The added value of high resolution in estimating the surface mass balance in southern Greenland

Abstract

Abstract. The polar version of the regional climate model RACMO2, version 2.3p1, is used to study the effect of model resolution on the simulated climate and surface mass balance (SMB) of south Greenland for the current climate (2007–2014). The model data at resolutions of 60, 20, 6.6, and 2.2 km are intercompared and compared to SMB observations using three different data refinement methods: nearest neighbour, bilinear interpolation, and a statistical downscaling method utilising the local dependency of fields on elevation. Furthermore, it is estimated how the errors induced by model resolution compare to errors induced by the model physics and initialisation. The results affirm earlier studies that SMB components which are tightly linked to elevation, like runoff, can be refined successfully, as soon as the ablation zone is reasonably well resolved in the source dataset. Precipitation fields are also highly elevation dependent, but precipitation has no systematic correlation with elevation, which inhibits statistical downscaling to work well. If refined component-wise, 20 km resolution model simulations can reproduce the SMB ablation observations almost as well as the finer-resolution model simulations. Nonetheless, statistical downscaling and regional climate modelling are complementary; the best results are obtained when high-resolution RACMO2 data are statistically refined. Model estimates in the accumulation zone do not benefit from statistical downscaling; hence, a resolution of about 20 km is sufficient to resolve the majority of the accumulation zone of the Greenland Ice Sheet with respect to the limited measurements we have. Furthermore, we demonstrate that using RACMO2, a hydrostatic model, at 2.2 km resolution led to invalid results as topographic and synoptic vertical winds exceed 10 m s−1, which violates the hydrostatic model assumptions. Finally, additional tests show that model resolution is as important as properly resolving spatial albedo patterns, correctly initialising the firn column, and uncertainties in the modelled precipitation and turbulent exchange.