Rapid formation of an ice doline on Amery Ice Shelf, East Antarctica

Surface meltwater accumulating on Antarctica’s floating ice shelves can drive fractures through to the ocean and potentially cause their collapse, leading to enhanced ice discharge from the continent. Surface melting in Antarctica is predicted to increase significantly during coming decades, but the implications for ice shelf stability are unknown. We are still learning how meltwater forms, flows and alters the surface, and that rapid water-driven changes are not limited to summer. The southern Amery Ice Shelf in East Antarctica already has an extensive surface meltwater system and provides us with an opportunity to study melt processes in detail. We present high-resolution satellite data (imagery, ICESat-2 altimetry and elevation models from WorldView stereo-photogrammetry) revealing an abrupt change extending across ~60 km² of the ice shelf surface in June 2019 (midwinter). We interpret this as drainage of an englacial lake through to the ocean below in less than three days. This left an uneven depression in the ice shelf surface, 11 km² in area and as much as 80 m deep, with a bed of fractured ice: an “ice doline”. The englacial lake had lain beneath the perennially ice-covered portion of a 20 km² meltwater lake. The reduced mass loading on the floating ice shelf after the drainage event resulted in flexure, with uplift of up to 36 m around the former lake. Applying an elastic flexural model to the uplift profiles suggests the loss of 0.75 km³ of water to the ocean. In summer 2020, we observed meltwater accumulating in a new lake basin created by the flexure. ICESat-2 observations profiled a new narrow meltwater channel (20 m wide and 3 m deep), rapidly incised inside the doline as meltwater spilled over from the new lake and started refilling the depression. This study demonstrates how high-resolution geodetic measurements from ICESat-2 and WorldView can explore critical fine-scale ice shelf processes. The insights gained will greatly improve our ability to model these processes, ultimately improving the accuracy of our projections.