Studies of well-observed glaciers indicate that glaciers in a wetter climate are more sensitive to changes in air temperature than glaciers in dry regions. This arises because the areaelevation distribution is different and the albedo feedback is more effective for glaciers with high precipitation snowfall. For the calculation of the glacier response to climatic change, all glaciers and small ice-caps on Earth have, therefore, been placed in one of 100 regions, each characterized by the present-day precipitation rate and glacierized area. For each region, the sensitivity of the glacier mass balance to changes in temperature depends on the mean annual precipitation (see Oerlemans and Fortuin, 1992). Model calculations start in 1990, although at present most glaciers are not in equilibrium. To account for the observed present-day thinning of several glaciers, projections of the contribution of glaciers and icecaps to sea level change include a constant long-term trend of 0.5 mm/yr sea level rise, which is consistent with observations.

The sea level contributions of the Greenland and Antarctic ice sheets are estimated using dynamic ice flow models. In the case of Greenland, a two-dimensional (latitude-longitude) model with a horizontal resolution of 20 x 20 km is used (Cadee, 1992), while a three-dimensional model of the Antarctic ice sheet with 20 km horizontal resolution and 14 layers is used (Huybrechts, 1992; Huybrechts and Oerlemans, 1990). Both ice sheet models are forced with the zonally-averaged temperature changes produced by the coupled atmosphere-ocean climate model. In the case of Greenland, the accumulation rate is held constant at the observationally based estimate for the present (Ohmura and Reeh, 1991), and changes in the rate of melting are computed using a simple surface energy balance model (van de Wal and Oerlemans, 1994). Model calculations start in 1990, at which time the Greenland ice sheet is assumed to have been in a state of equilibrium. In the case of Antarctica, a combination of observations and theory suggests that the accumulation rate should increase with increasing temperature, in proportion to the increase in the ability of air over Antarctica to hold moisture. The accumulation rate over Antarctica is therefore

estimate values obtained here differ significantly from the middle results shown in Section 4.3.1. Results obtained here are shown in Figure 14, and should be compared with the corresponding results in Figure 13. The largest difference is in the thermal expansion contribution to sea level, followed by the difference in the Antarctic contribution. Although the reasons for these differences were not entirely resolved at the time of publication of the SAR WGI, several differences in model features were identified (SAR WGI: Section 7.5.3.2). The differences likely to be important for the thermal expansion component of sea level rise include the meridional resolution of the two-dimensional model, the different model formulations of heat exchange between atmosphere and oceans, the absence of sea ice in the upwelling-diffusion model, different climate sensitivities (2.5°C for the one-dimensional model middle case, 2.2°C for the two-dimensional model, the latter not being adjustable), and the way in which the thermohaline circulation is represented. In the case of the Antarctic contribution, different temperature perturbations are used to force the ice sheet and smaller ice sheet sensitivities are used for the results presented in Section 4.3.1.

As is the case for the calculations presented in Section 4.3.1, a wide range of model input parameters is possible, giving a wide range of sea level results. However, the middle or "best"

Sea level change (cm)

Figure 14. The individual contributions to the "MID" sea level rise case for emission scenario IS924 as computed using the twodimensional upwelling-diffusion model described in Section 4.3.2 (reproduced from SAR WGI, Figure 7.11).

34

change are: the regional distribution of the temperature changes; the initial volume of the glaciers and ice-caps, and their sensitivity to increases in temperature; and the initial state of balance of the Greenland and Antarctic ice sheets and their sensitivity to temperature changes. Uncertainties in sea level rise cannot, therefore, be separated from uncertainties in global mean

temperature change. However, changes in accumulation will also affect the volume of land-based ice. For the glaciers and icecaps and for the Greenland ice sheet, accumulation has been assumed constant, where for the Antarctic ice sheet, accumulation is assumed to increase as temperature increases. Figures 11 and 12 express the uncertainty in temperature and sea level rise.

Simple climate models have been, and will continue to be, used for analysis of the global scale implications of alternative emissions scenarios or of alternative assumptions concerning the properties of individual model components. It is, therefore, pertinent to compare the global mean temperature and sea level projections as simulated by one- and two-dimensional upwelling-diffusion models on the one hand, and AOGCMs on the other hand.

Figure 15 compares the change in global mean surface air temperature as simulated by several different AOGCMS with that of the one-dimensional upwelling-diffusion model with a CO2 doubling climate sensitivity of 2.5°C and that of the twodimensional climate model (whose sensitivity is fixed at 2.2°C). The spread in the AOGCM results can be largely explained by the differences in the model climate sensitivity, which varies from 2.1 to 4.6°C. Note that the interannual variability in the AOGCM response is absent in the SCM response, which increases smoothly but is otherwise similar to the AOGCM response. Comparison of Figure 15 with Figure 11 illustrates the ability of upwelling-diffusion models to span the results of most AOGCMs when a range of values for the climate sensitivity is used.

A further illustration of the comparability of AOGCM and SCM time-dependent behaviour is given in Figure 16, which compares the global mean temperature change for the Geophysical Fluid Dynamics Laboratory (GFDL) AOGCM and the upwelling-diffusion climate model when both models are driven by various rates of increase in atmospheric CO2 concentration (see SAR WGI: Section 6.3.1). To ensure a valid comparison, the SCM climate sensitivity was set at the GFDL model value of 3.7°C. All other parameter values remained unchanged. The value of the land/ocean sensitivity differential (1.3), chosen on the basis of other GCM results (Raper, et al.. 1996), is similar to that for the GFDL model. The thermohaline circulation in the SCM was made to vary with surface warming in a manner that closely approximated the variation in the GFDL model (Manabe and Stouffer, 1994). The surface temperature responses are seen to agree well over a wide range of forcings.

Global temperature change (°C)

Global temperature change (°C)

50

100

150

200

250

301

Year from start of experiment

Figure 15. Comparison of global mean surface air temperature change as simulated by several different AOGCMs (with climate sensitivity varying from 2.1 to 4.6°C), the one-dimensional upwelling-diffusion climate model (climate sensitivity of 2.5°C), and the two-dimensional upwelling-diffusion model (climate sensitivity of 2.2°C), in each case driven by a 1 per cent per year (compounded) CO2 concentration

Year from start of experiment

Figure 16. Global mean surface air temperature increase as computed by the GFDL AOGCM (solid lines) and the one-dimensional upwelling-diffusion climate model with a CO2 doubling sensitivity of 3.7°C. Results are shown for cases in which the atmospheric CO2 concentration increases by 0.25, 0.5, 1, 2 and 4 per cent per year (reproduced from SAR WGI, Figure 6.13).