Mass, surface area and number metrics in diesel occupational exposure assessment

ABSTRACT

While diesel aerosol exposure assessment has traditionally been based on the mass concentration metric, recent studies have suggested that particle number and surface area concentrations may be more health-relevant. In this study, we evaluated the exposures of three occupational groups-bus drivers, parking garage attendants, and bus mechanics-using the mass concentration of elemental carbon (EC) as well as surface area and number concentrations. These occupational groups are exposed to mixtures of diesel and gasoline exhaust on a regular basis in various ratios. The three groups had significantly different exposures to workshift TWA EC with the highest levels observed in the bus garage mechanics and the lowest levels in the parking ramp booth attendants. In terms of surface area, parking ramp attendants had significantly greater exposures than bus garage mechanics, who in turn had significantly greater exposures than bus drivers. In terms of number concentrations, the exposures of garage mechanics exceeded those of ramp booth attendants by a factor of 5-6. Depending on the exposure metric chosen, the three occupational groups had quite different exposure rankings. This illustrates the importance of the choice of exposure metric in epidemiological studies. If these three occupational groups were part of an epidemiological study, depending on the metric used, they may or may not be part of the same similarly exposed group (SEG). The exposure rankings (e.g., low, medium, or high) of the three groups also changes with the metric used. If the incorrect metric is used, significant misclassification errors may occur.

METHODS

A. Instrumentation

The instrument suite used in this project enabled aerosol from <10 nm to 10 mm to be characterized and was uniquely capable of characterizing exposure.

i. Mass concentration (elemental carbon sampling). 

Elemental carbon (EC) accounts for a significant fraction of DPM emissions, and is readily detectable by thermal-optical analysis. This analytical method is described in the NIOSH Manual of Analytical Methods. The personal sampling train for EC samples consisted of a 10 mm MSA cyclone, 37 mm pre-fired ultra pure, quartz fiber, Pallflex filter held in a 3-piece 37 mm cassette, and a MSA ELF pump. Pumps were calibrated with a Gilibrator primary flow device (electronic bubble meter) and operated at 3.0 LPM so that the collected mass of EC was greater than the limit of detection for the method. At this flow rate, the cyclone had a 50% cut-point of 2.85 mm. The minimum sampling time was six hours. Diesel and spark-ignition aerosols are nearly all less than 1.0 mm in diameter, and thus the cyclone collected virtually the entire aerosol mass.

Quartz fiber filters are known to adsorb organic vapors during the sampling process. A correction factor was estimated for this adsorption by using dynamic blanks. A Teflon filter (Zefluor) was placed on top of the Pallflex filter in the three piece cassette and samples were collected as discussed above. The Teflon filter adsorbs little organic vapor but effectively removes particulate matter. The organic vapor passes through and adsorbs on the Pallflex filter, which is analyzed by the NIOSH 5040 method for EC. It is believed that the Pallflex filter reaches equilibrium with the adsorbed organic material thus allowing for a correction factor to be estimated. In addition to dynamic blanks, a limited number of field blanks were submitted for analysis.

ii. Surface area concentration (Diffusion Charger—DC). 

The diffusion charger measures the total active surface area of particulate matter. Positively charged ions are produced by a glow discharge, forming in the neighborhood of a very thin wire. These ions attach themselves to the sampled aerosol stream with a certain probability. The charged aerosol particles are then collected on a filter. The electric current flowing from the filter to ground potential is measured and is proportional to the number of ions attached to the particles. For particles in the free molecular range, the attachment is proportional to the surface area of the particles, but is independent of the composition of the particles. Siegmann et al. contend that the DC measures the so-called ‘‘active surface’’ in the size range from slightly above ten to a few hundred nm. The active surface is the effective surface area available for mass transfer in a kinetically limited situation and should be appropriate for describing the gas to particle mass transfer taking place in a diluting exhaust plume. However, for particles ~1 mm, the instrument response scales as particle diameter, and thus the relationship with surface area is lost.

A portable DC (EcoChem Analytics LQ1-DC) was used in this project. It was set to average data over a 10 s period, covering a range from 0 to 4000 fA with a resolution of 1 fA and a flow rate of 1.0 L min-1.

iii. Surface area concentration (Photoelectric Aerosol Sensor—PAS). 

The PAS responds to photoemitting substances on the surface of aerosol particles. Ultraviolet irradiation of the sampled aerosol leads to the emission of photoelectrons from surface material that readily undergoes photoemission. The remaining positively charged aerosol particles are separated from the electrons and collected on a filter connected to an electrometer. The measured current is a function of the UV irradiation wavelength and intensity, the total available surface area and the photoemission properties of the surface materials. Commercially available instruments usually use a wavelength of 222 nm. Diesel accumulation mode particles strongly respond to the PAS.

The PAS was originally promoted as a monitor for surfacebound polycyclic aromatic hydrocarbons (PAHs). Surface bound PAHs have a high photoelectric (PE) yield, causing a high PAS response. DPM found in the accumulation mode consists primarily of EC, and has a significantly lower PE yield, which results in a more moderate PAS response. For diesel accumulation mode particles, the PAS signal correlates with the accumulation mode surface area concentration. In mixed urban aerosols, on the other hand, where the PAH surface concentration (for example from spark ignition engines) may be much higher, the high PE yield of the PAH dominates the PAS response. Therefore, PAS signals can only be quantitatively interpreted in a certain context. Nevertheless, PAS signal observations are useful for monitoring of relative changes using the measured electrometer response (usually fA units). These values are corrected for lamp fluctuations, which enables good instrument-to-instrument comparability in commercially available instruments.

A portable PAS (EcoChem Analytics PAS 2000 CE) was used in this project. It was set to average data over a 10 s period, covering a range from 0 to 4000 fA with a resolution of 1 fA (estimated accuracy of 2 fA). The PAS limit of detection is about 10 fA, and it has a flow rate of 1.0 L min 1.

iv. Number concentration (Ultrafine Condensation Particle Counter—UCPC). 

The UCPC (Model 3025A, TSI Inc.) was used to obtain total particle number concentration. The UCPC counts particles in the range of 3–1000 nm, and operates by saturating the carrier air stream with a butyl alcohol vapor. The air stream is cooled which produces a supersaturation and causes vapor to condense onto the particles. The particles then rapidly grow to an optically detectable size (10–12 mm).

v. Size distribution (Scanning Mobility Particle Sizer— SMPS). 

The SMPS consists of a TSI 3071A Electrostatic Classifier and a TSI 3025A Ultrafine Condensation Particle Counter (UCPC). It was used to classify particles by an electrical mobility equivalent diameter. The SMPS was configured to cover the range of 8 to 300 nm in the high flow mode (10 L min-11 sheath air flow and 1.0 L min-1 aerosol flow). Data were collected using 60 s up scan and 30 s down scan times. Data were analyzed using version 3.2 of the TSI SMPS software. SMPS data presented here include the integrated total number concentration, surface area concentration, and geometric mean diameter. The SMPS active surface is calculated using expressions given by Pandis et al., Kasper et al., and Kasper et al. assuming spherical particles and that the species charging the particles is a hydrated proton which has a mean free path of 14.5 nm.51 This calculation should simulate the response of the DC although it does not include size dependent losses in either instrument.

B. Sampling strategy 

i. Sampling locations and occupational categories. 

The Twin Cities Metropolitan and the University of Minnesota transit systems together offer an environment in which both diesel and gasoline powered vehicles operate on a regular basis in various mixtures from almost all gasoline vehicles to nearly all diesel-fueled buses. This provided a unique opportunity to measure the different exposure scenarios.

Area and personal samples were collected depending upon the specific location and situation. Personal samples were collected in or near the breathing zone of the worker over a full work-shift. The PAS, DC, SMPS and UCPC were used to collect area samples as near to the breathing zone of the worker as possible. These data were collected in near real-time and downloaded to a computer.

Three occupational categories were selected for monitoring: bus drivers (personal samples), parking garage attendants (personal and area samples), and bus mechanics (personal and area samples). These were selected because they provided a mix of exposure scenarios based on the mix of gasoline and diesel exhaust in each scenario as well as the age of the aerosol, and they are representative of workers employed in the transportation industry. Table 1 summarizes the number of different types of measurements and their sampling durations for the three occupational categories. Table 1 shows the number of workshifts. For each workshift, there will be a large number of measurements gathered depending on the sampling frequency of the real-time instruments, and a single workshift average is then calculated (shown in Table 2). For example, for DC, Table 1 shows that 6 workshift average measurements were obtained in bus drivers (3 + 2 + 1). In Table 2, we see the summary statistics for the workshift whose average sampling time is given. We also see that these 6 workshifts correspond to 5595 10-second average measurements.

ii. Bus drivers. 

It was not always possible to obtain personal exposure measurements during this study because of objections raised by either management or union personnel. This was the case with the bus driver sampling. Two members of the project team acted as surrogates for the bus drivers and rode the Metropolitan Transit buses for 6–8 h wearing personal samplers.  It was assumed that their exposures would be similar to those of the bus drivers as they sat as close to the bus driver as possible. The surrogates got off the bus at the end of each route and waited for the next bus to take them back to their point of origin. The same bus typically would not run the same route all day long. In fact, each bus route would have several bus drivers in different buses throughout the day. The amount of time spent on the buses was recorded daily. Table 1 shows the average sampling time duration and standard deviation for each bus route. Four bus routes (16, 27, 75 and 80) were selected, prior to sampling, based on their traffic patterns.

The personal air monitoring of the bus routes comprised 39 air samples for EC divided, roughly equally, among four different bus routes in the Minneapolis-St. Paul metropolitan area (shown in Table 1). Route 16 traveled between downtown Minneapolis to downtown St. Paul where the bulk of the traffic was a mixture of spark ignition commuter automobiles with some diesel truck traffic. Route 27 traveled between downtown Minneapolis to suburban Anoka approximately 20 miles away. Traffic comprised spark ignition autos and diesel trucks that followed this route to various industrial sites. Route 75 was also a suburban corridor to downtown Minneapolis but with somewhat lighter diesel truck traffic. Route 80 went from downtown Minneapolis to the Mall of America, Bloomington, MN and comprised mostly of freeway traffic along an interstate highway. The Mall has a semi-enclosed bus terminal with constant bus traffic. Much of this traffic consisted of buses that were idling while unloading, loading, and waiting for passengers.

A total of 39 personal EC samples were collected on the four routes as shown in Table 1. Only a single set of portable, battery powered, PAS and DC instruments was available for this study. When used on the buses, the instruments were placed in the laps of the surrogate bus drivers with a common sampling tube extending into the breathing zone next to the EC sampler. The instruments stored a running 10 s average of the surface area or PAH concentration in memory. Data were downloaded daily to a computer. The PAS was used on eight bus sampling days and the DC was used on six bus sampling days. On these days, project personnel maintained a log of heavy and light-duty traffic encountered on the bus routes. Traffic counts were recorded every minute.

iv. Parking ramp attendants.

Parking ramp attendants at two ramps (the 4th Street ramp and the Washington Avenue ramp on the University of Minnesota campus) were monitored. They were exposed to a mixed aerosol, although the bulk of the parking ramp traffic was automobiles. The attendants were housed in booths with positive pressure ventilation and windows, but the intake vents for both ramps were located near potential sources of vehicle exhaust aerosol so workers may have been exposed to vehicular aerosol generated from sources that were not in the immediate vicinity of the parking ramp booth. The Washington Avenue ramp is one of the busiest campus parking facilities with constant in and out traffic. The 4th street ramp has periods with heavy commuter traffic in the morning and afternoon.

The EC samplers were placed near the breathing zone of the parking ramp attendants and close to the open window. There were several parking ramp attendants that worked in the booth throughout the day, to ensure that sufficient mass was collected for the chemical analysis of EC, it was decided to keep the cyclone near the breathing zone for several attendants rather than limit the sampling to one specific individual during a day. Typically, sampling began in the morning and ended in the late evening (average sampling times and standard deviations are shown in Table 1). In addition to EC samples, we also used the portable PAS and DC instruments as shown in Table 1. At the Washington Avenue ramp the SMPS and UCPC were also used. Twelve of the 23 EC samples were taken concurrently with the real-time measurements.

v. Bus garage mechanics. 

The garage phase of this study was selected to provide an environment thought to be dominated by diesel aerosol. Although exposure was primarily from diesel aerosol, cigarette smoke and other submicron aerosol such as welding fume may also have been present. The selected locations were the Camden garage in Minneapolis and the Cedar Avenue garage in Richfield, MN. The primary activity at both garages was diesel engine maintenance.

Instrumentation included EC samplers, the UCPC, and the portable PAS and DC samplers as shown in Table 1. Nineteen EC samples were taken at the Camden garage at three different locations. Seven samples were collected simultaneously with the real-time instruments located in the main garage area where most of the engines were maintained. The other nine were split between a maintenance bay that was separated from the main garage area and used by maintenance workers and a loft location that was one story above the instruments and used by clerical workers. The Cedar Avenue garage was relatively new, and the same instrumentation was used. Fifteen EC samples were obtained at the garage, and six of these samples were collected simultaneously with the real-time instruments.

RESULTS AND DISCUSSION

This section is divided into three parts, each part describing the exposure assessment results based on a different exposure metric-mass concentration, surface area concentration, and number concentration. Table 2 presents summary statistics for the three exposure groups using the different exposure metrics. We used Analysis of Variance (ANOVA) for statistical comparisons using the S-Plus software package (Insightful Corporation, 2002). 

A. Mass concentration measurements

Fig. 1 shows box plots of EC concentrations for all three occupational categories—bus drivers, parking ramp booth attendants, and garage mechanics. The specific locations at which measurements were made, i.e., four bus routes, two parking ramps, and two garages with three locations each further divide these categories. The line inside each of the boxes is the median, the ends of the boxes are the 25th and 75th percentiles, and the ends of the lines are the 10th and the 90th percentiles of the measurements. The solid circles are the 5th and the 95th percentiles. Each measurement is the average EC concentration over a work shift (typically 6–7 hours). A number of measurements in the ramps and buses were close to or below the limit of detection ~1 mg m 3). The three groups had statistically significant differences in exposures to EC as shown in Table 2 (p < 0.001). The least squares mean exposures and corresponding standard errors were 3.87 +- 0.60 mg m3 for garage mechanics, 1.98 +- 0.21 mg m 3 for bus drivers, and 1.10 +- 0.11 mg m 3 for ramp attendants. The highest levels of EC were seen in the maintenance garages that were dominated by diesel exhaust and the lowest levels were observed in the parking ramp booths that were dominated by spark ignition (gasoline) exhaust aerosol. The four bus routes had a hybrid of diesel/spark ignition exposure, and had intermediate levels of EC. The EC levels ranged (5th to 95th percentiles) from the method detection limit (MDL = 1 mg m3) to 5.6 mg m for the four bus routes, MDL to 2.30 mg m 3 in the ramps, and MDL to 17.7 mg m 3 in the garages.

Very low EC concentrations in the parking ramp booths are consistent with the facts that there was very little diesel traffic, that the booths were under positive pressure and the measurements were made inside the booth near the breathing zone of the attendants.

The Camden garage had statistically significant higher concentrations than the Cedar Avenue garage. The mean and standard errors of EC levels were 3.4 0.5 mg m 3 for Camden, and 1.1 0.5 mg m 3 for Cedar Avenue garage. Camden is an older garage with less ventilation. The Cedar Avenue garage is relatively new with better ventilation, which is reflected in the lower EC concentrations.

The results from the bus garages are consistent with a previous study conducted by NIOSH of trucking industry workers.52 In that study the geometric mean EC exposure was 3.8 mg m 3 for highway truck drivers, 4.0 mg m 3 for local truck drivers, 12.1 mg m 3 for bus mechanics, and 27.2 mg m 3 for dockworkers using diesel-powered forklift trucks. In the same study, background levels measured in a residential area and near the highway ranged from 1–3 mg m 3, similar to the levels we measured in the buses and parking ramps.

B. Surface area concentration measurements

As mentioned earlier, the diffusion charger (DC) measures the total surface area of the particles irrespective of their chemistry, while the photoelectric aerosol sensor (PAS) measures the surface adsorbed PAHs—a surrogate for the surface area of diesel exhaust aerosol specifically.

Fig. 2 shows a box plot comparison of the 10-second average DC measurements in buses, parking ramps, and bus garages. Table 2 shows that the arithmetic mean 10-second average surface area concentrations in parking ramps are significantly greater than those in bus garages, which in turn are significantly greater than those in buses. However, the much larger exposures in ramp attendants are due to a few large exposure occasions. This can be seen by comparing the GSDs of the exposure distributions for the three groups.  The large standard deviation for ramp attendants results from a few very high exposure conditions.  Table 2 also shows that the work shift average surface area concentrations (arithmetic as well as geometric means) for the three occupational groups are not much different from the 10-second averages. However, the three groups are not significantly different for work shift average surface area concentrations. As expected, the range of values for work shift average values is narrower than for 10-second averages.

Fig. 3 shows a box plot comparison of the 10-second average PAS measurements in buses, parking ramps, and bus garages. Table 2 shows that using 10-second average PAS measurements, the ramp booths had the highest arithmetic mean exposures of 98 nanograms (ng) PAH m 3 compared with 25 and 22 ng PAH m 3 for bus drivers and garage mechanics. The three groups had statistically significant differences in PAS measurements as shown in Table 2 (p o 0.001). As Fig. 3 shows, the distribution for the ramp attendants is skewed so that even though the three groups have similar geometric means, the arithmetic means are very different. As in the case of the DC measurements, the much larger exposures in ramp attendants are due to a few large exposure occasions. Again, this can be seen by comparing the GSDs of the exposure distributions for the three groups. Interestingly, if work shift average PAS measurements are considered (arithmetic as well as geometric means), then the ramp attendants have the lowest exposures and bus drivers have the highest exposures. The differences are still significantly different. Thus, the sample averaging times have a significant effect on the exposure rankings.

C. Number concentration measurements

At selected sampling locations particle number concentrations were measured using the UCPC, that provided total counts, and the SMPS that provided size segregated particle counts.

An integrated number concentration is obtained from the SMPS size distribution by summing the number of particles counted in each scan channel. The UCPC data are averaged o na 1-minute time base to match the time base of the SMPS. The UCPC concentrations are consistently higher than the SMPS concentrations. The difference in total number concentrations from the SMPS and the UCPC is explained by the different measurement range of the instruments. The UCPC is designed to detect particles as small as 3 nm, while the lower sizing limit of the SMPS as configured in this study is nominally 8 nm. The difference in total and integrated number concentrations between the UCPC and the SMPS is partially due the concentration of particles ranging between 3 and 8 nm in diameter. In addition, differences in the total and integrated particle number concentrations are often attributed to diffusive particle loss inside of the SMPS column. With the SMPS configuration used here, such losses were significant for particles less than 50 nm. 

Fig. 4 shows a box plot comparison of the 1-minute average aerosol number concentration (as measured by the UCPC)inside and outside the parking ramp booth and in the two garages. The concentrations inside the booth ranged between 1x104 to 2.5x104 particles cm-3. As is to be expected, the number concentrations inside the booth (which is where the workers would be exposed) are much lower than outside. This is because the booth is under positive air pressure. The concentrations outside the booth are quite similar to the levels measured in the two garages, and ranged between 2.5x104 to2.25x105particles cm3. The levels in garages were significantly higher than the levels inside the booths. Table 2 shows that the arithmetic and geometric mean levels in garages are 5 to 6 times the levels in the booths. Thus the exposure distributions of the two corresponding cohorts (ramp booth attendant sand garage mechanics) to 1-minute average particle number concentrations were significantly different (p  <0.001). The difference is only marginally significant for work shift averages(p<0.076). No significant differences were noted between the levels in the two garages. 

Fig. 5 shows a plot of the integrated SMPS surface area concentration and the corresponding 1-minute average diffusion charger (DC) reading at the Washington ramp. The diffusion charger provides a direct measurement of surface area, while the SMPS provides the size distribution of the aerosol by particle number. The surface area of each size bin is calculated using the midpoint diameter of the bin range and the particle count in that bin. The total surface area is determined by adding up the surface areas of all bins. All days show a very close match between the measurements of the two instruments with the exception of the first half of February 28. This is most likely due to a sampling system malfunction on the first day of sampling. Fig. 6 is a comparison between the SMPS calculated geometric mean aerosol number diameter, Dg, and data from the UCPC and DC. The Dg is determined from an SMPS particle size distribution using the following relationship:

where Di and ni is the particle diameter and number concentration of channel i respectively, and N is the total number concentration.The total number count N and the total surface area S are used to determine the diameter of average surface, Das as

where S is the total surface area (mm2cm3) measured by the DC and N is the total number concentration (# cm3) mea-sured by the UCPC. Thus, not only do the DC and UCPC give real time concentrations, combined, they also give a meaningful representation of the particle size in real time.
