← The Constellation
The Seminar · one field, examined

The Cliff That Might Not Fall

Monday · August 3, 2026 · A decade ago two scientists modeled the edge of Antarctica collapsing fast enough to raise the sea a meter by 2100. Their own field has spent the years since arguing the number down.
I · Seminar

The Cliff That Might Not Fall

glaciology · how fast a doomed glacier comes apart

Ten years ago two scientists built a model in which the edge of Antarctica could come apart fast enough to raise the world’s oceans by a meter within this century. The mechanism was a row of ice cliffs collapsing one after another, like dominoes. Much of their field has spent the years since arguing that the dominoes do not fall that fast, and the alarming number has quietly shrunk, even in the original authors’ own later work. It is a rare chance to watch a science revise one of its most dramatic claims almost in real time.

>1 m2016 projection, by 2100
~34 cmsame authors, revised 2021
65 cmif Thwaites fully goes
≥135 mcliff height needed to shatter

The field in brief

Ice-sheet science studies the two great ice masses, Antarctica and Greenland, and how fast they raise the sea as they shed ice, chiefly by melting and by calving, the breaking-off of icebergs at the margin.[1] Much of West Antarctica is what glaciologists call a marine ice sheet: its bed lies below sea level and slopes downward as you move inland, an arrangement that is inherently unstable.[7] An ice shelf is the floating tongue of a glacier extending over the ocean, and it acts as a cork, holding back the grounded ice behind it through a back-pressure called buttressing.[5]

Two instabilities need to be kept apart. The older and widely accepted one is marine ice-sheet instability: on a bed that deepens inland, once the grounding line begins to retreat, the retreat feeds itself, slowly, over centuries.[7] The contested one is marine ice-cliff instability, MICI for short: once an ice shelf is gone, it can leave behind a tall, unsupported wall of ice that is too high to bear its own weight, so it collapses, exposing a taller cliff behind it, and runs away.[4] The whole fight is about that second mechanism.

The fight

The opening shot came in 2016, when Robert DeConto and David Pollard published a model in Nature that, for the first time, wrote ice-cliff collapse into the equations.[1] Under high emissions it found that Antarctica alone could add more than a meter of sea level by 2100 and more than fifteen meters by 2500, and the result ran on the front page of the New York Times.[1][2] The number carried weight partly because the model was tuned to the deep past, to warm periods when seas stood six to nine meters higher than today.[2]

>1002016with MICI452019with MICI152019no MICI342021revisedcm by 2100
The headline projection for Antarctica’s contribution by 2100, deflating across a decade of studies. The 2016 model included the ice-cliff mechanism; the 2019 reanalysis showed it was not needed to fit the past; the same authors’ 2021 revision brought their own high-emissions figure down to about a third of a meter.[2][3][6]

The first serious rebuttal arrived in 2019, when Tamsin Edwards and colleagues reanalyzed the same model’s own runs and found that the cliff mechanism was not required to reproduce those ancient sea levels; take it out, and the model still fit the record, with a 2100 projection several times smaller.[3]

Then the physics of the cliff itself came under the microscope. Between 2021 and 2024 a group that included Jeremy Bassis, Douglas Benn, and Anna Crawford modeled how an ice cliff actually fails and found it must be far taller than the original study assumed, roughly 135 meters, before it shatters; below that it crumbles gently, and fast-flowing ice tends to thin itself before it ever reaches the dangerous height.[4] Others showed that the jumble of icebergs and sea ice that piles up in front of a cliff, called mélange, can press back and help hold it in place.[5]

The proponents themselves moved. In 2021 DeConto, Pollard, and colleagues published a revised model whose high-emissions figure for 2100 had fallen from over a meter to about thirty-four centimeters, though they added a warning that above roughly three degrees of warming, retreat could still become unstoppable over the longer run.[6]

It broke fully open in 2024. Mathieu Morlighem led a study, its coauthors including the very researchers who had worked out the cliff physics, that reran the Antarctic simulations with the better-constrained failure rule. This time the cliff instability never switched on during this century: when the shelf went, the glacier sped up and thinned, and its cliffs never grew tall enough to collapse.[7] In 2025 the International Thwaites Glacier Collaboration, an eight-year field program on the glacier itself, published its closing verdict: catastrophic cliff-driven collapse this century is less likely than feared, though the glacier will keep retreating and losing ice for centuries.[8]

Where it stands

On the narrow question the papers actually contest, whether ice-cliff instability will drive catastrophic, meter-scale rise this century, the weight of evidence has moved decisively toward the skeptics, and this is not a case for splitting the difference.[7] The rebuttals came from several directions at once, and the mechanics were worked out in part by the same people the 2024 modelers then relied on.[4][7] The most telling fact is that the original authors revised their own high-end 2100 figure downward by roughly two-thirds.[6]

What stays genuinely unresolved is that no one has ever watched marine ice-cliff instability happen; the mechanism lives entirely in models, and how a tall ice cliff really fails remains open.[7] And ruling it out does not make the glacier safe. Thwaites holds about sixty-five centimeters of sea level on its own, the slow instability still grinds forward, and the question has narrowed from whether the coast is doomed by 2100 to which mechanism governs, and on what century’s clock.[8]

Sources
  1. DeConto & Pollard, “Contribution of Antarctica to past and future sea-level rise,” Nature 531:591 (2016) — the paper that introduced ice-cliff collapse into the model; source for the >1 m by 2100 and >15 m by 2500 figures. Primary.
  2. Carbon Brief, “Studies shed new light on Antarctica’s future contribution to sea-level rise” — side-by-side of the 2016 and 2019 numbers and the New York Times front-page reception.
  3. Edwards et al., “Revisiting Antarctic ice loss due to marine ice-cliff instability,” Nature 566:58 (2019) — the reanalysis showing the cliff mechanism is not required to reproduce past sea levels. Primary.
  4. Crawford, Benn, Bassis et al., “Marine ice-cliff instability modeling shows mixed-mode ice-cliff failure,” Nature Communications 12:2701 (2021) — the cliff-failure physics and the ~135 m threshold. Primary.
  5. Schlemm et al., “Ice-cliff failure … ice mélange buttressing,” The Cryosphere 16:1979 (2022) — mélange can exert back-pressure that helps halt the instability. Primary.
  6. DeConto, Pollard, Alley et al., “The Paris Climate Agreement and future sea-level rise from Antarctica,” Nature 593:83 (2021) — the proponents’ revision to a ~34 cm high-emissions 2100 figure and the ~3 °C threshold. Primary.
  7. Morlighem et al., “The West Antarctic Ice Sheet may not be vulnerable to MICI during the 21st century,” Science Advances 10:eado7794 (2024) — the study in which the cliff instability never switches on this century. Primary (open PDF).
  8. International Thwaites Glacier Collaboration, concluding findings (2025) — the field program’s verdict that catastrophic cliff-driven collapse this century is less likely than feared, and the ~65 cm Thwaites figure. Primary.