The influence of convection-permitting regional climate modeling on future projections of extreme precipitation: dependency on topography and timescale

Abstract

Owing to computational advances, an ever growing percentage of regional climate simulations are being performed at convection-permitting scale (CPS, or a horizontal grid scale below 4 km). One particular area where CPS could be of added value is in future projections of extreme precipitation, particularly for short timescales (e.g. hourly). However, recent studies that compare the sensitivity of extreme hourly precipitation at CPS and non-convection-permitting scale (nCPS) have produced mixed results, with some reporting a significantly higher future increase of extremes at CPS, while others do not. However, the domains used in these studies differ significantly in orographic complexity, and include both mountain ranges as well as lowlands with minimal topographical features. Therefore, the goal of this study is to investigate if and how the difference between nCPS and CPS future extreme precipitation projections might depend on topographic complexity and timescale. The study area is Belgium and surroundings, and is comprised of lowland in the north (Flanders) and a low mountain range in the south (Ardennes). These two distinct topographical regions are separated in the analysis. We perform and analyze three sets of 30 year climate simulations (hindcast, control and end-of-century RCP 8.5) at both nCPS (12 km resolution) and CPS (2.5 km resolution), using the regional climate model COSMO-CLM. Results show that for our study area, the difference between nCPS and CPS future extreme precipitation depends on both timescale and topography. Despite a background of general summer drying in our region caused by changes in large-scale circulation, the CPS simulations predict a significant increase in the frequency of daily and hourly extreme precipitation events, for both the lowland and mountain areas. The nCPS simulations are able to reproduce this increase for hourly extremes in mountain areas, but significantly underestimate the increase in hourly extremes in lowlands, as well as the increase in the most extreme daily precipitation events in both the lowland and mountain areas.