Determining Prone Areas to Rill Erosion Using Maximum Entropy Method (Case Study: Kaji Wetland Sourth Khorasan Province)

IntroductionSoil erosion is a serious threat to human well-being and life, especially in arid and semi-arid regions, and is one of the important issues in land management. Rill erosion is one of the most significant events in water erosion that affect soil loss, landscape, water resources, and land degradation can cause significant loss of soil in different climates. Identifying the effective processes that lead to the creation and expansion of rill erosion is necessary, and finding effective solutions to prevent rills is essential. In the meantime, one of the management solutions is determining the prone area to rill erosion. The high sensitivity of the lands of Nehbandan city (Kaji wetland watershed) to erosion is the reason for determining the prone areas to rill erosion. The maximum entropy method was used to identify the area that is susceptible to rill erosion. Material and MethodsThe modeling process used 9 effective factors, including height, slope steepness, slope direction, rainfall, land use, land cover, soil texture, geomorphology, and geology, based on similar research. Factors affecting the occurrence of rill erosion were analyzed as independent variables. The first step in preparing a rill erosion sensitivity map was to determine the location of rill erosions in the Kaji wetland watershed using Google Earth and then to conduct field surveys. The basin was monitored in the field using GPS and 138 cases of rill erosion were recorded. The occurrence points were divided into two groups: training and validation, with a 70:30 ratio. The total occurrence points were divided into 97 incident points that were randomly selected for model training (validation stage) and 41 incident points that were used for validation purposes. The MaxEnt model relied on the data set used for training as independent variables. In order to use the maximum entropy model to determine rill erosion, first the independent variables (factors affecting the occurrence of rill erosion) and the dependent variable (identification of points with rill erosion) was converted to the required format and introduced to MaxEnt software. To evaluate the effectiveness of the model in detecting occurrence points (rill erosion) from pseudo-non- occurrence points, the area under the ROC curve was used. The Jack Knife test was utilized to investigate the identification and prioritization of 9 influential factors (independent variables) that influence the results. The model was implemented using the remaining variables as input factors after removing the independent variables separately for this purpose. The efficiency of the model built using all independent variables was measured in comparison to the case where the model was built based on other variables. To determine its effect on the output, the contribution of the omitted independent variable was examined. Results and DiscussionAccording to the validation results, the sensitivity map for rill erosion has a high efficiency. The test stage should have a ROC curve of 0.859 (very good) and the test stage should have an average curve of 0.6 (moderate). The Jack Knife test revealed that the slope's steepness was the most significant environmental factor in the predicted sensitivity map for rill erosion in the study area. Geology and land cover were also recognized as other important factors. The MaxEnt model was found to be an effective model for preparing the rill erosion susceptibility map, according to the results. According to the findings, the slope steepness factor, which is 25%, is the main factor that affects the rill erosion of the Kaji wetland watershed. The high frequency of the slope class with a slope class below 20% suggests that this slope class is a significant factor in the development of rill erosion. The geological map of the region indicates that the majority of the region is dedicated to Quaternary formations, which is crucial for the development and creation of erosion in the region. The watershed's proneness to rill erosion is caused by poor rangeland usage. Management of vegetation and rangeland is necessary to reduce the potential for soil erosion in the region. According to the results, the soil texture of the region had less effect on the development of rill erosion; because most of the soil in the area is related to sand-gravel texture, which has a low effect on rill erosion. The MaxEnt model's high accuracy in modeling the sensitivity of rill erosion is evidenced by the results of the present study.