The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method Issue 2 | 27 September 2018 This report takes into account the particular instructions and requirements of our client. It is not intended for and should not be relied upon by any third party and no responsibility is undertaken to any third party. Job number 252746-00 Ove Arup & Partners International Ltd. 13 Fitzroy Street London W1T 4BQ United Kingdom www.arup.com Document Verification Job title Sierra Leone Multi-City Hazard Review and Risk Job number Assessment 252746-00 Document title Final Report (Volume 1 of 5): Technical File reference Methodology and Summary of Results Document ref 20180927-DOC-05A_V1of5_Method Revision Date Filename 20171110-REP-02A-252746_Vol1_DRAFT Draft 1 10 Nov Description First draft 2017 Prepared by Checked by Approved by Peter Redshaw James Bottomley Name Grace Campbell Anna Morley Matthew Free Anna Morley Matthew Free Signature Draft 2 23 Mar Filename 2018 Description Issue 1 Prepared by Checked by Approved by Peter Redshaw James Bottomley Anna Morley Name Grace Campbell Matthew Free Matthew Free Anna Morley Matthew Free Signature Draft 3 31 May Filename 2018 Description Draft final report updated Prepared by Checked by Approved by Peter Redshaw James Bottomley Anna Morley Name Grace Campbell Matthew Free Matthew Free Anna Morley Matthew Free Signature Issue Document Verification with Document  20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX Document Verification Page 2 of 3 Job title Sierra Leone Multi-City Hazard Review and Risk Job number Assessment 252746-00 Document title Final Report (Volume 1 of 5): Technical Methodology File reference and Summary of Results Document ref 20180927-DOC-05A_V1of5_Method Revision Date Filename Issue 14 Sep Description Final Issue incorporating World Bank comment 2018 Prepared by Checked by Approved by Grace Campbell Anna Morley Anna Morley Name Matthew Free Matthew Free Matthew Free Peter Redshaw James Bottomley Signature Issue Document Verification with Document  20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX Document Verification Page 3 of 3 Job title Sierra Leone Multi-City Hazard Review and Risk Job number Assessment 252746-00 Document title Final Report (Volume 1 of 5): Technical Methodology File reference and Summary of Results Document ref 20180927-DOC-05A_V1of5_Method Revision Date Filename Issue 2 27 Sep Description Final Issue (2) incorporating World Bank comment 2018 Prepared by Checked by Approved by Grace Campbell Anna Morley Anna Morley Name Matthew Free Matthew Free Matthew Free Peter Redshaw James Bottomley Signature Issue Document Verification with Document  20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Contents Page 1 Introduction 12 1.1 Project background 12 1.2 Outline of this report 13 1.3 Where are the hazard and risk maps for each city? Where to find more details and data? 13 1.4 Project scope 14 2 Sierra Leone: Country Profile and Context 18 2.1 Introduction 18 2.2 Recent history 18 2.3 Geographic setting and natural hazards in Sierra Leone 19 2.4 Urbanization in Sierra Leone 29 2.5 Socio-economics in Sierra Leone 29 2.6 Governance in Sierra Leone 30 3 Exposure Identification and Classification Methodology 34 3.1 Introduction 34 3.2 Buildings 34 3.3 Infrastructure 39 3.4 Population 39 4 Qualitative Hazard and Risk Assessment Methodology 42 4.1 Introduction 42 4.2 Qualitative hazard assessment 42 4.3 Qualitative exposure assessment 47 4.4 Qualitative risk assessment 48 5 Quantitative Hazard and Risk Assessment Methodology 50 5.1 Introduction 50 5.2 Quantitative flood hazard and risk assessment 50 5.3 Quantitative landslide hazard and risk 70 5.4 Quantitative coastal erosion hazard and risk 89 5.5 Quantitative sea-level rise hazard and risk 94 6 DRR/DRM Recommendation Methodology 97 6.1 Introduction 97 6.2 Methodology 97 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 7 Cost-Benefit Analysis Methodology 106 7.1 Introduction 106 7.2 Cost benefit analysis methodology – DRR/DRM direct costs and benefits 106 7.3 Cost benefit analysis methodology – including valuation of reduction in human fatalities 108 8 Summary of Hazard and Risk Results 110 8.1 Introduction 110 8.2 Flood hazard and risk 110 8.3 Landslide hazard and risk 111 8.4 Coastal erosion hazard and risk 113 8.5 Sea-level rise hazard and risk 113 9 References 116 Appendices Appendix A Building Replacement Costs in Sierra Leone 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Acknowledgements The Sierra Leone Multi-City Hazard Review and Risk Assessment would not have been possible without the dedication and support of different partners and stakeholders at national and local levels, who contributed both time and expertise. The assessment was prepared by the Project Team, which consisted of experts from Arup (as the lead organisation), the British Geological Survey (BGS), JBA flood risk consultancy, and the Integrated Geo-information and Environmental Management Services (INTEGEMS) consultancy. Arup wishes to extend great thanks to each member organisation and individual of the Project Team. The dedication, creativity, technical capacity and enthusiasm of the team members made the completion of this project to the best possible quality and practical use possible. The assessment was carried out in partnership with the World Bank, the Government of Sierra Leone, ministries, City and District Councils, and the community. The financial support for this assessment was provided by the Global Facility for Disaster Reduction and Recovery (GFDRR) and the European Union, in the framework of the Africa Caribbean Pacific–European Union Natural Disaster Risk Reduction (ACP–EU NDRR) Program, managed by GFDRR. Arup wishes to acknowledge the World Bank Task Team Leader for this project, Dr. Isabelle Celine Kane, for her commitment, vision, critical feedback, and leadership throughout this project. We would also like to thank a number of other key World Bank staff and consultants for their critical feedback, direction and support in completing this project, including: Deepali Tewari, Sokhna BA, Swati Sachdeva, Robert Reid, and Megha Mukim. We thank Mr. Parminder P. S. Brar, former Country Manager, World Bank Group, for his commitment to the project and in-country support, Gayle Martin, Country Manager, World Bank Group, and Sheik Sesay, Operations Officer, World Bank Group. Arup would like to extend its appreciation and acknowledge the numerous ministries and organizations for their assistance in granting access to information, providing support to the report and for their availability for discussions during the assessment. The Office of National Security (ONS) played a critical role in co- ordinating in-country meetings and workshops with members of the ministries and councils. We thank Mr. Ismail Sheriff Tarashid Tarawali, National Security Coordinator, Mr. John Vandy Rodgers and Mr Nabie Kamara, and many other senior officials from all participating ministries for their immense contribution to the process. In particular, we would like to acknowledge the mayors of each city Sunkari Kabba-Kamara (Makeni city), Mr. Harold Tucker (Bo city), and Franklyn Bode Gibson (Freetown), for their commitment to this project, and to the local stakeholders and community members we met during our time in each of the cities who shared important knowledge and feedback. During the course of this project, the Regent-Lumley flood and landslide disaster occurred on August 14th 2017 in Freetown. Subsequently, the Project Team spent two weeks in country collaborating with the World Bank and other government representatives and numerous experts from UN agencies and development 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 1 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results partners to complete a Rapid Damage and Loss Assessment (DaLA). Arup would like to acknowledge and thank the extreme effort and quality contributions from all those involved in this mission, as acknowledged in the official World Bank DaLA Report published following the Disaster1. To all the contributors, the team expresses its deepest gratitude and appreciation, especially to the local communities and affected populations, who experience annual flooding during the wet season, and who experienced the devastating effects of the August 14th Disaster. This report would not have been possible without their trust and engagement. 1 http://documents.worldbank.org/curated/en/523671510297364577/Sierra-Leone-Rapid-damage- and-loss-assessment-of-August-14th-2017-landslides-and-floods-in-the-western-area 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 2 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Abbreviations and Acronyms AAL Average annual loss ALARP As low as reasonably possible (or practicable) B/C Benefit cost ratio BGS British Geological Survey CIDMEWS Climate Information, Disaster Management, and Early Warning System DaLA Damage and Loss Assessment DR Discount rate EU European Union EPA Environmental Protection Agency FCC Freetown City Council GDP Gross Domestic Product GFDRR Global Facility for Disaster Reduction and Recovery GVWC Guma Valley Water Company INTEGEMS Integrated Geo-information and Environmental Management Services IRR Internal rate of return JBA JBA Consulting is part of the JBA Group, an environmental, engineering and risk group. km Kilometres MCC Makeni city council m a.s.l. Metres above sea level mm Millimetre No. Number MoWR Ministry of Water Resources NGO Nongovernmental organization NMA Nationals Minerals Agency NPAA National Protected Area Authority NPV Net present value ONS Office of National Security 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 3 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results OSM OpenStreetMap SLL Sierra Leone Leone (currency) SLPP Sierra Leone Peoples Party SuDS Sustainable Urban Drainage UN United Nations UNDP United Nations Development Programme UNICEF United Nations International Children’s Emergency Fund UNOPS United Nations Office for Project Services USD United States Dollar VSL Value of statistical life WASH Water Supply, Sanitation, and Hygiene WFP World Food Programme WHO World Health Organization 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 4 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 5 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Executive Summary This report describes the methods and data used to carry out the different analyses for the natural hazard and risk assessment undertaken for each of the three cities (Freetown, Makeni and Bo) assessed as part of this project. This report also provides an at-a-glance overview of the quantitative hazard and risk results from this study. The key stages of the project are outlined in Figure 1. The report provides guidance to the Government of Sierra Leone and other key stakeholders to prioritise a range of DRR/DRM options that will save lives, reduce the potential for damage to critical buildings and infrastructure, and reduce the potential economic losses caused by flooding and other hazards. Figure 1 – Overview of the key stages in this project. The following sections summarise the key points for each of the input datasets developed and analyses undertaken in this project for each of the three cities. Exposure: people, buildings and infrastructure Digital buildings geographic location data for this project is initially from OpenStreetMap (OSM, www.openstreetmap.org). Using the Humanitarian OpenStreetMap team (HOT OSM, https://www.hotosm.org/) Arup set up a ‘HOT task’ enabling us to extend the coverage of this initial exposure model by digitally mapping thousands more buildings, roads and bridges to this dataset for Freetown, Makeni and Bo, specifically for this project (for example Figure 2). This is now an open source, verified, and freely available dataset, which can be used by all the cities engaged in this project. In addition to the building locations, this project has also modelled building type, usage and replacement value, the latter being 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 6 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results informed by the World Bank Damage and Loss Assessment Report (World Bank, 2017). For further details of the methodology please refer to Section 3 of this report. For additional details of the city-specific exposure please refer to report Volumes 2, 3, and 4 of this project. Figure 2 – Example of the digital exposure model for an area of Freetown city, the pink polygons represent the real building locations and types, as shown in the photograph. Qualitative and Quantitative Natural Hazard and Risk Assessment Methodology Two types of hazard and risk assessment have been undertaken for this project – qualitative and quantitative. The qualitative hazard and risk assessments follow simple relationships between hazard, exposure, vulnerability and risk and are informed by expert scientific judgement and valuable local knowledge to identify areas of low, medium and high hazard and risk (flooding, landslides, coastal erosion and sea-level rise). The quantitative hazard and risk assessments provide, for example, the modelled depths of flood waters in meters, the number of people affected by different natural hazards, or numbers of buildings damaged by a particular hazard, in each of the three cities. The following sections summarise the methodology developed for the quantitative natural hazard and risk assessments. For further details of qualitative and quantitative hazard and risk results for each city, refer to report Volumes 2, 3, and 4 of this project. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 7 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Quantitative flood hazard and risk assessment To quantify the potential losses in Freetown, Makeni and Bo due to flooding, a flood catastrophe model was built. Firstly, the characteristics of rainfall and river flow were analysed and a spatial and temporal model of flooding across the region was created. Potential future flood events were simulated based on this model, and more extreme and widespread events than have been observed in recent history were created. The simulated events form the multi-peril event set. For each event in the event set, the spatial pattern of severity was provided, and this was converted into a depth and local flood extent using flood hazard maps. Vulnerability functions enable the damage to properties and human life to be estimated for each event in a stochastic event set, given the depth and location of flooding. The functions describe the expected mean damage ratio (expressed as a proportion of total value) of a risk for a given hazard intensity. The compiled risk results present both the average annual losses associated with flooding and the modelled losses at different return periods. For further details refer to Section 5.2 of this report. Quantitative landslide hazard and risk assessment The quantitative landslide hazard and risk assessment is sub-divided into four main parts:  Landslide susceptibility assessment;  Landslide hazard assessment;  Vulnerability assessment; and  Landslide risk assessment. Landslide susceptibility analysis for this project uses elevation, slope angle, slope aspect and the distance to the nearest river to describe the susceptibility of the terrain to hosting landslides. This is calibrated using a bivariate statistical analysis of the >400 historical landslides which were identified as part of the project literature review. The analysis is calibrated against a random sample of the landslides from the inventory and then blind-tested against the remainder to quantify goodness-of-fit. The susceptibility analysis therefore identifies slopes which could give rise to slope failures based on historical observations in a statistically robust way. Landslide hazard assessment is sub-divided into two main steps:  Landslide frequency-magnitude analysis; and  Landslide runout modelling. Landslide frequency-magnitude analysis describes how often landslides of different sizes occur within the study area. Landslide runout is modelled using a Gravitational Process Path (GPP) model (Wichmann, 2017), which is used to determine the probability that a given pixel within a digital elevation model will be affected by landslide debris which is initiated from a different pixel. This approach can provide a reasonable approximation of landslide runout probability 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 8 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results at city-scale and provides sensible results when used with a 30m spatial resolution digital elevation model. Landslide vulnerability analysis has been informed by a broad literature review and has been calibrated against losses from the 2017 Regent-Lumley Disaster. Landslide risk is expressed as the product of the probability of hazard occurrence (e.g. a damaging landslide event) and its adverse consequences (Lee and Jones, 2004). Landslide risk is calculated to buildings, infrastructure and population. For further details refer to Section 5.3 of this report. Quantitative coastal erosion hazard and risk assessment This project has determined 2050 scenario-based estimates of coastal erosion hazard and risk for the Freetown Peninsular. The estimated coastal cliff recession rate has been estimated from the interpretation and mapping of historical satellite imagery and historical maps. The recession rate takes both the specific present- day water levels surrounding the Freetown Peninsular (provided by local data) and future climate change into account by using the current IPCC global Atmosphere-Ocean General Circulation Model (AOGCM). To estimate the losses from the 2050 scenario-based estimates of coastal erosion hazard, the coastline recession rate has been projected inland and a range of losses are estimated based on the current location of buildings and population. For further details refer to Section 5.4 of this report. Quantitative sea-level rise hazard and risk assessment Sea-level rise projections have been estimated for Freetown from using a combination of both local water level data for the Peninsular and the Intergovernmental Panel on Climate Change (IPCC) most recent version of the global Atmosphere-Ocean General Circulation Models (AOGCM). The local sea- level rise (SLR) is calculated, relative to a baseline value determined from 1995, using the global mean sea-level rise for the future climate-change scenario and time horizon of interest. The resulting 2050 scenario estimated land-loss maps generated from this analysis show areas of the city (and wider Freetown Peninsular) which may be lost to sea- level rise for 2050. The location of exposed assets within the potential land-loss areas have been identified. The response of these assets within the land-loss areas (i.e. their vulnerability) will be total loss. For further details refer to Section 5.5 of this report. Disaster Risk Reduction and Disaster Risk Management Recommendation Methodology The method of developing city-specific DRR/DRM measures was to first consider a broad range of measures within the context of the priorities of the Sendai Framework for Disaster Risk Reduction (United Nations, 2015), and within the context of the specific cross-cutting issues faced by each of the cities, such as the lack of solid waste management and sand mining. The broad range of DRR/DRM measures were then considered on a catchment-by-catchment basis for each 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 9 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results different hazard in each different city. The aim of this approach was to capture how the nature of the hazard and risk changes depending on the location of, and location within, the catchments in each of the cities. A final suite of DRR/DRM options (each of which includes a combination of practically appropriate and cost- effective measures) were developed for each city, along with a proposed budget for each option. The final recommended options and budgets have been informed by a quantitative cost-benefit analysis (CBA). For further details refer to Section 6 of this report. Details of each of the broader DRR/DRM measures and the specific catchment-by-catchment analysis for each city can be found in the latter sections of report Volumes 2, 3 and 4 of this project. Cost-Benefit Analysis (CBA) Methodology The CBA methodology developed for this project includes the following six steps. A further description of this methodology has been provided in Section 7 of this report.  Step 1: Undertake quantitative risk assessment as discussed in Section 4;  Step 2: Risk calculation output without DRR measures will be cost of damage in USD for buildings and infrastructure and number of fatalities;  Step 3: Risk-reduction calculation with DRR measures in place will be cost of damage in USD for buildings and infrastructure and number of casualties;  Step 4: Benefit is equal to the difference in risk calculation output without DRR measures (Step 2) minus the difference in risk calculation output with DRR measure (Step 3). Benefit is determined by estimating the percentage risk reduction with DRR measures;  Step 5: Cost estimation for DRR measures including capital expenditure and operational expenditure over 33 years. Allow for discount rate of 6% over 33 years;  Step 6: Calculate Benefit Cost Ratios (BCR), Net Present Value (NPV) and Internal Rate of Return (IRR). Summary of Results for Each City Finally, this report also includes a high-level, at-a-glance summary of the hazard and risk results for each city. However, the full context, explanations, details and maps have been presented in each corresponding city report (Volumes 2, 3, and 4) and therefore these other report Volumes should be referred to for detail alongside this Volume. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 10 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 11 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 1 Introduction 1.1 Project background The geographical location of Sierra Leone makes it prone to intense and recurring natural hazards such as flooding and landslides, which cause loss of life, devastate, and result in damage and significant economic impact. Increasing urbanisation together with the effects of climate change are compounding the natural hazard related problems faced by cities across the country. The lack of available and reliable data on the frequency and impact of natural hazards on these cities is hindering DRR and DRM, urban planning and investment. To better understand and quantify natural hazard and disaster risk in Sierra Leone, the World Bank and Global Facility for Disaster Reduction and Recovery (GFDRR) are supporting, under Africa Caribbean Pacific – European Union (ACP-EU) funding, the development of new natural hazard and risk information in Sierra Leone for targeted cities namely, Freetown, Makeni and Bo (Table 1). The World Bank have commissioned Ove Arup and Partners International Ltd (Arup), Integrated Geo-information and Environmental Management Services (INTEGEMS), JBA Risk Management (JBA) and the British Geological Survey (BGS) (collectively, the ‘Project Team’) to undertake this consultancy assignment. Table 1 – Natural hazard and risk assessment scope of work in Freetown, Makeni and Bo Freetown Makeni Bo Flooding    Landslides    Sea level rise    Coastal erosion    The results of this study will help to inform the understanding of natural hazards and risk for the three cities and build on ongoing DRR and DRM work in Sierra Leone by recommending simple but practical and effective solutions to natural hazard-risk to reduce risk and increase resilience in each of the three cities. Throughout the assignment the Project Team has been working closely with the Office of National Security (ONS) and local stakeholders through ongoing engagement, workshops and sharing of information and findings. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 12 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 1.2 Outline of this report This report is Volume 1 of 5 of the Final Report. Volume 1 contains:  Section 1 – An introduction to the project, including aims, objectives, definitions and limitations;  Section 2 – An overview of Sierra Leone, with specific reference to recent history and the context of natural disasters within Sierra Leone;  Section 3 – The data sources and methodology used to provide exposure data (buildings and infrastructure) for the project;  Section 4 – The methodology used to conduct a qualitative natural hazard and risk assessment for each city and peril;  Section 5 – The methodology used to undertake a quantitative natural hazard and risk assessment for each city and peril;  Section 6 – The methodology used to inform the DRR/DRM options for each hazard in each city;  Section 7 – The methodology used to undertake a cost-benefit analysis to support the recommendation of DRR/DRM measures; and  Section 8 – An at-a-glance summary of the results from the qualitative and quantitative risk assessments. Volumes 2 – 4 contain city-specific information relating to each of Freetown, Makeni and Bo respectively. These volumes contain further literature reviews specifically relating to each city, a more detailed review of the results from the natural hazard and risk assessments for each city and a review of city-specific Disaster Risk Management/Reduction options. Volume 5 contains a series of A3 page size maps showing the spatial distribution of natural hazard and risk in each of Freetown, Makeni and Bo. This report should be read alongside the relevant other report Volumes for this project. 1.3 Where are the hazard and risk maps for each city? Where to find more details and data? Throughout this report, example maps showing the spatial distribution of natural hazard and risk in the cities are included as examples or when specifically referred to. These maps have been reduced in size to fit an A4 page. For the full series of A3 maps, see report Volume 5. Whilst the maps provide an overview of natural hazard and risk in each city, most value can be gained from this study by viewing and interrogating the data produced by this study using a Geographic Information System (GIS) such as ArcGIS or QGIS (QGIS is open-source and freely available2). Throughout report Volumes 2, 3 and 4 of this project, example screenshots from GIS are 2 QGIS available for free download at: https://qgis.org/en/site/ 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 13 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results included to demonstrate this. All data produced by this project is open source and are available from the World Bank. Please contact World Bank Task Team Leader for this project Dr. Isabelle Celine Kane (ikane@worldbank.org), or Arup Project Manager, Anna Morley (anna.morley@arup.com) for the website link to the online data. 1.4 Project scope 1.4.1 Objectives The objective of the assignment is to support the Government of Sierra Leone through the World Bank and GFDRR to develop city level natural hazard and risk assessments for three cities in Sierra Leone, namely Freetown, Makeni and Bo. The risk assessment will commence with a qualitative review of the risks to these cities as well as a review of the existing DRM and urban planning policies currently in place. This review will feed into quantitative risk assessment for flooding, landslides, sea-level rise and coastal erosion. The results of these hazard and risk assessments will be used to identify priority DRM investments through a high-level cost-benefit analysis. The results will also be used to feed into high-level DRM and urban planning policy recommendations within the wider context of urban master-planning for Freetown, Makeni and Bo. 1.4.2 Terms of reference The scope of the project includes:  A review and assimilation of relevant and up to date hazard, exposure, vulnerability and risk information, current DRR initiatives at city level and current urban planning policy/guidance in Freetown, Bo and Makeni;  A qualitative review of the natural hazards and risks in the form of maps and tables;  A preliminary action plan for DRM;  Quantitative reviews of natural hazards and risks in the form of maps and tables (landslides, coastal erosion and sea-level rise in Freetown only);  Final DRM suggestions;  High-level cost benefit analysis for five DRM options;  To lead the delivery of risk information to the ONS and the city councils in each of the study cities;  To frame the results of the qualitative and quantitative risk assessments in the context of urban planning;  Progress meetings with the World Bank and ONS; and  To undertake three field visits to Sierra Leone to gather data for the project, liaise with stakeholders and lead the delivery of risk information to the key 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 14 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results stakeholder as specified above. These field visits are referred to as Mission #1, Mission #2 and Mission #43. 1.4.3 Variations to the terms of reference Over the course of the project there have been three variations to the terms of reference and scope of work. 1.4.3.1 Variation 1 During Mission #1 it was identified that one of the key issues from the city stakeholder’s perspective is how urban development policy and practice is contributing to risk. The main point of adjustment following Mission #1 is a shifting focus from natural hazard and risk assessment leading on to prioritisation of disaster risk reduction options, to natural hazard and risk assessment within the context of wider urban planning policy. There is a clear need to understand how better knowledge of risk can directly inform city disaster management options through improved:  Urban policy and planning; and  Urban master-planning and development. Based on the results of the natural hazard and risk assessments, the Project Team has provided urban planning advice consisting of:  A rapid high-level city resilience profile based on limited desktop qualitative and quantitative data;  Summary table of city plans and policies for each city; and  High level DRM/DRR advice set within the context of the Sendai Framework. 1.4.3.2 Variation 2 On Monday 14th August 2017 a devastating landslide occurred in Regent, Freetown, Sierra Leone. The landslide, which occurred in multiple phases, was located in an area which was already understood to have been affected by severe flooding. As a direct result of the landslide and flooding, approximately 6,000 people were affected, of which 1,141 have been declared dead or missing. The disaster is referred to in this report as the Regent-Lumley Disaster. Following the event, the Government of Sierra Leone requested the support of the World Bank to conduct a comprehensive Damage and Loss Assessment (DaLA, World Bank, 2017) in partnership with the United Nations (UN). The World Bank requested the assistance of the Project Team owing to the knowledge already developed about exposure, hazard and risk in Freetown as part of this wider study. The DaLA was conducted between 24/08/2017 and 08/09/2017, with the objective 3 Mission #3 was conducted in response to the Regent-Lumley Disaster of 14/08/2017, where the Project Team mobilised to Sierra Leone to support the World Bank in conducting a Damage and Loss Assessment. Further details of this are provided in Volume 2. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 15 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results of estimating damages and losses and making preliminary estimations for mobilizing funds and launching immediate recovery. The findings from the DaLA are reported separately by the World Bank (World Bank, 2017). 1.4.3.3 Variation 3 Following the Regent-Lumley Disaster, new information was made available relating to the nature of the flood and landslide hazard, exposure, vulnerability and risk in Sierra Leone (and Freetown in particular). Much of this information is summarized in the DaLA. It was decided that the information from the DaLA should be fed back into this project. This included:  Revised building replacement costs to accord with those used during the DaLA;  Estimation of the return period of events similar to the Regent-Lumley Disaster to properly frame the disaster within the context of the wider project; and  A qualitative review of the nature of the Regent-Lumley Disaster and comment on how it is represented in the wider project. Of the ~1000 fatalities, how many could be attributed to the landslide in Regent in comparison to the flooding some ~6km away in Lumley? This is a complex combination of flooding and landsliding. In order to better understand this interaction, the Project Team conducted research into the nature of the event, and furthermore to how well it is represented by either the flood or landslide model (or indeed if the two must be somehow used together to fully model the Regent-Lumley Disaster). 1.4.4 Limitations Limitations to this project include (but are not limited to):  This study addresses specific natural hazards. It is recognised that Sierra Leone suffers many other hazards including epidemics, wild fire, land degradation, household fires etc. Information has been compiled in this project which could be used to study these hazards in the future, however the focus of this project is on addressing the hazard and risk associated with flooding, landslides and coastal erosion and sea-level rise.  Risk to agriculture is not included in this multi-city project given the primarily urban context.  Data input at city scale does not allow for output at detailed resolution and would not be appropriate for design of individual engineering structures. Specific limitations of data input and modelling practices are addressed in the individual sections of this report.  The Terms of Reference have specifically asked for a cost benefit analysis on DRM/DRR options. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 16 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 17 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 2 Sierra Leone: Country Profile and Context 2.1 Introduction The following sections provide the overview and context of the project at country scale. Report Volumes 2, 3, and 4 provide the city-specific context for each hazard and risk assessment. Sierra Leone is situated on the western coast of Africa and shares borders with Guinea, Liberia and the Atlantic Ocean. In 2015, Sierra Leone had a per capita GDP of US$684, making it one of the poorest countries within Sub-Saharan Africa and globally. Sierra Leone is ranked 179 out of 188 countries on the United Nations 2016 Human Development Index and chronic malnutrition is still on the rise with 44% of children below the age of 5 being stunted in 2010, up from 40 percent in 2005. Per capita GDP stagnated in the period after independence in 1961, contracted by 3.4 percent on average between the years of the civil war (1991-2001) and increased by an average of 5.9 percent from 2002 to 2014. In 2014 Sierra Leone was severely affected by the twin shocks of the Ebola outbreak and the downturn of international prices of iron ore. This caused the economy to contract by more than 20 percent, sending the country and its 7.1 million inhabitants into economic and social turmoil, from which it is yet to recover (World Bank, 2017). 2.2 Recent history Sierra Leone has had an unstable modern history, enduring major natural and man-made shocks which have rocked the nation and compounded underlying vulnerabilities (Table 2). These events include significant periods of political instability and conflict and more recently a health crisis that debilitated the whole country as well as neighbouring West African countries. The civil war resulted in lack of technical capacity in the country related to lack of school and university attendance during these years. Table 2 – An overview of recent history in Sierra Leone (BBC, 2017). Event Description 1961 – 1991: Sierra Leone becomes independent from British rule in 1961. Democratic Independence governance exists up to 1967. After this commences a 40 year period that is and political characterised military coups, repressive rule and varying levels of instability corruption. 1991: Civil war The Revolutionary United Front (RUF), led by Foday Sankoh and supported begins by Liberian fighters loyal to then Liberian President Charles Taylor invades Eastern Sierra Leone. Fierce fighting takes place across the country during an 11-year conflict including several key battles in Freetown. RUF controls more than 60% of the country at times during this period. They take control of lucrative diamond mines, carry out brutal attacks and recruit child soldiers. At this time the country goes through several presidents with a further series of military coups. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 18 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Event Description 2000 – 2002: UK military intervention assists other international forces to end the Civil war ends conflict. Sankoh is captured in 2000 and UK forces leave the country in 2002. An international peacekeeping force of 17,500 troops oversee the transition to peace. Ahmad Tejan Kabbah (first elected in 1996 but ousted by coups) wins a landslide 2002 election. 2007 – 2012: Ernest Bai Koroma wins a majority in parliament in 2007. He secures a Presidential second and final term in 2012 in the first election held without UN oversight elections since conflict ended. In 2012 former Liberian President Charles Taylor is sentenced at The Hague to 50 years in prison for war crimes during the 1991-2002 conflict. 2014 – 2016: The Sierra Leone government issue a state of emergency in July 2014 after Ebola crisis the outbreak of Ebola. The epidemic persists over a two-year period, claiming almost 4,000 lives. 2017: Regent- On Monday 14th August 2017 a landslide occurred in Regent, Freetown, Lumley Disaster Sierra Leone. The landslide, which occurred in multiple phases, was located in an area which was already understood to have been affected by severe flooding. As a direct result of the landslide and flooding, approximately 6,000 people were affected, of which 1,141 have been declared dead or missing. 2.3 Geographic setting and natural hazards in Sierra Leone Sierra Leone has a humid, tropical climate and a highly seasonal rainfall pattern that peaks between July and September (McSweeney et al., 2010). Temperatures are relatively uniform throughout the year, ranging from 24 to 28 degrees Celsius. Lowest temperatures are from July to September, in the middle of the rainy season, and highest temperatures are in February and March, near the end of the dry season as shown in Figure 3 (Lapworth et al., 2015). The national average annual rainfall is 2746mm, but varies across the country: 3659mm in Bonthe in the south, 2979mm in Lungi in the west, and 2618mm in Kabala and Bo (UNDP, 2012). Recently, periods of drought have occurred due to the delayed onset of the monsoon rains, and when the heavy rain has arrived there has been extensive flooding (UNDP, 2012). This rainfall season is largely controlled by the movement of the tropical rain belt, also known as the Inter-Tropical Convergence Zone (ITCZ), which oscillates between the northern and southern tropics over the course of a year and affects Sierra Leone when in its northern position. The ITCZ produces the West African Monsoon, resulting in exceptionally high coastal rainfall in the wet season (exceeding 1000mm on the coast, and decreasing inland to 300mm). Seasonal rainfall varies on inter-annual and inter-decadal timescales, in part due to the El Niño Southern Oscillation (ENSO) (UNDP, 2012). The main rivers in Sierra Leone are the Moa, the Sewa, the Taia, the Little Scarcies and the Rokel which all flow from northeast to southwest across the country and drain most the surface of the land. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 19 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 3 – Average monthly temperature and precipitation in Sierra Leone. 2.3.1 Geology Sierra Leone forms the central part of the West African Craton whose counterpart forms the Guyana Shield of northern South America (Morel, 1979). The most recent assessment of geology and mineral resources was by the Sierra Leone Geological Survey (MacFarlane et al, 1979), which focused on north of the country. The Freetown Layered Complex, is a 65 km long, 14 km wide and 7 km thick tholeiitic intrusion, which intruded the West African Craton during the Early Jurassic ~190 Ma (Chalokwu et al, 1995; Chalokwu, 2001). The intrusion has an arcuate outcrop towards the west and extends out under the Atlantic Ocean. It is composed of a layered complex of gabbro, norite, troctolite and anorthosite. Platinum occurs in the gravels of many of the streams (Morel, 1979) that cut the outcrops of anorthosite and anorthositic gabbro. The relationship of this complex with the other units is obscured by the coastal veneer of Tertiary sediments of the Bullom Group (formally Bullom Series) which lies unconformably on the basement. Tertiary and more recent weathering has led to lateritisation across a large part of Sierra Leone, affecting mainly the greenstone belts and the extensive dolerite intrusions. The bauxite deposits formed within the Kasila Group, which crops out extensively to the east of Freetown, are a result of this weathering process. Diamonds have been a major export for the county since first being discovered in the 1930’s (Morel, 1979). 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 20 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 2.3.2 Natural hazards in Sierra Leone 2.3.2.1 Flooding The UNDP published a paper in 2012 on the topic of climate change and disaster management in Sierra Leone (UNDP, 2012). The key findings that relate to flooding were:  Of the total number of people affected by disasters in Sierra Leone in the last 30 years, 90% of were affected by flooding;  From 1980 to 2010, over 220,000 people were affected by floods, and 145 killed (EMDAT, 2009);  Floods occur frequently and can occur any time but are particularly common in the rainy season between May and October;  Floods are usually the result of heavy rainfall combined with high tides; ocean surges, blocked, diverted or narrowed drains; and increasing river volumes;  The population of Sierra Leone is particularly vulnerable due to land reclamation from the sea and swamps, and buildings constructed on floodplains and in the bottoms of valleys;  In the last few years, the areas worst affected by flooding included: Kroo Bay, Susan’s Bay, Granville Brook, Lumley in Western Area, Regent, Port Loko and Kambia Districts, the Newton catchment area, Pujehun and Bo areas, Kenema and Moyamba Districts, and coastal beaches of the Western Area Peninsular;  In general, floods recede within an hour as flooding is primarily driven by intense, localised rainfall over a short time. However, there have been occasions when it has taken up to a week or month to recede due to longer- lasting, lower intensity rainfall over a wide area. For example, in the Port Loko and Kambia Districts, floodwater in 2003 and 2004 lasted for about a month;  Heavy rainfall in neighbouring countries may cause floods in Sierra Leone due to the overflowing of three rivers: Great Scarcies and Little Scarcies rivers from Guinea and Mano from Liberia;  The most common consequences of flooding are loss of life, disease outbreak, and damage to crops, livestock, infrastructure and housing;  Flood water depth can exceed 1m, and if standing for a prolonged period, can cause damage to newly planted rice crops in both the valley bottoms and the nurseries up-slope; and  In the future, rising sea levels will contribute to increased flood frequency probabilities and inundation of coastal lands and wetlands. Flooding in Sierra Leone is a regular occurrence (EMDAT, 2016, e.g. Figure 4). Kroo Bay in Freetown, one of the largest coastal slums, has flooded every year since 2008 due to heavy rains. This is exacerbated by the expansion onto the floodplains (Africa Research Institute, 2015). The historic record shows that 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 21 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results events are often of a very short duration, (hours to days rather than weeks) and are frequently coupled with landslides. Figure 4 – Examples of flooding in Freetown (upper two photographs) and in Bo city (lower photograph). 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 22 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 3 – Historical flood events which have affected the study area. The 14/08/2017 Regent-Lumley Disaster is discussed in the accompanying Freetown-specific report (Volume 2) and is omitted from this table. Date Flood type Weather Locations People affected Financial Comments References conditions affected cost 29/07/1996 River - Freetown 12 deaths and many more - Intense rainfall caused the EMDAT injured. 200,000 evacuated. flooding of the Rokel River (2009) Many reported missing. Estuary which caused a densely-populated area of Freetown to get flooded where the Rokel River flows through the city. 28/08/2002 – River (Tibel ‘Four days of Senehun and No deaths. 500 people - Water flooded a bridge on the Dartmouth 01/09/2002 River) heavy rainfall’ Jamma, near Bo. homeless. At least 125 highway linking Bo to (2017) Njagabehun homes destroyed. Four Freetown. 180 hectares of rice Town in newly constructed guest fields were swept away as Moyamba houses collapsed. Makeshift well as 28 hectares of District, Freetown huts built by refugees and cassava, groundnuts and market stalls destroyed. vegetable fields were waterlogged and uprooted. Total area of 9,150sqkm affected. 07/2005 - - Kissy, Freetown Seven deaths. - Severe floods swept through a AllAfrica quarry (2017) 09/2006 River Heavy rain Lumley, Many houses destroyed and - People were warned about Freetown many badly damaged. A building residences in the area man nearly lost his son as because it was a waterway water swept through the that was prone to flooding. house. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 23 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Date Flood type Weather Locations People affected Financial Comments References conditions affected cost 08/2007 River Heavy rain Kroo Bay, 4,500 affected. - - EMDAT Freetown (2009) 11/09/2008 Coastal/river Heavy days of Kroo Bay, 1,000 people homeless and - 80% of Kroo Bay could not AllAfrica ‘torrential rain’ Freetown 10,000 affected. sleep in their homes the night (2017) of the 11th. The magnitude of the flood ranges from the worst in 20 years to some people thinking it was the worst flood Kroo Bay has ever experienced. 11th was the heaviest day’s rainfall of the year. 13/08/2009 – River Heavy, Kroo Bay, 103 deaths, 1,470 affected. - - EMDAT 20/08/2009 prolonged rain Freetown (2009) 08/2010 River Heavy rain Kailahun, Kono 234 affected - - EMDAT Districts (Eastern (2009) province) Kambia, Tonkolili Districts (Northern province), Western Area Urban, Southern province 08/2013 - Heavy rain Kroo Bay, 50 informal properties ‘Millions of - AllAfrica Freetown destroyed. Leones (2017) worth of property’ 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 24 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Date Flood type Weather Locations People affected Financial Comments References conditions affected cost 16/08/2014 Surface water Heavy rain Makeni 100s of people displaced, 50 ‘Millions of - Awoko houses destroyed. Leones (2017) worth of property’ 04/09/2015 – River and Heavy rain. Bo: 3,293 people 12 fatalities. 24,303 people USD In the Bo District flooding ReliefWeb 17/09/2015 surface water ‘147mm rain affected. affected. 4,051 families $20million+ was caused by the Sewa River (2015) fell in 6 hours’. Freetown: 14,050 affected. Freetown: 1,400 to repair and bursting its banks. people affected. homes destroyed. Bo: 339 rehabilitate Bonthe: 4,650 houses destroyed. Pujehun: some of the people affected. 16 houses destroyed. roads and Port Loko: 1,510 buildings people affected. destroyed. Pujehun: 800 people affected. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 25 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 2.3.2.2 Landslides In Sierra Leone, most of the landslide hazard is concentrated around the steep hills of the Freetown Peninsula and to the far east of the country where the topography rises into Guinea. Landslide hazards in the Freetown area are frequent (Wells, 1962) and landslide hazard is widely recognised by government and the resident population in Freetown. Figure 4 a photograph of the massive Regent landslide that occurred in Freetown in August 2017 with catastrophic consequences. This disaster is discussed in detail in report Volume 2 of this project. It is essential to understand the causes of past landslides and their physical processes in order to select appropriate slope assessments methodologies for future landslide hazard probability and risk. Preliminary understanding can be derived from a desk-based study of historic and recent landslides including, palaeo-climate, previous land use change and vegetation cover changes. Rainfall is the main trigger factor for landslides in Sierra Leone (and indeed was the trigger for the 2017 Regent Landslide), with human activities the second most common cause (Thomas, 1998). In reality, these conditioning and triggering factors often combine, resulting in landslides. Figure 5 – The Regent landslide, Freetown August 2017. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 26 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results The rapid population increase and expansion of Freetown, during and after the civil war applied pressure on the landscape around the Freetown settlements, resulting in unregulated deforestation. The high rate of forest degradation on the mountains of Freetown has been linked to recent environmental disasters and several natural hazards are prevalent including, increased gully erosion, landslides, mudflow, rock fall, and flooding (Mansaray et al., 2016). The seismic hazard in Sierra Leone is very low and earthquake triggering of landslides is extremely unlikely within a 50-year return period. 2.3.2.3 Coastal erosion and sea-level rise The coastline of Sierra Leone measures some 460 km of which 150 km is significantly developed (including Freetown) and 190 km is relatively sheltered by extensive mangrove systems and mudflats. Sea level rise and coastal erosion is a major concern for Sierra Leone both due to natural and man-made causes. Natural causes include storms (wave and swell actions) with mechanical and chemical weathering to cliffs. Man-made pressures including alluvial mining, sand mining, haphazard use of individual coastal defences, poor coastal management, and deforestation have also contributed to the susceptibility to coastal hazards. Coastal erosion has resulted in loss of private and public property along the northern sector of the coastline as well as posing threat to beaches, settlements and other shoreline facilities such as hotels, clubs resorts along the coastline of the Freetown peninsula. Erosion of the sandy beaches is accelerated by sand extraction activities for construction and building purposes. Efforts made in the past to halt shoreline retreat along certain portions of the Freetown Peninsula have proved unsuccessful (information received by project team during in country workshops, although no specific efforts were cited). 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 27 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 28 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 2.4 Urbanization in Sierra Leone Sierra Leone has an urban population of approximately 2.6 million people (as at 2015; 40% of the total population of 7.1 million people, Statistics Sierra Leone, 2016). The population of Sierra Leone has been increasing gradually since the 1963 census, however, from 2004 to 2015 the population has increased rapidly from ~5 million to ~7 million, representing an inter-census increase of 43% (Statistics Sierra Leone, 2016). Following the civil war, many people relocated to the cities in Sierra Leone, sometimes settling in informal settlements which can be highly vulnerable to natural hazards. In Sierra Leone, this accelerated urbanisation has not been accompanied by sufficient resources to plan and manage this fast growth; cities have lacked the financing to make the necessary investments to cope with the accelerated demand for infrastructure and services. Additional factors associated with increasing urbanization in Sierra Leone which have led to increased natural hazard-risk include:  Deforestation: Deforestation increases both landslide and flood hazard, by removing the reinforcing effect of tree-roots to stabilise the ground, reducing evapotranspiration and increasing surface runoff. Deforestation is also indicative of development into areas which might historically have been preserved, free from development for example protected parklands such as in Regent, Freetown.  Lack of zoning and city-planning: A lack of natural hazard and risk-based zoning has meant that in some cases large settlements have been able to develop in high hazard-risk areas, for example in Kroo Bay, Freetown.  Lack of building regulations: A lack of building regulations and poor enforcement mean that buildings and structures in many parts of Sierra Leone are vulnerable to the effects of natural hazards.  Inadequate provision of services. As a result of urbanization and some of the factors associated with it as listed above, cities in Sierra Leone are increasingly vulnerable to natural hazards. 2.5 Socio-economics in Sierra Leone Although the incidence of poverty decreased between 2003 and 2011 by almost 13%, the number of poor remained nearly constant at around 3.3 million due to population growth. In 2011 the estimated incidence of poverty was 53.8% (SLIHS, Statistics Sierra Leone, 2014), with three quarters of the poor residing in rural areas, despite the gains made in poverty reduction attributed to agriculture. Urban areas outside Freetown, experienced the most significant decline in poverty from 79.9% in 2003 to 39.9% in 2011. Freetown was the only area to experience an increase in poverty between 2003 and 2011, from 14% to 21%, however poverty rates remain well below the rest of the country. The increase in poverty in Freetown is believed to have been driven a lack of employment opportunities, high corruption, the Ebola crisis, massive displacement due to the civil war, high rates of illiteracy, and inadequate infrastructure (Ravichandran, 2011). 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 29 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results The country’s young, urbanizing and growing population needs employment opportunities. According to the 2015 census (Statistics Sierra Leone, 2016), the population of Sierra Leone is approximately 7.1 million, with 45.8% of the population under the age of 15 and 74.8% below the age of 35. The share of the population living in urban areas almost doubled from 21% in 1967 to almost 40% in 2015, with a high concentration in the capital Freetown, which has grown to a population of more than 1 million. Although jobs in manufacturing are concentrated in Freetown, the majority are either informal (72%) or unpaid (8%). The service sector accounted for about 33% of the labour force in 2014 (mostly in the capital Freetown), though its contribution to GDP declined from 30% in 2001 to 20% in 2015. More than half of the individuals aged between 15 and 35 participate in the labour force, and 91% percent of these are self-employed. Concentration of 15% of Sierra Leone’s population in Freetown could generate significant benefits from economies of scale, but the combination of population growth, fiscal space, authority and capacity has posed challenges in allocating land and providing services. In Freetown, there are already large deficits in municipal infrastructure and services, and whilst the responsibility is vested with local councils, in practice central government ministries continue to deliver them, thus breaking direct accountability route of service provision. Whilst city councils are autonomous legal entities governed by elected councils with their own expenditure budgets and revenue resources, they have limited budgetary opportunities that could advance densification and therefore more cost-effective service delivery. Within that limited fiscal envelope and with no incentives, local governments are struggling to deliver services to standards and levels commensurate with their budgets. 2.6 Governance in Sierra Leone In Sierra Leone there are three spheres of government: Central, Local and Chiefdoms. Local government is provided for by the Local Government Act 2004, and not by the constitution. The Ministry of Local Government and Rural Development is responsible for Local government, which comprises six City councils (Freetown, Bo, Kenema, Makeni, Koidu and Bonthe) and 13 District councils. In the third sphere of government there are some 150 Chiefdom councils. Local council elections are held every four years (Commonwealth Governance, 2017). The Government of Sierra Leone has enacted the National Security and Central Intelligence Act 2002 (NSCIA). Section 18, subsection IV of the NSCIA mandates the Office of National Security (ONS) to co-ordinate the management of national emergencies, including both natural and man-made disasters. The ONS Disaster Management Department (ONS-DMD) was established under the ONS for Sierra Leone. 2.6.1 Disaster management and emergency response The Office of National Security are mandated to co-ordinate the management of national disaster mitigation and the preparation of a National Disaster Management Plan, with a specific Disaster Management Department under ONS. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 30 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results The Disaster Management Plan has been undertaken with the Ministry of Lands, Country Planning and the Environment (MLCPE), Freetown City Council (FCC), and Ministry of Works, as some of the main stakeholders. ONS objectives include to:  Ensure the integration of disaster-risk management into sustainable development programmes and policies, warranting a holistic approach to disaster management;  Improve the identification, assessment, monitoring, and early warning of risks; and  Improve effectiveness of response through stronger disaster preparedness. (Frazer-Williams, 2014). The United Kingdom’s influence in Sierra Leone appears to have had some effect on the country’s command and control emergency response structure. It shares the UK’s distinction of 'Gold' (strategic), 'Silver' (tactical) and 'Bronze' (operational) response to emergency situations. However, whilst in the UK Gold, Silver and Bronze response agencies are usually local to the emergency situation, in Sierra Leone emergencies of any significant scale have a nationally-led response. A National Strategic Situation Group (NSSG) is activated to manage the Gold response during a ‘Level 2 or 3’ emergency. The NSSG has core staff from ONS but also includes representatives from all involved ministries/departments/agencies (MDAs) including the armed forces (RSLAF), police (SLP) and Ministries of Finance (MOFED), Homeland Security (MHS) and Social Welfare, Gender and Children (MSWGCA) and local government (MLGRD) (NSSG, 2016). Depending on the scale of the event the Silver (tactical) response will be led by the District Disaster Management Committee (DDMC). The DDMC is a multi- organizational grouping, which exists within each district, bringing together District Council leaders, security representatives, key local representatives, the health organizations and any active international partners within the area. It meets on a regular basis to support DM preparations and risk assessments, and will stand up on a 24 hour basis during an emergency (NSSG, 2016). At the operational level, response is led by the District Emergency Operations Centre (DEOC). This links to the DDMC. A report by World Vision on the Ebola response observed that DEOC management during that crisis was enhanced by the support of various NGOs. The report questioned the sustainability of such centres should that support scale back (World Vision, 2017). In fact, much of the national disaster management structure described appears to have only recently been finalised in aftermath of the 2015 floods across Bo, Pujeuhn and Bonthe which displaced an estimated 15,000 people. The flood response utilised the presence of international partners still in country post-Ebola and there was a realisation that as international support scales back, there is a need to reinforce national-district coordination. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 31 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 2.6.2 Zoning, building regulations and urban planning The Ministry of Lands, Country Planning and the Environment, MLCPE, has traditionally been responsible for planning activity in Sierra Leone and the Ministry of Works, Housing and Infrastructure (MWHI) for building inspection, permitting and development control. The Local Government Act (LGA, 2004), was enacted in 2004 which provided city powers for strategic local planning, preparation of land-use plans, and issuance of building permits amongst other responsibilities. Freetown, Makeni and Bo are governed by elected city councils. However, it has been suggested that development control is weak at present (Frazer-Williams, 2014). A Freetown development plan for 2016 – 2018 (FCC, 2016) was developed by the Freetown City Council (FCC) although is lacking in implementation due to limited resources. A Freetown pilot plan was developed by MLCPE and FCC with European Union support in 2014, with on the job training of FCC staff of the project’s objectives (MLCPE/FCC, 2014) but this has yet to be implemented. 2.6.3 NGOs and community based organizations There is a great deal of national and international NGO activity in Sierra Leone. This significantly increased in scale during the 2014 Ebola crisis, particularly with regard to response and recovery activities but there is also a strong permanent presence within the county that contributes to disaster management activities. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 32 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 33 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 3 Exposure Identification and Classification Methodology 3.1 Introduction This section of the report provides a description of the methodology used for compiling data on the location and nature of buildings, infrastructure and population in Freetown, Makeni and Bo. 3.2 Buildings 3.2.1 Building location data Buildings location data for this project have been sourced from OpenStreetMap (OSM, www.openstreetmap.org). Although the location of many buildings were already well represented in the OSM database, Arup liaised with the Humanitarian OpenStreetMap team (HOT OSM, https://www.hotosm.org/) to set up a ‘HOT task’ to allow the verification and supplementation of the existing OSM database for this project. Building location polygons are manually drawn around each building by viewing the built in imagery in OSM, which is typically sourced from Bing, DigitalGlobe or Mapbox. This Sierra Leone Multi-City Hazard Review and Risk Assessment Project has added in the order of 25,000 – 30,000 buildings to the OSM dataset since project commencement in January 2017. One of the main advantages of undertaking the exposure mapping in this way, is that this data is now fully open-source and available to use by anyone (and is downloaded through the OSM website). This dataset includes the following limitations:  Many new buildings are being built in Sierra Leone all the time so this dataset is reasonably representative at the time the imagery used to digitize the buildings was taken. The majority of the imagery used is dated from 2017, although this cannot be guaranteed at all times. In areas where multiple dates of imagery covered a particular area, the most recent available imagery was used to digitize buildings. Buildings without roofs were not included in the dataset.  Digitisation of data in the OSM environment can be done by anyone with internet connection globally therefore input quality could vary. Arup have selectively reviewed the buildings in the study area, although the accuracy or completeness of the dataset cannot be guaranteed, but is considered suitably accurate for this city-scale project.  Only building polygons which lie entirely within the study area are included.  It is likely that building digitization is poorer in unplanned developments where distinguishing between buildings from imagery is difficult and they are dynamic, changing frequently. Often multiple buildings are digitized as one larger building. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 34 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results  For this project it is assumed that building information is reasonably complete to the end of 2016. 3.2.2 Building typology To further investigate the distribution of risk within Freetown, Makeni and Bo, the buildings datasets are subdivided to create a model which classifies buildings into the following broad categories:  Planned residential;  Unplanned residential;  Industrial;  Educational facilities;  Government facilities;  Healthcare facilities; and  Utility facilities. To identify the distribution of planned residential, unplanned residential and industrial buildings, the following process is undertaken:  On a ward by ward basis, the approximate proportions of planned residential, unplanned residential and industrial buildings is estimated using an array of satellite imagery, ground-based photographs and land-use maps. A specific distribution of each is assigned to each ward to capture the spatial variation in building types across each city.  Large areas of unplanned developments are specifically identified e.g. Kroo Bay.  Using these proportions, the largest buildings within each ward are modelled to be industrial buildings. The smallest buildings within each ward are modelled to be unplanned residential buildings, and the remainder of all buildings within each ward are assigned to be planned residential buildings. Areas specifically designated as unplanned developments (e.g. Kroo Bay) are excluded from this process and are assigned as unplanned residential regardless of size. A number of sources were used to identify the location of educational, government, healthcare and utility facilities within the buildings dataset (Table 4). 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 35 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 4 – Data sources used to identify educational, government, healthcare and utility facilities within the buildings datasets for Freetown, Makeni and Bo. Facility type Freetown Makeni Bo Educational Freetown basemaps Maps prepared by San OpenStreetMap provided by the World Pablo University (01/06/2017). Bank (25/07/2017) and (received 14/07/2017) OpenStreetMap and OpenStreetMap (21/05/2017). (19/05/2017). Government OpenStreetMap OpenStreetMap OpenStreetMap (21/05/2017) and (19/05/2017) and maps (01/06/2017). Freetown basemaps prepared by San Pablo provided by the World University (received Bank (25/07/2017). 14/07/2017). Healthcare OpenStreetMap OpenStreetMap Sierra Leone Ministry of (21/05/2017), Freetown (19/05/2017), maps Health and Sanitation basemaps provided by prepared by San Pablo 2015 District Health the World Bank University (received Information System (25/07/2017) and 14/07/2017) and Sierra (Accessed using Sierra Leone Ministry Leone Ministry of CIDMEWS, data of Health and Health and Sanitation provided by Sanitation 2015 2015 District Health INTEGEMS, District Health Information System 01/03/2017). Information System (Accessed using (Accessed using CIDMEWS, data CIDMEWS, data provided by provided by INTEGEMS, INTEGEMS, 01/03/2017). 01/03/2017). Utility OpenStreetMap OpenStreetMap OpenStreetMap (21/05/2017) and (19/05/2017) and maps (01/06/2017). Freetown basemaps prepared by San Pablo provided by the World University (received Bank (25/07/2017). 14/07/2017). In the first instance, all information on the location of educational, government, healthcare and utility facilities was digitized to vector (point) data if required and duplicates were manually removed. Points were then overlain onto the main building polygon dataset and assigned to nearby buildings using the following order of precedence:  If a building polygon within 50m of the point was tagged as a school, hospital, government department etc.;  To the largest building previously classified as industrial within 50m;  To the largest building previously classified as planned residential within 50m; or  To the largest building previously classified as unplanned residential within 50m. Following this assignment, each facility was inspected in a GIS using an array of satellite imagery, maps and ground-based photographs to ensure that a reasonable 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 36 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results distribution had been achieved. This process allowed the additional manual identification of additional buildings within (say) a school or hospital compound. The number of storeys per residential building is modelled using information from the 2011 Sierra Leone Integrated Household Survey (SLIHS, Statistics Sierra Leone, 2014), which provides approximate proportions of the number of storeys per residential property. All unplanned residential buildings are assumed to be single storey, whilst planned residential buildings are assigned a modelled number of storeys in accordance with the SLIHS (Statistics Sierra Leone, 2014), where the largest (by footprint area) are assigned three storeys and the next largest are assigned two storeys. The remainder of planned residential buildings are modelled as single storey. The number of storeys for educational, government, healthcare and utility facilities is estimated by identifying the minimum size (by footprint area) three storey planned residential building and assigning all educational, government, healthcare and utility facilities greater than or equal to this size as having three storeys. Similarly, two storey educational, government, healthcare and utility facilities are distributed by assigning all facilities with a footprint area greater than the minimum size two storey planned residential building (but less than the minimum sized three storey planned residential building) as having two storeys. The remainder of educational, government, healthcare and utility facilities are modelled as having one storey. Within each city, all modelled building storey information was subsequently manually reviewed and modified if required. This was done using a combination of oblique aerial photography, vertical satellite imagery (if possible) and expert local knowledge by INTEGEMS. 3.2.3 Building replacement value During Mission #2 and throughout the preparation of the Interim Report a wide range of sources were consulted to estimate the replacement value of buildings (Appendix A). This review formed the basis of the replacement values initially used to quantify natural hazard risk in Freetown, Makeni and Bo. Following the Regent-Lumley Disaster, a wealth of new information was made available about the replacement value of buildings. Much of this information is summarized in the DaLA Report (World Bank, 2017). To ensure consistency between this study and the DaLA, it was agreed that the replacement values used to produce the DaLA should also be used for this study (Variation 3, Section 1.4.3.3). Table 5 – Table 7 summarise the building replacement values used to characterise buildings in Freetown, Makeni and Bo, respectively. These values are in line with those used during the DaLA, however are significantly higher than those initially proposed based on the research summarised in Appendix A. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 37 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 5 – Building replacement values (Freetown). Building category Lower estimate Upper estimate Average estimate (USD/m2) (USD/m2) (USD/m2) Educational facilities $129 $631 $380 Formal residential $129 $631 $380 Government facilities $129 $631 $380 Healthcare facilities $129 $631 $380 Industrial facilities $129 $631 $380 Informal residential $17 $26 $21 Utility facilities $129 $631 $380 Table 6 – Building replacement values (Makeni). Building category Lower estimate Upper estimate Average estimate (USD/m2) (USD/m2) (USD/m2) Educational facilities $73 $228 $151 Formal residential $73 $228 $151 Government facilities $73 $228 $151 Healthcare facilities $73 $228 $151 Industrial facilities $73 $228 $151 Informal residential $19 $25 $22 Utility facilities $73 $228 $151 Table 7 – Building replacement values (Bo). Building category Lower estimate Upper estimate Average estimate (USD/m2) (USD/m2) (USD/m2) Educational facilities $69 $227 $148 Formal residential $69 $227 $148 Government facilities $69 $227 $148 Healthcare facilities $69 $227 $148 Industrial facilities $69 $227 $148 Informal residential $27 $33 $30 Utility facilities $69 $227 $148 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 38 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 3.3 Infrastructure 3.3.1 Infrastructure location and alignment OpenStreetMap is the primary source of information about the location and nature of roads for this study. The Global Roads Open Access Data Set (gROADS) was also consulted, however was found to be less complete than the OSM roads dataset. The one freight-only railway in Sierra Leone is not considered by this study. 3.3.2 Infrastructure typology The surface-type for roads was determined by comparison of the OSM roads dataset with satellite imagery and selected ground-based photographs. 3.3.3 Infrastructure replacement value The lower estimate rebuild values for roads are set in accordance with the values used for the World Bank Project, National-Level Landslide Risk Profiles for Sub- Saharan Africa (Stages 1 and 2) (Arup-BGS, 2017). The upper estimate rebuild value for roads is set in accordance with the World Bank Project, Measuring Seismic Risk in Kyrgyz Republic and expert judgement. These values are summarised in Table 8. Table 8 – Rebuild value for roads. Road surface type Lower estimate USD/km Upper estimate USD/km Paved $227,800 $455,600 Unpaved $9,600 $96,000 3.4 Population Population data for this project is taken from the 2015 Sierra Leone Census (Statistics Sierra Leone, 2016). This census information is available disaggregated to Chiefdom level only. To identify the spatial distribution of natural-hazard risk, it was necessary to statistically model the sub-Chiefdom level distribution of the population. This was done using building density as a proxy. To estimate the population in Chiefdoms only partly included in the study area, the average number of households per residential building and then the average number of people per household was used (where the number of residential buildings was taken from the buildings exposure dataset and the household statistics were taken from the 2015 Sierra Leone Census, Statistics Sierra Leone, 2016). To account for the fact that people are less likely to reside in industrial, educational or government facilities at night, the distribution was determined based on: 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 39 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results  60% the distribution of all buildings and roads, representing daytime conditions; and  40% the distribution of just residential and healthcare buildings, representing night-time conditions. An average of these two conditions was used to statistically distribute the Chiefdom level population to a modelled spatial resolution of 30m. In some cases, it was necessary to manually correct the population distribution, for example in the area of the Freetown National Stadium, which showed an unrealistically high population due to high building density. For distributing the population either into buildings or to outside areas (which is required for some aspects of the natural hazard and risk calculations), it is assumed that during the day 50% of the population are indoors (in a given 30m pixel) and 50% are outdoors. At night, it is assumed that 95% of the population are indoors and 5% are outdoors. As above, it is assumed that 60% of the day is daytime and 40% is night-time. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 40 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 41 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 4 Qualitative Hazard and Risk Assessment Methodology 4.1 Introduction This section of the report provides a high-level summary of the methodology used for the qualitative hazard and risk assessment undertaken for Freetown, Makeni and Bo. 4.2 Qualitative hazard assessment The general approach for the qualitative risk assessment follows the simple relationship between hazard, exposure, vulnerability and risk, whereby risk is approximately equal to the product of hazard, exposure and vulnerability. This Section describes how the qualitative hazard assessment is produced and combined with information on exposure to qualitatively assess the risk to each of Freetown, Makeni and Bo to flooding, landslides, sea level rise and coastal erosion. Each qualitative hazard assessment is a general appraisal of the likelihood of the given area to experience a particular hazard (Table 9). This general approach and use of descriptive terms and indicative return periods follows the approach of the Australian Geomechanics Society (2000) and the Institution of Structural Engineers (in the United Kingdom, 2013) amongst many others. The exact method of assessment is specific to each individual hazard and is described in Sections 4.2.1 – 4.2.3. Table 9 – General approach for estimating qualitative hazard. Hazard Descriptor Description Indicative return period Very low Rare The event is conceivable, but only under > 50,000 years exceptional circumstances Low Unlikely The event might occur under very 5,000 – 50,000 adverse circumstances years Medium Possible The event could occur under adverse 500 – 5,000 conditions years High Likely The event will probably occur under 50 – 500 years adverse conditions Very high Almost The event is expected to occur 5 – 50 years certain 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 42 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 4.2.1 Flood hazard Qualitative flood hazard was assessed based on the spatial extent of the modelled flood hazard levels and return periods which have been produced for the quantitative flood hazard assessment. The development of these quantitative flood hazard levels and return periods is discussed in Section 5.1. Each modelled flood hazard level is associated with a given return period, so for each of 20, 50, 100, 200, 500 and 1500-year return period, a dataset is generated showing the modelled depth of flood water over the study area. For this qualitative assessment, the return periods of each full flood extent are categorized according to the general approach outlined in Table 10. Table 10 – General approach for estimating flood hazard. Hazard Descriptor Description Modelled flood return periods Very low Rare The event is conceivable, but only under NA exceptional circumstances Low Unlikely The event might occur under very 1500 years adverse circumstances Medium Possible The event could occur under adverse 100, 200 and conditions 500 years High Likely The event will probably occur under 20 and 50 years adverse conditions Very high Almost The event is expected to occur NA certain 4.2.2 Landslide hazard Qualitative landslide hazard was assessed using a weighted scoring system which classified and then combined slope angle and a built environment density factor with a weighting of 75:25. In Freetown, consideration was also made to the potential for large landslides to become channelized as they run-off from the steep upland slopes. This was done using the spatial extent of the 100 year return period flood dataset. Slope angle was used as the primary factor which contributes to landslide hazard. Slope angle is widely accepted to be one of the most important factors in determining slope stability. Slope angle was reclassified and scored as described in Table 11. Slope angle was determined from the 30m spatial resolution SRTM1 dataset. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 43 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 11 – Categorization of slope angle for qualitative landslide hazard assessment. Score Range of slope angles 1 0° – 5° 2 5° – 10° 3 10° – 15° 4 15° – 30° 5 >30° It was also decided to include a separate built environment density factor in the hazard part of the qualitative landslide risk assessment. This is because it was identified during Mission #1 and Mission #2 and the landslide hazard literature review, that the built environment puts great pressure on the landscape in Sierra Leone and can directly trigger, or increase susceptibility to landslides. For example, in areas where roads are constructed, the hillslopes are flattened to accommodate the road. This means steepening the slope in some way, which can cause landslides. Additionally, in many parts of Freetown in particular, to permit house-building it is first necessary to construct a relatively flat platform. This is done by removing hillslope material, which can in turn cause or promote landslides. The density of the built environment was estimated by the process described in Section 4.3. The density of the built environment was reclassified and scored as described in Table 12. Table 12 – Categorization of the density of the built environment for qualitative landslide hazard assessment. Score Range of density of built environment 1 0% – 20% 2 20% – 40% 3 40% – 60% 4 60% – 80% 5 80% – 100% The score given to slope angle and the density of the built environment were then multiplied together with a weighting of 75:25 and the resulting scores were classified as shown in Table 13 to give the qualitative assessment of landslide hazard. All areas within Freetown which resided within the spatial extent of the 100 year flood return period were classified as ‘medium hazard’, unless the combination of slope angle and built environment density yielded a score which would be classified as ‘likely’ or ‘almost certain’, in which case the higher hazard case was prioritised. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 44 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 13 – Scoring for qualitative landslide hazard assessment. Hazard Descriptor Description Combined score of slope angle and built environment density Very low Rare The event is conceivable, but only 1.0 – 1.8 under exceptional circumstances Low Unlikely The event might occur under very 1.8 – 2.6 adverse circumstances Medium Possible The event could occur under 2.6 – 3.4 adverse conditions High Likely The event will probably occur 3.4 – 4.2 under adverse conditions Very high Almost The event is expected to occur 4.2 – 5.0 certain 4.2.3 Coastal erosion hazard The qualitative hazard assessment comprised an analysis of the variation in likelihood of coastal erosion occurring around the Freetown coastline, including the harbour shoreline and along the creeks. This was done by reviewing the nature of the whole coastline, and dividing into section lengths (cells) of similar composition or morphology. These cells were then assigned a hazard ranking based on the analysis (Table 14). To identify the nature of the coastline and divide into separate cells, satellite imagery available through Google Earth Pro was used. The spatial resolution of this imagery is typically around 1m, although the quality can vary depending on climatic conditions at the time of imaging with some areas more blurred, over exposed or in shadow, with resulting different levels of discernible detail. The satellite imagery of Freetown is made up of a mosaic of overlapping image scenes which were captured at different times. It was not possible to ground truth the satellite image interpretation work in the field, and this would have allowed some sections to be better categorised – but would have been a lengthy process, with difficult access to the shoreline in places. The assessment is therefore considered to be a ‘first-pass’, based on the particular context and constraints of the project, and not a definitive categorisation of the hazard. There was some difficulty in the more built-up parts of the city identifying the exact type of sea or harbour frontage. This was due either to the satellite image resolution, the density of structures near the shoreline or the nature of the seawall, where present. In places it was possible to make reasonable assumptions about the sea-frontage, particularly where these form straight boundaries, with corners. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 45 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 14 – Hazard ranking for coastal cells. Hazard Type of frontage Example Low Rocky headland Figure 6 Concrete or masonry seawall Revetment Vertical engineered quay wall Medium Mangrove swamps Figure 7 Sheltered coves protected by their orientation to the coast Rip rap High Sandy beaches Figure 8 Areas of sand mining Inter-tidal areas and coastlines along estuary and river mouths Areas of fine sediment build-up Figure 6 – Example of low Figure 7 – Example of Figure 8 – Example of high hazard ranking coastal cell medium hazard ranking hazard ranking coastal cell coastal cell 4.2.4 Sea-level rise hazard The potential and possible frequency of sea level inundation events is assessed qualitatively by comparing the elevation of the low-lying areas around the coastline with the potential effects of incremental increases in sea-level. This is based largely on expert judgement, but also on the basis that good practice in the UK dictates that properties should not be constructed at an elevation which is less than the level of the highest astronomical tide + some allowance for sea level rise + an overtopping allowance + 0.3m. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 46 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 15 – General approach to sea level rise hazard assessment. Hazard Descriptor Description Inundation elevation Very low Rare The event is conceivable, but only under 5 – 8m AMSL exceptional circumstances Low Unlikely The event might occur under very 4 – 5m AMSL adverse circumstances Medium Possible The event could occur under adverse 3 – 4m AMSL conditions High Likely The event will probably occur under 2 – 3m AMSL adverse conditions Very high Almost The event is expected to occur < 2m AMSL certain 4.3 Qualitative exposure assessment Exposure data sources are described in Section 3. For the qualitative risk assessment, the buildings and roads datasets were simplified to generate a built environment density map. This built environment density map also acts as a proxy for the location of population, and hence for qualitative risk assessment, only one set of risk maps is produced per hazard and per city. Built environment density maps represent the proportion of each 30m2 area which is covered by building and/or road, where 100% means that the entirety of the area is covered by buildings and/or roads, and 0% means that none of the area is covered by buildings or roads. Qualitative risk maps are not produced differentiated by exposure type and are intended to present a general overview of natural hazard-risk only. Table 16 shows how the different densities of built environment might contribute to the potential consequences should a damaging event occur within a given area. This categorization is used for the qualitative risk assessments for all of the natural hazards. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 47 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 16 – Categorization of built environment density for qualitative risk assessment. Level Descriptor Description Density of built environment 1 Insignificant Little damage 0 – 20% 2 Minor Limited damage to part of structure, or part of 20 – 40% site, requiring some reinstatement 3 Medium Moderate damage to some of structure, or 40 – 60% significant part of site. 4 Major Extensive damage to most of structure, or 60 – 80% extending beyond site boundaries, requiring significant engineering works to remediate/repair. 5 Catastrophic Structure completely destroyed or large-scale 80 – 100% site damage requiring major engineering works to remediate. 4.4 Qualitative risk assessment The qualitative assessment of each hazard is combined with the assessment of exposure/vulnerability (expressed simply as building density) to produce qualitative estimates of risk. This is done using a risk matrix approach as shown by Table 17. Table 17 – A simple risk matrix is used to compare hazard and exposure/vulnerability to qualitatively assess natural-hazard risk. Exposure/vulnerability (density of built environment) 0 – 20% 20 – 40% 40 – 60% 60 – 80% 80 – 100% Very low Very low Very low Very low Low Low Low Very low Very low Low Medium Medium Hazard assessment Medium Very low Low Medium High High High Low Medium High High Very high Very High Medium High High Very high Very high 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 48 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 49 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 5 Quantitative Hazard and Risk Assessment Methodology 5.1 Introduction This section of the report provides a summary of the methodology used for the quantitative hazard and risk assessment undertaken for Freetown, Makeni and Bo. 5.2 Quantitative flood hazard and risk assessment In order to quantify the potential losses in Freetown, Makeni and Bo due to flooding, a catastrophe model was built. Catastrophe models are typically used to estimate the potential financial losses due to events that have the potential to occur in the near future, and those that have occurred in the past. The model development process is illustrated in Figure 9. First, the characteristics of rainfall and river flow were analysed and a spatial and temporal model of flooding across the region was created. Potential future flood events were simulated based on this model, and more extreme and widespread events than have been observed in recent history were created. The simulated events form the multi-peril event set. The severity of these events was then mitigated by the presence of flood defences (where they exist). For each event in the event set, the spatial pattern of severity was provided, and this was converted into a depth and local flood extent using flood hazard maps. Flood hazard maps were produced by computationally routing water depths and volumes over a digital terrain model using a 2D hydraulic model, and allowing the water to flow to the natural low points where flooding occurs. Once the depths and extents of each of the simulated events in the event set were produced, vulnerability functions were used to convert depth into expected percentage damage. The final input into the model was the exposure data, or the portfolio. The exposure data detail the location and value of the properties at risk, or the location and number of people at risk. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 50 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 9 – Schematic diagram of the flood model development and portfolio analysis process. A portfolio of buildings and their values and a population dataset were analysed against the model in JBA's catastrophe modelling platform, JCalf. 5.2.1 Flood modelling grid A grid was required to provide a framework for the model and a resolution on which to carry out the damage and loss assessment. As flood intensity can vary significantly on the order of metres, maintaining as high a resolution as possible in the areas of interest was important. For this reason, a variable resolution grid was developed with high resolution square cells in the study area (Freetown, Makeni and Bo), and lower resolution polygons elsewhere. In addition to this, aggregation polygons have been created to allow probabilistic analysis and reporting of results at administrative boundaries. To allow analysis at the highest resolution possible in the study area, a 30m grid was created that aligned with the 30m flood hazard data, as shown in Figure 10. Outside the study area, the administrative polygons of 'Sections' have been used. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 51 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 10 – Example of 30m grid for Freetown. Inset C shows the grid aligned with the flood hazard maps. 5.2.2 Stochastic flood event set A catalogue of multi-peril stochastic flood events formed a key input to the model. Simulated rainfall over a stationary 10,000-year period was the main driver of both rainfall-induced surface water flood events and river flood events in the model. Importantly, the variability within the catalogue of simulated events is not limited to the range of values in the observation period and by using extreme value theory, contains some events that are more extreme in terms in spatial extent and severity. The event set was output on a network of locations around the country. At each affected location, for each event, the simulated rainfall and river flow levels were converted to return periods, before being further interpolated onto the model grid cells. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 52 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results A return period signifies the average amount of time between an event of a certain magnitude recurring when considering a long time period. For example, a 10-year return period is equivalent to an annual probability of occurrence of 10%, a 100- year return period is an annual probability of 1%. The primary steps to generate the event set were:  The calibration of rainfall-runoff models for gauged catchments based on historical rainfall and river flow data;  The generation of simulated rainfall across the country for a long period;  The simulated precipitation was then used to drive a set of dynamic, conceptual rainfall-runoff models producing daily simulated flow values. The resulting simulated daily precipitation and flow data were then converted into return period estimates (as a measure of severity) using extreme value theory. 5.2.2.1 Input meteorological data A key requirement for the event set is a gridded global dataset of meteorological data, including precipitation and near-surface temperature. This is provided by the Climate Forecast System Reanalysis (CFSR) dataset which is produced and made available by the Environmental Modeling Center, National Centers for Environmental Prediction (NCEP), National Weather Service, NOAA. The main component relevant to the event set is the precipitation field. A temperature estimate for a given catchment is required to model evapotranspiration processes. The CFSR data were used to estimate this series using the near surface atmospheric temperature value averaged over the catchment. River flow data are required for the calibration of rainfall-runoff models. Ideally, for each modelled region, the best available data are sought, with good spatial and temporal coverage, at daily or monthly resolution. Unlike the precipitation or the temperature data, the river flow data exhibit greater inconsistency in quality, depending on the availability of river gauge record. There were no adequate river flow records available for Sierra Leone, so rainfall-runoff models were calibrated using flow data from nearby countries in terms of their similarity. One global dataset used in this project is the monthly river flow data from National Center for Atmospheric Research (NCAR) Computational & Information Systems Lab, Research Data Archive (CISLRDA). A catchment's seasonal temperature summary and climate conditions provide a key indicator of its expected rainfall-runoff characteristics. The event set uses the Köppen climate classification scheme as formulated by Kottek et al (2006). The Köppen climate classification for Sierra Leone is Tropical Monsoon Climate (‘Am’). The dominant land use for a catchment is also a key indicator of its expected properties. This was defined with reference to the European Space Agency (ESA) GlobCover map that uses 23 classifications for land use (Arino et al., 2012). The 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 53 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results majority of Sierra Leone is covered by mosaic vegetation, with some semi- deciduous forestry in coastal areas. A catchment's dominant soil type was defined using the 27 Great Group level type classification system and map defined by Zobler (1999). The dominant soil types range from gleysol along the coast, to lithosol in higher altitude regions. Several studies, such as the UK Centre for Ecology and Hydrology (CEH) Flood Estimation Handbook (Baylis, 1999), suggest elevation and slope percentiles can be useful for categorising catchments. To this end, the event set method includes collecting the elevation and slope for all cells within a catchment using the NOAA GLOBE 1 km resolution digital elevation model data. The 10th, 50th and 90th percentile values are stored as catchment attributes. Where necessary, the bounding box of very small catchments is increased until it contains a minimum of 25 cells. As the slope requires the difference in elevation between adjacent cells, a cube of 25 cells allows 2 concentric squares around a central cell. The WorldClim gridded precipitation data are used to estimate monthly mean precipitation for each catchment. The data are described in Hijmans et al. (2005), and were used as a catchment attribute in the nearest neighbour interpolation of catchment properties. 5.2.2.2 Simulation of stochastic rainfall The next stage of the event set generation process was the simulation of stochastic rainfall. These data were generated using a statistical model that combined non- parametric probability distributions with semi-parametric extreme value models. This choice of approach reflects our interest in both extreme rainfall (which directly causes surface water flooding, and is a key driver of river flooding) but also non-extreme rainfall (which may affect catchment conditions such as soil moisture), and therefore can also affect river flood frequency and severity. Once completed, the rainfall outputs were taken forward in two ways:  To directly describe the surface water flood events in the model; and  As the key input to the estimation of river flow events using optimised rainfall-runoff models. 5.2.2.3 Calibration of rainfall-runoff models There was inadequate river flow data available in Sierra Leone (one gauge in Bo for a limited time), so a rainfall-runoff approach to modelling river flooding was taken. This had the benefit of being able to predict river flows at ungauged locations. Rainfall-runoff models have been created and calibrated for several thousand gauged catchments across the world, distributed across a wide range of climatic, area, topographic, soil and land use types. The most appropriate model was selected for each catchment in Sierra Leone based on the similarity of the catchment characteristics to the pre-calibrated models. To derive the parameter values that relate rainfall to flow for each gauged river catchment, the Identification of unit Hydrographs And Component flows from 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 54 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Rainfall, Evaporation and Streamflow data (IHACRES) rainfall-runoff modelling framework, as implemented by Jakeman et al. (1990) was used. Since IHACRES parameters represent the physical properties of river catchments, plausible parameter ranges could be enforced. This provided assurance that model responses remained sensible beyond the range of calibration data (i.e. when applied to simulate the river flows associated with synthetic rainfall events). Figure 11 shows the two model components used by IHACRES in orange. Firstly, rainfall and temperature time-series were input into each catchment model. The soil moisture accounting component then apportions the rainfall between catchment storage, evapotranspiration and runoff. Finally, the runoff, together with any drainage from catchment storage, is then conveyed dynamically to stream flow via the flow routing component. Figure 11 – The two components of the IHACRES modelling framework (in orange) together with their respective inputs and outputs (in grey). 5.2.2.4 Ungauged river catchments A key requirement of the event set method was to provide flow estimates for points along a river network that do not coincide with the location of a river gauge. This is a common challenge in flood risk management (Sivapalan, 2003). The approach taken here is to relate ungauged sites to their nearest-neighbour gauged sites using their catchment attributes. The rainfall and temperature for the ungauged site is then applied to these models and their output is merged using a weighted mean. The weighting is inversely proportional to a measure of dissimilarity among the output series; i.e., if one of the gauged sites produces a flow series that is very different from the others, it will make a smaller contribution to the final merged output. 5.2.2.5 Estimation of flood level return periods Simulated physical intensities (i.e. rainfall depths and river flow levels) of each event per location per peril were converted to return periods using a peaks-over- threshold method. This enabled the depths from the correct flood maps to be selected for each simulated event to produce flood depth footprints for each event. A peak-over-threshold method was used due to its insensitivity to the rare years where no extreme values exist. Under this method, a high threshold was first selected, and the individual extreme values (flow or rainfall) were identified using Smith's runs method (Davison and Smith, 1990). Then, a Generalised Pareto Distribution (GPD) was fitted to the extreme values, before being combined with the extremal index (from Smith’s runs method) to convert physical quantities to 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 55 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results return periods. GPD parameters were estimated for each 10,000-year data series (i.e. at each location, for each peril) using the R evd package (Stephenson, 2002). Figure 12 illustrates the translation of one river flow time series into return periods. Figure 12 – Illustration of the translation between a river flow time series (top) to a return period time series (bottom). 5.2.2.6 Flood event definition Events were selected into the final event set based on the following principles:  An event can last for multiple days, and across multiple flood perils (surface water or river);  The start of an event is the day when the physical intensity at any modelled location, for any peril, exceeds a fixed return period (specifically the 1 in 10- year level);  An event ends when the physical intensities at all modelled stations stay below the return period threshold for 3 consecutive days. 5.2.2.7 Flood event interpolation The maximum severities for each event and for each peril were interpolated onto the model grid. This was done via a series of intermediate 'interpolation points', as illustrated by the conceptual diagram (Figure 13). Inverse distance weighted 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 56 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results approaches were taken based on the locations of the observation stations relative to the cells. The interpolation points used in Sierra Leone for river and surface water are displayed in Figure 14 to Figure 17. Figure 13 – First a dense network of gauges is derived, then the events at the original observation points are interpolated onto these points, and finally onto the analysis cells. Figure 14 – Surface water observation points (orange) and interpolation points (grey) across Sierra Leone. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 57 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 15 – Surface water interpolation points across Freetown. Figure 16 – River observation points (orange) and interpolation points (grey) across Sierra Leone. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 58 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 17 – River interpolation points (grey) along river drainage lines (blue) across Freetown. 5.2.3 Flood hazard mapping Once the spatial pattern and severity of each event had been defined in the event set, water depth information at each location was extracted from the flood hazard maps. The hazard maps provide indicative flood extents and depths for undefended river and surface water flood hazard for multiple return periods for every 30m grid-cell across the country. Table 4 shows the return periods modelled for Sierra Leone. Table 18 – Modelled flood hazard map return periods for Sierra Leone. Peril Return period modelled River flooding 1 in 20, 1 in 50, 1 in 100, 1 in 200, 1 in 500, 1 in 1,500 years. Surface water flooding 1 in 20, 1 in 50, 1 in 100, 1 in 200, 1 in 500, 1 in 1,500 years. Figure 18 shows the overview of the hazard mapping method. First, the digital elevation model (DEM) was processed to determine the location of the river centrelines and the catchment characteristics. Next, statistical models were fitted to historical river gauge and rainfall data to estimate flows for specified return periods. Finally, the input flows for each return period were run in the hydraulic model allowing water to spread over the terrain model and provide the flood extents and depths. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 59 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 18 – Global Flood Map process chart. 5.2.3.1 Digital terrain data The underlying digital elevation data is the NEXTMap World 30 Digital Surface Model (DSM) provided by Intermap Technologies Inc. A DSM represents the earth’s surface as detected by radar and therefore includes features such as tree canopies and tall buildings. 5.2.3.2 River flow and rainfall estimation Small rivers have a different hydrologic and hydraulic behaviour to large rivers, so were modelled slightly differently. The inputs into a hydrological model for small rivers need to be of a higher precision than those used for large rivers. The different approaches used by JBA are explained below. To identify the large rivers in Sierra Leone, a variety of GIS tools were used to enable analysis of the flow directions and accumulations over the digital terrain data. Validation was carried out to ensure the derived drainage network closely matched watercourses visible in background mapping and aerial imagery and where the network was found to be incorrect, it was adjusted. The drainage network is shown for Sierra Leone, Freetown, Makeni and Bo in Figure 19 – Figure 22. Major drainage lines correspond to large rivers, and minor drainage lines correspond to small rivers. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 60 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 19 – Major (dark blue) and minor (light blue) river drainage lines across Sierra Leone. Figure 20 – Minor (light blue) river drainage lines for Freetown and the surrounding areas. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 61 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 21 – Minor (light blue) river drainage lines in Bo. Figure 22 – Minor (light blue) river drainage lines in Makeni. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 62 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results To estimate river flow along the river network, an empirically based rainfall- runoff approach was used that linked design rainfall to design river flood magnitude. The model was based on rainfall data provided by the Climatic Research Unit (CRU) at the University of East Anglia (UK). The approach made use of the CRU TS 3.2 dataset which presents gridded rainfall data at 0.5° spatial and monthly temporal resolution for the 1901-2011 period. The CRU dataset has global coverage, is consistent and is of good quality, enabling a reasonably reliable 110-year long monthly rainfall record to be derived for the catchment upstream of any point along the drainage network. The data were readily analysed to provide design rainfall statistics for a range of return periods. Calibration was then required to transform design rainfall to a design flood peak. To do this, an empirical relationship calibrated on observed river discharges was used. The impact of different climatic processes and meteorological regimes (for example the monsoon) is inherently captured in the rainfall data. Small rivers are much more sensitive to short intense rainfall events than larger rivers, so a historical rainfall record of higher temporal and spatial resolution was used. The Climate Forecast System Reanalysis (CFSR) precipitation data available from the National Center for Environmental Prediction (NCEP) provides hourly rainfall totals on a 0.3° grid for the 1979-2009 period. The duration of a rainfall storm is important to consider in this analysis. Short, intense rainfall events tend to generate more flooding in steep river valleys, whereas flatter regions are often more severely affected by slower moving storms. Design rainfall storms for three storm durations were generated – the 1-hour, 3- hour and 24-hour storm. This was done by carrying out extreme value statistical analysis of the CFSR data, resulting in 18 design rainfall totals for each 0.3° tile (three storm durations for each of the six return periods). These were smoothed and interpolated to generate a continuous rainfall surface for each return period and storm duration. Where available, local Intensity-Duration-Frequency (IDF) curves and national rainfall statistics were used to supplement the CFSR data and to calibrate the rainfall estimates. The rainfall estimates were then used to generate storm hyetographs for each individual model catchment, using a triangular profile. A triangular profile is a simple representation of short rainfall events but is important because it captures the increasing rate of rainfall to a peak. A steady-state model would distribute the total event volume equally over the storm duration therefore missing that peak intensity which is often the main feature of a short-intense storm that results in flooding. JBA has developed two cutting edge hydraulic models to enable flood mapping to be carried out at a country scale and both were applied in Sierra Leone. Choosing the most appropriate model is important as both models have their own advantages in different situations. Table 19 provides highlights of the two models. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 63 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 19 – Overview of JBA’s hydraulic models. RFlow JFlow® 1D normal depth model with 2D flood 2D hydrodynamic, solves shallow water spreading equations Optimised for use on DSM Ideal for use on bare-earth DTM Less affected by noise and artificial high Requires structures such as bridges to be ground removed prior to modelling Extremely rapid Runs on largest dedicated flood modelling computer grid in the world Calibrated against JFlow (2D) and ISIS The first hydrodynamic flood modelling (detailed 1D) – found to perform similarly in software to repurpose gaming technology most terrains Independent peer reviews Finalist for the UK’s most prestigious award for innovation in engineering (The Royal Academy of Engineering’s MacRobert Award) RFlow has been used to model all of the large rivers in Sierra Leone where a bare- earth DTM is unavailable. It is a 1D model based upon normal depth calculations and custom 2D flood spreading tools. RFlow is optimised for modelling areas where input data are sparse, for example where there is limited hydrological information like in Sierra Leone. RFlow is designed to work with DSM features that would otherwise block flow in conventional 2D models causing build-ups of water (e.g. buildings, trees). RFlow requires three main inputs – elevation data, the drainage network and river flow estimates (in m3/s) at the upstream and downstream end of each watercourse section. The model interpolates between each flow estimation point at intervals between 200m and 5km, and creates a cross section at each interpolation point. Manning’s equation is then used to estimate flood depth at each cross section, which is spread downstream using JBA’s flood spreading algorithms. JBA’s full 2D hydrodynamic model has been used to model the following:  Large rivers where bare-earth DTM is available;  Large rivers where no DTM is available but 1D model is inappropriate, such as very flat, wide floodplains;  All small rivers; and  Surface water flooding. JFlow can be run in different configurations for different purposes. For the large rivers, a traditional fluvial model configuration has been used. This requires regularly spaced points along the drainage network. At each point, a flood hydrograph is generated for each return period. The volume of water that can be held within the river channel is estimated and removed from the flood simulation. JFlow then spreads the remaining volume of water for each return period across the floodplain by solving full shallow water equations. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 64 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results For the small rivers and surface water, a direct-rainfall configuration has been used. This approach applies the design hyetographs to an entire catchment. Different runoff rates are applied across each catchment to reflect spatial variations in climate and land use. In towns and cities, urban drainage systems are accounted for by removing the 5-year rainfall total from the flood simulations. The assumptions used for the hydraulic modelling are as follows:  All rivers are assumed to be undefended;  Artificial changes to the channel capacity are not accounted for; and  Structures such as bridges, culverts, weirs, dams or pumping stations are not accounted for. The output from the hydraulic models was the maximum depth of flooding for each 30m cell for a range of return periods. 5.2.4 Combining the event set and flood hazard maps Flood footprints indicating the depth and extent of the flooding for each stochastic event in the event set were required in order to calculate flood damage to people and buildings. To generate this dataset, return period hazard maps were summarised by the analysis cells, and then combined with the stochastic event set. For each modelled return period hazard map, the proportion of area affected (PAA) was calculated (the area of each cell or polygon flooded), and the 5th and 95th percentile flood depths were returned. As the return periods per event from the stochastic event set had already been transferred onto the same cells (as described in Section 5.2.2.7), these two datasets were merged to produce the PAA, 5th and 95th percentile flood depths per cell, per event. Where the return period of a stochastic event did not match a design return period from the hazard maps, linear interpolation between modelled return period depths was conducted. 5.2.5 Flood defences Aerial imagery suggests a lack of river flood defences in the cities of Freetown, Makeni and Bo. In Freetown, buildings and informal settlements are situated within and adjacent to the river channels. 5.2.6 Flood vulnerability functions Vulnerability functions enable the damage to properties and human life to be estimated for each event in a stochastic event set, given the depth and location of flooding. The functions describe the expected mean damage ratio (expressed as a proportion of total value) of a risk for a given hazard intensity. When considering flood, depth is the primary explanatory variable with respect to damage (e.g. Büchele, et al., 2006; Kreibich, et al., 2010; Merz, et al., 2013) so depth-damage functions were employed in this model. However, significant 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 65 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results epistemic uncertainty is present due to the impact of other flood characteristics such as flood duration and water velocity on the amount of damage experienced. In addition to this, aleatory uncertainty in how a structure will react to a given water depth is also present. These sources of uncertainty are accounted for in the standard deviation of the vulnerability functions. For buildings, the mean damage ratio expresses the cost of damage against the total reconstruction cost for each building. For human vulnerability, the mean damage ratio expresses the estimated chance of a person being directly affected or killed by a certain depth of water. In this model, a person is deemed to be affected if their property is inundated with a water depth of above 15cm. Figure 23 shows an example of a buildings vulnerability function and the associated uncertainty (shaded in orange). Figure 23 – A typical depth-damage vulnerability function. The orange line represents the mean value, and +/- one standard deviation is shown shaded. The damage ratio is the percentage of the property value (Total Inventory Value, TIV) lost. For this model, building vulnerability functions were created for each building type. Surface water (flash flood) and river flood employ the same vulnerability functions. 5.2.6.1 Data sources for vulnerability functions Vulnerability functions were developed based on peer-reviewed academic research, and the functions are calibrated against available claims and rebuild cost data where available. For example, vulnerability functions from Penning-Rowsell 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 66 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results et al. (2010) were used as a basis for the commercial vulnerability functions. The values were adapted based on rebuild costs, however, the shape of the functions remained the same. Very little information is available on industrial property types and vulnerability, but Admiraal (2011) provide data for different industrial property types. A number of other papers including Messner et al. (2007) and Su et al. (2009) were used to support and further develop the vulnerability functions along with claims data and rebuild costs where available. The vulnerability functions for Sierra Leone were informed by previous experience of model development and claims data in India, Vietnam, Sri Lanka and Thailand. The vulnerability functions for people affected were developed based on the assumption that floodwater starts to enter a property when it is approximately between 1cm and 30cm deep, so a 15cm threshold was implemented here. This assumption is based on research by the UK Environment Agency and supported by experience from past flood events. The vulnerability functions assume that people are affected if the water depth around their property exceeds 15cm. The human fatality vulnerability function used was based on research by Waarts (1992, quoted in Jonkman, 2002): . 0.665 10 This relationship only considers the direct fatalities due to drowning, and does not take into account indirect flood fatalities due to issues such as contaminated drinking water. 5.2.7 Estimation of future flood risk In addition to calculating the annual economic and social impact of flood damage, future flood risk has also been modelled to provide a quantitative risk assessment of the impact climate change may have on losses in 2050 in Freetown, Makeni and Bo. Projected rainfall return periods for 2050, under scenario RCP8.5, were generated using two CMIP5 models (NORESM1-M and MPI_ESM_MR) for each city, as well as current rainfall return periods. A multi-model mean was calculated for both datasets, and the projected results compared with the current rainfall return periods per city (Figure 24 – Figure 26). 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 67 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 24 – Current and future rainfall return periods based on a multi-model mean projection for 2050 for Freetown. Figure 25– Current and future rainfall return periods based on a multi-model mean projection for 2050 for Makeni. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 68 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 26 – Current and future rainfall return periods based on a multi-model mean projection for 2050 for Bo. The average of the multi-model mean was calculated for each return period across the three cities, i.e. an average projected rainfall for the 2-year return period, for both the current and future values. The projections were then aligned to the current rainfall return period values to gauge the scale of the change between now and 2050 (Table 20). It was beyond the scope of the project to adjust the rainfall inputs to the flood mapping or the event set. Instead, the change in the frequency of high rainfall values was used to adjust the frequency of high loss events that were output from the model. It was not possible to apply the adjustment factors to the three cities independently due to a lack of data at the low return periods, so the adjustment factor was applied to the overall loss results for the three cities combined. Table 20 – Average changes in current and future rainfall return periods with future values aligned to the closest current value. Return period 10 100 500 1500 Average historic Precipitation (mm/day) 93 115 127 133 Return period 2 5 10 25 Average projected Precipitation (mm/day) 92 113 125 137 The average projected return periods were then aligned with the current mean loss results for each exposure type (people affected, fatalities, buildings loss) to 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 69 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results provide projected losses under RCP8.5. Using this method, projected return periods for 2050 were only available for up to 25-years. Although this method may be useful to plan and mitigate against a worst-case scenario, it is important to recognise that key social and economic factors are not accounted for. One example of a non-stationary variable is population; Sierra Leone's population has steadily increased and records show that, as of 2016, total population stood at ~7,000,000 people (UNDP, 2017). Projections indicate this rise in total population to continue up to and beyond 2050, which is not factored in to the climate change loss projections. It should be noted that RCP8.5 is widely acknowledged to be the most extreme climate scenario and so it may be beneficial to consider using simulations under RCP4.5 or RCP6.0 to gain an alternative and broader picture of the impact climate change may have on natural hazard-risk in Sierra Leone in future studies. 5.3 Quantitative landslide hazard and risk The quantitative landslide hazard and risk assessment is sub-divided into four main parts:  Landslide susceptibility assessment;  Landslide hazard assessment;  Vulnerability assessment; and  Landslide risk assessment. 5.3.1 Landslide susceptibility assessment Landslide susceptibility assessment is based on the premise that a range of parameters need to be combined to obtain an approximation of the conditions in the landscape that determine the propensity of slopes to generate landslides. The hierarchy of properties and processes that determine the stability of slopes is complex and variable in both time and space. However, it is possible to capture the most influential factors in a simple set of algorithms that can reflect the spatial changes in slope stability that correspond with reality (Arup-BGS, 2017). For this project, a bivariate statistical analysis approach was applied and validated to derive a landslide susceptibility map for Freetown at city-scale. The approach, which calculates Landslide Susceptibility Index (LSI) is described by Chalkias et al. (2014). The following factors were used to represent landslide susceptibility:  Elevation;  Slope angle;  Slope aspect; and  Distance to rivers. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 70 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Each of these factors was categorized (e.g. for slope angle, 0° – 10°, 10° – 20° etc.) and compared to a random subset of the landslide inventory compiled for this project4. It is then possible to determine how frequently landslides are triggered in particular categories for each susceptibility factor (for example, it might be found that 75% of landslides are triggered where slope angle is in the range of 30° - 40°). This process is repeated for each susceptibility factor. LSI is calculated for each category (j) of each susceptibility factor (i) as: ln Nij is the number of landslides in category j of the factor i, Aij is the area of this category, Nr is the total number of landslides and Ar is the total study area. Categories which are highly correlated to landslide occurrence produce high relative LSI scores, whilst categories which are negatively correlated with landslide occurrence produce low negative LSI scores. It is therefore desirable to identify a range of LSI scores for each susceptibility factor, showing both areas which are well correlated with landslide occurrence and areas which are not. The overall susceptibility score for each pixel is determined by: 1 LSIi is the susceptibility for the factor i and n is the total number of factors. This calculation is all undertaken using spatial raster datasets to give an indication of the spatial distribution of landslide susceptibility. All calculations are completed at a spatial resolution of 30m. The relative success of the landslide susceptibility mapping was determined by an Area Under Curve (AUC) approach which allowed identification of both success and prediction rates. To identify the best possible landslide susceptibility map from the most appropriate combination of factors and category sub-divisions, a systematic approach was undertaken to trial 4096 different combinations of factors and categories. This is exemplified by Table 21. 4 Randomly sub-setting the landslide inventory used to calibrate the LSI allowed determination of both the success rate and prediction rate curves. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 71 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 21 – Overview of the systematic approach used to identify the best combination of landslide susceptibility factors and the number of categories used. Iteration Number of categories per factor Success Prediction rate rate Elevation Angle Aspect Distance to rivers 1 3 3 3 3 66% 62% 2 4 3 3 3 67% 65% 3 5 3 3 3 64% 60% - Trial all possible combinations and rank based on success rate - 4096 10 10 10 10 71% 67% Using this approach, the highest success rate was identified to be (Table 22): Table 22 – The best combination of landslide susceptibility factors and the number of categories used as identified by the highest success rate of the 4096 trials. Iteration Number of categories per factor Success Prediction rate rate Elevation Angle Aspect Distance to rivers 3064 9 10 8 10 84% 78% Figure 27 shows the success rate and prediction rate curves associated with this combination of factors and categories. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 72 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Success Predict Proportion of landslides source areas in the susceptiblity classes 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 Proportion of study area in susceptiblity classes Figure 27 – Success rate and prediction rate curves associated with the landslide susceptibility map developed for Freetown using the landslide inventory. Following the approach of Jaiswal et al. (2011), the high landslide susceptibility class is classified as the highest 25% scoring pixels. This high landslide susceptibility class hosts approximately 75% of the landslides identified in the landslide inventory. The medium landslide susceptibility class is defined as the 25% - 40% highest scoring pixels. More than 80% of all landslides in the full landslide inventory are identified by the high and medium landslide susceptibility classes. 5.3.2 Landslide hazard assessment Landslide hazard assessment is sub-divided into two main steps:  Landslide frequency-magnitude analysis; and  Landslide runout modelling. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 73 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 5.3.2.1 Frequency-magnitude analysis Landslide frequency-magnitude analysis describes how often landslides of different sizes occur within the study area. Because the landslide inventory compiled for this project focuses on the spatial extent of the rupture surface of landslides in Freetown, this analysis defines the magnitude of landslides by the area of the rupture area. Landslide debris depth (and hence volume) is not specifically considered, however it is implicit that typically landslides which occur over a wider rupture area may mobilise a greater thickness of debris. Five landslide area categories are considered based on the distribution of landslide rupture surfaces identified in the landslide inventory:  Magnitude 1: Landslides with a rupture surface area of between 100m2 and 1,000m2;  Magnitude 2: Landslides with a rupture surface area of between 1,000m2 and 10,000m2;  Magnitude 3: Landslides with a rupture surface area of between 10,000m2 and 30,000m2;  Magnitude 4: Landslides with a rupture surface area of between 30,000m2 and 100,000m2; and  Magnitude 5: Landslides with a rupture surface area of more than 100,000m2. Because the compiled landslide inventory was not able to identify the exact age of each landslide, it is necessary to make an assumption about the probable time- period which the inventory represents to understand the frequency of landslides of different sizes. That is, if the inventory represented 1,000 years and contained 100 landslides of a particular size, the approximate annual frequency attributed to landslides of that size would be 0.1, or 1 in 10 years. Due to the uncertainty associated with this step, a range of inventory ages are input into the frequency- magnitude analysis. Based on the literature review (included in the accompanying Freetown-specific report) it is assumed that the inventory represents between 500 and 3,000 years of landslide observation. Additionally, it is unlikely that the compiled landslide inventory contains all landslides of each size class. Smaller landslides are more likely to be under- represented in the inventory because they may be less readily identified from satellite imagery or may have become re-vegetated, leaving little visible rupture surface. To account for this in the frequency-magnitude analysis, a completeness factor is applied to factor-up the number of landslides within each magnitude class. The following assumptions have been made about landslide inventory completeness within each magnitude class:  Magnitude 2: Between 25% and 50% of all landslides which have occurred are represented in the landslide inventory;  Magnitude 3: Between 35% and 60% of all landslides which have occurred are represented in the landslide inventory;  Magnitude 4: Between 45% and 70% of all landslides which have occurred are represented in the landslide inventory; and 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 74 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results  Magnitude 5: Between 55% and 80% of all landslides which have occurred are represented in the landslide inventory. As detailed above, a range of completeness estimates have been used to allow modelling of the uncertainty associated with this step. Note that Magnitude 1 landslides are not well represented in the landslide inventory. This is because:  These smaller landslides are extremely difficult to identify from historical satellite imagery due to their small size;  These landslides are typically unreported unless they occur in populated areas; and  The occurrence of these small landslides is often associated with anthropogenic processes, for example slope over-steepening to accommodate buildings or infrastructure. Since this is a relatively recent process (certainly in comparison to the ~500 – 3,000 year old landslide inventory), the landslide inventory cannot well represent these features. The annual frequency of Magnitude 1 landslides is modelled as being between 10 and 100 events across the whole study area. This is based on limited and incomplete information available from the literature review. As above, by introducing a range it is possible to define a range of plausible risk estimates from the landslide risk assessment. It is assumed that landslides <100m2 cannot be identified or accounted for by this city scale study. The spatial distribution of the annual frequency of Magnitude 2 – 5 landslides was distributed between the areas identified as high (75%) and medium (25%) landslide susceptibility. 5.3.2.2 Runout modelling Landslide runout is modelled using a Gravitational Process Path (GPP) model (Wichmann, 2017), which is used to determine the probability that a given pixel within a digital elevation model will be affected by landslide debris which is initiated from a different pixel. This approach can provide a reasonable approximation of landslide runout probability at city-scale and provides sensible results when used with a 30m spatial resolution digital elevation model. The logic of the GPP model is as follows:  A landslide is initiated from a given pixel within the digital elevation model;  Flow may then occur between the initial pixel and any surrounding pixel which is at a lower elevation that the starting pixel;  The probability that flow occurs into any surrounding pixel is dictated by the elevation difference to the initial pixel. Generally, it is more likely that flow will occur between adjacent pixels with greater elevation difference, however the model can be calibrated to control this; 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 75 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results  Based on this probability, a weighted-random flow path is taken from the initial pixel into successive pixels;  The process is repeated until such a time that the model reaches some pre- determined limit (e.g. maximum landslide length, maximum angle of reach) or flows into a depression in the digital elevation model;  This whole flow path process is then iterated a number of times, starting again at the initial pixel and again choosing a weighted-random flow path until the model limit is reached again, and so on and so forth; and  The probability that flow will reach a given pixel within the digital elevation model is then determined by the number of times which a pixel was incorporated into the flow path divided by the number of model iterations. The annual frequency with which a given pixel is affected is equal to the annual frequency of triggering (i.e. the annual frequency of landslide occurrence at the initial pixel) multiplied by the probability that the pixel would be affected by flow form the initial pixel. This process is repeated for all pixels classified as having high or medium landslide susceptibility and the resulting annual frequencies are summed on a pixel-wise basis. This process is repeated a total of eight times, using the lower and upper bound estimates of annual frequency of occurrence of landslide magnitude classes 2 – 5. Flow paths are not considered for Magnitude 1 landslides because the spatial resolution of the digital elevation model exceeds the landslide size. The flow path characteristics of different magnitude class landslides (2 – 5 only) are controlled using the model variables (Table 23). 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 76 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 23 – Model variables used to control the modelled landslide flow paths. Parameter Description Slope A slope angle above which flow is always from the initial pixel to the steepest threshold available neighbour. Divergent A model factor which controls the amount of divergent flow when below the flow factor slope threshold angle. Persistent flow A model factor which can be used to achieve greater fixation in the direction factor of movement of the flow path. Horizontal A model factor which can allow flow between adjacent pixels at the same flow control elevation. Fahrboschung A limiting factor on the model which can cause the flow path to end when a angle critical angle is reached between the initial pixel and the pixel at the end of the current flow path. Number of This factor controls how many times the flow path is iterated to determine the paths probability that a given pixel will be affected by flow. Maximum A limiting factor on the model which can cause the flow path to end when it number of had made a specified number of steps between adjacent pixels. steps Maximum A limiting factor on the model which can cause the flow path to end when it length has travelled a specified horizontal distance from the initial pixel. Different parameters are defined for each magnitude class based on comparison of model results with the 5 entries in the landslide inventory which have information on both rupture area and runout. Runout from the Regent-Lumley Landslide was also used to calibrate the flow path model parameters. An example of the quantitative landslide hazard assessment results for Freetown are shown in A4 on Page 78. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 77 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 78 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 5.3.3 Vulnerability assessment The impact of a landslide event on the population, structures and infrastructural assets of an area varies from no loss (typically assigned the value 0), to total loss (assigned the value 1). Corominas et al. (2014) define two major types of vulnerability, namely:  Physical vulnerability, which refers to direct damage caused to structures and infrastructural assets; and  Vulnerability of people, referring to whether or not a landslide event will result in injuries or fatalities. 5.3.3.1 Vulnerability/fragility of buildings An assessment of the vulnerability of structures and infrastructure (or ‘physical vulnerability’ analysis) is an important step in the landslide risk analysis process. Consider a single structure at the foot of a slope – susceptibility and hazard analyses (including triggering and reach probability) estimate the probability that a landslide of a given size will affect the structure. The physical vulnerability of the structure describes its probable response to being affected by a landslide of a given magnitude. This response might be superficial damage, functional damage, or structural damage (Table 24) and logically will vary according to the construction and condition of the structure (Table 25). The ability of the structure to resist damage therefore controls not just the economic losses which result from having to rebuild or repair it, but the vulnerability of the persons within it. Table 24 – Damage classes of exposed assets (Cardinali et al., 2002). Damage state Description Superficial/cosmetic Superficial (aesthetic/minor) damage where the functionality of structures/infrastructure is not compromised and the damage can be readily repaired rapidly and at low cost. Functional Functional (medium) damage where the functionality of structures/infrastructure is compromised and the damage takes time and large resources to be fixed. Structural Structural (severe or total) damage where structures/infrastructure are severely or completely damaged, requiring extensive, costly repairs. Demolition may be necessary. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 79 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 25 – Relative structural resistance of buildings of different typologies (Heinimann, 1999). Structure construction Resistance Lightest simple constructions (timber) No resistance Light construction Very weak Mixed structure (concrete and timber) Weak Brick walls, concrete Medium resistance Reinforced concrete Strong Reinforced Very strong For site specific landslide risk assessment this concept of physical vulnerability can be greatly expanded upon to consider intricacies such as the diameter of falling boulders in a rock-fall or the point of impact of a boulder relative to the structural layout of a building (Pitilakis et al., 2011). Region and city-scale studies such as this are better suited to methodologies which incorporate broad categories of structure typology and maintenance state (e.g. Du, Nadim and Lacasse, 2013). There are two paradigms for quantifying physical vulnerability:  The application of bulk vulnerability functions, which estimate the damage state which will occur as a result of a structure being affected by a landslide of a certain size (e.g. Table 26); or  The use of fragility functions, which estimate the probability that a certain damage state will occur as a result of a structure being affected by a landslide of a certain size. This study utilises the second approach, defining fragility functions which are guided by published bulk vulnerability functions and are informed by expert judgement. Fragility functions and loss ratios are defined in Section 5.3.3.1, 5.3.3.2 and 5.3.3.3. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 80 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 26 – Bulk vulnerability functions of exposed structural and infrastructural assets (Cardinali et al., 2002) (modified by Lee and Jones, 2004). Landslide intensity Buildings Main Secondary Minor Railway roads roads roads lines Light Rockfall Cosmetic Cosmetic Cosmetic Cosmetic Cosmetic Debris flow Cosmetic Cosmetic Functional Functional Cosmetic Slide Cosmetic Cosmetic Functional Cosmetic Cosmetic Medium Rockfall Functional Functional Functional Functional Functional Debris flow Functional Functional Functional Functional Functional Slide Functional Functional Structural Structural Functional High Rockfall Structural Structural Structural Structural Structural Debris flow Structural Structural Structural Structural Structural Slide Structural Structural Structural Structural Structural V. high Rockfall Structural Structural Structural Structural Structural Debris flow Structural Structural Structural Structural Structural Slide Structural Structural Structural Structural Structural Table 27 to Table 29 present the simple fragility functions defined to characterise the buildings in Freetown. Table 27 – Probability of cosmetic damage or higher to buildings (showing range). Building category <1,000m2 1,000m2 - 10,000m2 - 30,000m2 - >100,000m2 2 2 10,000m 30,000m 100,000m2 Educational facilities 0.55 - 0.75 0.9 - 1 1-1 1-1 1-1 Formal residential 0.55 - 0.75 0.9 - 1 1-1 1-1 1-1 Government facilities 0.55 - 0.75 0.9 - 1 1-1 1-1 1-1 Healthcare facilities 0.55 - 0.75 0.9 - 1 1-1 1-1 1-1 Industrial facilities 0.55 - 0.75 0.9 - 1 1-1 1-1 1-1 Informal residential 1-1 1-1 1-1 1-1 1-1 Utility facilities 0.55 - 0.75 0.9 - 1 1-1 1-1 1-1 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 81 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 28 – Probability of functional damage or higher to buildings (showing range). Building category <1,000m2 1,000m2 - 10,000m2 - 30,000m2 - >100,000m2 2 2 10,000m 30,000m 100,000m2 Educational facilities 0.35 - 0.55 0.7 - 0.9 0.95 - 1 1-1 1-1 Formal residential 0.35 - 0.55 0.7 - 0.9 0.95 - 1 1-1 1-1 Government facilities 0.35 - 0.55 0.7 - 0.9 0.95 - 1 1-1 1-1 Healthcare facilities 0.35 - 0.55 0.7 - 0.9 0.95 - 1 1-1 1-1 Industrial facilities 0.35 - 0.55 0.7 - 0.9 0.95 - 1 1-1 1-1 Informal residential 0.9 - 1 1-1 1-1 1-1 1-1 Utility facilities 0.35 - 0.55 0.7 - 0.9 0.95 - 1 1-1 1-1 Table 29 – Probability of structural damage to buildings (showing range). Building category <1,000m2 1,000m2 - 10,000m2 - 30,000m2 - >100,000m2 2 2 10,000m 30,000m 100,000m2 Educational facilities 0.15 - 0.35 0.5 - 0.7 0.75 - 0.95 0.95 - 1 1-1 Formal residential 0.15 - 0.35 0.5 - 0.7 0.75 - 0.95 0.95 - 1 1-1 Government facilities 0.15 - 0.35 0.5 - 0.7 0.75 - 0.95 0.95 - 1 1-1 Healthcare facilities 0.15 - 0.35 0.5 - 0.7 0.75 - 0.95 0.95 - 1 1-1 Industrial facilities 0.15 - 0.35 0.5 - 0.7 0.75 - 0.95 0.95 - 1 1-1 Informal residential 0.7 - 0.9 0.8 - 1 0.9 - 1 1-1 1-1 Utility facilities 0.15 - 0.35 0.5 - 0.7 0.75 - 0.95 0.95 - 1 1-1 5.3.3.2 Vulnerability/fragility of infrastructure Pitilakis et al. (2011) present a methodology for assessing the fragility of roads of different construction based on expert judgement pooled from 47 academic, government and industry experts from 17 countries around the world. Some physical vulnerability relationships originate from the field of seismic risk assessment; Giovinazzi and King (2009) present a methodology for characterising the performance of ‘geographically distributed lifelines’ (transport networks, electric power systems and potable water systems) with respect to ground displacement due to landslides, liquefaction and fault rupture of seismic origin. Fragility functions to describe the fragility of roads are based on those developed by Pitilakis et al. (2011) for low and high-speed roads. For this project unpaved roads are assigned the fragility of ‘low-speed roads’ and paved roads are assigned the fragility of ‘high-speed roads’. Table 30 to Table 32 present the simple fragility functions defined to characterise roads in Freetown. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 82 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 30 – Probability of cosmetic damage or higher to roads (showing range). Road typology <1,000m2 1,000m2 - 10,000m2 - 30,000m2 - >100,000m2 2 2 10,000m 30,000m 100,000m2 Paved 0.5 - 0.7 0.75 - 0.95 0.85 - 1 0.95 - 1 1-1 Unpaved 0.75 - 0.95 0.85 - 1 0.95 - 1 1-1 1-1 Table 31 – Probability of functional damage or higher to roads (showing range). Road typology <1,000m2 1,000m2 - 10,000m2 - 30,000m2 - >100,000m2 2 2 10,000m 30,000m 100,000m2 Paved 0.35 - 0.55 0.6 - 0.8 0.75 - 0.95 0.9 - 1 1-1 Unpaved 0.45 - 0.65 0.7 - 0.9 0.8 - 1 0.95 - 1 1-1 Table 32 – Probability of structural damage to roads (showing range). Road typology <1,000m2 1,000m2 - 10,000m2 - 30,000m2 - >100,000m2 10,000m2 30,000m2 100,000m2 Paved 0.2 - 0.4 0.5 - 0.7 0.65 - 0.85 0.85 - 1 1-1 Unpaved 0.4 - 0.6 0.65 - 0.85 0.8 - 1 0.95 - 1 1-1 5.3.3.3 Loss ratios A loss ratio is used to estimate what proportion of an assets value is lost as a result of a particular damage state occurring. The loss ratios shown in Table 33 are used for the buildings and infrastructure. These loss ratios are derived from review of the various examples presented by Lee and Jones (2004). It should be noted that although the same loss ratio is applied for all assets, the fragility of assets is differentiated by type (Section 5.3.3.1 and 5.3.3.2) and hence the vulnerability of assets to loss from landslide damage is differentiated by asset type. Table 33 – Loss ratios for general building stock, critical facilities, roads and railways. Damage state Loss ratio Cosmetic 0.2 Functional 0.5 Structural 1 5.3.3.4 Vulnerability of persons indoors The vulnerability of occupants within a building which is damaged during a landslide is directly correlated with the level of structural damage. Buildings which are more likely to collapse when affected by a landslide therefore leave their occupants more vulnerable to death or injury (Finlay, Mostyn and Fell, 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 83 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 1999). Du, Nadim and Lacasse (2013) expand upon this concept by noting that not only does structural typology control physical vulnerability, it also influences the vulnerability of the persons within. This is because different construction materials produce rubble with different characteristics (e.g. weight and cavity size). Table 34 to Table 36 present the vulnerability functions which describe the probability of death for a single occupant within different building types subject to different damage states caused by landslides of varying magnitudes. Table 34 – Probability of death for an occupant in a building given cosmetic damage (showing range). Building category <1,000m2 1,000m2 - 10,000m2 - 30,000m2 - >100,000m2 2 2 2 10,000m 30,000m 100,000m Educational facilities 0-0 0-0 0-0 0-0 0 - 0.05 Formal residential 0-0 0-0 0-0 0-0 0 - 0.05 Government facilities 0-0 0-0 0-0 0-0 0 - 0.05 Healthcare facilities 0-0 0-0 0-0 0-0 0 - 0.05 Industrial facilities 0-0 0-0 0-0 0-0 0 - 0.05 Informal residential 0-0 0-0 0-0 0 - 0.05 0 - 0.1 Utility facilities 0-0 0-0 0-0 0-0 0 - 0.05 Table 35 – Probability of death for an occupant in a building given functional damage (showing range). Building category <1,000m2 1,000m2 - 10,000m2 - 30,000m2 - >100,000m2 2 2 2 10,000m 30,000m 100,000m Educational facilities 0 - 0.05 0 - 0.05 0.05 - 0.1 0.05 - 0.15 0.1 - 0.2 Formal residential 0 - 0.05 0 - 0.05 0.05 - 0.1 0.05 - 0.15 0.1 - 0.2 Government facilities 0 - 0.05 0 - 0.05 0.05 - 0.1 0.05 - 0.15 0.1 - 0.2 Healthcare facilities 0 - 0.05 0 - 0.05 0.05 - 0.1 0.05 - 0.15 0.1 - 0.2 Industrial facilities 0 - 0.05 0 - 0.05 0.05 - 0.1 0.05 - 0.15 0.1 - 0.2 Informal residential 0.05 - 0.1 0.1 - 0.15 0.15 - 0.2 0.2 - 0.25 0.25 - 0.3 Utility facilities 0 - 0.05 0 - 0.05 0.05 - 0.1 0.05 - 0.15 0.1 - 0.2 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 84 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 36 – Probability of death for an occupant in a building given structural damage (showing range). Building category <1,000m2 1,000m2 - 10,000m2 - 30,000m2 - >100,000m2 2 2 2 10,000m 30,000m 100,000m Educational facilities 0.35 - 0.55 0.4 - 0.6 0.45 - 0.65 0.5 - 0.7 0.55 - 0.75 Formal residential 0.35 - 0.55 0.4 - 0.6 0.45 - 0.65 0.5 - 0.7 0.55 - 0.75 Government facilities 0.35 - 0.55 0.4 - 0.6 0.45 - 0.65 0.5 - 0.7 0.55 - 0.75 Healthcare facilities 0.35 - 0.55 0.4 - 0.6 0.45 - 0.65 0.5 - 0.7 0.55 - 0.75 Industrial facilities 0.35 - 0.55 0.4 - 0.6 0.45 - 0.65 0.5 - 0.7 0.55 - 0.75 Informal residential 0.55 - 0.75 0.6 - 0.8 0.65 - 0.85 0.7 - 0.9 0.75 - 0.95 Utility facilities 0.35 - 0.55 0.4 - 0.6 0.45 - 0.65 0.5 - 0.7 0.55 - 0.75 In Regent, where the majority of the fatalities associated with the Regent-Lumley Disaster might be attributed to the landslide (rather than the downstream flooding effects), approximately 950 people were reportedly in buildings which suffered structural damage (data collected by house-to-house surveys to inform the DaLA, World Bank, 2017). Of these 950 people, approximately 400 were recorded as fatalities, corresponding to a high-level estimate of the probability of death given being in a building which was destroyed of ~0.4. This estimate does not (and cannot) include buildings where all of the occupants were killed as a result of the landslide (and hence could not be reported for in the house-to-house survey results). The Regent-Lumley Disaster landslide rupture area represents a Magnitude 4 landslide (30,000m2 – 100,000m2) by the classification of this project. The corresponding probability of death for occupants in a building subject to structural damage is modelled as 0.5 – 0.7 for planned buildings (Table 36), which gives reasonable agreement with the observations made during the DaLA (~0.4, which is likely to be an underestimate due to under-reporting, World Bank, 2017). 5.3.3.5 Vulnerability of persons outdoors It is widely agreed throughout the literature that persons affected by a landslide, whilst not protected by a structure are more vulnerable to death or serious injury, than those afforded protection by a structure. Finlay, Mostyn and Fell (1999) further note that the probability of death or serious injury for persons affected by a landslide is directly linked to if the individual is buried by landslide debris or not. Table 37 summarises some values from the literature for the vulnerability of persons affected by a landslide whilst not within a structure. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 85 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 37 – Example values describing the vulnerability of persons outdoors affected by different types of landslides. Reference Situation Vulnerability function Finlay, Mostyn Person in open space struck by a rockfall 0.5 and Fell (1999) Person in open space buried by debris 1.0 Person in open space not buried by debris 0.1 DNV Technica Person in open space affected by a 50m3 landslide 0.03 (1996); Halcrow Person in open space affected by a 5000m3 landslide 0.48 Asia Partnership (1999). Bell and Glade Person in open space affected by a low magnitude debris 0.2 (2004) flow Person in open space affected by a med. magnitude debris 0.3 flow Person in open space affected by a high magnitude debris 0.5 flow Person in open space affected by a low magnitude rock fall 0.2 Person in open space affected by a med. magnitude rock 0.4 fall Person in open space affected by a high magnitude rock 0.5 fall Table 38 summarises the vulnerability functions used to describe the probability of death for a person outdoors affected by landslides of varying sizes. Table 38 – Probability of death for a person outdoors affected by landslides of different sizes (showing range). <1,000m2 1,000m2 - 10,000m2 - 30,000m2 - >100,000m2 2 2 2 10,000m 30,000m 100,000m Person 0.5 – 0.7 0.6 – 0.8 0.7 – 0.9 0.8 – 1 0.9 – 1 outdoors 5.3.4 Landslide risk assessment Landslide risk is expressed as the product of the probability of hazard occurrence (e.g. a damaging landslide event) and its adverse consequences (Lee and Jones, 2004). This section of the report provides a step by step description of how the various aspects of the landslide risk calculation are combined to estimate the average annualised landslide risk to Freetown. All estimates of landslide risk are presented as a range, whereby the lower estimate relates to the lower estimate of the annual frequency of landslides of different sizes, and the upper estimate relates to the upper estimate of the annual 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 86 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results frequency of landslides of different sizes. The derivation of this range is discussed in Section 5.3.2. 5.3.4.1 Landslide risk to buildings Landslide risk to buildings is calculated on a building-by-building basis. The following process describes the approach to calculating landslide risk to a single building. This process is undertaken simultaneously for all buildings within the buildings exposure dataset (these results are then aggregated for presentation):  Determine the annual frequency at which the 30m pixel in which the building is located is affected by landslides of different magnitudes. This is done for both the lower and upper estimate of annual frequency associated with each landslide magnitude class;  Calculate the probability that the building is affected by landslides of different sizes. If the landslide is Magnitude 2 – 5, this value is 1 (because the landslide size is greater than the pixel). If the landslide is Magnitude 1, this is approximately equal to the building area divided by the pixel area (~900m2)5.  Multiply the probability that the building is affected by the probability of the building suffering a given damage state resulting from landslides of different sizes (as determined by the fragility assessment, Section 5.3.3.1).  Multiply the annual frequency at which the 30m pixel in which the building is located is affected by landslides of each magnitude class, the probability that the building is affected and suffers a given damage state resulting from landslides of different sizes by the loss ratio associated with that damage state and the rebuild value of the building. Sum for all possible damage states and landslide magnitudes to give the AAL (USD). 5.3.4.2 Landslide risk to roads Landslide risk to roads is calculated on a pixel-wise basis at a spatial resolution of 30m. The following process describes the approach to calculating landslide risk to a section of road within a given 30m pixel. This process is undertaken simultaneously for all roads within the roads exposure dataset:  Determine the annual frequency at which the 30m pixel in which the road section is located is affected by landslides of different magnitudes. This is done for both the lower and upper estimate of annual frequency associated with each landslide magnitude class;  Calculate the probability that the road section is affected by landslides of different sizes. If the landslide is Magnitude 2 – 5, this value is 1 (because the landslide size is greater than the pixel). If the landslide is Magnitude 1, this is 5 When estimating the AAL number of buildings affected, the probability that a building within a given pixel will be affected is equal to 1 regardless of landslide size. This is to partially account for the indirect effects of a landslide occurring nearby a given building. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 87 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results approximately equal to the road area divided by the pixel area (~900m2)6. It is assumed that all roads have a width of 9m.  Multiply the probability that the road is affected by the probability of the road suffering a given damage state resulting from landslides of different sizes (as determined by the fragility assessment, Section 5.3.3.2).  Multiply the annual frequency at which the 30m pixel in which the road is located is affected by landslides of each magnitude class, the probability that the road is affected and suffers a given damage state resulting from landslides of different sizes by the loss ratio associated with that damage state, the rebuild value of the road per kilometre and the length of road within the pixel. Sum for all possible damage states, landslide magnitudes and for both paved and unpaved roads to give the AAL (USD). 5.3.4.3 Landslide risk to population Landslide risk to the indoor population is calculated on a building-by-building basis. The following process describes the approach to calculating landslide risk to the population within a single building. This process is undertaken simultaneously for all buildings within the buildings exposure dataset (these results are then aggregated for presentation):  Determine the annual frequency at which the 30m pixel in which the building is located is affected by landslides of different magnitudes. This is done for both the lower and upper estimate of annual frequency associated with each landslide magnitude class;  Calculate the probability that the building is affected by landslides of different sizes. If the landslide is Magnitude 2 – 5, this value is 1 (because the landslide size is greater than the pixel). If the landslide is Magnitude 1, this is approximately equal to the building area divided by the pixel area (~900m2).  Multiply the probability that the building is affected by the probability of the building suffering a given damage state resulting from landslides of different sizes (as determined by the fragility assessment, Section 5.3.3.1).  Determine the population of the building by calculating the indoor population for the pixel and distributing evenly amongst all buildings within the pixel.  Multiply the probability of the building being affected by landslides of different sizes and suffering a given damage state by the probability of death for a single occupant. Sum the probability of death for one occupant resulting from cosmetic damage, functional damage and structural damage to determine the probability of death for one occupant in a building affected by landslides of different sizes. Multiply this value by the population of the building to estimate the AAL number of fatalities for the building. 6 When estimating the AAL kilometre length of road affected, the probability that a road section within a given pixel will be affected is equal to 1 regardless of landslide size. This is to partially account for the indirect effects of a landslide occurring nearby a given road section. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 88 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Landslide risk to the outdoor population is determined on a pixel-wise basis at a spatial resolution of 30m. The following process describes the approach to calculating landslide risk to the outdoor population.  Determine the annual frequency at which the 30m pixel is affected by landslides of different magnitudes. This is done for both the lower and upper estimate of annual frequency associated with each landslide magnitude class;  Calculate the probability that a single person outdoors is affected by landslides of different sizes. If the landslide is Magnitude 2 – 5, this value is 1 (because the landslide size is greater than the pixel). If the landslide is Magnitude 1, this is approximately equal to an assumed person-area (5m2) divided by the pixel area (~900m2).  Multiply the probability that the person is affected by the probability of death given being affected by landslides of different sizes. Multiply by the corresponding annual frequency of landslides of different sizes to determine the AAL number of fatalities for the pixel. Sum the AAL number of fatalities from indoor and outdoor scenarios to estimate the AAL number of fatalities. Note that this is effectively completed twice, once for a daytime scenario and once for a night-time scenario, each with different distributions of people indoors and outdoors (Section 3.4). 5.3.4.4 Estimation of future landslide risk Landslide risk in 2050 is estimated by factoring up the current-day landslide risk estimates in proportion to the increase in estimated flood risk, as determined by the method described in Section 5.2.7. It should be noted that this provides an extremely high-level estimate of future landslide risk only and is subject to the limitations set out in Section 5.2.7. 5.4 Quantitative coastal erosion hazard and risk The 2050 scenario-based estimates of coastal erosion hazard and risk have been based on the interpretation of historical imagery and mapping to estimate a recession rate. An example of the coastline mapping from this variety of data in GIS is shown in Figure 28. This rate is then projected inland and losses are estimated based on the current location of buildings. 5.4.1 Coastline data sources Google Earth Pro satellite images and historical maps of Freetown were used to analyse cliff recession. In total, three Google Earth images were interpreted, taken in March 2017, January 2013 and February 2006. The most recent image (March 2017) was used at the present-day coastline, and this acted as a fixed reference to measure the historical information against. Satellite images taken prior to 2006 could not be used because the resolution of the images is too coarse to resolve the coastline for accurate, peninsular-wide mapping of the coastline. Four historical maps were also used, which show 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 89 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results historical coastlines in 1959, 1966, 1969 and 1971. Each of these maps only covers a small area of the Freetown peninsula, so they do not provide a continuous source of coastline data like the Google Earth Pro satellite imagery. Other historical maps that do provide a continuous source of data were not at a suitable scale to be used in this study. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 90 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 28 – Two satellite images taken in 2017 (top) and 2006 (middle), and a 1969 historical map (bottom) of north east Freetown. The green line represents the 2017 coastline, the orange line represents the 2006 coastline, and the blue line shows the 1969 coastline. Satellite images © DigitalGlobe and NES/Airbus. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 91 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 5.4.2 Coastline analysis The coastlines in the satellite images and historical maps were first digitised using GIS. The present-day coastline was also converted to points at 50m intervals, which would be the interval at which the recession rates would be calculated. A perpendicular line was then extrapolated seawards from each point along the current coastline to find where this line intersects the historical coastlines (Figure 29). The distance between the point at the current coastline and the intersection coordinate at the historical coastline could then be calculated. Rate of cliff recession was then obtained by dividing the distance by the time difference (i.e. 2017 minus the data source date) to get an estimated rate at 50m intervals in metres per year. Figure 29 – Overview of the technical approach used to determine cliff recession rate at 50m intervals around the Freetown peninsula. To calculate an upper bound and a lower bound rate of recession, all rates were plotted against chainage around the coastline of Freetown and a maximum and minimum rate of recession were interpreted at 1 km intervals (Figure 30). It was then possible to take the lower bound and upper bound recession rates and create two predicted 2050 coastlines by multiplying the rates by 33 years (the time difference between 2050 and 2017). 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 92 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 30 – Rate of coastal-recession relative to chainage around the Freetown peninsula (example only). 5.4.3 2050 scenario loss assessment It is assumed that all buildings currently seaward of the 2050 projected coastline are lost to coastal recession. Estimates are determined for both the lower and upper estimate. 5.4.4 Assumptions and limitations The following assumptions and limitations apply to this method of estimating 2050 scenario cliff retreat:  Erosion is taking place perpendicular to the present-day coastline.  The coastline along beaches is defined by the boundary between sand and vegetation.  The coastline around rocky outcrops is defined by the high tide water mark, where visible.  The coastline along mangrove areas is defined by the mangrove and open water boundary, and not the mangrove and land boundary.  The coastline in Cockle Bay is defined by the road bridge from Murray Town to Aberdeen. The mangrove area that lies between Murray Town and Aberdeen has not been considered in this cliff recession study because the shorelines are not exposed to wave action and, therefore, recession through wave-action alone is likely to be negligible. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 93 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results  Piers/jetties/other man-made protrusions have been ignored when digitising the coastlines.  Only the Google Earth Pro satellite imagery provides continuous coastline information for the whole of the Freetown peninsular. Therefore, for large stretches of coastline, the recession rates are based on the 2006 and 2013 coastlines only. This means that the recession rates are calculated from recent changes to the coastline and do not represent a longer-term trend.  Areas where the coastline may be migrating seawards through mostly man- made developments, i.e. areas that have a negative recession rate, have not been considered and recession rates have been set to zero.  The Google Earth Pro satellite imagery appears to be spatially accurate to ~2m. This restricts the accuracy of the recession rates, and will have a greater effect on rates calculated from the most recent sources of data.  Newly built structures, such as docks/ports, will slow down the rate of recession markedly. The predicted 2050 coastline will be affected by these. 5.5 Quantitative sea-level rise hazard and risk Table 39 summarises present-day water levels around the Freetown peninsula (Environment Protection Agency, 2015). These levels will be used as the base levels for all sea-level rise estimations. Table 39 – Present-day water levels around the Freetown peninsula. Level Elevation relative to chart datum HAT (Highest Astronomical Tide) +3.50m MHWS (Mean High Water Springs) +3.00m MHWN (Mean High Water Neaps) +2.30m MSL (Mean Sea Level) +1.77m MLWN (Mean Low Water Neaps) +1.00m MLWS (Mean Low Water Springs) +0.40m LAT (Lowest Astronomical Tide) +0.00m Sea-level rise projections are based on the Atmosphere-Ocean General Circulation Models (AOGCM) as part of the Intergovernmental Panel on Climate Change (IPCC) Fifth assessment report (AR5). The AOGCM is sampled at the required location using the nearest grid point (resolution of 1°), the relative sea-level rise factor (DZOSloc) is then determined. The local sea-level rise (SLR) is then calculated using the global mean sea-level rise for the future scenario and time horizon of interest. The values of SLR are relative to the 1995 baseline. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 94 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results , , , The 2050 scenario land loss maps are then generated which show areas which may be lost to sea-level rise for 2050 as informed by the analysis. The location of exposed assets within the potential land-loss areas will be identified. The response of these assets within the land-loss areas (i.e. their vulnerability) will be total loss. 5.5.1 Assumptions and limitations  The primary limitations of the sea-level rise projections are the use of global climate models which use a relatively coarse resolution. Therefore, any specific localised effects may not be captured.  It is also widely accepted that using globally available digital elevation models (such as SRTM30, used for this project) may underestimate sea-level rise predictions (van de Sande et al., 2012; Kulp and Strauss, 2016). For these reasons, a range of estimates of sea-level rise will be generated.  It will be assumed that sea level rise occurs at a sufficiently slow rate that no persons will be killed by cliff failures.  The total value of those assets within the land loss areas will represent the 2050 scenario risk. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 95 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 96 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 6 DRR/DRM Recommendation Methodology 6.1 Introduction The DRR/DRM measures have been informed by the specific geography, terrain, climate, and exposure in each city, in addition to discussion with the stakeholders, field observations, expert judgment and research, and by the qualitative and quantitative natural hazard and risk maps that have been developed by this study specifically for each city. 6.2 Methodology Firstly, a broad range of DRR/DRM measures were considered, within the context of the priorities of the Sendai Framework for Disaster Risk Reduction (United Nations, 2015), and within the context of the specific cross-cutting issues faced by each of the cities, such as the lack of solid waste management and sand mining. The broad range of DRR/DRM measures were then considered on a catchment- by-catchment basis for each different hazard in each different city. Table 40 – Table 42 show the range of measures considered in each city on a catchment-by- catchment basis. Figure 31 to Figure 34 illustrate the catchment analysis of the DRR/DRM measures. The aim of this approach was to capture how the nature of the hazard and risk changes depending on the location of, and location within, the catchments in each of the cities. This is particularly true for Freetown, because of the steep mountainous terrain and well-established natural river valleys that characterise the city’s landscape. As an example, Map FT-0117 on Page 104 shows catchments 1 to 13 defined for Freetown. The areas of the peninsular that are shaded a darker purple indicate the distribution of existing development in what have been defined as proposed hazard zones, informed by the quantitative hazard results from this project (refer to report Volumes 2, 3 and 4 for further details regarding the hazard zones). Identification of these hazard zones has been a critical first step in enabling specific DRR/DRM recommendations to be developed for each city. Volume 5 of this project contains all maps for all cities at the scale of A3 page size so they can be observed clearly and in detail. Additionally, the GIS data for the project is open access and freely available from the World Bank (refer to Section 1.3 for details). Details of each of the broader DRR/DRM measures and the specific catchment- by-catchment analysis for each city can be found in the latter sections of report Volumes 2, 3 and 4 of this project. A high-level qualitative cost-benefit analysis (CBA) of the measures in Table 40 – Table 42 has been carried out to shortlist which measures should be taken forward for the quantitative CBA. The quantitative CBA has informed the prioritized DRR/DRM recommendations and estimated budgets for each city. However, it is emphasised that not all DRR/DRM recommendations can, or should be, justified on cost-benefit analyses alone. The CBA methodology has been summarised in Section 7 of this report and further details of the CBA for each city can be found in the latter sections of report Volumes 2, 3, and 4 of this study. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 97 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 40 – Summary of the broad range of DRR/DRM measures considered for flood hazard in each city where applicable # Proposed Flood DRR/DRM Measures 1 Flood hazard and risk communication and engagement 2 Early warning systems 3 Reforestation of upper catchment areas 4 Revegetation of the natural channels 5 Community drainage implementation 6 Check dam (with instrumentation and monitoring) 7 Flood water storage ponds 8 Flood hazard signage 9 Land use plans, zoning and building regulations 10 Rooftop rainwater harvesting 11 Engineered green channels 12 Delta rehabilitation 13 Engineered concrete channels 14 Engineered concrete culverts 15 Drainage channel clearance 16 Multi-purpose disaster response shelters Table 41 – Summary of the broad range of DRR/DRM measures considered for landslide hazard # Proposed Landslide DRR/DRM Measures 1 Landslide hazard and risk communication and engagement 2 Early warning systems 3 Reforestation of upper catchment areas 4 Landslide-hazard signage 5 Community drainage implementation 6 Check dam (with instrumentation and monitoring) 7 Land use plans, zoning and building regulations (with geotechnical guidance) 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 98 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 42 – Summary of the broad range of DRR/DRM measures considered for coastal hazard in Freetown where applicable # Proposed Coastal DRR/DRM Measures7 1 Mangrove preservation 2 Sand mining prevention 3 Riprap 4 Rock armour revetment 5 Vertical seawall 6 Groynes 7 Detached breakwaters In this summary of the DRR/DRM methodology, we emphasize that it is essential the proposed DRR/DRM recommendations for each city are aligned with comprehensive urban planning and strategy for the city. It is also essential that these plans are co-ordinated within a broader Urban Resilience Framework and zoning regulations that take into consideration the specific characteristics and vulnerabilities of the location, the diverse livelihoods of the citizens and social and economic requirements of the city infrastructure. The aims of the DRR/DRM recommendations are to save lives, reduce the number of people impacted, and to reduce the direct losses and economic impact caused by damage and disruption to the built environment from natural hazards. The recommendations have been presented in terms of a holistic strategy at city scale. The holistic strategy takes into consideration the geographic setting and terrain of each of the cities with an emphasis placed on re-establishing green and environmental solutions throughout the city to manage the risk associated with natural hazards. We also emphasize that the DRR/DRM measures that have not been carried forward for quantitative CBA in this project can still be considered for implementation, for example, at a different scale or on a longer term. Not all DRR/DRM recommendations can, or should be, justified on cost-benefit analyses alone. Furthermore, similar ongoing DRR/DRM measures that are currently being carried out across the city by varied development partners or government actors on the ground should continue to be encouraged and supported. 7 Landsliding and Coastal DRR/DRM measures do not apply to Makeni and Bo cities because they are located within the interior flat-lying pains of Sierra Leone, hundreds of kilometres away from the coast with low topographic relief. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 99 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 31 – Schematic illustration of a mountain catchment and flood hazard DRR/DRM measures recommended for Freetown. Illustrations of measures adapted from the Sustainable Urban Drainage Manual (Woods-Ballard et al., 2007). 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 100 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 32 – Schematic illustration of a mountain catchment and landslide hazard DRR/DRM measures recommended for Freetown. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 101 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 33 – Schematic illustration of a typical river catchment in Freetown and proposed coastal hazard DRR/DRM options appropriate for Freetown. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 102 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Figure 34 – Schematic illustration of a typical river catchment in Makeni/Bo and proposed flood hazard DRR/DRM options appropriate for Makeni and Bo. Illustrations of measures adapted from the Sustainable Urban Drainage Manual (Woods-Ballard et al., 2007). 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 103 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 104 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 105 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 7 Cost-Benefit Analysis Methodology 7.1 Introduction This section of the report provides a high-level summary of the cost-benefit analysis (CBA) methodology used for each of the hazards and associated proposed disaster risk reduction measures. Cost-benefit analysis is a useful tool to evaluate, in economic terms, the outcomes of proposed DRR measures. This approach allows an objective assessment and prioritization of each of the risk reduction measures for flooding, landslides, coastal erosion or sea-level rise in terms of their inherent advantages and disadvantages, by comparing the respective costs and benefits. Cost corresponds to the value of initial investment necessary for the implementation of a given risk reduction measure, and benefit is represented by the achieved reduction of economic and/or human losses as a result of the implementation of the proposed retrofit measure calculated over a specific time- span (e.g. 33 years). Benefits are expressed in monetary terms (e.g. total economic losses avoided in a period of 33 years) to be directly comparable with the associated costs. A CBA can be more challenging to apply when dealing with social, environmental and human risk. In this project, CBA analysis is used specifically to determine the direct economic benefits to compare different DRR/DRM recommendations. In addition to direct economic benefits, the reduction in human fatalities expected to be achieved through the implementation of the proposed DRR measures is also estimated. 7.2 Cost benefit analysis methodology – DRR/DRM direct costs and benefits In this project, the economic costs are the direct cost of establishing and implementing DRR/DRM measures as well as the estimated operating costs. The economic benefits are the avoided direct losses due to building or infrastructure damage. Losses to building contents are not included. Business interruption and other indirect economic impacts are also not included in the CBA. The CBA methodology proposed for this project is based on the recommendations of Mechler (2005). The following steps will be followed for this approach:  Step 1: Undertake quantitative risk assessment as discussed in section 4;  Step 2: Risk calculation output without DRR measures will be cost of damage in USD for buildings and infrastructure and number of fatalities;  Step 3: Risk-reduction calculation with DRR measures in place will be cost of damage in USD for buildings and infrastructure and number of casualties;  Step 4: Benefit is equal to the difference in risk calculation output without DRR measures (Step 2) minus the difference in risk calculation output with 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 106 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results DRR measure (Step 3). Benefit is determined by estimating the percentage risk reduction with DRR measures;  Step 5: Cost estimation for DRR measures including capital expenditure and operational expenditure over 33 years. Allow for discount rate of 6% over 33 years;  Step 6: Calculate Benefit Cost Ratios (BCR), Net Present Value (NPV) and Internal Rate of Return (IRR). The cost-benefit of the proposed DRR/DRM recommendations can be compared using three main CBA decision metrics. These metrics are described in detail by Nas (2016). The Net Present Value (NPV): 1 Where is the initial investment cost, is the discount rate, is the benefit stream that begins at year 1 ( =1) and is the project’s lifetime. The Benefit-Cost Ratio (B/C): Where PV is the present value and and are the steams of benefits and costs, respectively. The Internal Rate of Return (IRR): 0 1 Where is the IRR. Other variables are defined as above in the NPV equation. These CBA metrics are calculated over an extended time period but it is necessary to calculate the results in terms of present year cost and benefit values and to use an appropriate discount rate to adjust future costs and benefits to the current year (Nas, 2016). For the purposes of this study the present year is assumed to be 2017 (i.e. the year in which the natural hazard and risk calculations for this project were undertaken) and therefore costs and benefits are all presented in terms of US$ values in 2017. The choice of discount rate for cost-benefit is always uncertain and has a significant impact on the cost-benefit analysis results (Nas, 2017). A discount rate of 6% is used in accordance with recommendations by The World Bank and Institute for Health Metrics and Evaluation (2016) for analysis in low to middle-income countries. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 107 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 7.3 Cost benefit analysis methodology – including valuation of reduction in human fatalities In addition to direct economic benefits of avoiding damage to buildings and infrastructure, the reduction in human fatalities expected to be achieved through the implementation of the proposed DRR/DRM measures is also estimated. The number of human fatalities estimated to be avoided is reported and an economic value for a human life is used to allow the value of a statistical life to be used in the CBA. A country-specific value of statistical life (VSL) has been used in the cost-benefit analysis to quantify the economic benefit of recommended DRR/DRM options. The assignment of VSL is an uncertain and sensitive issue and therefore cost-benefit analyses have been undertaken with and without the analysis of the benefit of avoided potential loss of life. The equation of VSL (from The World Bank and Institute for Health Metrics and Evaluation, 2016) is as follows: , , Where , is the VSL for country in year , is the average base VSL estimate from the sample of willingness to pay studies in OECD countries (in 2011 U.S. dollars at purchasing power parity, PPP, rates), , is the GDP per capita for country in year (adjusted for price inflation and converted to 2011 U.S. dollars at PPP rates), is the average GDP per capita for the base sample of OECD countries, and is the income elasticity of the VSL. The World Bank and Institute for Health Metrics and Evaluation (2016) report indicates an average value of statistical life for OECD countries VSLOECD of US$3.83 million and a GDP per capita for OECD countries of US$37,000. Sierra Leone had a per capita GDP in 2015 of US$684. Using these input values, a VSL for Sierra Leone is calculated to be approximately US$70,800. The assignment of VSL is uncertain and therefore the Sierra Leone country-specific value of statistical life of US$100,000 has been used in the cost-benefit analysis to quantify the economic benefit of recommended DRR/DRM options. Assigning a value of statistical life is a sensitive issue and therefore cost-benefit analyses have been undertaken with and without the analysis of the benefit of avoided potential loss of life. Indirect economic and non-economic losses from natural disasters will of course also have many and varied wide-spread impacts on the city and its residents and these wider impacts should also inform decision makers. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 108 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 109 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 8 Summary of Hazard and Risk Results 8.1 Introduction This section of the report provides a high-level summary of the risk results for each of Freetown, Makeni and Bo. Detailed results relating to each city are provided in the accompanying city-specific reports. 8.2 Flood hazard and risk 8.2.1 Flood hazard and risk in Freetown In Freetown, areas of medium to high flood hazard are typically confined to inland watercourses flowing from areas of high relief towards the coast. Broad areas of high flood hazard are also located on the coastal flats on the western side of the peninsula and the wetland areas to the east. As a result, areas with a medium to high risk of flooding are located where watercourses flow through the urbanized areas on lower-lying terrain. There is a notable area with a high-risk flooding around Tower Hill, Kingtom and Congo Town. A number of watercourses flowing from the terrain above New England converge on these densely developed areas. Other notable areas with a high risk of flooding are located around the watercourses flowing through Kissy, Wellington, Samura Town and Fonima, including intersecting the main coastal roads such as Bai Bureh and Peninsular roads. Table 43 – Overview of quantitative flood risk assessment results for Freetown. Risk metric Average estimate AA number fatalities 9 AA number persons affected 3,011 AAL all buildings ($) $2,547,000 8.2.2 Flood hazard and risk in Makeni In Makeni, areas of medium to high flood hazard are associated with several networks of inland watercourses in the study area. There are two major watercourses extending through Makeni: one to the east of the city, and one flowing through the city centre. In addition, there are a number of tributaries and other minor watercourses associated with areas of medium to high flood hazard. Areas with a medium to high risk of flooding are located along the banks where the major watercourses intersect densely urbanised areas such as Robani and Makama to the southwest. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 110 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 44 – Overview of quantitative flood risk assessment results for Makeni. Risk metric Average estimate AA number fatalities <1 AA number persons affected 148 AAL all buildings ($) $47,000 8.2.3 Flood hazard and risk in Bo In Bo, areas of medium to high hazard area associated with an extensive network of inland watercourses in the study area. Major watercourses extend through the densely urbanised areas at the city centre as well as to the west in New York, New London and Chinese Farm. As such, areas with a medium to high risk of flooding are located along the banks of the watercourses in these locations. The low-lying terrain and shallow relief in the study area means the flood hazard has the potential to encroach land relatively far from watercourses. Table 45 – Overview of quantitative flood risk assessment results for Bo. Risk metric Average estimate AA number fatalities 2 AA number persons affected 607 AAL all buildings ($) $195,000 8.3 Landslide hazard and risk 8.3.1 Landslide hazard and risk in Freetown In Freetown, areas of medium to very high landslide hazard are located in areas of steeper terrain. Much of the steep terrain in characterised by a medium hazard and the areas of high hazard are associated with the encroachment of the built environment onto surrounding hillsides. Large areas of high landslide hazard are located on steeper terrain in Mount Aureol, New England, Kissy, Wongo Town, Wellington, Samura Town, Wilberforce and Hill Station with localised areas of very high landslide hazard in Mount Aureol. Some broad areas of high landslide hazard are located in the upland areas in the central region of the peninsula; these are associated with the steepest terrain rather than encroachment of the built environment. Areas of medium to very high landslide risk are associated with areas where the built environment encroaches onto steeper terrain, such as in Mount Aureol, New England, Kissy, Wilberforce and Hill Station. As a result, the areas of high landslide hazard in the upland areas have a low to medium landslide risk, due to the lack of development. Landslide risk is lower along the western extent of the peninsula due to the lower built environment density. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 111 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 46 gives an overview of the quantitative landslide risk assessment results for Freetown. Note that quantitative landslide risk assessment was undertaken for Freetown only. Table 46 – Overview of quantitative landslide risk assessment results for Freetown. Risk metric Lower Upper Average estimate estimate estimate AA number fatalities 1 20 11 AA number persons affected 25 255 140 AAL all buildings ($) $9,000 $701,000 $355,000 AA number all buildings affected 3 29 16 AAL roads ($) < $1,000 $9,000 $5,000 AA road length affected (km) 0.03 0.27 0.15 The quantitative landslide hazard analysis has highlighted that the areas which are probabilistically most frequently affected by mobile landslides are river channels and drainage lines within mountainous catchments, not necessarily just the steep upper slopes of mountainous catchments. This is because, many hillside landslides will travel down to the same drainage channel and therefore the hazard will be cumulatively higher in the channel. A landslide could have initiated anywhere upstream of Kamayama and still travelled down the Regent-Lumley channel, affecting Kamayama. Conversely, the houses on the steep slopes adjacent to Regent could only be affected by a landslide which initiated there, or upslope. 8.3.2 Landslide hazard and risk in Makeni In Makeni, areas of medium to high landslide hazard are associated with the areas of steeper terrain near the Mena and Wusum areas. The landslide hazard in the remaining part of the study area is very low with localised areas of low hazard. As the development of the available land does not typically extend onto steeper terrain, the landslide risk remains very low, with localised areas of low hazard, over all of the study area. 8.3.3 Landslide hazard and risk in Bo Due to the low relief in the Bo area, the landslide hazard is typically very low. Some localised areas of slightly higher relief, such as the Kandi Hill, Simbo Town and Chinese Farm areas, have low to medium landslide hazard. As the development of the available land does not typically extend onto steeper terrain, the landslide risk remains very low, with localised areas of low hazard, over all of the study area. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 112 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 8.4 Coastal erosion hazard and risk Large areas of medium to high coastal erosion hazard are located along the western extent of the Freetown Peninsula at Cockerill Bay and the beaches at Adonkia Village and Angola Town. Lumley Beach is considered particularly vulnerable to coastal erosion owing to it forming a narrow sandy spit. Areas of high coastal erosion hazard are also associated with the developed alluvial estuaries where the major watercourses flow into White Man’s Bay, Kroo Bay, Destruction Bay and Cline Bay. Table 47 – Overview of 2050 scenario coastal erosion risk results for Freetown. Risk metric Lower Upper Average estimate estimate estimate 2050 scenario buildings loss ($) $1,121,000 $261,711,000 $131,416,000 2050 scenario number buildings affected 243 2222 1233 8.5 Sea-level rise hazard and risk Areas of high to very high inundation hazard are associated with the low-lying coastal areas at Aberdeen and Lumley Beach at the north-eastern extent of Freetown as well as on the eastern extent of the peninsula at Orogu Village and Wellington. A large area of medium inundation hazard is located at the Sussex Beach area on the western extent of the peninsula. The inundation risk is relatively low in these areas due to these low-lying wetland areas being unsuitable for urban development. Areas of medium to high risk are located where the built environment encroaches the periphery of these areas. Developed coastal areas in the Kingtom area, particularly in White Man’s Bay and Kroo Bay, have a high to very high risk of inundation as these low-lying areas have been densely urbanised. In Freetown, SLR to 2050 using RCP8.5 is estimated to be +0.26m, which is slightly higher than the global mean sea-level rise calculated using RCP8.5 of +0.25m. This means that for a calm sea state at the highest astronomical tide, including allowance for SLR to 2050, the sea level would be approximately 2m above the current global mean sea level. This projection does not include allowance for storm surge or wave run-up, which could increase the level (although this allowance would usually be added to the mean high-water spring level, rather than the highest astronomical tide). For this reason, and considering that the use of global elevation models typically underestimates the effects of sea- level rise, particularly in urban settings (Section 5.5), the 2050 scenario sea-level is estimated to be 3 – 4m above the current global mean sea level. This estimate includes allowance for SLR, highest astronomical tidal conditions and a nominal allowance for wave-action8. 8 Based on a high-level review of the tail ends of historical hurricane tracks off the coast of West Africa, we can conservatively assume a storm surge that relates to a category 1 on the Saffir- Simpson Hurricane Intensity Scale. Using this scale, this relates to a storm surge of 1.0m – 1.7m. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 113 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table 48 – Overview of 2050 scenario sea-level rise risk results for Freetown. Risk metric Lower Upper Average estimate estimate estimate 2050 scenario buildings loss ($) $9,925,000 $83,731,000 $46,828,000 2050 scenario number buildings affected 1481 2280 1881 A conservative range of 1m – 2m can therefore be assumed for potential storm surge as a high level estimate. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 114 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 115 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 9 References [1] Admiraal, J. (2011) Flood damage to port industry. Case study: vulnerability of the port of Rotterdam to climate change. Faculteit der Aard- en Levnswetenschappen, Vrije Universieit. [2] AllAfrica (2017) AllAfrica. Available online: http://allafrica.com/ [Accessed 28/02/2017]. [3] Africa Research Institute (2015) Flooding in Freetown: a failure of planning? Available online: http://www.africaresearchinstitute.org/newsite/blog/flooding-in- freetown-a-failure-of-planning/ [Accessed 02/03/2017]. [4] Arino, O., Ramos Perez, J. J., Kalogirou, V., Bontemps, S., Defourny, P. and Van Bogaert, E. (2012) Global land cover map for 2009, GlobCover 2009. European Space Agency & Université catholique de Louvain, PENGEA. [5] Arup-BGS (2017) National-Level Landslide Risk Profiles for Sub-Saharan Africa (Stage 2 Countries: Cabo Verde, Malawi, Mali and Mozambique): Final Report. Report produced for the World Bank/GFDRR, Issue 1, 26/05/2017. [6] Australian Geomechanics Society (2000) Landslide risk management concepts and guidelines. Australian Geomechanics, 35, 49-52. [7] Awoko (2017) Awoko. Available online: http://awoko.org/ [Accessed 28/02/2017]. [8] Bayliss, A. (1999) Flood Estimation Handbook: Procedures for flood frequency estimation, Volume 5: Catchment descriptors. Wallingford: Centre for Ecology and Hydrology. [9] BBC (2017) Sierra Leone profile – Timeline. Available online: http://www.bbc.co.uk/news/world-africa-14094419 [accessed 26/10/2017]. [10] Bell, R. and Glade, T. (2004) Quantitative risk analysis for landslides – Examples from Bildudalur, NW-Iceland. Natural Hazards and Earth Systems Sciences, 4, 117-131. [11] Büchele, B., Kreibich, H., Kron, A., Thieken, A., Ihringer, J., Oberle, P., Merz, B. and Nestmann, F. (2006). Flood-risk mapping: contributions towards an enhanced assessment of extreme events and associated risks. Natural Hazards and Earth System Science, 6, 4, pp.485-503. [12] Cardinali, M., Reichenbach, P., Guzzetti, F., Ardizzone, F., Antonini, G., Galli, M., Cacciano, M., Castellani, M. and Salvati, P. (2002) A geomorphological approach to the estimation of landslide hazards and risks in Umbria, Central Italy. Natural Hazards and Earth System Sciences, 2, 57-72. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 116 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results [13] Chalkias, C., Ferentinou, M. and Polykretis, C. (2014) GIS-Based Landslide Susceptibility Mapping on the Peloponnese Peninsula, Greece. Geosciences, 4, 176-190. [14] Commonwealth Governance (2017) Local Government of Sierra Leone. Available online: http://www.commonwealthgovernance.org/countries/africa/sierra_le one/local-government/ [accessed 01/11/2017]. [15] Corominas, J., van Westen, C., Frattini, P., Cascini, L., Malet, J-P., Fotopoulou, S., Catani, F., Van Den Eeckhaut, M., Mavrouli, O., Agliardi, F., Pitilakis, K., Winter, M. G., Pastor, M., Ferlisi, S., Tofani, V., Hervas, J. and Smith, J. T. (2014) Recommendations for the quantitative analysis of landslide risk. Bulletin of Engineering Geology and the Environment, 73, 209-263. [16] Dartmouth Flood Observatory (DFO) (2017) Global Active Archive of Large Flood Events. Available online: http://floodobservatory.colorado.edu/Archives/index.html [Accessed 28/02/2017]. [17] Davison, A.C. and Smith, R.L. (1990). Models for Exceedances over High Thresholds. Journal of the Royal Statistical Society: Series B, 52, 3, pp.393–442. [18] DNV Technica (1996) Quantitative landslip risk assessment of Pre-GCO man-made slopes and retaining walls. DNV Technica Report to Geotechnical Engineering Office, Hong Kong Government. [19] Du, J., Nadim, F. and Lacasse, S. (2013) Quantitative vulnerability estimation for individual landslides. Proceedings of the 18th International Conference on Soil Mechanics and Geotechnical Engineering, Paris 2013, 2181-2184. [20] EMDAT (2009) The OFDA/CRED International Disaster Database. Available online: http://www.emdat.be/ [Accessed 02/03/2017]. [21] EMDAT (2016) Sierra Leone Country Profile. Available online: http://www.emdat.be/country_profile/index.html [Accessed 02/03/2017]. [22] Environment Protection Agency (2015) Sierra Leone State of the Marine Environment Report. Freetown, Sierra Leone. [23] FCC [Freetown City Council] (2016) Freetown City Development Plan 2016 – 2018. Available online: http://www.slurc.org/uploads/1/0/9/7/109761391/freetown_city_deve lopment_plan.pdf [Accessed 26/02/2018]. [24] Finlay, P. J., Mostyn, G. R. and Fell, R. (1999) Landslides: Prediction of Travel Distance and Guidelines for Vulnerability of Persons. Proceedings of the 8th Australia/New Zealand Conference on 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 117 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Geomechanics, Hobart. Australian Geomechanics Society, 1, 105- 113. [25] Frazer-Williams, R. (2014) Environmental Assessment and Evaluation of Natural Disaster Risk and Mitigation in Freetown. Ministry of Lands, Country Planning and the Environment (MLCPE) and Freetown City Council (FCC), European Union. [26] Giovinazzi, S. and King, A. (2009) Seismic performance of geographically distributed lifelines: an overview. NZSEE 2009 New Zealand Society for Earthquake Engineering Annual Conference, Christchurch, New Zealand. Available online: ir.canterbury.ac.nz/handle/10092/3667 [Accessed 03/04/2017]. [27] Halcrow Asia Partnership (1999) Stage 2 Detailed Site-Specific QRA Report. Technical report to Geotechnical Engineering Office, Hong Kong Government. [28] Hallegatte, S. (2015) The Indirect Cost of Natural Disasters and an Economic Definition of Macroeconomic Resilience. Impact Appraisal for Sovereign Disaster Risk Financing and Insurance Project: Phase 1, Public Finance and Macroeconomics, Paper 3. Report produced for the DFID, World Bank Group and GFDRR. [29] Heinimann, H. R. (1999) Risikoanalyse bei gravitativen Naturgefahren- Fallbeispiele & Daten. Umwelt-Materialen, 107/I, Bern. [30] Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones P. G. and Jarvis, A. (2005) Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology, 25, 15, pp.1965- 1978. [31] Institute for Structural Engineers (2013) Manual for the systematic risk assessment of high-risk structures against disproportionate collapse. IStructE Ltd. [32] Jaiswal, P., van Westen, C. J. and Jetten, V. (2011) Quantitative estimation of landslide risk from rapid debris slides on natural slopes in the Nilgiri hills, India. Natural Hazards and Earth System Sciences, 11, 1723 – 1743. [33] Jakeman, A. J., Littlewood, I. G., and Whitehead, P. G. (1990) Computation of the instantaneous unit hydrograph and identifiable component flows with application to two small upland catchments. Journal of Hydrology, 117, 1, pp.275-300. [34] Jonkman, S. (2002) Loss of Life Models for Sea and River Floods. In: K. Wu, ed., Flood Defences. New York: Science Press. [35] Kottek, M., Grieser, J., Beck, C., Rudolf, B. and Rubel, F. (2006) World map of the Köppen-Geiger climate classification updated, Meteorologische Zeitschrift, 15(3), pp.259-263. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 118 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results [36] Kreibich, H., Seifert, I., Merz, B. and Thieken, A. H. (2010) Development of FLEMOcs–a new model for the estimation of flood losses in the commercial sector. Hydrological Sciences Journal, 55, 8, pp.1302- 1314. [37] Kulp, S. and Strauss, B. H. (2016) Global DEM errors underpredict coastal vulnerability to sea-level rise and flooding. Frontiers in Earth Science, 4:36. doi: 10.3389/feart.2016.00036. [38] Lapworth, D. J., Carter, R. C., Pedley, S. and MacDonald, A. M. (2015) Threats to groundwater supplies from contamination in Sierra Leone, with special reference to Ebola care facilities. British Geological Survey Technical Report OR/15/009, Nottingham, UK. [39] Lee, E. M. and Jones, D. K. C. (2004) Landslide risk assessment. Thomas Telford Ltd, London. [40] LGA. (2004), http://www.sierra-leone.org/Laws/2004-1p.pdf [Accessed 31 May 2018] [41] Mansaray, L. R., Huang, J., Kamara, A. A. (2016) Mapping deforestation and urban expansion in Freetown, Sierra Leone, from pre- to post- war economic recovery. Environ. Monit. Assess, p188-470. [42] McSweeney, C., New, M., Lizcano, G. (2010) UNDP Climate Change Country Profiles: Sierra Leone. Available online: http://country- profiles.geog.ox.ac.uk/ [Accessed 12/06/2017]. [43] Mechler, R. (2005) Cost-benefit Analysis of Natural Disaster Risk Management in Developing Countries. Manual. Deutsche Gesellschaft fur Technische Zusammenarbeit (GTZ) GmbH. Commissioned by the Federal Ministry for Economic Cooperation and Development. [44] Merz, B., Kreibich, H., & Lall, U. (2013) Multi-variate flood damage assessment: a tree-based data-mining approach. Natural Hazards and Earth System Science, 13, 1, pp.53-64. [45] Messner, F., Penning-Roswell, E., Green, C., Volker, M., Tunstall, S. and van der Veen, A. (2007) Evaluating flood damages: guidance and recommendations on principles and methods. Report number T09- 06-01 Revision number 2_2_P44. Available online: http://www.floodsite.net/html/partner_area/project_docs/T09_06_01 _Flood_damage_guidelines_d9_1_v2_2_p44.pdf [Accessed 01 March 2017]. [46] MLCPE [Ministry of Lands, Country Planning and the Environment]/FCC [Freetown City Council] (2014) Sierra Leone Preparatory Components and Studies for the Freetown Development Plan “The Urban Planning Project”, Freetown Structure Plan 2013–2028. European Union. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 119 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results [47] Morel, S. W. (1979) The geology and mineral resources of Sierra Leone. Economic Geology, 74, pp1563-1576. [48] Nas, T.F. (2016). Cost-Benefit Analysis - Theory and Application. Second Edition, Lexington Books. [49] NSSG [National Strategic Situation Group] (2016) The National Flood Response Plan for Sierra Leone. NSSG. [50] Penning-Roswell, E., Viavattene, C., Pardoe, J., Chatterton, J., Parker, D. and Morris, J. (2010) The benefits of flood and coastal risk management. London: Middlesex University Press. [51] Pitilakis, K., Fotopoulou, S., Argyroudis, S., Pitilakis, D., Senetakis, K., Treulopoulous, K., Kakderi, K. and Riga, E. (2011) Deliverable 2.5: Physical vulnerability of elements at risk to landslides: Methodology for evaluation, fragility curves and damage states for buildings and lifelines. SafeLand: Living with landslide risk in Europe: Assessment, effects of global change, and risk management strategies. [52] Ravichandran, S. (2011) Factors of Persistent Poverty in Sierra Leone. E- International Relations Students. Available online: https://www.e- ir.info/2011/07/28/factors-of-persistent-poverty-in-sierra-leone/ [Accessed 25/09/2018]. [53] ReliefWeb (2015) Sierra Leone: Sections Affected by Flood in Western Area Urban. Available online: https://reliefweb.int/sites/reliefweb.int/files/resources/ocha_sle_flood _affected_september2015.pdf [Accessed 28/02/2017]. [54] Sivapalan, M. (2003) Prediction in ungauged basins: a grand challenge for theoretical hydrology. Hydrological Processes, 17, 15, pp.3163- 3170. [55] Statistics Sierra Leone (2014) Sierra Leone Integrated Household Survey 2011. Available online: https://www.statistics.sl/wp- content/uploads/2017/04/sierra_leone_integrated_household_survey_ 2011-1.pdf [Accessed 02/02/2017]. [56] Statistics Sierra Leone (2016) 2015 Population and Housing Census, Summary of Final Results, Available online: https://www.statistics.sl/wp-content/uploads/2017/01/final-results_- 2015_population_and_housing_census.pdf [Accessed 02/02/2017]. [57] Stephenson, A. G. (2002) evd: Extreme value distributions. R News, 2, 2, pp.31-32. [58] Su, M-D. Kang, J-L. and Chang, L-F. (2009) Industrial and commercial depth-damage function assessment. Wseas Transactions on Environment and Development, 2, 5, pp.199-208. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 120 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results [59] Thomas, M. F. (1974) Tropical Geomorphology: A Study of Weathering and Landform Development in Warm Climates. Focal Problems in Geography Series. Macmillan. 332 pp. [60] Thomas, M. F. (1983) Contemporary denudation systems and the effects of climatic change in the humid tropics – some problems from Sierra Leone. In: Studies in Quaternary Geomorphology (Ed. D. J. Briggs & R. S. Waters), pp. 195-214. Geo Books, Norwich. [61] Thomas, M.F. (1994) Geomorphology in the Tropics. John Wiley & Sons Ltd, Chichester. ISBN 0-471-93035-0. [62] Thomas, M. F. (1998) Landscape sensitivity in the humid tropics: a geomorphological appraisal. In: Maloney, B. K. (ed.) Human Activities and the Tropical Rainforest. Past, Present and Possible Future. Springer-Science Business Media, B.V. ISBN 978-90-481- 4952-0. [63] Thomas, M, F., Thorp M. B. (1996) The response of geomorphic systems to climate and hydrological change during the Late Glacial and early Holocene in the humid and sub-humid tropics. In: Branson, J., Brown, A. G. & Gregory, K. J. (eds), 1996, Global Continental Changes: the Context of Palaeohydrology, Geological Society Special Publication No. 115, pp 139-153. [64] UNDP (2012) Diagnostic analysis of climate change and disaster management in relation to the PRSP III in Sierra Leone. Available online: http://www.sl.undp.org/content/dam/sierraleone/docs/focusareadocs/ undp_sle_analysisclimatechangeDM.pdf [Accessed 02/03/2017]. [65] UNDP (2017) World Population Prospects 2017. Available online: https://esa.un.org/unpd/wpp/Download/Probabilistic/Population/ {Accessed 11/09/2017]. [66] van de Sande, B., Lansen, J. and Hoyng, C. (2012) Sensitivity of coastal flood risk assessments to digital elevation models. Water, 4, 568-579. [67] Wells, M. K. (1962) Structure and petrology of the Freetown Layered Basic Complex of Sierra Leone. Overseas Geology and Mineral Resources Supplement Series. Bulletin Supplement No 4. HMSO: London. [68] Wichmann, V. (2017) The Gravitational Process Path (GPP) model (v1.0) – a GIS-based simulation framework for gravitational processes. Geoscientific Model Development, Manuscript under review. [69] World Bank (2017) Sierra Leone: Rapid Damage and Loss Assessment of August 14th, 2017 Landslide and Floods in the Western Area (August 24 – September 8, 2017). World Bank Group. [70] World Vision (2017) Protecting the living, honouring the dead. World Vision/Cafod. Available online: 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 121 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results http://cdn.worldvision.org.uk/files/7714/9149/5615/WV_SierraLeone _Report_WEB.pdf [accessed 19/09/2017]. [71] Zobler, L. (1999) Global Soil Types, 1-Degree Grid (Zobler). Available online: http://www.daac.ornl.gov [Accessed 02/03/2017]. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page 122 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX Appendix A Building Replacement Costs in Sierra Leone The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results A1 Construction costs desk study In order to provide a range of estimated construction costs for Freetown, Makeni and Bo, the following should be considered:  Building types in each city;  Building materials used in each city; and  Cost of building materials ($ / m2). Prior to Mission #2, a desk study of construction costs in Sierra Leone was undertaken, which provided cost estimates that formed the basis of Arup’s in- country workshop discussions with the local resident stakeholders. The desk study identified two main reports which provided information on construction costs in Sierra Leone:  The ‘Housing Finance in Africa 2016 Yearbook’ by the Centre for Affordable Housing Finance in Africa (CAHF) and;  The ‘Freetown Construction Manual Ivor Leigh School Design Analysis’ by Amanda Rashid and Jonathan Buckland from London Metropolitan University. The findings from the desk study are summarised in Tables A1 and A2 below. Table A1 – Cost of construction for different building types and cost to buy. Cost to build / buy Building Cost (US $) Cost (US $ Source type / m2) Build 2 bed flat $ 15,000 $75 / m2 Housing Finance in 200 m2 Africa 2016 Yearbook 3 bed flat $ 25,000 $100 / m2 Housing Finance in 250 m2 Africa 2016 Yearbook School $ 21,500 $72 / m2 Freetown ~300 m2 Construction Manual Ivor Leigh School Design Analysis Buy House $ 50,000 $625 / m2 Housing Finance in (Cheapest newly built 80 m2 Africa 2016 house by a formal Yearbook developer or contractor) 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page A1 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table A2 – Cost of building materials. Building material Cost (US $) Source Sandcrete Block 0.4 Freetown Construction Manual Ivor Leigh School Design Ventilation Block 2 Analysis Cement (50 kg bag) 10 Housing Finance in Africa 2016 Yearbook Sand (1t delivery) 71 Freetown Construction Manual Ivor Leigh School Design Aggregate (30 kg bag) 4 Analysis Zinc Roofing (8’ x 3’) 8 R-bar Beams / Columns (40’/12 m of ½”) 11 Doors 71 Water Butt 187 The desk study did not provide much information on the building types or building materials used within the three cities. Missions #1 and #2 enabled first- hand visual observation in Freetown, Makeni and Bo. Photographs in Section A2 are from Mission #1 and show selected buildings within the cities. A1.1 Government Compensation Formula The Planning Unit of the Development Department within the Sierra Leonean Government possesses a formula for calculating housing compensation for properties within the defined Right-of-Way: Estimated Cost = (Floor Area in m2 x NS Factor x TR Factor x CI) NS Factor: Takes into account the nature of the structure, for: Sandcrete blocks NS factor = 1.04 Mud block NS factor = 1.0 ‘Pan body*’ NS factor = 1.0 * Note that ‘pan body’ is corrugated metal TR Factor: Takes into account type of roof, for: High quality corrugated iron zinc = 1.0 Medium quality corrugated iron zinc = 0.96 Appreciable quality corrugated iron zinc = 0.85 Unroofed = 0.5 CI: Common Index = Common rate for the type of structures based on existing construction cost The CI is based on 2010 construction cost estimates, shown in Table A3. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page A2 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results The compensation formula indicates the value which the Sierra Leonean Government assigns to each building type. The construction costs (Table A3) are of particular relevance as they cover a wide range of building materials and structure types. The construction costs provided in Table D3 are less than the $100/m2 value of an average three-bed flat in Sierra Leone, as stated in the ‘Housing Finance in Africa 2016 Yearbook’. The difference in values may be due to the six-year time interval between the costing or a different estimation of material and/or labour costs. Table A3 – Common Index values. 2010 construction cost estimates. Construction cost* Construction cost* Structure type (Le / m2) (US $ / m2) 120,000 16 ‘Pan body’ structure 150,000 20 Mud structure without plaster 170,000 23 Mud structure with plaster 440,000 59 Plastered sandcrete structure with tarpaulin roof 450,000-480,000 60-64 Sandcrete structure with corrugated iron sheet roof 500,000-550,000 67-73 Ordinary multi-storey building 550,000-680,000 73-91 Modern multi-storey building *Assumed exchange rate of 1 US $ to 7500 Sierra Leonean Leone (Le), rounded to nearest $. A2 Building Photographs The following photographs were taken by Arup during Mission #1 in March 2017, and provide an indication of the different building types within Sierra Leone. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page A3 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results A2.1 Freetown Figure A1 – Hill-side buildings in Freetown. The majority of buildings are multi-storeyed with tiled roofs. Figure A2 – Different building types in Freetown. The buildings range from ‘pan body’ structures in the foreground, to multi-storey blocks in the background. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page A4 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results A2.2 Makeni Figure A3 – Single-storey residential building in Makeni. The structure appears to be constructed with plastered brick, and a corrugated metal roof. Figure A4 – The majority of buildings in Makeni are single-storeyed structures. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page A5 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results A2.3 Bo Figure A5 – Commercial buildings in Bo. Figure A6 – Residential buildings in Bo, with a mixture of single and multi-storey buildings. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page A6 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results A3 Construction Costs from Workshops The desk study estimates of construction cost were used in the workshops during Mission #2 in June 2017. The findings provided benchmark figures to which workshop participants could suggest higher or lower estimates based on their knowledge. A workshop was run in each of the three cities and the findings are discussed in the following sections. A3.1 Freetown Workshop The Freetown workshop resulted in discussions on construction costs, as well as the cost of land. The cost of land within Freetown varies with location, and increases with proximity to:  Utilities;  Beach;  Markets;  City centre;  Roads; and  Building materials. Workshop participants suggested that the majority of buildings within Freetown are made from concrete blocks, with timber buildings being mostly historical. The estimated cost of different building types and plots of land from the workshop are summarised in Table A4. Table A4 – Cost estimates provided during the Freetown workshop. Building type Cost (US $) Large reinforced concrete house $ 60,000* Single storey 3 bed duplex $ 350,000* Minimum construction costs $ 1,000-2,000 Maximum construction costs single storey $ 100,000 Plot of land (750 sq. ft.) East: $ 10,000 +/- $ 2,000 West: $ 10,000 +/- $ 5,000 Hills: $ 4-5,000 * Cost estimates provided by participants during the workshop have not been rigorously verified. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page A7 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results A3.2 Makeni Workshop The Makeni workshop discussions led to material costs being estimated, as well as building costs. The main outcomes from the workshop were:  Buildings are mostly single-storeyed;  Buildings are made from wattle and daub, sandcrete, mud bricks, or concrete;  Many buildings are own-builds;  Location is important (see cost of land in Table A5); and  There were residential houses in the centre which have now been converted into commercial buildings. Table A5 – Cost estimates provided during the Makeni workshop. Building type Cost A mud brick $ 0.07* (Le 500) 3,000-5,000 bricks needed for 3 bedroom house Cost of bricks for 3-bed house – assume 4,000 bricks = $ 267* (Le 2,000,000) Mud brick 6-room house $ 17,333* (Le 130,000,000) Duplex $ 18,667* (Le 140,000,000) Concrete 4 bed house $ 40,000* (Le 300,000,000) Most expensive houses $ 200,000-400,000* (Le 1.5-3 billion) Cost of land Centre: $ 8,000* (Le 60,000,000) Outskirts: $ 1,333-2,000* (Le 10-15,000,000) *Assuming an exchange rate of 1 US $ to 7500 Sierra Leonean Leone (Le) A3.3 Bo Workshop The Bo workshop mostly focused on the types of houses within the city and did not look at material costs or land prices. The workshop participants determined that:  The most common structures in Bo are made from wattle and daub;  Few homes are made from mud brick, although there are some mud brick buildings with concrete; and  Informal developments are not very common in Bo. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page A8 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table A6 – Cost estimates provided during the Bo workshop. Building type Cost Wattle and daub house 2 rooms: $ 667* (Le 5,000,000) High end: $ 2,667* (Le 20,000,000) Concrete block 2-3 bed Normal concrete block house: house $ 13,333* (Le 100,000,000) Concrete block in swamp  more construction required $ 20,000* (Le 150,000,000) High quality concrete $ 26,667* (Le 200,000,000) 2 storey concrete 4-5 bed $ 80,000-133,333* (Le 0.6-1 billion) house *Assuming an exchange rate of 1 US $ to 7500 Sierra Leonean Leone (Le) A3.4 Construction Costs Summary The comparison of the desk study and workshop-estimated costs is not a simple process. Workshop responses did not include floor area, so direct comparisons of cost per square metre cannot be made. However, comparisons based on the number of bedrooms were possible. Table A7 compiles all of the estimated construction costs. The desk study provided an average construction cost ranging from $15,000 to $25,000 for two- to three-bed flats. The estimate for a two- to three-bed concrete house in Bo falls within a very similar range of $13,000-27,000. The comparable desk study and workshop estimate suggests that an average construction cost of $75-100/m2 is suitable for a concrete building within Sierra Leone. However, a more conservative range may be around $70-120/m2. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page A9 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table A7 – Cost of construction for different building types. Building type Cost Cost Freetown Makeni Bo 2 (US $) (US $ / m ) cost (US $) cost (US $) cost (US $) 2-bed flat $15,000 $75 / m2 200m2 3-bed flat $25,000 $100 / m2 250m2 School $21,500 $72 / m2 ~300m2 Large $60,000 reinforced concrete house Concrete 4-bed $40,000 house Concrete 2-3 $13,000- bed house $27,000 Mud brick 6- $17,000 room house Wattle & daub 2 rooms: house $1,000 High end: $3,000 Minimum $1,000- construction $2,000 costs Maximum $100,000 construction costs single storey * Estimates from desk study (rows 1-3) and workshops (row 4+). Note costs have been rounded to the nearest $1,000. Building types across the three cities varied. The workshop discussions, along with Missions #1 and #2, have shown that Freetown structures are predominantly made of concrete/sandcrete, and are often multi-storeyed. Makeni has mostly single-storeyed buildings, as shown in Figure A4, which are made from a range of materials, including wattle and daub, sandcrete, mud bricks and concrete. Bo, however, has few mud brick buildings, but instead has a large proportion of wattle and daub. Construction costs identified in the desk study are shown in Table A8. 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page A10 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results Table A8 – Construction costs from desk study in USD$ / m2. Construction cost Structure type (US $/m2) $16 / m2 ‘Pan body’ structure 2 $20 / m Mud structure without plaster 2 $23 / m Mud structure with plaster $59 / m2 Plastered sandcrete structure with tarpaulin roof 2 $60-64 / m Sandcrete structure with corrugated iron sheet roof $67-73 / m2 Ordinary multi-storey building 2 $73-91 / m Modern multi-storey building $75 / m2 2 bed flat $100 / m2 3 bed flat 2 $72 / m School Table A9 provides a range of approximate cost estimates, based on the desk study and workshop values. Table A9 – Approximate construction cost range in US $ / m2 Construction cost Building type City commonly (US $ / m2) present in $ 15-25 / m2 ‘Pan body’ structure F $ 20-40 / m2 Mud structure (assume includes wattle & daub M, B structures) $ 30-50 / m2 Mud brick structure M $ 60-80 / m2 Sandcrete structure F, M $ 70-120 / m2 Concrete structure F, M, B 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page A11 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX The World Bank Sierra Leone Multi-City Hazard Review and Risk Assessment Final Report (Volume 1 of 5): Technical Methodology and Summary of Results 20180927-DOC-05A_V1of5_Method | Issue 2 | 27 September 2018 Page A12 G:\250000\252746-00\60_OUTPUT\1_REPORTS\6_FINAL REPORT\20180927-DOC-05A_V1OF5_METHOD.DOCX