Transfer learning unlocks FY-4A satellite for high-accuracy solar radiation data

A groundbreaking study has unlocked the potential of China's Fengyun-4A (FY-4A) geostationary satellite to accurately estimate surface solar radiation (SSR) and its global, direct, and diffuse components. Researchers from the Aerospace Information Research Institute, Chinese Academy of Sciences; Sichuan University of Science and Engineering; and the Institute of Atmospheric Physics, Chinese Academy of Sciences, reported (DOI: 10.34133/remotesensing.1044) in the Journal of Remote Sensing on April 29, 2026. The team developed a novel transfer learning framework that adapts knowledge from the Himawari-8-based Cloud, Atmospheric Radiation and Renewal Energy Application (CARE) product to FY-4A, enabling high-accuracy SSR estimates without heavy reliance on auxiliary meteorological datasets. This innovation addresses the critical need for reliable solar radiation data, which is essential for solar power forecasting, climate research, land-surface modeling, and sustainable energy planning.

The study's core advance lies in a deep neural network (DNN) model pretrained on Himawari-8 Level 1 data and the CARE product, then fine-tuned with FY-4A data. Using top-of-atmosphere reflectance and solar-satellite geometry as inputs, with Bayesian optimization for hyperparameter tuning, the model achieved impressive validation results across 33 ground stations from the Baseline Surface Radiation Network (BSRN), Bureau of Meteorology (BOM), and Global Tropical Moored Buoy Array (GTMBA) during 2018–2020. At representative BSRN sites, instantaneous root mean square errors (RMSEs) were 102.2, 117.5, and 83.1 W m⁻² for global, direct, and diffuse radiation, respectively. Daily mean RMSEs dropped to 28.5, 30.1, and 22.6 W m⁻², demonstrating robust performance across temporal scales.

The authors emphasized that this framework turns FY-4A into a powerful resource for energy and climate applications by separately estimating direct and diffuse solar radiation, which are crucial for solar photovoltaic (PV) and concentrating solar power systems. Direct radiation is key for concentrating solar power, while diffuse radiation affects PV output under cloudy or hazy conditions. By reducing dependence on auxiliary data, the method is more practical for near-real-time monitoring. Looking ahead, the same transfer learning strategy could be extended to other Chinese geostationary satellites like Fengyun-4B (FY-4B), enhancing solar-energy monitoring across East Asia and beyond. This research provides a stronger data foundation for PV site assessment, power forecasting, grid management, and climate modeling, driving the clean-energy transition forward.