Billfish CPUE standardization in the Hawaii longline fishery: Model selection and multimodel inference

Abstract

This paper presents catch per unit effort (CPUE) standardizations and model selection procedures for four billfish species (Family Istiophoridae) caught primarily as bycatch in the Hawaii-based pelagic longline fishery during 1995–2011: Blue marlin Makaira nigricans; Striped marlin Kajikia audax; Shortbill spearfish Tetrapturus angustirostris; and Sailfish Istiophorus platypterus. The first three species were analyzed on a fishery-wide basis. For sailfish, the fishery data came exclusively from tuna-targeted longline sets in the deep-set sector of the Hawaii-based fishery. We used fishery observer data from the NOAA Fisheries Pacific Islands Regional Observer Program to fit the CPUE standardization models. In this context, our objective was to investigate the quality of model fit for five types of generalized linear models (GLMs: Poisson; negative binomial; zero-inflated Poisson; zero-inflated negative binomial; delta-Gamma). Each of these models represented a different hypothesis about the capture process for a bycatch species for which the catch data primarily consisted of zero catch observations. The five GLMs were fitted by forward entry variable selection, and the best fitting GLM for each species was selected on the basis of Akaike Information Criterion values and calculated Akaike weights. The best-fitting model selected for each species was a zero-inflated negative binomial GLM (ZINB). The ZINB model was comprised of a negative binomial counts model for expected zero catch sets and a positive catch per set distribution along with a binomial inflation model to account for excess zeros. For each species, the important explanatory variables for standardizing CPUE were fishing year, fishing (i.e., calendar) quarter, and fishing region. The best-fitting models indicated that standardized CPUE for striped and blue marlins decreased significantly during the study period. Because the ZINB model was selected as the best fitting model for all species, we suggest that longline CPUE for incidentally caught billfishes is best represented as a process characterized by zero inflation and overdispersion in the positive catches and expected zero catches. We therefore recommend that ZINB models be considered as an a priori model for CPUE standardizations of billfishes and other bycatch species in longline fisheries.