Since the early 2000s, GRACE and GRACE-FO satellite missions monitor mass change of the Antarctic ice sheet (AIS) from space. They show a clear mass loss trend, that is expected to continue under future climate warming. Mass loss is concentrated in coastal West Antarctica and results from increased ice discharge (D) across the ice sheet grounding line (blue contour in Fig. 1), where land ice transitions into floating ice shelves. Satellites also detect large year-to-year mass fluctuations, that are driven by the ice sheet surface mass balance (SMB), i.e., the difference between mass gained from snowfall, and mass lost from sublimation and meltwater runoff to the ocean.
Although most meltwater refreezes in the cold Antarctic snow, thus not significantly contributing to mass loss, floating ice shelves are vulnerable to surface melt (Fig. 1). The reason is that meltwater can trigger a destructive mechanism called hydrofracturing, i.e., when meltwater fills pre-existing crevasses and pushes them further open. When ice shelves break up, this reduces the buttressing effect on the grounded ice sheet, further enhancing mass loss.
Surface processes are therefore very important to understand and predict ice sheet mass change. However, our knowledge of these processes largely relies on climate models that typically run at 25-100 km spatial resolution (Fig. 2A-B), too coarse to capture mountain ranges of e.g., the Antarctic Peninsula, or the steep ice sheet grounding lines. To address this, we present a statistical downscaling procedure that refines a low-resolution regional climate model at 27 km to a high-resolution 2 km grid, resolving rugged terrains in much greater detail (Fig. 2C).
The downscaling procedure enhances orographic snowfall over steep mountain ranges, that were spatially smoothed in the low-resolution climate model (Fig. 2B-C). This pattern is clearly visible over the Antarctic Peninsula (Fig. 3).
When integrated over the grounded ice sheet, snowfall and thus the surface mass balance (SMB) increase by 2%, sufficient to reconcile satellite-observed mass change (MB = SMB - D) from GRACE and GRACE-FO sensors (Fig. 4).
The downscaling procedure also increases surface melt by no less than 46%, notably near grounding lines that were not well resolved at low resolution (Fig. 1). This is clearly visible across the grounding line of Larsen C ice shelf fringing the Antarctic Peninsula (Fig. 5).
We applied statistical downscaling to model runs until 2100 using climate projections from the global climate model CESM2 (Fig. 3A) under a low-end, moderate and high-end warming scenario (Fig. 6). Our results show that the higher surface melt near grounding lines persists under all warming scenarios, further endangering the stability of vulnerable ice shelves.
Our study highlights the importance of capturing snowfall and surface melt with high spatial detail to yield reliable estimates of Antarctic ice sheet mass change. Until models can be run at such high resolutions, statistical downscaling is a powerful tool to better resolve and understand these surface processes.