Research Concentrations and Emissions of Polybrominated Diphenyl Ethers from U.S. Houses and Garages S T U A R T A . B A T T E R M A N , * ,† SERGEI CHERNYAK,† CHUNRONG JIA,† CHRISTOPHER GODWIN,† AND SIMONE CHARLES‡ Department of Environmental Health Sciences, University of Michigan, Ann Arbor, Michigan 48109-2029, and Jiann-Ping Hsu College of Public Health, Georgia Southern University, Statesboro, Georgia 30460
Received October 24, 2008. Revised manuscript received February 7, 2009. Accepted February 12, 2009.
Concentrations of polybrominated diphenyl ethers (PBDEs) and other brominated flame retardants (BFRs) have been rapidly increasing in fish, birds, sediments, indoor environments, and humans, but emission sources and exposure pathways of these pollutants remain poorly understood. The many BFRcontaining materials in buildings constitute a large reservoir of these compounds, and in-use releases from this reservoir may be a significant environmental source. To estimate in-use releases from building materials and contents in residences, we monitored 12 houses and garages in two seasons and combined measurements of BFRs in air and settled dust, air exchange rates, and other information in an approach that utilized the building as a “natural” test chamber. Results were scaled to provide a first estimate of aggregate emission rates from U.S. houses. PBDE releases total about 4 µg h-1 per house or 20 ng m-2 h-1, and U.S. houses and garages collectively release about 4100 kg y-1. Most of these releases are settled floor dust, but about 20% are released directly to the ambient environment via airborne vapor and particulate matter. These screening-level estimates are subject to considerable uncertainty, but they have an advantage in that they reflect realworld conditions based on mass balance calculations.
Introduction The widespread incorporation of polybrominated diphenyl ethers (PBDEs) as brominated flame retardants (BFRs) in plastics, textiles, foams, and electronic equipment in houses, workplaces, and vehicle interiors has been associated with rapidly increasing and elevated levels of these pollutants in indoor environments, biota, and people, especially in North America (1). Emission sources and transport pathways of these chemicals are poorly understood. Their widespread occurrence suggests that these chemicals enter the environment at multiple product use and disposal stages, not just at the few sites where they are produced, e.g., Arkansas in the United States (2). Suggested release points include manufacturing facilities (2-5), recycling and disposal facilities (2, 6), wastewater treatment and sewage sludge facilities (8, 9), * Corresponding author e-mail:
[email protected]. † University of Michigan. ‡ Georgia Southern University. 10.1021/es8029957 CCC: $40.75
Published on Web 03/19/2009
2009 American Chemical Society
and storm drains (10), and from volatilization from products in use (3, 7), dust formed during the use of treated products (2), debromination of decabromodiphenyl ether (deca-BDE) (11), transport by ambient air (12), and global cycling (13). Aggregate emission estimates of PBDEs are very approximate, lack spatial resolution, and have other deficiencies (14, 15). In-use emissions are believed to be the dominant sources, at least for the more volatile congeners (3). Many studies have shown that concentrations of PBDEs in indoor air and dust greatly exceed levels outdoors, presumably due to the presence of PBDE-emitting materials and the limited dispersal and environmental degradation occurring indoors. Collectively, the many PBDE-containing materials in buildings constitute a large reservoir of these compounds (3, 16). An unknown portion of PBDEs from this reservoir will be released into the indoor environment and incorporated into dust, become airborne as vapor, or become absorbed on airborne particulate matter (PM). Five pathways are suggested for such releases from houses and garages: (1) housefoutside air: house materials/furnishings f airborne BFRs in house f direct airborne releases; (2) garagefoutside air: garage materials/contents f airborne BFRs in garage f direct airborne releases; (3) housefdust: house materials/ furnishings f settled dust in house; (4) garagefdust: garage materials/contents f settled dust in garage; (5) houseffilter, house materials/furnishings f airborne BFRs in house f capture on furnace filter. The first two pathways represent direct releases into ambient air. The dust and filter pathways (pathways 3 and 4 and pathway 5, respectively) represent indirect releases into settled dust or captured onto heating, ventilating, and air conditioning (HVAC) filters. These materials will be removed from the building by mopping, vacuuming, or replacing filters, and ultimately a portion of these removals will be released to the environment. This paper is aimed at quantifying accumulations and releases of PBDEs in houses. We studied 12 houses, garages, and vehicles in two seasons to gauge the significance of the air, dust, and filter pathways.
Experimental Section Emissions to Air and Dust. We utilized buildings as “natural” test chambers to characterize vapor and particulate emissions, an approach previously described for volatile organic compounds (VOCs) (17, 18). In the two-zone model (house and garage), the rate of emission of PBDEs from household BFR sources that enters the outdoor environment, SH (pg h-1 house-1), is Q |SS | ) |-Q H G
H HfG
| | |
-QGfH ∆CH × QG ∆CG
(1)
where QH ) air flow rate leaving the house to the outdoors (m3 h-1), CH ) BFR concentration in the house (pg m-3), and QGfH ) air flow from the garage to the house (m3 h-1) (Figure 1 and Supporting Information). Similarly, emissions from garage sources that enter the environment, SG, depend on the concentration in the garage (CG), air flows exiting the garage to the outdoors (QG), and house to garage air flow (QHfG). ∆CH and ∆CG denote that outdoor air concentrations are subtracted from the house and garage concentrations. These calculations assume that (1) BFR measurements are representative of the zone, (2) air is well mixed in each zone, (3) measured flows are representative, (4) concentrations are at a steady state, and (5) the sample is representative of VOL. 43, NO. 8, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2693
FIGURE 1. BFR concentrations in dust from houses, garages, and car interiors. The median and interquartile range is shown. houses and garages. Given these strong assumptions, only screening level estimates are possible. However, the mixing assumption is often reasonable, especially in buildings with forced air HVAC systems. Furthermore, to ensure representative measurements, 7-day averages of concentrations and flow rates are used. Finally, because ambient concentrations are generally much lower than indoor levels, we focused on accurate measurements of CH and QH. Emissions to settled dust, Sdust (pg y-1 house-1), are estimated as Sdust)Cdust × DL × A × CF
(2)
where Cdust ) BFR concentration in dust (pg g-1), DL ) dust loading (g m-2), A ) building area (m2), and CF ) cleaning frequency (times y-1 house-1). This calculation is performed separately for houses and garages. Key assumptions include (1) dust measurements are representative, (2) dust loadings and concentrations are uniform in the zone, (3) the cleaning frequency applies throughout the zone, (4) floors are relatively dust-free after cleaning, (5) other areas of dust accumulation are negligible, (6) concentrations in dust and air are at steady state, and (7) the sample is representative. These are stronger assumptions than those used to derive airborne emissions; thus, again the estimates are approximate. Loadings of BFRs onto the HVAC filters, SF (pg y-1 house-1), are derived as SF)CF × CPM × εF × QF × T × 10-6
(3)
where CF ) BFR concentration in filter dust (pg g-1), CPM ) indoor PM concentration (µg m-3), εF ) filter capture efficiency, QF ) filter air flow rate (m3 h-1), and T ) time when the HVAC system was operating (h y-1 house-1). The total emission value from BFR sources in the house is the sum of SH, SG, Sdust, and SF. Study Sites. We recruited owners of 12 private houses in southeastern Michigan and provided a descriptive recruitment letter, informed consent document, and a modest financial incentive. Houses were monitored in two seasons. Four participants dropped out, and follow-ups were not possible. Sampling was conducted from March 2006 to August 2007 at three locations in each house: indoors in the central family/living room, in the garage, and outside. Most houses were in suburban neighborhoods. All were single-family houses with attached garages, basements, and forced air heating systems; all but one had central air conditioning. House age varied from 4 to 50 years. Half of the houses were ranch-style (single-occupied floor); the other half had two 2694
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 8, 2009
floors. Garages shared at least one wall with the house, and a passage door permitted access. A walk-through audit was conducted to assess indoor air quality and factors potentially associated with PDBEs, e.g., presence of foam cushions and electronic equipment. The floor area was obtained from the homeowner and confirmed. Ceiling heights and building and garage dimensions were measured. Sampling and Analysis. Dust was vacuumed from representative and central areas, usually from the family/living room and garage and from the car parked in the garage. In the houses, a central 1 m2 area was vacuumed, usually in the family/living room, which in most cases was on the ground floor. In the garages, two or three smaller areas, totaling 1 m2, were sampled to account for different types and loadings of dust and to avoid obstacles. Floors and seats were vacuumed in vehicles. When possible, the same area was resampled on the second visit. Dust was collected using a modified and precleaned Oreck XL compact canister vacuum onto a precleaned and preweighed filter, which was weighed after conditioning at 33% relative humidity. At least one field blank was collected at each house. Air was sampled for a one-week period at three locations at each house: the family/living room, the garage (usually the far side from the garage door), and outdoors. The medium flow samplers (15 L min-1) used a precleaned polytetrafluoroethylene (PTFE) 1 µm filter, followed by a precleaned polyurethane foam (PUF) element (19). At least one filter and PUF field blank were collected at each house. After spiking with labeled PBDE and unlabeled PCB-65 and PCB-166 (polychlorinated biphenyls) surrogate standards, we Soxhlet extracted the filters and PUFs separately for 24 h using hexane and dichloromethane. They were then analyzed with gas chromatography-mass spectrometry (GC-MS) using selected ion monitoring, a negative chemical ionization mode, a 30 m DB-5 column, and a 2 µL splitless injection. A first run quantified tetrabromobisphenol-A (TBBPa) and 19 BDEs using a GC oven temperature of 80 °C for a 2 min hold and then a 10 °C min-1 ramp to 300 °C for 46 min hold. A second run quantified and confirmed BDE209 using a different temperature program, again starting at 80 °C for a 2 min hold and then a 50 °C min-1 ramp to 300 °C for a 40 min hold. Calibration standards included 20 BDE congeners, and internal standards included labeled PCB32 and PCB-138 (Cambridge Isotopes, Inc., Andover, MA). Dust was similarly extracted and analyzed. Method Validation and Quality Assurance. Each sample batch included blanks, linearity, drift (repeat analyses of the
standard injected into every fifth sample), and spike and surrogate recovery tests (19, 20). Linearity was within 15% over a wide (>1000-fold) concentration range. Surrogate and spike recoveries were 77-97%. Regular checks were made using standard reference materials (National Institute of Standards and Technology, Standard Reference Material 2585, Organic Contaminants in House Dust, and an integrated air extract derived from multiple long duration samples). Standards and reference materials were confirmed by the Analytical Center “Typhoon” Laboratory in Obninsk, Russia, and were within 10% of those expected. For most dust samples, the limited mass precluded duplicate extractions and analyses; however, duplicate samples analyzed during method development agreed within 20%. For air samples, a comprehensive assessment showed a wide dynamic range, minimal breakthrough for the sample volume used, little blank contamination on PUFs ( 0.75) were observed between BDE-28 and BDE-47, BDE-47 and BDE99, BDE-100 and BDE-85 and BDE-190, BDE-208 and BDE207, and BDE-207 and BDE-206 and BDE-209, following homologue orders and the composition of common PBDE mixtures. Similar groupings have been found in dust samples from Boston houses (22). Household dust elsewhere has shown elevated PBDE levels. Concentrations in Dallas, TX, houses averaged 12 µg g-1 (23). Washington, D.C., and Charleston, SC, houses had ∑BDE levels from 0.7 to 30 µg g-1; deca-BDE typically constituted >50% of the total and was associated with computers or new carpets (24). Three BDE congeners were reported in over half of Cape Cod, ME, houses (25). Our results were very close to the median concentrations of BDE47, BDE-99, and BDE-100 measured in California houses, although the California maxima were considerably higher (26). In 20 Boston houses, the geometric mean ∑BDE levels were 14 and 6 µg g-1 in living areas and bedrooms, respectively (maximum of 192 µg g-1) (22). Our median ∑BDE level in household dust, 21 µg g-1, is in the range reported for U.S. houses. We note the concentrations among the tested houses varied widely (1-290 µg g-1). Few studies have examined TBBPa in dust, but our house dust (and a lower concentration of car dust) measurements are comparable with a recent U.K. study (27). Measured dust loadings (median ) 0.21 g m-2, geometric mean ) 0.26 g m-2) appear typical. Essentially, the same loading was shown in a large number of Syracuse, NY, houses (geometric mean ) 0.31 g m-2) (28). Dust loadings are important for calculating emissions, which are described later. Recently, active PUF sampling of vehicle interiors in Greece showed a median ∑BDE concentration of 200 pg m-3 and a dominance of BDE-47, BDE-99, and BDE-209. The sampling train did not use a particle filter, although some particulate-phase PBDEs would have been collected (29). The elevated levels of BDE-209 in our vehicle samples are consistent with these results. Passive air sampling in 25 United Kingdom cars showed higher ∑BDE concentrations than in United Kingdom houses and offices, most of which was BDE99 and BDE-47 (BDE-209 was not reported) (30). Our dust measurements in cars are roughly comparable with those reported in the United States (32) and the United Kingdom (27), although, on the basis of the medians, we found lower ∑BDE levels (median of 28000 versus 57000 pg g-1) and more comparable levels of BDE-47, BDE-99, and BDE-209 than in the U.K. study. However, on the basis of averages, especially in newer cars, our samples were also dominated by BDE209. Airborne BFRs. In the vapor phase, we detected TBBPa, tri-BDEs, and some tetra-BDEs and penta-BDEs in most of the houses and garages, and occasionally more brominated congeners (Table 1 and Figure 2). No seasonal effects were observed, which was not surprising because indoor temperatures were relatively constant. Outdoor levels were much VOL. 43, NO. 8, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2695
TABLE 1. BFR Concentrations in Dust, Vapor, and Particulate Matter (PM)a dust in house (ng g-1) BFR
mean
median maximum
TBBPa 260 57 BDE-17 39 14 BDE-28 50 26 BDE-75 43 26 BDE-49 130 85 BDE-71 3100 65 BDE-47 6400 2000 BDE-66 860 49 BDE-100 5900 1200 BDE-99 11000 4600 BDE-85 650 130 BDE-154 430 190 BDE-153 230 110 BDE-138 21 5 BDE-166 860 18 BDE-183 840 220 BDE-190 520 170 BDE-203 1400 290 BDE-208 600 110 BDE-207 630 150 BDE-206 6400 64 BDE-209 11000 190 Total PBDE 49000 21000
2500 220 130 180 470 47000 46000 13000 78000 79000 7000 1300 790 160 14000 7600 4100 9700 2700 3000 110000 66000 290000
dust in garage (ng g-1) observed > MDL
mean
18 17 17 17 17 17 20 16 20 20 18 17 19 15 17 16 17 16 17 17 17 18 _
92 8 11 4 32 16 1300 54 290 1900 70 69 29 64 35 350 130 410 2800 4400 2600 200000 210000
median maximum
vapor in house (pg m-3)
16 2 3 1 18 7 160 6 39 280 7 7 6 2 3 8 7 22 31 32 18 7 2600
dust in car (ng g-1)
observed > MDL
940 43 74 18 130 80 7900 660 1800 11000 630 950 150 1000 230 3300 700 3800 40000 65000 32000 3900000 4100000
18 16 16 13 16 16 18 15 18 18 16 16 18 15 11 14 14 12 12 12 12 9 _
mean
median
maximum
observed > MDL
110 28 570 74 17 750 45 13 310 140 20 1700 3500 74 53000 7400 85 130000 5000 1800 30000 390 98 4400 1600 790 8100 9300 2600 63000 170 100 800 330 120 2000 1000 77 7200 180 11 2600 500 80 6500 2700 73 31000 6000 120 60000 8700 110 100000 72000 580 750000 130000 490 1,400000 81000 250 950000 15000000 3100 210000000 15000000 28000 210000000
vapor in garage (pg m-3)
19 18 17 18 17 17 19 17 19 19 18 18 19 16 18 17 16 17 15 14 13 16 _
vapor in outside air (pg m-3)
BFR
mean
median
maximum
observed > MDL
mean
median
maximum
observed > MDL
mean
median
maximum
observed > MDL
TBBPa BDE-17 BDE-28 BDE-75 BDE-49 BDE-71 BDE-47 BDE-66 BDE-100 BDE-99 BDE-85 BDE-154 BDE-153 Total PBDE
29 240 200 99 150 190 2500 51 180 760 10 20 32 4500
7 160 140 12 11 92 1400 3 110 410 1 6 10 3200
280 800 590 1400 1200 710 7200 250 970 4500 75 110 260 15000
12 14 15 12 9 15 20 7 20 20 6 8 10 -
68 40 59 12 46 95 1300 42 140 470 15 8 17 2500
4 14 25 4 28 43 430 5 31 140 1 4 6 1200
900 180 210 110 220 670 9400 380 820 4000 140 44 86 15000
10 9 11 6 10 10 19 9 15 18 6 5 6 -
9 91 99 5 26 27 150 8 21 74 2 4 9 710
1 5 5 3 4 14 170 3 17 59 1 4 5 550
78 1300 1800 23 130 110 300 42 67 310 13 10 44 3600
7 10 6 1 7 13 18 8 15 18 3 2 3 -
airborne PM in house (pg m-3)
airborne PM in garage (pg m-3)
airborne PM in outside air (pg m-3)
BFR
mean
median
maximum
observed > MDL
mean
median
maximum
observed > MDL
mean
median
maximum
observed > MDL
TBBPa BDE-17 BDE-28 BDE-75 BDE-49 BDE-71 BDE-47 BDE-66 BDE-100 BDE-99 BDE-85 BDE-154 BDE-153 Total PBDE
29 3 6 8 14 8 390 9 73 340 9 31 17 1200
13 1 4 4 5 2 110 3 26 110 2 11 7 530
150 11 25 36 99 46 1800 43 370 1500 44 230 110 3900
14 4 3 3 5 6 18 4 14 18 9 11 7 -
25 2 4 5 4 3 89 3 48 270 3 23 34 720
1 1 4 4 4 2 53 3 13 64 1 6 6 400
310 9 6 24 8 20 400 15 420 2900 23 310 360 4500
9 1 0 1 1 3 14 1 9 13 3 6 7 -
9 2 3 4 14 4 58 7 26 91 4 9 30 480
3 1 3 3 4 2 26 3 11 38 1 7 7 370
60 11 5 12 54 17 240 89 180 390 18 34 300 1300
12 1 0 2 6 5 12 1 8 11 4 9 8 -
a
Data collected from houses, garages, car interiors, and outdoors. Number of observations over MDLs is shown.
lower, and only tetra- and pentacongeners BDE-47, BDE100, and BDE-99 exceeded MDLs. Concentrations decreased from indoor to garage to outdoor microenvironments for TBBPa and most tri-BDEs through penta-BDEs. (BDE-49 and BDE-66 were below MDLs in about half of the measurements; 2696
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 8, 2009
the inconsistent gradient in Figure 2 should be ignored.) The variation in vapor concentrations, especially in garages, was large. The few studies measuring airborne BFRs in houses have reported a wide range of concentrations. In Ottawa, ON, the
FIGURE 2. BFR concentrations in airborne vapor in houses, garages, and outside air. The median and interquartile range is shown. median vapor-phase ∑BDE concentration was 100 pg m-3 (7). In Boston, ∑BDE concentrations ranged from 220 to 3500 pg m-3 in living rooms and from 170 to 1500 pg m-3 in bedrooms (34). In Europe, compositions differ and levels are generally lower (31, 32). Our vapor phase ∑BDE concentration (median of 3200 pg m-3, range of 340-15000 pg m-3) is at the high end of the literature. While we measured a few more congeners than most of the literature, this is not expected to cause a significant difference. Indoor levels of ∑BDE exceeded outdoor levels by about 8-fold, compared to two earlier studies reporting ratios of 15 and 50; however, these studies were conducted in the winter, when outdoor PBDE levels are low (7, 35). In ambient air, rural areas around the Great Lakes showed low concentrations (5 pg m-3) and higher levels in urban areas (52 pg m-3 in Chicago) (16). Seasonal trends and higher concentrations were noted at a southern Ontario site (90-1250 pg m-3) (12). Our ∑BDE measurements (medians of 514 and 382 pg m-3 in vapor and PM, respectively) are at the high end of previous measurements, probably reflecting the urban nature of the sites. In the particulate phase, airborne mass concentrations were low, typically 13 µg m-3. Several BFRs were detected in nearly all PM samples, including TBBPa and tetra- and pentacongeners BDE-47, BDE-100, and BDE-99 (Table 1). Occasionally other congeners were detected, most commonly hexacongeners BDE-154 and BDE-153, especially in garages. The few BFRs detected is a consequence of the little PM collected (1-2 mg), which increased MDLs. In the houses, the most prevalent BFR in airborne PM was BDE-47, and indoor concentrations exceeded outdoor levels by an order of magnitude. Hexacongeners and above did not show indoor-outdoor differences, which is a result of measurements at or near MDLs. In the garages, concentrations fell between indoor and outdoor levels. Indoor PM measurements across seasons showed moderate agreement, with somewhat lower temporal variation than dust or vapor. In the houses and garages, the interquartile range was about 10-fold; ambient levels had about half as much variation. Most studies measuring airborne BFRs indoors have used passive sampling, which does not effectively capture the particulate fraction. Moreover, few studies have detected the more brominated BDEs in indoor PM, despite its strong partitioning to PM (35), which is a consequence of high MDLs for these congeners (∼100 pg m-3). However, we have shown that deca-BDE is among the most predominant congeners in airborne PM when collected on ventilation system filters, which collect large quantities of PM and thus permit low
TABLE 2. Emission Rates for ∑BDE from Tested Buildings and Garages and U.S. Estimatesa test buildings (mg y-1 house-1) national estimate (kg y-1)
house garage total
air
dust
filters
total
air
dust
4.6 1.9 6.5
21.2 4.1 25.4
3.0 3.0
28.8 6.1 34.9
585 2715 137 292 722 3007
filters total 382 382
3682 429 4111
a On the basis of average across measurements, except for dust in the garage, which uses a median emission estimate.
MDLs (20). Therfore, our ∑BDE measurements are underestimates because they do not account for congeners with high MDLs. Mixing and Air Flows. VOC measurements indicated that the air was well-mixed in 18 of 20 house visits and in 17 of 20 garage visits. In the houses, AERs averaged 0.34 ( 0.18 h-1, a low value representing a relatively “tight” structure. Seasonal variation was not statistically significant (p ) 0.31, Wilcoxon signed rank test). In four visits, all during very cold or very hot periods, we found very low AERs (
