Global evaluation of simulated surface shortwave radiation in CMIP6 models

Surface solar radiation (SSR) is the main energy source of Earth system, and has direct influence on climate change, the hydrological cycle, and the development of renewable energy technologies. In this study, we compared simulated monthly SSR profiles generated from 62 Coupled Model Intercomparison Project Phase 6 (CMIP6) models against natural SSR observational data collected at 226 GEBA sites during 1990–2014. The accuracy with which CMIP6 models simulated spatial features of SSR was further assessed using ISCCP-ITP satellite data. Our results show that most CMIP6 models overestimate the monthly SSR averaged over all GEBA sites, with an averaged mean bias error (MBE) of approximately 7.8 W/m2. The root mean square error (RMSE) for individual models ranged from 26 to 38 W/m2, and the RMSE for their ensemble mean decreased by an average of 8.1 W/m2. Averaging of the multi-model ensemble led to a significant improvement in the correlation coefficient, which increased from 0.90 to 0.96. Ground-based data showed that the accuracy and annual trend for the CMIP6 multi-model mean (MMM) simulations were more accurate than any individual model. The spatial distribution of CMIP6 MMM multi-year average SSR values agrees well with ISCCP-ITP satellite data retrievals; however, the overall model performance is overestimated in most locations on Earth (e.g., the southern Sahara Desert, the land along the Mediterranean Sea, and the Yangtze River Delta plain), but underestimated in some others (e.g., the Rocky Mountains, the Andes Mountains, Indonesia). The results of this study will allow more accurate predictions to be made of future SSR data produced by CMIP6 models.