Predictive Modeling of Daily Evapotranspiration in Arid Regions Using Artificial Neural Networks

Abstract

For thousands of years, people have understood the importance of evapotranspiration (ET) in maintaining the hydrologic cycle and replenishing the world's freshwater supplies. The process of estimating evapotranspiration with a high accuracy in arid regions is considered one of the most important processes in hydrological studies. It has a great importance in the efficient management of water resources and hydrological modeling, as well as in the management of irrigation operations, because these areas are permanently linked to the issue of water scarcity. The prediction of evapotranspiration is a vital step towards management of water resource. This study aims to develop an artificial neural network model to predict the evapotranspiration in arid regions. The RBFNN and GRNN models with six input data were used in present study. The input data are Max. temperature, Min. temperature, Ava. Temperature, Humidity, Wind speed and Solar radiation. The ANN modelling was achieved by using MATLAB with hyperbolic sigmoid transfer function for both input and output layers. Several statistical indicators have been used for examining the model's prediction accuracy. Results show that the current model is a powerful model which has the capability to predict the evapotranspiration with high accuracy. The superiority of the GRNN model is very obvious in comparison to the RBFNN model, where the coefficient of determination for GRNN model was more than 96% in comparison with 94% for RBFNN and the mean square error for GRNN was 0.4 in comparison with 0.52 for RBFNN model.