The western Mediterranean is a very cyclogenetic area and many of the cyclones developed over this region are associated with high-impact weather phenomena that affect the society of the coastal countries. Two ensemble prediction systems based on multiphysics and perturbed initial and boundary conditions are designed in order to improve the forecast of these heavy rain episodes. The MM5 mesoscale model nested in the ECMWF forecast fields provides the simulations, run at 22.5 km resolution for a two-day period. The multiphysics ensemble combines different model physical parameterizations while the other ensemble uses a potential vorticity (PV) inversion technique to perturb the initial state and boundary forcing of the model (PVperturbed). A PV error climatology derived from the large-scale fields allows to perturb the ECMWF PV fields using the appropriated error range. The results show an improvement of the prediction skill for both ensembles over the traditional deterministic prediction. Moreover, the PV-perturbed ensemble performs better than the multiphysics EPS. The verification procedure highlights the difficulties of evaluating the rainfall field, since its discontinuities and the limited samples of extreme events affect the scores negatively. Despite this, the results reinforce the idea that an EPS represents an improvement over a deterministic prediction.