Influence of cloud microphysics schemes on weather model predictions of heavy precipitation

Abstract. Cloud microphysics is one of the major sources of uncertainty in numerical weather prediction models. In this work, the ability of a numerical weather prediction model to correctly predict high-impact weather events, i.e., hail and heavy rain, using different cloud microphysics schemes is evaluated statistically. Polarimetric C-band radar observations over 30 convection days are used as observation dataset. Simulations are made using the regional-scale Weather Research and Forecasting Model (WRF) with five microphysical schemes of varying complexity (double moment, spectral bin (SBM), and particle property prediction (P3)). Statistical characteristics of heavy rain and hail events of varying intensities are compared between simulations and observations. All simulations, regardless of the microphysical scheme, predict heavy rain events that cover larger average areas than those observed by radar. The frequency of these heavy rain events is similar to radar-measured heavy rain events, but still scatters by a factor of 2 around the observations, depending on the microphysical scheme. The model is generally unable to simulate extreme hail events with reflectivity thresholds of 55 dBZ and higher, although they have been observed by radar during the evaluation period. For slightly weaker hail/graupel events, only the P3 model is able to reproduce the observed statistics. Analysis of the raindrop size distribution in combination with the model mixing ratio shows that the P3, Thompson 2-mom, and Thompson aerosol-aware models produce large raindrops too frequently, and the SBM model misses large rain and graupel particles.