For comparison purposes, the convective adjustment parameterization was introduced into an H model with fixed sea-surface temperatures and was found to reduce appreciably the penetration of water vapor and cloud to high levels (J.Hansen, NASA Goddard Institute for Space Studies, personal communication). Since the original penetration was probably too high because of lack of noncloud air entrainment, we conclude that the surface ΔT’s due to the upper water-vapor-cloud feedback may very well have been overestimated in the H series, whereas, because of insufficient penetration, they were probably underestimated in the M series. Since, moreover, the snow-ice boundary is too far equatorward in H1 and too far poleward in M1 and M2 (see Appendix), we believe that the snow-ice albedo feedback has been overestimated in the H series and underestimated in M1 and M2. For the above reasons, we take the global or hemispheric surface warmings to approximate an upper bound in the H series and a lower bound in the M series (with respect to positive water-vapor-cloud and snow-ice albedo feedback effects). These are at best informed guesses, but they do enable us to give rough estimates of the probable bounds for the global warming. Thus we obtain 2°C as the lower bound from the M series and 3.5°C as the upper bound from H1, the more realistic of the H series. As we have not been able to find evidence for an appreciable negative feedback due to changes in low- and middle-cloud albedos or other causes, we allow only 0.5°C as an additional margin for error on the low side, whereas, because of uncertainties in high-cloud effects, 1°C appears to be more reasonable on the high side. We believe, therefore, that the equilibrium surface global warming due to doubled CO2 will be in the range 1.5°C to 4.5°C, with the most probable value near 3°C. These estimates may be compared with those given in our discussion of feedback effects in one-dimensional, radiative-convective models. There the range was 1.6°C to 4.5°C, with 2.4°C estimated as a likely value.
We recall that the snow-ice albedo feedback is greater in the northern than in the southern hemisphere because of the greater land area and the lack of albedo change over Antarctica. Hence we estimate that the warming will be somewhat greater in the northern hemisphere and somewhat less in the southern hemisphere.
The existing general circulation models produce time-averaged mean values of the various meteorological parameters, such as wind, temperature, and rainfall, whose climate is reasonably accurate in global or zonal mean. Their inaccuracies are revealed much more in their regional climates. Here physical shortcomings in the treatments of cloud, precipitation, evaporation, ground hydrology, boundary-layer turbulent transport phenomena, orographic effects, wave-energy absorption and reflection in the high atmosphere, as well as truncation errors arising from lack of sufficient resolution combine to produce large inaccuracies. Two models may give rather similar zonal averages