Probabilistic Approach to Estimating Indoor Air Concentrations of

One potential source of volatile organic compounds is vapors from underlying .... The primary difference is that the EPA model uses estimated soil vap...
0 downloads 0 Views 4MB Size
Download PDF
Environ. Sci. Technol. 2011, 45, 1007–1013

Probabilistic Approach to Estimating Indoor Air Concentrations of Chlorinated Volatile Organic Compounds from Contaminated Groundwater: A Case Study in San Antonio, Texas JILL E. JOHNSTON* AND JACQUELINE MACDONALD GIBSON Department of Environmental Sciences & Engineering, Gillings School of Global Public Health, University of North CarolinasChapel Hill, CB 7431 Chapel Hill, North Carolina 27599, United States

Received June 21, 2010. Revised manuscript received October 11, 2010. Accepted November 23, 2010.

Thispaperdescribesaprobabilisticmodel,basedontheJohnsonEttinger algorithm, developed to characterize the current and historic exposure to tricholorethylene (TCE) and tetrachlorethylene (PCE) in indoor air from plumes of groundwater contamination emanating from the former Kelly Air Force Base in San Antonio, Texas. We estimate indoor air concentration, house by house, in 30 101 homes and compare the estimated concentrations with measured values in a small subset of homes. We also compare two versions of the Johnson-Ettinger model: one used by the Environmental Protection Agency (EPA) and another based on an alternative parametrization. The modeled mean predicted PCE concentration historically exceeded PCE screening levels (0.41 ug/m3) in 5.5% of houses, and the 95th percentile of the predicted concentration exceeded screening levels in 85.3% of houses. For TCE, the mean concentration exceeded the screening level (0.25 ug/m3) in 49% of homes, and the 95th percentile of the predicted concentration exceeded the screening level in 99% of homes. The EPA model predicts slightly lower indoor concentrations than the alternative parametrization. Comparison with measured samples suggests both models, with the inputs selected, underestimate indoor concentrations and that the 95th percentiles of the predicted concentrations are closer to measured concentrations than predicted mean values.

Introduction Urban residents spend 85-90% of their time indoors, where elevated concentrations of contaminants are commonly present (1). One potential source of volatile organic compounds is vapors from underlying plumes of contaminated groundwater (2-10). While detailed case studies are limited, indoor air with concentrations of contaminants above recommended exposure levels has been attributed to vapor intrusion from plumes of chlorinated volatile organic compounds (CVOCs) at several sites (3-7). This paper develops a probabilistic model, based on the Johnson-Ettinger algorithm (11), to characterize the current * Corresponding author phone: 919-843-5786; e-mail: jillj@ unc.edu. 10.1021/es102099h

 2011 American Chemical Society

Published on Web 12/16/2010

and historic exposure to TCE and PCE in indoor air from plumes of groundwater contamination emanating from the former Kelly Air Force Base in San Antonio, Texas. We model probability distributions for indoor concentrations of these contaminants in 30 101 private homes adjacent to the base. Our model characterizes the exposure potential separately for each home to account for small-scale variability in characteristics of the site that affect the potential for vapor intrusion. Johnson-Ettinger Model. Because of political, technical, and monetary constraints on directly monitoring indoor air quality in private homes, a mathematical screening tool is necessary to identify at-risk areas. The JEM is widely used for regulatory guidance on vapor intrusion in the United States. The JEM estimates the vapor attenuation ratio, R, a unitless parameter that relates the indoor air concentration to the concentration in the vapor phase at equilibrium with the contaminated groundwater: Cindoor ) R × Csource

(1)

where R is the vapor attenuation ratio, Cindoor is the contaminant concentration in indoor air (mass/volume), and Csource is the contaminant vapor-source concentration (mass/ volume). The JEM couples one-dimensional steady-state diffusion of volatile compounds through porous media with diffusion and advection through the building foundation. The output is intended to serve as a conservative screening-level estimate of the potential influence of groundwater contamination on indoor air and to identify sites for further testing. Important parameters that influence vapor intrusionsand are included in the modelsare soil characteristics, building characteristics (air exchange rate, foundation type, and volume), and pressure differentials between the indoor and subsurface environments. The JEM uses the following equation to estimate R (11):

( (

) ( ) ( )[ (

) )

eff QsoilLcrack Dtotal Ab exp eff QbuildingLt DcrackηAb R) eff QsoilLcrack Dtotal Ab + + exp eff DcrackηAb QbuildingLt

(

eff Dtotal Ab QsoilLt

exp

QsoilLcrack eff Dcrack ηAb

(2)

) ] -1

eff is total overall effective diffusion coefficient where Dtotal (cm2/s), Ab is the area of enclosed space below grade (cm2), Qbuilding is the building ventilation rate (cm3/s), Lt is the sourcebuilding separation (cm), Qsoil is the volumetric flow rate of soil gas into the enclosed space (cm3/s), Lcrack is the enclosed space foundation or slab thickness (cm), η is the fraction of eff is foundation surface area with cracks (unitless), and Dcrack the effective diffusion coefficient through the cracks (cm2/s). Comparisons between modeled and measured R values indicate that with reasonable input parameters the JEM can predict within 1 order of magnitude the expected actual indoor air concentrations (12, 13). While designed to be conservative, the JEM has under-predicted R, and thus indoor air concentrations, in certain cases (14-17). Nonetheless, a comparison with six other algorithms used in Europe found that the JEM was the most accurate and least conservative (10). Table 1 details results from studies that measured R in a variety of settings. The values range over several orders of

VOL. 45, NO. 3, 2011 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1007

TABLE 1. Comparison of Measured Vapor Attenuation Ratios Measured between Groundwater Vapor and Indoor Air Concentrations Site

r, low

Chemicala

vapor intrusion sites in Massachusetts (winter)

PCE (n ) 6) TCE (n ) 11)

CDOT-MTL Denver, Colorado 50 houses in Redfield, Colorado Colorado and California Stafford, New Jersey CDOT-MTL Denver, Colorado EPA draft Vapor Intrusion database (2008)

TCE (n ) 111) 1,1 DCE (n ) 50) CVOCS (5 sites) Benzene (n ) 1) TCE, 1,1- DCE, TCA (n ) 7) VOCs (n ) 1,229)

r, high

r, average

Source

0.00013 0.00009

0.1 0.097

0.028 0.02

(16)

0.000001 0.0000048

0.00007 0.001 0.00077

0.000036 0.00005 (estimated) 0.00014 0.0000043

0.0000048 0.0000099

0.00034 0.074

(34) (5) (13) (35) (36) (37)

median: 0.00011

a BTEX: benzene, toluene, ethylbenzene, and xylene; PCE: tetrachloroethene; TCE: trichloroethene; DCE: dichloroethene; TCA: trichloroethane; n: number of sites sampled.

magnitude and indicate large differences not only between sites but also between buildings at the same site. However, current federal guidance does not require the characterization of variability and uncertainty in inputs to the JEM and thus fails to consider the variability observed in empirical studies. Stochastic modeling can better represent both the uncertainty and variability associated with predicted indoor air concentrations. Two previous studies have utilized Monte Carlo simulations to account for the uncertainty and variability of the parameters used to estimate indoor air concentrations and indicated reasonable agreement with measured values (14, 15). This study expands on previous work by characterizing uncertainty and variability in predicted indoor concentrations at the scale of individual houses across an entire community. Site Description. The case study site is a low-income neighborhood adjacent to Kelly Air Force Base. The neighborhood overlies extensive plumes of PCE and TCE in groundwater emanating from the base. Prior to remediation, these plumes extended 5 miles to the southeast of the base and occupied 12 square miles (18, 19). In 2008-2009, the EPA evaluated a small cohort of houses for vapor intrusion. Prior to sampling, EPA staff searched for and removed potential indoor sources of the contaminants. The EPA then verified that all indoor sources had been removed by scanning each home with a real-time trace atmospheric gas analyzer (TAGA). After confirming the removal of indoor sources, the EPA placed SUMMA canisters in the homes for 24 h. Afterward, EPA analyzed these canisters for TCE and PCE following the EPA Modified Method TO-15. The detection limits were 0.04 ug/m3 and 0.14 ug/m3 for TCE and PCE, respectively. TCE and PCE levels in some homes exceeded the EPA human health screening levels. However, the sample of homes tested was too small to represent the communitywide exposure or to characterize the historic risk.

Modeling Method This analysis uses site-specific data coupled with information available through the literature to develop probabilistic versions of the JEM that produce probability distribution functions for the vapor attenuation coefficient and the predicted indoor air concentrations of PCE and TCE in 30 101 homes in the case study area in both 1998 and 2007. Contaminant Source Data. Concentrations of TCE and PCE in groundwater were obtained from periodic measurements at 913 wells (a total of 3436 and 3876 samples for TCE and PCE, respectively) from 1997-2007 as reported by the DoD Air Force Real Property Agency (AFRPA). Groundwater elevations (7380 readings) also were obtained. We used these data, along with the Bayesian Maximum Entropy (BME) method (of which kriging techniques are special cases) to estimate the concentrations of PCE and TCE in groundwater 1008

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 45, NO. 3, 2011

beneath each home in the two study years. Details of the BME method are provided elsewhere (20). Primary and Secondary Inputs to the JEM. All the inputs for the JEM algorithm (eq 1) were estimated using procedures and secondary inputs described in the 2004 EPA guidance document “User’s Guide for Evaluating Subsurface Vapor Intrusion into Buildings.” Soil Diffusion Properties. The dominant soil types underlying the study area were acquired from the U.S. Department of Agriculture Soil Conservation Service classification scheme. Every residential land parcel was assigned a soil type. If more than one soil type was present, the house was assigned the soil class that occupied the greatest area. Probability distributions for these soil parameters were assumed to be log-normal. Air Exchange Rate. Data from over 100 homes at the Houston, Texas, site in the Relationship of Indoor, Outdoor, and Personal Air (RIOPA) study served as a proxy for air exchange rates at the case study site (21, 22). The RIOPA data set is a suitable substitute for site-specific air exchange rates because the case study site and RIOPA site lie on a similar latitude, share a similar number of days of extreme heat and freezing temperatures, and have similar demographic characteristics. Alternative Version of the JEM. While the EPA draft guidance on vapor intrusion employs the version of the JEM in eq 2, Johnson suggested an alternative algorithm to express the R coefficient that may reduce uncertainty, misuse, and inconsistencies in the screening model. The algorithm can be reduced into three dimensionless groups, simplifying the equation to three primary parameters (23): R)

Aexp(B) exp(B) + A + (A/C)[exp(B) - 1]

A) B)

(3)

eff Dtotal Ab

Eb(Vb/Ab)Lt (Qsoil/Qbuilding)Eb(Vb/Ab)Lcrack

(4)

eff Dcrack η

C ) (Qsoil/Qbuilding) where Eb is the indoor air exchange rate (1/h), Vb is the house volume (cm3), and all other variables are as previously defined. Comparison of the Two JEM Versions. For this case study, the EPA guidance method (hereafter referred to as the EPA model) and the reparameterized JEM (hereafter referred to as the JEM alternative) are calculated independently and compared. The primary difference is that the EPA model uses estimated soil vapor permeability factors and pressure differentials to calculate Qsoil, while the JEM alternative uses

FIGURE 1. Predicted spatial distribution of PCE in 1998. (a) Estimated PCE groundwater concentrations (ug/L); (b) Predicted vapor attenuation ratio of PCE by home; (c) Mean estimated indoor air concentration (ug/m3) of PCE by home; (d) 95th percentile predicted PCE indoor air concentration (ug/m3) by home. a generic Qsoil/Qbuilding distribution. Johnson (23) proposes that the literature offers better insight into reasonable values for this ratio than calculating individual flow rates. Simulation Method. The JEM algorithm and input data were encoded in a simulation model using Analytica software. The Supporting Information provides details on probability distributions used to represent each of the input variables (which were specified separately for each house to reflect house-to-house variation). To generate probability distributions for the R values and indoor air concentrations of TCE and PCE, we used the Latin hypercube method to sample from the input distributions and conducted 800 simulations for each house.

Results and Discussion Vapor Attenuation Ratios. The range of predicted R values, averaged over all the homes, is typical of observations at previous sites (Table 1) and spans just over an order of magnitude. Bimodal clustering is present; a group of homes has mean R values around 10-5 for both PCE and TCE, whereas a larger segment has mean values closer to 4 × 10-5 (Figure 2). In general, R for TCE is slightly higher than for PCE, which can be attributed to TCE’s chemical properties, including higher diffusivities in both air and water. Groundwater Concentrations. Figure 1a shows the spatial distribution of PCE in groundwater; TCE exhibits a similar spatial pattern. Groundwater concentrations range over 6 orders of magnitude, with peak concentrations at 72 000 µg/L and 54 800 µg/L for PCE and TCE, respectively. The highest concentrations are to the northeast of the former base. As expected, the concentration decreases as the groundwater moves farther from the base. Indoor Air Concentrations. As suggested by previous studies, this model indicates a great deal of variability in the

predicted indoor air concentrations across the study areasvariability that is not adequately captured with a single point estimate for the entire region. The range for both chemicals extends over 2 orders of magnitude. While the majority of the predictions cluster around low levels, there are still a significant number of homes with predicted concentrations above EPA Region 6 screening-level values. The indoor air screening level threshold established for PCE is 0.41 µg/m3 (based on a 1 in 106 increase in risk for cancer). The threshold for TCE is still under review, but the latest research indicates a level of 0.25 µg/m3 for a 1 in 106 increase in cancer risk (primarily due to kidney cancer) (24). Table 2 shows the predicted number of homes exceeding the threshold values. These estimates use annual means for the indoor air exchange rate. Also shown are estimates using the EPA model with the lower air exchange rates expected in summer (see Supporting Information). This summer scenario provides a worst-case estimate of contaminant levels in indoor air. Table 2 shows that, historically, at least half the homes are predicted to have had mean estimated TCE concentrations above the threshold, and at least 1,600 homes had PCE concentrations above the threshold. Indoor concentrations for PCE and TCE may remain elevated in thousands of homes, although decreases in exposure have occurred due to ongoing groundwater remediation. Figure 1c and d shows the spatial distribution of the predicted mean and 95th percentile values of the indoor concentrations for PCE. The distribution is similar for TCE. The spatial analysis suggests several correlations. First, regions with the highest groundwater concentrations indicate higher indoor air concentrations as seen closest to the perimeter of the base. However, the overall indoor air trend did not follow the groundwater trend, suggesting that VOL. 45, NO. 3, 2011 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1009

TABLE 2. Predicted Number of Homes to Exceed Screening Levels by Chemical and Year 1998 3

PCE (homes above 0.41 ug/m )

mean

95 percentile

mean

95th percentile

EPA EPA Model JEM Alternative Summer EPA Model

1644 (5.5%) 2079 (6.9%) 10261 (34.1%)

25688 (85.3%) 26260 (87.2%) 29099 (96.7%)

880 (2.9%) 1184 (3.9%) 3730 (12.4%)

20740 (68.9%) 21823 (72.5%) 27193 (90.3%)

TCE (homes above 0.25 ug/m3)

mean

95th percentile

mean

95th percentile

EPA Model JEM Alternative Summer EPA Model

14859 (49.4%) 16419 (54.5%) 26140 (86.8%)

29992 (99.6%) 30007 (99.7%) 30100 (99.9%)

3469 (11.5%) 3833 (12.7%) 8065 (26.8%)

23595 (78.4%) 24583 (81.7%) 28917 (96.1%)

groundwater concentrations alone are not a sufficient predictor of exposure. Second, regions with more sandy soil have higher estimated indoor air concentrations. The combination of either of these factors with a shallow groundwater table also increases the likelihood of higher indoor levels of the CVOCs. The strip along the eastern border of the study region showed the largest concentration of at-risk homes despite its distance from the contaminant source, likely due to the presence of sandy loam soil, which has higher soilvapor permeability than the clay-rich soil typical of the area. Thus soil type appears to be a strong indicator of vapor intrusion risk, a finding echoed in previous studies (13, 25, 26). Variations in soil characteristics are known to contribute greatly to the output uncertainty related to solute transport (27). Based on previous studies at Kelly Air Force Base (28), soil type is known to vary greatly, so using U.S. Department of Agriculture soil classifications as a proxy for soil type directly beneath the foundation of each home may not be appropriate. Further, soil may be altered significantly (including by importing of soil and/or gravel) during construction. The clay loam and sandy clay loam soil on average have a vapor attenuation ratio that is almost an order of magnitude higher than for the clay and silty clay varieties (3.5 × 10-5 vs 5 × 10-6). Garbesi et al. (1989) found, however, that measured soil gas intrusion rates were actually higher than predicted and attributed this difference in part to the existence of highpermeability flow paths, a phenomenon not appropriately captured in current mathematical models (29). Comparison of Modeling Approaches. The modeling suggests that the JEM alternative produces slightly higher estimated R values and thus is more conservative than the EPA version (Figure 2). The JEM alternative does not require site-specific soil vapor permeability calculations, and the Qsoil to Qbuilding ratio represents the range of clays, loams, and sandy soils. Twenty to 30% of the Qsoil to Qbuilding ratios (depending on the soil texture) from the EPA model were below the lower end of the expected distribution (10-4). However, approximately three-fourths of the values fell within the distribution used in the JEM alternative. Summer Scenario. As shown in Table 2, a reduction in the mean air exchange rate to reflect conditions during the summer results in a 5-fold increase in the estimated number of homes at risk (at mean concentrations) for PCE and a doubling for TCE in 1998. This is further evidence that the air exchange rate is an important predictor of indoor air quality and suggests the need to develop more site-specific and house-specific estimates to improve upon the models’ predictive ability and identify homes at risk. Sensitivity and Uncertainty Analysis. Since the JEM response to parameter changes is nonlinear, we analyzed how R changes when each uncertain variable is set at its 5th and 95th percentile value. The results shown in Figure 3 (for PCE in 1998) indicate that the air exchange rate has the largest 1010

2007

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 45, NO. 3, 2011

th

influence on R, explaining over 90% of model uncertainty. A lower air exchange rate results in a higher indoor concentration (increased exposure) because less air flows through the space to dilute the indoor or subsurface emission source (30). In fact, Pennell et al. (25) argue that R, with its reliance on air exchange rate, is a poor substitute for measuring vapor intrusion potential. As Figure 3 shows, other important variables include the total effective diffusion, airfilled porosity, and effective diffusion coefficient, in part due to the uncertainty of these parameters as well as their importance in the model. The model is relatively insensitive to building foundation properties, a result similar to that found in other analyses (13, 23). These results agree with Johnson’s findings that the JEM’s most critical inputs are Eb (air exchange rate), Deff (effective diffusion coefficient, which includes air-filled porosity as an input), and depth to groundwater (Lt), which is a primary input in calculating total effective diffusion. Based on the parametrization by Johnson (23) it is possible to determine the relative importance of different processes that influence R at the site. Our analysis suggests that advection is the dominant mechanism of transport across the foundation and that diffusion through the soil is the overall rate-limiting process. Comparison with Measured Values from an EPA Study. As mentioned in the introduction, in May 2008 and February 2009 the EPA collected PCE and TCE samples in a small number of homes (15 for PCE and 21 for TCE). Several studies have shown that even when vapor intrusion is occurring, indoor contaminant and ambient background sources can also significantly affect indoor air quality (6, 31, 32). As

FIGURE 2. Comparison of the vapor attenuation ratio of the EPA and the JEM Alternative model.

FIGURE 3. Sensitivity analysis of uncertainty variables for predicted r in EPA JEM model. The diagram shows how r changes as each uncertain input is changed from its lower 5th percentile to its upper 95th percentile value.

FIGURE 4. Comparison of modeled and measured indoor air concentrations. (a) PCE measured values compared with EPA model (2007); (b) PCE measured values compared with JEM Alternative model (2007); (c) PCE measured values compared with EPA model, summer air exchange rate (2007); (d) TCE indoor air concentration with EPA predicted model values in 2007. described in the introduction, the EPA removed all indoor sources prior to sampling. In addition, ambient air samples taken outside each home showed no detectable TCE or PCE. Therefore, it is reasonable to assume that the TCE and PCE detected indoors were exclusively from vapor intrusion. Figure 4 compares the concentrations measured by EPA with the corresponding predictions in each home by the EPA model and JEM alternative. As shown, both the EPA model and JEM alternative, given the inputs, consistently under-predicted measured values for PCE, although the measured value typically was within the 90% confidence

interval of the predicted value. On average, the predicted concentrations of PCE are five times lower than measured values. The use of summer air exchange rates rather than year-long average rates in the model resulted in the best match for the measured PCE data. As Figure 4d shows, the JEM demonstrated increased accuracy for TCE. The EPA model with selected inputs under-predicted the TCE concentration by approximately 130%, on average. The JEM alternative and summer air exchange rate scenarios generally predicted slightly higher than measured values for TCE. Since TCE concentrations in groundwater were generally lower VOL. 45, NO. 3, 2011 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1011

than the corresponding PCE concentrations, the improved accuracy of the predictions for TCE, in comparison to TCE, is consistent with previous findings that the JEM is more accurate in predicting lower than higher concentrations (15). In general, for both PCE and TCE the prediction was within 1 order of magnitude in accuracy, as in previous studies. This comparison, though limited by the quantity of measured data, points to two insights. First, local characterization of soil type may be needed to improve predictions using JEM. In this analysis, parameters for the soil properties were based on native soil type national data. These data fail to include changes in soil permeability and porosity due to construction, such as what might result from the addition of sand or gravel beneath the foundation. The soil data also do not capture preferential pathways, such as sewage pipelines or geological heterogeneities, that may be important sources of soil vapor intrusion. On-site measurements of soil properties, including permeability and better characterization of geological features in the community, would likely improve the predictive capacity of the model. Other studies suggest that this algorithm likely predicts lower than actual water-filled porosity in soil, resulting in conservative diffusion estimates (13, 33). Second, while this model improves on the spatial resolution of the screening prediction and decreases the uncertainty associated with groundwater levels and contaminant concentrations, the air exchange rate also is a critical variable. In order to refine a house-by-house approach, it is necessary to develop better tools for estimating or local measurements of the indoor air exchange rate to better characterize household risk. Implications of this Research. This analysis presents a framework by which to identify houses most at risk to the vapor intrusion exposure pathway. Our approach is more spatially detailed than the traditional community-averaged estimates. In this case, the model, given the inputs used, identified an area situated at the farthest edge of the groundwater plume at highest risk to the vapor intrusion pathway, houses that would be deemed low risk under traditional methods. The probabilistic output provides a more complete picture that incorporates uncertainties and site variability. Compared with the known measured values, this model seems to systemically underestimate high exposures, at least in homes atop fine-grain soils. Results suggests that use of the 95th percentile of the predicted value may more accurately screen for potential indoor air exposure risk. The JEM alternative, which requires fewer inputs, particularly in terms of soil properties, closely resembles the EPA output and is in this case slightly more conservative and closer to known indoor air values using simpler and fewer inputs. This work reiterates the importance of identifying appropriate soil parameters as well as demonstrating the need to more accurately characterize air exchange rates in order to develop a tool to analyze house-by-house risk. Finally, this analysis concludes that homes above the shallow groundwater plume of PCE and TCE surrounding the Kelly base were historically and are currently at risk of vapor intrusion, particularly where the groundwater concentration is very high, the soil type has fewer fine particles, and/or the air exchange rate is low. This modeling approach may prove increasingly useful as more sites are identified as at risk for vapor intrusion. The probabilistic approach can better identify priority areas for further sampling more accurately than the currently used deterministic approach.

Supporting Information Available Details of the model inputs and equations are provided. This material is available free of charge via the Internet at http:// pubs.acs.org. 1012

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 45, NO. 3, 2011

Literature Cited (1) Klepeis, N. E.; Nelson, W. C.; Ott, W. R.; Robinson, J. P.; Tsang, A. M.; Switzer, P.; Behar, J. V.; Hern, S. C.; Engelmann, W. H. The National Human Activity Pattern Survey (NHAPS): A resource for assessing exposure to environmental pollutants. J. Exposure Anal. Environ. Epidemiol. 2001, 11, 231–252. (2) Schuver, H. Intrusion: Risks and Challenges. Air Waste Manage. Assoc. 2007, 6–9. (3) Renner, R. A case of the vapors. Sci. Am. 2002, 287, 27. (4) EerNisse, P. A.; Steinmacher, S. J.; Mehraban, T.; Case, J. D.; Hanover, R. Vapor intrusion mitigation system efficacy in residential communities surrounding Hill Air Force Base, Utah. 2009. (5) Folkes, D.; Wertz, W.; Kurtz, J.; Kuehster, T. Observed spatial and temporal distributions of CVOCs at Colorado and New York vapor intrusion sites. Ground Water Monit. Rem. 2009, 29, 70– 80. (6) McDonald, G. J.; Wertz, W. E. PCE, TCE, and TCA vapors in subslab soil gas and indoor air: A case study in upstate New York. Ground Water Monit. Rem. 2007, 27, 86–92. (7) Kliest, J. Relationship between soil contaminated with volatile organic compounds and indoor air pollution. Environ. Int. 1989, 15, 419. (8) Little, J. C.; Daisey, J. M.; Nazaroff, W. W. Transport of subsurface contaminants into buildings. Environ. Sci. Technol. 1992, 26, 2058–2066. (9) Ferguson, C.; Krylov, V.; McGrath, P. Contamination of indoor air by toxic soil vapours: A screening risk assessment model. Build. Environ. 1995, 30, 375–383. (10) Provoost, J.; Reijnders, L.; Swartjes, F.; Bronders, J.; Seuntjens, P.; Lijzen, J. Accuracy of seven vapour intrusion algorithms for VOC in groundwater. J. Soils Sed. 2009, 9, 62–73. (11) Johnson, P. C.; Ettinger, R. A. Heuristic model for predicting the intrusion rate of contaminant vapors into buildings. Environ. Sci. Technol. 1991, 25, 1445–1452. (12) Johnson, P. C. Identification of critical parameters for the Johnson and Ettinger (1991) Vapor Intrusion Model. Am. Pet. Inst. Tech. Bull. 2002, 17–38. (13) Hers, I.; Zapf-Gilje, R. Evaluation of the Johnson and Ettinger model for prediction of indoor air quality. Ground Water Monit. Rem. 2003, 23, 119–133. (14) Hers, I.; Zapf-Gilje, R.; Evans, D.; Li, L. Comparison, validation, and use of models for predicting indoor air quality from soil and groundwater contamination. Soil Sediment Contam. 2002, 11, 491–527. (15) Schreuder, W. A. Uncertainty Approach to the Johnson and Ettinger Vapor Intrusion Model.: Principia Mathematica: Lakewood, CO, 2006. (16) Fitzpatrick, N. A.; Fitzgerald, J. J. An evaluation of vapor intrusion into buildings through a study of field data. Soil Sediment Contam. 2002, 11, 603–623. (17) Mills, W. B.; Liu, S.; Rigby, M. C.; Brenners, D. Time-variable simulation of soil vapor intrusion into a building with a combined crawl space and basement. Environ. Sci. Technol. 2007, 41, 4993–5001. (18) Mittal, A. K. Groundwater Contamination: DOD Uses and Develops a Range of Remediation Technologies to Cleanup Military Sites.; U.S. General Office of Accountability: Washington, DC, 2005. (19) Installation Restoration Program: Phase II Stage 1 Field Evaluation, Kelly AFB, Texas: Radian Corp.: Texas, 1984. (20) Christakos, G.; Bogaert, P.; Serre, M. L. Temporal GIS: Advanced Functions for Field-Based Applications.: Springer: New York, 2001. (21) Meng, Q. Y.; Turpin, B. J.; Korn, L.; Weisel, C. P.; Morandi, M.; Colome, S.; Zhang, J.; Stock, T.; Spektor, D.; Winer, A. Influence of ambient (outdoor) sources on residential indoor and personal PM2. 5 concentrations: analyses of RIOPA data. J. Exposure Sci. Environ. Epidemiol. 2004, 15, 17–28. (22) Yamamoto, N.; Shendell, D. G.; Winer, A. M.; Zhang, J. J. Residential air exchange rates in three major US metropolitan areas: Results from the RIOPA Study 1999-2001. Indoor Air 2009, 20, 85–90. (23) Johnson, P. C. Identification of application-specific critical inputs for the 1991 Johnson and Ettinger vapor intrusion algorithm. Ground Water Monit. Remediat. 2005, 25, 63–78. (24) Charbotel, B.; Fevotte, J.; Hours, M.; Martin, J. L.; Bergeret, A. Case-control study on renal cell cancer and occupational exposure to trichloroethylene. Part II: Epidemiological aspects. Ann. Occup. Hyg. 2006, 50, 777–87.

(25) Pennell, K. G.; Bozkurt, O.; Suuberg, E. M. Development and application of a three-dimensional finite element vapor intrusion model. J. Air Waste Manag. Assoc. 2009, 59, 447– 460. (26) Tillman, F. D.; Weaver, J. W. Uncertainty from synergistic effects of multiple parameters in the Johnson and Ettinger (1991) vapor intrusion model. Atmos. Environ. 2006, 40, 4098–4112. (27) Carsel, R. F.; Parrish, R. S. Developing joint probability distributions of soil water retention characteristics. Water Resour. Res. 1988, 24, 755–769. (28) Final Semiannual Compliance Plan Report: January through June 2007, Former Kelly AFB, San Antonio, TX.; HydroGeoLogic, Inc.: San Antonio, TX, 2007. (29) Garbesi, K.; Sextro, R. G. Modeling and field evidence of pressuredriven entry of soil gas into a house through permeable belowgrade walls. Environ. Sci. Technol. 1989, 23, 1481–1487. (30) Chan, W. R.; Nazaroff, W. W.; Price, P. N.; Sohn, M. D.; Gadgil, A. J. Analyzing a database of residential air leakage in the United States. Atmos. Environ. 2005, 39, 3445–3455. (31) Hers, I.; Zapf-Gilje, R.; Li, L.; Atwater, J. The use of indoor air measurements to evaluate intrusion of subsurface VOC vapors into buildings. J. Air Waste Manage. Assoc. 2001, 51, 1318– 1331.

(32) McHugh, T.; Connor, J.; Ahmad, F. An empirical analysis of the groundwater-to-indoor-air exposure pathway: The role of background concentrations in indoor air. Environ. Forensics 2004, 5, 33–44. (33) Fischer, D.; Uchrin, C. G. Laboratory simulation of VOC entry into residence basements from soil gas. Environ. Sci. Technol. 1996, 30, 2598–2603. (34) Johnson, P.; Ettinger, R.; Kurtz J.; Bryan, R.; Kester, J. Migration of soil gas vapors to indoor air: determining vapor attenuation factors using a screening-level model and field data from the CDOT-MTL Denver, Colorado site. Am. Pet. Inst. Bull. 16, 2002. (35) Sanders, P. F.; Hers, I. Vapor intrusion in homes over gasolinecontaminated ground water in Stafford, New Jersey. Ground Water Monit. Remediat. 2006, 26, 63–72. (36) Johnson, P.; Ettinger, R.; Kurtz, J.; Bryan, R.; Kester, J. Empirical assessment of ground water-to-indoor air attenuation factors for the CDOT-MTL Denver Site. Ground Water Monit. Rem. 2009, 29, 153–159. (37) US Environmental Protection Agency. U.S. EPA’s Vapor Intrusion Database: Preliminary Evaluation of Attenuation Factors, Draft; U.S. Environmental Protection Agency: Washington, DC, 2008.

ES102099H

VOL. 45, NO. 3, 2011 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1013