Estimation Downward Longwave Radiation Using Neural Networks in Baghdad City

Abstract

Abstract The aim of the present study is to compare the accuracy and precision of the traditional method and artificial neural network (ANN) to estimate downward longwave radiation (DLR) under clear-sky conditions. DLR is useful for agriculture, global warming detection, and ecology. In order to train and estimate DLR in Baghdad, Iraq, an artificial neural network (ANN) model was developed. The activation functions (Sigmoid, Hyperbolic secant, Hyperbolic tangent), hidden layers (1, 2, 3, and 4), alteration 10000 to 50000 with interval 10000 were employed for both hidden and output layers. ANN best model was compared to eight other models based on root mean square error (RMSE) and correlation coefficient (R2). The comparative statistic for performance of the DLR model calculations during day-time and night-time have shown that the parameterization has the best results compared to the measured data. It was found that the optimum model is the (105) model where R and RMSE were 0.999 and 0.6 respectively. The results in this study indicate that the ANN can successfully be used for the estimation of DLR for Baghdad city and is more suitable than the traditional method.