Impact of the horizontal resolution on the simulation of extremes

The simulation of extremes using climate models is still a challenging task. Currently, the model grid horizontal resolution of state-of-the art regional climate models (RCMs) is about 11–25 km, which may still be too coarse to represent local extremes realistically. In this study we use dynamically downscaled ERA-40 reanalysis data of the RCM COSMO-CLM at 18 km resolution, downscale it dynamically further to 4.5 km and finally to 1.3 km to investigate the impact of the horizontal resolution on extremes. Extremes are estimated as return levels for the 2, 5 and 10‑year return periods using ‘peaks-over-threshold’ (POT) models. Daily return levels are calculated for precipitation and maximum 2 m temperature in summer as well as precipitation and 2 m minimum temperature in winter. The results show that CCLM is able to capture the spatial and temporal structure of the observed extremes, except for summer precipitation extremes. Furthermore, the spatial variability of the return levels increases with resolution. This effect is more distinct in case of temperature extremes due to a higher correlation with the better resolved orography. This dependency increases with increasing horizontal resolution. In comparison to observations, the spatial variability of temperature extremes is better simulated at a resolution of 1.3 km, but the return levels are cold-biased in summer and warm-biased in winter. Regarding precipitation, the spatial variability improves as well, although the return levels were slightly overestimated in summer by all CCLM simulations. In summary, the results indicate that an increase of the horizontal resolution of CCLM does have a significant effect on the simulation of extremes and that impact models and assessment studies may benefit from such high-resolution model output