Is the cloud absorption of solar radiation still underestimated significantly by current climate models?

Comparing with observations, the abnormally smaller cloud absorption (to solar radiation) given by climate models (namely alleged cloud absorption anomaly) ever raised widespread concerns in the mid-1990s and early 2000s but was seldom mentioned thereafter. Based on three state-of-the-art modeled products, NCEP CFSv2, ECMWF ERA5 and NASA MERRA2, and the newest collocated satellite-surface observation in the last 12 years (2012–2023), we reinvestigate this controversial issue. Our results demonstrate the observed cloud absorption of solar radiation still significantly exceeds the modeled (regardless of modeled products), but their systematic discrepancy has dropped a lot, especially for NCEP CFSv2. NCEP CFSv2 has the lowest bias with the observation, followed by ECMWF ERA5, and the bias of NASA MERRA2 is largest. This implies that cloud absorption anomaly fluctuates with not only sites (as reported by previous studies) but also models. Models’ radiation schemes that introduce the Monte Carlo Independent Column Approximation (McICA) may mitigate the systematic discrepancy between observation and modeling essentially. Additionally, it is noteworthy that there is not a perfect approach to obtaining the observed cloud absorption and particularly the water vapor difference between clear and cloudy skies often would result in its unrealistic overestimation. If the influence from the water vapor difference is neglected, NCEP CFSv2, ECMWF ERA5 and NASA MERRA2 underestimate globally-mean cloud absorption by approximately 10.07 W/m², 16.65 W/m² and 18.67 W/m², respectively; and if it is corrected, the underestimations will be reduced to 7.75 W/m², 14.33 W/m² and 16.35 W/m², respectively.