# SPRI study: climate models overestimate Arctic winter sea ice by nearly 1 million km²

- Published: 2026-11-07 17:15 UTC
- Section: Planet › Climate
- Tags: Arctic winter sea ice, climate model bias, March sea ice extent, mid-Holocene sea ice, sediment core algae, sea ice biomarkers, Scott Polar Research Institute, Crystal Fu, sea ice projections, Nature Geoscience
- URL: https://realtech.media/news/rt-02202f072158be7f7d12ad92_spri_study_climate/

**Algae traces in 6,000-year-old Arctic sediments were used to test 27 climate models; the authors say future winter ice loss may be underestimated.**

![SPRI study: climate models overestimate Arctic winter sea ice by nearly 1 million km²](https://realtech.media/media/2026/11/ffa767955a313d40.jpg)
*Earth.com*

A team led by Crystal Fu of the Scott Polar Research Institute reports that climate models overestimate Arctic winter sea ice, according to spri.cam.ac.uk. Dr Matthew Osman of the University of Cambridge and Marco Aquino-López of CIMAT in Mexico are co-authors. The study, published in Nature Geoscience, used traces of algae preserved in Arctic sediments from about 6,000 years ago to evaluate model performance. The authors conclude that models may underestimate future winter ice loss by nearly one million square kilometres [1].

## Algae in seafloor mud as a test

Algae that live inside sea ice leave a distinctive molecule behind, and algae that live in open water leave others. The ratio of the two in a layer of seafloor mud records how much ice covered that spot. The team compiled 46 such records from sediment cores across the Arctic, which they describe as the largest set of its kind.

The reference period was chosen because between 5,000 and 7,000 years ago a change in Earth's orbit gave the Arctic more summer sunlight. The extra heating was comparable in size to what future greenhouse gases are expected to add. The authors converted the ancient Arctic ice simulations of 27 climate models into the same molecular ratios and weighted each model by how well it matched the cores. Related models shared credit.

## What changes in the projections

With these weights, the models' average March ice for the recent past matched the satellite record, and the bias of nearly one million square kilometres disappeared [1]. The spread among models narrowed by up to 85%, about 6.3 million square kilometres, to a range comparable to ordinary year-to-year swings. Weighting by the mid-Holocene alone corrected the models more than weighting by the period 1950 to 1999.

Applied to the models' futures, the same weights lowered projected March ice by 4% to 6% for the 2090s under low, middle and high emissions. According to the authors, the corrected projections show about one million square kilometres less winter ice in every coming decade and under every emissions path [1]. These figures are model projections re-weighted by evidence from the past, not new measurements of present-day ice.

The focus on winter matters because the models' largest disagreements and errors fall in March rather than September, while most testing has concentrated on summer. In March 2025 the Arctic's annual maximum ice extent was the lowest on record. Open water in the Arctic winter releases heat and moisture into the air, which is why the extent of winter ice is of interest beyond the polar region itself.

## Sources

1. [spri.cam.ac.uk](https://www.spri.cam.ac.uk/)
2. [Earth.com](https://www.earth.com/climate/arctic-winter-sea-ice-is-being-overestimated-by-climate-models)

---

Source: RealTech, realtech.media — https://realtech.media/news/rt-02202f072158be7f7d12ad92_spri_study_climate/