How does the assessment of extreme precipitation profit from convection permitting climate ensembles?

In response to global warming, an intensification of extreme precipitation has been observed, and models project this trend to continue. Since the return values of extreme precipitation events are regularly used in practice in the form of heavy rainfall hazard products, a reliable update of these products is required. Moreover, for resilient planning a projection of future conditions is urged by practice stakeholders.

A promising tool for projection are convection permitting climate simulations, which have been shown to better represent extreme precipitation events compared to coarser simulations and thus provide higher confidence in future extreme estimates. However, due to the large computation time of convection permitting simulations, evaluations are mostly based on single time slice experiments. Therefore, we explore the potential of an unique transient convection permitting (2.8 km) ensemble with COSMO-CLM regional simulations (1971-2100) over Germany, with four ensemble members driven by MPI-ESM-LR, EC-EARTH, CNRM-CM5, and HadGEM2-ES with the emission scenario RCP8.5. Extreme precipitation is derived over 30-year running time slices and the scales investigated span from hourly to 3-day event duration and return periods from 1 year to 100 years, representing the wide range of events considered for application.

Within the historical period (1971-2005) we found adequate agreement between the simulations and the observation data set KOSTRA with increasing bias with longer event duration. Furthermore, the climate change signal, derived as a relative value with regard to the historical period of the simulation, was found to increase with return period and for shorter durations. Strongest relative changes lie within the range of Clausius-Clapeyron-scaling with global warming. Analysis of the uncertainty revealed a substantial residual standard deviation of the linear approximation of the change signal over global warming, highlighting the benefit of a transient ensemble that enables a more robust estimation of the change signal of extreme events. Moreover, the results indicate an increased variance of future extreme precipitation.