This study explores the benefit that can be drawn from incorporating local potential vorticity (PV) modifications into a numerical forecast. The case under study is the severe cyclogenesis that occurred in the Western Mediterranean from 9 to 12 November 2001. This case was shown to be extremely sensitive to the upper-level flow, which governed to a great extent the deepening of the depression and the location and intensity of its associated precipitation. Corrections of PV were guided by METEOSAT-7 water vapour observations and restricted to the upper troposphere, mainly altering the topography of the dynamical tropopause. Using both the French operational global model ARPEGE and the non-hydrostatic mesoscale model Meso-NH, it is shown that carefully designed PV corrections lead to a substantial improvement in the simulation of the storm, both in terms of surface pressure, cloud cover and precipitation forecasts. Furthermore, the impact of the modifications is shown to be a maximum when they are introduced at the time corresponding to the incipient stage of cyclogenesis.