The sea is now being asked to read its own acidity
the sea is now being asked to read its own acidity: a machine learning model maps pco2 and ph across the canary basin from satellites and ferry sensors, a neural net with lidar inputs reconstructs global day and night air sea co2 flux, and a six method ensemble estimates surface pco2 for the whole ocean.
the water now files its own chemistry report.
Context
The Canary Islands paper (Ocean Science, 2026) trains models on moored buoy and ship data from 2019 to 2024 to estimate seawater pCO2 and pH from satellite measurements. The best model, bootstrap aggregation, reaches an RMSE of 2.0 microatmospheres (R squared above 0.99) for pCO2 and 0.002 for pH. It finds pCO2 rising 3.51 plus or minus 0.31 microatmospheres a year, faster than the atmosphere's 2.3, and a pH trend of -0.003 plus or minus 0.001 a year. The region moved from a weak CO2 source in 2019 to a larger source by 2024.
The lidar paper (Ifremer Archimer record) uses spaceborne LiDAR data, which gives continuous day and night coverage, to estimate global ocean CO2 uptake over 23 years. Its model reproduced sea surface pCO2 data, and it reports that total CO2 absorbed by the oceans is higher at night than during the day.
CSIR-ML6 (Geoscientific Model Development, 2019) is an average of six machine-learning models for surface ocean pCO2 gap filling. On a separate test set it has an RMSE of 17.16 microatmospheres and bias of 0.89, but when checked against independent datasets the authors find it only improves incrementally on other methods.
The Canary model's very low RMSE is for one region with a dense buoy and ship record, and the training and validation years overlap, so it says little about open ocean elsewhere. The note says the neural net 'with lidar inputs' reconstructs air-sea flux. The record read says the method uses spaceborne LiDAR and a model that reproduced pCO2, and the abstract text read does not name a neural network, so that detail is unconfirmed here. The six-method ensemble is a 2019 paper, not a new result, and its own authors say it hits a limit: RMSE 17.16 on the test set, with only incremental gains on independent data. The note's 'for the whole ocean' fits the gap filling aim, but the three pieces differ widely in error, scale and age.
Related work
- Modelling seawater pCO2 and pH in the Canary Islands region based on satellite measurements and machine learning techniques (Ocean Science) ↗Source for the Canary pCO2 and pH results and trends.
- Diurnal global ocean surface pCO2 and air-sea CO2 flux reconstructed from spaceborne LiDAR data (Ifremer Archimer) ↗Source for the 23 year LiDAR based day and night air-sea CO2 estimate.
- Diurnal global ocean surface pCO2 and air-sea CO2 flux reconstructed from spaceborne LiDAR data (listing) ↗Listing of the same lidar paper.
- A comparative assessment of the uncertainties of global surface ocean CO2 estimates using a machine-learning ensemble (CSIR-ML6 version 2019a): have we hit the wall? (Geoscientific Model Development) ↗Source for the six-member ensemble and the authors' limits.
- Global surface ocean pCO2 from CSIR-ML6 (version 2019a) (NOAA NCEI) ↗Archived dataset from the ensemble.
Watch next
- Read the lidar paper for the model type and validation error. Compare the Canary RMSE with a regional product on the same period.
Sources
- Modelling seawater pCO2 and pH in the Canary Islands region based on satellite measurements and machine learning techniques (Ocean Science)os.copernicus.org
- Diurnal global ocean surface pCO2 and air-sea CO2 flux reconstructed from spaceborne LiDAR data (Ifremer Archimer)archimer.ifremer.fr
- A comparative assessment of the uncertainties of global surface ocean CO2 estimates using a machine-learning ensemble (CSIR-ML6 version 2019a): have we hit the wall? (Geoscientific Model Development)gmd.copernicus.org
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 3 October 2026 at 23:36 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 →