Two months on, with the season's data complete, the Arctic's 2026 sea ice minimum stands at 4.60 million square kilometres, reached on 12 September. NSIDC ranks it tied for 10th lowest in 48 years of satellite records. All 20 of the lowest minimums came in 2007–2026. The winter maximum is still falling: 2026's, 14.29 million km², is tied with 2025's for the lowest in the 1979–2026 record.
Summer losses have been slower since 2007, but how much slower depends on where the trend starts: from 2007 the minimum's trend is −0.03 million km² per decade, from 2005 −0.26, against −0.73 over the full 1979–2026 record.
What these numbers are
Sea ice extent is the area of ocean where at least 15% of the surface is ice, measured by satellite since October 1978. The figures here are from the US National Snow and Ice Data Center's Sea Ice Index, the series behind NSIDC's announcements of each year's minimum and maximum. Following NSIDC, each day's value is averaged with the four days before it, and a year's minimum and maximum are the lowest and highest of those 5-day averages. The minimum falls in September every year in this record; the maximum fell in March in 40 of the 48 years, in February in 7 and in April once. Extent says where the ice is, not how thick it is; this data says nothing about ice volume.
Since 1 January 2025 the index has used a different satellite instrument, AMSR2. NSIDC's analysis of the change found that for January–June 2025 the new Arctic extents differed from the old by 0.05 million km² on average and by less than 0.2 at most, mostly but not always lower. Both 2025 and 2026 are measured this way.
A step down, then a slower decline
Before 2007 the September minimum fell at 0.54 million km² per decade. The lowest year before 2007 was 2005, at 5.32 million km². Every year since has been lower. As NASA's Earth Observatory put it, each of the past 20 summers ranks among the 20 lowest.
The average minimum was 4.49 million km² in 2007–2016 and 4.45 in 2017–2026, against 7.00 in 1979–1988. The 2007–2026 trend is −0.03 per decade, with a 95% range of −0.38 to +0.32.
That figure is the flattest of the nearby start years: from 2004 the trend to 2026 is −0.36 per decade, from 2005 −0.26, from 2006 −0.20 and from 2008 −0.09. We selected 2007 after inspecting the record; these fits do not establish a unique breakpoint or a permanent plateau. The 1979–2026 minimum trend is −0.73 million km² per decade. A recent trend indistinguishable from zero does not demonstrate that ice loss or human influence has stopped. The intervals here assume independent regression errors and exclude breakpoint-selection and instrument uncertainty.
Where 2026 ranks
Sorting the published figures puts 2026 11th of 48, a few thousandths of a million km² below 2025 and above 2008. NSIDC treats extents within 0.04 million km² of each other as ties, and ranks the 2026 minimum tied for 10th with 2008, 2010 and 2025. The lowest minimum in the 1979–2026 record is still 2012's, at 3.39 million km².
Instrument differences vary by day, so there is no simple correction. As an illustrative sensitivity test, not a correction or uncertainty bound: adding 0.1 million km² to both AMSR2 years, 2025 and 2026, would place 2026 14th by numerical ordering. On a different index, from the EUMETSAT OSI SAF, the 2026 minimum was 5.06 million km² and 14th lowest, according to the Met Office.
Winter ice is still shrinking
The fitted winter decline is −0.31 million km² per decade for 2007–2026 (95% range −0.54 to −0.08), against −0.44 in 1979–2006; excluding 2025–2026 gives −0.21, with an interval including zero (−0.49 to +0.06). The average maximum went from 16.18 million km² in 1979–1988 to 14.96 in 2007–2016 and 14.66 in 2017–2026; the 2017–2024 average, before the new instrument, was 14.75.
The 2026 maximum was 14.29 million km², on 15 March, statistically tied with 2025's 14.31 for the record low; NSIDC treats differences within 0.04 million km² as ties. Both years use AMSR2. They are 0.12 and 0.10 million km² below 2017's 14.41, the lowest maximum before the instrument change. NSIDC's March monthly averages put the two years the other way round, with March 2025 the lowest.
2026, day by day
Through 6 November 2026, daily extent was below the middle 80% of 1981–2010 values on 306 of 310 days. On 12 September it was 1.83 million km² below that period's median for the date. By the standard of the past 20 years 2026 was an ordinary year; against 1981–2010 it was below the usual range almost every day.
Why, according to others
This data shows the slowdown, not its cause. M. R. England, L. M. Polvani, J. Screen and A. C. Chan found "no statistically significant decline in September sea ice area since 2005", robust across the NSIDC and OSI SAF indices, and that climate models produce similar periods while greenhouse gas emissions rise. They conclude that natural variability has substantially offset human-driven loss, that the slowdown could plausibly persist another five to ten years, and that the chances of a faster-than-average decline are increased in the near future. They stress that it does not mean human-caused change has stopped, and that the ice would likely have grown over the period without human influence. Their study ends in 2024; on this data, with 2025 and 2026 added, the 2005–2026 trend in the annual minimum, −0.26 per decade, is also not statistically significant.
How this was made
The data is NSIDC's Sea Ice Index, Version 4 (Fetterer, F., Knowles, K., Meier, W. N., Savoie, M., Windnagel, A. K. and Stafford, T., 2025, doi:10.7265/a98x-0f50), Northern Hemisphere daily extent from 26 October 1978 to 6 November 2026, retrieved on 8 November 2026. fetch.mjs downloads the daily file, the 1981–2010 daily climatology and NSIDC's own table of annual minima and maxima, used as a check; calculate.mjs computes the annual minima and maxima and every number in this post (statistics.json); arctic-sea-ice-make-charts.mjs builds the charts, and re-running it reproduces them exactly. From 1988 on, the computed minima and maxima match NSIDC's table to 0.001 million km²; the 20 lowest years are the same. Method, choices and licence: DATA.md and PROVENANCE.md. From the download folder, run:
npm ci
node arctic-sea-ice-src/calculate.mjs
node arctic-sea-ice-make-charts.mjs
Credit: Sea Ice Index, National Snow and Ice Data Center.