Cloud feedbacks strongly affect models’ ability to reliably predict future climate, and the research community has worked hard to reduce this uncertainty. The science is now mature enough that quantitative “ground truth” values of several key cloud feedback components were recently established based on expert assessment of a very wide body of evidence from theory, observational analyses, and high-resolution modeling. Somewhat paradoxically, the spread in cloud feedbacks across global climate models continues to grow, with some of the latest models simulating strong amplifying cloud feedbacks leading to very high effective climate sensitivity (ECS). Quantitatively evaluating climate models’ individual cloud feedback components against this ground truth reveals that models with smallest errors have moderate feedbacks and climate sensitivity, a result consistent with the increasing body of evidence ruling out high and low extremes of ECS. Moreover, skillful simulation of mean-state clouds does not guarantee skillful simulation of cloud feedback, highlighting the need for advanced process-level evaluation of how clouds respond to their environment. Applying such analyses to the E3SM model reveals how specific parameterization choices lead to a reduced cloud feedback from the very high values of version 1 to more reasonable values in version 2. Leveraging satellite-derived sensitivities of clouds to their environment, we also estimate the degree to which cloud feedback depends on the pattern of surface warming, an uncertainty that complicates our ability to infer climate sensitivity from the observed climate record. A recurring theme is the critical importance of high quality global observations and advanced diagnostics including satellite instrument simulators for rigorous model evaluation and process-level understanding.

This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. It is supported by the Regional and Global Model Analysis Program of the Office of Science at the DOE. IM Release #LLNL-ABS-838336.