Ancient Algae in Arctic Mud Sharpen the Forecast for Winter Sea Ice

Three researchers have used chemical traces of algae buried in Arctic seabed mud to judge which climate models to trust on winter sea ice, and report that doing so narrows the uncertainty around the models' combined projection by 85%.

The work, published Oct. 16, 2026, in Nature Geoscience, was aimed at a gap the authors describe in how models get graded. The satellite record is short and strongly shaped by year-to-year swings, so a model can look skillful on it without capturing how sea ice responds to sustained warming. The mid-Holocene, a warm spell 5,000 to 7,000 years ago when more summer sunlight reached the Arctic, offers a test the models were never fitted to.

The team compiled 46 measurements from Arctic seabed cores, each pairing a molecule made by ice-dwelling algae with one made by open-water plankton, a ratio that records how much ice sat above the site. A calibration model converted the ice cover simulated by 27 climate models into the same units, so the two could be compared directly. The paper reports that weights that bring the models' combined answer onto the cores also bring it onto the observed March ice extent, and remove a long-standing tendency in the group to simulate too much ice.

Applied to the same models' projections, those weights shift the winter outlook downward. The paper states that future winter sea ice extent may be nearly 1 million square kilometers below prior estimates, across coming decades and under every emissions scenario tested. The 85% describes the precision of that estimate and not an amount of ice. No new observation of melting is involved; what moved is the projection.

The figure that reaches the news each year is the September minimum, the low point at the end of summer, while this work is about March, when Arctic ice is at its widest. The authors say winter has been passed over in the effort to pin down the first ice-free summer. In the weighted projections, the greatest losses fall in the Barents, Greenland and Labrador Seas. The paper is open access.