Simulation and comparison of runoff estimation by neural network methods, regression and fuzzy inference(A case study:Dez catchment)

Due to prolonged droughts in the recent decades. The importance of predicting the flow rate of surface water in rivers for water resources management increases. In this regard the flow rates in the natural water ways. Which is the most important supplement source for water in dam storages. are considered as the most vital factors in predicting surface water. To evaluate the accuracy of different models have been used runoff data of thirteen stations without trend in dez catchment. In this study, artificial neural network, support vector regression and adaptive neuro-fuzzy inference system methods with clustering and grid partitioning approaches were used to the simulation of run-off in dez catchment. Simulation results of runoff using the mentioned methods were compared using the statistical indicators of R, RMSE and NSE. Comparison between ANN, ANFIS and SVR showed that although the difference in the accuracy of the models is very small. It can be said that all three models have acceptable answers. The results also show that the ANN and ANFIS with clustering approach models with NSE=0.66 and 0.66 respectively during the test period have the ability to simulate two models SVR and ANFIS with grid partitioning approach.