Spring area reduced Antarctic summer forecast error against a mean in 2010-2025, while gains over prior summer were uncertain

We tested whether Antarctic spring sea ice area improves summer forecasts beyond a recent mean or prior summer. We used the five southern sectors of Sea Ice Index Version 4 and 16 summer forecasts for 2010-2025. Each fit used the 15 most recent valid earlier annual pairs. Spring area reduced mean squared error by 59.3% against the matched mean. The conditional 95% three-year block interval was 17.5% to 80.7%. Root mean squared error fell from 0.712 to 0.454 million square kilometres. The previous summer value alone had an error of 0.472. The gain against that benchmark was 7.5%, with an interval of -121.4% to 65.8%. Gains after matched trend and prior-summer regression controls were also uncertain. A one-month target gap retained the area gain against a mean. A winter predictor and an extent outcome gave weaker evidence. The result shows a benefit against a long mean, but no resolved extra gain beyond a simple prior-state benchmark. It does not identify a physical cause or establish future skill. The short test, changing conditions, current data snapshot, and conditional uncertainty limit inference. The package includes source hashes, monthly inputs, code, all forecasts, scores, bootstrap draws, an audit, and a figure.

Antarctic sea ice · seasonal prediction · sea ice area · forecast baselines · climate persistence · chronological validation