Accounting for vessel effects when standardizing catch rates from cooperative surveys

Abstract

Interpretation of fishery-dependent and independent-survey data requires accounting for changes in the proportion of local individuals that are caught by fishing gear (“catchability”). Catchability may be influenced by measured characteristics of fishing gear, and even standardized fishing techniques may experience changing catchability over time due to changes in fishing vessel characteristics and personnel. The importance of vessel power has long been recognized in the analysis of fishery dependent catch per unit effort data, but less-studied in the analysis of fishery independent data collected by research vessel surveys. Here we demonstrate how differences in catchability among vessels (“vessel effects”), as well as random variation in vessel-specific catchability over time (“vessel-year effects”) can be incorporated into generalized linear mixed models through their treatment as random effects. We apply these methods to data for 28 groundfish species caught in a standardized survey using contracted fishery vessels and personnel in the Northeast Pacific. Model selection shows that vessel, vessel-year, and both effects simultaneously are supported by available data for at least a few species. However, vessel-year effects generally have a larger effect on catch rates than vessel-effects and hence abundance indices estimated using both vessel- and vessel-year effects are generally similar to estimates when using just vessel-year effects. Additionally, models indicate little support for the hypothesis that characteristics such as length and displacement of the contracted vessels used in this survey have a substantial impact on catch rates. Finally, inclusion of vessel- or vessel-year effects generally results in wider estimates of credible intervals for resulting indices of abundance. This increased credible interval width is consistent with statistical theory, because vessel effects will result in non-independence of different sampling occasions, thus decreasing effective sample sizes. For this reason, we advocate that future analyses include vessel- and/or vessel-year effects when standardizing survey data from cooperative research programs.