Bering Sea | Fishes and Fish Habitats MASTER’S ORAL PRESENTATION A comparison of species distribution model structures for describing juvenile salmon distribution and abundance in the eastern and northern Bering Sea Presenter: Lilian Hart , lkhart2@alaska.edu, University of Alaska Fairbanks Curry Cunningham , cjcunningham@alaska.edu Ellen Yasumiishi , Ellen.Yasumiishi@noaa.gov, Auke Bay Laboratories, National Oceanic and Atmospheric Administration Wild Pacific salmon ( Oncorhynchus spp. ) fisheries are integral to Alaska and Alaskans for economic health, for food, and for sustaining traditional cultures. Fisheries management agencies invest significant time and effort into the sustainable management of Alaska’s wild fisheries. However, the majority of past research effort and data collection has focused on adult life stages, and there remains a paucity of knowledge about juvenile salmon ecology during their first months after ocean entry. Information gathered in marine ecosystem surveys conducted by the NOAA Alaska Fisheries Science Center and Alaska Department of Fish and Game provide an opportunity to better understand juvenile salmon distribution and habitat utilization during this important life history stage. The object of this study was to describe the distribution and relative abundance of juvenile Pacific salmon species in the Bering Sea. Our first objective was to quantify the temporal variation in core habitat areas for juvenile salmon in the Bering Sea; specifically, whether core habitat areas persist or shift across time. Our second objective was to compare the performance of the two classes of applicable species distribution models (SDMs): generalized additive models (GAMs) and Vector Autoregressive Spatiotemporal (VAST) models. Maps of predicted juvenile salmon distributions in the eastern Bering Sea showed the greatest densities in nearshore habitats along the 50m isobath of the continental shelf. Predicted distributions differed among species independent of the SDM used. Although the GAM and VAST models that estimated time-varying spatial distributions performed better in terms of information theoretic metrics, visual analysis found that core areas largely persisted across the timespan of the survey. VAST models outperformed GAMs in terms of AIC and percent deviance explained across all model configurations, for all species in the study. We therefore conclude that a species-distribution model approach can be effective for estimating both static and dynamic fish distributions, and that VAST is an especially appropriate tool for this application in conditions where abundance observations are spatially and/or temporally heterogeneous.
Alaska Marine Science Symposium 2023 48
Made with FlippingBook flipbook maker