Global climate models are unable to reproduce cloud cover response to aerosol

Anthropogenic aerosols and their impact on clouds are one of the largest uncertainties in climate assessment. A major challenge lies in the lack of robust large-scale observational constraints on aerosol-cloud interactions (ACI), leaving models poorly constrained. Our recent work has successfully disentangled cloud property responses to large-scale aerosol perturbations from natural variability of clouds during the Holuhraun-2014 volcanic event, where plumes spanned thousands of km over an otherwise near-pristine North Atlantic. Here, for the first time, this large-scale ACI signal is used to constrain six diverse global climate models of how clouds respond to changes in aerosols, and transform our understanding of uncertainty. We show that marine liquid cloud albedo responses to aerosols are reasonably well simulated, although through the compensation of uncertainties. However, all models largely underestimated cloud cover responses to aerosols, with five out of six models falling outside the 90% confidence level of the observational constraint. This issue persists even when tuning five distinct cloud schemes and sensitivity tests on key cloud processes. Such bias in cloud cover response is a major driver of simulation uncertainty in ACI cooling and needs to be narrowed urgently to improve climate projections, and to support Net-Zero and climate intervention policymaking.