ESR 13

ESR 13 - Analyses and modelling of changes in vulnerability with focus on private sector

Nivedita Sairam

Nivedita Sairam, home country India

Host Institute: German Research Centre for Geosciences GFZ, Section 5.4 Hydrology

Contact details:
Email: nivedita@gfz-potsdam.de
Tel: +: +49 331 288-1596

Research

The research is aimed at answering three particular questions in concepts relating to vulnerability of private households and estimated flood losses. The vulnerability of households are determined based on their adaptation to flood events. The first phase of the research aims at determining the efficiency of private precaution in reducing flood losses and also how this relationship is accounted for in existing flood loss estimation models (FLEMO). Following these analyses, the second phase aims are developing a flood loss estimation model based on causal Bayesian Networks to account for changes in vulnerability in flood loss estimation. Integrating these results in the German wide Regional Flood Model (RFM) would be the final step in my research.

Additionally, I would be analyzing the spatial transferability of my Bayesian Network based model by learning and updating the model with datasets collected from the UK floods and also comparing the model built based on empirical data collection with synthetic data for flood loss estimation as a part of my secondament in MiddleSex University.

Studies about flood risk change are currently focused on changes in flood hazard and exposure. Changes in vulnerability, e.g. changes in precaution, are so far not taken into account. The project aims at analysing past changes in vulnerability and their drivers, as well as modelling these changes for future flood risk projections. The focus is on private precaution as a main determinant of vulnerability of private households. Quantitative analyses of past changes in private precaution and their drivers will be undertaken based on a unique object-specific flood damage database (HOWAS21). To this end, data-mining approaches, in particular Random Forests and Bayesian Networks will be used yielding uncertainty estimates as well. Based on these analyses, an agent-based model will be developed to simulate changes in vulnerability for projections in the future (links with ESR12). The developed model and results will be integrated into the Regional Flood Model for Germany (RFM) undertaking flood risk analyses under past (together with ESR14) and future (together with ESR1, ESR14) changing conditions.