Sensitivity of convection-permitting regional climate simulations to the level of microphysics parameterization complexity

van Ypersele de Strihou, Jean-Pascal [UCL] Marbaix, Philippe [UCL] et al.

Over the last decades, bulk microphysics schemes with prognostic hydrometeors were developed for numerical weather prediction (NWP) models. As enhanced computational resources enable climate models to operate at horizontal grid resolution comparable to NWP, it also becomes desirable to integrate them using improved microphysics approximations. However, given the typically long time scales of climate simulations, such schemes still impose a large computational burden. Hence, an important question to address is: what level of complexity is required to simulate the essential processes relevant to convective-resolving climate studies? In this study, we investigate the sensitivity of simulated convective precipitation characteristics to the level of complexity used in the microphysics parameterization. From a 10 year climate simulation, 20 intense convective events were selected and evaluated against state-of-the-art observations. Factors examined include the number of predicted moments, the number of predicted ice categories, and formulations of collisional raindrop breakup. First, it was shown that 2-moment schemes (2M) do not outperform 1-moment schemes (1M) in terms of surface precipitation, due to compensating effects. Second, we found a large sensitivity of peak precipitation to the drop breakup in 2M schemes. Intense breakup results in small drops, slower fallout and more evaporation compared to weak breakup. Using Joint Probability Density Functions of radar reflectivity and drop diameter versus rain rates, we could evaluate the breakup formulation against observations. Last, we found that the addition of hail to a 2M has a larger impact on radar reflectivity than on surface precipitation. In general, our study suggests that improved understanding of basic processes (such as breakup and graupel-to-hail conversion) is required, rather than increased complexity and higher resolution, to bear significant improvement for deep-convective climate simulations.