In this paper we propose a hierarchical spatio-temporal model for daily mean concentrations of PM10 measured in 11 monitoring sites located in the main cities of the Emilia-Romagna Region. Main aims of the proposed model are: the identification of the sources of variability characterising the PM10 process, the imputation of missing observations in order to obtain time series free from missingness that can be used in ecological regression studies, the estimation of pollution levels in unmonitored spatial locations. The modelling approach is fully Bayesian, the implementation has been performed via Monte Carlo Markov Chain algorithms. The model has been carefully checked using Bayesian p-values and graphical posterior predictive checks.