Reference Evapotranspiration (ETo) is one of the key components of the hydrological cycle that is effective in water resources planning, irrigation and agricultural management and, other hydrological processes. Accurate estimation of ETo is valuable for various applications of water resource engineering, especially in developing countries such as Iran, which has no advanced meteorological stations and lacks facilities and information. The present study investigates the ability of the hybrid artificial neural network- Gray Wolf Optimization (ANN-GWO) model to estimate ETo for Iran. The accuracy of ANN-GWO was evaluated versus least square support vector regression (LS-SVR) and standalone ANN. The development of models is based on meteorological data of Iran’s 31 provinces consists of 5 different climates. Based on the Shannon entropy and correlation, seven different input scenarios were introduced and Penman-Monteith reference evapotranspiration was considered as the output of the models. Several statistical indicators including SI, MAE, U95, R2, Global Performance Indicator (GPI), and Taylor diagram were used to evaluate the performance of the models. The results showed that the GWO algorithm acted as an efficient tool in optimizing the structure of the ANN and the ANN-GWO model was more accurate than ANN and LS-SVR in all scenarios. ANN-GWO6 with inputs of wind speed, maximum and minimum temperatures, had the lowest error and decreased in terms of SI index by 42% (compared to ANN6) and 30% (compared to LS-SVR6). Furthermore, based on GPI, it is in the first place with a 99% reduction, compared to ANN6 and SVR6. Finally, ETo interpolated maps were plotted for Iran based on the values estimated by ANN-GWO6. The hybrid approach used in this study can be developed as a trustful expert intelligent system for estimating ETo in Iran.
