Assessment of Artificial Neural Network Models and Maximum Entropy in Zoning of Gully Erosion Sensitivity of Golestan Dam Basin

Zoning of gully erosion susceptibility and determining the factors controlling gully erosion is very important and vital. The aim of this study was to investigate the spatial distribution of gully erosion using two models of ANN and MaxEnt and to determine the factors affecting this type of erosion in Golestan Dam basin. Therefore, 14 factors in the form of three divisions, including topographic factors, other factors and combination of factors (14 factors) were considered as predictors of sensitivity. Out of 1042 gully erosion points, 30 and 70 percent were randomly classified as validation and test data, respectively. The results of Jackknife test showed that the parameters of height, rainfall and depth of valley are the most important variables affecting the prediction of gully erosion. The results of the modeling showed that the best accuracy of the model based on the ROC curve in the training model (0.923) and in the validation, stage (0.902) was the artificial neural network model, and this condition is achieved when all the factors in the modeling be involved. According to this model, about more than 20 percent of the domain (45633 ha) has a high sensitivity and is very susceptible to gully erosion.