Variability in modeled cloud feedback tied to differences in the climatological spatial pattern of clouds

Abstract

Despite the increasing sophistication of climate models, the amount of surface warming expected from a doubling of atmospheric CO $$_2$$ (equilibrium climate sensitivity) remains stubbornly uncertain, in part because of differences in how models simulate the change in global albedo due to clouds (the shortwave cloud feedback). Here, model differences in the shortwave cloud feedback are found to be closely related to the spatial pattern of the cloud contribution to albedo ( $$\alpha$$ ) in simulations of the current climate: high-feedback models exhibit lower (higher) $$\alpha$$ in regions of warm (cool) sea-surface temperatures, and therefore predict a larger reduction in global-mean $$\alpha$$ as temperatures rise and warm regions expand. The spatial pattern of $$\alpha$$ is found to be strongly predictive ( $$r=0.84$$ ) of a model’s global cloud feedback, with satellite observations indicating a most-likely value of $$0.58\pm 0.31$$ Wm $$^{-2}$$ K $$^{-1}$$ (90% confidence). This estimate is higher than the model-average cloud feedback of 0.43 Wm $$^{-2}$$ K $$^{-1}$$ , with half the range of uncertainty. The observational constraint on climate sensitivity is weaker but still significant, suggesting a likely value of 3.68 ± 1.30 K (90% confidence), which also favors the upper range of model estimates. These results suggest that uncertainty in model estimates of the global cloud feedback may be substantially reduced by ensuring a realistic distribution of clouds between regions of warm and cool SSTs in simulations of the current climate.