Heavy precipitation events in the western Mediterranean area often lead to large social and economic impacts. Especially in late summer and in autumn, rainfall systems can be fed efficiently by warm and moist air masses in low tropospheric levels. Due to complex interactions of various lifting mechanisms and the steep Mediterranean coastal orography, the initiation and the development of those heavy precipitation events is not fully understood. Even more, the accurate prediction of severe weather events remains challenging. The improvement of the predictability of those events is a main subject of current meteorological research. Within this thesis, the physical processes and the predictability of prototypic heavy precipitation events in the western Mediterranean area are studied.
A novel methodology was applied to assess and to improve the predictability of this heavy precipitation event by numerical ensemble simulations, based on diverse percentages of two different sources of atmospheric uncertainty. Synoptic-scale uncertainty was introduced by various initial and boundary conditions, whereas convective-scale uncertainty (i.e. small-scale uncertainty) was gained by a stochastic parameterization scheme for cumulus convection.
Regarding the synoptically stronger forced event, a proportionally larger influence of synoptic-scale uncertainty led to a better ensemble performance. Concerning the weaker forced case, an overdispersive ensemble resulted for the
settings with more than 50% of synoptic-scale uncertainty. The best performance was achieved for the ensemble
with equally distributed sources of uncertainties.