Multispecies methods to facilitate the transition from heuristics to statistics in ecosystem-based fisheries management

Abstract

Since 1985, fisheries management has undergone vast changes. Requests for new policies that evaluate impacts beyond individual removals of fish have outpaced the scientific tools available to implement them. Tools that were or became available were often implemented without rigorous testing. Here, simulations were used to assess the ability of two multivariate time-series models to provide information relevant to ecosystem-based fisheries management. Simulations highlighted the need to account for time-series properties of data. Multivariate autoregressive state-space models successfully estimated interactions between two time series when observation error was small and the length of the time series was sufficient. Prewhitening procedures were also successful in estimating interactions. The framework proved useful for estimating synchrony between indicators of ecosystem status and the true state of the ecosystem. Spatiotemporal multivariate models successfully recovered trends in the data. Models were robust to model misspecification and estimates of covariates related to linear relationships between habitat and relative abundance were unbiased. In contrast, squared terms were biased, particularly for the catch-rate component of the model. Incorrectly including a habitat covariate when it did not govern the data-generating process was less problematic than not including the covariate when it should have been included. Results from simulations guided the selection of quantitative tools used to provide inference on management uncertainty related to gear switching in the US West Coast Groundfish LE trawl fishery. Fishers in this fishery were allowed to switch from using trawl-gear to fixed-gear to land sablefish in 2011, and it was expected that bycatch of species with restrictive quotas would influence their decision to switch. However, bycatch species were not a major factor in their decision. Instead, unmodeled factors related to the major port groups included in the analysis were the best predictor of the proportion of landings that were caught using fixed-gear compared to trawl-gear. It was hypothesized that social and logistical challenges specific to each port group could limit their ability or desire to switch gear. Additionally, vessels could be limited to using trawl-gear because ports may not have adequate facilities to process catch from fixed-gear.