Grantee Research Project Results

On-road motor vehicle emissions were measured during the summer of 2010 in a highway tunnel in the San Francisco Bay area, by a team that included researchers from UC Berkeley and Aerodyne Research. Measurements included both gaseous and particle-phase air pollutants emitted by vehicles. Air pollutant emissions were measured in different tunnel tubes, at different times of day, and on weekends as well as weekdays, to provide separate emission factor results for light-duty passenger vehicles and heavy-duty diesel trucks. Emission factors were calculated using carbon balance methods, by normalizing measured concentrations of various pollutants of interest to total carbon (mainly carbon dioxide) concentrations present in vehicle exhaust. Emission factor results are calculated and reported in units of mass of pollutant emitted per unit mass of fuel burned.

 

Fast-response instrumentation was used to measure concentrations of pollutants present in the exhaust plumes of individual heavy-duty diesel trucks as they drove by our sampling location, located near the east end of the Caldecott tunnel on California state highway 24, with vehicles traveling uphill on a 4% grade.

 

Fleet-average emission factors as well as emission factor distributions have been derived and reported for heavy-duty diesel truck emissions of nitric oxide (NO), nitrogen dioxide (NO₂), carbon monoxide (CO), formaldehyde, ethene, black carbon (BC), organic aerosol (OA), as well as optical properties of the emitted particles. The total number of trucks providing usable data varied depending on the pollutant, due to instruments being online to varying degrees, and in some cases due to problems distinguishing diesel emissions from high background levels of gasoline engine emissions present inside the tunnel. The sample size exceeded 500 heavy-duty trucks for all of the above pollutants except organic aerosol, ethene, and light absorption at λ = 630 nm. These are large sample sizes compared to what can be measured in laboratory studies using chassis or engine dynamometers, and more extensive in terms of the list of pollutants measured compared to on-road remote sensing studies.

 

Diesel truck emission factor distributions were skewed, with the highest emitting 20% of trucks observed on the road responsible for fractions ranging from 40 to over 70% of total diesel-related emissions. The emission factor distribution for NO_x was less skewed than corresponding distributions for OA and BC. This finding is consistent with expectations that emission factor distributions become increasingly skewed as emission control efforts advance, given that there has been more progress to date in reducing primary PM compared to NOx from diesel engines. The high-emitting trucks are not necessarily the same for different pollutants. For example, we observed that trucks with high primary NO₂ emissions tended to have low emissions of BC, OA, and organic gases, consistent with required deployment of diesel particle filters on trucks with 2007 or newer engines.

 

Further measurements of vehicle emissions in separate lanes of the tunnel where heavy-duty diesel trucks are prohibited were used to derive fleet-average emission factors for light-duty passenger vehicles, which are almost entirely powered by gasoline engines. We found that accounting for pollution contributions from seemingly small numbers of mostly 2-axle/6-tire diesel-powered trucks led to substantial downward adjustments of ~20% in OA and NO_x and ~45% for BC emission factors from gasoline engines on a fleet-average basis.

 

Based on our analysis of fleet-average emission factors for gasoline and diesel engines as of 2010, we developed a graphic that depicts the gasoline/diesel split in contributions to total on-road emissions of various air pollutants. The vertical axis of the plot is the ratio of emission factors (diesel/gasoline) expressed per unit of fuel burned in both numerator and denominator. As shown below in Figure 1, these ratios vary widely depending on the pollutant, from < 1 for CO, near unity for CO₂, in the range of 10-20 for OA and NO_x, up to 40-60 for BC. In other words, per unit of fuel burned, the average on-road heavy-duty diesel engine emitted 40-60 times more BC than the corresponding average gasoline engine, as observed at the Caldecott tunnel in summer 2010. The horizontal axis of Figure 1 shows the fraction of total on-road fuel use in the region of interest (i.e., in a specific city, state, or nationally) that is due to diesel fuel combustion, with gasoline accounting for the remainder. The labeled curves show the resulting diesel contribution to total on-road vehicle emissions as a function of emission factor ratio and diesel fuel use fraction. Clearly, there is no single answer to gasoline/diesel split questions: the answer varies depending on pollutant, location, and spatial scale of interest. It is nevertheless our finding that diesel engines accounted as of 2010 for half or more of total on-road running emissions of BC, OA, and NO_x. In the case of NO_x especially, this result can be surprising even to air quality specialists. The situation for NO_x is a result of success in controlling gasoline emissions and the lack of comparable progress in reducing diesel emissions.

 

 

 

Figure 1. Contributions to total on-road running emissions due to diesel emissions as of 2010 (labeled curves). The upper right corner of the diagram is heavily diesel-dominated, whereas at the lower left, emissions are dominated by contributions from gasoline engines. This graphic does not account for excess emissions associated with cold engine starting, which are higher during winter months. Source: Dallmann et al. (ES&T 2013).

 

 

NO_x emission factors derived from our 2010 field study at the Caldecott tunnel, together with data from other on-road emission studies both at Caldecott and elsewhere were critically reviewed and used to assess long-term trends in on-road vehicle NO_x emissions at air basin, state and national scales. Figure 2 presents U.S. national emission trends, showing on-road gasoline (green), diesel (blue), and on-road total (black) emissions of NO_x with associated uncertainties in the fuel-based emission inventory estimates depicted as shaded bands. Also shown in Figure 2 are corresponding estimates from the motor vehicle emission simulator (MOVES) model developed by EPA. The fuel-based and MOVES estimates agree as of 2010 in terms of total on-road NO_x emissions. The estimates differ in their assessments of the relative importance of gasoline versus diesel contributions to the on-road total, as well as in trends over time between 2000 and 2010. Installation of advanced emission control systems to control NO_x from diesel engines may therefore have a greater effect on emissions and air quality, and further light-duty vehicle control efforts may be less useful, relative to what is indicated in MOVES-based assessments of the national emission inventory and gasoline versus diesel contributions to NO_x.

 

Figure 2. National trends in NO_x emissions from on-road vehicles, 1990-2010. Vertical axis is plotted in units of metric tons emitted per day. NO_x mass is reported in NO₂ equivalents. Shaded bands represent effects of emission factor and fuel sales uncertainties on fuel-based emission inventory estimates. Dot-dashed lines are EPA MOVES model estimates and are shown for comparison. Source: McDonald et al. (JGR 2012).

 

 

In addition to measuring vehicle emissions at the Caldecott tunnel, liquid samples of unburned gasoline and diesel fuel were collected and analyzed as part of this research. A novel gas chromatography-mass spectrometry (GC-MS) analytical technique was used for both unburned diesel fuel and tunnel organic aerosol analyses in this study. We used a bright and tunable synchroton light source at Lawrence Berkeley National Laboratory to generate vacuum ultraviolet (VUV) photons with energies comparable to the ionization potential of hydrocarbons present in diesel fuel. Our soft ionization technique greatly reduces fragmentation of parent ions compared to the traditional higher-energy electron ionization (EI) approach. Knowing the molecular weights of the parent ions made it possible for us to achieve a detailed accounting of the hydrocarbons present in diesel fuel, as shown below in Figure 3. In prior analyses of diesel fuel, n-alkanes and polycyclic aromatics are commonly quantified and reported, whereas branched and cyclic hydrocarbons tend to co-elute in an unresolved mixture and therefore are left unidentified.

 

 

 

Figure 3. Measured chemical composition of unburned diesel fuel samples collected during summer 2010 in California. Note for the alkanes, the total is subdivided into normal (straight-chain, dark red) and branched (lighter red) isomers. Likewise for the monocyclic alkanes, the subdivision shows whether alkyl substituents attached to the ring structures are straight (dark blue) or branched (lighter blue). Source: Gentner et al. (PNAS 2012).

 

 

Volatile organic compound (VOC) emissions were measured on-road from both gasoline and diesel vehicles using an online GC-MS system that was operated at the Caldecott tunnel (Gentner et al., ES&T 2013). We combined the tunnel data with fuel composition data to calculate overall VOC emission rates including compounds known to be present in fuel but not measured directly in the tunnel. As shown in Figure 1, we found that as of 2010, exhaust VOC mass emission factors from diesel engines were comparable to or higher than corresponding gasoline engine emission factors, when expressed per unit of fuel burned. Gasoline engines remain the dominant source of VOC emissions, since the use of gasoline greatly exceeds the volume of diesel fuel sold for on-road use. However, the diesel contribution to VOC mass is non-negligible, and has become more significant over time, as a result of large reductions in gasoline-related VOC emissions that have occurred over the 20 years. We calculated secondary organic aerosol yields per unit mass of unburned gasoline and diesel fuel, and found that the SOA yield for diesel fuel was 6.5 times that for gasoline. When combined with our findings about relevant VOC precursor emissions, we concluded that diesel engines make a contribution that is comparable in importance to the SOA formed from VOC emissions from on-road gasoline engines (Gentner et al., PNAS 2012).

 

GC-MS with soft ionization was also used to measure the chemical composition of vehicular primary organic aerosol (POA) emissions. In parallel, tunnel aerosol was analyzed using a new aerosol mass spectrometer (SP-AMS) that responds to refractory aerosol components such as black carbon and trace elements. Based on separate analytical results from both methods, and multiple consistent lines of evidence, we found gasoline and diesel emissions of POA to be of similar chemical composition, consistent with a common lubricating oil-related source for OA emissions from both engine types.

 

AMS-derived OA mass spectra include fragments that are characteristic of unsaturated hydrocarbons (i.e., two, four, or six missing hydrogen atoms relative to the expected hydrogen abundance for saturated alkanes). Furthermore, SP-AMS data show trace elements such as zinc and phosphorus/phosphate, measured at high time resolution in the exhaust plumes of individual diesel trucks. The abundance of trace elements, when stated as a fraction of total OA mass emitted by individual trucks, is consistent with levels of zinc dialkyl dithiophosphate (ZDDP) additive known to be present in engine lubricating oil.

 

As shown below in Figure 4, vehicular POA emitted by both light-duty gasoline and heavy-duty diesel vehicles consists mainly of cycloalkanes with one or more branched alkyl side chains (≥80%), with low abundances of alkanes and aromatics (<5%), similar to “fresh” lubricating oil. The ring structures are most likely dominated by cyclohexane and cyclopentane rings rather than larger cycloalkanes. High molecular weight combustion byproducts, including alkenes, oxygenates, and aromatics, were not present in significant amounts. The observed carbon number distribution and chemical composition of motor vehicle POA was consistent with lubricating oil being the dominant source from both gasoline and diesel-powered vehicles, with an additional smaller contribution from diesel fuel, and a negligible contribution from unburned gasoline. Note that cyclic alkanes are desired and deliberately selected for use in engine lubricating oils, whereas waxy paraffins and char-forming aromatics are removed during oil processing at refineries.

 

Figure 4. Chemical composition of tunnel POA emissions (upper panels) and commercial samples of lubricating oil (lower panels), as a function of carbon number (NC) and molecular structure (see legend for details). NDBE in the number of double bond equivalents; values correspond to the number of alkane ring structures that are present in a molecule. Aromatic hydrocarbons have NDBE = 4, but this category is dominated by tetracyclic alkanes rather than aromatics. See Worton et al. (ES&T 2014) for further details.

 

 

In past studies focusing on the analysis of the molecular composition of organic aerosol samples, it has been typical for only a minor fraction of total organic mass to be identified. Much of the remainder is typically referred to as an unresolved complex mixture or UCM, of unknown chemical structure. In the present research sponsored by EPA, using VUV-GC/MS analytical techniques, our analyses determined carbon numbers and classified molecular structures at an unprecedented level of > 60% of the total mass of vehicular POA emissions.

 

We caution against the use of fuel-derived tracers in receptor-based source apportionment studies for organic aerosol, which has been common practice in past receptor modeling/chemical mass balance studies. We find that the bulk of vehicular POA emissions consists of lubricating oil, not unburned fuel. If the abundance of fuel-derived hydrocarbons associated with POA emissions varies depending on the rate of lubricating oil emission, this could make fuel-derived hydrocarbons unreliable as tracers for vehicular POA emissions.

 

Application of similar analytical techniques to sampling of ambient OA at Bakersfield and Pasadena (supported by other sponsors), allowed for tracking of the unresolved complex mixture and its dynamics in transit from emission sources to receptor monitoring sites. Published work by Chan et al. (JGR 2013) highlights the importance of branched versus linear alkanes in POA emissions, and finds the branched alkanes to be both more abundant and more reactive in the atmosphere. Therefore, branched alkanes play a more important role in SOA formation.

 

Use of redundant measurements in this study allowed for numerous cross-comparisons and consistency checks of the measurements to ensure quality and robustness of the analytical results and findings that emerged. Although our sampling strategy emphasized use of online instrumentation, we also collected quartz and Teflon filter samples, which proved to be useful not only for quality assurance of online instrument calibrations, but also for later use in offline VUV-GC/MS analyses of tunnel POA. Similarly, we made numerous independent measurements of black carbon concentrations, which again allowed for measurement cross-comparisons and helped with quality assurance of measurements derived from online instruments.