|
Abstract:
|
Coral reefs provide an important ecosystem for life underwater. We developed an approach to modeling the benthic coral reef dynamics for the data of the Abrolhos bank. The Abrolhos reefs, off the coast of Southern Bahia, Brazil, are the largest and the richest reef complexes of the Southwestern Atlantic. The reef structures are under severe stress from climate and anthropogenic stressors and the unequivocal contamination. Reef structures are built by categories such as corals, macroalgae, turf, bry-ozoans, sponge, fire coral, cyano-bacteries, and others. Each one of these categories decribe the benthic community and can be expressed as proportions of a whole. To understand this benthic cover composition over different reef locations, we extended the Dirichlet regression model including hierarchical effects by sites. The inference procedure was done under the Bayesian approach using Hamiltonian Monte Carlo (HMC) method to obtain the approximations to the posterior marginal distributions of interest. Decision theory supported choosing one component as a reference to avoid model identifiability issues. The main results exhibit skewness and heteroscedasticity for specific components.
|
