Office of Research and Development
National Exposure Research Laboratory
-------
Table of Contents
Table of Contents i
List of Figures iv
List of Tables xiii
Acronyms and Abbreviations xvi
Authors and Acknowledgements xviii
Executive Summary ES-1
Project Background ES-1
Model Structure ES-2
Model Scenarios ES-3
Main Findings ES-4
Limitations of the Model ES-8
Intended Community Value ES-9
Next Steps ES-9
1 I ntroduction 1
1.1 Project Overview 1
1.2 Project Background 2
The Sustainable and Healthy Communities (SHC) Research Program 2
Criteria for Selecting a Modeling Approach 3
System Dynamics (SD) Models 3
The Durham-Orange Light Rail Project (D-O LRP) 4
1.3 Potential Uses for the D-O LRP SD Model 5
Integrated Assessments for Environmental Review 6
I ntegrated, Multi-Sector Approaches to Planning 7
Education and Enhancing Community Engagement 8
2 Model Development and Stakeholder Process 12
2.1 Overview 12
2.2 Problem Definition 12
2.3 Conceptual Model Development 13
Overview of Causal Loop Diagrams (CLDs) 13
Initial CLD for the D-O LRP SD Model 14
2.4 Quantitative Model Construction 16
2.5 Scenario Design and Evaluation 18
Final CLD for the D-O LRP SD Model, Version 1.0 19
-------
Model Structure 21
3.1 Overview of Data Processing and Primary Data Sources 21
Data Clipping Method 21
Primary Data Sources 22
3.2 Overall Model Structure 25
Model Specifications 25
Structural Overview 25
Inter-Sector Feedback Loops 26
3.3 Model Sector Descriptions 28
Land Use 29
Transportation 34
Energy 41
Economy 44
Equity 50
Water 53
Health 56
Scenario Descriptions 59
4.1 Main Scenario Descriptions 59
Business As Usual (BAU) 59
Light Rail 63
Light Rail + Redevelopment 65
4.2 Additional Scenario Descriptions 65
Land Use Planning Changes 65
Transportation Planning Changes 69
Market Trends 74
Demographic Trends 77
Environmental Policy Changes 79
Model Results 84
5.1 I ntroduction 84
5.2 Business as Usual (BAU) Scenario: Model Fit to Data and Projections 84
Tier 2 85
Tierl 91
BAU Results Summary 96
5.3 Light Rail Scenarios: Model Behaviors 96
5.4 Scenario Results by Sector 106
Land Use 106
-------
Transportation 114
Energy 125
Economy 134
Equity 145
Water 154
Health 162
5.5 Summary of Results 170
6 Quality Assurance 176
6.1 QA Overview 176
6.2 Model Parameterization (Calibration) 177
Data Collection and Analysis 177
Systemic Model Creation 182
Validation of Historical Simulations 192
6.3 Model Corroboration (Validation and Simulation) 212
Direct Structure Tests 214
Model Behavior and Sensitivity Tests 247
6.4 Computational Reproducibility 272
7 Summary & Conclusions 274
7.1 I ntegrated Assessments 274
7.2 Integrated Planning/Coordinated Agency Decision-Making 275
7.3 Support for Community Participation in Complex Decisions 276
7.4 Next Steps 277
Final Word 278
Appendix A: Advanced User Guide for the Durham-Orange Light Rail Project System
Dynamics Model 280
Appendix B: Detailed Data Documentation 304
Appendix C: M odel Input Characterization Tables 305
Appendix D: Data Sources Not Used 349
Appendix E: Stakeholder Meeting Materials 357
Bibliography 368
-------
List of Figures
Figure 1-1. Map of the Proposed Durham-Orange Light Rail Project Station Locations 14
Figure 2-1. Example CLD showing reinforcing (R) and balancing (B) feedback loops 14
Figure 2-2. Initial CLD for the D-O LRP SD Model, Presented at 2nd Stakeholder Meeting 15
Figure 2-3. Map of the D-O LRP SD Model Geographic Tiers 16
Figure 2-4. Depiction of the Iterative Model Construction Process 17
Figure 2-5. Stakeholder Agencies Overlaid on Final CLD for the D-O LRP Model 18
Figure 2-6. Results of Third Stakeholder Meeting Feedback Survey on the Top Priorities for the
D-O LRP SD Model: (A) Alternative Policies or Interventions to Test and (B) Output
Indicators to Add 19
Figure 2-7. Final CLD for the D-O LRP SD Model 20
Figure 3-1. Example of Methodology for Clipping in ArcGIS (Blue Circles Indicate Tier 1) 21
Figure 3-2. Simplified CLD of the D-O LRP SD Model with Core Sectors (Blue) and Output-
Oriented Sectors (Yellow) 26
Figure 3-3. Simplified CLD of R1: A Reinforcing Feedback Loop Between the Economy and
Land Use Sectors 26
Figure 3-4. Simplified CLD of B1: a Balancing Feedback Loop Between the Economy and
Transportation Sectors 27
Figure 3-5. Simplified CLD of B2: A Balancing Feedback Loop Between the Economy, Equity,
and Transportation Sectors 27
Figure 3-6. Simplified CLD of R1, R2, and B3: Balancing and Reinforcing Feedback Loops
Between the Economy, Land Use, and Energy Sectors 28
Figure 3-7. CLD for the Land Use Sector 32
Figure 3-8. CLD for the Transportation Sector 37
Figure 3-9. CLD for the Energy Sector 42
Figure 3-10. CLD of the Tier 2 Economy Sector 44
Figure 3-11. CLD of the Tier 1 Economy Sector 46
Figure 3-12. D-O LRP Cost and Revenue Assumptions (2010 Dollars) 49
Figure 3-13. CLD for the Equity Sector 51
Figure 3-14. Factors Affecting Single-family, Multifamily, and Commercial Property Values in the
D-O LRP SD Model 52
Figure 3-15. Storm Event Mean Concentrations (EMC) of Nitrogen in Both Tiers, Modeled for
2015 54
Figure 3-16. CLD for the Water Sector 55
Figure 3-17. CLD for the Health Sector 57
Figure 5-1. Additional Decision Support Scenarios (Bullets) Listed Under the Corresponding
Main Scenario With Which They Were Run 83
Figure 5-2. Population - Tier 2: BAU and Data, 2000-2040 84
iv
-------
Figure 5-3. Developed Land-Tier 2: BAU and Data, 2000-2040 85
Figure 5-4. Nonresidential Sq. Ft. - Tier 2: BAU and Data, 2000-2040 85
Figure 5-5. Total Employment-Tier2: BAU and Data, 2000-2040 85
Figure 5-6. Total Retail Consumption - Tier 2: BAU and Data, 2000-2040 86
Figure 5-7. Gross Regional Product - Tier 2: BAU and Data, 2000-2040 86
Figure 5-8. Vehicle Miles Traveled (VMT) Tier 2 BAU and Data, 2000-2040 87
Figure 5-9. Congestion - Tier 2: BAU and Data, 2000-2040 88
Figure 5-10. Public Transit Person Miles - Tier 2: BAU and Data, 2000-2040 88
Figure 5-11. Building Energy Use-Tier2: BAU and Data, 2000-2040 89
Figure 5-12. CO2 Emissions - Tier 2: BAU and Data, 2000-2040 89
Figure 5-13. Impervious Surface - Tier 2: BAU and Data, 2000-2040 90
Figure 5-14. Population Tier- 1: BAU and Data, 2000-2040 90
Figure 5-15. Developed Land - Tier 1: BAU and Data, 2000-2040 91
Figure 5-16. Nonresidential sq ft-Tier 1: BAU and Data, 2000-2040 91
Figure 5-17. Total Employment - Tier 1: BAU and Data, 2000-2040 92
Figure 5-18. Gross Regional Product-Tier 1: BAU and Data, 2000-2040 92
Figure 5-19. Total Retail Consumption - Tier 1: BAU and Data, 2000-2040 92
Figure 5-20. VMT-Tier 1: BAU and Data, 2000-2040 93
Figure 5-21. Congestion - Tier 1: BAU and Data, 2000-2040 93
Figure 5-22. Public Transit Person Miles - Tier 1: BAU and Data, 2000-2040 94
Figure 5-23. Impervious Surface-Tier 1: BAU and Data, 2000-2040 95
Figure 5-24. Total Nonresidential Sq Ft - Tier 1: Main Policy Scenarios Compared to BAU 96
Figure 5-25. Total Nonresidential Sq Ft - Tier 2: Main Policy Scenarios Compared to BAU 97
Figure 5-26. Total Employment - Tier 1: Main Policy Scenarios Compared to BAU 97
Figure 5-27. Total Employment - Tier 2: Main Policy Scenarios Compared to BAU 97
Figure 5-28. Population - Tier 1: Main Policy Scenarios Compared to BAU 98
Figure 5-29. Population - Tier 2: Main Policy Scenarios Compared to BAU 99
Figure 5-30. VMT-Tier 1: Main Policy Scenarios Compared to BAU 99
Figure 5-31. VMT, Tier 2: Main Policy Scenarios Compared to BAU 100
Figure 5-32. Congestion - Tier 1: Main Policy Scenarios Compared to BAU 101
Figure 5-33. Congestion - Tier 2: Main Policy Scenarios Compared to BAU 101
Figure 5-34. Percent of Population in Poverty - Tier 1: Main Policy Scenarios Compared to BAU ..102
Figure 5-35. Percent of Population in Poverty - Tier 2: Main Policy Scenarios Compared to BAU ..102
Figure 5-36. Percent of Households with Zero Cars - Tier 1: Main Policy Scenarios Compared to
BAU 103
-------
Figure 5-37. Percent of Households with Zero Cars - Tier 2: Main Policy Scenarios Compared to
BAU 103
Figure 5-38. Total Energy Spending - Tier 1: Main Policy Scenarios Compared to BAU 104
Figure 5-39. Total Energy Spending - Tier 2: Main Policy Scenarios Compared to BAU 104
Figure 5-40. Percent Change in Land Use Sector Model Outputs Between 2020 and 2040 for
Three Main Scenarios 105
Figure 5-41. Developed Land: Main Policy Scenarios Compared to BAU 106
Figure 5-42. Developed Acres by Use in 2000: BAU 107
Figure 5-43. Nonresidential Sq. Ft. by Use in 2000 and 2040-Tier 1: BAU 108
Figure 5-44. Percent Change in Density Measures Between 2020 and 2040 for Three Main
Scenarios 108
Figure 5-45. HHI of Mixed Use: Main Policy Scenarios Compared to BAU 109
Figure 5-46. Jobs-Housing Balance: Main Policy Scenarios Compared to BAU 110
Figure 5-47. Percent Change in Selected Land Use Sector Outputs Between 2020 and 2040 for
BAU + Redevelopment - Tier 1 111
Figure 5-48. Developed Land - Tier 1: BAU + Redevelopment 111
Figure 5-49. Developed Land - Tier 1: Bold Redevelopment 112
Figure 5-50. Percent Change in Housing & Transportation Costs Between 2020 and 2040 for
Bold Redevelopment 113
Figure 5-51. Public Transit Ridership Under Three Main Scenarios 114
Figure 5-52. Modal Person Miles of Travel Per Day in 2020 and in 2040 Under Three Main
Scenarios 115
Figure 5-53. Percent Change in Transportation Sector Model Outputs Between 2020 and 2040
for Three Main Scenarios 116
Figure 5-54. Percent Change in Modal Person Miles of Travel by Residents Per Day Per Capita
Between 2020 and 2040 for Three Main Scenarios 118
Figure 5-55. (A) VMT - Tier 2 and (B) Percent Change in Modal Person Miles of Travel by Tier 2
Residents Per Day Per Capita between 2015 and 2040 for Higher Gas Prices
Scenario vs. BAU 119
Figure 5-56. (A) Congestion - Tier 2 and (B) Percent Change in Modal Person Miles of Travel by
Tier 2 Residents Per Day Per Capita Between 2017 and 2040 for No Road Building
Scenario vs. BAU and Light Rail Scenarios 120
Figure 5-57. (A) Change in Modal Person Miles of Travel by Tier 2 Residents Per Day Per Capita
Between 2020 and 2040 for Fare Free Transit Scenario vs. BAU and Light Rail
Scenarios and (B) Hours of Congestion Delay Per VMT-Tier 2 121
Figure 5-58. (A) Percent Change in Modal Person Miles of Travel by Tier 1 Residents Per Capita
Between 2020 and 2040 for High Parking Price Scenario vs. Light Rail +
Redevelopment Scenario and (B) Transportation and Renter Costs - Tier 1 122
Figure 5-59. (A) Nonmotorized Travel Facilities - Tier 2 and (B) Percent Change in Modal Person
Miles of Travel by Tier 2 Residents Per Day Per Capita Between 2020 and 2040 for
Sidewalk Building Scenario vs. BAU 123
Figure 5-60. Energy Use: Light Rail, Light Rail + Redevelopment Scenarios Compared with BAU . 124
vi
-------
Figure 5-61. CC>2 Emissions: Light Rail, Light Rail + Redevelopment Scenarios Compared with
BAU 124
Figure 5-62. Energy Use in 2020 and 2040 Under the Three Main Scenarios 125
Figure 5-63. CO2 Emissions Over Time in the BAU Scenario, Color Coded by Source 126
Figure 5-64. Percent Change in Energy Spending Between 2020 and 2040 for Three Main
Scenarios 127
Figure 5-65. Percent Change in Energy Intensity Between 2020 and 2040 for Three Main
Scenarios 128
Figure 5-66. (A) Tier 2 CO2 Emissions, (B) Tier 2 GRP, and (C) Percent Change between 2020
and 2040 (Tier 2): Light Rail + Redevelopment Scenario with Energy Efficiency 129
Figure 5-67. (A) Light Rail Effect on Public Transit Usage - Tier 2 and (B) Energy and
Transportation Indicators - Tier 1: Light Rail + Redevelopment With Higher Light
Rail Effect on Public Transit Ridership 130
Figure 5-68. CC>2 Emissions: Light Rail + Redevelopment Scenario With and Without Clean
Power Plan 131
Figure 5-69. (A) Solar Capacity, (B) Annual CC>2 Emissions, and (C) Solar as a Fraction of
Building Electricity Use: Light Rail + Redevelopment Scenario Compared With
Increased Solar Capacity Scenarios in Tier 2 132
Figure 5-70. Change in Overall Economic Growth Indicators between 2020 and 2040 133
Figure 5-71. Percent Change in Economy Sector Employment Indicators between 2020 and
2040 for Three Main Scenarios 134
Figure 5-72. Time Series of Unemployment Rate over the Entire Study Period (2000-2040) 135
Figure 5-73. Percent Change in Per Capita Economic Growth Indicators between 2020 and 2040
for Three Main Scenarios 136
Figure 5-74. Time Series between 2000 and 2040 of Total Real Property Value: Model Scenario
Outputs and Historical Data 137
Figure 5-75. Absolute Change in Real Property Value by Type between 2020 and 2040 for Three
Main Scenarios 138
Figure 5-76. Percent Change in Property Tax (PT) Revenues between 2020 and 2040 for Three
Main Scenarios 139
Figure 5-77. Percent Change in D-O LRP Revenue Sources between 2020 AND 2040 for Three
Main Scenarios 139
Figure 5-78. Cumulative Nominal D-O LRP Revenues between 2010 and 2040, Three Main
Model Scenarios vs. Local Projections 140
Figure 5-79. D-O LRP Revenues, LRT Costs, and Net Budget between 2010 and 2040 141
Figure 5-80. (A) Total Nonresidential Sq. Ft. - Tier 2 and (B) Economic Indicators - Tier 2 142
Figure 5-81. (A) Gross Operating Surplus Tier 2 and (B) Change (%) in Economic indicators -
Higher Rent Compared to Three Main Scenarios 143
Figure 5-82. Percent Change in Equity Outcomes between 2020 and 2040 for Three Main
Scenarios 144
Figure 5-83. Change (%) in Property Values between 2020 and 2040 for Three Main Scenarios... 145
Figure 5-84. Multifamily and Single-Family Unit Property Values - Tier 1 146
vii
-------
Figure 5-85.
Figure 5-86.
Figure 5-87.
Figure 5-88.
Figure 5-89.
Figure 5-90.
Figure 5-91.
Figure 5-92.
Figure 5-93.
Figure 5-94.
Figure 5-95.
Figure 5-96.
Figure 5-97.
Figure 5-98.
Figure 5-99.
Figure 5-100.
Figure 5-101.
Figure 5-102.
Figure 5-103.
Figure 5-104.
Figure 5-105.
Figure 5-106.
Percent Change in Housing and Transportation Costs between 2020 and 2040 for
Three Main Scenarios 147
Percent Change in Affordability Indices between 2020 and 2040 for Three Main
Scenarios 147
Households in Poverty at Risk of Displacement 148
Housing Gap for Households in Poverty - Tier 1 149
Multifamily and Single-Family Acres Under Light Rail + Redevelopment and More
Multifamily Households Scenarios - Tier 1 150
Percent Change in Developed Land, Residential Impervious Surface, and
Residential Energy Use Between 2020 and 2040 for More Multifamily Households
Scenario Compared to Light Rail + Redevelopment - Tier 1 150
Potential Population in Poverty Displaced - Tier 1 151
(a) Retail Earnings Tier 1 and (b) GRP Tier 1 - Retail Wage Increase Scenario vs.
Light Rail + Redevelopment Scenario 152
(a) Resident Per Capita Net Retail Earnings Tier 1 and (B) Affordability Index Tier 1
- Retail Wage Increase Scenario vs. Light Rail + Redevelopment Scenario 152
Water Demand: Light Rail, Light Rail + Redevelopment Scenarios Compared to
BAU
153
Tier 1 Water Demand (MGAL Per Day) in (A) 2000 and (B) 2040, BAU Scenario
Compared to Light Rail Scenarios 154
Tier 2 Water Demand (MGAL Per Day) in (A) 2000 and (B) 2040, BAU Scenario
Compared to Light Rail Scenarios155
Days of Supply in Durham County Water Reservoirs: Three Main Scenarios 155
Stormwater N Load by Land Use: BAU Scenario 156
Stormwater N Load: Light Rail, Light Rail + Redevelopment Scenarios Compared to
BAU 157
(A) Total Impervious Surface - Tier 1 and (B) Total Developed Land - Tier 1: Light
Rail, Light Rail + Redevelopment Scenarios Compared to BAU 157
Stormwater P Load: Light Rail, Light Rail + Redevelopment Scenarios Compared to
BAU 158
Annual Stormwater Runoff Volume and Percent Impervious Surface, Modeled for
2015
158
Stormwater N Load - Tier 1: (A) Percent Reduction Targets and (B) Reduction to
2.2 Ib/acre/year 160
Time Series of Three Exogenous Inputs Affecting Health Outcomes - Same for all
Three Main Scenarios 161
Percent Change in VMT and VMT-Related Health Indicators between 2020 and
2040 for Three Main Scenarios 162
Time Series (2000-2040) of Nonmotorized Travel for Transportation by Residents
Per Day Per Capita for Three Main Scenarios 163
VIII
-------
Figure 5-107. Percent Change in Nonmotorized Travel by Residents Per Day and Associated
Percent Change in Avoided Premature Mortalities between 2020 and 2040 for Three
Main Scenarios: Total and Population-Normalized 163
Figure 5-108. Net Premature Mortalities Avoided Per Year from PM2.5 and NOX Vehicle Emissions,
Crash Fatalities, and Nonmotorized Travel Combined between 2020 and 2040,
Departure from BAU 165
Figure 5-109. Cumulative Premature Mortalities Avoided by Cause and Net Cumulative Premature
Mortalities Avoided between 2020 and 2040 for the Light Rail and Light Rail +
Redevelopment Scenarios: Departure from BAU 165
Figure 5-110. (A) PM2.5 Vehicle Emissions Per VMT: Vehicle Emissions Reduced Scenario
Compared to BAU and (B) Cumulative Premature Mortalities Avoided from PM2.5
and NOX Vehicle Emissions - Tier 2: Vehicle Emissions Reduced Scenario
Departure from BAU 166
Figure 5-111. (A) Nonmotorized Travel by Residents Per Day Per Capita - Tier 2: Higher Gas
Price Scenario Compared to BAU and (B) Cumulative Premature Mortalities
Avoided Due to Nonmotorized Travel - Tier 2: Higher Gas Price Scenario Departure
from BAU 167
Figure 5-112. (A) Nonmotorized Travel by Residents Per Day Per Capita - Tier 2: Sidewalk
Building Scenario Compared to BAU and (B) Cumulative Premature Mortalities
Avoided Due to Nonmotorized Travel - Tier 2: Sidewalk Building Scenario
Departure from BAU 168
Figure 6-1. Schematic Illustrating Cross-Sectoral Linkages Established After Model Integration .. 181
Figure 6-2. Service Square Feet-Tier 2: Before and After Model Integration 182
Figure 6-3. Percent Change in Total Nonresidential Sq. Ft. - Tier 2: 20% Increase in Demand
Scenario Over BAU, Pre- and Post-Integration 183
Figure 6-4. Transportation Sector in Tier 2, Before and After Sectoral Integration 184
Figure 6-5. Energy Model in Tier 2, Before and After Sectoral Integration: (A) Commercial
Energy Use and (B) Commercial Square Footage 185
Figure 6-6. Energy Model in Tier 2, Before and After Sectoral Integration: (A) Vehicle Fuel
Consumption and (B) VMT 186
Figure 6-7. Unemployment Rate - Tier 1: Unemployment Effect on Migration in Tier 1 Removed 187
Figure 6-8. BAU Scenario vs. Historical Data for Population 192
Figure 6-9. BAU Scenario vs. Historical Data for Developed Land 193
Figure 6-10. BAU Scenario vs. Historical Data for Nonresidential Sq. Ft 193
Figure 6-11. BAU Scenario vs. Historical Data for Single Family Dwelling Units 194
Figure 6-12. BAU Scenario vs. Historical Data for Multifamily Dwelling Units 194
Figure 6-13. BAU Scenario vs. Historical Data for Impervious Surfaces 195
Figure 6-14. BAU Scenario vs. Projections for VMT 196
Figure 6-15. BAU Scenario vs. Projections for Congestion 197
Figure 6-16. BAU Scenario vs. Data and Projections for Public Transit Person Miles 198
IX
-------
Figure 6-17. Building Energy Use - Tier 2: BAU Model Output Compared to Data 198
Figure 6-18. CO2 Emissions - Tier 2: BAU Model Output Compared to Data 199
Figure 6-19. BAU Scenario vs. Historical Data and Projections for Total Employment 200
Figure 6-20. Historical Employment Data and Projections Used as Inputs in the Tier 2 Economy
Model for (A) Shares of Employment by Category, and (B) Average Earnings Per
Employee by Category 201
Figure 6-21. Historical Employment Data and Projections Used as Inputs in the Tier 1 Economy
Model for (A) Shares of Employment by Category, and (B) Average Earnings Per
Employee by Category 202
Figure 6-22. Total Earnings: Model Fit to Calculated Historical Data and Projections 202
Figure 6-23. Gross Regional Product and Gross Operating Surplus: Model Fit to Values
Calculated from Data 203
Figure 6-24. Total Retail Consumption - Tier 2: Model Fit to Historical Data and Projections 204
Figure 6-25. Total Retail Consumption -Tier 1: Model Fit to Historical Data 205
Figure 6-26. BAU Scenario vs. Historical Data for Tier 2 Residential Property Values 206
Figure 6-27. BAU Scenario vs. Historical Data for Tier 1 Residential Property Values 206
Figure 6-28. BAU Scenario vs. Historical Data for Nonresidential Property Values 206
Figure 6-29. Water Demand - Tier 2: BAU Model Output Compared to Data 207
Figure 6-30. Illustration of Average Percent Departure Above or Below vs. Average Absolute
Percent Departure from a Reference Run 212
Figure 6-31. Acres by Land Use Type - Tier 1: Before and After Land Use Sector Restructuring... 213
Figure 6-32. Developed Land-Tier 1: Land Development Restructure Test 214
Figure 6-33. Multifamily Dwelling Units - Tier 2: Effect of Vacancy on Dwelling Units Removed 215
Figure 6-34. Person Miles of Travel by Different Modes Not Normalized vs. BAU 217
Figure 6-35. Automobile Driver Person Miles Per Day: No Post-Normalization Congestion Effect
on Automobile Driver Person Miles Percent Difference from BAU 219
Figure 6-36. NMT Facilities 220
Figure 6-37. Public Transit Person Miles of Travel Per Day - Tier 2: Light Rail Scenario with Light
Rail Treated Like Bus Service for Determining Transit Use vs. Regular Light Rail
Scenario 223
Figure 6-38. Public Transit Person Miles Of Travel Per Day - Tier 2: Light Rail Scenario with Light
Rail Treated Like Bus Service For Determining Transit Use and No Assumed Direct
Effect of Light Rail On Demand For Commercial Floor Space Or Migration Vs. Light
Rail Scenario With Light Rail Treated Differently from Bus Service for Determining
Transit Use But Still No Assumed Direct Effect of Light Rail On Demand for
Commercial Floor Space or Migration 223
Figure 6-39. Structural Analysis for Unlinking Gross Operating Surplus from Energy Spending:
Effect in Tier 2 on "Energy Spending Share of GRP Factor Affecting GOS" 225
-------
Figure 6-40. Structural Analysis for Unlinking Gross Operating Surplus from Energy Spending:
Effect in Tier 2 on (A) GRP and (B) GRP Growth Rate 226
Figure 6-41. Structural Analysis for Using Relative Energy Spending, Not Energy Spending/GRP,
in Feedback to GRP: Effect on (A) Relative Energy Spending and (B) GRP in Tier 2. 227
Figure 6-42. Structural Test of Removing Effect of Congestion on Economic Productivity:
Comparison to BAU of Two Economic Indicators for (A) Tier 2 and (B) Tier 1 228
Figure 6-43. Resident Percent of the Working Population for Tier 2 and Tier 1, Calculated from
Historical Data and Projections of Employment and Population 229
Figure 6-44. Structural Test of the Economic Effects of Three Different Formulations for Resident
Labor Force Tier 1: Comparison of (A) Unemployment Rate (%) Tier 1 and (B)
Population Tier 1 231
Figure 6-45. Structural Test of the Economic Effects of Three Different Formulations for Resident
Labor Force Tier 1: Comparison of (A) Total Earnings Tier 1 and (B) Resident Per
Capita Net Earnings Tier 1 231
Figure 6-46. Multifamily Property Value Per MF Dwelling Unit - Tier 1: BAU and Structural Test.... 233
Figure 6-47. Structural Analysis for Water Demand: Sectoral Formulation Compared to Per
Capita Formulation 234
Figure 6-48. Structural Analysis for Stormwater N Load: Land Uses Aggregated Compared to
Disaggregated 235
Figure 6-49. Structural Analysis for Health Effects of Prr^.s and Nox Vehicle Emissions: Total
Premature Mortalities Relative to BAU -Tier2 236
Figure 6-50. Extreme Value Test of Zero Person Miles by Any Mode from Start of Model Run vs.
BAU 237
Figure 6-51. Extreme Value Test of 70% Population Drop at 2020 vs. BAU 239
Figure 6-52. Extreme Value Test of Unemployment Becoming 50% During 2020-2040 vs. BAU.... 240
Figure 6-53. Extreme Value Test of an Economic Crash Wherein GRP Declines 20% in 2020 vs.
BAU 242
Figure 6-54. Extreme Value Test of Gas Price Spike to $1,000 Per Gallon During 2016-2018 vs.
BAU 244
Figure 6-55. Total Impervious Surface - Tier 2: Sensitivity to Nonresidential Impervious Surface
Coefficients 247
Figure 6-56. Tier 1 Person Miles Per Day by Mode: Percent Difference from BAU 249
Figure 6-57. Sensitivity Analysis for Energy Efficiency: Effect on CC>2 Emissions in the BAU
Scenario - Tier 2 251
Figure 6-58. Sensitivity Analysis for Elasticity of Consumption to GRP: Effect on Total Retail
Consumption-Tier 2 251
Figure 6-59. Sensitivity Analysis for Employment Per Dollar of Consumption: Effect On Desired
Employment (A) and Total Employment (B) for Tier 2 (C) and Tier 1 (D) 252
Figure 6-60. Sensitivity Analysis for Average Precipitation with Drought Year: Effect On Tier 1
Stormwater N Load 253
Figure 6-61. Sensitivity Analysis for Impervious Surface - Tier 1: Effect On (A) Stormwater N
Load and (B) Stormwater N Load Per Acre 254
XI
-------
Figure 6-62. Sensitivity Analysis of Doubling the Average Person Miles of NMT Per PIT: Effect
On (A) Person Miles of NMT by Residents Per Day and (B) Avoided Premature
Mortality Due to NMT-Tier 1 255
Figure 6-63. Sensitivity Analysis of Doubling the Average Person Miles of NMT Per PTT: Effect
on (A) Person Miles of NMT by Residents Per Day and (B) Avoided Premature
Mortality Due to NMT-Tier 2 255
Figure 6-64. Tier 2 Person Miles Per Day by Mode: Longer Reaction Times vs. BAU 258
Figure 6-65. Sensitivity Analysis for Earnings Per Industrial Employee Tier 1 Reduced by 50%
Between 2000 and 2040: Effect on (A) Industrial Earnings Tier 1 and (B) Relative
GRPTieM 259
Figure 6-66. Sensitivity Analysis for Average Precipitation with Random Factor: Effect on (A)
Projected Precipitation and (B) Tier 2 Stormwater N Load 260
Figure 6-67. Nonresidential Sq. Ft. - Tier 1: Monte Carlo Simulation for Sensitivity to Effect of
LRT on Nonresidential Sq. Ft. Demand in Tier 1 261
Figure 6-68. Unemployment Rate - Tier 1: Monte Carlo Simulation for Sensitivity to Effect of LRT
on Nonresidential Sq. Ft. Demand in Tier 1 262
Figure 6-69. Affordability Index- Tier 1: Monte Carlo Simulation for Sensitivity to Effect of LRT on
Nonresidential Sq. Ft. Demand in Tier 1 262
Figure 6-70. Net Migration - Tier 1: Monte Carlo Simulation for Sensitivity to Effect Of LRT on
Nonresidential Sq. Ft. Demand In Tier 1 262
Figure 6-71. Net Migration - Tier 2: Monte Carlo Simulation for Sensitivity to Percent of Net
Migration Due to Light Rail in Tier 1 That Is External to Tier 2 265
Figure 6-72. Unemployment Rate - Tier 2: Monte Carlo Simulation for Sensitivity to Percent of
Net Migration Due to Light Rail in Tier 1 That Is External to Tier 2 265
Figure 6-73. Traffic Congestion: Congestion Per Weekday Peak-Period VMT Per Lane Mile Held
Constant vs. No Road Building, Light Rail, and BAU Scenarios 268
Figure 6-74. Sample "Runs Compare" Report 271
XII
-------
List of Tables
Table 3-1. Summary of the D-O LRP SD Model's Specifications 25
Table 3-2. Summary of Primary Stocks in the Land Use Sector 29
Table 3-3. Variables that Affect Modal Person Miles 34
Table 3-4. Economy Sector Variables Affecting Variables in Other Sectors 45
Table 3-5. Job Types by 2-digit NAICS Codes Aggregated into Four Employment Categories
Used in the D-O LRP SD Model 47
Table 4-1. Values for "Finished LRT Line Miles" in Tier 1 and Tier 2 under the Light Rail
Scenario 62
Table 4-2. Values for "Urban Nonhighway Lane Miles in Disruption from LRT Construction" in
Tier 1 and Tier 2 under the Light Rail Scenario 63
Table 4-3. Values for "Building Energy Intensity Policy" in Tier 1 and Tier 2 under the Light Rail
+ Redevelopment Scenario 67
Table 4-4. Values for "MPG Policy" in Tier 1 and Tier 2 under the Light Rail + Redevelopment
Scenario 67
Table 4-5. Values for "MPG without Congestion" under the Light rail + Redevelopment
Scenario 67
Table 4-6. Values for "Functioning Lane Miles" in Tier 1 and Tier 2 under the No Road Building
and BAU Scenarios 68
Table 4-7. Values for "Public Transit Fare Price" under the Fare Free, BAU, and Light Rail
Scenarios 70
Table 4-8. Values for "Parking Cost of AverageTtrip" in Tier 1 and Tier 2 under the High
Parking Price and Light Rail + Redevelopment Scenarios 71
Table 4-9. Values for "Desired Nonmotorized Travel Facilities per Developed Acre" in Tier 1
and Tier 2 under the Sidewalk Building and BAU Scenarios 72
Table 4-10. Values for "Change in Person Miles of Public Transit Travel Per Year Due to Adding
Fixed Guideway Transit" in Tier 1 and Tier 2 Under the Higher LRT Effect on Public
Transit Ridership Scenarios and the Light Rail + Redevelopment Scenario (Million
miles per year) 73
Table 4-11. Values for "Price of Gasoline" Under the High Gas Price and BAU Scenarios (USD
2010 per gallon) 74
Table 4-12. Values for "Nominal Hourly Retail Wage" and "Retail Earnings per Employee" Under
the Retail Wage Increase and Light Rail + Redevelopment Scenarios 75
Table 4-13. Values for "Gross Operating Surplus per sq. ft." Under the Higher Rent and Light
Rail + Redevelopment Scenarios 76
Table 4-14. Values for "Percent of People in SFDU Table Historical Tier 1" Under the More
Multifamily Households and BAU Scenarios 77
Table 4-15. Values for "Percent of MF Dwelling Units Below 77 Percent of Median Renter Costs
Table Tier 1" Under the Fewer Organically Affordable Units and BAU Scenarios 78
XIII
-------
Table 4-16. Values for "PIVb.s Emissions per VMT" and "NOX Emissions per VMT" Under the
Vehicle Emissions Reduced and BAU Scenarios 79
Table 4-17. Values for "Desired Future Solar Capacity" Under the increased Solar Capacity and
Light Rail + Redevelopment Scenarios 80
Table 4-18. Electricity Emissions Factors "Ton CO2 per kWh" and "Lb CO2 per MWh" Under the
Clean Power Plan and Light Rail + Redevelopment Scenarios 81
Table 4-19. Percent of Stormwater N Load from Post-2015 Development Treated Onsite, and
Percent from pre-2015 Development Treated Onsite Under the Stormwater
Management Scenarios 82
Table 5-1. Summary of Changes in Key Indicators between 2020 and 2040 for the BAU, Light
Rail, and Light Rail + Redevelopment Scenarios 169
Table 5-2. Summary of Key Indicators in 2040 for the BAU, Light Rail, and Light Rail +
Redevelopment Scenarios 172
Table 6-1. Selected Variables and Data Sources for Historical and Projected Data 177
Table 6-2. Existing Simulation Models Used in the D-O LRP Model 178
Table 6-3. Equations Obtained from Existing Studies and Models 179
Table 6-4. Average Yearly Percent Departure from Data, 2000-2014: Unemployment Effect on
Migration Tier 1 Removed 187
Table 6-5. Selected Variables from Model Input Characterization Table 189
Table 6-6. Model Input Uncertainty Characterization Summary by Model Sector 191
Table 6-7. Statistical Analysis of Tier 1 Variables Compared to Historical Data and Projections.. 208
Table 6-8. Statistical Analysis of Tier 2 Variables Compared to Historical Data and Projections.. 209
Table 6-9. Average Yearly Percent Departure of Test from Restructured Land Sector and from
BAU, 2000-2040: Land Development Structure Test 215
Table 6-10. Average Yearly Percent Departure from BAU, 2000-2040: Effect of Vacancy on
Dwelling Units Removed 216
Table 6-11. Average Yearly Percent Departure from BAU, 2000-2040: Not Normalizing Person
Miles by Mode 217
Table 6-12. Average Yearly Percent Departure from BAU, 2000-2040: No Post-Normalization
Congestion Effect on Automobile Driver Person Miles 219
Table 6-13. Average Yearly Percent Departure from BAU, 2000-2040: Building of Pedestrian
and Bicycle Facilities Driven by Exogenous Construction Schedule Based on TRM
VSSEData 221
Table 6-14. Average Yearly Percent Departure from Light Rail Scenario, 2026-2040: Light Rail
Treated the Same as Bus Service for The Purpose of Determining Ridership 223
Table 6-15. Average Yearly Percent Departure from Light Rail Scenario with No Assumed Direct
Effect of Light Rail On Demand for Commercial Floor Space or Migration, 2026-
2040: Light Rail Treated the Same As Bus Service for the Purpose of Determining
Ridership + No Assumed Direct Effect of Light Rail On Demand for Commercial
Floor Space or Migration 224
XIV
-------
Table 6-16. Elasticities Used to Calibrate Property Values in the BAU and the Property Value
Structure Test 232
Table 6-17. Average Yearly Percent Departure from BAU, 2000-2040, Property Value Structure
Test 233
Table 6-18. Selected Quantitative Indicators and Unit of Measure 245
Table 6-19. Average Yearly Percent Departure from BAU, 2000-2040: Sensitivity to
Nonresidential Impervious Surface Coefficients 248
Table 6-20. Average Yearly Percent Departure from BAU, 2000-2040: Sensitivity to Tier-1-to-
Tier-2 Weighting Factors 250
Table 6-21. Average Yearly Percent Departure From BAU, 2000-2040 252
Table 6-22. Sensitivity of Stormwater Runoff to Impervious Surface: Average Yearly Percent
Departure from BAU From 2000 to 2040 254
Table 6-23. Input Variable Changes Made to Test Model Sensitivity to Longer Reaction Times
Associated with Elasticities, Measured in Years 256
Table 6-24. Yearly Percent Departure from BAU: Sensitivity to Longer Reaction Times
Associated with Elasticities from Literature in Transportation Sector 258
Table 6-25. Average Yearly Percent Departure from BAU, 2000-2040 260
Table 6-26. Average Yearly Percent Departure from BAU, 2000-2040: Sensitivity Effect of LRT
on Nonresidential Sq. Ft. Demand Tier 1 264
Table 6-27. Average Yearly Percent Departure from BAU, 2000-2040: Sensitivity to Percent of
Net Migration Due to Light Rail in Tier 1 that is External to Tier 2 266
Table 6-28. Average Yearly Percent Departure from BAU, Light Rail, and No Road Building
Scenarios, 2010-2040: Congestion Per Weekday Peak-Period VMT Per Lane Mile
Held Constant 269
Table 6-29. Sensitivity Analysis for Energy System Variables: Effect on Cross-Sector Indicators
in Tier 2 in 2040 270
xv
-------
Acronyms and Abbreviations
Abbreviation
ACS
AEO
BAU
BEA
CAMPO
CATS
CLD
CPP
CV2
DATA
DCHC MPO
DEIS
D-OLRP
EIA
EIS
EMC
FAR
FTA
FY
GHG
GIS
GOS
GREET
GRP
HEAT
HHI
HIA
HUD
IEA
LEHD
LINC
LODES
LRP
LRT
MF
MOVES
MPG
MPO
MTP
Description
American Community Survey (from U.S. Census Bureau)
Annual Energy Outlook
Business-As-Usual
U.S. Bureau of Economic Analysis
Capital Area Metropolitan Planning Organization (Raleigh, NC)
Charlotte Area Transit System
Causal Loop Diagram
Clean Power Plan
Community Viz 2.0
Durham Area Transit Authority (now called GoDurham)
Durham-Chapel Hill-Carrboro Metropolitan Planning Organization
Draft Environmental Impact Statement
Durham-Orange Light Rail Project
Environmental Impact Analysis
Environmental Impact Statement
Event Mean Concentration
Floor Area Ratio
Federal Transit Administration
Fiscal Year
Greenhouse Gas
Geographic Information System
Gross Operating Surplus
Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation Model
Gross Regional Product
Health Economic Assessment Tools
Herfindahl-Hirschman Index
Health Impact Assessment
U.S. Dept. of Housing and Urban Development
International Energy Agency
Longitudinal Employer-Household Dynamics (from U.S. Census Bureau)
Log Into North Carolina
LEHD Origin-Destination Employment Statistics (from U.S. Census Bureau)
Light Rail Project
Light Rail Transit
Multifamily
Motor Vehicle Emission Simulator
Miles Per Gallon
Metropolitan Planning Organization
Metropolitan Transportation Plan
XVI
-------
MW Megawatt
N Nitrogen
NAICS North American Industrial Classification System
NC DENR North Carolina Department of Environment and Natural Resources
NC DHHS North Carolina Department of Health and Human Services
NC DOR North Carolina Department of Revenue
NC DOT North Carolina Department of Transportation
NC DST North Carolina Department of State Treasurer
NC ESC North Carolina Employment Security Commission
NEPA National Environmental Policy Act
NERL National Exposure Research Laboratory
NHGIS National Historical Geographic Information System
NOAA National Oceanic and Atmospheric Administration
NMT Nonmotorized Travel
NTD National Transit Database
O&M Operations and Maintenance
ORD Office of Research and Development
P Phosphorus
PTT Public Transit Trip
QA Quality Assurance
SD System Dynamics
SE Socioeconomic
SF Single-family
SHC Sustainable and Healthy Communities Research Program
SIA Sustainability Impact Assessment
SLD Smart Location Database
SMP Sustainable Mobility Proj ect
TAZ Traffic Analysis Zone
TCRP Transit Cooperative Research Program
TIGER Topologically Integrated Geographic Encoding and Referencing
TJCOG Triangle J Council of Governments
TRM Triangle Regional Model
TTA Triangle Transit Authority (later changed to Triangle Transit, then GoTriangle)
UNC University of North Carolina
US BLS U.S. Bureau of Labor Statistics
USD United States Dollar
US DOE U.S. Department of Energy
US DOT U.S. Department of Transportation
US EIA U.S. Energy Information Administration
US EPA U.S. Environmental Protection Agency
USGS U.S. Geological Survey
VMT Vehicle Miles Traveled
WHO World Health Organization
XVII
-------
Authors
This report was jointly produced by U.S. EPA's Office of Research and Development, National Exposure
Research Laboratory, and Industrial Economics, Inc:
Rochelle Araujoa, Andrea Bassib, Llael Coxc, Nicholas Flanders0, Jenna Kollingc, Andrew Procter0, Nadav
Tannersd
aU.S. EPA ORD National Exposure Research Laboratory, bKnowlEdge Sri, C0ak Ridge Institute for Science
and Education, dlndustrial Economics, Inc.
Acknowledgements
The model and report would not have been complete without the efforts and contributions of many individuals
and organizations. The following individuals provided information, data, advice, or review:
Ben Bearden (TJCOG), Todd BenDor (University of North Carolina at Chapel Hill), Amelia Bryan (NC Dept.
of Revenue), Mel Downey-Piper (Durham County Public Health), Krystyna Stave (University of Nevada, Las
Vegas), Tobin Freid (Durham City/County Sustainability Office), Geoff Green (GoTriangle), Jennifer Green
(GoTriangle), Teresa Hairston (Durham County Tax Administration), Kristen Hassmiller Lich (University of
North Carolina at Chapel Hill), Andy Henry (Durham City/County Transportation Planning), John Hodges-
Copple (TJCOG), Joe Huegy (ITRE), Hannah Jacobson (Durham City/County Planning Department), Amy
Lamson (U.S. EPA), Patrick McDonough (GoTriangle), Timothy Mulrooney (North Carolina Central
University), Felix Nwoko (DCHC MPO), Gavin Poindexter (AECOM), Kimberly H. Simpson (Durham
County Tax Administrator), Jeff Weisner (AECOM), Laura Dale Woods (Durham City/County Planning
Department), and Yanping Zhang (DCHC MPO).
The following organizations participated in our stakeholder meetings, informing the development of the
model:
Capital Area Friends of Transit, Chapel Hill Office of Planning and Sustainability, Chapel Hill Transit,
Durham City/County Planning Department, Durham City/County Sustainability Office, Durham County Public
Health Department, Durham-Chapel Hill-Carrboro Metropolitan Planning Organization (DCHC MPO),
Durham Coalition for Affordable Housing and Transit, Durham Neighborhood Improvement Services,
GoTriangle, Institute for Transportation Research and Education (ITRE), and Triangle J Council of
Governments (TJCOG).
Dr. Gary Foley and Dr. Marilyn TenBrink of US EPA, Office of Research and Development have championed
the application of system dynamics modeling to Sustainability issues among EPA partners and stakeholders
and provided support to the present project.
Although these individuals and organizations were helpful in the preparation or review of this model and
report, these efforts do not constitute an endorsement of the results of this report or of any U.S. EPA policies
and programs.
Questions concerning this report should be directed to:
Rochelle Araujo
National Exposure Research Laboratory (NERL), U.S. Environmental Protection Agency,
Mail Code D305-01, 109 T.W. Alexander Drive, Research Triangle Park, NC 27709, 919-541-4109
Araujo.rochelle@Epa.gov
XVIII
-------
Executive Summary
Project Background
EPA's Sustainable and Healthy Communities Research Program (SHC) is conducting
transdisciplinary research to inform and empower decision-makers. EPA tools and approaches
are being developed to enable communities to effectively weigh and integrate human health,
socioeconomic, environmental, and ecological factors into their decisions to promote community
sustainability. To help achieve this goal, EPA researchers have developed systems approaches to
account for the linkages among resources, assets, and outcomes managed by a community.
System dynamics (SD) is a member of the family of systems approaches and provides a
framework for dynamic modeling that can assist with assessing and understanding complex
issues across multiple dimensions. To test the utility of such tools when applied to a real-world
situation, the EPA has developed a prototype SD model for community sustainability using the
proposed Durham-Orange Light Rail Project (D-O LRP) as a case study.
The EPA D-O LRP SD modeling team chose the proposed D-O LRP to demonstrate that an
integrated modeling approach could represent the multitude of related cross-sectoral decisions
that would be made and the cascading impacts that could result from a light rail transit system
connecting Durham and Chapel Hill, NC. In keeping with the SHC vision described above, the
proposal for the light rail is a starting point solution for the more intractable problems of
population growth, unsustainable land use, environmental degradation, and the persistence of
economic, social, and health inequities. To achieve the maximum potential benefits from the
light rail across all of the dimensions of sustainability while reducing its potential negative
consequences, concurrent policies must be weighed in combination with the light rail to assess
the tradeoffs associated with these decisions. Therefore, the D-O LRP SD modeling team
developed many concurrent policy scenarios in addition to the light rail that can aid stakeholders
in finding leverage points within the system where interventions can have the largest impact.
In the first phase of this modeling effort, a conceptual model for the D-O LRP was designed with
a high degree of input from stakeholders, including representatives from the regional transit
authority, county health department, stormwater management department, and city and regional
land use and transportation planning departments, among others. This conceptual model served
as a framework for the operational SD model, which was built to evaluate a number of policy
scenarios, many of which were also suggested by stakeholders. The operational model was
subjected to rigorous quality assurance tests, including the sensitivity of the model to
assumptions and inputs, and the evaluation of outcomes - social, economic, and environmental -
resulting from actions that emanate from or impinge on the D-O LRP.
ES-1
-------
Model Structure
The D-O LRP SD Model was calibrated using
historical data and local projections, when
available, for its two geographic boundaries:
Tier 2 - the area defined as being within the
boundaries of the Durham-Chapel Hill-
Carrboro Metropolitan Planning Organization
(DCHC MPO); and Tier 1 - the combined area
of/^-mile-radius zones surrounding each of the
proposed light rail stations (/^-mile radius was
chosen because it is common practice among
urban planners to regard a half mile as the
greatest distance that most people are willing to
walk to a public transit station).
Figure ES-1. Map of the D-O LRP SD Model
Geographic Tiers
Tier 2: DCHC MPO
ccording to TRM TAZs)
Tier 1: V? mile radius
around proposed light
rail stations
Model variable outputs are reported for each
Tier annually between 2000 and 2040. The
model is designed to explore dynamic
interactions among sectors of the urban system, including land use, transportation, energy,
economics, equity, water, and health. These sectors are visualized in Figure ES-2, with plus (+)
signs indicating a positive association between variables (an increase in A produces an increase
in B, and a decrease in B produces a decrease in B), and minus (-) signs indicating a negative
association between variables (an increase in A produces a decrease in B, and vice versa). Model
scenarios run in a few seconds, and users can edit inputs or equations for any variable in the
model.
Health^
Stormwater Pollutant
Loading
Nonmotorized
Person Miles
^k.
Public Transit
Person Miles
H & T Costs
Property
Values
•^
L
Households
in Poverty
ortation
Equity/
Figure ES-2. Causal Loop Diagram (CLD) for the D-O LRP SD Model
ES-2
-------
Model Scenarios
Three main scenarios were run in the D-O LRP SD Model to reflect the most likely
transportation and land use plans.
1. Business-As-Usual (BAU) scenario
The BAU scenario represents expected results if current demographic, land use, and
transportation trends continue and serves as a baseline to contrast with the other scenarios
described below.
2. Light Rail scenario
The Light Rail scenario represents the implementation of the 17-mile light rail transit (LRT)
line by 2026 between Durham and Chapel Hill and also deviates from the BAU scenario as
follows:
• Assumes LRT motivates more people to use public transit than an equal number of bus
service miles;
• Assumes a 10% increase in demand for developed nonresidential (excluding industrial)
floor space in Tier 1, gradually phased in during the six-year period of light rail
construction; and
• Assumes a higher share of Tier 1 employees will choose to move to Tier 1 rather than
commute from elsewhere in Tier 2.
3. Light Rail + Redevelopment scenario
The Light Rail + Redevelopment scenario represents the implementation of the LRT line
with additional changes to zoning to encourage land redevelopment and increased density
around the station areas.
• Assumes 20% of developed land is redeveloped to almost three times its existing density
by 2040, starting in 2020 in anticipation of the rail
Testable Interventions
The results of 17 additional policy, demographic, and market scenarios, were also analyzed to
demonstrate the breadth of policies and other factors that can be tested with the model. Some of
the main policy, demographic, and market levers that can be modified by users include:
• Policy Interventions
o Density and redevelopment
o Fare free transit
o Parking price changes
o Sidewalk building
o Clean Power Plan
o Stormwater management
• Demographic and Market Shifts
o More multifamily households
o Higher gas prices
o Change in wages
• Technology Changes
o Building energy efficiency
o Vehicle fuel efficiency
o Solar capacity
ES-3
-------
Main Findings
Findings from Model Construction
The model uses defensible explanatory mechanisms to approximate historical trends. We
used documented causal relationships between variables to drive behavior in the model. After
calibration, these mechanisms were adequate to reproduce historical trends, and they form the
basis of future projections. While each variable in the model can potentially influence most other
variables, Table ES-1 represents a select set of variables whose values are of high interest to
decision makers (labeled "Indicators") and the set of additional variables (labeled "Drivers") that
influence them most strongly in the model.
Table ES-1. A Selection of Model Indicators and their Drivers
Indicator
GRP
Employment
Productivity loss due to congestion
Nonresidential property values
Residential Property values
Poverty rate
Transit-dependent population
Affordability index
Net premature mortalities avoided
Person miles of public transit travel
Person miles of nonmotorized
travel
Energy use
CO2 emissions
Stormwater N and P loading
VMT
Plvhsand NOX emissions
Model Drivers
Economic
Earnings, nonresidential sq ft, gross operating surplus persq ft,
energy spending, congestion
Labor force, GRP, retail consumption
VMT, congestion, per capita earnings
Employment growth, retail density, building size
Land availability, income growth, commute time, population growth,
lot size, retail density
Social
Unemployment rate
Population in poverty
Renter costs, vehicle costs, transit costs
VMT, NOx and PM2 .5 emissions per VMT, accidents per VMT, person
miles of nonmotorized travel
GRP, population, fare price, revenue miles, price of gasoline, MPG,
traffic congestion, travel by other modes
GRP, population, nonmotorized travel facilities, jobs-housing balance,
price of gasoline, MPG, traffic congestion, travel by other modes
Environmental
Building stock, building energy intensity, VMT, MPG
Energy use, emissions intensity of electricity generation, solar
capacity
Developed land, impervious surface, stormwater mitigation (e.g. rain
gardens)
GRP, population, population in zero-car households, price of
gasoline, MPG, traffic congestion, travel by other modes
VMT, emissions per VMT
The model has facilitated interactions among diverse stakeholders to address complex,
interconnected community issues. The CLD was developed based on the input of a diverse
group of stakeholders, including local land use and transportation planners, sustainability
experts, and public health leaders. This stakeholder group also provided feedback on preliminary
model capabilities and results, which drove revisions and additions to the model.
ES-4
-------
Selected Scenario Results
Economic
Job growth in Tier 1 due to the light rail is accompanied by greater traffic congestion
despite decreases in per capita vehicle miles traveled (VMT). When the light rail opens in
2026, there is a shift towards more transit use, decreasing VMT per capita by residents of Tier 1,
as shown in Figure ES-3 A. However, the D-O LRP SD Model assumes the light rail will increase
demand for nonresidential floor space in the station areas, which leads to more employment
growth between 2020 and 2040 than the BAU scenario (53% vs. 35%, Figure ES-4). This
economic growth spurs population growth by encouraging immigration to the area and leads to
increases in total VMT and congestion in Tier 1. Congestion sharply declines in 2026 due to the
introduction of the light rail line (Figure ES-3B); however this decline is offset within four years
by the increased traffic due to economic and population growth.
(A)
(B)
1.30
1.25
1.20
1.15
1.10
1.05
1.00
2000 2010 2020 2030 2040 2°°° 2010 2020 2030
Light Rail + Redev Light Rail BAU Light Rail + Redev ——Light Rail
Figure ES- 3. Light Rail Scenarios Compared to BAU for: (A) VMT by Tier 1 Residents per Day per
Capita, and (B) Tier 1 Congestion
GRP Employment Population Total VMT
125%
BAU
• BAU • Light Rail • Light Rail + Redev
Figure ES- 4. Change in Tier 1 Overall Growth Indicators Between
2020 and 2040
More of the potential economic benefits of light rail are realized in the Light Rail +
Redevelopment scenario, due to compact redevelopment in Tier 1 Denser development
allows demand for nonresidential square feet to be met, causing gross regional product (GRP) to
grow by 114% between 2020 and 2040, compared to only 89% under Light Rail alone (Figure
ES-4).
ES-5
-------
Social
Compact redevelopment increases nonresidential property values far more than residential
property values in Tier 1, providing a win-win for tax revenues and housing affordability. In
real terms, residential property values in the Light Rail + Redevelopment scenario are no more
than 7% higher than BAU in 2040, while nonresidential property values increase by 136% more
than under the Light Rail scenario in Tier 1 between 2020 and 2040 (Figure ES-5A). The rise in
multifamily property values, coupled with an increase in vehicle costs, drives decline in the
affordability index of 4.8% by 2040 in the Light Rail + Redevelopment scenario relative to BAU
(Figure ES-5B). However, the rise in nonresidential property values leads to $660M more in
cumulative real property taxes (PT) levied between 2020 and 2040 than under the Light Rail
scenario in Tier 1 alone (Figure ES-5A).
(A)
Single Family Multifamily
Property Value Property Value
per DU per DU
• BAU
Nonresidential
property value
per Sq Ft
Cumulative City
+ County PT
Levied
2000
Light Rail • Light Rail + Redev
•Light Rail + Redev
2020
2030
•Light Rail
•BAU
Figure ES- 5. Tier 1 (A) Change in Single Family, Multifamily, and Nonresidential Property Values
between 2020 and 2040 in Tier 1, (B) Affordability Index: Main Policy Scenarios Compared to BAU
Residents' health improves due to more walking and cycling under the light rail scenarios.
Within the context of an overall declining trend in nonmotorized travel per capita, the light rail
encourages more walking and cycling relative to BAU (Figure ES-6A). Consequently, premature
mortalities avoided due to an active lifestyle increase, resulting in 46 (Light Rail) and 54 (Light
Rail + Redevelopment) cumulative additional avoided premature mortalities (Figure ES-6B).
This net health improvement reflects that the benefits of increased physical activity outweigh the
negligible impacts of increased PIVb.s and NOX vehicle emissions and the slight increase in
vehicle crash fatalities.
(B)
PM2.5 and NOx
Vehicle
Emissions
Net Premature
Walking and Mortalities
Cycling Avoided
2000
2010
•Light Rail + Redev
•Light Rail
•BAU
iht Pail 4. Da/Ho
Figure ES- 6. Health Effects in Tier 1: (A) Nonmotorized Travel by Residents per Day per Capita
and (B) Cumulative Premature Mortalities Avoided by Cause and Net Cumulative Premature
Mortalities Avoided between 2020 and 2040 for the Light Rail Scenarios: Departure from BAU
ES-6
-------
Environmental
Intensity indicators demonstrate improvements in resource use efficiency under the Light
Rail + Redevelopment scenario in Tier 1 despite economic growth. In 2040, impervious
surfaces per capita in Tier 1 are 19% lower in the Light Rail + Redevelopment scenario than in
BAU (Table ES-2). Daily water demand per capita in Tier 1 is also improved in the Light Rail +
Redevelopment scenario, at 1.8% lower than BAU. On the other hand, CO2 emissions per dollar
of GRP increase in the Light Rail + Redevelopment scenario by 3% relative to the BAU. This is
due to redevelopment increasing nonresidential use, which is more energy intensive than
residential use.
Table ES-2. Selected environmental intensity measures across the BAU, Light Rail, and Light
Rail + Redevelopment scenarios
2040 Value
Light Rail Light Rail + Redev
2040Value %diff from BAD 2040Value %diff from BAU
Tierl
Intensity Measu
Impervious surface (acres) per capita
CO2 Emissions per GRP (tons/million USD 2010)
Daily water demand (Mgal/year) per capita
Determining the cumulative environmental impacts of scenarios requires viewing model
results as a time series. Although Light Rail and Light Rail + Redevelopment would appear to
have the same impact on CO2 emissions by 2030 (Fig ES-7A), they diverge afterward, leading
Light Rail + Redevelopment to have the highest cumulative CO2 emissions by 2040. The
situation is opposite for stormwater N load (Fig ES-7B): although the two light rail scenarios
have similar stormwater N load in 2040, Light Rail sustains this load for a longer time, so its
cumulative stormwater N load is higher than Light Rail + Redevelopment by 2040.
(A) CC>2 Emissions
(B) Stormwater N Load
2000 2010 2020 2030 2040
•Light Rail+Redev — —Light Rail
•BAU
2000 2010
•Light Rail+Redev — —Light Rail
•BAU
Figure ES- 7. Environmental impacts in Tier 1: Light Rail, Light Rail + Redevelopment Scenarios
Compared to BAU
Compact redevelopment increases energy use and COi emissions intensity in Tier 1, due to
unlocking growth potential and increasing nonresidential uses. CO2 emissions reductions
strategies and stormwater management policies such as the Durham GHG Plan, the Clean Power
Plan, and the Jordan/Falls Lake Rules can help offset the environmental impacts of growth
ES-7
-------
described above. If applied to our model, emissions goals set by the Clean Power Plan would
drop Tier 1 emissions by 16% from their projected 2030 level (Figure ES-8A). A stormwater N
mitigation plan that would treat 30% of the stormwater N load from development after 2015 and
15% of the load from development existing before 2015 could cause stormwater N load to level
off in Tier 1 (light blue line in Figure ES-8B).
(A) Clean Power Plan (CPP)
(B) Stormwater N Mitigation
70
60
™
40
30
20
10
2000 2010 2020
-CPP Light Rail+Redev
2030
2040
-Light Rail+Redev
0
2000 2010 2020 2030 2040
—30pct new 15pct existing N load treated Light Rail+Redev
^30pct new N load treated Light Rail+Redev
Light Rail+Redev
Figure ES- 8. Mitigating the Environmental Effects of Light Rail + Redevelopment in Tier 1
Regional impacts of the light rail are quantified by the D-O LRP SD Model, but those
impacts are less pronounced than in the station areas. In terms of population, energy use, and
water demand, Tier 2 is between seven and eleven times the scale of Tier 1. Therefore, the
cascading impacts of the light rail are diluted in Tier 2. For example, projected annual energy use
in 2040 is 10% higher in the Light Rail scenario compared to BAU in Tier 1, but only 3.9%
higher than BAU in Tier 2. Similarly, nonmotorized travel by residents per capita is 15% higher
in the Light Rail scenario than BAU in 2040 in Tier 1, but only 1% higher in Tier 2.
Limitations of the Model
The D-O LRP SD Model illustrates trends and relative magnitude, not predictions. System
dynamics models are intended to explore the complexity and interactions within a system, rather
than produce an exact answer to a given question. Although our model explores policies and
scenarios related to the role of light rail transit in sustainable regional development, our model
does not provide specific directions to urban planners. Rather, it shows potential future trends,
relative magnitudes of impact, and interactions among different sectors of the urban system.
The model is not designed to work under extreme conditions or past the year 2040. Model
results have been extensively tested for inputs within reasonable value ranges, and model
parameters have been set to work within these boundaries. Extreme values, changes to historical
inputs, or extrapolation of results past the model timespan could produce unrealistic outputs.
The model is not spatially explicit. All inputs and indicators are aggregated to the level of the
two modeled tiers.
ES-8
-------
Intended Community Value
The system dynamics modeling approach can add value to three types of community processes:
1) Regulatory process - applying a multisector SD model like the D-O LRP model could
allow the Indirect and Cumulative Impacts section of an Environmental Impact Statement
(EIS) to have more quantitative projections. Currently the Indirect and Cumulative
Impacts of the D-O LRP EIS are described as general increases or decreases. Applying a
SD model would allow the regulatory process to consider the relative size of increases or
decreases, so that tradeoffs could be weighed.
2) Urban planners and planning process - interactions between transportation, health and
sustainability planners could be facilitated through a model that integrates their various
sectors. In urban planning, actions in one department may compete with or counteract the
interests of another department or agency. Developing and using a stakeholder-invested
model could help departments to coordinate their efforts, or at least visualize how the
actions of one department influence the interests of other departments.
3) Public discussion - an SD model can help explore issues that citizens raise at public
meetings. A large-scale public project such as light rail transit requires an education and
outreach effort to maintain public support. An SD model like the D-O LRP model can
help citizens visualize the various ways in which a project like light rail transit can affect
regional development.
Next Steps
• Model Version 2.0. An updated version of the model is currently under development. It
will include, an updated calibration of employment for the BAU scenario in Tier 1,
updated estimates of the elasticities of person miles (by each mode of transportation) to
GRP per capita, and additional model enhancements pending our interactions with
stakeholders.
• User Interface. A more user-friendly interface will be developed to allow stakeholders to
test assumptions and policy interventions, access relevant background information, and
view indicators for side-by-side comparison.
• Transferable Tools. Ultimately, the experiences gleaned from this and other ORD and
regional efforts to employ systems approaches will be consolidated into tools and
guidance for communities and regions to apply these methods to a wide range of
sustainability issues.
ES-9
-------
1 Introduction
1.1 Project Overview
As our understanding of the nature of sustainability issues evolves, so must the tools and approaches we
use to address them. It is now widely recognized that systems previously treated discretely—air, water,
land, transportation, energy, health and well-being, quality of place, the economy, and social equity—
are inextricably intertwined. By acknowledging that human society exists within highly complex and
interdependent systems, each potentially valued subjectively, scientists can provide approaches that
transcend the deficiencies of traditional reductionist approaches, such as failure to account for linkages,
cross-system impacts, and unanticipated consequences. The United States Environmental Protection
Agency (US EPA) Sustainable and Healthy Communities Research Program (SHC) has committed to
developing systems approaches that enable communities to act on an enhanced understanding of these
interconnections and to comprehensively account for the full costs and benefits of community decisions
in the social, economic, and environmental dimensions. This report documents the development and
testing of a system dynamics (SD) model as a decision support tool for community sustainability with
the proposed Durham-Orange Light Rail Project (D-O LRP) in Durham and Orange Counties, North
Carolina, as a case-study.
The proposed D-O LRP, a 17-mile light rail transit system that would connect the city of Durham, NC
and the town of Chapel Hill, NC, provides a concrete example for understanding how a SD model could
be used to explore the interactions among decision sectors within a community and presents an
opportunity for identifying points in the system where coordinated actions could yield a greater net
value and possibly offset unintended consequences. The primary goals of the D-O LRP are to enhance
mobility, capture untapped markets for transit use, and support desired development patterns in the
region (Triangle Transit 2012a). If these goals are achieved, this project could have many positive
impacts along the rail corridor. There are concerns, however, that the project could have negative
impacts in some areas, such as displacement of low- to moderate-income households by increasing
property values and intensification of urban stormwater runoff through higher-density development. It is
therefore crucial for decision-makers to be able to identify how the project might interact with other
sectors, such as land use change, housing, and water resource management, to maximize benefits and
avoid or limit negative consequences.
To illustrate the interconnections in the complex system that would be perturbed by the D-O LRP, the
D-O LRP SD modeling team has developed a prototype SD model that can help address the concerns
listed above. In Phase I of this project, the modeling team developed a conceptual model for the system
potentially affected directly and indirectly by the D-O LRP. This effort was made possible by many
collaborations with local stakeholders, described in detail in Chapter 2 of this report. In Phase II of the
project, the modeling team built a computer model using relationships derived from literature and
calibrated it to historical data and projections (local, to the extent possible) for each of the seven high-
priority decision sectors identified by stakeholders: land, transportation, energy, economy, equity, water,
and health. Chapter 3 of this report includes an overview of the primary data sources used in the model,
a description of the overall model structure, focusing on inter-sector relationships and feedbacks, and
by-sector descriptions of intra-sector relationships and feedbacks, data sources and processing, and
-------
calibration steps.1 In Phase III of the project, the modeling team constructed model scenarios based on
stakeholder input, described in detail in Chapter 4. These scenarios were designed to not only test the D-
O LRP as a policy against a "business as usual" (BAU) case, but also test additional decisions across
sectors that could potentially maximize the benefits of the light rail or minimize undesired
consequences. The results of these scenarios are given in Chapter 5, and the results of extensive QA
model testing are shown in Chapter 6. Ultimately, the goal of this modeling effort is to demonstrate that
a SD model constructed for a representative issue of community sustainability (such as the D-O LRP)
will contribute to a more generalizable modeling approach that can help community decision-makers
and stakeholders explore alternative policy scenarios for a wide variety of sustainability problems, as
well as test the assumptions on which those scenarios are based.
1.2 Project Background
This section describes how this project fits within the US EPA SHC program and how systems
approaches are necessary for understanding complex, large-scale problems. It then introduces SD
models and discusses why such models are appropriate for complex problems. Finally, it provides
background information on the D-O LRP, the project to be examined by this model.
The Sustainable and Healthy Communities (SHC) Research Program
American communities face complex challenges as they try to strengthen their economies, meet
changing demand for housing and transportation, and protect the environment and public health.
Following the strategic realignment of the US EPA Office of Research and Development (ORD)
research programs in 2010, the agency has placed a stronger focus on transdisciplinary research to
support the growing interest of communities in sustainable practices (Anastas 2012). Providing the
scientific foundation to address the complex and multi-dimensional problems communities face is at the
heart of SHC's mission. As stated in the US EPA's 2012 report, "Sustainable and Healthy Communities
Strategic Action Plan 2012-2016," the overall vision of SHC is:
"to inform and empower decision-makers in communities, as well as in federal, state and tribal
community-driven programs, to effectively and equitably weigh and integrate human health,
socio-economic, environmental, and ecological factors into their decisions in a way that fosters
community sustainability." (US EPA 2012)
At the Federal level, the recognition that these issues are interconnected has resulted in the formation of
the Federal Partnership for Sustainable Communities, which includes the US EPA, the U.S. Department
of Transportation (US DOT), and the Department of Housing and Urban Development (HUD).
However, Federal agencies and policies represent only a fraction of the interests that must be engaged to
solve community problems. Thus, SHC has committed to developing the information, tools, and
approaches needed to promote the collaboration required for communities to evaluate problems,
proactively assess decision alternatives, implement more effective solutions, and track results (US EPA
2012). While much progress has been made in the development, use, and integration of science-based
1 For a detailed more detailed description of the structure of each sector, including calibrated variables, exogenous inputs, and
equations from literature, see Appendix B.
-------
assessment tools and approaches to solve complex environmental challenges, there remains a need for
additional assessment tools to gain a greater understanding of the inter-connected relationships that exist
among social, economic, and environmental dimensions when addressing these challenges.
Criteria for Selecting a Modeling Approach
The goal of this SHC research project was to develop and evaluate the use of a tool that captured the
dynamics among community decisions and the trends underlying the challenges that communities face
in such a way that individual decisions could be evaluated in context and multiple decisions could be
aligned to greater net benefit. In reviewing the range of candidate model types, we identified several key
capabilities that the model needed to have:
1. The model must represent social, economic and environmental processes.
2. The model must account for actions that lead to multiple consequences, as well as outcomes that
are fed by multiple actions.
3. Social and economic processes must have a more complex structure than simple "feed-forward"
- that is, social and environmental consequences must have feedbacks to the system, allowing
them to be drivers as well as consequences.
4. The model must allow for nonlinear responses, i.e. tipping points that cause a shift in the
magnitude of a relationship.
5. Feedback mechanisms (with an emphasis on mechanisms that can be modified by policies or
actions) are included in at least the transportation, economy, and land use sectors.
6. The model must incorporate and reflect stakeholder concerns while simplifying them sufficiently
enough to be integrated into a complex model depicting many community processes.
As indicated by the criteria listed above, an integrated assessment of sustainable systems cannot be
accomplished by simply linking together a collection of domain-specific models. Assessing the higher-
level interactions among interdependent systems requires tools that specialize in capturing the emergent
behaviors and dynamic relationships that characterize complex, adaptive systems (Fiksel 2006). One
such tool is system dynamics.
System Dynamics (SD) Models
System dynamics (SD) is a member of the family of systems approaches and provides a framework for
dynamic modeling that can assist with assessing and understanding complex issues across multiple
dimensions. SD models can be critical tools for investigating the interdependence, nonlinear responses,
feedback loops, and dynamic behaviors of complex systems. The process of constructing a SD model
lends itself well to stakeholder engagement because it promotes the understanding and conceptualization
of a problem, allowing those who participate in the process to view the behavior of the system and
understand how the feedbacks and mechanisms within the model reflect and give rise to its behavior.2
SD is a policy-oriented modeling technique that provides a framework for the design of policies and
2 See Chapter 2 for a more detailed description of the process that was followed in building our SD model with the help of
local stakeholders.
-------
management of systems to achieve improved system behavior. However, SD models do not provide an
explicit answer to a problem and are not meant to replace decision-makers or to inhibit their role in
making decisions; rather, they are developed to help and support people in achieving better decisions
(Abbas and Bell 1994).
SD modeling offers a whole-system approach to transportation and land use planning, illustrating how
decisions regarding such large-scale projects will have cascading direct and indirect impacts across
many sectors, including human health and the natural environment. Because SD models meet the criteria
listed above, we decided that they offered the best modeling style for developing a tool that meets the
research goals of SHC. To ensure that such a tool would have utility when applied to real-life problems,
we elected to develop our SD model based on a light rail transit system proposed for our own
community, Durham, NC, thus achieving two objectives: (1) providing immediate, practical decision
support to real community issues related to sustainability, and (2) providing a test case that motivates the
development of the model, demonstrates the "proof-of-concept" for this modeling approach, and
provides "lessons learned" for transferring the approach to other communities and/or problem types.
The Durham-Orange Light Rail Project (D-O LRP)
The Triangle Region of central North Carolina (which connects the cities of Raleigh, Durham, and
Chapel Hill) has been a hot spot for population growth and development over the past few decades due
to a boom in employment and educational opportunities. While this rapid expansion has boosted
economic prosperity for the region, it has also put significant stress on infrastructure and the natural
environment. In a region that has become known for high average vehicle miles traveled (VMT) for
work commuters, the vast majority of people are solely reliant on automobiles for transportation,
overtaxing the existing roadway infrastructure. Projected levels of population growth and land
conversion will not only exacerbate this problem, but will also lead to increased consumption of
materials, energy, and water, and increased levels of air and water pollution. To address these issues,
and to promote health and prosperity, a light rail transit system has been proposed to connect the town of
Chapel Hill (located in Orange County, NC) and city of Durham (located in Durham County, NC) along
a heavily-used corridor. The proposed light rail station locations are pictured in Figure 1-1.
Planning for fixed-guideway transit in the Triangle Region began over 20 years ago, and a number of
transit studies have been conducted to advance major transit investments in the area.3 Evaluation of
fixed-guideway alternatives such as bus rapid transit (BRT) and light rail transit (LRT) in the Durham-
Orange corridor began with the 1998-2001 US Highway 15-501 Major Investment Study, which
included extensive public involvement and resulted in the establishment of an adopted transit corridor
between Chapel Hill and Durham that continues to be protected and preserved for transit use by the local
governments (Triangle Transit 2012b). In 2011 and 2012, respectively, Durham and Orange County
residents voted in favor of a half-cent sales tax dedicated to transit, and in 2012, Triangle Transit (now
referred to as GoTriangle), the local project sponsor for the D-O LRP, completed an alternatives analysis
for the Durham-Orange corridor as the first step in the Federal Transit Administration (FTA) Project
Planning and Development process. The alternatives analysis concluded by identifying a locally
preferred alternative, LRT, as the only technology that satisfied the criteria of enhancing mobility,
expanding transit options between Durham and Chapel Hill, serving populations with a high propensity
for transit use, and fostering compact development and economic growth (Triangle Transit 2012a). Soon
after, the Durham-Chapel Hill-Carrboro Metropolitan Planning Organization (DCHC MPO)
3 A collection of the transit studies done over the past twenty years can be found at http://ourtransitfuture.com/library/.
4
-------
Transportation Advisory Committee unanimously adopted the LRT alternative as the locally preferred
alternative for further study through Preliminary Engineering and the National Environmental Policy
Act (NEPA) review process.
Legend
Preliminary NEPA Preferred Alignment
Q Park and Ride
© Proposed station
—,, Rail Operations and Maintenance
Facility site under consideration
^^^— LRT alignment
•^^^ Aerial sections of LRT
Other alternatives evaluated
Other evaluated station
R Other ROMF site alternative
C1 Alternative
dAAIternative
C2 Alternative
New Hope Creek LPA Alternative
New Hope Creek Alternative 1
Durham
Martin Luther King jr Parkway
^v Transit
FUTURE-
Durham-Orange
Light Rail Project
For more information, visit:
» www.ourtransitfuture.com
HI P.O. Box 530 | Morrisville, NC | 27560
V. 800.816.7817
(5) info@ourtransitfuture.com
Mii«s NORTH
Map Data h. Esn. AECOM and oltiore Cop^qM MIS C
Figure 1-1. Map of the Proposed Durham-Orange Light Rail Project Station Locations
As part of the approval process for receiving federal grant money under the FTA New Starts program
(US DOT 2013), which includes an extensive environmental review process, GoTriangle released a
Draft Environmental Impact Statement (DEIS) on August 28, 2015, pursuant to NEPA (FTA 2015a).
The primary purpose of the DEIS is "to assist decision-makers and the public in assessing potential
impacts associated with the implementation of the proposed D-O LRP" (GoTriangle 2015). The DEIS is
circulated for review during a 45-day public comment period to interested parties, including members of
the public, community groups, the business community, elected officials, and public agencies. This
public comment period of the NEPA process provides a significant opportunity for the community to
question and engage with the underlying information and assumptions of the project and to express
concerns about direct and indirect impacts. Through the development of the D-O LRP SD Model, the
project team sought to provide useful input into this process.
1.3 Potential Uses for the D-O LRP SD Model
The D-O LRP SD Model has multiple potential uses in the D-O LRP planning and assessment process;
the model supports the current assessment process by augmenting the ability of NEPA's environmental
review process, also called an environmental impact assessment (EIA), to address cumulative and
indirect impacts. In the more general sense, we believe that models such as the D-O LRP SD model can
support the evolution of the EIA to better address a wider suite of sustainability-oriented considerations,
-------
enhance the integration of information and actions in the planning process, and better educate and
engage communities about the broader implications of projects and policies. In this section, we consider
the larger debate about integrated assessments, integrated planning, and community participation in
decision-making as reflected in the literature and also present, in text boxes, specific examples of how
those issues play out in the local context of the D-O LRP.
Integrated Assessments for Environmental Review
Since its incorporation into environmental decision making by NEPA in 1969, EIA has been adopted by
100 countries worldwide. Intended to be both anticipatory and participatory, EIA is a systematic process
that considers whether a project will have adverse environmental impacts, whether it should proceed,
and—if so—what mitigatory measures should be adopted. An Environmental Impact Statement (EIS) is
a full-disclosure document that forms part of the EIA process in the U.S.; a Draft EIS is made available
for public comment, followed by a Final EIS, and, finally, a Record of Decision. In this section, we will
refer to the EIA as the general process, in order to encompass its uses beyond the U.S., reflect the
ongoing debates about its uses and suitability in the scientific and policy literature, and maintain a
parallelism with alternate forms of Integrated Assessments, such as Health Impact Assessments (HIA)
and Sustainability Impact Assessments (SIA).
Increasingly, EIA is being positioned in a broader context of Sustainability, raising questions about its
effectiveness and suitability for that purpose (e.g. Jay et al. (2007), as well as the extent to which it
reflects evolving community values and engagement in the assessment process itself (Weston 2004).
Although NEPA calls for the use of "all practical means and measures.. .to create and maintain
conditions under which man and nature can co-exist in harmony and fulfill the social, economic and
other requirements of present and future generations of Americans" (FT A 1969), Section 101(a)), the
EIA process has, for the most part, been narrowly focused on and based on technical, "objective" tools
and approaches. The push to address Sustainability concerns, which harks back to the original NEPA
language, has largely taken the form of elucidating indirect and cumulative impacts, including social and
economic impacts, within the EIA (or the product of the EIA - the EIS). However, because social and
economic impacts reflect a wider array of endpoints, many of which are difficult to measure objectively
or with certainty, the treatment of indirect and cumulative impacts tends to be less quantitative than the
treatment of the direct effects of a project.
NEPA legislation directs federal agencies to examine indirect and cumulative effects but does not
prescribe a specific methodology. The Federal Transportation Agency's Planning Assistance and
Standards (Title 23 CFR Part 450) indicate that indirect and cumulative effects should be sufficiently
detailed such that consequences of different alternatives can be readily identified, based on current data
and reasonable assumptions, and based on reliable and defensible analytic methods (FTA 1969). Indirect
effects are those that occur either later in time or outside of the immediate geography of the project, and
may include changes in the pattern of land use, changes in population, and related effects on air, water,
and other natural systems. Cumulative impacts can result from individually minor but collectively
significant actions taking place over time (D-O LRP DEIS, Chapter 4.17) (GoTriangle 2015). When
impacts are understood to encompass social and economic considerations in addition to environmental
ones, the challenges of assessing indirect and cumulative impacts increase many-fold (Gibson 2006).
In recent years, numerous authors have critiqued the ability of the EIA process to address the broader
implications of environmental projects on Sustainability, a challenge made even more difficult when the
subject of the assessment is a policy or set of actions rather than a concrete and well-defined
construction project (Jay et al. 2007, Weston 2004). The same authors also critique the EIA process as
-------
prioritizing procedure over substance and as being inherently adversarial in nature. In recent years, HIAs
have been increasingly employed to assess and communicate a broader set of outcomes, including many
social and economic outcomes that contribute to community well-being and environmental justice, but
that may be difficult to quantify or predict using conventional indicators (Dannenberg et al. 2008). The
strengths of HIAs - especially their ability to assess programs or policies, (as opposed to discrete
projects) and their ability to consider community values as well as objectively-determined impacts -
contrast with the more technical, project-specific orientation of EIAs, leading some authors to speculate
whether HIAs might take the place of EIAs in the NEPA process (Cole et al. 2004).
MORE QUANTITATIVE TREATMENT OF INDIRECT AND CUMULATIVE IMPACTS WOULD SUPPORT
BETTER ESTIMATES OF NET BENEFITS AND UNDESIRED CONSEQUENCES
The D-O LRP DEIS, Chapter 4.17, addresses the potential indirect impacts of transportation on Land Use,
Economic Development, Visual and Aesthetic, Historic Resources, Natural Resources, Water Resources,
Hazardous and Regulated Materials Acquisitions, and Relocations and Displacements. The cumulative
impacts considered include Parking, Pedestrian and Bicycle Conditions, Land Use (Community Character),
Economic Development, Visual and Aesthetic, Habitat, and Water Quality. Whereas the bulk of the DEIS
addresses the expected direct impacts of the rail, this section of the document views transportation as a
potential catalyst for induced growth and e))conomic development in the area within 1/4 mile of the rail and
through the year 2040. The DEIS does not use In the absence of tools to evaluate feedbacks and estimate
the magnitude of deviations from projections due to such catalysis, so it does not include any quantitative
predictions of catalyzed growth or its consequences are included, nor does it characterize uncertainties.
Rather, it concludes that the quantity, type, location, and pace of growth are deemed to be consistent with
local land use plans based on narrative treatments and best judgment. Moreover, such descriptive
treatments of all indirect and cumulative impacts do not permit summing them with more quantitatively
estimated impacts, as would be desired determinations of net cost/benefits or the evaluation of ancillary
policies that might mitigate undesired or enhance desired indirect or cumulative impacts. (GoTriangle 2015)
Essentially, EIAs struggle to extend their data-driven, techno-rational approaches to the assessment of
indirect and cumulative impacts and to the broader suite of outcomes that contribute to emergent
properties such as sustainability. HIAs, in contrast, embrace a more value-driven and contextual
assessment of programs and policies, but struggle to bring quantitative approaches and rigor to the
assessment. Though each has its strengths, both fall short of an ideal process that would bring scientific
rigor to a decision that ultimately reflects societal values, that is both responsive enough to engage
community participation and robust enough to withstand legal challenges. What we present here can
help address the shortcomings of either EIAs or HIAs in the near term, but perhaps more importantly,
can combine the strengths of each to move toward the ideal process described above. In essence, a
systems dynamics model is a semi-quantitative, testable narrative. The story it tells is built on both
community perspectives and well-documented scientific information, creating a whole that is informed
by stakeholder values and wherein the magnitude and sensitivity of individual quantities and
relationships can be tested and evaluated with respect to the overall behavior of the system.
Integrated, Multi-Sector Approaches to Planning
Urban planning and environmental management communities have long acknowledged a range of
interactions among the economic, transportation, and land use sectors, as well as other sectors. To
inform planning efforts, managers have invested significant resources in understanding and modeling
these interactions. There is a rich literature documenting the impacts of one sector on another and
-------
supporting the utility of models for translating the dynamics of one sector into the inputs or drivers for
another. Wegener (2004) reviewed twenty models for the interactions of transport and land-use change,
noting the challenges of integrating processes that occur at very different time scales (e.g., network
construction vs goods movement). That study reviewed the models for comprehensiveness, structure,
theoretical foundations, dynamics, data requirements, calibration and validation, operationality, and
applicability. Based on this overview, Wegener identified challenges in linking the models to
environmental impacts and to social processes. In the years since this review, the challenges of linking
land use and transportation to environmental impacts (air emissions) have been addressed by numerous
air emission exposure models, as reviewed by Jerrett et al (2005). In most cases, regardless of the
approach taken to estimate emissions, such models do not dynamically represent interactions between
land use and transportation, but rather take various configurations of those two sectors as exogenous
inputs.
For the most part, the models reviewed by Wegener (2004) and Jerrett et al (2005) tend to be feed-
forward, linearly designed with drivers as contributory modules culminating in the ultimate outputs,
usually that sector of primary interest to the author. When feedbacks or bi-directionality of mechanisms
are represented, it is often as an iterative refinement of feed-forward methods. Few integrated models
capture the dynamic and multi-directional flow of information and impacts among the multiple
concurrent processes that determine the shape and the impact of urban systems.
Rickwood et al. (2007) made the case that population, land use, transport, water, and energy use all need
to be examined in an integrated fashion, not only to optimize or evaluate individual sectors, but rather to
improve the sustainability of urban systems as a whole. They propose integrating a number of topical
models within a framework (UrbanSim) to evaluate policies and trends as they affect urban
sustainability. Duran-Encalada and Paucar-Caceres (2009) also take a modeling approach to evaluate
urban sustainability, but rather than integrating separate models, they developed a system dynamics
model for Puebla, Mexico which they link to a range of policy impact techniques, such as
Environmental Impact Analysis, Community Impact Analysis, and Financial Evaluation, thus placing
the assessment of individual projects or policies in the broader context of urban sustainability. Fiksel et
al. (2014) outlined a systems approach that treats sustainability as a function of the net creation of value
among social, economic and environmental dimensions. They demonstrated that SD modeling was
useful in elucidating which processes within and among those dimensions interacted to yield net value.
So far, models have been examined for their ability to bring together information from different sectoral
domains to inform decisions, which could be achieved without contact among practitioners in each of
the sectors. However each knowledge domain also has its corresponding institutions and agencies; for
complex decisions, it may not be obvious which agencies have bearing on the issue, especially for
agencies primarily associated with indirect or cumulative impacts. Construction of an integrated systems
dynamics model has, as its first step, the development of a conceptual model, which can serve the
double purpose of linking issues together conceptually, but also mapping the relationships of individual
agencies and characterizing the interactions among them, thus allowing them to be brought into the
process early and in a way that focuses their participation on the high-value linkages and synergies.
Education and Enhancing Community Engagement
As models have become more complex frameworks for the assemblage and integration of disparate
information, their use has shifted somewhat away from direct application to calculate answers and
toward constructing narratives. Pfaffenbichler (2011) captures this shift through four identified uses for
models in the planning process:
8
-------
1. Models as eye-openers: can put new environmental issues on the political agenda
2. Models as arguments in dissent: used to challenge assessment by way of counter-expertise or
visualization of alternatives
3. Models as vehicles in creating consensus: used to help create a consensus view of a problem
among different stakeholders
4. Models for management: used to assist stakeholders in concrete policy decisions
These narratives, in turn, have come to be viewed as potent tools for engaging stakeholders in the
planning process (Foth et al. 2008).
EXPLICIT TREATMENT OF FEEDBACKS AND SYNERGIES AMONG SECTORS CAN ENHANCE
INTEGRATED PLANNING
Although the planning for the D-O LRP reflects a high degree of communication and coordination among
the regional business, utility and planning institutions, the ability of this process to account for feedbacks
and synergies is limited by the tools available. An example of this lack of feedbacks and synergies can
be seen in the model used to calculate the transportation demand for the light rail as part of the region's
2040 Metropolitan Transportation Plan (MTP), the Triangle Regional Model (TRM). The TRM used as
inputs information derived from a number of other sectors, notably demographics, economy and land
use. Based on community plans and data from local planning departments, the Office of State Budget
and Management, the U.S. Census Bureau, and independent forecasters, estimates of "base year"
(2010) and "plan year" (2040) population and jobs were developed for the area covered by the TRM.
Using a software tool called CommunityViz, "plan year" people and jobs were allocated to locations
deemed suitable for residential and nonresidential development (CAMPO and DCHC MPO 2013). The
suitability of locations for development reflected best professional judgment by local urban and
transportation planners as part of the Imagine 2040 effort and was consistent, in corresponding
scenarios, with existing comprehensive plans and watershed policies (TJCOG 2013). The allocations of
people and jobs were coupled with differing assumptions about transportation investments to produce a
limited number of scenarios for the TRM, which yielded travel demand by mode for each scenario.
The feed-forward nature of the process was not able to capture or capitalize on the potential for
dynamics in economy, land use, urban form and transportation to generate alternative representations of
population or employment as inputs. As a result, the TRM was run for a finite number of static scenarios,
and lacked a capability to test assumptions about how model internal processes might affect those
inputs. Moreover, processes that accounted for indirect and cumulative impacts of the Light Rail Project
(e.g. impacts on health, environmental quality) would also be expected to respond to dynamic feedbacks
within and among population, land use, economy, etc. Failing to include them in the model meant that
they could only be accounted for in the DEIS as non-quantitative descriptions of likely impacts.
-------
Stave (2002, 2010) has suggested that systems dynamics models, especially when developed in a
participatory setting, can enhance the role of stakeholders in environmental decision making. In most
cases of environmental decision making, stakeholders have divergent views on root causes of the
problem, goals for possible solutions, and which aspects of the problem to prioritize. Stave (2002) notes
that when public input is elicited for decision making, it often takes the form of decision-makers
distributing technical information, gathering public views, and then retreating behind closed doors to
make critical trade-offs among stakeholder interests. Uses of system dynamics models as a tool for
engaging with stakeholders creates a structure that allows decision makers and stakeholders together to
learn about the system, recognize the structural origins of the problem, and identify key policy levers for
the solutions. Moreover, stakeholder participation in the model construction can be documented by the
model process, especially if iterative versions of the model are retained, this creating a record of
stakeholder contributions to an evolving understanding of the problem and refinements of the policy
solutions.
Stave (2010) compared the quality of the decision-making process and the decisions that emerged from
four cases with differing degrees of public participation in the building of system dynamics models. She
found that greater participation in the model-building process generated more trust in the model, greater
buy-in to the consensus solutions and greater understanding of feedback interactions among system
components. Moreover, the decision process itself was enhanced and streamlined, because the
stakeholders did not continually re-visit model assumptions and scenarios, once they had been
incorporated into the model.
10
-------
WORKING FROM A COMMON MODEL COULD ENABLE AGENCIES TO ENGAGE COMMUNITY AND
CONVEY HOLISTIC VIEW OF THE PROJECT
A delegate at a recent meeting of the Durham Inter-Neighborhood Council (INC) recently remarked "When
we voted on the tax increase to support the construction of the Light Rail in 2012, we were told that it would
reduce congestion, reduce air pollution, increase jobs and increase mobility for the transit-dependent
population. Since then, we've come to realize that the rail itself isn't required to achieve those goals, which
we still support. Rather, the city is being transformed by a massive shift in planning, with high-rise, dense,
mixed-use required to justify the Light Rail overtaking our residential neighborhoods."
In the years since the idea of the LRT was introduced, it has become apparent that social, environmental,
and economic aspirations that had been attached to the rail will likely only be partly delivered by the rail
itself. Rather, the LRT will be a nucleating central event for the timing and location of a greater density of
urban development, and that that urban development, while disruptive, is actually required for and the
source of a necessary precondition for the greater proportion of the desired benefits. This realization is
causing a bit of dismay has caused concern among stakeholders that the environmental impact statement
and the meetings held by the transit agency (Triangle Transit, since renamed GoTriangle) to inform and
consult with public have focused rather narrowly on the impacts of the rail, neglecting with the attendant
redevelopment, which is being planned and communicated by the City/County Planning Departments.
Planning activities related to redevelopment include the designation of high-density, mixed-use Compact
Neighborhoods in the 1/2-mile vicinity of the proposed rail station locations. Complementary to Compact
Neighborhood designs are efforts to modify the non-motorized travel infrastructure in the form of the Station
Area Strategic Infrastructure Initiative (SASI), (Durham City-County Planning Department 2014), and
adoption of a Complete Streets approach (by the city, county and state) to street design. For the most part,
the public supports the goals of greater walkability/bike ability, and greater traffic safety, while grappling with
expected negative impacts, including the disruption to current residential and sometimes historic
neighborhoods, rising land prices and costs of housing, and the realization that likely increases in congestion
and environmental impacts in some areas congestion and environmental impacts are likely to increase.
While the interconnectedness of these issues was, to a large extent, acknowledged by the agencies
involved, it is only now starting to be appreciated by the public, which is decrying the lack of tools to evaluate
and mechanisms to influence the process.
11
-------
2 Model Development and Stakeholder
Process
2.1 Overview
Constructing a SD model is an iterative process where the modeler or model group, here referred
to as the D-O LRP SD modeling team, follows a step-by-step process, which can progress
through multiple cycles in which each step is refined in subsequent iterations (Abbas and Bell
1994). According to John D. Sterman, author of Business Dynamics (2000), "there is no
cookbook recipe for successful system dynamics modeling, no procedure you can follow to
guarantee a useful model." Even so, best modeling practices recommend a disciplined process
that involves the following phases:
1. Problem Definition: Articulating the problem to be addressed;
2. Conceptual Model Development: Formulating a dynamic hypothesis or theory about the
causes of the problem and designing a conceptual model that represents that problem and
its linkages in a system;
3. Quantitative Model Construction: Translating the causal relationships in the
conceptual model into tested, research-based mathematical equations, calibrating model
variables to historical data and projections, and extensive model testing;
4. Scenario Design and Evaluation: Testing policies and interventions that help people
visualize the systemic effects of actions.
This chapter describes how the D-O LRP SD modeling team proceeded through each of the four
phases listed above. At each phase of the SD modeling process, we consulted with local and
regional stakeholders, both individually and jointly at three large stakeholder meetings, who
were either directly involved with the D-O LRP or had concerns about the impacts that it would
have on the region.4 These stakeholders had a critical and evolving role in developing the D-O
LRP SD Model, and their contributions to each step in model development are also discussed in
this chapter.
2.2 Problem Definition
As described in Chapter 1, the D-O LRP SD modeling team focused on the development of the
D-O LRP in Durham and Orange counties, both as a proposed solution to existing problems and
as a perturbation to the system that could itself give rise to harmful impacts that might need to be
addressed with further policies. To identify the high-priority decision sectors within the three
pillars of sustainability (social, economic, and environmental) that would be impacted by the
4 More information on stakeholder meetings, including a list of attendees, meeting agendas, and workshop handouts,
can be found in Appendix E.
12
-------
light rail and subsequent development, we convened an initial meeting of stakeholders involved
with the D-O LRP on February 24, 2014. Stakeholders in attendance included:
• City and regional planners from Durham City/County Planning and Triangle J Council of
Governments (TJCOG);
• Representatives from GoTriangle and URS Corp. (now AECOM), who were directly
involved with the DEIS; and
• Representatives from the Durham-Chapel Hill-Carrboro Metropolitan Planning
Organization (DCHC MPO).
Stakeholders at the first meeting gave several presentations on the sustainability challenges the
community expects to face over the next 25 years if existing patterns of urban sprawl continue,
including increased pressure on local watersheds that would interfere with meeting nutrient
reduction goals, increased automobile dependency and energy use making residents less healthy
and more vulnerable to energy price spikes, and government budgets being stretched thin with
increasing demand to build roads and other infrastructure to new developments. In addition,
stakeholders presented concerns associated with the proposed D-O LRP, including a lack of
public funds available to build the infrastructure necessary to make the light rail safely and easily
accessible and the risk of gentrification and lack of affordable housing in the light rail station
areas due to increased property values.
2.3 Conceptual Model Development
With our stakeholders' concerns in mind, our modeling team sought to construct a conceptual
model that would represent both the direct transportation problems that the D-O LRP is intended
to relieve (namely, reducing automobile dependency and enhancing mobility in the areas located
near the proposed light rail stations), and the long-term issue of how the light rail might help or
hinder the community's efforts to achieve its sustainability goals over the next 25 years.
Overview of Causal Loop Diagrams (CLDs)
Defining and describing the structure of a system with conceptual models helps organize
information and highlight connections and possible unintended consequences within the system.
A conceptual model is best formed with the input of relevant stakeholders and is usually
communicated with causal loop diagrams (CLDs). A CLD is a map of the system analyzed - a
way to explore and represent the interconnections between the key indicators in the analyzed
sector or system (Probst and Bassi 2014). Involving stakeholders in this process ensures that the
CLD includes variables and relationships that are high priorities for people living within the
system and creates opportunities for group learning in which different stakeholders may discover
aspects of the system that they had not previously understood. This then contributes to the next
step in the SD model development process, building a simulation model. When simulation
models are built, the causal relationships in CLDs are translated into tested, research-based
mathematical equations, which can provide an internally consistent tool for comparing the
effects of alternative policy options. With the focus of attention on leverage points within the
system that can be influenced by decision-makers, rather than on external causes, SD helps
13
-------
people see connections between their actions and systemic effects, and guides participants to
answer the question: where can we intervene in the system? (Stave 2002).
In a CLD, different variables in a system are connected through simple, "all-else-equal" causal
relationships that indicate only the direction of causality and whether the relationship is positive
(i.e., both variables move in the same direction) or negative (i.e., the variables move in opposite
direction). In Figure 2-1, the relationship between population and vehicles on the road is positive
- as population increases, vehicles on the road also increases, and vice-versa. On the other hand,
the relationship between functioning roads and road congestion is negative - as functioning
roads increase, road congestion decreases (all else equal), and if functioning roads were to
decrease, road congestion would increase.
population
vehicles on the
road
road congestion
functioning roads
roads under
construction
Figure 2-1. Example CLD showing reinforcing (R) and balancing (B) feedback loops
Most relationships in CLDs are assumed to occur within a single specified period of time, though
delayed responses can be represented in CLDs as well (e.g., the relationship between roads
under construction and functioning roads is marked with "//" to indicate that it may take some
time before roads under construction are ready for use). As the diagram develops - preferably
through an interactive process involving modelers and stakeholders - these simple, unidirectional
relationships give rise to feedback loops. These loops are either reinforcing (R) or balancing (B)
feedback loops. Reinforcing loops amplify what is happening in the system, while balancing
loops tend towards maintaining equilibrium, balancing the forces in a system. When systems
contain both balancing and reinforcing loops, dynamic changes may move the system to a new
equilibrium (Shepherd 2014).
Initial CLD for the D-O LRP SD Model
Immediately after our first stakeholder meeting, the D-O LRP SD modeling team began drafting
a CLD using a software program called Vensim® (http://vensim.com/). The CLD construction
process allowed us to explore and represent the interconnections between the key social,
environmental, and economic components that stakeholders identified as being potentially
positively or negatively affected by the D-O LRP. To ensure that the CLD included variables and
relationships that were high priorities for stakeholders and to make sure they were being
accurately represented, we met with several D-O LRP stakeholders and local experts from the
University of North Carolina at Chapel Hill, Duke University, and North Carolina Central
14
-------
University, on an individual basis throughout the CLD construction process. After several rounds
of editing and refining, we invited our stakeholders back for a second meeting, held on June 4,
2014, and presented our initial CLD, shown in Figure 2-2.5 In addition, we presented more
detailed, "zoomed-in" CLDs of the seven key decision sectors (not shown) that were planned to
guide model development: land use, transportation, energy, economy, equity, water, and health.
For each of the key decisions sectors, we asked for feedback from stakeholders on potential
policies and interventions that the SD model could be built to test. At this second stakeholder
meeting, we heard from stakeholders that, although they did not expect the D-O LRP on its own
to mitigate the environmental, social, and economic issues facing the community, believed that it
could be a catalyst for change and for exploring how concurrent policies and interventions could
help maximize the benefits of the D-O LRP and minimize its potential consequences.
distance> \ /+J *"<+ /
\ / / X / co hsions> -.housinj
^
- mixed use
duration ,~
V +/ public transpo
transportat,ontnpS>
new road
demand road capacity ^^^ traffic
^ new road *T collisions
expenditure
le energy
public +/ """ consumption
+ transportation^^ /^
expenditure
.vehicle fueL
efficiency
Figure 2-2. Initial CLD for the D-O LRP SD Model, Presented at 2nd Stakeholder Meeting
5 This initial CLD does not represent the final version for the D-O LRP SD Model, which is shown at the end of this
chapter in Figure 2-7.
15
-------
2.4 Quantitative Model Construction
With the lessons from the first two stakeholder meetings in mind, the D-O LRP SD modeling
team began the task of translating the causal relationships in the CLDs we had built into tested,
research-based mathematical equations. After hearing from our stakeholders the desire to see the
simulated effects that the D-O LRP would have on both the immediate light rail station areas and
the region as a whole, we decided that our model should have outputs for both of these
geographic areas. Thus, the D-O LRP SD modeling team delineated two model geographies,
called "Tiers," shown in Figure 2-3.
Tier 2: DCHC MPO
(according to TRM TAZs)
Tier 1: !/> mile radius
around proposed light
rail stations
Figure 2-3. Map of the D-O LRP SD Model Geographic Tiers
Tier 1 consists of the combined half-mile-radius zones around each of the proposed light rail
stations (shown in yellow circles), since the areas in the immediate vicinity of the proposed light
rail stations are where the light rail will likely have its most pronounced direct impacts. Half-
mile radii were chosen to define Tier 1 because it is common practice among urban planners to
regard a half mile as the greatest distance that most people are willing to walk to a public transit
station. The boundary of Tier 2, which encompasses Tier 1, is defined to be equal to the
boundary of the Durham-Chapel Hill-Carrboro Metropolitan Planning Organization (DCHC
MPO). We expect that the indirect effects of the D-O LRP and subsequent development on
population growth, land use and zoning changes, and resource consumption, among others, are
likely to be felt throughout this region, which surrounds the light rail corridor. We defined Tier 2
16
-------
to include the suburban and rural areas of Orange and Durham Counties, as well as the northern
portion of Chatham County, which includes Jordan Lake, a significant regional water supply.
The D-O LRP SD modeling team then began an extensive collection of local data and
projections, aggregated for each Tier, to serve both as exogenous inputs to the model (e.g., birth
rate and death rate) and as targets for calibration purposes, used to ensure that endogenous
formulations of model variables (e.g., population) were able to reproduce historical or projected
trends.6 Where local data and projections were not available, we used state or national data
instead. In addition, we used equations and elasticities from the academic literature to inform the
mathematical connections between variables represented in the CLDs7. These data sources
played key roles in the process of building and testing the D-O LRP SD Model.
Like all steps in the SD modeling process, construction of the quantitative model was an iterative
process, depicted in Figure 2-4 below, that began with the building of the four main "core"
sectors: land use, transportation, energy, and economy. These sectors formed the engine of the
model and were built individually at first for Tier 2 and then fully integrated and calibrated so
that values for the main indicator variables closely matched historical data and projections, when
available, for our "business as usual" (BAU) scenario, which is meant to represent a continuation
of current trends without any policy interventions.
D-O LRP SD Model Construction Process
Calibration to Historical Data
Conceptual Model
Model Testing
\^K Data, Projections, & Equations
2000 2001 2002
VI 45 8 19
V2 62 0.3 3.4
V3 15 12 9
Build Core Sectors, Link Core Sectors
Model Outcome-Oriented Sectors
Figure 2-4. Depiction of the Iterative Model Construction Process
After we connected and calibrated these core sectors for Tier 2, we then duplicated all variables
for Tier 1. When necessary, and where data sources permitted, we adjusted the model structure
and parameter values for Tier 1 so that mechanisms more closely represented reality and outputs
6 The primary data sources for the D-O LRP SD Model are described in detail in Section 3 A of this documentation.
7 The data sources used to inform equations and elasticities in the model are described in detail in Appendix B.
17
-------
could more closely match data and projections for the proposed light rail station areas. We also
created connections between the two tiers so that any changes that occurred in Tier 1 were
accounted for in Tier 2.
Once the connections between the tiers were in place and indicator variables from both tiers were
calibrated to historical data and projections, we added the three outcome-oriented sectors (water,
health, and equity), including feedbacks from these sectors to the core sectors. At each stage of
the quantitative model construction process and with each new addition, key model variables
were monitored and recalibrations were made when necessary to ensure that the BAU scenario
represented, as best as possible, the social, environmental, and economic outcomes of extended
historical trends and community plans (excluding those that include the light rail project). Model
structure and sensitivity tests, described in detail in Chapter 6 of this report, were performed
throughout the quantitative model construction to ensure the model functioned properly under
varying conditions.
2.5 Scenario Design and Evaluation
For the third stakeholder meeting, held on May 13, 2015, we invited additional representatives
from community organizations that, would likely be interested in how the secondary and
cumulative impacts of the proposed light rail would impact their own areas of concern, although
they were not themselves directly involved with transportation and land use planning. Additional
stakeholders who attended included representatives from the Durham County Department of
Public Health, Durham Neighborhood Improvement Services, the Durham Coalition for
Affordable Housing, and Durham Economic and Workforce Development. Their input and
perspectives helped shape the final D-O LRP SD Model modifications and scenarios. Their
influence on the model is shown in Figure 2-5, where the names of stakeholder agencies are
overlaid on the Final CLD for the D-O LRP SD Model, which is displayed in full in Figure 2-7.
- - - . r-
" ' - I wily Sprain
t-
Durham Economic &
Workforce Development
't y
I
Opening I
\J °?
ABCrtatlc K«orl»:0i tn
HouxniStiKk 7>'
-»
Durham Neighborhood
Improvement Services
DCHC MPO
& TJCOG
-'
Sustai liability Off ice of
'' Durham City/County
Durham City/County Planning
?^.a,(H.o*Jsii»
Durham Water Management
.
Triangle Regional Model
'- '-
Service Bureau
Durham County Dept. of Public Health
Figure 2-5. Stakeholder Agencies Overlaid on Final CLD for the D-O LRP Model
18
-------
At the third stakeholder meeting, we administered a survey to assess stakeholders' priorities in
terms of scenarios for the D-O LRP SD model to run and output indicators that the model should
include. Feedback from the survey, summarized in Figure 2-6, helped shape the final three main
scenarios and additional scenarios that were chosen for the D-O LRP SD model, described in
detail in Chapter 4 of this report. For the three main model scenarios, stakeholders suggested that
we include the redevelopment of already developed land to higher densities in the proposed light
rail station areas instead of only increasing the density of new development. Participants listed
other high priority policies interventions to add, including parking costs and green infrastructure.
We were able to add parking costs as an adjustable policy lever. In addition, stakeholders listed
as their top priorities output indicators related to affordability and mortality health impacts, all of
which were added to the model. Unfortunately, quantifying the displacement of poor households
and creating a quality of place index, both of which stakeholders had interest in, proved to be too
challenging and time consuming to add in this version of the model.
(A) (B)
10 10
9
Redevelopment Green Land Reduced Telecommuting Affordability Health Outcomes in Health Outcomes Quality of Place
Infrastructure Conservation Parking Terms of Mortality in Dollar Terms Index
Strategies Requirements
• Most Important • Somewhat Important Less Important • Most Important • Somewhat Important Less Important
Figure 2-6. Results of Third Stakeholder Meeting Feedback Survey on the Top Priorities for the D-
O LRP SD Model: (A) Alternative Policies or Interventions to Test and (B) Output Indicators to Add
Final CLD for the D-O LRP SD Model, Version 1.0
The D-O LRP SD Model, version 1.0, was finalized shortly after the third stakeholder meeting in
July of 2015. A final CLD showing the main causal connections in the model is presented in
Figure 2-7, with model variables in black and potential policy interventions in red. As described
above, stakeholder involvement from many individuals and agencies in the region helped shape
the D-O LRP SD Model. Their willingness to not only attend our meetings, but be active
participants in our discussions helped demonstrate the potential for a SD model like the D-O
LRP SD Model to foster collaborative decision-making between local and regional organizations
in support of a more economically, socially, and environmentally sustainable future.
19
-------
< Vehicle FuelUse>
GRP Road Congestion Total Energy Spending
Property Tax
Revenues
S
Affordable Households in
Housing Stock ^ Poverty
Emp. Growth
i-/,
CO2 Emissions
___ Sales Tax Revenues ^^~~— Retail -^
*^~~~ Consumption T"^ Productivity Loss by
Government + ^__^Resident Per Capita-*-y/ „ - UJxr - /
Revenues /-" + Net Earnings -/Earnings />^ \ ~~T~~~~^ ?+
Resident i _ / ^y .^ + \ "'ff. - Energy Spending
Employment^ — /__ A^^ ^Parking Gross Operating Share of GRP ^ 1T,
Employment /Requirements Surplus _;^ x ^^-lotal Energy Use
Unemployment
Rate • ^JSfet Migration ^s$S \7^^. ' T / ^__^-Building Energy Use ^r
/V Parking Costs^-y/^,-.- J*^J? > \ Light Rail
*+ \. + /^^ ~~~^^~~^ / - Building Energy\ Electricity Use
GRP Per Capita \^ /^ .^ ^-/^ Efficiency
•Nonresidential ^ / ^, Nonresidential
Light Rail -i"/- ' D-eloped Land^
\ Residential
+ Developed Land
Dwelling Units \ %
\ ^^~ \
TTTT • n ^ i \ / \ \ X^-v X—Dwelling Units
HHs in Poverty at _ + ^1 * ^-" \ \ \ ^\ PerAcre
Dispkcement
Pop. Growth
+
i
Property Values
+ \ ^+
+ ^ Total Person Miles
/ Medkn Renter of Travel
Costs V \
iiKai
city I
+1
Floor Area \
Ratio
Energy Use for Water
:Land Shortage
+ Jobs-Housing
Nonmotonzed
Travel Facilities
Treatment + Dist.
Stormwater
+ « Pollutant Loadings
Impervious
Surfaces
Stormwater BMPs
Gasoline Price
Figure 2-7. Final CLD for the D-O LRP SD Model
20
-------
3 Model Structure
This chapter discusses the structure of the D-O LRP SD model. The information presented in this
chapter is meant to provide a general understanding of the model's structure and function; more detailed
documentation of model inputs, data sources, data processing, and calibration can be found in
Appendices B and C, and Chapter 6 - Quality Assurance. In the first section, we review primary local
data sources that were used in the model. In the second section, we provide an overview of the model's
structure, including the organization of the model into seven sectors and a brief description of some of
the model's major inter-sector third section, we provide additional information on each of the seven
sectors, including the major relationships in each sector, the data sources and processing steps used to
develop each sector, and any calibration done to ensure that key variables in each sector produced values
consistent with historical data and future projections from highly-regarded forecast models.
3.1 Overview of Data Processing and Primary Data Sources
This section presents an overview of the data processing method, called "clipping," used in ArcGIS by
the D-O LRP SD modeling team to scale the data we collected at various geographic levels to our model
Tiers (Tier 2 = DCHC MPO; Tier 1 = combined !/2 mile radii of proposed light rail stations), and gives
an overview of the primary data sources used in the D-O LRP SD Model. More detailed descriptions of
the data processing and analysis done for particular data sources can be found in Section 3.3, Chapter 6,
and in Appendix B.
Data Clipping Method
An abundance of local data sources were available for the study region thanks to the variety of
university modeling groups, local planning agencies, and other city organizations in the area devoted to
the collection and organization of local data. While a few local data sources provided outputs in GIS
shapefiles that perfectly conformed to the boundaries of our Tiers, the vast majority of the data sources
had to be "clipped" in ArcGIS for either one or both Tiers.
Census block
groups
Figure 3-1. Example of Methodology for Clipping in ArcGIS (Blue Circles Indicate Tierl)
When clipping was necessary, the values for each geographic zone in the source data (e.g., traffic
analysis zone, census tract, or census block group) were multiplied by an area multiplier to scale down
21
-------
the results to the portion of the geography contained within the Tier. For example, Figure 3-1 shows the
block groups in Durham County overlaid by our Tier 1 area in blue. We calculated the acres of each
block group that fell within the blue Tier 1 overlay and divided this by the total acres in the block group
to obtain the area multiplier for that block group. We note that this scaling process made the necessary
assumption that the scaled values were evenly distributed across the geographic zone. Non-countable
variables, such as median renter costs, could not be multiplied by an area multiplier directly. Instead,
relevant weighting variables, such as renter-occupied units in each geographic zone, were multiplied by
the area multiplier, and the resulting values were used to weight the median value for each geographic
zone.
Primary Data Sources
2040 Metropolitan Transportation Plan and Associated Modeling Efforts
Every five years, the Durham-Chapel Hill-Carrboro Metropolitan Planning Organization (DCHC MPO;
www.dchcmpo.org), the organization responsible for transportation planning for the western part of the
Research Triangle in North Carolina, is tasked with developing a Metropolitan Transportation Plan
(MTP), which is a "guiding document for future investments in roads, transit services, bicycle and
pedestrian facilities and related transportation activities and services to match the growth expected in the
Research Triangle Region" (CAMPO and DCHC MPO 2013). The 2040 MTP is a synthesis of many
individual studies, presenting results from the local transportation and land use modeling efforts, and a
detailed financial analysis of the costs associated with those transportation projects chosen for the plan.
The results of these modeling efforts make up a large part of the data used in the D-O LRP SD Model
and are described in further detail here.
The Triangle Regional Model Version 5
The Triangle Regional Model (TRM) is a travel demand forecasting tool that was developed for
understanding how future growth in the Triangle region of North Carolina will impact transportation
facilities and services.8 The TRM can help identify the location and scale of future transportation
problems, and the proposed solutions to those problems can be tested using the TRM, making it a
valuable tool for local planners. The TRM is an aggregate trip-based model that uses four steps: trip
generation (number of trips made and for what purpose), trip distribution (where the trips go), mode
choice (what transportation mode is used to make the trip), and trip assignment (what route and facilities
are used to make the trip) (TRM Service Bureau and TRM Team 2012). In the primary model step, trip
generation, trips are calculated based on population and employment data for each of the model's traffic
analysis zones, or TAZs.9
For the 2040 MTP, version 5 of the TRM (TRM v5) was used to test the transportation impacts of
several different hypothetical 2040 transportation network and land use scenario alternatives.
Performance evaluation measures were computed for each alternative based on the TRM v5 output,
including vehicle miles of travel, vehicle hours traveled, degree of traffic congestion, number of trips
8 The TRM is developed and maintained by the TRM Service Bureau, housed at the Institute for Transportation Research and
Education (ITRE) at North Carolina State University on behalf of the Durham-Chapel Hill-Carrboro MPO, Capital Area
MPO, North Carolina DOT, and GoTriangle, formerly Triangle Transit.
9 TAZs are defined as the geographic areas that a given travel demand model represents as a single node in the transportation
system, which trips can either begin or end at; there are 774 TAZs in Tier 2 and 178 TAZs are either partly or entirely
contained in Tier 1.
22
-------
taken by travel mode, and public transit ridership, for use by its client agencies (TRM Service Bureau
2008). Based on these performance evaluation measures, a "preferred option" transportation network
was chosen for the 2040 MTP that incorporated a road network, a bus transit network, and light rail and
commuter rail transit investments (CAMPO and DCHC MPO 2013).
For the D-O LRP SD Model, we used transportation data from both the raw output of the TRM v5, or
"TRM v5 result shapefiles," which provided information on traffic volume, distribution, and speed for
2010, 2017, and 2040 at the level of individual road links, and inputs to the TRM that were provided in
the socioeconomic (SE) data shapefiles, including parking prices and the availability of nonmotorized
travel facilities (DCHC MPO 2013). The 2040 MTP document itself provided TRM v5 performance
evaluation measures for the DCHC MPO transportation system under the preferred growth land use
scenario. Measures were reported for the base year (2010) and for two projections: (1) the transportation
system in 2040 based on the MTP, and (2) the 2040 "Existing + Committed" scenario, essentially a "no-
build" scenario, which included only those transportation projects that will be operational by 2017.
These performance evaluation measures provided information for Tier 2 on trips by mode and the
average distance of a person trip (CAMPO and DCHC MPO 2013).
Imagine 2040
As part of the process for developing the 2040 MTP and for use in the TRM v5, a base year (2010)
inventory of population and employment and forecasts of regional population and job growth by 2040
were compiled based on community plans and data from local planning departments, the Office of State
Budget and Management, the US Census Bureau, and independent forecasts from Woods & Poole
Economics, Inc. (CAMPO and DCHC MPO 2013). With these forecasts, the DCHC MPO, with the help
of Triangle J Council of Governments (TJCOG), launched an initiative called Imagine 2040 that used a
land use allocation software program called Community Viz to distribute the forecasted population into
single-family and multifamily housing units and to distribute forecasted employment by category into
each of the TAZs under several different land-use scenarios: trend development, community plans, and
all-in-transit (TJCOG 2013).10 A hybrid of the tested land use scenarios called the "preferred growth"
scenario was chosen for the 2040 MTP, and the results of the Community Viz modeling were then used
in the TRM v5 to forecast 2040 travel behavior.
The "preferred growth" land use scenario concentrated development in the region's activity centers,
including along the proposed light rail line between Durham and Chapel Hill. For the D-O LRP SD
Model, we made use of GIS shapefiles with socioeconomic (SE) data by TAZ for several interim years
(2010, 2017, 2020, 2030, 2035 and 2040), referred to hereafter as the "TRM v5 SE data."
CommunityViz 2.0
As noted above, the land use modeling for the 2040 MTP used CommunityViz to forecast future growth
of population and employment to the year 2040. Unfortunately, this modeling work did not produce an
inventory scenario for the 2010 base year due to funding constraints. In the second round of
CommunityViz modeling for the 2045 MTP (CommunityViz 2.0 or "CV2"), however, the modelers
included a 2013 inventory scenario of all parcels in the study area. This inventory scenario categorized
10 CommunityViz is an extension of ESRI's ArcGIS desktop software that facilitates the visualization and comparison of
alternative development scenarios. More information on CommunityViz and its capabilities for regional planning is available
on their website: placeways.com/communityviz
23
-------
parcels by one of 30 "place types" that best describe the current development on the site today and by
one of eight possible "development status" options. Together, the place type and development status
designations form a kind of current land use map (TJCOG 2014e). From this parcel data, we were able
to estimate baseline stocks of acres of land by development status (e.g., vacant, agricultural, protected
open space, and developed) and by use (e.g., single-family, multifamily, retail, office, service, and
industrial) for 2013. Note that these files were originally posted online with the parcel data pre-
populated with place type and development status designations that were used for the 2040 MTP for the
purpose of allowing local planning offices to review and edit the place type and development status for
2013 (TJCOG 2014e). At the time that we downloaded the shapefiles, 29% of parcels in our study area
had still not been reviewed, and therefore may have reflected the forecasted 2040 use and status rather
than that for 2013.
Community Property Tax Databases
The Durham, Orange, and Chatham County Offices of Tax Administration are responsible for listing,
appraising, and assessing all real property in the counties. They therefore maintain detailed property
maps and records at the parcel level. Because the databases included fine-grained parcel-lev el data, the
process of generating values for the model's Tiers was straightforward, with no need for area
multipliers. We obtained data on developed commercial square feet and property values, categorized by
use, from these parcel-level databases. Databases from the most recent year, 2014, were available for all
three counties, in varying levels of detail. For Durham County, sufficient data were available to establish
a time series for purposes of calibrating the model, with data from the years 2000-2008, 2010, and 2014.
U.S. Census Bureau
Most historical demographic data used to calibrate the D-O LRP SD model came from the U.S. Census
Bureau, including the Decennial Census (2000 and 2010) and the American Community Survey (ACS).
When possible, we downloaded the Census data through the ESRI Community Analyst online software
program (https://communityanalyst.arcgis.com/esriCA/login/), which allowed us to upload the
boundaries of our Tiers as GIS shapefiles and clip the Census data for the exact area we needed.
Community Analyst produces more accurate Census information for our Tiers than our own clipping
methodology described above since the program's clipping method (ESRI 2015) weights the data within
clipped sections of block groups by Census block-level population or employment, as opposed to our
own method which weights data by population, employment, or dwelling units at the Census block
group or Census tract level.11 However, because Community Analyst only reported Census data for most
of their variables from the 2010 Decennial Census and later, we used several other internet applications
to download historical Census data, including SimplyMap (Geographic Research Inc. 2015a), and the
National Historical Geographic Information System (NHGIS) Data Center (Minnesota Population
Center 2015, Abbas and Bell 1994).
For historical Census data that were not available from one of these three sources, we downloaded
Census data directly from the American Fact Finder website (http://factfinder.census.gov/), joined the
data to TIGER/Line shapefiles for the appropriate geographic zone (block group or census tract), and
Census blocks are the smallest geographic level of Census data, then block groups, then tracts.
24
-------
clipped the geographic zones in ArcGIS to our two model Tiers.12 As noted above, model inputs
prepared in this fashion made the necessary assumption that the scaled values were evenly distributed
across the block group or census tract.
3.2 Overall Model Structure
This section provides an overview of the D-O LRP SD Model structure, including specifics about the
model and an overall description of the main inter-sector feedback loops that make up the "engine" of
the model.
Model Specifications
Table 3-1 presents a summary of the model's specifications. As noted in Chapter 2, the model's
geographic scale includes two "Tiers":
• Tier 1 is defined as the combined half-mile radius zones around each of the proposed light rail
stations. Tier 1 is located entirely within Tier 2.
• Tier 2 is defined to equal the area of the Durham-Chapel Hill-Carrboro Metropolitan Planning
Organization (DCHC MPO), which includes parts of Orange, Durham, and Chatham Counties.
The model begins calculations in the year 2000 and continues simulating through 2040. Though the
model calculates the values of variables every l/16th of a year (or 0.0625 years), it reports outputs only
once per year.
Table 3-1. Summary of the D-O LRP SD Model's Specifications
MODEL SPECIFICATION
Model Scale and Boundaries
Model Time Frame
Model Time Step
D-O LRP SD MODEL VALUE
Two model "Tiers," for which all variables are aggregated:
• Tier 2 - the DCHC MPO boundary (includes Tier 1)
• Tier 1 - the combined 1/4 mile radii of proposed light
rail stations
2000-2040, with outputs every year
.0625 years
Structural Overview
As mentioned in Chapter 2, the D-O LRP SD Model has four "core" sectors - land use, transportation,
energy, and economy - and three output-oriented sectors -equity, water, and health. Figure 3-2 presents
a simplified CLD of the model with the core sectors in blue and the output-oriented sectors in yellow.
The CLD also shows key indicator variables with arrows representing the main intra- and inter-sectoral
connections. Note that for the sake of simplicity, the CLD shows only a small subset of the most
important variables and relationships in the model, and that the variable names shown in this diagram
may not be representative of the actual model variable names.
12 TIGER/Line shapefiles do not include demographic data, but they do contain geographic entity codes (GEOIDs) that can
be linked to the U.S. Census Bureau's demographic data and are available at: www.census.gov/geo/maps-data/data/tiger-
line.html
25
-------
Land Use
^x
Population] »•[ Dwelling Units
Net Migration!
Developed Land I ' ^
7-
Health
Stormwater Pollutant
Loading
\\\ ^
| Employment]
Nonmotorized
Person Miles
+ ^k.
Public Transit
Person Miles
I
z
Economy^
Property
Values
3
L
Households
in Poverty
Transportation
Equity
Figure 3-2. Simplified CLD of the D-O LRP SD Model with Core Sectors (Blue) and Output-Oriented
Sectors (Yellow)
Inter-Sector Feedback Loops
The core sectors are linked by several strong inter-sector feedback loops, meaning that a change to the
value of a model parameter in one of these sectors will have cascading impacts throughout the rest of the
model. As a result, these are the places where policies and interventions can have the largest impact in
multiple sectors. The outcome-oriented sectors have fewer strong feedbacks to the core sectors, meaning
that the variables in these sectors tend to respond to changes in the core sectors, rather than themselves
driving change in other sectors. Nevertheless, even the output-oriented sectors have some inter-sectoral
feedback loops. The remainder of this section is devoted to explaining the main inter-sectoral feedback
loops; intra-sectoral feedback loops will be described in the following section (3.3).
Employment
GRP
Nonresidential
SqFt
Figure 3-3. Simplified CLD of R1: A Reinforcing Feedback Loop Between the Economy and Land Use
Sectors
26
-------
R1: Economy -> Land Use -> Economy
The feedback loop shown in Figure 3-3, labeled Rl, is a reinforcing loop that links employment growth
to a growth in nonresidential square feet, which increases gross regional product (GRP), which (after a
two-year delay) contributes to an increase in total employment. Absent other factors affecting these
variables, this feedback loop will lead to continued growth in employment, developed commercial area,
and GRP.
GRP-
Person Miles
Congestion
Figure 3-4. Simplified CLD of B1: a Balancing Feedback Loop Between the Economy and Transportation
Sectors
B1: Economy -> Transportation -> Economy
The feedback loop shown in Figure 3-4, labeled Bl, is a balancing loop that links economic growth and
mobility. With an increase in GRP, transportation by automobile modes (labeled here as "person miles")
increases, hence increasing vehicle miles traveled (VMT), which increases congestion, which decreases
economic productivity, thus decreasing GRP and total person miles. Absent other factors affecting these
variables, this feedback loop serves as a limit on economic growth.
GRP'
/^
Congestion
\'
VMT
V
\.
+ ' Resident
Employment
\
A \
^K ~Df) \
\ y Percent of Population
in Poverty
^
. Transit- Dependent
Population
Figure 3-5. Simplified CLD of B2: A Balancing Feedback Loop Between the Economy, Equity, and
Transportation Sectors
B2: Economy -> Equity -> Transportation -> Economy
The feedback loop shown in Figure 3-5, labeled B2, is a balancing loop that links economic growth and
employment to equity and underserved populations who are more likely to be transit dependent. The
model assumes that as GRP increases, resident employment also increases, which lowers the
27
-------
unemployment rate. A drop in the unemployment rate is assumed to reduce the percent of the population
living in poverty, which decreases the portion of the population that is transit dependent. Decreasing the
number of transit-dependent people (i.e., increasing the number of people who drive cars) increases
VMT, which consequently decreases travel by other modes, thus increasing congestion and decreasing
GRP. This feedback loop also serves as a limit on economic growth.
Energy Spending
Factor Effect on GRP
Energy Spending
Share of GRP
Initial Energy Spending
Share of GRP
Rl
Employment
Energy Spending
onresidential Energy Price
SqFt
Figure 3-6. Simplified CLD of R1, R2, and B3: Balancing and Reinforcing Feedback Loops Between the
Economy, Land Use, and Energy Sectors
R1, R2, and B3: Economy -> Land Use -> Energy -> Economy
The balancing and reinforcing loops shown in Figure 3-6 depict the relationships that exist in the D-O
LRP SD Model between the land use, energy, and economy sectors. In the balancing portion (B3), an
increase in nonresidential square feet increases spending on energy, which - all else equal - increases
the share of GRP that is devoted to energy spending. If GRP is not increasing due to other factors, then
an increase in the share of GRP devoted to energy spending reduces the value added through the gross
operating surplus (a portion of GRP, not pictured), leading to a reduction in overall GRP. This negative
impact on GRP can be reinforced if energy spending does not decrease, since a reduction in GRP further
increases the share of GRP that is devoted to energy spending, indicated by the top reinforcing feedback
loop (R2) in Figure 3-6. Since energy spending and GRP may both increase (indicated in Rl through the
direct effect of nonresidential sq ft on both energy spending and GRP), GRP is negatively impacted only
if energy spending grows faster than GRP. Note that Rl, the reinforcing feedback loop from Figure 3-3,
is also included in Figure 3-6 in order to illustrate how the relationship with the energy sector provides a
limit to the reinforcing relationship between the land use and economy sectors.
3.3 Model Sector Descriptions
This section is meant to be an overview of the intra-sector relationships, data sources and processing
steps, and calibration steps taken for each of the seven sectors in the D-O LRP SD model. The secondary
bullets below give instructions on where to find more detailed information in this report for each of the
topics. For each sector, we describe the following:
• The Sector Relationships section gives an overview of the sector's main relationships - both
relationships among variables within the sector and relationships to other sectors in the model.
To illustrate the primary feedback loops in each sector, we use sector-specific CLDs. Unless
otherwise noted, the relationships described within each sector apply to both Tier 2 and Tier 1.
28
-------
o Alternative intra-sector relationships and structures that were considered are presented in
Section 6.2 of this report.
• The Data Sources and Processing section briefly describes the primary data sources used in
each sector and summarizes the data processing steps that were taken to prepare the values used
in the model. Where applicable, we describe how we adjust values taken from external data
sources to apply to the geographic areas used in the model - both Tier 1 and Tier 2 - as well as
any assumptions that were used in that process.
o Alternative data sources and equations that were tested but not ultimately chosen for the
model are described in Section 6.2 of this report and Appendix D.
• The Calibration section describes how values for some variables in the model were adjusted in
order to produce results that aligned with historical data, projections produced by other models,
or both.
o For a more detailed discussion of the steps taken to calibrate the model to historical data
and projections, and a quantitative analysis of how model results fit to data, see Chapter 6
of this report. Appendix B of this report lists all of the model variables that were
calibrated and the parameters that were adjusted during calibration.
Land Use
Sector Relationships
The Land Use sector comprises three types of stocks: (1) land disaggregated by use and development
status, (2) dwelling units disaggregated into single-family and multifamily, and (3) developed
nonresidential square feet disaggregated by category (retail, office, service, and industrial). Broadly
speaking, changes in the stocks of dwelling units are driven by changes in population; changes in the
stocks of each category of developed nonresidential square feet are driven by changes in category-
specific employment; and changes in these two types of stocks together drive changes in the stocks of
land use types. These changes are summarized in Table 3-2.
Table 3-2. Summary of Primary Stocks in the Land Use Sector
STOCK TYPE
Land (in
acres)
Dwelling Units
Developed
Nonresidential
Square Feet
CATEGORIES OF
DISAGGREGATION
Vacant, Agricultural,
Protected Open Space,
Right of Way, Retail,
Office, Service,
Industrial, Single-
Family, Multifamily
Single-Family,
Multifamily
Retail, Office, Service,
Industrial
VARIABLES DRIVING
CHANGES IN THE STOCK
Dwelling Units and
Developed
Nonresidential Floor
Area (in square feet)
Population
Employment
(disaggregated by
category)
GENERAL EQUATIONS
(Nonresidential Square Feet / Floor
Area Ratio) + (Dwelling Units /
Residential Density) = Total Land
Development (capped by Total Amount
of Developable Land).
Household Population * Average Single-
Family and Multifamily Household Sizes
= Minimum Dwelling Units.
Employment * Employee Space Ratios
= Demand for Developed Nonresidential
Square Feet.
As the figure shows, additional variables mediate these changes. Employee space ratios determine how
many square feet of each category are required for each employee in that same category. Average
29
-------
single-family and multifamily household sizes determine how many dwelling units are required by the
"household population" (i.e., the population adjusted for the percent of the population who do not live in
households, such as students in college dormitories, or those living in institutions). How much land
development is required for dwelling units and developed nonresidential floor area is determined by two
types of input tables: (1) the floor area ratio input table (square feet of floor area per square foot of land
area) determines the total land required for developed nonresidential square feet, and (2) the residential
density tables (dwelling unit/acre) determine the total land required for dwelling units. Total land
development is capped by the amount of developable land, broken down into stocks of vacant land and
agricultural land. These are assumed to be converted to developed use in equal proportion to their supply
(i.e., the model holds the ratio of vacant to agricultural land constant). While there is ample available
land in Tier 2 for years to come, this cap on land does come into play in Tier 1.
Negative land conversion can also occur if, for instance, population or employment declines, density is
increased, or shifts in the share of employment by category leads to a drop in space required for one of
the four categories of developed commercial area (e.g., desired industrial land could decline due to a
reduction in industrial jobs). Except in extreme testing, this rarely leads to an overall decline in
developed land, rather causing declines just in particular categories of developed land, leaving more land
to be developed for other categories. This essentially functions as an endogenous redevelopment
between uses, responding to demand. See the Economy sector for an explanation of how the model
estimates changes in the share of employment by category over time.
In the study area, the total number of dwelling units exceeds the number of households, since there is
always some degree of vacancy in housing markets. To capture this in the model, the construction of
dwelling units is not only determined by the required units given the number of households, but also by
the development of second homes (in the case of single-family units), dwelling unit turnover, and
anticipatory development in advance of their use. This leads to an endogenous vacancy rate in the
model, which goes on to affect renter costs in the Equity sector.
When the exogenous redevelopment switch is turned on (as in the Light Rail + Redevelopment scenario
described in Chapter 4), a portion of the current stocks of each category of developed land are
redeveloped to a higher density (by default set to almost 3 times the current). This density was chosen
because it is the overall weighted average increase in density (by jobs and housing) across the station
areas that was achieved in the Preferred Growth Scenario of the Imagine 2040 Regional Model (Green
2015). This high-density redevelopment has two effects in the model: (1) it boosts the number of
dwelling units and/or developed commercial square feet, and (2) this increased supply in turn reduces
the demand for regular-density greenfield development, ensuring that land development does not exceed
demand.
The land use sector creates many outputs that are then used throughout the model:
1. In the transportation sector, total developed land is used to calculate intersection density and
population density, which in turn affect choice of mode of transportation.
2. In the economy sector, total nonresidential square footage impacts gross operating surplus
3. In the energy and water sectors, the quantity and balance of developed land by land use affects
energy use and pollutant runoff from impervious surfaces.
4. In the equity sector, nonresidential square footage is used to calculate an endogenous
nonresidential density, and retail square footage is used to calculated retail density, both of which
30
-------
affect property values.
The primary feedback loops involving the land use sector are shown in the CLD presented in Figure 3-7.
The reinforcing loop shown in green (Rl) illustrates that employment (by category) increases developed
nonresidential square feet, which leads to increases in GRP, which then eventually increases
employment, completing the loop. Another reinforcing loop is shown in purple (R2): In Tier 1, where
there are more jobs than households, increasing dwelling units increases the jobs-housing balance (a
higher value is more balanced), which increases walking, eventually feeding back into population and
increasing dwelling units, as a result of improved health. Therefore, the loop that runs through the jobs-
housing balance is reinforcing in Tier 1, as increases in dwelling units increases walking (though it
should be noted that this loop has only a very small impact on population). In Tier 2, jobs and dwelling
units are close to balanced at the start of the simulation, so any large shift in either direction would
reduce the jobs-housing balance, which would decrease walking, and eventually decreasing (in a very
small way) population and dwelling units, making the loop potentially balancing in Tier 2. The larger
purple loop running through intersection density (Bl) shows that as population and dwelling units
increase, developed land also increases, affecting pedestrian accessible (excluding highways and major
arterial roads) intersection density in two ways. First, intersection density is calculated as pedestrian-
accessible intersections divided by developed land, so increasing developed land (all else equal)
decreases intersection density. Second, intersections are assumed to grow along with land development,
after a delay, due to both the construction of new intersections, and the addition of pedestrian
infrastructure to existing intersections. This causes a minimum increase in intersections as developed
land increases. Nonetheless, the net effect of increased land development on intersection density is
generally negative in the model (and in cases where land development stops or even declines,
intersection density increases). A decrease in intersection density, which is an indicator of walkability,
goes on to decrease walking and physical health in the model, which finally feeds back into population
and dwelling units by increasing the death rate (albeit very little), making this a balancing loop.
Differences between the Tier 2 and Tier 1 Land Use Sector
The only structural difference between Tier 2 and Tier 1 relevant to the Land Use sector is in what
forces drive migration into the area. In Tier 2, the developed portion of residential land drives net
migration following an inverted U-shaped curve; a lot of vacant residential land leads to relatively little
immigration; as more residential land gets developed (acting as a proxy for popularity of the region),
immigration increases; once the percent of land that is developed crosses a specified threshold,
immigration slows, reacting to scarcity (Vina-Arias 2013). We decided this relationship would not apply
to a small area such as Tier 1. Instead, net migration in Tier 1 is linked to the employment gap (the gap
between desired employment and total labor force, which includes the resident labor force and
commuters) and the unemployment rate. When the employment gap is positive (desired employment
exceeds total labor force), these desired new workers are added to the total labor force after a one year
delay and it is assumed that 5% of the new workers filling these jobs are new immigrants to Tier 1,
while the remaining 95% commute. After the introduction of the light rail, it is assumed that the
percentage of desired new workers who will choose to move to Tier 1 rather than commute will increase
from 5% to 10%. Finally, the unemployment rate provides a balancing effect in Tier 1 - if it rises too
high, immigration will slow, which reduces the size of the labor force; conversely, if it is very low,
immigration will increase. In both tiers, the relationships governing net migration are examples of how
nonlinearities are introduced in a system dynamics model: via the U-shaped table function in Tier 2 and
via the limits and feedbacks from employment, labor force, and unemployment in Tier 1.
31
-------
intersection
density
light rail project
redevelopment
nonresidential
developed square feet
(by categories
physical health labor force
residential
^^developed landW dwelling units„
per acre
redevelopment?
Figure 3-7. CLD for the Land Use Sector
Data Sources and Processing
Historical Demographic Data
Most historical demographic data came from the U.S. Census Bureau, including the Decennial Census
(U.S. Census Bureau 2000), and the American Community Survey (U.S. Census Bureau American
Community Survey 2014). Methods used to download and scale data to the model's Tiers are described
in Chapter 2.
Projected Demographic Data
Most long-term projections for demographic data came from GIS shapefiles with socioeconomic data
from modeling done for the "Preferred Growth Scenario" in the 2040 MTP (GoTriangle 2015). Methods
used to scale these data to the model's Tiers are described in Section 3-1.
Land Use Data
We obtained parcel-level data on acres of land by use and development status from shapefiles posted on
the TJCOG website, labeled CV2 Parcel Geodatabase for Place Type & Development Status Editing
(TJCOG 2014b). Note that these files were posted for the purpose of allowing local planning offices to
review and update the land use and development status since the 2010 version. At the time of use, 29%
of parcels in our study area had still not been reviewed, and therefore may have reflected the forecasted
2040 use and status rather than that for 2013.
To arrive at estimates for developed land in 2013, we clipped the parcel data to the Tiers. Because the
32
-------
parcel-level data were already fine-grained, we did not need to apply area multipliers in the scaling
process. For each Tier, we then totaled acres of land for each development status, using land use tables
from TJCOG used for the 2010 Community Viz modeling effort. These tables contained floor-area
ratios (FAR), residential density, percent residential, percent single-family/multifamily, and percent
retail/office/service/industrial for each place type in each jurisdiction in the study area. To estimate acres
of land by category, we first calculated an average of each value (by place type) from the land use
tables, weighted by the proportion of each Tier occupied by each jurisdiction. Within the "developed"
development status, we then applied these values to the total acres of each place type to obtain
developed acres by use and place type, and then summed these to get final estimates of developed acres
of single-family, multifamily, retail, office, service, and industrial use. Finally, we applied an adjustment
to the estimates of developed land provided by this source, subtracting out the acreage of parks and open
space, which we define as non-developed land for purposes of the D-O LRP SD model.
Nonresidential Square Footage
Property tax databases from the three counties provided parcel-level data on developed nonresidential
square feet, organized by land use category (Durham County Tax Administration 2000-2014, Orange
County Tax Administration 2014, FTA 1969). Methods used to scale these data to the model's Tiers are
described in Section 3-1. Heated building square feet was summed by land use code (residential,
commercial, etc). For Durham, the detailed land use codes from the source data were translated into the
categories used in the model (single-family and multifamily for housing units, and service, office, retail,
and industrial for developed commercial area), using NAICS codes. As noted in Chapter 2, Orange and
Chatham County did not have data for years prior to 2014, so we calculated 2000 values for developed
nonresidential square feet by category by first calculating per-capita parameters of developed
nonresidential building area (based on 2014 data) and then multiplying those parameters by each
county's population within each Tier in 2000. Since Orange and Chatham County did not have detailed
land use codes, we developed an allocation weighting scheme by dividing total jobs in each category
and year (office, service, retail and industrial), obtained from Woods and Poole Economics, Inc by
employee-space ratios calculated from the Durham county database. For example, the allocation weight
for the service sector in the year 2000 was equal to:
= [service jobs in 2000]/[service employee space ratio in Durham]
The share of total developed nonresidential square footage allocated to each category was determined by
dividing each category's allocation weight by the sum of allocation weights across all four categories.
Calibration
While full coverage of the calibration done for this sector can be found in Appendix B: Detailed
Documentation by Sector, two primary decisions made during model construction are discussed here.
First, population, total employment, and employment by job category were all considered as potential
drivers for nonresidential square footage. When we estimated nonresidential square feet as a function of
population alone, the results fit historical data on square feet well, but we rejected this formulation
because it would not allow the model to forecast different trends for different categories of
nonresidential developed land. Comparing the trends in total employment and employment by category
to data on developed nonresidential square feet produced very similar R2 values, so we chose
employment by category so that the model could estimate separate trends by category (though the retail
category of developed land in Tier 2 is still driven by population based on the assumption, also made by
the Durham City-County Planning Department (2012), that demand for retail space is driven more by
33
-------
consumer demand than by employment).
Second, early in the model construction process, the land development sector aggregated demand for
categories of developed land into one function that drove overall land development, which was then
allocated to different uses based on fixed zoning percentages. However, this created mismatches
between demand and supply and led to dramatic drops in retail density as retail uses were undersupplied
relative to demand, which then adversely affected property values. We then restructured the Land Use
sector to treat demand for different types of land and land conversion for each use separately and to
allow negative land conversion (i.e., conversion of developed to undeveloped land) to take place. This
resulted in slightly more variation in the distribution of developed land by category over time, along
with more developed land overall, particularly in Tier 1. We then needed to recalibrate gross operating
surplus per square foot (which regulates the strength of the relationship between nonresidential sq ft and
gross operating surplus and eventually employment), reducing the value of that parameter so that the
model's estimate of land development dropped back to historical levels.
Transportation
Sector Relationships
The heart of the transportation model sector is the calculation of mode shares, or the percentage of
person miles that are traveled by each transportation mode: automobile driver, automobile passenger,
public transit, and nonmotorized. The model first establishes baseline projections of the number of
person miles traveled per day by each mode (driven by population and real GRP per capita). These
baseline values are adjusted by a series of elasticities (with respect to both exogenous and endogenous
variables) and then normalized so that the total person miles of travel per day match the baseline
projections. This normalization is done to keep overall person miles of travel per day within
expectations and guarantee that an increase or decrease in the use of any one mode will also affect usage
rates of the other modes. Variables that drive deviations from the baseline estimates of modal person
miles are listed in Table 3-10. The table also notes whether each variable is an exogenous policy
variable or calculated endogenously in the model. For endogenous variables, the table notes in which
model sector the calculation takes place.
Table 3-3. Variables that Affect Modal Person Miles
VARIABLE
AFFECTING MODAL
PERSON MILES
Average
automobile speeds
Fuel costs
Parking costs
SUMMARY OF PRIMARY EFFECT
Lower speeds lead to reduced
automobile driver and passenger
travel and greater travel by public
transit and nonmotorized modes.
Higher costs lead to reduced
automobile driver travel and greater
travel by all other modes.
Higher costs lead to reduced
automobile driver travel and greater
travel by all other modes.
ENDOGENOUS OR
EXOGENOUS
Endogenous
Partly endogenous
(total fuel costs are
affected by the
impact of traffic
congestion on fuel
efficiency)
Endogenous
MODEL SECTOR
WHERE ESTIMATED (IF
ENDOGENOUS)
Transportation
Transportation
Land Use, Economy,
and Transportation
34
-------
VARIABLE
AFFECTING MODAL
PERSON MILES
Population density
Jobs-housing
balance
Density of
intersections that
are not purely
automobile
oriented
Public transit fare
prices
Proportion of
people who are not
in zero-car
households
Public transit
revenue miles per
day
SUMMARY OF PRIMARY EFFECT
Higher densities lead to increased
nonmotorized and public transit travel
and reduced automobile travel.
Greater jobs-housing balance leads to
more nonmotorized travel
Higher densities lead to increased
nonmotorized and public transit travel
and reduced automobile travel.
Higher fares lead to reduced public
transit travel.
Higher proportions lead to increased
automobile driver travel.
More revenue miles lead to increased
public transit travel.
ENDOGENOUS OR
EXOGENOUS
Endogenous
Endogenous
Endogenous
Exogenous
Endogenous
Exogenous
MODEL SECTOR
WHERE ESTIMATED (IF
ENDOGENOUS)
Land Use
Land Use
Transportation and
Land Use
N/A
Equity
N/A
In scenarios where the light rail line is built, the size of its effect on public transit travel is determined by
an equation that is driven by traffic volumes, population, and employment rates. The model assumes that
the majority of new public transit users resulting from the light rail line will be people who would have
otherwise traveled by automobile. In other words, most of any increase in public transit travel caused by
the light rail line is offset in the model by decreases in travel by automobile drivers and passengers, but
not by decreases in nonmotorized travel.
As noted above, the adjusted modal person miles are normalized so that total person miles in the model
match baseline projections. However, there are additional adjustments that are applied after the
normalization process is completed. First, traffic congestion is taken to have an additional dampening
effect on person miles of automobile driver travel. This adjustment is applied after the normalization
because we assume that this decrease in automobile driver travel does not necessarily translate into an
increase in use of any of the other modes. Second, for each public transit trip that occurs in the study
area, a certain additional amount of nonmotorized travel is taken to occur as a result of people walking
to and from transit stops. Similarly, this increase in nonmotorized travel does not translate to an
equivalent reduction in travel by any other mode.
The key feedback loops in this sector are illustrated in the CLD presented in Figure 3-8. The most
significant feedback loop within the sector is the way in which increasing automobile person miles of
travel increases vehicle miles traveled (VMT), which increases traffic congestion, which both reduces
vehicle speeds and increases fuel consumption and spending. These two effects both reduce automobile
person miles of travel, forming a balancing loop (labeled Bl and B2 in Figure 3-8). Most other
significant feedback loops relating to the transportation sector are inter-sectoral. For example, traffic
congestion decreases GRP (by first decreasing Gross Operating Surplus, not shown in the diagram),
which has a negative effect on person miles of automobile driver travel, and hence VMT, creating
another (less significant) balancing loop with traffic congestion (labeled B3 in Figure 3-8).
35
-------
Some other factors that change the distribution of travel by mode include the construction of roads and
nonmotorized travel facilities (e.g., sidewalks, paths, and bike lanes), the light rail line, and changes to
the bus transit system. These variables are mostly exogenous policy inputs in the model (the exception is
that construction of nonmotorized travel facilities is taken to be reactive to land development), with
increases or improvements in facilities or services related to a particular mode of transportation taken to
drive an increase in the use of that mode, usually at the expense of the other modes. However, the effect
of facility and service increases and improvements on person miles is not modeled in the same way for
every travel mode. When the construction of new roadway lane miles increases automobile travel, it is
because capacity has been increased relative to demand, which reduces traffic congestion. However, like
in most U.S. communities, public transit vehicles in the DCHC MPO rarely reach capacity. Even though
the proposed light rail line would likely attract more riders than most bus routes and would consequently
be more likely than bus routes to encounter capacity issues in the future, this model does not
disaggregate public transit use by type of transit vehicle, given how common it is for riders to transfer
between buses and trains. Therefore, this model does not consider the possibility of public transit use
being limited by how many people can fit on a transit vehicle. Instead, expanding or improving the
public transit system is taken to increase transit use by increasing the number of destinations that can be
reached via public transit, expanding the times during which those destinations can be reached, and
reducing wait times at transit stops, all of which are factors that make public transit more desirable
regardless of whether or not vehicle capacities are adequate to meet demand. In contrast, roadway
systems already provide access to most potential destinations in urban areas and rarely ever have limits
on when they can be used, making congestion relief the predominant benefit to automobile users of
expanding or improving those systems.
Impacts of the transportation sector on other model sectors, some of which are shown in Figure 3-8,
include:
• In the health sector, increases in the per capita use of nonmotorized modes of transportation drive
increases in physical health outcomes;
• In the energy sector, reductions in VMT reduce vehicle energy use and air emissions (including
greenhouse gas emissions);
• In the water sector; building new transportation facilities contributes to increases in impervious
surface area;
• In the equity sector, vehicle ownership/maintenance costs (which are a function of vehicle
ownership, itself affected by resident net earnings per capita), fuel costs, parking costs, and
public transit fares all contribute to overall household spending on transportation.
36
-------
land used for light
rail project
automobile
automobile driver passenger person
persqnmiles^- ,
lane miles
road construction
expenditure
public transit
fare
^^overall person-j"
miles
-population
Figure 3-8. CLD for the Transportation Sector
Data Sources and Processing
VMT, Traffic Congestion, and Roadway Lane Miles
The TRM v5 provides data for 2010, 2017, and 2040 at the level of individual road links, which we
clipped to the study area of the D-O LRP SD Model (DCHC MPO 2013). We multiplied the TRM's
reported traffic volumes on each road link by those links' respective lengths to determine VMT in peak
periods, in off-peak periods, and during the entire day. The TRM also reports average vehicle speeds
during peak periods and off-peak periods (which we assume represent freeflow conditions). To arrive at
a measure of traffic congestion, we divided freeflow speed by peak-period speed, weighted according to
the peak-period VMT on each relevant road link. In addition, we used the TRM's data on the number of
lanes on each road link to estimate lane miles in each Tier.
Public Transit Use and Operations
The National Transit Database (www.ntdprogram.gov) provides data on entire transit systems on a
yearly basis for 1995-2013 (FTA 2015b). For the D-O LRP SD Model, we obtained transit-related data
on ridership, person miles of travel, revenue miles, VMT, consumption of diesel and gasoline, agency
operating expenses, and agency fare revenues from the NTD entries for Chapel Hill Transit, GoDurham
(formerly the Durham Area Transit Authority), and GoTriangle (formerly Triangle Transit). Because
only part of GoTriangle's service area is within the DCHC MPO, we adjusted the numbers reported for
that agency based on data received from Transit Service Planner Jennifer Green regarding how many of
their boardings and revenue miles they attribute to Durham and Orange Counties, as opposed to Wake
County (Green 2014).
37
-------
The Bus and Rail Investment Plan in Orange County (Triangle Transit et al. 2012) and the Durham
County Bus and Rail Investment Plan (DCHC MPO et al. 2011) provide information on the expected
costs of building and operating the future light rail line. These documents also estimate expected
increases in the number of vehicle revenue hours of non-light-rail public transit in the study area. We
converted these revenue-hour figures to revenue miles using the NTD's reported ratio of revenue miles
per revenue hour in the year 2013 for the transit agencies in question.
We used the Our Transit Future website (http://ourtransitfuture.com) to ascertain the expected length of
the future light rail line, as well as the timeframe for its construction and opening. The website also
included the light rail line's expected schedule of operations at the time of its opening, which we used to
estimate the light rail line's expected vehicle revenue miles per day and per year (Triangle Transit
2015b).
Projections of Person Miles by Mode
As noted above, a key element of the transportation sector of the D-O LRP SD Model is the projection
of person miles by mode (automobile drivers, automobile passengers, nonmotorized travel, and public
transit). With the exception of base-year (2010) values for Tier 2 person miles of public transit travel
(for which we instead used NTD data for 1995-2013), we used the following methodology to generate
projections of modal person miles to which to calibrate the model:
(A) Obtain estimates of the number of person trips by mode (drive-alone, carpool, nonmotorized, and
public transit) and of the overall average distance per person trip from the DCHC MPO 2040
MTP for 2010 and 2040 (CAMPO and DCHC MPO 2013).
(B)Estimate an average carpool size, using 2008-2012 ACS data from the U.S. Census Bureau
(which reports the number of workers by census block group who commute in a carpool with 2,
3, 4, 5-6, or 7+ total people), as reported by ESRI Community Analyst (2014).
(C)Use the average carpool size to convert the MTP's numbers of drive-alone and carpool person
trips to driver and passenger person trips.
(D)Use average trip distances by mode from the Household Travel Survey Final Report (originally
created to generate data for the Triangle Regional Model) (NuStats 2006) to calculate expected
ratios between trip distances by each of the four modes.
(E) Multiply total person trips by average person-trip distance to yield total person miles of travel.
(F) From the number of trips by each mode, the total person miles traveled by all four modes
combined, and the ratios between the distances of individual trips by any given mode, calculate
the average distance of one trip by each of the modes in question, as well as the number of total
person miles by each mode per day.
(G)Use ACS data reported by Community Analyst to scale the Tier 2 person-mile-by-mode results
to Tier 1. The methodology for this scaling process can be found in Appendix B.
Nonmotorized Travel Facilities and Parking Costs
We obtained estimates of the number of miles of nonmotorized travel facilities in the area and of the
parking cost of an average trip from the TRM v5 SE data (DCHC MPO 2013). The data source provided
separate average parking cost estimates for work and non-work trips. From these estimates, we
generated a single average parking cost estimate for each year in each TAZ, weighted by the numbers of
work and non-work trips ending in each TAZ (estimated using trip-generation equations found in the
TRM Version 5 documentation (TRM Service Bureau and TRM Team 2012).
38
-------
Intersections
The Smart Location Database (SLD) supplies 2010 data on location efficiency at the census-block-group
level, including information on intersection densities (US EPA 2013d). We multiplied these density
estimates by total land area in the census block groups within the Tiers to estimate the total number of
intersections, excluding those that are purely automobile-oriented.
Vehicle Stock
We used SimplyMap to obtain summarized data from the U.S. Census Bureau, clipped to specified
geographic areas in accordance with a standardized methodology (Geographic Research Inc. 2015a).
From this source, we obtained vehicle-stock figures for the years 2000, 2010, 2011, 2012, and 2014.
SimplyMap provides the number of households with 1, 2, 3, or 4+ vehicles; for purposes of this model,
we assume that all households in the "4+ vehicles" bin have exactly four vehicles.
Transportation Infrastructure Costs
The DCHC MPO 2040 MTP provides expected overall costs of road construction and road maintenance
for the period 2010-2040 (CAMPO and DCHC MPO 2013). We divided these estimates by the reported
number of lane miles to be built and the number of lane miles existing, respectively, in the TRM v5
"preferred" scenario to yield estimates of construction and maintenance costs per lane mile. Similarly,
we estimated costs per mile of nonmotorized travel facilities by dividing the MTP's expected 2010-2040
spending on such facilities by the TRM v5 SE data's projected increase in nonmotorized travel facilities
during that period (DCHC MPO 2013).
Calibration
The transportation sector was originally constructed in isolation from the other sectors of the model. At
that stage, we calibrated the outputs from this sector to be consistent with results of the Triangle
Regional Model (Version 5), in light of the fact that the TRM was the largest source of this sector's
important inputs (including inputs from the DCHC MPO 2040 Metropolitan Transportation Plan, which
itself relies mostly on the TRM). During this calibration stage, inputs that were to eventually come from
other sectors (e.g., GRP, population, developed land area, and jobs-housing balance) were drawn from
TRM data, and those elasticities that were not drawn from literature were based on TRM projections. In
addition, we set all lookup tables representing future policy decisions (such as transportation facility
construction) to be consistent with the TRM's 2017 and 2040 projections. Finally, the initial year of the
model was set to 2010, the same initial year used by the TRM. With these measures in place, major
outputs of the transportation sector (VMT, congestion, etc.) were successfully brought within a
reasonable range of values projected by the TRM.
When we connected the transportation sector to the economy, land use, equity, energy, water, and health
sectors, we reset the initial year to 2000 and adjusted initial values of dynamic variables to (1) produce
approximately the same 2010 values as were used during the previous calibration stage, and (2) maintain
the same trends out to the year 2040 (with key inputs coming from other sectors of the D-O LRP SD
Model, rather than from TRM projections). For the most part, we did not need to change equations in the
transportation sector as part of this step. In one exception, we changed the calculation of population
density and intersection density to use developed land as the denominator, rather than total land area.
Although we made further changes to the model after connecting its various sectors, those changes were
mainly about making the model more closely match real-world cause-and-effect relationships, rather
than about fitting the model to data. For example, we later adopted a new, more authoritative elasticity
39
-------
for the effect of public transit fare prices on public transit person miles and made the percentage of the
population that is not in zero-car households an input to both vehicle stock and miles of automobile
travel per day.
As additional changes were made, both to the transportation sector and to the rest of the model, we
checked for differences in the values of certain variables that we had designated as key variables to
ensure that the transportation sector remained well calibrated. For these variables, we ensured that all
modeled values were close to reference data and projections (usually within 5% in the most recent year
for which there is historical data). In most cases, we were able to keep modeled values for these
variables within the preferred range of their reference data and projections by changing initial-year
values. When it was not possible to achieve such a match to all of the reference data and projections for
a given variable, we prioritized matching the most recent historical data point. We performed calibration
checks in a specific order, beginning with variables that are least affected by the rest of the
transportation sector and ending with variables that have the least effect on the rest of the transportation
sector. The variables used to calibrate the transportation sector, in the order in which they were routinely
checked, are listed below:
1. Vehicle stock
2. Functioning nonmotorized travel facilities
3. Intersections excluding those that are purely automobile oriented
4. Parking cost of average trip
5. Person miles of automobile driver travel per day
6. Person miles of automobile passenger travel per day
7. Person miles of public transit travel per day
8. Person miles of exclusively nonmotorized trips per day
9. Through traffic VMT
10. VMT
11. Congestion
12. Vehicle trip distance
13. People in traffic accidents per year
40
-------
Energy
Sector Relationships
The energy sector has four main components: (1) estimating total energy used by buildings (divided into
commercial, industrial, and residential categories) and for water treatment and distribution; (2)
estimating total CO2, PM2.5, and NOX emissions caused by energy used by buildings, transportation
(divided into passenger vehicles, buses, and the light rail), and water treatment and distribution; (3)
estimating the total cost of energy use by buildings and transportation; and (4) projecting growth in
generation of alternative energy sources within the region.
For the first component, the energy used by buildings in both Tier 2 and Tier 1 is driven by economic
and land use factors, including the stocks of single-family and multifamily dwelling units, developed
commercial building square feet (defined as the sum of developed office, service, and retail square feet),
and developed industrial building square feet. Population growth also drives building energy use, both
by increasing the stocks of buildings and by increasing water consumption (because energy is required
for water treatment and distribution).
For the second component, the energy sector calculates CO2 emissions from building energy use and
from transportation, using miles traveled by LRT, buses, and automobiles, as calculated by the
transportation sector. For each category of energy use, the model calculates CO2 emissions using fuel-
specific emissions factors, assuming a fixed mix of fuel types for each category (e.g., the model assumes
that 35 percent of commercial energy use is from natural gas, and this value is held constant throughout
the modeled time period). The model also estimates emissions of PIVh.s and NOX, by multiplying
automobile VMT by pollutant-specific emissions factors. In turn, emissions of these two pollutants
affect outcomes in the health sector.
For the third component, regional energy spending is calculated for four categories (residential,
commercial, industrial, transportation) based on energy use, fuel type (divided into gasoline, electricity,
and natural gas), and fuel price. As noted above, for each category, the model holds the mix of fuel types
constant.
For the fourth component, the model simulates the use of alternative energy sources within the region,
namely solar and landfill gas. Currently, most energy used in the region - electricity, natural gas,
gasoline and diesel - is generated or produced outside the region. Solar electric capacity is small but
growing in the region (about 20 MW capacity as of 2014 (North Carolina Sustainable Energy
Association 2015)), and North Carolina has a thriving solar industry, ranked fourth in the US for
installed solar capacity (GoTriangle 2015). Based on data provided by the North Carolina Sustainable
Energy Association, our model estimates conservatively that solar capacity grows in Tier 2 from 2.5 kW
in 2000 to level off at 40 MW by 2020. Our model also explores future scenarios with higher solar
capacity. Durham County also recently began generating electricity from landfill gas (~3MW capacity),
and our model projects that this capacity is sustained until 2028, representing the 20-year purchased
power contract for one landfill. Increased production from these two alternative energy sources
decreases CO2 emissions from building electricity use, as the alternative sources are assumed to be
carbon-neutral.
Figure 3-9 presents a CLD for the energy sector. As the figure shows, total energy use in the model is
affected by (1) changes in the energy intensity of buildings and vehicles, (2) changes in building stock
and vehicle miles traveled (VMT), (3) changes in the use of public transportation, and (4) changes in the
proportion of building types (e.g., single-family vs multifamily residential) through development. These
41
-------
energy changes, along with any growth in renewable energy, also affect modeled air pollutant emissions.
The key feedback loops in this sector are labeled in Figure 3-9 Energy spending represents either a
balancing or reinforcing loop on electricity use (B1/R1). The loop is balancing if energy spending
increases relative to GRP, and reinforcing if the energy spending decreases relative to GRP. Public
transit ridership, nonmotorized travel trips, and congestion each form balancing feedback loops through
vehicle energy spending, opposing either an increase or decrease in VMT (B2, B3, B4). In the B2 loop,
an increase in VMT increases vehicle energy consumption and vehicle energy spending, which
stimulates public transit ridership, which reduces VMT. Increased vehicle energy spending due to
increased VMT also stimulates nonmotorized travel (in the B3 loop), reducing VMT. In the B4 loop, an
increase in VMT produces congestion, which reduces VMT.
nonresidential sq ft
SF and MF dwelling units
light rail
project
solar and LFG
capacity
redevelopment
residential and
nonresidential density
building energy
intensity
electricity
price
jblic transit ( B2 f^r
ridership \^_^J^
non-motorized ( B3
travel trips
energy for DW and
WW conveyance
price of
gasoline
-VMT
vehicle energy"
consumption
Figure 3-9. CLD for the Energy Sector
Data Sources and Processing
Energy Prices and Energy Intensities
We obtained both historical and projected future data on energy prices, building energy intensities, and
passenger vehicle fuel efficiency from the EIA. For projected values, rather than using the specific
values forecast by EIA, we converted the projected values from the Annual Energy Outlook 2015
(AEO2015) to annual percent changes through 2040 and applied them to the most recent-year historical
data. EIA provided historical energy prices specific to North Carolina (US EIA 2015b) and projected
future energy prices for the US South Atlantic region (US EIA 2015a). EIA also projected trends in
residential, commercial, and industrial energy intensities (e.g., MMBtu per household or per square
42
-------
foot), which we applied to local historical energy intensity data when possible, or to literature values (for
example, Residential Energy Consumption Survey 2009 (US EIA 2009) for national average residential
energy intensities). Finally, AEO2015 provided passenger vehicle fuel efficiency trends, in miles per
gallon, for the light duty vehicle stock, which assumes a range of vehicle age classes.
PM2.5 and NOX Emissions
We obtained county-level historical PM2.5 and NOX emissions from EPA's National Emissions Inventory
(US EPA 2015d). We scaled emissions data from 2002, 2005, 2008, and 2011 from Durham County to
Tier 2 based on the 2010 ratio of Durham County VMT to Tier 2 VMT (0.586), assuming that the
relationship between VMT and emissions would be consistent across Durham county and the other areas
of Tier 2. The ratio of Durham County VMT to Tier 2 VMT was calculated from Durham City-County
Sustainability Office data and Triangle Regional Model v5 data (Durham City-County Sustainability
Office 2015, DCHC MPO 2013). We recognize that a limitation of these historical data are that they
come from two different models (2002 and 2005 emissions data were generated by the MOBILE model
but 2008 and 2011 emissions data came from the MOVES model (Driver 2015). Passenger vehicle
PM2.5 and NOX emissions rates were estimated from the Argonne National Laboratory GREET model
(Cai et al. 2013) which is based on the EPA MOVES model. For purposes of projecting future emissions
from vehicles, we estimated overall PM2.5 and NOX emission rates for the passenger vehicle stock by
applying emission rates by model year in the GREET model to the age distribution of the US fleet
(Jackson 200Ic). To estimate emission rates for model years prior to 1990 and after 2020 (the endpoints
of the GREET data), we applied a linear extrapolation to the GREET data. We assumed the fleet age
distribution provided by EPA data, which includes age classes from 1-17 years, would remain constant
through 2040.
Energy consumption and CO2 emissions
Ms. Tobin Freid, the Sustainability Manager for Durham City-County, provided energy consumption
and CO2 emissions data for Durham County for the years 2006-2013 (Freid 2015). Ms. Freid estimated
CO2 emissions for four categories (residential, commercial, industrial, and transportation) based on
energy consumption data provided by local utility companies. For the D-O LRP SD model, we scaled
Durham County energy and CO2 emissions data to Tier 2 for each category as follows: residential data
were scaled using the ratio of Durham residential sq ft to Tier 2 residential sq ft (0.703); commercial
data were scaled using the ratio of Durham commercial sq ft to Tier 2 commercial sq ft (0.835);
industrial data were scaled using the ratio of Durham industrial sq ft to Tier 2 industrial sq ft (0.986);
and transportation data were scaled using the ratio of Durham VMT to Tier 2 VMT (0.586 in 2010).
Calibration
When calibrating the variables in the energy sector, we first calibrated total energy use in buildings by
adjusting building energy use intensity by category (residential, commercial, industrial) in order to
improve the fit between the model's estimated energy use by category and historical data provided by
the Durham City-County Sustainability Office for the years 2006-2013. We then adjusted the natural gas
fraction of energy use by category in order to improve the fit between the model's estimated CO2
emissions from natural gas combustion and historical data. Finally, we adjusted average automobile fuel
efficiency to calibrate passenger vehicle emissions. We also tested the internal consistency of Durham
energy and emissions data, since these were scaled up to generate Tier 2 data.
Calibration allowed modeled total CO2 emissions and total building energy consumption to match
historical data to within 5% on average. Total building energy consumption is the sum of residential,
43
-------
commercial, and industrial energy use; and modeled commercial energy use fit historical data better than
modeled residential energy use. Although residential energy use matched the average value of data (5%
average deviation), the residential energy data contained "noise" or unexplained variability. Passenger
vehicle CO2 emissions had similar uncertainty in model fit; the model matched the average value of data
(10% average deviation) but the vehicle emissions data contained high unexplained variability. These
observations about model fit suggest that the D-O LRP SD model projections of aggregate energy
variables such as total CO2 emissions and total building energy consumption may be more accurate than
building or vehicle sector-specific projections.
Economy
Sector Relationships
To estimate regional economic activity (i.e. gross regional product or GRP, disaggregated into total
earnings and gross operating surplus), the economy sector of the D-O LRP SD Model uses feedback
loops within the economy sector and responds to connections from the land use, transportation, and
energy sectors. The inter-sector relationships affect gross operating surplus, an indicator that represents
the difference between the total monetary value earned by businesses for the production of goods and
services in the region and the total earnings of their employees. These relationships are described in the
Section 3.2 of this paper, while this section discusses the intra-sector relationships within the economy
sector.
unemployment rate
resident labor force
resident
employment
population
employment by
category
cumulative half cent
transit sales tax
revenues
resident percent of the
working population
half cent transit sales-
tax revenues
local sales tax
revenues
transit sales tax share of
local sales tax revenues
local government
revenues
ital nonresidential
sqft
total earnings by
category
;otal earnings
gross operating _ effect ofroad
surplus "^ congestion on
V" productivity
effect of energy
spending on productivity
+J
-------
shows this loop highlighted in green (Rl). Total employment is divided into four categories (industrial,
office, retail, and service), each with different earnings per employee. Earnings per employee are
multiplied by the number of employees in each category and combined into total earnings, the largest
component of GRP. Changes in GRP drive total retail consumption (as does any change in the resident
percent of the working population in Tier 2), which is in an indicator of desired employment. The
relationship between GRP, total retail consumption, and desired employment is widely accepted in
economic theory (Keynes 1936). The standard theory holds that if retail consumption grows
(representing demand), the desired production of goods and services also increases along with
employment (two key drivers of supply). At the same time, employment only increases as long as there
are enough labor force participants to fill the desired employment positions. This constraint is
represented by the link between total labor force and total employment in Figure 3-10 (though in the
model, this relationship takes the form of a "minimum" function between desired employment and total
labor force). Retail consumption also drives the collection of sales tax revenue in the economy sector.
Besides local sales tax revenue, the other primary component of local government revenues is property
tax revenue, based on property values (described in detail in the Equity sector. This main reinforcing
loop (Rl) that drives economic growth also affects several variables outside of the economy sector. A
summary of these relationships is presented in Figure 3-10.
Table 3-4. Economy Sector Variables Affecting Variables in Other Sectors
DIRECTION OF
ECONOMY SECTOR VARIABLE INFLUENCE
Relative GRP per capita
Relative total employment
Relative resident per capita net
earnings
Relative resident per capita net
earnings
Relative retail per capita net earnings
Unemployment rate
(+)
(+)
(+)
(+)
(+)
(+)
SECTOR AND VARIABLE IT INFLUENCES
Transportation: Baseline relative person miles of
travel per day per capita (all modes)
Land Use: Nonresidential property value persq ft
Equity: Single-family and multifamily property
value per dwelling unit
Transportation: Income factor affecting desired
vehicle ownership per person not in a zero car
household
Equity: Affordability index
Equity: Percent of population in poverty
Differences Between the Tier 2 and Tier 1 Economic Model
When the Tier 2 economy sector was duplicated for Tier 1, it became apparent that the simple
formulation used for total employment was going to be insufficient. Unlike Tier 2, where population and
employment grew historically at about the same rate (and were projected to do the same going forward),
historical data for Tier 1 show that employment has grown more rapidly than population (U.S. Census
Bureau 2015). In addition, whereas 74% of jobs in Tier 2 were held by residents of Tier 2 in 2010 (NC
ESC 2014), only about 20% of the jobs in Tier 1 were held by residents of Tier 1 in that year
(Geographic Research Inc. 2015a). This result is not surprising, given that the two major employment
45
-------
centers in Durham and Orange County, Duke and UNC Hospitals, are located within Tier 1, and many
of the employees that work there commute from outside of Tier 1. Consequently, the "relative resident
percent of the working population" variable that was used to influence total retail consumption in Tier 2
did not increase retail consumption in Tier 1 enough to match historical trends because much of the retail
consumption that takes place in Tier 1 is done by people who reside elsewhere.13 In addition, a new
formulation that allowed a dynamic portion of the desired employment to be filled by commuters, shown
in Figure 3-11, allowed employment to grow more rapidly than population in Tier 1. Lastly, the resident
employment and unemployment rate calculations were made dynamic for Tier 1, with unemployment
rate having an influence on net migration for Tier 1.
resident labor
force Tier 1
^
unemployment rate
Tierl
population Tier 1
immigration due to
employment gap Tier 1
resident total net
earnings Tier 1
total employment
Tierl
, desired
employment gap
~. i + employment Tier 1
p total retail
desired additional consumption Tier 1
commuters Tier 1
share of desired
employment that is net
migration Tier 1
Economy Sector Tier 1
-employment by
category Tier 1'
total earnings by
category Tier 1
gross
regional
product
Tierl
total earnings
Tierl
gross operating
surplus Tier 1
Figure 3-11. CLD of the Tier 1 Economy Sector
Data Sources and Processing
Total Employment and Employment by Job Category
Historical employment data and projections were available from several sources, but we chose to
calibrate the model to the employment data (2010) and projections (2011-2040) from the TRM v5 SE
data (DCHC MPO 2013). For the preceding model years, 2000-2009, annual employment growth trends
from the U.S. Bureau of Economic Analysis (BEA) for Durham and Orange County for Tier 2 (BEA
2014d) and from the U.S. Census Bureau's Longitudinal Employer-Household Dynamics (LEHD)
Program Origin-Destination Employment .Statistics (LODES) dataset for Tier 1 (U.S. Census Bureau
2015) were applied to the 2010 value for total employment from the TRM. Although we were able to
use total employment data and projections from the TRM v5 SE data files to calibrate total employment
in the model for Tier 2 and Tier 1, which also provided delineations of employment into five categories
(industrial, office, service, retail, and highway retail), specific job types within a 2-digitNAICS category
13 A more detailed description of this formulation and why it could not be used for Tier 1 can be found in the " Structure
Confirmation Tests" Subsection of Chapter 6, Sections.
46
-------
were divided among job categories in the TRM v5 SE data, making it impossible to combine this jobs by
category data with Woods & Poole's earnings per employee by job category data and projections to get
total earnings. Thus, instead of using the employment by job category numbers from the TRM v5 SE
data, we aggregated the 20 job types provided by Woods & Poole14 (Woods & Poole Economics Inc.
Copyright 2014) into four categories, industrial, office, retail, and service, shown in Table 3-5, based off
of the 5-tier employment classification scheme used in the Smart Location Database (Ramsey and Bell
2014), and multiplied total employment with the share of employment that fell into each category for the
entire study period (2000-2040).
Table 3-5. Job Types by 2-digit NAICS Codes Aggregated into Four Employment Categories Used in the
D-O LRP SD Model
JOB TYPE BY CATEGORY
NAICS CODE
Retail Jobs
Retail Trade
Accommodation and Food Services
NAICS 44-45
NAICS 72
Office Jobs
Information
Finance and Insurance
Real Estate and Rental and Leasing
Management of Companies and Enterprises
NAICS 51
NAICS 52
NAICS 53
NAICS 55
Industrial Jobs
Agriculture, Forestry, Fishing, and Hunting
Mining
Utilities
Construction
Manufacturing
Wholesale Trade
Transportation and Warehousing
NAICS 1 1
NAICS 21
NAICS 22
NAICS 23
NAICS 31-33
NAICS 42
NAICS 48-49
Service Jobs
Professional, Scientific, and Technical
Administrative and Support and Waste Management
and Remediation Services
Educational Services
Health Care and Social Assistance
Other Services [Except Public Administration]
Arts, Entertainment, and Recreation
Public Administration
NAICS 54
NAICS 56
NAICS 61
NAICS 62
NAICS 81
NAICS71
NAICS 92
Total Earnings and Earnings per Employee by Job Category
Since total employment in the model for Tier 2 and Tier 1 was calibrated to historical data and
14 Woods & Poole does not guarantee the accuracy of this data. The use of this data and the conclusions drawn from it are
solely the responsibility of the US EPA.
47
-------
projections to match the total employment from the TRM v5 SE data, and the total employment numbers
in the TRM v5 SE data were lower than total employment numbers from historical data and projection
sources for Durham and Orange County combined (e.g. BEA and Woods & Poole Economics, Inc.),
alternative calculations of total earnings were done to calibrate the model. Woods & Poole provided
average earnings per employee per year for Durham and Orange County, both historical data and
projections, for each of the 2-digitNAICS code classifications. To calculate total earnings, we took the
average earnings per employee per year in each of the four employment categories and multiplied them
by the number of jobs in each employment category for each year and summed them.
Gross Regional Product (GRP)
Although Woods & Poole provides historical GRP and projections by county, we were unable to use
these values because our estimates of total earnings were lower than their estimates. To calculate
historical GRP for the three counties in Tier 2 based on our total earnings estimate, we used a variation
of an equation from the BEA Regional Economic Accounts Office, GDP -by-metropolitan-area
methodology that uses at top-down approach, distributing state-level output to metropolitan areas (Panek
et al. 2007):
| ,, „ .
I „ , x ^ arnmdsec, county, y
\c arnmgsecstateiy j
Where: ec = employment category (industrial, office, retail, service), and y = year.
To use this equation, we first obtained GDP and earnings by employment category for North Carolina
for 2000-2013 from the BEA(BEA 2014a, c). We then summed the GRP per employment category per
county per year to get the total GRP per year for Tier 2 for 2000-2013. We used the same equation to
calculate GRP for Tier 1, with Tier 1 earnings replacing county earnings. We projected GRP in Tier 2
from 2014 to 2040 based on our projections of total earnings (calculated from TRM v5 SE data and
Woods & Poole data, as described above), holding the earnings percent of GRP constant at its 2013
level of 60%. For Tier 1, we used the same process, though the earnings percent of GRP in 2013 was
about 70%, due to the high percentage of service employment in this area.
Total Retail Consumption
Three sources of historical retail consumption data were available to calibrate the Tier 2 economy
model, all with data for Orange and Durham County. The first source was "total taxable sales" data for
2000-2014 from NC DOR (NC DOR 2014). The second source was Woods & Poole Economics, Inc.
with "total retail sales" estimates between 2000 and 201 1, with 2002 and 2007 being actual historical
data from the U.S. Department of Commerce (Woods & Poole Economics Inc. Copyright 2014). The
third source was "total retail sales" estimates downloaded from SimplyMap for 201 1-2014 (data were
also downloaded from 2008-2010, but were found to be erroneous), benchmarked from the 2007
Economic Census (Geographic Research Inc. 2015b). Since Woods & Poole' s projections began in
2012, we decided to calibrate the model as closely as possible to the most recent historical data point
from 2014 by adjusting the initial value (2000) for retail consumption to a value between the two
historical data sources and then used the Woods & Poole total retail sales growth rate to calibrate the
model to projections for 2015 to 2040.
Government Revenues (Including D-O LRP Costs and Revenues)
We obtained historical property tax and sales tax data from NC DOR Statistical Abstracts (NC DOR
48
-------
2013b, 2004-2014, 2013a, 2003-2013, 2013c) and Annual Financial Reports from Durham County,
Orange County, the City of Durham, and the Town of Chapel Hill, and assumed that all tax rates were
held constant at 2014 values between 2015 and 2040. (City of Durham Department of Finance 2009,
2014, Durham County Finance Department 2009, 2014, The Orange County Financial Services
Department 2009, 2014, Town of Chapel Hill Business Management Department 2009, 2014). We
obtained information related to the expected costs (including capital costs and operation and
maintenance costs) and revenues information for the D-O LRP from the Durham County and Orange
County Rail Investment Plans (DCHC MPO et al. 2011, Triangle Transit 2014, Triangle Transit et al.
2012). A summary of the cost and revenue assumptions used in our model for the D-O LRP is presented
in Figure 3-12.
Light Rail Capital Cost: $1.345 Billion (split evenly per year of light rail construction 2020-2025)
Light Rail O&M Cost: $14.09 Million/year (beginning in 2025)
1/2 Cent Sales Tax Revenue: Begins collection in FY2013 for two months (model year 2013), then all of 2014
= 18% of total local sales tax collected (Durham and Orange County).
$3 and $7 Vehicle Registration Fees: Begins collection in FY 2014 and increases at a 2% growth rate ($10
combined fee in nominal dollars is converted to 2010 dollars for 2014-2040).
Rental Car Tax: Begins collection in FY 2014 and increases at a 4% growth rate.*
Sources: Durham and Orange County Bus and Rail Investment Plans, NC DOR Local Option Sales Tax Reports.
Figure 3-12. D-O LRP Cost and Revenue Assumptions (2010 Dollars)
Calibration
We adjusted several exogenous inputs in the Tier 2 and Tier 1 economy sectors to improve the fit
between estimated values for key variables in the business-as-usual (BAU) scenario and historical data
and projections. "Gross operating surplus per sq ft" was calibrated so that, with the effects of congestion
and energy spending included, gross operating surplus in Tier 1 and Tier 2 would make up 40% and
30% of GRP, respectively. This was required to ensure that the model's historical estimates of gross
operating surplus, which are influenced by several endogenous factors, coherently represent the trends
observed in the data and do not under- or over-estimate change over time. We made additional
adjustments to ensure that the model's estimates of total retail consumption for Tier 2 more closely
match historical trends. This included slightly adjusting "resident percent of the working population"
from the values we originally calculated from historical data and increasing the initial value for "retail
consumption." We slightly modified the value originally calculated from historical data for
"employment per dollar of consumption" in both Tiers in order to calibrate total employment. Finally,
we added the variable, "effect of unemployment on net migration tier 1" to ensure that the model did not
project unemployment levels to fall below zero (this addition is discussed in greater detail in the
following section).
49
-------
Equity
Sector Relationships
The Equity sector is primarily composed of indicators and has relatively few feedbacks to other sectors
of the model. Some of the key indicators are shown in blue in Figure 3-13, including property values
(disaggregated into single-family, multifamily, and nonresidential), renter costs, transportation costs,
population in poverty, zero-car households (a proxy for the transit-dependent population), and
affordability indices. The green balancing loop highlights how the number of zero car households, which
is affected by the population in poverty, feeds back into the transportation sector of the model, where it
reduces VMT and ultimately increases transit ridership. Though the sector does not have many
feedbacks to the wider model, the sector does respond to many other sectors in the model. From the
economy sector, unemployment rate affects the population in poverty, and resident net earnings, GRP
growth rate, and job density affect affordability indices; from the transportation sector, transportation
costs affect affordability, and commute time (which is affected by congestion) affects residential
property values; and from the land use sector, residential density, nonresidential density, retail density,
and available land affect property values. Property values in particular are driven by several other
variables in the model. Single-family property value is affected by the population growth rate, vacant
land, income, lot size, job density, commute time, and retail density. Multifamily property values are
impacted by many of the same factors (minus population growth rate, lot size, and job density), as well
as commercial building size and retail density. Nonresidential property values are affected by just three
factors: building size, employment growth, and retail density.
There are two affordability indices in the model, both of which are a factor of both transportation and
renter costs. The primary affordability index is calculated as the ratio of per-capita earnings (indexed to
the initial value) and transportation and renter costs (indexed to the initial value). Therefore, the higher
the affordability index, the more affordable it is to live in the Tier for someone who makes average
earnings. A second index is calculated as the ratio of a constant housing and transportation cost
threshold (set at 45% of the poverty threshold for the average household size in the Tier) and 75% of the
actual transportation and renter costs. This index compares forecasted costs to a threshold based on the
assumption that wage increases for the average household may not affect households in poverty, and it
accounts for the fact that housing and transportation costs will be lower for such households than for
average households.
The other two primary output indicators shown in Figure 3-13 are transit-dependent households, which
uses the proxy of zero-car households in the model, and the households in poverty at risk of
displacement, which is defined as the number of households in poverty not accommodated in subsidized
dwelling units. The housing gap is a related indicator, defined as the number of households neither
accommodated in subsidized dwelling units nor in organically affordable dwelling units (i.e., the percent
of market-rate multifamily units that cost no more than 30% of the monthly income for a household
making 60% of the Area Median Income).
50
-------
Data Sources and Processing
Historical Demographic Data
We obtained most of the data for this sector from the same sources as for the Land Use sector. One
exception is SimplyMap, which provided summarized data from the U.S. Census Bureau. We used
Census data from SimplyMap (Geographic Research Inc. 2015a) to corroborate other data and, in the
cases of median annual renter costs, percent of population in poverty, and zero-car households, for
calibration purposes.
Property Value Data
Estimates for property values were derived from the Durham County Tax Administration Real Property
Database (Durham County Tax Administration 2000-2014), Orange County Parcel Database (Orange
County Tax Administration 2014), and Chatham County Tax Parcel Database (Chatham County Tax
Administration Office 2014). To obtain estimates for our Tiers, property values in 2014 from Durham,
Orange, and Chatham were summed by category (commercial, single-family, multifamily, and vacant).
In Tier 2, we back-cast estimates for previous years by applying the percent change in total residential
and commercial real property value for Orange and Durham Counties from the Counties' Financial
Reports to the 2014 value. In Tier 1, the percent change in the Durham portion of Tier 1 year to year was
applied to the 2014 value for the entire tier. Finally, values were converted to 2010 dollars. More detail
on the methods used to download and scale data to the model's Tiers is provided in Chapter 6.
congestio:
transit person
miles
transit-
dependent
population
households
in poverty at
risk of
displacement
light rail project
nonresidential
developed land (by
categories)
employment
unemployment
rate
resident net
earnings
density
median housing
costs
developed land
' median household
transit costs
Housing and
— Transporation costs
Figure 3-13. CLD for the Equity Sector
51
-------
Calibration
This sector makes heavy use of elasticities and effect tables, both those supported by the literature and
those derived from other variables. That fact combined with the challenge of capturing location-specific
market dynamics and complex relationships between property values and other variables not yet well
documented in the literature, make this one of the more uncertain sectors. Nonetheless, we are confident
that the key indicators in this sector move in the right direction in response to an array of endogenous
model variables. The most significant calibration that took place in this sector was the calibration of
property values. A large amount of literature exists linking various unit-level (e.g., number of
bedrooms), local (e.g., accessibility from the home to retail), and regional (e.g., population growth)
characteristics to single-family and commercial property values. The literature on factors affecting
multifamily property values, on the other hand, is relatively sparse. We therefore borrowed many
relationships from the literature on single-family and commercial property values to develop an
endogenous model for multifamily property values. Furthermore, the research that does exist typically
focuses on a particular city and one point in time, and we did not identify any meta-analyses that
summarized the literature. The literature therefore provides a range of elasticities for many
characteristics, and we made several judgment calls to determine which characteristics could be modeled
with any certainty, what values linking a particular characteristic to property values are reasonable in the
local context, and how to account for factors that cannot be captured at the model's scope (e.g., number
of bedrooms or quality of nearby schools). In a few cases, elasticities had to be modified from literature
values to fit the study area (see Appendix B for details on the modifications made). In addition, a
housing crash was exogenously introduced, starting in 2005 to better reproduce this global phenomenon
and therefore better fit data on housing costs. Figure 3-14 illustrates the factors affecting property values
that we ultimately chose to include in the model for the three categories of properties.
Single
family
only
Commute
Time
Available
Land
Retail
Density
Residential
property
values
Multi
family
only
Building Size
Employment
Growth
Nonres
property
values
Figure 3-14. Factors Affecting Single-family, Multifamily, and Commercial Property Values in the D-O
LRP SD Model
52
-------
To a smaller degree, a similar calibration process took place for median annual renter costs, though far
less literature is available on the topic. Therefore, with one exception, the elasticities governing the
relationships with renter costs were borrowed from the literature on property values, derived from
available data, or exclusively calibrated. The one elasticity found in the literature is that linking renter
costs to renter vacancy, for which we used the median value from a study of 17 U.S. cities (Rosen and
Smith 1983). In Tier 1, the elasticity of income to property values used in the property value section was
borrowed and applied to renter housing costs as well. (Capozza et al. 2002a). We calibrated the table
governing the effect of GRP growth rate on renter housing costs based on the assumption of a balanced
S curve shape to the effect. The elasticity of renter costs to multifamily property values was purely
calibrated to match historical data on renter housing costs (U.S. Census Bureau 2000, Geographic
Research Inc. 2015a (Tier 1), U.S. Census Bureau American Community Survey 2014) (Tier 2)). See
Appendix B: Detailed Documentation by Sector, for a variable by variable accounting of the calibration
steps taken.
The addition of the variable "effect of unemployment on net migration tier 1," was another significant
calibration step that affected unemployment and therefore the percent of the population in poverty. In
the absence of this variable, unemployment in the model was driven primarily by economic activity
(specifically, total retail consumption). Because the model forecasts economic activity to rise faster than
population, it forecasted unemployment to fall below zero. Adding a nonlinear link between
unemployment and migration, which boosts migration when the unemployment rate is below 3% and
lowers it sharply as the unemployment rate goes above 6%, allowed the model to better represent reality,
in that a decrease in unemployment in an area causes more people to move to the area for jobs, which
raises the resident population and increases unemployment until it reaches a new equilibrium. This
restructuring allowed a more accurate endogenous connection to the percent of the population in poverty
in Tier 1.
Water
Sector Relationships
The water model simulates water supply, demand, and stormwater runoff. Within stormwater runoff, the
model simulates loadings of nitrogen (N) and phosphorus (P) in stormwater, as well as runoff volume.
Water supply is affected by annual rainfall, reservoir volume, and water demand. Average annual
rainfall is kept constant, but a random number generator is used to project future annual rainfall,
reflecting historic variation around the average. This random projected rainfall is used for calculating
water supply, but for ease of trend visualization, stormwater runoff is simulated using a constant average
rainfall. Water demand is driven by several outputs from the land and economy sectors, including the
number of single-family and multifamily dwelling units and total employment. Water use intensity for
nonresidential buildings and for single-family and multifamily dwelling units factor into water demand,
with a slight effect of building density on single-family residential water use intensity
(gallons/day/dwelling unit).15 Water demand is an output to the energy sector, where it determines the
energy used for water treatment and distribution. Water demand is modeled for both Tier 1 and Tier 2,
15 The elasticity of single-family residential water use to residential density (assumed -0.1) is based on a study in Portland,
OR (Chang et al. 2010). Therefore a doubling (100% increase) in single-family residential density would cause a 10%
decline in single-family residential water use.
53
-------
while water supply is modeled only for Tier 2, because local reservoirs operate at this regional scale.
Stormwater is modeled for Tier 1 and 2. Using the Simple Empirical Model for stormwater runoff
(Shaver et al. 2007), the model calculates runoff volume and N and P loadings as a function of annual
rainfall, impervious coefficient (impervious area/total land area, equivalent to percent impervious, an
output from the land use sector), land area, and event mean concentrations (EMCs) for each pollutant:16
Stormwater Pollutant Loading = Annual Rainfall Volume x Runoff Coefficient x EMC
Annual Rainfall Volume = Annual Precip x Land Area
Runoff Coefficient = 0.05 + (0.009 x impervious coefficient x 100)
The model estimates EMCs for N and P based on average values from local and national literature and
percent impervious cover; EMCs for N are shown in Figure 3-15. Stormwater is modeled separately by
land use type, then summed for total N and P load for each Tier.
We estimated EMCs for each land use using the projected impervious coefficient for that land use and
literature-based EMCs taken from the Jordan Lake Stormwater Load Accounting Tool User's Manual
(NCDENR 2011). In some cases we calculated weighted average EMCs; for example, we estimated the
EMC for SF residential stormwater N as a weighted average of the EMC for lawns and roofs:
EMC N SF Residential
= EMC N lawns or open space x (1 — SF impervious coefficient)
+ EMC N roofs x SF impervious coefficient
Modeled EMCs for different land use types in 2015 are shown in Figure 3-15, ordered from largest
(agricultural and open space) to smallest (roads and nonresidential). Although agricultural land and open
space are assumed to have the same EMCs in Tier 1 and Tier 2, the other EMCs are 6-13% lower in Tier
1 than Tier 2 due to higher impervious coefficients (less open space) in Tier 1.
3.50
• Tier 1
i Tier 2
Figure 3-15. Storm Event Mean Concentrations (EMC) of Nitrogen in Both Tiers, Modeled for 2015
16 EMCs are converted from mg/L to Ib/L when calculating stormwater pollutant loading. In the loading formula, the runoff
coefficient is the fraction of rainfall volume that becomes runoff.
54
-------
As Figure 3-16 shows, several variables from other sectors can affect indicators in the water sector.
From the land use sector, population growth and changes in the share of single-family or multifamily
residences can affect water demand, as can changes in employment from the economy sector. Changes
in impervious cover (resulting from land development in the land use sector) can affect loadings of N
and P in stormwater. The model can also run user-specified scenarios that would affect variables in the
water sector, including changes in annual rainfall, changes in reservoir capacity, or implementation of
low-impact development technologies such as bioretention (rain gardens).
light rail
project
event mean
concentrations, N
andP
SW green
infrastructure
population
annual
precipitation
overflow
SF and MF dwelling units
employment
energy for water
treatment and
distribution
water consumption
intensities: SF, MF,
nonresidential
Figure 3-16. CLD for the Water Sector
Data Sources and Processing
Water supply
We obtained historical annual rainfall data for the Northern Piedmont region (covering the years 2000-
2013) from the State Climate Office of North Carolina (State Climate Office of North Carolina 2015).
Data for the combined volume of Durham County's drinking water reservoirs in 2015 came from the
City of Durham Department of Water Management (City of Durham Department of Water Management
2015). To estimate reservoir volume in 2015, we multiplied current days of supply by current daily use
rate (Mgal/day).
Water Demand
Data on historical water demand for Durham County came from the North Carolina Department of
Environment and Natural Resources (NCDENR) (NC State Data Center 2015). We obtained projected
future water demand data from the TJCOG Triangle Regional Water Supply Plan (TJCOG 2012, 2014f),
which projects residential water use rate (gallon/day/person), nonresidential use rate
(gallon/day/employee), total demand, and demand by sector (residential, nonresidential, and
nonrevenue, in Mgal/day). These projections cover the years 2010-2060, but our model uses their
55
-------
projections to 2040. For both historical and projected Durham County water use data, we scaled values
for Durham County to Tier 2 based on the ratio of Durham County population to Tier 2 population
(0.67).
Impervious Surfaces and Stormwater
The model calculates stormwater runoff using precipitation data (noted above) as well as impervious
surface data drawn from the EPAs EnviroAtlas online database (US EPA 2015e). This source provided
impervious surface data with a resolution of one square meter for about 86 percent of the study area.
Once we estimated total impervious surface area in the portion of Tier 2 covered by EnviroAtlas, we
multiplied that estimate by 1.17 to produce an estimate of total impervious surface area for Tier 2
(assuming that the area of overlap is representative of percent impervious surface for all of Tier 2). We
corroborated this estimate using impervious surface data shared by the Durham City/County Planning
Department, which only addressed impervious surfaces on parcels within the City of Durham (e.g.
excluding roads). To estimate runoff loadings of N and P, we obtained event mean concentration (EMC)
data from the Jordan Lake Stormwater Load Accounting Tool User's Manual (NCDENR 2011). The
EMC values in that source are based on a literature review which included many values from the
Piedmont region of North Carolina.
Calibration
We calibrated key output variables from the water sector to historical (before 2015) and projected (after
2015) data for Durham County, scaled to Tier 2 (assuming a constant Durham County/Tier 2 population
ratio of 0.67). These variables include total water demand, residential water use, nonresidential water
use, and nonrevenue water use. For each of these variables, we were able to achieve a close fit with
historical data and projections (R2 values were greater than 0.94 and average deviations were less than
five percent for the combined dataset, which included both historical and projected data).
Residential water use makes up the majority of Tier 2 water use, and we identified data sources for total
residential water use, but not separate water use by single-family and multifamily units. Instead, we
obtained average single-family and multifamily residential water use rates from national averages in the
2010 Buildings Energy Data Book (US DOE 2011) and applied one calibration factor to all residential
water use so that it matched local data.
Health
Sector Relationships
As with the equity and water sectors, the health sector of the D-O LRP SD Model is primarily output-
oriented. For any scenario run in the model, the health sector translates changes in variables in the
transportation and energy sectors into positive and negative premature mortality and morbidity
outcomes. Directly comparing the health effects of the business-as-usual (BAU) scenario to alternative
scenarios allows the user to isolate the estimated health benefits and detriments above and beyond those
of other actions that would be taken anyway.
The health sector quantifies the net avoided premature mortality that results from scenario changes
relative to BAU in: 1) vehicle emissions of PM2.5 and NOX, 2) walking and cycling for transportation,
and 3) vehicle crash fatalities. The main outcome variable for the health sector is "net premature
mortalities avoided per year from PM2.5 and NOX vehicle emissions, physical activity, and crash
fatalities relative to BAU," which is the sum of the positive and negative effects on premature mortality
56
-------
of all three transportation- and energy-related outcomes. The only feedback loop within the health sector
is a reinforcing loop (labeled Rl in Figure 3-17) in which changes in mortality risk due to increases or
decreases in physical activity affect the overall death rate for the population of the study area.1?
N Ox vehicle
emissions
NOx vehicle
emissions BAU
premature mortalities
avoided/caused per ton of
PM2.5 removed/added
total crash fatalities per
year rektive to BAU
premature mortalities avoided per
year due to waking and eye
rektive to BAU
premature mortalities
avoided/caused per ton of
NOx removed/added
premature mortalities avoided per
year from PM2.5 and NOx vehicle
emissions rektive to BAU
person miles of
nonmotonzed travel per day
reduction in mortality as a result
of walking and cyclinj
transportation
premature mortalities avoided
per year due to walking and
cycling BAU
Figure 3-17. CLD for the Health Sector
Data Sources and Processing
Premature Mortality and Morbidity Estimates from Changes in Vehicle Emissions
To estimate the health impacts of changes in vehicle emissions of PlVh.s and NOX on mortality and
morbidity (due to changes in vehicle use), we used incidence-per-ton estimates from an EPA nationwide
study of the health impacts of removing directly emitted PlVb.s and PlVb.s precursor emissions, including
NOX, from 17 sectors, including on-road mobile sources (US EPA 2013c). Because the values used for
avoided mortality and morbidity per reduced ton of PlVh.s or NOX vehicle emissions were from a
nationwide study and were not meant for smaller geographic scales, there is considerable uncertainty
about the use of these figures in our model, including translating changes in emissions to changes in
ambient pollutant concentrations, and relating changes in ambient pollutant concentrations to changes in
health outcomes. Nevertheless, we chose to include the numbers from this study to illustrate the
magnitude of the health impacts of air emissions, relative to the magnitude of other health effects in the
model (e.g., changes in physical activity and vehicle-crash fatalities).
17 We decided not to include the premature mortality effects of changes in emissions reductions and crash fatalities (relative
to the BAU) on the overall death rate, because testing indicated that the effect of these factors on the death rate was
negligible.
57
-------
Health Benefits of Physical Activity
To estimate the health impacts of model changes in physical activity, or, more specifically, "person
miles of nonmotorized travel by residents per day per capita," we used equations from the World Health
Organization/Europe Health Economic Assessment Tool (HEAT) (WHO 2014), which is freely
available online. The HEAT equations estimate the health effects of active transport over an entire
population based on daily average amounts of walking and cycling. Although HEAT is meant to be used
for daily average amounts of walking and cycling for all purposes, which the D-O LRP SD Model does
not estimate, we instead applied the HEAT equations in the model to the average walking and cycling
distances traveled by residents per day solely for transportation purposes (i.e., from trips for which the
main mode of transportation was nonmotorized), as well as an assumption of 0.25 miles of walking or
cycling per one-way public transit trip. Because the HEAT model assumes a delay between increases in
physical activity and measurable benefits to health, we also applied a five-year delay between a change
in walking and cycling and the associated avoided premature mortality. We also note that the estimates
of premature mortalities avoided due to walking and cycling predicted by the HEAT equations are likely
to be conservative estimates of the total health benefits of physical activity, since they do not account for
the beneficial effects of physical activity on many aspects of morbidity. This is because current evidence
on morbidity, both for walking and cycling, is more limited than that on mortality (WHO 2014).
Premature Mortality and Injury Estimates from Vehicle Crashes
We obtained local data on historical occurrences of traffic accidents, injuries, and fatalities by county
from the NC Crash Data Query Web Site (Highway Safety Research Center at UNC Chapel Hill 2015),
which we used to calibrate variables related to traffic accidents and fatalities in the D-O LRP SD model.
The D-O LRP SD Model estimates values for these variables based on historical numbers of people
involved in reportable traffic accidents and the percentages of them who either died or were injured,
under the assumption that the relationship between VMT and people in traffic accidents (including
people not in vehicles) is constant over time. We investigated the possibility of having the model's
estimated number of vehicle-crash fatalities per year be affected by other variables, including
nonmotorized travel volumes, vehicle speeds, the frequency of intersections, and traffic-control
measures, but we were not able to find data or studies that would allow us to relate these factors to crash
and crash-fatality rates at a regional scale.
Calibration
We did not perform any calibration to data or projections for the vehicle-emissions and physical-activity
components of the health sector of the model. In the vehicle-crashes component, we adjusted the input
variable "people in traffic accidents per VMT" so that the number of people in traffic accidents per year
in Tier 2 ("people in traffic accidents per year") would match data from the NC Crash Data Query Web
Site for Orange and Durham Counties in the year 2013, the most recent year for which data are available
from that authoritative source (data are available for 2001-2013) (Highway Safety Research Center at
UNC Chapel Hill 2015). We assume that the number of people in traffic accidents within Tier 2 is the
same as the number of people in traffic accidents in the combined area of Orange and Durham Counties,
even though Tier 2 (the DCHC MPO) does not conform to county boundaries. Because crash statistics
are not available for Tier 1, we assume that the Tier 2 value of "people in traffic accidents per VMT"
resulting from this calibration process also applies to Tier 1. We also assume that the percentage of
people involved in traffic accidents who are either injured or killed is the same in both Tiers.
58
-------
4 Scenario Descriptions
This chapter describes the scenarios that were run in the D-O LRP SD Model. We designed these
scenarios to explore the effects of the light rail project and a range of policies and other factors that
could affect land use, transportation, economic and demographic trends, and energy use. The chapter
begins with a description of the three main scenarios - Business As Usual (BAU), Light Rail, and Light
Rail + Redevelopment (Light Rail + Redev). It then describes 17 policy, demographic, and market
scenarios, grouped into the following categories:
• Land Use Planning Changes
• Transportation Planning Changes
• Market Trends
• Demographic Trends
• Environmental Policy Changes
For the additional scenarios, we list primary variables to review for changes under each scenario.
However, these are not listed for the main scenarios, due to the large number of key variables and
indicators affected under those scenarios. Most of the indicators on the model dashboard (the summary
of policy switches and model indicators in the Vensim model, described in greater detail in Appendix A)
would be interesting to review for changes under the two light rail scenarios.
4.1 Main Scenario Descriptions
The three main scenarios serve as the reference points to which the 17 additional decision support
scenarios are compared. The BAU scenario represents the without-light rail baseline, the Light Rail
scenario represents the implementation of the light rail transit (LRT) line without any additional
changes, and the Light Rail + Redevelopment scenario represents the implementation of the LRT line
with additional changes to zoning to encourage redevelopment and increased density around the station
areas.
Business As Usual (BAU)
Scenario Switch: Change the "Main policy switch" to 0.
Scenario(s) for Comparison: Not Applicable (baseline case)
Scenario Summary
The Business As Usual (BAU) scenario is modeled on the assumption that current demographic, land
use, and transportation trends continue, with values either remaining constant at current values or
extending along historical trends. The scenario is compared to outputs from other projection models for
purposes of calibrating the D-O LRP SD Model, and it serves as a baseline against which to compare the
effects of the changes described in the other scenarios.
59
-------
Assumptions and Variable Changes
The BAU scenario makes the following main assumptions:
• Land Use
o Employee-space ratios remain constant at 2014 values, except for ratios for office and
industrial uses, which are extended upward and downward along historical trends,
respectively.
o Floor-area ratios and residential densities remain constant at 2014 values.
o The division of dwelling units into single-family and multifamily units and their
respective household sizes remain constant at values equal to the average values in 2014.
• Transportation
o Road building between 2010 and 2040 matches the DCHC MPO 2040 Metropolitan
Transportation Plan (CAMPO and DCHC MPO 2013), as projected by the Triangle
Regional Model (DCHC MPO 2013).
o The ratio of miles of nonmotorized travel facilities per developed acre that data indicate
for 2013 (the only year for which historical data are available for developed acres
(TJCOG 2014b); 2013 facility miles are interpolated between 2010 and 2020 figures
(DCHC MPO 2013) is a result of the builders of nonmotorized travel facilities attempting
to maintain a ratio of miles per developed acre that is assumed to have existed in 2000,
with a two-year delay between when additional acres are developed and when the
consequent additional miles of nonmotorized travel facilities are completed. In the event
of a decline in developed acres, miles of nonmotorized travel facilities hold steady.
o All changes in public transit non-rail revenue miles per day are in accordance with the
Bus and Rail Investment Plans for Orange and Durham Counties (Triangle Transit et al.
2012, DCHC MPO et al. 2011).
o Most determinants of person miles of travel by mode (automobile driver, automobile
passenger, public transit, and nonmotorized) change the distribution of total person miles
of travel among these modes, rather than changing the overall number of person miles of
travel by all modes.
o Between 2010 and 2040, the ratio between traffic congestion and the amount of weekday
peak-period VMT per lane mile decreases linearly by the amount predicted by the TRM
projections created for the DCHC MPO 2040 MTP (DCHC MPO 2013).
o Barring the effects of changes in the cost of vehicle fuel and the average speed of vehicle
traffic, the VMT of through traffic (i.e., vehicle trips that pass through one or both Tiers
without stopping) increases by 1.0% per year during 2014-2040, after matching the
historical rate of population growth in North Carolina during 2000-2013. The 1.0%
assumption for through-traffic VMT growth is based on the growth rate of North
Carolina's population in 2013 (NC Office of State Budget and Management 2015).
60
-------
o The number of people involved in traffic accidents per VMT stays constant at a value that
is calibrated to match reported people in traffic accidents per year in 2013, and the
percentages of accident-involved people who are injured or killed remain constant at
2013 levels, after matching historical percentages injured or killed during 2001-2013
(Highway Safety Research Center at UNC Chapel Hill 2015).
Economy
o Total employment is consistent with projections developed by local planners for the
DCHC MPO area (Tier 2) for use in the TRMvS, which allocated the expected
employment growth (2010-2040) by TAZ according to the "preferred growth scenario"
option generated by local planners using Community Viz. (DCHC MPO 2013) (TJCOG
2013).
o Employment and earnings by job category (industrial, office, retail, and service) are
consistent with projections (2012-2040) by Woods and Poole Economics, Inc. for
Durham and Orange County18. (Woods & Poole Economics Inc. Copyright 2014).
o Total earnings are projected to continue to make up 60% of GRP in Tier 2 (70% for Tier
1) between 2012 and 2040. Therefore, the gross operating surplus constitutes the
remaining 40% of GRP in Tier 2 (30% in Tier 1).
o Retail consumption grows similarly to retail sales growth rate projections from Woods &
Poole Economics, Inc. for Durham and Orange County, being influenced by relative
growth in GRP and the resident working population (Tier 2 only).
o Local sales tax rates, property tax rates (both city and county), and the portion of the local
sales tax collected that goes toward the D-O LRP transit fund remain at 2014 levels. The
percent of nonresidential land that is tax exempt in both Tiers also remains constant at
2014 values.
• Energy
o Building energy use intensities decline from 2015 to 2040 according to US EIA Annual
Energy Outlook 2015 (AEO2015) projections (US EIA 2015a).
o Passenger vehicle MPG increases from 2015 to 2040 according to US EIA AEO2015
projections (US EIA 2015a).
o Energy use in buildings continues to be primarily electricity and natural gas.
o Energy prices (electricity, natural gas, gasoline) rise in real terms from 2015 to 2040
according to US EIA AEO2015 projections (US EIA 2015a).
See Chapter 5: Model Results to view detailed results regarding the calibration process and
18 Woods & Poole does not guarantee the accuracy of this data. The use of this data and the conclusions drawn from it are
solely the responsibility of the US EPA.
61
-------
consequences of these baseline model assumptions.
Light Rail
Scenario Switch: Change the "Mainpolicy switch" to 1.
Scenario(s) for Comparison: BAU
Scenario Summary
The Light Rail scenario is meant to show what would happen if the planned 17-mile light rail transit
(LRT) line between Durham and Chapel Hill were built, but no other policy changes were instituted that
deviate from the BAU case. In this scenario, any differences in land use between the BAU and Light
Rail scenarios are due entirely to increases in demand that the rail line itself causes, rather than to any
changes in zoning (which will be part of the Light Rail + Redevelopment scenario).
Assumptions and Variable Changes
This scenario is consistent with the BAU scenario, with the following changes:
• "LRT line mile construction" is set to be equal to "LRT construction schedule," which dictates
that construction of the light rail line will take place during 2020-2026, such that "finished LRT
line miles" will reach 17 in 2026 (Triangle Transit 2015b) and the value of the variable "LRT
open" will change from 0 to 1. The values for "finished LRT line miles" resulting from this
change for both Tiers are shown in Table 4-1.
Table 4-1. Values for "Finished LRT Line Miles" in Tier 1 and Tier 2 under the Light Rail Scenario
SCENARIO NAME
FINISHED LRT LINE MILES
2015
2019
2020
2023
2025
2026
2040
Tierl
Light Rail
BAU
0
0
0
0
0.41
0
7.15
0
11.36
0
11.55
0
11.55
0
Tier 2
Light Rail
BAU
0
0
0
0
0.73
0
10.73
0
16.80
0
17
0
17
0
• To reflect the expectation that the construction of the LRT line will require the temporary closure
of nearby roadways in order to accommodate construction equipment, this scenario changes the
values for the stock variable "urban nonhighway lane miles in disruption from LRT
construction." During the period 2020-2026, this variable will have the values shown in Table
4-2, based on the assumption that two lane miles of roadway will be disrupted for each LRT line
mile currently under construction. These lane miles in disruption represent a temporary reduction
in "functioning urban nonhighway lane miles."
62
-------
Table 4-2. Values for "Urban Nonhighway Lane Miles in Disruption from LRT Construction" in Tier 1 and
Tier 2 under the Light Rail Scenario
SCENARIO NAME
URBAN NONHIGHWAY LANE MILES IN DISRUPTION FROM
LRT CONSTRUCTION
2015 | 2019
2020
2023
2025 | 2026 | 2040
TieM
Light Rail
BAU
0
0
0
0
1.35
0
1.73
0
0.38
0
0
0
0
0
Tier 2
Light Rail
BAU
0
0
0
0
1.73
0
2.13
0
0.40
0
0
0
0
0
• "LRT construction cost per line mile" is set at $79,112,700 per mile, and once the rail line is near
completion, "LRT O and M cost per year" is set to be $14,093,300 per year (both in 2010 USDs)
(Triangle Transit et al. 2012, DCHC MPO et al. 2011).
• The value of "land use per LRT line mile" is assumed to be 2.42 acres per line mile, a portion of
which is counted towards the study area's impervious surface.
• Once the rail line is open, it will add 2,310 revenue miles per day to the public transit system.
We calculated this value on the basis of anticipated service frequencies (Triangle Transit 2015b).
• The light rail trains are assumed to consume 62,800 Btu of electricity per revenue mile. This
factor was previously used to estimate energy consumption of the LYNX light rail line in
Charlotte, NC (FTA and CATS 2011).
• The effect of the rail line on "person miles of public transit travel" is taken to be determined by a
different formula than the effect of other public transit services. Once the LRT is open, the
variable "change in person miles of public transit travel per year due to adding fixed guideway
transit" is activated, which is driven by "population Tier 1," "total employment Tier 1," "retail
plus entertainment employment Tier 1," "jobs earning $3,333 per month in 2010 USDs Tier 1,"
and Tier 2 "VMT per highway lane mile," all of which indicate greater demand for fixed-
guideway public transit (rail or bus routes on exclusive rights-of-way) (Chatman et al. 2014).
The entire change in Tier 2 person miles of public transit travel that results from the LRT line is
taken to come from Tier 1. In contrast, changes in non-rail public transit revenue miles on the
overall transit system affect both Tier 1 and Tier 2 person miles through an elasticity derived
from a different source (Sinha and Labi 2007).
• The "share of desired employment that is net migration Tier 1" is allowed to rise from a
maximum of 5% to a maximum of 10%.
• The "Effect of LRT on retail/office/service sq ft tier 1" variables are set to 10, indicating that
along with the light rail comes a 10% increase in demand for developed commercial (excluding
industrial) floor space in the station areas. This increase is gradually phased in during the 6-year
period of light rail construction.
63
-------
Light Rail + Redevelopment
Scenario Switch: Change the "Mainpolicy switch" to 2.
Scenario(s) for Comparison: BAU and LRT
Scenario Summary
The Light Rail + Redevelopment scenario aims to show a case similar to the Preferred Growth Scenario
of the Imagine 2040 Regional Model. In this scenario, the light rail is constructed, zoning is altered to
encourage Transit-Oriented Development, and new demand spurs more and denser growth around the
station areas. Local planners believe this to be the most likely land use and transportation scenario,
provided the light rail is built.
Assumptions and Variable Changes
This scenario maintains all changes from the Light Rail scenario and adds the following changes:
• The value for "overall target percent of land redeveloped table tier 1" increases gradually from 0
percent in 2020 to 20 percent in 2040. This change begins several years before station
construction is completed, in anticipation of increased demand for residential and commercial
space.
• The value for "redeveloped density multiplier Tier 1" is changed from 1 (its value in the BAU
and Light Rail scenarios) to 2.93, indicating that all redeveloped land gets developed at a density
193 percent higher than previous levels in Tier 1. We used the value of 193 percent to match the
weighted-average increase in jobs and housing density realized across the station areas in the
Preferred Growth Scenario of the Imagine 2040 Regional Model (Green 2015).19
4.2 Additional Scenario Descriptions
Land Use Planning Changes
BAU + Redevelopment Scenario
Scenario Switch: Change the "Mainpolicy switch" to 4.
Scenario(s) for Comparison: BAU and Light Rail + Redevelopment
Scenario Summary
This scenario isolates the effects of redevelopment from the introduction of the Light Rail. It conforms
to the BAU scenario, plus the changes made for the redevelopment scenario.
19 The Preferred Growth Scenario of the Imagine 2040 Regional Model projects that the density of all developed land around
the light rail station areas (not just the redeveloped land) will increase by 193 percent. Because a change of this magnitude
produced unusual development patterns in our model, we did not use it for our primary scenario but instead illustrated it in
one of the Land Use Planning scenarios.
64
-------
Assumptions and Variable Changes
This scenario keeps all assumptions the same as the BAU scenario, with the following exceptions:
• The value for "overall target percent of land redeveloped table tier 1" increases gradually from 0
percent in 2020 to 20 percent in 2040.
• The value for "redeveloped density multiplier Tier 1" is changed from 1 (its value in the BAU
and Light Rail scenarios) to 2.93, indicating that all redeveloped land gets developed at density
193 percent higher than previous levels in Tier I.20
The primary variables to review for changes under this scenario include:
• Population Tier 1
• Total nonresidential sq ft Tier 1
• Developed land Tier 1
• SF property value per SF DU Tier 1
• Total impervious surface Tier 1
• Impervious surface per capita tier 1
• Transportation related costs incurred by residents per year per capita Tier 1
• Congestion Tier 1
Bold Redevelopment Scenario
Scenario Switch: In the Redevelopment Switches box, increase the "overall target percent of land
redeveloped table tier 1" value to 0.50 in 2040 and increase the "redeveloped density multiplier tier 1"
from 2.93 to 6.
Scenario(s) for Comparison: Light Rail + Redevelopment
Scenario Summary
In this scenario, the percent of land redeveloped and the density at which is it redeveloped is set high
enough to reach an overall increase in density of 193% for all land in Tier 1, developed and redeveloped.
This change matches the results of the Preferred Growth Scenario of the Imagine 2040 Regional Model
(Green 2015).
20 A Ml listing of the policy, demographic, and market switches meant to be modified by users can be found in the User's
Guide (Appendix A, Table A-l).
65
-------
Assumptions and Variable Changes
This scenario is consistent with the Light Rail + Redevelopment scenario, with the following changes:
• The "overall target percent of land redeveloped table tier 1" value in 2040 is increased from 0.20
to 0.50, indicating that 50% of all land in the Tier will be redeveloped by 2040.
• The "redeveloped density multiplier tier 1" is increased from 2.93 to 6, indicating a 500%
increase in the density of land that is redeveloped.
Taken together, these changes lead to an approximately 193% increase in the average density of all
developed land in Tier 1.
The primary variables to review for changes under this scenario include:
• Developed land Tier 1
• Total nonresidential sq ft Tier 1
• Nonresidential property value per sq ft Tier 1
• SF property value per SF DU Tier 1
• VMT per capita Tier 1
• Total impervious surface Tier 1
• Impervious surface per capita tier 1
• Transportation related costs incurred by residents per year per capita Tier 1
• Congestion Tier 1
Energy Efficiency
Scenario Switch: In the Policy Switches box, change "policy switch building energy intensity" from 0 to
1. Also change "policy switch MPG" from 0 to 1.
Scenario for Comparison: Light Rail + Redevelopment
Scenario Summary
Buildings and vehicles represent the majority of energy use and CO2 emissions in the model.
Redevelopment in Tier 1 presents an opportunity to improve the energy efficiency of buildings. This
scenario reduces building energy use intensity by 10% in Tier 1 relative to the Light Rail +
Redevelopment scenario. Although redevelopment in the SD model is concentrated in Tier 1, this
scenario applies the same building energy efficiency trend to Tier 2 for comparison. This scenario also
includes a policy encouraging the use of more fuel-efficient vehicles in both Tiers, increasing average
MPG by 10% between 2015 and 2040, compared to the AEO2015 projected trend over that time.
66
-------
Assumptions and Variable Changes
This scenario is consistent with the Light Rail + Redevelopment scenario, with the following changes:
• The variable "building energy intensity policy" decreases linearly from 1.0 in 2015 to 0.9 in
2040 (Table 4-3). "Building energy intensity policy" is a multiplier that the model applies to
energy intensity projections.
• The variable "MPG policy" increases linearly from 1.0 in 2015 to 1.1 in 2040 (Table 4-4). "MPG
policy" is a multiplier that the model applies to MPG projections, as seen in Table 4-5.
Table 4-3. Values for "Building Energy Intensity Policy" in Tier 1 and Tier 2 under the Light Rail +
Redevelopment Scenario
SCENARIO NAME
BUILDING ENERGY INTENSITY POLICY
2015
2020
2030
2040
Both Tiers
Energy Efficiency
Light Rail + Redevelopment
1
1
0.98
1
0.94
1
0.9
1
Table 4-4. Values for "MPG Policy" in Tier 1 and Tier 2 under the Light Rail + Redevelopment Scenario
SCENARIO NAME
MPG POLICY
2015
2020
2030
2040
Both Tiers
Energy Efficiency
Light Rail + Redevelopment
1
1
1.02
1
1.06
1
1.1
1
Table 4-5. Values for "MPG without Congestion" under the Light rail + Redevelopment Scenario
SCENARIO NAME
MPG WITHOUT CONGESTION
2015
2020
2030
2040
Both Tiers
Energy Efficiency
Light Rail + Redevelopment
16.5
16.5
18.6
18.2
25.0
23.6
29.7
27.0
The primary variables to review for changes under this scenario include:
• CO2 emissions from buildings and transportation
• Gross regional product
• Total energy spending
67
-------
VMT
• Congestion
• Nonresidential sq ft
Transportation Planning Changes
Wo Road Building Scenario
Scenario Switch: Change the "Mainpolicy switch" to 3.
Scenario(s) for Comparison: BAU
Scenario Summary
In the DO-LRP SD Model, the number of lane-miles of roadways built each year is determined entirely
by exogenous projections, based on the DCHC MPO 2040 Metropolitan Transportation Plan and the
Triangle Regional Model. Whether or not the supply of roadway lane-miles keeps pace with the growth
of VMT in the study area determines the average severity of traffic congestion, which has implications
for people's transportation mode choices, as well as for environmental, economic, and health outcomes.
To illustrate the effect of the MPO's projections for road building, this scenario tests what would happen
if a choice were made to not build any new lane-miles beyond what is already committed to be built by
the year 2017.
Assumptions and Variable Changes
The No Road Building scenario includes the following changes:
• Starting in 2016, the values of the exogenous variables "urban nonhighway lane mile
construction," "urban highway lane mile construction," "rural nonhighway lane mile
construction," and "rural highway lane mile construction" are set to zero.
• Because the model applies a one-year delay between the start and finish of any given lane-mile
under construction, "functioning lane miles" reaches its maximum value in 2017 and remains
constant thereafter. This is in contrast to the BAU scenario, where lane miles continue to grow
between 2017 and 2040. The differences between the values for "functioning lane miles" in the
No Road Building and BAU scenarios is illustrated in Table 4-6.
Table 4-6. Values for "Functioning Lane Miles" in Tier 1 and Tier 2 under the No Road Building and BAU
Scenarios
SCENARIO NAME
FUNCTIONING LANE MILES
2010 | 2017 | 2020 | 2025 | 2030 | 2035
2040
TieM
No Road Building
BAU
265
265
267
267
267
269
267
274
267
278
267
282
267
287
Tier 2
No Road Building
BAU
3,444
3,444
3,520
3,520
3,520
3,553
3,520
3,609
3,520
3,665
3,520
3,721
3,520
3,777
68
-------
Main Variables Affected
The primary variables to review for changes under this scenario include:
• Total impervious surface
• Congestion
• VMT
• Public transit unlinked passenger trips per day
• Person miles of nonmotorized travel by residents per day per capita
• Gross regional product
• CO2 emissions from buildings and transportation
• Cumulative premature mortalities avoided due to walking and cycling relative to BAU
Fare Free Transit Scenario
Scenario Switch: Change the "fare free transit system in 2026" switch from 0 to 1.
Scenario(s) for Comparison: BAU
Scenario Summary
One of the three major public transit agencies in the study area, Chapel Hill Transit, has been a fare-free
system since 2002, a policy decision which significantly increased its ridership. Therefore, this scenario
illustrates what would happen if all public transit in the study area were free to users, as opposed to the
assumption in the BAU scenario that the average price per trip remains constant in inflation-adjusted
dollars from 2013 forward. If fare-free transit shifts travelers to public transit from other modes, it will
produce effects in the areas of health, economics, and the environment. Note that the model does not
account for changes in public finances that would result from a loss of fare revenue.
Assumptions and Variable Changes
This scenario is consistent with the BAU scenario, with the following changes:
• Between 2025 and 2026, the value of the exogenous variable "public transit fare price" decreases
to $0.01 (we cannot set its value to $0.00 because that would produce a "divide by zero" error
within the model). In all other scenarios, this variable remains at a steady real-dollar value during
2013-2040. This change is illustrated in Table 4-7. The same fare price applies to both Tiers.
• "Money spent by residents on public transit fares per year per capita" and "money spent on
public transit fares per year per capita" both become zero in 2020.
69
-------
Table 4-7. Values for "Public Transit Fare Price" under the Fare Free, BAD, and Light Rail Scenarios
SCENARIO NAME
PUBLIC TRANSIT FARE PRICE
2015
2026-2040
Both Tiers
Fare Free Transit
BAU
Light Rail
$0.30
$0.30
$0.30
$0.01
$0.30
$0.30
The primary variables to review for changes under this scenario include:
• Public transit unlinked passenger trips per day
• Person miles of nonmotorized travel by residents per day per capita
• VMT
• Transportation and renter costs per year per household
• CO2 emissions from buildings and transportation
• Cumulative premature mortalities avoided due to walking and cycling relative to BAU
High Parking Price Scenario
Scenario Switch: Change the "parking price hike instituted Tier 1" switch from 0 to 1.
Scenario(s) for Comparison: Light Rail + Redevelopment
Scenario Summary
Stakeholders have expressed an interest in seeing what the implications would be of the cost of parking
going up in the areas around the future light rail stations (Tier 1). Such a change could affect people's
transportation mode choices, with indirect impacts on the economy, the environment, and public health.
This scenario therefore tests what would happen if future parking prices in the light rail station areas
experienced a sudden, dramatic increase.
Assumptions and Variable Changes
This scenario is consistent with the Light Rail + Redevelopment scenario, with the following changes:
• In 2020, the variable "parking cost of average trip Tier 1" increases by $4.00, relative to the
Light Rail + Redevelopment scenario. Otherwise, this variable increases in proportion to "jobs
per commercial acre Tier 1."
• In 2020, the variable for parking costs in Tier 2 ("parking cost of average trip") increases by an
amount equal to $4.00 times "Tier 1 percent of Tier 2 employment," which is meant to
approximate how much of overall Tier 2 parking demand is concentrated in Tier 1. As with Tier
1, this effect is on top of whatever changes occur due to forces that are endogenous to the model.
70
-------
These variable changes are summarized in Table 4-8.
Table 4-8. Values for "Parking Cost of AverageTtrip" in Tier 1 and Tier 2 under the High Parking Price
and Light Rail + Redevelopment Scenarios
SCENARIO NAME
PARKING COST OF AVERAGETRIP
2015
2020 | 2025 | 2030
2035
2040
TieM
High Parking Price
Light Rail + Redevelopment
$1.06
$1.06
$5.06
$1.06
$5.01
$1.01
$5.09
$1.09
$5.16
$1.16
$5.26
$1.26
Tier 2
High Parking Price
Light Rail + Redevelopment
$0.36
$0.36
$1.52
$0.36
$1.50
$0.36
$1.54
$0.36
$1.58
$0.36
$1.64
$0.36
The primary variables to review for changes under this scenario include:
• VMT
• Public transit unlinked passenger trips per day
• Person miles of nonmotorized travel by residents per day per capita
• Transportation and renter costs per year per household
• CO2 emissions from buildings and transportation
• Cumulative premature mortalities avoided due to walking and cycling relative to BAU
Sidewalk Building Scenario
Scenario Switch: Change the "decision to increase desired nonmotorized travel facilities per developed
acre " switch from 0 to 1.
Scenario(s) for Comparison: BAU
Scenario Summary
In the D-O LRP SD Model, the construction of sidewalks, bike lanes, and other facilities for
nonmotorized travel is determined by an assumption that the builders of such facilities attempt, on a
delay, to maintain at least a certain ratio of facility miles per developed acre and that that target ratio
remains constant. Currently, large portions of the study area do not have many nonmotorized travel
facilities. Therefore, this scenario tests what would happen if a policy were instituted to dramatically
increase the amount of nonmotorized travel facilities, potentially leading to more travel by nonmotorized
modes and less by other modes, which carries implications for people's health, among other outcomes.
Assumptions and Variable Changes
This scenario is consistent with the BAU scenario, with the following change:
71
-------
• In 2020, the variable "desired nonmotorized travel facilities per developed acre" is set to twice
the value that it has under the BAU scenario.
This variable change is summarized in Table 4-9.
Table 4-9. Values for "Desired Nonmotorized Travel Facilities per Developed Acre" in Tier 1 and Tier 2
under the Sidewalk Building and BAU Scenarios
SCENARIO NAME
DESIRED NONMOTORIZED TRAVEL FACILITIES
PER DEVELOPED ACRE
2015
2020-2040
TieM
Sidewalk Building
BAU
0.0385
0.0385
0.0770
0.0385
Tier 2
Sidewalk Building
BAU
0.00734
0.00734
0.01470
0.00734
The primary variables to review for changes under this scenario include:
• Person miles of nonmotorized travel by residents per day per capita
• VMT
• Public transit unlinked passenger trips per day
• Cumulative premature mortalities avoided due to walking and cycling relative to BAU
• Total impervious surface
• CO2 emissions from buildings and transportation
Higher LRT Effect on Public Transit Ridership
Scenario Switch: Increase the variable "policy test change in public transit person miles due to LRT"
from 1 to 1.1. For a larger effect, increase it to 2.
Scenario for Comparison: Light Rail + Redevelopment
Scenario Summary
We recognize that there is a great deal of uncertainty about projected light rail ridership in this model.
Instead of projecting person miles on light rail trains, we project the net effect of the introduction of
light rail on person miles of travel on all forms of public transit (including the rail line itself). In the
Light Rail + Redevelopment scenario, the light rail increases total public transit use by 16 million person
miles per year in 2026 and 45 million person miles per year in 2040 (Table 4-8). This increase in
ridership once the light rail is built is calculated using an equation whose inputs include employment,
population, and VMT, all of which are positively correlated with greater demand for fixed-guideway
transit service. In contrast, changes in vehicle revenue miles of bus and demand-response public transit
services affect person miles through an elasticity (i.e., increases in revenue miles lead to increases in
public transit person miles and decreases in travel by other modes). The Higher LRT Effect on Public
72
-------
Transit Ridership scenarios increase the effect of the light rail line on public transit person miles by a
constant factor. This represents an increase in light rail ridership without explicitly modeling ridership
on the rail line, itself. This change is implemented identically in Tiers 1 and 2, because the effect of the
light rail on public transit ridership in Tier 2 is assumed to come entirely from the effect in Tier 1.
Assumptions and Variable Changes
These scenarios are consistent with the Light Rail + Redevelopment scenario, with the following
change:
• Beginning with rail operation in 2026, the effect of light rail on public transit person miles per
year is increased by 10% or 100% (doubled) compared to its value in the Light Rail +
Redevelopment scenario.
The variable change for these two scenarios is summarized in Table 4-10.
Table 4-10. Values for "Change in Person Miles of Public Transit Travel Per Year Due to Adding Fixed
Guideway Transit" in Tier 1 and Tier 2 Under the Higher LRT Effect on Public Transit Ridership Scenarios
and the Light Rail + Redevelopment Scenario (Million miles per year)
SCENARIO NAME
CHANGE IN PERSON MILES OF PUBLIC TRANSIT TRAVEL
PER YEAR DUE TO ADDING FIXED GUIDEWAY TRANSIT
2025
2026
2030
2035
2040
Both Tiers
LRT Public Transit Effect +10pct
LRT Public Transit Effect +100pct
Light Rail + Redevelopment
0
0
0
17.9
32.5
16.2
25.0
45.6
22.8
35.4
64.6
32.2
49.3
90.2
44.8
The primary variables to review for changes under this scenario include:
• CO2 emissions from buildings and transportation
• Gross regional product
• VMT
• Congestion
Market Trends
Higher Gas Prices
Scenario Switch: Change the "gasoline price scenario" switch from 0 to 1.
Scenario(s) for Comparison: BAU
Scenario Summary
The D-O LRP SD Model relies on an exogenous projection from AEO2015 to estimate the price of
gasoline for the years 2015-2040, with the figures for the years 2000-2014 taken from historical data.
During the period where the model uses historical data, gasoline prices were at their highest in 2012,
73
-------
before dropping dramatically over the following three years. In the model, these exogenous price-of-
gasoline figures are a major driver of what mode of transportation people use, as well as the
environmental, economic, and health implications of each mode choice. In order to explore the
significance of the impact of gasoline prices on the model's outputs, we designed this scenario to test
what would happen if future gas prices turned out to be substantially higher than what is currently
projected, since energy-price trends are often changed by unforeseen events outside the control of local
policy-makers.
Assumptions and Variable Changes
This scenario is consistent with the BAU scenario, with the following change:
• In 2016, the value of the exogenous variable "price of gasoline" is set to be equal to the value it
had in 2012 (which is much higher than its values for 2013-2015). It then proceeds to increase at
the same year-over-year rate as in the BAU scenario. The price of gasoline is the same in both
Tiers, and is summarized in Table 4-11.
Table 4-11. Values for "Price of Gasoline" Under the High Gas Price and BAU Scenarios (USD 2010 per
gallon)
SCENARIO NAME
PRICE OF GASOLINE
2015 | 2020
2025
2030
2035
2040
Both Tiers
High Gas Price
BAU
$2.21
$2.21
$3.67
$2.62
$3.95
$2.83
$4.29
$3.06
$4.73
$3.38
$5.23
$3.74
The primary variables to review for changes under this scenario include:
• VMT
• Public transit unlinked passenger trips per day
• Person miles of nonmotorized travel by residents per day per capita
• Transportation and renter costs per year per household
• Gross regional product
• CO2 emissions from buildings and transportation
• Cumulative premature mortalities avoided due to walking and cycling relative to BAU
Retail Wage Increase Scenario
Scenario Switch: Change the "retail wage increase switch Tier 1" from 0 to 1.
Scenario(s) for Comparison: Light Rail + Redevelopment
Scenario Summary
The D-O LRP SD Model relies on an exogenous projection from Woods & Poole Economics, Inc. to
74
-------
estimate retail earnings per employee per year. Although the projections from Woods & Poole include
an increase in retail earnings per employee per year, they do not increase as much as the earnings per
employee in the other job categories (i.e., office, industrial, and service). In the model, retail earnings
per employee are one component used to determine the affordability of housing and transportation for
lower income residents, which is expected to decrease (i.e., become less affordable) in Tier 1 under the
Light Rail + Redevelopment scenario due to the increased property values and increased renter costs
around the station areas. This scenario therefore tests what would happen to affordability in Tier 1 if
future retail wages were increased above and beyond projections made by Woods & Poole.
Assumptions and Variable Changes
This scenario is consistent with the Light Rail + Redevelopment scenario, with the following change:
• Retail earnings per employee Tier 1 remain the same as the Light Rail + Redevelopment scenario
until 2016, when a $2.00 increase in the nominal hourly retail wage is added to the original
nominal retail wage. For the next three years (2017-2019), an additional $1.00 nominal raise is
added to the yearly increase already projected by Woods & Poole. For 2020-2040, the yearly
nominal wage increase remains the same as the Woods & Poole projection. These nominal retail
hourly wages are then converted to 2010 dollars and multiplied by 2000 hours/year to yield retail
earnings per employee per year. This change is summarized in Table 4-12.
Table 4-12. Values for "Nominal Hourly Retail Wage" and "Retail Earnings per Employee" Under the Retail
Wage Increase and Light Rail + Redevelopment Scenarios
SCENARIO NAME
NOMINAL HOURLY RETAIL WAGE
2015
2016
2017 | 2018 | 2019
2040
TieM
Retail Wage Increase
Light Rail + Redevelopment
$15.18
$15.18
$17.65
$15.65
$19.15
$16.15
$20.68
$16.68
$22.26
$17.26
$43.42
$38.42
RETAIL EARNINGS PER EMPLOYEE
TieM
Retail Wage Increase
Light Rail + Redevelopment
$27,800
$27,800
$31,600
$28,000
$33,400
$28,200
$35,100
$28,300
$36,700
$28,500
$34,700
$30,700
The primary variables to review for changes under this scenario include:
• Retail earnings Tier 1
• Resident per capita net retail earnings Tier 1
• Affordability index Tier 1
Higher Rent for Nonresidential Land Scenario
Scenario Switch: Change the "higher rent switch" from 0 to 1.
Scenario(s) for Comparison: BAU, Light Rail, and Light Rail + Redevelopment
Scenario Summary
"Gross operating surplus per sq ft" is a variable in the D-O LRP SD Model that represents the monetary
75
-------
profits gained by owning or renting commercial land through the sale of goods and services after all
expenses have been subtracted, including payment for employees and payments for owning or renting
the space. This variable is an exogenous lookup table with different values for each year, calibrated for
the BAU scenario to make up 40% of GRP between 2012 and 2040. For most scenarios, we leave the
values of this table unchanged, under the assumption that profits will not change in response to the
changes tested in each scenario. In this scenario, however, we test the possibility that nonresidential
property values (and thus property tax payments) will increase under the "Light Rail + Redevelopment"
scenario, consequently increasing commercial rent at a higher rate than the increase of consumption of
goods and services from businesses. This change would cause GOS per square foot to fall below the
levels used for the BAU scenario.
Assumptions and Variable Changes
This scenario is consistent with the Light Rail + Redevelopment scenario, with the following change:
• "Gross Operating Surplus per Sq Ft" is reduced by $5, relative to the values in the Light Rail +
Redevelopment scenario, for each year beginning in 2025. This change is illustrated in Table
4-13.
Table 4-13. Values for "Gross Operating Surplus persq. ft." Under the Higher Rent and Light Rail +
Redevelopment Scenarios
SCENARIO NAME
GROSS OPERATING SURPLUS PER SQ. FT.
2020
2025
2030
2035
2040
Tier 2
Higher Rent
Light Rail + Redevelopment
134
134
134
139
138
143
143
148
151
156
The primary variables to review for changes under this scenario include:
• Gross operating surplus
• GRP
• Total retail consumption
• Total employment
• Total nonresidential sq. ft.
Demographic Trends
More Multifamily Households Scenario
Scenario Switch: Change the "More Multifamily Households Tier 1" switch from 0 to 1.
Scenario(s) for Comparison: BAU
Scenario Summary
The D-O LRP SD model uses data from the American Community Survey (U.S. Census Bureau
76
-------
American Community Survey 2014) to divide total dwelling units between single-family and
multifamily from 2000 to 2014. Though the data for these years showed a shift toward more households
living in multifamily dwelling units, the BAU scenario keeps the distribution of households between
single-family and multifamily constant at levels in the ACS 2008-2012 5-yr estimate for the years 2015-
2040. In this scenario, we explore the impacts of extending the linear decline in the percent of dwelling
units that are single-family between 2000 and the ACS value (assumed to apply to 2014) out to 2040.
Extending this trend results in a 10.6 percent increase in the portion of households that live in
multifamily dwelling units by the year 2040.
Assumptions and Variable Changes
This scenario is consistent with the BAU scenario, with the following change:
• The "percent of people in SFDU table historical Tier 1" was changed from a constant value after
2014 to values reflecting the linear extension of the historical trend from 2000 through 2014.
This change is illustrated in Table 4-14.
Table 4-14. Values for "Percent of People in SFDU Table Historical Tier 1" Under the More Multifamily
Households and BAU Scenarios
SCENARIO NAME
PERCENT OF PEOPLE IN SFDU TABLE HISTORICAL TIER 1
2000
2010
2014
2020
2030
2040
TieM
More Multifamily Households
BAU
0.342
0.342
0.309
0.309
0.309
0.309
0.292
0.309
0.264
0.309
0.236
0.309
The primary variables to review for changes under this scenario include:
• MF and SF dwelling units Tier 1
• Multifamily and single-family acres Tier 1
• Total residential impervious surface Tier 1
• Energy use residential Tier 1
• Jobs housing balance Tier 1
Fewer Organically Affordable Units Scenario
Scenario Switch: On the gentrification Tier 1 view, reduce the "percent of MF dwelling units below 77
percent of median renter costs table tier 1 "from 0.26 in 2040 to 0.15.
Scenario(s) for Comparison: BAU
Scenario Summary
In the absence of accurate historical data and proj ections for the percent of organically affordable
multifamily units (i.e., the percent of market-rate multifamily units that cost no more than 30% of the
monthly income for a household making 60% of the Area Median Income (AMI)), the D-O LRP model
must make assumptions. First, the cost of organically affordable housing was calculated to be
77
-------
approximately 77% of the median renter costs in 2000 and from 2010-2014 in Tier 1. In the BAU
scenario, we therefore assume that values for this parameter in the years 2015-2040 remain constant at
the pre-2014 level. We recognize that this assumption is likely incorrect for Tier 1, particularly in the
face of redevelopment and changes to property values and renter costs expected to result from the light
rail. In particular, we expect that, in addition to total renter costs increasing, it is likely that the
distribution of renter costs around the median will change. This scenario therefore explores the impacts
of reducing the percent of multifamily units that are organically affordable between 2020 and 2040.
Assumptions and Variable Changes
This scenario is consistent with the BAU scenario, with the following change:
• The value for "percent of MF dwelling units below 77 percent of median renter costs table tier 1"
is reduced from 0.26 in 2020 to 0.15 by 2040. This change is illustrated in Table 4-15.
Table 4-15. Values for "Percent of MF Dwelling Units Below 77 Percent of Median Renter Costs Table Tier
1" Under the Fewer Organically Affordable Units and BAU Scenarios
SCENARIO NAME
PERCENT OF MF DWELLING UNITS BELOW 77 PERCENT OF MEDIAN
RENTER COSTS TABLE TIER 1
2010 |~~2020
2025
2030
2035
2040
TieM
Fewer Organically Affordable Units
BAU
0.26 0.26
0.26 0.26
0.23
0.26
0.21
0.26
0.18
0.26
0.15
0.26
The primary variables to review for changes under this scenario include:
• Organically affordable dwelling units tier 1
• Housing gap for households in poverty tier 1
• Potential population in poverty displaced tier 1
Environmental Policy Changes
Vehicle Emissions Reduced Scenario
Scenario Switch: In the Policy Switches box, beside Health, change "vehicle emissions reduced switch"
from 0 to 1.
Scenario for Comparison: BAU
Scenario Summary
The D-O LRP SD model treats vehicle emissions per VMT for both PIVh.s and NOX as exogenous inputs,
using data from a study that reports average vehicle emissions by vehicle model year (Cai et al. 2013)
for years up to 2020, and applying a linear extrapolation method to extend these average vehicle
emissions for 2020-2040. Though Cai et al. assume that emissions from PIVh.s and NOX remain constant
78
-------
at 2010 levels through 2020, recent EPA regulations for stricter fuel economy standards suggest further
reductions in PIVh.s and NOX emissions per VMT for vehicles in the future (US EPA 2015g). Although
we did not identify any study projecting future vehicle emissions based on these new standards, this
scenario explores the possible impacts of reducing vehicle emission rates of PIVh.s and NOX by 10
percent from 2020-2040, relative to the values projected by Cai et al.
Assumptions and Variable Changes
This scenario is consistent with the BAU scenario, with the following change:
• PM2.5 and NOX vehicle emissions per VMT (gram/mile) are decreased by 10 percent relative to
the BAU scenario for vehicle model years between 2020 and 2040. Note that this does not
translate directly into 10 percent reductions in total vehicle emissions, as the emission factors are
weighted by the fraction of vehicles in the U.S. vehicle fleet that are ages 1 to 17 (Jackson
200 la).
Table 4-16. Values for"PM2.5 Emissions per VMT" and "NOX Emissions per VMT" Under the Vehicle
Emissions Reduced and BAU Scenarios
SCENARIO NAME
PM2.5 EMISSIONS PER VMT
2020
2025
2030
2035 2040
Tier 2
Vehicle Emissions Reduced
.0070
.0067
.0063
.0059
.0055
BAU
.0070
.0069
.0067
.0064
.0062
NOx EMISSIONS PER VMT
Tier 2
Vehicle Emissions Reduced
.1623
.1167
.1075
.1006
.0947
BAU
.1623
.1205
.1151
.1104
.1053
The primary variables to review for changes under this scenario include:
• PM2.5 vehicle emissions tons and NOX vehicle emissions tons
• Premature mortalities avoided from PIVh.s and NOX vehicle emissions relative to BAU
Increased Solar Capacity
Scenario Switch: In the Policy Switches box, beside Energy and Water, increase "Desiredsolar
capacity"from 40MWto the appropriate level.
Scenario for Comparison: Light Rail + Redevelopment
Scenario Summary
Solar electricity generation is one strategy for reducing CO2 emissions, and solar capacity (MW) in Tier
2 has grown by an annual factor of 3.2 between 2005 and 2014, on average (North Carolina Sustainable
Energy Association 2015). In 2005, Tier 2 solar capacity was 2.5 kW, and by 2014 it had reached 22
MW. The D-O LRP SD model assumes electricity generated by solar is carbon neutral, and replaces
electricity from the conventional grid. In the BAU, Light Rail, and Light Rail + Redevelopment
scenarios, the model assumes that solar capacity increases at historical growth rate, but growth slows as
79
-------
it approaches a desired solar capacity of 40 MW, roughly double the current capacity. "Increased Solar
Capacity" scenarios have higher desired solar capacity and explore the timeframe in which solar
capacity would reach 10% of regional building electricity demand.
Assumptions and Variable Changes
These scenarios are consistent with the Light Rail + Redevelopment scenario, with the following
changes:
• Desired solar capacity is increased from 40 MW to 80, 320, and 640MW. These changes are
illustrated in Table 4-17.
Table 4-17. Values for "Desired Future Solar Capacity" Under the increased Solar Capacity and Light Rail
+ Redevelopment Scenarios
SCENARIO NAME
DESIRED FUTURE SOLAR CAPACITY
2000-2040
Tier 2
640MW solar LRT+redev
320MW solar LRT+redev
80MW solar LRT+redev
Light Rail + Redevelopment
640
320
80
40
The primary variables to review for changes under this scenario include:
• Solar capacity
• Solar energy production in kWh
• Solar fraction of building electricity use
• CO2 emissions from buildings and transportation
Clean Power Plan - Decreased Electricity Emissions Factor
Scenario Switch: In the Policy Switches box, beside Energy and Water, change "Policy switch
electricity emissions factor "from 0 to 1.
Scenario for Comparison: Light Rail + Redevelopment
Scenario Summary
This scenario reflects the Clean Power Plan for reducing carbon pollution from fossil fuel-fired power
plants, announced in August 2015 (US EPA 2015a, f). This scenario tests applies Clean Power Plan
emissions reduction goals specific to North Carolina (US EPA 2015c). These goals represent
technological changes in (1) fossil fuel-fired steam plants and (2) natural gas-fired combined cycle
plants.
Assumptions and Variable Changes
This scenario is consistent with the Light Rail + Redevelopment scenario, with the following change:
80
-------
• The variable "ton CC>2 per kWh" decreases linearly from 2022 to 2030, reaching 77% of its 2022
level by 2030. In contrast, the electricity emissions factor stays constant in the Light Rail +
Redevelopment scenario. After 2030, the electricity emissions factor is assumed to stay constant
in both scenarios, although future policy may reduce the emissions factor further between 2030
and 2040. Changes to the electricity emissions factor are illustrated in Table 4-18.
Table 4-18. Electricity Emissions Factors "Ton CO2 per kWh" and "Lb CO2 per MWh" Under the Clean
Power Plan and Light Rail + Redevelopment Scenarios
SCENARIO NAME
Both Tiers
Clean Power Plan
Light Rail + Redevelopment
Clean Power Plan
Light Rail + Redevelopment
TON CO2 PER KWH
2022
0.000778
0.000778
2025
0.000711
0.000778
2030
0.000599
0.000778
2040
0.000599
0.000778
LB CO2 PER MWH
1,560
1,560
1,420
1,560
1,200
1,560
1,200
1,560
The primary variables to review for changes under this scenario include:
• CO2 emissions from buildings and transportation
Stormwater Management Scenarios
Scenario Switch: In the Policy Switches box, beside Energy and Water, change "Percent N treated
onsite " to 0.3 or 0.4, and/or change "Percent reduction of existing N load" to 0.15
Scenario for Comparison: Light Rail + Redevelopment
Scenario Summary
Stormwater runoff is a significant and growing problem in the D-O LRP region. Both Jordan Lake and
Falls Lake, which receive Stormwater runoff from Tier 2, are considered in nonattainment of nutrient-
related water quality standards (NCDENR 2014, 2015). Local governments are currently subject to the
Jordan Lake and Falls Lake Rules, which require N and P reductions from new and existing
development; for simplicity, we focus on N reductions. According to the Falls Lake and Jordan Lake
rules, new developments in the Durham County area must plan to reduce their Stormwater N load to 2.2
Ib/acre/year, through a combination of onsite treatment and mitigation banking (purchased credits for
offsite watershed treatment) (NCDENR 2010, 2014, Woolfolk 2015). Durham county has the additional
requirement that new developments treat at least 30% of their Stormwater N load onsite, with the
remainder treated through mitigation banking (Woolfolk 2015). We modeled Stormwater management
scenarios to explore different targets for reducing Stormwater N load. These scenarios simulate growth
of Stormwater N load as impervious surface increases, as well as treatment of Stormwater N load from
development after 2015, possibly along with treatment of N load from development before 2015.
Reductions of 30% and 40% of the N load from new development are simulated, as well as 30% from
new development plus a 15% reduction in N load from existing development. For reference, the 2.2
Ib/acre/year target for Stormwater N from new development is also simulated.
81
-------
Assumptions and Variable Changes
This scenario is consistent with the Light Rail + Redevelopment scenario, with the following changes to
only Tier 1, described in Table 4-19:
Table 4-19. Percent of Stormwater N Load from Post-2015 Development Treated Onsite, and Percent from
pre-2015 Development Treated Onsite Under the Stormwater Management Scenarios
SCENARIO NAME
PERCENT OF STORMWATER N LOAD FROM POST-
2015 DEVELOPMENT TREATED ONSITE
2020
2040
TieM
30 pet new N load treated
40 pet new N load treated
30 pet new 15 pet existing N load treated
New N load treated to 2_2 lb_ac_y
Light Rail + Redevelopment
.3
.4
.3
-
0
.3
.4
.3
-
0
PERCENT FROM PRE-2015 DEVELOPMENT
TREATED ONSITE
TieM
30 pet new N load treated
40 pet new N load treated
30 pet new 15pct existing N load treated
New N load treated to 2_2 lb_ac_y
Light Rail + Redevelopment
0
0
.15
-
0
0
0
.15
-
0
The primary variables to review for changes under this scenario include:
• Total N load after treatment Tier 1
• Total N load after treatment to target Tier 1. Note that this variable is for comparison with, but
does not affect, "Total N load after treatment Tier 1"
82
-------
5 Model Results
5.1 Introduction
This chapter presents the results of the 20 scenarios that we ran in the D-O LRP SD Model, including
the three main scenarios (BAU, Light Rail, and Light Rail + Redevelopment) and 17 additional decision
support scenarios. These additional scenarios, described in detail in the previous chapter, are listed in
Figure 5-1 underneath the main scenario that they were each added onto. This chapter begins by briefly
discussing the results of the BAU scenario, including the model fit to historical data and projections
(when available) for both Tier 2 and Tier 1. It then provides an overview of the dynamic behaviors that
result from the model's inter-sector feedbacks under the Light Rail and Light Rail + Redevelopment
scenarios, collectively known as "the light rail scenarios." The last section of this chapter provides an in-
depth examination of the three main scenario results by sector, with summaries of the results of the 17
additional scenarios displayed in text boxes. Percent changes in variables under the BAU scenario are
typically presented between the most recent year of historical data (in most cases 2014) and 2040, while
percent changes under the light rail scenarios are presented from 2020-2040 to isolate the changes that
come with the introduction of the light rail line, as both scenarios are identical to the BAU before 2020.
Three Main Scenarios
BAU
Redevelopment
No Road Building
Sidewalk Building
Fare Free Transit
Higher Gas Prices
More Multifamily
Households
Fewer Organically
Affordable Units
Vehicle Emissions
Reduced
Light Rail
Additional Decision
Support Scenarios
Light Rail +
Redevelopment
Bold Redevelopment
Energy Efficiency
High Parking Price
Higher LRT Effect on
Public Transit Ridership
Retail Wage Increase
Higher Rent for
Nonresidential Land
Increased Solar Capacity
Clean Power Plan
Stormwater Management
Figure 5-1. Additional Decision Support Scenarios (Bullets) Listed Under the Corresponding Main
Scenario With Which They Were Run
5.2 Business as Usual (BAU) Scenario: Model Fit to Data and Projections
The BAU scenario was the basis on which the D-O LRP SD Model was built and for which extensive
validation tests were performed (described in more detail in Chapter 6 and Appendix B of this report).
For the BAU scenario, we recalibrated the D-O LRP SD Model each time a new connection or
adjustment was made in the model. Once all major connections between model sectors were in place, we
83
-------
validated the BAU scenario against historical data and projections. The aim of these validation tests was
to ensure that the model generated results consistent with historical data and projections from other
sources. Where projections from local data sources were available, the decision to calibrate the model to
either match or deviate from those projections (or to match one projection over another when multiple
projections were available) was made on a case-by-case basis and was dependent upon the assumptions
that went into the modeling that produced those projections. Where projections were not available, we
calibrated the BAU scenario so that historical trends were carried into the future. This validation process
supported the assertion that the model accurately represents important real-world dynamics in the
system that it seeks to represent. The main assumptions that went into the calibrations for the BAU
scenario are listed in Chapter 4, and results for key model variables that were calibrated are presented in
this chapter. We first present results for Tier 2 (DCHC MPO), followed by results for Tier 1 (combined
l/2 mile radii around proposed light rail station locations). For each tier, we discuss the behavior of
several indicators in the BAU scenario and note the degree to which the model's estimates match
historical data and projections from other sources. Graphs in this section present the model's estimates
(solid red lines) along with exogenous sources (dashed grey lines and dotted or dashed yellow lines, in
the case of multiple sources).
Tier 2
Under the BAU scenario, Tier 2 population increases by 53% between 2014 and 2040, reaching
660,000 in 2040 (shown in Figure 5-2). We calibrated this variable to align with population projections
for the DCHC MPO used in version 5 of the Triangle Regional Model (TRM v5), and the model's
results are within 5% of the TRM v5 Socioeconomic (SE) data projection in the year 2040. The driver
for the increase in population in the model, aside from births and deaths, is net migration, which
increases by 30% between 2014 and 2040. By 2040, approximately 820 people per year are projected to
move to Tier 2.
700
_o;600
§500
-0400
(0
3300
o
H200
2000
2010
2020
2030
2040
•BAU —— —TRM v5 SE Data U.S. Census Bureau
Figure 5-2. Population - Tier 2: BAU and Data, 2000-2040
As a result of the increase in population, total developed land in Tier 2 (shown in Figure 5-3) increases
by 54%, or 69,000 acres, between 2014 and 2040, with an additional 53 million nonresidential sq ft
built (shown in Figure 5-4). Figure 5-3 also shows two projections of developed land, both derived from
the CommunityViz2 (CV2) Parcel Geodatabase for Place Type & Development Status Editing; the
model's estimates of developed land very closely matches the lower of the two projections, with only a
1.9% deviation in 2040. Developed land in the model was not calibrated to these estimates; the fact that
they fit is a natural outgrowth of our assumptions and model projections for population, employment,
and initial developed land values.
84
-------
2000
2010 2020 2030
— BAD — — — Estimate 1 .... Estimate 2
2040
Figure 5-3. Developed Land - Tier 2: BAD and Data, 2000-2040
Figure 5-4 compares the model's estimate for total nonresidential square feet in Tier 2 to historical data
derived from the County Office of Tax Administration databases; as the figure shows, the model's
estimate deviates from the historical trend by no more than 8.5% during 2000-2014. In the D-O LRP SD
Model, the growth in nonresidential sq ft is driven by additional employment.
2010
2020
2030
2040
-BAU — — — County Offices of Tax Administration
Figure 5-4. Nonresidential Sq. Ft. - Tier 2: BAD and Data, 2000-2040
Total employment in Tier 2 (shown in Figure 5-5) increases by 55% between 2014 and 2040 under the
BAU scenario, adding 169,000 jobs during that time. We calibrated total employment to match TRM v5
SE data (2010) and projections (2011-2040), with historical employment growth rates from the U.S.
Bureau of Economic Analysis (BEA) from 2000-2010 applied to the 2010 employment value from the
TRM to get total employment for 2000-2009.
2000 2010 2020 2030
BAU — — — Data and Projections
2040
Figure 5-5. Total Employment - Tier 2: BAU and Data, 2000-2040
85
-------
The change in total employment in turn is driven by total retail consumption in the model, an
indicator of economic growth. Figure 5-6 shows the model's estimates of total retail consumption,
together with historical retail consumption data (2000-2014) from the NC Dept. of Revenue (NC DOR)
and a projected retail sales growth rate (2015-2040) from Woods & Poole Economics, Inc. We
developed the projections shown in Figure 5-6 for comparison with the BAU scenario output by
applying an inflation-adjusted retail sales growth rate of 2.5-2.8%/year from Woods & Poole
Economics, Inc. to the 2014 retail sales data from the NC DOR. As the figure shows, between 2000 and
2014, the historical retail consumption values were highly volatile from year to year. Though the model,
which targets a longer time frame for the analysis, does not replicate the short-term trend in the data, it
does capture the overall medium-to-long-term upward historical trend. The model's estimates of total
retail consumption for the BAU scenario demonstrate a growth rate of 2.6-3.1%, which is slightly
higher than the Woods & Poole Economics, Inc. growth rate that we used for the projections (2015-
2040). As a result, the model's estimate under the BAU scenario rises more quickly than projections, but
the modeled value in 2040 is still only 3% higher than the projected value.
o
o
2000 2010 2020 2030 2040
BAU — — — Data + Projections
Figure 5-6. Total Retail Consumption - Tier 2: BAU and Data, 2000-2040
Total retail consumption in the model is largely driven by an increase in Gross Regional Product
(GRP), shown in Figure 5-7. We calculated historical values (2000-2013) for this economic indicator in
Tier 2 based on a methodology from the BEA. In the model, GRP is the sum of two variables: total
earnings and gross operating surplus. Historically, total earnings composed between 59-64% of GRP in
Tier 2, with a 2013 share of 60%. For the BAU scenario, the model holds the total earnings share of
GRP constant at 60% and increases GRP at the same rate as total earnings from 2014 to 2040. The
model uses projections of earnings per job (by category) from Woods & Poole and projections of
employment from the TRM v5 SE data. Gross Operating Surplus, the second component of GRP, was
calibrated to make up 40% of GRP, in order to hold the total earnings share of GRP constant at 60%.
2000 2010 2020 2030
^^^—BAU — — — Data and Projections
2040
Figure 5-7. Gross Regional Product - Tier 2: BAU and Data, 2000-2040
86
-------
The increasing population in Tier 2 drives an increase in overall person miles of travel, including
nonmotorized travel, public transit travel, automobile driver travel, and automobile passenger travel. At
the same time, GRP per capita is projected to grow over time, which increases the growth rate of
automobile travel, relative to the growth rates of all other transportation modes. Various other variables
in the model, including traffic congestion, gasoline prices, intersection densities, and public transit fare
prices, affect the distribution of person miles of travel among the four modes without substantially
affecting total person miles of travel by all modes. The model estimates overall automobile vehicle miles
traveled (VMT) by combining its estimates of automobile driver travel related to trips that either begin
or end in the study area (affected by the process described above) with a separate projection of vehicle
travel that passes through the area without stopping (based on North Carolina population growth rates).
As shown in Figure 5-8, the model estimates that overall VMT increases by 55% during 2014-2040 in
the BAU scenario. This forecasted increase is in line with TRM v5 projections, both for the official
"preferred" scenario generated for the DCHC MPO 2040 Metropolitan Transportation Plan (2040 MTP,
which assumes a light rail line is built in the area, unlike our BAU scenario) and for the "Existing +
Committed" scenario (which assumes that no light rail is built but which still differs from the BAU
scenario in that it assumes there is no new road building after 2017).
2000 2010 2020 2030 2040
_ BAU —TRM v5 Result Data - —TRM v5 E+C Result Data
Figure 5-8. Vehicle Miles Traveled (VMT) Tier 2 BAU and Data, 2000-2040
The most significant feedback from increasing VMT is that increasing the ratio of VMT to roadway lane
miles (which is driven by an exogenous policy input in the model) increases traffic congestion,
mitigated by the assumption that during 2010-2040 traffic-management improvements will decrease the
amount of congestion that results from any given amount of peak-period VMT per roadway lane mile (in
accordance with projected changes in this ratio from the TRM's "preferred" scenario). Congestion in the
model is defined as the ratio of travel time in peak traffic to travel time under freeflow conditions, such
that a value of "1" indicates a congestion-free state. As shown in Figure 5-9, during 2014-2040,
congestion increases from 1.06 to 1.14, with a low point in 2015 and a high point in 2035. The non-
linear shape of this projected trend cannot be compared to other projections. The only other available
projection of future traffic congestion in the DCHC MPO is from the TRM v5's "preferred" scenario,
which only provides figures for 2010 and 2040. However, the D-O LRP SD Model's 2010 and 2040
congestion values in the BAU case are both within 1% of the TRM v5 figures. Changes in VMT are the
primary driver of traffic congestion. As such, congestion is partially driven by a balancing loop in the
model: When congestion increases, VMT goes down and the use of all other transportation modes goes
up.
87
-------
2000
2010
2020
2030
2040
-BAU ---TRMv5 Result Data
Figure 5-9. Congestion - Tier 2: BAU and Data, 2000-2040
Among the other variables affected by this balancing loop is person miles of public transit travel per
day, which is shown in Figure 5-10. This variable, which is also affected by population, GRP, and other
factors, as described above, increases by 40% between 2014 and 2040. During 2000-2013, this variable
fits historical data from the National Transit Database with an R-squared value of 0.81. In future years, it
deviates from both of the TRM v5's projections, but, as discussed above, the assumptions in both TRM
v5 projections differ from the assumptions in the D-O LRP SD Model's BAU scenario (the light rail is
built in TRM v5's "preferred" scenario, and road building does not continue after 2017 in TRM's
"Existing + Committed" scenario). As one would expect, the D-O LRP SD Model's BAU scenario
projects fewer public transit person miles than TRM v5's "preferred" scenario. Somewhat
counterintuitively, the D-O LRP SD Model's BAU scenario projects more future public transit person
miles than the TRM v5's "Existing + Committed" scenario, despite the fact that our model's BAU
scenario includes a higher road capacity than TRM v5's "Existing + Committed" scenario. This
difference is the result of the D-O LRP SD Model using more recent data from the National Transit
Database, which indicate an historical trend (2000-2013) of faster growth in public transit use than what
the TRM projects forward from 2010. Furthermore, the D-O LRP SD Model accounts for the positive
effects on public transit use of increases in public transit bus service that are anticipated by "The Bus
and Rail Investment Plan in Orange County" (September 2012) and "The Durham County Bus and Rail
Investment Plan (June 2011), plans published after the TRM v5's 2010 base year.
2000 2010
BAU
_ —TRM v5 E+C Result Data
2020
2030
2040
•TRM v5 Result Data
National Transit Database
Figure 5-10. Public Transit Person Miles - Tier 2: BAU and Data, 2000-2040
Energy and water use are also projected to increase as a result of increased population, land use, and
economic activity. As shown in Figure 5-11, building energy use is projected to increase by 33%
between 2014 and 2040, reaching 47 million MMBtu/year in 2040. Though building energy use is
driven by commercial activity (indicated by nonresidential sq ft) and population growth (indicated by
88
-------
dwelling units), the model projects a lower increase for building energy use than for those other
variables (projected to increase by 55% and 53%, respectively, over the same time period) because it
assumes that energy intensity will drop by 13-15% during that time (consistent with AEO projections, as
described in Chapter 4). Building energy use is calibrated to be within 2 percent of historical data from
the Durham City-County Sustainability Office.
2000
2010
2020
2030
2040
•BAU
— — — Durham Sustainability Office data scaled to Tier 2
Figure 5-11. Building Energy Use - Tier 2: BAU and Data, 2000-2040
COi emissions are projected to increase by 21% over the same period, reaching 11 million tons/year in
2040, as shown in Figure 5-12). Though CO2 emissions are driven largely by population, this projected
change is smaller than the 53% increase in population over this period due to the combined effects of
decreasing building energy intensity (noted above) and increasing passenger vehicle fuel efficiency.
Before accounting for the effects of congestion (which tends to reduce fuel efficiency), average fuel
efficiency in the study area is projected to increase from 16 MPG to 27 MPG between 2014 and 2040.
The D-O LRP's historical estimates of CO2 emissions are within 4% of historical data from the Durham
City-County Sustainability Office. Because our BAU scenario includes projected energy efficiency
improvements, our model projects slower growth in CO2 emissions than the BAU scenario of the
Durham GHG Plan (ICLEI2007). Between 2005 and 2030, the Durham GHG Plan projects a 48%
increase, but the D-O LRP model projects a 28% increase. If energy efficiency improvements were
excluded from our model (as detailed in Chapter 6), CO2 emissions would increase by 47% between
2005 and 2030, similar to the Durham GHG Plan projections.
12
£ 6
o
c 4
o
I2
0
2000 2010 2020 2030 2040
^—BAU
— — — Durham Sustainability Office data scaled to Tier 2
Figure 5-12. CO2 Emissions - Tier 2: BAU and Data, 2000-2040
89
-------
Besides increasing energy use, land development causes a 35% increase in impervious surface during
2014-2040, reaching 65,000 total impervious acres (33% of developed land) by 2040, as shown in
Figure 5-13. The growth of impervious surface causes increased stormwater runoff, which is discussed
later in this chapter. Impervious surface is calibrated to within 2% of the one-meter land cover data from
EPA EnviroAtlas for 2010 in Tier 2.
2010
^—BAU
2020 2030
-EPA EnviroAtlas
Figure 5-13. Impervious Surface - Tier 2: BAD and Data, 2000-2040
Tier"!
In Tier 1, population is projected to increase by 18% during 2014-2040 (compared to 53% in Tier 2),
reaching 50,000 in 2040, as shown in Figure 5-14. In Tier 1, we chose to calibrate the model to the
historical population trend from the U.S. Census Bureau, rather than the TRM v5 SE data population
projection, because the latter source was based off of a "preferred growth" land use scenario with
assumptions that differed from our BAU scenario for Tier I.21 One cause of the population growth is net
migration, which increases by 164% within Tier 1 over this time period. About 310 people per year are
projected to move into Tier 1 by 2040.
30
•BAU TRMV55E Data
U.S. Census Bureau
Figure 5-14. Population Tier -1: BAU and Data, 2000-2040
21 Specifically, the TRMv5 SE "preferred growth" land use scenario assumed higher density and a higher percentage of
multi-family dwelling units in the traffic analysis zones (TAZs) in Tier 1 than the D-O LRT SD model assumes in the BAU
scenario. Note that these assumptions are similar to the assumptions that we use in the D-O LRT SD model's Light Rail
scenario; as a result, the population projections in that scenario more closely match the TRM population projections. In Tier
2, the TRMv5 SE data population projection was essentially identical to the U.S. Census Bureau data, and therefore no
choice had to be made regarding to which to calibrate..
90
-------
Developed land in Tier 1 is projected to increase by 31% during 2014-2040 (compared to 54% in Tier
2), as shown in Figure 5-15, or about 1,200 acres, with an additional 7.8 million nonresidential building
sq ft. The model's projection for developed land closely fits the lower of two estimates derived from the
CV2 Parcel Geodatabase for Place Type & Development Status Editing (TJCOG 2014a), with a value
within 4% of that source in 2040. The model's projection for total nonresidential square feet, shown in
Figure 5-16, deviates by no more than 2% in Tier 1 from the estimate derived from the three County
Offices of Tax Administration databases (Durham County Tax Administration 2000-2014, Orange
County Tax Administration 2014, Chatham County Tax Administration Office 2014).
7
6
5
4
3
2
1
0
20
«•— — ^— —
,!••"•
••'
• ••I*"'""
00 2010 2020 2030 20
BAU — — — Estimate 1 • Estimate 2
Figure 5-15. Developed Land - Tier 1: BAU and Data, 2000-2040
2000 2010 2020 2030 2040
^^^—BAU — ——County Offices of Tax Administration
Figure 5-16. Nonresidential sq ft - Tier 1: BAU and Data, 2000-2040
Under the BAU scenario, total employment in Tier 1, shown in Figure 5-17, increases by 47% between
2014 and 2040 (compared to 65% in Tier 2). Although we did not use the TRM v5 SE data to calibrate
our population projections in Tier 1, we used TRM v5 SE data to calibrate total employment projections,
since we did not identify any other employment projections by employment category at the appropriate
geographic scale for Tier 1. To calibrate historical employment in Tier 1, we used employment growth
rates (2000-2010) from the U.S. Census Longitudinal Employment-Household Dynamics (LEHD).
The combination of increasing employment and nonresidential sq ft in Tier 1 causes a 115% increase in
GRP between 2010 and 2014 in Tier 1 (shown in Figure 5-18), reaching $14 billion (USD 2010) in
2040. However, with the slow increase in population in Tier 1 under the BAU scenario between 2014
and 2040, the model assumes that the majority of available new jobs are filled by people who reside
outside of Tier 1 (though we note that this dynamic changes in the Light Rail scenario). As in Tier 2, the
model calculates GRP for Tier 1 as the sum of total earnings and gross operating surplus; the average
growth rate for GRP between 2014 and 2040 (which is set to be the same as the growth rate for total
earnings) is 2.74% per year. In Tier 1, the total earnings share of GRP is higher than in Tier 2 (70%, as
opposed to 60%), since this Tier has a higher percentage of service jobs, which generate less gross
operating surplus than industrial, office, and retail jobs.
91
-------
2000
2010 2020 2030
^—BAU — — — Data and Projections
2040
Figure 5-17. Total Employment - Tier 1: BAD and Data, 2000-2040
2000
2010
2020
2030
2040
• BAU — — — Data and Projections
Figure 5-18. Gross Regional Product - Tier 1: BAU and Data, 2000-2040
We calibrated total retail consumption in Tier 1 to retail sales data estimates from 2008-2014
downloaded from SimplyMap. As shown in Figure 5-19, this variable increases by 114% between 2014
and 2040 under the BAU scenario, driven by the increase in GRP.
3 0
2000
2010 2020 2030
BAU — ——SimplyMap
2040
Figure 5-19. Total Retail Consumption - Tier 1: BAU and Data, 2000-2040
In the transportation sector, we used the same variable calculations and calibration sources in Tier 1 as
in Tier 2. In some cases, such as VMT and congestion, the source of calibration data is a GIS shapefile
that we clipped to each of the two Tiers. In other cases, calibration data for many Tier 1 variables could
only be estimated by scaling down Tier 2 data. In some of these cases, such as person miles of travel by
individual modes, we were only able to use values from 2010 for calibration purposes, since the scaling
process involved deriving ratios from historical Tier-1-scale and Tier-2-scale data for which Tier-1-scale
projections were not available. As shown in Figure 5-20, during 2014-2040, VMT in Tier 1 increases by
37% (compared to 55% in Tier 2), reaching 1.8 million miles per day in 2040. This represents a greater
92
-------
deviation (-8.0%) from the 2040 MTP's "Preferred" infrastructure case, as projected by the TRM v5,
than our estimate of VMT in Tier 2. This deviation is likely caused by the fact that the 2040 MTP's
"Preferred" infrastructure case involves the light rail being built (which is expected to lead to an increase
in population and VMT), whereas the D-O LRP SD Model's BAU scenario does not.
>,
-o ^-u
B-15
ffl
E 1 n
°
5 °'5
n n
-^*C*=
^_j^_"" - -
2000 2010 2020
BAU — — — TRM v5 Result Data •
2030 2040
TRM v5 E+C Result Data
Figure 5-20. VMT - Tier 1: BAU and Data, 2000-2040
The 37% increase in VMT estimated by the D-O LRP SD Model for Tier 1 (which is significantly less
than the increase in Tier 2 VMT) leads to an increase in traffic congestion of only 0.42% between 2014
and 2040 (compared to an 8.0% increase in Tier 2), as shown in Figure 5-21. In large part, the reason
that congestion increases so little relative to the amount that VMT increases is because the D-O LRP SD
Model assumes (in both Tiers) that traffic management systems will get better over time, decreasing the
amount of congestion that is produced by a given level of peak-period VMT per lane mile. This is in
addition to congestion being alleviated by the construction of new roadway lane miles over time. As in
Tier 2, Tier 1 congestion in the D-O LRP SD Model can only be compared to other data/projections for
the years 2010 and 2040 from the TRM v5's "preferred" scenario, which assumes that a light rail line is
built through Tier 1, whereas this model's BAU scenario does not. As mentioned above, we assume that
a light rail line would increase the number of people and jobs in Tier 1, meaning that it also increases
Tier 1 VMT and traffic congestion. For this reason, 2040 Tier 1 congestion is 6.3% less than the TRM
v5 projection used for calibration.
1.25
1.20
1.15
1.10
i
1.05
1.00
2000
2010 2020 2030 2040
__ BAU - - -TRM v5 Result Data
Figure 5-21. Congestion - Tier 1: BAU and Data, 2000-2040
As shown in Figure 5-22, due in part to the small growth in traffic congestion, as well as the projected
increase in GRP per capita (among other factors), the model projects that Tier 1 person miles of public
transit travel per day actually declines by 7.2% during 2014-2040, compared to a 40% increase in Tier
2. Because Tier 1 employment grows faster than population in the BAU scenario, even though overall
person miles of travel (encompassing all modes) increase faster than population between 2014 and 2040
93
-------
(22% growth for total person miles of travel vs. 18% growth for population), the number of person miles
by residents increases much more slowly (2.6% growth in the same period). As a result, person miles of
travel by residents per capita actually decrease by 13% in Tier 1 over this period (partially because
projected increases in densities of jobs and retail reduce the distances that people must travel to get to
work or run errands), compared to an increase of 1.9% in Tier 2.
•3C
^n
1 R
n
1
-S
^^
r —
—
•— ^— ™^_
2000
2010
•BAU
2020
2030
2040
-TRMvS Result Data
Figure 5-22. Public Transit Person Miles - Tier 1: BAU and Data, 2000-2040
The model's estimates of increased energy and water impacts in Tier 1 reflect the same increases in
population, land use, and economic activity as in Tier 2. Due to increased building square footage,
building energy use is projected to increase 22% between 2014 and 2040, reaching 6.1 million
MMBtu/year by 2040. Since there were no historical data for Tier 1 energy use or CO2 emissions, we
used the energy and CO2 emissions calibration factors from Tier 2 for Tier 1 as well. We assumed
technological improvement would cause building energy use intensity to decrease by 13-15% between
2014 and 2040, as in Tier 2. COi emissions are projected to increase by 12% during this time, reaching
1.3 million tons/year in 2040. As in Tier 2, the increase in CO2 emissions follows an increase in VMT
and building sq ft. With land development, impervious surface in Tier 1 is projected to increase by
27% in the next 25 years (compared to a 35% increase in Tier 2), reaching 3,900 total impervious acres
(79% of developed land) by 2040, as shown in Figure 5-23. Growth in impervious surface drives an
increase in stormwater runoff, which is discussed in the Scenario Results for the Water sector. As with
Tier 2, we calibrated the model's estimates of impervious surface in Tier 1 to one-meter land cover data
from EPA EnviroAtlas for 2010; for Tier 1, the model's estimate of impervious surface in 2010 is within
1% of the EnviroAtlas data.
See Table 6-6 and Table 6-7 in Chapter Six on Quality Assurance for a summary of the Tier 1 and Tier 2
indicators discussed in this section, among others. For each indicator, the table notes any external data
sources used for calibration purposes - for historical estimates, future projections, or both - and
provides the R-squared fit and average percent deviation between the model's estimates and the external
data sources.
94
-------
40
35
30
20
10
2000
2010
2020
2030
2040
•BAU
-EPA EnviroAtlas
Figure 5-23. Impervious Surface - Tier 1: BAU and Data, 2000-2040
BAU Results Summary
As discussed above, the changes projected to occur in the BAU scenario present several opportunities
and challenges to the region. With expected growth in population and economic activity, developed land
and the attendant impervious surfaces and runoff are expected to increase significantly over the next 25
years. In addition, in both Tier 1 and Tier 2, public transit use and walking and cycling rates are
projected to decline relative to automobile travel, in the absence of any major investment in improving
public transit. Though population will continue to grow rapidly in Tier 2, population growth in Tier 1
will not match that expected by the TRM v5 SE Data without greater attractions to the area, such as
employment and retail growth. Population growth will also increase energy use and CO2 emissions,
though these increases are mitigated somewhat by increased energy efficiency of buildings and vehicles.
The Light Rail and Light Rail + Redevelopment scenarios provide opportunities to explore how two
possible approaches for managing growth in the region might affect these outcomes. The remainder of
this chapter focuses on those two scenarios, beginning with an overview of inter-sectoral model
behaviors, followed by a sector-by-sector discussion of scenario results.
5.3 Light Rail Scenarios: Model Behaviors
The model's fidelity to historical data and projections, as discussed above, suggests that the results of
the other two main scenarios in the model - the Light Rail and Light Rail + Redevelopment scenarios -
would be indicative of the effects of those alternatives on the modeled system.
As noted in in Chapter 3 (in the "Overview of Model Structure" section), the inter-sector feedback loops
that connect the three core sectors (Land Use, Economy, and Transportation) have the largest impacts on
the modeled system. Consequently, any scenario that introduces a change to one of the variables in these
main inter-sector loops will have cascading impacts on variables throughout the model. This section
illustrates this dynamic by presenting an overview of model results for the two main scenarios: (1) the
Light Rail scenario and (2) the Light Rail + Redevelopment Scenario.
In the Light Rail scenario (as described in Chapter 4), 17 miles of light rail are added to the public
transit system over the six-year period between 2020 and 2026. The additional public transit person
miles of travel resulting from the light rail are determined by a formula that takes into account
population, retail and entertainment jobs, high wage jobs, and overall employment in Tier 1, and VMT
per highway lane mile in Tier 2 (Chatman et al. 2014). In addition, demand for commercial square
footage in the station areas is assumed to increase by 10% (though this variable can be modified by users
in the user interface of the model). In the Light Rail + Redevelopment scenario, on top of these changes,
95
-------
20% of developed land in Tier 1 is gradually redeveloped to a density almost triple the density in the
BAU scenario. The discussion below highlights several of the results of these scenarios by exploring
how the changes described above affect the main inter-sector feedback loops described in Chapter 3.
Light Rail scenario:
17 miles of light rail transit added
between 2020 and 2026
10% more demand for retail,
office, and service sq ft in Tier 1
Economy -> Land Use -> Economy
In this reinforcing feedback loop, employment growth
leads to growth in nonresidential square feet, which
increases gross regional product (GRP), which eventually
increases total employment. In both scenarios,
nonresidential sq ft in Tier 1 increases relative to the BAU
scenario, though this increase is larger in the Light Rail +
Redevelopment scenario. In the Light Rail scenario, total
nonresidential sq ft in Tier 1 increases by 16% and 8% relative to the BAU scenario in 2033 and 2040,
respectively. Figure 5-24 shows that growth in nonresidential square feet in this scenario levels off
around 2033, which is when the maximum allowed expansion of developed land in Tier 1 is reached. In
the Light Rail + Redevelopment scenario, the increase in density allows for greater development of
nonresidential sq ft per acre of developed land, so the maximum allowed expansion of developed land in
Tier 1 is not reached during the model's timeframe. As a result, total nonresidential sq ft in Tier 1
follows a similar path to the Light Rail scenario through 2033, but continues increasing relative to the
BAU scenario, reaching a level 34% higher than the BAU value by 2040. As shown in Figure 5-24, in
2040, the Light Rail + Redevelopment scenario projects 34
million nonresidential sq ft, compared to 25 million sq ft in
the BAU scenario and 27 million sq ft in the Light Rail
scenario. In Tier 2, the increase in nonresidential square
feet is limited to the amount added in Tier 1 as a result of
development of the Light Rail, peaking at 6.5% higher than
BAU in 2040 for both the Light Rail and Light Rail +
Redevelopment scenarios (Figure 5-25).22
Light Rail + Redevelopment scenario:
• 20% of developed land in Tier 1
redeveloped to densities 193%
more dense than originally
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
• BAU
Figure 5-24. Total Nonresidential Sq Ft - Tier 1: Main Policy Scenarios Compared to BAU
22 Note that the impacts of redevelopment on land use in Tier 1 are not linked to Tier 2. This means that while the Light Rail
+ Redevelopment scenario results in an increase in nonresidential sq ft even as land development declines in Tier 1, in Tier 2,
both land development and square footage remain very similar to the Light Rail scenario. This can be understood as a
concentration of square footage - during redevelopment, some businesses decide to move from Tier 2 to Tier 1.
96
-------
2000
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
•BALI
Figure 5-25. Total Nonresidential Sq Ft - Tier 2: Main Policy Scenarios Compared to BAD
As described above, an increase in nonresidential sq ft leads to an increase in employment and GRP, so
all three indicators increase under both the Light Rail and the Light Rail + Redevelopment scenarios. As
shown in Figure 5-26. Total Employment - Tier 1: Main Policy Scenarios Compared to BAU, in the
Light Rail scenario, employment in Tier 1 is about 5.4% higher and 14% higher than BAU in 2030 and
3040, respectively, while in the Light Rail + Redevelopment scenario, employment in Tier 1 is 6.9%
higher and 23% higher than BAU in 2030 and 2040, respectively. GRP follows a similar trend. Relative
to the Light Rail scenario, the increase in density in the Light Rail + Redevelopment scenario allows for
greater economic growth in Tier 1, resulting in 15,000 more jobs created in 2040 (150,000 vs. 135,000).
As shown in Figure 5-27 employment is about 6% higher than BAU in Tier 2 in 2040 under both the
Light Rail and Light Rail + Redevelopment scenarios (and GRP is about 6.5% higher).
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
• BAU
Figure 5-26. Total Employment - Tier 1: Main Policy Scenarios Compared to BAU
2000
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
•BAU
Figure 5-27. Total Employment - Tier 2: Main Policy Scenarios Compared to BAU
97
-------
Economy -> Population -> Economy
This pair of balancing and reinforcing loops link employment and population in Tier 1; more
employment drives more migration to the area, increasing the population and therefore the labor force.
This has two counterbalancing impacts. First, a higher labor force increases the potential employment in
the region, so with sufficient demand for employment, an increased labor force increases employment
and creates a reinforcing loop. Second, if demand for labor is not sufficient to employ all new additions
to the labor force, an increasing labor force increases unemployment, which reduces net migration,
creating a balancing loop. In the Light Rail and Light Rail + Redevelopment scenarios, the net result of
these two feedback loops is a large increase in population for two reasons: (1) the increase in GRP
described above leads to an increase in demand for employment, which directly leads to an increase in
net migration to Tier 1; and (2) a decrease in the unemployment rate, with the additional resident
population having a higher rate of employment, independently also increases net migration.
Consequently, population in Tier 1 is about 7.3% higher and 22% higher than BAU under the Light Rail
scenario in 2030 and 2040, respectively, and 8.3% higher and 29% higher in Light Rail +
Redevelopment in 2030 and 2040 (Figure 5-28). By 2040, the population of Tier 1 under the Light Rail
+ Redevelopment scenario reaches 64,910, closely approaching the TRM v5 SE Data projection of
66,980 for that year (see Figure 5-28). In Tier 2, the majority of the change in population comes from
the change in Tier 1, leading to increases of 2.8% and 3.7% over BAU by 2040 under the Light Rail and
the Light Rail + Redevelopment scenarios, respectively (Figure 5-29). The light rail is the primary
attractor for additional population growth, so the model assumes that only 150% of the increase in net
migration in Tier 1 is applied to Tier 2 (meaning all of the migration into Tier 1, plus an additional 50%
of that migration to areas of Tier 2 outside of Tier 1), despite it covering a much larger area.
In general, percent changes in model variables relative to the BAU scenario are much smaller in Tier 2
than Tier 1. This is due to the fact that the changes implemented in these two scenarios are centered in
Tier 1, which only represents a fraction of the overall population, economy, and traffic of Tier 2.
Therefore, the effects of the light rail scenarios are proportionally much smaller in the context of Tier 2
than in the context of Tier 1.
u 60
Q.
o Rn
Q.
| 40
| 30
£ 20
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
• BAU
Figure 5-28. Population - Tier 1: Main Policy Scenarios Compared to BAU
98
-------
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
• BAU
Figure 5-29. Population - Tier 2: Main Policy Scenarios Compared to BAU
Economy -> Transportation -> Economy
In this balancing feedback loop, an increase in GRP per capita causes an increase in person miles of
automobile travel and VMT, which increases congestion, which decreases economic productivity,
resulting in a decrease in GRP per capita. Even though the addition of the light rail line increases transit
mode share and ridership, which serves to mitigate the effect of a rising GRP on VMT and congestion,
the net effect of the shift to more transit use and the feedback loop described above is that increases in
GRP cause increases in VMT and congestion in the Light Rail and Light Rail + Redevelopment
scenarios. VMT in Tier 1 is about 1.5% higher and 6.3% higher than BAU under the Light Rail scenario
in 2030 and 2040, respectively, and 1.7% higher and 9.4% higher in Light Rail + Redevelopment in
2030 and 2040, respectively (Figure 5-30). As discussed above, the policy interventions that distinguish
the Light Rail and Light Rail + Redevelopment scenarios from the BAU scenario are centered in Tier 1,
which is just one part of Tier 2, causing Tier 2 effects of the light rail scenarios to be less pronounced
than the corresponding Tier 1 effects. Therefore, in Tier 2, the increase in VMT relative to BAU is
smaller in both light rail scenarios than in Tier 1, reaching 1.6% higher than BAU in 2040 under the
Light Rail scenario and 2% higher under the Light Rail + Redevelopment scenario (Figure 5-31).
2000 2010 2020 2030 2040
Light Rail + Redev — —Light Rail BAU
Figure 5-30. VMT - Tier 1: Main Policy Scenarios Compared to BAU
99
-------
2000 2010 2020 2030
Light Rail + Redev — —Light Rail
2040
•BALI
Figure 5-31. VMT, Tier 2: Main Policy Scenarios Compared to BAD
In the Light Rail scenario, Tier 1 congestion increases by 6.2% relative to BAU by 2040 (slightly less
than the corresponding increase in VMT), and in the Light Rail + Redevelopment scenario, it increases
by 12% relative to BAU (Figure 5-32) (greater than the corresponding increase in VMT). This higher
growth in congestion under the Light Rail + Redevelopment scenario is due both to the higher rates of
growth in GRP per capita under this scenario and to land being redeveloped with higher Floor Area
Ratios (FARs), which also drive traffic congestion in Tier 1, as high FARs are correlated with a greater
density of activity, and hence more concentrated traffic. As with VMT, increases in congestion in the
light rail scenarios are smaller in Tier 2 than in Tier 1, with only 1.6% and 2% increases over BAU in
2040 in the Light Rail and Light Rail + Redevelopment scenarios, respectively (Figure 5-33).
As shown in Figure 5-32, Tier 1 congestion sharply declines in 2026 due to the introduction of the light
rail line and the consequent increase in public transit usage. But the economic and population growth
that also result from the light rail line lead to an increase in traffic that reverses this decline starting the
following year. This nonlinear response stems from the two parallel causal chains driving the output's
value in opposite directions, where there is a delay in the causal chain with the stronger effect (economic
and population growth) but not in the one with the weaker effect (light rail's effect on transit usage). As
another product of nonlinearities in the model, congestion starts to decline after 2037 in the Light Rail
scenario but not in the Light Rail + Redevelopment scenario. One reason for this is the aforementioned
difference in FARs between the scenarios. Another reason is that there is a finite amount of land in Tier
1 that can be developed. Whereas greater-than-BAU economic growth causes this limit to be reached in
the Light Rail scenario, the higher development densities under the Light Rail + Redevelopment
scenario cause developable Tier 1 land to not be exhausted before 2040. Since limiting developed land
also limits nonresidential floor space, and therefore employment, GRP growth in the Light Rail scenario
slows down in the latter years of the model run, hence also slowing down VMT growth. Even though
Tier 1 FARs and VMT do not decline in the latter years of the model run in the Light Rail scenario, the
average FAR stops increasing and VMT increases little enough that the continued construction of
roadway lane miles and the improvements we assume will occur in traffic-control measures over time
are able to produce a net decline in congestion after 2037 that is not mirrored in the Light Rail +
Redevelopment scenario.
100
-------
20
S~^
X
00 20
Light Ra
V. >*
^•^
10 20
il + Redev •
»— N, --^
•^-*V£.
20 20
- —Light Rail
^^
<***^ "" ™" •
^
30 20
^— BAU
Figure 5-32. Congestion - Tier 1: Main Policy Scenarios Compared to BAD
/~
?
00 20
Light Ra
.X
-^
10 20
I + Redev •
_^— — •
20 20
- —Light Rail
*r- ..
30 20
BALI
Figure 5-33. Congestion - Tier 2: Main Policy Scenarios Compared to BAD
Economy -> Equity -> Transportation -> Economy
This balancing feedback loop links economic growth and employment to equity and underserved
populations, who are more likely to be transit dependent. Economic growth decreases the percent of
people who are transit dependent, which increases VMT (at the expense of travel by other modes),
which increases congestion and decreases GRP. As employment increases due to the Light Rail
scenarios, unemployment declines, leading to a decline in the percent of the population in poverty in
both Tiers. In the Light Rail scenario, the percent of the population in poverty in Tier 1 is about 6.4%
lower and 7.5% lower than BAU in 2030 and 2040, respectively; in the Light Rail + Redevelopment
scenario, these values are and 7.7% lower and 17% lower, respectively (Figure 5-34). The poverty rate
rises near the end of the Light Rail simulation ultimately in response to the cap on developable land
being reached. This stops expansion of nonresidential sq ft, which slows employment growth, leading to
higher unemployment and poverty, demonstrating how systemic limits produce nonlinear results in the
model. In Tier 2, the percent of the population in poverty under the light rail scenarios drops more
relative to the BAU (21%) than in Tier 1, because the rate starts at a lower level overall (Figure 5-35).
101
-------
C0.35
8 0.3
8.0.2
§0.15
-------
20
3 18!
-g 16?
I 145>
o 10?
a 87
•a 0/
o 6?1
K 4%
o 2%
1 0%
2010
2020
• Light Rail + Redevelopment
2030
• Light Rail
2040
•BAU
Figure 5-36. Percent of Households with Zero Cars - Tier 1: Main Policy Scenarios Compared to BAU
1056
y, 2%
2000
2010
2020
2030
•Light Rail + Redevelopment — — Light Rail
2040
•BAU
Figure 5-37. Percent of Households with Zero Cars - Tier 2: Main Policy Scenarios Compared to BAU
Economy -> Land Use -> Energy -> Economy
In this feedback loop, as GRP and nonresidential sq ft increase due to the development of the light rail,
total energy spending increases. At the same time, residential energy spending increases as population
and dwelling units increase. As shown in Figure 5-38, total energy spending in Tier 1 is about 9.4%
higher and 10% higher than BAU under the Light Rail scenario in 2030 and 2040, respectively, and 11%
higher and 26% higher under Light Rail + Redevelopment in 2030 and 2040, respectively. As described
in Chapter 2, if energy spending grows fasters than GRP, gross operating surplus is negatively affected.
This is not the case in either Tier in the two light rail scenarios: In Tier 1, GRP grows at an annual rate
of 3.2% and 4.0% in Light Rail and Light Rail + Redevelopment, respectively between 2026 and 2040.
During this same time period, energy spending grows at the slower rates of 1.8% and 2.7% for the two
respective scenarios. In Tier 2, the growth in total energy spending is also surpassed by the growth in
GRP, with about 3.1% annual GRP growth between 2026 and 2040 in both light rail scenarios,
compared to 2.1% annual growth in total energy spending (Figure 5-39).
Along with energy spending, the light rail scenarios cause an increase in energy consumption and CO2
emissions in both Tiers. Compared to BAU, Tier 1 energy consumption is 10% higher in the Light Rail
scenario in both 2030 and 2040; and 12% and 28% higher in those respective years in the Light Rail +
Redevelopment scenario. Tier 1 CO2 emissions are 12% and 11% higher than BAU under the Light Rail
scenario in 2030 and 2040, respectively, and 14% and 31% higher than BAU under the Light Rail +
Redevelopment scenario. In Tier 2, the light rail causes a lower percentage increase in energy
consumption and CO2 emissions than in Tier 1. In both light rail scenarios, energy consumption is about
2% higher in 2030 than in the BAU scenario. By 2040, energy consumption is about 4% higher than
103
-------
BAU in the Light Rail scenario and about 5% higher than BAU in the Light Rail + Redevelopment
scenario. CO2 emissions are about 2% higher than BAU in 2030 and 5% higher in 2040 in both light
rail scenarios.
2010
•Light Rail + Redev
2020
2030
2040
• Light Rail
•BAU
Figure 5-38. Total Energy Spending - Tier 1: Main Policy Scenarios Compared to BAU
2500
2000
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
• BAU
Figure 5-39. Total Energy Spending - Tier 2: Main Policy Scenarios Compared to BAU
Light Rail Scenarios Results Summary
The previous section outlined several opportunities and challenges expected to occur by 2040 in the
BAU scenario, and this section highlights several areas in which the light rail and redevelopment might
respond to these changes. The largest impacts of the light rail scenarios are seen in Tier 1, since this
where the light rail is located and has the largest effect. Under the BAU scenario, population in Tier 1
falls short of the TRM v5 SE Data projection by 25% in 2040. The growth stimulus provided by the
Light Rail scenario goes a long way towards matching this projection, leading to a level 8.5% lower in
2040, while the TRM v5 SE Data projection for population in Tier 1 is essentially matched under the
Light Rail + Redevelopment scenario (with a 3% difference in 2040). Total employment rises as well,
reaching 23% higher than BAU under the Light Rail + Redevelopment scenario by 2040. Along with
this boom in population and economic growth however, comes increased land development, impervious
surfaces, nitrogen loadings due to stormwater runoff, energy use, and emissions. In 2040, CO2
emissions per year in Tier 1 are 11% and 31% higher than the BAU scenario in the Light Rail and Light
Rail + Redevelopment scenarios, respectively. However, the densification and centralization of
development that occurs under the Light Rail + Redevelopment scenario offers an opportunity to
mitigate some of these effects. On a per capita basis, impervious surfaces, nitrogen loadings, water
demand, energy use, and emissions all decline. Furthermore, walking and bicycling rates in Tier 1 are
about 15% higher than BAU in both light rail scenarios, reversing the declining trend shown in the BAU
104
-------
scenario. Nonetheless, congestion in Tier 1 remains 12% higher than BAU in 2040 under the Light Rail
+ Redevelopment scenario, and 6.2% higher than BAU under the Light Rail scenario.
Section 5.4 presents a sector-by-sector discussion of scenario results. Section 5.5 summarizes these
results and includes several tables presenting the change in key indicator variables over time and
between scenarios.
5.4 Scenario Results by Sector
This section presents detailed sector-specific results showing the percent change in model outputs
between 2020 and 2040 for the three main scenarios and walks through the impacts of the primary
feedbacks in each sector. Following each sector are text boxes which present the results of relevant
additional decision support scenarios.
Land Use
The Light Rail scenario directly impacts land use in the D-O LRP SD Model through a 10% increase in
the demand for retail, service, and office square feet in Tier 1. This small change is compounded in the
model due to the reinforcing feedback loop involving developed nonresidential sq ft, GRP, and total
employment (see Figure 3-3). In addition, the higher population in the light rail scenarios in Tier 1
causes an increase in demand for residential development, leading to a higher number of single-family
and multifamily dwelling units, relative to BAU. As seen in Figure 5-40, both light rail scenarios show a
larger percent increase in nonresidential and residential development than BAU over the period from
2020 to 2040. For the Light Rail + Redevelopment scenario, it is worth noting that although the percent
growth in nonresidential sq ft is larger than the percent growth in dwelling units over this time, the
change relative to the BAU scenario is much larger for dwelling units (dwelling units increase by 300%
more than BAU while nonresidential sq ft increase by 130% more than BAU).
(A) Tier 1
(B) Tier 2
Nonresidential Sq Ft
Single Family Dwelling Multifamily Dwelling Units
Units
Nonresidential Sq Ft
Single Family Dwelling Multifamily Dwelling Units
Units
• BAU • Light Rail • Light Rail + Redev
i BAU • Light Rail • Light Rail + Redev
Figure 5-40.
Scenarios23
Percent Change in Land Use Sector Model Outputs Between 2020 and 2040 for Three Main
23 Note: Data labels are the absolute change in the model output between 2020 and 2040 for the BAU scenario. All dollar
values are in constant 2010 dollars. B=Billion, M=Million. Model changes due to the Light Rail and Light Rail +
Redevelopment begin in 2020.
105
-------
There is no internal feedback mechanism in the model that would cause density to increase
endogenously. As a result, demand for nonresidential sq ft in the Light Rail scenario leads directly to
increased land development. As shown in Figure 5-41 (see the green dashed line), this land development
leads the Tier 1 areas to reach the maximum allowable expansion at around 2033. In the Light Rail +
Redevelopment scenario, we increase density exogenously, allowing more nonresidential sq ft and
dwelling units to be developed on the same amount of land. More nonresidential sq ft leads to growth in
employment, which increases migration to Tier 1, resulting in a higher population than the Light Rail
Scenario by 2040. Because demand for nonresidential sq ft in the Light Rail + Redevelopment can be
satisfied by adding to previously developed lots, the expansion of developed land slows considerably. In
the Light Rail scenario, developed land in Tier 1 is 15% and 9.2% higher relative to the BAU case in
2030 and 2040, respectively. In the Light Rail + Redevelopment scenario, developed land in Tier 1 only
exceeds the BAU around 2033 (see the blue line in Figure 5-41), reaching 4.9% higher than BAU in
2040. Specifically, developed land is forecasted to reach the maximum allowable expansion of 5,300
acres by 2040 in the Light Rail scenario. In the Light Rail + Redevelopment scenario, on the other hand,
developed land only increases by 200 acres over BAU (5,100 acres vs. 4,900 acres), while allowing for
more nonresidential sq ft, employment, and economic activity.
(A) Tier 1
2010 2020 2030 2040
• Light Rail + Redev — — Light Rail BAU
(B) Tier 2
2010 2020 2030 2040
-Light Rail + Redev — —Light Rail BAU
Figure 5-41. Developed Land: Main Policy Scenarios Compared to BAU
At the introduction of the light rail, the impacts to population and nonresidential sq ft in Tier 2 are
merely equal to the absolute changes in Tier 1, since Tier 1 is contained within Tier 2. However, over
time, the feedbacks from this initial push cause a slightly greater increase in both variables (e.g., the
initial population increase from migration leads to increased births and subsequent population growth),
as shown in Figure 5-39B. The Light Rail + Redevelopment scenario does not significantly impact either
variable in Tier 2, due to the fact that redevelopment in Tier 1 is not connected to Tier 2.24
24 While the Ml effect of redevelopment in Tier 1 does not show up in Tier 2, it does result in a small increase in total
nonresidential sq ft in Tier 2 under the Light Rail + Redevelopment scenario, relative to the Light Rail scenario.
106
-------
Developed land is split between six categories: four nonresidential categories, which are driven by
employment, and two residential categories, which are driven by the split between single-family and
multifamily homes and their respective household sizes. Figure 5-42 shows the distribution of developed
land by category in both Tiers in the BAU scenario in 2000. In Tier 1, the shares vary in small ways,
both over time and between scenarios, largely reflecting a shift towards more nonresidential land as
economic activity increases. In the BAU scenario, the portion of land dedicated to residential use drops
from 57% in 2000 to 52% in 2040, driven by job growth increasing more quickly than population in this
scenario. In the Light Rail scenario, it is assumed that more workers will move to Tier 1 for jobs rather
than commuting, and the resulting increased population growth slows the trend of declining residential
land, with to the portion of land dedicated to residential use declining only to 53% in 2040. In Tier 2, the
shares of land by use remain largely constant over time and between scenarios, with more than 75% of
developed land dedicated to single-family residences.
Tier 1
Tier 2
• Industrial acres, BAU
• Multifamily acres, BAU
• Office acres, BAU
Retail acres, BAU
• Service acres, BAU
• Single family acres, BAU
Figure 5-42. Developed Acres by Use in 2000: BAU
Nonresidential sq ft are divided into four categories (industrial, office, retail, and service), with growth
in each category driven by growth in employment, the shares of employment in each category, and the
employee-space ratios which determine how many square feet are necessary for each employee in a
given category. In Tier 2, between 2000 and 2040, the industrial category declines from 20% to 12% of
the total, while the retail category declines from 33% to 26% of the total. Service remains at about 25%
for the duration, while the office category increases from 22% to 36% of the total by 2040. In Tier 1,
larger shifts in the shares of square footage occur, as shown in Figure 5-43. Because the light rail
scenarios do not affect employment shares and employee-space ratios, the distribution of nonresidential
sq ft by category is not significantly impacted by the main policy scenarios in either Tier.
107
-------
• Industrial sq ft Tier 1, BAU • Office sq ft Tier 1, BAU
• Retail sq ft Tier 1, BAU Sen/ice sq ft Tier 1, BAU
2000
42%
2040
Figure 5-43. Nonresidential Sq. Ft. by Use in 2000 and 2040 - Tier 1: BAU
Developed Land Nonresidential Single Family
Per Capita Floor Area Ratio Density Multifamily Density
(A) Tier 1
2.1 sq ft floor
area per sq ft of
land
(B) Tier 2
• BAU • Light Rail • Light Rail + Redev
Developed Land Nonresidential Single Family
Per Capita Floor Area Ratio Density Multifamily Density
• BAU • Light Rail • Light Rail + Redev
Figure 5-44. Percent Change in Density Measures Between 2020 and 2040 for Three Main Scenarios
The unique impacts of the Light Rail + Redevelopment scenario can most clearly be seen in its impact
on several land use intensity measures in the model, shown in Figure 5-44. In Tier 1, there is little
change in most of the intensity measures under the BAU scenario between 2020 and 2040, though
developed land per capita increases slightly. In the Light Rail scenario, residential density increases by
108
-------
about 14% between 2020 and 2040 due to the increase in population and the fact that the cap on
developed land is reached.25 The overall nonresidential floor area ratio (FAR) does not change, on the
other hand, as development of nonresidential space is capped along with the cap on developed land.
Because it forces both residential and nonresidential density to increase through the process of
redevelopment, the Light Rail + Redevelopment scenario has a more dramatic effect on the land use
intensity variables, relative to BAU, with a 30% increase in all three density measures, and a 13% drop
in developed land use per capita. On the residential side, this increased density corresponds to an
increase from almost 11 dwelling units (du) per acre in 2020 to over 14 du/acre in 2040. The situation in
Tier 2 is more complex. In the BAU scenario, single-family residential units grow from 65% of all units
in 2000 to 67% in 2040; because single-family units use more land per unit than multifamily, residential
density decreases overtime. In both light rail scenarios, developed land per capita increases by 1.1%, a
larger increase than the BAU scenario. This change is because land in the service category grows
slightly relative to other categories in the light rail scenarios, and because this category has a small FAR,
this growth leads to higher land use.
(A) Tier 1
(B) Tier 2
^0.23
6 0.23
S0.23
2 0.22
=> 0.22
n ??
^^^
X .
^^
>*>,.
/ v,
^
7^^
v
>*^-
^^
2000
2010
2020
2030
Redev
— —Light Rail
•BAU
0.61
•Light Rail* Redev
Light Rail
•BAU
Figure 5-45. HHI of Mixed Use: Main Policy Scenarios Compared to BAU
The impact of the change in shares of land use can be seen more clearly in Figure 5-45, which presents a
zoomed-in view of the Herfindahl-Hirschman Index (HHI), a dissimilarity index that has been applied to
measuring land use mix (Song and Rodriguez). A value of 1 in this index indicates that all the land is
25 During model development, we decided it would be unrealistic for the model to force population growth to stop once the
cap on developed land is reached. We instead allowed the model to increase residential density to meet demand, since the
current level of residential density in the area is below limits established by zoning (unlike nonresidential density, which
generally requires approval prior to redevelopment to a higher density). After the cap on developed land is reached, the model
therefore assumes that any increases in dwelling units take place on already developed land, increasing the endogenously
calculated density indicator that affects several other variables in the model, including property values, impervious surface
coefficients, and residential water usage.
109
-------
occupied by only one land use; the lower the value, the more equally mixed are the uses. The overall
increase in mixed use (indicated by a decline in the HHI) under the BAU scenario is due both to changes
in employment shares (which drive demand for nonresidential land) and to the different densities applied
to new land development. The industrial and service categories account for the largest percentages of
land use historically, and both decline in land use between 2015 and 2040, creating a more equal split. In
the Light Rail scenario in Tier 1, the HHI reaches its lowest point in 2026, with a value 1.8% lower than
BAU, due to the 10% increase in demand for retail, service, and office uses under this scenario. In Tier
2, the measure declines slightly under both light rail scenarios (for the same reason as in Tier 1),
reaching a value 1.1% lower than BAU by 2040.
(A) Tier 1
(B) Tier 2
2000
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
• BAU
0.91
0.91
2000
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
• BAU
Figure 5-46. Jobs-Housing Balance: Main Policy Scenarios Compared to BAU
The shifts in the ratios of residential and nonresidential land also impact the jobs-housing balance, an
indicator of the equality of mix between residential and commercial uses that is often used in the
transportation planning literature as an indicator of the propensity for commuting between regions. If an
area has many more homes than jobs, it is likely that many people will need to commute long distances
to other areas. This balance is used in the model to affect the automobile driver mode share. The
measure ranges from 0 for zones with only jobs or housing, not both, to 1 for zones with a nominal
balance of jobs and housing (determined by the number of workers per household, e.g., in Tier 2, there
are 1.2 workers per household, so 1.2 jobs for every household would lead to a jobs-housing balance of
1). Figure 5-46 shows the job-housing balance for both tiers in the BAU and light rail scenarios. In both
Tiers, there are more jobs than workers in households, leading to jobs-housing balance values less than
one. The imbalance is stronger in Tier 1, however; as of 2014 (the most current year of data on jobs and
housing), Tier 1 had a balance value of 0.42, while Tier 2 had a value of 0.90. In Tier 1, the introduction
of the Light Rail brings new residents and accompanying dwelling units in higher proportions than it
does jobs, increasing the balance by 8.8% over BAU by 2040. The Light Rail + Redevelopment scenario
110
-------
does not significantly impact the balance. In Tier 2, jobs grow more quickly than population under the
Light Rail Scenario, leading to a small drop in the balance of-1.6% relative to BAU in 2040. These
changes also impact rates of walking and driving, which contribute to a minor feedback loop connecting
to population by affecting health through physical activity and vehicle emissions.
Alternative Policy Scenario: BAU + Redevelopment
This scenario isolates the effects of redevelopment from the introduction of the Light Rail. As
described in Chapter 4, it conforms to the BAU scenario plus the density-related changes made for
the Light Rail + Redevelopment scenario. In Tier 1, land use density measures increase at roughly
the same pace and magnitude as in the Light Rail + Redevelopment scenario, but there is relatively
little increase in demand for land (which in other scenarios is primarily driven by the revitalization
of Tier 1 from the light rail). In fact, total nonresidential sq ft, total employment, and population all
look much more similar to the BAU scenario than the Light Rail + Redevelopment scenario (Figure
5-47). Because this scenario combines BAU-level demand for developed land with increased
density in development, it leads to a flattening in the growth of developed land (Figure 5-48), and a
steeper drop in developed land per capita (Figure 5-47), since densities increase more than enough
to compensate for the lower population growth. Figure 5-47 also shows that housing costs decline
relative to the BAU (the increase from 2020 to 2040 is 11% lower than the BAU) as land scarcity
declines and the desirability of living in Tier 1 does not increase (due to the absence of the light
rail). In this scenario, Tier 2 looks almost identical to the BAU scenario.
Nonresidential Developed Land Annual Renter
Sq Ft Employment Population per Capita Costs
• BAU • Light Rail + Redevelopment • BAU + Redevelopment
Figure 5-47. Percent Change in Selected Land Use Sector Outputs Between 2020 and 2040 for BAU
+ Redevelopment - Tier 1
2000 2005 2010 2015 2020 2025 2030 2035 2040
^—BAU ^—Light Rail + Redevelopment ^—BAU + Redevelopment
Figure 5-48. Developed Land - Tier 1: BAU + Redevelopment
111
-------
Alternative Policy Scenario: Bold Redevelopment
In this scenario, which is run on top of Light Rail scenario, the percent of land redeveloped and
the density at which is it redeveloped is set high enough to reach an overall increase in density
of 193% for all land in Tier 1, developed and redeveloped. As noted in Chapter 4, this change
matches the results of the Preferred Growth Scenario of the Imagine 2040 Regional Model. Due
to the initial boost to nonresidential sq ft allowed by the high-density redevelopment,
nonresidential sq ft, employment, population, and GRP rise above their levels in the Light Rail +
Redevelopment scenario beginning in 2020, reaching between 6.3% (population) and 9.1%
(nonresidential sq ft) higher by 2040. However this increase in factors that contribute to demand
for development is smaller than the increase in density, so the model actually forecasts a
decrease in developed land, implying that some land is returned to vacant or park use (Figure
5-49). Developed land per capita drops precipitously, reaching a low of 0.04 acres per person in
2040, compared to 0.079 acres per person under the Light Rail + Redevelopment scenario. The
dramatic shifts in density and land use under this scenario impact the three categories of
property values in very different ways (Figure 5-50). Single-family property values decline
relative to the Light Rail + Redevelopment scenario due to decreasing lot sizes and the increased
supply of land. Multifamily property values show relatively little change from the reference
scenario, pushed in opposite directions by the increased supply of land on the one hand and
increasing building size and retail density on the other. Finally, nonresidential property values
increase dramatically, with an increase nearly three times as large (598%) as the increase seen in
the Light Rail + Redevelopment scenario (207%) between 2020 and 2040 due to increasing
building size, retail density, and employment growth. Affordability worsens relative to the
reference scenario, but interestingly, this is not primarily due to any change in housing costs,
which increase by only 0.05% over the Light Rail + Redevelopment scenario between 2020 and
2040. By contrast, transportation-related costs per multifamily household increase by 245%, due
almost entirely to a spike in parking costs, which are affected by the number of jobs per
commercial acre.
8 5
2010
2020
•Light Rail + Redevelopment
2030 2040
•Bold Redevelopment
Figure 5-49. Developed Land - Tier 1: Bold Redevelopment
112
-------
Single Family Multifamily Nonresidential Transportation Parking Cost of An
Property Value Property Value Property Value Related Costs Average Trip
I Light Rail + Redevelopment Bold Redevelopment
Figure 5-50. Percent Change in Housing & Transportation Costs Between 2020 and 2040 for
Bold Redevelopment
Transportation
In the Light Rail scenario, the opening of the light rail line in 2026 produces an increase in the number
of public transit person miles that exceeds the next-best realistic alternative transit development (i.e., the
introduction of a new bus line without an exclusive right-of-way). The size of this increase in public
transit use is a function of population and employment levels in the station areas (Tier 1), as well as
VMT per highway lane mile in the broader metropolitan area (Tier 2). The literature suggests that
people with high-income jobs (those making at least $40,000 per year in 2010 dollars) and jobs in the
retail and entertainment sectors are particularly responsive to light-rail-induced improvements to public
transit (Chatman et al. 2014). As shown in Figure 5-51, the increase in public transit ridership is
approximately the same in both Tiers, meaning that all public transit ridership caused by the light rail in
Tier 2 is attributed to trips that either begin or end in Tier 1. In the first year of light rail service, 2026,
there are 11,000 more transit trips per day in the Light Rail scenario than BAU, an increase of 133% for
Tier 1 and 22% for Tier 2. By the year 2040, public transit ridership in the Light Rail scenario is 375%
higher than BAU in Tier 1 and 48% higher in Tier 2. This continued widening of the transit-ridership
gap between the Light Rail and BAU scenarios occurs primarily because Tier 1 population and
employment continue to grow in the Light Rail scenario between 2026 and 2040, due in part to the
assumed increase in demand for nonresidential development in Tier 1 caused by the light rail (as noted
in Chapter 4). In the Light Rail + Redevelopment scenario, the increased density of development leads
to increases in population and employment. Therefore, by 2040, Tier 1 and Tier 2 transit ridership are
8.8% and 4.1% higher in the Light Rail + Redevelopment scenario than in the Light Rail scenario,
respectively.
113
-------
(A) Tier 1
(B) Tier 2
51 w
S. /0
£. 60
£ m
o
00 20
Light Rail
10 20
+ Redev —
__ m—
/
20 20
— Light Rail
»***^
30 20.
BAU
Figure 5-51.
2000 2010 2020 2030 2040
Light Rail + Redev ——Light Rail BAU
Public Transit Ridership Under Three Main Scenarios
Note: Model changes due to the Light Rail and Light Rail + Redevelopment scenarios begin in 2020 and the rail line opens
in 2026.
In all three main scenarios, population consistently increases over time, causing overall person miles of
travel per day (by all modes) to increase over time, as well. Because Tier 1 population growth
significantly exceeds BAU in both light rail scenarios after 2020, the overall number of person miles of
trips that either start or end in Tier 1 is also greater than BAU in these scenarios (Figure 5-51). In 2040,
overall Tier 1 person miles of travel are 22% and 30% greater than BAU in the Light Rail and Light Rail
+ Redevelopment scenarios, respectively. For Tier 2, 2040 overall person miles of travel are only 3.1%
and 3.9% greater than BAU in the Light Rail and Light Rail + Redevelopment scenarios, respectively,
since the impact on Tier 2 population in these scenarios is a result of changes that only occur in Tier 1.
As discussed in Chapter 3, overall person miles of travel by all modes are determined primarily by
population and GRP, while other factors influencing travel behavior merely shift travel from one mode
to another. As shown in Figure 5-52, though the two light rail scenarios increase person miles of travel
by all modes, they also increase both absolute public transit use, relative to BAU, as well as the
percentage of overall person miles of travel that are on public transit. In 2040, the Tier 1 public transit
mode share (by person miles) is 8.1% and 8.3% in the Light Rail and Light Rail + Redevelopment
scenarios, respectively, as opposed to only 2.1% in the BAU scenario. In Tier 2, the difference in public
transit mode shares is less dramatic (1.8% and 1.9% of overall person miles in the Light Rail and Light
Rail + Redevelopment scenarios, respectively, vs. 1.3% in the BAU scenario). The model assumes that
the average one-way public transit trip involves 0.25 miles of additional nonmotorized travel, since
public transit users generally walk or bicycle some distance to and from transit stations. As a result, the
two light rail scenarios also cause the nonmotorized share of person miles to increase in Tier 1 (from
6.0% in the BAU scenario to 6.6% in both light rail scenarios). At the Tier 2 level, the difference
between scenarios in the nonmotorized share of person miles is negligible.
114
-------
(A) Tier 1
2.5
2020
2040 - BAU 2040 - Light Rail 2040 - Light Rail -
Redev
(B) Tier 2
• Automobile Driver BAutomobile Passenger • Public Transit Nonmotorized
MM
2020
2040 - BAU 2040 - Light Rail 2040 - Light Rail -
Redev
• Automobile Driver BAutomobile Passenger • Public Transit Nonmotorized
Figure 5-52. Modal Person Miles of Travel Per Day in 2020 and in 2040 Under Three Main Scenarios
Note: Model changes due to the Light Rail and Light Rail + Redevelopment scenarios begin in 2020.
Although the two light rail scenarios increase the public transit and nonmotorized mode shares at the
expense of vehicle travel, relative to BAU, the simultaneous increases in population, employment, and
GRP under these scenarios lead to a net increase in VMT relative to BAU (Figure 5-53). Between 2020
and 2040, VMT in Tier 1 increases by 33% and 37% in the Light Rail and Light Rail + Redevelopment
scenarios, respectively, compared to a 25% increase in the BAU scenario. This greater VMT in the two
light rail scenarios leads to greater traffic congestion than in the BAU scenario, despite the much greater
public transit ridership seen in these two scenarios (Tier 1 2020-2040 increases of 337% in the Light
Rail scenario and 376% in the Light Rail + Redevelopment scenario, compared to a decrease of 8% in
the BAU scenario; Tier 2 increases of 85% in the Light Rail scenario and 92% in the Light Rail +
Redevelopment scenario, compared to an increase of 25% in the BAU scenario). In the BAU scenario,
traffic congestion in Tier 1 actually decreases by 2% during 2020-2040, due to assumed improvements
in traffic management and the building of new roadway lane miles, but it increases by 3% and 9% in the
Light Rail and Light Rail + Redevelopment scenarios, respectively. Similar trends are seen in Tier 2,
though the difference between the two light rail scenarios and BAU is much less dramatic. Between
2020 and 2040, Tier 2 VMT increases by 40% in the Light Rail scenario, 41% in the Light Rail +
Redevelopment scenario, and 38% in the BAU scenario. In the same period, Tier 2 traffic congestion
increases by 5% in the Light Rail scenario, 6% in the Light Rail + Redevelopment scenario, and 4% in
the BAU scenario.
115
-------
(A) Tier 1
Congestion
Nonmotorized
Transit Ridership Person Miles
i BAU • Light Rail • Light Rail + Redev
Nonmotorized
Congestion VMT Transit Ridership Person Miles
(B) Tier 2
• BAU • Light Rail • Light Rail + Redev
Figure 5-53. Percent Change in Transportation Sector Model Outputs Between 2020 and 2040 for Three
Main Scenarios
Note: Data labels are the numeric change in the model output between 2020 and 2040 for the BA U scenario. Congestion
(unitless) is measured as the ratio of peak-period travel time to travel time under freeflow conditions. All other outputs are
aggregate counts of miles or trips per day. M=Million. Model changes due to the Light Rail and Light Rail + Redevelopment
scenarios begin in 2020.
These single-digit-percentage increases in traffic congestion in both Tiers show that, even though traffic
congestion is a product of GRP, employment, and population, it is not especially sensitive to changes in
these variables.26 This is due in part to the balancing feedback loop between roadway congestion and
VMT: as roadway congestion increases, automobile traffic decreases, mitigating the effects of any factor
otherwise increases congestion. Two additional factors that prevent dramatic increases in traffic
congestion in the model are as follows: First, all three scenarios share the common assumption that
gradual traffic-management improvements will decrease the amount of congestion that results from any
given amount of peak-period VMT per roadway lane mile. Second, the three scenarios also all assume
that there will be increases in roadway lane miles (of 6.4% in Tier 1 and 6.3% in Tier 2) during 2020-
2040. Furthermore, traffic congestion forms another balancing loop by reducing Gross Operating
Surplus and GRP, which decreases VMT and reduces the upward pressure on congestion.
26 By contrast, Tier 1 GRP increases by 69%, 89%, and 114% in the BAU, Light Rail, and Light Rail + Redevelopment
scenarios, respectively, during 2020-2040; Tier 1 employment increases 35%, 53%, and 66% in the BAU, Light Rail, and
Light Rail + Redevelopment scenarios, respectively; and Tier 1 population increases 14%, 40%, and 48% in the BAU, Light
Rail, and Light Rail + Redevelopment scenarios, respectively, during the same time period. At the Tier 2 scale, GRP
increases 70%, 82%, and 83% during 2020-2040 in the BAU, Light Rail, and Light Rail + Redevelopment scenarios,
respectively; employment increases 41%, 50%, and 51% in the BAU, Light Rail, and Light Rail + Redevelopment scenarios,
respectively; and population increases 38%, 42%, and 44% in the BAU, Light Rail, and Light Rail + Redevelopment
scenarios, respectively, during the same period.
116
-------
As noted in Section 5.2, total employment in Tier 1 grows faster than population in the BAU scenario.
As a result, even though overall person miles of travel in Tier 1 increase faster than population between
2020 and 2040 (17% growth for total person miles of travel vs. 14% growth for population), the number
of person miles by residents increases much more slowly (2.4% growth in the same period).
Consequently, person miles of travel by residents per capita actually decrease in Tier 1 over this period
(Figure 5-54). In the two light rail scenarios, the public transit mode reverses this trend, increasing by a
dramatic 189% and 191% in the Light Rail and Light Rail + Redevelopment scenarios, respectively,
compared to a 29% decrease in the BAU scenario. Because public transit in general, and rail transit
especially, is most often used for trips that the traveler considers to be too long for the use of
nonmotorized modes, the increase in public transit person miles resulting from the opening of the light
rail line comes primarily at the expense of automobile travel, rather than at the expense of nonmotorized
travel. Furthermore, the additional nonmotorized travel related to public transit use in the two light rail
scenarios significantly mitigates the reduction in Tier 1 nonmotorized travel that occurs in the BAU
scenario, carrying potential health benefits. Nonmotorized person miles by Tier 1 residents per capita
decrease by only 12% and 13% in the Light Rail and Light Rail + Redevelopment scenarios,
respectively, during 2020-2040, compared to a 24% decrease in the BAU scenario. As one would expect
from the shifts toward public transit and nonmotorized travel modes described above, automobile driver
person miles by Tier 1 residents per capita decline by more in the Light Rail and Light Rail +
Redevelopment scenarios between 2020 and 2040 (7.9% and 9.9%, respectively) than in the BAU
scenario (5.1%). As a result, Tier 1 residents do not spend as much on vehicle fuel in the two light rail
scenarios in 2040 as they would in the BAU scenario. Because of increased expenditures on public
transit fares, however, the average amount of money spent on transportation by Tier 1 residents per year
per capita in 2040 is 1.5% and 2.7% greater than BAU in the Light Rail and Light Rail +
Redevelopment scenarios, respectively.
Whereas person miles traveled by residents per capita decline for all travel modes in the BAU scenario
in Tier 1, Tier 2 automobile-driver person miles by residents per capita increase by about 3% between
2020 and 2040 (Figure 5-54). The two light rail scenarios feature similar increases because all three
scenarios see increases in GRP per capita, which leads people to favor automobile travel at the expense
of other modes. In the BAU scenario, this leads to negative changes in person miles of travel by
residents per capita for all other travel modes, including a 10% decrease in public transit person miles of
travel by residents per capita. In the Light Rail and Light Rail + Redevelopment scenarios, however,
Tier 2 public transit person miles of travel by residents per capita increase by 29% and 34%,
respectively, during 2020-2040. In all three scenarios, automobile passenger and nonmotorized person
miles by residents per capita decrease during 2020-2040 in both Tiers, though the two light rail
scenarios feature slightly smaller reductions in walking and cycling than BAU.
117
-------
(A) Tier 1
(B) Tier 2
Automobile Driver
200%
175%
150%
125%
100%
-0.34
50% mi./day/person
25%
0%
-25%
-50%
Automobile
Passenger
Public Transit Nonmotorized
IBAU • Light Rail • Light Rail + Redev
Automobile Driver
Automobile
Passenger
Public Transit Nonmotorized
I BAU • Light Rail • Light Rail + Redev
Figure 5-54. Percent Change in Modal Person Miles of Travel by Residents Per Day Per Capita Between
2020 and 2040 for Three Main Scenarios
Note: Data labels are the numeric change in the model output between 2020 and 2040 for the BA U scenario. Model changes
due to the Light Rail and Light Rail + Redevelopment scenarios begin in 2020.
118
-------
Higher Gas Prices
The Higher Gas Prices scenario, which is run on top of the BAU scenario, tests what would happen
if future gas prices turned out to be substantially higher than what is currently projected. As noted in
chapter 4, the model sets gas prices at 2016 to be equal to 2012 levels and then applies the same
annual increase to gas prices as in the BAU scenario. This change produces gas prices that are
consistently 40% higher than BAU between 2016 and 2040. In both this scenario and the BAU
scenario, gas prices increase by 48% during 2016-2040.
Compared to BAU, the Higher Gas Prices scenario produces a 5.7% reduction in Tier 2 VMT by
2030 and a 6.1% reduction by 2040 (Figure 5-55), due to people driving less in order to save money
on vehicle fuel. In Tier 1, the reduction in VMT relative to BAU is 7.2% in 2030 and 8.0% in 2040.
Because traveling less by one mode results in people traveling more by other modes, the Higher Gas
Prices scenario also results in public transit person miles by Tier 2 residents per capita increasing by
5.8% during 2015-2040, as opposed to an 11% decrease under the BAU case (Figure 5-55).
Nonmotorized person miles by Tier 2 residents per capita go down during 2015-2040 in the Higher
Gas Prices scenario, but only by 1.8%, vs. 13% with BAU.
(A)
(B)
Automobile
Automobile Driver Passenger Public Transit Nonmotorized
2010 2020
-Higher Gas Prices
BAU
BAU
Higher Gas Prices
Figure 5-55. (A) VMT - Tier 2 and (B) Percent Change in Modal Person Miles of Travel by Tier 2
Residents Per Day Per Capita between 2015 and 2040 for Higher Gas Prices Scenario vs. BAU
Note: Data labels in the bar chart are the numeric change in the model output between 2015 and 2040 for the BA U
scenario. Model changes due to the Higher Gas Prices scenario begin in 2016.
119
-------
Wo Road Building
This scenario, which is run on top of the BAU scenario, tests the effects of not investing in new road
construction. Whereas the BAU scenario uses an exogenous projection of continuous future road-
building activity, no new roadway lane miles are built after 2017 in this scenario. By 2040, there are
257 fewer Tier 2 lane miles than BAU (6.8% less) and 20 fewer Tier 1 lane miles (7.0% less).
Building lane miles under BAU causes Tier 2 congestion to be 3.2% lower than in the No Road
Building scenario in 2030 and 5.3% lower in 2040 (Figure 5-56), because there is more road capacity
for comparable demand. In Tier 1, BAU congestion is 2.9% less than in the No Road Building
scenario in 2030 and 4.8% less in 2040. Meanwhile, the Light Rail scenario's Tier 2 congestion is
less than the No Road Building scenario's by 2.7% in 2030 and 3.8% in 2040, and Tier 1 congestion
is 1.4% less than in the No Road Building scenario in 2030 and 1.1% greater in 2040. This means
2040 Tier 1 population, GRP, and VMT in the Light Rail scenario increase congestion (by increasing
VMT) by more than the congestion reduction caused by (1) people switching modes due to the rail
line and (2) the congestion-mitigation effect of lane miles built after 2017. Congestion has a relatively
small, negative effect on GRP. The congestion relief of road building that happens in the BAU
scenario but not the No Road Building scenario increases 2040 GRP by only 0.44% in Tier 2 and
0.47% in Tier 1.
The increased congestion under the No Road Building scenario relative to BAU does not greatly
change travel by individual modes. Automobile driver person miles by Tier 2 residents per capita
increase by 5.3% during 2017-2040 in the No Road Building and Light Rail scenarios, vs. 5.9% with
BAU (Figure 5-56). Public transit and nonmotorized person miles by residents per capita in the No
Road Building scenario see slightly smaller declines during 2017-2040 than BAU (9.0% and 9.5%,
respectively, vs. 12% for both modes under BAU). However, nonmotorized travel by residents per
capita decreases by less in the No Road Building scenario (9.5%) than in the Light Rail scenario
(11%), possibly because the Light Rail scenario affects mode shares in a smaller area and is instituted
in a later year.
(B)
Automobile
Automobile Driver Passenger Public Transit Nonmotorized
2000 2010 2020 2030
^^^No Road Building — —Light Rail ^
nJ~U/d°y°/pn£| m£°,f^on \
No Road Building
• Light Rail
Figure 5-56. (A) Congestion - Tier 2 and (B) Percent Change in Modal Person Miles of Travel by Tier
2 Residents Per Day Per Capita Between 2017 and 2040 for No Road Building Scenario vs. BAU and
Light Rail Scenarios
Note: Data labels in the bar chart are the numeric change in model outputs during 2017-2040 for the No Road Building
scenario. Model changes due to the No Road Building scenario start after 2017. Model changes due to the Light Rail
scenario begin in 2020 and the rail line opens in 2026.
120
-------
Fare Free Transit
This scenario, which is run on top of the BAU scenario, tests what would happen if public transit
agencies stopped charging fares on all transit vehicles in both Tiers, rather than building a light rail
line with stations in Tier 1. For this scenario, the model changes fare prices from $0.30 per trip (the
value in 2010 USDs where they are held constant after 2013 in the BAU and light rail scenarios) to
zero in 2026, the year that rail transit service would otherwise commence in the Light Rail scenario.
The Fare Free Transit scenario has a far more dramatic effect on people's transportation mode choices
relative to BAU than does the Light Rail scenario, both because of the price elasticity of demand and
because it affects the entire public transit system, not just the fraction that is in Tier 1. In 2030 and
2040 the Fare Free Transit scenario causes Tier 2 ridership to be 274% and 272% greater than BAU,
respectively, compared to increases relative to BAU of 29% and 48% in the Light Rail scenario. In
2040, public transit person miles by Tier 2 residents per day per capita in the Fare Free Transit
scenario are 0.94 miles per day per person greater than BAU, compared to 0.15 miles per day per
person greater in the Light Rail scenario (Figure 5-57). Similarly, the increase in 2040 nonmotorized
person miles by Tier 2 residents per day per capita (relative to BAU) is about eight times larger in the
Fare Free Transit scenario than in the Light Rail scenario (0.0388 vs. 0.0049 miles per day per
person). Finally, both the Fare Free Transit and Light Rail scenarios yield 2020-2040 increases in
automobile driver person miles by Tier 2 residents per day per capita, but these increases are 0.62 and
0.11 miles per day per capita less than BAU, respectively.
Because automobile driver miles per capita are lower than BAU in the Fare Free Transit scenario,
traffic congestion is lower than BAU as well. The amount of time lost to congestion delay per vehicle
mile of travel in the Fare Free Transit scenario is 18% and 15% less than BAU in 2030 and 2040,
respectively (Figure 5-57). By contrast, because the Light Rail scenario significantly increases
population relative to both the BAU and Fare Free Transit scenarios, it results in congestion delay per
vehicle mile of travel that is 3.5% and 12% greater than BAU in 2030 and 2040, respectively.
(A)
(B)
1.0
I °-8
g. 0.6
fc 0.4
§• 0.2
fc 0.0
S" -0.2
_OJ
:£ -0.4
-0.6
Automobile
Automobile Driver Passenger Public Transit Nonmotorized
Jl
0.0014
0.0012
tj 0.0010
£ 0.0008
S" 0.0006
£ 0.0004
0.0002
0.0000
l Light Rail
Fare Free Transit
2000 2010 2020 2030 2040
Fare Free Transit — —Light Rail BAU
Figure 5-57. (A) Change in Modal Person Miles of Travel by Tier 2 Residents Per Day Per Capita
Between 2020 and 2040 for Fare Free Transit Scenario vs. BAU and Light Rail Scenarios and (B) Hours
of Congestion Delay Per VMT - Tier 2
Note: In the three scenarios shown here, 2020 modal person miles by Tier 2 residents per day per capita are as follows:
Automobile Driver = 18; Automobile Passenger = 6.1; Public Transit = 0.38; Nonmotorized = 0.54. Model changes due
to the Fare Free Transit scenario start in 2026.
121
-------
High Parking Price
This scenario tests the effects of a sudden, steep increase in parking costs. In Tier 1, the scenario
assumes an increase of $4.00 per trip (in 2010 USDs) in 2020, in addition to increases caused by
endogenous factors in the model, while Tier 2 parking costs are only affected to the extent that Tier 1 is
in Tier 2. Otherwise, this scenario's inputs are identical to the Light Rail + Redevelopment scenario.
We selected the Light Rail + Redevelopment scenario as a reference because its dense Tier 1
development would produce conditions most likely to result in high demand for parking.
Because higher parking costs discourage driving, the High Parking Price scenario results in a larger
2020-2040 decrease in automobile driver person miles by Tier 1 residents per capita (15%) than the
Light Rail + Redevelopment scenario (10%, see Figure 5-58). The model incorporates a tradeoff among
transportation modes, so this larger decline in automobile driver miles per capita leads to a smaller
decline in automobile passenger miles by Tier 1 residents per capita (11% vs. 18%) and a smaller
decline in nonmotorized person miles by Tier 1 residents per capita (6.4% vs. 13%). Because both of
these scenarios incorporate the boost to transit ridership that comes from the light rail line and the land
use changes associated with it, they also both feature significant increases in public transit person miles
by Tier 1 residents per capita, but the increase in the High Parking Price scenario is slightly larger
(194% vs. 191%). The reduction in driving, and hence congestion, that results from higher parking
prices produces a small increase in GRP (<1% above the Light Rail + Redevelopment scenario in
2040), with corresponding increases in jobs, net migration, and population, in both Tiers.
In the first year of increased parking prices in the High Parking Price scenario (2020), they increase
combined transportation and renter costs, relative to the Light Rail + Redevelopment scenario, by 18%
and 12% in Tier 1 and Tier 2, respectively (see Figure 5-58 for Tier 1 impacts). During 2020-2040,
renter costs increase faster in Tier 1 than Tier 2, so the Tier 1 impact of higher parking costs as a
percent of total costs declines over time. As a result, by 2040, the percent difference in combined
transport and renter costs relative to the Light Rail + Redevelopment scenario is 13% in both Tiers.
(A)
(B)
200%
175X
150X
125%
10CK
75%
50%
25%
Automobile i_ 25
Automobile Driver Passenger Public Transit Nonmotorized
mi. /day/person
-0.64
mi. /day/person
l Light Rail + Redev • High Parking Price
2000 2010 2020 2030 2040
High Parking Price Light Rail + Redev
Figure 5-58. (A) Percent Change in Modal Person Miles of Travel by Tier 1 Residents Per Capita
Between 2020 and 2040 for High Parking Price Scenario vs. Light Rail + Redevelopment Scenario and (B)
Transportation and Renter Costs - Tier 1
Note: Data labels in the bar chart are the numeric change in the model output between 2020 and 2040 for the Light Rail +
Redevelopment scenario. Model changes due to the High Parking Price scenario begin in 2020. The line graph is specific to
households in multifamily housing.
122
-------
Sidewalk Building
This scenario tests the effects of doubling the demand for nonmotorized travel facilities (e.g.,
sidewalks, bike lanes, and paths) per developed acre in 2020. Otherwise, the Sidewalk Building
scenario is identical to the BAU scenario. Due to a delay built into the model to account for
construction time, the increase in demand for nonmotorized travel facilities does not produce an
increase in sidewalks, bike lanes, and paths in the study area until 2022. In both Tiers, by 2026, there
are approximately twice as many nonmotorized-travel facility miles in the Sidewalk Building scenario
as in the BAU scenario, and they stay within 1% of twice the BAU number of facility miles through
2040 (Figure 5-59).
The primary effect of the Sidewalk Building scenario is an increase in nonmotorized travel, which
carries health benefits for residents. When nonmotorized travel facilities are limited, even if a
particular origin and destination are close enough together for people to walk or bicycle between them,
they may still choose to make the trip by other modes out of concern for their safety, given the risk
inherent in traveling by a slow, nonmotorized mode in a transportation corridor that is only designed
for fast, motorized travel. Consequently, the doubling of the amount of nonmotorized-travel facility
miles in this scenario causes nonmotorized person miles of travel by Tier 2 residents per capita to
increase by 26% during 2020-2040, as opposed to decreasing 7.6% under the BAU scenario (Figure
5-59). Since the model assumes that using public transit entails some amount of walking or cycling to
and from transit stops, policy actions that increase nonmotorized travel tend to also increase public
transit travel, and vice-versa. As a result, in the Sidewalk Building scenario, public transit person
miles by Tier 2 residents per capita increase by 10% during 2020-2040, vs. a 10% decrease in the
BAU scenario. Finally, as people travel more by walking, cycling, and using public transit, they travel
less by automobile. In the Sidewalk Building scenario, automobile driver person miles by Tier 2
residents per capita increase by only 0.98% during 2020-2040, vs. a 3.8% increase in the BAU
scenario.
(B)
Automobile
Automobile Driver Passenger Public Transit Nonmotorized
-Sidewalk Building
Sidewalk Building
Figure 5-59. (A) Nonmotorized Travel Facilities - Tier 2 and (B) Percent Change in Modal Person
Miles of Travel by Tier 2 Residents Per Day Per Capita Between 2020 and 2040 for Sidewalk Building
Scenario vs. BAU
Note: Data labels in the bar chart are the numeric change in the model output between 2020 and 2040 for the BA U
scenario. Model changes due to the Sidewalk building scenario begin in 2020.
123
-------
Energy
In Tier 1, energy use is projected to grow by 7% between 2020 and 2040 in the BAU scenario and by
17% and 36% in the Light Rail and Light Rail + Redevelopment scenarios, respectively (Figure 5-60).
In all three scenarios, this increase is driven mainly by increases in passenger vehicle VMT, dwelling
units, and nonresidential sq ft. Dwelling units and nonresidential sq ft increase due to population growth
and the higher demand for nonresidential sq ft stimulated by the light rail. As noted above, the two light
rail scenarios have higher growth in population and GRP per capita, which together lead to increased
VMT (higher GRP per capita increases vehicle use because wealthier people are assumed to drive
more). These increases are larger than the factors that reduce energy use in the future (improved vehicle
fuel efficiency and reduced building energy intensity), leading to the net increase in energy use seen in
Figure 5-60A. Despite this increase in total energy use, regional energy intensity (energy use per GRP
and energy use per capita) is projected to decrease in both Tiers.
(A) Tier 1
(B) Tier 2
-Light Rail + Redev — —Light Rail
-Light Rail + Redev — —Light Rail
Figure 5-60. Energy Use: Light Rail, Light Rail + Redevelopment Scenarios Compared with BAU
Tier 2 energy use is projected to grow by 11% between 2020 and 2040 in the BAU scenario and by 15%
and 16% in the Light Rail and Light Rail + Redevelopment scenarios, respectively (Figure 5-60). The
smaller Tier 2 impact of the Light Rail + Redevelopment scenario is expected, as the model only
changes development density in Tier 1 in this scenario. The increase in energy use in Tier 2 is driven by
the same factors as in Tier 1 (increases in VMT, population, dwelling units, and nonresidential sq ft,
offset by improvements in vehicle fuel efficiency and building energy intensity), though the increases
are smaller in proportion to the larger baseline values in Tier 2.
(A) Tier 1
(B) Tier 2
1.8
1.6
1.4
1.2
1
2000
2010
•Light rail+redev
2020
2030
2040
•Light rail
•BALI
2010
•Light rail+redev
2020
2030
2040
• Light Rail
•BAU
Figure 5-61. CC>2 Emissions: Light Rail, Light Rail + Redevelopment Scenarios Compared with BAU
124
-------
Between 2020 and 2040, Tier 1 CO2 emissions increase by 12% in the BAU scenario and by 24% and
47% in the Light Rail and Light Rail + Redevelopment scenarios, respectively (Figure 5-61). As the
figure shows, CO2 emissions level off around year 2033 in the Light Rail scenario (due to the halt in
land development as the Tier 1 cap is reached) but continue to increase in the Light Rail +
Redevelopment scenario, reaching 1.7 million tons/year in 2040. Because buildings make up the
majority of CO2 emissions, the scarcity of space for new dwelling units and nonresidential sq ft that
occurs in the Light Rail scenario explains why CO2 emissions plateau and why they continue to increase
in the Light Rail + Redevelopment scenario, where this scarcity is less of a factor. In Tier 2, CO2
emissions increase by 17% between 2020 and 2040 in the BAU scenario and by 22% and 23% in the
Light Rail and Light Rail + Redevelopment scenarios, respectively (Figure 5-61). Because land scarcity
is not a factor in any of the scenarios in Tier 2, CO2 emissions continue to grow through 2040, reaching
11 million tons/year in the Light Rail + Redevelopment scenario.
(A) Tier 1
(B) Tier 2
2020 2040 - BAU 2040 - Light Rail 2040 - Light Rail +
Redev
• water treatment and distribution • public transport, incl. rail after 2020
• building natural gas use • building electricity use
• passenger vehicle fuel
• water treatment and distributic
• building natural gas use
• passenger vehicle fuel
2040 - Light Rail 2040 - Light Rail -
Redev
public transport, incl. rail after 2020
• building electricity use
Figure 5-62. Energy Use in 2020 and 2040 Under the Three Main Scenarios
Figure 5-62 shows the distribution of total energy use in 2020 (same for all scenarios) and in 2040 for
the BAU and light rail scenarios. As the figure shows, building natural gas, building electricity, and
passenger vehicles compose the largest shares of total energy use in all scenarios in both Tiers, with only
minor contributions from water treatment and distribution and public transportation (which includes
energy use by the light rail after 2020). In Tier 1, passenger vehicles and building electricity are 40%
and 39% of 2020 energy use, respectively. Energy consumption by passenger vehicles declines slightly
between 2020 and 2040 in all scenarios, due to a projected 50% increase in fuel efficiency (from 18
MPG in 2020 to 27 MPG in 2040). This decline brings vehicle energy use from 3.4 million MMBtu in
2020 to 2.9 million, 3.1 million, and 3.2 million MMBtu in 2040 in the BAU, Light Rail, and Light Rail
+ Redevelopment scenarios, respectively. Despite decreased building energy use intensity between 2020
and 2040 (by about 7-12%, depending on building type), building electricity use increases between 2020
and 2040, by 21%, 34%, and 63% in the BAU, Light Rail, and Light Rail + Redevelopment scenarios,
respectively.
Tier 2 consumes about eight times as much energy as Tier 1. As in Tier 1, passenger vehicles and
building electricity constitute the largest shares of regional energy use with 48% and 33% of the BAU
total in 2020, respectively (Figure 5-62). Despite an increase in VMT between 2020 and 2040, passenger
vehicle energy use declines slightly in Tier 2 in the three main scenarios over this period, as it does in
Tier 1, due to increased fuel efficiency. Tier 2 passenger vehicles consume 35 million MMBtu in 2020,
compared with 33 million, 33 million, and 34 million MMBtu in the 2040 BAU, Light Rail, and Light
Rail + Redevelopment scenarios, respectively. Building electricity use in Tier 2 grows by 27% between
2020 and 2040 in BAU, and by 34% and 35% in Light Rail and Light Rail + Redevelopment,
125
-------
respectively. The largest difference between the Tiers in terms of energy use is that building energy use
increases by about the same amount in both light rail scenarios in Tier 2, whereas the increase in
building energy use in Tier 1 is much higher in the Light Rail + Redevelopment scenario than in the
Light Rail scenario. As noted above, this difference is due to the fact that the density assumptions in the
Light Rail + Redevelopment scenario are felt almost entirely in Tier 1, since land scarcity is modeled as
a constraint in Tier 1 but not in Tier 2.
(A) Tier 1
(B) Tier 2
1.4
buses Tier 1
building natural gas use Tier 1
building electricity use Tier 1
passenger vehicles Tier 1
water treatment and distribution
buses
building natural gas use
building electricity use
passenger vehicles
Figure 5-63. CC>2 Emissions Over Time in the BAD Scenario, Color Coded by Source
Note: Colors in this figure match energy budget components in the previous figure.
Figure 5-63 presents the distribution of CO2 emissions by source over time in the BAU scenario for Tier
1 (Figure 5-63) and Tier 2 (Figure 5-63). In both Tiers, the largest source of CO2 emissions is building
electricity use (colored in orange in Figure 5-61). In Tier 1, building electricity use in the BAU scenario
grows from 57% to 69% of total emissions between 2000 and 2040. Building electricity use is a larger
share of CO2 emissions in Tier 1 than in Tier 2, because nonresidential buildings (which tend to have a
higher energy intensity than residences) compose a higher percentage of developed land in Tier 1 than in
Tier 2 (about 40% vs. about 20%). Due to increases in fuel efficiency, passenger vehicle emissions
decline from 33% of total Tier 1 emissions in 2000 to 19% in 2040.
Similar patterns are seen in CO2 emissions in Tier 2 in the BAU scenario (Figure 5-63), with building
electricity use growing from 49% to 63% of total CO2 emissions between 2000 and 2040. As in Tier 1,
CO2 emissions from passenger vehicles decrease over this same period, from 42% of the total in 2000 to
26% by 2040. Meanwhile, municipal water treatment and distribution accounts for less than 1% and
buses account for less than 0.5% of Tier 2 CO2 emissions. If light rail were added, the electricity used to
power the rail line would represent less than 0.5% of Tier 2 CO2 emissions.
126
-------
(A) Tier 1
Total Energy
Spending
Per Capita
Energy
Spending
IBAU
Per Capita
Transportation
Energy
Spending
• Light Rail
Per Capita
Residential
Energy
Spending
Per Capita
Commercial
Energy
Spending
Per Capita
Industrial
Energy
Spending
i Light Rail + Redev
(B) Tier 2
70%
50%
30%
10%
-10%
-30%
Total Energy
Spending
i — : —
Per Capita
Energy
Spending
J ! 1
Per Capita
Transportation
Energy
Spending
Per Capita
Residential
Energy
Spending
Per Capita
Commercial
Energy
Spending
Per Capita
Industrial
Energy
Spending
IBAU
l Light Rail
l Light Rail + Redev
Figure 5-64. Percent Change in Energy Spending Between 2020 and 2040 for Three Main Scenarios
Figure 5-64 summarizes the percent change in several categories of energy spending between 2020 and
2040 in the three main scenarios in Tier 1 (A) and Tier 2 (B). As the figure shows, total energy spending
increases over this time period in both Tiers for the three main scenarios. In Tier 1, energy spending
increases by 32% ($58M in USD 2010) between 2020 and 2040 in the BAU scenario, and by 45% and
66% in the Light Rail and Light Rail + Redevelopment scenarios, respectively. Overall per capita energy
spending in Tier 1 increases in all three scenarios (by 15% in BAU, 4% in Light Rail, and 12% in Light
Rail + Redevelopment, respectively), but dividing total energy spending into transportation, residential,
commercial, and industrial categories reveals some differences. Transportation energy spending per
capita decreases in the two light rail scenarios, where population increases faster than vehicle energy
spending, but increases in BAU, where the reverse is true. Per capita energy spending in residential,
commercial, and industrial buildings increases in all three scenarios, but the Light Rail scenario has the
lowest per capita commercial and industrial energy spending due the land scarcity constraint on
development of nonresidential sq ft around 2033.
Tier 2 energy spending follows many of the patterns of Tier 1, although differences between the three
main scenarios are proportionately smaller. Tier 2 energy spending increases between 2020 and 2040 by
41% ($650M in USD 2010) in the BAU scenario, and by 46% and 47% in the Light Rail and Light Rail
+ Redevelopment scenarios, respectively. Overall per capita energy spending increases by 1.6%, 2.3%,
and 2.2% in the BAU, Light Rail, and Light Rail + Redevelopment scenarios, respectively. Tier 2
transportation energy spending per capita declines in all three scenarios, but by more in the Light Rail
127
-------
and Light Rail + Redevelopment scenarios (-4.7% and -5.0%, respectively) than in the BAU scenario (-
3.8%). This decline is due to increases in vehicle fuel efficiency and decreases in VMT per capita, with
larger VMT per capita decreases in the light rail scenarios than in the BAU scenario. In all three
scenarios, per capita energy spending in residential, commercial, and industrial buildings increases due
to increased development and increased electricity and natural gas prices, despite a decrease in building
energy intensity (see Figure 5-65 below).
(A) Tier 1
Energy use per capita Nonresidential energy Total building energy
Energy use/GRP Tier 1 Tier 1 intensity Tier 1 intensity Tier 1
-60% -0.38 MMBtu per
-70% thousand USD 2010
-90%
-100%
• BAU • Light Rail • Light Rail + Redev
1 1ler *• Nonresidential energy Total building energy
Energy use/GRP Energy use per capita intensity intensity
-0.017 MMBtu per
year per sq ft
-60% _0 73 MMBtu per
»™/° I thousand USD 2010
-80/o
-90%
-100%
-21 MMBtu per year
Per acre
• BAU • Light Rail • Light Rail + Redev
Figure 5-65. Percent Change in Energy Intensity Between 2020 and 2040 for Three Main Scenarios
Figure 5-65 presents percent changes in four measures of energy intensity between 2020 and 2040 in
both Tiers for the three main scenarios, including overall energy use per GRP, energy use per capita,
nonresidential energy intensity (energy use per year per sq ft), and total building energy intensity
(energy use per year per developed acre). In almost every measure, energy intensity decreases in both
Tiers over this time period for the three main scenarios. This reflects assumptions in the model that
technological improvements will improve efficiency of energy use in buildings and vehicles. Energy use
per GRP decreases by about a third in both Tiers across all three scenarios, meaning that the economy
becomes significantly more energy efficient. Energy use per capita declines in both Tiers in all three
scenarios. In Tier 1, energy use per capita sees the largest drop (-16%) in the Light Rail scenario, due to
continued population growth and a plateauing of building energy use around 2033 as a result of land
scarcity. In Tier 2, energy use per capita declines by about 20% in all three scenarios, since land scarcity
is less of a factor in this Tier. In Tier 1, total building energy intensity (MMBtu per year per acre)
declines by about 1-2% in the BAU and Light Rail scenarios, but increases by 26% in the Light Rail +
Redevelopment scenario. In that scenario, higher density development causes an increase in building
energy use without any corresponding increase in developed land, leading to an increase in energy
intensity (per developed acre) that dominates over the decline in energy intensity caused by energy
efficiency improvements.
128
-------
Energy Efficiency
In this scenario, beginning in 2015, energy intensity of residential, commercial, and industrial
buildings drops linearly to a value 10% lower than the Light Rail + Redevelopment projection by
2040. At the same time, passenger vehicle MPG is multiplied by a factor that increases linearly to
reach a value 10% higher than the Light Rail + Redevelopment projection by 2040. This scenario
tests the effects of energy efficiency on overall energy use and CO2 emissions, as well as on
economic effects which themselves indirectly affect energy consumption.
The Energy Efficiency scenario causes CO2 emissions from buildings and transportation in Tier 2 to
drop by 9.1% relative to the Light Rail + Redevelopment scenario by 2040 (Figure 5-66). This
decrease in energy spending leads to increased economic growth (see the Energy-^Economy-^Land
Use-^Energy loop in Chapter 3), causing GRP to increase by 1.3% by 2040 (Figure 5-66). The same
feedback loop that connects energy spending to GRP also has a balancing effect, however, as GRP
growth leads to increased development of nonresidential sq ft (Figure 5-66), which causes energy
spending to rise again. These feedback effects are small compared to the overall effect on energy
use, as the increase in nonresidential sq ft between 2020 and 2040 is only 0.8% higher in the Energy
Efficiency scenario (at 51.5%) than in the Light Rail + Redevelopment scenario (at 50.7%).
(A)
(B)
(Q
2000 2005 2010 2015 2020 2025 2030 2035 2040
2000 2005 2010 2015 2020 2025 2030 2035 2040
-Energy efficiency LRT+redev
Nonresidential sq ft
-Light Rail + Redev
Total energy spending
• Energy efficiency LRT+redev
-Light Rail + Redev
Energy efficiency LRT+Redev
• Light Rail + Redev
Figure 5-66. (A) Tier 2 CO2 Emissions, (B) Tier 2 GRP, and (C) Percent Change between 2020 and
2040 (Tier 2): Light Rail + Redevelopment Scenario with Energy Efficiency
129
-------
Higher LRT Effect on Public Transit Ridership
Estimates of future light rail usage are highly uncertain, but these estimates play an important role in
the D-O LRP model and affect energy variables. In two scenarios, beginning with rail operation in
2026, the effect of light rail on public transit person miles per year is increased by 10% or by 100%
(doubled) compared to its value in the Light Rail + Redevelopment scenario. Whereas the Light Rail
+ Redevelopment scenario sees 45 million person miles of public transit use in 2040, the 10%
increase in ridership leads to 50 million person miles of public transit use, and the 100% increase
leads to 90 million person miles in that same year (Figure 5-67). These increases in public transit
usage due to the light rail are assumed to occur entirely in Tier 1.
In these scenarios, the increase in Tier 1 VMT between 2020 and 2040 is smaller than in the Light
Rail + Redevelopment scenario, with increases of 36.6% and 32.3% in the 10% ridership increase
and 100% ridership increase scenarios, respectively, compared to 37.1% in the reference scenario
(Figure 5-67). Slower VMT growth leads to slower growth in congestion, which increases by 8.5%
and 5.0% in the two high-ridership scenarios, compared to 8.8% in the reference scenario. Despite
the smaller increase in VMT, these scenarios show little difference in total Tier 1 CO2 emissions,
which increase by only 0.3% less in the doubled-ridership scenario than the Light Rail +
Redevelopment scenario. Although doubling the LRT ridership effect on public transit substantially
reduces VMT, even higher LRT ridership levels would be needed to reduce CO2 emissions by 1% or
more.
(A)
2000 2010 2020 2030
^—LRT public transit effect +100pct LRT+redev
^—LRT public transit effect +10pct LRT+redev
^—Light Rail + Redev
(B)
Congestion
C02 emissions
LRT public transit effect +100pct LRT+redev
LRT public transit effect +10pct LRT+redev
• Light Rail* Redev
Figure 5-67. (A) Light Rail Effect on Public Transit Usage - Tier 2 and (B) Energy and Transportation
Indicators - Tier 1: Light Rail + Redevelopment With Higher Light Rail Effect on Public Transit
Ridership
130
-------
Clean Power Plan
This scenario approximates the Clean Power Plan goal for North Carolina (US EPA 2015c), which
includes technology change in (1) fossil fuel-fired steam plants and (2) natural gas combined cycle
plants. To represent these changes, this scenario decreases the electricity emissions factor linearly
between 2022 and 2030 to reach 77% of its level in the Light Rail + Redevelopment scenario. This
percentage change results in a drop from 1,560 Ib CC>2 per MWh in 2022 to 1,200 Ib CO2 per MWh
in 2030, after which it remains constant. As a result, CO2 emissions from buildings and transportation
drop by 7.7% in Tier 2 between 2022 and 2030, then increase between 2030 and 2040 due to
economic and population growth (Figure 5-68A, orange line). Without any further reductions to the
electricity emissions factor between 2030 and 2040, CO2 emissions reach 9.7 million tons per year,
slightly above the 2022 emissions rate but still 15% below the Light Rail + Redevelopment emissions
rate in 2040.
In Tier 1, the Clean Power Plan would decrease CO2 emissions by 1.7% between 2022 and 2030
(Figure 5-68B, orange line). This lesser decline relative to Tier 2 is due to the steeper growth in
emissions after 2020 in the Light Rail + Redevelopment scenario in Tier 1. As in Tier 1, CO2
emissions increase between 2030 and 2040 due to economic and population growth, reaching 1.4
million tons per year, a value 16% below the Light Rail + Redevelopment emissions rate.
(A) Tier 2
(B) Tier 1
2000 2005 2010 2015 2020 2025 2030 2035 2040
10 2015 2020 2025 2030 2035
-Clean Power Plan LRT+Redev
-LRT + Redevelopment
-Clean Power Plan LRT+Redev
-LRT + Redevelopment
Figure 5-68.
Plan
CC>2 Emissions: Light Rail + Redevelopment Scenario With and Without Clean Power
131
-------
Increased Solar Capacity
In these scenarios, desired solar capacity is increased from 40 MW to 80, 320, and 640 MW, such that
solar capacity reaches a level 2, 8, and 16 times as high as in the Light Rail + Redevelopment scenario
by 2022, at which point it remains constant. This scenario is meant to examine the effects of rapid
growth in solar capacity on CO2 emissions (continuing current trends where solar capacity has roughly
doubled annually between 2010 and 2014). Increasing solar capacity 16-fold (blue line in Figure
5-69A) results in CO2 emissions from buildings and transportation that are 5.4% lower than the Light
Rail + Redevelopment scenario by 2040 (blue line in Figure 5-69B). Assuming a solar capacity factor
of 0.15 for North Carolina (Kaplan and Ouzts 2009) and 640 MW desired solar capacity, the 640-MW
solar scenario results in solar meeting almost 12% of building electricity demand by 2022, dropping to
9% by 2040 due to growth in building sq ft and overall building electricity use (Figure 5-69C).
Because solar represents a fraction of the Tier 2 energy mix (about 0.8% of total building electricity
use in 2015) it can begin reducing regional CO2 emissions more than a few percent if solar capacity
grows about tenfold.
(A)
(B)
400
300
2000 2010
640MW solar LRT+redev
80MW solar LRT+redev
2020 2030 2040
—320MW solar LRT+redev
^—Light Rail + Redev
2000 2010
640MW solar LRT+redev
80MW solar LRT+redev
Light Rail + Redev
2020
2030 2040
— 320MW solar LRT+redev
- — — Durham Sustainability Office
(Q
0.14
2000 2010
640MW solar LRT+redev
80MW solar LRT+redev
2020 2030 2040
—320MW solar LRT+redev
Light Rail + Redev
Figure 5-69. (A) Solar Capacity, (B) Annual CC>2 Emissions, and (C) Solar as a Fraction of Building
Electricity Use: Light Rail + Redevelopment Scenario Compared With Increased Solar Capacity
Scenarios in Tier 2
132
-------
Economy
The model outputs in the economy sector illustrate the importance of redevelopment in order to fully
realize the direct and indirect benefits of the light rail in the station areas and the DCHC MPO region as
a whole.27 The light rail itself attracts economic activity that has cascading impacts throughout the
region, but the ability of people to reap the economic benefits of the light rail in Tier 1 is limited by the
amount of space in this area. In order for economic growth in Tier 1 to continue, high-density
redevelopment must therefore occur in this area. In the Light Rail + Redevelopment scenario, the
economic benefits of redevelopment in Tier 1 are not felt in Tier 2 during the time frame of our model
because of the current assumption that an increase in density in Tier 1 leads to a decrease in density in
the area in Tier 2 that is outside of Tier 1 (a.k.a. "Tier 2 donut"). We ran the "Higher Tier 2 Density"
scenario to explore the impacts of changing this assumption.
(A) Tier 1 Absolute Change 2020-2040
(B) Tier 1 Percent Change 2020-2040
120%
100%
Total Earnings Gross Operating Gross Regional Total Retail
Surplus Product Consumption
• BAU iLightRail Blight Rail + Redev
(C) Tier 2 Absolute Change 2020-2040
Total Earnings Gross Operating Gross Regional Total Retail
Surplus Product Consumption
• BAU • Light Rail • Light Rail + Redev
(D) Tier 2 Percent Change 2020-2040
Total Earnings Gross Operating Gross Regional Total Retail
Surplus Product Consumption
• BAU • Light Rail • Light Rail + Redev
Total Earnings Gross Operating Gross Regional Total Retail
Surplus Product Consumption
• BAU BLightRail • Light Rail + Redev
Figure 5-70. Change in Overall Economic Growth Indicators between 2020 and 2040
27 Figures in the economy sector results section present model outputs of the main economic indicators in the D-O LRP SD
Model between 2020 and 2040, with 2020 being the initial year that model outputs for the three main scenarios begin to
diverge. The majority of the figures presented are bar graphs showing either the absolute or percent change (relative to 2020
values) in model outputs between 2020 and 2040 for the three main scenarios.
133
-------
Figure 5-70 A-D shows the absolute and percent change between 2020 and 2040 of four economic
growth indicators for Tier 1 and Tier 2. The increase in demand for commercial sq ft (nonresidential sq
ft that is retail, office, or service, but not industrial) that occurs in Tier 1 under the Light Rail scenario is
the primary driver of economic growth in that Tier between 2020 and 2040. Under the Light Rail +
Redevelopment scenario, the additional nonresidential sq ft added to Tier 1 due to high-density
redevelopment compounds this economic growth. Because nonresidential sq ft is directly linked to gross
operating surplus (GOS), Tier 1 GOS increases by 140% under the Light Rail + Redevelopment scenario
(Figure 5-70B) compared to a 90% increase in the Light Rail scenario and an 80% increase in the BAU
scenario. This increase in the GOS in Tier 1 leads to an increase in GRP and total retail consumption,
which lead to an increase in employment, total earnings, and finally an additional increase in GRP (due
to a reinforcing feedback loop within the economy sector). These increases can be seen in Figure 5-70,
where all four economic growth variables show larger increases in the Light Rail + Redevelopment
scenario than in the Light Rail scenario, which in turn shows larger increases than in the BAU scenario.
Because redevelopment does not occur in Tier 2, the difference between the Light Rail and Light Rail +
Redevelopment scenarios in the four overall economic growth indicators in Tier 2 (shown in Figure
5-70) is not as pronounced as in Tier 1. In Tier 2, the difference between the BAU and light rail
scenarios in the four indicators is mostly attributable to the Tier 1 changes mentioned above. The slight
difference between the Light Rail + Redevelopment scenario and Light Rail scenario in overall
economic growth indicators in Tier 2 is due to a mechanism in the model that links the effect of the light
rail on demand for nonresidential sq ft in Tier 1 to Tier 2. The model endogenously calculates the ratio
of nonresidential sq ft in Tier 1 to nonresidential sq ft in Tier 2 (varying between about 17% and 21%),
so an increase in nonresidential development in Tier 1 leads to a higher Tier I/Tier 2 nonresidential sq ft
ratio. As a result, this higher ratio increases the demand for nonresidential sq ft in Tier 2 under the light
rail scenarios by the same amount as in Tier 1. The increase is magnified over time due to feedback
loops, leading to a slightly higher increase in Tier 2 nonresidential sq ft than in Tier 1. We used this
model structure on the rationale that 1) Tier 2 includes the changes in Tier 1, and 2) additional new
development is expected to take place in the portion of Tier 2 just outside of Tier 1 in response to the
light rail, though to a lesser extent than development in Tier 1.
(A) Tier 1
(B) Tier 2
Unemployed Resident Labor Resident
Total Employment Population Force Employment
Unemployed Resident Labor Resident
Total Employment Population Force Employment
I BAU • Light Rail • Light Rail + Redev
I BAU • Light Rail • Light Rail + Redev
Figure 5-71. Percent Change in Economy Sector Employment Indicators between 2020 and 2040 for
Three Main Scenarios
Total employment in Tier 1 (Figure 5-71) increases by nearly twice as much between 2020 and 2040
under the Light Rail + Redevelopment scenario as in the BAU scenario, with an increase in jobs of 66%
and 35%, respectively. The increase in the amount and density of nonresidential sq ft added to Tier 1
leads to increased employment creation, immigration, and residential development. In Tier 2, the largest
134
-------
increase in total employment (Figure 5-71) also happens under the Light Rail + Redevelopment
scenario, but the increase is due more to economic growth caused by the Light Rail than to
redevelopment, since the redevelopment that occurs in Tier 1 in the Light Rail + Redevelopment
scenario essentially relocates the nonresidential development that took place in the Tier 2 donut in the
Light Rail scenario to Tier 1. In other words, the overall employment growth potential remains the same
in Tier 2 for both scenarios but Tier 1 becomes more attractive than the Tier 2 donut in the Light Rail
scenario.
For both Tier 1 and Tier 2, the total unemployed population (Figure 5-72) and the unemployment rate
(Figure 5-72) are reduced significantly in both the Light Rail and Light Rail + Redevelopment scenarios
due to resident employment increasing faster than the resident labor force (Figure 5-72). This is due in
part to the way the model calculates resident labor force by using a fixed share of the resident population
in the labor force. This share is based on historical data and is held constant between 2020 and 2040 at
52% for Tier 2 and 42% for Tier 1. By assuming that the percent of the population in the labor force
remains constant, the model does not allow for the possibility that more people might enter the labor
force if additional jobs are created, so it is possible that it overestimates the decline in unemployment
associated with a given increase in economic development. The decrease in the unemployment rate in
Tier 1 under the light rail scenarios causes an increase in net migration to Tier 1, which causes the
population in Tier 1 to grow significantly faster than the BAU under both scenarios (Figure 5-73). It is
difficult for the model to predict unemployment in Tier 1, because its small geographic area makes it
less certain how many of the new residents to Tier 1 would both live and work there, but the larger size
of Tier 2 makes it much more likely that the percent of the residential population in the labor force will
remain consistent with historical trends.
(A) Tier 1
(B) Tier 2
8.0%
6.0%
4.0%
2.0%
0.0%
20
1
,\
1 \
1 V^- -^ST •
^X^*L
/"
^
^^
^
00 2010 2020 2030
^^— Light Rail + Redevelopment — —Light Rail
^^BAU — — — U.S. Census
2010
2020
• Light Rail + Redevelopment
• BAU
2030
• Light Rail
• NC ESC
2040
Figure 5-72. Time Series of Unemployment Rate over the Entire Study Period (2000-2040)
135
-------
(A) Tier 1
(B) Tier 2
GRP per Capita Earnings per Capita Resident Net Total Population
Earnings per Capita
• BAU • Light Rail • Light Rail + Redev
GRP per Capita Earnings per Capita Resident Net Total Population
Earnings per Capita
• BAU • Light Rail • Light Rail + Redev
Figure 5-73. Percent Change in Per Capita Economic Growth Indicators between 2020 and 2040 for
Three Main Scenarios
Just as the unemployment rate in Tier 1 is harder to predict than in Tier 2, per capita economic growth
variables, shown in Figure 5-73, carry a great deal of uncertainty in Tier 1. In Tier 1, GRP per capita,
earnings per capita, and resident per capita net earnings are higher in the BAU scenario than in the Light
Rail and Light Rail + Redevelopment scenarios because employment grows much faster than population
in the BAU scenario.28 In the Light Rail scenario, the rail makes Tier 1 more attractive for businesses
and families, but without redevelopment to higher densities, nonresidential sq ft is capped, so population
continues to grow, but businesses do not, leading to lower growth for GRP and earnings per capita
between 2020 and 2040. In the Light Rail + Redevelopment scenario, redevelopment allows
nonresidential sq ft to increase in Tier 1 relative to the Light Rail and BAU scenarios, but this increase is
still slower than the growth in population caused by the light rail, leading to per capita economic growth
that is lower than in the BAU scenario. Resident per capita net earnings growth in Tier 1 is also highest
in the BAU scenario because this variable only represents the earnings of residents that are also working
in Tier 1, and the model assumes that the family members of new workers migrating to Tier 1 would
likely work outside of Tier 1, causing their income to be excluded from this variable. Another factor that
contributes to the lower per capita growth in earnings in the light rail scenarios is that the model assumes
that average earnings per job by category is not affected by economic growth. Therefore, even though
Tier 1 becomes more economically attractive and total employment increases under the light rail
scenarios, the average earnings per job in each employment category, which are exogenous lookup
tables with data and projections from Woods & Poole Economics Inc. (Copyright 2014), remain the
same for all three main scenarios.
In Tier 2, the per capita economic growth indicators, shown in Figure 5-73, demonstrate the increased
economic potential that the light rail scenarios bring to individuals living in the DCHC MPO region.
Relative to the BAU scenario, GRP, earnings, and resident net earnings per capita increase at a higher
rate under both light rail scenarios in Tier 2, though the Light Rail scenario increases slightly more than
28 BAU scenario employment growth in Tier 1 was calibrated to closely match projections to 2040 from the TRM v5 SE data,
which was based off of the "preferred growth" land use scenario that included additional employment in the light rail station
areas proposed under the 2040 MTP. These projections were the only employment projections available to us by TAZ at the
time that version 1.0 of the D-O LRP SD Model was created, but we subsequently received employment growth projections
to 2040 by TAZ under a "trend development" scenario, which more closely matches our BAU scenario, and we will be using
them to re-calibrate the BAU scenario in the 2.0 version of our model.
136
-------
the Light Rail + Redevelopment scenario because redevelopment causes population to increase more
than employment in Tier 2.
Property Tax Revenues
Real property (which includes land and buildings) has been rising in value steadily over the past 15
years for both Tiers, as shown in Figure 5-74. The D-O LRP SD Model was calibrated in the BAU
scenario to match this historical rise in real property values, which includes both residential (single-
family and multifamily) and nonresidential (not tax exempt) properties.29 In 2022, real property values
diverge in the light rail scenarios in anticipation of completion of the light rail and (in the Light Rail +
Redevelopment scenario) high-density redevelopment, as shown in Figure 5-74, with significant
increases in property values relative to the BAU scenario.
Figure 5-75 shows the absolute change in each category of real property value between 2020 and 2040
for the three main scenarios in Tier 1 and Tier 2. As the figure shows, the largest increase in property
values in Tier 1 occurs in nonresidential property value, with a $16 billion increase (2010 dollars) in the
Light Rail + Redevelopment scenario. In both Tiers, the greatest increase in each property value
category occur in the Light Rail + Redevelopment scenario.
(A) Tier 1
(B) Tier 2
2000 2010 2020 2030
Light Rail + Redev
— —Light Rail
^—BAU
— — — Durham and Orange County Property Tax Database
Light Rail + Redev
Light Rail
^—BAU
— — — Durham and Orange County AFP
Figure 5-74. Time Series between 2000 and 2040 of Total Real Property Value: Model Scenario Outputs
and Historical Data
29 For a detailed explanation of the factors influencing property values in the model, see the Land Use Sector Overview in
Chapters.
137
-------
(A) Tierl
20
18
16
14
12
10
8
6
4
2
Single-Family Mulitfamily Nonresidential Total Combined
• BAU • Light Rail • Light Rail + Redev
(B) Tier 2
20
10
Single-Family Mulitfamily Nonresidential Total Combined
• BALI • Light Rail • Light Rail + Redev
Figure 5-75.
Scenarios
Absolute Change in Real Property Value by Type between 2020 and 2040 for Three Main
The increase in real property values in both Tiers leads to significant increases in property tax revenues
collected between 2020 and 2040 for both city (City of Durham and Town of Chapel Hill only) and
county (Durham and Orange County) governments, as shown in Figure 5-76. In Tier 2, this results in a
cumulative increase in combined (city and county) property taxes collected between 2020 and 2040 of
$1.1 Billion and $1.2 Billion (2010 dollars) for the Light Rail and Light Rail + Redevelopment
scenarios, respectively. In Tier 1, these totals are $410 Million and $1.1 Billion (2010 dollars),
respectively. Per acre of developed land, an additional $28,000 per year (2010 dollars) in property tax
revenues are collected in Tier 1 in 2040 under the Light Rail + Redevelopment scenario, an increase of
225% over 2020 levels (Figure 5-76).
138
-------
(A) Tier 1
(B) Tier 2
Total County PT
Levied per Year
Total City PT Levied
per Year
Cumulative City + PT Revenues per Acre
County PT Levied of Developed Land
Total County PT
Levied per Year
Total City PT Levied Cumulative City + PT Revenues per Acre
per Year County PT Levied of Developed Land
• BAU • Light Rail • Light Rail + Redev
I BAU • Light Rail • Light Rail + Redev
Figure 5-76. Percent Change in Property Tax (PT) Revenues between 2020 and 2040 for Three Main
Scenarios
D-O LRP Budget
Figure 5-77 presents the model-estimated changes in D-O LRP revenues between 2020 and 2040,
disaggregated by revenue source. Vehicle registration fees (set at $10 per vehicle) are projected to
decrease during that period due to inflation rising faster than the vehicle stock. However, increases in
public transit fares and transit sales taxes (driven by increases in total retail consumption in both light
rail scenarios) lead to a net increase in combined D-O LRP revenues collected per year that is larger in
both light rail scenarios (81% and 82%, respectively) than in the BAU scenario (71%) during that same
period. Rental car tax revenues aren't shown in Figure 5-77 because they are the same for all three main
scenarios.
Vehicle Registration
Fees
Half Cent Transit Sales Combined LRP
Public Transit Fare Tax Revenues
-$740,000 per year |
• BALI • Light Rail • Light Rail + Redev
Figure 5-77. Percent Change in D-O LRP Revenue Sources between 2020 AND 2040 for Three Main
Scenarios
Figure 5-78 compares D-O LRP SD Model outputs for the three main scenarios with two local
projections for the cumulative D-O LRP revenues collected between 2013 and 2040 in nominal or year-
of-expenditure dollars (not inflation-adjusted). The "MIN" projection is from the Durham and Orange
County bus and rail investment plans and assumes an annual growth rate in the half cent sales tax
revenues collected for transit services of 3.0% in Durham County and 3.5% in Orange County (DCHC
MPO et al. 2011, Triangle Transit et al. 2012). The "MAX" projection assumes annual growth rates of
4.65% in Durham County and 4.4% in Orange County (CAMPO and DCHC MPO 2013). As the figure
139
-------
shows, all three main scenarios for the D-O LRP SD Model project cumulative D-O LRP revenues that
far exceed local projections, with the cumulative divergence exceeding $1.2 Billion by 2040. This
difference is because the local projections assume much lower annual growth rates in retail sales than
the D-O LRP Model, which is calibrated in the BAU scenario to closely match retail sales projections
from Woods & Poole Economics Inc. (Copyright 2014).30
2015
2020
2025
2030
2035
-Light Rail + Redev
-Light Rail
-BAU
-MAX Projected Half Cent Sales Tax Growth Rate (4.65% Durham Co.; 4.4% Orange Co.)
MIN Projected Half Cent Sales Tax Growth Rate (3.5% Durham Co.; 3.6% Orange Co.)
Figure 5-78. Cumulative Nominal D-O LRP Revenues between 2010 and 2040, Three Main Model
Scenarios vs. Local Projections
Figure 5-79 presents model estimates for two D-O LRP budget scenarios. The first scenario (shown in
Figure 5-79) assumes that 25% of the total capital cost to build the light rail will be paid for by local
revenue sources, while the second scenario (shown in Figure 5-79) assumes that 50% of the capital costs
will be paid for locally. Both scenarios assume that 100% of the light rail's operations and maintenance
(O&M) costs will be paid for locally. The reason for this comparison is to show how the D-O LRP
budget would be affected if the State of NC did not cover their expected share of the light rail capital
cost (25%). The figures show the net D-O LRP budget (blue solid line) in each scenario, calculated by
subtracting the cumulative local expenditure for the light rail (capital + O&M, red dotted line) from the
cumulative D-O LRP revenues collected in the Light Rail + Redevelopment scenario (green dotted line).
In the second scenario, the budget becomes negative between 2022 and 2033, an indicator that
additional revenues or funding sources would be necessary if local funding sources were responsible for
50% of the light rail's capital costs.
30 Woods & Poole does not guarantee the accuracy of these data. The use of their data and the conclusions drawn from it are
solely the responsibility of the US EPA.
140
-------
D-O LRP Budget with
Capital Cost = 25% Local
(A) D-O LRP Budget with
Capital Cost = 50% Local
on USD 2010
D O O ^ -
i cf- bo b f-
15
0.2
n n
^ f
.-x^
^ ^^^^"^
^
**»
^
^-^
X
X
X
_ .
^
X
X
X
^^
-s^
2015 2020 2025 2030 2035 2C
— Light Rail + Redev Cumulative LRP Revenues
—Cumulative Local LRT Expenditure: 25S; Capital Cost + OSM
— Net LRP Budget
2020
2025
2030
2035
• Light Rail + Redev Cumulative LRP Revenues
• Cumulative Local LRT Expenditure: 50% Capital Cost + 0&M
-Net LRP Budget
Figure 5-79. D-O LRP Revenues, LRT Costs, and Net Budget between 2010 and 2040.
141
-------
Higher Tier 2 Density Scenario
In this scenario, which is run on top of the Light Rail + Redevelopment scenario, density in Tier 2
increases starting in 2020 (when redevelopment begins in Tier 1), reaching a level 6.5% higher
than the reference scenario by 2040 (Figure 5-80A). This causes an increase in nonresidential sq ft
approximately equal to the increase in nonresidential sq ft caused by redevelopment in Tier 1. In
this scenario, growth in the Tier 2 economy exceeds the growth caused by the light rail, with all
four economic indicators in Figure 5-80B increasing more between 2020 and 2040 than in the
reference scenario, though two of these indicators (earnings per capita and resident per capita net
earnings) increase at the same rate in this scenario as in the Light Rail scenario.
(A)
(B)
100.0%
+$21,000 per +iU,000 per +8,500 per
person person person
mi mi mi
2020
2025
• Higher Tier 2 Density
• Light Rail
2030
2035
• Light Rail + Redev
• BALI
2040
GRP GRP per Capita Earnings per Resident per
Capita Capita Net
Earnings
• BALI • Light Rail • Light Rail + Redev • Higher Tier 2 Density
Figure 5-80. (A) Total Nonresidential Sq. Ft. - Tier 2 and (B) Economic Indicators - Tier 2
142
-------
Higher Rent for Nonresidential Land Scenario
In this scenario, run on top of the Light Rail + Redevelopment scenario, we decrease gross
operating surplus (GOS) per sq ft by $5 each year from 2025 to 2040 to simulate the economic
consequences of higher property values leading to higher rent that is not being offset by increases
in profits. This reduction in GOS per sq ft has profound impacts on the economy, reducing the
overall GOS per year to levels below BAU (Figure 5-81), essentially offsetting all of the economic
gains from the light rail and redevelopment. Between 2020 and 2040, the Higher Rent scenario
GRP per year increases by only 62% compared to 83% in the Light Rail + Redevelopment
scenario, and the smaller increase in employment leads to 50,000 fewer jobs in 2040.
(A)
(B)
2020
• Higher Rent'
• Light Rail* Redev •
• Light Rail-
• BAU
GOS GRP Employment
• BAU • Light Rail • Light Rail + Redev • Higher Rent
Figure 5-81. (A) Gross Operating Surplus Tier 2 and (B) Change (%) in Economic indicators -
Higher Rent Compared to Three Main Scenarios
143
-------
Equity
Though the two main scenarios do not directly affect the equity sector, the changes they produce lead to
large indirect impacts, partially due to the balancing feedback loop that connects the economy, equity,
and transportation sectors (see Chapter 3). Through this feedback loop, a decline in the unemployment
rate reduces the percent of the population in poverty. A reduction in the overall number of households in
poverty reduces the number of zero-car households (a proxy for transit-dependent households), which in
turn increases VMT and congestion. Increased congestion affects the economy sector by reducing
productivity, grp, and employment, completing the loop by increasing unemployment. Under both light
rail scenarios, unemployment rates decline significantly, leading to declines in the percent of the
population in poverty in Tier 1 between 2020 and 2040 of 8% in the Light Rail scenario and 18% in the
Light Rail + Redevelopment scenario (Figure 5-82). In Tier 2, the percent of the population in poverty
declines by 29% between 2020 and 2040 in both light rail scenarios (Figure 5-82). The impact on the
percent of households in zero-car households is similar, with a 30% drop in Tier 2 under both light rail
scenarios, and a 26% and 27% decline in Tier 1 between 2020 to 2040 under the Light Rail and Light
Rail + Redevelopment scenarios, respectively.
(A) Tier 1
(B) Tier 2
Percent of Population Percent of Households Percent of Population Percent of Households
Unemployment Rate in Poverty with Zero Cars Unemployment Rate in Poverty with Zero Cars
0%
-10%
0% points!
IBAU • Light Rail • Light Rail + Redev
I BAU • Light Rail • Light Rail + Redev
Figure 5-82. Percent Change in Equity Outcomes between 2020 and 2040 for Three Main Scenarios
The majority of the light rail scenario's impacts in the equity sector are in the sector's output indicators,
rather than variables which affect other sectors through feedback loops. The primary indicator variables
in this sector are property values, the percent of the population in poverty, the transit-dependent
population, the households in poverty at risk of displacement, and the housing + transportation (H+T)
affordability index, which is calculated in different ways for the average household and for those in
poverty.
In Tier 2, all three categories of property values respond similarly to the main scenarios between 2020
and 2040, with a moderate increase in value over BAU in the Light Rail scenario, and a small additional
increase with the Light Rail + Redevelopment scenario. Of the three categories, the largest change in
Tier 2 is in single-family property values in the Light Rail + Redevelopment scenario, with per dwelling
unit values increasing by 95% (Figure 5-82, all values are in 2010 dollars and are inflation adjusted). In
the Light Rail + Redevelopment scenario, multifamily property values see a larger increase relative to
the BAU scenario, however, with an increase 43% larger than the increase in BAU. This is because the
value of multifamily homes is more strongly affected by retail density, which grows faster under the
light rail scenarios. Tier 1 has more dramatic changes both between scenarios and between land use
categories. Nonresidential unit property values (expressed per sq ft rather than per dwelling unit for
144
-------
residential property values) in Tier 1 see the largest change, with a 207% increase under the Light Rail +
Redevelopment scenario (Figure 5-82). This is primarily due to the almost tripling of density for the
20% of land that gets redeveloped, which leads to a large increase in building size, one of the factors
linked to nonresidential property values. In the Light Rail scenario, however, nonresidential unit
property values actually show a smaller increase than under the BAU scenario, with increases of 71%
and 77%, respectively. This smaller growth, which is reflected among multifamily property values as
well, is due to the drop in retail density that occurs in Tier 1 once all available land is built out at the
allowable density and development of nonresidential sq ft stops (property values are higher than BAU
between about 2020 and 2035 but lower from 2035 to 2040). The light rail has very little net impact on
single-family unit property values, relative to BAU, with unit property values increasing by slightly less
than BAU in the Light Rail + Redevelopment scenario. This is due to the fact that factors that increase
single-family property values relative to BAU in these scenarios (e.g., population growth and retail
density) are offset by factors that decrease value (e.g., decreasing lot sizes, relatively more vacant land
in the Light Rail + Redevelopment scenario).
(A) Tier 1
(B) Tier 2
200%
Single Family Property
Value per DU
Multifamily Property
Value per DU
Nonresidential property
value per Sq Ft
Single Family Property
Value per DU
Multifamily Property
Value per DU
Nonresidential property
value per Sq Ft
I BAU • Light Rail • Light Rail + Redev
I BALI • Light Rail • Light Rail + Redev
Figure 5-83. Change (%) in Property Values between 2020 and 2040 for Three Main Scenarios
The overall percent change values shown in Figure 5-83 do not tell the whole story of how property
values change in each scenario over time. Multifamily property values in Tier 1 in particular follow a
nonlinear trend (see Figure 5-83). In the Light Rail scenario, multifamily property values first increase
to levels higher than the other two scenarios, and then decline after 2030. As described in Chapter 3,
property values in the model are calculated by several different variables, and the changing trajectory of
multifamily property values in this scenario is the result of different variables and feedback loops
dominating during different periods. At first, the steep decline in available land is the dominant factor,
initially pushing multifamily unit property value to levels 1.3% above values in the Light Rail +
Redevelopment scenario and 8.1% above the BAU scenario, between 2025 and 2030. Starting around
2030, however, the maximum impact of limited land availability is reached, and other factors become
more dominant. Slower growth in relative incomes, commercial building size, and job density all
contribute, but the most important factor becomes retail density per capita, which declines starting in
2033, as population continues to increase while nonresidential development is stopped by the land cap.
Ultimately, multifamily property values are 7.3% lower in the Light Rail scenario than in the BAU
scenario in 2040. Without the effects of the land cap, property values under the Light Rail +
Redevelopment scenario follow a steadier increase as land is developed more slowly and retail density
per capita grows steadily, reaching values 6.5% higher than the BAU scenario in 2040.
145
-------
(A) Multifamily
(B) Single-Family
2000
2010
• Light Rail+ Redev
2020
2030
2040
• Light Rail
•BALI
2000
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
• BALI
Figure 5-84. Multifamily and Single-Family Unit Property Values - Tier 1
The changes in multifamily property values affect renter costs, which is one of the two affordability
measures tracked in the model (Figure 5-84).31 In Tier 1, annual renter costs grow the most between
2020 and 2040 in the BAU scenario, but because the increase accelerates earlier in the Light Rail +
Redevelopment scenario, that scenario actually has the highest cumulative costs per household over that
period ($254,000 versus $250,000 in the BAU scenario). The Light Rail scenario shows the lowest
increase in renter costs and therefore the lowest increase in overall housing and transportation costs
(14%). Though vacancy rates and GRP growth rates also affect estimates of annual renter costs, this
discrepancy is driven primarily by the differences in multifamily property values described above,
meaning that renter costs in the Light Rail scenario exceed costs in the Light Rail + Redevelopment
scenario between 2024 and 2030 but end up below BAU costs by 2040. By 2040, annual renter costs are
forecasted to grow by 28%, 18% and 27% under the BAU, Light Rail, and Light Rail + Redevelopment
scenarios, respectively, with the lowest cumulative costs in the Light Rail scenario. Conversely, in Tier
2, the increases in both renter and transportation costs are highest in the Light Rail scenario, though the
magnitude of cost increases in all scenarios is much smaller (e.g., 6% increase in renter costs in Tier 2
versus 18% in Tier 1 in the Light Rail scenario). The Light Rail scenario has the highest increase in
renter costs in Tier 2 because multifamily property values consistently rise, rather than falling as
described above in Tier 1.
In Tier 1, transportation costs in the two light rail scenarios show a larger increase than in the BAU
scenario primarily due to the expected increase in money spent on transit fares. Vehicle related costs per
capita drop below BAU to a low around 2027, though by 2040 they are higher than in the BAU scenario.
The initial per capita drop is due to a drop in VMT per capita with the introduction of the light rail. The
eventual increase in vehicle-related costs is due to two primary factors: 1) higher vehicle ownership per
capita due to the drop in the transit-dependent population, and 2) increased parking costs (though this is
a much smaller share of total vehicle costs). The extra increase in transportation costs under the Light
Rail + Redevelopment scenario is largely due to higher parking costs, which is driven by changes in job
density. In Tier 2, the increase in transportation costs is almost equal across all three scenarios because
vehicle related costs per capita do not change significantly relative to the BAU and the increase in
money spent on transit fares is negligible compared to spending for vehicles.
31 Single-family property values were not considered as a part of affordability.
146
-------
(A) Tier 1
(B) Tier 2
10%
Annual Renter Costs
Annual Transportation
Costs
Transportation and
Renter Costs
Annual Renter Costs
Annual Transportation
Costs
Transportation and
Renter Costs
• BAU • Light Rail • Light Rail + Redev
I BAU • Light Rail • Light Rail + Redev
Figure 5-85. Percent Change in Housing and Transportation Costs between 2020 and 2040 for Three
Main Scenarios
Transportation and renter costs are used in the calculation of two affordability indices in the model. The
primary affordability index (for average households, as described in Chapter 3.3 in the Equity sub-
section) shows improvement under every scenario in Tier 2 between 2020 and 2040 (Figure 5-86), but
especially under the light rail scenarios (5% increases vs. 2% increases under the BAU), due to
increased earnings and relatively flat housing and transportation costs. In Tier 1, this measure changes
significantly only in the Light Rail + Redevelopment scenario (where it declines by 5%), due to
relatively slow per-capita earnings that do not increase fast enough to keep up with increases in housing
and transportation costs (which are equivalent to the increases in the BAU scenario).32 In contrast, the
affordability index for households in poverty worsens under every scenario and Tier. This is to be
expected, since the index is essentially the inverse of patterns in overall housing and transportation
costs; moderate increases in these costs in Tier 2 lead to moderate drops in affordability (roughly 5% in
all scenarios), and larger increases in costs in Tier 1 lead to more pronounced affordability drops, of
15%, 11%, and 15% in the BAU, Light Rail, and Light Rail + Redevelopment scenarios, respectively.
Affordability Index Affordability Index
for Households in Affordability Index - for Households in Affordability Index-
Poverty - Tier 1 Tier 1 Poverty - Tier 2 Tier 2
I Light Rail + Redev • Light Rail • BAU
Figure 5-86. Percent Change in Affordability Indices between 2020 and 2040 for Three Main Scenarios
Finally, the model's estimates of the percent of the population in poverty suggest that the risks of
32 The percent of the population in the labor force in Tier 1 declines given the assumption that when workers move to the area
to fill a job opening, they bring with them families and spouses who may still work outside the Tier, resulting in more
families living in the area. Thus per capita earnings decline slightly, even though per household earnings do not.
147
-------
gentrification vary between the two Tiers in the three main scenarios. In Tier 2, the decline in the
percent of the population in poverty under the light rail scenarios leads to a corresponding decline in the
number of households at risk of displacement (defined as the number of households in poverty not
accommodated in subsidized dwelling units). In Tier 1 however, the decline in the percent of the
population in poverty is offset by the rise in population, leading to an overall rise in the number of
households in poverty, and an increase in the number of households at risk of displacement. Figure 5-86
shows that, at current levels of subsidized dwelling units, the number of households at risk of
displacement will surpass the BAU scenario starting in 2024 in the Light Rail scenario and 2031 in the
Light Rail + Redevelopment scenario, reaching 4,800 households in risk of displacement in 2040 under
the Light Rail scenario, a level 18% higher than the BAU.
(A) Tier 1
(B) Tier 2
o
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
• BAU
A
/
y /
r
""">^— ***
-^
1^
"^^^^
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
• BAU
Figure 5-87. Households in Poverty at Risk of Displacement
148
-------
Alternative Policy Scenario: Fewer Organically Affordable Units Scenario
This scenario, which is run on top of the Light Rail + Redevelopment scenario, is meant to explore
the uncertainty around the percent of multifamily units that will be organically affordable (defined
as the number of dwelling units that cost no more than 40% of the poverty threshold income) in the
future, since it is unlikely that the percent of units that are affordable will remain constant (as the
model assumes), particularly in the face of the light rail and redevelopment. In this scenario, the
percent of multifamily units that are assumed to be organically affordable in Tier 1 declines
gradually from 26% in 2020 (the value derived from data for 2014) to 15% by 2040, with the
decline starting in anticipation of the light rail. With this change, the absolute number of dwelling
units still grows, but the portion that are affordable declines. This altered assumption directly leads
to a rise in the housing gap for households in poverty, which is defined as the number of
households neither accommodated in subsidized dwelling units nor in organically affordable
dwelling units. The housing gap is initially negative under the Light Rail + Redevelopment
scenario, only becoming positive between 2016 and 2025 before becoming negative again as
poverty levels fall (Figure 5-88). In contrast, under this scenario, the housing gap continues to rise,
reaching a high of about 1,400 households in 2040. On the other hand, in order to prevent the
housing gap from exceeding zero in the Light Rail + Redevelopment scenario, the percent of
multifamily units that are affordable would need to rise from 26% to 27% by 2020.
2040
-Fewer Organically Affordable ^^—Light Rail + Redevelopment
Figure 5-88. Housing Gap for Households in Poverty - Tier 1
149
-------
Alternative Policy Scenario: More Multifamily Households Scenario
This scenario, which is run on top of the Light Rail + Redevelopment scenario, extends the
demographic shift towards more households living in multifamily dwelling units in Tier 1.
Extending the linear trend in the percent of single-family households between 2000 and 2010 to
2040 results in a 10.3% increase in the portion of households that live in multifamily dwelling units
by the year 2040, relative to the reference scenario. As a result of this shift, multifamily acres grow
by 30% between 2020 and 2040 (relative to a 20% increase in the reference scenario), while single-
family acres decline by 5% (relative to a 21% increase in the reference scenario) (Figure 5-89).
This leads to a drop in total developed land (and correspondingly, residential impervious surface),
falling to a level 2.5% below the reference scenario by 2040 (Figure 5-90). Residential energy use
also declines over this period, due to a lower assumed energy use intensity for multifamily
households, with 31% growth, compared to 38% growth in the reference scenario. The increase in
multifamily dwelling units provides more organically affordable units (since the percent that is
organically affordable is assumed to remain constant), causing the housing gap to disappear, and
leading the potential population in poverty displaced to remain at zero for the entire simulation
(Figure 5-91).
Mjltifamily acres Tier 1, Wore WFHHs
I Single family acres Tier!, Wore WFHHs
-Mjltifamily acres Tier 1, Li^Tt Rail + Redevebprrert
•Single family acres Tier 1, Lght Rail + Redevebprrent
Figure 5-89. Multifamily and Single-Family Acres Under Light Rail + Redevelopment and More
Multifamily Households Scenarios - Tier 1
Developed Land Residential Impervious Residential Energy Use
Surface
• Light Rail + Redevelopment BMore Multifamily Households
Figure 5-90. Percent Change in Developed Land, Residential Impervious Surface, and Residential
Energy Use Between 2020 and 2040 for More Multifamily Households Scenario Compared to Light
Rail + Redevelopment - Tier 1
150
-------
120
100
80
60
40
20
2000
2010
2020
2030
2040
•More Multifamily Households
•Light Rail + Redevelopment
Figure 5-91. Potential Population in Poverty Displaced - Tier 1
151
-------
Retail Wage Increase Scenario
In this scenario, which is run on top of the Light Rail + Redevelopment scenario, retail earnings per
employee in Tier 1 experience an increase in the nominal hourly retail wage of $2.00 in 2016, with
additional $1.00/hour increases added over the following three years (2017-2019), on top of the
yearly increase already projected by Woods & Poole.
As a result, total retail earnings in Tier 1 (Figure 5-92) increase by 23% more than the Light Rail +
Redevelopment scenario by 2040 (note that the percentage difference between this scenario and the
reference scenario actually peaks at 30% in 2019 but decreases over time due to inflation). This
increase in retail wages lead to an increase in GRP in Tier 1 (Figure 5-92) that starts at 1.0% in 2019
and grows to 8.4% in 2040.
(A)
(B)
-Retail Wage Increase
-Light Rail + Redev
— Retail Wage Increase
-Light Rail + Rede
Figure 5-92. (a) Retail Earnings Tier 1 and (b) GRP Tier 1 - Retail Wage Increase Scenario vs. Light
Rail + Redevelopment Scenario
The retail wage increase also has an immediate impact on resident per capita net earnings Tier 1
(Figure 5-93) and the affordability index in Tier 1, shown in Figure 5-93. Since the retail wage
increase is nominal, its value depreciates over time, thus affordability peaks in 2019 at 1.2 under the
Retail Wage Increase scenario, a 28% increase from the affordability index in Tier 1 in 2020 under
the Light Rail + Redevelopment scenario.
(A)
(B)
,_ 1,400
2 1,200
fc 1,000
o 800
8 600
a 400
200
^>^~~
£-
-
.
•— ^•^•"•-™
— — —
15 2020 2025 2030 2035 20
Retail Wage Increase Light Rail + Redev
• Retail Wage Increase
-Light Rail + Redev
Figure 5-93. (a) Resident Per Capita Net Retail Earnings Tier 1 and (B) Affordability Index Tier 1 -
Retail Wage Increase Scenario vs. Light Rail + Redevelopment Scenario
152
-------
Water
Municipal Water Demand and Supply
Figure 5-94 shows the modeled estimates of water demand in both Tiers for the three main scenarios, for
the period 2000-2040. Municipal water demand dropped in both Tiers prior to 2010 due to conservation
practices that began during the 2007 drought and continued afterward. In Tier 1, the model projects
daily water demand in 2040 that is 1.2 and 1.9 Mgal/day higher than BAU in the Light Rail and Light
Rail + Redevelopment scenarios, respectively (representing increases of 17% and 27% over the 2040
BAU value of 6.9 Mgal/day). Light rail scenarios increase water demand due to increases in dwelling
units, employment, and population, which drive the residential, nonresidential, and nonrevenue
components of water demand, respectively. In Tier 1, dwelling units increase by 44% and 57% in the
light rail scenarios between 2020 and 2040 (compared to 14% in BAU), while employment increases by
53% and 66% (compared to 35% in BAU). Nonrevenue demand, which is tied to changes in population,
increases 40% and 48% in the light rail scenarios between 2020 and 2040, compared to 14% in BAU. In
the Light Rail scenario, water demand increases more slowly after 2035 due to land scarcity, which
reduces residential construction and also slows down Tier 1 GRP growth, including employment growth
(which affects nonresidential demand).
(A) Tier 1
(B) Tier 2
2000
2010
2020
• Light Rail + Redevelopment
2030
• Light Rail
2040
•BAU
2000
2010
2020
• Light Rail + Redevelopment
2030
• Light Rail
2040
•BAU
Figure 5-94. Water Demand: Light Rail, Light Rail + Redevelopment Scenarios Compared to BAU
In Tier 2, the changes in the light rail scenarios also increase water demand relative to BAU, but
proportionately less than in Tier 1, and with less difference between the two light rail scenarios. In 2040,
water demand is 2.5 and 3.1 Mgal/day higher than BAU in the Light Rail and Light Rail +
Redevelopment scenarios, respectively (Figure 5-94), representing increases of 4.1% and 5.0% over the
2040 BAU value of 62 Mgal/day. As in Tier 1, development of the light rail stimulates water demand
due to increases in dwelling units, employment, and population. Between 2020 and 2040, dwelling units
and population both increase by 42-44% in the light rail scenarios compared to 38% in BAU, while
employment increases 50-51% in the light rail scenarios, compared to 41% in BAU.
153
-------
(A) 2000 (B) 2040, BAU
5.6 Mgal/day total demand 6.9 Mgal/day total demand
(C) 2040, LIGHT RAIL AND LIGHT RAIL + REDEVELOPMENT
8.1 and 8.7 Mgal/day total demand, respectively
i MF residential
i SF residential
i nonresidential
nonrevenue
Figure 5-95. Tier 1 Water Demand (MGAL Per Day) in (A) 2000 and (B) 2040, BAU Scenario Compared to
Light Rail Scenarios
In the BAU scenario, total Tier 1 water demand grows from 5.6 Mgal/day to 6.9 Mgal/day between 2000
and 2040 (Figure 5-95 A-B), representing an increase of 23%. During this time, nonresidential
(commercial and industrial) facilities remain the largest portion of water demand, increasing from 50%
of total demand in 2000 to 66% in 2040. Though the residential share of water demand decreases during
this period, the shift from single-family to multifamily dwelling units in Tier 1 is also reflected in the
distribution of water demand, with single-family residential demand decreasing from about 44% of
residential demand in 2000 to about 38% in 2040. Nonrevenue demand remains less than 10% of total
demand between 2000 and 2040.
The two light rail scenarios change the proportions of water demand only slightly, with slight increases
in the multifamily residential and nonrevenue shares of total water demand (Figure 5-95C). These shifts
are due to the light rail causing a proportionately larger increase in MF dwelling units than in
employment. Relative to the BAU change from 2000-2040, MF dwelling units increase by 26% and
36% more than BAU in the Light Rail and Light Rail + Redevelopment scenarios, respectively, while
total employment increases by 13% and 24% more than BAU.
154
-------
(A)2000
45 Mgal/day total demand
(B)2040
BAU: 62 Mgal/day total demand
Light Rail: 64 Mgal/day
Light Rail + Redevelopment: 65 Mgal/day
i MF residential
i SF residential
i nonresidential
nonrevenue
Figure 5-96. Tier 2 Water Demand (MGAL Per Day) in (A) 2000 and (B) 2040, BAU Scenario Compared to
Light Rail Scenarios
Tier 2 water demand grows from 45 Mgal/day to 62 Mgal/day between 2000 and 2040 in the BAU
scenario, representing an increase of 38%, a larger percent increase than in Tier 1 (Figure 5-96). Unlike
Tier 1 water demand, which is mostly nonresidential, the majority of Tier 2 water demand is residential,
with most of that demand coming from single-family homes, as opposed to multifamily homes using
more water in Tier 1. In 2040, nonresidential demand represents a larger share of total demand in all
three scenarios (growing from 28% to 36%, as shown in Figure 5-96), because water use intensity drops
more for residential than for nonresidential users over this period, even though dwelling units grow
faster than employment in Tier 2 (100% vs. 87% growth between 2000 and 2040). Residential water use
per dwelling unit drops by 55% between 2000 and 2040, while nonresidential water use per employee
drops by 10%. Despite increasing total water demand relative to BAU, the two light rail scenarios have
no effect on the distribution of water demand by use (Figure 5-96B), because the drivers of demand
(numbers of dwelling units and employed people) grow at roughly equal rates among the three main
scenarios.
350
200
2000
2010
•Light Rail + Redev
2040
•Light Rail
•BALI
Figure 5-97. Days of Supply in Durham County Water Reservoirs: Three Main Scenarios
155
-------
In the absence of supply augmentation, water supply in Durham County's combined water reservoirs is
projected to drop in the BAU scenario from 250 days of supply in 2000 to 180 days in 2040 (Figure
5-97), due to growing water demand from population and employment growth. Further residential and
employment growth in the light rail scenarios reduce days of supply by 7 and 9 days in 2040 relative to
BAU in the Light Rail and Light Rail + Redevelopment scenario, respectively.33
Stormwater N Load
(A) Tier 1
(B) Tier 2
i N load open space Tier 1
N load roads Tier 1
IN load nonresidential Tier 1
IN load sf residential Tier 1
• N load mf residential Tier 1
• N load agricultural
IN load open space
N load roads
IN load nonresidential
IN load sf residential
IN load mf residential
Figure 5-98. Stormwater N Load by Land Use: BAU Scenario
Figure 5-98 presents the model's BAU projections of Stormwater load for five different land use types
(open space, roads, nonresidential, single-family residential, and multifamily residential). As shown in
Figure 5-98 A, in Tier 1, nonresidential land (gray area) has the largest Stormwater N load, with about
41% of total load by 2040, while roads (yellow) represent 28% of the total, and multifamily residential
(light blue) represents 18% of the total. Nonresidential land also shows the largest growth in Stormwater
load over time, increasing from 32% of total load in 2000 to 41% by 2040. This disproportionate
increase in Stormwater N from nonresidential land reflects growth in nonresidential land, due to
employment growing more quickly than population in this Tier under the BAU scenario. This result is
consistent with the decline in the jobs-housing balance in Tier 1 (mentioned in the Land Use section),
which also results from nonresidential land use growing faster than residential land use.
As shown in Figure 5-98B, single-family residential land (orange area) has the largest Stormwater N
load in Tier 2, with about 43% of total load by 2040, while nonresidential land (gray) contributes 21%,
and roads (yellow) contribute 15% of the total. Single-family residential land shows a significant
increase in N load, growing from 29% of total loads in 2000 (the same as loads from open space) to 43%
by 2040, as open space is primarily converted to single-family residential use.
33 The model's projection of days of supply is a rough estimate, as it assumes constant rainfall (45 inches/year) and constant
reservoir supply (7,568 Mgal, based on March 2015 data). Further, the model assumes that Durham County demand remains
at 67% of Tier 2 water demand, based on the historical ratio of Durham County to Tier 2 population.
156
-------
In Tier 1, the Light Rail and Light Rail + Redevelopment scenarios increase stormwater N load relative
to BAU by 6% and 4%, respectively, in 2040 (Figure 5-99A). In the Light Rail scenario, stormwater N
load increases rapidly starting in about 2022 and levels off by 2035, while in the Light Rail +
Redevelopment scenario, it shows gradual, continual growth through 2040. The nonlinear growth
pattern in Stormwater N load in Tier 1 largely reflects the growth pattern for impervious surface (Figure
5-100A), which increases by 930 acres and 830 acres in Light Rail and Light Rail + Redevelopment,
respectively, between 2020 and 2040 (compared to an increase of 650 acres in the BAU scenario). These
changes in impervious surface in turn reflect underlying changes in land development (reproduced in
Figure 5-1OOB), which reflect the effects of the developed land cap in 2033 in the Light Rail scenario,
but not in the Light Rail + Redevelopment scenario, which shows more gradual, but steady growth. In
Tier 2, stormwater N load is about 2% higher than BAU in Light Rail and Light Rail + Redevelopment
in 2040 (Figure 5-99B). As in Tier 1, this reflects growth in impervious surface due to land
development. Stormwater P load follows many of the same patterns as stormwater N load, since both are
driven by increases in land development and impervious surface. In Tier 1, the Light Rail and Light
Rail + Redevelopment scenarios increase stormwater P load by 4.9% and 3.2%, respectively, relative to
BAU in 2040 (Figure 5-101 A). In Tier 2, stormwater P load is 1-2% higher in the two light rail
scenarios compared to BAU in 2040 (Figure 5-10IB). Annual stormwater P load is about one-fifth of
annual stormwater N load.
(A) Tier 1
(B) Tier 2
1.6
1.4
=, 1-2
2L 1
3 0.8
| 0.6
^ 0.4
2000
2010
2020
2030
2040
2010
2020
2030
2040
Light Rail+Redev Light Rail BALI Light Rail+Redev Light Rail BAU
Figure 5-99. Stormwater N Load: Light Rail, Light Rail + Redevelopment Scenarios Compared to BAU
(A)
(B)
6000
5000
4000
3000
2000
1000
2000 2010
Light Rail+Redev
2020 2030
— —Light Rail
2040
-BAU
• Light Rail+Redev
• Light Rail
Figure 5-100. (A) Total Impervious Surface - Tier 1 and (B) Total Developed Land - Tier 1: Light Rail,
Light Rail + Redevelopment Scenarios Compared to BAU
157
-------
(A) Tier 1
(B) Tier 2
I 4
2
300
| 250
S.200
£
- 150
ro
3 100
£ 50
2000
2010
• Light Rail+ Redev
2020
2030
2040
2000
• Light Rail
•BALI
2010
• Light Rail + Redev
2020
2030
2040
• Light Rail
•BALI
Figure 5-101. Stormwater P Load: Light Rail, Light Rail + Redevelopment Scenarios Compared to BAD
(A) Tier 1
i Light Rail + Redev 2015 • Light Rail + Redev 2040
(B) Tier 2
(C) Percent Impervious Surface
h II., II
ll
• Light Rail + Redev 2015 • Light Rail + Redev 2040
• Tier 1 BTier 2
Figure 5-102. Annual Stormwater Runoff Volume and Percent Impervious Surface, Modeled for 2015
In Tier 1, nonresidential land has the highest Stormwater runoff, at nearly six billion L/year in 2015,
growing to nearly nine billion L/year by 2040 (Figure 5-102A). There is no agricultural land in Tier 1,
so it has the lowest (zero) runoff volume at all years. Tier 2 Stormwater runoff volume is roughly ten
times that of Tier 1, and its largest components are single-family residential, nonresidential, and open
158
-------
space (Figure 5-102B). As open space is converted to developed land, open space runoff volume
decreases between 2015 and 2040, while SF residential and nonresidential land have 52 billion and 40
billion L/year more runoff by 2040, respectively. Since rainfall volume is assumed to be constant over
the region, differences in storm water runoff volume by land use are driven by land area and percent
impervious cover. Like runoff volume in Tier 1, impervious surface by land use is lowest for agricultural
land and highest for nonresidential land, where it reaches over 80% in 2015 (Figure 5-102C). With the
exception of agricultural land and open space, all land uses had higher impervious surface in Tier 1 than
in Tier 2, so that aggregate impervious surface is 43% of Tier 1 land area and 14% of Tier 2 land area in
2015. As impervious cover changes, the model calculates new EMCs for each land use type (see graph
of stormwater N EMCs in Chapter 3) and multiplies them by runoff loading to calculate stormwater N
loading. Modeled EMCs change by less than 2% between 2000 and 2040.
159
-------
Stormwater Management Scenarios
These scenarios, which are run on top of the Light Rail + Redevelopment scenario, simulate
compliance with rules designed to protect water quality in Jordan Lake and Falls Lake. Accordingly,
they reduce stormwater N loads from developed land to reflect (1) 30% onsite stormwater treatment
in new developments, (2) 40% onsite stormwater treatment in new developments, or (3) 30% onsite
stormwater treatment in new developments plus 15% onsite stormwater treatment in existing
developments. For the purposes of this scenario, new development is defined to be everything
developed after 2015. In the Light Rail + Redevelopment scenario, a 30% reduction in stormwater N
from new development equals a decrease in N load of 3,500 Ib/year by 2040 (yellow line, Figure
5-103A), a 40% reduction causes an additional 1,200 Ib/year decrease (orange line), and a 30%
reduction in new development stormwater N with a 15% reduction in existing development
stormwater N keeps the stormwater N load at roughly 2015 levels (51,000 Ibs/year) through 2040
(light blue line). For reference, the orange line in Figure 5-103B shows the projected stormwater N
load from a 2.2 Ib/acre/year target for new development (Falls Lake and Jordan Lake development
rules).
(A)
(B)
30
2010
2020
2030
-30pct new 1 5pct existing N load treated LRT+redev
-40pct new N load treated LRT+redev
30pct new N load treated LRT+redev
-Light Rail+Redev
-New N load treated to 2.2 Ib/ac/y LRT + redev
-Light Rail+Redev
Figure 5-103.
Ib/acre/year
Stormwater N Load - Tier 1: (A) Percent Reduction Targets and (B) Reduction to 2.2
As a scoping estimate, if the 30% reduction in stormwater N from new development were
accomplished with bioretention (rain gardens), it would cost $510,000/year in 2020 and $3.8
million/year in 2040, in USD 2010. A 40% reduction in new development stormwater N would cost
$680,000/year in 2020 and $5.1 million/year in 2040. A bioretention strategy that gradually reduces
stormwater N from existing development from 0% in 2015 to 15% by 2040 would cost an additional
$3.1 million/year in 2020 and $15 million/year in 2040, in USD 2010. These bioretention estimates
assume a USD 2010 cost of $l,064/lb N/year removed from new development and $2,038/lb N/year
removed from existing development (The Center for Watershed Protection 2013, Cox 2015).
160
-------
Health
In the D-O LRP SD Model, the contributions to avoided premature mortalities resulting from changes
between scenarios in vehicle air emissions, transportation-related physical activity (cycling and
walking), and vehicle crash fatalities are estimated individually and summed to yield net premature
mortalities avoided (per year and cumulatively) for Tier 1 and Tier 2. Health sector outcomes from the
D-O LRP SD Model for the three main scenarios are therefore largely dependent on changes in the
transportation and energy sectors. Although the Light Rail and Light Rail + Redevelopment scenarios
result in net health benefits for the population in both Tier 1 and Tier 2, projected increases in vehicle
fuel efficiency (and therefore reductions in the cost of driving per VMT), which are the same in the two
light rail scenarios as in the BAU scenario, have a much larger impact on health outcomes (by shifting
travel toward more vehicle use and away from nonmotorized travel) than do the changes in either of the
light rail scenarios. For this reason, we present three scenarios that correspond to additional health
interventions which create stronger health responses than either the Light Rail or Light Rail +
Redevelopment scenario alone (see text boxes), two of which are already discussed in this chapter in the
context of the transportation sector (Higher Gas Prices and Sidewalk Building scenarios).
(A) PM2.5 Emissions
per VMT
(B) NOx Emissions
per VMT
(C) Crash Fatalities per
Billion VMT
0.030
0.025
0.020
0.015
0.010
0.005
2000 2005 2010 2015 2020 2025 2030 2035 2040
^^—PM2.5 Emissions per VMT
2000 2005 2010 2015 2020 2025 2030 2035 2040
Emissions per VMT
2000 2005 2010 2015 2020 2025 2030 2035 2040
^^—Crash Fatalities per Billion VMT
Figure 5-104. Time Series of Three Exogenous Inputs Affecting Health Outcomes - Same for all Three
Main Scenarios
161
-------
The model estimates vehicle air emissions and crash fatalities as a proportional function of VMT, using
exogenous parameters, some of which are projected to change in value over time.34 The values for these
parameters are the same for the three main scenarios and are shown in Figure 5-104A-C. Despite steep
reductions in historical vehicle emissions of PM2.5 and NOX per VMT between 2000 and 2015, projected
PM2.5 emissions per VMT (Figure 5-104A) plateau after 2015, causing total PM2.5 vehicle emissions to
increase slightly in the model between 2020 and 2040 in both Tiers in all three main scenarios, as shown
in Figure 5-105. Due to higher VMT in both scenarios, PM2.5 vehicle emissions increase more than BAU
in the Light Rail and Light Rail + Redevelopment scenarios in both Tiers, with increases of 18% and
22% in Tier 1, and 24% and 25% in Tier 2, respectively (compared to BAU increases of 11% in Tier 1
and 22% in Tier 2). Because NOX emissions per VMT are projected to decline more than PM2.5
emissions during the model timeframe (Figure 5-104B), overall NOX emissions from vehicles in all three
scenarios decline even as VMT increases, although NOX emissions do not decline as much in the Light
Rail and Light Rail + Redevelopment scenarios (decreases of 14% and 11% in Tier 1, and 9% and 8% in
Tier 2, respectively, vs. BAU decreases of 19% and 11%), due to VMT increasing more than in the
BAU scenario. Since the model assumes that crash fatalities per VMT after 2013 remain constant at
2013 levels (Figure 5-104C), percent increases in crash fatalities in Tier 1 and Tier 2 between 2020 and
2040 are the same as percent increases in VMT, resulting in 33% and 37% increases in crash fatalities
per year in Tier 1 for the Light Rail and Light Rail + Redevelopment scenarios, respectively, compared
to a BAU scenario increase of 25% (Figure 5-105). In Tier 2, annual crash fatalities increase by 40% and
41% between 2020 and 2040 in the Light Rail and Light Rail + Redevelopment scenarios, respectively,
compared to a BAU scenario increase of 38% (Figure 5-105B).
(A) Tier 1
(B) Tier 2
VMT
PM2.5 Vehicle NOx Vehicle
Crash Fatalities Emissions Emissions
VMT
PM2.5 Vehicle NOx Vehicle
Crash Fatalities Emissions Emissions
I BAU • Light Rail • Light Rail + Redev
I BAU • Light Rail • Light Rail + Redev
Figure 5-105. Percent Change in VMT and VMT-Related Health Indicators between 2020 and 2040 for
Three Main Scenarios
34 Explanations of the calculations for PM2 5 and NOX Emissions per VMT can be found in the Energy Sector portion of
Appendix B, and Crash Fatalities per VMT calculations are explained in the Health Sector portion of Appendix B.
162
-------
(A) Tier 1
(B) Tier 2
ro
Q. 0.9
5
°- n 7
> u-/
20
00 20
- \
10 20
* ^:
20 20
«•—
" ^
30 20
•BAU
a 1'°
& °-9
£0.8
> 0.7
to °-6
| 0.5
5 n 4
^
—
^^^^^^^
2000
2010
2020
2030
2040
• Light Rail + Redev ——Light Rail
• BAU
Figure 5-106. Time Series (2000-2040) of Nonmotorized Travel for Transportation by Residents Per Day
Per Capita for Three Main Scenarios
Figure 5-106 shows the declining trend in nonmotorized travel (walking and cycling) by residents per
day per capita in both Tiers between 2015 and 2040. This decline is due in large part to increases in fuel
efficiency that reduce the cost of driving. On average, residents of Tier 1 walk or cycle significantly
more per day for transportation purposes than do residents of Tier 2. In Tier 1, nonmotorized travel for
transportation declines by about 24% in the BAU scenario between 2020 and 2040, though this decline
is somewhat mitigated in the Light Rail and Light Rail + Redevelopment scenarios, where it decreases
by only 12% and 13%, respectively (Figure 5-106A). In Tier 2, the two light rail scenarios have little or
no effect on nonmotorized travel by residents per day per capita for transportation purposes relative to
the BAU scenario, since the effects the light rail line and associated land use changes are concentrated in
Tier 1 (Figure 5-106B). In all three main scenarios, nonmotorized travel by residents per capita declines
more rapidly in Tier 1 than Tier 2. This is because Tier 1 employment grows faster than Tier 1
population in all three scenarios, which increases GRP, resulting in more people traveling by automobile
and fewer traveling by nonmotorized modes. This change indicates that a greater proportion of travel is
by nonresidents, and is associated with people having to travel shorter distances to get to work or run
errands. Also, this model does not reflect nonmotorized travel at the beginning or end of a trip whose
primary mode is automobile travel; nonmotorized travel of this sort is more common in highly
developed areas, such as Tier 1.
(A) Tier 1
(B) Tier 2
Premature
Total Nonmotorized Nonmotorized Mortalities Avoided
Travel by Residents Total Premature Travel by Residents per Ten Thousand
per Day Mortalities Avoided per Day per Capita People
Premature
Total Nonmotorized Nonmotorized Mortalities Avoided
Travel by Residents Total Premature Travel by Residents per Ten Thousand
per Day Mortalities Avoided per Day per Capita People
• BAU • Light Rail • Light Rail + Redev
• BAU • Light Rail • Light Rail + Redev
Figure 5-107. Percent Change in Nonmotorized Travel by Residents Per Day and Associated Percent
Change in Avoided Premature Mortalities between 2020 and 2040 for Three Main Scenarios: Total and
Population-Normalized
163
-------
Despite the decreases in nonmotorized travel by residents per day per capita in both Tiers for all
scenarios, the higher population growth in the light rail scenarios in Tier 1 increases total daily
nonmotorized travel by Tier 1 residents by 23% and 28% under the Light Rail and Light Rail +
Redevelopment scenarios, respectively, during 2020-2040 (Figure 5-107A), compared to a 13%
decrease in the BAU scenario. These 2020-2040 increases in walking and cycling for transportation
purposes in Tier 1 increase the total number of premature mortalities avoided by 15% and 22% under
the Light Rail and Light Rail + Redevelopment scenarios, respectively, compared to a 14% decrease in
the BAU scenario (Figure 5-107A). Even though nonmotorized travel by residents per capita declines
more quickly in Tier 1 than Tier 2 in all three main scenarios, Tier 1 nonmotorized travel rates are still
higher than Tier 2 rates for the entire model run. Therefore, because the two light rail scenarios increase
the proportion of people living in Tier 1 (in addition to resulting in more walking by residents per capita
than the BAU scenario), the average health of Tier 1 residents is improved in these scenarios. In Tier 2,
total nonmotorized travel by residents increases under all three scenarios during 2020-2040, but the
increase is larger in the light rail scenarios, at 33% and 34% in the Light Rail and Light Rail +
Redevelopment scenarios, respectively, compared to 28% in the BAU scenario (Figure 5-107B). This
results in increased avoided premature mortalities of 26% in the BAU scenario, 31% in the Light Rail
scenario, and 32% in the Light Rail + Redevelopment scenario. However, the additional avoided
premature mortalities in Tier 2 in the two light rail scenarios relative to BAU are all in the portion of
Tier 2 that constitutes Tier 1. Regardless, these results suggest that the addition of the light rail line and
associated redevelopment policies in Tier 1 could have lasting health benefits for the population living
in that Tier.
Figure 5-108 presents a summary of the net health impacts of the Light Rail and Light Rail +
Redevelopment scenarios relative to BAU, calculated by summing the effect of all three health
outcomes (air emissions, traffic accidents, and nonmotorized travel) on premature mortalities avoided
per year. When the health effects of PIVb.s and NOX vehicle emissions, differences in crash fatalities, and
nonmotorized travel are combined, the two light rail scenarios increase avoided premature mortalities
relative to BAU in both Tiers. By 2040, an additional 5.7 net premature mortalities per year are
prevented in Tier 2 in the Light Rail scenario and 7.2 premature mortalities per year are prevented in the
Light Rail + Redevelopment scenario, relative to BAU (Figure 5-108B). Interestingly, the combination
of larger population growth relative to BAU in Tier 1 under the light rail scenarios and the larger relative
increase in nonmotorized travel (compared to BAU) results in more premature mortalities avoided per
year in 2040 in Tier 1 than in Tier 2, with 6.3 additional premature mortalities avoided per year under
the Light Rail scenario and 7.8 avoided per year in the Light Rail + Redevelopment scenario.35 Because
Tier 2 includes Tier 1, this implies that under the light rail scenarios premature mortality rates in the
portion of Tier 2 that is outside Tier 1 are slightly worse than BAU. The net increase in avoided
mortality in Tier 1 (and, by extension, Tier 2) relative to BAU occurs despite a slight increase in
premature mortalities due to vehicle air emissions and crash fatalities in the two light rail scenarios.
35 The health benefits of nonmotorized travel are estimated based on the average amount of walking and cycling per day by
residents, which is much lower in Tier 2 than Tier 1. Consequently, the two light rail scenarios result in slightly fewer
premature mortalities avoided per year in Tier 2 than in Tier 1 in 2040, even though Tier 2 includes Tier 1.
164
-------
(A) Tier 1
(B) Tier 2
S -
O g.
7
6
5
4
3
2
1
^***
x*^
^x
_>*^<»
^^>
s^*
S
/ .
/+'
r *
*
20 2025 2030 2035 20
0)
0) _
s •,
20 20
^-^
^^
25 20
^^
.S&
<**'
30 20
^S
,X^^ *
X^ **
35 20.
•Light Rail* Redev
•Light Rail
•BAU
•BAU
Figure 5-108. Net Premature Mortalities Avoided Per Year from PM2.5 and NO* Vehicle Emissions, Crash
Fatalities, and Nonmotorized Travel Combined between 2020 and 2040, Departure from BAU
(A) Tier 1
(B) Tier 2
Net Premature
PM2.5andNOx Walking and Mortalities
Vehicle Emissions Crash Fatalities Cycling Avoided
Net Premature
PAA2.5 and NOx Walking and Mortalities
Vehicle Emissions Crash Fatalities Cycling Avoided
I Light Rail • Light Rail + Redev
I Light Rail • Light Rail + Redev
Figure 5-109. Cumulative Premature Mortalities Avoided by Cause and Net Cumulative Premature
Mortalities Avoided between 2020 and 2040 for the Light Rail and Light Rail + Redevelopment Scenarios:
Departure from BAU
Finally, Figure 5-109 summarizes the health benefits of the light rail scenarios by showing the difference
in cumulative premature mortalities avoided for each health outcome and for all three outcomes
combined between 2020 and 2040, relative to BAU. In Tier 1, the Light Rail and Light Rail +
Redevelopment scenarios result in 46 and 54 additional avoided premature mortalities, respectively
(Figure 5-109A). In Tier 2, the cumulative totals are 44 and 50 additional avoided premature mortalities,
respectively (Figure 5-109B). The largest difference relative to BAU in avoided premature mortality
comes from additional walking and cycling for transportation in the light rail scenarios, which offsets
slight reductions in avoided premature mortalities caused by increased crash fatalities. Between 2020
and 2040, in both light rail scenarios, PIVb.s and NOX vehicle emissions increase premature mortalities
by fewer than 0.5 cumulative deaths relative to the BAU scenario. The insignificant effect of differences
in vehicle emissions, in spite of increased VMT, between scenarios on health outcomes in the model
may be credited to the higher emissions standards on new vehicles that are represented in every scenario
and which make it less likely for changes in VMT to have significant health impacts in the future, as
each mile of vehicle travel is projected to generate less pollution than historical levels.
165
-------
Vehicle Emissions Reduced
In this scenario, rather than having PM2.5 and NOX vehicle emissions per VMT plateau between
2020 and 2040, both vehicle emission factors are reduced by an additional 10% for all new vehicles
built between 2020 and 2040 (PM2.5 emissions factors per VMT are shown in Figure 5-110; NOX
emissions per VMT are not shown). This change does not translate into an immediate 10%
reduction in vehicle emissions in the model in 2020, but instead causes a gradual reduction over 17
years, as vehicle model year average emission factors (Cai et al. 2013) are weighted each year by
the fraction of vehicles in the U.S. vehicle fleet that are aged 1 year through 17 years (Jackson
200Ib); we assume that the average vehicle remains in use for 17 years (International Energy
Agency 2009). Between 2020 and 2040, the additional reductions in vehicle emissions in this
scenario result in a cumulative four premature mortalities being prevented between 2020 and 2040
relative to BAU (Figure 5-110).
(A) Tier 1
(B) Tier 2
0.025
£ 0.020
0)
% °'015
tC
o 0.010
0.005
2000 2010 2020 2030
^^—Vehicle Emissions Reduced ^^—BAU
•Vehicle Emissions Reduced
-BAU
Figure 5-110. (A) PM2.5 Vehicle Emissions Per VMT: Vehicle Emissions Reduced Scenario Compared
to BAU and (B) Cumulative Premature Mortalities Avoided from PM2.5 and NOX Vehicle Emissions -
Tier 2: Vehicle Emissions Reduced Scenario Departure from BAU
166
-------
Higher Gas Prices
In this scenario (which is also described in the Transportation section), gas prices are set to be
consistently 40% higher than BAU between 2016 and 2040. This increase in the cost of automobile
fuel makes other transportation modes more attractive to travelers. As a result, Tier 2 nonmotorized
travel by residents (walking and cycling) stays near its peak of 0.56-0.57 miles per day per capita
between 2015 and 2040, rather than dropping to less than 0.50 miles per day per capita by 2040, as in
the BAU scenario (Figure 5-111). Between 2015 and 2040, the health benefits of maintaining the
higher rates of walking and cycling for transportation in this scenario lead to a cumulative 280
additional premature mortalities avoided in Tier 2 relative to BAU (Figure 5-111).
(A) Tier 1
(B) Tier 2
3 ~V
%%
SI
°-U
0) 0)
> '^
'^ —
fO fO
0.49
2000
2010
2020
2030
2040
2000
• High Gas Prices
2010 2020
—High Gas Prices
2030
-BAU
2040
Figure 5-111. (A) Nonmotorized Travel by Residents Per Day Per Capita - Tier 2: Higher Gas Price
Scenario Compared to BAU and (B) Cumulative Premature Mortalities Avoided Due to Nonmotorized
Travel - Tier 2: Higher Gas Price Scenario Departure from BAU.
167
-------
Sidewalk Building
In this scenario (which is also described in the Transportation section), demand for nonmotorized
travel facilities (e.g., sidewalks and bike lanes) per developed acre, which is normally held constant,
is set to double its value in BAU in 2020. Due to sidewalk and bicycle path/lane construction time
and a lag in the health effects of walking and cycling, there is a delay of about five years before this
change causes any noticeable health effects. As nonmotorized travel facilities are increased, walking
and cycling levels increase as well, to a lesser degree (the model includes this relationship on the
assumption that increased nonmotorized travel facilities improve the perceived safety of
nonmotorized travel). In Tier 2, sidewalk and bicycle path/lane building leads to an increase in
nonmotorized travel by residents, up to a plateau of about 0.68 miles per day per capita by 2033
(when the increase in nonmotorized travel facilities has been fully realized and travel mode shares
have reach a new equilibrium in response), as opposed to dropping to less than 0.50 nonmotorized
person miles by residents per day per capita by 2040, as in the reference scenario (Figure 5-112).
Between 2020 and 2040, the health benefits of the increase in daily walking and cycling for
transportation by residents lead to a cumulative 560 additional premature mortalities avoided relative
to BAU (Figure 5-112).
(A) Tier 1
(B) Tier 2
-------
5.5 Summary of Results
As the discussion throughout this chapter shows, both of the light rail scenarios result in increased
population growth, land use and development, economic activity and employment, and VMT and
congestion, relative to the BAU scenario in the D-O LRP SD Model. Some of these changes result in
improvements to local quality of life (e.g., a reduction in the percent of the population living in poverty),
while others have more negative effects (e.g., increased traffic congestion and stormwater runoff from
increased impervious surfaces). In this section, we summarize results for some of the model's main
social, economic, and environmental indicators. We first summarize model results by isolating the
changes in model outputs between 2020 and 2040 for the three main scenarios in Table 5-1 to emphasize
the magnitude of the impacts that our SD model estimates for the light rail scenarios. Next, we
summarize the overall differences in model outputs in 2040 for the three main scenarios in Table 5-2 to
show the end result of the model's many feedbacks. Finally, we summarize and list benefits and
tradeoffs of the light rail scenarios, with a third list summarizing the Light Rail Scenario tradeoffs that
are mitigated in the Light Rail + Redevelopment Scenario.
In Table 5-1, we summarize the magnitude of the changes to indicators for each light rail scenario
between 2020 and 2040, relative to the BAU scenario. For each indicator, the table presents the absolute
change in each Tier in each scenario, as well as the relative change (as a percentage of the BAU change)
for the Light Rail and Light Rail + Redevelopment scenarios. For example, Tier 1 population increases
by 6,160 people in the BAU scenario between 2020 and 2040, while the population change in the Light
Rail scenario is 182% higher, at 17,400 people. Of the 12 indicators listed in Table 5-1, all moved in the
same direction as BAU in Tier 1 in both of the light rail scenarios, with the exception of public transit
use by residents per capita (which increased rather than declining) and impervious surfaces per capita
(which declined rather than increasing). For most variables, the relative difference between the BAU and
the light rail scenarios was larger in Tier 1 than Tier 2, consistent with the expectation that light rail
impacts will be locally concentrated around the station areas.
Table 5-1. Summary of Changes in Key Indicators between 2020 and 2040 for the BAU, Light Rail, and
Light Rail + Redevelopment Scenarios
Variable -Tier 1
BAU (A 2020-
2040)
Light Rail
Light Rail + Redev
% diff from A in BAU A 2020-2040 % diff from A in BAU
Population
Nonresidential sq ft
Developed land (acres)
Employment (jobs in area)
Zero car households
+6,200
Rates
GRP (USD2010/year)
Total retail consumption (USD2010/year)
VMT (miles/day)
Intensity Measures
Energy spending share of GRP (percentage points)
Public transit use by residents (person miles/day/capita)
CO2 emissions per GRP (tons/million USD 2010)
Impervious surfaces per capita (acres/thousand people)
-48
+4
-5.7
169
-------
Variable-Tier 2
BAU (A 2020-
2040)
Light Rail
Light Rai I + Rede v
% diff from A in BAU A 2020-2040 % diff from A in BAU
+180,000 +201,000
+230
+25B
+5.8B
+28B
+6.7B
S16%
14%
+29B
+6.7B
| 17%
| 15%
+5.8M
+6.1M
5.7%
Population
Nonresidential sq ft
Developed land (acres)
Employment (jobs in area)
Zero car households
GRP (USD2010/year)
Total retail consumption (USD2010/year)
VMT (miles/day)
Intensity Measures
Energy spending share of GRP (percentage points)
Public transit use by residents (person miles/day/capita)
CO2 emissions per GRP (tons/million USD 2010)
Impervious surfaces per capita (acres/thousand people)
* Percent difference not visualized due to scale.
The Light Rail + Redevelopment scenario suggests that it is possible to maintain the economic benefits
of expansion while mitigating some harmful effects of increased land development (i.e., it slows the
growth of impervious surface). Between 2020 and 2040, the increase in Tier 1 developed land is 47%
higher than BAU in the Light Rail scenario but only 26% higher than BAU in the Light Rail +
Redevelopment scenario (see Table 5-1). This reduction in land development does not limit economic
growth in the Light Rail + Redevelopment scenario, as it still sees a GRP increase in Tier 1 that is 66%
higher than BAU (compared to 29% higher in the Light Rail scenario). The increased economic growth
does increase the carbon intensity of the economy in Tier 1 under the Light Rail + Redevelopment
scenario relative to BAU, but within an overall trend of decreased carbon intensity. Due to increasing
energy efficiency in buildings and vehicles, CO2 emissions per GRP are projected to decline by 48 tons
per million USD 2010 between 2020 and 2040 in the Tier 1 BAU scenario (Table 5-1). CO2 emissions
per GRP decline 2.2% more, and 6.6% less, in the Light Rail and Light Rail + Redevelopment scenarios,
respectively.
Besides stimulating GRP, the two light rail scenarios project dramatic increases in Tier 1 public transit
use by residents per day per capita between 2020 and 2040 (189% in the Light Rail scenario and 191%
in the Light Rail + Redevelopment scenario). This is counter to the declining trend in Tier 1 public
transit use in the BAU scenario. However, the increase in population and economic activity in both light
rail scenarios leads to growth in overall VMT. During 2020-2040, Tier 1 VMT in the Light Rail and
Light Rail + Redevelopment scenarios increases by 31% and 46% more than in the BAU case,
respectively. In addition to increased CO2 emissions, greater VMT results in more traffic congestion and
more emissions of PM2.5 and NOX. Thus, even though the development of the light rail results in a larger
number of people using public transit and nonmotorized travel modes (with corresponding reductions in
VMT per person), it does not fully mitigate the negative consequences of population and economic
growth in the region.
Finally, below we present a summary of the main benefits, tradeoffs, and mitigating effects of the two
light rail scenarios evident in the year 2040. Table 5-2 visually summarizes these effects, showing the
values of several indicators in 2040 in the BAU and two light rail scenarios, together with percent
differences from the BAU values.
170
-------
Benefits of the Light Rail Scenarios
• The light rail attracts more economic growth to the region, increasing employment in 2040
by 28,000 jobs in Tier 2 and by 16,000 jobs in Tier 1 in the Light Rail scenario relative to the
BAU scenario (see Table 5-2).
• The light rail creates more mixed uses more quickly in Tier 1, as measured both by the HHI
(reaching its maximum divergence at a level 1.8% lower than BAU in 2026) and the jobs-
housing balance (peaking at 8.8% higher than BAU in 2040). Redevelopment does not
significantly affect mixed uses, however (i.e., no significant difference between Light Rail and
Light Rail + Redevelopment scenarios).
• Because it improves the public transit system through the addition of fixed-guideway service on
an exclusive right-of-way, the Light Rail scenario causes 2040 Tier 2 public transit ridership
to be 48% greater than what it would have been in the BAU scenario, even though it only
causes Tier 2 population to be 2.8% greater than what it would have been in the BAU scenario.
Because it takes the additional step of allowing denser development in light rail station areas
(and hence allowing more people and businesses to locate there), the Light Rail +
Redevelopment scenario causes 2040 Tier 2 public transit ridership to be 54% greater than
the BAU scenario, accompanied by an increase in population that is only 3.7% greater than
BAU.
• As the Light Rail and Light Rail + Redevelopment scenarios increase public transit use, they
have the added benefit of increasing nonmotorized travel. In 2040, the average resident of Tier 1
walks or bicycles 0.095 more miles per day in the Light Rail scenario than in the BAU scenario
(a difference of 15%) and 0.084 more miles per day in the Light Rail + Redevelopment scenario
(a difference of 14%).
171
-------
Table 5-2. Summary of Key Indicators in 2040 for the BAD, Light Rail, and Light Rail + Redevelopment
Scenarios
Stocks
Population
Employment (jobs in the Tier)
GRP(USD2010/year)
VMT miles/day
Peak period automobile speed (miles/hour)
Total energy use (MMBtu/year)
CO2 emissions from buildings and transportation (tons/year)
Stormwater runoff N load (Ibs/year)
SF property value (USD 2010/dwelling unit)
MF property value (USD 2010/dwelling unit)
Nonresidential property value (USD 2010/sq ft)
Net premature mortalities avoided peryear*
Indices
Jobs-housing balance (1 is balanced)
HHI Index (lower is more mixed uses)
Cumulative Measures
BfffflBBBBHBB
Intensity Measures
Light Rail Light Rail+Redev
2040 Value I %diff from BAUl 2040 Value I % diff f ram BAU
Cumulative city and county real property tax levied (USD 2010)
Developed land (acres) percapita
Publictransit ridership by residents (trips/year) percapita
Nonmotorized travel by residents (person miles/day) percapita
Daily water demand (Mgal/year) percapita
Percent of the population in poverty (percentage points)
50,000
119,000
14B
61,000
135,000
/O
14%
0.10
65,000
147,000
Population
Tier 2
Employment (jobs in the Tier)
IBAU Light Rail Light Rail + Redev
2040Value | 2040Value | % diff f rom BAU | 2040Value |%diff from BAU
• 660,000
430,000
680,000
455,000
12.7%
fc.9%
690,000
459,000
! 3.5%
1 6.7%
Rates
GRP USD2010/year
VMT (miles/day)
Peak period automobile speed (miles/hour)
Total energy use (MMBtu/year)
CO2 emissions from buildings and transportation (tons/year)
Stormwater runoff N load (Ibs/year)
SF property value (USD 2010/dwelling unit)
MF property value (USD 2010/dwelling unit)
Nonresidential property value (USD 2010/sq ft)
Net premature mortalities avoided peryear*
Jobs-housing balance (1 is balanced)
HHI Index (lower is more mixed uses)
Cumulative LRP revenues (USD 2010)
Cumulative city and county real property tax levied (USD 2010)
Intensity Measures
Developed land (acres) percapita
Publictransit ridership by residents (trips/year) percapita
Nonmotorized travel by residents (person miles/day) percapita
Daily waterdemand (Mgal/year) percapita
Percent of the population in poverty (percentage points)
60B
21M
47
80M
10.8M
1.46M
480,000
93,000
543
125
0.30
63 B
21.3M
46
83M
11.3M
1.49M
510,000
100,000
577
130
1.5%
-1.5%
3.9%
1.9%
1.9%
.1%
i.3%
L5%
64B
21.4M
46
84M
11.4M
1.49M
520,000
101,000
583
132
7.2%
1-1.9%
4.6%
6.2%
3.4%
r.3%
5.7%
0.89
0.62
0.88
0.62
!-1.7%
-1.1%
0.88
0.62
1-1.7%
-1.1%
LIB
28.9B
1.2B
29.9B
S2.0%
3.5%
1.2B
30.0B
[2.1%
|3.8%
* The net effect ofPM2.s andNOx vehicle emissions, physical activity, and crash fatalities.
172
-------
Tradeoffs Due to the Light Rail Scenarios
• The economic growth produced by the light rail scenarios increases motor vehicle travel
and hence increases traffic congestion, in spite of the light rail line shifting a portion of
travelers from automobiles to public transit. In Tier 1, the average peak-period automobile speed
in 2040 is 2.5 mph less than BAU in the Light Rail scenario and 4.7 mph less than BAU in the
Light Rail + Redevelopment scenario. In Tier 2, the average peak-period automobile speed in
2040 is 0.72 mph less than BAU in the Light Rail scenario and 0.94 mph less than BAU in the
Light Rail + Redevelopment scenario.
• The light rail scenarios increase energy consumption and COi emissions relative to BAU,
with a larger relative increase in Tier 1 than in Tier 2. In Tier 1, energy consumption and CO2
emissions in 2040 are about 10% higher in the Light Rail scenario than BAU, and 30% higher in
the Light Rail + Redevelopment scenario than BAU. In Tier 2, energy consumption and CO2
emissions increase by less than 5% over BAU in both light rail scenarios. Due to the effects of
land scarcity on development in the Light Rail scenario, energy consumption and emissions
level off around 2033 in Tier 1, but they continue to increase out to 2040 in the Light Rail +
Redevelopment scenario.
• In all three main scenarios, the energy intensity of the economy is projected to decrease. Energy
use per real dollar of GRP decreases by about 35% between 2020 and 2040 in the three main
scenarios, in both Tiers. However, whereas area-based building energy intensity in Tier 1
(MMBtu/year/acre) declines by 1-2% in the BAU and Light Rail scenarios between 2020 and
2040, it increases by 26% in the Light Rail + Redevelopment scenario, due to higher-
density development.
• Due to the growth in impervious surface, the light rail scenarios increase stormwater
runoff N load in Tier 1 by 6% over BAU in 2040 in the Light Rail scenario and 4% over BAU
in the Light Rail + Redevelopment scenario. Despite the effect of land scarcity on development
starting in 2033, the Light Rail scenario has the highest stormwater N load in Tier 1 by 2040.
• Due to growth in population and employment, the light rail scenarios increase total
municipal water demand relative to BAU, with a larger relative increase in Tier 1 than in Tier
2. In 2040, projected daily water demand in Tier 1 is 17% and 27% higher than BAU in the Light
Rail and Light Rail + Redevelopment scenarios, respectively. In Tier 2, daily water demand in
2040 is forecasted to be 4-5% higher in the light rail scenarios than BAU. Land scarcity after
2033 in Tier 1 causes slower growth in economic and residential development, leading to slower
growth in water demand in the Light Rail scenario compared to the Light Rail +
Redevelopment scenario. Although total water demand increases under the light rail scenarios,
daily water demand per capita is 2-4% lower in 2040 relative to BAU in Tier 1, and nearly
unchanged in Tier 2.
• The Light Rail + Redevelopment scenario causes the largest decrease in the percent of the
population in poverty in Tier 1 (17% lower than BAU in 2040) due to a drop in
unemployment, which results in a smaller increase in the number of transit-dependent
households in Tier 1 than in the Light Rail scenario (1.5% and 2.6% higher than BAU in 2040,
respectively), which reduces transit ridership relative to what it otherwise would be.
173
-------
Light Rail Scenario Tradeoffs Mitigated in the Light Rail + Redevelopment Scenario
• The Light Rail + Redevelopment scenario is assumed to significantly increase the density of
land use, and this change is evident through a 13% drop in developed land per capita between
2020 and 2040 in Tier 1.
• The Light Rail + Redevelopment scenario offers a benefit of reducing annual stormwater N
load, relative to the Light Rail scenario, in Tier 1, with only a 4% higher annual N load relative
to BAU in 2040 compared to 6% higher under the Light Rail scenario.
• Although the Light Rail + Redevelopment scenario only marginally increases economic
growth in Tier 2 relative to the Light Rail Scenario, it concentrates employment in Tier 1,
adding 12,000 additional jobs to Tier 1 in 2040 relative to the Light Rail Scenario.
• Increasing property values in Tier 1 under the Light Rail + Redevelopment scenario provide a
win-win for tax revenues and affordability; in real terms, residential property values are no more
than 7% higher than BAU in 2040 (compared to 7% lower under the Light Rail scenario), while
nonresidential property values increase by 136% more than under the Light Rail scenario in Tier
1 between 2020 and 2040. This leads to $660M more in cumulative real property taxes levied
between 2020 and 2040 than under the Light Rail scenario in Tier 1 alone.
• Since more people live in Tier 1 in the Light Rail + Redevelopment scenario (compared to the
BAU and Light Rail scenarios), more people realize the health benefits of increased walking and
cycling for transportation purposes caused by the light rail. This results in approximately 7.8
more premature mortalities avoided per year in 2040 in Tier 1 than in the BAU scenario
(compared to 6.3 avoided per year in the Light Rail scenario relative to BAU), despite the small
increase in deaths due to vehicle air emissions (0.03 per year) and crash fatalities (0.5 per year)
in Tier 1 due to increased VMT.
174
-------
6 Quality Assurance
This chapter presents a detailed description of the quality assurance (QA) steps taken to calibrate and
validate the D-O LRP SD model. At the onset of model development, the modeling group developed a
Quality Assurance Project Plan (QAPP) (dated September 16, 2014), which was approved by the
National Exposure Research Laboratory (NERL) and Office of Research and Development (ORD)
Directors of Quality Assurance. After we developed the conceptual model in Phase I of this research
project, we constructed and calibrated the operational model (Phase II) and presented its results for
several transportation and land use scenarios to stakeholders. Based on their reactions, we made
modifications to the model and identified a number of alternative scenarios that could maximize the
benefits of the D-O LRP or minimize its consequences, which we analyzed in Phase III of the project.
This chapter presents the results of thorough model testing at both Phases II and III in accordance with
the QAPP. Additional exploration of the D-O LRP SD Model's structure and functionality can be
conducted in the Vensim file itself. For assistance in navigating the model in Vensim, please refer to the
User Guide provided in Appendix A.
6.1 QA Overview
Models can be classified in many different ways and assessed according to different criteria, such as
physical versus symbolic; dynamic versus static; deterministic versus stochastic, and others. As it relates
to the notion of validity, a crucial distinction must be made between models that are "correlational" (i.e.,
purely data-driven or "black-box") and models that are "causal-descriptive" (i.e., theory-like or "white-
box").
In correlational models, since there is no claim of causality in structure, what matters is the aggregate
output behavior of the model; the model is assessed as valid if its output matches the "real" output
within a specified range of accuracy, without any questioning of the validity of the individual
relationships that exist in the model. This type of "output" validation can often be cast as a classical
statistical testing problem. Models that are built primarily for forecasting purposes (such as time-series
or regression models) belong to this category.
On the other hand, causal-descriptive models make statements about how real systems actually operate
in some aspects. In this case, generating an "accurate" output behavior is not sufficient for model
validity; what is crucial is the validity of the internal structure of the model. A causal-descriptive model,
in presenting a theory about the real system, must not only reproduce or predict its behavior, but also
explain how the behavior is generated, and possibly suggest ways of changing the system's behavior.
System dynamics models, such as the Durham-Orange Light Rail model, fall into the causal-descriptive
category of models. Such models are built to assess the effectiveness of alternative policies or design
strategies at improving the behavior of a given system. This is only possible, of course, if the model has
an internal structure that adequately represents those aspects of the system that are relevant to the
problem behavior at hand. In short, it is often said that a system dynamics model must generate the
"right output behavior for the right reasons."
This section discusses model parameterization (i.e., calibration), corroboration (i.e., validation and
simulation, and sensitivity analysis), and computational reproducibility of the results of the model. The
main purpose of these procedures is to ensure that the model is accurate and precise enough to meet the
175
-------
project needs. Because the model is intended to provide information about causal pathways and the
magnitudes and direct!onalities of policy impacts, rather than yielding numerical results that in
themselves form the basis for decisions, there are no formal data quality objectives. However, the ability
of the model to reproduce historical data and to match projections where applicable is an important
indication of its structural validity. We have identified a target of 10% for goodness of fit, which, in
most cases, was easily met. The sections that follow describe the testing of the model and interpret the
results of those tests with respect to model structure, performance and reliability. Section 6.1, Model
Parameterization (Calibration), describes sources of historical data and projections used (including other
models); the integration of D-O LRP model sectors; the characterization of uncertainty among inputs to
the D-O LRP model; and calibration statistics comparing model output to data. Section 6.2, Model
Corroboration (Validation and Simulation), describes tests of model structure (equations, linkages); and
sensitivities of model behavior to the value of uncertain parameters. Section 6.3 (Computational
Reproducibility) describes how users can obtain the D-O LRP model source code, which includes
documentation of the model equations.
6.2 Model Parameterization (Calibration)
This section describes the calibration steps that took place during two stages of model development:
1. Intra-sectoral calibration, which took place separately within each sector; and
2. Inter-sectoral calibration, which took place after the sectors were linked into a single model
framework.
In both stages, model outputs were compared against external data sources and projections. Where
necessary, we adjusted parameter values in order to improve this fit. We first describe data sources
(including other models) used to calibrate the D-O LRP model, and then describe the results of model
integration (inter-sector calibration) and the fit of major variables to data in the business-as-usual
scenario (intra-sector calibration).
Data Collection and Analysis
In the initial phase of the development of the D-O LRP model, we selected historical data from existing
literature for use in populating and calibrating the model. When relevant historical data were not
available, the modeling team consulted experts in order to determine baseline assumptions that would
allow for further calibration and validity testing. These assumptions include parameter values as well as
equations for relationships between model variables that would allow the model to reproduce historical
or projected data trends.
Historical Data and Projections
Table 6-1 lists a selection of the main variables of the model, together with data sources that provided
historical and (where applicable) projected trends used for calibration purposes. Further detail on the
calibration and data processing steps used for these variables can be found in Appendix B.
176
-------
Table 6-1. Selected Variables and Data Sources for Historical and Projected Data
VARIABLE
DATA SOURCES (HISTORICAL)
DATA SOURCES (PROJECTIONS)
LAND USE SECTOR
Population
Developed Land
Nonresidential Sq Ft
Dwelling Units
ESRI Community Analyst
CV2 Parcel Geodatabase for Place Type &
Development Status Editing
Durham County Tax Administration Real Property
Database; Orange County Parcel Database;
Chatham County Tax Parcel Database
ESRI Community Analyst
TRM v5 SE Data
Imagine 2040 Results CIS Data
None used
ESRI Community Analyst
TRANSPORTATION SECTOR
VMT
Person Miles of Public
Transit Travel per Day
Traffic Congestion
TRM v5 travel demand result shapefiles
FTA. 2015. National Transit Database.
Supplemented by information from Jennifer
Green at Triangle Transit (now called GoTriangle)
TRM v5 travel demand result shapefiles
TRM v5 travel demand result
shapefiles
(1) CAMPO and DCHC MPO. 2013.
"2040 Metropolitan Transportation
Plans.";
(2) TRM v5 travel demand result
shapefiles;
(3) Greater Triangle Travel Study,
Household Travel Survey Final Report
(conducted for the TRM);
(4) ESRI Community Analyst. 2014.
TRM v5 travel demand result
shapefiles
ENERGY SECTOR
Building Energy Use
CO2 Emissions
Energy prices
Durham City-County Sustainability Office
Durham City-County Sustainability Office
USEIA. 2015:
(1) "Weekly Retail Gasoline and Diesel Prices."
(2) "North Carolina Price of Natural Gas
Delivered to Residential Customers."
(3) "Electricity: Sales (consumption), revenue,
prices & customers."
None used
Durham City-County Sustainability
Office
US EIA. 2015. "Annual Energy Outlook
2015."
ECONOMY SECTOR
Total Employment Tier 2
Total Employment Tier 1
Total Retail Consumption
Tier 2
Total Retail Consumption
Tierl
Gross Regional Product
(GRP)
U.S. Bureau of Economic Analysis (BEA) and
TRM v5 SE data
LODES (U.S. Census Bureau) and TRM v5 SE
data
North Carolina Dept. of Revenue and Woods &
Poole Economics, Inc. Copyright 2014
U.S. Census Economic Census data downloaded
from SimplyMap
Methodology from BEA (Panek et al. 2007)
TRM v5 SE data
TRM v5 SE data
Woods & Poole Economics, Inc.
Copyright 2014
None used
None used
EQUITY SECTOR
Property Values
Durham County Tax Administration Real Property
Database; Orange County Parcel Database;
Chatham County Tax Parcel Database
None used
WATER SECTOR
Impervious Surface
Average Precipitation
Total Water Demand
U.S. Environmental Protection Agency
EnviroAtlas
State Climate Office of North Carolina
NC Department of Environment and Natural
Resources
None used
None used
Triangle Regional Water Supply Plan
Vol. 1
HEALTH SECTOR
Crash fatalities per year
Highway Safety Research Center at UNC Chapel
Hill. 2015.
None used
177
-------
Other Existing Simulation Models
A variety of models were used to obtain additional information to be used in the D-O LRP model. These
were needed either to fill gaps in historical data or to provide a higher degree of disaggregation for
projecting behavior (or outcomes) that otherwise cannot be measured. These include the Triangle
Regional Model for transportation planning, the Community Viz 2.0 model for land use planning, and
the Jordan Falls Stormwater Load Accounting Tool for water infrastructure planning, as indicated in the
Model Description report:
Table 6-2. Existing Simulation Models Used in the D-O LRP Model
MODEL
SECTOR WHERE
USED
INFORMATION USED
REFERENCE
1. Triangle Regional
Model (TRM) v5
Land Use,
Economy, and
Transportation
Projections for population,
households, employment,
VMT, and many other
transportation-related
variables
DCHC MPO. 2013. "Triangle Regional Model
version 5: Socioeconomic data and
projections for the preferred growth scenario
and travel demand result shapefiles."
2. CommunityViz 2.0
Land Use
Acres by development
status and type in 2013
TJCOG. 2014. "CommunityViz 2 (CV2) Parcel
Geodatabase for Place Type & Development
Status Editing."
3. IEA/SMP
Transportation Model
Transportation
Equation (including an
elasticity) describing effect
of economic indicators on
vehicle ownership
International Energy Agency and World
Business Council for Sustainable
Development. 2004. "IEA/SMP Transportation
Model." Spreadsheet model discussed in:
Fulton, Lew, and G. Eads. 2004. "IEA/SMP
Model Documentation and Reference Case
Projection."
4. "Draft Spreadsheet
Tool: Estimated
Ridership and Cost of
Fixed-Guideway
Transit Projects,"
created as part of
TCRP Project H-42
Transportation
Equation for the change in
person miles of public
transit travel per year due
to adding fixed-guideway
transit.
"Draft Spreadsheet Tool: Estimated Ridership
and Cost of Fixed-Guideway Transit Projects,"
created as part of TCRP Project H-42:
Chatman, Daniel G., Robert Cervero, Emily
Moylan, Ian Carlton, Dana Weissman, Joe
Zissman, ErickGuerra, Jin Murakami, Paolo
Ikezoe, Donald Emerson, Dan Tischler, Daniel
Means, Sandra Winkler, Kevin Sheu, and Sun
Young Kwon. 2014. "TRCP Report 167:
Making Effective Fixed-Guideway Transit
Investments: Indicators of Success."
5. Jordan Falls
Stormwater Load
Accounting Tool
Land Use, Water
Coefficients for impervious
surfaces by residential
density; event mean
concentration N and P.
NCDENR. 2011. "Jordan Lake Stormwater
Load Accounting Tool User's Manual."
6. Simple Method for
Calculating
Stormwater Runoff
Water
Equation for Stormwater
nitrogen and phosphorous
loading.
Shaver et al. 2007. "Fundamentals of urban
runoff management: technical and institutional
issues."
7. National Energy
Modeling System
(NEMS)
Energy
The basis for EIA Annual
Energy Outlook 2015
projections, which were
used for future building
and vehicle energy
intensity as well as future
energy prices.
US EIA. 2009. "The National Energy Modeling
System: an overview."
8. Health economic
assessment tools
(HEAT) for walking
and for cycling (from
the World Health
Organization)
Health
Equation for number of
deaths avoided due to
walking for transportation,
and number of deaths
avoided due to cycling for
transportation.
WHO. 2014. "Health economic assessment
tools (HEAT) for walking and for cycling:
Methodology and user guide, 2014 update."
178
-------
The use of these models, including (1) the data inputs employed by each model, (2) the results each
model generated, and (3) the specific equations used to estimate these results, allowed us to include
sectors in the D-O LRP model for which data on parameter values and structural relationships would
otherwise not have been available.
Other Existing Studies Providing Selected Model Equations
Certain parameters for model setup and calibration were obtained from studies of other areas, primarily
due to the lack of data to inform particular relationships for the Durham and Orange County area. These
studies generally focus on specific relationships, such as that between employment and person miles of
public transit travel. Table 6-3 provides a list of all equations taken from the literature and other models,
with their sources. Note that this table only includes sources for equations. For a complete listing of all
other inputs (e.g. elasticities and effect tables) taken from the literature, as well as on the actual
equations used from the sources below, see Appendices B and C.
Table 6-3. Equations Obtained from Existing Studies and Models
VARIABLE
MODEL PARAMETERS USED TO
CALCULATE
SOURCE
LAND USE SECTOR
Jobs-housing balance
Herfindahl-Hirschman
Index (HHI)
Total employment, workers per household,
total dwelling units
Industrial, office, retail , service, multifamily,
and single family percents of developed land
Ewing, R., et al. 1996. "Land use impacts on trip
generation rates."
Song, Y. and D. A. Rodriguez. "The
measurement of the level of mixed land uses: A
synthetic approach."
TRANSPORTATION SECTOR
Change in person miles
of public transit travel per
year due to adding fixed
guideway transit
Desired vehicle
ownership per person not
in a zero-car household
Total employment Tier 1; population Tier 1;
retail plus entertainment employment Tier 1;
jobs earning $3,333 per month in 2010 USDs
Tier 1 ; VMT per highway lane mile (Tier 2)
Initial vehicle ownership per person not in a
zero-car household, relative resident per
capita net earnings
"Draft Spreadsheet Tool: Estimated Ridership
and Cost of Fixed-Guideway Transit Projects,"
created as part of TCRP Project H-42:
Chatman, Daniel G., Robert Cervero, Emily
Moylan, Ian Carlton, Dana Weissman, Joe
Zissman, ErickGuerra, Jin Murakami, Paolo
Ikezoe, Donald Emerson, Dan Tischler, Daniel
Means, Sandra Winkler, Kevin Sheu, and Sun
Young Kwon. 2014. "TRCP Report 167: Making
Effective Fixed-Guideway Transit Investments:
Indicators of Success."
International Energy Agency and World
Business Council for Sustainable Development.
2004. "IEA/SMP Transportation Model."
Spreadsheet model discussed in: Fulton, Lew,
and G. Eads. 2004. "IEA/SMP Model
Documentation and Reference Case
Projection."
ECONOMY SECTOR
Desired employment
(both Tiers)
Employment per dollar of consumption, total
retail consumption
Keynes. 1936. "General theory of employment,
interest and money"; Trends. 2010.
WATER SECTOR
Nitrogen loading ("Total
N load")
Phosphorus loading
("Total P load")
Event mean concentrations of N, average
precipitation per year, total land, impervious
coefficient
Event mean concentrations of P, average
precipitation per year, total land, impervious
coefficient
Shaver et al. 2007. "Fundamentals of urban
runoff management: technical and institutional
issues;" NCDENR. 2011. "Jordan Lake
Stormwater Load Accounting Tool User's
Manual;" State Climate Office of North Carolina.
2015. "Historical Data."
Shaver et al. 2007. "Fundamentals of urban
runoff management: technical and institutional
issues;" NCDENR. 2011. "Jordan Lake
179
-------
VARIABLE
MODEL PARAMETERS USED TO
CALCULATE
SOURCE
Stormwater Load Accounting Tool User's
Manual;" State Climate Office of North Carolina.
2015. "Historical Data."
HEALTH SECTOR
Number of premature
mortalities avoided due
to walking for
transportation & number
of premature mortalities
avoided due to cycling
for transportation
Person miles of walking, cycling for
transportation by residents per day per capita
WHO. 2014. "Health economic assessment
tools (HEAT) for walking and for cycling:
Methodology and user guide, 2014 update."
Given that the D-O LRP model uses causal relationships to generate projections for all the variables
included in the model, it is important to ensure that the model built with the information received from
various data sources represents a coherent system and produces consistent results. To accomplish this
goal, we employed an integrated framework in which we evaluated equations and parameters obtained
from existing studies by incorporating them into the model in an iterative fashion, and - at each step -
comparing model outputs to historical data and projections.
To illustrate this framework, the team performed structural and sensitivity tests on the equity sector to
quantify the variability of property value estimates in response to the many factors affecting them. A
literature search revealed a wide variety of relationships based on diverse geographies: from small
studies based in one neighborhood to papers looking at trends in all major metro areas of the U.S. Most
studies that created relationships based on spatially explicit measures could not be used (such as
proximity to retail) since our model is not spatially explicit.36 In addition, relationships that rely on
associations with variables not included in our model (e.g., housing age or number of rooms) could not
be used. Since there was very little literature available on factors affecting multifamily property values,
we borrowed relevant elasticities from the literature on both single-family and nonresidential property
values to complete the relationships in this sector.37 Through this process, we found that some
relationships had to be modified to maintain a close fit between model outputs and external data sources.
For example, an elasticity between available land and home prices (Capozza et al. 2002b) produced
reasonable outputs for Tier 2, but in Tier 1, the use of this parameter value produced results that
diverged significantly from historical data. We found that this result was caused by the fact that land
availability in Tier 1 approaches zero, which is outside of the range of values where the estimate from
the literature applies. Since this represents a somewhat artificial restraint, we replaced this elasticity with
an effect table which simulates the same elasticity for moderate values of land availability, but dulls the
effect at the extremes. For a complete listing of all changes made during calibration, and the adjustments
made to data and equations drawn from the literature, see Appendix B.
36 The exception is average commute time to work (Kockelman 1997), for which we were able to create a proxy in the model
based on person miles of peak period automobile travel by residents per day and peak period vehicle speed.
37 For example, we used the elasticity between job density and single-family home values and the elasticity between building
size and nonresidential property value (Srour et al. 2002) to relate job density and building size, respectively, to multifamily
property values.
180
-------
Systemic Model Creation
The second phase of calibration takes place after various model sectors are linked, and exogenous
assumptions (or drivers) of certain sectors are replaced by the outputs (endogenous variables) of others,
creating a cross-sectoral causal network. In this section we (1) describe the effects of this model
integration, summarizing how the integration process adds functionality and realism to the model; (2)
present a model input characterization table assessing the quality of variables used within each sector;
and (3) summarize the uncertainty within each sector, based on the input characterization table.
In the model integration stage, the modeling team once again compared model outputs to historical data
and projections and adjusted parameter values to improve the model's fit with external data sources.
This calibration process is systematic, as there are precise steps to follow and validation tests to perform,
as well as systemic, given that various modules have to be linked horizontally (i.e., across sectors). This
process allowed the modeling team to identify any incorrect sectoral parameters, as errors in one sector
would be propagated to others, and to carry out a more precise and comprehensive calibration. Figure
6-1 illustrates the cross-sectoral linkages in green (both Tier 2 and Tier 1) and purple arrows (Tier 1
only) that were established when the core model sectors (land use, transportation, energy, and economy)
were integrated. Intra-sector relationships that already existed are shown in dotted black arrows.
Land Use
Net Migration
^^ I I 1
Population I +~\ Dwelling Units | | N oni esidential Sq Ft |
/~
^Transportation
Figure 6-1. Schematic Illustrating Cross-Sectoral Linkages Established After Model Integration
Improvements Resulting From Model Integration
The integration of the seven sectors in the D-O LRP SD Model is one of the key features that sets this
model apart from sector-specific models. This section discusses the major changes that occurred in each
of the four core sectors as a result of this integration process, including completing feedback loops that
allowed changes in one sector to affect outcomes in another sector. For each sector, we describe the
differences between the pre-integration and post-integration versions of the model and discuss how
those differences add functionality and realism to the model.
181
-------
Integrating the Land use and Economy Sectors
Integrating the land use sector with the other model sectors involved replacing an exogenous projection
for employment (which drives demand for nonresidential sq ft) by the endogenously calculated
employment from the economy sector. This step completed a reinforcing feedback loop from
employment to nonresidential square feet, to gross operating surplus to GRP, and back to employment
(See Figure 6-1). This change also affected the calculation of the jobs-housing balance. Through this
integration, outputs from the land use sector, including endogenous population, developed land, housing
units, and nonresidential sq ft were linked to most of the other sectors of the model.
As shown in Figure 6-2 (for the service sector), this integration caused very little change to
nonresidential sq ft in the short term, but the reinforcing feedback loop with employment lead to rising
demand starting in 2014. As the figure shows, non-residential sq ft in the service sector was 34% higher
post-integration.
2000 2010
^^— BAU Post-Integration
--- Tier2 DATA
2020
2030
2040
BAU Pre-Integration
Figure 6-2. Service Square Feet - Tier 2: Before and After Model Integration
Completing this loop makes land use responsive to changes in the economy, allowing alternative
scenarios to have effects that ripple through the sectors. To illustrate this, the test shown in Figure 6-3
below shows the percent change in total nonresidential sq ft caused by a 20% increase to the demand for
nonresidential sq ft in Tier 2. While the 20% increase in demand leads to a boost in total nonresidential
sq ft of about 2.3% over the model time period before integrating the Land Use and Economy sectors,
the same change leads to an increase of about 2.8% post integration.38 The feedback loop means that the
increased nonresidential sq ft increase economic activity and employment, which in turn increases
demand for nonresidential sq ft over what it would have otherwise been.
38 The percent increase in total nonresidential sq ft is much lower than the percent increase in demand due to two factors: 1)
the 20% boost does not apply to industrial sq ft, and 2) this test was conducted using the models as they existed immediately
pre and post-integration, before the land development sector was restructured to allow development to better reflect demand,
as described in Structure Confirmation Tests subsection of Section 6.3.
182
-------
0.0%
2000
2010 2020 2030
—Percent change in Post-Integration
—Percent change in Pre-lntegration
2040
Figure 6-3. Percent Change in Total Nonresidential Sq. Ft. - Tier 2: 20% Increase in Demand Scenario
Over BAD, Pre- and Post-Integration
Integrating the Transportation Sector with Economy. Land, and Energy Sectors
When we integrated the transportation sector, we replaced exogenous population and GRP projections
with endogenous estimates of these variables. Before model integration, we assumed in the
transportation-sector model that population would exactly match the Triangle Regional Model's
projections. Pre- and post-integration outputs for these variables, as well as several other variables in the
transportation sector, are presented in Figure 6-4. So as to avoid calibrating to data and projections that
might be based on inconsistent assumptions, other exogenous projections in the transportation-sector
model before integration (e.g., for employment, dwelling units, nonmotorized travel facilities, and
parking prices) were, as often as possible, derived from the TRM. Because population in the integrated
model is driven by feedback loops for births, deaths, and migration, its growth follows a slightly
exponential trajectory, while the TRM projections reflect more linear growth (Figure 6-4A). Meanwhile,
the pre-integration exogenous projection of inflation-adjusted GRP per capita (which was based on the
BEA's estimates of the average annual rate of change for the Durham-Chapel Hill, NC Metropolitan
Statistical Area) showed higher growth than the post-integration endogenous results, though the overall
trend was otherwise fairly similar (Figure 6-4B). The model integration process also created a balancing
feedback loop wherein traffic congestion reduces GRP, which discourages automobile travel, which
reduces congestion (see Section 3.2). However, the effect of traffic congestion on GRP is small
compared to total GRP, so the impact of this feedback loop is minor.
Baseline travel volumes increase when population goes up, but baseline automobile travel decreases
when GRP goes down. As such, the effects on VMT of the greater population and lesser GRP that
resulted from model integration mitigate one another, hence reducing the net effect on VMT. In part for
this reason, Tier 2 VMT in 2040 in the integrated D-O LRP SD Model is 1.4% less than in the pre-
integration transportation-sector model (Figure 6-4C). Tier 2 VMT in the present, integrated model is
also 8.8% greater than in the pre-integration model in the year 2000, but that is largely due recalibrations
that we performed after updating the assumed lookup table for gasoline prices in the model, rather than
being attributable to the model-integration step. Tier 2 public transit person miles in 2040 are 25.5% less
in the present, integrated model than in the pre-integration transportation-sector model (Figure 6-4D),
largely thanks to changes that we made to the model after integration. If not for these post-integration
model changes, we would expect the present, integrated model to produce higher public transit person
mile values than the pre-integration transportation-sector model.
183
-------
Other variables affecting the transportation sector that became dynamic as a result of the model-
integration process include employment, dwelling units, and earnings by residents, whose pre-
integration lookup tables were all derived from the TRM. We use dwelling units and employment to
calculate the jobs-housing balance, which helps to drive levels of nonmotorized travel. The present,
integrated model's Tier 2 jobs-housing balance only differs from the pre-integration model by an
average of 0.8% in any given year and is 2.3% greater than in the pre-integration model in 2040 (Figure
6-4E). We also use employment per commercial acre to represent demand for parking, and hence drive
parking prices, replacing an unrealistic lookup table for parking prices that we calculated from TRM
data and projections. By 2040, Tier 2 parking prices in the integrated model are 28.7% greater than in
the unintegrated transportation-sector model (Figure 6-4F). Meanwhile, the endogenous variable of
resident per capita net earnings replaces an exogenous lookup table for the purpose of driving demand
for automobiles, and hence the actual stock of vehicles. The endogenous earnings variable increases
more quickly than the exogenous one. By 2040, Tier 2 vehicle stock in the integrated model is 29.2%
greater than in the unintegrated transportation-sector model (Figure 6-4G). Finally, we added a
mechanism whereby nonmotorized-travel-facility-building is driven by developed land, rather than a
lookup table of TRM-derived figures, assuming that demand for sidewalks and bike lanes mostly exists
around developed land, because that is where there are destinations to walk or bicycle to. As a result of
this change, Tier 2 miles of nonmotorized travel facilities (which affect travel by nonmotorized modes)
now increase more quickly and in 2040 are 15.3% greater than in the pre-integration model (Figure
6-4H).
(A) Population
(B) Gross Regional Product
2000 2010
BAU Post-Integration
___ TRMvSSE Data
• BAU Pre-integration
•BAU Post-Integration
•BAU Pre-integration
(C) VMT
— — — Data and Projections
(D) Public Transit Person Miles
BAU Pre-integration
— ——TRMvS Result Data
BAU Post-Integration
— —TRM v5 E+C Result Data
2010
BAU Post-Integration
___TRMvS Result Data
• • National Transit Database
2030 2040
•BAU Pre-integration
TRM v5 E+C Result Data
Figure 6-4. Transportation Sector in Tier 2, Before and After Sectoral Integration
184
-------
(E) Jobs-Housing Balance
(F) Cost of Parking
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
00 2010 2020 2030 20
^— BAU Post-Integration ^— BAU Pre-lntegration
BAU Pre-lntegration
(G) Vehicle Stock
BAU Post-Integration
___ TRMvS SE Data Adjusted
(H) Nonmotorized Travel Facilities
500
•BAU Post-Integration
• U.S. Census
•BAU Pre-lntegration
BAU Post-Integration
___ TRMvSSE Data
•BAU Pre-lntegration
Figure 6-4 (continued). Transportation Sector in Tier 2, Before and After Sectoral Integration
Integrating the Energy Sector with Land and Transportation Sectors
(A)
(B)
140
2000
2010
2020
2030
2040
2000
2010
2020
2030
2040
-BAU - before integration
-BAU - after integration
-Durham City-County Sustainability Office
• BAU - before integration
• BAU - after integration
• Durham County Property Tax Data
Figure 6-5. Energy Model in Tier 2, Before and After Sectoral Integration: (A) Commercial Energy Use
and (B) Commercial Square Footage
Integrating the energy model with other sectors resulted in three main improvements: (1) endogenously
modeled building stock, which reflects economic and land use feedbacks; (2) endogenously modeled
VMT, which reflects economic and transportation feedbacks; and (3) endogenously represented
feedback between the energy system and the economy. These improvements add nuance and detail to
the energy sector projections. For example, before integration, commercial energy use was projected to
have linear growth based on linear growth in commercial square footage (Figure 6-5 A-B, orange line).
185
-------
Before model integration, historical data on commercial energy use and commercial square footage do
not suggest that this linear trend will change (Figure 6-5 A-B, grey dashed line). After model
integration, commercial energy use grows slower after 2014 partly due to slower growth in commercial
square footage after 2010 (Figure 6-5A-B, red line).
As another example, before model integration, vehicle fuel consumption was based on VMT, which
includes effects of population growth and elasticity to fuel price (Figure 6-6A-B, orange line). After
model integration, the growth of vehicle fuel consumption is stimulated by increased GDP per capita,
and includes balancing feedback from traffic congestion, in addition to the population growth, VMT,
and fuel price effects found in the model before integration (Figure 6-6a-b, red line). The integrated
model represents feedback between the transport and economy sectors which was not captured in the
energy model before integration: economic growth (GDP) stimulates VMT, but congestion rises with
VMT, giving balancing feedback to economic growth. Before integration, the energy model projected
VMT by multiplying population by a constant VMT per capita, and applying an elasticity of VMT to
fuel price. The integrated model is less sensitive to the drop in gasoline price in 2008, and VMT
projections from the integrated model better match TRM v5 projections compared to the model before
integration (Figure 6-6b). Vehicle fuel consumption declines after 2009 in the integrated model due to
improvements in vehicle fuel efficiency based on EIA Annual Energy Outlook 2015 projections (Figure
6-6a). Local data were available for VMT, but not vehicle energy consumption, so we can only assess
the accuracy of these fuel consumption projections in terms of their assumptions. To summarize the
benefits of model integration for vehicle energy consumption: the integrated model includes feedbacks
from economic growth and traffic congestion not present in the non-integrated model; and further, the
integrated model has an endogenous formulation for VMT per capita, which was assumed constant in
the non-integrated model.
(A)
(B)
350
300
250
200
0
2000
2010 2020 2030
^^—BAU - before integration
^^—BAU - after integration
2040
2000 2010 2020 2030
^^^BAU - before integration
BAU - after integration
— — —TRMvS Projections
2040
Figure 6-6. Energy Model in Tier 2, Before and After Sectoral Integration: (A) Vehicle Fuel
Consumption and (B) VMT
Integrating the Economy Sector and the Population Component of the Land Use Sector
Integrating the Economy and Population sectors in Tier 1 resulted in the discovery of the need for a
balancing feedback between unemployment and net migration. Formerly, Tier 1 migration was driven
exclusively by changes in desired employment, as restricted by available dwelling units. However, as
can be seen in Figure 6-7, when run on top of the BAU scenario, the unemployment rate rises
186
-------
continuously and unreal!stically in Tier 1. In the Light Rail + Redevelopment scenario, unemployment
drops below zero, since without a link, employment rises with the economy, reducing unemployment.
The result is that before the integration was made, the BAU policy scenario was far worse in terms of
unemployment and poverty rates relative to the Light Rail + Redevelopment.
20:
155
-55
2C50 2005 2010 2015 2020 2025 2030 2035 2040
Redev No Unemployment Effect
LRT + Redev
BAU No Unemployment Effect
BAU
Figure 6-7. Unemployment Rate - Tier 1: Unemployment Effect on Migration in Tier 1 Removed
The addition of the effect of unemployment on migration in Tier 1 represents reality more truly by
including a balancing feedback to migration - as unemployment drops very low in an area, more people
move to the area for jobs, which raises the resident population, and increases unemployment until it
reaches a new equilibrium. Table 6-4 displays the average yearly percent departure from the data, in
comparison to the BAU, for the years data are available (2000-2014). The addition of the unemployment
effect on migration in the BAU scenario improves the model fit with data for the Tier 1 unemployment
rate and percent of the population in poverty. Even where the change worsens fit with the data
historically, the effect is minor in comparison to the improvement in the model trends shown in Figure
6-7.
Table 6-4. Average Yearly Percent Departure from Data, 2000-2014: Unemployment Effect on Migration
Tier 1 Removed
Average percent deviation from data for the
years 2000, 2005-2014
BAU v Data
Unemployment
Effect Removed
Tier 1
Unemployment rate Tier 1
Population Tier 1
Percent of population in poverty Tier 1
-17.3%
-7.6%
-1.8%
-19.2%
-3.9%
-2.7%
Tier 2
Unemployment rate
Population
Percent of population in poverty
0.8%
0.7%
9.5%
0.3%
1.4%
9.3%
187
-------
Model Input Characterization and Assessment
As part of the quality assurance process in developing the D-O LRP model, we assessed the quality of
model inputs, both data sources and any methods used to manipulate them for use in the model. The
results of this assessment are a qualitative ranking of the level of confidence in the model input (high,
medium or low) and a description of the associated uncertainties. The process for determining the level
of confidence in an input is based on applying a weight of evidence approach to the following criteria:
• Is the input based on information from one or more externally peer reviewed documents?
• Is there agreement in the literature or within the relevant community of practitioners about the
underlying data or method for the input? Or are there conflicting viewpoints?
• Do the characteristics of the input make it suitable for use in the context of the study area? For
example, is the input based on data from either Durham or Orange Counties or another area with
similar characteristics; or is it based on data from another area that is highly site-specific?
• If the model input is based on manipulation of a data set, is the method used in developing the
input an established and widely applied approach? If the method applies equations developed
from external models, are those equations applied to local data in an appropriate manner?
A sample of the assessment results is presented in Table 6-5 (the full table can be found in Appendix C).
For each input, we provide the source, how the input is used in the model, the rationale for selecting the
input, and the confidence level in the input, along with a description of uncertainties associated with it.
188
-------
Table 6-5. Selected Variables from Model Input Characterization Table
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Acres of developed, vacant,
agricultural, and protected
open space land in 2000
TJCOG. 2014. "CommunityViz 2
(CV2) Parcel Geodatabase for
Place Type & Development Status
Editing." Place type and
development status of parcels
calculated for the year 2000 by
back-casting 2013 per-capita
values.
Initial values for acres by category,
which affects developed land.
Created in part to inform the LRP
process; best inventory of current land
use available for entire study area.
MEDIUM-HIGH: Each parcel was
reviewed by local planning staff, but
values do not quite match local
comprehensive plan estimates, at
least in Durham for which these were
available.
Elasticity of public transit
travel to fare price
McCollom, Brian E., and Richard
H. Pratt. 2004. "TCRP Report 95:
Traveler Response to
Transportation System Changes:
Chapter 12—Transit Pricing and
Fares." (page 12-9)
Used to determine how much
changes in the average public transit
fare price affect person miles of
travel by public transit. Greater public
transit use reduces VMT (and hence
congestion, fuel consumption, and
traffic accidents) and helps drive
nonmotorized travel, and hence
physical activity.
Source adapted this elasticity from the
Simpson & Curtin formula, which is
commonly used among public transit
planners. Also, TCRP reports are well-
regarded in their own right.
MEDIUM-HIGH: Not local.
VMT (calibration)
DCHCMPO. 2013. "Triangle
Regional Model version 5: Travel
demand result shapefiles." (TRM)
Used to estimate traffic congestion,
fuel consumption by vehicles, and
traffic accidents.
The TRM is the primary source of
VMT and traffic congestion projections
used by local transportation planning
agencies; spatial nature allows
clipping to both Tiers.
HIGH: Authoritative source with
straightforward application to the
study area, but the TRM only models
weekday traffic. We assume that
VMT on a weekend day is the same
as VMT on a weekday.
Carbon dioxide (CO2)
emissions (calibration)
(1) Durham City-County
Sustainability Office. 2015.
(2) Freid, Tobin. Email message to
authors on January 9, 2015.
CO2 emissions is an endpoint
indicator variable.
Authoritative local source; emissions
are calculated by the Sustainability
Office based on energy data supplied
by utility companies.
MEDIUM: For buildings, only
emissions from electricity and natural
gas (which represent the large
majority of energy use in buildings)
are currently tracked.
Total water demand
(calibration)
NC State Data Center. 2015.
"LINC: Log Into North Carolina."
Used to determine withdrawals from
water reservoirs and calculate
energy used by the municipal water
system.
Authoritative government source with
multiple time points.
HIGH: Authoritative source for
historical water demand data.
Reduction in mortality per
person mile of walking for
transportation per day per
capita; reduction in mortality
per person mile of cycling
for transportation per day
per capita
WHO. 2014. "Health economic
assessment tools (HEAT) for
walking and for cycling:
Methodology and user guide, 2014
update."
Used to calculate avoided premature
mortalities due to changes in the
amount of walking or cycling for
transportation per day per capita.
The HEAT model equations are
simple and based off of many
epidemiological studies and
associated correlations between
walking/cycling and health benefits.
MEDIUM-HIGH: The source's
recommended applicable age range
for walking is 20-74 and for cycling is
20-64, but we applied it to the
average rate of walking and cycling
for transportation over the entire
population. Also, the accuracy of the
HEAT calculations should be
understood as estimates of the order
of magnitude of the expected effect
rather than the precise effect.
189
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Shares of total employment
by employment category
(industrial, office, retail,
service)
(1) Tier 2: Historical data (2000-
2011) and projections (2012-2040)
from: Woods & Poole Economics,
Inc. Copyright 2014. "Durham and
Orange County, NC Data
Pamphlet."
(2) Tier 1: Historical (2002-2009):
U.S. Census Bureau. 2015.
"LODES Data. Longitudinal
Employer-Household Dynamics
Program."
(3) Tier 1: Historical data (2010)
and projections (2011-2040) from:
DCHCMPO. 2013. "Triangle
Regional Model version 5:
Socioeconomic data and
projections for the preferred
growth scenario."
Multiplied by total employment in the
model to determine the total number
of jobs by employment category
which are used to calculate "total
earnings," a component of "GRP,"
and are multiplied by employee
space ratios for each employment
category to determine "total
nonresidential sq ft."
(1) Woods & Poole's historical data
are from the BEA, but they fill in gaps
in certain employment categories that
the BEA omits. Projections were used
by the DCHC MPO to help create
employment guide totals for the
CommunityViz modeling that
generated the output for the TRM v5
SE data.
(2) and (3) Only sources available for
historical and projected employment
by category at a small enough
geographic scale for Tier 1.
MEDIUM-HIGH for Tier 2: High
confidence level for historical data
since its original source is the BEA,
and medium-high confidence level for
projections due to the inherent
uncertainty associated with future
projections.39
MEDIUM-LOW for Tier 1: LODES
data has built-in noise that distorts
the data on small scales, and
changes in census block group
geographies between 2002 and 2009
also caused inconsistencies in the
data in Tier 1. Also, TRM v5 SE
data's definitions of employment
categories differed from those used
in the model.
39 Woods & Poole does not guarantee the accuracy of this data. The use of this data and the conclusions drawn from it are solely the responsibility of the US EPA.
190
-------
Model Input Uncertainty Characterization
As an extension of characterizing model inputs, we summarized the degree of confidence in each input.
Table 6-6 provides a qualitative scale of confidence for model inputs (summarized by model sector) that
is meant to complement the quantitative validation and sensitivity analyses presented later in this
section. In addition, this uncertainty characterization can be used by model users to target specific model
inputs for scenario-specific sensitivity analyses in order to determine the extent to which uncertainty in
their values can affect scenario results.
Table 6-6. Model Input Uncertainty Characterization Summary by Model Sector40
MODEL SECTOR
Land Use
Transportation
Energy
Economy
Equity
Water
Health
Total
SUMMARY OF DATA INPUTS (COUNTS OF INPUTS BY LEVEL OF CONFIDENCE)
HIGH
2
5
-
6
3
5
-
21
MEDIUM-
HIGH
1
13
4
10
-
-
3
31
MEDIUM
8
12
8
3
6
8
1
46
MEDIUM-
LOW
1
5
-
1
4
-
2
13
LOW
3
10
-
-
1
-
-
14
TOTAL
15
45
12
20
14
13
6
Validation of Historical Simulations
The calibration of the model starts with the evaluation of the results generated when using specific
parameters or equations obtained from literature. This is accomplished through the comparison of
historical data and the results of the baseline simulation.
The examples provided below show the D-O LRP SD Model Business-As-Usual simulation (blue line)
40 This table reflects the uncertainty characterization for Tier 2 inputs, except in cases where an input applied only to Tier 1.
In a few cases where an input applied to both Tiers, inputs for Tier 1 had different (often lower) confidence levels assigned to
them.
191
-------
and historical data (red line) for the period 2000 - 2014. The model starts simulating in 2000, and runs
differential equations to project results for subsequent years; it does not use historical data to generate
projections. Therefore, the modeling team was able to use historical data to check whether the structure
of the model is capable of reproducing the historical observed behavior.
Population
Population is driven by a combination of birth, death, and migration rates. In Tier 2, calibration was
achieved through the combination of a slightly declining birth rate, constant death rate, and net
migration, which was linked by an inverted-U-shaped function to the availability of residential land. In
Tier 1, the team chose to calibrate more closely to the historical population trend from the U.S. Census
Bureau, thus the BAU projection does not match the TRM projection in later years. The birth and death
rates in Tier 1 were assumed to be identical to those used for Tier 2 (no data were available for such a
small area), and net migration is driven by a link with the demand for employment and the
unemployment rate, capped by the availability of dwelling units. Population in Tier 2 is calibrated very
well and is within 5% of the TRM v5 SE projection in the year 2040 (Figure 6-8). In Tier 2, the R2
equals 0.99 for the projection and 1.00 for the data, while in Tier 1, the R2 equals 0.99 for the projected
data and 0.87 for the historical data. This small difference is due to the more realistic exponential growth
pattern rather than a linear one. Population in Tier 1 varies by no more than 3.5% from the U.S. Census
data between 2000 and 2014. While population in BAU is about 25% lower than the TRM population
projection in 2040, it is much more closely matched under the Light Rail scenario.
(A) Tier 2
(B) Tier 1
o
— -— •"•"*
00 20
„*****"*"
10 20
— '
20 20
^t^f^^^^
****^
30 20
80
B 70
3- 60
*• 50
i 40
1 30
: 20
2000
2010
2020
2030
2040
.BAU ———TRM v5 SE Data ••
U.S. Census Bureau
-BAU — — —TRMvSSE Data
U.S. Census Bureau
Figure 6-8. BAU Scenario vs. Historical Data for Population
Developed Land
Since there were no explicit projections of land development, developed land could not be calibrated
with great confidence. The current value was derived from the CV2 Parcel Geodatabase for Place Type
& Development Status Editing by summing all parcels assigned a development status of developed,
minus parcels assigned a place type of protected open space. The team estimated total developed land in
2040 in two ways. The first estimate (Estimate 1) applies average floor area ratios and densities for each
land use type to the employees and households added in the preferred growth scenario Imagine 2040
Grid output file, and added this to the acres of developed land in 2013. The second method (Estimate 2)
calculates the acreage of grids with any allocation of employment or households. In Tier 1, Estimate 2
was in fact above the total parcel acreage (excluding protected open space), so we used that value as the
maximum threshold. In both tiers, the BAU scenario is just above the lower estimate in 2040, leaving
room for reasonable growth in subsequent scenarios (Figure 6-9).
192
-------
In both Tiers, the R2 equals 0.99 for both estimates. In 2040, developed land deviates by -8.1% and by
1.9% from Estimate 1 and Estimate 2 in Tier 2, respectively, and in Tier 1 by 4.4% and -27% from
Estimate 1 and Estimate 2, respectively.
(A) Tier 2
2000
2010
2020
2030
• BAU — — — Estimate 1 • •• Estimate 2
(B) Tier 1
•BAU — — — Estimate 1 •
Estimate 2
Figure 6-9. BAU Scenario vs. Historical Data for Developed Land
Nonresidential Sa. Ft.
To create a historical data series for the Tiers, data from three counties had to be combined. Most years
were available for Durham County, and this accounts for the majority of nonresidential sq ft in both
Tiers. For Orange and Chatham counties, we back-cast data from the 2014 estimate from the county tax
administration databases using per capita rates from each county. Therefore the speed of growth shown
in the data is somewhat uncertain. Since Orange and Chatham County also did not have detailed land
use codes, we developed an allocation weighting scheme, which is described in the Land Use Sector
description in Section 3.3, to arrive at estimates for each subcategory of square feet, including retail,
office, service, and industrial. Employee space ratios for each subcategory were calculated from the final
estimates and from the historical data on employment. In the model, employment and employee space
ratios drive demand for development of nonresidential sq ft, leading to the close calibration seen below.
Calibration of the subcategories of nonresidential square feet are similarly close, with somewhat more
delays and swings visible due to the delays and feedbacks inherent in the model structure.
Total nonresidential square feet deviates by no more than 8.5% in Tier 2 from the estimate derived from
the County Office of Tax Administration databases, and by no more than 2% in Tier 1 (Figure 6-10). In
Tier 2, the R2 equals 0.95 for the data, while in Tier 1, the R2 equals 0.99 with the data.
(A) Tier 2
(B) Tier 1
• BAU — — —County Offices of Tax Administration
• BAU — — —County Offices of Tax Administration
Figure 6-10. BAU Scenario vs. Historical Data for Nonresidential Sq. Ft.
193
-------
Dwelling Units
Dwelling units, shown in Figure 6-11 and Figure 6-12, were calibrated in the BAU scenario to closely
match the historical data (2000, 2010) and projection (2014, 2019) estimates for total dwelling units
obtained from Community Analyst, following the methodology explained in Section 3.1. Total dwelling
units were separated in the model into single family and multifamily stocks. The percent of dwelling
units that are single and multifamily was obtained from Decennial Census 2000; SF3 DP4 and the ACS
estimate 2008-2012, as clipped by Community Analyst.
Calibration of single and multifamily dwelling units to the historical data and projections used initial
values of households by category, with four factors added to approximate adequate vacancy to match the
data for dwelling units. First, an average housing lifetime simulates the necessary turnover due to natural
degradation. Second, the percent of dwelling units that are second homes was added to single family
properties, boosting the needed construction in excess of households. Third, an endogenous calculation
of population growth over the next five years is used to begin construction in anticipation of demand.
Finally, an effect of vacancy on equilibrium dwelling units was added. This table was made as a gentle L
shape with a long tail - indicating that if vacancy is very low, there is a boost to the demand for dwelling
units. As vacancy increases, demand gradually declines and eventually has a slightly negative effect.
The shape of this table follows basic economic supply and demand, however the specific values were
calibrated after the previous factors were established.
Dwelling units deviate by no more than 1.5% in Tier 2 and 3.2% in Tier 1 in any given year. In Tier 2,
the R2 equals 0.99 for both single family and multifamily, while in Tier 1, the R2 equals 0.95 and 0.99,
for single family and multifamily, respectively.
(A) Tier 2
(B) Tier 1
— — — ESRI 2010 Census Profile
— — — ESRI 2010 Census Profile
Figure 6-11. BAU Scenario vs. Historical Data for Single Family Dwelling Units
30
(A) Tier 2
(B) Tier 1
— — — ESRI 2010 Census Profile
- — — ESRI 2010 Census Profile
Figure 6-12. BAU Scenario vs. Historical Data for Multifamily Dwelling Units
194
-------
Total Impervious Surface
To estimate impervious surfaces in the model, we used impervious surface coefficients (ISCs) for
residential land use by density as reported in the Jordan Lake Stormwater Accounting Tool User's
Manual (NC State Bio & Ag Engineering and NCDENR, 2011). However, coefficients for
nonresidential land and roads had to be obtained from the User's Guide for the California Impervious
Surface Coefficients, whose applicability to the local region is unknown. Therefore some calibration
was necessary. While in Tier 1, mean values for nonresidential uses and coefficients for highway and
rural roads were used as given, in Tier 2, to reach the 2010 value for total impervious surfaces, the
nonresidential coefficients were reduced to 85% of the mean values. Urban roads had to be calibrated
down from .91 to .7 in Tier 2 and to .8 in Tier 1. The ISC for nonmotorized travel facilities was not
addressed, and therefore was assumed to be 1.
Total acres of impervious surfaces are calibrated to within 2% of the one-meter land cover data from
EPA EnviroAtlas for 2010 in Tier 2, and to within 1% in Tier 1 (Figure 6-13). Since there is only one
data point, an R2 could not be obtained, and uncertainty remains regarding the speed of change in
impervious surfaces over time.
(A) Tier 2
(B) Tier 1
0
0
0
0
0
0
0
20
^^^^^^^«
^^^^^^
— —
.
00 2002 2004 2006 2008 2010 2012 20
45
40
35
30
25
20
15
10
5
0
-EPA EnviroAtlas
-EPA EnviroAtlas
Figure 6-13. BAD Scenario vs. Historical Data for Impervious Surfaces
VMT
The largest drivers of VMT are population and GRP. Other important drivers include the disincentives
to drive that come from traffic congestion (endogenous, forming a balancing feedback loop) and
gasoline prices (exogenous, determined by a lookup table). Some other, less impactful drivers include
parking prices, population density, and intersection densities. During the model-building process,
changes occasionally resulted in the values of these various drivers, requiring us to adjust the model
inputs for initial-year travel volumes (person miles) for the sake of keeping VMT calibrated to the
Triangle Regional Model's data and projections.
Version 5 of the TRM includes both a projection of future travel behavior under an official "Preferred"
infrastructure-building scenario generated for the MTP and a projection of future travel behavior in an
"Existing + Committed" infrastructure-building scenario. As discussed in Section 5.2, neither of these
195
-------
projections is entirely a match to the conditions assumed in the D-O LRP SD Model's BAU scenario.41
These two TRM-generated scenarios only diverge from one another after the year 2017. Therefore, in
the D-O LRP SD Model, we calibrated VMT principally to the values indicated by the TRM for the
years 2010 and 2017, with a lower priority placed on matching it to the TRM 2040 projections. 2010,
2017, and 2040 are the only years for which the TRM provides VMT figures.
In the BAU scenario, Tier 1 and Tier 2 VMT both have an R2 value of 0.999 with the MTP "Preferred"
infrastructure case and an R2 value of 0.997 with the "Existing + Committed" infrastructure case. In
large part, these high R2 values are due to them being based on only three points in time (2010, 2017,
and 2040), in addition to the TRM's numbers having been used to calibrate VMT in this model. Tier 1
VMT deviates from the TRM's figures by +0.4% in 2010 and -2.4% in 2017; in 2040, it is 8.0% less
than what the TRM projects in the Metropolitan Transportation Plan "Preferred" infrastructure case and
2.7% less than what the TRM projects in the "Existing + Committed" infrastructure case. Tier 2 VMT
deviates from the TRM's figures by a margin of less than 0.1% in 2010 and by -2.6% in 2017; in 2040,
it is 0.6% less than what the TRM projects in the MTP "Preferred" infrastructure case and 2.2% less
than what the TRM projects in the "Existing + Committed" infrastructure case.
(A) Tier 1
(B) Tier 2
2.5
1.5
2000 2010 2020 2030 2040
BAU — — — TRM v5 Result Data — —TRM v5 E-t-C Result Data
2000 2010 2020 2030 2040
BAU — ——TRM v5 Result Data — —TRM v5 E-t-C Result Data
Figure 6-14. BAU Scenario vs. Projections for VMT
Congestion
In both Tiers, congestion is primarily driven by the ratio of VMT to functioning roadway lane miles,
with an exogenous lookup table determining how much congestion results from a given amount of
weekday peak-period VMT per lane mile. In Tier 1, congestion is also driven, to a lesser extent, by
commercial floor area ratios. Since roadway lane miles are determined by exogenous policy inputs,
VMT is the primary endogenous determinant of traffic congestion. Therefore, if VMT is well-calibrated,
congestion is usually also well-calibrated, given that they are both calibrated to the TRM's projections.
As discussed in Section 5.2, neither the TRM's Metropolitan Transportation Plan "Preferred"
infrastructure-building scenario nor its "Existing + Committed" infrastructure-building scenario is
entirely a match to the conditions assumed in the D-O LRP SD Model's BAU scenario. Furthermore, the
TRM does not report 2040 peak-period traffic speeds for the "Existing + Committed" scenario.
Therefore, only the "Preferred" infrastructure scenario could be used to calibrate traffic congestion in
41 The BAU scenario assumes no light rail is built and roadway lane miles will continue to be built after 2017, while the TRM
"Preferred" scenario assumes that a light rail is built, and the TRM "Existing + Committed" scenario assumes that no further
roadway lane miles will be built after 2017.
196
-------
the D-O LRP SD Model. Regardless, since the major infrastructure-building assumptions of the TRM-
generated scenarios only diverge from those of the BAU case after the year 2017, we calibrated
congestion principally to the values indicated by the TRM for the year 2010, with a lower priority placed
on matching it to the available TRM MTP "Preferred" infrastructure scenario 2040 projections. 2010
and 2040 are the only years for which the TRM provides congestion figures.
Because there are only two points in time at which the BAU case can be compared to data or projections
(2010 and 2040), an R2 value cannot be calculated (or, rather, the R2 value would automatically be one).
The availability of data and projections for only two points in time also obscures whether the trend of
the data and projections is linear, exponential, or otherwise. Therefore, the fact that the BAU congestion
forecasts in Figure 6-15 do not match the seemingly linear trend of the data lines is not meaningful. Tier
1 congestion deviates from the TRM's figures by +2.2% in 2010 and -6.3% in 2040. Tier 2 congestion
deviates from the TRM's figures by a margin of less than 0.1% in 2010 and by +0.6% in 2040.
(A) Tier 1
(B) Tier 2
1.25
1.00
2010 2020 2030 2040
BAU - - -TRM v5 Result Data
2010 2020 2030 2040
_ BAU — ——TRM v5 Result Data
Figure 6-15. BAU Scenario vs. Projections for Congestion
Public Transit Person Miles
The largest endogenous drivers of public transit person miles are population (goes up as population goes
up) and GRP (goes down as GRP per capita goes up). Like all other modal person miles, public transit
person miles are significantly affected by changes in gasoline prices per mile of automobile travel
(which are determined by an exogenous lookup table), as well as by the endogenous input of traffic
congestion, given that congestion affects fuel efficiency. Some other, less impactful drivers include
parking prices, population density, and intersection densities. In addition, public transit person miles are
driven to a significant degree by the exogenous policy variables of public transit fare prices and public
transit vehicle revenue miles. Meanwhile, in scenarios where a light rail line is built (unlike the BAU
case), Tier 1 population and employment and Tier 2 VMT per highway lane mile assume additional,
significant influence over public transit person miles per day.
In Tier 2, both historical data and projections were available during the calibration of public transit
person miles. The primary basis for the calibration of this variable was historical data from the National
Transit Database (NTD), which provided yearly data for the period 2000-2013. We placed top priority
on calibrating to the most recent historical data, for 2013, so that the D-O LRP SD Model would project
forward from an accurate representation of present-day conditions, as opposed to matching data from
earlier years and extending a trend from that data that does not recreate the present. During 2000-2013,
Tier 2 public transit person miles in the BAU case have an R2 value of 0.81 with NTD data. The average
percent deviation above or below NTD data during this period is 9.1% and the percent deviation in
197
-------
2013, the last year for which NTD data were available, is -0.4%. The yearly percent deviation ranges
from +20.4% in 2000 to -17.1% in 2008.
The only years for which the MTP provides figures on public transit use are 2010 and 2040, so, again,
we could not calculate an R2 value for projected values of this variable. In 2010, Tier 2 public transit
person miles in the BAU case deviate from the TRM-generated scenarios by -1.6%. In 2040, they
deviate from the TRM-generated MTP "Preferred" scenario by -15.5% and from the "Existing +
Committed" scenario by +21.6%, meaning the 2040 value is in between the two projections, which
differ from one another by a wide margin. Furthermore, since historical data from the NTD was the
primary basis for calibrating this variable, it would not be surprising for there to be some amount of
deviation between public transit person miles in the BAU case and in the TRM-generated scenarios.
In Tier 1, only one data point, for the year 2010, was available for calibrating public transit person miles
per day, derived from figures reported in the Durham-Chapel Hill-Carrboro Metropolitan Planning
Organization's Metropolitan Transportation Plan (MTP) and scaled down to Tier 1 with the help of
information from version 5 of the Triangle Regional Model (TRM v5). For this reason, it was not
possible to calculate an R2 value for this Tier. Tier 1 public transit person miles in 2010 in the BAU case
are 0.4% greater than this data point. Since the MTP-derived 2010 data point was the sole basis for
calibrating Tier 1 public transit person miles, the percent deviation of the BAU from it is small.
(A) Tier 1
(B) Tier 2
35
30
15
10
0
t
-S
^^
r-
—
-^-^— ^-—
00 2010 2020 2030 20'
-TRM v5 Result Data
BAU
—TRMvS E-t-C Result Data
__—TRMvS Result Data
National Transit Database
Figure 6-16. BAU Scenario vs. Data and Projections for Public Transit Person Miles
Building Energy Use
2000
2010
2020
2030
2040
•BAU
Figure 6-17. Building Energy Use - Tier 2: BAU Model Output Compared to Data
198
-------
Historical data for building energy use in Tier 2 are from the Durham City-County Sustainability Office
(Figure 6-17), originally sourced from local utility companies. Historical building energy use was
scaled from Durham County to Tier 2 using the fraction of Tier 2 residential, commercial, and industrial
square feet found in Durham County. Year-to-year variations in the data are likely due to a variety of
factors, some of which are included in the model (e.g. population and sq ft) while others are not (e.g.
temperature fluctuations which affect heating and cooling energy use).
In the model, building energy use is driven by the stock of dwelling units and the stock of nonresidential
square feet, multiplied by the energy use intensity of each. The stock of single-family and multifamily
dwelling units, as well as nonresidential square feet are each calibrated to historical data. Energy use
intensities for the residential, commercial, and industrial sectors are also calibrated to historical data and
projections (which are based on EIA Annual Energy Outlook 2015 projections (US EIA 2015a)).
Building energy use is calibrated to an average annual deviation of 2.0% from historical data (2006-
2013),42 and modeled building energy use matches historical data with R2 equal to 0.87.
2010
2020
2030
2040
•BAD
— — — Durham Sustainability Office data scaled to Tier 2
Figure 6-18. CC>2 Emissions - Tier 2: BAD Model Output Compared to Data
Historical CO2 emissions data are from the Durham City-County Sustainability Office (Figure 6-18),
based on building energy consumption data sourced from local utility companies, and based on local
VMT data for vehicles. Historical CO2 emissions were scaled from Durham County to Tier 2 using the
fraction of Tier 2 building square feet found in Durham County (0.73). This scaling approach was
chosen because buildings represent the majority of regional CO2 emissions, with vehicles representing
most of the remainder. As with the energy use data, year-to-year variation in the CO2 emissions data
may have a variety of causes, including increases in building square footage and VMT, which are
modeled, and year-to-year temperature fluctuations, which are not modeled.
Modeled CO2 emissions are the sum of building and vehicle emissions, as well as emissions from
municipal water treatment and distribution. Vehicle emissions include buses, passenger vehicles, and
light rail. Within each of these categories, CO2 emissions are the product of energy use and a fuel-
specific emissions factor. For example, burning a gallon of gasoline in a car is assumed to produce
0.00889 metric tons of CO2 (US EPA 2015b).
42 This is the average absolute value of annual % deviation from data. Omitting absolute value, model projections are on
average 0.91% higher than data.
199
-------
CO2 emissions are calibrated to an average annual deviation of 4.1% from historical data (2006-2012),43
and modeled CO2 emissions match historical data with R2 equal to 0.81.
Total Employment
Total employment for Tier 2 and Tier 1, shown in Figure 6-19, was calibrated in the BAU scenario to
closely match total employment data (2010) and projections (2011-2040) from the TRM v5 SE data
files, clipped in ArcGIS to the two geographies. For the preceding model years, 2000-2009, annual
employment growth trends from the U.S. Bureau of Economic Analysis (BEA) for Durham and Orange
County for Tier 2 and from the U.S. Census LODES dataset (clipped to Tier 1 geography) for Tier 1
were used to back-cast total employment from the TRM v5 SE data 2010 value for total employment.
The R2 equals 1.0 for both sources of data and projections for both Tier 2 and Tier 1, with the Tier 2
average annual deviation being 0.27% between 2000 and 2009, and 0.35% between 2010 and 2040, and
the Tier 1 total employment average annual deviation being 0.61% between 2000 and 2009, and 0.16%
between 2010 and 2040.
Total employment (see Figure 6-19) is driven in the D-O LRP SD Model by total retail consumption,
which is also calibrated to fit historical data and projections, and the exogenous input "employment per
dollar of consumption," which was calculated based on historical data and projections for total
employment and retail consumption in both Tier 2 and Tier 1 and adjusted slightly to improve the model
fit.
(A) Tier 2
• -U.S BEA Historical Employment Growth Trend (2000-2009)
TRMvS SE Data Et Projections (2010-2040)
(B) Tier 1
• LODES Historical Employment Growth Trend (2000-2009)
TRMvS SE Data Et Projections (2010-2040)
Figure 6-19. BAU Scenario vs. Historical Data and Projections for Total Employment
43 This is the average absolute value of annual % deviation from data. Omitting absolute value, model projections are on
average 0.02% higher than data.
200
-------
Total Earnings
Since total employment in the model for Tier 2 and Tier 1 was calibrated to historical data and
projections to match the total employment from the TRM v5 SE data, and the total employment numbers
in the TRM v5 SE data were lower than total employment numbers from historical data and projection
sources for Durham and Orange County combined (e.g., BEA and Woods & Poole Economics, Inc.),
alternative calculations of total earnings were done to calibrate the model. Woods & Poole provide
employment and average earnings per year, both historical data and projections, for each of the 20 job
types that are categorized by the North American Industrial Classification System (NAICS) 2-digit code
classification. We aggregated these 20 jobs types into four categories: industrial, office, retail, and
service, and multiplied the total employment from the TRM v5 SE data with the share of employment
that fell into each category for the entire study period (2000-2040), shown in Figure 6-20a. We then took
the average earnings per employee per year in each of the four categories, shown in Figure 6-20b and
multiplied them by the number of jobs in each employment category for each year and summed them up
to get total earnings.
(A) Tier 1
(B) Tier 2
30%
2000 2005 2010 2015 2020 2025 2030 2035
2000 2005 2010 2015 2020 2025 2030 2035 2040
•Industrial Share of Employment
•Retail Share of Employment
•Office Share of Employment
-Service Share of Employment
•Industrial Earnings per Employee
•Retail Earnings per Employee
•Office Earnings per Employee
•Service Earnings per Employee
Figure 6-20. Historical Employment Data and Projections Used as Inputs in the Tier 2 Economy
Model for (A) Shares of Employment by Category, and (B) Average Earnings Per Employee by Category
The same methodology that was used to calculate total earnings for Tier 2 was applied to Tier 1, but
since employment by category and earnings by category from Woods & Poole are only available at the
county level, a hybrid approach was used to calculate both. For Tier 1 employment by category, shown
in Figure 6-2la, the percent of employment by category from the TRM v5 SE data in Tier 1 compared to
Tier 2 (also from the TRM) was multiplied by the total amount of employment by category calculated
from Woods & Poole data for Tier 2. For earnings per employee by category in Tier 1, the same average
earnings per category were taken from Tier 2, but weighted by the number of jobs per category in the
Durham and Orange County portions of Tier I.44
Figure 6-22 shows the B AU scenario model output for (A) Tier 2 total earnings and (B) Tier 1 total
earnings compared to the total earnings calculated from data and projections. Because total employment
44 This was also done in Tier 2, where industrial earnings per employee were much higher in Durham County than Orange
County due to the large number of high wage jobs classified as manufacturing located in RTF, NC (the outskirts of Durham
County).
201
-------
is the only direct endogenous connection affecting earnings in the model, the fit for total earnings is
about the same as total employment, with the R2 value being 1.0 for both Tier 2 and Tier 1 for calculated
historical and projected earnings and the average % deviation being 0.26% and 0.38% for Tier 2
calculated historical and projected earnings, respectively, and 0.54% and 0.19% for Tier 1 calculated
historical and projected earnings, respectively.
(A) Tier 1
70%
(,0%
5CK
40%
—
2000 2005
2010
2015 2020 2025 2030
2035
2040
• Industrial Share of Employment Tier 1
• Retail Share of Employment Tier 1
e Share of Employment Tier 1
ce Share of Employment Tier 1
(B) Tier 2
2000 2005 2010 2015 2020 2025 2030 2035
2040
•Industrial Earnings per Employee Tier 1
•Retail Earnings per Employee Tier 1
•Office Earnings per Employee Tier 1
•Service Earnings per Employee Tier 1
Figure 6-21. Historical Employment Data and Projections Used as Inputs in the Tier 1 Economy
Model for (A) Shares of Employment by Category, and (B) Average Earnings Per Employee by Category
(A) Tier 2
2000 2010 2020 2030
BAU
— — -Calculated Historical Earnings
• • Calculated Projected Earnings
2040
(B) Tier 1
- — -»
•> "~
00 20
^»***^
10 20
********
20 20
.^
****^
30 20
— -Calculated Historical Earnings Tier 1
• 'Calculated Projected Earnings Tier 1
Figure 6-22. Total Earnings: Model Fit to Calculated Historical Data and Projections
202
-------
Gross Operating Surplus and Gross Regional Product
To calibrate Gross Regional Product (GRP) in the model, GRP was calculated for 2000-2013 using a
methodology from the BEA (Panek et al. 2007) from the calculated earnings data for Tier 2 and Tier 1.
The difference between GRP and total earnings was considered the gross operating surplus (GOS).
GOS is the sum of three factors in the model: (1) GOS per sq ft (multiplied by total nonresidential sq ft),
(2) productivity loss by road congestion, and (3) profit gain or loss due to energy spending (relative to
GRP).45 GOS per sq ft is an exogenous lookup table and was calculated externally by dividing the
calculated GOS for Tier 2 and Tier 1 (calculated from endogenously calculated GRP and total earnings)
by total nonresidential sq ft for Tier 2 and Tier 1 under the BAU scenario, then adjusted so that the
effects of congestion and energy spending were taken into account.
Figure 6-23 shows the BAU model output for GOS and GRP for Tier 2 (Figure 6-23A) and Tier 1
(Figure 6-23B). The R2 value for GRP and GOS in both Tier 2 and Tier 1 is 1.0, with the average %
deviation from data for Tier 2 being 0.32% for GOS and 0.28% for GRP, and the average % deviation
from data for Tier 1 being 0.45% for GOS and 0.33% for GRP.
(A) Tier 2
30
in 25
= 5
2000
2002
2004
2006
2008
2010
2012
• GOS BAU
• COS Calculated from Data
• GRP BAU
GRP Calculated from Data
(B) Tier 1
2000 2002 2004
^—GOS Tier 1 BAU
1 — -GOS Tier 1 Calculated from Data
2006 2008 2010 2012
^^GRP Tier 1 BAU
• -GRP Tier 1 Calculated from Data
Figure 6-23. Gross Regional Product and Gross Operating Surplus: Model Fit to Values
Calculated from Data
45 The effects on GOS of removing the connections to productivity loss due to road congestion and the profit gain or loss due
to energy spending are shown in Figures 6-39 through 6-42.
203
-------
Total Retail Consumption
Three sources of historical retail consumption data were available for reference to calibrate the Tier 2
economy model, all with data for Orange and Durham County. The first data reference source was "total
taxable sales" data for 2000-2014 from the North Carolina Department of Revenue (NC DOR). The
second reference source was Woods & Poole Economics, Inc. with "total retail sales" estimates between
2000 and 2011, with 2002 and 2007 being actual historical data from the U.S. Department of
Commerce. The third reference source was "total retail sales" estimates downloaded from SimplyMap
for 2011-2014 (data were also downloaded from 2008-2010, but was found to be erroneous),
benchmarked from the 2007 Economic Census. Since Woods & Poole's last year of historical data was
2011, we decided to use the more up-to-date data for Tier 2 from NC DOR and apply the annual retail
sales growth rate predicted by Woods & Poole for 2015 through 2040. This reference data and
projections combination is shown by the green dashed line in Figure 6-24. We then calibrated the model
to match (as closely as possible) the most recent reference data point (2014) from this combination by
adjusting the initial value (2000) for retail consumption.
2000 2005 2010 2015 2020 2025 2030
— Woods a Poole Data and Projections
^^^—NC Dept. of Revenue Historical Data
— — — NC DOR w/ WaP Growth Rate
BAU
• • • • SimplyMap
2035
2040
Figure 6-24. Total Retail Consumption - Tier 2: Model Fit to Historical Data and Projections
Two other factors (besides the initial value) were adjusted to calibrate Tier 2 retail consumption: (1) the
elasticity of consumption to GRP, and (2) the resident percent of the working population. Together with
the embedded delay function that delays the impact of relative GRP on retail consumption by two years,
the slow and gradual increase of the resident percent of the working population allows for Tier 2 retail
consumption in the model to reproduce historical trends and extend them into the future, though it is not
able to reproduce the peaks and dips that are shown by the historical data to be typical for the region.
For the historical data (2000-2014), the R2 value for Tier 2 total retail consumption is 0.63 with an
average deviation of 8.8%. For the projections (2015-2040), the R2 value is 1.0 with an average
deviation of 1.6%.
For total retail consumption in Tier 1, only one source of historical retail sales reference data was
available for calibration: SimplyMap, which takes the county-level retail sales data from the 2002 and
2007 Economic Census and uses a computer model to output retail sales at the census block group level
between 2008 and 2014 based on business locations, among other variables. No projections of retail
sales were available for Tier 1. Shown in Figure 6-25, the Tier 1 economy model was calibrated to
204
-------
match the most recent year of retail sales data for Tier 1, 2014, by adjusting the initial retail
consumption value and the elasticity of consumption to GRP, which is equal to the elasticity used for
Tier 2(1.1). The R2 value for the model fit to historical retail sales data for Tier 1 is 0.74 and the
average deviation is 5.3%.
2000 2002 2004 2006 2008 2010 2012 2014
^^^BAU • • • 'Simply Map
Figure 6-25. Total Retail Consumption - Tier 1: Model Fit to Historical Data
Property Values
Estimates for historical property values were derived from the Durham, Orange, and Chatham County
Tax Administration Databases, as described in the Equity Sector of Section 3.3. Orange County Parcel
Database (parview), and Chatham County Tax Parcel Database (ASOUTR). Nonresidential property
values are expressed per square foot, obtained from the same databases. Residential property values are
expressed per single or multifamily dwelling units, obtained from the ESRI 2010 Census Profile clipped
by Community Analyst.
Calibration of single family, multifamily, and nonresidential property values was accomplished through
a combination of a diverse set of elasticities obtained from the literature. A housing crash was also
exogenously introduced, starting in 2005, to better reproduce this global phenomenon and therefore
better fit data on housing costs. This adjustment reduces property values across the board by 15% by
2007, with full recovery by 2009. For a full listing of the elasticities used and explanation of the process,
see the Equity Sector in Section 3.3. In a few cases, these elasticities had to be modified to fit the study
area, as many studies provide elasticities for either one metro area or an average for the nation, and none
were found that were specific to the study area. For details on the modifications made to elasticity
values, see Appendix B.
Because property values are responsive to many factors, including both local and national, using only
endogenous mechanisms available in the model led to a less than ideal fit with historical trends.
Residential property values deviate from the data by an average of 9.2% in Tier 2 and 4.2% in Tier 1 for
single family properties, and 10% in Tier 2 and 2.0% in Tier 1 for multifamily properties (Figure 6-26
and Figure 6-27). Nonresidential property values have more divergence, with an average absolute
deviation of 12% in Tier 2 and 31% in Tier 1 (Figure 6-28). In Tier 2, the R2 equals 0.53, 0.34, and 0.21
for single family, multifamily, and nonresidential property values, respectively, while in Tier 1, the R2
equals 0.46, 0.72, and 0.013, respectively.
205
-------
(A) Single Family
(B) Multifamily
e
£ 200
0
S 150
8
° 100
^
-u
i 50
D
5 n
t
X
j
---^
^r
»-«r:
— —
Q
!| 70
oi 60
Q.
o 50
Q T_
i/i 30
•a 20
c
5 in
SL n
J
X
— ^••w
—
^^^^-^"
2000 2002 2004 2006 2008 2010 2012 2014
BAU — — — County Property Databases
2000 2002 2004 2006 2008 2010 2012 2014
^^^BAU — ——County Property Databases
Figure 6-26. BAU Scenario vs. Historical Data for Tier 2 Residential Property Values
(A) Single Family
(B) Multifamily
Q 10U
to 160
oi 140
Q.
o 120
° 100
" °u
3 60
C AC,
a 40
T ^n
1
J= n
***»•
/
— #^
^r-
Q °"
fe 7n
S /u
Q. 60
!5 30
"^ 20
re
O
-c n
MM^M
• •• *
•^^-^
^^
S*"""«
2000 2002 2004 2006 2008 2010 2012 2014
2000 2002 2004 2006 2008 2010 2012 2014
•BAU
— — — County Property Databases
•BAU
— — — County Property Databases
Figure 6-27. BAU Scenario vs. Historical Data for Tier 1 Residential Property Values
(A) Tier 2
(B) Tier 1
sr 300
i 250
S ?nn
5 ^0°
N 150
= 100
n
/""
— —
^
^~
^ '
. -
-- — .
2000 2002 2004 2006 2008 2010 2012 2014
•BAU
•County Property Databases
2000 2002 2004 2006 2008 2010 2012 2014
• BAU
— — — County Property Databases
Figure 6-28. BAU Scenario vs. Historical Data for Nonresidential Property Values
206
-------
Water Demand
to
T3
2000
2010
2020
2030
2040
• Historical and Projected Data
• BAU
Figure 6-29. Water Demand - Tier 2: BAU Model Output Compared to Data
Historical water demand data for 2000, 2005, and 2010 are from the NC Department of Environment
and Natural Resources (NC State Data Center 2015). Projected data for 2020, 2030, and 2040 are from
the Triangle Regional Water Supply Plan, Vol 1 (Triangle J Council of Governments 2012).
Municipal water demand is the sum of residential, nonresidential, and nonrevenue water demand.
Nonrevenue demand includes uses such as distribution system maintenance (such as line flushing), and
water lost through system leakage. Each of these sectoral demands is calibrated to historical and/or
projected data from the same sources as total water demand. In each case, demand is the product of an
endogenously modeled stock (such as SF housing units) and a water demand intensity factor (such as
water use per SF household).
Municipal water demand is calibrated to an average annual deviation of 0.64% from historical and
projected data46. Modeled water demand matches the combined historical and projected data with R2
equal to 0.999. The close match between modeled water demand and data is due to the assumed trend in
residential water conservation ("conservation improvement residential water use") which is manually
calibrated using an exogenous input.
Summary
Below, in Table 6-7 and Table 6-8, we present the R2 and average absolute percent deviation for each of
the variables discussed above, compared to both data and projections.
46 This is the average absolute value of annual % deviation from data. Omitting absolute value, model projections are on
average 0.55% higher than data.
207
-------
Table 6-7. Statistical Analysis of Tier 1 Variables Compared to Historical Data and Projections
TIER1 MODEL
VARIABLE BAU
SCENARIO
Population
Developed land
Nonresidential
sqft
Single family
dwelling units
Multifamily
dwelling units
VMT
Person miles of
public transit
travel per day
Traffic
congestion
Total
Employment
Total Earnings
Gross Operating
Surplus
Gross Regional
Product (GRP)
Total Retail
Consumption
Single family
property value
Multifamily
property value
HISTORICAL
DATA SOURCE
U.S. Census
Bureau
Not Available
County Office of
Tax Administration
databases
U.S. Census
Bureau
U.S. Census
Bureau
Not Available
Not Available
Not Available
LODES total
employment growth
rate (2000-2009)
applied to TRMvS
SE Data total
employment for
2010
Calculated from
Woods & Poole
earnings per job by
category and TRM
v5 SE data percent
of employment by
category in Tier 1
Difference between
calculated GRP
and Total Earnings
Calculated from
BEA methodology
(Panek et al. 2007)
SimplyMap (2008-
2014)
County Office of
Tax Administration
databases
County Office of
Tax Administration
HISTORICAL
DATA R2
VALUE
0.87
N/A
0.99
0.95
0.99
N/A
N/A
N/A
1.00
1.00
1.00
1.00
0.74
0.46
0.72
AVERAGE PERCENT
DEVIATION FROM
HISTORICAL DATA
1.7%
N/A
0.57%
2.05%
1.73%
N/A
N/A
N/A
0.61%
0.54%
0.45%
0.33%
5.3%
4.2%
2.0%
PROJECTED
DATA
SOURCE
TRM v5 SE
Data
CV2 Parcel
Geodatabase
for Place Type
& Development
Status Editing
Not Available
Not Available
Not Available
TRM v5 travel
demand result
shapefiles
TRM v5 travel
demand result
shapefiles and
DCHC MPO
Metropolitan
Transportation
Plan
TRM v5 travel
demand result
shapefiles
TRM v5 SE
Data total
employment
(2010-2040)
Calculated from
Woods & Poole
earnings per
job by category
and TRM v5 SE
data percent of
employment by
category in Tier
1
Not Available
Not Available
Not Available
Not Available
Not Available
PROJECTED
DATA R2
VALUE
0.99
Insufficient data
points
N/A
N/A
N/A
MTP
"Preferred"
infrastructure
case: 0.999
"Existing +
Committed"
infrastructure
case: 0.997
Insufficient data
points
Insufficient data
points
1.00
1.00
N/A
N/A
N/A
N/A
N/A
AVERAGE
PERCENT
DEVIATION FROM
PROJECTED
DATA
13%
2.6%
N/A
N/A
N/A
VITP "Preferred"
nfrastructure case:
3.6%
'Existing +
Committed"
nfrastructure case:
1. 8%
0.41%
4.2%
0.16%
0.19%
N/A
N/A
N/A
N/A
N/A
208
-------
TIER1 MODEL
VARIABLE BAU
SCENARIO
Nonresidential
property value
Impervious
surface
HISTORICAL
DATA SOURCE
databases
County Office of
Tax Administration
databases
EPA EnviroAtlas
HISTORICAL
DATA R2
VALUE
0.01
Insufficient data
points
AVERAGE PERCENT
DEVIATION FROM
HISTORICAL DATA
31%
1 .0%
PROJECTED
DATA
SOURCE
Not Available
Not Available
PROJECTED
DATA R2
VALUE
N/A
N/A
AVERAGE
PERCENT
DEVIATION FROM
PROJECTED
DATA
N/A
N/A
Table 6-8. Statistical Analysis of Tier 2 Variables Compared to Historical Data and Projections
TIER 2 MODEL
VARIABLE BAU
SCENARIO
Population
Developed land
Nonresidential
sqft
Single family
dwelling units
Multifamily
dwelling units
VMT
Person miles of
public transit
travel per day
Traffic
congestion
Building energy
use
CO2 emissions
Total
Employment
HISTORICAL
DATA SOURCE
U.S. Census
Bureau
Not Available
County Office of
Tax Administration
databases
U.S. Census
Bureau
U.S. Census
Bureau
Not Available
National Transit
Database
Not Available
Durham City-
County
Sustainability
Office
Durham City-
County
Sustainability
Office
U.S. BEA Total
Employment
Growth Rate
(2000-2009)
applied to TRMvS
SE Data 2010 total
employment
HISTORICAL
DATA
R2 VALUE
1.00
N/A
0.95
0.99
0.99
N/A
0.81
N/A
0.87
0.81
1.00
AVERAGE PERCENT
DEVIATION FROM
HISTORICAL DATA
0.0085%
N/A
3.4%
0.42%
0.22%
N/A
9.1%
N/A
2.8%
4.4%
0.27%
PROJECTED
DATA
SOURCE
TRM v5 SE
Data
CV2 Parcel
Geodatabase
for Place Type
& Development
Status Editing
Not Available
Not Available
Not Available
TRM v5 travel
demand result
shapefiles
TRM v5 travel
demand result
shapefiles and
DCHC MPO
Metropolitan
Transportation
Plan
TRM v5 travel
demand result
shapefiles
Not Available
Not Available
TRM v5 SE
Data Total
Employment
(2010-2040)
PROJECTED
DATA R2
VALUE
0.99
Insufficient data
points
N/A
N/A
N/A
MTP "Preferred"
infrastructure
case: 0.999
"Existing +
Committed"
infrastructure
case: 0.997
Insufficient data
points
Insufficient data
points
N/A
N/A
1.00
AVERAGE
PERCENT
DEVIATION
FROM
PROJECTED
DATA
1 .5%
5.0%
N/A
N/A
N/A
MTP "Preferred"
nfrastructure
case: 1.1%
'Existing +
Committed"
nfrastructure
case: 1.6%
MTP "Preferred"
nfrastructure
case: 8.7%
'Existing +
Committed"
nfrastructure
case: 11.5%
0.31%
N/A
N/A
0.35%
209
-------
TIER 2 MODEL
VARIABLE BAU
SCENARIO
Total Earnings
Gross Operating
Surplus
Gross Regional
Product
Total Retail
Consumption
Single family
property value
Multifamily
property value
Nonresidential
property value
Impervious
surface
Water demand
HISTORICAL
DATA SOURCE
Calculated from
Woods & Poole
earnings per job by
category and
shares of
employment by
category
Difference between
GRP and Total
Earnings
Calculated from
U.S. BEA
methodology
(Panek et al. 2007)
NC DOR Total
Taxable Sales by
County (2000-
2014)
County Office of
Tax Administration
databases
County Office of
Tax Administration
databases
County Office of
Tax Administration
databases
EPA EnviroAtlas
NC Dept. of
Environment and
Natural Resources
HISTORICAL
DATA
R2 VALUE
1.00
1.00
1.00
0.63
0.53
0.34
0.21
Insufficient data
points
1.00
AVERAGE PERCENT
DEVIATION FROM
HISTORICAL DATA
0.26%
0.32%
0.28%
8.8%
9.2%
10%
12%
1.8%
0.15%
PROJECTED
DATA
SOURCE
Calculated from
Woods & Poole
earnings per
job by category
and shares of
employment by
category
Not Available
Not Available
Woods & Poole
Retail Sales
Growth Rate
(2015-2040)
applied to 201 4
NC DOR Total
Taxable Sales
Not Available
Not Available
Not Available
Not Available
Triangle
Regional Water
Supply Plan,
VoM.
PROJECTED
DATA R2
VALUE
1.00
N/A
N/A
1.00
N/A
N/A
N/A
N/A
1.00
AVERAGE
PERCENT
DEVIATION
FROM
PROJECTED
DATA
0.38%
N/A
N/A
1 .6%
N/A
N/A
N/A
N/A
1.1%
6.3 Model Corroboration (Validation and Simulation)
The ultimate objective of system dynamics model validation is to establish the accuracy of model
structure—that is, are the processes in the model an accurate reflection of processes in the real world?
Accuracy of the model's reproduction of real behavior is also evaluated, but this is meaningful only if
we first have sufficient confidence in the structure of the model. Thus, we test the validity of the model
structure prior to testing its behavioral accuracy. In this Model Corroboration section, we first describe
direct tests of model structure, in which equations and linkages are changed, or parameters are set at
extreme values, and the resulting output is compared to data or expectations. The direct structure tests
section is composed of three parts:
• Structure Confirmation Tests: In these tests, we changed the equations that determine the values
of variables (as opposed to only changing the values of model inputs), in order to confirm that
the model structure we ultimately chose to use does a better job both of reproducing data and of
representing the way actual systems work.
210
-------
• Extreme-Condition Tests: In these tests, we set the values of certain key model inputs to values
far removed from what is ever likely to happen in real life, such as modeling a sudden 70% drop
in population, or a spike in gasoline prices to $1,000 per gallon. Then, we assessed the
plausibility of the resulting output values against the knowledge or anticipation of what would
happen under a similar condition in real life.
• Unit Consistency: Using a functionality built into the modeling software Vensim, we confirmed
that each variable calculated in the model had units assigned to it that were mathematically
consistent with the units of its inputs and also represented what the variable was intended to
represent (e.g., "population" has units of "person" and "developed land" has units of "acre," so
"population density" has units of "person/acre").
Next, we analyze model behavior through sensitivity tests, wherein we determine how sensitive model
outputs are to the values of uncertain parameters. The model behavior and sensitivity tests section has
three parts, corresponding to three categories of model sensitivity:
• Numerical Sensitivity exists when a change in assumptions changes the numerical values of the
results, without necessarily changing the trend of the output values. It is an inherent property of
models to exhibit numerical sensitivity; the purpose of numerical sensitivity testing is to assure
responsiveness consistent with the functions and feedbacks of the model.
• Behavior Mode Sensitivity exists when a change in assumptions changes the patterns of behavior
generated by the model. For example, if plausible alternative assumptions changed the behavior
of a model from smooth adjustment to random oscillation about a mean value or reduced the
sizes of the "peaks" and "valleys" in the trend of a given output by lengthened reaction times to
exogenous shocks, the model would exhibit behavior mode sensitivity.
• Policy Sensitivity exists when a change in assumptions reverses or heightens the impacts or
desirability of a proposed policy. For example, if a change in assumptions allowed untapped
demand to be realized under one scenario but not under another, the model would exhibit policy
sensitivity.
With each test, we present a table showing the average percent departure from the BAU scenario (or
other base case, as applicable) between 2000 and 2040, which may be either a positive or negative value
(expressed as "percent above or below" in the tables). We chose to compare test results to the BAU
scenario rather than to data, because we have already established the BAU fit against data, for the
variables for which data are available. In some cases, in addition, we report the average absolute-value
percent departure between 2000 and 2040. The first measure takes the percent deviation in each of the
forty years (both positive and negative), and averages them. The second takes the absolute value of the
percent deviation in each of the forty years, and averages them.
The two measures will have the same magnitude if the variable in question is always above or always
below the BAU scenario. However, in cases where there is variation both above and below the BAU
scenario, the average percent departure can obscure variation, as it averages out negative and positive
differences. Figure 6-30 illustrates an example of the difference between the two measures. The arrow
indicates the percent departure for a given year, and the average percent departure is the average over all
years. In the figure, Run 2 values are higher than Run 1 values for about half of the time and below than
Run 1 values for about half of the time, and by about the same magnitude in each case; therefore its
211
-------
average percent departure is zero. However, the average absolute percent departure is different from
zero, since Run 2 always deviates from Run 1, except at the central time point. Therefore, we have also
chosen to show the average absolute percent deviation in tests where output values vary both above and
below B AU over time, resulting in differences in the magnitudes of average percent above or below
BAU and average absolute percent departure from BAU. Note that the average absolute-value percent
deviation can have the effect of exaggerating differences between the test case and BAU, if, for
example, the test causes oscillation slightly above and below the BAU case.
Value
— Run 1 — Run 2
Time
Figure 6-30. Illustration of Average Percent Departure Above or Below vs. Average Absolute
Percent Departure from a Reference Run
Direct Structure Tests
Direct structure tests assess the validity of the model structure by direct comparison with knowledge
about the structure of the real system. This involves assessing each relationship within the model
individually and comparing it with available knowledge about the real system. Several structural tests
have already been illustrated above, including the analysis of improvements resulting from model
integration. The tests below include structure and parameter confirmation tests and extreme-condition
tests, which test equations and data that are obtained from other sources or taken as assumed.
Structure Confirmation Tests
In this section, we present the results of structure confirmation tests for all the key variables of the
model, which we tested first against existing literature and second against historical data. The
comparison with available historical data in the model development phase, when exogenous
assumptions were replaced with endogenous formulations, allowed the modeling team to carry out
parameter confirmation tests by checking the impacts of each parameter used (regardless of the source)
in the integrated structure of the D-O LRP model. We present fourteen structure confirmation tests,
covering all seven sectors in the model.
212
-------
Land Development Aggregated
This structural test compares early versions of the model before and after restructuring of the land
sector, while running the BAU scenario at the time. The test aims to discover if disaggregating land
demand and supply by land use type improve the precision of the model's estimates of land
development.
Prior to restructuring, the land development sector aggregated all demand for land to feed one
conversion stream, which was then allocated to different uses with static percentages based on those
calculated from the CV2 Parcel Geodatabase for Place Type & Development Status Editing. This
created mismatches between demand and supply and led to dramatic drops in retail density (which
affected property values). The former structure also led to inconsistencies when exploring alternative
scenarios. For example, reducing the percent of people in single family households lowered the demand
for residential land (because multifamily homes use less land), which in turn reduced overall land
development, including commercial development, since the ultimate land developed by category was not
controlled by demand for that category, but instead determined by multiplying a constant percentage by
the total land development flow.
In the restructured land development sector, demand for development and conversion is disaggregated
by the six land use types and negative land conversion is allowed to occur. The negative land conversion
implicitly represents redevelopment from one use to another, following demand. Figure 6-31 shows the
change in the trajectory of acres by category; the paler lines show the former results and the darker lines
show the restructured results. This makes clear that office and retail acres were under allocated
throughout the simulation prior to restructuring, while service, multifamily, and single-family acres were
over allocated in the short term and subsequently under allocated in the medium and longer term.
Therefore, disaggregating land demand and supply by land use type does improve the precision of the
model's estimates of land development.
-A/ultifamily acres, Restructured
-A/Utifamily acres, Prior to Restructuring
-Single family acres, Restructured
-Single family acres, Prior to Restructuring
Service acres, Restructured
Service acres, Prior to Restructuring
—Office acres, Restructured
-Office acres, Prior to Restructuring
- Retail acres, Restructured
Retail acres, Prior to Restructuring
• Industrial acres, Restructured
Industrial acres, Prior to Restructuring
2000 2005 2010 2015 2020 2025 2030 2035 2040
Figure 6-31. Acres by Land Use Type - Tier 1: Before and After Land Use Sector Restructuring
213
-------
The change resulted in a little more nonlinearity over time in total developed land, along with more
developed land overall, particularly in Tier 1. We then calibrated GOS per sq ft to bring land
development back to historical levels. Figure 6-32 shows the trajectory of land development under the
former structure, immediately after the restructure, and in the most recent model.
2000 2010 2020 2030 2040
Prior to Restructuring — — Restructured Land Sector BAU
Figure 6-32. Developed Land - Tier 1: Land Development Restructure Test
In addition to improving the estimates of developed land by use, the structural change improved the
consistency of other land use metrics. Table 6-9 summarizes the average yearly percent departure from
BAU 2000-2040 for several affected outcomes. Nonresidential sq ft and total impervious surface follow
the trend in land and are more variable, and retail density avoids the dramatic drop in density that had
been artificially imposed by the constant share of acres applied, maintaining a relative constant value for
the model duration. Since insufficient land had been allotted to meet growing residential demand in Tier
1, the change allowed dwelling units to increase there. As more dwelling units are available, more
people may move to the area, marginally increasing net migration and population in Tier 1.
Effect of Vacancy on Single Family and Multifamily Dwelling Units Removed
This structural test removes the effect of vacancy on equilibrium dwelling units, which serves to boost
dwelling unit construction when vacancy is very low, and decrease construction when vacancy is very
high. This effect table was not found in the literature and was therefore calibrated to move the model
projection of dwelling units closer to historical data. Since the table was based purely on assumptions,
this test was performed to verify that it improves model fit and behavior.
The biggest immediate impact of this change is on multifamily dwelling units in Tier 2, as shown in
Figure 6-33. The BAU scenario almost perfectly matches historical data; however, the model's estimates
without the effect of vacancy on equilibrium dwelling units fall short of the BAU by an average of 4.2%.
While the impact is not as large for single-family dwelling units and dwelling units in Tier 1, for
consistency, the same table was applied for the other dwelling units, and always improves the fit, even if
not by much.
214
-------
Table 6-9. Average Yearly Percent Departure of Test from Restructured Land Sector and from BAD, 2000-
2040: Land Development Structure Test47
Prior to Restructuring v
Restructured Land Sector Prior to Restructuring v. BAU
Percent above
Percent above
Tier 1
Developed land Tier 1
Retail acres Tier 1
Single family acres Tier 1
Nonresidential sq ft Tier 1
Retail density Tier 1
Total impervious surface Tier 1
Total dwelling units Tier 1
Net migration Tier 1
Population Tier 1
Tier 2
Developed land
Retail acres
Single family acres
Nonresidential sq ft
Retail density
Total impervious surface
Total dwelling units
Net migration
Population
2.4%
4.1%
1.2%
4.6%
3.6%
2.2%
0.9%
34%
0.7%
1.8%
10%
1.4%
5.7%
10%
1.3%
0.0%
1.0%
0.0%
-1.7%
-4.1%
0.7%
-4.4%
-3.5%
-1.8%
-0.7%
17%
-0.6%
-1.8%
-10%
-1.4%
-5.7%
-10%
-1.3%
0.0%
-0.3%
0.0%
3.7%
3.3%
7.5%
1.1%
7.0%
1.0%
7.2%
164%
6.0%
1.4%
10%
1.4%
4.0%
10%
0.6%
0.1%
4.1%
0.1%
3.6%
-1.4%
7.4%
-0.3%
-6.6%
1.0%
6.7%
148%
5.8%
-1.4%
-10%
-1.4%
-3.8%
-10%
-0.1%
0.1%
4.0%
0.1%
2000 2010 2020 2030 2040
BAU No Effect of Vacancy on DU BAU — — -TierS DATA
Figure 6-33. Multifamily Dwelling Units - Tier 2: Effect of Vacancy on Dwelling Units Removed
The addition of the table also improves the fit and behavior of other variables in the sector. Annual
renter costs more closely fit the data with the effect of vacancy on dwelling units table included in the
model, particularly in Tier 2. While the BAU deviates on average by 2.1% from the data for the years
available, without the effect table, it deviates by an average of 17% in Tier 2. Table 6-10 summarizes the
average yearly percent departure from BAU 2000-2040 for several affected outcomes, including the
annual renter costs.
47 Note that the large percent departure from BAU in net migration in Tier 1 is due to later corrections in the net migration
rate, and not to model changes made in this test, which are instead reflected in the percent departure from the "Restructured
Land Sector" scenario.
215
-------
Table 6-10. Average Yearly Percent Departure from BAD, 2000-2040: Effect of Vacancy on Dwelling Units
Removed
No Effect of Vacancy on DU Test v.
Absolute Percent Percent above or
Departure below
Tier 1
Multifarrrily dwelling units Tier 1
Single family dwelling units Tier 1
Multifamily vacancy rate Tier 1
Single family vacancy rate Tier 1
Annual renter costs Tier 1
Tier 2
Multifamily dwelling units
Single family dwelling units
Multifamily vacancy rate
Single family vacancy rate
Annual renter costs
0.7%
0.9%
7.1%
8.5%
0.1%
-4.2%
-1 .5%
-55%
-16%
19%
0.7%
0.9%
6.5%
8.5%
-0.1%
-4.2%
-1.5%
-55%
-16%
19%
Not Normalizing Person Miles by Mode
The heart of the transportation model sector is the calculation of mode shares, or the percentage of
person miles that are taken by each transportation mode: automobile driver, automobile passenger,
public transit, and nonmotorized. The model first establishes baseline projections of the number of
person miles traveled per day by each mode (driven by population and real GRP per capita). These
baseline values are adjusted by a series of elasticities (with respect to both exogenous and endogenous
variables) and then normalized so that the total person miles of travel per day match the baseline
projections. This normalization is done to keep overall person miles of travel per day within
expectations and guarantee that an increase or decrease in the use of any one mode will also affect usage
rates of the other modes. In this structure confirmation test (hereafter called the Travel Not Normalized
test case), we compared the BAU results of the final model to what they would be if the normalization
step were removed, meaning that there is no limit on how high or low the total number of person miles
of travel may be and that factors that increase or decrease person miles by one mode need not
necessarily also decrease or increase person miles by other modes, respectively. We did this in order to
determine whether the normalization step was necessary, in this particular modeling effort, to prevent
the generation of unreasonable person-mile results.
There is little difference between the trends of the modal-person-mile variables in the Travel Not
Normalized test case and the BAU scenario (Figure 6-34) or between their magnitudes (Table 6-11).
Therefore, leaving the person-mile-normalization step out of the model would have only required slight
increases in the initial values set during the calibration process in order to match data and projections
about as well as in the BAU scenario. Otherwise, given the D-O LRP SD Model's default input values
and other assumptions, it may have not significantly changed the accuracy of the model to leave out the
normalization step. However, if, under a given scenario, the model featured more dramatic drivers of
modal person miles, leaving out the normalization step could potentially result in numbers of person
miles that are either unrealistically high or unreal!stically low for an area with a given population and
GRP.
216
-------
The greatest difference between the test case and the BAU scenario is around 2015, when, in both cases,
the model is reacting to a recent period of high gasoline prices, producing a reduction in the growth rate
of automobile driver person miles and increases in the growth rates of person miles by all other modes.
In the BAU scenario, the normalization step moderates the negative effect on automobile driver person
miles and enhances the positive effect on person miles by all other modes. This is because, before
normalization, the negative elasticity of automobile driver person miles to gasoline prices has a greater
magnitude than the positive elasticities of all other types of person miles to gasoline prices. After
normalization, the effective elasticity of person miles by any given mode to any given input is a function
of both its own pre-normalization elasticity and the pre-normalization elasticities of person miles by all
of the other modes, since any increase or decrease in person miles by one mode relative to the baseline
must be made up for by opposite-direction changes in person miles by one or more of the other modes.
Therefore, the normalization step causes the large, negative pre-normalization elasticity of automobile
driver person miles to gasoline prices to effectively be moved closer to the magnitudes of the smaller,
positive pre-normalization elasticities of person miles by any other mode to gasoline prices, and vice
versa. In other words, the normalization step reduces the negative effect of high gasoline prices on
automobile driver person miles and increases the positive effect of high gasoline prices on all other types
of person miles. As shown in Figure 6-34, in the Travel Not Normalized test case, automobile driver
person miles and public transit person miles show a worse fit with data and available projections than
they do in the BAU case, mostly for want of the readjustment performed by the normalization step. This
suggests that the forecast generated by the normalization step improves model accuracy. The overall
effect is that the Travel Not Normalized test case results in fewer person miles by all four travel modes
than the BAU scenario, as shown in Table 6-11.
(A) Automobile Driver Person Miles - Tier 2
(B) Public Transit Person Miles - Tier 2
2000 2010
^^^ Travel Not Normalized
---TRMvS Result Data
2020 2030 2040
BAU
— —TRM v5 E+C Result Data
2000 2010
^^Travel Not Normalized
-_TRMvS Result Data
• • National Transit Database
2020 2030 2040
^—BAU
— —TRMvS E+C Result Data
Figure 6-34. Person Miles of Travel by Different Modes Not Normalized vs. BAU
Table 6-11. Average Yearly Percent Departure from BAU, 2000-2040: Not Normalizing Person Miles by
Mode
VARIABLE
AVERAGE YEARLY
ABSOLUTE PERCENT
DEPARTURE
AVERAGE YEARLY PERCENT
ABOVE OR BELOW
Tierl
Automobile driver person miles
Automobile passenger person miles
Public transit person miles
Nonmotorized person miles
Total person miles
1.6%
1.8%
2.4%
2.3%
1.8%
-1.6%
-1.7%
-2.3%
-2.2%
-1.7%
Tier 2
Automobile driver person miles
Automobile passenger person miles
Public transit person miles
2.6%
2.9%
4.0%
-2.5%
-2.7%
-3.9%
217
-------
VARIABLE
Nonmotorized person miles
Total person miles
AVERAGE YEARLY
ABSOLUTE PERCENT
DEPARTURE
3.7%
2.7%
AVERAGE YEARLY PERCENT
ABOVE OR BELOW
-3.5%
-2.6%
No Post-Normalization Congestion Effect on Automobile Driver Person Miles
One exception to the normalization process described above is the effect of traffic congestion on
automobile driver person miles. The model assumes that if traffic congestion causes driving a car to be
less attractive, some people will choose to instead travel by other modes while other people will choose
to reduce the amount they drive without traveling more by any other mode. Therefore, traffic congestion
is modeled as leading to reduced automobile driver travel both before and after the normalization step.
Before the normalization step, average vehicle speeds affect automobile driver person miles (and person
miles by all other modes) through elasticity values taken from literature. After the normalization step,
automobile person miles are affected further by congestion, through an S-shaped lookup function.
"Congestion" is defined as the ratio of peak-period travel time to freeflow travel time, wherein a value
of one represents a congestion-free state. If congestion is equal to one, no further change is made to
automobile driver person miles. If congestion is equal to 1.75, the inflection point of the S-shaped
function, automobile driver person miles will be reduced by 10%, with a one-year delay; if congestion is
equal to 2.5 or greater, automobile driver person miles will be reduced by 20%, again with a one-year
delay. In this structure confirmation test (hereafter called the No Additional Congestion Effect test case),
we compared the BAU results of the final model to what they would be if this post-normalization effect
of traffic congestion on automobile driver person miles were not included, in which case the effect of
traffic congestion on automobile driver person miles would come entirely in the form of people
switching to or from other travel modes. We did this in order to determine how much the post-
normalization effect of traffic congestion on automobile driver person miles contributes to keeping
person-mile trends close to projections.
As shown in Figure 6-35 and Table 6-12, the No Additional Congestion Effect test case results in
slightly more automobile driver person miles of travel than the BAU scenario. In neither the BAU
scenario nor either of the light rail scenarios does congestion exceed 1.27 in either Tier in any year
during 2000-2040. Relative to the inflection point of the S-shaped function that determines the post-
normalization effect of traffic congestion on automobile driver person miles, this is a low level of
congestion. Therefore, with the model's default inputs and other assumptions, leaving out the post-
normalization effect of congestion on automobile driver person miles would not have a major impact on
the magnitudes or trends of model outputs. However, if congestion were much greater, the presence or
absence of that post-normalization effect would make a significant difference.
In the No Additional Congestion Effect test case, automobile driver person miles are higher, meaning
that VMT per lane mile is higher, which increases congestion, which reduces automobile driver and
automobile passenger person miles and increases public transit and nonmotorized person miles. This
feedback is an additional reason why the No Additional Congestion Effect test case and the BAU
scenario do not have significantly different outputs. However, this feedback effect only partially
mitigates the increase in automobile driver person miles that results from removing the post-
normalization effect of congestion, so that automobile driver person miles in the No Additional
Congestion Effect test case are consistently higher than in the BAU scenario, though only by a small
amount. At the same time, the feedback effect exacerbates the assumption under the No Additional
218
-------
Congestion Effect test case that a congestion-induced reduction in automobile driver person miles
necessarily results in the same number of person miles being cumulatively added to the remaining
modes. Without a post-normalization effect of traffic congestion on automobile driver person miles, an
unusually high level of traffic congestion could result in unrealistically large increases in person miles
by public transit and nonmotorized modes relative to population and GRP. The post-normalization effect
of congestion therefore helps to make the model more consistent with expectations by avoiding such
unrealistic conditions.
(A) Tier 1
(B) Tier 2
0.3%
0.2%
0.1%
^~+
7
^r
/
^V >
^^^/^
^- •
^s^
^^^^
1
1
2000 2010 2020 2030 20
Figure 6-35. Automobile Driver Person Miles Per Day: No Post-Normalization Congestion Effect on
Automobile Driver Person Miles Percent Difference from BAD
Table 6-12. Average Yearly Percent Departure from BAD, 2000-2040: No Post-Normalization Congestion
Effect on Automobile Driver Person Miles
VARIABLE
NO ADDITIONAL CONGESTION EFFECT V. BAU
Tierl
Congestion
Automobile driver person miles
Automobile passenger person miles
Public transit person miles
Nonmotorized person miles
Total person miles
+0.25%
+0.55%
-0.06%
+0.10%
+0.06%
+0.32%
Tier 2
Congestion
Automobile driver person miles
Automobile passenger person miles
Public transit person miles
Nonmotorized person miles
Total person miles
+0.28%
+0.48%
-0.01%
+0.18%
+0.12%
+0.36%
NMT Facility Construction Not Driven by Developed Land
Originally, the construction of sidewalks, bike lanes, and other nonmotorized travel (NMT) facilities in
the D-O LRP SD Model was driven entirely by exogenous projections from the Triangle Regional
Model version 5 SE Data files. In the current model, we replaced this exogenous function with one
driven by the amount of developed land. We made this change because, if NMT facility construction is
completely exogenous while land development is endogenous, it could produce a scenario where the
amount of NMT facilities per developed acre is unrealistic if development does not closely match levels
assumed in the TRM SE Data files. We did not apply the same reasoning to make road construction
driven by land development, however. We did this because roads are more likely than NMT facilities to
be built through areas that are not currently developed but may be developed in the future, in accordance
with transportation plans with decades-long time horizons, since land is rarely developed in the absence
219
-------
of motor-vehicle access. As a result, it is not unrealistic to suppose that the ratio of roads to developed
acres would be greater in a low-development scenario and less in a high-development scenario.
Furthermore, because NMT facilities can generally be built faster and more cheaply than roads can,
NMT facilities' construction is able to be more responsive to changes in demand. In this structure
confirmation test (hereafter called the B AU Exogenous Path Building test case), we examined whether
keeping nonmotorized-travel-facility construction exogenous rather than endogenous would have
produced unreasonable results in the model, relative to data and projections.
The TRM SE Data files feature development, population, and employment projections that are based on
the assumption that the planned light rail line between Durham and Chapel Hill is built. Therefore, using
that source's nonmotorized-travel-facility projections in the BAU case, wherein no light rail line is built,
would produce inaccurate results. In fact, with NMT facility construction driven by developed land, the
Light Rail scenario much more closely matches projections of Tier 1 NMT facilities than does the BAU
scenario (Figure 6-36A). However, in Tier 2, the BAU, Light Rail, and Light Rail + Redevelopment
scenarios all exceed the exogenous NMT facility projections used in the BAU Exogenous Path Building
test case (Figure 6-36B). This may be attributed either to the TRM SE Data files assuming different
ratios of NMT facility construction to land development than we did for this particular Tier or to the fact
that the TRM SE Data files do not include projections of developed acres, such that there may be some
difference between the amount of developed land implicit in that data source and the developed-land
projections (calculated from Imagine 2040 outputs) that were used to calibrate the D-O LRP SD Model.
As shown in Table 6-13, no significant feedback effects are created by NMT facilities being driven by
developed land in the D-O LRP SD Model: replacing this mechanism with an exogenous nonmotorized-
travel-facility construction schedule does not change developed land. This lack of a meaningful
feedback effect is expected, since the model assumes that the construction of NMT facilities is done in
response to land development, as opposed to the other way around. Furthermore, even if the presence of
sidewalks, bike lanes, and paths attracts people to an area, hence making further development in that
area more attractive, people are usually only attracted by those facilities if there are already developed
parcels nearby that the facilities may be used to reach.
Table 6-13 also shows the expected outcome that changing the amount of NMT facilities in an area
produces a same-direction change in NMT and an opposite-direction change in automobile driver travel.
However, because increases in the use of one transport mode are offset by decreases in other modes,
overall person miles of travel are not affected.
(A) Tier 1
(B) Tier 2
2000 2010 2020
^^^—BAU Exogenous Path Building —
— —Light Rail -
— — —TRMvSSE Data
2030
- Light Rail + Redev
-BAU
2000 2010 2020
^^^—BAU Exogenous Path Building —
— — Light Rail —
— — —TRM v5 SE Data
2030
-Light Rail + Redev
-BAU
Figure 6-36. NMT Facilities
220
-------
Table 6-13. Average Yearly Percent Departure from BAD, 2000-2040: Building of Pedestrian and Bicycle
Facilities Driven by Exogenous Construction Schedule Based on TRM V5 SE Data
VARIABLE
BAU EXOGENOUS PATH BUILDING V. BAU
Tierl
NMT facilities
Developed land
Nonmotorized person miles
Automobile driver person miles
Automobile passenger person miles
Public transit person miles
Total person miles
+4.8%
0.0%
+ 1.4%
-0.2%
+0.1%
+0.9%
0.0%
Tier 2
NMT facilities
Developed land
Nonmotorized person miles
Automobile driver person miles
Automobile passenger person miles
Public transit person miles
Total person miles
-7.3%
0.0%
-2.7%
+0.2%
-0.5%
-1.8%
0.0%
Light Rail Treated the Same as Bus Service for the Purpose of Determining Ridership
Adding light rail to a public transit system has the potential to attract more riders per revenue mile than
adding more bus service, which, unless it is Bus Rapid Transit (BRT), does not have the advantage of
traveling on an exclusive right-of-way. Originally, the D-O LRP SD Model did not contain a mechanism
to account for the difference in attractiveness to travelers between light rail and bus service. Instead, the
only direct effect on public transit usage of the light rail line opening in 2026 in the Light Rail and Light
Rail + Redevelopment scenarios was that it represented an increase in the total number of revenue miles
on the public transit system, which is one of the drivers of public transit person miles of travel. In other
words, travelers were assumed to regard the light rail line in the same way that they would the opening
of a new non-BRT bus route. In the current model, instead of the opening of the light rail line increasing
public transit person miles through the increase in revenue miles that it represents, it activates an
equation that determines the number of additional public transit person miles resulting from the light rail
as a function of Tier 1 population and jobs and Tier 2 highway VMT per lane mile, derived from a
spreadsheet tool found in literature (Chatman et al. 2014). The use of this equation leads to higher
impacts on public transit person miles of travel from the light rail line than would assuming that
additional light rail revenue miles would have the same effect as non-BRT bus revenue miles, as seen in
Figure 6-37. This structure confirmation test (hereafter called the "Light Rail Treated Like Bus" test
case) evaluates whether this equation produces results that are more consistent with projections from the
TRM than an alternative formulation that treats light rail revenue miles like bus revenue miles.
The DCHC MPO Metropolitan Transportation Plan uses TRM v5 outputs to project how much public
transit use will increase between 2010 and 2040 in a scenario where the light rail line between Durham
and Chapel Hill is built, but without describing the shape of the trend between 2010 and 2040 ("TRM v5
Result Data" in Figure 6-37). In the year 2040, Tier 2 public transit person miles in the Light Rail
scenario are 24.7% greater than the MTP projection, and Tier 2 public transit person miles in the Light
Rail Treated Like Bus test case are 9.8% less than the MTP projection (both scenarios are calibrated to
the MTP figures for 2010). One possible interpretation of this data is that the public-transit-person-mile
equation used in the Light Rail and Light Rail + Redevelopment scenarios performs its intended
221
-------
function, representing light rail service (or any other fixed-guideway transit service) as being more
attractive to travelers than bus service, but because it produces an output that deviates from projections
by a greater margin than the alternative formulation, it may overestimate the attractiveness of light rail
relative to bus service. This could be because the equation used in the Light Rail and Light Rail +
Redevelopment scenarios to determine additional public transit person miles resulting from the light rail
line does not account for public transit use in the study area before the light rail line opened. Instead, it
uses population, employment, and highway VMT per lane mile to estimate the amount of demand for
fixed-guideway transit services, using equations derived from data gathered from metropolitan areas
across the United States. Therefore, it could be that the population of the study area is more resistant to
switching to public transit than the national average and the D-O LRP SD Model fails to reflect this.
The difference between the Light Rail scenario and TRM projections shown in Figure 6-37 may also be
caused by the D-O LRP SD Model assuming that the light rail line will have a larger impact on
commercial development and net migration than does the MTP, which would also cause public transit
person miles to be higher. To explore this possibility, the Light Rail scenario and the Light Rail Treated
Like Bus test case are rerun, both with the added change of removing the assumptions that the light rail
line will necessarily cause greater demand for Tier 1 commercial floor space and that net migration will
necessarily be affected, renamed the Light Rail No Sq Ft Effect and Light Rail Like Bus No Sq Ft test
cases, respectively, as shown in Figure 6-37. In the year 2040, Tier 2 public transit person miles in the
Light Rail No Sq Ft Effect case are 18.7% greater than the MTP projection (compared to 24.7% greater
for the Light Rail scenario), and Tier 2 public transit person miles in the Light Rail Like Bus No Sq Ft
case are 10.7% less than the MTP projection (compared to 9.8% less for the Light Rail Treated Like Bus
case). As expected, removing the assumed effects of the light rail on commercial floor space and
migration brings the Light Rail scenario closer to the MTP-derived projections for public transit person
miles, while making the Light Rail Treated Like Bus case farther from MTP-derived projections.
However, the Light Rail Like Bus No Sq Ft case is still closer to the MTP-derived 2040 Tier 2 public
transit person mile projection than is the Light Rail No Sq Ft Effect case. This suggests that, even if the
figures derived from the MTP assume that the opening of the light rail will have less of an impact on
land development, jobs, and population than does the D-O LRP SD Model, the D-O LRP SD Model may
still overestimate how much light rail will increase public transit use, unless, conversely, the MTP's
projections were to turn out to be too low in this regard.
A change in public transit person miles produces a same-direction change in nonmotorized person miles
(because each unlinked, one-way public transit trip is assumed to result in an average of 0.25 miles of
additional nonmotorized travel) and an opposite-direction change in automobile person miles (Table
6-14 and Table 6-15). The Light Rail Treated Like Bus test case also has fewer overall person miles of
travel after 2026 than the regular Light Rail scenario, due to there being less public-transit-induced
nonmotorized travel. Table 6-14 also shows that the large difference in public transit person miles
between the Light Rail scenario and the Light Rail Treated Like Bus test case has relatively minor
feedback effects on population, employment, and VMT per highway lane mile, the inputs that drive
light-rail-induced public transit person miles in the Light Rail scenario. The small size of these
feedbacks at least partially reflects the fact that public transit represents a very small proportion of
overall person miles of travel, regardless of whether or not the light rail line is built. If a much larger
percentage of person miles were traveled on public transit or these feedback effects were stronger, the
model's results could potentially be rendered less accurate, as the underlying equation used in the model
originally came from a model that did not feature feedbacks and made assumptions about the pre-light-
rail public transit usage rates of areas that may differ from the study area of our model.
222
-------
400
2000 2010
^^ Light Rail Treated Like Bus
^—BAD
. —TRMvS E+C Result Data
2020
2040
2030
•Light Rail
•TRMvS Result Data
National Transit Database
Figure 6-37. Public Transit Person Miles of Travel Per Day - Tier 2: Light Rail Scenario with Light Rail
Treated Like Bus Service for Determining Transit Use vs. Regular Light Rail Scenario
400
2000 2010 2020
Light Rail Like Bus No Sq Ft — <
^—BAU --
- — TRM v5 E+C Result Data
2030 2040
• Light Rail No Sq Ft Effect
•TRM v5 Result Data
National Transit Database
Figure 6-38. Public Transit Person Miles Of Travel Per Day - Tier 2: Light Rail Scenario with Light Rail
Treated Like Bus Service For Determining Transit Use and No Assumed Direct Effect of Light Rail On
Demand For Commercial Floor Space Or Migration Vs. Light Rail Scenario With Light Rail Treated
Differently from Bus Service for Determining Transit Use But Still No Assumed Direct Effect of Light Rail
On Demand for Commercial Floor Space or Migration
Table 6-14. Average Yearly Percent Departure from Light Rail Scenario, 2026-2040: Light Rail Treated the
Same as Bus Service for The Purpose of Determining Ridership
VARIABLE
LIGHT RAIL TREATED LIKE BUS V.
LIGHT RAIL
Tierl
Public transit person miles
Automobile driver person miles
Automobile passenger person miles
Nonmotorized person miles
Total person miles
Population
Employment
VMT per highway lane mile
-64.6%
+4.9%
+5.0%
-4.2%
-0.5%
-0.1%
-0.2%
+2.5%
Tier 2
Public transit person miles
Automobile driver person miles
Automobile passenger person miles
Nonmotorized person miles
Total person miles
-21.2%
+0.3%
+0.4%
-1.3%
-0.1%
223
-------
VARIABLE
Population
Employment
VMT per highway lane mile
LIGHT RAIL TREATED LIKE BUS V.
LIGHT RAIL
0.0%
0.0%
+0.2%
Note: Difference between Light Rail scenario and Light Rail Treated Like Bus test case starts in 2026, when the light rail line
opens.
Table 6-15. Average Yearly Percent Departure from Light Rail Scenario with No Assumed Direct Effect of
Light Rail On Demand for Commercial Floor Space or Migration, 2026-2040: Light Rail Treated the Same
As Bus Service for the Purpose of Determining Ridership + No Assumed Direct Effect of Light Rail On
Demand for Commercial Floor Space or Migration
VARIABLE
LIGHT RAIL LIKE BUS NO SQ FT V. LIGHT RAIL NO
SO FT EFFECT
Tierl
Public transit person miles
Automobile driver person miles
Automobile passenger person miles
Nonmotorized person miles
Total person miles
Population
Employment
VMT per highway lane mile
-65.2%
+5.0%
+5.2%
-4.2%
-0.4%
-0.1%
-0.2%
+2.4%
Tier 2
Public transit person miles
Automobile driver person miles
Automobile passenger person miles
Nonmotorized person miles
Total person miles
Population
Employment
VMT per highway lane mile
-19.5%
+0.3%
+0.3%
-1.1%
0.0%
0.0%
0.0%
+0.2%
Note: Difference between Light Rail No Sq Ft Effect and Light Rail Like Bus No Sq Ft cases starts in 2026, when the light
rail line opens.
Structure of Energy-Economy Feedback
In the real world, the link between energy spending and gross regional product is complicated and
difficult to summarize in a model. In two structural tests we explore alternative formulations of the
energy spending-GRP link to determine the effect of this structure on our results. The current D-O LRP
model assumes energy spending can have either a balancing or reinforcing effect on GRP, depending on
whether energy spending grows or decreases as a fraction of GRP.48 This structure represents energy
affordability; if energy spending grows proportionately more than GRP does, energy becomes less
affordable, and vice-versa. Increasing affordability of energy in the model leads to an increase in GOS,
which is set to be equal to 40% of GRP.
To test the effects of this structure, we removed the connection from energy spending to GOS. This
48 As context, other system dynamics modelers have assumed a balancing feedback between energy spending and economic
growth at the US (Bassi et al. 2010) level.
224
-------
removes the balancing and reinforcing feedbacks from energy spending to GRP, to determine the size of
these energy spending effects.
In the current version of the model, energy spending tends to decrease GOS (and therefore GRP)
between 2000 and 2022, but it increases GOS from 2023 to 2040 (red line in Figure 6-39). Unlinking
energy spending from GOS therefore causes GOS to grow faster in the first half of the time series and
slower in the second half, compared to BAU. Unlinking energy spending leads to less growth in GOS
and GRP in the long run due to the effect of increasing energy efficiency. As building and vehicle
energy efficiency increases into the future, energy spending decreases as a fraction of GRP, which in the
BAU scenario leads to higher GRP growth.
0.92
2000
2010
2020
2030
2040
•BAU energy_econ_unlinked
•BAU
Figure 6-39. Structural Analysis for Unlinking Gross Operating Surplus from Energy Spending:
Effect in Tier 2 on "Energy Spending Share of GRP Factor Affecting GOS"
Figure 6-40 shows the effect of this structural change on (a) GRP and (b) GRP growth rate in Tier 2. In
(a), the orange line of unlinked GRP is larger than in BAU between 2005 and 2033, but smaller than in
BAU afterward. The unlinked case differs from reference data and projections by 6% in 2010 and 4% in
2040. In contrast, the BAU scenario differs from reference data and projections by 0.3% in both years.
This suggests that feedbacks from energy spending cause a 5% fluctuation in GRP in our model. These
fluctuations can also be viewed in terms of growth rate; Figure 6-40B shows that the unlinked GRP (the
orange line) grows faster than in BAU between 2003 and 2013, but afterward grows slower than in
BAU.
225
-------
(A)
2000
2010
2020
2030
2040
• BAU energy_econ_un linked
• BAU
• Data and Projections
-0.01
(B)
2040
BAU energy_econ_unlinked
BAU
— — — Data and Projections
Figure 6-40. Structural Analysis for Unlinking Gross Operating Surplus from Energy Spending:
Effect in Tier 2 on (A) GRP and (B) GRP Growth Rate
In a second structural test, relative energy spending, rather than energy spending/GRP, is used to create
feedback from the energy sector to the economy. Again, this test explores an alternative formulation of
the energy spending-GRP link to determine the effect of this structure on our results. (Here, "relative
energy spending" is relative to the 2000 value of energy spending (Figure 6-4la). Using relative energy
spending leads to much slower GRP growth (orange line compared to red line, Figure 6-4Ib), with 38%
lower GRP by 2040. Reducing the elasticity of GRP to energy spending from 0.2 to 0.1 (yellow line)
improves the fit to historical data for this structure, but only up to 2017.
In general, using energy spending as a fraction of GRP leads to a better fit to historical data. This
suggests energy spending/GRP may better represent feedback from the affordability of energy to GRP.
As the economy grows, a feedback structure using only relative energy spending would create negative
feedback on GRP (Figure 6-4la), but relative energy spending/GRP can create positive feedback when
GRP grows faster than energy spending. This is the case here; in the BAU scenario, energy spending
grows by a factor of 2.3 between 2000 and 2040 while GRP grows by a factor of 2.75.
226
-------
(A)
(B)
2.5
1.5
0.5
s
2000
2010
2020
2030
2040
2000
2010
2020
2030
2040
BAU 215 rel_energy01
•BAU 215 rel_energy
•BAU
BAU 215 rel_energy01
•BAU 215 rel_energy
• BAU
•Data and Projections
Figure 6-41. Structural Analysis for Using Relative Energy Spending, Not Energy Spending/GRP,
in Feedback to GRP: Effect on (A) Relative Energy Spending and (B) GRP in Tier 2
Removing the Effect of Congestion on Productivity
Congestion negatively impacts the economy in the D-O LRP SD Model by reducing the GOS. The
rationale for having congestion only impact GOS and not total GRP is that even though the earnings of
salaried employees would not be affected from being stuck in traffic, the business or company that
employs the worker is losing profits from the work that the employee would be doing, in addition to the
profit loss from the transport of goods and services to and from the business being delayed. This
structural test shows the impact of removing the effect of congestion on GOS and the subsequent effect
on GRP in Tier 2 and Tier 1.
Although removing the effect of congestion seems to have a very little impact in the short term on GOS
and GRP in both Tier 2 (Figure 6-42, Top) and Tier 1 (Figure 6-42, Bottom), the internal feedbacks in the
economy sector (GOS -^ GRP -^ Retail Consumption -> Employment -^ GRP, described in the
economy sector narrative in Section 3.3) cause these small differences to become more substantial over
time - with GRP being $1.8 Billion (2010 dollars) higher in 2040 in Tier 2 compared to the BAU
scenario and GRP Tier 1 being $1.1 Billion (2010 dollars) higher in 2040 compared to the BAU
scenario.
227
-------
(A)
(B)
2000 2010 2020
^^^GOS No Congestion Effect ^
- —-GOSBAU —'
2030
2040
•GRP No Congestion Effect
•GRP BAD
5 12
2010
2020
2030
2040
• COS Tier 1 No Congestion Effect
• COS Tier 1 BAU
• GRP Tier 1 No Congestion Effect
• GRP Tier 1 BAU
Figure 6-42. Structural Test of Removing Effect of Congestion on Economic Productivity:
Comparison to BAU of Two Economic Indicators for (A) Tier 2 and (B) Tier 1
Tier 1 Retail Consumption Calculation With and Without Connection to Resident % of the Working Population
This section is not a quantitative structure test, but instead discusses why the Tier 1 model structure for
estimating retail consumption differs from Tier 2. In Tier 2, retail consumption changes, after a two year
delay, relative to an initial (2000) retail consumption value that is multiplied by two factors:
(1) The relative change in the resident percent of the working population, calculated exogenously
based on historical data and projections for the following equation:
i-
(total employment — resident employment')
population
(2) The relative change in GRP (endogenous) raised to the power of a calibrated elasticity of consumption to
GRP (1.1)
The first relative factor, resident percent of the working population, is meant to represent the change in
the number of people that both live and work in an area, assuming that those who work in an area but
don't also live there tend to do their shopping elsewhere (closer to home). Thus, as long as population
grows faster than the percent of jobs that are filled by non-residents, the relative value (relative to the
initial 2000 value) for resident percent of the working population will increase, allowing retail
consumption to increase.
228
-------
When calculating the resident percent of the working population for Tier 1, we found that, since
nonresident employment (i.e. commuters) is actually higher than population in Tier 1, this parameter
was negative and grew to be much more negative in the BAU scenario, shown in Figure 6-43, where
total employment actually grows faster than resident employment and population. Consequently, the
relative resident percent of the working population in Tier 1 decreases, changing by much more than the
change observed in Tier 2. We therefore decided not to use this formulation to estimate retail
consumption in Tier 1. We reasoned that, despite the decline of resident workers relative to total
employment, Tier 1 residents and employees will still do their shopping in Tier 1 and continue to make
up a portion of the retail consumption that happens there. As a result, we decided not to connect retail
consumption in Tier 1 to the resident percent of working population in that Tier. On the other hand, if
we were to create this connection, we would need to modify the formulation of retail consumption (to
account for the negative and declining resident percent of working population). This would require us to
divide relative GRP by the resident percent of working population, and to add an elasticity that would
reduce the strength of the relationship (to account for the fact that, as mentioned above, resident
population and all workers would still spend money in Tier 1).
100%
-100%
2000 2005 2010 2015 2020 2025 2030 2035
^^"Resident Percent of the Working Population Tier 1
^^Resident Percent of the Working Population Tier 2
2040
Figure 6-43. Resident Percent of the Working Population for Tier 2 and Tier 1, Calculated
from Historical Data and Projections of Employment and Population
Tier 1 Resident Labor Force Calculations
The equation we chose for calculating the resident labor force in Tier 1 (Option 2 below), which directly
affects unemployment rate and thus net migration to Tier 1, involved several assumptions that, at the
time the model was completed, seemed to be the best assumptions. However, two other options for
calculating resident labor force in Tier 1 were tested and we present here the results of all three
calculations and how the different formulations affect the rest of the model.
At an early stage in model development, we simply used the same formulation for Tier 1 as in Tier 2:
Option 1: Population Tier 1 *percent of residential population in labor force Tier 1 with the percent of
the residential population in the labor force Tier 1 (42%) calculated exogenously based on historical data
and held constant for the entire time period.
229
-------
However, since the additional employment added to Tier 1 in the Light Rail and Light Rail +
Development scenarios brings additional residents to Tier 1 (compared to the BAU scenario), we
decided that the family of the additional residents who were moving to Tier 1 for work should be
subtracted out of the calculation for residential labor force under the assumption that it was more likely
that they would either not be seeking employment or would be employed outside of the Tier 1
geographical area. Thus, we developed a second formulation for the Tier 1 labor force:
Option 2: population Tier l*percent of residential population in labor force Tier l-(resident
employment Tier 1 - initial resident employment Tier l)*(additional family members per worker tier 1)
After an analysis of the results of the model, we realized that this formulation actually subtracts too
many residents from the labor force, decreasing the unemployment rate, which increases the population
in Tier 1 and causes the "per capita" parameters in Tier 1 (e.g. resident net earnings per capita) to
decrease in the Light Rail and Light Rail + Redevelopment Scenarios (relative to the BAU). As a result,
we created a third option for possible use in future versions of the D-O LRP SD Model:
Option 3: (population Tier 1 - ((resident employment Tier 1 - initial resident employment Tier
l)*additional family members per worker tier l))*percent of residential population in labor force Tier 1
Option 3 subtracts the additional family members that move along with newly employed residents from
the population before multiplying that population by the "percent of residential population in labor force
Tier 1."
Results of the structural tests for the three options for calculating the resident labor force in Tier 1 are
shown in Figure 6-44 and Figure 6-45. For both figures, the three options are shown for the Light Rail +
Redevelopment scenario, with the default option (Option 2) also shown for the BAU scenario for
reference. As was mentioned previously, the unemployment rate drops very low in Tier 1 in the Light
Rail + Redevelopment scenario (Figure 6-44A) when Option 2 is used to calculate resident labor force
Tier 1. As a result, the population in Tier 1 increases the most with this combination (Figure 6-44b) due
to the connection between unemployment rate Tier 1 and net migration to Tier 1. Consequently, the
additional population in Tier 1 due to the unemployment rate being underestimated by Option 2 causes
the resident per capita net earnings Tier 1 (Figure 6-45b) to decrease compared to the BAU scenario,
which affects affordability in the model. This was obviously an unintended consequence of the equation
that we chose for our model and one that we hope to fix in future versions of the model. Because of this,
and to allow for more flexibility when using the model, a switch was added, called the "resident labor
force calculation Tier 1 switch," to allow the user to easily switch between the three formulations for
resident labor force and assess the outcomes of each simulation on several indicators across sectors.
230
-------
> 2010 2015 2020 2025 2030 2035 2040
Light Roil + Redev + Resident LF Option 3
— — — Light Roil + Redev + Resident LF Option 2
— — — BAU + Resident LF Option 2
^^^^ Light Roil + Redev + Resident LF Option 1
2000 2005 2010 2015 2020 2025 2030 2035 2040
Light Roil + Redev + Resident LF Option 3
— — — Light Roil + Redev + Resident LF Option 2
— — — BAU + Resident LF Option 2
^^^^ Light Roil + Redev + Resident LF Option 1
Figure 6-44. Structural Test of the Economic Effects of Three Different Formulations for Resident Labor
Force Tier 1: Comparison of (A) Unemployment Rate (%) Tier 1 and (B) Population Tier 1
2005 2010 2015 2020 2025 2030
^—^— Light Rail + Redev + Resident LF Option 3
— — — Light Rail + Redev + Resident LF Option 2
— — — BAD + Resident LF Option 2
^—^— Light Rail + Redev + Resident LF Option 1
2020 2025 2030
v + Resident LF Option 3
2010 2015
Light Rail + Rede
Light Rail + Redev + Resident LF Option 2
BAD + Resident LF Option 2
Light Rail + Redev + Resident LF Option 1
Figure 6 45. Structural Test of the Economic Effects of Three Different Formulations for Resident Labor
Force Tier 1: Comparison of (A) Total Earnings Tier 1 and (B) Resident Per Capita Net Earnings Tier 1
Property Value Structure Test
This test sets all elasticities and effect tables affecting property values to the exact values found in the
literature, in both Tiers, and removes any relationships that were not explicitly quantified in the
literature. We ran this test due to the uncertainty of the model projection of property values and to verify
the need for the multiple changes made to elasticities found in the literature during calibration.
Elasticities had to be changed for various reasons including studies that were specific to very different
contexts, elasticities established for one property type that had to be borrowed for another, and
relationships established at the metropolitan level that were too strong for Tier 1. For example, while
the elasticity of+1.09 between population growth and single family property values from a study at the
metropolitan level (Jud and Winkler, 2002) worked well in Tier 2, the relationship was too strong in Tier
1 and had to be adjusted down to +0.5. Table 6-16 compares the values used in the BAU, in this test, and
all those found in the literature. For the BAU scenario, values shown in orange were borrowed from
literature on another building type. Values shown in red were either altered from their original literature
value during calibration or were not available from the literature and were instead derived from data.
These red and orange values are the ones changed during this test. No test values were given from lot
size with multifamily (MF) and nonresidential (Nonres) property values, as this would have been
duplicative with the relationships with building size. Values that were not changed from the literature
values during the calibration process also remain the same in this test.49
49 When multiple values were available from the literature, we chose values for this test from studies conducted at a similar
scale (citywide rather than on a neighborhood basis). If there were multiple values available at a citywide scale, we chose the
value that better matched the data.
231
-------
Table 6-16. Elasticities Used to Calibrate Property Values in the BAD and the Property Value Structure
Test
VARIABLE
Lot size
(1 /density)
SF Density
Income
Population
growth
Vacant land
Job density
Retail
density
Building
size
Employmen
t
Avg time to
work
SF
(BAU)
0.321
50
(T1)
+1.5
(T2)
+0.17
+1.09
(T2)
+0.5
(T1)
0.38"
(T2),
1.2
max
multip
Her
(T1)
+0.29
1
+.062
-
-
-0.343
SF
(TEST)
-0.321
(T1)
-
+0.17
+1.09
-0.38
+0.291
+.062
-
-
-0.343
SF
(LIT)
+0.00
5;
+0.27
6;
+0.32
1;
+1.45
-
+0.17;
+0.293
; +0.45
+1.09;
+1.53
-0.38
+0.291
+.062;
+.044
-
-
-0.343
MF
(BAU)
-
-
-
(T2),'
1.2
max
multipli
er(T1)
+1.44
(T2)
+0.35
(T1)
(T2)
+0.3
(T1)
-
-0.108
MF
(TEST)
-
-
-
-
-
-
+1.44
-
-
-0.108
MF
(LIT)
-.009;
+0.06*
-
-
-
-
-
+1.44
-
-
-0.108
NONRES
(BAU)
-
-
-
-
-
-
+0.05096
(T2)
+1 .44(11)
+0.994
+1.09
-
NONRES
(TEST)
-
-
-
-
-
-
+0.05096
+0.994
+1.09
-
NONRE
S (LIT)
-0.920
-
-
-
-
-
+0.05096
+0.994
+.00671 ;
+1.09;
+1.92;
-146
-
SOURCES AND
NOTES
(Kain and Quigley,
1 970) (Srour, 2002)
(Kockelman, 1997)
N/A
(Jud and Winkler,
2002); (Hiekkila,
1989); (Capozza et
al, 2002)
(Jud and Winkler,
2002); (Capozza et
al, 2002
(Capozza et al,
2002)
(Srour, 2002)
(Kain and Quiqley,
1970) (Srour, 2002)
(Srour, 2002)
(Srour,
2002)(Dobson and
Goddard, 1992)
(Kockelman, 1997)
Figure 6-46 shows the effect of the changes on multifamily property values in Tier 1. Without the
changes made for the BAU scenario, property values fall well below historical data, and average 20%
below data over the years for which historical data are available, versus only 1.1% above the data under
the BAU scenario.
50 The elasticity of 0.321 found was for single family lot size. Since it is connected with density instead in our model, we
have used the inverse, -0.321.
232
-------
2000 2010 2020
Uncalibrated property values —
2030 2040
•BAU Tierl DATA
Figure 6-46. Multifamily Property Value Per MF Dwelling Unit - Tier 1: BAU and Structural Test
Finally, Table 6-17 summarizes the average yearly percent departure for the three property value classes
and several variables which are affected by them from BAU 2000-2040. Since the least literature was
available to support relationships with multifamily property values, the most changes had to be made
and they are therefore the most affected by the test.
Table 6-17. Average Yearly Percent Departure from BAU, 2000-2040, Property Value Structure Test51
Property Value Test v BAU
Variable
Absolute
percent
deviation or below
Percent above
Tier 1
Nonresidential property value per sq ft Tier 1
SF property value per SF DU Tier 1
MF property value per MF DU Tier 1
Median annual renter costs Tier 1
Tier 2
Nonresidential property value per sq ft
SF property value per SF DU
MF property value per MF DU
Median annual renter costs
19%
8.6%
26%
14%
0.0%
10.2%
27.2%
6.4%
16%
8.6%
-25%
-13%
0.0%
-10%
-27%
-6.4%
Structure of Water Demand
The current model calculates municipal water demand by disaggregating demand into sectors such as
residential and nonresidential demand. In this test, we compare the current sectoral structure with an
earlier structure which used constant per-capita demand in order to evaluate whether the sectoral
structure is more realistic. The per capita formulation is the blue line in Figure 6-47a-b, in which
municipal water demand equals per-capita demand (169 gal/person/day, based on 2005 Durham County
water use) times total population. In contrast, the current model estimates demand separately in the
51 Nonresidential property value per sq ft in Tier 2 does not change because all elasticities remain the same in the test (none
differed from the literature values).
233
-------
residential, nonresidential, and nonrevenue sectors, then sums them. Each sector has an exogenous trend
for water use intensity. In addition, residential demand is proportional to number of dwelling units,
nonresidential demand is proportional to number of employees, and nonrevenue demand is proportional
to population.
(A) Tier 1
2OOO
2O10
2O20
2O30
2O40
-sectoral structure
-constant per capita demand
(B) Tier 2
2OOO
2O10
2O20
2O30
2O40
sectoral structure
^^^^—constant per capita demand
Historical Data and TRWSP Projections
Figure 6-47. Structural Analysis for Water Demand: Sectoral Formulation Compared to Per
Capita Formulation
The per capita structure yields, on average, 25% higher water demand than the sectoral structure in Tier
1, and on average 63% higher demand than the sectoral structure in Tier 2. Since the sectoral structure
is calibrated to historical water demand data as well as to projected water demand in the Triangle
Regional Water Supply Plan, it is not surprising that it gives a more accurate view of demand. In Tier 2,
the sectoral structure differs from historical data and TRWSP projections by an average absolute value
of 0.64% (Figure 6-47b). Moreover, assuming a constant per capita water demand would ignore the
effects of more water efficient technology and/or conservation practices. These effects are seen in Tier
2 in the drop in demand after 2005 and the lower slope of demand growth between 2020 and 2040 in the
sectoral structure compared to the per capita structure. Since historical data and projections were only
available for Tier 2, the fact that the sectoral structure shows a better fit with Tier 2 data suggests that
this structure is also more accurate than a per capita structure in Tier 1.
Structure of Stormwater Runoff
Our model estimates Stormwater runoff using the Simple Empirical Method (Shaver et al. 2007); in this
structural test we explore how aggregating vs disaggregating the Stormwater structure by land use affects
results. An earlier version of our model estimated Stormwater N load by aggregating all land uses within
each Tier and applying an aggregate percent impervious surface to calculate Stormwater runoff volume
and event mean concentration. The orange line in Figure 6-48 shows this aggregated form of the model.
The current model disaggregates Stormwater N load by land use type (SF residential, MF residential,
nonresidential, etc.) and calculates Stormwater runoff volume and event mean concentration separately
for each land use type. The total Stormwater N load for the disaggregated structure is the blue line in
Figure 6-48.
234
-------
(A) Tier 1
(B) Tier 2
8
>-.
to
2OOO
2O10
2O20
2O30
2O40
2OOO
2O10
2O20
2O30
2O40
•BAD, land uses aggregated
•BAU
•BAD, land uses aggregated
•BAU
Figure 6-48. Structural Analysis for Stormwater N Load: Land Uses Aggregated Compared
to Disaggregated
The two structures show similar behavior in N load over time, but N load for the aggregated structure is
on average 20% higher than for the disaggregated structure in Tier 1, and 11% higher than the
disaggregated structure in Tier 2. One reason for this is how event mean concentrations (EMCs) are
calculated. The aggregated structure assumes all impervious cover has an EMC of 1.44 mg N/L, but the
disaggregated structure has different EMCs for each land use. In the disaggregated structure, the
residential and nonresidential land uses representing the majority of Tier 1 and Tier 2 storm water N load
assume 1.08 mg N/L for roofs and 1.44 mg N/L for parking lots and driveways.
Stormwater runoff volume is nearly identical in the aggregated structure compared to the disaggregated
structure (graphs not shown), and this consistency is expected, given that runoff volume is modeled as a
function of impervious surface. In Tier 1, runoff volume from the aggregated structure differs from the
disaggregated structure by an average absolute value of 0.04%, while in Tier 2 they differ by 0.58%. We
take these small differences as rounding errors, and as support that we are applying the equations
consistently.
Structure of Vehicle Emissions Effects on Health
We tested several alternative structures for modeling the health effects of vehicle emissions in the D-O
LRP SD Model. The EPA photochemical modeling study (US EPA 2013c) that estimated the national
health benefits per ton of removing pollutants from various sources, including on-road mobile sources,
used health benefit estimates from two epidemiological studies: Krewski et al. (2009) and Lepeule et al.
(2012). Based on advice given by one of the authors of that report (US EPA 2013c), the default benefit-
per-ton estimate chosen for the model is from the former, by Krewski et al. At first we were only going
to model the difference between scenarios in avoided premature mortalities from reducing (or
increasing) direct PM2.5 emissions from vehicles; however, under the advice of the EPA report author,
we included effects of NOX emissions (a PM2.5 precursor) from vehicles as well.
The structural tests in Figure 6-49 show the three different options for estimating the health effects of
changes in vehicle emissions (relative to BAU) in the model. The Light Rail + Redev scenario (default,
red) shows the combined effects of changes in PM2.5 and NOX vehicle emissions (relative to the BAU)
on premature mortality based on Krewski et al. (2009), whereas the "Light Rail + Redev Lepeule" case
(orange) uses estimates from Lepeule et al. (2012). The Lepeule values yield 123% higher premature
235
-------
mortalities than the Krewski values, on average, between 2020 and 2040 in Tier 2, the time period
reflecting effects of the D-O LRP. The other structural test of excluding NOX vehicle emissions effects
on premature mortality (leaving only PM2.5 effects), is shown in Figure 6-49 in yellow. The PIVh.s-only
case, "Light Rail + Redev PIVh.s only," uses the health effects estimates from Krewski et al. (2009) and
produces 26% fewer premature mortalities on average during 2020-2040, compared to including both
PM2.5 and NOX effects (Light Rail + Redev). Therefore PIVh.s represents the majority of vehicle-
emissions-related premature mortalities in this model, but leaving outNOx emissions would
significantly underestimate emissions-related premature mortalities. Further, the default Krewski et al.
mortality-per-ton estimate results in roughly half the number of premature mortalities as the Lepeule et
al. estimate. However, because air emissions are projected to have a very small impact on overall
premature mortalities (relative to other factors such as walking and cycling), the downstream impacts of
these structural changes are negligible.
0.14
0.12
03
S, 0.1
£. 0.08
f °'06
°- 0.04
0.02
0
-n n?
/
| ^£^f^^
t
/
/
/
/ *
/ S^
^^X
^
2000 2010 2020 2030
Light Rail + Redev PAA2.5 only
tail + Redev Lepeule
tail + Redev
2040
Figure 6-49. Structural Analysis for Health Effects of Prri2.5 and Nox Vehicle Emissions: Total Premature
Mortalities Relative to BAD - Tier 2
Extreme-Condition Tests
Direct extreme-condition testing is a very important step in the validation of the D-O LRP SD Model.
Testing the model for extreme conditions involves evaluating the validity of the model's structure under
extreme (not necessarily plausible) conditions, by assessing the coherence of the resulting values against
the knowledge or anticipation of what would happen under a similar condition in real life. We present
five extreme-condition tests below, including 1) Zero person miles, 2) Extreme population drop, 3)
Extreme unemployment, 4) Economic crash, and 5) Extreme gas price spike.
Zero Person Mile Test
In this test, we set initial person miles of travel per day to zero for all modes of transportation. All
drivers of person miles of travel in the D-O LRP SD Model are reliant upon elasticities that cause an
initial value to be multiplied by a dynamically-determined product. Therefore, if person miles of travel
are zero in the year 2000, we expect them to remain at zero for the entire model run. As expected, this
change resulted in person miles of travel by all modes to remain at zero for the entirety of the model run.
An example of this, for public transit person miles per day, is shown in Figure 6-50A. This is consistent
with the real-world expectation that people's past travel behavior (or lack thereof) predicts their future
travel behavior, unless events transpire to change it.
236
-------
In this test, VMT is consistently reduced, but does not go to zero (Figure 6-50). This is because, even
though VMT for trips that either start or end in the study area go to zero in this test case, through-traffic
VMT (which, in the BAU case, is over a third of all Tier 2 VMT throughout 2000-2040) does not. In
fact, due to reduced traffic congestion, through-traffic VMT (not shown) is slightly greater than BAU.
As expected, because nobody does any walking or cycling in this test case, the absence of the health
benefits of those nonmotorized travel modes results in far fewer premature fatalities avoided (Figure
6-50). This test case produces a GRP slightly greater than BAU, because there is no traffic congestion
(Figure 6-50). This result is consistent with the fact that the model does not contain a feedback for the
effect of health outcomes on the economy, as well as the fact that the model does not assume that
transportation is a necessary precondition for engaging in economic activities above a certain level, both
of which are limitations of the model.
(A) Public Transit Person Miles - Tier 2
(B) VMT - Tier 2
200
0
2000 2010 2020
^^—Zero Person Mile Test
BAU
2010 2020
-Zero Person Mile Test
BAU
(C) Cumulative Premature Mortalities
Avoided Due to Walking and Cycling
Relative to BAU - Tier 2
(D) Gross Regional Product - Tier 2
3 1
fO
v~° °
35-4
1 5
' — •— --^^^
^•^
^^^
'
^^
^^^^^
^^
£ "°
H 2000 2010 2020 2030 20
^— Zero Person Mile Test ^— BAU
-Zero Person Mile Test
-BAU
Figure 6-50. Extreme Value Test of Zero Person Miles by Any Mode from Start of Model Run vs. BAU
Extreme Population Drop Test
In this test, we created new outflows from Tier 1 and Tier 2 population, set to reduce population by
approximately 70% around the year 2020. Births, deaths, and net migration are all functions of current
population. Therefore, in this test, Tier 1 and Tier 2 population both remain far below BAU during
2020-2040. In Tier 1, this dramatic population drop has very little effect on GRP (Figure 6-51 A). This is
because the primary determinants of GRP are employment and nonresidential square feet (both of which
exist in feedback loops with GRP), and whatever portion of Tier 1's labor force needs cannot be met by
Tier 1 residents are presumed to be met by people commuting into Tier 1 . However, Tier 2 GRP is
dramatically less than BAU after 2020 (Figure 6-5 IB). This is because the Tier 2 labor force is assumed
to be a direct function of Tier 2 population, in keeping with the expectation that the larger a geographic
237
-------
area is, the smaller is the percentage of its workers who commute from outside the area. When
population and employment drop, so do earnings, and hence GRP. As GRP drops, so does retail
consumption, and hence desired employment. Around 2025, this feedback causes Tier 2 desired
employment to be less than the Tier 2 labor force, which decreases employment and GRP even further.
This forms a vicious cycle, such that, even after its dramatic post-2020 drop, Tier 2 GRP continues to
decline through the year 2040, as opposed to increasing during 2020-2040, as it would in the BAU case.
Tier 2 net migration is modeled as being mostly insensitive to economic influences, while Tier 1 net
migration is sensitive to economic influences. Since Tier 1 population declines while Tier 1 GRP does
not, Tier 1 unemployment goes down and the Tier 1 employment gap increases, causing an increase in
Tier 1 migration. As a result, 2040 Tier 1 population is 44% of BAU, whereas 2040 Tier 2 population is
32% of BAU.
The numbers of dwelling units and square feet of nonresidential floor space do not automatically change
when population changes. Therefore, the usual expectation is that a sudden drop in population will cause
real estate demand to decrease relative to supply, meaning that property values will decrease. However,
in this test case, the opposite occurs in Tier 1, with a sudden drop in population triggering an increase in
the value of nonresidential square feet of floor space, single-family dwelling units, and multifamily
dwelling units (Figure 6-51C). This is because, since Tier 1 population decreases but Tier 1 GRP does
not, the amounts of jobs, earnings, and retail establishments/?^ capita all become much higher in 2020
in this test case, driving up Tier 1 property values, an unrealistic outcome that represents a shortcoming
of the model. In the context of Tier 1, it is reasonable that economic activity would not necessarily
decline when population declines, or at least not decline by the same proportion, since, unlike in Tier 2,
most Tier 1 workers and a significant portion of Tier 1 retail customers reside outside of Tier 1.
Meanwhile, Tier 2 property values drop below BAU as a result of population being reduced, in keeping
with expectations. However, in the case of Tier 2 multifamily dwelling units (but not single-family
dwelling units or nonresidential floor space), property values are far above BAU during 2020-2021,
before dropping below BAU for the remainder of the model run (Figure 6-5 ID). This is because the
value of multifamily dwelling units is highly elastic to retail density (the number of retail establishments
per capita). After population drops in 2020, retail density spikes upward, then reacts to reductions in
Tier 2 employment and GRP, ultimately stabilizing at a value slightly below BAU. This follows the
expectation that after people leave the metropolitan area, demand for retail goods declines, after which
there is a delay of some amount of time before businesses start closing due to a lack of customers.
However, the increase in multifamily dwelling unit property values during that period of delay is much
less realistic.
Because of the aforementioned effects on GRP and property values, it is not surprising in this test case
that Tier 1 and Tier 2 transportation and renter costs per year per household stabilize at values above and
below BAU, respectively. However, in both Tiers, these costs spike upward in 2020 and take several
years to come down from that spike (Figure 6-51E-F). Renter costs change because they are, in part,
affected by the increase in multifamily property values described above. Transportation costs respond
this way because the number of motor vehicles in the study area does not immediately respond to
reductions in population, even though the excess cars are not being driven. Whereas vehicle purchases
are a function of vehicle demand (such that demand is driven by population and earnings), vehicles are
assumed to only be retired from the study area when they reach the end of their useful life (assumed to
be an average of 17.5 years after they were purchased), rather than also moving out of the study area in
the event of their owners moving away. As a result, there is a period of time during which it is assumed
that all of the cars owned by the 70% of the population who moved away are still in the study area, and
their fixed costs of ownership (including car payments, insurance, and maintenance, and not including
238
-------
fuel or parking) continue to be paid by the people who still live in the study area. However, increases in
population much more immediately produce increases in vehicle stock. Therefore, in scenarios where
population mostly increases over time (including the BAU, Light Rail, and Light Rail + Redevelopment
scenarios), this difference between expectations and model results does not appear. This test
demonstrates that the model is designed to accommodate increases in population more realistically than
sharp decreases in population.
(A) Gross Regional Product - Tier 1
(B) Gross Regional Product - Tier 2
0
0
0
0
0
0
0
•—•••——
^_— — — —
•
""
00 2010 2020 2030 20
= 10
-Extreme Population Drop
-BAU
•Extreme Population Drop
-BAU
(C) Multifamily Property Value Per
Multifamily Dwelling Unit - Tier 1
(D) Multifamily Property Value Per
Multifamily Dwelling Unit - Tier 2
350
-Extreme Population Drop
-BAU
-Extreme Population Drop
-BAU
(E) Transportation and Renter Costs Per
Year Per Household - Tier 1
(F) Transportation and Renter Costs Per
Year Per Household - Tier 2
-BAU
Extreme Population Drop BAU ^—Extreme Population Drop
Figure 6-51. Extreme Value Test of 70% Population Drop at 2020 vs. BAU
Extreme Unemployment Test
In this test, we created a test case where Tier 1 and Tier 2 unemployment are both set at 50% during the
period 2020-2040, regardless of the drivers that would normally cause unemployment to increase or
decrease. In both Tiers, the unemployment rate directly determines the poverty rate. However, in Tier 1,
it also affects net migration. In the D-O LRP SD Model, if Tier 1 unemployment is greater than 7.5%,
net migration drops to zero, approximating a situation where it is no longer attractive to move to Tier 1.
239
-------
Therefore, unsurprisingly, this test case causes Tier 1 population to be increasingly less than BAU after
2020 (Figure 6-52A). Since Tier 1 unemployment does not exceed 7.5% in any of the scenarios we ran
outside of extreme-condition testing, the model was not designed to consider the possibility of
emigration due to high unemployment. Meanwhile, because Tier 2 net migration is assumed to not be
affected by unemployment rates, Tier 2 population in this test case differs from BAU by very little
(Figure 6-52B).
As increased unemployment increases the poverty rate, it also increases the number of zero-car
households. In the D-O LRP SD Model, zero-car households have a negative effect on VMT. However
this effect is small, owing to the very small percentage of people who live in zero-car households in the
BAU case. Because baseline zero-cars households are so few, the increase in such households that
results from 50% unemployment has a very small effect on VMT. In Tier 1, VMT is less than BAU
mostly because of population being lower (Figure 6-52). In Tier 2, VMT's difference from BAU is
negligible (Figure 6-52).
(A) Population - Tier 1
(B) Population - Tier 2
50
_O)
8 40
-5096 Unemployment 2020
-5096 Unemployment 2020
(C) VMT - Tier 1
(D) VMT - Tier 2
2000
-5096 Unemployment 2020
2010 2020
096 Unemployment 2020
2030
-BAU
Figure 6-52. Extreme Value Test of Unemployment Becoming 50% During 2020-2040 vs. BAU
Economic Crash Test
In this test, we modeled a sudden 20% drop in GRP in the year 2020, with the assumption that the
recovery from this economic crash will take at least a decade. To do this, we added an input variable to
Tier 1 and Tier 2 GRP that multiplies the pre-existing GRP formula by a value that is equal to one
through the year 2019, declines to 0.8 during 2019-2020, then ramps back up to one during 2020-2030.
In both Tiers, GRP remains well below BAU through the year 2040 (Figure 6-53). This is because of a
reinforcing feedback loop, wherein lowering GRP lowers retail consumption, which lowers desired
employment, which lowers actual employment, which lowers both earnings and nonresidential floor
space, hence lowering GRP, as discussed in Section 3.2. This is consistent with the expectation that a
temporary economic shock may dampen economic activity for a very long time afterwards, unless other
events intervene.
240
-------
Although Tier 2 GRP does not get closer to BAU after the economic crash, it does continue to increase,
albeit at a growth rate that is slightly less than BAU. However, Tier 1 GRP continues to decline in
absolute terms after the end of the economic crash, thanks to a difference between the Tiers in how
nonresidential floor space is determined. In both tiers, a reduction in overall square feet of nonresidential
floor space, including retail, office, service, and industrial floor space, reduces GRP. Also in both Tiers,
the addition or subtraction of office, service, and industrial land and floor space is driven by the number
of office, service, and industrial workers, respectively. This makes sense, given that businesses must
acquire whatever amount of floor space is necessary for the particular size of their operations, which is
usually proportional to how many workers they have, depending on the type of business. However, Tier
1 and Tier 2 differ in how the amount of retail floor space is determined. Whereas the customers of
businesses that use office, service, or industrial floor space may or may not be located in the same
metropolitan area, the customers of retail establishments may be assumed to mostly come from the same
metropolitan area, unless tourism is an unusually large proportion of the economy, which it is not in the
DCHC MPO. Therefore, while other types of businesses may have an unlimited number of customers
from anywhere in the world, the amount of retail that a metropolitan area can support is limited by its
population. For that reason, Tier 2 retail floor space is assumed to be driven by population (i.e., the retail
customer base), rather than the number of retail employees. Since the economic crash is assumed to not
significantly affect Tier 2 population, it does not significantly affect Tier 2 retail floor space either,
hence mitigating the reduction in overall Tier 2 nonresidential floor space that the economic crash
triggers. Meanwhile, because Tier 1 is one small part of a much larger metropolitan area, its retail
customers need not come from Tier 1. Therefore, in the D-O LRP SD Model, Tier 1 retail floor space is
driven by employment levels in the same fashion as all other types of nonresidential floor space. That
means that as the economic crash depresses Tier 1 retail employment, it also reduces the amount of Tier
1 retail land and floor space, hence pushing down Tier 1 GRP even more, in a vicious cycle that is
mirrored in the cases of office, service, and industrial employment and floor space. It is because this
vicious cycle only applies to three of the four major nonresidential employment/land use types that Tier
2 GRP is able to eventually assume an upward trend in the years after the economic crash, rather than
continuing to decline over time, as Tier 1 GRP does. Unfortunately, because population is used as a
proxy for the amount of demand for retail establishments in Tier 2, the model does not account for the
possibility of the amount of retail demand per capita going down, as it would be expected to do during
an economic depression. Therefore, the D-O LRP SD Model's Tier 2 GRP results after the economic
crash may be too high, even though it is also unrealistic for economic activity to decline indefinitely in
the decades after a crash, as in the case of Tier 1.
Unsurprisingly, the economic crash increases the unemployment rate. However, the unemployment rate
increases far more at the Tier 2 level than at the Tier 1 level (Figure 6-53). In Tier 1, prior to the
economic crash, a large percentage of jobs were held by people commuting from outside of Tier 1. The
model assumes, unrealistically, that the total number of jobs held by residents of Tier 1 can never
decrease, making the Tier 1 unemployment rate much less than the Tier 2 unemployment rate in the
event of an economic crash, reflecting the fact that the model was constructed with an aim of modeling
periods of economic growth, rather than periods of economic contraction. In addition, because Tier 1 net
migration declines in response to unemployment (unlike Tier 2 net migration), there come to be fewer
Tier 1 residents competing for the same number of jobs. However, mirroring the GRP trends discussed
above, Tier 2 unemployment eventually plateaus, whereas Tier 1 unemployment does not, since GRP
drives the number of jobs in each Tier. Therefore, conceivably, if the model were extended far enough
into the future, Tier 1 unemployment could eventually overtake Tier 2 unemployment, unless it
eventually also plateaued.
241
-------
GRP is a driver of VMT, given that economic activity provides a reason for travel and higher incomes
make automobile travel more affordable. Therefore, in keeping with expectations, the economic crash
causes VMT to be less than BAU in both Tiers, even though the sensitivity is small (Figure 6-53).
Because the economic crash also causes Tier 1 population to decline relative to BAU, there is a greater
proportional difference in Tier 1 VMT than in Tier 2 VMT.
(A) Gross Regional Product - Tier 1 (B) Gross Regional Product - Tier 2
SJ, 60
S, 50
= 10
-Economy Crash
(C) Unemployment Rate - Tier 1
50%
45%
40%
35%
30%
25%
15*
-Economy Crash
(E) VMT-Tier 1
2010 2020
"Economy Crash
—BAU
(D) Unemployment Rate - Tier 2
40%
35!6
30%
10%
5%
IW
^r
/
I
/
2000 2010 2020 2030 2040
^^—Economy Crash ^^—BAU
(F) VMT - Tier 2
25
-Economy Crash
2010 2020
—Economy Crash
—BAU
Figure 6-53. Extreme Value Test of an Economic Crash Wherein GRP Declines 20% in 2020 vs. BAU
Extreme Gas Price Spike Test
In this test, we adjusted the price of gasoline, which is exogenous in the model, to rise to $1,000 per
gallon during the period 2016-2018, before dropping back down to its BAU value in 2019. Not
surprisingly, this produces a large dip in VMT. After gasoline prices return to normal, VMT goes back
up, but never rebounds all the way to BAU, since the shock of the fuel price spike causes a permanent
contraction of the economy. This economic contraction is worse in Tier 1 than in Tier 2 for the same
reasons cited in the Economic Crash Test section: retail space is sensitive to declines in employment in
Tier 1 but not in Tier 2. Of greater note, however, is that the exorbitant spike in gasoline prices fails to
make VMT drop to zero (or practically zero), as would normally be expected. This is because the D-O
LRP SD Model assumes that the effect of fuel price changes on travel behavior is phased in over a
period of five years. Since the spike in gas prices lasts less than five years, by the time VMT has fully
242
-------
reacted to the first year of unaffordable fuel, the return to BAU gas prices in 2019 has already started
mitigating the effect of the price spike on VMT. The five-year reaction time of travel behavior to
gasoline prices assumes that the elasticity value is a "long-term" elasticity (Litman 2013). If gas prices
change gradually enough (relatively speaking) that people do not necessarily change their driving habits
immediately (such as in the gasoline-price lookup table that is assumed for the BAU, Light Rail, and
Light Rail + Redevelopment scenarios), the outputs of the D-O LRP SD Model are likely to be more
realistic.
In both Tiers, the D-O LRP SD Model produces an output called the affordability index, defined as
relative resident per capita net retail earnings over relative transportation and renter costs per year per
household. Not surprisingly, during the period of the gas price spike, the affordability index is far lower
than at any other time, since the cost of the transportation component of household spending is far
greater than at any other time. However, after gasoline prices return to normal, the affordability index
does not match BAU. During 2019-2040, the Tier 2 affordability index is less than BAU, and the Tier 1
affordability index is greater than BAU. In Tier 2, because the gas price spike makes the economy
permanently contract without population being significantly changed, earnings per capita are far below
BAU, hence reducing the numerator of the affordability index. However, it is assumed in Tier 1 that the
economic contraction will first cause commuters into Tier 1 to lose their jobs, as opposed to people who
both live and work in Tier 1. It is also assumed that economic downturns have more of a dampening
effect on Tier 1 population than on Tier 2 population. For both of these reasons, Tier 1 earnings per
capita are reduced far less relative to BAU than are Tier 2 earnings per capita, which means that the
numerator of the Tier 1 affordability index is much larger than the numerator of the Tier 2 affordability
index. At the same time, the denominator of the Tier 1 affordability index (relative transportation and
renter costs per household) declines more relative to BAU than does that of the Tier 2 affordability
index, resulting in the Tier 1 affordability index being greater than BAU during 2019-2040. This is
because renter costs are driven by GRP, and the economic contraction resulting from the gas price spike
is worse in Tier 1 than in Tier 2. Because the model unrealistically assumes that an economic
contraction like the one in this test case cannot cause the total number of jobs held by Tier 1 residents to
decline (at most, it can be made to plateau) even as overall jobs in Tier 1 decline, this test case results in
a higher percent of Tier 1 jobs being held by Tier 1 residents (relative to BAU), which also increases the
percent of VMT in and out of Tier 1 that is by Tier 1 residents. Therefore, even though overall Tier 1
VMT remains less than BAU for the entire period after the gas price spike and Tier 1 resident
unemployment consistently grows relative to BAU for the remainder of the model run, VMT by Tier 1
residents is greater than BAU from 2024 to 2040, meaning that Tier 1 residents surprisingly end up
spending more on gasoline per household than BAU, five years after fuel prices return to normal.
The inconsistent impacts of a gas price spike on overall Tier 1 VMT and VMT by Tier 1 residents
described above suggest either that too weak of an elasticity is assumed for the effect of GRP on overall
VMT in and out of Tier 1 (by residents and nonresidents) or that too large a percent of workers and
potential workers residing in Tier 1 are assumed to also work in Tier 1. This latter possibility is
especially likely when, as in this test case, the ratio of jobs in Tier 1 to the size of the Tier 1 resident
labor force is particularly low (due to the economic contraction and the moderate effect it has on
population, even in Tier 1). In that event, there are lower odds of any given Tier 1 resident being
qualified to fill one of the subset of Tier 1 jobs that are not already held by someone else, hence making
them more likely to either seek employment outside of Tier 1 or remain unemployed.
243
-------
(A) VMT - Tier I
(B) VMT - Tier 2
-Extreme Gas Price Spike
(C) Affordability Index - Tier 1
Relative in come /relative cos
o o o o
ON)kc>CO-i^
1
\
\
\
u
00 2010 2020 2030 20
^^— Extreme Gas Price Spike ^^^BAU
2010 2020
-Extreme Gas Price Spike
(D) Affordability Index - Tier 2
• Extreme Gas Price Spike
Figure 6-54. Extreme Value Test of Gas Price Spike to $1,000 Per Gallon During 2016-2018 vs. BAD
Unit Consistency
Unit consistency was checked and ensured both during model development and after the completion of
the D-O LRP model. (See Table 6-18 for a list of selected indicators and their unit of measure.) Further,
Vensim has a specific feature, called "units check," that allows users to quickly identify errors or
inconsistencies in the units used in the model. Still, for models of this type and size, it is very likely that
Vensim will identify unit "errors" even when the units are correct. Below are examples for why this may
happen, but it should be also noted that unit errors do not impact the simulations and the quality of the
results generated.
• Vensim requires that every argument of an equation be represented by a variable with a unit of
measure. As an example, if we use an equation such as "A = B+10," Vensim will give us a unit
error because a unit for "10" is not provided. To avoid unit errors we would need to have the
following equation: "A = B+C", with C=10 and the same unit specified for each of the three
variables (given that we are adding B and C). This can happen when a system dynamics model
uses equations based on literature without specifying a unit for each number in the equations.
• Vensim will also return a warning if a lookup table is used with an input containing units such as
Time. Such warnings do not affect model behavior and are more a convention of Vensim
software. To fix this, the input would need to be replaced with a unitless version of that variable
(such as time divided by 1 year, so that it is unitless).
244
-------
Table 6-18. Selected Quantitative Indicators and Unit of Measure
SECTOR
Land use
Land use
Land use
Economy
Economy
Economy
Transportation
Equity/Transportation
Energy
Energy
Energy/Economy
Land use/Water
INDICATOR (VARIABLE NAME IN QUOTES, IF DIFFERENT)
Population
Nonresidential sq ft ("Total nonresidential sq ft")
Developed land
Total retail consumption
Employment ("Total employment")
Gross regional product - GRP
VMT
Zero car households
Total energy spending
Cumulative CO2 emissions ("CO2 emissions cumulative")
CO2 emissions per GRP
Impervious surfaces ("Total impervious surface")
UNIT
person
sqft
acre
USD2010/year
person
USD2010/year
mile/day
dwelling unit
USD2010/year
ton
ton/USD 2010
acre
Model Behavior and Sensitivity Tests
The direct structure tests discussed above are designed to evaluate the validity of the model structure.
Once these tests have established an adequate level of confidence in the validity of the D-O LRP
model's structure, we apply model behavior and sensitivity tests. These examine the extent to which
model behavior changes when certain parameters are adjusted, to determine which parameters influence
model behavior most and least. It is crucial to note that the emphasis is on pattern prediction (periods,
frequencies, trends, phase lags, amplitudes, etc.) rather than point (event) prediction. These analyses can
be carried out by modifying one or more model inputs, to test the impacts of changing assumptions for
several variables simultaneously. Three different types of model behavior and sensitivity analyses have
been carried out to test the D-O LRP model: numerical, behavior mode, and policy sensitivity. We
present sixteen tests, categorized by type, below.52
52 In addition to the tests described in this report, model users can conduct additional sensitivity analyses within the Vensim
software itself. For users of Vensim DSS (but not PLE), several tools are provided to evaluate behavioral validity against
historical data, such as minimum, maximum, mean, median, standard deviation. This information is available in the
"Statistics" tool of Vensim, and this type of result can be estimated for every variable and simulation in the model.
245
-------
In conducting behavior and sensitivity tests, we have applied the same criteria as we described in the
model parameterization (calibration) section above. In fact, several projections have been presented and
evaluated in this document already. These include, among others:
• A modification of the structure governing land development, analyzing the impact on the
distribution of land use by type, economic performance, and property values (see Figure 6-31,
Figure 6-32, and Table 6-9).
• Changes in assumptions concerning the drivers of nonmotorized travel facilities, analyzing the
impact on mode choice and person miles (see Figure 6-36 and Table 6-13).
• An overview of the impact of removing the link between GOS and energy spending, analyzing
the impact on GRP (see Figure 6-39).
Several indicators in the model were not tested against historical data due to limitations in data
availability. If additional data were to become available - either historical data or projections from other
models - we could conduct additional behavior pattern tests to validate and further calibrate the model's
structure. The following list provides a sample of the type of data that would allow for such tests:
Data "Wish List"
• Developed land use by type covering additional historical years;
• Single family and multifamily household sizes covering additional historical time periods;
• Property values by land use from Orange and Chatham counties covering prior years;
• Historical VMT and traffic congestion data from before 2010;
• Tier 1 data/projections for person miles of travel by mode, comparable to those provided for the
DCHC MPO in the Metropolitan Transportation Plan;
• Local stormwater N and P load data as a historical time series;
• Local impervious surface data as a historical time series; and
• Local data on average passenger vehicle MPG (since MPG in the model had to be calibrated
lower than the national average MPG).
Numerical Sensitivity
Numerical sensitivity exists when a change in assumptions affects the numerical values of the results.
All models exhibit numerical sensitivity. We present nine numerical tests below, covering inputs in the
Land Use, Transportation, Energy, Economy, Water, and Health sectors.
246
-------
Sensitivity to Nonresidential Impervious Surface Coefficients
This test runs on top of the BAU scenario and tests the sensitivity of impervious surface and water
quality indicators to the impervious surface coefficients (ISCs). The model applies coefficients for each
of the six land use types, as well as for agricultural land, vacant and park land, and four classifications of
roadways (highway, urban nonhighway, rural nonhighway, and nonmotorized travel facilities).
The Residential ISC table from the Jordan Falls stormwater load accounting tool manual fit the data on
impervious surfaces better than the California residential ISC table, despite being less detailed (it was
decidedly lower at each density). However no local ISCs were available for the commercial uses, so we
used the values from the California source, which had to be calibrated to match the data. They were
reduced from the mean values cited in the report by 15% across the board in Tier 2, though they
remained the same in Tier 1. Because there was no precedent for the values chosen and historical data
were very limited for purposes of verification, we ran this test to see the sensitivity of other model
variables to the change.
This test sets all nonresidential ISCs to their mean values in the California report in Tier 2 to explore the
effects of undoing the calibration step described above. This causes an overall increase in estimated total
impervious surface and percent impervious surface, on average 4.8% higher than the BAU over the
model duration (Figure 6-55). This difference is entirely due to a 17.5% higher nonresidential
impervious surface projection on average over the model time frame.
2000
2010
• ISC Sensitivity Test
2020
2030
2040
• BAU
EPA EnviroAtlas
Figure 6-55. Total Impervious Surface - Tier 2: Sensitivity to Nonresidential Impervious
Surface Coefficients
The test change in impervious surfaces leads to a smaller than proportional average % change in N load
due to nonresidential uses, 5.4% average deviation from BAU, indicating that it is not very sensitive to
the impervious surface coefficients. The volume of runoff due to nonresidential uses changes
approximately proportionally, with an average deviation from BAU of 16%, compared to the 17.5%
deviation of nonresidential impervious surface.
247
-------
Table 6-19. Average Yearly Percent Departure from BAD, 2000-2040: Sensitivity to Nonresidential
Impervious Surface Coefficients
Variable
Tier 2
Total impervious surface
Total nonresidential impervious surface
Percent impervious surface
N load nonresidential
Volume of runoff nonresidential
ISC Test v BAD
4.8%
17.5%
4.8%
5.4%
16%
Weighting the Effects of Tier 1 and Tier 2 Congestion. Fuel Cost, and Parking Cost on Tier 1 Travel Behavior
When people decide what mode of transportation to use on the basis of traffic congestion levels (which
can affect mode choices in the model either directly, through the effect of congestion on travel speed, or
indirectly, by affecting vehicle fuel efficiency and overall fuel costs per VMT), conditions at any point
along their entire travel route may influence their decision. Meanwhile, even though the cost of parking
at someone's destination has more of an effect on their travel decisions than the cost of parking at their
origin point, the D-O LRP SD Model does not distinguish between the beginnings and ends of trips, so
both trip ends are assumed to be equally likely to be the destination for purposes of calculating parking
costs. Because Tier 2 encompasses an entire metropolitan area, it may be reasonably assumed that most
of the trips that either start or end in Tier 2 have their entire route and both of their ends within Tier 2.
However, since Tier 1 is much smaller than Tier 2, it is probable that a large percentage of trips that
either begin or end in Tier 1 have their other end somewhere else in Tier 2. Therefore, the D-O LRP SD
Model assumes that traffic congestion, fuel costs per VMT, and parking costs in all of Tier 2 have an
influence on Tier 1 modal person miles, vehicle stock, and trip distances. The model also assumes that
Tier 1 travel behavior is affected more by traffic congestion, fuel costs per VMT, and parking costs in
Tier 1 than by the same factors in the rest of Tier 2. To incorporate these assumptions into the model, we
developed two sets of factors to represent the effects on any given variable of vehicle speed, fuel cost
per VMT, and the parking cost of an average trip: one for Tier 2 and one for Tier 1 as it would be if Tier
1 travel behavior were solely a function of Tier 1 inputs. Then, we created a 50-50 weighted average
factor from these two separate factors, which we applied to Tier 1 baseline values. This sensitivity test
examines two cases: (1) Tier 1 travel behavior being solely a function of Tier 1 conditions (Tier 1
Weighting Factor All Tier 1 test case, with the weighting-factor parameter set to one), and (2) Tier 1
travel behavior being solely a function of conditions in all of Tier 2 (Tier 1 Weighting Factor All Tier 2
test case, with the weighting-factor parameter set to zero).
As shown in Figure 6-56 and Table 6-20, there is very little apparent sensitivity of any variable to the
Tier-l-to-Tier-2 weighting factors just described. In part, this is a result of Tier 1 and Tier 2 vehicle
speeds and fuel costs per VMT differing from their year-2000 values by proportions that are not far
removed from one another, thanks to their comparable inputs. Meanwhile, Tier 1 and Tier 2 parking
costs per trip differ from their year-2000 values by significantly different proportions over the course of
2000-2040, but modal person miles of travel have very small elasticities to the cost of parking, another
cause of low sensitivity. Figure 6-56 and Table 6-20 also show that a 100% decrease in the weighting
factors and a 100% increase in the weighting factors produce proportional deviations from BAU that are
very near to mirror images of one another. This is an expected result, given that most variables are
calculated with either the same or very similar equations in both Tiers.
248
-------
Because the primary effect of high traffic congestion, low vehicle speeds, high fuel costs per VMT, and
high parking costs per trip on travel behavior is to discourage automobile driver travel and encourage
travel by other modes, the usual expectation is that a change in the model that increases or decreases
automobile driver person miles will have an opposite-direction effect on the other modes. However, as
shown in Table 6-20, the average yearly deviations of Tier 1 public transit and nonmotorized person
miles from BAU in the Tier 1 Weighting Factor All Tier 1 and Tier 1 Weighting Factor All Tier 2 cases
are in the same direction as the average yearly deviations of Tier 1 automobile driver person miles from
BAU in those same two cases, respectively. The reason for this result may be seen in Figure 6-56. In
each of these two test cases, automobile driver, public transit, and nonmotorized person miles first
become either more or less than BAU, then follow a trend in the opposite direction. In keeping with
expectations, the initial deviation from BAU for public transit and nonmotorized person miles is in the
opposite direction from the deviation for automobile driver person miles. However, because traffic
congestion and vehicle speeds have a larger and more direct impact on automobile driver person miles
than on person miles by other modes, the changes in automobile person miles relative to BAU in the
early years of the period 2000-2040 take longer to be entirely reversed than do the smaller early-year
changes in public transit and nonmotorized person miles relative to BAU. Because the deviation from
BAU for public transit and nonmotorized person miles switches polarity in an earlier year, their values
after that point in time have a greater influence on their average yearly percent departure from BAU
during 2000-2040 than do their values prior to that point in time. Meanwhile, the opposite is true for
automobile driver person miles, which explains the fact that the average yearly percent departures from
BAU are in the same direction for automobile driver, public transit, and nonmotorized person miles in
Table 6-20.
(A) Automobile Driver Person Miles - Tier 1 (B) Automobile Passenger Person Miles - Tier 1
1.0%
0.5%
0.0%
-0.556
-1.056
-1 S%
^
"^ -*
2000 2010 2020 2030 2040
-Tier 1 Weighting Factor All Tier 1 ^—Tier 1 Weighting Factor All Tier 2
(C) Public Transit Person Miles - Tier 1
2000 2010 2020 2030 2040
Tier 1 Weighting Factor All Tier 1 Tier 1 Weighting Factor All Tier 2
(D) Nonmotorized Person Miles - Tier 1
1.5
2000 2010 2020 2030 2040
—Tier 1 Weighting Factor All Tier 1 ^^—Tier 1 Weighting Factor All Tier 2
2000 2010 2020 2030 2040
-Tier 1 Weighting Factor All Tier 1 Tier 1 Weighting Factor All Tier 2
Figure 6-56. Tier 1 Person Miles Per Day by Mode: Percent Difference from BAU
249
-------
Table 6-20. Average Yearly Percent Departure from BAD, 2000-2040: Sensitivity to Tier-1-to-Tier-2
Weighting Factors
VARIABLE
TIER 1 WEIGHTING FACTOR
ALL TIER 1 V. BAU
ABSOLUTE
PERCENT
DEPARTURE
PERCENT
ABOVE OR
BELOW
TIER 1 WEIGHTING FACTOR ALL
TIER 2V. BAU
ABSOLUTE
PERCENT
DEPARTURE
PERCENT
ABOVE OR
BELOW
Tierl
Tier 1 to Tier 2 weighting factors
Vehicle speed
Fuel cost per VMT
Parking cost of average trip
Automobile driver person miles
Automobile passenger person miles
Public transit person miles
Nonmotorized person miles
Total person miles
VMT
Congestion
Vehicle stock
Vehicle trip distance
GRP
Population
100%
0.1%
0.0%
0.0%
0.2%
0.3%
0.3%
0.2%
0.1%
0.2%
0.2%
0.0%
0.6%
0.0%
0.0%
+ 100%
0.0%
0.0%
0.0%
-0.2%
+0.3%
-0.3%
-0.1%
-0.1%
+0.1%
+0.1%
0.0%
+0.5%
0.0%
0.0%
100%
0.1%
0.0%
0.0%
0.2%
0.3%
0.4%
0.2%
0.1%
0.2%
0.2%
0.0%
0.6%
0.1%
0.0%
-100%
0.0%
0.0%
0.0%
+0.2%
-0.3%
+0.3%
+0.1%
0.0%
-0.1%
-0.1%
0.0%
-0.6%
-0.1%
0.0%
Note: All Tier 2 differences from the BA U scenario in these test cases had magnitudes of less than 0.1 %, both in terms of
average percent above or below BA U and in terms of average absolute percent departure. Therefore, Tier 2 outputs are not
shown in this table.
CO2 Emissions Sensitivity to Increases in Energy Efficiency
The Annual Energy Outlook 2015 projects decreases in building energy intensity and increases in
passenger vehicle MPG between 2015 and 2040 (US EIA 2015a). The D-O LRP model incorporates
these projections, but it is useful to determine what effect they have on the BAU scenario. A test case
with no building energy efficiency improvement after 2015 (yellow line, Figure 6-57) results in a 36%
increase in CO2 emissions between 2005 and 2030, compared to 28% increase over the same time in the
BAU scenario (red line). The 2005-2030 timeframe is significant because the Durham County GHG
Plan projects a 48% increase in CO2 emissions during this time, in the absence of energy efficiency
improvement (ICLEI2007). In the D-O LRP model, a test case with neither building nor vehicle
efficiency improvement after 2015 (orange line) closely matches the Durham County GHG Plan
projections, with a 47% increase in emissions between 2005 and 2030. This consistency is expected,
given that the D-O LRP model uses population and VMT projections similar to the Durham County
GHG Plan to calculate energy and emissions.
250
-------
2OOO
2O1O
2O2O
2O3O
2O4O
Figure 6-57.
Tier 2.
• No bldg or vehicle efficiency improvement
No bldg efficiency improvement
• BAU
• Durham City/County Sustainability Office
Sensitivity Analysis for Energy Efficiency: Effect on CC>2 Emissions in the BAU Scenario -
Elasticity of Consumption to GRP
Elasticity of consumption to GRP is an exogenous input that is used in the model to regulate the strength
of the relationship between changes in GRP and retail consumption. The value chosen for both Tier 1
and Tier 2 for this elasticity was 1.1 based on the calibration of retail consumption to historical data.
This sensitivity test shows the behavior of retail consumption and other economic variables when the
elasticity of consumption to GRP is modified. Five different test cases were simulated, with the
elasticity ranging from 1 to 1.2. Figure 6-58 below shows the variance in total retail consumption in Tier
2 as a result of the different elasticities.
2010 2020 2030
• BAU elasticity of consumption to GRP = 1.20
BAU elasticity of consumption to GRP = 1.15
BAU
BAU elasticity of consumption to GRP = 1.05
• BAU elasticity of consumption to GRP = 1.00
Figure 6-58. Sensitivity Analysis for Elasticity of Consumption to GRP: Effect on Total
Retail Consumption - Tier 2
Changing the elasticity of consumption to GRP has a more pronounced effect on total retail
consumption when the elasticity is reduced (rather than when it is increased), due to the balancing
feedback that occurs in the economy sector with employment, which is calculated in the model by
multiplying total retail consumption by the exogenous input "employment per dollar of consumption"
(explained below). The model projects employment per dollar of consumption to decrease over time
(indicating an increase in labor productivity), so because reducing the elasticity of consumption to GRP
decreases retail consumption, employment, earnings, and GRP are reduced further, thus the more
pronounced decline in total retail consumption and other economic variables, shown in Table 6-21.
251
-------
Table 6-21. Average Yearly Percent Departure From BAD, 2000-2040
TIER 2 VARIABLE
Elasticity of Consumption to GRP
Total Retail Consumption
Total Employment
GRP
ELASTICITY =
1.0
-9.1%
-17%
-17%
-16%
ELASTICITY =
1.05
-4.6%
-11%
-11%
-10%
ELASTICITY =
1.15
+4.6%
+5.5%
+3.1%
+2.9%
ELASTICITY =
1.2
+9.1%
+8.4%
+3.4%
+3.2%
Employment Per Dollar of Consumption
Employment per dollar of consumption is an exogenous input to the model that drives desired
employment, a variable that represents the demand for employment that could be filled if the supply of
labor force workers meets the demand. Employment per dollar of consumption was calculated For Tier 2
and Tier 1 based on historical employment and retail consumption data (2000-2010) and projections
(2011-2040). Since total employment projections from the TRMvS SE data grow at a slower rate than
retail consumption projections (Woods & Poole Economics, Inc.), the value for employment per dollar
of consumption decreases each year between 2010 and 2040 (indicating an increase in labor
productivity, as mentioned above). This simulation tests the sensitivity of the model to an alternative
BAU test case where employment per dollar of consumption is held constant at its 2010 calculated value
for both Tier 2 and Tier 1.
•Employment per dollar of consumption constant after 2010
-BAU
•Employment per dollar of consumption constant after 2010
-BAU
•Employment per dollar of consumption constant after 2010
•BAU
•Employment per dollar of consumption constant after 2010
•BAU
Figure 6-59. Sensitivity Analysis for Employment Per Dollar of Consumption: Effect On
Desired Employment (A) and Total Employment (B) for Tier 2 (C) and Tier 1 (D)
Several interesting results emerge from this sensitivity test. First, although desired employment in Tier
2 increases to about 670,000 jobs in 2040 (Figure 6-59, top left), total employment in Tier 2 only
increases to 470,000 jobs in 2040 (Figure 6-59, top right). This is because total employment is limited
by total labor force and, unlike in the Tier 1 economy model, there is no link in Tier 2 that increases net
migration to Tier 2 when there is a (positive) gap between desired employment and total labor force.
When this feedback loop is implemented in Tier 1, employment grows considerably, reaching possibly
unrealistic values. Regarding the internal consistency of the model, given that no explicit links exist
252
-------
between Tier 1 and Tier 2 employment, we note that employment in Tier 1 becomes larger in 2040
(570,000 jobs) than total employment in Tier 2, which encompasses Tier 1. In addition to the link
between desired employment and net migration in Tier 1 (mentioned above), this is due in part to the
2010 value for employment per dollar of consumption in Tier 1 being larger than Tier 2, which causes
employment to grow exponentially faster in Tier 1 due to the reinforcing feedback loop between
employment, GRP, retail consumption, and again employment. In addition, despite the fact that
nonresidential sq ft in Tier 1 hits a cap in this case around 2028, employment in Tier 1 continues to
grow, exposing the need for a balancing loop that inhibits employment growth in Tier 1 when available
nonresidential sq ft runs out. This test highlights that the model is well suited to work with values that
are consistent with historical trends, and important deviations in the area of employment creation (both
for Tier 1 and Tier 2, but for different reasons) have to be carefully analyzed and interpreted.
Drought Year Test
In another sensitivity test, we simulated a drought year for 2030 to determine the effect on stormwater N
load. In this year, annual precipitation drops to 77% of its average value, which is equivalent to the
historical drought in 2007. Because stormwater N load is a linear function of stormwater runoff volume,
and therefore rainfall, stormwater N load also drops to 77% of its projected value in 2030 (Figure 6-60).
In the BAU scenario this is 43,000 Ib N/year, which is lower than the 2000 value of 47,000 Ib N/year.
Therefore the projected increase in Tier 1 N load due to increased development after 2000 is less than
the variation due to a major drought.
30
0
2000
2010
2020
2030
2040
Light Rail + Redevelopment_drought
— Light Rail_drought
BAU_drought
BAU
Figure 6-60. Sensitivity Analysis for Average Precipitation with Drought Year: Effect On Tier
1 Stormwater N Load
Sensitivity to Percent Impervious Surface
As another way of gauging the uncertainty in stormwater runoff, we tested how a 10% decrease or
increase in impervious surface affects Tier 1 stormwater N load. A 10% decrease in impervious surface
causes a 4.6% drop in stormwater N load (Figure 6-61, Table 6-22), while a 10% increase in impervious
surface causes a 3.8% increase in stormwater N load. Stormwater N load is the product of stormwater
runoff volume and event mean concentration. Stormwater N load changes by less than 10% because it is
determined by a runoff coefficient that increases less than in direct proportion with impervious surface.
Furthermore, event mean concentrations increase with a decrease in impervious surface (and vice-versa)
because impervious surface is assumed to have a lower EMC (1.08 mg N/L) than open space or lawns
(2.24 mg N/L).
253
-------
(A)
(B)
7
3
2
1
2000
2010
-BAU -10pct impervious
-BAU
2020 2030 2040
BAU +10pct impervious
2000
2010
2020
2030
2040
-BAU -10pct impervious
-BAU
-BAU +10pct impervious
Figure 6-61. Sensitivity Analysis for Impervious Surface - Tier 1: Effect On (A) Stormwater N Load and
(B) Stormwater N Load Per Acre
Table 6-22. Sensitivity of Stormwater Runoff to Impervious Surface: Average Yearly Percent Departure
from BAU From 2000 to 2040
TIER 1 VARIABLE
Total impervious surface Tier 1
Total N load Tier 1
Total N load Tier 1 Ib per acre
Total volume of runoff Tier 1
Event mean concentration N nonresidential Tier 1
Event mean concentration N roads Tier 1
DEPARTURE FROM BAU
-10%
-4.6%
-4.6%
-8.8%
+7.3%
+3.8%
+ 10%
+3.8%
+3.8%
+8.8%
-7.3%
-3.8%
Health Benefits of NMT: Sensitivity to Miles of NMT per PTT
To test the sensitivity of modeled health benefits from nonmotorized travel, we doubled the NMT per
public transit trip (PTT) from 0.25 to 0.5 miles in the Light Rail + Redevelopment scenario (Light Rail +
Redev Double NMT per PTT test case). In the D-O LRP SD Model, each PTT is normally assumed to
generate an average of 0.25 miles of NMT. This is meant to account for the average distance walked or
biked to a bus stop or light rail station as well as the distance walked or biked to a destination once the
passenger disembarks from the bus or light rail train. Both this additional NMT associated with PTTs
and the NMT generated from exclusively nonmotorized trips are counted towards the model calculation
of health benefits from NMT.
The results of doubling the NMT per PTT in Tier 1 are shown in Figure 6-62, compared to the three
main scenarios. Values are in person miles of nonmotorized travel by residents per day (Figure 6-62 A)
and avoided premature mortalities due to nonmotorized travel (Figure 6-62B). Figure 6-63 A and B show
the results in Tier 2. Although this test was run for the Light Rail + Redevelopment scenario, the Light
Rail and BAU scenarios are also presented to show how the health effects of doubling NMT per PTT
compare to the increases in NMT that result from the Light Rail and Redevelopment.
254
-------
(A)
(B)
.
0)
E 20
2000
2010
2020
2030
2040
Light Rail + Redev Double NMT per PIT
Light Rail* Redev
Light Rail
BAU
2000
2010
2020
2030
2040
• Light Rail + Redev + Double NMT per PIT
• LightRail* Redev
• Light Rail
• BAU
Figure 6-62 Sensitivity Analysis of Doubling the Average Person Miles of NMT Per PIT: Effect On (A)
Person Miles of NMT by Residents Per Day and (B) Avoided Premature Mortality Due to NMT - Tier 1
2000
2010
2020
2030
2040
• Light Rail + Redev + Double NMT per PIT
• Light Rail + Redev
• Light Rail
•BAU
2000 2010 2020 2030
Light Rail + Redev + Double NMT per PTT
Light Rail + Redev
— —Light Rail
BAU
2040
Figure 6-63. Sensitivity Analysis of Doubling the Average Person Miles of NMT Per PTT: Effect on
(A) Person Miles of NMT by Residents Per Day and (B) Avoided Premature Mortality Due to NMT - Tier
2
In Tier 2, doubling NMT per PTT increases total person miles of NMT by residents per day by an
average of 4.1% before the light rail line opens (2000-2025) and by an average of 6.0% after the line
(2026-2040), relative to the Light Rail + Redevelopment scenario. This translates into an average
increase in avoided premature mortality due to NMT of 4.0% before the light rail opens (2000-2025)
and 5.6% after it opens (2026-2040), relative to the Light Rail + Redevelopment scenario. The reason
that the health benefits realized from the increase in NMT are slightly less is due to a model function
that delays the realization of the full health benefits of changes in NMT by 5 years. In Tier 1, doubling
NMT per PTT increases total person miles of NMT by residents per day by an average of 2.2% before
the light rail opens (2000-2025) and by an average of 6.6% after the light rail opens (2026-2040),
relative to the Light Rail + Redevelopment scenario. This translates into an average increase in avoided
premature mortality due to NMT in Tier 1 of 2.3% before the light rail opens (2000-2025) and 5.2%
after it opens (2026-2040), relative to the Light Rail + Redevelopment scenario. Doubling NMT per
PTT increases person miles of NMT by residents by roughly the same percent in both Tiers, because the
Tiers have comparable ratios of public transit use to NMT. In the Light Rail + Redevelopment scenario,
traveling to and from public transit stops represents 4.9% and 3.8% of total nonmotorized person miles
in Tier 2 and Tier 1 on average, respectively.
255
-------
Behavior Mode Sensitivity
Behavior mode sensitivity exists when a change in assumptions changes the patterns of behavior
generated by the model. For example, if plausible alternative assumptions changed the behavior of a
model from smooth adjustment to oscillation or from s-shaped growth to overshoot and collapse, the
model would exhibit behavior mode sensitivity. We present three behavior mode sensitivity tests below,
covering inputs into the Transportation, Economy, and Water sectors.
Lengthening Modal Person Mile Reaction Times to Drivers
We used many elasticities from literature to model relationships in the transportation sector, especially
in regard to drivers of person miles of travel by mode. However, even though some of the sources of
these elasticities indicated whether they were "long-term" or "short-term" elasticities, none of them
made a more precise indication of what amount of time it took for a change in a given driver to produce
the amount of change in a given output variable that the relevant elasticity indicates. Therefore, we
adopted a rule of thumb for transportation-sector elasticities from literature whereby all elasticity-based
cause-and-effect relationships are modeled as either occurring over a five-year period ("long-term"
elasticities) or over a one-year period ("short-term" elasticities). However, in earlier versions of the
model, we assumed that transportation-sector "long-term" elasticities were associated with a fifteen-year
reaction time and "short-term" elasticities were associated with a two-year reaction time, as indicated by
Sinha and Labi (2007), which we deemed to be unrealistically long reaction times for the particular
relationships to which we were applying them. In this sensitivity test (hereafter called the Longer
Reaction Times test case), those previous, longer reaction times are reinstated. By running this test case,
we demonstrated how choosing shorter reaction times increased the model's sensitivity to shocks and
shortened the time needed to recover from those shocks. Because the D-O LRP SD Model cannot run
with reaction times of zero and some of the reaction times in the transportation sector are already only
one year, we did not run an alternate scenario of setting reaction times to less than B AU. The eight long-
term and three short-term elasticities adjusted in this test are listed in Table 6-23.
Table 6-23. Input Variable Changes Made to Test Model Sensitivity to Longer Reaction Times Associated
with Elasticities, Measured in Years
VARIABLE
through traffic reaction time to automobile speed
through traffic reaction time to fuel cost
vehicle trip distance reaction time to vehicle speed
mode share reaction time to automobile speed
mode share reaction time to fuel cost
mode share reaction time to parking cost
mode share reaction time to population density
mode share reaction time to intersection density
nonmotorized travel reaction time to jobs housing balance
public transit travel reaction time to fare price
public transit travel reaction time to vehicle revenue miles
BAU
5
5
5
5
5
5
5
5
1
1
1
LONGER REACTION TIMES
SCENARIO
15
15
15
15
15
15
15
15
2
2
2
Note: For this sensitivity test, input changes in Tier 1 and Tier 2 are identical
256
-------
The primary effect of the Longer Reaction Times test case is to make changes in the trends of output
variables occur more gradually. In the BAU scenario, during the period of approximately 2008-2015,
automobile driver person miles experience a lesser growth rate than during the rest of the model run,
while automobile passenger, public transit, and nonmotorized person miles have significant increases in
their growth rates during that period (Figure 6-64). This occurs due to a period of mostly high gasoline
prices (but still with some ups and downs during that period), with the price of a gallon of gasoline being
an exogenous input. Meanwhile, in the Longer Reaction Times case, the effect of gasoline prices on
transportation by mode is more muted in the short term but extends for a longer period of time.
In the BAU scenario, all modal person miles are calibrated to the most recent available historical data.
Unsurprisingly, reverting to the Longer Reaction Times scenario causes the model to no longer be
calibrated to these values, such that adopting those longer reaction times into the base scenario would
require altering the inputs for year-2000 person miles by each mode. After such a recalibration,
however, the fit of modal person miles to projections derived from the DCHC MPO's Metropolitan
Transportation Plan would be slightly improved relative to the BAU scenario. On the other hand, the fit
of public transit person miles to historical data from the National Transit Database for the years 2000-
2013 would be worsened by reverting to longer reaction times, even after recalibration.
As shown in Table 6-24, the Longer Reaction Times test case only creates moderate average yearly
deviations from BAU. The largest average yearly deviation from BAU in absolute terms is 4.2%, for
Tier 2 public transit person miles per day. It is not surprising that these deviations from BAU are not
large, since the only difference between the scenarios is how much time it takes for the same cause-and-
effect relationships to manifest. Because any period of time when a given output variable is more or less
than BAU in the Longer Reaction Times scenario is followed by a period when the opposite is true,
average yearly deviations from BAU tend towards zero, given enough time.
Even though average deviations of the Longer Reaction Times test case from BAU during 2000-2040
are small, larger deviations still occur in individual years within that span (Figure 6-64 and Table 6-24).
In 2015, mostly-high gasoline prices during the previous few years result in automobile driver person
miles being significantly lower in the BAU scenario than in the Longer Reaction Times scenario and in
person miles by all other modes being significantly greater in the BAU scenario(Table 6-24). Based on
the model's exogenous inputs, gasoline prices decline sharply in the year 2015, then resume growing at
a more gradual rate until 2040. During this period of increasing gasoline prices, the Longer Reaction
Times scenario again lags in responding to this change, so modal-person-mile values in the BAU
scenario and the test case eventually reconverge with one another. In the Longer Reaction Times test
case, even though gasoline prices drop in 2015, that event is immediately preceded by a period of
relatively high gas prices and is followed by a period of gradually rising gas prices. Because of the long
time it takes in this case for modal person miles to respond to changes in gas prices, the effects of the
2015 drop in gas prices are felt simultaneously with the effects of the higher gas prices that exist in the
years before and after 2015. Therefore, the growth rates of modal person miles during 2015-2020 in the
Longer Reaction Times test case are little different from those in the preceding and following years,
even though the growth rates of modal person miles in the BAU scenario experience a much more
noticeable change during that same period. This is because the growth rates of modal person miles in the
BAU scenario at any given point in time are reacting to the gasoline prices that existed during a narrow
portion of the recent past, such that the impacts of sudden changes in fuel prices are less diluted by
values in neighboring years. As a result of the mechanisms just described, the outputs of scenarios and
test cases that assume different reaction times may converge more quickly after a shock to the system
than the difference between their assumed reaction times would suggest (as indicated in Table 6-23, the
257
-------
reaction time of modal person miles to fuel prices is five years in the BAU scenario and fifteen years in
the Longer Reaction Times scenario). In addition, when the D-O LRP SD Model calls for the reaction of
a variable to a driver to not be immediate, we do not phase it in in a linear fashion. Instead, even though
the entire indicated reaction time is needed for the effect of a change in a driver to be fully felt, the
majority of the impact occurs in the first half of the indicated reaction period.
(A) Automobile Driver Person Miles - Tier 1
(B) Automobile Passenger Person Miles - Tier 1
•Longer Reaction Times
-BAU
-Longer Reaction Times
-BALI
(C) Public Transit Person Miles - Tier 1
(D) Nonmotorized Person Miles - Tier 1
^—Longer Reaction Times ^^BALI Longer Reaction Times ^^BAU
Figure 6-64. Tier 2 Person Miles Per Day by Mode: Longer Reaction Times vs. BAU
Table 6-24. Yearly Percent Departure from BAU: Sensitivity to Longer Reaction Times Associated with
Elasticities from Literature in Transportation Sector
VARIABLE
AVERAGE YEARLY
ABSOLUTE PERCENT
DEPARTURE
AVERAGE YEARLY
PERCENT ABOVE
OR BELOW
2015
2020
Tierl
Automobile driver person miles
Automobile passenger person miles
Public transit person miles
Nonmotorized person miles
Total person miles
Through traffic VMT
VMT
Congestion
Vehicle trip distance
GRP
Population
2.4%
2.6%
3.6%
2.7%
0.6%
3.4%
2.9%
2.9%
0.8%
1.5%
0.5%
-0.1%
-1.2%
-2.1%
-1.0%
-0.6%
+ 1.4%
+0.6%
+0.6%
+0.8%
-1.5%
-0.5%
+5.5%
-8.1%
-9.2%
-7.9%
-0.5%
+ 10.5%
+7.8%
+7.8%
+ 1.0%
-1.2%
-0.1%
-0.1%
-1.6%
-1.2%
-0.5%
-0.6%
+ 1.0%
+0.4%
+0.4%
-0.1%
-1.8%
-0.5%
Tier 2
Automobile driver person miles
Automobile passenger person miles
Public transit person miles
Nonmotorized person miles
Total person miles
Through traffic VMT
1.3%
3.5%
4.2%
3.2%
0.2%
3.3%
+0.1%
-0.9%
-2.1%
-1.0%
-0.2%
+ 1.6%
+3.4%
-9.7%
-10.8%
-9.4%
-0.4%
+ 10.7%
0.0%
-1.3%
-1.1%
-0.5%
-0.3%
+ 1.3%
258
-------
VARIABLE
VMT
Congestion
Vehicle trip distance
GRP
Population
AVERAGE YEARLY
ABSOLUTE PERCENT
DEPARTURE
2.0%
2.0%
1.1%
1.2%
0.1%
AVERAGE YEARLY
PERCENT ABOVE
OR BELOW
+0.7%
+0.7%
+ 1.1%
-1 .2%
-0.1%
2015
+6.1%
+6.1%
+ 1.4%
-1.3%
0.0%
2020
+0.5%
+0.5%
+0.3%
-1.8%
-0.1%
Earnings per Industrial Employee Tier 1 Reduced by 50%
Earnings per industrial employee Tier 1 is an exogenous input taken from historical data and projections
for Durham and Orange County, since no earnings data were available at a smaller geographic scale. But
due to the large amount of high wage, white-collar jobs classified as industrial in the outskirts of
Durham County, average earnings per industrial employee were very high and likely not representative
of the average earnings per industrial employee in Tier 1, where more blue-collar industrial jobs exist.
Thus, this scenario tests the sensitivity of the model to reducing earnings per industrial employee Tier 1
by 50%, which reduces the average earnings per industrial employee in Tier 1 in 2010 from $92,000 to
$46,000, during the entire model period (2000-2040).
(A)
(B)
-BAU
-BAU
2010 2020 2030 2040
Industrial Earnings per Employee Tier 1 Reduced 509c
•BAU Industrial Earnings per Employee Tier 1 Reduced 509c
-BAU
Figure 6-65. Sensitivity Analysis for Earnings Per Industrial Employee Tier 1 Reduced by 50% Between
2000 and 2040: Effect on (A) Industrial Earnings Tier 1 and (B) Relative GRP Tier 1
Although total industrial earnings Tier 1, shown in Figure 6-65 A, is considerably lower for this
sensitivity scenario than the BAU throughout the study period, relative GRP Tier 1, shown in Figure
6-65B, is higher than the BAU. This is due to the formulation used to estimate GRP and its main drivers,
where relative GRP (rather than absolute GRP) is used to influence other variables in the model that are
impacted by economic performance, including retail consumption, which has a built-in two year delay
before relative GRP begins to influence it. As a result, retail consumption, and thus employment,
earnings, and GRP are not affected initially by the reduction in industrial earnings per employee,
meaning that the GOS also does not change relative to the BAU simulation. Consequently, the economy
remains unaffected by the initial change for two years, but because the gap between the initial GRP in
2000 and the GRP in 2001 and 2002 is larger than in the BAU scenario (because of the reduction in
industrial earnings), relative GRP is larger in these years, which increases retail consumption,
employment, earnings, and again GRP above the BAU values. This leads to a snowball effect on
variables throughout the model, with examples shown in Table 6-25, creating an unrealistic situation
where a decrease in industrial earnings actually improves overall economic performance.
259
-------
This sensitivity test revealed a weakness in the model relationships that depend on relative changes in
other variables following delays, which is that any scenario that causes initial (2000) model values to be
reduced below levels in the BAU scenario will cause the relative values of that variable following the
delay to be higher than in the BAU scenario.
Table 6-25. Average Yearly Percent Departure from BAU, 2000-2040
TIER 1 VARIABLE
Industrial Earnings Tier 1
Total Retail Consumption Tier 1
GRPTierl
Population Tier 1
VMTTierl
AVERAGE YEARLY % DEPARTURE FROM
BAU
-39%
+24%
+17%
+9.0%
+5.1%
Stormwater N Load Sensitivity to Rainfall Variability
Although stormwater runoff outcomes presented the Water Sector of Section 5.4 assume constant
precipitation at the historical average level, realistic rainfall varies from year to year. This test explores
the impacts on stormwater N load from using randomly-generated precipitation that reflects patterns in
historical data. Data from 2000-2013 show that the Northern Piedmont region receives 45 inches of
precipitation per year with standard deviation of 7.0 inches per year (Figure 6-66A). Projecting this
forward using a random factor causes Tier 2 stormwater N load to vary by as much as 600,000 Ib N per
year from the BAU level, a 41% difference (Figure 6-66B). In comparison, BAU stormwater N load
grows by 350,000 Ib between 2000 and 2040. Therefore the projected increase in stormwater N load
due to regional development is within the range of variability due to realistic precipitation. If local data
on stormwater N load were collected for management and policy purposes, it would be difficult to detect
trends unless the data were adjusted for rainfall.
2000
2010
2020
2030
2040
2000
2010
2020
2030
2040
•BAU random_preci'p
• BAU
• NC State Climate Office
•BAU random_precip
•BAU
Figure 6-66. Sensitivity Analysis for Average Precipitation with Random Factor: Effect on (A)
Projected Precipitation and (B) Tier 2 Stormwater N Load
Note: the effect of the random precipitation factor on stormwater N load looks nearly identical in Tier 2 (b) and Tier 1 (not
shown), because rainfall volume, and therefore its variability, scales with area.
260
-------
Policy Sensitivity
Policy sensitivity exists when a change in assumptions reverses the impacts or desirability of a proposed
policy. We present four policy sensitivity tests below, covering inputs in the Land Use, Transportation,
and Energy sectors.
Sensitivity to the Effect of the LRT on Nonresidential Square Feet Demand
This scenario is run on top of the Light Rail scenario and aims to test the sensitivity of the model to the
effect of the light rail on the demand for nonresidential square feet in Tier 1, since this value is an
assumption, and is intended to be modified by users. The default value in the model is 10, indicating that
with the introduction of the light rail, demand for nonresidential commercial square footage will
increase by 10% over the Light Rail scenario. This change is phased in during the construction of the
light rail, allowing demand to begin in anticipation of its completion. Since there are so many
confounding variables affecting demand for commercial space in cities, no relevant literature could be
found to quantify the impact of light rail lines on the demand for nonresidential square footage in the
surrounding areas.
In this test, we report the sensitivity for a minimum value (0) for the percentage increase in demand for
nonresidential sq ft, a maximum value (100), and a moderate increase (20); we also display the results of
a Monte Carlo simulation. The simulation was set as a random-uniform distribution with minimum of 0
and maximum of 100, run 1,000 times. Immediate effects are seen on total nonresidential square footage
in Tier 1, which under the maximum value quickly hits a cap based on the maximum density and land
expansion allowable, around 2025 (Figure 6-67). In the Monte Carlo simulation figures, the 50%, 75%,
95%, and 100% color-coded bands refer to the probability of the projection landing within that range,
given the minimum and maximum values set, and the assumption that the likelihood of each value for
the tested input variable is the same within the specified range.
total nonresidential sq fl Tier 1
40 M
Figure 6-67. Nonresidential Sq. Ft. - Tier 1: Monte Carlo Simulation for Sensitivity to Effect of LRT
on Nonresidential Sq. Ft. Demand in Tier 1
261
-------
unemployment rate tier 1
0.08
2010 2020
Time (Year)
Figure 6-68. Unemployment Rate - Tier 1: Monte Carlo Simulation for Sensitivity to Effect of LRT
on Nonresidential Sq. Ft. Demand in Tier 1
affordability index Tier 1
1
0.925
0.85
0.775
2040
Figure 6-69. Affordability Index- Tier 1: Monte Carlo Simulation for Sensitivity to Effect of LRT on
Nonresidential Sq. Ft. Demand in Tier 1
net migration Tier 1
2,000
Figure 6-70. Net Migration - Tier 1: Monte Carlo Simulation for Sensitivity to Effect Of LRT
on Nonresidential Sq. Ft. Demand In Tier 1
The additional portion of demand in Tier 1 is also added to Tier 2, since Tier 2 is inclusive of Tier 1,
which is why effects are seen in Tier 2 as well. Nonresidential sq ft in Tier 1 diverges from the Light
Rail on an average yearly basis by -4.6% in the 0% increase case, +3% in the 20% case, and +10.4% in
the 100% case (Table 6-26). Unemployment is one of the most sensitive variables to this change. Since
employment is strongly connected to nonresidential sq ft in the model, it increases faster than population
in response, and the unemployment rate drops as the effect of the LRT on nonresidential sq ft increases,
by -9.3% and -32% in the 20% and 100% cases, respectively (Figure 6-68).
262
-------
The affordability index in Tier 1 also shows a sharp nonlinear reaction, with affordability spiking down
dramatically soon after the effect begins in 2022, particularly in the maximum case (Figure 6-69). This
is due almost entirely to a spike upwards in multifamily property values, driven primarily by the
connection with retail density, which increases as nonresidential sq ft increases, and then declines as
square footage is capped while population continues to grow. Therefore, despite the large swings as the
light rail is introduced, even in the maximum case (a 100% increase in demand, though developed land
is capped after a 67% increase in nonresidential sq ft), the average yearly percent deviation in the
affordability index from the Light Rail is only -1.7% in Tier 1. Net migration in Tier 1 also displays
nonlinearity. Under the maximum case, net migration spikes quickly as a result of the connection
between employment and immigration. The increase in desired employment at first outpaces
immigration due to the reinforcing feedback loop between the increase in nonresidential sq ft and
employment, but eventually immigration catches up, leading to the subsequent drop in the employment
gap and thus in net migration (Figure 6-70).
This test was also run on top of the Light Rail + Redevelopment scenario for comparison. It
demonstrates that if the effect of the light rail on demand for commercial building space is much larger
than assumed, the Light Rail + Redevelopment scenario becomes relatively more desirable. When the
test is run on top of the Light Rail scenario, a larger effect of the light rail on demand for commercial
development causes the local economy in general to improve more quickly with higher GRP and a lower
unemployment rate, while diminishing affordability and increasing greenhouse gas emissions.
Furthermore, the cap on developable land is hit much more quickly, indicating a much stronger case for
the kind of increased density allowed in the Light Rail + Redevelopment case. By contrast, under the
Light Rail + Redevelopment scenario, in the maximum case (100), GRP in Tier 1 averages 13.4% higher
than the Light Rail + Redevelopment baseline, versus only 8.9% under the Light Rail maximum case.
At the same time, affordability in Tier 1 is hurt less than under the light rail maximum case; 1.6% lower
than the Light Rail + Redevelopment baseline on average versus 1.7% lower on average than the Light
Rail baseline under the Light Rail maximum case (Table 6-26).
263
-------
Table 6-26. Average Yearly Percent Departure from BAD, 2000-2040: Sensitivity Effect of LRT on
Nonresidential Sq. Ft. Demand Tier 1
Variable
Min (0) v.
Light Rail
Med (20) v.
Light Rail
Max (100)
Light Rail +
Max (100) v. Redev v. Light
Light Rail Rail + Redev
Tier 1
Nonresidential sq ft Tier 1
Developed land Tier 1
Total employment Tier 1
Net migration Tier 1
Population Tier 1
GRP Tier 1
Unemployment rate Tier 1
Afford ability index Tier 1
C02 emissions from buildings and transportation Tier 1
Tier 2
Nonresidential sq ft
Developed land
Total employment
Net migration
Population
GRP
Unemployment rate
Afford ability index
C02 emissions from buildings and transportation
-4.6%
-2.8%
-2.9%
-16.3%
-2.0%
-3.4%
15.3%
0.9%
-3.6%
-1.4%
-0.4%
-1.1%
-3.0%
-0.3%
-1.3%
35.3%
-0.6%
-0.8%
3.0%
1.2%
2.3%
11.0%
1.6%
2.5%
-9.3%
-0.5%
2.4%
1.3%
0.4%
0.7%
1.9%
0.2%
0.9%
-13.1%
0.3%
0.7%
10.4%
3.2%
8.1%
31.6%
4.4%
8.9%
-31.9%
-1.7%
8.1%
7.7%
1.6%
1.5%
4.9%
0.6%
4.1%
-21.6%
-0.4%
4.1%
16.1%
7.0%
11.9%
33.7%
4.9%
13.4%
-60.7%
-1.6%
12.3%
8.9%
1.8%
1.6%
6.1%
0.7%
4.7%
-22.8%
-0.5%
4.7%
Sensitivity to Percent of Net Migration Due to Light Rail in Tier 1 That is from Areas External to Tier 2
This test is run on top of the Light Rail + Redevelopment scenario and tests the sensitivity of the model
to changes in the percent of net migration due to the light rail in Tier 1 that is external to Tier 2, an
assumed ratio for which there were no historical data available. This variable makes an assumption
about how many of the people who move to the station areas are coming from outside the study region,
and adds that portion to the Tier 2 population, since Tier 2 is inclusive of Tier 1. In the model, the
default value is 1.5, indicating that 100% of those who move to the station areas as a result of the light
rail are from outside Tier 2 (and therefore must be added to the Tier 2 population as well), and in
addition, 50% more people will move into Tier 2, likely to areas near, but not within, the /^ mile station
area radii. This value was chosen because it made the unemployment rate more reasonable in Tier 2; as
jobs increase dramatically, population must also to some degree fill those jobs. In contrast to Tier 1, Tier
2 does not have a link between jobs and migration. Instead, the primary relationship in Tier 2 that
governs migration is a link with vacant residential land, but since available land is plentiful, it doesn't
increase population under the Light Rail scenarios to keep the unemployment rate above zero.
264
-------
In this test, we report the sensitivity for a minimum (0) and a maximum value (3) as well as display the
results of a Monte Carlo simulation. The simulation was set as a random-uniform distribution with
minimum of 0 and maximum of 3, run for 1000 times. The immediate effects are seen on net migration
(Figure 6-71) and population. Net migration varies from 6,920 to 10,300 people in 2040, with a percent
deviance from the BAU of between -20% and +20% in the maximum case. Because population varies
considerably more than total employment (see Table 6-27), the unemployment rate is significantly
affected, at 11% below the BAU on average in the minimum case and 21% above in the maximum case.
Figure 6-71. Net Migration - Tier 2: Monte Carlo Simulation for Sensitivity to Percent of Net
Migration Due to Light Rail in Tier 1 That Is External to Tier 2
Figure 6-72. Unemployment Rate - Tier 2: Monte Carlo Simulation for Sensitivity to Percent of
Net Migration Due to Light Rail in Tier 1 That Is External to Tier 2
Changes to these variables propagate throughout the model, eventually affecting the percent of the
population in poverty. Smaller effects are seen in developed land, total employment, VMT, and GRP.
The particularly large impact on unemployment means a low value for the percent of net migration due
to the light rail in Tier 1 that is from areas external to Tier 2 makes the light rail scenarios less attractive
from the perspective of the economic prospects of the residents. On the other hand, GRP, and therefore
overall economic performance, is not very sensitive to changes in this variable.
265
-------
Table 6-27. Average Yearly Percent Departure from BAD, 2000-2040: Sensitivity to Percent of Net
Migration Due to Light Rail in Tier 1 that is External to Tier 2
Min (0) v Max (3) v
Variable
Tier 2
Net migration
Population
Developed land
GRP
Total employment
Unemployment rate
Percent of population in poverty
VMT
-6.8%
-0.656
-0.5%
-0.2%
-0.3%
-10.B::
-2. IS
-0.3%
6.7$
Q.&
0.4::
0.1%
0.1S
21.4':
5.6::
0.3i:i
Congestion Per Weekday Peak-Period VMT per Lane Mile Held Constant
For both Tier 1 and Tier 2, the D-O LRP SD Model features an exogenous input variable called
"congestion per weekday peak period VMT per lane mile" that determines how much traffic congestion
(defined as the ratio of peak-period travel time over travel time in freeflow conditions) results from a
given density of vehicles on the roadway system. Changes in its value may be due to changes in the
effectiveness of traffic-control measures in the study area, but may also be due to traffic becoming more
or less concentrated, both spatially across the roadway system and temporally throughout the peak travel
periods. The variable consists of a lookup table, with values derived from the TRM v5 projections that
were used in the DCHC MPO Metropolitan Transportation Plan. Those projections provide figures for
the years 2010 and 2040, with no indication of what trend is followed between those years. Therefore,
the D-O LRP SD Model assumes that congestion per weekday peak-period VMT per lane mile holds at
a constant value from 2000 to 2010, then decreases linearly from 2010 to 2040, in both Tier 1 and Tier
2. For this sensitivity test, we modified the model's inputs so that the constant 2000-2010 value
continues to hold constant until 2040. In Tier 1, it remains constant at 0.000388, instead of ramping
down to 0.000296 during 2010-2040 (a 23.8% reduction). In Tier 2, it remains constant at 0.000483,
instead of ramping down to 0.000347 during 2010-2040 (a 28.1% reduction). In order to analyze the
policy implications of this sensitivity test, we examined the effects of the change in inputs on the BAU
scenario, the Light Rail scenario, and the No Road Building scenario, wherein all new road construction
stops after 2017, subsequently reducing the capacity of the roadway system relative to the BAU and
Light Rail cases. The test cases created on the basis of these scenarios are called BAU + Cong per VMT,
Light Rail + Cong per VMT, and No Road Building + Cong per VMT, respectively.
Not surprisingly, in the cases where congestion per weekday peak-period VMT per lane mile does not
decline after 2010, traffic congestion is much higher during 2010-2040 than it would be with the
model's default inputs (Figure 6-73). In each test case, this difference is approximately the same
percentage that each test case increases congestion relative to the scenario that was modified to create it.
Consequently, this sensitivity test does not change the proportion by which building roads reduces
traffic congestion or the proportion by which adding a light rail line increases congestion, mostly due to
the increase in Tier 1 land development that it is assumed to cause. Instead, the constant rate of
congestion per weekday peak-period VMT per lane mile assumed in this test increases the baseline
266
-------
congestion that those policy decisions modify. If baseline traffic congestion is more severe, it makes
mitigating it through road building a more attractive option and makes building a light rail line that
increases congestion a less attractive option than it otherwise would be. However, the fact that the D-O
LRP SD Model predicts that the building of a light rail line will increase traffic congestion is mostly due
to the assumption that the presence of a light rail line will increase demand for Tier 1 commercial floor
space by 10%, as well as increase the proportion of Tier 1 workers who choose to also live in Tier 1. In
the absence of those assumptions, opening a light rail line would be expected to decrease traffic
congestion. Furthermore, even if those assumptions are left in the model, creating a light rail line would
still create economic benefits, in addition to the transportation-related effects discussed here.
The downstream effects of this sensitivity test are mostly unsurprising. Since traffic congestion is more
severe in the test cases, there is also less automobile driving and more travel by public transit and
nonmotorized modes (Table 6-28), indicating a greater amount of latent demand for public transit
improvements, such as the addition of a light rail line. Also unsurprisingly, increased traffic congestion
has a small, negative effect on GRP. From 2025 to 2040, the negative effect on Tier 2 GRP of holding
congestion per weekday peak-period VMT per lane mile constant is less than the Tier 2 GRP benefit of
creating the light rail line, and has less than half the magnitude of that benefit during 2032-2040. In Tier
1 (but not Tier 2), the small negative effect on GRP creates a smaller negative effect on population, by
way of the unemployment rate and net migration. Table 6-28 also shows that, whereas the increased
traffic congestion in this test causes Tier 1 automobile passenger miles per day to slightly decrease, it
also causes Tier 2 automobile passenger miles per day to increase. This is due to fact that, in the D-O
LRP SD Model, the effects of most of the drivers of modal person miles are normalized, so that overall
person miles track a baseline trend that is driven by population and GRP. As a result, when person miles
by one mode go up or down, an opposite-direction change must occur in person miles by one or more of
the other modes. Without normalization, traffic congestion would reduce automobile passenger person
miles in both Tiers, due to fewer people being willing to travel in congested conditions. However, some
people are assumed to respond to traffic congestion by carpooling, so that the effect of traffic congestion
on automobile passenger person miles is less than the effect on automobile driver person miles. Since
congestion has a stronger effect on automobile driver travel than on any other mode and automobile
driver travel accounts for the majority of person miles in both Tiers and in all scenarios and test cases,
the normalization step requires that there be a net increase in person miles by all modes other than
automobile driver travel. Due to how the normalization step is carried out, how much of this net increase
comes from each of the remaining three modes is determined by how large of a proportion of overall
person miles each of them represented to begin with. In Tier 1, public transit and nonmotorized modes
account for a large enough fraction of overall person miles that the reduction in automobile driver
person miles merely results in the reduction in automobile passenger person miles being partially
mitigated. In Tier 2, though, public transit and nonmotorized modes account for a significantly smaller
fraction of overall person miles, meaning that, after normalization, a larger percentage of the reduction
in automobile driver person miles must be made up for by increased automobile passenger person miles.
In effect, this means that travelers in the portion of Tier 2 that is outside of Tier 1 are more likely to start
carpooling in response to greater traffic congestion than are those whose trips either begin or end in Tier
1.
Holding congestion per weekday peak-period VMT per lane mile constant has about the same effect on
model outputs in the BAU, Light Rail, and No Road Building scenarios (Table 6-28). The largest
exception to this is that it does not increase public transit person miles by as large of a proportion in the
Light Rail + Cong per VMT test case (relative to the Light Rail scenario) as in either of the other test
cases relative to their respective baselines. Since people walk or bicycle to and from transit stops, Tier 1
267
-------
nonmotorized person miles are also increased less in the Light Rail + Cong per VMT scenario than in
the other test cases. This happens because the increase in public transit person miles that is assumed to
occur when the light rail line opens under the Light Rail scenario does not account for changes in
congestion, as we define it in the model. Rather than a function that expressly mentions traffic
congestion, light-rail-induced person miles are driven by Tier 1 population and employment and Tier 2
VMT per highway lane mile, in accordance with an equation from literature (Chatman et al. 2014). In
this equation, the input "VMT per highway lane mile" is both an indicator of travel demand in the study
area and a stand-in for traffic congestion, as decided upon by the originators of the equation. However,
this way of representing traffic congestion does not account for changes (or the lack thereof) in the
amount of traffic congestion that results from a given density of vehicles on the road. Therefore,
adjusting congestion per weekday peak-period VMT per lane mile in scenarios where the light rail line
is built may lead to unrealistic results.
(A) Tier 1
(B) Tier 2
2000 2010 2020 2030
• No Road Building + Cong per VMT No Road Building
•Light Rail + Cong per VMT Light Rail
• BALI + Cong per VMT — BALI
2000 2010 2020 2030
• No Road Building + Cong per VMT — No Road Building
• Light Rail + Cong per VMT Light Rail
• BALI + Cong per VMT — BALI
Figure 6-73. Traffic Congestion: Congestion Per Weekday Peak-Period VMT Per Lane Mile Held
Constant vs. No Road Building, Light Rail, and BAD Scenarios
268
-------
Table 6-28. Average Yearly Percent Departure from BAD, Light Rail, and No Road Building Scenarios,
2010-2040: Congestion Per Weekday Peak-Period VMT Per Lane Mile Held Constant
VARIABLE
LIGHT RAIL + NO ROAD BUILDING +
BAU + CONG PER CONG PER VMT CONG PER VMT V. NO
VMT V. BAU V. LIGHT RAIL ROAD BUILDING
Tierl
Congestion
VMT
Automobile driver person miles
Automobile passenger person miles
Public transit person miles
Nonmotorized person miles
Total person miles
GRP
Population
+ 11.7%
-2.1%
-1.7%
-0.1%
+3.6%
+2.7%
-0.8%
-0.9%
-0.3%
+ 11.6%
-2.2%
-1.9%
-0.2%
+0.8%
+2.3%
-0.9%
-0.7%
-0.4%
+ 11.6%
-2.2%
-1.9%
-0.1%
+3.5%
+2.7%
-0.9%
-1.0%
-0.3%
Tier 2
Congestion
VMT
Automobile driver person miles
Automobile passenger person miles
Public transit person miles
Nonmotorized person miles
Total person miles
GRP
Population
+ 15.1%
-1.9%
-1 .4%
+0.5%
+4.7%
+3.7%
-0.8%
-1.0%
0.0%
+ 15.0%
-1.9%
-1.5%
+0.5%
+3.4%
+3.6%
-0.9%
-1.0%
-0.1%
+ 15.0%
-2.0%
-1 .6%
+0.5%
+4.7%
+3.6%
-0.9%
-1.0%
0.0%
Note: Difference between scenarios with and without constant congestion per weekday peak-period VMT per lane mile starts
after 2010.
Sensitivity of Cross-Sector Indicators to Changes in the Energy Sector
In this test, we asked which energy variables had the largest proportional effect on CO2 emissions, given
the Durham County goal of reducing GHG emissions 30% from a 2005 baseline by 2030. We also
examined "side effects" on variables representing travel behavior, economic performance, and
impervious area. Energy variables (MPG, building energy intensity, LRT effect on person miles of
public transit, desired solar capacity, and electricity emissions factor) were increased or decreased by
10% or more relative to the Light Rail + Redevelopment scenario (Table 6-29). Note that these
sensitivity tests apply a constant % change to energy system variables between 2000 and 2040. This is in
contrast to the policy tests described in the Energy Sector of Section 5.4, which often apply a change
which starts in 2015 and gradually increases to 2040.
CO2 emissions were most sensitive to changes to building energy intensity (MMBtu/year per sq ft or per
dwelling unit) and electricity emissions factor (tons CO2 per kWh delivered), with a 10% decrease in
building energy intensity reducing Tier 2 emissions by 7.2% by 2040, and a 10% decrease in electricity
emissions factor reducing Tier 2 emissions by 6.4%. The relationship between CO2 emissions and both
building energy intensity and electricity emissions factors is not 1:1 because buildings represent 73% of
Tier 2 CO2 emissions, and electricity represents about 63% of Tier 2 CO2 emissions in the model. CO2
emissions were next most sensitive to automobile fuel efficiency, with a 10% increase in MPG causing a
2.5% decrease in Tier 2 emissions.
GRP was most sensitive to changes in MPG and building energy intensity. Counterintuitively, a 10%
increase in MPG causes a 0.51% decrease in GRP. Although this increase in MPG reduces total energy
269
-------
spending/GRP, the reduction is 5% in 2000 but 3.8% in 2040, so that by 2040 total energy
spending/GRP is higher relative to its updated initial (year 2000) value, which causes the slight decline
in GRP. A 10% decrease in building energy intensity causes a 0.18% increase in GRP due to declining
energy cost as a fraction of GRP.
Increasing the effect of Light Rail on public transit person miles by 200% (representing an increase in
rail ridership) has a slight effect on cross-sector indicators. Both VMT and congestion decrease by
0.71% in response, while GRP increases by 0.28% and impervious surface increases by 0.12%. VMT
decreases due to mode shifting from automobiles to public transit, and congestion decreases as a result
of VMT decreasing. GRP increases due to decreased energy spending, and impervious surface increases
due to land development stimulated by increased GRP.
In summary, this set of energy system sensitivity tests demonstrates that decreases in building energy
intensity and electricity emissions factor will have the largest proportional effect towards reducing GHG
emissions, followed by increases in automobile MPG. Further, cross-sector side effects of changing
these energy variables by 10% were negligible; GRP and impervious surface changed mostly by less
than 0.2% in response (with the exception of GRP decreasing by 0.5% in response to a 10% increase in
MPG). Although increasing the LRT ridership effect on public transit reduced CO2 emissions, the effect
was small; tripling the LRT effect reduced Tier 2 CO2 emissions by less than 0.1%. Our model suggests
that decreased building energy intensity, decreased electricity emissions factor, and increased MPG
would make the largest proportional impact towards achieving a GHG reduction target, with minimal
cross-sector side effects.
Table 6-29. Sensitivity Analysis for Energy System Variables: Effect on Cross-Sector Indicators in Tier 2
in 2040
Energy system variable Change
MPG +10%
'-10%
Building energy intensity +10%
r-io%
LRT effect on person miles +10%
of public transit r-10%
+200%
Solar capacity +10%
+100%
Electricity emissions factor +10%
-10%
Cross-system indicator affected
VMT Congestion GRP C02 emissions impervious surface
-0.067% -0.068% | -0.509%
0.062%
-0.007%
0.013%
0.062%
-0.007%
0.013%
-0.035% -0.035%
0.035% 0.034%
-0.706%
| -0.707%
0.431%
• -2.467%
2.992%
-0.092% Ht30%
| 0.184% -7.246%
0.014% -0.003%
-0.014%
0.280%
0.003%
-0.066%|
0.000% 0.000% 0.000% -0.024%
0.000%
0.000%
0.000%
0.000%
0.000%
0.000%
0.000%
0.000%
0.000%
-0.240%
-0.134%
0.149%
-0.004%
0.027%
0.006%
-0.006%
0.116%
0.000%
0.000%
6.388% 0.000%
-6.388%
0.000%
Note: orange indicates a decrease, and blue an increase, relative to the Light Rail + Redevelopment scenario.
6.4 Computational Reproducibility
Making model computations reproducible is critical to help researchers examine and use the results of
simulation exercises. Further, it greatly helps in carrying out follow-up studies, saving time in
identifying and interpreting the parameters used in the initial model and in subsequent scenarios. In
general, computational reproducibility can be helpful in the following ways in the context of the D-O
LRP project:
270
-------
• It will help the technical team that has developed the D-O LRP model to reproduce scenario
results and check underlying assumptions incorporated in the model. This will be particularly
important as the model is used over time.
• It will also support other stakeholders who wish to use the D-O LRP model to simulate
alternative scenarios or apply this type of model to other locations by allowing them to fully
understand the data and relationships incorporated in it through documented code; and thereby
avoid spending substantial periods of time trying to figure out how model results are produced.
In order to ensure computational reproducibility, the modeling team provided EPA with the full source
code of the D-O LRP model and the Agency can make it available to qualified users. Further, model
users can obtain full documentation of any scenarios they simulate, the parameters used to run them, and
a summary of results from Vensim. Figure 6-74 provides an example of the Vensim documentation for
the Light Rail + Redevelopment scenario, in comparison to the BAU scenario. "Runs compare" is a
button at the left of the Vensim screen (see Appendix A: User Guide) which generates a list of
differences between two loaded scenarios. The list may be very simple, as in Figure 6-74, or detailed if
the differences are extensive.
Comparing Light Rail + Redevelopment and BAU
******Constant differences between Light Rail + Redevelopment and BAU*****
Main policy switch - has changed in value
2 Light Rail + Redevelopment
0 BAU
Figure 6-74. Sample "Runs Compare" Report
271
-------
7 Summary & Conclusions
The premise of building the D-O LRP SD Model was to demonstrate, by means of a particular but
generalizable case, that integrated approaches yield a broader and more contextual basis for community
decisions, and that system dynamics models are a useful tool for achieving integrated decision-making.
In assessing the value added of the model, we look at both the model results and the modeling process,
as evaluated against the goals of (1) better integrated and more contextual assessment processes, (2)
more integrated and efficient cross-sectoral planning, and (3) enhanced community comprehension of
and participation in complex decisions. Here, we discuss features of the model both as they perform in
the present context of the D-O LRP, but also how they would likely perform across the spectrum of
envisioned usages.
7.1 Integrated Assessments
Key features of the model that enhance its ability both to assess the success of the D-O LRP project and
to evaluate its impacts include the ability to: 1) alter and test input assumptions, 2) account for
feedbacks of the project on those inputs, 3) explore the effects of nonlinearities of project impacts over
time and 4) characterize the nature, magnitude, and interdependences of indirect and cumulative
impacts. Regional projections of rapid population and employment growth over the next 25 years were
important drivers for the local land use and transportation models used to evaluate the D-O LRP as a
future transportation alternative. In the TRM v5, for example, the amount, type, and location of
population and employment in 2040, along with additional assumptions, drove the total demand for
transportation. However, the population projections used as inputs to the TRM, once selected, were
static and not subject to endogenous feedbacks, as they were in our model. Furthermore, in our model,
confidence in the causal mechanisms underlying the projections was increased by starting the model
from 2000 and calibrating the model's outputs to historical data. Achieving close calibrations for drivers
such as population using underlying mechanisms rather than pure data fitting increased our confidence
in the model's ability to forecast the future.
Representing processes mechanistically not only allows the user to directly test assumptions about
model inputs, but also permits the tracing of causal pathways by which they deliver their impacts.
Endogenous population growth, based on birth, death and migration rates, allows the population
assumptions to be tested, but also, perhaps more importantly, allows the population growth to respond to
internal processes in other sectors of the model. For instance, premature mortalities avoided due to the
positive health benefits of increased physical activity under the Light Rail scenarios feed back into the
death rate, leading to fewer deaths per year than the BAU scenario. Of greater consequence, however,
are the model feedbacks that impact migration rates, namely employment, housing, and the availability
of residential land, which can alter population to a much greater extent.
Explicitly representing the endogenous variables that influence the drivers not only lets us alter
projections and test the model's sensitivity to the altered inputs, but also allows us to separate processes
that may be lumped together in a simpler model. This can be helpful in distinguishing the relative
contributions of otherwise coupled processes as well as isolating leverage points in the model where
policy interventions might alter the initial projected inputs. For example, while the Light Rail +
Redevelopment scenario reflects what is expected to occur (in the real world) in Tier 1 should the light
rail be built, there is debate as to whether the economic and transportation benefits could be realized
with redevelopment to higher densities without the light rail. The model affords us the ability to examine
272
-------
the impacts of the light rail and redevelopment separately as well as superimpose redevelopment on top
of the light rail. As a result, individually, both redevelopment and the light rail increase population and
employment above BAU in Tier 1, though to varying degrees. Redevelopment by itself increases
population and employment in Tier 1 slightly more than does BAU, but to a much lesser extent than the
combination of redevelopment and the light rail. The attractiveness of the light rail is expected to bring
more nonresidential development and jobs to the area, while at the same time decreasing VMT per
capita and increasing non-motorized travel per capita relative to either BAU or redevelopment alone.
Similarly, being able to isolate causal pathways associated with drivers or sectors allows us to test the
effectiveness of targeted policies. For example, impacts of wage increases or non-residential rents
propagate through the model from their point of intervention within the economy sector, as do the
impacts of constructing non-motorized infrastructure from its transportation origins.
The temporal dynamics of the SD model are well-suited to examine changes or interventions that occur
at some point in between the model start date and end date, or that change in value throughout the model
period. The model could be used to assess the implications of the timing of light rail or road
construction, especially for outcomes that have time lags, synergies, and/or feedbacks, such as
congestion. In a similar vein, technological changes may be viewed as continuous changes of a quantity
(e.g. increasing building efficiencies or vehicular fuel efficiencies) or discontinuous shifts (e.g. shift to
solar power), where timing and rate of change can significantly alter outcomes. For as much as total
VMT increased in all scenarios, a projected simultaneous increase in vehicle fuel efficiency meant
energy use and CO2 emissions by passenger vehicles stayed fairly level, and PIVh.s and NOX emissions
from passenger vehicles decreased due to projected reductions in PM2.5 and NOX vehicle emissions per
VMT.
The ability to identify and quantify indirect and cumulative benefits, singly or in combination, is an
important feature of the model that could substantially enhance a comprehensive, integrated assessment
of projects such as the D-O LRP. While indirect and cumulative impacts, in the context of EIAs, are
usually thought of as negative, we have also demonstrated positive impacts that may lead to increased
community support for a project, either as independently supported objectives or by affecting the
perceived net cost-benefit of the project itself. The D-O LRP DEIS identifies a number of indirect and
cumulative consequences. Some, such as visual and aesthetic impacts and impacts on historic resources,
are predominantly site-specific in nature, and are not addressed by models such as ours. Others, such as
consumption of water resources and production of stormwater, can be viewed in the model from the
perspectives of timing and magnitude, but also of causation. Causal processes, in turn, can be viewed
from a risk-benefit or cost-benefit perspective either directly from model dynamics, or by expressing
outcomes as intensity measures, where the denominator of intensity measures can be selected to
highlight the nature of trade-offs to be considered. So, for example, in Tier 1, although both light rail
and redevelopment increase total stormwater N loadings, they decrease stormwater N loadings per capita
by increasing population density. Stormwater per unit of GRP could also be a useful intensity measure
that highlights possible tradeoffs or suggests a potential mitigation strategy.
7.2 Integrated Planning/Coordinated Agency Decision-Making
A project of the magnitude of the D-O LRP has a scope that extends beyond the immediate bounds of
the construction of the rail itself. Its dependence on long-term demographics and economics for its
success, and its ramifications in many spheres of community life by necessity invoke the participation
and the coordination of numerous agencies. That scope presents a challenge to modeling. The adage to
"model the problem, not the system" suggests that a model should be no more complex than it needs to
273
-------
be for the issue at hand. However, when a model such as the D-O LRP SD Model has a range of
potential uses, and the D-O LRP itself has such a range of consequences, parsimony needs to be
weighed against utility.
In the D-O LRP SD Model, we've strived to bound and structure the model to reflect not only the key
causal and consequential dynamics, but also to capture key points of intervention. These points of
intervention occur within sectors that are themselves the domains of agencies. In some cases, such as for
the transit and planning agencies, their objectives and actions are almost inextricable, and their
collaboration is a given from the inception of the project. Those interactions are captured in our three
main scenarios, which reflect assumptions about drivers and magnitudes of interdependences more than
they establish novel interconnections among agencies. The expanse of the model to capture such issues
as affordability, health, and resource intensity brings in agencies who had not seen themselves as central
to the issue, yet whose missions could be very much impacted by it - either negatively or positively. At
first glance, a health agency might not see a role for itself in a transportation issue. Yet the enhancement
of nonmotorized travel that results from the transportation and land planning decisions may improve
health outcomes on a level commensurate with other, more classically "in-house" health actions.
Conversely, the priority of constructing additional nonmotorized travel infrastructure might fail to meet
funding thresholds until the health benefits are included in the cost-benefit analysis and advocated by the
health agencies. In similar fashion, the model equips the housing agencies with critical information
about property values, affordability, employment and economically-driven (i.e. private sector)
residential development, allowing them to elevate such issues as equity and affordability from vague
concerns to projectable trends and estimates of magnitudes, subject to testable policies. In practice, we
have seen the process of building the D-O LRP SD Model and testing of scenarios result in enhanced
participation of housing, health, water, and sustainability agencies at the municipal and county level.
7.3 Support for Community Participation in Complex Decisions
Issues that, by either their intrinsic nature or the nature of the decision process invoked, require a high
degree of public participation could also benefit from a tool like the D-O LRP SD Model. In our case,
largely as a function of us beginning the modeling project during the late stages of the D-O LRP
planning process, the development of the conceptual model, the vetting of the operational model, and
the design of the scenarios to be tested all occurred with the input of stakeholder agencies, rather than
the general public. Nevertheless, the construction of the conceptual model lends itself to public
participation. Issues, concerns, and beliefs or understandings about connections can all be captured in
model diagrams, for discussion in a public forum. They can then later be tested for importance or
validity when developing the operational model, with feedback to the community group. Once the
operational model has been constructed to incorporate community perspectives, assumptions and
scenarios can be tested, either off-line or in real time. The SD model runs very fast, permitting "on-the-
fly" model runs in settings such as stakeholder meetings or public fora. Having the public's questions or
challenges addressed and analyzed in an open forum is helpful in demonstrating transparency,
maintaining focus on the most influential issues, and developing buy-in to the decision process.
These features of the model could be even more useful if employed earlier in the design and deliberative
process than occurred with the D-O LRP SD Model. In our case, the D-O LRP had already been
designed and subject to a referendum for funding before the development of the SD model, with the
public gradually becoming aware of the full suite of potential impacts over the course of several years.
Community concerns and questions that have emerged in the planning and DEIS processes include:
274
-------
1) the project planners' assumptions for population growth, ridership, and costs;
2) the impact of the project on affordability of housing, especially for the transit-dependent;
3) the likelihood of technological innovations diminishing the need/demand for fixed-guide way
transit;
4) the resource carrying capacity of the region; and
5) the potential to generate public revenues through indirect impacts on jobs and property.
All of these concerns can be explored with the SD model, and while no model can pretend to head off all
community objections to a project, a fuller representation of the project from the beginning would
introduce transparency earlier in the process and possibly lessen the entrenched objections that arise in
any project of this magnitude.
Admittedly, the construction of a model such as the D-O LRP SD Model, at least with the tools available
at this time, does require considerable time and data. The effort may not be justified for all issues
(although we hope, as outlined in next steps, to lower the threshold of effort required); it is more likely
to be undertaken for larger projects with longer time frames. Given that it is just such projects that are
most likely to invoke the requirement for an EIS, to involve multiple agencies, and to engender public
concerns, we see great potential for such models to inform and enhance such projects from inception to
execution. Furthermore, once built, such models can be used for years to come to evaluate policies or
actions that, in and of themselves, would not have warranted the investment in a model.
7.4 Next Steps
The D-O LRP SD Model has been developed for and applied to the analysis of the Durham-Orange
Light Rail Project. In its current (beta) form, with little packaging between the user and the model
inputs, outputs and equations, a technically adept user can run alternate scenarios, test assumptions, and
even modify data and/or equations. Less technically-oriented stakeholders can likewise interact with the
model with the help of a capable intermediary.
After thorough review of the "naked" model, we do intend to develop an interface that supports easier
access to the model as well as transferability to other applications. We have solicited stakeholder inputs
on features that they would find desirable in a user interface, and engaged EPA's Environmental Model
Visualization Lab (EMVL) in a two-stage development process. The first stage of the effort will be to
develop a model interface that allows users to: 1) construct scenarios that reflect policy-driven changes
or other "what-ifs" (e.g. population or economic trends) as inputs, 2) visualize and contrast scenarios
with regards to multiple outputs, 3) perform sensitivity analyses on model inputs, parameters and
assumptions, 4) visually explore the dynamics of causality within the model, and 5) archive scenarios
with associated input and output values.
The second stage of the effort will to construct a "model builder" that will streamline and simplify the
construction of this type of model for other issues, locations, or users. Many communities face similar
decisions that require integration of multiple sectors, just as they also strive to maximize net benefits
across social, economic, and environmental outcomes. Community-level data to populate and calibrate
models are often extracted from larger scale data sets (e.g. census, economic, land cover) or are
developed and held by communities in digital forms that are increasingly similar across communities
275
-------
(e.g. property inventories). Features of such a model builder would include: tools for structured
elicitation (from users and stakeholders) of issues and linkages for conceptual model development; web-
based linkages to both general and specific data and models; tools for dynamic model development; and
a wrapper that combines user-interface, analytic, and visualization tools with the above.
Final Word
Ultimately, the strength of the D-O LRP SD Model is that it is a defensible, testable narrative - a
narrative that speaks to a number of audiences. It allows proposed projects, policies, or aggregations of
the two to be evaluated in a fully contextual mode, as required for determining net costs and benefits and
sustainability. It demonstrates the connectivity and synergies of actions spread out across multiple
entities, thus allowing them to develop collaborations that might not have previously existed, and to
focus their efforts where net gains may have been previously undetected because they were diffuse
across endpoints. And, finally, by placing the multiple concerns of agencies and stakeholders in a
common frame of reference, and subjecting the connections and assumptions to rigorous testing,
decisions can be made more inclusive, comprehensive, and transparent. Making synergies and conflicts
more visible, and testable, early in the process ultimately benefits all and fulfills the mandate of NEPA
"... to use all practical means and measures ... to create and maintain conditions under which man and
nature can exist in productive harmony and fulfill the social, economic, and other requirements of
present and future generations of Americans (NEPA, Sec 101(a))."
276
-------
Appendices
277
-------
Appendix A: Advanced User Guide for the
Durham-Orange Light Rail Project System
Dynamics Model
278
-------
TABLE OF CONTENTS - ADVANCED USER GUIDE
INTRODUCTION TO THE D-0 LRP SYSTEM DYNAMICS MODEL 280
INSTALLATION GUIDE 280
Installing Vensim Software 280
VENSIM FEATURES 2sr
Navigating Through the Model 281
Variables and Relationships 283
Tier 1 and Tier 2 of the Model 283
Shadow Variables and Causes/Uses Trees 284
Equations and Constants 284
Running a Simulation 289
Viewing Results 289
Graphs and Tables 289
Comparing Multiple Simulations 297
Loops Tool 293
Causes Strip Tool 293
D-0 LRP MODEL VIEWS 294
Overview 294
Interventions and Indicators 294
Dashboard View 295
279
-------
INTRODUCTION TO THE DURHAM-ORANGE LIGHT RAIL PROJECT SD MODEL
The Durham-Orange Light Rail Project System Dynamics (D-O LRP SD) Model is a collaborative effort
led by EPA's Sustainable and Healthy Communities (SHC) research program. The project uses system
dynamics to explore the impact of light rail transit and associated development in the rapidly growing
Research Triangle region of North Carolina. The model boundary is the Durham-Chapel Hill-Carrboro
Metropolitan Planning Organization (DCHC MPO) area. Variables in the model are simulated annually
between 2000 and 2040, with a time step of 0.0625 years, allowing projections to produce relatively
smooth curves. The model is designed to explore dynamic interactions among seven sectors of the urban
system, including Land Use, Transportation, Energy, Economy, Equity, Water, and Health. These sectors
are visualized in Figure A-l, with plus (+) signs indicating positive association, and minus (-) signs
indicating negative association between two variables. Model scenarios run in a few seconds, and users
can edit any variable in the model. This allows users to experiment with and test different aspects of the
model's representation of a complex urban system.
Redevelopment
Light Rail
Developed Land
* A
se / f \ .
Dwelling Units
tonnwate
ollutant Load!
Transit
Riders hip
Nonresidntial
jNonnwtonzed
Travel
Transportation
Person Miles
Air Emissions
Energy
Ene rgySpe nding
Transit-Dependent
Population
Population/in'
Poverty
FIGURE A-1. SIMPLIFIED CAUSAL LOOP DIAGRAM (CLD) FOR THE D-O LRP SD MODEL
INSTALLATION GUIDE
INSTALLING VENSIM SOFTWARE
Vensim comes in several different versions. The D-O LRP model will run fully on the Professional
Learning Edition (PLE) version of Vensim and can also be explored (without the possibility to modify
variables and equations) with Vensim Reader. Vensim PLE and Reader are free for educational use and
can be downloaded (in Windows or Macintosh OSX versions) from the Vensim website at
http: //vensim. com/download/.
280
-------
Macintosh users: when installing Vensim, you may get a message that the program cannot be opened because
it is not by a registered developer. In this case, users will need to click on the apple symbol in the navigation
bar, click on "System preferences, " and open "Security and Privacy. "Near the bottom of the window, you
should see the option to "Open anyway" the Vensim application. From this point on, Vensim and the D-O LRP
model operate much the same as in Windows.
Advanced users: although the model can be run and edited fully on Vensim PLE, it was built partly using
Vensim DSS, which has additional functionality for model editing and sensitivity tests. Users interested in
Vensim DSS should contact the authors of this report for more information.
OPENING THE MODEL
After opening Vensim, go to the File menu -$ Open Model and select the model file you wish to open. If
a dialog box pops up saying "The current display scaling is different from the computer the model was
last modified on...," click Yes to rescale the sketch. This will pop up another box, "Rescale Sketches,"
and click OK to accept defaults. The model should now be open and ready to run.
VENSIM FEATURES
Figure A-2 presents an annotated illustration of the Vensim window through which the user accesses the
D-O LRP model. This figure serves as a guide that will be referred to throughout this document as
different features of the model and Vensim software are described. Specific features of the Vensim
window are labeled with green circles, including some of the controls available on the vertical and
horizontal toolbars across the top and left side of the window. These features and controls include:
1) Menu for navigating across views 6) Graphing tool
2) An example of an Auxiliary/Constant 7) Table tool
variable 8) Runs Compare tool
3) Causes Tree and Uses Tree tools 9) Control Panel
4) Equation tool 10) Loops tool
5) Where to name and run a simulation n) Causes Strip tool
NAVIGATING THROUGH THE MODEL
When first opening the model, the user may want to zoom out or zoom into the sketch. This is done by
holding CTRL and turning the mouse wheel backward or forward. The user can also reposition the sketch
using the scroll bars at the bottom and right edges of Figure A-2, or by holding the right mouse button and
moving the mouse.
Vensim is capable of separating a large model into individual sketches or views, making it easier to
organize and view a complex model. The D-O LRP model contains over 80 views which can be used to
access different elements of the model. To navigate through the model by selecting a view, use the menu
at the bottom left of the Vensim window, identified as^l) in Figure A-2. Views are grouped by sector;
in Figure A-2, the displayed view "normalized modal person miles" is part of the Transportation sector.
The next sectors are Land Use, Equity, etc. The Dashboard view at the bottom of the view navigator
menu is described later in this User Guide.
281
-------
fif Vensim:LRP_v247_2015Q831.rridl Vartuel cost factor affecting adjusted automobile driver travel
£i te £d it View Layout Model Qptions Windows He IP
esTier 1
dgap
d gap Tier 1
oymnenteap
loym9r>tcepTier1
using
using Tier 1
perty value
property value tier 1
impervious surface
ous surface Tier 1
EQUITY
dabilityTier 1
iTficatJon
Irif ication
ENERGY
use in buildings
rgyuse in buildings Tier 1
gybill
gy bill Tier 1
iss ions Tier 1
lesions - curnui[ativB
supply
ECONOMY
irnment Revenues
Government Revenues Tier 1
LRP Financing
WATER
water supply and consumption
water consurmption Tier 1
stormwater N
stormwaler N Tier 1
StormwatBrP
StqrmwaterPTier 1
HEALTH
TreJlic Accidents
Traffic Accidents Tier 1
hearth
health Tier 1
DASHBOARD
dashboard
dashb o ard - rel aii^e i n dicators
' dashboard-relalr/e indicators Tier 1
ii density factor
affecting adjusted
automobik drK'er tra%'el>
FIGURE A-2. VENSIM WINDOW
282
-------
VARIABLES AND RELATIONSHIPS
There are three different types of variables in Vensim: box variables, flow variables, and
auxiliary/constant variables (an example of an auxiliary/constant variable is labeled as •» in Figure A-2).
A box variable represents an accumulating quantity (or stock) and has associated inflows and outflows,
which are flow variables. Auxiliary/constant variables are constants or calculated values that can affect
inflows and outflows or other auxiliary/constant variables. Arrows in the model view represent
relationships between variables; a given auxiliary or flow variable can be affected by all other variables
with arrows pointing to it (e.g., in Figure A-2, "fuel cost factor affecting adjusted automobile driver
travel" affects "adjusted person miles of automobile driver travel per day"). A variable can be located by
name by going to the Edit menu -> Find. To find the next instance of a variable in the model, press the F3
key.
TIER 1 AND TIER 2 OF THE MODEL
The D-O LRP SD model is divided into two geographic Tiers (Figure A-3). Tier 1 is the 1A mile radius
zones around proposed light rail stations, representing the region accessible to stations by walking. Tier 2
is the entire DCHC MPO region, inclusive of Tier 1. Most views of the model are represented for both
Tiers. For example, the view "normalized modal person miles" in Figure A-2 represents Tier 2, and the
next view is "normalized modal person miles Tier 1." The variable at the right hand side of the screen in
Figure A-2, "vehicle ownership factor affecting adjusted automobile travel," has a Tier 1 version,
"vehicle ownership factor affecting adjusted automobile travel Tier 1."
Tier 2: DCHC MPO
(according to TRM TAZs)
Chatham ( mim>
Tier 1: Vj mile radius
around proposed light
rail stations
FIGURE A-3 GEOGRAPHIC BOUNDARIES OF THE MODEL
Wore: in Figure A-3, TRM TAZs = Triangle Regional Model Traffic Analysis Zones
283
-------
SHADOW VARIABLES AND CAUSES/USES TREES
Variables between angle brackets () are shadow variables, or copies of variables that are
calculated in a separate view. Many of the variables in Figure A-2 are shadow variables, in addition to
being auxiliaries, constants, box variables, or flows. These variables are included in a specific view
because they interact with other variables in the view. Because they are calculated in another view,
however, it is not immediately apparent what variables affect them. Similarly, the use of shadow variables
can make it difficult to discern whether a variable in a given view influences variables in other views.
Vensim provides tools to identify which upstream variables affect a particular variable or which
downstream variables are affected by it: the Causes Tree and the Uses Tree. To use these tools, click on
the variable and then click on the Causes Tree or Uses Tree button, labeled as in Figure A-2.
Figure A-4 and Figure A-5 show how these tools can be used to trace causes or uses up to two steps away
from the variable. This is especially useful for looking at shadow variables, which can have relationships
that span multiple views.
EQUATIONS AND CONSTANTS
Each variable in Vensim is defined by either an equation or a predetermined value, which may vary with
time. To see how a variable is defined, first right-click on the variable name to open the variable options
window, shown in Figure A-6. Clicking the equation button will open the equations window which will
show either a single value, a lookup range which varies with time, an equation based on other variables,
or an equation based on a built-in function. Examples of each of these types of variable definitions are
provided in Figure A-7 through Figure A-l 1. The equations window can also be opened by clicking the
equations tool, labeled as 4 in Figure A-2, and then clicking on a variable.
Lookup variables such as "county property tax rate" mentioned below, are input as tables but can be
viewed as a graph for easier editing. Clicking on the "As Graph" button in Figure A-8 (circled in red) will
launch the window shown in Figure A-9. The user can enter values in the Input (x axis) and Output (y
axis) boxes on the left side of the window and the model will extrapolate values in a straight line between
the specified points. Users can also click and drag points on the graph to change their values. When done
editing, click OK to return to the main equation window.
284
-------
>G? Bin S3 number of deaths avoided due to cycling for transportation: Causes Tree
basslm* death rate
baseline number of deaths per ye
initial person rr.iles of cycling
icT. by resident: par 4ay per capital
parson miles of cycling isr transportation by rasida-.ti par diy par capita -4 raduction in morttlity IE t rasult of cycling for transportation'
xfuction in mratility par parson mila of cycling Sir transportation par day par capiu
FIGURE A-4. CAUSES TREE
person miles of public transit travel per day: Uses Tree
/person miles of public transit travel by residents per day •
/person miles of public transit travel per day per capita —
person miles of public transit tra^'el per day
person miles of public transit travel by residents per day per capita
-relative person miles of public transit travel per day per capita
person miles of travel per day
automobile driver share of person miles
automobile passenger share of person miles
automobile share of person miles
nonmotorized share of person miles
(public transit share of person miles}
public transit share of person miles
public transit unlinked passenger trips per dav
-•person miles of nonmotorized travel generated by public transit travel per day
"public transit unlinked passenger trips per year
FIGURE A-5. USES TREE
285
-------
CUana
o nape
f* None
f By Type
f* Box
(• Clear Box
(* Inside
P Below
Shape Color
r
r
r
r
r
f
i
Options for nonfatal crash injuries per VMT
Circle
Hexagon
Diamond
Triangle
Up Triangle
Left
Right
Face: Times New Roman Size (Points)
OArial Unicode^ * Bold |-|2 _rj
@Batang i |~~ "a'ic
@BatangChe y |~ Underline
@DFK.ai-SB p Strikethrough
@Dotum r- vertical
@DotumChe |- ^ Color
TrueType Times New Roman
1 Thickness
P~ Background Color [
OK
J<
FIGURE A-6. VARIABLE OPTIONS WINDOW
Edit elasticity of non motorized travel to jobs housing balance
Var iab 1 e I n f or ma t 1 on Edi t a D i f f er en t Var i ab 1 e
Name (elasticity of nonmotorized travel to jobs housing ba.
Type | Constant j^| Sub-Type | Normal j^J
Units [Drnnl
T| Check Units F Sup
ance |A11 ^ 1 /' evapotranspiration t
q_,, _h M_j=l % o£ rainwater used
bearch npdel 5 -ear pnpulation gro-th Tier if
nlementarv New Variable S year population growth
- • Back to Prinr- Edlt £2 Percent or annual renter cos
Group 1 irp v224 20150702 T| Min 1 Max 1 Incrl . .. - '£ P«i'^«in, LJI ^nnu^x reutar UUB
Jump to Hilite absolute change in person miles'
Equat ions
0.19
Functions |rr,M,,n _-J Keypad Buttons Variables i ]r,,,«e= _-J
ABS
DELAY FIXED
DELAY 1
DELAY 11
DELAY 3
DELAY 3 I
EXP
GET 123 CONS
GET 123 DATA
GET 123 LOOK
IGET DIEECT c
Comment
l~ Expand
Errors: | Equat
OK
7 | 8 J 9 | + :AHD:
IT 4 | 5 6 - :OR:
1 1 | 2 3 * :NOT:
0 | E . / :HA:
( 1 ) , * 0
™TS -^J^d^d^JlE
UPS _LJ_LJ_Ll_Ll >
OHSTAHTS - — „ : ' ' .. ' ...
Undo -> {[()]}
Ewing, Reid and Robert Cervero. 2010. Travel and the Built Environment: A Meta— Analysis . Journal of the *
American Planning Association 6, no. 3 (Summer): 265-294. DOT 10.1080/01944361003766766.
ion OK
Check Syntax | Check Model
J
Delete Variable | Cancel | Help |
FIGURE A-7. SINGLE VALUE VARIABLE EXAMPLE
286
-------
dft county property tax rate
Variable Information
Name ] county property tax rate
Type | lookup _»] Sub-Type |
Units |Dmnl j-J Chec
Group | irp v224 20150702 _^j Min |
1 -Edit a Different Variable
^~—~*^ I*11 - '<: eyapot
-3 C^ «*"!* a Search Modsl r °L.r"
k Units ^^«p
~ Max |
frf^ntary Hew Variable 5 year p
Back to Prior Edit '\ peroe
Jump to Hi lite absolute
ranspiration *
a water used
Dpulation growth Tier r*
Dpulation growth
at of annual renter co;
it of annual renter co;
change in person mile; '
Equations [(2000, 0.007)-(2040, 0.01)], (2000. 0.00926), (2001, 0.00929), (2002, 0.00755), (2003, 0.0078), (2004, 0.00792),
(2005, 0.00822), (2006, 0.00322), (2007, 0.00845), (2008. 0.00878), (2009, 0.00803), (2010, 0.00764). (2011, 0.00787),
(2012, 0.00788), (2013, 0.00786), (2014, 0.00805), (2040, 0.00805)1
Functions — [common ^1 — Keypad Buttons
ABS » 7 1 8
DELAY FIXED
DELAY1 =
DELAY1I 1 1 2
| 9 | + | :AND: |
J 6 - :OR:
| 3 « :HOT:
maAU n _ | .„, .
DEIAY3I ° 1 E 1 • \ / \ -HA- 1
EXP (J )J ,1 "J <>
GET 123 CONSTANTS . . i
GET 123 DATA > 1 > 1 1 < 1 <"
GET 123 LOOKUPS [ | ] | 1 | { | }
GET DIRECT CONSTANTS - Undo _>J{Im
Variables | |rauses
Comment Historical values are a weighted average tax rate between Durham and Orange Counties. 2014-2040 is kept
constant at 2014 rate.
~ Expand
Errors: | Equation OK
OK Check Syntax ]
Check Model
Delete Variable | Cancel |
J
Help J
FIGURE A-8. TIME-BASED VARIABLE EXAMPLE
Graph Lookup - county property tax rate
Historical values are a weighted average tax rate between Durham and Orange Counties. 2014-
2040 is kept constant at 201 4 rate.
nput Output
2000
2001
2002
2003
2004
0,00926 ±_
0.00929
0.00755
0.0078
0.00792
2005 JO. 00822
2006 1 0.00822
2007
2008
2009
2010
Ne
OK |
0.00845
0.00878
0.00803
0.00764 T
w
X-mi
Clear Poin
•
1=1
s
~?
. ' _
_ _i —
r\
J \r-J~
2000 _r] x=2028 v=O.C
Clear All Points | Cur->Ref
1009 X-max:|
Clear Reference
>040 _^\
Ref->Cur
Print
Y-max'
|0.01 _wj
Dmnl
Y-min:
0.007 _^J
Reset 9caling
Cancel
FIGURE A-9. TIME-BASED VARIABLE EXAMPLE - VIEWED AS GRAPH
287
-------
Edit: agriculture and open space impervious surface
Variable Information
Name [agriculture and open space impervious surface
Type [Auxiliary j^J Sub-Type [Normal j^J
Units jacre
Group | irp ,,
Equations
Functions
ABS
DELAY FIXED
DELAY 1
DELAY 11
DELAYS
DELAY 3 1
EXP
GET 123 CONS
GET 123 DATA
GET 123 LOOK
IGET DIRECT c
Comment
F Expand
T| Check Units 1 T Suppleme
224 20150702 _^| Mm ] Max |
1 rEdit a Different Variable 1
All T ^ evapotranspiration *
Search Model | "• °'ar po^ulSion gro-th Tier J •
t Hew Variable 5 year population growth
Back to Prior Edit ^ P"==« °| an""»| "^ =°'
Jump to Hilite j absolute change in person mile; "
agricultural land*agricultural impervious coef f icient+(parks and protected open space+vacant land)*open space
impervious coefficient
ICommon jj Keypad Buttons Variables hjcauses jj
,. 7 | 8 | 9 +| :AND: agr:
04 5 | 6 - :OR: ^
1 1 2 | 3 * :NOT: par!
0 | E | .|/| :HA: | wara
TANTS ' i* '~|
ONSTANTS - ,. i —
cultural impervious coef f icient
cultural land
i space impervious coefficient
.s and protected open space
mt land
-*
Errors: | Equation OK
OK
Check Syntax | Check Model |
Delete Variable | Cancel | Help |
FIGURE A-10. CALCULATED VARIABLE EXAMPLE FUNCTION
Edit delay total county real property value
Edit a Different Variable
Name jdelay total county real property value
Type [Auxiliary
hits [USD 2010
:Group |.lrp V247 20
^] Sub-Type |Normsl V]
~^~\ Check Units |
[All
. .
Search Model
New Variable
evapotranspiration
of rainwater used
year population growth Tier 1
5 year population growth
75 percent of annual renter costs tier
percent of annual renter costs
absolute change in person miles of eye'
Functions
ABE
DELAY FIXED
DELAY1
DELAY1I
DELAYS
EXP
GET 123 CONE
OET 123 DAT;
OET 123 LOOt
OET DIRECT C
DELAY FIXED (
total county real property value,
'
* 7
IJL
J i
0
(
TANTS
UPS [
ONSTANTS " — ,
Un
sypa
8
5
2
E
)
> =
]
Jo
d Bu
6
3
-;-
!
1, 1. 9354e+010L^
ttoi
-
»
/
-J-
{[
is
:AN
D:
:OR:
:NOT:
:NA:
-^~
)]}
total county real property value
FIGURE A-11. VARIABLE USING A BUILT-IN FUNCTION
288
-------
Some variables in the model use Vensim's built-in functions. Figure A-l 1 shows an example using the
Delay Fixed function (circled in red), which is one of a variety of built-in functions. Here the function
delays "total county real property value" by one year, after an initial value of 19B USD 2010.
RUNNING A SIMULATION
When the D-O LRP model is first opened, all variables are initialized to the values that define the
Business As Usual (BAU) scenario. Clicking on the "Simulate" button (to the right of Q) in Figure A-2)
will run this scenario. Once variables in the model are changed to define a new scenario, enter the
scenario's name in the scenario name window (to the left of ^B in Figure A-2) and click the "Simulate"
button. Since multiple scenarios can be stored and viewed at one time, it is useful to use a descriptive
name so that they can be easily distinguished.
VIEWING RESULTS
Once a simulation has been run, results can be viewed for any variable in the model. Results for all loaded
simulations can be viewed as either tables or graphs. Though Vensim provides several graphing options,
this guide only describes line graphs, which display variables as a time series.
Graphs and Tables
To view a graph, first click on the desired variable and then click on the graph button (see ^p in Figure
A-2). To view a table, first click on the desired variable and then click on the table button (see ^^ in
Figure A-2). Figure A-12 and Figure A-13 show a sample graph and table, respectively. The scenario run
most recently will appear in blue. Results are shown for the Light Rail scenario (blue line and text) and
the BAU scenario (red line and text). The red circle indicates the export feature that will copy the graph or
table to the user's clipboard, which can then be pasted into another program on the user's computer, such
as Word or Excel.
For graphs, the user can zoom in to see more detail. By holding down the shift key while clicking and
dragging to the side a new time range can be set. Similarly, by holding down the control key while
clicking and dragging up or down a new vertical range can be set. The selected ranges will appear once
you close and reopen the graph. To refresh just the vertical range of the graph, either click the graph
button again (see (^ in Figure A-2) or close and reopen the graph. The Time Axis tab in the Control
Panel window (see Figure A-14) can be used to alter the time range and to reset both the time and vertical
axes of output graphs. The "Reset to Full Range" button, circled in red, will make all graphs display the
original full range of values.
289
-------
firanh for vehicle fup.l cost incurred tw residents ner dav
vehicle fuel cost incurred by residents per day
2M
1,5 M
1M
500,000
0
2000 2006 2012 2018 2024 2030 2036
Time (Year)
vehicle fuel cost incurred by residents per day : Light Rail
vehicle fuel cost incurred by residents per day : BAU -
FIGURE A-12. SAMPLE GRAPH
TimVtffear) 2000 2001 2002 2003
"vehicle fuel cost incurred by residents per day" Runs: Light Rail BAU
vehicle fuel cost incurred by TesideiOsTp334)fy 745027 711321 824222
:BAU 777302 745027 711321 824222
FIGURE A-13. SAMPLE TABLE
290
-------
Control Panel
Variable TimeAxis 1 Scaling] Datasets
1 2000
Special <^ ^ End
Time (2040 " Time (2040
f~ Keep on top
FIGURE A-14. CONTROL PANEL WINDOW - TIME AXIS TAB
Comparing Multiple Simulations
If a user has run multiple simulations, the Runs Compare tool can help track how the inputs differ
between the simulations (see Qj) in Figure A-2). Figure A-15 displays an example scenario with a 10%
linear reduction in building energy use intensity between 2015 and 2040. The tool compares this scenario
to the Light Rail + Redevelopment scenario and shows which inputs were changed when the scenario was
run.
= d1ilinil Cnmnarinn I nwerhlrin FIJI I RT+redev and I RT + rfidffvelonment
Comparing Lower bldg EUI LRT+redev arid LRT + redevelopment
******Lookup differences between Lower bldg EUI LRT+redev and LRT + redevelopment******
^policy test building energy1 intensity Tier 1# - has changed in value
X Lower bldg EUI LRT+redevLRT + redevelopment
2000 1 1
2015 1 1
2040 0.9 1
#policy test building energy intensity^ - has changed in value
X Lower bfdg EUI LRT+redevLRT + redevelopment
2000 1 1
2015 1 1
2040 0.9 1
FIGURE A-15 EXAMPLE OUTPUT OF RUNS COMPARE TOOL
Users can create and save multiple D-O LRP SD model simulations in Vensim and select which of them
will be included in graphs and tables. Clicking on the Control Panel button (see fy in Figure A-2) will
open the Control Panel window, which has multiple tabs; the Datasets tab will list all loaded simulations,
as shown in Figure A-16. Clicking on one of the loaded simulations (red circle) will change the order of
these simulations, listing the selected simulation first. The simulation listed first will appear in graphs and
tables as a blue line or text. The second simulation will appear as a red line or text; and other simulations
will also have standardized colors. Simulations can be moved between the "Available info" and "Loaded
info" windows, and those in the "Loaded info" window will be displayed in graphs and tables. A
simulation file must be located in the same folder as the model, in order to appear under "Available info."
291
-------
If another simulation is required, it can be found by selecting "Load From ...," which prompts a pop-up
window that serves as a browser across local folders.
The last four files shown under "Available info" in Figure A-16 are actually historical datasets and
projections from models other than the D-O LRP SD Model, rather than being simulations (Excel files
and other datasets can be converted into Vensim simulation files (.vdf) using Vensim DSS). When any of
these files are moved to "Loaded info," their numbers may be compared with model runs-especially the
BAU scenario, which was calibrated to this data. Tierl DATA and Tier2 DATA contain historical and
projected data for variables in all major sectors of the model; their transportation data (VMT, congestion,
road-building, public transit, nonmotorized travel) come from the Triangle Regional Model's Existing +
Committed scenario (no road building after 2017 and no light rail). In contrast, TierlRoadMTP and
Tier2RoadMTP only contain data for transportation-related variables, which come from the "Preferred"
scenario that the TRM generated for the DCHC MPO Metropolitan Transportation Plan. The MTP
numbers assume that the light rail line will be built, unlike the BAU scenario, but they also assume the
same road-building projections as the BAU. Users interested in comparing transportation variables to
local data should load both the DATA and RoadMTP datasets, as no single dataset matches BAU
assumptions regarding both road construction and light rail construction. Scenarios analyzed by the D-O
LRP SD modeling team are described further in Chapter 4.
Control Panel
Variable Time Axis | Scaling Datasets Graphs]
Available-Info.. |
BAU
Tierl DATA
Tierl RoadMTP
Tier2 DATA
Tier? Road MTP
Lower bldg EUI LRT+redev
T+ redevelopment
Delete
Load From..
Keep on top
I Close i
FIGURE A-16. CONTROL PANEL WINDOW - DATASETS TAB
292
-------
Loops tool
The Loops tool shows the number of feedback loops that contain a selected variable (see I
A-2). Figure A-17 displays some of the 2822 loops that population is involved in.
in Figure
nomilatinn : I nons
Loop Number 1 of length 1
population
net migration
Loop Number 2 of length 1
population
deaths
Loop Number 3 of length 1
population
births
FIGURE A-17. LOOPS TOOL
Causes Strip
The Causes Strip tool shows the behavior of variables that directly affect a selected variable (see n
Figure A-2). This is a useful way to determine why a variable behaves the way it does. Figure A-18
shows the Causes Strip for "VMT Tier 1." Three scenarios are shown (blue, red, and green lines), and the
two variables that directly affect "VMT Tier 1" are "through traffic VMT Tier 1" and "VMT of trips
starting or ending in area Tier 1."
=tf ane
Light Rail
Light Rail
VMT Tier
3M
2.5 M
2M
1.5 M
1M
through tr£
900,000
800,000
700,000
600,000
500.000
VMT of tr
2M
1.65M
1.3 M
950,000
600,000
VMT Tier 1: Can?
+ Redevelopme
PS Strin Sl2"
ips starting or ending in area Tier 1
^^^•MI
p^
— ' —
-^
00 2020 2040
Time (Year)
FIGURE A-18. CAUSES STRIP TOOL
293
-------
DO-LRP SYSTEM DYNAMICS MODEL VIEWS
OVERVIEW
The DO-LRP SD model has separate views for different interventions, indicators, and sectors. Chapter 3
and Appendices B-C present more details about the sectors and the relationships among their variables. In
addition to the sector views described above, the model includes four customized "dashboard" views that
allow users to easily access the main features of the model.
INTERVENTIONS AND INDICATORS
The main "dashboard" view shown in Figure A-17 is a collection of the main input and output variables
related to policy interventions and model assumptions. This dashboard makes it easier for the user to view
a graph for each variable or change the value of input variables to define new scenarios. Recall that
variables can be edited by right-clicking the variable and then clicking the Equation button in the Options
window. The most significant output variables are organized by sector to allow the user to review
scenario outcomes more easily. There are four main sections in this view (the numbers below correspond
to the green circled numbers in Figure A-2):
1) Policy switches: These can be viewed using the equation editor to quickly review which policy
interventions have been activated. In particular, "Main policy switch" is a master switch used to
select among preset scenarios. Policy variables found in this section and elsewhere in the model
are detailed in Table A-l below.
2) Key indicators: All of the main outputs of the model are contained in this section, organized by
sector. These can be viewed as tables or graphs as described in the Viewing Results section
above. Key indicators colored blue summarize the behavior of a particular sector. A full list of
indicator variables, together with the sustainability dimension of each variable, is presented in
Table A-2 below.
3) Redevelopment switches: These can be viewed to quickly review or change model behavior
associated with land redevelopment.
4) Extreme value test switches: used in model validation, these set key variables such as GRP or
population to extreme values.
In addition to the main "dashboard" view, the model includes the "dashboard - relative indicators" view,
the "dashboard - relative indicators Tier 1 view," and the "dashboard - intensity indicators" view. The
"dashboard - relative indicators" view shown in Figure A-18 visualizes a selection of key Tier 2
indicators, relative to their value in the year 2000. The view contains a custom graph that displays results
for the most recently-run scenario. The box to the right of the graph lists the relative values of key
indicators in the year 2040. These values show whether a given indicator has grown faster or slower
compared to other indicators. The "dashboard - relative indicators Tier 1" view (not shown) reproduces
the "dashboard - relative indicators" view for Tier 1. The "dashboard - intensity indicators" view (not
shown) presents some key intensities such as GRP per capita and CO2 emissions per GRP.
294
-------
Durham-Orange Light Rail Project
System Dynamics Model
US EPA
t 5wttcli« ! : 7
. .
" -,:.
^^k
::
K -=«-
FIGURE A-17. DASHBOARD VIEW
295
-------
3
2.25
1 1.5
P
0.75
0
20
relative po
congestion
relative grc
relative CC
Relative indicators
— -*-
^
00
julation
:BAU
ss regie
)2 emis
&
_
2006
:BAU
^
_ —
— —
— =
^
^
_- — -
_-• **
— -— '
_^~
^
—
^
-~
/
-
,-"
--
2012 2018 2024 2030 2036
Time (Year)
nal procluc
3ions : BAL
2010 :BAU
Time (Year) 2040
relative population 1.958
congestion 1.144
relative gross regional product 2010 2.756
relative CO2 emissions 1.606
FIGURE A-18. DASHBOARD - RELATIVE INDICATORS VIEW
TABLE A-1. POLICY VARIABLES
VARIABLE NAME
TIER
NOTES
MAIN POLICY SCENARIOS
Main policy switch
Both
LAND USE
Overall target percent of land redeveloped table tier 1
Redeveloped density multiplier Tier 1
Density switch
Density multiplier table tier 1
Tier 1 only
Tier 1 only
Both
Tier 1 only
0 = BAU (with MTP-prescribed
amounts of road building and no
light rail)
1 = Light Rail
2 = Light Rail + Redevelopment
3+ = other, less-featured scenarios
The portion of developed land
that will be redeveloped by 2040
(e.g. 0.20 = 20%). Must be used in
conjunction with a Main Policy
switch setting that includes
Redevelopment.
The increase in initial density
applied to redeveloped land (e.g.
3 = 200% increase)
0 = default, 1 = density multiplier
table applied to new
development. Must be used in
conjunction with a Main Policy
switch setting that includes
Density.
The increase in initial density
applied to newly developed land
(e.g. 3 = 200% increase)
296
-------
VARIABLE NAME
More multifamily households Tier 1 switch
Percent of people in SFDU table historical Tier 1
TIER
Tier 1 only
Tier 1 only
NOTES
0=default, 1=More Multifamily
Households scenario with steady
decline in the percent of the
household population that lives in
single-family dwelling units.
Allows the percent of the
household population that lives in
single-family dwelling units to be
adjusted as desired.
TRANSPORTATION
Main policy switch
Gasoline price scenario
Fare free transit system in 2026
Decision to increase desired nonmotorized travel facilities
per developed acre
Parking price hike instituted Tier 1
Both
Both
Both
Both
Tier 1 (primary), Tier
2 (to the extent that
Tier 1 is part of Tier
2)
0, 1, 2, 4, 8, 9, 10, 12 = MTP-
prescribed amounts of road
building (otherwise, differentiated
according to decisions about the
light rail, redevelopment, and
density)
3, 5, 6, 7, 11, 13, 14, 15 = no road
building after 2017 (otherwise,
differentiated according to
decisions about the light rail,
redevelopment, and density)
0 = default
1 = increase in price of gasoline
effective 201 6
0 = default fare price
1 = all public transit becomes
fare-free in 2026
0 = default
1 = desired nonmotorized travel
facilities per developed acre
doubles in 2020, leading to
increased construction of
nonmotorized travel facilities
0 = default
1 = cost of parking within Tier 1
increases by $4.00 in 2020
ECONOMY
Higher rent switch
Earnings per retail employee with retail wage increase
switch Tier 1
Both
Tier 1 only
0 = gross operating surplus per sq
ft BAU
1 = gross operating surplus per sq
ft REDUCED ($5 less per sq ft than
BAU, starting in 2025)
Gross operating surplus per sq ft
represents how profitable local
companies are, per unit building
area.
0 = earnings per retail employee
BAU Tier 1
1 = earnings per retail employee
with retail wage increase Tier 1
($1 increase in the NOMINAL
hourly wage in 2016 and an
additional $.25 for each year
thereafter)
297
-------
VARIABLE NAME
Property tax rate revenue neutral switch
TIER
Both
NOTES
0 = county /city property tax rate
BAU
1 = county /city property tax rate
revenue neutral 2040
EQUITY
Percent of MF dwelling units below 77 percent of median
renter costs table tier 1
Tier 1 only
The percent of multifamily
dwelling units that are organically
affordable (In 2010 dollars, 77% of
the median housing costs equaled
about 40% of the poverty threshold
- the assumed maximum
households in poverty could spend
on housing).
ENERGY
Policy switch building energy intensity
Policy switch MPG
Policy test solar capacity
Policy switch electricity emissions factor
Tier 2 only
Both
Tier 2 only
Both
0 = no effect on building energy
intensity trend
1=10% decrease in building
energy intensity trend between
2015-2040
0 = no effect on MPG trend
1=10% increase in MPG trend
between 2015-2040
A multiplier on the desired future
solar electric capacity. 1 = no
change. 2 = double, 3 = triple,
etc.
0 = no effect on electricity CO2
emissions factor
1 = 23% reduction in electricity
CO2 emissions factor between
2022-2030 (approximates North
Carolina goal within Clean Power
Plan)
WATER
Percent N treated onsite
Percent reduction of existing N load
Target N load
Tier 1 only
Tier 1 only
Tier 1 only
Percent reduction in stormwater
runoff N from post-201 5
development due to onsite
stormwater treatment
Percent reduction in stormwater
runoff N from land developed
before 2015, due to onsite
stormwater treatment (such as
retrofits).
Desired stormwater runoff N
(Ib/acre/year) after onsite or
offsite stormwater treatment.
Offsite could include purchasing
mitigation credits.
HEALTH
Gasoline price scenario
Both
0 = default
1 = increase in price of gasoline
effective 201 6
Tests how gasoline price affects
walking and cycling
298
-------
VARIABLE NAME
Vehicle emissions reduced switch
NOx emissions per VMT with reductions
PM2.5 emissions per VMT with reductions
TIER
Both
Both
Both
NOTES
0 = no effect on PM2.5 or NOx
emissions rate
1=10 percent reduction in PM2.5
and NOx emission rate in new
vehicles, between model year
2020 and 2040
Tests reductions in passenger
vehicle NOx emissions between
model year 2020 and 2040
Tests reductions in passenger
vehicle PM2.5 emissions between
model year 2020 and 2040
TABLE A-2. KEY INDICATORS
VARIABLE NAME
TIER
SUSTAINABILITY DIMENSION(S)
LAND USE
Population
Net migration
Developed land
SF property value per SF DU
MF property value per MF DU
Nonresidential property value per sq ft
HHI index
HHI index for nonresidential sq ft
Total nonresidential sq ft
Retail density
Total impervious surface
Both
Both
Both
Both
Both
Both
Both
Both
Both
Both
Both
Economy, Society
Society
Economy, Environment
Economy, Society
Economy, Society
Economy
Economy, Society
Economy, Society
Economy
Economy
Environment
TRANSPORTATION
VMT
VMT per capita
Congestion
MPG with congestion
Person miles of nonmotorized travel per day
Public transit unlinked passenger trips per day
Vehicle stock
Vehicle trip distance
Vehicle trip duration
Parking cost of average trip
Functioning lane miles
Functioning nonmotorized travel facilities
Population not in zero car households
Both
Both
Both
Both
Both
Both
Both
Both
Tier 2 only
Both
Both
Both
Both
Economy, Environment
Economy, Environment
Economy, Environment
Economy, Environment
Environment, Society
Economy, Society, Environment
Economy
Economy
Economy
Economy
Economy
Economy
Economy, Society
ENERGY
299
-------
VARIABLE NAME
CC>2 emissions from buildings and transportation
CO2 emissions from buildings
CO2 emissions from passenger vehicles
SF energy intensity
MF energy intensity
Commercial energy intensity
Industrial energy intensity
PM2.s vehicle emissions tons
NOX vehicle emissions tons
Solar capacity
LFG to energy capacity
Total energy spending
TIER
Both
Both
Both
Both
Both
Both
Both
Both
Both
Tier 2 only
Tier 2 only
Both
SUSTAINABILITY DIMENSION(S)
Environment
Environment
Environment
Environment
Environment
Environment
Environment
Environment
Environment
Environment
Environment
Environment, Economy
ECONOMY
Gross regional product
GRP growth rate
Real property tax revenues per acre of developed land
Total retail consumption
Cumulative LRT revenues
Total LRT expenditure
Total employment
Unemployment rate
Both
Both
Both
Both
Both
Both
Both
Both
Economy
Economy
Economy
Economy
Economy
Economy
Economy
Economy
EQUITY
Affordability index
Affordability index for households in poverty
Median annual renter costs
Transportation related costs incurred by residents per year
per MF household
Transportation and renter costs per year per household
Population in poverty
Households in poverty at risk of displacement
Zero car households
Both
Both
Both
Both
Both
Both
Both
Both
Society
Society
Society, Economy
Society, Economy
Society, Economy
Society, Economy
Society
Society
WATER
Average precipitation
Total N load, Total P load
Total N load kg per ha
Total volume of runoff
Event mean concentration N, P (disaggregated by land use:
nonresidential, SF residential, MF residential, etc. )
Tier 2 only
Both
Both
Both
Both
Environment
Environment
Environment
Environment
Environment
300
-------
VARIABLE NAME
Total water demand
Residential water use, Nonresidential water use,
Nonrevenue water use
SF water use per household
MF water use per household
TIER
Both
Both
Both
Both
SUSTAINABILITY DIMENSION(S)
Environment
Environment
Environment
Environment
HEALTH
Total number of deaths avoided due to walking and cycling
for transportation
Delta premature mortality PM2.5 + NOx Krewski Lepeule
Crash fatalities per year
Both
Both
Both
Society
Society
Society
301
-------
Appendix B: Detailed Data Documentation
Please see the accompanying spreadsheet (separate file) for detailed documentation of the
model's data and relationships. The three tabs in this spreadsheet include:
• Calibrated Variables: contains information on variables in each sector that were
calibrated to external data sources, including historical data and projections.
• Exogenous Inputs: contains information on variables in each sector that were drawn
from exogenous sources.
• Equations from Literature: contains information on equations used in the model to
calculate selected variables, in cases where the equations themselves (rather than
parameter values or calibration targets) were drawn from external sources.
For each variable in the Calibrated Variables and Exogenous Inputs tabs, the appendix provides
information on the variable's units, the data source(s) used for initial, historical, and/or projected
values, and notes on subcategories, data processing, and calibration. For each equation in the
Equations from Literature tab, the appendix provides information on the variable calculated, the
variable's units, the equation itself, notes on the calculation process, and the data source from
which the equation was drawn.
302
-------
Appendix C: Model Input Characterization Tables
MODEL INPUT CHARACTERIZATION AND ASSESSMENT
To assist users of the D-O LRP SD Model, the project team assessed the quality of key model inputs, both
data sources and any methods used to manipulate them for use in the model. The results of this
assessment are an indicator of the confidence level in the model input (high, medium or low) and a
description of the uncertainties associated with it. Following the Quality Assurance Project Plan, the
process for determining the confidence level in an input are based on applying a weight of evidence
approach to the following criteria:
• Is the input based on information from one or more externally peer reviewed documents?
• Is there agreement in the literature or within the relevant community of practitioners about the
underlying data or method for the input? Or are there conflicting viewpoints?
• Do the characteristics of the input make it suitable for use in the context of the Durham-Chapel
Hill-Carrboro MPO? For example, is the input based on data from either the DCHC MPO or
another area with similar characteristics; or is it based on data from another area that is highly
site-specific?
• If the model input is based on manipulation of a data set, is the method used in developing the
input an established and widely applied approach? If the method applies equations developed
from external models, are those equations applied to local data in an appropriate manner?
The assessment results are presented in Tables C-l through C-7. For each input, we provide the source,
how the input is used in the model, the rationale for selecting the input, and the confidence level in the
input, along with a description of uncertainties associated with it. If the model input is used for calibration
purposes, that fact is noted in the "Model Input" column.
This information is made available to users of the model so that they can evaluate the relative confidence
level in model results. For example, for any output indicator produced from a scenario run, it is possible
to determine how many of the model inputs that are causally linked to it are classified as having high,
medium or low confidence levels. In addition, the assessment results can be used to help target specific
model inputs for sensitivity analyses that can help determine the extent to which uncertainty in their
values can affect the model results.
303
-------
TABLE C-1. MODEL INPUT CHARACTERIZATION FOR THE LAND USE SECTOR
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Acres of developed,
vacant, agricultural,
and protected open
space land in 2000
TJCOG. 2014. "CommunityViz 2 (CV2)
Parcel Geodatabase for Place Type &
Development Status Editing." Place
type and development status of
parcels calculated for the year 2000
by back-casting 2013 per-capita
values.
Initial values for acres by
category, which affects
developed land.
Created in part to inform
the LRP process; best
inventory of current land
use available for entire
study area.
MEDIUM-HIGH: Each parcel was
reviewed by local planning staff, but
values do not quite match local
comprehensive plan estimates, at least
in Durham for which these were
available.
Division of developed
commercial acres by
category (retail, office,
service, and industrial)
and residential acres by
category(single-family
and multifamily)
TJCOG. 2013. "Imagine 2040: The
Triangle Region Scenario Planning
Initiative Final Summary Document."
Used in the Imagine 2040
CommunityViz model.
Initial values for acres of
developed land by
category. Also used to
calculate floor area ratios
(FAR) by land use type, and
residential density per
acre, which affect the rate
of land development.
Created in part to inform
the LRP process; best
inventory of current land
use available for entire
study area.
MEDIUM-LOW: Each parcel was
reviewed by local planning staff, but
the dataset was still in editing at the
time of analysis, and the land use
tables are broad estimates for each
jurisdiction's zoning classifications, and
therefore do not necessarily reflect
use.
Developed land
estimates in 2040
(calibration)
DCHCMPO. 2013. "Triangle Regional
Model version 5: Socioeconomic data
and projections for the preferred
growth scenario and travel demand
result shapefiles." We developed two
estimates of total developed land
based on data from TRMvS: 1)
projected jobs and dwelling units
multiplied by average space needed
per unit of each, 2) acreage of all
parcels with either employment or
dwelling units allocated in 2040.
Used to calibrate land
development to within
reasonable range, which
affects impervious surfaces
and non-motorized
facilities.
Created in part to inform
this LRP process; best
inventory of current land
use available for entire
study area.
LOW: Development status was never
assigned for future projections in
Imagine 2040, so we had to derive
estimates consistent with Imagine 2040
projections based on allocations that
weren't meant to estimate land
development.
Total impervious
surface (calibration)
US EPA. 2013. "EnviroAtlas."
Affects loadings of
phosphorus and nitrogen
from storm water runoff.
Provided the broadest
coverage of the study
area; consistent with
impervious surface data
for the City of Durham
shared by the Durham
City/County Planning
Department, which
excluded impervious
surfaces from roads.
MEDIUM: Input comes from an
authoritative source for land cover
data, but it does not cover the entire
study area and was only available for
one year (2010).
304
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Population, households,
and housing units
(calibration)
ESRI Community Analyst. 2014.
"Census Data 2000, ACS 1-yr and 5-yr
Estimates 2005-2013, and ESRI
Demographic Projections for 2014 and
2019"
Initial values and
calibration for population,
drivers of demand for
residential land.
Community Analyst uses
Census and ACS data, and
the clipping method
weights it by block level
population data, making it
more accurate than
personally downloading
Census data and
performing area-based
weighting in excel.
HIGH: The U.S. Census Bureau is the
authoritative source for demographic
data and the Community Analyst
clipping method more accurate than by
hand.
Birth, death, and
migration rates
NC Department of Health and Human
Services Vital Statistics Database.
2015. "Resident Live Births, Deaths,
and Estimated Net Migration 2000-
2014." NC State Data Center,
Accessed from Log Into North Carolina
(LINC) (Durham/Orange County
average)
Average death rate is held
constant throughout the
model period; the average
historical birth rate trend
is extended into the
future.
Most geographically
specific data source with
longest time series.
MEDIUM: the population levels
simulated by using historical birth,
death, and migration rates did not
match historical population trends, so
additional calibration was applied to
birth and migration rates.
Proportion of dwelling
units that are single-
family
U.S. Census Bureau. 2000. "Census
2000, Summary File 3." And
ESRI Community Analyst. 2014.
"Census Data 2000, ACS 1-yr and 5-yr
Estimates 2005-2013, and ESRI
Demographic Projections for 2014 and
2019."
Used, along with household
sizes to calculate the
percent of people in single-
family dwelling units,
which affects residential
land usage and energy
efficiency.
Only source found;
property database doesn't
include number of units in
multifamily structures.
MEDIUM: The U.S. Census Bureau is the
authoritative source for this input, but
the data were only available for two
years, so a consistent trend could not
be established.
Single-family and
multifamily household
sizes
Hodges-Copple, John. 2012. "FINAL
2040 Population Forecast by County
for CommViz -EPA,"(Year 2010, based
on ACS 2010 data by county) .
Used, along with the
percent of people in single-
family dwelling units, to
calculate demand for
single-family and
multifamily dwelling units
based on population.
Only sources available for
the study area.
HIGH for Tier 2; MEDIUM for Tier 1 as
there are no good benchmarks to check
the distribution except estimates of
owner-occupied and renter-occupied
household sizes.
305
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Retail, industrial,
office, and service
building square feet
(calibration)
Durham County Tax Administration.
2000-2014. Durham County Tax
Administration Real Property
Database, Orange County Tax
Administration. 2014. Orange County
Parcel Database, Chatham County Tax
Administration Office. 2014. Chatham
County Tax Parcel Database
Initial year 2000 values and
calibration. Also used to
calculate employee space
ratios.
Most detailed source; used
by Tax Office and Planning
departments.
MEDIUM: According to Durham Planning
Department representative Laura
Woods, the data may underestimate
actual values; there are some blank
properties in database, as well as
duplicate records and inaccuracies.
Effect of developed
portion of residential
land on migration table
Vina-Arias, Laura Beatriz. 2013.
"Understanding Patterns of Growth at
Kendall Square Using a System
Dynamics Approach."
Used in Tier 2 to relate
land supply to immigration.
Only source found to
address the relationship
between land
development and
immigration; used
elsewhere in a system
dynamics model.
LOW: First, the effect table was
originally used to relate land supply to
residential construction rather than
migration - showing the attractiveness
of construction increases gradually
until land becomes very scarce, when
it gradually decreases (sharper than
increase). Second, the context is very
different - an urban core, rather than a
suburban metro area. Finally, the
strength of the relationship had to be
decreased during the calibration
process.
Effect of
unemployment on net
migration Tier 1
N/A
Used in Tier 1 to relate the
unemployment rate to
immigration.
No literature found
quantifying the topic.
LOW: No source available, calibrated
to historical data on population.
Average housing
lifetime
US EPA. 2013. Analysis of the Life
Cycle Impacts and Potential for
Avoided Impacts Associated with
Single-Family Homes. Page 4.
Used as in input to the
calculation of single family
dwelling units.
Authoritative source that
reviewed the research on
the topic to provide a
reasonable range.
MEDIUM: The meta-analysis provided a
wide range of estimates which were
only given on a national scale. We
chose the lowest estimate.
Percent second homes
Economic and Strategic Research.
2014. "Second homes: Recovery post
financial crisis."
Used as an input to the
calculation of single-family
dwelling units.
Authoritative source that
reviewed the research on
the topic to provide a
reasonable range.
MEDIUM: Value is an average for 1998-
2014 for the nation and includes both
single-family and multifamily homes
with a mortgage, while we used the
value only for single-family homes.
306
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
MEDIUM: The most comprehensive local
source, but values are averages for
broad categories of land use.
Residential impervious
surface coefficients
NCDENR. 2011. Jordan Lake
Stormwater Load Accounting Tool
User's Manual. See Table 5 on page
20.
Used to calculate
impervious surface due to
residential land use.
This source relates
residential density to
impervious surface cover
based on CIS land cover
data from several counties
in MD, VA, and PA for use
in the Triangle region of
North Carolina. More
representative of the local
context than source used
for nonresidential and
road impervious surface
(Washburn et al, 2010).
Nonresidential and road
impervious surface
coefficients
Washburn, Barbara, Katie Yancey, and
Jonathan Mendoza. 2010. User's Guide
for the California Impervious Surface
Coefficients. See pages 20-17.
Used to calculate
impervious surface due to
land use of nonresidential
uses and roads.
Cited in the Imagine 2040
model documentation;
best available source for
coefficients for
nonresidential uses.
MEDIUM: Used in local regional
planning efforts, but based upon
California land uses, which may not be
representative of land uses in the
study area.
307
-------
TABLE C-2. MODEL INPUT CHARACTERIZATION FOR THE TRANSPORTATION SECTOR
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
(1) Annual change in
desired vehicle
ownership per person
not in a zero-car
household that is
independent of explicit
inputs (i.e., not
attributed to any
specific cause) and (2)
Elasticity of desired
vehicle ownership per
person not in a zero-car
household to per-capita
income
International Energy Agency and
World Business Council for Sustainable
Development. 2004. "IEA/SMP
Transportation Model." Spreadsheet
model discussed in: Fulton, Lew, and
G. Eads. 2004. "IEA/SMP Model
Documentation and Reference Case
Projection."
(1) Used to account for
factors influencing vehicle
stock other than resident
per capita net earnings and
fuel cost; and (2) Used to
determine how much
influence resident per-
capita net earnings has on
vehicle stock. Vehicle
stock helps drive household
transportation costs and
vehicle registration fee
revenue.
Best identified equation for
estimating desired vehicle
ownership; produces results
consistent with other data
sources used in the model.
LOW: The equation was created for
the combined area of the U.S.,
Canada, and Mexico, so it may not be
representative of the study area. As
part of the downscaling, we assumed
resident per-capita net earnings could
be used instead of GDP per capita.
We also assumed that counting the
effect of fuel cost separately would
not be double-counting. Whereas the
source's equation was for vehicles per
capita, it is included in the D-O LRP
SD Model in terms of vehicles per
person not in a zero-car household.
Baseline rate of change
in through-traffic VMT
NC Office of State Budget and
Management. 2015. "Population
Estimates and Projections."
Since through-traffic VMT
is, by definition, largely
the result of factors
external to any given study
area, this rate of change
serves as a stand-in for
those external factors.
VMT is used to estimate
traffic congestion, fuel
consumption by vehicles,
and traffic accidents.
Baseline rate of change is
based on the rate of
population change in North
Carolina, since VMT and
population tend to follow
similar trends. Local data on
through-traffic VMT is
scarce. Produces VMT
results that are consistent
with projections.
LOW: The relationship between
through-traffic VMT and North
Carolina population is assumed, not
based on empirical evidence.
308
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Change in person miles
of public transit travel
per year due to adding
fixed guideway transit
"Draft Spreadsheet Tool: Estimated
Ridership and Cost of Fixed-Guideway
Transit Projects," created as part of
TCRP Project H-42: Chatman, Daniel
G., Robert Cervero, Emily Moylan, Ian
Carlton, Dana Weissman, Joe
Zissman, Erick Guerra, Jin Murakami,
Paolo Ikezoe, Donald Emerson, Dan
Tischler, Daniel Means, Sandra
Winkler, Kevin Sheu, and Sun Young
Kwon. 2014. "TRCP Report 167:
Making Effective Fixed-Guideway
Transit Investments: Indicators of
Success."
Equation from literature
used to dynamically
determine new public
transit person miles of
travel resulting from the
LRT line. Greater public
transit use reduces VMT
(and hence congestion,
fuel consumption, and
traffic accidents) and helps
drive nonmotorized travel,
and hence physical
activity.
TCRP is a well-respected
source; the equation
produces results that appear
reasonable when compared
to the projections in the
Metropolitan Transportation
Plan.
Congestion (calibration)
DCHCMPO. 2013. "Triangle Regional
Model version 5: Travel demand result
shapefiles." (TRM)
Used to estimate vehicle
speed, which affects daily
modal person miles using
elasticities. Modal person
miles affect household
transportation costs,
traffic accidents, fuel
consumption, and health
outcomes.
CIS format, highly detailed,
with per-link variables for
traffic volume, capacity, and
freeflow and peak-period
travel speeds. Also, it is the
official travel-behavior
model for the Durham-
Chapel Hill-Carrboro
Metropolitan Planning
Organization.
MEDIUM: The equation is not specific
to this particular study area and does
not account for the number of
revenue miles run on the new LRT
line. The equation is also for fixed-
guideway transit in general, as
opposed to LRT specifically. Finally,
the equation does not factor in any
change over time.
HIGH: It is an authoritative source,
but many modeling assumptions go
into the numbers. "Congestion" is
defined here as the ratio of freeflow
speed on a road link to the traffic
speed on that link during "peak"
periods (6-10 AM and 3:30-7:30 PM on
weekdays), weighted by the peak-
period VMT on each link. This
definition does not account for
congestion during the rest of the day,
or for variations in congestion or
traffic volumes during the designated
"peak" periods.
Congestion per
weekday peak period
VMT per lane mile
DCHCMPO. 2013. "Triangle Regional
Model version 5: Travel demand result
shapefiles." (TRM)
Used to translate modeled
peak-period-VMT and lane-
mile figures into a measure
of traffic congestion that
can feed back into the
model.
CIS format, highly detailed,
with per-link variables for
traffic volume, capacity, and
freeflow and peak-period
travel speeds. Also, it is the
official travel-behavior
model for the Durham-
Chapel Hill-Carrboro
Metropolitan Planning
Organization.
MEDIUM: We took ratios between
congestion and peak period VMT per
lane mile in 2010 and 2040 and
assumed a direct, causal, proportional
relationship. Other factors influencing
congestion are ignored.
309
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Construction cost per
lane mile; maintenance
cost per lane mile per
year; construction cost
per nonmotorized
facility mile
CAMPO and DCHCMPO. 2013. "2040
Metropolitan Transportation Plans."
(MTP); DCHCMPO. 2013. 'Triangle
Regional Model version 5:
Socioeconomic data and projections
for the preferred growth scenario and
travel demand result shapefiles."
(TRM)
Constant values are
multiplied by numbers of
roadway lane miles to
arrive at transportation
facility expenditure totals.
Resulting values for costs per
lane mile are comparable to
what transportation planners
in the local MPO use.
MEDIUM-HIGH: Does not disaggregate
the cost of urban vs. rural roadwork,
the cost of highway vs. nonhighway
roadwork, or facilities built by the
government vs. by developers. Does
not account for variations in the cost
of right-of-way acquisition. We
assume that the MTP's anticipated
nonmotorized-facility spending is all
construction and not maintenance
(due to sidewalks being the
responsibility of private land owners).
Elasticity of congestion
to commercial FAR Tier
1
N/A
Used to estimate the
effect of average building
floor area ratio (FAR) on
traffic congestion in Tier 1.
Traffic congestion reduces
VMT and increases the use
of other travel modes.
Modal person miles affect
household transportation
costs, traffic accidents,
congestion, fuel
consumption, and health
outcomes.
No source available.
Although the model also has
a relationship wherein VMT
increases traffic congestion,
Tier 1 traffic congestion is
also affected by
nonmotorized traffic,
including nonmotorized
travel within Tier 1 by
people who entered Tier 1
by automobile. FAR is used
as a proxy for these more
immediate drivers. This
effect is represented only in
Tier 1 because Tier 2
nonmotorized travel is very
low, so it is assumed to have
a negligible effect on Tier 2
congestion.
LOW: No source available.
Elasticity of desired
vehicle ownership per
person not in a zero car
household to fuel price
Johansson, Olof, and Lee Schipper.
1997. "Measuring the long-run fuel
demand of cars: Separate estimations
of vehicle stock, mean fuel intensity,
and mean annual driving distance."
Used to determine how
much influence fuel price
per VMT has on vehicle
stock, which helps drive
household transportation
costs and vehicle
registration fee revenue.
Best available equation for
this purpose; produces
results consistent with other
data sources used in the
model.
LOW: Came from a study that looked
at entire countries, including
countries that are much farther from
vehicle-ownership saturation than the
U.S.
310
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Elasticity of person
miles of travel per day
per capita to GRP per
capita (separate values
for automobile driver,
automobile passenger,
nonmotorized, and
public transit travel)
(1) BEA. 2014. "Gross Domestic
Product (GDP) by Metropolitan Area:
Per Capita Real GDP, 2001-2013, for
the Durham-Chapel Hill, NC
Metropolitan Statistical Area (MSA)."
(2) CAMPO and DCHC MPO. 2013.
"2040 Metropolitan Transportation
Plans." (MTP)
(3) DCHC MPO. 2013. 'Triangle
Regional Model version 5: Travel
demand result shapefiles." (TRM)
(4) NuStats. 2006. Greater Triangle
Travel Study, Household Travel Survey
Final Report.
(5) ESRI Community Analyst. 2014.
"ACS 1 -yr and 5-yr Estimates 2005-
2013."
Used to create a baseline
of overall modal person
miles of travel that fits
projections and which all
other drivers of modal
person miles may modify in
a process where most
other inputs serve to
change the mode shares,
as opposed to the overall
quantity of person miles
across modes. Modal
person miles affect
household transportation
costs, traffic accidents,
congestion, fuel
consumption, and health
outcomes.
(1) BEA is an authoritative
government source that
looks at individual
metropolitan areas (as
Metropolitan Statistical
Areas). (2) The Metropolitan
Transportation Plan (MTP) is
the authoritative
transportation planning
document in the DCHC MPO.
Produces results consistent
with other data sources used
in the model. (3) The TRM is
the primary source of travel-
behavior projections for the
MTP. (4) The Household
Travel Survey Final Report is
one of the TRM's key inputs.
(5) Community Analyst is
able to clip ACS data from
the U.S. Census Bureau to
specific geographic areas.
MEDIUM-LOW: This is a custom-made
elasticity, calculated by assuming
that two separate projections will
both come to pass. As such, the
actual elasticity between person
miles of travel per capita and GRP per
capita might be greater or lesser,
with the apparent outcome shown
here being the result of the effects of
that elasticity being either offset or
compounded by other factors not
accounted for in the model.
Furthermore, it is unconfirmed
whether the two projections used
here are based on compatible sets of
assumptions.
Elasticity of public
transit travel to fare
price
McCollom, Brian E., and Richard H.
Pratt. 2004. "TCRP Report 95:
Traveler Response to Transportation
System Changes: Chapter 12—Transit
Pricing and Fares." (page 12-9)
Used to determine how
much changes in the
average public transit fare
price affect person miles
of travel by public transit.
Greater public transit use
reduces VMT (and hence
congestion, fuel
consumption, and traffic
accidents) and helps drive
nonmotorized travel, and
hence physical activity.
Source adapted this
elasticity from the Simpson
& Curtin formula, which is
commonly used among public
transit planners. Also, TCRP
reports are well-regarded in
their own right.
MEDIUM-HIGH: Not local.
311
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Elasticity of public
transit travel to vehicle
revenue miles
Sinha, Kumares Chandra, and Samuel
Labi. 2007. Transportation Decision
Making: Principles of Project
Evaluation and Programming, (page
55)
Used to determine how
much changes in the
number of daily public
transit vehicle revenue
miles affect person miles
of travel by public transit.
Greater public transit
ridership reduces VMT (and
hence congestion, fuel
consumption, and traffic
accidents) and helps drive
nonmotorized travel, and
hence physical activity.
Provided an elasticity
between a variable that data
could easily be obtained for
(revenue miles) and public
transit use.
MEDIUM: Results appear reasonable,
but we do not possess information on
the original study that produced the
elasticity.
(1) Elasticity of through
traffic to automobile
speed;
(2) Elasticity of through
traffic to fuel cost
Litman, Todd. 2013. "Understanding
Transport Demands and Elasticities:
How Prices and Other Factors Affect
Travel Behavior."
Used to determine how
much changes in (1)
average vehicle speed and
(2) fuel cost affect
through-traffic VMT. VMT is
used to estimate traffic
congestion, fuel
consumption by vehicles,
and traffic accidents.
Same source provided
multiple related elasticities
for this model, creating a
greater likelihood of
consistency. Comes with
note reading "in areas with
high vehicle ownership
(more than 450 vehicles per
1,000 population)," which
suggests it is applicable to
the study area for this
model.
MEDIUM-LOW: Elasticity is not specific
to this region. Also, using this
parameter for through-traffic VMT
implies that VMT of trips originating
outside of the study area are just as
sensitive to average vehicle speeds
and the cost of vehicle fuel per VMT
in the study area as VMT of trips
starting or ending in the study area.
Elasticity of travel to
(1) automobile speed,
(2) fuel cost, and (3)
parking cost (separate
values for automobile
driver, automobile
passenger,
nonmotorized, and
public transit travel)
Litman, Todd. 2013. "Understanding
Transport Demands and Elasticities:
How Prices and Other Factors Affect
Travel Behavior."
Used to determine how
much changes in (1)
average vehicle speed, (2)
fuel cost, and (3) parking
cost affect person miles of
travel by mode. Modal
person miles affect
household transportation
costs, traffic accidents,
congestion, fuel
consumption, and health
outcomes.
Same source provided
multiple related elasticities
for this model, creating a
greater likelihood of
consistency. Comes with
note reading "in areas with
high vehicle ownership
(more than 450 vehicles per
1,000 population)," which
suggests it is applicable to
the study area for this model
MEDIUM: Elasticities are not specific
to this region.
312
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Elasticity of travel to
(1) intersection density
and (2) population
density (separate
values for automobile
driver, nonmotorized,
and public transit
travel); (3) elasticity of
nonmotorized travel to
jobs housing balance
Ewing, Reid and Robert Cervero.
2010. 'Travel and the Built
Environment: A Meta-Analysis."
Used to determine how
much changes in (1) the
density of intersections
that are not entirely
automobile-oriented, (2)
population density, and (3)
the jobs-housing balance
affect person miles of
travel by mode. Modal
person miles affect
household transportation
costs, traffic accidents,
congestion, fuel
consumption, and health
outcomes.
Journal article synthesizes
results of multiple studies on
the same causal
relationship(s) and is a highly
regarded work by highly
regarded researchers in the
transportation planning
field.
MEDIUM: Distinctions made in
individual studies are obscured by
composite numbers, possibly leading
to misuse.
Elasticity of vehicle trip
distance to vehicle
speed
DCHCMPO. 2013. "Triangle Regional
Model version 5: Travel demand result
shapefiles." (TRM); CAMPO and DCHC
MPO. 2013. "2040 Metropolitan
Transportation Plans."
Used to help determine the
average distance of a
vehicle trip. Along with a
separate calculation of
overall VMT by residents,
this is used to arrive at an
aggregate number of
vehicle trips by residents,
which drives how much
residents spend on parking.
We employed this ad-hoc
elasticity because we could
not find an elasticity or
formula for the distance of
an average trip, as opposed
to number of trips or
cumulative distance of trips.
Because this elasticity was
created from TRM/MTP data,
it replicates the trends in
that data.
LOW: The elasticity produces results
consistent with TRM/MTP data, but it
is still an ad-hoc elasticity lacking the
usual rigor of a scientific study.
Finished LRT line miles
Triangle Transit. 2015. "Our Transit
Future."
Used in calculating the
land use of the LRT, its
building expense, its
revenue miles, and the
amount of roadway lane
miles that are disrupted
during LRT construction,
which affects traffic
congestion.
Source is the agency in
charge of planning the light
rail line.
HIGH: Most of the LRT alignment has
been determined at the time of this
writing.
313
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Off-peak-period vehicle
speed; freeflow speed
of the roadway link on
which the average peak
period vehicle mile of
travel is performed
DCHCMPO. 2013. "Triangle Regional
Model version 5: Travel demand result
shapefiles." (TRM)
Freeflow speed and
congestion are used to
estimate peak period
vehicle speed, which,
together with off-peak-
period vehicle speed,
determines average vehicle
speed. Higher vehicle
speeds increase
automobile travel and
reduce public transit and
nonmotorized travel,
affecting fuel
consumption, household
transportation costs, and
health outcomes.
The TRM is the primary
source of VMT and traffic
congestion projections used
by local transportation
planning agencies; spatial
nature of the data allows
clipping to both Tiers.
HIGH: Authoritative source with
straightforward application to the
study area.
Functioning lane miles
DCHCMPO. 2013. "Triangle Regional
Model version 5: Travel demand result
shapefiles." (TRM)
Along with VMT, this is
used to calculate traffic
congestion. It is also used
to determine land use and
impervious surface area
due to roadways.
The TRM result CIS
shapefiles provide detailed
counts of both existing lane
miles and those planned to
be built by 2040, according
to theMTP.
MEDIUM-HIGH: Some lower-order
roads are represented as "centroid
connectors," which only estimate
their cumulative lane miles.
314
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Functioning
nonmotorized travel
facilities (calibration)
DCHCMPO. 2013. "Triangle Regional
Model version 5: Socioeconomic data
and projections for the preferred
growth scenario shapefiles." (TRM)
Indicated percent
reduction in driver
person miles due to
congestion
N/A
Used to drive roadway
intersection density
(excluding those that are
purely automobile
oriented), which affects
modal person miles. Modal
person miles affect
household transportation
costs, traffic accidents,
congestion, fuel
consumption, and health
outcomes.
Best available source that
quantifies miles of
nonmotorized facilities;
spatial nature of the data
allows clipping to both Tiers.
Used to determine how
much changes in traffic
congestion affect person
miles of automobile driver
travel. Automobile driver
person miles affect
household transportation
costs, traffic accidents,
fuel consumption, and
health outcomes, as well
as have a feedback effect
on traffic congestion.
No source was available that
indicated the degree to
which traffic congestion
causes people to drive less
without traveling more by
other modes, as opposed to
traffic congestion causing
people to switch travel
modes (for which a source
was used).
MEDIUM-HIGH: Does not distinguish
between sidewalks, bike lanes, etc.,
even though these different types of
facilities attract different amounts of
nonmotorized person miles, with
different proportions of those person
miles being walking or cycling. Does
not indicate how well-connected the
nonmotorized travel facilities are,
even though well-connected facilities
are much more likely to drive
nonmotorized travel than
disconnected facilities. When clipping
data to the Tier 1 boundary, it was
necessary to assume that
nonmotorized facilities were evenly
distributed throughout any given
Traffic Analysis Zone that was
clipped.
LOW: No source available.
315
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Initial nonmotorized
travel facilities per
developed acre
(1) DCHCMPO. 2013. 'Triangle
Regional Model version 5:
Socioeconomic data and projections
for the preferred growth scenario
shapefiles." (TRM)
(2) TJCOG. 2014. "CommunityViz 2
(CV2) Parcel Geodatabase for Place
Type & Development Status Editing."
Estimates of developed land in 2013
are obtained for the CV2 parcels
assigned a development status of
developed, minus parcels assigned a
place type of POS (Protected Open
Space).
Intersections excluding
those that are purely
automobile-oriented
(calibration)
US EPA, Office of Policy, Office of
Sustainable Communities. 2013. Smart
Location Database version 2.0. (SLD)
Land use per lane mile
(separate values for
rural highways, rural
nonhighways, urban
highways, and urban
nonhighways)
NC DOT. 2007. Roadway Design
Manual. (Part II, Chapter 9: Right of
Way)
Multiplied by initial
developed land to get
initial functioning
nonmotorized travel
facilities. Used as the
default value for desired
nonmotorized travel
facilities per developed
acre. Nonmotorized travel
facilities drive the density
of intersections that are
not purely automobile-
oriented, helping to
increase nonmotorized and
public transit person miles
and decrease VMT,
affecting fuel
consumption, household
transportation costs, and
health outcomes.
(1) Best available source
that quantifies miles of
nonmotorized facilities;
spatial nature allows clipping
to both Tiers.
(2) Provided the most
detailed accounting of land
usage by acres that was
available for the entire study
area; provided a consistent
categorization system for
the study area, and was used
in the development of the
Imagine 2040 model, and
therefore was assumed to be
consistent with the data and
projections used for the
transportation sector.
MEDIUM-LOW: For the sake of
calibration, this value had to be
adjusted in Tier 2 (but not in Tier 1).
Some interpolation was required to
get the numerator and denominator
values to be for the same year. Due to
a lack of land development data from
prior to 2013 and a lack of
nonmotorized facility data from prior
to 2010, we had to back-cast to
estimate a value for 2000.
The density of
intersections affects modal
person miles through
elasticities. Modal person
miles affect household
transportation costs,
traffic accidents,
congestion, fuel
consumption, and health
outcomes.
Used to determine the land
use of roadways in the area
and the amount of
impervious surface (and
hence runoff) that they
contribute.
Provides a tally of
intersections that excludes
purely automobile-oriented
ones, which makes it more
useful for examining effects
on mode choice; spatial
nature allows clipping to
both Tiers.
MEDIUM-HIGH: Block Groups are a
courser unit than TAZs; data are
available for 2010 only
Authoritative source on road
rights-of-way.
MEDIUM: The source is authoritative,
but we did not check it against actual
local conditions, and the source
provides wide allowable ranges that
apply to new road construction.
316
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Land use per LRT line
mile
N/A
Used to determine the land
use of the light rail line,
which is a component of
impervious surface, which
determines runoff
volumes.
No source available.
Estimated from an assumed
right-of-way width.
LOW: No source available.
LRT construction cost
per line mile; LRT O&M
cost per year
Triangle Transit, Chapel Hill Transit,
Orange County, DCHC MPO, and
Orange County Public Transportation.
2012. 'The Bus and Rail Investment
Plan in Orange County." DCHC MPO,
Triangle Transit Board of Trustees,
and Durham Board of County
Commissioners. 2011. "The Durham
County Bus and Rail Investment Plan.'
Used to quantify spending
on construction and
operation of the light rail
line.
Authoritative planning
documents for matters of
public transit spending.
HIGH.
LRT revenue miles per
line mile per day
Triangle Transit. 2015. "Our Transit
Future."
Contributes to overall
public transit revenue
miles, which affects public
transit person miles; drives
LRT electricity use.
Source is the agency in
charge of planning the light
rail line.
MEDIUM: This input is from a back-of-
the-envelope calculation based on
anticipated service hours and
frequencies for the future LRT line.
This calculation has not been verified
by the transit agency that will
operate the line.
Ownership and
maintenance costs for
one automobile for one
year
Litman, Todd Alexander, and Eric
Doherty. 2009. "Transportation Cost
and Benefit Analysis: Techniques,
Estimates and Implications." 2nd ed.
(Chapter 5.1: Vehicle Costs)
Used to estimate vehicle
ownership and
maintenance costs per
capita, which is one
component of annual per-
capita transportation
related costs incurred by
residents.
Best available source that
disaggregates fuel- and non-
fuel-related vehicle costs
and reduces purchase costs
to an average per year
instead of a one-time
expense every several years.
MEDIUM: The source uses non-local
data from 2007.
317
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Parking cost of average
trip (calibration)
DCHCMPO. 2013. "Triangle Regional
Model version 5: Socioeconomic data
and projections for the preferred
growth scenario shapefiles." (TRM);
TRM Service Bureau and TRM Team.
2012. 'Triangle Regional Model
Version 5 Model Documentation
Report."
Affects modal person miles
through elasticities. Modal
person miles affect
household transportation
costs, traffic accidents,
congestion, fuel
consumption, and health
outcomes.
Best available source that
quantifies the average cost
per trip of parking; spatial
nature allows clipping to
both Tiers.
Peak period percent of
VMT
DCHCMPO. 2013. "Triangle Regional
Model version 5: Travel demand result
shapefiles." (TRM)
Determines how much of
the modeled VMT is taken
to be during peak periods,
which drives congestion.
Congestion decreases VMT
and increases the use of
public transit and
nonmotorized modes.
CIS format, highly detailed,
with per-link variables for
traffic volume, capacity, and
freeflow and peak-period
VMT and travel speeds; the
source is the official travel-
behavior model for the
Durham-Chapel Hill-Carrboro
Metropolitan Planning
Organization.
MEDIUM-LOW: Source does not
account for the opportunity cost that
is felt when parking is free yet scarce.
Processed data used here assumes
that the only source of change in
"parking cost of average trip" during
2010-2040 is the proportion of trips
that end in existing paid-parking
areas, as opposed to new paid-parking
areas being created or prices in the
existing paid-parking areas increasing
at a rate that is not equal to the
inflation rate.
MEDIUM: By using this input and
determining the value of this variable
through a lookup table, we assume
that it is entirely exogenous, even
though there likely are additional
factors within the model that would
cause people to do more or less of
their driving during peak periods.
318
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Percent of person miles
of public transit travel
that is on transit
systems entirely
contained within the
DCHC MPO
FTA. 2015. National Transit Database.
(NTD, 2013 value held constant going
forward); Green, Jennifer. 2014.
"FY2014 TTA Stats by County.xlsx."
(data on the percent of Triangle
Transit boardings attributed to any
given county; Triangle Transit is now
known as GoTriangle)
Used in Tier 2 only; helps
determine the percent of
public transit travel
starting or ending in area
that is by residents, which
is used with projections of
overall public transit use to
determine public transit
use by residents, which
determines the public-
transit-fare component of
household transportation
spending. Also, resident
public transit travel
increases resident
nonmotorized travel,
which drives health
outcomes.
The NTD provides detailed
service and usage data on
most of the public transit
systems in the U.S. However,
additional data from
Jennifer Green at Triangle
Transit (now known as
GoTriangle) was needed to
estimate how many of their
person miles are in the DCHC
MPO.
MEDIUM-HIGH: We assume that the
value of this variable remains
constant after 2013.
Percent of VMT that is
on highways
DCHC MPO. 2013. "Triangle Regional
Model version 5: Travel demand result
shapefiles." (TRM, 2010 value held
constant going forward)
Used to calculate VMT per
highway lane mile, which
helps determine the
change in person miles of
public transit travel per
year due to adding fixed-
guideway transit. Greater
public transit use reduces
VMT (and hence
congestion, fuel
consumption, and traffic
accidents) and helps drive
nonmotorized travel, and
hence physical activity.
The TRM is the primary
source of VMT and traffic
congestion projections used
by local transportation
planning agencies; spatial
nature allows clipping to
both Tiers.
MEDIUM: We assume that the value of
this variable remains constant over
time, though it is likely affected by
other variables in the model.
319
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Person miles of
automobile driver
travel per day;
through-traffic VMT;
person miles of
automobile passenger
travel per day; person
miles of nonmotorized
travel per day
(calibration)
(1) CAMPO and DCHC MPO. 2013.
"2040 Metropolitan Transportation
Plans." (MTP);
(2) DCHC MPO. 2013. "Triangle
Regional Model version 5: Travel
demand result shapefiles." (TRM);
(3) NuStats. 2006. Greater Triangle
Travel Study, Household Travel Survey
Final Report, (conducted for the
TRM);
(4) ESRI Community Analyst. 2014.
"ACS 1 -yr and 5-yr Estimates 2005-
2013."
Person miles of automobile
driver travel per day and
through-traffic VMT drive
VMT; person miles of
automobile driver and
passenger travel per day
determine average vehicle
occupancy; person miles of
nonmotorized travel per
day determines walking
and cycling activity by
residents and associated
health benefits.
(1) The MTP is the
authoritative transportation
planning document in the
DCHC MPO.
(2) The TRM is the primary
source of travel-behavior
projections for the MTP;
spatial nature allows clipping
to the Tiers.
(3) The Household Travel
Survey Final Report is one of
the TRM's key inputs.
(4) Community Analyst is
able to clip ACS data from
the U.S. Census Bureau to
specific geographic areas.
Tier 2 = MEDIUM; Tier 1 = LOW: A long
sequence of processing steps was
required in order to convert the
modal trip counts in the MTP into
person-mile counts, as well as to
scale the data to Tier 1, requiring
several assumptions that may not be
correct. For example, we used
average carpool sizes from the ACS,
which are specific to journey-to-work
trips, as opposed to travel for all
purposes. In addition, for Tier 1
scaling, we assume that there is a
proportional relationship between the
percent of VMT that is through traffic
and the percent of VMT that is on
highways.
Person miles of
nonmotorized travel
generated by average
public transit trip
N/A
Used to estimate how
much nonmotorized travel
is increased by public
transit travel, through
people walking or biking to
and from public transit
stops. Nonmotorized travel
affects health outcomes.
No source available. Input
value is in keeping with
general rules of thumb
employed by public transit
planners and researchers.
MEDIUM-LOW: No source available,
but input value is in keeping with
general rules of thumb employed by
public transit planners and
researchers.
320
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Person miles of public
transit travel per day
(calibration)
Historical Values (Tier 2 only):
(1) FTA. 2015. National Transit
Database. (NTD)
(2) Green, Jennifer. 2014. "FY2014
TTA Stats by County.xlsx."
2010 (Tier 1 only) and Projected
Values (Tier 2 only):
(3) CAMPOand DCHCMPO. 2013.
"2040 Metropolitan Transportation
Plans." (MTP)
(4) DCHCMPO. 2013. 'Triangle
Regional Model version 5: Travel
demand result shapefiles." (TRM)
(5) NuStats. 2006. Greater Triangle
Travel Study, Household Travel Survey
Final Report, (conducted for the TRM)
(6) ESRI Community Analyst. 2014.
"ACS 1 -yr and 5-yr Estimates 2005-
2013."
Used to estimate public
transit ridership and the
amount of additional
nonmotorized travel that
results from traveling to
and from public transit
stops. Greater public
transit use reduces VMT,
and hence congestion, fuel
consumption, and traffic
accidents. Greater
nonmotorized travel drives
physical activity, with its
associated health benefits.
(1) The NTD provides
detailed service and usage
data on most of the public
transit systems in the U.S.,
provided by public transit
agencies themselves to the
federal government
(supplemented by data from
personal communication
with (2) Jennifer Green at
Triangle Transit, now known
as GoTriangle).
(3) The MTP is the
authoritative transportation
planning document in the
DCHC MPO.
(4) The TRM is the primary
source of travel-behavior
projections for the MTP.
(5) The Household Travel
Survey Final Report is one of
the TRM's key inputs.
(6) Community Analyst is
able to clip ACS data from
the U.S. Census Bureau to
specific geographic areas.
Tier 2 = MEDIUM-HIGH (Historical =
High; Projected = Medium)
Tier 1 = LOW (Historical only)
A long sequence of processing steps
was required in order to convert the
modal trip counts in the MTP into
person-mile counts, as well as to
scale the data to Tier 1, requiring
several assumptions that may not be
correct. For example, to convert the
MTP's weekday-only transit-use
projections to an average of all seven
days of the week, we held the NTD's
reported 2013 ratio of weekend-day
transit use to weekday transit use
constant until 2040.
Public transit fare price
FTA. 2015. National Transit Database.
(NTD, 2013 value held constant going
forward); supplemented by: Green,
Jennifer. 2014. "FY2014 TTA Stats by
County.xlsx." (data on the percent of
Triangle Transit boardings attributed
to any given county, obtained so that
that transit agency's ridership in the
study area could be compared to that
of other transit agencies in the study
area to arrive at a weighted average
of their fare prices; Triangle Transit is
now known as GoTriangle)
Helps determine person
miles of public transit
travel per day, through an
elasticity; directly affects
public transit fare revenue
per year. Greater public
transit use reduces VMT
(and hence congestion,
fuel consumption, and
traffic accidents) and helps
drive nonmotorized travel,
and hence physical
activity.
The NTD provides detailed
service and usage data on
most of the public transit
systems in the U.S., provided
by public transit agencies
themselves to the federal
government. However,
additional data from
Jennifer Green at Triangle
Transit was needed to
estimate how many of their
boardings are in the DCHC
MPO, in order to create a
weighted average.
MEDIUM-HIGH: We assume that the
value of this variable remains
constant after 2013, though it may
change in response to changes in
ridership or operation costs.
321
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Public transit nonfuel
expenditure per road
based VMT
FTA. 2015. National Transit Database.
(NTD, 2013 value held constant going
forward); supplemented by: Green,
Jennifer. 2014. "FY2014 TTA Stats by
County.xlsx." (data on the percent of
Triangle Transit revenue miles
attributed to any given county;
Triangle Transit is now known as
GoTriangle)
Used to estimate one
component of annual
public transit operating
expenditure.
The NTD provides detailed
service, usage, and
operations data on most of
the public transit systems in
the U.S., provided by public
transit agencies themselves
to the federal government.
Additional data from
Jennifer Green at Triangle
Transit was needed to
estimate how many of their
revenue miles and VMT are
in the DCHC MPO.
MEDIUM-HIGH: The calculation in the
model subtracts an estimate of
public-transit-vehicle fuel (based on
NTD fuel-consumption data and
Energy Information Administration
figures for gasoline and diesel prices
at a national level) from NTD
operating expenditure figures, so it
combines local and non-local data.
Public transit road-
based vehicle revenue
miles per day; ratio of
VMT to revenue miles
for public transit road
based vehicles
(1) FTA. 2015. National Transit
Database. (NTD, 2013 value held
constant going forward)
(2) Green, Jennifer. 2014. "FY2014
TTA Stats by County.xlsx." (data on
the percent of Triangle Transit
revenue miles attributed to any given
county; Triangle Transit is now known
as GoTriangle);
(3) Triangle Transit, Chapel Hill
Transit, Orange County, DCHC MPO,
and Orange County Public
Transportation. 2012. "The Bus and
Rail Investment Plan in Orange
County.";
(4) DCHC MPO, Triangle Transit Board
of Trustees, and Durham Board of
County Commissioners. 2011. 'The
Durham County Bus and Rail
Investment Plan."
Revenue miles influence
person miles of public
transit travel per day
through an elasticity.
Greater public transit use
reduces VMT (and hence
congestion, fuel
consumption, and traffic
accidents) and helps drive
nonmotorized travel, and
hence physical activity.
The product of public
transit road-based vehicle
revenue miles per day and
the ratio of VMT to
revenue miles for public
transit road based vehicles
determines road-based
public-transit VMT, fuel
use, and operating costs.
The NTD provides detailed
service and usage data on
most of the public transit
systems in the U.S., provided
by public transit agencies
themselves to the federal
government (supplemented
by data from personal
communication with Jennifer
Green at Triangle Transit,).
County Bus and Rail
Investment Plans are
authoritative planning
documents for public transit
spending.
MEDIUM-HIGH: The input comes from
an authoritative source, but we
needed to convert gross revenue-hour
figures from the county bus and rail
investment plans to road-based
revenue miles by assuming the NTD's
2013 ratio of road-based revenue
miles per revenue hour will hold
constant.
322
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Public transit trip
distance
FTA. 2015. National Transit Database.
(NTD, 2013 value held constant going
forward); supplemented by: Green,
Jennifer. 2014. "FY2014 TTA Stats by
County.xlsx." (data on the percent of
Triangle Transit boardings attributed
to any given county; Triangle Transit
is now known as GoTriangle)
Reaction times of
modal person miles and
vehicle trip distance to
(1) vehicle speed, (2)
fuel cost, (3) parking
cost, (4) population
density, (5) jobs-
housing balance, (6)
intersection density,
(7) public transit fares,
and (8) public transit
revenue miles.
N/A
Used to estimate daily
public transit passenger
trips from person miles of
public transit travel per
day. The number of public
transit trips drives public
transit fare revenue and
nonmotorized travel to and
from transit stops, which
contributes to physical
activity, with its attendant
health benefits.
Used to determine effects
on person miles of travel
by mode and vehicle trip
distance. Modal person
miles affect household
transportation costs,
traffic accidents,
congestion, fuel
consumption, and health
outcomes. VMT by
residents is divided by
vehicle trip distance to
arrive at vehicle trips per
resident, which determines
parking spending per
resident.
The NTD provides detailed
service, usage, and
operations data on most of
the public transit systems in
the U.S., provided by public
transit agencies themselves
to the federal government.
Additional data from
Jennifer Green at Triangle
Transit was needed to
estimate how many of their
boardings and person miles
are in the DCHC MPO.
MEDIUM-HIGH: The input comes from
an authoritative source and does not
require any assumptions to apply to
the model, except that we assume
that the value of this variable remains
constant over time.
Sources of elasticity values
for these relationships did
not indicate specific
amounts of time that it takes
for a given output to react to
changes in a given input.
Indications were only
provided of whether the
elasticities were "long-term"
or "short-term."
LOW: No source available. Instead,
inputs are based on system-dynamics
rules of thumb for what to consider
"long-term" versus "short-term"
elasticities.
323
-------
MODEL INPUT
Tier 1 to Tier 2
weighting factors for
effect of (1 ) congestion
(2) vehicle speed, (3)
fuel cost, and (4)
parking cost on Tier 1
desired vehicle
ownership per person
not in a zero-car
household, vehicle trip
distance, through-
traffic VMT, and modal
person miles (separate
values for automobile
driver, automobile
passenger,
nonmotorized, and
public transit travel).
Urban nonhighway lane
miles disrupted per LRT
line mile under
construction
SOURCE
N/A
N/A
HOW INPUT IS USED IN THE
MODEL
Used to determine how
much changes in Tier 1 and
Tier 2 (1 ) traffic
congestion (2) average
vehicle speed, (3) fuel
cost, and (4) parking cost
affect Tier 1 person miles
of travel by mode, VMT,
trip distances and vehicle
ownership. These outputs
affect household
transportation costs,
traffic accidents,
congestion, fuel
consumption, and health
outcomes.
Used to estimate
reductions in functioning
roadway lane miles
resulting from light rail
line construction.
Temporarily reducing
functioning lane miles
increases traffic
congestion, which reduces
VMT and increases the use
of other travel modes.
Modal person miles affect
household transportation
costs, traffic accidents,
congestion, fuel
consumption, and health
outcomes.
RATIONALE FOR SELECTING
SOURCE
Sources were not available
that estimated how much
Tier 1 travel behavior and
vehicle ownership are
affected by neighborhood-
scale (Tier 1 ) congestion,
vehicle speed, fuel cost, and
parking cost, relative to the
metropolitan-area scale
(Tier 2). We decided that
both of these scales were
important for the
relationships discussed here.
No source available. Used an
assumed value.
CONFIDENCE LEVEL NOTES
LOW: No source available.
LOW: No source available.
324
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Vehicle lifespan
International Energy Agency,
Directorate of Sustainable Energy
Policy and Technology. 2009.
Transport, Energy, and €62: Moving
Toward Sustainability. (Table 3.1)
Each addition to the stock
of vehicles is eventually
removed (unless the end of
the model run intervenes)
after the average vehicle
lifespan. Vehicle stock
helps drive household
transportation costs and
vehicle registration fee
revenue.
The International Energy
Agency is an authoritative
source; the result is
consistent with information
found in other sources.
MEDIUM-HIGH: Data are not specific
to the study area (so we assume that
average vehicle lifespans in the study
area are the same as national
averages) and are drawn from the
period 2000-2005.
Vehicle stock
(calibration)
Geographic Research Inc. 2015.
"Census Data 2000, ACS Estimates
2008-2014." (data by Census Block
Group)
Multiplied by vehicle
ownership and
maintenance costs to form
one component of
transportation costs
incurred by residents; also
drives vehicle registration
fee revenues.
Chosen because it provides
multiple years' data up to
2014. Unlike the Smart
Location Database (which
only has one year's data),
counts households with 0, 1,
2, 3, or 4+ vehicles, as
opposed to just households
with 0, 1, or 2+ vehicles.
MEDIUM-HIGH: Only counts vehicles
belonging to households. Clipping
Census Block Groups to the Tiers
assumes that vehicles are evenly
distributed within any given block
group. Does not count any vehicles in
a household beyond four.
Vehicle trip distance
(calibration - Tier 2
only)
(1) CAMPO and DCHC MPO. 2013.
"2040 Metropolitan Transportation
Plans." (MTP);
(2) DCHC MPO. 2013. "Triangle
Regional Model version 5: Travel
demand result shapefiles." (TRM);
(3) NuStats. 2006. Greater Triangle
Travel Study, Household Travel Survey
Final Report, (conducted for the
TRM);
(4) ESRI Community Analyst. 2014.
"ACS 1 -yr and 5-yr Estimates 2005-
2013."
Used along with VMT by
residents to determine the
number of resident vehicle
trips. The number of
vehicle trips by residents
and the average cost of
parking determine overall
and parking expenditures,
a component of household
transportation costs.
Same collection of sources
and data processing steps
used to arrive at data for
person miles of travel by
mode per day, which helps
impart consistency.
(1) The MTP is the
authoritative transportation
planning document in the
DCHC MPO.
(2) The TRM is the primary
source of travel-behavior
projections for the MTP.
(3) The Household Travel
Survey Final Report is one of
the TRM's key inputs.
(4) Community Analyst is
able to clip ACS data from
the U.S. Census Bureau to
specific geographic areas.
MEDIUM: A long sequence of
processing steps was required in order
to convert the average distance of all
person trips reported in the MTP into
average person-trip distances by
mode, requiring several assumptions
that may not be correct. For
example, we used average carpool
sizes from the ACS, which are specific
to journey-to-work trips, as opposed
to travel for all purposes. We also
assumed that the ratios between
average trip distances by different
modes reported in the Household
Travel Survey could be held constant
going forward for the purpose of
calculating vehicle trip distances to
calibrate to.
325
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
VMT (calibration)
DCHCMPO. 2013. "Triangle Regional
Model version 5: Travel demand result
shapefiles." (TRM)
Used to estimate traffic
congestion, fuel
consumption by vehicles,
and traffic accidents.
The TRM is the primary
source of VMT and traffic
congestion projections used
by local transportation
planning agencies; spatial
nature allows clipping to
both Tiers.
HIGH: Authoritative source with
straightforward application to the
study area, but the TRM only models
weekday traffic. We assume that VMT
on a weekend day is the same as VMT
on a weekday.
326
-------
TABLE C-3. MODEL INPUT CHARACTERIZATION FOR THE ENERGY SECTOR
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Carbon dioxide (CO2>
emissions (calibration)
(1) Durham City-County Sustainability
Office. 2015.
(2) Freid, Tobin. Email message to
authors on January 9, 2015.
CC>2 emissions is an
endpoint indicator
variable.
Authoritative local source;
emissions are calculated by
the Sustainability Office
based on energy data
supplied by utility
companies.
MEDIUM: For buildings, only
emissions from electricity and
natural gas (which represent the
large majority of energy use in
buildings) are currently tracked.
Building energy use
(calibration)
(1) Durham City-County Sustainability
Office. 2015.
(2) Freid, Tobin. Email message to
authors on January 9, 2015.
Used to estimate CO2
emissions and energy
costs.
Authoritative local source;
energy data is supplied by
utility companies.
MEDIUM: For buildings, only
emissions from electricity and
natural gas (which represent the
large majority of energy use in
buildings) are currently tracked.
We assume that the data collection
methods used by the companies
(their definition of county
boundaries, etc.) are consistent
year-to-year.
Energy use intensity -
buildings (calibration)
(1) US EIA. 2009. "Residential Energy
Consumption Survey (RECS)."
(2) Durham County Property Tax
Database. 2000-2014.
(3) US DOE. 2008. "Energy Efficiency
Trends in Residential and Commercial
Buildings."
(4) US EIA. 2015. "Annual Energy
Outlook 2015."
Light rail electricity use
FTA and CATS. 2011. "Lynx Blue Line
Extension, Northeast Corridor Light
Rail Project, Charlotte-Mecklenburg
County, North Carolina, Final
Environmental Impact Statement."
Historical and projected
trends in building energy
use intensity are
multiplied by local data on
housing stock and building
square footage to
calculate building energy
use. A calibration factor is
applied to better fit
building energy use to
historical data.
EIA and DOE are
authoritative national
sources for energy data and
projections. Durham County
Property Tax Database is a
comprehensive and detailed
record of buildings in Durham
County.
MEDIUM. We assume that regional
or national trends are appropriate
for the study area.
Used to estimate CO2
emissions and energy costs
of the light rail.
This environmental impact
statement is for light rail in
the same state as the D-O
LRP.
MEDIUM. We assume that the D-O
LRP rail will consume electricity at
the same rate per vehicle mile
traveled as the Charlotte Lynx Blue
Line.
327
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Energy prices: gasoline,
diesel, natural gas,
electricity
(1) US EIA. 2015. "Weekly Retail
Gasoline and Diesel Prices."
(2) US EIA. 2015. "Annual Energy
Outlook 2015."
(3) US EIA. 2015. "North Carolina
Price of Natural Gas Delivered to
Residential Customers."
(4) US EIA. 2015. "Electricity: Sales
(consumption), revenue, prices &
customers."
Used to calculate historical
and projected energy
spending
US EIA is an authoritative
national source on historical
and projected fuel prices.
MEDIUM-HIGH: Historical prices:
HIGH. Future prices: MEDIUM.
Projections are inherently
uncertain.
CC>2 emissions factors
US EPA. 2015. "Clean Energy,
Calculations and References."
Used to calculate CO2
emissions by fuel type
Authoritative national source
on pollutant emissions
factors.
MEDIUM-HIGH: Authoritative source
with straightforward application to
the model, but local emission
factors (such as for electricity
generation) may differ from the
regional average.
PM2.s emissions per
VMT; NOx emissions per
VMT
(1) Cai et al. 2013. "Updated Emission
Factors of Air Pollutants from Vehicle
Operations in GREET™ Using MOVES."
(GREET = Greenhouse gases,
Regulated Emissions, and Energy use
in Transportation model. MOVES =
Motor Vehicle Emission Simulator)
(2) Jackson, Tracie R. 2001. "Fleet
Characterization Data for MOBILE6:
Development and Use of Age
Distributions, Average Annual Mileage
Accumulation Rates and Projected
Vehicle Counts for Use in MOBILE6."
Used to calculate PM2.5 and
NOX emissions from
passenger vehicles, which
affect premature
mortality.
In the absence of local data
for vehicle fleet emissions,
we chose to use the most
recent available data on
average PM2.s and NOX
emissions per VMT by vehicle
model year (1) and weight
each model year by the only
U.S. vehicle fleet
distribution data by vehicle
age that we could find (2).
MEDIUM: MEDIUM-HIGH for
historical data since the two studies
are peer-reviewed and highly
regarded, but we assume that
national averages are appropriate
for the study area. MEDIUM-LOW for
projections because Cai et al. (1)
shows PM2.s emissions per VMT flat-
lining for vehicle model years after
2010, which may not be accurate
based on more recent vehicle
emissions reductions, and the study
only goes until 2020, so emissions
per VMT for NOxand PM2.5 were
extrapolated to 2040.
Vehicle fuel efficiency
(MPG)
US EIA. 2015. "Annual Energy Outlook
2010-2015."
Used to calculate
passenger vehicle energy
use, fuel costs, and
emissions
Authoritative national source
for energy data and
projections.
MEDIUM: Projections are inherently
uncertain.
328
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING
SOURCE
CONFIDENCE LEVEL NOTES
Additional fuel
consumed per hour of
congestion delay
Sinha, Kumares Chandra, and Samuel
Labi. 2007. Transportation Decision
Making: Principles of Project
Evaluation and Programming, (page
389, citing FHWA)
Used to determine how
much additional vehicle
fuel is consumed as a
result of the reduced fuel
efficiency that results
from traffic congestion.
Cites an authoritative source
and uses inputs that are
easily calculated in the
model.
MEDIUM: Does not account for
changes in vehicle technology (that
might affect fuel efficiency) over
time.
Ratio of fuel
consumption per public
transit road-based VMT
over gasoline gallon
equivalent consumption
per non-public transit
VMT
(1) FTA. 2015. National Transit
Database. (NTD, 2013 value held
constant going forward);
supplemented by: Green, Jennifer.
2014. "FY2014 TTA Stats by
County.xlsx." (data on the percent of
Triangle Transit revenue miles
attributed to any given county;
Triangle Transit is now known as
GoTriangle);
(2) US EIA. 2015. "Annual Energy
Outlook 2010-2015."
Used to estimate public
transit vehicle diesel and
gasoline consumption per
VMT (of buses and
demand-res ponse/vanpool
vehicles, respectively),
after accounting for
different fuel efficiencies
by vehicle type.
(1) The NTD provides
detailed service, usage, and
operations data on most of
the public transit systems in
the U.S., provided by transit
agencies themselves to the
federal government
(supplemented by data from
personal communication with
Jennifer Green at Triangle
Transit).
(2) The US EIA is an
authoritative source on fuel
efficiency.
MEDIUM-HIGH: The input comes
from authoritative sources, but we
implicitly assume that local public
transit vehicle fuel efficiency will
increase along a similar trend to
U.S. light-duty-vehicle fuel
efficiency.
Solar capacity
NC Sustainable Energy Association.
2015. "Installed solar projects in NC.
Used to determine the
historical trend of solar
electricity installation, and
its effect on CC>2 emissions
in Tier 2.
NCSEA has a comprehensive
state-wide database of solar
energy installations, which
we used to calculate solar
capacity in Tier 2.
MEDIUM-HIGH: The source is
comprehensive and regularly
curated. Data are available by
county, and the sum of solar
capacity in Durham and Orange
county was taken to approximate
the solar capacity of Tier 2.
LFG energy capacity
(1) Duke Energy. 2008. "Duke Energy
Carolines signs deal to turn landfill
gas into energy."
(2) City of Durham. 2009. "Durham
landfill gas-to-energy green power
project."
Used to determine the
effect of current
renewables on CO2
emissions in Tier 2
Duke Energy is a major
electricity utility in the
region. The City of Durham is
a partner in this LFG project,
and reference (2) is a press
release.
MEDIUM: An alternate source states
that the landfill will generate 3MW
rather than 2MW:
http://icma.org/en/icma/knowledg
e_network/documents/kn/Docume
nt/102318/Durham_Landfill_GastoE
nergy_Green_Power_Project
329
-------
TABLE C-4. MODEL INPUT CHARACTERIZATION FOR THE ECONOMY SECTOR
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Total employment
(calibration)
(1) Tier 2: Historical (2000-2009)
employment growth rate calculated
from: BEA. 2014. "Local Area
Personal Income and Employment:
Table CA25N, Total Full-Time and
Part-Time Employment by NAICS
Industry, 2001-2013, for Durham and
Orange County, NC."
(2) Tier 2: Historical data (2010) and
projections (2011-2040) from: DCHC
MPO. 2013. "Triangle Regional Model
version 5: Socioeconomic data and
projections for the preferred growth
scenario."
(3) Tier 1: Historical (2002-2009)
employment growth rate calculated
from: U.S. Census Bureau. 2015.
"LODES data. Longitudinal Employer-
Household Dynamics Program."
(4) Tier 1: Historical data (2010) and
projections (2011-2040) from: DCHC
MPO. 2013. "Triangle Regional Model
version 5: Socioeconomic data and
projections for the preferred growth
scenario."
Total employment is one
of the main drivers of
economic growth in the
model, generating
earnings, the largest
component of GRP, which
has cascading impacts
throughout the model.
(1) The BEA is a prominent
government source for historical
economic data and updates historical
employment data regularly.
(3) Only source of employment data
available at a small enough
geographic scale for Tier 1.
(2) and (4) TRM v5 SE data total
employment are the benchmark we
wanted to match because these
employment numbers were used to
generate travel demand in the TRM
v5. In addition, the data are output
by traffic analysis zones (TAZs) which
are small enough in the proposed light
rail station areas to use for Tier 1.
HIGH for Tier 2: (1) The most
accurate, up-to-date source
for total employment data by
county, which was verified to
be a good surrogate for Tier 2
by calculating total
employment from TRM v5 SE
data for Durham and Orange
County, which was less than
1% different than total
employment for Tier 2. (2)
HIGH for historical data (2010)
and MEDIUM-HIGH for
projections due to the
inherent uncertainty
associated with future
projections.
MEDIUM-HIGH for Tier 1: (3)
LODES data has built-in noise
that distorts the data on small
scales, but total employment
values were mostly consistent
year-to-year. (4) HIGH for
historical data (2010) in Tier
1, but MEDIUM-HIGH for the
projections because, although
the regional employment
growth projections were
estimated by local planners, a
model (CommunityViz) was
used to allocate that
employment growth to smaller
geographies (TAZs) based on a
subjective measure of the
attractiveness of those areas.
330
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Shares of total
employment by
employment category
(industrial, office,
retail, service)
(1) Tier 2: Historical data (2000-
2011) and projections (2012-2040)
from: Woods & Poole Economics, Inc.
Copyright 2014. "Durham and Orange
County, NC Data Pamphlet."
(2) Tier 1: Historical (2002-2009):
U.S. Census Bureau. 2015. "LODES
Data. Longitudinal Employer-
Household Dynamics Program."
(3) Tier 1: Historical data (2010) and
projections (2011-2040) from: DCHC
MPO. 2013. 'Triangle Regional Model
version 5: Socioeconomic data and
projections for the preferred growth
scenario."
Multiplied by total
employment in the model
to determine the total
number of jobs by
employment category,
which are used to
calculate "total earnings,"
a component of "GRP,"
and are multiplied by
employee space ratios for
each employment category
to determine "total
nonresidential sq ft."
(1) Woods & Poole's historical data
are from the BEA, but they fill in gaps
in certain employment categories that
the BEA omits. Projections were used
by the DCHC MPO to help create
employment guide totals for the
CommunityViz modeling that
generated the output for the TRM v5
SE data.
(2) and (3) Only sources available for
historical and projected employment
by category at a small enough
geographic scale for Tier 1.
MEDIUM-HIGH for Tier 2: HIGH
for historical data since its
original source is the BEA, and
MEDIUM-HIGH for projections
due to the inherent
uncertainty associated with
future projections.53
MEDIUM-LOW for Tier 1: LODES
data have built-in noise that
distorts the data on small
scales, and changes in census
block group geographies
between 2002 and 2009 also
caused inconsistencies in the
data in Tier 1. Also, TRM v5 SE
data's definitions of
employment categories
differed from those used in
the model.
Percent of jobs
earning $3,333 per
month in 2010 USDs
Tier 1
Historical (2010): U.S. Census
Bureau. 2015. "LODES Data.
Longitudinal Employer-Household
Dynamics Program."
Used to determine the
change in person miles of
public transit travel per
year due to adding fixed-
guideway transit.
LODES is based on Census Bureau
data, clipped to the study area.
MEDIUM-LOW: We assume that
this source was using 2010
USDs but were unable to verify
this assumption. We also
assume that this value remains
constant over time, though it
is likely to change in response
to other factors in the model.
53 Woods & Poole does not guarantee the accuracy of this data. The use of this data and the conclusions drawn from it are solely the responsibility of the US
EPA.
331
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Ratio of
entertainment jobs to
retail jobs Tier 1
Historical (2010): U.S. Census
Bureau. 2015. "LODES Data.
Longitudinal Employer-Household
Dynamics Program."
Is a component of retail
plus entertainment
employment in Tier 1,
which is used to determine
the change in person miles
of public transit travel per
year due to adding fixed-
guideway transit.
LODES is based on Census Bureau
data, clipped to the study area.
MEDIUM: Entertainment jobs
are technically a subset of
service employment, but we
estimate them using a ratio to
retail employment. In
addition, we assume that this
ratio remains constant over
time.
Earnings per
employee by
employment category
Historical data (2000-2011) and
projections (2012-2040) from: Woods
& Poole Economics, Inc. Copyright
2014. "Durham and Orange County,
NC Data Pamphlet."
Multiplied by employment
by category to estimate
the total earnings by
category, which is then
summed to yield total
earnings.
Historical earnings data (2000-2011)
from Woods & Poole is from the U.S.
Bureau of Economic Analysis (BEA),
but fills in gaps in certain earnings
categories that the BEA omits. No
earnings per employee by
employment category data were
available at the Tier 1 level, so
earnings by category for Orange and
Durham County were used, weighted
by the number of Tier 1 jobs in each
county.
HIGH for Tier 2: The historical
data are reported by
employers, not employees,
and the projections in real
dollars are reasonable when
compared with historical
trends.
MEDIUM-LOW for Tier 1: Using
county-level earnings per
employee estimates likely
overestimates earnings per
employee for industrial jobs in
Tier 1, because the North
American Industrial
Classification System (NAICS)
considers high-wage jobs in
RTP to be manufacturing.
Percent of residential
population in labor
force
(1) Tier 2: NC ESC. 2014. "Local Area
Unemployment Statistics, 2000-2014
for Durham and Orange County, NC.
(2) Tier 1: Labor force statistics by
block group from: Geographic
Research Inc. 2015. "Census Data
2000, ACS Estimates 2008-2014."
Accessed from SimplyMap Database.
Multiplied by total
population to determine
the civilian labor force in
Tier 2 and Tier 1, which is
used to determine the
unemployment rate of
residents.
(1) Best available source that had
county-level data for all years
covered by the model (U.S. Bureau of
Labor and Statistics only had data
from 2004 onward).
(2) Only labor force data at small
enough geographies for Tier 1 (census
block group) and Simply Map provides
the data from both sources in one
convenient download.
HIGH: Both historical data
sources have high accuracy
and reliability. The only small
source of uncertainty is for
Tier 1 because the ACS warns
that the 5-year average data
may not be comparable to the
data from the decennial
census, though Simply Map
accounts for this in their
methodology.
332
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Percent of
employment held by
commuters (Tier 2
only)
(1)NCESC. 2014. "Local Area
Unemployment Statistics, 2000-2014
for Durham and Orange County, NC."
(2) Calculated "total employment"
(see above).
Used to determine the
number of people that
work AND live in the
region, which is used to
determine the
unemployment rate of
residents.
(1) Best available historical data
source for unemployment statistics.
(2) No single data source gave
employment statistics for the number
of jobs held by residents of a
particular area, so total employment
was used.
MEDIUM: While confidence in
the historical data for
employed residents is high, we
divide these historical data by
our calculated total
employment data to
determine the percent of
employment held by
commuters. Assumptions that
introduce uncertainty in this
calculation are: (1) that all
employed residents work
within the study area, and (2)
that all employed residents
hold only one job. Since our
total employment numbers
include both part-time and
full-time employment, it is
likely that some residents hold
multiple jobs, thus the
percent of employment held
by commuters is likely over-
estimated.
333
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Total retail
consumption
(initial value and
calibration)
(1) Tier 2: Historical data (2000-
2013) from: NC DOR. 2013. Table
37A. State Sales and Use Tax: Retail
Taxable Sales by County. 2014 value
from: NC DOR. 2014. "Gross Retail
Sales for Durham and Orange County,
FY2014."
(2) Tier 2: Historical retail sales data
(2000-2011) and projected retail
sales growth rate (2012-2040) from:
Woods & Poole Economics, Inc.
Copyright 2014. "Durham and Orange
County, NC Data Pamphlet."
(3) Tier 1: Historical data from:
Geographic Research Inc. 2015.
"Total Retail Sales (2008-2014) by
Census Block Group."
Used to determine
"desired employment"
which adds additional
employment yearly in the
model, and to determine
all sales tax revenues.
(1) and (2): The initial retail
consumption value used was in
between both sources of historical
retail sales data for 2000 since the
NCDOR data (total taxable sales)
slightly under-estimates total retail
sales, and total retail sales from
Woods & Poole slightly over-estimates
the amount for tax purposes
historically, though the two data
sources converged around 2010.
Woods & Poole was the only source of
retail sales projections.
(3) Retail sales data are only output
by the Economic Census by county,
but Simply Map estimates retail sales
by block group using a computer
model.
HIGH for Tier 2: (1) Historical
data for retail sales for
Durham and Orange County
are a very reasonable
substitute for Tier 2 since
almost all of the retail
businesses in Tier 2 are
located in those counties. (2)
Woods & Poole retail sales
growth rate projections are
modest compared to the high
growth rates in retail sales for
Durham and Orange County
over the past couple of years,
but this balances out with
other periods of slow growth
or decline over the past 15
years.
MEDIUM-HIGH for Tier 1: (3)
Retail sales data at the Tier 1
level (Simply Map) are
estimated by a computer
model, so it is not directly
from the source (businesses).
Elasticity of
consumption to GRP
N/A
Used to determine the
magnitude of the annual
increase of retail
consumption based on the
relative increase in GRP.
No particular source was available for
this elasticity, thus the value for both
Tier 2 and Tier 1 (1.1) was
determined through the calibration.
MEDIUM-HIGH: The elasticity
was calculated based on
historical data.
334
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING SOURCE
CONFIDENCE LEVEL NOTES
GRP
(calibration)
Historical GRP (2000-2014)
calculated using a top-down
methodology from: Panek, Sharon
D., et al. 2007. "Introducing New
Measures of the Metropolitan
Economy: Prototype GDP-by-
Metropolitan-Area Estimates for
2001 -2005," using NC State GDP
from: BEA. 2014. "Gross Domestic
Product (GDP) by State: GDP in
Current Dollars, 2000-2013, for North
Carolina." And earnings data from:
BEA. 2014. "Annual State Personal
Income and Employment: Table
SA05N - Personal Income by Major
Source and Earnings by NAICS
Industry, 2000-2013, for North
Carolina." Local earnings data
were calculated for Tier 2 and Tier 1
using "total employment," the
"share of jobs by employment
category," and "earnings per
employee by category," (see above).
GRP influences many
model variables, most
directly total person miles
(all modes), CC>2 emissions,
and stormwater pollutant
loadings.
The methodology provided allowed us
to calculated GRP based on our own
employment data and earnings
calculations, rather than using GRP
directly from the BEA or Woods &
Poole (See Appendix D: Data Sources
Not Used).
MEDIUM-HIGH: This
methodology is used by the
BEA to calculate GDP (GRP) by
county, which is very difficult
to calculate based on a
"bottom-up" approach. For
the purposes of this model,
the GRP we calculated is the
best possible fit.
Elasticity of GRP to
energy spending
Bassi, Andrea M., Robert Powers,
and William Schoenberg. 2010. "An
integrated approach to energy
prospects for North America and the
rest of the world." Energy Economics
no. 32 (1): 30-42.
Used to determine the
magnitude of the influence
of relative increases in
energy spending on "gross
operating surplus," which
is one component of GRP.
Most relevant source for the data
needed and gives a range of values
for the magnitude of the elasticity.
MEDIUM: The range for this
value given in the report was
0.1-0.3, so we chose 0.2, but
the actual impact for our
study area is unknown.
335
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING SOURCE
CONFIDENCE LEVEL NOTES
GRP deflator 2010
(1) Historical (2000-2014): BEA.
2015. "National Income and Product
Accounts Tables: Section 2 - Table
2.3.4. Price Indexes for Personal
Consumption Expenditures (PCE),
Annual Data for 2000-2014."
(2) Projected (2015-2040): Woods a
Poole Economics, Inc. Copyright
2014. "Durham and Orange County,
NC Data Pamphlet."
Used to convert 2010
constant dollars to current
(nominal) dollars, and was
also used for all data
sources to convert
historical dollar values
that were not adjusted for
inflation to 2010 dollars.
The PCE index, rather than the CPI
index, is used by Woods & Poole for
historical data, though we
downloaded the more recent PCE
from the BEA website since it is
updated regularly.
MEDIUM-HIGH: (1) The PCE is a
national price index, but is a
widely used price index that
can apply to different
consumption categories,
making it ideal for our model
rather than using several
different price indexes. (2)
Woods & Poole projects future
inflation indices based on
historical trends among many
other predictive variables
making it the best available
source for a variable that has
a wide range of uncertainty in
the future.
Social insurance
contribution
(1) Tier 2: Woods a Poole
Economics, Inc. Copyright 2014.
"Durham and Orange County, NC
Data Pamphlet."
(2) Tier 1: BEA. 2014. "Local Area
Personal Income Methodology."
Reduces "earnings per
capita" along with a
residence adjustment to
determine "resident net
earnings per capita,"
which affects vehicle
ownership and property
values.
(1) Woods & Poole provide historical
data and projections of social
insurance contributions for Orange
and Durham County.
(2) No data was available at the Tier
1 level, so a national average from
2013 (11%) was used for the entire
study period from the BEA.
MEDIUM-HIGH: The percent
social insurance contribution
did not vary widely historically
for Durham an Orange County,
providing a reasonable
assumption that it will not
vary widely in the future,
however, in Tier 1 where per
capita income is lower, the
national average percent may
be slightly high.
336
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Percent of tax exempt
nonresidential
property value; total
nonresidential
property value
(calibration); total MF
property value
(calibration); total SF
property value
(calibration); total
commercial real
property value
(calibration)
(1) Total tax-exempt commercial
and residential (SF only) property
values from: NC DOR. 2004-2014.
"County Taxable Real Property
Valuations" for Durham and Orange
County.
(2) Breakdown of total commercial
property value into multifamily and
single-family housing from: "Durham
County Tax Administration Real
Property Tax Database." 2000-2014.
Used to determine the
amount of taxable real
property value in Tier 2
and Tier 1, which affects
property tax revenues.
The data from the NC DOR give %
residential and % commercial real
property value, but they do not divide
commercial into multifamily housing
(apartments) and nonresidential
property value. The Durham County
property tax database did not make it
easy to separate tax exempt and non-
tax exempt property, therefore, both
datasets were necessary to determine
the amount of non-residential
property value that was tax exempt.
MEDIUM-HIGH for Tier 2, with
the only sources of error being
the limited amount of
property value available for
Orange and Chatham County
(2014 only) and the use of
Orange County entire for the
combined portions of Orange
and Chatham County in the
DCHC MPO for the NC DOR
data. MEDIUM-LOW for Tier 1,
with quite a bit of uncertainty
around these calculations with
data unavailable at the Tier 1
level.
County property tax
rate; city property tax
rate; percent of
county real property
value in cities (Tier 2
only), total county
real property value
(calibration)
Durham County, Orange County, City
of Durham, and Town of Chapel Hill
Comprehensive Annual Financial
Plans (AFPs) for FY 2009 and 2014.
Used to determine
property tax revenues for
the county and city local
government budgets.
Combined sales tax
rate; local sales tax
rate; percent of
collected local option
sales tax distributed
(1) NC DOR. 2013. "Table 36A. State
Sales and Use Tax: Gross Collections
by County, FY 2000-2013."
(2) NC DOR. 2003-2013. "Table 56.
Summary of Local Sales and Use Tax
Collections, Tax Allocations, and
Distributable Proceeds by County."
Used to determine state
and local sales tax
collected and local sales
tax revenues received back
from those collections for
the county and city local
government budgets.
AFPs are the official city and county
financial documents, making them
the most reliable, direct source for
tax revenue information and each AFP
provided 10-year summary tables of
property tax information, requiring
only 2 documents to be downloaded
(2009 and 2014) to collect historical
data for 2000-2014.
MEDIUM-HIGH: HIGH for
historical property value and
tax rate data for Tier 2 and
MEDIUM for projected values
because tax rates are held
constant at 2014 values for
2015-2040 even though tax
rates are highly variable
historically.
Only NC DOR had data on the amount
of state and local sales and use tax
collected AND distributed.
HIGH: Authoritative
government source.
337
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Transit sales tax share
of local option sales
tax revenues
(1) NC DOR. 2013. "Table 60A.
Article 43 Local Government Sales
and Use Taxes for Public
Transportation for FY 2013."
(2) Triangle Transit. 2014. "Durham
and Orange County Bus and Rail
Investment Progress Reports for FY
2014."
Used to calculate "half
cent transit sales tax
revenues," a component of
"total LRP revenues,"
which is used to calculate
the D-O LRP budget.
NC DOR has not yet released data for
FY 2014, but Triangle Transit did
report the half cent transit tax
revenues for FY 2014, so FY 2013 was
checked for both sources to make
sure the values matched.
MEDIUM-HIGH: While
confidence is high for the
historical years of data
available (2013 and 2014),
2014 was the only full fiscal
year that the transit tax was
collected, thus only providing
one data point for forecasting
future values of this model
input.
County share of local
option sales tax
revenues; city share
of local option sales
tax revenues (both
Tier 2 only)
NC DST. 2015. "Local Government
Revenues by Source: Local Option
Sales Tax Revenues for Chapel Hill
and the City of Durham, FY 2000-
2014."
Used to divide local option
sales tax revenues into
county and city revenues
for the county and city
local government budgets.
Several different sources, including
county and city AFPs and annual
budgets, had conflicting values for
these percentages, but NC DST
reviews AFPs and budgets to compile
their statistics.
MEDIUM-HIGH: Authoritative
government source, however a
large amount of variability
historically makes projecting
these percentages less
certain.
Rental car tax
revenues (Tier 2 only)
(1) Triangle Transit. 2014. "Durham
and Orange County Bus and Rail
Investment Progress Reports for FY
2014."
(2) Triangle Transit et al. 2012. "The
Bus and Rail Investment Plan in
Orange County."
(3) DCHC MPO et al. 2011. "The
Durham County Bus and Rail
Investment Plan."
A component of "total LRP
revenues," which is used
to calculate the D-O LRP
budget.
(1) Only source for rental car tax
revenues collected for the D-O LRP in
2014 (the first year it was collected).
(2) and (3) Bus and rail investment
plans gave an expected annual growth
rate for rental car tax revenues
collected in both Durham and Orange
County (4.0%).
Nominal revenues per
vehicle (Tier 2 only)
(1) NC DOT. 2014. "Auto and Truck
Registrations in Durham County and
Orange County, 2014."
(2) Triangle Transit. 2014. "Durham
and Orange County Bus and Rail
Investment Progress Reports for FY
2014."
(3) Triangle Transit. 2015. "FY 2015
Annual Budget & Capital Investment
Plan: XI. Durham-Orange Bus and
Rail Investment Plan."
Used to calculate "vehicle
registration fee revenues,"
a component of "total LRP
revenues," which is used
to calculate the D-O LRP
budget.
(1) and (2) were used to calculate the
revenues collected for the year 2014,
and were the most authoritative
sources for both data needs. The
Triangle Transit 2015 budget (3) gave
an estimate of vehicle registration fee
revenues for Durham and Orange
County in 2015, the first full year in
which the fees were collected, with
one vehicle registration fee being $7
per vehicle and another being $3 per
vehicle, for a combined total of $10.
MEDIUM-HIGH: Only one year
of actual collected data was
available (2014), but Triangle
Transit does extensive budgets
using historical data.
HIGH: Calculated from
historical data for 2014 and no
reason to believe it will
change from $10 total per
vehicle between 2015 and
2040.
338
-------
TABLE C-5. MODEL INPUT CHARACTERIZATION FOR THE EQUITY SECTOR
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR
SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Property values
(calibration)
Durham County Tax Administration.
2000-2014. Durham County Tax
Administration Real Property
Database, Orange County Tax
Administration. 2014. Orange County
Parcel Database, Chatham County Tax
Administration Office. 2014. Chatham
County Tax Parcel Database.
Used to determine
affordability and property
tax revenue.
Most detailed sources;
used by Durham planning
department for their
comprehensive plan
HIGH: Authoritative source used for
property taxation.
Median annual renter
costs (calibration)
Geographic Research Inc. 2015.
"Census Data 2000, ACS Estimates
2008-2014." Accessed from SimplyMap
Database by county and block group;
U.S. Census Bureau. 2000. "Census
2000, Summary File 3" by block
group;
U.S. Census Bureau American
Community Survey. 2014. American
Community Survey 1-yr estimates by
county and 5-yr Estimates by block
group 2005-2013.ACS 5- yr estimates
by block group; Zillow by zip code
Used to determine
affordability.
Multiple sources were
used to corroborate
trends, since each source
had either the right
spatial or temporal scale,
not both.
MEDIUM: Different sources show
different values and somewhat
different trends.
Percent of population
in poverty (calibration)
Geographic Research Inc. 2015.
"Census Data 2000, ACS Estimates
2008-2014." Accessed from SimplyMap
Database by block group;
U.S. Census Bureau. 2000. "Census
2000, Summary File 3" by block group;
U.S. Census Bureau American
Community Survey. 2014. American
Community Survey 1-yr estimates by
county and 5-yr Estimates by census
tract 2005-2013.
Used to determine the
population in poverty and
zero-car households.
Multiple sources were
used to corroborate
trends, since each source
had either the right
spatial or temporal scale,
not both.
MEDIUM: Different sources show
different value and somewhat
different trends.
339
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR
SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Zero-car households
(calibration)
Geographic Research Inc. 2015.
"Census Data 2000, ACS Estimates
2008-2014." Accessed from SimplyMap
Database by block group.
Used to determine transit-
dependent population
which affects
transportation mode
choice.
Consistent with the
source used for the
population in poverty.
HIGH: The U.S. Census Bureau is the
authoritative source for demographic
data.
Percent of multifamily
dwelling units below 75
percent of median
renter costs
U.S. Census Bureau American
Community Survey. 2014. American
Community Survey 1 -yr Estimates
2005-2013 at county level
Used to indicate
organically affordable
units, which affects the
"housing gap for
households in poverty;" in
the absence of reliable
projections for this value,
model users can vary this
estimate going forward.
Consistent with data
source for renter housing
costs
LOW: The initial value has high
confidence, but we were unable to
identify either reliable projections or a
modeling method for estimating how
the distribution of rental units at
various cost units will change over
time. Accordingly, we have
designated this variable as a user-
specified input.
Subsidized total
dwelling units
Bearden, Ben. 2015. Durham and
Orange County Subsidized Housing
Point Shapefiles received from Ben
Bearden at Triangle J Council of
government.
Initial value for a user-
modifiable table which
determines the number of
households not at risk for
displacement.
Most comprehensive list
of publicly subsidized
housing units maintained
by governments in
Durham and Orange
county.
MEDIUM: Data are only available for
the current year; an undetermined
percentage of units are not guaranteed
to remain subsidized in the future, so
we are unable to project changes in
this variable over time.
Poverty threshold
U.S. Census Bureau. 2013. "National
Poverty Standards for 2013 by Size of
Family and Number of Related
Children Under 18 Years."
Forty-five percent of this
value is used as the
budget/income value in
the affordability index for
households in poverty.
The U.S. Census Bureau
uses an absolute poverty
threshold, which allows
comparison over time.
HIGH: Authoritative source for defining
the poverty threshold.
Elasticity of single-
family (SF) property
value to SF density
(Tier 1); elasticity of
nonresidential property
value to building size;
elasticity of SF
property value to job
density; elasticity of
nonresidential property
value to retail density
(Tier 2)
Srour, IssamM., KaraM. Kockelman,
and Travis P. Dunn. 2002.
"Accessibility indices: Connection to
residential land prices and location
choices."
Used to relate several
variables (including
density, a building size
proxy, and a retail density
proxy) to property values,
calibrated to match
historical trends.
Clear elasticities
available for a variety of
variables we estimate in
the model. One of best
sources for nonresidential
property values.
MEDIUM-LOW: The study examined
Dallas-Fort Worth during the 1990s and
derived elasticities on a neighborhood
basis. Elasticities from the study
therefore reflect variation across
locations, but we apply them in the
model to variation over time.
340
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR
SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Elasticity of single-
family (SF) property
value to income;
elasticity of SF
property value to
population
Jud, G. Donald, and Daniel T.
Winkler. 2002. "The dynamics of
metropolitan housing prices."
Used to relate income and
population to property
values, calibrated to match
historical trends.
Values provided by this
source yielded a better
fit to historical trends
than other available
sources (e.g., Srour,
2002).
MEDIUM: The study derived elasticities
for 130 metro areas around the U.S.,
so values may not be representative of
the study area.
Elasticity of single-
family (SF) and
multifamily (MF)
property value to
vacant land (used to
create the effect table
of vacant land on SF
property value (in Tier
1)
Capozza, Dennis R., Patric H.
Hendershott, Charlotte Mack, and
Christopher J. Mayer. 2002.
"Determinants of real house price
dynamics."
Used to relate vacant land
to property values,
calibrated to match
historical trends.
Best available source that
characterized this
relationship.
MEDIUM: The study derived elasticities
for 62 metro areas around the U.S., so
values may not be representative of
the study area.
Elasticity of single-
family (SF) property
value to retail density;
elasticity of multifamily
(MF) property value to
retail density; elasticity
of nonresidential
property value to retail
density (Tier 1)
Kain, John F., and John M. Quigley.
1970. "Measuring the value of housing
quality."
Used to relate a retail
density proxy to property
values, calibrated to match
historical trends.
Values provided by this
study yielded a better fit
to historical trends than
other available sources,
particularly in Tier 1.
MEDIUM-LOW: The study examined St.
Louis in the 1970s and derived
elasticities on a neighborhood basis.
Elasticities therefore reflect variation
across locations, but we apply them in
the model to variation over time.
Elasticity of single-
family (SF) property
value to commute
time; elasticity of MF
property value to
commute time
Kockelman, K. M. 1997. "The effects
of location elements on home
purchase prices and rents: Evidence
from the San Francisco Bay Area. "
Used to relate a commute
time proxy to property
values, calibrated to match
historical trends.
Best available source that
characterized this
relationship separately
for renter-occupied and
owner-occupied housing.
MEDIUM-LOW: The study examined San
Francisco in the 1990s and derived
elasticities on a neighborhood basis.
Elasticities therefore reflect variation
across locations, but we apply them in
the model to variation over time.
Elasticity of
nonresidential property
value to employment
Dobson, S.M., and J. A. Goddard.
1992. "The determinants of
commercial property prices and
rents." (Dobson and Goddard)
Used to relate employment
growth to property values,
calibrated to match
historical trends.
Best available source that
characterized the
relationship over time,
consistent with the
intended use in the
model.
MEDIUM-LOW: The study examined
Great Britain in the 1970s-80s, though
the elasticities were derived to reflect
change over time in response to
changes in employment.
341
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR
SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Elasticity of housing
costs to renter vacancy
Rosen, Kenneth T., and Lawrence B.
Smith. 1983. "The price-adjustment
process for rental housing and the
natural vacancy rate."
Used to relate renter
vacancy to renter housing
costs, calibrated to match
historical trends.
Best available source that
characterized this
relationship for renter-
occupied housing.
MEDIUM: The study derived elasticities
for 17 metro areas around the U.S. (we
use the median reported value), so
values may not be representative of
the study area.
342
-------
TABLE C-6. MODEL INPUT CHARACTERIZATION FOR THE WATER SECTOR
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR
SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Historical precipitation
State Climate Office of North
Carolina. 2015. "Historical Data.
Used to calculate drinking
water reservoir volume and
stormwater runoff
Authoritative government
source with regional data.
HIGH: Authoritative source at the same
geographic scale as the study area.
Percent
evapotranspi ration
Wilson. 2011. Constructed Climates:
A Primer on Urban Environments.
Chicago: The University of Chicago
Press.
Used to calculate drinking
water reservoir volume
Authoritative local
source; author is a
professor at a local
university.
MEDIUM: Water lost by
evapotranspiration will vary due to
many site-specific factors that are not
accounted for in this estimate.
Impervious Surface
Area
US EPA. 2015. "EnviroAtlas.
Used to calculate
stormwater runoff and
stormwater pollutant
loading. (Impervious
surface area is calibrated
within the land use
sector.)
Authoritative national
source with 1-meter-
resolution land cover.
MEDIUM: The 1-meter land cover
generally follows the Durham study
area but does not extend into Chapel
Hill.
Event mean
concentration (EMC) for
nitrogen and
phosphorus
NCDENR. 2011. Jordan Lake
Stormwater Accounting Tool User's
Manual (Version 1.1.).
Used to calculate
stormwater pollutant
loading.
Authoritative local
source. The Jordan Lake
Stormwater Accounting
Tool is a spreadsheet tool
used to calculate the
stormwater runoff impact
of new local
development.
MEDIUM: Data from this source are
based on local and national sources,
but EMCs have large variability due to
fine-scale differences in land cover.
Our use of EMCs for different land use
types approximates that variability.
Energy intensity for
water treatment and
distribution
American Council for an Energy-
Efficient Economy. 2010. "North
Carolina's Energy Future: Electricity,
Water, and Transportation Efficiency.
Report E102."
Used to calculate energy
used by the municipal
water system.
Authoritative source that
applies national numbers
on the energy intensity of
water systems to North
Carolina.
MEDIUM: The source provided two
energy intensity estimates: 3,972.5
kWh/Mgal and 3,239 kWh/Mgal. We
adopted a conservative approach in
the context of reducing CO2 emissions,
using the higher of the two intensity
estimates to prevent underestimating
CO2 emissions.
Total water demand
(calibration)
NC State Data Center. 2015. "LINC:
Log Into North Carolina."
Used to determine
withdrawals from water
reservoirs and calculate
energy used by the
municipal water system.
Authoritative government
source with multiple time
points.
HIGH: Authoritative source for
historical water demand data.
343
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR
SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Durham reservoirs -
days of supply
(calibration)
City of Durham Department of Water
Management. 2015. "Water Supply
Status." Available at:
http://durhamnc.gOV/ich/op/dwm/P
ages/Water-Supply-Status.aspx
Used (together with water
demand) to determine
water reservoir volume. In
addition, future trends in
reservoir volume are
modeled, assuming a
constant water supply rate
but a growing demand.
Authoritative government
source.
HIGH: Authoritative source for
historical reservoir supply data.
Durham reservoirs -
average demand
(calibration)
City of Durham Department of Water
Management. 2015. "Water Supply
Status." Available at:
http://durhamnc.gOV/ich/op/dwm/P
ages/Water-Supply-Status.aspx
Used (together with
reservoir levels) to
determine water supply.
Authoritative government
source.
HIGH: Authoritative source for
historical water demand data.
Projected water
demand - Durham and
other counties
(calibration)
TJCOG. 2012. "Triangle Regional
Water Supply Plan, Vol I: Regional
Needs Assessment."
Used (together with
reservoir levels) to
determine water supply.
Total water demand is also
used to calculate energy
and emissions for the
municipal water system.
Authoritative government
source.
MEDIUM: Future projections are
inherently uncertain.
Single-family
residential water use
rate per person
(calibration)
US DOE. 2011. 2010 Buildings Energy
Data Book.
Used to calculate
residential water use,
which is the largest
component of local
drinking water demand.
Authoritative government
source.
MEDIUM: Source provides a national
average that may not be
representative of our study area.
Multifamily water use
rate per household
(calibration)
US DOE. 2011. 2010 Buildings Energy
Data Book.
Used to calculate
residential water use,
which is the largest
component of local
drinking water demand.
Authoritative government
source.
MEDIUM: Source provides a national
average that may not be
representative of our study area.
Elasticity of water use
to residential density
Chang et al. 2010. "Spatial variations
of single-family residential water
consumption in Portland, Oregon."
Urban Geography 31, 7, 953-972.
Used to relate residential
density to residential
water use.
Has data relating single-
family residential density
(DU/acre) to household
water consumption.
MEDIUM: Source examined Portland,
OR, so the relationship it estimates
may not be representative of our study
area.
Nonresidential use rate
(calibration)
TJCOG. 2012. "Triangle Regional
Water Supply Plan, Vol I: Regional
Needs Assessment."
Used to calculate
nonresidential water use,
the second largest
component of local
drinking water demand.
Authoritative government
source
HIGH: Authoritative local government
source for nonresidential water use.
344
-------
TABLE C-7. MODEL INPUT CHARACTERIZATION FOR THE HEALTH SECTOR
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR
SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Mortality cases per ton
of PM2.5; morbidity
cases per ton of PM2.5;
Mortality cases per ton
of NOX; morbidity cases
per ton of NOX
US EPA. 2013. "Technical support
document: estimating the benefit per
ton of reducing PM2.5 precursors from
17 sectors."
Used to estimate the
mortality and morbidity
impacts of changes in PM2.5
and NOx emissions from
vehicles (positive or
negative) relative to the
BAU scenario.
Source involved extensive
health benefits research
and source apportionment
photochemical modeling;
most current available
study with estimates of
health impacts of PM2.5
and NOx emissions from
vehicles, expressed in
cases per ton.
MEDIUM-LOW: Though these are the
best available estimates for health
impacts per ton of PM2.5 and NOx
emissions reduced from vehicles, they
are national averages intended to be
used for a scoping-level analysis and
may therefore not be appropriate at
the scale used in our model.
Reduction in mortality
per person mile of
walking for
transportation per day
per capita; reduction in
mortality per person
mile of cycling for
transportation per day
per capita
WHO. 2014. "Health economic
assessment tools (HEAT) for walking
and for cycling: Methodology and user
guide, 2014 update."
Used to calculate avoided
premature mortalities due
to changes in the amount
of walking or cycling for
transportation per day per
capita.
The HEAT model
equations are simple and
based off of many
epidemiological studies
and associated
correlations between
walking/cycling and
health benefits.
MEDIUM-HIGH: The source's
recommended applicable age range for
estimating benefits from walking is 20-
74 and for cycling is 20-64, but we
applied it to the average rate of
walking and cycling for transportation
over the entire population. Also, the
accuracy of the HEAT calculations
should be understood as estimates of
the order of magnitude of the
expected effect rather than the
precise effect.
People in traffic
accidents per year;
crash injuries and
fatalities per year;
crash fatalities per
year; nonfatal crash
injuries per year
(calibration, Tier 2
only)
Highway Safety Research Center at
UNC Chapel Hill. 2015. "NC Crash
Data Query Web Site."
Used to estimate mortality
and morbidity impacts due
to changes in
transportation by vehicles.
Based on official crash
statistics from the North
Carolina government
(2001-2013), which are
highly detailed and able
to be parsed out in
numerous ways.
MEDIUM-HIGH: Statistics are for
Orange and Durham Counties
combined, as opposed to the exact
extent of the study area. Data are
limited to "reportable" crashes,
defined as those that occur on
publicly-maintained roadways and
which also result in at least $1,000
worth of damage.
345
-------
MODEL INPUT
SOURCE
HOW INPUT IS USED IN THE
MODEL
RATIONALE FOR
SELECTING SOURCE
CONFIDENCE LEVEL NOTES
Percent of people in
traffic accidents that
are injured or killed;
fatal percent of crash
injuries
Highway Safety Research Center at
UNC Chapel Hill. 2015. "NC Crash
Data Query Web Site."
Used to determine crash
fatalities and injuries per
year from the overall
number of people in
crashes during a year.
Data come from official
source and are highly
detailed.
Tier 2 = MEDIUM-HIGH; Tier 1 =
MEDIUM: The numbers are for all of
Durham and Orange Counties. It is
assumed that the Durham + Orange
figure will be fairly close to the MPO
figure. In the absence of data for Tier
1, we assume that values for Tier 1 are
equal to those for Tier 2.
People in traffic
accidents per VMT
(1) Highway Safety Research Center
at UNC Chapel Hill. 2015. "NC Crash
Data Query Web Site."(numerator);
(2) DCHCMPO. 2013. "Triangle
Regional Model version 5: Travel
demand result shapefiles." (TRM)
(denominator)
Used to estimate the
number of people in traffic
accidents per year, from
which are calculated the
numbers of people killed
and injured in traffic
accidents per year.
(1) Data come from
official sources and are
highly detailed.
(2) Primary source of VMT
and traffic congestion
projections used by local
transportation planning
agencies; spatial nature
allows clipping to both
Tiers.
Percent of person miles
of nonmotorized travel
by residents that is
cycling
(1) US DOT, Federal Highway
Administration. 2009. "2009 National
Household Travel Survey."
(2) ESRI Community Analyst. 2014.
"ACS 1 -yr and 5-yr Estimates 2005-
2013." (Tierl only)
Used to divide
nonmotorized person miles
by residents into walking
and cycling miles, which
the model uses to estimate
health benefits from
walking and cycling (using
the HEAT model).
(1) Authoritative source
with detailed data;
(2) Community Analyst is
able to clip ACS data from
the U.S. Census Bureau to
specific geographic areas.
Tier 2 = MEDIUM-LOW; Tier 1 = LOW:
Data were only available from both
sources in 2010; therefore, we
assumed the 2010 ratio to hold
constant for all model years. Statistics
are for Orange and Durham Counties
combined, as opposed to the exact
extent of the study area. Data are
limited to "reportable" crashes
(defined as those that occur on
publicly-maintained roadways and
which also result in at least $1,000
worth of damage), so they may
underestimate total traffic accidents
with health risks. In the absence of
data for Tier 1, we assume that values
for Tier 1 are equal to Tier 2.
MEDIUM: Data are not specific to the
study area, so we assume that the
value for this variable for the DCHC
MPO matches the national average.
Community Analyst/ACS data that
were used to scale NHTS data to Tier 1
only examine journey-to-work trips,
and do not consider the number of
miles traveled.
346
-------
Appendix D: Data Sources Not Used
As we developed the D-O LRP SD Model, we identified several data sources that we were not able to
include in the model, either because they were inferior to other data sources or because they did not
contain data at the appropriate scale for our model's purposes. The table presented in this appendix lists
data sources that we did not use to develop the model, together with the rationale for excluding each data
source.
347
-------
TABLE D-1. DATA SOURCES NOT USED
SECTOR
MODEL INPUT
DATA SOURCE
WHY NOT USED
Acres by land use
type
(1) Durham County Tax Administration.
2000-2014. Durham County Tax
Administration Real Property Database.
(2) Orange County Tax Administration.
2014. Orange County Parcel Database.
(3) Chatham County Tax Administration
Office. 2014. Chatham County Tax
Parcel Database.
When we examined the data from the tax property databases, we
found that they showed that total developed acres declined between
2001 and 2014, contrary to common sense. Laura Woods at the Durham
County Planning Department, who has analyzed the same dataset,
mentioned that many errors, particularly in earlier years, were
evident. Finally, Orange and Chatham counties did not have nearly as
detailed information as Durham, whereas the CV2 database was of
equal detail for the whole study area.
Land Use
Developed
nonresidential
square feet by type
DCHCMPO. 2013. "Triangle Regional
Model version 5: Socioeconomic data
and projections for the preferred
growth scenario"
Imagine 2040's projections for developed nonresidential square feet by
type are based on many strong assumptions (including constant
employee space ratios), and only provide estimates for 2040
Single-family home
values
U.S. Census Bureau American
Community Survey. 2014. "American
Community Survey 1 -yr and 5-yr
Estimates 2005-2013."
Since property values of multifamily homes and nonresidential
properties were only obtainable from the County Property Tax
Databases, we obtained values for single-family homes from the same
source for the sake of consistency.
Employee space
ratios
TJCOG. 2013. "Imagine 2040: The
Triangle Region Scenario Planning
Initiative Final Summary Document"
Uses constant employee space ratios from the literature that were very
inconsistent with those calculated from our data sources for
employment and square feet.
Transportation
Congestion
CAMPO and DCHCMPO. 2013. 2040
Metropolitan Transportation Plans.
Though it provides measures of congestion, this source does not provide
the specific measure of congestion that was ultimately chosen for the
model (namely, the ratio of peak-period travel time to freeflow travel
time, weighted by VMT on each road link).
348
-------
SECTOR
MODEL INPUT
DATA SOURCE
WHY NOT USED
Transportation
Construction cost
per lane mile
American Road and Transportation
Builders Association. 2014.
"Transportation FAQs."
Elasticity of desired
vehicle ownership
per person not in a
zero-car household
to per-capita
income
Johansson, Olof, and Lee Schipper.
1997. "Measuring the long-run fuel
demand of cars: Separate estimations
of vehicle stock, mean fuel intensity,
and mean annual driving distance."
We chose to use the Metropolitan Transportation Plan, which is both
authoritative and local, instead of this source, which only provides data
from other states.
The elasticity value estimated by this source is much higher than the
value we used, largely because it comes from a study that included
many developing nations, which are much farther from vehicle-
ownership saturation than the United States is.
Elasticity of public
transit travel to fare
price
Functioning lane
miles
Functioning lane
miles
LRT construction
cost per line mile
Sinha, Kumares Chandra, and Samuel
Labi. 2007. Transportation Decision
Making: Principles of Project
Evaluation and Programming (page
55).
Though we considered using the elasticity provided by this source, we
instead followed the recommendation of a staffer from a local planning
agency to use a formula based on the Simpson & Curtin formula, which
is commonly used by public transit planners.
CAMPO and DCHCMPO. 2013. 2040
Metropolitan Transportation Plans.
This source only provides data for the DCHC MPO as a whole and not at
a scale that could be used for Tier 1, so we used data from the TRM
instead. While this source's numbers are based on the TRM, it does not
count the TRM's "centroid connectors," which serve as stand-ins for a
variety of lower-order roadways.
US EPA, Office of Policy, Office of
Sustainable Communities. 2013. Smart
Location Database version 2.0. (SLD)
This source only provides densities of road miles within Census Block
Groups and does not provide the number of lanes on the average
roadway. Rather than clipping Block Groups to the study area (which
could introduce error), we used road-link-based data from the Triangle
Regional Model.
Triangle Transit. 2015. "Our Transit
Future."
This source provides numbers for the casual observer without much
explanation, in contrast to the planning documents that we used for
this variable.
349
-------
SECTOR
MODEL INPUT
DATA SOURCE
WHY NOT USED
Person miles of
travel per day
US DOT, Office of Policy, Office of
Transportation Policy, Office of
Economic and Strategic Analysis. 2014.
The Value of Travel Time Savings:
Departmental Guidance for Conducting
Economic Evaluations Revision 2 (2014
Update).
This source provides estimates of the monetary values that people
place on the time they spend traveling. These values could be used to
calculate what the appropriate ratio should be between the respective
effects of travel time and monetary travel costs on mode choice. The
source is authoritative, but not local, and it does not provide a clear
equation to input, but just numbers that could be used to figure out
such an equation. Because of these limitations, we chose not to
expressly include the tradeoff between travel time and monetary costs
in the model, although modal person miles are partially driven by
factors that influence travel time and monetary costs of travel.
Public transit trip
distance
NuStats. 2006. Greater Triangle Travel
Study, Household Travel Survey Final
Report, (study conducted to generate
inputs for the Triangle Regional Model)
The National Transit Database contains more recent and more detailed
data for this variable.
Transportation
Public transit
unlinked passenger
trips per day
Ramsey, Kevin, and Alexander Bell.
2014. Smart Location Database Version
2.0 User Guide.
The National Transit Database provides data for this variable that
covers multiple years, is more detailed, and is authoritative.
Vehicle stock
US EPA, Office of Policy, Office of
Sustainable Communities. 2013. Smart
Location Database version 2.0. (SLD)
For this variable, we relied on SimplyMap, which provides calibration
data for multiple years, whereas the SLD only provides data for a single
year. Both SimplyMap- and SLD-derived vehicle stock figures are based
on counts of households with given numbers of vehicles, but SimplyMap
includes more categories (up to 4+ vehicles) than SLD (up to 2+
vehicles), making it a more accurate source for our purposes.
Vehicle trip distance
NuStats. 2006. Greater Triangle Travel
Study, Household Travel Survey Final
Report, (study conducted to generate
inputs for the Triangle Regional Model)
The values provided by this source are inconsistent with the
Metropolitan Transportation Plan (MTP), which, although partially
based on this source, also contains more recent data. We calculated
the values to which this variable is calibrated primarily on the basis of
the MTP, instead.
350
-------
SECTOR
MODEL INPUT
DATA SOURCE
WHY NOT USED
VMT
CAMPO and DCHCMPO. 2013. 2040
Metropolitan Transportation Plans.
For this variable, we relied on direct outputs from the Triangle
Regional Model (which the numbers in this source are based on). Doing
so allowed us to get data in both Tiers, instead of just Tier 2.
Developed
nonresidential floor
area (square feet)
(1) Durham City-County Sustainability
Office. 2015. (citing Durham City-
County Planning Department data)
(2) TJCOG. 2014. CommunityViz 2.
For developed nonresidential floor area, we relied on data from the
Durham County property tax database over data from the Durham
planning department due to longer coverage (2000-2014 vs. 2005-2012).
We also considered using projections from CommunityViz 2, but those
values would have been calculated based on general floor area ratios
rather than being based on data.
Vehicle Fuel
Efficiency (MPG)
US DOT. 2013. "New and Small Starts
Evaluation and Rating Process Final
Policy Guidance."
Does not equate Btu/VMT to gallons of fuel/VMT. We instead needed to
calculate gallons of fuel to model energy spending. For this, we relied
on data from the Energy Information Administration's Annual Energy
Outlook (AEO), which has projected MPG every year from 2012-2040.
The "New and Small Starts" document, by contrast, only projects 10
and 20 years into the future.
Energy
PM2.s and NOx
emissions per VMT
US DOT. 2013. "New and Small Starts
Evaluation and Rating Process Final
Policy Guidance."
The New and Small Starts document gives PM2.s and NOx emissions rates
for three years only (2013, 2023, 2033), and while the NOx emissions
rates decrease over that time period, the PM2.5 emissions rates remain
constant. We were advised through personal correspondence with Dr.
Amy Lamson that this was erroneous and we instead calculated annual
emissions rates using GREET model emissions rates and a US fleet age
distribution, which shows a steady decline in PM2.5 emissions rates.
Solar capacity
(1) Oleniacz. 2014. "Company Puts
Last Panel on Durham County Solar
Farm." The Herald Sun (local
newspaper)
(2) Crowley and Quinlan. 2011. North
Carolina Clean Energy Data Book. p. 55
These two sources provided numbers for local solar capacity at two
years within the D-O LRP model timeframe. However, the North
Carolina Sustainable Energy Association more recently provided county-
level data on solar capacity, 2005-2014, which we used instead.
351
-------
SECTOR
MODEL INPUT
DATA SOURCE
WHY NOT USED
GRP a GRP Growth
Rate
Woods & Poole Economics, Inc.
Copyright 2014. "Durham and Orange
County, NC Data Pamphlet."
Economy
GRP: Woods & Poole use their own formulation to calculate GRP by
county based on their earnings data, but do not disclose this formula.
Because our calculated earnings estimates were less than Woods &
Poole (our employment numbers were lower), we could not use Woods
& Poole's GRP without knowing their methodology.
GRP Growth Rate: Woods & Poole's projected GRP growth rate (2014-
2040) of about 3% per year could not be applied to our historically
calculated GRP in 2013 because this would cause GRP to grow faster
than calculated "total earnings" between 2014 and 2040, with the
share of earnings decreasing from 60% in 2013 to less than 55% of GRP
in 2040. Without knowing Woods & Poole's rationale for why GRP would
grow faster than earnings, we decided instead to hold the earnings
share of GRP constant at 60% between 2014 and 2040 for the BAU
scenario (Tier 2), and use this calculated GRP to calibrate the Tier 2
GRP in the model.
GRP a GRP Growth
Rate
BEA. 2014. "Gross Domestic Product
(GDP) by Metropolitan Area: GDP in
Current Dollars, 2001-2013, for the
Durham-Chapel Hill, NC Metropolitan
Statistical Area (MSA)."
Historical GDP data from the BEA could not be used for Tier 2 because
they were only available by Metropolitan Statistical Area (MSA), not by
county.
Total Employment
Woods & Poole Economics, Inc.
Copyright 2014. "Durham and Orange
County, NC Data Pamphlet;"
BEA. 2014. "Local Area Personal
Income and Employment: Table CA25N,
Total Full-Time and Part-Time
Employment by NAICS Industry, 2001-
2013, for Durham and Orange County,
NC."
We could not use Woods & Poole's historical or projected data for total
employment (historical employment data from Woods & Poole are from
the BEA), even for the model years not covered by the TRMvS (2000-
2009), because Woods & Poole (BEA) total employment numbers are
much higher in 2010 than the TRMvS SE data total employment
numbers. This difference is because they measure more kinds of
employment, including proprietors, private household employees, and
miscellaneous workers, which the data collection for the TRMvS did not
include.
352
-------
SECTOR
MODEL INPUT
DATA SOURCE
WHY NOT USED
Employment by Job
Category
DCHCMPO. 2013. "Triangle Regional
Model version 5: Socioeconomic data
and projections for the preferred
growth scenario."
The TRMvS SE data provided employment by job categories, however
some job types within a certain 2-digit NAICS category were divided
among job categories in the TRM, making it impossible to combine
these jobs by category with earnings per job by category data from
Woods & Poole to get total earnings. E.g. The job type "Library &
Archives," which is categorized by NAICS as an information job (2-digit
NAICS code = 51), is considered a service job in the TRMvS SE data,
while the job type "Newspaper Publishers," also categorized as an
information job with a 2-digit NAICS code of 51, is considered an
industrial job in the TRMvS SE data.
Equity
Elasticities of
property values to
housing age, home
sq ft, # of rooms,
and interest rates
Dobson, S.M., and J. A. Goddard. 1992.
"The determinants of commercial
property prices and rents."
Heikkila, E., P. Gordon, J.I. Kim, R.B.
Peiser, and H.W. Richarson. 1989.
"What happened to the CBD-distance
gradient?: Land values in a policentric
city."
Kain, John F., and John M. Quigley.
1970. "Measuring the value of housing
quality."
Kockelman, K. M. 1997. "The effects of
location elements on home purchase
prices and rents: Evidence from the
San Francisco Bay Area."
Srour, IssamM., KaraM. Kockelman,
and Travis P. Dunn. 2002. "Accessibility
indices: Connection to residential land
prices and location choices."
While these variables are significantly associated with single-family or
nonresidential property values in the literature, they were either not
associated with property values in our time series data (e.g., housing
age did not correlate with property values over time and had very low
R2), or were not likely to show any changes in response to other
variables in our model, so we decided not to include these
relationships.
Median Rent
Zillow Real Estate Research. 2010-
2014. "Median Rent by County and Zip
Code."
The data were only available for a short timespan, were based on
advertised rent rather than rent paid, and were not standardized as to
the costs they include (e.g. utilities), as are renter costs reported by
the U.S. Census Bureau.
353
-------
SECTOR
MODEL INPUT
DATA SOURCE
WHY NOT USED
Average
Precipitation
NOAA. 2015. Durham, NC station
312515
NOAA's precipitation data for the region comes from a single weather
station in the city of Durham. Because the model is regional in scope,
we instead chose a data source that provided regional precipitation
data.
Water
Per Capita Water
Use
USGS. 2015. "Water use in North
Carolina: 2005 data tables."
http://nc.water.usgs.gov/infodata/wa
teruse/data/Data_Tables_2005.html
The USGS source has data for 2005 only. In addition, the source that we
used for this model input provided separate water use intensities for
different land use types (namely, single-family, multifamily, and
nonresidential).
Lake Water Quality
NCDENR. 2015. "Falls Lake and Jordan
Lake monitoring."
http://portal.ncdenr.org/web/wq/fall
sjordan
The NC DENR source includes data on nitrogen (N) and phosphorus (P)
concentrations in the lakes receiving stormwater runoff from Tier 2,
which is different (though related to) the N and P loading calculated by
the model (estimated as Ibs N or Ibs P/year).
Health
Premature mortality
and morbidity per
ton for PM2.s and
NOX
Chan and Jackson. 2011.
http://www.lung.org/associations/stat
es/california/assets/pdfs/advocacy/sm
art-growth/tiax-full-technical-
slides.pdf
We relied on incidence-per-ton estimates from EPA's technical support
document (EPA 2013) because they are more recent and cover a larger
time period (2016-2030, compared to 2025 only in Chan and Jackson
2011).
354
-------
Appendix E: Stakeholder Meeting Materials
FIRST STAKEHOLDERS MEETING ON FEBRUARY 24, 2014 AT EPA RTP
Meeting Announcement and Agenda
Workshop on Developing Integrated Approaches to Sustainability
EPA's Research Program in Sustainable and Healthy Communities (SHC) - Durham Pilot Project
Feb24, 1:00-4:30 pm
EPA RTP Campus Room C400A
Objective: To design an approach that uses dynamics systems modeling to explore the interplay among
actions and decisions that lead to (or fail to lead to) healthier and more sustainable communities.
Background: EPA is engaging with its Federal (Department of Transportation and Housing and Urban
Development) and local partners to develop approaches that enable communities to better consider
interactions among decisions in the context of sustainability. Durham has been selected as one of the
communities where approaches are developed and applied to real-world issues.
The present activity focuses on the proposed Durham-Orange Light Rail plan as a driver for these
interactions - within the context of projections for regional demographic and economic growth - and
Durham's goals for a sustainable and healthy community.
1:00 - 1:15 Welcome and Overview of SHC
1:15 - 1:30 Systems Approaches to Sustainability
1:30 - 1:45 Durham/Chapel Hill Sustainability Goals
1:45-2:00 Overview of Durham - CH Light Rail
2:00 - 2:15 Population and Economic Growth in the Region
2:15-2:30 Infrastructure and Growth -Structural and
Fiscal Issues
2:30-2:45 Break
2:45 - 3:00 Strategic Planning for Infrastructure
Improvements
3:00 - 3:15 Urban Growth, Watersheds and Water
Resources
3:15 - 3:30 Transportation, Energy and Air Quality
3:30 - 4:30 Discussion - Key Interactions
Meeting Attendees - February 24, 2014
Rochelle Araujo, EPA ORD
Joseph Fiksel, EPA ORD
Tobin Freid, Durham Sustainability
Office
JeffWeisner, URS
Felix Nwoko, DCHC MPO
John Hodges-Copple, Triangle J
Hannah Jacobson, Durham City-
County Planning
Sarah Bruce and Heather Saunders,
Triangle J
Brennan Bouma, Triangle J
ALL
355
-------
EPA
Name
Almeter, Andrew
Araujo, Rochelle
Cox, Llael
Flanders, Nick
Foley, Gary
Hughes, Sara
Kolling, Jenna
Lefew, William
Loughlin, Dan
McCullough, Melissa
Procter, Andrew
Rimer, Linda
Rudder, Charles
Smith, Betsy
Ten Brink, Marilyn
Affiliation
ORISE
EPA
ORISE
ORISE
EPA
ORISE
ORISE
EPA
EPA
EPA
ORISE
EPA
ORISE
EPA
EPA
Email
Almeter.Andrew@epa.gov
Araujo.Rochelle@epa.gov
Cox.Llael@epa.gov
Flanders.Nick@epa.gov
Foley.Gary@epa.gov
Hughes.Sara@epa.gov
Kolling.Jenna@epa.gov
Lefew.William@epa.gov
Loughlin.Dan@epa.gov
Mccullough.Melissa@epa.gov
Procter.Andrew@epa.gov
Rimer.Linda@epa.gov
Rudder.Charles@epa.gov
Smith.Betsy@epa.gov
Tenbrink.Marilyn@epa.gov
D-O LRP Stakeholders
Name
Affiliation
Email
Bassi, Andrea
Bouma, Brennan
Bruce, Sarah
Carol, Megan
Fiksel, Joseph
Fried, Tobin
Henry, Andrew
Hodges-Copple, John
Jacobson, Hannah
Nwoko, Felix
Saunders, Heather
Tanners, Nadav
Wilson, Will
Weisner, Jeff
Youngblood, Helen
KnowlEdge Sri
Triangle J
Triangle J
Durham City/County
Ohio State University/EPA
Durham City/County - Sustainability
DCHCMPO
Triangle J
Durham City/County - Planning
DCHCMPO
Triangle J
Industrial Economics
Duke
URS
Durham City/County
andrea.bassi@ke-srl.com
bbouma@tjcog.org
sbruce@tjcog.org
mcarroll@dconc.gov
fiksel.2@osu.edu
tfreid@dconc.gov
andrew.henry@durhamnc.gov
johnhc@tjcog.org
Hannah.Jacobson@durhamnc.gov
felix.nwoko@durhamnc.gov
hsaunders@tjcog.org
NTanners@indecon.com
wgw@duke.edu
jeff.weisner@urs.com
helen.youngblood@durhamnc.gov
356
-------
SECOND STAKEHOLDERS MEETING ON JUNE 4, 2014 AT TRIANGLE J COUNCIL OF GOVERNMENTS
Meeting Announcement and Agenda
EPA Project - Integrated Approaches to Sustainability:
Developing a Dynamic Systems Model
Focus: Durham-Chapel Hill Light Rail Project
Event: Stakeholders Meeting
Wednesday, June 4th l:00pm
Large Conference Room, Triangle J Council of Governments
4307 Emperor Blvd, Suite 110 Durham, NC 27703
Purpose:
• An update on the progress of our conceptual model since our first stakeholders meeting
(February 24th, 2014).
• Updates on the light rail project development phase and draft environmental impact statement.
• An interactive discussion on:
o How well our conceptual model represents the system holistically.
o Aligned decisions or policies that could be tested in our model and what we can do to
incorporate them.
o Availability of local data that could be useful to our model.
o How we can make our model useful to the stakeholders.
Schedule:
1:00 - 1:10 Welcome and Introductions
1:10 - 1:20 Meeting Objectives, Project Overview and Approach (Araujo)
1:20 - 1:40 Demo of 3VS Model (Bassi)
1:40 - 1:45 Model organization, Outline of Desired Stakeholder Input (Araujo, Rolling)
1:45 - 2:00 Land use change, Equity (Cox)
2:00 - 2:10 Stakeholder Inputs to Land use change, equity
2:10 - 2:20 Water resources and infrastructure (Almeter)
2:20 - 2:30 Stakeholder Inputs to water resources and infrastructure
2:30 - 2:45 Health, quality of place, economics (Rolling)
2:45-3:00 Stakeholder inputs to health, quality of place, economics
3:00 - 3:10 Transportation, Energy (Flanders)
3:10 - 3:20 Stakeholder input to Transportation, Energy
3:20 - 3:30 Update on light rail project, draft EIS and other potential model uses
3:30-4:00 Group discussion on model structure, data sources, applications, future interactions
357
-------
Meeting Attendees - June 4, 2014
EPA
Name
Almeter, Andrew
Araujo, Rochelle
Cox, Llael
Flanders, Nick
Geller, Andrew
Rolling, Jenna
McCullough, Melissa
Mulvaney, Kate
Procter, Andrew
Smith, Betsy
Ten Brink, Marilyn
Thompson, Bob
Towery, Meredith
Email
almeter.andrew@epa.gov
araujo.rochelle@epa.gov
cox.llael@epa.gov
flanders.nick@epa.gov
geller.andrew@epa.gov
kolling j enna@epa. gov
mccullough.melissa@epa.gov
mulvaney .kate @epa. gov
procter.andrew@epa.gov
smith.betsy@epa.gov
tenbrink.marilyn@epa.gov
thompson.bob@epa.gov
towery.meredith@epa.gov
Affiliation
ORISE
EPA
ORISE
ORISE
EPA
ORISE
EPA
ORISE
ORISE
EPA
EPA
EPA
EPA
D-O LRP Stakeholders
Name
Email
Affiliation
Bendor, Todd
Bonk, David
Bouma, Brennan
Bruce, Sarah
Cain, Aaron
Fried, Tobin
High, William
Hodges-Copple, John
Huegy, Joe
Legaard, Roy Jr.
Makoid, Meghan
Nwoko, Felix
Poindexter, Gavin
Richardson, John
Saunders, Heather
Shiflett, Michael
Watterson, Bergen
Zhang, Yanping
bendor@unc.edu
dbonk@townofchapelhill.org
bbouma@tj cog. org
sbruce@tjcog.org
aaron .cain@durhamnc. gov
tfreid@dconc.gov
williamhigh@gmail.com
johnhc@tjcog.org
jbhuegy@ncsu.edu
roy@wavedash.co
mmakoid@triangletransit.org
felix.nwoko@durhamnc.gov
gavin.poindexter@urs.com
jrichardson@townofchapelhill.org
hsaunders@tj cog.org
mwshiflett@hotmail .com
bwatterson@tj cog .org
vanping. zhang@durhamnc. gov
UNC-CH
Chapel Hill Transportation
TJCOG
TJCOG
Durham Planning
Durham City-County
UNC
TJCOG
NCSU-TRM
Wavedash Digital, LLC
Triangle Transit
DCHC MPO
URS
Chapel Hill - Sustainability
TJCOG
DO Friends of Transit
TJCOG
DCHC MPO
358
-------
Stakeholder Meeting Handout - June 4, 2014
LRP Stakeholder Feedback Sheet
Tentative Key Indicators
Land:
• relative change in developed land
• developed land per capita
• agricultural land converted to
developed use
• quantity of multifamily housing and
compact development
• degree of mixed-use
Equity:
• displacement of poorer residents
• housing affordability
• transportation affordability
• access to jobs
Water:
• pollutant Loading
• water Consumption
• water Treatment Cost
• infrastructure Cost
Economy:
• public revenues due to increased sales
tax revenues, property tax, and public
transport fees
• economic productivity gains due to
increase in public transport trips and
reduced congestion
• public expenditure changes due to
reduced infrastructure costs
• private spending changes due to
housing and transportation affordability
Health:
• air emissions
• traffic collisions
• physical activity increases from
compact development and public
transportation trips
• stress
Quality of Place:
• walkability
• housing and transportation affordability
• traffic collisions
• access to ...
• public transportation capacity and trips
Transportation:
• relative change in Vehicle Miles
Traveled (VMT)
• traffic congestion
• public transportation trips
• non-motorized trips
• transportation-related government
expenditures
Energy:
• total energy consumption
• air emissions
• spending on motorized transportation
• electricity spending
Questions for stakeholders:
1. Are these indicators of interest?
359
-------
2. Which other indicators would be valuable to your organization?
3. Which elements or assumptions do you expect to be the most significant drivers of change due to the
light rail? Have we addressed them?
Possible interventions that could be testable
• Land and Equity:
o reduced parking requirements for affordable TOD
o increased height/density allowances for affordable housing in the Yi mile zone
o flexible-use zoning near rail stations
o expedited approval processes for mixed-use development
o agricultural land conservation strategies
• Water:
o green infrastructure incentives
• Economics:
o strategies to promote vibrant walkable places
• Health and Quality of Place:
o strategies to increase walkability of the station areas
• Transportation:
o improved pedestrian and bicycle facilities
o expanding/improving bus services that connect with the light rail line
o increasing the level of traffic congestion that must be projected in order to warrant roadway-
system capacity expansions, based on the assumption that some travelers will switch to either
public transit or non-motorized transportation
• Energy:
o actions by local government agencies to reduce energy consumption in their operations, such as
drinking water and wastewater conveyance and providing public transit services
o strategies promote energy-efficiency of buildings near light rail stations, through compact
design
Questions for stakeholders:
1. Are the system-wide impacts of these strategies and interventions of interest?
2. What specific interventions would be valuable to your organizations to test?
360
-------
THIRD STAKEHOLDERS MEETING ON MAY 13, 2015 AT TRIANGLE J COUNCIL OF GOVERNMENTS
Meeting Announcement and Agenda
EPA Project - Integrated Approaches to Sustainability:
Developing a Dynamic Systems Model
Focus: Durham-Orange Light Rail Project
Event: Stakeholders Review of Model Development and Analyses
Wednesday, May 13th l:00pm
Large Conference Room, Triangle J Council of Governments
4307 Emperor Blvd, Suite 110 Durham, NC 27703
Purpose:
• To present and discuss the "beta" version of the working integrated systems model for
Durham -Orange Light Rail and associated regional and local issues:
o integrated, calibrated and tested interactions for land use, economic, and
transportation sectors.
o pathways, available information and potential model representation of energy, air
emissions, water quantity, stormwater runoff, health, equity and quality of place.
• To discuss scenarios or policy options for use of the model.
• To discuss ways of interacting with the model (inputs/outputs/interfaces} and setting in
which model can be most useful.
Schedule (Approximately):
1:00 - 1:10 Welcome and introductions
1:10 - 1:20 Meeting objectives, project overview and approach
1:20-2:15 Model Demo with land and transportation scenarios
2:15 - 2:30 Discussion of model construction
2:30-2:45 Break
2:45 - 3:30 Discussion of policies, scenarios
3:30 - 3:45 Demonstration of interfaces developed for similar models
3:45 - 4:00 Discussion - Interfaces, General
The team can be available for an additional half hour after the meeting for folks to "poke around under the
hood". Similarly, we will be distributing comment sheets for attendees to use - either at the meeting - or to
send along later - to provide feedback on the model, data/projections/models used, scenarios to be
analyzed, additional functionalities desired of the model, etc.
361
-------
Meeting Attendees - May 13, 2015
In Person: Llael Cox, Jenna Rolling, Nick Flanders, Andrew Procter, Rochelle Araujo (EPA ORD);
Nadav Tanners (ffic); Hannah Jacobson (Durham City/County Planning Dept); Patrick McDunnough
(GoTriangle); Katherine Eggleston (GoTriangle); Geoff Green (GoTriangle); David Bonk (Chapel Hill);
Mila Vega (Chapel Hill Transit); Jeff Weisner (AECOM); Marissa Mortiboy (Durham County); Helen
Youngblood (Durham City/County Planning); Mike Shiflett (Coalition for Affordable Housing &
Transit); Roy LeGard (Chapel Hill); Chris Dickey (Durham); Constance Stancil (City of Durham); Andy
Henry (MPO); Andy Gillespie (EPA ORD); Elizabeth Doran (Grad Student at Duke); Todd BenDor
(UNC Chapel Hill); John Hodges-Copple (Triangle J); Laura Jackson (EPA ORD);
On the Phone: Andrea Bassi (KnowlEdge Sri); Marilyn Ten Brink (EPA ORD); Melissa McCullough
(EPA ORD); Gary Foley (EPA ORD); Kate Mulvaney (EPA ORD)
362
-------
Stakeholder Meeting Handouts- May 13, 2015
EPA D-0 LRP System Dynamics Model
Summary of main assumptions in current model scenarios - May 13, 2015
BAD (Business as Usual)
Land
• Share of acres in each land use type held constant at 2013 values (from CV2 Parcel Geodatabase)
• Floor area ratios and dwelling units per acre remain steady at 2013 values (acres from CV2 Parcel Geodatabase, sq
ft from County Tax Property Databases, and dwelling units from the Census)
• Employee space ratios remain relatively steady at 2013 value, except for:
o industrial which declines following the historical trend
o retail and office in Tier 1, which rise following the historical trend
• The percent of people in single and multi-family dwelling units are held constant at 2014 values
• Single family and multifamily household sizes are held constant at the 2014 value
Economy
• Total employment calibrated to match employment in the TRM v5 SE Data
• Total earnings is 60% of GRP, while gross operating surplus is 40%
• Share of employment by category and earnings per job by category from Woods & Poole Economics, Inc.
Transportation
• No light rail built
• No road building after 2017
• Bus system expansion conforms to Orange County and Durham County public transit plans
• Sidewalk building conforms to Triangle Regional Model projections
Energy and Water
• Passenger vehicle MPG increases according to US Energy Information Administration projections
• Annual precipitation stays constant over time
• Water consumption rates (per dwelling unit or per job) are taken from literature, including Triangle Regional Water
Supply Plan
Light Rail
• Light Rail service begins 2026
• Tier 1 commercial sq ft increases in 2020 (10% increase in desired retail/office/service sq ft)
Light Rail + Density
• Increases density of new development (sq ft per acre and dwelling units per acre), by:
o 10%inTier3(DCHCMPO)
o 20% in Tier 1 (combined 1/2 mile radius of all proposed rail stations)
Light Rail + Road Building
• Future road building increased, conforming to DCHC MPO 2040 MTP "Preferred Scenario"
• Still introduce light rail in 2026, but no increased density
363
-------
Light Rail + Density + Road Building
• Light Rail service begins 2026
• Tier 1 commercial sq ft increases in 2020 (10% increase in desired retail/office/service sq ft)
• Increases density of new development (sq ft per acre and dwelling units per acre), by:
o 10%inTier3(DCHCMPO)
o 20% in Tier 1 (combined 1/2 mile radius of all proposed rail stations)
• Future road building increased, conforming to DCHC MPO 2040 MTP "Preferred Scenario"
364
-------
System Dynamics Model Stakeholder Feedback Sheet - May 13, 2015
Name and contact:
Organization:
We plan to add the ability to test several policies, interventions, or future trends in the model.
Please rank the top three options in order of importance to your organization (1 as most
important).
Redevelopment that changes density and use of existing land
Green infrastructure and its effect on impervious surface and stormwater
Land conservation strategies
Reduced parking requirements and the effect on affordability and VMT
Telecommuting and the effect on congestion
Other:
Other:
Please rank your three highest priority indicators to add to the model (1 as most important).
Income brackets and housing costs brackets to better assess household budgets and affordability
Health outcomes in terms of morbidity
Health outcomes in dollar terms
Quality of place index
Other:
Other:
Other:
Many assumptions about future trends are already easily modifiable in the model, such as those
below. Are there other future trends that we might want to consider?
1. Densification (retail, office, service, industrial, single family, and multifamily separately)
2. Increase in nonmotorized facilities
3. Shifts in the share of employment
4. Increases in energy efficiency
What features of a user interface would you find most useful?
In what settings could you imagine using this model?
Please let us know any other feedback you have (feel free to elaborate on the back)
365
-------
Bibliography
Abbas, Khaled A., and Michael G.H. Bell. 1994. "System dynamics applicability to transportation modeling."
Transportation Research Part A 28 (5):373-390.
American Council for an Energy Efficient Economy. 2010. "North Carolina's energy future: electricity, water, and
transportation efficiency." American Council for an Energy Efficient Economy, http://aceee.org/north-
carolina%E2%80%99s-energy-future-electricitv-water-and-transportation-efficiencv.
American Road and Transportation Builders Association. 2014. "Transportation FAQs." Accessed July 10.
http://www.artba.org/about/transportation-faqs/.
Anastas, Paul T. 2012. "Fundamental changes to EPA's research enterprise: The path forward." Environ Sci
Techno/46 (2):580-6. doi: 10.1021/es203881e.
Bassi, Andrea M., Robert Powers, and William Schoenberg. 2010. "An integrated approach to energy prospects
for North America and the rest of the world." Energy Economics 32 (l):30-42. doi:
10.1016/j.eneco.2009.04.005.
BEA. 2014a. "Annual State Personal Income and Employment: Table SA05N - Personal Income by Major Source
and Earnings by NAICS Industry, 2000-2013, for North Carolina." Last Modified September 30.
http://www.bea.gov/iTable/iTableHtml.cfm?reqid=70&step=30&isuri=l&7022=2&7023=0&7024=naics
&7033=-1&7025=0&7026=37000&7027=-1&7001=42&7028=-1&7031=0&7040=-
l&7083=levels&7029=28&7090=70.
BEA. 2014b. "Gross Domestic Product (GDP) by Metropolitan Area: Per Capita Real GDP, 2001-2013, for the
Durham-Chapel Hill, NC Metropolitan Statistical Area (MSA)." Last Modified September 16.
http://www.bea.gov/iTable/iTable.cfm?reqid=70&step=10&isuri=l&7003=200&7035=-
l&7004=naics&7005=-l&7006=20500&7036=-l&7001=2200&7002=2&7090=70&7007=-
l&7093=levels#reqid=70&step=10&isuri=l&7003=1000&7035=-
l&7004=naics&7005=l&7006=20500&7036=-l&7001=21000&7002=2&7090=70&7007=-
l&7093=levels.
BEA. 2014c. "Gross Domestic Product (GDP) by State." Accessed January 2015.
http://www.bea.gov/iTable/iTableHtml.cfm?reqid=70&step=10&isuri=l&7003=200&7035=-
l&7004=naics&7005=-l&7006=37000&7036=-l&7001=1200&7002=l&7090=70&7007=-l&7093=levels.
BEA. 2014d. "Local Area Personal Income and Employment: Table CA25N, Total Full-Time and Part-Time
Employment by NAICS Industry, 2001-2013, for Durham and Orange County, NC." Last Modified
November 20.
http://www.bea.gov/iTable/iTableHtml.cfm?reqid=70&step=30&isuri=l&7022=ll&7023=7&7024=naics
&7033=-1&7025=4&7026=37063.37135&7027=-1&7001=711&7028=-1&7031=37000&7040=-
l&7083=levels&7029=33&7090=70
BEA. 2014e. "Local Area Personal Income Methodology." http://www.bea.gov/regional/pdf/lapi2013.pdf.
BEA. 2015. "National Income and Product Accounts Tables: Section 2 - Table 2.3.4. Price Indexes for Personal
Consumption Expenditures (PCE), Annual Data for 2000-2014." Last Modified April 29.
http://www.bea.gov/iTable/iTableHtml.cfm?reqid=9&step=3&isuri=l&904=2000&903=64&906=a&905
=2014&910=x&911=0.
Bearden, Ben. 2015. Durham and Orange County Subsidized Housing Point Shapefiles. edited by Triangle J
Council of Governments.
Cai, Hao, Andrew Burnham, and Michael Wang. 2013. "Updated Emission Factors of Air Pollutants from Vehicle
Operations in GREET Using MOVES." Argonne National Laboratory. https://greet.es.anl.gov/publication-
vehicles-13.
CAMPO and DCHC MPO. 2013. "2040 Metropolitan Transportation Plans."
http://www.dchcmpo.org/programs/transport/2040.asp.
Capozza, Dennis R., Patric H. Hendershott, Charlotte Mack, and Christopher J. Mayer. 2002a. "Determinants of
real house price dynamics." NBER Working Paper Series (9262).
366
-------
Capozza, Dennis R., Patric H. Hendershott, Charlotte Mack, and Christopher J. Mayer. 2002b. Determinants of
Real House Price Dynamics. In NBER Working Paper Series. Cambridge, MA: National Bureau of
Economic Research.
Chan, Michael, and Michael Jackson. 2011. "Health impact of land use VMT and in-use vehicle fleets." TIAX
Accessed September 8.
http://www.lung.org/associations/states/california/assets/pdfs/advocacy/smart-growth/tiax-full-
technical-slides.pdf.
Chang, H, GH Parandvash, and V Shandas. 2010. "Spatial variations of single-family residential water
consumption in Portland, Oregon." Urban Geography 31 (7):953-972.
Chatham County Tax Administration Office. 2014. Chatham County Tax Parcel Database (ASOUTR). edited by
GIS/Mapping Division. Chatham County, NC.
Chatman, Daniel G., Robert Cervero, Emily Moylan, Ian Carlton, Dana Weissman, Joe Zissman, Erick Guerra, Jin
Murakami, Paolo Ikezoe, Donald Emerson, Dan Tischler, Daniel Means, Sandra Winkler, Kevin Sheu, and
Sun Young Kwon. 2014. "TCRP Report 167: Making Effective Fixed-Guideway Transit Investments:
Indicators of Success." Transportation Research Board, http://www.trb.org/main/blurbs/170972.aspx.
City of Durham. 2009. Durham landfill gas-to-energy green power project.
City of Durham Department of Finance. 2009, 2014. "Comprehensive Annual Financial Report."
http://durhamnc.gov/Archive.aspx?AMID=60.
City of Durham Department of Water Management. 2015. "Water supply status." Accessed August 31.
http://durhamnc.gov/1213/Water-Supply-Status.
Cole, Brian L, Michelle Wilhelm, Peter V Long, Jonathan E Fielding, Gerald Kominski, and Hal Morgenstern. 2004.
"Prospects for health impact assessment in the United States: new and improved environmental impact
assessment or something different?" Journal of Health Politics, Policy and Law 29 (6):1153-1186.
Cox, John. 2015. Email message to authors on July 14, 2015.
Crowley, Rich, and Paul Quinlan. 2011. "2011 North Carolina clean energy data book." North Carolina
Sustainable Energy Association.
Dannenberg, Andrew L, Rajiv Bhatia, Brian L Cole, Sarah K Heaton, Jason D Feldman, and Candace D Rutt. 2008.
"Use of health impact assessment in the US: 27 case studies, 1999-2007." American journal of
preventive medicine 34 (3):241-256.
DCHC MPO. 2013. "Triangle Regional Model version 5: Socioeconomic data and projections for the preferred
growth scenario and travel demand result shapefiles." Accessed May 2014.
http://www.dchcmpo.org/programs/regional/travel demand /.
DCHC MPO, Triangle Transit Board of Trustees, and Durham Board of County Commissioners. 2011. "The
Durham County Bus and Rail Investment Plan." http://ourtransitfuture.com/durham-countv-bus-and-
rail-investment-plan/.
Dobson, S.M., and J. A. Goddard. 1992. "The determinants of commercial property prices and rents." Bulletin of
Economic Research 44 (4):301-321.
Driver, Laurel. 2015. Email message to authors on June 12 2015.
Duke Energy. 2008. Duke Energy Carolinas signs deal to turn landfill gas into energy. Charlotte, NC: Duke Energy.
Duran-Encalada, Jorge A., and Alberto Paucar-Caceres. 2009. "System dynamics urban sustainability model for
Puerto Aura in Puebla, Mexico." Systemic Practice and Action Research 22 (2):77-99.
Durham City-County Planning Department. 2012. "Durham Comprehensive Plan, Appendix A: Existing
Conditions." Durham City-County Planning Department.
http://durhamnc.gov/DocumentCenter/Home/View/1168.
Durham City-County Planning Department. 2014. "Station Area Strategic Infrastructure Study (SASI)." Accessed
November 5. http://durhamnc.gov/373/Station-Area-Strategic-lnfrastructure-SA.
Durham City-County Sustainability Office. 2015. GHG_BaselineValues. Durham, NC.
Durham County Finance Department. 2009, 2014. "Comprehensive Annual Financial Report."
http://dconc.gov/government/departments-f-z/finance/annual-financial-reports.
Durham County Tax Administration. 2000-2014. Durham County Tax Administration Real Property Database.
edited by Land Records Division. Durham County, NC.
ESRI. 2015. "Esri Demographics: Data allocation method." Accessed January 2015. http://doc.arcgis.com/en/esri-
demographics/reference/data-allocation-m ethod.htm.
367
-------
ESRI Community Analyst. 2014. "Census Data 2000, ACS 1-yr and 5-yr Estimates 2005-2013, and ESRI
Demographic Projections for 2014 and 2019." Accessed December.
http://www.esri.com/software/arcgis/communitv-analyst.
Ewing, Reid, and Robert Cervero. 2010. "Travel and the built environment: A Meta-Analysis." Journal of the
American Planning Association 76 (3):265-294. doi: 10.1080/01944361003766766.
Ewing, Reid, Marybeth DeAnna, and Shi-Chiang Li. 1996. "Land use impacts on trip generation rates."
Transportation Research Record No 1518:1-6.
Fiksel, Joseph. 2006. "Sustainability and resilience: Toward a systems approach." Sustainability: Science Practice
and Policy 2 (2):14-21.
Fiksel, Joseph, Randy Bruins, Annette Gatchett, Alice Gilliland, and Marilyn Ten Brink. 2014. "The triple value
model: A systems approach to sustainable solutions." Clean Technologies and Environmental Policy 16
(4):691-702.
Foth, Marcus, Helen G Klaebe, and Gregory N Hearn. 2008. "The role of new media and digital narratives in
urban planning and community development." Body, Space & Technology 7 (2).
Freid, Tobin. 2015. Email message to authors on January 9.
FTA. 1969. "National Environmental Policy Act." 42nd United States Congress.
http://www.fta.dot.gov/15154 225.html.
FTA. 2015a. "The Environmental Process." U.S. Department of Transportation, Accessed August 2015.
http://www.fta.dot.gov/15154 224.html.
FTA. 2015b. National Transit Database.
FTA and CATS. 2011. "Lynx Blue Line Extension, Northeast Corridor Light Rail Project, Charlotte-Mecklenburg
County, North Carolina, Final Environmental Impact Statement." Federal Transit Administration
Accessed August 31. http://charmeck.org/city/charlotte/cats/planning/BLE/Pages/feis-toc.aspx.
Fulton, Lew, and G. Eads. 2004. "IEA/SMP Model Documentation and Reference Case Projection." International
Energy Agency and World Business Council for Sustainable Development.
http://www.wbcsd.org/web/publications/mobility/smp-model-document.pdf.
Geographic Research Inc. 2015a. "Census Data 2000, ACS Estimates 2008-2014." Accessed June 2015 from
SimplyMap Database http://geographicresearch.com/simplymap/.
Geographic Research Inc. 2015b. "Total Retail Sales (2008-2014) by Census Block Group." Accessed March 2015
from SimplyMap Database, http://geographicresearch.com/simplymap/.
Gibson, Robert B. 2006. "Sustainability assessment: basic components of a practical approach." Impact
Assessment and Project Appraisal 24 (3): 170-182.
GoTriangle. 2015. "Durham-Orange Light Rail Transit Project Draft Environmental Impact Statement."
http://ourtransitfuture.com/deis/.
Green, Geoffrey. 2015. Comparo-2005-2035-2040.xlsx. Email Message from GoTriangle personnel to authors on
May 22, 2015.
Green, Jennifer. 2014. FY2014 TTA Stats by County.xlsx. Email message from Triangle Transit personnel to
authors on November 6, 2014.
Gulliver, J., and D. J. Briggs. 2004. "Personal exposure to particulate air pollution in transport
microenvironments." Atmospheric Environment 38 (l):l-8. doi: 10.1016/j.atmosenv.2003.09.036.
Heikkila, E., P. Gordon, J.I. Kim, R.B. Peiser, and H.W. Richarson. 1989. "What happened to the CBD-distane
gradient?: Land values in a policentric city." Environment and Planning A 21:221-232.
Highway Safety Research Center at UNC Chapel Hill. 2015. "NC Crash Data Query Web Site." Accessed June 15.
http://nccrashdata.hsrc.unc.edu/index.cfm.
Hodges-Copple, John. 2012. FINAL 2040 Population Forecast by County for CommViz -EPA. Email message from
TJCOG personnel to authors on November 26, 2014.
ICLEI. 2007. "City of Durham & Durham County greenhouse gas and criteria air pollutant emissions inventory and
local action plan for emission reductions." ICLEI Energy Services.
International Energy Agency and World Business Council for Sustainable Development. 2004. "IEA/SMP
Transportation Model." www.wbcsd.org/web/publications/mobility/smp-model-spreadsheet.xls.
International Energy Agency, Directorate of Sustainable Energy Policy and Technology. 2009. Transport, Energy,
and CC>2: Moving Toward Sustainability. Paris: International Energy Agency.
Jackson, TR. 2001a. "Fleet characterization data for MOBILE6: development and use of age distributions, average
annual mileage accumulation rates, and projected vehicle counts for use in MOBILE6." Office of
Transportation and Air Quality, U.S. Environmental Protection Agency.
368
-------
Jackson, Tracie. 2001b. "Fleet Characterization Data for MOBILES: Development and Use of Age Distributions,
Average Annual Mileage Accumulation Rates and Projected Vehicle Counts for Use in MOBILES." United
States Environmental Protection Agency.
Jackson, Tracie R. 2001c. "Fleet Characterization Data for MOBILES: Development and Use of Age Distributions,
Average Annual Mileage Accumulation Rates and Projected Vehicle Counts for Use in MOBILES." U.S.
Environmental Protection Agency.
Jay, Stephen, Carys Jones, Paul Slinn, and Christopher Wood. 2007. "Environmental impact assessment:
Retrospect and prospect." Environmental impact assessment review 27 (4):287-300.
Jerrett, Michael, Altaf Arain, Pavlos Kanaroglou, Bernardo Beckerman, Dimitri Potoglou, Talar Sahsuvaroglu,
Jason Morrison, and Chris Giovis. 2005. "A review and evaluation of intraurban air pollution exposure
models." Journal of Exposure Science and Environmental Epidemiology 15 (2):185-204.
Johansson, Olof, and Lee Schipper. 1997. "Measuring the long-run fuel demand of cars: Separate estimations of
vehicle stock, mean fuel intensity, and mean annual driving distance." Journal of Transport Economics
and Policy 31 (3):277-292.
Jud, G. Donald, and Daniel T. Winkler. 2002. "The dynamics of metropolitan housing prices." Journal of Real
Estate Research 23 (1/2): 29-45.
Kain, John F., and John M. Quigley. 1970. "Measuring the value of housing quality." Journal of the American
Statistical Association 65 (330):532-548.
Kaplan, Siena, and Elizabeth Ouzts. 2009. "Growing solar in North Carolina: Solar power's role in a clean energy
future." Environment North Carolina Research & Policy Center.
Keynes, John Maynard. 1936. General theory of employment, interest and money: Atlantic Publishers & Dist.
Kochtitzky, Chris S., H. Frumkin, R. Rodriguez, A. L. Dannenberg, J. Rayman, K. Rose, R. Gillig, T. Kanter, and
Prevention Centers for Disease Control and. 2006. "Urban planning and public health at CDC." MMWR.
Morbidity and mortality weekly report 55 Suppl 2.
Kockelman, K. M. 1997. "The effects of location elements on home purchase prices and rents: Evidence from the
San Francisco Bay Area." Transportation Research Record No. 1606:40-50.
Kotchen, M. J., and S. L. Schulte. 2009. "A meta-analysis of cost of community service studies." Int. Reg. Sci. Rev.
International Regional Science Review 32 (3):376-399.
Krewski, D, M Jerrett, RT Burnett, R Ma, E Hughes, Y Shi, MC Turner, CA Pope, G Thurston, EE Calle, and MJ
Thun. 2009. "Extended Follow-Up and Spatial Analysis of the American Cancer Society Study Linking
Particulate Air Pollution and Mortality." Health Effects Institute.
Lepeule, J, F Laden, D Dockery, and J Schwartz. 2012. "Chronic Exposure to Fine Particles and Mortality: An
Extended Follow-Up of the Harvard Six Cities Study from 1974 to 2009." Environmental Health
Perspectives 120 (7):965-970.
Litman, Todd. 2013. "Understanding Transport Demands and Elasticities: How Prices and Other Factors Affect
Travel Behavior." Victoria Transport Policy Institute, http://vtpi.org/elasticities.pdf.
Litman, Todd Alexander, and Eric Doherty. 2009. "Transportation Cost and Benefit Analysis: Techniques,
Estimates and Implications." Victoria Transport Policy Institute, http://www.vtpi.org/tca/.
Loo, Becky P. Y., and S. Y. Chow. 2006. "Sustainable urban transportation: Concepts, policies, and
methodologies." Journal of Urban Planning & Development 132 (2):76-79. doi: 10.1061/(asce)0733-
9488(2006)132:2(76).
McCollom, Brian E., and Richard H. Pratt. 2004. "TCRP Report 95: Traveler Response to Transportation System
Changes: Chapter 12—Transit Pricing and Fares." Transportation Research Board.
http://onlinepubs.trb.org/onlinepubs/tcrp/tcrp rpt 95cl2.pdf.
Minnesota Population Center. 2015. "Census Data 2000, ACS 1-yr and 5-yr Estimates 2008-2014." University of
Minnesota. Accessed May 2015 from National Historical Geographic Information System Version 2.0.
http://www.nhgis.org.
NC DHHS. 2015. "Resident Live Births, Deaths, and Estimated Net Migration 2000-2014." NC State Data Center,
Accessed from Log Into North Carolina (LINC) February 2015. http://linc.state.nc.us/.
NC DOR. 2003-2013. "Table 56. Summary of Local Sales and Use Tax Collections, Tax Allocations, and
Distributable Proceeds by County." Statistical Abstract of North Carolina Taxes, Part IV: Local
Government Sales And Use Tax Revenues, http://www.dornc.com/publications/abstract/.
NC DOR. 2004-2014. "County Taxable Real Property Valuations."
http://www.dornc.com/publications/taxable real property.html.
369
-------
NC DOR. 2013a. "Table 36A. State Sales and Use Tax: Gross Collections by County, FY 2000-2013." Statistical
Abstract of North Carolina Taxes, Part III: State Tax Collections.
http://www.dornc.com/publications/abstract/2013/table 36a.pdf.
NC DOR. 2013b. "Table 37A. State Sales and Use Tax: Retail Taxable Sales by County for FY 2000-2013."
Statistical Abstract of North Carolina Taxes, Part III: State Tax Collections.
http://www.dornc.com/publications/abstract/2013/table 37a.pdf.
NC DOR. 2013c. "Table 60A. Article 43 Local Government Sales and Use Taxes for Public Transportation for FY
2013." Statistical Abstract of North Carolina Taxes, Part IV: Local Government Sales And Use Tax
Revenues, http://www.dornc.com/publications/abstract/2013/table 60a.pdf.
NC DOR. 2014. "Gross Retail Sales for Durham and Orange County, FY 2014." Accessed from Log Into North
Carolina (LINC) December 2014. http://data.osbm.state.nc.us/pls/linc/dyn line topic reports.show.
NC DOT. 2007. Roadway Design Manual.
NC DOT. 2014. "Auto and Truck Registrations in Durham County and Orange County, 2014." Accessed from Log
Into North Carolina (LINC) February 2015.
http://data.osbm.state.nc.us/pls/linc/dyn line topic reports.show.
NC DST. 2015. "Local Government Revenues by Source: Local Option Sales Tax Revenues for Chapel Hill and the
City of Durham, FY 2000-2014." Accessed from Log Into North Carolina (LINC) February 2015.
http://data.osbm.state.nc.us/pls/linc/dyn line main.show.
NC ESC. 2014. "Local Area Unemployment Statistics for Durham and Orange County, NC." Accessed February
2015. http://esesc23.esc.state.nc.us/d4/LausSelection.aspx.
NC Office of State Budget and Management. 2015. "Population Estimates and Projections." Accessed February 9.
http://www.osbm.state.nc.us/ncosbm/facts and figures/socioeconomic data/population estimates.sh
tm.
NC State Data Center. 2015. "LINC: Log Into North Carolina." NC State Data Center Accessed August 19.
http://data.osbm.state.nc.us/pls/linc/dyn line main.show.
NCDENR. 2010. "Falls Lake Nutrient Management Strategy." N.C. Department of Environment and Natural
Resources Accessed September 8. http://portal.ncdenr.org/web/fallslake/read-the-rules.
NCDENR. 2011. Jordan Lake Stormwater Load Accounting Tool User's Manual. Raleigh, NC: North Carolina
Department of Environment and Natural Resources.
NCDENR. 2014. "Jordan Lake Rules." N.C. Department of Environment and Natural Resources Accessed
September 2. http://portal.ncdenr.org/web/iordanlake/home.
NCDENR. 2015. "Falls Lake and Jordan Lake monitoring." Division of Natural Resources, N.C. Department of
Environment and Natural Resources Accessed September 8.
http://portal.ncdenr.org/web/wq/fallsiordan.
NOAA. 2015. "Monthly summaries station details: Durham, NC US." National Centers for Environmental
Information, National Oceanographic and Atmospheric Administration Accessed September 8.
http://www.ncdc.noaa.gOV/cdo-web/datasets/GHCNDMS/stations/GHCND:USC00312515/detail.
North Carolina Center for County Research. 2015. "Basics of North Carolina Local Option Sales Taxes."
http://www.ncacc.org/DocumentCenter/View/1175.
North Carolina Sustainable Energy Association. 2015. Installed Solar Projects in NC. Raleigh, NC.
NuStats. 2006. "Greater Triangle Travel Study, Household Travel Survey Final Report." Prepared for: CAMPO,
DCHC MPO, NC DOT, Triangle Transit Authority.
http://www.dchcmpo.org/programs/regional/travel demand /.
Oleniacz, Laura. 2014. "Company puts last panel on Durham County solar farm." The Herald-Sun, May 12.
Orange County Tax Administration. 2014. Orange County Parcel Database (parview). edited by Land Records/GIS
Office. Orange County, NC.
Panek, Sharon D., Frank T. Baumgardner, and Matthew J. McCormick. 2007. "Introducing New Measures of the
Metropolitan Economy: Prototype GDP-by-Metropolitan-Area Estimates for 2001-2005." Survey of
Current Business 87 (11):79-114.
Pfaffenbichler, Paul. 2011. "Modelling with systems dynamics as a method to bridge the gap between politics,
planning and science? Lessons learnt from the development of the land use and transport model
MARS." Transport Reviews 31 (2):267-289. doi: 10.1080/01441647.2010.534570.
Probst, Gilbert, and Andrea M. Bassi. 2014. Tackling Complexity: A Systemic Approach for Decision Makers.
Sheffield, UK: Greenleaf Publishing.
370
-------
Ramsey, Kevin, and Alexander Bell. 2014. "Smart Location Database Version 2.0 User Guide." US EPA, Office of
Policy, Office of Sustainable Communities, http://www2.epa.gov/smartgrowth/smart-location-database-
technical-documentation-and-user-guide.
Rickwood, Peter, Damien Giurco, Garry Glazebrook, Alex Kazaglis, Leena Thomas, Michelle Zeibots, Spike
Boydell, Stuart White, Graziella Caprarelli, and Janet McDougal. 2007. "Integrating population, land-use,
transport, water and energy-use models to improve the sustainability of urban systems." State of
Australian Cities Conference.
Rosen, Kenneth T., and Lawrence B. Smith. 1983. "The price-adjustment process for rental housing and the
natural vacancy rate." The American Economic Review 73 (4):779-786.
Shaver, E, R Horner, J Skupien, and C May. 2007. Fundamentals of urban runoff management: technical and
institutional issues. 2nd ed. Madison, Wl: North American Lake Management Society.
Shepherd, SP. 2014. "A review of system dynamics models applied in transportation." Transportmetrica B:
Transport Dynamics 2 (2):83-105.
Sinha, Kumares Chandra, and Samuel Labi. 2007. Transportation Decision Making: Principles of Project
Evaluation and Programming. Hoboken, NJ: John Wiley & Sons.
Song, Van, and Daniel A. Rodriguez. "The measurement of the level of mixed land uses: A synthetic approach."
Carolina Transportation Program White Paper Series:l-38.
Srour, Issam M., Kara M. Kockelman, and Travis P. Dunn. 2002. "Accessibility Indices: Connection to Residential
Land Prices and Location Choices." Transportation Research Record: Journal of the Transportation
Research Board 1805 (2002 Travel Demand and Land Use):25-34. doi: 10.3141/1805-04.
State Climate Office of North Carolina. 2015. "Historical Data." Accessed August 31. http://www.nc-
climate.ncsu.edu/climate/drought/historical.
Stave, Krystyna. 2002. "Using system dynamics to improve public participation in environmental decisions."
System Dynamics Review 18 (2):139-167.
Stave, Krystyna. 2010. "Participatory system dynamics modeling for sustainable environmental management:
Observations from four cases." Sustainability 2 (9):2762-2784.
Sterman, John D. 2000. Business Dynamics. New Delhi: McGraw-Hill Companies, Inc.
The Center for Watershed Protection. 2013. "Cost-effectiveness study of urban stormwater BMPs in the James
River Basin." James River Association.
The Orange County Financial Services Department. 2009, 2014. "Comprehensive Annual Financial Report."
http://www.orangecountync.gov/departments/financial audit.php.
TJCOG. 2012. "Triangle regional water supply plan, volume I: regional needs assessment."
TJCOG. 2013. "Imagine 2040: The Triangle Region Scenario Planning Initiative Final Summary Document."
http://www.ticog.org/imagine2040/downloads.aspx.
TJCOG. 2014a. "CommunityViz 2 (CV2) Parcel Geodatabase for Place Type & Development Status Editing."
Accessed Nov, 2014. http://www.ticog.org/imagine2040/downloads.aspx.
TJCOG. 2014b. "CommunityViz 2 (CV2) Parcel Geodatabase for Place Type & Development Status Editing."
Accessed November, 2014. http://www.ticog.org/imagine2040/downloads.aspx.
TJCOG. 2014c. "CommunityViz 2 (CV2) Parcel Geodatabase for Place Type & Development Status Editing."
Accessed November 2014. http://www.ticog.org/imagine2040/downloads.aspx.
TJCOG. 2014d. "Imagine 2040 Results GIS Data." Accessed November 2014.
http://www.ticog.org/imagine2040/downloads.aspx.
TJCOG. 2014e. "Triangle CommunityViz2: Place Type & Development Status-Key Points."
http://ticog.Org/Data/Sites/l/media/imagine-2040/cv2/triangle-cv2—pt-and-ds-kev-points.pdf.
TJCOG. 2014f. "Triangle regional water supply plan, volume II: regional water supply alternatives analysis."
Town of Chapel Hill Business Management Department. 2009, 2014. "Comprehensive Annual Financial Report."
http://www.townofchapelhill.org/town-hall/departments-services/business-management/financial-
statements.
Triangle J Council of Governments. 2012. "Triangle regional water supply plan, volume I: regional needs
assessment." TJCOG.
Triangle Transit. 2012a. "Alternatives Analysis Final Report: Durham-Orange County Corridor." Prepared by URS.
Triangle Transit. 2012b. "Scoping Report: Durham-Orange Light Rail Transit Project."
http://ourtransitfuture.com/do scoping report/.
Triangle Transit. 2014. "Durham and Orange County Bus and Rail Investment Progress Reports for FY2014." Our
Transit Future, http://ourtransitfuture.com/durham-county-bus-and-rail-investment-plan/.
371
-------
Triangle Transit. 2015a. "FY 2015 Annual Budget & Capital Investment Plan: XI. Durham-Orange Bus and Rail
Investment Plan." http://www.triangletransit.org/content/publications.
Triangle Transit. 2015b. "Our Transit Future." Accessed February 9. http://ourtransitfuture.com/.
Triangle Transit, Chapel Hill Transit, Orange County, DCHC MPO, and Orange County Public Transportation. 2012.
"The Bus and Rail Investment Plan in Orange County." http://ourtransitfuture.com/durham-county-bus-
and-rail-investment-plan/.
TRM Service Bureau. 2008. "User's Guide: TRM LRTP Evaluation Measures Tool V.16."
TRM Service Bureau and TRM Team. 2012. "Triangle Regional Model Version 5 Model Documentation Report."
Prepared for: NC DOT, Triangle Transit, CAMPO, DCHC MPO.
http://www.dchcmpo.org/programs/regional/travel demand /.
U.S. Census Bureau. 2000. "Census 2000, Summary File 3." Accessed from American FactFinder May 2015.
http://factfinder2.census.gov.
U.S. Census Bureau. 2013. "National Poverty Standards for 2013 by Size of Family and Number of Related
Children Under 18 Years." Accessed May 2015.
https://www.census.gov/hhes/www/povertv/data/threshld/.
U.S. Census Bureau. 2015. "LODES Data. Longitudinal Employer-Household Dynamics Program." U.S. Census
Bureau, Center for Economic Studies. Accessed March 2015. http://lehd.ces.census.gov/data/lodes/.
U.S. Census Bureau American Community Survey. 2014. "American Community Survey 1-yr and 5-yr Estimates
2005-2013." Accessed from American FactFinder May 2015. http://factfinder2.census.gov.
US DOE. 2008. "Energy efficiency trends in residential and commercial buildings." U.S. Department of Energy.
http://appsl.eere.energy.gov/buildings/publications/pdfs/corporate/bt stateindustry.pdf.
US DOE. 2011. 2010 Buildings Energy Data Book. Washington, D.C.: U.S. Department of Energy.
US DOT. 2013. "New and Small Starts evaluation and rating process final policy guidance." U.S. Department of
Transportation. http://www.fta.dot.gov/documents/NS-SS Final PolicvGuidance August 2013.pdf.
US DOT, Federal Highway Administration. 2009. "2009 National Household Travel Survey." http://nhts.ornl.gov.
US DOT, Office of Policy, Office of Transportation Policy, Office of Economic and Strategic Analysis. 2014. The
Value of Travel Time Savings: Departmental Guidance for Conducting Economic Evaluations Revision 2
(2014 Update). Washington, DC.
US EIA. 2009. "2009 RECS Survey Data." U.S. Energy Information Administration Accessed August 31.
http://www.eia.gov/consumption/residential/data/2009/.
US EIA. 2015a. "Annual Energy Outlook 2010-2015." U.S. Energy Information Administration Accessed August 19.
http://www.eia.gov/forecasts/aeo/.
US EIA. 2015b. "Electricity: Sales (consumption), revenue, prices & customers." U.S. Energy Information
Administration Accessed September 9. http://www.eia.gov/electricity/data.cfmtfsales.
US EIA. 2015c. "North Carolina Price of Natural Gas Delivered to Residential Customers." U.S. Energy Information
Administration Accessed September 9. http://www.eia.gov/dnav/ng/hist/n3010nc3A.htm.
US EIA. 2015d. "Weekly Retail Gasoline and Diesel Prices." U.S. Energy Information Administration Accessed
September 9. http://www.eia.gov/dnav/pet/pet pri gnd dcus nus a.htm.
US EPA. 2012. "Sustainable and Healthy Communities: Strategic Research Action Plan 2012-2016." US
Environmental Protection Agency, http://www.epa.gov/research/docs/shc-strap.pdf.
US EPA. 2013a. Analysis of the Life Cycle Impacts and Potential for Avoided Impacts Associated with Single-
Family Homes. Washington, DC.
US EPA. 2013b. "Sustainability Analytics: Assessment Tools & Approaches." U.S. Environmental Protection
Agency, http://www.epa.gov/sustainabilitv/analvtics/docs/sustainabilitv-analytics.pdf.
US EPA. 2013c. "Technical support document: estimating the benefit per ton of reducing PM2.s precursors from
17 sectors." U.S. Environmental Protection Agency, Office of Air and Radiation.
US EPA. 2015a. "Carbon Pollution Emission Guidelines for Existing Stationary Sources: Electric Utility Generating
Units." U.S. Environmental Protection Agency. Accessed August 24.
http://www2.epa.gov/sites/production/files/2015-08/documents/cpp-final-rule.pdf.
US EPA. 2015b. "Clean Energy: Calculations and References." U.S. Environmental Protection Agency Accessed
August 19. http://www.epa.gov/cleanenergv/energv-resources/refs.html.
US EPA. 2015c. "Clean Power Plan: State at a Glance: North Carolina." U.S. Environmental Protection Agency.
Accessed August 24. http://www.epa.gov/airqualitv/cpptoolbox/north-carolina.pdf.
US EPA. 2015d. "Emission Inventories." U.S. Environmental Protection Agency. Accessed August 31.
http://www.epa.gov/ttn/chief/eiinformation.html.
372
-------
US EPA. 2015e. "EnviroAtlas -Durham, NC- One Meter Resolution Urban Area Land Cover Map (2010)." U.S.
Environmental Protection Agency. Accessed August 31.
https://edg.epa.gov/metadata/rest/document?id=%7b2FF66877-A037-4693-9718-
D1870AA3F084%7d&xsl=metadata to html full.
US EPA. 2015f. "Overview of the Clean Power Plan: cutting carbon pollution from power plants." U.S.
Environmental Protection Agency, http://www2.epa.gov/sites/production/files/2015-08/documents/fs-
cpp-overview.pdf.
US EPA. 2015g. "Regulations & standards: light duty." U.S. Environmental Protection Agency. Accessed
September 8. http://www.epa.gov/otaq/climate/regs-light-duty.htm.
US EPA, Office of Policy, Office of Sustainable Communities. 2013d. Smart Location Database version 2.0.
USGS. 2015. "Water use in North Carolina: 2005 data tables." South Atlantic Water Science Center - North
Carolina office. Accessed September 8.
http://nc.water.usgs.gov/infodata/wateruse/data/Data Tables 2005.html.
Vina-Arias, Laura Beatriz. 2013. "Understanding Patterns of Growth at Kendall Square Using a System Dynamics
Approach." Master of Science in Transportation and Master in City Planning, Department of Civil and
Environmental Engineering and Department of Urban Studies and Planning, Massachusetts Institute of
Technology.
Wang, Tingting, and Cynthia Chen. 2014. "Impact of fuel price on vehicle miles traveled (VMT): Do the poor
respond in the same way as the rich?" Transportation 41 (1):91-105. doi: 10.1007/sllll6-013-9478-l.
Washburn, Barbara, Katie Yancey, and Jonathan Mendoza. 2010. User's Guide for the California Impervious
Surface Coefficients, edited by Office of Environmental Health Hazard Assessment Ecotoxicology
Program: California Environmental Protection Agency.
Wegener, Michael. 2004. "Overview of land-use transport models." Handbook of Transport Geography and
Spatial Systems 5:127-146.
Weston, Joe. 2004. "EIA in a risk society." Journal of environmental planning and management 47 (2):313-325.
WHO. 2014. "Health economic assessment tools (HEAT) for walking and for cycling: Methodology and user
guide, 2014 update." http://www.euro.who.int/en/health-topics/environment-and-health/Transport-
and-health/publications/2011/health-economic-assessment-tools-heat-for-walking-and-for-cycling.-
methodologv-and-user-guide.-economic-assessment-of-transport-infrastructure-and-policies.-2014-
update.
Wilson, WG. 2011. Constructed climates: a primer on urban environments. Chicago, IL: The University of Chicago
Press.
Woods & Poole Economics Inc. Copyright 2014. Durham and Orange County, NC Data Pamphlet. Washington,
D.C.
Woolfolk, Michelle. 2015. Personal communication with authors on July 17, 2015.
Zillow Real Estate Research. 2010-2014. "Median Rent by County and Zip Code." Accessed March 2015.
http://www.zillow.com/research/data/.
373
-------
United States
Environmental Protection
Agency
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35
Office of Research and Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300
------- |