Bayesian mixed latent factor model for multi-response marine litter data with multi-source auxiliary information

This work aims at analyzing the composition and spatio/temporal distribution of marine litter amounts found at the sea-floor in a region of the central Mediterranean. Inspired by common multi-species distribution modeling problems, we propose a suitable Bayesian multivariate approach to model litter data inferring possible environmental covariates while controlling for correlation between different litter categories and providing a method for residual ordination. A combined environmental information coming from multiple sources at different spatio/temporal scales is considered to investigate environmental drivers that might affect dynamics of marine litter assemblages at local scale.