Ethylbenzene

Pollutant modeling approach. Models were developed in two stages using different predictor variables and methodology to capture background, regional, and local-scale pollution variation. First, for each National Air Pollution Surveillance (NAPS) fixed-site monitoring station, we derived satellite-based estimates (PM_2.5 and NO₂ only) and geographic variables (e.g., road length, population density, proximity to large emitters) using ArcGIS (version 9.3; ESRI, Redland, CA, USA). We used forward stepwise regression to develop LUR models and retained variables that corresponded to hypothesized effect directions; we maximized the sums of squares explained by Akaike’s information criterion. Spatial autocorrelation was also evaluated using the Moran’s I statistic in ArcGIS. We sought to develop parsimonious models rather than traditional predictive models that maximize prediction but make interpretation of individual variable contributions difficult. Only variables significant at the p < 0.05 level were included in the final models. As expected, NAPS monitoring locations in Canada did not display sufficient variability to estimate model coefficients for important local-scale parameters, such as proximity to major roadways, because of monitor siting. Thus, local-scale predictors were underpowered in the LUR modeling approach.
In the second stage, we conducted comprehensive literature reviews to identify deterministic factors to represent local-scale gradients in pollutant concentrations associated with specific sources (i.e., highways, major roads, gas stations). For each pollutant, we identified concentrations near these selected sources in relation to local background levels and developed deterministic multipliers with distance decay rates (together referred to as gradients in this paper) to apply to the background and regional concentrations predicted by our LUR models. All statistical analyses were conducted using SAS (version 9.1; SAS Institute Inc., Cary, NC, USA).
Air quality data. Annual average concentrations of PM_2.5 (177 monitoring stations), NO₂ (134 monitors), and benzene, ethylbenzene, and 1,3-butadiene (53 monitors) were calculated using data from unique NAPS monitoring sites that were operating during 2006 (see Figure 1). Continuous monitoring data from a given monitor were included if at least 50% of hourly observations were available for a 24-hr period and at least 50% of days were available in a month. Monthly averages from filter-based PM_2.5 measurements required a minimum of three of five valid measurements per month. Annual averages for 2006 were not calculated for individual monitors unless there were at least 6 months of complete data with one valid month per quarter.
NAPS includes different monitor types for PM_2.5, including tapered element oscillating microbalances (TEOMs), dichotomous partisol samplers (Thermo Fisher Scientific Inc., Waltham, MA, USA), and beta-attenuation mass monitors (Met One Instruments Inc., Grants Pass, OR, USA). Multiple monitors are often present at one location, and our comparative analysis found differences in levels measured by TEOMs, which are known to underpredict PM_2.5 because of nitrate evaporation (Dann T, personal communication). We therefore selected other monitor types when they were available at the same location. Those stations with only TEOMs available were adjusted based on yearly calibration between collocated dichotomous and TEOM monitors during 2006 [n = 14, dichotomous = 1.640 + 1.089 × (TEOM), R² = 0.89, p < 0.001]. NO₂, benzene, ethylbenzene, and 1,3-butadiene were measured using standard methods (NAPS 2004).
Predictor variables. PM_2.5 and NO₂ satellite data. Canada-wide concentrations of PM_2.5 and NO₂ were estimated using satellite atmospheric composition data combined with local, coincident scaling factors from a chemical transport model [Goddard Earth Observing System (GEOS)-Chem 2011]. Ground-level PM_2.5 estimates were derived from aerosol optical depth data from the Terra satellite [National Aeronautics and Space Administration (NASA) 2011b], in combination with output from GEOS-Chem simulations to estimate the relationship between aerosol optical depth over the atmospheric column and ground-level PM_2.5 (van Donkelaar et al. 2010 (link)). Ground-level NO₂ concentrations were estimated from tropospheric NO₂ columns retrieved from the ozone monitoring instrument on the Aura satellite (NASA 2011a); GEOS-Chem was also used to calculate the relationship between the NO₂ column and ground-level concentration (Lamsal et al. 2008 (link)). Both PM_2.5 and NO₂ were estimated at a 0.1 × 0.1° resolution (~ 10 × 10 km). Estimates for PM_2.5 were calculated from 2001–2006 data to ensure sufficient observations. For NO₂ estimates, we used data from 2005 and 2006, because ozone monitoring instrument measurements began in late 2004.
Geographic data. We modeled regional pollutant variation using geographic predictor variables potentially relevant to pollutant sources, emissions, and dispersion. To capture varying spatial influences of predictors, all variables were calculated for circular buffer distances ranging from 50 m to 50 km. Classes of variables included population density derived from census block-face points (Statistics Canada 2006 ); 1-km land use classifications (Global Land Cover Characterization 2008 ); high-resolution (30 m) land-use classifications (DMTI Spatial Inc., Markham, Ontario, Canada); sources of large industrial emissions from the Canadian National Pollutant Release Inventory (NPRI; Environment Canada 2010 ); small point source locations extracted from the Dun and Bradstreet (D&B) Selectory database of businesses (Hoovers, Austin, TX, USA) in Canada; length of and distance to specific road classifications using the DMTI Spatial road network, such as freeway, highway, major road, and minor road (DMTI Spatial Inc.); length and density of railroads; elevation; and meteorological variables (precipitation and temperature). Any geographic variables with > 30% zero values—those with no predictive features in proximity to a monitor—were recoded as binary (i.e., present/absent). In total, 10 variable classes and 270 buffer-specific variables were explored in the LUR models.
Deterministic gradients. Gradients were developed with a focus on mobile sources and gas stations. We conducted a comprehensive literature review of published studies to identify the distance from sources at which pollutant concentrations typically return to background levels, and an expected ratio of near-source pollutant levels compared with background pollutant levels for each source and pollutant. We searched PubMed (2010), Web of Science (Thomson Reuters 2010 ), and Google Scholar (2010) using a range of keywords to identify studies with measurements of pollutant gradients. Studies varied widely in terms of location, date, methods, duration of measures, number of samples, and definition of near source and background. We developed linear gradients using the steepest portion of the exponential decay curves typically found in the literature, as the tails of the decay functions were very sensitive to local parameters. Gradients were also selected to represent Canadian conditions. Table 1 summarizes the gradients developed for Canada and applied to the LUR models.
To identify the distance of each NAPS monitor from the nearest highway, major road, local urban road, and gas station, we used DMTI road network data and D&B commercial data for point sources. If a monitor was close enough to one of these features for the source to influence pollutant levels, we modified the corresponding LUR model results (not including point source industrial variables) to account for the deterministic gradients. For example, based on our review of the literature, we assumed that NO₂ concentrations at the side of a highway would be 1.65 times higher than LUR-based background concentrations but consistent with background levels 300 m from the highway; this assumption resulted in a distance decay rate of 0.33% per meter that was applied to the model to estimate NO₂ levels within the 300-m gradient buffer.
Model evaluation. We used three approaches for model evaluation. Due to the small number of NAPS monitoring stations for PM_2.5, NO₂, benzene, ethylbenzene, and 1,3-butadience, we did not leave out a percentage for independent postmodel evaluation, because we wanted to capture the greatest range of model predictors possible. Therefore, we first evaluated all LUR models using a bootstrap approach to determine the sensitivity of model prediction and parameter estimates to monitor sampling. Random selection of monitors was conducted, with replacement, and variable coefficients and model R² values were recorded from the new full sample. This was repeated for 10,000 iterations to estimate the 95% confidence interval (CI) for overall model prediction and individual variable coefficients. Next, we conducted a leave-one-out analysis where each LUR model was repeatedly parameterized on n – 1 data points and then used to predict the excluded monitor measurement. The mean differences between the predicted and measured values were used to estimate model error.
Finally, we evaluated the NO₂ and benzene LUR models, with and without gradients, against independent data (35–196 monitoring sites per city) previously collected for LUR models in seven Canadian cities (for a full description of data collection and modeling see Allen et al. 2010 ; Atari and Luginaah 2009 (link); Crouse et al. 2009 ; Henderson et al. 2007 (link); Jerrett et al. 2007 (link); Su et al. 2010 ). Briefly, in each city, monitoring took place over a 2-week period; data from fixed-site monitors, monitoring during yearly average concentration periods, or multiple measurement periods were used to estimate yearly averages [see Supplemental Material, Table 1 (doi:10.1289/ehp.1002976) for the city-specific data used for model evaluation]. These pollution measurements were collected at much higher spatial densities than were NAPS and from monitors that were located to specifically capture spatial pollutant gradients. Consequently, these data were reasonable for use as a gold standard to determine how well the two national NO₂ and benzene models (the LUR models and the LUR models with gradients) predicted within-city variation. In addition, we compared the city-specific data with estimates based on inverse distance weighting (IDW) of annual average NO₂ and benzene concentrations measured at NAPS monitors (with and without deterministic gradients). Because of NAPS monitor density in Canada, kriging could not be applied.
Population exposure assessment. The national pollutant models were applied to each of the 478,831 Statistics Canada street block-face centroid locations in 2006 to estimate population exposures. First, we applied the LUR models to each block point to derive a unique predicted pollutant concentration for each point, representing the average exposure level for 89 and a SD of ± 158 individuals. We used a GIS to identify the distance of each block centroid to the nearest highway, major road, local urban road, and gas stations and adjusted the corresponding LUR model estimate when the street block point was located within an associated gradient. We then estimated population-weighted exposures to PM_2.5, NO₂, benzene, ethylbenzene, and 1,3-butadiene in the Canadian population as a whole, and we estimated uncertainty using the 95% confidence limits for LUR model predictions. Because there was insufficient information in the literature to examine uncertainty for specific gradients, we selected ± 50% for all gradients (values shown in Table 1).

OSRC workers performed a range of jobs/tasks, from stopping the leak to administrative support, with different exposure profiles (Table 1). Initially, jobs and tasks were the basis of a preliminary exposure assessment. Due to the weathering of the oil, vessel, vessel type, location, and time periods were later identified as possible determinants of individual exposure levels. The ultimate goal of the GuLF STUDY is to have quantitative exposure estimates for total hydrocarbons (THC) and BTEX-H (benzene, toluene, ethylbenzene, xylene, hexane) as these oil-related chemicals comprised most of the air measurements taken during the spill and are generally considered to be the more toxic components. Exposure estimates for dispersants and particulates from burning were also desired because of their association with some health effects and because of concerns raised by the public. An ordinal job–exposure matrix (JEM) was developed based on jobs or tasks/vessel or vessel type/location/time period to estimate THC exposures for study participants (Stewart et al., in press ). THC is a composite of the volatile chemicals from the oil and, as such, can be thought of as a surrogate for the “OSRC oil experience.” In the development of the questionnaire and the ordinal and quantitative JEMs, study industrial hygienists (IHs) relied on BP measurement data and their accompanying documentation, federal and BP contractor reports, numerous other spill-related documents, and interviews with key personnel managing the OSRC effort and some workers.
The exposure section of the enrollment interview was structured to capture detailed information about the participants’ OSRC activities and served as the link to the JEM. Participants provided the start/stop dates for any OSRC work and then for each OSRC job/task queried, start/stop dates, average number of days worked/week, average number of hours worked/day, use of personal protective equipment, and dermal contact with chemical agents. Participants also provided information on heat stress and other work-related exposures and on sleeping quarters.
More than 28,000 full-shift, personal air monitoring samples were collected on workers by BP contractors to characterize exposure to OSRC chemicals from April 2010 through June 2011. Because multiple chemicals were analyzed on each sample, 160,000 measurements were available on THC, BTEX-H, and other toxicants. A large proportion of these measurements was below the reported limits of detection when analyzed based on occupational exposure limits. When these monitoring data were recalibrated by one of the BP contractors and the study IHs to reflect the analytical methods’ limits of detection, it was possible to quantify levels below the initially reported LODs. The effort substantially decreased the amount of censored data; for example, THC censored data went from 80% to ~ 20%. The proportion of censored data for the other chemicals was still relatively high (~ 70%) but was substantially lower than the original 95% censoring. We evaluated strategies for dealing with censored data and developed methods to leverage the censored data on THC to develop estimates for other BTEX-H chemicals (Huynh et al. 2014 (link), 2016 (link); Quick et al. 2014 (link)).
Our team of experienced IHs used the recalculated air measurement data to identify factors associated with exposure levels to characterize exposures: jobs/tasks, vessel/vessel type, location, and time period. Unique combinations of these factors were identified that were expected to have similar distributions of THC exposure. The measurement data were used to determine average THC exposures for each job or task/vessel or vessel type/location/time period combination (n = 2,385 “exposure groups”), which was translated to ordinal values (1–7). The resulting JEM was linked to the OSRC work reported in the questionnaire to estimate THC exposures for each participant in the cohort. Different metrics can be developed for different exposure–response scenarios and assumptions. For example, we estimated the maximum exposure by identifying the maximum level across all estimates assigned to an individual to create a person-specific maximum exposure metric. Exposure averages (mean or median) within and across jobs/tasks or in specific time periods (e.g., before the well was capped) or locations also can be developed.
Specific questionnaire responses were also used to identify, based on tasks, vessels, locations, and dates, workers with likely exposure to dispersants (yes/no) and to particulates (low, medium, high) from burning of oil. Quantitative exposure estimates for inhaled THC and specific chemicals (e.g., BTEX-H) are being developed, as are semiquantitative estimates of dermal exposure, estimates for dispersants, and estimates for particulate matter from burning.

Ventilation and air quality measurements were concurrently obtained in the living area and the child’s bedroom of each house during baseline and each seasonal assessments (summer was defined as June, July and August; fall as September, October and November; winter as December, January, February, and spring as March, April and May.) ACRs were estimated using the multizone constant injection method [16 ,17 ], which allows determination of “local” ACRs. The method used two different perfluorocarbon tracers (PFTs) and concentration measurements of the tracers in both zones. Two passive emitters of hexafluorobenzene (HFB) were placed in the living area, and two emitters of octafluorotoluene (OFT) in the sleeping area. Emitters were typically placed in opposite corners of rooms, releasing the PFT at a constant rate over the weeklong sampling period. PFT concentrations were measured using passive samplers at central locations in the two rooms, placed away from the emitters [18 (link)]. The passive samplers were analyzed using thermal desorption, cryofocusing and GC-MS analysis [19 (link)]. Figure 1 depicts flows Q₁ to Q₄ determined by this method; flows Q₅ and Q₆ are obtained by flow balance. Note that zone 1, called the living area in this paper, refers to all rooms other than the (study) bedroom.
ACRs for the house and bedroom, and flows between these zones, were determined using the measured concentrations and the volumes of the house and bedroom as follows [20 (link)]:

where Q₁ and Q₂ = air flows into zone 1 and 2, respectively, from outdoors (m³·h⁻¹); Q₃ = air flow rate from zone 1 to 2 (m³·h⁻¹); Q₄ = air flow rate from zone 2 to 1 (m³·h⁻¹); C_HFB,1 and C_OFT,1 = concentrations of PFT tracers HFB and OFT in zone 1 (mg·m⁻³); C_HFB,2 and C_OFT,2 = concentrations of HFB and OFT in zone 2 (mg·m⁻³); and E_HFB,1 and E_OFT,2 = emission rates of HFB and OFT in zones 1 and 2 (mg·h⁻¹), respectively. This result assumes that outdoor PFT concentrations are zero, and that the PFTs are inert (removed only by airflows and not by settling, deposition, filtration or reaction). The ACR_i (h⁻¹) in zone i was calculated as Q_i/V_i (i = 1,2), where V_i = volume of zone i (m³).
Interzonal flows. Interzonal flows transport pollutants between zones, e.g., cigarette smoke emitted in the living area that is brought to the bedroom. Interzonal flows Q₃ and Q₄ from the two zone model are expressed as interzonal flow proportions 𝛼_HB and 𝛼_BH (dimensionless, ratios between 0 and 1):


where 𝛼_HB = fraction of the air coming into the bedroom that arises from the (remainder of the) house, and similarly, 𝛼_BH = fraction of house air coming from the bedroom. These proportions provide a simple way to compare the magnitude of interzonal flows among buildings of different sizes.
Indoor and outdoor measurements. PM concentrations in bedrooms were measured as sequential 24-h filter samples during the sampling week collocated with the PFT samplers in each season. As detailed elsewhere [15 (link)], PM samples were collected at 15 L/min on 1 µm-rated PTFE filters installed in static-free polypropylene cassettes (Omega Specialty Instruments Co., Houston, TX, USA) for gravimetric analysis. The inlets on these cassettes are not designed to be size selective, and they essentially capture the total suspended particulate (TSP) fraction. In addition, particle number counts (PNCs) in 0.3–1.0 µm and 1–5 µm dia size ranges were measured continuously using an optical particle counter (GT-521, MetOne, Grants Pass, OR, USA). PM concentrations and 0.3–1.0 µm PNCs were significantly correlated [15 (link)]. The PM data were reduced to weeklong averages.
Outdoor PM_2.5 measurements were obtained from air quality monitoring sites in Detroit selected to be representative of population exposure. These included daily data from four sites (Allen Park, Ambassador Bridge, Dearborn, Newberry School), and every third day data from five additional sites (Southwest High School, Linwood, East 7 Mile, Livonia, Wyandotte). These sites were operated by the Michigan Department of Natural Resources and the Environment using protocols that followed standard federal reference methods. Meteorological data, including wind speed, direction, temperature, humidity, and barometric pressure, were obtained from the Detroit City Airport site located near the middle of the study area.
In addition to the PFT tracers, the same passive samplers measured two tracers of environmental tobacco smoke (ETS), i.e., 2,5-dimethylfuran and 3-ethenylpyridine used to confirm the presence of ETS, along with about 100 other volatile organic compounds (VOCs), e.g., naphthalene, BTEX (sum of benzene, toluene, ethylbenzene and xylenes), and total volatile organic compounds (TVOC, sum of target compounds) [15 (link),18 (link),20 (link),21 (link)]. These samplers were placed in bedrooms and living areas for a 1-week period, and analyzed using thermal desorption, gas chromatography and mass spectrometry. Temperatures and relative humidity also were monitored in both bedrooms and living rooms, and CO₂ was monitored in the bedroom using an infrared sensor. These variables were monitored continuously and reduced to 1-week averages.
Quality assurance (QA). PFT, VOC and ETS tracer measurements used duplicate samplers and showed good agreement, i.e., replicate precision averaged 11 ± 12% for the PFTs, 15 ± 16% for VOCs, and 14 ± 13% for the ETS tracers. Field blanks for passive sampling tubes were deployed at each household each week, and showed negligible contamination. Emitters were weighed periodically to determine emission rates, and samplers were temperature corrected. ACR measurements that were excessively large (≥10 h⁻¹) or unrealistically small (≤0.1 h⁻¹) probably resulted from incomplete mixing or other reasons, and thus were omitted from analyses. (As shown later, such values constituted a very small fraction of measurements.) Further description of QA is provided elsewhere [15 (link)].