The Redfield site is located in Denver, Colorado, along the southern edge of the Cherry Creek valley, at the “headwaters” of a bedrock channel that runs down the valley sideslope. The alluvial deposits within the channel range from coarse sand and gravels near the base to overlying fine sand, silt, and clay, and are covered by about 10 to 20 feet of silt and clay loess. Near the top of the valley sideslope, ground water generally flows through weathered claystone, siltstone, and fine-grained sandstone bedrock of the Denver Formation (Robson 1987) at depths of approximately 25 to 30 feet below ground surface (bgs). The bedrock channel begins to collect and control ground water flow further down the valley sideslope, with depths to ground water ranging from about 30 to 50 feet bgs. Near the bottom of the valley, ground water discharges from the channel deposits into alluvial terrace deposits at depths ranging from about 30 to 10 feet bgs.
Historical releases of 1,1,1-trichloroethane (TCA), trichloroethylene (TCE), and tetrachloroethylene (PCE) at the former Redfield facility impacted underlying soil, weathered bedrock, and shallow ground water. Investigations conducted between 1998 and 2001 by EnviroGroup Limited found that the resultant CVOC ground water plume extended more than 13,000 feet downgradient of the former Redfield facility and was up to 1200 feet wide and 70 feet deep in certain portions of the site. The principal compound of concern was 1,1-dichloroethene (1,1-DCE), a breakdown product of TCA and, to a lesser extent, TCE. 1,1-DCE concentrations in ground water were approximately 3000 μg/L at the site boundary, decreasing with distance downgradient of the facility.
Monitoring wells were typically constructed of 2-inch I.D. schedule 40 flush jointed, threaded PVC with a screen slot size of 0.020 inch. Screen lengths were typically 10 feet and installed to straddle the water table (and collect samples representative of the upper few feet of the aquifer); although in a few cases the water table fell slightly below or rose slightly above the screened interval. It should be noted that the Redfield site is in a semi-arid region, with an average precipitation of approximately 15.8 inches per year (NWS 2006); therefore, only limited infiltration is available to create a clean water lens over the plume. Ground water samples were collected following standard industry practices and quality control procedures, under the oversight of the Colorado Department of Public Health and Environment (CDPHE), and were analyzed by EPA Method 8260b.
Indoor air samples were collected over a 24-hour period in 6-L Summa® canisters at more than 800 homes overlying or adjacent to the ground water plume to determine the extent of vapor intrusion impacts due to 1,1-DCE. The canisters were dedicated to the project and batch certified clean to the project reporting limits. Samples were collected from the lowest potential living space within the house, away from vents and windows. Houses with basements were sampled in the basement and homes without basements were sampled at a central location on the first floor. The indoor air samples were analyzed in accordance with EPA Toxic Organic Method TO-15 following CDPHE’s Guidance for Analysis of Indoor Air Samples (CDPHE 2000) since October 1998, using a mass spectrometer operated in the selective ion monitoring (SIM) mode with 1,1-DCE reported down to the method detection limit (varying from 0.011 to 0.04 μg/m3).
The single-family residential units in the vicinity of the Redfield ground water plume were constructed predominately between 1950 and 1965 (92.4%). A large majority (94%) are single-story homes ranging from 800 to 1750 square feet in area. Approximately 67.1% have basements, 31.5% have crawl spaces, 0.43% have slab-on-grade construction, and 1% are undefined. Most are heated by natural gas–heated warm air (86.4%), with the majority of the remainder being heated by gas-heated hot water (12.7%). Approximately half of the units have one or more fireplaces. Slightly more than half of the homes have swamp coolers or air conditioners.
CDPHE required the installation of subslab depressurization systems at 387 residences where the initial indoor air concentration exceeded the action level of 0.49 ug/m3 (the action level was increased to 5 μg/m3 in 2004). Periodic indoor air testing was required in mitigated homes (performance monitoring), typically on a quarterly basis for the first year, semi-annually for the second year, and annually thereafter. In addition, periodic indoor air testing was required in select homes around the edges of the mitigated area (verification monitoring), at frequencies ranging from quarterly to annually. Indoor air samples have been collected at annual or greater frequencies in a large number of homes for up to 10 years. Background sources of 1,1-DCE are relatively rare (Kurtz and Folkes 2002); therefore, most indoor air concentrations of 1,1-DCE are likely attributable to vapor intrusion.
Observed correlation between ground water and indoor air concentrations
The estimated boundary of the ground water plume (defined as the 7 μg/L MCL for 1,1-DCE) and the location of homes with 1,1-DCE above the action level of 0.49 ug/m3 in indoor air are essentially coincident at the Redfield site (Figure 1). No vapor intrusion impacts have been observed at distances greater than about one or two houses beyond the estimated extent of the ground water plume. Similar relationships have been observed during extensive vapor intrusion investigations conducted by one of the authors in Endicott, New York, and in Cortlandville, New York. The observed vapor intrusion impacts beyond the estimated plume boundary may be due to lateral diffusion of vapors in the vadose zone, which appears to be generally limited to a distance of about 100 feet (Folkes et al. 2007; Lowell and Eklund 2004), or they may be due to imprecision in the interpolated edge of the plume, or to impacts associated with the presence of 1,1 DCE at concentrations less than 7 μg/L.
Figure 1. Correlation between extents of ground water plume and vapor intrusion impacts due to 1,1-DCE, Redfield site. •= monitoring well, lines = 1,1-DCE groundwater concentration contour (μg/L), shaded homes = indoor air DCE above 0.49 μg/m3.
Download figure to PowerPoint
The initial indoor air 1,1-DCE concentrations (prior to mitigation) for each home sampled are plotted vs. interpolated ground water derived soil-gas concentration (ground water concentration multiplied by the Henry’s law constant [HLC] for 1,1-DCE) at the time of indoor air sampling on Figure 2 for 515 indoor air tests conducted between May 1998 and August 2003. The indoor air concentrations are observed to increase with ground water concentration, although the concentrations of 1,1-DCE were very low for a number of homes across the full range of ground water concentrations. Ground water attenuation factors ([1,1-DCE μ/m3] indoor air/[1,1-DCE μ/m3] ground water × HLC)] varied from approximately 10-3 to less than 10-6, consistent with observations at other sites (Hers et al. 2003). Although the plot indicates a positive correlation between ground water and indoor air concentrations, the correlation is weak. Setting aside temporal variability in building ventilation and transport processes over the 5-year period of monitoring, some of this imprecision may be due, in part, to ground water concentrations that are based on interpolation of concentrations between wells separated by distances of several hundred feet (not unusual for a site of this size) and use of paired ground water and indoor air data from sample dates that varied by several weeks to months. It is important to note that ground water concentrations were generally stable over the time period of the indoor air measurements, typically varying by less than a factor of 2 or 3 near the majority of the homes tested. The average distance between the homes included in Figure 2 and the nearest well is 220 feet. Filtering the data to include indoor air data only for homes within 200 feet of a well (average distance 126 feet) does not significantly affect the nature of the plot shown on Figure 2.
Figure 2. Ground water derived soil gas concentration versus premitigation indoor air concentration of 1,1-DCE, all data 1998–2003.
Download figure to PowerPoint
We attempted to reduce the imprecision introduced by ground water data interpolation and different sampling dates by selecting a subset of homes that (1) are within approximately 100 feet of wells, (2) have not been mitigated, (3) and have several sets of indoor air and ground water data collected within 90 days of each other over a period of several (2.5 to 7.5) years. Over this long time frame, ground water concentrations for several of these homes varied by approximately an order of magnitude. Plots of paired indoor air and ground water data for the five homes that met these criteria are shown on Figure 3. Again, a positive correlation between ground water and indoor air concentrations is noted (in composite, and for individual homes), with significant data scatter represented by ground water attenuation factors that range from about 10-4 to 10-6. Note that the ground water attenuation factors range by approximately one order of magnitude for several of these homes over time (see later section of paper).
Figure 3. Ground water derived soil gas concentration versus premitigation indoor air concentration of 1,1-DCE, five homes.
Download figure to PowerPoint
Other factors that could influence this correlation, or the ground water attenuation factor, include but are not limited to spatial variations in geology, temporal and spatial variability in vadose zone transport processes, depth to ground water, and building construction. We qualitatively examined the potential influence of these factors on the correlation between indoor air and ground water concentrations in a portion of the Redfield site where ground water concentrations and geologic conditions appeared to be relatively consistent, as shown on Figure 4. Indoor air concentrations, largely measured over a 3-month period, between homes within this area varied by up to three orders of magnitude (0.15 to 120 μg/m3), with variations of more than two orders of magnitude occurring between immediately adjacent homes (e.g., 0.15 vs. 21 μg/m3) that were sampled within days of each other. Depth to ground water, vadose zone geology, and variations in ground water concentrations over the time period that indoor air concentrations were measured are provided in Table 1 for the wells shown on Figure 4. Although depths to ground water are fairly consistent both spatially and over time, spatial variations in soil conditions in the capillary fringe and vadose zone may explain, in part, the spatial variation in indoor air concentrations compared to the more consistent (apparently) ground water concentrations.
Table 1. Ground Water and Soil Data, MW-35, -36, -38, -41 circa 2001 and 2002.
|Depth to GW, feet||25.5 ± 0.5||22.0 ± 0.4||22.2 ± 0.5||28.2 ± 0.3|
|Screen depth, feet||21.8–31.8||22.8–32.8||21.1–26.1||23.2–33.2|
|Vadose zone material (top to bottom)||–14 feet sandy to clayey silt||–14 feet silt to clay||–13.5 feet silt and clay||–11 feet silt|
| ||–4 feet sand|| ||–3.5 feet sand, f-c|
|–10 feet sand and clay||–4 feet clay||–2 feet sand, fine to clayey||–3 feet silt|
|–1.5 feet wx siltstone|| ||–7 feet m-c sand||–2.5 feet sand, f-m|
| ||–4 feet clay|
| ||–4.5 feet sand, f-m|
|DCE, μg/L(min, max, average)||120, 320, 223||200, 260, 224||58, 250, 138||72, 82, 78|
The spatial variability in the indoor air data compared to ground water data may also be due in part to differences in building construction and ventilation. Foundation type varied among the homes shown in Figure 4 between full basements, full crawl space, crawl space with basement, crawl space with slab on grade, and trilevel with basement (basement with slab on grade). However, no correlation between building style and indoor air concentration is readily evident on Figure 4. When examined statistically on a site-wide basis, we find a slight trend, with attenuation factors for slab-on-grade buildings being, on average, about half (i.e., twice the attenuation) the average attenuation factor for basements or crawl space homes (Folkes et al. 2004). Nevertheless, the range in attenuation factors calculated for each building type is similar and broad (about three orders of magnitude).
Temporal variability in parameters other than ground water concentration may contribute to the variability in attenuation factors observed between houses, as evidenced by the variation in attenuation factors at several individual homes over time of approximately one order of magnitude in Figure 3. Differences in ventilation (i.e., both depressurization and air exchange rate) may contribute more to spatial variability in the indoor air data than building construction (see below). This variability over time would result in a similar degree of imprecision in the spatial data when based on single indoor air tests conducted over one 24-hour period. The contributions of temporal changes in moisture content in the vadose zone are not known for the Redfield site.
Observed Temporal variations in indoor air concentrations
Indoor air monitoring has been conducted on a quarterly to annual basis at a number of unmitigated homes at the Redfield site, where detectable (but below action level) concentrations of 1,1-DCE have been routinely observed. These homes are generally located along the outside edge of the ground water plume, adjacent to areas with mitigated homes. For the purposes of evaluating the variability of vapor intrusion impacts over time, we selected 45 homes where indoor samples have typically been collected on a quarterly or semiannual basis for a period of approximately 2 to 10 years (average of 7.8 years per property), and where 1,1-DCE was detected above the reporting limit (0.04 to 0.011 μg/m3, depending on the time frame) at least 50% of the time.
The indoor air data set for the 45 monitored homes includes 715 individual indoor air measurements. Approximately 70% of the measurements were above the reporting limit. Values below the reporting limits were replaced with values equal to half the reporting limit. Although the inaccuracy in the standard deviations introduced by this substitution method is recognized, the statistical evaluations presented herein do not appear to benefit from the more rigorous substitution methods (e.g., the Kaplan-Meier method). The individual 1,1-DCE concentrations ranged from below reporting limits (see above) to 1.4 μg/m3. Average annual 1,1-DCE concentration in each of the 45 homes ranged from 0.023 to 0.27 μg/m3 (Figure 5). The data set is limited to this general concentration range because homes with significantly higher 1,1-DCE concentrations were typically mitigated.
Figure 5. Probability plot of annual average indoor air 1,1-DCE concentrations at 45 unmitigated properties, Redfield site.
Download figure to PowerPoint
The indoor air data were then normalized by dividing individual concentrations by the average annual concentration in that home, so that the variability in concentrations over time could be compared for all 45 homes on an equal basis. The average annual concentration in each home was calculated by first determining the seasonal arithmetic average 1,1-DCE values (i.e., winter, spring, summer, and fall) for each property. The seasons were defined as the winter months of December through February, the spring months of March through May, the summer months of June through August, and the fall months of September through November. The average annual 1,1-DCE concentration at each property was then calculated as the arithmetic average of the four seasonal values. The plotting position formula used in evaluating seasonal indoor air concentration values was 1 – (rank – 0.326)/(n + 0.348), which is reported to provide a distribution free plotting position (Yu and Huang 2001).
A log-probability plot of the normalized individual 1,1-DCE concentration values (i.e., values divided by the property average annual concentration) is provided in Figure 6, showing the overall log-linear distribution of the normalized observed values. The normalized values range from about 10% of the annual average to about 10 times the annual average concentration of the home being tested. However, most of the time (e.g., plus or minus one standard deviation, or 68% of the data), individual test results are within a factor of 2 or 3 of the average annual concentration.
Figure 6. Probability plot of indoor air 1,1-DCE concentrations divided by the property average annual concentration, Redfield site.
Download figure to PowerPoint
The seasonal trends were evaluated by separating the data by season and plotting the distribution of the normalized values for each season, as shown on Figure 7. A clear difference among seasons is observed, with median winter concentrations that were 15% higher than the annual average and median summer concentrations that were 52% lower than the annual average. Spring and fall median concentrations were 23% and 17% lower than the annual average, respectively. Although seasonal trends are observed, it is also important to note that nearly the full range of concentrations is observed in any given season, so that short-term variability can overwhelm any seasonal trend. Compared to the general population of single-family homes on the site, the 45 homes used in the seasonal evaluation have a slightly greater percentage of basements, a greater number of natural gas warm air heating systems, and a slightly smaller percentage of fireplaces.
Figure 7. Seasonal probability plots of indoor air 1,1-DCE concentrations divided by the property average annual concentration, Redfield site.
Download figure to PowerPoint
Finally, a plot of the average monthly values divided by the average annual values is shown on Figure 8. Also plotted are monthly values normalized to the annual mean for VOCs measured in apartments in Leipzig, Germany (Rehwagen et al. 2003), normalized monthly values for radon (based on 6-day tests) in England (Camplin and Henshaw 2004), and normalized monthly radon values in homes near Bialystok, northeastern Poland (Karpinska et al. 2004). The normalized trends for both VOCs and radon at all sites, including Redfield, are strikingly similar, with the lowest concentrations (i.e., less than the annual average) occurring from June through September and the highest concentrations occurring from November through April. Concentrations were closest to the annual average in April and October. Based on the group average for these sites, summer concentrations are, on average, about 20% to 40% less than the annual average concentration, and winter concentrations are, on average, about 20% to 50% higher than the annual average. More extreme seasonal variations have been observed in other studies; for example, monthly averages varied by up to a factor of 9 at a site in India (Ramu et al. 1992). Seasonal variations are likely to be related to local climate, ventilation practices (e.g., use of swamp coolers vs. air conditioners), and building construction; therefore, although the Redfield site data compared well with data from other temperate climate sites, including England and Poland, they may not compare as well to sites in more tropical climates (or more severe winter climates).