Federal Test Procedure Review Project: Preliminary Technical Report TABLE OF CONTENTS AND EXECUTIVE SUMMARY May 1993 EPA 420-R-93-007 Certification Division Office of Mobile Sources Office of Air & Radiation U.S. Environmental Protection Agency DISCLAIMER: Although the information described in this preliminary technical report has been funded in part by the United States Environmental Protection Agency under contracts to Radian Corporation and to Sierra Research, Inc., it has not been subjected to the Agency's peer and administrative review processes. It is being released for information purposes only and could be used in potential regulation development. Table of Contents Executive Summary. . . . . . . . . . . . . . . . . . . . . . . 1 Chapter 1. Introduction . . . . . . . . . . . . . . . . . . . 8 Chapter 2. Background . . . . . . . . . . . . . . . . . . . . 10 2.1. The Air Quality Problem. . . . . . . . . . . . . . . 10 2.2. History of the Federal Test Procedure. . . . . . . . 11 2.3. Statutory Provision. . . . . . . . . . . . . . . . . 14 2.4. Areas of Potential Concern . . . . . . . . . . . . . 14 2.4.1. Fuel . . . . . . . . . . . . . . . . . . . 14 2.4.2. Temperature. . . . . . . . . . . . . . . . 16 2.4.3. Altitude . . . . . . . . . . . . . . . . . 17 2.4.4. Driving Behavior (Including Acceleration). . . . . . . . . . . . . . . 18 2.4.5. Test Procedure Modifications . . . . . . . 20 2.5 Heavy-Duty Vehicles and Engines. . . . . . . . . . . 20 Chapter 3. Project Overview . . . . . . . . . . . . . . . . . 23 3.1. Research on Driving Behavior . . . . . . . . . . . . 24 3.2. Emission Assessment of In-Use Driving. . . . . . . . 27 3.2.1. Cycle Development. . . . . . . . . . . . . 27 3.2.2. Emission Simulation Model. . . . . . . . . 29 3.2.3. Vehicle Testing Programs . . . . . . . . . 31 3.3. Notice of Proposed Rulemaking Development. . . . . . 31 Chapter 4. Driving Survey Methods . . . . . . . . . . . . . . 33 4.1. Survey Options . . . . . . . . . . . . . . . . . . . 33 4.1.1. Choosing Survey Approach . . . . . . . . . 33 4.1.2. Selecting Survey Sites . . . . . . . . . . 35 4.2. Instrumented Vehicle Approach. . . . . . . . . . . . 36 4.2.1. Key Features . . . . . . . . . . . . . . . 36 4.2.2. Instrumented Vehicle Field Results . . . . 39 4.2.3. Atlanta Instrumented Vehicle Study . . . . 42 4.3. Chase Car Approach . . . . . . . . . . . . . . . . . 43 4.3.1. Key Features . . . . . . . . . . . . . . . 44 4.3.2. Chase Car Field Results. . . . . . . . . . 46 4.3.3. Los Angeles Chase Car Study. . . . . . . . 47 4.4. Analyses of Potential Bias . . . . . . . . . . . . . 47 4.4.1. Analysis of Instrumented Vehicle Method. . 48 4.4.2. Analysis of Chase Car Method . . . . . . . 52 4.5. Selection of Principal Data Set. . . . . . . . . . . 55 4.5.1. Comparison of the Two Surveys Results. . . 55 4.5.2. Rationale for Choosing 3-parameter Instrumented Vehicle data. . . . . . . . . 58 Chapter 5. Analytical Methods and Considerations. . . . . . . 60 5.1. Driving Behavior Variables . . . . . . . . . . . . . 60 5.1.1. Speed-Based Measures of Driving Behavior . . . . . . . . . . . . . . . . . 60 5.1.2. Trip-based Measures. . . . . . . . . . . . 62 5.1.3. Vehicle-Based Measures . . . . . . . . . . 65 5.2. Descriptive Methods. . . . . . . . . . . . . . . . . 66 5.3. Statistical Accuracy . . . . . . . . . . . . . . . . 68 5.4. Computer Resource Issues . . . . . . . . . . . . . . 69 5.5. Driving Conditions . . . . . . . . . . . . . . . . . 70 5.5.1. Engine and Catalyst Cooldown . . . . . . . 70 5.5.2. Road Grade . . . . . . . . . . . . . . . . 75 Chapter 6. Discussion and Analysis of Driving Survey Results . . . . . . . . . . . . . . . . . . . . 77 6.1. Summary of In-Use Driving Results for Four Cities . . . . . . . . . . . . . . . . . . . . . . . . . . 77 6.1.1. Driving Behavior - Second-by-Second Analysis . . . . . . . . . . . . . . . . . 77 6.1.2. Driving Behavior - Trip Analysis . . . . . 82 6.1.3. Choosing Representative Survey Data. . . . 83 6.2. Analysis of Baltimore Driving Conditions . . . . . . 87 6.2.1. Speed-based Measures of Driving Behavior . . . . . . . . . . . . . . . . . 87 6.2.2. Trip Measures. . . . . . . . . . . . . . . 98 6.2.3. Vehicle-based Measures . . . . . . . . . .108 6.2.4. Vehicle Soak . . . . . . . . . . . . . . .109 6.2.5. Trip Start Driving Activity. . . . . . . .115 6.3. Comparisons to the Federal Test Procedure. . . . . .127 6.3.1. Speed-based Measures . . . . . . . . . . .127 6.3.2. Trip Comparison. . . . . . . . . . . . . .140 6.3.3. Vehicle Soak . . . . . . . . . . . . . . .143 6.3.4. Trip Start Driving Activity. . . . . . . .145 6.3.5. Road Grade . . . . . . . . . . . . . . . .154 Chapter 7. Test Cycle Development Methods and Approach. . . .155 7.1. Test Cycle Methods . . . . . . . . . . . . . . . . .155 7.2. Method Types . . . . . . . . . . . . . . . . . . . .156 7.2.1. Segment-splicing Methods . . . . . . . . .156 7.2.2. Monte Carlo Simulation . . . . . . . . . .157 7.2.3. Engineering Cycles . . . . . . . . . . . .158 7.3. Cycle Criteria . . . . . . . . . . . . . . . . . . .159 7.4. Cycle Validation . . . . . . . . . . . . . . . . . .159 7.5. Current EPA Test Cycle Development . . . . . . . . .160 List of Appendixes . . . . . . . . . . . . . . . . . . . . . .161 Appendix A. Chase Car Method Bias Analysis: Supplementary Tables Appendix B. Baltimore 3-Parameter Vehicle Characteristics Appendix C. Summary Statistics, Distributions, and Graphs Appendix D. Baltimore Soak Periods Appendix E. Trip Start Driving Activity Executive Summary Pursuant to Section 206(h) of the Clean Air Act, as amended in 1990 (Act, or CAA), EPA has undertaken a review of the Federal Test Procedure (FTP) used to test light duty vehicle and light duty truck emissions to determine whether it adequately represents actual current driving conditions. The driving cycle used for the FTP was adopted over twenty years ago and accumulated research suggests that it may no longer adequately represent overall vehicle emission control performance under current driving conditions. Project Overview The principal subject of this report is driving behavior, including acceleration and trip start patterns. After reviewing existing research on driving behavior, the Agency determined that new surveys were needed to assess current driving in U.S. urban nonattainment regions. A parallel study of current technology vehicle emissions under the full range of driving conditions is also in progress. Results from these efforts will be combined in determining the need for test procedure revisions. This preliminary technical report discusses how the driving surveys were conducted, presents analyses of the results, and compares the data to the existing FTP. Quantitative assessments of the emission impacts are still in progress and are not discussed in this report. With support from the American Automobile Manufacturers Association (AAMA) and the Association of International Automobile Manufacturers (AIAM), EPA conducted surveys of driving behavior in Baltimore, MD, and Spokane, WA. Two methods of data collection were employed. In an instrumented vehicle study, 113 Baltimore and 102 Spokane vehicles were equipped with "3- parameter" datalogger packages that recorded second-by-second speed and two other variables during periods of operation. As part of the same surveys, the manufacturers recruited 79 vehicles for study using "6-parameter" instruments designed to measure additional variables. The instrumented vehicles were observed for seven to ten day periods. A separate chase car study collected similar speed data in the two cities using a laser device mounted on a patrol car that tracked in-use target vehicles. This produced relatively short sequences of data on a much larger sample of vehicles. The Baltimore and Spokane surveys are supplemented by data collected in two other cities. EPA's Office of Research and Development sponsored an instrumented vehicle study in Atlanta, GA. Finally, the California Air Resources Board (CARB) sponsored a chase car study in Los Angeles similar to the chase car studies in Spokane and Baltimore. For reasons relating to representativeness, availability, and precision of the survey data, most of the discussion in this report is confined to driving observed in the Baltimore 3- parameter instrumented vehicle study. Participants in the instrumented vehicle study in Baltimore were recruited from two different centralized Inspection/Maintenance stations, one in central Baltimore and one in a suburb just outside Baltimore. The urban site yielded driving characteristics similar to those in Spokane and the suburban site was similar to Atlanta and Los Angeles. As it was not possible to obtain detailed analyses of the Atlanta and Los Angeles data in time for this report, EPA decided to use just the Baltimore data for purposes of this report. This should yield a more representative picture than also including the Spokane data. Future research will include further analyses of data obtained from the other sources. EPA also entered into a cooperative agreement with the New York State Bureau of Air Research to obtain test data and analyses of the engine and catalyst cooldown processes and factors affecting the cooldown rates. These data are used to assess the condition of the engine and catalyst during engine start-up. Preliminary Indications of Actual Driving Behavior Speed and Acceleration Speeds were much higher in Baltimore than are represented on the FTP. The average speed in Baltimore was 24.5 mph (median speed was 23.7). The speeds observed ranged to almost 95 mph; 6.4% were above 60 mph and 2.6% above 65 mph. By comparison, the FTP has an average speed of 19.6 mph with a maximum of 56.7 mph. About 8.5% of all speeds in Baltimore exceeded the FTP maximum. Acceleration rates in Baltimore were also significantly higher than those on the FTP. The acceleration rates observed ranged up to 15 mph/sec, with a standard deviation of 1.5. The FTP has a maximum acceleration rate of 3.3 mph/sec and a standard deviation of 1.4. About 2.5% of all driving in Baltimore exceeded 3.3 mph/sec. Power-related measures also indicate that the observed driving behavior was more aggressive than the FTP. Specific power for the Baltimore sample ranged up to 558 mph2/sec and averaged 46.0, with a median of 34.7. The FTP has a maximum power of 192, average of 38.6, and median of 21.6. An analysis was also done of the scatter of speed-acceleration points occurring in the Baltimore sample outside the FTP envelope of speed and accelerations. These points represent about 18% of total Baltimore driving time. Driving Behavior Determinants Vehicle Type - Speed distributions were fairly similar for each of the three categories analyzed; trucks, sedans, and high performance vehicles. However, high performance vehicles demonstrated more aggressive driving behavior than the other classes, with over twice as much operation at power levels above 200 mph2/sec. Vehicle Age - Newer vehicles (1983 and later) had higher average speeds than older vehicles (25.1 mph v. 21.2 mph), were driven somewhat longer and farther per day, and averaged fewer trips and slightly fewer stops per mile. The data indicate that newer vehicles spend more time at high speeds and are used for longer trips than older vehicles. However, analyses of the aggressiveness of the driving behavior, as measured by acceleration and power distributions, indicate very little difference between older and newer vehicles. Time of Day - Average speeds were lowest during the evening rush period of 4-7 pm (23.0 mph) and highest during the morning rush period of 6-9 am and night driving period of 9 pm to 1 am (25.4 mph). Extremes of acceleration and specific power were not highly associated with time of day. About twice as many trips began during the evening rush period as during the morning rush period. The morning peak period had the longest average trip lengths (6.8 miles), while mid-day trips averaged only 4.4 miles. Time of Week - Average weekend speeds were substantially higher than on weekdays; 26.3 mph and 23.9 mph, respectively. This is due, in part, to increased high speed driving; 13.6% of all driving during the weekend exceeded 55 mph, compared to 9.8% during weekdays. Weekend driving also produced substantially higher average time, distance, and number of trips than weekdays. However, acceleration and specific power measures do not indicate that weekend driving is more aggressive than on weekdays. Trip Patterns Average in-use trip lengths are much shorter than the FTP, which represents a 7.5 mile trip. The average observed trip covered 4.9 miles. The median value of trip distance indicates that "typical" trips are even shorter, only 2.5 miles. One of the in-use impacts of shorter trips is that a much higher proportion of overall driving is done within 0.67 mile of vehicle starts (12.0% v. 8.9% on the FTP), prior to engines and catalysts reaching normal operating temperatures. The frequency of stops on the FTP is also uncharacteristic of in-use trips; the average distance between stops on the FTP is only 0.41 miles compared to 0.87 in Baltimore. Despite these differences, the FTP and Baltimore trips disagree only slightly in the proportion of time spent in the four operating modes: idle, cruise, acceleration, and deceleration. Vehicle Soaks The in-use data contains a large proportion of intermediate soak periods (that is, the time between the end of a previous trip and the beginning of the next one) that are not reflected on the FTP. The FTP contains soak periods of 10 minutes and 12-36 hours; almost 40% of all soak periods in Baltimore were between 10 minutes and 2 hours. As catalysts cool off much faster than engines and most are almost completely cold in about 45-60 minutes, this is a potential emission concern. Analyses indicate that only about 30% of all in-use starts occur with catalysts hot enough to be immediately effective; the FTP implicitly assumes that 57% of all starts occur with hot catalysts. On the other hand, the FTP implicitly assumes that 43% of all starts occur with cold engines, while less than 25% of in-use starts occur with cold engines. Trip Start Driving Activity While the FTP has lower speeds and is less aggressive than in-use driving behavior, overall, the reverse occurs for the first few minutes after a vehicle start. The average observed speed during the first 80 seconds of all trips (the initial idle period was not included in this period) was only 14.4 mph, compared to 23.1 mph for the first micro-trip on the FTP. The average in-use speed 81-240 seconds into the trip was 22.8 mph, compared to 29.8 for a comparable period on the FTP. The aggressiveness of the FTP was also off substantially, with the first micro-trip on the FTP substantially less aggressive than in-use driving and the second micro-trip greatly overaggressive. Under similar ambient air temperature conditions, the initial idle time on the FTP after a cold start is similar to observed data. However, after a hot start the initial idle time on the FTP is much longer than observed data. The FTP uses an initial idle time of 20 seconds after both cold and hot starts; observed initial idle times after cold starts averaged 28 seconds with a median of 9 seconds, while initial idle times after hot starts averaged only 12 seconds with a median of 5 seconds. Emission Impact Assessment Plans The analysis of data obtained from the driving surveys described here indicates significant differences exist between actual driving behavior and the FTP. The driving survey data will serve as the primary input into programs to assess the difference between emissions predicted by the FTP and emissions that occur in actual driving. This assessment requires the development of driving cycles that are more representative of the driving behavior information obtained from the surveys. EPA, in cooperation with the California Air Resources Board, has investigated cycle generation alternatives and developed improved methods. Cycles generated from the driving survey data using these methods will be used in test programs to quantify in-use emissions.