Objective Soil erosion poses a global threat to ecological and agricultural sustainability. In 2024, soil erosion affected 15.70% of Hubei province's total land area. While the RUSLE model is widely used as an empirical statistical model, the value of its key parameter—the soil and water conservation practice factor (P)—is mainly determined based on literature-derived experience and land use types, leaving room for improvement in accuracy. Although field runoff plot observations can provide critical data for the P factor, existing studies are often limited to single measures or short-term observations. Moreover, the data are mostly used for model validation, failing to systematically optimize the P factor. This study aims to systematically optimize the P factor values in the RUSLE model, enhance the simulation accuracy of the RUSLE model, and evaluate its impact on the simulation accuracy using measured data.

Methods This study obtained a series of P values based on the definition of P, utilizing runoff plot monitoring data from different soil and water conservation practices across Hubei province. Layer-by-layer integration of DEM, soil type, land use, vegetation cover, and terraced field datasets was performed, and the obtained P values were assigned to corresponding grid cells to optimize the P factor values. To validate the accuracy of the optimized model, the predicted values were compared with measured values using the coefficient of determination (R²) and root mean square error (RMSE) for each runoff plot. Additionally, the simulation accuracy was evaluated by comparing pre- and post-optimization model predictions with measured soil loss quantities across small watersheds.

Results The results showed that: 1) high P values (P ≥ 0.600) were distributed in western, northeastern, and southeastern Hubei province, while low P values (P < 0.300) were concentrated in the central plain area. 2) The average value of the optimized P factor was 0.457, representing a decrease of approximately 10.42% compared with the original value. 3) The accuracy assessment based on runoff plot observation data showed that R² was 0.816, RMSE was 214.6, and the relative error was 11.22%, with 98% of the data points in the optimized RUSLE predictions falling within the 95% prediction band. 4) The accuracy assessment based on small watershed monitoring data showed that the optimized simulation accuracy reached 86.630%, an improvement of 9.798% over the pre-optimization results. The proportions of severe erosion, extremely intense erosion, and intense erosion before optimization were all higher than those after optimization. This was due to the excessively high values assigned to the P factor before optimization.

Conclusions It is feasible to use the measured data from runoff plots combined with multi-source data to optimize the P factor values. The optimized P factor values are more consistent with the configuration of soil and water conservation practices in the study area, which significantly improves the simulation accuracy of the RUSLE model. The findings provide strong support for optimizing RUSLE model factors and for regional soil erosion research.