Application of multivariate statistical analysis using organic compounds: Source identification at a local scale (Napajedla, Czechia)

Abstract

The study aimed to apply novel source classification tool for local scale air pollution assessment reducing the total number of organic compounds in the model. Samples of particulate matter (PM) were collected in the town of Napajedla (South-eastern Czech Republic) in 2016. The industrial sector of the town is represented by plastics processing and manufacturing, as well as by mechanical engineering. Analytical technique of pyrolysis chromatography with mass spectroscopy detection was employed to identify organic species in the PM 10 fraction. Two datasets (465 determined organic compounds and 50 selected organic markers) were used and compared by multivariate analysis - principal component analysis followed with hierarchical clustering on principal components incorporating compositional data approach. Three resulting clusters were observed in both cases. The cluster representing measurements near plastic processing and manufacturing plants was identical in both the analysed datasets with the same organic compounds that characterized resulting cluster Consequently, leading markers for plastic processing and manufacturing sources were suggested (bumetrizole, bis(tridecyl)phthalate, mono(2-ethylhexyl)phthalate). Other two clusters varied among the analysed datasets, however, dataset with selected markers showed more reliable outcomes. The results imply that concept of using only selected organic marker species with the compositional approach in multivariate statistical methods is sufficient and allows properly distinguishing the main air pollution sources between sampling locations even at a small urban scale.