Measuring In-Cabin School Bus Tailpipe and Crankcase PM_2.5: A New Dual Tracer Method

ABSTRACT

Exposures of occupants in school buses to on-road vehicle emissions, including emissions from the bus itself, can be substantially greater than those in outdoor settings. A dual tracer method was developed and applied to two school buses in Seattle in 2005 to quantify in-cabin fine particulate matter (PM_2.5) concentrations attributable to the buses' diesel engine tailpipe (DPM_tp) and crankcase vent (PM_ck) emissions. The new method avoids the problem of differentiating bus emissions from chemically identical emissions of other vehicles by using a fuel-based organometallic iridium tracer for engine exhaust and by adding deuterated hexatriacontane to engine oil. Source testing results showed consistent PM:tracer ratios for the primary tracer for each type of emissions. Comparisons of the PM:tracer ratios indicated that there was a small amount of unburned lubricating oil emitted from the tailpipe; however, virtually no diesel fuel combustion products were found in the crankcase emissions. For the limited testing conducted here, although PM_ck emission rates (averages of 0.028 and 0.099 g/km for the two buses) were lower than those from the tailpipe (0.18 and 0.14 g/km), in-cabin PM_ck concentrations averaging 6.8 μg/m³ were higher than DPM_tp (0.91 μg/m³ average). In-cabin DPM_tp and PM_ck concentrations were significantly higher with bus windows closed (1.4 and 12 μg/m³, respectively) as compared with open (0.44 and 1.3 μg/m³, respectively). For comparison, average closed- and open-window in-cabin total PM_2.5 concentrations were 26 and 12 μg/m³, respectively. Despite the relatively short in-cabin sampling times, very high sensitivities were achieved, with detection limits of 0.002 μg/m³ for DPM_tp and 0.05 μg/m³ for PM_ck.

IMPLICATIONS

PM_2.5 measurements in two Seattle school buses showed average concentrations of 26 and 12 μg/m³ with windows closed and open, respectively. Virtually all PM_2.5 was car bonaceous. Tracer measurements showed that bus self-pollution contributed approximately 50% of total PM_2.5 concentrations with windows closed and 15% with windows open, with over three-quarters of these contributions attributed to crankcase emissions. Maintaining ventilation in buses clearly reduces total PM_2.5 exposures and that from the buses' own emissions. The dual tracer method now offers researchers a new technique for explicit identification of single source contributions in settings with multiple sources of carbonaceous emissions.

INTRODUCTION

In-vehicle exposures to vehicle-related emissions of fine particulate matter (PM_2.5) and other species inside vehicles have been shown to be high relative to urban outdoor environments.1,2 Particular emphasis has been given to concentrations of diesel exhaust particulate matter (PM) inside school buses that are attributable to the buses' own exhaust. Such source attribution efforts are difficult in purely observational studies because of the presence of emissions from other vehicles, which are chemically identical to those of the bus, and the presence of chemically similar PM emitted by nonvehicular sources. A method was therefore sought that could allow definitive identification of the emissions of a single bus. Previous studies3–8 have produced estimates of exhaust PM inside school buses that are attributable to the buses' own exhaust ranging from 0.2 to as much as 20 μg/m³. With the exception of Ireson et al.,5 these studies inferred bus exhaust contributions from measurements of species emitted by diesel engines and other sources and did not measure the emission rates of the buses tested. They were therefore unable to explicitly identify the buses' contribution relative to other sources of the marker species. One study did use a tracer in exhaust, but it did not attempt to maintain a constant ratio of tracer to emissions.6,7 As a result, although suggestive, these studies do not definitively quantify bus emission contributions.

To allow specific identification of a single bus' emissions, a previous study5,9 extended the use of a fuel-based iridium (Ir) tracer method for diesel tailpipe exhaust (DPM_tp)10,11 by conducting chassis dynamometer measurements of the tailpipe emission rates of PM_2.5 and particle-bound Ir. The mass ratio of these two species in the exhaust (DPM_tp:Ir) was used in conjunction with in-cabin Ir measurements to calculate the in-cabin concentration of DPM_tp. The study presented here, an initial phase of a larger monitoring campaign,12 tested two buses using the Ir tracer and added a new second tracer to the bus lubricating oil to specifically quantify crankcase PM emission contributions. An on-board dilution tunnel system was used to measure PM and tracer emission rates from the tailpipe and crankcase under actual operating conditions. The primary objective of this paper is to document this new method for use and extension by others. Results for two buses are presented to demonstrate the capabilities of the method.

Historically, relatively little attention has been given to crankcase contributions, which were generally considered to be significantly lower than tailpipe emissions. Hill et al.8 conducted a series of experiments using nonspecific optical methods to measure PM in individual buses with and without retrofit PM_ck controls. They showed that crankcase emissions may contribute significantly to in-cabin concentrations of particles at the upper end of the PM_2.5 size range. To accurately quantify this contribution, a deuterated al-kane (d-alkane) was used as an explicit marker for lubricating oil. Emissions testing and analysis of crankcase (PM_ck) particulate emissions13 and DPM_tp were conducted to determine the PM mass-to-tracer ratios for both Ir and the d-alkane. In-cabin concentrations attributable to the crankcase and tailpipe emissions were quantified using direct measurements of the two tracers during routine bus operations.

Design of the study was guided by two primary objectives: (1) to unambiguously quantify DPM_tp and PM_ck concentrations in school buses under typical operating conditions, and (2) to ensure adequate sensitivity to measure ambient contributions of tailpipe and crankcase emissions at levels well below the minimum concentrations of interest.

EXPERIMENTAL PROCEDURES

A sampling campaign was conducted during summer 2005 to evaluate several independent methods for estimating penetration of bus emissions into the cabin, as described by Liu et al.12 The tracer study presented here was conducted as part of that campaign to provide the in-cabin DPM_tp and PM_ck concentrations against which the other methods were evaluated. In-cabin sampling and source testing were conducted using two conventional (front-engine) diesel school buses—a 2003 Bluebird and a 2000 International. Odometer readings for the buses were 48,900 and 78,400 mi, respectively, and both were equipped with diesel oxidation catalysts. Buses were taken in “as-received” condition, and no repairs (e.g., door or window seal replacement) were made, even when some evidence of wear was observed. The buses operated over a 10-km Seattle school bus route with each complete in-cabin sampling run consisting of forward and reverse routes. Typical ambient PM_2.5 concentrations in the study area are lower than many urban areas, ranging from approximately 5 to 15 μg/m³. Each route was of approximately 1 hr in duration. Drivers were instructed to follow their standard practices, including the number and duration of door-opening events to on- and off load students. This route is predominately on residential streets, with little arterial and no freeway driving, and average speeds were less than 15 km/hr. There were frequent periods of low-speed, high engine load operations. Emissions testing was conducted on route segments selected to characterize, as closely as possible, the duty cycle of the school buses during in-cabin sampling.

Tracers, Equipment, and Sampling Media

Several studies have used metals as intentional fuel-based tracers for atmospheric dispersion and transport and exposure experiments.14–17 Wu et al.10 demon-strated the use of Ir in fuel as a means of investigating the exposures of students to diesel exhaust from the public transit fleet used by urban high school students. A fuel-based Ir tracer was initially selected by Ireson et al.5 for three primary reasons: (1) individual Ir atoms are bound in exhaust particles and emitted at the same rate as fuel combustion, which is closely related to the diesel PM (DPM) emission rate; (2) the natural abundance of Ir is extremely low, and it is not commonly used in commercial or industrial processes that lead to its emission; and (3) it can be quantified in amounts as low as 50 fg (1 fg = 10⁻¹⁵g) using instrumental neutron activation analysis (INAA).18

In the selection of a tracer for lubricating oil, trace metals were considered but rejected because it was unclear if a metallic oil additive would behave similarly to the naturally occurring components of lubricating oil. PM_ck emissions were assumed to consist primarily of the higher molecular weight hydrocarbons in oil that were aerosolized by blow-by past the piston rings or valve guides or were vaporized and then condensed into fine particles. Deuterated organic compounds are commonly used in gas chromatography (GC)/mass spectroscopy (MS) because (1) they do not occur naturally, (2) they elute from GC columns at nearly the same time as their nondeuterated analogs, and (3) they are readily differentiated from their nondeuterated analogs by MS. Normal (i.e., unbranched, straight carbon chain) hexatriacontane (n-C₃₆H₇₄) is a naturally occurring higher molecular weight component of lubricating oils, and its fully deuterated form, n-C₃₆D₇₄, was selected as the tracer.

Analyses of source and in-cabin samples included gravimetric mass, organic carbon (OC) and elemental carbon (EC), INAA for Ir, and GC/MS for the d-alkane. These analyses require Teflon (for gravimetric mass and INAA) and quartz (for particulate carbon and GC/MS) filter media. Both media were used for all types of samples.

The nominal sensitivity achievable for DPM_tp on an integrated filter sample depends on several factors, including the Ir concentration in background PM, the analytical sensitivity of INAA, the sample volume, and the DPM_tp:Ir ratio. The latter is determined in part by the DPM emission rate of the engine and the spike level in the fuel. Previous studies10,11 showed that urban ambient concentrations of Ir were typically on the order of 100 fg/m³. High flow rate samplers designed and constructed at the University of Maryland were used to increase the sensitivity achievable for relatively short-duration samples. This “UMd sampler” uses an inlet impactor and 47-mm filter holders. The impactor inlet is designed to achieve a particle size cut point of approximately 2.3 μm at a flow rate of 120 L/min.19 Sample flow rates ranged from 98 to 120 L/min (average 114 L/min), resulting in actual sample-specific cut points between 2.3 and 2.5 μm. On the basis of typical school bus fuel consumption rates, emission rates, and speeds, a conservative nominal spike level of 2.6 mg of Ir per liter of fuel was selected to achieve a target sensitivity of 0.001 μg/m³ DPM_tp.

The crankcase d-alkane tracer spike level selected was 100 g of the tracer for the lubricating oil volume of typical school bus engines (∼22 L). This level was selected based on preliminary calculations and assumptions regarding the expected mass fraction of tracer in crankcase PM emissions. For 1-hr integrated samples at a flow rate of 120 L/min and an assumed nominal abundance of tracer in crankcase emissions, the anticipated sensitivity for ambient PM_ck measurements was approximately 0.1 μg/m³.

Source testing of vehicle emissions has traditionally been conducted using stationary engine or chassis dynamometers with large constant volume sampling (CVS) dilution tunnels. For this study, an alternative approach was used that allowed collection of PM emissions samples under the same bus operating conditions as those of the in-cabin sampling. Tailpipe and crankcase vent emissions testing was conducted using the Ride Along-Vehicle Emissions Measurement (RAVEM) system developed by Engine Fuel & Emissions Engineering, Inc. The RAVEM is a portable dilution tunnel system based on proportional partial-flow CVS from the vehicle exhaust pipe. Isokinetic sampling in the tailpipe and subsequent dilution produces a diluted exhaust stream comparable to that of full-flow CVS systems used in chassis and engine dynamometer testing, but it allows measurements to be made on-board a moving vehicle. Teflon and quartz filter samples were collected from the dilution tunnel using University Research Glass PM_2.5 cyclone inlets and 47-mm filter holders operated at their design flow rate of 16.7 L/min. Comparative testing of the RAVEM with full-flow CVS systems shows good general agreement between the two methods.20–22 For carbon dioxide (CO₂) and oxides of nitrogen emissions, the agreement is generally very close. For PM emissions, the differences between RAVEM results and those of specific full-flow CVS laboratories are generally between 0 and 25%, which is less than the range of differences between different full-flow CVS systems in round-robin testing.23 Although the RAVEM shows some tendency toward under-reporting of PM emission rates, the key calculations from RAVEM samples are the PM: tracer ratios, which are based on analyses of the same filter. As discussed later in this paper, this minimizes potential bias in tracer-based estimates of in-cabin DPM_tp and PM_ck concentrations that might result from calculations based on concurrent measurements from different media or methods.

For testing, buses were operated on ultralow sulfur diesel fuel to which a solution of tris(norbornadiene)iridium-(III)acetylacetonate in toluene was added. This tracer is an organometallic chelate that is readily soluble in toluene and soluble in diesel fuel. Because single Ir atoms in a positive oxidation state are bound in the complex and distributed uniformly through the fuel, the Ir atoms will be captured and bound within the fine carbonaceous particles formed in the diesel combustion chamber. For each bus, a solution of approximately 2 g of tracer compound (0.66 g of Ir) in 200 mL of toluene was added to the nearly empty fuel tank (expected capacity of 190–280 L), which was then refueled to mix the tracer. The bus was then driven for several miles to mix the spiked fuel through the system before testing.

The n-C₃₆D₇₄ tracer (100 g) was dissolved in 19 L of new motor oil and added to the first test bus, with new oil added as needed to bring the oil to its normal operating level (20 –25 L). At the conclusion of testing of the first bus, the spiked oil was drained and transferred to the second bus, again with the addition of new oil to bring the level up to its normal operating range.

In-Cabin On-Road Sampling

For each bus, in-cabin sampling was conducted during two sets of forward- and reverse-route pairs with windows closed and another two sets with windows open. The windows-closed tests were conducted with all windows, vents, and bus ventilation system closed and turned off. Integrated filter samples were collected inside of the bus and inside of a lead vehicle (LV) that traveled the same route a few minutes ahead of the bus. Samples from the LV could then be used to determine the background level (if any) of the tracers. Near-roadway PM_2.5 samples were collected and analyzed by INAA several months before the tracer study to verify that background Ir levels were sufficiently low so as not to compromise the desired sensitivity for DPM_tp.

In-bus and LV integrated samples were collected using the UMd samplers. Power for the sampling pumps was provided by 12 VDC inverters and banks of four lead-acid marine batteries. Concurrent samples were collected on Teflon and prefired quartz filters. Two pairs of samplers were operated on the bus, with one pair collecting samples for the approximately 1-hr duration of each forward and reverse route and the other pair collecting samples for the approximately 2-hr duration of full forward and reverse routes. Thus, for each bus, four Teflon/quartz filter pairs were collected for each 2-hr forward/reverse route run, and eight filter pairs were collected during the 1-hr forward and reverse routes. A single pair of samplers was operated in the LV, with a pair of samples collected for each route (forward or reverse). For bus 1, the samplers for 1-hr forward and reverse runs were mounted in the second and third seats from the rear of the bus on the left side, and the sampler pair for 2-hr samples was mounted in the second seat from the rear on the right side. For bus 2, the 1-hr samplers were mounted in the second and fourth seats from the front on the right side, and the 2-hr sampler pair was mounted in the third seat from the front on the right side. This change was made for bus 2 to bring these samplers closer to the other sampling equipment that was part of the larger study.12 Filters were stored at −20 °C between sampling and laboratory analyses. Teflon filters were analyzed gravimetrically and by INAA. Quartz filters were analyzed for OC and EC by thermal optical reflectance (TOR) for a punch taken from each filter and for a range of alkanes (including the d-alkane tracer), hopanes, steranes, polycyclic aromatic hydrocarbons, and nitroarenes by extraction and GC/MS.13 Teflon and quartz field blanks were also collected and analyzed.

Source Testing

Tailpipe and crankcase vent emissions testing using the RAVEM was conducted after the conclusion of in-cabin sampling to avoid the possibility of inadvertent contamination of the bus interior with either tracer. Tail-pipe sampling was conducted using the standard RAVEM proportional partial flow approach, with a small fraction of the total exhaust being extracted into the dilution tunnel. The low flow rate from crankcase vents precludes this approach, so the PM_ck samples were collected using the RAVEM as a conventional full-flow CVS system. The RAVEM dilution tunnel module was mounted on the front bumper of each bus, and the full-flow CVS system from the crankcase vent was ducted directly into the tunnel. For each test run, Teflon and quartz samples were collected from the dilution tunnel. In addition, a third sample (alternately collected on Teflon and quartz filters) of the tunnel dilution air was collected upstream of the emission injection point in the tunnel to serve as a field blank. All filters were stored at −20 °C between sample collection and laboratory analyses (gravimetry and INAA for Teflon, OC/EC and GC/MS for quartz).

For each bus, samples were collected for three runs for tailpipe and crankcase emissions. Testing was conducted on three segments of the bus routes used in on-road sampling that were selected as being representative of average driving conditions over the full length of the two routes. Tailpipe emissions test runs ranged from 30 to 40 min each. Tailpipe tests ducted 0.6% of total exhaust flow into the tunnel with a dilution flow rate of 600 L/min, yielding a dilution equivalent to a full-flow CVS operating at 100 m³/min. Because this was the first test of its kind, crankcase emissions test durations were limited to 15–20 min for the first bus to prevent overloading of filters but were increased to 30 min each for the second bus. Tunnel dilution flows were set to 1500 L/min for the first two crankcase tests but were reduced to 800 L/min after it was determined that filter loading was not a problem.

Calculation of In-Cabin DPM_tp and PM_ck Concentrations

The use of intentional tracers as quantitative markers for tailpipe and crankcase PM emissions allows direct calculation of DPM_tp and PM_ck concentrations from in-cabin measurements. The mass ratios of these two parameters to the two tracers provide, in effect, a source profile for the two sources. Because there are no other sources of the tracers, the calculation of the two parameters' concentrations can be accomplished by solving a linear two-equation system:

(1)

(2)

The known values are the measured in-cabin concentrations of the tracers [Ir] and [d-alkane] and the mass:tracer ratios DPM_tp :Ir, DPM_tp :d-alkane, PM_ck :Ir, and PM_ck :dalkane. This approach explicitly accounts for the amounts of each tracer contributed by PM from both emission sources.

RESULTS

Source Testing

Table 1 shows the summary results of emission testing of the two buses for tailpipe and crankcase emissions. The observed DPM_tp ranged from 0.13 to 0.20 g/km, which is typical of expected rates for buses of this age and mileage. Bus 1 PM_ck averaged 0.028 g/km, or approximately 16% of the tailpipe emission rate. Crankcase emission rates for bus 2 were approximately 3 times higher, averaging 0.099 g/km, approximately 70% of the bus' tailpipe emission rate. The PM:tracer ratios were quite consistent for the primary tracer for each emissions source. For DPM_tp:Ir ratios, the coefficients of variation (CV; the relative standard deviation across three samples for each bus) were 8%, and for the PM_ck:d-alkane ratio the CVs were 5% and 7%. For the two buses, the average DPM_tp:d-alkane ratios were 10 and 24 times higher than the corresponding PM_ck:d-alkane ratio, and the PM_ck:Ir ratios were 500 and 1200 times higher than the corresponding DPM_tp:Ir ratios. This suggests that, although there is some small amount of unburned lubricating oil being emitted from the tailpipe (in the range of 4–9% of mass emissions), there are virtually no diesel fuel combustion products in the crankcase emissions (<0.2%).

Table 1a. Tailpipe PM emission rates and PM:tracer mass ratios

Bus	Speed, km/hr (CV %)	Fuel Use, km/L (CV %)	DPMtp, g/km (CV %)	DPMtp:Ir, g:g (CV %)	DPMtp:d-alkane, g:g (CV %)
1	14.4 (8)	2.22 (7)	0.18 (13)	1045 (8)	6939 (57)
2	14.1 (3)	2.04 (7)	0.14 (14)	975 (8)	4082 (24)
Notes: Data shown are averages of three runs each.

Table 1b. Crankcase PM emission rates and PM:tracer mass ratios

Bus	Speed, km/hr (CV %)	PMck, g/km (CV %)	PMck:Ir, g:g (CV %)	PMck:d-alkane, g/g (CV %)
1	12.0 (21)	0.028 (7)	5.69 × 10^5(56)	283 (5)
2	13.5 (23)	0.099 (33)	1.21 × 10^6 (4)	360 (7)
Notes: Data shown are averages of three runs each.

For most sample pairs, there was good agreement between gravimetric and carbonaceous mass data. There was some evidence of vacuum-side sampling manifold leaks for 4 of the 24 total samples: 1 Teflon tailpipe sample, 1 quartz tailpipe sample, and 2 quartz crankcase samples. However, the effects of these problems were minimal because paired samples and observed consistency between filter pairs provided acceptable surrogates for the suspect data. The gravimetric mass for one bus 1 run showed tailpipe emission rates approximately one-half of the concurrent carbonaceous mass (estimated as EC + 1.2 × OC to reflect mass associated with hydrogen and other element fractions of OC). Other tailpipe tests showed gravimetric mass to be consistently approximately 7% higher than the carbonaceous mass. To avoid underestimating PM emissions for this run, the mass emission rate was estimated from the carbonaceous mass. The two suspect quartz crankcase samples had carbonaceous mass from 35 to 70% lower than gravimetric mass. All other crankcase samples showed gravimetric mass to be within 10% of the carbonaceous mass. Manifold leaks did not affect the PM_ck emission rates, which were determined from gravimetric analyses of Teflon filters. Although these possible leaks may have resulted in a sampler cut point larger than 2.5 μm, they are not expected to compromise the DPM_tp:Ir or PM_ck:d-alkane because these were based on gravimetric and INAA analysis of the same Teflon filter and on OC/EC and GC/MS analysis of the same quartz filters. It is also possible that OC could include positive artifact mass,24 resulting in some overestimation of actual carbonaceous mass.

The mass ratios of PM emissions to the tracers should ideally be calculated based on laboratory analyses of the same filter to avoid introducing errors because of sample volume or particle size cut point differences for two concurrent samples. PM:Ir ratios were calculated from the gravimetric and INAA analyses of individual Teflon filters. Because quartz filters were not analyzed gravimetrically, the PM mass loading of the quartz filters was calculated from the OC/EC data as the carbonaceous mass multiplied by the average ratio of gravimetric to carbonaceous mass. As noted above, the gravimetric:carbonaceous mass ratios averaged 1.07 for tailpipe sample pairs and 1.04 for crankcase samples. DPM_tp:d-alkane ratios were calculated by multiplying carbonaceous mass:tracer ratios by 1.07. PM_ck:d-alkane ratios were calculated by multiplying carbonaceous mass:tracer ratios by 1.04.

The PM:tracer mass ratios were derived from several measurements. For the PM:tracer mass ratios, the CVs for the primary tracer ratios (DPM_tp:Ir and PM_ck:dalkane) are indicative of the level of uncertainty carried in these parameters. DPM_tp:Ir CVs were 8% for both buses, and the PM_ck:d-alkane CVs were 4 and 7% for the two buses. The gravimetric, INAA, and GC/MS analyses for source samples reported analytical uncertainties ranging from less than 1% (gravimetry) to 15–20% (dalkane). Reported EC and OC measurement uncertainties for source test filters were between 10 and 15%, and Ir measurement uncertainty was approximately 5%.

Fuel samples collected from each bus at the conclusion of testing were analyzed by INAA. This analysis showed the Ir spike level in both buses to be approximately 3.2 mg/L. Source tests for both buses showed the Ir emission rate to be only approximately 0.32 mg/L. Most of the Ir in the fuel (∼90%) appears to be captured in the engines or exhaust systems (which included diesel oxidation catalyst emission controls). This effect was noted, although to a lesser degree, in the single tracer study; however, that bus was not equipped with a diesel oxidation catalyst.5 There are several possible mechanisms by which Ir could be captured, including being trapped in lubricating oil after making contact with cylinder walls; impaction on turbocharger blades; deposition on exhaust valves, manifolds, or exhaust pipes; and capture by the diesel oxidation catalyst. Once captured, Ir would not be collected on exhaust samples unless it were re-emitted in the PM_2.5 size range.

The potential effect of this Ir loss cannot be definitively determined but would be expected to be small. It is possible that the DPM_tp:Ir ratio varies with engine operating mode on a short-term basis. However, the time scale of engine operating transients is small (on the order of seconds), whereas the time scales of dispersion processes outside of the bus are substantially longer. Because the PM:tracer ratios were determined under typical operating conditions, such variability would be expected to have little effect on longer-term averages. Also, because the DPM_tp:Ir ratio was empirically determined from source testing at the tailpipe, the loss of Ir does not affect the accuracy of estimates of DPM_tp derived from on-road test data, and the use of such average ratios is consistent with current practices in receptor modeling.

In-Cabin Sampling

Table 2 presents the mass- and tracer-based results for all samples, and Figure 1 shows the time-weighted run average in-bus concentrations. Because of time constraints, the Teflon and quartz filter pair for the reverse portion of one run (bus 1 run 4) could not be prepared, so two sample pairs were collected for forward and reverse routes.

Figure 1. In-bus concentrations of PM_2.5, OC/EC, DPM_tp, and PM_ck.

Table 2. Bus 1 in-cabin PM_2.5, carbonaceous PM, DPM_tp, and PM_ck concentrations (μg/m³)

Run and Direction	Windows	PM2.5		EC + 1.2 ×OC		DPMtp		PMck
		2-hr^a	1-hr^a	2-hr	1-hr	2-hr	1-hr	2-hr	1-hr
1 F^a	C	14.8	19.1	15.8	24.3	0.24	0.54	9.8	6.8
1 R^a	C		23.0		22.6		0.29		10.9
2 F	C	10.4^b	60.2	22.9	26.0	0.36^b	0.67	12.5	9.3
2 R	C		34.9		38.0		0.29		10.1
Average closed	–	14.8	34.1	19.4	27.8	0.24	0.43	11.1	9.4
3 F	O	13.8	17.5	7.5	45.6^c	0.17	0.29	0.5	0.6
3R	O		14.0		12.3		0.18		0.4
4 F & R^d	O	13.8	13.7^d	8.2	9.9^d	0.50	0.43^d	0.2	0.4^d
Average opened	–	13.8	14.8	7.9	20.0	0.34	0.33	0.4	0.5
Notes: C = closed, O = open. ^aEach run consisted of a forward (F) route (from residences to schools) and a reverse (R) route (from schools to residences). A nominal 2-hr duration sample set was collected over the course of a run, as well as a nominal 1-hr duration sample set during the forward and reverse portions. ^bSuspected leak; data not included in calculated averages. ^cOne OC fraction for this sample is suspect, but it has been included in the average shown. ^dSampling during run 4 included only two 2-hr sample sets.

The observed PM_2.5 concentrations in the buses and the LV averaged between approximately 14 and 22 μg/m³. This is substantially lower than in-vehicle concentrations reported in other areas (some in excess of 100 μg/m³).1–3,5 This is certainly due in part to the intentional selection of a light-traffic, residential route and the lower urban background concentrations in Seattle as compared with other areas. PM_2.5 concentrations were roughly comparable between buses and between the buses and the LV. However, PM_2.5 concentrations were lower in the bus than the LV when windows were open and higher when windows were closed. Concentrations of EC, OC, and other parameters exhibit different relationships, as discussed by Liu et al.12 Tracer measurements on samples collected in the LV were insignificant with the exception of a 0.05-ng/m³ d-alkane concentration observed during the first bus 2 run. This was assumed to be the result of the bus at some point being close enough and upwind of the LV to generate a measurable level of tracer, so this value was not treated as actual background and was not subtracted during the course of the tracer analyses.

Tables 2 and 3 present results for bus 1 and 2, respectively, and the averages of all tests are presented in Table 4. The right-most columns in these tables present the calculated carbonaceous mass, DPM_tp, and PM_ck. The carbonaceous PM_2.5 concentrations are close to, and in some cases higher than, the corresponding gravimetric mass. This is particularly apparent for the 2-hr samples for run 2 of bus 1 and for run 2 of bus 2. The generally close agreement between carbonaceous mass and PM_2.5, and the generally large differences between carbonaceous mass and the mass explained by DPM_tp and PM_ck, suggest that the dominant sources of in-cabin PM_2.5 concentrations are emissions from other vehicles or other sources of carbonaceous material.

Table 3. Bus 2 in-cabin PM_2.5, carbonaceous PM, DPM_tp, and PM_ck concentrations (μg/m³)

Run and Direction	Windows	PM2.5		EC + 1.2 ×OC		DPMtp		PMck
		2-hr^a	1-hr^a	2-hr	1-hr	2-hr	1-hr	2-hr	1-hr
1 F^a	C	28.4	28.1	31.2	36.1	2.17	2.22	20.8	22.5
1 R^a	C		26.8		33.0		2.75		15.9
2 F	C	9.4^b	33.8	19.9	24.9	1.03^b	3.66	8.9	7.9
2 R	C		22.9		30.4		2.16		6.9
Average closed	–	28.4	28.0	25.5	31.1	2.17	2.71	14.8	13.3
3 F	O	10.5	8.7	8.1	10.2	0.43	0.54	2.5	2.1
3 R	O		2.5		11.6		0.30		1.8
4 F	O	12.5	11.0	12.7	12.2	0.67	0.81	2.1	2.4
4 R	O		16.0		21.0		0.30		2.9
Average opened	–	11.5	9.1	10.4	13.0	0.55	0.53	2.3	2.2
Notes: ^aEach run consisted of a forward (F) route (from residences to schools) and a reverse (R) route (from schools to residences). A nominal 2-hr duration sample set was collected over the course of a run, as well as a nominal 1-hr duration sample set during the forward and reverse portions. ^bSuspected leak; data not included in calculated averages.

Table 4. Average in-cabin PM_2.5, carbonaceous PM, DPM_tp, and PM_ck concentrations (μg/m³)

Bus 1 and Bus 2	PM2.5	EC + 1.2 × OC	DPMtp	PMck
Average closed	26	26	1.4	12
Average opened	12	13	0.44	1.3
Average all	19	19	0.91	6.8

During testing, some difficulty was reported in loading and sealing filter cassettes, and it is possible that undetected leaks in the cassettes resulted. It is also possible that the carbonaceous mass estimates are influenced by positive OC artifact.24 The bus 1 run 3 forward-route sample exhibited an abnormally high value for one of the four OC fractions, suggesting it may have been affected by contamination. Other than these three cases, PM_2.5 and carbonaceous mass are in general agreement for all filter pairs. The 2-hr sample concentrations are often lower but in some cases higher than those of the corresponding 1-hr samples. There were not enough sampling locations in each bus to establish consistent patterns, but it was suspected that the observed differences were related to differences among the pathways by which different pollutants enter the bus. For example, because the PM_ck release point is at the front of the bus, its concentration gradients within the bus will differ from those of DPM_tp (emitted at the rear bumper of the bus) and PM from other sources. This suggests that the different locations of the sampling equipment within the buses may have contributed to the observed differences between the 2-hr and the two 1-hr sampler concentrations.

Individual DPM_tp sample concentrations in this study ranged from 0.24 to 3.7 μg/m³ with windows closed (averages of the 2-hr samples and two 1-hr samples across runs were of 0.33 and 2.6 μg/m³, respectively) and from 0.17 to 0.81 μg/m³ with windows open (averages of the 2-hr samples and two 1-hr samples across runs were 0.33 and 0.54 μg/m³, respectively). The averages of all DPM_tp concentrations for the two buses were 0.33 and 1.5 μg/m³. PM_ck concentrations are substantially higher than those of DPM_tp for both buses. Hill et al.8 suggested this result. However, the magnitude is somewhat surprising considering the relative mass emission rates of the two sources, particularly for bus 1 (see Table 1). With windows closed, PM_ck concentrations ranged from 6.8 to 23 μg/m³ (averages of 10 and 14 μg/m³ for bus 1 and 2, respectively), and from 0.2 to 2.9 μg/m³ with windows open (averages of 0.4 and 2.3 μg/m³ for bus 1 and 2, respectively). The overall average PM_ck concentrations for the two buses were 5.3 and 8.2 μg/m³—approximately 16 and 5 times greater than the corresponding DPM_tp concentrations, respectively. Factors likely to contribute to this result include (1) the crankcase vent location inside of the engine compartment, below the engine; (2) the tailpipe location at the rear of the bus; and (3) the substantial positive air pressure gradient from inside of the engine compartment to inside of the bus while driving, resulting in enhanced flow of air from the engine compartment through the firewall in spaces left for engine and bus controls. The passenger door at the front right of the cabin is also a possible point of entry for crankcase emissions that is well away from the tailpipe. Intrusion via the firewall and the passenger door would also be likely to carry emissions from vehicles ahead of the bus and re-entrained road dust.

In-cabin PM_2.5 concentrations from all sources averaged 10 μg/m³ with windows open and 26 μg/m³ with windows closed. As expected, these over-roadway concentrations are higher than typical ambient concentrations in Seattle but are low when compared with other large urban areas. For example, in-bus PM_2.5 concentrations during the original Los Angeles Ir study5 averaged 72 μg/m³.

As for source testing, PM_2.5, OC/EC, DPM_tp, and PM_ck concentrations in the buses are subject to uncertainties in PM:tracer mass ratios, total sample volumes, and the individual analyses. Again, a 5% uncertainty can be assumed for sample volumes, and gravimetric and INAA analysis reported uncertainties were also approximately 5%. OC uncertainties were between 5 and 10%, and EC uncertainties were between 10 and 15%.13 Uncertainties reported for d-alkane were 15–20% for most measurements,13 increasing to 25% for some of the lower concentrations (open-window runs).

A question has been raised regarding the validity of the Ir tracer method because of possible suppression of DPM formation by Ir in the fuel because no emissions testing of the bus was conducted with unspiked fuel.25 However, Clark et al.26 conducted engine dynamometer testing using fuel with and without the Ir tracer and found that the tracer had no measurable effect on DPM emission rates (the final report is published at http://secure.awma.org/onlinelibrary/samples/10.3155-1047-3289.61.5.494_supplmaterial.pdf). Further, although busspecific DPM_tp:Ir ratios were used to calculate in-cabin DPM_tp concentrations for the specific buses tested, the Ir measurements themselves are definitive indicators of the amount of exhaust intrusion into the buses. They can be used to calculate the expected in-cabin DPM_tp concentrations for any assumed DPM emission rate.

DISCUSSION

The results of this study are consistent with those of previous studies in showing that the highest in-cabin PM_2.5 concentrations occur when buses are operated with windows closed. However, the DPM_tp results of this and the previous Ir study differ from other results with regard to estimates of exhaust intrusion into school bus cabins. The 0.33- and 1.5-μg/m³ DPM_tp concentrations found in the Seattle study are higher than the 0.22-μg/m³ average observed in the first Ir tracer study of a single Los Angeles bus.5 Nevertheless, they are still substantially lower than estimates derived using concentrations of nonspecific markers or a tracer for which the relationship to DPM_tp was variable and unquantified.4,7 This discrepancy is attributed to the inadvertent misattribution of some fraction of nonspecific indicator species (e.g., black carbon [BC]) to the bus, when it actually came from another source, or to undetected instrument calibration bias.12 The substantially larger contribution of PM_ck, despite its lower emission rates, confirms the identification by Hill et al.8 of crankcase emissions as potentially significant. The testing of the two Seattle buses does not fully resolve questions of representativeness of buses tested, but the low-wind, low-speed, high engine load test conditions encountered in Seattle are arguably more likely to result in high levels of DPM_tp in bus cabins. Further study of additional buses in different settings is needed to determine whether the tracer-based findings to date are representative of the full range of in-cabin DPM_tp and PM_ck concentrations.

Larson et al.27,28 used this study's data to explore the use of receptor modeling techniques for tailpipe and crankcase emissions, using OC, EC, and GC/MS-derived marker species including hopanes and steranes as potential tracers of opportunity for lubricating oil. In the evaluation of inferential methods against the tracer results presented here, Liu et al.12 found that under the conditions of these tests, differential concentrations between in-bus and LV measurements of nonspecific markers such as EC, OC, and PM₁ (PM ≤ 1 μm in aerodynamic diameter) may provide some useful information regarding self-pollution, although with a positive bias of up to 20%. In the absence of definitive evidence to the contrary, such approaches cannot be considered reliable or accurate without concurrent verification by tracers. The dual tracer method allowed explicit identification of a single engine's emissions at very low detection limits. The authors are unaware of any other method capable of accurately quantifying tailpipe and crankcase self-pollution in school buses or for addressing similar source attribution problems.

SUMMARY

Two conventional front-engine school buses were tested in a relatively clean urban setting using separate intentional tracers for diesel exhaust and crankcase vent PM. The results for in-bus concentrations of its tailpipe and crankcase vent emissions showed self-pollution to be higher with windows closed, and despite their lower emission rate, crankcase PM concentrations were substantially higher than those of tailpipe exhaust PM. The results also confirmed the specificity and sensitivity achievable by this method. The complete absence of ambient interferences in the Ir and d-alkane analyses and the sensitivities of analytical methods for these species made it possible to unambiguously quantify DPM_tp and PM_ck concentrations as low as 0.002 and 0.05 μg/m³, respectively. For comparison, detection limits for EC or BC, the most commonly used indicator of DPM_tp, are typically 0.1 μg/m³ or greater. Before this study, no methods were available to quantitatively distinguish PM_ck concentrations from those of similar carbonaceous PM sources.

The use of high flow rate in-cabin samplers allowed an approximately 6-fold increase in sensitivity relative to the standard 16.7-L/min samplers. The use of the RAVEM system allowed emissions testing under actual bus operating conditions. The samplers and the RAVEM system generally performed well, but they exhibited occasional evidence of leaks. Although some in-bus and source samples were potentially compromised by these leaks, a sufficient number of samples were collected, and there was sufficient redundancy of samples (including concurrent Teflon and quartz sampling) to generate acceptable surrogates for suspect values and to establish the validity of the overall findings of the field program.

It should be noted that future applications of the dual tracer method may not require the detection limits achieved here. Options include using lower flow rate in-cabin samplers, reducing sample durations, and reducing the tracer spike levels. In particular, the d-alkane tracer is extremely expensive. If lower sensitivity is acceptable, material cost savings from reducing the spike level in lubricating oil could be substantial, especially if multiple engines/vehicles are being tested. Also, if appropriate, quartz filter extractions could be composited to reduce GC/MS analytical costs.

The use of the d-alkane lubricating oil tracer provided clear confirmation and quantification of the apparent significance of PM_ck intrusion in bus cabins, as initially reported by Hill et al.8 Although PM_ck is not diesel exhaust, in-vehicle concentrations of particles from all sources are of concern. PM_ck concentrations in the buses were disproportionately higher than DPM_tp based on a comparison of the emission rates of the two pollutants. This suggests that the location of the crankcase vent may be an important factor in the amount of intrusion that occurs.

ACKNOWLEDGMENTS

This study was partially sponsored by the National Institute of Environmental Health Sciences (1R01ES12657-01A1). Inclusion of the dual tracer experiments was made possible by funding provided by Navistar, Inc., and the U.S. Department of Energy Office of Vehicle Technologies through the National Renewable Energy Laboratory. The assistance of David Anderson of the Seattle School District Transportation Department, First Student, Inc., and the Puget Sound Clean Air Agency is gratefully acknowledged.