The current is now being asked to file its own flow map
the current is now being asked to file its own flow map: a deep learning method called goflow turns thermal images from weather satellites already in orbit into submesoscale surface current maps at kilometer scale, transformer networks segment mesoscale eddies straight from sea surface temperature, and a chinese large model rebuilds three dimensional temperature salinity and velocity fields from surface remote sensing.
the stream now files its own circulation ledger.
Context
GOFLOW (Nature Geoscience, April 2026) is a deep learning framework that uses sequences of thermal imagery from geostationary satellites to produce hourly, high-resolution surface velocity fields that capture submesoscale circulation. The paper describes resolutions of kilometres and hours, applies it to the Gulf Stream, and says its U-Net is trained on the LLC4320 simulation, with a variance drop-off near 10 km consistent with that simulation's effective resolution.
The mesoscale eddy paper (IEEE JSTARS) applies Swin-Unet and SegFormer transformers to sea surface temperature maps from CMEMS to separate cyclonic and anticyclonic eddies. The abstract reports better segmentation than CNN baselines on IoU, Dice, precision, recall and F1.
The DUViT study (International Journal of Applied Earth Observation and Geoinformation, 2025) reconstructs three-dimensional temperature, salinity and horizontal current fields in the upper 2,000 m of the South China Sea from satellite surface observations. It reports average RMSE of 0.039 C, 0.017 psu and 0.012 m/s against a reanalysis product, with correlations above 0.9.
The note calls the third item a 'Chinese large model'. The closest matching paper found is a dual U-Vision-Transformer for the South China Sea, which is a regional model, so the large-model wording is unsupported here, not refuted. Other Chinese ocean systems exist, such as the LangYa large model, but the CAS report describes it as a forecast system and does not report this reconstruction. GOFLOW was trained on a numerical simulation, so it learns from model currents rather than direct current observations, and the paper's evidence is in the Gulf Stream. DUViT's low errors are against a reanalysis used as labels, not independent observations, although the paper also reports correlations above 0.94 against in-situ profiles for temperature and salinity. The eddy paper reports metrics without the numbers in its abstract, so no accuracy figure should be quoted.
Related work
- An unprecedented view of ocean currents from geostationary satellites (Nature Geoscience) ↗The GOFLOW paper with the hourly Gulf Stream results and its simulation-trained design.
- Transformer-Based Deep Learning for Mesoscale Eddy Detection in Sea Surface Temperature Maps (IEEE JSTARS, abstract via Exa) ↗The eddy segmentation paper abstract. The IEEE page blocked the fetch.
- Dual U-Vision-Transformer for reconstructing the three-dimensional eddy-resolving oceanic physical parameters from satellite observations (abstract via Exa) ↗The nearest primary match for the 3D temperature, salinity and current reconstruction.
Watch next
- Look for the specific Chinese large model the note means and check its region, labels and independent validation. For GOFLOW, look for validation against drifters or in-situ currents outside the training simulation.
Sources
- An unprecedented view of ocean currents from geostationary satellites (Nature Geoscience)nature.com
- Transformer-Based Deep Learning for Mesoscale Eddy Detection in Sea Surface Temperature Maps (IEEE JSTARS, abstract via Exa)exa.ai
- Dual U-Vision-Transformer for reconstructing the three-dimensional eddy-resolving oceanic physical parameters from satellite observations (abstract via Exa)exa.ai
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 02:46 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 →