Rutgers researchers use AI and ocean data to predict whale locations near offshore wind areas
Rutgers researchers built an AI lookout for endangered whales
Fewer than 370 North Atlantic right whales still swim the Atlantic — and until recently, the ships, fishing vessels, and offshore wind operators sharing those waters had no reliable way to know where any of them were at any given moment.
For decades, ocean scientists had been quietly building up the kind of data that could change that: whale sightings, underwater glider readings, satellite measurements stretching back to 1992. The problem wasn’t a lack of information. It was the missing link between what the ocean recorded and what decision-makers could actually use in time to matter.
“This approach can support a wise and environmentally responsible use of these waters,” said Ahmed Aziz Ezzat, the project’s lead researcher.
A species on the edge, sharing a crowded ocean
North Atlantic right whales have been listed as endangered under the Endangered Species Act since 1970. Today, roughly 370 individuals remain — and only about 70 of those are reproductively active females, according to NOAA. That margin leaves almost no room for error.
The Atlantic coast they inhabit is anything but quiet. Shipping lanes, commercial fishing operations, and a rapidly expanding offshore wind industry all compete for the same water. Each brings its own collision and disturbance risks. A single vessel strike can be fatal, and the cumulative pressure on such a small population is severe.
Existing monitoring methods haven’t kept pace. Aerial surveys and acoustic buoys provide snapshots, not continuous coverage. By the time a sighting gets reported, the whale has moved on and a ship is already on course. What was missing wasn’t concern — it was prediction.
Two databases, one powerful connection
The breakthrough came from a decision to stop treating two rich data sources as separate problems. Rutgers researchers merged whale detection records with decades of environmental ocean data — two datasets that had never been formally linked before.
The environmental side of that equation is substantial. Since 1992, the Rutgers University Center for Ocean Observing Leadership has deployed autonomous underwater gliders along the Mid-Atlantic coast. These torpedo-shaped vehicles measure temperature, salinity, current strength, chlorophyll levels, and fish density — and they record the underwater calls of whales, pinpointing animals in time and space.
Satellite data filled in the surface picture. The University of Delaware provided publicly available measurements of sea surface temperature, water color, and ocean fronts — the kinds of features that shape where prey concentrates and, by extension, where whales feed. Together, those streams gave the AI something it could actually learn from: a long, detailed record of what the ocean looked like every time a whale was — and wasn’t — detected.
How the machine learning model works
Standard computer programs follow explicit instructions. This one doesn’t. The machine learning model analyzed the combined datasets and found patterns on its own — connections between whale presence and specific environmental conditions that no researcher had manually coded in.
Josh Kohut, a marine sciences professor at Rutgers, described the logic with a simple analogy: tracking where people are in a house by noticing whether there’s food in the kitchen or a TV on in the den. Context predicts behavior. The same principle applies underwater. Certain water temperatures, chlorophyll concentrations, and acoustic signatures reliably signal that right whales are likely nearby.
The output is what researchers call a “probability map” — a spatial and temporal picture of where whale encounters are most likely. As the model processes more data over time, its predictions get sharper. “We can predict the time and location that represents a higher probability for whales to be around,” Kohut said. “This will enable us to implement different mitigation strategies to protect them.”
From wind farms to the wider blue economy
The tool was originally built with a specific problem in mind: supporting responsible offshore wind farm development in the U.S. Mid-Atlantic, where turbine construction and vessel traffic were adding pressure on right whale habitat. High-resolution habitat models were needed before operations could be planned responsibly.
The researchers quickly recognized, though, that its reach extended well beyond wind energy. Shipping companies, fishing fleets, and any industry operating in the blue economy could use the probability maps to adjust routes or operations when whale presence is predicted to be high — making mitigation strategies actionable rather than reactive.
The study was published in Nature Scientific Reports, with full technical details made publicly available alongside the paper — a deliberate choice to encourage adoption across sectors. “This approach can support a wise and environmentally responsible use of these waters,” said Ahmed Aziz Ezzat, the project’s lead researcher.
What comes next for whale protection
Probability maps are useful on paper — but their real value lies in integration. Researchers envision the tool being embedded into real-time maritime navigation systems and regulatory frameworks, giving ship operators advance warning before entering high-risk zones rather than after.
The model’s architecture isn’t limited to right whales. The same approach could be adapted for other marine species, extending its conservation reach well beyond a single endangered population. Broader adoption, though, will require more than good science — shipping companies, federal regulators, and conservation organizations would all need to coordinate around a shared system, which is a significant institutional challenge.
Still, the foundation is now in place. For the first time, the ocean’s own environmental signals can be read as a predictive guide to where its rarest animals will be. Whether that capability translates into fewer collisions and smarter development will depend on who picks it up next.
Discover more here: Jiaxiang Ji, Jeeva Ramasamy, Laura Nazzaro, Josh Kohut, Ahmed Aziz Ezzat. Machine learning for modeling North Atlantic right whale presence to support offshore wind energy development in the U.S. Mid-Atlantic. Scientific Reports, 2024; 14 (1) DOI: 10.1038/s41598-024-80084-z
Daniel García is an Editor-in-Chief with strong expertise in structural work and engineering principles. He combines this technical foundation with deep knowledge of energy, spatial design, and emerging technologies, bringing a forward-thinking and analytical approach to editorial leadership.
