Project Lead: Andrew Berdahl, UW School of Aquatic and Fishery Sciences and Ben Koger, University of Wyoming School of Computing and Department of Zoology and Physiology

Data Science Lead: Valentina Staneva

From daily foraging movements to long-distance migrations, animals often travel in groups, and their navigational decisions are influenced by those around them. Theory suggests that by pooling noisy estimates of travel direction, groups can improve navigational accuracy. Pacific salmon, renowned for their ability to return to their natal streams, are a classic model of animal navigation. Yet despite often migrating in schools, salmon navigation is typically studied in a solitary context, overlooking the potential benefits of collective decision-making. Recent work supports a role for collective navigation: salmon home more accurately to natal streams and navigate river barriers more efficiently when migrating in larger numbers. The mechanism underlying these improvements, however, remains unknown. During their upstream migrations, salmon repeatedly encounter binary choices at river confluences. These decision points are directly analogous to the classic Y-maze animal behavior paradigm, providing a natural system in which to test theories of collective navigation.

To uncover the mechanisms underlying collective navigation in salmon, we collect high-resolution movement data from groups of fish passing through river confluences. Using drones, we record overhead video and then extract the trajectories of individual salmon with computer vision tools developed in collaboration with the eScience Institute. The resulting dataset is equivalent to placing a GPS tag on every fish, providing centimeter-scale spatial resolution at sub-second temporal resolution. These data allow us to test how salmon use social information when making navigational decisions and to determine how collective movement enhances migratory performance.