QPF and QPE performance during the July 2021 flood event

Last July, the Ardennes/Eifel regions and the wider Moselle, Meuse and Rhine basins were hit by severe flooding. Large precipitation depths were forecast as of the weekend prior to the flooding, some 5 days in advance. Post the event, however, it turned out that precipitation forecasts were significantly lower than the observed precipitation depths. This caused various hydrological forecasts to initially underestimate river flow. That, in turn, had all sorts of fall-out, including a significant additional level of incredulity on the part of flood responders, who were already taken by surprise due to the unusual timing of the flood. 

The analyses show the extent to which various quantitative precipitation forecast (QPF) products (including DWD-ICON and ECMWF deterministic and ensemble products) coincide and deviate from the ‘observational’ quantitative precipitation estimates (QPE), both in terms of quantity, timing and location. The analysis shows that the true uncertainty in the weather forecasts was significantly larger than what the ensemble forecasts suggested. 

The analyses also show that the various observational and reanalysis products (including ERA5, E-OBS, HYRAS, RADOLAN and those originating from various operational hydrological forecasting systems) greatly vary in their estimate of actual/true/observed precipitation amounts. 

The analysis is conducted both on the level of the original grids as well as on the level of various Meuse and Rhine tributaries including the Vesdre and Ahr river basins. Also, the effect of the different QPE and QPF estimates on the simulated and forecasted discharge is shown. 

The present case study provides anecdotal evidence of the performance of QPF originating from numerical weather prediction products in severe flood events. Additional emphasis will be given to the propagation of these uncertainties to river flow forecasts produced by the Dutch national fluvial forecasting service, where the lead author was acting as flood duty officer in the week preceding the floods.