To evaluate the bearing capacity of water resources objectively, a PNN (probabilistic neural network)-based model was established in line with the evaluation criteria for rational water allocation in water-abundant areas. The model was applied to evaluate the rationality of water allocation in different target years in Wenshan prefecture. Results showed that the water allocation in different target years was between level 3 - level 7, which suggests a basically-rational to rational water allocation. It also indicated that the fairness of water allocation, the efficiency of water consumption and the coordination between utilization and supply needs to be improved. The results largely reflect the status quo and medium-to-long-term condition of water allocation in the prefecture. Moreover, the PNN-based model is superior to the error backpropagation model in terms of classification precision. The model is simple and feasible, with short operation time and no local optimum value, thereby could effectively evaluate the rationality of regional water allocation.