Because DPM concentrations cannot be measured directly in the ambient air, it is necessary to estimate concentrations based on surrogate measures and source emissions data linking the DPM to its surrogates. MATES-II monitored elemental carbon (EC) as a surrogate; other studies have applied source apportionment methodologies. This appendix discusses two approaches to estimating DPM concentrations, source emissions data issues, and the DPM emissions estimation approach used in MATES-II.
As mentioned above, there are two primary methodologies for estimating DPM concentrations. Both require that the DPM be analyzed for its constituents and that emissions of other sources in the area being studied be known.
In the case of EC, samples of particulate matter from diesel exhaust are analyzed to determine the extent to which they contain EC. Regional emission inventories are then made for EC, and the extent to which diesel sources are contributing to the EC emission inventory is assessed. DPM concentrations in the ambient air are then determined by multiplying measured EC concentrations by a conversion fraction:
% of EC inventory from diesels/% of EC in diesel exhaust
One complication with this methodology is that EC itself is not a well-defined quantity. Some researchers have assumed that EC corresponds to the fraction of particulate that cannot be extracted using a strong solvent; others use thermal methods to remove organics and optical methods to analyze EC. All give different results; as Chow et al. (1993, pp. 1185-1186) note:
There are several variations on [the thermal/optical analytical method for EC/OC] with respect to: (1) the temperatures to which the samples are subjected, (2) the length of analysis time at each temperature, (3) the rate of temperature increase, (4) the composition of the atmosphere surrounding the sample, (5) optical monitoring of pyrolysis/volatilization/combustion and (6) calibration standards. . . .these differences in the application of thermal/optical methods have little effect on the total carbon measured on a sample, but can have a significant effect on the point of delineation between organic and elemental carbon. The same is true for other carbon analysis methods.
For DPM, the most common source apportionment method has been to apply chemical mass balance (CMB) techniques to trace constituents of the DPM. CMB analyses include a number of assumptions such as: compositions of source emissions remain constant over time, there are no interactions among the chemical species used to represent specific sources, and all sources are known for the substances under evaluation (Watson et al., 2000). Good source profiles are needed, and ideally ambient and source sampling techniques would be based upon the same methodologies. One of the advantages of the CMB approach over the EC approach is that CMB uses a number of directly measured pollutant concentrations, rather than just EC, as a method of estimating DPM concentrations, and those pollutants are chemically defined, rather than being defined based upon details of the analytical methodology. A key disadvantage, however, is that the assumptions needed to successfully apply the methodology are not always met.
Both the EC- and CMB-based methodologies for estimating DPM have been applied to ambient air quality data measured at four sites in southern California in 1982 (Gray, 1986; Schauer et al., 1996); results are shown in Table A-1.
|Site||EC Surrogatea||Chemical Mass Balanceb||Percent Difference|
|a Determined by multiplying EC data from Gray (1986), Table 2.2, by 0.67/0.64 = 1.04.
b Schauer et al. (1996)
|Pasadena||4.11||5.27 ± 0.72||28%|
|Downtown Los Angeles||5.06||11.6 ± 1.19||129%|
|West Los Angeles||3.75||4.36 ± 0.64||16%|
|Rubidoux||3.15||5.35 ± 0.51||70%|
In this case, the CMB methodology was based upon the relative amounts of various polyaromatic hydrocarbons (PAHs) in the vehicle exhaust (Schauer et al., 1996); results are comparable but slightly higher than those from the EC-surrogate approach (except in the case of downtown Los Angeles, where results are significantly higher). It could be argued either that the EC approach assumed too low of an EC-to-DPM conversion fraction, or that the CMB approach was inferior because it was based upon chemicals that account for less than 1% of the total DPM mass (U.S. Environmental Protection Agency, 2002c) and the DPM PAH profile was based upon a relatively small number of vehicle tests. The following section addresses complications associated with characterization of DPM, which both of these methods are highly dependent upon.
Clark et al. (2002) have noted that various factors affect the amount of emissions produced by a diesel, estimating driving cycles as having a 1,500% effect and vehicle age as having a 1,200% effect. Often, for purposes of emission inventories or the methods described above, diesel emissions are characterized and estimated based on a limited number of tests, possibly at only one drive cycle. Clark et al. (2002) did not study differences in DPM composition as a function of drive cycle; however, Amann and Siegla (1982) tested a diesel vehicle idling and at a speed of 48 km/h, and showed that the overall hydrogen-to-carbon (H/C) ratio for the compounds in the extractable organic carbon (OC) portion of DPM was 1.26 at idle but 1.63 at 48 km/h. Heywood (1988) noted that radioactive tracer studies in a light-duty diesel vehicle showed lubricating oil (instead of the diesel fuel) to be responsible for 2-25% of DPM mass and 16-80% of the extractable organic portion, with the greatest percentages being measured at the highest engine speed studied (3000 rpm). Given that these differences are large for even a relatively simple change in drive cycle, it is difficult to estimate the extent to which DPM composition varies for other, more complex drive cycles.
Lighty et al. (2000) have shown that typically, only a small fraction of the particle-phase organics in DPM are identifiable; therefore, simply correlating DPM to EC is favorable. However, the extent to which DPM is composed of EC is also not fixed. In a 1982 emission inventory, Gray (1986) assumed that 64% of DPM mass was EC, based upon the assumption that 83.7% of the fine particulate matter ("fine" was defined as being smaller than 2.1 mm) exhausted by diesels was carbon and that 76.6% of the carbon was EC. However, Christoforou et al. (2000) cite evidence that the percentage of EC in DPM decreased from 55% in the late 1970s to 40.5% in the late 1980s; and Hayes et al. (1998) assume that DPM was 28-34% EC as of the late 1990s. More recently, Lev-On (2002) has shown that when particulate filters are in place, exhausted DPM is essentially 0% EC. (This study did not investigate the extent to which these differences are affected by differences in EC determination methodologies.)
As mentioned previously, the MATES-II study estimated DPM concentrations by scaling EC concentrations using a conversion factor. However, the scaling factor used was derived from the 1982 emission inventory by Gray (1986). That inventory, developed for a 50 mile by 50 mile grid in the South Coast, showed that 67% of the EC emitted in the area was attributable to mobile diesel sources. As mentioned in the previous section, it was assumed that 64% of DPM mass was EC, so the EC-to-DPM scaling factor used was 0.67/0.64 = 1.04.
The relationship between DPM and EC has changed over time. Table A-2 shows that the 1998 emission inventory developed for the 210-km by 210-km area of the MATES-II study appears to have assumed that the percentage of DPM equal to EC was approximately 31%, considerably lower than the 64% value assumed by Gray (1986). However, diesel sources were also shown to account for just 28% of the total EC emissions (51,736 lb/day), much less than the 67% that was assumed for 1982. Therefore, the EC-to-DPM scaling factor based upon the MATES-II emission inventory (0.28/0.31 = 0.90) is 13% lower than the scaling factor of 1.04 that was actually used in MATES-II to estimate DPM concentrations (i.e., the scaling factor based on the 1982 inventory was 16% higher).
|Source Category and Code
(Codes from SCAQMD, 2000)
|DPM (lb/day)||Elemental Carbon (EC) (lb/day)||% of Diesel that is EC|
|a The total DPM from these source categories accounts for 95% of the total DPM emission inventory in MATES-II (47,113 lb/day). Several of the more minor source categories of DPM emitters included a mixture of diesel and non-diesel sources, rendering it impossible to determine the percentage of diesel that SCAQMD assumed was EC. Based on the data published in the MATES-II reports, we therefore estimate total EC from diesels as being approximately (13,951)/(0.95) = 14,700 lb/day.
b These values are maximums (i.e., assuming that all of the EC emissions from trains, ships, and other mobile equipment are from diesel engines, with no contribution from spark-ignition engines or other engine types).
|700 On-Road Vehicles
740 Heavy-Duty Diesel Trucks
|800 Other Mobile
870 Mobile Equipment
Studies such as MATES-II may associate a specific estimated excess cancer risk to DPM. In the case of MATES-II, the SCAQMD estimated excess cancer risk associated with DPM exposure to be approximately 1000 excess cancers per million exposed people. However, uncertainty associated with at least two of the factors important to DPM risk assessments, emissions and URFs, means that DPM-related excess cancer risk estimates from MATES-II and other studies need to be considered more as a potential range of risk, rather than as a point estimate. DPM emission inventory estimates may vary due to:
EC vs. CMB-based approaches that may yield DPM estimates differing by as much as a factor of two (Table A-1).
Outdated EC source apportionment characterizations that may inadequately represent the fraction of EC attributable to diesel exhaust. Limited evidence indicates that assumptions regarding the fraction of EC in diesel exhaust differ on the order of a factor of two; in the case of MATES-II, this results in potential error of approximately 10% to 15%.
A full exposition on health risk assessment is beyond the scope of this white paper, and there are a number of publications that address in greater detail the limitations associated with risk assessment (e.g., ENVIRON, 1986). This discussion touches only briefly on example issues. Lifetime risks for cancer are typically calculated by multiplying pollutant concentrations, measured in either parts per million (ppm) or micrograms per cubic meter (mg/m3), by a pollutant-specific Unit Risk Factor (URF), which is based upon the assumption of a lifetime (70 years) of exposure to that concentration. For example, if the measured concentrations of pollutants A, B, and C are CA, CB, and CC, and an individ ual is exposed to those concentrations continuously for 70 years, the overall cancer risk would be calculated as
Total risk = CAURFA + CBURFB + CCURFC (Equation B-1)
Because this additive approach does not take any synergistic effects between different chemicals into account, it may under or overestimate risk. This approach can also be complicated when categories of pollutants are considered as a single pollutant. For example, if DEOG and benzene (a component of DEOG) were considered to be separate pollutants, and a URF were developed for each, application of Equation B-1 would double-count the cancer risk associated with the benzene.
For purposes of comparing the toxicity of different air pollution scenarios, URFs are usually calculated based on the assumption that a person will be exposed to the concentrations of interest for their entire lifetime. However, actual exposures depend upon the amount of time spent by people in various settings, and the pollutant concentrations in those settings over a 70-year period. In the case of pollutant classifications, the relative amounts of the different chemicals in the class may also be variable. The composition of DPM, for example, is affected by driving patterns, model year, terrain driven, and other variables (see Appendix A for more information regarding DPM composition); insufficient information is available to state with any degree of certainty how these factors affect the URF for DPM.
Even for pure, well-defined substances, uncertainties associated with URFs can often be an order of magnitude or more, due to difficulties associated with exposure assessments, varying results from different studies, and the means by which they are calculated (e.g., extrapolating data from high doses and high cancer incidence down to very low exposures, and making assumptions regarding the shape of the dose-response curve near zero dose).
This appendix provides background information to explain the emissions modeling results included in Table 5-1. Because California has historically had different vehicle emissions standards than the other 49 states, two different models for determining emission factors exist. California's EMFAC2002 model incorporates data such as fleet age distributions, distribution of vehicle types, ambient temperatures and humidities, and other factors for various subregions within California. EPA's MOBILE6 model, on the other hand, requires the user to either input the area-specific data, or utilize default national data developed by the EPA. A calendar year of 2000 was chosen for comparing the results from the two models; emissions of CO and VOC were chosen for comparison because CO is the pollutant which is currently used to assess microscale air quality impacts from transportation projects, and the VOC category includes many of the MSAT pollutants.
EMFAC2002 can be operated in one of three modes:
"Burden mode," which estimates emissions in tons per weekday;
"Emfac mode," which estimates g/mi emission factors as a function of specific temperatures, humidities, and speeds; and
"Calimfac mode," which estimates detailed emission rates for each vehicle class and model year as a function of vehicle mileage.
Given the comparison being conducted for this study--i.e., regional fleetwide differences between the area covered by the MATES-II study and Las Vegas--EMFAC2002 was run in "Burden mode" for the South Coast region of California. Because EPA's MOBILE6 model only allows the user to evaluate the months of January and July, EMFAC2002 was run for these two months. With respect to emissions of hydrocarbons and organics, the classification "Reactive Organic Gases" (ROG) was chosen, which is the same as EPA's (VOC) definition: i.e., the classification includes gases such as aldehydes (which are not detected by the commonly-used flame ionization detectors used for measuring hydrocarbon emissions) and does not include hydrocarbons that are exempt from the definition of VOC, such as methane. (It should be noted that VOC emissions from EMFAC2002 include both exhaust and evaporative losses, but not refueling emissions.) No other inputs needed to be provided to the model.
EMFAC2002 outputs information regarding the vehicle miles traveled (VMT) per day; therefore, average emission factors (g/mile) were calculated by dividing the daily emissions by the VMT (and converting mass units). The EMFAC modeling runs and calculations of emission factors are shown in attachments at the end of this appendix.
MOBILE6 has several input parameters, which are summarized in Table C-1. More detailed information can be found in the MOBILE6 input files included at the end of this appendix. As with EMFAC2002, MOBILE6 was run for a weekday during the months of January 2000 and July 2000. Readily available meteorological data for the months of January and July were obtained from the National Oceanic and Atmospheric Administration (NOAA). Several input parameters related to the local vehicle fleet and travel patterns were obtained from the Clark County Department of Air Quality Management (DAQM). Hydrocarbons and organics were expressed as "VOC" for consistency with the EMFAC2002 modeling, and refueling emissions were excluded. MOBILE6 output files are included at the end of this appendix.
As indicated in Section 5.1, the scope of work for this project did not allow us to quality assure the accuracy of the data obtained from the Clark County DAQM.