Harvest models and stock co-occurrence: probabilistic methods for estimating bycatch

Abstract

A primary goal of ecosystem-based fishery management is to reduce non-target stock impacts, such as incidental harvest, during targeted fisheries. Quantifying incidental harvest has generally incorporated fishery-dependent catch data, yet such data may be biased by gear non-retention, observation difficulties, and non-random harvest patterns that collectively lead to an impartial understanding of non-target stock capture. To account for such issues and explicitly recognize the combined influence of ecological and harvest factors contributing to incidental capture within targeted fisheries, we demonstrate a probabilistic modelling framework that incorporates: (i) background rates of target and non-target stock co-occurrence as the primary ecological basis for incidental harvest; (ii) the probability of harvesting at localities exhibiting co-occurrences; (iii) the probability of selecting for non-target species with fishery gear; and, (iv) as a function of harvest effort, the overall probability of incidental capture for any non-target stock contained in the species pool available for harvest. To illustrate application of the framework, simulation models were based on fishery-independent data from a freshwater fishery in Ontario, Canada. Harvest simulations of empirical stock data indicated that greatest species-specific capture values were over 4000 times more likely than for species with lowest values, indicating highly variable capture probabilities because of the combined influence of stock heterogeneity and harvest dynamics. Estimated bycatch–effort relationships will allow forecasting incidental harvest on the basis of effort to evaluate future shifts in fishing activity against specific ecosystem-based fishery management objectives, such as reducing the overall probability of bycatch while maintaining target landings.