Environ. Sci. Technol. 2005, 39, 3261-3268
A Field Comparison of Volatile Organic Compound Measurements Using Passive Organic Vapor Monitors and Stainless Steel Canisters G R E G O R Y C . P R A T T , * ,† D O N B O C K , † THOMAS H. STOCK,‡ MARIA MORANDI,‡ JOHN L. ADGATE,§ GURUMURTHY RAMACHANDRAN,§ STEVEN J. MONGIN,§ AND KEN SEXTON| Environmental Outcomes Division, Minnesota Pollution Control Agency, 520 Lafayette Road, St. Paul, Minnesota 55155, Health Science Center at Houston, School of Public Health, University of Texas, Houston, Texas 77030, School of Public Health, University of Minnesota, Minneapolis, Minnesota 55455, and School of Public Health, Brownsville Regional Campus, University of Texas, Brownsville, Texas 78520
Concurrent field measurements of 10 volatile organic compounds (VOCs) were made using passive diffusionbased organic vapor monitors (OVMs) and the U.S. Federal Reference Method, which comprises active monitoring with stainless steel canisters (CANs). Measurements were obtained throughout a range of weather conditions, repeatedly over the course of three seasons, and at three different locations in the Minneapolis/St. Paul metropolitan area. Ambient concentrations of most VOCs as measured by both methods were low compared to those of other large metropolitan areas. For some VOCs a considerable fraction of measurements was below the detection limit of one or both methods. The observed differences between the two methods were similar across measurement sites, seasons, and meteorological variables. A Bayesian analysis with uniform priors on the differences was applied, with accommodation of sometimes heavy censoring (nondetection) in either device. The resulting estimates of bias and standard deviation of the OVM relative to the CAN were computed by tertile of the canister-measured concentration. In general, OVM and CAN measurements were in the best agreement for benzene and other aromatic compounds with hydrocarbon additions (ethylbenzene, toluene, and xylenes). The two methods were not in such good agreement for styrene and halogenated compounds (carbon tetrachloride, p-dichlorobenzene, methylene chloride, and trichloroethylene). OVMs slightly overestimated benzene concentrations and carbon tetrachloride at low concentrations, but in all other cases where significant differences were found, OVMs underestimated relative to canisters. Our study indicates that the two methods are * Corresponding author phone: (651) 296-7664; fax: (651) 2977709; e-mail:
[email protected]. † Minnesota Pollution Control Agency. ‡ University of Texas, Houston. § University of Minnesota. | University of Texas, Brownsville. 10.1021/es0497328 CCC: $30.25 Published on Web 03/24/2005
2005 American Chemical Society
in agreement for some compounds, but not all. We provide data and interpretation on the relative performance of the two VOC measurement methods, which facilitates intercomparisons among studies.
Introduction Measurements of volatile organic compounds (VOCs) in ambient air are of interest because of concerns about their possible health and environmental effects (1-3). Such measurements are now routinely made by regulatory agencies using a variety of techniques. The U.S. Environmental Protection Agency developed reference methods (including TO-14A and TO-15) for measuring speciated VOCs in ambient air (4, 5). These methods use stainless steel canisters, collection periods of 24 h, and laboratory analysis by gas chromatography and mass spectrometry. The methods are straightforward, but require significant planning and resources to implement, especially on a routine basis. Furthermore, canister-based methods are not well suited for certain applications, such as occupational and personal sampling. Passive samplers such as charcoal-based organic vapor monitors [OVMs; the term OVM is typically applied specifically to the 3M Corp. (Maplewood, MN) passive sampler] have been used for many years for sampling VOCs in occupational settings with relatively high concentrations (6, 7). They have recently grown in popularity for indoor and personal sampling because they are relatively inexpensive, nonintrusive, and more easily deployed than canisters or other active samplers and because analytical methods have been developed for determining low concentrations of VOCs sampled with passive samplers. Brown (8) discussed the theory and practical considerations involved in using passive, diffusive samplers, concluding that reliable measurements can be obtained by following recommended principles. Reports are now appearing in the literature on studies in outdoor, indoor, and personal air done with passive samplers. Initial results from several studies have demonstrated the ability to measure levels found in ambient air provided they are deployed for sufficient time. For instance, OVMs were used to measure concentrations of VOCs in Canadian homes (9). This study found that OVMs were suitable for measuring VOCs in homes and that the presence of indoor VOC sources had a greater influence on concentrations than ventilation or meteorological conditions. A German study (10) used OVMs to measure benzene, toluene, ethylbenzene, and xylenes at multiple locations within homes. The OVMs proved efficacious, reliable, and sensitive enough to detect higher concentrations in kitchens and living rooms than in bedrooms. A French study (11) measured indoor, outdoor, and personal benzene concentrations using passive air samplers (not OVMs). The methodology was found to be practical for widespread multisite monitoring in the urban environment. Sexton et al. (12), and Adgate et al. (13, 14) found that OVMs were appropriate for detecting differences among indoor, outdoor, and personal VOC concentrations. Given the increasing use of OVMs and other passive air samplers, it is important to understand the comparability of this methodology with active sampling methods. Unfortunately, relatively little work has been done comparing passive samplers with other measurement methodologies, especially under field conditions. Dobos (15) compared OVMs and lowflow, active sampling charcoal tubes for measurements of styrene and found a strong linear relationship between the VOL. 39, NO. 9, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3261
TABLE 1. Detection Limits (µg/m3), Number of Observations below Detection, and Precision Estimates canister (n ) 129)
total n ) 129 pairs
ADL
n e ADL
ne0
lab dup MRADb (%) (n ) 60)
benzene carbon tetrachloride p-dichlorobenzene ethylbenzene methylene chloride styrene toluene trichloroethylene o-xylene m/p-xylene
0.13 0.16 0.21 0.12 0.18 0.16 0.09 0.32 0.16 0.16
0 0 15 1 12 15 0 74 0 1
0 0 0 0 9 (6%) 3 (2%) 0 9 (6%) 0 0
3 5 34 5 7 1 2 10 5 4
OVM (n ) 142)a colocated MRADb,c (%) (n ) 53)
MDL
ADL
n e ADL
ne0
colocated MRADb (%) (n ) 13)
4 0 15 5 8 9 7 15 21 8
0.40 0.51 0.58 0.29 0.78 0.26 7.68 0.26 0.34 0.86
0.11 0.12 0.16 0.16 0.73 0.19 0.11 0.10 0.10 0.12
0 1d 107 12 127 94 27 72 0 4
0 0 60 (42%) 2 (1%) 26 (18%) 80 (56%) 26 (18%) 37 (26%) 0 2 (1%)
6 9 21 11 19 8 19 36 8 8
a The OVM measurements included 13 duplicates (colocated samplers). b Median relative absolute differencesdefined as the median of the ratios of within-pair absolute differences divided by the within-pair mean. c The canister colocated data were taken from a separate set of measurements made during the study period by the monitoring agency at a site not included in this study. The colocated site was operated in the same metropolitan area as the study sites, and the measured concentrations were similar in magnitude. d This one measurement of carbon tetrachloride was judged to be in error and was deleted from the analysis.
two methods, although the OVM styrene concentrations averaged 31% higher than those of the charcoal tubes. Pfeffer and Breuer (16) used a passive sampler to measure benzene, toluene, and xylenes near a cokery plant and found good agreement (R2 from 0.82 to 0.97) with an active sampling method (a commercial thermodesorption instrument with manual pumped 24 h sampling following the draft German standard). Chung et al. (17) exposed OVMs to known concentrations of nine VOCs under a variety of conditions in a test chamber. They found that OVM performance was compound specific and dependent upon temperature and humidity in addition to concentration. For most substances the OVMs showed a negative bias and were within (25% of the calculated chamber concentration. Stock et al. (18) compared OVM measurements with canister measurements at low temperatures for 12 target compounds. They found good agreement (correlations above 0.64) between the methods for eight compounds. For all but two VOCs the OVM measurements were biased low (mean bias ranged from +11.7% to -56.6%). The work described here was conducted within the context of a larger study of personal, indoor, and outdoor VOCs and particulate matter (12, 19). The study design called for concurrent outdoor VOC measurements using canisters and OVMs. This database of concurrent measurements is useful for comparing the two methods in ambient air under field conditions where relatively low concentrations are expected.
Experimental Section Monitoring stations were established in three communities in the Minneapolis/St. Paul metropolitan area. Sampling sites were located in residential areas away from localized industrial and roadway sources. A map of the sampling locations together with breakdowns of emissions sources for each pollutant can be found in Pratt et al. (20). VOC concentrations at each site were measured concurrently using two methods, the U.S. Environmental Protection Agency (EPA) Federal Reference Method consisting of stainless steel canisters, and charcoal-based diffusive samplers referred to as organic vapor monitors (OVMs, model 3500, 3M Corp., Maplewood, MN). Sampling was done during the spring (April to June), summer (June to August), and fall (September to November) of 1999. The canister (CAN) measurement methodology was described previously (3). Briefly, VOC samples were collected following the U.S. Federal Reference Method (TO-14A4) in evacuated, summa-polished, stainless steel canisters, twovalve model (Scientific Instrumentation Specialists, Moscow, 3262
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 9, 2005
ID). The canisters were deployed using a Xon Tech model 910A canister sampler housed in an enclosure that allowed heating during the cold season (Xon Tech, Inc., Van Nuys, CA). Evacuated canisters were filled by mass flow controllers over 48 h periods, which ran concurrently with the sampling period for OVMs. Sample analysis was done using a Varian Saturn model 2000 gas chromatograph/mass spectrometer (Varian, Inc., Palo Alto, CA). The analytical detection limit (ADL) was defined as the standard deviation of seven replicate analyses of a standard prepared to 5 times the estimated detection limit divided by the square root of n (i.e., 7) and multiplied by the Student’s t value appropriate for a 99% confidence level with n - 1 (i.e., 6) degrees of freedom:
ADL ) t(n - 1, 0.99)
[
1
n-1
n
∑ i)1
]
1/2
(xi - xj )2
where xi are the replicate analyses. The QA procedures including the method used to determine ADLs, are specified by EPA (4). Duplicate laboratory analyses (n ) 60) were highly comparable. Using imputation of half the ADL for nonpositive values, Pearson’s correlation coefficients were >0.94 for all analytes except p-dichlorobenzene (0.78), trichloroethylene (0.84), and carbon tetrachloride (0.88). We define the median relative absolute difference (MRAD) as the median of the ratios of within-pair absolute differences divided by the within-pair mean. The MRAD (see Table 1) was 0.91 for all analytes except trichloroethylene (0.53) and styrene (0.76). The MRAD (see Table 1) was 0.91 for all analytes except p-dichlorobenzene (0.86), trichloroethylene (0.47), and styrene (0.73). The MRAD (see Table 1) was
