The impacts of Cloud-Radiation Bias on Circulations and Temperatures Simulations in CMIP5 and NCAR CESM Sensitivity Experiments

Conventional global climate models (GCMs), including all Coupled Model Intercomparison Project (CMIP) phase three (CMIP3) and most phase five (CMIP5) models, consider radiation interactions only with small-particle/suspended cloud mass, ignoring large-particle/falling and convective core cloud mass. However, constraints on models' global radiation balance, clouds, and related quantities are made with measurements sensitive to a broader range of hydrometeors. Our previous observation-based modeling studies led to the hypothesis that the typical practice of ignoring the impacts of precipitating hydrometeors would account for at least a portion of this systematic bias. We found that this systematic bias of cloud-radiation leads to persistent systematic biases in the CMIP3 and CMIP5 (compared to observations) that include the overestimation of downward shortwave (SW) at the surface and outgoing longwave (OLR) at the top of the atmosphere (TOA) in heavily precipitating regions (e.g., ITCZ, warm pool). This leads to biases in the vertical radiative heating structure and in turn the convective strength, and extends to the atmospheric circulations resulting in bias in atmospheric temperatures simulations in CMIP5. We explore and characterize the impacts of the bias from the interaction of radiation and precipitating hydrometeors (i.e. snow) on temperatures, water vapor and circulation simulations, using the NCAR coupled GCM by conducting sensitivity experiments that turn off the radiation interaction with snow. The changes (without snow-radiation interaction minus snow-radiation on) associated with the typical exclusion of precipitating hydrometeors are consistent with those biases in CMIP5, which are more OLR at TOA and more downward SW flux at the surface in the precipitating and convectively active regions (e.g., ITCZ, warm pool) in conjunction with significantly underestimating the amount of cloud water content. Neglecting large hydrometeors also enhances the net radiative cooling near the cloud top and triggers more unstable convective updrafts in the precipitating and convectively active regions. We conclude that the clouds in the no-snow simulations result in too much SW energy reaching the surface with excessive LW cooling aloft. The differential vertical heating leads to a stronger local Hadley circulation over the mid- and east Pacific, in conjunction with changes including low-level winds, upper level winds, and surface wind stress as well as the vertical zonal winds structures, and the associated vertical temperatures bias (compared to AIRS) following the thermal wind relationship which are all consistent with those found in the CMIP5 ensemble averages.