Résumé
The aim of this study is to characterise the hydrologic behaviour of a sahelian catena in western Niger. From previous studies, it is acknowledged that soil surface characteristics (crust type. micro-relief, vegetation density), referred to as sail surface features, play a determinant role in infiltration and runoff processes. We intend here to quantify the horizontal redistribution of water on the catena (that is the partition infiltration-runoff and subsequent surface transfer), as affected by the spatial distribution of soil surface features. Experimental data from runoff plots (100 m²), representative of one type of soil surface feature, show that infiltration is low on areas with permanent crusting (plateau bare soil, fallow) ; runoff is estimated respectively to 50 and 25 % of the annual rainfall. On cultivated areas, crusts destruction by the weeding operations and subsequent restoration induce a high temporal variability of infiltration. and runoff is estimated to less than 12 % of the rainfall depth . On the catena, infiltration is limited to the first meters of the soil , except in the areas fed with high quantities of water (plateau vegetation, flood spreading zone, gully) where the infiltration depth is higher than 2 to 3 times the rainfall depth . There is no water storage in the soil from one season to the next, and only the zones of water accumulation are likely to feed deep drainage towards the aquifer. In order to strengthen the experimental results and to establish a seasonal water balance on the catena, the physically based, distributed, SWATCH hydrologic model was modified to allow the simulation of infiltration in a two- layer soil system (crust-subsoil). The model is first calibrated at the plot scale, assumed to be representative of the hillslope processes. The model parameters defined on each plot (i.e. for each sail surface feature) are then affected to zones of homogeneous surface feature defined on the catchments from a surface feature map. After calibration of the channel transfer parameters, this simplified representation of the watersheds lead to a satisfactory simulation of observed discharges. The deep infiltration depth computed by the model over the wider catchment (0,9 square km) at the end of the rainy season represents 12 percent of the rainfall depth. But, as a part of this water amount (not quantified) is evaporated in dry season, the fraction of water draining towards the aquifer is probably low. It is concluded that the contribution of this type of catena to annual groundwater recharge is low regarding that provided by the pools, studied elsewhere.