Surface Daytime Net Radiation Estimation Using Artificial Neural Networks

摘要: Net all-wave surface radiation (Rn) is one of the most important fundamental parameters in various applications. However, conventional Rn measurements are difficult to collect because high cost and ongoing maintenance recording instruments. Therefore, empirical estimation models have been developed. This study presents results two artificial neural network (ANN) (general regression networks (GRNN) Neuroet) estimate globally from multi-source data, including remotely sensed products, measurements, meteorological reanalysis products. estimates provided by ANNs were tested against in-situ obtained 251 global sites between 1991–2010 both mode (all data used fit models) conditional (the divided into four subsets fitted separately). Based on extensive experiments, it has proved that superior linear-based modes GRNN performed better was more stable than Neuroet. The had a determination coefficient (R2) 0.92, root mean square error (RMSE) 34.27 W∙m−2, bias −0.61 W∙m−2 based validation dataset. concluded ANN methods potentially powerful tool for estimation.