Fauna characterization of a cold-water coral community network along the Apulian coasts by Bayesian mixed models

The exploration of a cold-water coral (CWC) community network connecting the Southern Adriatic fish populations with those of the Northern Ionian Sea has many challenging implications involving biodiversity conservation and fisheries management. To characterize the species assemblages of the CWC community network, we analyze the size of six mostly abundant species to highlight the main differences among five CWC areas along the Apulian coasts. Data are surveyed by experimental longlines casted in all CWC areas between 2013 and 2014. Bayesian Generalized Additive Mixed Models (GAMMs) are applied to analyze the variation of the fishes length in the five CWC areas according to species and covariates such as depth and abundance. GAMMs allow to account for various effects of variability components on the overall length of fishes. Appropriate smooth functions are available to describe the physical effect of each covariate and specific random area effects are allowed to induce correlation among individuals of all species captured in the same area. Parameter estimation is carried out by maximization of the joint posterior probability distribution, where marginal posterior probabilities associated to specific model terms are obtained by a spike-and-slab prior structure, that can be viewed as a scale mixture of Gaussians. This approach is implemented in the R package spike SlabGAM, able to deal with most common distributional assumptions and allowing efficient variable selection and model choice. Results show that the size of fishes is collectively affected by random effects of the CWC areas and by smooth effects of their depth and abundance.