Hierarchical analysis of a remote, Arctic, artisanal longline fishery

• Published in: ICES Journal of Marine Science
• Main Authors: Dennard, Susan T., MacNeil, M. Aaron, Treble, Margaret A., Campana, Steven, Fisk, Aaron T.
• Format: Article in Journal/Newspaper
• Language: English
• Published: Oxford University Press (OUP) 2009
• Subjects: Arctic; Baffin Island; Baffin; Cumberland Sound; Greenland; North Atlantic; North Atlantic oscillation
• Online Access: http://dx.doi.org/10.1093/icesjms/fsp220; http://academic.oup.com/icesjms/article-pdf/67/1/41/29134955/fsp220.pdf

Abstract Dennard, S. T., MacNeil, M. A., Treble, M. A., Campana, S., and Fisk, A. T. 2010. Hierarchical analysis of a remote, Arctic, artisanal longline fishery. – ICES Journal of Marine Science, 67: 41–51. This is the first paper to explore trends in catch per unit effort (cpue) through time of a Greenland halibut Reinhardtius hippoglossoides stock targeted by an artisanal, winter fishery in Cumberland Sound on southern Baffin Island, Canada. We modelled cpue data from 1987 to 2003, looking at two questions: what factors have driven cpue trends, and is cpue an accurate index of a stock's abundance? In the context of limited data availability, we used generalized linear models (GLMs) and hierarchical models to assess important predictors of cpue. Hierarchical models with multiple fixed environmental effects contained fishing location or individual fisher as random effects. A month effect showed greatest catch rates during February and March; the monthly North Atlantic Oscillation index was positively associated with catch rates; and a change from decreasing to increasing cpue after 1996 was linked to reduced fishery participation following a large storm. The best Akaike's information criterion-ranked GLM identified a negative relationship of cpue with shark bycatch. Although data limitations precluded conventional stock assessment, our models implicated the environment and fisher behaviour as drivers of cpue trends. Additionally, using multiple hierarchical models to predict cpue provided a more informative analysis for understanding trends in cpue than a GLM alone.