Performance of methods used to estimate indices of abundance for highly migratory species

Abstract

Estimating indices of abundance from fishery-dependent data requires that catch-per-unit-effort (CPUE) be standardized to account for factors that may have affected CPUE but are not related to changes in abundance. Such standardization is particularly important for highly migratory species (e.g., tunas, pelagic sharks, and billfishes), because of time-varying mismatches between distributions of abundance and the distribution of fishing effort. Two commonly applied methods for standardizing CPUE are generalized linear models (GLMs), which can account for changes in fishing practices in a straightforward linear fashion, and habitat-based standardizations (e.g., statHBS), which use nonlinear analysis to relate the distribution of fishing effort to the species distribution. We evaluated the accuracy of these methods over three patterns in vertical catchability as related to ocean temperature profiles, and 50 possible biomass trajectories using a simulation framework that followed the general effort dynamics of the Japanese longline fishery in the Atlantic Ocean from 1956 to 2009. Additionally, we propose a method for directly incorporating vertical habitat information into the linear models. Overall, we found the most accurate approach to be a delta-lognormal GLM with our unique habitat factor. The statHBS approach was the most accurate when catchability was simulated to peak in surface waters. However, statHBS was much more sensitive to errors in estimates of longline hook depths (i.e., habitats exploited). Based on these results, we recommend that relative abundance be estimated for highly migratory species following a delta-GLM approach that considers vertical habitats fished.