Analysis of precipitation extremes using generalized extreme value theory in convection-resolving climate simulations

Abstract:

Extreme value analysis provides a useful tool for the estimation of the probability of unusually large precipitation events that may cause devastating floods. Climate modeling studies, as well as theory and observations suggest that such extreme precipitation events will intensify with global warming. Current projections of extreme precipitation events are based on conventional global and regional climate models, which due to the coarse resolution need to parameterize convective precipitation. The use of convection parameterization leads to a poor representation of extreme precipitation especially on the sub-daily time scale.

Here we present an analysis of extreme precipitation events in convection-resolving climate simulations. The simulations are performed with the COSMO-CLM model at a convection-resolving resolution of 2.2 km across an extended region covering the Alps and its larger-scale surrounding from northern Italy to northern Germany. Ten-year long control and scenario simulations (1991-2000 and 2081-2090) are conducted, driven by a CMIP5 coupled climate model (MPI-ESM-LR) under an RCP8.5 greenhouse gas scenario. Previous work on validation has shown a dramatic improvement in the representation of sub-daily precipitation statistics (Ban et al. 2014).

Here we apply generalized extreme value theory to address projections of 5-day mean, daily and hourly extreme precipitation events in all seasons. We present a detailed intercomparison of the convection-resolving model against a conventional (12 km grid spacing) climate model in terms of model performance and differences between climate change signals. Analysis shows negligible differences between the two simulations for winter precipitation on all time scales, while in other seasons and especially for hourly precipitation the 2km model shows smaller changes and is more confident than the 12km model. The results are generally consistent with a previous analysis using percentile indices (Ban et al. 2015) and indicate that changes in extreme summer precipitation qualitatively scale with the Clausius-Clapeyron relationship. In other seasons, when large-scale precipitation becomes more dominant type of precipitation, the change in extreme precipitation exceeds the Clausius-Clapeyron rate.