Soil erosion is one of the most important environmental issues in developing countries, including Iran that there is inaccurate information about its amount and distribution. For this purpose, the accuracy and distribution of erosion classes obtained from EPM and MPSIAC models as compared to BLM as ground truth values were evaluated in Shahriari watershed. First, the required data and information for MPSIAC, EPM and BLM models for basin erosion status was stored in a database. Then, using ARC/GIS9.3 software, erosion class maps were prepared based on each model formula and they have been intersected with the ground-truth map. Also a point map containing 1400 random points was also prepared for sampling the maps obtained from the models. Then, based on random sampling points map, each of the statistical indices of RMSE, RRMSE, MAE, MBE, NSE, CD, CRM and MSE were calculated. Also, to verify the spatial accuracy of the erosion class maps derived from the error matrix models and the Kappa coefficient of agreement, overall accuracy, producer accuracy and user accuracy. The results of the erosion of the models showed that the mean erosion intensity in MPSIAC model was 589 and in EPM model 287 m3/ km2.year that the basin erosion status is estimated to be moderate for the basin with the help of MPSIAC model and EPM model. The analysis of the above statistical indices shows that the RSME, MAE, MBE and MSE statistics in MPSIAC model have lower values. As a result, the MPSIAC model has less error than the EPM model. Overall, the results of the statistical indices showed that the MPSIAC model's erosion classes were more in line with the BLM output (as a ground truth map) than the EPM model. Also the kappa coefficient in MPSIAC is higher than in EPM model. Overall, the results show that the MPSIAC model erosion class maps overlap with the BLM model and in fact with the actual conditions of the area.
