Simulation and Comparison of Potential Evapotranspiration by Artificial Neural Networks, ANFIS (Fuzzy Neural Network) and Decision Making M5 (Case Study; Synaptic Station of Shiraz)

The proper estimation of evapotranspiration in designing, managing irrigation and drainage systems is very important. One of the methods of estimation of evapotranspiration, which is widely used in solving these problems and its prediction, are Neuro-Fuzzy Methods (ANFIS), Artificial Neural Networks (ANNs) and decision making tree M5. The purpose of this study was to evaluate the efficiency of the mentioned methods in estimating the reference evapotranspiration in the Shiraz meteorological station. For this purpose, the 5 yearly climatic data of the station were selected as inputs of the models. To implement artificial neural network model, Nero fuzzy model and decision tree M5 were used respectively from Qnet2000, MATLAB and WEKA software. In order to evaluate the results of these models, the mean squared error (RMSE), coefficient of determination (R2) and the criterion of the mean power of relative error (MAE) were used. The results of Artificial Neural Network model and ANFIS model with the help of statistical indices R2, RMSE and MAE were 0.0999, 0.099, 0.0500 and 0.0999, 0.051, and0.01119, respectively the accuracy of both models in simulation is high. Also, the correlation coefficient (R2), RMSE and MAE of decision tree model were calculated to be 0.7064, 0.0935 and 0.0414 respectively, which indicates the proper performance of the M5 tree model in predicting the reference evapotranspiration rate.