Flood loss estimation Models for companies and households affected by flash floods

This study introduces FLEMOflash, a new multivariate probabilistic flood loss estimation model designed specifically to capture the drivers and impacts of flash floods, addressing key gaps in loss assessments that traditionally focus on fluvial flooding. Developed using survey data from households and companies affected by flash flood events in Germany in 2002, 2016, and 2021, the model integrates machine learning techniques, Elastic Net, Random Forest, and XGBoost, for data-driven feature selection with Bayesian networks to estimate losses probabilistically while accounting for uncertainty. 

The findings highlight the critical role of emergency response and preparedness in reducing losses, showing that effective emergency measures can reduce building losses by up to 47% for large companies under extreme hazard conditions. Similarly, households with clear knowledge of appropriate actions during high water depths reduced building losses by 77% and contents losses by 55%. Overall, FLEMOflash provides a robust tool to support flash flood risk communication and management by enabling more accurate and uncertainty-aware loss estimation.