Evaluation of the Accuracy and Trend Consistency of Hourly Surface Solar Radiation Datasets of ERA5, MERRA-2, SARAH-E, CERES, and Solcast over China

The global ambition to achieve carbon neutrality by the mid-21st century is driving a transition towards clean energy. Accurately assessing solar energy potential necessitates high-quality observations of hourly surface solar radiation (SSR). The performance of hourly SSR data from two reanalysis products (ERA5 and MERRA-2) and three satellite-derived products (CERES, SARAH-E, and Solcast) is validated against 22 years of continuous surface observations over 96 stations across China. The accuracy (in %) and trend consistency (in % decade−1) of estimates from gridded products in reproducing the diurnal cycle and trend of SSR are generally lower at sunrise and sunset than at noon, and they are also reduced in the cold season (October to next March) compared with the warm season (April to September). Regionally, accuracy is generally lower in the southwestern plateau region, and the trend consistency of most products is lowest in the rugged and cloudy southern part of China. Among the evaluated datasets, Solcast and MERRA-2 exhibit the highest accuracy and trend consistency in capturing the diurnal pattern of SSR, respectively, while CERES demonstrates the best overall performance.