Simulation of Daily Evaporation Using Adaptive Neuro-Fuzzy Inference Systems (ANFIS) and Multivariate Regression (MR) IN Tabriz Synoptic Satation

Using empirical models for estimating evaporation requires a lot of variables that some of them can not be measured in the stations. Therefore, this study aimed to simulate the daily evaporation of Tabriz synoptic satation using meteorological data including average temperature of air (úc), wind velocity mean (m/s), relative humidity (%) and sun light hours by Adaptive Neuro-Fuzzy Inference Systems (ANFIS) and Multivariate Regression (MR) in the different architectures and input variables. After standardization of data, 85% of the data was used for network training and the efficiency of models (with indicators RMSE and R2 indicators) was coducted on testing data that included 15% of data.The results illustrated that the optimal model of ANFIS were obtained grid method (with three Gaussian membership functions) when one and two variables used as inputs and gained cluster method when three and four variables used as inputs. Adding relative humidity variable to the multivariate regression model didnot cause a significant changes in validation criterias of the training and testing data and also sun light hour's variable was excluded from the multivariate regression model. The results showed that ANFIS simulation compared to multivariate regression can increase the coefficient of determination of model to more than 10 percent, which requires using of cluster method and four input variables (avearage temperature of air, wind velocity mean, relative humidity and sun light hours).