| Abstrakt: |
Introduction and Objective: Gully erosion is one of the factors threatening the environmental balance and its stability and plays a prominent role in increasing the sedimentation capacity of watersheds and changing the hydrological characteristics of surface streams. The purpose of this research is to evaluate the effectiveness of the maximum disorder machine learning model in order to prepare a gully erosion sensitivity map in the Semnan watershed. Identifying the most important environmental factors affecting the occurrence of gully erosion using the jackknife method and examining the importance of each environmental factor in the study area by analyzing the response curves are other goals of this research. Material and Methods: This research consist of five main steps: the first step is choosing the study area, collecting and preparing maps of the effective factors, the second step is preparing the distribution map of the ditch erosion event, the third step is the multiple collinearity test with the index of the tolerance coefficient and the variance inflation factor in order to check the information overlap of the effective factors. and checking the importance of the factors, the fourth step is to implement the maximum irregularity or maximum irregularity model, prepare a zoning map of sensitivity to the event of gully erosion and classify it into five categories: very low, low, medium, high and very high, the fifth step is to evaluate the classification accuracy and validate the zoning map and Prediction of sensitivity to gully erosion. Results: According to the results, among the 23 initial factors or variables, the factors of the waterway slope length index, surface curvature, and waterway power index had colinearity or overlapping information, and in the next stages, they were removed from entering the modeling process, and modeling was carried out with 20 independent factors or variables. The gully erosion sensitivity map of the studied area showed that the southern and southern parts of the Semnan watershed are prone to gully erosion. Also, the zoning results obtained from the implementation of the maximum irregularity model indicate that the areas prone to gully erosion are in high areas, sensitive rock units (quaternary clay and marl zones), average annual rainfall, Eridesville soil type, ultra-arid climate, high drainage density classes, low slope, low lands without geographical slope, high topographic moisture index, pasture land use and low land surface texture are observed. Also, the maximum irregularity model has 91% and 89% accuracy in the calibration and validation stages, respectively, which in terms of efficiency is in the very good category for predicting gully erosion prone areas. Conclusion: About 35% of the area of high and very high sensitivity areas resulting from the implementation of the maximum irregularity model include more than 90% of the ditches in the region. Also, the maximum irregularity model in the stages of implementation or development and prediction or validation with the area under the receiver operating characteristic curve with the values of 0.91 and 0.89, respectively, is effective in zoning and predicting the occurrence of gully erosion. This result will be useful for local managers and planners as well as executive experts in order to identify areas prone to gully erosion and determine the best implementation methods of watershed operations for soil protection approaches. [ABSTRACT FROM AUTHOR] |
