Bayesian Spatiotemporal Models for Counts of Tropospheric Ozone Daily Exceedances

In Northern Italy, air quality has recently become a major concern. The Lombardy region is classified as one of the most polluted areas in Europe due to a combination of geographical factors, high population density, and industrial activities. In this paper, we focus on ground-level ozone (O3), which is subject to regulation by the European Union through air quality thresholds that must be complied with. In Lombardy, ozone levels frequently exceed these thresholds, raising concerns about the associated health risks and environmental impacts, and making it crucial to control the number of days when these levels are exceeded. Since ozone air pollution becomes concerning only in the warmest period of the year, we consider ozone exceedances only from April to October from 2000 to 2023. We focus on a bivariate response, the number of days in each month when at least a prolonged and immediate ozone exceedance has occurred, respectively. To study the trend in the number of days with ozone exceedances over the years, we fit a bivariate mixed-e!ects model for the responses, using a binomial likelihood for each. We incorporate territorial and meteorological factors as covariates associated to fixed e!ects parameters, in addition to latent spatio-temporal random e!ects. In particular, we model the year and the month associated to each count data as random grouping factors. Our analysis shows that the probability of exceeding the long-term ozone exposure threshold at least once a day has remained constant over the years. On the other hand, there has been a steady decline over the years in the probability of occurrences of extreme ozone levels.