West Antarctica lost its Amundsen Sea ice in past warm periods, seabed clay suggests

In deep-sea drill cores, a clay mineral that comes mostly from beneath Thwaites Glacier falls from 12% of the clay today to under 5% in every earlier warm period of the past 800,000 years, which the authors read as repeated loss of that ice, worth at least about 1.5 m of sea level.

Peer reviewed · Sediment-core proxy study · Interpretation compared with published models

The part of West Antarctica that drains through Thwaites and Pine Island glaciers appears to have disappeared during each earlier warm period of the past 800,000 years, according to a study published in Nature Communications on 9 November 2026. Claus-Dieter Hillenbrand of the British Antarctic Survey and 37 co-authors traced a clay “fingerprint” in seabed mud and found it almost vanished in those warm spells. That matters because this sector is losing ice fastest today. The evidence is indirect, and it cannot say how much of the wider ice sheet went with it.

Evidence at a glance

Is this a big deal?

It is a solid step, not the final word. Sea-level records from far away and computer models have long hinted that West Antarctica shrank in past warm periods, but the paper itself notes that direct evidence from near the ice “has remained inconclusive.” This is a record taken right off the coast, and it repeats across many warm periods. What it can’t do is settle the bigger question of whether the whole ice sheet collapsed, so it narrows the debate rather than ending it.

What scientists found

Today, mud on the seabed off the eastern Amundsen Sea is rich in kaolinite. In the modern seabed sample nearest drill site U1533 it makes up 12% of the clay, and it stays at 8% or more across at least 400 km around both drill sites.

In sediment laid down during warm interglacial periods, the picture flips. At U1533, kaolinite fell below 5% in every interglacial before the present one, back to 800,000 years ago, and in some layers it was almost absent, under 2%. The second site, U1532, shows the same pattern more weakly, because its warm-period layers are thin (about 1 cm per 1,000 years). The authors interpret the drop as a halt in erosion beneath Thwaites Glacier, which requires its catchment to have lifted off its bed and largely melted. Pine Island Glacier also supplies kaolinite (14–19% of its clay), so they infer it retreated a long way too.

How they found it

In the austral summer of 2019 the drill ship JOIDES Resolution cored two sites on Resolution Drift, a mound of sediment on the deep sea floor off the eastern Amundsen Sea. U1533 sits under 4,180 m of water and U1532 under 3,960 m, 62 km apart. Mud piles up slowly there: at U1533, the past 773,000 years fit into less than 20 m of core.

The trick is that kaolinite works like a postmark. It forms only by chemical weathering in warm, acidic soils or swamps, so in Antarctica it comes from rocks older than the widespread glaciation that began about 34 million years ago. Off the eastern Amundsen Sea, the team’s map of 63 newly reported and many older seabed samples showed that erosion of the bed underneath the Antarctic ice at Thwaites Glacier is the most important source. When the postmark stops showing up offshore, the glacier that stamped it has stopped grinding its bed. The analogy has a limit: a missing postmark tells you a sender went quiet, not how far the sender moved.

The clay in 200 samples from U1533, 102 from U1532 and 64 from a nearby piston core was measured by X-ray diffraction at the University of Leipzig. A team at Yonsei University in Seoul then measured the chemistry of individual smectite clay grains with an electron microscope. Warm-period layers held magnesium-rich smectite that comes from the deep sub-Antarctic Pacific to the north, while cold-period layers appeared to hold only the low-magnesium type found on the Antarctic shelf. That points to more warm Circumpolar Deep Water, a relatively warm deep current, flowing toward the ice in interglacials.

A third line came from diatoms, single-celled algae whose glass shells change shape with temperature. Using shells of Fragilariopsis kerguelensis, the team estimated past summer surface temperatures at U1533 between 0.6–1.4 °C (MIS 7, 243,000–191,000 years ago) and 3.3–4.1 °C (MIS 13, 533,000–478,000 years ago) warmer than today. MIS stands for Marine Isotope Stage, the standard numbering of Earth’s past cold and warm periods.

Why it matters to you

This won’t change anything for you this year. It matters for coastlines over centuries. The authors estimate that losing the Amundsen Sea sector would have raised global sea level by at least about 1.5 m. For scale, the paper cites a West Antarctic contribution of about 7.5 mm from 2003 to 2019, so 1.5 m is 200 times that amount.

The authors also note that the last interglacial, MIS 5e (130,000–115,000 years ago), was about 1.0–1.5 °C warmer than pre-industrial times globally, and the paper treats it as a key analogue for a near-future world. They argue that warming of that size “can constitute a tipping point” for major West Antarctic ice loss. That is their interpretation of past events. The paper does not forecast when, or whether, the same thing will happen this time.

What the evidence does not show

The clay record shows that mud from Thwaites and Pine Island stopped arriving. It does not measure ice thickness, extent or timing directly. The authors write that their results “do not reveal the full extent” of retreat and cannot confirm or rule out model estimates of a 3–4 m Antarctic contribution to sea level in the last interglacial.

Some ice almost certainly survived. Clay from the western Amundsen Sea kept arriving from the Marie Byrd Land highlands, and illite chemistry at U1533 suggests ice caps persisted in the Thurston Island area. The Weddell and Ross Sea sectors may have survived as well. The ages come from lining up productivity peaks with a global reference record of past climate, checked against magnetic records, so the timing within each warm period is approximate. Two other suggestions in the paper, a late peak in retreat during MIS 5e and the link to deep-water warming, are interpretations rather than measurements.

How this fits what we already knew

Ice-sheet models disagree widely about the last interglacial. Some lose the whole West Antarctic Ice Sheet, some lose only certain sectors, and some keep it close to today’s size. Earlier evidence from ice cores, seabed records and genetic data that point to a past seaway across West Antarctica was described in the paper as “inconclusive or disputed.”

The new data favour the models in which the Amundsen Sea sector goes first, driven by ocean heat. They also agree with model runs, cited in the paper, in which an ocean-forced Amundsen Sea collapse added about 3 m of sea level from Antarctica around 116,000 years ago.

Independent perspective

SciQuest has not yet obtained comment from a researcher outside the author team, and the request is logged. Within the published literature, the strongest counterweight is an earlier study of the same two drill sites that used neodymium isotopes and concluded that Pleistocene retreat of the ice sheet was rather limited. The authors say their findings may not rule out survival of the Weddell and Ross Sea sectors. A June 2026 Geophysical Research Letters paper by an overlapping team, using iceberg-dropped mineral grains from two other Amundsen Sea cores, also inferred a lower ice sheet in MIS 5, 11 and 15.

How other headlines framed it

At the time of SciQuest’s check (08:10 UTC, 15 November 2026), no news outlet coverage of this paper could be found, so there is no framing to compare yet.

What happens next

The open question is how far the ice went. The authors point to provenance tools from their Pliocene work on the same cores, neodymium and lead isotopes and argon dating of single mineral grains, which traced retreat all the way to the Ellsworth-Whitmore Mountains during the warmer Early to Mid-Pliocene. Applying those finer fingerprints to the Pleistocene layers could show whether the ice retreated that far in recent interglacials too. The clay data are public at the NERC Polar Data Centre for other teams to test. For how glaciers that end in the ocean are losing ice today, see why glaciers ending in lakes or the sea account for half of glacier ice loss.

In one sentence

A clay mineral that comes from under Thwaites Glacier nearly disappears from deep-sea mud in every earlier warm period of the past 800,000 years, which suggests the Amundsen Sea part of West Antarctica melted away repeatedly, while leaving open how much of the rest went with it.

Read next

• What is underneath the ice in Antarctica?
• Glaciers ending in lakes or the sea: 3% of glaciers, half of the ice loss
• Arctic winter sea ice may shrink more than models project, ancient clues suggest

Sources and disclosure

1. Hillenbrand, C.-D., Ehrmann, W., Park, Y. K. et al. Repeated loss of the West Antarctic Ice Sheet’s Amundsen Sea sector during late Pleistocene interglacials. Nature Communications 17, 10450 (2026), published 9 November 2026. SciQuest read the full open-access text, figure legends, methods and data statement.
2. Simões Pereira, P., Hillenbrand, C.-D., Hemming, S. R. et al. Reduced West Antarctic Ice Sheet size during prominent Quaternary interglacials constrained by iceberg-rafted debris provenance in the Amundsen Sea. Geophysical Research Letters (2026), published 24 June 2026 (abstract read).
3. NERC Polar Data Centre. Clay mineral data, IODP Hole U1533A (one of four datasets listed in the paper).

The authors declare no competing interests. Funders listed with the paper include the UK Natural Environment Research Council and the US National Science Foundation. Seabed samples near Thwaites came from the International Thwaites Glacier Collaboration’s THOR project.

Disclosure: This article was drafted with AI assistance and checked against the sources above. The graphic plots only values stated in the paper; “below” values are drawn at their upper limit, and the 200-times comparison is SciQuest’s arithmetic (1.5 m ÷ 7.5 mm). Spotted an error? See our corrections policy.