Agricultural burning smoke in Eastern Washington: Part II. Exposure assessment

Several studies have documented potential health effects due to agricultural burning smoke. However, there is a paucity of literature characterizing community residents’ exposure to agricultural burning smoke. This study assesses personal exposures to particulate matter (PM) with aerodynamic diameters <2.5 μm (PM2.5) from agriculture burning smoke (Eb) for 33 asthmatic adults in Pullman, WA. PM2.5 concentrations were measured on 16 subjects, inside of all but four residences, outside of 6 residences, and at a central site. The mean±standard deviation of personal exposure to PM2.5 was 13.8±11.1 μg m−3, which was on average 8.0 μg m−3 higher during the agricultural burning episodes (19.0±11.8 μg m−3) than non-episodes (11.0±9.7 μg m−3). The levoglucosan (LG, a unique marker for biomass burning PM) on personal filter samples also was higher during the episodes than non-episodes (0.026±0.030 vs. 0.010±0.012 μg m−3). We applied the random component superposition model on central-site and home indoor PM measurements, and estimated a central-site infiltration factor between 0.21 and 2.05 for residences with good modeling performance. We combined the source apportionment and total exposure modeling results to estimate individual Eb, which ranged from 1.2 to 6.7 μg m−3 and correlated with personal LG with an r of 0.51. The sensitivity analysis of applying the infiltration efficiency estimated from the recursive model showed that the Eb (range: 1.3–4.3 μg m−3) obtained from this approach have a higher correlation with personal LG (r=0.75). Nevertheless, the small sample size of personal LG measurements prevents a comparative and conclusive assessment of the model performance. We found a significant between-subject variation between episodes and non-episodes in both the Eb estimates and subjects’ activity patterns. This suggests that the LG measurements at the central site may not always represent individual exposures to agricultural burning smoke. We recommend collecting more microenvironmental samples to model the Eb and more personal samples to validate the Eb estimates.