Abstract
Clouds constitute a large portion of uncertainty in predictions of equilibrium climate sensitivity (ECS), with tropical high cloud feedbacks exhibiting considerable spread across models. This study applies the cloud radiative kernel technique of Zelinka et al. (2012a; 2013) to 22 models across the CMIP5 and CMIP6 ensembles to survey tropical high cloud feedbacks and analyze their relationship to ECS, mean state properties, and changes to the tropical overturning circulation, precipitation efficiency, and deep convective organization across scales. First, the intermodel spread in tropical high cloud net, altitude, and optical depth feedbacks exhibit significant correlations to ECS in the tropical mean and on convective margins while the spread in the high cloud amount feedback is uncorrelated to ECS, motivating further exploration of physical mechanisms driving the intermodel spread in high cloud feedbacks. Intermodel variability in deep convective organization – at both the mesoscale and planetary scales – relates to the intermodel spread in high cloud feedbacks along convective margins. Most notably, decreases in tropical ascent area and increases in mesoscale organization of deep convection relate to more positive high cloud feedbacks, particularly within weak ascent and weak descent regimes. Increases in mesoscale organization also coincide with a greater weakening of the Pacific Walker circulation. Precipitation efficiency, on the other hand, does not appear to systematically drive much spread in high cloud feedbacks across the tropics. Finally, connections between mean state high cloud properties and high cloud feedbacks are explored in an attempt to place observational constraints on high cloud feedbacks and ECS. High ECS models are cloudier in the upper troposphere but have a thinner high cloud population. Having more thin high clouds in the mean state generally yields more positive high cloud altitude and optical depth feedbacks and it either amplifies or dampens the high cloud amount feedback depending on the large-scale dynamical regime. A clear link between mean state high cloud characteristics and deep convective onset behavior across the CMIP ensemble highlights how model diversity in convective processes contributes systematically to diversity in mean state high cloud properties. In summary, this analysis highlights the importance of tropical high cloud feedbacks for driving intermodel spread in ECS, underscores the influence of dynamical regime shifts on the response of high clouds to warming, and suggests that mean state high cloud characteristics might provide a unique opportunity for observationally constraining high cloud feedbacks.
