A novel approach for characterizing neighborhood-level trends in particulate matter using concentration and size fraction distributions: a case study in Charleston, SC

Abstract

In this study, we aim to illustrate how novel technologies and methodologies can be used to enhance neighborhood level studies of ambient particulate matter (PM). This is achieved by characterizing temporal and spatial features of PM levels and by assessing patterns in particle size composition using simultaneous measures across multiple size fraction ranges in Charleston, SC, USA. The study is conducted in three stages: (1) we monitor real-time PM concentrations for the following: PM ≤ 15 μm, PM ≤ 10 μm, PM ≤ 4 μm, PM ≤ 2.5 μm, and PM ≤ 1 μm at five locations during February–July, 2016; (2) we apply a generalized additive model (GAM) to assess temporal and spatial trends in PM2.5 after controlling for meteorology, instrument, and temporal confounders; and (3) we employ a self-organizing map (SOM) to identify hourly profiles that characterize the types of size fraction distribution compositions measured at our sites. Monitoring results found that average PM2.5 levels during our ‘snapshot’ monitoring were 6–8 μg/m3 at our sites, with 95th percentiles ranging from 9 to 13 μg/m3. GAM results identified that temporal peaks for PM2.5 occurred during the early morning hours (6–8 am) across all sites and that the marginal means for four of our inland sites were significantly different (higher) than a waterfront site. SOM results identified six hourly profiles, ranging from hours when all size fractions were relatively low, to hours dominated by single size fractions (e.g., PM1), and to hours when multiple size fractions were relatively high (e.g., PM15–10 and PM10-PM2.5). Frequency and duration distributions show variability in the occurrence and persistence of each hourly type. Collectively, our findings reveal the complexity of PM behavior across a relatively small geographic region and illustrate the potential usefulness of using size fraction composition to better understand air quality. However, it is important to note that this study only presents a snapshot of air quality and that longer monitoring periods are recommended for more definitive characterizations.