Artificial intelligence helping to prevent whale strikes

Arial view of a North Atlantic right whale. Image credit: Nick Hawkins / Ocean School
June 24, 2024

Atlantic Canada serves as a hub for the shipping industry and an access point to the Atlantic and Arctic oceans. With ports spread across four provinces, shipping vessels are vital for the transfer of goods.

However, collisions between ships and marine mammals are an ongoing concern. The good news: researchers are finding new ways of predicting shipping movements to avoid encounters with marine life.

Dr. Gabriel Spadon, an international postdoctoral fellow with the Ocean Frontier Institute, is building a new Artificial Intelligence (AI) model to better predict the future movement of ships in the Gulf of St. Lawrence. The goal is to minimize the likelihood of whale strikes from passing ships.

The research is being done in partnership with the smartWhales initiative, led by DHI Water & Environment and WSP, and funded by the Canadian Space Agency (CSA) and Mitacs, in collaboration with Fisheries and Oceans Canada and Transport Canada.

Predicting movement patterns

Dr. Spadon began his academic career at home in Brazil, where he studied computer science, including how machine learning can be used to predict the movement of objects through space and time.

“Computer science can be applied in different areas, and it opened the door for me to work on geography and cartography research projects,” says Dr. Spadon.

“This gave me a broad view of what computer science can do and its possibilities of application in other research fields, which led me to work with oceans today.”

Dr. Spadon noticed a gap in researchers’ knowledge. Despite current vessel tracking methods and known migration paths, whales were still being struck by these shipping vessels. He was determined to create a new computer modeling system to better predict the shipping paths and, through the partnership with the smartWhales initiative, prevent encounters between whales and ships.

“We focused less on the specific techniques and more on the problem itself: How do we prevent whale strikes?” says Dr. Spadon.  After meeting with stakeholders, his research team started creating AI models to better understand ships. “If the system can predict where a ship and whale will meet, the operator can notify the vessel and reduce the likelihood of a casualty,” he says. The Institut des Sciences de la mer de Rimouski (ISMER) and the Canadian Whale Institute collaborated on the development of a whale habitat model to predict whale movement. These models were later merged into a single decision support system, which combines all the results into a unified and cooperative solution for preventing strikes.

Whale watching in the St. Lawrence

The Gulf of St. Lawrence is a particularly important region for researchers to study, as many whale species migrate into the Gulf for the summer months, often crossing known shipping routes. North Atlantic Right Whales are one of the species found in this area. Their population was nearly brought to extinction during the commercial whaling period in the late 1800s, and the numbers have yet to recover. “In the Gulf of St. Lawrence, there were more than ten North Atlantic Right Whale fatalities in 2017 alone,” explains Dr. Spadon. With less than 400 of them remaining, many projects are working towards minimizing the mortality of North Atlantic Right Whales.

Diagram of the Gulf of St. Lawrence, showing the various shipping ports, vessels and known whale hotspots, with high uncertainty about the vessel's future location. Diagram extracted from

The smartWhales initiative is one of the projects working to mitigate the risk of collision between whales and ships. “The smartWhale project involves many partners and experts in different areas that can help us achieve a better future for North Atlantic Right Whales,” says Dr. Spadon.

The future of computer modeling

In the computer modeling world, state-of-the-art ocean models often involve extensive and complex programming, which sometimes fails to capture basic relationships derived from the ocean domain knowledge.

To address this issue, Dr. Spadon utilized historical Automatic Identification System (AIS) data and maritime knowledge on the routes shipping vessels tend to travel throughout the changing seasons. Using domain knowledge and machine learning, he developed a new model that can produce top-notch results faster than the current models used for this purpose.

Dr. Spadon divided the Gulf of St. Lawrence into a grid to extract the most likely routes for the vessels. Using this information, the model accounts for changing vessel speeds and how this might impact travel paths, resulting in a more accurate prediction of the vessel’s trajectory. Creating an accurate curve of the vessel’s path is challenging. Still, Dr. Spadon’s model aligns more closely with the movement observed in the real world, which in turn helps researchers identify the vessels at risk of encountering a whale.

Diagram showing previous model predictions (top image) compared to Spadon’s more accurate model prediction (bottom image). Both images show where the ship was tracked (in Green), where the ship was predicted to go (in red), and where the ship ended up going (in blue). Diagram extracted from

“I hope for a future where we have better integration of AI and ocean research,” says Dr. Spadon.

“I see AI being used to improve ocean sensors and give us better information for researchers to keep track of what is happening on and under the water.”

Dr. Spadon will be starting a new position at Dalhousie University, taking on the role of assistant professor in the Faculty of Computer Science.

His post-doctoral fellowship with OFI has created connections and opportunities that have helped to solidify his drive for a more sustainable ocean. “I intend to keep working with [AI computer models] to see its applications for our oceans,” says Dr. Spadon. “We know too little about our ocean compared to its size, and AI can help us understand more about it.”