• (1) This study provides the first estimation of hourly-scale grid- sunshine duration (SD). • (2) Estimating incident solar radiation (R s ) for the first time using hourly grid-SD. • (3) Look-up table (LUT) generated in conjunction with the 6S model. • (4) The effect of environmental factors to R s is quantitatively assessed. High-spatiotemporal-resolution sunshine duration (SD) data has consistently been a significant concern in critical domains and provides strong support for estimating incident solar radiation (R s ). To achieve highly precise hourly estimations of both SD and R s , this study developed a ‘multimodal hourly SD model to SD-based hourly R s model’ framework using FY-4 and Advanced Himawari Imager (AHI) data spanning 2019–2024. Gradient Boosting Decision Tree (GBDT) classification model combined with transformer regression model was constructed for hourly SD model, and a radiation transfer empirical model based on grid SD data was constructed for grid SD-based hourly R s model. The models were trained, tested, and validated using ground measured data from Chinese Meteorological Administration (CMA) stations. Validation results demonstrated high performance: the estimated hourly SD showed a correlation coefficient (R) of 0.96 and root mean square error (RMSE) of 0.11 h, while the estimated hourly R s achieved an R of 0.91 and an RMSE of 97.31 W/m 2 against ground-measured data. Furthermore, an analysis of the impacts of climatic conditions on SD-based hourly R s model showed that water vapor content and temperature were positively correlated with both R and RMSE. Our methodology offered a new perspective for hourly SD and R s inversion, which combined multi-source geostationary satellite remote sensing and ground-based observation data.
Zhang et al. (Fri,) studied this question.