Environ. Sci. Technol. 2009, 43, 4737–4743
Determinants of Personal Exposure to PM2.5, Ultrafine Particle Counts, and CO in a Transport Microenvironment S. KAUR† AND M . J . N I E U W E N H U I J S E N * ,†,‡ Center for Research in Environmental Epidemiology (CREAL), IMIM, CIBERESP, Parc de Recerca Biome`dica de Barcelona PRBB, Barcelona, Spain, and Imperial College London
Received November 12, 2008. Revised manuscript received February 7, 2009. Accepted February 27, 2009.
Short-term human exposure concentrations to PM2.5, ultrafine particle counts (particle range: 0.02-1 µm), and carbon monoxide (CO) were investigated at and around a street canyon intersection in Central London, UK. During a four week field campaign, groups of four volunteers collected samples at three timings (morning, lunch, and afternoon), along two different routes (a heavily trafficked route and a backstreet route) via five modes of transport (walking, cycling, bus, car, and taxi). This was followed by an investigation into the determinants of exposure using a regression technique which incorporated the sitespecific traffic counts, meteorological variables (wind speed and temperature) and the mode of transport used. The analyses explained 9, 62, and 43% of the variability observed in the exposure concentrations to PM2.5, ultrafine particle counts, and CO in this study, respectively. The mode of transport was a statistically significant determinant of personal exposure to PM2.5, ultrafine particle counts, and CO, and for PM2.5 and ultrafine particle counts it was the most important determinant. Traffic countexplainedlittleofthevariabilityinthePM2.5 concentrations, but it had a greater influence on ultrafine particle count and CO concentrations. The analyses showed that temperature had a statistically significant impact on ultrafine particle count and CO concentrations. Wind speed also had a statistically significant effect but smaller. The small proportion in variability explained in PM2.5 by the model compared to the largest proportion in ultrafine particle counts and CO may be due to the effect of long-range transboundary sources, whereas for ultrafine particle counts and CO, local traffic is the main source.
1. Introduction As individuals move around within a transport microenvironment on different modes of transport, they are inadvertently exposed to air pollutants primarily from vehicle emissions released by traffic and/or influenced by dispersion conditions (1, 2). These can potentially cause numerous adverse health effects including cardiovascular and respiratory disease and mortality (3). Health effects such as myocardial infarction have been specifically linked with presence in the transport environment (4), and McCreanor * Corresponding author phone: (+34) 93 316 0646; fax: (+34) 93 316 0575; e-mail:
[email protected]. † Imperial College London. ‡ CREAL. 10.1021/es803199z CCC: $40.75
Published on Web 03/20/2009
2009 American Chemical Society
et al. (5) showed an acute reduction in lung function and an increase in inflammation associated with being in a transport environment. The transport microenvironment needs to be characterized since people spend a substantial component of their outdoor time in this environment (6). Exposures to pollutants in the transport microenvironment are often highly elevated compared to elsewhere; consequently, individuals may gain a significant contribution to their daily total exposure in a short period of time. In an urban transport microenvironment, the individuals exposed are not only restricted to those in motor vehicles, it also includes people waiting around traffic-congested streets, people working along busy streets, people commuting on bicycles, and people waiting to use public transport (e.g., buses). An understanding of factors that influence personal exposures is therefore vital in order to develop targeted control strategies in urban air quality management and to have a better understanding of the health risks posed by air pollutants in different conditions. However, few field studies have been performed that measure personal exposure concentrations along with other influencing variables and subsequently quantitative data on determinants of personal exposure are very limited. The influence of variables on exposure concentrations to date has largely been assessed qualitatively (e.g., refs 1, 2, 7, 8) or using ambient pollutant measurements as surrogates for personal exposure levels instead (e.g., refs 1, 2, 9). As part of the 2003 DAPPLE field campaign (10) we collected personal measurements (11) along with additional data for meteorological and traffic variables to investigate possible factors that impacted personal exposure concentrations. The aim of this paper is to assess the determinants of personal exposure concentrations using regression analyses to quantify and understand the contribution of key factors that may explain the variability of PM2.5, ultrafine particle count, and carbon monoxide (CO) personal exposure levels at and around an urban street intersection situated in Central London, UK.
2. Materials and Methods 2.1. Study Background. The first DAPPLE field campaign (10) was conducted in Central London between 28th April 2003 and 23rd May 2003. During the four week field campaign, groups of four volunteers collected data at each of the three timings (morning, lunch, and afternoon) with the exception of the first week when additional early evening measurements were made at a fourth timing. They traveled along two different routes (a heavily trafficked route and a backstreet route) via five modes of transport (walking, cycling, bus, car, and taxi) - the mode of transport and route followed by each volunteer at each timing were randomly designated. Exposure measurements were made for three pollutants: PM2.5, ultrafine particle counts (particle diameter range: 0.02-1 µm), and CO in conjunction with comprehensive meteorology and traffic measurements (11, 12). The exposure assessment measurements were made at the DAPPLE field study site, centered at Westminster Council House on the intersection of Marylebone Road and Gloucester Place in Central London, incorporating the surrounding area of approximately 250 m in radius (Figure 1). 2.2. Data Sets. 2.2.1. Personal Exposure Data. The personal exposure measurements collected in the 2003 DAPPLE field campaign are described in detail by Kaur et al. (11); however, a very brief summary is given below. VOL. 43, NO. 13, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4737
FIGURE 1. Map of the DAPPLE field site showing the exposure routes followed. The DAPPLE exposure assessment involved data collection while walking, cycling, or traveling by bus (number 18 and 27 modern double-decker buses depending on route), car (petrol with three way catalyst: Toyota Starlet, 1996) and taxi (diesel Black Cabs). Measurements were made around the field site on two routes: Route 1 was a heavily trafficked route along Marylebone Road and Route 2 was a figure-ofeight circuit that included Gloucester Place and backstreets to the north and south of Marylebone Road (Figure 1). Ultrafine particle count and CO data was collected in the latter three weeks of the field campaign alongside PM2.5 measurements at three timings: morning (timing 1 at 8.30 a.m.), lunchtime (timing 2 at 12 pm), and afternoon (timing 3 at 3.15 pm). PM2.5 personal exposure measurements were made throughout the 4 week field campaign at three timings with additional early evening measurements (timing 4 at 5.15 pm) during the first week. At each timing, four samples were collected by volunteers who were randomly designated a route and mode of transport. The PM2.5 personal exposure measurements were made using a gravimetric high-flow personal sampler developed by Adams et al. (13, 14) with minimum sample duration of 18 min. Alongside, ultrafine particle counts at 1 s resolution were recorded using P-TRAK ultrafine particle counters (model 8525) manufactured by TSI Inc., and personal exposure measurements of CO were made using the Langan (T15 and T15v) CO Measurers (Langan Products, Inc., San Francisco, CA). Ultrafine particle count and CO data, after downloading from their respective equipment, were converted into Microsoft Excel XP files, while PM2.5 data were entered manually into the software. The start and end times of the sample were used to select and extract the corresponding ultrafine particle count and CO data streams for each sample. The data streams were averaged to give a single average value of the ultrafine particle count and CO personal exposure for each sample, corresponding to the PM2.5 exposure value. 2.2.2. Meteorology Data. A reference automatic weather station constructed by the School of the Environment, 4738
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 13, 2009
University of Leeds, located on the roof of Westminster City Council House (approximately 15 m height) (Figure 1), collected meteorological data throughout the field campaign (10, 12). Meteorological parameters monitored included wind speed (m/s), wind direction, relative humidity (%), and temperature (°C). Hourly averages for each meteorological parameter corresponding approximately to the personal exposure timings were provided for use in the determinant analysis. 2.2.3. Traffic Data. Postprocessed traffic flow data, collected by the split, cycle and offset optimisation technique (SCOOT) network, was provided by the Institute of Transport Studies, University of Leeds, for the DAPPLE study site (10). Hourly bidirectional traffic flow averages for each route were calculated from the SCOOT loop network along the Marylebone Road and Gloucester Place sections of the study site corresponding approximately to the exposure timings for use in the determinant analysis 2.3. Determinant Analysis. The statistical software package SPSS for Windows (SPSS Inc. version 11.5) was used to produce descriptive statistics and to perform the general linear model (GLM) univariate procedure (analysis of covariance (ANCOVA)), which encompasses both regression analysis and analysis of variance (ANOVA) and allows the input of categorical and continuous variables. The personal exposure concentrations were the dependent variables while other variables such as mode of transport, temperature, and traffic counts were the independent variables. Although the distributions of the personal exposure concentrations was slightly skewed, we used untransformed concentrations to make the interpretation of the results easier and because the models appeared to be robust. The statistical assumptions underlying the ANCOVA, highlighted and described in detail by Munro (15) and Pallant (16), were checked carefully prior to performing the procedure. The checks on the data to ensure that the assumptions were not violated are not presented, but examples of the checks performed include homogeneity of variance, correlations among covariates, normality, lin-
TABLE 1. Summary Statistics of the Pollutants by Mode of Transport, Route, and Time of Day PM2.5 (µg/m3) N all
AM
141 32.9
median STD 31.4
14.7
range
ultrafine particulates (pt/cm3) N
6.0-77.5 55
AM 85123
median
STD
39 29 33 22 18
27.1 33.8 33.1 33.4 43.4
25.3 32.8 34.1 32.4 45.0
mode of transport 14.8 7.9-64.4 16 63065 59798 15084 15.8 9.7-77.5 8 77621 81500 20701 11.3 6.0-50.6 14 100018 100629 20349 13.1 15.2-58.5 9 101770 100152 28034 15.3 17.9-71.8 8 91947 97101 22849
Marylebone Road Gloucester Place
69 34.2 72 31.7
34.6 28.6
14.6 14.8
8:30 a.m. 12:00 p.m. 3:15 p.m. 5:15 p.m.
33 47 49 12
32.2 28.5 35.0 43.9
13.8 12.5-66.5 16 12.5 6.0-55.9 20 14.6 10.2-77.5 19 18.4 8.9-71.8
32.4 27.6 36.0 42.9
range
N
AM median STD
82480 25800 36679-151810 141 0.9
walk cycle bus car taxi
6.0-77.5 30 8.3-71.8 25
carbon monoxide (mg/m3)
0.6 dl< 2.1
16 14 19 10 13
0.7 0.9 0.8 1.3 1.2
0.6 0.7 0.8 1.4 1.2
0.6 0.6 0.5 0.6 0.5
route 95074 94343 23655 47300-151810 40 73182 69400 23436 36679-134995 32
1.2 0.6
1.1 0.4
0.5 0.4-2.1 0.5 dl< 1.9
time of 95996 80516 80816
1.3 0.7 0.8
1.4 0.6 0.8
0.5 0.2-2.1 0.6 dl
