Testing robustness of CPUE standardization and inclusion of environmental variables with simulated longline catch datasets

Abstract

Environmental variability changes the distribution, migratory patterns, and susceptibility to various fishing gears for highly migratory marine fish. These changes become especially problematic when they affect the indices of abundance (such as those based on catch-per-uniteffort: CPUE) used to assess the status of fish stocks. The use of simulated CPUE data sets with known values of underlying population trends has been recommended by ICCAT (International Commission for the Conservation of Atlantic Tunas) to test the robustness of CPUE standardization methods. A longline CPUE data simulator was developed to meet this objective and simulate fisheries data from a population with distinct habitat preferences. The simulation was used to test several statistical hypotheses regarding best practices for index standardization aimed at accurate estimation of population trends. Effort data from the US pelagic longline fleet was paired with a volume-weighted habitat suitability model for blue marlin (Makaira nigricans) to derive a simulated time series of blue marlin catch and effort from 1986-2015 with four different underlying population trends. The simulated CPUE data were provided to stock assessment scientists to determine if the underlying population abundance trend could accurately be detected with different methods of CPUE standardization that did or did not incorporate environmental data. While the analysts' approach to the data and the modeling structure differed, the underlying population trends were captured, some more successfully than others. In general, the inclusion of environmental and habitat variables aided the standardization process. However, differences in approaches highlight the importance of how explanatory variables are categorized and the criteria for including those variables. A set of lessons learned from this study was developed as recommendations for best practices for CPUE standardization.

Long‐Term Trends in Juvenile Atlantic Sturgeon Abundance May Signal Recovery in the Hudson River, New York, USA

“…Therefore, standardized abundance indices are often used to account for factors (e.g., environmental conditions, fishing effort, and migratory patterns) that may lead to changes in catch rates rather than changes in true abundance (Maunder and Punt 2004; Mateo and Hanselman 2014; Forrestal et al. 2018).…”

The environment drives Atlantic bluefin tuna availability in the Gulf of Lions

“…A clear path to provide a robust and reliable abundance index is therefore to remove the effect of changes in availability to only retain the signal related to changes in abundance. Appropriate standardisation of fisheries‐dependent indices for the varying environment might help to deal with this aspect, providing that the right processes are appropriately detected and dealt with (Forrestal et al., 2019; Teo & Block, 2010).…”

Seasonal Distribution and Habitat Preferences of Crimson Seabream Parargyrops edita: Implications for a Marine Protected Area in Beibu Gulf, Northern South China Sea

“…How to more reliably identify and quantify the relative importance of ecological factors in influencing fish distribution from limited scientific survey data still remains challenging (Agenbag et al 2003;Basille et al 2008;Gill et al 2019). However, GAMs do not have specific underlying assumptions for relationships between response and predictor variables, and therefore such relationships could be linear, polynomial, logarithmic, or take on other nonlinear forms (Cury et al 1995;Swartzman et al 1995;Maury et al 2001;Grüss et al 2014;Forrestal et al 2019). As a nonparametric extension of the traditional generalized linear models, generalized additive models (GAMs) are increasingly used to identify the relative importance of environmental and ecological variables, such as predictors for fish distribution at broad spatial and temporal scales.…”

“…As a nonparametric extension of the traditional generalized linear models, generalized additive models (GAMs) are increasingly used to identify the relative importance of environmental and ecological variables, such as predictors for fish distribution at broad spatial and temporal scales. However, GAMs do not have specific underlying assumptions for relationships between response and predictor variables, and therefore such relationships could be linear, polynomial, logarithmic, or take on other nonlinear forms (Cury et al 1995;Swartzman et al 1995;Maury et al 2001;Grüss et al 2014;Forrestal et al 2019).…”