Commuter types identified using clustering and their associations with source-specific PM2.5

Abstract

Traffic-related fine particulate matter air pollution (tr-PM2.5) has been associated with adverse health outcomes such as cardiopulmonary morbidity and mortality, with in-vehicle tr-PM2.5 exposure contributing to total personal pollution exposure. Trip characteristics, including time of day, day of the week, and traffic congestion, are associated with in-vehicle PM2.5 exposures. We hypothesized that some commuter characteristics, such as whether commuters travel primarily during rush hour, would also be associated with increased tr-PM2.5 exposures. The commute data consisted of unscripted personal vehicle trips of 46 commuters in the Washington, D.C. metro area over 48-h, with a total of 320 trips. We identified commuter types using sparse K-means clustering, which identifies the hours throughout the day important for clustering commuters. Source-specific PM2.5 over 48 h was estimated using Positive Matrix Factorization. Linear regression was used to estimate differences in source-specific PM2.5 by commuter cluster. Two commuter clusters were identified using the clustering approach: rush hour commuters, who primarily travelled during rush hour, and sporadic commuters, who travelled throughout the day. The hours given the largest weights by sparse K-means were 7–8 a.m. and 6–7 p.m., corresponding to peak travel times. Integrated black carbon (BC) was higher for rush hour commuters (median = 3.1 μg/m3 (IQR = 1.5)) compared to sporadic commuters (2.0 μg/m3 (IQR = 1.9)). Mobile PM2.5, consisting primarily of tailpipe emissions and brake/tire wear, was also higher for rush hour commuters (2.9 μg/m3 (IQR = 1.6)) compared to sporadic commuters (2.1 μg/m3 (IQR = 2.4)), though this difference was not statistically significant in regression models. Estimated differences between commuter types for secondary/mixed PM2.5 and road salt PM2.5 were smaller. Further research may elucidate whether commuter characteristics are an efficient way to identify individuals with highest tr-PM2.5 exposures associated with commuting and to develop effective mitigation strategies.