Antarctic Sea Ice Hits Its 2026 Maximum: Reading the Numbers

Antarctic sea ice likely peaked on 14 September 2026 at 17.59 million square kilometers, according to NASA and the National Snow and Ice Data Center. That ranks as the third-lowest winter maximum since satellites began watching in 1978 (2023 and 2024 were lower). NASA Earth Observatory published the finding this month with an updated map and extent chart.

The number is the easy part. The harder part is knowing what kind of number it is. Two agencies, using different sensors and different filters, put the peak on different days and at different sizes, and both are correct for what they measure. Then there is what happened afterward: by early October the ice was shrinking faster than in any previous year on the same calendar date.

What happened

NASA’s account, built on NSIDC data, says the extent reached its annual maximum on 14 September at an estimated 17.59 million square kilometers (6.79 million square miles). Carbon Brief reported the same date and value on 25 September, describing the figure as provisional NSIDC data, and carried NSIDC’s warning that big swings are normal in Antarctic ice near the seasonal peak.

The season did not run in a straight line. NSIDC’s Sea Ice Today analysis says extent stayed below the lowest interdecile range from the start of the growth season in late March, yet remained above the 2023 record low for that stretch. September averaged 17.32 million square kilometers in the NSIDC record. From 2 October, daily extent dropped to record-low values for the date, passing 2023.

Region by region, the picture was patchy. Post-maximum losses showed up in the Ross and Amundsen Seas and in the Indian Ocean sector. A Korea Polar Research Institute scientist quoted by Carbon Brief said the Bellingshausen Sea was nearly ice-free in April but had largely recovered its deficit by late August, while the Amundsen Sea and much of East Antarctica held unusually little ice.

The numbers in one place

Every value below comes from a source I opened. Where two datasets disagree, both appear, because the disagreement is part of the story.

How a maximum is defined, and why the date moves

Extent is not the area of ice. It is the total area of grid cells with at least 15 percent ice concentration, a threshold NASA says scientists commonly use. A cell that is 16 percent covered counts fully. The Sea Ice Index derives this from passive microwave data, and the instruments see through cloud because microwaves pass through it.

Next comes the averaging. According to NSIDC’s data notes, NSIDC announces its minimums and maximums using a five-day trailing average, the given day plus the four before it, specifically to smooth out storm-driven false retrievals and coastal or surface effects. A single-day spike therefore never becomes the official peak. Before April 2012 the average was centered instead of trailing, which shifts dates by two days when old and new records are compared.

Different sensors and algorithms add another layer. The Met Office and Copernicus numbers rest on the EUMETSAT OSI SAF Sea Ice Index, version 3.0, not on NSIDC’s product. As of its 21 September briefing, the Met Office had a daily maximum of 18.12 million square kilometers on 11 September, and it cautioned that many years have set maxima after 20 September, some as late as early October. Copernicus later put the annual daily maximum at 18.19 million square kilometers on 10 September.

So the same winter has a 10 September peak, an 11 September peak and a 14 September peak, depending on whose ruler is used. The gap of roughly half a million square kilometers between NSIDC and OSI SAF figures is a difference of product, not an error by either group. That gap is my arithmetic from the published values, and I did not find an agency explaining its cause, so treat the cause as unverified. What does hold across both is the ranking: Copernicus and the Met Office place 2026 behind only 2023 and 2024, and NSIDC also calls it third lowest.

Against the long record and the post-2016 slide

NASA’s account of the long arc is blunt about its own uncertainty. From 1979 to 2014 total Antarctic sea ice grew by about 1 percent per decade, though NASA says it is unclear whether that was meaningful change given how much yearly extent varies. Satellites logged new winter highs from 2012 to 2014, just as Arctic ice was falling steeply.

Then it turned. The growth ended in 2015, and significant decreases began in 2016. Some years rebounded a little, but coverage has mostly stayed below normal. Both the record-low minimum and the record-low maximum fell in 2023, and the 2026 maximum sits close behind them.

On the monthly measure, Copernicus puts September 2026 at 17.8 million square kilometers, 1.2 million below the 1991–2020 average, or 6.3 percent, the third-lowest September in 48 years. I could not confirm a single-number comparison of the 2026 peak against the 1981–2010 average, so I have not given one. NSIDC’s charts show the 1981–2010 line and the 2026 curve below it for most of the season, which is a visual reading, not a published figure.

One more data point: Carbon Brief reports July and August 2026 extents as the fifth- and fourth-lowest on record, and notes a near-average February minimum before winter coverage fell back to low levels. A single near-normal summer did not reverse the drift.

Why the Antarctic does not behave like the Arctic

Geography does most of the explaining, and this part is established. The Arctic is an ocean basin ringed by land. Antarctica is a continent ringed by open ocean. That leaves Antarctic ice free to spread outward each winter, but it also carries the ice into warmer latitudes, where much of it melts each summer. NASA notes the ice usually peaks in September and bottoms out in February.

The consequence is a different rhythm. Compare the Arctic, which is a slower, steadier story. This year’s Arctic minimum, 12 September, tied for tenth lowest at 4.81 million square kilometers for the month, 1.6 million below the 1981–2010 mean. That is a steady downward trend. Antarctica swung from record highs to record lows in roughly a decade. For how the two poles and the agencies line up on the northern side, see why agencies rank the 2026 Arctic minimum differently.

Where scientists still disagree

Facts first. The extent figures, the dates and the rankings above are measurements. The trend break after 2015 is also in the record. Everything after that is interpretation.

The main open question is whether the post-2016 decline is a lasting shift or a long swing in natural variability. NASA says scientists will keep observing each year to decide whether it is a long-term change or a statistical fluctuation. Lettie Roach of the Alfred Wegener Institute told Carbon Brief the attribution to human-caused warming is less clear than in the Arctic, and that more years of data are needed before calling it a structural shift.

Short-term drivers draw less argument. Clare Eayrs of the Korea Polar Research Institute attributed this season’s regional shifts to changes in atmospheric pressure and wind. NSIDC’s analysis, notably, offers no weather or ocean explanation for the Antarctic result, so I am not claiming one. Carbon Brief presents no direct clash between experts here. The disagreement is mostly about how much the record can support, not about what the record says.

A gap I should flag: the sources I read do not link this season to the 2026 El Niño. Whether tropical Pacific conditions played a role is not something any of them addressed, so it stays an open question.

What to watch next

NSIDC called the peak “likely,” and the 14 September date and 17.59 million figure are preliminary. NSIDC itself labels the data preliminary, so the numbers may move, and the OSI SAF and NSIDC numbers will not converge, since they are different products.

The nearer-term test is the spring retreat. Record-low daily values since 2 October mean the ice is melting back from a low peak at an early pace. If that continues into November and December, the February 2027 minimum becomes the next marker. NASA notes the minimum typically arrives in February; this year’s near-average February minimum shows how little a single summer settles.

Regional patterns are worth following alongside the totals. The Ross and Amundsen Seas and the Indian Ocean sector led the post-peak losses, and NSIDC notes that in recent low years some sectors still ran above average. A total that looks stable can hide big swings by sector.

Sources

1. NASA Earth Observatory, Antarctic Sea Ice Grows to Its 2026 Maximum
2. NSIDC Sea Ice Today analysis, October 2026
3. NSIDC Sea Ice Today, about the data
4. NSIDC Sea Ice Index
5. Carbon Brief, ‘Concerning’ low sea ice persists as Antarctic hits third-lowest winter peak
6. Met Office, briefing on Arctic sea ice minimum and Antarctic sea ice maximum, September 2026
7. Copernicus Climate Change Service, sea ice cover September 2026