We investigate the adiabatic and diabatic thermal balance of an atmospheric general circulation model (GCM) under two conditions: the control case, representing today's atmosphere, and a ''nuclear winter'' scenario in which virtually all sunlight in northern hemisphere and mid-latitudes is absorbed in the upper troposphere by prescribed dense smoke clouds hypothesized to result from the burning of many cities in a nuclear war. We also examine the changes in moisture and cloudiness simulated by the model. Our object is to examine the reliability of existing simulations of the climatic response to assumed dense, widespread, high-altitude smoke and to identify improvements needed in model parameterizations. We find that in the smoke-perturbed case our model simulation of land surface temperature is particularly influenced (i.e., warmed) by parameterized diffusion of heat downward from the lower troposphere. In turn the lower troposphere over land is supplied with heat transported from the relatively warm oceans. Thermal balance in the perturbed atmosphere as a whole is dominated by intense solar heating of the upper troposphere smoke layer in mid-latitudes balanced by parameterized dry convection and large-scale dynamical heat transport. Clouds largely disappear in the mid to upper troposphere in smoke-affected regions as a consequence of a decrease in local relative humidity that results from temperature increases and, to a smaller extent, from a reduction of vertical moisture transport. The computation of substantial downward vertical heat diffusion into the lowest model layer is almost certainly an overestimate for the smoke-perturbed conditions of high vertical stability. Consequently, the use of the present diffusive parameterization will, in the absence of other errors, result in an underestimate of the magnitude of land surface cooling in the ''nuclear winter'' scenario. Current three-dimensional model simulations, however, contain numerous additional omissions and approximations that make it difficult to say whether ''nuclear winter'' simulations, to date, produce effects that are more or less severe than those that might really occur given the hypothesized smoke amounts and distribution.

We believe that the most important areas for GCM enhancement to study climatic effects of nuclear war-generated aerosols include improved surface and planetary boundary layer processes and incorporation of radiatively active aerosol tracer transport and removal.