Flood damage modelling in the residential sector

Flood damage modelling in the residential sector

© @ULiège

Performing a flood risk assessment requires flood damage models, which guide the development of flood risk reduction strategies. Despite increasing efforts in the development of these models, spatial transferability and validation remain challenging due to a lack of reliable empirical data and the heterogeneity of potentially affected assets and economic contexts (Scorzini et al., 2022).

In July 2021, Belgium experienced an extreme flood event, with rainfall volumes equivalent to three months' worth of precipitation falling within just two days in the eastern part of the country. The event caused severe damage to residential buildings, industries, and infrastructure, including railways and roads, with an estimated total cost exceeding EUR 3 billion. A research initiative was launched to collect damage data, as well as information on hazard characteristics, the vulnerability of exposed assets, socio-economic factors, and the coping capacity of inhabitants and emergency services (i.e., emergency and precautionary measures) in various municipalities along the Vesdre River, one of the most impacted sub-catchments. The resulting database is now being used to analyse flood damage mechanisms and to support the calibration and validation of a flood damage model for the region.

In addition to the Belgian data collection efforts, similar initiatives have been undertaken in the affected areas of Germany and the Netherlands. Machine learning techniques are being used to identify the most important factors contributing to flood damage in the residential sector, as well as to develop a transboundary flood damage model that predicts both high-frequency and low-frequency extreme events.