Shaibal
Mukerjee
*a,
Karen D.
Oliver
b,
Robert L.
Seila
a,
Henry H.
Jacumin Jr.
b,
Carry
Croghan
a,
E. Hunter
Daughtrey Jr.
b,
Lucas M.
Neas
c and
Luther A.
Smith
b
aNational Exposure Research Laboratory E205-02, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711, USA
bAlion Science and Technology, Inc., Durham, NC 27713, USA
cNational Health and Environmental Effects Research Laboratory, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711, USA
First published on 8th October 2008
This study evaluates performance of nitrogen dioxide (NO2) and volatile organic compound (VOC) passive samplers with corresponding reference monitors at two sites in the Detroit, Michigan area during the summer of 2005. Ogawa passive NO2 samplers and custom-made, re-useable Perkin-Elmer (PE) tubes with Carbopack X sorbent for VOCs were deployed under week-long sampling periods for six weeks. Precise results (5% relative standard deviation, RSD) were found for NO2 measurements from collocated Ogawa samplers. Reproducibility was also good for duplicate PE tubes for benzene, toluene, ethylbenzene, and xylene isomers (BTEX species, all ≤ 6% RSD). As seen in previous studies, comparison of Ogawa NO2 samplers with reference chemiluminescence measurements suggested good agreement. Generally good agreement was also found between the PE tubes and reference methods for BTEX species.
Ogawa passive samplers have been used in ambient air monitoring networks to monitor urban ozone, assess ozone, nitrogen dioxide (NO2), and sulfur dioxide trends at national parks and forested areas, determine distance of nitrogen dioxide emissions from highway traffic, and monitor personal/indoor air.4,5,8–10Ogawa samplers can measure these varying gaseous species depending on the reactants used on the collection filter substrate.1,5 For this study, Ogawa passive samplers were used for the collection of NO2.
Limited evaluations of the Perkin-Elmer (PE) tubes packed with various sorbents for sampling of volatile organic compounds (VOCs) have been performed in other studies in controlled exposure chamber and field settings.11–15 In terms of field comparisons, diffusive sampling PE tubes packed with Carbopack X sorbent have been found to be comparable to pumped methods and canister monitoring, with apparent over- and under-estimation of concentrations. A cost advantage in using these samplers is that they are re-usable for field deployment after thermal desorption (see methods section).
During the summer of 2005, a study was conducted by the U.S. Environmental Protection Agency (EPA) in the Detroit, Michigan area to measure NO2 and VOCs at elementary schools as part of an overall study assessing respiratory effects in children due to traffic and related urban air pollution. As part of the study, concentrations obtained with passive samplers were compared to NO2 and VOC measurements obtained with continuous monitors at two compliance sites. A motivation for using these passive samplers was their evaluation and application in a concurrent EPA study called the Detroit Exposure and Aerosol Research Study (DEARS).14,16 Also, similar methods were used in other cities (El Paso17 and Dallas, Texas) for planned intercity comparisons. We hypothesized that if the passive samplers were comparable in performance to reference measurements in the field, then this could provide further validation of passive sampling networks for monitoring community exposures in other locations on a cost-effective basis.
The first site was in a residential section of northeast Detroit called the East 7 Mile site (AQS Site 26-163-0019) at the corner of Linnhurst and Dresden Streets on the Bessy Playground and near Von Steuben Elementary School. The second site was in neighboring Dearborn, Michigan (AQS Site 26-163-0033) at 2842 Wyoming Avenue. This site was next to Paul Costea Park and Salina Elementary School, approximately 500 m from the River Rouge Industrial Complex and 200 m from an associated railroad yard.
As part of routine monitoring conducted by MDEQ, the East 7 Mile site measured NOx species (including NO2) on a continuous basis using a TECO Model 42C gas-phase chemiluminescent NOx analyser (Thermo Environmental Instruments, Franklin, MA); these data were compared to NO2 measurements from the passive samplers. The chemiluminescent NOx analysers used were Federal Reference Method (FRM) samplers (Designated Method RFNA-1289-074) in accordance with Title 40, Part 53 of the Code of Federal Regulations (40 CFR Part 53).18,19 All chemiluminescence data were reported in units of parts per billion by volume (ppbV) and above the method detection limit (MDL) of 5 ppbV.
Volatile organic compounds were continuously monitored by MDEQ at the East 7 Mile Site. The site is designated as an EPA Photochemical Assessment Monitoring Station (PAMS) to assess trends in ozone-related pollutants; an automated gas chromatograph (auto-GC) was established according to PAMS network monitoring protocol.19,20 VOCs were sampled hourly using a AutoSystem GC, and a Model ATD400 Automatic Thermal Desorption unit (Perkin-Elmer Life and Analytical Sciences, Shelton, CT) fitted with an air sampling accessory. This system cryogenically (−30 °C) pre-concentrated air on a proprietary dual bed sorbent trap at a flow rate of 15 ml min−1 for 40 min. The trap was heated to 320 °C to inject the cryogenically pre-concentrated sample onto a non-polar 50 m × 0.22 mm ID BP1 column (SGE, Austin, TX) which was connected through a Deans switch to a second column, a 50 m × 0.32 mm ID Al2O3/Na2SO4 (alumina) porous layer open tubular column (Varian, Palo Alto, CA). After passing through the BP1 column, light hydrocarbons (C2 to C4) were further separated on the alumina column and quantified by a flame ionization detector (FID). Activation of the Deans switch sent the heavy fraction (C5 to C12 hydrocarbons) directly to a second FID. The columns were temperature programmed from 46 °C for 15 min to 170 °C at 5 °C min−1, then to 200 °C at 15 °C min−1 held for 6 min. Total analysis time was 48 min. All auto-GC measurements were above the MDL (0.1 ppbC).
The MDEQ did not conduct VOC/NO2 sampling at Dearborn. Thus, VOCs were collected at this site in 32 L Summa-polished stainless steel canisters (BRC/Rasmussen, Hillsboro, OR) that were cleaned, evacuated, and subsequently sampled to one-half atmosphere at 1.5 ml min−1 using a Model 423 Series Precision Low Flow Controller (Veriflo Division, Parker Hannifin Corporation, Richmond, CA) over a one week period. Pumps were not used for canister sampling. Analyses of VOCs were performed using GC-FID. Volatile organic sampling and analysis was based on U.S. EPA Compendium Methods TO-14a.21 All VOC analyses with the canisters were above the MDL (0.1 ppbC). No reference NO2 measurements were performed at Dearborn.
Analysis of the tubes was performed using a TurboMatrix ATD and Saturn 2000 GC/MS system. Further details on the analytical method and experiments conducted to evaluate response of the selected VOCs under varying temperature and humidity conditions are presented elsewhere.13 All VOC samples from the PE tubes were field blank-corrected and reported in units of parts per billion carbon (ppbC) to match the original measurement units of the reference methods. The tubes were ready for re-use after thermal desorption and were re-deployed in the field.
The Model 3300 Ogawa Passive Samplers (Ogawa & Company, Pompano Beach, FL) were used for collecting NO2. This two-sided sampler consists of a cylindrical polymeric body (2 cm in diameter and 3 cm long) with a diffusion barrier and two stainless steel screens on each side. The device holds a glass-fiber collection pad coated with triethanolamine (TEA) at each end for sampling. The TEA-coated pads were loaded in the lab just prior to deployment to minimize contamination and degradation. All components, except the collection pad, are re-useable. Samples were field blank-corrected and reported in units of ppbV to match the original measurement units of the chemiluminescence analyser reported in AQS. The NO2 content was determined by an ion chromatograph (IC, Model 500, Dionex, Sunnyvale, CA) equipped with an IonPac™ AS-14 anion-exchange analytical column with 1.2 mL min−1eluent flow, a 50 µL sample loop, an anion self-regenerating suppressor (ASRS-1), and NaHCO3/Na2CO3-eluting buffers to quantify nitrite ion as a measure of NO2 to analyse the water extracts of the pads. Additional details on the Ogawa sampler for NO2 are discussed elsewhere.4
![]() | ||
Fig. 1 Week-long passively loaded PE tube series for benzene using 2 ppbv TO-14 VOC mixture in exposure chamber. |
To make the comparisons described here, hourly AQS data from the continuous samplers were averaged on the same time frames as the six passive sampling periods. Missing hourly values were not imputed; averages were calculated only from available data. Measurements from the continuous monitors are typically quality assured prior to being reported in AQS. The AQS data were required to exhibit a data completeness level of at least 75% (i.e., 75% of valid hourly data) within each of the sampling periods. The Dearborn canister sample on week 2 was not used owing to flow controller problems, thus yielding only five sampling periods for VOC comparison purposes at that site.
For both NO2 and VOC data collected, simple linear regressions of the passive sampling measurements on the averaged values from AQS data were utilized to assess the performance of the passive samplers at each site. Regression analysis was chosen because it is an efficient way to compare the two methods across the range of observed values. Given that duplicate sampling was part of the study design, precision error (as percent relative standard deviation) was also calculated. All statistical analyses were performed with SAS 9.1 software.24
Compound | East 7 Mile passivesa (N = 12)b | Dearborn passives (N = 12) | MDLc | East 7 Mile referenced (N = 6 for NO2; N = 5 for VOCs) | Dearborn referencee (N = 5) | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Median | Min | Max | %RSDf | Median | Min | Max | %RSD | Median | Min | Max | Median | Min | Max | ||
a Samples were field blank-corrected with data summarized based on pollutants below detection limit assigned a value equal to ½ the detection limit. b Numbers in parentheses are the number of observations designated at the top of site column. c MDL: method detection limit for passive samplers. d Reference methods at E-7 Mile site were auto-GC for VOC and chemiluminescence analyser for NO2 operated by MDEQ and reported in the EPA Air Quality System; these methods were continuous and are summarized based on their average value matched to the same time integral as the co-located passive samplers. Co-located NO2 and VOC measurements from reference samplers in same units. e Reference methods at Dearborn were passive 32 L summa-polished canister samplers operated by EPA for this study; values measurements are matched to the same time integral as the co-located passive samplers. f %RSD: relative pooled standard deviation (standard deviation ÷ mean as percentage) for analysis of precision of duplicate samples. g NM: not measured. | |||||||||||||||
NO2 | 14 | 11 | 17 | 3 | 21 | 20 | 28 | 5 | 0.7 | 13.7 | 12.1 | 17.8 | NM g | NM | NM |
1,3-Butadiene | 0.3 | 0.2 | 0.4 | 10 | 0.2 | 0.1 | 0.4 | 13 | 0.2 | NM | NM | NM | 0.2 | 0.1 | 0.3 |
Benzene | 2.6 | 1.8 | 3.5 | 4 | 2.2 | 1.4 | 2.9 | 6 | 0.1 | 2.3 | 2.0 | 4.7 | 3.5 | 2.2 | 5.5 |
Toluene | 9.3 | 7.6 | 14 | 2 | 7.6 | 5.4 | 9.2 | 2 | 0.1 | 8.1 | 7.0 | 10.9 | 8.7 | 6.0 | 10.3 |
Ethylbenzene | 1.4 | 1.2 | 1.9 | 1 | 1.5 | 1.0 | 2.2 | 4 | 0.1 | 1.4 | 1.1 | 1.7 | 1.8 | 1.1 | 2.4 |
m,p-Xylene | 4.5 | 4.1 | 6.5 | 2 | 5.2 | 3.2 | 7.4 | 4 | 0.3 | 4.8 | 4.1 | 5.9 | 5.4 | 3.3 | 6.4 |
o-Xylene | 1.6 | 1.4 | 2.3 | 1 | 1.5 | 1.0 | 1.8 | 3 | 0.1 | 2.5 | 2.1 | 4.0 | 1.7 | 1.2 | 2.3 |
Styrene | 0.2 | 0.1 | 0.2 | 5 | 0.2 | 0.2 | 0.3 | 9 | 0.2 | 0.6 | 0.5 | 0.8 | 1.2 | 0.5 | 3.0 |
1,3,5-Trimethylbenzene | 0.6 | 0.5 | 0.9 | 4 | 0.7 | 0.5 | 1.8 | 6 | 0.1 | 0.8 | 0.7 | 1.0 | 1.1 | 0.4 | 1.5 |
![]() | ||
Fig. 2 Average concentrations of duplicate Ogawa passive samplers (as horizontal lines) during the study period. Hourly NO2 concentration from the chemiluminescence analyser (as gray area peaks) are also averaged per week (as dots in the middle of the sampling week). |
![]() | ||
Fig. 3 Regression of PE tube vs. auto-GC at East 7 Mile: (a) benzene; (b) toluene; (c) ethylbenzene; (d) o-xylene; (e) m,p-xylene. ap < 0.05 for y intercept significantly different from zero and/or slope significantly different from 1. |
![]() | ||
Fig. 4 Regression of PE tube vs. 32 L canister sample at Dearborn: (a) benzene; (b) toluene; (c) ethylbenzene; (d) o-xylene; (e) m,p-xylene. ap < 0.05 for y intercept significantly different from zero and/or slope significantly different from 1. |
Initial examination of Fig. 3 suggest that PE tubes tended to track the auto-GC for all BTEX species. While the regressions for benzene and o-xylene were statistically different from the 1:1 line, note that the regression lines for benzene (Fig. 3a) and o-xylene (Fig. 3d) were heavily influenced by a single outlier in each case. Review of the individual hourly data from the auto-GC revealed that the outliers for benzene and o-xylene were driven by extremely elevated readings over a particular five hour period. For this five hour period, benzene was elevated to a level approximately ten times higher than any other level seen outside this period. Similarly, o-xylene levels were at least five times higher than any other observation outside this five hour period. This illustrates the distortion of the mean value from the auto-GC as a representative concentration for the entire week when a relatively few elevated values are present. The levels for the other BTEX species were relatively elevated during this period but not nearly to the degree as for benzene and o-xylene. Note that without this outlier, the agreement with the PE tubes and auto-GC was very good for benzene (R2 = 0.82) and o-xylene (R2 = 0.98). However, review of quality assurance data for both the PE tubes and auto-GC did not indicate a problem with the observations and, therefore, all observations are reported in Fig. 3. The overall results from Fig. 3 indicate good agreement of PE tubes and auto-GC with the exception of weeks when extremely elevated concentrations occur during a few hours.
For BTEX measurements compared between PE tubes and 32 L canisters (Fig. 4), agreement as indicated by R2 was generally good except for benzene. Though the regression lines for benzene, ethlybenzene and o-xylene were statistically different from the 1:1 line and the R2 for benzene was only 0.32, note that levels for all species were reported to be very low by both methods. The largest discrepancy was observed for benzene and this difference was < 3 ppbC. Note that the median differences for the PE tubes and 32 L canisters at Dearborn did not exceed 1 ppbC (Table 2).
Compound | Bias (%)a | Differenceb |
---|---|---|
a Bias ((passive–reference) ÷ reference × 100) reported as the median of all sampling period mean biases. b Median difference between passive method minus reference method; units of ppbV for NO2 and ppbC for VOCs. c NO2 comparison from Ogawavs. chemiluminescence analyser. d VOC comparisons from Carbopack X tubes vs. auto-GC at East 7 Mile. e VOC comparisons from Carbopack X tubes vs. 32 L canister samples at Dearborn. | ||
East 7 Mile | ||
NO2c | −2 | −0.2 |
Benzene d | −7 | 0.4 |
Toluene | 19 | 3 |
Ethylbenzene | 5 | 0.2 |
m,p-Xylene | 0.5 | 0.3 |
o-Xylene | −40 | −0.8 |
Styrene | −70 | −0.4 |
1,3,5-Trimethylbenzene | −24 | −0.2 |
Dearborne | ||
1,3-Butadiene | 19 | 0.03 |
Benzene | −38 | −0.8 |
Toluene | −7 | −0.3 |
Ethylbenzene | −12 | −0.2 |
m,p-Xylene | 1 | −0.3 |
o-Xylene | −18 | −0.3 |
Styrene | −84 | −1 |
1,3,5-Trimethylbenzene | −17 | −0.5 |
Biases and absolute differences in Table 2 indicate that the PE tubes tended to under-predict VOC levels in comparison to 32 L canisters. For the PE tubes versus auto-GC, slight over-predictions and under-predictions were observed for VOC species (Table 2). Fig. 5a to c display the agreement between the PE tubes and reference methods.
![]() | ||
Fig. 5 Average concentrations of duplicate Carbopack X tubes and canisters (as horizontal lines) during the study period with hourly average concentrations from the auto-GC. (a) Benzene; (b) toluene; (c) m,p-xylene. Hourly VOC concentration from auto-GC (as gray area peaks) are also averaged per week (as dots in the middle of the sampling week). |
An assessment of quality of the auto-GC VOC data was done by plotting total non-methane organic compound (TNMOC) concentration versus individual VOCs from the hourly averaged auto-GC data. In urban airsheds dominated by traffic, BTEX compounds have been found to be highly correlated with TNMOC.30 Scatterplots (not shown) suggested that BTEX compounds were properly identified in the chromatograms. Separate analysis of the AQS and passive lab data did not reveal any quality issues with the data used.
Nitrogen dioxide measurements reported from the Ogawa samplers and chemiluminescence analyser at the East 7 Mile site were comparable. Fig. 2 shows the good agreement between Ogawas and weekly mean chemiluminescence data at East 7 Mile. The NO2 regression results (Fig. 6) revealed a reasonable R2 (0.6) with the y intercept and slope (3.6 and 0.7, respectively) indicating that the Ogawa and reference method were close to the 1 : 1 line; p values indicated that the slope was not different from 1 and intercept not different from 0. The Ogawas at East 7 Mile tracked NO2 levels well with the chemiluminescence analyser (Fig. 2) and the bias was only 2% (Table 2).
![]() | ||
Fig. 6 Regression of Ogawa sampler vs. chemiluminescence analyser at East 7 Mile. |
Finally, Fig. 2 and 5 also show respective data from the chemiluminescence analyser for NO2 and from the auto-GC for benzene, toluene and m,p-xylene superimposed on the week-long passive and week-long canister sample results; duplicate passive sampler data were averaged for presentation purposes. Potential episodic or short-term influences are smoothed out by the passive samplers and canisters relative to the hourly-averaged reference values. It is conceivable that these episodic ‘spikes’ may have affected the comparison results. Although week-long periods using the hourly data that had < 75% data capture were not used (such as week 1 for the auto-GC—see Fig. 5), it is possible that the reduced sample sizes affected the power of these regressions.
The samplers were generally comparable to reference methods given the low concentrations and small sample sizes. Absolute differences between passive and reference methods for VOCs were small. Results from this study indicate that the auto-GC and the PE tubes could show discrepancies for a week-long sampling period. However, these discrepancies were driven by a few hours of highly elevated concentrations. Thus, the passive samplers are useful for integrated samples while continuous monitors are necessary to assess short term temporal variations.
The primary caveat attached to these results is that this was a short-term study with collection of few samples, and only one measurement method available for the reference samplers. Comparisons of data were, thus, limited. Because of this, it is recommended that additional method evaluations be conducted with these passive samplers to more definitively establish their precision and accuracy. Accordingly, comparison of passive samplers with corresponding reference methods should be a routine quality assurance component where such samplers are deployed for air pollution measurement and modeling studies.
This journal is © The Royal Society of Chemistry 2009 |