Quantification of multifactorial effects on particle distributions at urban neighborhood scale using machine learning and unmanned aerial vehicle measurement

Abstract

This study contributes to quantifying the coupling impacts of multiple factors including traffic, meteorology, and built environment on particle (PM) distributions at urban neighborhood scales. Increasing knowledge of the PM distribution characteristics and the quantification of multifactorial effects could assist in alleviating pollution. However, the contribution of multiple factors and their coupling effects to PM concentrations have not been quantified adequately. Meanwhile, the vertical distribution patterns above 10 m in urban near-road areas, which are closely related to the residents’ exposure in high-rise buildings, are rarely revealed. To bridge the research gaps, this study developed an integrated method of unmanned aerial vehicle measurement and Explainable Boosting Machine to reveal the vertical PM distributions and quantify the coupling effects of traffic, meteorology, and built environments. Results show that the PM concentrations at 9–18 m are significantly higher than those at ground level. Moreover, background concentration (20.9%), traffic volume (15.6%), and street morphology (11.4%) are the top three important features. The coupling effects of selected features, especially the interplay between wind factors and monitoring position, account for 8.9% of PM concentrations. This study provides insights into source identification of particles and improving livability and sustainability in urban residential areas.