Performance evaluation of different cloud products for estimating surface solar radiation

Abstract

The presence and variability of clouds have a significant effect on surface solar radiation (SSR). The range of cloud products currently available for SSR estimation vary in spatial and temporal resolution and accuracy. Since the effect of different cloud products on the accuracy of SSR estimation has not been adequately quantified in existing studies, this study evaluates the performance of four cloud products (Himawari-8, ISCCP, CERES, and MERRA-2) in estimating SSR and analyzes them in comparison with the MODIS cloud product. The accuracy of SSR estimation of the four cloud products is verified using measured data from BSRN and CERN ground-based observatories. The results show that Himawari-8 has the best performance with R-squared (R²) values of 0.94 and 0.74 and root mean square errors (RMSE) of 71.03 W/m² and 141.36 W/m² on the sub-daily scales at the BSRN and CERN sites, respectively. CERES and ISCCP have similar performances, but they vary by site and month. While MERRA-2 grossly underestimates SSR, probably related to misclassification of clear skies as cloudy and overestimation of cloud optical thickness under cloudy conditions. Compared to MODIS, Himawari-8 provides better agreement with MODIS results in cloud classification and cloud phase identification, while CERES provides better agreement with MODIS results in cloud optical thickness. Overall, Himawari-8 performs best in SSR estimation. This comprehensive assessment not only highlights the crucial role of cloud observations on SSR estimations but also details the strengths and weaknesses of each cloud product in enhancing the understanding of solar radiation dynamics.