Assessing drivers of estuarine debris using a Bayesian spatial modelling approach (INLA-SPDE)

Abstract

Estuaries are under increasing pressure from urbanization, including through the release of domestic wastes such as anthropogenic debris. Growing concern over environmental contamination has led to management interventions, aimed at reducing the further release of debris. Effective strategies, however, require an understanding of the factors influencing the abundance and distribution of debris. Estuaries pose a challenge to assess debris trends due to the myriad of factors that influence debris patterns. Here we present a Bayesian approach to assess spatial patterns in debris across four estuaries of New South Wales, Australia. We provide an overview of debris composition and employ Integrated Nested Laplace Approximation (INLA) alongside Stochastic Partial Differential Equation (SPDE) to produce three models testing predictors of debris abundance: (i) non spatial, (ii) a model accounting for spatial autocorrelation and (iii) accounting for spatial autocorrelation and introducing the estuarine foreshore as a barrier. Data was sourced from cleanups reported within the Australian Marine Debris Initiative (AMDI), a national citizen science program coordinated by the Tangaroa Blue Foundation. The data, from 169 sampling events between 2012 and 2017 was used. All models tested predictors against the abundance of hard plastic, expanded plastic, plastic bags, and plastic drink bottles. Plastic was the dominant component, ranging from 69% to 82% of total debris within the study estuaries. Important predictors differed across debris types and included ‘distance to estuary mouth’, access to foreshore, population, and site substrate. Specific factors differed in importance when accounting for spatial autocorrelation, demonstrating the need for spatial approaches when assessing debris trends within estuaries. The present Bayesian approach could be used to better understand drivers of estuarine debris and inform interventions for effective management.