Statistics of extremes in poorly gauged basins: application of a lumped rainfall-runoff model in five catchments area in Walloon region (Belgium). The RS-PDM© model generates flows in time series on the basis of observed rainfall after self-calibration which is based on short period of observed rainfall and flow. This is done with daily or hourly data. The study highlights the capacity of the RS-PDM© model to simulate extreme flows in poorly gauged basins in Walloon region on the basis of observed rains and of a short period of observed flows. Concerning the time series, it was highlighted an important dispersion of the flows simulated for a given observed flow even for daily or hourly data. The Nash criteria varied between 0.59 and 0.92. However, the flows simulated via RS-PDM© make it possible to adjust a statistical distribution of extremes flows (Gumbel law) close to the statistic obtained with a long period of gauging. So, if the simulations realized by mean of this model present poor to acceptable performances concerning the time series simulations, they permit to obtain rapidly quite good statistics of extremes flows on the basis of a short gauging period.