Multivariate Analyses to Predict and Avoid Bycatch Events

Fisheries managers need to find effective and efficient approaches to managing bycatch of unmarketable fish species. Bycatch of unmarketable species occurs because target species and bycatch species overlap in time and space.  Thus, the best management response depends on a number of factors including understanding the interacting physical, ecological, environmental, and temporal factors that influence bycatch.  However, bycatch can be highly variable in space and time and therefore difficult for managers and fishers to predict.  Abnormally large bycatch events, although rare, do occur, often account for a large proportion of annual species bycatch, and can have serious negative impacts on both the bycatch species population and the fishery.  Rare events, such as abnormally large bycatch events, require special modeling techniques.  To date few studies have compared the efficacy of alternative modeling methods for predicting rare bycatch events.  The goal of this study is to compare the ability of three models (logistic regression, zero-inflated models, and choice-based sampling models) to predict the probability of bycatch events using multiple spatial, temporal and environmental variables.  To facilitate this comparison, I will present analyses of the relationship between environmental variability and bycatch variability of rebuilding fish species in the U.S. Pacific limited entry trawl groundfish fishery.  Utilizing data collected by the West Coast Groundfish Observer Program, this work will provide insight into the predictability and rarity of bycatch distribution and events, inform management tools for reducing bycatch, and provide a basis for understanding preferred modeling approaches for rare bycatch events.