Purpose <p>This study evaluates long-term soil loss in the small semi-arid Keskin watershed of central Anatolia, Türkiye, by testing whether Revised Universal Soil Loss Equation (RUSLE)-based estimates, after correction for sediment delivery using Sediment Delivery Ratio (SDR), are consistent with independently measured sediment accumulation in the Keskin Reservoir over the same period. Reservoir sediment accumulation is treated as an independent, field-based benchmark, while its uncertainties are explicitly acknowledged. The main research question is therefore whether the RUSLE/SDR-derived sediment-yield estimate provides a reasonable first-order approximation of the measured reservoir sediment mass in a small, relatively well-connected basin.</p> Materials and methods <p>Soil loss for 1998–2015 was estimated in a GIS framework using the RUSLE with spatially distributed R, K, LS, C and P factors derived from meteorological data, soil analyses, topographic maps, CORINE land cover, orthophotos and field observations. A distributed SDR approach was applied to convert gross soil loss into effective sediment delivery. Model results were evaluated against direct measurements of reservoir-floor sediment thickness obtained during an exceptional low-water stage in 2015, when nearly half of the lake floor was exposed. Sediment thickness data were filtered to remove small-scale microtopographic effects, extrapolated across submerged sectors using a basin-scale depositional model, and converted to sediment volume. Wet bulk density and moisture content were measured to estimate dry sediment mass.</p> Results and discussion <p>The total dry sediment mass accumulated in the reservoir over 1998–2015 is approximately 22,000 ± 1,800 t, considering only the propagated uncertainty from wet bulk-density variability, whereas the RUSLE/SDR-derived sediment-yield estimate for the same period is about 23,750 t. This close correspondence indicates first-order consistency between the modelled sediment-yield estimate and the independently measured reservoir sediment mass. In this small basin, the agreement suggests that the RUSLE/SDR workflow provides a reasonable first-order approximation of sediment yield, although it does not eliminate uncertainties related to sediment-volume extrapolation, dry bulk density, trapping efficiency and possible compensating errors. The relatively modest sediment yield should not be interpreted as evidence of limited land degradation. Field observations and previous work instead indicate widespread degradation associated with shallow, stony soils, poor land management and overgrazing, suggesting that low sediment output partly reflects limited production of transportable fine material in an already degraded semi-arid system.</p> Conclusions <p>The Keskin case demonstrates that independent comparison with measured reservoir sediment accumulation can substantially strengthen the interpretation of empirical erosion models in semi-arid Anatolian catchments. In small and relatively well-connected basins, reservoir sedimentation provides a practical field-based benchmark for evaluating RUSLE/SDR-derived sediment-yield estimates. At the same time, low sediment yield should not automatically be equated with low land degradation, because exhausted, coarse-textured soils may export only limited fine sediment despite substantial environmental degradation.</p>