For decades, scientists have struggled to observe some of the ocean’s most important currents. Large ocean currents such as the Gulf Stream can be tracked from space, but much smaller currents—often only a few kilometres wide and changing within hours—have remained difficult to measure.
Now, researchers led by Luc Lenain have introduced a new artificial intelligence system called Geostationary Ocean Flow (GOFLOW) that could dramatically change how we observe these hidden movements.
Using frequent thermal images from geostationary satellites and deep learning, GOFLOW can reconstruct ocean-surface currents at kilometre-scale resolution and roughly hourly intervals. The technology could provide scientists with an unprecedented view of rapidly changing ocean circulation and help improve our understanding of climate, marine ecosystems and ocean pollution.
The Ocean Is Far More Dynamic Than It Looks
The ocean may appear relatively calm when viewed from the shore or from space, but beneath its surface lies a constantly changing network of currents, fronts, filaments and eddies.
Among these are submesoscale currents, which typically range from hundreds of metres to tens of kilometres in size. Unlike large ocean currents that can persist for months, these smaller features can develop and change within approximately a day.
Despite their relatively small size, they have an enormous influence.
They help transport heat, carbon dioxide, nutrients and other chemicals between the surface and deeper layers of the ocean. They also influence how floating materials—including marine debris and pollutants—spread across the sea.
Understanding these currents is therefore important not only for oceanography but also for climate science, marine ecology and pollution monitoring.
The problem is that observing them is extremely difficult.
Why Traditional Satellites Struggle
Modern satellites have transformed ocean science. Missions such as TOPEX/Poseidon made it possible to map ocean surface height globally and revolutionized our understanding of large-scale ocean circulation.
Today, satellite altimetry can measure sea-surface height and use the principle of geostrophic balance to estimate ocean currents.
But there is a major limitation: time.
Many orbital satellites revisit the same location only every several days. The SWOT mission, for example, provides remarkable spatial information, but its approximately 21-day repeat cycle cannot continuously follow ocean features that may evolve substantially within a single day.
Imagine trying to understand the movement of a rapidly changing traffic system by taking only one photograph every three weeks. You might see where the roads are, but you would miss most of the actual movement.
Submesoscale currents create essentially the same problem.
There is another complication. Satellite measurements of sea-surface height can contain signals produced by internal ocean waves and tides. Separating those signals from actual surface circulation can be extremely challenging.
A Different Approach: Watch the Ocean Every Hour
This is where GOFLOW takes a different approach.
Instead of relying primarily on sparse measurements of sea-surface height, the researchers use sea-surface temperature imagery from geostationary satellites.
Unlike satellites that orbit Earth and periodically pass over the same location, geostationary satellites remain positioned above approximately the same region. This allows them to repeatedly observe the ocean throughout the day.
GOFLOW takes advantage of this continuous stream of thermal imagery.
A deep-learning architecture known as a U-Net is trained using high-resolution ocean circulation model data. The system learns how patterns in changing sea-surface temperature correspond to the movement of water.
It can then use satellite imagery to reconstruct detailed surface velocity fields.
The result is effectively a high-frequency map of how the ocean surface is moving.
Seeing Currents That Were Previously Hidden
One of the most important capabilities of GOFLOW is that it does not simply estimate the speed and direction of currents.
It can also estimate velocity gradients—how rapidly the current changes from one location to another.
This is particularly important for studying submesoscale features such as narrow fronts, filaments and small eddies.
When the researchers applied GOFLOW to the Gulf Stream, the system revealed detailed patterns in the ocean's surface circulation.
The reconstructed currents showed statistical characteristics of vorticity and divergence that had previously been documented mainly through high-resolution ocean models.
In simple terms, scientists were able to observe the small-scale twisting, stretching and converging motions of ocean water from satellite observations.
That is a significant step because these motions influence how energy and materials are redistributed throughout the ocean.
GOFLOW Also Captures Ageostrophic Motion
Another important advantage is that GOFLOW is not restricted to the traditional assumption of geostrophic flow.
Large-scale ocean circulation can often be approximated using geostrophic balance, where pressure gradients and Earth's rotation largely determine the movement of water.
But at smaller scales, ocean currents can behave differently.
Submesoscale flows can contain significant ageostrophic components, meaning their motion cannot be fully explained by the simplified geostrophic relationship.
GOFLOW's machine-learning approach can capture these components.
This allows researchers to study rapidly evolving currents that conventional satellite altimetry may struggle to resolve.
Why This Matters for Climate
The significance of GOFLOW goes far beyond creating prettier maps of the ocean.
The ocean absorbs enormous amounts of heat and carbon from Earth's atmosphere. Small-scale ocean processes influence how these materials move between the surface and deeper waters.
If scientists cannot properly represent these processes, climate models may miss important aspects of the Earth's climate system.
High-resolution observations from GOFLOW could therefore help researchers test, improve and constrain next-generation climate models.
As climate and weather models become increasingly sophisticated and operate at finer spatial and temporal scales, they require observations with comparable resolution.
GOFLOW could provide an important new source of such observations.
A Potential Tool Against Ocean Pollution
The technology could also have practical applications.
When oil, plastic, chemicals or other floating materials enter the ocean, their movement is strongly influenced by surface currents.
Large-scale current maps can provide a general idea of where these materials might travel. But small-scale currents and rapidly changing eddies can dramatically alter their trajectories.
Hourly, high-resolution current maps could potentially help scientists and authorities better predict the movement of marine pollutants and floating debris.
This could improve the response to future pollution events and help researchers understand how contaminants spread through marine environments.
There Is Still One Major Challenge: Clouds
GOFLOW is not without limitations.
Because it relies on thermal satellite imagery, cloud cover can interrupt observations. At any given moment, a large fraction of the ocean can be hidden beneath clouds.
However, the researchers found that the system could still perform well when observations were intermittently interrupted by clouds. Comparisons with ship-based measurements showed strong agreement even under heavily cloudy conditions.
The high frequency of geostationary observations is crucial here. Even when clouds block the ocean at one moment, gaps can appear later, providing additional information.
Future versions could combine GOFLOW with other observations, including microwave instruments and satellite altimetry, to help fill these gaps.
A New Era of Ocean Observation
For generations, scientists have had to choose between broad coverage and fine detail. Traditional satellites offered global observations but often lacked the temporal resolution needed to follow rapidly changing currents.
GOFLOW offers a different possibility.
By combining geostationary satellites, thermal imagery and artificial intelligence, researchers can begin watching the ocean's small-scale movements unfold almost in real time.
The technology does not simply provide more data. It provides a new way of seeing processes that were previously difficult to observe.
From climate forecasting and marine ecosystem monitoring to pollution tracking and ocean modelling, the potential applications are enormous.
The ocean is not simply a vast body of water moving in a few predictable directions. It is a constantly changing system filled with countless smaller motions that collectively influence Earth's climate and ecosystems.
GOFLOW could finally give scientists the ability to watch many of those hidden movements—hour by hour, kilometre by kilometre.
And in the race to understand a changing planet, being able to see the ocean move at this scale could prove to be a major breakthrough.
Reference: Lenain, L., Srinivasan, K., Barkan, R. et al. An unprecedented view of ocean currents from geostationary satellites. Nat. Geosci. 19, 526–533 (2026). https://doi.org/10.1038/s41561-026-01943-0
Comments
Post a Comment