Item

Abstract

Compound events-combinations of multiple hazards contributing to societal or environmental risk-can significantly exacerbate disaster impacts, yet their effect on flood-related losses remains poorly quantified. Using a pan-European multi-hazard dataset spanning 1981-2020 at sub-national resolution, we find that more than 70% of recorded flood events involve compounding hazards, including meteorological extremes such as heatwaves and windstorms, alongside anomalous river discharge, with an increasing trend over time. The top 1% of events by economic losses are all compound, with total losses exceeding 167 billion euros above single-hazard floods. We introduce a compound hazard complexity metric and combine it with regional exposure and vulnerability data. Applying an ensemble machine learning model with explainable AI and a Double Machine Learning Causal Forest, we show that regions with higher complexity experience greater losses, even after controlling for flood magnitude and vulnerability, highlighting the importance of compound hazard information in risk modeling.