Deriving fuel-based emission factor thresholds to interpret heavy-duty vehicle roadside plume measurements

ABSTRACT

Remote sensing devices have been used for decades to measure gaseous emissions from individual vehicles at the roadside. Systems have also been developed that entrain diluted exhaust and can also measure particulate matter (PM) emissions. In 2015, the California Air Resources Board (CARB) reported that 8% of in-field diesel particulate filters (DPF) on heavy-duty (HD) vehicles were malfunctioning and emitted about 70% of total diesel PM emissions from the DPF-equipped fleet. A new high-emitter problem in the heavy-duty vehicle fleet had emerged. Roadside exhaust plume measurements reflect a snapshot of real-world operation, typically lasting several seconds. In order to relate roadside plume measurements to laboratory emission tests, we analyzed carbon dioxide (CO₂), oxides of nitrogen (NO_X), and PM emissions collected from four HD vehicles during several driving cycles on a chassis dynamometer. We examined the fuel-based emission factors corresponding to possible exceedances of emission standards as a function of vehicle power. Our analysis suggests that a typical HD vehicle will exceed the model year (MY) 2010 emission standards (of 0.2 g NO_X/bhp-hr and 0.01 g PM/bhp-hr) by three times when fuel-based emission factors are 9.3 g NO_X/kg fuel and 0.11 g PM/kg using the roadside plume measurement approach. Reported limits correspond to 99% confidence levels, which were calculated using the detection uncertainty of emissions analyzers, accuracy of vehicle power calculations, and actual emissions variability of fixed operational parameters. The PM threshold was determined for acceleration events between 0.47 and 1.4 mph/sec only, and the NO_X threshold was derived from measurements where after-treatment temperature was above 200°C. Anticipating a growing interest in real-world driving emissions, widespread implementation of roadside exhaust plume measurements as a compliment to in-use vehicle programs may benefit from expanding this analysis to a larger sample of in-use HD vehicles.

Implications: Regulatory agencies, civil society, and the public at large have a growing interest in vehicle emission compliance in the real world. Leveraging roadside plume measurements to identify vehicles with malfunctioning emission control systems is emerging as a viable new and useful method to assess in-use performance. This work proposes fuel-based emission factor thresholds for PM and NOx that signify exceedances of emission standards on a work-specific basis by analyzing real-time emissions in the laboratory. These thresholds could be used to prescreen vehicles before roadside enforcement inspection or other inquiry, enhance and further develop emission inventories, and potentially develop new requirements for heavy-duty inspection and maintenance (I/M) programs, including but not limited to identifying vehicles for further testing.

Introduction

Extensive human epidemiological and animal exposure studies have demonstrated acute and chronic health effects from exposure to both ozone (Lippmann 1989; Touloumi et al. 1997) and fine particulate matter (PM) (Brook et al. 2010; CARB 1998; Lloyd and Cackette 2001; Pope and Dockery 2006). The California Air Resources Board (CARB) and the regional air quality management districts throughout California are required to demonstrate compliance with National Ambient Air Quality Standards (NAAQS) for the federal criteria pollutants, including ozone and PM, in state implementation plans (SIP). Because of the significant contribution of the transportation sector to air and climate pollution, air quality attainment plans necessarily include control of emissions from on-road heavy-duty vehicles (HDVs)—those having gross vehicle weight ratings (GVWR) greater than 14,000 lb. These vehicles currently represent 50% of statewide diesel PM, 45% of nitrogen oxides, 6% of greenhouse gas emissions, and are the mobile source category, after passenger cars, with greatest fuel consumption in the United States (2011a; CARB 2015c). New heavy-duty (HD) on-road engines manufactured to the model year (MY) 2010 emission standards have 97% lower NO_X and PM emissions than engines manufactured to the MY 1987 emission standards; these limits are currently 0.2 g NO_X/bhp-hr and 0.01 g PM/bhp-hr when tested over the Federal Test Procedure (FTP) and Supplemental Emission Test (SET) cycles. The heavy-duty fleet has a long average lifetime; only 10% of heavy-duty engines are rebuilt after operating 800,000 miles (CARB 2016b). In order to accelerate the transition to a cleaner fleet, California has uniquely adopted a number of in-use diesel fleet rules. The most comprehensive on-road heavy-duty in-use rule is the Truck and Bus Regulation in 2008 (CARB 2011), which requires nearly all HD vehicles operating in California to have MY 2010 and newer engines by 2023. As a result of this regulation and natural vehicle turnover, the majority of the on-road fleet soon will be equipped with diesel particulate filters (DPF) and selective catalytic reduction (SCR) after-treatment systems to control PM and NO_X emissions, respectively.

In May 2015, CARB released a study identifying a new high-PM emitter problem in the HD fleet. A small fraction of HD vehicles—8%—in the fleet have damaged or malfunctioning DPFs, and these vehicles contribute disproportionately (approximately 70%) to the fleet-wide PM emissions (CARB 2015b). The study also showed that some engines and emission control systems are not sufficiently durable to effectively operate and control emissions over the lifetime of the vehicle. The findings were further evaluated and emphasized more broadly in CARB’s Mobile Source Strategy released in May 2016 (CARB 2016a), which called for more stringent emission certification standards, more comprehensive in-use requirements, and expanded inspection and maintenance (I/M) requirements for HDVs that operate in California.

The performance of DPF and SCR systems has been evaluated using fuel-based emission factors measured by entraining diluted exhaust from individual HDV plumes at the roadside. These include the plume-based sampling method employed by Harley and colleagues (Preble et al. 2015), and the On-Road Heavy-Duty Measurement System (OHMS, or “tent”) study developed by Stedman and colleagues (Bishop et al. 2015; Bishop, Schuchmann, and Stedman 2013; Haugen and Bishop 2017). These measurement approaches have been leveraged by CARB for various assessments of rule benefits, such as evaluating the emissions benefits of the Drayage Truck Regulation at the Port of Oakland and Port of Los Angeles, as well as the Truck and Bus Regulation at other roadside locations through the state. Unlike snap acceleration opacity tests that can be performed by pulling over a vehicle at the roadside according to the SAE J1667 procedure (SAE 1996), roadside plume-based measurements do not require any interaction with the vehicle operator and could theoretically be performed autonomously without dispatching field staff.

Initial evaluations of roadside plume measurements have shown that vehicles with improperly functioning DPFs can be distinguished from vehicles with properly functioning DPFs under a wide range of duty cycles. In addition to PM, roadside plume capture measurements can also be adopted to measure gaseous emissions, including nitric oxide (NO), nitrogen dioxide (NO₂), or total NO_X. Considerations that must be applied to fuel-based NO_X emission factors, especially for vehicles equipped with SCR, include after-treatment temperature and duty cycle during plume capture (Bishop, Schuchmann, and Stedman 2013; Quiros et al. 2016). In past decades, light-duty hydrocarbon and carbon monoxide emissions have been compared to laboratory test cycles (CARB 1994; EPA 1991); however, no such work has been released presenting these relationships specifically for HDVs. This work is new and unique for HD vehicles; in contrast to light-duty vehicles, evaluation of emissions from HD vehicles must consider a larger range of vehicle weights, and emissions assessment must be performed relative to engine rather than vehicle certification standards.

This paper uses emissions data collected from four HD vehicles over several chassis dynamometer cycles to address three main objectives regarding the applicability, validity, or representativeness of roadside plume measurements for quantifying PM and NO_X emissions. First, because fuel-based emissions use carbon dioxide (CO₂) emissions as a proxy for engine work, we define CO₂ emissions per unit work as a function of engine power that can be used to correlate engine fuel consumption and work performed. Second, laboratory chassis dynamometer data are used to compare real-time PM and NO_X emissions relative to composite emission factors over standard test cycles. Third, this paper discusses considerations for fuel-based NO_X and PM thresholds that may be used to identify high-emitting vehicles on the road based on a test fleet of four engines in the Class 7–8 range. The uncertainties of the overall approach, including roadside plume measurement accuracy, uncertainties in approximating the vehicle power, and variability in PM and NO_X emissions for a given operation, are assessed when proposing fuel-based emission factor thresholds that can be applied to field measurements.

Methods

Vehicle selection

Table 1 lists three MY 2010 and newer vehicles in our analysis, which were each equipped with a diesel oxidation catalyst (DOC), DPF, and SCR system and certified to the 0.01 g/bhp-hr PM standard and at least a 0.3 g/bhp-hr family emission limit (FEL) for NO_X. Also listed in Table 1 is a MY 1998 vehicle (Vehicle D) that was not equipped with any after-treatment system, and is used to assess trends in real-time engine-out PM. Among the four vehicles, two major original equipment manufacturers (OEMs) were represented: One was a medium-HD engine (MHD) with a peak power rating of 325 hp designed for operation in a vehicle with a gross vehicle weight rating (GVWR) between 19,001 and 33,000 lb, and three were heavy-HD (HHD) engines with peak power ratings between 400 and 505 hp, intended for operation in a vehicle with a GVWR greater than 33,000 lb and less than 80,000 lb. The 2014 release of California’s on-road emissions model (EMFAC2014) defines the average Urban Dynamometer Driving Schedule (UDDS) emission rate for vehicles with MY 2014 engines at 1.16 and 1.97 g/mile NO_X (0.31 and 0.52 g/bhp-hr assuming 3.8 bhp-hr/mile for a UDDS) for MHD and HHD engines, respectively, before applying any field-aging or emission control deterioration factors. Table 1 also provides a description for how vehicles A through D were analyzed in this paper. For example, vehicles A and C exhibited NO_X emissions to within a factor of 1½ of the fleet-average reported by the EMFAC model and were labeled “typical” NO_X emissions, whereas vehicle B had NO_X emissions more than three times the EMFAC prediction and was labeled “higher” NO_X emissions. Vehicle D was not equipped with a DPF, and only its PM emissions were analyzed to better understand trends of “engine-out” PM emissions that, for example, pass through a DPF when leaking. Vehicle D was not analyzed for NO_X emissions because it was equipped with a MY 1998 engine that was certified to a standard 20 times greater (4.0 g NO_X/bhp-hr) than the MY 2010 standard, and it is likely the engine calibration and control strategies would not represent NO_X trends of modern HD vehicles operating in California. It should be noted when vehicle D was retrofitted with a DOC and DPF, the PM emissions during UDDS cycle were below 0.01 g/bhp-hr (Herner et al. 2009).

Table 1. List of selected vehicles to represent a range of NO_X and PM emissions. Emissions are reported for the heavy-duty urban dynamometer driving schedule (UDDS).

Vehicle	Class	Engine and after-treatment	Peak power (bhp)	UDDS PM (g/mile)	UDDS NOX (g/mile)	Description**
A	MHD	MY 2010 (DOC + DPF + SCR)*	325	0.001	1.56	“Typical” NOX
B	HHD	MY 2011 (DOC + DPF + SCR)*	485	0.003	7.51	“Higher” NOX
C	HHD	MY 2014 (DOC + DPF + SCR)*	505	0.002	1.58	“Typical” NOX
D†	HHD	MY 1998 (No after-treatment)	400†	0.670	21.1†	“Engine-out” PM

Notes.*DOC = diesel oxidation catalyst, DPF = diesel particulate filter, SCR = selective catalytic reduction.

†Only evaluated for PM emissions, NO_X emissions provided for reference(Herner et al. 2009). Peak power is rated for its engine family.

**Description pertains to how the data were used in analysis, not the vehicle or engine itself.

Laboratory testing and emissions measurement

Vehicles A, B, and C were tested at CARB’s Depot Park Facility located in Sacramento, CA. The laboratory is equipped with a single-roll 72-inch, 600-hp chassis dynamometer manufactured by Burke E. Porter Machinery Company (model 4700, Grand Rapids, MI) designed to simulate road loads on vehicles with GVWR between 10,000 and 80,000 lb. The MHD and HHD vehicles tested in the laboratory were tested at 80% of their respective GVWRs, which were 26,500 lb for vehicle A and 65,000 lb for vehicles B and C. Drive axle inertia was set to 1.5%. Target road loads were derived following guidelines in SAE J1263 and J2263. Vehicle A was a box truck configuration, and road load targets were 160.59 lbsf, −0.883 lbsf/(mph), and 0.1781 lbsf/(mph); vehicles B and C were dual-axle sleeper cab tractors; coast-downs were performed with a 50-foot trailer with low rolling resistance tires, and road load targets were 268.08 lbsf, 6.605 lbsf/(mph) 0.1001 lbsf/(mph).

Emissions were measured and are reported based on raw exhaust measurements. Exhaust flow was measured with a 5-inch high-speed exhaust flowmeter (EFM-HS) manufactured by Sensors Incorporated (Saline, MI); gaseous emissions of carbon monoxide (CO), carbon dioxide (CO₂), nitric oxide (NO), total nitrogen oxides (NO_X), and total hydrocarbons (THC) were measured by an AMA i60 bench analyzer (AVL, Graz, Austria); and gravimetric PM measurements were performed using a BG-3 partial flow sampling system manufactured by Sierra Instruments (Monterey, CA). Gaseous sampling lines were heated to 191°C. PM measurements were made with an inlet orifice probe downstream of a cyclone separator to remove coarse PM larger than ~2.5 µm. The BG-3 uses a proportional sampling system, with constant sampling fractions ranging between 0.03 and 0.06% depending on the cycle, to achieve an approximate minimum dilution ratio between 5 and 7 times according to CFR 1065 (CFR 2011). Analyzer signals were postprocessed according to CFR guidelines for performing drift correction (CFR 1065.672), intake-air humidity NO_X correction (CFR 1065.670), and performing dry-to-wet conversion of analyzers operating downstream of a chiller (CFR 1065.659). On-board diagnostic (OBD) broadcast data were recorded, and additional parameters were requested over the SAE J1939 protocol and recorded with a Dearborn Protocol Adapter (DPA5, DG Tech, Farmington Hills, MI) and an in-house software package. Acquired parameters used in this analysis include suspect parameter number (SPN) 190 for engine speed (rpm), SPN 513 for actual engine torque, SPN 514 for frictional engine torque, SPN 544 for engine reference torque, and SPN 4360 for SCR intake temperature (°C). Brake-specific power (brake-horsepower, bhp) was calculated as broadcast from the engine, and for this analysis is assumed equivalent to dynamometer roll-set power and road power.

Six chassis dynamometer driving cycles with average speeds ranging between 1.8 and 62 mph were used to collect a wide range of emissions behavior, including extended low-speed acceleration to extended higher speed and load operation. Table 2 lists these cycles and descriptive parameters, and Figure 1 presents the time series of the various cycles. Real-time emissions analysis considered results from all cycles; however, comparisons of real-time data were made to the average emissions measured over the UDDS cycle (average speed 18.9 mph) because it was the basis for the development of the transient engine dynamometer FTP that is used for new engine certification and in-use emissions compliance (Smith 1978). Real-time gaseous emissions were time aligned to within 1 sec of exhaust flow measurements for each gaseous species separately, depending on the response time measured by the AVL AMA i60 system. Cycles were run in triplicate, and for vehicles A, B, and C, approximately 11 hr of real-time data was generated.

Table 2. List of chassis dynamometer driving cycles with duration, distance, and speed parameters. Each cycle was run at least once and up to three times per vehicle. The time series of these plots are shown in Figure 2.

Cycle name	Duration (sec)	Distance (mi)	Maximum speed (mph)	Average speed (mph)
Creep × 8	2024	0.994	8.20	1.77
Near-dock drayage	3047	5.59	40.5	6.61
Local drayage	3362	8.70	46.4	9.31
Urban dynamometer driving Schedule (UDDS) × 2	2120	11.1	58.0	18.9
Cruise	2083	13.6	52.4	39.9
High-speed cruise	1385	23.1	72.4	62.0

Figure 1. Chassis dynamometer test cycles used to measure emissions under controlled laboratory conditions. Circles in panel (a) indicate the locations of two real-time analyses presented and discussed with Figure 9.

Figure 2. Theoretical flow schematic of roadside plume-based detection systems for identifying vehicles with emissions exceeding relevant certification or in-use emissions standards.

Vehicle D was tested at ARB’s Metropolitan Transit Authority (MTA) laboratory as part of a previously published test program discussed elsewhere (Herner et al. 2009). Testing procedures were similar to those for vehicles A, B, and C; however, testing was conducted at a simulated inertial weight of 53,230 lb, and the MTA laboratory is equipped with a constant volume sampler (CVS) operating at ~2600 cubic feet per minute (CFM), rather than a raw exhaust system. Vehicle D was equipped with a MY 1998 engine, certified to a 0.1 g/bhp-hr PM standard, and real-time PM emissions were evaluated during a UDDS cycle using size distribution data measured from a TSI engine exhaust particle sizer (EEPS, 5.6–560 nm, TSI, Incorporated, Shoreview, MN), and the integrated particle size distribution (IPSD) method (Liu et al. 2012; 2009; Quiros et al. 2015b). An effective density function was applied based on results and methodologies described in Quiros et al. (2015a) for a diesel vehicle operating under simulated-transient operation (Quiros et al. 2015a). The IPSD method reported 63% (0.425 g/mile) of the filter-based gravimetric mass value (0.670 g/mile) for the UDDS evaluated; however, because this analysis is only evaluating the ratio of real-time to composite UDDS emissions, a linear but not necessarily one-to-one relationship is required between the methods. Here, real-time PM mass is reported to characterize PM formation of an engine as a function of the duty cycle.

Calculations

Table 3 presents percentile statistics of vehicle speed, acceleration, and power for the chassis dynamometer UDDS. The speed versus time traces for chassis dynamometer cycles do not specify vehicle power. Instead, power is determined by the road load coefficients described in the preceding, mass of the vehicle (80% of GVWR, the assumption used in California’s EMFAC model [CARB 2015a]), the driving cycle, and other dynamometer settings.

Table 3. List of Newtonian vehicle operation conditions for the chassis dynamometer UDDS for typical Class 7 and Class 8 trucks. Vehicle speeds and accelerations are identical for all weight classifications because the same speeds were applied. Vehicle power was calculated using road load force and inertial forces. Percentiles are calculated from positive torque values only converted back to a scale from 0 to 100% by multiplying by the fraction of nonzero torque values (46%). Weights were obtained from the curb weight of the Class 7 vehicle with a 23-foot flatbed trailer attached (12,000 lb) and a Class 8 vehicle with a 53-foot curtain trailer attached (33,000 lb).

Truck classification	Vehicle power (bhp)			Vehicle speed (mph)			Vehicle acceleration (mph-sec)
Percentile	25th	50th	75th	25th	50th	75th	25th	50th	75th
Class 7 trucks
Minimal load (12,000 lb)	9.12	24.6	36.8	6.8	14.5	24.9	0.47	0.9	1.4
Typical load (26,500 lb)	16.3	38.3	52.8	6.8	14.5	24.9	0.47	0.9	1.4
Maximum load (33,000 lb)	19.7	43.9	64.4	6.8	14.5	24.9	0.47	0.9	1.4
Class 8 trucks
Minimal load (33,000 lb)	23.9	57.9	76.2	6.8	14.5	24.9	0.47	0.9	1.4
Typical load (65,000 lb)	35.2	77.2	121	6.8	14.5	24.9	0.47	0.9	1.4
Maximum load (80,000 lb)	40.9	82.1	143	6.8	14.5	24.9	0.47	0.9	1.4

Fuel-based emissions factors were calculated using real-time concentrations of the pollutant, here either nitrogen oxides (NO_X) or PM determined by the IPSD method, CO₂ as measured by laboratory analyzers, fuel carbon content of 0.869 for diesel fuel based on parameters listed in CFR 1065.655 (CFR 2011), and the molecular weights of CO₂ (44.01 g/mol) and C (12.01 g/mol). This analysis derives and presents initial values for performing two additional calculations or modifiers that can be applied to fuel-based emission factors for improved characterization relative to HD engine operation standards. These two parameters, brake-specific fuel consumption (BSFC) and duty cycle factor, are presented in eq 1 and defined in the following:

(1)

Here, X is either NO_X or PM, but can also be extended to other pollutants or chemical species. The term “fuel based EF” would be reported directly from the calculated value based on analyzer response. The term “brake-specific fuel consumption (BSFC)” defines the relationship between fuel consumption and brake-specific work, which we present as a function of average engine power in the first objective of this work. BSFC is expected to decrease as greenhouse gas (GHG) engine standards become more stringent. For instance, the GHG engine standard decreases by 9% from MY 2014 engines (475 g CO₂/bhp-hr) and MY 2027 engines (432 g CO₂/bhp-hr) (EPA 2011b; EPA 2015). Therefore, the values of BSFC derived in this study may need to be updated over the next decade as engine efficiencies and vehicle fuel economy improves. The “duty cycle factor” is presented as a function of power as part of the second objective of this work to better relate real-time NO_X and PM to cycle-average emission factors.

The uncertainty of combining the three terms in eq 1 is evaluated as the third objective of this paper, and the sources of uncertainties are grouped into four categories listed in Table 4. For each category, the coefficient of variation (CV) is calculated from the reported uncertainty (± 1 standard deviation, SD) divided by the expected value or mean. The uncertainty was propagated among multiple sources assuming that data followed a normal distribution. Further, the 5th and 95th percentiles of actual data were assumed equivalent to ±2 SD.

Table 4. Sources of uncertainty using advanced detection systems measuring diluted exhaust plumes at the roadside.

Source of uncertainty	UDDS median value*	CV (%)	Notes/reference
Group 1—Analyzer uncertainties (eq 2)
CO2	500 ppm	10.8%	Assumes n = ± 3% of measured range, B = 400 ppm, and P = 600 ppm
NOX	500 ppb	0.147%	Assumes n = ± 1 ppb, B = 40 ppb, and P = 1000 ppb
PM/BC	9,000 ng/m^3	3.72%	Assumes n = ± 500 ng/m^3 at 1 Hz, B = 1000 ng/m^3, and P = 20,000 ng/m^3
Group 2—Newtonian parameters (eqs 3–5)
Speed, V	10 mph	0.363%	Eq 3, based on speed-based uncertainty function of Doppler radar sensor
Acceleration, A	0.5 mph-sec	10.5%	Eq 4, based on speed-based uncertainty function of Doppler radar sensor
Weight, W	65,000 lb	4.47%	Eq 5, based on uncertainty of commercially available weigh-in-motion sensors, assuming 10% uncertainty per axle, and 5 axles per truck)
Group 3—Vehicle power
Road load, L	77 hp	3%	Assumed uncertainty from SAE J1263 method for determining road load for each coefficient a, b, and c.
Road grade, θ	0%	0.00%	Assumed 0% uncertainty in determining precise value at field site
Power zero grade	77 hp	13.0%	Propagated from each term in eq 6, excluding road grade
Power nonzero grade	77 hp**	13.4%	Propagated from each term in eq 6, assumes 10% road grade
Group 4—Uncertainty in vehicle emissions
CO2/work	77 hp	23.5%*	Figure 3
PM/fuel	77 hp	19.8%	Figure 6b
NOX/fuel	77 hp	29.7%	Figure 8b
Total uncertainty for PM/BC measurement		35.2%
Total uncertainty for NOX measurement		41.7%

Notes.* UDDS median values were determined from measurements at 77 hp, and approximations of expected values from existing systems deployed in the field today for the purpose of estimating real-world uncertainty and calculating CV. For Group 1, P and B are peak and baseline concentrations (ppm or mass per volume) and n is the reported uncertainty fraction of the analyzer at measured concentrations (eq 2).

**Actual power value would be greater for nonzero grade. This term remains the same to show that introducing another instance of CV_W results in an increased propagated CV for vehicle power when road grade is nonzero.

Figure 3. Median brake-specific CO₂ emissions as a function of vehicle power. A power-fit model was applied to the average of the median Class 7 and Class 8 truck data for broadcast power ranges between 5 and 75 bhp, and separate power-fit models were applied for broadcast power between 75 and 325 hp. Modeled equations are annotated within the plot areas.

Figure 4. Histogram of engine or vehicle power during (a) a chassis UDDS cycle at 65,000 lb test weight and (b) an engine FTP cycle using specifications for vehicle C. Darker bars indicate the 25th and 75th percentiles for adjusted positive values. Both distributions have the majority of the targeted range overlapping with power values.

Figure 5. Real-time PM emissions during a selected UDDS from vehicle D that was not equipped with a DPF.

Figure 6. Real-time emission rates as a function of vehicle power for vehicle D. Bars are bin-average values for each 20 hp. Observations where vehicle power was less than 10 hp were are not shown. A 5-sec running average was applied when binning data to correspond with roadside plume measurement over a similar time basis (from a few to several seconds).

Figure 7. Real-time PM emissions during a selected UDDS from vehicle C equipped with a properly function DPF.

Figure 8. Real-time NO_X emission rates as a function of vehicle power. Bars are bin-average values for each 20 hp. Observations where vehicle power was less than 10 hp were are not shown. A 5-sec running average was applied when binning data to correspond with roadside plume measurement over a similar time basis (from a few to several seconds).

Figure 9. Fuel-based NO_X emissions during acceleration events from (a) 0 to 15 mph and (b) 20 to 35 mph during a UDDS cycle. Measured NO_X emission rates are reported using a 5-sec running average, and the corresponding broadcast brake-specific power was used in the equation defined in the panel of Figure 8b to generate transformed NO_X emission rates. Data were obtained from a randomly selected test run from vehicle B.

Results and discussion

Concept overview

Figure 2 illustrates the theoretical flow schematic of data inputs and processing that can be used to assess real-time work-specific vehicle emissions in the field. Although this work does not describe the physical design or software requirements in order to deploy a roadside plume sampling system, it describes the scope and potential inputs to the processing and interpreting of emissions results for use as part of a number of programs or purposes.

The first step is to determine vehicle power, which can be calculated using vehicle weight, speed, and acceleration, and using typical road load equations for a HD vehicle. The calculation approach is discussed subsequently in this paper (see eq 6). Speed and acceleration can be measured with a variety of sensors, such as a Doppler RADAR, light detection and ranging (LIDAR), and ultrasonic sensors. Vehicle weight can be measured by deploying a weigh-in-motion (WIM) sensor, or vehicle power can be bounded by assuming a weight based on typical weights of minimally and maximally loaded trucks.

Next, fuel-based emission factors (e.g., g PM/kg fuel) should be calculated (possibly in real time) using emissions analyzer responses above baseline levels. Possible approaches for fuel-based emission factor calculation include integrating the areas of the measured emissions peaks above baseline levels, or by using a best-fit line between CO₂ and the pollutant(s) of interest. Fuel-based emission factors are then converted into work-based emission factors according to the BSFC, which is defined using the chassis dynamometer data in Figure 3.

Finally, emissions data are stored alongside vehicle identifiable information, which may allow identification of the vehicle identification number (VIN), and engine family to assess the measured emissions against the expected emission levels extrapolated from certification standards of the engine.

Objective 1—Define BSFC as a function of vehicle power

The BSFC, or CO₂ emissions per unit work, is used to convert fuel-based emission factors to work-based emission factors, which is the basis upon which emission standards are established. Previous groups have reasonably assumed 0.15 kg fuel/bhp-hr (Burgard et al. 2006), which equates to about 38% thermal efficiency assuming 45.6 MJ/kg diesel fuel, and is equivalent to about 480 g CO₂/bhp-hr. Using vehicle power derived from engine broadcast data, for real-time periods when speed was nonzero and acceleration was greater than 0.1 mph/sec, we calculated work-specific CO₂ emission factors as a function of average power bin.

Figure 3 presents bin-average data for Class 7 (MHD) and Class 8 (HHD) vehicles separately as piecewise functions, first for power between 5 and 75 hp (a), and then for power between 75 and 325 hp (b). Figure 3a shows that the CO₂ per unit work for transient operation up to 75 hp is similar for both engine groups, and therefore CO₂ emissions are presented using a single power-fit equation. Additionally, it should be noted that lower thermal efficiencies (down to 15%) and higher fuel-to-work ratios (up to 0.4 kg fuel/bhp-hr) are observed within this power range. This range of BSFC is substantially higher at lower power than the previously applied value of 0.15 kg fuel/bhp-hr. Figure 3b also shows the BSFC and CO₂ emissions for MHD and HHD spread somewhat with increasing power above 75 bhp up to about a 10%  difference at the highest power bins, and therefore the emission rates are quantified and annotated separately.

No error bars are presented in Figure 3, and we show later in this work that the CV of BSFC (24.3%) is larger than the difference between the MHD and HHD values (~10%). Larger scatter in real-time data within groups (e.g., MHD data points only) than difference in means of the groups is not surprising because each power bin represents a combination of transient operation with various speeds and acceleration rates. Acknowledging some data from the two groups overlap, we still report the MHD and HHD lines separately because the difference still signifies a difference in BSFC for engines rated to different peak horsepower. Specifically, the MHD vehicle emits more CO₂ than the HHD vehicle as power approaches the maximum rated horsepower for the smaller MHD engine.

Objective 2—Determine the ranges of vehicle speed, acceleration, and power that instantaneously best represent composite emissions measured over the UDDS

Engine and vehicle power during the FTP and UDDS

Figure 4 presents a histogram of the power distributions for vehicles B and C. The engine FTP schedule is a function of minimum and maximum speed and torque, and values were calculated using the parameters of the engine. Power is below 5 hp for more than 50% of the FTP and UDDS, the 25th to 75th percentiles (the interquartile range, or IQR, shaded area) of positive-power observations are overlapping (35 to 120 hp for UDDS, 35 to 140 hp for FTP), and the positive-power median and average power values are within 10% or less (77 and 70 hp respectively for the UDDS, and 85 and 84 hp respectively for the FTP). These similarities are expected because the cycles were developed from the same data set(Smith 1978), and confirm that the UDDS can be used to roughly represent the activity behavior if an engine were removed from a truck and tested on an engine dynamometer FTP cycle. The interquartile ranges (IQR) for Newtonian parameters (e.g., vehicle speed, acceleration, and power) were chosen arbitrarily to capture a middle range of typical parameters of the chassis-based UDDS.

The objective of analyzing chassis dynamometer data is to determine how a brief snapshot of emissions, which may only capture a few seconds of operation, can adequately represent the emissions that would be assessed over a full-length test cycle such as the FTP or UDDS. Of course, the long history of comparison of relative standard deviation (RSD) measurements to the performance of the on-road passenger car fleet suggests that the preceding objective for HDVs is entirely plausible. For this reason, the real-time fuel-based emission factors measured on the dynamometer are compared as a ratio to the composite emission factors over the full UDDS. Using this approach, real-time emission ratios less than 1 imply that the snapshot underrepresents, that is, predicts lower emissions than would be measured over the full duty cycle. Conversely, real-time emission ratios greater than one imply that the snapshot may overrepresent, that is, predict higher emissions than would be measured over the full duty cycle. The inverse of these values can be considered the “duty cycle factor” as outlined in eq 1.

Fuel-based PM emissions with respect to vehicle power

Analysis of real-time PM emissions was conducted using emissions measured from vehicle D and the IPSD method; vehicles A, B, and C were equipped with properly functioning DPFs and had measured PM emissions at emission levels at least 10 times lower than the stringent 0.01 g/bhp-hr standard during all test cycles. For DPF-equipped vehicles with compromised or physically damaged filters, tailpipe PM emissions would more closely resemble engine-out PM, albeit the actual emission rate may vary depending on the severity of the failure. We further acknowledge that engine-out PM emissions of an HD engine originally equipped with a DPF have not been widely reported, and when combined with potential upstream engine failures, the filtration efficiency of a DPF remains largely unknown. Figure 5 illustrates the real-time variability in PM emissions rates during a UDDS for vehicle D, which typically either fall below or above the average emission rate of 0.63 g/mile or 1.029 g PM/kg fuel (dashed line). Accordingly, a brief snapshot of on-road operation may more likely under- or overrepresent cycle-average behavior, and why a duty cycle factor (see eq 1) may be helpful for incorporating a wider range of real-world measurements to determine whether an emission control failure has occurred. However, it is possible that specific instances of real-world operation or portions of the UDDS are quantitatively similar to the composite UDDS emission rate (such as when the solid red line and dashed red line intersect in Figure 5), and the initial fuel-based emission factor alone would be sufficient to determine whether a vehicle exceeded a given threshold. Additionally, Figure 5 shows how PM is consistently created and emitted during acceleration events, which generally provide for good opportunities for measurement and assessment to cycle-average emission rates.

Figure 6 presents PM emissions from vehicle D as a ratio of real-time (1 Hz) to composite UDDS emissions in bins of the nearest 20 hp. When considering all of the real-time data points for the entire UDDS as shown in Figure 6a, average PM emission rates are reasonably predicted by a polynomial fit equation annotated within the figure panel. Although power is narrowly defined within each bin, vehicle acceleration and engine conditions varied greatly, and therefore PM emissions also vary widely. For example, the bin centered at 80 hp had an average emission ratio of ~5 (where the real-time emission rate was 5 times the value of the composite UDDS emission rate), but the range of real-time emission factors ranged between ~0.5 and ~8 times the UDDS ratio for 5th and 95th percentiles of the data, respectively.

When considering observations only with acceleration rates within the IQR of the UDDS (0.47 to 1.4 mph-sec, Table 3) as shown in Figure 6b, the variability about the bin averages is more narrow while the bin averages are approximately the same. In the example of the 80-hp bin, the average emission ratio remained at ~5, and the 5th to 95th percentile ranges reduced to ~3 to ~8 times the UDDS emission rate. Although more narrow ranges of acceleration may provide less variability in the real-time emission ratios than presented in Figure 6b, the incidental speed and acceleration rates from on-road vehicles would likely result in a lower overall capture rate.

The majority of HD vehicles operating today are still diesel fueled, and by 2023 virtually all HD vehicles operating should be equipped with a DPF. Deploying roadside plume measurement systems could, at an initial implementation level, serve as a screening compliance tool for trucks either (1) equipped with DPFs that are improperly functioning or (2) not equipped with DPFs at all, in potential violation of in-use fleet regulations. When properly functioning, DPFs are at least 85% efficient in removing PM, but laboratory testing by various groups has consistently shown in-field effectiveness is typically 98% or more (CARB 2015b). MY 2007 and newer engines are originally equipped with DPFs, and are certified to emissions levels typically at least five times lower than the 0.01 g/bhp-hr standard. Accordingly, if roadside PM measurements report fuel-based emission factors equate anywhere near 0.01 g/bhp-hr, PM emissions are already at least five times higher than the engine’s certified condition, which suggests that the DPF is either damaged, tampered with, or otherwise not properly functioning. The emissions from vehicle D (not equipped with a DPF because it was certified to the MY 1998 standards) were approximately 0.17 g/bhp-hr on a brake-specific basis, which is 17 times greater than the MY 2007 standard of 0.01 g/bhp-hr, and would readily be flagged as a potential vehicle without a properly functioning DPF.

Conversely, PM emissions from vehicle C (equipped with a DPF) were 0.002 g/mile, which is approximately 20 times below the 0.01 g/bhp-hr standard when using a 3.8 bhp-hr/mile conversion. If a plume from this vehicle were captured at a low power bin, which has the greatest ratio of average UDDS emissions at the 95th percentile, Figure 6b shows the resulting emissions would be 7 times higher than the UDDS emission rate, but still about 50% below the MY 2007 standard of 0.01 g/bhp-hr standard. Figure 7 shows real-time suspended PM emissions from vehicle C (MY 2014, DOC + DPF + SCR); data show this vehicle with a properly functioning DPFs never emitted PM at levels greater than the MY 2007 emission standard.

Fuel-based NO_X emissions with respect to vehicle power

Figure 8 presents the ratio of real-time NO_X emissions to composite UDDS rate for all operation at vehicle speeds less than 50 mph. Data collected at vehicle speeds greater than 50 mph consistently underreported the composite UDDS emission ratios by a factor of 5 or more during both lower and higher power outputs. Therefore, we analyzed emission rates captured at vehicle speeds below 50 mph to best estimate composite UDDS emission rates. Figure 8a presents all data, which include periods where SCR catalysts are both below and above the light-off activation temperature. For the purposes of this analysis, we assume that “light off” occurs when SCR temperatures are >200°C (Misra et al. 2013), and emissions ratios corresponding to these conditions where SCR light off has been achieved are shown in Figure 8b. The average real-time emission ratios not only are lower when restricting the analysis to SCR temperatures greater than 200°C, but also are bounded more tightly by the lines fit to the 5th and 95th percentiles. Using this approach, we can develop operational criteria that can be used to characterize a properly functioning SCR system, which can be used in the field to identify real-world driving events without sufficient NO_X control. NO_X emissions were not dependent upon acceleration rate, and the data suggest that unlike PM emissions, a power-based model alone, without an acceleration filter, should be sufficient for characterizing a NO_X emissions snapshot at the roadside.

Certification and in-use standards have not yet incentivized the development of thermal management strategies to achieve SCR light off during extended low-load operation. Thus, near-term remote sensing efforts could focus on identifying vehicles with damaged SCR systems. When deploying roadside plume measurement systems, operators should maximize opportunities for capturing snapshots of vehicles when SCR temperatures are greater than 200°C by installing instrumentation at locations downstream of road segments where vehicles typically operate at higher load (e.g., after exiting from a major highway, or climbing an extended road grade). Additionally, roadside plume measurement systems could be setup in conjunction with infrared-based temperature sensors to detect exhaust temperature in real time to verify that after-treatment temperature was likely above a given SCR light-off threshold. Measurement of exhaust plumes with temperatures above these thresholds could help assess after-treatment catalyst and DEF douser performance, rather than full after-treatment control strategy resilience. Over time when more advanced thermal management strategies are used by engine and emission control manufacturers, the fuel-based NO_X thresholds derived using this methodology should be applicable to this broader application of roadside sampling systems. For this future evaluation, instrumentation and sampling systems should be set up at locations that are not typically conducive to higher engine load.

Figure 9 presents the fuel-specific NO_X emissions rates for vehicle B for two separate acceleration events lasting approximately 30 sec from 0 to 15 mph, and from 20 to 35 mph during the second repetition of the 2xUDDS. It should be noted that the acceleration over this speed range required shifting over multiple gears because most Class 8 HDD trucks are equipped with a 10- or 13-speed transmission and are capable of wheel speeds between 0 and at least 60 mph; it is the load dropout and return during gear shifts that likely causes the observed real-time variability. A roadside plume capture system measuring emissions at any point during this acceleration would still report fuel-based emission factors to within three times the average UDDS emission rate (dashed line). The measured emissions from these two acceleration events illustrate how NO_X emissions can be reasonably assessed when accelerating from a stop, either from an intersection entering or leaving an inspection facility, assuming SCR light off has been achieved.

When applying the equation derived and quantified from Figure 8b, transformed NO_X emission rates (solid black line) more closely adhere to the composite UDDS rate, typically within a margin of two times the average UDDS emission rate. This example shows how removing the effect of vehicle power by considering that the impacts of the duty cycle can result in emission factors that resemble a composite cycle-average value.

Objective 3—Assess considerations for fuel-specific emission thresholds for identifying “high emitters” and suggesting threshold values based on controlled testing of actual in-use vehicles

Uncertainty in assessing brake-specific emissions from HDVs passing roadside emissions measurement systems is categorized into four groups as listed in Table 4. These include uncertainty in emissions analyzer measurements (Group 1), uncertainty in Newtonian parameters (Group 2), uncertainty in engine or vehicle power (Group 3), and uncertainties in the emission factors from the vehicles (Group 4). Table 4 lists UDDS median values for each source of uncertainty, or input parameter that is used to calculate brake-specific emission factors. We assigned common measured values to propagate numerical error and calculate total uncertainty. For instance, 77 hp is the UDDS median value for vehicle power, but does not necessarily represent the median power of a HD vehicle operating at a given roadside location. In the case of vehicle power and several other parameters in Table 4, equations indicated in the next several paragraphs can be used to calculate the uncertainty as a function of the input parameter. Uncertainty calculations this paper use represent values shown in Table 4 that likely correspond to typical values associated with a snapshot at 77 hp.

Group 1—Analyzer uncertainties

This group includes uncertainty with the measured concentrations of CO₂ and pollutants (e.g., NO_X and PM), which can be calculated for the plume integration technique where the measured peak concentration is subtracted from the baseline according to eq 2:

(2)

where P and B are peak and baseline concentrations [ppm or mass per volume] and n is the reported uncertainty fraction of the analyzer at measured concentrations. Peak and baseline concentrations were obtained from additional measurements conducted with a prototype automated roadside collection system. Peak concentrations are defined as the typical maximum measured concentration on the roadside when typical high-emitter emissions were detected. In the case of PM and NOx emissions, potential on-road concentrations vary by the dilution factor of exhaust and actual emission factor from the truck. Concentrations P could be much greater than assumed for the uncertainty analysis (e.g., 100,000 ng/m³ instead of the assumed 20,000 ng/m³ for black carbon [BC] or PM); however, typical or median values were identified and chosen to represent typical high NO_X- and high PM-emitting HD vehicles. The assumed uncertainties from commercially available analyzers for CO₂, NO_X, and BC are listed in Table 4.

Group 2—Newtonian parameters

Here we discuss the uncertainty parameters associated with the Newtonian parameters of the vehicle, including speed, acceleration, and weight. These inputs are used subsequently to approximate road load power and gravitational force acting upon the vehicle. Uncertainty associated with vehicle speed (V) measurement can be calculated according to eq 3:

(3)

where the equation can be simply reduced to n, the uncertainty fraction of speed V. Using currently available Doppler RADAR sensors such as the Delta DRS1000 manufactured by GMH Engineering (Orem, UT), n may be a speed-dependent parameter, and is typically less than 1% of speed V. The uncertainty associated with acceleration A, the first derivative of speed V, can be propagated from two speed measurements speed V, at time i and i + j according to eq 4:

(4)

where A is acceleration, which can be calculated by the difference between two speeds at times i and i + j. The uncertainty associated with vehicle weight can be calculated from propagating the uncertainty with a real-time “weigh-in-motion” sensor for each axle, or from the assumption of a single value of weight for a broad class of vehicles. Commercially available weigh-in-motion sensors, such as Strip Sensors manufactured by Intercomp Company (Medina, MN), report typical uncertainties of ±10% of measured weigh per axle z. Given the load distribution among axles, and that the total number of axles will vary by configuration, we propose calculating total vehicle uncertainty according to eq 5:

(5)

which assumes total vehicle weight, W, is evenly distributed among a total of z = 5 axles for each truck, and each measurement has uncertainty of n. Table 4 reports CV_Weight is 4.47% or 2,910 lb for the combined tractor and trailer. We acknowledge that road grade or other factors could result in an uneven distribution of weight on vehicle axles, and the assumption that uncertainty is evenly distributed among the axles may not always be applicable in the real world. Without weigh-in-motion measurements, the vehicle weight could be assumed to be the median of loaded trucks (65,000 lb) or midway between a minimally loaded (33,000 lb) and maximally loaded (80,000 lb) Class 8 tractor-trailer combination (56,500 lb). Using this approach, maximal and minimal weights could be assumed to equal mean plus or minus two times the SD, which would equate to an uncertainty of 11,780 lb and a CV_Weight of 18.1%. Although the uncertainty of vehicle weight is four times greater when estimated compared to when it is measured, final propagated uncertainty associated with not actually weighing a vehicle is only 3.8% and 4.5% greater for final estimates of brake-specific NO_X and PM emissions, respectively. Uncertainty with measured vehicle weight is used for the remainder of this analysis, but uncertainty with estimating weight can be introduced into the final calculations if weigh-in-motion measurements are not available.

Group 3—Vehicle power

Vehicle power can be derived from the road load equation, vehicle mass, and the road grade according to eq 6:

(6)

where a, b, and c are target road load coefficients for a Class 8 truck (0.1001 lbsf/mph², 6.605 lbsf/mph, and 268.08 lbsf, respectively, as derived following SAE J1263), θ is the road incline angle (if present), and g is the gravitational acceleration constant equal to 21.94 mph/sec. We assume there is no uncertainty in measuring the actual road angle θ or in the gravitational acceleration constants. We assume the combined uncertainty in road load coefficients a, b, and c is 3% (CV_L) based on the criteria of SAE J1263. The uncertainty values for V, A, and W were previously derived in the previous section (Group 2). Propagating the uncertainty of these dependent variables, the combined uncertainty in vehicle power is 13.0% with a neutral road grade, and generally no more than 13.4% with nonzero road grade because the additional term reintroduces uncertainty of vehicle weight W and speed V a second time. In this analysis, g and θ are assumed to be known with no uncertainty, and the 0.4% increase in uncertainty in vehicle power assumes a maximum real-world value for road grade angle of 5.7 degrees (a 10% road grade).

Group 4—Uncertainty in vehicle emissions

In addition to the uncertainty of the analyzer measurement, a critical source of variability is the uncertainty in the actual emissions emitted from a vehicle. The uncertainty in the actual emissions is derived from the analysis presented in Figure 3 for CO₂ emissions, in Figure 6b for PM emissions, and in Figure 8b for NO_X emissions. The uncertainty of PM predictions is based on the measurement accuracy of an aethalometer that reports the BC fraction only. Given that the majority of PM emissions from a compromised DPF is solid particles and detectable as BC for modern combustion engines (Kamboures et al. 2013, 2015), this assumption is reasonable for predicting error bounds of the instrumentation. Moreover, measurements of only BC emissions need to assume a BC to PM ratio to apply to proposed emissions thresholds. For modern diesel engines, organic carbon, sulfate, or other inorganics may contribute the majority of the mass fractions under some conditions that need to be considered when calculating total PM mass (Quiros et al. 2016). The shaded regions of Figures 6b and 8b indicate the 5th to 95th percentile extent of the modeled emissions data, which were assumed to span from minus to plus two times the SD.

Table 4 presents the CVs for the median power of the UDDS cycle (77 hp for these Class 8 test vehicles); however, as shown by Figures 6 and 8, means and uncertainties are both a function of power and need to be calculated accordingly. The values in Table 4 and values used in discussion are applicable for the median power of the UDDS horsepower.

The CV for BSFC can be calculated using the mean as expressed by the equation annotated within Figures 3a and 3b, and one SD can be calculated according to 5250 × [hp]^−0.842, which was empirically derived from data presented in Figure 3. Using a similar approach, the CV for the PM duty cycle factor can be calculated using the mean as expressed by the equation annotated within Figure 6b, and the SD can be expressed as a polynomial according to 2.91 × 10^–5 [hp]² – 0.0137[hp] + 1.794. The CV for the NOx duty cycle factor can be calculated using the mean as expressed by the equation annotated within Figure 8b, and one SD can be expressed according to the equation 6.48 × 10^–5 log[hp] + 0.0477.

Overall method uncertainty

The overall uncertainty of a roadside plume measurement system was calculated by propagating the uncertainty described in Groups 1–4 just described. We considered uncertainty from the following categories: the integrated concentration of CO₂ (eq 2), the integrated concentration of the pollutant (here, PM or NO_X, eq 2), the power estimated with measuring vehicle weight (eq 6), BSFC emissions (eq 4), the BSFC reported in Figures 3a and 3b, and the duty cycle factors derived from data presented in Figures 6b and 8b using the equations in the panels and the previous paragraph. Table 4 presents the uncertainty of the overall method for emission snapshots at the median UDDS value (77 hp, 35.3% for PM and 41.6% for NO_X).

Figures 10 and 11 show fuel-based emission factors corresponding to certification standards of 0.01 g/bhp-hr for PM and 0.2 g/bhp-hr for NO_X emissions, and their corresponding OBD emissions thresholds that are defined here as 3 times the certification standards. Existing regulations in California require detection of in-use emissions in excess of 2 or 2½ times the certification standards. The fuel-based PM emission factor is highly dependent upon vehicle power, where emission factors as low as 0.05 g PM/kg fuel or as high as 0.5 g PM/kg fuel could indicate exceedances of the OBD threshold at the three-sigma (>99%) confidence level. Fuel-based NO_X emission factors are more uniformly distributed over a wider power range, where emission factors between ~8 and 9 g NO_X/kg fuel more consistently indicate exceedances of the OBD threshold at the 99% confidence level.

Figure 10. Fuel-based emission factor thresholds equivalent to the certification (0.01 g/bhp-hr) and OBD emissions limits (0.03 g/bhp-hr) as a function of vehicle power. Shaded regions indicate +3 × CV, which contains >99% of real-time observations measured during controlled laboratory conditions for this evaluation.

Figure 11. Fuel-based emission factor thresholds equivalent to the certification (0.2 g/bhp-hr) and OBD emissions limits (0.6 g/bhp-hr) as a function of vehicle power. Shaded regions indicate +3 × CV, which contains >99% of real-time observations measured during controlled laboratory conditions for this evaluation.

Application and other considerations

Although the bulk of the discussion in this paper is to identify vehicles that emit cumulatively over specific emissions levels, there are many potential and legally valid reasons why a vehicle may be flagged but be operating within its certified specification. Vehicles can emit, at least temporarily, emissions in excess of the cycle-average certification limits (such as cold-start emissions or hard accelerations), and roadside plume detection systems may capture and report emission factors that do not reflect typical operation of a given vehicle. Likewise, any disclosed auxiliary emission control devices (AECDs) that temporarily reduce the effectiveness of an emission control system, but are fully disclosed and approved during certification, create temporary but allowed increases in emission rates. One such situation includes when a vehicle may be undergoing a DPF regeneration event, where elevated PM emissions are temporarily permitted to exceed average certification limits. Excess emissions during AECDs and DPF regeneration are quantified and accounted for during the engine certification process with the EPA and CARB. The reported thresholds indicate the 99% confidence levels of exceeding the OBD thresholds; during proper operation, 1% of the observations may exceed a given OBD threshold limit due to uncertainty of some parameter in the remote sensing system.

This analysis was limited to periods where SCR catalysts had achieved light off and were reducing tailpipe NOx emissions. Site selection and vehicle behavior can increase the probability of this occurring; however, a variety of real-world factors, such as the interaction of current NOx control strategy limitations with cold ambient temperatures or unforeseen traffic congestion, could result in emissions being measured from trucks with after-treatment temperatures below SCR light off. Such measurements of deactivated SCR systems as a result of duty cycle are important to the overall inventory, and assessment of OEM design robustness, but are not strictly diagnostic of catalytic activity loss or dosing equipment failures.

The use of roadside plume measurements hinges upon the assumption that an exhaust plume from the engine of a vehicle was entrained into the sampling system at sufficient concentration that the pollutants can be distinguished from background levels. The uncertainty of the analyzers was considered in uncertainty; however, even highly sensitive laboratory analyzers may not be able to resolve or measure small increases in exhaust plumes that have been diluted too greatly. Possible interferences could include fluctuating or high levels of near-road or ambient background of PM, NO_X, and CO₂ emissions, simultaneous entrainment of plumes from multiple vehicles, or mixtures of plumes from multiple sources on the same vehicles. A number of on-road vehicles are equipped with transport refrigeration units (TRUs), which are generally diesel powered and may exhibit emission profiles similar to those of the larger motive engines. The data analysis and calculated emission thresholds presented in this paper assume adequate capture of a single exhaust plume from the truck’s main propulsion engine.

In a situation where the preceding considerations do not apply, where an exhaust plume is adequately captured, and the vehicle operation aligned with suitable criteria outlined in this analysis, there are four additional limitations of this work that must be considered:

1. The analysis is based on four HD vehicles equipped with engines built by two engine manufacturers, and trends observed here are not guaranteed to be repeated by other engine makers today. A greater number of vehicles could be tested on the chassis dynamometer to identify more widely applicable emission thresholds, although the proposed thresholds are based on more than 10 hr of real-time laboratory testing data.

2. Emission thresholds for PM were obtained from a vehicle without a DPF, but not a damaged, malfunctioning, or mal-maintained DPF. PM emissions from a damaged DPF may exhibit different trends as a function of power or exhaust flow compared to an engine without a DPF, and could be further explored. A focused study could easily introduce intentional physical damage to a DPF, or other upstream engine component, as a means of better quantifying the impacts on PM emissions.

3. Our values of BSFC were derived from engine broadcast data, but vehicle power calculated at the roadside is based on Newtonian measurements and the road load equation. The difference between these two parameters is not significant (EPA’s HD GHG and Fuel Efficiency Standards, Phase 2 rule estimates 6% driveline losses [EPA 2015; EPA 2017]).

4. The propagation of error includes several elements and assumes the distribution of measurements and uncertainties follow a normal distribution. Alternative types of probability distributions could be more appropriate for some variables and thereby more accurate when characterizing the spread of potential input data. However, total uncertainty defined by the coefficient of variation for PM and NO_X emissions is 35 and 42%, respectively, when calculated for operation at 77 hp. A three-sigma outlier at these uncertainty levels of measured values would be 106% and 125% of measured emission factors for NO_X and PM. Three-sigma uncertainty estimates of roadside detection systems are still substantially lower than the 200% increase above certification levels that CARB’s HD OBD regulation defines as the threshold for illuminating the dashboard malfunction indicator lamp (MIL).

Summary

This analysis proposes a new, innovative approach for fuel-based emissions thresholds, as a function of vehicle power, which can be used for advanced detection systems measuring exhaust plumes at the roadside from passing HD vehicles. The work quantifies sources of uncertainty associated with deriving fuel-based emission factors from roadside plume measurements. This analysis quantifies uncertainty from the following sources: analytical pollutant analyzers; Newtonian parameters including vehicle speed, acceleration, and mass; road load and vehicle power calculations; and uncertainty in emissions from four fleet-representative HD vehicles measured on the laboratory chassis dynamometer. This analysis propagates total system uncertainty to calculate the magnitude of emissions required to achieve 99% confidence that a given work-based emissions factor has been exceeded.

Analysis indicates that a typical HD vehicle will exceed the MY 2010 emission standards (of 0.2 g NO_X/bhp-hr and 0.01 g PM/bhp-hr) by three times when fuel-based emission factors are 9.3 g NO_X/kg fuel and 0.11 g PM/kg using the roadside plume measurement approach. Analysis also indicates that the MY 2010 emission standards will have exceeded the certification standards, with 99% confidence, once fuel-based emission factors exceed 3.1 g NO_X/kg fuel and 0.035 g PM/kg fuel. These fuel-based emission factor thresholds are valid for snapshots of HD vehicles operation at 77 hp. This paper presents empirically derived equations for calculating the relevant fuel-based emission thresholds for a wider range of operating HD vehicle power.

Acknowledgment

The authors acknowledge the contributions of Robert Ianni, Wayne Sobieralski, Arlmon Vanzant, and Donald Chernich for performing emissions testing in the laboratory, and Tai Sea Yen for his assistance with procuring vehicles for this test program. The authors also thank Dr. Michael Benjamin and Kim Heroy-Rogalski for their thoughtful review and insightful comments on this paper.

The statements and opinions expressed in this paper are solely the authors’ and do not represent the official position of CARB or the SMAQMD. The mention of trade names, products, and organizations does not constitute endorsement or recommendation for use. CARB is a department of the California Environmental Protection Agency. CARB’s mission is to promote and protect public health, welfare, and ecological resources through effective reduction of air pollutants while recognizing and considering effects on the economy. CARB oversees all air pollution control efforts in California to attain and maintain health-based air quality standards. SMAQMD is the local public authority with responsibility for advancing the clean air and climate actions in the California capital region.

Additional information

Notes on contributors

David C. Quiros

David C. Quiros is a manager at the California Air Resources Board working on a variety of mobile and stationary source measurement, monitoring, and regulatory programs.

Jeremy D. Smith

Jeremy D. Smith is an Air Pollution Specialist at the California Air Resources Board, working on technical development of remote sensing systems.

Walter A. Ham

Walter A. Ham is a manager at the California Air Resources Board working on a variety of mobile and stationary source measurement, monitoring, and regulatory programs.

William H. Robertson

William H. Robertson is a Vehicle Program Specialist at the California Air Resources Board, dedicated to leading heavy-duty vehicle programs and regulatory strategies among multiple divisions.

Tao Huai

Tao Huai is a Branch Chief, overseeing all of the California Air Resources Board’s heavy-duty vehicle emissions work.

Alberto Ayala

Alberto Ayala is Executive Director and Air Pollution Control Officer of the Sacramento Metropolitan Air Quality Management District and former Deputy Executive Officer of the California Air Resources Board in charge of mobile sources.

Shaohua Hu

Shaohua Hu is a manager at the California Air Resources Board working on a variety of mobile and stationary source measurement, monitoring, and regulatory programs.