Daily evapotranspiration estimation using artificial neural networks and classicial methods

Evaporation is regarded as an important parameter of hydrological and meteorological studies. Correct evaporation estimation is crucial for various purposes such as development, control and management of water resources. In this study, daily evaporation estimation (ET_0), has been made according to Penman-Monteith method recommended as the standard method by FAO (Food and Agriculture Organization), and evaporation estimates found according to Penman Monteith method were accepted as reference. The evaporation amount by comparing the performances of Artificial Neural Networks (ANN), Radial Based Artificial Neural Networks (RBNN), Multiple Linear Regression (MLR) methods and Auto-regressive model AR (p) with empirical methods such as Hargreaves-Samani, and Turc equations has been estimated.A station in Hartwell lake in Anderson region, South Carolina (USD) was chosen as the study area. Average daily air temperature (T_mean), highest (T_max), and lowest daily air temperatures (T_min), wind speed (u), sunshine amount (SR) and relative humidity (RH) were used for the estimation of the average daily evaporation amount. All daily data are divided into training and test data. The feedforward-back-propagation ANN model working according to the principle of back propagation has been used for the optimization of ANN. ANN, Radial Based Artificial Neural Networks (RBNN), Multiple Linear Regression (MLR) method and Auto-regressive model AR (p) results were compared with the results of traditional Hargreaves-Samani, and Turc method. The comparison has shown that the ANN model performed better than other methods in estimating the evaporation amount.