The ECHAM model generally reproduces well the local features of the accumulation pattern over Greenland and Antarctica (Wild and Ohmura (1999)). The precipitation field derived from the MIT model is obtained by multiplying the zonal average precipitation amount with an array consisting of the observed snow accumulation over the ice sheet, normalized over each latitude band in order to conserve the amount of precipitation predicted by the climate model. The total snow accumulation integrated over the ice sheets predicted by the two climate models and the measured estimates are summarized in Table 1. Because the observed accumulation is for the most part derived from snow pit measurements and not from rain or snow gauge data, the evaporation estimated by the snowpack model is subtracted from the model's total snowfall before comparing it to the observed accumulation. The snow accumulation predicted for the Greenland ice sheet is within the range of uncertainty of observations for both climate models. The high resolution of the ECHAM model does however allow some improvement over the value predicted by the MIT model in Antarctica but both are significantly higher than the most recent observations (Vaughan et al. (1999)). Note that the accumulation calculated on the original ECHAM 4 T106 grid by using that model's snowfall and evaporation is 20% smaller than the number obtained on the 40 km grid (Wild and Ohmura (1999)). The difference stems from the larger evaporation calculated by the ECHAM model when compared to the snowpack model and from differences in snowfall in the coastal areas of the continent which are due to the interpolation procedure and the resolution. As noted by Genthon (1994), the overestimate of precipitation in high latitudes is a problem encountered by many GCM's. In the case of the MIT model, the excess precipitation in Antarctica is to some degree due to the absence of topography. This allows the surface temperatures and specific humidities to be higher than they should, thereby leading to excessive precipitation. Adding the zonally averaged topography of Antarctica at the model's lower boundary does in fact reduce the total accumulation by 15%. The second source of error is associated with the presence of a vertical wall at the southernmost grid point. This wall induces excessive upward motion and precipitation. This is a problem specific to two-dimensional models which is not encountered by their three-dimensional counterparts.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The estimates of snowfall and evaporation and thus accumulation are virtually unaffected by the choice of the grid resolution, 40 or 20 km, on the Greenland ice sheet.