Stewart, D.R. and J.M. Long. 2016. Using hierarchical Bayesian multi-species mixture models to estimate tandem hoop-net based habitat associations and detection probabilities of fishes in reservoirs. Transactions of the American Fisheries Society 145:450-461. DOI: 10.1080/00028487.2016.1143395
Species distribution models are useful tools to evaluate species-habitat relationships, though more often developed to evaluate factors controlling fish assemblages in lotic environment and less often in lentic systems. We evaluated species-habitat interactions for commonly captured fishes in six Oklahoma reservoirs using tandem hoop nets. We used hierarchical Bayesian multi-species N-mixture models to evaluate detection- and abundance-habitat relationships. Twelve species were captured, equating to 7,212 fish with channel catfish (46%), bluegill (25%), and white crappie (14%) comprising the majority of the catch. Detection estimates ranged from 8% to 69%, and modeling results suggested that fishes were primarily influenced by reservoir size and context, water clarity and temperature, and land-use types. Species were differentially abundant within and among habitat types, consisting of some fishes being found more abundant in turbid, less impacted (urbanization and agriculture) reservoirs with greater shoreline lengths; whereas, other species were found more often in clear, nutrient rich impoundments with generally less shoreline length and increased percentage agriculture. Our results demonstrate that habitat and reservoir characteristics may differentially benefit species and assemblage structure. This study provides a useful framework for evaluating capture efficiency for not only hoop nets but other gears used to sample fishes in reservoirs.