Leveraging Artificial Intelligence for Hydrological Response: Mapping Water Erosion in a Tunisian Context

International audience This paper aims to map and quantify water erosion in the El Gouazine watershed using four empirical models integrated into a Geographic Information System (GIS). The models studied are the Universal Soil Loss Equation (USLE), its revised version (RUSLE), its development (MUSLE) and the FAO equation. Using these tools, a synthetic map of soil losses was created, providing an overview of the intensity and distribution of erosion. The results show that more than 74% of the basin suffers from low erosion, with losses less than 2.5 tons per hectare per year, while 20% of the area is affected by moderate erosion (2.5 to 10 tons/ha/year), and 13.74% by severe erosion (greater than 15 tons/ha/year). After the implementation of sustainable soil and water management (SWM) practices, a significant reduction in erosion was observed, illustrating the effectiveness of the interventions. The RUSLE and MUSLE models showed substantial reductions in soil losses, confirming their essential role in preserving natural resources and limiting erosion. Soil erosion is a major threat to agricultural land in Tunisia, affecting approximately 3 million hectares and posing a risk to the sustainability of small hilly lakes. In the El Gouazine watershed, located in central Tunisia, this phenomenon is amplified by a semi-arid Mediterranean climate, rugged topography, fragile soils, and increasing anthropogenic pressure.