Performance evaluation of Dempster-Shafer model for erosion potential mapping in Kakareza watershed, Lorestan province

Identifying susceptible areas for erosion can be considered as one of the most important soil conservation measures. In this study, the capability of Dempster-Shafer (DS) model for mapping potential areas for erosion was investigated in Kakareza watershed in Lorestan province. First thematic layers of influential factors in soil erosion, including altitude, slope, slope, aspect, plan curvature, lithology, land use, distance from the river, soil and topographic wetness index were prepared. In addition, 29 eroded positions in the study area that their positions were obtained from GPS and Google earth on 10 July 20016 were mapped and then were divided into a training (70%) and testing (30%) points. The layers of environmental variables were classified into different classes according to and then based on the density of eroded points in the study area and DS analysis, the weight of each class was determined and the potential map of vulnerable areas to erosion was obtained according to the DS model. The accuracy of a generated map was also investigated using testing points and receiver operating characteristic (ROC) curve. The result showed that the produced map has the success rate of 21%, that means the poor capability of the DS model for mapping susceptible areas of erosion. In addition, according to the DS map, areas with the highest potential to surface erosion are located in the central and eastern part of the study area. Therefore, it can be indicated that this model has a poor ability in identifying potential and vulnerable areas to surface erosion compared to other phenomena such as flood and gully erosion.