Reference evapotranspiration estimation using artificial neural network for Tarai region of Uttarakhand (India)

Reference evapotranspiration (ET0) prediction is required in irrigation water management for the calculation of crop water requirement and its scheduling. Reference evapotranspiration (ET0) is however, a complex and non-linear phenomenon because it depends on several interacting climatological factor, such as temperature, humidity, wind speed and sun shine hours. Artificial neural network (ANN) is effective tool that is able to solve accurately complex and non-linear input-output relationships. This paper has used feed-forward multilayer artificial neural networks of single and two hidden layer(s). These networks were trained with fixed iteration 1000 and tested for the prediction of daily evaporation for Tarai region of Uttarakhand (India). The daily temperature, relative humidity, wind velocity, sunshine hour of current and previous two day as well as evaporation of previous two days as input variables and current day evaporation as the output variable, thus there were 17 input nodes in the input layer and one node in the output layer in ANN model development. The daily meteorological data during period 1/1/2001 to 9/8/2007 were used for the training (calibration) of the model, whereas the data of years 10/ 8/2007 to 31/12/2011 were used for verification (testing) of the developed model. Training was conducted using back-propagation algorithm. ET0 estimation performance of the ANN models was compared with most reliable and standard FAO-56 PM method. This study reveals that the use of ANN model resulted better performance for estimation of reference evapotranspiration (ET0). The coefficient of correlation values between predicted ET0 from ANN model and calculated from Penman-Monteith model was obtained as 0.9527. Based on this result, it can be concluded that the ANN can predict ET0 better than conventional methods.