On the sensitivity of precipitation in convection-permitting climate simulations in the Eastern Alpine region

This study evaluates the representation of precipitation in a set of multi-year convection-permitting sensitivity experiments over the European Alpine region. In the last few years, studies have consistently demonstrated the added value of convection-permitting regional climate models (RCMs) over coarser resolved RCMs with parametrized convection for the representation of precipitation. They allow unprecedented insights in the role of mesoscale processes in the climate system and are hoped to provide more realistic climate change projections. However, their uncertainties due to variations in the models' configurations are still a matter of ongoing research. The present study addresses this issue using a set of hindcast simulations with CCLM v5.0 at 0.0275 degrees grid spacing (similar to 3 km), from January 2006 to December 2009. Six configuratio parameters are chosen amongst the following categories: parametrization of turbulence, parametrization of microphysics, surface orography, lateral boundary forcing, and driving data. They are tested individually with regards to a reference experiment and evaluated against two high-resolution gridded (1 km grid spacing) observational datasets over Austria, for winter and summer seasons. Also, a simulation with WRF v3.7.1, using a similar experimental set up, provides an estimate for the model-dependency of precipitation biases. The added-value regarding coarser resolved (0.11 degrees grid spacing, i.e. similar to 12.5 km grid spacing) CCLM and WRF simulations from the EURO-CORDEX initiative, which were used as driving data, is discussed. In agreement with previous studies, convection-permitting experiments show added-value compared to their driving data regarding precipitation extremes, the height-dependency and the mean diurnal cycle of precipitation in summer. However, CCLM at convection-permitting resolution suffers from a predominant wet bias in mountainous regions in winter, and a dry bias in the eastern Alpine forelands during summer. The latter is related to a significant underestimation of the spatial extent of the precipitation events that cannot be compensated by overestimated intensities as it happens mostly in the mountains. This interplay is insensitive to the parameters tested. In contrast, WRF at convection-permitting resolution largely overestimates precipitation because of too large and too intense precipitation events, though there are improvements in summertime. Using a direct nesting strategy with an operational high-resolution (0.225 degrees grid spacing, i.e. similar to 20 km grid spacing) analysis product from numerical weather prediction as driving data improves seasonal biases in the 3 km CCLM domain, thanks to reduced (enhanced) frontal precipitation in winter (summer).