Soil Erosion Susceptibility Mapping in Kozetopraghi Catchment, Iran: A Mixed Approach Using Rainfall Simulator and Data Mining Techniques

摘要: Soil erosion determines landforms, soil formation and distribution, fertility, land degradation processes. In arid semiarid ecosystems, is a key process to understand, foresee, prevent desertification. Addressing throughout watersheds scales requires basic information develop control strategies reduce degradation. To assess remediate the non-sustainable rates, restoration programs benefit from knowledge of spatial distribution losses maps erosion. This study presents Support Vector Machine (SVM), Random Forest (RF), adaptive boosting (AdaBoost) data mining models map susceptibility in Kozetopraghi watershed, Iran. A inventory was prepared field rainfall simulation experiments on 174 randomly selected points along watershed. previous studies, this has been using indirect methods such as Universal Loss Equation Direct measurements for mapping have so far not carried out our site past. The rate generated by simulated 1 m2 plots at 40 mmh−1 used map. Of available data, 70% 30% were classified calibrate validate models, respectively. As result, RF model with highest area under curve (AUC) value receiver operating characteristics (ROC) (0.91), lowest mean square error (MSE) (0.09), most concordance differentiation. Sensitivity analysis Jackknife IncNodePurity indicates that slope angle important factor within showed areas located center near watershed outlet can be support development sustainable plans more accuracy. Our methodology evaluated also applied other regions.