Remote Sensing of Mobile Source Air Pollutant Emissions: Variability and Uncertainty in On-Road Emissions Estimates of Carbon Monoxide and Hydrocarbons for School and Transit Buses Report No. FHWY/NC/97-005
Prepared by: H. Christopher Frey and David A. Eichenberger Center for Transportation Engineering Studies Department of Civil Engineering North Carolina State University Raleigh, NC 27695-7908
Prepared for: Division of Highways North Carolina Department of Transportation P.O. Box 25201 Raleigh, NC 27611-5201
June 1997
Table of Contents List of Figures
v
List of Tables
ix
Acknowledgments
xiii
EXECUTIVE SUMMARY 1.0
2.0
3.0
ES-1
INTRODUCTION
1
1.1
Mobile Source Emissions
2
1.2
Emission Regulations
2
1.3
Emissions Contributions of "Non-Controlled" Vehicles
3
1.4
Conventional Approaches To Estimating Motor Vehicle Emissions
5
1.5
Remote Sensing of On-Road Emissions
6
REMOTE SENSING: THEORY AND OPERATION
10
2.1
Remote Sensing Device (RSD) Equipment
10
2.2
Non-Dispersive Infrared Absorption
16
2.3
Data Acquisition
17
2.4
Calibration
19
2.5
Remote Sensor Accuracy
20
2.6
System Setup and Operation
22
SELECTION OF FLEETS FOR REMOTE SENSING OF BUS EMISSIONS24 3.1
Bus Fleets Identified For Study
24
3.2
School Bus Fleet Characterization 3.2.1 Vehicle Numbers/Fuel Classification 3.2.2 Passenger Loadings 3.2.3 Miles Traveled 3.2.4 Fuel Consumed 3.2.5 Summary of County Data
25 25 28 29 31 32
i
4.0
5.0
3.2.6 Area of Investigation 3.2.7 School Bus Maintenance Records
32 33
3.3
Triangle Transit Authority
33
3.4
Raleigh-Durham Airport Authority Bus Fleet
34
3.5
City of Raleigh Capital Area Transit Buses
34
SELECTION OF REMOTE SENSING MEASUREMENT SITES
36
4.1
Site Selection Strategies
36
4.2
Site Selection Criteria
37
4.3
School Bus Site Selection
38
4.4
TTA RSD Transit Bus Site Selection
43
4.5
Raleigh-Durham International Airport Authority RSD Transit Bus Site 44
4.6
City of Raleigh CAT Transit Bus Sites
44
MOTOR FUELS AND EMISSION FACTORS
46
5.1
Factors Affecting Pollutant Emissions from Highway Vehicles
46
5.2
Production of Vehicle Fuels
48
5.3
Gasoline 5.3.1 Antiknock Quality 5.3.2 Volatility 5.3.3 Hydrocarbon Composition 5.3.4 Sulfur Content 5.3.5 Density 5.3.6 Heating Value 5.3.7 Additional Considerations
50 50 52 53 55 55 56 56
5.4
Diesel Fuel 5.4.1 General Characteristics of Diesel Fuel 5.4.2 Ignition Quality 5.4.3 Additives 5.4.4 Classifications 5.4.5 Hydrocarbon Composition
57 58 58 59 59 59
5.5
Derivation of Emission Factors 5.5.1 Simplified Combustion Model 5.5.2 Emission Factors on a Grams Per Gallon Basis
60 60 64
ii
5.5.3 Emission Factors on a Grams per Vehicle Mile Basis 5.5.4 Bias in the Hydrocarbon Emission Factor 6.0
7.0
REMOTE SENSING MEASUREMENTS AND ESTIMATED EMISSION FACTORS FOR SCHOOL BUSES 68 6.1
Data Collection and Database Development Activities
68
6.2
Summary of School Bus Emissions Estimates
70
6.3
Fuel Economy Data for School Buses Observed at the Rock Quarry Road Site 75
6.4
Diesel School Buses at the Rock Quarry Road Site 6.4.1 CO and HC Emissions for All Observed Buses 6.4.2 Individual Buses 6.4.3 Effect of Bus Characteristics on Emissions 6.4.4 Correlation Between CO and Hydrocarbon Emissions 6.4.5 Effect of Vehicle Direction on Emissions 6.4.6 Summary
6.5
Gasoline-Fueled Buses at the Rock Quarry Road Site 98 6.5.1 CO and HC Emissions for All Observed Gasoline-Fueled School Buses 98 6.5.2 Effect of Bus Characteristics on Emissions 101 6.5.3 Effect of Vehicle Direction on Emissions 105
6.6
Gasoline-Fueled School Buses at Other Sites
105
6.7
Diesel School Buses at Other Sites
108
76 76 79 85 90 90 94
REMOTE SENSING MEASUREMENTS AND ESTIMATED EMISSION FACTORS FOR TRANSIT BUSES 114 7.1
Data Collection and Database Development Activities
114
7.2
Summary of Transit Bus Emissions Estimates
117
7.3
Transit Bus Emissions Estimates for the Observed Triangle Transit Authority Buses
118
Transit Bus Emissions Estimates for the Observed Raleigh-Durham International Airport Buses
126
7.4
8.0
65 65
DISCUSSION, CONCLUSIONS, AND RECOMMENDATIONS
136
8.1
136
Limitations and Caveats
iii
9.0
8.2
Estimates of Per-Passenger Emissions
138
8.3
Estimates of the High-Emitting Fraction of Vehicles
139
8.4
Sample Size and Statistical Significance
140
8.5
Recommendations
140
REFERENCES
142
APPENDIX A: WAKE COUNTY OBSERVED SCHOOL BUS MAINTENANCE RECORDS 143 APPENDIX B: SUMMARY OF REMOTE SENSING DATA AND CALCULATED EMISSION FACTORS FOR GASOLINE-FUELED SCHOOL BUSES 149 APPENDIX C: SUMMARY OF REMOTE SENSING DATA AND CALCULATED EMISSION FACTORS FOR DIESEL-FUELED SCHOOL BUSES AT THE ROCK QUARRY ROAD SITE 151 APPENDIX D: SUMMARY OF REMOTE SENSING DATA AND CALCULATED EMISSION FACTORS FOR DIESEL-FUELED SCHOOL BUSES AT LAURA DUNCAN ROAD, GARNER, WAKE FOREST, AND WOODCROFT SITES 163
iv
List of Figures Figure 1-1.
Example of Remote Sensing Data Obtained in Raleigh, NC for 1,027 Vehicles Using the RES-100 "Smog Dog"
8
Figure 2-1.
Simplified Schematic of a Deployed Remote Sensing Device
11
Figure 2-2.
Photograph of the Interior of the RES-1 Smog Dog
11
Figure 2-3.
Close up of the Infrared Source, Infrared Receiver, Portable Electric Generator, and Calibration Gas Cylinders
11
Figure 2-4.
Set Up of the Video Camera for Recording Images of Each Vehicle
12
Figure 2-5.
Schematic of Method for Adjusting Beam Height
14
Figure 4-1.
Schematic of Location of Remote Sensing Site at RDU Airport
37
Figure 4-2.
View from Inside the Entrance to Rock Quarry Road Site
38
Figure 4-3.
View of School Bus Staging Area at the Rock Quarry Road Site
38
Figure 4-4.
View of the Rock Quarry Road Site: Outbound Bus Near the Inside Gate39
Figure 4-5.
Southbound View of the Laura Duncan Road Site
Figure 6-1.
Variability in Fuel Economy for 209 Diesel and 21 Gasoline School Buses Observed at Rock Quarry Road. 75
Figure 6-2.
Variability and Uncertainty in 984 Estimates of Diesel School Bus CO Emissions (grams/gallon) based Upon Remote Sensing Measurements at the Rock Quarry Road Site. 77
Figure 6-3.
Variability and Uncertainty in 984 Estimates of Diesel School Bus CO Emissions (grams/mile) based Upon Remote Sensing Measurements at the Rock Quarry Road Site. 77
Figure 6-4.
Variability and Uncertainty in 984 Estimates of Diesel School Bus Propane-Equivalent Hydrocarbon Emissions (grams/gallon) based Upon Remote Sensing Measurements at the Rock Quarry Road Site. 78
Figure 6-5.
Variability and Uncertainty in 984 Estimates of Diesel School Bus Propane-Equivalent Hydrocarbon Emissions (grams/mile) based Upon Remote Sensing Measurements at the Rock Quarry Road Site. 78
Figure 6-6.
Comparison of Empirical Cumulative Distribution Functions for CO Emissions of 14 Diesel-Fueled School Buses with 10 or More Observations at the Rock Quarry Road Site.
Figure 6-7.
39
81
Comparison of 95 Percent Confidence Intervals for Mean CO Emissions of 14 Diesel-Fueled School Buses with 10 or More Observations at the Rock Quarry Road Site. 81
v
Figure 6-8.
Comparison of Empirical Cumulative Distribution Functions for PropaneEquivalent Hydrocarbon Emissions of 14 Diesel-Fueled School Buses with 10 or More Observations at the Rock Quarry Road Site. 84
Figure 6-9.
Comparison of 95 Percent Confidence Intervals for Mean PropaneEquivalent Hydrocarbon Emissions of 14 Diesel-Fueled School Buses with 10 or More Observations at the Rock Quarry Road Site. 84
Figure 6-10.
Scatter Plot of Estimated CO Emissions Versus Odometer Readings for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site. 87
Figure 6-11.
Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Odometer Readings for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site. 87
Figure 6-12.
Scatter Plot of Estimated CO Emissions Versus Chassis Year for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site. 88
Figure 6-13.
Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Chassis Year for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site. 88
Figure 6-14.
Scatter Plot of Estimated CO Emissions Versus Capacity for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site. 89
Figure 6-15.
Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Bus Capacity for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site. 89
Figure 6-16.
Scatter Plot of Estimated CO Emissions Versus Fuel Economy for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site. 91
Figure 6-17.
Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Fuel Economy for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site. 91
Figure 6-18.
Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Estimated CO Emissions for 984 Observations of 209 DieselFueled School Buses at the Rock Quarry Road Site.
92
Figure 6-19.
Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Inbound Versus Outbound Diesel Buses at Rock Quarry Road (n = 984). 93
Figure 6-20.
Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Inbound Versus Outbound Diesel Buses at Rock Quarry Road (n = 984). 93
vi
Figure 6-21.
Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Inbound Versus Outbound Chevrolet Diesel Buses at Rock Quarry Road (n = 375). 95
Figure 6-22.
Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Inbound Versus Outbound Chevrolet Diesel Buses at Rock Quarry Road (n = 375).95
Figure 6-23.
Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Inbound Versus Outbound Ford Diesel Buses at Rock Quarry Road (n = 329). 96
Figure 6-24.
Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Inbound Versus Outbound Chevrolet Diesel Buses at Rock Quarry Road (n = 329).96
Figure 6-25.
Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Inbound Versus Outbound International Diesel Buses at Rock Quarry Road (n = 280). 97
Figure 6-26.
Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Inbound Versus Outbound International Diesel Buses at Rock Quarry Road (n = 280). 97
Figure 6-27.
Variability and Uncertainty in 68 Estimates of Gasoline School Bus CO Emissions (grams/gallon) based Upon Remote Sensing Measurements at the Rock Quarry Road Site. 99
Figure 6-28.
Variability and Uncertainty in 68 Estimates of Gasoline School Bus CO Emissions (grams/mile) based Upon Remote Sensing Measurements at the Rock Quarry Road Site. 99
Figure 6-29.
Sampling Distribution for Mean Propane-Equivalent Hydrocarbon Emissions (g/gal) for Gasoline-Fueled School Buses Observed at the Rock Quarry Road Site based Upon an Assumed Lognormally Distributed Population for Inter-Vehicle Variability in Emissions. 100
Figure 6-30.
Comparison of Sampling Distributions for Mean Propane-Equivalent Hydrocarbon Emissions (g/mi) for Gasoline-Fueled School Buses Observed at the Rock Quarry Road Site based Upon Resampling, Normal Distribution and Lognormal Distribution for Inter-Vehicle Variability in Emissions. 100
Figure 6-31.
Scatter Plot of Estimated CO Emissions Versus Odometer Readings for 68 Observations of 21 Gasoline-Fueled School Buses at the Rock Quarry Road Site. 102
Figure 6-32.
Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Odometer Readings for 68 Observations of 21 Gasoline-Fueled School Buses at the Rock Quarry Road Site. 102
vii
Figure 6-33.
Scatter Plot of Estimated CO Emissions Versus Chassis Year for 68 Observations of 21 Gasoline-Fueled School Buses at the Rock Quarry Road Site. 103
Figure 6-34.
Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Chassis Year for 68 Observations of 21 Diesel-Fueled School Buses at the Rock Quarry Road Site. 103
Figure 6-35.
Scatter Plot of Estimated CO Emissions Versus Fuel Economy for 68 Observations of 21 Gasoline-Fueled School Buses at the Rock Quarry Road Site. 104
Figure 6-36.
Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Fuel Economy for 68 Observations of 21 Diesel-Fueled School Buses at the Rock Quarry Road Site. 104
Figure 6-37.
Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Inbound Versus Outbound Gasoline Buses at Rock Quarry Road (n = 68). 106
Figure 6-38.
Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Inbound Versus Outbound Gasoline Buses at Rock Quarry Road (n = 68). 106
Figure 6-39.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Gasoline Buses at Laura Duncan Road (n=16). 107
Figure 6-40.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Gasoline Buses at Laura Duncan Road (n=16). 107
Figure 6-41.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel Buses at the Woodcroft site (n=38). 109
Figure 6-42.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel Buses at the Woodcroft site (n=38). 109
Figure 6-43.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel Buses at the Laura Duncan Road site (n=33) 110
Figure 6-44.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel Buses at the Laura Duncan Road site (n=33). 110
Figure 6-45.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel Buses at the Wake Forest site (n=9) 111
Figure 6-46.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel Buses at the Wake Forest site (n=9). 111
Figure 6-47.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel Buses at the Garner site (n=5). 112
viii
Figure 6-48.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel Buses at the Garner site (n=5). 112
Figure 7-1.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel TTA Buses (All Sites, n=37) 116
Figure 7-2.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel TTA Buses (Woodcroft Site, October 16, 1996, n=6)116
Figure 7-3.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel TTA Buses (Woodcroft Site, October 23, 1996, n=5)117
Figure 7-4.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel TTA Buses (Briggs Avenue Site, n=12) 117
Figure 7-5.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel TTA Buses (Davis Drive Site, n=14) 118
Figure 7-6.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel TTA Buses (All Sites, n=37) 119
Figure 7-7.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel TTA Buses (Woodcroft Site, October 16, 1996, n=6) 119
Figure 7-8.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel TTA Buses (Woodcroft Site, October 23, 1996, n=5) 120
Figure 7-9.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel TTA Buses (Briggs Avenue Site, n=12) 120
Figure 7-10.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel TTA Buses (Davis Drive Site, n=14) 121
Figure 7-11.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Gasoline TTA Buses (n=6) 122
Figure 7-12.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Gasoline TTA Buses (n=6) 122
Figure 7-13.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for New Diesel Buses at RDU Airport (n=106) 124
Figure 7-14.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for New Diesel Buses at RDU Airport (n=106) 124
Figure 7-15.
Inter-Measurement Variability and Uncertainty in Average for Calculated CO Emissions for New RDU Bus No. 3 (n=25) 125
ix
Figure 7-16.
Inter-Measurement Variability and Uncertainty in Average for Calculated Propane-Equivalent Hydrocarbon Emissions for New RDU Bus No. 3 (n=25) 125
Figure 7-17.
Inter-Measurement Variability and Uncertainty in Average for Calculated CO Emissions for New RDU Bus No. 4 (n=28) 126
Figure 7-18.
Inter-Measurement Variability and Uncertainty in Average for Calculated Propane-Equivalent Hydrocarbon Emissions for New RDU Bus No. 4 (n=28) 126
Figure 7-19.
Inter-Measurement Variability and Uncertainty in Average for Calculated CO Emissions for New RDU Bus No. 7 (n=25) 127
Figure 7-20.
Inter-Measurement Variability and Uncertainty in Average for Calculated Propane-Equivalent Hydrocarbon Emissions for New RDU Bus No. 7 (n=25) 127
Figure 7-21.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Old Diesel Buses at RDU Airport (n=34) 129
Figure 7-22.
Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Old Diesel Buses at RDU Airport (n=34) 129
Figure 7-23.
Inter-Measurement Variability and Uncertainty in Average for Calculated CO Emissions for Old RDU Bus No. 1 (n=9) 130
Figure 7-24.
Inter-Measurement Variability and Uncertainty in Average for Calculated Propane-Equivalent Hydrocarbon Emissions for Old RDU Bus No. 1 (n=9) 130
Figure 7-25.
Inter-Measurement Variability and Uncertainty in Average for Calculated CO Emissions for Old RDU Bus No. 5 (n=10) 131
Figure 7-26.
Inter-Measurement Variability and Uncertainty in Average for Calculated Propane-Equivalent Hydrocarbon Emissions for Old RDU Bus No. 5 (n=10) 131
x
List of Tables Table ES-1.
Summary of Estimated Average Emission Factors for Diesel-Fueled School and Transit Buses ES-4
Table ES-2.
Summary of Estimated Average Emission Factors for Gasoline-Fueled School Buses ES-5
Table 3-1.
Top Ten N.C. Counties - Public School Bus Fleets for 1994-1995
Table 3-2.
Top Ten N.C. Counties - Total Public School Bus Riders for 1994 to 199528
Table 3-3.
Top Ten N.C. Counties - Public School Bus Vehicle Miles Traveled for 1994 to 1995 29
Table 3-4.
Top Ten N.C. Counties - Gallons of Fuel Used for 1994 to 1995
31
Table 5-1.
Fuel Composition - Examples of Detailed Hydrocarbon Speciation
54
Table 5-2.
Summary of Fuel Properties
60
Table 5-3.
Equivalent Molecular Formula
62
Table 6-1.
Summary of School Bus Emissions Measurements
69
Table 6-2.
Summary of School Bus RSD Data Collection Dates and Locations
69
Table 6-3.
Summary of Estimated Average Emission Factors for Diesel-Fueled School Buses
72
26
Table 6-4.
Summary of Estimated Average Emission Factors for Gasoline-Fueled School Buses 74
Table 6-5.
Summary of CO Emissions for 14 Individual Diesel Buses at Rock Quarry Road 80
Table 6-6.
Summary of T-Test Results for CO Emissions of 14 Individual Diesel Buses
77
Table 6-7.
Summary of Propane-Equivalent Hydrocarbon Emissions for 14 Individual Diesel School Buses at Rock Quarry Road 80
Table 6-8.
Summary of T-Test Results for Propane-Equivalent Hydrocarbon Emissions of 14 Individual Diesel Buses at Rock Quarry Road
80
Table 7-1.
Summary of Data Collection Sites for TTA Transit Buses
115
Table 7-2.
Summary of Estimated Average Emission Factors for Diesel-Fueled Transit Buses of the Triangle Transit Authority
115
Summary of Estimated Average Emission Factors for Diesel-Fueled Transit Buses at Raleigh Durham International Airport
116
Table 7-3.
xi
xii
Acknowledgments This project would not have been possible without the significant amount of support that we have received from a number of key individuals. We would like to thank the sponsors at North Carolina Department of Transportation, and in particular members of the project advisory committee. The committee members include Dr. Ron Poole, David Hyder, and Pat Strong of NCDOT, Donald Redmond of the Air Quality Section (AQS) of the NC Department of Environment, Health, and Natural Resources, and John Schrohenloher of the Federal Highway Administration. Donald Redmond of NC DEHNR provided access to remote sensing equipment and personnel utilized in this study. We especially thank Mark Smith of the AQS for the many hours of his time which he provided in the field collecting data, including early and late days. We also thank numerous other individuals at NC DEHNR who assisted in the data collection effort. Niranjan Vescio of the U.S. Environmental Protection Agency in Research Triangle Park (now of Remote Sensing Technologies, Inc.) provided valuable advice and kindly loaned us replacement equipment at several key points during field data collection. We are grateful to the many people who provided data and information for the study, including the NC Department of Public Instruction, the Wake county school system, the Institute for Transportation Research and Engineering, and others. We especially thank Doug White and Derrick Graham of NCDPI, and Wyatt Currin of the Wake County Public Schools for providing data regarding school bus characteristics, and James Ritchey for providing information regarding Triangle Transit Authority bus characteristics. Melissa Harden and Tim Soles provided information regarding transit buses operated by the Raleigh-Durham International Airport Authority. We thank those individuals who provided permissions for field data collection work at specific sites. We thank Tim Johnson and Ken McIntire, district engineers for the NC DOT, who provided permission for site locations, and helped to get approval to work on state roadways. Carl Dawson of the City of Raleigh provided permissions for siting and measurements in downtown Raleigh. Ellis Cayton of the RDU Airport Authority reviewed potential sites and coordinated obtaining permissions for data collection. Ken Wright and Wesley Carom provided permissions for site locations inside the City of Durham. Riley Riener and Wyatt Harper provided permissions for work on school bus data collection at the Rock Quarry Road site. Robert Teer of Teer Associates provided permission for data collection at the shopping center where the TTA central bus staging area is located. We also thank Melissa Benson, who participated in this project with partial support from the Research Experiences for Undergraduates (REU) program of the National Science Foundation, for her extensive data analysis work. The authors are solely responsible for the use of information and the content of this report.
xiii
EXECUTIVE SUMMARY Current approaches to estimating highway vehicle emissions are based upon laboratory emissions testing for prescribed driving cycles. The driving cycle emissions data may not be representative of real world, on-road emissions. An example of a vehicle category whose emissions may not be well-characterized by the laboratory-based emissions data are school buses. In North Carolina, there are over 12,000 public school buses which travel approximately 147 million vehicle miles annually. In addition, there are numerous transit buses in cities and counties throughout the state. The overall purpose of this project is to characterize the on-road emissions of these types of vehicles. The technology employed for this purpose is infrared remote sensing. Remote sensing has been used primarily to help improve inspection and maintenance programs. Remote sensing can be used to obtain instantaneous measurements of tailpipe emissions. The remote sensing measurements are in the form of ratios of carbon monoxide (CO) to carbon dioxide, and hydrocarbons (HC) to carbon dioxide, in the exhaust plume of a passing vehicle. These ratios can be used, in combination with a combustion mass balance model, fuel properties such as composition and density, and vehicle fuel economy, to estimate emissions of CO and HC on a grams per gallon of fuel consumed or grams per mile of vehicle travel basis. The objectives of this project were to: (1) Conduct on-road remote sensing of carbon monoxide (CO) and hydrocarbon (HC) pollutant emissions from selected types of vehicles (i.e. school and transit buses); (2) Determine the on-road emission rates of such vehicles; (3) Estimate the number of passengers carried by selected non-controlled vehicles such that emission rates can be related to a per passenger basis; (4) Estimate the fraction of high emitting vehicles in the traffic stream; and (5) Collect sufficient data to satisfy statistical significance tests. Bus Fleets Selected for Study Objectives (1) and (2) have motivated the bulk of the effort in this project. Based upon the capabilities of remote sensing with respect to maximum beam length and safe deployment of the equipment in the field, numerous candidate remote sensing sites were evaluated. A key objective was to identify sites at which there was sufficient frequency of passing school or transit buses to obtain a large amount of data in a relatively short amount of time. Therefore, it was necessary to identify and characterize selected bus fleets for focus in this study. After collecting
ES-1
data regarding the characteristics of the statewide public school bus fleet, it was decided to focus data collection efforts in Wake County. Wake County has over 600 school buses which travel over 11 million miles per year. Wake County was selected because it offers a representative bus fleet and facilitiated logistics with respect to Remote Sensing Device (RSD) deployment. After evaluating school bus routes and acceptable RSD sites, five general locations were selected for school bus data collection. Transit buses operated by the Triangle Transit Authority (TTA), Raleigh Durham International Airport (RDU) authority, and Capital Area Transit (CAT) of Raleigh were also selected for data collection. Three sites were selected for measuring TTA bus emissions, one site was selected for measuring RDU bus emissions, and one site was selected for measuring CAT bus emissions. Data regarding the bus fleets, such as maintenance records and vehicle manufacturers, were obtained from relevant agencies. Development of Emission Factor Models The selected bus fleets are comprised primarily of diesel-fueled buses. However, the public school bus and TTA bus fleets include some gasoline-fueled buses. A detailed review of fuel properties was undertaken in order to characterize key physical and chemical properties of both gasoline and diesel fuels, such as fuel density and fuel composition in terms of weight percent of carbon and hydrogen. A mass balance-based combustion model was developed based upon a stoichiometric chemical equation for the conversion of fuel and air to carbon monoxide, carbon dioxide, water vapor, hydrocarbons, and nitrogen. Based upon the fuel properties and the measured emission ratios obtained from remote sensing, the model can be used to calculate CO and HC emission factors in units of grams of pollutant emitted in the exhaust per gallon of fuel consumed. Emission factors in units of grams per mile of vehicle travel can be estimated based upon vehicle-specific fuel economy data, which was available for most of the observed buses. Remote Sensing Data Collection Activities A total of 1,340 valid remote sensing measurements of on-road emissions ratios of CO/CO2 and HC/CO2 were obtained for 265 diesel-fueled school buses, 36 gasoline-fueled school buses, 19 diesel-fueled buses of the TTA, 3 gasoline-fueled buses of TTA, and 12 dieselfueled transit buses at RDU over the course of 22 days of field work. The development of databases based upon the observed ratios of CO/CO2 and HC/CO2 and available information regarding characteristics of the observed buses involved detailed review of both data and video records from the remote sensing device (RSD), as well as interactions with several agencies. Numerous quality assurance checks were performed on the data sets, which were analyzed by three different people and reviewed several times for validity. ES - 2
Emission Estimates Based Upon Remote Sensing The estimated fleet average emission factors for each of the observed diesel bus fleets are given in Table ES-1. The school bus data include measurements obtained at five different sites. It was not possible to identify statistically significant differences in average emissions among the five sites, even though average speeds varied from approximately 15 to 45 mph from one site to another. However, for some sites, very few data points were collected, leading to wide confidence intervals on the average emissions estimates. The largest number of data points, 984, were collected at one site. The school bus data were analyzed to attempt to identify explanatory variables for differences in emissions. No statistically significant findings were obtained. This is because the variability in emissions for individual buses was typically similar to the variability in emissions for the entire fleet. Therefore, factors such as vehicle age, odometer reading, vehicle size, manufacturer, and fuel economy were found to be statistically insignificant in explaining variability in emissions. The emission measurements for the TTA buses are based upon observations at three different sites. The TTA buses are based upon a Ford 350 chassis. It appears that this fleet has the lowest emission rates of the fleets that were observed in this study. However, differences in average emissions may be attributable to site characteristics, which differed among the four fleets. Two different bus fleets were observed at the RDU site. In late summer of 1996, measurements were taken on an older bus fleet based upon a Ford 350 chassis. This fleet was replaced with a new bus fleet in the early fall of 1996 based upon substantially larger buses built by Blue Bird. All measurements at RDU were taken at one site. The 106 measurements of new buses include 25 repeat values for each of two buses and 28 repeat values for a third. The variability in emissions estimates for each individual bus are approximately the same as the overall variability in emissions for all of the measurements taken at the site. This implies that the instantaneous emission estimates obtained by remote sensing are highly variable due to the short averaging time involved (approximately 0.6 seconds). For the older bus fleet, there are 10 repeat measurements on one bus, 9 repeat measurements on a second bus, and smaller number of repeat measurements on other buses in the older fleet. Due to the small sample size obtained for the older fleet, the confidence intervals for the mean emissions of the fleet and of each individual bus are relatively wide. For this reason, it is not possible to draw conclusions regarding comparisons between the emissions of the newer and older fleets.
ES - 3
Table ES-1. Summary of Estimated Average Emission Factors for Diesel-Fueled School and Transit Buses 95 Percent Standard Confidence Interval Description Mean Deviation Lower Upper Count All School Buses (All Sites) CO g/gal 97.1 139.5 88.7 105.5 1069 HC g/gal 25.2 58.8 21.7 28.7 1069 CO g/mi 13.0 18.6 11.9 14.1 1069 HC g/mi 3.4 7.8 2.9 3.9 1069 Triangle Transit Authority Buses (All Sites) CO g/gal 59.1 43.1 45.2 73.0 37 HC g/gal 14.8 12.4 10.7 18.8 37 CO g/mi 7.8 5.5 6.0 9.5 37 HC g/mi 1.9 1.6 1.4 2.5 37 Raleigh Durham International Airport: New Bus Fleet CO g/gal 92.1 40.7 84.3 99.8 106 HC g/gal 15.1 9.5 13.3 16.9 106 CO g/mi 12.1 5.3 11.1 13.1 106 HC g/mi 2.0 1.2 1.7 2.2 106 Raleigh Durham International Airport: Old Bus Fleet CO g/gal 123.8 186.3 61.2 186.4 34 HC g/gal 33.1 70.3 9.5 56.8 34 CO g/mi 16.2 24.5 8.0 24.5 34 HC g/mi 4.4 9.2 1.2 7.5 34
Valid data could not be obtained for the CAT buses based upon the existing RSD equipment. This is because the CAT buses emit from the top of the bus, requiring the use of scaffolding to elevate the infrared beam. Because the space between the top of the windows and the top of the roof of the bus is relatively small, it was difficult to achieve a beam height which would allow for proper triggering of the beam. The estimated fleet average emission factors for each of the observed gasoline school bus fleets are summarized in Table ES-2. Insufficient data were collected regarding gasoline fueled transit buses. Gasoline-fueled school buses are a declining portion of the total number of public school buses in North Carolina, and also represented a declining portion of vehicle miles ES - 4
Table ES-2. Summary of Estimated Average Emission Factors for Gasoline-Fueled School Buses 95 Percent Standard Confidence Interval Description Mean Deviation Lower Upper Count All School Buses (All Sites) CO g/gal 970.3 808.4 801.4 1139.2 88 HC g/gal 42.7 165.9 8.0 77.3 88 CO g/mi 206.7 172.2 170.7 242.7 88 HC g/mi 9.2 35.4 1.8 16.6 88
travelled by public school buses. The data clearly indicate that CO emissions are much higher for gasoline than for diesel fueled school buses. Hydrocarbon emissions may be significantly higher as well. Limitations of the Emissions Data The emissions estimates developed in this study must be used with some caution. Some of the key factors to consider include: • Vehicle emissions may depend upon site conditions and driver behavior. • Although several sites were used for both school buses and TTA buses, nonetheless only a limited sample of sites were included in this study. Thus, the emissions estimates may not be representative of other types of sites not included here. • Because the RSD typically measures the emissions that occur over a 0.6 second period, there is substantially variation from one measurement to another, even for the same bus at the same site. The variation may be attributable to differences in driver behavior that are not observable with the current equipment. • The wide range of variability from one emission measurement to another contributes to uncertainty regarding estimates of vehicle or fleet average emissions. • The grams per mile emission factors were calculated using average annual fuel economy estimates in combination with emissions data taken over 0.6 seconds. Variability in fuel economy for the same averaging times as the emissions data is likely to be larger than indicated by the annual averages. • For the most part, the measurements in this study were taken during summer weather conditions. These data may not be representative of wintertime emissions even for the same sites as were used in this study. • The measurements of hydrocarbon emissions account for only a portion of total hydrocarbon emissions. This is because the RSD is designed and calibrated for hydrocarbons similar to alkanes and alkenes. The HC emission factors produced in this study, therefore, are not inclusive of all possible hydrocarbon compounds emitted from ES - 5
the tailpipe. The amount of bias in these emissions factors is not well known, but may be as much as 50 percent or more lower than the actual total hydrocarbon emissions. • The measurements of hydrocarbon emissions do not include particulate matter. Especially for diesel buses, a portion of the hydrocarbons in the engine exhaust may be contained in or condensed upon particulate matter. This leads to additional bias in the hydrocarbon emission factors. • Because the RSD is calibrated using a cylinder gas that represents a high-emitting gasoline vehicle, the accuracy of the measurements may not be as good for other emissions values. Preliminary work at the U.S. Environmental Protection Agency suggests that some RSDs have a nonlinear response to different emission levels, which introduces a bias in the observed values. Based upon these considerations, the CO emission factors used in this study can be considered to be more reliable than the HC emission factors. The HC emission factors should be viewed as a lower bound on the true emissions of hydrocarbons from the observed vehicles. Because variability in the speciation of total hydrocarbons from engine exhaust is not well known, the amount of bias in the hydrocarbon emission factors is not readily quantifiable. Estimates of Per-Passenger Emissions Given the caveats of the preceding section, a preliminary estimate is made regarding emissions of CO and hydrocarbons for both school buses and TTA transit buses on a per passenger basis. The average CO emissions attributable to each school bus rider are approximately 12 kg of CO per daily rider per year. Approximately 80 percent of this total is due to gasoline-fueled school buses, which comprise a decreasing share of the public school bus fleet. If all the school buses were diesel, then the average CO emissions would be approximately 2.8 kg per rider per year. The annual average propane-equivalent hydrocarbon emissions would be approximately 1.0 kg per rider per year. It is possible that the total hydrocarbon emissions are double or more than this amount. For the TTA bus fleet, the approximate annual emissions are 4.9 kg CO per passenger per year based upon diesel bus emissions. For propane-equivalent hydrocarbons, the emissions are approximately 1.2 kg per passenger per year. The actual total hydrocarbon emissions may be double or more than this amount. Although the TTA buses have lower average emission factors than the school buses, the per passenger totals are larger. This is likely to be due to longer bus routes for the TTA buses compared to school buses, since the TTA provides intercity service.
ES - 6
Fraction of High-Emitting Vehicles One of the original objectives of the study was to identify the fraction of high emitting vehicles in the fleet. However, it is difficult to provide a quantitative answer in response to this objective. Commonly, remote sensing measurements are misinterpreted to indicate the fraction of vehicles that are high emitters. In fact, the distributions of emissions merely indicate that a fraction of the observations occurred during a time period in which a vehicle was producing high emissions. The study of emissions of individual school buses and of individual RDU transit buses illustrates that even individual vehicles can produce a wide range of emissions readings. Therefore, it is difficult to classify an individual vehicle as a high emitter based upon a small number of remote sensing emissions measurements. It is not possible to draw rigorous conclusions in this study regarding the presence or absence of vehicles that are systematically high emitters. To do so would require more detailed measurements on individual vehicles for a much larger number of individual vehicles. Sample Size and Statistical Significance The wide range of variability in the data obtained in this study suggests that much larger data sample sizes are needed to obtain narrow confidence intervals for the mean values and for the purpose of identifying explanatory variables. In analyzing the RSD data, confidence intervals were estimated and considered when making comparisons between data sets. In the course of data analysis, a large number of pair-wise t-tests were performed. These tests typically yielded negative results, indicating that there were not significant differences between the datasets being compared. Recommendations The emission factors developed in this study represent estimates of on-road emissions of selected types of vehicles for selected sites. The variability and uncertainty in the emissions factors are quantified. Due to the wide range of variability and uncertainty in the emission factors, it was difficult to identify explanatory variables which could be used to disaggregate the datasets For example, the variability in emissions for individual vehicles was found to be comparable to the variability in emissions for all observed vehicles, as in the case of dieselfueled school buses and new transit buses at RDU airport. These findings suggest that additional data are needed regarding the operating conditions for each individual vehicle in order to find potentially meaningful explanations for differences in emissions observations.
ES - 7
The procurement of a speed and acceleration measurement system for use with the RSD may enable the development of better explanations for differences in vehicle emissions, even for individual vehicles. It is recommended that additional data collection be conducted using the new speed and acceleration measurement capabilities to determine whether these yield significant insights regarding variability in emissions. A preliminary study, such as at RDU airport, could be used to evaluate the utility of the speed-acceleration data prior to more extensive data collection. For example, if variability in emissions measurements for individual buses can be explained in part by differences in speed and acceleration for each measurement, then it is likely that speed and acceleration would be useful in explaining differences in emissions for other vehicles or fleets. If initial testing under the relatively controlled conditions at RDU is successful, then applications of the new speed-acceleration measurement capabilities to school buses at the Rock Quarry Road site would be a logical next step. Both of these sites can yield significant amounts of data points for individual buses, and the Rock Quarry Road site also can yield information regarding approximately 200 buses. As more information becomes available regarding the nonlinearity in the response of the RSD to different values of the emissions ratios, it may become necessary to re-analyze the data contained in this study to correct for potential biases in the instrumentation. Because the data taken in this study were primarily for summertime conditions, it would be useful to collect comparable data for winter time conditions to evaluate the effect of ambient temperature on emissions estimates. This study focused upon collection of data for CO and hydrocarbons. Recently, NOx measurement capabilities for RSDs have become commercially available. A potential limitation of the new NOx measurement capabilities is that they may be less precise than the capability for CO measurements. Because NOx emissions from diesel vehicles are typically higher than for gasoline vehicles, it is important to consider acquisition of a NOx measurement capability and subsequent application to measurement of the bus fleets studied here.
ES - 8
1.0
INTRODUCTION
Highway vehicles are a major emission source for several key pollutants, including nitrogen oxides (NOx), carbon monoxide (CO), volatile organic compounds (VOCs), and particulate matter (PM). VOCs are often referred to as hydrocarbons (HC). Carbon monoxide poses direct health effects. Nitrogen oxides and VOCs are precursors to the formation of tropospheric ozone (O3). Exposure to ozone leads to acute respiratory illness. The ambient concentrations of NOx, CO, PM, and O3 are regulated under the National Ambient Air Quality Standards (NAAQS). Emission rates for individual motor vehicles are also regulated by Federal standards. There is a growing recognition of the need to obtain on-road emissions data that are representative of actual driving conditions. Furthermore, relatively little attention has been given to emissions from specialized categories of vehicles, such as school and transit buses. Good information about the emissions from such vehicles is necessary to develop emission inventories and transportation plans. The objectives of this project are to: (1) Conduct on-road remote sensing of pollutant emissions from selected types of vehicles (i.e. school and transit buses). (2) Determine the on-road emission rates of such vehicles. (3) Estimate the number of passengers carried by selected non-controlled vehicles such that emission rates can be related to a per passenger basis. (4) Estimate the fraction of high emitting vehicles in the traffic stream. (5) Collect sufficient data to satisfy statistical significance tests. This project employed state-of-the-art remote sensing technology to obtain onroad emissions data for these vehicles. The research program featured a procedure by which remote sensing field measurements studies were developed and by which the data obtained were analyzed. The results of this project are estimates of on-road CO and HC emissions for school and transit buses, in units of grams of pollutant per mile of vehicle travel. Statistical techniques, including classical analytical approaches and numerical modeling techniques, have been used to characterize the precision and accuracy of the estimated emission factors, and to distinguish inter-vehicle variability in emissions from uncertainty in fleet average emissions. The latter is of concern in developing area-wide emission inventories.
1
In the rest of this chapter, we provide background information regarding emissions from highway vehicles and the need for measurement of emissions for on-road vehicles under actual driving conditions. In Chapter 2, we discuss the remote sensing technology in more detail. In Chapter 3, we discuss the procedure by which sites were selected for remote sensing. In Chapter 4, we derive a model by which the remote sensing observations, which are obtained as ratios of pollutant concentrations, are used to calculate emissions factors on a grams per mile basis. Remote sensing was used to measure emissions from three different fleets of buses: (1) school buses; (2) transit buses used at Raleigh-Durham International Airport; and (3) transit buses operated by the Triangle Transit Authority. Each of these fleets are addressed in Chapters 5, 6, and 7, respectively. In Chapter 8, a discussion of the results of emission factor estimates for all three bus fleets is presented. Conclusions and recommendations for future work are provided in Chapter 9. 1.1
Mobile Source Emissions
From 1970 to 1991, the national highway emission inventory for hydrocarbons, which includes VOCs, has decreased by 66 percent, CO has decreased by 59 percent, and NOx has decreased by 21 percent. During this same time, the number of registered vehicles has increased by 80 percent, while total vehicle miles traveled (VMT) has doubled. The emissions reduction is due in large part to increasingly stringent emission control technologies included as part of new engine design and exhaust gas treatment (Leonard and Formby, 1994). Although the total emissions from motor vehicles has declined as the result of improved technology, they nonetheless comprise a substantial portion of the national emission inventory. These emissions typically account for about 40 percent of total VOCs, 40 percent of total NOx, 60 percent of total CO, and 20 percent of total particulate matter emissions in any particular area. These numbers can be misleading, because it is often that case that emission inventories underestimate VOC and CO emissions (NRC, 1992a). 1.2
Emission Regulations
The NAAQS impose maximum allowable atmospheric concentrations of six "criteria" pollutants, of which mobile sources are major emitters of four. The NAAQS are intended to protect public health, and they are implemented via State Implementation Plans (SIPs). Areas for which the ambient concentration of a criteria pollutant exceeds 2
the NAAQS are said to be in nonattainment for that pollutant. Such areas are subject to severe restrictions on permitting of any new emission sources, and they must find ways to reduce emissions to acceptable levels. While there are a number of air pollution regulations that affect mobile sources, one of the most important is the "conformity" rule. The conformity provisions of the 1990 Clean Air Act Amendments (CAAA) require that any federally funded project or transportation plan must demonstrate an improvement in air quality. Conformity requirements apply to nonattainment areas and maintenance plan areas. A maintenance plan area is one that has been redesignated from nonattainment status since the CAAA of 1990. In the future, conformity requirements may be extended to attainment areas within 85 percent of the NAAQS. To demonstrate conformity, a transportation plan or project must improve air quality with respect to one or more of the following: (1) motor vehicle emission budget in the SIP; (2) emissions that would be realized if the proposed plan or program is not implemented; and/or (3) emission levels in 1990. There are many specific aspects of the conformity rule with respect to when and how assessments must done (Austin, 1994). Conformity requirements have made air quality a primary constraint on transportation planning (Sargeant, 1994). Essentially, a conformity determination is intended to be a guarantee that a transportation plan and a Transportation Improvement Program (TIP) conforms to the goals of the SIP. Conformity requirements differ depending on the type of project. For transportation plans and TIPs, a regional analysis is required. Regional analyses must typically consider primary pollutants including NOx, VOC, PM10 (particulate matter less than 10 microns in diameter), and CO. In contrast, a conformity analysis for a project need only consider localized impacts, which typically focuses on CO and PM10 (Sargeant, 1994). TIPs are updated annually in North Carolina. Transportation plans are required to be updated, at a minimum, every three years for metropolitan planning areas that lie in a nonattainment area. There must be a conformity determination on transportation plans and TIPs any time that they are revised. 1.3
Emissions Contributions of "Non-Controlled" Vehicles
As emission limits on light duty vehicles become increasingly stringent, the proportion of total mobile source emissions contributed by other types of vehicles is likely to be increasing. The development of emission control technologies has been focused primarily on light duty gasoline vehicles (LDGVs). The new vehicle emission standards for LDGVs are currently 3.4 g/mi for CO, 0.41 g/mi for HC, and 0.4 g/mi for 3
NOx. New vehicle emissions are measured for a subset of newly made cars to gauge whether manufacturers are in compliance with U.S. Environmental Protection Agency standards. LDGV certification testing is done a chassis dynamometer. For heavy duty vehicles, which are typically vehicles over 8,500 pounds gross vehicle weight (GVW), the emission regulations are expressed in terms of grams of pollutant per brake-horsepowerhour from the engine. Thus, testing is done with an engine dynamometer. The current emission standards for heavy duty diesel vehicles (HDDVs) are 15.5 g CO/bhp-hr, 1.3 g HC/bhp-hr, and 5.0 g NOx/bhp-hr. The standards for heavy duty gasoline vehicles (HDGVs) are slightly more stringent for CO and HC (Black, 1992; Cooper and Alley, 1994). When HDGV emissions are estimated on a gram/mile basis and compared to LDGVs, it is typically the case that for the HDGV: (1) CO emissions are higher by a factor of two or three, with more pronounced differences at low speeds and low ambient temperatures; (2) HC emissions are higher by approximately 20 percent at high speed and a factor of two at low speed; and (3) NOx emissions are higher by a factor of two to three over a range of speeds. Compared to LDGVs, HDDVs are believed to have: (1) 20 to 40 percent lower HC emissions, but higher particulate matter emissions; (2) substantially lower CO emissions by a factor of three to five; and (3) substantially higher NOx emissions by a factor of seven to ten. These comparisons are based on averages. Emissions are highly variable as a function of driving conditions. Heavy duty vehicles that are of special concern in the development of emissions inventories include school buses, service fleets, emergency medical vehicles, fire service, law enforcement, and miscellaneous other categories. These vehicles comprise an increasing share of the vehicle fleet on the streets and highways of North Carolina. For example, school buses have traditionally operated from late August through June of the school year. However, many school systems are considering year-round schooling, which would increase school bus traffic during the summer. Wintertime emissions of CO are of concern because high-emitting vehicles contribute significantly to the possibility of nonattainment and may need to be considered and addressed in conformity analyses. Summertime emissions of hydrocarbons and nitrogen oxides are of concern because they are precursors to the photochemical formation of ozone. Ozone formation is greatest during the longer daylight and warmer summer months, and is particularly acute during stable atmospheric conditions (NRC, 1992a).
4
Of the various types of unusual and exceptional uncontrolled vehicles, school buses are perhaps the most ubiquitous. In North Carolina, there are over 12,000 public school buses which travel more than 140 million miles per year. The State of North Carolina is responsible for estimating mobile source emissions in seventeen metropolitan planning areas, ranging in size from large cities such as Charlotte and Raleigh to smaller towns such as Hickory and Concord. Because school buses are not well-characterized by existing emission inventory tools such as the Mobile5a emission factor model, it is important to develop emission estimates that are representative of actual on-road emissions. This project will address this need. 1.4
Conventional Approaches To Estimating Motor Vehicle Emissions
Emission inventories for motor vehicles are developed based on emission factors and activity data. Emission factors are emission rates normalized to some measure of activity. For mobile sources, emission factors are generally expressed as grams of pollutant per mile driven. Some emissions, such as refueling and evaporative emissions, and CO emissions during idling, are not associated with travel. However, these are often averaged with the other emissions on the basis of an assumed trip profile. Emission factors for mobile sources are developed based primarily upon Federal certification testing of new motor vehicles. The groupings typically include light duty vehicles (e.g., cars, pickup trucks), heavy duty vehicles (e.g., trucks), buses, and motorcycles. For light duty vehicles, testing is performed with the complete vehicle on a chassis dynamometer within a climate controlled test cell. This enables the determination of exhaust emissions as a function of speed, driving cycle, ambient temperature, and ambient humidity. A standard driving cycle, which consists of a changing speed profile with time, is the Federal Test Procedure (FTP). From the driving cycle test data, emission factors are developed. For heavy duty vehicles, testing is performed using only the engine on an engine dynamometer. Emission factors depend on speed, temperature, vehicle technology, trip profiles, and many other factors. Because of the complexity of estimating mobile source emission factors, several computer codes have been developed for this purpose. The most widely used is Mobile5a, which is a U.S. Environmental Protection Agency model. The California Air Resources Board (CARB) has developed its own model, called EMFAC7 (NRC, 1992). The emission factors obtained from models such as Mobile5a must be multiplied by the estimated vehicle miles traveled (VMT) for each vehicle category. Ideally, VMT estimates would be based on long-term monitoring of roads, and would be disaggregated 5
by road type, time of day, time of year, speed, and other attributes (NRC, 1992). Data are rarely available at this level of detail to support the development of emissions inventories. For example, traffic volume, speeds, turning movements, and level of service are commonly available, but usually only for one day or one week of sampling (O'Connor and Ireson, 1994). There are several important shortcomings to the emission factor models. First, the emission factors are based primarily on certification testing for new vehicles and supplemental driving tests. Vehicle emissions change with age and depend on many factors, such as driver behavior, road grade, fuels used, loads carried, and so on. Thus, laboratory conditions for new vehicles are not likely to be representative of actual driving conditions and actual emissions on the road. Second, the emission factors are based on specific driving cycles. There is uncertainty regarding the representative of these driving cycles with respect to actual driving. Third, emissions are sensitive to factors such as speed, acceleration, temperature, age, maintenance, condition of the emission control systems, type of fuel used, and other factors. The approach used in the emission factor models is to employ "correction factors" to adjust for changes in some of these conditions. For example, there is a speed correction factor which is used to adjust tailpipe emission factors for differences in average speed over a driving cycle. However, these corrections are based on comparisons of various driving cycles which have different average speeds. For a given average speed, there may in fact be many alternate types of driving cycles which are not accounted for by this approach. Techniques for collecting on-road emissions data offer the promise of developing better emission estimates that are representative of actual driving conditions. The leading technology for obtaining such data is remote sensing. 1.5
Remote Sensing of On-Road Emissions
Remote sensing is a technology that has emerged over the last few years as a means to collect on-road emissions data. Current commercially-available remote sensing devices (RSDs) are capable of measure the carbon monoxide (CO), hydrocarbon (HC), and carbon dioxide (CO2) concentrations in the exhaust plume of vehicles as they operate on the road (SBRC, 1994). Remote sensing data can be used, therefore, to develop emission estimates that are representative of actual emissions. To date, however, no studies have been identified which specifically address emissions from uncontrolled vehicles and, in particular, school buses.
6
The Air Quality Section of the NC Department of Environmental, Health, and Natural Resources (DEHNR) has a RSD. The Air Quality Section agreed to make this device available to NCSU for the purposes of collecting data on uncontrolled vehicles. In this project, the RSD has been employed to measure emissions from both school and transit buses. Remote sensing has been employed over the last few years primarily to help improve inspection and maintenance (I/M) programs. Conventional I/M programs require all of the light-duty vehicles in a given region be tested annually or biennially for emission performance. They do not discriminate between clean and dirty vehicles and therefore all the motorists are required to bear the cost and inconvenience of testing. In California's 1991 decentralized, biennial inspection and maintenance (I/M) program, only 20 percent of the vehicles tested required repairs under the program rules (Rueff, 1992). Thus 80 percent of the vehicle owner time and the testing costs resulted in no air quality benefit. Therefore selective I/M programs, utilizing a simple, reliable means of identifying high emitting vehicles from the in-use fleet would be vastly more efficient then the conventional I/M programs. Such programs would allow the regulatory authorities to ease the cost burden for the motorists who purchase and maintain low emitting vehicles. In studies conducted at Denver, Los Angeles, and Chicago, it has been observed that roughly half of the mobile source CO can be traced to a small fraction of the general fleet (Lawson et al., 1990). California's 1989 random roadside survey indicated that 10 percent of the light-duty fleet is responsible for 60 percent of mobile source CO emissions and that 10 percent of the fleet is responsible for 60 percent of exhaust HC (Rueff, 1992). Data collected by the Department of Environmental, Health and Natural Resources (DEHNR) of Raleigh, North Carolina in September 1994 using a remote emissions sensor supports the results of the roadside survey carried out in California. DEHNR’s remote sensor is a RES-1 “Smog Dog” developed by the Santa Barbara Research Center of Hughes Aircraft. The results of the "Smog Dog" remote sensor are shown in Figure 1-1. The data in Figure 1-1 indicate that of the population of vehicles for which instantaneous emission measurements were taken, only a small percentage appear to have very high emissions. Similarly, a small portion of the population accounted for HC concentrations representative of over half of the total. However, it is important to bear in mind that the remote sensor captures nearly instantaneous emissions readings for vehicles
7
which may be operating in a wide variety of different modes (e.g., speed, acceleration) and under different conditions (e.g., fuel composition, maintenance, etc.). Therefore, these estimates may overstate the total contribution of the highest emitting vehicles. Nonetheless, the results do indicate that a relatively small portion of the on-road vehicle fleet accounts for a substantial portion of total emissions. In light of these findings, selective I/M programs could also provide a cost effective means of reducing fleet wide motor vehicle emissions.
CO concentrations Versus Fraction of Vehicles (ranked by emissions) 14 12 10 8 6 4 2 0 0
10
20
30
40
50
60
70
80
90
100
80
90
100
Fraction of vehicles (Ranked by Emissions) (%)
(a) Remote Sensing Data for CO Emissions HC concentrations Versus Fraction of Vehicles (ranked by emissions) 6 5 4 3 2 1 0 0
10
20
30
40
50
60
70
Fraction of vehicles ( Ranked by Emissions) (%)
(b) Remote Sensing Data for HC Emissions Figure 1-1. Example of Remote Sensing Data Obtained in Raleigh, NC for 1,027 Vehicles Using the RES-100 "Smog Dog" Remote sensing can be used to monitor the emissions from thousands of vehicles on the road. It can constitute a very effective method of targeting high emitting vehicles. Besides serving as an effective I/M tool, remote sensing of vehicle exhaust emissions would also help improve our understanding of the real-world emissions since all the 8
emissions measurements would be carried out on the road (Cadle and Stephens, 1994).
9
2.0
REMOTE SENSING: THEORY AND OPERATION
Remote sensing equipment (RSD) for highway vehicles was developed to provide a method via which exhaust emission resulting from actual operating conditions might more accurately be estimated (Stephens and Cadle, 1991; Stedman, 1989). The University of Denver introduced remote sensing devices (RSDs) for exhaust carbon monoxide (CO) in 1987. They were followed by General Motors (GM) R&D Center in 1988. Simultaneous determination of hydrocarbon (HC) emissions was added to the sensors in 1990. The need for measurement of on-road emissions is driven by concerns of inherent inaccuracies in currently available technologies, such as the use of chassis and engine dynamometer testing for the purpose of measuring average emissions for a specified driving cycle under laboratory conditions. A limited number of standardized driving cycles underlie emissions factor models such as the U.S. Environmental Protection Agency’s Mobile models (e.g., Kini, 1996). 2.1
Remote Sensing Device (RSD) Equipment
Figure 2-1 shows the schematic diagram of a remote sensor deployed at a measurement site. The components shown in the diagram are generic to highway vehicle remote sensing technologies. The emission data taken in the course of this study utilized NC DEHNR’s RES-1 “Smog Dog.” Remote sensing equipment is typically housed in a cargo van which contains on-board computer, calibration, and video systems (including video monitors and video cassette recorder), as indicated in Figure 2-2. An infrared (IR) source is used to transmit an infrared beam that is detected by an infrared receiver. The source and receiver for the RES-1 “Smog Dog” are shown in Figure 2-3, and the deployment of these devices is illustrated in Figure 2-1. Also shown in Figure 2-3 is the portable generator used to provide power to the infrared source, and the calibration gas cylinders. In addition, remote sensing systems have a receiver-alignment module which is used for alignment of receiver to source, a camera for viewing the license plates of passing vehicles, and a video camera for recording images of the back of each passing vehicle. The set up of the two cameras is shown in Figure 2-4. The portable electric generator and the exhaust tube for the on-board generator in the van are located downwind of the infrared beam, to avoid interference with the vehicle emissions measurements. Traffic cones are used on both sides of the road to warn traffic away from the equipment. Remote sensing is a fair weather technology, and it is not deployed under rainy conditions.
10
On-board Generator Exhaust Video Camera
Infrared Receiver
Cables
Traffic Cones
Cables
Camera for "Tag" Reader
RSD Van Passing Bus Infrared Beam
Infrared Source
Wire
Portable Electric Generator
Figure 2-1. Simplified Schematic of a Deployed Remote Sensing Device (not to scale). The infrared measurement system for remote vehicle emission measurements operates using a technology known as Non-Dispersive Infrared Absorption (NDIR), which is described in more detail in Section 2.2. The infrared detector measures the amount of infrared energy traveling through the exhaust plume (SBRC, 1994). The detectors are equipped with narrow band optical filters to isolate the absorption band of the gas to be detected. Post-car signals are acquired immediately after a vehicle passes through the infrared beam, and ambient air signals are acquired when there is no traffic flow for several seconds (Cadle and Stephens, 1991). The infrared beam is set at a height where it will be obstructed when a vehicle passes in front of it. The height and initial alignment of the beam is determined manually through a sometimes difficult iterative process. A low power laser is mounted inside the infrared source. Both the source and receiver are initially set up based upon an estimate of what the road-side elevation of each device must be in order for the beam to be a specified height from the pavement at the location of a typical vehicle exhaust, as illustrated in Figure 2-5. With experience, it is possible to initially align the source so that the laser beam can be seen on the front edge of the box containing the infrared receiver. When the laser beam is located at the top right corner of the receiver box, then a measurement of the beam height over the pavement is made. This is done by first ensuring that there is no vehicular traffic, and then by manually placing a card or piece of
11
Figure 2-2. View of the RSD setup at the inner gate of the Rock Quarry Road school bus depot. In nearground is the infrared source. In front of the van is the infrared receiver. To the right in front of the fence are the video cameras.
Figure 2-3. Photograph of the interior of the RES-1 “Smog Dog.” Shown are the computer console (center), video monitor (right), and monitor for the license plate tag reader (left).
12
Figure 2-4. Close up of the infrared source (black box in upper lefthand corner), infrared receiver (white box in lower left hand corner), portable electric generator (near center) and calibration gas cylinders (right hand side).
13
Infrared Receiver Infrared Source Laser Beam Used for Initial Alignment and Height Adjustment Beam Height
Figure 2-5. Schematic of Method for Adjusting Beam Height paper in the path of the laser beam at the location over the pavement where the exhaust pipes are typically expected. Using a yardstick, the distance from the pavement surface to the laser beam is measured. This distance is the estimate of the beam height. The height of the beam must be such that it will both be blocked by the passing vehicle and pass through the subsequent exhaust plume. For automobiles, the beam height is typically 17 inches. For school buses, the beam height was typically adjusted to approximately 24 inches. Because the RSD equipment was not designed for use at the significantly higher beam height required for buses, it was necessary to improvise in order to further elevate the beam. Cinder blocks were found to provide a stable and inexpensive means for elevating the source and the receiver, as illustrated in Figure 2-6 for the receiver and Figure 2-7 for the source. Figure 2-7 also illustrates that the beam must be high enough to be blocked by the vehicle as it passes. Thus, the beam is typically higher than the centerline of the tailpipe. The final alignment of the beam is done using an electronic alignment module. This module provides a visual representation of the strength of the infrared signal. The final alignment is typically accomplished by making minor adjustments to the height and angle of the receiver using the tripod shown in Figure 2-6. The strength of the infrared signal is also displayed on the computer console as a voltage. RSD equipment provides the user with an inference regarding the ratio of carbon monoxide (CO) to carbon dioxide (CO2) in the exhaust and the ratio of unburned Hydrocarbons (HC's) to Carbon Dioxide (CO2). The on-board computer in the RSD also estimates the volume percentage of CO, CO2, and unburned HC's present in the exhaust. These latter calculations are based upon assumptions regarding fuel composition and the fuel-to-air ratio during combustion. Information on individual vehicles and fleet composition greatly assist in the effective use of the remote sensing technology. Therefore many remote sensing systems have incorporated a video camera that is focused on the license plates of the passing
14
Figure 2-6. View of the infrared receiver elevated to a height of approximately 24 inches above the centerline of the road at the Laura Duncan Road site.
Figure 2-7. View of the infrared source and a passing school bus. The source is elevated slightly above the tail pipe so that the beam is broken by the passing bus.
15
vehicles. A "frame grabber" captures the image of the license plate and, using a character recognition program, records the license plate number. The license plate data are integrated with the remote sensor output. The license plate numbers can be used with registration data to obtain a vehicle's age, make, and model. Vehicle identification numbers can be obtained from the state registration database to provide more information about the vehicle. In early versions of remote sensor devices (RSDs), the license plate numbers were read manually from the video records. A reliable license plate reader eliminates this laborious process (Stephens, 1994). An Automatic License Plate Recognition (ALPR) system which includes a separate computer, software, video camera and monitoring system is included, for example, with the commercially available Hughes RES-100 "Smog Dog" RSD (Gorse et al, 1994). The RES-100 Smog Dog instrument used in the surveys carried out in North Carolina have experienced some difficulty with the ALPR system. In early measurements, approximately 30 percent of the license plates could not be read by the ALPR. This problem is in the process of being corrected by Hughes. The difficulty stems from the differences in lettering styles on North Carolina license plates versus California plates, for which the ALPR system was originally calibrated. For this study, the license plate reader was not used for data collection. Instead, the video camera was used to record images of each passing vehicle, which were reviewed to confirm data logbook entries made in the field regarding the identification number of each bus. The identification numbers for school buses, for example, are painted on the sides of the bus near the front and on the center in the back. In the following sections the theory behind remote sensing measurement and interpretation will be discussed. The setup and operation of the specific equipment utilized for this study is also discussed. 2.2
Non-Dispersive Infrared Absorption
The exhaust emissions data yielded by the RSD equipment results from the measurement of infrared absorption that occurs within a well-defined region of the infrared spectrum. Every molecule is made up of atoms which are in constant motion with respect to one another. For every specific molecular compound there are characteristic frequencies at which the atoms, which make up the structure, vibrate. The vibrational frequencies of a molecule are a function of the characteristics of the bonds existing between the atoms of the structure. The atomic vibrational spectrum corresponds closely with the infrared light spectrum. When infrared light is exposed to a particular compound 16
that part of the spectrum which has the same wavelength as the vibrational frequencies of the atomic bonds is absorbed (Brady and Humiston, 1978). By carefully observing how much specific regions of the infrared spectrum are absorbed one can deduce as to the structure and relative concentration of molecules responsible for the absorption. The various RSDs operate on the same general principle. An infrared beam is continuously directed across a lane of traffic to a detector. The detector works on the principle of selective absorption of infrared radiation by CO, CO2 and HC gases. It uses band pass filters to isolate the HC, CO, and CO2 absorption regions of the spectrum. It also employs a non-absorbing region which is used to monitor the beam intensity. A vehicle passing through an infrared beam triggers the receiver to collect transmission data on the exhaust plume of the passing vehicle. A computer analyzes the data and determines the effective percentage of each pollutant. Multiple detectors can be used to continuously monitor the frequency regions so that integrated concentration over the beam path can be determined simultaneously. The beam is typically positioned approximately 17 inches above the road surface, which is the average height of the lightduty tailpipes. This height can be adjusted for other types of vehicles, such as school buses. 2.3
Data Acquisition
The infrared source and receiver, along with a computer for control and data storage, are the principle components of the RSD system. The system utilizes a highly coherent infrared source combined with mercury-cadmium-telluride infrared detectors (receivers). Under normal conditions the data from the receiver is sent to the computer every one-thousandth of a second. The receiver has four channels; one for each of the species of interest, and one for calibrating out any signal attenuation due to atmospheric interference. When processing data, the RSD must compare the transmittances of the beam through the exhaust plume with the transmittances of the beam during background conditions. This is so that the exhaust readings can be corrected for background levels of CO, CO2, or HC. The RSD can operate in several modes with respect to background readings. One is to take a background reading each time before a vehicle passes. In this mode, the computer does not process the signal from the receiver unless it has received a trigger signal that occurs when the beam is broken. This occurs when the front end of a vehicle enters the path of the beam. At this point, the computer maintains in memory data for the most recent time interval specified by the user, which is typically two seconds minus the 17
user-specified time interval for obtaining data from the exhaust plume. For example, if the user specifies that data regarding the exhaust plume should be collected for 0.6 seconds, then the computer will retain the 1.4 seconds of data preceding re-establishment of the beam. If the vehicle takes 1.0 second to pass through the beam, then there will be 0.4 seconds of background data (i.e. 0.4 seconds of data before the beam is broken, 1.0 seconds during which the beam is broken, and 0.6 seconds during which the beam is reestablished and passes through the exhaust plume). When a vehicle exits from the beam path, a second trigger signal is sent to the computer. The second trigger has two purposes; it alerts the computer when to collect data for the exhaust plume, and it invalidates a measurement in the event that the second trigger is not received within a given time period from the first trigger. Thus, for example, if 0.6 seconds of exhaust plume data collection are required, and if the vehicle takes more than 1.4 seconds to pass through the beam, then no background data will be available and the measurement will be invalid. The time limitation on the measurements can cause difficulty in the event that the vehicles under investigation are long and speeds at which they are traveling are slow. After the second signal has been received there is a user selectable window of time of between 0.4 and 1.5 seconds during which data from that vehicle is acquired and analyzed. An alternative to taking background readings for each individual vehicle is to manually take background readings separately from the measurement of each vehicle. In the manual background reading mode, it is possible to acquire exhaust data even for slow moving or long vehicles. However, the trade-off is that the time between the manual background readings and the measurement of the exhaust plume can be large. Therefore, the representativeness of the background reading for the ambient atmospheric conditions at the time a vehicle passes by may be unknown and contribute additional uncertainty to the measurement. During the study, efforts were made to take manual background readings on a regular basis to minimize this problem. The on-board computer calculates the ratios of CO/CO2 and HC/CO2 using regression analysis of the exhaust plume data. Although not specifically documented by the manufacturer, our understanding is that the background readings are used to provide an estimate of the receiver voltage levels associated with the average ambient concentrations of CO, CO2, and HC. These voltage levels are then compared to those obtained during the measurement of the exhaust plume. The difference between the instantaneous exhaust plume readings and the ambient levels indicates the additional
18
amount of CO, CO2, and HC in the path of the beam due to emissions from the vehicle. The differences between the exhaust and ambient readings are plotted on the computer screen as CO versus CO2 and HC versus CO2. The actual levels of CO, CO2, and HC in the exhaust plume are continuously changing as the exhaust plume enters the path of the beam and, subsequently, as the exhaust plume disperses. However, the ratio of the pollutants should remain constant. Therefore, linear regressions of the two plots are used to estimate the slope, which yields the ratios for CO/CO2 and HC/CO2. If the regression is considered to be valid, the slopes are recorded in the data file. If the results are considered to be non-valid, the slopes are not recorded and an error code of 9999 is entered into the data file. The on-board computer also performs calculations regarding the estimated volume percentages of each pollutant in the vehicle exhaust. These calculations are based upon assumptions regarding fuel composition and the air-to-fuel ratio. The assumptions are based upon gasoline-fueled vehicles. The assumptions are not valid for diesel vehicles, and probably are not valid for gasoline vehicles operating in specific driving modes. Therefore, we did not make use of the emission concentration estimates generated by the on-board computer. In Chapter 4, we discuss the theoretical basis for further analysis of the RSD data in order to make inferences regarding emission factors. In addition to the numerical data obtained from the infrared sensor and the onboard computer, video cameras are used to provide a visual record of the measured vehicle. Together this equipment provides the user with a data file consisting of emission measurements of specific vehicles and a videotape by which vehicles may be identified. Error checking within the computer insures that, in the event of a near miss of the exhaust plume due to a high release point, heavy traffic, etc., readings outside of normal bounds will be marked invalid. Other important equipment in the RSD-100 system includes equipment for calibration. 2.4
Calibration
Three separate calibration routines help to insure the accuracy of the RSD system. The system is first calibrated at the factory. It is then calibrated before each use. Finally, the system calibration can be verified during use. The factory calibration consists of the measurement of a commercially available mixture of CO, CO2, and HC gases. These gases, in concentrations accurate to within one percent, are measured with different diluent concentrations by an optical cell of 19
known transmittance. The resulting photovoltaic detector voltages vs. gas concentration data are stored. A curve-fitting polynomial equation is then generated for each detector, and this equation is stored in the on-board computer system for use in the RSD system (SBRC, 1994). Field calibration must be performed daily, prior to data taking, and as warranted by changes in ambient conditions. Field calibration is necessary to compensate for variations in the background CO2 levels. In this procedure the CO sensor is used as the absolute standard, and the system is exposed to a calibration gas consisting of 15 percent CO, 15 percent CO2, and 2.5 percent HC. The calibration gas is introduced to the path of the infrared beam via a tube from the on-board calibration gas cylinders to the infrared receiver. The calibration gas corresponds to a CO/CO2 ratio of 1.0 and a HC/CO2 ratio of 0.167. The resulting RSD measurements of the calibration gas are required to be within a specified tolerance of these ratios of gas concentrations. This forces the concentrationversus-time curves for CO and HC's to be in the correct ratio at all times. The resulting voltage-versus-concentration data is used to modify the factory-generated polynomial curve-fit equations to account for variations in CO2 levels. A third calibration method, the 'Puff In Vehicle Mode' (PIV), offers a method to verify accurate system operation while data taking is underway. When manually selected, this procedure causes a sample of the field calibration gas to be 'puffed' through the receiver optical cell. The resulting CO/CO2 and HC/CO2 ratios must agree with the calibration gas to within 10 percent. If not, then a calibration must be done. The PIV mode offers a means by which the precision of measurements may be ascertained, by allowing for repeated readings of a gas of known composition. The calibration gas composition is intended to be typical of the ratios of pollutants that would be observed in a high emitting vehicle. Thus, the measurements are likely to be most accurate for vehicles with exhaust pollutant ratios similar to that of the calibration gas. 2.5
Remote Sensor Accuracy
The accuracy of detection of individual measurements are a function of how the exhaust plume disperses and what part of the plume is intersected by the RSD beam. Thus it is difficult to give absolute accuracy for the RSD. Remote sensing measurements compare well with measurements made using on-board instrumentation. For example, in one study by the University of Denver it was found that CO emissions were measured
20
within ± 5 percent and HC emissions were measured within ± 15 percent using the two techniques (Ashbaugh et al., 1992). Overall accuracies have been reported by GM as ± 15 percent for the CO/CO2 and HC/CO2 ratios (Cadle and Stephens, 1994). In another study, the performance of remote sensors in detecting hydrocarbons was compared to that of other devices. These included a gas chromatograph (GC), a flame ionization detectors (FID), a Fourier transform infrared spectrometers (FTIR), and a commercially produced non-dispersive infrared analyzer (NDIR). These instruments were used to measure HC concentrations in a variety of samples, including ten HC species, twelve vehicle exhaust samples, and three volatilized fuel samples. The exhaust samples were obtained under various vehicle operating conditions (e.g., fuel rich). The baseline instrumentation was that which is typically used in dynamometer test cell. Such test cells typically employ a FID device and a NDIR for measurements of CO and CO2, respectively. Compared to several of the other techniques, remote sensing techniques tend to do a good job of measuring alkane hydrocarbon emissions, but are not as good at measuring other types of hydrocarbons such as alkenes and aromatics. Examples of aromatics include toluene and xylene. Remote sensors are good at measuring alkanes because the hydrocarbon measurements are based on a single infrared region which is most sensitive to this type of hydrocarbon, and to propane in particular. Thus, the response of remote sensors to the presence of straight-chain alkanes is very similar to that of other measurement techniques. However, the ratio of straight-chain alkanes to total hydrocarbons is not constant from one vehicle to another or as a function of operating conditions. Therefore, the accuracy of remote sensing measurements of HC is variable, depending upon the speciation profile of the exhaust. This is a result of the variability of the infrared absorption coefficients for the various HCs over the range of infrared wavelengths used in the HC measurement. Therefore, remote sensors that use different infrared regions for HC measurements will yield different results with potentially different accuracies (Cadle and Stephens, 1991). Remote sensing for hydrocarbons could be improved if the devices used multiple infrared regions for measurements of HC and multi-component HC mixtures for calibration (Stephens et al., 1994). However, this will require further theoretical and experimental work on the part of the developers and vendors of the remote sensing technology. The RES-1 system has been shown by the manufacturer to provide sensitive and highly repeatable results. The detectors used in the system have been reported to have a sensitivity for CO of 50 percent by volume). Olefins generally account for between 0 and 15 percent of the hydrocarbons by volume. Aromatics account for between 15 and 40 percent, and napthenes account for the remainder (SAE, 1988, Schoonveld and Marshall, 1991). Table 5-1 documents the major hydrocarbons identified in examples of two fuel analyses. The ASTM D-1319 standard, Test Method for Hydrocarbon Types in Liquid Petroleum Products by Fluorescent Indicator Absorption, provides a breakdown of the relative concentration of major hydrocarbon families in the fuel. Non-standardized tests exist which permit the identification of specific hydrocarbon compounds and an ultimate analysis for the ratio of carbon to hydrogen in the fuel. A quantitative analysis of hydrocarbon types in motor fuels is beyond the capability of the North Carolina Motor Fuels Laboratory (Sutton, 1995). The elemental composition of gasoline is generally considered to consist of between 84 and 86 weight percent carbon and 14 to 16 weight percent hydrogen (Ward, 1979; Robert Bosch GmbH, 1986). In analyzing RSD measurements, investigators at the University of Denver used a molecular formula of CH2 (85.6 percent C, 14.4 percent H) for gasoline to derive the percentage measurement for CO2, CO, and HC's (Bishop, 1996). Similarly Obert (1973) uses isooctane, C8H18 (84.1 percent C, 15.9 percent H), as a gasoline equivalent. Gasoline has been characterized as a mixture of hydrocarbons primarily within a range of four to ten carbons atoms per molecule (Hirao and Pefley, 1988). An estimate of the range of elemental compositions can be developed by looking at the chemical formulation for the constituent members of the hydrocarbon families found to make up gasoline.
53
Table 5-1 Fuel Composition - Examples of Detailed Hydrocarbon Speciation ________________________________________________________________________ Fuel Fuel A B A B Compound Wt% Wt% Compound Wt% Wt% ________________________________________________________________________ Isobutane Isobutene 1-Butene Butane trans-2-Butene 2,3,-Dimethylhexane cis-2-Butene 3-Methyl-1-butene Isopentane 1-Pentane Pentane 2-Methyl-1-butene 5-Methylnonane 4-Methylnonane trans-2-Pentene cis-2-pentene 2-Methyl-2-butene 2,2-Dimethylbutane 4-Methyl-1-pentene Cyclopentane 2,3-Dimethylbutane 2-Methylpentane 3-Methylpentane Hexane 2,4-Dimethylpentane 2-Methylhexane Benzene Heptane Methycyclohexane Toluene 2-Methylheptane 3,4-Dimethylhexane Unknown C6 Alkene Unknown C7 Alkene Unknown C8 Alkene Unknown C9 Alkene Unknown C10 Alkene Unknown C10 Alkane Unknown C11 Alkane Methyl-t-butyl ether
0.1 0.0 0.0 5.3 0.0 0.2 0.0 0.0 6.1 0.2 1.0 0.3 0.1 0.1 0.3 0.6 1.6 0.2 0.2 0.2 1.1 3.0 1.8 0.7 1.9 1.2 1.4 0.6 0.8 10.5 0.3 1.0 0.7 2.2 2.2 1.4 0.9 2.2 1.7 0.0
0.1 0.0 0.0 3.0 0.1 0.1 0.0 0.0 9.7 0.2 1.3 0.2 0.2 0.2 0.2 0.4 1.2 0.2 0.1 0.3 1.5 4.1 2.1 0.7 2.6 1.2 0.9 0.6 0.6 4.4 0.2 0.8 0.7 1.8 1.8 1.3 0.8 1.9 1.4 5.3
2,2,5-Trimethylhexane Octane Ethylbenzene Paraxylene Metaxylene 1,1,2-Trimethylcyclepentane Orthoxylene Nonane n-Propylbenzene 1-Methyl-3-ethylbenzene 3,4,4-Trimethylheptane 3,4,5-Trimethylheptane 2-Methyl-1,3-butadiene 1-Methyl-2-ethylbenzene 3-Methyl-cis-2-pentene 5-Methylnonane 4-Methylnonane 1,3,5-Trimethylbenzene 1,2,4-Trimethylbenzene Decane 1,2,3-Trimethylbenzene Undecane 3-Methylheptane Methylcyclepentane 2,2,4-Trimethylpentane 2,3-Dimethylpentane 2,2,3-Trimethylpentane 2,3,4-Trimethylpentane 2,3,3-Trimethylpentane 2-Methyl-3-ethylpentane 4-Methylheptane Unknown C5 Alkene Unknown C6 Alkane Unknown C7 Alkane Unknown C8 Alkane Unknown C9 Alkane Unknown C10 Aromatic Unknown C11 Aromatic Unknown C12 Alkane
0.9 0.4 1.7 3.6 1.6 0.9 2.1 0.2 1.5 0.8 0.4 0.4 0.3 0.1 0.2 0.1 0.1 0.1 2.1 0.2 0.1 0.1 0.1 1.1 6.4 1.9 1.1 1.2 1.2 0.9 0.2 0.5 0.3 2.3 1.3 1.8 2.4 2.4 1.2
1.2 0.4 1.4 2.4 1.3 1.0 1.4 0.2 1.0 0.6 0.3 0.3 0.2 0.1 0.1 0.2 0.2 0.1 1.5 0.2 0.0 0.0 0.1 1.0 9.5 2.6 1.5 1.6 1.6 0.7 0.3 0.4 0.2 1.8 1.0 1.7 2.0 2.0 1.1
________________________________________________________________________ Fuel A - Unleaded, 89 Octane Fuel B - Unleaded, 'Reformulated' Gasoline w/MTBE, 89 Octane (Source: Schoonveld, et.al, SAE 910380) 54
The paraffin (alkane) family accounts for most of the hydrocarbons found in gasoline. The members of this family exist as either straight-chain molecules or as branched-chain molecules (isoparaffins). The general elemental formula for the members of this family follows the form CnH2n+2. Thus, for long chain paraffins, the carbon-tohydrogen ratio asymptotically approaches 1:2. Using the general formula for alkanes, and considering alkanes containing between 4 and 10 carbons, we find that carbon accounts for between 82.8 percent and 84.5 percent of the total mass in these hydrocarbons. Constituents of the olefin and naphthene families of hydrocarbons have different molecular structures. Naphthenes are characterized by carbon atoms arranged in a ring structure. However both olefins and naphthenes have the same general elemental formulation of CnH2n. Thus these compounds have a carbon content of 85.6 percent of the total mass for these molecules. Aromatics are the final major hydrocarbon family that comprises gasoline. Elemental formulations of aromatic compounds generally following the elemental form CnH2n-6. Carbon accounts for between 89.5 and 96 percent of the total mass of an aromatic for values of n between 4 and 10. The elemental compositions presented for these individual hydrocarbon families are close to the range of carbon and hydrogen found in unrefined crude oils. For crude oils, carbon has been found to make up between 84 and 86 percent of the weight, hydrogen between 10 and 14 percent, and the remainder of the fuel is made up of small percentages of sulfur, nitrogen, and oxygen. 5.3.4
Sulfur Content
Unleaded gasolines are limited to 0.1 percent sulfur by mass in accordance with ASTM D-439. In practice the average sulfur content of gasoline sold in the United States is approximately 0.3 percent by mass (SAE, 1988). Transportation is not seen as a major contributor of sulfur oxides to the atmosphere (Flagan and Seinfeld, 1988). However, even low sulfur level can lead to the formation of acids, which can have a detrimental affect on engine components. The copper strip tarnish test, ASTM D-130 is used to detect the presence of corrosive sulfur compounds. 5.3.5
Density
Density of motor fuels is usually given in terms of specific gravity or 'relative density'. This measurement refers to the ratio between the weight of a specific volume of gasoline to an equivalent volume of water. For the purposes of establishing standards between manufacturers this measurement is made at 15 oC (60 oF), 101.325 kPa (1 atm.). The density of gasoline can be determined using ASTM D-1298, Test Method for Relative Density (Specific Gravity). An older term which is still commonly applied, 55
degrees API, is based on an arbitrary hydrometer scale developed by the American Petroleum Institute, which is also related to specific gravity (Owen and Coley, 1995) as follows: Degrees API =
141.5 – 131.5 Specific Gravity (at 60 o F, 1 atm)
For example, water has a specific gravity of 1.000 and a degrees API gravity of 10o API. The relative density is useful for converting volumes to weight and is sometimes used as a way of identifying gasolines. The relative density of the motor fuel can affect the air-tofuel ratio at which a vehicle runs. Fuel injection systems are design to precisely meter the volume of fuel which is injected into each cylinder. Thus as the density of the fuel changes the mass of fuel injected into the cylinder will change unless the engine is specifically design to account for this discrepancy (Owen and Coley, 1995). Gasoline is generally considered to have a specific density of between 0.72 and 0.75; this corresponds to an API of between 65 to 57 (Owen and Coley, 1995; Robert Bosch, GmbH, 1976; Ward, 1979). The density is highly dependent on the type of blending used to produce the final product, as well as the temperature. 5.3.6 Heating Value The energy content of gasoline is expressed as its heating value. The heating value of gasoline can be determined using ASTM D-240, Test Method for Heat of Combustion of Liquid Hydrocarbon Fuels by Bomb Calorimeter or by ASTM D-2382, Test Method for Heat of Combustion of Hydrocarbon Fuels by Bomb Calorimeter (High Precision Method). The typical value for the lower heating value of gasoline is approximately 44 MJ/kg (SAE, 1988). Variation in the volumetric heating value of gasolines has been found to vary with specific gravity. It has been found that the typical extremes found in the heating value, and the associated variation in the specific gravity, of gasoline can account for a three percent variation in fuel economy (SAE, 1988). 5.3.7
Additional Considerations
A variety of additives, including marking dyes, corrosion inhibitors, drag reducing compounds, antistatic additives, and biocides are mixed with gasoline to promote its handling characteristics for distribution through pipelines and other transport methods. Antioxidants are added to increase the storage life. Corrosion inhibitors, anti-icing compounds, and fuel system detergents are added to enhance vehicle performance. Though a wide range of standards apply to the quality of gasoline, the wide range of
56
blending stocks and additives used by distillers leads to a high degree of variability in the chemical makeup in different manufacturers gasolines. A number of specifications are used to regulate the quality of gasoline. The ASTM standard D4814-93a was released in 1988 to cover gasoline and its blends with oxygenates such as alcohols and ethers (Owen and Coley, 1995). This standard assigns volatility classes to different mixtures of gasoline, and it sets other limits for the lead content, sulfur content, and the oxidation stability, and copper content, among other things. There are dozens of ASTM standards which apply to specific characteristics of gasoline. Regulations have been developed to limit the corrosivity of gasoline as well as its heating value and conductivity. An important characteristic of gasoline is its fugability, or its compatibility with other suppliers formulations. This is important because refineries often exchange products to make up for shortcomings in their own processes. In addition, the gasolines from various suppliers must have the capability to mix in the consumers gas tank. 5.4
Diesel Fuel
The same refining processes used for gasoline production also apply to diesel fuel production. Diesel fuels are typically pulled from the primary distillation stream from the refinery, but production yields may be improved by blending with fuel streams from more complex processes, such as thermal cracking or catalytic hydrogenation. The diesel engine was developed in the late 19th Century, and early applications were limited to industrial and marine uses. Initial interest in diesel engines was promoted by the difficulty in obtaining gasoline, primarily in Europe. Before long it was recognized that diesel powered vehicles were capable of better fuel economy than similar gasoline fueled vehicles. Diesel engines utilized for highway use were generally higher revving, and of a higher compression ratio, than their counterparts used in industrial settings. This lead to more stringent diesel fuel requirements for highway vehicles than for their industrial counterparts. Initial improvements in diesel fuel quality focused on reducing viscosity and the level of hard combustion residues. As engines improved further, the focus was shifted toward improving the ignition quality of the fuel as well as its cold weather performance. The diesel fuels in use today are the result of crude oil processor's efforts to maximize the yield from their raw materials while maintaining the fuel quality required by modern engines. Much of the information presented in Section 5.3 on the properties of gasoline
57
also applies to diesel fuels. In the sections that follow, some of the other important qualities of diesel fuel will be discussed. 5.4.1
General Characteristics of Diesel Fuel
Ignition quality is an important characteristic for diesel fuels, but this property must be balanced with the fuel density, viscosity, and volatility characteristics to function properly. In the operation of a diesel engine, a precisely metered amount of fuel is injected into a combustion chamber containing air under pressure. The fuel is injected near the end of the compression stroke of the engine, and as the mixture is compressed further the additional heat of compression induces a spontaneous combustion reaction of the mixture. In many ways, the fuel delivery system of early diesel engines was more sophisticated than the gasoline fueled cars of the same era, and the fuel injection systems being used on many of today's gasoline powered cars are similar to the fuel delivery systems long used on diesel engines. On its path from the fuel tank to the engine, the diesel fuel is filtered, pumped, and precisely metered. Thus, there is a narrow range for the viscosity of a properly performing fuel. Diesel fuel must also possess satisfactory qualities for cleanliness, corrosivity, and have a low tendency to produce engine deposits. For the consumer, the most common measure of a diesel fuel is its ignition quality, or cetane number. 5.4.2
Ignition Quality
The cetane number is a measure of the ignition quality of a diesel fuel; the higher the cetane number, the better the ignition quality. The cetane number represents the ratio of the percentage by volume of n-cetane, C16H34 (n-hexadecane) to that of alphamethylnapthalene, C11H10, with which the same ignition lag is determined in a test engine as with the test diesel fuel. The cetane number 100 is assigned to pure cetane, which has very good ignition qualities, while methylnapthalene (poor ignition qualities) is assigned the cetane number 0 (Robert Bosch GmbH, 1976). In essence, the cetane number is a measure of the ignition delay of the fuel. The ignition lag is usually measured in degrees of crank angle the engine turns through between the time the fuel is injected and combustion occurs. Poor engine performance occurs when a fuel of poor ignition quality, or too low of a cetane number, is used and pre-ignition of the fuel occurs. Conversely, usage of fuels with a much higher cetane number than required results in excess emissions of soot and unburned hydrocarbons due to incomplete combustion. In general the cetane requirement of a diesel engines increases slightly as the engine wears, due to a
58
drop in the compression ratio with age. In the United States the cetane numbers for diesel fuel typically range between 42 and 50. 5.4.3
Additives
Early diesel fuels, before the 1970's were largely a blend of products produced by simple distillation with little or no performance additives. In the 1960's additives began to be used to enhanced the cold weather performance of fuel. The oil embargo of the 1970's lead to changes in refinery practices which reduced diesel fuel yield. In addition, reduced demand for lower quality oil products lead to the increase usage of these fuels for diesel production. The lower quality of these raw materials often resulted in a diesel fuel with a poor ignition quality. Thus a variety of chemicals have been developed for use as cetane enhancers. Additionally additives are combined with diesel fuel to prevent corrosion in the fuel system. Antioxidants and stabilizers are added to the fuel to prevent breakdown in storage. Dispersants are added to prevents deposit formation of gums or sludge which could clog filters, injectors, and engine parts. 5.4.4
Classifications
Three types of diesel fuel are commonly found in the United States; diesel No. 1 (1-DA), diesel No. 2 (2-DA), and diesel No. 4 (4-DA). Diesel No. 1 and Diesel No. 2 are used for highway vehicles and industrial application, with 1-DA diesel being most common for highway use. No. 4-DA diesel is a lower quality blend of distillates, compared to 1-DA and 2-DA diesel, which is used for low speed engines or nonautomotive applications. In the United States these fuels are regulated for the amount of sulfur they contain as well as the residual carbon content, viscosity, water and sediment, cetane number, and the flash and cloud temperature points (ASTM D975-91). 5.4.5
Hydrocarbon Composition
Diesel fuel, like gasoline, is a complex mixture of many different hydrocarbons. The hydrocarbon composition influences many of the fuel's properties, including ignition quality, heating value, volatility, gravity, and oxidation stability. Diesel fuel has been found to consist of 86.25 percent carbon and 13.25 percent hydrogen (Robert Bosch, GmbH, 1976). Approximately 0.51 percent of the mass of the fuel is allowed to consist of a combination of ash and sulfur; water and sediment are allowed to make up 0.2 percent of the volume (ASTM D-975). The density of diesel fuel is commonly expressed in units of API gravity. Diesel fuel has been found to possess a density of between 0.81
59
and 0.85 kg/l (Robert Bosch, GmbH, 1976), and this corresponds to an API gravity of between 43o and 35o. Table 5-2 Summary of Fuel Properties ________________________________________________________________________ Fuel Property Gasoline Diesel ________________________________________________________________________ Density (kg/l) 0.72 - 0.78 0.81 - 0.85 Chief Constituents Carbon (wt-%) 86 86.25 Hydrogen (wt-%) 14 13.25 Heating value (kJ/kg) 43.5 40.6 - 44.4 Stoichiometric air requirement (lb air/lb fuel) 14.8 14.8 ________________________________________________________________________ 5.5
Derivation of Emission Factors
Remote sensing devices, like the one used for this study, measure the absorption of an infrared beam in the plume of a moving vehicles exhaust to determine the ratio of carbon monoxide (CO) to carbon dioxide (CO2) and the ratio of unburned hydrocarbons (HC) to CO2 present in the exhaust. From the measured ratios of CO to CO2 and HC to CO2, and by assuming an elemental composition for the fuel, the RSD equipment also calculates a theoretical volume percentage for the gases in question. The exhaust gas volume percentages estimated by the remote sensing equipment are based on a fixed ratio of carbon to hydrogen in the fuel of 1:2 and assume a specific air-to-fuel ratio. The following derivation will demonstrate how other fuel compositions might be investigated. Since our interest is in the development of mass, rather than concentration, based emission factors, we do not require information regarding the volume percentage of the exhaust gas components. 5.5.1
Simplified Combustion Model
In order to develop emission factors, a simplified combustion model is derived. The combustion model is intended to represent the conversion of fuel and air to the main products of combustion. For complete combustion, the main products of combustion are carbon dioxide and water vapor. In addition, since air contains primarily nitrogen, and since most of the nitrogen is unreacted during combustion, the exhaust from a combustor typically contains large amounts of nitrogen. If a combustor is operated with excess amounts of oxygen, then there would also be oxygen in the exhaust products. Diesel
60
engines tend to operate with excess air (fuel lean), while gasoline engines typically operate with a slight excess amount of fuel (fuel rich). In an actual vehicle engine, processes occur which lead to products of incomplete combustion. These processes are due to heterogeneity in the fuel, air, and exhaust product mixture in the cylinders, and differences in temperature and pressure throughout the power stroke of the engine. These differences lead to variation in mixing and reaction rates throughout the combustion process (Flagan and Seinfeld, 1988). The typical products of incomplete combustion are carbon monoxide, unburned fuel, and intermediate hydrocarbon species associated with the oxidation of the fuel. In addition, nitrogen oxides are formed from the nitrogen and oxygen in the inlet air. Remote sensing data provides information regarding the ratios of CO to CO2 and HC to CO2. Since CO2, CO, and HCs are the products of combustion which contain carbon, it is possible to develop a mass balance regarding the amount of carbon contained in the fuel and the amount of carbon contained in these three products of combustion. The total number of carbon atoms in each case should be equal. Because carbon atoms must be conserved, the observed CO to CO2 and HC to CO2 from the remote sensing data enable the calculation of stoichiometric coefficients for the amount of CO2, CO, and HCs formed per mole of fuel combusted. Knowledge regarding the molecular weight and density of the fuel then enables calculation of the amount of CO and HCs emitted per gallon of fuel combusted. Finally, information regarding individual vehicle fuel economy enables calculation of the amount of CO and HCs emitted per mile of vehicle travel. Several assumptions are made to arrive at the composition of the exhaust products. These assumptions are: (1) complete consumption of the fuel is assumed; (2) the oxidation of nitrogen from the combustion inlet air is ignored; (3) the carbon released from the fuel is emitted as either CO2, CO, or as an unburned hydrocarbon equivalent to propene (C3H6); (4) the ratio of hydrogen to carbon in the fuel is specified based upon typical fuel compositions as described in the previous sections; and (5) only enough oxygen is consumed to covert the fuel to CO2, CO, HC, and water vapor. For simplicity, our combustion calculations are based upon stoichiometric amounts of air. The assumptions regarding the air-to-fuel ratio do not affect the ratios of CO to CO2 and HC to CO2. If excess oxygen were combusted, the only effect on the mass balance would be that there would also be oxygen in the exhaust products. The starting point for the combustion calculation is to develop an equivalent molecular formula for the fuel. For example, we assume that the fuel consists primarily
61
of carbon and hydrogen, with negligible amounts of other species for the purposes of the mass balance. In Table 5-3, the development of an equivalent molecular formula is illustrated for a fuel containing 86 weight percent of carbon and 14 weight percent of hydrogen. For the basis of the calculation, 100 pounds of fuel are assumed. The pounds of each component per 100 pounds of fuel are divided by their respective molecular weights, to obtain the lbmoles of each component per 100 pounds of fuel. Then, the lbmoles of each component per 100 pounds of fuel are divided by the lbmoles of carbon per 100 pounds of fuel. The result is the lbmoles of each component per lbmole of carbon in the fuel. In this example, the equivalent molecular formula is CH1.95. Table 5-3. Equivalent Molecular Formula lb per 100 lb Molecular lbmole per Component fuel Weight 100 lb of fuel Carbon 86 12 7.17 Hydrogen 14 1 14.00
lbmole per lbmole C 1.00 1.95
In general, we assume that the molecular weight of the fuel is given by the equivalent molecular formula CHy, where: MWC y = wt–% H wt–% C MWH
(5-1)
where MWC is the molecular weight of carbon, and MWH is the molecular weight of hydrogen. The fuel is combusted in air, which contains a mixture of approximately 21 volume percent of oxygen and 79 volume percent of nitrogen. The products of combustion are assumed to be CO, H2O, C3H6, CO2, and N2. The mass balance for combustion, neglecting excess oxygen, is given by: CHy + m 0.21 O 2 + 0.79 N2 → a CO + b H2 O + c C3 H6 + d CO2 + e N2
(5-2)
where the variables a, b, c, d, e, and m are unknown stoichiometric coefficients defined as follows: m = moles of “air” (mixture of O2 and N2) consumed per mole of fuel consumed a
= moles of CO formed per mole of fuel consumed
b
= moles of H2O formed per mole of fuel consumed
c
= moles of C3H6 formed per mole of fuel consumed
62
d
= moles of CO2 formed per mole of fuel consumed
e
= moles of N2 in the product per mole of fuel consumed
Since the atoms which make up the compounds involved in the combustion equation are neither created or destroyed, mass balance equations based upon conservation of atoms can be written for each elemental species: Element
Reactants
=
Products
Carbon (C)
1
=
a + 3c + d
(5-3)
Hydrogen (H)
y
=
2b + 6c
(5-4)
Oxygen (O)
0.42 m
=
a + b + 2d
(5-5)
Nitrogen (N):
1.58 m
=
2e
(5-6)
For example, in Equation (5-3), there is one mole of carbon atoms per mole of fuel consumed, a moles of carbon in the CO produced per mole of fuel consumed, 3c moles of carbon atoms in the C3H6 produced per mole of fuel consumed, and d moles of carbon in the CO2 produced per mole of fuel consumed. Since the oxidation of nitrogen is neglected the problem is reduced to a system of three equations and five unknowns: a, b, c, d and m. However, from the RSD equipment additional information is known regarding the ratios of CO to CO2 and HC to CO2. These ratios are on a volume basis, which is also a molar basis for an ideal gas. Therefore, additional equations can be introduced: RCO = CO = a CO2 d
(5-7)
RHC = HC = c CO2 d
(5-8)
From these equations, the stoichiometric coefficients for CO and C3H6 can be expressed in terms of the stoichiometric coefficient for CO2 and the ratios of CO to CO2 and HC to CO2: a = RCO d
(5-9)
c = RHC d
(5-10)
Based upon Equations (5-9) and (5-10), Equation (5-3) can be rewritten as: 63
d=
1 RCO + 3 RHC + 1
(5-11)
Thus, the moles of CO2 produced per mole of fuel can be estimated based upon the emissions ratios from the remote sensing measurements, assuming that closure on the carbon mass balance can be obtained by considering only CO2, CO and C3H6. Once the value of d is known, Equations (5-9) and (5-10) can be used to solve for the moles of CO and C3H6, respectively, formed per mole of fuel combusted. As an example, assume that we obtained RSD measurements of RCO = 0.25 and RHC = 0.003. The estimated moles of CO2 produced per mole of fuel would be: d=
1 = 0.79 0.25 + 3 (0.003) + 1
lbmole CO 2 lbmole fuel
(5-12)
Based upon this result, the molar amounts of CO and C3H6 emitted per mole of fuel combusted can be estimated as: a = (0.25) (0.79) = 0.20
lbmole CO lbmole fuel
c = (0.003) (0.79) = 0.0024
5.5.2
lbmole C3H 6 lbmole fuel
(5-13) (5-14)
Emission Factors on a Grams Per Gallon Basis
The next step is to calculate an emission factor on a grams per gallon basis. For this purpose, it is necessary to estimate the density of the fuel. For gasoline, we assume a density of 0.742 g/cm3. For diesel fuel, we assume a density of 0.835 g/cm3. The equivalent molecular formula is used to estimate an equivalent molecular weight. The equivalent molecular weight is given by: MWfuel = 12.011
gC gmole C gH gmole H 1 + 1.0079 y gmole C gmole fuel gmole H gmole fuel
(5-15)
For y = 1.95, the equivalent molecular weight is 13.976 grams of fuel per gmole of fuel. Thus, the estimated molar amount of fuel per gallon of fuel for gasoline is given by: ρ' =
ρ 3785 cm3 MWfuel gallon
(5-16)
where ρ is the density of the fuel in g/cm3 and ρ’ is the density in gmol/gallon. For the example of gasoline, the estimated density is: ρ' =
0.742 g cm3
gmol 13.976 g
3785 cm3 = 201 gmol gallon gallon
64
(5-17)
The emission factor in grams of CO emitted per gallon of fuel consumed is given by: ' EFCO = ρ ' a MWCO
(5-18)
For the example, the emission factor for CO is: gmol CO g CO ' EFCO = 201 gallon 0.20 gmole 28 gmole gmole CO = 1,100
g CO gallon
(5-19)
Similarly, the emission factor for hydrocarbons is given by: ' EFHC = ρ ' c MWHC
(5-20)
For the example, the hydrocarbon emission factor is: gmol HC ' EFHC = 201 gallon 0.0024 gmole 42 gmole
5.5.3
g HC gmole HC
= 20
g CO gallon
(5-21)
Emission Factors on a Grams per Vehicle Mile Basis
Emission factors on a grams per vehicle-mile travel basis may be estimated based upon assumptions regarding the vehicle fuel economy: EFCO =
' EFCO MPG
(5-22)
EFHC =
' EFHC MPG
(5-23)
For example, if the fuel economy of a gasoline bus is 5.2 miles per gallon, then the CO and HC emission factors would be estimated as: g CO
EFCO =
1,100 gallon 5.2
miles gallon
= 211
g CO mile
(5-24)
Similarly, the emission factor for propane-equivalent hydrocarbons is estimated to be 3.8 grams per mile. 5.5.4
Bias in the Hydrocarbon Emission Factor
The RSD measures hydrocarbons that are similar to the straight-chain molecules used for calibration. However, it is less likely to detect other types of hydrocarbon species. In order to consider the potential effect of hydrocarbon species in the exhaust which are not detected by the RSD equipment the preceding combustion equation and emission factor derivation must be modified. To do this, the combustion equation (5-2) will be changed to include an undetected amount of unburned hydrocarbons. The undetected HC's will be assumed to be a volume percentage of the total unburned HC's in 65
the exhaust. The composition of the undetected hydrocarbons is treated as a variable in order to determine how this might effect the resulting measurements. The modified combustion equation is: CHy + m 0.21 O 2 + 0.79 N2 → a CO + b H2 O + c C3 H6 + d CO2 + e N2 + f Cu Hv (5-25) where the stoichiometric coefficient, f, is the molar amount of undetected hydrocarbons produced per mole of fuel consumed. The undetected hydrocarbons have an unknown equivalent molecular formula. The stoichiometric coefficient, f, is related to the variable z, which is the fraction of total hydrocarbons, on a volume basis, which are undetected. Therefore, the relationship between f and z is: f = 1 z– z c (5-26) For example, if the measured amount of hydrocarbons represents only 50 percent of the total volume of hydrocarbons, then z = 0.5, and f is equal to the stoichiometric coefficient, c, for the measured hydrocarbons. The equations for conservation of carbon and hydrogen atoms must be modified to account for the undetected hydrocarbons: Element
Reactants
=
Products
Carbon (C)
1
=
a + 3 c + d + 1 z– z c u
(5-27)
Hydrogen (H)
y
=
2 b + 6 c + 1 z– z c v
(5-28)
This modified combustion model illustrates conceptually how undetected hydrocarbons could be considered in developing adjusted hydrocarbon emission factors. However, little quantitative basis exists at this time for making the adjustment. More analysis is needed regarding the sensitivity of the RSD to different species in the hydrocarbon exhaust and regarding the variability in hydrocarbon speciation in vehicle exhaust emissions in order to develop reasonable estimates of the ratio of undetected to detected hydrocarbon emissions measurements.
66
67
6.0
REMOTE SENSING MEASUREMENTS AND ESTIMATED EMISSION FACTORS FOR SCHOOL BUSES
In this chapter, the on-road measurements of school bus emissions are summarized. First, we provide an overview of the logistics associated with data collection. The development of emissions measurements data bases and quality assurance for both data base development and emission factor calculations is described. The rest of the chapter is devoted to detailed presentations of the results of the measurements for school buses. 6.1
Data Collection and Database Development Activities Emissions measurements were obtained over 16 days of data collection for 265
diesel school buses and 36 gasoline school buses at five major locations. Table 6-1 summarizes the data collected, and Table 6-2 summarizes the dates and locations of school bus emissions data collection activities. At two of these locations, measurements were taken at more than one site. These locations include “Rock Quarry Road” and “Woodcroft.” A total of 1,457 measurements were obtained. Of these, 300 were considered to be invalid or sufficiently questionable that they were not included in the final data set used for emission factor development. Based upon consultation with Dr. Michael Jack of the Santa Barbara Research Center, it was decided to exclude from the data base any data for either the CO/CO2 or HC/CO2 ratios when one or both of the values were reported as invalid by the RSD on-board computer. Furthermore, it was also decided to exclude from the data base any data for which invalid estimates of pollutant volume percentages were estimated by the on-board computer. Although the volume percentage estimates would not be valid in any case for diesel vehicles, the error code reported by the on-board computer is based upon an array of statistical tests. The details of these statistical tests are not reported by the manufacturer. The use of the error codes to screen out potentially invalid data is believed to be conservative: it is possible that valid data may have screened out by this process, but it is unlikely that any truly invalid measurements were included in the emission factor data base. As indicated in Table 6-1, there were a total of 1,157 data points for 301 school buses that were used to develop the emission factors presented in this section. Of these measurements, there were 1,069 for diesel school buses, and 88 for gasoline school buses. Most of the data points were obtained at the Rock Quarry Road site. For diesel buses, 30 or more data points were obtained at the Rock Quarry Road, Laura Duncan Road, and Woodcroft locations. For gasoline buses, which comprise a relatively small portion of the 68
observed bus fleet, only the Rock Quarry Road site yielded more than 30 data points. The Garner and Wake Forest sites had relatively low volumes of school buses. Unfortunately, the Wake Forest and Woodcroft sites also had a relatively large proportion of data points which did not survive the data quality screening criteria. Table 6-1. Summary of School Bus Emissions Measurements Good Data Points Bad Data Points Percent Percent Location/Fuel Number of Total Number of Total Rock Quarry - gas 68 88.3 9 11.7 Rock Quarry - diesel 984 82.0 216 18.0 Laura Duncan - gas 16 76.2 5 23.8 Laura Duncan 33 71.7 13 28.3 diesel Garner - diesel 5 83.3 1 16.7 Wake Forest - gas 3 60.0 2 40.0 Wake Forest - diesel 9 32.1 19 67.9 Woodcroft - gas 1 33.3 2 66.7 Woodcroft - diesel 38 53.5 33 46.5 Total - diesel 1069 79.1 282 20.9 Total - gas 88 83.0 18 17.0 Total - all 1157 79.4 300 20.6
Total 77 1200 21 46 6 5 28 3 71 1351 106 1457
Table 6-2. Summary of School Bus RSD Data Collection Dates and Locations Date Site April 16, 1996 Rock Quarry Road May 3, 1996 Rock Quarry Road May 9, 1996 Rock Quarry Road May 15, 1996 Rock Quarry Road May 17, 1996 Rock Quarry Road May 21, 1996 Rock Quarry Road May 22, 1996 Garner May 29, 1996 Rock Quarry Road May 31, 1996 Rock Quarry Road June 4, 1996 Laura Duncan Road June 6, 1996 Laura Duncan Road June 10, 1996 Rock Quarry Road June 11, 1996 Rock Quarry Road June 11, 1996 Laura Duncan Road October 16, 1996 Woodcroft October 17, 1996 Wake Forest October 23, 1996 Woodcroft
69
A summary of the gasoline-fueled school bus data for all of the sites is given in Appendix B. This summary includes the school bus identification number, the site at which the measurements were taken, the observed CO/CO2 and HC/CO2 ratios, the average fuel economy for each bus, and the calculated emission factors. The fuel economy for each bus was estimated based upon the school bus maintenance record data summarized in Appendix A. For each bus that was observed, the bus identification number was manually recorded in the field and entered into the emissions data base. The bus identification numbers were also verified by reviewing the video tapes that were recorded in the field. The video tapes were also used to fill in entries that may have been missed in the field. Once the bus identification data were entered into a spreadsheetbased data base, lookup functions were used to automatically look up needed school bus characteristics from the maintenance record data, which also was converted to a spreadsheet. A summary of the diesel-fueled school bus data for the 984 valid observations at the Rock Quarry Road site is given in Appendix C. This summary includes the bus identification number, the date, the chassis year, the direction the bus was traveling (into or out of the staging area), the bus manufacturer, the most recently available odometer reading from the maintenance record data base, the passenger capacity of the bus, the average fuel economy, the observed CO/CO2 and HC/CO2 emission ratios, and the calculated emission factors on both a grams per gallon and grams per mile basis. A summary of the diesel-fueled school bus data for the Laura Duncan Road, Garner, Wake Forest, and Woodcroft locations is given in Appendix D. The summary includes the school bus identification number, the site at which the measurements were taken, the observed CO/CO2 and HC/CO2 ratios, the average fuel economy for each bus, and the calculated emission factors. 6.2
Summary of School Bus Emissions Estimates
The emissions data were analyzed several different ways to provide insights regarding variability from one measurement to another and regarding uncertainty in fleet average emissions. The variability between individual measurements typically spanned several orders-of-magnitude. However, we are mainly interested in fleet average emission factors that can be used as a basis for developing emissions inventories. Because of the large variation in individual measurements and, in some cases, small data
70
set sizes, there is uncertainty regarding the mean. confidence intervals for the mean.
Therefore, we have calculated
Table 6-3 is a summary of the estimated average fuel economy and emission factors for diesel-fueled school buses. The first set of data in the table are based upon all 1,069 observations of diesel buses. Fuel economy and emissions estimates for each of the remote sensing locations are also summarized. The CO and HC emissions estimates for the diesel school buses observed at Rock Quarry Road and Laura Duncan Road sites are similar, even though the operating conditions and the bus fleets observed at these sites were different. At the Rock Quarry Road site, the buses have relatively low speeds of approximately 15 mph compared to the Laura Duncan Road, which has a posted speed limit of 45 mph. The buses which operate at the Laura Duncan Road site use Apex High School as a staging area. Therefore, this bus fleet does not include any buses in common with that observed at Rock Quarry Road. Thus, the similarities in average emissions imply that on-road diesel bus emissions may not be particularly sensitive to inter-vehicle variability for similar fleets or to differences in speed. Both site have relatively flat road grades. It is likely that the buses at the Rock Quarry Road site had higher accelerations than at the Laura Duncan Road site. The average emissions at the Garner site appear to be higher than for the Rock Quarry Road and Laura Duncan Road sites. However, because there are only five data points for the Garner site, the confidence intervals for the mean emissions are wide and enclose the confidence intervals for the Rock Quarry Road and Laura Duncan sites. Thus, it is not possible to conclude that the observations at the Garner site are due to emissions actually being higher than for the other two sites. The Wake Forest site, for which there are only nine observations, has higher average emissions than for the other three sites. However, because of the small sample size in this case, the confidence intervals for these averages are wide enough to enclose the confidence intervals for the means of the previously described three sites. The high averages and wide confidence intervals for the Wake Forest site are due in part to two observations that both generated large emission factor estimates.
71
Table 6-3.
Summary of Estimated Average Emission Factors for Diesel-Fueled School Buses 95 Percent Confidence Standard Interval on the Mean Description Mean Deviation Lower Upper Count All School Buses MPG 7.6 1.0 7.5 7.7 1069 CO g/gal 97.1 139.5 88.7 105.5 1069 HC g/gal 25.2 58.8 21.7 28.7 1069 CO g/mi 13.0 18.6 11.9 14.1 1069 HC g/mi 3.4 7.8 2.9 3.9 1069 Rock Quarry Rd. MPG 7.6 1.0 7.5 7.7 984 CO g/gal 92.1 116.8 84.8 99.4 984 HC g/gal 22.5 42.5 19.9 25.2 984 CO g/mi 12.4 16.1 11.4 13.4 984 HC g/mi 3.0 6.1 2.7 3.4 984 Laura Duncan Road MPG 8.0 1.0 7.6 8.3 33 CO g/gal 83.3 81.2 55.6 111.0 33 HC g/gal 20.2 13.1 15.8 24.7 33 CO g/mi 10.4 9.9 7.0 13.8 33 HC g/mi 2.5 1.5 2.0 3.0 33 Garner MPG 8.3 0.4 8.0 8.7 5 CO g/gal 157.1 88.5 79.5 234.8 5 HC g/gal 45.8 37.6 12.9 78.8 5 CO g/mi 19.0 10.8 9.5 28.4 5 HC g/mi 5.6 4.8 1.4 9.8 5 Wake Forest MPG 8.2 0.5 7.8 8.5 9 CO g/gal 244.9 277.8 63.4 426.4 9 HC g/gal 190.0 344.3 0.0 415.0 9 CO g/mi 29.1 32.3 8.1 50.2 9 HC g/mi 22.8 41.3 0.0 49.8 9 Woodcroft MPG 6.8 1.4 6.3 7.2 38 CO g/gal 194.5 399.9 67.4 321.7 38 HC g/gal 57.0 133.6 14.5 99.5 38 CO g/mi 28.2 48.9 12.6 43.7 38 HC g/mi 8.2 16.5 2.9 13.4 38
72
The average CO emission factor for the Woodcroft site is nearly the same as for the Wake Forest Site. Because of the much larger number of data points for Woodcroft (38) versus Wake Forest (9), the confidence intervals for the average emission factors based upon the Woodcroft site are narrower than those based upon the Wake Forest site data. However, the Woodcroft data set includes one observation in particular that leads to a high emission estimate of 300 g/mi for CO and 100 g/mi for HC. If this observation is removed, then the average CO emission factor decreases from 28 g/mi to 21 g/mi, and the 95 percent confidence interval for the mean becomes (15.2, 26.3) g/mi. However, there is no specific basis for excluding this data point from the analysis. Of the five sites tested, the Woodcroft site appears to be somewhat different from the other four in terms of the average emission estimates. The data in Tables 6-3 and 6-4 indicate that the emissions for gasoline and diesel buses are significantly different, with gasoline buses typically having higher emissions of CO and HC than diesel buses. This result is expected. In contrast to the comparison of emissions for diesel buses at the Rock Quarry Road and Laura Duncan Road sites, the emissions for the gasoline buses appear to be more sensitive to the site conditions. For example, the CO emissions at the Rock Quarry Road site, which has relatively low bus speeds and in some cases significant accelerations, tend to be substantially higher than for the Laura Duncan Road site, where the buses are operating in a cruise mode at a much higher speed. Similarly, the HC emissions appear to be significantly higher at the Rock Quarry Road site than for the Laura Duncan Road site. Because only three gasoline buses were observed at the Wake Forest Site and only one gasoline bus was observed at the Woodcroft site, these data are not reported separately in Table 6-4. However, these four observations are included the combined data set for “All School Buses.” The school bus emissions estimates are described in more detail in the following sections.
73
Table 6-4. Summary of Estimated Average Emission Factors for Gasoline-Fueled School Buses 95 Percent Confidence Standard Interval on the Mean Description Mean Deviation Lower Upper Count All School Buses MPG 4.7 0.6 4.6 4.9 88 CO g/gal 970.3 808.4 801.4 1139.2 88 HC g/gal 42.7 165.9 8.0 77.3 88 CO g/mi 206.7 172.2 170.7 242.7 88 HC g/mi 9.2 35.4 1.8 16.6 88 Rock Quarry Rd. MPG 4.8 0.6 4.7 4.9 68 CO g/gal 1142.9 810.6 950.2 1335.5 68 HC g/gal 51.3 188.0 6.6 96.0 68 CO g/mi 240.9 170.7 200.3 281.4 68 HC g/mi 11.0 40.1 1.5 20.5 68 Laura Duncan Road MPG 4.7 0.9 4.2 5.1 16 CO g/gal 310.6 287.6 169.7 451.5 16 HC g/gal 12.5 7.7 8.8 16.3 16 CO g/mi 67.9 60.8 38.1 97.7 16 HC g/mi 2.8 1.6 2.0 3.6 16
74
6.3
Fuel Economy Data for School Buses Observed at the Rock Quarry Road Site
The fuel economy data used to calculated the grams per mile emission factors is illustrated in Figure 6-1. These data clearly indicate that, on average, diesel buses have better fuel economy than gasoline buses. Here, we focus on the fuel economy data for the Rock Quarry Road site, since this is the most abundant data set for the emission factor analysis. For the 209 diesel buses that were observed at the Rock Quarry Road site, the average estimated fuel economy is 7.6 miles per gallon. For the 21 gasoline buses that were observed, the average fuel economy was 4.5 miles per gallon. One of the gasoline bus fuel economy values was adjusted from 43 miles per gallon to the average of the remaining values. This adjustment was made because it was deemed improbable that even a well-maintained school bus would have such a high fuel economy. This fuel economy estimated was associated with Bus No. 58, which had consumed only 31 gallons during the time period for which the fuel economy estimate was made. Thus, this high value fuel economy was judged to be due to either a data entry error in the school bus maintenance record data base or to inconsistency in how the fuel tank was topped off between odometer readings. The cumulative distributions for fuel economy indicate that for both diesel and gasoline buses there is a small percentage of the fleet with very low fuel economy. A review of the data specific to these buses did not reveal any obvious errors in the data base. For example, the buses with the lowest fuel economies had all consumed more than 1,500 gallons of fuel during the time period over which the fuel economy estimate was based.
Cumulative Frequency
1.0 0.8 0.6 0.4 Diesel Buses
0.2
Gasoline Buses
0.0 1
2
3
4
5 6 7 8 Fuel Economy (Miles per Gallon)
9
10
11
Figure 6-1. Variability in Fuel Economy for 209 Diesel and 21 Gasoline School Buses Observed at Rock Quarry Road.
75
6.4
Diesel School Buses at the Rock Quarry Road Site
In this section, we analyze in detail the diesel school bus emissions estimates for the Rock Quarry Road site. We focus on this site because of the large number (n=984) data points. 6.4.1
CO and HC Emissions for All Observed Buses
The Rock Quarry Road site diesel bus emissions were analyzed on both a grams per gallon and grams per mile basis. A summary of the mean, standard error, and 95 percent confidence bounds for the mean of both CO and propane-equivalent hydrocarbon emissions is given in Table 6-3. Figure 6-2 displays the calculated CO emission factors, in grams per gallon, for all 984 valid measurements for diesel school buses at the Rock Quarry Road site. Figure 6-3 displays similar result in units of grams per mile. The emissions varied over nearly four orders-of-magnitude from a minimum value of 0.066 grams/mile to a maximum of 219 grams/mile. The average emissions are 12.36 grams/mile, and the standard deviation for variability in emissions is 16.13 grams/mile. As noted in Table 5-1, the 95 percent confidence interval for the mean ranges from 11.35 g/mi to 13.37 g/mi (plus or minus 9 percent). Figures 6-4 and 6-5 display the calculated propane-equivalent hydrocarbon emission factors in units of grams per gallon and grams per mile, respectively. The estimated emissions varied from a minimum of 0.0094 grams/ mile to a maximum of 116.1 grams/mile, which is a range of over four orders-of-magnitude. The average emissions estimate is 3.05 grams/mile, and the standard deviation for the individual estimates is 6.13 grams/mile. The 95 percent confidence interval for the uncertainty in the mean is plus or minus 12 percent of the mean value.
76
Cumulative Frequency
1.0 Diesel School Buses Rock Quarry Road n = 984
0.8
Calculated from Data Set
0.6 0.4
Mean
0.2
95 % Confidence Interval for the Mean
0.0 0
50
100 150 CO Emission Factor (g/gal)
200
250
Figure 6-2. Variability and Uncertainty in 984 Estimates of Diesel School Bus CO Emissions (grams/gallon) based Upon Remote Sensing Measurements at the Rock Quarry Road Site.
Cumulative Frequency
1.0 Diesel School Buses Rock Quarry Road n = 984
0.8 0.6
Calculated from Data Set
0.4
Mean
0.2
95 % Confidence Interval for the Mean
0.0 0
5
10
15 20 25 CO Emission Factor (g/mi)
30
35
40
Figure 6-3. Variability and Uncertainty in 984 Estimates of Diesel School Bus CO Emissions (grams/mile) based Upon Remote Sensing Measurements at the Rock Quarry Road Site.
77
Cumulative Frequency
1.0 Diesel School Buses Rock Quarry Road n = 984
0.8 0.6
Calculated from Dataset
0.4
Mean
0.2
95 % Confidence Interval for the Mean
0.0 0
10 20 30 40 50 Propane-Equivalent Hydrocarbon Emission Factor(g/gal)
60
Figure 6-4. Variability and Uncertainty in 984 Estimates of Diesel School Bus PropaneEquivalent Hydrocarbon Emissions (grams/gallon) based Upon Remote Sensing Measurements at the Rock Quarry Road Site.
Cumulative Frequency
1.0 Diesel School Buses Rock Quarry Road n = 984
0.8
Calculated from Dataset
0.5 Mean
0.2
95 % Confidence Interval for the Mean
0.0 0
3 6 9 12 Propane-Equivalent Hydrocarbon Emission Factor (g/mile)
15
Figure 6-5. Variability and Uncertainty in 984 Estimates of Diesel School Bus PropaneEquivalent Hydrocarbon Emissions (grams/mile) based Upon Remote Sensing Measurements at the Rock Quarry Road Site.
78
6.4.2
Individual Buses
At the Rock Quarry Road site, the 984 observations of diesel bus emissions represent 209 different buses. To gain insight into whether the observed variability for the 984 data points can be attributed to discernible differences in emissions from one bus to another, we analyzed data for individual buses. There are 14 buses in the data set for which 10 or more measurements were obtained, and these buses were the focus of our analysis. 6.4.2.1 CO Emissions Table 6-5 summarizes the CO emissions data for these 14 buses. Figure 6-6 shows a comparison of the cumulative distribution functions of the CO emissions estimates for each of the 14 buses. Figure 6-7 displays the 95 percent confidence intervals for the average CO emissions for each of these buses. In Table 6-6, the results of t-tests for the means of each combination of pairs of buses are reported. These data indicate that the average emissions for one bus versus another are in most cases not significant statistically. In all cases for CO, the range of variation in emissions for each bus and the 95 percent confidence intervals for the means both overlap, as illustrated in Figures 6-6 and 6-7. The results of the t-tests for the means indicate that only one bus, No. 2802, has emissions that are significantly different from some of the other buses, which include Buses No. 98, 594, and 774. On average, the emissions of these three buses are higher than for Bus No. 2802. In Table 6-5, the upper confidence bound for the mean of CO emissions for Bus 2802 is 9.2 g/mi, while the lower confidence bounds for Buses No. 98, 594, and 774 are 7.4 g/mi, 8.1 g/mi, and 7.6 g/mi, respectively. Thus, in all three cases, there is some overlap in the confidence intervals with that of Bus No. 2802. As the overlap becomes greater, the significance level of the t-test increases. For example, the ttest between the means of CO emissions for Buses 2802 and 266 has a significance level of 0.36, and the confidence interval for Bus No. 266, which ranges from 5.5 g/mi to 10.0 g/mi, has a large amount of overlap with the 95 percent confidence interval for the mean of the CO emissions from Bus No. 2802. As another example, although the average emissions for Bus No. 915 are nearly twice that for Bus No. 2802, these two averages are not significantly different at the 0.05 significance level.
79
Table 6-5.
Bus No. 83 86 89 98 119 235 266 381 523 594 716 774 915 2802
Summary of CO Emissions for 14 Individual Diesel Buses at Rock Quarry Road Confidence Bounds for Number of Average Standard Standard the Mean Data Emissions Deviation Error Lower Upper Points, N (g/mi) (g/mi) (g/mi) (2.5%) (97.5%) 13 11 11 10 12 10 12 14 10 10 10 10 10 10
8.0 8.7 9.8 14.1 8.1 10.2 7.8 11.0 11.2 12.7 13.8 15.4 12.0 6.2
2.5 4.6 4.3 9.5 4.8 4.9 3.5 8.2 7.0 6.4 10.0 10.9 9.0 4.1
0.7 1.4 1.3 3.0 1.4 1.6 1.0 2.2 2.2 2.0 3.2 3.4 2.9 1.3
6.4 5.6 6.9 7.4 5.1 6.6 5.5 6.2 6.1 8.1 6.6 7.6 5.5 3.3
9.5 11.8 12.7 20.9 11.1 13.7 10.0 15.7 16.2 17.3 21.0 23.2 18.4 9.2
Table 6-6. Summary of T-Test Results for CO Emissions of 14 Individual Diesel Buses (Identified by Bus Number) at the Rock Quarry Road Site a 83 86 89 98 119 235 266 381 523 594 716 774 915 2802 Significance Level for t-test of the Null Hypothesis that Means are the Same 83 1.00 86 .63 1.00 89 .22 .56 1.00 98 .07 .12 .21 1.00 119 .33 .36 .40 .56 1.00 235 .22 .49 .87 .26 .41 1.00 266 .87 .58 .22 .07 .33 .21 1.00 381 .21 .40 .66 .40 .44 .78 .20 1.00 523 .19 .36 .61 .44 .44 .72 .18 .95 1.00 594 .05 .12 .25 .70 .50 .33 .05 .56 .61 1.00 716 .10 .17 .27 .94 .55 .33 .10 .47 .51 .78 1.00 774 .06 .10 .16 .78 .62 .19 .06 .29 .32 .51 .73 1.00 915 .20 .33 .51 .60 .48 .59 .19 .79 .83 .83 .67 .45 1.00 2802 .26 .21 .06 .03 .29 .07 .36 .08 .07 .02 .05 .03 .09 1.00 a A significance level of less than 0.05 indicates that the null hypothesis that the mean emissions of the two buses are the same should be rejected at the 0.05 significance level.
80
Cumulative Frequency
1 0.8 0.6 0.4 0.2 0 0
5
10
15
20
25
30
35
CO Emission Factor (g/mile) Bus No. 83
Bus No. 119
Bus No. 523
Bus No. 774
Bus No. 86
Bus No. 235
Bus No. 594
Bus No. 915
Bus No. 89
Bus No. 266
Bus No. 716
Bus No. 2802
Bus No. 98
Bus No. 381
Figure 6-6. Comparison of Empirical Cumulative Distribution Functions for CO Emissions of 14 Diesel-Fueled School Buses with 10 or More Observations at the Rock Quarry Road Site. Mean CO Emission Factor (g/mile)
25 Lower Confidence Range
20
Upper Confidence Range
15 10 5 0 83
86
89
98 119 235 266 381 523 594 716 774 915 2802 Bus Number
Figure 6-7. Comparison of 95 Percent Confidence Intervals for Mean CO Emissions of 14 Diesel-Fueled School Buses with 10 or More Observations at the Rock Quarry Road Site.
81
40
In Table 6-6 there are a total of 91 comparisons of the means of CO emissions of different pairs of buses. At a 0.05 significance level, we expect by random statistical fluctuation that five percent of the pairs (approximately 5 out of 91) would yield a t-test at less than 0.05 significance level even if all of the true means were identical. In this case, we have three results that are outside of the specified significance level. Thus, it is possible that these results are an artifact of random statistical fluctuation due to small sample size, and not due to real differences in the mean emissions from one bus to another. Based on a review of the information contained in Figures 6-6 and 6-7 and Tables 6-5 and 6-6, we conclude that there are no significant differences in observed CO emissions from one bus to another within the sample of 14 buses for which we had 10 or more valid measurements. Furthermore, the range of variability and uncertainty in the individual bus data is consistent with that observed for all 984 measurements. These results indicate that the variability in bus emissions has less to do with bus design, age, maintenance, and so on and perhaps is more dependent upon site conditions at Rock Quarry Road and driver behavior. 6.4.2.2 HC Emissions The mean propane-equivalent hydrocarbon emissions for the 14 diesel buses at the Rock Quarry Road site for which there were 14 or more data points are given in Table 67, along with estimates of the 95 percent confidence intervals for the mean. The variation in emissions for each individual bus is shown in Figure 6-8. The confidence intervals for the means are displayed in Figure 6-9. A summary of t-tests of the means for each possible pairwise combination of the 14 buses is given in Table 6-8. For Bus No. 119, one extreme outlier was discarded. The results indicate that, in most cases, there is no apparent difference in the mean emissions between buses. Of the 14 buses, only Bus No. 235 and Bus No. 266 have significant differences from one or more of the other 12 buses. Bus No. 235 has significantly higher emissions than Bus No. 83, 266, and 716. Bus No. 266 has significantly higher emissions than five other buses, including Bus No. 235, 381, 774, 915, and 2802.
82
Table 6-7. Summary of Propane-Equivalent Hydrocarbon Emissions for 14 Individual Diesel Buses at Rock Quarry Road Confidence Bounds for Number of Average Standard Standard the Mean Data Emissions Deviation Error Lower Upper Bus No. Points, N (g/mi) (g/mi) (g/mi) (2.5%) (97.5%) 83 86 89 98 119 235 266 381 523 594 716 774 915 2802
13 11 11 10 12 10 12 14 10 10 10 10 10 10
1.8 4.0 2.5 2.6 2.1 4.1 1.5 2.7 2.8 6.2 2.0 3.3 3.9 2.6
1.0 5.8 1.5 1.3 1.2 2.1 0.7 1.2 2.1 6.5 1.2 2.1 2.7 1.2
0.3 1.8 0.4 0.4 0.3 0.7 0.2 0.3 0.7 2.1 0.4 0.7 0.8 0.4
1.2 0.1 1.5 1.6 1.3 2.6 1.0 2.0 1.3 1.5 1.2 1.8 2.0 1.7
2.4 7.9 3.5 3.5 2.8 5.6 1.9 3.4 4.3 10.9 2.9 4.8 5.8 3.5
Table 6-8. Summary of T-Test Results for Propane-Equivalent Hydrocarbon Emissions of 14 Individual Diesel Buses (Identified by Bus Number) at the Rock Quarry Road Site a 83 86 89 98 119 235 266 381 523 594 716 774 915 2802 Significance Level for t-test of the Null Hypothesis that Means are the Same 83 1.00 86 .25 1.00 89 .23 .42 1.00 98 .15 .44 .91 1.00 119 .32 .51 .37 .37 1.00 235 .01 .97 .06 .07 .50 1.00 266 .31 .18 .06 .03 .30 .00 1.00 381 .05 .49 .69 .77 .39 .08 .00 1.00 523 .18 .54 .68 .73 .40 .19 .07 .88 1.00 594 .06 .43 .11 .11 .73 .35 .05 .13 .15 1.00 716 .65 .30 .44 .35 .34 .02 .19 .18 .31 .08 1.00 774 .07 .71 .34 .36 .43 .42 .03 .45 .62 .21 .12 1.00 915 .04 .95 .17 .18 .49 .85 .02 .22 .34 .32 .07 .60 1.00 2802 .12 .45 .85 .94 .38 .07 .02 .83 .77 .12 .31 .39 .19 1.00 a A significance level of less than 0.05 indicates that the null hypothesis that the mean emissions of the two buses are the same should be rejected at the 0.05 significance level.
83
Cumulative Frequency
1 0.8 0.6 0.4 0.2 0 0
3 6 9 12 Propane Equivalent Hydrocarbon Emission Factor (g/mile)
15
Bus No. 83
Bus No. 119
Bus No. 523
Bus No. 774
Bus No. 86
Bus No. 235
Bus No. 594
Bus No. 915
Bus No. 89
Bus No. 266
Bus No. 716
Bus No. 2802
Bus No. 98
Bus No. 381
Figure 6-8. Comparison of Empirical Cumulative Distribution Functions for PropaneEquivalent Hydrocarbon Emissions of 14 Diesel-Fueled School Buses with 10 or More Observations at the Rock Quarry Road Site. Mean Propane-Equivalent Hydrocarbon Emissions (g/mile)
12.5 Lower Confidence Range
10 Upper Confidence Range
7.5 5 2.5 0 83
86
89
98 119 235 266 381 523 594 716 774 915 2802 Bus Number
Figure 6-9. Comparison of 95 Percent Confidence Intervals for Mean PropaneEquivalent Hydrocarbon Emissions of 14 Diesel-Fueled School Buses with 10 or More Observations at the Rock Quarry Road Site.
84
Bus No. 266 has the lowest average emissions of any of the 14 buses. It also has the smallest standard error. Thus, the upper confidence bound for this bus is the lowest of all of the 14 buses. This illustrates why it appears to be statistically significant from a large portion of the other 13 buses. Bus No. 266 is a Ford with a 1986 chassis and a fuel economy of 9.4 miles per gallon. Other than its relatively high fuel economy, which is at approximately the 90th percentile of the fleet of 209 diesel buses observed at the Rock Quarry Road site, there are no other apparent reasons why this particular bus should be a relatively low emitter. Bus No. 235 is an International make with a 1984 chassis model year and a belowaverage fuel economy of 5.8 miles per gallon. The average emissions of 4.1 g/mi for this bus are similar to a few others, including Bus Nos. 86 (4.1 g/mi) and 915 (3.9 g/mi). However, Bus No. 235 has a lower standard error than these other two buses. Thus, it is more likely to be significantly different from the lower emitting buses than Bus Nos. 86 and 915. Although there are a total of 12 buses that have lower average emissions than Bus No. 235, only three of these appear to be statistically significant differences. Similar to the analysis of CO emissions from individual buses, there are a total of 91 comparisons of the means of HC emissions of different pairs of buses. At a 0.05 significance level, we expect by random statistical fluctuation that five percent of the pairs (approximately 5 out of 91) would yield a t-test at less than 0.05 significance level even if all of the true means were identical. In this case, we have seven results that are outside of the specified significance level. Thus, it is possible that these results are an artifact of random statistical fluctuation due to small sample size, and not due to real differences in the mean emissions from one bus to another. Based upon review of the results for the HC emissions of individual buses we conclude that, although there may be some significant differences among a relatively small number of buses, the observed variation in emissions for individual buses is comparable to that for the observed fleet as a whole. It appears that, based upon the limited amount of data currently available, it is difficult to attribute variability in all of the 984 measurements to factors specific to individual buses. We consider this issue from a different perspective in the next section. 6.4.3
Effect of Bus Characteristics on Emissions
In this section, we explore whether characteristics of each bus can be used to explain at least a portion of the observed variability in emissions. The bus characteristics that we have considered include odometer reading, chassis year, capacity, and fuel 85
economy. In a separate analysis presented later, we also evaluate whether emissions differ significantly depending upon the manufacturer. We have used regression analysis and statistical hypothesis tests to try to identify statistically significant explanatory variables for the emissions. However, in this section, we will focus on presenting scatter plots, which provide a visual indication of trends in the data set. The scatter plots, regression analyses, and statistical hypothesis tests all lead to the same conclusions. The large variation within the data set cannot be explained by any of the variables that we have considered here. The odometer readings provide an indication of both the age of the bus and possible wear associated with the amount of actual service the bus has had. It was hypothesized that emissions would increase with odometer reading. However, no significant trend was observed for either CO or HC emissions, as shown in Figures 6-10 and 6-11, respectively. These figures clearly illustrate that the 209 observed buses had a wide range of odometer readings, from close to zero for some new buses to nearly 250,000 miles for one of the older buses. In spite of the variation in odometer readings, it was not possible to obtain a meaningful regression model for emissions. For example, a log-linear model for CO emissions versus odometer reading had a coefficient of determination (R2) of only 0.00014. Similar results were obtained for the other hypothesized explanatory variables. Figures 6-12 and 6-13 are dot plots of CO and HC emissions, respectively, versus chassis year. The chassis year, like the odometer reading, provides an indication of the age of the bus. It was hypothesized that emissions would be larger for older buses. However, there is no significant trend of emissions with respect to chassis year. The size of the bus was hypothesized to affect emissions. Typically, similar size engines are used in both the “short” buses which seat approximately 35-40 students and in the “long” buses which seat approximately 60 students. Thus, the engine load on the larger buses would be expected to be proportionally larger, leading to potentially higher emissions. However, it was not possible to identify a statistically significant trend of emissions with respect to bus size, as illustrated in Figures 6-14 and 6-15 for CO and HC emissions, respectively.
86
Calculated CO Emissions (g/mile)
1000 100 10 1 0.1 0.01 0
50000
100000 150000 Odometer Reading (miles)
200000
250000
Figure 6-10. Scatter Plot of Estimated CO Emissions Versus Odometer Readings for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site. Calculated Propane-Equivalent Hydrocarbon Emissions (g/mile)
1000 100 10 1 0.1 0.01 0.001 0
50000
100000 150000 Odometer Reading (miles)
200000
250000
Figure 6-11. Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Odometer Readings for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site.
87
Calculated CO Emissions (g/mile)
1000 100 10 1 0.1 0.01 83
84
85
86
87
88
89 90 91 Chasis Year
92
93
94
95
96
97
Figure 6-12. Scatter Plot of Estimated CO Emissions Versus Chassis Year for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site.
Calculated Propane-Equivalent Hydrocarbon Emissions (g/mile)
1000 100 10 1 0.1 0.01 0.001 83
84
85
86
87
88
89 90 91 Chasis Year
92
93
94
95
96
97
Figure 6-13. Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Chassis Year for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site.
88
Calculated CO Emissions (g/mile)
1000 100 10 1 0.1 0.01 30
40
50 60 Capacity (Number of Passengers)
70
Figure 6-14. Scatter Plot of Estimated CO Emissions Versus Capacity for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site.
Calculated Propane-Equivalent Hydrocarbon Emissions (g/mile)
1000 100 10 1 0.1 0.01 0.001 30
40
50 60 Capacity (Number of Passengers)
Figure 6-15. Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Bus Capacity for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site.
89
70
Fuel economy was also considered as a possible explanatory variable for differences in emissions. Fuel economy may be an indicator of both the efficiency of the engine and of vehicle maintenance. However, as illustrated in Figure 6-16 and 6-17 for CO and HC emissions, respectively, most of the variability in emissions cannot be explained by this variable. Note that in this case emission factors on a grams per gallon basis are shown in the figures. The grams per mile emission factors were not used because one of the inputs to calculate these is fuel economy. Thus, the grams per mile emission factors are not independent of the fuel economy data. These findings indicate that it is not worthwhile to pursue further development of a regression model for emissions versus these hypothesized explanatory variables. For example, the following regression model was developed: log(CO, g/mi) = 1.84 - 0.00895 (Chassis Year) - 0.0674 (Odometer Reading) - 0.00017 (Passenger Capacity) The adjusted coefficient of determination (R2) for this model is 0.000. Thus, it cannot explain any of the variation in the emission factors. None of the coefficients are statistically significant except for the constant. Similar findings were obtained for HC emissions. 6.4.4
Correlation Between CO and Hydrocarbon Emissions
The emissions of CO and hydrocarbons appear to be moderately correlated. For example, as shown in Figure 6-18, there is some dependence between HC emissions versus CO emissions. The linear sample correlation coefficient in this case is 0.77. These data suggest that a vehicle which is a high emitter of one of the two pollutants may also tend to be a high emitter of the other pollutant. 6.4.5
Effect of Vehicle Direction on Emissions
At the outset of this study, it was hypothesized that the emissions of school buses may depend upon the engine temperature, which in turn is a function of how long the engine has been operating during a given trip since startup. At the Rock Quarry Road site, outbound buses typically have been sitting in the staging area for several hours. Thus, even in the afternoon, the buses leaving the site typically were started within a few minutes prior to departure. The incoming buses, in contrast, were typically returning from a route and may typically have been operating continuously for at least a half hour before returning to the staging area. Thus, to the extent that engine temperature may
90
affect emissions, it was hypothesize that such differences would be apparent in comparing emissions estimates for inbound and outbound buses.
Calculated CO Emissions (g/gal)
10000 1000 100 10 1 0.1 0
2.5
5 7.5 Fuel Economy (Miles per Gallon)
10
12.5
Figure 6-16. Scatter Plot of Estimated CO Emissions Versus Fuel Economy for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site.
Calculated Propane-Equivalent Hydrocarbon Emissions (g/gal)
1000 100 10 1 0.1 0.01 0
2.5
5 7.5 Fuel Economy (Miles per Gallon)
10
12.5
Figure 6-17. Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Fuel Economy for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site.
91
Calculated Propane-Equivalent Hydrocarbon Emissions (g/mile)
1000 100 10 1 0.1 0.01
0.001 0.01
0.1
1 10 Calculated CO Emissions (g/mile)
100
1000
Figure 6-18. Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Estimated CO Emissions for 984 Observations of 209 Diesel-Fueled School Buses at the Rock Quarry Road Site. In Figures 6-19 and 6-20, the cumulative distribution functions, means, and 95 percent confidence intervals for the means of CO and HC emissions, respectively, are compared for inbound and outbound diesel school buses at the Rock Quarry Road site. The results indicate that the inter-individual variability in emissions was similar for both the inbound and outbound buses. Although the means appear to be slightly different, there is a substantial amount of overlap in their confidence intervals. Thus, given uncertainty in the fleet average emissions, it is not possible to conclude that the inbound emissions for the entire observed fleet are significant different than the outbound emissions for the fleet. The differences in inbound versus outbound emissions were investigated in more detail by disaggregating the bus fleet with respect to bus manufacturer. The purpose of this disaggregation was to reduce inter-vehicle variability in the design of the bus to identify whether differences in emissions due to operating conditions would become more apparent. Figure 6-21 shows the cumulative distribution functions, means, and 95 percent confidence intervals for the means of CO emissions for inbound and outbound buses assembled by Chevrolet. While there is relatively little overlap in the confidence intervals for the means of the inbound and outbound emissions estimates, the difference in the means is not significant at the 5 percent significance level. Furthermore, there is similarity in the range and shapes of the cumulative distribution functions, suggesting
92
Cumulative Frequency
1.0 0.8
Inbound Inbound Mean 95 % Confidence Interval on Inbound Mean Outbound Outbound Mean
0.5
0.2
95 % Confidence Interval on Outbound Mean
0.0 0
5
10 15 20 25 Calculated CO Emission Factor (g/mile)
30
35
Figure 6-19. Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Inbound Versus Outbound Diesel Buses at Rock Quarry Road (n = 984).
Cumulative Frequency
1.0
0.8 Inbound Inbound Mean 95 % Confidence Interval on Inbound Mean Outbound Outbound Mean
0.5
0.2
95% Confidence Interval on Outbound Mean
0.0 0
2 4 6 8 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mile)
10
Figure 6-20. Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Inbound Versus Outbound Diesel Buses at Rock Quarry Road (n = 984).
93
little qualitative differences in the emissions as a function of direction. Similar results were obtained in Figure 6-22 when the HC emissions were compared for inbound versus outbound buses. The emissions for inbound and outbound diesel school buses assembled by Ford are shown in Figure 6-23 for CO and Figure 6-24 for propane-equivalent hydrocarbons. These results do not indicate any significant differences in emissions due to the direction of the buses. Similarly, Figures 6-25 and 6-26 do not indicate any significant differences in inbound versus outbound emissions of CO or HC, respectively, for diesel school buses assembled by International. In fact, for the International bus fleet, there appear to be greater similarities between the inbound and outbound observations and estimates than for the other two manufacturer’s fleets. The average emissions do not differ significantly among the three manufacturers. 6.4.6
Summary
The analyses presented here suggest that, although there is a substantial amount of variability in emissions from one observation to another, there are no observed explanatory factors which are significant in explaining the differences in emissions from one vehicle to another or for one observation to another. These findings suggest that unobserved factors are responsible for the differences in emissions. Unobserved factors that are likely to be important include differences in vehicle operating conditions due to differences in driver behavior. However, such factors are difficult to measure other than through the use of on-board instrumentation.
94
Cumulative Frequency
1.0 Rock Quarry Road Diesel Fuel Chevrolet Buses Inbound n = 140 Outbound n = 235
0.8 0.6
Inbound Inbound Mean 95 % Confidence Interval on Inbound Mean
0.4
Outbound Outbound Mean
0.2
95 % Confidence Interval on Outbound Mean
0.0 0
25
50
75 100 125 Calculated CO Emissions (g/gal)
150
175
200
Figure 6-21. Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Inbound Versus Outbound Chevrolet Diesel Buses at Rock Quarry Road (n = 375).
Cumulative Frequency
1.0 Rock Quarry Road Diesel Fuel Chevrolet Buses Inbound n = 140 Outbound n = 235
0.8 0.6
Inbound Inbound Mean 95 % Confidence Interval on Inbound Mean Outbound Outbound Mean 95 % Confidence Interval on the Mean
0.4 0.2 0.0 0
10 20 30 40 Calculated Propane-Equivalent Hydrocarbon Emissions (g/gal)
50
Figure 6-22. Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Inbound Versus Outbound Chevrolet Diesel Buses at Rock Quarry Road (n = 375).
95
Cumulative Frequency
1.0 Rock Quarry Road Diesel Fuel Ford Buses Inbound n = 126 Outbound n = 203
0.8 0.6
Inbound Inbound Mean 95 % Confidence Interval on Inbound Mean
0.4
Outbound Outbound Mean
0.2
95 % Confidence Interval on Outbound Mean
0.0 0
25
50
75 100 125 150 Calculated CO Emissions (g/gal)
175
200
Figure 6-23. Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Inbound Versus Outbound Ford Diesel Buses at Rock Quarry Road (n = 329).
Cumulative Frequency
1.0 Rock Quarry Road Diesel Fuel Ford Buses Inbound n = 126 Outbound n = 203
0.8 0.6
Inbound Inbound Mean 95 % Confidence Interval on Inbound Mean Outbound Outbound Mean 95 % Confidence Interval on Outbound Mean
0.4 0.2 0.0 0
10 20 30 40 Calculated Propane-Equivalent Hydrocarbon Emissions (g/gal)
50
Figure 6-24. Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Inbound Versus Outbound Chevrolet Diesel Buses at Rock Quarry Road (n = 329).
96
Cumulative Frequency
1.0 Rock Quarry Road Diesel Fuel International Buses Inbound n =116 Outbound n = 164
0.8 0.6
Inbound Inbound Mean 95 % Confidence Interval on Inbound Mean Outbound
0.4
Outbound Mean
0.2
95 % Confidence Interval on the Mean
0.0 0
25
50
75 100 125 Calculated CO Emissions (g/gal)
150
175
200
Figure 6-25. Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Inbound Versus Outbound International Diesel Buses at Rock Quarry Road (n = 280).
Cumulative Frequency
1.0 0.8 Inbound
0.6
Inbound Mean
0.4
Rock Quarry Road Diesel Fuel International Buses Inbound n =116 Outbound n = 164
0.2
95 % Confidence Interval on Inbound Mean Outbound Outbound Mean 95 % Confidence Interval on Outbound Mean
0.0 0
10 20 30 40 Calculated Propane-Equivalent Hydrocarbon Emissions (g/gal)
50
Figure 6-26. Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Inbound Versus Outbound International Diesel Buses at Rock Quarry Road (n = 280).
97
6.5
Gasoline-Fueled Buses at the Rock Quarry Road Site
In this section, we discuss in more detail the estimated emission factors for gasoline-fueled school buses at the Rock Quarry Road site. 6.5.1
CO and HC Emissions for All Observed Gasoline-Fueled School Buses
The variability and uncertainty in estimated CO emissions for the observed gasoline buses at the Rock Quarry Road site is shown in Figure 6-27 on a grams per gallon basis, and in Figure 6-28 on a grams per mile basis. These graphs indicate that the emissions estimates for individual observations span nearly two orders-of-magnitude. The uncertainty in the mean is approximately 20 percent. Because the standard deviation of the hydrocarbon emission estimates for gasoline school buses observed at Rock Quarry Road is very large, a numerical simulation method for developing confidence intervals was employed. The method is bootstrap simulation (Efron and Tibshirani, 1993). Given a data set of sample size n, the general approach in bootstrap simulation is to assume a distribution which describes the quantity of interest, to perform r replications of the data set by randomly drawing, with replacement, n values, and then calculate r values of the statistic of interest. For the first step of assuming a distribution for the data set, there are many options. One approach is to use the actual data set itself, and to randomly select, with replacement, the actual values of the data set. This is sometimes referred to as resampling. Another approach is to assume a parametric distribution, such as normal or lognormal, to represent the data. Each approach will lead to a different estimate of the confidence interval. Both approaches are considered. The confidence interval for the Rock Quarry Road fleet average hydrocarbon emission factor depends upon the assumption made regarding the underlying population probability distribution for inter-vehicle variability in emissions. The 95 percent confidence interval for the mean based upon bootstrap resampling of the 68 calculated emission factors is 22 to 100 g/gal. If the population of vehicles are assumed to have a lognormal distribution for emissions, then the 95 percent confidence interval for the mean is 25 to 101 g/gal. Because of the good agreement between resampling and parametric bootstrap simulation based upon the lognormal distribution, it appears reasonable to use a confidence interval of approximately 25 to 100 g/gal. The sampling distribution for the mean based upon bootstrap simulation of a lognormal distribution fitted to the calculated emissions data is shown in Figure 6-29. 98
Cumulative Frequency
1.0 Gasoline School Buses Rock Quarry Road n = 68
0.8 0.6
Calculated From Data Set
0.4
Mean
0.2
95 % Confidence Interval for the Mean
0.0 0
500
1000 1500 2000 Calculated CO Emissions(g/gal)
2500
3000
Figure 6-27. Variability and Uncertainty in 68 Estimates of Gasoline School Bus CO Emissions (grams/gallon) based Upon Remote Sensing Measurements at the Rock Quarry Road Site.
Cumulative Frequency
1.0 0.8 Calculated from Data Set
0.6
Mean
0.4 Gasoline School Buses Rock Quarry Road n = 68
0.2
95 % Confidence Interval on the Mean
0.0 0
200
400 600 Calculated CO Emissions(g/mile)
800
1000
Figure 6-28. Variability and Uncertainty in 68 Estimates of Gasoline School Bus CO Emissions (grams/mile) based Upon Remote Sensing Measurements at the Rock Quarry Road Site.
99
Cumulative Probability
1.0 0.8 0.6 0.4 0.2 0.0 0
25 50 75 100 125 Mean Propane-Equivalent Hydrocarbon Emissions (g/gal)
150
Figure 6-29. Sampling Distribution for Mean Propane-Equivalent Hydrocarbon Emissions (g/gal) for Gasoline-Fueled School Buses Observed at the Rock Quarry Road Site based Upon an Assumed Lognormally Distributed Population for Inter-Vehicle Variability in Emissions.
Cumulative Probability
1.0
0.8 Lognormal
0.5
Resampling Normal
0.2
0.0 0
5
10
15
20
25
Lognormal
Figure 6-30. Comparison of Sampling Distributions for Mean Propane-Equivalent Hydrocarbon Emissions (g/mi) for Gasoline-Fueled School Buses Observed at the Rock Quarry Road Site based Upon Resampling, Normal Distribution and Lognormal Distribution for Inter-Vehicle Variability in Emissions.
100
The 95 percent confidence interval for the mean HC g/mi emission factors for Rock Quarry Road is 4.9 to 21.6 g/mi based upon resampling and 5.4 to 22.3 g/mi based upon a lognormal distribution. In contrast, the confidence interval is 1.5 to 22.6 g/mi based upon a normal distribution. The normal distribution is the least appropriate basis for constructing a confidence interval in this case, since it includes the possibility of negative values at the extreme lower table, as shown in Figure 6-30. These data suggest a range of approximately 5 to 22 g/mi are appropriate to represent the uncertainty in the mean. 6.5.2
Effect of Bus Characteristics on Emissions
In this section, we explore whether characteristics of each bus can be used to explain at least a portion of the observed variability in emissions for gasoline-fueled buses at the Rock Quarry Road site. The bus characteristics that we have considered include odometer reading, chassis year, capacity, and fuel economy. Scatter plots were used to determine whether any trends are apparent in the data. For example, Figure 6-31 shows a scatter plot of CO emissions versus odometer reading for the 68 observations of gasoline-fueled buses. No particular trend is apparent from this graph. Figure 6-32 provides similar information for hydrocarbon emissions versus odometer reading. Attempts were made to develop curve fits in which the odometer reading was used as an explanatory variable. The coefficients of determination for these curve fits were small and not significant. For example, for hydrocarbons, the coefficient of determination for an exponential curve fit was only 0.079. The chassis year was not statistically significant as an explanatory variable for emissions. Figures 6-33 and 6-34 are scatter plots of CO and HC emissions, respectively, versus the chassis year. There are no statistically significant trends in these figures. There were insufficient data to compare emissions of gasoline buses based upon the bus capacity. This is because all of the gasoline-fueled buses are of a similar size. Of the 68 data points, 63 were for 54-passenger buses. Four data points were for 48passenger buses, and one data point was for a 60-passenger bus. Fuel economy was not a statistically significant explanatory variable, as indicated in the scatter plots in Figures 6-35 and 6-36 for CO and HC emissions, respectively.
101
Calculated CO Emissions (g/mile)
10000
1000
100
10 100000
125000
150000 Odometer Reading (miles)
175000
200000
Figure 6-31. Scatter Plot of Estimated CO Emissions Versus Odometer Readings for 68 Observations of 21 Gasoline-Fueled School Buses at the Rock Quarry Road Site.
Calculated Propane-Equivalent Hydrocarbon Emissions (g/mile)
1000
100
10
1
0.1 100000
125000
150000 Odometer Reading (miles)
175000
200000
Figure 6-32. Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Odometer Readings for 68 Observations of 21 Gasoline-Fueled School Buses at the Rock Quarry Road Site.
102
Calculated CO Emissions (g/mile)
10000
1000
100
10 78
79
80
81 82 Chasis Year
83
84
85
Figure 6-33. Scatter Plot of Estimated CO Emissions Versus Chassis Year for 68 Observations of 21 Gasoline-Fueled School Buses at the Rock Quarry Road Site.
Calculated Propane-Equivalent Hydrocarbon Emissions (g/mile)
1000
100
10
1
0.1 78
79
80
81 82 Chasis Year
83
84
85
Figure 6-34. Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Chassis Year for 68 Observations of 21 Diesel-Fueled School Buses at the Rock Quarry Road Site.
103
Calculated CO Emissions (g/gal)
10000
1000
100
10 1
2
3 4 Fuel Economy (Miles per Gallon)
5
6
Figure 6-35. Scatter Plot of Estimated CO Emissions Versus Fuel Economy for 68 Observations of 21 Gasoline-Fueled School Buses at the Rock Quarry Road Site.
Calculated Propane-Equivalent Hydrocarbon Emissions (g/gal)
10000
1000
100
10
1 1
2
3 4 Fuel Economy (Miles per Gallon)
5
6
Figure 6-36. Scatter Plot of Estimated Propane-Equivalent Hydrocarbon Emissions Versus Fuel Economy for 68 Observations of 21 Diesel-Fueled School Buses at the Rock Quarry Road Site.
104
6.5.3
Effect of Vehicle Direction on Emissions
Vehicle direction was not found to be a statistically significant explanatory factor for differences in gasoline-fueled school bus emissions at the Rock Quarry Road site. Figures 6-37 and 6-38 show CO and HC emissions, respectively, cumulative distribution functions, means, and 95 percent confidence intervals for the mean for inbound and outbound buses. The inbound and outbound distributions for each pollutant are not significantly different from each other. 6.6
Gasoline-Fueled School Buses at Other Sites
Aside from the Rock Quarry Road site, the Laura Duncan Road site was the only other location at which an appreciable number of data points were obtained for gasolinefueled school buses. Figures 6-39 and 6-40 display cumulative distribution functions, means, and 95 percent confidence intervals for the mean of the estimated CO and HC emissions, respectively. Because of the small sample size (n=16), there is substantial uncertainty in the estimate of the fleet average emissions. For example, the average CO emissions are uncertain by plus or minus 44 percent, and the average propane-equivalent hydrocarbon emissions are uncertain by plus or minus 30 percent. As previously noted, the CO emissions at the Laura Duncan Road site appear to be significantly lower than for the Rock Quarry Road Site. The hydrocarbons emissions also appear to be substantially lower. The differences in emissions between the two sites may be attributable to differences in site conditions, such as vehicle speed.
105
Cumulative Frequency
1.0 0.8 0.6
Inbound
0.4
Inbound Mean 95 % CI on Inbound Mean
Gasoline School Buses Rock Quarry Road n = 68
0.2
Outbound Outbound Mean 95 % CI on Outbound Mean
0.0 0
200
400 600 Calculated CO Emissions(g/mi)
800
1000
Figure 6-37. Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Inbound Versus Outbound Gasoline Buses at Rock Quarry Road (n = 68).
Cumulative Frequency
1.0 0.8 0.6
Inbound Inbound Mean
0.4
95 % CI on Inbound Mean Outbound
0.2
Outbound Mean 95 % CI on Outbound Mean
0.0 0
10 20 30 40 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mile)
50
Figure 6-38. Comparison of Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Inbound Versus Outbound Gasoline Buses at Rock Quarry Road (n = 68).
106
Cumulative Probability
1.0 0.8 Gasoline Buses at Laura Duncan Road, n = 16
0.6
Calculated from Data Set
0.4
Mean
0.2 95 Percent Confidence Interval on the Mean
0.0 0
50
100 150 Calculated CO Emissions (g/mi)
200
250
Figure 6-39. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Gasoline Buses at Laura Duncan Road (n=16).
Cumulative Probability
1.0 0.8 0.6
Gasoline Buses at Laura Duncan Road, n = 16
0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval on the Mean
0.0 0
1 2 3 4 5 6 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
7
Figure 6-40. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Gasoline Buses at Laura Duncan Road (n=16).
107
6.7
Diesel School Buses at Other Sites
Data regarding diesel school buses were obtained at four other locations aside from the Rock Quarry Road location. These locations are Woodcroft (n=38), Laura Duncan Road (n=33), Wake Forest (n=9), and Garner (n=5). The variability in individual emission factor estimates, the average emission factor estimate, and the 95 percent confidence interval for the average for both CO and HC emissions are shown in a series of figures. Emissions estimates for CO and HC based upon data collected at the Woodcroft site are shown in Figures 6-41 and 6-42, respectively. These figures indicate that there is a substantial amount of variability in the individual emission estimates. For both pollutants, there are a small number of relatively high estimates, as seen above the 90th percentile of the cumulative distribution functions. For example, there is an individual emissions estimate of over 200 g CO/mile, and a HC emissions estimate of over 50 g HC/mile. These high numbers influence both the mean and the confidence interval for the mean. However, there is no basis for excluding these large values from the data set. The emissions estimates for the Laura Duncan Road site exhibit less variability from one measurement to another, and do not include as large extreme values as does the Woodcroft data set. Therefore, the means for Laura Duncan Road are smaller than for Woodcroft, and the confidence intervals on the means are narrower, as illustrated in Figures 6-43 and 6-44 for CO and HC emissions, respectively. The emissions values for Laura Duncan Road are similar to those for Rock Quarry Road, as previously discussed. The Wake Forest and Garner sites yielded only small amounts of data (n=9 and n=5, respectively). The resulting emission estimates for CO and HC for the Wake Forest data are shown in Figures 6-45 and 6-46, respectively. The corresponding graphs for the Garner site are given in Figures 6-47 and 6-48. These graphs illustrate that, for data with small sample sizes, the confidence intervals in the mean can be comparable in magnitude to the overall variation in the individual emission estimates. Because of the wide range of uncertainty in the mean values, it is difficult to draw any conclusions regarding potential similarities or differences in emissions compared to the other sites.
108
Cumulative Probability
1.0 0.8 0.6
Calculated from Data Set
0.4
Mean
Woodcroft Site n = 38
0.2
95 % Confidence Interval on the Mean
0.0 0
50
100 Calculated CO Emissions (g/mi)
150
200
Figure 6-41. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel Buses at the Woodcroft site (n=38).
Cumulative Probability
1.0 0.8 0.6
Calculated From Data Set
0.4
Mean
Woodcroft Site n = 38
0.2
95 % Confidence Interval for the Mean
0.0 0
10 20 30 40 Calculated Propane-Equivalent Hydrocarbon Emissions(g/mi)
Figure 6-42. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel Buses at the Woodcroft site (n=38).
109
50
Cumulative Probability
1.0 0.8 Laura Duncan Road, n = 33
0.6
Calculated from Data Set
0.4
Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
10
20 Calculated CO Emissions (g/mi)
30
40
Figure 6-43. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel Buses at the Laura Duncan Road site (n=33)
Cumulative Probability
1.0 0.8 Laura Duncan Road, n = 33
0.6 0.4
Calculated From Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
1 2 3 4 5 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
6
Figure 6-44. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel Buses at the Laura Duncan Road site (n=33).
110
Cumulative Probability
1.0 0.8 Wake Forest Site, n = 9
0.6 0.4
Calculated from Data Set Mean
0.2 95 Percent Confidence Interval for the Mean
0.0 0
25
50 Calculated CO Emissions (g/mi)
75
100
Figure 6-45. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel Buses at the Wake Forest site (n=9)
Cumulative Probability
1.0 0.8 Wake Forest Site, n = 9
0.6
Calculated from Data Set
0.4
Mean
0.2 95 Percent Confidence Interval for the Mean
0.0 0
25 50 75 100 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
125
Figure 6-46. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel Buses at the Wake Forest site (n=9).
111
Cumulative Probability
1.0 0.8 0.6 Calculated from Data Set
0.4
Mean
Garner Site n=5
0.2
95 Percent Confidence Interval for the Mean
0.0 50
100
150 200 Calculated CO Emissions (g/mi)
250
300
Figure 6-47. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel Buses at the Garner site (n=5).
Cumulative Probability
1.0 0.8 0.6 0.4
Calculated from Data Set
0.2
Mean
Garner Site n=5
95 Percent Confidence Interval for the Mean
0.0 0
25 50 75 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
100
Figure 6-48. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel Buses at the Garner site (n=5).
112
113
7.0
REMOTE SENSING MEASUREMENTS AND ESTIMATED EMISSION FACTORS FOR TRANSIT BUSES
Remote sensing measurements were obtained for three fleets of transit buses. These fleets include a portion of the Triangle Transit Authority bus fleet, older transit buses in use at Raleigh-Durham International Airport just prior to their retirement in 1996, and new transit buses placed into service at Raleigh-Durham International Airport in the fall of 1996. As previously noted, an attempt was made to collect data on transit bus of the Capital Area Transit system of the city of Raleigh. However, no valid data were obtained for these latter buses. 7.1
Data Collection and Database Development Activities
Emissions measurements were obtained over seven days of data collection for 31 diesel transit buses and three gasoline transit buses at four major locations. Table 7-1 summarizes the locations and dates at which data were collected for buses of the Triangle Transit Authority. In addition, data were collected on two dates at RDU airport, and attempts were made to collect data for CAT buses on an additional date. Summaries of the emissions estimates for the TTA and RDU buses are given in Tables 7-2 and 7-3, respectively. As documented in the tables, the number of valid measurements obtained were 37 for TTA diesel buses, 34 for the retired RDU buses, and 106 for the new RDU buses. In addition, six data points were obtained for TTA gasoline buses. Thus, a total of 183 valid measurements were obtained for transit buses. Spreadsheet-based databases were developed in a manner similar to those for school buses. However, less information was available regarding the characteristics of each bus. The Triangle Transit Authority provided information regarding the fuel economy of each bus. However, only a fleet average estimate of fuel economy was available for the RDU transit buses. Through some computer simulations based upon judgments regarding plausible ranges of uncertainty in fuel economy, it was determined that the lack of information regarding fuel economy has only a minor effect on the emissions estimates. Typically, the variability in the emission factor estimates is dominated by variability in the measured ratios of CO to CO2 and HC to CO2, which spans orders-of-magnitude. In contrast, variability in fuel economy estimates typically span a range of less than a factor of two from the lower 5th percentile to the upper 95th percentile.
114
Table 7-1.
Summary of Data Collection Sites for TTA Transit Buses Date Site October 16, 1996 Woodcroft October 23, 1996 Woodcroft November 11, 1996 Briggs Ave. November 12, 1996 Davis Drive
Table 7-2.
Summary of Estimated Average Emission Factors for Diesel-Fueled Transit Buses of the Triangle Transit Authority 95 Percent Confidence Standard Interval on the Mean Description Mean Deviation Lower Upper Count All TTA Bus Measurements CO g/gal 59.1 43.1 45.2 73.0 37 HC g/gal 14.8 12.4 10.7 18.8 37 CO g/mi 7.8 5.5 6.0 9.5 37 HC g/mi 1.9 1.6 1.4 2.5 37 Woodcroft Site, October 16, 1996 CO g/gal 52.0 31.9 26.4 77.5 6 HC g/gal 11.8 6.2 6.9 16.7 6 CO g/mi 7.0 4.2 3.6 10.4 6 HC g/mi 1.6 0.8 0.9 2.3 6 Woodcroft Site, October 23, 1996 CO g/gal 86.6 34.3 56.5 116.7 5 HC g/gal 16.4 7.8 9.6 23.3 5 CO g/mi 11.9 4.3 8.1 15.6 5 HC g/mi 2.2 1.0 1.4 3.1 5 Briggs Avenue Site, November 11, 1996 CO g/gal 51.4 43.9 26.6 76.3 12 HC g/gal 9.1 5.2 6.1 12.0 12 CO g/mi 6.8 6.1 3.4 10.3 12 HC g/mi 1.2 0.7 0.8 1.6 12 Davis Drive Site, November 12, 1996 CO g/gal 59.0 49.0 33.3 84.6 14 HC g/gal 20.3 17.5 11.2 29.4 14 CO g/mi 7.4 5.6 4.5 10.3 14 HC g/mi 2.6 2.3 1.4 3.8 14
115
Table 7-3.
Summary of Estimated Average Emission Factors for Diesel-Fueled Transit Buses at Raleigh Durham International Airport 95 Percent Confidence Standard Interval on the Mean Description Mean Deviation Lower Upper Count All Measurements of New RDU Buses CO g/gal 92.1 40.7 84.3 99.8 106 HC g/gal 15.1 9.5 13.3 16.9 106 CO g/mi 12.1 5.3 11.1 13.1 106 HC g/mi 2.0 1.2 1.7 2.2 106 New RDU Bus No. 3 CO g/gal 88.5 51.6 68.3 108.8 25 HC g/gal 12.4 7.3 9.5 15.3 25 CO g/mi 11.6 6.8 9.0 14.3 25 HC g/mi 1.6 1.0 1.2 2.0 25 New RDU Bus No. 4 CO g/gal 88.3 38.4 74.1 102.6 28 HC g/gal 15.4 9.9 11.8 19.1 28 CO g/mi 11.6 5.0 9.7 13.5 28 HC g/mi 2.0 1.3 1.5 2.5 28 New RDU Bus No. 7 CO g/gal 105.5 34.5 92.0 119.0 25 HC g/gal 17.9 12.1 13.2 22.6 25 CO g/mi 13.9 4.5 12.1 15.6 25 HC g/mi 2.3 1.6 1.7 3.0 25 All Measurements of Old RDU Buses CO g/gal 123.8 186.3 61.2 186.4 34 HC g/gal 33.1 70.3 9.5 56.8 34 CO g/mi 16.2 24.5 8.0 24.5 34 HC g/mi 4.4 9.2 1.2 7.5 34 Old RDU Bus No. 1 CO g/gal 56.8 12.0 49.0 64.6 9 HC g/gal 14.2 10.3 7.4 20.9 9 CO g/mi 7.5 1.6 6.4 8.5 9 HC g/mi 1.9 1.4 1.0 2.7 9 Old RDU Bus No. 5 CO g/gal 187.6 280.9 13.5 361.7 10 HC g/gal 53.4 112.6 0.0 123.2 10 CO g/mi 24.6 36.9 1.8 47.5 10 HC g/mi 7.0 14.8 -2.2 16.2 10
116
7.2
Summary of Transit Bus Emissions Estimates
The estimated emission factors based upon observations of TTA buses are given in Table 7-2. The estimated emission factors based upon observations of RDU buses are given in Table 7-3. Based upon all valid measurements at all sites, the average CO emission factor for the observed TTA buses is estimated to be 7.8 g/mile, with a range of uncertainty of plus or minus 25 percent. The propane-equivalent hydrocarbon emission factor is estimated to be 1.9 g/mile, with an uncertainty range of plus or minus 25 percent. Emissions data were obtained at three sites over four days. While there are some differences in the averages for each day of data collection, the range of average values is relatively narrow. For example, the average CO emission estimates based upon each day of observations range from 6.8 g/mi to 11.9 g/mi. The confidence intervals for these two values overlap. The average emission estimates based upon observations of TTA buses are significantly lower than for the diesel school buses. For example, the average emission factors based upon all 1,069 diesel school bus measurements is 13.0 g CO/mile, versus 7.8 g CO/mile for the TTA measurements. The average propane-equivalent hydrocarbon emission factor for all diesel school buses is 3.4 g/mile, versus 1.9 g/mile for the TTA measurements. It appears that the TTA bus fleet, which is comprised of smaller vehicles than the school bus fleet, has lower emissions per vehicle-mile of travel. However, the differences in emissions may be attributable to differences in site characteristics or vehicle operation. The old RDU buses are similar in design to the TTA buses. The average emissions of the old RDU buses appear to be higher than for the TTA buses. However, a direct comparison of the RDU emissions estimates with the TTA estimates may be misleading. The site at which the RDU buses were measured has a much steeper road grade than at many of the other sites used in this project. Therefore, it is likely that the higher emissions from the older RDU bus fleet is due to site characteristics and not due to differences in vehicle design or operation compared to the TTA bus fleet. Because it was possible to get repeated readings on a small number of buses at the RDU site, more detailed insight can be obtained regarding intra-vehicle versus intervehicle variability in emissions. For the old RDU bus fleet, a total of 34 measurements were obtained for 5 buses. Of these, 19 measurements were obtained on two buses. The average CO emissions for the two buses for which the largest number of data points were 117
obtained differ by a factor of three: old RDU Bus No. 5 yielded an average emission estimate of 25 g/mi versus 7.5 g/mi for old RDU Bus No. 1 at the same site. However, as will be illustrated later, old RDU Bus No. 5 had one large data value which substantially impacts the mean. Although the average emissions appear to differ, the difference is not likely to be significant since the confidence interval for the mean of old RDU Bus No. 5’s CO emissions encloses that for old RDU Bus No. 1. More reliable data were obtained for three of the buses in the new RDU parking shuttle fleet. Of the 106 data points obtained for the new buses, three buses accounted for approximately equal shares of 78 of the data points. New RDU Buses Nos. 3 and 4 had very similar average emissions estimates of approximately 11.6 g CO/mile and 2 g HC/mile. New RDU Bus No. 7 had slightly higher emissions than the other two. However, there is overlap in the confidence intervals for all three buses. The average of the CO emissions estimates for the new buses is approximately 12 g/mile. This average is approximately 25 percent less than the average emissions for the old RDU fleet measured at the same site. Because of the uncertainty regarding the mean emissions estimates for both the old and new fleets, it is possible that the true difference in emissions may be different than this estimate. The RDU data provide insight that the variability in emissions for individual buses can be large. Thus, intra-vehicle variability in emissions may be the dominating factor, for this fleet. The emissions estimates for both the TTA and RDU fleets are described in more detail in the following sections. 7.3
Transit Bus Emissions Estimates for the Observed Triangle Transit Authority Buses
Because the TTA buses run on an infrequent schedule, a relatively small number of data points were obtained for these buses compared to the other two bus fleets for which emissions data were collected. Most of the measurements (n=37) of TTA buses were for diesel buses. In addition, there are six measurements for gasoline buses. For the calculated CO emissions for diesel buses, cumulative distribution functions, means, and 95 percent confidence intervals for the mean are shown in Figures 7-1 through 7-5 for the following data sets: (1) all measurements of TTA diesel buses; (2) Woodcroft site on October 16, 1996; (3) Woodcroft site on October 23, 1996; (4) Briggs Avenue site; and (5) Davis Drive site. These figures illustrate that the confidence
118
intervals for the means of the CO emissions estimates for each of the four data collection days overlap. Most of the intervals, for example, include values from 5 g/mi to 10 g/mi.
Cumulative Probability
1.0 0.8
All TTA Diesel Bus Measurements, n = 37
0.6
Calculated from Data Set
0.4
Mean
0.2 95 Percent Confidence Interval for Mean
0.0 0
5
10 15 20 Calculated CO Emission Factor (g/mi)
25
30
Figure 7-1. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel TTA Buses (All Sites, n=37)
Cumulative Probability
1.0 TTA Diesel Bus Measurements at Woodcroft Site on October 16, 1996 n=6
0.8 0.6 0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
5
10 15 20 Calculated CO Emissions (g/mi)
25
30
Figure 7-2. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel TTA Buses (Woodcroft Site, October 16, 1996, n=6)
119
Cumulative Probability
1.0 TTA Diesel Bus Measurements at Woodcroft Site on October 23, 1996 n=5
0.8 0.6
Calculated from Data Set
0.4
Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
5
10 15 20 Calculated CO Emissions (g/mi)
25
30
Figure 7-3. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel TTA Buses (Woodcroft Site, October 23, 1996, n=5)
Cumulative Probability
1.0 0.8
TTA Diesel Bus Measurements, Briggs Ave Site, n = 12
0.6 0.4
Calculated from Data Set Mean
0.2 95 Percent Confidence Interval for the Mean
0.0 0
5
10 15 20 Calculated CO Emission Factor (g/mi)
25
30
Figure 7-4. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel TTA Buses (Briggs Avenue Site, n=12)
120
Cumulative Probability
1.0 0.8 0.6
TTA Diesel Bus Measurements, Davis Drive Site, n = 14
0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
5
10 15 20 Calculated CO Emissions (g/mi)
25
30
Figure 7-5. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Diesel TTA Buses (Davis Drive Site, n=14)
For the calculated propane-equivalent hydrocarbon emissions for diesel buses, cumulative distribution functions, means, and 95 percent confidence intervals for the mean are shown in Figures 7-6 through 7-10 for the following data sets: (1) all measurements of TTA diesel buses; (2) measurements taken at the Woodcroft site on October 16, 1996; (3) measurements taken at the Woodcroft site on October 23, 1996; (4) measurements taken at the Briggs Avenue site; and (5) measurements taken at the Davis Drive site. The mean HC emission estimate is typically a value near 2 g/mi. The exception is for the Briggs Avenue site, for which the upper bound of the 95 percent confidence interval for the mean is only 1.6 g/mi. These results imply that the conditions at the Briggs Avenue site are such that the buses tend to produce less propane-equivalent HC than for the other sites at which measurements were taken. However, the CO emissions at the Briggs Avenue site were similar to other sites, such as Davis Drive. These results imply that there may be differences in the variability of CO and HC emissions from one site to another.
121
Cumulative Probability
1.0 0.8 All TTA Diesel Bus Measurements, n = 37
0.6
Calculated from Data Set
0.4
Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
2 4 6 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
8
Figure 7-6. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel TTA Buses (All Sites, n=37)
Cumulative Probability
1.0 TTA Diesel Bus Measurements at Woodcroft Site on October 16, 1996 n=6
0.8 0.6 0.4
Calculated from Data Set Mean
0.2 95 Percent Confidence Interval for the Mean
0.0 0
1
2 3 4 5 6 7 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
Figure 7-7. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel TTA Buses (Woodcroft Site, October 16, 1996, n=6)
122
8
Cumulative Probability
1.0 TTA Diesel Bus Measurements at Woodcroft Site on October 23, 1996 n=5
0.8 0.6 0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
1
2 3 4 5 6 7 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
8
Figure 7-8. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel TTA Buses (Woodcroft Site, October 23, 1996, n=5)
Cumulative Probability
1.0 0.8
TTA Diesel Bus Measurements, Briggs Ave Site, n = 12
0.6 0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
1
2 3 4 5 6 7 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
8
Figure 7-9. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel TTA Buses (Briggs Avenue Site, n=12)
123
Cumulative Probability
1.0 0.8 0.6
TTA Diesel Bus Measurements, Davis Drive Site, n = 14
0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
2 4 6 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
8
Figure 7-10. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Diesel TTA Buses (Davis Drive Site, n=14)
The six data points for gasoline buses in the TTA fleet are shown in Figure 7-11 for CO emissions estimates and in Figure 7-12 for propane-equivalent hydrocarbon emissions estimates. These data are for three buses observed at the Woodcroft and Briggs Avenue sites. Because of the small sample size, there is substantial uncertainty in the mean values, as indicated by the wide confidence intervals for the mean. The wide confidence intervals are also influenced by a few large emissions estimates within the data set. The emissions estimates in these figures are qualitatively similar to the gasolinefueled school bus estimates for the Laura Duncan Road site, in that the confidence intervals for the means of both datasets for both pollutants overlap. The Woodcroft and Briggs Avenue sites both typically had vehicle traffic moving at 30 to 40 mph, whereas the posted speed limit on Laura Duncan Road is 45 mph. Thus, it is likely that the buses at these three sites were moving at similar speeds, which may account for the apparent similarities. On the other hand, because there are only six data points for the gasolinefueled TTA buses, the large amount of uncertainty in the mean emissions does not enable development of any definitive conclusions regarding comparisons.
124
Cumulative Probability
1.0 0.8 TTA Gasoline Bus Measurements at All Sites n=6
0.6 0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
50
100 150 Calculated CO Emissions (g/mi)
200
250
Figure 7-11. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Gasoline TTA Buses (n=6)
Cumulative Probability
1.0 0.8 0.6
All TTA Gasoline Bus Measurements, n = 6
0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
2 4 6 8 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
Figure 7-12. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Gasoline TTA Buses (n=6)
125
10
7.4
Transit Bus Emissions Estimates for the Observed Raleigh-Durham International Airport Buses
The measurement of buses at Raleigh-Durham International Airport is a unique aspect of this study. Because a small fleet of transit buses is continuously looping around the airport, it was possible to collect a relatively large amount of data regarding several buses. Also, because measurements were taken at the same site before and after the older transit bus fleet was replaced with a new transit bus fleet, it is possible to compare emissions for different buses at the same site. Furthermore, because the site features a relatively steep road grade, there is less variability in engine load than would be the case for a level site. The largest amount of data was collected for the new transit buses at the RDU site. A summary of the estimates of the CO and propane-equivalent hydrocarbon emissions for all 106 data points are given in Figures 7-13 and 7-14, respectively. These data indicate that the overall range of variability in CO emissions is similar to that estimated for the TTA diesel buses, but with a slight upward shift in values. The mean CO emissions have a 95 percent confidence interval of less than plus or minus 10 percent. There were three new RDU buses for which 25 or more data points were obtained. These are New RDU Bus Nos. 3, 4, and 7. The cumulative distribution functions for variability in estimated CO emissions based upon each individual emission measurement are approximately similar in all three cases. These cdfs are shown in Figures 7-15, 7-17, and 7-19. New Bus No. 7 appears to have slightly less variation from the smallest to the largest observed values than do the other two buses. New Bus No. 7 also tends to have a slightly larger mean emission value. However, the 95 percent confidence intervals for the means of all three buses overlap, making it difficult to draw conclusions that emissions are significantly different from one bus to another. These results imply that, even for the relatively controlled conditions of this specific study, there is substantial variability from one measurement to the next due to the nearly instantaneous nature of the remote sensing measurements. Similar findings are obtained for the propane-equivalent hydrocarbon emissions of the new RDU buses, as indicated by a comparison of the results for New Buses Nos. 3, 4, and 7 in Figures 7-16, 7-18, and 7-20, respectively. The variability in individual emissions estimates is approximately one to two orders-of-magnitude, and the range of uncertainty in the mean, as represented by a 95 percent confidence interval, is approximately plus or minus 25 percent.
126
Cumulative Probability
1.0 0.8
All New RDU Diesel Bus Measurements, n = 106
0.6
Calculated from Data Set
0.4 Mean
0.2 95 Percent Confidence Interval for Mean
0.0 0
5
10 15 20 25 Calculated CO Emission Factor (g/mi)
30
35
Figure 7-13. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for New Diesel Buses at RDU Airport (n=106)
Cumulative Probability
1.0 0.8 All New RDU Diesel Bus Measurements, n = 106
0.6
Calculated from Data Set
0.4
Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
2 4 6 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
Figure 7-14. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for New Diesel Buses at RDU Airport (n=106)
127
8
Cumulative Probability
1.0 0.8 New RDU Bus No. 3 n = 25
0.6 0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
5
10 15 20 Calculated CO Emissions (g/mi)
25
30
Figure 7-15. Inter-Measurement Variability and Uncertainty in Average for Calculated CO Emissions for New RDU Bus No. 3 (n=25)
Cumulative Probability
1.0 0.8 0.6
New RDU Bus No. 3 n = 25
0.4
Calculated from Data Set Mean
0.2 95 Percent Confidence Interval for the Mean
0.0 0
1
2 3 4 5 6 7 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
8
Figure 7-16. Inter-Measurement Variability and Uncertainty in Average for Calculated Propane-Equivalent Hydrocarbon Emissions for New RDU Bus No. 3 (n=25)
128
Cumulative Probability
1.0 New RDU Bus No. 4 n = 28
0.8 0.6
Calculated from Data Set
0.4
Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
5
10 15 20 Calculated CO Emissions (g/mi)
25
30
Figure 7-17. Inter-Measurement Variability and Uncertainty in Average for Calculated CO Emissions for New RDU Bus No. 4 (n=28)
Cumulative Probability
1.0 New RDU Bus No. 4 n = 28
0.8 0.6 0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
1
2 3 4 5 6 7 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
8
Figure 7-18. Inter-Measurement Variability and Uncertainty in Average for Calculated Propane-Equivalent Hydrocarbon Emissions for New RDU Bus No. 4 (n=28)
129
Cumulative Probability
1.0 0.8 0.6
New RDU Bus No. 7 n = 25
0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
5
10 15 20 Calculated CO Emission Factor (g/mi)
25
30
Figure 7-19. Inter-Measurement Variability and Uncertainty in Average for Calculated CO Emissions for New RDU Bus No. 7 (n=25)
Cumulative Probability
1.0 0.8 New RDU Bus No. 7 n = 25
0.6 0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
1
2 3 4 5 6 7 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
8
Figure 7-20. Inter-Measurement Variability and Uncertainty in Average for Calculated Propane-Equivalent Hydrocarbon Emissions for New RDU Bus No. 7 (n=25)
130
The emissions estimates for the old RDU bus fleet exhibited more variability than did the emissions estimates for the new RDU bus fleet. For example, in Figure 7-21, the range of variability in CO emissions for the old buses is from approximately 4 g/mi to over 125 g/mi, compared to an estimated range of variation of approximately 3 g/mi to 30 g/mi for the new buses. Because of the wider variability and the smaller number of data points, the uncertainty in the mean CO emissions is larger for the old bus, with a range of plus or minus 50 percent. The uncertainty in the mean propane-equivalent hydrocarbon emissions is approximately plus or minus 75 percent. While the mean emissions estimates for the old buses appear to be larger than for the new buses for both pollutants, it is not possible to draw definitive conclusions regarding this comparison due to the wide range of uncertainty in the mean values for the old buses. Because of the relatively small number of data points obtained for the old RDU buses, there are fewer data available to characterize the emissions of individual buses. For old RDU Bus No. 1, there were 9 observations, and for old RDU Bus No. 5, there were 10 observations. Estimated emissions for CO and propane-equivalent hydrocarbon emissions for each of these buses are shown in Figures 7-23 through 7-26. Based upon the available data, it appears that RDU Bus No. 1 may have significantly lower emissions of CO than the other older buses. The ranges of uncertainty in the mean emissions for old Bus No. 5 are wide due to the small sample size and the presence of a small number of large values in the data set. Therefore, no definitive comparison is possible for this bus.
131
Cumulative Probability
1.0 0.8
All Old RDU Diesel Bus Measurements, n = 34
0.6
Calculated from Data Set
0.4 Mean
0.2 95 Percent Confidence Interval for Mean
0.0 0
25
50 75 Calculated CO Emission Factor (g/mi)
100
125
Figure 7-21. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated CO Emissions for Old Diesel Buses at RDU Airport (n=34)
Cumulative Probability
1.0 0.8 All Old RDU Diesel Bus Measurements, n = 34
0.6
Calculated from Data Set
0.4
Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
10 20 30 40 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
50
Figure 7-22. Inter-Vehicle Variability and Fleet Average Uncertainty for Calculated Propane-Equivalent Hydrocarbon Emissions for Old Diesel Buses at RDU Airport (n=34)
132
Cumulative Probability
1.0 0.8 Old RDU Bus No. 1 n=9
0.6 0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
5
10 15 20 Calculated CO Emissions (g/mi)
25
30
Figure 7-23. Inter-Measurement Variability and Uncertainty in Average for Calculated CO Emissions for Old RDU Bus No. 1 (n=9)
Cumulative Probability
1.0 0.8 0.6
Old RDU Bus No. 1 n=9
0.4
Calculated from Data Set Mean
0.2 95 Percent Confidence Interval for the Mean
0.0 0
1
2 3 4 5 6 7 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
8
Figure 7-24. Inter-Measurement Variability and Uncertainty in Average for Calculated Propane-Equivalent Hydrocarbon Emissions for Old RDU Bus No. 1 (n=9)
133
Cumulative Probability
1.0 0.8
Old RDU Bus No. 5 n = 10
0.6
Calculated from Data Set
0.4
Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
10
20
30 40 50 60 70 Calculated CO Emissions (g/mi)
80
90
100
Figure 7-25. Inter-Measurement Variability and Uncertainty in Average for Calculated CO Emissions for Old RDU Bus No. 5 (n=10)
Cumulative Probability
1.0 0.8 0.6
Old RDU Bus No. 5 n = 10
0.4
Calculated from Data Set Mean
0.2
95 Percent Confidence Interval for the Mean
0.0 0
10 20 30 40 Calculated Propane-Equivalent Hydrocarbon Emissions (g/mi)
50
Figure 7-26. Inter-Measurement Variability and Uncertainty in Average for Calculated Propane-Equivalent Hydrocarbon Emissions for Old RDU Bus No. 5 (n=10)
134
135
8.0
DISCUSSION, CONCLUSIONS, AND RECOMMENDATIONS As noted in the introduction, the objectives of this project were to: (1) Conduct on-road remote sensing of pollutant emissions from selected types of vehicles (i.e. school and transit buses). (2) Determine the on-road emission rates of such vehicles. (3) Estimate the number of passengers carried by selected non-controlled vehicles such that emission rates can be related to a per passenger basis. (4) Estimate the fraction of high emitting vehicles in the traffic stream. (5) Collect sufficient data to satisfy statistical significance tests.
Objectives (1) and (2) have motivated the bulk of the effort in this project. A total of 1,340 valid measurements of on-road emissions ratios of CO/CO2 and HC/CO2 were obtained for 265 diesel-fueled school buses, 36 gasoline-fueled school buses, 19 dieselfueled buses of the Triangle Transit Authority, 3 gasoline-fueled buses of TTA, and 12 diesel-fueled transit buses at Raleigh-Durham International Airport over the course of 22 days of field work. The development of databases based upon the observed ratios of CO/CO2 and HC/CO2 and available information regarding characteristics of the observed buses involved detailed review of both data and video records from the RSD, as well as interactions with several agencies, such as the North Carolina Department of Public Instruction, Wake County Schools Department of Transportation, TTA, and the RDU airport authority. Information regarding the number of passengers carried were provided by the NCDPI and TTA. Such data were not readily available for the RDU buses. An emission factor model was developed based upon a detailed review of factors affecting fuel composition and the development of a simplified combustion model. 8.1 Limitations and Caveats The emissions estimates developed in this study must be used with some caution. Some of the key factors to consider include: • Vehicle emissions may depend upon site conditions and driver behavior • Although several sites were used for both school buses and TTA buses, nonetheless only a limited sample of sites were included in this study. Thus, the emissions estimates may not be representative of other types of sites not included here. • Because the RSD typically measures the emissions that occur over a 0.6 second period, there is substantially variation from one measurement to another, even for the same bus at the same site. The variation may be attributable to 136
differences in driver behavior that are not observable with the current equipment. • The wide range of variability from one emission measurement to another contributes to uncertainty regarding estimates of vehicle or fleet average emissions. • The grams per mile emission factors were calculated using average annual fuel economy estimates in combination with emissions data taken over 0.6 seconds. Variability in fuel economy for the same averaging times as the emissions data is likely to be much larger than indicated by the annual averages. • For the most part, the measurements in this study were taken during summer weather conditions. These data may not be representative of wintertime emissions even for the same sites as were used in this study. • Emissions of CO differ significantly for diesel and gasoline fueled school buses. Gasoline-fueled buses have significantly higher CO emissions than diesel-fueled school buses. • The measurements of hydrocarbon emissions account for only a portion of total hydrocarbon emissions. This is because the RSD is designed and calibrated for hydrocarbons similar to alkanes and alkenes. The HC emission factors produced in this study, therefore, are not inclusive of all possible hydrocarbon compounds emitted from the tailpipe. The amount of bias in these emissions factors is not well known, but may be as much as 50 percent or more lower than the actual total hydrocarbon emissions. • The measurements of hydrocarbon emissions do not include particulate matter. Especially for diesel buses, a portion of the hydrocarbons in the engine exhaust may be contained in or condensed upon particulate matter. This leads to additional bias in the hydrocarbon emission factors. • Because the RSD is calibrated using a cylinder gas that represents a highemitting gasoline vehicle, the accuracy of the measurements may not be as good for other emissions values. Some preliminary work at the U.S. Environmental Protection Agency suggests that RSDs have a nonlinear response to different emission levels, which introduces a bias in the observed values. Based upon these considerations, the CO emission factors used in this study can be considered to be more reliable than the HC emission factors. The HC emission factors should be viewed as a lower bound on the true emissions of hydrocarbons from the observed vehicles. Because variability in the speciation of total hydrocarbons from engine exhaust is not well known, the amount of bias in the hydrocarbon emission factors is not readily quantifiable. Extensive research was required prior to field data collection in order to identify suitable sites. A number of practical issues were encountered during data collection, but 137
overall it was possible to obtain a significant number of data points over the course of the study for the purposes of performing various statistical analyses. The large amount of inter-vehicle variability appears to be a confounding factor in the search for explanations regarding the observed differences in emissions. For example, a number of factors associated with diesel and gasoline-fueled school bus characteristics and site characteristics were explored as potential explanations for differences in emissions. Due to the large variation within the data sets, there is a relatively large amount of uncertainty regarding statistics based upon the data, such as the mean. Similarly, for the fleet of observed transit buses at Raleigh Durham International Airport, efforts were made to determine whether emissions differ significantly from one vehicle to another. However, the relatively large amount of uncertainty in the average emission factors precludes the formation of definitive conclusions in this regard. Because of the wide range of variability in individual measurements, is often difficult to identify differences in the mean emissions associated with different subpopulations of vehicles. It is likely that there are important but unobserved factors that are influencing emissions. For example, speed and acceleration are likely to be useful explanatory variables. However, the currently available RSD does not include capability to measure speed and acceleration. New equipment with this capability is currently being procured by NC DEHNR. Speed and acceleration would provide a more direct indication of the effect of driver behavior on emissions. For example, two vehicles operating at the same speed as they pass the RSD may have different emissions because of differences in acceleration. 8.2
Estimates of Per-Passenger Emissions
Given the caveats of the preceding section, a preliminary estimate is made regarding emissions of CO and hydrocarbons for both school buses and TTA transit buses on a per passenger basis. In Table 3-3, the total number of school bus vehicle-miles traveled by public school buses in North Carolina was 147 million for the 1994 to 1995 school year. Based upon school bus data for the Rock Quarry Road site, the average fuel economy of gasoline-fueled school buses is assumed to be 4.5 miles per gallon, while the average fuel economy of diesel buses is assumed to be 7.6 miles per gallon. From Table 3-4, the total amount of diesel fuel consumed in the 1994 to 1995 school year was 14,756,240 gallons. Using the average fuel economy estimate of 7.6 miles per gallon, the estimated vehicle miles traveled by diesel vehicles is 112 million miles. From Table 3-4, the amount of gasoline used in the same time frame was 7,013,753 gallons. Using the 138
average fuel economy estimate of 4.5 miles per gallon, the estimated vehicle miles traveled by gasoline vehicles is 32 million miles. The estimated total of 145 million miles agrees closely with the reported total of 147 million miles. The total public school bus ridership in North Carolina was reported to be 694,210. Thus, based upon an average CO emission factor of 13.0 g/mi for diesel school buses from Table 6-3, and an average CO emission factor of 207 g/mi for gasoline school buses from Table 6-4, the average CO emissions attributable to each rider are approximately 12 kg of CO per daily rider per year. Approximately 80 percent of this total is due to gasoline-fueled buses, which comprise a decreasing share of the public school bus fleet. If all the school buses were diesel, then the average CO emissions per rider per year would be approximately 2.8 kg per rider per year. The annual average propane-equivalent hydrocarbon emissions per public school bus rider per year is approximately 1.0 kg per rider per year. It is possible that the total hydrocarbon emissions are double or more than this amount. For the TTA bus fleet, based upon an estimate of 85,000 vehicle miles per month, we assume that the fleet travels approximately one million vehicle-miles per year. Most of the fleet is diesel-fueled, although insufficient data are readily available in order to estimate the disaggregation of fuel consumption by fuel type. If we assume for simplicity that all of the vehicle-miles-traveled are due to diesel buses, then we can produce a lower bound on the annual per-passenger emissions. For CO, using an average emission factor of 7.8 g/mi from Table 7-2, the approximate annual emissions are 4.9 kg CO per passenger per year. For propane-equivalent hydrocarbons, the emissions are approximately 1.2 kg per passenger per year, based upon an average emission factor of 1.9 g/mi. The actual total hydrocarbon emissions may be double or more than this amount. Although the TTA buses have lower average emission factors than the school buses, the per passenger totals are larger. This is likely to be due to longer bus routes for the TTA buses compared to school buses, since the TTA provides intercity service. There are a variety of other possible explanatory factors that would require more extensive measurements in order to evaluate. For example, the effect of ambient temperature on emissions would require collection of substantial amounts of data under different ambient conditions. 8.3
Estimates of the High-Emitting Fraction of Vehicles
It is difficult to provide a quantitative answer in response to this original objective of the study. Commonly, remote sensing measurements are misinterpreted to indicate the fraction of vehicles that are high emitters. In fact, the distributions of emissions merely 139
indicate that a fraction of the observations occurred during a time period in which a vehicle was producing high emissions. The study of emissions of individual school buses and of individual RDU transit buses illustrates that even individual vehicles can produce a wide range of emissions readings. Therefore, it is difficult to classify an individual vehicle as a high emitter based upon a small number of remote sensing emissions measurements. It is not possible to draw rigorous conclusions in this study regarding the presence or absence of vehicles that are systematically high emitters. To do so would require more detailed measurements on individual vehicles for a much larger number of individual vehicles. What is clear from this study is that any of the vehicles for which repeated measurements were obtained can produce high emissions, on an instantaneous basis. Therefore, any of these vehicles could potentially be misidentified as a high emitter if only one measurement were taken that coincides with an instantaneous episode of high emissions. 8.4
Sample Size and Statistical Significance
The wide range of variability in the data obtained in this study suggests that much larger data sample sizes are needed to obtain narrow confidence intervals for the mean values and for the purpose of identifying explanatory variables. In analyzing the RSD data, confidence intervals were estimated and considered when making comparisons between data sets. In the course of data analysis, a large number of pair-wise t-tests were performed. These tests typically yielded negative results, indicating that there were not significant differences between the datasets being compared. 8.5
Recommendations
The emission factors developed in this study represent estimates of on-road emissions of selected types of vehicles for selected sites. The variability and uncertainty in the emissions factors are quantified. Due to the wide range of variability and uncertainty in the emission factors, it was difficult to identify explanatory variables which could be used to disaggregate the dataset. For example, the variability in emissions for individual vehicles was found to be comparable to the variability in emissions for all observed vehicles, as in the case of diesel-fueled school buses and new transit buses at RDU airport. These findings suggest that additional data are needed regarding the operating conditions for each individual vehicle in order to find meaningful explanations for differences in emissions observations. 140
The procurement of a speed and acceleration measurement system for use with the RSD may enable the development of better explanations for differences in vehicle emissions, even for individual vehicles. It is recommended that additional data collection be conducted using the new speed and acceleration measurement capabilities to determine whether these yield significant insights regarding variability in emissions. A preliminary study, such as at RDU airport, could be used to evaluate the utility of the speedacceleration data prior to more extensive data collection. For example, if variability in emissions measurements for individual buses can be explained in part by differences in speed and acceleration for each measurement, then it is likely that speed and acceleration would be useful in explaining differences in emissions for other vehicles or fleets. If initial testing under the relatively controlled conditions at RDU is successful, then applications of the new speed-acceleration measurement capabilities to school buses at the Rock Quarry Road site would be a logical next step. Both of these sites can yield significant amounts of data points for individual buses, and the Rock Quarry Road site also can yield information regarding approximately 200 buses. As more information becomes available regarding the nonlinearity in the response of the RSD to different values of the emissions ratios, it may become necessary to reanalyze the data contained in this study to correct for potential biases in the instrumentation. Because the data taken in this study were primarily for summertime conditions, it would be useful to collect comparable data for winter time conditions to evaluate the effect of ambient temperature on emissions estimates. This study focused upon collection of data for CO and hydrocarbons. Recently, NOx measurement capabilities for RSDs have become commercially available. A potential limitation of the new NOx measurement capabilities is that they may be less precise than the capability for CO measurements. Because NOx emissions from diesel vehicles are typically higher than for gasoline vehicles, it is important to consider acquisition of a NOx measurement capability and subsequent application to measurement of the bus fleets studied here.
141
9.0
REFERENCES
Ashbaugh, L.L., D.R. Lawson, G.A. Bishop, et al. (1992), "On-Road Remote Sensing of Carbon Monoxide and Hydrocarbon Emissions During Several Vehicle Operating Conditions", Presented at AWMA/EPA Conference on PM10 Standards and Nontraditional Particulate Source Controls, Phoenix, AZ, January. Austin, B.S. (1994), "The Conformity Challenge: Finding Conformity Under The New Transportation/Air Quality Conformity Guidelines," Paper No. 94-MP13.01, Proceedings, 87th Annual Meeting, Air and Waste Management Association, Pittsburgh, PA. Austin, T.C., F. J. DiGenova, T.R. Carlson (1994), Analysis of the Effectiveness and Cost-Effectiveness of Remote Sensing Devices. Prepared by Sierra Research, Inc. for U.S. Environmental Protection Agency. May 1994. Biggs, D.C. (1989). Vehicle Consumption Estimation Program ARFCOM User Guide. Technical Manual ATM 28, Australian Road Research Board, South Victoria, Australia. Bishop, G. A. (1994), "Enhancements of Remote Sensing for Vehicle Emissions in Tunnels", Journal of the Air and Waste Management Association, 44(2) pp. 169 Bishop, G.A., D.A. Stedman, J.E. Peterson, et al. (1993), "A Cost-Effectiveness Study of Carbon Monoxide Emissions Reduction Utilizating Remote Sensing," Journal of the Air and Waste Management Association, 43(7):978-988. Bishop, G.A. (1996), Personal Communication with David Eichenberger. Black, F. (1992), "Control of Motor Vehicle Emissions: The U.S. Experience," manuscript prepared for submission to J. of Environmental Control Technology, by U.S. Environmental Protection Agency, Research Triangle Park, NC. Brady, J.E., and G.E. Humiston (1978). General Chemistry - Principle and Structure. John Wiley and Sons: New York. 1978. Cadle S. and R. D. Stephens (1991), "Remote Sensing Measurements of Carbon Monoxide Emissions from On-Road Vehicles", Journal of the Air and Waste Management Association, 41(1) pp 39 Cadle S. and R. D. Stephens (1994), "Remote Sensing of Vehicle Exhaust Emissions", Environmental Science and Technology, 28(6) pp 258. Cooper, C.D., and F.C. Alley (1994). Air Pollution Control: A Design Approach. Waveland Press: Prospect Heights, Illinois. Decisioneering (1993). Crystal Ball User's Manual, Version 3.0. Decisioneering, Inc: Denver, Colorado. 142
Elliott, D., C. Kaskavaltizis, and T. Topaloglu (1992). "Evaluation of the Stedman (FEAT) Vehicle Emission Sensing System." Society of Automotive Engineers Paper No. 922314, Warrendale, PA, October 1992. Efron, B., and R.J. Tibshirani (1993). An Introduction to the Bootstrap. Chapman and Hall: New York. EPA (1993). Mobile5 User's Guide. U.S. Environmental Protection Agency. Ann Arbor, Michigan. Frey, H.C. (1992). Quantitative Analysis of Uncertainty and Variability in Environmental Policy Making. American Association for the Advancement of Science. Washington, DC. Flagan, R.C., and J.H. Seinfeld (1988). Fundamentals of Air Pollution Engineering. Prentice Hall: New York. Gorse, R. A., et al. (1994), "Real World Vehicle Emissions: A Summary of the Fourth Annual CRC_APRAC On-Road Vehicle Emissions Workshop", Journal of the Air and Waste Management Association, 44(10) pp. 1180. Hirao, O., and R.K. Pefley (1988). Present and Future Automotive Fuels: Performance and Exhaust Certification. Johne Wiley and Sons: New York. Iman, R.L.; Shortencarier, M.J. (1984) "A Fortran 77 Program and User's Guide for the Generation of Latin Hypercube and Random Samples for Use With Computer Models". Report No. SAND83-2365; Sandia National Laboratories: Albuquerque, NM, March, 1984. Kini, M.D. (1996), “Probabilistic Modeling of Exhaust Emissions From Light Duty Gasoline Vehicles,” Master’s Thesis, Department of Civil Engineering, North Carolina State University, Raleigh, NC. Knapp, Kenneth T. (1995). U.S. Environmental Protection Agency, Research Triangle Park, NC. Personal Communication with H.C. Frey, North Carolina State University. January. Lawson, D. R. et al (1990), "Emissions from in-use motor vehicles in Los Angeles: A Pilot Study of Remote Sensing and the Inspection and Maintenance Program", Journal of the Air and Waste Management Association, 40(8) pp 1096. Leonard, S.A., and T.A. Formby (1994), "Mobile Source Emission Controls — Progress and Prognosis," Paper No. 94-TP53.03, Proceedings, 87th Annual Meeting, Air and Waste Management Association, Pittsburgh, PA.
143
Lumina (1994). Demos User's Manual. Lumina Decision Systems: Palo Alto, California. NCDPI School Bus Specification, Sept. '95, personal conversation with Jim Ritchie, TTA. NRC (1992a). Rethinking the Ozone Problem in Urban and Regional Air Pollution. National Academy Press: Washington, DC. NRC (1992b). Automotive Fuel Economy: How Far Should We Go? National Research Council, National Academy Press: Washington, DC. Obert, E.F. (1968). Internal Combustion Engines. International Textbook Company: Scranton, PA. O'Connor, K., and R.G. Ireson (1994), "Survey on Current Practices in Air Quality Modeling," Paper No. 94-RP107B.03, Proceedings, 87th Annual Meeting, Air and Waste Management Association, Pittsburgh, PA. Owen, K., and T. Coley (1995). Automotive Fuels Reference Book, Society of Automotive Engineers: Warrendale, PA. Redmond, Donald (1995). NC DEHNR. Personal Communication with H.C. Frey. Ritchie, J. (1996), personal communication with David Eichenberger. Robert Bosch, GmbH (1986), “Automotive Handbook,” Robert Bosch, GmbH, Stuttgart, Germany. Rueff , M. R. (1992), "The Cost of Reducing Emissions from Late Model High-Emitting Vehicles Detected Via Remote Sensing", Journal of the Air and Waste Management Association, 42(7) pp 921. SAE (1988), “Automotive Fuels — Recommended Practice,” J312, Report of the Fuels and Lubricants Division, published in SAE Surface Vehicle Emissions Standards Manual, SAE HS-2400, Society of Automotive Engineers, Warrendale, PA, pp. 72-82, 1993. Sargeant, K.A. (1994), "Transportation Conformity Final Rule: Summary of Provisions and Their Implications," Paper No. 94-MP13.02, Proceedings, 87th Annual Meeting, Air and Waste Management Association, Pittsburgh, PA. SBRC (1994). Remote Emissions Sensor RES-100 Smog Dog Userís Manual, Prepared by Hughes Santa Barbara Research Center for State of North Carolina Dept. of Environmental, Health, and Natural Resources, Raleigh, NC. WA No LG73, SBRC Reference NO. 94-0327.
144
Schoonveld, G.A., and W.F. Marshall (1991), “The Total Effect of a Reformulated Gasoline on a Vehicle Emission by Technology (1973 to 1989),” Society of Automotive Engineers, Paper No. 910380, 24 pp. Seinfeld, J.H. (1988). Fundamentals of Air Pollution Engineering. Prentice-Hall: New York. Stedman, D.H. (1989), “Automobile Carbon Monoxide Emissions,” Environmental Science and Technology, 23:147-149. Stedman, D.H., G.A. Bishop, J.E. Peterson, et al. (1991). On-Road Carbon Monoxide and Hydrocarbon Remote Sensing in the Chicago Area. Prepared by University of Denver for Illinois Department of Energy and Natural Resources, Springfield, IL, October. Stephens R. D. (1994), "Remote Sensing Data and a Potential Model of Vehicle Exhaust Emissions", Journal of the Air and Waste Management Association, 44(11) pp 1284. Stephens, R.D., and S.H. Cadle (1991), “Remote Sensing Measurements of Carbon Monoxide Emissions from On-Road Vehicles,” J. Air & Waste Management Association, 41:39-46. Stephens, R.D, P.A. Mulawa, M.T. Giles, et al. (1994). An Experimental Evaluation of Remote Sensing Based Hydrocarbon Measurements: A Comparison to FID Measurements. Prepared by General Motors, ManTech Environmental Technologies, and U.S. Environmental Protection Agency for Coordinating Research Council, Contract VE11-4. September. Sutton, W. (1995), North Carolina Department of Agriculture, Standards Division, Personal Communication with David Eichenberger. Ward, C.C. (1979), “Petroleum and Other Liquid Fuels,” in Mark’s Standard Handbook for Mechanical Engineers. Baumeister, Avallone, and Baumeister, eds. McGraw-Hill: New York. pp. 7.14-7.21.
145
Appendices are not available in the Adobe PDF version of this report at this time.
146
