CLUBB Reduces Low-Cloud Dependency on Shallow Convective Mixing in the Community Atmosphere Model: Insight From Stable Water Isotopes

Abstract

Modeling experiments and field campaigns have evaluated shallow convective mixing as a potential constraint on the low-cloud climate feedback, which is critical for establishing climate sensitivity. Yet the apparent relationship between low-cloud fraction and shallow convective mixing differs substantially among general circulation models (GCMs), large eddy simulations, and both remote sensing and in situ observations. Here, we consider how changes in GCMs' representations of subgrid-scale vertical moist fluxes can alter the cloud-mixing relationship. Using vertical profiles of water vapor isotope ratios (δD) to characterize the strength of shallow convective mixing in a manner that can be compared directly to satellite observations, we evaluate the cloud-mixing relationship produced in tiered experiments with the Community Atmosphere Model (CAM). From versions 5 to 6 of CAM, the most notable physics change is CLUBB, a scheme that unifies the representation of shallow convection and boundary layer turbulence through a joint probability density function (PDF) for subgrid velocity and moisture. CLUBB reduces the covariance between low-cloud fraction and shallow convective mixing, producing a bivariate distribution that is more similar in character to monthly averaged satellite observations. Using parameter sensitivity experiments, we argue that CLUBB's ability to simulate skewness in the distribution of vertical velocity produces more isolated but stronger moist updrafts, which reduce the grid-mean low-cloud fraction while maintaining efficient hydrological connectivity between the boundary layer and the free troposphere. These results suggest that mixing is not an effective predictor of low-cloud feedback in GCMs with PDF closure schemes.