CHSSDR-Updated: Daily Surface Solar Diffuse Radiation Dataset in China (1980-2022, 10km) based on REST2_v9.1 and Integrated Machine Learning techniques

Diffuse solar radiation is an essential component of surface solar radiation that contribute to photovoltaic power generation and carbon cycle. Based on REST2_v9.1 model and integrated machine learning model, we updated and constructed a national 43-year (1980-2022) daily diffuse solar radiation dataset (CHSSDR-updated) using ERA5 and MERRA2 reanalysis data, with a spatial resolution of 10 km. Compared with the previous diffuse solar radiation dataset, the improved dataset effectively improves the accuracy of diffuse solar radiation by introducing physical model, improving machine learning models and optimizing model inputs. The validation results with ground-based measurements showed that the CHSSDR-updated dataset had a high and stable performance with the correlation coefficient (R), the root-mean-square error (RMSE), and the mean absolute error (MAE) for the sampled-based cross-validation of 0.943,13.901 Wm-2 and 10.466 Wm-2, which are increased by 0.063 and decreased by 5.639 Wm-2 and 4.404 Wm-2.