The water resources are severely affected by hydrological cycle. Estimation of evapotranspiration which is the main component of the hydrological cycle plays an important role in water resources management. This phenomenon is non-linear and many factors affect on that and its estimation is very difficult. Various methods have been employed to estimate evapotranspiration although they have some limitations or problems. Some of these methods are costly and time-consuming such as lysimeters, and other empirical methods have local authority. Accordingly, applying a method that can be able to model the evapotranspiration regard to the nature of the gathered data and usage of minimum climate parameters is necessary. Nowadays, Artificial Neural Network (ANN) as a novel intelligent method are used in various sciences. In this study, the daily data of two climatological stations, namely Farokhshahr and Shahrekord airport in the interval of 2004-2013 including minimum temperature, maximum temperature, average relative humidity, sunshine, and wind speed at the height of two meters under different scenarios were utilized. Initially, empirical methods of reference evapotranspiration were approximated. The used empirical methods in this research have been Hargreaves, Blany Criddle, Priestley Taylor, and Jensen Hayes. The ANN model has been designed based on different scenarios of input data through MATLAB (R2012 b) Software. In this step, different ANN architectures were evaluated based on sensitivity and accuracy So, threshold functions such as tangent sigmoid and log sigmoid in hidden layers, linear function output layer were tested in topology where as Levenberg Marquardt employed as learning function. To evaluate the models, Penman Monteis FAO 56 model was employed. The statistical indexes, namely RMSE, MAE and R were calculated. Ten scenarios have been examined, and the results demonstrated that Scenario one with five parameters had the lowest error in comparision to FAO 56 technique. Furthermore, the perposed model show superior performance than empirical methods. However, between the empirical methods, Priestley Taylor and Hargreaves had better performance. On the other hand, the sensitivity analysis illustrated that the maximum temperature and wind speed had the greatest influence on reference evapotranspiration in these regions.