Cloud Feedbacks in the Climate System

Abstract

It is well established that inter-model differences in cloud feedbacks are the leading cause of differences in equilibrium climate sensitivity, the change in global mean near-surface temperature that eventually follows an instantaneous doubling in carbon dioxide concentrations. This thesis presents the contribution from four peer reviewed publications which seek to understand cloud feedback mechanisms. Webb et al. (2015a) investigated the diurnal cycle of marine cloud feedbacks in seven climate models using high frequency outputs at selected locations, and found that reductions in marine low-cloud fraction in the warmer climate are in almost all cases largest in the mornings when more cloud is present in the control simulations. Webb et al. (2015b) assessed the impact of convective parametrization on cloud feedback by analysing SST forced climate change experiments performed with ten climate models with convective parametrizations deactivated. This reduced the range of longwave cloud feedback but not shortwave or net cloud feedback, indicating that differences in convective parametrizations are not primarily responsible for the overall range in cloud feedback. Webb et al. (2018) perturbed surface evaporation and radiative cooling independently in a climate model, quantifying their individual contributions to changes in stability and low-cloud responses with climate warming, as well as to those in the near-surface atmospheric properties which regulate the hydrological sensitivity. Enhancing evaporation at the surface increased atmospheric stability and low-cloud fraction, while enhancing atmospheric radiative cooling destabilised the atmosphere and reduced low cloud. Webb and Lock (2020) investigated the finding of \cite{tian2015} that CMIP3 and CMIP5 climate models with larger double-ITCZ biases have lower climate sensitivities. It was hypothesized that deep convection encroaching into subtropical low-level cloud regions disrupts the formation of low clouds and inhibits positive low-cloud feedback. Results from sensitivity tests with a single model were consistent with this, but not all of the predicted regional correlations were statistically significant in the multi-model ensemble.