Environmental Erosion Research Journal

1- Introduction
Soil erosion is one of the major global environmental challenges that threatens soil quality, natural resources, ecosystems, agriculture, and water resources. This phenomenon has particularly severe negative impacts on environmental sustainability and food security in arid and semi-arid regions such as Iran. Numerous models have been developed to predict and manage soil erosion, and the RUSLE model has gained widespread use due to its effective performance in Iran's dry and semi-arid areas. Implementing this model at large scales requires accurate spatial data, which can be obtained by combining ground data with remote sensing and applying geostatistical methods such as Digital Soil Mapping (DSM).
DSM, by integrating field data, remote sensing, and auxiliary variables, provides high accuracy in identifying spatial patterns of erosion and facilitates the production of maps with reduced uncertainty. The aim of this study is to improve the accuracy of soil erosion hazard assessment in the Damghan Rud watershed using the RUSLE model and integrating it with DSM and machine learning methods. This approach can help identify erosion-prone areas and enhance natural resource management.
.
2. Methodology
The present study was conducted in the Damghan Rud watershed area, located in the northwest of Damghan along the Damghan-Kiasar road, with an area of 1300 square kilometers. This region exhibits topographic diversity, including mountainous areas in the north and plains in the south, with an elevation range between 1390 and 1538 meters. The annual precipitation in the watershed varies between 57.75- and 249.22-mm. Soil sampling was performed using the Cubic Latin Hypercube Sampling (cLHS) method, and auxiliary variables such as Landsat 8 spectral bands, vegetation indices, and topographic features were utilized. In this study, 112 soil samples were collected from the depth of 0 to 30 cm, and physical and chemical analyses including clay, silt, sand, pH, EC, and organic matter content were conducted. To predict soil properties and the erodibility factor (K), Digital Soil Mapping (DSM) was employed using remote sensing data from Landsat 8, Digital Elevation Model (DEM), and Random Forest models. Model evaluation using statistical indicators such as R², RMSE, MAE, and ME showed that combining local data and remote sensing improved modeling accuracy and reduced errors. In estimating soil loss using the RUSLE model, the main formula A=RKLSCP was applied, where the rainfall factor (R) was calculated from the regional precipitation map and synoptic data, the soil erodibility factor (K) was determined based on the clay, silt, sand, organic matter content, and soil structure, the topographic factor (LS) was extracted using the DEM and empirical relations, and the vegetation cover factor (C) was assessed using the NDVI index and field data.

3- Results
The results of the present study in the region indicate that the factors of rainfall (P), rainfall erosivity (R), soil erodibility (K), slope length and steepness (LS), and vegetation cover (C) play a key role in the extent of soil erosion. The average annual precipitation is 285 mm, and the average rainfall erosivity index is 226 MJ mm ha⁻¹ h⁻¹ y⁻¹, increasing from east to west and from south to north. The soil in the region has a high sand content (52.93%) and low organic carbon content (0.42%), which increases its erodibility (with an average K value of 0.038). The region's topography, with an average slope of 12.96% and an average elevation of 2018 meters, plays an important role in the intensity of erosion. Vegetation cover, with a C factor ranging from 0.54 to 1 and an average of 0.92, is effective in reducing erosion in areas with adequate vegetation cover. Overall, the erosion mapping revealed that the erosion risk class exceeding 30 tons per hectare covers the largest area (40.97%), while areas with less than 4 tons of erosion per hectare are primarily concentrated in the central and lower parts of the watershed. These findings highlight the importance of sustainable soil management and conservation measures in the region.
4- Discussion & Conclusions
The present study was conducted to assess the potential of Digital Soil Mapping (DSM) in improving the accuracy of soil erosion hazard evaluation in the Damghan Rud watershed using the RUSLE model. The findings indicated that soil erosion is significantly influenced by various factors such as soil texture, topography, vegetation cover, and conservation practices. The RUSLE model estimated soil loss to be approximately 19.81 tons per hectare per year. However, when compared with the actual conditions of the region, it was observed that this model overestimates the erosion rates. The use of more accurate data, including vegetation cover percentage maps and gravel percentage, improved the model's accuracy.
The study confirmed the significant impact of high sand content, slopes exceeding 10%, and reduced vegetation cover on increasing soil erosion. Additionally, conservation practices such as terracing and cultivation along contour lines reduced the P factor and were effective in controlling erosion. The results demonstrated that the RUSLE model has a high capability in estimating erosion, and combining field data with Digital Soil Mapping (DSM) enhances the modeling accuracy. This approach is recommended for similar regions.