The potential tipping of key elements of the Earth system, which would drastically impact the living conditions on Earth, is a growing concern of the scientific community. While anticipation of the future is generally based on a vision of progressive transition, the possibility of abrupt transition should lead to a change of perspective. In the Sahel, long term monitoring data have shown that runoff has continuously increased since the 1950s despite the severe regional drought of the 70s-90s. This paradoxical evolution has been attributed to changes in watershed properties due to drought-induced vegetation collapse and land used change. This type of hydrological response to perturbations suggests a regime shift, during which the sahelian watersheds have tipped from a state of ”low runoff” to a state of ”high runoff”. In the Sudanian, sub-humid, zone a gradual and reversible response between river flow and rain was observed. Although the region is also subjected to strong expansion of cultivated areas and deforestation, no regime shift has been observed. Tipping points in land surface hydrology are not well understood and documented nowadays, although it could lead to irreversible shifts and could strongly impact water security and flood risks. Feedbacks between flows, vegetation, land cover, soil structure and agro-pastorale practices are first order drivers of the hydrological changes observed at the decade scale. In this complex system, feedback loops can lead to tipping points explaining regime shifts. To better understand past changes and explore future trajectories, we are developing a systemdynamics model representing the first order processes and feedbacks. Each relation is built based on the availability of long observation series giving a unique opportunity to go beyond a theoretical approach. Observations are coming from the AMMA-CATCH long term observatory, complemented by historical rainfall and discharges measurements (1940 – present), and land use / land cover maps (1950s onwards). The approach has been validated at the local scale ( ˜ 1 km2) with only a limited set of feedbacks interacting (rainfall - soil- vegetation). Here we present first results obtained at the watershed scale (> 100 km2), accounting for landscape heterogeneity and channel processes. The approach is proved to be relevant at this scale. It allows to identify domains within the watershed where regime shifts occur, and to define the associated equilibrium states. Ultimately, we aim at deciphering where the risk of irreversible hydrological regime shift is high under various climatic and socio-economical possible futures.