One of the main challenges encountered in studying water erosion in many regions of the world lies in the scarcity and unavailability of validation data, which limits the accuracy and reliability of the applied erosion models. In this regard, this study aims to apply an approach for assessing the spatial distribution of soil loss and sediment yield (SY), and for identifying critical erosion-prone areas in a semi-arid mountainous context. The approach adopted is based on the use of the Revised Universal Soil Loss Equation (RUSLE) and the Sediment Delivery Ratio (SDR), combined with the spatial processing technologies offered by Google Earth Engine (GEE) and a geographic information system (GIS). To validate the results, the performance of the model used was evaluated using ROC (AUC) analysis, based on field observations, while sediment yield data from bathymetric surveys conducted in the studied dams were used to supplement and support this validation. This study reveals that the R and LS factors, followed by the C factor, contribute most to soil erosion. The spatial combination of the most contributory classes of these three dominant factors revealed that critical areas represent 66.35% and 67.96% of the total surface areas of the N'fis and Assif el-Mal watersheds, respectively. These areas correspond to high-risk erosion zones, with soil loss rates exceeding 10 t/ha/yr. The results confirmed not only the usefulness of the RUSLE model in the arid and semi-arid mountainous context but also highlighted its remarkable reliability, thanks to the validation of its predictions using recent bathymetric data, complemented by field observations and satellite imagery. This study provides valuable insights for the application of the RUSLE/SDR model to support sustainable soil management and reduce reservoir siltation.