We proceed now to a discussion of the three-dimensional model simulations on which our conclusions are primarily based. Some of the existing general circulation models have been used to predict the climate for doubled or quadrupled CO2 concentration. The results of several such predictions were available to us: three by S.Manabe and his colleagues at the NOAA Geophysical Fluid Dynamics Laboratory (hereafter identified as M1, M2, and M3) and two by J.Hansen and his colleagues at the NASA Goddard Institute for Space Studies (hereafter identified as H1 and H2). Some results obtained with the British Meteorological Office model (Mitchell, 1979) were also made available to us but will not be described here because both the sea-surface temperature and the sea-ice distribution were prescribed in this model, thus placing strong constraints on the surface ΔT, whereas it is just the surface ΔT that we wish to estimate.
The only one of the five predictions available in published form is M1. M2 is described in a prepublication manuscript, and H1 in a research proposal. We learned of M3 and H2 through personal communication.
The Geophysical Fluid Dynamics Laboratory and the Goddard Institute for Space Studies general circulation models, which are the basic models used in the M and H series, respectively, were independently constructed and differ from one another in a number of physical and mathematical aspects. They also differ in respect to their geographies, seasonal changes, cloud feedbacks, snow and ice properties, and horizontal and vertical grid resolutions. These differences are summarized in Table 1. In this table “swamp” means that the model ocean has no heat capacity though it provides a water surface for evaporation, and “mixed layer” means that the model ocean has a heat capacity corresponding to that of an oceanic mixed layer of constant depth. Heat transport by ocean currents is neglected in both model oceans. This is one of the weaknesses of all the predictions, as discussed in Section 3.3.
The horizontal resolution of the H series is rather coarse and perhaps only marginal for meaningful climate prediction. On the other hand, these models take into account more physical factors, such as ground heat storage, sea-ice leads, and dependence of snow-ice albedo on snow age, than do the models of the M series.
The models M1, H1, and H2 were run for doubled CO2 concentrations, M2 for both doubled and quadrupled concentrations, and M3 for quadrupled concentrations. The temperature changes for doubled CO2 in M2 were approximately half of those for quadrupled CO2. Since it can be expected that a similar result would have been obtained for M3, we have halved the M3 temperature changes.*