Ice-core-based reconstructions show that increased snow accumulation on the Antarctic Ice Sheet mitigated global sea level rise by ~ 10 mm during 1901-2000 (Medley and Thomas, 2019). Here, we attribute this trend by evaluating a suite of single-forcing, all-forcing and nudged ensembles from a climate model, along with dynamically consistent reconstructions of sea level pressure, temperature and wind from paleoclimate data assimilation (PDA). The single-forcing ensembles reveal that rising concentrations of greenhouse gasses (GHGs) have been the dominant driver of the historical snow accumulation increase, but acting alone, GHGs would have caused twice the observed increase. We investigate possible explanations for this over-prediction: a) The uncertain cooling effects of anthropogenic aerosols; b) Extreme internal variability; c) Atmospheric circulation trends; and d) Sea surface temperature (SST) trends evident in the PDA reconstructions (and observed SSTs) but not simulated by the model. The latter best explains the spatial and temporal evolution of snow accumulation, including the lack of an Antarctic-wide accumulation increase since 1980. This SST trend pattern resembles the previously modeled response to Antarctic meltwater, and its emergence coincides with the mid-Twentieth-Century onset of ice shelf thinning and retreat of Thwaites and Pine Island glaciers. Aerosols have also damped the accumulation increase and contributed to the global-scale SST pattern, which includes long-term cooling in the central tropical Pacific that cannot be explained by internal variability. Our results imply that including Antarctic meltwater in models would substantially improve projections of Antarctic snowfall, global sea level, and SSTs in the Southern Ocean and tropical Pacific.

The simulation of ice sheet-climate interaction such as surface mass balance fluxes are sensitive to model grid resolution. Here we simulate the multicentury evolution of the Greenland Ice Sheet (GrIS) and its interaction with the climate using the Community Earth System Model version 2.2 (CESM2.2) including an interactive GrIS component (the Community Ice Sheet Model v2.1 [CISM2.1]) under an idealized warming scenario (atmospheric CO2 increases by 1% yr−1 until quadrupling the pre-industrial level and then is held fixed). A variable-resolution (VR) grid with 1/4◦ regional refinement over broader Arctic and 1◦ resolution elsewhere is applied to the atmosphere and land components, and the results are compared to conventional 1◦ lat-lon grid simulations to investigate the impact of grid refinement. An acceleration of GrIS mass loss is found at around year 110, caused by rapidly increasing surface melt as the ablation area expands with associated albedo feedback and increased turbulent fluxes. Compared to the 1◦ runs, the VR run features slower melt increase, especially over Western and Northern Greenland, which slope gently towards the peripheries. This difference pattern originates primarily from the weaker albedo feedback in the VR run, complemented by its smaller cloud longwave radiation. The steeper VR Greenland surface topography favors slower ablation zone expansion, thus leading to its weaker albedo feedback. The sea level rise contribution from the GrIS in the VR run is 53 mm by year 150 and 831 mm by year 350, approximately 40% and 20% smaller than the 1◦ runs, respectively.