SCIScience NO. ---

Thwaites Glacier sea-level rise could reach 30 mm by 2100: study

By modelling the melting band beneath Thwaites Glacier as a zone a few kilometres wide, researchers matched decades of satellite-observed retreat far better. The model projects up to 30 mm of sea-level rise by 2100, even without further ocean warming.

By Rowan Ashby Published 21.10.26 10:47 UTC 7 min read

NASA image of Thwaites Glacier's ice in West Antarctica — NASA / JPL via Wikimedia Commons / Public domain · Source

Key takeaways

A new ice-sheet model from researchers at NASA's Jet Propulsion Laboratory (JPL), UC Irvine, UCLA, Dartmouth College and UC San Diego treats the base of Antarctica's Thwaites Glacier as a melting "grounding zone" a few kilometres wide, not as a single line.
Calibrated against roughly 25 years of satellite observations (1996–2023), the model shows that versions without a grounding zone are about four times less likely to reproduce the glacier's observed retreat.
In its highest well-fitting case, Thwaites adds up to 30 ± 6 mm to global mean sea level by 2100. That is about twice the 15.2 mm average of the ISMIP6 model ensemble run under the same conditions, and it assumes no further ocean warming.
The study puts the chance that the glacier detaches from the stabilising Mouginot Ridge before 2100 at about 45% to 60%, depending on the zone width. In some runs it detaches as early as 2038.
The authors list important limits. The zone width is fixed, the ice front is held in place, and ocean melt is simplified. Outside scientists say the processes under the glacier are still poorly understood.

What the study did

Thwaites Glacier sea-level rise could reach up to 30 ± 6 mm by 2100, about twice the 15.2 mm ISMIP6 average, according to a NASA JPL-led ice-sheet model published in Nature Communications on 12 October 2026. It assumes no further ocean warming. NASA's sea-level data is also tracking El Niño 2026, which closely resembles the record 1997-98 event.

The team used the Ice-sheet and Sea-level System Model (ISSM), which was developed at JPL. Their focus was a detail that most continental-scale models simplify. A glacier like Thwaites rests on bedrock and then lifts off to float on the ocean. Models usually draw that transition as a sharp boundary called the grounding line. Satellite radar studies over the past decade suggest the real picture is messier. Tides flex the ice up and down, and seawater can push several kilometres inland beneath ice that is still technically grounded, melting it from below.

Earlier UC Irvine-led research helped establish this. In May 2024 the university reported a study in the Proceedings of the National Academy of Sciences that used commercial ICEYE radar satellite data from March to June 2023. It described warm, pressurised seawater intruding kilometres beneath the grounded ice of Thwaites. The new paper takes that observation and builds it into a forward-looking ice-flow model.

Why grounding zones matter

In the model, the grounding zone is a band of a set width in which ocean melting is allowed to act on grounded ice. The researchers did not choose the width or the melt rate by hand. They ran a Bayesian calibration that started from 10,000 prior samples and used a grid of 105 simulations. Each run was compared with what satellites recorded between 1996 and 2023: retreat of the grounding line on the main trunk and the Eastern Ice Shelf, ice discharge, thinning, and surface speed-up. The observations came from records including ICESat, ICESat-2 and CryoSat.

The best fits clustered around grounding zones 2 to 5 km wide, with melt rates of 90 to 120 metres of ice per year. The single most likely width was about 3 km. Runs that used a traditional sharp grounding line, with no zone, were about four times less likely to reproduce the historical trend. NASA's summary describes this as a markedly closer match to satellite-observed retreat since the 1990s. Averaged across the cases studied, adding any grounding zone raised projected sea-level contribution by 51% compared with the traditional approach.

The numbers: projections by grounding-zone width

For its projections the team held ocean and climate forcing steady at the 1996–2023 average and applied a melt rate of 110 m per year. The table shows how the 2100 result depends on the width of the zone.

Grounding-zone width	Projected sea-level contribution by 2100	Chance of detaching from Mouginot Ridge by 2100
None (sharp grounding line)	8.3 ± 3.0 mm	No retreat past the ridge
1 ± 1 km	14.3 ± 3.1 mm	No retreat past the ridge
3 ± 1 km (most likely width)	19.4 ± 5.2 mm	About 45% (2–3 km cases)
5 ± 1 km	30.1 ± 5.7 mm	About 60% (4–6 km cases)
ISMIP6 ensemble mean, same forcing (reference)	15.2 ± 21.8 mm	n/a

Source: Poinelli et al., Nature Communications (2026), open-access full text via PubMed Central. "±" values are the uncertainties reported by the authors.

The headline figure therefore needs careful reading. "Double" refers to the 5 km case, the upper end of the best-fitting range. It produces about 30 mm, roughly twice the ISMIP6 average. The statistically most likely 3 km case gives 19.4 mm. That is lower than 30 mm but still about 30% above the ensemble mean. Every case that includes a grounding zone lands in the upper half of the earlier ensemble's wide uncertainty range.

The most important difference between the cases is the Mouginot Ridge, a seafloor high that currently pins part of the glacier in place. With no zone, or a narrow one, the modelled glacier stays on the ridge through 2100. With wider zones, melting at the base speeds up the "unpinning." The mean year of detachment is 2088 for a 2 km zone and 2065 for a 3 km zone. At 5–6 km it can happen as early as 2038. Once the ice comes free of such a ridge, retreat into deeper basins can speed up.

Warming oceans are reshaping ecosystems as well as ice sheets. The first global IUCN Red List of Ecosystems assessment of coral reefs rates 96% of them Critically Endangered and finds the Western Galápagos system already collapsed, with marine heatwaves the main driver; see our explainer on the IUCN coral reef assessment.

What it means

Thirty millimetres, about 1.2 inches, sounds small. But it comes from one glacier, over about 75 years, without any extra ocean warming. The paper puts the total sea-level equivalent of Thwaites at about 65 cm. In the model, adding 0.5–0.7 °C of ocean warming raised projected ice loss by about another 15%. So the result is best read as a guide to how sensitive projections are to a single process, not as a worst case.

For planners, the main point is about method. If projections that rely on sharp grounding lines systematically underestimate retreat, then regional sea-level estimates used for coastal defences may also be biased low. Rignot told CNN the results "predict greater risk of and exposure for coastal flooding than anticipated in the coming decades." Ted Scambos of the University of Colorado Boulder, who was not involved, described the work to CNN as potentially "a real gamechanger for sea level forecasting from Thwaites."

Remaining uncertainties

The authors list several simplifications that could push results in either direction:

Fixed zone width: The grounding-zone width is a time-averaged constant. It does not vary with bed slope, tides or water pressure beneath the ice, and the authors call for a version that evolves over time.
No calving: The ice front is held fixed by design, so calving and ice-front retreat are not modelled.
Simplified ocean: Melt follows a depth-based formula, not a full ocean-circulation model. Changes in the depth of warm water layers and year-to-year variability are not included.
Bed and friction: Friction at the bed does not change over time, subglacial water flow is not included, and bed topography remains uncertain. The model overestimates Eastern Ice Shelf retreat by about 15 km.
Sparse data: Direct measurements beneath the ice shelf are rare, which limits how tightly melt rates can be constrained.

Independent researchers urged caution. Robert Larter, a marine geophysicist at the British Antarctic Survey, told CNN that "there is a lot we still need to find out about grounding zone processes." He added that conditions beneath Thwaites are more complex than the paper portrays. According to CNN's report, the team plans to deploy instruments at the glacier in 2028 to gather the kind of field data that could test the model's assumptions.

Frequently asked questions

What is a grounding zone?

It is the area where a glacier moves from resting on bedrock to floating on the ocean. Seawater can reach several kilometres beneath grounded ice there and melt it from below. Most models simplify this area as a single line.

How much could Thwaites Glacier add to sea level by 2100?

The study projects 8.3 mm with no grounding zone and up to 30.1 ± 5.7 mm with a 5 km zone. Its most likely 3 km case gives 19.4 ± 5.2 mm. All of these assume ocean conditions stay at their 1996–2023 average.

Why does the new model say 'double'?

Under the same forcing, the ISMIP6 multi-model ensemble averages 15.2 mm by 2100. The upper well-fitting case in the new study, about 30 mm, is roughly twice that figure.

What is the Mouginot Ridge?

It is a seafloor ridge that currently pins part of Thwaites Glacier in place. The study estimates a 45% to 60% chance that the glacier detaches from it before 2100, depending on the grounding-zone width. In some runs this happens as early as 2038.

How certain are the results?

The authors note several simplifications: a fixed grounding-zone width, no calving, simplified ocean melt, and sparse data from beneath the ice. Independent scientists say much about grounding-zone processes remains unknown.