Jointly modelling air quality and meteorological variables using the D-STEM software

This work discusses a general framework for mapping pollutant concentrations over a region of interest considering both concentration and meteorological data. The framework is based on a multivariate space-time model able to handle missing data and non-colocated monitoring networks. The model is implemented and estimated by the D-STEM (Distributed Space Time Expectation Maximization) software. As a case study, the model is used to produce daily maps of particulate matters and nitrogen dioxide concentrations for the Italian region Apulia. The official data from the air quality monitoring network of Apulia and the meteorological data from the Meteorological Service of Apulia are used, for a total of 8 variables and thus a 8-variate model. In order to deal with the computational complexity, the Bari INFN high performance grid computing infrastructure is used.