2023 AMSS Abstract Book

Bering Sea | Fishes and Fish Habitats

Predictive distribution models to support flexible management of Bristol Bay red king crab Presenter: Emily Ryznar , emily.ryznar@noaa.gov, NOAA - AFSC Mike Litzow , mike.litzow@noaa.gov, Alaska Fisheries Science Center, NOAA Fisheries The Bristol Bay stock of red king crab ( Paralithodes camtshaticus , hereafter “BBRKC”) has experienced a decadal-scale period of very low recruitment, resulting in a persistent, gradual decline in the abundance of mature animals that triggered directed fishery closures in 2021/22 and 2022/23. Northward shifts in distribution have also been observed for BBRKC in recent years, creating concern around the suitability of statically-defined management boundaries and awareness of the possible need for flexible management to account for observed distribution shifts. However, there is a large disconnect in our understanding of BBRKC distribution between the data-rich summer period and data-poor fall fishing seasons when management boundaries and closure areas are most relevant. In addition, the low abundance of the stock has increased interest in the role that bycatch may play as a source of BBRKC mortality, but this has yet to be explored in a spatiotemporal predictive capacity. The goal of this project is to develop dynamic species distribution models (SDMs) to improve the scientific understanding of the factors regulating BBRKC distribution as a basis for possible development of flexible management structures. Our specific objectives are: to 1) to predict fall distribution; and 2) to predict bycatch in non-pelagic trawl (NPT) groundfish fisheries. For the first objective, we used summer bottom trawl survey data (BBRKC catch, bottom temperature, and depth), sea ice percent cover, sea surface temperature, and sediment grain size to predict the distribution of BBRKC during the fall directed fishing season. For the second objective, we used the same process covariates to predict BBRKC bycatch probability in NPT fisheries by year in September- October and April-May, as 10-year average BBRKC bycatch rates were highest in these months. For both objectives, we trained Boosted Regression Trees (BRTs) and evaluated predictive performance using area under receiver-operator curve (AUC-ROC) metrics. Preliminary results indicate that summer bottom temperature is an important predictor for fall BBRKC distribution and bycatch. We will present spatiotemporal predictions, covariate response curves, and AUC-ROC metrics for both objectives. Future directions will integrate ongoing BBRKC tagging data to help resolve habitat use and distribution. Further, other model frameworks may be explored for their predictive ability and out-of-sample predictive performance.

Alaska Marine Science Symposium 2023 217

Made with FlippingBook flipbook maker