JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 99, NO. D10, PAGES 20,829-20,845, OCTOBER 20, 1994 Response of an atmospheric general circulation model to radiative forcing of tropical clouds Steven C. Sherwood, • V. Ramanathan,•, 2 Tim P. Barnett, • Mary K. Tyree, • and Erich Roeckner 3 Abstract. The effects of upper tropospheric cloud radiative forcing (CRF) on the atmospherehave been examined using a recent version of the atmospheric general circulation model (AGCM) developed by the Max Planck Institute of Meteorology and the University of Hamburg. This model reproducessatellite-observed radiative forcing of clouds well overall, except that model maxima somewhat exceed those of observations.Three simulations have been performed where the clouds above 600 mbar have been rendered transparent to all radiation: first, throughoutthe tropics in the "NC" experiment; then only over oceans warmer than 25øC in the "NCW" experiment; and finally, only over the western Pacific warm oceans in the "NCWP" experiment. The local radiative effects of these clouds when they are present in the model are radiative heating of the middle to upper troposphere due to convergence of longwave and solar radiation; radiative coolingof the tropical atmosphere near and above the tropopause; a large reduction of solar radiation (50 to 100 W/m 2)reaching the surface; and a slight increase (5 to 20W/m 2)in the downward longwave radiation at the surface. The removal of cloud radiative forcing significantlyalters the circulation of the model atmosphere,as in previous AGCM studies, showingthat a seemingly moderate heat source such as CRF is nonetheless capable of widespread influence over the global circulation and precipitation. The experiment responses include a significant weakening (in NCW) or rearrangement (in NCWP) of the Walker circulation. Zonal mean cloud cover, rainfall, and low-level convergencechange modestly in the experiments, while zonal departures of these from their tropical means shift considerably. Regions over the warmest oceans which lose CRF become much less cloudy, indicating a positive local feedback to convection. The experiment circulation changes are diagnosed in terms of simple energy budget arguments,which suggest that the importance of CRF is enabled by the small magnitudeof the atmosphericmoist energy transport in the tropics. They also suggest that the responseof the zonal mean atmospheremay be strongly dependenton the response of zonal eddies and on interactions between surface fluxes and tropospheric lapse rates. The response of the zonal eddies itself should be relatively independentof these interactions. 1. Introduction The desire to predict climate changehas spurred increas- ing effortsin recent years to developand understand general circulation models of the atmosphere.While these models start with similar representations of the basic equations of fluid dynamics, they parameterize unresolved processes using a variety of strategies. The models, furthermore, exhibit widely varying responses to perturbationssuch as global warming scenarios. Recently, intercomparison stud- ie,q have attrihnted the ,qnnrce of di.qcrepancy to the param- eterizations of clouds [Cess et al., 1989]. A better under- standing of cloud processes is needed to produce more reliable models of climate. 1 Scripps Institution ofOceanography, University ofCalifornia at San Diego, La Jolla. 2Also at Center for Clouds, Chemistry andClimate, Scripps Institution of Oceanography, La Jolla, California. 3Max-Planck-Institut f/ir Meteorologie, Hamburg, Germany. Copyright 1994 by the American GeophysicalUnion. Paper number 94JD01632. 0148-0227/94/94JD-01632505.00 Clouds and convection influence the Earth's climate by redistributingradiation and releasinglatent heat. The latent heat release is a nonlocal transfer of heat from the oceans to the atmosphere,while the radiative effects, althoughperhaps smaller in absolute magnitude, are potentially more far- reaching since they can change the balance of the energy budgetfor the planet as a whole. In both respects the clouds tend to act as a time-varying, nonuniformheat sourcefor the atmosphere and (usually) sink for the surface. The exact spatial dependenceof this heating is poorly known, particu- larly in the vertical, and its statistics may depend signifi- cantly on a large variety of atmospheric parameters. A number of simpler models of the tropical atmosphere and its response to heating precede this study, a notable early model being that of Gill [1980], who considered a linearized shallow-water equation model on a beta plane. The flow resulting from a tropical heat source off the equator included a pattern of ascent similar to that of the source and a zonally broad pattern of descent. Hoskins and Karoly [1981] developed a five-layer, spherical, baroclinic model linearized about the observed mean state. They found that subtropical heating would excite Rossby waves which 20,829