Summer ice deficit reduced Kara Sea advance-date test error beyond September extent in 2005-2024

We tested whether summer ice history adds information after the ice extent observed on September 1. We used five Arctic shelf seas in Sea Ice Index Version 4. Each estimate used only earlier years to fit the model. In the Kara Sea, adding May-August extent deficit reduced 2005-2024 mean squared error by 48.0%. Root mean squared error fell from 19.88 to 14.34 days. The target was the first return to 80% of a fixed winter reference. Conditional block limits for mean loss reduction were above zero, including a five-region family check. The other four seas had intervals that included zero. This is a test of added forecast information, not a measurement of ocean heat storage or cause. The record has only 20 test years and uses final data, not real-time data versions.

sea ice · Kara Sea · seasonal memory · forecast validation · ocean heat