The German flood cooking book: ingredients and regional flavors of floods

Floods are generated by complex interactions between different processes that are still poorly understood especially at the large scale. Here we investigate 177,267 rainfall-runoff events across 373 German catchments to identify which spatial and temporal properties of precipitation events and wetness state of catchments lead to the emergence of floods by applying two machine learning approaches based on decision trees and random forests. We found that in wet mountainous catchments with rather shallow soils, rainfall characteristics play an ultimate role in modulating flood occurrence. Instead, in the drier catchments the occurrence of flood events is driven primarily by the pre-event wetness conditions and spatial interplay between soil moisture and rainfall within the catchment. Snow cover additionally enhances flood occurrence in all study catchments. The identified ingredients and regional flavors of flood events in Germany provide new insights on the spatial dynamics of hydro-meteorological processes leading to floods and might become an essential tool to foster regionally-coordinated flood management strategies and early warning systems.