Areas at Risk – Concept and Methods for Urban Flood Risk Assessment: A Case Study of Santiago de Chile 351510092X, 9783515100922

Table of contents :
ACKNOWLEDGEMENTS
CONTENT
LIST OF FIGURES
LIST OF TABLES
LIST OF ABBREVIATIONS
1 INTRODUCTION
2 DISASTER RISK RELATED TERMS AND CONCEPTDEVELOPMENT
3 EXISTING FLOOD RISK-RELATED RESEARCH
4 DESCRIPTION OF THE STUDY AREA
5 DEFINITION OF A CASE SPECIFIC INDICATOR SET FOR THERISK ANALYSIS
6 DATA BASE AND DATA PROCESSING
7 REMOTE SENSING DATA ANALYSIS
8 HYDROLOGIC MODELING
9 FLOOD RISK ANALYSIS AND ASSESSMENT
10 PREVENTION AND MITIGATION MEASURES
11 DISCUSSION AND CONCLUSIONS
REFERENCES
APPENDIX 1 – ADDITIONAL TABLES AND CHARTS
APPENDIX 2 – COLOURED FIGURES

Citation preview

Annemarie Müller

Areas at Risk – Concept and Methods for Urban Flood Risk Assessment A Case Study of Santiago de Chile

Geographie

Megacities and Global Change Megastädte und globaler Wandel

Franz Steiner Verlag

Band 3

Annemarie Müller Areas at Risk – Concept and Methods for Urban Flood Risk Assessment

megacities and global change megastädte und globaler wandel herausgegeben von Frauke Kraas, Jost Heintzenberg, Peter Herrle und Volker Kreibich

Band 3

Annemarie Müller

Areas at Risk – Concept and Methods for Urban Flood Risk Assessment A Case Study of Santiago de Chile

Franz Steiner Verlag

Gedruckt mit freundlicher Unterstützung der Helmholtz-Zentrum für Umweltforschung (UFZ) GmbH

Umschlagabbildung: High water levels in the upper Mapocho River during winter rainfall © James McPhee

Bibliografische Information der Deutschen Nationalbibliothek: Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über abrufbar. Dieses Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist unzulässig und strafbar. © Franz Steiner Verlag, Stuttgart 2012 Druck: AZ Druck und Datentechnik, Kempten Gedruckt auf säurefreiem, alterungsbeständigem Papier. Printed in Germany. ISBN 978-3-515-10092-2

ACKNOWLEDGEMENTS I express my sincere thanks to Prof. Dr. Ulrike Weiland from the University of Leipzig and Prof. Dr. Frido Reinstorf from the Magdeburg-Stendal University of Applied Sciences for their supervision and guidance through the thesis work. I would also like to thank all members of the research initiative “Risk Habitat Megacity” in Germany and Chile and my colleagues at the Helmholtz Centre for Environmental Research (UFZ) in Leipzig for the support and fruitful discussions. Especially, I would like to thank Sonia Reyes, Dr. James McPhee, and Gerhard Schleenstein for their valuable help during my field stays and René Höfer, Christian Gadge, and Prof. Dr. Bernd Hansjürgens for their support at the UFZ. This research would not have been possible without the numerous experts from various institutions in Santiago de Chile that I interviewed. I appreciate their willingness to corporate, to provide insight in their work, and to answer all my questions. I likewise thank all the people interviewed during the household surveys for their participation and readiness to support this work. I sincerely thank Jessica Reiter for carrying out these field surveys and for delivering a valuable contribution to this thesis. I also thank Javiera Perez for sharing her research results and for allowing me to use them as input for my analyses. This work is supported by the Initiative and Networking Fund of the Helmholtz Association. I also thank the German Academic Exchange Service for financing my longest research stay in March and April 2009. I thank the HIGRADE office and the Helmholtz Centre for Environmental Research (UFZ) for financially supporting field stays and additional trainings. I am very grateful to my husband and my family for their continued support.

CONTENT 

List of Figures..................................................................................................11 List of Tables .................................................................................................. 13 List of Abbreviations ...................................................................................... 15 1 Introduction ................................................................................................. 19 1.1 Background and problem description ................................................... 19 1.1.1 Urbanization trends: Chile in a worldwide context ....................... 19 1.1.2 The influence of urban expansion on flood risk ............................ 20 1.1.3 Floods in Santiago de Chile ........................................................... 22 1.2 Research goals and research questions ................................................. 27 1.3 Methodology and structure of the thesis .............................................. 27 1.3.1 Development of a conceptual risk framework............................... 27 1.3.2 The choice of an appropriate study area ........................................ 28 1.3.3 The methodologies applied in this research .................................. 29 1.3.4 Exchange with stakeholders .......................................................... 31 2 Disaster risk related terms and concept development ................................. 33 2.1 A conceptual flood risk framework for a complex urban setting ......... 33 2.1.1 Identification of the components of risk ........................................ 33 2.1.2 Relation between the components of risk ...................................... 35 2.2 Hazard................................................................................................... 36 2.3 Elements at risk .................................................................................... 36 2.4 Vulnerability ......................................................................................... 37 2.4.1 Exposure and its role in the vulnerability concept ........................ 40 2.4.2 Resilience and coping capacities in the vulnerability concept ...... 40 2.5 Risk management ................................................................................. 41 2.6 General approaches to face flood risk .................................................. 42 3 Existing flood risk-related research ............................................................. 45 3.1 State of research: RS in flood risk analysis .......................................... 45 3.2 State of research: GIS in flood risk analysis ........................................ 48 3.3 State of research: Hydrological modeling for flood risk analysis ........ 49 3.4 Land-use/land-cover changes for flood risk reduction ......................... 51 3.5 Empirical methods ................................................................................ 52 3.6 Conclusion from the current states of research .................................... 53 4 Description of the study area ....................................................................... 55 4.1 Human-geographic description ............................................................ 55 4.1.1 Overview about location and administrative boundaries .............. 55 4.1.2 Social structure .............................................................................. 57 4.1.3 Urban development ....................................................................... 57 4.2 Planning institutions, instruments & processes relevant for flood risk 62 4.2.1 Institutions, laws, and instruments in the prevention stage ........... 62 4.2.2 Institutions, laws, and instruments in the mitigation stage ............ 64 CONTENT

8

Content

4.3 Physio-geographic description ............................................................. 65 4.3.1 Geology, Geomorphology, and Soil .............................................. 66 4.3.2 Climate .......................................................................................... 68 4.3.3 Hydrology ...................................................................................... 71 4.3.4 Vegetation and urban green spaces ................................................ 73 4.3.5 In-depth study area: Quebrada San Ramón................................... 74 5 Definition of a case specific indicator set for the risk analysis ................... 77 5.1 Function and characteristics of indicators ............................................ 77 5.2 Selection of variables relevant for the analysis of flood hazard ........... 77 5.3 Selection of variables referring to the elements at risk ........................ 79 5.4 Selection of vulnerability-related variables .......................................... 79 5.5 Bringing together the relevant variables............................................... 81 5.6 From descriptive variables to indicators............................................... 84 6 Data base and data processing ..................................................................... 87 6.1 Hydro-meteorological data base ........................................................... 88 6.2 Processing of the hydro-meteorological data ....................................... 89 6.3 GIS and remote sensing data base ........................................................ 91 6.4 Pre-processing of the GIS data ............................................................. 92 6.5 Pre-processing of the remote sensing data ........................................... 93 6.5.1 Geometric correction and co-registration ...................................... 93 6.5.2 Data coverage and data gaps ......................................................... 93 6.6 Socio-economic data base .................................................................... 94 6.7 Pre-processing of the socio-economic data .......................................... 95 6.8 Expert interviews, household surveys, and questionnaires .................. 96 6.8.1 Expert interviews ........................................................................... 96 6.8.2 Household surveys......................................................................... 96 6.8.3 Questionnaires ............................................................................... 97 6.9 Processing of the empirical data ........................................................... 98 6.9.1 Expert interviews ........................................................................... 98 6.9.2 Household surveys......................................................................... 99 6.9.3 Questionnaires ............................................................................... 99 7 Remote sensing data analysis .................................................................... 101 7.1 The theory of remote sensing-based land-use classifications ............. 101 7.1.1 Pixel-based classification ............................................................ 102 7.1.2 Object-oriented classification ...................................................... 103 7.2 Classification of the ASTER images .................................................. 107 7.2.1 Classification process .................................................................. 108 7.2.2 Classification results .....................................................................110 7.2.3 Accuracy assessment .................................................................... 111 7.3 Classification of the Quickbird image .................................................113 7.3.1 Classification process ...................................................................113 7.3.2 Classification results .....................................................................117 7.3.3 Accuracy assessment ....................................................................117

Content

9

8 Hydrologic modeling ..................................................................................119 8.1 Theoretical background and selection of methods ..............................119 8.1.1 Relevant processes with respect to LULC changes ......................119 8.1.2 Theoretical background of hydrological models ......................... 120 8.1.3 Selection of model and methods.................................................. 122 8.2 Data preparation with HEC-GeoHMS................................................ 126 8.2.1 Terrain pre-processing ................................................................. 128 8.2.2 Hydrologic processing ................................................................. 128 8.2.3 Delineation of Hydrological Soil Groups (HSG) ........................ 130 8.2.4 Creation of a gridfile matching the SHG ..................................... 131 8.2.5 HMS Project Setup ...................................................................... 133 8.3 Hydrological modeling using HEC-HMS .......................................... 134 8.3.1 Model description ........................................................................ 134 8.3.2 Parameterization .......................................................................... 135 8.3.3 Simulation.................................................................................... 139 8.3.4 Sensitivity analysis ...................................................................... 140 8.3.5 Model optimization ..................................................................... 143 8.4 Modeling alternative land-use/land-cover scenarios .......................... 146 8.4.1 The characteristics and functions of scenarios ............................ 146 8.4.2 Land-use/land-cover scenarios for the San Ramón basin............ 147 9 Flood risk analysis and assessment ........................................................... 153 9.1 Delineation of information to feed the indicators............................... 153 9.2 Flood hazard assessment .................................................................... 155 9.2.1 The relation between runoff and flood extents ............................ 155 9.2.2 Quantification of the hazard ........................................................ 157 9.2.3 Analysis of the hazard maps ........................................................ 158 9.3 Assessment of the elements at risk ..................................................... 159 9.3.1 Quantification of the elements at risk .......................................... 159 9.3.2 Analysis of the map of elements at risk ....................................... 160 9.4 Flood vulnerability assessment........................................................... 161 9.4.1 Evaluation of the vulnerability-related variables ........................ 161 9.4.2 Quantification of vulnerability .................................................... 165 9.4.3 Analysis of the vulnerability maps .............................................. 166 9.4.4 Sensitivity analysis for the weighting.......................................... 168 9.5 Comprehensive risk analysis .............................................................. 169 9.6 The development and assessment of flood risk .................................. 171 9.6.1 Number of new residential sites in flood-prone areas ................. 171 9.6.2 Number of people living in flood-prone areas ............................ 172 9.6.3 Development of the components of risk in the future ................. 173 10 Prevention and mitigation measures ........................................................ 177 10.1 Analysis of the previous deficits....................................................... 177 10.2 Flood risk scenarios .......................................................................... 180 10.2.1 Scenario I  Increasing aridity .................................................. 180 10.2.2 Scenario II  Afforestation ........................................................ 182 10.2.3 Scenario III  Construction activities ........................................ 183

10

Content

10.3 Recommendations ............................................................................ 185 10.3.1 Prevention measures .................................................................. 185 10.3.2 Mitigation measures .................................................................. 189 10.4 Taking action .................................................................................... 192 11 Discussion and conclusions ..................................................................... 193 11.1 Discussion ......................................................................................... 193 11.1.1 Vulnerability .............................................................................. 193 11.1.2 Hazard ........................................................................................ 195 11.1.3 Elements at risk ......................................................................... 197 11.1.4 Risk ............................................................................................ 198 11.1.5 The applicability of the concept for multi-hazard studies ......... 199 11.1.6 The spatial transferability of the concept................................... 200 11.1.7 The role of flood risk and flood risk awareness ........................ 201 11.1.8 Further ecological consequences of urban growth .................... 202 11.2 Conclusions ...................................................................................... 203 References .................................................................................................... 207 Appendix 1 – Additional Tables and Charts ................................................. 221 Appendix 2 – Coloured Figures ................................................................... 235

LIST OF FIGURES Figure 1:Schematic diagram of the water cycle .................................................... 21 Figure 2:The impact of land-use/land-cover changes on flood risk ...................... 22 Figure 3:Situation after a flood event in La Reina, June 2008.............................. 24 Figure 4: Location of the in-depth study area ....................................................... 29 Figure 5: Work flow. ............................................................................................. 30 Figure 6: The components of risk ......................................................................... 35 Figure 7: The Pressure and Release Model ........................................................... 38 Figure 8: Disaster risk management cycle ............................................................ 41 Figure 9: Construction site above 1,000 m ........................................................... 61 Figure 10: Cross-sectional profile of the main geological units ........................... 66 Figure 11: Average temperature and precipitation in Santiago ............................. 69 Figure 12: Main hydrological network in the Metropolitan Region ..................... 72 Figure 13: The influence of aspect on vegetation coverage .................................. 73 Figure 14: Geologic map of the San Ramón catchment ....................................... 75 Figure 15: Biodiversity and water intake station in the catchment. ...................... 75 Figure 16: Schema of variables relevant for flood risk analysis ........................... 83 Figure 17: Overview about input data and their usage ......................................... 88 Figure 18: Average evaluation of each vulnerability variable by experts ........... 100 Figure 19: The system of image levels with image segments ............................. 104 Figure 20: Overview about hydrological simulation models .............................. 120 Figure 21: Process flow for the application of HEC-HMS & HEC-GeoHMS. .. 127 Figure 22: Basin model of the catchment of Quebrada San Ramón. .................. 133 Figure 23: Exemplary result from the runoff simulation .................................... 140 Figure 24: Map of subbasins. .............................................................................. 141 Figure 25: Result of the hydrologic modeling using HEC-HMS I ..................... 145 Figure 26: Result of the hydrologic modeling using HEC-HMS II .................... 145 Figure 27: Land-use/land-cover data based on classification results .................. 149 Figure 28: Overview about people in the AMS in flood hazard zones ............... 172 Figure 29: Schema of variables relevant for flood risk analysis ......................... 175 Figure 30: Congestion of the storm water infrastructure .................................... 179 Figure 31: Schema of variables relevant for flood risk prevention..................... 188 Figure 32: Schema of variables relevant for flood risk mitigation ..................... 190 Figure 33: Recent construction activities in San Ramón channel ....................... 191 Figure A34: Workflow for the delineation of the ASCII-file ............................ 226 Figure A35: The urban expansion between 1970 and 2000 .............................. 235 Figure A36: The administrative units of the Metropolitan Region ................... 236 Figure A37: The administrative units in Santiago de Chile .............................. 237 Figure A38: LULC changes in La Reina and Peñalolén ................................... 238 Figure A39: Example for urban land-use/land-cover changes .......................... 239 Figure A40:. Location of gaging stations .......................................................... 240 Figure A41: Overview over the available satellite images. .............................. 241

12

List of Figures

Figure A42: Daily precipitation values. ............................................................ 242 Figure A43: The process of the object-oriented classification approach .......... 242 Figure A44: Land-use/land-cover classification results ASTER. ..................... 243 Figure A45: Land-use/land-cover classification results Quickbird .................. 244 Figure A46: Standard Hydrological Grid .......................................................... 245 Figure A47: Hydrograph for the rainfall event in July 2001. ........................... 246 Figure A48: Hydrograph for the rainfall event in May 2002. ........................... 246 Figure A49: Hydrograph for the rainfall event in June 2002. ........................... 247 Figure A50: Hydrograph for the rainfall event in June 2005. ........................... 247 Figure A51: Hydrograph for the rainfall event in August 2005. ....................... 248 Figure A52: Hydrograph for the first rainfall event in May 2008. ................... 248 Figure A53: Hydrograph for the second rainfall event in May 2008. ............... 249 Figure A54: Hydrograph for the rainfall event in September 2009. ................. 249 Figure A55: Indicator sheet: Street level. ......................................................... 250 Figure A56: Indicator sheet: Construction material .......................................... 251 Figure A57: Indicator sheet: Green spaces. ...................................................... 252 Figure A58: Indicator sheet: 2.5 people sharing one bedroom ......................... 253 Figure A59: Indicator sheet: No or incomplete basic education ....................... 254 Figure A60: Indicator sheet: No employment or no income............................. 255 Figure A61: Indicator sheet: Female population. .............................................. 256 Figure A62: Indicator sheet: People below five or above 65 years .................. 257 Figure A63: Flood hazard map for a 100 year return period ............................ 258 Figure A64: Changes in the extent of flood hazard zones ................................ 258 Figure A65: Available flood hazard maps for La Reina.................................... 259 Figure A66: Distribution of people, built-up area and infrastructure ............... 260 Figure A67: Distribution of the combined elements at risk .............................. 260 Figure A68: Overview about location of the surveyed households .................. 261 Figure A69: Vulnerability map based on evaluation of experts. ....................... 261 Figure A70: Vulnerability map based on the evaluation households. ............... 262 Figure A71: Normalized risk map for La Reina and Peñalolén (PRMS) ......... 262 Figure A72: Normalized risk map for La Reina and Peñalolén (Perez 2009) .. 263 Figure A73: Normalized risk map for La Reina and Peñalolén (Scenario B2). 263 Figure A74: Development of new construction in flood hazard zones (93-02) 264 Figure A75: Development of new construction in flood hazard zones (02-09).265

LIST OF TABLES Table 1: Methods for flood risk-related research. ................................................. 46 Table 2: Selected terms in the context of urban planning in Santiago. ................. 56 Table 3: Population characteristics in the Metropolitan Area ............................... 57 Table 4: Overview about institutions relevant for prevention and mitigation. ..... 64 Table 5: Probabilities for the daily maximum precipitation after Perez (2009) .... 71 Table 6: Probabilities for the daily maximum precipitation after AC (2009). ...... 71 Table 7: Location of the 0°C isotherm in the RM ................................................. 76 Table 8: Set of variables used for the flood risk analysis ...................................... 82 Table 9: Variables and resulting indicators for the analysis of risk. ...................... 84 Table 10: Data needs for flood risk assessment in the present research. .............. 87 Table 11: Overview over available hydro-meteorological data. ........................... 89 Table 12: Vector data used for this study............................................................... 91 Table 13: Raster data used for this study. .............................................................. 91 Table 14: Census variables available from 2002 survey. ...................................... 94 Table 15: LULC classes and their content description (ASTER data)................ 109 Table 16: Results from the accuracy assessment of the ASTER data 2002 ......... 111 Table 17: Results from the accuracy assessment of the ASTER data 2005 .........112 Table 18: Results from the accuracy assessment of the ASTER data 2009 .........112 Table 19: Segmentation parameters. ....................................................................113 Table 20: LULC classes and their content description (Quickbird data) .............114 Table 21: Results from the accuracy assessment of the Quickbird data. .............118 Table 22: Comparison of deterministic and stochastic models ........................... 121 Table 23: Parameters used for the hydrologic pre-processing ............................ 129 Table 24: Geologic formations and hydrological soil groups ............................. 132 Table 25: ASTER classification results with their hydrological properties ........ 132 Table 26: Selected hydrologic elements available in HEC-HMS ....................... 134 Table 27: Global parameters used for the hydrologic model HEC-HMS. .......... 135 Table 28: The influence of the CN values on the model output .......................... 143 Table 29: Maximum daily precipitation intensities [mm]. .................................. 143 Table 30: Modeling results for the San Ramón catchment. ................................ 144 Table 31: Modeling results for the three LULC scenarios .................................. 150 Table 32: Definition of flood risk levels. ............................................................ 158 Table 33: Results from the sensitivity analysis ................................................... 168 Table A34: Overview about floods in Santiago between 1990 and 2002. ........ 222 Table A35: Projection transformation parameters. ........................................... 222 Table A36: Partners for the expert interviews during field work...................... 223 Table A37: Preprocessing steps to be performed using HEC-GeoHMS........... 224 Table A38: Albers projection file (prj.adf) ........................................................ 225 Table A39: Header information of Curve Number (CN) ASCII-file ................ 225 Table A40: Rule set used for the LULC classification of the Quickbird data. . 227 Table A41: Individual parameters used for subbasins in HEC-HMS ............... 232

14

List of Tables

Table A42: Sensitivity analysis for Ia and Retention scale factor. .................... 233 Table A43: Runoff volume for San Ramón catchment [m³/s] .......................... 234

LIST OF ABBREVIATIONS ADPC AMS B CN CONAMA COREMA CRED DEM DGA DLR DOH DOM DSM DSS Dur EaR-I EM-DAT ET GIS GMES HEC-DSSVue HEC-HMS HH HI HQ100 HSG Ia IDW IGM INE IPCC IR IS

LIST OF ABBREVIATIONS Asian Disaster Preparedness Centre Área Metropolitana de Santiago de Chile, Metropolitan Area of Santiago de Chile Built-up Area Curve Number Comisión Nacional de Medio Ambiente, National Environmental Commission Comisión Regional de Medio Ambiente, Regional Environmental Commission Centre for Research on the Epidemiology of Disasters Digital Elevation Model Dirección General de Aguas, Central Water Directorate German Aerospace Centre Dirección de Obras Hidráulicas, Hydraulic Works Directorate Departamento de Obras Municipales, Municipal Works Department Digital Surface Model Decision Support System Duration of rainfall event [in hrs] Normalized Elements-at-Risk Index Emergency Events Database Evapotranspiration Geographic Information System Global Monitoring for Environment and Security HEC Data Storage System Visual Utility Engine Hydrologic Engineering Center  Hydrologic Modeling System Household Normalized Hazard Index Flood with a return period of 100 years Hydrological Soil Group Initial abstraction ratio Inverse Distance Weighted Instituto Geográfico Militar, Geographic Institute of the Military Instituto Nacional de Estadísticas de Chile, National Statistics Institute of Chile Intergovernmental Panel on Climate Change InfraRed Infrastructure

16 LGUC LUCK LULC LUT MINAGRI MINVU MLC MOP ms MTT n.a. NDVI NEXRAD NRCS ONEMI OOA OTAS

P pan PAR-Modell Pct P-GIS PMALS Pmax Pop PRC PRMS Protege PSAD Ptotal Q Qmax R

List of abbreviations

Ley General de Urbanismo y Construcciones, General Law for Urbanism and Construction Land Use Change Modelling Kit Land Use/Land Cover Lookup Table Ministerio de Agricultura, Ministry for Agriculture Ministerio de Vivienda y Urbanismo, Ministry for Housing and Urbanism Maximum Likelihood Classifier Ministerio de Obras Públicas, Public Works Ministry multi-spectral Ministerio de Transportes y Telecomunicaciones, Ministry for Transport and Telecommunication not available Normalized Differenced Vegetation Index Next-Generation Radar Natural Resources Conservation Service Oficina Nacional de Emergencia, National Emergency Office Object-oriented Analysis Bases para un Ordenamiento Territorial Ambientalmente Sustentable en la Región Metropolitana de Santiago, Basis for an Environmentally Sustainable Territorial Planning in the Metropolitan Region of Santiago de Chile Precipitation pansharpened Pressure and Release Modell Percentage Participatory GIS Plan Maestro de Evacuación y Drenaje de Aguas Lluvias del Gran Santiago, Master Plan for the Evacuation and Drainage of Storm Water for Gran Santiago Maximum precipitation within 24 hours [in mm] unless otherwise noted Population Plan Regulador Comunal, Municipal Regulatory Plan Plan Regulador Metropolitano de Santiago, Metropolitan Regulatory Plan Proyecto de Conservación y Protección de la Cordillera de Santiago de Chile, Project for the Conservation and Protection of the Cordillera in Santiago de Chile Provisional South American Ellipsoid Total precipitation [in mm] Runoff Maximum runoff [in m³/s] Storage coefficient

List of abbreviations

RIMAX

17

Risikomanagement extremer Hochwasserereignisse, Risk Management of Extreme Flood Events RM Región Metropolitana, Metropolitan Region RS Remote Sensing SCS Soil Conservation Service SECPLAN Secretaría de Planificación, Planning Secretariat SEIA Sistema de Evaluación del Impacto Ambiental, System for the Evaluation of Environmental Impact SEREMI Secretaría Regional Ministerial, Regional Ministerial Secretariat SERNAGEOMIN Servicio Nacional de Geología y Minería, National Service for Geology and Mining SERVIU Servicio de Vivienda y Urbanismo, Office/service for housing and urbanism SHG Standard Hydrological Grid SMCA Spatial Multi-Criteria Analysis SRTM Shuttle Radar Topography Mission St Storage SUBDERE Subsecretaría de Desarollo Regional y Adminstrativa, Subsecretary of Regional and Administrative Development (Ministry of the Interior) tc Time of concentration, travel time TTA Test and Training Areas UH Unit Hydrograph UN-ISDR United Nations International Strategy for Disaster Reduction USACE United States Army Corps of Engineers USDA United States Department of Agriculture USGS United States Geological Survey UTC Coordinated Universal Time UTM Universal Transverse Mercator VI Normalized Vulnerability Index WFS Web Feature Service WGS World Geodetic System WPS Web Processing Service

1 INTRODUCTION 1.1 BACKGROUND AND PROBLEM DESCRIPTION Between January 2008 and September 2010  the approximate time frame for this thesis work  floods affected 133,730,961 people and caused a monetary damage of 39,372,391 US$ worldwide (CRED 2010). Most of the damage occurred because people settled in flood prone areas and interrupted the natural catchment and flowpath of the waterways in the course of urban expansion. 1.1.1 Urbanization trends: Chile in a worldwide context Urbanization and especially mega-urbanization is one of the most prominent phenomena and at the same time one of the great challenges of the 21st century (Hansjürgens et al. 2008, Hazel & Miller 2006, Kraas et al. 2005). The percentage of urban population worldwide grew from 29.0 % of the total population in 1950 to 48.7 % in 2005. Estimations predict nearly 60 % of the total population to live in urban areas within the next two decades (UN Habitat 2008). The current average urban growth rate in developing countries is 5 million people per month (UN Habitat 2008). The geographic focus of this research is on Latin America and Chile because Latin America is with 77 % (UN Habitat 2008) the most urbanized developing region in the world. In some Latin American countries, the urban population has reached 90 % and beyond (ECLAC/CEPAL 2000). In Chile, 87 % of the population lived in urban areas in 2008 (INE 2008). Urban growth, and particularly the development of megacities, is a phenomenon of growing importance for social and environmental scientists (UN Habitat 2008, Hansjürgens et al. 2007, UNDP 2004). Megacities can be defined according to their size (minimum population 5 million and upwards depending on the definition) or according to their function as an economic, political, and cultural center of the country that has global importance (Hansjürgens et al. 2008, Wenzel et al. 2007, Bronger 1996). Megacities concentrate demographic and economic growth, resources, and labor and do therewith provide “a space of opportunity”. At the same time, they are “a space of risk”, as the development of mega-agglomerations with high densities of buildings, urban infrastructure, industry, and people poses a number of risks for the inhabitants and for protected goods (Büscher 2008, Romero et al. 2008, UN Habitat 2008, Hansjürgens & Heinrichs 2007, Wenzel et al. 2007, Heinrichs & Kabisch 2006, Romero & Vásquez 2005). The study area for this research is Santiago de Chile, the capital city of Chile and a megacity because of its high dynamics and functional value. It is the political and economic center of the country with approximately 6.7 million inhabitants (INE 2008) and it is undergoing a rapid process of urbanization with changes in

20

1 Introduction

land use and urban morphology in a planned but also in an informal way (Romero et al. 2008). With respect to the speed of growth and development in combination with the insufficiently regulated urban planning, Santiago de Chile can be regarded as a representative example for urbanization and mega-urbanization in a Latin American country (Hansjürgens et al. 2007). Its urban area grew from 31,419 ha in 1970 to 64,140 ha in the year 2002 (Figure A35). The area open for urbanization after the Metropolitan Regulatory Plan (Plan Regulador Metropolitano de Santiago, PRMS) has been as large as 110,319 ha since 2007. Thus, a continuation of the spatial expansion of the city can be expected during the next years. More detailed information about the past urbanization process of Santiago is given in Chapter 4.1.3. 1.1.2 The influence of urban expansion on flood risk Spatially seen, urban growth is frequently associated with an expansion of impervious surfaces, the loss of natural areas, and therefore a severe impact on the ecosystem. Agricultural land is converted into residential areas; bushland makes way for transport infrastructure and recreation areas. Land-use/land-cover (LULC) changes frequently lead to an increase of the occurrence of hazardous events, such as floods, landslides, and heat stress (Romero et al. 2008, Ducci & González 2006, Solway 2004, Pelling 2003, Bronstert et al. 2002). Accompanying the issue of urban expansion is the lack of ecological and environmental awareness in the urban planning process and the deficit of adequate planning instruments and competencies. The time lines of physical growth and the necessary adaption of urban governance structures do in many cases diverge, resulting in malfunctioning steering and control mechanisms with respect to urban planning. Hansjürgens et al. (2008) discuss that large urban agglomerations face higher risks to experience damage from hazardous events as they are constructed in a complex and dense way. The factors relevant for the generation of flood risk have not been quantified yet, thus are not sufficiently known. Rather, adaptation measures were taken without considering the root causes of hazard generation on the one hand and the susceptibility to experience damage on the other hand. The issues resulting from urban expansion with regard to flood risk are twofold and represent a lack of sustainable urban planning: The first impact affects the urban water cycle (Figure 1) and the flood hazard (read the box “The water cycle” in Appendix 1), the second affects the increasing exposure of people, properties, infrastructure, ecologic, and economic values (in the following only referred to as values) in flood-prone areas. The changes in LULC lead to a reduction of infiltration and interception capacities and a higher amount of surface runoff. The likelihood of a flood event (i.e. the flood hazard) is consequently increasing in those watersheds that are prone to or subject to anthropogenic disturbances. In the course of urbanization, the construction of mitigation measures for extreme flood events and the setup of flood risk management programs provide the notion of security.

1.1 Background and problem description

21

Figure 1: Schematic diagram of the water cycle (Laude Horton Watkins Highschool 2010).

This false sense of security, the lack of alternatives or the pure lack of knowledge lets people perpetually get settled in flood-prone areas along river beds and floodplains (GFZ 2006). Thus, besides the increasing flood hazard, urban expansion leads to a higher exposure of people and values to these potentially damaging events  and thus to an increase of risk. The schematics in Figure 2 demonstrate these relationships and consequences. Official disaster statistics show a fluctuating but increasing amount of reported flood events and an increasing number of people affected. The largest share in terms of the total number of disasters and affected people do in fact have hydrometeorological disasters (floods, wet mass movements, and storms (Vos et al. 2009), p. 5). Millions of people, especially in Asia, are affected each year. However, the number of fatalities has been relatively constant during the last years and shows a decreasing trend in the Americas (Vos et al. 2009), which expresses better disaster preparation and disaster risk management.

22

1 Introduction

Major hydrological processes during storm. Urban growth and new land-use patterns.

Less infiltration capacities lead to higher surface runoff.

Higher surface runoff and more people and buildings exposed result in high flood risk.

Figure 2: The impact of land-use/land-cover changes on flood risk in the scope of urban expansion. Schematic representation of the main influencing factors and consequences.

1.1.3 Floods in Santiago de Chile Floods are a regularly occurring phenomenon in Santiago de Chile (Table A34 in the appendix). It is now examined what type of floods occur, how they can be characterized, and what research was carried out with respect to flood generation and flood risk.

1.1 Background and problem description

23

1.1.3.1 Triggers for floods Three main triggers lead to floods in the Metropolitan Region: -

Water emerging to the surface arising from a high-lying groundwater table, overflow of rivers and channels due to high water levels in the water course (river floods), accumulation of storm water on the roads and in local depressions, e.g. under bridges (urban floods).

A map showing the flood hazard zones of Santiago de Chile was created by Fernández & Montt (2004). In this study, river floods and urban floods are of interest. During orographic storms with high precipitation values especially in the Andean foothills the Metropolitan Region experiences floods. The increasing amount of sealed surface is adding to the quantity and velocity of surface runoff into the lower-lying parts of the city, but also in the urbanized hills, what leads to even larger areas of potential flood risk (Hansjürgens et al. 2007, Reyes 2003). The flood hazard has been increasing additionally through inappropriate urban planning in the past and partly still today. A complete canalization system only exists in the historic city center as the construction of storm water infrastructure has for a long period been regarded as irrelevant. Today, storm water sewers have to be built in all new construction sites. To construct an appropriate storm water sewer system for the RM, 1,500 million US$ would have to be invested. Estimations claim that Santiago is losing 43 million US$ annually under the assumption that 50,000 people are affected by floods every year (Reyes 2003). The Hydraulic Works Directorate (DOH) identified 117 points in 33 municipalities where floods occurred (Fernández & Montt 2004). 1.1.3.2 Characteristics of floods The floods in Santiago are typically slow (i.e. no flash floods), carrying large amounts of sediment and branches from deforested slopes that remain on the streets after the water disappears (Figure 3, Reyes 2003, Silva 2008). In contrast to other countries, such as the Philippines, where urban floods have a severe impact on the economic growth of the area, floods in Santiago de Chile do not have such a very severe negative economic impact. The floods do rather occur regularly (almost annually) and people have in most cases adapted to a certain hazard level or adopt as soon as they have suffered damage once (Reiter 2009). Nevertheless, floods  even with a lower intensity  cause monetary costs as well as alternative costs, e.g. if people are hindered going to work. The flood height seldom exceeds 20 cm, but regularly interrupts the urban functioning and harms vulnerable households in one way or the other. Most frequently, front yards, outer walls, floors, and furniture are affected, moisture remains in the walls (physical damage) or people are trapped in gated communities and cannot go to work (immaterial and economic damage) (Reiter 2009). Also, floods in other regions of the world are the disasters that affect the highest numbers of fatalities per event. That is not the case in

24

1 Introduction

the study area. Even though there are single events that affect a very high number of people, the average flood does fortunately not cause that many victims. A field survey conducted by Reiter (2009) showed that 69 % of interviewed people in the study area declared that they suffered immaterial damage after a flood with limited mobility being the most important concern. Sufficient food, energy, and water supply, health and mental problems resulting from moisture in walls and floor, as well as fear are other issues (compare Section 9.4.1.2). With all the negative impacts, the importance of flood hazard diminishes during the dry times of the year where problems like crime and job security or earthquakes prevail (Reiter 2009).

1

Helpers evacuating water from the streets.

Wall of sandbags protecting property from damage.

Sediment load remaining on the streets after water disappears.

Small branches accumulating along the sandbags.

Figure 3:Situation one day after a flood event in the municipality of La Reina, June 2008.

1.1 Background and problem description

25

1.1.3.3 Flood risk related research done in Santiago de Chile Hazard maps for floods resulting from natural stream and canal overflow, high ground water tables, and accumulation of storm water on streets are published in a study carried out by Ayala et al. in 1987. Maximum runoff values were determined based upon a hydrological analysis. Return periods of 2, 5, 20, 25, 50, and 100 years were considered. The study also comprises a hydraulic study for all streams and creeks. Associated to the results is a proposal for appropriate land-use planning and official hazard zoning. Respective maps were incorporated in the fundamental planning instrument, the Metropolitan Regulatory Plan (PRMS) of Santiago de Chile. The study was carried out by order of the Ministry of Housing and Urbanism (MINVU) and is referenced as such in all flood hazard-related maps of this study. The hazard maps developed by Ayala et al. (1987) were updated in 1996 (Ayala & Cabrera 1996). Nevertheless, the new information has not yet been incorporated in the PRMS as it has not yet been renewed with respect to hazard zoning (Paredes 2009). Flood hazard analyses were furthermore conducted in the scope of previous research projects, such as OTAS (Bases para un Ordenamiento Territorial Ambientalmente Sustentable en la Región Metropolitana de Santiago, Basis for an Environmentally Sustainable Territorial Planning in the Metropolitan Region of Santiago de Chile, map at the scale of 1:100,000, (Ferrando 1998, Ferrando & Cueto 1998)), and by SERNAGEOMIN (Servicio Nacional de Geología y Minería, National Service for Geology and Mining (Antinao et al. 2003)). These studies are based on findings from older studies and did not deliver updated hazard maps. To develop solutions for the still prominent issue of floods in Santiago, a master plan for storm water evacuation and drainage (Plan Maestro de Evacuación y Drenaje de Aguas Lluvias del Gran Santiago, Master Plan for the Evacuation and Drainage of Storm Water for Gran Santiago (CADE IDEPE 2001)) was developed by order of the Public Works Ministry (MOP), DOH in 2001 (CADE IDEPE 2001) (compare Section 4.2). An updated flood hazard map comprising information on location, causes, and effects of previous flood events was created by the Departamento de Ingeniería Hidráulica y Ambiental Universidad Católica de Chile in cooperation with Mapcity and DICTUC (Ingenería DICTUC 2005, Mapcity 2005, Fernández & Montt 2004). Respective information about previous flood events was overlaid for the entire city area (for details refer to Fernández & Montt 2004). The most important sources of information were the areas listed in the Plan Maestro de Evacuación y Drenaje de Aguas lluvias del Gran Santiago, verified and updated using data of the Servicio de Vivienda y Urbanismo (SERVIU, Office/Service for Housing and Urbanism), municipalities, the National Emergency Office (ONEMI), the city hall, and press releases (Fernández & Montt 2004). This map, however, has also not been used to update the PRMS. Detailed flood hazard studies for those parts of the eastern municipalities that are located below 1,000 m at the Andean piedmont were published in 2008 (AC

26

1 Introduction

Ingenieros 2008) although at least the interviewed decision makers in the respective municipalities did not obtain the results of this investigation (Quezada 2009). Fuentes & Romero (2007) carried out a LULC change detection analysis for three river catchment along the Andean piedmont. The goal was to investigate the influence of LULC changes in the course of urban expansion on the storm water runoff behavior. The study analyses the urban growth towards the Andean mountains by means of remote sensing and GIS analysis in the time frame between 1975 and 2007. The authors applied the Curve Number method to quantitatively estimate the changing runoff coefficients in the area of interest. Even though the chosen approach implies some methodological drawbacks it is found that in the catchment areas of the three water courses the runoff coefficients increased by one tenth every year. This finding can most likely be transferred to the present study area and plays a role in the further course of this research. In addition, several studies on flood hazard assessment in Santiago de Chile were carried out in the scope of Master and Diploma theses. Perez (2009) investigated the influence of climate change on flood hazard in the catchment of Quebrada San Ramón applying the hydraulic models HEC-RAS and MOUSE for hazard analysis. The results from the hazard analysis were combined with data on LULC for a generalized vulnerability analysis to form a flood risk map. The role of regional climate change was investigated in this context. These findings are partly implemented in this research and further discussed in Chapters 6 and 9. 1.1.3.4 Research gaps with respect to flood risk Despite the large number of flood related studies in Santiago, except of the work of Perez (2009) who investigated the influence of climate change on flood hazard, all flood hazard and risk studies are solely based on the analysis of previous events. Changes in LULC that are currently taking place and that are expected to continue occurring in the future were so far only evaluated by Fuentes & Romero (2007). Even though this study delivers valuable insights into the changes taking place in the urban part of the catchment, this study does primarily focus on current runoff coefficients and does not yet make any predictions about the future. That means that the influence of LULC changes associated with the ongoing urban expansion on flood risk has not been quantified yet. Besides that, more comprehensive flood risk studies are so far only focused on land use as the only determinant of vulnerability. More specific research on coping capacities and exposure issues to further characterize and analyze vulnerability towards floods has not yet been carried out. Recommendations to minimize flood risk are proposed in the before mentioned Plan Maestro de Evacuación y Drenaje de Aguas Lluvias del Gran Santiago but focus on mitigation rather than on prevention measures (see Section 2.5). What is lacking so far for the study area is a comprehensive, systemoriented, and integrated approach to analyze and assess flood risk with its causes and interdependencies of determining factors.

1.2 Research goals and research questions

27

1.2 RESEARCH GOALS AND RESEARCH QUESTIONS As the analysis and assessment of risk and its causes are crucial prerequisites for the development of risk prevention measures, the goals of this research are to investigate the annual problem of floods in Santiago de Chile in a comprehensive way and to develop and apply a methodology to analyze and assess flood risk. The main focus is thereby set on the influence of LULC changes in the scope of urban expansion on flood risk in the study area. It is thus another research goal to get insight into institutions and instruments relevant for urban planning. With respect to the goals of this research, a preliminary analysis of the problem of floods in Santiago de Chile, and a revision of the research that has previously been conducted on the subject, the following key research questions were formulated: -

Which conceptual framework can be used to capture, analyze, and assess the flood risk in a growing urban agglomeration? (Chapter 2) What are the components influencing the flood risk in the study area on the communal and household level? (Chapter 5) How can indicators be used as a tool for flood risk analysis? (Chapters 5 and 9) What is the influence of LULC types and their changes on flood risk in the study area? (Chapters 8 and 9) How can flood risk be assessed in the study area? (Chapters 9 and 11) What measures concerning LULC could be taken to decrease flood risk? (Chapter 10) 1.3 METHODOLOGY AND STRUCTURE OF THE THESIS

Flood risk has its origins on various dimensions that are sometimes hard to capture and to describe precisely and even harder to measure and to evaluate. The deficit of most projects is that they are for the most part focused on just one component of risk and that they lack the multi-disciplinary character that risk studies require. This study connects both physical and social aspects in order to get a comprehensive understanding of the generation of flood risk in Santiago de Chile. The main steps are described in the following subsections. 1.3.1 Development of a conceptual risk framework Chapter 2 discusses the various definitions of risk, hazard, and vulnerability as well as the approaches for their practical assessment. It was found necessary to adapt the existing conceptual frameworks to match the specific conditions of a complex, megaurban environment and to show at the same time how a conceptual model can be operationalized. The concept proposed here can be formulated as: Risk = Hazard * Elements at Risk * Vulnerability.

28

1 Introduction

The main steps of this investigation are oriented on the 8-step-approach for vulnerability assessment, which is adapted to be a 9-step-approach for risk assessment after Polsky et al. (2003), p. 5): -

Select people and places carefully: choose scale, select stakeholders (Chapter 4) Get to know places over time: review literature, contact researchers, spend time in field, explore nearby areas (Chapter 4) Hypothesize who is at risk and why: identify people, identify drivers (Chapter 5) Develop a causal model of risk: describe factors, describe pathways, examine adaptation, formalize into a model (Chapters 2 and 5) Find indicators for the components of risk: exposure indicators, sensitivity indicators, adaptive capacity indicators (Chapter 5) Feed indicators: collect data, review different data sources, conduct interviews, bring data in suitable format to match the indicators (Chapters 6 to 9) Weight and combine the indicators: combine rigorously, represent results, validate results (Chapter 9) Project future risk: choose scenarios, run model (Chapters 8 and 9) Communicate risk creatively: be rigorous about uncertainty, trust stakeholders, use multiple media (Chapters 10 and 11) 1.3.2 The choice of an appropriate study area

The flood risk analysis was not carried out for the entire city of Santiago de Chile but for a specific study area in the eastern part of the city: the Quebrada San Ramón (Figure 2). The study area was selected for its location along the Andean foothills and its frequent affectedness through floods. The catchment and flowpath of the Quebrada San Ramón comprises parts of the municipalities Las Condes, La Reina, and Peñalolén. For hazard studies, this area is interesting as further urbanization, i.e. land-use changes and sealing of the soils in higher parts of the foothills can be expected for the future. This is a relevant process for flood origination and therefore forms an interesting research aspect. For vulnerability studies, this area is interesting as the inhabitants living in the hazard prone areas cover all existing social classes with all of them being affected by floods.

1.3 Methodology and structure of the thesis

29

Figure 4: Location of the in-depth study area San Ramón catchment with the adjacent municipalities La Reina and Peñalolén.

1.3.3 The methodologies applied in this research The methodological outline is shown in Figure 5. It is based on the before mentioned concept of risk. For the practical analysis and assessment of flood risk, a set of multi-scale (individual, household, municipal level) indicators was developed for the study area (Chapter 5). The main intention besides showing the complexity of the processes associated with flood risk is to communicate interdependencies and to show how the flood risk as a whole changes if single indicators, such as land use/land cover, number of elements at risk etc. are altered. What follows is the geodata processing and analysis. As an introduction to that, Chapter 3 shows how various methods of geoinformatics have successfully been applied in the past to analyze flood hazard, elements at risk or their vulnerability. The main advantages of the use of geodata for flood risk analysis are the large spatial coverage, the homogeneity of the information, the repeatability and transferability of the methods, and the efficiency in terms of time and costs. To account for non-tangible information, the data were complemented by census data, household surveys, and expert interviews.

30

Figure 5: Work flow.

1 Introduction

1.3 Methodology and structure of the thesis

31

Chapter 6 provides detailed information on the variety of data used for this study and the required pre-processing tasks. Within the scope of the geodata processing, high resolution multi-temporal ASTER satellite data (15 m geometric resolution) of the watershed are analyzed in combination with geologic and geomorphologic information to delineate updated information that support the analysis of runoff characteristics and their changes over time (Chapters 7 and 8). In addition to that, very high resolution remote sensing data (Quickbird satellite) and data of geographic information systems (GIS) are analyzed to identify elements at risk and their vulnerability in the study area (Chapters 7 and 9). Data from field surveys, expert interviews, and the census are analyzed to obtain in-depth information on vulnerability and elements at risk (Chapters 5 and 9). Chapter 8 relates the current and previous land-use information to the runoff generation and develops possible future land-use scenarios. It investigates the resulting runoff and finally the flood hazard using the hydrological precipitation-runoff model HEC-HMS. The resulting runoff predictions are then combined with existing very recently generated hazard maps for the study area (Perez 2009) to estimate the influence of LULC changes on the future flood hazard. Chapter 9 shows how the indicators are applied for the analysis and assessment of flood risk and how new flood risk maps are generated. Chapter 10 provides recommendations for risk prevention and mitigation that are elaborated on the basis of adequate land-use management in the interests of sustainable urban development. It is made use of scenario techniques to illustrate the possible impacts of changes of the flood-relevant variables. Chapter 11 discusses the applied methods and the concept and content of this research. Concluding remarks finalize this thesis. 1.3.4 Exchange with stakeholders A crucial prerequisite for a successful and realistic planning of the research, a good cooperation with the Chilean partners, and a fair evaluation of land-use development trends is indeed the understanding of the administrative and legal frameworks that deal with flood risk management. Several personal interviews with local decision makers and stakeholder workshops were carried out to discuss aspects of this research, to obtain more in-depth knowledge, and to draw the decision maker’s attention to the present research project in order to facilitate the knowledge transfer at the end of this project. One of the very latest developments with respect to GIS, i.e. WebGIS technologies, was applied to allow for an up-todate communication and presentation of the research results on the Internet (Ebert [Müller] & Müller 2010b). That means that the research results can be accessed via a web browser.

2 DISASTER RISK RELATED TERMS AND CONCEPT DEVELOPMENT Fundamental for the further work is the clarification of central terms related to risk analysis. It needs to be defined how the terms “hazard” and “risk” are distinguished and which role the concepts of “vulnerability” and “elements at risk” play. Section 2.1 outlines how flood risk as a concept is treated in this research. The development of an adapted conceptual flood risk framework is a central point of this study. The concepts of hazard, elements at risk, and vulnerability are described in the Sections 2.2, 2.3, and 2.4, respectively. Section 2.5 introduces disaster risk management as a tool for risk prevention, reduction, and appropriate reaction. Finally, general methods to face the flood risk are outlined in Section 2.6. 2.1 DEVELOPMENT OF A CONCEPTUAL FLOOD RISK FRAMEWORK FOR A COMPLEX URBAN SETTING 2.1.1 Identification of the components of risk Risk calculations have their origin in statistics (Luhmann 1991) and have over time been taken up by various sciences. When assessing the risk of suffering damage from natural hazards, in general at least two main factors are taken into consideration: either probability and consequence (Meyer et al. 2009, Apel et al. 2006, Brooks 2003, Luhmann 1991, Knight 1921, or hazard and people/economic values (Europäische Union 2007) or hazard and vulnerability (ISDR 2004, Wisner et al. 2004). The first two options, mostly applied in technocratic or economic studies, explain that there is a certain damage purely through the occurrence of a hazardous event and the existence of people and goods with a specified value that are located in the area affected by the hazard. They are purely technocratic approaches. But especially in large, complex, and heterogeneous urban settings, a hazard hits people and places unequally and causes different types of damage on a large scale. Thus, the conceptualization of risk needs to include a component that allows for the consideration of the reasons for why there is a certain, unequal type of damage. The above presented third option, broadly used in studies with a background in social sciences, includes the concept of vulnerability that exactly serves this purpose. This third option is followed here, as the man-made changes of hazard magnitude and the man-made system-intrinsic reasons for a disaster to happen (i.e. the vulnerability) will be investigated. If a hazard interferes with vulnerable conditions the risk of a disaster to happen emerges. That means that the preparedness to prevent a disaster is no longer given and measures for minimizing the adverse effects of the hazardous event need to be taken. Where exactly the deficits in disaster preparedness are rooted is especially important to be known in that case

34

2 Disaster risk related terms and concept development

when prevention and mitigation measures are planned, as those are always related with financial investments. That in turn means that they always need proper justification. Thus, the better the knowledge about the hazard and the amount and the capacities of people and values that are potentially affected, the more appropriate can the measures be planned. Since the present study is conducted for an urban agglomeration  a part of a megacity  an analysis at a higher spatial resolution needs to be carried out. It is not sufficient to have largely generalized vulnerability information as contained in the most conceptual risk frameworks. Rather, a strong focus of the analysis is laid on the population and constructed urban environment. That requires a preferably exact and clear representation of these “elements at risk” in the risk calculations. For that reason people, values, and urban infrastructure are considered as a separate, third contributing factor “elements at risk” in the risk concept (Kron 2005, Siebert 2004, Cardona 2003a, Alexander 2000). It was also already stated by Birkmann (2005d, p. 4), that people should be a central part of the risk analysis as they contribute to it but also have a potential to minimize it if risk management is human-oriented, i.e. starts where action can directly be taken. The risk in areas which accommodate a high number of people has a higher potential to be minimized rapidly  e.g. simply through information campaigns  than in e.g. industrially used areas (Cross 2001). By considering elements at risk as a separate component of risk, the risk assessment and potential risk management operations become more transparent: If the number of people potentially affected remains part of the vulnerability framework (e.g. under the term exposure) it will not become clear, where exactly the vulnerability values comes from. It either results from a high number of people and values potentially affected with comparatively high coping capacities or from a low number of people and values potentially affected with comparatively low coping capacities. This is visually exemplified in Figure 6. The figure depicts the concept of flood risk that was developed for this study. Four maps can be seen in the figure: A hazard map at the bottom showing different levels of flood hazard, a map containing elements at risk showing the number of buildings and people located in the study area, and a vulnerability map showing the different levels of vulnerability against floods. The uppermost map is the resulting risk map indicating the place-dependent level of risk resulting from floods. The risk is high (dark boxes in upper level of the figure) if the hazard is high or if there are many elements at risk exposed or if the vulnerability is high. Boxes in the image symbolize examples. As shown by the two examples in the middle, the level of risk differs with different levels of vulnerability even if the hazard and the amount of elements at risk is equal. In reality this could be the case if two buildings located in the same hazard prone area are constructed of different materials: the one with the lower vulnerability value is made of stone, the one with the high vulnerability value has wooden walls and therewith shows a higher susceptibility to suffer damage. The different vulnerability could also result from personal characteristics of the resident: the one with the lower vulnerability could be a midaged male; the one with the higher vulnerability could be a retired female that has less physical capacities to escape in case of a flood event and has less financial

2.1 Development of a conceptual flood risk framework for a complex urban setting

35

means to prepare for the hazardous event. As the example on the left side of the figure shows, there is no risk if there is no hazard but vulnerable elements at risk.

Figure 6: The components of risk exemplified for the case of floods: hazard, vulnerability, elements at risk.

2.1.2 Relation between the components of risk Besides the selection of components that are needed for the assessment of risk, the relation between these factors has to be clarified. In theoretical concepts, risk in form of a mathematical equation either equals the sum or the product of the two or three influencing factors (Wisner et al. 2004, UNDP 2004, Rashed & Weeks 2003). Respective risk equations can frequently be found in literature. If normalized numbers cannot be provided for the contributing factors, risk can alternatively be considered being a function of its contributing components (Birkmann & Wisner

36

2 Disaster risk related terms and concept development

2006, Briguglio 2003). The risk equation applied for this research treats risk as the product of its three components: Risk = Hazard * Elements at Risk * Vulnerability And the term risk is defined for this research as: the location-specific damage potential dependent on the flood hazard level, the amount of elements at risk, and their vulnerability. It has to be noted, that the resulting risk value is not the potential damage value as for example in Apel et al. (2006). It is rather a product that is later normalized and thus a relative value. The three factors on which risk mainly depends upon are explained in more detail in the following. 2.2 HAZARD Hazard is understood as the probability of a certain potentially damaging (and therewith hazardous) event with a certain magnitude to occur in a certain period of time (ISDR 2004, Cardona 2003a). To increase the reliability and credibility of the results from the hazard analysis, not just the expected magnitude but also the expected return period of an event has to be pointed out. That will significantly increase the probability that stakeholders consider the recommendations for risk reduction provided. A hazard is the probability of occurrence of a potentially damaging natural phenomenon, within a specific period of time in a given area (Cardona 2003b, p. 37.) Hazard maps for floods with certain magnitudes are obtained from a combined analysis of the watersheds using hydrological and hydraulic models. A certain degree of uncertainty that is contained in these maps has to be communicated to the stakeholders. Hazard maps can also be derived from GIS-based terrain analyses or the statistical analysis of historic data. 2.3 ELEMENTS AT RISK The following definition after Alexander et al. (2000) is being applied: The elements at risk comprise all people, the built environment, the natural environment, economic activities and infrastructure that are located in a hazard zone. The amount of people can be expressed in absolute numbers, the urban elements can be expressed based on their importance for the urban functioning, in financial values, if possible, or in absolute numbers as well (e.g. number of buildings, transport infrastructure, etc.).

2.4 Vulnerability

37

In economic analyses, the usage of “elements at risk” as a separate factor enables the assessment and estimation of economic losses and costs caused by a hazardous event by assigning damage functions (Meyer et al. 2009, Messner & Meyer 2006). Due to limited data availability the loss estimations are focused on the pure number of people and infrastructure affected after a flood event struck the area. Where elements at risk are not considered as a separate component in the risk equation, they are included in the concept of vulnerability under the term exposure (Fedeski & Gwilliam 2007, Schneiderbauer 2007). 2.4 VULNERABILITY Vulnerability has its origin in poverty research and  generally spoken  explains why the same hazardous event has different effects on each element at risk. A variability of types of vulnerability exist: social, physical, ecological, economic, individual, and urban vulnerability amongst others (Adger 2006, Luers et al. 2003). The high amount of definitions for vulnerability that can be found in literature is a corollary. The discussion of a selected number of concepts it thus provided some space at this stage. The common base line of all approaches is that they refer to the conditions that make an individual or a system susceptible to experience harm as a consequence of an external shock. What differs is the explanation for that aforementioned susceptibility as that depends on the type of shock, on the considered scale, the reference objects, and the location-specific conditions. The concept of vulnerability is non-tangible and it is a practical challenge to quantitatively capture it. A range of elementary concepts were generated which do all have a high explanatory value and represent interdependencies that are more or less universally valid but can be specified for specific case studies by choosing vulnerability indicators accordingly (Bogardi 2006, Birkmann 2005d, Brooks 2003, Cutter et al. 2003, Turner et al. 2003). A very prominent concept to capture the multi-dimensional character of vulnerability is the “Pressure and Release (PAR) Modell” developed by Blaikie et al. (1994) and republished by Wisner et al. (2004, Figure 7) that emphasizes the diversity of relevant scales for vulnerability research. Besides physical and social characteristics of an individual or household level, institutional, economic, and systemic conditions that influence the vulnerability are included in the proposed concept. The inclusion of institutional features on various levels  as shown in the PAR-Model  initially allows for a very comprehensive explanation of the generation of vulnerable conditions and is very suitable to explain the multiple dimensions and the complexity of the phenomenon. Encompassing two dimensions, Clark et al. (1998) define vulnerability as “people’s differential incapacity to deal with hazards, based on the position of groups and individuals within both the physical and social worlds”. During field surveys it was investigated if there is a relation between the geographic location of a household, its social position and the level of coping capacities and risk

38

2 Disaster risk related terms and concept development

knowledge (Chapter 9). Thus, the definition given by Clark et al. (1998) coincides with the research direction followed in this study.

Figure 7: The Pressure and Release Model to explain the generation of vulnerability and risk after (Wisner et al. 2004). Vulnerable conditions are generated on three dimensions reaching from political and institutional conditions (root causes) over current development processes on a regional level (dynamic pressures) to local site-specific conditions (unsafe conditions).

Pelling (2003) understands vulnerability as a concept comprising exposure (location relative to hazard, environmental surrounding), resistance (livelihood, health), and resilience (adjustments, preparation). Especially the consideration of resistance  referring also to the economic, psychological, and physical health of individuals  make the approach very realistic, but at the same time very costly and complex as in-depth household studies are needed. Wisner et al. (2004, p. 11) define vulnerability as: “the characteristics of a person or group and their situation that influence their capacity to anticipate, cope with, resist and recover from the impact of a natural hazard”. This definition is associated with the PAR-Modell (Figure 7). The explanatory power of the PARModell indeed enables the detection of other causes of vulnerability apart from the physical exposure, social conditions, and individual risk perception. Applying the PAR-Model for this study would leave the option to include institutional and economic features at the scale of a municipality, region, entire city or country. This becomes important when communal or regional prevention and mitigation measures are involved in the analysis.

2.4 Vulnerability

39

Nevertheless, the proposed model is very challenging to be applied for a practical vulnerability assessment with indicators as influencing factors might be doubled when working with the three proposed dimensions. Parameters belonging to the dimension of “unsafe conditions” are just a logical consequence of certain “root causes” and would in an approach comprising indicators referring to all three levels be statistically over-represented. Besides that, to allow for a complete quantitative vulnerability assessment, uniform measures need to be found to capture the different root causes, dynamic pressures, and unsafe local conditions for one or more locations. Thus, the model is used as a general orientation to formulate a site-specific risk model (read further Chapter 5). The basic structure of the PAR-Modell approach can also be found in the holistic approach of Cardona (2003a). %, which is for this research context considered to be the best suited definition of vulnerability. In contrast to Clark et al. (1998), Cardona (2003a) further distinguishes the social exposure into (i) the socio-economic fragility and (ii) the lack of resilience. Same as Birkmann (2005a), vulnerability is defined as the internal risk factor  in contrast to hazard which is defined as the external risk factor, a concept also carried by the PAR-Model. For Cardona vulnerability originates as a consequence of three factors: -

Physical fragility or exposure, linked to the susceptibility of human settlements to be affected by natural or social phenomena due to its location in a hazard-prone area; Socio-economic fragility, linked with the predisposition to suffer harm due to marginalization, social segregation in human settlements, and due to poverty and similar factors; and Lack of resilience, related to the limitations of access and mobilization of resources, and incapacity to respond when it comes to absorbing the impact of a disaster. It can be linked with under-development and the lack of riskmanagement strategies.

It has to be stressed out that independent from its exact definition vulnerability is a highly dynamic component. The physical fragility, the social vulnerability, and also the resilience can change rapidly  e.g. after the impact of a disastrous event  or slowly with changing personal, communal or national conditions (e.g. individual aging process, political changes, economic development, ...). Vulnerability is to a large part dependent on the hazard: in terms of construction material for example that means that a certain construction material shows a higher fragility against floods than against earthquakes. In terms of physical abilities of humans that means that people might more likely be able to wade through flooded terrain than to escape from an earthquake. The definition that is found appropriate for this research is a symbiosis of the previous theoretical findings: Vulnerability refers to the social and physical conditions that make parts of an urban system susceptible to experience damage from a flood event.

40

2 Disaster risk related terms and concept development

This vulnerability research is based on different data sources from different years (compare Chapter 6). Nevertheless, each individual type of data (e.g. census data, questionnaires, remote sensing data) was only available from one time step each. Therefore the dynamics of vulnerability can hardly be assessed quantitatively within the scope of this PhD research. Nevertheless, the transferability and reproducibility of the approach as well as the applicability for a multi-hazard study will be discussed in Chapter 11. 2.4.1 Exposure and its role in the vulnerability concept Exposure here is understood as a part of vulnerability, of physical vulnerability or physical fragility. That means that elements at risk are exposed if they are located in an area potentially affected by floods. Furthermore, they are only exposed if they do not have structural or private measures against floods (e.g. walls, backflow flaps). In other words, a building is not exposed i.e. not physically vulnerable if it is surrounded by a solid high stone wall that keeps all water out. The degree of exposure is not to be confused with the number of elements at risk. Being exposed or not is a feature of the elements at risk, expressed through their level of vulnerability. Exposure has in other studies been regarded as separate component of risk (Fedeski & Gwilliam 2007, Schneiderbauer 2007). Fedeski & Gwilliam (2007) define exposure as “the extent and value of buildings exposed to risk”. Schneiderbauer (2007) refers to the number of people located in the hazard zone. Both definitions come closest to the component “elements at risk” (Section 2.3) used in this study but do not completely match it. After Messner & Meyer (2006) exposure indicators comprise the type of exposure of each element at risk (e.g. proximity to river) as well as the flood characteristics (e.g. flood duration). While the exposure of the elements at risk is understood in a comparable manner in this research, the exposure-related indicators of Messner & Meyer (2006) are in the present research treated as hazard indicators. 2.4.2 Resilience and coping capacities and their role in the vulnerability concept Resilience is widely understood as the counterpart of vulnerability (Bohle 2007). According to Adger (2006, p. 268 f.) resilience includes the “capacity to selforganise and the capacity for adaptation to emerging circumstances”. That means that the elements at risk with a low level of vulnerability automatically show a high level of resilience (Watson & Albritton 2002). In other words, those elements at risk can overcome a disastrous event much more easily. Resilience implements a high level of coping capacities and therewith lowers the physical and/or social fragility. Coping capacities comprise measures and methods to reduce the negative impact of a hazardous event, e.g. the support of social networks or financial resources, and should according to Bohle (2007) be a strong focus of risk reduction measures (e.g. awareness and preparedness before the flood, capabilities to cope

2.5 Risk management

41

during the flood, and the potential to recover after a flood event (Messner & Meyer 2006). Resilience is here understood as a result from coping capacities that minimizes the level of vulnerability. It is expressed through low vulnerability values. Information about coping capacities have predominantly been collected during field surveys (questionnaires). Thus they can only quantitatively be incorporated in the practical assessment. That the assessment of vulnerability through its fuzzy nature remains an ill-structured problem has been stated numerous times (Taubenböck et al. 2008, Villagrán de León 2006, Rashed & Weeks 2003). 2.5 RISK MANAGEMENT After all the components contributing to the generation of risk are identified and are ideally also quantified, measures for minimizing risk can be taken. Strategies on how to prevent and mitigate risks resulting from hazardous events, how to encounter and react on them, how to return life back to normal, and to prepare for the next hazardous event are the main goals of disaster risk management. It comprises the five components prevention, mitigation, preparedness, response, and recovery which are briefly explained in Figure 8. Although floods do in Santiago not lead to disasters as defined by the Centre for Research on the Epidemiology of Disasters (CRED) (2007), the concept of disaster risk management seems useful to describe the flood risk management activities.

Figure 8: Disaster risk management cycle, modified after Felgentreff 2008.

This research is focused on the part of the risk management cycle that focuses on actions and implementations related to land-use changes previous to a disaster, i.e. on the prevention stage, and partly on the mitigation stage. Traditionally, the main

42

2 Disaster risk related terms and concept development

focus of disaster risk management was laid on the response and recovery after a disaster. Experts though were calling for a paradigm shift from pure damage assessment and hazard monitoring towards the identification and assessment of the populations’ vulnerability to natural hazards prior to a disaster, i.e. for a reinforcement of preparedness and mitigation (UNU-EHS 2005, UN-ISDR 2005). Especially because an increasing number of people is exposed to hazards and is affected by disasters, it is of urgent need to strengthen the human component in the first stages of the disaster risk management cycle and to work out prevention, mitigation, and preparedness strategies, also on a political level. The most popular example for the call of a paradigm shift is the Hyogo Framework of Action, an outcome of the World Conference on Disaster Reduction in Kobe, Japan in January 2005 (UN-ISDR 2005). It has also been stated e.g. in the Stern-Report (Stern 2007) that economic losses resulting from global climate change can be minimized if investments focus on measures for risk minimization at an early stage. This is also valid for the present study: losses through floods could be minimized with targeted investments in prevention and mitigation measures. Even though only parts of the entire risk management cycle are being touched in this research, its growing importance worldwide and the recognition of its importance by international institutions has to be pointed out. 2.6 GENERAL APPROACHES TO FACE FLOOD RISK Coming back from a more general perspective of risk analysis, the following section outlines approaches to face flood risk. Regarding the growing importance of flood risk research as a result of a changing global climate and an increase of hydro-meteorological extreme events, respective studies were carried out for numerous potentially affected countries. One example is the FLOODsite project that was carried out for seven case studies in Europe to obtain information about flood risk generation and to develop an integrated practical flood risk management methodology for European rivers, estuaries, and coasts (FLOODSite 2009). Further European examples are the FLOWS Project (Flood Plain Land Use Optimizing Workable Sustainability, finished in 2006) and the ORCHESTRA Project. The FLOWS project was carried out by researchers from Great Britain, the Netherlands, Sweden, Norway, and Germany to develop adapted flood risk management strategies with respect to a changing climate along the North Sea shores. The ORCHESTRA Project aims at harmonizing the technical underpinning of risk management. FLOOD-ERA as another European example project evaluates different flood risk mitigation measures for different risk perceptions. European flood risk-related research is funded by the European Commission amongst others. A manual for flood risk management was developed by the European Investment Bank (EIB 2007). A guideline for reducing flood losses worldwide was published by the UN-ISDR (Pilon 2004) and a guideline for settlement planning in areas prone to flood disasters was published in 1995 by the United Nations Centre for Human Settlements (UN-HABITAT 1995). An example for initiatives to reduce flood risk in Africa is the disaster preparedness manual by the Sudanese Red

2.6 General approaches to face flood risk

43

Crescent Society (Elmustafa et al. 2004). The Asian Disaster Preparedness Center (ADPC) focuses on building safer communities and sustainable development to reduce disasters in Asia. One example project is the Program for HydroMeteorological Disaster Mitigation in Secondary Cities in Asia (PROMISE) that amongst others investigates potentials to reduce flood losses through private and public mitigation (ADPC 2009). Another example of the ADPC is a very community-oriented project for flood risk reduction in Cambodia (Apikul 2002). A project to face flood risk in Argentina is currently carried out by the World Bank (World Bank 2006). A further international initiative is the Global Risk Identification Programme (GRIP). Besides specific projects, several national and international programs and institutions, such as the Helmholtz EOS Natural Disasters Networking Platform (NaDiNe), the DLR (German Aerospace Centre) disaster management centre, the RISK-EOS network (a network of European service providers, which is part of the GMES (Global Monitoring for Environment and Security) Service Element Program), the Dresden Flood Research Centre, the UK Flood Risk Management Research Consortium, several national research agencies such as the British Environment Agency, and the United States Geological Survey (USGS) that focus on flood risk management do exist and work on different spatial levels. Existing studies to prevent and mitigate measures for flood risk reduction with respect to land use/land cover (without going into detail about structural measures) were carried out by Naef et al. (2002), Rahman & Alkema (2007), and Im et al. (2009) and are discussed further in the following Chapter 3. In Europe, a legal guideline for flood management was established in 2007 that obliges the member countries of the European Union to identify areas at flood risk, to analyze the risk with maps, and to assess and reduce the risk using flood risk management plans (EU 2007). The law defines risk as a combination of the occurrence of a flood and the adverse effects of the event for the human health, the environment, the cultural heritage, and the economic activities. Thus the risk maps that need to be generated must involve different flood scenarios and have to show the potential damage of a flood. It is then recommended to set the focus of flood risk management on risk avoidance, protection, and prevention, and to protect or rebuild the natural retention areas. Finally, the member states are asked to define targets for a minimization of the negative impact of flood events. It has to be emphasized that the guideline explicitly points out that no structural flood mitigation measures are allowed if they have a significantly negative impact of the flood hazard on neighboring countries. In general, the possibilities to manage the risk are manifold and do in a first row depend on the amount of knowledge about the risk (i.e. risk analysis and assessment), on the local capacities (human, financial, temporal, ...) as well as the risk awareness and the willingness to approach that problem. Especially the last two points are very non-tangible and opened up new research directions which will not be deepened here though.

3 EXISTING FLOOD RISK-RELATED RESEARCH The following chapter gives an overview on how remote sensing, geographic information systems (GIS), and hydrological & hydraulic modeling and empirical methods were used for studies related to risk assessment with a major focus on flood risk studies. Table 1 summarizes the main fields of applications for each methodology. The structure of the chapter is accordingly: Section 3.1 lines out the current state of research in remote sensing applications with respect to flood risk, Section 3.2 shows the applications of GIS, and Section 3.3 summarizes studies based on hydrological and hydraulic modeling in the scope of flood hazard and risk research. As a focus of this research is set on investigating the influence of land-use/land-cover changes on flood risk, Section 3.4 lines out research that has previously been done in this specific field. In the last part of the chapter (Section 3.5), empirical methods for the assessment of vulnerability and elements at risk are discussed. 3.1 STATE OF RESEARCH FOR USING RS IN FLOOD RISK ANALYSIS Depending on the spectral properties, the geometrical and temporal resolution, remote sensing data can provide a range of flood risk-related information (Table 1). This information can be processed using image analysis software and GIS and can then be analyzed directly or that can be used as input data for hydrological and hydraulic models to estimate the flood hazard in a certain area. Slack & Welch (1980) used Landsat-1 data to obtain LULC information to use them for runoff estimations based on the curve number (CN) method. Van der Sande et al. (2003) have used IKONOS images to delineate information about LULC and inherent surface roughness  a key parameter in flood hazard assessment  and come to the result that the geometric resolution of 10 m is sufficient to distinguish between several urban LULC classes, even though it is not sufficient to identify buildings and roads. Canters et al. (2006) investigated the potential of optical satellite data of different geometric resolutions (IKONOS 10 m and Landsat ETM+ 30 m) for the quantification of impervious surface in an urban environment. These data are then further used as input data for a hydrological model (WetSpa). Stating that orthorectified aerial photographs would deliver the most accurate results, the authors name time constraints for their analysis as well as a small spatial coverage as main hindering factors for their usage. Instead, the semi-automated classification approaches that can be performed for multi-spectral very high resolution satellite data could yield comparable results at a better cost-benefit-ratio. The IKONOS data were used in combination with a DEM and texture measures to delineate eleven urban LULC classes.

46

3 Existing flood risk-related research

Table 1: The use of remote sensing, hydrological & hydraulic modeling, GIS, and empirical methods for flood risk-related research. Method Remote sensing

Component of risk Hazard

Extracted information Relief information Land use/Land cover

Surface roughness Soil characteristics Rainfall data Flood extent

Hydrological and hydraulic modeling

Vulnerability

Housing quality

Elements at risk

Location of elements at risk

Hazard

Runoff

Water heights Influence LULC GIS

Hazard Vulnerability

Elements at risk Risk assessment Overall Empirical methods

Vulnerability

Elements at risk

Water heights Spatial distribution of vulnerable conditions Location of elements at risk Multi-criteria anal ysis Mapping & Spatial analysis Reasons and causal relations of vulnerable conditions Location of elements at risk

References Bates & de Roo 2000 Im et al. 2009, Jacquin et al. 2008, Canters et al. 2006, Sandholt et al. 2003, Nirupama & Simonovic 2002 van der Sande et al. 2003 Corbane et al. 2008, Ray & Jacobs 2007, Thiel et al. 2001 Artan et al. 2007, Hossain et al. 2007, Hong et al. 2007 Heremans et al. 2003, Sandholt et al. 2003, Horritt & Bates 2002 Taubenböck et al. 2008, Ebert [Müller] et al. 2009 van Westen et al. 2005, Taubenböck et al. 2008, Aubrecht et al. 2009 Horritt & Bates 2002, Knebl et al. 2005, Weichel et al. 2007 Knebl et al. 2005 Im et al. 2009, Naef et al. 2002, Rahman & Alkema 2007 Lastra et al. 2008 Meyer et al. 2009

Ebert [Müller] et al. 2009, Meyer et al. 2009 Meyer et al. 2009 ESRI 2008 Kienberger & Steinbruch 2005, Cutter et al. 2003, Pelling 1997 Azar & Rain 2007, Claire 2007, Schneiderbauer 2007, Taubenböck 2007

47

3.1 State of research for using RS in flood risk analysis

Comparable results were achieved applying sub-pixel analysis for the Landsat ETM+ data. As input for the hydrological model, three scenarios were created: (i) a scenario with one averaged coefficient for impervious surface, (ii) a scenario with different coefficients for different land-use types, and (iii) a spatially fully distributed scenario in which each cell in the model was assigned a value depending on the spatial proportions of all LULC types within that cell. Modeling results using the distributive model WetSpa showed that the calculated peak discharge reacted very sensitive on the coefficients (i.e. the degree of imperviousness) and their spatial arrangement. The results proof the high value of additional detailed information about degrees of imperviousness that can be derived from the analysis of very high resolution satellite data (Canters et al. 2006). The potential of different optical and radar sensors for mapping flood extent and flood duration was tested by Sandholt et al. (2003) for a study area in Senegal. It was found that radar as well as optical (Landsat) data were useful to map flood extents and flood duration depending on the revisit time and spatial resolution. Nevertheless, it was stated that radar data in that special case are much more useful as they can penetrate the clouds that are usually associated with precipitation events. At this stage, commercially available satellite data have a geometric resolution of higher than 1 m (Quickbird, TerraSAR-X) and can therewith be used to obtain detailed ground information, i.e. it becomes possible to map elements at risk and to give information about their vulnerability. Taubenböck et al. (2008) investigated on the potential of remote sensing data (IKONOS, Landsat, SRTM) for vulnerability and risk assessment against earthquakes in Istanbul. A hierarchical holistic conceptual framework was constructed to break down the complexity of the issue and to allow for a delineation of a number of measurable indicators. Amongst those indicators, eight vulnerability and two hazard-related indicators were delineated and added up to indicate the sitespecific risk during day time and night time. For the transformation of the geospatial information into vulnerability indicators (indexing of indicators), S-curves were used to assign values between 0 (no influence on vulnerability) and 1 (high influence on vulnerability). The resulting risk index was delineated using the equation ଵ

ܴ݅‫ ݇ݏ‬ൌ ‫ כ‬ሺܸ‫ ݔ݁݀݊ܫݕݐ݈ܾ݅݅ܽݎ݈݁݊ݑ‬൅ ‫ݔ݁݀݊ܫ݀ݎܽݖܽܪ‬ ଶ

Equation 1

(Taubenböck et al. 2008), p. 41.

The study shows the potential of remote sensing data for the delineation of information regarding exposure, susceptibility, and coping that do at the same time contain spatial information. The limitation is clearly that social, ecological, and economic aspects have not been considered using this approach. The proposed methodology can most likely be transferred to other hazards. Ebert [Müller] et al. (2009) used Quickbird data for a study area in Tegucigalpa, Honduras, to map elements at risk (roads, buildings, green spaces) and their social vulnerability to landslides and floods using proxy variables. These proxies were delineated using object-oriented analysis and they were found to be valuable supporting information in social vulnerability assessment.

48

3 Existing flood risk-related research

Jacquin et al. (2008) tried to develop an operational method for mapping urban objects as they play a major role in flood risk studies. SPOT-5 images (5 m and 10 m geometric resolution) for the Touch River catchment in southwest Toulouse (France) were applied and analyzed using an object-oriented analysis (OOA) approach at different spatial scales. Even though the results showed that the quality of the delineation of urban objects have a strong negative correlation with increasing density, the derived maps are found to have a sufficient level of accuracy to be used in flood risk studies. The contextual information obtained during OOAs are highly valuable for the vulnerability and risk analysis as the position and surrounding environment of each object in the image adds to its vulnerability characteristics. Space-dependent properties are for example the position of a building in relation to the hazard zone or the distance to green spaces. Stow et al. (2007) for example showed how OOA was successfully used to classify Quickbird data and to delineate information about the socio-economic status of the residents in Accra, Ghana by using the V-IS (Vegetation-Impervious Surface-Soil) concept after Ridd (1995). 3.2 STATE OF RESEARCH FOR USING GIS IN FLOOD RISK ANALYSIS Geographic information systems can broadly be used in flood risk analysis as they offer a range of functions and opportunities. Applications start with inventorying and mapping the river network, structural flood protection measures, infrastructure, the distribution of elements at risk, and their vulnerabilities. These data can be stored, managed, analyzed spatially, updated if needed, and shared amongst institutions and decision makers. The potential negative impact of a hazard can be modeled, simulated, visualized, and communicated. During a flood event, GIS technologies offer the possibility to update hazard maps based on field work e.g. using mobile GIS devices, to derive evacuation pathways, to map damage and to visualize the impact of the flood. Finally, it offers the possibility to share and provide data worldwide via the Internet with WebGIS applications or services such as web feature services (WFS) or web processing services (WPS). That means that spatial data can be accessed without a GIS but via a web browser that amongst others allows carrying out spatial analyses on the fly using WPS (ESRI 2008). Dunn et al. (2000) used the ArcView Extension HEC-GeoHMS to prepare terrain and hydrological data for hydrologic modeling with HEC-HMS (Hydrologic Engineering Center  Hydrologic Modeling System). The remarkable capabilities of the GIS tool compensated the large amount of field measurements that was necessary for the chosen approaches for the modeling of a larger basin. Merkel et al. (2008) also point out the tremendous increase in efficiency of using GIS for preparing data for their application in a hydrological model. Krätzschmar & Böhm (2006) used a rich GIS data base in combination with different high resolution satellite data to identify elements located in hazard prone areas and to assign different levels of risk to these areas based on the respective function of the previously identified elements. Using a GIS to provide the derived information to the public, Krätzschmar & Böhm suggest to further use their re-

3.3 State of research in using hydrological modeling for flood risk analysis

49

sults for emergency planning or the design of an urban master plan for the respective study area. Kienberger et al. (2009) developed a raster cell based object-oriented approach for the assessment of vulnerability against floods. Indicators were found to determine the vulnerability towards floods for a catchment in Austria independent from administrative units. Weights for all indicators were defined based on input from experts. This method proved to be suitable for the combination of data from different sources into Vulnerability Units (VulnUs) by bringing all of them in raster format and summing them up to a vulnerability index. Meyer et al. (2009) developed a GIS-based multi-criteria flood risk assessment and mapping approach implemented in the software tool FloodCalc. Using economic, social, and environmental risk criteria, one- and multi-criteria analyses can be performed for mapping flood risk and for quantifying the flood risk reduction after the implementation or construction of mitigation measures. Raaijmakers et al. (2008) applied GIS for a spatial multi-criteria analysis (SMCA) to evaluate flood risk perception based on the components preparedness, awareness, and worries. Fernández & Lutz (2010) similarly applied GIS-based SMCA to construct a flood hazard map based on the five parameters distance to the drainage channels, topography (height and slopes), ground water table depths, and urban land use. Weights for each parameter were assigned using the analytic hierarchical process method. Tyrna & Hochschild (2010) used the GRASS GIS application r.sim.water to model urban flash floods in Jungingen, Germany. The goal is to simulate the behavior of runoff after extreme precipitation events independent from a water way. A very high resolution digital surface model (DSM) was used in combination with very high resolution remote sensing data. The curve number method after the USDA (1986) was applied for runoff estimation. Butts et al. (2006) apply a Web-GIS for real-time flood forecasting including the current meteorological conditions (FLOODRELIEF DSS). The application is based on meteorological and hydraulic models and was adapted to match the needs of the stakeholders to better serve as a decision support system. In that scope the importance of early warning and sufficient information about the threat to minimize the adverse impacts is pointed out. Furthermore, the importance of also communicating the issue of uncertainty (resulting from various sources) is highlighted. 3.3 STATE OF RESEARCH IN USING HYDROLOGICAL MODELING FOR FLOOD RISK ANALYSIS A variety of hydrological and hydraulic models are being used for flood hazard assessment. While a hydrological model is used to analyze the relation between precipitation and runoff, a hydraulic model is applied to investigate flow paths of the water in a stream and in the adjacent area in case of a flood event (Rahman & Alkema 2007).

50

3 Existing flood risk-related research

Krause (2002) applies the process-oriented distributed hydrological model J2000 to simulate land-use/land-cover changes in a large basin. He points out that it is especially important in large basins to analyze where the land-use changes take place, i.e. to apply a distributed approach (compare further Section 8.1.2). Knebl et al. (2005) received good results for flood modeling from combining the hydrological models HEC-HMS and the hydraulic model HEC-RAS in the San Antonio Basin in Texas. The RIMAX (Risk management of extreme flood events) sub-project HORIX (Development of an Operational Expert System for Flood Risk Management considering Prediction Uncertainty) is focused on the development of an efficient flood management tool for meso-scale watersheds in Germany. It consists of three components, namely the forecast of precipitation events, the application of a precipitation-runoff model (WaSiM-ETH), and the application of a hydraulic model (Sobek Rural). As uncertainties are always inherent in modeling approaches, the quantification of those is attempted by performing individual uncertainty analysis of all hydrological and hydraulic parameters (GFZ 2006). By providing dynamic hazard maps, this management tool then forms an input for risk assessment, particularly for damage estimations. Apel et al. (2006) used stochastic models and Monte Carlo analysis for the flood hazard analysis. A very high number of different flood scenarios in a part of the German Rhine catchment were analyzed to obtain probabilistic risk values. Several models were used to calculate the river runoff and to develop levee breach scenarios. The flood risk was then calculated by multiplying the flood probability with a stage-damage function. Crooks & Naden (2007) developed a semi-distributed conceptual rainfallrunoff model called CLASSIC to evaluate the impact of land-use and climate change on the flood hazard in various catchments in the UK. Weichel et al. (2007) used the 2-dimensional model TRIMR2D to estimate flood extents in the Elbe river basin. One goal in this research was to investigate the influence of the scale (i.e. the geometric resolution) of LULC information on the modeling results. First results showed that the geometric resolution of the respective input data does not have a significant influence on the model output. Lastra et al. (2008) compared a hydro-meteorological  and for a sub-area additionally a hydraulic model  with a traditional geomorphologic approach for flood extent estimations in an ungaged basin in Northern Spain. The results show that the modeling approach tends to overestimate the flood extent whereas the geomorphological approach delivers well-fitting maps. Breuer et al. (2009) present one of several initiatives that apply the method of ensemble modeling to compare different models. While many initiatives are focused on the comparison of models with respect to climate change, this project aims at evaluating the capabilities of different hydrological models to account for changes in land use and land cover under identical boundary conditions. The model performances were compared using objective functions and frequently applied coefficients such as the Nash-Sutcliffe-Efficiency. It was found that the conceptual models (see Section 8.1) generally outperformed the physically based fully distributed models. One explanation might be that the conceptual models func-

3.4 Land-use/land-cover changes for flood risk reduction

51

tion much more automatically, also in terms of calibration. It is recommended that the process of model calibration should be better standardized for further model comparisons. 3.4 LAND-USE/LAND-COVER CHANGES FOR FLOOD RISK REDUCTION Naef et al. (2002) analyzed the processes of storm runoff generation for a range of soil types and LULC types doing field measurements in the German Sulzbach catchment. While the modification of the soil type  e.g. through drainage  is costly and frequently not practicable, the surface cover could be adapted more rapidly. Higher root density, tillage and lush surface cover have a positive impact in terms of flood prevention. It was found that type, lot size as well as the spatial distribution of LULC types is relevant for the catchment response to heavy precipitation events. Nevertheless, the effectiveness of the LULC changes is limited by the soil properties of the basin. The hydrological properties of areas with soil layers with a low permeability cannot significantly be improved by LULC changes. Niehoff et al. (2002) developed a land use change modeling kit (LUCK) to simulate spatially distributed land-use changes on a grid cell basis. The land-use information (categories) is inserted in the LUCK on a raster basis and possible changes in LULC are simulated based on their topology and true position. Thus, alternative LULC scenarios were created and used as input for an adapted version of the hydrological model WaSiM-ETH. The adaptations of the physically-based model allowed for the inclusion of fast infiltration processes, rainfall intensity, vegetation coverage, and the definition of the proportion of actually sealed area per raster cell. For the study area Lein catchment in southern Germany it was found that the simulated changes in LULC were only relevant in terms of changing infiltration patterns for convective rainfalls which in reality only occur locally. They do therewith not significantly affect the water balance in the analyzed mesoscale catchment. Ashley et al. (2007) present AUDACIOUS, a decision making aid that fosters sustainable development of urban storm water drainage systems by various stakeholders in England and Wales. The systems does primarily account for expected changes in (i) the urban land use/land cover and (ii) the regional climate, i.e. the precipitation pattern. The very comprehensive system includes amongst others very detailed hydrological and hydraulic simulation models (e.g. roof drainage simulations) to evaluate how the urban system behaves under changing conditions. The goal is to adapt urban land use in that way that it supports the mitigation of adverse effects through floods. Respective measures and practical examples are currently being implemented in the scope of urban planning. Rahman & Alkema (2007) investigated the influence of larger scale urban LULC changes, such as the construction of highways, on the flood hazard as small changes in the physical setting of an urban environment can lead to changes in the flow pattern during floods. It was found that physical changes in the urban morphology, such as building new highways or other flow obstacles, are suited to

52

3 Existing flood risk-related research

control the flood hazard in one region but at the same time enlarge the hazard in adjacent regions. Im et al. (2009) used satellite data and the hydrological model MIKE-SHE to analyze the effect of land-use changes in the Gyeongancheon watershed in Korea on the hydrological cycle. Using historical data from the Landsat satellite, real land-use changes were analyzed and could be simulated through the model. The same analyses were carried out for different future land-use scenarios (involving urbanization), predicting a significant increase in direct surface runoff  a process that had also been verified for the previous land-use changes that predominantly involved an expansion of the urban areas. Karmakar et al. (2010) developed an information system for decision makers and investigated the influence of LULC changes using a GIS. The focus was hereby set on a changing risk as a consequence of more elements being exposed to a (constant) flood hazard. 3.5 EMPIRICAL METHODS Practical approaches to assess vulnerability and elements at risk Findings from the Expert Working Group ‘Measuring Vulnerability’ show that there are basically two approaches to measure vulnerability, (i) expert approaches, such as indices based on expert knowledge and (ii) participatory approaches, e.g. self-assessment that allows to raise awareness of the own situation of affected people (Birkmann 2005d). Another differentiation is the viewing perspective. While the ‘bird eye view’ on vulnerability looks at the national or international scale, the ‘frog eye view’ focuses on characteristics of individuals or a smaller community. Since there is often a lack of criteria applicable to assess vulnerability on a larger scale, most of the studies found in literature refer to a specific site with a certain threat (Haki et al. 2004, Briguglio 2003, Rashed & Weeks 2003). In contrast, the study by Cutter et al. (2003) comprises the entire area of the US, because data availability is not a constraint in that area. However, the coverage of such a large area comes at the expense of lacking detail. Cutter et al. (2003) developed a Social Vulnerability Index to evaluate vulnerability to natural disaster of the US at a county level, using a range of criteria obtained from US census data. Vulnerability can hardly be assessed directly. Villagrán de León (in: Birkmann 2005b) proposes an indirect way of measuring vulnerability through sectors applying a community based approach. The basic idea is to analyze different sectors that are present in each community (e.g. health, trade, education) and to weight and compare them. Kienberger & Steinbruch (2005) used a participatory GIS (P-GIS) for vulnerability assessment in Mozambique with incorporating a close exchange (semistructured interviews, transect walks, and community mapping) with the population living in the flood hazard-prone areas. The willingness to accept and adopt risk reduction measures and to enrich these measures using indigenous knowledge was shown.

3.6 Conclusion from the current states of research

53

With the goal to investigate vulnerability, Steinführer et al. (2009) conducted semi-structured and narrative interviews with experts and people affected by floods in Germany in the scope of the FLOODsite project. A clear relationship between social characteristics and vulnerability could not be found. The difference was rather the ownership status (renters versus owners). Besides that it could be shown that floods are very site-specific problems that vary between cities and regions, for example with respect to their history. Fekete (2009) applied census data validated with a second independent data set about damage after flood events in 2002 to analyze social vulnerability in Germany. The analysis was done on the spatial level of counties for the entire country with the type of region (population density, housing type), social conditions (living space per person, (un)employment ratio, education type) and fragility (people above 64) being used as the most relevant social vulnerability indicators. Pelling (1997) published a study on vulnerability to floods where field surveys were conducted seven days after a flood event in Georgetown, Guyana, in order to find out what types of people were affected by floods. The findings showed that those households experienced the highest damage were the ones that had a low income, poor housing quality, and little community organization (Pelling 1997). The high vulnerability of children was pointed out as they on the one hand could not attend school and on the other hand had a higher probability of contracting a disease. An interesting finding was that the housing status played a role as the ability to cope with the damage seemed to be higher in informal settlements with private ownership and an apparently stronger social cohesion. In other cases the responsibilities of taking action after flood events were not always clear as they were shared between private owners and the city, amongst others. The access to health infrastructure could not be clearly identified as a vulnerability indicator but its importance is pointed out in the scope of environmental hazard studies. Finally, a functioning institutional network is regarded to be crucial for a more effective risk management, especially with regard to the installation of prevention measures. 3.6 CONCLUSION FROM THE CURRENT STATES OF RESEARCH With respect to remote sensing, the latest data sources (very high resolution optical data) and methods such as OOA and fusion of different input data have previously successfully been applied for risk studies. The challenge for this study is thus to make use of the available data sets and methods to derive information that are useful for the analysis of flood risk in the present study area. GIS data and hydrologic and hydraulic modeling were used to obtain a broad range of flood hazard-related information. It was shown that applying a distributed approach for hydrological modeling delivers best results. However, no studies about the application of the distributed modeling approach implemented in HECHMS were encountered so far for studies outside the United States of America. It is shown in the scope of this research that the transfer of the standardized hydro-

54

3 Existing flood risk-related research

logical grid that was developed especially for the conterminous United States can also be applied in other regions of the world. What is lacking in most studies, however, is the development of a conceptual framework of risk that can (i) capture the wide and case-dependent range of causes of risk generation in a specific urban environment and that can (ii) be applied in practice. Most flood risk-related studies do so far only focus on the assessment or analysis of single components of risk. That is a valuable contribution but it is not sufficient to serve as a planning or decision making tool, especially in large and dynamic urban areas where LULC changes take place continuously and where they interrupt the urban ecosystem. Risk prevention and mitigation measures can in these cases only be delineated with a disciplinary perspective. The present research thus tries to bridge this gap, especially for the area of Santiago de Chile, and provides an interdisciplinary approach for the assessment of flood risk including hydrologic, geologic, geographic, meteorological, socio-demographic, and political data. This research combines empirical and measured data on different scales. Moreover, the demand for more transdisciplinary research is taken up through the development of a WebGIS application that serves as tool to communicate research results and to allow for the interaction and participation of stakeholders in the flood risk analysis.

4 DESCRIPTION OF THE STUDY AREA The first part of this chapter (Section 4.1) provides a human-geographic characterization of the Santiago de Chile with a focus on administrative units, population, and urban development. That provides knowledge for the assessment of vulnerability and elements at risk. Planning instruments and institutions which are relevant for flood risk prevention are described in Section 4.2 as they become important when analyzing strategies and measures for flood risk reduction and risk prevention. The third part of this chapter (Section 4.3) comprises the physiogeographic description of the area, providing information relevant for the flood hazard analysis. 4.1 HUMAN-GEOGRAPHIC DESCRIPTION No official urban boundaries are associated with the term “Santiago de Chile”. Therefore, Section 4.1.1 gives an overview about the administrative divisions of the city. Section 4.1.2 provides a brief characterization of the population living in Santiago de Chile based on census data from 1992 and 2002. The people living in the study area can be considered as one of the central components within this flood risk analysis. The expansion process of the urban areas is sketched in Section 4.1.3, where the extent of built-up areas and its alterations during the last decades is shown, and where relevant drivers of the directions of urban development are listed. Learning about former development directions is important for two reasons: first, the urban expansion is considered being one of the major causes for floods and therewith an important component in the hazard and risk analysis and second, knowledge about the past is indispensable for the construction of future land-use scenarios (Chapter 8). 4.1.1 Santiago in brief: Overview about location and administrative boundaries Santiago de Chile is located in the central part of Chile at a latitude between 32°55’ and 34°19’ S and at a longitude between 69°46’ and 71°43’ W (Figure A36). With approximately 6.7 million inhabitants (40 % of the total population) it is the largest urban agglomeration of Chile (INE 2008) and its capital city. The special about Santiago de Chile is that it is not a traditional city forming one administrative unit, but a composition of municipalities with individual mayors. Although the municipalities have their own administrative bodies, regional, and national institutions are responsible for issues exceeding the municipal limits, such as urban planning (Reyes 2003).

56

4 Description of the study area

The Metropolitan Region of Santiago de Chile (Región Metropolitana de Santiago, RM, black line in Figure A36)  covering an area of 1,550.66 km² (Reyes 2003)  is composed of six provinces (provincias) which are administratively subdivided into 52 municipalities (comunas). The municipalities can further be subdivided into districts (distritos) and finally into building blocks (manzanas, see Figure A37) (Galetovic & Poduje 2006). The building blocks form the smallest administrative level and also the smallest areal unit for all spatial analyses done with respect to mapping the three components of risk within this research. Figures A36 and A37 show the different administrative levels at the example of the comuna La Reina, a municipality within the study area, which is part of the province Santiago in the RM. For the human-geographic description in this chapter it is referred to the Area Metropolitana de Santiago (AMS) as census data were not available for the entire RM (Galetovic & Poduje 2006). The boundary of the AMS is represented by the green line in Figure A36. It consists of 34 municipalities, comprising all comunas of the province Santiago, and the adjacent municipalities San Bernardo and Puente Alto. Section 4.3 refers to the entire RM as the physio-geographic setting of the study area is introduced. For the analysis of the hydrological conditions it is important to not only focus on the urban body but also on the urban fringe and the rural surrounding. Table 2 gives a short overview with only brief explanations for terms that are relevant within this context. Table 2: Overview about selected terms that are used in the context of urban planning in Santiago. Term

Meaning

Región Metropolitana de Santiago (RM)

52 comunas (see Figure A36)

Area Metropolitana de Santiago (AMS)

34 comunas (see Figure A36)

Province (provincia) Santiago

Comprises 32 of 34 comunas of the AMS

Municipality (comuna) Santiago

Historic city centre and one comuna within the province Santiago Regulatory plan for the RM

Plan Regulador Metropolitano de Santiago (PRMS) Plan Regulador Comunal (PRC)

Regulatory plans for the municipalities

Mancha urbana

Actual physical boundary

Limite urbano

Normative urban limit, areas open for construction after PRMS (Figure A36)

57

4.1 Human-geographic description

4.1.2 Social structure The current population of the RM was estimated to be 6.8 million in June 2009 (INE 2009). According to the census data from 2002, the AMS had a population of 5,456,326 (compared to 4,791,520 in 1992). Together with the rest of the population of the RM a total amount of 6.06 million inhabitants resulted for 2002 (INE 2008). Between 1992 and 2002, the province Chacabuco in the northern part of the city (Figure A36) has experienced the highest annual growth with 3.89 %. Second highest was the annual growth in the rest of the RM with 2.31 %, and lowest was the annual growth between 1992 and 2002 with 1.31 % in the AMS, i.e. in the central municipalities (Galetovic & Poduje 2006). These numbers are pointing at the spatial development directions that form part of the problem analysis in this research. Table 3 gives a brief overview about relevant population characteristics of the AMS and the municipalities La Reina and Peñalolén. Table 3: Population characteristics in the Metropolitan Area of Santiago de Chile, the municipality La Reina and the municipality Peñalolén. Variable

AMS

La Reina

Peñalolén

Size in km² Amount of areas facing high flood hazard level after PRMS [km²] Total population 2002 Total population 1992 Pop. growth 9202 (%) Population density per km² in 2002 Number of male population in 2002 Number of female population in 2002 People above 65 years (%)

2,271.29 160.24

23.4 3.87

54.2 4.68

5,408,150 4,758,655 13.65

96,762 92,410 4.71

216,060 179,781 20.18

2,379 2,606,894 2,801,256 7.33

4,135 44,293 52,469 10.39

3,986 105,528 110,532 5.83

People under 5 years (%) Incomplete basic or no education (%) University education (%) Population living in flood prone areas (%) New buildings constructed in flood prone areas between 1993 and 2002(%)

n.a. n.a. n.a. 10.55 11.63

5.95 20.71 32.72 27.80 11.46

9.25 32.53 12.37 17.99 8.12

4.1.3 Urban development 4.1.3.1 A brief history of urban development in Santiago de Chile Since its foundation in 1541, Santiago de Chile has been growing significantly. The highest growth took place between 1960 and 1970 and between 1990 and 2000 (Reyes 2003). From 1940 to 2002, the urban built-up body mancha urbana grew from 11,017 ha to 64,140 ha (Galetovic & Jordán 2006).

58

4 Description of the study area

In the particular study area around the Quebrada San Ramón and the two adjacent municipalities La Reina and Peñalolén, the analysis of LULC changes shows a growth of the amount of built-up area by 7.47 % in La Reina and by 13.46 % in Peñalolén between 1993 and 2009. Urban expansion took place towards the natural areas of the Andean foothills but also in areas that had previously been used agriculturally, e.g. for the cultivation of vine and crops. Figure A38 shows that a large part of the new built-up areas was constructed on formerly agricultural land causing a significant change of the infiltration capacities of the soil. While agricultural land, especially vineyards in Peñalolén had mostly been covered by vegetation and therewith had high infiltration capacities and positive effects of interception, the newly evolving intermediate density housing areas have significantly lower vegetation coverage and higher degrees of sealed surface. Further significant changes occurred in the western part of Peñalolén where commercial sites were constructed and resulted in an almost complete sealing of the surface. In La Reina, the urban expansion was less strong and mainly resulted in a loss of urban green spaces (densification of the existing built-up structures). Figure A38 furthermore shows the location of flood hazard zones from 1987 that are used in the PRMS (blue hatched polygons) and the hazard-prone areas that were derived from a flood hazard study of the water ways coming from the Andean mountains carried out in 2008 (AC Ingenieros 2008). Both data sets have different formats with the older study showing the entire outline of the hazard zone and the more recent study only showing the points of overflow. However, it can be seen that the flood hazard changed during the last two decades (more areas potentially affected). Especially in La Reina, where construction activities occurred on former green spaces, almost all creeks coming from the Andean mountains (blue lines in Figure A38) imply a flood hazard. The analysis of the LULC changes in Peñalolén shows that construction occurred in areas declared as hazardous according to the PRMS, which signifies a non-sustainable development in the municipality (compare further Section 9.6). Important for this research is the fact that the progress of the urban expansion in Santiago de Chile does not take the natural courses of creeks and rivers into serious consideration. In former times, pathways from the old city center towards the agriculturally used areas on the Andean foothills emerged parallel to the waterways. With further expansion of the city, these pathways were consolidated into roads (Romero et al. 2007). It is explained in Section 4.3.2 that the dry climate in the RM can lead to little or no water in the creeks at most days of the year. For that reason, roads were constructed in the river beds and some of them are today still named after the creeks (e.g. Avenida Antupiren). Besides the low land prices and the non-presence of water courses in large parts of the year, another reason for construction in river beds and flood plains might be the lack of knowledge about the negative consequences resulting from a disruption of that natural hydrological network in the catchment (Reyes 2003) and the economic benefit that can be yield (Carvacho 2010). Following the laws of physics though, storm water finds its way no matter if the surface is covered by asphalt or gravel, and leads to urban floods on roads during winter precipitation and to river floods in numerous parts of the city (Section 1.1.3).

4.1 Human-geographic description

59

These “acts of man” do indeed lead to an acceleration of urban floods in the study area. While some of the vineyards that still exist in the western part of the municipality will be maintained, it can already be observed today that remaining open spaces are being built upon: For example, the new residential sites evolving at the westernmost point of overflow of the Lo Hermida creek (Figure A38) where the creek is first channeled and then continues below surface. These newly evolving residential sites are located directly in a flood prone area. Even though a smaller channel that is capable of directing a certain amount of storm water around the new residential sites was constructed above that area, the loss of retention areas resulting from the ongoing construction work will amplify the amount of surface runoff reaching the lower western parts of the municipality. An increasing exposure of the population is the result. Between 1991 and 2000, the entire mancha urbana has grown by 12,049.6 ha, which equals an annual growth of 1,339 ha (13.39 km² Ducci & González 2006). Caracas grew 462.9 ha per year between 1991 and 2001 and São Paulo grew 2,423 ha per year between 1988 and 2000 (Angel et al. 2005). The more recent expansion in Santiago de Chile took place towards all directions, whereby the eastern parts of the city were transformed from agricultural or natural into residential areas and the western and north-western parts are now covered by industrial areas or areas with mixed use (residential and industrial). Especially in those areas in the south and southwest where the low class population is settled, a decent amount of agricultural area can be found and will probably be converted into further low class residential areas (Ducci & González 2006). That the rapid urban expansion and partly also densification of the city in this case has more drawbacks than benefits for the city itself as well as for the rest of the country was stated numerous times (Romero et al. 2008, Hansjürgens et al. 2007, Ducci & González 2006, Galetovic & Jordán 2006, Walter 2005, Reyes 2003). 4.1.3.2 Normative urban boundary vs.mancha urbana It is outlined in Section 4.2.1 that areas open for construction are defined by the PRMS (límite urbano, see also Figure A36). Thereby, it has to be distinguished between the normative limit set by the PRMS of which 69 % is currently built-up area and the total physical límite urbano (mancha urbana) (Galetovic & Poduje 2006). Construction outside the normative limit can be permitted for housing and commercial projects with a maximum density of 150 inhabitants/ha in the AMS and 85 inhabitants/ha in Chacabuco. Industrial or hazardous developments as well as construction in green spaces, parks, river beds, and hills cannot be authorized without special permission and by imposing certain conditions. For that reason, the physical boundary, e.g. the area already covered with buildings and urban infrastructure has a larger extent than the official limits as defined by the PRMS (Figure A36). This development can especially be noted in the southern region of the RM and in the comuna Colina.

60

4 Description of the study area

The main reasons for the difference between the official and the actual built-up area are: (i) Parcelas de agrado: areas smaller than 5,000 m² where private construction is allowed outside of the normative urban limit as defined by PRMS without the need of constructing infrastructure and services (summer houses) (ii) Viviendas sociales: social housing for families in need, which is mainly financed by subsidies of the state and can due to high land prices not be financed in those areas officially open for construction (Artículo 50 de la Ley General de Urbanismo y Construcciones (Galetovic & Jordán 2006, Ducci & González 2006). 4.1.3.3 Drivers for urban growth The general driver for urban growth is most likely the socio-demographic development in a city as a result of the economic development. The rising number of urban dwellers worldwide but also in Santiago de Chile was lined out above. The expectations of a higher quality of life and better opportunities to make a living increase the attractiveness of cities. The economic development can be regarded as another relevant factor determining the amount of urban growth as it defines the need for production, service or retail sites, infrastructure, and larger properties for those dwellers benefitting from the economic prosperity amongst others. With the decline in urban density and the strong role of the private sector today, the economic situation and again the socio-economic development are the main engines for urban sprawl. These tendencies are further strengthened through the neoliberal economy in Chile. Realizing the increase in income and car ownership, the demand for more individual space to live on the one hand and the means to finance that on the other hand confirm this argument (Petermann 2006). Moreover the technical development regarding transportation, communication, and service infrastructure reduce the need for a compact urban setting and favor space and flexibility when buying real estate. This can consequently also be named as a driving factor for recent urban development. Poduje (2006) even states that under Jaime Ravinet (at that time minister for housing) and Luis Eduardo Bresciani (Head of the Department of Urban Development, MINVU) the urban boundary was extended in 1997 in order to demonstrate the increasing income of at least some parts of the population and to allow those people to make use of their opportunities. Thus, today, the land prices are seen as another main driver for urban growth as they show the value of the land, the potential social classes that will get settled and the demand for commercial and service infrastructure. Last, the influence of private and public institutions can be regarded as a driving and directing factor for urban growth according to Augustin Perez, head of the construction department in the municipality of La Florida (Reyes & Ebert [Müller] 2009). During a stakeholder workshop carried out in November 2009 he brought the example of the construction of one of the major highways in Santiago de Chile (Americo Vespucio) which was superimposed to the initial PRC the respective municipality.

4.1 Human-geographic description

61

4.1.3.4 Possible future urban development There was no urban expansion in 18 out of 40 municipalities between 1991 and 2000 as those 18 have already been 100 % urbanized according to the urban limits set by the PRMS (Ducci & González 2006). Nevertheless, the option of constructing social housing or consolidating peri-urban cabins Parcelas de agrado still remains where space allows, e.g. in La Reina, which has according to the PRMS technically no growth capacities (Quezada 2009). It is under debate whether or not the currently valid maximum elevation for construction should be set higher than 1,000 m. That would result in another spatial extension and in a loss of biodiversity and retention areas amongst others. As expected, the main driver for that discourse is the pressure on the land market and profitable land prices (Reyes 2003). Construction is already possible above the 1,000 m limit (cota mil) if the usage is neither residential nor industrial (see Figure 9). The protection of ecologically valuable areas is legally being ignored in such cases.

Figure 9: Construction site above 1,000 m in the municipality La Reina. The building is used as a recreation and service area and is therewith not subject to the currently valid construction limit after the Metropolitan Regulatory Plan.

According to the MINVU-SEREMI, the regional ministerial secretaries (Secretaría Regional Ministerial) of the Ministry of Housing and Urbanism (Carvacho & Paredes 2009), the most significant urban sprawl will take place in the municipality of Colina in the northern part of the city, in San Bernardo towards the south, in Maipú and Cerrillo towards the west. In the eastern municipalities the degree of

62

4 Description of the study area

expansion towards the mountains depends on whether or not the construction limit of 1,000 m will be maintained or not. Densification of the existing settlement can already be observed in these municipalities, as exemplified by Figure A-39. 4.2 PLANNING INSTITUTIONS, INSTRUMENTS, AND PROCESSES RELEVANT FOR FLOOD RISK In order to do justice to the complete flood risk management cycle, the five stages prevention, mitigation, preparedness, response, and recovery (Section 2.5) would have to be considered in the following analysis. As the focus of this work lays on the analysis of land-use/land-cover patterns and their impacts on flood risk the main focus will in this research be laid on the prevention stage comprising longterm measures (i.e. urban planning) with some analyses being carried out for the mitigation stage (short-term measures). It lies beyond the scope of this research to analyze all actions taken shortly before or after the impact of a flood event and to work on all responsible institutions and instruments. 4.2.1 Relevant institutions, laws, and instruments in the prevention stage On the national level, the ministries for housing (MINVU), for public works (MOP), for agriculture (MINAGRI), for transport and telecommunication (MTT), for geology and mining (SERNAGEOMIN), and for the environment (CONAMA) are involved in LULC decision making. The urban planning on the regional level is regulated by the General Law of Urbanization and Construction (Ley General de Urbanismo y Construcciones, LGUC). It contains directions on how the competencies, authorities, and functionalities are distributed and through which normatives these organs that deal with urban planning and construction are directed and controlled (Article 1 & 2, LGUC). The associated regulation (ordenanza) describes the processes of administration, planning, and urbanization, including construction and technical standards (Article 2, LGUC). The Metropolitan Regulatory Plan of Santiago, the PRMS, is a legally binding document and forms the framework for urban development in Santiago de Chile on the regional level. The responsibility for the PRMS has the MINVU. The PRMS was initially developed for the AMS (1994) and in 1997 it was extended to also include the province Chacabuco (see Figure A36). Since 2007, the PRMS applies for the entire RM (i.e. all 52 municipalities) and mainly defines whether an area is open for construction or not (Poduje 2006). The PRMS consists of maps at the scale of 1:50,000, a compilation of regulations (ordenanza) and their interpretation (memoria). It therewith builds a framework on a smaller spatial scale that has to be refined by the municipalities at a larger scale (1:2,000 or 1:5,000) through the obligatory and binding communal regulatory plans (PRC). Changes in the PRMS are time consuming. They can be requested by the municipalities and have to be permitted by the SEREMI of the MINVU (Galetovic & Poduje 2006).

4.2 Planning institutions, instruments, and processes relevant for flood risk

63

Modifications were done in 1997, 2003, and 2007 (Petermann 2006, Reyes 2003). The current limit of the area open for construction according the PRMS (2007) is shown by the red outline in Figure A36. Those areas within the RM that are not open for construction officially comprise (i) agriculturally used areas, (ii) parks, and (iii) ecologically protected areas (Galetovic & Poduje 2006). Most important for this study is Título 8 (Article 8) of the PRMS, the section where areas with construction restrictions resulting from risk are defined. Among the flood prone areas after Article 8.2.1.1. it is distinguished between floods coming from natural water ways and floods from water emerging from phreatic water tables. With respect to the first type the width of the areas along both sides of the water way on which construction restrictions are imposed are defined for each water course (e.g. 100 m for the Quebrada San Ramón, 40 m for the Quebrada Lo Hermida). Even though the values are defined individually for each stream or channel, local hot spots (points that have a disproportionate high flood probability) are not considered. The zoning is in the case of the municipalities La Reina and Peñalolén directly transferred into the PRC. The only types of construction allowed in those areas are green spaces (sportive, recreational or touristy usage) with a minimum proportion of built-up areas that only imply shortterm stays of people. Upon special request, regular construction activities can be admitted if an official survey proves that no negative impacts on the water ways and riparian zones result. Respective requests do frequently come from real estate traders or construction companies (compare listing in AC Ingenieros 2008). These exceptions get permitted in a number of cases. Regrettably, the plan does not include updated or verified information on flood hazards (Carvacho & Paredes 2009). The first reason for that is that the most comprehensive study on floods in Santiago de Chile has only been published in 1996, after the PRMS was designed. Reasons for not updating neither the regional nor the communal plans with information coming from that study are most likely decreasing land prices in the respective areas as well as a bad reputation of a municipality if it becomes public that the equipment with storm water infrastructure and flood mitigation measures is insufficient (Carvacho & Paredes 2009). Those areas lined out as flood hazard zones are derived from studies carried out in 1986 and 1987 (Ayala et al. 1987). This research refers to those areas as areas facing a certain hazard level after the PRMS. The link to the CONAMA (since October 2010 transformed into an environmental ministry) with respect to urban planning is established over the green spaces. The task of this institution is to prevent a decrease of urban green spaces. The regional representatives of the National Environmental Commission (CONAMA), the COREMA, are responsible for the environmental impact studies (SEIA) of planned urban expansion, a point being criticized by the MINVUSEREMI in Santiago as the liberalism (i.e. lack of normative regulations) and pluralism (i.e. too many parties involved) of that system complicates decision making in the planning unit (Carvacho & Paredes 2009). On the municipal level, the urban planning and development is steered by the PRCs. The PRC have to be in line with the zoning of the PRMS and are designed by the communal planning secretariats (SECPLAN). They comprise those areas

64

4 Description of the study area

officially open for construction but not those areas above 1,000 m. That means that these areas can be changed on regional level only. After Article 142 of the LGUC, the Board for Municipal Construction (DOM, Dirección de Obras Municipales) is responsible for the implementation and compliance of these regulations on the municipal level. Changes in the zoning of the PRC have to be requested at the MINVU-SEREMIs and require sufficient financial resources as they can only be done after environmental impact assessments of the envisaged changes are delivered (Quezada 2009). 4.2.2 Relevant institutions, laws, and instruments in the mitigation stage To get a first overview about the actors relevant prior to flood events, Table 4 provides an overview about the institutions, laws, and instruments of importance for the prevention and mitigation stages. Table 4: Overview about institutions in Santiago de Chile that are relevant for the stages of flood risk prevention and mitigation (Carvacho & Paredes 2009, Metz & Weiland 2009, Retamal & Estellé 2009). Stage

Level

Relevant institutions

Instruments

Prevention long-term measures

National

MINVU, MOP, MINAGRI, MTT, SERNAGEOMIN, CONAMA GORE

Law 19,3001

Regional

Regional

COREMA MINVU-SEREMI SECPLAN DOM MOP-DOH

Intercommunal

MINVU-SEREMI

Intercommunal Communal Mitigation medium-term measures

Law 19,1752 PROT not yet published SEIA PRMS PRC (Design) PRC (Application) Law 19,000 Law 19,5253 Plan Maestro de Evacuación y Drenaje Aguas Lluvias para el Gran Santiago

The Dirección de Obras Hidráulicas (DOH) started working in the field of storm water management fairly recently in the year 1997 following the Chilean storm water law 19,500. This law defines that the DOH of the MOP carries the responsibility for the primary storm water evacuation network. In addition to the primary 1 Ley de Bases del Medioambiente 2 Ley Organica Constitucional sobre Administración y Gobiernos Regionales, Article No. 16 3 Regula Drenaje y Evacuación de Aguas Lluvias

4.3 Physio-geographic description

65

net there is a secondary net for which the MINVU SEREMIs carry the responsibility. To define the two networks that same law obligates the two institutions to define a master plan for the drainage and evacuation of storm water in the urban zone (PMALS). This plan had been worked out between the years 2000 and 2002. All natural water ways, all canals, and all tubes with a diameter of 800 mm or more were defined as the primary network. The secondary network hence comprises all pipes with a diameter of less than 800 mm. The design of the secondary network is not part of the master plan, it is a separate responsibility of the MINVU SEREMIs. The plan also includes basic proposals to adapt the canals and the evacuation network to avoid the accumulation of storm water in undesirable locations. More detailed studies are carried out through consultants who also propose more specific solutions. These studies are financed by the MOP-DOH. Luis Estellé, head of the department for storm water projects stated in an interview in April 2009 (Retamal & Estellé 2009) that a new policy is currently being brought on the way that aims at reducing the amount of sealed surface in the newly built-up parts of the city to avoid negative consequences in the lower parts. Unfortunately, this new policy is not yet incorporated in the master plan, hence is not being put into practice, but will be considered in the ongoing modification process. The secondary net is being established patchy where new construction (mainly roads and new residential sites) occurs as it is now defined to be a mandatory part of all construction activities to offer solutions for the storm water evacuation in the law of construction (Ley de Construcciones). As neither the primary nor the secondary network is complete yet, intermediate solutions have to be found where the newly constructed part of the drainage network cannot be connected to existing parts (Retamal & Estellé 2009). In addition, the MOP-DOH has a consulting function with respect to urban planning. Whenever regional or municipal land-use plans are modified, the MOPDOH has to evaluate and agree with the changes from the point of view of storm water management. According to Estellé (Retamal & Estellé 2009), the MOPDOH does primarily point out deficits by comparing the proposed modifications with the master plan for the drainage and evacuation of storm water in the urban zone and revises the necessary environmental impact studies carried out by external consultants. It is noteworthy that no responsibilities in the mitigation stage are anchored on the communal level. This lack might partly explain why floods occur. Leonardo Céspedes stated that the government provides sufficient storm water infrastructure for the average rainfall events in 90 to 95 % of the country (Céspedes 2010). It is then the responsibility of the municipalities to ensure a proper functioning of the infrastructure during a storm event, i.e. that the gullies are not clogged and that there is no garbage left in the water courses. 4.3 PHYSIO-GEOGRAPHIC DESCRIPTION The following section contains a physio-geographic description of the floodrelevant aspects of the study area. The relief, geology, and soil conditions within a

66

4 Description of the study area

catchment (Section 4.3.1) determine the flow characteristics of rainfall and melting water. Climate (Section 4.3.2) plays an important role in hydrological studies as it  amongst others  describes temperature and precipitation patterns as well as characteristic circulation patterns. Section 4.3.3 gives an overview about the hydrological network, about natural water ways and canals in the city. Natural as well as cultivated vegetation needs to be considered as it is relevant for the processes of surface runoff, infiltration, transpiration, and interception, which all play a major role in flood hazard studies (Section 4.3.4). Finally, Section 4.3.5 sets the focus on the in-depth study area for the hydrologic analysis: The catchment of the San Ramón River. 4.3.1 Geology, Geomorphology, and Soil Santiago de Chile is located in a bowl-shaped valley between the Andean mountains on the eastern side, the Coastal Cordillera to the west, the mountain range Cordón de Chacabuco in the northern part and the Angostura de Paine, which forms an elongation of the Andean mountains towards the Coastal Cordillera in the south. The three main landscape units in the area  the Andean mountains, the Central Valley, and the Coastal Cordillera  are products of tectonic activity, volcanism, and processes of transformation and reshaping during past glacials and interglacials, and explain the climatic and hydrological conditions of today in the study area. Figure 10 shows the main profile at the latitude of Santiago de Chile with the three major relief components.

Figure 10: Cross-sectional profile of the main geological units in the Santiago region: CC = Coastal Cordillera, CD = Central Depression (Central Valley), PC = Principal Cordillera (Andean mountains). Adapted from Charrier et al. 2007, p. 22.

4.3.1.1 Geology Out of the 15,506.58 km² of the RM, 56 % are mountainous surface, to a large part granitic and to small parts limnic rock (Reyes 2003). The major part of the study area  the San Ramón catchment  is covered by rigid bedrock, slope sediments or blockschutt of the Abanico Formation (Formación Abanico). This is a

4.3 Physio-geographic description

67

strongly folded and thick (up to 2,000 m) succession of tuff, volcanic breccia, andesitic lavas, and rhyolites with numerous subvolcanic intrusions (AC Ingenieros 2008, Charrier et al. 2007). The Abanico Formation is dated to have emerged during the late Eocene and early Miocene. As a result of tectonic activity and triggered through heavy rainfalls parts of the basin are covered by pleistocene and younger landslide and also alluvial deposits with higher permeabilities than the bedrock formations. The geological structure of the basin is furthermore cut and influenced by the existence of the active San Ramón fault line that is located along the boundary between the Central Valley and the Andean mountains (Charrier et al. 2007). 4.3.1.2 Geomorphology The Andean mountains  reaching elevations higher than 6,000 m  started forming during the Tertiary and today function as headwater areas for the hydrological network. Their steep relief is very characteristic for areas of solid  and to a large extent impermeable  volcanic rock. The slopes become significantly smoother after reaching the Andean piedmont (AC Ingenieros 2008). The sediments eroded in the higher areas (alluvial sediments) are consecutively deposited in the the Central Valley (Valle Central), where the city of Santiago is located. Several large alluvial fans give evidence for the relocation of sediments through the hydrological network. The third landscape unit, which is the Coastal Cordillera, is a younger and lower (around 2,000 m) mountain range extending parallel to the Andean mountains from north to south. It is cragged by a number of valleys resulting from erosional forces of rivers coming from the Andean mountain range. Characteristic are the associated single scattered mountains which can be found in the westerly Central Valley (AC Ingenieros 2008). 4.3.1.3 Soil Most soil-related research in Chile is carried out for those areas that can be used for agricultural purposes. These areas are generally plain and do neither include urban built-up areas nor the steep or dynamic regions of the Andean mountains. For that reason, only little information and no maps about the soils in the San Ramón catchment are available (Stumpf 2009). Field studies showed that the soils in the catchment are to a large part just developing and do not yet show distinct horizons. According to the USDA soil taxonomy that uses the group “order” to distinguish different soils according to their genesis, Entisols and Andisols can be found (USDA 1999). Entisols predominantly consist of mineral soil materials and do not show distinct pedogenic horizons. The lack of distinct soil horizons can either be caused by the underlying material that does not support the quick formation of soil because it is inert or very solid (e.g. quartz sand, limestone) or be-

68

4 Description of the study area

cause of the environmental circumstances: Erosion, high deposition rates, plowing or animals can prevent or significantly slow down soil development. Andisols do often form on volcanic ashes or volcanoclastic materials and are characterized by high phosphorus retention, available water capacity, and cationexchange capacity (USDA 1999, p. 120). This soil order is furthermore described as having a „a dominance of shortrange-order minerals or Al-humus complexes that result from weathering and mineral transformation with a minimum of translocation” (USDA 1999, p. 120). That means that in contrast to Entisols these soils develop in more stable environments and do already show organic layers where water can be stored for vegetation or after rainfall events. In regions above 2,100 m where the slopes are steeper, the organic matter content is less than 3 % and further decreases with elevation (Cavieres et al. 2000). The relation between soil, geology, and the infiltration behavior in the watershed is further established in Section 8.2.3. 4.3.2 Climate The Metropolitan Region of Santiago de Chile has a subtropical climate with hot and dry summers (November to March) and heavy periodic rainfalls during the colder winter months (May to August) (Weischet 1970). The climate above 3,200 m can be characterized as tundra climate (Cavieres & Arroyo 1999). Arctic climate is present on the high glaciated peaks with an elevation of more than 5,000 m. 4.3.2.1 Precipitation Its geographical position in the Central Valley between the Coastal Cordillera and the major Andean mountain range influences the citys’ precipitation pattern. The Coastal Cordillera impedes a strong influence of the maritime climate conditions associated with the spatial proximity to the Pacific Ocean on the west unless the thermal inversion layer rises above 1,000 m and west winds occur. That would lead to an influx of clouds via the valleys of the rivers and therewith to a higher amount of precipitation reaching Santiago de Chile. If west winds carry clouds above the Coastal Cordillera rain starts falling in the central part of the city. The precipitation intensity then increases with elevation and proximity to the Andean piedmont through uprising of the clouds. Annual average precipitation values from the time span between 1979 and 2007 proof this pattern: the average rainfall at the station “Terraza Oficinas Centrales DGA” located at 560 m was 332.3 mm compared to 442.9 mm at the station “Antupiren” located in the eastern part of the city at 920 m asl (own calculations of rainfall statistics). In the Andean mountains the amount of precipitation generally rises with elevation (DGAC 2009). Almost every rainfall event leads to floods in at least some parts of the city. In the lower region, 80 % of the annual average precipitation is recorded during winter time (Figure 11), whereby the precipitation from above 1,500 m is like-

4.3 Physio-geographic description

69

ly to fall as snow (DGAC 2009). Precipitation below the 0°C-isotherm frequently falls as torrential rainfall (AC Ingenieros 2008). Snow melt is not a significant cause of floods. The amount of precipitation is mostly influenced by the El Niño/Southern Oscillation (ENSO) phenomenon that leads to an increase of rainfall. In contrast, dry winters can generally be recognized during La Niña years. The amount of snow melting during the spring and summer months varies accordingly (Weischet 1996, Weischet 1970). 4.3.2.2 Temperature With an average maximum temperature of 29°C and an average minimum temperature of 12°C, the month of January is the warmest of the year. With average high and low temperatures of 14°C and 3°C, the month of June is the coldest month in average years (BBC 2009), see Figure 11. The average annual humidity is 70 % and the average wind direction is south-west.

Figure 11: Average temperature and precipitation in Santiago de Chile (BBC 2009).

Important with respect to the topic is the location of the 0°C-isotherm, which defines whether the precipitation falls as snow or as rain. The elevation of the permanent snow line (equilibrium line altitude, ELA) is calculated via radiosonde measurements to be located at around 3,954 m +/- 236 m and expected to increase in the future (Carrasco et al. 2008). In the case that warm sub-tropical air is coming from the north, the snow line can rise at a scale of a several hundred meters. If associated precipitation falls as rain instead of snow the amount of direct surface runoff is higher than the retarded runoff resulting from snowfall. The relevance of these climatic factors for flood events was shown in the past numerous times. Best known in the case of Santiago is the example of the floods at Quebrada de Macul

70

4 Description of the study area

in May 1993 (Vargas 1999, Naranjo & Varela 1996). The creek is comparable to Quebrada San Ramón, which is located around 6 km north of Quebrada de Macul. During the events in May 1993, the snow line rose due to high temperatures above 16°C to about 3,240 m. Thus precipitation with a ten-year return period caused discharge with a 100 year return period (Vargas 1999). 4.3.2.3 The role of climate change in the study area Climate change studies based on the scenarios A2 and B2 developed by the Intergovernmental Panel on Climate Change (IPCC 2007) predict a general reduction of precipitation in the central region of Chile. A study for the entire country which is based on the IPCC scenario A2 and projections from the English global climate change model HadCM3 predicts a temperature increase of 1°C for the period of 20102039, of 2 to 2.5°C until 2069 and of 2.5 up to 3°C for the period 20702099 (Bárcena et al. 2009). Temperature increase projections for the alternative scenario B2 are slightly lower. The IPCC scenario A2 assumes strong population growth and the highest greenhouse gas emissions while the scenario B2 assumes sustainable development and a moderate increase of greenhouse gas emissions. The total amount of precipitation is predicted to decrease in scenario A2 by 10 to 20 % until 2039, by 20 % until 2069 and by up to 30 % until 2099. The predictions for scenario B2 are lower and to only reach 20 % until 2099 (Bárcena et al. 2009). At the same time, the probability of hydro-meteorological extreme events  droughts as well as floods  increases clearly with scenario B2. Even though the total number of extreme precipitation events will decrease, respective events will become more threatening as the amount of rainfall will increase with an increasing height of the 0°C-isotherm (Bárcena et al. 2009, Perez 2009). Perez (2009) investigated the influence of climate change on flood risk in the San Ramón basin. Projections of the IPCC scenarios A2 and B2 that were regionalized by the Geophysics Department (DGF) of the Universidad de Chile using the PRECIS (Providing Regional Climates for Impact Studies) model as a starting point. The regionalization is conducted for a 25 by 25 km² grid. In order to obtain values for the San Ramón catchment the values derived for a grid point close to a central meteorological station (Quinta Normal) were used and modified with respect to the changes in elevation between the grid point, and the gravity center of the San Ramón catchment. In practice that means that the precipitation is 20 mm higher in the area of interest in the Andean piedmont. Table 5 shows the maximum precipitation values that were delineated for the catchment based on the adapted values (Perez 2009).

71

4.3 Physio-geographic description Table 5: Probabilities for the daily maximum precipitation (mm) at San Ramón station (800 m altitude) for different return periods under climate change conditions (Perez 2009). Return period [years] 2 5 10 20 50 100

Base line [mm]

A2 [mm]

B2 [mm]

63.8 82.5 94.9 106.8 122.2 133.8

61.6 79.9 91.9 103.5 118.5 129.8

64.8 94.5 114.2 133.1 157.5 175.9

The rainfall probabilities for the current daily maximum in the San Ramón catchment as shown in Table 6 are calculated based on statistical tests (Chi Square, Filiben, Kolmogorov Smirnov) by AC Ingenieros 2008. Comparing these numbers with the climate change projections as shown in Table 5 indicates the strong negative influence of climate change on the intensity of single precipitation events in the study area. Table 6: Current probabilities for the daily maximum precipitation (mm) at San Ramón station (800 m altitude) for different return periods (AC Ingenieros 2008). Return period [years] 2 5 10 25 50 100

Precipitation [mm] 45.8 62.8 74.1 88.3 98.9 109.5

4.3.3 Hydrology Santiago de Chile is characterized by a broad network of rivers, creeks, and irrigation canals (Reyes 2003). The main rivers are the Río Maipo in the southern part of the city and the Río Mapocho which is crossing the city center (see Figure 12). Several smaller creeks which are not all shown in the figure developed on the western side of the Andes flowing westwards towards the Central Valley and to the Pacific Ocean. Those rivers and creeks are characterized by a generally high slope in the headwaters resulting in relatively high sediment fluxes. Most of rivers in the RM are characterized by a pluvial regime with maximum runoff during winter time. The exceptions are those watersheds that are partly located in higher altitudes, such as the catchment of San Ramón, showing a pluvio-nival regime (AC Ingenieros 2008). Due to the dry climate (Section 4.3.2), a large number of the smaller creeks coming from the Andean mountains only carry water periodically during winter time or during snow melt in spring.

72

4 Description of the study area

Figure 12: Main hydrological network in the Metropolitan Region of Santiago de Chile.

The two main rivers are connected through the Canal San Carlos which was constructed between 1772 and 1818 east of the city center. The main function of the Canal San Carlos is the transport of storm water coming from the Andean mountains into the parallel flowing Río Mapocho (Reyes 2003). Amongst other creeks the San Ramón River discharges into the Canal San Carlos. A characteristic phenomenon of the northern municipalities Huechuraba, Quilicura, and Colina, the western municipalities Pudahuel and Maipú and La Florida and La Pintana in the south are the wetlands resulting from a very highlying groundwater table. Especially after winter precipitation, the subsurface water emerges on the surface (Reyes 2003). Like numerous other sites in the RM with ecological value, these areas fall victim to the continuous construction activities, do influence the water balance and lead to a loss of biodiversity (Reyes 2003).

4.3 Physio-geographic description

73

4.3.4 Vegetation and urban green spaces For hydrological studies, vegetation cover and its distribution and pattern plays an important role as it influences the local water balance in various ways: -

Infiltration of rainwater, rainwater interception, effect of shading, evaporative cooling, storage of rainwater.

In summary, the proportion of green spaces in an area determines the amount of interception, infiltration, and runoff.

Figure 13: The influence of aspect on vegetation coverage exemplified for bushland (north-facing slope, right side) and woodland (south-facing slope, left side) in the catchment area of Quebrada San Ramón.

The vegetation cover in the rural part of the Metropolitan Region mainly depends on aspect, slope, elevation, and within the urban boundary on irrigation activities. While the urban green is dominated by irrigated grassland (parks), deciduous woodland, and private gardens, the vegetation in the rural areas is dominated by shrubs and bushes (e.g. Colligueja odorifera, Acacia caven, Lithrea caustico, Quillaja saponaria, Porlieria chilensis) in the sun exposed areas (north-facing slopes) and woodland in the remaining regions. Figure 13 shows the influence of aspect on vegetation at the example of bushland and woodland. Between 2,300

74

4 Description of the study area

and 3,000 m, the vegetation cover only consists of grasses, lichen, and moss. No vegetation can be found above 3,000 m (AC Ingenieros 2008). The creation of new public urban green spaces in Santiago de Chile was reluctant during the last decade. Three main reasons are: (i) not enough resources are foreseen for the maintenance of existing and construction of new green spaces, (ii) the existing green spaces that are open to the public can frequently be found in an undesirably poor conditions (garbage, destruction, etc.), and (iii) private resources from the users of the green spaces are requested for maintenance plus the public resources that are foreseen for the creation of green spaces are allocated to different institutions which do not always cooperate (Reyes 2003). The poor municipalities (e.g. Quinta Normal, La Cisterna, Pedra Aguirre Cerda) provide less than 2 m² of green spaces per inhabitant, where a “healthy” city possesses at least 9 m² (Reyes 2003). The flood hazard is accordingly higher in these areas. 4.3.5 In-depth study area: Quebrada San Ramón The catchment of Quebrada de Ramón (Figure 14 and 2 in Chapter 1 is located in the municipality of Las Condes and has a size of around 36.72 km² with a river length of 12.60 km. The river flows westwards into the urban area into the municipality of La Reina. The highest point in the catchment is Cerro San Ramón with an elevation of 3,253 m; the medium elevation is 1,400 m (AC Ingenieros 2008). At an elevation between 1,450 and 1,650 m material from a quaternary mass movement is deposited, resulting in a very flat slope in that area (Figure 14). Slopes outside the before mentioned flat part vary between 10° and 30°. Mass movements and erosion have previously been and still are frequent phenomena in the study area. That results in patches with accumulated material (Stumpf 2009) (Figure 14). The streams in the watershed do mainly have their origin on bedrock of the Abanico formation (compare Section 4.3.1). Very solid underground results in the formation of single cascades. A further description of the soils and geologic formations in the catchment is provided in Section 8.2.3 where the delineation of the hydrological soil groups for the modeling process is described.

4.3 Physio-geographic description

75

Figure 14: Geologic map of the San Ramón catchment after (Stumpf 2009).

The upper part of the basin is ecologically rich, mostly undisturbed, and constitutes habitat for a wide range of plant and animal species. The only installation in this part is a water intake station of the water supplier Aguas Andinas (Figure 15). This station is ignored for the modeling of single rainfall events.

Biodiversity in the catchment

Water intake station in the catchment

Figure 15: Biodiversity and water intake station in the catchment.

The lower part of the river (after entering the mancha urbana) is canalized. A range of hydraulic equipment was installed in this part in and along the water way: embankment stabilization with walls of stone, pipelines, water intake points, etc. (AC Ingenieros 2008). The canalized river ends after 4.5 km with its entrance in the Canal San Carlos at an elevation of 600 m. The hydrologic modeling process

76

4 Description of the study area

is solely carried out for the non-canalized part of the river, i.e. from its origin in the catchment to the location of the gaging station. Even though the regime of the catchment is pluvio-nival with the 0°C isotherm occasionally laying below 3,253 m a solely pluvial regime is assumed for the simplified modeling process applied in this research. However, this is partly being compensated by an adapted parameterization of the hydrological model (Chapter 8). Perez (2009) summarizes the projections of the elevation of the 0°C isotherm under climate change conditions as follows: Table 7: Location of the 0°C isotherm in the RM under climate change conditions (Perez 2009). Return period [years]

Base line [m]

A2 [m]

B2 [m]

2 5 10 20 50 100

2,225 2,550 2,700 2,815 2,940 3,015

2,730 3,120 3,325 3,495 3,685 3,815

2,715 3,095 3,285 3,440 3,605 3,715

5 DEFINITION OF A CASE SPECIFIC INDICATOR SET FOR THE RISK ANALYSIS The flood risk analysis and assessment carried out in the scope of this research is based on set of 23 indicators. The functions, importance, and strengths of indicators in general are described in Section 5.1. The process of selecting variables for flood risk analysis in the study area is lined out in Section 5.2, 5.3, and 5.4. The sections also give a short description of the content and relevance of each variable with respect to the research goals. How the variables are brought together in a functional scheme and how they are used for the practical analysis is lined out in Section 5.5. Section 5.6 explains how these variables were converted to indicators and implemented for their use in practice. 5.1 FUNCTION AND CHARACTERISTICS OF INDICATORS Birkmann (2005d) states that the employment of indicators requires an overall goal and guiding vision, which is in this case showing the complexity of the problem of flood risk in Santiago de Chile and the possibilities for reducing that risk. A good set of indicators for risk assessment is easy to understand, policy relevant, based on available data sets, is able to capture root causes of risk, statistically reproducible, and representative (Atteslander et al. 2008 p. 214, Birkmann 2005d, p. 12. Birkmann (2005d) names the ability to set priorities, to give a background for action, to raise awareness, and analyze trends as the most important functions of risk-related indicators. Atteslander et al. 2008 point out that the indicators need to be up to date and very close to reality in order to avoid obtaining neutral answers or evaluations from the interviewed people. Grunwald & Kopfmüller (2006) refer especially to sustainability indicators and sets of indicators when listing the following key functions: (i) information, (ii) orientation, (iii) steering, (iv) communication, and (v) integration (Grunwald & Kopfmüller 2006). Indicators are therewith different from variables. While indicators can be quantified and be assigned with a threshold value, variables are rather descriptive and are in this study regarded as a pre-stage of an indicator. That means that variables relevant for flood risk analysis are compiled and then transferred into indicators that can fulfill the functions as mentioned before. 5.2 SELECTION OF VARIABLES RELEVANT FOR THE ANALYSIS OF FLOOD HAZARD For the selection variables that physically determine the hazard, literature research was conducted. The characteristics of an urban system that lead to flood hazard

78

5 Definition of a case specific indicator set for the risk analysis

can be described straightforward by referring to measurable geographic and topographic components. The respective variables are described in the following. Amount of precipitation: In the case of Santiago, the amount of precipitation plays an important role in flood risk analysis, (i) as it determines the runoff in the river and (ii) as it determines the amount of runoff on the streets because storm water collectors are missing in a large part of the city. The higher the precipitation the more negative is the impact on flood risk. The critical precipitation value per event depends besides the slope and the current LULC on the value of all forthcoming hazard-related indicators, and the precipitation intensity. Generally, precipitation events with a total amount of rainfall of more than 65 mm per event cause floods in the study area. However, if the precipitation intensity is very high, floods result from events with approximately 50 mm within less than 24 hours (own finding). Infiltration capacity of the land use/land cover: The LULC is important as the occurrence and duration of flood events to a large part depends on the physical characteristics of the surface. This variable refers to the proportion of precipitation that infiltrates into the soil. With respect to the flood hazard, the classes which have low infiltration rates and/or little interception potential are to be rated critically (Canters et al. 2006). Relief: The slope and exposition of the terrain in the catchment physically determine the amount, direction, and the speed of surface runoff. The lower the slope, the slower the runoff, and the higher the likelihood of infiltration. Runoff: This variable refers to the measured amount of runoff in m³/s at the gaging station. The higher the runoff and the higher the excess, the larger the flooded areas or the deeper the flood. Thus, the higher the value, the more negative the rating. The delineation of a target value for that indicator depends on the location-specific capacity of river bed or the channel and the availability of flood protection measures. Slope of street: The slope and exposition of streets determine flow directions and local depressions possibly lead to an accumulation of water on the streets (urban flooding). Water height: The water height is the level of water (in m) outside the river during a flood situation. The water height has a direct influence on the damage and leads to a disturbance of the urban functioning (e.g. while people have troubles crossing streets with standing or flowing water, some vehicles can still pass). The affectedness of people, buildings, green spaces (habitats), and further urban infrastructure depends on their exposure, which is being analyzed in the scope of the vulnerability analysis. The variables presented above are not all equally relevant for the determination of the flood hazard. Rather, their interplay is important. A large part of the mentioned variables is brought together in a hydrologic precipitation-runoff model (Chapter 8): (i) the amount of precipitation, (ii) the infiltration capacities of the LULC, and (iii) the slope. One single indicator is derived from this hydrologic analysis: the runoff in m³/s at the gaging station. The target value for this indicator mainly depends on the capacity of the water course and determines the location-specific

5.3 Selection of variables referring to the elements at risk

79

high water threshold. The remaining variables, i.e. (i) the slope of the street, (ii) the existence of flow barriers, and again (iii) the infiltration capacity of the LULC finally affect the extent and water height of the flooded area. This last step is not part of the present study as previous hydraulic studies were carried out on this subject for the study area and the results can be used readily (Perez 2009). 5.3 SELECTION OF VARIABLES REFERRING TO THE ELEMENTS AT RISK Variables referring to the elements at risk can be identified straightforward: Number of people: The number of people in the hazard zone is used to quantify the amount of elements at risk. The higher their number the higher the risk (Cardona 2007, Taubenböck 2007). As it becomes obvious in the risk equation, there is no risk if there are no elements at risk or if the elements at risk are not vulnerable. Number of infrastructure: The number of infrastructure in the hazard zone is another component for the analysis of elements at risk. Cardona (2007) proposes to distinguish more detailed variables such as rupture in gas network, fallen lengths on HT power lines or electricity substations affected. Due to data availability this is not entirely feasible for this study, but the concept allows including such detailed information at a later stage. 5.4 SELECTION OF VULNERABILITY-RELATED VARIABLES While a range of widely accepted relevant characteristics and indicators is being presented in literature the actual conditions that determine flood vulnerability are to a certain degree very site-specific and location-dependent. The review of relevant literature and a number of field surveys and interviews carried out showed that the following variables are relevant for flood vulnerability analysis in the study area: Main construction material for roof, walls, and floor: The main construction material for roof, walls, and floor determines the physical fragility towards flood events and indicates the resistance against damage (Schneiderbauer 2007, Taubenböck 2007, Briguglio 2003, Clark et al. 1998), and also the social status (Cutter et al. 2003, Ranganath 2000). Moreover, some types of construction material allow humidity to remain in the walls or floor after flood events which can lead to health problems  an issue which was raised during field surveys (Reiter 2009). Position of building in relation to street level: The position of buildings in relation to the street level partly determines the likelihood of constructions to suffer damage in case of a flood event (Schneiderbauer 2007). In general, those buildings that are located above street level are less likely to experience adverse effects as the streets often function as waterways or channels and the water flows

80

5 Definition of a case specific indicator set for the risk analysis

by. Field surveys showed that people that live below or at street level were disadvantaged as the exposure to the flood is higher (Reiter 2009). Availability of flood protection on buildings: The availability or application of flood protection infrastructure such as small walls and backflow flaps lead to less exposure and therewith less physical vulnerability (Schneiderbauer 2007). With regard to the characteristics of floods in Santiago, even the availability of small protection measures is helpful. Nevertheless, the collection of data about individual flood protection measures is time consuming as field surveys are required. Respective information is only available for a small amount of the study area that was covered by the field surveys (Reiter 2009). Age: Especially the young and the elderly people are vulnerable towards natural hazards both because of their physical condition and their financial dependence (Schneiderbauer 2007, Dwyer et al. 2004, Haki et al. 2004, Cutter et al. 2003. Clark et al. (1998) name the dependence on care facilities as a further aspect: Workers might be forced to stay at home in order to take care of infants or elderly without getting paid. Nevertheless, the vulnerability of the elderly is minimized by their experience (Clark et al. 1998). Gender: Because of their stronger involvement in family life, sector-specific jobs, and lower wages, women are generally described as more vulnerable towards natural hazards than men (Wisner et al. 2004, Haki et al. 2004, Cutter et al. 2003). Above that, it was stated during the expert surveys that women are more emotional what makes them more vulnerable to suffer from adverse effects of hazardous events. Proportion of green spaces per neighborhood: This indicator can be used to describe the social status (Stow et al. 2007) as it is also used as an indicator for quality of life studies (Vásquez et al. 2007, van Herzele & Wiedemann 2003). In addition to the social significance, green spaces are relevant from a hydrological point of view. The higher the amount of green spaces in an area, the higher the retention potential, and the lower the flood hazard (Niehoff et al. 2002). This fact, however, is being covered by the flood hazard variable LULC. As this indicator in this shape has a positive effect on vulnerability its inverse value is used when calculating the vulnerability index after Equation 13. Level of education: Education plays a multi-dimensional role for flood vulnerability. On the one hand, good education offers better chances to find good employment and to reach higher living standards and therewith more resources to recover from the adverse impact of a hazardous event. On the other hand, specific education on environmental awareness and ecological issues leads to a better hazard preparedness and more resilience. On a long-term, a good education of broad parts of the population will even lead to a more sustainable way of thinking and planning (Schneiderbauer 2007, Velasquez & Tanhueco 2005, Haki et al. 2004). The higher the level of education, the lower the vulnerability. Household size vs. number of bedrooms: This variable is used indirectly to delineate the social status and to approximate the number of dependent people. The higher the household size, the lower is generally the social status and the higher is the amount of people affected  and therewith the damage (Haki et al. 2004, Cutter et al. 2003). Nevertheless, large households embody intrinsic social

5.5 Bringing together the relevant variables

81

networks and manpower which can in emergency situations be valuable (Velasquez & Tanhueco 2005). The household size in conjunction with the number of bedrooms seems to be a more valuable indicator than the population density. High population density alone is no one-to-one indicator, e.g. when comparing dense slum settlements with exclusive urban high rise buildings. Employment status: The employment status indicates the regularity of income and therewith the possibilities of a household to spare money for flood mitigation measures or preparedness. Interviews with residents during field surveys showed, that the employment status plays an important role for vulnerability analysis (Reiter 2009). Briguglio (2003) proposes indicators such as employment rate and income distribution  but the data with the required level of detail are not available for the study area. Experience with floods: Experience with damage resulting from flood events  which does to a certain degree resemble the “lessons learned” concept  does on the one side sensibilize people to the problem and does on the other side lead to the generation of private flood mitigation measures (Reiter 2009). Experience with damage has a positive influence on preparedness, especially those cases where people could completely recover (Birkmann 2005d, Velasquez & Tanhueco 2005, Wisner et al. 2004, Cardona 2003a). Knowledge about flood hazard: With respect to the concepts of resilience and coping capacities, (Cardona 2003a) names knowledge about flood hazards as an important variable. The more knowledge and information are available, the lower the vulnerability  and the higher the resilience and coping capacities. Knowledge about private protection measures: The knowledge about private protection measures leads to a diminishment of vulnerability as field surveys showed that financial resources are not real constraint for the construction of protection measures  at least not for short term protection measures such as sandbags (Reiter 2009). 5.5 BRINGING TOGETHER THE RELEVANT VARIABLES Bringing all variables together, results in a schema like Figure 16 shows. It is distinguished between direct variables and descriptive background variables. Direct variables are expected to have a direct influence on the hazard, vulnerability or elements at risk. Descriptive background variables in turn are characteristics that represent important factors influencing the primary, direct variables (compare Figure 16). Several variables are only available for a small part of the study area, i.e. for those households that were interviewed. That comprises (i) the availability of flood protection on buildings, (ii) the experience with floods, (iii) knowledge about the flood hazard, and (iv) knowledge about private protection measures. All selected variables are listed in Table 8. Their relation is pictured in Figure 16 and further discussed in Chapter 10. It is important to note that a large part of the variables depend on the governance structures, the institutions, instruments, and the value system and attitude of the decision makers. These factors are not expressed

82

5 Definition of a case specific indicator set for the risk analysis

through own variables but need to be considered when analyzing risk. Chapter 10 refers to these “framework conditions” in more detail. Table 8: Set of variables used for the flood risk analysis. Variables printed italic are only available for a limited amount of households that was interviewed during field surveys (Reiter 2009) Variable

Type

Data source & reference

Amount of precipitation Runoff Infiltration capacity of LULC Relief Slope of streets

Direct Direct Descriptive Descriptive Descriptive

Water height

Direct

Measured data (DGA 2010) Measured data (DGA 2010) RS data (Mockus et al. 2007) GIS, DEM (Saa et al. 2005) GIS, high resolution DEM (AC Ingenieros 2008) Previous study (Perez 2009)

Hazard-related variable

Vulnerability-related variables Main construction material of wall, floor & roof Position of building in relation to street Availability of flood protection on buildings Age Gender Level of education Household (HH) size Proportion of green spaces Employment status Experience with floods Knowledge about hazard Knowledge about private protection measures

Direct

Direct

Based on Arriagada & Moreno 2006, using census data (INE 2002) LULC classification GIS (AC Ingenieros 2008) Questionnaire (Reiter 2009)

Direct Direct Descriptive Descriptive Descriptive Direct Direct Direct Direct

Census data (INE 2002) Census data (INE 2002) Census data (INE 2002) Census data (INE 2002) LULC classification Census data (INE 2002) Questionnaire (Reiter 2009) Questionnaire (Reiter 2009) Questionnaire (Reiter 2009)

Direct

Variables referring to the elements at risk Number of people Number of buildings & infrastructure

Direct Direct

Census data (INE 2002) LULC classification, GIS (Saa et al. 2005)

5.5 Bringing together the relevant variables

83

Figure 16: Schema of variables relevant for flood risk analysis in Santiago de Chile, distinguished between descriptive background variables and direct variables.

84

5 Definition of a case specific indicator set for the risk analysis

5.6 FROM DESCRIPTIVE VARIABLES TO INDICATORS For the practical risk analysis, the variables need to be transformed into indicators and can then be applied in a GIS-based multi-criteria analysis (Chapter 9). As mentioned above, only one indicator is derived for the flood hazard analysis: the amount of runoff in m³/s at the gaging station. In the case of the elements at risk, the proposed variables are at the same time already indicators: the number of people in the hazard zone and the number of infrastructure in the hazard zone. The vulnerability variables show a very diverse nature. To make all of them comparable they are best translated into ratio scale indicators, i.e. all indicators are expressed as relative frequencies per building block in order to treat them as quantitative measures in a vulnerability index. The indicators are thereby always formulated in that way that a high indicator value represents high vulnerability. The variable age as an example was transformed into the indicator “Proportion of people below 5 or above 65 years old per neighborhood” to capture those age groups that show a higher vulnerability against floods. Likewise, the indicator “Proportion of households that have not experienced floods” refers to households with a higher vulnerability as field surveys showed that households that have already experienced a flood event are better prepared and thus less vulnerable than others (Reiter 2009). Table 9 shows how the variables/characteristics were translated into indicators and also describe the content of each indicator for vulnerability analysis and for the analysis of the elements at risk. How values from the analysis of the available data sets were finally assigned to each indicator is described in Section 9.1. Table 9: Variables and resulting indicators for the analysis of hazard, vulnerability, and elements at risk with content description. Variable

Indicator

Content

Runoff at the gaging station in m³/s

Amount of river runoff measured at the gaging station

Main construction material of wall, floor & roof

Proportion of buildings with poor construction material per building block

Position of building in relation to street

Proportion of buildings at or below street level per building block Proportion of people under 5 and above65 years old per building block

Buildings with roofs made of garbage, floor made of plastics, concrete or soil and walls made of garbage or clay Buildings at or below street level

Hazard-related indicators Runoff Vulnerability indicators

Age

People below 5 and above65 years old

5.6 From descriptive variables to indicators Gender Level of education

Household (HH) size

Proportion of green spaces

Employment status

Income status

Proportion of female population per building block Proportion of heads of household with poor education per building block Proportion of households with more than 2.5 people per room per building block Proportion of green spaces per building block Proportion of people without employment per building block Proportion of people without permanent income per building block

85

Female population Households with heads of household having no or incomplete basic education Household with more than 2.5 people sharing one bed room All types of public and private green spaces brownfields only if covered with vegetation People that are seeking work or that are permanently unemployable People that are working for the family, having employment but are not working, students; retired, homemaker

Indicators referring to the elements at risk Number of people Number of infrastructure & buildings Vulnerability indicators only available from household surveys Availability of flood protec- Proportion of buildings withtion on buildings out flood protection Experience with floods Proportion of households that have not experienced floods Knowledge about hazard Proportion of households without knowledge about floods Knowledge about private Proportion of households protection measures without knowledge about private protection

Absolute amount of inhabitants per building block Absolute amount of infrastructure & buildings in [m²] Buildings without flood protection measures Households that have no experience with floods Households without knowledge about flood hazard Households without knowledge about private flood protection measures

6 DATA BASE AND DATA PROCESSING A central part of this study consists of the analysis of digital geodata (remote sensing, GIS, and census data) and empirical data. An overview about the data needs and data used in this research for each of the three main steps in risk analysis is given in Table 10. Table 10: Data needs for flood risk assessment in the present research. Step

Method

Input data needs

Hazard analysis

Hydrological modeling

Mapping of elements at risk

Remote sensing (RS) & GIS analysis Statistics RS & GIS Interviews & field surveys Statistics

-

Vulnerability analysis

-

Measured precipitation data Measured runoff data Elevation & slope (GIS data) LULC (ASTER) Soil and/or geology data Data about location of buildings & infrastructure (Quickbird & GIS data) Census data LULC (Quickbird) Completed questionnaires

-

Census data

This chapter comprises the description of all digital geodata and empirical data used in this study (compare Figure 17). First, the hydro-meteorological data used for the hazard analysis are described in Section 6.1. Section 6.2 gives an overview about the processing of the hydro-meteorological data. Second, an overview about all available GIS and remote sensing data used for all three parts of the risk analysis is given in Section 6.3. The following Sections 6.4 and 6.5 line out which preprocessing operations were done with the GIS and remote sensing data. The census data base used for the analysis of elements at risk and their vulnerability is presented in Section 6.6 and the processing in Section 6.7. Finally, the methodology for the conduction of expert interviews, household surveys (Reiter 2009), and questionnaires is presented in Section 6.8. This chapter concludes with the processing of these data that are likewise used for the vulnerability analysis (Section 6.9).

88

6 Data base and data processing

Figure 17: Overview about input data and their usage for the further analysis (DEM = digital elevation model).

6.1 HYDRO-METEOROLOGICAL DATA BASE Runoff [m³/s] and precipitation [mm] values are required amongst others as input for the application of the hydrological model HEC-HMS (Section 8.3). Runoff data are available for the study area at a daily temporal resolution. However, a comparison with the precipitation data showed that the data quality is unsatisfactory and not reliable (Perez 2009). These data do thus need to be omitted for the modeling process. As the rainfall intensity is considered being an important parameter in hazard assessment, the data should have the highest possible temporal resolution. In this study, hourly measured data were available for the time frame between June 01, 2000 and July, 31 2010 for the station Cerro Calán. Artan et al. (2007) proof that measured rainfall data are much better suited for the analysis of short-term events than satellite derived rainfall data.

6.2 Processing of the hydro-meteorological data

89

Figure A40 in the appendix shows the location of the Cerro Calán station and also shows the location of the gaging station Quebrada Ramón de Recinto Emos right within the study area. This station would be most valuable for this study but sufficient data are not available. Table 11 gives an overview over the available hydro-meteorological data. Table 11: Overview over available hydro-meteorological data. Content

Name of station

Time frame

Precipitation hourly

Cerro Calán

20002010

Precipitation daily

Quebrada Ramón en Recinto EMOS Quebrada de Macul Cerro Calán Antupiren Terrazas Oficinas Centrales DGA

AprJuly 2008 2006 – 2008 1975 – 2008 1979 – 2008 19702008

As of the only station that lies within the catchment of Quebrada San Ramón  Quebrada Ramón en Recinto Emos  precipitation data are only available from 01.04.2008 to 31.07.2008, data from comparable stations were used for data substitution (see Table 11). For the time frame between April and July 2008, daily data of the stations Quebrada Ramón en Recinto Emos, Quebrada de Macul, Terraza Oficinas Centrales DGA, Antupiren, and Cerro Calán were available and were analyzed statistically. The correlation coefficient of the precipitation data for the respective time frame ranges from 0.983 (Quebrada Ramón en Recinto EMOS with Terraza Oficinas Centrales DGA) to 0.994 (Quebrada Ramón en Recinto EMOS with Quebrada de Macul). The correlation coefficients with the two remaining stations Antupiren (0.986) and Cerro Calán (0.993) lay inbetween. Figure A42 pictures the values in a graph. In terms of absolute values, the maximum difference between the stations is 18.1 mm which means that there was an almost 61 % higher amount of precipitation at the station Antupiren than at Quebrada Ramón en Recinto Emos. Because of high correlation and sufficient data availability with a high temporal resolution (daily, and partly hourly data), the station Cerro Calán was selected for data substitution. As data from Quebrada de Macul  which is located nearby in a catchment physically most similar to the catchment of San Ramón  were only available from 2006 onwards, the station has not been used in the further analysis. 6.2 PROCESSING OF THE HYDRO-METEOROLOGICAL DATA HEC-HMS requires input data without data gaps. For the modeling process, only periods with significant precipitation (above 65 mm per event) are being considered as the focus is laid on simulating runoff after extreme events. Thus, events were selected both after amount of rainfall and data availability.

90

6 Data base and data processing

For the application of a distributed modeling approach in HEC-HMS the data need to be made available in spatially equal grid cells even if only one precipitation station is used for the basin. In a first step, the time series containing the precipitation data that are available in MS Excel format need to be converted in the DSS-format supported by HEC-HMS using the program HEC-DSSVue, which is the graphical user interface for the HEC data storage system. The program GageInterp, that was developed for interpolating rainfall data, was applied to bring the rainfall data time series in the required grid-format (*.dss) with a cell size of 50 m (USACE 2006). The interpolation process and file generation parameters are defined in a control file that has to be written by the user. The control file contains three sections: Time frame and time interval for the interpolation process, type, content, extent, and cell size of the resulting output file, and third, information about the gaging station (USACE 2006). From the example control file for a storm event in May 2002 is shown below it already becomes apparent that the geographic location of the basin is on purpose defined to be within Standard Hydrological Grid (SHG) of the conterminous United States (compare further Section 8.2.1 and Universidad Politecnica de Catalunya). The output of this interpolation process are hourly rainfall data from the station Cerro Calán covering the time slot between 24 may, 2002, 24:00 hr and 28 may, 2002, 23:00 hr. Lines starting with an asterisk are comments only. *CONTROLFILEforinterpolationofrainfalldatausingGageInterp  *ŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞ *FIRSTSECTION: *Interpolationprocess TimeStart:24May2002,2400 TimeEnd:28May2002,2300 TimeStep:60 Weight:NEAR Range:UNLIMITED *LapseandBiasareleftoutatthispoint *ŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞ *SECONDSECTION: *Outputgridparameters: GridType:SHG CellSize:50 GridOrigin:645080125 GridRows:200 GridCols:300 OutUnits:mm OutDataType:PERŞCUM OutParameter:precip OutFile:May02INT2.dss OutPath:/SHG50/CALAN/PRECIP///INTERPOLATED/ *ŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞŞ

91

6.3 GIS and remote sensing data base *THIRDSECTION: *Precipitationgageparameters: DSSFile:May2002.dss Gage:CALAN,6300000,360000 Timezone:CST,Y Path:/RAMON/CALAN/PRECIPITATION/24May2002/1HOUR/OBS/ Datum:NAD83 Text:CALAN

6.3 GIS AND REMOTE SENSING DATA BASE Tables 12 and 13 give a brief overview about the vector and raster data used in this study. Table 12: Vector data used for this study. Content

Scale

Data source

Contour lines

1:50,000 1:2,500 n. a. n. a. n. a. n. a.

DOH-DOF DGA INE OTAS project (Saa et al. 2005) OTAS project (Saa et al. 2005)

Location of gaging stations Administrative units (comunas, manzanas) Urban street network Infrastructure (Sport, education, health facilities, police stations, municipal services, supermarkets, banks, etc.) Table 13: Raster data used for this study. Content/source

Date

Geometric resolution

Quickbird pansharpened ASTER

December 19, 2006 February 01, 2002 February 09, 2005 February 04, 2009

0.61 m 15 m (band 13) 15 m (band 13) 15 m (band 13)

DEM delineated from 50 m contour lines for catchment area from 2.5 m contour lines for urban area

n. a.

10 m

Flood hazard maps (Perez 2009)

2009

2.5 m 5m

Perez (2009) derived flood hazard maps in the scope of a Master’s thesis at the Universidad de Chile. These maps are used as input for the flood hazard and risk calculations in this study. In a first step, runoff volumes for the present area of interest (Quebrada San Ramón) were delineated using a synthetic unit hydrograph

92

6 Data base and data processing

(Linsley) and assuming conditions of climate change. Therefore, estimations about the future development of temperature, maximum daily precipitation, and the location of the 0°C isotherm were formed using the IPCC scenarios A2 and B2. The hydraulic model HEC-RAS was in a second step applied to process the runoff volumes estimated from the climate change scenarios. Hydraulic transversal profiles available every 50 m enabled in a third step the preliminary estimation of points of overflow from the channel. The overflow from the channel was then used to calculate a flow pattern in the street network. Simulating the movement of the water through the street network though was not possible using HEC-RAS, thus the hydraulic model MOUSE was applied for this part of the modeling process. A methodological drawback still is that the transversal street profiles need to be simplified to a rectangular shape. As a consequence the water does not pass over the sidewalk as it would be in reality (Perez 2009). However, as the simulations indicate, the amount of runoff overflowing into the street network is three to seven times as high as the runoff remaining in the channel indicating the highly insufficient capacity of the water courses. The precipitation that is falling on the street during a storm is thereby not even included in the modeling process and would thus additionally add to the street runoff. Section 9.2 further explains the generation of the hazard maps, their analysis, and implementation in the risk calculation. 6.4 PRE-PROCESSING OF THE GIS DATA To create a common spatial reference, all available GIS data that had been projected to Universal Transverse Mercator (UTM) coordinate system Zone 19S with PSAD 1956 ellipsoid were converted to UTM with the ellipsoid WGS84 Zone 19S to spatially match the applied satellite data. Therefore, transformation parameters that were made available by the Instituto Geográfico Militar (IGM) were used (see Table A35 in the appendix). The further pre-processing of the GIS data that were used for the hydrological modeling was done using HECGeoHMS (USACE 2003) (Section 8.2). The first prerequisite for the application of HEC-GeoHMS is the existence of a digital elevation model (DEM), which was created prior to the application of HEC-GeoHMS based on 50 m contour lines that were available for the entire catchment area. The contour lines were interpolated to a 10 x 10 m grid using an adapted inverse distance weighted (IDW) algorithm after Hutchinson 1989, which has a strong emphasis on drainage enforcement by removing sinks and pits. This is of value as this approach is more time-efficient and accurate than removing sinks from an existing DEM. As Hutchinson (1989) states, the ex-post removal of sinks frequently leads to a smoothing of existing terrain features and therewith to a falsification of the terrain. The drainage enforcement algorithm considers local sinks, saddles, and their surrounding values. If either the sink or its neighboring saddle point is not associated with a measured elevation point, the interpolated value is adapted to the measured value to ensure drainage. If none of the two points is associated with a drainage value, both values are modified until drainage

6.5 Pre-processing of the remote sensing data

93

is ensured. In case both points have real, measured values, the user decides whether or not the elevation data are altered in order to enforce drainage (Hutchinson 1989). Stream line data can be incorporated to further improve the results. The accuracy of the river network and the exact method of how the data were generated are not precisely known. Thus, respective data were not used and the DEM was generated based on the contour line values only. The delineated DEM was exported as a grid in ASCII-format. 6.5 PRE-PROCESSING OF THE REMOTE SENSING DATA Figure A41 gives an overview over the images from ASTER and Quickbird. The pre-processing of the data comprises the geometric correction and co-registration (Section 6.5.1) and the closure of data gaps (Section 6.5.2). 6.5.1 Geometric correction and co-registration When satellite images are acquired they usually show a geometric distortion caused by variance in altitude, attitude, and velocity of the sensor platform, relief influence, atmospheric refraction, and systematic errors (Erdas 1999, p. 253, Lillesand & Kiefer 2004, p. 474. All used data were already geometrically corrected (i.e. projected to a map coordinate system) with the main errors removed (quality level 1B for the ASTER data, *.til-format for the Quickbird data). However, they still showed some minor random errors and did consequently not exactly match each other. The irregular mismatch of the ASTER data had a maximum value of 1 pixel (15 m) which was found to be acceptable for this research. The accuracy of the Quickbird data was likewise acceptably low. When importing the satellite data into the image analysis software ENVI the ground control points used as reference for the correct geometric positioning of the image are automatically read from the metadata information of the image (header file). Thus, no additional effort was taken by the interpreter as no other reference information for a better geometric correction of the data was available. 6.5.2 Data coverage and data gaps For the further processing, all available raster data were subsetted to fit the size of the catchment area and the urban parts of the municipalities La Reina and Peñalolén, respectively. The complete catchment of the Quebrada San Ramón is only covered by the ASTER image from 2005. The data gap in the two remaining ASTER images comprise mainly areas with open rock underlying minimal changes over time so the temporal gap between the acquisition dates can be neglected for that part. For that reason, the ASTER data from 2002 and 2009 data were mosaicked with the ASTER image from 2005 to cover the missing part of the watershed.

94

6 Data base and data processing

The Quickbird data in turn do only cover the urban areas and the adjacent urban fringe up to an elevation of approximately 1,700 m, and small parts of the watershed of San Ramón above the gaging station. Therefore the Quickbird data are used for the vulnerability analysis only (Sections 7.1.2 and 9) and the ASTER data are used for the hydrological modeling process. 6.6 SOCIO-ECONOMIC DATA BASE Selected data from the 17th official census of the population and the sixth census of the apartments/dwellings of Chile collected with reference to April 24th, 2002 are available on manzana, i.e. building block level (INE 2002). The census data are not collected to assess vulnerability, but even though they are only available for one point in time every decade on the maximum accuracy of a building block, they have in previous studies proven to show a potential to contribute to vulnerability assessment (Ebert [Müller] et al. 2009, Claire 2007, Cutter et al. 2003, Javed 2005, Clark et al. 1998). Using the software RETADAM  that was developed for the analysis of the Chilean census data  additional composite indicators can be derived (see Section 6.7, e.g. Heinrichs et al. 2009). The information used in this study comprises the variables derived from the census data base as listed in Table 14. Table 14: Census variables available from 2002 survey. Content Settlement density: Number of households and apartments Number of households Gender and age groups Number of male population Number of female population Population under 5 years Population from 65 to 69 years Population from 70 to 74 years Population from 75 to 79 years Population older than 80 years Social status I: Employment and education Working with income Not working, but having employment Seeking employment, having worked before Working for family without payment Seeking employment, not having worked before Homemaker Studying Retired Permanently unemployable Other situation

Source INE 2002 INE 2002

INE 2002

95

6.7 Pre-processing of the socio-economic data Number of households where head of household is without education with incomplete basic education with complete basic education with incomplete medium level education with complete medium level education with secondary level technical education with university education Social status II: Building quality Amount of households with roofs made of bricks (adobe, metal, cement) shingles (wood, asphalt) concrete slabs zinc schist glass fiber phonolite garbage (cardboard, plastics, etc.) Amount of households with floors made of parquet ceramic tiles timber flooring carpet cement tiles plastics (linoleum, etc.) concrete soil Amount of households with outer walls made of reinforced concrete, stone bricks Panels, blocks (prefabricated) wood or partition wall plasterboard & adobe/clay bricks garbage (cardboard, plastics, etc.)

delineated from census Heinrichs et al. 2009

INE 2002

INE 2002

INE 2002

6.7 PRE-PROCESSING OF THE SOCIO-ECONOMIC DATA Pre-processed census data were made available in table format in the scope of the research project. The maximum spatial resolution and therewith the degree of accuracy of the socio-economic data is the administrative level of building blocks. These units  independent from their size  were treated as being homogeneous. To bring both data sets  the manzana outlines and the contents of the census data table  together, the census codes from 2002 used by the official statistics office (INE) were used as linking point. As both data sets did not match in some

96

6 Data base and data processing

parts because of administrative changes, several census blocks were omitted for the further analysis in order to avoid errors. This applies primarily to nonpopulated areas such as transport infrastructure as well as newly constructed areas. 6.8 EXPERT INTERVIEWS, HOUSEHOLD SURVEYS, AND QUESTIONNAIRES 6.8.1 Expert interviews To obtain non-spatial information on how flood risk is being generated in Santiago de Chile, several expert interviews were conducted with professionals that are working in fields and institutions that were assumed to be relevant in this aspect. All experts were interviewed personally in their offices and did in most cases invite colleagues working in the same field. Depending on the partners, the interviews were conducted in Spanish or English language. To not lose information and to minimize errors/gaps of acoustic understanding all interviews were recorded with the interviewees and transcribed by Chilean native speakers. Interviews carried out in English language were not transcribed. Dates for the meetings were made by telephone and email, a file containing some open core questions to be asked in the interview was provided to each interviewee beforehand to allow for a better preparation, and to already give some more detailed information about the envisaged content of the talk. The interviews always started with a short introduction about the scope of this research and were then conducted according the prepared manual, making it a semi-structured approach with a non-standardized question catalogue (Atteslander et al. 2008, p. 124 ff.). Each interview was prepared individually depending on the respective background, responsibilities, position, and area of work of each interview partner. Where more than one interview partner was present, questions were generally answered by the person that was most involved in the respective topic or could according to the principal interview partner provide most information. Group discussions did in most cases not evolve. In some cases, non-structured, non-standardized narrative interviews were conducted (Atteslander et al. 2008, p. 133). As these interviews were mostly not scheduled they have not been recorded but notes were taken. Table A36 in the appendix shows which type of interview was conducted with whom. 6.8.2 Household surveys While information derived from geodata (GIS, remote sensing data) or from census data cover large areas and are mostly readily available, they have generally not been acquired specifically to obtain information about flood risk. Especially for the analysis of vulnerability, additional data sources are required to get more causal information on how vulnerability is being generated. Geodata alone do also only provide a limited level of spatial detail, i.e. individual characteristics cannot

6.8 Expert interviews, household surveys, and questionnaires

97

be delineated (Ebert [Müller] et al. 2009). To get knowledge on why and how households are vulnerable against flood events, which variables and dimensions are most important and how the affected people personally evaluate the issue of floods, interviews were carried out with households living in flood prone areas according to the PRMS (Reiter 2009). As the goal for the data acquisition was therewith fixed, the most appropriate method to obtain the desired information is a structured interview (Atteslander et al. 2008). As in the scope of this research it was practically impossible to interview all households potentially affected through floods, samples needed to be selected. The selection of the household was mainly dependent on the accessibility, the presence of a household member, and their willingness to participate in the survey (Reiter 2009). A total of 82 household surveys was conducted in the municipalities of La Reina and Peñalolén by Reiter (2009) using a standardized questionnaire. A pre-test carried out in the field ensured that the closed and open questions that were posed in Spanish language were clear and understandable. Possible additional questions could be clarified during the interview as all questionnaires were completed personally with the interviewees. The selection of variables to be contained in the questionnaire was based on previous flood vulnerability studies, results from expert interviews (Section 6.8.1), and hypothetic assumptions (Reiter 2009). Following the vulnerability concept of Cardona (2003a), the variables were partly adapted and grouped using the categories “physical components”, “socio-economic dimension”, and “lack of resilience”. The list of variables contained in the questionnaire comprises: Number of floors, Position of building in relation to street level (elevation), Slope, Structural flood protection measures, Income groups, Level of education, Employment status, Household size (number of dependent people), Experience with damage (material or immaterial), Knowledge about protection measures, Occupancy, Taking of permanent protection measures, Ownership status, Social networks, and Insurance. 6.8.3 Questionnaires An additional questionnaire was designed to investigate the specific importance of vulnerability indicators in the study area from the experts’ point of view. During the household surveys, a range of variables and characteristics related to flood vulnerability were obtained but their importance and weights that need to be known for the calculation of a vulnerability index (Section 9.4) did not entirely and reliably become clear from the household survey alone. Therefore, a standardized questionnaire with closed and open questions (Atteslander et al. 2008, page 136 ff.) was designed. The closed questions comprise a list of vulnerability indicators and five possibilities to rank their importance (100 % = very important, 75 % = important, 50 % = medium important, 25 % = little important, and 0 % = not important).

98

6 Data base and data processing

In detail, the vulnerability indicators are the following: Predominant construction material of the building, Location of building in relation to street level (elevation), Availability of structural flood protection measures, Age, Gender, Socio-economic level, Employment status, Usage of building (commercial, residential, industrial, ...), Experience with floods, Knowledge about floods, Knowledge about private protection measures, Proportion of green spaces per building block, Distance to channel. The open questions in the questionnaire leave the opportunity to add comments to the experts’ evaluation of each indicator and to list further aspects and variables relevant for the analysis of vulnerability. The Spanish questionnaires were sent out together with a cover letter by email to experts that were found to have relevant knowledge about flood-related vulnerability in the study area and that had previously been in touch with the research initiative “Risk Habitat Megacity” in which this PhD research is embedded. It was assumed that this would enlarge the response rate. Eleven out of 50 questionnaires were returned by email within a period of three months between November 2009 and February 2010, whereby a reminder was sent out after seven weeks leading to a doubling of the previous response rate. All closed questions were always completely answered; the open (voluntary) questions have only partly been answered. The cover letter for the questionnaire contained important background information about the research and goal of this expert survey, and did furthermore ask the expert to name additional people if they did not complete the questionnaire by themselves. This opportunity was seized in one case. In two cases apart from the eleven returned questionnaires the experts returned their emails with the remark to send it to other institutions as the content lies outside their scope of knowledge or responsibility. The experts got the opportunity to declare their interest for the final findings of this survey. 6.9 PROCESSING OF THE EMPIRICAL DATA 6.9.1 Expert interviews The expert interviews carried out in Spanish language were analyzed only based on their transcription. The advantage of that method is that almost no content information got lost. The disadvantage is that emotions expressed through the voice and tonation of the interviewee could not be included and interpreted. All relevant aspects and information given by each expert or group of experts with respect to the research question of this research were extracted from the Spanish documentation (transcription) of the conversation. The interviews carried out in English language were analyzed directly from the recorded audio file. All relevant qualitative and quantitative information were noted to be used for the further research.

6.9 Processing of the empirical data

99

6.9.2 Household surveys The answers given during the household surveys were entered into statistic software (SPSS) for their further analysis and interpretation. The goal of the data analysis was mainly to investigate how the different characteristics of a household relate and if and how they determine the level of vulnerability. Therefore, methods of descriptive multivariate statistics were applied for the data analysis: Correlation and regression analyses were carried out. The correlation analysis delivers a correlation coefficient r which shows if there is a linear relation between the variables. It is purely descriptive and does neither give evidence about causal relations nor about strength and direction of the relation (Atteslander et al. 2008). To identify the strength of that linear relation, a regression analysis was applied. The challenge and hindering factor in this statistical analysis was the lack of a reference value that can be used as dependent variable. Vulnerability values would be the optimal measure but do not exist for the study area. Therefore, and with respect to the regularity of the flood events, the evaluation of the personal affectedness of the households was used as reference. Bivariate regression analysis was used for the evaluation of the relevance of each variable obtained from the questionnaire and the weights of the variables have then been determined using a logistic regression analysis (Reiter 2009). The findings from this statistical analysis are presented in Section 9.4.1.2 in the scope of the complete flood risk analysis. 6.9.3 Questionnaires The results from questionnaires that contain the evaluation of each vulnerability variable were entered in a table. Besides the quantitative evaluation of each indicator several questionnaires contained further qualitative judgments of the experts. This additional information was entered in the table alike. The goal of this investigation was to determine the weight, i.e. relevance of each variable. Thus, the answers given by the different experts were compared in a first step. To start with, the arithmetic mean and standard deviation is calculated. Figure 18 shows the outcome of the survey: the evaluations vary  depending on the variable  among the experts but also show a range of similar evaluations (values with small standard deviations). The lower the general mean value for each indicator  i.e. the lower its importance in average  the higher the standard deviations. The final analysis concentrates on three values with high average ratings and low standard deviations to deliver more stable results. The importance of green spaces has a standard deviation of 0.3 which is comparably high. However, this indicator was still included in the final analysis as its relevance is practically physically defined. The varying perception of the role of green spaces is then discussed in Chapter 11.

100

6 Data base and data processing

Figure 18: Average evaluation of each vulnerability variable and its standard deviation according to the experts.

The results of the quantitative part of the analysis are furthermore summarized in Section 6.8.3. The qualitative statements given in the questionnaires support the interpretation of the overall flood risk analysis and assessment, and the suggestion of flood prevention and mitigation measures (compare Chapters 9 and 10).

7.1 The theory of remote sensing-based land-use classifications

101

7 REMOTE SENSING DATA ANALYSIS This chapter describes the processing of the remote sensing data and explains how they are used for the flood risk analysis. This chapter starts with an introduction to the methods of land-use/land-cover (LULC) classifications using remote sensing data (Section 7.1). One of the central geodata processing steps is the LULC classification and its changes over time using remote sensing data and the delineation of information relevant for the flood risk analysis and assessment. Section 7.2 delineates the LULC and its changes in the catchment area as the influences of these changes on the flood hazard will be explored in the following Chapter 8. Section 7.3 describes the derivation of the current LULC pattern of the urban area as this information are used to describe the elements at risk and parts of their vulnerability in Section 9.1 (compare Figure 17). 7.1 THE THEORY OF REMOTE SENSING-BASED LAND-USE CLASSIFICATIONS The following section will give a brief theoretical background about pixel-based and object-oriented image classification, two approaches that are in the first instance based on analyzing the spectral characteristics of an image. The spectral characteristics reflect the conditions of the earth’s surface as a result of the material-specific radiation properties. The sensors installed on earth observation satellites measure the electromagnetic radiation and transform them into digital signals. Passive sensors (e.g. sensors on Landsat, Quickbird, and Terra satellites) measure the proportion of radiation that is naturally emitted or reflected (reflected sunlight) by the surface. Active sensors (e.g. sensors on radar platforms such as TerraSARX) emit energy and measure the backscatter to obtain information about surface characteristics (Richards & Jia 1999). The information received by the sensor can then be transformed into grey values and displayed in a histogram. The goal of the image classification is the generation of a data set in which pixels with similar spectral characteristics are assigned to the same thematic class, i.e. the generation of a LULC map. During the pixel-based classification, each pixel in the image is analyzed individually while the object-oriented analysis (OOA) approach is based on the generation of relatively homogeneous objects from a group of similar pixels that are being analyzed as units (called objects or segments) with spectral, textural, informational, shape, and contextual information. Pixel-based classification approaches can be both unsupervised if no or little ground truth data are available for the study area or supervised  if sufficient information about the study area are available  and unsupervised. An unsupervised classification only relies on the statistic image content while supervised classifications include class descriptions by the interpreter and can be both parametric and

102

7 Remote sensing data analysis

non-parametric. OOA approaches are always supervised and require a higher interaction of the user, as a segmentation of the image has to be performed prior to its classification. The two main types of image classification approaches will be presented with their advantages and disadvantages in the following. 7.1.1 Pixel-based classification To perform a supervised pixel-based classification, informational LULC classes need to be created. The definition of the classes is oriented at the natural or manmade setting of the study area. Classes have to be defined and described before the classification process is performed. Most commonly, the description of the classes is achieved by selecting representative samples of each class from the image, also referred to as training areas (Richards & Jia 1999). Training areas should be homogeneous areas with a total size of at least 100 pixels, they should be evenly spread over the entire scene, and they should be located closed to landmarks to enable a verification of their position in nature. The class descriptions have to be unique and unambiguous, that means that overlap in their feature space need to be reduced to a minimum. Thus if the grey value histogram of the training areas shows a bimodal distribution in one of the spectral bands the class should be split up in two or more classes to avoid confusion during the assignment of pixels to classes (Richards & Jia 1999). Class signatures are then developed from the training areas. Separability measures can be calculated to validate the spectral distinction between the class signatures (Erdas 1999). Different decision rules are available to assign pixels to classes based on their signatures: (i) maximum Likelihood classifier (MLC), (ii) Parallelepiped classifier, and (iii) minimum Distance/Nearest Neighbor classifier are the most commonly applied options. Each of them is using different parametric or non-parametric approaches for the class assignment (Erdas 1999, Richards & Jia 1999). The MLC, which was selected for the pixel-based classification of the ASTER images, is a parametric approach, i.e. statistic measures are used for the class assignment. The basic assumption is that the probability of a pixel to belong to a certain class is equal throughout all classes and that the gray values of each input band have normal distributions. Several other classification methods exist for the supervised image classification, e.g., expert or rule-based classification (Suzuki et al. 2001), sub-pixel classification, artificial neural networks (Ripley 1996), and local binary patterns (Pietikaeinen 2005). The expert classifier allows for the incorporation of additional vector and raster data to enlarge the possibilities for the description of image pixels. All class descriptions are expressed through hypotheses with certain probabilities. This approach was used for the classification of the ASTER data and is further described in Section 7.2.

7.1 The theory of remote sensing-based land-use classifications

103

7.1.1.1 Advantages The main advantage of pixel-based classification is that every pixel is treated individually and that no generalization is done during the entire analysis. Scenes acquired by different sensors or taken at different dates remain more likely comparable to each other as long as they show the same geometric resolution. The main argument for that is that no differing outlines of object borders can disturb. 7.1.1.2 Disadvantages The main disadvantage of pixel-based approaches also results from the fact that every pixel is regarded individually and independently from its surrounding pixels. Suzuki et al. (2001) argue that pixel-based classification reaches its limits when it comes to structural knowledge (shape and size of objects, (Suzuki et al. 2001)). Classifications using the pixel-based approach very often show the salt and pepper effect, which describes single wrongly classified pixels that result from small spectral inhomogeneities of objects (e.g. grass on a roof). That might for some applications complicate the further processing of classification results in a GIS, since an even higher number of polygons (sometimes at the size of single pixels) is resulting from a very heterogeneous classification result. Another limitation is that the signal that is represented by a pixel in the image does not necessarily represent the spectral properties of exactly the area shown in that particular pixel, as influences from the surrounding area are possible (Blaschke & Strobl 2001). This phenomenon can especially be observed in the strong reflecting urban built-up area. 7.1.2 Object-oriented classification It is obvious for humans that objects that look like homogeneous areas in an image belong to the same class in nature. The imitation of this human perception equals the concept of OOA. A group of adjacent pixels with similar properties (e.g. spectral values) are aggregated to non-overlapping image objects or segments. This process is called image segmentation (compare Figure 19 and A43). There are several methods to perform the segmentation of an image, e.g. thresholding, edge detection, and region-based segmentation (Pal & Pal 1993) or point-based, edge-based, region-based or mixed approaches (Schiewe 2002). All methods will lead to a completely segmented image with differing segment outlines  depending on the characteristics of the input image, the chosen algorithm and the parameters defined. All approaches follow the principle of value similarity, referring to the homogeneity of pixels (Schiewe 2002). The formation of segments (i.e. the type and degree of fusion of pixels) depends on the definition and accepted threshold value of a homogeneity criterion (Schiewe 2002). The larger the threshold value for a homogeneity measure, the higher the likelihood that adjacent pixels are merged, and the smaller the number of resulting image

104

7 Remote sensing data analysis

objects. In the case of eCognition Developer 8 classification software, which was used for this research, the desired average sizes of the image objects (compare Figure 19) have to be specified as one first part of the segmentation process. Each created object consequently represents the average gray value of all pixels contained in one segment. Figure 19 exemplifies the concept of various image levels implemented in the software. Each image level contains differently sized image objects. Each object does thereby know its shape, position, spectral properties, as well as its neighboring, sub-, and super-objects (Definiens 2007). Once the first image level with small image objects is created, the objects can be further aggregated in higher image levels. The original segment outlines thereby remain, but they are aggregated. The way of fusing the small segments to bigger segments depends on the choice of segmentation parameters (average segment size, spectral, and shape properties) that have to be specified individually for every image level. Depending on the image resolution, the smallest objects that can be identified in the image (e.g. cars and single trees on the pansharpened Quickbird image) can be classified using small image objects. Larger, homogeneous areas such as lakes and grassland can be classified using larger image objects on a higher image level. A hierarchical network of image objects can be constructed where the resulting objects know their sub-, super-, and neighbor-objects (Figure 19). If the image content is rather heterogeneous  as in the present case  the classification can also be carried out on one single image level. Relations between different classes in different image levels or on the same image level can be used as a criterion for the class description. For instance, it can be defined that cars (identified in a low image level) do always have to be parts of roads or parking lots (identified on a higher image level or at the same image level).

Figure 19: The system of image levels with image segments of different sizes. Each image object knows its neighboring segments, sub-, and superobjects (Definiens 2007, p. 27).

7.1 The theory of remote sensing-based land-use classifications

105

After the segmentation of the image is completed the supervised classification of the newly created objects can be performed on one or more image object levels. Therefore, a class hierarchy containing all informational classes has to be developed. The classes can then be described and characterized by selecting samples or by creating knowledge based fuzzy sets. Fuzzy sets are mathematical functions. The mathematical functions refer to object features, such as layer values, shape characteristics, information from thematic layers, or relations to other image objects. They describe the membership of an object to one of the defined classes based on the object features (membership functions) (Baatz et al. 2004). That means that they determine a range of values for a certain object feature that is characteristic for the respective class. For example, roads do always have a minimum ratio of width and length (as one object feature) that is larger than 3. In addition, their grey values in the near infrared band always range between 5 and 50. To describe these characteristics of a class, two membership functions are generated that refer to the object features “Ratio width/length” and “Band near infrared” and that mathematically express the desired value range. Multi-dimensional membership functions referring to all spectral bands are automatically generated by the software during the sample selection process. For each class description, different membership functions can be combined using logical operators such as ‘and’ and ‘or’. For the assignment of segments to classes, fuzzy, instead of crisp values are used. The combination of fuzzy rules is also called fuzzy rule base (Baatz et al. 2004). The concept of fuzzy logic assigns values of possible membership to all available classes to each image segment. Thus, the binary values of membership or non-membership are replaced by continuous values between 0 and 1. It is important to distinguish between the possibility and the probability of membership. Fuzzy classification defines the possibility of membership to several classes, which means that the added-up value can be larger than 1 in comparison to the probability that has always the value 1 in sum. For instance, a segment can have a possible membership value of 0.9 for Class A and 0.85 for Class B. The degree of possible membership depends on the degree to which the class characteristics are met by each segment. The segment will be assigned to the class with the highest possibility of membership. The higher the value of possible membership of a segment to a class and the larger the difference to the second highest value, the more stable the class assignment (Baatz et al. 2004). The detailed classification process for this application is described in Section 7.3. Figure A43 gives a brief overview about the image segmentation and analysis procedures and the potential outcomes.

106

7 Remote sensing data analysis

7.1.2.1 Advantages The most important advantage of this approach is the creation of meaningful objects with semantic or context information. Buildings, green spaces or a river for instance will be treated as objects and not just as a group of adjacent pixels. They have a certain shape, size, and context information that can be used to further describe them. Blaschke (2010) compiled numerous case studies reporting improved classification results using an OOA. The position of the object in the environment (e.g. its slope), as well as its relation to other image objects (such as distance to hazard zones) is known. Another advantage of the object-oriented approach is the reduction of the salt and pepper effect because spectral variances are averaged out during the segmentation process. This will potentially lead to an increase of the classification accuracy for very high resolution data and it will also allow a more convenient postprocessing of the classification results in a GIS. The classified polygons can be exported to a GIS and analyzed further. 7.1.2.2 Disadvantages The segmentation is the most important step in the object-oriented classification process. Consequently, if the image objects do not represent the conditions of the ground properly, the classification accuracy will not increase, but decrease. The creation of segments is associated with the potential to significantly improve the classification result and the risk that the objects are not meaningful or that important details get lost. The two main points here are the existential uncertainty (Does an object or a boundary really exist?) and the extensional uncertainty (What is the spatial extent of an object?) (Lucieer & Stein 2002, Molenaar 1998). Thus a validation of the quality of segments by a visual comparison with objects in the image has to be done iteratively by the analyst in order to have a solid starting point for the classification process. As no automatic method for this step yet exists, local knowledge is indispensable (Zhang & Maxwell 2006, Meinel & Neubert 2004, Schiewe 2002). Another disadvantage can be that information contained in a single pixel are aggregated or averaged out during the segmentation process. This might be an advantage in some cases, but also results in a loss of detailed information. Especially in satellite data with a lower spatial resolution, e.g. in the ASTER data used for this study (geometric resolution: 15 m), the issue of mixed pixels already arises without image segmentation. That means that reflectance information from more than one object at the Earth’s surface is contained in one pixel. When further aggregating these data, the information content becomes even more blurry. Even though the selection of the segmentation parameters can limit the degree of averaging out the information during the image segmentation this phenomenon occurs.

7.2 Classification of the ASTER images

107

7.1.2.3 Summary of supervised classification options The two main classification approaches were presented and discussed, whereby both pixel-based and object-oriented approaches follow completely different philosophies. Pixel-based classification is based on the analysis of single pixels values to classify a scene. Object-oriented analysis creates meaningful homogeneous segments before starting the classification. Predetermined segments are selected as samples for each class using the OOA, while samples have to be collected independently from any pre-structuring of the image when using the supervised pixel-based approach. Sufficient knowledge about the setting of the study area is indispensable in both cases. Only spectral values and texture are taken into consideration for the classification using the pixel-based approach. Using the object-oriented approach, the user has the possibility to use spectral, spatial, and shape characteristics as well as contextual information for the class description which goes far beyond the analysis of spectral values of single pixels. The existing limitations concerning the automation of the segmentation process make the OOA time-consuming and less easy transferable. Nevertheless, the obtained classification results are more comprehensive and offer more possibilities for further analyses. Schiewe (without page numbering) stated in 2002 that “traditional multi-spectral classification methods on pixel basis are no longer suited for the evaluation of high-resolution and multisource data from remote sensing”. Following the discussions in the literature there is no special classification mode that is optimal for every application. The choice of the classification method depends on the data type, the available hard- and software, and the purpose of the classification (Richards & Jia 1999). However, a significant increase in the application of object-oriented approaches has taken place during the last decade (Blaschke 2010). The ASTER data were in this study classified using a pixelbased approach as most spectral information content remains in the image and the LULC maps can directly be used for the hydrologic modeling. The main purpose of the classification of the ASTER data is to derive information about the vegetation content and to compare them between the three points in time that were considered for this analysis (2002, 2005, and 2009). The Quickbird data were classified using an OOA to make use of the numerous advantages and tools for the classification of a heterogeneous image with a large amount of detailed information implemented in this approach. 7.2 CLASSIFICATION OF THE ASTER IMAGES As for the hazard analysis and hydrological modeling the current pattern and the changes of LULC were of interest, the pixel-based approach was used for the classification of the ASTER images from 2002, 2005, and 2009. The LULC information of each year was consequently delineated for each picture element. The chosen hydrological model HEC-HMS is likewise grid cell based. The classification results from the LULC classification using the pixel-

108

7 Remote sensing data analysis

based knowledge-based classifier can therewith directly be used to (i) derive hydrological soil groups (compare Chapter 8) and to (ii) form a direct input for the hydrological model. 7.2.1 Classification process The pre-processed ASTER images of 2002, 2005, and 2009 (Table 13) covering the catchment area of San Ramón were also classified using ERDAS 9.2 Expert Classifier (Erdas 2001). To match the capacities of the hydrological model, which distinguishes between limited numbers of LULC types, the delineated LULC classes were chosen according to their hydrological properties. Most relevant was in this case the availability, type, and density of vegetation within the catchment area. A feature to describe the vegetation content in satellite data is the Normalized Differenced Vegetation Index (NDVI). It is the ratio shown by Equation 2: ܰ‫ ܫܸܦ‬ൌ

ே௘௔௥ூ௡௙௥௔௥௘ௗିோ௘ௗ ே௘௔௥ூ௡௙௥௔௥௘ௗାோ௘ௗ

Equation 2

with the Near Infrared Band representing the wavelengths 7001,300 nm and Red representing 630700 nm (Richards & Jia 1999).

The ratio makes use of the “red edge” that describes the significant increase of the reflectance values between the red and near infrared part of the electro-magnetic spectrum in the case of healthy and green vegetation. The absorbance of rays of the red spectral region and their reflectance in the near-infrared spectral regions are determined by the proportion of chlorophyll in the surface coverage. The ratio is higher the higher the chlorophyll content of the vegetation. That means that a high NDVI value represents regions with dense and healthy vegetation coverage that are active in photosynthesis (Richards & Jia 1999). The application of the Expert Classifier requires knowledge about the spectral characteristics of each LULC class without providing tools for an in-depth analysis of spectral characteristics and separability between the different desired LULC classes. Therefore, as a first step, a regular supervised classification using the MLC was performed to obtain better knowledge about the spectral properties of the image and the separability of the different LULC classes. Training areas for most classes were selected based on knowledge obtained during field campaigns as well as visual interpretation of the image. No training areas were selected for the classes “Water courses”, “Dense built-up areas”, and “Intermediate built-up areas”. The water courses were directly taken as a GIS layer from the hydrologic preprocessing; the built-up areas were manually digitized (building mask).

109

7.2 Classification of the ASTER images

Table 15: LULC classes and their content description as delineated from the ASTER data using a pixel-based approach (all unsigned 8-bit data). Class name

Content & Description

Open rock

Barren, rocky surface without soil or vegetation coverage, mostly volcanic rock; Class description: NDVI < 0

Barren land

Barren surface with soil coverage but without or just minimal vegetation coverage; Class description: NDVI 00.08, Green > 40, Red > 30, IR > 30

Sparse vegetation

Barren surface covered by shrubs, grasses or small trees; Class description: NDVI 0.080.23

Woodland

Forested areas with pine and eucalypt trees, little vegetation between trees; Class description: NDVI > 0.23

Water courses

Creeks and rivers; Class description: areas covered by river mask derived during data preprocessing with HEC-GeoHMS

Example

110

7 Remote sensing data analysis

Dense built-up areas

Built-up areas > 70 % impervious, in this case solely water uptake infrastructure (see right column); Class description: areas covered by digitized building mask

Intermediate urban areas

Built-up areas > 30 % impervious; Class description: areas covered by digitized building mask

Undefined

Areas that could not be classified

Seven LULC classes with unimodal histograms that occur in the catchment area of Quebrada San Ramón and one class to capture those pixels that could not be assigned to one of the informational classes (including shadow) were created. They are mainly discriminated through different vegetation and soil coverage (Table 15). Taking this classification based on spectral properties as a basis, the main classification was performed using the ERDAS Expert Classifier to improve the results. First, a file containing the class descriptions in the form of hypotheses containing a certain confidence value was created using the knowledge engineer. Basically, all classes were defined based on the outcome of the purely spectral classification. In addition, GIS and elevation data were used for the class description. A LULC map with an increased accuracy in comparison to the regular supervised classification is then delineated based on the evaluation of all hypotheses in the class descriptions. The class to which the hypothesis with the highest confidence belongs is assigned to each pixel. The Normalized Differenced Vegetation Index (NDVI)  a measure frequently applied to detect different types of vegetation from remotely sensed data  and elevation data were incorporated to distinguish between different types of vegetation. Table 15 compiles all class descriptions in more detail. 7.2.2 Classification results The resulting classifications (Figure A44) form the basis for the delineation of hydrological soil groups (HSG, Section 8.2.3) and the description of LULC changes. The table contained in Figure A44 shows that almost no construction activities took place in the basin between 2002 and 2009. Likewise, the space covered by the river remained stable. With a basin size of 35.8 km² the proportion of built-up area and water courses is relatively small.

111

7.2 Classification of the ASTER images

Changes did occur in the vegetation coverage: the amount of woodland decreased from approximately 1.5 km² in 2002 to about 1 km² in 2009. The most likely explanation is the amount of precipitation recorded during the previous rainy season(s). In dry years (La Niña years), the amount of rainfall is significantly smaller and thus results in a decreasing activity of photosynthesis of the forest. The forest contains less biomass and changes its hydrological relevance (less interception and evapotranspiration). For the same reason, the amount of sparse vegetation decreased even more significantly from 12.5 km² in 2002 to 7.3 km² in 2009. Simultaneously, the amount of barren land increased from 7.8 km² in 2002 to 10 km² in 2009 and the space covered by open rock increased from 13.6 km² to 17.2 km² in the same period (all approximate values). 7.2.3 Accuracy assessment Like the classification of all three ASTER images the accuracy assessment was carried out using Erdas Imagine 9.3. The accuracy assessment is based on a number of randomly selected points that are assigned with a value based on the classification result. The interpreter of the map assigns values based on which the accuracy is calculated. For the three images the accuracy assessment was only carried out for the classes Open rock, Barren land, Sparse vegetation, and Woodland as the three remaining classes were classified based on a digitized mask. The accuracy for these classes is therefore assumed to be 100 %. For the four remaining classes 100 random points were generated. Hereby, the number of random points was proportionate to the total proportion of the respective class in the image (Erdas 1999). Classes were assigned to the randomly selected points based on expert knowledge. Confusion matrices as shown in Table 16 to Table 18 result from the comparison of original and assigned value and represent the accuracy assessment. Table 16: Results from the accuracy assessment of the ASTER data from February 2002. Abbreviations in the first row correspond to the class names in the first column. Class

Op

37 Open rock 1 Barren land 0 Sparse vegetation 0 Woodland 38 Sum 0.97 Producer 0.97 User Overall Accuracy: 0.92 Overall Kappa Statistics: 0.88

Ba

Sp

Wo

Sum

1 17 2 0 18 0.85 0.77

0 4 33 0 37 0.89 0.94

0 0 0 5 5 1.0 1.0

38 22 35 5

The matrices in Table 16 to Table 18 show the correct class assignments in the main diagonal and the false assignments in the remaining parts of the matrix.

112

7 Remote sensing data analysis

Those random points that show wrong assignments in the horizontal direction indicate the user accuracy, i.e. indicate the number of points that are samples of the class in each row but were classified as other classes (error of commission). The numbers in the columns of the matrices represent the producer accuracy, i.e. they indicate the number of random points that were classified as the class in the column even though they were samples of other classes (error of omission). The overall accuracy interprets the values of the main diagonal and is the ratio of the correctly classified random points and the total number of random points. The overall Kappa statistics indicates how the classification result is different from a random assignment of classes to the pixels and refers to the entire matrix. The value in Table 16 signifies that the classification result is by approximately 88 % better than a random assignment (Lillesand & Kiefer 2004). Table 17: Results from the accuracy assessment of the ASTER data from February 2005. Abbreviations in the first row correspond to the class names in the first column. Class

Op

42 Open rock 1 Barren land 0 Sparse vegetation 0 Woodland 43 Sum 0.97 Producer 1.0 User Overall Accuracy: 0.92 Overall Kappa Statistics: 0.88

Ba

Sp

Wo

Sum

0 19 4 0 23 0.82 0.86

0 2 26 0 28 0.93 0.84

0 0 1 5 6 0.83 1.0

42 22 31 5

Table 18: Results from the accuracy assessment of the ASTER data from February 2009. Abbreviations in the first row correspond to the class names in the first column. Class

Op

48 Open rock 0 Barren land 0 Sparse vegetation 0 Woodland 48 Sum 1.0 Producer 1.0 User Overall Accuracy: 0.94 Overall Kappa Statistics: 0.90

Ba

Sp

Wo

Sum

0 26 3 0 29 0.89 0.92

0 2 17 0 19 0.89 0.81

0 0 1 3 4 0.75 1.0

48 28 21 3

The classes Open Rock and Woodland did in all cases yield very high accuracies. The class Open rock can very well be described and separated using the NDVI as it contains no vegetation at all. The class Woodland has a very high vegetation content and is therewith distinct from the other vegetation classes. Errors mainly

7.3 Classification of the Quickbird image

113

occurred between the classes Barren land and Sparse Vegetation. An appropriate threshold value was defined (confirmed by high classification accuracies) but for the interpreter it becomes difficult to clearly distinguish between these two classes with medium vegetation content solely by interpreting the satellite image for the accuracy assessment. Another issue is in this case the occurrence of mixed pixels that partly contain shrubs and barren soil. It can in those cases not intuitively be defined to which class the pixel belongs. Thus, the accuracy and quality of the NDVI-threshold value finally defines the proper class assignment. 7.3 CLASSIFICATION OF THE QUICKBIRD IMAGE 7.3.1 Classification process The classification of the Quickbird data was performed using eCognition Developer 8.0. The software classifies the satellite data by using the supervised Nearest Neighbor classifier based on the selection of samples and the creation of membership functions. Before the classification process was started, meaningful objects were generated from the image. The segmentation parameters comprise a scale factor to determine the average size of the segment, a shape factor, and a compactness factor. If the shape value is set small, the main information for the creation of segments is taken from the spectral content of the image. If the compactness value is set low, the resulting image objects will contain a very irregular shape. If this value is set high, the resulting image objects approach a rectangular shape. The process of finding proper parameters as an input for the segmentation cycle usually requires some trial-and-error runs. The segmentation parameters for the present case are displayed in Table 19. The main classification was carried out on image level 1. Table 19: Segmentation parameters. Parameter Scale Factor Shape Compactness

Level 1 30 0.1 0.3

Parallel to the segmentation a classification scheme with class descriptions was developed. This was a new classification scheme matching the different setting of the urban area. The scheme contains 15 urban LULC classes matching the very high geometric resolution of the satellite data (0.6 m) was developed to capture all relevant LULC types in the study area (Table 20). Once the class hierarchy is established, samples can be selected and fuzzy sets can be developed. The class description is in this case based on the spectral characteristics of an object, on its shape, its elevation, information about their relations to neighbor objects, values obtained from additional GIS-Layer, and texture information. One of the most

114

7 Remote sensing data analysis

relevant classes in this study, the class “Buildings” for example only occur in areas that are defined as census units (to exclude open rock or barren land in the higher lying non-populated areas of the Andean mountains) and have a low NDVI. The class comprises five different roof types which are not separately listed in Table 20 but in Table A40 in the appendix where all detailed class descriptions and the classification scheme are shown. After running the first classification, a second run can be done using classrelated features. That means that object features will be considered that depend on the classification result of the other image segments. That is of value to assign classes to very small objects that remained unclassified or were wrongly classified in a first run. For example, some segments were classified as certain roof types but were surrounded by barren land by more than 90 %. In this case a rule set was developed that assigned the class “Barren Land” to the respective objects to correct for the false initial assignment. Other very small wrongly classified or nonclassified segments were treated alike (compare complete rule set shown in Table A40). Table 20: LULC classes and their content description as delineated using the object-oriented approach. Class name

Content

Buildings

Built-up areas predominantly comprising buildings of different usage and construction material (5 roofs types groups merged to one), more than 70 % impervious

Streets

Including all paved streets in the study area

Swimming pools

Artificially created swimming pools, both private and public usage

Example

7.3 Classification of the Quickbird image Urban unpaved

Unpaved areas in urban part of the city (dirt roads, open spaces with very little or no vegetation coverage)

Unpaved park

Walking paths in green spaces that are covered with gravel and sand

Sand

Purely sandy patches on golf courses

Trees

Natural areas with groups of pine and eucalypt trees and little vegetation in between trees as well as singlestanding trees in the urban part of the study area, NDVI 0.551

Grassland

Areas covered by healthy, in most cases irrigated grassland with different usage (e.g. parks, sport facilities), NDVI 0.50.7 and 0.740.85 for stadiums

Dry vegetation

Areas covered by smaller natural shrubs or grasses, areas outside the urban built-up body or urban green spaces with insufficient irrigation, NDVI 0.250.5

115

116

7 Remote sensing data analysis

Barren land

Barren surface with soil coverage but without or just minimal vegetation coverage, areas outside the urban built-up body

Bushland

Areas covered by larger shrubs or small trees, areas outside the urban built-up body

Agricultural areas

Areas with agricultural usage, predominantly vineyards, different levels of cultivation

Water bodies

Natural water courses, such as creeks and rivers and (irrigation) ponds

Open rock

Barren, rocky surface without soil or vegetation coverage, mostly volcanic rock

Undefined

Areas that could not be classified, including shadow

The required rule set for the LULC classification (Table A40) was developed for a subset of the study area, has then been tested and adapted in other parts of the image, and could finally be applied to classify the entire study area in 5,000 by 5,000 pixel patches (3 * 3 km, appropriate size to ensure acceptable computing times of less than 10 minutes with stable performance).

7.3 Classification of the Quickbird image

117

7.3.2 Classification results The result of the classification of the satellite data is a map showing the LULC at a spatial resolution of 0.6 m for the municipalities of La Reina and Peñalolén (Figure A45). The map shows the classes as described in Table 20 and in addition to that the single roof types that could be distinguished during the classification. The roof types were for the further processing grouped to one class “Buildings”. It can be seen that the proportion of green spaces is higher in the northern part of the image and lower in the central-western and south-western part. The agricultural areas in the south-western part of the image are vineyards. It can in the eastern part be seen where approximately the cota mil is located, i.e. where construction activities stop and where the natural green spaces begin. Especially in the south-eastern part of the study area, the gap between the urban built-up body and the cota mil remains large, theoretically providing space for further urban expansion. How the LULC information is used to feed the indicators for the determination of elements at risk and their vulnerability describes Section 9.1. To allow for a better evaluation of the provided LULC maps, the following section provides a description of the accuracy assessment that was carried out prior to the further processing of the data. 7.3.3 Accuracy assessment The accuracy assessment was carried out using a TTA mask (Test and Training Areas). That means that sample segments were selected manually and as randomly as possible for each LULC class from the segmented but unclassified image. The class assignments of these segments were then compared with the class assignment during the semi-automatic classification process. As the number of classes in the image was initially very high (compare Figure A45) and as the classes were strongly aggregated for the further processing, the classes were already pre-grouped before carrying out the accuracy assessment. That means that all roof types were grouped to one class “Buildings”. Trees and Bushland were grouped into one class “Trees”. Urban unpaved areas, Unpaved parks, Sand, Open rock, and Barren land were aggregated to one group “Barren land”. All remaining classes were left as they were. The comparison of the class assignments delivers the results as shown in Table 21. The classes Barren land, Streets, Trees, Grassland, Dry vegetation, and Buildings are all classified with producer accuracies higher than 0.85 which is satisfying. The class Water bodies that constitutes only small parts of the overall image shows very low accuracies. Water bodies were very difficult to derive as they are spectrally very similar to streets without showing distinctive shape parameters. A large part also remained unclassified. This class is for this specific study area and for the further use of the data of minimal importance. The goal of this classification was to delineate the urban built-up area and green spaces, which was very successful. Streets could be classified very well using shape characteristics, amongst others (compare Table A40). Buildings could likewise be classified with a high accuracy. Most use-

118

7 Remote sensing data analysis

ful were here the green band, the GIS layer containing the outlines of the building blocks, and the NDVI as initial criteria to extract built-up areas. Further fuzzy rules were then developed to distinguish between different roof types. However, these roof types were at this stage not used for the further analysis as no sufficient data for a detailed validation of the results were available. While all of the selected TTAs for Agricultural land were classified correctly (high user accuracy), some of the TTAs selected for Trees and Grassland got likewise classified as agriculturally used land. This gap results from a completely unambiguous description of the class Agricultural areas. The class descriptions for all other types of green spaces were not as unambiguous, thus did not exclude that the agricultural class was assigned to any other type of green space. This moderate producer accuracy of the class Agricultural area does however not significantly influence the further computations as they are in the next steps aggregated to “Green spaces”, i.e. merged with the classes Trees (consisting of Trees and Bushland), Grassland, and Dry vegetation. The overall accuracy of 0.85 is satisfying and sufficient for an unhesitating further use of the results. Table 21: Results from the accuracy assessment of the Quickbird data. Abbreviations in the first row correspond to the class names in the first column. Class

Ba

3,111 Barren land 0 Water bodies 0 Streets 0 Trees 0 Grassland Swimmingpools 0 0 Dry vegetation 0 Agriculture 0 Buildings 0 Unclassified 3,111 Sum 1 Producer 0.88 User Overall Accuracy: 0.85

Wa

St

Tr

Gr

Sw

Dr

Ag

Bu

Sum

0 495 277 0 0 0 0 0 59 578 1,409 0.35 1

0 0 2,626 0 0 0 0 0 151 0 2,777 0.95 0.90

0 0 0 1,800 116 0 0 0 0 0 1,916 0.94 0.72

0 0 0 25 1,734 0 0 0 0 0 1,759 0.99 0.77

0 0 0 12 0 470 0 0 67 39 588 0.80 1

318 0 0 0 0 0 3,646 0 0 0 3,964 0.92 1

0 0 0 678 398 0 0 2,166 0 0 3,242 0.67 1

89 0 0 0 0 0 0 0 1,336 97 1,522 0.88 0.83

3,518 495 3,028 2,515 2,248 470 3,646 2,166 1,613 714

8 HYDROLOGIC MODELING The goal of this Section is to explore the influence of LULC changes on the runoff behavior of the San Ramón watershed using a hydrological model. Section 8.1 first provides some theoretical background about the impact of land-use changes on the hydrological processes and the hydrological modeling process. The selection of methods is explained based on that. Section 8.2 gives detailed information on how the terrain and hydrological processing were carried out using HECGeoHMS. Included in this section is the delineation of the Hydrological Soil Groups (HSG) and the model setup. The modeling process including model description, parameterization, calibration, validation, and a sensitivity analysis is described in the subsequent Section 8.3. Section 8.4 contains the modeling of alternative LULC scenarios of the catchment area. Chapter 8 concludes with the establishment of a relationship between the runoff estimations and the existing flood hazard maps (Section 9.2.1). 8.1 THEORETICAL BACKGROUND AND SELECTION OF METHODS The hydrological model HEC-HMS (Hydrologic Engineering Center  Hydrologic Modeling System) was developed by the hydrologic engineers of the U.S. Army Corps of Engineers (USACE). It has mainly been applied for urban flooding studies, flood-frequency studies, flood-loss reduction studies, flood-warning system planning studies, reservoir design studies, and environmental studies in the USA (Ford et al. 2008, p. vii). In this study it is applied to simulate the relationship between precipitation and runoff in the catchment of the San Ramón River. As the future can only be predicted if past is understood, the hydrological model is applied to reconstruct past events and to therewith enable a better understanding of the processes and interrelations of different components in the specific catchment and for the occurrence of floods. Regarding the influencing factors, different LULC scenarios are developed to simulate the impacts of LULC changes on the flood hazard over time. 8.1.1 Relevant processes with respect to LULC changes In the course of urbanization some of the most relevant effects of changes in LULC with respect to hydrology are changes in the amount of rainfall stored on and in biomass (interception storage, litter storage, and root zone storage) and changes in the infiltration capacities of the soil due to sealing of the surface (Bronstert et al. 2002, p. 517). To obtain valid modeling results and insight in the changing response of the watershed after LULC changes, these relevant processes need to be represented by

120

8 Hydrologic modeling

the model (compare further Bronstert et al. 2002). How HEC-HMS takes these parameters into consideration is outlined from Section 8.1.3 onwards after the following section provides a more general overview about model types and the selection of the appropriate model for this study. 8.1.2 Theoretical background of hydrological models Hydrological models are a simplified representation of a real catchment system and are applied to simulate hydrological processes. They consist of a hydrological core (e.g. definition of variables, process description) and a technological shell (e.g. programming, user interface) (Refsgaard 1996, p. 17). As shown in Figure 20, different types of hydrological simulation models can be distinguished.

Figure 20: Overview about different types of hydrological simulation models. Modified after Reefsgard (1996) and Piepho (2003).

The first distinction refers to the main theoretic philosophy, the degree of causality, and the process representation. It is distinguished between deterministic models that are based on exactly defined (presumably known) conditions and rules as well as stochastic models that include one or more random variables. It is assumed that the system contains unknown factors, thus the algorithms for the description of the processes are based on statistical analyses of the measured input data. While a deterministic model always delivers the same results when using the same input variables, the stochastic model most likely yields variable or differing results. The stochastic models can further be subdivided into probabilistic models that represent single hydrologic parameters, e.g. extreme events only based on a statistical analysis of their occurrence without considering dependencies from other parame-

8.1 Theoretical background and selection of methods

121

ters (Carrera 2005, Dyck & Peschke 1995). The second group of stochastic models constists of time series that deliver extrapolations of variables. A famous example is the group of autoregression models. Table 22 compares the advantages and disadvantages of deterministic and stochastic models. Table 22: Comparison of deterministic and stochastic models after Dyck & Peschke (1995) and Carrera (2005). Deterministic model

Stochastic model

Advantages Causality between variables and parameters and complexity of system considered

Few input data needed Fast results

Disadvantages Model parameters describing physical properties and initial and boundary state variables unknown; Error propagation resulting from unknown data

All hydrological parameters are random in time and space Purely statistical relationship between considered variables

Process-describing equations not always appropriate for envisaged scale and representation of dominant process; Not all important variables & parameters represented

Deterministic models can be further classified according to various aspects. With regard to the spatial representation of the catchment and its characteristics it can be distinguished between lumped models, distributed models, and semi-distributed models. Lumped models treat the basin as one single unit, e.g. assume one average land-use type. They are also referred to as mean value models. Distributed models in contrast maintain the spatial heterogeneity of a basin, i.e. in the case of land use they assign different land-use patterns within the basin. Semi-distributed models are spatially explicit but do not have a common spatial unit (e.g. have differently sized sub-watersheds) and consider data that are artificially distributed over space even though only one measurement exists (e.g. precipitation data). Deterministic models can furthermore be classified into physically-based white box models where all processes and mass flows are described using mathematical equations and all input parameters can be measured. The second subcategory, conceptual models, are also physically-based but additionally include semi-empirical estimated parameters (gray box models). Solely empirical data are used as input for the third type of deterministic models, the black box empirical models, where measured input and output data are analyzed and brought into statistical relation without considering the physical processes in the catchment.

122

8 Hydrologic modeling

While white box models are mostly distributed models and black box models can only be lumped models, conceptual models can be both and are a classic example for precipitation-runoff models (Refsgaard 1996). With respect to time it can be distinguished between continuous models and event-based models. Continuous models simulate the hydrological system behavior on a long-term base. Event-based models in turn are often used to simulate single flood events whereby parameters such as the initial loss and the continuous loss during a storm event are estimated. In the case of continuous models, these parameters would be delineated from long-term system observations (Boughton & Droop 2003). 8.1.3 Selection of model and methods 8.1.3.1 Selection of the model and general approaches Processes and their interrelations, especially land-use changes and resulting effects on the runoff are analyzed in the scope of this research. Thus, a deterministic model with stochastic components (e.g. including the certain probability of an extreme event) was chosen. Because this study does focus on the analysis of single flood events, the event-based approach is used. A continuous simulation is also considered less appropriate in the present case as precipitation is few in the study area, does not yield significant changes in runoff and therewith limits the benefits of effective and permanent system monitoring and analysis. This research develops a methodology for a megaurban environment. As often only little or no financial means for additional software investments are available and data availability is limited, a model that is free of charge and works with a limited amount of input data was searched for. The reasoning is that the proposed methodology should be kept transferable and repeatable. Thus, the local conditions in the study area need to be considered when choosing the analysis tools. After literature research and discussions with experts from universities in the study area the Hydrologic Modeling System of the USACE Hydrologic Engineering Center (HEC-HMS) was selected to be an appropriate tool to reach the objective of this research. HEC-HMS is a deterministic, conceptual basin model. Its capacities comprise: -

Simulation of spatial and temporal patterns for precipitation and evaporation within a catchment Simulation of infiltration and runoff volumes Simulation of direct surface runoff and interflow Simulation of baseflow Simulation of channel flow, including the prediction of time series of downstream flow, stage, or velocity, given upstream hydrographs (Ford et al. 2008, p. 4).

Different methods exist for the calculation of each simulation. A detailed overview is given on p. 5 of the HEC-HMS Applications Guide (Ford et al. 2008). The

8.1 Theoretical background and selection of methods

123

following subsection now explains the selection of the specific methods that are required for the modeling with HEC-HMS. From the explanation it also becomes obvious where each method can be located in the scheme of Figure 20. 8.1.3.2 Selection of the loss method First, it has to be defined how the amount of excess precipitation (runoff) is determined. Again, it has to be referred to the main goal of this research: The analysis of the influence of LULC changes on runoff generation. As this clearly requires an approach with a spatial reference that is as explicit as possible, the distributed or semi-distributed modeling approach was found to be the only choice. HEC-HMS is a lumped basin model but has one component that supports a distributed approach. The empirical and distributed gridded Soil Conservation Service (SCS) curve number (CN) loss method that works on a grid basis was selected as the loss method for surface runoff estimation: All methods considering evapotranspiration values (deficit and constant loss method, gridded deficit constant loss, gridded soil moisture accounting were neglected, as evapotranspiration does not play a significant role for the modeling of short-term heavy precipitation events (Scharffenberg & Fleming 2008, p. 33). The green and ampt loss method requires information about the active soil porosity which is not available and the exponential loss method requires field experiments which have not been feasible during that research. Other methods such as Smith Parlange loss method, soil moisture accounting loss method require input information with a level of detail or parameters that were not available. A very simple method for runoff estimation is the initial and constant loss method, which distinguishes between one initial loss value directly after the precipitation event and one following constant loss factor during the precipitation event after the initial loss is satisfied. Even though the impervious proportion of the basin can be explicitly specified, all other values are averaged out for each the subbasin. The non-gridded SCS CN loss method has the same disadvantage of only assuming average CN values for each subbasin. As data from the satellite imagery have a higher spatial resolution than that, the gridded SCS CN method was chosen. The CN method was developed by the Natural Resources Conservation Service (NRCS, formerly known as Soil Conservation Service SCS) and is a quantitative descriptor of the land cover/soil complex. This information can be derived from remote sensing data (Tekeli et al. 2007, Slack & Welch 1980). The CN values take the land cover type, treatment (especially relevant for agricultural studies), the hydrologic conditions, antecedent runoff conditions and the Hydrological Soil Groups (HSG) into consideration (USDA 1986). While only few input data are required, one of the limitations is that it does not consider time and rainfall duration or intensity (USDA 1986).

124

8 Hydrologic modeling

The rainfall-runoff relation using the standardized ratio that is used for runoff estimation can be described as follows: ܳൌ

ሺ௉ି଴Ǥଶௌሻ; ௉ା଴Ǥ଼ௌ

Equation 3

with Q as the total runoff or excess precipitation [mm], P as the total amount of precipitation [mm] and S as the maximum potential retention value after runoff begins. The spatial reference is a grid cell (Mockus 1972, Feldmann 2000). The values 0.2 and 0.8 result from the assumption that 20 % of the precipitation is used for initial storage (Johnson & Miller 1997).

S can be calculated as: ܵൌ

ଵ଴଴଴ ஼ே

െ ͳͲ

Equation 4

with CN as a quantitative descriptor of hydrologic soil group, cover type, treatment, hydrologic condition, and antecendent runoff condition (USDA 1986).

The derivation of the equations is described in more detail in Atkinson (2001). 8.1.3.3 Selection of the transform method As a second step it has to be defined how the resulting excess precipitation is transformed into runoff. Using an empirical method implies referring to the unit hydrograph (UH) models. They assume a linear runoff generation in a basin and a causal relation between excess precipitation and runoff. The UH reflects the logical relation that runoff results from excess precipitation in a certain unit. The total runoff is then calculated by adding up the number of units and the respective runoff that was produced there. Physical processes are not considered. The only available option for a distributed modeling approach in HEC-HMS is the empirical ModClark method (Kull & Feldman 1998). ModClark is a method based on the Clark hydrograph method that considers translation (movement of the excess rainfall from origin to outlet) as well as attenuation (storage effects of excess rainfall) processes (Feldmann 2000, p. 60). The storage over time is expressed in a linear reservoir model by: ௗௌ ௗ௧

ൌ ‫ܫ‬௧  െ  ܱ௧

Equation 5

dS/dt is thereby the time rate of change of water storage at time t, It is the average inflow to, and Ot the average outflow from the storage at the time t (Feldmann 2000).

Out of seven options implemented in the model, the ModClark transform method was identified to be the best suited transform method as it is the only one that supports distributed modeling. Besides a number of other empirical methods, HEC-HMS also supports the conceptual kinematic wave method which was developed for urban areas but cannot be implemented when using a distributed modeling approach.

8.1 Theoretical background and selection of methods

125

8.1.3.4 Selection of the baseflow method The baseflow is also known as the “fair-weather” runoff in a channel (Feldmann 2000) and refers to the runoff during storms resulting from stored water and/or slow subsurface drainage of water into the channels of a basin. There are several approaches applicable for event-based short-term and for long-term continuous modeling. The two mathematical models Constant Monthly Baseflow and Bounded Recession Baseflow are used for long-term simulations and are thus not suited for the modeling of single events. The Linear Reservoir Method uses a linear reservoir to model the recession baseflow after a storm event (Scharffenberg & Fleming 2008) and requires information about groundwater storage which was not available for the present study in a sufficient temporal resolution. The Nonlinear Boussinesq Baseflow simulates the behavior of the watershed after a storm event but requires soil conductivity and porosity data which are both not available. The Recession Baseflow model can be used for both short- and long-term modeling. It requires the initial discharge before the storm event and a recession constant k that is defined as the current baseflow divided by the baseflow one day earlier (Scharffenberg & Fleming 2008, Feldmann 2000). As this information is available, this method was chosen for the present application. 8.1.3.5 Selection of the routing method The routing method defines how the water flows through the stream elements (reaches), i.e. how they are connected. If no routing method is selected, the incoming flow will automatically and instantaneously be transferred to the next subbasin. One of the models is the Kinematic Wave method that approximates the full steady open flow (Scharffenberg & Fleming 2008). It is best suited for nonurban settings with significant relief. Besides the average slope and the Manning coefficient of the area along the reach it requires data about channel width, height, and geometry which are not available for the major part of the basin. The Lag Routing method is suitable for short stream segments as attenuation and diffusion processes are not considered (Scharffenberg & Fleming 2008). It solely requires one parameter: The lag time that defines the time between the points when the inflow enters and then again leaves the reach. The Modified Puls Routing in turn considers attenuation by involving a storage-discharge function. Attenuation is also approximated by the Muskingum Routing method. It assumes a linear water surface. Its parameters are the travel time through the reach (Muskingum K), which is calculated during the data processing using HEC-GeoHMS, and the Muskingum X parameter that describes the attenuation by referring to the nonlevel water surface. The weighting between inflow and outflow, which is reflected in the Muskingum X coefficient, can take values between 0.0 (maximum attenuation, high weight of inflow) and 0.5 (no attenuation). The Muskingum-Cunge Routing method is an extended and more precise version of the previous method and requires information about channel shape, slope, and roughness along the river sides as well as data about channel cross sections. The last available routing

126

8 Hydrologic modeling

method is the Straddle Stagger Routing, which is a simple empirical method that again involves lag parameters to describe the delay and travel time of the flood wave through the reach. For this application, the Muskingum Routing method was selected because it is the most accurate method that can be used with regard to data availability. 8.1.3.6 Selection of the loss/gain method The loss and gain methods describe losses from the channel to the groundwater, gains from the groundwater or bi-directional movements. If no method is selected, routing is carried out without the consideration of losses and gains to/from groundwater. The Constant Loss/Gain method can only capture losses at a fix rate from the channel. The Percolation Loss/Gain method assumes a constant infiltration and can only be used with the Modified Puls Routing and the MuskingumCunge method. As a result of data constraints and as infiltration along the shoreline does not play a major role in this context, no Loss/Gain method was selected. 8.1.3.7 Description of the meteorological data With the selection of ModClark as transform method, gridded precipitation data need to be used. The preprocessing of the meteorological data, i.e. their selection and transformation in grid format are described in Section 6.2. 8.2 DATA PREPARATION WITH HEC-GEOHMS Before HEC-HMS can be applied, a range of input data need to be generated as outlined in the following sections. The main steps of the data preprocessing operations and the subsequent modeling with HEC-HMS are illustrated in Figure 21. It provides an overview about the main steps: Data preprocessing, hydrologic modeling, and scenario analysis. The modeling extension HEC-GeoHMS is a geospatial hydrology toolkit that can be used as an extension in ArcView and ArcGIS. In combination with the ESRI Spatial Analyst extension it supports the development and derivation of hydrological and terrain data that can directly be used as input for HEC-HMS. Its functions and features comprise the visualization and storage of data, the terrain, and hydrological pre-processing of the digital terrain data. Resulting products from the application of HEC-GeoHMS for data preprocessing are: -

a background map file containing the position of streams and the subbasin boundaries (see Figure 22), a basin model containing the position of all hydrologic elements (Table 26), their alignment and connectivity and empty fields in which hydrologic parameters can be added,

8.2 Data preparation with HEC-GeoHMS

-

127

tables containing physical characteristics of streams and watersheds, a grid-cell parameter file in case of using a distributed approach, a distributed basin model in case of using a distributed approach (USACE 2003, p. 2).

The pre-processing using HEC-GeoHMS can thereby be grouped into two major steps: (i) Terrain pre-processing (Section 8.2.1) and (ii) hydrologic processing (Section 8.2.2) (USACE 2003).

Figure 21: Process flow of the hydrologic modeling process with HEC-HMS and pre-processing steps with HEC-GeoHMS.

128

8 Hydrologic modeling

8.2.1 Terrain pre-processing The terrain pre-processing for the catchment area above the gaging station  i.e. for the area that is being modeled  is based on the DEM delineated from the 50 m contour lines %and for the urban area of La Reina on the 5 m and 2.5 m contour lines. The following steps were performed using HEC-GeoHMS to prepare the hydrological modeling process: -

filling of sinks, calculation of flow direction, calculation of flow accumulation, definition of streams in the basin, segmentation of streams, delineation of watershed.

Table A37 shows a listing of all performed terrain pre-processing steps with input and output data as well as parameters that were used. The final step of the terrain pre-processing is the delineation of the Quebrada San Ramón watershed based on the location of the gaging station (station Quebrada Ramón en Recinto EMOS, BNA Code: 057300086, UTM: 359226 Easting, 6299667 Northing, Zone 19S). The catchment size for that gaging station is 35.77 km². Since a distributed modeling approach is used for this analysis, the elevation, LULC, and soil/geology data are required in grid format. The application of gridded data is based on the spatial reference of the standard hydrological grid (SHG). This is a pre-defined grid by the USACE for the conterminous United States with various cell sizes. In order to apply HEC-HMS outside this area all projection information associated with the available raster data (elevation data, land-use data, etc.) have to be deleted and replaced by Albers projection  which is also used for the SHG. The respective Albers projection file has to have the content as shown in Table A38 in the appendix (Universidad Politécnica de Catalunya). 8.2.2 Hydrologic processing Basin information such as river length and slope, flow path segments with respective length and slope, longest flow path (12,444.5 m), and the basin centeroid were calculated during the hydrologic processing. Based on that, the time of concentration that is later needed as a basin descriptor is calculated for each subbasin. That is the “time of flow from the most hydraulically remote point in the watershed to the watershed outlet” (Feldmann 2000, p. 58). This travel time calculated in HEC-GeoHMS represents a weighted estimate of overland and in-channel flow. HEC-GeoHMS automatically calculates the travel time according to the TR55 method (USDA 1986) as a sum of sheet flow, shallow concentrated flow, and channel flow in a separate MS Excel-file (standard name: Tc.xls). A number of required values are automatically calculated during the terrain and hydrologic

8.2 Data preparation with HEC-GeoHMS

129

analysis. That comprises flow length for all three flow types, land slope, watercourse slope, channel slope, average velocity, hydraulic radius, and the flow times for each of the three flow parts. In addition, some more parameters need to be specified by the user (Table 23). The method requires the usage of English units. Table 23: Parameters used for the hydrologic pre-processing of the San Ramón catchment using HEC-GeoHMS. Parameter

Value

2-year 24-hour average rainfall Wetted perimeter

0.071 in 1215 ft, depending on river width 12 ft², for all subbasins

Cross-sectional area Manning coefficient of river bed Manning coefficient for upper 300 ft in rocky subbasins 300 ft in vegetated subbasins Surface description

0.03 0.035 0.25 unpaved

The sheet flow path starts at the outer basin limit and by default ends after 300 ft along the longest flowpath. The shallow flow begins where the sheet flow ends and ends where the longest theoretical flowpath first intersects the channel (USACE 2003). Required information therefore are flow path lengths as well as land, watercourse, and channel slopes which were calculated during the hydrologic processing. Another parameter is the two-year 24-hour average rainfall which is in the present case 0.071 inches (daily average over two years from the Cerro Calán station from September 1, 2000 to August 31, 2002). The following parameters have additionally been defined for the basin area: Average Manning’s roughness coefficient for the sheet and channel flow, surface description (paved/unpaved) for the shallow flow as well as the cross-sectional flow area (ft²), and the wetted perimeter (ft) of the channel for each subbasin. The wetted perimeter (cross-sectional length of river bed that is covered with water) during a rainfall event was estimated to be between 12 and 15 ft from field surveys and aerial photographs. The cross-sectional flow area respectively is estimated to be 12 ft² in average. For this study, these values can only be estimations but testing the two parameters showed that they affect the resulting travel times only marginally when varying the values in a range of 23 ft and ft², respectively. The Manning coefficient for the river bed that is mostly covered by gravel was set to 0.03 (Arcement & Schneider 1989). The coefficient for the upper 300 rocky feet was set to 0.035 for rocky subbasins and to 0.25 for vegetated subbasins. Travel times from 0.48 to 2.05 hours per subbasin result from that. The travel times for each subbasin are shown in Table A41 in the appendix. These data are the topographic and hydrologic foundation for the hydrological modeling process with HEC-HMS. The following sections line out additional pre-

130

8 Hydrologic modeling

processing operations that need to be performed when using a distributed modeling approach with the gridded SCS CN method. 8.2.3 Delineation of Hydrological Soil Groups (HSG) Figure A34 in the appendix shows the workflow described in the following sections. The LULC classification results and possible LULC scenarios are one of the main input data sets for the hydrological model. In order to estimate the surface runoff based on a certain LULC pattern in the basin, the gridded CN method is applied (USACE 2003). Therefore, a so-called CN-file  a map with hydrologically similar classes  has to be created (Merwade 2008), compare Section 8.1.3. The hydrological classes are defined through the concept of Hydrological Soil Groups (HSG) and therewith based on the minimum infiltration rate of the barren soil after prolonged wetting (Mockus et al. 2007, USDA 1986). Four different types of HSGs exist, whereby the proportion of each of the four HSGs has to be defined for each land-use and soil class (Mockus et al. 2007). HSG Type A:  soils have low runoff potential and high infiltration rates even when thoroughly wetted. They consist of deep, well to excessively drained sand or gravel, and have a high rate of water transmission (greater than 0.30 in/hr). HSG Type B:  soils have moderate infiltration rates when thoroughly wetted and consist chiefly of moderately deep to deep, moderately well to well drained soils with moderately fine to moderately coarse textures. These soils have a moderate rate of water transmission (0.150.30 in/hr). HSG Type C:  soils have low infiltration rates when thoroughly wetted and consist chiefly of soils with a layer that impedes downward movement of water and soils with moderately fine to fine texture. These soils have a low rate of water transmission (0.050.15 in/hr). HSG Type D:  soils have high runoff potential. They have very low infiltration rates when thoroughly wetted and consist of clay soils with a high swelling potential, soils with a permanent high water table, soils with a claypan or clay layer at or near the surface, and shallow soils over nearly impervious material. These soils have a very low rate of water transmission (00.05 in/hr) (Mockus et al. 2007). Soil data are not available for the study area but a geologic map was compiled by Stumpf (2009), compare Figure 14. Based on the description of the geologic formations, the proportions of each hydrological soil group were estimated. A very low infiltration capacity, i.e. high runoff potential (HSG D = 90) is assigned to the class Rigid Bedrock Abanico Formation. Faults where rainfall is captured and stored are accounted for by setting the value for HSG A (high infiltration potential) to 10. Active and old alluvial deposits contain gravel and sand and in the eastern part of the basin they also contain sand, silt, and clay (Stumpf 2009). They are therewith highly to moderately permeable. The process of consolidation of the

8.2 Data preparation with HEC-GeoHMS

131

older deposits was accounted for by setting the HSG A value to 30 instead of 70 for active deposits. The HSG B value was defined accordingly to reach 100 in sum (compare Table 24). The values for HSG A and B of the young alluvial deposits were set to 50 as this formation represents an intermediate stage of the two classes described beforehand. Old landslide deposits show a high proportion of sandy gravels with volcanic ashes and do therewith get a higher HSG A value of 70 and a complementary HSG B value of 30. The quaternary deposits in the Santiago basin show according to Stumpf (2009) a high proportion of alluvial, fluvial, lacrustine, and evaporitic deposits. The deformed quaternary deposits in the western part of the basin are therefore described with HSG A and B values of 50. The assignments of all HSG values are also displayed in Table 24. The HSG values for each LULC class were taken from literature (USDA 1986). Table 25 describes the contents of each of the seven LULC classes derived from the ASTER satellite data with their hydrological properties expressed through the HSGs. Based on these data, a CN theme in raster format can automatically be created using HEC-GeoHMS. Therefore, the attribute table of the geological map needs to be extended by four columns that refer to the percentage of the four HSGs as shown in Table 24. The land-use data from all three time steps need to be brought to vector format. Next, a look-up table (LUT) in dbf-format containing the proportionate number for each HSG for all LULC types needs to be created based on Table 25. The land-use and geology information are then unified to have all information in one data base. Based on the percentage of each HSG from the geology map and the HSG numbers of each LULC type, an average CN is calculated for each unique combination of one land-use and one geology type (Figure A46). The calculation is based on the following equation that also shows the relation between HSGs and CNs: CN=(PctA)(CNA) + (PctB)(CNB) + (PctC)(CNC) + (PctD)(CND)l Equation 6 with CN as the curve number for each unique combination of one land-use and one geology type, PctA as the percentage of HSG A from the geology map (Table 24), and CNA as the curve number for HSG A for each land-use class (here from the look-up table based on literature research). Variables for B, C, D accordingly (Atkinson 2001).

The average CN values are finally calculated for each subbasin (Merwade 2008). 8.2.4 Creation of a gridfile matching the SHG A grid spatially matching the SHG has to be defined as a spatial reference for the further grid-based analysis using HEC-HMS. Therefore, a ModClark grid (Kull & Feldman 1998) was created and populated with the CN values. As mentioned before, the Albers projection file has to be used instead of previously existing projection files in order to apply the model outside the US and to simulate a wrong spatial reference (compare Section 8.2.1). The cell size can be chosen in predefined steps reaching from 10 m to 2,000 m. A cell size of 50 m was selected for

132

8 Hydrologic modeling

the present analysis (Figure A46) to generalize land-use information as little as possible and to still enable acceptable computing times. The grid represents the subbasins with grid cells for the distributed modeling approach (USACE 2003). Table 24: Geologic formations and hydrological soil groups. The columns PctA to PctD show the proportion of the four hydrological soil groups for each type of geologic formation accordingly. Formation

Content

PctA

PctB

PctC

PctD

Rigid bedrock Abanico Formation Old landslide deposits Active and old alluvial deposits Young alluvial deposits Slope sediments Intrusiva Blockschutt

2000 m thick succession of basic to intermediate volcanic rocks

0

0

10

90

Sandy gravels, volcanic ashes, and platy gravels Stratified, moderately consolidated sediments Fluvial, colluvial, alluvial, and landslide deposits Dense, fine sand with clay Volcanic rock Blocks and gravel with finer soil material Alluvial, fluvial, lacustrine, evaporitic deposits

70

30

0

0

70 30 50

30 70 50

0 0 0

0 0 0

0 0 100

0 0 0

100 0 0

0 100 0

50

50

0

0

Quaternary deformed

Table 25: Land-use/land-cover classes of ASTER classification results with their hydrological properties. The columns PctA to PctD show the proportion of the four hydrological soil groups for each LULC type accordingly. Class

Content

PctA

PctB

PctC

PctD

Water bodies Open rock

Water courses, water bodies Barren surface without soil or vegetation coverage Barren surface with soil but without or minimal vegetation coverage Areas covered by shrubs, grasses, and small trees Forested areas with pine and eucalypt trees, little vegetation on soil Built-up areas with more than 70 % imperviousness Built-up areas with more than 30 % imperviousness

100 98

100 98

100 98

100 98

63

77

85

88

55

72

81

86

36

60

73

79

81

88

91

93

57

72

81

86

Barren land Sparse vegetation Woodland Dense built-up areas Intermediate built-up areas

8.2 Data preparation with HEC-GeoHMS

133

8.2.5 HMS Project Setup After calculating the relevant basin and hydrological parameters, the information obtained using HEC-GeoHMS has to be brought in the appropriate format that can be used with HEC-HMS. Therefore, all delineated subbasins, streams, and junctions are labeled and checked for consistence and ambiguity. Legend symbols and coordinates are assigned to all hydrological elements (compare Figure 22 and Table 26). The background map file and grid parameter file storing topological and stream information for the information transformation into HEC-HMS have to be created before the distributed basin model and the HMS project framework are set up, compare (USACE 2003). The information generated using HEC-GeoHMS are then transferred to HEC-HMS.

Figure 22: Basin model of the catchment of Quebrada San Ramón in HEC-HMS as generated through the pre-processing operations using HEC-GeoHMS.

134

8 Hydrologic modeling

8.3 HYDROLOGICAL MODELING USING HEC-HMS 8.3.1 Model description The model HEC-HMS consists of the four main components (i) basin model component, (ii) meteorological model component, (iii) control specifications component, and (iv) input data components, which are listed and characterized in the following. 8.3.1.1 Basin Model Component The function of the basin model component is to describe the physical conditions of the watershed using hydrologic elements (e.g. subbasins, junctions). Table 26 shows a list of those hydrologic elements that were used for this research. Mathematical models are then used to characterize and calculate the hydrologic elements which describe the physical processes in the watershed. They define the methods for infiltration calculations (loss), surface runoff calculations (transform), and subsurface flow calculations (baseflow) in each subbasin. For each reach the routing method is defined. A reference (observed) flow can be defined for each junction. Which exact parameter values were found to be appropriate is lined out in Section 8.3.2. Table 26: Selected hydrologic elements available in HEC-HMS. Modified after Scharffenberg & Fleming (2008). Hydrologic element Subbasin

Reach

Junction

Description The subbasin is used to represent the physical watershed. Given precipitation, outflow from the subbasin element is calculated by subtracting precipitation losses, calculating surface runoff, and adding baseflow. The reach is used to convey streamflow in the basin model. Inflow to reach can come from one or many upstream elements. Outflow from the reach is calculated by accounting for translation and attenuation. The junction is used to combine streamflow from elements located upstream of the junction. Inflow to the junction can come from one or many upstream elements. Outflow is calculated by summing up all inflows.

8.3.1.2 Meteorologic Model Component The meteorological model component contains information on how to include which data about precipitation. Evapotranspiration and snowmelt data can be defined in this component if applicable.

8.3 Hydrological modeling using HEC-HMS

135

8.3.1.3 Control Specifications Component The control specifications component defines the modeling time frame and the time interval of the modeling process. 8.3.1.4 Input Data Components All data (time series, paired data, grid data) can be entered in the respective data managers of the input data components. 8.3.2 Parameterization Before the provided model can be used for runoff estimation, its parameters need to be specified to fit the model to the local conditions of the watershed (Ford et al. 2008). The choice of the appropriate mathematical model is outlined in Section 8.1.3. How the mathematical models are applied to parameterize each hydrologic element in the basin model component is subject of this section. The parameterization of the model also comprises the definition of boundary and initial conditions. In this case, initial conditions refer to the initial flow conditions. Boundary conditions are the “forces that act on the hydrological system and cause it to change” (Ford et al. 2008, p. 7). Table 27 provides an overview about the parameters set to describe these conditions in the model. Some of the parameters were individually specified for each subbasin and are separately listed in Table A41 in the appendix. Table 27: Global parameters used for the hydrologic model HEC-HMS. Parameter

Value

Initial abstraction ratio (Ia) Retention scale factor Time of concentration (Tc) Storage coefficient (R) Threshold baseow (Q0) Initial discharge value

0.010.4 1 0.482.05 0.482.05 0.1 m³/s 0.03 m³/s

Recession constant

0.95

Muskingum X Number of subreaches

0 1

8.3.2.1 Basin Model Component The initial input for the basin model component is the basin model generated during the processing with HEC-GeoHMS (Figure 22). In the present study the basin

136

8 Hydrologic modeling

model as pictured in Figure 22 consists of 11 subbasins, 5 reaches, and 6 junctions of which one is the outflow (gaging station San Ramón). 8.3.2.2 Basin Model Component I  Subbasin Loss method: Gridded SCS CN Loss: Besides the 50 x 50 m CN grid (Section 8.2.3), this method requires the definition of the initial abstraction ratio Ia and a scale factor. The scale factor is used to adjust the retention for each grid cell calculated by the curve number before it is multiplied by the initial abstraction ratio. The initial abstraction ratio (initial abstraction/potential maximum retention (Mockus 1972)) was standardized to 0.2 after extended tests (Scharffenberg & Fleming 2008, Mockus 1972) (compare Equation 3). It comprises losses through infiltration, interception, and surface storage that vary especially after a series of precipitation events when the infiltration capacities of the soil reach their limit (Mockus 1972). Several studies though encountered more realistic modeling results after lowering the initial abstraction ratio to values between 0.014 for the modeling of single events in a Greek watershed (Baltas et al. 2007) and 0.01 for long-term modeling in India (Mishra & Singh 2004). Woodward et al. (2003) claim that in 90 % of the cases a ratio smaller than 0.2 is appropriate and propose a standard value of 0.05 to be used instead. No infiltration experiments could be done in the scope of this research nor were they available from other studies. Still, the standard values were changed in the subbasins during the model calibration. It was found that it has a noticeable influence on the simulated discharge (Section 8.3.4) and has therefore been used to adjust the discharge values of each basin. Based on knowledge about the individual physio-geographic setting of each subbasin the values were estimated to be between 0.01 and 0.4. In general, the best modeling results were obtained for high ratios in the upper part of the basin and generally lower values in the rest of the basin. Differences were made with respect to snow and vegetation coverage. High abstraction ratios in the upper part of the catchment (0.4) account for possible storage on snow coverage or snowfall in these areas. Lower initial abstraction ratios in the subbasins with north-exposed slopes account for the sparse vegetation coverage in these areas. The values are set slightly higher in the areas with the south-facing slopes in the northern parts of the basin. The central subbasins which the river crosses all have low ratio values as most of the water directly discharges to the river. Refer to Table A41 in the appendix for a detailed listing of the values for each subbasin. The retention scale factor was left with the value of 1, meaning that the maximum retention that was delineated from the CN grid is multiplied directly by the value of the initial abstraction ratio. The higher the retention scale factor the lower the simulated discharge at the gaging station (compare Table A42).

8.3 Hydrological modeling using HEC-HMS

137

Transform method: ModClark: ModClark allows for a distributed simulation of rainfall data and consequent runoff generation based on a predefined grid (compare Section 8.2.4). Further parameters that need to be specified are the time of concentration tc and the storage coefficient R. By default, R is set equal to tc (see Table A41). When using the ModClark model, the translation (travel time) is accounted for in a grid-based travel-time model. For each cell of the 50 m grid the time until precipitation reaches the outlet can be calculated as: ‫ݐ‬

೏ ௖௘௟௟ୀ௧೎ ೎೐೗೗

Equation 7

೏೘ೌೣ

with tcell as the travel time for the cell, tc as time of concentration for the watershed, dcell as the travel distance from the cell to the outlet, and dmax as the longest possible travel distance in the basin (Feldmann 2000).

The travel time tc is defined individually for each subbasin. Respective values were calculated during the hydrologic pre-processing (Section 8.2.2, Table A41). The storage coefficient is mathematically expressed as: ܴൌ

ௌ೟ ை೟

Equation 8

with R as the storage coefficient, St as the storage at time t, and Ot as the outflow from storage at time t (Feldmann 2000).

Baseflow method: Recession Baseflow: The baseflow model is applied both before and after the storm. It requires the user to define a threshold value from which onwards it simulates the runoff in the channel (after the runoff from excess precipitation is modeled). There are two ways of defining this threshold: Either as a ratio to the peak (e.g. 10 % of the maximum flow) or directly via a threshold value. After an analysis of the hydrograph from measured data, the threshold runoff value was set to 0.1 m³/s in all subbasins as no detailed information for the individual subbasins was available. Setting a fix threshold value rather than a relative value seems plausible for this basin as the hydrograph is fairly regular. The baseflow Qt at any time is then calculated with: ܳ௧ ൌ ܳ଴ ݇௧ Equation 9 with Qt as the baseflow at the given time, Q0 as the “threshold” baseflow (or the initial baseflow), and k as an initial decay constant that is calculated by dividing Qt by Qt-1 (Feldmann 2000).

The model furthermore requires an initial discharge value that is the runoff observed one day before the rain starts. This value was set to 0.03 m³/s after analyzing the runoff behavior of the basin from measured data. The recession constant as a last parameter needed for the baseflow component is the ratio between the baseflow of the current time and the baseflow of one day earlier (Scharffenberg & Fleming 2008, p. 139). This value was for the present analysis calculated to be 0.95 in average.

138

8 Hydrologic modeling

8.3.2.3 Basin Model Component II: Reach The reach element is used to model river and streams. The amount of outflow is calculated by simulating open channel flow which is an appropriate method for river runoff simulations. Routing method: Muskingum: Three parameters are required for the Muskingum routing method: (i) Muskingum K, (ii) Muskingum X, and (iii) the number of subreaches. Muskingum K is the travel time in hours through the reach that has to be known or estimated. The same travel time as used for the ModClark calculations  that was calculated during the HEC-GeoHMS processing  was used here. Muskingum X, as described in Section 8.1.3 defines the level of attenuation. Setting this value to zero delivers most plausible results and represents high attenuation in the basin. The number of subreaches, which also affects attenuation, was set to 1. The number can be increased if attenuation needs to be decreased. 8.3.2.4 Basin Model Component III: Junction The junctions are functional elements that are not parameterized. 8.3.2.5 Meteorologic Model Component Evaporation is not considered when simulating single extreme events (Scharffenberg & Fleming 2008). Also snowmelt does not have a significant influence on the generation of runoff during heavy rainfall events. Precipitation is thus the only data set that is referred to in the meteorologic model component. Section 6.2 describes how the rainfall data from the nearby station Cerro Calán were transferred to the HEC-HMS supported grid-format *.dss to match the distributed modeling approach. Even though in this case rainfall data from just one station were used, the interpolation approach implemented in GageInterp was applied to generate a grid with uniform values at a cell size of 50 m (Section 6.2). Like the general modeling time interval, the precipitation data have a temporal resolution of 1 hour. Besides selecting the appropriate grid containing precipitation data, a time shift can be set. During the interpolation process with GageInterp the time zone was set to Central Standard Time (CST) with summer/winter time shifts, i.e. to 6 hours behind Coordinated Universal Time (UTC). As the time zone of Santiago de Chile is only 4 hours behind UTC, a time shift of -2 hours needs to be set.

8.3 Hydrological modeling using HEC-HMS

139

8.3.2.6 Control specification component The simulation time frame covers the duration of the rainfall event and ends after the runoff values return to normal level. The model is run for the following precipitation events: -

July 17, 2001; 11:00July 21, 2001; 08:00 May 25, 2002; 12:00May 29, 2002; 03:00 June 02, 2002; 15:00June 06, 2002; 12:00 June 27, 2005; 08:00June 29, 2005; 15:00 August 26, 2005; 10:00August 30, 2005; 15:00 May 22, 2008; 21:00May 23, 2008; 22:00 May 26, 2008; 18:00May 29, 2008; 08:00 September 05, 2009; 09:00September 08, 2009; 23:00

The events were initially selected based on the total amount of precipitation. During field stays however, it could be noted that precipitation events with less than 65 mm maximum rainfall also resulted in floods of the San Ramón channel. Thus, most important is the intensity of the rainfall. To illustrate this, two events from 2008 and one from 2009 were additionally selected for modeling. The first rainfall event in May 2008 with a total amount of recorded precipitation of 46.75 mm within 14 hours resulted in a flood while the second rainfall event in May 2008 with a total amount of recorded rainfall of 55 mm of which 38.25 mm fell within 24 hours did not cause river floods. The event in September 2009 with a peak rainfall of 30.25 mm in 24 hours and a total of 49.5 mm caused minor floods. The time interval for the simulation runs is set to 1 hour. The simulation starts few hours before the rainfall begins and ends a few hours after the rainfall has stopped when the discharge values decrease and return back to the baseflow level. 8.3.3 Simulation Using data and parameters as described above, the precipitation-runoff simulation can be started. The following subsections outline how the computation of results was performed and optimized. Runoff values are not available for the validation of the model output. However, input and output data are at a later stage related to existing hydrological investigations from the study area. The main issue with the input data was the CN grid that has to contain a header file exactly as shown in Table A39 in the appendix. After successful completion of the simulation run, the results can be viewed in graphical or table format (Figure 23).

140

8 Hydrologic modeling

Figure 23: Exemplary result from the runoff simulation at the gaging station Ramón.

8.3.4 Sensitivity analysis The sensitivity analysis is carried out to investigate the performance of the model under the condition of changing parameter values. That means that single parameters of the model are altered in a realistic range and that the resulting changes of the simulated runoff values are analyzed. The output of the model (discharge values) is in many cases evaluated using the Nash-Sutcliffe coefficient (Nash & Sutcliffe 1970). However, as measured runoff values are not available at a reliable quality level, the output values are solely compared with modeling results from other studies in this catchment. This sensitivity analysis focuses on the parameters “Initial abstraction ratio” (Ia), “Retention scale factor”, and “CN grid” (input raster data) as changes in all other parameters are not considered being realistic and comprehensible. Table A42 in the appendix shows the altered values for the first two parameters of interest for each subbasin. The last three columns of the table indicate the total outflow at the gaging station (Total), the peak outflow (Peak)  both in m³/s  and the time when the peak outflow is registered (Peak time). All remaining parameter values are left as indicated in Table 27 and Table A41. The names of the subbasin are abbreviated in Table A42 and can be located with basic relief information using Figure 24.

8.3 Hydrological modeling using HEC-HMS

141

Figure 24: Map showing the subbasins with names and shaded relief in the San Ramón basin.

First, Ia was tested. A general finding is that the parameter does influence the amount of the total and the peak discharge but not affect its time of occurrence. The parameters in up to three subbasins were altered in combination if the subbasins showed similar characteristics. The first block of Table A42 shows the analysis for the two highest-lying subbasins 90 and 91 which have only little vegetation coverage and together cover an area of around 11 km². The higher the value for Ia the higher the retention in the subbasin and the lower the total and peak outflow. The values for the total outflow reach from 26.25 m³/s to 27.58 m³/s, the peak values vary between 21.7 m³/s and 22.8 m³/s. High retention (Ia = 0.4) could for example result from snow coverage or precipitation falling as snow instead of rain. Low retention (Ia = 0.01) is assumed to represent the natural conditions of the subbasins. Changing the parameter values in subbasin 84, the south-central subbasin with predominantly sparse vegetation coverage in the same value range showed a stronger impact on the resulting runoff even though the subbasin has a smaller size (7 km²). The values for the total runoff range from 24.94 to 26.49 m³/s and the resulting peak runoff values lay between 20.4 and 21.9 m³/s. Applying a high value for subbasin 84 would imply that retention basins are installed in that area or that the snow line is very low. A low value for the initial abstraction ratio would imply a sparse vegetation coverage which seems to be realistic with respect to the exposition of the basin (compare Section 4.3.4). The analysis of the parameter values in the south-facing subbasin 79 with a size of 7.5 km² and a denser vegetation coverage resulted in peak flow values be-

142

8 Hydrologic modeling

tween 21.2 and 21.9 m³/s. Like in the previous example, assigning a high value (0.4) presumes the availability of retention basins or a low-lying snow line. With an average elevation in the subbasin of 1,900 m that might be the case under present climate conditions (Table 7). However, it does not seem to be realistic under climate change conditions. A low value again represents low vegetation coverage which might be the case if the climate gets dryer and less water is available for the plants. The parameters for the small subbasins 78 and 75 located in the central part of the basin did not show a very high influence on the resulting runoff. The two basins cover an area of less than 1 km², have a flat slope, and have comparably dense vegetation coverage as they form large parts of the river bed. Changing the values in the same range as in the subbasins before brought total runoff values between 25.98 and 26.29 m³/s and peak flow values between 21.6 and 21.7 m³/s. As the three north-lying subbasins 69, 74, and 66 in the central and lower part of the catchment show similar vegetation coverage and slopes they were analyzed as a group. The subbasins cover an area of about 5.6 km² and are covered by shrub and bush vegetation. Based on the same reasoning for the parameter selection as in the previous subbasins, peak runoff values between 20.7 and 21.8 m³/s result. The total runoff volume ranges from 25.22 to 26.39 m³/s. Subbasin 68 is located in the central eastern part of the basin and has a size of about 1.1 km². With its small size it does not show a significant impact on the resulting runoff volumes. Values for the total runoff between 25.97 and 26.27 m³/s and for the peak runoff between 21.4 and 21.7 m³/s support that assumption. To finish, subbasin 72 was analyzed. It covers the westernmost 5.2 km² in the catchment and contains the gaging station San Ramón. Changing its parameter values has a comparably high impact on the resulting total and peak outflow values. Assuming a high retention potential, e.g. as a result of newly constructed retention basins, would lead to total runoff volumes of 24.68 m³/s and a peak flow of 20.3 m³/s. With a low initial abstraction value of 0.01 the total runoff volume rises to 26.25 m³/s and the peak flow is 21.7 m³/s. Changing the retention scale factor was considered being a potential parameter to improve the model performance but did not yield realistic results. The impact on the resulting runoff values is evaluated to be too high. That means that a changing retention potential of single subbasins can most realistically be represented by changing the value of the initial abstraction ratio. Table A42 proofs that argumentation. Finally, the sensitivity of the applied CN grid was tested. The CN grid values result from the same soil and geology information combined with different landuse patterns depending on the year of investigation (compare Section 7.2). The test was carried out for the rainfall event on May 2223, 2008 with a maximum precipitation of 46.75 mm. Table 28 shows the resulting differences in peak and total runoff for the same precipitation event under different LULC conditions.

143

8.3 Hydrological modeling using HEC-HMS

Table 28: The influence of the CN values on the model output at the example of the rainfall event on May 2223, 2008. Year

Total runoff [m³/s]

CN grid 2002 CN grid 2005 CN grid 2009

Peak runoff [m3/s]

23.95 24.75 26.25

19.6 20.3 21.7

This parameter is a very important variable as the land-use/land-cover scenarios are represented through changing LULC maps and consequently through updated CN grids. Thus, the obvious sensitivity of the model regarding this parameter is very well suited to fulfill the purpose of this investigation. 8.3.5 Model optimization Optimization means that the model parameters are iteratively estimated to improve the modeling results. Most of the parameters were either physically defined or calculated during the data preprocessing steps. Thus, the values shown in Table 27 were not altered. However, the values shown in Table A41 were changed individually in each subbasin. Reliable reference runoff data were not available for the study area (Perez 2009, AC Ingenieros 2008). Thus, previous studies on runoff estimations for the area of interest were used to calibrate the model and to validate its results. Table 29 shows how precipitation intensities and modeled or calculated runoff volumes were associated to which return periods in previous studies (compare further Tables 5 and 6). AC Ingenieros (2008) had time series from the pluviometric station San Ramón available and used these time series to derive return periods. The values calculated by Perez (2009) are higher than the ones derived by AC Ingenieros (2008). Perez used precipitation time series from the station Quinta Normal located in the western part of the AMS and 380 m lower than the station Cerro Calán to delineate precipitation return periods and applied a factor to adapt the rainfall intensity values for the higher-lying station. Table 29: Maximum daily precipitation intensities [mm] with their return periods used for the modeling process in the San Ramón catchment: (a) derived by AC Ingenieros (2008) and (b) by Perez (2009). A&C Return period 2 5 10 25 50 100

Precipitation 45.8 62.8 74.1 88.3 98.9 109.5

Perez Runoff 19.8 27.1 33.1 43.6 52.6 63.7

Precipitation 63.8 82.5 94.9 106.8 122.2 133.8

Runoff 20.5 27.3 36.2 42.4 47.4

144

8 Hydrologic modeling

The study by the private consulting company AC Ingenieros (2008) used the empirical Rational Method to estimate runoff volumes of the San Ramón catchment for precipitation events with a return period between 5 and 100 years (Table A43). Perez (2009) applied the unit hydrograph method to determine minimum, maximum, and average runoff volumes under current climate conditions and for the IPCC scenarios B2 and A2 at the gaging station San Ramón (compare Section 6.3). More detailed modeling results are shown in Table A43 in the appendix. Table 30 shows the event characteristics with the hydrological model input (precipitation) and output (runoff) used for the present investigation. The duration (Dur.) comprises the number of hours where rainfall was recorded at the station. It is then distinguished between the maximum precipitation within 1 hour and 24 hours (Pmax) and the total precipitation of the event (Ptotal). Furthermore, the peak runoff (Qmax) is shown. The respective return period [in years] was associated based on the evaluation of the previous studies shown in Table 29. Thereby, the return period was assigned by relating the precipitation values with the maximum precipitation values within 24 hours and their return period from the study of AC Ingenieros (2008). Table 30: Modeling results for the San Ramón catchment.

Event July 1720, 2001 May 2528, 2002 June 0206, 2002 June 2729, 2005 Aug 2630, 2005 May 2223, 2008 May 2629, 2008 Sept 0508, 2009

Dur. [hrs] 47 49 51 25 48 14 38 41

Pmax [mm/hr] 6.75 12.00 13.25 6.75 10.25 5.75 4.50 8.75

Input

Output

Pmax [mm/24 hrs] 75.00 90.50 161.25 61.75 89.25 46.75 38.25 30.25

Ptotal [mm] 104.50 113.50 243.00 63.25 148.00 46.75 55.00 49.5

Qmax [m³/s] 38.50 31.90 73.50 27.50 45.60 21.70 14.70 14.50

Return Period 10 25 100 5 25 2 2 1

The resulting discharge values range from being almost similar to that same study to being rather similar to the values obtained by Perez (2009). As no final validation with measured reference data could be carried out in any of the studies and the obtained results show a plausible range, the parameters used in this modeling process are considered being appropriate for the calculation of runoff values. They are thus used for the scenario calculations as described in the following section. Figures 25 and 26 show the resulting hydrographs exemplified for two periods in time. Figure 25 depicts the rainfall event in May 2002 with a Pmax of 90.50 mm and Figure 26 represents a rainfall event in August 2005 with a comparable maximum rainfall but a higher peak runoff value (Table 30).

8.3 Hydrological modeling using HEC-HMS

145

Figure 25: Result of the hydrologic modeling using HEC-HMS: Precipitation data from Cerro Calán station and modeled hydrograph at San Ramón station for rainfall event in May 2002.

Figure 26: Result of the hydrologic modeling using HEC-HMS: Precipitation data from Cerro Calán station and modeled hydrograph at San Ramón station for rainfall event in August 2005.

146

8 Hydrologic modeling

Both rainfall events show two peaks in the precipitation and the runoff values. The hydrographs exemplify how the runoff recedes as soon as the precipitation stops or significantly minimizes. That stands for a fast drainage of the rainwater and a direct response of the runoff on precipitation. The temporal delay, i.e. the response time of the watershed is thereby approximately eight to ten hours. The difference between the two events is the duration of the rainfall as well as a changing LULC pattern in the basin that could be observed during the three years that lay between both modeling time steps. The influence of potential future LULC changes is investigated in the following section. 8.4 MODELING ALTERNATIVE LAND-USE/LAND-COVER SCENARIOS This section describes what the main characteristics and requirements of scenarios are and how they were used for this study. 8.4.1 The characteristics and functions of scenarios With respect to the goal of scenario work, two main types of scenarios can be distinguished: Exploratory and normative scenarios (van Notten et al. 2003). An exploratory scenario consists of a more or less creative storyline referring to driving factors and associated keywords that are meant to be capable and sufficient to describe thinkable future developments in a certain area with respect to a certain aim. While exploratory scenarios build on the uncertainties of future developments, normative scenarios refer to specific goals and describe ways how to reach these fixed goals, contemporaneously functioning as a type of decision support tool (McDowell & Eames 2006, van Notten et al. 2003, Kurz et al. 2000, Steinmüller 1997). Uncertainties and different options about the future development are meant to be included in the scenario process. In the current project, the aim is to show how flood risk related conditions might change under certain (varying) circumstances, to ponder which development directions are possible and what implications they bring. Therefore, exploratory, forecasting scenarios are applied (van Notten et al. 2003, Steinmüller 1997). Scenarios can either be created intuitively or formal (van Notten et al. 2003). While the formal approach often uses quantitative knowledge or computer simulations as prediction method, the intuitive approach is more creative, based on discussions and qualitative knowledge (van Notten et al. 2003). Several interviews were conducted during field work to obtain expert knowledge from different perspectives about possible future urban development (Section 6.8) as the intuitive approach is meant to be used for this study. The scenario creation is the pre-condition for the analysis and assessment of changes of the flood risk related conditions in that area and is meant to show potential futures and raise awareness for their consequences (Berkhout et al. 2002). Thus, different exploratory scenarios concerning the future appearance of the setting in the catchment of Quebrada San Ramón sketch the possible developments

8.4 Modeling alternative land-use/land-cover scenarios

147

of flood-hazard related conditions. In order to function as a learning machine, the developed scenarios need to be realistic. General quality criteria that each scenario should fulfill are highlighted in Steinmüller 1997, p. 62: -

Credibility: Scenarios need to be credible, plausible, and consistent. Utility: Scenarios need to fulfill the given purpose. Intelligibility: Scenarios need to be understandable and comprehensible. 8.4.2 Land-use/land-cover scenarios for the San Ramón basin

Three different possible LULC scenarios were developed for the San Ramón basin. The envisaged time horizon is the year 2030. Scenario I refers to the impacts of the projected climate change in the study area (Perez 2009) that is regarded being a very realistic future development. Scenario II refers to the ongoing afforestation activities that could be observed during field stays. Even though afforestation in the study area requires substantial irrigation activities and the seedlings need special protection as they are otherwise destroyed by fauna living in the basin it is very interesting to qualitatively assess the possible impact of an increase in tree coverage. Scenario III finally assumes that the construction restrictions are lifted and that leisure or residential areas would develop in the central parts of the basin that are characterized by flat slopes. This is an alternative that is for legal restrictions not officially being discussed. However, it is regarded being realistic if the neo-liberal economy system and low ecological awareness are maintained. The processes of describing the three different LULC patterns are based on the assumptions lined out below, respectively: Scenario I  Increasing aridity: The total amount of annual precipitation is decreasing and stronger drought periods need to be expected, thus there is less vegetation in the basin. However, the intensity of extreme events rises. The 0°C isotherm rises above 3,250 m (highest point in the basin) so there is no snow falling during precipitation events. Due to increasing temperatures and decreasing amounts of precipitation the vegetation stocks are reduced in the basin. Especially the north- and northwest-facing slopes in the lower and central parts of the basin show a reduced amount of bushes and scrubs. Rather, barren land dominates this region. The forested areas along the south-facing slopes in the lower part of the basin are substituted by sparse bush vegetation. However, no additional construction activities are permitted and the basin continues to be ecologically protected. Scenario II  Afforestation: The ongoing afforestation activities in the basin are successful and the amount of vegetation in the catchment area increases. The attempts to green the basin through tree planting activities are supported by the Regional Government. An irrigation network is set up in the lower parts of the basin. Water is partially taken from the river and supplementary taken from the public water supply network. The tree-planting activities are especially fruitful along the south-facing slopes as less of the strong sunshine during the growth period reaches the vegetation. Bush vegetation can develop along the north-facing slopes in the basin. No construction activities besides the negligible share of the

148

8 Hydrologic modeling

irrigation infrastructure occur and the status of a protected area remains valid. The changes in temperature are not as high as projected thus the 0°C isotherm is located at around 2,700 m in average, i.e. there is some snowfall recorded in the basin. Scenario III  Construction activities: The protection and construction restrictions according to the PRMS are changed and the development of residential areas above 1,000 m becomes possible. In terms of urban planning and environmental awareness, economic interests clearly dominate ecological value systems and attempts of environmental protection. The influence of the private sector on planning decisions, mainly through real estate traders, is high and proves the dominant force of the market. The Alvaro Casanova street that is located right south of the gaging station San Ramón is converted into a new highway and attracts investors. The construction limit of 1,000 m vanishes and allows for a significant urban expansion towards the Andean mountains. Medium-density residential areas develop along the foothills and also in the central part of the San Ramón basin as flat slopes, high air quality, and an excellent view on the city of Santiago de Chile offer a favorable spot for homeowners. However, irrigation infrastructure is installed in the basin to create private green spaces. The number of non-private green spaces minimizes as a result of longer drought periods that occur in parallel. The 0°C isotherm rises above 3,250 m so there is no snow falling during precipitation events. Figure 27 illustrates the mentioned alternatives and provides quantitative information about the amount of LULC changes. The LULC changes in the basin were assigned manually and using knowledge about the basin instead of using a standardized reduction rate for vegetation. 8.4.2.1 Modeling result The simulation runs using the alternative LULC scenarios were carried out for the same precipitation events as introduced before. The model parameters were left as indicated in Table 23 and Table A41, only the CN grid and consequently the retention scale factor, both representing the new LULC pattern, were changed. For Scenario I and III, the retention scale factor was set to 0.01 in all basins, representing little vegetation coverage in most parts of the basin. Only in the subbasins 68, 69, and 74 the factor was set to 0.05, representing bushland coverage or built-up area with a medium density, i.e. large proportion of private green spaces. Scenario II expects some snow fall in the upper part of the basin, thus, the retention scale factors were set to 0.2 in the subbasins 90 and 91. Depending on the exposition (north/south), the values were set to 0.1 and 0.05 in the remaining parts of the watershed. A dense vegetation coverage and consequently higher retention is assumed for the south facing subbasins 66, 68, and 69. The only exception is the east-lying subbasin 74, which is due to its currently sparse vegetation coverage not expected to experience a large increase of greening. Therefore, the retention scale factor is set to 0.01 in this area.

8.4 Modeling alternative land-use/land-cover scenarios

149

Figure 27: Land-use/land-cover data based on classification result of the ASTER satellite image from 2009 and scenario assumptions.

150

8 Hydrologic modeling

Table 31 summarizes the original model outputs based on the CN grids of the respective years and the results of the modeling process for Scenario I (dry conditions), Scenario II (afforestation), and Scenario III (construction). The peak runoff volumes are in bold. For a better visual comparison, the four model results for each time step are depicted in Figure A47 to 54 in the appendix. Table 31: Modeling results for the three LULC scenarios in the San Ramón catchment. Original Event July 2001 May 2002 June 2002 June 2005 August 2005 May 2008 (a) May 2008 (b) Sept 2009

Pmax [mm/24 hrs] 75.00 90.50 161.25 61.75 89.25 46.75 38.25 30.25

Return period 10 25 100 5 25 2 2 1

Peak [m³/s] 38.5 31.9 73.5 27.5 45.6 21.7 14.7 14.5

Scenario I Peak [m³/s] 41.5 35.6 78.5 30.0 49.2 23.3 15.4 15.2

Scenario II Peak [m³/s] 39.6 33.4 75.4 28.0 46.2 21.0 14.3 13.4

Scenario III Peak [m³/s] 41.3 35.4 78.2 29.9 49.0 23.2 15.3 15.1

8.4.2.2 Interpretation of the results The changing land-use pattern in the basin  either man-made or resulting from a changing climate  impacts the runoff behavior in the basin after extreme precipitation events. Interpreting the obtained results yields some interesting insights. The first modeled event, the rainfall event in July 2001, shows the lowest peak runoff values for the LULC pattern from 2002. As expected, increasingly dry conditions (Scenario I, compare Figure 27) or construction activities in the basin (Scenario III) would increase the amount of surface runoff. But even Scenario II, representing the afforestation efforts currently going on in the basin, results in higher runoff values than during the original conditions from 2002. The reason is that it still contains less forested areas and bushland than it used to be present in the year 2002. Denser vegetation coverage than already contained in Scenario II, however, is not regarded being realistic with regard to the projected impact of climate change. The same conclusions apply for the rainfall events in May and June 2002 and for the storms in June and August 2005. Only for the events in 2008 and 2009, an improvement, i.e. a reduction of the total and peak runoff in comparison with the current conditions, could be achieved with Scenario II. Another interesting finding is that the large-scale loss of vegetation due to increasing aridity would have an even more negative impact than construction activities in the central parts of the basin. The difference is not very large, but it can on this scale not be proven that potential future construction activities in the basin have a more negative impact on the runoff behavior than the impact of climate

8.4 Modeling alternative land-use/land-cover scenarios

151

change. These results represent the negative impact of the ongoing changes in vegetation coverage in the basin. The type of construction assumed in Scenario III though is medium density residential areas with a large amount of private green spaces. The private green spaces associated with the potential building developments would be most likely irrigated and maintained thoroughly throughout the year. That offers a higher interception and infiltration potential than barren land as long as it compensates for the loss of infiltration capacities due to construction activities in the lots. Even though that seems at a first glance like a support for the development of settlements in the basin it should be clearly kept in mind which ecological consequences building development in that area would have (compare Section 11.1.8). To conclude, if precipitation events like the ones modeled here would reoccur, the maximum runoff value is in all scenario cases higher than in the original setting, partly as a result of changing land use and as a result of the rising snow line. That means that the influence of a changing LULC pattern will be further accelerated by a changing climate. As changing land-use patterns are  most obviously in Scenario I  strongly interlinked with a changing climate, the potential influence of a changing precipitation behavior should be explored in future studies. The increase predicted by Perez (2009) in Pmax is around 2 % for a 2 year return period, 15 % for a 5 year return period, 20 % for a 10 year return period, 29 % for a 50 year, and 31 % for a 100 year return period in the case of B2. The changes are -3 % for all return periods the scenario A2.

9 FLOOD RISK ANALYSIS AND ASSESSMENT This chapter starts with the delineation of information from remote sensing, GIS, and census data to feed the indicators for the assessment of vulnerability and elements at risk (Section 9.1). Section 9.2 and 9.3 then describe the calculation of the index for the analysis and assessment of the flood hazard and the elements at risk, respectively. Section 9.4 introduces the equation used for the creation of a vulnerability index based on the selected indicators and describes how the index is used to calculate vulnerability on a neighborhood (manzana) level. Section 9.5 brings all three components of risk together according to the risk equation and Section 9.6 assesses the flood risk in the previous and current states and discusses the influence of LULC changes on flood risk today and in the future. 9.1 DELINEATION OF INFORMATION TO FEED THE INDICATORS Measured hydro-meteorological, GIS, and census data and the classification results of the remote sensing data are used to derive information for the indicators for vulnerability, elements at risk, and the flood hazard. This section presents how data for each indicator were derived. Runoff measured at gaging station in m³/s: The delineation of this indicator requires the employment of a hydrologic precipitation-runoff model and is as the only indicator described in the previous Chapter 8. Proportion of buildings with poor construction material per manzana: The base for this indicator is the composite index referring to the construction material of each building developed by the MINVU (Arriagada & Moreno 2006). The index defines garbage, adobe, clay bricks, and soil as poor construction materials for roof and walls. Plastics, concrete, and soil are classified as poor construction materials for floors. The relative frequency of respective buildings was calculated for each building block. Proportion of buildings located at or below street level per manzana: This vulnerability indicator can be derived from the analysis of remote sensing data in combination with GIS data. Buildings that are located at or below the elevation of the adjacent streets are prone to floods as some of the main roads in the study area regularly turn into water ways during extreme precipitation events. Respective streets were digitized using 2.5 m contour lines, the Quickbird satellite data, results from previous flood hazard studies (AC Ingenieros 2008), and local knowledge. The LULC map delineated from the Quickbird data (Section 7.3) was used as a base to delineate a building mask that comprises all image objects that are classified as buildings. Elevation values from the DEM delineated from 2.5 m contour lines were then assigned to the building mask (average elevation for each building) and to the hazard prone street network. Using spatial analysis applica-

154

9 Flood risk analysis and assessment

tions, the distance of a building to the closest street was determined within a radius of 100 m. The radius was chosen in that size to also include the front yards that are part of the property and might take damage but are not included in the building mask. Data of lot outlines were not available. The indicator thus comprises the relative frequency of the buildings located at or below the elevation level of the flood-prone streets per manzana. The following major roads in the study are prone to floods as they are located in former creek beds: -

Avenida Príncipe de Gales (Municipalidad La Reina 2008, p. 150) Avenida Larraín, for physical location (AC Ingenieros 2008) La Cañada, Aguas Claras, Monseñor Edwards, Pdte Ovalle, PRMS Carlos Silva Vildósola, El Sendero (AC Ingenieros 2008) María Monvel, Julia Bernstein (AC Ingenieros 2008) Talinay (Municipalidad Peñalolén 2006, p. 30) Avenida José Arrieta (Reiter 2009, AC Ingenieros 2008, Municipalidad Peñalolén 2006, p. 204) Las Vertientes (AC Ingenieros 2008) Antupirén (AC Ingenieros 2008, (Municipalidad Peñalolén 2006, p. 204) El Buen Camino (AC Ingenieros 2008) Avenida Departamental (AC Ingenieros 2008)

The indicator finally includes the relative proportion of buildings per block that is located at or below street level. Proportion of people under 5 and above 65 years old per manzana: The relative number of all people below five and above 65 years was calculated for each building block based on census data. Proportion of female population per manzana: The relative number of the female population was calculated for each building block based on census data. Proportion of heads of households with poor education, per manzana: Based on the number of school years that the head of the household attended, Heinrichs et al. (2009) delineated a composite index with seven groups referring to the level of education. The indicator thus contains the relative number of heads of household with no or incomplete basic education (two lowest groups) per building block. The original data were available from the census. Proportion of households with more than 2.5 people per bedroom per manzana: The relative number of households per manzana with more than 2.5 people sharing one bedroom is an indicator delineated from the census data base after MINVU (Arriagada & Moreno 2006). Proportion of green spaces per manzana: The GIS-based combination of all types of vegetated areas as derived from the analysis of the Quickbird data, i.e. trees, grassland, dry vegetation, bushland, and agriculturally used land leads to one class “vegetation” (compare Table 20). This is used as a base to calculate the relative vegetation coverage per manzana. The inverse number was used for the indicator as the lack of vegetation determines the vulnerability (i.e. 1 minus proportion of green spaces). Proportion of people without employment per manzana: This indicator refers to the relative proportion of the people in each building block without em-

9.2 Flood hazard assessment

155

ployment, i.e. those people that are according to the census seeking employment and have worked before, seeking employment without having worked before, or are permanently unemployable. Proportion of people without permanent income per manzana: The relative proportion of the people in each building block without income, i.e. those people that are according to the census working for the family, students, retired, homemaker, or that have employment but are not working is contained in this indicator. Amount of people per manzana: The amount of population per manzana is counted based on the location where they have spent the night before the acquisition of the census data (de hecho, Welz 2005). This number describes the indicator for each building block. Amount of infrastructure per manzana: GIS data showing the location of certain types of infrastructure were used to involve the amount of infrastructure per building block. The infrastructure information comprise the location of metro stations, roads, sport facilities, shopping centers and malls, banks, supermarkets, health facilities, public services, public security (police stations, fire fighters, etc.), ponds, and high voltage lines. The amount of infrastructure per building block equals the sum of infrastructure facilities in each building block. In addition, it comprises the amount of built-up area [in ha] per building block that could be delineated from the Quickbird image. Descriptive sheets for each vulnerability-related indicator are attached in the Appendix (Figures A55 to A62). 9.2 FLOOD HAZARD ASSESSMENT After the previous Chapter 8 explained the determination of the hazard-related variables this section relates the runoff values to the existing flood hazard maps and does then explain how the hazard maps are incorporated in the further risk analysis. 9.2.1 The relation between runoff and flood extents The six flood hazard maps available from the study of Perez (2009) were generated based on the following runoff volumes: 27.3 m³/s, 47.4 m³/s (baseline scenario), 38.2 m³/s, 64.6 m³/s (IPCC scenario B2), 35.1 m³/s, and 50.9 m³/s (IPCC scenario A2). While these different runoff volumes result from changes in precipitation intensity and frequency, the different runoff volumes calculated in the scope of this research result from changing land-use and land-cover pattern. Leaving the precipitation probability out at this stage allows for a combination of existing hazard maps and newly calculated runoff volumes. As a matter of fact, the new runoff values do not exactly match the runoff values on which the calculations of Perez (2009) were based. Nevertheless, a combination of both data sets delivers an in-

156

9 Flood risk analysis and assessment

sightful output. With an increase of the runoff values from 27.3 to 35.1 to 38.2 m³/s the affected area changes from 188 to 214 ha (runoff + 7.8 m³/s) and from 214 to 217 ha (runoff + 3.1 m³/s). Assuming higher discharge values the changes are accordingly: With an increase from 47.4 to 50.9 to 64.6 m³/s the flooded areas (only outlines) would increase from 230 to 243 (runoff + 3.5 m³/s) and from 243 to 396 ha (runoff + 13.7 m³/s) (Perez 2009). In addition, the water depth changes with an increasing amount of overflow (Perez 2009), Figure A64. The changes are not linear but indicate that in that dense urban environment the damage would increase notably. The higher the absolute peak discharge values the higher the absolute changes in the spatial extent of the hazard zones with the same absolute increase in discharge. That means that the absolute changes are higher for high runoff values, i.e. low frequency events. The runoff volumes calculated for a changing land-use pattern are shown in Table 31. Taking the rainfall event from August 2005 as an example shows that the runoff would increase from 45.6 m³/s to 49.2 m³/s in Scenario I. That means an absolute increase of approximately 3.6 m³/s. Comparing these numbers with the ones presented above (i.e. 100 year return period for baseline scenario and scenario A2) shows that the areas affected only by the San Ramón channel would increase by approximately 13 ha (compare Figure A64). Even though this number seems not to be threatening at the first sight it is considerably high when referring back to the research area, i.e. a dense urban environment with a high density of people and values. In addition to the increase of the spatial extent of the affected areas, most of the flooded areas would face a higher water depth which requires more investments in measures to reduce the physical exposure and most likely results in longer duration of the floods. Furthermore, additional regions of the study area are stronger affected by river floods from other creeks or by urban floods if they are located along the before mentioned flood-prone streets. The projected increase in precipitation intensity for the extreme events is in this analysis not yet incorporated. The impact is less for events with a lower return period, e.g. for the example of May 2008 (a) with a return period of two years. The runoff would increase from around 21.7 to 23.3 m³/s, i.e. by 1.6 m³/s. This would also increase the size of the affected areas and the water depths, but to a lower degree.

9.2 Flood hazard assessment

157

9.2.2 Quantification of the hazard The flood hazard maps showing water heights with a spatial resolution of 5 m were generated for events for the current climate conditions and for the IPCC scenarios A2 and B2 with return periods of 10 and 100 years (Perez 2009), compare Section 6.3. Maps resulting from these calculations that contain a normalized hazard index are used for the further risk analysis. Equation 10 demonstrates how the normalization of the flood depths was carried out to obtain a flood hazard index with values between 0 and 10: ‫ ܫܪ‬ൌ ͳͲ ‫כ‬

௛ ௛೘ೌೣ

Equation 10

with HI as the flood hazard index between 0 and 10, h as water height at the given location (m), and hmax as the maximum water height in the area (m).

The hazard maps shown in Figure A63 are the most recent products available for that region but only include the flood hazard coming from the Quebrada San Ramón. However, it was mentioned before that the hazard in the study area comprises several smaller creeks (AC Ingenieros 2008) and that floods also result from water accumulating on the streets. Therefore, more data would need to be considered for a more complete and comprehensive risk analysis. Figure A65 shows in addition to the map of Perez two more flood hazard maps with a return period of 10 years for the municipality of La Reina. One is from the private consulting company AC Ingenieros (2008), the other one is the map from 1987 anchored in the PRMS. The information content and data format is different in each case but the identified flood extent is comparable amongst the maps. The upper map, which is also used as a decision making base for regional planning and zoning, shows the highest amount of potentially flooded areas but does not provide any information on water depth. The second map in the middle of Figure A65 only shows the points of overflow and no spatial extent of the hazard zones. However, it comprises the hazard of all smaller creeks coming from the Andean mountains. It can practically not be applied though for a spatial risk analysis. The lower map shows the hazard map for a return period of 10 years after Perez (2009). Spatial extent as well as flood depth (in this figure normalized numbers) are available. This type of information, however, is only available for the limited area as displayed in Figure A65. The final risk calculations are thus based on two hazard maps: the very recent map of Perez (2009) and the hazard maps anchored in the PRMS (MINVU 2005). While the numbers of the normalized flood hazard map by Perez can readily be used the maps of MINVU first need to be normalized to become comparable to the first set of maps. The flooding probabilities were used for the normalization and two maps were generated: When considering events with a 10 year return period only, the value 10 is directly assigned to all areas that have previously been flooded at least once every 10 years (medium and high hazard level) and the value zero is assigned to all other areas.

158

9 Flood risk analysis and assessment

For a more general analysis the values between 0 and 10 are assigned equally between the different hazard levels: high hazard level = 10, medium hazard level = 6.7, low hazard level = 3.3, no hazard = 0 (compare Table 32). Table 32: Definition of flood risk levels after Fernández & Montt (2004). Risk level

Explanation

High risk

Areas that are prone to floods during normal precipitation events, flooded between several times per year and once every two years Areas that are only flooded during above average precipitation events, occurring between once every two years and once every ten years Areas that are only flooded after exceptionally meteorological events occuring less than once every decade Areas that have never been flooded and that are not expected to be flooded in future

Medium risk Low risk No risk

9.2.3 Analysis of the hazard maps The flood heights are generally lowest at the gaging station of San Ramón under the current climate conditions (most areas do not exceed flood heights of 30 to 40 cm) and highest for scenario B2 (maximum of 160 cm at a runoff volume of 64.6 m³/s). While single affected areas can no longer be made out in the most negative scenario B2 for a 100 year return period (Figure A63) all other simulation results allow the distinction of three main affected areas: in the northern part the two zones around Avenida Príncipe de Gales and the zone between Avenida Larraín, and Avenida Tobalaba in the western part of the study area (Perez 2009). As further described by Perez (2009), the northernmost zones (north of Avenida Príncipe de Gales) need to expect flood heights with a 10 year return period ranging from 10 to 20 cm under current conditions (precipitation: 94.9 mm, maximum runoff: 27.3 m³/s) and 40 to 50 cm for the scenario B2 (precipitation: 114.2 mm, maximum runoff: 38.2 m³/s). The flood heights with a 100 year return period are 20 to 30 cm (precipitation: 133.8 mm, maximum runoff: 47.4 m³/s) and 80 to 100 cm (precipitation: 174.9 mm, maximum runoff: 64.6 m³/s), respectively. The areas south of Avenida Príncipe de Gales face slightly higher depths (approximately +10 cm) for the 10 year return period and flood depths up to 1.6 m in all three cases (current, A2 and B2) for the 100 year return period. Worst affected is the frequently flooded area north of Avenida Larraín and east of Avenida Tobalaba. The flood heights are simulated to be 60 to 80 cm under current conditions and up to 100 cm under both climate change conditions for the 10 year return period. Like in the northern areas, flood heights exceed 150 cm in all scenarios for the 100 year return period. As already explained in Section 6.3 a methodological drawback is that the modelation conducted by Perez assumes that the water does not flow over the sidewalk which would be the representation of real conditions. The maximum

9.3 Assessment of the elements at risk

159

water heights calculated above the streets are thus not realistic. However, Perez (2009) concludes that the spatial extent of the hazard zones coincides well with observations from field from previous years and that the most adverse effects need to be expected for scenario B2. The affects are accordingly lower for scenario A2 and the current climate conditions. The maps anchored in the PRMS (MINVU 2005) show similar flood extents, but are more generalized in terms of their outline and water depth information. However, they also include areas affected by urban floods, shown by the four smaller polygons in the upper part of Figure A65. How reliable the information content is can not be determined as sufficient validation data are not available. 9.3 ASSESSMENT OF THE ELEMENTS AT RISK 9.3.1 Quantification of the elements at risk The number of elements at risk per building block equals the sum of people, builtup area, and infrastructure located in that area (Figure A66). With respect to flood risk analysis it needs to be quantified how many elements at risk are located in the respective region. As the amount of affected infrastructure can hardly be set equal to the number of affected people and the built-up area in hectares, a scaling factor is applied. It is assumed that the considered infrastructure serves 1,000 people each day in average. The value is roughly approximated and based on the fact that there is one infrastructure element per 1,000 people, whereby no distinction is made between different types of infrastructure (compare Section 9.1). Respective data about the absolute or relative importance of infrastructure elements was not available. It is a matter of fact that the average person makes demand on more than one type of infrastructure in his or her daily routine in the course of a month but given the fact that the floods do usually not last longer than few days this value is considered appropriate for this purpose. The number of buildings can either be represented by the amount of built-up area in hectares or as the number of buildings. The first option is used here, initially for methodological reasons: the built-up area was delineated from very high resolution satellite data. While the classification accuracy for the delineation of the area covered by buildings is high (0.88 producer, 0.83 user, Section 7.3.3) it is still challenging to correctly determine the exact number of buildings, especially in densely settled environments. A second reason for the choice of this method is that the walls and floors of small buildings suffer less absolute damage than the ones of large buildings: A fact that would not be considered when using absolute building numbers. Therefore, it is most appropriate to use the area in hectares rather than the number of buildings. Thus the number of elements at risk (EaR) equals the sum of the population living in that area (Pop), the built-up area [in ha] (B), and the number of infrastructure (IS) times 1,000. Bringing all these components together implies adding up different units. As this is mathematically incorrect, the built-up area is for the further calculations considered to be unitless rather than being used as an index or normalized value.

160

9 Flood risk analysis and assessment

The goal of this study is to directly indicate the number of affected people, builtup area, and infrastructure and to leave as many information included and transparent as possible. In a final step the amount of elements at risk per building block needs to be normalized to allow for a risk index calculation. The original information does thereby remain available in an adjacent database. The counting of the elements at risk takes place in a first step using Equation 11: ‫ܴܽܧ‬௠௔௡௭ ൌ ܲ‫ ݌݋‬൅ ‫ ܤ‬൅ ‫ͲͲͲͳ כ ܵܫ‬

Equation 11

with EaRmanz as the absolute number of elements at risk per building block, Pop as the total population living in the building block, B as the built-up area in ha, and IS as the number of infrastructure.

In a second step the absolute numbers are normalized and multiplied by 10 after Equation 12: ‫ ܫܴܽܧ‬ൌ ͳͲ ‫כ‬

ா௔ோ೘ೌ೙೥ ா௔ோ೘ೌೣ

Equation 12

with EaR-I as the normalized elements at risk index between 0 and 10, EaRmanz as the absolute number of elements at risk per building block, and EaRmax as the maximum number of elements at risk in the area.

The resulting map shows the overall distribution of the elements at risk in Figure A67 in the appendix. When the infrastructure facilities were point locations (banks, supermarkets, metro stations, etc.), their numbers were summed up for each building block. The numbers were likewise summed up for each building block when the infrastructure type was not a point but a linear element (e.g. metro lines, high voltage lines). One infrastructure per building block is counted where the elements intersect with the outlines of the building blocks. In several cases, infrastructure elements are not directly located within the administrative outline of a building block. Respective elements were counted to the nearest building block if the distance did not exceed 20 m to avoid false assignments. As pictured in Figure A66, some infrastructure elements remain without assignment as the respective blocks have in a previous step been excluded for the vulnerability analysis (because of data unreliability). These elements are omitted in the further analysis  a methodological drawback that is addressed in the discussion of the results (Chapter 11). 9.3.2 Analysis of the map of elements at risk The distribution of people and built-up area is rather heterogeneous over the study area indicating a very diverse settlement structure. While some of the building blocks show a high building density and number of population (central parts) some other building blocks show a high density of buildings and low number of people (eastern areas, newly built-up). Even though some of the building blocks occupied with newly constructed buildings were excluded for this analysis, a reason for this phenomenon might also be the time gap between the different data

9.4 Flood vulnerability assessment

161

sources: the census (population) data are from 2002 while the built-up areas were delineated from a satellite image taken in December 2006. Given that the study area has a predominantly residential character the infrastructure equipment mainly consists of education, health, and leisure facilities. Metro stations can only be found in the very western part of the study area and one local airport exists in the central part of La Reina. Supermarkets and banks are unequally scattered along the main roads. Informal or very small shops in the ground floor of private homes are abundant in the study area but not registered in a data base and therefore omitted in this analysis. Police stations, fire fighters, and communal service locations are mostly located at or close to public squares and spaces. As shown in Figure A66 in the appendix, built-up area and a high number of people are located in flood-prone areas in La Reina and Peñalolén. Besides that, a number of education facilities and banks are located in areas facing a high flood hazard level in the north-eastern part of the study area. In the central part of the study area, police stations and fire fighters, a medical practice, the municipality of Peñalolén, and several educational facilities are again located in flood-prone areas. Only few infrastructure facilities are located in the flood-prone areas in the southern part of the study area. Sport facilities are situated in flood-prone areas in all parts of the study area. This is a very useful planning strategy as they might function as retention areas. Planning sport facilities away from flood-prone areas however would enable their usage as tent camps or gathering points as shown after Hurricane Katrina hit New Orleans, USA, in August 2005. All roads and streets would theoretically need to be taken into consideration as infrastructure as they serve as corridors for private and public transport. For methodological reasons this is not done here. This study carries out a risk analysis on the level of the administrative unit of building blocks which do not comprise any type of roads. Nevertheless, as the results are presented in a map every expert and local decision maker can visually obtain information about the spatial distribution and content of the potentially flooded area which minimizes this methodological disadvantage. The display in Figure A67 in the appendix contains the summed up absolute numbers of elements at risk rather than values normalized using the total size of the area. The rationale is again to maintain the largest content-informational value as possible for potential users of the maps. The goal is to show the absolute number of people and likewise built-up area [in ha] affected and to therewith provide a decision-making aid for the planning and implementation of risk reduction measures. 9.4 FLOOD VULNERABILITY ASSESSMENT 9.4.1 Evaluation of the vulnerability-related variables While the relevance of the variables related to hazard and elements at risk are physically or mathematically defined, the relevance of the variables referring to vulnerability with respect to flood risk is more challenging to evaluate. Interviews

162

9 Flood risk analysis and assessment

carried out with local decision makers and affected households allowed a broader estimation of respective information and deeper insight into the local, site-specific conditions. Questionnaires sent out to experts working in the Regional Government or branches of ministries, in non-governmental organizations as well as communal planning institutions in the research area and 82 household surveys carried out in the municipalities of La Reina and Peñalolén (Reiter 2009) brought the findings lined out in the following. 9.4.1.1 Results of the expert surveys The evaluation of eleven questionnaires (Section 6.8.1) showed that parameters referring to the location of a building (position of building in relation to street level and distance to water way) as well as the construction material of a building and the availability of private protection or mitigation measures are expected to be most relevant for the assessment of vulnerability. These variables refer to the exposure and physical part of the vulnerability construct. In addition, variables referring to the coping capacities, e.g. knowledge about the hazard and experience with floods, were also ranked important. In more detail, the variables were rated in average as follows: -

Position of building in relation to street level (0.85) Availability of flood protection measures on building (0.85) Distance to channel (0.8) Construction material of building (0.75) Experience with floods (0.725) Knowledge about floods (0.725) Knowledge about flood protection measures (0.7) Socio-economic level of household (0.65) Proportion of green spaces per building block (0.675) Age (0.425) Building usage (commercial, residential, industrial) (0.4) Occupation status (0.325) Gender (0.3)

The variable distance to the channel was introduced here to test the understanding of flood occurrence. It was proven before  and confirmed in a number of questionnaires  that this intuitively important factor is only valid for river floods but not for urban floods (e.g. resulting from lacking storm water drainage system), which are also a frequent phenomenon and which are likewise considered in this analysis. This high value thus refers to an exposure variable that is in its content not sufficient to cover all hazard aspects and is therefore left out in the further analysis. While age groups, gender, and building type/usage were rated little important (42.5 %, 30 % and 40 %, respectively), the socio-economic level of a household was in average evaluated to be of (in this case only) 65 % importance. This last

9.4 Flood vulnerability assessment

163

value and the interpretation of comments from the expert interviews show that not only low income groups, but also middle and high income groups are affected by floods, a fact supported by the results from the household surveys (Reiter 2009). The difference between the socio-economic groups is rather the type of damage. However, that already indicates that socio-economic variables are not useful for the explanation of vulnerability. 9.4.1.2 Results of the household surveys Answers from the residents (Section 6.8.2) about material and immaterial damages they suffered during flood events improved the understanding about what it practically means to be affected and to suffer damage from floods. 34 out of 82 households declared they suffered physical damage, i.e. that parts of the exterior (garden furniture, plants) or interior equipment (floor, documents, electric equipment, furniture, etc.) got wet and destroyed to different degrees. Three cases reported that the sewer system broke and excrements could enter their houses. In total, 53 households declared they suffered immaterial damage. With more than one answer possible, 24 suffered limited mobility, 18 isolation, eight had financial losses as they could not go to work or as their shops were inundated, and four in each case declared power and water outages, illness, and mental stress as the main problems (Reiter 2009). According to the local residents, the location of a building in relation to the street level (above, at or below street level), the employment and income status (no, sporadic or permanent occupation, no permanent income), and the number of dependent people do best explain the households’ evaluation of own affectedness (compare Section 6.9.2). Looking at variables such as structural flood protection measures, the relation with the affectedness is diverse: on one hand, it prevents from material damage, on the other hand it results that people are captured in their buildings (mostly in the case of gated communities) and suffer immaterial damage. In the questionnaires, 37 households stated they apply temporary mitigation measures such as sandbags (23 people) and cleaning of drains and gutters before the raining season starts (22 households) (Reiter 2009). A central finding from the field surveys that could not statistically be proven though is that the experience with floods plays a central role for the level of vulnerability. Households that had suffered any type of damage during floods did take precaution measures as it is widely known that floods are a regularly reoccurring phenomenon. Twenty of the households took permanent measures: six heightened their houses, e.g. with cement, 15 constructed walls or watergates, four constructed a private drainage system on their property (multiple answers possible) (Reiter 2009). No relation could be shown for the variables level of education and income groups with knowledge about protection measures and taking of measures. The reason is most likely that information is largely circulated on informal networks working independent from the social status.

164

9 Flood risk analysis and assessment

Figure A68 in the appendix shows the areas where the inhabitants suffered predominantly immaterial damage and the regions where people suffered material damage. Out of the interviewed households, 34 suffered material damage. Twelve of these households (35.3 %) are located in building blocks that belong to the highest socio-economic group ABC1 (average over the building block, data from the Adimark survey Adimark 2003). The same amount of households belongs to the lowest socio-economic groups E and D. That means that households of all socioeconomic strata are prone to suffer physical damage from floods. Nine of the poor households and eleven of the rich households also suffered immaterial damage. Nine additional households of the low socio-economic strata and seven additional households of the high socio-economic strata claimed that they only suffered immaterial damage during the previous flood events. That clearly proofs that households of all socio-economic levels are located in flood-prone areas and that all of them suffer damages. At the same time it needs to be noted that not all households located in flood-prone areas automatically suffered damage. Finally, it is again stated by Reiter (2009) that the frequently used socioeconomic indicators are not sufficient for the explanation of the generation of vulnerability. 9.4.1.3 Comparison of the results Interesting is the comparison between the results from the household survey carried out by Reiter (2009) and from the expert survey. Although the results cannot be compared directly as the applied methods were different and the set of variables taken into consideration was not identical, some findings can be highlighted: While the location of a building in relation to the street level and the slope of the street as a hazard indicator have in both cases been ranked important, the employment status of the people (permanent contract or sporadic work) was ranked as very important in the household survey but only little important (32.5 %) in the expert survey. The same applies for the availability of private protection measures: While experts rank this parameter as very important (85 %), the household survey showed no significant relation between the households that have private flood mitigation measures (e.g. walls or watergates) and households that suffered damage. However, even though it is not statistically significant, the high relevance of this indicator was experienced during the field survey. As shown before (Figure 18) the ranking of the importance of each characteristic varies amongst the experts, depending on their professional background, the administrative level they are working on and possibly also the interests that each institution follows. The results show the different viewing points and also the knowledge gaps in the scope of vulnerability assessment. To investigate the relevance of the different elevations (i.e. the weights) of each indicator for the final vulnerability analysis, a sensitivity analysis was carried out at the end of the analysis (Section 9.4.4).

9.4 Flood vulnerability assessment

165

9.4.2 Quantification of vulnerability A vulnerability index adapted from Haki et al. (2004) and also used by Kienberger et al. (2009) is applied to calculate the vulnerability of each neighborhood towards flood events using the selected indicators. For the practical implementation, the index is normalized by dividing the vulnerability score by the number of vulnerability items, i.e. the maximum vulnerability value is 1. The normalized composite vulnerability is thus calculated based on the formula: ܸ௠௔௡௭ ൌ

σ೘ ೔సభ ௩೔ ௤೔ ௜

Equation 13,

with Vmanz as the vulnerability score, vi as the weight of each variable (ranging from 0 to 1) and qi as the relative frequency of the variable per manzana (ranging from 0 to 1), and i as the total number of indicators.

No absolute weights for the indicators were available from the household survey. Thus, the weighting options from the expert survey were used as a base for the assignment of weights: The most important indicator in both cases obtained the weight 1.0 to make use of the maximum value and to underline its importance expressed through both surveys, even though the average value was only 0.85 during the expert survey. The values 0.75 and 0.7 were then assigned to the second and third indicator as a clear distinction between their importance was very hard to determine from both surveys. The value for the second indicator exactly equals the number obtained from the expert survey and the evaluation of the third indicator is basically the rounded value from the expert survey for the respective variable. The same weights were applied for the indicators rated important during the household survey to allow for consistence and higher comparability. The index is applied using a GIS with all relevant input data being available in a digital spatial database (polygon shape file). As a tool for the application of the vulnerability index was previously not available in a GIS it was created using the ArcGIS Model Builder. Using pre-defined components from library a tool was created that asks the user to enter weights for each vulnerability indicator. After all weights are entered in a valid format (ranging from 0 to 1) the index is calculated on the fly based on Equation 13. The application of the tool requires an empty column in the database with the vulnerability information. This column is during the execution of the process filled with values based on which a vulnerability map can be displayed. Using simple GIS-query functions, the vulnerability-relevant information can be retrieved from the database. To later use Vmanz for the overall risk calculations the vulnerability results need to be normalized to have values between 0 and 10, just as the results for the elements at risk and hazard calculations. This is done using Equation 14: ܸ‫ ܫ‬ൌ ͳͲ ‫כ‬

௏೘ೌ೙೥ ௏೘ೌೣ

Equation 14

with VI as the normalized vulnerability index between 0 and 10, Vmanz as the vulnerability index per building block, and Vmax as the maximum vulnerability index in the area.

166

9 Flood risk analysis and assessment

9.4.3 Analysis of the vulnerability maps Using all indicators (also the ones with a low weight according to the evaluation of the experts) leads to a rather homogeneous distribution of vulnerability. For the chosen level of investigation, the variables age and gender can generally not be considered to be a valuable contribution or input information for the vulnerability map. As the census data are aggregated on a building block level the proportion of male and female population is in almost all cases approximately 50 %. Information about where exactly the “old, single, female”  classically exemplifying the most vulnerable individual  lives is impossible to obtain. For that reason, only those three indicators that are according to (i) the experts and (ii) the affected households most relevant for the determination of vulnerability were used. In the case of the experts that are the position of buildings in relation to street level, the construction material, and the amount of green spaces. According to the household survey it concerns the position of the building in relation to street level, the employment status, and the household size. To describe the employment status, the proportion of people without employment but also the proportion of people without permanent income were considered. Household size is here expressed through the number of people per bedroom as that describes the critically large households which this indicator targets. Vulnerability maps containing the most relevant information according to each group were calculated based on the three indicators each. Figure A69 shows the vulnerability map for the two municipalities La Reina and Peñalolén. Green colors indicate low, yellow  orange colors medium, and red colors high levels of vulnerability. The spatial unit for the analysis is the building block. The maps are derived using the normalized vulnerability index and the weights as evaluated by the experts. Most vulnerable to suffer damage from floods are those building blocks along the roads that are constructed in former creek beds and that are located on the lower-lying part of that street. A comparison with the indicator sheets (Figures A55 to A62 in the Appendix) further enlarges the sitespecific knowledge and eases the interpretability of the maps with respect to the origin of vulnerability. A likewise high level of vulnerability show those building blocks that contain a low amount of green spaces and a high amount of buildings with bad construction material, whereby the number of buildings with bad construction material is comparably low in the study area (compare Figure A56). The vulnerable areas thus are the low-income settlements in the south-western and north-western part of Peñalolén (former tomas de tereno, compare Section 4.1.3, middle and lower black boxes in Figure A69) and several building blocks located along those before mentioned large streets that do until now not have a functioning storm water drainage system. Figure A70 shows the vulnerability map for the same municipalities derived using the same index but with weights as determined from the household surveys, i.e. from the point of view of the affected population. According to this evaluation, the building blocks along the before mentioned large streets (former creek beds) are rated vulnerable. Likewise vulnerable are the building blocks located in high-

9.4 Flood vulnerability assessment

167

er-lying regions of the Andean piedmont. These regions were not included in the household survey, but have a comparably high number of inhabitants without permanent employment. In this case however, this is not an indicator for a low social status of the household as these buildings to a large part belong to people preferring an alternative lifestyle. Further field surveys are required to determine if these households also face a higher vulnerability towards floods. In this region that only becomes relevant if the lots are affected by floods at all. It becomes obvious that the resulting vulnerability is similar to the expertbased map in some regions but differs in other regions. Differences occur for example in the region of the local airport in La Reina (upper box in the image) where overpopulation (large household sizes) are the main determinant for vulnerability when using the results of the household surveys. Comparing these maps with the results of the household surveys carried out by Reiter (2009) indicates that the building blocks are most likely to a large degree heterogeneous. In other words, the values obtained for each building block generalize the actual conditions. The vulnerability values for the households that suffered both material and immaterial damage range from 0.94 to 2.4 according to the evaluation of the households and from 1.02 to 2.99 according to the evaluation of the experts. The maximum vulnerability scores are 2.4 and 3.1, respectively. That means that all affected people are in the vulnerability analysis rated as being vulnerable but the maps generated on a building block level can only provide an orientation and are not sufficient to carry out a detailed vulnerability analysis. For the further analyses and the assessment of flood risk, the vulnerability maps derived using the three indicators that are ranked important according to the expert survey are used as they only focus on exposure indicators and physical vulnerability which are more reliable on the considered spatial scale. The experts rate the personal statistic information with a rather low relevance. With respect to the individuals, their specific risk-relevant knowledge and experiences are rated to be more important (even though this can in any way not be considered in this study for data availability reasons). Besides the individual characteristics, the group of the professionals broadly agrees that the information that are available on a building or building block level e.g. amount of green spaces, construction material, are most important for flood vulnerability. Going back to the definition of vulnerability, this proofs that the physical (or exposure) side seems to have a much higher influence on vulnerability than the social characteristics. The possible reasons for this judgment are double-edged: it is either really true or the social aspects are not sufficiently perceived and considered in the expert’s understanding. Coming back to the social aspect, the results of the household’s evaluation show that the importance of social characteristics is higher there. While the different viewing perspective might partially explain this finding, there is also a methodological explanation. The surveys were carried out with individuals so the equipment of the building block with green spaces was not surveyed. The construction material was part of the questionnaire but did not show statistical significance in the analysis of the results (Reiter 2009). A reason for that might be that the construction material is rather heterogeneous amongst the interviewees so

168

9 Flood risk analysis and assessment

there was no direct link between the general damage that the affected people suffered and the individual building qualities. It has to be recalled though that the availability of private or public structural mitigation measures should be rated important, too, but can for data availability reasons not be included in this study. 9.4.4 Sensitivity analysis for the weighting Figures A69 and A70 in the appendix showed that the vulnerability maps change with changing indicators. It still needs to be determined how sensitive the applied weights are. Therefore, a sensitivity analysis was carried out for the weights of the three most relevant indicators based on the evaluation of the experts. The weights of the indicators “Proportion of buildings located at or below street level per building block” (Weight 1), “Proportion of buildings with poor construction material per building block” (Weight 2), and “Proportion of green spaces per building block” (Weight 3) were altered and the resulting vulnerability scores were analyzed. The weights were altered according to the weighting options in the questionnaire: 1.0, 0.75, and 0.5 for the two first indicators, 1.0, 0.75, 0.5, 0.25, and 0 for the last indicator. These values cover the answers given by the experts in the questionnaire and are therefore regarded as being plausible. The resulting vulnerability indices were then analyzed statistically using a simple correlation analysis. Table 33 displays the results. The first column indicates the name for the weight combination, columns 2, 3, and 4 show the respective weights for the three indicators, and the last column indicates the correlation coefficient with the original values, i.e. the average values that were delineated from the expert questionnaires and that were then used for the vulnerability analysis. Table 33: Results from the sensitivity analysis: The weights for the vulnerability indicators and the resulting correlation coefficient r² with the vulnerability scores applying the original set of weights. Alternative

Weight 1

Weight 2

Weight 3

r²

Original 1A 1B 2A

1.0 0.75 0.5 1.0

0.75 0.75 0.75 1.0

0.7 0.7 0.7 0.7

1.00 0.99 0.96 1.00

2B

1.0

0.5

0.7

1.00

3A 3B 3C 3D

1.0 1.0 1.0 1.0

0.75 0.75 0.75 0.75

1.0 0.5 0.25 0.0

0.99 0.99 0.91 0.72

The results of this analysis proof that the correlation between the resulting vulnerability scores is very high, except of the alternative 3D, where the indicator

169

9.5 Comprehensive risk analysis

weight was set to zero. The absolute values, however, differ. That is not very relevant for the vulnerability analysis, as the differences between the regions remain stable. For the final risk analysis this becomes relevant though. The maximum vulnerability value is in the original calculation 5.47, it is 5.75 in alternative 2B (highest), and 3.51 in alternative 3D (lowest). It depends on the weights of the two other components hazard and elements at risk that influence the final risk values. To conclude, the results of the analysis show that the chosen weights do not play a significant role for the description of the relative vulnerabilities between the building blocks. They rather influence the final risk score. This influence is to a certain degree minimized as the three components are all indexed (hazard, elements at risk, and vulnerability) and are multiplied by 10 before they are combined to the risk index. For this analysis the average values from the expert surveys are used as that seems to be the most plausible solution. For future analysis and for the usage of the information for practitioners however, an interactive webbased GIS application was developed based on the practical vulnerability assessment used in this research that allows for a more customized evaluation of vulnerability according to the priorities of each decision maker (Ebert [Müller] & Müller 2010b, Ebert [Müller] & Müller 2010a). 9.5 COMPREHENSIVE RISK ANALYSIS The risk which is defined here as the the location-specific damage potential dependent on the flood hazard level, the amount of elements at risk, and their vulnerability is calculated after Equation 15: ܴ௠௔௡௭ ൌ ‫ܫܴܽܧ כ ܫܸ כ ܫܪ‬

Equation 15

with Rmanz as the flood risk level in a building block, HI as the hazard index for the building block, VI as the vulnerability index per building block, and EaR-I as the index for the elements at risk per building block.

The value range for this calculation is 0 to 1,000. As values with such a wide range are very hard to interpret and understand, Equation 16 was applied to again normalize the risk values and to obtain a risk index with a maximum value of 10. ܴ‫ ܫ‬ൌ ͳͲ ‫כ‬

ோ೘ೌ೙೥ ோ೘ೌೣ

Equation 16

with RI as the flood risk index between 0 and 10, Rmanz as the total risk for the building block, and Rmax as the maximum flood risk per building block in the area.

As discussed before, the hazard map anchored in the PRMS was used for the risk map generation as this is still the official decision making base and as it covers the entire study area. With respect to the elements at risk, one single map containing the Elementsat-Risk-Index was generated and used for all risk maps. The Vulnerability-Index based on the evaluation of the experts was applied.

170

9 Flood risk analysis and assessment

The first final risk map shown in Figure A71 in the appendix describes much more precisely the flood risk distribution in the study area than the official maps of the PRMS which only consider the hazard level. Detailed information on the amount of affected people and infrastructure and their vulnerability can help decision makers to better focus investments to reduce the flood risk. While the maps in the PRMS only show large polygons with homogeneous flood hazard values, the flood risk map provides detailed information on the level of a building block. It becomes obvious that especially those areas that show a high number of elements at risk face a higher risk than areas that are less densely populated. Clear differences become visible in all hazard zones, for example in the central western part of the study area where a homogeneous hazard zone now turned into a heterogeneous risk zone that is determined by the number of potentially affected people and infrastructure and their vulnerability. Also the large hazard zone resulting from the San Ramón channel contains now more detailed information. However, as the example in the southwestern part of the study area (1 km north of the southwestern tip) shows, entire building blocks become classified at highly risky even though only a fraction of them is affected by floods. This is the drawback of the applied methodology and level of detail. The large advantage of this method, however, is that the generated maps are dynamic and interactive and that they can be updated as soon as new or more detailed information becomes available. Using the query function of the corresponding GIS data base allows obtaining detailed information on the variables leading to the generation of flood risk in the specific place. Making use of WebGIS technologies allows for making the risk maps available on the Internet (Ebert [Müller] & Müller 2010b). An alternative risk maps with updated flood hazard information was generated for a subset of the study area using the hazard map after Perez (2009). Figure A72 contains the flood hazard with a 10 year return period for the current climate conditions. Comparing the two risk maps shows quite comparable risk index values for most of the building blocks. However, as the information provided by Perez (2009) shows a significantly higher spatial resolution with more detailed information, some of the building blocks that have medium risk values when using the data of the PRMS do not face any flood risk after the calculations of Perez (2009). Further differences can be observed in the northern parts of the study area: The areas that face a high level of risk using the hazard map in the PRMS do now face a lower level of risk. That proofs the added value of all updated and more detailed information on the study area.

9.6 The development and assessment of flood risk

171

9.6 THE DEVELOPMENT AND ASSESSMENT OF FLOOD RISK The goal of this last section of the chapter is to assess the current risk levels in the entire AMS and to also provide some estimation on how the risk conditions could develop in the future. To provide an overview about the current status two indicators are chosen that specifically illustrate the large amount of people and buildings affected by floods. 9.6.1 Number of new residential sites in flood-prone areas The first indicator refers to the amount of buildings located in areas facing a high flood hazard level according to the PRMS. Areas facing a high flood hazard level are defined as regions that are flooded at least once every two years. These regions contain construction restrictions according to the PRMS. However, special regulations apply if it can be proven that the specific site is protected against floods. Thus, even though it is known that the respective areas are subject to regular floods, a number of new residential sites evolved there during the last decades. The proportion of residential sites and infrastructure constructed in areas per municipality facing a high flood hazard level for the period between 1993 and 2002 is shown in Figure A74 in the appendix. The development of new construction in flood-prone areas between 2002 and 2009 shows Figure A75 in the appendix. The data about construction activities were derived based on an existing LULC map from the analysis of ASTER satellite data for the large parts of the AMS and adjacent municipalities of the RM (Vásquez et al. 2010). It has to be noted that the spatial extent of the available remote sensing data equals the area of the red rectangle in Figures A74 and A75 in the appendix. All calculations do consequently only refer to the area covered by the remote sensing data. This is in some cases not the entire municipality but only parts of it, e.g. Lampa and Colina in the northern part of the city. For a better orientation the figures show the urban builtup area in grey color and the newly constructed residential sites during the respective periods in purple color as a background map. The areas facing a high flood hazard level according to the PRMS are drawn in blue. The columns in every municipality indicate the amount of buildings constructed inside (red) and outside (grey) the flood-prone areas. New developments of residential, commercial or industrial usage in floodprone areas exemplify the clearly negative tendencies of elements at risk. As a quantitative indicator it should certainly have the target value “0” but does in fact exceed this value in almost all municipalities (compare numbers in Figure A74 and Figure A75). While the development of new residential sites in flood-prone areas is comparably low in the central municipalities, it used to be high in the areas of the Andean piedmont in general, but consolidated recently as physically only little room for further expansion is left. The growth is alarmingly high in those municipalities that are predicted to experience extensive growth in the near future. Even though local mitigation measures such as dams along channels need to be constructed in

172

9 Flood risk analysis and assessment

newly developed areas if the land is prone to flooding, a system-oriented complex impact assessment of the construction activities on the local water balance and an update of flood hazard maps are not undertaken. In practice that means that the areas located at lower elevations than the newly established buildings are subject to an increasing flood hazard as the overall infiltration capacities in each subbasin decline. That proofs that the ongoing activities are clearly not sustainable. That furthermore proofs that the number of elements at risk in combination with the flood hazard is the main determinant of an increasing flood risk in the future. That the level of vulnerability in these regions can be reduced to zero is not considered being realistic when looking at the current conditions. 9.6.2 Number of people living in flood-prone areas A second indicator refers to the amount of people currently living in highly floodprone areas. Figure 28 shows that the values exceed 20 % in six municipalities, amongst others in La Reina. Even though the highly dynamic municipalities in the northern part of the city such as Lampa and Colina are not included in this map as they do not belong to the AMS and respective census data are thus not available, the comparison with the findings from the previous indicator make it easy to estimate the high number of people living in flood-prone areas, also in these municipalities. Even though current structural mitigation measure might be able to protect people and values from floods today that might change in the future, for example through a changing land use or a changing climate.

Figure 28: Overview about the amount of people in the AMS living in areas facing a high flood hazard in 2002.

9.6 The development and assessment of flood risk

173

9.6.3 Development of the components of risk in the future The projections for the hazard development are clearly negative, both due to climate change and as a result of a physically changing land-use pattern. The trend of constructing new buildings for residential and commercial use at the Andean foothills is continuing because of profitable land prices. An increasing demand for personally-owned real estate through inner-urban migration, preferably into gated communities or into newly constructed high-rise buildings equipped with private amenities such as swimming pools and gymnasiums, additionally fosters urban sprawl and urban densification at the same time. Besides that, the possible climate-change related loss of vegetation leads to a further decrease of retention areas and less infiltration capacities. As a result, the surface runoff and therewith the flood hazard increases. This effect is most likely being worsened by the projected increase of the maximum precipitation values. Possible climate-related direct changes of the flood hazard zones are further discussed by Perez (2009) and are depicted for example in Figure A64. Figure A73 provides another example of the possible impact of climate change (increasing precipitation intensity) on the flood hazard. Even though this map does only show the in-depth study area and does not contain spatially explicit projections of changes in the urban morphology, the predictions are to a large part transferable to the rest of the RM. It becomes clear that the worst case IPCC scenarios would result in a dramatic increase in the hazard and consequently the risk. As mentioned before, the hazard will be further increased with a loss of vegetation both as a result of a dryer climate and through continuous sealing of the surface. Together with the current development of the elements at risk that leads to an overall increase of risk that is rated considerably high. It was just indicated that completely updated risk maps based on a changing land use cannot yet be provided. To provide these updated maps, spatially explicit information about the future distribution of elements at risk, their vulnerability, and the updated flood extent (considering new constructions and thus flow barriers) would be required as an input. The hazard maps generate so far include changing runoff volumes but do not included possible physical changes of the urban morphology and still have methodological drawbacks that limit their accuracy. More reliable hazard maps can at this point not be provided as they would require a very high resolution digital surface model and more detailed hydraulic calculations that lay beyond the scope of this research. The hazard maps provided by AC Ingenieros (2008) show points of overflow, but do not show the extent of the potentially flooded area. Thus, while the impact of a changing land-use pattern can be quantified using a hydrological model, at this point only qualitative estimations about the future development of the overall risk can be given. As the current land-use planning instruments, mainly the PRMS, leaves backdoors to circumvent the regulations that forbid the development of construction in flood prone areas the number of elements at risk and therewith the overall risk is going to further increase in the future. This is going to get even worse if the cota mil is lifted and construction above 1,000 m becomes possible without restrictions.

174

9 Flood risk analysis and assessment

In practice, positive changes in the hazard-related variables and in the variables relevant for the elements at risk are for several  but mainly financial  reasons not expected. The only component of risk that might realistically experience a neutral or positive development is thus the vulnerability. The only efforts carried out so far are structural measures in certain parts of the city to reduce exposure. However, it was explained before that the money that would be needed to install sufficient storm water infrastructure in all regions of the city exceeds the available financial resources by orders of magnitude. Awareness raising campaigns and the support of private flood protection and mitigation measures might take place if the damage after floods exceeds a certain value. Figure 16 in Chapter 5 showed which main variables determine the flood risk generation in Santiago de Chile. All of them are embedded in an institutional planning framework, governance structures and a societal value system with certain preferences that influence the decision making with respect to urban development. An adapted version of Figure 16 shows which risk-influencing parameters apart from the framing factors can in principle be altered through human influence on regional, local or individual level (Figure29). The amount of variables that cannot be influenced is small, again indicating the significant contribution of human activities to the generation of flood risk. Land use and land cover, the construction of adequate storm water evacuation infrastructure, the type and location of buildings, knowledge and preparedness for floods have to be named as influential factors in a first place. As pointed out before, also the governance structures and societal values are very important in this context as they determine all other human-influenced variables. Some hazards, above all geological hazards, can barely be influenced by anthropogenic activity: earthquakes, volcano eruptions or tsunamis as secondary hazards after earthquakes. That means that they are justifiably called “natural” events. It was shown that the occurrence of floods as an example for hydrometeorological hazards however can on various points be influenced by human interaction. Thus it does not seem justifiable to call them “natural hazards”, they are rather natural events that are being turned into man-made hazards. This terminology (riesgos naturales) however is still used by the MINVU, amongst other institutions, indicating the lack of understanding for the generation of flood risk. Reflecting on the previous deficits and using exploratory scenario techniques again, the following section shows what could happen if some of the variables that influence the flood risk are altered. It needs to be questioned where exactly the deficits are to identify starting points of possible improvement and to develop measures to minimize these negative tendencies.

9.6 The development and assessment of flood risk

175

Figure 29: Schema of variables relevant for flood risk analysis in Santiago de Chile. Those variables printed in bold are the ones that can directly influenced by human activities.

10 PREVENTION AND MITIGATION MEASURES Chapter 10 first summarizes the current deficits with respect to flood prevention and mitigation in Santiago de Chile that can be observed or reconstructed today (Section 10.1). That forms a basis for the identification of starting points for improvements. Subsequently, Section 10.2 refers back to the scenarios developed in Chapter 8. While Chapter 8 only referred to the impact of a changing landuse/land-cover pattern, Section 10.2 describes how possible future developments with respect to relevant institutions and instruments, flood prevention, mitigation, and the actual flood event could look like in more detail. The descriptions are using the same three scenario alternatives and their general assumptions and it becomes obvious how the potential impacts of certain activities would affect the flood risk. Drawing on the vivid illustrations about future development alternatives, Section 10.3 provides specific recommendations to prevent and mitigate the flood risk in the study area. Section 10.4 finally points out how and when the respective recommendations should best be put into practice. 10.1 ANALYSIS OF THE PREVIOUS DEFICITS The main points named by stakeholders with respect to the current deficits that lead to flood risk are the decentralized governance structures and  along with that  the lack of communication between the institutions and the lack of coordinated planning. As a result of the existing communication structure there is only little or no exchange between decision makers about common topics. Besides, responsibilities are not clearly defined, for example in the field of maintenance of the storm water infrastructure and green spaces. What is lacking further is a common vision about the development of the city. But even if there was a common vision: with the deficits named before it would be challenging to put it into practice with the existing structures. Another deficit is the main planning instrument, the PRMS. During its creation in 1994, three direct objectives were defined to assure the best possible living conditions (Metz & Weiland 2009): -

the preservation of agricultural areas, the densification of the existing settlement (vs. urban sprawl), and the conservation of an urban limit while allowing construction within the limits in another 59,330 ha with a buffer area potentially open for construction of 13,000 ha around the limits.

These limitations were meant to be valid until 2020 and  through the planned densification of the urban settlement of up to 150 inhabitants/ha  meant to enable the accommodation of the expected 8.7 million inhabitants in 2020 (Petermann 2006). The compact urban city with a preservation of the existing agricultural land

178

10 Prevention and mitigation measures

as retention areas around the urban fringe would have limited the increase of flood risk. However, in 1997 construction activities outside the normative urban limit became possible in private garden land with a maximum size of 5,000 ha (parcelas de agrado). In 2003, the MINVU SEREMI made it possible to convert agriculturally used areas of a size of at least 300 ha to residential areas if also educational and commercial infrastructure were set up on that area in order to avoid an increase of traffic towards the city center (Reyes 2003). High standard residential areas where the investors could afford constructing the required infrastructure evolved along the fringe of the city (gated communities). However, there is no control organ that ensures the availability of these infrastructure facilities. Thus, the traffic is not necessarily reduced. Therewith, one of the original goals of the PRMS, the conservation of agriculturally used land, vanished. As a result and another proof of failure of the PRMS, the density of the mancha urbana decreased from 96.5 inhabitants/ha in 1992 to 85.1 inhabitants/ha in 2002 (Petermann 2006). Other deficits include that neither roads nor green spaces (especially in the poor areas) were constructed in a structured and sufficient manner and that the subcenters as designed in the PRMS have not been consolidated yet (Reyes 2003). Today, construction is possible in flood prone areas if certain conditions apply. In fact, real estate traders earn large profits by buying flood-prone land at very low prices, installing local structural measures to protect the lot of interest, and to then sell the land at higher prices. One example from the study area is the construction of social housing units in former flood-prone areas in the municipality of Peñalolén on Avenida Antupiren. While the financial benefit from these areas is comparably low, the benefit gets higher with a rising quality of the buildings. Buildings of various types are constructed in formerly flood prone areas as exemplified in the work of Carvacho (2010). Even though it is foreseen by law that compensatory retention areas are created this is not controlled by any instance. That means that it does in practice not happen (Carvacho 2010). In addition, there is no comprehensive assessment of flood risk carried out if new construction projects are planned or after they are finished. The analysis of flood risk is rather carried out pointwise and does not consider the complexity of the processes that are triggered in the ecosystem of a catchment. Even if the flood risk is minimized for a certain lot, the new sealing of surfaces reduces the infiltration capacities in the basin and thus only transfers the hazard to other people or values that did previously not face any hazard. The negative development of constructing in flood-prone areas, however, started earlier with the construction of roads in the river beds during the last centuries (compare Section 4.1.3). This ignorance of water ways and flood risk in urban planning was intensified during the last century. The existing storm water infrastructure is not sufficiently maintained. That means that the infrastructure is clogged by sediments and organic material in numerous cases (Figure 30). In addition, the river network can be blocked by solid waste, such as old electrical equipment which is thrown in rivers or dry river beds before winter precipitation starts (Figure 30). That leads to a congestion of the water ways, which can also cause floods.

10.1 Analysis of the previous deficits

Congestion of the drains by organic material.

179

Congestion of the creeks by solid waste.

Figure 30: Congestion of the storm water infrastructure and creeks by organic material and solid waste in a the municipality of Peñalolén.

This even aggravates the fact that a range of channels in the RM are not wide enough to accommodate the large amounts of storm water after extreme events. It becomes clear that flood risk in Santiago de Chile is largely determined by anthropogenic disturbances of the urban and peri-urban ecosystem and that it is then combined with a  voluntarily or not  self-exposition in hazard prone areas. Even though floods are part of the natural water cycle, the urban development in the study area reinforced the negative impacts of the regular occurring floods at different stages. Summing up, the most important deficits with respect to flood risk prevention stage are: -

Lack of vertical and horizontal collaboration of relevant planning authorities in the field of land-use planning No sufficient consideration of flood risk in planning instruments and regulations and no controlling body Unclear responsibilities and competencies Insufficiently comprehensive urban planning, i.e. pointwise SEIAs Sealing of catchment areas and loss of infiltration capacities Conversion of river beds into streets, i.e. sealing of natural water courses Insufficient maintenance and conservation of retention areas Development of residential sites in flood-prone areas and consequent exposure of people to flood hazard

180

10 Prevention and mitigation measures

The most important deficits in the field of flood mitigation that can be observed or reconstructed today are: -

Insufficient width of constructed channels Insufficient maintenance of existing storm water infrastructure Congestion of channels through waste Insufficient availability and capacities of storm water infrastructure 10.2 FLOOD RISK SCENARIOS

Based on the previous findings, i.e. the accentuation of relevant variables (Section 9.6) and the inclusion of the deficits (Section 10.1), the following section describes possible consequences of the efforts that are taken  or not  to prevent and/or mitigate the negative impacts of floods. Explorative scenarios are again used to draw an image of possible futures. The descriptions depict possible developments that are based on the current governance structures, planning instruments, and attitudes as well as the current measures for flood risk prevention and mitigation. They then show possible changes. The storylines of the scenarios are inspired from discussion results of a workshop with stakeholders (Reyes & Ebert [Müller] 2009) or with experts interviewed during field stays. The storylines match the assumptions of the LULC scenarios from Section 8.4.2. 10.2.1 Scenario I  Increasing aridity Scenario I assumes a climatic development as projected by the IPCC scenarios. That means that the climate is getting dryer over time with less annual precipitation and more extreme events. It reflects the currently most likely development (business as usual) both in terms of land-use development and climate change. The efforts with regard to flood prevention and mitigation activities remain rather limited. The land-use management system is characterized by superposition of different levels of administration but there is no common vision about the development of the city, which prevents from a synchronized and target-driven development. Another hindering factor is the lack of coordination and collaboration between the different responsible institutions on a horizontal and on a vertical level. Different ministries, such as the MINVU and the MOP-DOH that do work in overlapping areas have a negligible amount of interaction and almost no direct contact. The involvement of the municipalities in decision making is minimal as they do only follow the planning decisions predetermined by the ministries (national level). The involvement of municipalities is reduced to the duty of delivering information and to retrieve opinions, but citizens cannot influence the decision-making process. The Metropolitan Regulatory Plan is published through the MINVU but it does not have to be approved by the municipalities, thus there is no public debate about it.

10.2 Flood risk scenarios

181

10.2.1.1 Flood prevention As the ecological awareness is not high enough to allow for a sustainable urban planning, land-use and land-cover patterns are not optimized in terms of flood prevention. The infiltration capacities of the soil decrease on the one hand as a result of less vegetation coverage due to a changing climate and on the other hand with the loss of agriculturally used or natural areas in the scope of urban expansion. Thus, more direct surface runoff is generated after extreme precipitation events. Land that is already equipped with sufficient infrastructure and possesses a good connection to the city center remains a scarce resource. Thus, new residential sites do still evolve in hazard prone areas. This process is partly also caused by the lack of knowledge about hazards and their ignorance throughout the dry season. Risk management aims at reducing damage from disasters and identifying risk areas but no risk analyses are being performed in a sufficient manner. Also, results from existing studies are not sufficiently shared with the respective municipalities. 10.2.1.2 Flood mitigation Efforts are done especially by the MOP-DOH to reduce the hazard by increasing the capacity of channels or by elevating or removing bridges. As only limited financial resources are available, these efforts do predominantly comprise small investments in different poor spots in the RM. No comprehensive measures are taken to mitigate floods from official sites. The part of the population that has previously been affected by floods though takes private measures, such as the construction of a higher foundation to put the house upon. As soon as the building is located higher than the street level, the exposure is significantly reduced and the probability to experience damage is little. Other private measures comprise the construction of low walls around the property. The possibility to take such measures and to choose an appropriate construction material for the house does mainly depend on the financial possibilities. Persons belonging to the group of low social status are disadvantaged in this regard as they cannot afford construction material with low physical fragility. Instead they use material such as cardboard, wood or clay where the humidity remains in the walls and causes long-term damage. 10.2.1.3 Flood scenario The amount of runoff is going to increase with increasing construction activities and the loss of natural retention areas. That will also result in increasing water heights during flood events. Nevertheless, some mitigation measures are installed by the MOP-DOH in the very poor or very frequently affected areas to reduce the damage. Public institutions such as fire fighters, municipal workers, and sponta-

182

10 Prevention and mitigation measures

neous neighborhood alliances help to remove the damage after rainfall. The monetary damage after floods is lower in low income settlements as there are less values exposed. The damage in rich areas is rather non-monetary with people being trapped behind the walls of gated communities. Nevertheless, the potential amount of people exposed is higher in low income areas as the population density is higher there. In general, these new residential sites that substitute natural retention areas and ecologically valuable zones do not suffer much from floods, rather the poor settlements in flood-prone areas throughout the city and on a larger scale in the lower part of the city suffer (increasingly) after extreme precipitation. 10.2.2 Scenario II  Afforestation Scenario II assumes that technical equipment is installed that fosters the development of green spaces and vegetation in all parts of the city. The importance of vegetation and urban water courses is recognized in urban planning. This scenario assumes more ecological awareness and the willingness to invest in prevention and mitigation measures prior to extreme events. Fundamental structural changes and rethinking would be required to arrive at these conditions. In Scenario II, environmental protection, social justice, and sustainable development are the determining factors in decision making processes with respect to urban planning. This new orientation leads to a stronger collaboration of institutions and the creation of new planning instruments to improve the environmental and living conditions. The responsibilities and working processes of regional and local governments become transparent and complementary to each other. The planning decisions are taken on multiple levels, beginning with national tendencies provided through the ministries. On a lower level, the regional development strategy including strategic planning on regional level (RM) based on joint decisions of GORE, private actors and the civil society leads to a moderate urban development. Local planners depend on the next higher level, the regional government and the Intendente (mayor) and have the chance to put adapted development goals into practice. Compensation payments are made by the government to encourage municipalities with ecologically valuable areas to protect and maintain them. Instead of a further spatial expansion of the city, central areas are being renovated and make lucrative spots to live for all levels of the society. There is no longer a severe scarcity of space that enforces settlements in hazard prone areas. 10.2.2.1 Flood prevention The possibilities to influence the probability of a flood event through measures of urban planning have been recognized by the respective institutions. Economic interests no longer superimpose attempts to conserve the ecological balance. Instead, the model of the compact city with as little expansion towards the Andean mountains and the North of Santiago as possible results in a protection of natural

10.2 Flood risk scenarios

183

and agriculturally used areas and therefore minimizes the flood hazard. Natural retention areas are preserved, artificial retention areas are created along the shores of the rivers and along the dams of the channels. Settlements with a sufficient amount of green spaces to provide space for recreation, infiltration capacities, and cooling effects can be found throughout the city. 10.2.2.2 Flood mitigation In those parts where land use and land cover cannot be adapted adequately, the MOP-DOH receives enough money and man-power to adopt constructions or build new measures to minimize the flood hazard. The capacity of channels such as the Canal San Ramón is enlarged to be prepared for the current and for rising runoff values. The communal and regional governments identify hot spots in each municipality where floods can cause damage and provide money where needed to improve the local conditions. Investments comprise public measures such as removable barriers on streets and private protection measures yielding at the reduction of the physical fragility of a building. 10.2.2.3 Flood scenario The general probability of a flood to happen is reduced through activities in flood prevention and flood mitigation. Nevertheless, the hazard of floods still exists as not all causes (such as the construction of main roads in former river beds) can be undone. Through an increasing knowledge of how and when floods occur and what can be done to minimize their negative impact, the damage that is caused by floods is reduced significantly throughout the city. Neighborly help is the main ideology to commonly repair the damage caused by floods. 10.2.3 Scenario III  Construction activities Scenario III assumes an increase of urban expansion towards all possible directions. Financial benefits clearly dominate the environmental conscience in urban and regional planning processes. This scenario assumes an impairment of the existing structures and attitude. A decent sense of environmental awareness got wiped away by the chance to make money selling land and real estate. Conservation restrictions are loosened in order to encourage and ease construction activities. Planning instruments on national, regional or communal level exist but do not have the power or validity to prevent negative impacts on the ecological environment. The main goal in urban planning is to please the wishes of the rich, i.e. large lot sizes in suburban regions with better air quality, good connection to the city center, and a fair equipment of infrastructure. At the same time, the existing built-up areas are further densified resulting from the construction of new low- and medium-class settlements. With

184

10 Prevention and mitigation measures

the legally defined public planning framework and the private institutions which are gaining more and more importance the amount of actors is increasing tremendously. As individual responsibilities are hard to be identified, the lack of synchronization and coordination grows further. Planning and development decisions are steered on a national level through the representatives of the ministries. The governance structure literally prevents adapted local and regional planning as the local and regional institutions directly depend on the ministries and not on a local or regional decision maker such as the Intendente (mayor) or Regional Government. The influence of the private sector is legally not clearly defined and thus further hinders the understanding of decision making procedures. 10.2.3.1 Flood prevention The relation between land use/land cover and flood hazard is not recognized. Urban planners are mainly architects with close relations to the economy and no or just little education on environmental issues. The potentials of possible prevention measures such as the conservation of natural retention areas are not identified. Instead, the options to prevent flood risk are ignored and superimposed by economic benefits of a small part of the population. 10.2.3.2 Flood mitigation The need to construct flood mitigation measures is only perceived in areas where rich people live, as high values are exposed and the willingness and possibilities to pay for such measures are much higher in these regions. Governmental money for the construction of public and private measures is reduced to a minimum. The distribution of this money does not leave the chance to significantly reduce the problem of floods. Especially with respect to construction material of houses there is almost no chance for the urban poor to improve their homes before or also after adverse impacts of a flood. Private green spaces remain in areas where their maintenance can be paid for, i.e. in the rich areas. In the poor areas of the city, green spaces become home for the very poor and spaces with rising delinquency rates or are misused as waste dumps. 10.2.3.3 Flood scenario Increasing losses of retention areas through an increasing amount of sealed surface in all parts of the city and to a large part also in the peri-urban areas lead to changing infiltration patterns. Less infiltration capacities result in higher amounts of surface runoff from the mountainous surrounding that is being collected in current and ancient river beds and channels. While the newly constructed rich areas do have stormwater infrastructure to protect their own houses, poor areas do not have sufficient infrastructure available. As a result, the area where physical dam-

10.3 Recommendations

185

age is caused clearly correlates with the social status of a neighborhood. After flood events, the households are responsible for their own debris and damage removal. The problem negative impact and damage is much higher in areas where capacities to recover are low. As space becomes more and more a scarce and also an expensive resource, poor people are forced to settle or stay in hazard-prone areas. Spontaneous neighborly help is being created after flood events in affected neighborhoods. 10.3 RECOMMENDATIONS The following two sections now outline a number of possibilities and starting points to minimize the existing problems with a focus on long-term (prevention) and medium-term (mitigation) measures. Longer-term flood risk prevention measures that from their physical and social root causes avoid risks are in general related to land-use planning. Short-term mitigation measures comprise structural measures and awareness raising campaigns that aim at minimizing the risk and the negative impact of the hazard. 10.3.1 Prevention measures What is needed most is a more comprehensive urban planning that includes (i) ecosystem processes such as the processes of the water cycle, (ii) temporal changes of these processes, e.g. after changes of the ecosystem, and (iii) people and their vulnerabilities. Planning decisions need to be made considering the entire city or an entire ecological unit, for example a watershed or a subbasin. Pointwise SEIAs most no longer be carried out. Rather models should be used to obtain knowledge about the impacts of changes in one part of the watershed on the other parts of the same watershed and adjacent regions. Ashley et al. (2007) developed strategies for an urban drainage system with respect to climate change and urban land-use changes and the central finding is that the “key to capacity building is stakeholder understanding of the problems, roles and assumption and distribution of responsibilities” (p. 80). To foster that, the communication structure needs to be improved. In general, it is important  though difficult to realize  to ease the access to information to broadly disseminate knowledge. Highly relevant is a better networking amongst stakeholders as that strongly influences decision making. Platforms for dialogues between regional and communal policies could for example be round tables and discussion groups where current topics are explored and discussed with people working in different sectors and different administrative levels. Exchanges with experts from other fields, i.e. between architects, engineers, geographers, soil scientists, hydrologists, social or political scientists would additionally enforce a system-oriented thinking. Therewith a more comprehensive planning could be established as a variety of relevant aspects, for example in the field of flood risk management, would become known. Bringing relevant decision makers together

186

10 Prevention and mitigation measures

would most likely also bring more light into the problem of unknown responsibilities. One example with respect to flood risk is that the responsibilities for the maintenance of the storm water evacuation structure need to be clearly defined. Another example is green spaces. Green spaces of all types (also agriculturally used areas) need to be maintained as retention areas. A minimum amount of green spaces should be defined for each municipality or even for each census district to further improve the environmental value of the city. Municipalities that maintain their green spaces in favor of construction should receive compensation payments from the state to further encourage the ecological awareness and sustainable development. So far, different institutions are responsible (or not) for different types of green spaces. Especially this field needs more coordination and very well defined responsibilities, and in addition to that: an economic incentive. With regard to urban growth, the city needs a common vision about its development. The compact city model should be the preferred development goal. A second option is a more equal equipment of different urban areas with infrastructure to minimize the need for road construction and transport. The ongoing construction activities towards the periphery without an equal provision of infrastructure throughout the city have a severe negative impact on the environment and therewith need to be stopped. A measure is also the renovation of central areas of the city to reduce the sprawl. Furthermore, water courses should be conserved in the scope of urban development and recognized as one valuable component of the urban space. It could be of value if the municipalities become more involved in urban planning, e.g. that they need to approve the PRMS and/or that they are clearly made responsible for the maintenance of their storm water infrastructures. More participation yields a better identification with the planning process and therewith more identification with the existing issues. The government structures need a single person or a specific institution that is responsible for the coordination of the large amount of actors and for the control of the compliance with restrictions. If that cannot be provided the further development of the city will be inefficient and will miss many chances and positive effects of mega-urbanization. Regulatory stipulations as for example the creation of compensatory retention areas in the case of constructing in flood-prone areas will never be followed. Synchronization, coordination and better control over different levels and across all sectors is of utmost importance and should clearly be the focus of restructuring the existing governance system. It lays beyond the scope of this research to give specific recommendations on how the governance network should be re-structured precisely. However, it is suggested that the governments on the various administrative levels should at least be made more use of (i) its existing possibilities and (ii) the methodological framework for flood risk analysis and assessment presented in this thesis to obtain more comprehensive information on the subject. What is required by the European directive, i.e. a large-scale flood risk zoning, should also be available for the study area. The governmental institutions have the possibility to employ private consulting companies, for example for carrying out SEIAs. It should be made use of the same companies to request an updated flood risk assessment for the entire RM.

10.3 Recommendations

187

Figure 31 displays how strategic urban planning would decrease the hazard by leaving natural retention areas and river beds and by eliminating the elements at risk, which in the case of buildings also negatively function as flow barriers. If strictly followed in the most appropriate way, land-use planning is oriented on avoiding flood risk and would be capable of entirely prohibiting settling down in hazard prone areas. Looking at Figure 31 and the concept of risk applied in this research it becomes clear that with no elements at risk and an additionally minimized hazard, the risk of suffering damage from flood events is non-existent. Even if such a sustainable and flood-risk focused land-use planning is highly unrealistic, the potential of appropriate land use for flood hazard reduction becomes obvious. At a minimum, the flood hazard should be re-assessed with every new construction activity or new zonation within the RM. In addition to that it is recommended to carry out a new complete risk assessment whenever new census data, hazard maps, cadastral data about elements at risk or any type of information that would lead to an improvement of the existing maps become available. Furthermore, the risk should be reassessed after taking risk reduction measures. Comprehensive risk assessments would improve the knowledge about flood generation at an early stage  for decision makers and affected people likewise. The presented approach for the analysis and assessment of flood risk is most likely transferable to the rest of the RM and can also be applied at different points in time. It is based on data that are to a large part free of costs or that can be substituted by low-cost data. For example, the Quickbird satellite data that are used for the derivation of green spaces could be substituted by ASTER data that are much more economic. The building mask that was also delineated from the very high-resolution Quickbird data is most likely also available from an official cadaster. The GIS-tool that was developed in the scope of this research (Ebert [Müller] & Müller 2010b, Ebert [Müller] & Müller 2010a) is one method to communicate research results, to disseminate knowledge, and to allow for an exploration of the complex generation of flood risk in the study area. As another measure, results from existing studies should (i) be made available and (ii) be incorporated in planning decisions. Very site-specific recommendations with respect to the study area for this research would be to foster afforestation activities or the development of green spaces of any type. Especially important are the north-facing subbasin 84 in the lower part of the basin and the westernmost subbasin 72 (compare Section 8.3.4). If the amount of green spaces in the basin further decreases the flood hazard in the adjacent municipalities will increase. This is also valid for the neighboring subbasins along the Andean foothills that face the additional problem of not being ecologically protected. It was shown in Section 8.4 though that a limited amount of urban expansion towards the Andean mountains would decrease the urban flood hazard in comparison to the dry scenario if a large amount of green spaces is included. In any way it needs to be ensured that the existing vegetation coverage along the Andean foothills does not decrease any further.

188

10 Prevention and mitigation measures

Figure 31: Schema of variables relevant for flood risk analysis in Santiago de Chile. Variables printed in bold black and surrounded by a box are points where prevention measures (land-use planning) can be taken to reduce both hazard and elements at risk an d elements at risk and therewith the overall risk of suffering damage from flood event.

10.3 Recommendations

189

10.3.2 Mitigation measures When looking at the mitigation stage, which comprises short-term measures to not avert a hazard but to minimize its damage potential, Figure 32 highlights the points where action can be taken. If the capacity of the water ways is enlarged, if flow barriers are constructed (e.g. walls, dams) or if private protection measures (e.g. backflow flaps) are installed, the hazard can be better controlled and the exposure as part of the vulnerability is minimized. Nevertheless, it has to be considered, that structural measures to evacuate or redirect the storm water runoff need to be designed and constructed in a way that the entire city is better protected. Also, it needs to be pointed out that these measures are associated with large investments from both governmental and individual actors and are thus in practice hard to be implemented to a sufficient degree. The MOP-DOH has an investment volume of around 10 million US$ annually. However, 1,000 million US$ are needed to do all construction works necessary to mitigate the flood risk in the AMS (Retamal & Estellé 2009). Analyzing the results from the vulnerability analysis reveals that the investment in better construction material and physical mitigation measures in buildings would improve the situation (e.g. buildings that are located in highly flood-prone areas can be elevated). Field surveys showed that experience with floods frequently decreases vulnerability as people are better prepared. Focused information campaigns in the potentially affected areas, raising awareness, and generation and dissemination of knowledge could also reduce the vulnerability. These information campaigns should include an explanation of the system behavior and address the questions why do floods occur and what can every individual do to make it clear that floods are not a natural but a man-made phenomenon. Public financial support for the installation of private mitigation measures could be provided to households which financial means are insufficient to install appropriate structural flood protection. To be prepared for a flood event, removable barriers could be provided in addition to the sandbags. Along with that a plan to coordinately evacuate superficial storm water should be established for areas in which subsurface systems are not available to a sufficient degree. In a publication of the Centro de Aguas Urbanas of the Catholic University of Chile from 2003 (Centro de Aguas Urbanas 2003) it is proposed to follow the best management practices known from various countries worldwide that include natural water courses in urban planning and in the design of public green spaces. If the existing natural hydrological network is maintained and the riparian zones are protected, the necessary investments in storm water infrastructure would decrease significantly. Large green spaces for public use could be used as retention areas along zones facing high flood hazards to avoid additional costs for the construction of new storm water infrastructure or to decrease the damage potential. It is proposed that new construction activities do not exceed the natural amounts of runoff, if necessary through the construction of respective storm water infrastructure.

190

10 Prevention and mitigation measures

Figure 32: Schema of variables relevant for flood risk analysis. Variables printed in bold black and surrounded by a box are points where mitigation measures can be taken to reduce both hazard and vulnerability and therewith the overall risk of suffering da mage from flood event.

10.3 Recommendations

191

Several specific structural measures are also proposed by AC Ingenieros (2008). The DOH is considering improving the situation in the eastern part of Santiago by enlarging the runoff capacity of the Canal San Ramón from 7 m³/s to 20 m³/s is under discussion (Silva 2008). Latest field surveys (November 2010) showed that construction activities are currently ongoing in this region (Figure 33). With all the planning of structural measures, it has to be kept in mind that the maximum runoff, i.e. the maximum load that a channel or water way needs to accommodate will increase in future.

Figure 33: Recently ongoing construction activities at the San Ramón channel that will lead to an enlarged capacity of the water way. Photos: René Höfer.

What have been proven to be an appropriate instrument for flood risk reduction in Europe are the introduction of an insurance system and the development of legal frameworks that enforce flood risk management (Siebert 2004, Wagner 2008, Europäische Union 2007). That forces all involved parties to deal with the problem of floods, to develop risk maps, to take prevention and mitigation measures, and to include the hazard and risk zonation in planning. Kron (2005) states that risk reduction requires a sound cooperation of governments, affected population and insurance companies. The role of insurance companies is thereby to minimize damage for the individual and to increase the willingness for home protection and awareness. In any case, the flood-related damage should in the future be mapped as accurate as possible to carry out cost-benefit analyses for the construction of mitigation measures  either paid by the government or even by the local residents. If none of these calculations or in-depth knowledge about the potential negative impact of a flood exist it is very difficult to plan measures and to sufficiently adapt to the situation.

192

10 Prevention and mitigation measures

10.4 TAKING ACTION Apparently the most appropriate moment for taking action to reduce risks of suffering damage from a hazardous event is the “window of opportunity” that can be located between the recovery and the prevention stages of the disaster risk management cycle (Figure 8) (Steinführer et al. 2005). Another possibility would be to improve things with political changes, e.g. with a new government or a planned restructuring of regulations. Grothmann & Reusswig (2006) conducted a study for Cologne, Germany where they found out that those people who have previously been affected by floods or that are home owners are most willing to take precautionary measures. People that have not yet been affected, have rented homes or rely on structural mitigation measures barely take private protection measures, a finding also stressed by Steinführer et al. (2009).

11 DISCUSSION AND CONCLUSIONS This section is divided into two parts. Section 11.1 starts with a discussion of the applied methods and puts them into the context of general achievements, limitations, and challenges in urban flood risk assessment. Research that still needs to be done is pointed out. Moreover, the transferability of the approach with regard to content and area is discussed. The section does finally address the issue of the role of flood risk in Santiago de Chile. Section 11.2 draws the final conclusions of this research. 11.1 DISCUSSION 11.1.1 Vulnerability The assessment of vulnerability is the mostly discussed part of risk assessment. The recognition of the importance of vulnerability in disaster risk assessment and the conceptual advances that have been made during the last decades, however, are major general achievements. At least in the scientific community, vulnerability is now understood as the explaining factor for risk generation that is highly dynamic and very place- and hazard-dependent. This research builds on that. It involves relevant variables from various spatial levels as it affects people as well as buildings and entire systems. However, a challenge that still remains is the consideration of vulnerability as a main cause of risk generation by urban planners and decision makers. In the present case of Santiago de Chile, vulnerability is not at all considered in the urban development and planning process. This research shows for the first time which variables are both according to the affected population and according to the understanding of selected experts relevant for the generation of vulnerability against floods in the study area. Having compiled this new site- and hazard-specific knowledge might foster a consideration of vulnerability in future decision making processes. Another general success are the methodological and technological advances made with respect to vulnerability assessment. Indicator-based top-down as well as bottom-up and community-based approaches evolved and are combined. With an increasing data quality and quantity, more accurate, comprehensive, and updated risk-relevant information can be derived. Area-wide field surveys were often too time-consuming and expensive if logistically possible at all. Carrying out field surveys  especially in poor areas  is to some degree unsafe. An example to overcome this drawback is the use of very high resolution remote sensing and GIS data that were applied for this research. This research shows how latest technologies in the field of geoinformatics can be combined with traditional methods (census data, field surveys) to obtain vulnerability information in a homogeneous format that cover large areas and that allow for an updated vulnerability analysis with

194

11 Discussion and conclusions

the availability of updated information. Besides the advantages previously mentioned, using geodata enables the analysis of spatial relations: Especially measures of proximity and exposure are important factors that influence vulnerability and that are considered in the presented approach. But even though new technologies constitute a valuable advantage in terms of time efficiency, area coverage, and repeatability, they cannot fully substitute traditional assessment approaches. This research is carried out on the spatial level of a building block. That means that all information is aggregated to this level. Information about individuals, however, is not considered. Thus, the vulnerability assessment applied in this study lacks relevant individual characteristics and is primarily based on physical and exposure indicators. That signifies that the strong dependence on input data remains a hindering factor, especially for local-scale approaches. A further drawback is that updated data in all formats are often not available in fast-growing and rapidly changing settings. That leads to the disadvantage that data from different points in time are used. This problem was in this research accounted for by analyzing only those building blocks that were populated in 2002 (the year of the census). Consequently, built-up areas that evolved after 2002 and that can be seen on the satellite images from 2006 could for methodological reasons not be considered in this research, what clearly constitutes an information sink. This is a clear deficit as especially the new dwellers mostly face the disadvantage of not knowing their surrounding and thus being more vulnerable as they do not have any experience with floods. When applying an indicator-based approach as proposed in this research, the selection and weighting of the indicators is a fundamental problem, especially in studies where local knowledge is left out. Therefore, experts as well as household surveys were carried out to obtain more insight on the relevance of each factor in the study area and to be able to assign validated weights. In addition to that, the web-based tool developed especially for the study area (Ebert [Müller] & Müller 2010b) allows each user (most likely urban planner) to assign individual weights to the indicators according to his/her planning premises. Even though this is associated with the risk of rating certain variables too little important this approach is considered to enlarge the acceptance of the results. Applying methods of quantitative statistics instead of qualitative statistics for the evaluation of the weights of the indicators was in this study not possible as no independent variable that describes the vulnerability level was available. Multivariate methods such as factor analysis or principal component analysis were also not regarded useful in this case as these methods construct new variables that can no longer be reproduced and used as a decision making aid. The most crucial problem in vulnerability assessment though is the broad lack of validation of the results which aggravates the development of generally accepted methods. This issue also applies for this study. Not even damage maps are available to allow for an approximation of the vulnerability. Reiter (2009), however, used information from field surveys to better validate the vulnerability variables. The evaluation of the own affectedness was used as an independent variable in a regression analysis to derive information about the relevance of the remaining surveyed variables. The results were presented in the scope of this research.

11.1 Discussion

195

Over time, local population and decision makers around the globe became increasingly involved in risk and vulnerability assessment and analysis, most recently even through interactive web-based projects (Ebert [Müller] & Müller 2010a, Mioc et al. 2008, Butts et al. 2006). A web-based communication of research findings enhances the understanding and eases the accessibility of knowledge for local decision makers. This research also contributes in that way as an interactive tool for planners in the study area was made available to explore risk-related data and to obtain knowledge about contributing factors. It not only includes the possibility to view dynamic maps online but it also provides the possibility of user interaction by generating own vulnerability maps. The likelihood that risk-reducing measures are taken thus rises as at least the awareness is hopefully rising and the negative consequences are noted. In addition, the positive consequences of taking measures to prevent or mitigate floods can be explored. A challenge with respect to communication of research results though is that vulnerability maps might be difficult to understand and interpret by urban and regional planners that might not be intuitively familiar with the concept of vulnerability. The expert interviews also showed that the evaluation and understanding of vulnerability by the expert varies between the institutions but also between the different spatial scales they are responsible for. Therefore, a sound documentation of the results is indispensable. At least in general it is gradually being recognized that the addressees in practice need to be identified and contacted (often policy makers and decision makers in urban planning). To conclude, the presented approach does for the first time show the relevant variables for flood vulnerability assessment in Santiago de Chile and explores their interrelations in depth using the latest technologies. However, it does still lack validation and does not include personal characteristics. It is recommended to carry out more household surveys to obtain a broader understanding on the deficits of knowledge about flood generation and flood risk prevention. 11.1.2 Hazard The general problem with hazard maps is the actuality and the spatial accuracy. Although the technologies for the generation of hazard maps  also for urban areas  have advanced significantly, the major constraints remain data availability and the possibility to purchase accurate hydraulic models. Both requirements are very costly. Required data for a fairly precise mapping of flood hazard zones is amongst others a very high resolution digital surface model that contains detailed information about the Earth’s surface and all built-up structures located on it. Partly for the same reasons and because it would go beyond its scope, this research does also not provide any new hazard maps but focuses on the changes in runoff that directly determine the flood hazard level: The application of the precipitation-runoff model HEC-HMS using a distributed modeling approach does for the first time in that study area quantitatively show how previous and possible future changes in land use and land cover influence the flood hazard. The chosen grid cell size of 50 m is small enough to account for a very good spatial represen-

196

11 Discussion and conclusions

tation of the current, previous, and future land-use patterns. Remote sensing data from the ASTER sensor proofed to be a valuable data base for the time efficient and semi-automatic derivation of LULC information covering the entire catchment at various points in time. ASTER data and the meteorological data are lowcost data, the elevation and geologic data are available to most decision makers in the city. The model HEC-HMS can be downloaded from the website of the USACE at no charge. Therewith the applied methodology can (i) be repeated at low costs if new data become available and can (ii) be comparatively easily be transferred to other regions of interest. The sensitivity analysis showed that the model reacts sensitive to changing LULC data. Consequently, the influence of LULC changes on the peak runoff volume can be explored and the new knowledge obtained can be regarded as a valuable and new aid in decision making for planners. However, it has to be kept in mind that the modeling process remains a simulation of the reality. The main drawback of the applied modeling process is that the resulting data cannot be validated as no measured runoff data are available for that area at a sufficient level of quality. That means that the modeling error remains unknown. Assessing the results from the sensitivity analysis shows that the range of values varies by +/- 3%. Besides the lack of runoff data, rainfall data are only available from the station Cerro Calán outside the catchment. The precipitation data used as model input have a temporal resolution of one hour, which is acceptable. Even though the data are measured at a station outside the basin about 5 km away from the outlet, statistical analyses showed that these data show a high correlation with the rainfall data available for the basin for a short period of three months. As only this one pluvial station was used as a basis for precipitation data no spatial interpolations were carried out. That means that the precipitation is assumed to be homogeneous in the catchment. This is not realistic as the differences in elevation within the catchment are higher than 2,000 m. Further modeling runs should include a factor to account for these differences as they practically result in higher rainfall intensities, especially in the study area. Another deficit is that the location of the snow line is not explicitly incorporated in the modeling process. Rather, significantly higher retention capacities are assigned to the two uppermost subbasins to account for snow storage. When precipitation falls as rain it directly infiltrates or runs off. If temperatures are low and the precipitation falls as snow above the 0°C isotherm it behaves differently, i.e. it is stored on the surface and infiltrates or runs off with a temporal delay. The basin area contributing to the runoff is thus smaller during cold weather conditions. For further modeling processes in that area it should be included explicitly in the modeling process as the exact location of the snow line in combination with the precipitation intensity again influences the amount of runoff generated (Perez 2009). The LULC pattern is incorporated in the model using the SCS CN method. This is a rather simple method and can thus be applied almost readily but it was developed for the use in the United States of America. The modeling results show that the calculated discharge values are within a plausible range. However, this

11.1 Discussion

197

research is not capable of certifying the validity of the approach in areas outside the United States as a proper validation of the runoff data could not be done. The results thus rely on the assumption that the CNs developed for the certain LULC types in the US are also valid for the LULC types in Chile. Unfortunately, also the required soil data were not available for the study area. Instead, a geologic map was used to derive respective information. Given the fact that the soil development is very low in the catchment this method is regarded being valid. However, the percentages of each HSG in the geologic and geomorphologic formations is derived based on their content description and genesis and not based on validated field studies. A further drawback of the SCS CN method is that the rainfall intensity is not considered (Feldmann 2000). The temporal resolution of the rainfall data of one hour already implies a certain process of averaging, thus this deficit is considered being less significant but remains existing. Finally, it was shown that the model can be applied outside the conterminous US using a distributed approach. It has to be noted though that this is a methodological challenge. The input data must fit into a standardized grid that is only defined and available for the conterminous United States. That means that the user needs to pretend to model a catchment located inside the United States. That does in no way affect the model output but is cumbersome to transform all data (grid data as well as time series) in a way that they can be read by the model. In this scope it needs to be considered that HEC-HMS only accepts the DSS-format as input format which requires the application of specific data conversion software. This cannot directly be downloaded but can be requested by the HEC-HMS support team. The modeled runoff values were then combined with existing flood hazard maps generated in the scope of a Master thesis by Perez (2009). The methodological drawbacks of the maps are outlined in Chapter 6. To conclude, this research generated a range of new information with respect to flood hazard generation, especially as it quantifies impacts of the previous and possible future LULC changes on the amount of runoff. However, the modeling procedure inhibits a range of methodological drawbacks that were addressed and should be improved. The generation of new hazard maps laid beyond the scope of this research but should despite the economic costs be carried out for the entire RM. 11.1.3 Elements at risk The main novelty with respect to the elements at risk is their inclusion in the theoretical concept and practical assessment of urban flood risk next to the component vulnerability. Mapping the spatial location and distribution of people, buildings, and infrastructure is commonly achieved using various types of geodata. Its success generally depends on data availability rather than on the availability of suitable methodologies. However, knowledge about the interpretation of very high resolution multi-spectral satellite data and the application of geographic information systems is indispensable in most cases. The elements at risk were also here mapped using various input data: census data (people), GIS data (infrastructure),

198

11 Discussion and conclusions

and remote sensing data (buildings). A commonly known drawback when using census data is that the distribution of people needs to be assumed homogeneous for one building block. This again explains why individual characteristics cannot be addressed. With respect to the infrastructure data the main drawback is that only those infrastructure installations can be considered for which data are available. Small businesses that are located in the ground floor of private buildings  as one example  are not included in this analysis. An accuracy of 0.88 could be achieved for the delineation of the building mask. This is satisfying for this area. The delineation of buildings from remote sensing data to a large part depends on the geometric resolution of the satellite data. Data with a geometric resolution of 0.6 m were available for this study. To obtain best results, such high resolution data should be analyzed with object-oriented classification software as it was done for this research. The acquisition and use of respective software is costly and time-consuming. Also purchasing respective satellite data requires considerable economic investments, but contains most recent information at a large spatial coverage. Alternatively, GIS data containing information about the distribution of buildings could be used if available. As the elements at risk are a purely quantitative component the proposed assessment method, i.e. the use of geodata for mapping their spatial distribution, serves it purpose. The resulting maps can also be updated readily as soon as new information becomes available. What is still lacking, however, is an assessment of the economic, ecologic, and functional losses. Stage-damage functions as applied in comparable analyses were not available for the study area, thus could not be incorporated. It is consequently recommended to include respective information as soon as they become available to get a more comprehensive understanding of risk. 11.1.4 Risk As a contribution to and further development of the existing approaches to assess flood risk (Chapter 2) this research introduces a conceptual framework that can be practically applied to assess flood risk in a fast-growing megaurban environment. It incorporates the hazard, the elements at risk, and their vulnerability as the three main determining components of risk. It was shown how the approach can be practically applied to generate risk maps by using indicators. The resulting maps are not just the most recent but also the most detailed information generated for the area of interest. As they include the elements at risk and their vulnerability, the final risk maps show a clear improvement of the existing hazard maps that have so far been used as the only planning and decision making tool. However, the chosen approach contains a number of drawbacks. The first is that it introduces a concept of risk that does not originate from the risk debates in Chile. The stakeholders for this research are mainly urban planners on different spatial levels. The traditional understanding of risk, especially in urban planning, mostly equaled the understanding of hazard as interviews proved. That means that the risk concept proposed in this research has the potential to improve the stakeholder’s understanding and the awareness of flood risk generation but might also

11.1 Discussion

199

be rejected as it is hardly comprehensible in the first instance and as it previously played no role in any decision making process or planning instrument. Further aspects with regard to the content of risk are discussed in Section 11.1.7. The approach for risk analysis and risk assessment applied in this research is pragmatic and to a large part technocratic. That has basically the large advantage of being able to obtain results covering a large area, of being transferable, repeatable, and being able to be displayed cartographically. However, real economic data are not included in the analysis as respective data were not available. This does not allow for an economic evaluation of the risk versus the benefits, e.g. of the costs and benefits of constructing in hazard-prone areas. Raaijmakers et al. (2008) point out that risk is always a tradeoff between a potential damage and the expected benefits. Thus it must be analyzed and evaluated in a bi-directional way. Also Dikau (2008) highlights the importance of also incorporating the potential benefits of taking a risky decision. A cost benefit-analysis for a flood defense program in Dresden, Germany was for example carried out in the scope of the MEDIS project (GFZ 2006). Respective analyses could not be carried out in the scope of this research which is a clear lack of information as economic interests have a very high importance in any planning decision taken in Chile (Carvacho 2010). In this context, the statement of Dikau (2008) should be highlighted. He claims that dealing with public risks is clearly a negotiation process of the society that has besides a technical and scientific perspective a social and subjective dimension and thus needs to involve various perspectives. For this research that means that the ecological point of view is not sufficient to carry out a risk assessment. Rather, the potential economic benefits as well as social benefits of having a space to live should for future analyses be included in the risk calculations. The risk assessment is carried out on the spatial scale of a building block that for methodological reasons needs assumed to be homogeneous. That means again that inaccuracies as a result of averaging out all input values were generated and that individual information cannot be provided. To conclude the discussion of concept and methodology, it needs to be pointed out again that  even though the results seem plausible  the level of accuracy of the risk maps can due to the lack of validation data not be defined. This point needs to be made clear to every user of the results. 11.1.5 The applicability of the concept for multi-hazard studies This concept was also applied to explore its potential for the investigation of thermal stress. It was found that some of the indicators used for flood vulnerability analysis are also valid for the analysis of vulnerability towards thermal stress (Ebert [Müller] & Höfer 2010). The main practical requirements are that the indicators needed for the assessment of vulnerabilities towards other hazards are spatially explicit and that the data are available in the same format as used for this study.

200

11 Discussion and conclusions

However, the transfer of the concept to different hazards and its practical assessment requires special care with respect to the content. In other words, the choice and weighting of the vulnerability indicators is hazard-dependent and placedependent. It needs to be validated which adaptations need to be made to the existing indicator set. If adaptations are necessary, the concept developed for this research allows the inclusion of further variables such as slope or building height if they can be made available as relative frequencies per neighborhood. Once the vulnerability is determined to a satisfying degree, the vulnerability index, the hazard index, and the index referring to the elements at risk can be calculated and then multiplied to obtain a risk map. If the risk towards more than one hazard is going to be assessed it is theoretically possible to feed more than one hazard map into the risk model. However, the transparency of the risk map cannot be maintained completely: even though elements at risk and hazard can clearly be identified, the vulnerability values would need to account for two hazards or more which makes it a little more challenging to plan vulnerability reduction measures. It is thus recommended to carry out more research on indicators relevant for other hazards and to apply the presented concept to derive new risk maps. It is also recommended to focus on one hazard at the time. 11.1.6 The spatial transferability of the concept The general risk concept was developed for the application in a fast-growing, complex, urban environment. It is thus per se valid to be transferred to comparable urban regions. The practical feasibility to transfer the results usually depends on data and software availability. It was attempted to use as many low-cost or freely available data and software as possible. Nevertheless, besides hardware (PC with at least 2 GB RAM, GPS, photo camera), software (HEC-HMS, HEC-GeoHMS, HECDSSVue, GageInterp, ArcView, ArcGIS (both with spatial analyst extension), Erdas Imagine, Definiens Developer, MS Office or alike) and data (compare Chapter 6) expert knowledge for data handling and processing is required. The transferability of the concept is enhanced by the structure of the risk model that allows for the inclusion of additional or new data sets. A further advantage is the flexibility of the weighting of indicators that can be taken into consideration for the risk assessment, and especially for the vulnerability assessment. It needs to be kept in mind though that one of the methodological questions in vulnerability assessment that has not been answered yet is if the same set of indicators can be used when comparing the vulnerability between regions. While a set of indicators can be considered valid for one area it does not imply that it is also valid to represent the vulnerability against the same hazard in another area. The decision for one or the other approach is a tradeoff between objective comparability and conceptual security. When transferring the theoretical concept presented here into practice (in one or the other region), vulnerability indicators can be added if applicable, if data are available, and if they can be brought to the same format (relative frequency per building block).

11.1 Discussion

201

In the present case the approach with the proposed weighting of the vulnerability indicators is considered to be valid for the entire Metropolitan Region of Santiago de Chile as input came from experts of different spatial levels across the RM. With regard to the people’s evaluation of their own vulnerability the results from the household surveys can also be transferred to the entire RM because the heterogeneity of the interviewed people within this study is regarded as being sufficiently representative for the rest of the city. Hazard maps can be added from existing data sources if available. The goal of the present study though was to show the variety of influencing factors that contribute to the generation of flood risk. Even if existing hazard maps are used, it should be noted and considered where this information comes from and which variables they depend upon. 11.1.7 The role of flood risk and flood risk awareness The research also showed that the perception of risk is a very individual characteristic and differs amongst the actors  with actors that can be individuals as well as institutions. It was explained that for methodological reasons the individual perceptions of risk of affected people could not be captured in the scope of this research. Nevertheless, field work, interviews, and findings from literature gave evidence about the different perceptions of the risk to suffer damage from flood events. Without being further investigated upon, these different perceptions are lined out in the following as they can to a certain part also be used to explain why the problem of flood risk exists and why it will most likely continue existing in the future. Most of the people living in or close to hazard-prone areas are aware of the risk they are facing to suffer damage from flood events. That becomes evident for example as they use their financial resources and start taking their own mitigation measures to minimize their personal risk of suffering damage as far as they can (compare Section 9.4.1.2). But even amongst the group of affected people the perception of risk is different, e.g. amongst different age groups or amongst men and women (compare Reiter 2009). Also the planners and authorities on the local, communal level  at least in the case of the two municipalities that were researched  are aware of the threat that the regularly occurring flood events pose and would like to change the situation as they lower the quality of life and security within the space they are responsible for (Quezada 2009). In the municipality of La Reina for example information about updated hazard zones and a damage inventory (in monetary values or number of affected households) would be highly appreciated (Quezada 2009). Even though the next issue is that sufficient financial resources for really minimizing the flood risk are scarce, such information would help to better address the issue and to maybe incorporate it in planning decisions. As no insurance system against damage from flood events exists in Chile, numbers about material and immaterial losses after floods are not inventoried. Affected people help themselves and find own means to turn life back to normal.

202

11 Discussion and conclusions

Investments in mitigation, preparedness, and response measures are financed by the municipality to a certain degree, e.g. by cleaning the storm water evacuation system if existent, providing sand bags, and removing water, depending on their funds. Structural measures need to a large part to be financed by the MOP-DOH  where sufficient financial means are not available. In addition, communication and collaboration between the actors on horizontal as well as vertical level shows a large potential to be fostered and optimized. In the case of the the area along the Andean foothills in the RM, solving the flood problems including the nuisance associated with the high sediment left behind after flood events, investments of about 15,000 million Peso would be necessary (Silva 2008). Required investments would focus on the widening of the channels to increase their capacity and the construction of sedimentation tanks to reduce the sediment load in the river when it overflows (Silva 2008). As the state does not provide sufficient money for investments in structural measures and as no respective insurance system exists, it becomes clear that the perception of risk is different on that level. Furthermore, the existing regional and urban planning framework is not adapted to this lack of sustainable urban development (Section 10.1). Risk and vulnerability have been and still are to a large extent ignored in the process of urban planning and urban expansion. The risk of flooding needs to be set in comparison to other risks. The large earthquake and tsunami that hit vast parts of Chile in February 2010 showed vividly that floods are not the only hazard that people are facing. Further risks go along with daily life in Santiago de Chile. Results from the household surveys (Section 6.8.2 & 9.4.1.2) showed that other threats such as crime and lack of housing space that are relevant throughout the entire year are also very important (Reiter 2009). Dikau (2008) furthermore brings up the point of when it becomes necessary to take action as it is very difficult to define a respective threshold value and as changes occur gradually. This requires further research and more in-depth and also more personal interviews. 11.1.8 Further ecological consequences of urban growth Construction activities towards the urban fringe and the densification of the urban built-up area do not just have an impact on the infiltration capacities of the soil and the resulting increase of the flood hazard in the urban area. The anthropogenic changes of the ecosystem associated with urban expansion induce the loss of further ecosystem services: Green spaces, also within the built-up area, provide shadow and a cooling function during days of heat stress. Romero et al. already showed in 1999 (Romero et al. 1999) the importance of urban ventilation patterns and local wind systems which is in the case of Santiago de Chile a main determinant of the urban air quality. Through its topographic location in a valley surrounded by mountain ranges and the frequent inversion layers combined with a permanently increasing car fleet (as one prominent source of

11.2 Conclusions

203

pollution) the problem of air quality and increased ozone concentration is ubiquitous. With increasing construction activities, especially towards the north-east, the ventilation corridors disappear when residential areas are constructed (Romero 2008). Green spaces and especially the ones in the peri-urban area function as habitats for a range of flora and fauna. The project Protege (Proyecto de Conservación y Protección de la Cordillera de Santiago de Chile) for example has set up a website and several educational services to raise awareness for the ecological function of the peri-urban fringe amongst the population (Protege 2010). As discussed in Section 4.3.4 the connectivity of green spaces is important for its functioning as habitat and pathway and should therefore be fostered. Besides their ecological value, green spaces do also offer a recreational function for the residents of Santiago de Chile. The tightrope walk is now to decide between a compact urban development with less green spaces in the built-up area to function as cooling or recreational areas or an extensive spatial expansion with more space for green areas but a higher fragmentation of the landscape and a much higher volume of traffic  in turn affecting the air quality and leading to a higher demand of road construction. Besides all calls for green spaces which are certainly valid and justified as their ecological and social value shows it has to be pointed out that Santiago de Chile is located in subtropical semi-arid environment which requires additional enormous efforts for irrigation. When planning and creating green spaces it should be considered that local species instead of palm trees are planted as climatic projections predict less availability of water within the next decades as glaciers that supply the city with water are predicted to melt (Bárcena et al. 2009, CONAMA 2007) . 11.2 CONCLUSIONS While the positive economic consequences of urban expansion are evident and appreciated, the negative environmental consequences are often neglected, especially at the scale of the entire municipality or city. This deficit becomes obvious when consulting existing flood hazard maps used in the Metropolitan Regulatory Plan PRMS: the official hazard zoning has not been updated since 1987, although catchment characteristics have changed significantly with respect to land use. Even though flood hazard studies were carried out in parts of Santiago de Chile and also for the municipalities of La Reina and Peñalolén (e.g. AC Ingenieros 2008, Perez 2009), the results obtained are hardly ever transferred into practice  not only for financial reasons but also as a result of poor communication among stakeholders and decision makers. The latest information on flood hazards is therefore never considered in decision-making with respect to urban development, resulting in a continuation of construction in flood-prone regions, a loss of retention areas, and an increasing flood risk. A conceptual framework that can be used to capture, analyze, and assess the flood risk in a growing urban agglomeration was developed in the scope of this

204

11 Discussion and conclusions

study. At the example of the San Ramón watershed and the two adjacent municipalities La Reina and Peñalolén in the eastern part of Santiago de Chile it was shown how the conceptual framework can be applied in practice to assess flood risk. The three main components of the framework are the hazard, the elements at risk, and their vulnerability. The inclusion of the elements at risk, i.e. the amount of people, infrastructure, and built-up area as well as their vulnerability constitutes an added value to the existing flood hazard maps in the study area. The main conclusions from this research are that land-use and land-cover changes in the scope of urban expansion do in various ways influence the flood risk in Santiago de Chile: The infiltration capacities of the soil decrease with a loss of various types of green spaces. That leads to an increasing flood hazard. Chapter 8 did clearly show how the discharge values of the San Ramón River change with a changing landuse pattern. The most threatening land-use scenario is hereby the loss of vegetation due to longer drought periods in combination with an increasing amount of rainfall during extreme events. Most efforts with respect to flood prevention should thus focus on the maintenance and ideally the increase of green spaces in the catchment. Even a limited amount of construction activities with a large amount of private green spaces would be better than leaving the basin with a decreasing amount of green spaces as it is currently the trend. Afforestation as the clearly preferable alternative should comprise native species that are adapted to low water availability in the dry season. As another aspect, the growing number of people, buildings, and infrastructure located in hazard-prone areas result in a permanently growing exposure of values to floods. The main reason for that is that flood-prone areas are economically highly interesting and that the Metropolitan Regulatory Plan allows for construction in these areas if certain conditions apply. The central issue is that open spaces within the study area that are used as retention areas or that are protected because of flood hazards will likely be converted into waste dumps if they are not maintained. The maintenance as green spaces costs around 6 US$ per m² per month which is very costly for a municipality (Carvacho 2010). Thus, the most beneficial use is to carry out a very local environmental impact assessment study, to install flood protection measures around the lot of interest, and to open the land for construction. The regulations then prescribe that compensatory green areas need to be constructed. However, there is currently no controlling body ensuring that these regulations are followed. As a consequence, they are not created. This regulation  or exception from the regulations  urgently needs to be changed as the flood damage will otherwise continue to increase. What would help in this situation in addition is either a stronger regulation system or a sudden increase of environmental awareness. It has to be made clear to the population that structural flood protection measures do in no case prevent from all types of damage. Furthermore, it has to be considered that all types of new constructions lead to changes in the local water balance. Even if one lot is protected through newly constructed measures the overall hazard is (i) increasing through a higher percentage of sealed surfaces in the basin and (ii) transferred to other lots that did not face a hazard before. However, it was stated in a report of the ECLAC (Bárcena et al.

11.2 Conclusions

205

2009, p. 54) that the expected impact of climate change in Santiago de Chile will cause significant economic costs of which some also result from the reconstruction of urban infrastructure that has been destroyed during flood events. Most interesting is that not just the urban poor but also middle and upper class households are affected by floods. The criteria that are relevant for vulnerability are rather the physical than the socio-economic criteria (Section 9.4). Both experts and affected population agreed that the location of a building in relation to street level and the amount of green spaces per building block are the most relevant variables. It was found from the household surveys that the household size and the type of employment (permanent vs. sporadic) also determine the vulnerability. The experts do in average think that the construction material plays a major role for vulnerability. Given the fact that the flood hazard in urban areas is most influenced by urban expansion and associated changes in LULC, reinforced by a periodical increase in intense rainfalls due to global and subsequently regional climate change, the ongoing transformations in the case study predict an increase in flood risk. The research about flood risk generation, its changes in the past, its possible development in the future, and its perception by the local population and stakeholders brought some interesting insights and results that were lined out and discussed in the course of this thesis. Advances were made with respect to the risk concepts, to methodologies, to content, and to the communication of information. However, a critical evaluation of discussions with decision makers and findings from literature and field surveys makes the researcher assume that the costs of changing the attitude of the main decision makers from short-term to long-term planning and thinking, of re-structuring the decision making processes with respect to land-use planning, and of creating a hazard-safe and attractive urban environment for the future generations are much higher than the damage caused by the regularly occurring flood events that the affected people suffer even though they are predicted to increase in the future. Further research should therefore not only be carried out with respect to the research gaps brought up in the previous Section 11.1, but also on how to foster environmental awareness, the readiness to plan long-term investments, to prepare for the future, and to create a sustainable environment for the next generations.

REFERENCES REFERENCES AC Ingenieros Consultores Ltda. (2008): Diagnóstico de Cauces Naturales, Sector Pie Andino. Región Metropolitana, Santiago de Chile. Adger, W. (2006): Vulnerability. Global Environmental Change 16, 268281. Adimark (2003): Mapa socioeconómico de Chile. Modelo estimativo del N.S.E. en los hogares de Chile, Adimark. Investigaciones de Mercado y Opinión Pública, Santiago de Chile. ADPC. Program for Hydro-Meteorological Disaster Mitigation in Secondary Cities in Asia (PROMISE). http://www.adpc.net/udrm/promise/default.htm, Access: 2009-08-16. Alexander, D. (2000): Confronting Catastrophe  New Perspectives on Natural Disasters. Oxford University Press, New York, 288 p. Angel, S., Sheppard, S., Civco, D., Buckley, R., Chabaeva, A., Gitlin, L., Kraley, A., Parent, J., and Perlin, M. (2005): The dynamics of global urban expansion, The World Bank. Transport and Urban Development Department., Washington D.C. Antinao, J., Fernandez, J., Naranjo, J., and Villarroel, P. (2003): Peligro de remociones en masa e inundaciones de la cuenca de Santiago. Carta Geologica de Chile, Servicio Nacional de Geología Ambiental, No. 2. 1 mapa escala 1:100.000. Servicio Nacional de Geología y Minería. Santiago de Chile. Apel, H., Thieken, A., Merz, B., and Blöschl, G. (2006): A probabilistic modelling system for assessing flood risks. Natural Hazards 38, 79100. Apikul, C. (2002): Mitigating flood risk in Cambodian communities. Empowering communities to manage disaster risk, Asian Urban Disaster Mitigation Program, ADPC, Pathumthani, Thailand. Arcement Jr., G. and Schneider, V. (1989): Guide for Selecting Manning’s Roughness Coefficients for Natural Channels and Flood Plains. United States Geological Survey Water-supply Paper 2339. Metric Version, Washington, D.C., 38 p. Arriagada, C. and Moreno, J. (2006): Atlas de la evolución del déficit habitacional en chile 19922002 323, Ministerio de Vivienda y Urbanismo. División Técnica de Estudio y Formento Habitacional  DITEC, Santiago de Chile. Artan, G., Gadain, H., Smith, J., Asante, K., Bandaragoda, C., and Verdin, J. (2007): Adequacy of satellite derived rainfall data for stream flow modeling. Natural Hazards 43, 167185. Ashley, R., Blanksby, J., Cashman, A., Jack, L., Wright, G., Packman, J., Fewtrell, L., Poole, T., and Maksimovic, C. (2007): Adaptable urban drainage: Addressing change in intensity, occurrence and uncertainty of stormwater (audacious). Built Environment. Climate Change and Cities 33, 1, 7084. Atkinson, E. (2001): Curve number analysis for Texas. Master’s thesis, Texas Tech University, Dept. of Civil Engineering, Lubbock. Atteslander, P., Cromm, J., Grabow, B., Klein, H., Maurer, A., and Siegert, G. (2008): Methoden der empirischen Sozialforschung, vol. 12. Erich Schmidt Verlag, Berlin, 359 p. Aubrecht, C., Steinocher, K., Hollaus, M., and Wagner, W. (2009): Integrating earth observation and GIScience for high resolution spatial and functional modeling of urban land use. Computers, Environment and Urban Systems 33, 1525.

208

References

Ayala, L. and Cabrera, G. (1996): Análisis de vulnerabilidad des sector oriente de la cuidad de Santiago ante la occurencia de aluviones y crecidas, Región Metropolitana, Ministerio de Obras Públicas, Santiago de Chile. Ayala, L., Espinoza, G., and Saragoni, R. (1987): Estudio de áreas de Riesgo por Inundación. Segunda Parte, Ministerio de Vivienda y Urbanismo, Santiago de Chile. Azar, D. and Rain, D. (2007): Identifying population vulnerable to hydrological hazards in San Juan, Puerto Rico. GeoJournal 69, 2343. Baatz, M., Benz, U., Dehghani, S., Heynen, M., Haltje, A., Hofmann, P., Lingenfelder, I., Mimler, M., Sohlbach, M., Weber, M., and Willhauck, G. (2004): eCognition User Guide 4. Definiens Imaging GmbH, München, Germany. Baltas, E., Dervos, N., and Mimikou, M. (2007): Technical Note: Determination of the SCS initial abstraction ratio in an experimental watershed in Greece. Hydrology and Earth System Sciences 11, 18251829 Bárcena, A., Prado, A., Samaniego, J., and Malchik, S. (2009): La economía del cambio climático en Chile 288, CEPAL, Gobierno de Chile, Santiago de Chile. Bates, P. and De Roo, A. (2000): A simple raster-based model for flood inundation simulation. Journal of Hydrogeology 236, 5477. BBC. Weather. Average conditions. Santiago de Chile. http://www.bbc.co.uk/weather/world/city guides/results.shtml?tt=tt001770, Access: 01-28-2009. Berkhout, F., Hertin, J., and Jordan, A. (2002): Socio-economic futures in climate change impact assessment: using scenarios as ‘learning machines’. Global Environmental Change 12, 8395. Birkmann, J. (2005a): Danger need not spell disaster. But how vulnerable are we? Research Brief UNU-EHS, Bonn, 1, 8. Birkmann, J. (2005b): Measuring Vulnerability. Expert Workshop in Kobe, Japan. Report on the 1st meeting of the expert working group “Measuring Vulnerability” of the United Nations University Institute for Environment and Human Security (UNU-EHS), United Nations University, UNU-EHS, Bonn. Birkmann, J. (2005c): Report on the 1st meeting of the expert working group “Measuring Vulnerability”. Working Paper No. 1, United Nations University, UNUEHS, Bonn. Birkmann, J. (2005d): Vulnerability assessment. Applicability and usefulness of indicators to measure vulnerability. In IIASA-DPRI Forum (Beijing, China,), p. 37. Birkmann, J. and Wisner, B. (2006): Measuring the un-measurable. The challenge of vulnerability. SOURCE 5/2006, United Nations University, UNUEHS, Bonn. Blaikie, P., Cannon, T., Davis, I., and Wisner, B. (1994): At Risk. Natural hazards, people’s vulnerability, and disasters., first ed. Routledge, London, New York. Blaschke, T. (2010): Object based image analysis for remote sensing. ISPRS Journal of Photogrammetry & Remote Sensing 65, 1. Blaschke, T. and Strobl, J. (2001): What’s wrong with pixels? Some recent developments interfacing remote sensing and GIS. GIS 6, 1217. Bogardi, J. (2006): Measuring Vulnerability to Natural Hazards: Towards Disaster Resilient Societies. Introduction. United Nations University Press, Bonn. Bohle, H.-G. (2007): Geographien von Verwundbarkeit. Geographische Rundschau 59, 10, 2025. Boughton, W. and Droop, O. (2003): Continuous simulation for design flood estimation – a review. Environmental Modelling & Software 18, 4, 309318.

References

209

Breuer, L., Huisman, J., Willems, P., Bormann, H., Bronstert, A., Croke, B., Frede, H.-G., Gräff, T., Hubrechts, L., Jakeman, A., Kite, G., Lanini, J., Leavesley, G., Lettenmaier, D., Lindström, G., Seibert, J., Sivapalan, M., and Viney, N. (2009): Assessing the impact of land use change on hydrology by ensemble modeling (LUCHEM). I: Model intercomparison with current land use. Advances in Water Resources 32, 129146. Briguglio, L. (2003): Indicators for disaster risk management. Methodological and practical considerations for constructing socio-economic indicators to evaluate disaster risk Operation ATN/JF-7907-RG, Inter-American Development Bank; Universidad Nacional de Colombia  Sede Manizales; Instituto de Estudios Ambientales  IDEA , Manzales, Colombia. Bronger, D. (1996): Megastädte. Geographische Rundschau 48, 5, 7481. Bronstert, A., Niehoff, D., and Bürger, G. (2002): Effects of climate and land-use change on storm runoff generation: Present knowledge and modeling capabilities. Hydrological Processes 16, 509529. Brooks, N. (2003): Vulnerability, risk and adaption: A conceptual framework. Tyndall Centre Working Paper 38, Centre for Social and Economic Research on the Global Environment (CSERGE), School of Environmental Sciences, University of East Anglia, Norwich, 01/06. Büscher, C. (2008): Durch Luhmanns Brille: Herausforderungen an Politik und Recht in Lateinamerika und in der Weltgesellschaft. Nomos Verlag, Baden-Baden, Risiko-Lebensraum Megastadt. Eine soziologische Perspektive. Butts, M., Klinting, A., Ivan, M., Larsen, J., Brandt, J., and Price, D. (2006): A flood forecasting system: Intergrating Web, GIS and modeling technology. In ESRI International User Conference Proceedings (August 711), ESRI. CADE IDEPE Consultores en Ingeniería. (2001): Plan Maestro de Evacuación y Drenaje de Aguas Lluvias del Gran Santiago. Ministerio de Obras Públicas, Dirección de Obras Hidraúlicas, Santiago de Chile. Canters, F., Chormanski, J., Van de Voorde, T., and Batelaan, O. (2006): Effects of different methods for estimating impervious surface cover on runoff estimation at catchment level. In 7th International Symposium on Spatial Accuracy Assessment in Natural Resources and Environmental Sciences (Lisbon, Portugal, July 57), M. Caetano and M. Painho, Eds., pp. 557566. Cardona, O. (2003a): Indicators for disaster risk management. In First expert meeting on disaster risk conceptualization and indicator modelling (Manizales), vol. March 2003, p. 37. Cardona, O. (2003b): Mapping Vulnerability: Disasters, Development and People. Earthscan Publishers, The need for rethinking the concepts of vulnerability and risk from a holistic perspective: A necessary review and criticism for effective risk management, p. 356. Cardona, O. (2007): Indicators of disaster risk and risk management. Program for Latin America and the Caribbean. Summary Report 363.348 C774-dc22, Inter-American Development Bank. Sustainable Development Department, Pittsburgh, 04/06. Carrasco, J., Osorio, R., and Casassa, G. (2008): Secular trend of the equilibrium-line altitude on the western side of the southern Andes, derived from radiosonde and surface observations. Journal of Glaciology 54, 538550. Carrera, J. (2005): Coupled models for the hydrological cycle. Springer, Berlin, Heidelberg, New York, Systems approach: The Nature of Coupled Models, pp. 75164. Carvacho, A. and Paredes, M. (2009): MINVU/SEREMI. Head of the Regional Planning Unit (Carvacho). Architect, Member of the Regional Planning Unit (Paredes) of Santiago de Chile Metropolitan Region. Interview carried out by A. Ebert (A. Müller) on March 24, 2009.

210

References

Carvacho, D. (2010): Recuperación de áreas inundables urbanas. Caso: Cuidad de San Vicente de Tagua Tagua. Master’s thesis, Universidad de Chile, Santiago de Chile. Cavieres, L. and Arroyo, M. (1999): Lapse-rate in the Río Molina valley, province of Santiago, central Chile (33°S). Revista Geográfica Terra Australis 44 (1999), 7986. Cavieres, L., Peñaloza, A. P., and Arroyo, M. (2000): Altitudinal vegetation belts in the highAndes of central Chile (33°S). Revista Chilena de Historia Natural 73, 331344. Centro de Aguas Urbanas. (2003): ¿Quién Debe Pagar por las Aguas Lluvias? Santiago de Chile, Noviembre. Céspedes, L. (2010): SUBDERE, Consultant of the Department of politics and investigations of the Ministry of the Interior. Santiago de Chile. Discussion on November 28, 2010. Charrier, R., Pinto, L., and Rodríguez, M. (2007): The Geology of Chile. The Geological Society, Tectonostratigraphic evolution of the Andean Orogen in Chile, pp. 21114. Claire, I. (2007): Role of census data & population in disaster management, RMSI Pvt. Ltd. India, 5 p. Clark, G. E., Moser, S., Ratick, S., Dow, K., Meyer, W. B., Emani, S., Jin, W., Kasperson, J. X., Kasperson, R. E., and Schwarz, H. E. (1998): Assessing the vulnerability of coastal communities to extreme storms: the case of Revere, MA., USA. Mitigation and Adaptation Strategies for Global Change 3, 1, 5982. CONAMA (2007): Plan de Acción Nacional de Cambio Climático, CONAMA, Santiago de Chile. Corbane, C., Raclot, D., Jacob, F., Albergel, J., and Andrieux, P. (2008): Remote sensing of soil surface characteristics from a multiscale classification approach. Catena 75, 308318. CRED. EM-DAT: The International Disaster Database. Centre for Research on the Epidamiology of Disasters  CRED. http://www.emdat.be/database, Access: 09-16-2010. Crooks, S. and Naden, P. (2007): CLASSIC: A semi-distributed rainfall-runoff modelling system. Hydrology and Earth System Sciences 11, 1, 516531. Cross, J. (2001): Megacities and small towns: different perspectives on hazard vulnerability. Environmental Hazards 3, 6380. Cutter, S., Boruff, B., and Shirley, W. (2003): Social vulnerability to environmental hazards. Social Science Quarterly 84, 2, 242260. Definiens AG. (2007): Definiens developer 7. User guide, Definiens AG, München. Dikau, R. (2008): KatastrophenRisikenGefahren: Herausforderungen für das 21. Jahrhundert. In Umgang mit Risiken: KatastrophenDestabilisierungSicherheit (Bayreuth, Berlin), E. Kulke and H. Popp, Eds., Deutsche Gesellschaft für Geographie, pp. 4768. Dirección General de Aeronáutica de Chile. Dirección Meteorológica de Chile. Región Metropolitana. Descripción climatológica. http://www.meteochile.cl/climas/climas Region Metropolitana.html, Access: 01-28-2009. Dirección General de Aguas. Ministerio de Obras Públicas. Servicios satelitales en tiempo real. http://dgasatel.moptt.cl, Access: 03-15-2009 to 03-31-2010. Ducci, M. and González, M. (2006): Santiago. Dónde estamos y dónde vamos. Centro de Estudios Públicos, Santiago de Chile, Anatomía de la expansión de Santiago, pp. 123146. Dunn, C., Ackerman, C., Doan, J., and Evans, T. (2000): Hydrologic Modeling for the Sacramento and San Joaquin Basins Using GeoHMS. In Proceedings of the ESRI User Conference. (San Diego, CA). Dwyer, A., Zoppou, C., Nielsen, O., Day, S., and Roberts, S. (2004): Quantifying social vulnerability: A methodology for identifying those at risk to natural hazards GeoCat No. 61168, Geoscience Australia, Canberra, 01/06.

References

211

Dyck, S. and Peschke, G. (1995): Grundlagen der Hydrologie. Verlag für Bauwesen, Berlin. Ebert, A. [Müller, A.] and Höfer, R. (2010): Mapping urban vulnerability from a multihazard perspective. In GI-Forum 2010. Workshop: Spatial assessment and analysis of vulnerability. (Oral presentation and extended abstract submitted for publication in conference proceedings). Ebert, A. [Müller, A.], Kerle, N., and Stein, A. (2009): Urban social vulnerability assessment with physical proxies and spatial metrics derived from air- and spaceborne imagery and GIS data. Natural Hazards 48, 275294. Ebert, A. [Müller, A.] and Müller, A. (2010a): Bewertung von Vulnerabilität gegenueber Hochwasser in Santiago de Chile mithilfe einer dienstebasierten WebGISAnwendung  reviewed paper. In Angewandte Geoinformatik 2010. Beiträge zum 22. AGIT-Symposium Salzburg (Berlin), J. Strobl, T. Blaschke, and G. Griesebner, Eds., Wichmann Verlag, pp. 840849. Ebert, A. [Müller, A.] and Müller, A. (2010b): Offene, dienstebasierte GIS-Architekturen zur Kommunikation von Forschungsergebnissen am Beispiel eines Teilprojektes der internationalen Forschungsinitiative „Risk Habitat Megacity“. In GeoForum MV 2010  Vernetzte Geodaten: vom Sensor zum Web, R. Bill, G. Flach, U. Klammer, and C. Niemayer, Eds., pp. 915. ECLAC/CEPAL (2000): From rapid urbanization to consolidation of human settlements in Latin America and the Caribbean: A territorial perspective, Santiago de Chile, 97 p.. EIB  HR Wallingford. (2007): Guide for preparation of flood risk management schemes. European Investment Bank. Elmustafa Hayati, O., Hamad Khogali, M., Abedallah Abdelrahaman, B., and Abbas Hilmi, U. (2004): Disaster preparedness training manual, Sudanese Red Crescent Society, Program Department. Disaster Management Section. Flood Risk Reduction Project. Erdas (1999): ERDAS Field Guide, 5th ed. ERDAS, Inc., Atlanta, Georgia. ERDAS (2001): ERDAS Imagine Expert Classifier. Overview, Erdas Inc., Atlanta, Georgia. ESRI (2008): Geographic Information Systems Providing the Platform for comprehensive Emergency Management, ESRI, Redland. Europäische Union (2007): Richtlinie 2007/60/EG des Europäischen Parlaments und des Rates vom 23. Oktober 2007 ueber die Bewertung und das Management von Hochwasserrisiken, Bundesministerium für Umwelt, Naturschutz und Reaktorsicherheit. Fedeski, M. and Gwilliam, J. (2007): Urban sustainability in the presence of flood and geological hazards: The development of a GIS-based vulnerability and risk assessment methodology. Landscape and Urban Planning 83, 5061. Fekete, A. (2009): Validation of a social vulnerability index in context to river-floods in Germany. Natural Hazard and Earth System Science 9, 393403. Feldman, A. (2000): Hydrologic Modeling System HEC-HMS. Technical Reference Manual, US Army Chorps of Engineers. Hydrologic Engineering Centre, HEC, Davis. Felgentreff, C. (2008): Naturrisiken und Sozialkatastrophen, vol. XVIII. Spektrum Akademischer Verlag, Wiederaufbau nach Naturkatastrophen, pp. 281293. Fernández, D. and Lutz, M. (2010): Urban flood hazard zoning in Tucumán Province, Argentina, using GIS and multicriteria decision analysis. Engineering Geology 111, 14, 9098. Fernández, B. and Montt, J. (2004): Mapas de riesgo de inundación para el Gran Santiago. Etapa I: Caracterización inicial, AACh, Mapcity, Santiago de Chile. Ferrando, F. (1998): Proyecto Regional Bases Para El Ordenamiento Territorial Ambientalmente Sustentable De La Región Metropolitana De Santiago I y II Etapa, Santiago de Chile.

212

References

Ferrando, F. and Cueto, S. (1998): Proyecto Regional Bases Para El Ordenamiento Territorial Ambientalmente Sustentable De La Región Metropolitana de Santiago. Segunda Etapa: Inventario Ecológico y Evaluación del Territorio. Santiago de Chile, Carta de riesgos naturales: inundación, pp. 117126. FLOODSite (2009): FLOODsite. Integrated Flood Risk Analysis and Management Methodologies. http://www.oodsite.net. Ford, D., Pingel, N., and DeVries, J.J. (Updated by HEC) (2008): Hydrologic modeling system HEC-HMS. Applications guide. Computer software application guide, US Army Corps of Engineers, Davis. Fuentes, C. and Romero, H. (2007): Análisis de los cambios de usos y coberturas de suelo y de la escorrentía superficial en tres cuencas del piedemonte de Santiago, entre 1975 and 2007. In Coloquio Internacional “Construir la Resiliencia de los Territorios”, Pontificia Universidad Católica de Valparaíso. Galetovic, A. and Jordán, P. (2006): Santiago. Dónde estamos y dónde vamos. Centro de Estudios Publicos, Santiago de Chile, Santiago: ¿Dónde estamos? ¿Hacia dónde vamos?, pp. 2571. Galetovic, A. and Poduje, I. (2006): Santiago. Dónde estamos y dónde vamos. Centro de Estudios Publicos, Santiago de Chile, ¿Quién es Santiago?, pp. 323. GFZ Potsdam (2006): RIMAX  Risikomanagement extremer Hochwasserereignisse. GeoForschungsZentrum Potsdam -GFZ- in der Helmholtz Gesellschaft (publisher) and Bundesministerium für Bildung und Forschung -BMBF-, Berlin (supporter), Potsdam. Grothmann, T., and Reusswig, F. (2006): People at risk of flooding: Why some residents take precautionary action while others do not. Natural Hazards 38, 101120. Grunwald, A. and Kopfmüller, J. (2006): Nachhaltigkeit. No. 978-3-593-37978-4. Campus Verlag, Frankfurt/Main & New York. Haki, Z., Akyuerek, Z., and Duezguen, S. (2004): Assessment of social vulnerability using geographic information systems: Pendik, Istanbul case study. In 7th AGILE Conference on Geographic Information Science (Heraklion, Greece), Middle East Technical University of Ankara, Turkey, pp. 413423. Hansjürgens, B. and Heinrichs, D. (2007): Mega-Urbanisierung: Chancen und Risiken für Nachhaltige Entwicklung in Megastädten. Bundeszentrale für Politische Bildung. Online Dossier Megastädte. Hansjürgens, B., Heinrichs, D., Kopfmüller, J., Lehn, H., and Nuissl, H., Eds. (2007): Risk Habitat Megacity ¿Sostenibilidad en Riesgo? A Helmholtz Research Initiative 20072013. Research Plan. Leipzig. Hansjürgens, B., Heinrichs, D., and Kuhlicke, C. (2008): Megacities. Resilience and social vulnerability. No. 10. SOURCE  Studies of the University: Research Counsel, Education  Publication Series of UNU-EHS, Bonn, Mega-urbanization and social vulnerability, pp. 2028. Hazel, G. and Miller, D. (2006): Megacities und ihre Herausforderungen. Die Perspektive der Städte, GlobeScan und MRC McLean Hazel, München. Heinrichs, D. and Kabisch, S. (2006): Risikolebensraum Megacity. Strategien für eine nachhaltige Entwicklung in Megastädten und Ballungszentren. GAiA 2, 157160. Heinrichs, D., Rodriguez, J., and Welz, J. (2009): Migration and residential segregation in Santiago de Chile: Trends, relations and implications for policy on social integration. In 5th International Conference ‘Private Urban Governance and Gated Communities’ of the Research Network, Santiago de Chile.

References

213

Heremans, R., Willekens, A., Borghys, D., Verbeeck, B., Valckenborgh, J., Acheroy, M., and Perneel, C. (2003): Automatic detection of flooded areas on NVISAT/ASAR images using an object-oriented classification technique and an active contour algorithm. In RAST 2003: Proceedings of the International Conference on Recent Advances in Space Technologies (Istanbul, Turkey, November 2022), S. Kurnaz, F. Ince, and S. Onbasioglu, Eds., vol. IEEE Catalog Number 03EX743, Signal & Image Center, R. Mil. Acad., Brussels, Belgium, Institute of Electrical and Electronics Engineers, p. 311. Hong, Y., Adler, R., Negri, A., and Huffman, G. (2007): Flood and landslide applications of near real-time satellite rainfall products. Natural Hazards 43, 285294. Horritt, M. and Bates, P. (2002): Evaluation of 1d and 2d numerical models for predicting river flood inundation. Journal of Hydrology 268, 8799. Hossain, F., Katiyar, N., Y.Hong, and Wolf, A. (2007): The emerging role of satellite rainfall data in improving the hydro-political situation of flood monitoring in the under-developed regions of the world. Natural Hazards 43, 199210. Hutchison, M. (1989): A new procedure for gridding elevation and stream line data with automatic removal of spurious pits. Journal of Hydrology 106, 34, 211232. Im, S., Kim, H., Kim, C., and Jang, C. (2009): Assessing the impacts of land use changes on watershed hydrology using MIKE SHE. Environmental Geology 57, 231239. INE (2002 Censo 2002):, Instituto Nacional de Estadísticas, Santiago de Chile. INE (2008): 1.2 Estadísticas demográficas, Instituto Nacional de Estadísticas and Comisión Económica para América Latina y el Caribe, División de Población Centro Latinoamericano y Caribeño de Demografía, Santiago de Chile. INE (2009): Compendio Estadístico. 1.2 Estadísticas demográficas, Instituto Nacional de Estadísticas. Dirección Nacional, Santiago de Chile. Ingeniería DICTUC (2005): Mapas de riesgo de inundación para el Gran Santiago, Etapa II. Revisión y Actualización 2005, División Ingeniería Hidráulica y Ambiental. Pontificia Universidad Católica de Chile, Santiago de Chile. IPCC (2007): Climate change 2007: Synthesis report, IPCC, Pachauri, R.K. and Reisinger, A. (Eds.), Geneva. ISDR (2004): Living with risk: A global review of disaster reduction initiatives, United Nations, Geneva. Jacquin, A., Misakova, L., and Gay, M. (2008): A hybrid object-based classification approach for mapping urban sprawl in periurban environment. Landscape and Urban Planning 84, 152165. Javed, A. (2005): Vulnerability mapping: A GIS based approach to identify vulnerable regions to climate change. GIS Development December. Johnson, D. and Miller, A. (1997): A spatially distributed hydrologic model utilizing raster data structures. Computers & Geoscience 23, 3, 267272. Karmakar, S., Simonovic, S., Peck, A., and Black, J. (2010): An information system for riskvulnerability assessment to flood. Journal of Geographic Information System 2, 129149. Kienberger, S., Lang, S., and Zeil, P. (2009): Spatial vulnerability units  expert-based spatial modelling of socio-economic vulnerability in the Salzach catchment, Austria. Natural Hazard and Earth System Science 9, 767778. Kienberger, S. and Steinbruch, F. (2005): P-GIS and disaster risk management: Assessing vulnerability with P-GIS methods  Experiences from Búzi, Mozambique. In International Conference on Participatory Spatial Information Management and Communication. PGIS ’05.

214

References

Knebl, M., Yang, Z.-L., Hutchison, K., and Maidment, D. (2005): Regional scale flood modeling using NEXRAD rainfall, GIS, and HEC-HMS/RAS: a case study for the San Antonio River Basin Summer 2002 storm event. Journal of Environmental Management 75, 325336. Knight, F. (1921): Risk, Uncertainty, and Profit, 1st ed. Boston, MA: Hart, Schaffner & Marx; Houghton Mifflin Co. Kraas, F., Aggarval, S., Coy, M., Heiken, G., de Mulder, E., Marker, B., Nenonen, K., and Yu, W. (2005): Megacities  our global urban future. Earth sciences for society, Planet Earth, Trondheim. Krätzschmar, E. and Böhm, C. (2006): Urban risk assessment on the basis of land use and change detection mapping. In 1st EARSeL Workshop of the SIG Urban Remote Sensing, HumboldtUniversität zu Berlin, March 23. Krause, P. (2002): Quantifying the impact of land use changes on the water balance of large catchments using the J2000 model. Physics and Chemistry of the Earth 27, 663673. Kron, W. (2005): Flood Risk = Hazard * Value * Vulnerability. In Water International (March), vol. 30 of International Water Resources Association, pp. 5868. Kull, D. and Feldman, A. (1998): Evolution of Clark’s unit graph method to spatially distributed runoff. Journal of Hydrologic Engineering 3, 1, 919. Kurz, W., Beukema, S., Klenner, W., Greenough, J., Robinson, D., Sharpe, A., and Webb, T. (2000): TELSA: the tool for exploratory landscape scenario analyses. Computers and Electronics in Agriculture 27, 227242. Ladue Horton Watkins High School. Our rain garden. the issue: Shrinking wetland and aquatic habitats. http://www.teamleaf.org/aqua/ourraingarden.html, Access: 08-03-2010. Lastra, J., Fernández, E., and Marquínez, A. (2008): Flood hazard delineation combining geomorphological and hydrological methods: an example in the Northern Iberian Peninsula. Natural Hazards 45, 277293. Lillesand, T., Kiefer, R., and Chipman, J. (2004): Remote sensing and image interpretation, fifth edition ed. Wiley & Sons, New York etc. Lucieer, A. and Stein, A. (2002): Existential uncertainty of spatial objects segmented from satellite sensor imagery. IEEE Transactions on Geoscience and Remote Sensing 40, 11, 25182521. Luers, A., Lobell, D., Sklar, L., Addams, C., and Matson, P. (2003): A method for quantifying vulnerability, applied to the agricultural system of the Yakie Valley, Mexico. Global Environmental Change 13, 255267. Luhmann, N. (1991): Soziologie des Risikos. de Gruyter, Berlin. Mapcity (2005): Mapas de riesgo de inundación para el Gran Santiago. Análisis de Información de prensa El Mercurio 29 de agosto de 2005, AACh, Mapcity, Santiago de Chile. McDowell, W. and Eames, M. (2006): Forecasts, scenarios, visions, backcasts and roadmaps to the hydrogen economy: A review of the hydrogen futures literature. Energy Policy 34, 11, 12361250. Meinel, G. and Neubert, M. (2004): A comparison of segmentation programs for high resolution remote sensing data. International Archives of Photogrammetry and Remote Sensing 35, 10971105. Merkel, W., Kaushika, R., and Gorman, E. (2008): NRCS GeoHydro|A GIS interface for hydrologic modeling. Computers & Geosciences 34, 8, 918930. Merwade, V. (2008): Creating SCS Curve Number GRID using HEC-GeoHMS. School of Civil Engineering, Purdue University.

References

215

Messner, F. and Meyer, V. (2006): Flood risk management  Hazards, vulnerability and mitigation measures, 1st ed. Springer, Flood damage, vulnerability and risk perception  challenges for flood damage research, pp. 149166. Metz, K. and Weiland, U. (2009): Analysis of flood risk prevention systems in Santiago de Chile and the Metropolitan Region, Helmholtz-Centre for Environmental Research (UFZ), Leipzig; Dept. Urban Ecology, Environmental Planning and Transport. Meyer, V., Scheuer, S., and Haase, D. (2009): A multicriteria approach for flood risk mapping exempliéd at the Mulde river, Germany. Natural Hazards 48, 1, 1739. MINVU (2005): Plan Regulador Metropolitano de Santiago. Ordenanza, Secretaría Ministerial Metropoli tana de Vivienda y Urbanismo  Unidad Desarrollo Urbano  Área Planificación, Santiago de Chile. Mioc, D., Nickerson, B., MacGillivray, E., Morton, A., Anton, F., Fraser, D., Tang, P., and Liang, G. (2008): Early warning and mapping for flood disasters. In The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4, Beijing. Mishra, S. and Singh, V. (2004): Long-term hydrological simulation based on the Soil Conservation Service curve number. Hydrological Processes 18, 12911313. Mockus, V. (1972): National Engineering Handbook. Section 4: Hydrology, 2 ed. Soil Conservation Service (SCS), Estimation of direct runoff from storm rainfall. Chapter 10. Mockus, V., Werner, J., Woodward, D., Nielsen, R., Dobos, R., Hjelmfelt, A., and Hoeft, C. (2007): Part 630 Hydrology National Engineering Handbook. United States Department of Agriculture, Natural Resources Conservation Service, Hydrological Soil Groups. Molenaar, M. (1998): An introduction to the theory of spatial object modeling for GIS. Routledge. Municipalidad La Reina (2008): Plan de Desarrollo de La Reina 20082012, Municipalidad La Reina, Santiago de Chile. Municipalidad Peñalolén (2006): Plan de Desarrollo Comunal de Peñalolén (PLADECO), Municipalidad Peñalolén, Santiago de Chile. Naef, F., Scherrer, S., and Weiler, M. (2002): A process based assessment of the potential to reduce flood runoff by land use change. Journal of Hydrology 267, 7479. Naranjo, J. and Varela, J. (1996): Flujos de Detritos y Barro que afectaron al Sector Oriente de Santiago el 3 de mayo de 1993, vol. Boletin No. 47. Servicio Nacional de Geología y Minería  Chile, Santiago de Chile. Nash, J. and Sutcliffe, J. (1970): River flow forecasting through conceptual models Part I|A discussion of principles. Journal of Hydrology 10, 3, 282290. Niehoff, D., Fritsch, U., and Bronstert, A. (2002): Land-use impacts on storm-runoff generation: scenarios of land-use change and simulation of hydrological response in a meso-scale catchment in SW-Germany. Journal of Hydrology 267, 12, 8093. Nirupama, N. and Simonovic, S. (2002): Role of remote sensing in disaster management. Research Report Paper Series  No. 21, Institute for Catastrophic Loss Reduction, London, ON, 11/05. Reyes Paecke, S. and Ebert, A. [Müller, A.] (2009): Workshop “Planificación territorial y usos del suelo en Santiago 2030”. Stakeholder workshop to discuss the future development of Santiago de Chile with respect to land use, land use management, and resulting risks, Santiago de Chile, November 2. Pal, N. and Pal, S. K. (1993): A review on image segmentation techniques. Pattern Recognition 26, 9, 12771294.

216

References

Paredes Herranz, M. (2009): MINVU/SEREMI. Architect, Member of the Planning Unit of Santiago de Chile Metropolitan Region. Interview carried out by A. Ebert (A. Müller) on April 17. Pelling, M. (1997): What determines vulnerability to floods; a case study in Georgetown, Guyana. Environment and Urbanization 9, 1, 203226. Pelling, M. (2003): The vulnerability of cities: natural disasters and social resilience. Earthscan, London etc. Perez, J. (2009): Riesgo de inundación producto del cambio climático. Caso de estudio: Quebrada de Ramón. Master’s thesis, Universidad de Chile, Santiago de Chile. Petermann, A. (2006): Santiago. Dónde estamos y dónde vamos. Centro de Estudios Publicos, Santiago de Chile, ¿Quién extendió a Santiago? Una breve historia del límite urbano, 19531994., pp. 205230. Piepho, B. (2003): Untersuchungen zum hydrologischen Reaktionsverhalten eines kleinen Einzugsgebietes unter Verwendung des Wasserhaushaltsmodells WaSiM-ETH. Eine Fallstudie im Rahmen des IPMS-Projektes in Nepal. PhD thesis, Technischen Universität CaroloWilhelmina zu Braunschweig, Braunschweig. Pietikäinen, M. (2005). Image analysis with local binary patterns. In SCIA 2005, H. Kalviainen et al., Ed., vol. LNSC 3540, Springer-Verlag Berlin Heidelberg, pp. 115118. Pilon, P. (2004): Guidelines for reducing flood losses, United Nations, Geneva, New York, Bangkok, Harare. Poduje, I. (2006): Santiago. Dónde estamos y dónde vamos. Centro de Estudios Publicos, Santiago de Chile, El globo y el acordeón: planificación urbana en Santiago, 19602004, pp. 231276. Polsky, C., Schröter, D., Patt, A., Gaffin, S., Martello, M., Neff, R., Pulsipher, A., and Selin, H. (2003): Assessing vulnerabilities to the effects of global change: An eight-step approach, Belfer Center for Science and International Affairs, Kennedy School of Government, Harvard University, 08/06. Protege (2010): Proyecto de Conservación y Protección de la Cordillera de Santiago de Chile, http://www.protege.cl, Santiago de Chile. Quezada, M. (2009): SECPLAN La Reina, Santiago de Chile, Director of the comunal planning unit. Interview carried out by A. Ebert (A. Müller) on March 23. Raaijmakers, R., Krywkow, J., and van der Veen, A. (2008): Flood risk perceptions and spatial multi-criteria analysis: an exploratory research for hazard mitigation. Natural Hazards 46, 307322. Rahman, M. and Alkema, D. (2007): Digital surface model (DSM) construction and flood hazard simulation for development plans in Naga City, Philippines. GISdevelopment. Ranganath, P. (2000): Mitigation and the consequences of international aid in post disaster reconstruction. Natural Hazards ResearchWorking Paper 103, McGill University & University of Colorado. Rashed, T. and Weeks, J. (2003): Assessing vulnerability to earthquake hazards through spatial multicriteria analysis of urban areas. International Journal of Geographical Information Science 17, 6, 547576. Ray, R. and Jacobs, J. (2007): Relationships among remotely sensed soil moisture, precipitation and landslide events. Natural Hazards 43, 211222. Refsgaard, J. (1996): Distributed Hydrological Modelling. Kluwer Academic Publishers, Terminology, modelling protocol and classification of hydrological model codes, pp. 1739. Reiter, J. (2009): Vulnerabilität gegenueber ueberutungen in ausgewählten Stadtteilen von Santiago de Chile. Master’s thesis, Humboldt-Universität zu Berlin.

References

217

Retamal, J. and Estellé, L. (2009): Ministerio de Obras Publicas, Dirección de Obras Hidraulicas, Director of the comunal planning unit. Santiago de Chile. Interview carried out by A. Ebert (A. Müller) and G. Schleenstein on April 16. Reyes Paecke, S. (2003): Perspectivas del medio ambiente urbano. GEO Santiago, PNUMA/UNEP, Santiago de Chile. Richards, J. and Jia, X. (1999): Remote sensing digital image analysis: an introduction, 3rd ed. Springer Verlag, Berlin. Ridd, M. K. (1995): Exploring a V-I-S (Vegetation-Impervious Surface-Soil) model for urban ecosystem analysis through remote sensing: Comparative anatomy for cities. International Journal of Remote Sensing 16, 21652185. Ripley, B. (1996): Pattern Recognition and Neural Networks, 1st ed. Cambridge University Press. Romero, H. (2008): Professor for Geography. Universidad de Chile, Santiago de Chile. Interview carried out by A. Ebert (A. Müller) on May 20. Romero, H., Fuentes, C., López, C., Smith, P., and Vásquez, A. (2008): Environmental effects of the recently urban sprawl proposal at Santiago de Chile: Urban governance and public policy under a neoliberal context. In 3rd International Seminar on Environmental Planning and Management. Romero, H., Ihl, M., Rivera, A., Zalazar, P., and Azocar, P. (1999): Rapid urban growth, land-use changes and air pollution in Santiago, Chile. Atmospheric Environment 33, 40394047. Romero, H., Molina, M., Moscoso, C., Sarricolea, P., Smith, P., and Vásquez, A. (2007): Santiago de Chile. Movilidad espacial y Reconfiguracion Metropolitana. Eure libros  Geolibros, Caracterización de los cambios de usos y coberturas de suelos causados por la expansión urbana de Santiago, análisis estadístico de sus factores explicativos e inferencias ambientales. Romero, H. and Vásquez, A. (2005). Evaluación ambiental del proceso de urbanización de las cuencas del piedemonte andino de Santiago de Chile. EURE XXXI, 94, 97118. Saa, R., Rodríguez, J., Contreras, M., Opazo, D., Montecinos, T., and Ibáñez, J. (2005): Atlas socioeconómico Regón Metropolitana de Santiago. Proyecto “Bases para un Ordenamiento Territorial Ambientalmente Sustentable para la Región Metropolitana de Santiago” (OTAS), Universidad de Chile, Gobierno Regional Metropolitano de Santiago, Gesellschaft für Technische Zusammenarbeit, Santiago de Chile. Sandholt, I., Nyborg, L., Fog, B., Lo, M., Bocoum, O., and Rasmussen, K. (2003): Remote sensing techniques for flood monitoring in the Senegal River Valley. Danish Journal of Geography 103, 1, 7181. Scharffenberg, W. and Fleming, M. (2008): Hydrologic modelling system HEC-HMS. User’s guide. User’s Manual 0704-0188, US Army Corps of Engineers, Davis. Scheuren, J.-M., le Polain de Waroux, O., Below, R., Guha-Sapir, D., and Ponserre, S. (2007): Annual Disaster Statistical Review. The Numbers and Trends 2007, CRED, UN-ISDR, UCL. Schiewe, J. (2002): Segmentation of high-resolution remotely sensed data  concepts, applications and problems. In Joint ISPRS Commission IV Symposium: Geospatial Theory, Processing and Applications (July 912). Schneiderbauer, S. (2007): Risk and Vulnerability to Natural Disasters – from Broad View to Focused Perspective. Theoretical background and applied methods for the identification of the most endangered populations in two case studies at different scales. Phd, Freie Universität Berlin. Siebert, A. (2004): Risikomanagement von Naturgefahren in der Versicherungswirtschaft, München.

218

References

Silva, R. (2008): Desbordes terminarían con ensanche del canal. Entre Urbanos May 27th, Santiago de Chile. Slack, R. and Welch, R. (1980): Soil Conservation Service runoff curve number estimates from LandZat data. Water Resources Bulletin. American Water Resources Association 16, 5, 887893. Solway, L. (2004): Natural disasters and sustainable development. Springer, Reducing the effect of natural hazards on urban areas, pp. 303338. Steinführer, A., Kuhlicke, C., de Marchi, B., Scolobig, A., Tapsell, S., and Tunstall, S. (2009): Local Communities at Risk from Flooding. Social Vulnerability, Resilience and Recommendations for Flood Risk Management in Europe. 88 p., Leipzig. Helmholtz-Centre for Environmental Research. Steinmüller, K. (1997): Grundlagen und Methoden der Zukunftsforschung. WerkstattBericht 21, SFZ, Gelsenkirchen. Stern, N. (2007): The Economics of Climate Change. The Stern Review. Great Britain Treasury. Cambridge University Press. Stow, D., Lopez, A., Lippitt, C., Hinton, S., and Weeks, J. (2007): Objectbased classification of residential land use within Accra, Ghana based on QuickBird satellite data. International Journal of Remote Sensing 28, 51675173. Stumpf, A. (2009): Landslide susceptibility mapping in Central Chile. Application of a probabilistic, GIS-based method at eastern Santiago de Chile and the bordering Andes. Diploma thesis, Dresden Technical University. Suzuki, H., Matsakis, P., Andrefouet, S., and Desachy, J. (2001): Satellite image classification using expert structural knowledge: A method based on fuzzy partition computation and simulated annealing. In IAMG 2001, Cancun, Mexico. Taubenböck, H. (2007): Vulnerabilitätsabschätzung der erdbebengefährdeten Megacity Istanbul mit Methoden der Fernerkundung. PhD thesis, Bayerischen Julius- Maximilians Universität Würzburg. Taubenböck, J., Post, J., Roth, A., Zosseder, K., Strunz, G., and Dech, S. (2008): A conceptual vulnerability and risk framework as outline to identify capabilities of remote sensing. Natural Hazard and Earth System Science 8, 409420. Tekeli, Y. I., Akguel, S., Dengiz, O., and Akuezuem, T. (2007): Estimation of flood discharge for small watershed using SCS curve number and geographic information system. In River Basin Flood Management. Chapter IV: Flood Forecasting Models Integrated Flood Management Applications (Antalya, Turkey, March 2224), International Congress on River Basin Management, DSI, General Directorate of State Hydraulic Works. Thiel, C., Gruenler, S., Herold, M., Hochschild, V., Jäger, G., and Hellmann, M. (2001): Interpretation and Analysis of Polarimetric L-Band E-SAR-Data for the Derivation of Hydrologic Land Surface Parameters. In Proceedings CD of IEEE 2001 International Geoscience and Remote Sensing Symposium IGARSS’01, Scanning the Present and Resolving the Future (Sydney, Australia, July 913). Turner, B., Kasperson, R., Matsone, P., McCarthy, J., Corell, R., Christensen, L., Eckley, N., Kasperson, J., Luerse, A., Martello, M., Polsky, C., Pulsipher, A., and Schiller, A. (2003): A framework for vulnerability analysis in sustainability science. PNAS 100, 14, 80748079. Tyrna, B. and Hochschild, V. (2010): Modellierung von lokalen Überschwemmungen nach Starkniederschlägen  reviewed paper. In Angewandte Geoinformatik 2010. Beiträge zum 22.

References

219

AGIT-Symposium Salzburg (Berlin), J. Strobl, T. Blaschke, and G. Griesebner, Eds., Wichmann Verlag, pp. 325334. UN-HABITAT (1995): Guidelines for settlement planning in areas prone to flood disasters HS/384/95E, United Nations Centre for Human Settlements, Nairobi. UN-HABITAT (2008): State of the World’s Cities 2008/2009  Harmonious Cities 4th edition, United Nations Human Settlements Programme (UNHABITAT), London, Sterling VA, 259 p. UN-ISDR (2005): Hyogo framework for action 20052015: Building resilience of nations and communities to disasters. In World Conference on Disaster Reduction (Kobe, Hyogo, Japan), UN/ISDR, p. 25. UNDP (2004): A Global Report. Reducing Disaster Risk. Challenge for Development, UNDP, New York, 05/06. Universidad Politecnica de Catalunya. BarcelonaTech. Using HEC-HMS and HEC-GeoHMS with Precipitation Grids outside the Conterminous US. UNU-EHS (2005): Verwundbarkeit messbar machen. Advancing human security through knowledge-based approaches to reducing vulnerability and environmental risk (Bonn), 6. USACE (2003): Geospatial Hydrologic Modeling Extension HEC-GeoHMS. User’s Manual, 1.1 ed. United States Army Corps of Engineers, Davis. USACE (2006): GageInterp. A Program for Creating a Sequence of HEC-DSS Grids from TimeSeries Measurements. User’s Manual. US Army Corps of Engineers, Davis. USDA (1986): Urban hydrology for small watersheds. TR-55 210-VITR- 55, United States Department of Agriculture, National Resources Conservation Service, Conservation Engineering Division. USDA (1999): Soil Taxonomy. A basic system of soil classification for making and interpreting soil surveys. United States Department of Agriculture & Natural Resources Conservation Service, Washington, D.C. van der Sande, C., de Jong, S., and de Roo, A. (2003): A segmentation and classification approach of IKONOS-2 imagery for land cover mapping to assist flood risk and flood damage assessment. International Journal of Applied Earth Observation and Geoinformation 4, 217229. van Herzele, A. and Wiedemann, T. (2003): A monitoring tool for the provision of accessible and attractive urban green spaces. Landscape and Urban Planning 63, 109126. van Notten, P., Rotmans, J., van Asselt, M., and Rothman, D. (2003): An updated scenario typology. Futures 35, 423443. van Westen, C., Kumar Piya, B., and Guragain, J. (2005): Geoinformation for urban risk assessment in developing countries; the SLARIM project. In 1st International Symposium on GeoInformation for Disaster Management (Delft), P. van Oosterom, S. Zlatanova, and M. Elfriede, Eds., ITC, The Netherlands; Department of Mines and Geology, Nepal; Department of Local Development, Nepal, pp. 379392. Vargas, X. (1999): Corrientes de detritos en la Quebrada de Macul, Chile. Estudio de caudales máximos. Ingeniería del Agua 6, 4, 341344. Vásquez, A., Salgado, M., and Romero, H. (2007): Desigualdades socioeconómicas y distribución inequitativa de los peligros ambientales en las comunas de Peñalolen y San Pedro de la Paz. Coloquio Del País Urbano a País Metropolitano, Diciembre 46, Santiago de Chile. Vásquez, A., Schmidt, A., and Banzhaf, E. (2010): Land-use and land-cover classification of the Aster images from 1993, 2002, and 2009. Internal Working Paper with maps, Helmholtz Centre for Environmental Research – UFZ, Leipzig.

220

References

Velasquez, G. and Ma T Tanhueco, R. (2005): Know Risk. United Nations ‘World Conference on Disaster Reduction’, Incorporating social issues in disaster risk assessment. Villagrán de León, J. (2006): Vulnerability. A conceptual and methodological review. Source 4/2006, United Nations University, UNU-EHS, Bonn. Vos, F., Rodríguez, J., Below, R., and Guha-Sapir, D. (2009): Annual disaster statistical review 2009. The numbers and trends, CRED, WHO, UCL. Wagner, K. (2008): Der Risikoansatz in der europäischen Hochwassermanagementrichtlinie. Natur und Recht 30, 774779. Walter, R. (2005): Politics and Urban Growth in Santiago, Chile, 18911941. No. 0-8047-4982-5. Stanford University Press. Watson, R. and Albritton, D. (Core Writing Team) (2002): Climate Change 2001: Synthesis Report: Third Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press. Weichel, T., Pappenberger, F., and Schulz, K. (2007): Sensitivity and uncertainty in flood inundation modelling  concept of an analysis framework. Advances in Geosciences 11, 3136. Weischet, W. (1970): Chile. Seine Länderdkundliche Individualität und Struktur. Wissenschaftliche Länderkunden Bd. 3. Weischet, W. (1996): Regionale Klimatologie. Teil 1. Die Neue Welt: Amerika, Neuseeland, Australien. G.B. Teubner Verlag, Stuttgart. Welz, J. (2005): Die CEPAL  Wirtschaftskommission für Lateinamerika und Karibik. Die Zensus von Lateinamerika  eine vergleichende Analyse. unpublished. Wenzel, F., Bendimerad, F., and Sinha, R. (2007): Megacities  megarisks. Natural Hazards 42, 481491. Wisner, B., Blaikie, P., Cannon, T., and Davis, I. (2004): At Risk. Natural hazards, people's vulnerability and disasters. 2nd ed. Routledge, London, New York. Woodward, D., Hawkins, R., Jiang, R., Hjelmfelt Jr., A., Van Mullem, J. A., and Quan, Q. (2003): Runoff curve number method: Examination of the initial abstraction ratio. In World Water & Environmental Resources Congress. World Bank (2006): Urban prevention and drainage project, TheWorld Bank, Washington, D.C. Zhang, Y. and Maxwell, T. (2006): A fuzzy logic approach to supervised segmentation for objectoriented classification. In ASPRS 2006 Annual Conference, Reno, Nevada, May 15.

APPENDIX 1 – ADDITIONAL TABLES AND CHARTS APPENDIX 1 – ADDITIONAL TABLES AND CHARTS The water cycle: The main processes of the water cycle are precipitation (P), runoff (Q), and evapotranspiration (ET). Their relation can be expressed as: P = Q + ET + ǻS

Equation 17

with ǻS indicating changes in the storage. These main processes are also schematized in Figure 1. Runoff can be further distinguished into surface runoff (Hortonian flow) and subsurface runoff (interflow). The amount of surface runoff is determined by the infiltration capacities of the soil, which in turn highly depend on the surface porosity. The higher the porosity, the higher the infiltration capacity. The rainwater that is infiltrating the soil either flows towards the stream in a permeable layer below the surface (as interflow) or further percolates through the soil until it reaches the groundwater table. If the inflow to the groundwater is high during a storm, the ground water table might rise until it reaches the Earth’s surface. The water thus accumulates above the soil and  depending on the relief  moves towards the stream as saturated overland flow. If the amount of rainfall directly exceeds the infiltration capacities of the soil, Hortonian overland flow is generated. Large amounts of Hortonian flow are generated in largely impervious areas, e.g. in dense urban settings. The phenomenon of accumulation of water resulting from a surplus of Hortonian overland flow is referred to as urban flooding. If the accumulated runoff in a channel exceeds its physical capacities it overflows its banks. The result is a river flood. Floods are part of the natural water cycle but have already posed a threat to the early human settlements and will in the future still put at risk people and their assets, the urban (social and economic), and the ecological environment. Evapotranspiration as the second main component equals the sum of transpiration (under the influence of active life) and evaporation (from water or non-vegetated surfaces). As during storms the humidity reaches 100 % and the soil becomes saturated the amount of evapotranspiration in that period can be neglected (Scharffenberg & Fleming 2008, p. 33). The amount of rainfall that neither infiltrates nor runs off directly or delayed as subsurface and groundwater flows is stored on the surface and thus leads to a reduction of the runoff (ǻS). The rainfall stored on the vegetation cover is called interception. Water can also be stored in other natural (glacier, snow, lakes, etc.) or man-made retention ponds (dams, irrigation ponds, etc.). Thus, the most relevant components and processes for the analysis of the direct consequences of a storm event with respect to floods are: (i) the amount of rainfall, (ii) the amount of infiltration, (iii) the amount of interception, and (iv) the amount of surface runoff (Figure 1) (Dyck & Peschke 1995, Im et al. 2009).

222

Appendix 1 – Additional Tables and Charts

Table A34: Overview about floods in Santiago between 1990 and 2002. Location

Effects and damage

Cause

Quebrada de Macul, Eastern part of Peñalolén, Length: 1.5 km last event: May 1993

Flooding and destruction of residences and infrastructure by sediment fluxes in river Damage: 13 km of roads 26,000 people affected Flood in densely populated 2.5 km² of Ñuñoa & La Reina Damage: 4 km of roads 7,000 people affected Undercutting of river banks and floods of adjacent land (5075 ha) 7,000 people affected Floods of adjacent land (150180 ha) 15,000 people affected

Mass movements Construction in high risk areas, narrowing of river as a result of construction activities Insufficient canalization system, accumulation of sediments from higher part of catchment Insufficient capacity of channel

Quebrada San Ramón, La Reina Length: 5 km Zanjón de la Aguada Lo Errázuriz Zanjón de la Aguada between San Carlos Channel and Carmen St. Length: 8 km Río Mapocho Between La Dehesa bridge and San Enrique Bridge Length: 3 km Río Mapocho Between Bulnes bridge and Pudahuel bridge Length: 11 km Area Metropolitana de Santiago Southern and Western zones, July 2000 Area Metropolitana de Santiago June 2002

Las Cruces Creek northern part of Quilicura Length: 15 km

Undercutting of river banks(8 ha affected), 2 storm water detention basins damaged 2,000 people affected Erosion of vegetation along embankment Damage: 200 km of roads 12,000 people affected Storm water accumulation in the urban area 2,000 people affected Floods in urban and rural areas, Damage: 250 km of roads, low lying areas, metro stations, 9,100 households 80,000 people affected Periodic floods in this area(835 ha agricultural area) 800 people affected

Table A35: Projection transformation parameters. dx dy dz

-328 340 -329

Insufficient capacity of channel, Canalization & covering of channel (20022003) Settlement in flood plains

Dikes existing but not sufficient protection Occurred during road construction works Lack of storm water collectors

Heavy precipitation, Lack of storm water collectors and insufficient capacity of existing system Insufficient capacity because of vegetation and debris in river bed and narrowing of creek

223

Appendix 1 – Additional Tables and Charts Table A36: Partners for the expert interviews during field work stays in May/June 2008, March/April 2009 and November 2009. Name/Profession

Institution

Topic

Interview type

Alberto Carvacho Architect, Head of Planning Unit M. E. Paredes Architect Daniel Carvacho Architect

MINVU/SEREMI Depto. Desarrollo Urbano MINVU/SEREMI Unidad Planificación Universidad de Chile

PRMS Urban planning, Scenarios

Semi- structured

Semi- structured

Marcos Rivera Head of Dept. Planificación y Gestión del Territorio Pablo Fuentes Architect, Head of Regional Planning Department Gerardo Ubilla Geographer Carolina Infante Architect R. Katscher

MIDEPLAN

Article 8 of the PRMS Housing development in flood hazard zones Vulnerability

Marcela Quezada Planner Veronica Saud Architect Mónica Aguilera Head of municipal emergency section Luis Estellé Head of Dpt. Jaime Retamal Ing. Civil Milo Millan Head of Dept. Cauces y Drenaje Urbano Leonardo Céspedes Consultant of SUBDERE

Narrative

Questionnaire

GORE-DIPLADE División Planificación& y Desarrollo Regional GORE-DIPLADE

Planning, Scenarios

Semi- structured

PRMS, & planning

GORE

Bicentenario

Semi- structured, Questionnaire Questionnaire

GORE

Questionnaire

SECPLAN La Reina

Participative planning PRC La Reina

SECPLAN La Reina

PRC La Reina

Municipalidad San Bernardo

Vulnerability

Questionnaire Semi-structured Questionnaire

MOP-DOH Dept. de Proyectos de AguasLluvias MOP-DOH Dept. de Proyectos de Aguas Lluvias MOP-DOH

Storm water management

Semi- structured

Storm water management

Semi- structured,Quetionnaire

Storm water infrastructure

Questionnaire

SUBDERE

Risk-prone areas

Narrative

Semi- structured

224

Appendix 1 – Additional Tables and Charts

Marcelo Canales Geographer

Ingeniería 4 Consultancy

G. Schleenstein Engineer James McPhee Civil Engineer Javiera Perez Civil Engineer Hugo Romero Geographer Alexis Vásquez Geographer Sonia Reyes Biologist F. Baeriswyl Head of Dept. Natural resources protection Gonzalo Garcia Head of Technical Division Felipe Bañados

Ingeniería Alemana S.A. U de Chile, Dept. De Ing. Civil Universidad de Chile

Risk study Piedmont (AC Ingenieros 2008) Stormwater infrastructure Hydrology, Modeling San Ramon Modeling Urban environment

Narrative

Narrative Narrative Narrative

Universidad de Chile Dept. de Geografía Universidad de Chile, Dept. de Geografía PUC

Narrative

Peñalolén Environm. justice Urban ecology

Narrative

SAG

Vulnerability

Questionnaire

Parque Metropolitana Santiago

Vulnerability

Questionnaire

Protege

San Ramón basin

Questionnaire

Narrative

Table A37: Preprocessing steps to be performed using HEC-GeoHMS. Process 1 Fill Sinks 2 Flow Direction 3 Flow Accumultion 4 Stream Definition

5 Stream Segmentation 6 Watershed Delineation 7 Watershed Polygon Processing 8 Stream Segment Processing 9 Watershed Aggregation

Input data

Parameter

Output

original DEM Fillgrid FDirgrid

-

FillGrid FDirGrid FAccGrid

FAccGrid

StrmGrid

StrmGrid and FDirgrid

map units: meters min area: 2.00000e+016 -

StrLinkGrid

FDirgrid StrLnkGrid

-

WatShdGrid

WatShdGrid

-

WatShdShp.shp

WShdGrid FDirGrid

-

Rivers.shp

Rivers.shp WatShedShp.shp

-

WatShedmg.shp

Appendix 1 – Additional Tables and Charts

225

Table A38: Albers projection file (prj.adf) to enforce proper projection information. Text File Content ALBERS Projection NAD83 Zone NO Zunits METERS Units GRS1980 Spheroid Parameters 29 30 0.000 /* 1st standard parallel 30 45 30 0.000 /* 2nd standard parallel -96 0 0.000 /* central meridian 0 0 0.000 /* latitude of projection’s origin 0.00000 /* false easting (meters) 0.00000 /* false northing (meters)

Table A39: Header information of Curve Number (CN) ASCII-file for the specific study area. ASCII-file header ncols 149 nrows 172 xllcorner 322750 yllcorner 4006550 cellsize 50 NODATA_value -9999

226

Appendix 1 – Additional Tables and Charts

Figure A34: Workflow showing the single steps necessary for the delineation of the ASCII-file containing curve numbers with ArcView 3.2 and the extension HEC-GeoHMS.

Appendix 1 – Additional Tables and Charts Table A40: Rule set used for the LULC classification of the Quickbird data. Rule Set Documentation Classes: Agricultural areas and (min) Threshold: Num. of overlap: Agriculture > 0 0.40.6: NDVI Barren land or (max) and (min) 600800: Mean IR 8001000: Mean Green 800850: Mean Elevation -0.30.15: NDVI and (min) -0.30.15: NDVI 600800: Mean IR 8001000: Mean Green 1.31131e+0131.31131e+013: Census Code: Manzanas Built-up and (min) Threshold: Num. of overlap: NoBuiltup < 1 0.91.1: NDVI 850950: Mean Green Bushland and (min) 22002400: Mean IR 0.40.7: NDVI Dark gray and black roofs and (min) 01000: Mean IR (generated) 175950: Mean Blue (generated) 01: Census Code: Manzanas Dry vegetation or (max) and (min) 0.150.5: NDVI 800850: Mean Elevation and (min) 0.150.5: NDVI 01: Census Code: Manzanas Grassland 2000-2200: Mean IR 0.5-0.85: NDVI 900-1100: Mean Elevation

227

228

Appendix 1 – Additional Tables and Charts Gray-blue roofs and (min) 900-2100: Mean Red (generated) 900-2300: Mean Blue (generated) 950-2350: Mean Green (generated) 1900-2100: Mean IR Greenish roofs and (min) 0.08-0.1: NDVI 800-820: Mean Red 225-1600: Mean Green (generated) 300-1250: Mean Blue (generated) 0-1: Census Code: Manzanas 300-2500: Mean IR (generated) Impervious No description Light to medium gray roofs and (min) 0-1: Census Code: Manzanas 700-1350: Mean Green (generated) 300-1500: Mean IR (generated) 650-1500: Mean Red (generated) Open rock and (min) 475-800: Mean Blue (generated) 500-815: Mean Green (generated) 330-650: Mean IR (generated) Sand and (min) 1.3113e+013-1.3113e+013: Census Code: Manzanas 750-1300: Mean Red (generated) 600-1000: Mean Blue (generated) 700-1200: Mean Green (generated) 450-1100: Mean IR (generated) Shadow and (min) 0-1214.36: Mean IR (generated) 0-3.01292: Mean Dissimilarity (generated) 0-1981.71: Mean Blue (generated)

Appendix 1 – Additional Tables and Charts Streets or (max) and (min) 8001200: Mean Elevation 11001300: Mean IR 1.92.1: Length/Width and (min) 8251200: Mean Elevation 01: not Census Code: Manzanas 11001300: Mean IR Swimming pool and (min) 160260: Brightness 2.53: Length/Width 10001200: Mean Elevation 4503500: Mean Blue (generated) -1730: Mean IR (generated) 500600: Mean Red 390410: Mean Green Trees and (min) 0.51: NDVI 9002500: Mean IR Unpaved no description Unpaved park and (min) -0.10.1: NDVI 0.50.6: Rel. border to Gray-blue roofs 0.50.6: Rel. border to Greenish roofs 12802300: Mean Red (generated) 10001900: Mean Blue (generated) 12002100: Mean IR (generated) 0.50.6: Rel. border to Light to medium gray roofs Urban unpaved and (min) -0.10.25: NDVI 7001500: Mean Red (generated) 9002100: Mean IR (generated) 5001250: Mean Blue (generated) 6151400: Mean Green (generated) Vegetation no description

229

230

Appendix 1 – Additional Tables and Charts

Very bright roofs and (min) 20002100: Mean Blue 20002100: Mean Green 01: Census Code: Manzanas Water bodies and (min) 450550: Mean IR (generated) 375600: Mean Green (generated) 0650: Mean Red (generated) Customized Features: NDVI: ([Mean IR]-[Mean Red])/([Mean IR]+[Mean Red]) Process: Main: do Segmentation delete image object level: on main at Level1: delete delete image object level: on reclass at Level1: delete delete image object level: on H-UST at Level1: delete delete image object level: on General H-UST at Level1: delete delete image object level: on General H-UST at Manzana: delete delete image object level: on H-UST at Level1: delete delete image object level: on H-UST at Manzana: delete multiresolution segmentation: 30 [shape:0.1 compct.:0.3] creating ‘Level1’ Classification remove classification: at Level1: remove classification remove classification: at Level1: remove classification classification: at Level1: Built-up classification: Built-up at Level1: Greenish roofs, Blue-gray roofs, Light to medium gray roofs, Very bright roofs assign class: Built-up at Level1: unclassified

Appendix 1 – Additional Tables and Charts

231

classification: unclassified at Level1: Agricultural areas, Barren land, Bushland, Dry vegetation, Dark gray and black roofs, Grassland, Open rock, Sand, Shadow, Streets, Swimming pool, Trees, Unpaved park, Urban unpaved, Water bodies assign class: Greenish roofs with Rel. border to Barren land >= 0.6 at Level1: Barren land assign class: Dark gray and black roofs with Rel. border to Barren land >= 0.9 at Level1: Barren land classification: unclassified with Rel. border to Trees >= 0.5 at Level1: Trees classification: unclassified with Rel. border to Bushland >= 0.25 at Level1: Bushland classification: 3x: unclassified with Rel. border to Streets >= 0.35 at Level1: Streets classification: 2x: unclassified with Rel. border to Swimming pool >= 0.25 at Level1: Swimming pool classification: 2x: unclassified with Rel. border to Blue-gray roofs >= 0.25 at Level1: Bluegray roofs Reclassification copy map: on main : copy map to ‘reclass’ assign class: on reclass Bushland, Dry vegetation, Grassland, Trees at Level1: Vegetation

232

Appendix 1 – Additional Tables and Charts

Table A41: Individual parameters used for each subbasin in the hydrologic model HEC-HMS. Compare with Figure 24 to locate the subbasins based on their numbers. Parameter

Subbasin number

Initial abstraction ratio (Ia)

R900W900 R910W910 R840W840 R790W790 R780W780 R750W750 R690W690 R740W740 R660W660 R680W680 R720W720 R900W900 R910W910 R840W840 R790W790 R780W780 R750W750 R690W690 R740W740 R660W660 R680W680 R720W720

Time of concentration (Tc) = Storage coefficient (R) = Muskingum K

Value May 2002 0.40 0.40 0.05 0.10 0.05 0.05 0.05 0.05 0.05 0.02 0.01 0.82 0.54 0.56 0.69 1.74 2.05 1.18 0.48 0.48 1.31 1.70

233

Appendix 1 – Additional Tables and Charts

Table A42: Sensitivity analysis for the parameters “Initial abstraction ratio” (Ia) and Retention scale factor. 90

91

84

Initial abstraction ratio 0.05 0.4 0.40 0.05 0.30 0.30 0.05 0.20 0.20 0.05 0.10 0.10 0.05 0.05 0.05 0.05 0.01 0.01 0.40 0.40 0.40 0.40 0.40 0.30 0.40 0.40 0.20 0.40 0.40 0.10 0.40 0.40 0.05 0.40 0.40 0.01 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 040 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05 0.40 0.40 0.05

79

78

75

69

74

66

68

72

Total [m³/s]

Peak [m³/s]

0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.40 0.30 0.20 0.10 0.05 0.01 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10

0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.40 0.30 0.20 0.10 0.05 0.01 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05

0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.40 0.30 0.20 0.10 0.05 0.01 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05

0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.40 0.30 0.20 0.10 0.05 0.01 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05

0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.40 0.30 0.20 0.10 0.05 0.01 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05

0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.40 0.30 0.20 0.10 0.05 0.01 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.05

0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.40 0.30 0.20 0.10 0.05 0.01 0.02 0.02 0.02

0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.40 0.30 0.20

26.25 26.52 26.84 27.20 27.41 27.58 24.94 25.20 25.52 25.98 26.25 26.49 25.69 25.86 26.05 26.25 26.36 26.45 25.98 26.03 26.11 26.20 26.25 26.29 25.22 25.47 25.76 26.08 26.25 26.39 25.97 26.01 26.06 26.16 26.21 26.27 24.68 25.01 25.39

21.7 21.9 22.2 22.5 22.6 22.8 20.4 20.7 21.1 21.5 21.7 21.9 21.2 21.4 21.5 21.7 21.8 21.9 21.6 21.6 21.6 21.7 21.7 21.7 20.7 21.0 21.3 21.6 21.7 21.8 21.4 21.4 21.5 21.6 21.7 21.7 20.3 20.7 21.1

234

Appendix 1 – Additional Tables and Charts

0.40 0.40 0.40

0.40 0.05 0.10 0.40 0.05 0.10 0.40 0.05 0.10 Retention scale factor 1 1 1 1 2 2 2 2 3 3 3 3 5 5 5 5 10 10 10 10

0.05 0.05 0.05

0.05 0.05 0.05

0.05 0.05 0.05

0.05 0.05 0.05

0.05 0.05 0.05

0.02 0.02 0.02

0.10 0.05 0.01

25.83 26.06 26.25

21.5 21.6 21.7

1 2 3 5 10

1 2 3 5 10

1 2 3 5 10

1 2 3 5 10

1 2 3 5 10

1 2 3 5 10

1 2 3 5 10

26.25 20.35 16.92 12.74 7.70

21.7 16.1 12.7 8.8 4.6

Table A43: Runoff volume for San Ramón catchment [m³/s] derived by AC Ingenieros (2008) and Perez (2009). Current conditions Return period 5 10 25 20 100

A & 2008 27.1 33.1 43.6 52.6 63.7

C

B2

A2

min

av

max

min

av

max

min

av

max

15.5 20.2 26.3 29.6 30.6

20.5 27.3 36.2 42.4 47.4

25.6 34.7 46.4 55.2 64.3

20.1 22.4 24.0 24.8 25.4

27.9 38.2 47.4 56.6 64.6

35.7 54.0 70.8 88.7 104.5

22.4 26.6 28.2 29.0 29.7

26.8 35.1 40.8 46.0 50.9

31.3 43.7 53.3 63.0 72.2

APPENDIX 2 – COLOURED FIGURES APPENDIX 2 – COLOURED FIGURES

Figure A35: The urban expansion of Santiago de Chile between 1970 and 2000. The black outlines show the area open for construction according to the Metropolitan Regulatory Plan from 2007.

236

Appendix 2 – Coloured Figures

Figure A36: The administrative units of the Metropolitan Region of Santiago de Chile. PRMS is the Metropolitan Regulatory Plan and is described further in Section 4.2.1.

Appendix 2 – Coloured Figures

FigureA37: The administrative units in Santiago de Chile at the example of the municipality (comuna) La Reina.

237

238

Appendix 2 – Coloured Figures

Figure A38: LULC changes in La Reina and Peñalolén between 1992 and 2009 and changes in the flood prone areas between 1987 and 2008.

Appendix 2 – Coloured Figures

239

Barren land in municipality of Peñalolén Newly constructed houses in the same area as along Avenida Antupirén in December 2006, example for urban densification. scale 1:50,000. UTM 357133.81, 6294222.45. Photo from November 2009. Figure A39: Example for urban land-use/land-cover changes in the municipality of Peñalolén.

240

Appendix 2 – Coloured Figures

Figure A40: Map showing the location of gaging stations in the eastern part of the Metropolitan Region of Santiago de Chile.

Appendix 2 – Coloured Figures

Figure A41: Overview over the extent and resolution of the four available satellite images.

241

242

Appendix 2 – Coloured Figures

Figure A42: Daily precipitation values for the stations Quebrada Ramón en Recinto Emos, Quebrada de Macul, Terraza Oficinas Centrales DGA, Antupiren, and Cerro Calán from 2008/04/01 to2008/07/ 31.

Figure A43: The process of the object-oriented classification approach (OOA) as used in this study. The main procedures comprise the segmentation of the satellite image and the consequent LULC classification with reference to the various image object features.

Appendix 2 – Coloured Figures

243

Figure A44: Land-use/land-cover classification based on ASTER sensor data for the catchment of San Ramón from February 2002, 2005, and 2009.

244

Appendix 2 – Coloured Figures

Figure A45: Result of the land-use/land-cover classification of the Quickbird satellite image from December 19, 2006. The result is the product of an object-oriented classification approach containing the classes as described in Table 20.

Appendix 2 – Coloured Figures

245

Figure A46: Standard Hydrological Grid for the year 2002 with cell size of 50 m generated through the pre-processing operations using HEC-GeoHMS. Gridded Curve Number values in 50 m grid as background information. High curve numbers represent low permeability.

246

Appendix 2 – Coloured Figures

Figure A47: Hydrograph for the different LULC patterns for the rainfall event in July 2001.

Figure A48: Hydrograph for the different LULC patterns for the rainfall event in May 2002.

Appendix 2 – Coloured Figures

Figure A49: Hydrograph for the different LULC patterns for the rainfall event in June 2002.

Figure A50: Hydrograph for the different LULC patterns for the rainfall event in June 2005.

247

248

Appendix 2 – Coloured Figures

Figure A51: Hydrograph for the different LULC patterns for the rainfall event in August 2005.

Figure A52: Hydrograph for the different LULC patterns for the first rainfall event in May 2008.

Appendix 2 – Coloured Figures

249

Figure A53: Hydrograph for the different LULC patterns for the second rainfall event in May 2008.

FigureA54: Hydrograph for the different LULC patterns for the rainfall event in September 2009.

250

Appendix 2 – Coloured Figures

FigureA55: Indicator sheet: Proportion of buildings located at or below street level per building block.

Appendix 2 – Coloured Figures

Figure A56: Indicator sheet: Proportion of households with poor construction material (wall, floor, roof material) per building block.

251

252

Appendix 2 – Coloured Figures

Figure A57: Indicator sheet: Proportion of green spaces per building block (inverse values: 1 minus proportion of green spaces).

Appendix 2 – Coloured Figures

Figure A58: Indicator sheet: Proportion of households with more than 2.5 people sharing one bedroom per building block.

253

254

Appendix 2 – Coloured Figures

Figure A59: Indicator sheet: Proportion of people with no or incomplete basic education per building block.

Appendix 2 – Coloured Figures

255

Figure A60: Indicator sheet: Proportion of people with no employment or no income per building block.

256

Appendix 2 – Coloured Figures

FigureA61: Indicator sheet: Proportion of female population per building block.

Appendix 2 – Coloured Figures

Figure A62: Indicator sheet: Proportion of people below five or above 65 years per building block.

257

258

Appendix 2 – Coloured Figures

Figure A63: Flood hazard map for a 100 year return period under current and climate change conditions for the area of La Reina after Perez (2009).

Figure A64: Changes in the extent of flood hazard zones for a 100 year return period under current and climate change conditions (IPCC scenario A2) for the area of La Reina after Perez (2009).

Appendix 2 – Coloured Figures

Figure A65: Available flood hazard maps for La Reina.

259

260

Appendix 2 – Coloured Figures

Figure A66: Distribution of people, built-up area and infrastructure in the municipalities of La Reina and Peñalolén.

Figure A67: Distribution of the combined amount of people, built-up area [ha], and infrastructure in the municipalities of La Reina and Peñalolén.

Appendix 2 – Coloured Figures

261

Figure A68: Overview about location of the surveyed households, experienced damage during previous flood events, and the hazard zones with high flood probability according to the Metropolitan Regulatory Plan.

Figure A69: Vulnerability map based on the three most relevant indicators according to the evaluation of the surveyed experts.

262

Appendix 2 – Coloured Figures

Figure A70: Vulnerability map based on the three most relevant indicators according to the evaluation of the surveyed households.

Figure A71: Normalized risk map for La Reina and Peñalolén. Calculation based on the hazard maps from the PRMS (10 year return period).

Appendix 2 – Coloured Figures

263

Figure A72: Normalized risk map for La Reina and Peñalolén. Calculation based on the hazard maps after Perez (2009) (baseline scenario, 10 year return period).

Figure A73: Normalized risk map for La Reina and Peñalolén. Calculation based on the projected hazard index after Perez (2009) (IPCC scenario B2, 100 year return period).

264

Appendix 2 – Coloured Figures

Figure A74: Development of new construction in the AMS in areas facing a high flood hazard (flooded at least once every two years) between 1993 and 2002.

Appendix 2 – Coloured Figures

265

Figure A75: Development of new construction in the AMS in areas facing a high flood hazard (flooded at least once every two years) between 2002 and 2009.

Susanne Meyer

Informal Modes of Governance in Customer Producer Relations The Electronic Industry in the Greater Pearl River Delta (China) Megacities and Global Change / Megastädte und globaler Wandel - Band 1

Susanne Meyer Informal Modes of Governance in Customer Producer Relations The Electronic Industry in the Greater Pearl River Delta (China) 2011. 222 pages with 15 illustrations and 45 tables. Soft cover. ¤ 42,ISBN 978-3-515-09849-6

The Greater Pearl River Delta (China) is one of the world’s mega-city regions. Its electronics industry is an important driver for economic growth and prosperity in China with firms that are highly flexible and quickly adapting to changing markets. The objective of this volume is to understand how informal modes of governance in customer and producer relations contribute to a flexible production network. It investigates whether informal interactions make up for constantly changing formal regulations in China or smooth business operations as an innovative mean to manage a firm’s network. The findings show that firms in the Greater Pearl River Delta organise their network by using a combination of formal and informal means in recruit-ing, contracting and enforcing processes. This research contributes to a better understanding of the complex business model in China and its implications for organisational novelty as such.

The author Susanne Meyer has been working for the Centre for Economic and Innovation Research of Joanneum Research (Graz / Austria) since 2010. After three years of working as a research assistant and lecturer at the Institute of Economic and Cultural Geography at the Leibniz University of Hannover (Germany) she graduated with a PhD in economic geography.

Franz Steiner Verlag Birkenwaldstr. 44 · D – 70191 Stuttgart Telefon: 0711 / 2582 – 0 · Fax: 0711 / 2582 – 390 E-Mail: [email protected] Internet: www.steiner-verlag.de

Carsten Butsch

Zugang zu Gesundheitsdienstleistungen Barrieren und Anreize in Pune, Indien Megacities and Global Change / Megastädte und globaler Wandel – Band 2

Carsten Butsch Zugang zu Gesundheitsdienstleistungen Barrieren und Anreize in Pune, Indien 2011. 324 Seiten mit 24 Abbildungen, 11 Tabellen, 3 Karten und 49 Diagrammen. Kartoniert. ¤ 49,ISBN 978-3-515-09942-4

Durch den raschen Urbanisierungsprozess in Schwellenund Entwicklungsländern entstehen Gesundheitsprobleme in bisher unbekanntem Ausmaß. Meist werden städtische Räume in diesen Ländern vereinfachend als ausreichend versorgt beschrieben. Dieser Sichtweise wird hier eine differenzierte Untersuchung des Zugangs von unterschiedlichen Bevölkerungsgruppen entgegengesetzt. Aufbauend auf einer Analyse etablierter Konzepte der Zugangsforschung wird ein erweitertes Konzept entwickelt: Zugang wird als Produkt von Zugangsbarrieren und -anreizen in sechs Dimensionen definiert. Das Konzept wird in sechs unterschiedlich strukturierten Untersuchungsgebieten in Pune empirisch überprüft. Die Ergebnisse der Studie lassen auf erhebliche Unterschiede im Zugang zu Gesundheitsdienstleistungen verschiedener Bevölkerungsgruppen schließen: Wesentliche Zugangsbarrieren entstehen durch die komplexe Anbieterstruktur im indischen Gesundheitssystem und die fehlende Regulierung des Gesundheitswesens. Die Studie leistet zudem einen konzeptionellen Beitrag im Bereich der interdisziplinären Zugangsforschung.

Der Autor Carsten Butsch studierte Geographie, Umweltökonomie und Städtebau an der Universität Bonn. Er ist als Wissenschaftlicher Mitarbeiter an der Universität zu Köln tätig und hat die Projektkoordination „Geofaktoren und zivile Krisenprävention in Megastädten“ inne. Seit 2011 ist er Sprecher des Arbeitskreises Geographien Südasiens.

Franz Steiner Verlag Birkenwaldstr. 44 · D – 70191 Stuttgart Telefon: 0711 / 2582 – 0 · Fax: 0711 / 2582 – 390 E-Mail: [email protected] Internet: www.steiner-verlag.de

Urban expansion and associated land-use changes increase both flood hazards and exposure. This book provides a conceptual and methodological framework for the analysis of urban flood risk in dynamic and complex settings, and proposes a comprehensive, system-oriented, integrated approach for its assessment. The risk assessment is carried out using case-specific indicators on the sub-city scale in two municipalities of Santiago de Chile. Relevant information is derived from various geodata sources, and together with explorative scenarios, is used to estimate future risk development. All data about hazard, elements at risk, and their vulnerability are compiled to a GIS-based risk map to join the risk-relevant components, to show their interrelations, and to provide a tool for monitoring and evaluating their changes over time. Finally, previous deficits in flood risk prevention and mitigation are outlined and recommendations on risk reduction are made.

ISBN 978-3-515-10092-2

www.steiner-verlag.de Franz Steiner Verlag