INTRODUCTION

The paper investigates the energy and environmental impacts of route choice decisions using in-field collected Global Positioning System (GPS) data and simulation results. Traffic congestion has grown significantly in the past two decades. A recent study found that the total hours of delay in the United States increased by 528% (0.7 to 3.7 billion hours) from 1982 to 2003, and individual travelers spend about three times as many extra delay hours (16 to 47 hours) than they did twenty years ago. Furthermore, congestion affects more roads, trips, and times of day in most U.S. metropolitan areas (Schrank and Lomax 2005).

Due to congestion, motorists face a difficult trip-planning process when attempting to reduce delays and improve travel time reliability. This decision-making process is based on the drivers’ urgency, experience, and current information on travel time, trip distance, and other trip-related factors. However, energy and environmental impacts are not typically utilized in drivers’ decision-making process.

Motorists typically choose routes that minimize their travel cost (e.g., travel time). Therefore, drivers typically select longer routes if it can save travel cost. However, the question that needs addressing is whether taking a longer but faster route can result in energy and air quality improvements. This study investigates the impacts of route choice decisions on vehicle fuel consumption and emission rates using GPS data gathered during the morning commute near a suburb in the Washington, DC metropolitan area. The study analysis is further expanded by conducting a sensitivity analysis using the INTEGRATION microscopic traffic simulation software.

The objectives of this study were twofold. First, the study investigates the impact of two route choices on vehicle fuel consumption and emission rates. Second, the study demonstrates the effectiveness of utilizing User-Equilibrium (UE) and System Optimum (SO) traffic assignments for environmental improvement and energy saving considerations.

OVERVIEW OF ENVIRONMENTAL IMPACTS OF TRAFFIC ASSIGNMENT

The commonly used UE and SO traffic assignment models typically utilize minimum travel time as a generalized cost to assign traffic flows over a network. However UE and SO assignments are estimated based on travel time, the fuel consumption and emissions of UE and SO conditions may not produce optimum energy and emission rates. Several researchers have investigated traffic assignment methods using environmental cost functions.

Tzeng and Chen (1993) developed the multi-objective traffic assignment method. They formulated multi-objective functions using nonlinear programming techniques and produced various solutions to emit low CO emissions. By utilizing the eigenvector weighting method with pair-wise comparison, the researchers estimated compromised solutions for the flow patterns. They applied the case study of metropolitan Taipei to evaluate the developed traffic assignment model, which utilized the simplified travel time function and CO emission module. The study utilized a fixed amount of CO emissions per link and the total emissions were summed up across all vehicles on a link (Tzeng and Chen 1993).

Rilett and Benedek (1994) and Benedek and Rilett (1998) investigated an equitable traffic assignment with environmental cost functions. They emphasized the impacts of CO emissions when UE and SO traffic assignments were applied to a sample network, a simple network from Ottawa, Ontario, Canada and a calibrated network from Edmonton, Alberta, Canada. The studies utilized a simple macroscopic CO emission model used in the TRANSYT 7F software. The emission model utilized the average speed and the link length as input variables. The researchers showed that the traffic flows of the SO-CO (the traffic flows that have the minimum total CO emissions) condition were roughly equivalent to the flows of the UE and SO conditions within a small error range (Benedek and Rilett 1998; Rilett and Benedek 1994).

Sugawara and Niemeier (2002) developed an emission-optimized traffic assignment model that used average speed CO emission factors developed by the California Air Resources Board (CARB). The sample network case study concluded that emission-optimized trip assignments can reduce system-level vehicle emissions moderately when compared to the time-dependent UE and SO conditions. The research also found that the emission-optimized assignment is most effective when the network is under low to moderately congested conditions, saving up to 30% of total CO emissions; when the network is highly congested, the emission reduction is diminished to 8%. The authors explain that under emission-optimized conditions, less traffic volume is assigned to the freeway because emission levels are very high at freeway free-flow speeds (Sugawara and Niemeier 2002).

Nagurney and her colleagues developed a multi-class and multi-criteria traffic network equilibrium model with an environmental criterion and claimed that a desired environmental quality standard can be achieved by the proposed model through a particular weighting method. In the study, a fixed amount of CO emission rate per traveler per link was utilized to estimate the total CO emissions (Nagurney 2000; Nagurney and Dong 2002; Nagurney et al. 1998).

As demonstrated, a number of researchers have focused on traffic assignment methods that can improve the environment. However, these research efforts have utilized simplified travel time functions and simplified mathematical expressions to compute emission rates based on average link speeds without regarding transient changes in a vehicle’s speed and acceleration as it travels. These approaches have been accepted by many researchers due to their simplicity; however, these methods are not adequate to quantify the energy and environmental impact of route choice behavior, particularly on congested networks, due to vehicle operational behaviors. To overcome the limitations of current research methods of evaluating the impact of traffic assignment, this study adopted a microscopic fuel consumption and emission models using instantaneous speed and acceleration as explanatory variables.

STUDY CORRIDOR CHARACTERISTICS

To identify the energy and environmental impacts of route choice behavior, morning commute GPS data were collected in the Northern Virginia area. As shown in Figure 1, the arterial route, VA Route 7, extends over 22.6 km (17.25 mi) and covers 32 signalized intersections. The study section started at the intersection of VA 28 (Sully Road) to the west and extended to the intersection of I–66 to the east. The corridor’s entire length is divided, with a four-lane cross-section on the eastern side and a six-lane cross-section on the western side. The posted speed limits range from 56 km/h (35 mph) on the congested east side to 88 km/h (55 mph) on the west side.

The highway route connects two highway sections and two arterial sections as shown in Figure 1. The highway section extends from the intersection of VA 28 (Sully Road) and Route 7 to the south and connects to a section of VA 267 (Dulles Toll Road) and a section of I–495 and finally connects to Route 7. The total distance of the highway route is 35.85 km (22.41 mi) of which 22.56 km (17.25 mi) traverses highways (VA 267 and I–495). The arterial section of this route consists of a section of VA 28 that extends over 9.94 km (6.21 mi) and covers four signalized intersections and a section of Route 7, which is 3.35 km (2.10 mi) long and has six signalized intersections.

Traffic flows along the corridors are typically directional. During the morning peak, traffic along the study corridors generally moves eastbound, toward downtown Washington, DC and Fairfax, Virginia. It should be noted that the eastern portion of the study section, which has closely-spaced signalized intersections on VA 7, is typically more congested than other portions of the study section. The study corridors are controlled by a centralized, computerized signal system with an optimized cycle length of 180 seconds or 210 seconds depending on the time of day. Most of the signal cycle time is assigned to the main route VA 7 and VA 28. The directional distribution of signal timing varies according to the time-of-day and, during the morning peak, more signal timing is assigned to the eastbound direction. These signal timings are continuously optimized by the Virginia Department of Transportation (VDOT) staff and thus represent the state-of-practice in optimal signal control. Since the study only investigates the impact of different route choices on vehicle fuel consumption and emission rates, a detailed description of the traffic signal operations on the study corridor is beyond the scope of the paper. The study utilizes GPS data collected under current traffic signal operations on the study sections during the morning commute period.

GPS DATA COLLECTION PROCEDURES

GPS technology is increasingly being used for transportation-related applications. The study utilized a portable Wide Area Augmentation System (WAAS)-enabled GPS receiver to gather second-by-second vehicle trajectories along the study sections. The WAAS-enabled GPS receiver provides longitude and latitude data to an accuracy of 2 m, altitude data to an accuracy of 3 m, and speed measurements to an accuracy of 0.1 m/s.

The study used a portable GPS unit, GD30L, manufactured by LAIPAC Technology, Inc. The GPS unit is designed to record date, time, vehicle longitude, vehicle latitude, vehicle speed, vehicle heading, and the number of tracking satellites. The system is completely configurable, and the user can change the setup of the DIP switches to select the recording interval from 1 second to 30 min as well as the data recording format. The logged GPS data are recorded in a removable MultiMedia Flash Memory Card (MMC), and the 32MB MMC easily holds 30 days of passenger car operational data at 1-second intervals. A flash memory card reader was used to transfer the GPS data to a PC. The device is operated as a stand-alone unit without the need for a PC or other equipment. Once the GD30L is powered up, the GPS unit collects the data automatically.

The GPS floating-car travel data were collected using a test vehicle on weekdays (Monday through Friday) between March and May of 2006. The trip route (highway or arterial) was randomly selected on the day of data collection. In order to record the aggregate characteristics of traffic flow, the probe vehicle maintained the average speed of the traffic stream. The travel data were recorded at a 1-second resolution and downloaded to a personal computer. The minimum sample size (N) was calculated to satisfy the 95% confidence limits (Z value 1.96) using the standard deviation (σ) value and travel time error (δ) (see Equation 1). The GPS data that were gathered exceeded the required minimum sample size. In total, 39 valid trips were recorded, of which 21 traveled on the highway route and 18 traveled on the arterial route, while 10 trips on the highway route and 11 trips on the arterial route were required to satisfy the minimum sample size considering a 95% confidence limit.

(1)

It should be noted that the morning and evening commute data were gathered. However, this study focuses on the morning commute only. Since the GPS data included the entire morning commute travel data, the portion of the trips that covered the study sections was extracted from the entire trip for analysis purposes using a MATLAB code that was developed for this purpose. The software automatically identified the first and last GPS points within the study corridor using the coordinates of the boundary study sections. Following the data reduction, a unique trip number was assigned to each trip.

ENERGY AND EMISSION MODELS

In order to estimate emission and fuel consumption using the second-by-second GPS probe vehicle data, the VT-Micro model, the Comprehensive Modal Emissions Model (CMEM), and the Environmental Protection Agency’s (EPA) MOBILE6 model were utilized. The following sections briefly describe each model.

MOBILE6 Model

The MOBILE6 model was developed by the EPA Office of Transportation and Air Quality. MOBILE6 is the latest of the MOBILE models. MOBILE6 was developed using recent vehicle-emission testing data collected by the EPA, CARB, and automobile manufacturers, as well as inspection and maintenance tests conducted in various states. A major characteristic of the MOBILE6 model is the addition of so-called off-cycle emissions, which involve aggressive driving with various facility-type modeling. MOBILE6 estimates emission factors based on different roadway types (e.g., highways, arterials, locals). Emission factors can be adjusted for different facility types and different average speeds based on vehicle testing over a series of facility cycles. Also, MOBILE6 estimates emission factors for the start portion and the running portion of the trip separately (Miller et al. 2000).

It should be noted that in order to model the highway trips, the average speed of each section (a highway section: VA 267 and I-495, and two arterial sections: VA 28 and VA 7) was individually simulated and combined later in the analysis. Since the study only demonstrates the relative energy and emission differences associated with motorists’ route choices, only average speeds by facility type were utilized for the sensitivity analysis. The study assumed identical conditions for other input variables for both routes. Thus, default settings of vehicle model year, mileage rate, vehicle age, vehicle-type percentage, and altitude information were used in the models instead of the fleet characteristics of Northern Virginia. Furthermore, only exhaust running emissions of light duty gasoline vehicles (LDGV) without start emissions were utilized for the study.

CMEM Model

The Comprehensive Modal Emissions Model was developed by researchers at the University of California, Riverside. CMEM estimates light-duty vehicle (LDV) and light-duty truck (LDT) emissions as a function of the vehicle’s operating mode. The term “comprehensive” is utilized to reflect the ability of the model to predict emissions for a wide variety of LDVs and LDTs in various operating states. For the test data, both engine-out and tailpipe emissions of over 300 vehicles, including more than 30 high emitters, were measured second-by-second over three driving cycles, including the Federal Test Procedure (FTP), US06, and the Modal Emission Cycle (MEC). CMEM predicts second-by-second tailpipe emissions and fuel-consumption rates for a wide range of vehicle and technology categories. Vehicle operational variables (such as speed, acceleration, and road grade) and model-calibrated parameters (such as cold-start coefficients and engine-friction factor) are utilized as input data (Barth et al. 2000). In order to estimate fuel consumption and emissions, the CMEM vehicle categories 11 and 24 were utilized. Category 11 represents Tier 1, relatively new low-mileage vehicles, and category 24 represents Tier 1, relatively old high-mileage vehicles.

VT Micro (ORNL) Model

While the CMEM model was developed as a power-demand model, the VT-Micro model was developed as a regression model from experimentation with numerous polynomial combinations of speed and acceleration levels to construct a dual-regime model of the form

(2)

where L^e_i,j and M^e_i,j represent model regression coefficients for measures of effectiveness (MOE) “e” at speed exponent “i” and acceleration exponent “j”. It should be noted that the intercept at zero speed and zero acceleration was estimated using the positive acceleration model and fixed in order to ensure a continuous function between the two regression regimes. Consequently, the calibration of the model involves estimating a total of 32 parameters for each MOE.

The model was developed utilizing a number of data sources including data collected at the Oak Ridge National Laboratory (ORNL) (9 vehicles) and the EPA (101 vehicles). These data included fuel consumption and emission rate measurements (CO, HC, and NOx) as a function of the vehicle’s instantaneous speed and acceleration levels. CO₂ emissions were estimated using the carbon balance equation in conjunction with the fuel consumption measurements. In this study, an average composite vehicle for the nine ORNL vehicles was utilized. This composite vehicle included six light-duty automobiles and three light duty trucks. These vehicles were selected in order to produce an average vehicle that was consistent with average vehicle sales in terms of engine displacement, vehicle curb weight, and vehicle type at the time the data were gathered. Specifically, the average engine size was 3.3 liters, the average number of cylinders was 5.8, and the average curb weight was 1497 kg (3300 lbs). The VT-Micro model fuel consumption and emission rates were found to be highly accurate compared to the ORNL data with coefficients of determination ranging from 0.92 to 0.99. Given that the model utilizes the vehicle's instantaneous speed and acceleration levels as independent variables, the model is easy to use for the evaluation of the environmental impacts of operational-level projects including Intelligent Transportation Systems (ITS). A more detailed description of the model derivation is provided in the literature (Melosh et al. 1990; Rakha et al. 2004b).

GPS DATA ANALYSIS

The average travel time results of the collected GPS data demonstrate that highway trips result in travel time savings of 17% with a smaller travel time variability compared to arterial trips (travel time standard deviation of 4.17 versus 5.08 min., respectively), as summarized in Table 1. In order to confirm the results, t-tests were performed at a 5% significance level assuming identical mean travel times for both cases. The t-test produced a p-value of 0.003, which indicates that there is sufficient evidence to reject the null hypothesis of equal travel times. Thus, we conclude that the travel times of the highway trips are significantly shorter than the travel times of the arterial trips even though the highway trips are 30% longer than the arterial trips (35.9 km versus 27.6 km). Table 1 also demonstrates that the highway trips have a significantly higher average speed (85.42 km/h, 53.08 mi/h) than the arterial trips (56.62 km/h, 35.18 mi/h).

The fuel consumption and emission rates estimated by the VT-Micro (ORNL), MOBILE6, and CMEM models for the highway and arterial trips are illustrated in Figure 2. It should be noted that the absolute values of fuel consumption and emissions are not necessarily equal given the potential for different model characteristics; however, the intent of this paper is not to derive definitive emission inventories but to demonstrate the relative energy and emission differences associated with motorists’ route choices. Indeed, since each emission model is based on a different dataset, each model generates different fuel consumption and emission values. It should be noted that the CO₂ and fuel consumption values of MOBILE6 are not presented in Figure 2 because MOBILE6 only estimates HC, CO, and NO_x emissions.

The figure demonstrates that a 47% reduction in HC emissions is achievable if motorists use the arterial route instead of the highway route based on the VT-Micro (ORNL) model estimates. Similarly, the output of the CMEM model demonstrates that the arterial trips result in savings in the range of 47 to 63% in HC emissions compared to the highway trips even though the highway trips can save 17% in travel time. Alternatively, the MOBILE6 model results demonstrate that highway trips reduce HC emissions by 36%.

Figure 2 also illustrates the route choice impacts on other MOEs, including CO, NO_x, CO₂, and fuel consumption. Specifically, reductions in CO emissions in the range of 52, 71, and 64% were observed when motorists utilized the arterial based on the VT-Micro (ORNL vehicle), CMEM 11, and CMEM 24 models, respectively. The figure also shows that the use of the arterial route can produce reductions in NO_x and CO₂ emissions up to 45 and 20%, respectively. It is interesting to note that 18 to 23% of energy cost can be saved when motorists sacrifice 4.3 minutes (17%) of travel time in this case study. However, it should be noted that similar to the results of HC emission, the MOBILE6 model estimated that CO and NO_x emissions were increased 7 and 23% when drivers selected the arterial route.

Figure 3 illustrates the variations in HC emissions over the individual trips with travel time and average speed data for both highway and arterial trips. The figure illustrates that the trend of HC emissions from VT-Micro (ORNL vehicle) models are consistent with that of the CMEM models. It should be noted that both models are microscopic emission models developed to capture the second-by-second operational behavior of individual vehicles. The figure also shows that the HC emissions of the MOBILE6 model were found to be more sensitive to the average speed of each trip than to the intricate differences in speed profiles. Because the CMEM, VT-Micro, and MOBILE6 models were developed based on different emission sources, the absolute values of each emission model are different. However, it should be noted that the general trends are consistent for the micro models but inconsistent in the case of the MOBILE6 model.

CO, NO_x, and CO₂ emissions and fuel consumption rates of individual trips are also illustrated in Figure 4. As shown in the figure, the trend of CO emissions in the VT-Micro model (ORNL vehicle) and CMEM models along highway and arterial trips is analogous to that of HC emissions. In case of CO emissions, highway trip 12 (24.97 min., 86.03 km/hr) produces significantly higher CO emission (3.14 times) compared to highway trip 13 (24.98 min., 86.05 km/hr) with CMEM 24 model, demonstrating that CO emissions are more sensitive to engine loads than other air pollutants. It should be noted that highway trip 12 produced 50 high engine load occurrences that each produced emissions of CO in excess of 5 g/s while highway trip 13 had only 8 instances of such cases.

It should also be noted that the CO and NO_X emissions of the MOBILE6 model for arterial trips were found to be less sensitive to differences across the drive cycles, thus demonstrating that the MOBILE6 model cannot reflect the instantaneous changes in speed profiles. Additionally, note that the trends of CO₂ and fuel consumption are identical for both highway and arterial trips.

Correlation coefficients among travel time, average speed, fuel consumption, and emissions of individual trips are shown in Table 2. The table shows that it is difficult to find relationships between travel time (or average speed) and the estimated emissions and fuel consumption rates produced by the VT‑Micro (ORNL) and CMEM models considering the non-linear behavior of vehicle emissions, demonstrating that MOEs produced by microscopic emission models are closely related to instantaneous vehicle operations as opposed to aggregate trip characteristics. However, high correlation coefficients are observed between travel time and HC and NO_X emissions of MOBILE6 in Table 2. Also, the emissions and fuel consumption rates of the VT-Micro (ORNL vehicle) model are highly correlated with the results of the CMEM model due to the microscopic nature of the models.

Under congested conditions, motorists tend to choose the route with the minimum travel cost (e.g., travel time). Between each origin and destination, UE is reached when no driver can unilaterally achieve a reduction in time or cost by changing his/her route of travel; in other words, the travel times for all routes are equal. Emissions and fuel consumption rates produced by the VT-Micro (ORNL vehicle), CMEM, and MOBILE6 models are presented in Figure 5 when the travel times of the highway trip (28.8 min) and the arterial trip (28.8 min) are identical. It may not be appropriate to claim that the UE condition is reached even under these conditions because the two trips were recorded on two different days. Figure 5 illustrates that even when the travel times are identical, motorists can save significant emission and fuel consumption rates when the arterial route is selected. Specifically, the VT-Micro model estimates reductions in the range of 44%, 49%, 25% and 18% for HC, CO, NO_X, and CO₂ emissions, respectively. Furthermore, a driver can save up to 19% in fuel costs by using the arterial route.

Second-by-second emissions and fuel consumption of highway and arterial trips under UE conditions are illustrated in Figure 6. The figure also includes the speed profiles of the highway and arterial trips, which involve several full and partial stops in addition to high-speed travel (speeds of about 100 km/h or 62 mph for the arterial trip and 120 km/h or 75 mph for the highway trip). The average speeds of the highway and arterial trips were 74 km/h (46 mph) and 57 km/h (35 mph), respectively. Figure 6 illustrates the variations in the instantaneous vehicle emissions and fuel consumption rates as estimated by the VT-Micro (ORNL vehicle) and CMEM 24 models as the vehicle traveled along the highway and arterial routes. The instantaneous emissions show the numerous peaks and valleys of the vehicle emissions, demonstrating that the MOEs are sensitive to changes in a vehicle’s speed and acceleration profile. Also, it is noted that some peaks of HC and CO emissions represent a significant amount of total emissions. For example, a few peaks of CO emissions are several thousand times greater than those of other instantaneous emission rates.

Table 3 shows that 1% (17 seconds of 1,729 seconds) of the highway trip is responsible for 16%, 19%, 4%, 3%, and 4% of the total HC, CO, NO_X, CO₂ emissions, and fuel consumption rates, respectively, along the drive cycle shown in Figure 6 when the MOEs are estimated by the VT‑Micro (ORNL) model. The CMEM 24 model shows similar results: 20%, 38%, 30%, 3%, and 5% of total HC, CO, NO_X, CO₂ emissions, and fuel consumption rates, respectively, are produced by the 1% portion of the trip. The results of the CMEM 24 model shows that almost 100% of HC and NO_X emissions are emitted by 10% (173 seconds of 1,729 seconds) of the arterial trip, implying that a small fraction of the trip is responsible for a significant amount of the total emissions while the results of the VT-Micro show smaller contributions. Note that, in general, CO₂ emission and fuel consumption rates are impacted less than other MOEs; up to 25 and 28% of total CO₂ emissions and fuel consumption rates are caused by 10% of the trip profile. These small portions of a trip are caused by high engine-load conditions. Consequently, the study demonstrates that a reduction of high-emitting driving behavior can significantly improve air quality.

SIMULATION RESULTS

In order to examine the system-wide impacts of traffic assignment on air quality and energy, the study utilized the INTEGRATION microscopic traffic assignment and simulation software to conduct further analyses. The INTEGRATION software utilizes VT-Micro fuel consumption and emission models with the ORNL vehicle type and a detailed description of the energy and emission modules within the INTEGRATION model is provided in the literature (Rakha and Ahn 2004).

The sample network consists of two links and two nodes, as illustrated in Figure 7. The O-D (origin-destination) demand is 5,000 vehicles per hour (veh/hr), and there are two routes available. The highway route is 5 km (3.1 mi) long with three lanes for the first 4 km (2.5 mi) and two lanes for the remaining 1 km (0.6 mi). It has a free-flow speed of 100 km/hr (62.5 mph) and a capacity of 1,800 veh/hr/lane. The jam density of the highway route is set to 100 veh/km, and the speed-at-capacity is coded at 90 km/h (56.3 mph). The arterial route is a shorter two-lane, 4-km (2.5‑mi) corridor that has three signalized intersections located every 1 km (0.6 mi), as illustrated in Figure 7. The arterial route has a lower free-flow speed of 75 km/hr (47 mph) with a 67 km/h (42 mph) speed-at-capacity and an identical jam density and lane capacity as those for the highway route. Also, it should be noted that three signals have a 60-second cycle length with a 0.67 g/C ratio (effective green time to cycle length ratio), and they are partially coordinated.

Eleven traffic assignment scenarios were utilized in this case study. Figure 8 shows the traffic assignment scenarios simulated using the INTEGRATION software. The vehicles assigned to the arterial route were increased from 500 vehicles (scenario 2) to 4,500 vehicles (scenario 10) in increments of 500 vehicles; for example, scenario 5 shows that 2,000 vehicles were assigned to the arterial route and 3,000 vehicles were assigned to the highway route. Scenarios 1 and 11 were excluded from the rule; in other words, 100 vehicles and 4,900 vehicles were assigned to the arterial route and the highway route, respectively, in scenario 1 and vice versa for scenario 11. Each simulation was loaded for 30 minutes and continued simulation for 90 minutes in order to clear all loaded vehicles. Therefore, 2,500 vehicles were utilized for this analysis.

The simulation results show that the SO condition is attained in scenario 5, as illustrated in Figure 8, which has the smallest total travel time of the entire network. Figure 8 also shows the UE condition in scenario 4, which has the same average travel time for the two routes. It should be noted that the total travel times (or total delay) are significantly increased as 4,000 veh/hr or more are assigned to the arterial route.

The emissions- and/or fuel-consumption-optimized traffic assignments are illustrated in Figure 9. Scenario 8 is the HC and CO emission-optimized traffic assignment while scenario 5, which is the SO condition based on travel time, also produces the minimum CO₂ emission and fuel consumption rates. These results are inconsistent with the results of Rilett and Benedek (1994), in which the traffic flows of the CO-optimized assignment condition were roughly equivalent to the flows of the UE and SO conditions within a small error range. Also, scenario 11, which assigns most of the vehicles to the arterial route and has the greatest total delay among all scenarios (8.4 times more delay than SO condition), produces the least NO_X emissions. It is concluded that because more NO_X emissions are produced under stoichiometric air/fuel ratio conditions, low-speed congested traffic conditions, which are typically under fuel-rich or fuel-lean conditions, do not increase NO_X emissions. It is interesting to note that HC and CO emissions, NO_X emissions, CO₂ emissions, and fuel consumption rates have different emission- and/or energy-optimized assignments, which are scenarios 8, 11, and 5, respectively. It should be noted that each emission group has different characteristics; HC and CO emissions are extremely sensitive to vehicle acceleration behaviors, and the highest emissions are produced under a fuel-rich condition while more NO_X emissions are generated under a stoichiometric ratio condition. Fuel consumption and CO₂ emissions are also sensitive to acceleration behaviors but not as sensitive as HC and CO emissions.

Figure 10 illustrates emission- and energy-optimized traffic assignments for high-emitting (HE4) and low-emitting (LDV3) vehicles as well as the ORNL vehicles. High emitters are motor vehicles that produce higher emissions than the average-emitting vehicle under normal driving conditions. It is known that a small fraction of high emitters contributes significantly to total mobile source emissions. Detailed description of high-emitting (HE4) and low-emitting (LDV3) vehicles is provided in the literature (Melosh et al. 1990; Rakha et al. 2004b), and both emission models were incorporated into the INTEGRATION software. The figure clearly illustrates that high-emitting vehicles emit significantly higher HC, CO, and NO_X emissions compared to other vehicle types, and low-emitting vehicles produce lower emissions than ORNL vehicles. It is interesting to note that ORNL vehicles consume more fuel than high emitters. The figure shows that fuel consumption and CO₂ emissions have the same optimized assignment, scenario 5, for all vehicle types. Similarly, minimum NO_X emissions for all vehicle types are observed in scenario 11. However, HC and CO emissions have different emission-optimized assignments for each vehicle type. Figure 10 shows that the emission-optimized assignment of HC and CO emissions for high-emitting vehicles is scenario 5, which is the SO condition, while in terms of low-emitting vehicles, lower HC and CO emissions are observed as more traffic is assigned to the arterial route. This trend is caused by the fact that high-emitting vehicles typically generate more emissions at low speeds and acceleration-operational conditions than normal vehicles due to various reasons (e.g., malfunction of catalytic converter) while low-emitting vehicles produce significantly lower emissions under the same operational conditions. It should be noted that catalytic converters are typically well operated under low speed and acceleration-operational conditions than high speed operational conditions. Thus, low-emitting vehicles produce small amounts of HC and CO emissions when they are assigned to the low-speed arterial route. Therefore it is recommended that in order to reduce HC and CO emissions, more low-emitting vehicles should be assigned to lower-speed routes. From a societal standpoint this may not be an efficient approach, given that such an assignment would be providing better routes to environmentally less-efficient vehicles and thus would encourage drivers not to fix their vehicles.

Table 4 demonstrates that vehicle emission and fuel consumption rates can be reduced when emission- and energy-optimized assignments are utilized. As seen in the table, the emission-optimized assignments provide a significant reduction in emissions, particularly of HC, CO, and NO_X. However, the savings of CO₂ emission and fuel consumption rates of emission-optimized assignment over UE assignment are less than HC, CO, and NO_X emissions but are still in the range of 4.0 to 7.6%. The emission-optimized assignments produce significant emission reductions over the UE and SO assignments. The table shows that low-emitting vehicles can save more HC, CO, and NO_X emissions (38.87, 92.13, and 36.40%, respectively) when the emission-optimized assignment is utilized than other vehicle types over the SO and UE assignment because low-emitting vehicles have significantly different traffic flows compared to SO and UE conditions. Also, relatively smaller HC, CO, and NO_X emission reductions—9.03, 7.25, and 14.93%, respectively—are found for high emitters.

CONCLUSIONS

This study demonstrates the environmental and energy impacts of route choice behavior using second-by-second GPS commute data and simulation results. This study demonstrates that a UE and SO traffic assignment does not necessarily minimize vehicle fuel consumption and emissions based on second-by-second GPS commute field data and simulation results. The study demonstrates that, for the specific example illustration, motorists could save 17% in travel time on highway travel (35.9 km, 22.3 mi) over travel on an arterial route (27.6 km, 17.1 mi). However, significant improvements (savings up to 63%, 71%, 45%, and 20% of emissions for HC, CO, NO_X, and CO₂, respectively) to air quality are observed when motorists utilize the longer time arterial route. Moreover, the study found that 23% of energy cost can be saved when motorists travel on the arterial route. The study also demonstrated that the VT-Micro and CMEM models produce more reasonable fuel consumption and emission estimates when compared to the macroscopic emission model because it is unable to capture the impact of instantaneous speed and acceleration levels on the various MOEs. The study also demonstrated that a small portion of the entire trip (10%) that involves operation at high engine loads can produce up to 50% of the total trip emissions and consume up to 25% of the total trip fuel consumption. Consequently, significant improvements in air quality and energy consumption are achievable by educating drivers.

In order to examine the system-wide impacts of traffic assignment on air quality and energy, the study utilized the INTEGRATION software to extend the study. ORNL, high-emitting vehicles, and low-emitting vehicles were utilized in the analysis.

The findings are summarized as follows:

· Based on the traffic assignment technique, fuel consumption and emission rates can be significantly improved.

· UE and SO conditions may not be the optimum conditions for HC, CO, and NO_X emissions.

· In terms of fuel consumption and CO₂ emissions, emission- and energy-optimized assignment is identical to the SO assignment.

· Minimum NO_X emissions are achieved when most of the vehicles are assigned to the low-speed arterial route. This finding results because more NO_X emissions are produced under a stoichiometric air/fuel ratio condition; low-speed congested traffic conditions do not increase NO_X emissions.

· Each vehicle type has different HC and CO emission-optimized assignment conditions.

· Based on vehicle types, emissions are significantly different because high-emitting vehicles typically generate more emissions at low speed and acceleration-operational conditions than normal vehicles while low-emitting vehicles produce significantly low emissions under the same operational conditions. Thus, low-emitting vehicles produce small amounts of HC and CO emissions when they are assigned to the low-speed arterial route.

· It is recommended that to reduce HC and CO emissions, more low-emitting vehicles should be assigned to lower-speed routes.

· The emission-optimized assignments produce significant emission reductions, up to 92% of CO emission, over the UE and SO assignments.

· Low-emitting vehicles can save more HC, CO, and NO_X emissions when emission-optimized assignment is utilized than other vehicle types over the SO and UE assignments.

REFERENCES

Ahn, K., Rakha, H., and Trani, A. (2004). Microframework for modeling of high-emitting vehicles. Transportation Research Record. n 1880 2004, 39-49.

Ahn, K., Rakha, H., Trani, A., and Van Aerde, M. (2002). Estimating vehicle fuel consumption and emissions based on instantaneous speed and acceleration levels. Journal of Transportation Engineering, 128(2), 182-190.

Barth, M., An, F., Younglove, T., Scora, G., Levine, C., Ross, M., and Wenzel, T. (2000). Comprehensive modal emission model (CMEM), version 2.0 user's guide, Riverside, CA.

Benedek, C. M., and Rilett, L. R. (1998). Equitable traffic assignment with environmental cost functions. Journal of Transportation Engineering-ASCE, 124(1), 16-22.

EPA. (2002). User's Guide to Mobile6, Mobile Source Emission Factor Model. EPA420-R-02-001, Ann Arbor, Michigan.

Nagurney, A. (2000). Congested urban transportation networks and emission paradoxes. Transportation Research, Part D: Transport & Environment, 5(2), 145-151.

Nagurney, A., and Dong, J. (2002). A multiclass, multicriteria traffic network equilibrium model with elastic demand. Sage Urban Studies Abstracts, 30(4), 415-517.

Nagurney, A., Ramanujam, P., and Dhanda, K. K. (1998). Multimodal traffic network equilibrium model with emission pollution permits: Compliance vs noncompliance. Transportation Research, Part D: Transport & Environment, 3(5), 349-374.

Rakha, H., and Ahn, K. (2004). Integration modeling framework for estimating mobile source emissions. Journal of Transportation Engineering, 130(2), 183-193.

Rakha, H., Ahn, K., and Trani, A. (2004). Development of VT-Micro model for estimating hot stabilized light duty vehicle and truck emissions. Transportation Research Part D-Transport and Environment, 9(1), 49-74.

Rilett, L. R., and Benedek, C. M. (1994). Traffic assignment under environmental and equity objectives. Transportation Research Record (1443), 92-99.

Schrank, D., and Lomax, T. (2005). The 2005 Urban Mobility Report, Texas Transportation Institute, Texas.

Sugawara, S., and Niemeier, D. A. (2002). PART 2 - AIR QUALITY - How Much Can Vehicle Emissions Be Reduced? Exploratory Analysis of an Upper Boundary Using an Emissions-Optimized Trip Assignment. Transportation Research Record (1815), 9.

Tzeng, G. H., and Chen, C. H. (1993). Multiobjective Decision-Making for Traffic Assignment. IEEE Transactions on Engineering Management, 40(2), 180-187.

LIST OF TABLES

TABLE 1: Trip Characteristics of GPS Data

TABLE 2: Correlation Coefficients Between MOEs

TABLE 3: Emission Percent That Contributed to Total MOEs

TABLE 4: Benefit of Using Emission-Optimized Traffic Assignment

LIST OF FIGURES

FIGURE 1: Highway and arterial study corridors.

FIGURE 2: Estimated emissions and fuel consumptions on study corridors.

FIGURE 3: HC emissions, travel times, and average speed of each trip.

FIGURE 4: CO, NO_X, CO₂, and fuel consumption of each trip.

FIGURE 5: Emissions and fuel consumption with same travel time.

FIGURE 6: Instantaneous emissions and fuel consumption with same travel time.

FIGURE 7: Sample network utilized for simulation study.

FIGURE 8: Simulation scenario and results.

FIGURE 9: Fuel consumption and emissions versus various O-D Demands (ORNL vehicle type).

FIGURE 10: Fuel consumption and emissions of various vehicle types.

TABLE 1 Trip Characteristics of GPS Data

	Highway	Arterial	Difference
Average Travel Time (min)	25.63	29.9	-4.27
95 percentile of Travel time	36.25	37.86
5 percentile of Travel time	23.32	26.23
Std. Dev. of Travel Time	4.17	5.08	-0.91
Average Speed (km/hr)	85.42	56.62	28.8
Std. Deviation of Speed	10.23	7.91	2.32
95 percentile of Speed	94.16	63.11
5 percentile of Speed	59.26	43.94
Distance (km)	35.9	27.6	8.3
Number of Trip	21	18

TABLE 2 Correlation Coefficients Between MOEs

	Highway Trips				Arterial Trips
	ORNL	CMEM11	CMEM24	MOBILE6	ORNL	CMEM11	CMEM24	MOBILE6
HC
Correlated with Travel Time	-0.06	0.02	-0.19	0.73	-0.21	-0.23	-0.52	1.00
Correlated with Average Speed	0.02	-0.06	0.15	-0.73	0.16	0.19	0.52	-0.98
Correlated with VT-Micro(ORNL)		0.94	0.94	-0.15		0.93	0.40	-0.24
Correlated with CMEM11 Model			0.90	-0.03			0.51	-0.26
Correlated with CMEM24 Model				-0.25				-0.53
CO
Correlated with Travel Time	-0.06	0.03	-0.02	0.44	-0.39	-0.23	-0.25	-0.17
Correlated with Average Speed	0.02	-0.07	-0.02	-0.44	0.34	0.19	0.21	0.32
Correlated with VT-Micro(ORNL)		0.95	0.97	-0.15		0.90	0.87	-0.29
Correlated with CMEM11 Model			0.99	-0.05			0.95	-0.26
Correlated with CMEM24 Model				-0.09				-0.28
NO
Correlated with Travel Time	-0.01	0.33	0.05	0.40	-0.53	0.20	-0.11	0.56
Correlated with Average Speed	-0.01	-0.37	-0.09	-0.39	0.48	-0.22	0.08	-0.43
Correlated with VT-Micro(ORNL)		0.63	0.65	-0.21		0.62	0.78	-0.57
Correlated with CMEM11 Model			0.94	0.00			0.87	-0.09
Correlated with CMEM24 Model				-0.10				-0.33
CO2
Correlated with Travel Time	0.85	0.73	0.71		0.86	0.83	0.82
Correlated with Average Speed	-0.86	-0.73	-0.70		-0.88	-0.84	-0.85
Correlated with VT-Micro(ORNL)		0.93	0.93			0.99	0.99
Correlated with CMEM11 Model			0.96				1.00
Correlated with CMEM24 Model
Fuel
Correlated with Travel Time	0.75	0.59	0.52		0.79	0.77	0.74
Correlated with Average Speed	-0.77	-0.60	-0.53		-0.81	-0.79	-0.77
Correlated with VT-Micro(ORNL)		0.93	0.93			0.99	0.99
Correlated with CMEM11 Model			0.97				1.00
Correlated with CMEM24 Model

TABLE 3 Emission Percent That Contributed to Total MOEs

	HC	CO	NOx	CO2	Fuel
ORNL Highway
Top 1%	16%	19%	4%	3%	4%
Top 2%	24%	30%	7%	6%	7%
Top 5%	39%	47%	17%	13%	14%
Top 10%	54%	64%	32%	23%	25%
ORNL Arterial
Top 1%	15%	20%	5%	3%	4%
Top 2%	21%	29%	9%	6%	7%
Top 5%	34%	45%	21%	13%	14%
Top 10%	47%	60%	37%	24%	25%
CMEM24 Highway
Top 1%	20%	38%	30%	3%	5%
Top 2%	32%	63%	50%	6%	9%
Top 5%	52%	80%	73%	14%	17%
Top 10%	81%	84%	90%	25%	28%
CMEM24 Arterial
Top 1%	31%	63%	55%	4%	5%
Top 2%	47%	66%	64%	7%	8%
Top 5%	90%	71%	89%	15%	16%
Top 10%	100%	77%	100%	26%	27%

TABLE 4 Benefit of Using Emission Optimized Traffic Assignment

	HC	CO	NOx	CO2	Fuel
ORNL Vehicles
Saving over UE condition	44.32%	50.15%	15.11%	7.09%	7.76%
Saving over SO condition	27.26%	33.77%	8.96%	0.00%	0.00%
HE4 Vehicles
Saving over UE condition	9.03%	7.25%	14.95%	4.04%	4.61%
Saving over SO condition	0.00%	0.00%	14.12%	0.00%	0.00%
LDV3 Vehicles
Saving over UE condition	39.89%	91.78%	37.44%	6.09%	6.30%
Saving over SO condition	38.87%	92.13%	36.40%	0.00%	0.00%

FIGURE 1 Highway and arterial study corridors.

FIGURE 2 Estimated emissions and fuel consumptions on study corridors.

FIGURE 3 HC emissions, travel times, and average speed of each trip.

FIGURE 4 CO, NO_X, CO₂, and fuel consumption of each trip.

FIGURE 5 Emissions and fuel consumption with same travel time.

FIGURE 6 Instantaneous emissions and fuel consumption with same travel time.

FIGURE 7 Sample network utilized for simulation study.

FIGURE 8 Simulation scenario and results.

FIGURE 9 Fuel consumption and emissions versus various O-D demands (ORNL vehicle type).

FIGURE 10 Fuel consumption and emissions of various vehicle types.

Environmental Impacts of Catalytic Converter Malfunctions

ABSTRACT

In order to effectively reduce the total vehicle emissions, identifying and understanding the effects of high emitting vehicles plays an important role to improve the air quality in urban areas. While several studies have documented the environmental impacts of high emitting vehicles, these studies have not looked at the operational impacts of high emitting vehicles. This paper investigates the operational effects associated with catalytic converter malfunctions on vehicle emissions using field-collected and simulated data. In addition, the study develops microscopic emission models for vehicles with catalytic converter malfunctions using on-road emission measurement data. The study demonstrates that catalytic converter malfunctions produce increases in HC, CO, and NO_x emissions of 297%, 211%, and 378%, respectively, while cruising. Furthermore, increases of 63.4%, 26.9%, and 76.1% are observed during acceleration maneuvers. The simulation results demonstrate that the highest impact of catalytic converter malfunctions occurs in the 45 to 85 km/h speed range at mild acceleration levels for HC and CO emissions and between 20 and 55 km/h at high acceleration levels in the case of NO_x emissions. Simulation analysis of various driving cycles demonstrates that catalytic converter malfunctions produce overall increases in HC, CO, and NO_x emissions in the range of 251%, 225%, and 336%, respectively.

Introduction

This paper investigates the operational impacts of catalytic converter malfunctions using field-collected and simulation data. High-emitters or high-emitting vehicles are motor vehicles that produce higher emissions than the average-emitting vehicles under normal driving conditions. Most of the vehicles on the road produce relatively low emissions. However, a small number of vehicles, that are high emitters, are responsible for a significant amount of the fleet emissions (Wenzel et al. 2000). HC and CO emission levels of the dirtiest 10% are six times higher than the fleet average, and the NOx emission level of the dirtiest 10% is three times higher than the average (McClintock 1999). Schwartz (2000) also has shown that the dirtiest 10% of vehicles account for as much as 50% of HC and CO emissions on the road. Numerous studies have documented the effects of high emitters (Wenzel and Ross 1998; Wolfe 1990). However, the operational effects of high emitters have not been systematically investigated. In order to effectively reduce the total vehicle emissions, identifying and understanding the effects of high emitters must play an important role to improve air quality in urban area.

There are various types of high emitters. Researchers from University of California, Riverside classified the high emitter as being one of four vehicle types, determined according to their emission characteristics. These high emitters include vehicles that operate lean or rich, have a misfire, and have catalyst related problems. The lean high emitter (type1) is a vehicle that operates with a chronically lean fuel-to-air ratio at moderate power. This type of vehicle typically shows low HC and CO emissions and high NOx emissions. While the exact physical problem is unknown, an improper signal from the oxygen sensor or an improper functioning of the electronic engine control could be causes for such behavior. The rich type of high emitter (type2) is a vehicle with a rich fuel-to-air ratio at moderate power demands. Under these conditions, the engine-out HC emissions remain normal; however, the CO emission index and catalyst pass fraction are high, resulting in high tailpipe CO emissions. One possible reason for this enrichment failure is a leaking exhaust line. The leaking exhaust line imports oxygen before the oxygen sensor causing the sensor calling for more fuel from the injectors. The misfire high emitters (type3) have high engine-out HC emissions, high engine-out CO and a high CO catalyst pass fraction. These vehicles have incomplete combustion problems (misfire) and a poor catalyst performance, resulting in moderate to slightly high tailpipe CO, very high HC, and moderate to low NOx emissions. The fourth type of high emitters (type4) emits high tailpipe emissions of HC, CO, and NOx. These vehicles have chronically (burned-out or a missing catalyst) or transiently (high catalyst pass fraction) poor catalyst performance (Barth et al. 2000). In this study, the type4 high emitters which have problems with the catalytic converter are utilized to investigate all HC, CO, and NOx emissions.

In order to compare the emission levels of the high emitter and the normal vehicles, this study utilizes an On-road Emission Measurement (OEM) unit to collect field data. The OEM unit is a new technology that collects vehicle emissions using a portable instrument. The on-road emission-measurement method is considered a desirable approach for quantifying emissions from test vehicles because of its ability to collect emissions during real driving conditions; in contrast, dynamometer testing is typically utilized to check new vehicles and to inspect in-use vehicles for compliance with emissions standards.

The objectives of this paper are two-fold. First, the paper demonstrates the operational effects of catalytic converter malfunctions, investigating the conditions under which they produce high emissions. Second, the paper develops a high emitter emission model using on-board emission measurement data. The proposed emission model is validated using the collected field data from the on-road emission measurement system and utilized to identify the operational effects of catalytic converter malfunctions.

The most significant contribution of this paper is that the paper investigates the environmental impacts of catalytic converter malfunctions under various driving conditions using in-road field data. The research quantifies the additional emissions associated with catalytic converter malfunctions for various driving conditions on real-road and the U.S. Environmental Protection Agency’s (EPA) driving cycles. In particular, the research compares the performance of high emitting vehicles for all feasible driving ranges.

The paper is organized into five sections. The following section illustrates and describes the OEM unit that was utilized in the study. The third section describes the data collection procedures and data sources that were utilized to develop the proposed modeling approach. The fourth section describes the model development procedures and modeling results. The fifth section compares the emission levels of the catalytic malfunction vehicle to the normal vehicle emission rates for all feasible vehicle operational regions. Finally, the conclusions of the research and recommendations for future work are presented.

On-Board Emission Measurement Unit

A portable On-Road Emission Measurement unit, OEM-2100^TM, manufactured by Clean Air Technologies International, Inc. was utilized for data collection in this study. The OEM unit is designed to measure vehicle mass exhaust emissions, using vehicle and engine operational data and concentrations of pollutants in exhaust gas sampled from the tailpipe during actual driving on the road. The unit is comprised of two five-gas analyzers, an engine diagnostic scanner, and an on-board computer providing second-by-second emissions, fuel consumption, vehicle speed, engine rpm and temperature, throttle position, and other parameters. The unit is designed to connect with the On-Board Diagnostics (OBD) link of the vehicle, from which real-time engine and vehicle operational data may be obtained while the vehicle is in operation. Most of the 1996 year models and later are equipped with an OBD link that is located under the dashboard on the driver’s side.

The OEM unit can measure the concentrations of HC, CO, CO₂, NO_x, and O₂ in the vehicle exhaust gas by a functional equivalent of a five-gas analyzer. The vehicle exhaust gas is collected from the tailpipe using a repair-grade probe and sample line. Simultaneously, OBD data such as vehicle speed, engine rpm, intake air mass flow, coolant temperature, and other engine operating parameters are gathered. Using the intake air mass flow (or composition of intake air), measured composition of exhaust gas, and user-specified composition of fuel, a second-by-second exhaust mass flow is calculated from the on-board computer. Multiplying the exhaust mass flow by the concentrations of pollutants generates second-by-second emission data in grams. More detailed description regarding the OEM unit is found in the literature (Frey et al. 2001; Vojtisek-Lom and Allsop 2001).

The OEM unit displays the vehicle operational variables and emission data on a second-by-second basis. Each data collection of the OEM unit is also summarized as a tab-delimited formatted text file. The first section of the emission output file displays general test information such as test date, time, vehicle characteristics, driver information, weather conditions, and test site data. The second section displays a summary of each bag data. The bag data fields include the bag number, the distance traveled, the bag duration, the total exhaust flow, the fuel consumption, and HC, CO, NO_x, CO₂ emissions. Following the bag data, second-by-second emission data is displayed in the output file. Each record displays a time stamp, bag number, vehicle speed, acceleration, engine speed, throttle percent, fuel consumption, and HC, CO, NO_x, and CO₂ emissions.

Emission data collection

This section describes the data collection procedures that were utilized in comparing the emission levels of the high emitter and the normal emitting vehicle and in developing the high emitter emission models, which were collected individually. A 1999 Ford Crown Victoria was utilized for the study to gather both normal condition emissions and high emitter emissions. In order to assemble high emitting data, the test vehicle was professionally modified, adding special equipment to bypass the catalytic converters in order to mimic the behaviors of a high emitter. Detailed information about the test vehicle is provided in Table 1. Data were collected between March and April of 2002. The minimum test temperature was 5°C and the maximum was 26°C. However, most of the tests were performed at ambient temperatures that ranged between 12.8°C (55°F) to 23.9°C (75° F).

Emission data were collected on a restricted test road, the Virginia Smart Road, and on a local highway section, Route 460 in southwest Virginia, for comparison purposes. The Smart Road is a 3.0 km experimental test facility located at the Virginia Tech Transportation Institute in southwest Virginia, which is not accessible to public vehicles. The test road is fairly straight with some minor curvature that does not impact vehicle speeds and has a substantial upgrade ranging from 6% at the departing point to 2.8% at the end. High speed data collections with a wide range of speed and acceleration observations were conducted on this restricted test road, while various constant speed runs were tested on a public highway section.

The test vehicle was tested for mild, normal, and aggressive acceleration levels to assemble a wide range of speed and acceleration levels. It should be noted that the effect of upgrades was normalized by adding the component of gravitational acceleration to acceleration field measurements. Specifically, the test vehicle was accelerated from a complete stop and continued to accelerate for the duration of the test drive. All acceleration (aggressive, normal, mild) tests were repeated ten times in order to obtain multiple repetitions for various engine loads.

The emission data were also collected on Route 460 in southwest Virginia, from the Tom’s Creek Road exit to the North Main Street exit in Blacksburg, Virginia, which is a 5.9-km section. This roadway facility is designed with high geometric standards and is fairly flat and, thus, required no adjustments in vehicle accelerations for grade effects. Also, because the traffic volume was low on the roadway, the test vehicle collected emission data with minor traffic interruptions. Data were gathered for different cruise test runs that included cruising at speeds of 72, 88, and 104 km/h (45, 55, and 65 mph). As was the case with the acceleration test runs, each constant speed test run was repeated ten times in order to capture any potential variability within the data.

Figure 1 summarizes the speed/acceleration frequency distribution for the field data that were gathered. The figure demonstrates that the majority of data cover the constant speed and mild acceleration range (0 to 1.4 m/s²); however, the data also cover a wide range of speed and acceleration levels with a maximum speed of 129 km/h and a maximum acceleration of 3.6 m/s². The figure clearly demonstrates a nearly complete coverage of the speed/acceleration envelope of the test vehicle. The use of a controlled test facility enables the collection of data at the boundary of the speed/acceleration envelope of the vehicle, which is critical in developing models that are reflective of the entire operation envelope of a vehicle.

Development of high emitter Emission Models

In order to compare the operational effects of the high emitter and the normal vehicle, the study utilized microscopic emission models as well as field-collected emission data. This section describes the framework for developing microscopic high-emitting vehicle emission models using on-road second-by-second OEM measurements. The emission models for normal vehicles were developed in a previous study (Rakha et al. 2004a). It should be noted that the emission data utilized for both high emitter and normal vehicle emission modeling were collected from the same test vehicle in order to ensure identical vehicle characteristics.

The OEM data included approximately 13,480 records and each record included the time stamp of the record, the bag number, the instantaneous speed and acceleration levels, the engine speed (revolutions per minute), amount of in-take air, total exhaust flow, second-by-second fuel consumption and emissions (HC, CO, CO₂, and NO_x), and other engine-related data. The emission data were adjusted for temporal offsets which reflect the time lag between vehicle emissions and speed observations. The offset between vehicle speed and emission observations typically ranges between 5 and 15 seconds depending on the data collection procedure. The accurate estimation of the time-lag is important, because it ensures that the independent variable values are associated with their corresponding instantaneous vehicle emission rates. Secondly, the OEM data were aggregated and averaged within a speed and acceleration bin in order to reduce the noise in the data. The speed bins ranged from 0 to 129 km/h at increments of 1 km/h, while the acceleration bins ranged from -10 to 13 km/h/s at increments of 1 km/h/s.

The proposed high emitter emission model utilized a logarithmic transformation of a dual-regime third order polynomial model, as summarized in Equation 1. This multiple regression model (VT-Micro model) relates the dependent variables (instantaneous emission estimates) to a set of quantitative independent variables, namely instantaneous speed and acceleration levels. The derivation of the final models involved experimentation with numerous polynomial combinations of speed and acceleration levels. Specifically, linear, quadratic, cubic, and quartic terms of speed and acceleration were investigated. The final regression models included a combination of linear, quadratic, and cubic speed and acceleration terms because it provided the least number of terms with a relatively good fit to the data. A more detailed description of the derivation of the model is provided in the literature (Melosh et al. 1990; Rakha et al. 2000).

The use of polynomial speed and acceleration terms may result in multi-colinearity between the independent variables as a result of the dependency of these variables. The Variance Inflation Factor (VIF), which is a measure of multi-colinearity, can be reduced by removing some of the regression terms with, however, a reduction in the accuracy of the model predictions. Consequently, a trade-off between reducing the model multi-colinearity should be weighed against a potential reduction in model accuracy. The existence of multi-colinearity results in unreliable model estimations when the dependent variables are outside the bounds of the original data. Consequently, the model was maintained with the caveat that the model should not be utilized for data outside the feasible envelope of a typical vehicle.

[1]

Where;

MOE_e	= instantaneous fuel consumption or emission rate (l/s or mg/s),
L^e_i,j	= Model regression coefficient for MOE “e” at speed power “i” and acceleration power “j” for positive accelerations,
M^e_i,j	= Model regression coefficient for MOE “e” at speed power “i” and acceleration power “j” for negative accelerations,
u	= Instantaneous speed (km/h), and
a	= Instantaneous acceleration (km/h/s).

While most vehicles may be able to travel faster than 130 km/h (80 mph), the upper limit of the testing boundary, it is highly unlikely to observe speeds outside this range within typical driving cycles. Consequently, the envelope of data coverage appears to cover the full range of typical vehicle operation. However, in these cases, the authors recommend using boundary speed and acceleration levels, which are illustrated in Figure 2, in order to ensure realistic vehicle MOE estimates. It should be noted that the models were confined to speed and acceleration levels within the envelope of the OEM data. Figure 3 illustrates graphically the quality of fit between the model predictions and the field measurements. Each raw data point in Figure 3 represents an average value of all emission measurements within a speed/acceleration bin, while each line represents different acceleration levels (–3, 0, 3, and 6 km/h/s or -0.83, 0, 0.83, and 1.67 m/s²). The figure demonstrates a reasonable fit between the model and the field data. Specifically, the prediction lines accurately follow the average of the field measurements for HC, CO, NOx, and CO₂ emissions.

Effect of High emitters

This section compares the emission levels of the high emitter and the normal vehicle collected from real roads and from simulation studies which utilized the proposed emission models to demonstrate the operational effects of the high emitter. It should be noted that the emissions of the high emitter and the normal vehicle were collected in almost the same ambient conditions including temperature, altitude, and degree of humidity. Also, two datasets were collected at the same location (Route 460 and Smart Road) using the same test vehicle during the same period of time.

Emission Comparison using Field Data

Figure 4 illustrates the emission levels of the high-emitting vehicle and the normal vehicle at various cruising speeds. The figure clearly shows that during various cruising speeds, the high emitter produces significant amounts of emissions compared to the normal vehicle. In particular, the high emitter emits 269%, 297%, and 273% more HC emissions than the normal vehicle at the cruising speed operations of 72 km/h, 88 km/h, and 104 km/h, respectively. In case of CO emissions, the high emitter increases the emissions by up to 211% compared to the normal vehicle at the speed of 88 km/h. Also, it should be noted that CO emission levels were drastically increased when the cruise speeds were increased from 82 km/h to 104 km/h for both the high emitter and the normal vehicle. Also, the high emitter CO emission level at 104 km/h was dramatically increased in terms of the absolute emission amount. The figure demonstrates that the high emitter increases NOx emissions significantly (between 130% and 378%) during cruising driving conditions. Besides HC, CO, and NOx emissions, Figure 4 also illustrates fuel consumption and CO₂ emission data. As expected, the high emitter did not increase the CO₂ emissions and slightly raises the fuel economy.

Figure 5 also shows the emission levels for high-emitting and normal vehicles at various acceleration levels. The figure demonstrates that, similar to the cruising speed tests, the high emitter produces more emissions than the normal vehicles with average emission increases of 63.4%, 26.9%, and 76.1% for HC, CO, and NOx emissions, respectively. Particularly, the field test revealed 75.8%, 56.2%, and 58.1% of HC emission increases at mild, normal, and aggressive acceleration levels when a high-emitting vehicle was operated. However, the field test demonsrates a different result in the case of CO; the high emitter produces more emissions for mild and normal tests (55.0% and 29.3%), but reduces CO emission by 3.5%. It should be noted that both HC and CO emissions were significantly increased during the aggressive acceleration tests for both the high emitter and the normal vehicle. Thus, it is considered that under extremely aggressive driving conditions, even normal vehicles produce very high CO emission rates. Equal mean t-tests showed that the high emitter CO emission values during aggressive acceleration tests are not significantly different from the CO emissions from the normal vehicle. The figure also demonstrates an increase in NOx emissions (71.2%, 75.9, and 81.2% for mild, normal, and aggressive acceleration levels, respectively) when the catalytic converter is not operating. The NOx emission results illustrate that more NOx emissions are produced at mild acceleration conditions for both the high emitter and the normal vehicle. This observation is explained by the fact that more NOx emissions are produced under stoichiometric air/fuel ratios. The figure similarly demonstrates that the catalytic converter malfunction does not increase the CO₂ emissions or impact the vehicle’s fuel economy.

Figure 6 illustrates the effects of acceleration operations for both the high emitter and normal vehicle. The figure compares the total emissions and fuel consumption rates of the 104 km/h speed cruise operation test and the various acceleration test results. The figure shows that the acceleration operations are a significant contributor to high emissions and fuel consumptions, producing up to 1,090%, 2,836%, 241%, 116%, and 617% of extra HC, CO, NO_x, CO₂, and fuel consumption, respectively. In particular, during aggressive acceleration operations, the CO emissions produce 28.36 times more CO emissions compared to constant speed operations. In general, normal vehicles are impacted more by excessive acceleration levels.

Emission Comparison using Emission Models

This section quantifies the environmental impact of catalytic converter malfunctions using the models that were developed based on the in-field emission data collection effort. A detailed description of the normal emission model which utilizes the same regression modeling technique is described in the literature (Rakha et al. 2004b).

Figure 7 illustrates the differences between the estimated HC, CO, and NOx emission values of the high emitter model and the normal vehicle model. The figure clearly demonstrates that the high emitter produces significantly more emissions than the normal vehicle in most cases. It is noted that the shapes of the high-emitter emissions are very similar to the shapes of the normal vehicle emissions. However, the absolute values of high-emitter emissions and normal vehicle emissions are notably different. In particular, at the -3 and 0 km/h/s acceleration levels, as vehicle speed increases, the absolute differences of HC and CO emissions are on the increase. However, in the case of high acceleration levels (3 and 6 km/h/s, the absolute differences of the high emitter emissions and normal vehicle emissions decreases as vehicle speed increases. Specifically, in the case of CO emissions, the normal vehicle produces CO emission rates that are similar in magnitude to high emitter rates.

Figure 8 illustrates the relative emission differences of the high emitter and the normal vehicle, which was estimated from the proposed simulation models for HC, CO, and NO_x emissions. It should be noted that the contour plot was generated only for the feasible regime in Figure 2. As shown in the figure, the high emitter has significant impacts, which produce increases in vehicle emissions of HC, CO, and NO_x in the range of 400%, 500%, and 800%, respectively. It is noted that the most severe increases in HC and CO emissions are observed in the speed range of 45 to 85 km/h at mild acceleration levels. In the case of NO_x, the high emitter produces much higher NO_x emissions in the speed range of 20 to 55 km/h operational regime at high acceleration levels.

Figure 9 shows the simulation results of sixteen driving cycles for the normal vehicle and high emitter. The sixteen facility-specific and area-wide driving cycles were developed by the U.S. EPA during the process of development for MOBILE models and emission validation and inspection for vehicles based on real-world driving studies. In order to represent real-world driving conditions, the driving behavior in each drive cycle was developed using observed speed-acceleration profiles and specific power frequency distributions of chase car data. These drive cycles cover four roadway types that include: Freeways, Arterials/Collectors, Freeway Ramps, and Local Roadways. The additional cycles include the LA4 (urban dynamometer driving cycle), the California Air Resources Board (CARB) area-wide Unified Cycle (LA92), the New York City Cycle (a low speed cycle which was previously used for speed correction factors in the MOBILE5 model), and ST01 (the first 258 seconds of the vehicle certification air conditioning cycle). The maximum speed of all cycles was 119.5 km/h on the Freeway High Speed Cycle, while the maximum acceleration rate was 11.04 km/h/s (3.07 m/s2) on the Freeway, LOS F and LA92 cycles. It should be noted that the New York cycle had the lowest average speed of all drive cycles (11.4 km/h) (Brzezinski et al. 1999; Carlson and Austin 1997).

The study simulated the sixteen drive cycles in order to represent real-world driving conditions using the proposed emission models in Figure 9. The high emitter produces overall 251%, 225%, and 336% (0.71 g/km, 4.50 g/km, and 1.41 g/km) more HC, CO, and NOx emissions, respectively, when compared to a normal vehicle across all sixteen driving cycles. In particular, the figure illustrates that the New York City Cycle produces the highest emissions among the sixteen driving cycles for the high emitter and the normal vehicle for HC and CO emissions, while freeway cycles generally emit lower emissions. However, the relative difference between normal and high emitters for the New York City Cycle emissions was one of lowest among the 16 cycles with increases in the range of 225%, 193%, and 264% for HC, CO, and NOx emissions, respectively. It is noted that the high emitting vehicle has the most significant impact on the Freeway Ramp cycle for HC and CO emissions with a 271% and 285% increase and Freeway LOS D Cycle for NOx emission with a 374% increase. The study also demonstrates that high speed cycles (e.g., Freeway High Speed, Freeway LOS A-C cycle) have fewer negative effects on high emitters for HC and CO emissions with low relative difference rates in comparison to other cycles.

Figure 10 illustrates the variations in the instantaneous vehicle emission estimates of HC, CO, and NOx for the high emitter and the normal vehicle emissions based on instantaneous vehicle speed and acceleration levels for the Arterial LOS A cycle. The figure also shows the instantaneous speed profiles of the ARTA drive cycle, which involves several full and partial stops in addition to travel at a fairly high speed (in the range of 100 km/h). The figure demonstrates that the high emitter produces significant emission increases in the range of 264%, 260%, and 364% for HC, CO, and NOx emissions, respectively, along the ARTA drive cycle. The figure demonstrates that the shape of emission estimates are very similar. However, it is noted that when the normal vehicle emissions are high, the emission differences between the high emitter and the normal vehicle become larger. Instantaneous relative differences along the ARTA cycle are also illustrated in the figure showing that high emitters produce as much as 388%, 371%, and 564% increases in emissions compared to normal vehicles, depending on the vehicle speed and acceleration levels. The figure illustrates that the high emitter has more negative effects on NOx emissions than HC and CO emissions in this case study.

Study Conclusions

The paper quantified the negative environmental impacts of catalytic converter malfunctions using field-collected and simulated data. In addition, as part of the study microscopic emission models were developed for vehicles with a malfunctioned catalytic converter using on-road emission measurements. The proposed emission models were utilized to estimate the negative environmental impacts of high emitters at various driving conditions.

The field results demonstrate that vehicles with catalytic converter malfunctions produce increases in vehicle emissions in the range of 297%, 211%, and 378% for HC, CO, and NOx emissions, respectively, while cruising. Furthermore, increases in HC, CO, and NOx in the range of 63.4%, 26.9%, and 76.1% were observed during acceleration tests. The analysis demonstrated that both HC and CO emissions were significantly increased during aggressive acceleration tests for both the high emitter and the normal vehicles. Thus, under extremely aggressive driving conditions, normal vehicles produce very high CO emissions (28 times the normal CO emissions) and behave similar to high emitters.

The simulation study demonstrated that the most significant impacts of high emitters are observed in the speed range from 45 to 85 km/h under mild acceleration levels in the case of HC and CO emissions and between 20 and 55 km/h at high acceleration levels for NOx emissions. Vehicles with malfunctioned catalytic converters on average increase HC, CO, and NOx emissions in the range of 251%, 225%, and 336%, respectively, over the sixteen EPA driving cycles. The New York City Cycle produces the highest emissions among the sixteen driving cycles for the high emitter and the normal vehicles for HC and CO emissions, while the freeway cycles generally emit lower emissions. The study also found that high speed cycles (e.g., Freeway High Speed, Freeway LOS A-C cycle) have fewer negative effects on high emitters for HC and CO emissions, with low relative difference rates compared to other cycles.

References

Barth, M., An, F., Younglove, T., Scora, G., Levine, C., Ross, M., and Wenzel, T. (2000). Comprehensive modal emission model (CMEM), version 2.0 user's guide, Riverside, CA.

Benedek, C. M., and Rilett, L. R. (1998). "Equitable traffic assignment with environmental cost functions." Journal of Transportation Engineering-Asce, 124(1), 16-22.

Brzezinski, D., Enns, P., and Hart, C. (1999). "Facility-Specific Speed Correction Factors." MOBILE6 Stakeholder Review Document (M6.SPD.002), US EPA, Ann Arbor, Michigan.

Carlson, T. R., and Austin, T. C. (1997). "Development of Speed Collection Cycles." MOBILE6 Stakeholder Review Document (M6.STD.001), EPA, Prepared by Sierra Research, Ann Arbor, Michigan.

Frey, H. C., Rouphail, N. M., Unal, A., and Colyar, J. "Measurement of On-Road Tailpipe CO, NO, and Hydrocarbon Emissions Using a Portable Instrument." Annual Meeting of the Air & Waste Management Association, Orlando, Florida.

McClintock, P. (1999). "Remote Sensing Measurements of Real World High Exhaust Emitters." Coordinating Research Council, Inc., Tucson, AZ.

Melosh, H. J., Schneider, N. M., Zahnle, K. J., and Latham, D. (1990). "Ignition of global wildfires at the Cretaceous/Tertiary boundary." Nature, 343, 251-254.

Miller, T. S., Tompkins County (N.Y.) Dept. of Planning, and U.S. Department of the Interior. (2000). "Unconsolidated aquifers in Tompkins County, New York." U.S. Geological Survey, Branch of Information Services, [Reston, Va.?].

Nagurney, A. (2000). "Congested urban transportation networks and emission paradoxes." Transportation Research, Part D: Transport & Environment, 5(2), 145-151.

Nagurney, A., and Dong, J. (2002). "A multiclass, multicriteria traffic network equilibrium model with elastic demand." Sage Urban Studies Abstracts, 30(4), 415-517.

Nagurney, A., Ramanujam, P., and Dhanda, K. K. (1998). "Multimodal traffic network equilibrium model with emission pollution permits: Compliance vs noncompliance." Transportation Research, Part D: Transport & Environment, 3(5), 349-374.

Rakha, H., and Ahn, K. (2004). "Integration modeling framework for estimating mobile source emissions." Journal of transportation engineering, 130(2), 183-193.

Rakha, H., Ahn, K., Ihab, S., and Sebong, J. "Emissions Model Development using In-vehicle On-board Emission Measurements." Presented at 83rd Annual Meeting of the Transportation Research Board, Washington, D.C., Paper 04-3410.

Rakha, H., Ahn, K., and Trani, A. (2004b). "Development of VT-Micro model for estimating hot stabilized light duty vehicle and truck emissions." Transportation Research Part D-Transport and Environment, 9(1), 49-74.

Rakha, H., Van Aerde, M., Ahn, K., and Trani, A. A. (2000). "Requirements for evaluating traffic signal control impacts on energy and emissions based on instantaneous speed and acceleration measurements." Transportation Research Record. n 1738 2000, 56-67 00-1133.

Rilett, L. R., and Benedek, C. M. (1994). "Traffic assignment under environmental and equity objectives." Transportation Research Record(1443), 92-99.

Schrank, D., and Lomax, T. (2005). "The 2005 Urban Mobility Report." MOBILE6 Stakeholder Review Document (M6.EXH.009), Texas Transportation Institute, Texas.

Sugawara, S., and Niemeier, D. A. (2002). "PART 2 - AIR QUALITY - How Much Can Vehicle Emissions Be Reduced? Exploratory Analysis of an Upper Boundary Using an Emissions-Optimized Trip Assignment." Transportation research record(1815), 9.

Tzeng, G. H., and Chen, C. H. (1993). "Multiobjective Decision-Making for Traffic Assignment." Ieee Transactions on Engineering Management, 40(2), 180-187.

Vojtisek-Lom, M., and Allsop, J. E. "Development of Heavy-Duty Diesel Portable, On-Board Mass Exhaust Emissions Monitoring System With Nox, Co2 and Qualitative Pm Capabilities." SAE International Fall Fuels & Lubricants Meeting & Exhibition, Session: Real-World Emissions Measurement Technology for Internal Combustion Systems, San Antonio, TX, USA,.

Wenzel, T., and Ross, M. "Characterization of recent-model high-emitting automobiles." Proceedings of the 1998 SAE International Spring Fuel & Lubricants Meeting & Exposition, Dearborn, MI, USA, 51-66 981414.

Wenzel, T., Singer, B. C., and Slott, R. (2000). "Some Issues in the Statistical Analysis of Vehicle Emissions." Journal of Transportation Statistics, 3 (2).

Wolfe, J. A. (1990). "Palaeobotanical evidence for a marked temperature increase following the Cretaceous/Tertiary boundary." Nature, 343, 153-156.

List of Tables

Table 1. Test Vehicle Characteristics

List of Figures

Figure 1. Speed and acceleration frequency distribution

Figure 2. Speed and acceleration feasible range of application

Figure 3. High emitter emission model predictions

Figure 4. Effects of high emitter at various cruise speeds

Figure 5. Effects of high emitter at various acceleration levels

Figure 6. Effects of various acceleration operations

_{Figure 7. Microscopic emission comparisons using the
estimates of emission models}

Figure 8. Relative emission differences

Figure 9. Emission rates for 16 driving cycles

Figure 10. Instantaneous Emissions for ARTA driving cycle

TABLE 1 Test Vehicle Characteristics

Make	Ford
Model Year	1999
Model	Crown Victoria
VIN	2FAFP73W6XX191763
Mileage	13,000
Transmission	Automatic
Gross Weight	3741 lb.
No. Cylinders	8
Engine Size	4.6
Engine Type	V8
Horse Power	200 hp @ 4250 rpm
Fuel Capacity	19 gal