PMF-Bayesian Modeling for Spatial Source Apportionment of Airborne Particulate Matter

Mass concentrations of ambient particulate matter (PM) have been extensively monitored in urban areas worldwide. Despite the widespread availability of such data, it has rarely been utilized in receptor-based source apportionment studies, which predominantly rely on PM chemical speciation data. In this study, we used over one million data points of PM concentrations from more than 100 monitoring sites within a Chinese megacity to perform spatial source apportionment of coarse particles (PM_2.5-10). These particles are believed to primarily originate from local emissions and are often characterized by significant spatial heterogeneity. We employed an enhanced positive matrix factorization (PMF) approach, designed to handle large datasets, in combination with a Bayesian multivariate receptor model to determine spatial source impacts. Four primary sources were successfully identified: residential burning, industrial processes, road dust, and meteorologically-related sources. The combined methodology demonstrates substantial potential for broader application in other regions.