An Artificial Neuron Network With Parameterization Scheme for Estimating Net Surface Shortwave Radiation From Satellite Data Under Clear Sky—Application to Simulated GF-5 Data Set

Net surface shortwave radiation (NSSR) is a key parameter that drives the material exchange and energy balance. Herein, we propose an improved artificial neuron network (ANN) with parameterized (ANN-P) method to first calculate albedo at top of atmosphere (TOA) by considering non-Lambertian effect. Subsequently, NSSR estimated based on relationship between TOA broadband Earth's surface-absorbed using under clear sky. The modeling process implemented Chinese Gaofen-5 (GF-5) visible/near-infrared channels data simulated via MODTRAN. For comparison, previously reported lookup table (LUT) (LUT-P) ANN are also employed. performances all these methods evaluated. In terms model simulation part, root-mean-square errors (RMSEs) 15.01 (17.07), 10.04 (13.67), 20.39 (29.99) W/m 2 for land, water, snow/ice surfaces, respectively, ANN-P (versus LUT-P) method. Their mean bias (MBEs) within 0.9 . With respect direct method, it shows highest accuracy yet relatively large deviation water surface. Additionally, sensitivity analysis vapor content (WVC) confirms more stable than LUT-P is, thereby, recommended clear-sky estimation. Finally, ground validations indicate RMSEs LUT-P, ANN-P, 49.33 (-3.01), 47.55 (1.75), 104.24 (-75.72) , respectively.