Can a Robot With No Ship Track the Ocean's Carbon? Autosub Long Range, Explained
In May 2022, a robot sub called Autosub Long Range left Plymouth, on the south coast of England, with no ship following it. In five weeks it traveled about 2,000 km, out across the Celtic Sea to the deep Atlantic and back, and its two pH sensors took nearly 50,000 readings of how acidic the water was. Researchers at the UK's National Oceanography Centre and the University of Southampton used them to work out which way carbon dioxide was moving between the sea and the air.
On average, the sea was taking it in. But when the sea and the air were close to balance, a tiny difference in pH could blur the answer, especially in strong wind.
In this video
- Why coastal carbon is hard to pin down: the ocean takes in about a quarter of the carbon dioxide people release, but most coastal measurements come from ships that visit and leave
- Reading carbon from pH: pH plus alkalinity (estimated here, mostly from salinity) gives the water's carbon dioxide pressure, and the wind sets how fast gas crosses the surface
- The robot: 3.6 m long, 750 kg, 30 battery packs, dives of four hours to a day, then a GPS fix and a satellite check-in
- The voyage: May 10 to June 13, 2022, 54 missions, more than 800 hours of data, 90% of the time in the top 100 m
- Two pH sensors: a slow lab on a chip that reads a dye's color once an hour, and a fast SeaFET that reads every minute but drifts. Before any correction they agreed within their combined margin of error 87% of the time. After the fast one was anchored to the slow one, 98%, as expected once it was tuned to it
- Sensors versus models: the two sensors agreed with each other better than with computer models of pH
- Into the sea: in late spring 2022 the Celtic Sea took in carbon dioxide on average, and pH and oxygen rose and fell with the day
- Small pH, big flow: changing the alkalinity moved the carbon dioxide pressure by 3 microatmospheres at most, while changing the pH moved it by up to 42. In the windy week of June 7, the data couldn't say which way the carbon dioxide went
- The limits: one robot, one route, five weeks, no water samples, and the wind from a weather model
For engineers
A NOC Lab-on-Chip spectrophotometric pH sensor (858 measurements) and a Sea-Bird Deep SeaFET V2 ISFET (49,184). At the surface the intakes sat about 0.4 m below the waterline, and only those surface values went into the fluxes. The SeaFET's drift was corrected against the Lab-on-Chip with two breakpoints and three linear segments (k0-LOC, 1σ 0.007 pH); the same correction against the ESPER-NN neural network (k0-NN) gave 1σ 0.011.
pCO₂ came from CO2SYS with three pH records and four salinity-based total alkalinity estimates. Fluxes used a bulk formula with a Wanninkhof (2014) wind-squared transfer velocity, ERA5 winds and atmospheric CO₂ from Penlee Point; every method's mean was negative (k0-LOC −4.3 mmol m⁻² d⁻¹). The pH choice alone could shift the flux by about 25%, and the authors treat the spread between methods as a minimum bound on the uncertainty.
The paper
"Coastal Marine Carbon and Air-Sea Fluxes Quantified From pH Sensors on an Extended AUV Deployment"
Emily M. Hammermeister, Cathy Wimart-Rousseau, Stathys Papadimitriou, Pablo Trucco-Pignata, Edward Chaney, Robert Templeton, Alexander B. Phillips, Socratis Loucaides
National Oceanography Centre, Southampton, and the School of Ocean and Earth Sciences, University of Southampton
Journal of Geophysical Research: Oceans 131, e2025JC023213 (August 2026), doi:10.1029/2025JC023213. Open access, CC BY 4.0.
Autosub Long Range 1500's development and its Long-Distance Proving Trial were supported by the ISCF/NERC Oceanids Programme and NERC National Underpinning Autonomy Capability funding, and through the INSPIRE DTP. Atmospheric CO₂ data: Penlee Point Atmospheric Observatory, Plymouth Marine Laboratory. Sensor data: University of Southampton, doi:10.5258/SOTON/D3716 (CC BY 4.0).
This is an independent explainer. I'm not affiliated with or endorsed by the authors or their institutions, and the research is all theirs. The animations and board sketches are my own. The route and the pH charts are redrawn from the study's open data, with my own approximate drift correction; drawings marked schematic are illustrations. Any mistakes are mine.
I'm Bora Celik from Piccard. I explain new research in ocean science and robotics.