GIS‐ and RS‐based models for gully erosion susceptibility mapping using machine learning and remote sensing data

Abstract In Northern Pakistan, gullying‐related soil erosion results in severe soil degradation, detrimentally affecting the region's socioeconomic situation and environment. Gully erosion susceptibility mapping (GESM) is most effective at mitigating the negative impacts of soil erosion. Gully management strategies begin by creating suitable evaluation tools, identifying the causes of the gully and devising solutions for controlling it. This study aimed to evaluate the performance of two models, frequency ratio (FR) and information value (IV), to assess the soil erosion susceptibility of the study area. In the present study, we detected and identified 1287 inventory points of gully erosion using various satellite images, including Google Earth Pro and field surveys. This research considered topographic, geologic and climatic factors as causative parameters. The FR illustrated that a low class of soil erosion covered 21.4% of the area, and medium, high and very high classes covered 34%, 31.4% and 13.2%, respectively. The results of the variables using IV models showed that the areas covered by low, medium, high and very high classes are 28.5%, 37.6%, 15.2% and 18.7%, respectively. The models were validated using the AUC technique, revealing that the success rate curve (SRC) and predicted rate curve (PRC) for IV are 80% and 89%, respectively, whereas the SRC and PRC for FR are 86% and 93%, respectively. The validation results show that the FR model outperformed the IV model. This research is useful for assisting decision‐makers and policymakers in devising a strategic plan to mitigate the adverse impacts of soil erosion in this important geographic region.