Abstract

Deep convective clouds and the resulting heavy precipitation, hail, lightning and wind gusts can cause severe damage. Their forecasting with current numerical weather prediction models is a challenge, both because of limitations on spatial resolution and because of the low predictability of timing, location and properties of the storms. In numerical simulations, the impact of changes in meteorological and in microphysical conditions (e.g. aerosol particles acting as cloud condensation nuclei, strength of ice formation) can be investigated in perturbed parameter sensitivity experiments, and I will show results for different cases and different variables of interest. However, for realistic cases, such sensitivity experiments become very costly or not feasible at all. For some applications, statistical emulation can be used as an approach to address this problem. Furthermore, I will show attempts to constrain ice-phase processes, which are particularly uncertain, using satellite observations.