Spatial and temporal measurements of NO 2 in an urban area using continuous mobile monitoring and passive samplers

Transcription

1 Journal of Exposure Analysis and Environmental Epidemiology (1999)9, 586± 593 # 1999 Stockton Press All rights reserved /99/$ Spatial and temporal measurements of NO 2 in an urban area using continuous mobile monitoring and passive samplers GARY NORRIS AND TIMOTHY LARSON Department of Civil and Environmental Engineering, Box , University of Washington, Seattle, Washington This paper describes the use of a continuous mobile monitor and passive samplers to estimate the spatial distribution of NO 2 in an urban area for the purpose of siting a continuous monitor to measure population exposure. Monitoring sites were sites selected based on the State and Local Air Monitoring Stations ( SLAMS) and National Air Monitoring Station ( NAMS) siting criteria required by the U.S. Environmental Protection Agency ( U.S. EPA). SLAMS monitoring objectives define scales in which the NO 2 concentration and land use are homogeneous. The SLAMS scales relevant to NO 2 monitoring for NAMS NO 2 monitoring sites are neighborhood ( 0.5 to 4 km), and urban ( several to 50 km). SLAMS siting objectives also define four categories of sites: highest concentration, representative concentration, impacts of major sources, and background sites. Mobile monitoring with a Scintrex LMA- 3 luminal monitor was used on a neighborhood scale to measure the NO 2 concentration at sites that covered a large geographical area. Passive samplers were then located at candidate mobile monitoring locations for long- term sampling which covered the neighborhood to the urban scale. These two methods complement each other by combining short-term continuous measurements and integrated long-term measurements which reflect the National Ambient Air Quality Standard for NO 2 which is based on an annual average. The neighborhood site with the highest concentration was not only in the area of highest population density, but was also representative of the larger urban scale. The magnitude of this urban scale is approximately 20 km. Keywords: NAMS, nitrogen dioxide, passive samplers, Scintrex LMA- 3, SLAMS. Introduction The goal of this study was to characterize the NO 2 spatial concentration distribution in the Seattle ± Bellevue urban airshed for the purpose of siting an NO 2 monitor to measure the highest average population exposure. The study design incorporates the siting criteria for NO 2 monitors required by the U.S. Environmental Protection Agency (U.S. EPA). In particular, specific monitoring locations were identified that meet both the ambient air monitoring objectives and siting criteria for NO 2 designated by the U.S. EPA under the State and Local Air Monitoring Stations ( SLAMS), and the National Ambient Monitoring Stations ( NAMS) criteria (40 CFR, Pt. 58). The techniques used to evaluate sites focused on relatively inexpensive methods for evaluating large numbers of potential monitoring sites. This study provides a basis for siting an NO 2 monitor for measuring population exposure based on a combination of mobile monitoring, passive sampling, SLAMS and NAMS siting criteria. 1. Abbreviations: NO 2, nitrogen dioxide; ANOVA, analysis of variance; NAMS, National Air Monitoring Station; SLAMS, State and Local Air Monitoring Station. 2. Address all correspondence to: Timothy Larson, PhD, Department of Civil and Environmental Engineering, Box , University of Washington, Seattle, WA tlarson@u.washington.edu Received 17 November 1997; accepted 1 March The SLAMS monitoring objectives define spatial scales in terms of homogeneous pollutant concentrations. Typical spatial scales associated with NO 2 monitoring are middle, neighborhood, and urban. A properly chosen middle- scale site represents NO 2 concentrations in the surrounding 100 to 500 m. A corresponding neighborhood scale site represents a reasonably homogeneous NO 2 concentration region over a scale of 0.5 to 4 km. In contrast, an urban scale site represents NO 2 concentrations over a large portion of an urban area, i.e., over a scale of several to 50 km. Neighborhood and urban scale sites must be kept at a minimum separation distance from roadways. This distance varies from 10 m for an average daily traffic flow of 10,000 to 250 m for an average daily traffic flow of 110,000 (40 CFR, Pt. 58, Appendix E). These distances minimize the impact of nearby mobile sources on the measured concentrations. In addition, the SLAMS criteria contain four monitoring objectives that focus on measuring population exposure. These objectives are to determine the highest concentration in the monitoring network, representative concentrations in high population areas, the impact of local sources, and background concentrations (40 CFR, Pt. 58, Appendix D). The four objectives are coupled with spatial scales to better define the monitoring goals. Local source impacts range from microscale to a neighborhood scale. Samplings over spatial scales from microscale to neighborhood are recommended for finding the ``highest'' concentration.

2 Spatial and temporal measurements of NO 2 Norris and Larson The spatial scales for population exposure range from a neighborhood to an urban scale. NAMS stations are selected from sites meeting applicable SLAMS siting criteria ( 40 CFR, Pt. 58, Appendix D). These sites are located where the NO 2 concentration is expected to be the ``highest'' and in areas of high population density. The determination of highest average concentration is based on the appropriate averaging time for the National Ambient Air Quality Standard for NO 2, which is the annual average. The relevant scales for NAMS NO 2 monitoring sites are neighborhood and urban. A site meeting both the NAMS and SLAMS criteria will represent a neighborhood or urban scale average concentration. In general, the EPA siting approach is different from that employed in many European countries. There, the emphasis is on microscale sites (spatial scale of several to 100 m) to find the maximum personal exposure concentrations. Based on the European Community Directive 85/ 203 ( CEC, 1985), measurement of NO 2 should be made ``in zones predominantly affected by pollution from motor vehicles, particularly street canyons.'' These sites are similar to the SLAMS middle- scale sites that represent maximum public exposure. The spatial distribution of NO 2 has been measured in numerous cities using passive samplers. NO 2 concentrations were measured in Lancaster, England using passive samplers; and concentrations ranged from a mean of 33.5 ppb NO 2 (63 g/m 3 ) for the city center, to a mean of 16 ppb NO 2 (30 g/m 3 ) for a suburban residential street ( Hewitt, 1991). An additional study conducted in London, England found NO 2 concentrations in street canyons to be 11 to 21 ppb NO 2 (20 to 40 g/m 3 ) higher than the local background levels (Laxen and Noordally, 1987). The effect of mobile emissions on the nitrogen dioxide concentration was also investigated in Tokyo, Japan ( Nakai et al., 1995). Passive samplers were put in zones 0 to 20 m, 20 to 150 m from a busy roadway, and also in a suburban reference site. The average NO 2 concentrations for all the sampling periods were 46 ppb at 0 m, 35 ppb at 150 m, and 12ppb in the suburban location. Passive sampling studies are useful for demonstrating elevated NO 2 levels based on neighborhood or middle scales in central city areas and street canyons due to emissions from motor vehicles and other local sources. The studies also show that the proximity of the samplers to the roadways has a critical effect on the measured average concentrations. Methods Both the NAMS and SLAMS criteria were combined to provide guidance on siting a monitor to measure population exposure in the Seattle and Bellevue, Washington urban areas. The spatial scales for the SLAMS criteria and recommended distances from roadways were used to determine initial candidate sites. Locating sites in heavily populated areas near major sources of NO 2, such as the major interstates passing through Seattle, incorporated SLAMS criteria. Passive sampling sites were located at those mobile monitoring sites with high average neighborhood scale concentrations; additional passive sites were added to determine potential concentration gradients over the urban scale. Mobile monitoring was used to measure NO 2 concentrations on a neighborhood scale over a large spatial region. The mobile monitor was placed in a truck and driven to a given site where a 5- min average concentration was collected. The mobile monitoring equipment included the following components: sampling probe, Scintrex/ Unisearch Luminox 1 NO Converter, Scintrex/Unisearch Luminox 1 LMA-3 Monitor, car battery (12V), DC/AC inverter, Environmental Systems Corporation 8800 data logger, and a lap-top computer. The Scintrex LMA-3 directly measures NO 2, and has response time of 1 sec for a 10 ppb change in NO 2 (Drummond et al., 1990). Responses to O 3 and peroxyacetyl nitrate (PAN) are reported to be less than 1% and 26% of the response to NO 2, respectively (Drummond et al., 1990). The LMA-3 monitor was calibrated with an Environics 9100 computerized ambient calibration system. A certified 10 ppm NO tank and a Thermo Environmental Instruments Model 111 zero air system were used to provide the calibration and dilution gas, respectively. All calibrations were conducted through the sampling probe. After the field sampling, a zero and one- point calibration was run to evaluate response drift. At each site, the 5- min average of 1-sec readings from the LMA-3 was recorded using the Environmental Systems Corporation 8800 datalogger. The Seattle and Bellevue urban and suburban areas were divided into a North and a South region for mobile monitoring. Sites were selected within the grids of The Thomas Guide ( 1994) map of King County, using the detail maps that have a grid size of 5.6 km (N/S)7.2km (W/E). The Seattle study region had four grids north to south (22.4 km) and two grids east to west (14.4 km), with a total of 16 sampling locations. The Bellevue study region had four grids north to south (22.4 km) and two grids east to west (14.4 km), with a total of 25 sampling locations. These neighborhood scale mobile monitoring sites were chosen as those with a low standard deviation of the NO 2 measurements collected over the 5-min sampling period. This technique was used to exclude the presence of local sources in the vicinity of the monitoring site to ensure that the site represented a neighborhood scale rather than a micro- or middle-scale. Either the North or South region was selected on a given day, and the order in which the sites were visited was randomized. Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6) 587

3 Norris and Larson Spatial and temporal measurements of NO 2 Table 1. Passive sampling validation study conducted in Portland, OR. Sample Sampling period ( days) NO 2 ( ppb), continuous NO 2 ( ppb), passive Passive ( ppb)/ Chemiluminescent( ppb) Samples The NO 2 passive sampler or ``filter badge'' in this study was developed for personal exposure monitoring ( Yanagisawa and Nishimura, 1982). The passive sampler is small (5 cm4 cm1 cm), and has a high sensitivity (detection limit 66 ppb h). Passive samplers were placed at selected mobile monitoring locations. These sites represented either highest average mobile locations, or areas on the fringe of the sampling region. Samples were exposed for a 3- week sampling period. A shelter was used to protect the samplers from precipitation and wind effects and to allow samplers to be placed away from buildings. Sites were located away from NO 2 sources, and were placed in ventilated areas. An overall mass transfer coefficient (K og ) of 0.09 cm/sec was used to calculate the passive sampler concentration based on the pilot study described below. In this pilot study, the shelter design and samplers were evaluated against a continuous chemiluminescent analyzer operated by the Oregon State Department of Environmental Quality in Portland, Oregon during the summer of The results of the passive sampling validation study are displayed in Table 1. The passive sampling results were 61% to 84% of the continuous chemiluminescent analyzer at the Portland site when using an assumed K og of 0.14 cm/ sec ( Yanagisawa and Nishimura, 1982). A linear regression of the sampling rates (g/m 3 h) consisting of 1-, 2-, and 3- week average samples has the following relationship: passive NO 2 (g/m 3 h) =0.62(chemiluminescent (g/m 3 h)), R 2 =0.99. The intercept term was not statistically significant. A K og of 0.09 was calculated from relationship shown above (0.14 cm/sec0.62). In an earlier study by Berglund et al. (1992), the same passive sampler measured 61% to 93% of the concentrations measured by a chemiluminescent analyzer when using a K og of 0.14 cm/sec (Yanagisawa and Nishimura, 1982). The K og for these samplers has been further investigated and a lower K og of 0.10 was recommended for indoor sampling (Lee et al., 1992). The apparently low collection rate that was determined in the pilot study (0.09) might be due to losses on the surfaces of the shelter or to the systematically lower wind speeds inside the shelter. However, the correlation with the chemiluminescent analyzer is high and therefore, the reduced correlation rate is systematically accounted for in the estimation of K og. The passive samplers were analyzed according to the procedure described by ( Yanagisawa and Nishimura, 1982). Duplicate samples were collected at two or three sites during each sampling period. The average coefficient of variation (CV) for 48 duplicate samples was 3% or an average difference of 0.5 ppb. This duplicate error is comparable to the CVof 3% found for the same samplers by Treitman et al. (1990). Results The mobile and passive sampling results were used to estimate both spatial scale as a function of location and the NO 2 concentrations throughout the Seattle±Belleuve urban area. A model was first developed, which accounts for the variation in the measured concentration caused by differences in atmospheric mixing during different sampling periods. Second, the monitoring areas were evaluated using analysis of variance (ANOVA) to determine if any of the site means was significantly different. Third, the spatial data were plotted to show the relative NO 2 concentrations in the study regions. Last, the spatial scale was estimated for the permanent monitoring site that was selected as part of this study. A multiplicative air pollution model was used to describe the sampling data. This model attributes spatial variations in NO 2 concentrations by site to different source strengths of NO 2 by site and an overall daily dilution (Equation 1). This model describe the measured concentrations in our study given the limited geographical area covered during the sampling periods and the emission density of NO and NO 2 from combustion sources. The model also contains a reaction term for the loss or gain of NO 2 due to deposition or photodissociation of reactants in the tropospheric nitrogen cycle. Multiplicative model:! Source i g Ci;j ˆ Reaction ig Dilution j m 3 Eq. 1 where: C = measured concentration; i =site; j= sampling period or day. Table 2. F -statistics for the effect of site for mobile monitoring. Sampling area F- statistics for effect of site Significance of F North Seattle South Seattle 5.12< North Bellevue a South Bellevue a One site was removed from the analysis because the variance was low compared to the other sites. 588 Journal of Exposure Analysis and Environmental Epidemiology (1999) 9 (6)

4 Spatial and temporal measurements of NO 2 Norris and Larson Table 3. F -statistics for the effect of site for passive sampling periods 2to 5. Sampling period F - statistics for effect of site Significance of F < < < < Based on this multiplicative nature, removal of the dayeffect would be based on dividing the concentration at a site by the overall average of the sites on the measurement day or period. The site concentration without the effect of day or sampling period is the normalized concentration (C). Other techniques for evaluating spatial monitoring assuming transport as the major source of air pollution variation have used the concentration minus the predicted concentration from a smooth function for the site (Guttorp et al., 1994). Mobile Monitoring The mobile monitoring study was divided into four regions: North and South Seattle, and North and South Bellevue. The North and South Seattle study regions were each sampled eight times. A total of 8 and 9 days were measured for the North and South Bellevue study regions, respectively. The mean NO 2 concentrations between September 93 and November 93 for the North and South Seattle study regions were 13 and 18 ppb, respectively. Mean NO 2 concentrations for the North and South Bellevue study regions between September 93 and May 94 were 19 and 16 ppb, respectively. The normalized mobile monitoring data were analyzed using ANOVA. The effect of site was statistically significant ( p < 0.05) for the North Seattle, South Seattle, and South Bellevue sites. Table 2and Table 3 shows the F-statistics and the significance of the effect of site for the sampling areas. The use of temperature and wind speed covariates did not significantly improve the model. North Bellevue was the only region which did not have a significant site effect. North Bellevue also had one site with a relatively low variance in the normalized site concentrations. This site was located on the eastern edge of the sampling region, and not impacted by local NO 2 sources. The site was removed from the ANOVA analysis due to the lack of homogeneity of variance when all the sites were analyzed. After removal of the site from the analysis, the effect of site was not statistically significant ( p =0.25), indicating that the North Bellevue region has a homogeneous NO 2 concentration. The mobile monitoring data for Seattle and Bellevue are summarized using normalized site concentrations in Figure 1. If the average normalized value minus 1 standard error (SE) is greater than 1, the site is designated as high. If the average normalized value plus or minus 1 SE crosses 1, then the site is designated as average. Last, if the average normalized value plus 1 SE is less than 1, the site is designated as low. Major freeways and interstates are Figure 1. Seattle and Bellevue mobile monitoring results. The high concentration sites are located near the Interstates 5 ( I5) and 405 (I405) which are the major freeways which pass through Seattle and Bellevue. Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6) 589

5 Norris and Larson Spatial and temporal measurements of NO 2 included on the maps to show influences on the site from major mobile sources. The mobile monitoring results show that the highest average NO 2 concentrations are in Southeast Seattle. The high region extends approximately 12km (north/south) in Eastern Seattle. The highest concentration sites for the Bellevue region are in the proximity of Interstates 90 (I90) and 405 (I405). Passive Sampling The passive sampling data were collected in five phases, with the number of sites ranging from 7 to 14 for each phase. Each phase was designed to obtain more information about the airshed based on the findings of the earlier phase. Figure 2shows the sampling locations used for each phase of the passive sampling. ANOVA was used to determine if any of the passive sampling normalized site means was significantly different. One site was excluded from the analysis because it is a microscale site and does not meet the siting criteria for a NO 2 monitor as defined by the NAMS siting criteria. The effect of site is highly significant ( p < 0.001) in all of the sampling periods, which indicates that at least one of the sites has an average which is significantly different from the mean of all the sites. The statistically significant effect of site confirms the passive sampling site selection that contained locations to measure both the high and regional concentrations of NO 2. Figure 2. Locations of passive samplers in the study region. The high concentration sites are consistently high throughout the five phases of the passive sampling study. 590 Journal of Exposure Analysis and Environmental Epidemiology (1999) 9 (6)

6 Spatial and temporal measurements of NO 2 Norris and Larson The passive sampling sites and their relative concentrations are shown in Figure 2. The sites are identified as high, medium, or low based on normalized site concentrations and the site standard errors using the criteria discussed earlier for the data shown in Figure 1. The initial sampling phase between June 25 and August 27, 1993 identified the city of Seattle as the area of high average NO 2 concentrations. The sampling phase between September 23 and December 15, 1994 demonstrated that the high concentrations occurred in South Seattle. The third sampling phase between December 15 and February 16, 1994 extended the sampling region south while eliminating low concentration sites in North Seattle. The fourth sampling phase between February 16 and June 13, 1994 includes sites from Bellevue and Seattle, allowing for a large area to be studied. The last phase between June 13 and August 26, 1994 extended the study region further south along the corridor of prevailing wind direction. During all the passive sampling phases, the sites in Seattle showed the highest normalized concentrations. The passive sampling sites for the period between February 16 and June 13 (phase 4) are shown in Figure 3. This sampling phase covered a large spatial area and allows for evaluation of the spatial scale in the Seattle± Bellevue urban area. Sites 1 and 9 are on the fringe of the Seattle ± Bellevue urban area and are in low-density housing areas. Sites 2, 6, 7, 10, 11, 12, and 13 are in urban residential areas. Sites 3 and 5 are located in Central Seattle in areas of high traffic density. Site 4 is located in a park. The two highest average concentration sites are site 3, which is located at the University of Washington (UW), and site 5, which is located at the Beacon Hill Reservoir (BHR). The UW and BHR sites consistently had the highest average concentrations for all the sampling phases shown in Figures 2and 3. A street canyon site, which does not meet EPA Siting Criteria, was added near UW (+) to investigate the difference between the European siting criteria and the U.S. EPA siting criteria. The additional site was located at a carbon monoxide sampling probe used by the Washington State Department of Ecology. The average concentrations of the UW site and the street canyon site were 21 and 31 ppb, respectively. The correlation between the BHR site and the other sites in Figure 3 is plotted in Figure 4. The plot shows a correlation above 0.88 for distances up to 20 km. The correlation for the sites within Figure 3. Passive sampling locations between February 16 and June 13, 1994 ( n=6). The mean NO 2 concentrations range from 8 to 15 ppb, with the highest average concentrations at the UW and BHR sites. Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6) 591

7 Norris and Larson Spatial and temporal measurements of NO 2 Discussion Figure 4. Pair- wise correlation versus distance from site 5 ( February 16±June 13, 1994, n =6) (site numbers are shown). the 20-km distance range from a low of 0.88 for site 3 at a distance of 9 km to a high of 0.99 for site 7 at a distance of 10 km. Additional sites are required to better define the relationship between distance and correlation beyond 20 km. Based on the lack of measurements between 20 and 30 km, the area of homogeneous concentrations using this method is limited to 20 km. Another technique that was evaluated assumes that the decrease in concentration from the location of maximum concentration is proportional to the distance from the maximum concentration (Equation 2). The model uses the normalized concentration (C) which is based on the multiplicative model (Equation 1). This allows for the calculation of the characteristic length or the average distance at which the normalized concentration is 37% (100(1/e)) of the maximum concentration: L ˆ 1 n 1 X n 1 iˆ1! Ci 1 Xi Eq. 2 where: Ci =normalized concentration difference ( maximum value value at site i); L=characteristic length; Xi=distance between maximum site and site i; n=number of sites. L is an estimate of the spatial scale of the elevated NO 2 concentration in the urban area. The value of L was found to be 24 km (s.d.=11 km) from the maximum site (BHR). This agrees well with the correlation between the regional passive sampling sites and the BHR site, which was high (above 0.88) for spatial distances up to 20 km. The BHR site also meets the SLAMS objective for a high population density site. The average population density for the census tracts included in a 4-km radius (maximum neighborhood spatial scale) of the BHR site is 2720 people/km 2 (U.S. Department of Commerce, 1990). This density is high when compared to the average population density of 1697 people/km 2 for the Seattle± Bellevue urban area. This study incorporated mobile monitoring, passive sampling, SLAMS, and NAMS siting criteria for locating a monitoring station to estimate the population exposure to NO 2. A network of both mobile monitoring and passive sampling sites was established to measure the spatial distribution of NO 2 in the Seattle±Bellevue urban area. Both methods complemented each other by providing both short (5 min) and long-term (3 weeks) average NO 2 concentrations. A total of 41 mobile monitoring locations and 24 passive sampling sites were evaluated. The combination of the regional sampling and the guidance in the NAMS and SLAMS siting criteria identified two sites that had the highest average concentrations. The two passive sampling sites with the highest mean concentrations calculated from 57 weeks of passive sampling data were UW (24 ppb) and BHR (26 ppb). The BHR average concentration was significantly higher than the UW site ( p<0.05). The spatial scale of the BHR monitoring site meets the primary monitoring objectives for SLAMS of measuring the highest concentration in high population areas ( 40 CFR, Pt. 58, Appendix D). This site was chosen based upon EPA criteria for the neighborhood scale ( distance from roadways). Based upon our results, we can say that the BHR site is not only an appropriate neighborhood scale site, but also is an urban scale site representing an area of about 20 km radius. The passive sampling method can demonstrate the extent of the urban scale, i.e., 20 km in this case. Because this is an urban scale site, it is also by definition a neighborhood scale site, i.e., the NO 2 concentrations are uniform over a 20-km area and therefore also uniform over a smaller scale. The mobile monitoring was also used to verify the absence of short-term peaks due to small-scale plumes that might otherwise impact the site Ð additional evidence that this is a neighborhood scale site. This study demonstrates the use of mobile and passive sampling to select a monitoring location that satisfies the requirements in the NAMS and SLAMS siting criteria. A site was determined, which meets both the spatial scale and other objectives specified in siting criteria. The BHR site represents an urban scale site with a spatial scale between 20 and 24 km, which measures the highest population exposure in the Seattle ±Bellevue urban area. As a result of this study, a NO 2 analyzer was located at the BHR by the Washington State Department of Ecology. Acknowledgments The Washington State Department of Ecology funded this research. We thank Dr. Serap Erdal, the Washington 592 Journal of Exposure Analysis and Environmental Epidemiology (1999) 9 (6)

8 Spatial and temporal measurements of NO 2 Norris and Larson State Department of Ecology, and the Oregon State Department of Environmental Quality for their assistance during this study. References Berglund M., Vahter M., and Bylin G. Measurement of personal exposure to NO 2 in Sweden Ð evaluation of a passive sampler. J. Expos. Anal. Environ. Epidemiol. 1992: 2 ( 3 ) : 295±307. CEC. Council directive on air quality standards for nitrogen dioxide 85/ 203/EEC. Off. J. Eur. Commun. 1985: L87: 1 ±7. Drummond J.W., Castledine J., Green J., Denno R., Mackay G.I., and Schiff H.I. New technologies for use in acid deposition networks, ASTM STP In Zielinski W.L., and Dorko W.D. ( Eds.), Monitoring Methods for Toxics in the Atmosphere. American Society for Testing and Materials, Philadelphia, 1990, pp. 133± 149. Guttorp P., Meiring W., and Sampsono P.D. A space± time analysis of ground- level ozone data. Environmetrics 1994: 2: 241± 254. Hewitt C.N. Spatial variations in nitrogen dioxide concentrations in an urban area. Atmospheric Environment 1991: 25B ( 3 ) : 429±434. Laxen D.P.H., and Noordally E. Nitrogen dioxide concentration distribution in street canyons. Atmospheric Environment 1987: 21 ( 9 ) : Lee K., Yanagisawa Y., Spengler J.D., and Billick H.I. Wind velocity effects on the sampling rate of NO 2 badge. J. Expos. Anal. Environ. Epidemiol. 1992: 2 ( 2 ) : 207±219. Nakai S., Hiroshi N., and Maeda K. Respiratory health effects associated with exposure to automobile exhaust: II. Personal NO 2 exposure levels according to distance from the roadside. J. Expos. Anal. Environ. Epidemiol. 1995: 5 ( 2) : 125±136. The Thomas Guide. King/ Pierce/ Snohomish Counties Street Guide and Directory, 1994 Edn. Thomas Bros. Maps, Treitman R.D., Ryan P.B., Soczek M.L., Yanagisawa Y., Spengler J.D., and Billick I.H. Sampling and analysis of nitrogen dioxide and respirable particles in the indoor environment. In: Monitoring Methods for Toxics in the Atmosphere. ASTM STP 1990: 1052: 197±212. U.S. Department of Commerce Census of population and housing, population and housing characteristics for census tract and block numbering areas, Seattle- Tacoma, WA CMSA, Seattle, WA PMSA. CPH-3-301A, 1990, pp. 1±52. Yanagisawa Y., and Nishimura H. A badge- type personal sampler for the measurement of personal exposure to NO 2 and NO in ambient air. Environ. Int. 1982: 8: 235±242. Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6) 593