An INLA spatio-temporal model for zero-inflated marine plastic litter abundance

The marine plastic litter pollution is a worldwide growing environmental concern. Despite its negative effects on marine ecosystems, the phenomenon is still not well-known at global and local scale. This work aims at assessing the spatio-temporal distribution of plastic litter amounts found at the sea-floor in a region on the central Mediterranean (Ionian sea). Inspired by species distribution models, we propose a two-parts model to accommodate the excess of zeros and the spatio-temporal correlation characterizing abundance monitoring data. A common spatial effect that links the plastic abundances and the probabilities of occurrences is impremented with the Stochastic Partial Differential Equation approach extended to a non -stationary barrier model. The INLA methodology allows to efficiently perform Bayesian inference to fit complex spatio-temporal models including effects of environmental covariates and enables to investigate the assemblages of plastic litter over the study region.