In 2002, a large part of the Larsen B ice shelf — one of the freely floating platforms of ice that surround the Antarctic ice sheet — disintegrated in less than six weeks. This allowed the glaciers that previously fed it to flow more quickly to the ocean1. Neither the speed nor the timing of the disintegration was predicted by ice-sheet models used to project future sea-level rise. Glaciologists have spent the past two decades looking at the aftermath of ice-shelf disintegrations, to learn lessons that will help them predict which ice shelf will be the next to fall, and how this will contribute to the discharge of grounded ice to the ocean2,3. Writing in Nature, Lai et al.4 report progress in this area. The authors combined simple theories of fracture formation with machine-learning techniques to determine which portions of an ice shelf are most vulnerable to break-up and most likely to lead to sustained drawdown of the grounded ice sheet on collapse.
Ice shelves restrain the flow of ice from the grounded portions of the ice sheet into the ocean. The boundary between a grounded ice sheet and a floating ice shelf is called the grounding line. The demise of ice shelves around parts of the ice sheet where the underlying bedrock slopes downwards from the grounding-line sheet as it passes beneath the sheet can lead to an irreversible cycle of increased discharge of grounded ice to the ocean (Fig. 1). This cycle is called a marine ice-sheet instability5, and directly contributes to global sea-level rise.
The two main suspects in the ongoing demise of ice shelves are atmospheric and oceanic warming, but atmospheric warming is suspect number one in the collapse of Larsen B. The disintegration of Larsen B and its neighbouring ice shelves was preceded by substantial atmospheric warming that melted the top of the ice, inundating the surface of the shelf with meltwater1. This resulted in the formation of pervasive ponds that filled fractures in the ice (crevasses), creating additional stress that caused the crevasses to deepen. The crevasses are thought to have eventually broken through all the way from the surface to the bottom of the ice sheet, a process called hydrofracturing6. So, can the mapping of existing crevasses on ice shelves be used to assess the likelihood of future collapses?
Enter Lai et al., who have used a machine-learning algorithm to analyse satellite images of all Antarctic ice shelves, and thereby to accurately map the locations of crevasses. This provided an unprecedented data set that reveals where ice is visibly broken and where it remains intact. The authors then used a decades-old theory known as linear elastic fracture mechanics7 to predict where the stress in the ice shelves is large enough to allow fractures to penetrate the entire ice-shelf thickness, and directly compared the results with their data set of where fractures are observed to occur.
The researchers find that the stress pulling apart most water-free surface crevasses is currently too small to allow the crevasses to penetrate the entire ice-shelf thickness. Similarly, the stress in the few parts of the ice shelf at which water regularly accumulates on the surface is often compressive, and therefore prevents the water from squeezing its way through hydrofractures to the bottom of the ice shelf. Under present conditions, these regions of the ice shelf are stable and unlikely to collapse rapidly.
However, as atmospheric temperatures continue to rise, larger portions of the ice shelves are expected to undergo surface melting than at present. Lai et al. find that up to 60% of the area of ice shelves that buttress (block the flow of) the ice sheet could be destabilized if they become inundated with meltwater, as a result of crevasses being filled by the water. Taken together, the authors’ findings pinpoint the portions of ice shelves that are most vulnerable to atmospheric warming, and show that large sections that are currently stable could collapse as atmospheric temperatures continue to rise.
Lai et al. focus on atmospheric warming as suspect number one, but it remains unclear how tightly the fate of ice shelves is tied to suspect number two: oceanic warming. At present, atmospheric temperatures remain too cold over much of the Antarctic ice sheet to promote substantial surface melting6. By contrast, a warming ocean has been linked8,9 to the thinning and retreat of ice shelves in the Amundsen Sea Embayment in West Antarctica. This is especially true for the ice shelves fed by the Pine Island and Thwaites glaciers — warm ocean water is rapidly thinning these shelves and sculpting deep basal channels into their undersides. These channels have been linked to increased fracturing of the ice shelf10, but surface melt can also drain into surface depressions associated with the channels, forming rivers that efficiently remove water from the surface of the ice shelf and thereby prevent widespread inundation of the ice shelf11. What happens on the top of an ice shelf is thus tightly linked to what happens at the bottom.
Increasingly sophisticated models have been used to simulate (or re-enact) the retreat and disintegration of ice shelves in response to atmospheric warming (see refs 2 and 3, for example). However, a deeper understanding of the effects of both the ocean and the atmosphere is needed to accurately predict the fate of ice shelves in a warming climate, because ice shelves are vulnerable to attack from above and below. In other words, the chief suspects in the destabilization of ice shelves do not act in isolation — they are co-conspirators.
Nature 584, 527-528 (2020)