Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A trip-based approach to modelling urban transportation greenhouse gas emissions for municipalities Rahman, Md. Nobinur 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2016_november_rahman_md. nobinur.pdf [ 3.76MB ]
Metadata
JSON: 24-1.0314183.json
JSON-LD: 24-1.0314183-ld.json
RDF/XML (Pretty): 24-1.0314183-rdf.xml
RDF/JSON: 24-1.0314183-rdf.json
Turtle: 24-1.0314183-turtle.txt
N-Triples: 24-1.0314183-rdf-ntriples.txt
Original Record: 24-1.0314183-source.json
Full Text
24-1.0314183-fulltext.txt
Citation
24-1.0314183.ris

Full Text

A TRIP-BASED APPROACH TO MODELLING URBAN TRANSPORTATION GREENHOUSE GAS EMISSIONS FOR MUNICIPALITIES   by  Md. Nobinur Rahman  B.Sc., Rajshahi University of Engineering & Technology, 2010  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF APPLIED SCIENCE in THE COLLEGE OF GRADUATE STUDIES (Civil Engineering)  THE UNIVERSITY OF BRITISH COLUMBIA (Okanagan)  September 2016  © Md. Nobinur Rahman, 2016 ii The undersigned certify that they have read, and recommend to the College of Graduate Studies for acceptance, a thesis entitled:   A Trip-based Approach to Modelling Urban Transportation Greenhouse Gas Emissions for Municipalities   Submitted by               Md. Nobinur Rahman            in partial fulfillment of the requirements of the degree of          Master of Applied Science            .   Dr. Ahmed O. Idris, Faculty of Applied Science/School of Engineering  Supervisor, Professor (please print name and faculty/school above the line)  Dr. Gordon Lovegrove, Faculty of Applied Science/School of Engineering Supervisory Committee Member, Professor (please print name and faculty/school above the line)  Dr. Cigdem Eskicioglu, Faculty of Applied Science/School of Engineering  Supervisory Committee Member, Professor (please print name and faculty/school above the line)  Dr. Alex Bigazzi, Faculty of Applied Science/Department of Civil Engineering University Examiner, Professor (please print name and faculty/school above the line)    September 07, 2016 (Date submitted to Grad Studies) iii Abstract Transportation has always been a major source of Greenhouse Gas (GHG) emissions all over the world. The fifth assessment report of the Intergovernmental Panel on Climate Change (IPCC) has reported that the transportation sector was responsible for 14% of the total emissions in 2010. In Canada, transportation has been the second largest source of GHG emissions in 2012. Therefore, cutting transportation-related GHG emissions has become a top priority on the international agenda. Many international governments have announced aggressive GHG emissions reduction targets. In response, numerous research efforts have looked at developing tools to model the effect of various transportation and land use policies on GHG emissions reduction. However, most of the developed models are context specific and vary greatly in complexity (e.g. network-wide, corridor-wide, etc.) and level of detail (i.e. macro, meso, and micro). Thus, they cannot be used in other settings. In addition, little has been reported on monitoring progress towards meeting municipal GHG emissions reduction targets.   To contribute to this issue, this research presents a TRIp-Based Urban Transportation Emissions (TRIBUTE) model for municipalities. TRIBUTE integrates two main components: a discrete mode choice/shift model and an emissions forecasting model. Given personal, modal, and land use information, the mode choice/shift model calculates the proportion of trips made by different travel options (e.g. car, bus, walk, etc.). The total Vehicle Kilometres Travelled (VKT) by each mode is then calculated by multiplying the proportion of trips made by each mode by respective average VKT. Finally, total GHG emissions are calculated by multiplying the total VKT by each mode by respective average emissions factors. TRIBUTE is intended to assist municipalities (especially those with no detailed transportation network model) explore the impacts of various transportation and land use planning policies on changing travel behavior, and subsequently GHG emissions from passenger transportation. The City of Kelowna, BC, Canada is selected as a case study. The model validation results show a difference of only 0.3% in GHG emissions between the model prediction and the historical data.   iv Preface Excerpts of this thesis have been presented and submitted for publication in peer reviewed conference proceedings and technical journals. In addition, a technical report, utilizing the developed approach, has been submitted and presented to the City of Kelowna, BC, Canada to investigate the impacts of urban densities on the city’s GHG emissions targets.   The author developed the models, analyzed the results, and wrote initial drafts of the manuscripts listed below. The research supervisor provided valuable guidance and helped in the development of the final versions of the publications. The work presented in the thesis has led to the following publications:  Rahman, M. N. and Idris, A. O. (2016). TRIBUTE: A TRIp-Based Urban Transportation Emissions Model for Municipalities. Submitted to the Journal of Sustainable Transportation (Date of submission: April 4th, 2016).  Rahman, M. N., Ruparathna, R., Sekhon, S., Kumar, V., AlHashmi, M., Umer, A., Andrade, M. T. B., Dyck, R., Doberstein, C., Hewage, K., Hicky, R., Idris, A., Li, E., Culver, K., and Sadiq, R. (2016). Investigating the Impacts of Urban Densities on City of Kelowna’s GHG Emission Targets. Technical report submitted to the City of Kelowna in January, 2016.  Rahman, M. N. and Idris, A. O. (2016). Modelling On-Road Transportation Greenhouse Gas (GHG) Emissions under Various Land Use Scenarios. Paper presented at the 95th Transportation Research Board (TRB) Annual Meeting, Washington D.C.  Rahman, M. N. and Idris, A. O. (2015). Trip-based Urban Transportation Greenhouse Gas (GHG) Emissions Model. Paper presented at the 2nd Engineering Graduate Symposium, School of Engineering, UBC Okanagan, Kelowna, BC, Canada. v Table of Contents  Abstract ......................................................................................................................................... iii Preface ........................................................................................................................................... iv Table of Contents ...........................................................................................................................v List of Tables ................................................................................................................................ ix List of Figures .................................................................................................................................x List of Abbreviations ................................................................................................................. xiii Acknowledgements .................................................................................................................... xiv Chapter 1: Introduction ................................................................................................................1 1.1 Outline............................................................................................................................. 1 1.2 Problem Statement and Research Motivation ................................................................. 1 1.3 Research Goal and Objectives ........................................................................................ 2 1.4 Structure of the Thesis .................................................................................................... 3 Chapter 2: Literature Review .......................................................................................................4 2.1 Outline............................................................................................................................. 4 2.2 Greenhouse Gases ........................................................................................................... 4 2.3 Sources of GHG Emissions ............................................................................................ 6 2.4 Driving Factors of Transportation GHG Emissions ....................................................... 9 2.5 Land Use and Transportation Interactions .................................................................... 16 2.5.1 Density Indicators ..................................................................................................... 16 2.5.2 Diversity Indicators ................................................................................................... 18 2.5.3 Design Indicators ...................................................................................................... 19 2.6 Travel Demand Modelling ............................................................................................ 20 vi 2.6.1 Mode Choice Modelling ........................................................................................... 22 2.6.1.1 Model Validation and Forecasting Performance .............................................. 24 2.6.2 Distance Travelled Modelling ................................................................................... 25 2.6.2.1 Simple Linear Regression ................................................................................. 25 2.6.2.2 Multiple Regression Analysis ........................................................................... 27 2.6.2.3 Violation of Assumptions ................................................................................. 29 2.7 Summary ....................................................................................................................... 31 Chapter 3: TRIBUTE: A TRIp-Based Urban Transportation Emissions Model for Municipalities ...............................................................................................................................32 3.1 Outline........................................................................................................................... 32 3.2 Methodology ................................................................................................................. 32 3.2.1 Model Building (Base Case Scenario) ...................................................................... 35 3.2.2 Linking Passenger Trips to Vehicle Trips ................................................................ 37 3.2.3 Model Application (Future Scenarios)...................................................................... 40 3.3 Summary ....................................................................................................................... 41 Chapter 4: Case Study .................................................................................................................42 4.1 Outline........................................................................................................................... 42 4.2 Study Area .................................................................................................................... 42 4.2.1 Kelowna’s Growth Projections ................................................................................. 44 4.2.2 Kelowna’s Transportation Network .......................................................................... 44 4.2.3 Kelowna’s GHG Emissions ...................................................................................... 48 4.3 Dataset Description ....................................................................................................... 49 4.3.1 Commuting Trips ...................................................................................................... 54 vii 4.3.2 Non-commuting Trips ............................................................................................... 63 4.3.3 Vehicle Occupancy ................................................................................................... 72 4.3.4 GHG Emissions Factors ............................................................................................ 73 4.3.5 Land Use Data........................................................................................................... 74 4.4 Summary ....................................................................................................................... 76 Chapter 5: Modelling Results .....................................................................................................77 5.1 Outline........................................................................................................................... 77 5.2 Generating Level-of-Service Attributes ........................................................................ 77 5.3 Relationship between Travel Distance and Travel Time .............................................. 77 5.3.1 Error Analysis ........................................................................................................... 80 5.4 Passenger Kilometre Travelled (PKT) Estimation Model ............................................ 82 5.4.1 Dependent Variables ................................................................................................. 82 5.4.2 Explanatory Variables ............................................................................................... 82 5.4.3 Empirical Analysis .................................................................................................... 83 5.4.4 Model Prediction Error ............................................................................................. 85 5.5 Base Case Estimates ..................................................................................................... 86 5.5.1 Mode Choice Modelling ........................................................................................... 86 5.5.1.1 Modes of Travel ................................................................................................ 86 5.5.1.2 Empirical Analysis ............................................................................................ 89 5.5.1.3 Forecasting Performance .................................................................................. 96 5.5.2 Estimation and Validation......................................................................................... 98 5.6 Future Scenario Development..................................................................................... 100 5.7 Scenario Estimates ...................................................................................................... 104 viii 5.7.1 Reduction in PKT ................................................................................................... 104 5.7.2 Modal Shift ............................................................................................................. 106 5.7.3 Forecasting GHG Emissions ................................................................................... 108 5.8 Summary ..................................................................................................................... 110 Chapter 6: Conclusions and Future Research ........................................................................112 6.1 Outline......................................................................................................................... 112 6.2 Conclusions ................................................................................................................. 112 6.3 Recommendations to the City of Kelowna ................................................................. 114 6.3.1 Increase Mixed Land Use Diversity ....................................................................... 114 6.3.2 Improve Public Transit ........................................................................................... 115 6.3.3 Promote Active Transportation ............................................................................... 115 6.4 Limitations and Future Research ................................................................................ 115 Bibliography ...............................................................................................................................117 Appendices ..................................................................................................................................129 Appendix A : Biogeme Software User Interface .................................................................... 129 Appendix B : Sample Model on Biogeme Software ............................................................... 130 Appendix C : Sample Output Results from Biogeme ............................................................. 131 Appendix D : Residual Plots of the Developed Regression Models ...................................... 132 ix List of Tables Table 2-1 Global Warming Potential of GHGs (IPCC, 1995; IPCC, 2007; IPCC, 2014) .............. 5 Table 4-1 Kelowna’s Projected Growth Rates (Rahman et al., 2016) .......................................... 44 Table 4-2 Descriptive Statistics of 2013 Okanagan HHTS .......................................................... 50 Table 4-3 Descriptive Statistics of Commuting Trips .................................................................. 54 Table 4-4 Descriptive Statistics of Non-Commuting Trips .......................................................... 64 Table 4-5 Average Vehicle Occupancy ........................................................................................ 73 Table 4-6 GHG Emission Factors (Ministry of Environment, 2014b) ......................................... 73 Table 4-7 Characteristics of Built Environment and Land Use Variables .................................... 75 Table 5-1 Relationships between Travel Distance and Travel Time ............................................ 80 Table 5-2 PKT Estimation Models (Square Root of PKT) ........................................................... 83 Table 5-3 Resultant Impacts of Land Use and Built Environment on PKT ................................. 85 Table 5-4 Mode Choice Models ................................................................................................... 91 Table 5-5 Resultant Impacts of Land Use and Built Environment on Mode Choice ................... 95 Table 5-6 FPM of the Developed Mode Choice Models based on Vehicle Ownership Segmentation................................................................................................................................. 97 Table 5-7 FPM of the Developed Models based on Geographic Location Segmentation ............ 98 Table 5-8 TRIBUTE Validation Results....................................................................................... 99 Table 5-9 Reduction Interval in PKT.......................................................................................... 105 Table 5-10 Modal share Interval in Different Scenarios ............................................................ 107 Table 5-11 Emissions Estimation Window in Different Scenarios ............................................ 109  x List of Figures Figure 2-1 Proportion of Different Components of GHGs (IPCC, 2014) ...................................... 6 Figure 2-2 Sources of GHG Emissions ........................................................................................... 7 Figure 2-3 Sources of GHG Emissions in Canada in 2012 ............................................................ 8 Figure 2-4 Transportation-related Emissions in Canada (Environment Canada, 2014) ................. 8 Figure 2-5 Four-Step Transportation Planning Model.................................................................. 20 Figure 2-6 Classification of Trips (Commuting vs. Non-Commuting)......................................... 21 Figure 2-7 Classification of Trips (Home-Based vs. Non-Home Based) ..................................... 22 Figure 2-8 Expected pattern for a plot of residuals versus predicted ........................................... 30 Figure 3-1 Interaction among Trip Length Reduction, Mode Shift, and VKT Reduction ........... 33 Figure 3-2 TRIp-Based Urban Transportation Emissions Model ................................................. 34 Figure 3-3 Linking Passenger Trips to Vehicle Trips ................................................................... 39 Figure 4-1 Distribution of Population Density in the City of Kelowna ........................................ 43 Figure 4-2 Road Network in the City of Kelowna ........................................................................ 45 Figure 4-3 Active Transportation Network in the City of Kelowna ............................................. 46 Figure 4-4 Public Transit Network in the City of Kelowna .......................................................... 47 Figure 4-5 Sources of GHG Emissions in the City of Kelowna (City of Kelowna, 2012) ........... 48 Figure 4-6 On-Road Emissions from Different Vehicle Types (City of Kelowna, 2009) ............ 49 Figure 4-7 Distribution of Trips Originated in Kelowna .............................................................. 52 Figure 4-8 Trips Purposes ............................................................................................................. 53 Figure 4-9 Modal Share ................................................................................................................ 53 Figure 4-10 Gender Distribution (Commuting Trips) .................................................................. 57 Figure 4-11 Age Distribution (Commuting Trips)........................................................................ 58 Figure 4-12 Household Size (Commuting Trips) ......................................................................... 58 xi Figure 4-13 Vehicle Ownership (Commuting Trips).................................................................... 59 Figure 4-14 Driver License (Commuting Trips) ........................................................................... 60 Figure 4-15 Transit Pass (Commuting Trips) ............................................................................... 60 Figure 4-16 Dwelling Types (Commuting Trips) ......................................................................... 61 Figure 4-17 Occupation Distribution (Commuting Trips) ............................................................ 62 Figure 4-18 Income Distribution (Commuting Trips) .................................................................. 62 Figure 4-19 Modal Share (Commuting Trips) .............................................................................. 63 Figure 4-20 Gender Distribution (Non-Commuting Trips) .......................................................... 66 Figure 4-21 Age Distribution (Non-Commuting Trips) ............................................................... 67 Figure 4-22 Household Size (Non-Commuting Trips) ................................................................. 68 Figure 4-23 Vehicle Ownership (Non-Commuting Trips) ........................................................... 68 Figure 4-24 Transit Pass (Non-Commuting Trips) ....................................................................... 69 Figure 4-25 Driver License (Non-Commuting Trips) .................................................................. 69 Figure 4-26 Dwelling Type Distribution (Non-Commuting Trips) .............................................. 70 Figure 4-27 Occupation Distribution (Non-Commuting Trips).................................................... 71 Figure 4-28 Income Distribution (Non-Commuting Trips) .......................................................... 71 Figure 4-29 Modal Share (Non-Commuting Trips) ...................................................................... 72 Figure 5-1 Relationships between Travel Distance and Travel Time for Car .............................. 79 Figure 5-2 Relationships between Travel Distance and Travel Time for Transit ......................... 79 Figure 5-3 Error Envelope of Travel Time and Travel Distance Estimation for Car ................... 81 Figure 5-4 Error Envelope of Travel Time and Travel Distance Estimation for Transit ............. 81 Figure 5-5 Choice Set for Mode Choice Models .......................................................................... 87 Figure 5-6 Non-Motorized Trip Distance in Commuting Trips ................................................... 88 xii Figure 5-7 Non-Motorized Trip Distance in Non-Commuting Trips ........................................... 89 Figure 5-8 Mixed Land Use Entropy in BAU and Hypothetical Scenario ................................. 102 Figure 5-9 Mixed Employment Entropy in BAU and Hypothetical Scenario ............................ 103 Figure 5-10 Estimated PKT Reduction ....................................................................................... 105 Figure 5-11 Modal Shift (Estimated) .......................................................................................... 107 Figure 5-12 Forecasted GHG Emissions in the City of Kelowna ............................................... 109 Figure 5-13 Per Capita GHG Emissions ..................................................................................... 110 Figure A-1 User Interface of Biogeme Software ........................................................................ 129 Figure C-1 Output Results from Biogeme .................................................................................. 131 Figure D-1 Residual Plot of the Travel Distance and Travel Time for Car ................................ 132 Figure D-2 Residual Plot of the Travel Distance and Travel Time for Transit .......................... 132  xiii List of Abbreviations BAU Business-As-Usual  BC British Columbia CEEI Community Energy and Emissions Inventory  CO2 eq. Carbon dioxide equivalent  FPM Forecasting Performance Measure  GDP Gross Domestic Product GHG Greenhouse Gas GWP Global Warming Potential HHTS Household Travel Survey ImPACT Impact (Im), Population (P), Affluence (A), Intensity of use (C), and Efficiency (T) IPAT Impact (I), Population (P), Affluence (A), and Technology (T) IPCC Intergovernmental Panel on Climate Change MASc Most Aggressive Scenario MNL Multinomial Logit Model PKT Passenger Kilometres Travelled  SD School District STIRPAT STochastic Impacts by Regression on Population, Affluence, and Technology STPCO Sustainable Transportation Partnership of the Central Okanagan TAZ Traffic Analysis Zone TRIBUTE TRIp-Based Urban Transportation Emissions UNFCCC United Nations Framework Convention on Climate Change USEPA United States Environmental Protection Agency VKT Vehicle Kilometres Travelled VMT Vehicle Miles Travelled  xiv Acknowledgements In the name of Allah, the most Beneficent, the most Merciful. All praises to Allah for giving me the strength and courage to complete this thesis.  This research would not have been possible without the gracious support of many people. First and foremost, I would like to express profound gratitude to my supervisor, Dr. Ahmed Idris, for his thoughtful guidance, support, encouragements, and understanding throughout the progression of my graduate studies. I can proudly say that he has taught me everything from scratch.  I would also like to extend my enduring gratitude to the members of my examination committee: Dr. Gordon Lovegrove and Dr. Cigdem Eskicioglu for their insightful comments on my thesis. I am also thankful to Dr. Alex Bigazzi, my university examiner, for spending his valuable time reading my thesis, providing constructive feedback, and attending my final thesis defence. My special thanks to Lori Walter for her invaluable feedback, comments, and advice to improve my writing skills, and reviewing my thesis.  I am grateful to my amazing colleagues of my research group, Kristoffer Mattisson, Hassan Masood, Eric Ma, and John Rozema. It was a great pleasure to work with such a wonderful research group. I would like to thank all of my friends, especially Mohamed Hegazi and Adam Lee. Also my sincere thanks to my great friend and colleague, Abdul Rahman Masoud, for helping and supporting me from the very first day of my graduate studies.  I would like to thank all the Bangladeshis in Kelowna for their efforts and selfless care, which alleviated my loneliness and helped me to concentrate on my research.   I would like to express my heartfelt gratitude to my parents, my parents-in-law, my sister (Nowrin Akter) and sister-in-law (Tasnima Haque), and other members of my family for their continuous inspiration, sacrifice, and support to complete the research work successfully. I am especially grateful to my beloved wife Tasnuva Haque. Words cannot express her contribution in my life. Acknowledgments would be incomplete if I don’t mention my little angel Nuayma. You are the best gift during this research work. xv Dedicated to my beloved wife Tasnuva Haque1 Chapter 1: Introduction 1.1 Outline This chapter starts with a discussion of the problem statement and research motivation in Section 1.2, followed by research goal and objectives in Section 1.3. Finally, Section 1.4 demonstrates the structure of the thesis.  1.2 Problem Statement and Research Motivation Transportation has always been a major source of Greenhouse Gas (GHG) emissions all over the world. From 1990 to 2007, GHG emissions from the transportation sector increased substantially, reaching around 15% of the overall GHG emissions in 2010 (International Transportation Forum, 2010). This global trend is further exceeded in North America. In 2008, transportation accounted for 27% of the total GHG emissions in the United States (Mahendra et al., 2012). In Canada, transportation was the second largest source of GHG emissions in 2012, accounting for 24% of the total emissions nationwide (Environment Canada, 2014). At the provincial level, 37.9% of British Columbia’s GHG emissions came from the transportation sector (Ministry of Environment, 2012). Such high levels of GHG emissions can be attributed to the large expansion of the urban road network and automobile dependency in many North American cities. For example, GHG emissions from private automobiles in Canada in 2007 were 14% higher than that of 1990 levels (Terefe, 2010).   In light of the above, cutting transportation-related GHG emissions has become a top priority on the international agenda. Many international governments announced aggressive, sometimes optimistic, GHG emissions reduction targets. For example, the United Kingdom government set a target to reduce transportation-related GHG emissions by 60% by 2030 (Hickman and Banister, 2007). The Australian government intended to reduce GHG emissions by 20% below 2000 levels by 2020, and ultimately reach a 50% reduction by 2050 (Stanley et al., 2011). The Government of Canada has also committed to reduce Canada’s GHG emissions by 17% below 2005 levels by 2020 under the Copenhagen Accord (Environment Canada, 2013). At the provincial level, British Columbia went further beyond the Canadian threshold and intended to reduce its emissions below 2007 levels by a minimum of 33% by 2020 and 80% by 2050 (Ministry of Environment, 2014c). 2 In response to these international governments’ commitments, GHG emissions have become a major concern within the international scientific community. Numerous research efforts have looked at developing tools to model and forecast GHG emissions from the transportation sector. While useful, the developed tools are context specific and constrained by the availability of data and/or expertise, being of little or no use outside that context. In addition, the available tools for GHG emissions modelling and forecasting vary greatly in complexity (e.g. network-wide, corridor-wide, etc.), level of details (i.e. macro, meso, and micro), and sensitivity to transportation attributes (e.g. travel speed, fuel consumption, etc.) and land use factors (e.g. density, diversity, design, etc.).  It is apparent that there is no universal tool for GHG emissions modelling and forecasting. Unfortunately, most of the developed tools are suitable for estimating GHG emissions in large urban centres where traffic congestion and emissions are major concerns, being of little use in small municipalities where the required data and/or expertise are unavailable. Other tools require a detailed transportation network model to estimate emissions at the link level (i.e. link-based models), being useless in cases where such network model is unavailable. Therefore, more research is still needed to develop GHG emissions modelling and forecasting tools suitable for the context of small and mid-sized municipalities with appropriate complexity, level of details, and sensitivity to transportation attributes and land use factors.   1.3 Research Goal and Objectives This research aims at developing modelling and scenario-comparison tools to explore the impacts of various transportation and land use planning policies on changing travel behavior, and subsequently GHG emissions. The developed tools are intended for estimating GHG emissions from passenger transportation in the absence of a detailed transportation network model. Such goal is achieved through the completion of the following objectives:   Develop discrete mode choice model to estimate mode shift (i.e. switching modes from cars to public transit and active transportation) in response to changes in density and land use mix  3  Develop Passenger Kilometre Travelled (PKT) model to capture the reduction in trip length due to changes in land use attributes   Develop GHG emissions forecasting model to assist municipalities evaluate alternative policy scenarios and eventually select the scenario(s) that help them to meet their future GHG emissions targets   Assess the impacts of urban densities and land use policies on the City of Kelowna’s GHG emissions targets as a case study  1.4 Structure of the Thesis This thesis consists of six chapters. Chapter 1 introduces the problem statement along with the research goal and objectives, and thesis outline. Chapter 2 reviews the literature on Greenhouse Gas (GHG) emissions and their sources, the controlling factors that affect GHG emissions, and the effects of land use and built environment on travel behavior and GHG emissions reduction. Chapter 3  proposes a TRIp-Based Urban Transportation Emissions (TRIBUTE) model and describes the research methodology and the framework of the developed model. Chapter 4 presents the study area, datasets, and modelling efforts. Chapter 5 applies TRIBUTE to evaluate the impact of alternative land use and transportation policy scenarios on Kelowna’s GHG emissions targets. Finally, Chapter 6 concludes the thesis and discusses limitations and future research.  4 Chapter 2: Literature Review 2.1 Outline This chapter starts with an overview of Greenhouse Gases (GHGs) in Section 2.2, followed by the sources of GHG emissions in Section 2.3. The factors that affect GHG emissions are presented in Section 2.4. In Section 2.5, the effects of land use and built environment on travel behavior and GHG emissions reduction are discussed. The traditional travel demand modelling approach is presented in Section 2.6.   2.2 Greenhouse Gases Greenhouse gases (GHG) generally act as a global insulator that absorbs surface energy and slows down energy escapes, thus warms the earth’s surface and makes it hospitable to life (IPCC, 1996). This phenomenon, warming the earth’s surface, is known as the greenhouse effect (Murphy, 2000). The most important natural GHGs associated with the greenhouse effect are water vapor (H2O), carbon dioxide (CO2), nitrous oxide (N2O), methane (CH4), and ozone (O3). However, the increase in concentration of GHGs in the earth’s atmosphere has resulted in higher absorption rates of surface energy and subsequently, global temperature rising, leading to a rise in mean sea level (Murphy, 2000).  Human activities such as burning fossil fuels and deforestation can contribute to the increase of GHGs in the earth’s atmosphere. Besides the natural GHGs, the Kyoto protocol has dealt with the greenhouse gases sulphur hexafluoride (SF6), hydrofluorocarbons (HFCs) and perfluorocarbons (PFCs) (UNFCCC, 2008). The components of GHGs can be divided into two broad categories: long-lived gases and short-lived gases (IPCC, 2007). The long-lived gases persist a long time in the atmosphere and are chemically more stable. Carbon dioxide, methane, and nitrogen dioxide lies within this category. On the other hand, the short-lived gases are chemically more active and are washed out in the form of precipitation. Short-lived gases include sulphur dioxide and carbon monoxide (IPCC, 2007).   Greenhouse gas emissions for each gas are normally stated and/or measured in terms of their global warming potential (GWP), a measure that allows their impact to be stated in terms of one comparable gas (USEPA, 2016). GWP is an indicator of energy absorption of a particular gas 5 over a given period relative to carbon dioxide over the same period. Usually, the given time is 100 years. The unit of GWP is carbon dioxide equivalent (CO2 eq.), since all GHGs are measured in terms of carbon dioxide. The GWP of major GHGs is given in Table 2-1.   Table 2-1 Global Warming Potential of GHGs (IPCC, 1995; IPCC, 2007; IPCC, 2014) GHG 100-Year GWP1 100-Year GWP2 100-Year GWP3 Carbon Dioxide (CO2) 1 1 1 Methane  (CH4) 21 25 28 Nitrous Oxide  (N2O) 310 298 265 Nitrogen Trifluoride (NF3) - 17,200 16,100 Sulphur Hexafluoride (SF6) 23,900 22,800 23,500 Hydrofluorocarbon - 23 (CHF3) 11,700 14,800 12,400 Hydrofluorocarbon - 32 (CH2F2) 650 675 677 Perfluorocarbons – 116 (C2F6) 9,200 12,200 11,100  As shown in Table 2-1, the GWP is not the same in all annual reports. The GWP is influenced by many factors, such as the new lifetime of GHGs, change in CO2 impacts, the strength of indirect effects on concentration of GHGs, and change of GHGs over time (Trottier, 2015). Based on the 100-year global warming potential values from the Fifth Assessment Report (5AR), 62% of global emissions come from CO2 fossil fuel and industrial processes (IPCC, 2014). Figure 2-1 illustrates the components of GHGs based on the Fifth Assessment Report (5AR).                                                  1 IPCC. (1996). Climate change 1995: Economic and social dimensions of climate change. Cambridge University Press. 2 IPCC. (2007). Climate change 2007: The physical science basis. Intergovernmental Panel on Climate Change.  3 IPCC. (2014). Climate change 2014: Synthesis report. Contribution of working groups I, II and III to the fifth assessment report of the Intergovernmental Panel on Climate Change.  6  Figure 2-1 Proportion of Different Components of GHGs (IPCC, 2014)  2.3 Sources of GHG Emissions According to the IPCC (2014), the sources of GHG emissions can be broadly categorized as direct emissions and indirect emissions from different economic sectors. Direct emissions come from five economic sectors: AFOLU1, buildings, transport, industry, and other energy sources. Indirect emissions come from electricity and heat production. Indirect emissions are further classified as follows: energy, industry, transport, buildings, and AFOLU (IPCC, 2014). Figure 2-2 illustrates the sources of GHG emissions from different economic sectors in 2010.                                                   1 Agriculture, forestry, and other land use (AFOLU) 7  Figure 2-2 Sources of GHG Emissions1   In Canada, the transportation sector was responsible for 24% of total GHG emissions in 2012, coming second after oil and gas, as shown in Figure 2-3 (Environment Canada, 2014).  Within the transportation sector, road transportation came first with 65% of all transportation-related emissions, as shown in Figure 2-4 (Environment Canada, 2014).  According to the Copenhagen Accord, Canada has committed to reduce its GHG emissions by 17% below the 2005 levels by 2020 (Environment Canada, 2013). At the provincial level, British Columbia went further than the Canadian threshold and intended to reduce its emissions below 2007 levels by a minimum of 33% by 2020 and 80% by 2050 (Ministry of Environment, 2014c). In British Columbia, 37.9% of GHG emissions come from the transportation sector (Ministry of Environment, 2012).                                                    1 Modified from Figure 1.7, page 47 from IPCC, 2014: Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, Pachauri, R.K. and Meyer, L. (eds.)]. IPCC, Geneva, Switzerland 8  Figure 2-3 Sources of GHG Emissions in Canada in 20121    Figure 2-4 Transportation-related Emissions in Canada (Environment Canada, 2014)                                                   1 © Her Majesty The Queen in Right of Canada, Environment and Climate Change Canada, 2014, by permission 9 2.4 Driving Factors of Transportation GHG Emissions Understanding the factors influencing global GHG emissions is key to developing appropriate mitigation strategies. Many previous studies have reported on the environmental impact of global GHG emissions and their intensity (Bongaarts, 1992; Bristow et al., 2008; York et al., 2003; Davis and Caldeira, 2010; Jung et al., 2012; Kawase et al., 2006; Kwon, 2005; Myers, 1993; Raskin, 1995; Rosa and Dietz, 1998; Xiangzhao and Ji, 2008). According to the framework for sustainability, introduced by Ehrlich and Holdren (1971), the environmental impact (I) consists of population (P), affluence (A), and technology (T), which is known as IPAT. IPAT establishes a relationship between human activity and its impact on the environment, but it can be used on particular environmental issues (Kwon, 2005). IPAT measures population, affluence, and technology in terms of per capita, per capita consumption, and per unit of consumption, respectively. The first algebraic formulation and the application of IPAT to data analysis was introduced by Commoner et al. (1971):   Impact (I) = Population(P) × Affluence (A) × Technology (T), (2-1)  However, IPAT is unable to use intensity in calculating the impact of environmental issues such as GHG emissions. To resolve this issue, IPAT was further modified by Waggoner and Ausubel (2002) by disaggregating technology (T) into intensity of use (C) and efficiency (T). The modified equation of IPAT is known as ImPACT:   Impact (I) = Population(P) × Affluence (A) × Intensity (C) × Efficiency (T), (2-2)  In ImPACT, if the impact (I) is considered as emissions then the other components can be written as population (P), GDP per capita (A), and emissions per GDP (C × T). Many researchers have used IPAT and ImPACT to assess the environmental impact based on different driving factors (Bongaarts, 1992; Holdren, 1991; Myers, 1993; Raskin, 1995; Smil, 1990). The strength of both IPAT and ImPACT is in capturing the factors that have direct impact on the environment.    10 In addition, Raskin (1995) used quantitative analysis techniques to study the effect of population on environmental changes. In his study, another iteration of IPAT equation was used to find out the explicit role of population in environmental changes. The study concluded that change in population growth is a significant contributing factor to GHG emissions. This finding was further supported by York et al. (2003) who proposed a new stochastic model based on IPAT to study the driving forces of GHG emissions. The new model is known as stochastic impacts by regression on population, affluence, and technology (STIRPAT). The main strength of STIRPAT over IPAT and ImPACT is its ability to test hypotheses without relying on proportionality assumptions. STIRPAT has been used by several researchers in analyzing the effects of driving forces on environmental impacts (Rosa and Dietz, 1998; York et al., 2003). According to York et al. (2003), population has a proportional relationship to emissions, and affluence increases the emissions monotonically. In addition, the authors claimed that urbanization and industrialization have greater impacts on CO2 emissions.  Further, the application of the IPAT with energy and emissions has led to the Kaya framework (Ramanathan, 2006). According to the Kaya framework, global GHG emissions are influenced by the following four driving factors: global population, gross world product, global energy consumption, and carbon intensity of energy (Kaya, 1990), mathematically expressed as follows (Girod et al., 2009):   CO2 emissions = Population × GDPPopulation× Energy useGDP× CO2 emissionsEnergy use,  (2-3)  Many researchers relies upon the Kaya equation to estimate global as well as national emissions (Albrecht et al., 2002; Kawase et al., 2006; Mahony, 2013; Ramanathan, 2006; Raupach et al., 2007; Timilsina and Shrestha, 2009; Xiangzhao and Ji, 2008). Raupach et al. (2007) used the Kaya equation to estimate GHG emissions from fossil fuel combustion and industrial processes. The authors claimed that rapidly developing economics have the highest emissions growth rates. This claim was further supported by Xiangzhao and Ji (2008), who found that economic development is the major driving factor of CO2 emissions along with an increase in population. In their study, the authors used a revised version of the Kaya framework to analyze GHG 11 emissions. Improvement in energy efficiency can reduce global CO2 emissions (Xiangzhao and Ji, 2008). In addition, Kawase et al. (2006) studied long-term emissions scenarios in France, Germany, and the UK, as well as medium-term scenarios in Japan by analyzing emissions using an extended form of the Kaya equation to support Japan’s long-term emissions scenario development. In their study, the authors claimed that energy intensity and carbon intensity have an enormous effect on reaching emissions reduction targets. In addition, affluence and population growth play important role in global emissions increase. According to Mahony (2013), the extended Kaya equation suggests that energy intensity and fossil fuel also play a role in increasing global emissions. By introducing renewable energy, energy intensity can be significantly reduced and subsequently, emissions can be reduced (Mahony, 2013).   Transportation systems are the lifelines of nations’ economies and the preconditions for progress and development. Starting from this interdependent relationship between transportation and economy, Lu et al. (2007) and Wang et al. (2015) studied the driving factors of transportation emissions. The authors claimed that rapid economic growth and development are the most important factors for increasing transportation-related GHG emissions. Further, Timilsina and Shrestha (2009) considered several factors that contribute to transportation-related emissions, such as fuel mix, modal shift, per capita gross domestic product (GDP), population, emissions coefficient, and transportation energy intensity. Based on their study, the per capita GDP, population, and transportation energy intensity are the dominant factors in increasing emission growth rates in the transportation sector.   Furthermore, Mohareb and Kennedy (2012) developed the Pathways to Urban Reductions in Greenhouse Gas Emissions (PURGE) model to estimate emissions considering four sectors, namely, electricity, building, transportation, and waste by utilizing the Kaya framework. According to this PURGE, transportation emissions are a function of fuel consumption and emissions intensity. In a more recent study, Wang et al. (2015)  showed that vehicle stock, fuel consumption, and distance travelled by each vehicle are also considered of the factors controlling GHG emissions from the transportation sector.   12 Unlike global GHG emissions, which are mainly dependent on Gross Domestic Product (GDP), transportation GHG emissions are dependent on passengers’ activities across space and time. Over the years, researchers have used passengers’ activities as determinants of GHG emissions from the transportation sector. Yang et al. (2009) examined the GHG emissions target in California from the transportation sector. The authors used an updated form of the Kaya equation to calculate GHG emissions. The updated Kaya equation directly included transportation intensity into the equation instead of GDP. Accordingly, GHG emissions from the transportation sector were estimated by multiplying transportation intensity by the total population.  Scholl et al. (1996) studied passenger transportation emissions in nine different countries from 1973 to 1992. In their study, the authors illustrated the impact of changes in activity, modal composition, and energy intensity on increase in emissions from passenger transportation. Further, the authors concluded that the increase in activity, in terms of passenger kilometres travelled, has a great effect on passenger transportation emissions. In addition, Schipper et al. (2011) analyzed the long-term emissions from the transportation sector in the USA by considering population growth, modal shift, fuel intensity, emissions intensity, and economic growth as the driving factors in increasing emissions. The results suggested that due to changes in transportation activity, emissions have tripled since 1960. The results also showed that transportation activity was a function of economic activity and car ownership.   Lakshmanan and Han (1997) analyzed the factors that affect passenger transportation emissions in the USA. According to their study, the most affecting factors were people’s tendency to travel, population growth, and GDP. The results showed that people’s tendency to travel can affect more than 50% of total emissions from the passenger transportation sector if other factor remains fixed (Lakshmanan and Han, 1997). Kwon (2005) studied factors affecting car emissions from 1970 to 2000 in Great Britain. The author used the IPAT equation to break down the driving factors associated with car emissions. The results showed that car-driving distance per person (affluence) is the dominant factor in increasing car emissions.   However, mode shift from single occupancy vehicles to high occupancy vehicles (e.g. carpool, public transit, etc.) and non-motorized transportation (e.g. cycle, walk, etc.) results in a reduction 13 in GHG emissions. Schipper and Marie-Lilliu (1999) proposed a different framework to break down emissions from transportation. According to their framework, GHG emissions can be attributed to the total distance travelled by passenger (A), modal share among different transportation modes (S), energy intensity of each transportation mode (I), and fuel used by different modes (F). Energy intensity (I) in turn is a function of three different criteria: efficiency, vehicle characteristics, and inverse of capacity utilization for each mode of transportation. Their framework can be mathematically written as follows (Tiwari et al., 2011):   Emissions = A × S × I × F,             (2-4)     On the other hand, ASIF has evolved into a new descriptive framework that elaborately describes the components related to transportation emissions and emissions reduction (Tiwari et al., 2011). The new framework encompasses four descriptive components of CO2 emissions: Avoid, Shift, Improve, and Finance, and the new framework is called the ASIF2 paradigm. According to Tiwari et al. (2011), ‘avoid’ refers to avoiding emissions by introducing better urban and transportation planning, ‘shift’ refers to the modal shift to public and non-motorized transportation to reduce emissions, ‘improve’ refers to operation and efficiency of urban transportation, and finance refers to investment in sustainable transportation to reduce emissions from the transportation sector. The descriptive ASIF2 framework refers to the changes in urban form including the modal shift to public transportation and non-motorized modes.   Furthermore, Yang et al. (2015) studied CO2 emissions from the transportation sector in China. In their study, the authors showed that socio-economic factors, urban form, and transportation development are the three major driving forces of GHG emissions from the transportation sector. The socio-economic development and subsequently, higher residential income causes an increasing trend in total as well as per capita emissions from the transportation sector. Poumanyvong et al. (2012) has categorized different countries into three different types based on their income level to illustrate the impacts of urbanization on on-road emissions. In their analysis, they used the STIRPAT equation to develop a relationship between urbanization and energy use in road transportation. Their study found that the change in urbanization has a greater impact on transportation energy consumption and subsequently, GHG emissions. In addition, 14 Papagiannaki and Diakoulaki (2009) studied emissions from passenger cars in Greece and Denmark form 1990 to 2005. They decomposed CO2 emissions into five categories: vehicle ownership, vehicle kilometres travelled (annual mileage), fuel mix, engine capacity of the passenger cars, and technology of the available cars. The results concluded that for both countries, vehicle ownership is the most influential factor in terms of increasing emissions from passenger cars. Loo and Li (2012) studied emissions from passenger transportation in China that have been measured since 1949. According to their findings, income growth is the dominant factor in increasing emissions from passenger transportation in China, followed by transportation intensity and modal shift.  As summarized above, numerous tools have been developed by researchers to analyze the factors that influence GHG emissions from the transportation sector. Most of the previous studies have looked at global and/or national GHG emissions. However, it is essential to estimate emissions from the transportation sector at the municipal level to suggest planning policies to reduce municipal GHG emissions. To assess the effects of sustainable transportation and land use policies on municipal GHG emissions, it is necessary to integrate those policies into municipal-level GHG emissions models.   Research has shown that various policies that promote public transit and active transportation have substantial effects on urban transportation GHG emissions reduction (Derrible et al., 2010; Ewing and Cervero, 2001; Mahendra et al., 2012; Rodier, 2009; Sugar and Kennedy, 2013; Tiwari et al., 2011; Walsh et al., 2008; Washbrook et al., 2006; Zahabi et al., 2012). Studies have also shown that increasing population density and balancing land use diversity (i.e. mix of activities such as residential and commercial uses) can reduce trip rates and promote non-motorized travel (Cervero and Kockelman, 1997). Further, compact density in urban core areas leads to reduction in the distances between trip origins and destinations, which in turn results in a reduction in GHG emissions (Ewing and Cervero, 2010). Conversely, low residential density increases the distance driven by vehicles, fuel consumption, and emissions (Brownstone and Golob, 2009).   15 In a more recent study, Derrible et al. (2010) developed a conceptual framework for a model called MUNTAG to help municipalities estimate their GHG emissions. The developed framework was used by Sugar and Kennedy (2013) to calculate GHG emissions from private automobiles, buses, streetcar, light rail transit, and subway. In their model, passenger kilometres travelled were converted into vehicle kilometre travelled (VKT) and then multiplied by GHG emissions factors to get the total GHG emissions for each mode of travel. However, the developed model comprised several empirical equations in capturing the effect of land use on active transportation and travel distance. While insightful, the developed model is not able to capture the impacts on GHG emissions of single land use changes (density or diversity) in the calculation of GHG emissions from the transportation sector.  On a small scale study, Mathez et al. (2013) studied GHG emissions for the trips generated by commuters to McGill University. In their study, the authors used an EMME/3 traffic assignment model to measure the travel distance from an origin-destination (OD) matrix and used emissions factors from the U.S. Environmental Protection Agency’s (EPA’s) Motor Vehicle Emissions Simulator (MOVES) model to estimate emissions. In addition, by using the Trip Reduction Impacts for Mobility Management Strategies (TRIMMS) planning tool, developed at the University of Florida, and the MOVES model, Mahendra et al. (2012) calculated GHG emissions under various transportation control measures (TCM). However, working with computer related software like MOVES, EMME/3, and/or MOBILE6 requires an explicit amount of information about road networks and vehicle types as well as a substantial amount of data on travel behavior (Derrible et al., 2010).  Household travel surveys (HHTS) are one of the most reliable sources of transportation data (Travel Forecasting Resource, 2016). From such surveys, emissions forecasting models can be developed using various parameters including socio-economic and demographic characteristics of passengers. Ko et al. (2011) used household travel survey data to calculate GHG emissions in the Seoul Metropolitan Area. The household travel survey consisted of socio-demographic information, household data, and transportation service attributes. The authors found that GHG emissions from the transportation sector are dependent on various socioeconomic characteristics 16 of passengers. However, a major drawback of this study was that land use policies were not taken into consideration.  2.5 Land Use and Transportation Interactions Various transportation and land use policies, such as parking management, promoting public transit and active transportation, and increasing land use density and diversity, have shown substantial effect on reducing on-road transportation GHG emissions (Derrible et al., 2010; Ewing and Cervero, 2010; Mahendra et al., 2012; Rodier, 2009; Sugar and Kennedy, 2013; Tiwari et al., 2011; Walsh et al., 2008; Washbrook et al., 2006; Zahabi et al., 2012; Zhang, 2004). Bristow et al. (2008) showed that changes in travel behaviour are more important in reducing emissions than technological development. Cervero and Kockelman (1997) studied the effect of land use and built environment on travel demand in terms of trip rates and mode choices in the San Francisco Bay area. The authors categorized the built environment and land use parameters into three influential categories: density, diversity, and design, also known as the DDD or 3Ds model. In addition, accessibility to transit and distance to destination were also considered among the land use parameters (Cervero and Murakami, 2008). The following subsections discuss the impact of the 3Ds on travel behaviour as well as reducing transportation emissions.  2.5.1 Density Indicators Density can be measured in terms of per unit area of population, employment, dwelling units, etc. The density of the overall activity per unit area can also be measured by adding population and employment (Ewing and Cervero, 2010). Numerous studies have shown that higher densities can reduce trip lengths and promote non-motorized travel (Brownstone and Golob, 2009; Cervero and Kockelman, 1997; Chatman, 2003; Ewing and Cervero, 2010). According to Cervero and Kockelman (1997), high population density reduces trip rates and promotes non-motorized travel. Further, compact density in urban core areas leads to a reduction in the distances between trip origins and destinations, which in turn results in a reduction in GHG emissions (Ewing and Cervero, 2010). Conversely, low residential density increases the distance driven by vehicles, fuel consumption, and emissions (Brownstone and Golob, 2009). Residential density is a function of vehicle ownership, income, and accessibility (Badoe and Miller, 2000). 17 Holtzclaw et al. (2002) studied the relationship between car ownership and distance travelled by vehicles in response to urban form in the Chicago, Los Angeles, and San Francisco areas. The results concluded that a 33% ~ 40% decrease in car ownership and a 32% ~ 43% decrease in vehicle kilometres travelled can be achieved by doubling residential density.   Furthermore, employment density showed a strong correlation to a modal shift towards active transportation and public transit (Zhang, 2004). Zhang (2004) studied the effects of various land use parameters on travel mode choice in Boston and Hong Kong. In his study, he claims that with an increase in employment density, the active transportation and public transit modal share will increase, further decreasing private vehicle use. In addition, he concludes that a 100% increase in job density will result in an increase of 9% and 0.4% in transit share for work trips and non-work trips respectively.  In addition, Zhang et al. (2012) studied the effects of built environment on distance travelled by vehicles in Seattle, Virginia, Washington DC, and Baltimore. The authors suggested that the population density and employment density are highly correlated with the Vehicle Mile Travelled (VMT) per vehicle. For instance, a 123% increase in residential density in Washington DC resulted in a 19.91% reduction in VMT. Also, a 110% increase in residential density resulted in a 15.80% VMT reduction in Baltimore. Employment density is also responsible for VMT reductions, but to a lesser extent. Population density is much more important in reducing VMT than employment density (Zhang et al., 2012).  Higher population density and employment density have also been associated with higher non-work walk trips (Greenwald and Boarnet, 2001). According to Greenwald and Boarnet (2001), both population and employment densities show a positive influences on walk trips. For example, a 1% increase in population density was found to result in a 0.34% increase in non-work walk trips. The impact of population density on increased walk trips is much higher than employment density.  18 2.5.2 Diversity Indicators Land use diversity has a strong relationship with the reduction of distance travelled by each vehicle and modal shift towards public transit and active transportation (Cervero and Kockelman, 1997; Chatman, 2003; Ewing and Cervero, 2010; Tiwari et al., 2011; Zhang, 2004). The diversity of land use can be measured in terms of entropy that measures the degree to which different land uses are evenly distributed (Lavoie, 2012). The entropy value ranges from 0 to 1, where the perfect mix use of land is indicated by 1 in a given area. Ewing and Cervero (2010) quantify the effects of land use mix (entropy index) to the walking mode and transit use. According to their study, 1% increase in mixed land use parameters could result in a 0.15% increase in walking trips, and a 0.12% increase in transit use. Moreover, their research concluded that the influence of job-housing balance could be even greater, with an increase of 0.19% in walking trips predicted associated with a 1% increase in mixed land use (Ewing and Cervero, 2010). This finding has been further supported by Zhang (2004) who showed that land use balance at both trip ends is strongly correlated with an increase in transit share as well as a decrease in car modal share.  In addition, mean entropy of mixed land use is highly correlated with the reduction in vehicle miles travelled (VMT) per household (Kockelman, 1996). Kockelman (1996) studied the effect of land use on travel behaviour in the San Francisco Bay area. The author concluded that mean entropy is highly correlated with VMT. Further, mean entropy at both origin and destination has almost the same positive effects on the choice of active transportation, and the parameter values are higher than the other variables. According to Zhang et al. (2012), mixed land use measured in terms of entropy can reduce 7.18%, 0.15%, 3.63%, and 9.29% VMT in response to 44%, 27%, 45%, and 0.54% increase in entropy in Seattle, Virginia, Baltimore, and Washington DC, respectively.  Further, Ewing and Cervero (2010) conducted a meta-analysis of the built environment and land use variables based on the literature until 2009. Based on the analysis, the authors found that the elasticity of mixed land use (measure in terms of entropy index) over VMT is negative 0.09, which means that a 1% increase in mixed land use entropy can reduce VMT by 0.09%. This finding was more pronounced in walk and transit trips that were associated with elasticities of 19 0.15 and 0.12, respectively. These elasticities imply that a 100% increase in mixed land use entropy can increase walk and transit trips by 15% and 12%, respectively. Since elasticity is less than one, therefore the observed elasticity is considered inelastic.    2.5.3 Design Indicators Design indicators include the built environment and road network characteristics in a given area. Accessibility to bus stops, availability of active transportation facilities, and intersection density are examples of design indicators. Several studies have shown that design indicators are highly correlated with reduced vehicle kilometres travelled and increased transit and walk/bike mode choice (Cervero and Kockelman, 1997; Ewing and Cervero, 2010; Kockelman, 1996). Among all density indicators, the proportion of 4-way intersections – linked to redundancy of walking routes - has a high elasticity with walk trips (Cervero and Kockelman, 1997; Ewing and Cervero, 2010; Zhang, 2004). However, more recent studies have cautioned that it is not road intersection density but walking (and bicycling) route intersection density that is the key factor and that promotes a safer, more sustainable, built environment for all users (Masoud et al., 2015; Sun and Lovegrove, 2013).  Distance to transit stops can play an important role in reducing vehicle kilometre travelled and in increasing walk/bike and transit modal shares (Bento et al., 2005; Ewing and Cervero, 2010; Naess, 2006). Ewing and Cervero (2010) showed that distance to transit stops is a dominant factor in increasing transit share as well as walk trips. The authors claimed that the elasticity1 between distance to the transit stop and transit ridership is 0.29, which means that a 1% decrease in distance to the transit stop could increase transit ridership by 0.29%. On the other hand, better accessibility to transit could increase walking trips by 0.15%. Design indicators demonstrate an inelastic correspondence to the transit share, walk trips, and VMT.                                                   1 Elasticity estimation formulas (Ewing and Cervero, 2010):  For linear regression: Elaticity = β ×x̅y̅ For logistic regression: Elaticity = β × x̅ [1 − (y̅n)], β is the regression coefficient, y̅ is the mean value of the travel parameter, and x̅ is the mean value of the built environment variable. (y̅n) is the mean estimated probability. 20 2.6 Travel Demand Modelling Travel demand modelling is at the heart of the urban transportation planning process. For many years, the four-step model has been considered as the building block of travel demand modelling. The four-step model consists of the following four steps: trip generation, trip distribution, mode choice, and trip assignment, as shown in Figure 2-5. Trip generation is concerned with estimating the total number of trips produced in and attracted to a specific zone. Trip distribution deals with the distribution of trips between zones. Mode choice estimates the proportion of trips choosing each mode of travel. Finally, trip assignment focuses on the circulation of trips on the transportation network (McNally, 2007).   Figure 2-5 Four-Step Transportation Planning Model1   The main unit of analysis in traditional demand models is the individual trip. Trips are usually classified according to their purposes. Trip purposes refer to the type of activity taking place at trip destination. In general, trips can be classified, according to trip purpose, into commuting                                                  1 Republished with permission of Emerald Group Publishing Limited from The four step model, McNally, M. G., , Handbook of transport modelling Volume 1, Hensher, D. A. & Button, K. J. (Eds.), (pp. 35-53); permission conveyed through Copyright Clearance Center, Inc. Trip Distribution Trip Assignment Mode  Choice F Flows Equilibration Feedback Trip Generation T Transport System A Activity System 21 trips (e.g. work trips, school trips, etc.) and non-commuting or discretionary trips (e.g. shopping, entertainment, etc.), as shown in Figure 2-6. Commuting trips are defined as trips related to work activities (i.e. trips to work or school) (Day et al., 2010; Habib et al., 2009). Therefore, going to work/school from home, and the reverse trips (i.e. coming back to home from work/school), are considered as commuting trips (Habib et al., 2009). In addition, going to work after dropping off children at school is also considered as commuting trips (Acker and Witlox, 2011). However, the first trip of the previous example (i.e. going from home to school) is considered as a non-commuting trip (Acker and Witlox, 2011). On the other hand, non-commuting trips are related to non-work activities such as discretionary trips from home and the reverse trips. Trips can also be categorized into home-based trips and non-home based trips. Home-based trips are defined as trips that start or end at home (National Cooperative Highway Research Program, 2012). Non-home based trips are trips that neither start nor end at home, as shown in Figure 2-7.    Figure 2-6 Classification of Trips (Commuting vs. Non-Commuting)  NCT CT Home Work or school NCT Home Shopping, park, movies, etc. NCT Home Movies NCT NCT CT Drop off someone Work or school Shopping CT: Commuting Trip NCT: Non-Commuting Trip CT NCT 22  Figure 2-7 Classification of Trips (Home-Based vs. Non-Home Based)  2.6.1 Mode Choice Modelling Mode choice modelling is the third step in the traditional four-step model. Mode choice can be conceptualized, according to the Random Utility Maximization (RUM) Theory, such that passengers assign weights to the different attributes characterizing each mode of travel and finally select the travel option that maximizes their utilities (Banister, 1978; Ben-Akiva and Boccara, 1995; McFadden, 1974). Under the RUM Theory, random refers to the modeller’s lack of knowledge (i.e. individual travellers know their choices but the analyst does not). If analysts thoroughly understood all aspects of the internal decision-making process of individual travellers as well as their perception of alternatives, they would be able to describe that process and predict mode choice using deterministic utility models. Further, utility is a measure of individual travellers’ satisfaction. Being rational, individual travellers seek to maximize their utilities by choosing certain modes of travel. According to the RUM framework, utilities are comprised of two parts: a deterministic component that considers observed characteristics of the decision-makers as well as the alternative modes (e.g. socioeconomic and demographic attributes, level of service attributes, etc.) and a stochastic error term (Ben-Akiva and Lerman, 1985; Idris, 2013; HBT HBT Home Work or school HBT Home HBT Home Movies NHBT NHBT Work or school Shopping HBT: Home Based Trip NHBT: Non-Home Based Trip HBT HBT Shopping, park, movies, etc. 23 Koppelman and Bhat, 2006). The utility function of a particular mode of travel can be expressed mathematically as follows (Koppelman and Bhat, 2006):   Uit  =  Vit  + εit , (2-5)  where, Uit is the utility of alternative i to individual t, Vit is the observable component of utility, and εit is the random error term.   The random error term is included in the utility function to account for the fact that the analyst is unable to completely and correctly measure or specify all attributes that determine travelers’ mode utility assessment. This includes imperfect information, measurement errors, omission of modal attributes, omission of the characteristics of the individual that influence his/her choice decision, and/or other errors in the utility function. By definition, error terms are unobservable and unmeasurable (Koppelman and Bhat, 2006). Utility is the function of attributes of the alternatives, characteristics of individual, and the interaction between alternatives and individual. This can be represented by the following equation (Koppelman and Bhat, 2006):   Vt,i  =  V (St)  +  V (Xi)  +  V (St, Xi), (2-6)  where, V (St) is the portion of utility of individual t, V (Xi)  represents the portion of utility of alternative i, and V (St, Xi) is the portion of the utility from interactions between alternative I and individual t (Koppelman and Bhat, 2006).  Different assumptions with regard to the distribution of the error term lead to different types of mode choice models. Normally distributed error terms lead to the Multinomial Probit model (MNP), which can only be calculated using multi-dimensional integration that makes it difficult to use in mode choice analysis (Koppelman and Bhat, 2006). On the other hand, Multinomial Logit (MNL) is the most widely used mathematical model for making probabilistic predictions of mode choices as a function of the systematic portion of the utility of each alternative. Identically  and independently distributed (IID) error terms with Type I Extreme Value (Gumbel) distribution lead to the MNL model (Koppelman and Bhat, 2006). However, identically and 24 independently distributed error term fails to capture taste variation (i.e. unexplained variation in travel behaviour; for example, two identical individuals making different choices). Furthermore, MNL models suffer from the Independence from Irrelevant Alternatives (IIA) property. The MNL model can be mathematically expressed as follows (Koppelman and Bhat, 2006):    P(𝑖) =exp( 𝑉𝑖)∑ exp( 𝑉𝑗)𝐽𝑗=1   , (2-7)  where, P(i) is the probability of choosing alternative i and Vit is the systematic component of alternative j.   2.6.1.1 Model Validation and Forecasting Performance Ideally, mode choice model validation should be performed using a different dataset that was not involved in the model estimation process. In some cases, where a small sample size is available, the entire dataset is desirable for model estimation (i.e. no independent dataset will be available for model validation). In such cases, the limited disaggregate validation technique can be performed by using market segments of the same dataset used for model estimation (Travel Forecasting Resource, 2016). The market segmentation variables can be categorized as follows:   Socioeconomic and demographic variables (e.g. income level, vehicle ownership, gender, age, etc.)  Geographic stratification (i.e. counties, neighborhoods, traffic analysis zones, etc.)  Level-of-service variables (i.e. trip length, trip time, etc.)  To quantify the forecasting performance of a mode choice model, the developed model can be used to predict known modal shares of an independent subset or the same dataset used for model estimation after market segmentation. The forecasting performance can be quantified using the following formula (Idris et al., 2015):    Forecasting Performance Measure (FPM) =  ∑ [(Pmm− Om)/Om , (2-8) 25 where, Pm is predicted trips, Om is observed trips for each mode (m)  Smaller value of FPM represents lesser forecasting errors with better forecasting performance of the developed model (Habib et al., 2012).  2.6.2 Distance Travelled Modelling The distance travelled by a vehicle is referred to as vehicle kilometres travelled (or vehicle miles traveled), whereas the distance travelled by a person is referred to as passenger kilometres travelled. Researchers have used regression techniques to estimate vehicle kilometres travelled (VKT) and passenger kilometres travelled (PKT) considering various explanatory variables such as socio-economic and demographic attributes, land use and built environment data, etc. (Cervero and Kockelman, 1997; Fan, 2007; Kockelman, 1996; Zegras, 2010; Zhou and Kockelman, 2008).    2.6.2.1 Simple Linear Regression Relationship between a dependent variable y and a single independent variable x can be estimated by using simple linear regression models. Mathematically, a simple linear regression model can be expressed as follows (Mendenhall and Sincich, 2007):   y = β0 + β1x + ε, (2-9)  where, y is the dependent variable, x is the independent variable, ε is the random error component, β0 is the y-intercept of the line and β1 is the slope of the line.  Furthermore, the least squares method can be used to estimate unknown parameters (β0 and β1). The line of means of Equation (2-9) is shown in Equation (2-10) and the fitted line is shown in Equation (2-11), respectively (Mendenhall and Sincich, 2007).    E(y) = β0 + β1x, (2-10)   ŷ = β̂0 + β̂1x, (2-11) 26 where, ŷ is an estimator of mean value, and β̂ and β̂1 are estimators of β0 and β1, respectively.  For a given data point (xi, yi), the observed value of y is yi and the predicted value of y would be obtained from the prediction equation (Mendenhall and Sincich, 2007).    yî = β̂0 + β̂1xi, (2-12)  The deviation of the ith value of y from its predicted value is   (yi − yî) = [yi − (β̂0 + β̂1xi)], (2-13)  The sum squares of the deviation of y values with respect to their predicted values is    SSE = ∑[yi − (β̂0 + β̂1xi)]2ni=1, (2-14)  Formulas for the least-squares estimates are as follows (Mendenhall and Sincich, 2007):   Slope: β̂1 =SSxySSxx , (2-15)     y − intercept: β̂0 = y̅ − β̂1x̅, (2-16)  where,   SSxy = ∑(xi − x̅)(yi − y̅)ni=1, (2-17)     SSxx = ∑(xi − x̅ )2ni=1, (2-18)  27 The goodness of fit of the least square equation, as shown in Equation (2-19), can be measured by estimating coefficient of determination. Coefficient of determination can be mathematically expressed as follows (Mendenhall and Sincich, 2007):   R2 =SSyy − SSESSyy= 1 −SSESSyy , (2-19)  where,   SSyy = ∑(yi − y̅ )2ni=1, (2-20)  A (1-α)100% prediction interval for y is as follows (Mendenhall and Sincich, 2007):   ŷ ± tα/2[Estimated standard deviation of (y − ŷ)] , (2-21)  or,   ŷ ± tα/2s√1 +1n+(xp − x̅)2SSxx, (2-22)  where, tα/2 is based on (n -2) degrees of freedom and n represents the sample size.  2.6.2.2 Multiple Regression Analysis A general linear regression or multiple regression analysis represents relationship between a dependent variable y and two or more independent variables. Mathematically, the general linear model can be expressed as follows (Mendenhall and Sincich, 2007):   y = β0 + β1x1 + β2x2 + …+ βkxk + ε, (2-23) where, y is the dependent variable, x1, x2, ..., xk are the independent variables, and ε is the random error term.  28 Equation (2-23) can be written in a matrix form as follows (Mendenhall and Sincich, 2007):  𝐘 =[    y1y2y3⋮yn]     𝐗 =[    11𝑥11𝑥21𝑥12𝑥221⋮𝑥31⋮𝑥32⋮1 𝑥𝑛1 𝑥𝑛2    ⋯ … … …   𝑥1𝑘𝑥2𝑘𝑥3𝑘⋮𝑥𝑛𝑘]      ?̂? =[     β̂1β̂2β̂3⋮β̂n]      𝛆 =[    ε1ε2ε3⋮εn]      The least-squares matrix equation and solution are shown in Equation (2-24) and Equation (2-25) respectively (Mendenhall and Sincich, 2007).   (𝑿′𝐗)?̂? = 𝑿′𝒀, (2-24)     ?̂? = (𝑿′𝐗)−1𝑿′𝒀, (2-25)  The multiple coefficient of determination can be defined as (Mendenhall and Sincich, 2007):    𝑅2 = 1 −𝑆𝑆𝐸𝑆𝑆𝑦𝑦, (2-26)  where,   SSE = ∑(yi − ŷ𝑖  )2ni=1,  (2-27)  SSyy = ∑(yi − y̅ )2ni=1, (2-28)       29 A (1-α)100% prediction interval for y is as follows(Mendenhall and Sincich, 2007):   ŷ ± tα/2s√1 + 𝐚′(𝑿′𝐗)−𝟏𝐚,  (2-29) where, 𝐚 =[    1𝑥1x2⋮xk]      2.6.2.3 Violation of Assumptions To use linear regression models, the following assumptions should hold (Corpuz et al., 2005; Montgomery, 2009; Nau, 2016; Neter et al., 1996, Rawlings et al., 2001):   Linearity – The predicted value of the dependent variable is a straight-line function of each independent variable.  Normality – The error term is normally distributed with non-negative values  Statistical independence – No correlation between consecutive errors  Homoscedasticity – The errors have a constant variance in terms of time, prediction, and other independent variables.  The normality assumption does not affect estimating variances among all linear estimators; however, it does affect the significant test and confidence interval of the parameters. Normality assumptions can be checked by plotting observed residuals, and skewness and kurtosis coefficients (Rawlings et al. 2001). Furthermore, non-normal distributions can be eliminated by introducing transformations of the dependent variables, such as the arcsin, the square root, the logarithmic, and the logistic transformations.  Due to the violation of statistical independence assumption of the linear regression models, the estimates loses precision. As a result, estimates become biased and it nullifies the significant tests (Rawlings et al., 2001). However, residual plotting according to the order may reveal patterns of residuals, which represents correlated errors.  30 The constant variance (homoscedasticity) plays a significant role in ordinary least squares (Rawlings et al., 2001). The violation of this assumption causes similar effects to the estimator as correlated errors. Due to the heterogeneous variances, least square estimators become inefficient. This issue can be handled by two approaches, such as transformation of the dependent variables and weighted least squares (Rawlings et al., 2001).   Outliers, inconsistent observations with the rest of the observations in the data set, can affect the least square estimators and, subsequently, goodness of fit of the model.  Outliers can be detected by plotting observed residuals; however, to have a common variance it is recommended that the residuals first be standardized (Rawlings et al., 2001). If all the assumptions are satisfied then the expected plot of the residuals with respect to the fitted values of the dependent variables should look like the following Figure 2-8 (Rawlings et al., 2001).   Figure 2-8 Expected pattern for a plot of residuals versus predicted1                                                     1 © 1998 Springer-Verlag New York, Inc, from Applied regression analysis: A research tool, Rawlings, J. O., Pantula, S. G., & Dickey, D. A., 2001, with permission of Springer 31 2.7 Summary GHGs are responsible for warming the earth’s surface to make it livable. However, the increase of GHGs beyond the natural absorption capacity of the earth cause climate change. Human activities (e.g. fossil fuels burning) cause an increase of GHGs in the earth’s atmosphere. In addition, different economic sectors are responsible for increasing GHG concentrations. Among them, transportation is one of the major sources of GHG emissions. Further, road transportation is the dominant contributing factor in the transportation sector. Different driving forces are associated with on-road GHG emissions. However, proper use of built environment and land use, such as density, diversity, and design factors have a strong correlation with the travel behaviour and subsequently, reduction in GHG emissions from on-road transportation. 32 Chapter 3: TRIBUTE: A TRIp-Based Urban Transportation Emissions Model for Municipalities 3.1 Outline This chapter proposes a TRIp-Based Urban Transportation Emissions (TRIBUTE) model. The developed model is trip-based, i.e. it considers the individual trip as the unit of analysis. TRIBUTE is intended to help municipalities estimate road transportation GHG emissions due to changes in various land use and transportation attributes by capturing the relationship between travel behavior and Vehicle Kilometres Travelled (VKT). This chapter presents the methodology of the developed model including model building, linking passenger trips to vehicle trips, and model application.   3.2 Methodology This research introduces a TRIp-Based Urban Transportation Emissions (TRIBUTE) model. TRIBUTE is intended to assist municipalities evaluate the impacts of various transportation and land use planning policies on GHG emissions from passengers transportation at the macroscopic level. The fundamental unit of analysis in TRIBUTE is the individual trip. Data required for model building comes from household travel surveys and emissions inventories. Importantly, TRIBUTE does not need a detailed transportation network model, which is a major advantage for small municipalities where a detailed network model is unavailable.   GHG emissions from on-road transportation are directly proportional to the total Vehicle Kilometres Travelled (VKT). VKT in turn is affected by the built environment. In specific, increasing land use density, balancing diversity (i.e. land use mix), and improving design can significantly decrease VKT, reduce per capita energy use, and lower GHG emissions (Brownstone and Golob, 2009; Cervero, 2010; Cervero and Kockelman, 1997; Chatman, 2003; Ewing and Cervero, 2010; Kockelman, 1996; Taylor, 2001; Tiwari et al., 2011; Zhang, 2004).  In general, VKT reduction can be attributed to two main factors: mode shift (i.e. passengers switching from single occupancy vehicles to high occupancy vehicles or non-motorized options) and trip length reduction (i.e. driving less), as shown in Figure 3-1.   33  Figure 3-1 Interaction among Trip Length Reduction, Mode Shift, and VKT Reduction  Bringing trip ends closer to each other through effective land use policies (e.g. increasing density, diversity, bike paths, sidewalks, etc.) leads to VKT reduction both indirectly (through mode shift) and directly (through driving for shorter distances). For example, if trip origins and destinations get closer to each other after increasing land use diversity, passengers will be more likely to choose active transportation instead of private automobile use. Further, for those passengers who will keep driving, VKT will be less due to driving for shorter distances. While trip length reduction is a function of the built environment, mode shift depends on several variables including personal attributes (e.g. age, income, car ownership, etc.), modal attributes (e.g. travel time, travel cost, travel distance, etc.), and land use attributes (e.g. density, diversity, etc.). As shown in Figure 3-1, trip length has a dual effect on VKT reduction (i.e. an indirect effect being one of the determinants of mode shift and a direct effect through driving less).   Figure 3-2 depicts the conceptual framework of TRIBUTE, with model building (base case scenario) displayed in the upper part, and model application (future scenario) displayed in the lower part. TRIBUTE is composed of two main components: a mode choice/shift model and an emissions forecasting model. The first component accounts for the determinants of VKT reduction (i.e. mode shift and trip length reduction). The second component translates the results in terms of VKT reduction and subsequently GHG emissions. The following subsections describe the components of TRIBUTE in more details. 34                              Figure 3-2 TRIp-Based Urban Transportation Emissions ModelDescriptive Analysis(base case)Modal Attributes (e.g. travel cost, travel time, parking cost, transit fare, etc.)Discrete ModeChoice ModelPersonal Attributes (e.g. age, gender, income, household size, car ownership, etc.)AutoDriverAuto PassengerTransit RiderActive TransportLand Use Attributes (e.g. dwelling type, density, diversity, sidewalks, bike paths, etc.)Expansion FactorPredictive Analysis(future scenarios)-Total Vehicle Kilometre Travelled (VKT) in kmTotal GHG emitted from transportation in CO2e tonRespective avg. emissions factor for each mode (CO2e kg/km)Respective avg. vehicle occupancy for each mode (pass/veh)Respective avg. emissions factor for each mode (CO2e kg/km)Total Vehicle Kilometre Travelled (VKT) in kmTotal GHG emitted from transportation in CO2e tonTotal Passenger Kilometre Travelled (PKT) in kmRespective avg. PKT for each mode (km/pass)Expected reduction in GHG emissions as a result of policy change+Policy scenarios (e.g. density, diversity, travel time, travel cost, parking cost, tolls, etc.)Discrete ModeShift ModelAutoDriverAuto PassengerTransit RiderActive TransportGrowth Factor Total Passenger Kilometre Travelled (PKT) in kmExpected Reduction in avg. PKTRespective avg. vehicle occupancy for each mode (pass/veh)AB35 3.2.1 Model Building (Base Case Scenario) To capture the effect of land use on mode shift, a discrete mode choice model was developed using Biogeme software (Bierlaire, 2003).  The developed mode choice model calculates the proportion of trips made by each mode of travel (e.g. car driver, car passenger, transit rider, and active transportation) in response to changes in personal, modal, and land use attributes. Biogeme’s user interface, sample code, and sample output results are presented in Appendix A, Appendix B, and Appendix C, respectively. Equations (3-1) through (3-6) show the general structure of the utility functions of the mode choice model.    VCar = ASCCar + Beta1_Car × x1 + Beta2_Car × x2 + ⋯+ Betan_Car × xn,  (3-1)     VCar pass = ASCCar pass + Beta1_Car pass × x1 + Beta2_Car pass × x2 + ⋯+ BetanCarpass × xn, (3-2)     VTransit = ASCTransit + Beta1_Transit × x1 + Beta2_Transit × x2 + ⋯+ BetanTransit × xn, (3-3)     VSchool bus = ASCSchool bus + Beta1_School bus × x1 + Beta2_School bus × x2 + ⋯+ BetanSchoolbus × xn, (3-4)     VWalk = ASCWalk + Beta1_Walk × x1 + Beta2_Walk × x2 + ⋯+ Betan_Walk × xn, (3-5)   VCycle = ASCCycle + Beta1_Cycle × x1 + Beta2_Cycle × x2 + ⋯+ Betan_Cycle × xn, (3-6)  Where, V is the utility function, ASC is the alternative specific constant, Beta is the parameter value, and x is the explanatory variable.     36 Further, the proportion of trips made by each mode of travel can be determined using the Multinomial Logit (MNL) model, as shown in Equations (3-7) through (3-12).   PCar =exp(𝑉𝐶𝑎𝑟)∑ exp(𝑉𝑗)𝐽𝑗=1   (3-7)     PCar pass =exp(𝑉𝐶𝑎𝑟 𝑝𝑎𝑠𝑠)∑ exp( 𝑉𝑗)𝐽𝑗=1  , (3-8)     PTransit =exp(𝑉𝑇𝑟𝑎𝑛𝑠𝑖𝑡)∑ exp( 𝑉𝑗)𝐽𝑗=1  , (3-9)     PSchool bus =exp(𝑉𝑆𝑐ℎ𝑜𝑜𝑙 𝑏𝑢𝑠)∑ exp( 𝑉𝑗)𝐽𝑗=1  , (3-10)     PWalk =exp(𝑉𝑊𝑎𝑙𝑘)∑ exp( 𝑉𝑗)𝐽𝑗=1  , (3-11)     PCycle =exp(𝑉𝐶𝑦𝑐𝑙𝑒)∑ exp( 𝑉𝑗)𝐽𝑗=1  , (3-12)  After calculating the proportion of trips made by each mode of travel, the total PKT by each mode can be calculated given respective average modal PKT (calculated from emissions inventories). Total PKT can then be converted to VKT using respective average vehicle occupancy for each mode, as shown in Equation (3-13) (the details of linking PKT to VKT are discussed in Section 3.2.2.). Average vehicle occupancies were calculated from the household travel survey by analyzing the survey question: “when travelling by automobile, how many people travelled with you?”   VKT =Average PKT (km per person) ×  Modal share(%) ×  PopulationVehicle occupancy, (3-13) 37 Finally, total GHG emissions from passenger transportation for the base case scenario can be calculated through multiplying the total VKT by each mode by respective average emissions factors (calculated from emissions inventories for different vehicle and fuel types).    Total GHG Emissions (𝐶𝑂2 eq. )= Total VKT(km) × Average emissions factors (𝐶𝑂2 eq. per km), (3-14)  As previously discussed and presented in Figure 3-1, VKT reduction can be attributed to two main factors: mode shift and trip length reduction. To capture the effect of land use attributes on trip length reduction, a Passenger Kilometres Travelled (PKT) estimation model was developed.  Separate PKT estimation models were developed for each mode of travel (car, transit, walk, and cycle) using multiple linear regression analysis. The structure of the developed models is shown in Equations (3-15) through (3-18). Various land use attributes such as density and diversity were considered as explanatory variables in the PKT estimation models.    PKTCar = InterceptCar + Beta1_Car × x1 + Beta2_Car × x2 + ⋯+ Betan_Car × xn, (3-15)     PKTTrasnit = InterceptTransit + Beta1_Transit × x1 + Beta2_Transit × x2 + ⋯+ Betan_Transit × xn, (3-16)     PKTWalk = InterceptWalk + Beta1_Walk × x1 + Beta2_Walk × x2 + ⋯+ Betan_Walk × xn, (3-17)     PKTCycle = InterceptCycle + Beta1_Cycle × x1 + Beta2_Cycle × x2 + ⋯+ BetanCycle × xn, (3-18)  3.2.2 Linking Passenger Trips to Vehicle Trips As presented earlier, data required for developing TRIBUTE comes from two main sources: household travel surveys and emissions inventories (described in Section 3.2). Household travel surveys are reliable sources for socioeconomic, demographic, modal share, and average vehicle occupancy data. Data on average VKT, number of registered vehicles, and total GHG emissions 38 were extracted from emissions inventories. Linking household travel survey data to information from emissions inventories was not a straightforward task given that the first reports on passenger trips while the latter reports on vehicle trips. Figure 3-3 depicts the process used for linking passenger trips to vehicle trips and eventually feeding TRIBUTE.  Emissions inventories are considered reliable sources for calculating average PKT and average emissions factors, especially when they utilize the resident-based methodology using vehicle registration data for reporting on emissions (as opposed to other methodologies that rely on fuel sales or modelling). Given that BC’s emissions inventories are resident-based (Ministry of Environment, 2014c), they were used to calculate average PKT and average GHG emissions factors for each mode from information on total GHG emissions and total VKT, as shown in Figure 3.3. To differentiate between car drivers’ average PKT and car passengers’ average PKT, the PKT split was extracted from the household travel survey knowing the origins and destinations of each trip. 39                     Figure 3-3 Linking Passenger Trips to Vehicle TripsModal Share(pass)Household Travel SurveyEmissions InventoriesModal Share(%)Passenger Kilometres Travelled (PKT) Split (%)Avg. Vehicle Kilometres Travelled (VKT) (km/veh)Avg. Vehicle Occupancy for each Mode(pass/veh)Total GHG Emissions (CO2e ton)Number of Registered Vehicles (veh)Total Vehicle Kilometres Travelled (VKT) (km)Total Passenger Kilometres Travelled (PKT) (km)Expansion FactorAvg. GHG Emissions Factor for each Mode(CO2e kg/km)BAvg. Passenger Kilometres Travelled (PKT) for each Mode (km/pass)AFeeds TRIBUTEFeeds TRIBUTE40 3.2.3 Model Application (Future Scenarios) Upon completion of model building, TRIBUTE can be used to forecast GHG emissions from the transportation sector in response to future transportation and land use policy scenarios by running its joint mode shift and emissions forecasting models. Future estimates are required to account for population growth, new emissions factors for new vehicle technologies, as well as changes in personal, modal, and land use attributes. The following bullets describe the steps required for forecasting future GHG using TRIBUTE:   Trip length reduction in response to changes in land use attributes (e.g. density, diversity, and design) can be captured by running the PKT estimation model developed in Section 3.2.1.   Mode shift in response to changes in land use attributes can be captured by running the mode choice model, as shown in Section 3.2.1. Importantly, the trip length reduction model (previous step) is expected to feed in the mode shift model to account for the effect of trip length reduction on mode shift.     Total VKT by each mode in the future scenario can then be calculated using Equation (3-19) as follows:   Future VKT (km)= Change in  PKT (km) ×  Modal shift (%) ×  Expected populationVehicle occupancy, (3-19)   Next, future GHG emissions can be estimated using Equation (3-20).   Emissions in future scenario (𝐶𝑂2 eq. )= Future VKT (km)× Average emissions factors (𝐶𝑂2 eq. per km),  (3-20)   41  Finally, the change in passenger transportation GHG emissions from base case can be calculated by Equation (3-21). This process can be repeated for different future scenarios and/or land use policies and subsequently, the policy/scenario combination that best meets GHG emissions reduction targets can be selected.  Change(%) =Emissions in future scenario −  Emissions in base caseEmissions in base case× 100, (3-21)  3.3 Summary This chapter proposed a TRIp-Based Urban Transportation Emissions (TRIBUTE) model for estimating and forecasting GHG emissions from passenger transportation. TRIBUTE is a trip-based approach given that it uses the individual trip as its fundamental unit of analysis. TRIBUTE can capture the relationship between travel behavior and VKT in estimating as well as in forecasting emissions. As such, it can be used to assist municipalities evaluate alternative transportation and land use policy scenarios and eventually select the one(s) that help them meet their future GHG emissions targets.  42 Chapter 4: Case Study 4.1 Outline As described in Chapter 3, the developed emissions forecasting model (TRIBUTE) relies mainly on household travel survey and emissions inventory data. This represents a major advantage of TRIBUTE as it can be used in small municipalities where a detailed transportation network model is unavailable. This chapter presents the application of TRIBUTE to estimate GHG emissions from passenger transportation in the City of Kelowna, British Columbia as a case study. This chapter is comprised of two main parts. In section 4.2, detailed information on the study area is presented including location, demographics, transportation network, and emissions reduction target. In Section 4.3, an overview of the data used and its sources is presented.   4.2 Study Area The City of Kelowna lies in the Okanagan Valley located in the Southern Interior of British Columbia, Canada. With a total area of 211.8 square kilometres and population of 117,312 in 2011, Kelowna is considered the largest city in British Columbia’s Okanagan Valley (Statistics Canada, 2016). Similar to many North American cities, the City of Kelowna faces severe suburbanization and urban sprawl promoted by the popularity of low-density car-oriented developments.  As comparison, Kelowna’s average population density (117,312 residents/211.8 km2 = 554 residents/km2) is approximately one tenth (1/10) of Vancouver’s, which has a much higher active transportation modal share.  The City of Kelowna is comprised of 10 sectors with 191 traffic analysis zones. Figure 4-1 shows the distribution of population density among the city’s 10 sectors and 191 traffic analysis zones (TAZs).   43  Figure 4-1 Distribution of Population Density in the City of Kelowna116°0'0"W116°0'0"W120°0'0"W120°0'0"W124°0'0"W124°0'0"W128°0'0"W128°0'0"W54°0'0"N54°0'0"N52°0'0"N52°0'0"N50°0'0"N50°0'0"N48°0'0"NOverview of the Study Area: City of KelownaBritish Columbia, Canada0 3 6 9 121.5KilometersÜMd. Nobinur RahmanApril 2016Data Source: City of KelownaCoordinate System: NAD 1983 UTM Zone 11NOkanaganLakeOkanaganLakeCity of KelownaLegendSector BoundaryTraffic Analysis ZonePopulation per Hectares0 - 4.54.6 - 1314 - 2526 - 4950 - 190200 - 530540 - 1,4001,500 - 5,000City BoundaryCensus Subdivisions44 4.2.1 Kelowna’s Growth Projections The City of Kelowna is expected to accommodate additional 63,070 people by the end of 2040, on top of its 117,312 people, which will result in a total population of 180,382. Among all age groups, the senior population is expected to increase substantially. The average household size is expected to be 2.11 persons per unit by 2030 (City of Kelowna, 2013). Table 4-1 outlines Kelowna’s population growth projections (Rahman et al., 2016). In 2013, the City of Kelowna’s total population was 119,801, which was calculated by taking the same growth rates between 2011 and 2014.   Table 4-1 Kelowna’s Projected Growth Rates (Rahman et al., 2016) Time Period Average Annual Growth Rate Population at End of Period 2014  - 121,045 2015 - 2019 1.7% 131,690 2020 - 2024 1.6% 142,708 2025 - 2029 1.5% 153,889 2030 - 2034 1.5% 165,782 2035 - 2039 1.4% 177,891 2040 1.3% 180,382  4.2.2 Kelowna’s Transportation Network Over the past few years, the City of Kelowna has pursued several measures in constructing infrastructure for active transportation and public transit. For example, to promote walking in the City of Kelowna, different walking facilities such as sidewalks, crosswalks, and shared pathways have been developed (City of Kelowna, 2016). To promote bicycling, the City of Kelowna has installed one of the most widespread bicycle networks in Canada (City of Kelowna, 2016). The city has nearly 300 km on-street bicycle lanes. Moreover, during 2012-2014, the City of Kelowna has also carried out various transit infrastructure and service projects (e.g. improving transit facilities, increasing frequency of the transit service, and improving transit route coverage) (City of Kelowna, 2016). The road network, active transportation network, and public transit network of the City of Kelowna are illustrated in Figure 4-2, Figure 4-3, and Figure 4-4, respectively. 45  Figure 4-2 Road Network in the City of Kelowna   !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!! !!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!! !!!!! ! !! !!!!!!!!!!!!!!!!!!!!!!!! !! !!!!!!!!!! !!! !! !!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!! ! !!!!!!!!!!!!!!!!!! !!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!! !! !!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!! !!!!!!!! !!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!! !!!! !!!!!!!!!!!!!!!!!!!!! !!! ! !!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !! !!!! !!!!! !!!!! !!!!!!!!!!!!!!!!! ! !!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!! ! !!!! !!!!!!!!!!!!!!!! !!!!!!!!!! !!!! !!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!! !!!! !!! !!!!!!! !!!!!!!! !!! !!!!!!!! !!!!!!!!! !!!! ! !!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!! !! !!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!! !!!!!!!!!! !!!!!! ! !!!!!!!!!!!!!!!! !!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!! !!!! !! !! ! !!!!! ! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!! !!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!! !!!!!!!! !!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!! !!!!!!!!!!!!!!!! !!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!! !!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!! !!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!0 2.5 5 7.5 101.25KilometersÜMd. Nobinur Rahman                  Data Source: City of KelownaApril 2016                                    Coordinate System: NAD 1983 UTM Zone 11NOkanaganLakeOkanaganLakeRoad Network of the City of KelownaLegend! Road IntersectionsRoad CentrelinesSector BoundaryCity BoundaryCensus Subdivisions46  Figure 4-3 Active Transportation Network in the City of Kelowna   0 2.5 5 7.5 101.25KilometersÜMd. Nobinur Rahman                  Data Source: City of KelownaApril 2016                                    Coordinate System: NAD 1983 UTM Zone 11NOkanaganLakeOkanagan LakeActive Transportation Network in the City of KelownaLegendSidewalkBicycle NetworkSector BoundaryCity BoundaryCensus Subdivisions47  Figure 4-4 Public Transit Network in the City of Kelowna 0 2.5 5 7.5 101.25KilometersÜMd. Nobinur Rahman                  Data Source: City of KelownaApril 2016                                    Coordinate System: NAD 1983 UTM Zone 11NOkanaganLakeOkanagan LakePublic Transit Network in the City of KelownaLegendBus StopsBus RoutesSector BoundaryCity BoundaryCensus Subdivisions48 4.2.3 Kelowna’s GHG Emissions Much of the motivation for active transportation improvements in the city has arisen from Kelowna’s traditionally high car dependency. The high car dependency in Kelowna has produced the second highest per-capita road transportation carbon footprint in British Columbia, coming after Prince George (City of Kelowna, 2009). According to the 2007 Community Energy and Emissions Inventory (CEEI), on-road transportation contributed 506,640 CO2 eq. ton, representing about 65% of Kelowna’s community GHG emissions in 2007 (City of Kelowna, 2012).  Figure 4-5 depicts the three major sources of GHG emissions in the City of Kelowna.    Figure 4-5 Sources of GHG Emissions in the City of Kelowna (City of Kelowna, 2012)   Kelowna residents depend highly on personal vehicles (mainly light trucks, vans, and SUVs) as a mode of travel. As a result, personal vehicles were responsible for almost two thirds of the on-road emissions produced in 2007 (Figure 4-6). If the city continues business-as-usual, on-road transportation GHG emissions are expected to escalate to unprecedented levels by 2040, putting citizens’ health and quality of life in question. In response, the City of Kelowna has committed to finding ways to tackle the challenges posed by climate change and pledged to reducing its GHG emissions by 33% from 2007 levels by 2040 (City of Kelowna, 2012). 49  Figure 4-6 On-Road Emissions from Different Vehicle Types (City of Kelowna, 2009)  4.3 Dataset Description The data used in this analysis came from two main sources: a household travel survey and emissions inventories. Socioeconomic, demographic, modal shares, and average vehicle occupancy data was acquired from the 2013 Okanagan Household Travel Survey (HHTS). The Okanagan HHTS is a household-based survey conducted in fall 2013 and includes detailed socioeconomic and demographic information about residents of the Central Okanagan and the City of Vernon along with their travel choices.   A sample of 13,633 trips, which represents a total of 429,559 trips that originate in the City of Kelowna on a typical week day, is recorded in the Okanagan HHTS. Among these trips, 89.64% stayed in Kelowna and 10.36% crossed the city boundaries to different destinations in the Okanagan Valley including Vernon, Lake Country, West Kelowna, West Bank First Nation (WFN), Peachland, Central Okanagan East, Central Okanagan West, North Okanagan, and South Okanagan. For each trip, the trip purpose and mode of travel are reported. The descriptive statistics of the 2013 Okanagan HHTS are shown in Table 4-2. Among the cross boundary trips, 50 the highest number of trips (4.28%) ends in West Kelowna, and the second and third highest end in Lake Country (1.80%) and Westbank (1.36%), respectively, as shown in Figure 4-7.   Table 4-2 Descriptive Statistics of 2013 Okanagan HHTS Variable Value Sample Size Percentage Destinations Vernon 146  (4,145)* 1.07% (0.97%)* Lake Country 292  (7,715)* 2.14% (1.80%)* Kelowna 12,057  (385,055)* 88.44% (89.64%)* West Kelowna 613  (18,400)* 4.50% (4.28%)* WFN 207  (5,851)* 1.52% (1.36%)* Peachland 71  (1,564)* 0.52% (0.36%)* Central Okanagan East 141  (3,987)* 1.03% (0.93%)* Central Okanagan West 23  (648)* 0.17% (0.15%)* North Okanagan CMA (minus Vernon) 18  (538)* 0.13% (0.13%)* South Okanagan CMA 7  (161)* 0.05% (0.04%)* External 57  (1,494)* 0.42% (0.35%)* Trip Purposes To Work 2,127 (64,987)* 15.66% (15.17%)* To grade school 484 (17,828)* 3.56% (4.16%)* To post-secondary school 382 (12,225)* 2.81% (2.85%)* To Restaurant 397 (11,944)* 2.92% (2.79%)* For Recreation 627 (19,811)* 4.62% (4.62%)* *Values between parentheses represent the expanded sample size     51 Table 4-2 Descriptive Statistics of 2013 Okanagan HHTS (Continued) Variable Value Sample Size Percentage Trip Purposes For a Social outing 496 (15,193)* 3.65% (3.55%)* For Shopping 1,496 (45,499)* 11.01% (10.62%)* For Personal business 897 (27,199)* 6.60% (6.35%)* To Home 4,980 (158,023)* 36.66% (36.89%)* To drive or pick-up someone 1,062 (35,277)* 7.82% (8.24%)* Other 638 (20,383)* 4.70% (4.76%)* Mode of Transportation Car Driver 9,194 (284,122)* 67.57% (66.35%)* Car Passenger 2,041 (66,691)* 15.00% (15.57%)* Transit 600 (20,577)* 4.41% (4.81%)* Walk 1,087 (33,530)* 7.99% (7.83%)* Cycle 166 (6,122)* 1.22% (1.43%)* School Bus 426 (14,156)* 3.13% (3.31%)* Other 93 (3,015)* 0.68% (0.70%)* *Values between parentheses represent the expanded sample size  52  Figure 4-7 Distribution of Trips Originated in Kelowna  The most frequent trip purpose recorded in the Okanagan HHTS is “to home”. The trip purpose “to home” indicates that going back home may either come from work/school or other locations. The work trips, going to work regardless of origin locations, are the second highest trips at 15.17% of total trips, which is followed by shopping trips (10.62%), as illustrated in Figure 4-8.  According to the HHTS, there is a very high percentage (81.92%) of car users in the study area (66.35% car drivers and 15.57% car passengers), as shown in Figure 4-9. On the other hand, active transportation accounts for 9.26% (7.83% walk and 1.43% cycle), public transit represents 4.81%, and only 3.31% of all trips are made by school bus.   53  Figure 4-8 Trips Purposes   Figure 4-9 Modal Share  54 This research focused on trips that originated and ended within Kelowna (i.e. cross-boundary trips were outside the scope of this study). A subset of the trips originated in Kelowna (described in the Table 4-2) was extracted to include only trips that both originated and ended in Kelowna. The extracted subset included a sample of 12,057 trips which represents a total of 385,055 trips that stay within the boundaries of the City of Kelowna on a typical week day. Further, the trips within Kelowna (a sample of 12,057 trips) were divided into two broad categories based on the trip purposes, namely, commuting trips, and non-commuting trips. The following section describes the descriptive statistics of the commuting and non-commuting trips separately.   4.3.1 Commuting Trips Commuting trips refer to trips associated with work-related activities. The HHTS includes a total number of 4,299 commuting trips (i.e. home to work/school, work/school to home, and other than home locations to work/school). There are 3,533 home-based commuting trips, i.e. home to work/school and the reverse trips, in the 2013 HHTS. In addition, there are various complex tours that consist of a numbers of linked trips in the HHTS. For example, the tour “dropping children off at school from home before going to work” consists of two trips: going from home to school and from school to work. In such tour, only the second trip (i.e. school to work) is considered as a commuting trip (Acker and Witlox, 2011). There are 766 commuting trips starting at locations other than home and ending at the work place. The descriptive statistics of commuting trips are shown in Table 4-3.  Table 4-3 Descriptive Statistics of Commuting Trips Variable Value Sample Size Percentage Gender Male 2,055 (66,765)* 47.80% (47.95%)* Female 2,244 (72,482)* 52.20% (52.05%)* Age (In years) 05-14 616 (23,261)* 14.33% (16.70%)* 15-24 719 (24,428)* 16.72% (17.54%)* 25-34 753 (22,375)* 17.52% (16.07%)* *Values between parentheses represent the expanded sample size  55 Table 4-3 Descriptive Statistics of Commuting Trips (Continued) Variable Value Sample Size Percentage Age (In years) 35-44 729 (23,815)* 16.96% (17.10%)* 45-54 779 (25,393)* 18.12% (18.24%)* 55-64 612 (17,361)* 14.24% (12.47%)* 65 and over 91 (2,616)* 2.12% (1.88%)* Household Size 1 353 (14,804)* 8.21% (10.63%)* 2 1,642 (39,429)* 38.19% (28.32%)* 3 802 (27,459)* 18.66% (19.72%)* 4+ 1,502 (57,557)* 34.94% (41.33%)* Number of Vehicles 0 121 (4,134)* 2.81% (2.97%)* 1 1,169 (36,991)* 27.19% (26.56%)* 2 1,851 (58,036)* 43.06% (41.68%)* 3+ 1,158 (40,087)* 26.94% (28.79%)* Transit Pass Holding Yes 534 (18,019)* 12.42% (12.94%)* No 3,765 (121,229)* 87.58% (87.06%)* Driver's License Holding Yes 3,405 (106,732)* 79.20% (76.65%)* No 894 (32,516)* 20.80% (23.35%)* Dwelling Type Single Detached House 2,922 (96,691)* 67.97% (69.44%)* Apartment or Condo 676 (20,451)* 15.72% (14.69%)* Townhouse or Row House 477 (14,422)* 11.10% (10.36%)* Duplex 187 (6,338)* 4.35% (4.55%)* *Values between parentheses represent the expanded sample size  56 Table 4-3 Descriptive Statistics of Commuting Trips (Continued) Variable Value Sample Size Percentage Occupation Professional 1,291 (40,736)* 43.67% (44.32%)* Business 292 (8,642)* 9.88% (9.40%)* Skilled Technical Worker 208 (6,493)* 7.04% (7.06%)* Sales 148 (4,678)* 5.01% (5.09%)* Clerical 226 (6,835)* 7.65% (7.44%)* Trades 215 (6,407)* 7.27% (6.97%)* Commercial Driver 22 (711)* 0.74% (0.77%)* Other 554 (17,412)* 18.74% (18.94%)* Household Income Less than $25,000 300 (9,936)* 6.98% (7.14%)* $25,000 to Less than $45,000 514 (17,119)* 11.96% (12.29%)* $45,000 to Less than $65,000 675 (21,187)* 15.70% (15.22%)* $65,000 to Less than $100,000 1,252 (39,766)* 29.12% (28.56%)* $100,000 or more 1,558 (51,241)* 36.24% (36.80%)* Modes of Transportation Car Driver 2,547 (79,418)* 59.25% (57.03%)* Car Passenger 562 (19,629)* 13.07% (14.10%)* Transit 348 (11,701)* 8.09% (8.40%)* Walk 443 (14,499)* 10.30% (10.41%)* Cycle 235 (7,853)* 5.47% (5.64%)* School Bus 142 (5,425)* 3.30% (3.90%)* Other 22 (724)* 0.51% (0.52%)* *Values between parentheses represent the expanded sample size  57 In the HHTS, it is reported that females make slightly more commuting trips (52.05%) than males in the City of Kelowna, as shown in Figure 4-10. The highest percentage of commuting trips is made by people in the 45 to 54 years age category, followed by 15 to 24 years, and then 35 to 44 years, as shown in Figure 4-11.  It is also seen from the HHTS that four or more people living in a single household are responsible for more commuting trips (41.33%) than the lower number of people in a household, as illustrated in Figure 4-12. The second highest trip making percentage (28.32%) is made by households having two people, followed by households having three people, and then single occupant households. On the other hand, two vehicles households are recorded as making the highest percentage of commuting trips in the 2013 Okanagan HHTS, as shown in Figure 4-13.   Figure 4-10 Gender Distribution (Commuting Trips)  58  Figure 4-11 Age Distribution (Commuting Trips)   Figure 4-12 Household Size (Commuting Trips)  59  Figure 4-13 Vehicle Ownership (Commuting Trips)  In addition, 76.65% of people are recorded in the HHTS as having their driver’s license, while 12.94% have a transit pass, as shown in Figure 4-14 and Figure 4-15, respectively. Higher numbers of people having a driver’s licence and a lower percentage of people without a transit pass indicates higher auto dependency and lower transit ridership for commuting trips.  60  Figure 4-14 Driver License (Commuting Trips)    Figure 4-15 Transit Pass (Commuting Trips)  61 Different types of dwellings and occupancies give an idea about the standard of living. The HHTS shows that the dominant dwelling type is a single detached house, and that includes 69.44% of the total different dwelling types, as shown in Figure 4-16. This is followed by apartments then townhouses. On the other hand, the dominant occupation in the HHTS is professional, consisting of 44.32% of all commuting trips, as illustrated in Figure 4-17. In addition, more than 35% of people have an income level more than $100,000, which is almost equal to the people having salaries less than $65,000, as depicted in Figure 4-18. A person with a professional occupation, living in a single detached house, and having a higher income exhibits higher tendency to auto dependency.    Figure 4-16 Dwelling Types (Commuting Trips)  62  Figure 4-17 Occupation Distribution (Commuting Trips)   Figure 4-18 Income Distribution (Commuting Trips)  63 The City of Kelowna consists of over 71% of commuting trips using cars, including 57.03% car drivers and 14.10% car passengers, as shown in Figure 4-19. On the other hand, over 16% of commuting trips are made by active transportation that includes 10.41% of walking. In addition, transit ridership is 8.40% in 2013 for commuting trips.   Figure 4-19 Modal Share (Commuting Trips)  4.3.2 Non-commuting Trips Non-work-related activities were considered as non-commuting trips in this research. Non-commuting trips are comprised of home-based discretionary trips (e.g. shopping, entertainment, etc.), non-home-based discretionary trips, and work to discretionary trips. Home-based discretionary trips are those trips that either start or end at home (National Cooperative Highway Research Program, 2012). The total number of home-based discretionary trips are 5,046, as recorded in the HHTS. In addition, there are 1,933 trips categorized as non-commuting trips as their origins and destinations are neither home nor work or school. Furthermore, there are 779 trips originating at work or school and with destination at discretionary locations. By combining different types of non-commuting trips, there are a total of 7,758 non-commuting trips recorded in the HHTS. The descriptive statistics of non-commuting trips are shown in Table 4-4. 64 Table 4-4 Descriptive Statistics of Non-Commuting Trips Variable Value Sample Size Percentage Gender Male 3,054 (95,284)* 39.37% (38.76%)* Female 4,704 (150,559)* 60.63% (61.24%)* Age (In years) 05-14 547 (19,835)* 7.05% (8.07%)* 15-24 589 (20,473)* 7.59% (8.33%)* 25-34 1,015 (29,826)* 13.08% (12.13%)* 35-44 1,338 (45,564)* 17.25% (18.53%)* 45-54 1,309 (44,561)* 16.87% (18.13%)* 55-64 1,474 (42,569)* 19.00% (17.32%)* 65 and over 1,486 (43,017)* 19.15% (17.50%)* Household Size 1 900 (38,133)* 11.60% (15.51%)* 2 3,554 (85,363)* 45.81% (34.72%)* 3 1,270 (43,455)* 16.37% (17.68%)* 4+ 2,034 (78,893)* 26.22% (32.09%)* Number of Vehicles 0 294 (10,014)* 3.79% (4.07%)* 1 2,430 (77,073)* 31.32% (31.35%)* 2 3,312 (100,019)* 42.69% (40.68%)* 3+ 1,722 (58,736)* 22.20% (23.89%)* Transit Pass Holding Yes 526 (17,222)* 6.78% (7.01%)* No 7,232 (228,621)* 93.22% (92.99%)* *Values between parentheses represent the expanded sample size   65 Table 4-4 Descriptive Statistics of Non-Commuting Trips (Continued) Variable Value Sample Size Percentage Driver's License Holding Yes 6,842 (214,025)* 88.19% (87.06%)* No 916 (31,818)* 11.81% (12.94%)* Dwelling Type Single Detached House 5,367 (172,602)* 69.18% (70.21%)* Apartment or Condo 1,195 (37,277)* 15.40% (15.16%)* Townhouse or Row House 882 (26,542)* 11.37% (10.80%)* Duplex 234 (6,993)* 3.02% (2.84%)* Mobile Home 80 (2,431)* 1.03% (0.99%)* Occupation Professional 1,703 (55,040)* 52.50% (53.20%)* Business 385 (11,608)* 11.87% (11.22%)* Skilled Technical Worker 232 (7,145)* 7.15% (6.91%)* Sales 171 (5,229)* 5.27% (5.05%)* Clerical 331 (10,204)* 10.20% (9.86%)* Trades 229 (7,082)* 7.06% (6.85%)* Commercial Driver 25 (812)* 0.77% (0.78%)* Other 168 (6,340)* 5.18% (6.13%)* Household Income Less than $25,000 1,060 (35,462)* 13.66% (14.42%)* $25,000 to Less than $45,000 1,139 (37,162)* 14.68% (15.12%)* $45,000 to Less than $65,000 1,316 (39,151)* 16.96% (15.93%)* $65,000 to Less than $100,000 2,180 (66,079)* 28.10% (26.88%)* $100,000 or more 2,063 (67,990)* 26.59% (27.66%)* *Values between parentheses represent the expanded sample size  66 Table 4-4 Descriptive Statistics of Non-Commuting Trips (Continued) Variable Value Sample Size Percentage Mode of Transportation Car Driver 5,439 (170,971)* 70.11% (69.54%)* Car Passenger 1,258 (40,791)* 16.22% (16.59%)* Transit 192 (6,780)* 2.47% (2.76%)* Walk 651 (19,773)* 8.39% (8.04%)* Cycle 180 (6,170)* 2.32% (2.51%)* Other 38 (1,360)* 0.49% (0.55%)* *Values between parentheses represent the expanded sample size  Females are responsible for more non-commuting trips than males, as shown in Figure 4-20. In particular, females are responsible for 61.24% of non-commuting trips, which is even higher than their share of commuting trips. Males generated 38.76% of non-commuting trips as recorded in the HHTS. On the other hand, those aged 35 years and over generates over than 70% of non-commuting trips, as illustrated in Figure 4-21.   Figure 4-20 Gender Distribution (Non-Commuting Trips) 67  Figure 4-21 Age Distribution (Non-Commuting Trips)  From the HHTS, it is seen that the majority (34.72%) of non-commuting trips are coming from households having two persons living together, as shown in Figure 4-22, followed by those having four or more persons in a single household, and then three people in a single household. However, 40.68% of non-commuting trip makers have two vehicles in their household, while only 4.07% have no vehicles, as shown in Figure 4-23. In addition, 31.35% household have one vehicle, while 23.89% have three or more vehicles in their household.  A very high percentage of people (92.99%) does not have a transit pass of their own and 87.06% of people does have a driver’s license, as depicted in Figure 4-24 and Figure 4-25, respectively, which indicates a higher auto dependency for non-commuting trips in the City of Kelowna.  68  Figure 4-22 Household Size (Non-Commuting Trips)   Figure 4-23 Vehicle Ownership (Non-Commuting Trips)  69  Figure 4-24 Transit Pass (Non-Commuting Trips)   Figure 4-25 Driver License (Non-Commuting Trips)  70 The majority of non-commuting trip makers are living in a single detached house (70.21%), as shown in Figure 4-26, followed by those living in an apartment (15.16%), and then those living in a townhouse (10.80%). In terms of occupation, the majority of non-commuting trips are made by professionals (53.20%), as illustrated in Figure 4-27, while 11.22% of non-commuting trips were made by businesspersons. In addition, over 50% of people are involved in generating non-commuting trips have salaries over $65,000, as depicted in Figure 4-28. This indicates that professional people living in single detached houses and having comparatively higher salaries are making a higher percentage of non-commuting trips.    Figure 4-26 Dwelling Type Distribution (Non-Commuting Trips)  71  Figure 4-27 Occupation Distribution (Non-Commuting Trips)   Figure 4-28 Income Distribution (Non-Commuting Trips)  72 The majority of non-commuting trips (86.13%) are made by using cars where car drivers are responsible for 69.54% of non-commuting trips, as shown in Figure 4-29. However, slightly more than 10% of people take active transportation in making non-commuting trips. Furthermore, transit ridership is very low (2.76%) in non-commuting trips. The school bus option is absent in non-commuting trips.   Figure 4-29 Modal Share (Non-Commuting Trips)  4.3.3 Vehicle Occupancy Average car occupancies were estimated from the 2013 HHTS by analyzing the answer of the survey question: “When travelling by automobile, how many people travelled with you?” By summing up the number of passengers (i.e. people who rode with the driver) in a specific vehicle, average vehicle occupancies were calculated. If the driver drove alone then the occupancy of the vehicle was taken as 1. Average transit bus and school bus occupancies were acquired from the Sustainable Transportation Partnership of the Central Okanagan (STPCO) and School District (SD) No. 23 respectively. STPCO is a partnership of four local governments, a first nation, and the regional district of Central Okanagan. The average vehicle occupancies for different travel modes are shown in Table 4-5. 73 Table 4-5 Average Vehicle Occupancy Travel Mode Car Driver Car Passenger Transit Bus School Bus Vehicle Occupancy 1.39 2.69 10 50  4.3.4 GHG Emissions Factors Data on average vehicle kilometres travelled (VKT), number of registered vehicles, and total GHG emissions was extracted from the Community Energy and Emissions Inventory (CEEI) (Ministry of Environment, 2014b). Different vehicle types, fuel types, fuel consumption, and corresponding emissions were recorded in the CEEI report. The previous data was used to calculate weighted average emissions factors for the car and transit options. Emissions factors, however, have changed between 2007 and 2010 due to the changes in fuel content as well as the imposed legislation to use renewable fuel (Ministry of Environment, 2014a). The most updated emissions factors obtained from the CEEI report are shown in Table 4-6.  Table 4-6 GHG Emission Factors (Ministry of Environment, 2014b)   Total Consumptions (Litres) No. of Vehicles Average VKT (km/veh) GHG (CO2 eq. ton) EF1 (kg/km) Ave. EF2 (kg/km) Small Passenger Cars Hybrid 64,366 58 21,100 144 0.118 0.278 Gasoline 40,006,983 25,125 16,600 89,728 0.215 Diesel 1,098,410 682 23,900 2,913 0.179 Large Passenger Cars Hybrid 348,611 243 25,300 776 0.126 Gasoline 23,244,105 12,350 16,500 52,202 0.256 Diesel 166,165 123 14,500 440 0.247 Light trucks, Vans, SUVs Hybrid 271,452 118 26,400 612 0.196 Gasoline 93,983,001 32,497 20,100 212,846 0.326 Diesel 3,096,191 970 19,800 8,190 0.426 Other 238,570 120 11,700 365 0.260 Buses Gasoline 302,518 113 17,500 677 0.342 0.543 Diesel 604,612 115 18,900 1,579 0.726                                                  1 Emission Factors 2 Weighted Average Emission Factors 74 4.3.5 Land Use Data Land use data was quantified at the Traffic Analysis Zone (TAZ)’s level by combining information from census, BC Assessment, Canada Business Points, and enrolment counts from Central Okanagan School District (SD23), among other sources. Density, diversity, and design indicators (also known as Cervero’s 3D) were considered as the land use variables in this research. The variables corresponding to each indicator are as follows:   Density indicators included activity density and school job density in each neighborhood of the City of Kelowna. Activity density was measured by combining the density of population and jobs within a specific neighborhood. The descriptive statistics of density indicators are shown in Table 4-7.   Diversity indicators indicate the mixed use of land within a specific area. Mixed land use was measured in terms of entropy (a measure of the homogeneity of an area). Entropy for diversity indicators was calculated according to Equation (4-1)  (Zhang et al., 2012; Zhang, 2004). The value of entropy lies between zero and one, with higher values indicating a better balance of mixed land use. Different entropies were calculated to capture the effect of mixed land use (residential/non-residential) and mixed employment (retail/non-retail) at both trip origins and destinations. The descriptive statistics of diversity parameters are shown in Table 4-7.   Entropy = −∑ [ Pj × ln Pj  ]jln J (4-1)  Where, Pj is the proportion of land development of the jth type and J is the number of different land use types.    Design indicators included availability of bus stops, sidewalks, and bike paths within suitable proximity (400 m) from trip origins and destinations. These indicators provide an idea about the transit bus availability and the infrastructure availability for active transportation. 75 Table 4-7 Characteristics of Built Environment and Land Use Variables Variables Commuting Trips Non-commuting Trips Mean Median Standard Deviation Mean Median Standard Deviation Population density at origin (person/hectare) 10.26 9.37 4.71 11.31 14.12 4.65 Population density at destination (person/hectare) 10.42 9.37 4.70 11.23 14.11 4.67 Employment density at origin (person/hectare) 12.06 6.87 11.29 14.66 6.86 11.86 Employment density at destination (person/hectare) 12.60 6.87 11.28 14.38 6.87 11.90 Activity density at origin (person/hectare) 25.97 20.99 15.93 25.97 20.99 15.93 Activity density at destination (person/hectare) 25.61 20.99 16.00 25.61 20.99 16.00 School job density at origin (person/hectare) 0.18 0.13 0.10 0.18 0.13 0.10 School job density at destination (person/hectare) 0.18 0.13 0.10 0.18 0.13 0.10 Mixed land use at origin (entropy: residential, and jobs) 0.90 0.94 0.11 0.90 0.94 0.11 Mixed land use at destination (entropy: residential, and jobs) 0.90 0.94 0.11 0.90 0.94 0.11 Mixed employment at origin (entropy: retail, non-retail, and school jobs) 0.57 0.52 0.10 0.57 0.52 0.09 Mixed employment at destination  (entropy: retail, non-retail, and school jobs) 0.57 0.52 0.1 0.57 0.52 0.09     76 4.4 Summary This chapter described the study area and the different sources of data used in this research. The City of Kelowna was selected as the study area for this investigation. This chapter also described population growth projection of the city and its transportation networks including road, active transportation, and public transportation networks. Further, this chapter thoroughly described the data from household travel survey and emissions inventories. In addition, land use data was quantified at the Traffic Analysis Zone (TAZ)’s level by combining different sources, such as census, BC Assessment, Canada Business Points, and enrolment counts from Central Okanagan School District (SD23).   77 Chapter 5: Modelling Results 5.1 Outline This chapter presents the methodology to apply, and results of applying, TRIBUTE to explore the impacts of land use on passenger transportation GHG emissions in the City of Kelowna as a case study. The following sections of this chapter are arranged as follows: generating level-of-service attributes for the individual trips recorded in the HHTS is presented in Section 5.2. The relationship between travel time and travel distance is modelled in Section 5.3. Modelling the base case scenario using existing land use and built environment attributes is presented in Section 5.4. Different future scenarios are modelled in Section 5.5. After applying TRIBUTE, expected GHG emissions from future scenarios are presented in Section 5.6.   5.2 Generating Level-of-Service Attributes To develop the mode choice component of TRIBUTE, it was necessary to generate the Level-of-Service (LOS) attributes (i.e. travel time and travel distance) for each mode of travel for each trip, regardless of the actual chosen mode. In this study, LOS attributes were generated for car, transit, walk, and cycle options using the Google Directions API with the help of origin and destination postal codes collected from HHTS (Idris, 2013).   5.3 Relationship between Travel Distance and Travel Time  Link travel time is a function of traffic volume on that link (Spiess, 1990). Mathematically, link travel time can be estimated using a volume-delay function. A volume-delay function reflects the congested link travel time by taking into account the ratio of the total traffic volume to the maximum capacity of that link.  The most commonly used volume-delay function is the Bureau of Public Roads (BPR) function, as shown in Equation (5-1) (Spiess, 1990):    𝑡𝐵𝑃𝑅(𝑣) = 𝑡𝑜[1 + (𝑣𝑐)𝛼], (5-1)  where, to is free flow travel time, v is traffic volume, c is the capacity of the road. The higher value of α represents congestion effects.   78 A volume-delay function is applicable where a detailed network data is available. The developed emissions model (TRIBUTE, as described in Chapter 3) does not deal with link volumes and does not require a transportation network model as it mainly relies on HHTS data. As such, it was necessary to translate the reduction in VKT (i.e. travel distance), in response to land use policies, in terms of travel time to feed in the mode choice model.  Modelling the relationship between travel distance and travel time using linear regression is not uncommon in the literature. Harmel and Janda (2015) studied the correlation between straight-line travel distance and travel time in upstate New York. The developed models showed a linear relationship between straight-line travel distance and travel time with a goodness of fit (R-squared) range of 0.685 to 0.973.  In this research, the relationship between travel distance and travel time was modelled using linear regression, described in Section 2.6.2.1. Different functional forms (linear and non-linear) for the relationship between travel distance and travel time were examined before selecting the linear functional form as the best fit. Figure 5-1 and Figure 5-2 show the scatter plot of the relationship between travel distance and travel time along with the developed models for car and transit, respectively. The detailed models are shown in Table 5-1. To check the constant variance, residual plots of the developed regression models are shown in Appendix D.  79  Figure 5-1 Relationships between Travel Distance and Travel Time for Car   Figure 5-2 Relationships between Travel Distance and Travel Time for Transit  80 Table 5-1 Relationships between Travel Distance and Travel Time    Travel Time (Minutes) Car Transit  R2 0.96 0.75 Standard Error 0.87 5.81 Variable Parameter Parameter Intercept 1.42 (84.37)* 8.95 (17.40)* Travel Distance (Km) 1.21 (472.01)* 2.33 (36.58)* *Values between parentheses shows the t-statistics value  5.3.1 Error Analysis To assess the predictability of the developed models, the standard error of the regression analysis was used to generate a prediction interval (PI) based on 95% confidence level. This indicates that the predicted value (travel time) of 95% of the sample (travel distance) falls within the interval around the linear regression (shown in Table 5-1). The prediction interval for the developed model was calculated by using Equation (2-21). The ranges of the prediction intervals were ±25% and ±45% for the developed linear regression models for car and transit, respectively. The estimated error envelope for the developed models for car and transit (shown in Table 5-1) are shown in Figure 5-3 and Figure 5-4, respectively.   (𝑇?̂?Car − 𝑇?̂?Car × 25%)  ≤ PI of 𝑇?̂?Car ≤ (𝑇?̂?Car + 𝑇?̂?Car × 25%), (5-2)   (𝑇?̂?Transit − 𝑇?̂?Transit × 45%)  ≤ PI of 𝑃𝐾?̂?Transit ≤ (𝑇?̂?Transit + 𝑇?̂?Transit × 45%), (5-3)  81  Figure 5-3 Error Envelope of Travel Time and Travel Distance Estimation for Car   Figure 5-4 Error Envelope of Travel Time and Travel Distance Estimation for Transit 82 5.4 Passenger Kilometre Travelled (PKT) Estimation Model As discussed earlier, VKT reduction can be attributed to two main factors: mode shift and trip length reduction (as shown in Figure 3-1). As such, PKT estimation models were developed for four travel options: car, transit, walk, and cycle to capture trip length reduction in response to policy changes. Existing land use and built environment attributes were used to develop PKT models to demonstrate their effect on distance travelled by each passenger. SPSS computer software was used in developing PKT estimation models.  5.4.1 Dependent Variables PKT was the only dependent variable of the developed models to estimate distance travelled by different travel options. Actual PKT values were used in model development. However, the actual values did not satisfy homoscedasticity, i.e. the constant variance assumption. Therefore, to fix this problem, different transformations were tried and finally, the square root transformation satisfied homoscedasticity. Furthermore, it was kept in mind that the PKT output of the model needed to be squared to get actual predicted PKT values.  5.4.2 Explanatory Variables Different land use and built environment attributes such as, density and diversity were considered as explanatory variables in the models. Activity density, sum of population density and employment density, was included in the model as the density indicators. On the other hand, mixed use of land and balanced employment types (as described in Section 4.3.5) were considered as the diversity indicators.   The explanatory variables were chosen in such a way to minimize the effect of multicollinearity. Multicollinearity assumption was verified by using Pearson’s correlation coefficient on SPSS software. According to Gerstman (2016), the correlation coefficient can be classified as weak, moderate, and strong based on the range of less than 0.3, 0.3 to 0.7, and greater than 0.7, respectively. In this research, the threshold for the Pearson’s correlation coefficient was taken as less than 0.3.   83 5.4.3 Empirical Analysis Using ordinary least square analysis, PKT estimation models were developed by incorporating existing built environment and land use variables, as shown in Table 5-2. Although the models showed low goodness of fit, it is not uncommon to attain such R2 value given the inherently high amount of unexplainable variability in the dataset (Cervero and Kockelman, 1997; Kockelman, 1995; Kockelman, 1996). For example, Kockelman (1997) studied the influence of various land use parameters, namely, accessibility, mixed land use, and land use balance on household VKT in the San Francisco Bay area. The author developed a household-based VKT model using multiple regressions analysis. The goodness of fit of the developed regression models were very low (R2= 0.04 ~ 0.15) because of the large number of unexplained variability in the data.   Table 5-2 PKT Estimation Models (Square Root of PKT)   Car Transit Walk Cycle R2 0.137 0.178 0.119 0.104 Standard Error 0.79 0.80 0.83 0.87 Variable Parameter Parameter Parameter Parameter Intercept 6.13 (42.35)* 7.99 (9.90)* 5.79 (42.34)* 5.77 (40.27)* Density Indicators Activity density at origin (Sum of population and job density) -0.012 (-20.34)* -0.014 (-5.46)* -0.011 (-20.75)* -0.011 (-19.51)* Activity density at destination (Sum of population and job density) -0.011 (-18.55)* -0.018 (-7.23)* -0.011 (-19.11)* -0.011 (-17.82)* *Values within parentheses shows the t-statistics value       84 Table 5-2 PKT Estimation Models (Square Root of PKT) (Continued)   Car Transit Walk Cycle R2 0.137 0.178 0.119 0.104 Standard Error 0.79 0.80 0.83 0.87 Variable Parameter Parameter Parameter Parameter Intercept 6.13 (42.35)* 7.99 (9.90)* 5.79 (42.34)* 5.77 (40.27)* Diversity Indicators Mixed land use at origin (Entropy: residential, and jobs) -0.879 (-10.03)* -1.547 (-3.59)* -0.791 (-9.28)* -0.71 (-7.94)* Mixed land use at destination (Entropy: residential, and jobs) -1.060 (-12.09)* -0.553 (-13.33)* -0.898 (-10.53)* -0.836 (-9.35)* Mixed employment at origin (Entropy: retail, non-retail, and school jobs) -1.430 (-14.62)* -2.073 (-4.76)* -1.507 (-15.85)* -1.577 (-15.84)* Mixed employment at destination (Entropy: retail, non-retail, and school jobs) -1.386 (-14.19)* -2.220 (-5.08)* -1.440 (-15.15)* -1.431 (-14.37)* *Values within parentheses shows the t-statistics value  The modeling results, as shown in Table 5-2, indicated that the PKT has an inverse relationship with the built environment and land use indicators. The bullets below presents a detailed discussion on the developed models.   Activity density demonstrated negative influence on PKT for all travel options. In particular, the parameter value of activity density at trip origin (-0.012) was higher than that at the destination (-0.011) for the car option. Although the effect was very small, density at the origin was more important than at the destination. However, the transit option showed the opposite result of the car. For the transit option, density at the destination was more dominant than that at the origin. On the other hand, density showed the same effect for active transportation (i.e. walk/cycle) users at both origin and destination.   85  From the Table 5-2, it can also be said that the diversity indicators have a dominant effect on the PKT over density. The parameter values of the model for all travel modes were much higher than density. In particular, mixed use of land has a negative parameter value both at origin (-0.879) and destination (-1.060) for car travel modes. In addition, mixed use of land was more influential for transit, which has a parameter value -1.547 at origin. Furthermore, the PKT was greatly influenced by the balanced use of employment among different job categories. In general, diversity indicators showed higher PKT reduction for transit than that of other modes of travel.   The summary of the impacts of land use and built environment indicators on PKT is showing in Table 5-3. In general, land use indicators showed a negative impact on PKT. This also indicated that the increase in activity density, mixed use of land, and balanced use of different job categories would reduce the distance travelled by each passenger, and subsequently reduce VKT.  Table 5-3 Resultant Impacts of Land Use and Built Environment on PKT Variables Resultant Impacts on PKT for All Modes of Transportation (e.g. car, transit, walk, cycle)  Activity density (person/acre) Negative Mixed land use (Entropy: residential, and jobs) Negative Mixed employment (Entropy: retail, non-retail, and school jobs) Negative  5.4.4 Model Prediction Error To generate the prediction error envelope for the developed PKT estimation models, the same procedure was followed as described in Section 5.3.1. The range of the prediction error was 18% to 38% based on the different PKT estimation models for different travel modes. In particular, the PKT model for the transit travel option showed a smaller interval (±18%), while the PKT model for walk and cycle demonstrated larger prediction intervals (±38% and ±37%, respectively). Due to the disperse PKT values of walk and cycle options, the prediction interval was larger for walk and cycle. However, the prediction interval for the car fell between -24% and 86 +24%. The error envelopes for the developed PKT models are shown in equations (5-5) through (5-8):   (𝑃𝐾?̂?Car − 𝑃𝐾?̂?Car × 24%)  ≤ PI of 𝑃𝐾?̂?Car ≤ (𝑃𝐾?̂?Car + 𝑃𝐾?̂?Car × 24%),  (5-4)       (𝑃𝐾?̂?Transit − 𝑃𝐾?̂?Transit × 18%)  ≤ PI of 𝑃𝐾?̂?Transit≤ (𝑃𝐾?̂?Transit + 𝑃𝐾?̂?Transit × 18%), (5-5)     (𝑃𝐾?̂?Walk − 𝑃𝐾?̂?Walk × 38%)  ≤ PI of 𝑃𝐾?̂?walk ≤ (𝑃𝐾?̂?Walk + 𝑃𝐾?̂?Walk × 38%), (5-6)     (𝑃𝐾?̂?Cycle − 𝑃𝐾?̂?Cycle × 37%)  ≤ PI of 𝑃𝐾?̂?Cycle ≤ (𝑃𝐾?̂?Cycle + 𝑃𝐾?̂?Cycle × 37%), (5-7)  where, 𝑃𝐾?̂? indicates the predicted PKT value of the model.  5.5 Base Case Estimates 5.5.1 Mode Choice Modelling To calculate the base case estimates, the development of the mode choice model (as described in Section 3.2.1) was imperative. Separate mode choice models were developed for both commuting and non-commuting trips given the different choice behaviour associated with each of these trip purposes. The models were developed using land use and built environment indicators and transportation level-of-service attributes (e.g. travel time, travel distance, etc.), as discussed below.   5.5.1.1 Modes of Travel As mentioned earlier, two types of trip purposes were considered in this investigation, commuting and non-commuting trips. For commuting trips, a choice set of seven modes was used, including (1) car driver, (2) car passenger, (3) transit bus, (4) school bus, (5) walking, (6) cycling, and (7) others. For non-commuting trips, the school bus was excluded from the choice set given its irrelevance. The choice sets for the developed mode choice models are shown in Figure 5-5. 87       (a) Choice Set for Commuting Trips       (b) Choice Set for Non-commuting Trips Figure 5-5 Choice Set for Mode Choice Models  The availability of the car driver option in the choice set was assumed based on driver’s license holding and car ownership. If a person has a car as well as a driver’s licence, then car was added to the model as an available option. However, missing either a car or a licence removed the car as an option from the choice set. On the other hand, car passenger and public transit were assumed available to all travelers.   The mode choice models also dealt with active transportation. From the HHTS and generated LOS attributes (as described in Section 5.2), the cumulative percentage of commuting and non-commuting trips were plotted with respect to the walk distance and cycle distance, as shown in Figure 5-6 and Figure 5-7, respectively. To get the available active transportation options, trip distance for 95% of trips was taken as a threshold. From Figure 5-6, it is seen that 95% of commuting walk trips were within 3 km, while 95% of commuting cycle trips were within 7 km. In the development of the mode choice model for commuting trips, trip distances of up to 3 km and 7 km were considered as having potential to attract walk and cycle trips, respectively.   Choice Set Car Driver Car Passenger Transit School Bus Walk Cycle Other Choice Set Car Driver Car Passenger Transit Walk Cycle Other 88 In addition, the same threshold was applied for non-commuting trips to identify the travel distance for walk and cycle. For non-commuting trips, 95% of the walking trips were less than 3 km, while 95% of cycle trips were within 9 km. Therefore, the availability of walk and cycle as travel options in the mode choice model for non-commuting trips was determined based on 3 km and 9 km thresholds, respectively.    Figure 5-6 Non-Motorized Trip Distance in Commuting Trips  89  Figure 5-7 Non-Motorized Trip Distance in Non-Commuting Trips   5.5.1.2 Empirical Analysis The inclusion of parameters in the developed MNL models was determined based on the following three criteria (Koppelman and Bhat, 2006): a) informal tests, b) goodness of fit measures, and c) statistical tests. The informal tests consist of the sign of the parameters to evaluate reasonableness of the implications. The overall goodness of the fit measures were calculated based on the following equation (Ben-Akiva and Lerman, 1985):   Goodness of fit = 1 − Final loglikelihoodNull loglikelihood (5-8)  where, the null likelihood of the model shows a model with no parameter value considering all alternatives as being equally likely to be chosen. The range of the goodness of the fit varies between 0 to 1, and the closer to the value of 1, the better the fit of the model (Day et al., 2010).  90 An open source software (BIOGEME), which was developed by Bierlaire (2003), was used to developed the MNL models. In terms of statistical significant, the critical value (1.96) of the t-statistic with a 95% confidence limit was considered as the threshold value of considering variables in the model. A series of specifications was tested and final specifications were reached based on the accommodation of variables with proper signs, overall goodness of fit, and statistical significance, at a 95% level of confidence test. However, some parameters with t-stat values lower than 1.96 were retained in the model because the corresponding variables provide considerable insight into the behavioural process, shown in Table 5-4. According to (Koppelman and Bhat, 2006), a low t-statistic value does not require removing a variable from the model if the variable has a strong reason to be included and the sign of its parameter is correct.  Moreover, the inclusion of the insignificant parameters may improve the overall predictability of the model (Cervero, 2002). According to McFadden, Rho-squared values between 0.2 and 0.4 indicate excellent fit for discrete mode choice models (McFadden, 1977). Given the Rho-squared values of 0.443 and 0.524 for commuting and non-commuting trips, respectively, shown in Table 5-4, the developed models have acceptable goodness of fit and explanatory power. In addition, significance test of the model was tested by calculating confidence interval for parameters (Hosmer and Lemeshow, 2000), shown in Equation (5-9) and Equation (5-10).   A 100(1-α)% confidence interval for the slope coefficients:   β1̂ ± z1−α/2 SÊ(β1̂), (5-9)  And for the intercept:   β0̂ ± z1−α/2 SÊ(β0̂), (5-10)  where is the upper 100(1- α/2)% point from the standard normal distribution and SÊ(∙) denotes standard error of the respective parameter estimator.    91 Table 5-4 Mode Choice Models    Commuting Trips Non-Commuting Trips Null Log-Likelihood -8,299.307 -13,785.797 Final Log-Likelihood -4,624.26 -6,565.491 Rho-Squared 0.443 0.524 Variable Mode Parameter Std. Error t-Stat Parameter Std. Error t-Stat Alternative Specific Constant Auto 5.910 0.392 15.31 7.000 0.471 14.87 Auto Passenger 7.710 0.560 14.72 5.170 0.512 10.04 Transit Bus -4.660 1.240 -4.48 -0.303 1.180 -0.26 School Bus 1.860 0.229 8.14 No school bus Walk 5.890 0.749 7.91 5.680 0.502 11.30 Bike 3.040 0.854 4.98 3.110 0.326 9.53 Other Fixed Fixed Density Indicators Activity density at origin (Person/Acre) Auto Passenger -0.0115 0.00395 -2.39 -0.004 0.002 -1.75 Activity density at destination (Person/Acre) Auto Passenger -0.0236 0.00386 -6.05 -0.00316 0.0022 -1.44 Transit Bus -  - 0.00865 0.00564 1.53 School job density at origin (Job/Acre) Auto -0.926 0.365 -2.53 -  - Auto Passenger -  - -0.894 0.401 -2.23 Walk 0.019 1.000 0.02 0.110 0.685 0.16 Bike 0.167 0.690 0.24 0.371 0.807 0.46 School job density at destination (Job/Acre) Auto -0.470 0.559 -0.86 -  - Auto Passenger -  - -0.374 0.428 -0.87 Walk -  - 0.256 0.722 0.35 Bike -  - 0.652 0.833 0.78 Bold text indicates statistically significant values at the 95% confidence level (p<0.05)    92 Table 5-4 Mode Choice Models (Continued)    Commuting Trips Non-Commuting Trips Null Log-Likelihood -8,299.307 -13,785.797 Final Log-Likelihood -4,624.26 -6,565.491 Rho-Squared 0.443 0.524 Variable Mode Parameter Std. Error t-Stat Parameter Std. Error t-Stat Diversity Indicators Mixed land use at origin (Entropy: residential, and jobs) Auto Passenger -1.100 0.454 -2.08 -  - Transit Bus 4.360 0.888 4.98 1.210 0.941 1.28 Mixed land use at destination (Entropy: residential, and jobs) Auto Passenger -1.530 0.536 -2.67 -  - Transit Bus 3.320 0.881 5.00 0.186 0.904 0.21 Bike 1.810 0.867 2.39 -  - Mixed employment at origin (Entropy: retail, non-retail, and school jobs) Auto -  - -0.764 0.400 -1.91 Walk 0.270 1.600 0.17 -  - Mixed employment at destination (Entropy: retail, non-retail, and school jobs) Auto -0.482 0.658 -0.75 -0.828 0.611 -1.36 Auto Passenger -  - -0.651 0.717 -0.91 Walk 0.848 1.250 0.68 -  - Bold text indicates statistically significant values at the 95% confidence level (p<0.05)        93 Table 5-4 Mode Choice Models (Continued)   Commuting Trips Non-Commuting Trips Null Log-Likelihood -8,299.307 -13,785.797 Final Log-Likelihood -4,624.26 -6,565.491 Rho-Squared 0.443 0.524 Variable Mode Parameter Std. Error t-Stat Parameter Std. Error t-Stat Design Indicators Bus stop availability at origin (Yes) Auto -  - -0.297 0.129 -2.30 Auto Passenger -0.116 0.120 -0.97 -0.178 0.147 -1.21 Transit Bus 0.668 0.194 3.42 0.415 0.291 1.43 Bus stop availability at destination (Yes) Auto -  - -0.075 0.073 -1.03 Transit Bus 1.010 0.215 4.67 0.862 0.304 2.84 Sidewalk availability at origin (Yes) Walk 0.607 0.427 1.42 0.120 0.373 0.32 Sidewalk availability at destination (Yes) Walk -  - 0.006 0.360 0.02 Level-of-Service Attributes Travel Time (min) Auto -0.063 0.012 -5.20 -0.091 0.020 -4.56 Auto Passenger -0.194 0.016 -12.32 -0.078 0.021 -3.78 Transit Bus -0.005 0.004 -1.24 -0.022 0.006 -3.53 Travel Distance (km) Walk -1.640 0.079 -20.71 -1.170 0.056 -21.02 Bike -0.351 0.031 -11.51 -0.385 0.042 -9.08 Bold text indicates statistically significant values at the 95% confidence level (p<0.05)  The developed mode choice models are sensitive to land use variables such as, density, diversity, and design indicators in addition to LOS attributes. The effects of land use variables are discussed in the following bullets.    Activity density, sum of population density and job density, was considered as one of the density indicators in the mode choice models. Overall, activity density showed a positive impact on transit, but a negative impact on car users. In particular, activity density has a 94 stronger reciprocal relationship with car use for commuting trips than non-commuting trips. In addition, lower parameter values for activity density in non-commuting trips indicates lower effects of density on choosing travel modes. On the other hand, school jobs density showed strong negative impacts on car use for both commuting and non-commuting trips. In addition, school jobs density has a positive sign for active transportation, which indicates that an increase in school jobs density leads to an increase in active transportation modal shares. Furthermore, activity at trip destination was more important than that at its origin, whereas school job density showed more influence on promoting active transportation at trip origin than at trip destination.   Diversity indicators were measured in terms of mix use of land (entropy) calculation. The developed mode choice model was sensitive to two types of diversity indicators: mixed use of land in terms of population and employment, and mixed use of jobs in terms of retail, non-retail, and school jobs. The parameter values associated with the diversity indicators were very high for the transit options. On the other hand, diversity indicators showed a negative sign for car users and the parameter values were higher than the density indicators. From   Table 5-4 5-4 it can be seen that the mixed use of land was more important in terms of increasing public transit and active transportation than that of density indicators. Furthermore, diverse use of land at origin was more important than at destination for commuting trips. However, non-commuting trips showed mixed use of land at destination was more important than at origin.    In this investigation, the availability of bus stops and sidewalks at both origin and destination were considered as design indicators. The availability of bus stops demonstrated positive parameter value for transit and negative parameter value for car use indicating that the availability of bus stops is important to promote public transit. However, the lower parameter value for availability of bus stops at origin indicated the bus stop at destination was more important than bus stop at origin. On the other hand, 95 availability of sidewalks was associated with the positive parameter value for the walk option, and has higher parameter value at origin than destination.  The fourth explanatory variable of the mode choice model was modal attributes that include travel time for motorized travel modes and travel distances for non-motorized travel modes. The developed mode choice models showed a negative parameter value for modal attributes. In addition, the walking option was affected more by the increase in travel distance between origins and destinations.  The summary of the impacts of land use and built environment indicators on choosing travel modes is shown in Table 5-5. In general, there was a negative influence of land use indicators on using the car option. Conversely, a positive relationship was observed for transit and active transportation. This indicates that the increase in density and better mix of land use would promote transit and active transportation at the expense of car use.   Table 5-5 Resultant Impacts of Land Use and Built Environment on Mode Choice Variables Resultant Impact on Different Modes Cars Transit Active Transportation Activity density (person/acre) Negative Positive Positive School job density (job/acre) Negative Positive Positive Mixed land use (Entropy: residential, and jobs) Negative Positive Positive Mixed employment (Entropy: retail, non-retail, and school jobs) Negative Positive Positive Bus stop availability (yes/no) Negative Positive Positive Sidewalk availability (yes/no) Negative Positive Positive     96 5.5.1.3 Forecasting Performance In this study, the entire dataset was used for model estimation. As such, market segmented subsets were used to validate the developed mode choice models and measure their forecasting performance. In particular, two subsets were created based on vehicle ownership (including 0 vehicle, 1 vehicle, 2 vehicles, and 3+ vehicles) and geographic location (including trips originated in central city and trips originated outside central city), as shown in Table 5-6 and Table 5-7, respectively.   The Forecasting Performance Measure (FPM) was quantified for both subsets. For vehicle ownership segmentation, FPM values were between 0.566 and 1.617 (Shown in Table 5-6). The aggregate forecasting error of the developed models was smaller (0.566) for single vehicle availability than multi vehicles availability (1.617 and 1.314). On the other hand, the geographic location segmentation showed even smaller (0.029 and 0.186) FPM value than the market segment subset of vehicle ownership. Idris et al. (2015) studied the mode shift model forecasting performance of five different developed mode choice models, where the FPM values were between 0.003 and 1.974. Given the FPM values between 0.566 and 1.617 for vehicle ownership segment, and between 0.029 and 0.186 for geographic location segment, it can be concluded that the developed models have acceptable forecasting power.           97 Table 5-6 FPM of the Developed Mode Choice Models based on Vehicle Ownership Segmentation Mode Alternative Observed Mode Choice Predicted Mode Choice Mode choice difference (Predicted - Observed) FPM 0 Vehicle Car Driver 6.99% 12.62% 5.63% 1.346 Car Passenger 12.05% 10.37% -1.68% Transit Bus 38.80% 35.78% -3.02% School Bus 0.00% 0.00% 0.00% Walk 35.18% 36.33% 1.15% Cycle 4.82% 4.50% -0.32% Other 2.17% 0.40% -1.77% 1 Vehicle Car Driver 59.56% 65.21% 5.65% 0.566 Car Passenger 17.14% 14.56% -2.58% Transit Bus 5.01% 4.45% -0.56% School Bus 0.76% 0.96% 0.20% Walk 12.21% 10.73% -1.48% Cycle 5.04% 3.64% -1.40% Other 0.28% 0.45% 0.17% 2 Vehicles Car Driver 71.30% 66.70% -4.60% 1.617 Car Passenger 15.12% 15.49% 0.37% Transit Bus 2.32% 4.26% 1.93% School Bus 1.56% 1.23% -0.33% Walk 6.75% 8.49% 1.74% Cycle 2.67% 3.33% 0.66% Other 0.27% 0.51% 0.24% 3+ Vehicles Car Driver 74.95% 68.03% -6.92% 1.314 Car Passenger 12.84% 15.22% 2.38% Transit Bus 2.35% 4.53% 2.18% School Bus 1.26% 1.50% 0.23% Walk 5.26% 7.03% 1.77% Cycle 2.42% 3.17% 0.74% Other 0.91% 0.53% -0.38%   98 Table 5-7 FPM of the Developed Models based on Geographic Location Segmentation Mode Alternative Observed Mode Choice Predicted Mode Choice Mode choice difference (Predicted - Observed) FPM Trips Originated in Central City (Sector 3) Car Driver 66.36% 65.72% -0.64% 0.029 Car Passenger 13.91% 13.80% -0.11% Transit Bus 4.06% 4.65% 0.59% School Bus 0.52% 0.95% 0.43% Walk 10.82% 10.72% -0.10% Cycle 3.85% 3.72% -0.13% Other 0.49% 0.45% -0.04% Trips Originated Outside Central City Car Driver 66.58% 67.76% 1.18% 0.186 Car Passenger 17.22% 17.42% 0.20% Transit Bus 5.02% 3.94% -1.08% School Bus 2.42% 1.62% -0.79% Walk 5.71% 5.89% 0.18% Cycle 2.56% 2.80% 0.24% Other 0.50% 0.57% 0.07%  5.5.2 Estimation and Validation TRIBUTE was utilized by jointly running its mode choice and emissions forecasting models. Community Energy and Emissions Inventory (CEEI) reports were used for estimating and validating the TRIBUTE model. In the CEEI report, the total GHG emissions in 2007 and 2010 were calculated based on fuel consumption and number of vehicles in Kelowna, BC (Ministry of Environment, 2014b).  The mode choice component of TRIBUTE was first used to estimate the proportion of trips made by each mode. Next, the emissions forecasting component of TRIBUTE was used to calculate the total PKT by each mode given respective average modal PKT (calculated from CEEI reports). The total PKT was converted to total VKT. For this task, respective average vehicle occupancy was used (described in Section 4.3.3). Finally, the total GHG emissions from the transportation sector were calculated by multiplying the total VKT by each mode by respective 99 average emissions factors. For model estimation, respective emissions factors were calculated from the CEEI report (described in Section 4.3.4).   For model validation, TRIBUTE was utilized to forecast the 2010 GHG emissions. The forecasted results were then compared to the total emissions reported in the 2010 CEEI report. Table 5-8 shows a comparison between the GHG emissions calculated using TRIBUTE and those reported in the CEEI reports. The validation results showed a very small difference (-0.3%) between the forecasted emissions using TRIBUTE and the reported emissions in 2010 CEEI report.  Table 5-8 TRIBUTE Validation Results  TRIBUTE CEEI Report (Ministry of Environment, 2014b) Year 2007 2010 2007 2010 Car Emissions (CO2e ton) 370,229 362,307 370,229 368,250 Buses Emissions (CO2e ton) 3,224 7,588 3,224 2,280 Other Emissions (CO2e ton) 891 644 891 1,185 Total Emissions (CO2e ton) 374,344 370,539 374,344 371,715 Difference -0.3%  From Table 5-8, it is seen that the difference in total emissions calculation between TRIBUTE and CEEI report was very small (0.3%); emissions from individual modes (car, bus, and other), however, showed higher differences. In particular, TRIBUTE over-predicted bus emissions and under-predicted emissions from cars. This discrepancy was because of changes in the emissions factor in 2010. The emissions factor for buses became higher in 2010 than that of 2007. On the other hand, the car emissions factor went down in 2010 due to the introduction of hybrid cars as well as higher fuel efficiency. In addition to the emissions factor, modal share played an associating role for the discrepancy. Since TRIBUTE considered individual trips in estimating emissions, and transit share was higher in 2010 than 2007, TRIBUTE estimated higher emissions from buses in 2010. In general, TRIBUTE looks promising at estimating total emissions. However, estimated emissions from individual modes involved a higher degree of error.  100 5.6 Future Scenario Development It is worth mentioning that TRIBUTE was used to help the City of Kelowna evaluate and select the best future density scenario to reduce GHG emissions in 2040 (Rahman et al., 2016). However, such real-world exercise is not included in this thesis due to the confidential nature of the project. For illustration, a hypothetical scenario was developed and tested using TRIBUTE considering a better land use mix in comparison to the business-as-usual (BAU) scenario. The assumptions behind the developed scenarios are described below.   Existing Scenario (2013) – The existing scenario was developed based on the existing land use and built environment in the City of Kelowna in 2013. In this scenario, population density, job density, and mixed land use were calculated based on the present conditions.   Business-as-Usual (BAU) Scenario – In the BAU case, population was altered to account for the expected total growth in population from 2013 to 2040. It was also assumed that the total number of jobs will follow the same growth rate as population. All other factors (e.g. distribution of population, land use types, locations, design indicators, etc.) were fixed across the city. This change has resulted in the same land use diversity in the BAU scenario as in the existing scenario.   Hypothetical scenario – Similar to the BAU scenario, population and jobs were altered to account for the expected total growth in population and jobs from 2013 to 2040 in the hypothetical scenario, but with different distribution. In particular, population density was doubled in the urban core of the city and the remaining growth was distributed to the remaining sectors in a similar manner to the BAU. Moreover, employment density was re-allocated to balance the increase in population and maintain higher entropy (i.e. higher mix of land use) in all ten sectors. In addition, different types of jobs (retail, non-retail, and school jobs) were also re-allocated to attain the maximum mix of employment among different sectors in Kelowna. This hypothetical scenario can also be referred to as the Most Aggressive Scenario (MASc).   101 The developed scenarios were designed based on altering land use attributes (i.e. density and  diversity). However, the developed emissions forecasting model, TRIBUTE, is also sensitive to  changes in the transportation network, such as inclusion of high-bike-infrastructure, new transit route, etc. For simplicity, it was assumed that the public transit and active transportation infrastructure would be the same as the existing scenario in all future scenarios. This means that the number of bus stops, sidewalk coverage, bike paths coverage, road network, and transit route coverage of the City of Kelowna will be the same in 2040 as in 2013. This would control for confounding factors, and allow the developed MASc models to be focused on land use planning policies. Mixed land use and mixed employment in the BAU and hypothetical scenarios are shown in Figure 5-8 and Figure 5-9, respectively. 102  Figure 5-8 Mixed Land Use Entropy in BAU and Hypothetical Scenario 0.9110.9820.8530.7660.9910.7760.4710.9380.9120.80810.8840.9970.9010.7930.9480.8140.9370.9330.837Mixed Land Use Entropy in Each Sector under Different ScenariosExisting Scenario and Business as Usual Scenario  Hypothetical Scenario0 2 4 6 81KilometersµOkanaganLakeOkanaganLakeMd. Nobinur Rahman                  Data Source: City of KelownaApril 2016                                    Coordinate System: NAD 1983 UTM Zone 11NLegendMixed Land Use Entropy0.00 - 0.4500.451 - 0.8700.871 - 0.9300.931 - 0.9850.986 - 1.000OkanaganLakeOkanaganLake103  Figure 5-9 Mixed Employment Entropy in BAU and Hypothetical Scenario0.360.450.510.340.80.650.480.520.680.670.440.80.560.360.750.560.860.70.840.69Mixed Employment Entropy in Each Sector under Different ScenariosExisting Scenario and Business as Usual Scenario  Hypothetical Scenario0 2 4 6 81KilometersµOkanaganLakeOkanaganLakeMd. Nobinur Rahman                  Data Source: City of KelownaApril 2016                                    Coordinate System: NAD 1983 UTM Zone 11NLegendMixed Employment Entropy0.00 - 0.400.41 - 0.550.56 - 0.750.76 - 0.900.91 - 1.00OkanaganLakeOkanaganLake104 5.7 Scenario Estimates After validating the model (as described in Section 5.5.2), TRIBUTE was then used to forecast GHG emissions from the transportation sector in 2040 for three scenarios: the existing scenario, the business-as-usual (BAU), and a hypothetical future scenario (described in 5.6). The emissions forecasting steps for the future scenarios are discussed below.   5.7.1 Reduction in PKT To calculate the reduction in PKT, the developed PKT estimation models were used for each scenario (as described in Section 5.4). From Figure 5-10, it is seen that the hypothetical scenario was responsible for a higher reduction in PKT for all modes of travel. The hypothetical scenario has been developed based on the higher density and higher mix of land use, and the effect of them on PKT reduction was higher (Figure 5-10). On the other hand, the unbalanced distribution of population and employment densities in the BAU scenario led to a lower reduction in PKT (Figure 5-10). The only contributing factor to PKT reduction in the latter case was diversity in terms of both population and job mix as well as a proper mix of different types of jobs. The calculated reduction in PKT was applied to the level-of-service attributes (i.e. travel time for motorized modes and travel distance for non-motorized modes) before running TRIBUTE’s mode shift component. Based on the error analysis discussed in Section 5.5.1.3, the prediction range of PKT is given in Table 5-9.  105  Figure 5-10 Estimated PKT Reduction  Table 5-9 Reduction Interval in PKT   Car Transit Walk Cycle Business as Usual Highest Reduction -23.64% -23.97% -26.49% -25.38% Estimated Reduction -19.04% -20.28% -19.23% -18.54% Lowest Reduction -14.44% -16.59% -11.97% -11.70% Hypothetical Scenario Highest Reduction -62.22% -62.69% -71.28% -68.63% Estimated Reduction -50.11% -53.04% -51.75% -50.14% Lowest Reduction -38.00% -43.39% -32.21% -31.65%    106 5.7.2 Modal Shift The mode shift component of TRIBUTE was used to calculate modal shift in the BAU and hypothetical scenarios, as shown in Figure 5-11 and Table 5-10, respectively. The results showed that the modal share of car in the BAU scenario was lower and walk share was higher than that of the existing scenario, given the increase in population density and job density in the BAU scenario. In addition, car share was further reduced in the hypothetical scenario because of the proper distribution of population and job densities. Walking, in the hypothetical scenario, was also increased by almost three times compared to the existing scenario due to the shorter distances between trip origins and destinations, because of better density and diversity. In addition, transit share was slightly increased in the BAU scenario as the population and job density increase regardless of their proper distribution. Furthermore, due to the balanced distribution of population and jobs in the hypothetical scenario, the transit share was also increased by almost one and half times compared to that of the existing scenario. Since the design indicators were unaltered in both future scenarios, the transit share was only affected by density and diversity. The modal shares in 2007 and 2013 were extracted from the 2007 and 2013 Okanagan HHTS, respectively.   107  Figure 5-11 Modal Shift (Estimated)  Table 5-10 Modal share Interval in Different Scenarios   Car Driver Car Passenger Transit School Bus Walk Cycle Others Business as Usual Highest Shift 64.5% 10.0% 7.4% 1.1% 11.4% 5.1% 0.4% Estimated Shift 66.1% 12.8% 4.9% 1.3% 10.2% 4.1% 0.6% Lowest Shift 67.7% 15.5% 2.4% 1.5% 8.9% 3.1% 0.8% Hypothetical Scenario Highest Shift 51.8% 5.5% 8.8% 1.0% 23.1% 9.4% 0.3% Estimated Shift 53.4% 8.3% 6.3% 1.2% 21.8% 8.3% 0.6% Lowest Shift 55.1% 11.0% 3.9% 1.4% 20.6% 7.3% 0.8%  108 5.7.3 Forecasting GHG Emissions With the observed reduction in PKT and modal shift, the GHG emissions were forecast for all future density scenarios based on the methodology described in Section 3.2. Figure 5-12 shows the forecast GHG emissions from urban transportation in the City of Kelowna in the existing scenario (2013), as well as the two future scenarios for 2040; one for the BAU case and the other for the hypothetical scenario. Figure 5-12 also shows 2007 and 2010 GHG emissions that were used for model estimation and validation, respectively. By using the existing PKT values (generated in Section 5.4) and the observed mode share (from HHTS), the emissions from road transportation were calculated in 2013 (existing scenario). The difference between the estimated emissions by TRIBUTE and from inventories (shown in Table 5-8) was used to calculate the emissions estimation envelope for 2013. After 2010, there was an increasing trend in total GHG emissions because of population growth. Although population growth and car share were higher in 2013 than in 2010, total car users were higher in 2013 than in 2010. Since cars were primarily responsible for the majority of emissions, the total emissions in 2013 increased again from 2010. However, it is worth mentioning that the emissions factors were kept constant for forecasting emissions in 2040, but it can be changed to respond to new fuel technology, new travel modes (electric vehicle, telecommunication, etc.).  Most importantly, TRIBUTE can capture those changes in forecasting emissions by looking new emission factors and proportion of switched trips towards new travel options. By combining error envelopes in PKT reduction (described in Section 5.4.4) and modal shift (described in Section 5.7.2), the prediction error was calculated in estimating GHG emissions in future scenarios, as shown in Table 5-11. The BAU scenario resulted in 11.1% to 30.8% increase in emissions above the 2007 levels. However, the hypothetical scenario showed a 22.8% to 55.6% reduction in emissions below the 2007 levels.   109  Figure 5-12 Forecasted GHG Emissions in the City of Kelowna  Table 5-11 Emissions Estimation Window in Different Scenarios   Emissions (CO2 eq. ton) Difference from Base Level (%) Emissions Per Capita 2013 (Existing Scenario) Upper Limit 379,715 1.4% 3.17 Estimated 378,521 1.1% 3.16 Lower Limit 377,327 0.8% 3.15 Business as Usual Upper Limit 489,460 30.8% 2.71 Estimated 452,145 20.8% 2.51 Lower Limit 415,914 11.1% 2.31 Hypothetical Scenario Upper Limit 288,836 -22.8% 1.60 Estimated 225,695 -39.7% 1.25 Lower Limit 166,191 -55.6% 0.92  110 In particular, there was a 39.7% reduction in emissions below 2007 levels in the hypothetical scenario, while there was a 20.8% increase in emissions above 2007 levels in the BAU scenario (Figure 5-132). Although the BAU scenario was projected to achieve a 20.8% increase in total GHG emissions by 2040, the per capita emissions were expected to be lower than the 2007 levels. Figure 5-13 shows the per capita GHG emissions for all scenarios. The hypothetical scenario showed the lower per capita emissions in 2040 and the ranges of per capita emissions were between 1.60 and 0.92 CO2 eq. ton per person. On the other hand, per capita emissions in BAU were between 2.71 and 2.31 CO2 eq. ton per person.   Figure 5-13 Per Capita GHG Emissions  5.8 Summary This chapter demonstrated the applicability of TRIBUTE in estimating and forecasting GHG emissions under different land use policy scenarios. The validation results showed a very small difference between TRIBUTE forecasts and the emissions inventories which gave more confidence in using TRIBUTE for estimating emissions for future scenarios. The business-as-usual, do-nothing scenario, shows higher emissions from on-road transportation than the 111 hypothetical scenario. Since the hypothetical scenario is developed considering proper mix of land use and high density, it shows a higher modal shift to public transportation and active transportation at the expense of car than BAU. In addition, there is a higher reduction in trip length, due to the shorter trip distance between origin and destination, than BAU. Therefore, the combined effect of trip length reduction and modal shift further reduced the emissions in the hypothetical scenario. In light of the above, TRIBUTE can help transportation planners make decisions regarding the implementation of land use and transportation policies that aim to reduce GHG emissions from passenger transportation.   112 Chapter 6: Conclusions and Future Research 6.1 Outline This chapter is organized as follows: summary of the presented research is demonstrated in Section 6.2, followed by the recommendations to the City of Kelowna to reduce emissions from passenger transportation in Section 6.3. Finally, limitation of the presented research and ideas for future continuation of this research are discussed in Section 6.4.  6.2 Conclusions This research introduced a novel TRIp-Based Urban Transportation Emissions (TRIBUTE) model for municipalities. The developed model is comprised of two main components: a discrete mode choice model and an emissions forecasting model. The mode choice model considers personal, modal, and land use information as its input, and the output of this model is the proportion of trips made by each mode. The emissions forecasting model, on the other hand,  estimates and predicts emissions under various land use policies. TRIBUTE can be applied to those areas where extensive transportation network data is not available. TRIBUTE relies mainly on household travel survey and emissions inventories to estimate and validate passenger transportation emissions.   TRIBUTE was applied to the City of Kelowna to find the best alternative  scenario among four different future density scenarios to reduce GHG emissions from the transportation sector, and to assist the city in meeting its emissions reduction target. The City’s actual future density scenarios, however, could not be included in this thesis for confidentiality reasons. Therefore, a hypothetical and a business-as-usual scenario (BAU) were developed to show the applicability of TRIBUTE in estimating and predicting GHG emissions. The different scenarios were designed considering different land use policies, such as density and diversity along with population growth rates. The hypothetical scenario was developed by relocating people and jobs to achieve higher density and a better mix of land uses within a neighborhood. The BAU scenario represented the "do-nothing” scenario, in which the growth of population and job, and their locations will be the same as exists in Kelowna today.   113 TRIBUTE was then validated by using available emissions inventories for the City of Kelowna. The validation results showed that there was a very little difference (0.3%) between the estimated emissions using TRIBUTE and emissions inventories. This result suggests that TRIBUTE is a reliable tool for predicting emissions in future scenarios.   Trip length reduction is directly related to reducing GHG emissions. Therefore, a series of passenger kilometres travelled (PKT) estimation models were developed for car, transit, walk, and cycle options to capture trip length reduction by implementing transportation and land use policies. The results suggested that the distance travelled by each passenger would reduce due to  increased density and balanced mixed use of land. Density parameters showed negative influence in all PKT models for different travel options. Moreover, diversity and mixed use of land were the dominating factors in reducing PKT for transit mode as it has higher parameter values. The goodness of fit of the models, however, were between 0.10 and 0.17 indicating high unexplained variability within the dataset. Therefore, the prediction interval was measured for the developed models and the interval was comprised of 95% of predicted values for given observations. Using the developed PKT models, the reduction in length was measured for both hypothetical MASc and BAU scenarios. Due to the higher density and proper mix of land, the hypothetical scenario showed higher reduction (~50%) in PKT for all travel options (with ±24% to ±37% forecast error). The organic growth of the population reduced the trip length in the BAU scenario as well, which was smaller than in the hypothetical scenario.    Mode shift, switching modes from motorized vehicle to non-motorized vehicles, can also reduce GHG emissions from passenger transportation. Therefore, a discrete mode choice model was developed to estimate modal shift by introducing land use policies. The developed mode choice models showed that density, diversity, and design indicators have negative parameter values for transit and active transportation options. However, the car option was associated with positive parameter values in the utility functions. This indicates that the increase in density and proper mixed use of land will increase transit rider, and walk and cycle trips at the expense of the car option. In addition, improved design indicators, such as availability of bus stops within proximity of a trip maker, will increase transit riders. Further, availability of sidewalks will increase walk trips. The estimated modal share from the developed mode choice model was then compared 114 with the actual modal share from HHTS to calibrate the developed mode choice models. After calibrating and validating, the developed models were further applied to the future scenarios (hypothetical scenario and BAU) to capture the changes in land use parameters on modal share. The results showed that the hypothetical scenario had higher shift towards transit and active transportation than BAU.  By running the PKT estimation model and the mode shift model, TRIBUTE was used to forecast emissions from passenger transportation for both hypothetical and BAU scenarios. The results showed that the hypothetical scenario would be associated with a 22.8% ~ 55.6% reduction in GHG emissions below 2007 levels. On the other hand, the BAU scenario would realize a 11.1% ~ 30.8% increase in GHG emissions from 2007 levels.   Therefore, TRIBUTE was able to test different scenarios and estimate emissions from passenger transport. TRIBUTE can capture the influence of different built environment and land use indicators on reducing emissions. The main advantage of TRIBUTE is the relatively moderate data requirement and ease with which it can be applied. It can be effectively used where extensive data/expertise/transportation models are not available. TRIBUTE can assist municipalities in evaluating alternative policy scenarios and eventually selecting the one(s) that will help them meet future GHG emission targets.  6.3 Recommendations to the City of Kelowna TRIBUTE was applied to the future density scenarios of the City of Kelowna and the best alternative scenario was chosen to meet future GHG emissions targets. Based on the result of the investigation, the following recommendations are made for the City of Kelowna to reduce GHG emissions from passenger transportation.   6.3.1 Increase Mixed Land Use Diversity Mixed land use diversity refers to the proper balance among residential, commercial, and industrial activities within a neighbourhood. The mode choice model component of the developed TRIBUTE model has demonstrated that the proper mix of residential population and employment has a strong effect on transit and non-motorized transportation modal shares. In 115 addition, increasing land use diversity reduces the trip length between origins and destinations and subsequently, reduces the need to travel long distances. Furthermore, it would cause a modal shift towards public transit and active transportation as well as a reduction in trip length, both of which results in reducing GHG emissions from passenger road transportation. Due to these findings, it is recommended that the City of Kelowna should pay proper attention to mixed uses of land in developing future scenarios. The mixed use of land can be achieved by relocating residential locations and employment locations among ten sectors in the City of Kelowna, as well as compacting the urban cores (downtown area) considering both population and employment.  6.3.2 Improve Public Transit Mode shift from single occupancy vehicles to public transit is another way to lower GHG emissions. Public transit can compete with the automobile for intermediate and long trips if appropriate funding and service planning are provided. It is evident in the developed mode choice models that several transit design indicators, such as bus stop availability and travel time, have a strong influence on increasing mode shift to transit at the expense of automobile use. Accordingly, the City of Kelowna should explore opportunities to improve its transit service by providing better infrastructure, increasing frequency, improving access, etc. as part of all future scenarios. This should be complemented by adopting different Transportation Demand Management (TDM) policies to incentivize (disincentivize) transit (car) use.   6.3.3 Promote Active Transportation As per the developed mode choice models, the availability of sidewalks is the most important factor that influences walking modal shares. Walk and cycle can compete with the automobile for short and intermediate trips if better active transportation infrastructure and end-of-trip facilities are provided. In addition, public education about the benefits of switching to transit and active transportation is imperative.  6.4 Limitations and Future Research This section describes limitations and the possible future research that have been identified during this research as follows: 116  Since community energy and emissions inventory 2010 is the most updated inventory in BC until now, it was applied to the validation of TRIBUTE. In addition, the 2013 household travel survey was used in developing TRIBUTE. Therefore, it is necessary to validate TRIBUTE again by integrating BC’s next updated emissions inventories with the most recent HHTS in future research.   This research considered individual trips as a fundamental unit of analysis. However, activity based analysis can be taken into consideration in developing mode choice model in future studies.   The mode choice component of TRIBUTE consists of commuting and non-commuting trips separately. However, a series of mode choice model can be developed by considering other trip purposes such as home-based trips, non-home based trips etc. in future research.   This research mainly focused on the effect of land use parameters, such as density, diversity, and design indicators in reducing emissions. However, socio-economic and demographic variables were not included in the mode choice model and PKT estimation model. Therefore, socio-economic and demographic variables can be added to increase the predictability of the models in future studies.   This research considered various transportation and land use policies in the development of future scenarios. However, safety issues were not included to the scenario development. Therefore, in future research, various safety issues can be analyzed for the developed future scenarios.    TRIBUTE does not (yet) incorporate economic metrics in its scenario development and output measures; adding in economic metrics would provide a tangible means for the City of Kelowna to reflect on and make decisions related to its infrastructure investments and supporting policies in addition to more detailed scenario impact costs and benefits. 117 Bibliography Acker, V. V., and Witlox, F. (2011). Commuting trips within tours: How is commuting related to land use? Transportation, 38(3), 465–486. doi:10.1007/s11116-010-9309-6 Albrecht, J., François, D., and Schoors, K. (2002). A Shapley decomposition of carbon emissions without residuals. Energy Policy, 30(9), 727–736. doi:10.1016/S0301-4215(01)00131-8 Badoe, D. A., and Miller, E. J. (2000). Transportation–land-use interaction: empirical findings in North America, and their implications for modeling. Transportation Research Part D: Transport and Environment, 5(4), 235–263. doi:10.1016/S1361-9209(99)00036-X Banister, D. (1978). The influence of habit formation on modal choice: A Heuristic model. Transportation, 7(1), 5–33. doi:10.1007/BF00148368 Ben-Akiva, M., and Boccara, B. (1995). Discrete choice models with latent choice sets. International Journal of Research in Marketing, 12(1), 9–24. doi:10.1016/0167-8116(95)00002-J Ben-Akiva, M. E., and Lerman, S. R. (1985). Discrete choice analysis: Theory and application to travel demand. Cambridge, Massachusetts: MIT press. Bento, A. M. ., Cropper, M. L. ., Mobarak, A. M. ., and Vinha, K. . (2005). The effects of urban spatial structure on travel demand in the United States. Review of Economics and Statistics, 87(3), 466–478. doi:10.1162/0034653054638292 Bierlaire, M. (2003). BIOGEME: A free package for the estimation of discrete choice models. Paper presented at the 3rd Swiss Transport Research Conference, Monte Verita, Ascona.  Bongaarts, J. (1992). Population growth and global warming. Population and Development Review, 18(2), 299–319. doi:10.2307/1973681 Bristow, A. L., Tight, M., Pridmore, A., and May, A. D. (2008). Developing pathways to low carbon land-based passenger transport in Great Britain by 2050. Energy Policy, 36(9), 3427–3435. doi:10.1016/j.enpol.2008.04.029 Brownstone, D., and Golob, T. F. (2009). The impact of residential density on vehicle usage and energy consumption. Journal of Urban Economics, 65(1), 91–98.    Cervero, R. (2002). Built environments and mode choice: Toward a normative framework. Transportation Research Part D: Transport and Environment, 7(4), 265-284. Cervero, R., and Kockelman, K. (1997). Travel demand and the 3ds : Density, diversity and design. Transportation Research Part D: Transport and Environment, 2(3), 199–219. 118 Cervero, R., and Murakami, J. (2009). Rail and property development in Hong Kong: Experiences and extensions. Urban Studies, 46(10), 2019–2043. doi:10.1177/0042098009339431 Chatman, D. (2003). How density and mixed uses at the workplace affect personal commercial travel and commute mode choice. Transportation Research Record: Journal of the Transportation Research Board, 1831, 193–201.  City of Kelowna. (2016). Retrieved from http://www.kelowna.ca/CM/Page82.aspx (May 20, 2016). City of Kelowna. (2009). Kelowna community greenhouse gas emissions inventory and projections report. Retrieved from http://apps.kelowna.ca/CityPage/Docs/PDFs/ Environment Division/GreenhouseGasEmissionsInventory_2009.pdf City of Kelowna. (2011). 20-year servicing plan & financing strategy. Retrieved from http://apps.kelowna.ca/CityPage/Docs/PDFs/Financial Services Division/20_Year Servicing Plan_2011.pdf City of Kelowna. (2012). Community climate action plan: Working towards a 33% reduction in greenhouse gases. Retrieved from http://www.kelowna.ca/CityPage/Docs/PDFs/ Environment Division/Climate Change/2012-06-12 Climate Action Plan FINAL PUBLIC version reduced.pdf City of Kelowna. (2013). Official community plan: Growth projections (Chapter 3). Retrieved from http://apps.kelowna.ca/CityPage/Docs/PDFs/Bylaws/Official Community Plan 2030 Bylaw No. 10500/Chapter 03 - Growth Projections.pdf Commoner, B., Corr, M., and Stamler, P. J. (1971). The causes of pollution. Environment: Science and Policy for Sustainable Development, 13(3), 2–19. doi:10.1080/00139157.1971.9930577 Corpuz, G., Mccabe, M., and Ryszawa, K. (2005). The development of a Sydney VKT regression model. Paper presented at the 29th Australasian Transport Research Forum, Gold Coast. Davis, S. J., and Caldeira, K. (2010). Consumption-based accounting of CO2 emissions. Proceedings of the National Academy of Sciences of the United States of America, 107(12), 5687–92. doi:10.1073/pnas.0906974107 Day, N., Habib, K. N., and Miller, E. J. (2010). Analysis of work trip timing and mode choice in the greater Toronto area. Canadian Journal of Civil Engineering, 37(5), 695–705. 119 doi:10.1139/L10-008 Derrible, S., Saneinejad, S., Sugar, L., and Kennedy, C. (2010). Macroscopic model of greenhouse gas emissions for municipalities. Transportation Research Record: Journal of the Transportation Research Board, 2191, 174–181. doi:10.3141/2191-22 Ehrlich, P. R., and Holdren, J. P. (1971). Impact of population growth. Science, 171(3977), 1212–1217. doi:10.1007/BF02862495 Environment Canada. (2013). The 2012 progress report of the federal sustainable development strategy. Retrieved from http://www.ec.gc.ca/dd-sd/23E4714E-B774-4CC5-9337-F87B01556727/2012_Progress_Report_ofthe_FSDS.pdf Environment Canada. (2014). National inventory report: greenhouse gas sources and sinks in Canada, 1990-2012. Retrieved from https://ec.gc.ca/ges-ghg/3808457C-9E9E-4AE3-8463-05EE9515D8FE/NIR2014-Exec Sum-Web-Final.pdf Ewing, R., and Cervero, R. (2001). Travel and the built environment: A synthesis. Transportation Research Record, 1780(1), 87–114. doi:10.3141/1780-10 Ewing, R., and Cervero, R. (2010). Travel and the built environment. Journal of the American Planning Association, 76(3), 265–294. doi:10.1080/01944361003766766 Fan, Y. (2007). The built environment, activity space, and time allocation: An activity-based framework for modeling the land use and travel connection (Doctoral dissertation). The University of North Carolina at Chapel Hill. Gerstman, B. B. (2016). StatPrimer (Version 7.0). Retrieved May 20, 2016, from Part B (Intermediate) website: http://www.sjsu.edu/faculty/gerstman/StatPrimer/correlation.pdf  Girod, B., Wiek, A., Mieg, H., and Hulme, M. (2009). The evolution of the IPCC’s emissions scenarios. Environmental Science and Policy, 12(2), 103–118.  Greenwald, M., and Boarnet, M. (2001). Built environment as determinant of walking behavior: Analyzing nonwork pedestrian travel in Portland, Oregon. Transportation Research Record, 1780(1), 33–41. doi:10.3141/1780-05 Habib, K. M. N., Day, N., and Miller, E. J. (2009). An investigation of commuting trip timing and mode choice in the Greater Toronto Area: Application of a joint discrete-continuous model. Transportation Research Part A: Policy and Practice, 43(7), 639–653.  Habib, K. M. N., Swait, J., and Salem, S., (2012). Investigating structural changes in commuting mode choice preferences with repeated cross-sectional travel survey data: The contexts of 120 Greater Toronto and Hamilton (GTHA) Area. Hickman, R., and Banister, D. (2007). Looking over the horizon: Transport and reduced CO2 emissions in the UK by 2030. Transport Policy, 14(5), 377–387.  Holdren, J. P. (1991). Population and the energy problem. Population and Environment, 12(3), 231–255. doi:10.1007/BF01357916 Holtzclaw, J., Clear, R., Dittmar, H., Goldstein, D., and Haas, P. (2002). Location efficiency: Neighborhood and socio-economic characteristics determine auto ownership and use - studies in Chicago, Los Angeles and San Francisco. Transportation Planning and Technology, 25(1), 1–27. doi:10.1080/03081060290032033 Hosmer, D. W., and Lemeshow, S. (2000). Applied logistic regression (2nd ed.). New York: Wiley. Idris, A. (2013). Modal shift forecasting models for transit service planning (Doctoral dissertation). University of Toronto.  Idris, A. O., Habib, K. M. N., and Shalaby, A. (2015). An investigation on the performances of mode shift models in transit ridership forecasting. Transportation Research Part A: Policy and Practice, 78, 551–565. doi:10.1016/j.tra.2015.06.012 International Transport Forum. (2010). Reducing transport greenhouse gas emissions: Trends & data 2010. Retrieved from http://www.itf-oecd.org/sites/default/files/docs/10ghgtrends.pdf IPCC. (1995). IPCC second assessment, Climate change 1995: A report of the Intergovernmental Panel on Climate Change. Retrieved from https://www.ipcc.ch/pdf/climate-changes-1995/ipcc-2nd-assessment/2nd-assessment-en.pdf IPCC. (1996). Climate change 1995: Economic and social dimensions of climate change (Contribution of working group III to the second assessment report of the intergovernmental panel on climate change (Bruce, J. P., Lee, H., and Haites, E. F. (Eds.)]). Retrieved from https://www.ipcc.ch/ipccreports/sar/wg_III/ ipcc_sar_wg_III_full_report.pdf IPCC. (2007). Climate change 2007: The physical science basis (Contribution of working group I to the fourth assessment report of the intergovernmental panel on climate change [Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K. B., Tignor, M., and Miller, H. L. (Eds.)]). Cambridge University Press, Cambridge, United Kingdom and New York, USA.  Retrieved from https://www.ipcc.ch/pdf/assessment-report/ar4/wg1/ar4-wg1-121 frontmatter.pdf IPCC. (2014). Climate change 2014: Synthesis report (Contribution of working groups I, II and III to the fifth assessment report of the intergovernmental panel on climate change [Core Writing Team, Pachauri, R. K., and Meyer, L. A. (Eds.)]). Geneva, Switzerland. Retrieved from https://www.ipcc.ch/pdf/assessment-report/ar5/syr/SYR_AR5_FINAL_full.pdf Jung, S., An, K.-J., Dodbiba, G., and Fujita, T. (2012). Regional energy-related carbon emission characteristics and potential mitigation in eco-industrial parks in South Korea: Logarithmic mean Divisia index analysis based on the Kaya identity. Energy, 46(1), 231–241. doi:10.1016/j.energy.2012.08.028 Kawase, R., Matsuoka, Y., and Fujino, J. (2006). Decomposition analysis of CO2 emission in long-term climate stabilization scenarios. Energy Policy, 34(15), 2113–2122.  Kaya, Y. (1990). Impact of Carbon Dioxide emission control on GNP growth: Interpretation of proposed scenarios. In IPCC energy and industry subgroup, response strategies working group, Paris, 76. Ko, J., Park, D., Lim, H., and Hwang, I. C. (2011). Who produces the most CO2 emissions for trips in the Seoul metropolis area? Transportation Research Part D: Transport and Environment, 16(5), 358–364. doi:10.1016/j.trd.2011.02.001 Kockelman, K. (1997). Travel Behavior as Function of Accessibility, Land Use Mixing, and Land Use Balance: Evidence from San Francisco Bay Area. Transportation Research Record: Journal of the Transportation Research Board, 1607(970048), 116–125. doi:10.3141/1607-16 Kockelman, K. M. (1995). Which matters more in mode choice: Density or income? The relative effects of population density and income on commute-trip modal split in urban areas. In 1995 Compendium of Technical Papers. Institute of Transportation Engineers 65th Annual Meeting (pp. 844-867). Kockelman, K. M. (1996). Travel behavior as a function of accessibility, land use mixing, and land use balance (Master's thesis). University of California, Berkeley. Koppelman, F. S., and Bhat, C. (2006). A self instructing course in mode choice modeling : Multinomial and nested logit models. Technical report prepared for U.S. Department of Transportation, Federal Transit Administration. Retrieved from http://caee.utexas.edu/prof/bhat/COURSES/LM_Draft_060131Final-060630.pdf 122 Kwon, T.-H. (2005). Decomposition of factors determining the trend of CO2 emissions from car travel in Great Britain (1970–2000). Ecological Economics, 53(2), 261–275. doi:10.1016/j.ecolecon.2004.06.028 Lakshmanan, T. R., and Han, X. (1997). Factors underlying transportation CO2 emissions in the U.S.A.: A decomposition analysis. Transportation Research Part D: Transport and Environment, 2(1), 1–15. doi:10.1016/S1361-9209(96)00011-9 Lavoie, M. (2012). Characterizing land use and transportation for transit-oriented development in the Montreal metropolitan region (Master's thesis). McGill University. Loo, B. P. Y., and Li, L. (2012). Carbon dioxide emissions from passenger transport in China since 1949: Implications for developing sustainable transport. Energy Policy, 50, 464–476. doi:10.1016/j.enpol.2012.07.044 Lu, I. J., Lin, S. J., and Lewis, C. (2007). Decomposition and decoupling effects of carbon dioxide emission from highway transportation in Taiwan, Germany, Japan and South Korea. Energy Policy, 35(6), 3226–3235. doi:10.1016/j.enpol.2006.11.003 Mahendra, A., Bowen, B., Simons, M. E., and Adler, K. (2012). Impacts of implementing transportation control measures on travel activity and emissions. Transportation Research Record: Journal of the Transportation Research Board, 2287, 113–121. doi:10.3141/2287-14 Mahony, T. O. (2013). Decomposition of Ireland’s carbon emissions from 1990 to 2010: An extended Kaya identity. Energy Policy, 59, 573–581. doi:10.1016/j.enpol.2013.04.013 Masoud, A. R., Lee, A., Faghihi, F., and Lovegrove, G. (2015). Building sustainably safe and healthy communities with the Fused Grid development layout. Canadian Journal of Civil Engineering, 42(12), 1063–1072. doi:10.1139/cjce-2015-0086 Mathez, A., Manaugh, K., Chakour, V., El-Geneidy, A., and Hatzopoulou, M. (2013). How can we alter our carbon footprint? Estimating GHG emissions based on travel survey information. Transportation, 40(1), 131–149. doi:10.1007/s11116-012-9415-8 McFadden, D. (1974). Conditional logit analysis of qualitative choice behavior. Frontiers in Econometrics. doi:10.1108/eb028592 McFadden, D. (1977). Quantitative methods for analysing travel behaviour of individuals: Some recent developments. In D. A. Hensher and P. R. Stopher, (Ed.). (pp. 279-318). Taylor & Francis. 123 McNally, M. G. (2007). The four step model. In D. A. Hensher & K. J. Button (Ed.), Handbook of transport modelling Volume 1 (pp. 35-53). Mendenhall, W., and Sincich, T. (2007). Statistics for engineering and the sciences (5th ed.). Upper Saddle River, N.J: Pearson/Prentice Hall. Ministry of Environment. (2012). British Columbia greenhouse gas inventory report 2012. Retrieved from http://www2.gov.bc.ca/assets/gov/environment/climate-change/reports-and-data/provincial-ghg-inventory-report-bcs-pir/pir-2012-full-report.pdf Ministry of Environment. (2014a). Community energy and emissions inventory initiative: CEEI 2007-2010 Q&A. Retrieved from http://www2.gov.bc.ca/assets/gov/environment/climate-change/reports-and-data/community-energy-and-emissions-inventory-ceei/ceei_20072010_qa.pdf Ministry of Environment. (2014b). Kelowna city: 2010 community energy and emissions inventory, Monitoring and reporting on progress towards greenhouse gas emissions reduction targets. Retrieved from http://www2.gov.bc.ca/assets/gov/environment/climate-change/reports-and-data/community-energy-and-emissions-inventory-ceei/2010-ceei-reports/ceei_2010_kelowna_city.pdf Ministry of Environment. (2014c). Technical methods and guidance document 2007-2010 reports. Community energy and emissions inventory (CEEI) initiative. Retrieved from http://www2.gov.bc.ca/assets/gov/environment/climate-change/reports-and-data/community-energy-and-emissions-inventory-ceei/ceei_techmethods_guidance_final.pdf Mohareb, E., and Kennedy, C. (2012). Greenhouse gas emission scenario modeling for cities using the purge model: A case study of the greater Toronto area. Journal of Industrial Ecology, 16(6), 875–888. doi:10.1111/j.1530-9290.2012.00563.x Montgomery, D. C. (2009). Design and analysis of experiments (7th ed.). Hoboken, NJ: Wiley. Murphy, R. (2000). Analysis of measures for reducing transportation sector greenhouse gas (Master's thesis). Simon Fraser University. Myers, N. (1993). Population, environment, and development. Environmental Conservation, 20(03), 205–216. Naess, P. (2006). Accessibility, activity participation and location of activities: Exploring the links between residential location and travel behaviour. Urban Studies, 43(3), 627–652. 124 doi:10.1080/00420980500534677 National Cooperative Highway Research Program (NCHRP). (2012). Travel demand forecasting: Parameters and techniques (Report No. 716). Transportation Research Board of the National Academies. Retrieved from http://onlinepubs.trb.org/onlinepubs/nchrp/nchrp_rpt_716.pdf Nau, R. (2016). Regression diagnostics: Testing the assumptions of linear regression. Retrieved April 25, 2016, from Statistical forecasting: Notes on regression and time series analysis website: http://people.duke.edu/~rnau/testing.htm Neter, J., Kutner, M. H., Nachtsheim, C. J., and Wasserman, W. (1996). Applied linear statistical models. Irwin Chicago. Papagiannaki, K., and Diakoulaki, D. (2009). Decomposition analysis of CO2 emissions from passenger cars: The cases of Greece and Denmark. Energy Policy, 37(8), 3259–3267. doi:10.1016/j.enpol.2009.04.026 Phibbs, C. S., and Luft, H. S. (1995). Correlation of travel time on roads versus straight line distance. Medical Care Research and Review, 52(4), 532–542.  Poumanyvong, P., Kaneko, S., and Dhakal, S. (2012). Impacts of urbanization on national transport and road energy use: Evidence from low, middle and high income countries. Energy Policy, 46, 268–277. doi:10.1016/j.enpol.2012.03.059 Rahman, M. N., Ruparathna, R., Sekhon, S., Kumar, V., AlHashmi, M., Umer, A., Andrade, M. T. B., Dyck, R., Doberstein, C., Hewage, K., Hicky, R., Idris, A., Li, E., Culver, K., and Sadiq, R. (2016). Investigating the impacts of urban densities on city of Kelowna’s GHG emission targets. Technical report submitted to the City of Kelowna. Ramanathan, R. (2006). A multi-factor efficiency perspective to the relationships among world GDP, energy consumption and carbon dioxide emissions. Technological Forecasting and Social Change, 73(5), 483–494. doi:10.1016/j.techfore.2005.06.012 Raskin, P. D. (1995). Methods for estimating the population contribution to environmental change. Ecological Economics : The Journal of the International Society for Ecological Economics, 15(3), 225–233. doi:http://dx.doi.org/10.1016/0921-8009(95)00047-X Raupach, M. R., Marland, G., Ciais, P., Le Quéré, C., Canadell, J. G., Klepper, G., and Field, C. B. (2007). Global and regional drivers of accelerating CO2 emissions. Proceedings of the National Academy of Sciences of the United States of America, 104(24), 10288–93. 125 doi:10.1073/pnas.0700609104 Rawlings, J. O., Pantula, S. G., & Dickey, D. A. (2001). Applied regression analysis: A research tool. Springer Science & Business Media. Rodier, C. (2009). A review of the international modeling literature : Transit , land use , and auto pricing strategies to reduce vehicle miles traveled and greenhouse gas emissions. Strategies, 94804(510), 1–23. doi:10.3141/2132-01 Rosa, E. A., and Dietz, T. (1998). Climate change and society speculation, construction and scientific investigation. International Sociology, 13(4), 421–455. Schipper, L., and Marie-Lilliu, C. (1999). Transportation and CO2 emissions : Flexing the link - A path for the World Bank (Paper No. 69). Retrieved from Environment Department Papers, Climate Change Series, International Energy Agency, Paris. website: http://www-wds.worldbank.org/external/default/WDSContentServer/WDSP/IB/2000/02/09/000094946_00012505400755/Rendered/PDF/multi_page.pdf Schipper, L., Saenger, C., and Sudardshan, A. (2011). Transport and carbon emissions in the United States: The long view. Energies, 4(4), 563–581. doi:10.3390/en4040563 Scholl, L., Schipper, L., and Kiang, N. (1996). CO2 emissions from passenger transport. Energy Policy, 24(1), 17–30. doi:10.1016/0301-4215(95)00148-4 Smil, V. (1990). Planetary warming: Realities and responses. Population and Development Review, 16(1), 1–29. doi:10.2307/1972527 Spiess, H. (1990). Technical note - Conical volume-delay functions. Transportation Science, 24(2), 153–158. Stanley, J. K., Hensher, D. A., and Loader, C. (2011). Road transport and climate change: Stepping off the greenhouse gas. Transportation Research Part A: Policy and Practice, 45(10), 1020–1030. doi:10.1016/j.tra.2009.04.005 Statistics Canada. (2016). Focus on geography series, 2011 census. Retrieved April 25, 2016, from Census subdivision of Kelowna, CY - British Columbia website: http://www12.statcan.gc.ca/census-recensement/2011/as-sa/fogs-spg/Facts-csd-eng.cfm?LANG=Eng&GK=CSD&GC=5935010 Sugar, L., and Kennedy, C. (2013). A low carbon infrastructure plan for Toronto, Canada. Canadian Journal of Civil Engineering, 40(1), 86–96. doi:10.1139/cjce-2011-0523 Sun, J., and Lovegrove, G. (2013). Comparing the road safety of neighbourhood development 126 patterns: traditional versus sustainable communities. Canadian Journal of Civil Engineering, 40(1), 35–45. doi:10.1139/cjce-2012-0002 Taylor, J. (2001). Transportation and community design: The effects of land use and street pattern on travel behavior (Technical bulletin 11). Transportation and community design: The effects of land use, density and street pattern on travel behaviour. Retrived from http://www.jtc.sala.ubc.ca/bulletins/TB_issue_11_Transportation_edit.pdf Terefe, B. (2010). Greenhouse gas emissions from private vehicles in Canada , 1990 to 2007 (Catalogue No. 16-001-M, No. 12). Retrived from Environment accounts and statistics analytical and technical paper series, Statistics Canada website: http://www.statcan.gc.ca/pub/16-001-m/16-001-m2010012-eng.pdf Timilsina, G. R., and Shrestha, A. (2009). Transport sector CO2 emissions growth in Asia: Underlying factors and policy options. Energy Policy, 37(11), 4523–4539. doi:10.1016/j.enpol.2009.06.009 Tiwari, R., Cervero, R., and Schipper, L. (2011). Driving CO2 reduction by integrating transport and urban design strategies. Cities, 28(5), 394–405. doi:10.1016/j.cities.2011.05.005 Travel Forecasting Resource. (2016). Model Validation and Reasonableness Checking Manual. Retrieved July 10, 2016, from http://tfresource.org/Model_Validation_and_Reasonableness _Checking/Mode_Choice Trottier, S. (2015). Ecometrica: Understanding the changes to global warming potential (GWP) values. Retrieved May 20, 2016, from https://ecometrica.com/assets/Understanding-the-Changes-to-GWPs.pdf  UNFCCC. (2008). Kyoto protocol reference manual. On accounting of emissions and assigned amount, United Nations Framework Convention on Climate Change. Retrived from https://unfccc.int/resource/docs/publications/08_unfccc_kp_ref_manual.pdf USEPA. (2016). Understanding global warming potentials. Retrieved April 30, 2016, from Climate change website: https://www3.epa.gov/climatechange/ghgemissions/gwps.html Waggoner, P. E., and Ausubel, J. H. (2002). A framework for sustainability science: A renovated IPAT identity. Proceedings of the National Academy of Sciences of the United States of America, 99(12), 7860–5. doi:10.1073/pnas.122235999 Walsh, C., Jakeman, P., Moles, R., and O’Regan, B. (2008). A comparison of carbon dioxide emissions associated with motorised transport modes and cycling in Ireland. Transportation 127 Research Part D: Transport and Environment, 13(6), 392–399.  Wang, Q., Chiu, Y. H., and Chiu, C. R. (2015). Driving factors behind carbon dioxide emissions in China: A modified production-theoretical decomposition analysis. Energy Economics, 51, 252–260. doi:10.1016/j.eneco.2015.07.009 Washbrook, K., Haider, W., and Jaccard, M. (2006). Estimating commuter mode choice: A discrete choice analysis of the impact of road pricing and parking charges. Transportation, 33(6), 621–639. doi:10.1007/s11116-005-5711-x Xiangzhao, F., and Ji, Z. (2008). Economic analysis of CO2 emission trends in China. China Population, Resources and Environment, 18(3), 43–47. doi:10.1016/S1872-583X(09)60005-X Yang, C., McCollum, D., McCarthy, R., and Leighty, W. (2009). Meeting an 80% reduction in greenhouse gas emissions from transportation by 2050: A case study in California. Transportation Research Part D: Transport and Environment, 14(3), 147–156. doi:10.1016/j.trd.2008.11.010 Yang, W., Li, T., and Cao, X. (2015). Examining the impacts of socio-economic factors, urban form and transportation development on CO2 emissions from transportation in China: A panel data analysis of China’s provinces. Habitat International, 49, 212–220. doi:10.1016/j.habitatint.2015.05.030 York, R., Rosa, E. A., and Dietz, T. (2003). STIRPAT, IPAT and ImPACT: Analytic tools for unpacking the driving forces of environmental impacts. Ecological Economics, 46(3), 351–365. doi:10.1016/S0921-8009(03)00188-5 Zahabi, S. A. H., Miranda-Moreno, L. F., Patterson, Z., and Barla, P. (2012). Evaluating the effects of land use and strategies for parking and transit supply on mode choice of downtown commuters. Journal of Transport and Land Use, 5(2), 103–119. doi:10.5198/jtlu.v5i2.260 Zegras, C. (2010). The Built Environment and Motor Vehicle Ownership and Use: Evidence from Santiago de Chile. Urban Studies, 47(8), 1793–1817. doi:10.1177/0042098009356125 Zhang, L., Nasri, A., Hong, J. H., and Shen, Q. (2012). How built environment affects travel behavior: A comparative analysis of the connections between land use and vehicle miles traveled in US cities. Journal of Transport and Land Use, 5(3), 40–52. doi:10.5198/jtlu.v5i3.266 128 Zhang, M. (2004). The role of land use in travel mode choice - Evidence from Boston and Hong kong. Journal of the American Planning Association, 70(3), 344–360. doi:10.1080/01944360408976383 Zhou, B., and Kockelman, K. (2008). Self-Selection in Home Choice: Use of Treatment Effects in Evaluating Relationship Between Built Environment and Travel Behavior. Transportation Research Record: Journal of the Transportation Research Board, 2077, 54–61. doi:10.3141/2077-08 129 Appendices Appendix A  : Biogeme Software User Interface  Figure A-1 User Interface of Biogeme Software          130 Appendix B  : Sample Model on Biogeme Software  [Choice] // Column representing mode choice choice  [Beta] // Model parameters (Betas) // Name   Value   LowerBound UpperBound status (0=variable, 1=fixed) ASC_CAR  0.0  -100.0  100.0         0 ASC_RAIL      0.0    -100.0  100.0         1 BETA_COST     0.0    -100.0    100.0  0 BETA_TIME     0.0    -100.0      100.0          0  [Utilities] // Systematic utility specifications // Id  Name Avail   linear-in-parameter expression (beta1*x1 + beta2*x2 + ... ) 0 Car one ASC_CAR * one + BETA_COST * car_cost + BETA_TIME * car_ivtt 1 Rail one ASC_RAIL * one + BETA_COST * rail_cost + BETA_TIME * rail_ivtt      [Expressions]  // Define here arithmetic expressions for name that are not directly  // available from the data one =1  [Model] // Currently, only $MNL (multinomial logit), $NL (nested logit), $CNL // (cross-nested logit) and $NGEV (Network GEV model) are valid keywords // $MNL  131 Appendix C  : Sample Output Results from Biogeme  Figure C-1 Output Results from Biogeme 132 Appendix D  : Residual Plots of the Developed Regression Models  Figure D-1 Residual Plot of the Travel Distance and Travel Time for Car  Figure D-2 Residual Plot of the Travel Distance and Travel Time for Transit 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0314183/manifest

Comment

Related Items