LAUSR

Abstract Erosion is a natural phenomenon, which may have destructive effects on environments and human societies around the world. This study aims to identify and map soil erosion hazard in the Mellah Watershed (North-eastern Algeria), based on an empirical model (RUSLE: Revised Universal Soil Loss Equation), a semi quantitative model (AHP: Analytic Hierarchy Process) and a statistical model (FR: Frequency Ratio). Several erosion-conditioning factors were used as input data, namely rainfall, elevation, slope, land cover, soil type, Normalized Difference Vegetation Index (NDVI), Stream Power Index (SPI), Topographic Wetness Index (TWI) and distances from roads and streams. The study resulted in generating vulnerability maps with five erosion risk classes, ranging from very low to very high risk. Areas of very high erosion risk were shown to occupy 5.06%, 10.91% and 12.57%, of the watershed area, using RUSLE, FR and AHP models respectively. Results obtained by the three adopted models were validated using the Receiver Operating Characteristic curves (ROCs), and the Areas Under the Curve (AUC). The AUC values were 97.9%, 95.7% and 94.2% for RUSLE, AHP and FR models respectively, implying reasonably good performances for the three models.