Lead: Margreth Keiler
To assess flood risk of settlements, knowledge on physical vulnerability of flood exposed buildings is crucial. Physical vulnerability is commonly defined as the potential damage expressed as a percentage of the asset’s economic value. It is a popular approach to link the potential damage with key characteristics of the flood, like flood depth or duration. However, case studies show, that this link to flood characteristics is rather weak and that building features, as construction type, determine the physical flood vulnerability too. Moreover, the results on physical flood vulnerability of buildings are hardly transferable to other areas and the application on supra-local level is hampered by the detailed building data required. Further research at the regional to national scale is limited by availability of event data with regard to flood characteristics, building features and damages.
We use a database of several ten thousand flood claims settled by 15 cantonal insurance companies for buildings in Switzerland. The flood claims are georeferenced at individual building level and can thus be combined with other building data as building features and (potential or occurred) exposure to flood. We apply data mining techniques to detect patterns of relations between recorded losses on the one hand and, building values, characteristics and exposure to floods on the other hand. The results help to quantify the flood vulnerability of buildings in Switzerland and are useful for estimating damages of future flood events. This data driven approach improves quantitative flood risk assessments which in turn inform stakeholders in decision making.