A tropospheric general circulation model is coupled with a Lagrangian trace species transport and removal model to determine the climatic response to continental-scale smoke injections arising from fires generated following a hypothetical nuclear war. The hydrological response is found to be particularly sensitive, with significant reductions in precipitation over land areas occurring for smoke injections that yield only modest cooling. This sensitivity is traced to the suppression of surface evaporation caused by a smoke-induced reduction in the convective mixing of boundary layer and upper tropospheric air. A variety of sensitivity experiments is considered, including the sensitivity to smoke injection mass, location, day, season, duration and altitude of injection, and smoke composition and size distribution. For modest injections the land surface cooling in the first few weeks is found to be proportional to the smoke loading while for larger injections the amount of smoke affects only the duration of the surface cooling. After several weeks the northern hemisphere land surface cooling for an April injection is found to be comparable to that for a July injection.

A protracted smoke injection, which produces a more homogeneous horizontal distribution of smoke than an instantaneous injection, yields a greater surface cooling. The assumed initial aerosol size distribution of the injection is found to play a particularly important role in determining the climatic response, with injections composed primarily of submicron spherical particles producing a larger and longer surface cooling than an injection composed primarily of supermicron spherical aerosols. The importance of various feedback processes is also discussed. Smoke lofting associated with solar absorption by the aerosols enhances the aerosol residence time in the atmosphere by an order of magnitude, compared to aerosol lifetimes in the control atmosphere. The ground hydrology feedback in the tropics is found to amplify the sensitivity of precipitation over land areas and to enhance surface warming after several weeks. The treatment of a stability dependent rather than neutral boundary layer enhances the surfaces cooling in the first few weeks by about 30%. ¿ American Geophysical Union 1988