How Robotic Ocean Floats Are Tracking Deep Marine Life
Using robotic floats and machine learning to map subsurface light, scientists tracked changes in ocean primary production down to 200 metres from 2010 to 2026. They found notable declines in the North Atlantic and Southern Ocean, alongside increases in parts of the Pacific.
Reading level
The full story with proper science words explained.
Looking Beneath the Surface
Satellites watching Earth from orbit can only peer into the ocean's very skin. This leaves hundreds of metres of underwater plant-like growth completely hidden from view. Microscopic marine organisms called phytoplankton carry out photosynthesis beneath the waves. This process generates ocean primary production, turning carbon into food that powers entire marine food webs.
Gazing at a forest from an aeroplane only reveals the sunlit upper canopy, while profiling floats act like elevators descending through the undergrowth to measure growth at every level. Measuring these deeper layers has historically proved difficult. Satellites miss the action beneath the surface optical depth, and deep-sea research cruises cannot sample every corner of the globe at once.
Clever Computing Rescues the Fleet
To inspect the sunlit layer down to 200 metres, researchers turned to Biogeochemical-Argo robotic floats. These autonomous instruments drift across open seas and dive deep before resurfacing to beam back data. However, out of 1,092 floats active between 2010 and 2026, only 254 carried dedicated sensors for photosynthetically active radiation, the light wavelengths used for photosynthesis.
To fix this shortfall, scientists trained a neural network called SOCA. By pairing satellite ocean colour with water temperature and salinity readings, the model estimated underwater light for the floats lacking light sensors. This trick expanded the usable dataset from 26,726 profiles to 90,199. Using this modelled light, they adjusted for nonphotochemical quenching, a protective habit where phytoplankton shed excess sunlight as heat and skew measurements. They then ran the data through the Carbon-based Production Model to calculate plant growth down through the water column.
Winners and Losers Across the Basins
The deep measurements revealed that biological activity is shifting unevenly across 16 global ocean biomes. In the North Atlantic and Southern Ocean, net primary production declined at rates between 2.20% and 4.15% per year. Less biological growth in these vast waters could influence how much carbon sinks out of the atmosphere.
Other regions showed the opposite trend. In the subpolar North Pacific, production climbed by 3.44% per year, while the South Pacific subtropical gyre rose by 3.69% per year. The study also revealed that converting raw sensor fluorescence to chlorophyll requires distinct factors for different regions and seasons. Using a single global number can introduce major regional bias.
What We Still Do Not Know
These findings provide our first detailed look below the surface on a global scale, but big puzzles remain. A 10- to 15-year record is likely not long enough to clearly pinpoint human-driven climate change, as this timeframe overlaps with natural climate oscillations that swing over decades.
Floats also remain unevenly distributed across the globe, leaving high-latitude North Pacific waters uncovered and polar ice biomes sparsely measured. In addition, tiny minerals and non-algal debris drift through clear ocean gyres, creating uncertainties when scientists convert light backscattering signals into carbon biomass.
Science words
- Net primary production
- The rate at which phytoplankton build organic matter using photosynthesis, minus the carbon they burn during respiration.
- Biogeochemical-Argo
- A global array of robotic profiling floats that dive through the sea while measuring chemical and biological properties.
- Photosynthetically active radiation
- The specific wavelengths of sunlight that plants and algae can absorb to fuel photosynthesis.
- Nonphotochemical quenching
- A mechanism where phytoplankton dump excess solar energy as heat under bright light, artificially lowering optical sensor readings.
- Carbon-based Production Model
- A mathematical method that calculates marine plant production from phytoplankton carbon biomass and growth rates.
Check it yourself
This story is based on a real research paper in Science Advances by Ryan-Keogh, Thomalla. We write with AI help and check it against the paper, but the original is the final word.