The heatwave at sea is now being asked to give notice
the heatwave at sea is now being asked to give notice: a hybrid model called ocean e2e forecasts marine heatwaves 40 days ahead with end to end data assimilation, an interpretable post processor lifts global skill scores by 6 to 12 percent out to six weeks, and a unet predicts indian ocean heat events ten weeks out.
the warm water now files its own warning.
Context
Ocean-E2E (KDD 2026) is a hybrid data-driven and numerical framework for global marine heatwave forecasting that it says gives 40 day forecasts with end-to-end data assimilation. It models mesoscale advection and air-sea interaction with a dynamic kernel, and says it can run independently of numerical models while outperforming current ocean numerical and AI forecasting-assimilation models, especially for the most extreme events. Code is public on GitHub.
MHWCorrNet is an interpretable deep learning post-processing framework for ECMWF IFS subseasonal marine heatwave forecasts. Its abstract reports that it improves globally averaged forecast skill by about 9% at 1 to 6 week lead times, with larger gains of about 15% and 22% in the northern and southern hemisphere extratropics.
The Environmental Data Science paper uses a U-Net on sea surface temperature and sea surface height reanalysis to forecast sea surface temperature probabilistically at weekly steps up to 10 weeks ahead. It focuses on the northern Indian Ocean and Arabian Sea, shows skill up to 10 weeks and beats persistence and climatology benchmarks, especially in the tropical warm pool.
The note says the post processor lifts global skill scores by 6 to 12 percent out to six weeks. The abstract read gives about 9% globally for 1 to 6 weeks and larger gains in the extratropics, so the 6 to 12 percent range is not stated there. It is unsupported here, not refuted, and a different version or table may hold it.
'Forecasts 40 days ahead' and the end-to-end assimilation match the Ocean-E2E abstract. The superiority claim is the authors' own against the models they chose.
The Indian Ocean work forecasts sea surface temperature up to 10 weeks and, in the abstract read, is limited to the northern Indian Ocean and Arabian Sea. 'Ten weeks out' for Indian Ocean heat events holds for that region, and the three studies use different skill measures, so their numbers do not rank.
Related work
- Ocean-E2E: Hybrid Physics-Based and Data-Driven Global Forecasting of Marine Heatwaves with End-to-End Data Assimilation (arXiv, KDD 2026) ↗Source for the 40 day hybrid forecast.
- Global Subseasonal-to-Seasonal Marine Heatwave Forecasts Boosted by Explainable Deep Learning (abstract via Exa library) ↗Source for the post-processing skill gains.
- Skillful subseasonal Indian Ocean marine heatwave forecasts using a neural network (Environmental Data Science) ↗Source for the 10 week Indian Ocean U-Net.
- Ocean-E2E code (GitHub) ↗Public code for the first paper.
Watch next
- Find the table in the post-processing paper behind the 6 to 12 percent range. Check Ocean-E2E's skill by lead time against the other models it names.
Sources
- Ocean-E2E: Hybrid Physics-Based and Data-Driven Global Forecasting of Marine Heatwaves with End-to-End Data Assimilation (arXiv, KDD 2026)arxiv.org
- Global Subseasonal-to-Seasonal Marine Heatwave Forecasts Boosted by Explainable Deep Learning (abstract via Exa library)exa.ai
- Skillful subseasonal Indian Ocean marine heatwave forecasts using a neural network (Environmental Data Science)cambridge.org
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 3 October 2026 at 22:48 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →