Uncovering a hidden population in extreme summer rainfall: Process-specific extreme value analysis in the Meuse basin

A high intensity rainfall event in July 2021 caused catastrophic floods in several tributaries of the Meuse river resulting in widespread damage and a high number of casualties. These unprecedented summer discharges raise urgent questions on the return periods of extreme summer precipitation. Conventional frequency analysis usually relies on Extreme Value Analysis of observational records. However, these datasets typically span only 100 years, with limited coverage of rare, high magnitude summer events. Quantification of the return values of far tail summer extremes like the July 2021 event is therefore subject to large statistical uncertainty. Furthermore, standard extrapolation approaches hypothesise a single population, failing to account for the distinct drivers of rare summer extremes. This could potentially yield underestimations of both precipitation and discharge extremes, undermining flood risk assessment and resulting in poorly informed policy making.

In this study we use 10.000 years of independent ECMWF seasonal forecast (SEAS5) data to investigate the dynamics of extreme Meuse summer precipitation beyond the range of historical records. We attempt to identify common spatial and temporal patterns in the drivers of far tail extreme Meuse summer rainfall and contrast these with the drivers of typical, moderate rainfall extremes. Pronounced dynamical differences suggest the existence of a secondary statistical population. By partitioning annual extremes based on the uncovered drivers, we aim to provide more robust return values of extreme summer rainfall. This research is an attempt to move beyond standard extrapolation of observations and a first step towards a more robust framework for flood protection policies.