Discussion

Validation of model results

The CSLE model has been used widely on the Loess Plateau (Kou et al., 2020; Wei et al., 2021; Wu et al., 2016; Zhang et al., 2020), and advances in remote sensing data resolution and geographic information system technology have vastly improved model applicability and accuracy.
Current assessments of soil erosion models typically use measurements based on field monitoring and isotope tracer techniques, often relying on depleted radionuclide inventories to simulate medium/long-term soil redistribution rates. Other remote sensing means use high-resolution aerial imagery to assess erosion severity in a qualitative/semi-quantitative manner (Fischer et al., 2018). Due to the lack of detailed historical field observation data in this study, we collated the results of published papers in the study area and compared them to our model results. Huang et al. (2020) used the CSLE model to assess soil losses in the Jiuyuangou watershed from 1970 to 2015, drew soil erosion modulus grading maps (1977, 2004, and 2015), and analyzed the effects of land use and slope on soil erosion. We compared the soil erosion modulus in 2015 with our results for the slope classification statistics (Figure 8a), carried out a correlation analysis based on this comparison, and calculated the correlation coefficient (R2) and Nash coefficient (NSE) using the previous data as the verification value. The results of both studies differed for slopes less than 15° (MRE=68.13%) but were more consistent for slopes greater than 15° (MRE=10.84%). The difference is acceptable due to the different calculation methods used for the T factor. In Figure 8b, the R2 (0.93) and NSE (>0.8) indicate credible model results. We also compared our results with the soil erosion assessment conducted by Wang (2018) in the Wuding River Basin from 2000 to 2014, which reported that weak erosion dominated the watershed (88.35% of the total watershed area), consistent with this study.