Modelling community structure and species co-occurrence using fishery observer data

Abstract

Abstract In this study, we modelled fishery observer data to compare methods of identifying community structure using cluster analyses to determine stratifications and probabilistic models for examining species co-occurrence in the Gulf of Mexico deepwater reef fish fishery. Comparing cluster analysis methods, the correlation measure of dissimilarity in combination with average agglomerative linkage was the most efficient method for determining species relationships using simulated random species as a comparison tool. Cluster analysis revealed distinct species stratifications and in combination with multiscale bootstrapping generated probabilities indicating the strength of stratifications in the fishery. A more parsimonious approach with probabilistic models was also developed to quantify pairwise species co-occurrence as random, positive, or negative based on the observed vs. expected fishing sets with co-occurrence. For the most common species captured, the probabilistic models predicted positive or negative co-occurrence between 84.2% of the pairwise combinations examined. These methods provide fishery managers tools for determining multispecies quota allocations and offer insights into other bycatch species of interest.