The climatological input to the snow melt models is derived from the simulations performed with two climate models. Both model outputs are interpolated with a distance weighted scheme onto a 20 km grid for Greenland and a 40 km grid for Antarctica.
Because of the high resolution required to capture adequately the topography and climate of Greenland and Antarctica, the simulations performed with the 1.1o x 1.1o resolution (T106) version of the ECHAM 4 GCM could not be performed in transient mode with a time varying forcing over the 21st century (Wild and Ohmura (1999)). These authors chose instead to use the sea surface temperature and sea ice distribution provided by a lower resolution simulation of the IS92a transient scenario (Houghton et al. (1996)), with the same model (Roeckner et al. (1999)), for the current climate and at the time of doubling of the equivalent carbon dioxide level (i.e. allowing for increases in CO2 and other trace gases), as boundary conditions for 10 year integrations of the high resolution model. The terminology ``time-slice experiment'' will be used to decribe these simulations.
The MIT model is a zonally averaged version of the GISS GCM which does however distinguish between land, ocean, land-ice and sea-ice (Sokolov and Stone (1998)). This model's main advantage is its computational efficiency, it allows the simulation of the climate's transient response to a large set of emissions scenarios. By making what are thought to be reasonable assumptions about the emissions rate of greenhouse gases and by varying key model parameters, the objective was an assessment of the range of sea-level change which can be expected to accompany changes in the mass balance of Greenland and Antarctica over the next century. The scenarios are described in detail in section 5.2. Although one could assume that a zonally averaged model would be unable to give reliable estimates of the mass balance of an ice sheet, the MIT model's performance in reproducing the known features of the current state of the mass balance of Greenland and Antarctics was respectable (Bugnion (1999)). The temperature distribution over the ice sheets is reconstructed from the air temperatures at the sea-level by using known lapse rates (Ohmura (1987), Schwerdtfeger (1970)) and the incoming longwave radiation is interpolated to the altitude of the grid point. The precipitation distribution is obtained by weighting the zonally averaged precipitation with an array representing the observed accumulation normalized in order to conserve the amount of snow- and rainfall predicted by the model; all other variables are left unaltered.