A steady “snowfall” of tiny particles carries carbon, nutrients and pollution from the ocean’s surface toward the deep sea. Scientists can see those particles sinking, but determining exactly what they contain is much harder.
University of Maine researchers are turning to artificial intelligence to help solve that problem.
The researchers are developing AI tools that could determine the chemical contents of “marine snow,” particles made up of organic matter, minerals and other material that continually sink through the ocean.
If successful, the approach could help scientists learn more from the underwater images they already collect, while reducing the time spent manually classifying particles and improving estimates of how material moves through the ocean.
The National Science Foundation awarded nearly $700,000 to UMaine researchers Meg Estapa and Chaofan Chen for the three-year project, which is scheduled to begin in January 2027 and run through December 2029.
“That sinking marine snow is what makes the deeper part of the ocean have a different chemistry than the surface of the ocean,” Estapa said. “The acidity is different. The oxygen is different. The types of microorganisms and fish that live there are different.”
Underwater cameras can capture large numbers of marine snow particles and reveal characteristics such as their size, shape and transparency. Images alone, however, generally cannot tell scientists what the particles are made of.
Teaching AI to read the ocean
Estapa and Chen, along with collaborator Melissa Omand from the University of Rhode Island, will investigate whether AI can bridge that gap by predicting particle composition from characteristics visible in underwater images.
Researchers have used image-analysis tools to study marine snow for years, but much of that work remains labor intensive. Estapa said one of her graduate students spent months classifying particles and identifying what appeared in underwater images.
Her own work studying particles in the ocean has evolved over roughly two decades. Early in her career, Estapa said, the process involved collecting water samples at sea, bringing them back to the laboratory and spending significant amounts of time analyzing them for individual data points.
AI has the potential to accelerate that process, allowing scientists to collect and interpret more information while spending more time on scientific analysis and discovery.
The researchers will use data from six major oceanographic field campaigns to test whether groups of particles visible in images can accurately predict their chemical composition. The data include images collected from waters off West Africa, the North Atlantic and tropical regions.
Teaching AI what ‘marine snow’ is made of
The project will also provide interdisciplinary training for graduate students in artificial intelligence and oceanography, combining advanced research with hands-on learning in keeping with UMaine’s R1, learner-centered mission.
During the first stage of the project, Estapa’s group will compile a database pairing marine snow images with information about what the particles contain, how much microplastic is present and where the samples were collected. Some physical samples collected for the project will also undergo laboratory analysis.
Chen will develop AI models that can learn from those data while also helping scientists understand how the models reach their conclusions.
“The overall goal would be to broaden the application of these techniques,” Chen said. “We are more interested in designing neural networks that easily explain conclusions to humans.”
That emphasis on interpretability is central to the project. Rather than creating a “black box” system that produces predictions without explaining them, Estapa and Chen plan to develop a model that shows scientists which characteristics influenced its conclusions.
“Every new scientific tool you want to approach with care,” Estapa said. “You want to understand what the benefits are and the places where this tool will trip you up and mislead you.”
Tracking carbon, nutrients and microplastics
Microplastics will be one focus of the project. By identifying and quantifying plastic particles alongside naturally occurring marine snow, the researchers hope to better understand how plastic pollution moves from surface waters into the deep ocean and through marine food webs.
The same tools could help researchers better understand the natural cycling of carbon and nutrients through the ocean.
The approach could have applications outside oceanography. Scientific fields ranging from ecology to medicine increasingly rely on complex images that can be difficult and time-consuming to interpret.
For Chen, the project also represents a new application of his work in interpretable artificial intelligence.
“This will be one of my first projects actually using AI for scientific discovery,” Chen said. “It’ll be an exciting new thing for me.”
If successful, the approach could allow oceanographers to extract more information from large collections of underwater images gathered during research expeditions and better estimate how carbon, nutrients and pollutants move through the ocean.
“We hope this project will help us understand our ocean better — and our impacts on the ocean directly and indirectly,” Estapa said.
Contact: David Nordman, david.nordman@maine.edu
