Machine learning exploring the chemical compositions characteristics and sources of PM2.5 from reduced on-road activity

Abstract

Particulate nitrate pollution has emerged as a major contributor to haze events in urban environment, due to the rapid increase of vehicle emissions. However, a comprehensive formation mechanisms of PM2.5 responses to vehicle emissions control still remains high uncertainties. In our study, hourly criteria air pollutants, meteorological parameters and chemical compositions of PM2.5 were continuously measured with or without reduced on-road activity at the coastal city in southeast China. XG Boost-SHAP models analysis showed that increasing concentrations of NO3−, NH4+, and BC contribute to elevated PM2.5 levels, due to the influence of vehicle emissions. Based on PMF model results, there was a notable increase in the contributions of traffic-related emissions, industrial activities, and dust sources to PM2.5, with increments of 13%, 4%, and 7%, respectively. In addition, metal elements such as Mn emerged as the primary contributor to hazard quotient (HQ) values, originated from non-exhaust emissions of vehicles, which might cause the potential toxic risks on human health, particularly during haze events. Hence, this study improve the understanding of air quality and human health both direct and indirect responses to vehicle emissions control in future urban management.