Environ. Sci. Technol. 2004, 38, 2005-2010
Methane Emissions from Vehicles E. K. NAM, T. E. JENSEN, AND T. J. WALLINGTON* Research Laboratory MD-3083, Ford Motor Company, P.O. Box 2053, Dearborn, Michigan 48121
Methane (CH4) is an important greenhouse gas emitted by vehicles. We report results of a laboratory study of methane emissions using a standard driving cycle for 30 different cars and trucks (1995-1999 model years) from four different manufacturers. We recommend the use of an average emission factor for the U.S. on-road vehicle fleet of (g of CH4/g of CO2) ) (15 ( 4) × 10-5 and estimate that the global vehicle fleet emits 0.45 ( 0.12 Tg of CH4 yr-1 (0.34 ( 0.09 Tg of C yr-1), which represents 1997) vehicles. We also approximated the fraction of cold start emissions (before the engine has warmed) and attempted to estimate the effects of aging on the fleet.
Experimental Section Vehicle Emission Studies at Ford Motor Company. The emissions measurement sequence consisted of a series of * Corresponding author phone: (313)390-5574; fax: (313)3227044; e-mail:
[email protected]. 10.1021/es034837g CCC: $27.50 Published on Web 02/26/2004
2004 American Chemical Society
two to eight U.S. EPA Urban Dynamometer Driving Schedule (UDDS or Federal Test Procedure, FTP-75) tests carried out on nearly consecutive days for each vehicle. Vehicle tests were conducted at the Ford Research Laboratory (Vehicle Emissions Research Laboratory, VERL 1) in Dearborn, MI, on a 48-in., single-roll (Burke-Porter, Grand Rapids, MI) chassis dynamometer. The tailpipe of the vehicle was connected to a dilution tube through an insulated, heated transfer line. The dilution tunnel flow rate was 10 m3 min-1 for gasoline-fueled vehicles, 20 m3 min-1 for alternate-fueled vehicles, and 30 m3 min-1 for diesel-fueled vehicles. Methane and the regulated emissions (non-methane hydrocarbon (NMHC), CO, and NOx) were measured using standard vehicle dynamometer emissions bench procedures (7). Methane was also measured by direct spectroscopic (Fourier transform infrared spectroscopy, FTIR) and chromatographic (gas chromatography with flame ionization detection, GC/FID) methods. A range of vehicles (30 in total) were tested. The cars tested were (in alphabetical order) Ford Contour, Ford Crown Victoria, Ford Escort, Ford Mondeo (2 vehicles), Ford Taurus, Honda Accord, Lincoln Continental (2 vehicles), Lincoln Grand Marquis, Lincoln Mark VIII, Lincoln Town Car (2 vehicles), Mercury Sable (3 vehicles), Volkswagen Passat, and Volvo V70. The trucks tested were Ford Econoline, Ford Explorer (3 vehicles), Ford F-150, Ford Ranger, Ford Windstar (5 vehicles), and Dodge Ram. In addition, 3-s time-series data were collected on a subset of 4 vehicles: Escort, Windstar, Accord, and the Volvo V70. All vehicles were equipped with modern three-way catalyst systems. The vehicle model years ranged from 1995 to 1999. The emission levels of the regulated (NOx, CO, and hydrocarbon) pollutants for these vehicles were generally between those required for compliance with “U.S. Tier 1” and “California ULEV” (ultralow emissions vehicles) regulations. For the class of vehicles considered here (passenger vehicles and light-duty trucks), Tier 1 corresponds to 400 mg/mi NOx and 250 mg/mi NMHC emission rates. For ULEV vehicles, the NOx and NMHC emission rates would be below 200 and 40 mg/mi, respectively. Six fuels were tested, although individual vehicles used only one fuel. The fuels used were U.S. Certification fuel (cert), California Reformulated fuel (CRF), compressed natural gas (CNG), M-85 (85% methanol, 15% gasoline), E-85 (85% ethanol, 15% gasoline) and U.S no. 2 diesel fuel (diesel). The use of different fuels provides information concerning the sensitivity of methane emissions to fuel type. The diluted exhaust sample was filtered through a 142mm quartz fiber filter (Palliflex, Tissuquartz 2500QAO) with backing in a stainless steel holder at room temperature. A new filter was used for each four-UDDS series. The 1.27-cm (0.5-in.) o.d. sample transfer lines were either stainless steel or Teflon. CH4 was measured in real time by FTIR spectrometer (Mattson Instruments, Nova-Cygni 120) operated at a resolution of 0.25 cm-1, and the interferograms were zerofilled to an effective resolution of 0.125 cm-1. The data system was used to co-add, transform, and analyze spectra for 20 compounds in 3-s intervals. A rotary vane pump (Gast 0822) pulled the diluted exhaust sample through the FTIR gas cell at 30 L/min with 93.3 kPa absolute pressure in the gas cell. The sample component concentrations were determined from a linear relationship to the spectral line strength of standard reference spectra. A more complete description of this analytical system is available elsewhere (8). The limit of detection for methane, CO, and CO2 in the UDDS three-bag composite are 2, 4, and 3400 mg/km, respectively. VOL. 38, NO. 7, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2005
TABLE 1. Vehicle Measurements Performed in This Work vehicle identifier
fuel
methane (mg/km)
NMHC (mg/km)
CO (g/km)
CO2 (g/km)
car A car B car C car D car E car F car G car H car I car J car K car L car M car N car O car P car Q car R
cert cert CRF CRF CRF CRF CRF CRF CRF CNG M-85 diesel diesel E-85 cert diesel cert CRF
16 ( 1 63 ( 16 6 ( 1 67 ( 11 12 ( 1 85 ( 2 6 ( 4 66 ( 2 11 ( 1 97 ( 6 9 ( 1 41 ( 2 14 ( 1 94 ( 6 18 ( 9 31 ( 3 2 ( 1 14 ( 3 43 ( 4 8(1 3 ( 1 18 ( 2 8 ( 1 107 ( 19 7 ( 3 116 ( 86 11 ( 6 54 ( 2 12 ( 2 34 ( 3 23 ( 2 18 ( 2 3 ( 1 34 ( 2 4 ( 1 43 ( 11
0.84 ( 0.07 0.23 ( 0.02 1.16 ( 0.02 0.68 ( 0.02 0.59 ( 0.04 0.53 ( 0.02 0.86 ( 0.02 0.23 ( 0.02 0.15 ( 0.05 0.23 ( 0.02 0.59 ( 0.05 0.30 ( 0.02 0.53 ( 0.23 0.88 ( 0.06 0.31 ( 0.07 0.07 ( 0.02 0.30 ( 0.05 0.43 ( 0.28
293 ( 2 288 ( 5 239 ( 1 258 ( 2 285 ( 1 295 ( 3 286 ( 2 305 ( 4 217 ( 2 247 ( 2 244 ( 9 193 ( 11 161 ( 2 247 ( 2 201 ( 8 146 ( 2 199 ( 1 250 ( 2
truck A truck B truck C truck D truck E truck F truck G truck H truck I truck J truck K truck L
cert cert cert CRF CRF CRF CRF CRF CNG E-85 Cert CRF
6(1 20 ( 1 21 ( 1 22 ( 1 4(1 7(1 11 ( 1 11 ( 1 97 ( 5 33 ( 2 11 ( 1 13 ( 2
54 ( 11 61 ( 2 51 ( 3 83 ( 3 43 ( 2 83 ( 6 78 ( 8 61 ( 3 18 ( 2 73 ( 38 93 ( 11 55 ( 12
0.44 ( 0.06 1.16 ( 0.07 0.44 ( 0.03 2.16 ( 0.06 0.48 ( 0.04 0.69 ( 0.05 0.97 ( 0.06 1.36 ( 0.09 0.32 ( 0.09 0.50 ( 0.05 0.54 ( 0.14 0.57 ( 0.05
394 ( 3 289 ( 4 293 ( 4 259 ( 2 268 ( 2 339 ( 2 345 ( 2 381 ( 40 329 ( 4 266 ( 3 351 ( 5 260 ( 17
car truck car truck car
avg avg avg non-CNG avg non-CNG avg diesel
12 ( 2 21 ( 7 10 ( 1 14 ( 3 13 ( 5
55 ( 8 63 ( 12 58 ( 8 67 ( 6 80 ( 31
0.50 ( 0.07 0.80 ( 0.15 0.51 ( 0.07 0.85 ( 0.16 0.30 ( 0.13
242 ( 11 314 ( 14 242 ( 12 313 ( 15 166 ( 14
The diluted exhaust samples, collected in Tedlar bags for each phase of the UDDS test, were speciated and quantified by gas chromatography (GC) and flame ionization detection (FID), respectively. The AQIRP vehicle emissions hydrocarbon analysis method employed, described elsewhere in detail (9), uses one GC/FID to analyze C1-C4 hydrocarbons and a second one to analyze C4-C12 hydrocarbons. The method has a 5 ppb carbon limit of quantitation and was operated with a library of 154 hydrocarbon species, including methane. Dynamometer emissions bench, FTIR, and GC/FID methane results were compared for all vehicle tests and were similar for each UDDS sample (within 5%) or ambient bag (within 10%).
Results CH4 Emissions Measured over the UDDS Cycle. The results of the dynamometer studies are given in Table 1. This vehicle dynamometer emissions bench data list the CH4, NMHC, CO, and CO2 for a variety of vehicles using different fuels, while driven using the UDDS. The values at the bottom of the table show the averages with and without inclusion of the two CNG vehicles. CNG vehicles have significantly higher CH4 emissions due to the use of CH4 as fuel. As seen from Table 1, within the experimental uncertainties and with the exception of CNG, there is no discernible effect of fuel type on the level of methane emission from the vehicles. The uncertainty displayed in the table is the standard error of the population. On the basis of this sample (Table 1), we observe that truck CH4 emissions are roughly 40% higher than those of the cars. Light trucks comprise approximately 50% of vehicle sales in the United States, and we average the car and truck values to obtain an emission rate of 12 ( 3 mg/km (19 ( 5 2006
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 7, 2004
mg/mi) for the on-road U.S. fleet. Emissions rates are often stated as a fraction of a more commonly measured value such as CO2 or NMHC. When scaled to CO2 and NMHC, the CH4 emission factor is grams of CH4/grams of CO2 ) (5 ( 1) × 10-5 and grams of CH4/grams of NMHC ) 0.23 ( 0.04. This corresponds to a grams of CH4/grams of THC of about 0.19. The ratios (vs CO2 and NMHC) for cars and trucks are not statistically different. While there is a greater physical link between total hydrocarbons and CH4 production (as compared to CO2), it is easier to calculate vehicle CO2 than NMHC emissions. Hence, it is more practical to use the grams of CH4/grams of CO2 ratio to estimate the CH4 emissions of the global vehicle fleet. Cold versus Hot Running Emissions. In any modern emissions study, it is important to attempt to quantify the difference between cold start and hot running emissions. Before the emissions system has warmed, the emissions tend to be somewhat higher (at any given moment) than afterward. This additional emission is referred to as “cold start”. A large portion of the CH4 emissions measured during chassis tests occur during the cold start portion. In the past, many studies have calculated cold start emissions by subtraction of the hot soak bag 3 emissions of the FTP test from the cold start bag 1. A “bag” refers to a portion of the laboratory test. The first 505 s of the FTP is referred to as bag 1 and begins with a cold start. The next 865 s is referred to as bag 2 and is driven while the vehicle is “hot”, thus it is “hot running”. The next 505 s is conducted after a 10-min soak, when the engine is off. The vehicle is restarted (“hot start”) and is then driven on the same driving trace as bag 1. The final portion, bag 4, is the highway portion and is also a hot running cycle. It should be noted though that this methodology for calculating cold start has large uncertainties since bag 3 can have significant emissions prior to the period when the catalyst reaches a sufficient temperature to be active (typically referred to as “light-off”). Time-series data (taken in 3-s intervals) for four of the vehicles in this study were examined, so a direct measurement of cold versus hot running emissions could be used to estimate the hot portion for the entire data set. Figure 1 shows the time trace for a typical vehicle. The trace clearly indicates that a large fraction of the CH4 emissions occurs during the cold start period before the catalysts light-off. The occasional negative values originate from the low concentration levels; the instrument can report negative concentrations if the measured values are less than the background concentration. This indicates that some measurements are near the noise level of the instrumentation. For the four vehicles, the average fraction of FTP CH4 emissions from the cold start period (determined at the light-off point) is 52%. Figure 2 shows the CH4 emissions in the hot and cold periods for each of the vehicles. Table 2 shows the approximate catalyst light-off time for the vehicles. Applying this hot running fraction to the emissions rate in the previous section, we obtain a hot running CH4 emission rate of 6 ( 1 mg/km and a cold start factor of 111 ( 28 mg per start (the FTP test is 17.8 km). To further differentiate between cold cold start and hot cold start (bag 1 vs bag 3), it is found that the cold cold starts account for 82% of the overall start emissions on the FTP cycle. The hot CH4/CO2 ratio is determined from Figure 3. Each point represents 3 s of data excluding the cold start portion of the tests but including some highway schedules. For this subset of vehicles, hot CH4/CO2 ) (1.4 ( 0.1) × 10-5, and the overall (cold and hot) CH4/CO2 ) (5 ( 1) × 10-5. The overall CH4/CO2 emission factor for the four vehicles is indistinguishable from that of the broader data set, suggesting that the four vehicles are a representative sampling of the 30 vehicle data set.
FIGURE 1. Typical time trace of methane emissions. Highway cycle is included, and the 10 min soak between bags 2 and 3 has been omitted for ease of viewing.
FIGURE 2. Cold start (bags 1 and 3) and hot running methane emissions (g) from four vehicles. On average, 52% of the emissions were during the cold period.
TABLE 2. Approximate Catalyst Light-Off Timea for Four Vehicle Model Subsets vehicle
bag 1
bag 3
vehicle
bag 1
bag 3
1 2
75 285
60 0
3 4
123 135
96 78
a
In seconds.
Unfortunately, the statistical significance of the correlation (even for hot running emissions) is low, reflecting the difficulty of measuring such low levels of CH4. Moreover, CO2 emissions is not much higher during cold start as the other controlled pollutants are, thus one should be cautious when applying these rates. This is one of the reasons why it is more meaningful to use the CH4/NMHC ratio instead (although less practical from a global accounting standpoint). However,
even this ratio has its limitations. For the subset of vehicles, 86% of the NMHC emissions are formed during cold starts (as opposed to 52% for CH4). This is complicated by the fact that the catalytic converter on an automobile converts some of the NMHC into CH4. To determine the impact of automobiles on atmospheric levels of methane more accurately, it would be necessary to separate cold from hot start emissions (especially as emissions standards tighten globally). However, doing so would require knowledge of the number of vehicle starts and the amount of hot driving. This scale of analysis is beyond the scope of this paper. For the purposes of the present study of vehicle methane emissions, we combine cold and hot running emissions into one factor. Model Year and Manufacturer. To complement the present experimental study, we summarize data collected from a variety of other sources to reflect model year and VOL. 38, NO. 7, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2007
FIGURE 3. CH4 as a function of CO2 using hot running modal data from four vehicles. manufacturer comparisons. Historical CH4 emission data (direct and indirect measurements) were collected from a variety of sources and are presented for comparison with the present measurements. Emission data for 2000 model year vehicles were obtained from the Ford Emissions Certification Database. Emission data for the 1980-1995 model year vehicles were obtained from the EPA Emissions and Fuel Economy Data (10). Emission data for European 2001 model year vehicles were obtained from the EU certification database (11). Figure 4 shows methane emissions, expressed as [CH4]/ [CO2] histograms over the UDDS cycle for new (e1 year old) vehicles from 1980 to 2000 model years. Table 3 shows the manufacturer, data source, and number of data points (i.e., number of vehicles tested) used to generate the data shown in Figure 4. Since direct measurements of methane emissions were not available for 1980-1995 vehicles, a factor of 0.19 (see CH4 Emissions Measured over the UDDS Cycle) was applied to the reported total hydrocarbon data to obtain the CH4 emission rate results shown in Table 4 and Figure 4. Reported CH4 emission data were used directly from the 2000 data. Figure 4 indicates that methane emissions have been decreasing with model year for the past two decades. This is consistent with the introduction of new hardware on more recent model vehicles and clean fuels to comply with increasingly stringent emissions standards. From the data in Figure 4 and Table 4, it seems reasonable to conclude that over the time period 1980-2000 the methane emissions from the global vehicle fleet have probably declined substantially (despite an approximately 50% increase in fuel use; 12). Since most of the above data is from Ford vehicles, it is helpful to draw a comparison with other manufacturers. This is done in Figure 5 for a cross section of 1995 vehicles. We note that the CH4 emissions are fairly consistent among the fleet. For the purposes of generalizing to global emissions, it is helpful to compare with European data as well. Figure 6 and Table 4 show data for 2001 model year European vehicles derived from reported THC (total hydrocarbon) emissions and assuming the same CH4/THC factor of 0.19. Comparison 2008
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 7, 2004
FIGURE 4. New vehicle (Ford) methane emissions rate trends in North America. CH4/CO2 ratios as a function of model year (see Table 3 for data source, see Table 4 for statistical information concerning the distributions). of the data in Figures 4 and 6 shows that emissions from modern European and U.S. vehicles are low and of a similar magnitude (on a per vehicle basis).
TABLE 3. Data Source for Model Year Studya model year 1980 1985 1990 1995
data points (n ))
manufacturer
source
214 289 281 350
EPA website EPA website EPA website EPA website
2000
F/L/M F/L/M F/L/M, Jag F/L/M, Jag, Mazda F/L/M
240
2001
Audi-VW
207
Ford Certification Group Vehicle Certification Agency
a
F, Ford; L, Lincoln; M, Mercury; Jag, Jaguar; VW, Volkswagen.
TABLE 4. Model Year [Methane]/[CO2] Emission Ratea Results North America
max. min. median a
Europe
1980
1985
1990
1995
2000
2001
40.0 2.9 11.7
35.0 4.7 11.3
21.1 2.5 7.9
18.1 2.1 6.6
13.9 0.4 1.9
10.9 1.5 4.9
Units of 10-5.
FIGURE 5. 1995 model year methane emissions by test type. FTP CH4 emissions separated by manufacturer.
FIGURE 6. Methane emissions for 2001 model year European vehicles. Relationship to On-Road Fleet. The vehicles studied in this work were approximately 1 year old at testing and had been maintained properly. To relate these measurements to the likely emission of the on-road fleet, we need to make allowance for two important factors. First, the on-road fleet contains vehicles of a variety of ages. Second, a small fraction of the vehicles in the on-road fleet are likely to be in a poor state of repair and have higher emissions. To account for vehicle aging and the presence of poorly maintained vehicles, we use the approach adopted in the EPA MOBILE6 inventory model (13, 14) with the following assumptions: CH4 emissions
will be affected by vehicle age and maintenance condition at the same rate as NMHC emissions, light trucks will age at roughly the same rate as cars, the emissions rate modification is for Tier 1 vehicles only, and CO2 emissions rates do not change significantly with vehicle age. The EPA MOBILE6 inventory model predicts that the on-road U.S. fleet average CH4 emission will be approximately 3.4 times that measured from the vehicles studied here. Thus, we arrive at an estimate of the U.S. fleet average of CH4/CO2 ) (15 ( 4) × 10-5. The emission control technologies used on vehicles in Europe and Japan are similar to those in the United States, so it is reasonable to assume that this emission rate will also be appropriate for the fleets in these areas. It should be noted that these estimates are based on model predictions only. More measurements over a number of years are required for an improved estimate.
Comparison with Previous Work This paper presents results from chassis dynamometer emission measurements at the Ford Motor Company using 30 different Tier 1 cars and trucks (from four different manufacturers). The emission factor (g of CH4/g of CO2) measured in the “laboratory” was (5 ( 1) × 10-5 for nonCNG-powered vehicles. For (the two) CNG vehicles, the emission rate was (2 ( 1) × 10-4. From examining the timeseries emissions from a subset of four vehicles (consistent with the average set), a hot running emissions ratio was determined to be (1.4 ( 0.1) × 10-5. Approximately 52% of the CH4 emissions during an FTP test occurred during the cold starts. Of that, roughly 82% of the start emissions occur in bag 1, and 18% occur in bag 3 of the FTP cycle. This gives a cold start factor of 11 ( 2 mg/bag 1 start and 3 ( 1 mg/bag 3 start, respectively. Second-by-second investigation of four of the vehicles indicated that CH4 emissions increases only mildly with road load (speed and acceleration). This suggests that the catalyst oxidizes methane efficiently, even under the stresses of a highway drive schedule. We recommend use of an overall CH4/CO2 emission factor of (15 ( 4) × 10-5 when calculating traffic-related CH4 emissions for emissions inventory purposes. The emission rates measured in the present work (see Table 1) are significantly less than those reported by Koike and Odaka (5). There are three likely explanations for this discrepancy. First, the vehicles studied by Koike and Odaka were model year 1996 or prior. As seen from Figure 4 there has been a steady reduction in vehicle CH4 emissions since the 1980 model year. Although Koike and Odaka (5) do not specify the model year of the vehicles used in their study (published in 1996), inspection of the data in Figure 4 suggest that the magnitude of the difference in the reported CH4 emissions in the two studies is unlikely to be explained solely by vehicle model year. Second, statistical effects associated with the small sample size (four) of three-way catalyst equipped vehicles in the study of Koike and Odaka (5) could bias their data. In the present work 30 vehicles were tested. Third, Koike and Odaka employed an indirect methodology where methane emissions were inferred from the difference between the measured total hydrocarbons (THC) and NMHC rather than the direct measurement (using FTIR or GC) employed herein. This technique is inherently less precise than that employed in the present study. During the course of the present work, Heeb et al. (15) published the results of a study of methane emissions from gasoline passenger cars in Switzerland. Heeb et al. (15) reported methane emission rates from 49 cars tested using the same driving cycle as that used in our study. The fleet of 18 vehicles designated “EURO-2” were of model years 1997-1998 and were most similar to those investigated in the present work. The distance averaged CH4 emission rate VOL. 38, NO. 7, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2009
of 15 ( 9 mg km-1 for the EURO-2 fleet calculated from the data in Table 2 from ref 15 is in good agreement with the average (non-CNG) car emission rate of 10 ( 1 mg km-1 from the present work (see Table 1).
Implications for Atmospheric Chemistry and Climate Change It is useful to place our emission data into perspective in terms of vehicle contributions to the global CH4 budget and radiative forcing of climate change. Using values of 964 Tg (1 Tg ) 1012 g) for the annual global vehicle fuel consumption in 1995 (637 Tg of gasoline, 327 Tg of automotive diesel; 16), 0.855 for the average carbon content of gasoline by mass (17), 44 for the molecular weight of CO2, 12 for the atomic weight of carbon, and (15 ( 4) × 10-5 for the emission factor of CH4 (assuming that the U.S. fleet is broadly representative of global fleet), we arrive at an estimate for the contribution of vehicular traffic to the global CH4 budget of 964 × 0.855 × (44/12) × (15 ( 4) × 10-5 ) 0.45 ( 0.12 Tg. Global anthropogenic CH4 emissions (from natural gas facilities, coal mines, petroleum industry, coal combustion, enteric fermentation, rice paddies, biomass burning, landfills, animal waste, and domestic sewage) are approximately 360 Tg yr-1 (18). The global vehicle fleet is responsible for
