Oceanographic Influences on Tuna Dive Behaviors

Simon Dedman1, Michael Castleton2, Robert Schallert2, Mike Stokesbury3, James Ganong4 and Barbara Block2, (1)Stanford University, Hopkins Marine Station, Pacific Grove, CA, United States, (2)Stanford University, Oceans Department, Pacific Grove, United States, (3)Acadia University, Biology, NS, Canada, (4)Stanford University, Hopkins Marine Station, United States
Abstract:
The proliferation of high-resolution biologging data sets is changing pelagic fish biology from data poor to data rich. Novel data science approaches are crucial to maximally benefit from these riches. Machine learning analyses of such large datasets can identify key influential variables, elucidating biological and environmental linkages. We have deployed over 1200 archival and pop-up satellite tags on Atlantic bluefin tuna over two decades and have been examining oceanographic and environmental correlates in this rich >50,000 day data set. We are exploring using machine learning techniques to relate these correlates (e.g. bathymetry, SST, DO2, eddies) to specific dive behavior types. The results reveal how bluefin tuna tuna utilize the ocean: how they behave differently based on location, age and season, and how hunting, navigating, traveling, and spawning behaviors relate to environmental conditions. The spatio-temporal hotspots for behavior identified by our results have the capacity to improve fisheries management decisions regarding spatial and/or temporal closure discussions. By simultaneously considering the full suite of elements affecting bluefin tuna movements, we gain insight into how these fish can be impacted by anticipated changes in climate. For example, increasing temperature can shift both prey fish movements and tuna spawning timing, collectively disrupting annual movement patterns. However, increased eddy frequency and strength can enhance access to mesopelagic food resources. Our analyses show how tuna diving behavior is influenced by multiple oceanographic factors, age, location, and season. This approach represents a framework which can be applied to similar biologging large scale datasets (e.g. Pacific bluefin, TOPP sharks), and the results may lead to smarter programming decisions for future biologging tags.