This study assesses gully erosion susceptibility in Qara Toureh village, Bijar County, Kurdistan province, utilizing three methods: frequency ratio (FR), logistic regression (LR), and random forest (RF). A compar-ative analysis of these methods is performed to determine their effec-tiveness. Field surveys recorded the spatial coordinates of 157 gully points around the village. In the frequency ratio method, statistical cal-culations in Excel were used to assign weights and values to various factors, followed by raster calculations to create a susceptibility map for gully erosion. Logistic regression analysis identified five key factors in-fluencing erosion: land use, aspect, distance to stream, stream density, and soil type. These were selected from fourteen potential variables based on their statistical significance. The random forest algorithm, im-plemented in R and integrated with ArcGIS 10.8, was used to generate a gully erosion map, evaluating the importance of each factor. This analy-sis revealed that lithology, slope, land use, stream density, and soil type were the most significant predictors of erosion, while soil depth, rain-fall, distance to fault, and fault density were less influential. Model ac-curacy was assessed using the Receiver Operating Characteristic (ROC) curve. The frequency ratio method achieved the highest area under the curve (AUC) at 0.91, indicating superior predictive accuracy. The lo-gistic regression and random forest models also performed robustly, with AUC values of 0.87 and 0.84, respectively. The current research results provide a straightforward, rapid, and accurate approach to ero-sion mapping, offering significant benefits to environmental planners and decision-makers in addressing erosion-related challenges, such as land use planning and disaster risk reduction.
