New research shows that new AI models could predict currents in the Gulf of Mexico from laptops.
Developed by researchers at Met Office and Exeter University, AI-based models help redefine how new technologies can support marine operational decisions.
The latest paper extends the team's previous work to show how machine learning from a low-cost offshore modeling (Malcom) framework, originally designed for regional prediction of ocean waves in UK coastal waters, is applied and applied to predict Gulf of Mexico flows.
What makes this framework particularly exciting is its flexibility and efficiency, not only was its ability to directly utilize Hyperspar's observational measurements for the basis of prediction, but its ability to train and execute them using a laptop or desktop computer.
Furthermore, the architecture of the model allows for easy interrogation of its temporal and spatial behavior, allowing it to better de-describe and explain its properties.
Dr. Edward Steele, who has met a fellow at Data Science and lead author of the paper, said: “AI-based predictions can revolutionize ocean prediction in a variety of ways. Our research demonstrates the exciting potential for very low-cost observation-driven, AI-based, AI-based models to provide promising results, even when constrained by the range of available data.”
“We are still in the early stages of refinement, but further development is needed to fully realize the expected benefits, but we are poised to support a variety of use cases with key offshore energy, ocean search and rescue and defence applications (particularly) that show that AI can quickly and reliably support marine process predictions.”
Dr. Ajit Pillai, senior lecturer at the University of Exeter and Royal Engineering Research Fellow, said: “This is an exciting application of the MALCOM framework for new parameters and new geographical regions, demonstrating the versatility of an AI-based approach and providing new decision-making capabilities to support offshore safety and operation.”
Pioneering partnerships
This study is a milestone of teams' ongoing work that will help ensure safe, efficient and successful operations for people working in the ocean through more accessible, accurate and faster predictions.
Further upgrades of the model are planned to further adjust the Malcom framework for marine current prediction and complement the main focus of the team's research on regional wave prediction.
This first concept, which later evolved into the Malcolm framework, originally began as an experimental research project led by the University of Exeter almost five years ago, with the Met Office supporting participants.
Together, the team worked on developing an approach and benchmarking AI-based model results, achieving performance performance on operational MET Office Physics-based wave models under typical (non-extreme) conditions 12 hours prior to testing new approaches, use cases, regional domains and marine variables. Throughout, they worked directly with industry operators to ensure that development complements decision-making and workflow.
Dr. Steel continues. “This is an example of how academic, government and industry organisations work together to develop new approaches and capabilities that are useful, usable and used. Also, scientifically inspiring and enjoyable collaboration is essential in rapidly evolving areas such as machine learning, partnerships are especially vital to realizing AI ambitions.”
Dr. Pirai commented: “It's always exciting that research will have a real impact, and the team's perception of work through the ASCE OTC Best Paper Award 2025 reflects some of the advances so far. These are exciting times.”
