Compounding drivers amplify the severity of river floods

Estimating the risk of river flooding under global warming is challenging, mainly due to the compound nature of the various drivers, which is not yet fully understood. Our study aims to quantitatively unravel the complex dynamics of multiple factors that interact and influence river flooding. Using interpretable machine learning techniques, we analyzed thousands of global catchments to identify the role of compounding drivers in river flooding. Our results indicate that these compounding drivers have played a significant role in increasing the magnitude of river floods over the past four decades. In particular, the influence of the interaction effects between these drivers becomes more pronounced with increasing flood magnitude, and the degree is modulated by specific physioclimatic conditions. Importantly, traditional flood analysis will underestimate the magnitude of extreme floods due to insufficient consideration of the varying compounding effects in flood generation. Overall, our results emphasize the need to more carefully incorporate compounding factors to improve extreme flood estimates.