چكيده لاتين
Soil erosion is a major environmental challenge worldwide, leading to land degradation, reduced fertility, and increased sedimentation. Gully erosion, as a specific type of water erosion, causes significant environmental and economic damage through the formation of deep gullies and channels. This study aimed to identify gully erosion susceptibility areas and compare the performance of different data mining models in the Mian-Ab watershed, Khuzestan Province, Iran. Initially, the locations of existing gullies were recorded using satellite imagery, Google Earth, field work and GPS. Subsequently, 19 parameters influencing gully erosion were extracted, including topographic indices (slope, aspect, elevation, LS factor, and terrain roughness), vegetation indices (NDVI, SAVI, RVI, TGSI), soil texture, distance to roads, land use, lithology, soil erodibility, and topographical indices (topographic wetness index, stream power index (SPI), drainage density, rainfall erosivity, and distance to rivers). Four models, namely CART, MAXENT, SVM, and XGBoost, were employed for gully erosion susceptibility mapping. A 70/30 split was used for training and validating the models. Performance was evaluated using MSE, MAD, RMSE, and the coefficient of determination (R²). The results indicated that the XGBoost model achieved the highest accuracy with an R² of 0.902 and a correlation coefficient of 0.95. The MaxEnt model with an AUC of 0.909, followed by SVM and CART, showed relatively lower performance. Based on the findings, land use, vegetation cover, and soil texture were the most influential factors in gully erosion occurrence and development. High-risk areas for gully erosion were primarily concentrated in the eastern and southeastern parts of the watershed, comprising 4.11% of the total area.
