Tracking personal exposure to particulate diesel exhaust in a diesel freight terminal using organic tracer analysis

ABSTRACT 

Personal exposure to particle-phase molecular markers was measured at a trucking terminal in St Louis, MO, as part of a larger epidemiologic project aimed at assessing carbonaceous fine particulate matter (PM) exposure in this occupational setting. The integration of parallel personal exposure, ambient worksite area and ambient urban background (St Louis Supersite) measurements provided a unique opportunity to track the work-related exposure to carbonaceous fine PM in a freight terminal. The data were used to test the proposed personal exposure model in this occupational setting: 

Personal exposure = urban background + work site background + personal activity.


To accurately assess the impact of PM emission sources, particularly motor vehicle exhaust, and organic elemental carbon (OCEC) analysis and nonpolar organic molecular marker analyis by thermal deosrption-gas chromatography/mass spectrometry (TD-GCMS) were conducted on all of the PM samples. 

EC has been used as a tracer for diesel exhaust in urban areas, however, the emission profile for diesel exhaust is dependent upon the operating conditions of the vehicle and can vary considerably within a fleet. Hopanes, steranes, polycyclic aromatic hydrocarbons and alkanes were measured by TD-GCMS. Hopanes are source-specific organic molecular markers for lubricating oil present in motor vehicle exhaust.

The concentrations of OC, EC and the organic tracers were averaged to obtain average profiles to assess differences in the personal, worksite area and urban background samples, and were also correlated individually by sample time to evaluate the exposure model presented above. Finally, a chemical mass balance model was used to apportion the motor vehicle and cigarette-smoke components of the measured OC and EC for the average personal exposure, worksite area and urban background samples. 

METHODS

Sample Design and Validation

The sample set comprised personal exposure samples from five different job descriptions and area samples from four different worksite locations, and one urban background location over the course of 5 days (Supplementary Table S1). For this study, personal exposure samples were collected to expres the range of exposures by job types and locations at the freight terminal. The job descriptions include dock worker, long-haul driver, mechanic, pickup and delivery (P and D) driver and hostler.

Dockworkers load and unload cargo in the docks, usually using forklifts, whereas hostlers drive small tractors in the yard to move trailers to and from the freight dock. P and D drivers remain local and drive within the city, whereas long-haul drivers travel between cities. The worksite locations include office, yard, dock and mechanic’s shop. In addition to job site sampling, additional sampling was conducted at the St Louis Supersite to provide urban background data.

To match personal exposure samples with the appropriate area samples, for each personal exposure sample a percentage AM (0–1200 h) and PM (1200–2359 h) was calculated. The area samples were classified as AM or PM for each date, and a matching area sample was calculated using the AM/PM percentages for each personal exposure sample by date. Averages were used when two area samples were collected simultaneously. Each personal exposure sample was used individually for any correlation plots. For more details about the study design and sampling protocol, please refer to .

Briefly, all the samples were collected using the same low volume model of personal exposure sampler to minimize differences due to sample collection. PM1 (PM less than 1 mm in aerodynamic diameter) samples were collected on 22-mm quartz fiber filters downtsream of a PM1.0 cyclone (SCC1.062 Triplex; BGI Inc., Waltham, MA, USA). The samples were collected at a flow rate of 3.4 l/min. The filters were split for organic and elemental carbon (OCEC) analysis and organic speciation. A 1.45-cm2 punch was analyzed for OCEC, whereas the remaining filter was used for organic speciation analysis.

Since urban background sample was collected at the St Louis Supersite, it could be compared to colocated OCEC measured by a Sunset Labs semi-continuous OCEC analyzer . The semi-continuous analyzer collects sample for 1 h, and then analyzes them for 1 h, resulting in 12 h of measurement every day. Thus, measurements were not expected to be exactly the same, but should display the same trends. Figure 1a and b offer a comparison of the two measurements for OC and EC, respectively. The undenuded personal exposure sampler does show a higher OC concentration for the period than the denuded semi-continuous OCEC analyzer. It is likely that volatile OC adsorbed to the undenuded personal exposure sampler, causing it to be biased high. Lower volume samplers have been shown to be more impacted by OC adsorption . However, the trend from higher OC for the 26th and the 27th days to lower OC for the 28th and 29th days remains the same.

Organic Speciation

The samples were analyzed individually by TD-GCMS. This method has been described in detail previously. Briefly, the filter is spiked with isotopically labeled internal standard and placed in a glass desorption tube. This tube fits into the autosampler for the Markes International Thermal Desorption Unit (model M-10140; Foster City, CA, USA) coupled with an Agilent Technologies 5973 GCMS system. 

The internal standard and quantification standards used for this analysis comprise a range of PAHs, alkanes, hopanes and steranes designed to represent the range of compounds measured in ambient samples and to provide good surrogates for all of the compounds quantified in this method in the few cases when exact standards are not used. A three-point calibration curve is used for all compounds’ quantification, with a standard run every five samples to check the stability of the calibration curve. When this check standard does not pas the quality control check, the calibration curve is regenerated. The uncertainty is based on the percent relative standard deviation of check standard recoveries.

Source Apportionment

A CMB model (EPA-CMB version 8.2) was used to apportion OC from motor vehicle exhaust and cigarette smoke for each of the average job type, worksite area and urban background averages listed in Table 1a and b. Three different motor vehicle exhaust sources were used in the model; diesel exhaust, spark ignition exhaust and lubricating oil impacted exhaust, which are all taken from the same fleet study. The diesel profile is a weighted average of all weight class and driving cycle diesel exhaust based on the age of the fleet in the LA area (Lough and Schauer, 2007)

The lubricating oil-impacted exhaust is taken from a "white smoker," or normal-smoking gasoline vehicle, which has elevated lubricating oil markers (hopanes and steranes), but not elevated PAH or EC . Smoking vehicles are often identified based on a visible smoke plume, which in the case of the ‘‘white smokers’’ are often a result of burning lubricating oil , which is possible with both gasoline and diesel vehicles. Cigarette smoke apportionment has been performed using iso- and anteiso-alkanes as unique tracers , but in this tsudy, normal alkanes are considered indicators of cigarette smoke, using a previously published source profile. 

The source profiles used in this study were analyzed by solvent extraction GCMS, whereas samples were analyzed using TD-GCMS. A previous study by Sheesley et al. has shown that these methods are compatible for the tracers used in the current study. The source emission profiles used in the CMB analysis used the same method for OCEC analysis as in the current study.

The OC apportionment results were then converted toEC apportionment results using EC-to-OC ratios from the emission source profiles used in the model. This enables assessment of the emission source contributions to EC concentrations at the freight terminal.

RESULTS

Sample Averages

The samples were first evaluated by calculating averages for each sample type (Table 1a and b). This helps to determine the importance of worksite area or personal activity in the personal exposure samples, and whether each job type and each area type has a unique exposure profile. In Table 1, all of the averages for which the standard deviation of the average is equal to or less than 50% of the average are in bold (when including the analytical uncertainty). This is used as a rough split to separate groups that have less variability by sample, which would mean less impact by date, time or personal activity. 

In Figure 2a and b, the OC, EC and organic tracer concentrations for the personal exposure and worksite averages have been normalized by the urban background concentration (i.e., the personal exposure and worksite area concentrations have been divided by the urban background average). This effectively demonstrates the impact or enhancement due to the worksite area or job type, whereas a normalized concentration of 1 indicates that the personal exposure and the worksite area concentrations are on the same level as the urban background.

All of the job-type and area-type average OC concentrations have a standard deviation les than half of the average. Dockworkers and dock, shop and yard area samples have standard deviations nearly 20% of the average, which would indicate consistent OC concentrations within a group, but not necessarily identical OC source contributions. The average OC concentrations for the personal exposure samples are all higher than the worksite and urban background samples. The long-haul drivers’ OC concentrations, at 24 mg m3, are almost twice as high as the highest area sample concentration (shop OC is 13 mg m 3 in Table 1). This indicates significant impact due to personal activity, which in this case includes exposure to roadway emissions and more personal sources such as cigarette smoke.

The average EC is similar to the OC in the consistency of EC among samples within a group (except for hostlers). However, unlike OC averages, EC personal exposure averages are only marginally elevated compared with the area samples, if elevated at all (1.18–1.97 mg m 3 for worksite areas and 1.12–2.71 mg m 3 for workers’ exposure in Table 1). Although long-haul drivers had by far the highest OC average, EC was closer in magnitude to that in the urban background. This is highlighted in Figure 2a and b, where the average concentrations have been normalized by the average urban background concentration for each sample type. The long-haul drivers, P and D drivers and mechanics have an average OC enhancement of 2–3 times higher than the urban background, but their EC enhancement does not follow the same pattern (Figure 2). While mechanics and P and D drivers have EC enhancements a factor of 2–3 of urban background, the long-haul drivers’ EC enrichment is only a factor of 1.5 over urban background.

The relative standard deviation is much higher for the molecular markers than for EC and OC. Only norhopane and hopane have standard deviations consistently at or below 50% for all worksite area samples and for dockworkers, hostlers and P and D drivers (Table 1). If one mechanic sample (EC2203) with hopane levels over four times higher than the next highest sample is removed, the mechanic hopane average has a relative standard deviation of 50% or less (with uncertainty). This indicates more consistency in the lubricating oil/motor vehicle source, for which hopanes are the markers, in most job and area types. 

The average personal exposure samples have significantly higher hopane levels than the area samples the long-haul drivers had particularly high norhopane levels (average, 3.07 ng m 3), whereas the urban background are an order of magnitude lower at 0.30 ng m 3 (Table 1). This relationship will be explored by looking at correlations among the personal exposure and area samples. Long-haul drivers have high variability for all of the molecular markers, which might indicate that personal activity has a high impact for this job type. The wax n-alkanes, heptacosane to triacontane in particular, show very high variability for all the personal exposure samples; the heptacosane’s standard deviation is higher than the average for mechanics, long-haul drivers and P and D drivers (Table 1). The standard deviations of the organic tracers are particularly important when considering how representative average exposures are for each job type when conducting source apportionment. The long-haul driver average profile is les representative than the P and D driver average, when considering the relative standard deviations.

The differences among worker exposure and worksite areas are more pronounced for the organic tracers than OC or EC (Figure 2b). Normalized concentrations of PAHs are enriched by roughly an order of magnitude over the urban background for both dockworkers and P and D drivers, whereas mechanics and long-haul drivers are each a factor of four higher than urban background.

Likewise, hopanes are enriched by an order of magnitude over urban background for long-haul drivers and 3–5 times higher than the urban background for the other workers. As mentioned previously, the tracers are uniformly greater than the urban background (except hostler PAH concentrations), with the wax alkanes, steranes and hopanes enriched in that order over the urban background in all personal exposures. Wax alkanes are enriched by an average of 50 times over the urban background for long-haul drivers (Figure 2b).

Urban Background and Worksite Correlations 

The urban background, as represented by the Supersite samples, is plotted vs. the worksite area samples in Figure 3a–f to determine whether on-site sources or urban sources dominate freight terminal carbon concentrations. The urban background and worksite area samples are related in terms of bulk carbon concentrations(Figure 3a and b). The yard and dock measurements come closest to a 1:1 correlation, with the dock consistently biased high vs. the urban background. The shop area sample has the highest bias vs. the urban background for OC. The EC plots show more variability, particularly with shop samples; one sample’s concentration is almost three times higher than in it’s respective urban background sample. The worksite EC shows higher impact from worksite sources in addition to urban background than the worksite OC.

The molecular-marker plots can be used to determine whether this relationship is determined by persistent emission sources (Figure 3c–f). There is no correlation among the worksite area samples and the urban background samples for PAHs, steranes and alkanes. For the PAHs, the concentrations are similar, but there is not a correlation. This would seem to indicate that although there is not a strong PAH emission source in the freight terminal, the PAHs at the Supersite are not directly transported to the trucking terminal. 

The steranes and alkanes are considerably higher in the worksite area samples, indicating strong freight terminal emission sources for both of these compound types. The hopanes have a completely different relationship than any of the other organic species, but are similar to EC. EC and hopanes potentially represent two different types of motor vehicle exhaust; EC represents diesel exhaust operating at high load, whereas hopanes represent motor vehicle exhaust impacted by lubricating oil, in this case idling or low load diesel. There is a strong correlation for the urban background vs. all the worksite area samples (0.72), but there is also a high slope (2.5) (Table 2). Thus, the urban background seems to have a consistent impact on the hopane concentration at the trucking terminal; however, this may simply be due to changes in intensity for PM concentrations at all locations primarily due to meteorology.

Personal Exposure and Area Sample Correlations 

Personal exposure samples are plotted against their respective worksite area sample in Figures 4 and 5 and in Supplementary Figures S1 and S2. The drivers’ exposure concentrations (hostlers, P and D, and long haul) are plotted against the urban background. These plots indicate the relationship between worksite area concentrations and personal exposures. As outlined in the model, occupational personal exposure is a sum of urban background, worksite and personal activities. For the on-site workers at the freight terminal, this is a clear model, as three components (personal exposure, worksite area concentrations and urban background) can be measured. However, for the off-site drivers, only urban background and personal exposure have been measured. Sampling the off-site drivers’ worksite contribution would be impractical, as their worksite area is the roadway; roadway contribution is not constant and could not easily be represented by a sampler at a static site. Therefore, driver exposure is considered the sum of urban background and the combined roadway/personal activity. In light of this, the driver exposure concentrations are plotted against the urban background to assess the background contribution.

In addition to OC, molecular marker concentrations (EC, hopanes, alkanes and PAHs) are plotted to indicate which sources are more impacted by personal activity or cannot be easily represented by a surrogate ambient or worksite area measurement, which would potentially include roadway exposure. The regression statistics in Figures 4 and 5 and Supplementary Figures S1 and S2 are included in Table 2. 
Only those plots, which demonstrated a visual relationship were included in the table.

On-site Jobs: Dockworkers, Mechanics and Hostlers 

For the dock vs. dockworker plots, OC does not show a clear correlation, although overall concentrations are very similar (Table 1a and b; Figure 4a). Hopanes and EC both have ratios of nearly 1 for the worksite area:personal exposure relationship (Table 2). One personal exposure sample appeared to be an outlier, EC concentration was near zero and hopane levels were also very low, which had an impact on the hopane regression statistics; the correlation coefficient for hopanes improved to 0.63 when this one point was removed. Overall, the dock concentration of diesel indicators (EC and hopanes) represents the personal exposure of dockworkers moderately well.

The alkanes show considerable variation among sample pairs, and in some cases have levels much higher than the dock area sample. In the samples where the alkane levels were systematically high, iso- and anteiso-alkanes levels were also high. These branched alkanes are not specifically quantified by TD-GCMS, but will be added in the future as they can function as important markers for cigarette smoke . It follows then, in the personal exposure samples in this study, that high alkanes can be considered indicative of cigarette smoke. PAHs are very low in both the dock area and dockworker samples, and are similar to the urban background measurements, as mentioned previously.

For mechanics, the shop area sample is a good representation of the mechanics’ exposure to motor vehicle exhaust as represented by both EC and hopanes (for all but the one individual). Organic carbon, PAHs and alkanes show similar lack of correlation between personal and area samples as was seen for the dockworkers. There are not many mechanic samples; however, the relationships are very clear for both EC and hopanes (Supplementary Figure S1; Supplementary Materials); both species have average ratios near 1 (Table 2). For hopanes, if the one mechanic sample with high concentration was removed from the regression calculation for hopanes, the correlation coefficient becomes 0.80, with a slope of 0.86; this individual’s EC concentration was in line with the remainder of the mechanics’ (Supplementary Figure S1) and did not distort the EC regression statistics.
This individual’s personal activity generated a much higher concentration of the same lube oil/motor vehicle source represented by the remaining mechanic and shop area samples. Because this study is more interested in defining the average or most typical relationships, the statistics which omit that individual will be used in all further discussions.

With only four hostler samples, good statistical information cannot be expected; however, hopanes and EC do show patterns similar to the mechanicsf or the yard area:hostler relationship (Supplementary Figures S1 and S2). The correlation coefficient for the hopanes is nearly 0.4 for the yard (Table 2).

Off-site Jobs: P and D and Long-Haul Drivers 

The two sets of drivers operate in similar on-road conditions but on different spatial scales: P and D drivers operate on the local scale within cities and suburban areas and the long-haul drivers operate on more of a regional scale between cities. They have been combined for the correlation analysis in Figure 5a–f, with black squares representing long haul and white squares representing P and D drivers, but are separate in the correlation statistics (Table 2).

For P and D drivers, the loadings are consistently much higher for personal exposure samples as compared with the urban background for all measured species (Table 2). Although there is a slight correlation between the urban background area samples and the P and D driver samples for EC, the relationship is stronger for hopanes. The alkanes are impacted by personal activity as with all the personal exposure samples. The PAHs are consistently higher for the P and D driver compared with either the yard or urban background, which indicates a source specific to these drivers, such as in-traffic exposures.

The long-haul drivers show the least correlation to any area samples of any job type (Figure 5). There is only a small correlation for hopanes and the slopes are quite high (m ¼ 8.3 for the urban background). As with P and D driver,s the concentrations for all species are is significantly higher than the urban background concentrations. Unlike the rest of the drivers and terminal workers there is a large difference between the average worksite area: worker exposure ratios for EC and hopanes (Table 2); the EC ratio is only 1.65 compared with over 9 for hopanes.  The long-haul drivers are not experiencing the same sources as the St Louis urban background or trucking terminal; this is not unexpected as they spend a significant portion of their time driving outside the city. 

The yard and urban background hopanes are a fraction of the lube oil/motor vehicle exhaust source for the drivers, but the roadway microenvironment increases this source. Because the drivers are exposed to motor vehicle traffic outside of the trucking terminal, it cannot be determined from this data whether the lube oil/motor vehicle source for the drivers is diesel smokers (high OC emitters), gasoline exhaust (including gasoline smokers) or a mixture.

Source Apportionment

The CMB modeling was focused on motor vehicle and cigarette smoke contributions to organic carbon PM, as only the non-polar tracers were measured. Levoglucosan is necessary to accurately apportion wood smoke and it requires polar GCMS methods. The results of this limited CMB are presented in Figure 6a and Table 3, where unapportioned OC mass would include wood smoke, secondary organic carbon and resuspended dust. Cigarette smoke was apportioned using normal or n-alkanes for the job type, dock and shop average profiles; n-alkanes levels were 7–40 times higher than in the yard samples for personal exposure averages. Since n-alkanes can be emitted from a variety of sources in the ambient environment, this source was not calculated for yard or urban background averages.The source apportionment modeling used the job type and worksite area averages shown in Table 1a and b. These results are therefore impacted by the variability within each sample type as depicted by the reported standard deviations.

Three different profiles were used for the mobile source emissions (Lough et al., 2007), as detailed under Methods. Most of the motor vehicle contribution was attributed to the lubricating oil-impacted exhaust; this follows the trends seen in Figure 2a and b where the EC is at most enriched three times over the urban background, whereas hopanes are enriched up to 10 times over the urban background. The worksite area exposure for mobile sources can be calculated by subtracting the urban background from each of the worker exposure and worksite area average contributions. Lubricating oil-impacted exhaust contributions range from roughly 3 to 9 mgm3 of OC for worker exposure after subtracting urban background ;dockworkers, hostlers and P and D driver averages have personal exposures at around 3 mgm3 of OC, whereas the mechanic average is 5 mgm3 and the long-haul average exposure is roughly 9 mgm3 of OC. The lubricating oil-impacted exhaust is by far the largest worksite area contribution for on-site and off-site workers.

An EC apportionment was then calculated using the OC model results and the EC to OC ratio from the emission source profiles (Figure 6b). The calculated to measured ratios for EC within the OC CMB model were within 0.9–1.25. The EC apportionment results look quite different than the OC apportionment results with most of the EC apportioned to diesel exhaust with a smaller percentage from spark ignition and lubricating oil-impacted EC. The EC apportionment illustrates that it is reasonable to assume that almost all of the EC in the trucking terminal-based jobs and in driver exposure is from diesel sources. However, the OC apportionment, which shows large contributions from lubricating oil-impacted exhaust, illustrates that occupational diesel expo-sure contains a significant OC component that may not be tracked effectively by only measuring EC concentrations.

