Land-atmosphere coupling is theorized to play an important role in the intra-seasonal and inter-annual variability of the West-African monsoon. It is therefore fundamental to better understand and model the influence of the spatial and temporal variability of surface processes, including vegetation, on the atmospheric circulation patterns and on the water and energy cycles. The Sahelian Transpiration Evaporation and Productivity (STEP) model simulates herbaceous growth and soil water dynamics in a Sahelian environment. This model has been run within the 2nd phase of Amma Land surface Model Intercomparison Project (ALMIP2) over the Gourma and the Niger mesoscale sites for the period 2005-2008. STEP has been run using two different rainfall datasets, obtained by differently kriging the in-situ measurements by automatic rain-gauges, and, for the Mali site, two different soils descriptions, one from the ECOCLIMAP2 database and the other one based on the classification of LANDSAT satellite images. At the local scale the evaluation is carried out by comparison to in-situ local measurements (i.e. LAI by hemispheric photographs, biomass and NPP, soil moisture and flux measurements by eddy covariance) acquired over three local sites representative of the three different soil types typical of the study area. At meso scale, STEP results are evaluated by comparison to remote sensing data such as LAI, FaPAR and for the Mali site to estimations of water volume in the ponds over the Agoufou and Bangui Mallam watersheds. The results obtained are analysed with a particular focus on the model sensitivity to the soil type characterization and to the rainfall description. It is shown that a correct estimation of the soil properties (texture and depth) is fundamental to have good estimations of vegetation productivity and of the water balance components.