توصيفگر ها :
الگوريتم ژنتيك , بهينهسازي , جنگل تصادفي , ماشين بردار پشتيبان , منحني عملكرد نسبي
چكيده انگليسي :
Soil erosion is one of the main problems, especially in agriculture, natural resources and environment in Iran. Gully erosion is one of the most important water erosion which leads to waste of water and soil resources. For preventing or combating it needs to determine the areas with risk of gully. In present study, the potential of gully erosion in the Semirom basin in the south of Isfahan Province were evaluated using the intelligent methods of Random forest, Support vector machine and optimization of these two models by Genetic algorithm. For this purpose, at first the position of the gully was prepared using field observations, satellite imagery, topographic maps and geographic information systems, and from 311 identified gullies, 70% (218 gullies) and 30% (93 gullies) were selected randomly for modeling and validation respectively. In the next step, 21 controling factors in the gully occurrence in the study area including altitude, slope angle, slope aspect, plan curvature, profile curvature, curvature, hillshade, Convergence Index, lithology, land use, Normalized Difference Vegetaion Index, distance from channel, Topographic Wetness Index, Stream Power Index, slope length factor, drainage density, rainfall, vertical distance to chanel network, total catchment area, soil type and soil depth were determined for modeling and digited in GIS and SAGA softwares. Ultimately, the gully erosion sensitivity map was prepared using RF, SVM and GA in the case study and then performance of the models was evaluated using the receiver operating characteristic and area under curve. Area under curve for test data was calculated as 0.801, 0.807, 0.888 and 0.800 for RF, SVM, RF-GA and SVM-GA models, respectively. According to the results, in general, intelligent models used are good for predicting gully potential areas. Also, results showed that among the models used, RF-GA model, has the best performance.
