Determination of the Best Adaptive Neuro-Fuzzy Inference System (ANFIS) Model for Estimating Grass Reference Crop Evapotranspiration in Coastal Semi-arid Climate of Hormozgan

Accurate estimation of reference crop evapotranspiration (ETo) plays an important role in water resources management and planning in dry regions. In this study, accuracy and ability of Adaptive Neuro- Fuzzy Inference System (ANFIS) in estimating ETo was evaluated. Daily meteorological data, including air temperature, relative humidity, sunshine hours, vapor pressure deficit, wind speed and solar radiation ofMinab synoptic station, Hormozgan province, during 2006 to 2011 were used for modeling. The evapotranspiration values estimated by the FAO-56 Penman-Monteith equation (PM) were considered as the reference values for calibrating ANFIS model. The performance of the developed model with different input combinations was also compared with the empirical models, namely, Hargreaves-Samani (HS) and Blaney- Criddle (BC). The root mean square error (RMSE), mean absolute error (MAE) and the coefficient of determination (R2) were used for comparing the results of ANFIS, HS and BC methods with reference method (FAO-56 Penman- Monteith equation). The results showed that the ANFIS was a more appropriate method for estimating the ETo inMinab and this model with 6 inputs (with 3 membership functions and Gaussian mixture model) had a better performance than the other considered methods with the R2, MAE and RMSE values of 0.99, 0.03 (mm day-1) and 0.04 (mm day-1), respectively. Also, the ANFIS model with 2 inputs (with 3 membership functions and Gaussian mixture model) was the best model for the stations which had only the measured temperature data.