Abstract:
The representativeness of outdoor, or ambient, particulate matter (PM) concentrations, often used as exposure surrogates in epidemiological studies, has been the subject of scientific scrutiny. This dissertation examined the relationship between ambient concentrations and personal exposures based on the Seattle panel study of susceptible individuals and modeled exposure to ambient and nonambient PM. The goal was to quantify the ambient contribution to personal exposure and determine factors that influence contribution.
Residential and personal PM concentrations were measured using gravimetric and light scattering techniques. The light scattering data from 62 residences were screened for indoor sources and the infiltration efficiency (Finf) of each residence was estimated with a series of sensitivity analyses. Finf estimates and time-location data were then used to estimate each subject's ambient contribution fraction (α)
On average, ambient particles contributed 77% of the indoor concentration. As a result, although subjects spent about 90% of their time indoors, ambient PM2.5 accounted for about half of their total exposures, on average. Sensitivity analyses demonstrated that the Finf estimates were stable and agreed with the sulfur tracer estimates, but that agreement depended on the approach used to estimate Finf from the recursive model. Unlike Finf, estimates of particle penetration, air exchange rate, and particle deposition were unstable. Finf depended on residence type and ventilation status and could be predicted using meteorological and housing data.
The ambient-personal correlations depended on time spent outdoors, Finf, and the nonambient contribution, and the latter seemed to dominate. Subjects with minimal nonambient sources had good correlations, which verified the representativeness of ambient concentrations for population exposure to ambient-generated PM. This dissertation also demonstrated the wide range of Finf, which resulted in varying ambient contributions to exposure among individuals. Future epidemiological studies could account for such variations by estimating Finf for individual residences, perhaps using a prediction model similar to the prototype that I developed for Seattle. The techniques developed in this dissertation could be applied to the assessment of ambient and nonambient PM exposure in epidemiological studies and for regulatory modeling.