Insights into black-footed albatross bycatch in US west coast fisheries using Bayesian models

Globally, seabird populations are declining, due to factors such as pollution, climate change, introduction of predators, and interactions with fisheries. Incidental bycatch in U.S. West coast longline fisheries significantly contributes to the mortality of the black-footed albatross (BFAL). BFAL populations are stable but predicted to decline over the next half century. My project focuses on testing multiple Bayesian time series models to understand the factors that contribute to BFAL bycatch in a U.S. West Coast longline fishery. We tested model performance assuming two different distributions of the bycatch (negative binomial, Poisson), and also examined constant versus non-constant bycatch rates. We also incorporated various combinations of seven unique covariates thought to influence bycatch rates to determine if a relationship exists between these covariates and the bycatch rate. We used leave-one-out cross validation to quantitatively evaluate the models. Results indicated that the best model of seabird bycatch used a negative binomial distribution with a constant bycatch rate. These results indicate that the distribution of bycatch is dispersed and that the underlying bycatch rate from year to year remains relatively constant. Further, we tested five covariates to determine the factors that could affect bycatch. Of the covariates used, streamer lines, season, and use of floated longlines had the most predictive results. This predictive capacity indicated that the types of longlines used, along with streamer lines and the time of year that the fishing took place all have the capacity to affect the bycatch rate. The results of our work demonstrate clear improvements over previous methods of bycatch modeling and suggest factors that managers might use to reduce seabird bycatch.