The North Atlantic right whale is the rarest of all large whales. Today, only about 400 individuals remain. Also known as Eubalaena glacialis, it was hunted to the brink of extinction by the early 1890s. Though the North Atlantic right whale has been protected under the International Convention for the Regulation of Whaling since the 1930s, the species has still not recovered. Their decline has been driven by human impacts. At least 17 individuals died in 2017 alone, many due to vessel strikes and entanglements with fishing gear. Given the small population size, the threats the whales face,  and their endangered status, it is crucial that the population abundance is closely monitored.

Ideally, researchers would be able to track individual whales, but this is often not possible due to the difficulty of the feat and the vast resources that would be needed. As an alternative, researchers will usually estimate the population size of a species through mark recapture where a percentage of a population is captured, marked, and then released. As you can imagine, however, capturing and marking right whales would be incredibly invasive and difficult. With this method, researchers are then able to estimate population size by taking repeated samples and assuming that the ratio of marked to unmarked animals is proportional to the greater population.

Photo identification provides an alternative. This method distinguishes between individuals using natural markings. Manually telling the difference between individuals, however, requires significant training and time. In a field that is often underfunded, the time of these researchers is precious.

During a data science competition in 2018, researchers from Poland and the United States came up with a machine learning algorithm that offers a solution. Using distinctive white patches of rough skin on top of the right whale’s head, the algorithm can identify specific whales with 87% accuracy. As you can imagine, this makes the photo identification of right whales faster, easier, and less resource intensive.

Rough, white skin atop the head of a North Atlantic Right Whale.

The researchers are now working to put their algorithm on Flukebook, a free online resource that uses machine learning to identify and track whales and dolphins using photos. Additionally, they believe that a similar technique could be successfully applied to other species with distinct markings, from other species of whales to meerkats.

Though there is room for improvement, this application of machine learning is an important step forward for not only the protection of North Atlantic right whales, but also conservation at large.

Bogucki, R., Cygan, M., Khan, C.B., Klimek, M., Milczek, J.K. and Mucha, M. (2019), Applying deep learning to right whale photo identification. Conservation Biology, 33: 676-684. doi:10.1111/cobi.13226