During the past decade, significant progress has been made towards the development of surface schemes for Numerical Weather Prediction (NWP) models. In order to simulate surface processes in such models adequately, a realistic description of surface properties is necessary. This is achieved using time series of NOAA (National Oceanic and Atmospheric Administration) AVHRR (Advanced Very High Resolution Radiometer) satellite data, and the method relies on an automatic clustering of multi-temporal NDVI (Normalized Difference Vegetation Index) maximum values. This leads to eleven vegetation classes where the forests are readily identified from a thorough analysis of visible reflectances in early summer. This intermediate step and look-up tables allow for a final mapping of the roughness length, leaf area index, fractional vegetation cover and minimum stomatal resistance. These are primary parameters of the land surface parameterization scheme ISBA used at Meteo-France and are updated from the seasonality of NDVI. In a last step, parameters are spatially averaged at the grid box scale according to aggregation rules. The reliability of the mapping is demonstrated through the improvements of the French NWP model in short-range forecasts.