This year's Arctic minimum tied for the tenth-lowest on record. The 20 years since 2007 hold the 20 lowest minimums, yet the minimum has shown no significant decline across them — a slowdown scientists are still trying to explain.
Arctic sea ice reached its smallest extent of 2026 on or around 12 September, at 4.60 million square kilometres, according to the US National Snow and Ice Data Center (NSIDC). That ties for tenth-lowest in the satellite record, which began in 1979, alongside 2008, 2010 and 2025. NSIDC treats values within 40,000 square kilometres as tied, and it describes the figure as preliminary.1 NASA, which analyses the same satellite data with NSIDC, reported the same ranking.2
The record remains 2012, when extent fell to 3.39 million square kilometres. This year finished 1.21 million square kilometres above that mark and 1.62 million below the 1981–2010 average minimum — a shortfall NSIDC likens to the area of Alaska.1
Over the full record the direction is not in doubt. Since 1979 the September minimum has fallen by 11.8% per decade relative to the 1981–2010 average, or about 73,200 square kilometres a year.1 A year earlier, with data to 2025, NSIDC put the same rate at 12.1% per decade — the long-run figure eases slightly each year that a new minimum fails to fall further.5
The past 20 years tell a different story. Every minimum from 2007 to 2026 is lower than any year from 1979 to 2006, yet NSIDC says there has been no significant downward trend within those two decades, and that the cause is actively debated.1 A peer-reviewed study published in 2025 put numbers on the slowdown: over 2005–2024, September extent fell by roughly 0.29 to 0.35 million square kilometres per decade, depending on the dataset, and neither trend was statistically significant. The authors describe that as four to five times slower than the fastest 20-year rate of loss, over 1993–2012.4
One complication is that ranks depend on the dataset. The Met Office, using the European OSI SAF index, put the 2026 minimum at 5.06 million square kilometres and 14th-lowest, rather than tenth. Both series agree it sat roughly 1.5 to 1.6 million square kilometres below their own 1981–2010 averages.3
The 2025 study, led by Mark England of the University of Exeter with colleagues at Exeter and Columbia University, concluded that internal climate variability — natural fluctuations in ocean and atmospheric circulation, particularly in the Pacific — has likely offset much of the loss that rising greenhouse gases would otherwise have caused. More than 85% of the climate model runs they examined show a larger forced loss than was observed. The authors note that a forced contribution, for example from changes in Atlantic ocean circulation, cannot be ruled out.4
The same study reports that sea ice volume, which also reflects thickness, has been roughly flat for at least 15 years, with a decline since 2010 about seven times slower than over the whole period since 1979. Volume is not observed directly across the Arctic; it comes from a model-based reconstruction.4
Climate models do produce pauses of this kind, though not often. The figures below are model output, not observations, and individual models disagree widely.
In the authors' simulations, ice eventually resumes its long-term decline, and pauses tend to be followed by faster-than-average loss. They argue a continued pause makes an ice-free Arctic summer less likely in the near term, without changing the longer-run direction.4 That is a modelled scenario, not a forecast this piece endorses.
Why the decline has slowed is unsettled: the leading study attributes it mainly to natural variability, but a forced contribution is not excluded, and NSIDC calls the question actively debated. How long the plateau lasts is unknown; the only estimates are model probabilities with wide disagreement between models. The 2026 figure is preliminary and could be revised, and the rank itself depends on which satellite index is used.