Bayesian Spatiotemporal Modeling of Urban Air Pollution Dynamics

This work deals with the spatiotemporal analysis of urban air pollution dynamics in the town of Perugia (Central Italy) using high-frequency and size resolved data on particular matter (PM). Such data are collected by an Optical Particle Counter (OPC) located on a cabin of the Minimetro, a public transport system that moves on a monorail on a line transect of the town. Hierarchical Bayesian models are used that allow to model a quite large dataset and include an autoregressive term in time, in addition to spatially correlated random effects. Models are fitted for three response variables (fine and coarse particle counts, nitric oxide concentration) and using covariate information such as temperature and humidity. Results show a large temporal autocorrelation, relatively larger for particle counts; moreover, all variables show a significant spatial correlation, with larger ranges for fine PM rather than for coarse PM and nitric oxide concentration.