Mean shifts account for 72% of the rise in cross-sector winter sea ice anomaly products between 2007–2015 and 2017–2025 with a 1989–2015 reference

We test whether recent low Antarctic winter sea ice reflects stronger shared variation between sectors. We use five Sea Ice Index version 4 sectors and July–September means. We compare nine winters in 2007–2015 with nine in 2017–2025. Relative to the 1989–2015 sector means, changes in period means account for 72.4% of the rise in summed off-diagonal anomaly products. Changes in centered covariance account for 27.6%. Centered variance amplification rises from 0.802 to 1.391, but the 95% year-bootstrap interval for its change is -0.784 to 1.515. It falls to 1.168 without 2023 and to 1.080 after local linear trends are removed. The total mean falls by 1.148 million square kilometres. These data show a lower mean in each sector, but nine-year windows do not establish a change in centered covariance. The decomposition is a statistical audit, not a cause estimate. The mean contribution ranges from 54.5% to 84.5% across reference checks. A matched June–August check reproduces prior campaign covariance values. These are overlapping data, not independent confirmation. Code, derived data, raw-file hashes, and reproduction steps are included.

Antarctic sea ice · winter extent · regional covariance · mean shift · uncertainty