Source Proximity and Outdoor-Residential VOC ... - ACS Publications

Jun 6, 2006 - of New Jersey, Piscataway, New Jersey, Division of Science,. Research and Technology, New Jersey Department of. Environmental Protection...
0 downloads 9 Views 339KB Size
Download PDF
Environ. Sci. Technol. 2006, 40, 4074-4082

Source Proximity and Outdoor-Residential VOC Concentrations: Results from the RIOPA Study J A Y M I N K W O N , †,§ C L I F F O R D P . W E I S E L , * ,†,‡ B A R B A R A J . T U R P I N , †,§ J U N F E N G ( J I M ) Z H A N G , †,| L E O R . K O R N , ⊥ MARIA T. MORANDI,# THOMAS H. STOCK,# AND STEVEN COLOME@ Environmental and Occupational Health Sciences Institute, Piscataway, New Jersey, Robert Wood Johnson Medical School, University of Medicine & Dentistry of New Jersey, Piscataway, New Jersey, Department of Environmental Sciences, Rutgers University, New Brunswick, New Jersey, School of Public Health, University of Medicine & Dentistry of New Jersey, Piscataway, New Jersey, Division of Science, Research and Technology, New Jersey Department of Environmental Protection, Trenton, New Jersey, School of Public Health, Houston Health Sciences Center, University of Texas, Houston, Texas, and Integrated Environmental Sciences, Irvine, California

Ambient volatile organic compound concentrations outside residences were measured in Elizabeth, New Jersey as part of the Relationship of Indoor, Outdoor, and Personal Air (RIOPA) study to assess the influence of proximity of the residences to known ambient emissions sources. The closest distances between the outdoor samplers and emission sources were determined using Geographic Information Systems (GIS) techniques. Multiple regression models were developed for residential ambient concentrations of aromatic hydrocarbons (BTEX), methyl tert butyl ether (MTBE), and tetrachloroethylene (PCE). The natural log transformed ambient concentrations of BTEX were inversely associated with distances to major roadways with high traffic densities and gasoline stations, atmospheric stability, temperature, and wind speed. Ambient MTBE levels were associated with inverse distance to gas stations and interstate highways. Residential ambient PCE concentration was inversely associated with distance to dry cleaning facilities, atmospheric stability, temperature, wind speed, and relative humidity. The linear regression models that include proximity to emission sources and meteorological variables explained 16-45% of the overall variation of ambient residential VOC concentrations. Meteorological conditions, especially atmospheric stability and temperature, explained 60-90% of the total variation in the regression models. The residential ambient air * Corresponding author phone: 732-445-0154.; fax: 732-445-0116; e-mail: [email protected]. † Environmental and Occupational Health Sciences Institute. ‡ Robert Wood Johnson Medical School, University of Medicine & Dentistry of New Jersey. § Rutgers University. | School of Public Health, University of Medicine & Dentistry of New Jersey. ⊥ New Jersey Department of Environmental Protection. # University of Texas. @ Integrated Environmental Sciences. 4074

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 13, 2006

concentrations were 1.5-4 times higher than the urban background levels outside homes very close (50 m) from their location on the DOQQs. To improve the accuracy of the proximity data, TIGER map errors were corrected to follow the roadway centerlines observed on the DOQQs, before geocoding and distance calculation. The locations of each outdoor residential sampler, point sources, gas stations, and the dry cleaning facilities, were determined by address-matching techniques, and corrected by using street information recorded during actual visits, using digital orthophoto, the City engineer’s map, and the GPS coordinates readings. Coordinates of samplers and emission sources were determined by running an ArcScript “addxycoo”. Distances between samplers and sources were calculated by “the nearest features” extension patch available at the ESRI’s site for ArcScripts (22). Multiple Regression Analyses. Statistical analyses were performed on the log transform uncensored ambient VOC concentration using SAS (version 8.2, SAS Institute Inc.) and SPSS (version 12.0, SPSS Inc.) for Windows. When the blank corrected concentrations were zero or negative, onehalf the detection limit was substituted prior to the log transformation. Multiple linear, forward stepwise regression analysis was conducted to identify predictor variables with an inclusion criteria of p < 0.15. Scatter plots of residential ambient air concentrations of VOCs and each independent variable were examined visually for any obvious associations. Distances of residences to mobile, area, and point emission sources, wind speed, atmospheric stability, mixing height, temperature, relative humidity, precipitation, and atmospheric pressure were used as the independent variables. Heteroscedasticity was tested for the assumption of equal variances in error terms. The equal variances in error terms were assumed in the best-fitting model when the p-value of the chi-square test was greater than 0.05. Since the RIOPA sampling included sampling a residence twice, a test on the effect of repeated measurement was conducted using SAS “Proc Mixed”. Regression parameter estimates were considered significant at p < 0.05. Possible model outliers from the dataset were identified in order to improve the model’s median trend and stability and to determine if additional parameters would be included in the model for the data that best fit the model. An outlier was selected when the absolute value of studentized deleted residuals were larger than 1.654 (F test statistics of 0.95, df 170). These outliers were excluded from the finalized bestfitting models. The model outliers were data points that were not fit well by the model, suggesting that parameters not being considered in the model (distance to known emissions and selected meteorological variables) were controlling their concentrations. The outliers included concentration values across the entire spectrum measured during the study. To verify that removal of outliers did not introduce biases in the models calculated, the means of each predictor variable for the outlier samples were compared to the means of predictors for the nonoutliers using SAS “PROC ANOVA; PROC GLM” for unbalanced datasets.

Results and Discussion Model Selection. Descriptive statistics for all the variables are listed in Table 1A-C. The best-fitting regression model 4076

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 13, 2006

had a similar form for all compounds and the model for m,p-xylene is presented as an example:

Ln(mpX) ) 9.2 + 8.4FC14-1 + 18.5GS-1 + 0.92S 0.03K - 0.14U + 0.006P where Ln(mpX) is the natural log-transformed ambient air concentration of m,p-xylene (µg/m3), FC14-1 is the inverse distance (m) to the closest major arterial roadway, GS-1 is the inverse distance (m) to the closest gas station, S is the atmospheric stability, K is the temperature (Kelvin), U is the average wind speed (m/s), and P is the precipitation (mm). For each compound, the model residuals plotted against the predicted values were randomly distributed without obvious trends or particular patterns, and had approximately normal distributions with constant variances. The probabilityprobability (PP) plot was nearly linear so the error term of the model was consistent with a normal distribution. Based on these methods of visual diagnoses, there was no clear evidence of lack of fit or of significant unequal error variance for the regression model. Based on the variation inflation factors for all predictors, which were close to 1 (1.04-1.46), it was concluded that there is no significant collinearity between the predictors included in the model. The parameters for the source proximity and meteorological predictors included in the best-fitting models of the selected VOCs are summarized in Table 2. The linear regression models that include proximity and meteorological variables explained 16-45% of the overall variability in lntransformed ambient VOC residential concentrations in Elizabeth. While the inverse distance to major urban arterial roadways (FC14-1) was a significant predictor in the bestfitting models for residential ambient air concentrations of all of the aromatic VOCs (BTEX), the inverse distance to urban interstate highways (FC11-1) was a significant predictor in the best-fitting model for residential ambient air concentrations of MTBE. The inverse distance to the closest gas station (GS-1) was a significant predictor in the models of residential ambient air concentrations of m,p-xylene, o-xylene, benzene, and MTBE, but not ethylbenzene nor toluene. No inverse distance to any of the point sources in the area was included as a significant predictor of residential ambient VOC concentrations. The inverse distance to the closest dry cleaning facility (DCF-1) was a significant predictor variable in the model of the residential ambient air concentration of PCE in Elizabeth. Overall, the meteorological variables contributed more (58-91%) to the explanatory power of the regression models than the source proximity variables. Temperature and wind speed were statistically significant predictors in the bestfitting models for all target VOCs. Atmospheric stability was a significant predictor for all VOCs except MTBE. Relative humidity (RH) was significantly associated with o-xylene, toluene, and PCE, while precipitation was a significant predictor only for m,p-xylene. Neither mixing height nor atmospheric pressure was a significant predictor in any of the models. The model results obtained for the VOCs were examined to determine if a consistency in the parameter estimates of the proximity variables was observed based on emission rates and volatility. The annual emission from mobile sources in Union County, NJ is ordered as follows: toluene, xylenes, MTBE, benzene, and ethylbenzene (9). The magnitude of the regression coefficients in the models for proximity to major roadways were higher in MTBE, toluene, and xylene (when combined), while the coefficients in the benzene and ethylbenzene models were lower, consistent with the emission rates from automobiles. The ordering of chemicals for emissions from service station storage tanks (Phase I) in

TABLE 1. Descriptive Statistics of Dependent and Independent Variables (N ) 183)

compounds

percent above DLa

m,p-xylene o-xylene toluene benzene ethylbenzene MTBE PCE

95% (4) 91% (1) 41% (3) 58% (1) 59% (7) 90% (8) 24% (15)

A. VOC Concentrations in Residential Ambient Air (µg/m3) percentiles standard mean deviation MDLb 25 50 3.25 1.71 6.82 1.50 1.35 5.75 1.11

variable, unit

mean

temperature, K wind speed, m/s relative humidity, % atmospheric pressure, mmHg precipitation, mm mixing heights, km fraction of time atmosphere was stable or neutral

286.5 4.3 66.5 762.3 6.6 1.027 0.66

emission sources

4.29 6.51 5.83 1.54 2.74 5.33 3.08

1.4 0.85 6.7 1.1 0.74 0.68 0.42

B. Meteorological Variables standard deviation minimum 8.7 1.1 12.6 4.5 14.2 0.362 0.17

265.5 1.9 42.7 750.3 0.0 0.414 0.0

1.51 0.59 2.59 0.69 0.46 2.23 0.50

2.37 0.94 4.83 1.22 0.99 4.35 0.74

75

maximum

3.97 1.38 9.36 1.90 1.74 7.51 1.11

51.2 81.0 32.9 18.1 36.2 27.2 41.8

25

percentiles 50

75

maximum

279.5 3.6 59.2 759.6 0.0 0.767 0.59

288.3 4.4 66.5 761.6 1.0 0.948 0.65

294.1 5.1 75.9 765.5 5.8 1.214 0.77

303.3 8.0 91.8 773.1 84.1 2.099 1.0

C. Distances from a Sampler Location to the Emission Sources (km) percentiles standard mean deviation minimum 25 50

75

maximum

FC11 (interstate highways) FC12 (freeways/expressways) FC14 (major arterial) FC16 (minor arterial) FC17 (collector) FC19 (local)

1.529 2.529 0.500 0.193 0.288 0.034

1.052 1.162 0.538 0.167 0.216 0.021

0.037 0.024 0.013 0.005 0.020 0.002

0.677 1.472 0.114 0.066 0.106 0.023

1.327 2.866 0.330 0.128 0.247 0.031

2.277 3.440 0.651 0.320 0.391 0.038

3.698 5.578 2.489 0.782 0.967 0.130

gas station dry cleaning facilities refinery tanker terminal surface coating industry aviation service pharmaceutical production chemical industry POTWc

0.36 0.55 2.98 4.78 2.51 5.92 4.19 3.27 2.77

0.21 0.39 1.12 1.14 1.00 1.15 1.11 1.14 1.20

0.03 0.06 0.84 3.23 0.62 2.80 0.81 0.99 0.40

0.22 0.25 2.06 3.78 1.75 5.03 3.50 2.46 1.91

0.36 0.43 3.07 4.58 2.60 6.18 4.43 3.24 2.36

0.49 0.77 3.77 5.72 3.26 6.91 5.08 3.98 3.80

1.01 1.69 5.76 7.69 5.63 8.63 6.64 6.04 5.81

a Ref 31. Number of zero values that were substituted by the half of method detection limit is in parentheses. Treatment Works, wastewater treatment facility.

Union County, NJ, which is a function of fuel composition and volatility of individual compounds, were as follows: MTBE (3.9 t), toluene (3.6 t), benzene (2.1 t), xylenes (1.3 t), and ethylbenzene (0.29 t) (9). The order of the magnitude of the regression coefficients for proximity to gas station was similar to the emission amounts for MTBE, m,p-xylene, benzene, and o-xylene, while gas station proximity was not a significant predictor in the toluene and ethylbenzene model. The lack of proximity to gas station being included in the ethyl benzene model could reflect its low emission rate and the number of air concentrations below the detection limit reducing the resolving power of the regression analysis for that compound. Thus, with the exception of toluene the model estimates are consistent with the emission rates of gasoline. One possible reason that inverse distance to gas stations was not included for toluene may be that toluene has other larger area sources, such as coating related products and graphic arts in Union County, NJ which were not accounted for in the model runs. The final regression model of PCE included the proximity to dry cleaning facilities along with meteorological variables, consistent with PCE being the primary solvent used in the dry cleaning facilities (30.7 tons) and emission from those facilities accounting for almost 78% of total emission in Union County, NJ.

b

Ref 10. c POTW: Publicly Owned

Test of Model Outliers, Repeated Measures Analysis, and Left Censored Data Model. The results of PROC ANOVA; PROC GLM verified that no bias was introduced into the model parameters by excluding outlier values of the original model selected by the F-test statistic. The Duncan’s multiple range test results indicated that means of all of the predictor variables were not significantly different between the outliers and nonoutliers at R ) 0.05 for all compounds. The reason for a data point to be considered an outlier could not be determined, though possible explanations include an unknown source in the region and analytical uncertainty at the low concentration range. A repeated measure analysis was conducted to compare the model for the sample pairs collected at an individual home. The within home variability was similar to the between home variability, and there was no significant correlation between the within home measurements suggesting that it was appropriate to use the two samples collected at each home as distinct samples. Due to the large percentage (17%) of concentrations being below the detection limits for PCE, the differences between the models derived from uncensored and censored data were evaluated. A regression model generated by left censoring of residential ambient PCE data with reported method detection VOL. 40, NO. 13, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4077

TABLE 2. Summary of Finalized Best-fitting Models of Selected VOCs (Ln-Transformed Concentrations, µg/m3, N ) 183) VOC

m,p-xylene r2)0.32 n )169, e ) 14 o-xylene r2 ) 0.45 n ) 164, e ) 19 toluene r2 ) 0.27 n ) 165, e ) 18 benzene r2 ) 0.39 n ) 169, e ) 14 ethylbenzene r2 ) 0.16 n ) 170, e ) 13 MTBE r2 ) 0.24 n ) 171, e ) 12 PCE r2 ) 0.31 n ) 164, e ) 19

type of predictors

Xi βi(se) p-r2 Xi βi(se) p-r2 Xi βi(se) p-r2 Xi βi(se) p-r2 Xi βi(se) p-r2 Xi βi(se) p-r2 Xi βi(se) p-r2

β0 9.2(1.6)b β0 7.1(1.3)b β0 7.3(2.0)c β0 11.4(1.4)b β0 6.9(2.3)c β0 -0.9(2.0)g β0 3.0(1.2)d

mobile/area/point sources FC14-1 8.4(4.5)e 0.01 FC14-1 9.7(4.4)d 0.01 FC14-1 18.6(4.6)b 0.06 FC14-1 5.4(3.4)f 0.01 FC14-1 9.6(5.7)e 0.015 FC11-1 22.8(14.4)f 0.01

GS-1 18.5(6.4)c 0.05 GS-1 9.9(5.3)e 0.03

GS-1 17.9(5.6)c 0.05

GS-1 35.5(8.2)b 0.09 DCF-1 32.5(12.4)c 0.04

meteorological variables

S 0.92(0.30)c 0.11 S 1.04(0.25)b 0.24 S 0.76(0.37)d 0.12 S 0.54(0.26)d 0.02 S 1.18(0.43)c 0.08

S 0.38(0.25)f 0.01

K -0.03(0.005)b 0.11 K -0.03(0.004)ba 0.09 K -0.02(0.007)c 0.05 K -0.04(0.005)b 0.25 K -0.03(0.01)c 0.05 K 0.01(0.007)e 0.01 K -0.01(0.004)c 0.05

U -0.14(0.05)c 0.03 U -0.14(0.04)c 0.06 U -0.10(0.05)e 0.02 U -0.11(0.04)c 0.07 U -0.11(0.07)e 0.015 U -0.24(0.05)b 0.12 U -0.15(0.04)b 0.18

precip 0.006(0.003)d 0.02 RH 0.007(0.003)d 0.02 RH 0.02(0.005)c 0.03

RH 0.01(0.003)c 0.03

a X , i th predictor variable; β , intercept of model; β , parameter estimate of i th predictor; se, standard error of parameter estimates; r2, coefficient i 0 i of determination; p-r2, partial r square of the variable; -1, indicates inverse values; n, sample size of finalized model; e, number of model outliers removed. S, atmospheric stability. K, temperature. U, wind speed. Precip, precipitation. RH, relative humidity. b p < 0.0001. c p < 0.01. d p < 0.05. e p < 0.10. f p < 0.15. g p ) 0.64.

limit, “proc Lifereg”, was used. The parameter estimates for the independent variables in the model for left-censored data were very similar to the parameter estimates for the PCE model provided in Table 2. Parameter estimates for inverse distance to dry cleaning facilities were 63.6(22.2 for the censored dataset and 58.7(21.2 from the uncensored dataset. Chi square statistics were significant for all predictors in the model. Estimates of Model Predictions. To estimate the effect of the distance to emission sources and meteorological parameters on the residential ambient air concentrations of the selected VOCs, the expected ambient concentration of a compound was calculated for the distance to roadway and to gasoline stations and each meteorological parameter, by varying one variable in the regression equation at a time while the other predictors were held constant. The maximum distances to roadways (2489 m for FC14, 3698 m for FC11) or gas stations (1010 m) measured for the study homes were used as the constant values for the proximity variables. Use of the maximum value should result in the concentration of a compound with a mobile source being at its urban background level. The median values were used as the constant meteorological variable estimates to provide typical conditions (Table 1). The predicted residential ambient air concentrations with distance to major arterial roadways (FC14), interstate highways (FC11), gasoline stations, or dry cleaning facilities are illustrated in Figure 2. The predicted residential ambient air VOCs concentrations decreased with increasing distances from the emission sources. The changes in concentration of aromatic compounds above the urban background levels, which is dominated by mobile sources for BTEX and MTBE in the urban atmosphere, are related to distance to the major arterial roadways (FC14) and gas stations. For the major roadways and gas stations, the model predicts a rapid decline in concentration during the first 250 m with little further change above the urban background levels beyond that distance. Thus, the additional contribution by mobile source emissions above urban background ambient air concentrations is predicted to occur only at homes very close to the sources, though within that distance changes of several µg/m3 are predicted, dependent upon the strength of sources 4078

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 13, 2006

for each compound. Rapid decreases in VOCs concentration with distance from roadways have been previously reported for major urban roads in a monitoring study conducted in London, UK and Paris, France (23). In that study, VOCs concentrations were approximately 5 times higher at 10 m than at 20 m from the centerline of roadways, with higher concentrations observed near roadways with more traffic. Predicted outdoor concentrations near the RIOPA study homes closest to the sources (13 m) under the median atmospheric conditions observed among all samples were 3.9, 2.0, 2.0, 1.8, and 1.5 times higher than the concentration predicted at 250 m from the major urban arterial roadways (FC14) for toluene, o-xylene, ethylbenzene, m,p-xylene, and benzene, respectively, with the benzene increase not being statistically different from zero at the 90% confidence level. Predicted concentrations for RIOPA study homes closest to gas stations (25 m) were 3.6, 2.0, 1.9, and 1.4 times higher than the predicted concentration at 250 m for MTBE, m,pxylene, benzene, and o-xylene, respectively. The predicted concentration of MTBE outside the home closest to the interstate highways (37 m) was 1.7 times higher than that predicted at 250 m, though not statistically significant at a 90% confidence level. The predicted PCE concentration at 50 m from a dry cleaning facility was 1.7 times higher than the PCE concentration predicted at 250 m. Other studies that have measured residential ambient air concentrations have also suggested an impact of emissions from roadways. Principal component analysis apportioned up to 53% of total variation of VOC concentrations to automobile emissions generated either locally in Helsinki, Finland or from long-range-transport (24). The residential outdoor air concentrations were significantly higher for m,pxylene, toluene, and ethylbenzene for homes near roadways with continuous traffic compared to homes near to roadways with infrequent traffic within the EXPOLIS study conducted in Helsinki, Finland (25). The residential outdoor air concentrations of benzene and total VOC were almost twice as high for homes near roadways with high traffic compared to homes near to roadways with low traffic in Amsterdam, The Netherlands (26). Janssen et al. (27) reported that the concentration of benzene and other air toxics outside of schools were significantly associated with distance to road-

FIGURE 2. Effect of the proximity to emission sources and the meteorological variables on the residential ambient air concentration of the selected VOCs estimated by each compound’s best-fitting models: (A) effect of distance to the major arterial roadways (FC14), (B) effect of distance to the gas stations, (C) effect of distance to the interstate highways (FC11) on MTBE concentration and the effect of distance to the dry cleaning facilities (DCF) on PCE concentration, (D) effect of the atmospheric stability, (E) effect of the temperature, and (F) effect of the wind speed. ways, traffic density, and percent of time the wind direction was from those roadways. The ambient air concentrations of aromatic compounds were significantly higher: 1.3-2.1 times near homes within 30 m to a gas station when compared to homes 60-100 m from a gas station (p < 0.05) (28).

Meteorological Effects. The predicted residential ambient air concentrations of all VOCs increased as the atmosphere became more stable, the temperature decreased, except for MTBE (Figure 2E), and wind speed decreased (Figure 2). One explanation for higher residential ambient air concentrations VOL. 40, NO. 13, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4079

TABLE 3. Population Estimated Living within Close Proximity to Emission Sources in the City of Elizabeth, New Jerseya distances from sources 300 m 250 m 200 m 150 m 100 m 50 m a

major roads 56,496 47,677 38,414 28,499 17,863 7,561

47% 40% 32% 24% 15% 6.3%

gas stations 41,812 30,512 19,830 11,197 4,471 859

35% 25% 16% 9% 4% 0.7%

major roads &gas stations

dry cleaning facilities

29,997 20,383 12,416 6,444 2,318 375

26,533 19,013 12,566 7,152 3,169 721

25% 17% 10% 5% 2% 0.3%

22% 16% 10% 6% 3% 0.6%

Citywide residential area ) 4.49 square miles; citywide population ) 120,568 (US Census 2000).

TABLE 4. Mean Concentration (µg/m3) of VOCs Measured in Ambient Air of Elizabeth, NJ studya

sample years

m,p-xylene

o-xylene

toluene

benzene

ethyl benzene

MTBE

PCE

1 2 3 4 5

1999-2001 1988-1989 1981-1983 winter, 1982 summer, 1981

2.97 4.6 8.9b 8.1c 5.5c

1.51 1.7 3.0b

6.73 13.6

1.41 4.6 7.6b 11.1 5.1

1.15

5.69

0.87 1.4 3.1b 4.5 3.5

20.7 18.1

3.0b

a Study 1: RIOPA NJ, Weisel et al. (31); 2: SI/NJ UATAP, US EPA (32); 3: TEAM NJ sampled in Elizabeth and Bayonne, Wallace et al., (6); 4, 5: ATEOS, urban air study, Lioy et al. (33), and Harcov et al. (34); all concentrations are arithmetic mean of total observations, unless otherwise specified. b Weighted median. c Xylene mixtures.

of BTEX compounds on lower temperature days than the higher temperature days is the increased exhaust byproducts from automobiles’ cold start and incomplete combustion. While higher temperatures are expected to increase the amount of gasoline evaporation, there are changes in gas composition to reduce the evaporation from summer gasoline. Further, meteorological conditions, such as increased mixing heights and reduced atmospheric stability, facilitate turbulence and mixing processes in the atmosphere during the summer. Risk Analysis on Proximity to Ambient Sources. The model predictions of residential ambient air concentrations of benzene and PCE based on proximity to sources were used to estimate the lifetime excess cancer risk for an individual exposed to these concentrations for a 70-year lifetime using the 95th percentile upper-bound unit risk estimate (µg/m3)-1 (29, 30). Benzene is a known human carcinogen and PCE is suspected carcinogen. If PCE is not a human carcinogen, the lower-bound cancer risk would be 0. The benchmark airborne concentration calculated for a lifetime upper-bound cancer risk of 10 × 10-6 is 1.28 µg/m3 for benzene, and 1.79 µg/m3 for PCE. Among the residential ambient air concentrations measured in Elizabeth during RIOPA, 47% and 7% exceeded the risk threshold benchmark concentrations of benzene and PCE, respectively. To assess the size of population living within close proximity to sources, the residential areas at distances of every 50 m from roadways, gas stations, and dry cleaners up to 300 m were calculated (Table 3). Elizabeth, NJ is a highly urbanized location within the NY/NJ metropolitan area resulting in a relatively high proportion of the population estimated to be living near emission sources. About 40% of the city population is estimated to live within 250 m from major roadways, while 15% of the population is estimated to live within 100 m of major roadways. The residential area within 250 m from both major roadways and gasoline stations was 17% of the citywide residential area. The high percentage of the population close to both sources is due to the locations of gasoline stations near major roadways. The residential area included within 250 m from dry cleaning facilities is estimated to include 16% of population in the city. Living very near roadways, gasoline stations, and dry cleaning establishments, within 25 m, appears to increase the lifetime upper-bound cancer risk from ambient benzene and PCE concentrations by 50-200% compared to living 4080

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 13, 2006

further than 250 m. While not quantified in this analysis, other health endpoints should also be considered in a complete risk assessment from exposure to air toxics due to proximity to ambient sources when developing policy decisions to protect public health. The influence of proximity to sources on exposure did rapidly decline with distance, with the exposure to residents living in homes at approximately 250 m being indistinguishable from urban background values. It should be noted that the variability in exposures levels due to meteorological influences on background levels is greater than that due to proximity to sources. Further, since the contributions to air pollutants from mobile sources and dry cleaners affect both exposure close to these sources as well as being the major contributors to the overall urban background air toxic levels, steps taken to decrease source emissions to lower the background levels will also decrease the risk associated with living close to those sources. Simple comparison of means from many different studies must be integrated with caution due to the large short-term temporal and local spatial variations that can exist. However, the reported mean concentrations of VOCs measured in ambient air of or near Elizabeth, New Jersey were compared in Table 4 (6, 31-34). The weighted medians of m,p-xylene, o-xylene, and PCE in the TEAM study sampled in Elizabeth and Bayonne, New Jersey from 1981 to 1983 were almost 2 times higher than those observed in Elizabeth, New Jersey from 1988 to 1989. The benzene concentrations measured in the TEAM study in Elizabeth and Bayonne in the early 1980s (6, 7) was many times higher than the benzene concentration measured in RIOPA samples in Elizabeth two decades later. This may be evidence that the regulations in the Clean Air Act Amendment of 1990, which have greatly reduced the allowable benzene content of gasoline, have improved the air quality of New Jersey in 2000.

Acknowledgments We are grateful to all of the participants in the study. The authors acknowledge the hard work of all the students and technicians in the field and laboratories of RIOPA investigators and the hospitality of the RIOPA participants. Field and laboratory personnel included Shahnaz Alimokhtari, Krishnan Mohan, and Ingrid Blosiers. We thank Mr. Ray Papalski (NJDEP) for his valuable advises on emission inventory data, and Mr. Arnold Schmidt (Union County) for providing the up-to-date lists of local emission sources. This research was

supported by The Mickey Leland National Urban Air Toxics Research Center (NUATRC) (contract 96-01A/P01818769) and by The Health Effects Institute (HEI, contract 98-23-3). HEI is jointly funded by the USEPA (EPA: Assistance Agreement R828112) and automotive manufacturers. This data analysis was supported by the USEPA Office of Transportation and Air Quality (Contract 68-C-04-149). The contents of this article do not necessarily reflect the views of the Mickey Leland NUATRC, HEI, (and policies of) EPA or of motor vehicle and engine manufacturers. C.P.W., B.S.T., and J.Z. were supported in part by the National Institute of Environmental Health Center for Excellence (ES05022) and B.S.T. was supported in part by the NJ Agricultural Experiment Station. We appreciate three anonymous reviewers for their excellent comments and suggestions for improvements to this manuscript.

Supporting Information Available SAS output of the finalized model of m,p-xylene; residential area map generated for risk population estimate. This material is available free of charge via the Internet at http:// pubs.acs.org.

(10)

(11)

(12)

(13)

(14)

Abbreviations DOQQ

Digital Ortho Quad Quadrangle

EXPOLIS

Air Pollution Exposure Distributions of Adult Urban Populations in Europe

FC

functional class

NAD83

North American Datum 1983

NCDC

National Climatic Data Center

NEI

National Emission Inventory

NJDOT

New Jersey Department of Transportation

NOAA

National Oceanic and Atmospheric Administration

TIGER

Topologically Integrated Geographic Encoding and Referencing system

Literature Cited (1) Rosenbaum, A. S.; Axelrad, D. A.; Woodruff, T. J.; Wei, Y.-H.; Ligocki, M. P.; Cohen, J. P. National estimates of outdoor air toxics concentrations. J. Air Waste Manage. Assoc. 1999, 49, 1138-1152. (2) Suh, H. H.; Bahadori, T.; Vallarino, J.; Spengler, J. D. Criteria air pollutants and toxic air pollutants. Environ. Health Perspect. 2000, 108 (4), 625-633. (3) Edwards, R. D.; Jantunen, M. J. Benzene exposure in Helsinki, Finland. Atmos. Environ. 2001, 35, 1411-1420. (4) Hoffmann, K.; Krause, C.; Seifert, B.; Ullrich, D. The German environmental survey 1990/92 (GerES II): Sources of personal exposure to VOCs. J. Exposure Anal. Environ. Epidemiol. 2000, 10 (10), 115-125. (5) Gonzalez-Flesca, N.; Bates, M. S.; Delmas, V.; Cocheo, V. Benzene exposure assessment at indoor, outdoor and personal levels, the French contribution to the life MACBETH programme. Environ. Monit. Assess. 2000, 65 (1-2), 59-67. (6) Wallace, L. A.; Pellizzari, E. D.; Hartwell, T. D.; Sparacino, C. M.; Sheldon, L. S.; Zelon, H. Personal exposures, indoor-outdoor relationships, and breath levels of toxic air pollutants measured for 355 persons in New Jersey. Atmos. Environ. 1985, 19 (10), 1651-1661. (7) Wallace, L. A. The total exposure assessment methodology (TEAM) study: An analysis of exposures, sources, and risks associated with four volatile organic chemicals. J. Am. Coll. Toxicol. 1989, 8 (5), 883-895. (8) Wallace, L. A. Human exposure to environmental pollutants: a decade of experience. Clin. Exp. Allergy 1995, 25, 4-9. (9) USEPA. Technology transfer network, clearinghouse for inventories and emission factors (TTN CHIEF) for 1999, National emission inventory documentation and data - Final version

(15)

(16)

(17)

(18)

(19) (20) (21)

(22)

(23)

(24)

(25)

(26)

3.0; http://www.epa.gov/ttn/chief/net/1999inventory.html. (accessed date 10/4/2003) Weisel, C. P.; Zhang, J.; Turpin, B. J.; Morandi, M. T.; Colome, S.; Stock, T. H.; Spektor, D. M.; Korn, L.; Winer, A.; Alimokhtari, S.; Kwon, J.; Mohan, K.; Harrington, R.; Giovannetti, R.; Cui, W.; Afshar, M.; Maberti, S.; Shendell, D. The relationships of indoor, outdoor and personal air (RIOPA) study: Study design, methods and initial results. J. Exposure Anal. Environ. Epidemiol. 2005, 15, 123-137. Kwon, J.; Weisel, C. P.; Turpin, B. J.; Zhang, J.; Morandi, M. T.; Colome, S.; Stock, T. H. Organization of the RIOPA database, a large exposure database. J. Environ. Monı´t. Assess. 2006 (submitted). Naumova, Y. Y.; Eisenreich, S. J.; Turpin, B. J.; Weisel, C. P.; Morandi, M. T.; Colome, S. D.; Totten, L. A.; Stock, T. H.; Winer, A. M.; Alimokhtari, S.; Kwon, J.; Shendell, D.; Jones, J.; Maberti, S.; Wall, S. J. Polycyclic aromatic hydrocarbons in the indoor and outdoor air of three cities in the US. Environ. Sci. Technol. 2002, 36, 2552-2559. Naumova, Y. Y.; Offenberg, J. H.; Eisenreich, S. J.; Meng, QY.; Polidori, A.; Turpin, B. J.; Weisel, C. P.; Morandi, M. T.; Colome, S. D.; Stock, T. H.; Winer, A. M.; Alimokhtari, S.; Kwon, J.; Maberti, S.; Shendell, D.; Jones, J.; Farrar, C. Gas/particle distribution of polycyclic aromatic hydrocarbons in coupled outdoor/indoor atmospheres. Atmos. Environ. 2003, 37, 703-719. Reff, A.; Turpin, B. J.; Porcja, R. J.; Giovennetti, R.; Cui, W.; Weisel, C. P.; Zhang, J.; Kwon, J.; Alimokhtari, S.; Morandi, M.; Stock, T.; Maberti, S.; Colome, S.; Winer, A.; Shendell, D.; Jones, J.; Farrar, C. Functional group characterization of indoor, outdoor, and personal PM2.5: Results from RIOPA. Indoor Air 2005, 15, 53-61. Meng, Q. Y.; Turpin, B. J.; Korn, L.; Weisel, C. P.; Morandi, M.; Colome, S.; Zhang, J.; Stock, T.; Spektor, D.; Winer, A.; Zhang, L.; Lee, J. H.; Giovanetti, R.; Cui, W.; Kwon, J.; Alimokhtari, S.; Shendell, D.; Jones, J.; Farrar, C.; Maberti, S. Influence of ambient (outdoor) sources on residential indoor and personal PM2.5 concentrations: Analyses of RIOPA data. J. Exposure Anal. Environ. Epidemiol. 2005, 15, 17-28. Chung, C. W.; Morandi, M. T.; Stock, T. H.; Afshar, M. Evaluation of a passive sampler for volatile organic compounds at ppb concentrations, varying temperatures, and humidity with 24-hr exposures. 1. Description and characterization of exposure chamber system. Environ. Sci. Technol. 1999, 33 (20), 36613665. Chung, C. W.; Morandi, M. T.; Stock, T. H.; Afshar, M. Evaluation of a passive sampler for volatile organic compounds at ppb concentrations, varying temperatures, and humidity with 24hr exposures. 2. Sampler performance. Environ. Sci. Technol. 1999, 33 (20), 3666-3671. NOAA. Air resources laboratory, real-time environmental applications and display (READY) system. http://www.arl.noaa.gov/ ready.html (accessed date 7/8/2003). U.S. Census Bureau. Geography division. http://www.census. gov/geo/www/tiger (accessed date 4/8/2003). NJDEP. The bureau of GIS, image warehouse. http://njgin. nj.gov/OIT IW/index.jsp (accessed date 12/17/2003). NJDOT. The bureau of transportation data and development. http://www.state.nj.us/transportation/refdata (accessed date 4/18/2003). ESRI (Environmental Systems Research Institute). ArcScripts and ESRI’s global accounts system. http://arcscripts.esri.com (accessed date 4/10/2003). Jones, S. G.; Fisher, B. E. A.; Gonzalez-Flesca, N.; Sokhi, R. The use of measurement programmes and models to assess concentrations next to major roads in urban areas. Environ. Monit. Assess. 2000, 64, 531-547. Edwards, R. D.; Jurvelin, J.; Koistinen, K.; Saarela, K.; Jantunen, M. J. VOC source identification from personal and residential indoor, outdoor and workplace microenvironment samples in EXPOLIS-Helsinki, Finland. Atmos. Environ. 2001, 35, 48294841. Edwards, R. D.; Jurvelin, J.; Saarela, K.; Jantunen, M. J. VOC concentration measured in personal samples and residential indoor, outdoor and workplace microenvironments in EXPOLIS-Helsinki, Finland. Atmos. Environ. 2001, 35, 45314543. Fischer, P. H.; Hoek, G.; van Reeuwijk, H.; Briggs, D. J.; Lebret, E.; van Wijnen, J. H.; Kingham, S.; Elliott, P. E. Traffic-related differences in outdoor and indoor concentrations of particles and VOCs in Amsterdam. Atmos. Environ. 2000, 34 (22), 37133722.

VOL. 40, NO. 13, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4081

(27) Janssen, N. A. H.; van Vliet, P. H. N.; Aarts, F.; Harssema, H.; Brunekreef, B. Assessment of exposure to traffic related air pollution of children attending schools near motorways. Atmos. Environ. 2001, 35, 3875-3884. (28) Jo, W. K.; Moon, K. C. Housewives’ exposure to volatile organic compounds relative to proximity to roadside service stations. Atmos. Environ. 1999, 33, 2921-2928. (29) USEPA. Integrated risk information system (IRIS), health effects information used in cancer and noncancer risk characterization for the NATA. 1996, National-scale assessment. http://www.epa.gov/iris/subst/0276.htm (accessed date 9/3/2003). (30) California EPA. Office of environmental health hazard assessment. http://www.oehha.org/air.html (accessed date 9/3/2003) (31) Weisel, C. P.; Zhang, J.; Turpin, B. J.; Morandi, M. T.; Colome, S.; Stock, T. H.; Spektor, D. M. Relationships of Indoor, Outdoor and Personal Air (RIOPA). Part I. Collection methods and descriptive analyses; Health Effects Institutute Report Number 130 Part 1 and Mickey Leland National Urban Air Toxics Research Center Number 7; November, 2005.

4082

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 13, 2006

(32) USEPA. Staten Island/New Jersey urban air toxics assessment project report, vol. I (summary); USEPA/902/R-93-001a; USEPA: Washington, DC, 1993. (33) Lioy, P. J.; Daisey, J. M.; Atherholt, T.; Bozzelli, J.; Darack, F.; Fisher, R.; Greenberg, A.; Harkov, R.; Kebbekus, B.; Kneip, T. J.; Louis, J.; McGarrity, G.; McGeorge, L.; Reiss, N. M. The New Jersey project on airborne toxic elements and organic substances (ATEOS): a summary of the 1981 summer and 1982 winter studies. J. Air Pollut. Control Assoc. 1983, 33, 649-657. (34) Harcov, R.; Kebbekus, B.; Bozzelli, J. W.; Lioy, P. J. Measurement of selected VOCs at three locations in New Jersey during the summer season. J. Air Pollut. Control Assoc. 1983, 33, 11771183.

Received for review September 15, 2005. Revised manuscript received February 2, 2006. Accepted May 1, 2006. ES051828U