Abstract: In order to protect water and soil resources and the environment, evaluate the regional soil erosion intensity, this paper took the Lalin River Watershed of Heilongjiang black earth region as the studied area. Slope angle, aspect, soil types, landuse and normalized vegetation index were chosen as evaluation indexes, and the logistic regression model and generalized regression neural network model were adopted to assess soil erosion intensity with the ArcGIS platform. Then, the results of two models were evaluated by the receiver-operating characteristic curve (i. e. ROC curve) . Analyses results showed that areas under the curve (i. e. AUC) of logistic regression model and generalized regression neural network model were respectively 0.857 and 0.881, consistent with the actual soil erosion intensity. The evaluation results of two models were established for mutual and the results of generalized regression neural network model were more accurate.