June 9, 2026 — A University of Alaska Fairbanks research team has translated a trove of data from a Chinook salmon tagging program into a predictive model that could help reduce bycatch by fishing trawlers.
Chinook salmon range from the ocean’s surface to depths where trawl nets target groundfish species. The researchers’ model uses more than 700,000 data points between Southeast Alaska and the Bering Sea to predict how Chinook will be distributed across the water column. With that information, trawlers can potentially adjust their operations to reduce inadvertent salmon catches.
To develop the model, the team used 13 years of data from tagged Chinooks through an ongoing project led by Andrew Seitz, a professor at UAF’s College of Fisheries and Ocean Sciences. The tags collect data every 5-10 seconds and transmit it to satellites. That provides a much more comprehensive dataset than is available from catching tagged salmon.
Tagging data included details such as depth, time of day, temperature and light levels. A separate dataset from Copernicus Marine Service added environmental context.
Graduate fisheries student Marcel Gietzmann-Sanders used machine learning to detect patterns in that massive dataset. He turned them into detailed charts showing the depth where Chinooks are most likely to be found at various times and locations. The project is described in a recently published paper in the journal Animal Biotelemetry.
“It shows what you can find out when you collect enough data about a species over a long enough time,” Gietzmann-Sanders said.
Better tools for reducing bycatch are important for both the health of declining Chinook populations and the economics of the trawl fleet. Fishing for pollock, a type of groundfish that supplies the largest catch volumes in the U.S., can be shut down if bycatch limits are exceeded. As recently as 2024, two fishing boats had enough Chinook salmon bycatch over a single weekend to close the fishery.
Read the full article at University of Alaska Fairbanks News