IceBench-S2S: A Benchmark of Deep Learning for Challenging Subseasonal-to-Seasonal Daily Arctic Sea Ice Forecasting in Deep Latent Space
Jingyi Xu, Shengnan Wang, Weidong Yang✉, Siwei Tu, Lei Bai✉, Ben Fei✉ · 2026
Read IceBench-S2S: A Benchmark of Deep Learning for Challenging Subseasonal-to-Seasonal Daily Arctic Sea Ice Forecasting in Deep Latent Space by Jingyi Xu, Shengnan Wang, Weidong Yang✉, Siwei Tu, Lei Bai✉, Ben Fei✉ (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-02567.txt.
Description
Jingyi Xu and 5 co-authors create the first benchmark for daily Arctic sea ice forecasts stretching a full season ahead, in a 2026 arXiv paper. Sea ice regulates Earth's climate and governs polar ecology and coastal activity. Deep learning forecasters beat physics models on accuracy and cost, but their skill usually ends at daily subseasonal scales or monthly averages.
IceBench-S2S targets successive one-hundred-eighty-day forecasts of sea ice concentration. Its system first compresses daily spatial ice data into a deep latent space. Temporally concatenated features are then modeled by interchangeable forecasting backbones that predict the seasonal variation. A unified training and testing workflow lets different backbones compete fairly. The benchmark also gives practical guidance for model selection in polar monitoring. Operational uses, from Arctic shipping routes to scientific expeditions, need exactly this daily long-range skill.
Catalog record
| CREATOR | Jingyi Xu, Shengnan Wang, Weidong Yang✉, Siwei Tu, Lei Bai✉, Ben Fei✉ |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-02567