The variation of the incoming shortwave radiation with height was not taken into account because it depends for the most part on the model's forecast of cloud heights. The impact of changes in cloudiness are dramatically reduced over surfaces with high albedos because of multiple reflections between the cloud base and the surface (Schneider and Dickinson (1976)). Fig.7 shows the downwelling shortwave radiation at Thule (11 m.) located on the coast in Greenland and South Pole station ( 2835 m.) in Antarctica. The annual cycle of the downwelling shortwave radiation is generally well simulated, both climate models do however exhibit a tendency to underestimate the peak intensity of summer insolation.
The downwelling longwave radiation derived from the MIT climate model output is interpolated linearly between the model levels to the altitude of the grid points on the ice sheet, the values predicted by the ECHAM model are left unaltered. Neither climate model appears to have any systematic biases in estimating this component of the radiation balance (not shown).
Because the upwelling longwave radiation flux depends only on the surface snow or ice temperature (blackbody emission, the emissivity of snow and ice is very close to one in the infrared part of the spectrum), which is in turn determined by the net surface energy budget, the upwelling longwave radiative flux provides a very good indicator of the overall quality of that budget. The comparison (Fig. 8) between observations and model estimates at ETH Camp ( 1155 m.) for Greenland and South Pole ( 2835 m.) for Antarctica shows that the agreement is generally good, although the model's tendency to underestimate the peak insolation can lead to surface snow temperatures which are too cold and an underestimation of the upwelling longwave flux (this occurs for example at the Carrefour site, not shown). In the summer in the ablation zone this flux is constrained by the surface ice temperature which remains near the melting point. The agreement between the model predictions and the station data at lower elevations in Greenland is therefore also very good.
There exist few reliable measurements of the turbulent latent and sensible heat fluxes to compare the model's output with, most published estimates are themselves derived from bulk formulae and therefore depend on the details of those parameterizations. The model's relative humidity was chosen to give a good aggreement with the measurements taken at ETH Camp.