Towards operational flood forecasting in small river basins in Germany using AI-based prediction methods from large-sample hydrology

Doctoral Researcher
Name Role at KCDS
KCDS Fellow, member of Doctoral Researcher Representatives (DRR) 25/26
KCDS Supervisors
Name Role at KCDS
MATH Supervisor, member of the Steering Committee
SEE Supervisor

Abstract

Extreme precipitation events leading to floods in small river catchments (approx. 5 – 500 km²) represent one of the greatest natural hazards in Central Europe, with serious consequences for human life and infrastructure. Small river catchments react quickly to extreme weather conditions, which shortens warning times and greatly increases the uncertainty of forecasts (LAWA, 2021). Recent research shows that modern ML methods, especially from the field of deep learning (DL), offer potential to close this gap. The aim is to use modern machine learning methods to a) enable robust and uniform hydrological forecasts in small catchments throughout Germany and b) significantly increase the accuracy of forecasts of hydrological extremes in these areas.