Introduction

Greenland and Antarctica contain together over 99% of the volume of glacier ice on Earth or about 75 meters of sea-level equivalent, yet the current state of the mass balance of these ice sheets, let alone how it may evolve in the future, is poorly known. The estimated contribution of those two ice sheets to the sea level rise of the past century reflects these uncertainties: The Houghton et al. (1996) report gives a central value of 0 cm, with the uncertainty range estimated at $\pm$ 4 cm for Greenland and $\pm$ 14 cm for Antarctica.

The two most important processes determining the mass balance at the surface of an ice sheet are snow accumulation and the runoff of meltwater. The accumulation in Antarctica is balanced almost entirely by the calving of icebergs, Greenland however balances accumulation by approximately equal amounts of iceberg calving and runoff.

Because of the volumes of ice involved, even small modifications in the mass balance of these ice sheets will be of crucial importance to future changes in the level of the oceans. Obtaining reliable estimates of short-term changes in mass balance does however require climate models which capture adequately the essential features of the current Arctic and Antarctic climate, and snow and ice melt models which can be trusted to estimate melting and runoff accurately. This paper addresses both issues by combining climate and melt models of varying complexity and assessing the reliability of the results. The three melt models which will be used to estimate the runoff from Greenland and the extent of the melt zone in Antarctica are described in section 2, with particular focus on a new approach to capturing changes in the snow cover. The climatic forcing is derived from the output of current climate simulations performed with two climate models of substantially different complexity, the MIT 2D-LO and ECHAM 4 models. Section 3 focuses on the climatology of key variables such as accumulation, temperature and albedo produced by these two models. Section 4 assesses the performance of the combined climate - melt models by comparing the results obtained at individual sites with measured meteorological and ablation data, and by comparing the extent of the ``wet snow'' zone with a similar quantity derived from satellite remote sensing measurements. The estimates of runoff integrated over the entire ice sheets are compared to the best guess derived from field data and discussed in section 5. An outlook on the perspective for modeling future changes in the mass balance of Greenland and Antarctica is found in the conclusion.