The paradoxical evolution of runoff in the pastoral Sahel : analysis of the hydrological changes over the Agoufou watershed (Mali) using the KINEROS-2 model

In recent decades, the Sahel has witnessed a paradoxical increase in surface water despite a general precipitation decline. This phenomenon, commonly referred to as "the Sahelian paradox", is not completely understood yet. The role of cropland expansion due to the increasing food demand by a growing population has been often put forward to explain this situation for the cultivated Sahel. However, this hypothesis does not hold in pastoral areas where the same phenomenon is observed. Several other processes, such as the degradation of natural vegetation following the major droughts of the 1970s and the 1980s, the development of crusted topsoils, the intensification of the rainfall regime and the development of the drainage network, have been suggested to account for this situation. In this paper, a modeling approach is proposed to explore, quantify and rank different processes that could be at play in pastoral Sahel. The kinematic runoff and erosion model (KINEROS-2) is applied to the Agoufou watershed (245 km(2)), in the Gourma region in Mali, which underwent a significant increase of surface runoff during the last 60 years. Two periods are simulated, the "past" case (1960-1975) preceding the Sahelian drought and the "present" case (2000-2015). Surface hydrology and land cover characteristics for these two periods are derived by the analysis of aerial photographs, available in 1956, and high-resolution remote sensing images in 2011. The major changes identified are (1) a partial crusting of isolated dunes, (2) an increase of drainage network density, (3) a marked decrease in vegetation with the nonrecovery of tiger bush and vegetation growing on shallow sandy soils, and (4) important changes in soil properties with the apparition of impervious soils instead of shallow sandy soil. The KINEROS-2 model was parameterized to simulate these changes in combination or independently. The results obtained by this model display a significant increase in annual discharge between the "past" and the "present" case (p value < 0.001), which is consistent with observations, despite a slight overestimation of the past discharge. Mean annual discharges are estimated at 0.51 x 10(6) m(3) (2.1 mm yr(-1)) and 3.29 x 10(6) m(3) (13.4 mm yr(-1)) for past and present, respectively. Changes in soil properties and vegetation cover (tiger bush thickets and grassland on shallow sandy soil) are found to be the main factors causing this increase of simulated runoff, with the drainage network development contributing to a lesser extent but with a positive feedback. These results shed a new light on the Sahelian paradox phenomenon in the absence of land use change and call for further tests in other areas and/or with other models. The synergetic processes highlighted here could play a role in other Sahelian watersheds where runoff increase has been also observed