Reducing low-cloud bias in the GEOS-5 model: role of large-scale controls and ocean-atmosphere coupling

Cloud biases have been plaguing climate models for decades. Here, we investigate the17 controls of low-cloud bias in the Goddard Earth Observing System (GEOS) model. We18 first reduce the tropical and subtropical cloud bias over the ocean by employing a uni-19 fied turbulence and shallow convection parameterization based on the Eddy-Diffusivity/Mass-20 Flux approach. We then analyze in detail the model’s response to that change for both21 atmosphere-only (i.e., using prescribed ocean conditions) and fully-coupled ocean-atmosphere22 simulations. For the atmosphere-only simulation, certain regions benefit from using the23 new parameterization, such as the intertropical convergence zone and coastal stratocumulus-24 dominated regions. However, the overall response of the system is more complex, and25 compensating errors away from those regions emerge due to the presence of multi-scale26 feedbacks. The improvement is more pronounced for the coupled simulation, in which27 more significant (as compared to the atmosphere-only simulation) mesoscale biases in28 the shortwave cloud radiative effect and SST for the south Pacific Ocean are strongly29 reduced. Decomposition of model biases into the errors in the representation of (i) large-30 scale atmospheric states and (ii) cloud properties for those states shows that the new model31 configuration more accurately represents both components, confirming a more realistic32 parameterization of cloud regimes for a range of subtropical and tropical conditions. The33 results also indicate challenges in constraining climate models after reducing their key34 biases and the need for subsequent adjustments and tuning of the remaining interactions35 and feedbacks.