Testing state-of-the-art radar-based precipitation estimates to simulate the exceptional European floods of July 2021

The July 2021 floods costed up to 200 lives and 30 billion euros of damages in Germany alone. The aftermath of this exceptional event was a wake-up call for questioning the ability of the hydrometeorological tools in providing timely and reliable flood forecasts, knowing that this kind of events will become more frequent due to global warming. For these events, we simulated the hourly streamflows of seven catchments in the west of Germany, by combining nine, partly polarimetric, radar-based rainfall products with two hydrological models: GR4H (Ficchì et al., 2019, 10.1016/j.jhydrol.2019.05.084), a bucket-style, lumped model, and ParFlow-CLM (Kollet and Maxwell, 2006, 10.1016/j.advwatres.2005.08.006), a distributed model that couples 3D sub-surface and overland flows. GR4H parameters were calibrated using more than 10 years of rainfall-runoff data, with emphasis on reproducing high flows, whereas ParFlow-CLM parameters were estimated based on landscape and soil properties. The key findings are as follows: (1) In comparison with estimates from rain gauges, all radar-based precipitation products underestimate the total precipitation depths for the 14^th of July, with significant differences among the products. This is most likely due to strong coalescence processes below the height monitored by the radars. (2) The hydrographs simulated by GR4H and ParFlow-CLM are similar for 3 of the 7 catchments, whereas the differences in the remaining ones could be explained by the effect of human water management, not represented in ParFlow-CLM. (3) The ability to detect the exceedance of the 100-year flood highly depends on the precipitation product used, highlighting the need for reliable precipitation estimates to reproduce such exceptional events.