Evaluation of Representativeness of Site-Specific Fuel-Based

The average fuel-based emission factors increased with vehicle specific power (VSP) ... vehicles that are subsequently further tested under inspection...
0 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/est

Evaluation of Representativeness of Site-Specific Fuel-Based Vehicle Emission Factors for Route Average Emissions Taewoo Lee† †

National Institute of Environmental Research, Gyeongseo-dong, Seo-gu, Incheon 404-708, Korea

H. Christopher Frey*,‡ ‡

Department of Civil, Construction and Environmental Engineering, North Carolina State University, Campus Box 7908, Raleigh, North Carolina 27695-7908, United States S Supporting Information *

ABSTRACT: An approach to evaluate the representativeness of site-specific fuel-based vehicle emission factors, such as would be obtained using Remote Sensing Devices (RSDs) is demonstrated based on real-world data for 23 selected light duty gasoline vehicles. Real time vehicle route-average emissions rates were measured using a Portable Emissions Measurement System (PEMS) for a variety of road types and traffic characteristics. Several hypothetical remote sensing sites were selected to estimate site-specific fuel-based emission factors. The average fuel-based emission factors increased with vehicle specific power (VSP) and varied by a factor of 3 and 4 for NOx and CO, respectively. The route average emission factors varied by approximately 20% for either NOx or CO. The site-specific emission factors varied among specific sites by 20 and 30% for NOx and CO, respectively. Fuel-based HC emission rates had little variability with engine load, among routes, or between sites. Arbitrarily chosen sites can lead to potential biases for CO and NOx if measured emission factors are used for route average rates and, therefore, for area-wide inventories. However, site-specific emission factors have the potential to be representative of area-wide emission rates if the distribution of positive VSP at the site is similar to that of routes or area-wide cycles of interest.

■

INTRODUCTION This research provides insight regarding the robustness of sitespecific fuel-based highway vehicle emission factors as surrogates for route-average emission rates that are a typical component of an area-wide emission inventory. Area-wide inventories depend on driving cycles that, in combination, represent point to point travel along routes. Fuel-based emission factors [g/gal] are inferred from tunnel studies or Remote Sensing Devices (RSD).1−3 Fuel-based emission factors for a vehicle group are multiplied by the amount of fuel sold for a geographical area to obtain an emission inventory.4−6 Fuel-based approaches are also useful for benchmark comparison to other inventory approaches for verification purposes. A RSD measures the ratio of exhaust emissions, such as nitric oxide (NO), hydrocarbon (HC), and carbon monoxide (CO), to carbon dioxide (CO2). Fuel-based emission rates are calculated based on these ratios. A typical RSD site has only one lane of traffic, to avoid interference from passing vehicles, and a positive road grade to ensure that the engine is under load. 7 Fuel-based emission factors provide a location snapshot © 2012 American Chemical Society

that may not represent emissions during a trip or for point-topoint vehicle operations along a route in a geographic area.8 Fuel-based emission factors vary less with engine load than distance-based emission factors.9,10 However, site-specific emission factors nonetheless can vary with engine load.11−14 The representativeness of site-specific fuel-based emission factors can be assessed by comparison to route average vehicle emission rates, which can represent area-wide emissions for each vehicle. Light duty vehicles, including passenger cars and passenger trucks (minivans, SUVs, pickup trucks), comprise over 80% of the U.S. on-road fleet. In the U.S., over 99% of passenger cars and over 98% of passenger trucks use gasoline. Therefore, the focus here is a sample of light duty gasoline vehicles that represent various model years, engine sizes, and body types. Pokharel et al.15 reported that the U.S Environmental Protection Agency (EPA)’s MOBILE emissions model Received: Revised: Accepted: Published: 6867

December 12, 2011 April 23, 2012 May 8, 2012 May 8, 2012 dx.doi.org/10.1021/es204451z | Environ. Sci. Technol. 2012, 46, 6867−6873

Environmental Science & Technology

Article

Instruments. The key instruments used to measure vehicle activity and emissions include the following: (1) a PEMS; (2) a vehicle On-Board Diagnostic (OBD) data logger; and (3) a geographic position system (GPS) with barometric altimeter. The volume percentage of CO2, CO, HC, and NO undiluted tailpipe exhaust concentrations were measured using the Axion system PEMS manufactured by Clean Air Technologies International, Inc. HC, CO, and CO2 are measured using nondispersive infrared (NDIR). NO is measured using electrochemical cell. NOx is typically comprised of approximately 95 volume percent NO; therefore, NO emissions converted to an equivalent NO2 mass basis (using the molecular weight of NO2) are a good indicator of total NOx emissions. The performance of PEMS has been verified in comparison to that of a laboratory grade chassis dynamometer.21 For CO2, CO, and NO, when comparing cycle average PEMS to dynamometer measurements, the coefficients of determination were 0.86 or higher, and the slopes were between 0.92 and 1.05. NDIR measurements of vehicle exhaust HC are biased low because NDIR is well-known to respond only partially to a typical HC speciation profile of vehicle exhaust.22 NDIR responds accurately to low-molecular weight straight-chain alkanes. The response is biased low for branched alkanes and aromatic compounds. The reported comparisons of NDIR with flame ionization detection (FID) measurements imply that HC emissions measurements obtained using NDIR should be multiplied by approximately a factor of 2 to obtain a more accurate indication of total HC emissions. NDIR measurements are compared here only for the same vehicle, assuming that the bias correction would be the same for the site specific and route average emissions. The PEMS is calibrated in the laboratory using a cylinder gas and in the field periodically recalibrates to ambient air to prevent instrument drift. On a second-by-second basis, emission concentrations were combined with exhaust flow rate derived from OBD data to get mass emissions per second. Fuel use rate was estimated based on engine air flow estimated from OBD data and air-to-fuel ratio inferred from the concentrations of exhaust gases. GPS coordinates were used to match vehicle location relative to each test route and to estimate distances between elevation measurements. The differences in elevation over distance were used to estimate road grade. Road grade is strongly correlated with vehicle fuel use and emission rates.23 Emission Factors. Fuel-based emission factors were developed to evaluate variability in emission factors with engine load and among driving cycles. Several hypothetical remote sensing sites are identified for the purpose of estimating site-specific fuel-based emission factors. Modal Average Fuel-Based Emission Factors. Modal average emission factors are defined as the average of secondby-second emission rates stratified based on ranges of VSP. VSP is an indicator of engine load that takes into account kinetic energy, road grade, tire rolling resistance, and aerodynamic drag. VSP is computed using second-by-second speed, acceleration, and road grade. For a typical light-duty vehicle, VSP is estimated as24

estimated area-wide emission factors were 40 to 80% higher for NO, 40% lower for HC, and 30 to 70% higher for CO versus RSD measurements. Kuhns et al.14 compared RSD and MOBILE emission factors. Measured emission factors were about 50% greater for NOx, about the same for HC, and about 50% lower for CO, with larger differences among older vehicles. Ekstrom et al.16 compared RSD measurements to the European COPERT III emission model for gasoline passenger cars. The comparison was consistent for NO and HC but less consistent for CO. Bias in fuel-based emission inventories can result from incomplete consideration of the relationship between fuelbased emission factors and engine load.6,15 Remote sensing is used as a screening tool to identify possible high-emitting vehicles that are subsequently further tested under inspection and maintenance programs.17,18 Sjodin and Lenner19 compared RSD and chassis dynamometer data and reported that average RSD CO emission rates were considerably higher because of a small fraction of ‘highemitting’ cars. The high emission rates may be related in part to fuel enrichment, in which the vehicle electronic control unit (ECU) operates the engine with a fuel-rich mixture at high engine load. This, in turn, produces more power while reducing oxygen in the exhaust to prevent CO oxidation from overheating and damaging the catalyst. Therefore, fuel enrichment can lead to episodes of high emissions for an otherwise normally functioning vehicle.20 The objective of this research is to demonstrate an approach to evaluate the representativeness of site-specific fuel-based vehicle emission factors for route average emissions. The key supporting goals are as follows: (1) to quantify variability in fuel-based emission factors for varying engine load; (2) to quantify variability in route average fuel-based emission factors among different driving cycles that represent area-wide vehicle operation; (3) to quantify site-specific fuel-based emission factors typical of possible remote sensing sites; and (4) to demonstrate how well site-specific fuel-based emission factors estimate route average emission factors.

■

METHODOLOGY Second-by-second fuel use and emission rates for 23 light-duty gasoline vehicles were measured using a Portable Emissions Measurement Systems (PEMS). Vehicle specific power (VSP) is used to quantify the relationship between emission rate and engine load. Route average emissions rates for four selected routes were compared to site-specific fuel-based emission factors that simulate RSD measurements. Data Collection. The 23 selected gasoline vehicles have engine displacement of 1.6 to 3.8 L, 3K to 228K accumulated miles, and model years ranging from 1997 to 2008. The vehicles were measured on selected round-trip routes for two origin and destination (O/D) pairs: (1) North Carolina State University (NCSU) to the north Raleigh residential area (NR) and (2) NR to Research Triangle Park business district (RTP). Each O/D pair has two alternative routes: Routes A and C for the NCSU/NR O/D pair and Routes 1 and 3 for the NR/RTP O/D pair. Routes A and C are comprised primarily of signalized minor and major arterial roads and represent local commuting over a one-way distance of approximately 8 to 11 miles. Routes 1 and 3 are comprised primarily either of freeways or major arterials, respectively, over a one-way distance of approximately 16 to 20 miles. The four routes entail approximately 110 miles driven per vehicle and represent a variety of area-wide road types and traffic characteristics.

VSP = v × [1.1 × a + 9.81 × (sin(tan−1(φ))) + 0.132] + 0.000302 × v 3

(1)

where VSP = vehicle specific power (kW/ton); v = vehicle speed (m/s) each second; a = acceleration (m/s2) each second; and φ = road grade. 6868

dx.doi.org/10.1021/es204451z | Environ. Sci. Technol. 2012, 46, 6867−6873

Environmental Science & Technology

Article

location on each ramp is selected taking into consideration the VSP distribution along the path on the ramp and the physical feasibility, such as safety and RSD installation space.

Second-by-second fuel use and emission rates were categorized into fourteen VSP modes, as shown in Table 1.24

■

Table 1. Definition of Vehicle Specific Power Modes24 VSP mode

definition (kW/t)

1 2 3 4 5 6 7 8 9 10 11 12 13 14

VSP < −2 −2 ≤ VSP < 0 0 ≤ VSP < 1 1 ≤ VSP < 4 4 ≤ VSP < 7 7 ≤ VSP < 10 10 ≤ VSP < 13 13 ≤ VSP < 16 16 ≤ VSP < 19 19 ≤ VSP < 23 23 ≤ VSP < 28 28 ≤ VSP < 33 33 ≤ VSP < 39 39 ≤ VSP

RESULTS Variability in modal average fuel-based emission factors among the selected vehicles is estimated for a range of engine load. Based on the distribution of engine load for each of four routes, route average emission factors are estimated. Site-specific and vehicles-specific average fuel-based emission factors estimated at a site are compared to route average fuel-based emission factors for the same vehicle, to provide insight as to under what conditions a site-specific emission factor may have the same value as a route average emission factor. Modal Average Fuel-Based Emission Factors. The relative variability in time-based fuel flow rate and fuel-based emission factors among the 23 vehicles are shown in Figure 1. For each vehicle, the modal average rates were divided by the empirical average rate. The relative trend in fuel use rate versus VSP is similar among each of the vehicles. The wider range of variability at high VSP is attributed to smaller sample sizes. The fuel use rate is highly sensitive to engine load and varies by a factor of 8 when comparing the highest to lowest modal rate. The average normalized fuel-based NOx emission factor increases with VSP and varies by a factor of 3 from the highest rate at Mode 14 to lowest rate at Mode 3. Variability in emission rates of individual vehicles ranges from factors of 1.6 to 25, when comparing the highest to lowest modal rates. Intervehicle variability in emission rates within a VSP mode, as illustrated in the box-whisker plot, is lowest in the middle range of engine load and increases at higher and lower engine load. The modal average HC emission factors are less variable when comparing VSP modes compared to those of other pollutants. The ratio of the highest to lowest average emission factor among vehicles is 1.2 and ranges from 1.1 to 2.6 for individual vehicles. The average normalized CO emission factor varies only slightly among Modes 1 to 11, with a ratio of 1.4 for the highest to lowest average emission factor among these modes. The ratio of the highest to the lowest average modal rates among all modes is 4.3 and ranges from 1.6 to 52 for individual vehicles. The very high average normalized CO emission rates at Modes 13 and 14 for several vehicles are associated with “fuel enrichment” operation. Route Average Fuel-Based Emission Factors. Route average emission rates are inferred from a weighted average of modal emission rates. Variation in traffic conditions can influence a vehicle driving cycle. On average, Route A has a lower frequency of higher VSP modes than Route 1, as shown in Figure 2. Route 1 has the highest average engine load, and Route A has the lowest average. Routes C and 3 have similar VSP frequency distributions, which lie between those of Routes A and 1. The average VSP values for each vehicle and route fall into VSP Modes ranging from 4 to 7. Route average emission factors vary among Routes A, C, 1, and 3, as shown in Table 2. Route 1 has the highest average emission rate among the four routes for NOx and CO. HC has little variability among the routes. On average, the route average emission factors vary by approximately 20% when comparing the highest to lowest average rates, for either NOx or CO. For individual vehicles, the route average NOx emission factor varies by 2 to 66% among the routes, and by 2 to 72% for CO.

VSP-based modal emission models are developed for fuel use rate and fuel-based emission factors for each pollutant and vehicle. Modal average emission rates are used to avoid random artifacts. For site-specific data, the emission rate would typically be quantified based on the average of repeated measurements and, thus, is appropriately represented by a mean value. Variability in site-specific RSD data can span orders of magnitude when comparing individual measurements among passing vehicles.25 However, the goal here is not to characterize an entire on-road fleet but rather to compare average site specific emission rates to route average emission rates for the same vehicle, taking into account how much the ratio between these two emission rates varies among a sample of vehicles. Route Average Fuel-Based Emission Factors. The average fuel-based emission factors for each route are estimated based on Ej,k divided by Fk 14

Ej , k =

∑ (FER i ,j·FR i·Ti ,k /3600) i=1

(2)

14

Fk =

∑ (FR i·Ti ,k /3600) i=1

(3)

where Ej,k = total emission amount of pollutant j in Route k (g); Fk = total fuel use in Route k (gal); FERi,j = fuel-based emission factor of pollutant j in VSP mode i (g/gal); FRi = modal fuel use rate in VSP mode i (gal/h); Ti,k = time spent in VSP mode i in Route k [seconds (s)]; subscript i = VSP modes (1 to 14); subscript j = pollutants (NOx, HC, and CO); and subscript k = routes (A, C, 1, and 3). For route-average emission rates, emission predictions are typically made using emission factor models, such as MOVES, based on a weighted average of modal emission rates. Site-Specific Fuel-Based Emission Factors. Average sitespecific measurements for an individual vehicle, which result in site-specific fuel-based emission factors, are simulated using the VSP-based modal emission models and one second of VSP at hypothetical remote sensing sites. The selected hypothetical sites are located in single lane freeway ramps. One ramp is located between Hillsborough Street and I-440 beltline westbound. A second ramp is located between I-440 westbound and Six Folks Road. The specific 6869

dx.doi.org/10.1021/es204451z | Environ. Sci. Technol. 2012, 46, 6867−6873

Environmental Science & Technology

Article

Figure 2. Frequency distribution of typical VSP in four selected routes and VSP for 23 vehicles at hypothetical RSD sites. Typical VSP distribution means an average driving cycle on overall 23 vehicles. VSP data of each route were binned by 1 kW/t interval. Vertical lines denote intervals of 14 VSP modes.

VSP Modes range from 6 to 13 and 2 to 13, for Sites R1 and R2, respectively. Site R1, which has a higher average VSP than Site R2, has 20 and 30% higher site-specific fuel-based emission NOx and CO factors, respectively, as shown in Table 2. Site-Specific versus Route Average Fuel-Based Emission Factors. For each vehicle, there are eight site-to-route combinations when comparing each of Sites R1 and R2 to each of the four routes. Intervehicle variability in the ratios of sitespecific to route average emission factors are quantified using cumulative distribution functions (CDFs) as shown in the left panels of Figure 4. The right panels of Figure 4 show 95% confidence intervals for the mean ratio for each site-route pair estimated based on 23 vehicles. For NOx the cumulative frequency distribution ranges from 0.28 to 2.98 for the ratio of Site R1 to Route A emission rates. This implies that the site-specific NOx emission factor at R1 underestimates the route average as much as 72% for one vehicle and overestimates by up to a factor of 3 for another vehicle. For site-route ratios based on Site R2, the frequency distributions have a tighter cluster of central tendency values than those based on Site R1, and the range of intervehicle variability is substantial, from 0.51 to 1.83. For NOx, five of the eight site-route combinations have mean ratios that are significantly different than 1.0, which indicates that site-specific emission factors are biased compared to the route average. On average, site-specific fuel-based emission factors at Sites R1 and R2 overestimate route average NOx emission rates by 27% and 9%, respectively. Site-specific HC emission factors accurately estimate route average emissions rates. The average ratio of site-to-route average emission factors is not significantly different from 1.0 for all site-route pairs. There is also less intervehicle variability in the site-route ratios than for NOx, with values typically ranging only between 0.8 and 1.4. The site-route ratio of emission factor comparisons for CO has qualitatively similar trends as NOx with respect to intervehicle variability among the site-route pairs. However, there is substantially more intervehicle variability in the Site R1 comparisons than for the Site R2 comparisons. The frequency distributions for intervehicle variability in the R1-to-route comparisons range from 0.32 to 6.0, which are much wider than those for the R2-to-route comparisons. On an average basis, site-specific emission factors at R1 overestimate the route emission factors by 45%, versus only 7% for R2. However, because of the wide range of intervehicle variability, particularly for the Site R1 comparisons, the mean ratio for each site-route pair is not significantly different from 1.0.

Figure 1. Variability of normalized fuel use and fuel-based emission rates for 23 vehicles. Modal rate divided by empirical average rate for each vehicle. Circles denote average normalized emission rates for 23 vehicles. The bottom and top of the box are the 25th and 75th percentiles, respectively. The band near center of the box is the 50th percentile. The upper and lower ends of the whiskers represent the 95th and fifth percentiles, respectively.

Site-Specific Fuel-Based Emission Factors. Variability in site-specific emission factors depends on variability in modal emission rates, as shown in Figure 1, and variability in one second VSP values at specific sites. VSP values along the path at two selected ramps range from approximately −30 to 50 kW/t, as shown in Figure 3. The location of Sites R1 and R2 are indicated in the figure. The average road grade of Sites R1 and R2 is approximately 2.2% and 1.3%, respectively. Alternative Sites R3 and R4 are discussed later. VSP distributions differ among sites and between routes, as shown in Figure 2. Average VSP values at Sites R1 and R2 are 22.8 kW/t (Mode 10) and 12.8 kW/t (Mode 7), respectively. 6870

dx.doi.org/10.1021/es204451z | Environ. Sci. Technol. 2012, 46, 6867−6873

Environmental Science & Technology

Article

Table 2. Average and Intervehicle Variability in Route Average and Site-Specific Fuel-Based Emission Factors for Selected Vehicles routes

sites

pollutants

quantitya

A

C

1

3

R1

R2

R3

R4

NOxb

average emission factor (g/gal) coefficient of variance average emission factor (g/gal) coefficient of variance average emission factor (g/gal) coefficient of variance

6.03 1.09 3.38 0.53 31.6 1.22

6.48 1.10 3.38 0.53 34.6 1.24

7.11 1.12 3.35 0.52 37.0 1.32

6.63 1.11 3.38 0.53 34.9 1.28

8.94 1.13 3.37 0.51 49.6 1.38

7.42 1.11 3.29 0.48 38.3 1.42

7.50 1.15 3.43 0.49 53.2 1.71

5.05 1.10 3.45 0.60 27.4 1.20

HCc COd,e a

Coefficient of variation is standard deviation divided by average. bVehicle J was excluded because of its malfunctioning of the NOx control system. Vehicle I was excluded because its HC emission rates are below detection limit of analyzer. dVehicles F and G were excluded because their CO emission rates are below detection limit of analyzer. eIf Vehicle U is excluded from both route and site average, the route average CO emission factors are 29.3, 31.4, 33.0, and 31.2 g/gal for Routes A, C, 1, and 3, respectively, and site-specific CO emission factors are 42.6, 34.0, 36.8, and 26.7 g/gal for Sites R1, R2, R3, and R4, respectively. c

Remote Sensing Site Selection Criterion. The sitespecific emission factors at Site R2 are approximately representative of the route average emission factors for Route 1 to within 2 to 5% among NOx, HC, and CO, as indicated in Table 2. Conversely, site-specific emission factors at Site R1 were substantially higher than those for Route A by 48 and 57% for NOx and CO, respectively. The concordance between a specific site and a route average emission factor is hypothesized to be influenced by the VSP distribution of each. For example, the VSP distributions for Site R2 and Route 1 are similar at the upper tail, as indicated in Figure 2. In contrast, the VSP distribution for Site R1 is substantially higher than that of Route A. To further explore the hypothesis, two additional sites are introduced. Site R3 was selected to produce a VSP distribution similar to that of Site R2, to provide insight regarding whether site specific emission rates are similar if the VSP distributions are similar. Site R4 was selected to produce a VSP distribution that is more similar to that of many of the routes, to provide insight regarding whether a location-specific emission factor might be comparable to a route average emission factor. R3 has a similar VSP distribution as R2 but is at a different location on the ramp as indicated in Figure 3(a). Average VSP values at R2 and R3 are 12.8 and 12.5 kW/t, respectively, both of which are contained within VSP Mode 7. For NOx and HC, site-specific emission factors at R2 and R3 are each similar to within less than 5%, based in Table 2. However, the average Site R3 CO emission factor is 39% higher than that of R2. This difference results mainly from one high emitting vehicle. In particular, Vehicle U is one of the highest CO emitters among the vehicles and also has substantial intermodal variability in emission rates likely attributable to fuel enrichment at high VSP modes. Vehicle U was operating in VSP Mode 11 at Site R2 and Mode 13 at Site R3. The corresponding modal emission factors are 120 and 380 g/gal, respectively, which differ by a factor of 3. Thus, a high emitter can significantly affect the average emission factors at a given site. If Vehicle U is excluded, site-specific CO emission factors at Sites R2 and R3 are similar at 34.0 and 36.8 g/gal, respectively, a difference of only 8%. These results indicate that site-specific emission factors have the potential to be comparable if their VSP distributions are similar. However, these results also indicate that the average emission rates for a group of vehicles can be substantially influenced by a small proportion of high emitters. For example, based on RSD data collected in 2010 for nearly 13,000 vehicles at a site in Van

Figure 3. VSP distribution for 23 selected vehicles along the selected ramps and location of hypothetical remote sensing sites. “R1”, “R2”, “R3”, and “R4” mean the location of hypothetical remote sensing sites. Each data point is a VSP value for one vehicle at a given location.

The wide range of intervehicle variability in the site-route ratios for CO emission factors, particularly related to Site R1, is attributable to a small fraction of the sample vehicles. The few vehicles that have site-to-route emission factor ratios of greater than 2.0 at Site R1 were typically operating in VSP Modes 10 to 13 at the site. Furthermore, these same highly loaded vehicles also have high intermodal variability in emission rates. The combination of high power demand and high emission rates that may be influenced by fuel enrichment lead to high sitespecific emission factors and, thus, the high site-to-route ratio. The highly emitting vehicles in this situation have a wide range of model years, and thus the fuel enrichment effect is not likely to be a strong function of vehicle age. 6871

dx.doi.org/10.1021/es204451z | Environ. Sci. Technol. 2012, 46, 6867−6873

Environmental Science & Technology

Article

Figure 4. Frequency distributions, average, and variability for the ratio of site-specific to route average emission factors for 23 vehicles. HC result of Vehicle I and CO results of Vehicle F and Vehicle G are excluded from the data set because those data are below detection limit of analyzer. Vertical solid lines in panels denote ratio of 1.0. Error bars in part (b) denote 95% of confidence intervals.

Nuys, CA, the percent of total emissions from the dirtiest 10% of the fleet was as follows: 86% for CO; 90% for HC; and 72% for NOx.26 Of the 23 observed vehicles, the top 3 contribute up to 76, 28, and 65% of the NOx, HC, and CO emissions, respectively, depending on the site. The average emission rate among the 23 vehicles is about twice as high for NOx and CO and 15% higher for HC compared to when the top three emitting vehicles are excluded. Site R4 has the same average VSP value as the average value among the four routes at 5.0 kW/t. The VSP distribution at Site R4 overlaps those of each of the four routes. However, the Site R4 site-specific NOx emission factor underestimates the route averages by 16 to 29%. The route average CO emission factors are underestimated by 13 to 26%. Thus, an association of average VSP is not sufficient to ensure comparability. Modal emission rates are typically constant or subject to relatively little variability for negative VSP and increase monotonically and substantially with positive VSP. Thus, the average emission factor tends to be more sensitive to the distribution of positive VSP than the distribution of all VSP values. Sites R2 and R3 have less than 10% frequency of negative VSP values. The average VSP values of these two sites are similar to the average positive VSP of 11.8 kW/t for Route 1. Except for one high-CO emitting vehicle, the average emission rates at these two sites and at Route 1 are very similar. The work here can be expanded to take into account more vehicles, more sites, and more routes using the same methodology demonstrated here. A further desirable refinement would be to conduct a PEMS-based field study for route average emission factors that includes transects at specific

locations using RSD devices, to better assess the concordance between estimates of emissions at a candidate RSD site based on PEMS data versus what would be measured by an RSD device at the same location. Overall, the results indicate that site-specific HC emission factors are likely to be representative of route average emission factors, site-specific NOx emission factors may be subject to bias, and site-specific CO emission factors might either be biased, imprecise, or both depending on the proportion of high emitting vehicles or fuel enrichment events that affect the rates. To help identify RSD sites that have the potential to be representative of area-wide emission rates, a comparison should be made between the distribution of positive VSP at the potential RSD site versus that of routes or area-wide cycles of interest.

■

ASSOCIATED CONTENT

S Supporting Information *

Tables and figures pertaining to the list of test vehicles, map of four selected routes in Raleigh, North Carolina, maps of hypothetical Remote Sensing sites, frequency distributions of VSP for 23 selected vehicles, route average fuel-based emission rates based on empirical cycles observed for each vehicle versus typical average cycles based on activity measured for all vehicles, evaluation of sources of variability in site-to-route emission ratios, and average VSP values and VSP modes for four routes and average VSP values for the 23 selected vehicles at hypothetical RSD sites. This material is available free of charge via the Internet at http://pubs.acs.org. 6872

dx.doi.org/10.1021/es204451z | Environ. Sci. Technol. 2012, 46, 6867−6873

Environmental Science & Technology

■

Article

gasoline and diesel engine vehicles in Las Vegas, Nevada. Sci. Total Environ. 2004, 322, 123−137. (15) Pokharel, S. J.; Bishop, G. A.; Stedman, D. H. An on-road motor vehicle emissions inventory for Denver: an efficient alternative to modeling. Atmos. Environ. 2002, 36, 5177−5184. (16) Ekstrom, M.; Sjodin, A.; Andreasson, K. Evaluation of the COPERT III emission model with on-road optical remote sensing measurements. Atmos. Environ. 2004, 38, 6631−6641. (17) Pokharel, S. J.; Bishop, G. A.; Stedman, D. H. Analysis of remote sensing data for development of I/M program evaluation protocols; U.S. EPA Report; University of Denver, Denver, CO, 2000. (18) Austin, T.; Burnette, A. D.; Klausmeier, R.; Slott, R. Review of literature on remote sensing devices; Prepared by Eastern Research Group, Inc. for California Air Resource Board and California Bureau of Automotive Repair, 2004. (19) Sjodin, A.; Lenner, M. On-road measurements of single vehicle pollutant emissions, speed and acceleration for large fleets of vehicles in different traffic environments. Sci. Total Environ. 1995, 169, 157− 165. (20) Muske, K. R.; Peyton Jones, J. C. Feedforward/feedback air fuel ratio control for an automotive catalyst, Proceedings of the American Control Conference, Denver, Colorado, June 4−6, 2003, 1386-1391. (21) Battelle. Environmental Technology Verification Report: Clean Air Technologies International, Inc. REMOTE On-Board Emissions Monitor; U.S. Environmental Protection Agency: Columbus, OH, June, 2003. (22) Stephens, R. D.; Mulawa, P. A.; Giles, M. T.; Kennedy, K. G.; Groblicki, P. J.; Cadle, S. H.; Knapp, K. T. An experimental evaluation of remote sensing-based hydrocarbon measurements: A comparison to FID measurements. J. Air Waste Manage. Assoc. 1996, 46 (2), 148− 158. (23) Zhang, K.; Frey, H. C. Road grade estimation for on-road vehicle emissions modeling using light detection and ranging data. J. Air Waste Manage. Assoc. 2006, 56, 777−788. (24) Frey, H. C.; Rouphail, N. M.; Zhai, H. Speed- and facilityspecific emission estimates for on-road light-duty vehicles based on real-world speed profiles. J. Transport. Rec. 2006, 1987, 128−137. (25) Rouphail, N. M.; Frey, H. C.; Unal, A.; Dalton, R. ITS Integration of Real-Time Emissions and Traffic Management Systems, IDEA Project No. ITS-44, Prepared by North Carolina State University for the IDEA Program, Transportation Research Board, National Research Council, Washington, DC, May 2000. (26) Bishop, G. A.; Schuchmann, B.; Stedman, D. H. Multi-species On-Road Remote Sensing of Vehicle Emissions in Van Nuys, California − August 2010, Contract No. AEV-8-88609-01; Prepared by University of Denver for National Renewable Energy Laboratory, Golden, Colorado, 2010.

AUTHOR INFORMATION

Corresponding Author

*Phone: 1-919-515-1155. Fax: 1-919-515-7908. E-mail: frey@ ncsu.edu. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The vehicle emissions measurements at NCSU were supported by the National Science Foundation (NSF) Grant No. CBET0756263. Taewoo Lee’s contribution was supported by the Korean National Institute of Environmental Research (NIER). Any opinions, findings, and conclusions or recommendations expressed here are those of the authors and do not necessarily reflect the views of NSF or NIER.

■

REFERENCES

(1) Ban-Weiss, G. A.; McLaughlin, J. P.; Harley, R. A. Carbonyl and nitrogen dioxide emissions from gasoline- and diesel-powered motor vehicles. Environ. Sci. Technol. 2008, 42, 3944−3950. (2) Bishop, G. A.; Stedman, D. H. A decade of on-road emissions measurements. Environ. Sci. Technol. 2008, 42, 1651−1656. (3) Bishop, G. A.; McLaren, S. E.; Stedman, D. H. Method comparisons of vehicle emissions measurements in the Fort McHenry and Tuscarora Mountain tunnels. Atmos. Environ. 1996, 30, 2307− 2316. (4) Singer, B. C.; Harley, R. A. A fuel-based motor vehicle emission inventory. J. Air Waste Manage. Assoc. 1996, 46, 581−593. (5) Schifter, I.; Diaz, L.; Mugica, V.; Lopez-Salinas, E. Fuel-based motor vehicle emission inventory for the metropolitan area of Mexico City. Atmos. Environ. 2005, 39, 931−940. (6) Guo, H; Zhang, Q; Shi, Y; Wang, D. On-road remote sensing measurements and fuel-based motor vehicle emission inventory in Hangzhou, China. Atmos. Environ. 2007, 41, 3095−3107. (7) Cadle, S. H.; Stephens, R. D. Remote sensing of vehicle exhaust emissions. Environ. Sci. Technol. 1994, 28, 258−264. (8) Ropkins, K.; Beebe, J.; Li, H.; Daham, B.; Tate, J.; Bell, M.; Andrews, G. Real-world vehicle exhaust emissions monitoring: Review and critical discussion. Crit. Rev. Environ. Sci. Technol. 2009, 39, 79− 152. (9) Stedman, D. H.; Bishop, G. A.; Beaton, S. P.; Peterson, J. E.; Guenther, P. L.; McVey, I. F.; Zhang, Y. On-road remote sensing of CO and HC emissions in California; Prepared by University of Denver, Denver, CO for California Air Resource Board, Contract No. A032093, 1994. (10) Ashbaugh, L. L.; Lawson, D. R.; Bishop, G. A.; Guenther, P. L.; Stedman, D. H.; Stephens, R. D.; Groblicki, P. J.; Johnson, B. J.; Huang, S. C. On-road remote sensing of carbon monoxide and hydrocarbon emissions during several vehicle operating conditions; Presented at A&WMA International Specialty Conference on PM10 Standards and Non-traditional Source Controls, Phoenix, 1992. (11) McClintock, P. Remote sensing measurements of real world high exhaust emitters- Interim Report; Prepared by Remote Sensing Technologies Inc., Tucson, AZ for Coordinating Research Council, Atlanta, GA, 1999. (12) Pokharel, S. J.; Bishop, G. A.; Stedman, D. H. On-road remote sensing of automobile emissions in the Phoenix area: Year 3; Prepared by University of Denver, Denver, CO, for Coordinating Research Council, Alpharetta, GA, 2002. (13) Chan, T. L.; Ning, Z.; Leung, C. W.; Cheung, C. S.; Hung, W. T.; Dong, G. On-road remote sensing of petrol vehicle emissions measurement and emission factors estimation in Hong Kong. Atmos. Environ. 2004, 38, 2055−2066. (14) Kuhns, H. D.; Mazzoleni, C.; Moosmuller, H.; Nikolic, D.; Keislar, R. E.; Barber, P. W.; Li, Z.; Etyemezian, V.; Watson, J. G. Remote sensing of PM, NO, CO and HC emission factors for on-road 6873

dx.doi.org/10.1021/es204451z | Environ. Sci. Technol. 2012, 46, 6867−6873