Analysis of temporal trends and spatial distributions in top-of-atmosphere shortwave radiation based on multiple datasets from satellites observations, reanalyses, and climate models

Understanding the spatiotemporal patterns of global net shortwave radiation at the top-of-atmosphere is crucial for studying Earth's energy budget and associated radiative forcing from natural or anthropogenic events. However, the differences across datasets can result in divergent or even contradictory conclusions regarding radiation variability, making it difficult to accurately assess and project climate change based on any single dataset. This paper comprehensively evaluates and intercompares the spatiotemporal variation and trends of 25 radiation datasets from satellite observations, reanalyses and climate models. Using Clouds and the Earth Radiant Energy System Energy Balanced and Filled (CERES EBAF) as the reference, which maintains stable annual means of ∼241 ± 1 W m⁻² from 2003 to 2014, CM SAF cLoud, Albedo and Radiation dataset (CLARA) and Fifth Generation ECMWF Reanalysis (ERA5) exhibit the most similar annual mean variations. Satellite-based and some reanalysis products reproduce the spatial patterns well over most regions, but large deviations (up to ∼50 W m⁻²) remain in regions dominated by marine stratocumulus and convective clouds. Decomposition of reflected radiation reveals that atmospheric reflection dominates the total bias, with compensatory effects between atmospheric and surface components being particularly pronounced in high-latitude regions. CERES EBAF exhibits increasing trend in global annual mean net shortwave radiation of 0.12 W m⁻² per decade during 2003–2014, equivalent to the cumulative radiative forcing from ∼160 Gt of CO₂ emissions. Similar interannual trends are consistently observed across EBAF, CLARA, and ERA5 datasets, with a pronounced positive trend over parts of the western Pacific warm pool (∼2.19 W m⁻² yr⁻¹) and a negative trend over portions of the Maritime Continent and South Pacific (∼ −1.98 W m⁻² yr⁻¹). In contrast, climate models and some reanalyses datasets exhibit substantial discrepancies, primarily driven by biases in the trends of atmospheric reflection, while deviations in surface reflection trends are more evident in high-latitude regions. Overall, this findings have implications for improving climate models and enhancing the understanding of the Earth's energy balance in response to climate change.