Fraction of PM2.5 personal exposure attributable to urban traffic: A modeling approach

Personal exposure to fine particles (PM2.5) in a nonsmoking adult populatio n has been characterized in Grenoble, France, in the framework of the Europ ean EXPOLIS study The objective of this paper is to assess the fraction of the PM2.5 personal exposure attributable to urban traffic emissions. Volunt eers (n = 40) carried a personal exposure monitoring case and filled in que stionnaires on their outdoor and indoor environments, as well as time-activ ity diaries (15 min resolution), during 48 h (working days). Workplaces and places of residence were classified in two categories using a Geographic I nformation System (GIS): the atmospheric environment of some volunteers is best represented by PM ambient air monitors located in urban background sir es, and others by monitors situated close to high traffic density sites (pr oximity sites). A partial least-squares regression model estimated the PM2. 5 personal exposure (average = 36.6 mu g/m(3); standard deviation = 23.4 mu g/m(3)) as a function of time spent in proximity (at work, home, or commut ing), PM10 ambient air levels during the same days, and several confounders (passive smoking and indoor sources of particles). Six scenarios of "proxi mity" and "background" environments were accommodated, according to traffic intensity and road distance, in a sensitivity analysis; the best fitted mo del had R-2 =.7. Personal PM2.5 exposures predicted by this model for diffe rent segments of the study population were compared to the background perso nal exposure, thus providing an estimate of the additional contribution of time spent near traffic sources. On average (percent time spent in proximit y = 16.3; proximity scenario defined as the area located less than 50 m fro m a street with a traffic intensity greater than 20,000 vehicles/day), the PM2.5 personal exposure attributable to traffic equals 30%. For the lower t ercile of the population, this contribution is 26%, for the upper tercile, it is 45%. A very influential parameter of this modeling estimation is the proportion of background ambient air particulate concentrations associated with traffic emissions Based on local nighttime/daytime concentration ratio s, a 20% proportion has bi en derived and used for these results. In the li terature, this parameter ranges from 10% to 60%, yielding a proportion of p ersonal exposure attributable to traffic proximity between 20% and 60%, and high-exposure situations reaching 60 to 75%. While these estimates are bas ed only on winter data, they are in agreement with other results published in the literature. This modeling approach might be applied to other metropo litan situations, insofar as local data are used to assess the influence of traffic emissions on background ambient air PM10 concentrations.