Performance of the climate models in capturing the climate of the ice sheets

Snow melt models are ideally tested by providing as input observed meteorological data and comparing the predicted and observed ablation. There are however no stations in the ablation region of the Greenland ice sheet which provide year-round measurements of all the meteorological parameters required for the snowpack model. The two climate model simulations which were therefore used as input to the melt models are the current climate representations obtained with the ECHAM 4 GCM (Wild and Ohmura (1999)) and the MIT 2D-LO model (Sokolov and Stone (1995), Sokolov and Stone (1998)). The ECHAM model is a three-dimensional climate GCM which was run at very high resolution (T106 or 1.1^o x 1.1^o) in order to adequately resolve the topographic features of the Greenland and Antarctic ice sheets. The simulation was performed by using the sea surface temperature and sea-ice distribution predicted by a coarser resolution simulation with the same model (Roeckner et al. (1999)) as lower boundary conditions for a 10 year integration of the high resolution version. The MIT 2D-LO model is a zonally averaged, height vs. latitude, version of the GISS GCM (Hansen et al. (1983)). The model does however distinguish between land, ocean land-ice and sea-ice in each latitude band and has a resolution of 7.8^o in latitude. Because it has no topography, certain climatic input fields had to be adjusted, as described later, in order to obtain realistic distributions over the ice sheets. The advantage of the ECHAM model is its high resolution and physics, yet the simplicity of the MIT model allows the simulation of a range of transient climate change experiments which are described in a companion paper (Bugnion (1999)). The input variables from the climate models are: The downwelling shortwave and longwave radiation, wind speed, surface air temperature and the precipitation. They are interpolated onto the 20/40 km grid of the snow melt models.

Because the input data from the climate models was available as monthly means and the snowpack model's timestep is much shorter, random gaussian variability was added to the temperature ( $\sigma_{T}^{}$ = 2^o C) and wind records ( $\sigma_{\vec{v}}^{}$ = 4 m/s) to ensure an adequate climate variability. The precipitation was disaggregated into individual events with a simple stochastic rainfall model in order to allow the albedo to depend upon the time elapsed since the previous snowfall event. This model performs a random selection between a set of precipitation events of predertermined duration and intensity.

The reliability of the representation by those two climate models of the climate of the ice sheets is assessed in the following paragraphs by comparing the predicted mass and energy balance components to observations.