Hierarchical space-time modelling of PM 10 pollution

Abstract

In this paper, we propose a hierarchical spatio-temporal model for daily mean concentrations of PM 10 pollution. The main aims of the proposed model are the identification of the sources of variability characterising the PM 10 process and the estimation of pollution levels at unmonitored spatial locations. We adopt a fully Bayesian approach, using Monte Carlo Markov Chain algorithms. We apply the model on PM 10 data measured at 11 monitoring sites located in the major towns and cities of Italy's Emilia-Romagna Region. The model is designed for areas with PM 10 measurements available; the case of PM 10 level estimation from emissions data is not handled. The model has been carefully checked using Bayesian p-values and graphical posterior predictive checks. Results show that the temporal random effect is the most important when explaining PM 10 levels.