Associations of PM2.5 and black carbon concentrations with traffic, idling, background pollution, and meteorology...

An air quality study was performed outside a cluster of schools in the East Harlem neighborhood of New York City. PM(2.5) and black carbon concentrations were monitored using real-time equipment with a one-minute averaging interval. Monitoring was performed at 1:45-3:30 PM during school days over the period October 31-November 17, 2006. The designated time period was chosen to capture vehicle emissions during end-of-day dismissals from the schools. During the monitoring period, minute-by-minute volume counts of idling and passing school buses, diesel trucks, and automobiles were obtained. These data were transcribed into time series of number of diesel vehicles idling, number of gasoline automobiles idling, number of diesel vehicles passing, and number of automobiles passing along the block adjacent to the school cluster. Multivariate regression models of the log-transform of PM(2.5) and black carbon (BC) concentrations in the East Harlem street canyon were developed using the observation data and data from the New York State Department of Environmental Conservation on meteorology and background PM(2.5). Analysis of variance was used to test the contribution of each covariate to variability in the log-transformed concentrations as a means to judge the relative contribution of each covariate. The models demonstrated that variability in background PM(2.5) contributes 80.9% of the variability in log[PM(2.5)] and 81.5% of the variability in log[BC]. Local traffic sources were demonstrated to contribute 5.8% of the variability in log[BC] and only 0.43% of the variability in log[PM(2.5)]. Diesel idling and passing were both significant contributors to variability in log[BC], while diesel passing was a significant contributor to log[PM(2.5)]. Automobile idling and passing did not contribute significant levels of variability to either concentration. The remainder of variability in each model was explained by temperature, along-canyon wind, and cross-canyon wind, which were all significant in the models.