A comparison of approaches to estimate traffic-related PM2.5 in a study of commuters

Abstract

Background/Aim: Associations between exposure to traffic-related fine particulate matter (tr-PM2.5) and health are not fully understood. Exposure to tr-PM2.5 while commuting may reflect an important exposure environment relevant for health. However, estimating exposure to tr-PM2.5 is challenging because tr-PM2.5 is a spatially heterogeneous chemical mixture that cannot generally be directly observed. We compare two approaches to estimate personal exposure to tr-PM2.5 among commuters.Methods: We measured personal exposure to continuous PM2.5 and 48-hour integrated PM2.5 metals and black carbon (BC) for 49 women commuters in the DC metro area. We recorded vehicle trips across 48 hours using vehicle data loggers and dashboard cameras. We estimated tr-PM2.5 using two approaches: source apportionment and vehicle monitoring. The source apportionment approach applied Positive Matrix Factorization (PMF) to integrated PM2.5 constituents as in previous studies. The vehicle monitoring approach estimated average PM2.5 exposure during vehicle trips identified using vehicle monitors. We compared estimates of tr-PM2.5 using Spearman correlations and comparing tertile groups. We also compared vehicle monitoring tr-PM2.5 with two tracers of tr-PM2.5: 48-hour integrated BC and zinc.Results: We fitted PMF to the integrated data and identified three sources: tailpipe emissions, salt, and road dust. Vehicle monitoring estimated personal tr-PM2.5 exposure with a mean of 6.7ug/m3 (interquartile range: 2.5,7.7). The correlation between tailpipe emissions and vehicle monitoring was 0.22, with correlations for the other sources close to zero. The two approaches generally grouped commuters differently, with 62% of commuters in different tertiles for tailpipe emissions and vehicle monitoring. Results were similar when BC and zinc were used instead of estimated sources.Conclusions: There were substantial differences between the two approaches for estimating tr-PM2.5 among commuters, indicating that using only integrated PM2.5 constituents may not be representative of commute-specific exposures. Future work will combine these two approaches to better estimate exposure to tr-PM2.5.