Investigating soil erosion processes from source-to-sink to prioritise erosion hotspots in the Ethiopian highlands

• An emerging spatial-validation approach was used for evaluating model results. • Dominant soil erosion processes are dependent on spatial scale and total rainfall. • Small area erosion hotspots contribute more than half of the total soil loss. • Land use and cover changes threaten a reservoir's sustainability in the basin. • Leveraging remote sensing improves input data availability in data-limited areas. Addressing the adverse impacts of soil erosion through effective soil conservation measures (SDG15) requires a thorough understanding of the erosion processes involved. However, existing erosion prediction models are often compared solely to sediment outflow at catchment outlets without explicitly assessing these various processes. In such cases, the observed and simulated rates at the outlet may align well without effectively showing the spatial sediment redistribution patterns in the catchment. Moreover, data availability is often limited for remote catchments. This study combined the less data-intensive Unit Stream Power Erosion Deposition (USPED) model with gully erosion indices to analyse sediment redistribution and to identify erosion hotspots in a data-scarce catchment in the Ethiopian highland. The model performance was evaluated spatially at three scales, each characterised by distinct erosion processes. The main findings indicate that the model displayed overall agreement across the three scales (R² = 0.63, NSE = 0.40, KGE = 0.47, Pbias = 5%, RMSE = 1.39 t ha⁻¹ yr⁻¹). The catchment experienced an average soil erosion of 32.3 t ha⁻¹ yr⁻¹ from 2000 to 2023, resulting in a total annual loss of 0.32 million tons. The most erosion-prone areas, which comprise just 18% of the catchment area, contributed approximately 69% of the total soil erosion. The observed erosion processes vary by scale, emphasising the need for scale-aware modelling, with distinct erosion processes involved. In conclusion, the USPED model, combined with gully erosion indices, effectively captures the dominant erosion processes at various scales and identifies hotspots for targeted conservation amid land use changes.