Estimation of Daily Pan Evaporation By Using MLP, RBF and Recuurent Neural Networks

Evaporation is one of the important parameters in hydrology and water resource engineering which is attractive for researchers. Due to interaction between various meteorology parameters, the estimation of evaporation is complicated. There are some nonlinear relations for assessing the evaporation. Artificial neural networks are innovative approaches for estimation and prediction by using learning concept. In this study, by using the daily data of Hashem Abad synoptical station in gorgan, the multilayer perceptron networks, radial basis function networks and element neural networks have been used. The various combinations of daily climatic variables, i. e. air temperature, and duration of sunshine, wind speed and humidity were selected as input parameters and evaporation was used as output parameter. A comparison is made between the estimated provided by ANN models and experimental existence relations as well. To evaluate the effective input parameters on evaporation a sensitivity analysis was used. The results shows that air temperature and wind speed have the maximum and minimum effect on estimation of daily evaporation. The results also show that ANN models perform better than the available experimental relations.