Interpretable machine learning for predicting urban flash flood hotspots using intertwined land and built-environment features

Zhewei Liu, Tyler Felton, Ali Mostafavi

Abstract: Highlights•Interpretable machine learning models are utilized to provide reliable prediction of flash flood hotspots.•Built environment variables are extracted for constructing model features.•The models can achieve very good accuracy in identifying flooded/non-flooded locations.•Hydrological and topological features have greater impacts on flash flood risk than other features•Localized specifications of models are needed for accurate flash flood prediction for particular cities