Computationally Efficient Clustering of PM10 Time Series Data

Air pollution is a critical global health concern, responsible for millions of premature deaths each year. Particulate matter (PM), a major pollutant, comprises airborne particles from both natural and anthropogenic sources. In particular, fine particles such as PM10 pose significant health risks. Hence, continuous monitoring of PM10 levels is essential for mitigating hazardous exposure. This study employs a Bayesian spatial product partition model to analyze geo-referenced PM10 data. An efficient Markov Chain Monte Carlo algorithm is implemented to make posterior inference about the clustering of PM10 monitoring stations. To speed up the computation, we exploit the fact that the precision matrices in the proposed model are tridiagonal. The method is illustrated by analyzing daily PM10 levels collected in 2018 over Austria.