Source apportionment of urban road dust using four multivariate receptor models

Road dust is one of the biggest contributors to airborne particulate matter (PM) in many urban regions. Due to the inherent heterogeneity of road dust, it is important that its sources are identified and mitigated. Multivariate receptor models are used for source apportionment of PM in many cities. In recent years, these receptor models are finding more applications outside the scope of PM source apportionment. In this study, four multivariate receptor models (Unmix, Positive Matrix Factorization, Principal Component Analysis, and Multiple Curve Regression) are used for source apportionment of road dust at Vellore City, India. The elemental composition of road dust samples from 18 locations and for three seasons (summer, winter, and monsoon) are measured using acid digestion followed by Inductively Coupled Plasma–Optical Emission Spectroscopy. Irrespective of models, results showed that crustal material (100–68%) and resuspended road dust (82–15%) are the biggest contributors to road dust in the study region. Brake wear, tire wear, biomass combustion, vehicular emission, and industrial sources are some of the other sources identified by the receptor models. Receptor modeling performance of MCR and PCA models are unsatisfactory. PMF and Unmix models gave acceptable results. From comparing the performance characteristics, Unmix is found to be the ideal receptor model for this dataset. This research clarifies the constraints of different receptor models and the source apportionment information obtained is critical for development of future policy and regulation.