Air pollution is a major global environmental health risk, with fine particulate matter (PM10) strongly linked to severe respiratory and cardiovascular conditions. Analyzing and clustering spatio-temporal air quality data is therefore essential for understanding pollution dynamics and informing effective policy interventions.
This talk provides an overview of Bayesian nonparametric clustering methods, with particular emphasis on their application to spatio-temporal data commonly encountered in environmental sciences. Key modeling frameworks for point-referenced spatio-temporal data are presented, highlighting their flexibility in capturing complex spatial and temporal dependence structures. Recent advances in Bayesian clustering are then reviewed, focusing on spatial product partition models that explicitly incorporate spatial structure into the clustering mechanism.
The methodology is illustrated through an application to PM10 monitoring data from Northern Italy, where the approach identifies meaningful spatial and temporal patterns in pollution levels. The findings underscore the potential of Bayesian nonparametric methods for environmental risk assessment and point to promising directions for future research in spatio-temporal clustering with relevance to public health and environmental policy.