The Reuters news staff had no role in the production of this content. It was created by Reuters Plus, the brand marketing studio of Reuters.
Produced by Reuters Plus for
When the ocean changes, so does everything connected to it, from fish stocks and farming to coastal safety and tourism, and the biodiversity that underpins entire ecosystems. Knowing what it will do next matters more than ever. Predicting how the ocean will behave over the next ten days has traditionally meant running some of the world's most powerful computers for the best part of an hour. A new generation of AI models can do it in seconds.
How AI is revolutionizing ocean forecasting
The leap in speed is part of a wider shift now reaching ocean science. Mercator Ocean International, which runs the ocean monitoring and forecasting service, Copernicus Marine, on behalf of the European Commission, is among those leading the work -- effectively building a faster, AI-assisted version of the services that shipping, fisheries, climate scientists and coastal authorities already rely on. The stakes are broad. Speaking at the first ever AI Ocean Forum as part of Digital Ocean Week in Brussels, Christos Dermentzopoulos, Greece Deputy Minister for Digital Governance and Artificial Intelligence, said, “The sea is part of our economy. Part of our culture. Part of our security. This is why we see digital ocean systems as a new generation of public infrastructure -- helping authorities assess coastal risks, monitor water quality, protect ecosystems, and plan maritime activities.” The implications stretch beyond infrastructure. For Grazyna Piesiewicz, Head of Unit for High Performance Computing and Applications at the European Commission, the ocean's role in the broader Earth system makes AI's potential even more significant. “The ocean is central to climate resilience, to biodiversity, to food security, energy, maritime safety, economic security.”
228MM
Sea level rise between 1901-2024
AI, she argues, offers a chance to speed up progress by turning huge amounts of environmental data into usable intelligence. For decades, ocean forecasts have been built on what scientists call first principles -- solving the physical equations that govern how currents, temperature and salinity evolve, a process that demands enormous computing power. AI doesn't replace that approach so much as build on it. "Without physics, AI ocean forecasting wouldn't exist," says Anass El Aouni, Machine Learning Oceanographer at Mercator Ocean, who helped build GLONET -- one of the clearest examples of this approach already in use.
“AI is pretty much entering every single level of processing”
- Anass El Aouni, Machine Learning Oceanographer, Mercator Ocean International
Time it takes to forecast 10 days of ocean conditions
(Source: European Union, European Digital Twin Ocean, EDITO, Information (2026), © Mercator Ocean)
1/36
Ultra-high-resolution target for OceanLens globally
OceanBench and GLONET are already helping bring AI-based ocean forecasting toward operational scale. They’re part of a wider vision for a European Digital Twin of the Ocean (EDITO): a working replica of the sea that can be used to see what might happen digitally before anything changes in the real world. "The main interest is to test some actions -- for instance, reducing fishing in certain areas -- inside the computer first. You see the effect on fish stocks, and then you can decide on your policy," says Alain Arnaud, Director of Mercator Ocean's Digital Ocean Department. Other applications include coastal flood risk forecasting, shipping route optimization and tracking ocean plastic.
The underlying AI also allows these models to zoom in on a much finer scale, Arnaud says -- useful for assessing the impact of new developments before they're built. Europe's vision of a coordinated ocean observation and a digital ocean system now has a formal political framework in OceanEye, a European Commission initiative designed to transform how the ocean is observed and understood. With the US and China already major players in this space, Europe's digital ocean infrastructure is expected to be fully operational as a public service by 2030, with the EU aiming to become the world's leading provider of ocean intelligence by 2035. Policy ambition is already attracting private sector interest. "This is a major opportunity for European businesses, for start-ups and scale-ups," says Wiebke Pankauke, Head of Unit for Ocean, Seas and Waters at the European Commission. "Ocean observation in itself is already a significant market in its own right." One company already benefiting is Amphitrite, a French start-up whose AI ocean-current models run on Nvidia's Earth-2 platform, helping ships find more efficient routes. "AI is the core of what we do," says Théo Archambault, Head of AI R&D at Amphitrite. "It gives a much more accurate estimate of voyage time and fuel use -- and helps avoid storms, taking less risk." From policy to the private sector, AI is reshaping how the ocean is understood and used. Europe is well placed to build on that, says Pierre Bahurel, Director General of Mercator Ocean International. "How lucky we are in Europe to be in a situation where we can accelerate, innovate, and change the world safely because we have what really matters: skilled people and strong public frameworks to develop innovation." Mercator Ocean’s own next steps reflect this. GLONET 2, currently being evaluated through OceanBench, extends forecasts out to 40 days and adds a high-resolution regional model for European waters. OceanLens goes further still -- a downscaling tool that takes a broad global forecast and sharpens it to nine times the detail, representing a significant step forward in access to high-resolution ocean data.
Taken together, these tools point to something bigger than faster forecasts. Their value depends on the ecosystem around them -- and none of it works without people: the scientists who build these tools, and the next generation who will need to use them. "We need to train models, but also train the young generation, open their eyes, and encourage them to be creative and responsible so they can provide the best knowledge for humanity," says Michèle Barbier, Ethics Expert for the European Commission. The technology will keep moving fast. The people building it, and the choices they make about how to use it, will matter just as much.
(Source: Copernicus Ocean State Report, European Union, Copernicus Marine Service Information (2026) © Mercator Ocean)
Disclaimer: The Reuters news staff had no role in the production of this content. It was created by Reuters Plus, the brand marketing studio of Reuters. To work with Reuters Plus, contact us here.
Discover more about digital ocean systems
03
seconds with AI
HOUR with AI
01
3 SECONDS
1 HOUR
with AI
without AI
°
º
Trained on more than 30 years of ocean data, going back to 1993 and deployed in a pre-operational setting alongside Mercator Ocean's existing physical-equations models, GLONET produces ten-day global forecasts in a fraction of the time a traditional model takes. But speed and cost reduction alone don't answer the obvious question: can these AI models be trusted? To build that confidence, Mercator Ocean launched OceanBench in 2025 -- an open, collaborative framework that benchmarks GLONET against AI ocean models developed by research groups elsewhere in the world, checking not just their accuracy but whether they respect fundamental ocean physics. “We propose OceanBench as a community-driven, open-source framework for AI Ocean model verification and intercomparison,” says Simon van Gennip, Oceanographer at Mercator Ocean. It was this validation that allowed GLONET to become the first AI ocean model deployed in this way. The need to instil confidence is something that ocean AI can also learn from those further along the same path. "One of the things the ocean community needs to be incredibly mindful of is building trust in machine learning models," says Rachel Furner, Ocean Modelling Scientist at the European Centre for Medium-Range Weather Forecasts, where AI has been transforming atmospheric prediction for longer. "In oceanography, we're at the beginning part of that revolution still -- and the way we build these models needs to be done in a way that our community trusts."
Less energy used by AI than a traditional model
10,000x
By revealing fine-scale ocean features such as surface currents with unprecedented detail for machine learning-based forecasts, OceanLens helps deliver more precise forecasts of rapidly changing marine conditions.
Image created by Seascape Belgium for EDITO 2 — Funded by the European Union (GA no. 101227771)
MAKING WAVES
“AI is pretty much entering every single level of processing,”
Trained on more than 30 years of ocean data, going back to 1993 and deployed in a pre-operational setting alongside Mercator Ocean's existing physical-equations models, GLONET produces ten-day global forecasts in a fraction of the time a traditional model takes.
The underlying AI also allows these models to zoom in on a much finer scale, Arnaud says -- useful for assessing the impact of new developments before they're built. Europe's vision of a coordinated ocean observation and a digital ocean system now has a formal political framework in OceanEye, a European Commission initiative designed to transform how the ocean is observed and understood. With the US and China already major players in this space, Europe's digital ocean infrastructure is expected to be fully operational as a public service by 2030, with the EU aiming to become the world's leading provider of ocean intelligence by 2035. Policy ambition is already attracting private sector interest. "This is a major opportunity for European businesses, for start-ups and scale-ups," says Wiebke Pankauke, Head of Unit for Ocean, Seas and Waters at the European Commission. "Ocean observation in itself is already a significant market in its own right." One company already benefiting is Amphitrite, a French start-up whose AI ocean-current models run on Nvidia's Earth-2 platform, helping ships find more efficient routes. "AI is the core of what we do," says Théo Archambault, Head of AI R&D at Amphitrite. "It gives a much more accurate estimate of voyage time and fuel use -- and helps avoid storms, taking less risk."
From policy to the private sector, AI is reshaping how the ocean is understood and used. Europe is well placed to build on that, says Pierre Bahurel, Director General of Mercator Ocean International. "How lucky we are in Europe to be in a situation where we can accelerate, innovate, and change the world safely because we have what really matters: skilled people and strong public frameworks to develop innovation."
But speed and cost reduction alone don't answer the obvious question: can these AI models be trusted? To build that confidence, Mercator Ocean launched OceanBench in 2025 -- an open, collaborative framework that benchmarks GLONET against AI ocean models developed by research groups elsewhere in the world, checking not just their accuracy but whether they respect fundamental ocean physics. “We propose OceanBench as a community-driven, open-source framework for AI Ocean model verification and intercomparison,” says Simon van Gennip, Oceanographer at Mercator Ocean. It was this validation that allowed GLONET to become the first AI ocean model deployed in this way. The need to instil confidence is something that ocean AI can also learn from those further along the same path. "One of the things the ocean community needs to be incredibly mindful of is building trust in machine learning models," says Rachel Furner, Ocean Modelling Scientist at the European Centre for Medium-Range Weather Forecasts, where AI has been transforming atmospheric prediction for longer. "In oceanography, we're at the beginning part of that revolution still -- and the way we build these models needs to be done in a way that our community trusts."
Mercator Ocean’s own next steps reflect this. GLONET 2, currently being evaluated through OceanBench, extends forecasts out to 40 days and adds a high-resolution regional model for European waters. OceanLens goes further still -- a downscaling tool that takes a broad global forecast and sharpens it to nine times the detail, representing a significant step forward in access to high-resolution ocean data.
For decades, ocean forecasts have been built on what scientists call first principles -- solving the physical equations that govern how currents, temperature and salinity evolve, a process that demands enormous computing power. AI doesn't replace that approach so much as build on it. "Without physics, AI ocean forecasting wouldn't exist," says Anass El Aouni, Machine Learning Oceanographer at Mercator Ocean, who helped build GLONET -- one of the clearest examples of this approach already in use.
The underlying AI also allows these models to zoom in on a much finer scale, Arnaud says -- useful for assessing the impact of new developments before they're built.
Europe's vision of a coordinated ocean observation and a digital ocean system now has a formal political framework in OceanEye, a European Commission initiative designed to transform how the ocean is observed and understood. With the US and China already major players in this space, Europe's digital ocean infrastructure is expected to be fully operational as a public service by 2030, with the EU aiming to become the world's leading provider of ocean intelligence by 2035. Policy ambition is already attracting private sector interest. "This is a major opportunity for European businesses, for start-ups and scale-ups," says Wiebke Pankauke, Head of Unit for Ocean, Seas and Waters at the European Commission. "Ocean observation in itself is already a significant market in its own right." One company already benefiting is Amphitrite, a French start-up whose AI ocean-current models run on Nvidia's Earth-2 platform, helping ships find more efficient routes. "AI is the core of what we do," says Théo Archambault, Head of AI R&D at Amphitrite. "It gives a much more accurate estimate of voyage time and fuel use -- and helps avoid storms, taking less risk."
From policy to the private sector, AI is reshaping how the ocean is understood and used. Europe is well placed to build on that, says Pierre Bahurel, Director General of Mercator Ocean International. "How lucky we are in Europe to be in a situation where we can accelerate, innovate, and change the world safely because we have what really matters: skilled people and strong public frameworks to develop innovation." Mercator Ocean’s own next steps reflect this. GLONET 2, currently being evaluated through OceanBench, extends forecasts out to 40 days and adds a high-resolution regional model for European waters. OceanLens goes further still -- a downscaling tool that takes a broad global forecast and sharpens it to nine times the detail, representing a significant step forward in access to high-resolution ocean data.
AI, she argues, offers a chance to speed up progress by turning huge amounts of environmental data into usable intelligence.
The leap in speed is part of a wider shift now reaching ocean science. Mercator Ocean International, which runs the ocean monitoring and forecasting service, Copernicus Marine, on behalf of the European Commission, is among those leading the work -- effectively building a faster, AI-assisted version of the services that shipping, fisheries, climate scientists and coastal authorities already rely on.
The stakes are broad. Speaking at the first ever AI Ocean Forum as part of Digital Ocean Week in Brussels, Christos Dermentzopoulos, Greece Deputy Minister for Digital Governance and Artificial Intelligence, said, “The sea is part of our economy. Part of our culture. Part of our security. This is why we see digital ocean systems as a new generation of public infrastructure -- helping authorities assess coastal risks, monitor water quality, protect ecosystems, and plan maritime activities.”
The implications stretch beyond infrastructure. For Grazyna Piesiewicz, Head of Unit for High Performance Computing and Applications at the European Commission, the ocean's role in the broader Earth system makes AI's potential even more significant. “The ocean is central to climate resilience, to biodiversity, to food security, energy, maritime safety, economic security.”
seconds
hour
1/36°
