Résumé
West Africa experienced severe drought between the 1970s and 1990s. A shift in the hydrological regime was observed in most of the catchments of the central Sahel (the semi-arid fringe south of the Sahara): from the drought onward, runoff increased significantly despite the rainfall deficit, and continues to increase today. Long-term field observations by the AMMA-CATCH observatory and other research programs in this region suggest that this rainfall-runoff decoupling regime is associated with the crossing of a tipping point. In this study, we investigated whether the drought of the 1970s-1990s could have triggered this shift. We used a simple system dynamics model representing feedbacks between rainfall, herbaceous and woody vegetation and runoff at the catchment scale. These feedbacks allow to represent the local response of vegetation to rainfall, soil water and runoff as an internal, dynamical process. The model forced with observed rainfall since the 1950s proved robust, as it reproduced well the runoff dynamics observed during this period in a selection of catchments in the central Sahel and the southern sub-humid zone closer to the Gulf of Guinea. Using historical CMIP6 precipitation simulations, the probability of a regime shift was found to be twice as high when the model was forced with a rainfall series reproducing drought (fully coupled ocean-atmosphere climate simulations) than with a series without drought (AMIP - atmosphere only). We conclude that drought was a sufficient factor to trigger the observed regime shift, although other processes (not questioned here) may have been simultaneously involved, such as human-induced land cover changes. With a "conventional" hydrological model, the regime shift observed in the central Sahel could only have been reproduced by prescribing both rainfall and the evolution of surface conditions. Our results suggest that, at least in arid and semi-arid regions, the inclusion of rain-vegetation-runoff feedbacks in critical zone models could help improve simulations and better assess the likelihood of triggering hydrological tipping points under future climate and land cover change scenarios.