Cloud-resolving model simulations of a mock Walker circulation and its climate sensitivity

Two- and three-dimensional cloud-resolving model simulations of a mock Walker-type circulation are performed over both a fixed, sinusoidally-varying SST (as in Grabowski et al 2000) and a slab ocean with zonally-varying ocean heat transport. Such simulations provide a highly idealized realization of tropical atmospheric circulations, and tropical climate sensitivity is explored in this idealized setting through SST/SST+2 and 1x/2xCO2 simulations as in the work of Cess et al (1990). The cloud-resolving model employed is the System for Atmospheric Modeling, developed by Marat Khairoutdinov at Colorado State University, which includes prognostic equations for total water (vapor + cloud) and precipitating water, and diagnoses the individual phases of condensate and precipitate (e.g. cloud water, cloud ice) based on temperature. This model has been extended to include different radiation (from CAM3.0) and microphysical (Fu, Krueger & Liou 1995) parameterizations, and the sensitivities of the results to changes in resolution, dimensionality and parameterizations are studied as well.

The simulations exhibit a region of persistent convection over the warmest SSTs/weakest ocean cooling, as well as a region of subsidence over the coldest SSTs/strongest ocean cooling. The inflow to the convective region occurs primarily in the boundary layer. While this basic pattern is reproduced in the different simulations, differences in the extent of the convective region, the level of upper tropospheric relative humidity and the amount of low cloud can lead to different climate sensitivities depending on the model setup (i.e. fixed SST vs. slab ocean) and parameterizations. The simulations show 1--1.8K warming in 1x/2xCO2 experiments --- depending on the choice of parameterization --- but show stronger negative feedbacks in the SST/SST+2 experiments, where the convective region narrows substantially as the domain-average SST increases. Profiles of the cloud fraction and amount above the freezing level are found to be substantially similar in the different simulations when plotted as a function of temperature, supporting the fixed anvil temperature hypothesis of Hartmann and Larson (2002).