The research, led by Jay Kirkham of Dalhousie University in Canada together with colleagues in Canada and the United States, tackled a long-standing problem in right whale biology: scientists can rarely watch these animals eat. North Atlantic right whales (Eubalaena glacialis) spend their summers foraging in the coastal waters of the eastern United States and Canada, swimming slowly through clouds of zooplankton with their mouths agape. Specialized baleen plates lining their jaws sieve tiny copepods and other prey from the water in a process called skim feeding. Yet most of this foraging happens at depth, far below the reach of human observers, so researchers have had to rely on indirect clues, such as GPS tracks, to infer what the whales were doing underwater.

Interpreting those indirect signals has always been difficult. A whale that dives repeatedly might be feeding, traveling, resting, or socializing, and the movement signatures of these activities can overlap. Misreading the data matters, because feeding behavior is the foundation of conservation planning. Vessel exclusion zones, seasonal speed restrictions, and fishing gear closures are all timed and placed according to when and where whales are expected to be feeding. If the underlying behavioral assumptions are wrong, protections may be drawn in the wrong places or at the wrong hours, with potentially fatal consequences for a population that numbers only a few hundred individuals.

To close this knowledge gap, the team deployed sophisticated suction-cup attached biologging tags on 23 North Atlantic right whales in the Gulf of St. Lawrence off the coast of Canada. These multi-sensor packages were far more than simple location trackers. Each tag carried an accelerometer to record the fine-scale accelerations of the whale’s body, a gyroscope to capture rotation and posture, a magnetometer to log orientation relative to Earth’s magnetic field, a depth recorder to track vertical movements, an underwater microphone to capture the sounds of feeding and the surrounding environment, a light sensor to distinguish day from night, and, crucially, a video camera that recorded what the whale actually saw as it swam.

The video proved to be the key that unlocked everything else. By reviewing the camera footage, the researchers could identify moments when a whale opened its mouth and skimmed prey from the water, then match those confirmed feeding events to the precise kinematic patterns recorded by the accelerometers and gyroscopes at the same instants. In effect, the cameras provided ground truth for the motion sensors. The team then used this camera-validated dataset to train machine learning models to recognize the movement signatures of feeding from the kinematic data alone, allowing them to classify behavior across the full record, including the many hours when no camera footage was available or usable.

The best-performing model proved remarkably reliable, agreeing with the researchers’ manual identification of feeding behavior 95 percent of the time. That level of accuracy matters, because it transforms hours of ambiguous sensor data into a trustworthy behavioral diary. With the validated models in hand, the researchers could quantify, for the first time with this precision, how much of a right whale’s day is actually devoted to eating. The answer was striking: the tagged whales spent between 11 and 76 percent of their time feeding, averaging 11.4 hours of feeding per day, a schedule that leaves little room in the animal’s energy budget for anything other than finding enough plankton to survive.

The most consequential discovery, however, was not how much the whales ate but when and where they did it. During daylight hours, the whales typically fed close to the seafloor, skimming prey from near the bottom. But as darkness fell, they shifted their foraging upward, moving toward the surface to feed in the top layers of the water column. This vertical migration almost certainly follows the prey: zooplankton species undertake their own daily migrations, rising toward the surface at night to feed on phytoplankton under cover of darkness and descending during the day to avoid visual predators. The whales, in other words, track their food up and down the water column on a daily cycle.

Nighttime surface feeding had not previously been reported for North Atlantic right whales, and the oversight has serious implications. Surface waters are where whales and ships intersect. A whale feeding just beneath the surface at night is difficult for a vessel operator to see and difficult for the whale itself to detect and avoid. The study’s data showed that the tagged whales spent roughly 25 percent of their feeding time outside designated vessel exclusion zones, the areas from which motorized boats are prohibited, placing them directly in harm’s way. If a substantial share of that unprotected feeding occurs at night near the surface, the collision risk posed by shipping traffic may be significantly higher than current management frameworks assume.

The study also delivered a first for right whale science: the first video footage ever recorded of North Atlantic right whales feeding at the seafloor. The tag cameras documented whales descending to the bottom and foraging there, confirming that benthic feeding is a genuine part of the species’ repertoire rather than an inference from dive profiles. The footage captured other remarkable moments as well, including interactions between individuals, apparent synchronized dive timing between paired whales, subsurface body contact, and one whale that dove to the seafloor on three consecutive dives, rolled, and appeared to drag its back along the bottom. Together, these glimpses reveal a species whose underwater lives are richer and more varied than surface observations alone could ever suggest.

The authors emphasize that these previously unreported surface foraging trips at night may have reduced the effectiveness of existing conservation protections, and they hope the findings will help managers refine vessel speed restrictions and exclusion zones to reflect the whales’ true behavior around the clock. The work was funded by Fisheries and Oceans Canada through the Canada Nature Fund for Aquatic Species at Risk, and the authors declared no competing interests. For a species pushed to the brink by ship strikes and fishing gear entanglements, and whose zooplankton prey is being redistributed by climate change in ways that may draw whales out of protected areas altogether, understanding where the animals feed, hour by hour and depth by depth, is not an academic luxury. As the researchers note, being able to study one of the most endangered whales in the world is an extraordinary opportunity, and filling in the hidden hours of the right whale’s day may prove essential to its survival.

Subject of Research: Nighttime surface feeding behavior of North Atlantic right whales identified using camera-validated kinematic data and machine learning

Article Title: Artificial intelligence reveals novel nighttime feeding behaviors in right whales

Article References: Artificial intelligence reveals novel nighttime feeding behaviors in right whales. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: North Atlantic right whale, machine learning, biologging tags, nighttime feeding, ship strikes, vessel exclusion zones, Gulf of St. Lawrence, zooplankton, conservation, PLOS One, accelerometer, critically endangered