio-economic data on floods Socio-economic data relating to floods and flood risk in Europe are currently difficult to find, and with a few exceptions, are not widely or systematically collected and analysed. We include here the following data:

• The number of properties of different basic types (e.g. residential, non-residential) located in flood prone areas (e.g. riverine, tidal and coastal areas); • The breakdown of the above into protected and unprotected categories; • The estimated number of people living in areas at risk of flooding • The socio-economic characteristics of populations at risk from flooding; • The geographical coverage of (a) flood forecasting and (b) public flood warning services; • Data on the reliability of flood warning services in terms of the number of people targeted with flood warnings compared to the number receiving flood warnings; • Average or standard flood damage data for common property types; and • Estimates of average annual flood damage potential;

Without data of this kind, it is not currently possible to measure basic aspects of flood risk, and it is not possible to track trends, for example, in exposure of properties to flooding over time.

3.10.6.2 Data on the damage reducing effects of floods During the course of this research project we have developed a number of tools which help identify and systematise the type of data required to enable the damage reducing effects of flood warnings to be estimated. The partner survey tool, the expert interview tool and the FWRBP model are the prime examples. Each is based upon a review of the data required to model the damage reducing effects of floods.

The expert interview tool in particular is based upon the two flood warning benefit models and identifies those variables for which data is required in order to apply the models. The development of the FWRBP model was generated by a critical review of the existing UK FHRC model. It is particularly important in identifying the data needed to model the wider range of warning response and benefit pathways which should now be considered. This model moves data collection needs on from being focused only or solely upon householders, and for the reasons described in Section 4 which underlie the development of this model. In this sense the research has progressed beyond objective 5 which were drafted in the absence of a vision of a broader and more complete flood warning response and benefits model. Nevertheless, below we focus upon this objective.

The principal method used for collecting data on the damage reducing actions taken by households upon receipt of a flood warning has been a social survey tool. Such tools were used to generate the data which is reanalysed when refining the UK FHRC model. Dependence upon social survey tools, used in the post-flood period to find out how householders respond to flood warnings, is a time-

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 135 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 consuming and costly strategy. Such a strategy has been adopted in England and Wales because the Environment Agency wishes to measure its performance in flood warnings by measuring the responses of flood warning recipients, and the post-flood social survey is a reliable method of achieving this. Although this method is advocated for further use elsewhere in Europe it is recognised that such a labour intensive and data-rich approach may not always be preferred outside of in-depth research investigations for which social surveys remain a prime data collection tool.

3.10.7 The main scientific progress and advances made by this research project The main scientific advances made in this project are as follows:

1. two models enabling the estimation of the benefits of flood warnings (i.e. flood damages avoided by the use of flood warnings) have been developed and are now ready for practical application in Europe; 2. the Flood Warnings Response and Benefits Pathways model, developed in this project, provides a new flood risk management strategy which should now be applied practically in Europe; 3. a prescription for improving flood warning response, based upon examination of empirical evidence, has been created which should now be applied practically in Europe.

3.10.8 Gaps in research knowledge The Flood Warning Response and Benefits Pathways (FWRBP) model developed and demonstrated in this research project is a potentially useful tool to enhance integrated flood risk management. The potential for applying this tool within flood risk management strategies has only been indicated in this research project, and further research needs to be undertaken to determine how the tool can be embedded into flood risk management in a range of flood risk management situations.

Further in-depth, detailed research is needed on the flood damage reducing effects (actual and potential) of a) flood detention and storage, b) watercourse maintenance and c) business continuity planning when they are used in combination with a flood warning system. There is now a considerable amount of evidence about the adoption of business continuity plans and measures, but much less is known about how they actually save damages in flood events. Research is already underway on the beneficial effects of other measures such as contingent resilience measures (e.g. resilience measures that are activated once a flood warning is provided) and the results of such research will require integrating into the FWRBP model in due course.

When flood forecasting and warning systems are used in combination with mobile flood defences (either large scale fixed flood barriers with gates, or temporary demountable flood defences), the benefits which arise are attributable to both the structural elements and the non-structural warning elements of the measures. Within a benefit-cost framework, or for investment appraisal purposes, it may be necessary to find a means of apportioning the benefits to each element used in this portfolio approach. To address such circumstances further research is necessary on different ways of apportioning benefits.

The prescription for improving flood warning response, which is a product of the current research project, requires detailed application at national and sub-national levels in Europe. Application requires a translation process which takes into account national and local circumstances and which develops a detailed action plan. The next stage is to undertake this process of translation.

We identified the net benefits of flood warning systems as being a measure of the value of these measures to society, but examination of costs of flood warning systems is excluded from this project. Further work is required on these costs for investment appraisal and related decision-making purposes.

The relationship between evacuation of people from their flood risk properties and flood damage saving potential needs to be researched further. Immediate evacuation can mean that flood damage

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 136 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 savings are zero (since there is no time to move damageable property). Evacuation after a long lead time may mean that damage savings by moving property can be significant. There are a range of circumstances between these two ends of the spectrum.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 137 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

4 Activity 3: The effects of floods and flood-induced pollution on ecosystem health (WL/Delft)

Floods have impacts upon ecosystems through erosion, sedimentation, inundation and through alterations in habitat conditions to flora and fauna. In addition to this, floods may bring along and spread polluted sediments, and also cause the release and spread of pollutants within a flooded area, e.g. from chemical sites, oil tanks, sewage treating plants, and pesticide supplies. This Activity aims at improving the assessment of flood impacts and effects of flood induced pollution on ecosystems, individual species and on relevant indicators for ecological quality (e.g. for biodiversity, naturalness, ecosystem health, similarity to a reference system (standards, where appropriate related to the Water Framework Directive). It aimed to develop a model framework in which the eco-toxicological stress levels during flooding can be predicted using OMEGA.

4.1 Introduction to the OMEGA modelling framework

That certain chemical components may cause a toxic effect in organisms or plants is a well know fact. While banning all toxic components from water bodies is desirable, it’s often not possible to eliminate all sources of a specific toxicant. Therefore, to set environmental safety standards, the relation between toxic effects and chemical components concentrations has to be established.

Within the EU Water Framework Directive (WFD), the targets for priority substances for water bodies are set by risk levels (like AA-QS and MAQ-QS). These risk levels are based on observed effects in ecotoxicological risk assessments. The ecotoxicological risk assessments are carried out by establishing the concentration-effect relation for individual chemical components and individual test species (single-species toxicity tests, measuring effects to individuals) (Posthuma et al., 2002; Kooijman, 1987; Wagner and Løkke, 1991; Aldenberg and Slob, 1993).

However, the WFD is based on preserving or improving the ecological status of the whole water body. On the water body scale, ‘simple’ dose effect relations between individual toxicants and individual species have to be aggregated to an evaluation of the toxic effect of chemical components for the whole water body. To resolve this incongruity between individual-based data and the complex biological entities addressed in ecological risk assessment, a model framework (OMEGA) will be introduced to link individual species-sensitivity distributions to risk levels on the water body scale.

4.2 Risk assessment based on species sensitivity distribution

Single-species test data are combined to predict concentrations affecting only a certain percentage of species in a community. Single-species data (e.g., median lethal concentration [LC50] or no-observed- effect concentration [NOEC] values) for many species are fit to a distribution such as the lognormal or log-logistic. From this distribution of species sensitivities, a hazardous concentration (HCp) is identified at which a certain percentage (p) of all species is assumed to be affected. The most conservative form of this approach uses the lower 95% tolerance limit of the estimated percentage to ensure that the specified level of protection is achieved (Posthuma et al., 2002; Kooijman, 1987; Wagner and Løkke, 1991; Aldenberg and Slob, 1993).

Species-sensitivity distribution or extrapolation methods are being incorporated into assessments of ecological risk (Newman et al., 2000). In Figure 4.1 an example of the SSD distribution based on individual NOEC’s for one chemical component is given. In Figure 4.2, the individual NOEC level for species is transferred to a Potential Affected Fraction (PAF) level for the toxicant.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 138 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Species Sensitivities Distribution for Toxic Pressures:

% species at risk 100

80 Algae 60 Amphibians 40 Macroinvertebrates Fish 20

0 -1 -0.5 0 0.5 1 1.5 2 log NOEC (µg/l) Figure 4.1: Combining individual NOEC levels

Figure 4.2: From NOEC’s to Potential Affected Fraction (PAF)

4.3 Modelling framework

OMEGA stands for Optimal Modelling for Ecotoxicological Assessment, predicting effects on plants, animals and populations or ecological functions. The method is based on a stepwise approach:

1. Calculation of the potentially affected fraction of species (PAF). 2. Identification of sensitive species or species groups. 3. Calculation of accumulation in food chains. 4. Calculating effects on development of populations.

Within the current ecotoxicological risk predicament, the focus is on the first two steps within the model.

OMEGA is the aggregation of individual SSD curves to calculate a Potential Affected Fraction. The PAF within OMEGA is based on a wide variety of NOEC’s for individual test organisms.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 139 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

4.4 New developments

The OMEGA framework (Figure 4.3) has been expanded to include:

• More ecotoxicological data. • The use of QSAR’s to fill data gaps. • The selection of subsets of organisms. • Combination of toxic effects of different toxicants. • The use of measured or calculated freely dissolved concentrations (derived from surface water concentrations, whole water concentrations or pore water concentrations in sediments).

Database laboratory tests Substance – species – sensitivity (NOEC/ LC50/EC50)

Fill gaps with QSARs

Select subset e.g. phytoplankton or fish

Dissolved Calculate SSD curve PAF Concentration Potentially affected fraction Specific methods e.g. pore water, Summation of similar SEM/AVS, biomimetic acting compounds Combine to multi- adsorbents TU normalised substance PAF Partition coef Gives toxic impact as % species Chemical measurements that could be present in unaffected Total water or sediment reference that is at risk

Figure 4.3: Modelling framework for determining critical concentrations (bioaccumulation) in aquatic biota (OMEGA)

While the data in the OMEGA basis has been in use for quite some time (the current Dutch legislation for surface water and sediment quality (Forth National policy document on Water Management Government Decision, 2000) and part of the current Water Framework Directive draft fact sheets (Substance Data Sheets: DRAFT of 2004, European commission directorate general, Joint Research Centre) is based on this data), the increase in data availability and new developments like QSAR relations make it possible to improve on the relations between toxic exposure and predicted effects.

4.5 From effect on the total population to effect on specific groups

The OMEGA framework is currently based on setting a PAF risk level based on all available data. This approach is partly based on the lack of ecological cause-effect data, not using a part of the available data would yield insufficient data. Progress with regard to the availability of more ecotoxicological data and the filling of knowledge gaps with QSAR relations yield the possibility to be more restrictive in the data used to establish the NOEC level for a certain group of organisms.

At the moment, OMEGA is adopted to be able to calculate a NOEC for the following groups of organisms: • Phytoplankton;

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 140 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

• Macrophytes; • Phytobenthos; • Benthic Invertebrates; • Fish.

4.6 Combination of toxic effects of different toxicants / Mode of Action

OMEGA is currently mainly based on the effects of one toxicant. Combination of the toxic effects of different toxicants to calculate a multi-substance PAF (ms-PAF) is possible. This can be done according to two principles:

1. Response Addition (RA). 2. Concentration Addition (CA).

Response Addition is based on the absence of interaction between toxicants on the target site of toxic action. The mixture toxicity can be described by calculating the combined effect, assuming that there is no correlation between the uptake of compounds.

Developments of insight into the Mode of Action make it possible to develop a Concentration Addition model based on calculation and summation of Hazard Units for each toxicant.

Both pathways can be incorporated within OMEGA and can help to understand toxic stress in water bodies with more than one significant toxicant (Figure 4.4).

Figure 4.4: Flow schema of OMEGA input and output, including QSAR relations and decision rules

If all toxic stress relations are incorporated in OMEGA and if sufficient toxic data for the water system is present, the PAF for all toxic components can be calculated. This PAF can be split in the PAF for different groups of organisms (see above).

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 141 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

By quantifying toxic component concentrations to a PAF, the toxic component has become a water system stress parameter. This can be done as part of a multi-stress analysis for the water system.

The following Section discusses the results of applying the OMEGA model to a case study of the Western Scheldt.

4.7 Western Scheldt case study: toxic risk prediction in surface water

4.7.1 Introduction The Western Scheldt is a Dutch/Belgium Estuary. The Western Scheldt is an important ecological system and the estuary has a special protected status within the EU Habitat Conservation and Protection Guidelines, Bird directive. This places an extra focus on the chemical quality status of the estuary. The estuary has also a high economic value since it functions as the entrance to the harbor of Antwerp. Especially since the large scale flooding in Zealand in 1953, flood risk management of the Western Scheldt is also an important factor in the management of the estuary.

The combination of these factors and the dedicate balance in the morphological conditions of the estuary (which have an impact on the channel system) make it very complex to change any of the specific functions (like channel deepening for the entrance of bigger sea ships) without impacting the other functions.

From a water quality point of view, the estuary has seen a reduction in the load of pollutants over the last decades. However, historical polluted sites within the estuary (like small harbors) and the discharge of pollutants from the Belgium river Scheldt / Rupel still have an influence on the chemical status of the estuary (by the end of 2006 a new waste water treatment plant for Brussels will impact the pollutant load from the river Scheldt). The question which has been addressed in this case study is if the chemical status can be translated to a toxic risk.

4.7.2 Materials and methods

4.7.2.1 Chemical status The Western Scheldt has an intensive monitoring network for water quantity parameters (Figure 4.5).

Unfortunately, only for Schaar van Ouden Doel (in the East near Antwerp) a complete record on a two week measurement frequency is present for the dissolved concentrations of organic and heavy metal pollutants. For this location the heavy metals Zink and Copper are the dominant pollutants, but other pollutants have also been taken into account: • Cd • Cu • Zn • Naphtalene • HCB • α,β,γ, HCH

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 142 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 4.5: Monitoring points water quantity and quality Western Scheldt

Data from the year 2000 is used – Figures 4.6 and 4.7.

Heavy metals- Schaar van Ouden Doel

80 0.7

70 0.6

60 0.5

50 0.4 Zink 40 Koper Cadmium 0.3 Cadmium ug/l 30 Copper and Zink ug/l

0.2 20

0.1 10

0 0

2 1 2 1 1 -01 -31 -01 -02 -01 -01 1 6 -01 -03-0 -04-0 -05-0 07 -08-0 -09-0 -12 -01 0 0 0 0 0 000-0 000 000-0 000- 000 001 2 2 200 200 200 2 2 200 200 2000-10-01 2000-11-01 2 2 Figure 4.6: Trends in heavy metal concentrations at Schaar van Ouden Doel (2000)

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 143 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Organic pollutants - Schaar van Ouden Doel

0.02

0.018

0.016

0.014

0.012 naftaleen alpha HCH 0.01 beta HCH gamma-HCH HCB 0.008 concentrations (ug/l) 0.006

0.004

0.002

2 1 1 1 1 -01 -01 -01 -01 -01 -03-0 04 -05-02 -06-0 -08-0 -09-0 10 -11-0 -12 0 0 0 0 0 000 000 0 000 000 2 2000-01-31 2 2000- 200 2 2000-07-02 200 2 2000- 200 2 2001-01-01 Figure 4.7: Trends in organic pollutants concentrations at Schaar van Ouden Doel (2000)

4.7.2.2 Sobek 1D/2D The Sobek 1D/2D model was based on the national grid for the Dutch River systems, taking into account the bathometry of the Western Scheldt, the discharge on the river Scheldt and the tidal influence of the North Sea Boundary (Figure 4.8).

Figure 4.8: Modelled Western Scheldt area

After the calculation of the hydrological conditions, the water quality is calculated with the water quality model Delwaq. By taking the North Sea concentration as the Western boundary of the system (Schaar van Ouden Doel is the East boundary), an east-west concentration profile can be calculated as

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 144 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 a function of the water quantity. For this, the hydrological model Sobek 1D/2D has been used in combination with the chemical model Delwaq.

4.7.2.3 Ecotoxicological stress The calculated concentrations for Copper and Zink are then used in the OMEGA model to calculate the ms-PAF and the contribution of each model to the ms-PAF. Since the Sobek 1D/2D model calculates the total dissolved concentration, the output of this model could be used as direct input for OMEGA.

The variation between the different organism groups and the total population was also calculated. Groups of organisms in OMEGA are: • Fish • Benthic Invertibrates • Phyto Plankton • Macrophytes

The number of species and abundance of the population is dependant on the specific ecotope. The ecotope in the Western Scheldt estuary is not constant. Main factors determining the ecotope are: • Salinity • Mud content of the sediment • The inter-tidal zone (sub-littoral, mid-low-littoral and upper-littoral) • Water depth and flow velocity

Within an ecotope, a certain species diversity, species density and species biomass is expected. This defines the good ecological quality status.

4.7.3 Results and discussion

4.7.3.1 Chemical status The Sobek 1D/2D model was calibrated for a period of one year (2000):

• number of chemical observations of water quality on the boundaries: 26 observations; • partitioning between dissolved and suspended sediment of pollutants; • hydrological time step conditions: 2 hours.

The result was compared with the measured concentration – Figure 4.9.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 145 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Heavy metals- Schaar van Ouden Doel

80 3.5

70 3

60 2.5

50 Zink 2 Zn_Sobek Koper 40 Cu_Sobek 1.5 Cadmium Cadmium ug/l Cd_Sobek 30 Copper and Zink ug/l

1 20

0.5 10

0 0 1999-12-06 2000-01-25 2000-03-15 2000-05-04 2000-06-23 2000-08-12 2000-10-01 2000-11-20 2001-01-09

Figure 4.9: Measured (dots) and calculated concentrations (lines) at eastern calibration point (Schaar van Ouden Doel)

The tidal influence and difference in load on the boundaries translates into a time dependant concentration of pollutants – Figure 4.10.

Zn segment 400

26 25 24 23 22 21 20 19 18 17 16 15 14

Zn [] Zn 13 12 11 10 9 8 7 6 5 4 3 2 1 15-01-2000 14-02-2000 15-03-2000 14-04-2000 14-05-2000 13-06-2000 13-07-2000 12-08-2000 11-09-2000 11-10-2000 10-11-2000 10-12-2000 00:00:00 00:00:00 00:00:00 00:00:00 00:00:00 00:00:00 00:00:00 00:00:00 00:00:00 00:00:00 00:00:00 00:00:00

Figure 4.10: Example of zinc concentration in the middle of the Western Scheldt during one year (2000), including tidal fluctuation Besides the tidal influence, the pollutant concentrations also vary during the year and as a function of place – Figure 4.11.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 146 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Plot calculated heavy metal concentrations at different points within the Western Scheldt

40 Zn - eastern Zn - middle Zn - western 30 concentration (ug/l)

000 000 000 000 001 /2 /2 1 4 2/05/2000 1/08/2000 1/11/2000 01/0 31/01/2000 02/03/2000 01/0 0 01/06/2000 02/07/2 0 01/09/2000 01/10/2 0 01/12/2000 01/01/2

Figure 4.11: Example of zinc during the year and as function of the place Eastern part = Schaar van Ouden Doel, near Antwerp Middle part = Terneuzen Western part = Vlissingen, near North Sea

The total dissolved concentration of pollutants in the Western Scheldt can be plotted as a concentration contour map at different time intervals, as shown in Figures 4.12 to 4.15.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 147

Figure 4.12: Zink concentrations on 14-02-2000 (spring peak) Figure 4.13: Zink concentrations on 11-10-2000 (autumn low)

Figure 4.14: Copper conc. on 14-02-2000 (spring peak) Figure 4.15: Copper conc. on 11-10-2000 (autumn low)

Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

4.7.3.2 Ecotoxicological stress

The ms-PAF has been calculated based on the water quality for an area in the South-West part of the Western Scheldt, the nature protection zone ‘Verdronken Land van Saeftinghe’ – Figure 4.16.

Figure 4.16: Verdronken Land van Saeftinghe’

Figure 4.17 shows the ms-PAF which has been calculated with OMEGA based on the dissolved concentrations of Cd, Zn, Cu, (α-β-γ) HCH, HCB and naphtalene.

model point 'Land van Saeftinghe'

1.00 35

0.90 30 0.80

0.70 25

0.60 20

0.50

15 Zn Cu, 0.40 conc (ug/l): (ug/l): conc (ug/l): conc Flu, HCB, HCH, Cd Flu, 0.30 10

0.20 5 0.10

0.00 0

0 0 0 0 0 00 00 0 00 0 /200 /2 /200 6 /04/2 /06/2000 /07/2000 /08 /09/2 /10/2000 11 2 5 01/04/2000 2 13/05/200 03 24/0 15 0 26/08/2000 16 07/10/200 28 18/ Flu HCB HCH Cd Cu Zn Figure 4.17: Dissolved concentrations for reference point ‘Verdronken Land van Saeftinghe’

The ms-PAF is based on observed chronic stress. No additional relations (like QSAR or acute stress data) were used to fill missing data gaps.

The ms-PAF for chronic exposure was calculated for each of the specie groups: Phytoplankton, Macrophytes, Phytobenthos, Benthic Invertebrates, Fish and the Total ms-PAF – Figures 4.18 and 4.19.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 149 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

ms-PAF august 2000

12.0% 10.0%

8.0% chronic stress 6.0% acute stress ms-PAF 4.0%

2.0%

0.0%

T F B P M _ _ _ _ _ F F F F F A A A A A sP sP sP sP sP m m m m m

Figure 4.18: Comparison of chronic versus acute stress, August 2000 [msPAF_T = Total; msPAF_F = Fish; msPAF_B = Benthic; msPAF_P = Phyto Plankton; msPAF_M = Macrophytes]

Chronic ms-PAF results 'Land van Saeftinghe'

16.0%

14.0%

12.0% msPAF_T_C Total

10.0% msPAF_F_C Fish

msPAF_B_C benthic invert. 8.0% msPAF msPAF_P_C Phyto plankton

6.0% msPAF_M_C Macrophytes

4.0%

2.0%

0.0% Apr-00 May-00 Jun-00 Jul-00 Aug-00 Sep-00 Oct-00 Nov-00 Dec-00

Figure 4.19: ms-PAF for chronic exposure for different groups of organisms and total-ms PAF, location Land van Saeftinghe

Toxic stress is higher for phytoplankton and benthic invertebrates (an average ms-PAF of 10%), while for fish the ms-PAF is a bit lower (around 7%). For macrophytes insufficient data is present. The overall ms-PAF for chronic exposure is around 8%. A check of the ms-PAF for acute stress in a summer month (August 2000), yields that the acute stress level is lower then the chronic stress level.

The ms-PAF can be split to the contribution for each single toxicant. For this, the chronic total ms- PAF has been used. The break-down on the contribution of single toxicants has been done on 01-04- 2006 and has been repeated for every two months – Figure 4.20.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 150 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

% of mode of action on msPAF (total) % of mode of action on msPAF (total) 01-04 2000 01-06 2000

Cd % of msPAF Cd % of msPAF

Cu % of msPAF Cu % of msPAF

Zn % of msPAF Zn % of msPAF

Flu+HCB % of msPAF Flu+HCB % of msPAF

HCH % of msPAF HCH % of msPAF

% of mode of action on msPAF (total) % of mode of action on msPAF (total) 01-08 2000 01-10 2000

Cd % of msPAF Cd % of msPAF

Cu % of msPAF Cu % of msPAF

Zn % of msPAF Zn % of msPAF

Flu+HCB % of msPAF Flu+HCB % of msPAF

HCH % of msPAF HCH % of msPAF

% of mode of action on msPAF (total) 01-12 2000

Cd % of msPAF

Cu % of msPAF

Zn % of msPAF

Flu+HCB % of msPAF

HCH % of msPAF

Figure 4.20: Contribution of single toxicants to the ms-PAF for chronic exposure at different moments in time, location Land van Saeftinghe

The main contribution to the ms-PAF comes from copper (80% to 90%), followed by zinc (4% to 14%).

4.7.4 Conclusions on the Western Scheldt case study

The OMEGA model was used successfully and applied to calculate the mixture toxicity of pollutants in the Western Scheldt. The OMEGA model yielded a Potential Affected Fraction (ms-PAF) on a chronic exposure level of around 9% for the study area (Land van Saeftinghe). The ms-PAF varied between 11% in the winter to 7% in the summer. The main pollutant causing the ms-PAF is copper (80-90%).

Translating this outcome to risks, 9% of the species present in the study area might have an effect caused by toxicants. Since the definition of effect on a chronic level is based on NOEC (No Observed Effect Concentration) experiments, the toxic effect might vary from growth inhabitation to a higher mortality.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 151 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

The ms-PAF level of 9% (on average) might have an impact on the ecological status of the study area. For validation, the ms-PAF for the study area has been compared with the ecological status for the ecotopes (2 in total) within the study area. The ecological status is based on the index for the diversity, total macrofauna density and total macrofauna biomass. Only the biomass index for one of the two ecotopes has a bad ecological status, all other indexes meet the MEP (Maximal Ecological Potential) status. Therefore the evidence that there is a toxic effect in the studied area, based on a ms-PAF of around 9%, is inconclusive at the moment.

4.8 Middelburg Case Study: Spreading of polluted sediments due to flooding

4.8.1 Introduction

During flood events, huge amounts of sediment are transported to the inundated area. The irregular supply of thousands of tons of sediment may shape the inundated area. Due to the occasional sedimentation of massive amounts of sediments this situation is not only from the morphology point of view undesired, but also from the water quality point of view. The excess of sedimentation during flood events carries with it significantly high levels of pollutant fractions absorbed in the sediment. Moreover, the transport and distribution of pollutants from storage facilities or industries (calamity events) is another risk during flood events. High concentrations of pollutants offer high ecotoxicological risks not only for plants and animals, but also for humans.

WL | Delft Hydraulics has developed a water quality model framework Sobek 1D/2D for studying the pollutant distribution and predicting the dissolved concentration at different locations and at different stages in a flood plain. The stages can be inundation versus dry top layer or running water versus stagnant water. This is done in conjunction with the hydraulic models used in FLOODsite Sub-Theme 1.2 (flood inundation modelling). These result in a first approach (fundamental research) with a practical spin off for knowledge rules suitable for FLOODsite to be implemented in Sobek 1D2D.

The case study was carried out for Middelburg and the surroundings, in the province of Zeeland, based on the collapse of dikes and the inundation of the land (see Figure 4.21). The study area considers a surface area of about 199 km2 with elevations that go from 1.2 m below see level in the interior to higher than 13 m above sea level in the dike area close to the coast.

Middelburg

Figure 4.21: Overview of the study area Middelburg

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 152 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

4.8.2 Methods and Materials

In this paragraph the water quality model framework Sobek 1D/2D is presented. With this model the sediment transport as well as the distribution of pollutants during a flood event was studied. The model enhances the insight in the spatial distribution of sand, silt and clay on the inundated area. With regard to the pollutants, this study focuses on the heavy metals. The heavy metals cadmium (Cd), copper (Cu), and zinc (Zn) are the dominant pollutants in the Western Scheldt (Wijdeveld, 2006). A better understanding of the sediment deposition and the metal fractions absorbed in the sediment during flood events is of great use for the evaluation of measures that should avoid excess polluted sedimentation.

4.8.2.1 The hydrodynamic model 4.8.2.2 Schematisation The hydrodynamic model regards the northeast section of the hydrodynamic model for the Western Scheldt that has been used in the simulation of dam breaks and flooding (Sub-Theme 1.2: flood inundation modelling). The water quantity model has been setup in the 1D Channel Flow and the 2D Overland module of Sobek. Sobek is software set for the hydrodynamics and the water quality simulation in undimensional channels and flood plains (two dimensions).

The 1D model simulates the Channel over Walcheren. The 2D schematisation covers the whole study area (Middelburg and the surroundings) and has a 300*300 m2 grid. This model considers an elevation map, as well as a friction map. The friction values for the roughness are between 0.1 and 10 s/m1/3. For more information about elevation and friction maps refer to the report on flood inundation modelling from Task 8 of FLOODsite report of flooding inundation modelling.

The inundated area is about 50 km2, one quarter of the total area of study. This is shown in Figure 4.22. The simulation period for the hydrodynamic period is from January 4th to January 29th 2010. The dam break is simulated to happen on January 6th at 00:00. In order to study the pollutant distribution at different stages in the 2D flood plain an adaptation is done to the hydrodynamic model. From January 15th it is assumed that the dam is somehow repaired and that during the next 15 days (until January 30th) no more water is inundating the area. By doing this it is possible to study the stage ‘inundation versus dry top layer’ (or ‘running water versus stagnant water’).

Figure 4.22: Elevations in the study area (right). Inundated area (left)

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 153 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

4.8.2.3 Inflow and outflow The inflow water level and the outflow discharge are simulated as boundary conditions of the 1D model. The water level of the inflow 1D boundary is given in Figure 4.23. From January 15th the water level is set at 0 m above see level, which is a lower level than the elevation of the cell where the water flows into the 2D model.

The outflow discharge is set on 1 m3/s until January 14th at 24:00. For the following period (until January 30th) the outflow discharge is assumed to be zero.

Figure 4.23: Water level at the inflow boundary of the 1D model

4.8.2.4 The water quality model The level of detail and the number of substances added in the model depend on the water quality problem to be investigated and of course of the behaviour of the water system. The problems that are studied in this simulation refer to the transport of sediments and to the dispersion of heavy metals in sediment and in the water column.

The water quality model has been setup in the 1D2D Water Quality module of Sobek. The water quality model uses the process library Delwaq as the central core. The process library Delwaq contains different process equations with a large range of substances and problems of water quality. The Sobek version that has been used corresponds to 2.11.000.16. It is recommended to consider the installation of newer versions of Sobek in the future because the module Sobek 2D Water Quality is still in a development phase.

The 1D Water Quality model and the 2D Water Quality module can be used simultaneously. In this way the exchange of sediments as well as the water quality between channels and inundated areas can be simulated. The hydrodynamics are computed jointly for 1D and 2D in one integrated calculation. For water quality, the calculation procedure is different: the 1D domain and the 2D domain are calculated separately as two different domains. The two Delwaq simulations exchange information about the 1D-2D connections during every time step. The simulation procedure is shown in Figure 4.24; the 1D and 2D Water Quality calculations are running simultaneously.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 154 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

2D Water Quality Delwaq Overland Flow Sobeksim Channel Flow 1D Water Quality Delwaq

Figure 4.24: Configuration of the Sobek modules that are used for the sediment transport calculations.

4.8.2.5 Suspended and in the sediment solids The water quality model considers three fractions of sediment: sand (IM1), silt (IM2) and clay (IM3). Each fraction has its own sedimentation characteristics. The sediment is defined as one layer (S1).

The sediment transport is modelled with a re-suspension - sedimentation approach. The bed shear stress is the determining parameter for re-suspension and sedimentation. The bed shear stress is a function of the flow velocity, the roughness of the river bed and the water depth. At low bed shear stresses, sedimentation of suspended particles prevails. At intermediate bed shear stresses, sedimentation and re-suspension are in a state of equilibrium. When the flow velocity and bed shear stress exceed a certain threshold level, the sediment layer goes into re-suspension, see also Figure 4.25.

The sedimentation velocity and the critical shear stress for sedimentation are different for each suspended solids fraction. The re-suspension velocity and the critical shear stress for re-suspension, on the other hand, apply to all three sediment fractions. These three sediment fractions are assumed to be mixed within the sediment layer. The re-suspension flux of each sediment fraction depends on the relative amount of the particular fraction in the sediment layer.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 155 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

sediment at ion equilibr ium r esuspension

τ crit, sed τ crit, res

flow velocity

Figure 4.25: Re-suspension –sedimentation in the sediment model.

The sediment model has been set up by use of the processes library Delwaq. The sediment model is available as a pre-defined subset, so the sediment model can be coupled easily to other model schematisations. The re-suspension of the sediments is controlled by the tangential tension at the bottom caused by the flow and by the waves created by the wind. However in this model the influence of the wind is not considered.

4.8.2.6 Heavy metals In the water quality model three heavy metals are considered: cadmium (Cd), copper (Cu) and zinc (Zn). The heavy metals are absorbed to the suspended solids and to the particles of the sediments. The affinity of the heavy metal for a sediment fraction in particular can be different and each affinity can be specified in the water quality model. Moreover, it is possible to adjust the adsorption coefficient for each metal in the water and in the sediments. Figure 4.26 and Table 4.1show the processes in the model for copper. The processes are similar for cadmium and zinc.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 156 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 4.26: Model for copper in the water and in the sediment.

Table 4.1: General relation of the processes in the heavy metal model for copper.

Process Description PartS1_Cu Division Cu en S1 PartS2_Cu Division Cu en S2 PartWK_Cu Division Cu in the water column Res_Cu Re-suspension adsorbed Cu Sed_Cd Sedimentation Cd

4.8.2.7 Emission and calamity modelling During flooding events an important risk is the transport of pollutants from urban and rural areas. Storage facilities, industries, pump stations, farms located in the flood area then become sources of pollutants, as well as the pollutants accumulated in the (agricultural and urban) ground. In the water quality model these emission and calamity locations are modelled, as a first approach, with 2D boundary nodes. It is assumed that 6 hours after the cell has been inundated the pollutants are released with a flow of 1 m3/s. In order to know when the grid cell (where the risk location is situated) will be inundated, a first run of the hydrodynamic model is done for the entire period.

Figure 4.27 shows the locations of the risk groups with dangerous substances in Middelburg and surroundings (www.risicokaart.nl). This group is defined in the map as the companies that store, produce or work with dangerous substances. This map also shows the transport routes of dangerous substances through water (in blue) and land (in red), as well as locations with the risk of being inundated. The dangerous substances in the study area belong to three groups: LPG-tank stations, propane gas and companies/factories with dangerous substances (VROM, 2006). In this case it refers to the factory Dutch Cleaning Mill bv, which is a factory that cleans all kinds of beans, nuts and seeds by means of riddles and sieves.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 157 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 4.27: Map of risks for Middelburg and surroundings (www.risikokaart.nl)

For modelling the metals accumulated in agricultural and urban grounds, emission factors are used. The yearly national emission factors are found in RIZA (1996). Through the yearly emission factors (e.g. kg/year-house or kg/year-m2 industrial area) and the variables that defined the emission (e.g. number of houses or surface of industrial areas), gross heavy metal emissions (kg/year) can be estimated. The variables that defined the emission are found on the website of the Dutch Central Office of Statistics CBS (www.cbs.nl). The emissions of heavy metals (not including the calamity risks at the pump station) are amplified by a factor of 100, in order to simulate the accumulation of the pollutants in the ground – Figure 4.28.

Lake Veerse D3 D4 D2

R3 R4

B R2 Middelburg A R1 D1

Figure 4.28: Location of sources of heavy metal emissions. D-locations are considered urban locations, R-locations are agricultural locations, A and B are industries.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 158 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

4.8.2.8 Assumptions and limitations in the water quality model The water quality model has some limitations, which are summarised briefly below: • Bed load as such is not taken into account. However, a sediment fraction with a high density and a high sedimentation velocity can act as a sand fraction in the model. • The accumulation of sediment at a certain location in the model does not affect the bathymetry of the hydrodynamic model. In a river bed, a significant increase of the bed level causes an increasing flow velocity, thus preventing any further sedimentation. • As a first approach, the impact of wind induced waves is not taken into account in the model. Wind induced waves may be of importance. The extra shear stress may avoid the sedimentation of suspended solids or even cause re-suspension of the sediment. • The simulation of emissions through a 2D boundary node is not optimal, since information is missing in the three other directions where the water is not flowing into the 2D plain. • Complete information about the heavy metal emissions is not available. In order to make an estimation of the emissions many assumptions had to be made (see previous paragraph). The same is true for the calamity events. What is important to see here is the difference between an urban area from a rural area, as well as the spatial distribution of the sources.

4.8.2.9 Data for the water quality model 4.8.2.10 Initial conditions The simulation starts on January 5th of the year 2010, at 00:00 a.m. The simulation begins about one day before the dam break and when the first sediments are transported into the modelled area. Initially, suspended solids are not present in the inundated area (2D model). The sediment layer contains no sediments at the start of the simulation for the 1D model, while the sediment fractions in the 2D model are defined locally. This is done by doing a run of the water quality model only with suspended solids and the sediment. The sediment results in the 2D model of this run are used as initial conditions. The initial conditions are summarised in Table 4.2.

Table 4.2: Initial conditions in the water quality model.

substance description initial condition Units 1D 2D Cd cadmium in water 0.0002 0 Mg/l Cu copper in water 0.008 0 Mg/l Zn zinc in water 0.014 0 Mg/l Continuity 1 1 (-) IM1 sand in water 0 0 Mg/l IM2 silt in water 5 0 Mg/l IM3 clay in water 25 0 Mg/l IM1S1 sand in sediment 0 Local g/m2 IM2S1 silt in sediment 0 Local g/m2 IM3S1 clay in sediment 0 Local g/m2

4.8.2.11 Boundary conditions The inflow of sediments from the channel and through the broken dam are sources of metals and sediments in the model. The boundary in the Western Scheldt represents an inflow (Vlissingen keersluisbrug), while the boundary at Channel door Walcheren close to the lake Veerse represents an outflow (Veere schutsluis). Table 4.3 summarises the boundary conditions for the metals and Table 4.4 for the sediments in the model. Table 4.5 shows the total amount of sediment that enters the system.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 159 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 4.3: Boundary conditions for the heavy metals during the January 2010 event.

From Until Cd Cu Zn day Time Day time (µg/l) (µg/l) (µg/l) 05-01-2010 00:00 15-01-2010 00:00 0.19 8 14.4 inflow 1D Vlissingen keersluisbrug 05-01-2010 00:00 15-01-2010 00:00 0.09 12.1 12.8 outflow 1D Veere schutsluis

Table 4.4: Boundary conditions for suspended solids during the January 2010 event.

From Until sand – IM1 silt – IM2 clay – IM3 day time Day time (mg/l) (mg/l) (mg/l) 05-01-2010 00:00 15-01-2010 00:00 28 15094 15146 inflow 1D 05-01-2010 00:00 15-01-2010 00:00 6 3019 3029 outflow 1D

Table 4.5: Total amount of sediment in the model.

inflow sand Silt clay total percentage (ton) (ton) (ton) (ton) inflow 1D 15.22 8206.73 8235.00 16456.95 146 outflow 1D 5.18 2608.42 2617.06 5230.66 46 Total 10.04 5598.31 5617.95 11226.30

4.8.2.12 Emissions The gross yearly emissions of cadmium, copper and zinc in Middelburg are given in table Table 4.6. Because no data were found on the overflow of metals in the study area, the values were taken over from a study for the area Hunze en Aa’s (De Straat Milieu-adviseurs B.V., 2003). The emissions from built-up areas are calculated proportional to the emissions in Middelburg base on the surface between the built-up area and Middelburg (see Table 4.7)

Table 4.6: Gross yearly emissions (kg) of Cadmium, Copper and Zinc in Middelburg

Source Cd Cu Zn Corrosion roof houses (kg) - - 9.94E+02 Corrosion roof companies (kg) - - 3.77E+00 Corrosion zinc street ornaments (kg) - - 9.21E+01 Wear tires (kg) 1.28E-02 5.24E-05 2.66E-04 Overflow (Hunze en Aa's) (kg) 2.56E-04 4.42E-02 5.58E-02 1.30E- 4.43E- 1.09E+03 Load Middelburg Total (kg) 02 02

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 160 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 4.7: Gross yearly emissions (kg) of Cadmium, Copper and Zinc in five built-up areas

Built-up areas Fraction area with regard to Middelburg Cd Cu Zn D1 0.075 9.67E-01 3.97E-03 8.26E+04 D2 0.038 4.83E-01 1.98E-03 4.13E+04 D3 0.038 4.83E-01 1.98E-03 4.13E+04 D4 0.015 1.93E-01 7.93E-04 1.65E+04 D5 0.023 2.90E-01 1.19E-03 2.48E+04

The sources of metals considered in the rural areas are the wear of copper rails, oil leaks, pig manure and artificial manure. In the case of the wear of copper rails and oil leak, each rural area is assumed to emit 25% of load of heavy metals from both sources in the country side. This is summarized in Table 4.8.

Table 4.8: Gross yearly emissions (g) of heavy metals from the wear of copper rails and oil leaks

Fraction of total Rural areas emissions Cd (g) Cu (g) Zn (g) R1, R2, R3, R4 0.25 3.50E-01 3.93E+03 2.26E+02

Pig manure is a source of copper. According to literature, values of about 15 mg of copper are contained in each kg of pig manure (Foundation Biologic Pig Farming, 2003). The pig manure produce in Middelburg in 2005 was about 37231000 kg (CBS data). It is assumed that the load of copper due to this source is equally divided in the four rural areas. This is summarized in Table 4.9.

On the website www.kunstmest.nl information can be found about artificial manure. About 0.4 to 1.6 g Cd can be found per hectare. For this study a factor of 1 g Cd/hectare is assumed; this is summarized in Table 4.9.

Table 4.9: Cadmium and copper emissions from artificial and pig manure

Cd-Artificial manure Area (ha) Cu-Pig manure (g) Rural areas (g) R1 657 657 139616 R2 1440 1440 139616 R3 1683 1683 139616 R4 1134 1134 139616

The modelling of calamities focuses on the risks around the pump station and industries. Based on the emission factors defined by RIZA (1996), the fraction between metal and oil is estimated. The volume of the tank in a pump station is assumed to be 150000 litters. The density of oil is defined at 895 kg/m3. The loads of the metals are presented in Table 4.10. The resulting loads are added to rural area 2.

Table 4.10: Load of metals coming from a pump station in a calamity

Fraction Metal/Oil Mass (g) Cadmium 1.28E-06 172 Copper 3.37E-05 4530 Zinc 8.24E-04 111000

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 161 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

In the case of the industries, yearly emissions are taken from the results of the study in the area Hunze en Aa’s as a reference (De Straat Milieu-adviseurs B.V., 2003). This is shown in Table 4.11.

Table 4.11: Yearly loads (g) of heavy metals from industries

Industry Cd (g) Cu (g) Zn (g) A 1.00E+01 3.50E+03 1.45E+04 B 1.00E+01 3.50E+03 1.45E+04

The emissions are modelled in the 2D model as boundary nodes. This means that the loads have to be converted to concentrations. In order to estimate the concentrations of heavy metals from the sources here presented, the surface and the water depth of the respective inundated areas (urban and rural) are estimated based on the results of the hydrodynamic model. The volumes (m3) of the inundated areas are presented in Table 4.12. The final concentrations (g/m3) given in Table 4.13 and added to the model are a factor of 100 larger.

Table 4.12: Calculated volumes (m3) of the inundated areas. The water depth in the 2D plane is calculated with the hydrodynamic model. The area is estimated according to the grid.

Inundated areas m3 Middelburg 2806200 D1 18000 D2 1800 D3 54000 D4 108000 R1 816750 R2 2269350 R3 2841300 R4 1962900 Industry area A 701550 Industry area B 701550

Table 4.13: Total concentrations of heavy metals (g/m3) at the different areas (100 times amplificatied).

Cd Cu Zn Middelburg 9.58E-03 1.58E-00 4.08E+03 D1 (g/m3) 5.37E-03 2.20E-05 4.59E+02 D2 (g/m3) 2.69E-02 1.10E-04 2.29E+03 D3 (g/m3) 3.58E-04 1.47E-06 3.06E+01 D4 (g/m3) 2.69E-04 1.10E-06 2.29E+01 A (g/m3) 1.43E-03 4.99E-01 2.07E-00 B (g/m3) 1.43E-03 4.99E-01 2.07E-00 R1 (g/m3) 2.01E-02 4.39E-00 6.90E-03 R2 (g/m3) 1.59E-02 1.58E+00 1.47E-02 R3 (g/m3) 1.48E-02 1.26E-00 1.98E-03 R4 (g/m3) 1.44E-02 1.83E-00 2.87E-03

4.8.2.13 Process coefficients The number of process coefficients in the sediment model is limited. For each sediment fraction the density of the material, the sedimentation velocity and the critical shear stress must be defined. The re- suspension velocity and the critical shear stress for re-suspension are two coefficients that apply to all sediment fractions. The porosity applies to the entire sediment layer. The model assumes that the three sediment fractions are completely mixed within the sediment layer. Table 4.14 gives the Kd

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 162 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 coefficients of metals for the three sediment fractions. Table 4.15 shows the values of the process coefficients used in the model.

Table 4.14: Kd coefficients of metals for sand, silt and clay

IM1 IM2 IM3 Metal Default 10% 30% 60% Sand Silt Clay Cu 50 8 25 50 Zn 110 18 55 110 Cd 130 22 65 130

As a first approach, the process coefficients for the sedimentation of sand, silt and clay are the same as the default values. The density of the sediment particles and the porosity of the sediment layer determine the simulated thickness of the sediment layer.

Table 4.15: Process coefficients in the sediment model, with default (non-calibrated) coefficient values.

coefficient Name Default value Used value units VSedIM1 sedimentation velocity IM1 (sand) 24 30 m/day VsedIM2 sedimentation velocity IM2 (silt) 4 7 m/day VsedIM3 sedimentation velocity IM3 (clay) 0.01 0.3 m/day TauScIM1 critical shear stress sedimentation IM1 0.05 0.07 N/m2 TauScIM2 critical shear stress sedimentation IM2 0.01 0.05 N/m2 TauScIM3 critical shear stress sedimentation IM3 0.005 0.02 N/m2 VResDM first order resuspension velocity 0.01 0.01 1/day TauRSiDM critical shear stress resuspension 0.2 0.2 N/m2 RHOIM1 density IM1 1.3 106 1.3 106 g/m3 RHOIM2 density IM2 1.3 106 1.3 106 g/m3 RHOIM3 density IM3 1.3 106 1.3 106 g/m3 PORS1 porosity sediment layer 1 0.50 0.5 - MinDepth minimum water depth for sedimentation 0.01 0.01 m KdCdIM1 partition coefficient Cd-IM1 130 78 m3/kgDM KdCdIM2 partition coefficient Cd-IM2 130 65 m3/kgDM KdCdIM3 partition coefficient Cd-IM3 130 130 m3/kgDM KdCdIM1S1 partition coefficient Cd-IM1 in layer S1 130 22 m3/kgDM KdCdIM2S1 partition coefficient Cd-IM2 in layer S1 130 65 m3/kgDM KdCdIM3S1 partition coefficient Cd-IM3 in layer S1 130 130 m3/kgDM KdZnIM1 partition coefficient Zn-IM1 110 18 m3/kgDM KdZnIM2 partition coefficient Zn-IM2 110 55 m3/kgDM KdZnIM3 partition coefficient Zn-IM3 110 110 m3/kgDM KdZnIM1S1 partition coefficient Zn-IM1 in layer S1 110 18 m3/kgDM KdZnIM2S1 partition coefficient Zn-IM2 in layer S1 110 55 m3/kgDM KdZnIM3S1 partition coefficient zn-IM3 in layer S1 110 110 m3/kgDM KdCuIM1 partition coefficient Cu-IM1 50 8 m3/kgDM KdCuIM2 partition coefficient Cu-IM2 50 25 m3/kgDM KdCuIM3 partition coefficient Cu-IM3 50 50 m3/kgDM KdCuIM1S1 partition coefficient Cu-IM1 in layer S1 50 8 m3/kgDM KdCuIM2S1 partition coefficient Cu-IM2 in layer S1 50 25 m3/kgDM KdCuIM3S1 partition coefficient Cu-IM3 in layer S1 50 50 m3/kgDM

4.8.2.14 Simulation period The hydrodynamic model starts at January 4th 2010. The dam break occurs at January 6th at 00:00. The inflow of sediments at the 1D boundaries of the model lasts from January 5th until January 15th. At this moment it is assumed that the dam somehow is repaired. The sedimentation of suspended solids continues for some time until January 29th, after the inflow of material has ended. The

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 163 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 sedimentation velocity of the finest sediment fraction is very low and sedimentation may continue several days or weeks after the transport of sediments to the lake has ended. January 30th is the end date of the flow simulation.

4.8.3 Simulation results

In this chapter the results are given for the water quality model in terms of the solids and the heavy metals. They refer mainly to the concentrations in the sediment layer S1. The results of the concentration of the substances in the water column are not presented in this report in order to give emphasis to the deposition of the sediments and the pollutants absorbed on to them after the break of a dam.

The results also give the pollutant distribution at different stages in the 2D flood plain. The first stage is the dam break until January 15th. From January 15th it is assumed that the dam is some how repaired and during the next 15 days no more water inundates the area. By doing this it is possible to study the stage ‘inundation versus dry top layer’ or ‘running water versus stagnant water’.

4.8.3.1 Results of the solids in the sediment The results of the sediment model consider the mass of the three fractions of sediments: sand, silt and clay. A main factor here is the tangential tension. The results of the tangential tension are given in Figure 4.29.

Figure 4.29: Tangential tension results on January 15th and January 29th.

The sediment balance is shown in Table 3.5. The sediment balance was made for the period January 5th until January 29th. Figure 4.30 to 4.32 show the spatial distribution of sediment mass (g) in the inundated area for sand, silt and clay respectively.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 164 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 4.30: Amount of freshly deposited sand (g) on January 15th and January 29th. The sedimentation velocity of sand is 30 m/d.

Figure 4.31: Amount of freshly deposited silt (g) on January 15th and January 29th. The sedimentation velocity of silt is 7 m/d.

Figure 4.32: Amount of freshly deposited clay (g) on January 15th and January 29th. The sedimentation velocity of clay is 0.3 m/d.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 165 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

The mass balance for the three sediment fractions in the 2D water quality model is given in Table 4.16. Both stages are presented in the table: January 15th (running water) and January 29th (stagnant water).

Table 4.16: Mass balance for the sediment in the 2D model for both stages (January 15th and January 29th)

January 15th IM1 IM2 IM3 IM1S1 IM2S1 IM3S1 Total mass in system 1.8526E+08 2.8645E+11 5.1301E+11 7.2050E+08 2.3241E+11 2.7076E+10 Changes by processes -5.1890E+05 -2.0252E+08 -4.3489E+07 5.1890E+05 2.0252E+08 4.3489E+07 Boundary inflows 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 Boundary outflows 5.8544E-04 1.3977E+04 1.5808E+04 0.0000E+00 0.0000E+00 0.0000E+00

January 29th IM1 IM2 IM3 IM1S1 IM2S1 IM3S1 Total mass in system 4.7501E+04 1.9659E+08 9.6737E+10 9.0571E+08 5.1866E+11 4.4341E+11 Changes by processes -1.5720E+02 -2.7319E+06 -4.2389E+08 1.5720E+02 2.7319E+06 4.2389E+08 Boundary inflows 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 Boundary outflows 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00

The simulation results are quite different for sand, silt and clay, due to their differing properties. Table 4.17 summarizes the distribution of the solids in the water and in the sediment. The sand is deposited almost immediately, near the outflow of the broken dam and also close to the cells where the flow is overtopping the embankments. The small amount of sand that is still in suspension after almost one month is trapped in the water in the areas with a low water depth. Sedimentation stops when the water depth is lower than a minimum water depth of 1 centimetre and the sand remains in suspension.

About 45% of the silt is deposited in the inundated area on January 15th while on January 29th all of the silt is settled down.

Due to the low sedimentation velocity (0.3 m/day), most of the clay remains in suspension for a long time. On January 15th 95% of the clay remains in suspension and on January 29th the percentage goes down to 18%. Furthermore, the clay sediments only on locations with low flow velocities.

Table 4.17: Distribution (percentage) of the three sediment fractions during both stages.

IM1 IM2 IM3 IM1S1 IM2S1 IM3S1 January 15th 20 55 95 80 45 5 January 29th 0 0 18 100 100 82

The spatial distribution of sediment depends on:

• The spatial distribution of the water that is loaded with sediments; • The time scale of sedimentation versus the time scale of hydrodynamics.

The exact location where clay, silt and sand will settle depends on the sedimentation velocity of the particles and the flow velocity of the water.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 166 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

The thickness of the sediment layer and the volume of the sediment in the lake and the rice fields can be computed when the porosity and the density of the sediment is known. Figure 4.33shows the thickness of the newly formed sediment layer. The density of the three sediment fractions is assumed to be 1.3 ton/m3, the porosity of the sediment is 0.50.

Figure 4.33: The thickness (m) of the freshly formed sediment layer on January 15th and January 29th.

4.8.3.2 Results of the heavy metals in the sediment The results of the sediment model consider the mass of the heavy metals cadmium, copper and zinc absorbed in the sediment. Figure 4.34 to 4.36 show the spatial distribution of sediment mass (g) in the inundated area for cadmium, copper and zinc respectively. The spatial distribution of the heavy metals is quite similar. This is due to the fact that heavy metals are mainly absorbed in the organic matter (clay). Therefore where high concentrations of clay are distributed, also high concentrations of metals are presented.

Figure 4.34: Amount of freshly deposited cadmium (g) on January 15th and January 29th.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 167 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 4.35: Amount of freshly deposited copper (g) on January 15th and January 29th.

Figure 4.36: Amount of freshly deposited zinc (g) on January 15th and January 29th. The mass balance for the heavy metals in the 2D water quality model is given in Table 4.18. Both stages are presented in the table: January 15th (running water) and January 29th (stagnant water). Table 4.18 summarizes the distribution of cadmium, copper and zinc in the water and in the sediment.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 168 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Table 4.18: Mass balance for the heavy metals in the 2D model for both stages (January 15th and January 29th)

January 15th Cd Cu Zn CdS1 CuS1 ZnS1 Total mass in system 5.9265E+03 2.6581E+05 1.6250E+07 2.1377E+03 1.1019E+05 4.5281E+07 Changes by processes -5.2134E+00 -6.2356E+01 -3.2151E+05 5.2134E+00 6.2356E+01 3.2151E+05 Boundary inflows 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 Boundary outflows 1.9084E-04 8.0363E-03 1.4465E-02 0.0000E+00 0.0000E+00 0.0000E+00 January 29th Cd Cu Zn CdS1 CuS1 ZnS1 Total mass in system 8.4739E+02 3.6778E+04 1.6257E+05 7.2460E+03 3.3941E+05 6.1366E+07 Changes by processes -3.7395E+00 -1.6230E+02 -9.5853E+02 3.7395E+00 1.6230E+02 9.5853E+02 Boundary inflows 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 Boundary outflows 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00 0.0000E+00

Table 4.19: Distribution (percentage) of the heavy metals during both stages.

Cd Cu Zn CdS1 CuS1 ZnS1 January 15th 73 70 26 27 30 74 January 29th 10 10 0.3 90 90 99.7

4.8.4 Recommendations

• The present grid is 300*300 m2. In order to have a better description of the distribution of the heavy metals and sediments in the inundated area, the grid can be refined to 100*100 m2. • The suspended solids concentrations at the boundaries of the model are constant in the present model schematisation. In reality, the suspended solids concentrations are most likely a function of the discharge and therefore a function of time. This also goes for the heavy metals. • The effect of the wind has to be included. Wind induced waves may be of importance. The extra shear stress may avoid the sedimentation of suspended solids or even cause re- suspension of the sediment. • Sensitivity analysis. The uncertainty by modelling of a calamity situation is considerable, especially when we want the model to be calibrated. However, a sensitivity analysis may give more understanding of the effects of the flooding in the water quality.

Table 4.20 lists the parameters that can be included in a sensitivity analysis.

Table 4.20: Parameters in the sediment model

parameter effect sedimentation velocity spatial distribution of sediments overall sediment mass balance critical shear stress spatial distribution of sediments overall sediment mass balance dispersion coefficient spatial distribution of sediments Density thickness of sediment layer Porosity thickness of sediment layer inflow of sediments overall sediment balance hydrodynamic simulation spatial distribution of sediments overall sediment mass balance

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 169 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

5 Activity 4: Developing methodological foundations for GIS-based multicriteria evaluation of flood damage and risk (UFZ)

5.1 Background

Flood risk management can be roughly divided into two parts (Schanze 2006): Flood risk analysis & assessment on the one hand and risk mitigation on the other. Broadly speaking, the purpose of flood risk assessment is to establish where risk is unacceptably high, i.e. where mitigation actions would be necessary. Risk mitigation means to propose, evaluate and select measures to alleviate risks in these areas. Currently, the evaluation of alternative measures is mostly done by means of cost-benefit analysis (CBA). In this case, the costs of a certain measure are compared with their benefits in terms of risk reduction. In theory, this procedure leads to an efficient allocation of funds and finally to an optimised protection against flooding.

For both parts, risk assessment and the evaluation of risk mitigation measures, it is required to quantify flood risk as exactly as possible. In this context, three deficits in today’s practice of flood risk management can be identified:

1. Flood risk defined by the formula risk = probability * consequence (Gouldby & Samuels, 2005) comprehends all kinds of consequences of flooding. Nevertheless, current practice of risk assessment and cost-benefit analysis still focuses on damages that can be easily measured in monetary terms. More precisely, risk analysis mainly deals with damage to assets, while social and environmental consequences are often neglected. In consequence, flood risk management often manages only certain parts of flood risk. On that basis, an optimised allocation and design of flood mitigation measures cannot be ensured and is the more unlikely, the more social and environmental risks are spatially separated from economic risks. 2. The spatial distribution of risks as well as of the benefits of flood mitigation measures is rarely considered. E.g. the evaluation and selection of appropriate mitigation measures is mostly based on their overall net benefit. Therefore, it is often not considered which areas benefit most from a measure and which areas do not. This may lead to spatial disparities of flood risk which are not desirable or acceptable. 3. Uncertainties in the results of risk assessment are often ignored. Although sophisticated methods in all parts of risk analysis and assessment have been elaborated over the past decades in order to give a reasonably exact estimation of flood risk, the results of risk assessment are still to some degree uncertain or imprecise. These uncertainties are often not communicated to the decision makers, i.e. a non-existent precision of estimation is pretended. This might facilitate the decision for the decision maker but reduces the scope of decision and could lead to a solution which is not optimal.

The methodological framework presented in this paper tries to provide solutions for these three problems. Hereby, the focus is set on the first point. In this context multicriteria analysis (MCA) is an appropriate method of incorporating all relevant types of consequences without measuring them on one monetary scale. It provides an alternative to the complex monetary evaluation and internalisation of intangible consequences in a cost-benefit analysis.

The second point can be considered by mapping risks and risk reducing effects, respectively. Geographical information systems (GIS) with their ability to handle spatial data are an appropriate tool for processing spatial data on flood risk. In our framework we therefore describe and test approaches which combine MCA with GIS.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 170 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Regarding point three, we will present at least some possibilities of integrating the uncertainties in the results of risk analysis in this GIS-based MCA approach in order to provide good decision support for the responsible decision makers.

This research relates to the European “directive on the assessment and management of flood risk” (EU 2006/C 311 E/02), which requires in article 6 a risk assessment and mapping of social, economic and environmental flood risk. This report delivers an approach on how to deal with these different risk dimensions in an integrated manner.

Therefore, we firstly discuss and develop a methodological framework for spatial MCA for flood risk mapping and secondly apply and test our approach at the FLOODsite pilot site Mulde, a tributary to the Elbe River (Task 21). The risk calculation and mapping of single criteria as well as the multicriteria analysis are supported by a software tool (FloodCalc, Scheuer & Meyer, 2007) which was developed for this task.

5.2 Research Results

The process of MCA can be divided into different steps (based on Munda, 1995): 1. Problem Definition 2. Evaluation Criteria 3. Alternatives 4. Criteria Evaluation / Decision Matrix 5. Criterion Weights 6. Decision Rules 7. Results & Sensitivity

In the following we will briefly explain the different steps and hereby describe the approach we applied at the Mulde pilot site.

5.2.1 Problem Definition At the beginning of any decision making process the problem needs to be recognized and defined. Malczewski (1999) defines the decision problem broadly as “a perceived difference between the desired and existing states of a system”.

With regard to flood risk management the underlying problem can be structured into two parts:

1. Multicriteria risk assessment First of all, the problem is to identify, where the flood risk is too high. Often there is in the beginning only a vague awareness that flood risk might be high. I.e. the current magnitude and spatial distribution of flood risk needs to be identified in order to find out where further mitigation measures are necessary. This multicriteria assessment of different areas is therefore an important prerequisite for step 2 as it is an important part of the problem definition of step 2. 2. Multicriteria project appraisal After identifying high risk areas, the second part of the decision problem is to find the best strategies or measures to reduce flood risk to an appropriate level. These mitigation measures need to be evaluated in order to find the best alternative or combination of alternatives. Hereby, the spatial distribution of these risk reducing effects is rarely considered at present. I.e. in most cases only the overall effects of alternative measures are evaluated. A GIS-based mapping of the effects of each measure may also help to highlight who and where the winners (and perhaps losers) are.

In this report we concentrate mainly on the multicriteria risk assessment and mapping. Nevertheless, our approach can be also used as a basis for the evaluation of risk management measures.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 171 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

5.2.2 Selecting Evaluation Criteria In the second step of MCA the evaluation criteria have to be selected. The inclusion or exclusion of criteria can greatly influence the results of the evaluation process, so it is important that stakeholders and decision makers participate in this selection process. The evaluation criteria should be complete on the one hand to make sure that the whole problem is encompassed, on the other hand the set of criteria should be kept minimal to reduce the complexity of the evaluation process (Keeney & Raiffa, 1993).

The selection of criteria should be based on an examination of relevant literature. Furthermore, experts and stakeholders can be integrated in the selection procedure, e.g. by applying the Delphi Technique. Regarding flood risk analysis the criteria should cover the whole range of economic, social and environmental risks. We gathered examples for different risk criteria from different flood risk MCA studies.

For our multicriteria assessment of flood risks at the Mulde River we apply the following risk criteria: • Economic: - Annual Average Damage • Social: - Annual average affected population - Probability of social hot spots (hospitals, schools etc.) being affected • Environmental: - Erosion potential (of material) - Accumulation potential (of material) - Inundation of oligotrophic biotopes

On the one hand the intention is at least to cover the three main dimensions of flood risk: economic, social and environmental risks, on the other hand this list is kept minimal and simple for reasons of applicability. For a more sophisticated and comprehensive analysis it might be good to extend this set by more criteria and/or to improve the criteria.

5.2.3 Alternatives The next step is to define the alternatives to be compared. As mentioned before, flood risk management deals with two kinds of MCA problems. These both differ considerably concerning the kind of alternatives or options they compare.

1. Multicriteria risk assessment and mapping Multicriteria risk assessment does not really compare different actions or decision alternatives. It is an assessment of different areas regarding their risk status. Hence the alternatives to be compared in this case are different spatial units within the research area. Depending on the underlying spatial data, or the GIS-model chosen, these spatial units to be compared could be grid cells (raster GIS) or points, lines and polygons in a vector GIS. 2. Multicriteria project appraisal The second multicriteria problem deals with the comparison and selection of alternative flood mitigation measures. I.e. the decision problem is to choose among a given set of flood risk management measures ranging from structural measures like dikes and dams to non-structural measures such as land use changes or warning systems.

In our study we focus on 1) multicriteria risk mapping. I.e. in our pilot study the alternatives to be compared are grid cells of 10*10 m, which leads to a relatively high number of alternatives. This restricts the choice of MCA-approaches to methods which are able to deal with such a large number of alternatives.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 172 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

5.2.4 Criteria Evaluation: Risk Maps For each alternative or grid cell the performance of each criterion needs to be evaluated. Regarding GIS-based flood risk assessment, the result is a risk map for each criterion.

Our basis for the assessment of flood risk in the pilot study is the definition of risk expressed by the formula Risk = probability * negative consequence (see e.g. Gouldby & Samuels, 2005).

In other words this is the expected annual average negative consequence of flooding, where “negative consequences” covers economic, social as well as environmental consequences. For the practical application of flood risk assessment this means that the negative consequences have to be evaluated for flood events of different probability. Based on these damage evaluations for different events a damage-probability curve can be constructed (see Figure 5.1). The risk (or the annual average damage) is shown by the area or the integral under the curve.27

Damage

Rtotal D2

1/200 1/100 1/20 1/5 Exceedance Probability Figure 5.1: Damage-probability curve The basis for all our damage evaluations in the Mulde pilot site is inundation data for events of different exceedance probability (1:10, 1:25, 1:50, 1:100, 1:200, 1:500), calculated by a 1D- hydrodynamic modelling by UFZ (Rode & Wenk). For each of these events the inundation area and depth is mapped for a grid with a spatial resolution of a 10m (see Figure 5.2).

Damage is calculated for each of these grid cells, so that a damage map for each of the events mentioned above is produced. By using the risk formula described above, the annual average damage per grid cell can be computed. All computations are carried out by the software tool FloodCalc (Scheuer & Meyer, 2007). It allows the uploading of grid data of inundation depth, value of assets, inhabitants, environmental values and to combine them with different sets of depth/damage function and thereby producing damage and finally risk grids.

Please note that all methods chosen here to estimate the different risk criteria (inundation modelling as well as damage evaluation) are fairly approximate approaches. This was necessary due to time and budget restrictions of the project. This means risk estimations of single raster cells may have high uncertainties. Our results are therefore not appropriate to estimate risks of single properties or to carry out detailed project appraisals for flood protection measures, nor should they challenge official data sources, such as hazard or risk maps which are already at hand for the River Mulde.

27 For criticism on this quantitative definition of risk especially in social science see e.g. Banse & Bechmann (1998)

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 173 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 5.2: Expected inundation depth for a 200-year flood event (City of Grimma) Source: Topographic map: Landesvermessungsamt Sachsen; Inundation data: Calculation by Wenk & Rode 2007

For the economic risk criterion, flood damage for each of the events mentioned above is calculated by means of a meso-scale damage evaluation approach (Meyer, 2005). The general procedure is the following:

1. The total value of assets at risk and its spatial distribution are estimated based on data from official statistics (the net value of fixed assets for different economic sectors) which is then assigned to corresponding land use categories (ATKIS-DLM). 2. Relative depth/damage curves are then used to calculate the damaged share of the values, depending on inundation depth.

Methodological uncertainties in damage evaluation are shown by applying 1) different spatial modelling keys of asset value to land use categories and 2) different sets of depth/damage curves. For each grid cell a mean damage can be calculated based on these different damage values as well as a minimum and maximum damage value. According to the risk formula mentioned above an annual average damage per raster cell is calculated based on the different damage estimations for inundation events of different exceedance probabilities (1:10, 1:25, 1:50, 1:100, 1:200, 1:500). This is conducted for the mean as well as for the minimum and maximum damage estimations so that the final output is a mean, minimum and maximum annual average damage per grid cell, accordingly. The mean annual average damage is shown in Figure 5.3.

In order to assess the environmental risk of an inundation in the Mulde floodplains, three evaluation criteria have been selected (see Table 5.1). A simple yes/no damage function is applied for each criterion, depending on if the area is affected or not. Because the 3 criteria are different in terms of

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 174 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420 their impact functions but may occur simultaneously during one unique flood event, we suggest calculating a sum of the values given for each criterion to estimate a first environmental impact potential of a flood. The list is not complete and has to be amplified.

Figure 5.3: Annual Average Damage (AAD) (City of Grimma): mean estimation Source: Topographic map: Landesvermessungsamt Sachsen; damage estimation: own calculations

Table 5.1: Criteria of environmental risk assessment Indicator / Criterion Potential damage (risk) Explanation / notes yes no Erosion 1 0 where erosion of fine grain material occurs pollutants might be mobilised and transported (pollutants = heavy metals bond to clay minerals and organic matter; nutrients such as Phosphorus) Accumulation 1 0 same as erosion but creation of new polluted sites due to accumulation of the transported material Inundation of 1 0 a longer inundation (>1 hour) of oligotrophic oligotrophic biotopes biotopes (see list below) might negatively affect these biotopes in form of eutrophification or drop of the number of species final assessment ∑ ∑

Analogous to the calculation of economic damage, damage maps for environmental consequences can be produced for each flooding event. Each raster cell can hereby achieve “damage values” between 0- 3. Based on these different damage maps an environmental risk map is calculated by using the risk formula described at the beginning of this section. This risk value can be interpreted as annual average environmental consequence, expressed in the point scale described above. In Figure 5.4 these values are already standardised in values from 0 to 1.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 175 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 5.4: Environmental risk (City of Grimma): standardised values (0-1) Source: Topographic map: Landesvermessungsamt Sachsen; damage estimation: own calculations

The spatial distribution of the affected population is calculated by a meso-scale approach more or less in the same way as the asset values (Meyer 2005): Therefore, the number of inhabitants is taken from official statistics on municipality level and broken down to corresponding land use categories. By intersecting this population density map with the inundation data the number of affected people can be estimated for each event. According to the risk formula, the number of the annual average affected population can be calculated (Figure 5.5).

As “social hot spots” the locations of hospitals, schools, old people’s and children’s homes are identified. For reasons of simplicity we assume that each of the hot spots has the same vulnerability, no matter e.g. the size of a hospital or school or if it is a primary or secondary school. Such differentiations would of course make sense, but they should be conducted by experts in this field for each study. Anyhow, such more detailed information could be easily included in the dataset later on. By intersecting the map with the social hot spots with the inundation maps it can be determined for which inundation scenario the hot spots would be affected. By applying the risk formula an approximate estimation of the probability of being affected can be calculated for each hot spot (Figure 5.6).

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 176 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 5.5: Annual affected population (City of Grimma): Source: Topographic map: Landesvermessungsamt Sachsen; inhabitants: own calculations

Figure 5.6: Social hot spots at risk and their probability of being flooded (City of Grimma): Source: Topographic map: Landesvermessungsamt Sachsen; social hot spot risk: own calculations

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 177 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

5.2.5 Criterion Weights The weighting of the criteria decides what influence a certain criterion has in the aggregation process. This step therefore significantly affects the results of the overall evaluation. Hence, it is one of the parts of the MCA-process where stakeholder and decision maker participation is most crucial. Regarding a multicriteria flood risk assessment, the decision makers have to decide on the relative importance of the different economic, social and environmental risk criteria.

We discuss in our report several different weighting techniques, like ranking and rating approaches, the pairwise comparison approach and the swing weight approach (see also Malczewski, 1999). Our software tool provides the possibility to carry out the point allocation approach, a rating technique where 100 points have to be allocated among the criteria, as well as the swing weight approach. Here, the criterion ranked first is given 100 points and the following criteria receive points according to their relative importance to the preceding criterion.

5.2.6 Decision Rules The decision rule defines the way the unidimensional measurements are aggregated under consideration of the weights given to each criterion to an overall evaluation. It can be therefore considered as the core of MCA.

Several approaches exist which are capable for spatial MCA like e.g. the Analytical Hierarchy Process (AHP), Compromise Programming (CP), the Hasse-diagramm technique, the Disjunctive approach, Multi attribute utility theory approaches (MAUT) or Outranking approaches.28 In our report these approaches are described and their advantages and disadvantages are discussed. We finally select two, the Disjunctive approach and an MAUT approach (simple additive weighting) to be implemented in our software tool and tested for our pilot site at the Mulde.

The general idea of the Disjunctive approach is that the decision maker has to define a threshold level for each criterion. E.g. in order to select areas which have a high risk of flooding, the decision maker has to determine for each risk criterion a critical value which defines the border between low/acceptable risk and high/unacceptable risk. If this threshold value is exceeded in only one of the criteria the area is selected as a high risk area. This simple approach seems to be appropriate e.g. for a quick screening and pre-selection of high risk areas.

The general concept of additive MAUT approaches is to generate a weighted average of the single criterion values for each area (or alternative). Given a set of evaluation criteria and a set of alternatives to be compared as well as scores for each alternative in each criteria and a set of weights for each criterion the procedure for this is the following:

1. Standardise the criteria scores to values (or utilities) between 0 and 1. 2. Calculate the weighted values for each criterion by multiplying the standardised value with its weight. 3. Calculate the overall value (utility) for each alternative by summing the weighted values (utilities) of each criterion. 4. Rank the alternatives according to their aggregate value (utility).

5.2.7 Results & Sensitivity Figure 5.7 shows sample results of the disjunctive approach for some arbitrary chosen threshold values for each of the criteria. All red areas exceed the threshold value in at least one criterion, some areas (in darker red) in more than one criterion.

28 For a further description of these approaches see e.g. Zimmermann & Gutsche (1991); Malczewski (1999); Tkach & Simonovic (1997); Thinh & Vogel 2006); Soerensen et al. (2004); Keeney & Raiffa (1993); Drechsler (1999).

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 178 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 5.7: Example for selected “high risk areas” by the disjunctive approach Source: Topographic map: Landesvermessungsamt Sachsen; risk estimation: own calculations

A sample result of the additive weighting approach is shown in Figure 5.8. In this case a weight of 0.4 is given to both the population as well as the economic risk criterion, because both are often seen by decision makers as the primary protection goals. The environmental and social hot spot criteria are both provided with a weight of 0.1.

For the additive weighting approach we also tested how sensitive the results of ranking procedure are regarding changes or errors in the inputs of the analysis. These changes or errors can concern either the criterion values, i.e. uncertainties in risk assessment, or the weights given to the criteria.

As described above we tried to document the criteria score uncertainty at least during the estimation of the economic risk criterion by calculating a mean, minimum and maximum annual average damage, depending on the spatial modelling of asset values and the set of damage functions chosen. Figures 5.9 and 5.10 show how such a change of the value of one criterion affects the overall results. Both calculations for the minimum and maximum annual average damage value are carried out with the same weighting as in Figure 4.8, where the mean value of annual average damage was used.

Finally the sensitivity of the overall results to the weights given to the criteria is shown by assigning one of them a very large weight (five times higher than the others or 0.625). Figure 5.11 shows for example the results with a large weight given to the environmental criterion.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 179 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 5.8: Standardised multicriteria risk: large weight on economic & population criteria (40% each) Source: Topographic map: Landesvermessungsamt Sachsen; risk estimation: own calculations

Figure 5.9: Standardised multicriteria risk - criteria score sensitivity: minimum value of annual average damage (weights as Figure 5.8) Source: Topographic map: Landesvermessungsamt Sachsen; risk estimation: own calculations

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 180 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Figure 5.10: Standardised multicriteria risk - criteria score sensitivity: maximum value of annual average damage (weights as fig. 4.8) Source: Topographic map: Landesvermessungsamt Sachsen; risk estimation: own calculations

Figure 5.11: Standardised multicriteria risk - weight sensitivity: large weight on environmental criterion (0.625) Source: Topographic map: Landesvermessungsamt Sachsen; risk estimation: own calculations

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 181 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

5.2.8 Outlook: Multicriteria project appraisal As mentioned before, our research mainly focuses on multicriteria risk mapping and not so much on multicriteria project appraisal. Finally, we aim to give at least a short impression of how these multicriteria risk maps described above could also be applied for the appraisal of flood risk reduction measures.

For our pilot study we calculated risk maps for the status quo as well as for the situation after implementing all measures which are planned in the flood protection concept (HWSK) for the River Mulde. By calculating the difference in the standardised risk value between the situation with and without these measures the effect of the planned measures can be illustrated. This is done in figure 5.12 showing the benefiting areas in blue (positive values = risk reduction). It can be seen that in this example (and under these weighting conditions) especially the city centre of Grimma would profit from the planned flood protection measure by a decrease of the standardised multicriteria risk of about 0.2-0.3 points.

Figure 5.12: Change in standardised multicriteria risk due to HWSK-measures (weights as 4.8) Source: Topographic map: Landesvermessungsamt Sachsen; risk estimation: own calculations

5.2.9 Conclusions The objective of our research was to develop approaches to improve the flood risk management process in three ways:

• Firstly, to include non-monetary risks in the overall flood risk assessment and project appraisal. • Secondly, to do this in a spatially differentiated way, i.e. to describe also the spatial distribution of these multicriteria risks. • Thirdly, to show approaches which deal with the uncertainties associated with the criteria evaluation.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 182 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Therefore, we developed a framework for a GIS-based multicriteria analysis which can be applied for assessment and decision problems in the context of flood risk management. Here, the different steps of such an MCA-framework were explained and different methods and approaches were introduced. Furthermore, we tested our framework for the multicriteria risk assessment and mapping for a pilot site, the Vereinigte Mulde in the federal state of Saxony, Germany. Therefore, a sample raster-based GIS-dataset was developed for social, environmental and economic risk criteria. Tests showed that some MCA approaches, such as the Hasse-diagram technique and PROMETHEE can be applied to vector-based GIS-data with a relatively low number of alternatives/areas to be compared but not for raster-based GIS-data, which usually involves a very high number of spatial units (raster cells). For the latter we applied for our pilot site two different MCA approaches: a disjunctive approach and an additive weighting approach. Our pilot study showed that both are appropriate for use within the framework of multicriteria risk mapping. The additive weighting approach would be furthermore applicable to show the spatial distribution of benefits of certain flood risk reduction measures. Regarding the consideration of uncertainties, an approach was shown at least for the economic criteria, as to how such uncertainties can be documented and dealt with.

As a further result, a first version of a software tool was developed (FloodCalc; Scheuer & Meyer 2007), which supports not only the calculations and mapping of the different damage and risk criteria, but also the two different MCA-procedures mentioned above.

However, the approach we applied in the pilot study should be seen only as a first basic approach, which needs to be adapted when transferred to other studies. Furthermore, several points can be identified where further improvement of the approach seems to be desirable:

Our pilot study mainly focussed on multicriteria risk mapping. Concerning the multicriteria appraisal of flood risk reduction measures further research seems to be necessary, e.g. how to combine the overall project appraisal with the analysis of the spatial distribution of its impacts.

The set of risk criteria used in our pilot study can be considered only as a first attempt to cover the economic as well as the environmental and social dimension of risks. For a more comprehensive study it should be verified whether more criteria should be included in order to show a more complete picture of flood risk. It may for example be useful also to include cultural heritage sites as a criterion or to incorporate also the potential environmental benefits of flooding. Furthermore the criteria used could be further elaborated. E.g. information on vulnerable groups could be integrated in the population criterion. The environmental criterion should also be further developed e.g. by specifying the functional relationships between flood characteristics and environmental effects in more detail. Even the economic criterion, which required the most effort among the criteria used, is calculated by a meso-scale approach. I.e. if more precise results were required here, this could be replaced by a micro- scale approach.

With regard to the multicriteria decision rules applied, the additive weighting approach is a very basic form of a MAUT approach. In order to represent the stakeholders’ preferences on single criteria in a better way e.g. value functions could be developed together with the decision makers and integrated into the decision rule for criteria standardisation. Furthermore a comparison with other decision rules, e.g. Compromise Programming, would be interesting.

In our pilot study we did not carry out a stakeholder workshop in order to investigate the decision makers’ preferences regarding the weighting of the different criteria. For a “real” multicriteria decision support this would, of course, be an important step in the whole process. The question who is allowed and legitimised to participate in such a decision making process seems to be another important research task.

Finally, only a very basic approach has been used to document uncertainties in the criteria evaluation and their influence on the final results. For a real project it would also be necessary to document the uncertainties in all criteria and not only one.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 183 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

6 Overall conclusions

With the recognition that it is not possible to protect from all floods, new innovative methods need to be devised to manage the associated risks and likely consequences. A clear understanding of these risks and impacts of future flooding, the potential benefits of mitigating these impacts and appropriate methodologies to evaluate them are therefore necessary to inform future investment in flood risk management options. The results from the four research projects reported here provide such new knowledge and understanding to derive risk analyses for flood prone areas and present innovative methods and tools to understand, model and evaluate flood damages.

In particular, the findings reported focus on socio-economic and ecological evaluation and modelling methodologies i.e. that relating to receptors such as people, buildings, infrastructure and the natural environment. Thus the research results help to redress the balance within flood risk management research, which has historically focused on flood prevention, and predicting flood probability rather than consequences. The current focus in Europe upon non-structural flood risk management measures, and national flood management policies which emphasise ‘finding room or space for rivers or water’, such as Making Space for Water in England and Wales (Defra, 2005), has helped to strengthen the importance of flood receptor research and has highlighted the extent to which responses to floods with regard to such receptors are currently inadequate and need to be improved. The combined results from the research should therefore lead to a better understanding and quantification of flood impacts on receptors, and the provision of evaluation methodologies, tools and approaches to better guide end- users in decision-making.

In particular, much of the research reported here has focused on non-monetary risks which have historically been ignored in favour of monetary risks, which are easier to measure. Although each of the methodologies reported here stands alone, each focusing on different receptors, the GIS-based multi-criteria approach to evaluation of various types of flood damages developed in Activity 4 can encompass the results from the models developed in Activities 1 to 3 on risk to life, the benefits of flood warnings and assessment of ecological pollution, thus drawing together the diverse elements of the research in a multicriteria analysis without having to measure them on one monetary scale.

In particular, the results also demonstrate that the use of social science approaches, methods and tools can provide important benefits for the assessment and management of future flood risks. The historical legacy of the past emphasis on defending against floods for many years largely resulted in research that encouraged an engineering and natural/physical science focus. Where it was included, social science research was rarely embedded as part of a multi-disciplinary process, and the potential for social science to help address policy questions and aid decision-making was not given the same status as natural and physical science (SAC, 2006). The research reported in this Task report therefore contributes to addressing the balance by providing socio-economic modelling methodologies developed by social scientists to help inform long-term strategic needs, medium-term policy priorities and shorter-term operational requirements.

These methodologies become even more important in respect to the new European Directive on the Assessment and Management of Flood Risks (EU 2007/60/EC of 23 October 2007). In particular, the methodologies can:

• contribute to the preparation of flood risk maps for at-risk areas; • provide innovative evaluation and modelling techniques which aim at reducing the adverse consequences for human, ecosystem and species’ health, cultural heritage and economic activity associated with floods in the European Community; • take into account the tangible and intangible costs and benefits of flood risk management strategies; and • contribute to aspects of flood risk management plans which focus on prevention, preparedness planning, emergency response and early warning systems.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 184 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Finally, the research serves to highlight a number of issues which are commonly reported in this document such as the availability of data, the importance of the local context, and issues relating to uncertainty. Activities 1 and 2 in particular highlighted the need for the establishment of reliable, systematic and consistent methods for collecting data following flood events across Europe, as well as for the need to make available such data that is collected. Methods of data collection were heavily influenced by the availability of data, particularly on the socio-economic dimensions of flood risk and floodplain use, flood damage, flood warning systems and warning response. A key constraint relates to who is responsible for collecting such data, which at present varies from agencies at local, regional and national levels. The EU has recognised the need for greater European co-ordination on flood risk management, therefore it is suggested that protocols are needed to address the data issue in order that data collection methods can be greatly improved.

Local context is everything, therefore it might be questionable to try to apply models across the whole of Europe when there are such large differences, for example: in the types and spatial distribution of flood hazard; in area, environment and people characteristics; in levels of social and economic vulnerability; in socio-political and institutional factors and arrangements; and differences in data reliability and availability. Therefore local contexts need to be particularly considered when applying models such as those developed for this research.

The research results have also attempted to highlight and address the important uncertainties associated with the assessment of flood risk, in this case particularly with regard to the human and ecological dimensions of floods and flood risk assessment, and the use of economic risk criteria. These uncertainties are in many ways inherently more unpredictable than the uncertainties in the fields of, for example, hydrology and hydraulics. The causes of these uncertainties are multifarious (e.g. data issues; modelling biases; human behaviour; and conceptual problems), therefore each Activity has suggested further research that is necessary to promote greater understanding of the factors influencing these uncertainties and, ultimately, for the reduction of the level of these uncertainties.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 185 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

7 References

1. ABT, S.R., WITTLER, R.J., TAYLOR, A., and LOVE, D.J. (1989) Human Stability in a high flood hazard zone. Water Resources Bulletin, 25 (4), 881-890.

2. AKTER, T. AND SIMONOVIC, S. P. (2005). "Aggregation of fuzzy views of a large number of stakeholders for multi-objective flood management decision-making." Journal of Environmental Management 77(2): 133-143.

3. ALDENBERG T, SLOB W. (1993) Confidence limits for hazardous concentrations based on logistically distributed NOEC toxicity data. Ecotoxicol Environ Saf 25:48–63.

4. ANDRYSZEWSKI, A., EVANS, K., HAGGETT, C., MITCHELL, B. & WHITFIELD, D. (2005), Flood Warning – The Next Step-Change, 40th Defra Conference of River and Coastal Engineers, Keele, July 2005.

5. BANA E COSTA, C. A. (1990). Reading in Multiple Criteria Decision Aid. Berlin, Heidelberg, New York, Tokyo, Springer Verlag.

6. BANA E COSTA, C. A., DA SILVA, P. A. AND NUNES CORREIA, F. (2004). "Multicriteria Evaluation of Flood Control Measures: The Case of Ribeira do Livramento." Water Resources Management 18: 21 (263-283).

7. BANSE G, BECHMANN G (1998), Interdisziplinäre Risikoforschung - eine Bibliographie. Wiesbaden.

8. BBC NEWS WEBSITE (2005) North still fighting flood effect 11 January 2005. Available: http://news.bbc.co.uk/1/hi/uk/4160387.stm Accessed: 10/06/07.

9. BELTON, V. & STEWART, T. J. (2002). Multiple Criteria Decision Analysis - An Integrated Approach. Boston, Dordrecht, London, Kluwer.

10. BLONG, R. (1997) A geography of natural perils. Australian Geographer, 28(1), 7-27.

11. BROUWER, R. AND VAN EK, R. (2004). "Integrated ecological, economic and social impact assessment of alternative flood control policies in the Netherlands." Ecological Economics 50(1-2): 1-21.

12. BRÜGGEMANN, P., BÜCHERL, C., PUDENZ, S. AND STEINBERG, C. E. W. (1999). "Application of the Concept of Partial Order on Comparative Evaluation of Environmental Chemicals." acta hydrochimica et hydrobiologica 27(3): 9 (170-178).

13. BÜTTTNER, U. (no date) Die größten Hochwasser im Gebiet der Mulden. Available at http://www.umwelt.sachsen.de/lfug/documents/HW_Mulden.pdf Accessed 14th May 2006

14. BYRNES, J.P., MILLER, D.C. and SCHAFER, W.D. (1999) Gender differences in risk taking: A meta-analysis, Psychological Bulletin, 125 (3), 367-383.

15. CNS SCIENTIFIC AND ENGINEERING SERVICE (1991) The Benefit Cost of Hydrometric Data – River Flow Gauging, Report prepared for Department of Environment. CNS, Reading.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 186 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

16. COATES, L. (1999) Flood fatalities in Australia, 1788-1996. Australian Geographer, 30 (3), 391-408.

17. COLIJN, F.; HAMANN, M.; REESE, S.; ROHR, T., (2000): Wertermittlung für die potenziell sturmflutgefährdeten Gebiete an den Küsten Schleswig-Holsteins. Gutachten im Auftrag des Ministeriums für ländliche Räume, Landesplanung, Landwirtschaft und Tourismus des Landes Schleswig-Holstein. web-publication: www.uni-kiel.de/ ftzwest/ag4/ PROJEKT/Txtlinks/test.htm.

18. COLOMBO, A. and VETERE ARELLANO, A.L. (eds.) (2002b) Lessons learnt from flood disasters. Available: http://nedies.jrc.it/doc/Floods%20Report%20with%20logo.pdf Accessed: 22/01/07.

19. CORNELISSEN, G. (Ph.D. Thesis, 1999) Mechanism and consequences of slow desorption of organic com-pounds from sediments, Universiteit van Amsterdam, Amsterdam.

20. COULTARD, T., FROSTICK, L., HARDCASTLE, H., JONES, K., RODGERS, D. and SCOTT, M. (2007), The June 2007 floods in Hull. Interim report by the Independent review Body, 24th August 2007. Available: http://www.benfieldhrc.org/floods/Crichton_Hull_2007.pdf Accessed: 03/09/07.

21. CRICHTON, D. (2007) The Hull floods of June 2007: Some insurance industry Implications. Paper published by Benfield Grieg Hazard Research Centre. Available: http://www.benfieldhrc.org/floods/Crichton_Hull_2007.pdf Accessed: 22/10/07.

22. DE BRUIJN, K. M. (2005) Resilience and flood risk management. WL/Delft, Delft.

23. DEFRA (2005) Making space for water: Taking forward a new Government strategy for flood and coastal erosion risk management, Defra, London

24. DELDERFIELD, E. (1981) The Lynmouth flood disaster, Published by Aktins, J & Letheren, J, Devon.UK.DE STRAAT MILIEU-ADVISEURS B.V. (2003). Diffuse bronnen in Groningen: Deelrapport Waterschap Hunze en Aa’s. Eindrapport w01a0086.H&A

25. DE MARCHI, B., SCOLOBIG, A., DELLI ZOTTI, G. & DEL ZOTTO, M. (2007) Risk construction and social vulnerability in an Italian Alpine Region. Milestone report T11- 06-08 for the Integrated Project FLOODsite.

26. DE STRAAT MILIEU-ADVISEURS B.V. (2003). Diffuse bronnen in Groningen: Deelrapport Waterschap Hunze en Aa’s. Eindrapport w01a0086.H&A

27. DI TORO DM, ALLEN HE, BERGMAN HL, MEYER JS, PAQUIN PR, SANTORE RC, (2001) A biotic ligand model of the acute toxicity of metals I. Technical basis. Environ.Toxicol.Chem. 20, 2383-2396.

28. DI TORO DM, ALLEN HE, BERGMAN HL, MEYER JS, SANTORE RC, PAQUIN PR, (2000) The Biotic Ligan Model: a computational approach for assessing the ecological effects of copper and other metals in aquatic systems. International Copper Association, Ltd., New York, USA.

29. DKKV (2004) Flood risk reduction in Germany: Lessons Learned from the 2002 Disaster in the Elbe Region, DKKV Publication, 29e.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 187 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

30. DOWNING, T.E. (1977) Warning for flash floods in Boulder, Colorado, Natural Hazard Working Paper 31, Institute of Behavioral Science, University of Colorado, Boulder.

31. DTLR (2001). Multi-Criteria Analysis: a manual. DTLR, London.

32. DRACUP, J.A. and KENDALL, D.R. (1990) Floods and droughts, in Waggoner, P.E. (ed.), Climate change and US water resources, John Wiley & Sons, New York, Section 11, 243-267.

33. DROBOT, S.D., BENIGHT, C and GRUNFEST, E.C. (forthcoming) Risk factors for driving into flooded roads, Environmental Hazards doi:10.1016/j.envhaz.2007.07.003.

34. DRECHSLER, M. (1999). Verfahren zur multikriteriellen Entscheidungsunterstützung bei Unsicherheit. Naturressourcenschutz und wirtschaftliche Entwicklung Nachhaltige Wasserbewirtschaftung und Landnutzung im Elbeeinzugsgebiet. H. Horsch, Ring, I. Leipzig.

35. DRECHSLER, M. (2004). PRIMATE - an interactive software for Probabilistic Multi- Attribute Evaluation.

36. DUCLOS, P., VIDONNE, P., BEUF, P., PERRAY, P. and STOEBNER, A. (1991) Flash flood disaster- Nimes, France, 1988. European Journal of Epidemiology, 7 (4), 365-371.

37. EASTERN DAILY PRESS (2004) The days that the rains came down. Feature on the Norfolk floods, 3rd August 2004 written by Keiron Pim. Available: http://www.edp24.co.uk/content/features/norfolkdisasters/floods.asp Accessed: 20/08/07.

38. EDWORHTY, J. & ADAMS, A. (1996) Warning Design: Research Perspective, Taylor and Francis, London.

39. ENVIRONMENT AGENCY (2005c) Living with the risk: The floods in Boscastle and North Cornwall 16 August 2004, Exeter, Devon: Environment Agency. Available: http://www.environment- agency.gov.uk/commondata/acrobat/living_with_the_risk_1111062.pdf Accessed: 15/05/07.

40. ENVIRONMENT AGENCY (2006) Cumbria Floods Technical Report: Factual report on meteorology, hydrology and impacts of January 2005 flooding in Cumbria, Environment Agency, Bristol. Available: http://publications.environment- agency.gov.uk/pdf/GENW1106BLSF-e-e.pdf Accessed: 20/08/07.

41. ENVIRONMENTAL AGENCY NATIONAL FLOOD WARNING AGENCY (2003) Flood Warning Investment Strategy Appraisal Report 2003/04 to 2012/13, Environment Agency National Flood Warning Centre, Frimley, UK.

42. ESDS (ECONOMIC AND SOCIAL DATA SERVICE) INTERNATIONAL WEBSITE Available at http://www.esds.ac.uk/international/access/access.asp, Accessed 20th August 2007.

43. EU DIRECTIVE 2000/60/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL (2000) Establishing a framework for Community action in the field of water policy Available at http://eur- lex.europa.eu/LexUriServ/site/en/oj/2000/l_327/l_32720001222en00010072.pdf. Accessed 28th November 2007.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 188 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

44. EUROPEAN ENVIRONMENT AGENCY (2001) Sustainable water use in Europe, Part 3: Extreme hydrological events: floods and droughts, Environmental Issue Report No. 21, EEA, Copenhagen

45. EUROPEAN PARLIAMENT (2007) Directive of the European Parliament and of the Council on the assessment and management of flood risks, 5th July 2007, Available at http://register.consilium.europa.eu/pdf/en/07/st03/st03618.en07.pdf Accessed 22nd November 2007.

46. EU STATISTICS UK WEBSITE Available at http://www.eustatistics.gov.uk/, Accessed 20th August 2007.

47. EUROSTAT DATABASE (2007) Available at http://epp.eurostat.ec.europa.eu/portal/page?_pageid=1090,30070682,1090_33076576&_da d=portal&_schema=PORTAL Accessed 10th September 2007.

48. EVANS, E., ASHLEY, R., HALL, J., PENNING-ROWSELL, E., SAUL, A., SAYERS, P., THORNE, C. and WATKINSON, A. (2004) Foresight. Future Flooding. Scientific Summary: Volume I Future risks and their drivers. Office of Science and Technology, London.

49. FEW, R.. (2006) Flood Hazards, Vulnerability and Risk Reduction. In Few, R.. and Matties, F. (Eds.) Flood Hazards and Health: Responding to the Risks of Climate Change. Earthscan, London, 8-27.

50. FOUNDATION BIOLOGIC PIG FARMING (2003). Biologic Pig Farming: Chapter 5 Feeding. Animal Sciences Group of the Wageningen University and Research centre.http://www.biologischeveehouderij.nl/biobieb/Digitaalhandboek/Inhoud/05Voeding. pdf

51. FORTH NATIONAL POLICY DOCUMENT on Water Management Government Decision, (2000) Dutch Staatscourant 114.

52. FREISTAAT SACHSEN (2003), Bericht der Sächsischen Staatsregierung zur Hochwasserkatastrophe im August 2002, Report of the Federal Government of Saxony on the Flood Catastrophe in August 2002

53. GESUNDHEITSBERICHTERSTATTUNG DES BUNDES, (2006) Federal Health Monitoring information system Available at: http://www.gbe-bund.de/. Accessed 12th November 2007.

54. GOULDBY, B. and SAMUELS, P. (2005), Language of risk - project definitions. FLOODsite project report T32-04-01.

55. GRAHAM, W.J. (1999) A procedure for estimating loss of life caused by dam failure. Dam safety office report DSO-99-6.

56. GOVERNMENT OFFICE FOR THE NORTH WEST (2005) Carlisle storms and associated flooding: multi-agency debrief report 14th July 2005. Available: http://www.ukresilience.info/upload/assets/www.ukresilience.info/carlislereport.pdf Accessed: 20/08/07.

57. GUZZETTI, F., STARK, C. and SALVATI, P. (2005) Evaluation of flood and landslide risk to the population of Italy. Environmental Management, 36 (1), 15-36.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 189 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

58. HANDMER, J.W. (2000) Are Flood Warnings Futile? Risk communication in emergencies. The Australasian Journal of Disaster and Trauma Studies, Vol. 2000 (2), 1- 14, Available at http://www.massey.ac.nz/%7Etrauma/issues/2000-2/handmer.htm. Accessed 26th November 2007.

59. HANDMER, J.W. & ORD, K.D. (1986) Flood warning and response. In Smith, D.I. & Handmer, J.W. (eds.) Flood Warning in Australia, Section 17, 235-257, Centre for Resource and Environmental Studies, Australian National University, Canberra.

60. HEWITT, K. (1997) Regions of Risk: A Geographical Introduction to Disasters. Longman, Harlow, UK.

61. HORLITZ, T., AHRENS, H., HARTH, M., HILLERT, D. AND SANDER, A. (2004). Integriertes Entwicklungskonzept. Integration von Schutz und Nutzung im Biosphärenreservat Mittlere Elbe. P. Wycisk, Weber, M. Berlin.

62. HOYOIS, P. & GUHA-SAPIR, D. (2003) Three decades of floods in Europe: a preliminary analysis of EMDAT data, Centre for Research on the Epidemiology of Disasters (CRED), Catholic University of Louvain, Louvain.

63. HR WALLINGFORD (2003) Flood Risks to People Phase 1. Final Report Prepared for Defra/Environment Agency Flood and Coastal Defence R&D Programme.

64. HR WALLINGFORD (2005a) The Flood Risk to People Methodology. Flood Risks to People, Phase 2 R&D Output FD2321/TR1. Defra/ Environment Agency Flood and Coastal Defence R&D Programme.

65. HR WALLINGFORD (2005b) Flooding in Boscastle and North Cornwall, August 2004: Phase 2 studies report, Report EX5160, May 2005, HR Wallingford. Available: http://www.hrwallingford.co.uk/downloads/publications/EX5160_Boscastle_floodings.pdf Accessed: 26/06/07.

66. HR WALLINGFORD (2005c) Guidance Document. R&D Outputs: Flood Risks to People, Phase 2. FD2321/TR2 Defra/Environment Agency Flood and Coastal Defence R&D Programme.

67. HULSCHER, T.E.M. (Ph.D. Thesis, 2005) Availability of organic contaminants in lake Ketelmeer sediment., Universiteit van Amsterdam, Amsterdam.

68. IKSR (INTERNATIONAL COMMISION FOR THE PROTECTION OF THE RHINE), 2001: Übersichtskarten der Überschwemmungsgefährdung und der möglichen Vermögensschäden am Rhein. Abschlußbericht: Vorgehensweise zur Ermittlung der hochwassergefährdeten Flächen, Vorgehensweise zur Ermittlung der möglichen Vermögensschäden.

69. INTERNATIONAL COMMISSION FOR THE PROTECTION OF THE RHINE (2002) Non Structural Flood Plain Management – Measures and their Effectiveness, ICPR, Koblenz, 2002 Available at http://iksr.de/fileadmin/user_upload/documents/rz_iksr_engl.pdf, Accessed 20th November 2007.

70. JANSSEN, R., HERWIJNEN, M. V. AND BEINAT, E. (2003). Definite - Case Studies and User Manual. Amsterdam, Vrije Universiteit Amsterdam / IVM.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 190 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

71. JOINT RESEARCH CENTRE (2004) Substance Data Sheets: DRAFT of 01-03-04, European Commission Directorate General.

72. JONKMAN, S.N. (2003) Loss of life caused by floods: an overview of mortality statistics for worldwide floods. Available: http://www.library.tudelft.nl/delftcluster/theme_risk.html Accessed: 16/11/05.

73. JONKMAN, S.N. (2005) Global perspectives on loss of human life caused by floods. Natural Hazards, 34, 151-175.

74. JONKMAN, S.N. (2007) Loss of life estimation in flood risk assessment: theory and applications. PhD dissertation, Delft Cluster, Delft Netherlands.

75. JONKMAN, S.N. and KELMAN, I. (2005) An analysis of the causes and circumstances of flood disaster deaths. Disasters 29 (1), 75-97.

76. JONKMAN, S.N., VAN GELDER, P.H.A.J.M., and VRIJLING, J.K. (2002) Loss of life models for sea and river floods. In Wu et al. (eds.) Flood Defence 2002, Science Press, New York Ltd. Available: http://www.waterbouw.tudelft.nl/public/gelder/paper120b- v10210.pdf Accessed: 16/08/07.

77. JONKMAN, S.N. and PENNING-ROWSELL, E.C. (forthcoming) Human stability in flood flows. Journal of the American Water Resources Association.

78. KARVONEN, R.A., HEPOJOKI, A., HUHTA, H.K.AND LOUHIO, A. (2000) The use of physical models in dam-break analysis, RESCDAM Final Report, Helsinki University of Technology, Helskini Finland. Available: http://www.ymparisto.fi/download.asp?contentid=13494&lan=fi Accessed: 05/10/07.

79. KEENEY, R. L. AND RAIFFA, H. (1993). Decisions with Multiple Objectives - Preferences and Value Tradeoffs. Cambridge, Cambridge University Press.

80. KELMAN, I. (2004) Perspectives on researching flood health issues. Available: www.ilankelman.org/disasterdeaths/IlanKelmanFloodHealthIssues.rtf Accessed: 19/01/07.

81. KIESE, M.; LEINEWEBER, B., (2001) Risiko einer Küstenregion bei Klimaänderung. Ökonomische Bewertung und räumliche Modellierung des Schadenspotentials in der Unterweserregion. Hannoversche Geographische Arbeitsmaterialien, Nr. 25. Hannover.

82. KIRCHBACH (2002) Bericht der Unabhängigen Kommission der Sächsischen Staatsregierung Flutkatastrophe 2002,Available at http://home.arcor.de/schlaudi/Kirchbachbericht.pdf, Accessed 2003.

83. KLAUS, J., SCHMIDTKE, R. F. (1990), Bewertungsgutachten für Deichbauvorhaben an der Festlandsküste – Modellgebiet Wesermarsch, Untersuchungsbericht an den Bundes- minister für Ernährung, Landwirtschaft und Forsten

84. KOK, M.; HUIZINGA, H.J.; VROUWENFELDER, A.C.W.M.; BARENDREGT, A., (2004) Standard Method 2004. Damage and Casualties caused by Flooding. Client: Highway and Hydraulic Engineering Department.

85. KOOIJMAN SALM. (1987) A safety factor for LC50 values allowing for differences in sensitivity among species. Water Res 21:269–276.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 191 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

86. KREIBICH, H., MULLER, M., THIEKEN, A.H. & MERZ, B. (2007) Flood precaution of companies and their ability to cope with the flood in August 2002 in Saxony, Germany, Water Resources Research, Vol. 43, (3) 1-15.

87. KREIBICH, H., THIEKEN, A.H., PETROW, T.H., MULLER, M. & MERZ, B. (2005) Flood loss reduction of private households due to building precautionary measures – lessons learned from the Elbe flood in August 2002, Natural Hazards and Earth Systems Sciences, Vol. 5, 117-126.

88. KRON, W. (2002) Flood risk = hazard x exposure x vulnerability, in Wu et al. (eds.) Flood Defence 2002. Science Press, New York.

89. KUNDZEWICZ, Z.W. and KUNDZEWICZ, W.J. (2005) Mortality in flood disasters, in Kirch, W., Menne, B. and Bartollini (eds.) Extreme weather events and public health responses. WHO, Springer, EU.

90. KUNDZEWICZ, Z.W., SZAMALEK, K. and KOWALCZAK, P. (1999) The great flood of 1997 in Poland. Hydrological Sciences Journal, 44 (6), 855-870.

91. KUNDZEWICZ, Z.W., ULBRICH, U., BRÜCHER, T., GRACZYK, KRÜGER, A., PINSKWAR, I., RADZIEJEWSKI, M. and SZWED, M. (2005) Summer floods in Central Europe- Climate change track? Natural Hazards, 36, 165-189.

92. LEGOME, E., ROBBINS, A. and RUND, D.A. (1995) Injuries associated with floods: The need for an international reporting scheme. Disasters, 19 (1), 50-54.

93. LEWIS, J. (1999) The meaning of vulnerability, in Lewis, J. (ed.) Development in disaster prone places: Studies of vulnerability, Intermediate Technology Publications, London, 3- 11.

94. LIND, N., HARTFORD, D. and ASSAF, H. (2004) Hydrodynamic models of human stability in a flood. Journal of the American Water Resources Association, 40(1), 89-96.

95. LUHMANN N (1991), Soziologie des Risikos. Berlin/New York.

96. LUTOFF, C. and RUIN, I. (2007) Gard river case study, Report T10-07-03 prepared as part of Task 10 of the Integrated Project FLOODsit.

97. MAI, S., GRABEMANN, I., EPPEL, D.-P., ELSNER, A., ELSNER, W., GRABEMANN, H.-J., KRAFT, D., MEYER, V., OTTE, C., YU, I., WITTIG, S., ZIMMERMANN, C. (2007), KRIM: Methode der erweiterten Risikoanalyse. In: Land unter? Klimawandel, Küstenschutz und Risikomanagement in Nordwestdeutschland: die Perspektive 2050, Eds. Schuchardt, B., Schirmer, M., 2007 (in Print).

98. MALCZEWSKI, J. (1999). GIS and multicriteria decision analysis. New York.

99. MALCZEWSKI, J. (2006). "GIS-based multicriteria decision analysis: a survey of the literature." International Journal of Geographical Information Science 20(7): 703-726.

100. MALCZEWSKI, J., CHAPMAN, T., FLEGEL, C., WALTERS, D., SHRUBSOLE, D. AND HEALY, M. A. (2003). "GIS - multicriteria evaluation with ordered weighted averaging (OWA): case study of developing watershed management strategies." Environment and Planning A 35(10): 1769-1784.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 192 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

101. MERZ, R. AND BUCK, W. (1999). Integrierte Bewertung wasserwirtschaftlicher Maßnahmen. Materialien. D. V. f. W. u. K. (DVWK). Bonn.

102. MERZ, B., KREIBICH, H., THIEKEN, A. & SCHMIDTKE, R. (2004) Estimation uncertainty of direct monetary flood damage to buildings, Natural Hazards and Earth System Sciences, Vol. 4, 153-163.

103. MESSNER F., PENNING-ROWSELL E, GREEN C, MEYER V, TUNSTALL S, VAN DER VEEN A, (2007) Guide-lines for socio-economic Flood Damage Evaluation. Deliverable report T09-06-01 for the Integrated Project FLOODsite.

104. MESSNER, F., ZWIRNER, O. AND KARKUSCHKE, M. (2006). "Participation in multicriteria decision support for the resolution of a water allocation problem in the Spree River basin." Land Use Policy 23(1): 63-75.

105. METEOROLOGICAL OFFICE/POSFORD DUVIVIER (2003) Pluvial flooding impact strategy study, Met. Office, Bracknell.

106. MEYER, S.M. (1999) Literature review on the bioavailability and toxicity of selected metals in the aquatic environment relative to natural water chemistry. Draft submitted to The International Lead-Zinc Research Organization, Inc., August 1999.

107. MEYER, V. (2005), Methoden der Sturmflut-Schadenspotenzialanalyse an der deutschen Nordseeküste, Vom Fachbereich Geowissenschaften und Geographie der Universität Hannover genehmigte Dissertation, UFZ Dissertation 3/2005

108. MEYER V, SCHEUER S, HAASE D (2007), GIS-based Multicriteria Analysis as Decision Support in Flood Risk Management. UFZ-Discussion Papers 6/2007. www.ufz.de/data/DP_2007_06_Meyer7117.pdf

109. MEYER, V., HAASE, D. & SCHEUER, S. (2007) GIS-based Multicriteria Analysis as Decision Support in Flood Risk Management. Milestone report T10-07-06 for the Integrated Project FLOODsite.

110. MINISTERIE VAN BINNENLANDSE ZAKEN EN KONINKRIJKRELATIES VROM (2006). Leidreid risico Inventarisatie: Deel gevaarlijke stoffen (LRI-GS). Versie 3, december 2006.

111. MMWR (1989) Medical Examiner/Coroner Reports of deaths associated with Hurricane Hugo—South Carolina. Morbidity and Mortality Weekly Report, 38(44), 759-762.

112. MMWR (2000) Morbidity and mortality associated with Hurricane Floyd-North Carolina, September-October 1999. Morbidity and Mortality Weekly Report, 45 (2), 32-38.

113. MMWR (2005) Carbon Monoxide Poisoning After Hurricane Katrina—Alabama, Louisiana, and Mississippi, August—September 2005. Morbidity and Mortality Weekly Report, 54 (39), 996-998.

114. MUNDA, G. (1995). Multicriteria Evaluation in a Fuzzy Environment - Theory an Applications in Ecological Economics. Heidelberg, Physica Verlag.

115. MUNDA, G. (2006). "Social multi-criteria evaluation for urban sustainability policies." Land Use Policy 23(1): 86-94.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 193 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

116. MURL (MINISTERIUM FÜR UMWELT, RAUMORDNUNG UND LANDWIRTSCHAFT DES LANDES NORDRHEIN-WESTFALEN) (Hrsg.) (2000a), Hochwasserschadenspotentiale am Rhein in Nordrhein-Westfalen, Abschlussbericht

117. NACHTNEBEL, H. P. (2007). Cost-benefit evaluation of risk reduction options. In: Schanze J (Ed). Flood risk management Research – from extreme events to citizen involvement. Procceedings of European Symposium on Flood Risk Management Research (EFRM 2007), 6-7 February 2007, Dresden.

118. NEWMAN M.C. et al, (2000) Applying species-sensitivity distributions in ecological risk assessment: Assumptions of distribution type and sufficient numbers of species. Environmental Toxicology and Chemistry, 2000, Vol. 19, pp. 508–515.

119. NIRUPAMA, N. AND SIMONOVIC, S. P. (2007). "Increase of flood risk due to urbanisation: A canadian example." Natural Hazards 40(1): 25-41.

120. NOJI, K.E. (1993) Analysis of medical needs during disasters caused by tropical cyclones: anticipated injury patterns. Journal of Tropical Medicine and Hygiene, 96, 370-376.

121. OFFICE OF SCIENCE AND TECHNOLOGY (2004) Foresight Future Flooding, Executive Summary, OFT, London

122. OLFERT, A. (2006). Draft methodology for ex-post evaluation of measures and instruments. Floodsite-Report.

123. OMANN, I. (2004). Multi-Criteria Decision Aid As An Approach For Sustainable Development Analysis And Implementation. Graz, Graz: 272.

124. PANEQUE SALGADO, P., CORRAL QUINTANA, S., GUIMARAES PEREIRA, A., DEL MORAL ITUARTE, L. AND PEDREGAL MATEOS, B. (no date) "Participative multi-criteria analysis for the evaluation of water governance alternatives. A case in the Costa del Sol (Malaga)." Ecological Economics In Press, Corrected Proof.

125. PARKER, D.J. (1991) The Damage Reducing Effects of Flood Warnings. Report Prepared for Halcrow National and for the National Rivers Authority (Anglian Region) Regional Telemetry Scheme Appraisal, FHRC, Enfield, UK.

126. PARKER, D.J. (1999) Flood. In Ingleton, J. (ed.) Natural Disaster Management, Tudor Rose, London, 38-40.

127. PARKER, D.J. (ed.) (2000) Floods. Routlege, London.

128. PARKER, D.J. (2004) Designing flood forecasting, warning and response systems from a societal perspective, Meteorologische Zeitschrift, Vol. 13 (1), 5-11.

129. PARKER, D.J. (2007) Evaluating business interruption losses caused by flooding, In Penning-Rowsell E C (ed) The Flood Threat to London, Chapter 8.

130. PARKER, D.J. & BUDGEN, P. (1998) The tropical cyclone warning dissemination system in Mauritius, In UK National Coordination Committee for the International Decade for Natural Disaster Reduction, Forecasts and Warnings, Section 1, Thomas Telford, London

131. PARKER, D.J. & HAGGETT, C. (2001) The development of flood warning technologies in England and Wales, Paper presented at the International Emergency and Disaster Management Conference (TIEMS), Oslo, Norway, June 2001.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 194 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

132. PARKER, D.J. & NEAL, J. (1990) Evaluating the Performance of Flood Warning Systems. In HANDMER, J.W. & PENNING-ROWSELL, E.C. (Eds.) Hazard and the Communication of Risk, Section 8, 137-156, Gower Technical Press, Aldershot.

133. PARKER, D.J., GREEN, C.H., THOMPSON, P.M. (1987) Urban flood protection benefits: a project appraisal guide. Gower Technical Press, Aldershot, UK.

134. PARKER, D.J., PRIEST, S.J., SCHILDT, A. & Handmer, J. (2007) Modelling the damage reducing effects of flood warning. Milestone report T10-07-12 for the Integrated Project FLOODsite.

135. PARKER, D.J., TAPSELL, S.M. & MCCARTHY, S. (2007) Why warn? The human impacts of flood forecasting, warning and response systems, Natural Hazards, 43 (3), 397- 414.

136. PENNING-ROWSELL, E.C. & PEERBOLTE, B. (1994) Concepts, Policies and Research. In Penning-Rowsell E C and Fordham M. (eds.) Floods Across Europe: Flood Hazard Assessment, Modelling and Management. Section 1, Middlesex University Press, London.

137. PENNING-ROWSELL, E.C. & TAPSELL, S.M. (2002) Coping with extreme floods: warnings, impacts and response, The extremes of the extremes: extraordinary floods, International Association of Hydrological Sciences, Vol. 271, 379-383.

138. PENNING-ROWSELL, E.C., CHATTERTON, J.B. & PARKER, D.J. (1978) The Effect of Flood Warning on Flood Damage Reduction, A report for the Central Water Planning Unit, Flood Hazard Research Centre, Enfield.

139. PENNING-ROWSELL, E., JOHNSON, C., TUNSTALL, S., TAPSELL, S., MORRIS, J., CHATTERTON, J., COKER, A. and GREEN, C. (2003). The benefits of flood and coastal defence: techniques and data for 2003. Enfield, Flood Hazard Research Centre.

140. PENNING-ROWSELL, E.C., PARKER, D.J., CREASE, D. & MATTISON, C.R. (1983) Flood warning dissemination: an evaluation of some current practices in the Severn Trent Water Authority area, Flood Hazard Research Centre, Middlesex Polytechnic, Enfield, UK.

141. PENNING-ROWSELL, E.C., TAPSELL, S.M. & WILSON, T. (2004) Some policy implications of the health effects of floods. World Health Organisation Conference, Bratislava, Hungary.

142. POSTHUMA L., SUTTER G.W., TRAAS T.P. (2002) Species Sensitivity Distributions in Ecotoxicology. Lewis Publishers.

143. PRIEST, S.J., WILSON, T., TAPSELL, S., PENNING-ROWSELL, E., VIAVATTENE, C. and FERNANDEZ-BILBAO, A. (2007) Building a model to estimate Risk to Life for European flood events. Milestone report T10-07-10 for the Integrated Project FLOODsite

144. PROCTOR, W. AND DRECHSLER, M. (2006). "Deliberative multicriteria evaluation." Environment and Planning C-Government and Policy 24(2): 169-190.

145. PROSSER, T. (2001) The Lynmouth flood disaster. Lyndale Photographic Limited, Devon, UK.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 195 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

146. PUDENZ, S., BRÜGGEMANN, R., LUTHER, B., KAUNE, A. and KREIMES, K. (2000). "An algebraic/graphical tool to compare ecosystems with respect to their pollution V: cluster analysis and Hasse diagrams." Chemosphere 40: 10.

147. RAUSCHMAYER, F. (2001). Entscheidungshilfen im Umweltbereich - Von der mono- kriteriellen zur multi-kriteriellen Analyse. Ökonomische Naturbewertung. F. Beckenbach, Hampicke, U.: 21 (221-241).

148. RAUSCHMAYER, F. AND WITTMER, H. (2006). "Evaluating deliberative and analytical methods for the resolution of environmental conflicts." Land Use Policy 23(1): 108-122.

149. RAYNOR, S. and CANTOR, R. (1987) How fair is safe enough? The cultural approach to societal technology choice. Risk Analysis, 7(1), 3-9.

150. REITER, P (2000) International methods of risk analysis, damage evaluation and social impact studies concerning dam-break accidents. International Seminar on the RESCDAM Project, Sein joki, 1-5 October 2000.

151. RENN O (1998), Three decades of risk research: accomplishments and new challenges. Journal of Risk Research 1 (1), 49-71.

152. RIJKSINSTITUUT VOOR INTEGRAAL ZOETWATERBEHEER EN AFVALWATERBEHANDELING RIZA (1996). Emissiefactoren voor diffuse bronnen van waterverontreiniging. Werkdocument 96.087x. Directoraat-Generaal Rijkswaterstaat.

153. ROSENTHAL, U. and BEZUYEN, M.J. (2000) Flood emergency management in developed countries. The experience of 1993, 1997 and 1997 in Europe, in Parker, D.J. (ed.) Floods, volume 1, 2000, London, Routledge.

154. ROWE, D. (2004) Boscastle: 16th August 2004 – The day of the flood, Truran books, Truro, UK.

155. RPA (2004). Evaluating a multi-criteria analysis methodology for application to flood management and coastal defence appraisals. R&D Technical Report. DEFRA.

156. RUIN, I. (2007) Conduite à contre-courant. Les pratiques de mobilité dans le Gard : facteur de vulnérabilité aux crues rapides. Thèse de doctorat, Université Joseph Fourier, Grenoble. Available: http://www- sr.imag.fr/Les.Personnes/Martin.Heusse/Isa/theseRuin.pdf.

157. SAXON FLOOD CENTRE WEBSITE (2007) The reporting and information chain for flood related messages in Saxony. Available at http://www.umwelt.sachsen.de/de/wu/umwelt/lfug/lfug-internet/wasser_18849.html, Accessed 28th November 2007.

158. SAXONY STATE OFFICE FOR ENVIRONMENT AND GEOLOGY (2004) Event analysis: August 2002 flood in the watercourses of the Osterzgebirge Mountains, Published by Saxony State Office for Environment and Geology, Dresden, Germany.

159. SCANLON, J. & FRIZZELL, A. (1979) Old theories don’t apply: implications of communications in crisis, Disasters, Vol. 3 (3), 315-19.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 196 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

160. SCHANZE J (2006), Flood risk management - a basic framework. In: Schanze, J.; Zeman, E.; Marsalek, J. (Eds.), 2006: Flood Risk Management - Hazards, Vulnerability and Mitigation Measures, 149-167. Springer.

161. SCHEUER, S. & MEYER, V. (2007), FloodCalc. Software tool for the calculation of multicriteria flood damage and risk maps. Version 1.0 alpha

162. SCHLENGER, W.E., NORTON, G.M., WALKER, D.K., GOLDBERG, A. and LEWIS, C. (2006) Estimating the loss of life from hurricane-related flooding in the Greater New Orleans area: Health effects of Hurricane Katrina. Report prepared for the Army Corps of Engineers, by Abt Associates Inc. Bethesda, Maryland, US.

163. SCHRODER T.J. (Ph.D. Thesis, 2005) Solid-solution partitioning of heavy metals in floodplain soils of the rivers Rhine and Meuse. Universiteit van Wageningen, Wageningen.

164. SCHWARE, R. & LIPPOLDT, P. (1982) An examination of community flood warning systems: are we providing the right assistance?, Disasters, Vol. 6 (3), 195-201.

165. SCIENCE ADVISORY COUNCIL (2006) Increasing the capacity and uptake of social research in Defra. SAC Paper (06) 42. SAC Secretariat 14/12/2006, Defra, London

166. SCOTLAND AND NORTHERN IRELAND FORUM FOR ENVIRONMENTAL RESEARCH (2006) Assessing the benefits of flood warning: a scoping study, Project Code UKCC10, SNIFFER, Edinburgh, UK.

167. SHEAFFER, J.R. (1960) Flood proofing: an element in a flood damage reduction program, Research Paper No. 65, Department of Geography, University of Chicago, Chicago, UK.

168. SHEAFFER, J.R. (1967) Introduction to flood proofing, The Centre for Urban Studies, University of Chicago, Chicago, UK.

169. SIMON, U. (2003). Multikriterielle Bewertung von wasserwirtschaftlichen Maßnahmen aus gewässerökologischer Sicht - Beispiel Berlin.

170. SIMONOVIC, S. P. and NIRUPAMA, N. (2005). "A spatial multi-objective decision- making under uncertainty for water resources management." Journal of Hydroinformatics 7(2): 117-133.

171. SMUL, LTV & PGS (Planungsgemeinschaft Dr. Scholz), 2004: Hochwasserschutzkonzept für die Mulden im Regierungsbezirk Leipzig. Unveröffentlicht.

172. SOCHER, M., SIEBER, H.-U., MÜLLER, G. and WUNDRAK, P. (2006). "Verfahren zur landesweiten Priorisierung von Hochwasserschutzmaßnahmen in Sachsen." Hydrologie und Wasserbewirtschaftung 50(3).

173. SOERENSEN, P. B., GYLDENKAERNE, S., LERCHE, D., BRÜGGEMANN, R., THOMSEN, M., FAUSER, P. AND MOGENSEN, B. B. (2004). Probability approach applied for prioritisation using multiple criteria. Order theory in environmental science. S. e. al.

174. SOUTH WEST RESILIENCE FORUM (2006) Major flooding at Boscastle and surrounding areas of North Cornwall 16th August 2004: A summary of lessons learned. A multi-agency report sponsored by the South West Resilience Forum, Preparing for

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 197 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

Emergencies, 17th March 2006. Available: http://www.devon- cornwall.police.uk/v3/showpdf.cfm?pdfname=boscastledebrief.pdf Accessed: 15/05/07.

175. STADTVERWALTUNG GRIMMA (2007) Grimma river webcam, Available at http://www.grimma.de/12_webcams/webcam-pegel.php Accessed 21st Novemeber 2007

176. STATE RESERVOIR ADMINISTRATION (2007) The State Reservoir Administration of Saxony: Functon, Organisation, Projects. Available at http://www.smul.sachsen.de/de/wu/organisation/staatsbetriebe/ltv/downloads/LTV_Image_ engl.pdf, Accessed 27th November 2007.

177. STATISTICHES LANDESAMT DES FREISTAATES SACHSEN (2007) Schwerbehinderte Menschem in Sachsen. Available from http://www.statistik.sachsen.de/11/Fb_Schwerbehinderte_07.pdf Accessed 4th October 2007.

178. STEENWIJK, VAN J.M., G. CORNELISSEN en Th.E.M. ten Hulscher. Omgaan met ver- delingscoëfficiënten voor organische verbindingen; Variaties op een Constante; sorptie- onderzoek. RWS/RIZA 1994-1999; 1999; RIZA Nota nr. 99.023; AKWA nota nr. 99.004.

179. STEINFUHRER, A. & KUHLICKE, C. (2007) Social vulnerability and the 2002 flood, Country Report Germany (Mulde River). Report T11-07-08 for the Integrated Project FLOODsite.

180. STILLWELL, W. G., SEAVER, D. A. AND EDWARDS, W. (1981). "A Comparison of Weight Approximation Techniques in Multiattribute Utility Decision-Making." Organizational Behavior and Human Performance 28(1): 62-77.

181. STOKES, R. & MAY, P. (2004) The Severn Valley: An Alternative Solution, Evaluation of Temporary Flood Barriers During the February 2004 Flood, Environment Agency Midlands Region.

182. STRAGER, M. P. AND ROSENBERGER, R. S. (2006). "Incorporating stakeholder preferences for land conservation: Weights and measures in spatial MCA." Ecological Economics 57: 13 (627-639).

183. TAPSELL, S.M., TUNSTALL, S.M., GREEN, C., VIAVATTENE, C. (2008, in progress) A conceptual model of the health impacts of flooding. Report for the Integrated Project FLOODsite.

184. TAPSELL, S.M., BURTON, R., PARKER, D.J. & OAKES, S. (2004) The Social Performance of Flood Warning Communications Technologies, Technical Report W5C- 016 to the Environment Agency, Flood Hazard Research Centre, Middlesex University, London. Available from http://sciencesearch.defra.gov.uk/Document.aspx?Document=Social_Perfo__1236.PDF Accessed 29th November 2007.

185. TAPSELL, S.M., PENNING-ROWSELL, E.C., TUNSTALL, S.M. & WILSON, T.L. (2002) Vulnerability to flooding: health and social dimensions, Philosophical Transactions of The Royal Society Mathematical, Physical and Engineering Sciences, Vol. 360 (1796), 1511-1525.

186. THIEKEN, A.H., KREIBICH, H., MULLER, M. & MERZ, B. (2007) Coping with floods: preparedness, response and recovery of flood-affected residents in Germany in 2002, Hydrological Sciences, 52, (5), October, 1016-1037.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 198 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

187. THIEKEN, A.H., MULLER, M., KREIBICH, H. & MERZ, B. (2005) Flood damage and influencing factors: New insights from the August 2002 flood in Germany, Water Resources Research, Vol. 41, 1-16.

188. THINH, N. X. AND HEDEL, R. (2004). A Fuzzy Compromise Programming Environment for the ecological evaluation of land use options. EnviroInfo 2004.

189. THINH, N. X. AND VOGEL, R. (2006). GIS-based multiple criteria analysis for land-use suitability assessment in the context of flood risk management. InterCarto - InterGIS 12, Berlin.

190. TKACH, R. J. AND SIMONOVIC, S. P. (1997). "A new approach to multi-criteria decision making in water resources." Journal of Geographic Information and Decision Analysis 1(1): 25-43.

191. TUNSTALL, S.M., TAPSELL, S.M. & FERNANDEZ-BILBAO, A. (2005). The damage reducing effects of flood warnings: Results from new data collection. Flood Hazard Research Centre, Middlesex University, London

192. UNDRO (United Nations Disaster Relief Organisation) (1984) Disaster prevention and mitigation, UNDRO, Geneva

193. UNIVERSITÉ CATHOLIQUE DE LOUVAIN “EM-DAT: The OFDA/CRED International Disaster Database, Brussels, Belgium. Available: www.em-dat.net Accessed 22/01/07.

194. VAN DUIN, M.J. & BEZUYEN, M.J. (2000) Flood Evacuation During the Floods of 1995 in The Netherlands. In Parker, D.J. Floods, Vol. 1, Section 22, 351-360, Routledge, London.

195. VIJVER, M.A. (Ph.D. Thesis, 2005) The ins and outs of bioaccumulation. Universiteit van Amsterdam, Amsterdam.

196. VINCKE, P. (1992). Multicriteria Decision-aid. Chichester, Wiley.

197. WAARTS, P. (1992) Methode voor de bepaling van het aantal doden als gevolg van inundatie. Report TNO B-91-1099.

198. WAGNER C, LØKKE H. (1991) Estimation of ecotoxicological protection levels from NOEC toxicity data. Water Res 25:1237–1242.

199. WHO-EUROPE (2002a) Flooding: Health effects and preventive measures, Fact sheet 05/02. Available: http://www.euro.who.int/document/mediacentre/fs0502e.pdf Accessed: 22/01/07.

200. WHO-EUROPE (2002b) Floods: climate change and adaptation strategies for human health. Report on a WHO meeting. London, UK, 30 June-2 July 2002. Available: http://www.euro.who.int/document/E77096.pdf Accessed: 22/01/07.

201. WILSON, T. (2006) Les risques des blessures et de décès par imprudence lors des inondations. Responsabilité and Environnement, 43, July 2006.

202. WIJDEVELD, A. (2006). Toxics. Prepared for EU 6th Framework. Rebecca project. Final version November, 2006.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 199 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

203. WIJDEVELD, A. VAN DEN ENDE, F, VAN NOORT, P., HAASNOOT, M. & PENAILILLO, R (2007). Toxic Stress: The development and use of the OMEGA modelling framework in a case study. Milestone report T10-07-14 for the Integrated project FLOODsite.

204. WOLGATER, M.S. & SILVER, N.C. (1990) Arousal strength of warning signal words, Forensic Reports, Vol. 3, 407-420.

205. ZHAI, G., FUKUZONO, T., and IKEDA, S. (2006) An empirical model of fatalities and injuries due to floods in Japan. Journal of the American Water Resources Association. 42 (4) 863-875.

206. ZIMMERMANN, H.-J. AND GUTSCHE, L. (1991). Multi-Criteria Analyse - Einführung in die Theorie der Entscheidungen bei Mehrfachzielsetzungen. Berlin, Heidelberg, New York, Tokyo, Springer Verlag.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 200 Task 10 Deliverable D10-1 Contract No:GOCE-CT-2004-505420

8 Appendix A: Data collection template

Characteristics Data requirement Explanation variables Area(s) Type of land use/ E.g. majority is rural, urban, industrial or mixed use characteristics29 spatial development Flood warnings Please state if warning systems are in place and, if so, whether The affected area systems systems are:- needs to be defined Good (where majority of people at risk received warning with by you but should adequate lead time) be a discrete area Fair (warning received by some of the population with adequate e.g. a small town or lead time) part of a town, a None (no warnings received) village etc. See note below. Information on warning lead time is also useful Type of buildings E.g. multi-storey apartments, single storey houses, multi-storey houses, commercial/ industrial properties, mobile homes, campsites, schools. Could give approximate % of each type or what type majority are. Rate of rise/speed of Low risk: onset very gradual over many hours onset Medium risk: onset is gradual , one hour or so High risk: rapid onset in minutes Building collapse Number or (preferably) % of buildings in area collapsed or destroyed Evacuation (Linked to flood warnings) Please state if people evacuated or not. If evacuated please state how many or % of people evacuated, also when evacuated, and duration Flood Details of flood Date, location/name of area(s) affected. characteristics Depth In metres Velocity In metres per second Debris content Size and type reported e.g. trees, rocks, cars Time of Flood Can be approximate e.g. night-time, morning, afternoon, evening. Information on if weekday, weekend or holiday period also useful. Duration Minutes, hours, days or weeks People Number of people Area(s) to be defined by you, see above characteristics affected in the area(s)/zones Number of deaths + data on cause of death needed where available Number of seriously + data on cause/type of serious injuries needed where available injured Age 75+ % population in area aged 75 or over

Health status % population in area with long-term illness/disability Population with % ethnic minorities, tourists, foreign workers language constraints Awareness of flood risk E.g. % recent migrants and/or % second home owners, % tourists

29 For some flood events the area characteristics may vary in terms of levels of hazard for different parts of the flood hazard area, for example for depth of flood or rate of rise. Where this is the case, it would be useful to have all the data disaggregated for the different flood ‘zones’. Data for different variables may also vary according to the zone of flooding e.g. for flood warnings, type of buildings etc.

T10_07_13_Methodologies_D10_1_V_1_3_P01.doc 29th February 2008 201