The north-western Mediterranean region is susceptible to heavy precipitation events (HPEs) between the months of September and November with south-eastern France being one of the areas most affected. Numerical forecasting of such events has drastically improved in recent years. However, because of the complex and multi-scale interactions taking place in deep convective cloud systems, model errors remain, especially in the representation of the physical processes. In particular, the cloud physics and turbulence parameterisations rely upon unavoidable simplifications and assumptions which clearly limit the potential accuracy of deterministic forecasts in capturing extreme weather events. To represent the uncertainties associated with these parameterisations, a probabilistic methodology using ensemble prediction systems (EPSs) was proposed. Perturbations were introduced upon the time tendencies of the warm and cold microphysical and turbulence processes, as well as the adjustable parameters of the microphysical parameterisations. This approach was tested and assessed upon two idealised convective events with the aim of uncovering the processes which induced the greatest dispersion in the surface rainfall. Following these tests, ensembles were constructed for seven real world events which occurred over south-eastern France in the autumns of 2010, 2011 and 2012. Simultaneously perturbing the time tendencies of the warm and cold microphysical and turbulence processes was shown to produce the greatest degree of dispersion in the rainfall pattern with the level of that dispersion depending upon the nature of the precipitating event. The dispersion is generally more important when the incident flow is weak to moderate and precipitation is observed on the plains. For certain events, the level of dispersion introduced by perturbations upon the physical processes is similar to that generated by perturbing the initial and boundary conditions. It is concluded that, even though the impact of perturbing the physical parameterisations is moderate, it is sufficiently important to warrant an inclusion of such perturbations in an operational EPS.
