A hierarchical bayesian model for modelling benthic macroinvertebrates densities in lagoons

The ecological status classification of aquatic ecosystems requires separate quantification of natural and anthropogenic sources of environmental variability. A clustering of ecosystems into ecosystem types (i.e. Typology) is used in order to minimise natural variability. Among transitional water quality elements, benthicmacroinvertebrates are the most exposed to natural variability patterns due to their life cycles and space-use behavior. Here, we address the ecological status classification issue for Mediterranean and Black Sea lagoons, using benthic macroinvertebrates, from a set of 14 reference lagoons. Two main classification approaches have been proposed in literature: the a-priori approach and the a-posteriori approach. The a-priori approach classifies the ecological status of the lagoons according to standard classification boundaries of some multimetric indices applied at the level of a priori defined ecosystem types. On the other hand, in the a-posteriori approach linear mixed effect models are used in order to study the relationships of multimetric indices with some lagoon characteristics and to define a posteriori typologies and new reference values for the classification of lagoons based on the indices. However, it may happen that different indices are in disagreement with respect to lagoon classification. We propose a Bayesian hierarchical model in which the multimetric indices are jointly modeled through a multivariate normal mixture model. Each mixture component is estimated as function of covariates of interest and corresponds to an ecological status. We compare the proposed model with the a-priori and a-posteriori approaches highlighting pros and cons of each method.