Modeling Community Structure and Abundance Using Observer Data for the U.S. Gulf of Mexico Deepwater Reef Fishery

Abstract

Fishery observer data collected in the Gulf of Mexico deepwater reef fish fishery from July 2006 through December 2013 were examined for community structure using hierarchical cluster analyses to quantify species relationships and reveal stratifications in the fishery. The correlation measure of dissimilarity with average agglomerative linkage was the most efficient method using randomly fake species as a comparison tool between dissimilarity and linkage choices. This approach in combination with a multiscale bootstrapping revealed distinct stratifications and probabilities indicating the strength of species relationships in the fishery. For deepwater species managed under the individual fishing quota (IFQ) system, cluster analyses findings detected patterns in species co-occurrence on fishing sets that may be of interest to managers. Additionally, delta-lognormal boosted regression tree and zero-inflated negative binomial predictive models were compared for standardizing spatial abundance for the fishery. Deltalognormal boosted regression tree models were superior in representing fine-scale variations, however, zero-inflated negative binomial models were more representative in abundance observed on a larger spatial scale. An examination of the deepwater IFQmanaged species also found evidence for size selection of discards and differences in retention rates for some species managed under the same allocation category.