Warm water under the Doomsday Glacier could double the sea-level rise expected from it by 2100

Thwaites Glacier, the Antarctic ice stream nicknamed the Doomsday Glacier, could add as much as 30 millimeters to global sea level by 2100, about twice what earlier projections gave it, according to a study led by Mattia Poinelli. The new model reaches that figure even without any additional warming. It also puts the chance that Thwaites detaches from a key stabilizing ridge before the century ends at about 60 percent.

AI, explained every day: The Deep View is the go-to daily newsletter for 600,000+ founders, knowledge workers and professionals who want to stay up to date with artificial intelligence.

Get The Deep View →

We may earn a commission if you sign up.

Poinelli was a postdoctoral scholar at the University of California, Irvine when the work was done and is now a project scientist at UCLA and a research affiliate at NASA’s Jet Propulsion Laboratory. Eric Rignot, a UC Irvine professor of Earth system science, is a co-author, along with Helene Seroussi, Joshua Cuzzone, Tyler Pelle and Eric Larour. The paper, titled “Sea-level contribution from Thwaites Glacier doubles in an observation-constrained ice-sheet model,” appeared in Nature Communications in 2026.

Seawater in the grounding zone changes the arithmetic

The doubling concerns Thwaites’ own share. The glacier sits in the Amundsen Sea Embayment, which accounts for about one-third of Antarctic Ice Sheet mass loss, and the study asks how much of the sea-level rise expected from this one glacier earlier models were missing. According to the ScienceDaily report on the study, the answer comes from the grounding zone, the stretch where ice lifts off the seabed and ocean water can reach beneath it. Satellite radar data from ICEYE showed that this zone fluctuates across several kilometers, and conventional models treat it as a sharp line.

Letting seawater intrude kilometers under the ice, at melt rates of 90 to 120 meters a year in the best-fitting runs, lets the model reproduce changes observed since the 1990s. Rignot said most projections to date have struggled to match the rapidity of observed mass loss because of missing physics. The ocean water doing the melting is warm enough to cut ice from below at those rates, and the model adds that process to the glacier’s behavior.

The numbers behind the doubling

The detailed figures appear in a technical summary of the paper. A multi-model ensemble run under constant climate forcing averaged 15.2 millimeters of sea-level contribution from Thwaites by 2100, so the roughly 5 kilometer grounding-zone case at 30 millimeters is about double. A 3 kilometer zone gave 19.4 millimeters, give or take 5.2, and the 5 kilometer zone gave 30.1 millimeters, give or take 5.7. The spreads show that 30 millimeters is a result under one scenario and not a ceiling.

For scale, 30 millimeters is about 3 centimeters, a contribution from a single glacier and not total sea-level rise from all sources, which is why the sea-level rise expected from Thwaites, not from the whole planet, is the thing that doubles. Expansion of warming oceans and melt elsewhere add on top.

Why the Mouginot Ridge matters

The 60 percent figure applies to the 4 to 6 kilometer grounding-zone cases; for the 2 to 3 kilometer cases the probability of detaching from the Mouginot Ridge by 2100 is about 45 percent. The ridge sits roughly 950 meters below sea level, and Poinelli said that once it is unpinned the glacier will continually retreat. The western sector has already retreated about 25 kilometers since 1996, the period the simulations were run against, through 2023.

The caveats are substantial. The model holds the ice front fixed, so calving and retreat of the front are excluded, and it represents ocean circulation under the ice with simplified melting rules, not a fully coupled simulation. The study does not predict complete collapse of the glacier this century, and the probabilities belong to the model ensemble and its assumptions. The Nature Communications paper is the record to consult for the calibration, and the institutions behind the work, UC Irvine and NASA’s Jet Propulsion Laboratory, are the places to watch for follow-up. The paper’s DOI is 10.1038/s41467-026-77976-1.

The study also bears on how projections are built. Ice-sheet models are tuned against satellite records of ice speed and thickness, and the Irvine-led team constrained theirs with observations from 1996 to 2023, which is why the paper’s title calls the model observation-constrained. The finding that a physically plausible grounding-zone setup matches the past better than a sharp-line setup suggests that earlier ensembles may have underestimated Thwaites in similar ways, though the fixed ice front and simplified ocean melting leave room for the full answer to move in either direction when later work removes those limits. In Rignot’s account, the missing physics is the reason earlier projections lagged observed mass loss, and the grounding zone is where the new model puts it back in.

Whether the real grounding zone is nearer 2 kilometers or 6 is the unresolved variable, and the spread between those cases is the spread between roughly 19 and 30 millimeters of sea-level rise from this one glacier by 2100.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

AI, explained every day: Get The Deep View →

We may earn a commission if you sign up.