Regional Transportation Models
Click HERE for graphic.
Board of Directors
SAN DIEGO ASSOCIATION OF GOVERNMENTS
The San Diego Association of Governments (SANDAG) is a public
agency formed voluntarily by local governments to assure overall
areawide planning and coordination for the San Diego region.
Voting members include the incorporated Cities of Carlsbad, Chula
Vista, Coronado, Del Mar, El Cajon, Encinitas, Escondido, Imperial
Beach, La Mesa, Lemon Grove, National City, Oceanside, Poway, San
Diego, San Marcos, Santee, Solana Beach, Vista, and the County of San
Diego. Advisory and Liaison members include Caltrans, U.S. Department
of Defense, San Diego Unified Port District, and Tijuana/Baja
California/Mexico.
CHAIRMAN: Hon. Mike Bixler
VICE CHAIRMAN: Hon. Elliot Parks
SECRETARY-EXECUTIVE DIRECTOR: Kenneth E. Sulzer
CITY OF CARLSBAD
Hon. Bud Lewis, Mayor
(A) Hon. Ann Kulchin, Councilmember
(A) Hon. Julianne Nygaard, Mayor Pro Tem
CITY OF CHULA VISTA
Hon. Shirley Horton, Mayor
(A) Hon. Jerry Rindone, Mayor Pro Tem
CITY OF CORONADO
Hon. Mary Herron, Mayor
(A) Hon. David Blumenthal, Councilmember
CITY OF DEL MAR
Hon. Elliot Parks, Councilmember
(A) Hon. Mark Whitehead, Councilmember
(A) Hon. Henry Abarbanel, Deputy Mayor
CITY OF EL CAJON
Hon. Richard Ramos, Councilmember
(A) Hon. Mark Lewis, Councilmember
CITY OF ENCINITAS
Hon. Gail Hano, Councilmember
(A) Vacant
CITY OF ESCONDIDO
Hon. Jerry Harmon, Councilmember
(A) Hon. Lori Holt Pfeiler, Councilmember
CITY OF IMPERIAL BEACH
Hon. Mike Bixler, Mayor
(A) Hon. Gail Benda, Councilmember
CITY OF LA MESA
Hon. Art Madrid, Mayor
(A) Hon. Barry Jantz, Councilmember
(A) Hon. Jay LaSuer, Councilmember
CITY OF LEMON GROVE
Hon. Jerome Legerton, Mayor Pro Tem
(A) Hon. Craig Lake, Councilmember
CITY OF NATIONAL CITY
Hon. Rosalie Zarate, Councilmember
(A) Vacant
CITY OF OCEANSIDE
Hon. Dick Lyon, Mayor
(A) Hon. Colleen O'Harra, Deputy Mayor
CITY OF POWAY
Hon. Don Higginson, Mayor
(A) Hon. Bob Emery, Councilmember
(A) Hon. Mickey Cafagna, Councilmember
CITY OF SAN DIEGO
Hon. Judy McCarty, Councilmember
(A) Hon. Barbara Warden, Councilmember
(A) Hon. Valerie Stallings, Councilmember
CITY OF SAN MARCOS
Hon. Lee Thibadeau, Mayor
(A) Hon. Mark Loscher, Councilmember
CITY OF SANTEE
Hon. Jack Dale, Mayor
(A) Hon. Hal Ryan, Councilmember
CITY OF SOLANA BEACH
Hon. Marion Dodson, Deputy Mayor
(A) Hon. Teri Renteria, Mayor
(A) Hon. Joe Kellejian, Councilmember
CITY OF VISTA
Hon. Gloria E. McClellan, Mayor
(A) Hon. Ed Estes, Councilmember
COUNTY OF SAN DIEGO
Vacant
(A) Hon. Pam Slater, Chair
(A) Vacant
STATE DEPT. OF TRANSPORTATION
(Advisory Member)
James van Loben Sels, Director
(A) Gary Gallegos, District 11 Director
U.S. DEPARTMENT OF DEFENSE
(Liaison Member)
CAPT. Tom Gunn, CEC, USN
Commanding Officer Southwest Division
Naval Facilities Engineering Command
SAN DIEGO UNIFIED PORT DISTRICT
(Advisory Member)
Jess Van Deventer, Commissioner
TIJUANA/BAJA CALIFORNIA/MEXICO
(Advisory Member)
Hon. Hector G. Osuna Jaime
Presidente Municipal de Tijuana
Revised January 6, 1995
ABSTRACT
TITLE: Regional Transportation Models
AUTHOR: San Diego Association of Governments
DATE: January 1995
SOURCE OF COPIES: San Diego Association of Governments
401 B Street, Suite 800
San Diego, CA 92101
(619) 595-5300
NUMBER OF PAGES: 354
ABSTRACT: This report describes transportation
modeling procedures that are
currently used by the San Diego
Association of Governments to produce
regional highway and public transit
travel demand forecasts for the years
1990 to 2015. The operation of each
step of the modeling process is de-
scribed, along with data
requirements, products produced,
calibration procedures, and data file
formats.
ACKNOWLEDGEMENTS
The following staff of the San Diego Association of Governments
contributed to the preparation of this document.
Kenneth E. Sulzer, Executive Director
Stuart R. Shaffer, Deputy Executive Director
Bob Parrott, Director of Research
Lee F. Hultgren, Director of Transportation
Bill McFarlane, Senior Transportation Planner
Michael Hix, Senior Transportation Planner
Jeff Martin, Senior Research Planner
Dan Hildebrand, Assistant Transportation Planner
Andrew Abouna, Assistant Transportation Planner
Mike Calandra, Senior Transportation Technician
Julie Jamarta, Senior Transit Technician
TABLE OF CONTENTS
CHAPTER 1 INTRODUCTION. . . . . . . . . . . . . . . . . . . . .3
Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . .3
Software. . . . . . . . . . . . . . . . . . . . . . . . . . . .4
Hardware. . . . . . . . . . . . . . . . . . . . . . . . . . . .5
Process . . . . . . . . . . . . . . . . . . . . . . . . . . . .5
New Developments. . . . . . . . . . . . . . . . . . . . . . . .11
Model Results . . . . . . . . . . . . . . . . . . . . . . . . .13
CHAPTER 2 ZONE SYSTEM . . . . . . . . . . . . . . . . . . . . .19
Data File Documentation . . . . . . . . . . . . . . . . . . . .29
CHAPTER 3 SURVEYS . . . . . . . . . . . . . . . . . . . . . . .35
Travel Behavior Survey. . . . . . . . . . . . . . . . . . . . .35
Regional Transit Survey . . . . . . . . . . . . . . . . . . . .36
External Trip Surveys . . . . . . . . . . . . . . . . . . . . .38
1991 Visitor Survey . . . . . . . . . . . . . . . . . . . . . .39
Traffic Counts. . . . . . . . . . . . . . . . . . . . . . . . .39
Transit Passenger Counts. . . . . . . . . . . . . . . . . . . .40
Data File Documentation . . . . . . . . . . . . . . . . . . . .43
CHAPTER 4 GROWTH FORECASTS. . . . . . . . . . . . . . . . . . .61
Input Assumptions . . . . . . . . . . . . . . . . . . . . . . .62
Regional Growth Control Totals. . . . . . . . . . . . . . . . .65
Sub-Regional Employment Allocation. . . . . . . . . . . . . . .65
Sub-Regional Residential Allocation . . . . . . . . . . . . . .66
MGRA Allocation . . . . . . . . . . . . . . . . . . . . . . . .68
Data File Documentation . . . . . . . . . . . . . . . . . . . .71
CHAPTER 5 TRANSPORTATION NETWORKS . . . . . . . . . . . . . . .85
Data File Documentation . . . . . . . . . . . . . . . . . . . .91
CHAPTER 6 HIGHWAY NETWORKS. . . . . . . . . . . . . . . . . . .103
Network Procedures. . . . . . . . . . . . . . . . . . . . . . .105
Capacity. . . . . . . . . . . . . . . . . . . . . . . . . . . .106
Travel Time . . . . . . . . . . . . . . . . . . . . . . . . . .111
Tranplan Input File . . . . . . . . . . . . . . . . . . . . . .112
Turn Prohibitor File. . . . . . . . . . . . . . . . . . . . . .114
Tranplan Highway Network File . . . . . . . . . . . . . . . . .114
Tranplan Zone-to-Zone Travel Time Files . . . . . . . . . . . .114
Data File Documentation . . . . . . . . . . . . . . . . . . . .119
CHAPTER 7 TRANSIT NETWORKS. . . . . . . . . . . . . . . . . . .131
Input Files . . . . . . . . . . . . . . . . . . . . . . . . . .133
Network Processing. . . . . . . . . . . . . . . . . . . . . . .135
Network Validation. . . . . . . . . . . . . . . . . . . . . . .141
Access Procedures . . . . . . . . . . . . . . . . . . . . . . .143
Data File Documentation . . . . . . . . . . . . . . . . . . . .147
CHAPTER 8 TRIP GENERATION . . . . . . . . . . . . . . . . . . .163
Model Structure . . . . . . . . . . . . . . . . . . . . . . . .164
Model Outputs . . . . . . . . . . . . . . . . . . . . . . . . .169
Model Calibration . . . . . . . . . . . . . . . . . . . . . . .172
Data File Documentation . . . . . . . . . . . . . . . . . . . .179
CHAPTER 9 TRIP DISTRIBUTION . . . . . . . . . . . . . . . . . .199
Model Structure . . . . . . . . . . . . . . . . . . . . . . . .199
Model Outputs . . . . . . . . . . . . . . . . . . . . . . . . .200
Model Calibration . . . . . . . . . . . . . . . . . . . . . . .202
Data File Documentation . . . . . . . . . . . . . . . . . . . .215
CHAPTER 10 VEHICLE TRIP FACTORING . . . . . . . . . . . . . . .229
Model Structure . . . . . . . . . . . . . . . . . . . . . . . .229
Model Outputs . . . . . . . . . . . . . . . . . . . . . . . . .232
Model Calibration . . . . . . . . . . . . . . . . . . . . . . .233
Data File Documentation . . . . . . . . . . . . . . . . . . . .237
CHAPTER 11 MODE CHOICE. . . . . . . . . . . . . . . . . . . . .241
Model Structure . . . . . . . . . . . . . . . . . . . . . . . .241
Model Outputs . . . . . . . . . . . . . . . . . . . . . . . . .259
Model Calibration . . . . . . . . . . . . . . . . . . . . . . .260
Data File Documentation . . . . . . . . . . . . . . . . . . . .273
CHAPTER 12 EXTERNAL TRIPS . . . . . . . . . . . . . . . . . . .283
Model Structure . . . . . . . . . . . . . . . . . . . . . . . .283
Data File Documentation . . . . . . . . . . . . . . . . . . . .289
CHAPTER 13 HIGHWAY ASSIGNMENT . . . . . . . . . . . . . . . . .293
Model Structure . . . . . . . . . . . . . . . . . . . . . . . .293
Post-Assignment Processing. . . . . . . . . . . . . . . . . . .297
Model Outputs . . . . . . . . . . . . . . . . . . . . . . . . .299
Calibration . . . . . . . . . . . . . . . . . . . . . . . . . .301
Data File Documentation . . . . . . . . . . . . . . . . . . . .311
CHAPTER 14 TRANSIT ASSIGNMENT . . . . . . . . . . . . . . . . .315
Model Structure . . . . . . . . . . . . . . . . . . . . . . . .315
Model Outputs . . . . . . . . . . . . . . . . . . . . . . . . .316
Model Calibration . . . . . . . . . . . . . . . . . . . . . . .317
CHAPTER 15 MOTOR VEHICLE EMISSION MODELING. . . . . . . . . . .323
Model Structure . . . . . . . . . . . . . . . . . . . . . . . .326
Data File Documentation . . . . . . . . . . . . . . . . . . . .335
LIST OF TABLES
Table 1-1 Computer Resource Requirements. . . . . . . . . . . . 7
Table 1-2 Historical Demographic and Travel Indicators. . . . .14
Table 1-3 Forecasts of Demographic and Travel Indicators. . . .15
Table 3-1 Travel Behavior Survey Expansion Factors. . . . . . .37
Table 4-1 Series 8 Employment by Major Statistical Area . . . .67
Table 4-2 Series 8 Dwelling Units by Major Statistical Area . .68
Table 6-1 Highway Network Summary . . . . . . . . . . . . . . 103
Table 6-2 Default Roadway Attributes. . . . . . . . . . . . . 106
Table 6-3 Green-to-Cycle Time Ratios. . . . . . . . . . . . . 109
Table 6-4 Turn Lane Capacities. . . . . . . . . . . . . . . . 110
Table 6-5 Default Capacities. . . . . . . . . . . . . . . . . 110
Table 6-6 Tranplan Highway Inputs . . . . . . . . . . . . . . 112
Table 6-7 Assignment Group Definitions. . . . . . . . . . . . 113
Table 7-1 Transit Network Summary . . . . . . . . . . . . . . 131
Table 7-2 Transit Modes . . . . . . . . . . . . . . . . . . . 134
Table 7-3 Transit Company Descriptions. . . . . . . . . . . . 138
Table 7-4 Transit Fares by Company. . . . . . . . . . . . . . 140
Table 7-5 Light Rail Transit Fares. . . . . . . . . . . . . . 141
Table 7-6 Distribution of Route Time Errors . . . . . . . . . 141
Table 7-7 Assignment of Survey Transit Trips. . . . . . . . . 142
Table 8-1 Trip End Balancing Factors. . . . . . . . . . . . . 170
Table 8-2 Total Person Trips by Major Statistical Area. . . . 171
Table 8-3 Person Trips by Trip Type . . . . . . . . . . . . . 171
Table 8-4 Person Trip Rate Summary. . . . . . . . . . . . . . 172
Table 8-5 Observed and Estimated Total Person Trips . . . . . 174
Table 8-6 Observed and Estimated Home-Work Person Trips . . . 174
Table 8-7 Observed and Revised Estimated Home-Work
Person Trips. . . . . . . . . . . . . . . . . . . . 175
Table 8-8 Trip Generation Root Mean Square Error. . . . . . . 175
Table 9-1 Person Trip Lengths . . . . . . . . . . . . . . . . 201
Table 9-2 Commute Person Trips Between
Major Statistical Areas . . . . . . . . . . . . . . 203
Table 9-3 Total Person Trips Between
Major Statistical Areas . . . . . . . . . . . . . . 204
Table 9-4 Attraction Error Rates by Iteration . . . . . . . . 206
Table 9-5 Observed and Estimated Screenline Traffic Volumes . 206
Table 9-6 Observed and Estimated Intra-zonal Person Trip
Percentages . . . . . . . . . . . . . . . . . . . . 208
Table 9-7 Observed and Estimated Person Trip Lengths. . . . . 209
Table 9-8 Observed and Estimated Commute Person Trips
Between Major Statistical Areas . . . . . . . . . . 210
Table 9-9 Observed and Estimated Total Person Trips
Between Major Statistical Areas . . . . . . . . . . 211
Table 10-1 Time of Day Factors . . . . . . . . . . . . . . . . 231
Table 10-2 Directional Vehicle Trip Factors. . . . . . . . . . 232
Table 11-1 Mode Choice Coefficients. . . . . . . . . . . . . . 245
Table 11-2 Mode Constants. . . . . . . . . . . . . . . . . . . 253
Table 11-3 Income Constants. . . . . . . . . . . . . . . . . . 253
Table 11-4 Trip Length Constants . . . . . . . . . . . . . . . 254
Table 11-5 Transit Commute Major Statistical Area
Adjustment Factors. . . . . . . . . . . . . . . . . 256
Table 11-6 Transit Non-Commute Major Statistical Area
Adjustment Factors. . . . . . . . . . . . . . . . . 257
Table 11-7 Vehicle Occupancies for 3+ Person Autos . . . . . . 258
Table 11-8 Trips by Mode . . . . . . . . . . . . . . . . . . . 260
Table 11-9 Commute Transit Trips Between Major
Statistical Areas . . . . . . . . . . . . . . . . . 261
Table 11-10 Total Transit Trips Between Major
Statistical Areas . . . . . . . . . . . . . . . . . 262
Table 11-11 Observed and Estimated Trips by Mode. . . . . . . . 265
Table 11-12 Observed and Estimated Income Distribution
by Mode . . . . . . . . . . . . . . . . . . . . . . 266
Table 11-13 Observed and Estimated Trip Length Distribution
by Mode . . . . . . . . . . . . . . . . . . . . . . 267
Table 11-14 Observed and Estimated Commute Transit Trips
Between Major Statistical Areas . . . . . . . . . . 268
Table 11-15 Observed and Estimated Total Transit Trips
Between Major Statistical Areas . . . . . . . . . . 269
Table 12-1 External Vehicle Trips. . . . . . . . . . . . . . . 285
Table 13-1 Speeds by Volume/Capacity Ratio . . . . . . . . . . 294
Table 13-2 Signal Delay Times by Volume/Capacity Ratio . . . . 295
Table 13-3 Off-Peak Period Adjusted to Input Speed Ratios. . . 296
Table 13-4 Peak Period Adjusted to Input Speed Ratios. . . . . 297
Table 13-5 Vehicle Miles of Travel . . . . . . . . . . . . . . 300
Table 13-6 Average Speeds After Assignment . . . . . . . . . . 301
Table 13-7 Assignment Error by Number of Iterations. . . . . . 302
Table 13-8 Assignment Differences by Number of Iterations. . . 303
Table 13-9 Observed and Estimated Vehicle Miles of Travel. . . 304
Table 13-10 Root Mean Square Error by Functional Class. . . . . 305
Table 13-11 Root Mean Square Error by Volume Group. . . . . . . 306
Table 13-12 VMT Error by Functional Class . . . . . . . . . . . 307
Table 13-13 VMT Error by Volume Group . . . . . . . . . . . . . 308
Table 14-1 Transit Boardings . . . . . . . . . . . . . . . . . 316
Table 14-2 Access/Egress Mode Use. . . . . . . . . . . . . . . 317
Table 14-3 Observed and Estimated Transit Boardings. . . . . . 318
Table 14-4 Observed and Estimated Access/Egress
Mode Use . . . . . . . . . . . . . . . . . . . . . 319
Table 15-1 Hot Soak Emission Rates . . . . . . . . . . . . . . 325
Table 15-2 Vehicle Classification Groups . . . . . . . . . . . 327
Table 15-3 Simplified Temperature Assumptions. . . . . . . . . 327
Table 15-4 Motor Vehicle Emission Forecasts. . . . . . . . . . 330
LIST OF FIGURES
Figure 1-1 Transportation Modeling Process . . . . . . . . . . . 6
Figure 2-1 Block Split Example . . . . . . . . . . . . . . . . .21
Figure 2-2 MGRA and Zone Relationship. . . . . . . . . . . . . .21
Figure 2-3 Series 8 Zones. . . . . . . . . . . . . . . . . . . .22
Figure 2-4 External Zones. . . . . . . . . . . . . . . . . . . .23
Figure 2-5 Subregional Areas . . . . . . . . . . . . . . . . . .24
Figure 2-6 Major Statistical Areas . . . . . . . . . . . . . . .25
Figure 2-7 Average Vehicle Ridership Zones . . . . . . . . . . .26
Figure 6-1 Highway Networks. . . . . . . . . . . . . . . . . . 104
Figure 7-1 Transit Networks. . . . . . . . . . . . . . . . . . 132
Figure 8-1 Unique Generators . . . . . . . . . . . . . . . . . 167
Figure 9-1 Highway Screenlines . . . . . . . . . . . . . . . . 207
Figure 11-1 Parking Costs . . . . . . . . . . . . . . . . . . . 247
Figure 11-2 Walking Time. . . . . . . . . . . . . . . . . . . . 249
Figure 15-1 Air Quality Modeling Grids. . . . . . . . . . . . . 324
CHAPTER 1
INTRODUCTION
INTRODUCTION
The ability of San Diego's street, freeway, and public transportation
systems to accommodate travel in the face of continuing population
growth is a concern of many San Diego residents. Another issue is the
effectiveness of proposed transportation facilities in attracting new
users and reducing congestion. Other concerns deal with side effects
of large-scale transportation projects on San Diego's quality of life,
such as increased air pollution from vehicular emissions.
Transportation models have been developed to help answer these and
other questions. Models are computerized procedures for
systematically predicting travel changes in response to changes in
development patterns, transportation systems, and demographics given
certain assumptions about travel behavior based upon existing
conditions.
The many factors affecting travel make manual analysis prohibitively
time consuming for almost all applications. The complexity of the
problem can quickly exhaust computer capacity as well so that the
design of transportation models requires compromises on the level of
detail and number of factors that can be considered.
The last several years have seen increasing demands placed upon
transportation models. Part of this increased demand is the result of
recent air quality and congestion management legislation which
mandates transportation model analysis. In addition, as computers
become more ingrained in our culture, there is greater reliance placed
upon computer-generated data in any decision-making process.
Users should remember that transportation models are primarily
accounting tools and provide limited insight into the "right"
decision. The main advantage of a model is that it provides a
systematic analysis process so that alternatives can be evaluated in
an even-handed manner.
SETTING
San Diego has one transportation model that is used for all
applications in the Region. A single source of transportation data
not only reduces discrepancies, it also enables data files to be
better maintained, an important factor when many different users
utilize the data in a variety of ways. Model users include:
þ SANDAG (regional planning)
þ Caltrans (freeway planning)
þ Transit agencies (bus and rail patronage studies)
þ Local jurisdictions (circulation element studies)
þ Developers (site-specific impact reports)
þ Air Pollution Control District (vehicle emissions)
Forecasting traffic volumes in response to land use changes is
probably the most commonly requested model application. Proposed land
use changes can be quite detailed, such as a site-specific project
that a developer may need analyzed as part of an environmental impact
report. At the other extreme are generalized regionwide growth
alternatives.
Forecasting traffic volumes on proposed freeways and streets is
another common application. Typically, a study looks at several
alternative alignments for a proposed road. Models are run to
determine the differences in traffic volumes between alternatives.
Traffic impacts upon nearby roads are often a concern as well.
Models are frequently used to forecast transit patronage for light
rail extensions. These studies usually involve the evaluation of
alternative transit improvements in a corridor. For example, the
performance of improved express bus service might be compared to the
performance of new rail service. Models produce forecasts of transit
ridership on specific routes, overall transit trips, and traffic
volume impacts that assist in the selection of the best alternative.
SOFTWARE
SANDAG uses a transportation planning computer package called
Tranplan, distributed by the Urban Analysis Group. Tranplan provides
a framework for performing much of the computer processing involved
with modeling. Tranplan functions have numerous user options that
enable different urban areas to tailor the generalized software to
their specific needs.
Another software package used extensively in the modeling process is
Arc/Info, distributed by Environmental Systems Research Institute,
Inc. This geographic information system (GIS) maintains, manipulates,
and displays transportation, land use, and demographic data. The
power of Arc/Info lies in its ability to relate data contained in
different geographic files for shared uses. Arc/Info also has
enhanced computer mapping capabilities.
The two software vendors are working on developing common databases.
Until these procedures become operational, SANDAG Fortran programs
provide the linkage between Tranplan and Arc/Info. Data maintained in
Arc/Info is translated into temporary Tranplan input files. Results
from Tranplan are fed back to Arc/Info for generating plots and
reports. Other SANDAG programs manipulate data and perform some
modeling functions such as trip generation and mode choice.
SANDAG has extensive experience with both Tranplan and Arc/Info.
SANDAG has used Tranplan since 1981 for a wide range of modeling
applications. Arc/Info was first installed at SANDAG in 1985.
Tranplan and Arc/Info have been used in conjunction for transportation
modeling since 1987.
HARDWARE
SANDAG's transportation modeling and database maintenance is performed
on SUN workstations. SPARC-10 and SPARC-20 workstations are dedicated
exclusively for transportation modeling. Seven other SUN workstations
are shared between transportation and other agency users. Each work
station is accompanied by a two giga-byte disk drive for data storage.
Outside modelers use either workstations or PCs. Outside workstations
most often are IBM RISC-6000s that can share data with SUN
workstations. PCs are primarily used by smaller cities for running
subsets of the regional model focused on their city.
PROCESS
The figure on the next page illustrates SANDAG's transportation
modeling process. The process is broken down into four major steps of
trip generation, trip distribution, mode choice, and assignment. This
four-step process is widely used throughout the country. Additional
functions provide inputs to the transportation models.
As indicated by the figure, the models can be applied in two stages.
First-stage applications make use of simplified trip distribution and
mode choice procedures. Second-stage applications make use of peak
and off-peak period travel times from first-stage highway assignment.
These travel times are used to redistribute trips and determine mode
choice.
Processing would stop after the first stage for most applications.
Federal guidelines for modeling air quality and major investment
impacts require more elaborate procedures. These studies would use
second-stage transportation models in order to better match
transportation demand with transportation supply. While the
federally-mandated process may yield somewhat better results,
increased computer resource requirements prevent its use on a routine
basis.
A brief summary of each function follows. Table 1-1 summarizes
computer processing time and disk space requirements for each step of
the modeling process where appropriate. Computer times assume that
programs are processed on a SPARC-20 workstation. Programs require
about 50% more time on SPARC-10 workstations.
Zone System
Zones are geographic subareas that provide a method of geographically
summarizing land use, demographic and travel data. SANDAG has a 4,545
transportation zone system that is the basis for most transportation
modeling. A more detailed set of 25,929 Master Geographic Reference
Areas (MGRAs) underlies transportation zones. The MGRAs are used in
transit access procedures and special applications.
(Insert Figure 1-1 - Transportation Modeling Process)
Table 1-1
COMPUTER RESOURCE REQUIREMENTS
Function Computer Time Disk Space
(Minutes) (Kilobytes)
First Stage
Trip Generation 10 6,900
Build Highway Network 45 15,600
Highway Path-Building 55 84,000
Trip Distribution 95 83,000
Vehicle Trip Factoring 40 44,000
Highway Assignment 190 4,900
Post-Assignment Processing 10 7,000
Sub-Total 445 245,400
(7 1/2 Hours)
Second Stage
Highway Path-Building 110 186,200
Trip Distribution 280 124,000
Build Transit Network 80 7,600
Transit Path-Building 60 232,900
Mode Choice 210 68,700
Highway Assignment 390 4,900
Transit Assignment 15 4,300
Post-Assignment Processing 20 7,000
Vehicle Emissions 20 5,100
Sub-Total 1185 640,700
(20 Hours)
TOTAL 1630 886,100
(27 Hours)
Surveys
Surveys are conducted periodically to calibrate relationships used
within the transportation models, such as the number of trips per
dwelling unit. These relationships are assumed to remain constant
between surveys and over the forecast period. Other survey data and
counts are obtained to validate model results.
Growth Forecasts
Every three to five years, SANDAG produces a new "series" of
population, dwelling unit, employment, and land use forecasts for the
Region as a whole and for various geographic levels within the Region.
The most recent Interim Series 8 Growth Forecasts were approved for
use in February 1994. These forecasts cover a 1990 to 2015 time span.
Trip Generation
Household person trip generation rates by structure type are applied
to MGRA level occupied dwelling unit forecasts to calculate
residential trip ends by MGRA. Non-residential trip ends are
estimated by applying trip rates to MGRA level forecasts of non-
residential land use by 80 land use categories. MGRA trip ends are
aggregated to zones for use in transportation models. The trip
generation model estimates daily person trips by ten trip types:
home-work, home-college, home-school, home-shop, home-other, work-
other, other-other, serve passenger, visitor, and airport. Time of
day factors by trip type and land use category split daily trip ends
into peak period and off-peak period trip ends for second stage model
applications.
Transportation Networks
A master transportation database is maintained using Arc/Info
software. The database includes existing and proposed transit and
highway facilities. Attributes are coded for individual segments that
describe geometric, operational, and phasing characteristics.
Highway Networks (First Stage)
Highway facilities for an alternative are selected from the master
transportation database. Capacity and travel times are computed for
each link and a Tranplan highway network is created. Off-peak zone-
to-zone travel times and distances are estimated from the coded
highway network.
Trip Distribution (First Stage)
A gravity model form of the trip distribution model links trip
productions estimated by the trip generation model with trip
attractions in other zones to produce trip movements between zones
based upon zone-to-zone travel times from the highway network
function. SANDAG distributes daily person trips for ten trip types
using off-peak highway travel times.
Vehicle Trip Factoring (First Stage)
Person trip tables from the trip distribution model are factored to
obtain vehicle trip tables for highway assignment. Vehicle trip
factors vary by time period, location, distance, and trip type.
Transit service is represented in a general manner to speed processing
and simplify procedures. Time of day and directional factors are
applied to obtain peak and off-peak period vehicle trip tables for
assignment.
External Trips (First Stage)
Vehicle trips with an end outside of the San Diego Region are
estimated by factoring base-year trip tables from roadside surveys
conducted at cordon stations located where major roads cross boundary
of the Region. Factors are based upon the change in person trips by
subarea and trip type within the San Diego Region. Model estimates
are adjusted to control totals for major roads. External vehicle trip
tables are added to internal vehicle trip tables prior to highway
assignment.
Highway Assignment (First Stage)
Vehicle trips between zones are loaded onto specific links based upon
travel times via alternative routes and limitations imposed by roadway
carrying capacity. Four iterations of Tranplan's equilibrium highway
assignment model are performed for peak and off-peak time periods.
Assignments from the two time periods are merged to obtain daily
traffic forecasts.
Post-Assignment Processing (First Stage)
SANDAG procedures tailor Tranplan highway assignment outputs to San
Diego's specific needs. Model-estimated volumes are adjusted to
compensate for calibration error. Link speeds and times for peak and
off-peak period conditions are computed using Highway Capacity Manual
procedures. Plots, reports, and datasets are produced.
Highway Networks (Second Stage)
Tranplan networks are created with peak and off-peak period congested
link times from the post-assignment process. An additional peak
period high-occupancy vehicle (HOV) network is output reflecting the
higher operating speeds on HOV-only facilities. Zone-to-zone peak and
off-peak period mixed-flow travel times, and peak period HOV travel
times are generated.
Trip Distribution (Second Stage)
Peak and off-peak period person trip ends for the ten trip types are
distributed separately. Mixed-flow congested travel times for each
time period are produced from the first-stage post-assignment process.
Transit Networks (Second Stage)
Tranplan transit networks are generated from the master transportation
database and post-assignment highway travel times. Nodes are located
at transit access points (TAPs). Links connect transit access points
and represent the roadway or rail line over which transit vehicles
operate. Lines are coded over links and describe the frequency of
service, type of service, and path followed by transit routes. Peak
and off-peak period TAP-to-TAP transit paths and travel times are
generated from transit networks.
Transit access files are also created that specify walk and auto
connections between zones and TAPs. Walk access is based upon MGRA
level trip ends and Arc/Info transit coding. Auto access is based
upon mixed-flow peak period zone-to-zone travel times and park-and-
ride lot locations.
Mode Choice (Second Stage)
Person trips between zones are split into six forms of transportation
called modes: drive alone, 2 person autos, 3 or more person autos,
transit-walk, transit-auto, and other. The model determines mode
shares based upon the level of service provided by each mode and trip
maker characteristics. Mode use percentages are calculated separately
by time period, income level, and trip type. Highway vehicle
directional factors are applied to produce peak and off-peak vehicle
trip tables for highway assignment. Transit peak and off-peak period
trips by walk and auto access are produced for transit assignment.
Highway Assignment (Second Stage)
Second stage highway assignment procedures are the same as first stage
procedures. Vehicle trip tables from the mode choice model are input
instead of trip tables from the vehicle trip factoring process.
Post-Assignment Processing (Second Stage)
Second stage post-assignment procedures are the same as first stage
procedures.
Vehicle Emissions
Vehicle emissions are computed based upon zone level vehicle trip ends
from the mode choice model, highway link volumes and speeds from the
post-assignment process, and transit link bus volumes and speeds from
the transit network. Other data from outside of the transportation
models such as vehicle emission factors are also input. Regionwide
emission summaries by pollutant are produced. A file of emissions by
hour of the day, type of pollutant, and air quality modeling gridcell
is optionally produced for use by the Air Pollution Control District's
dispersion model.
NEW DEVELOPMENTS
Developing a new set of growth forecasts provides an opportunity to
implement enhanced modeling procedures. Transportation models
underwent a major over-haul during Series 8 development. The major
objective of these revisions was to merge SANDAG's regional and sub-
area transportation models into one system. Three factors led to this
decision.
Previously, SANDAG had a regional transportation model that was used
to produce forecasts for SANDAG's Regional Transportation Plan,
transit patronage studies, and air quality assessments. Much of the
Region was also covered by one or more special-purpose sub-area
highway models that had been developed at the request of member
agencies. Maintaining multiple databases and sorting out conflicting
forecasts grew to be a major problem as the number of sub-area models
increased.
Computer processing power has grown enormously in the last few years.
The two transportation SUN workstations have about 30 times the
capacity of SANDAG's old mainframe computer that was used for Series 7
transportation modeling. Enhanced computers coupled with Arc/Info's
ability to manage large amounts of data enable SANDAG to model the
entire region at a level of detail previously only possible for small
areas.
Finally, a growth management initiative was recently passed that
requires that transportation impacts of growth be considered more
rigorously for all circulation element roads. Regional models were
too general for this type of analysis and sub-area models covered only
part of the Region.
Compromises between procedures used under the two approaches were
necessary to establish one system applicable for all uses.
Differences between Series 7 and Series 8 are highlighted below. In
general, Series 8 procedures are more straightforward than Series 7
regional procedures although somewhat more complicated than Series 7
sub-area procedures.
Zone Structure
Series 7 regional zones were discarded. Series 8 zones are largely
built around sub-area zones that had previously been developed for
Series 7 subarea studies. There was some adjustment of sub-area zone
boundaries to match nearby Census block boundaries and accommodate
transit access considerations. New detailed zones were developed in
areas not previously covered by sub-area models. The definition of
the Region was expanded to encompass the entire County, not just the
western 40% of the County used in Series 7. The number of zones
increased from 773 zones in Series 7 to 4,545 in Series 8.
Highway Networks
Series 7 regional highway networks included freeways, prime arterials,
major arterials, and regional zone connectors in the western part of
the County. Sub-area networks contained all circulation element roads
within a study area, regional roads outside a study area, and sub-area
zone connectors.
Series 8 highway networks contain all circulation element roads
throughout the Region, as shown in each jurisdiction's adopted General
Plan. Freeways are now represented by one-way links for each freeway
direction in contrast to previous coding which used single, two-way
links. Schematic ramp representations have been replaced with ramp
coding that follows actual alignments. Direction codes have been
dropped in favor of turn prohibitors. The number of Tranplan links
has increased from 17,000 to 45,000 in a typical network.
Transit Networks
Sub-area models dealt only with vehicle trips so sub-area transit
networks were not coded. Series 8 transit networks are similar to
Series 7 transit networks. Procedures have been re-worked to make use
of the new zone system and other databases.
Trip Generation
The Series 7 regional model estimated person trip ends by five trip
types based upon dwelling units by income level by zone and employees
by five Standard Industrial Classification groups by zone. Dwelling
units and employment were phased in five-year increments between 1985
and 2010.
Sub-area models estimated vehicle trip ends by five trip types within
a study area based upon dwelling units by structure type and acres of
non-residential land use by 80 land use categories. Series 7 regional
trip ends were used outside of individual study areas. Full
development of general plans was assumed inside most study areas.
Series 8 combines both procedures. Person trip ends are estimated
based upon dwelling units by structure type and acres of non-
residential land use by 80 land use categories. Dwelling units and
acres are phased by five-year increments between 1990 and 2015. The
number of trip types has been expanded to ten. Trip ends by peak and
off-peak time periods have been added. Trip rates are generally
reduced from previous sub-area model levels.
Trip Distribution
The Series 7 regional model distributed person trips using weighted
peak and off-peak period highway times. Sub-area models distributed
vehicle trips based upon off-peak highway times.
Series 8 distributes daily person trips based upon off-peak period
highway times for most applications. A second stage gravity model is
available which distributes peak period person trips using peak period
highway times and off-peak trips using off-peak period highway times.
Friction factors have been re-calibrated for the new zone system.
Mode Choice
Series 7 sub-area models avoided the need for a mode choice model by
modeling only vehicle trips. Series 8 has a simplified person trip to
vehicle trip factoring process that is available for highway-oriented
applications. This procedure is not based upon transit network times,
so processing time is much less than incorporating the full mode
choice procedures.
For more detailed transit analysis, Series 8 uses a mode choice model
that is similar in structure to the Series 7 regional model. Model
parameters have been re-calibrated, and some procedures have been
modified.
Highway Assignment
Highway assignment procedures were the same between Series 7 regional
and sub-area models. Series 8 makes use of Tranplan's equilibrium
assignment instead of the capacity restraint procedures used
previously. Minor modifications have been made to input parameters.
Transit Assignment
Series 7 and Series 8 transit assignment procedures are similar.
MODEL RESULTS
Table 1-2 presents historical demographic and travel measures for the
years 1980, 1990, and 1993. The 1980-1990 time period was an era of
high growth, while the years 1991-1993 represent a time of poor
economic conditions. These historical measures provide a context for
evaluating Series 8 forecasts summarized in Table 1-3 for the Series 8
base year of 1990, an intermediate year of 2000, and the Series 8
horizon year of 2015.
As indicated in Table 1-2, the 1980's saw strong employment growth in
San Diego, averaging 4.7% per year, fueled by high defense-related
expenditures. San Diego's economy has been in a prolonged recession
since 1990, causing a regional employment loss through 1993. Table 1-
3 shows moderate employment growth through the forecast period after
bottoming out in 1994. These economic trends translate into fewer
employees per household in 2000 and 2015 than in 1990.
Relatively high population growth is expected to continue throughout
the forecast period, although the 3.4% annual average population
growth rate experienced during the 1980's slows considerably to 1.9%.
The slower growth rate is largely due to a drop in the rate of in-
migration linked to fewer job opportunities.
Average household size increased slightly from 1980 to 1990, but is
expected to decrease by a small amount over the forecast period. This
leads to a dwelling unit growth rate that is somewhat higher than
population growth rate.
Travel changes tend to mirror economic conditions. The annual average
growth in vehicle miles of travel (VMT) drops from 7.0% during the
1980s to 1.4% during the 1990's. VMT grows more rapidly after 2000,
averaging 2.3% per year. VMT increased dramatically during the
1980's, growing at twice the rate of population growth. However, VMT
has remained virtually unchanged since 1990, while population has
increased about 2% per year. The high VMT per capita growth rate in
the 1980s was due to unusually strong economic growth and population
growth in high travel age categories.
Historical trends show transit ridership increasing at the same rate
as VMT during the 1980's. Since 1990, transit ridership has dropped,
while overall travel has remained constant. Transit ridership is
expected to grow by 3.1% per year during the 1990's and 4.5% after
2000. Ridership forecasts are produced by SANDAG's mode choice model
and reflect a significant expansion of San Diego's light rail system
and the effects of a proposed Travel Demand Management Ordinance.
Table 1-2
HISTORICAL DEMOGRAPHIC TRAVEL INDICATORS
Indicator 1980 1990 1980-1990 1993 1990-1993
Population 1,873,000 2,520,000 +3.4% 2,668,000 +2.0%
Dwelling Units 674,000 892,000 +3.2% 923,000 +1.1%
Employment 764,000 1,121,000 +4.7% 1,081,000 -1.2%
Vehicle Miles 36,636,000 62,043,000 +6.9% 62,456,000 +0.2%
Miles per Capita 19.6 24.6 +2.6% 23.4 -1.6%
Transit Trips 85,000 146,000 +7.1% 143,000 -0.7%
% = Annual average percentage change
Table 1-3
FORECASTS OF DEMOGRAPHIC TRAVEL INDICATORS
Indicator 1990 2000 1990-2000 2015 2000-2015
Population 2,520,000 3,002,000 +1.9% 3,816,000 +1.8%
Dwelling Units 892,000 1,030,000 +1.6% 1,385,000 +2.3%
Employment 1,121,000 1,132,000 +0.1% 1,472,000 +2.0%
Vehicle Miles 62,043,000 70,778,000 +1.4% 95,589,000 +2.3%
Miles per Capita 24.6 23.6 -0.4% 25.0 +0.3%
Vehicle Trips 8,415,000 9,601,000 +1.4% 12,502,000 +2.0%
Trips per Capita 3.3 3.2 -0.3% 3.3 +0.2%
Average Trip Miles 7.4 7.4 +0.0% 7.6 +0.2%
Vehicle Hours 1,927,000 2,198,000 +1.4% 3,045,000 +2.6%
Average Trip Minutes 13.7 13.7 +0.0% 14.6 +0.4%
Transit Trips 146,000 191,000 +3.1% 320,000 +4.5%
% = Annual average percentage change
CHAPTER 2
ZONE SYSTEM
ZONE SYSTEM
The entire 4,200 square mile County of San Diego is defined as the San
Diego Region and is covered by SANDAG's modeling system. SANDAG
breaks down the Region into various sub-areas, depending upon the type
of application. Transportation models primarily make use of
transportation zones, although SANDAG's other levels of geography also
come into play.
SANDAG's smallest unit of geography is called a master geographic
reference area (MGRA), of which there are 25,929. The concept behind
MGRAs is to produce forecasts for very small areas that can later be
aggregated into higher levels of geography as needed.
MGRAs are based upon blocks as defined by the Census Bureau for use in
the 1990 Census. Blocks are polygons bounded by streets that existed
in 1989 and other Census Bureau features such as Census Tract
boundaries. There are 20,317 blocks averaging 600 feet on a side in
size, but ranging from 100 feet to over 10 miles on a side. The main
problem with using blocks directly for planning purposes is the size
of blocks in rural areas slated for development. Since blocks only
reflect on-the-ground features, they are not well suited for analyzing
proposed development sites. Another problem with blocks is that they
meander along ridge lines and other areas that lack regular street
patterns.
In order to overcome these problems, blocks were split by the
following features to create MGRAs:
þ Community plan boundaries
þ Zip code boundaries
þ City sphere of influence boundaries
þ Planned roadways
þ 1/2 mile buffer around transit routes
þ Transportation zone boundaries
Information is most often requested by zip code, community plan area,
city, and city spheres of influence. These boundaries were added
where they differed from census block boundaries.
Planned freeways and local circulation element roads were added to
further subdivide blocks. These facilities provide convenient
breakpoints and were brought into the MGRA boundary file from
transportation network files.
Specifying the amount of activity within walking distance of transit
is important when estimating transit patronage. SANDAG assumes 1/2
mile is the maximum distance people will walk to transit. Blocks were
manually split as necessary to delineate walk areas around existing
and proposed transit routes and rail stations. Ridge lines and other
topographic features that prevent walk access to transit were also
added.
MGRAs were further subdivided for highway modeling purposes. This
part of the process was done in conjunction with developing a
transportation zone system, which is the next highest level of
geography. One of the objectives of upgrading transportation models
was to allow analysis of traffic volumes for smaller streets and local
transit routes. This dictated having a large number of zones. Much
of the Region had previously been modeled at a detailed level in
various sub-area traffic studies. Zone boundaries from these studies
were added to the MGRA file where there were no nearby features
already in the file. MGRAs in the rest of the Region not covered by
past studies were split as necessary to separate dissimilar land uses
and specify access opportunities.
Figure 2-1 shows how blocks were split to form MGRA's in the Chula
Vista area. The western part of the mapped area is a developed, older
area and has few block splits. The eastern part of the mapped area
was only partially developed in 1990 and, therefore, has more block
splits to accommodate future development.
Once all additional features had been added to the block file,
creating transportation zones was simply a matter of aggregating MGRAs
to zones. A total of 4,545 transportation zones were created. Of
these, 4,536 zones are internal to the Region and nine are external
zones located where major roads cross the County line. Figure 2-2
illustrates how MGRAs are aggregated to form zones in the Chula Vista
area. Figure 2-3 depicts boundaries of the 4,545 zones. External
zones are shown in more detail in Figure 2-4.
Transportation models also make use of three higher levels of
geography which are shown in maps at the end of the chapter.
Transportation zones are nested into 45 sub-regional areas (SRAs), 8
major statistical areas (MSAs), and 3 average vehicle ridership (AVR)
zones. It should be noted that additional SRAs and MSAs have been
added for transportation modeling purposes that are not in SANDAG's
standard definitions.
All geographic files are maintained as Arc/Info coverages. This
enables other aggregations of forecasted data to be easily produced as
requested. Cross-reference files between different levels of
geography can be readily produced. Arc/Info also has sophisticated
mapping capabilities which are useful for error checking and display
purposes.
(Insert Figure 2-1 - Block Split Example
and
Insert Figure 2-2 - MGRA and Zone Relationship)
(Insert Figure 2-3 - Series 8 Zones)
(Insert Map/Figure 2-4 - External Zones)
(Insert Map/Figure 2-5 - Subregional Areas)
(Insert Map/Figure 2-6 - Major Statistical Areas)
(Insert Map/Figure 2-7 - Average Vehicle Ridership Zones)
DATA FILE
DOCUMENTATION
ZONES
The transportation zone coverage is called "zones" and is located
under the /max7/data/covs directory. The following is a list of
polygon attributes coded in the zones coverage.
AREA Arc/Info computed polygon area in square feet.
PERIMETER Arc/Info computed polygon perimeter in feet.
ZONES# Arc/Info assigned unique, sequential ID number for
polygon.
ZONES-ID User assigned unique, fixed ID number for polygon.
Zone number used in transportation modeling.
ZONE Same as ZONES-ID.
COLOR Number indicating color that would be used to shade
zone when producing plot.
TEMP Temporary variable.
SUB-REGIONAL AREAS
The sub-regional area (SRA) coverage used in transportation modeling
is called "pseudosra" and is located under the /max7/data/covs
directory. This coverage has additional SRAs delineating Centre City,
Mission Valley, and Otay Mesa needed for transportation modeling that
are not in SANDAG's standard SRA coverage. External zones have also
been added. The following is a list of polygon attributes coded in
the SRA coverage.
AREA Arc/Info computed polygon area in square feet.
PERIMETER Arc/Info computed polygon perimeter in feet.
PSEUDOSRA# Arc/Info assigned unique, sequential ID number for
polygon.
PSEUDOSRA-ID User assigned unique, fixed ID number for polygon.
SRA SANDAG's two-digit SRA number.
MSA SANDAG's standard Major Statistical Area in which SRA
is located.
NAME Name of SRA.
SEQSRA Sequential SRA number.
MAJOR STATISTICAL AREAS
The major statistical area (MSA) coverage used in transportation
modeling is called "pseudomsa" and is located under the
/max7/data/covs directory. This coverage has an additional MSA
delineating Centre City San Diego that is not in SANDAG's standard MSA
coverage. The following is a list of polygon attributes coded in the
MSA coverage.
AREA Arc/Info computed polygon area in square feet.
PERIMETER Arc/Info computed polygon perimeter in feet.
PSEUDOMSA# Arc/Info assigned unique, sequential ID number for
polygon.
PSEUDOMSA-ID User assigned unique, fixed ID number for polygon.
MSA SANDAG's one-digit MSA number, where:
1 = North City
2 = South Suburban
3 = East Suburban
4 = North County East
5 = North County West
6 = East County
8 = Centre City
9 = Central Area
NAME Name of MSA.
AVERAGE VEHICLE RIDERSHIP ZONES
The average vehicle ridership (AVR) zone coverage used in
transportation modeling is called "avrzone" and is located under the
/max7/data/covs directory. The following is a list of polygon
attributes coded in the AVR zone coverage.
AREA Arc/Info computed polygon area in square feet.
PERIMETER Arc/Info computed polygon perimeter in feet.
AVRZONE# Arc/Info assigned unique, sequential ID number for
polygon.
AVRZONE-ID User assigned unique, fixed ID number for polygon.
AVRZONE Average Vehicle Ridership zone number, where:
1 = Centre City
2 = Suburban
3 = Rural, unincorporated
CHAPTER 3
SURVEYS
SURVEYS
Surveys are needed at many points in the modeling process to establish
relationships between input variables and model-estimated variables.
At other points, model estimates need to be verified against
independent data. Data collection is costly and time consuming so
surveys are conducted relatively infrequently. This normally does not
create a problem since underlying model relationships are relatively
stable over time.
Four surveys listed below provide most of the data used to calibrate
transportation models.
þ 1986 Travel Behavior Survey
þ 1990 San Diego Regional Transit Survey
þ 1986 and 1991 External Trip Surveys
þ 1991 San Diego Visitor Survey
The 1990 Census Transportation Planning Package (CTPP) will provide
additional data for work trips, once it has been tabulated and
checked. A 500 household survey of San Diego residents was also
conducted as part of a 1990 California State-wide Survey. This survey
was not used for model calibration because it lacked some of the data
needed for SANDAG's models and produced trip totals that differed
significantly from local surveys.
Major sources of validation data are traffic counts from Caltrans and
local jurisdictions and transit passenger counts from SANDAG's Transit
Passenger Counting Program.
TRAVEL BEHAVIOR SURVEY
SANDAG's 1986 Travel Behavior Survey (TBS) is the primary source of
model calibration data. Survey procedures and results are documented
in 1986 Travel Behavior Surveys, Volumes 1 and 2 (SANDAG, 1987).
In this survey, 2,754 San Diego households were interviewed to obtain
household, household member, and household trip characteristics on a
survey day. Surveying was conducted on weekdays between February and
June 1986. A two-stage interviewing process was used. First,
households were telephoned using a directory-based digit dialing
method. The survey was explained and general information was
collected. Households agreeing to participate in the survey were
mailed a travel diary in which household members recorded all trips
made on a pre-determined survey day. Surveyors subsequently called
back survey households to record travel diary entries. Of the 8,000
households contacted, 34% agreed to participate in the survey and
completed travel diaries.
Once survey data had been collected, coders manually assigned Series 7
zone numbers at the start and end of each trip based upon addresses,
cross-street names, or place names from survey respondents. Survey
responses were entered into a hierarchical data file with a record for
each household containing household characteristics, followed by a
record for the first person in the household containing person
characteristics, followed by a record for each trip made by the person
on the survey day containing trip characteristics. Alternating person
and trip records follow for each person in the household.
Several modifications were made to the TBS dataset before use in
calibrating Series 8 models. Model-estimated travel times and
distances between start and end zones were added to survey records.
Another modification was to link trip records having a start or end
purpose of "change mode" with other trips. For example, the survey
would record a trip from home to work by driving to a park-and-ride
lot and then getting on a bus as two trips: a home-to-change mode
trip by auto and a change mode-to work trip by bus. The two trips
would be combined into one home-to-work trip by bus through the
linking process.
Another modification was to compute new survey expansion factors to
represent 1990 conditions. This was done by dividing 1990 Census
occupied households by Major Statistical Area, structure type, and
automobile availability by survey households in each category. Table
3-1 summarizes expansion factor calculations. Expansion factors by
structure type, auto availability, and MSA were averaged to compute
expansion factors in cells that had fewer than ten observations.
Finally, Series 7 zone numbers were replaced with Series 8 zone
numbers. The fact that Series 7 zones are often split into several
Series 8 zones complicated the task. Furthermore, Series 8 zones do
not nest into Series 7 zones. First, the fraction of trip ends of
each Series 7 zone in Series 8 zones was computed. TBS trip records
were read and multiple output records were created for each Series 7
and 8 zone combination at the start and end of the trip. Expansion
factors were modified to reflect the fraction of trip ends that the
new record represents.
The Travel Behavior Survey was tabulated to develop the following
calibration data:
þ Trip generation rates for the trip generation model
þ Time-of-day factors for the trip generation model
þ Trip length frequency distributions for the trip distribution
model
þ Vehicle trip factors for the person trip to vehicle trip factoring
process
þ Non-transit mode use percentages for the mode choice model
þ Time of day factors for the emissions model
REGIONAL TRANSIT SURVEY
Every five years SANDAG, in cooperation with transit operators,
conducts an on-board transit survey to obtain transit trip and transit
user characteristics. The most recent survey, conducted in fall of
1990, provides data used to calibrate the transit portion of the mode
choice model. Survey procedures and results are presented in 1990 San
Diego Regional Transit Survey, Volumes 1 and 2 (SANDAG, October 1991).
Table 3-1
TRAVEL BEHAVIOR SURVEY EXPANSION FACTORS
Click HERE for graphic.
Surveyors stationed on-board buses and trolleys distributed
questionnaires to passengers over 12 years of age as they boarded the
vehicle. Passengers filled out forms while they completed their trip
and dropped off forms as they got off vehicles. A total of 57,000
survey forms were distributed over a three month time span from
September to November 1990. This represents a sampling rate of about
27 percent of all passengers. About 41,000 surveys were returned with
useable information.
Survey responses were entered into a dataset. Coders added bus stop
numbers based upon boarding and alighting bus stop locations.
SANDAG's bus stop inventory was used to add state plane coordinates
based upon bus stop numbers. An automated geo-coding process was used
to assign state plane coordinates to the starting and ending address
of each trip as provided by survey respondents. Zone numbers were
attached by overlaying survey coordinates on zones using Arc/Info
procedures. Transit travel times were appended to trip records based
upon transit network estimates.
The survey sample was selected to provide a representative sample by
route configuration, geographic area, and time of day. An expansion
factor was calculated for each survey record to match Passenger
Counting Program boardings by routes, time period and 50 geographic
areas.
Transit survey records were tabulated to determine the following
calibration and validation data:
þ Transit trip shares by income level, trip type, and trip length
for mode choice calibration
þ MSA-to-MSA transit trips for mode choice calibration
þ Park-and-ride locations for coding transit network park-and-ride
nodes
þ Walk access distance distribution to set maximum walk access
distances
þ External transit trip table for external trip modeling
þ Relationship of total boardings to linked trips for transit
assignment validation
þ Access mode percentages for transit assignment validation
þ Zone-to-route trips for transit network validation
þ Zone-to-zone trip tables for transit network calibration
EXTERNAL TRIP SURVEYS
Roadside interview surveys are conducted periodically to determine the
travel characteristics of trips coming into or passing through the San
Diego Region from outside San Diego. Roadside interviews were
conducted in 1986 at five locations to obtain external trip data
consistent with 1986 Travel Behavior Survey data. Four new external
zones were added for Series 8 forecasts when the Region was expanded
to cover the entire County. These new stations were surveyed in 1991.
In addition, two zones on the Mexican border were re-surveyed in 1991
to account for new conditions. Figure 2-4 summarizes the location and
survey year of external zones.
Roadside surveys were conducted by stopping some vehicles as they were
leaving the Region. Surveyors noted vehicle characteristics and
questioned drivers about trip characteristics. Survey responses were
coded and expanded to match daily traffic counts by vehicle type for
each station.
External trip surveys are used to obtain base-year external vehicle
trip tables by trip type. These external trip tables are factored to
represent future year trips and added to internal trip tables from the
rest of the modeling process.
1991 VISITOR SURVEY
San Diego is a major convention and vacation destination. While
travel by business and vacation visitors to San Diego is significant,
other surveys pick up only partial visitor data. Travel behavior
surveys collect information about visitors staying with San Diego
residents. Transit surveys collect information about visitor transit
trips. External surveys collect information about the first San Diego
stop of visitors who drive to San Diego. A small-scale visitor survey
was conducted during the months of July, August and September 1991 to
obtain a more complete picture of visitor travel patterns.
The approach used was to tack on a travel survey to a periodic visitor
survey conducted for the San Diego Convention and Visitors Bureau
(CONVIS). Surveyors stationed outside selected hotels and tourist
attractions questioned passers-by about their trips made on the
previous day. Visitors who were on the first day of their stay in San
Diego were questioned about that day's trips. Approximately 1,400
visitors were surveyed, resulting in 3,200 useable survey trip
records. Survey responses were coded and expanded to the estimated
average daily visitors supplied by CONVIS.
The Visitor Survey was used to obtain visitor trip generation rates
and visitor trip length frequency distributions for gravity model
calibration.
TRAFFIC COUNTS
Traffic counts, used in model validation, are obtained in a variety of
ways, depending upon who has jurisdiction over a facility. Caltrans
counts traffic at relatively few freeway locations each year, but
produces annual count estimates for every freeway segment. There was
some question as to whether model validation should be based upon
actual freeway counts or upon the complete set of estimated counts.
After trying both approaches, it was decided that the more extensive
coverage of estimated counts provided a better basis for validation.
The City and County of San Diego conduct comprehensive traffic count
programs, together collecting approximately 3,500 counts each year.
SANDAG converts traffic count stations into an Arc/Info point coverage
and subsequently matches counts to network links.
Most of the other cities in the region also have good count programs.
For the last 15 years SANDAG, in cooperation with local jurisdictions,
has produced an annual Traffic Flow Map. Each year, cities provide
SANDAG with an updated listing of traffic counts on Traffic Flow Map
links within their jurisdiction. Counts from this program were used
for calibration purposes on surface streets in incorporated areas
outside the City of San Diego. A shortcoming with using Traffic Flow
Map counts is that one Traffic Flow Map link may cover several
Tranplan links. It is not apparent which Tranplan link to assign the
count to. Since traffic volumes usually vary by only small amounts
over a Traffic Flow Map link, this is not believed to be a serious
problem.
TRANSIT PASSENGER COUNTS
SANDAG has operated a Passenger Counting Program since 1979 in
response to Federal requirements and a need for passenger data at the
local level. The program is documented in Transit Passenger Counting
Program Technical Documentation, SANDAG, March 1994. Under the
program, every bus route is counted once a year. Routes are scheduled
for counting using a random selection process. A bus route is a
collection of individual bus trips. All regularly scheduled bus trips
on a route are counted. Light rail trips are counted using a
different sampling technique, where a random sample of five trips
every three days is selected for counting.
Trips are counted by stationing surveyors on-board transit vehicles.
Surveyors record the number of passengers boarding and alighting at
each transit stop. The number of passengers on-board vehicles between
stops is computed from the boarding and alighting data. Surveyors
also record arrival and departure times at selected time points along
a route. An up-to-date transit route and stop inventory is maintained
as part of the Passenger Counting Program.
Bus stop inventories from the Passenger Counting Program are used to
determine bus stop locations in transit network coding. The following
results of the program are used to validate transit assignment
estimates:
þ Ons and offs at stops
þ Screenline counts
þ Boardings by route and mode
DATA FILE
DOCUMENTATION
TRAVEL BEHAVIOR SURVEY
Travel Behavior Survey responses are contained in a file called
"survey" under "/max7/data/tbs." Other files that are derived from
the original survey responses are too numerous to document here. The
survey file is an ASCII file with one record for each surveyed
household, one record for each person, and one record for each trip.
HOUSEHOLD RECORD:
Columns Variable Type Description
1-1 I1 Record Type, where:
1 = Household Record
2 = Person Record
3 = Trip Record
3-9 I7 Survey Number
16-17 I2 Month of Survey
18-19 I2 Day of Survey
21-21 I1 Household Type, where:
1 = Single Family
2 = Multiple Family
3 = Condominium or Townhouse
4 = Mobile Home
5 = Group Quarters
6 = Other
22-22 I1 Tenure, where:
1 = Rent
2 = Own
23-24 I2 Household Size
25-26 I2 Vehicle Available
27-28 I2 Motorcycles Available
29-30 I2 Bicycles Available
31-31 I1 Income Group (1986$s), where:
1 = Less the $5,000
2 = $5,000 to $10,000
3 = $10,000 to $20,000
4 = $20,000 to $40,000
5 = $40,000 to $75,000
6 = More than $75,000
PERSON RECORD:
Columns Variable Type Description
1-1 I1 Record Type, where:
1 = Household Record
2 = Person Record
3 = Trip Record
3-9 I7 Survey Number
11-12 I2 Person Number
16-16 I1 Relationship, where:
1 = Related
2 = Not Related
3 = Visitor
17-17 I1 Gender, where:
1 = Male
2 = Female
19-20 I2 Age
21-21 I1 Drivers License, where:
1 = Licensed
2 = Not Licensed
22-22 I1 Employment Status, where:
1 = Employed Full-Time
2 = Employed Part-Time
3 = Employed with Multiple Jobs
4 = Unemployed
5 = Student
6 = Retired
7 = Not in Labor Market
9 = Student, Employed Part-Time
23-24 I2 Employment Industry
1 = Agriculture/Forestry/Fishing
2 = Mining
3 = Construction
4 = Manufacturing
5 = Transportation/Communications/Utilitie
s
6 = Wholesale Trade
7 = Retail Trade
8 = FIRE
9 = Service
10 = Government
11 = Military
25-26 I2 Employment Industry of Second Job
TRIP RECORD:
Columns Variable Type Description
1-1 I1 Record Type, where:
1 = Household Record
2 = Person Record
3 = Trip Record
3-9 I7 Survey Number
11-12 I2 Person Number
13-14 I2 Trip Number
16-16 I1 Purpose at Destination, where:
1 = Home
2 = Work
3 = Education
4 = Shop
5 = Work Related
6 = Social or Recreation
7 = Change Mode
8 = Serve Passenger
9 = Other
18-20 I3 Series 7 Zone at Destination
21-22 I2 Land Use at Destination, where:
10 = Residential
11 = Hotel/Motel
19 = Other Residential
20 = Regional Shopping Center
21 = Community Shopping Center
25 = Other Retail
26 = Gas Station
27 = Bank
28 = Restaurant/Bar
29 = Other Services
30 = Heavy Industry
31 = Light Industry
32 = High Rise Office
33 = Government Office
34 = Medical Office
39 = Other Office
40 = Nursery/Day Care
41 = Elementary School
42 = Junior High School
43 = Senior High School
44 = Junior College
45 = College/University
49 = Other Educational
Columns Variable Type Description
50 = Hospital, Nursing Home
51 = Church
52 = Cultural Center
53 = Military Base
54 = Transportation Terminal
59 = Other Institutional
60 = Beach/Bay
61 = Park
62 = Tourist Attraction
63 = Outdoor Recreation
64 = Theater/Movie
65 = Indoor Recreation
69 = Open Space
70 = Other
24-27 I4 Time (Hour/Minute) at Origin
28-28 A1 AM/PM at Origin
30-33 I4 Time at Destination
34-34 A1 AM/PM at Destination
36-37 I2 Mode, where:
1 = Automobile
2 = Pick-up/Light Truck
3 = Van
4 = Truck
5 = Motorcycle
6 = Bicycle
7 = Walk
8 = Taxicab/Limousine
9 = Public Transit
10 = School Bus
11 = Railroad
12 = Airplane
13 = Other
38-38 I1 Driver, where:
1 = Yes
2 = No
39-40 I2 Vehicle Occupancy
41-41 I1 Paid Parking at Destination, where:
1 = Yes
2 = No
42-42 I1 Primary Work Trip, where:
1 = Yes
2 = No
REGIONAL TRANSIT SURVEY
Transit survey responses are contained in a file called "survey" under
/max8/data/trs90. Other files that are derived from the original
survey responses are too numerous to document here. The survey file
is an ASCII file with one record for each unlinked transit trip that
was surveyed.
Columns Variable TypeDescription
1-6 I6 Survey Number
7-12 I6 Survey Date
13-15 I3 Route
16-19 I4 Terminal Time
20-20 I1 Direction Number
21-22 I2 Surveyor ID
23-23 I1 Language, where:
0 = English
1 = Spanish
24-24 I1 Purpose at Origin, where:
1 = Home
2 = Work
3 = School
4 = Shopping
5 = Social or Recreation
6 = Other
7 = Medical
25-25 I1 Address Status at Origin, where:
0 = No Information
1 = Good Information
26-83 A58 Address at Origin
84-85 A2 City at Origin
86-89 I4 Bus Stop Number at Origin
90-90 I1 Access Mode at Origin, where:
1 = Transferred from Bus
2 = Transferred from Rail
3 = Walked
4 = Drove Alone,
5 = Carpooled
6 = Dropped Off
7 = Other
8 = Transferred from Dial-a Ride
Columns Variable TypeDescription
91-93 I3 Route Transferred from at Origin
94-94 I1 Direction of Transfer Route
95-96 I2 Blocks Walked to Bus Stop at Origin
97-97 I1 Purpose at Destination
98-98 I1 Address Status at Destination
99-156 A58 Address at Destination
157-158 A2 City at Destination
159-162 I4 Bus Stop Number at Destination
163-163 I1 Egress Mode at Destination
164-166 I3 Route Transferred from at Destination
167-167 I1 Direction of Transfer Route
168-169 I2 Blocks Walked from Bus Stop at Destination
170-171 I2 Fare, where:
1 = Cash
2 = Monthly Pass
3 = Transfer Slip
4 = Ten Pack
5 = Other
172-172 I1 Auto Availability, where:
1 = Auto Available
2 = Auto Not Available
173-173 I1 Transit Use Frequency, where:
1-7 =
Days Used per Week
9 = Less Than Once a Week
174-174 I1 Gender, where:
1 = Male
2 = Female
175-175 I1 Visitor, where:
1 = Visitor
2 = Resident
176-176 I1 Military, where:
1 = Military
2 = Civilian
177-177 I1 Household Size, where:
1-4 =
Number in Household
5 = Five or more
178-178 I1 Ethnicity, where:
1 = Hispanic
2 = White/Non-Hispanic
3 = Black
4 = Asian
5 = Other
6 = American Indian
Columns Variable TypeDescription
179-179 I1 Quality of Transit Service, where:
1 = Good
2 = Average
3 = Poor
180-181 I2 Age
182-182 I1 Income (1990 $s), where:
1 = Less Than $10,000
2 = $10,000 to $20,000
3 = $20,000 to $30,000
4 = $30,000 to $40,000
5 = $40,000 to $50,000
6 = $50,000 to $60000
7 = More Than $60,000
183-183 I1 Knowledge of Route Schedules, where:
1 = Knowledgeable
2 = Not Knowledgeable
184-184 I1 Rail Border Crossing, where:
1 = Trip Crosses U.S. Border
2 = Trip Stays in U.S.
185-185 I1 Prior Mode Use of Rail Passengers, where:
1 = Drove Alone
2 = Carpooled
3 = Rode Bus
4 = Trip Not Made
5 = Other
186-188 I3 Prior Bus Route of Rail Passengers
189-191 I3 First Comment
192-194 I3 Second Comment
195-201 I7 X-Coordinate of Origin Bus Stop
202-208 I7 Y-Coordinate of Origin Bus Stop
209-209 A1 Location Code of Origin Bus Stop
210-211 I2 Expansion Zone at Origin Bus Stop
212-218 I7 X-Coordinate of Destination Bus Stop
219-225 I7 Y-Coordinate of Destination Bus Stop
226-226 A1 Location Code of Destination Bus Stop
227-228 I2 Expansion Zone at Destination Bus Stop
229-235 I7 X-Coordinate of Origin
236-242 I7 Y-Coordinate of Origin
243-244 I2 Expansion Zone at Origin
245-251 I7 X-Coordinate of Destination
252-258 I7 Y-Coordinate of Destination
259-260 I2 Expansion Zone at Destination
261-267 F7.3 Unlinked Trip Expansion Factor
268-269 I2 Number of Links (Transfers)
Columns Variable TypeDescription
270-270 I1 Transit District at Origin
271-271 I1 Transit District at Destination
272-277 I6 Census Tract at Origin
278-281 A4 Census Block at Origin
282-285 I4 Census Place at Origin
286-291 I6 Census Tract at Destination
292-295 A4 Census Block at Destination
296-299 I4 Census Place at Destination
300-305 I6 Census Tract at Origin Bus Stop
306-309 A4 Census Block at Origin Bus Stop
310-313 I4 Census Place at Origin Bus Stop
314-319 I6 Census Tract at Destination Bus Stop
320-323 A4 Census Block at Destination Bus Stop
324-327 I4 Census Place at Destination Bus Stop
1986 EXTERNAL TRAVEL SURVEY
Travel Behavior Survey responses are contained in a file called
"survey" under "/max7/data/ext86." Other files that are derived from
the original survey responses are too numerous to document here. The
survey file is an ASCII file with one record for each surveyed trip.
Columns Variable TypeDescription
1-5 I5 Survey Number
6-10 I5 Survey Station
11-11 I1 Direction of Travel, where:
1 = North
2 = South
3 = East
4 = West
12-13 I2 Beginning Hour of Interview
14-14 I1 Vehicle Occupancy
15-16 I2 Vehicle Type
17-20 A4 Residence, where:
X000 =
External Place Code
0000 =
External Place Code
M000 =
Mexican Zone
000 = Series 7 Zone
21-23 I3 Entry Cordon Station
24-25 I2 Length of Stay (Days)
26-29 I4 Place Code of External Last Stop
30-32 I3 Series 7 Zone of Internal Last Stop
33-34 I2 Purpose of Last Stop
35-36 I2 Purpose of Next Stop
37-40 I4 Location of Next Stop (See 17-20)
41-41 I1 Final Destination, where:
1 =
Yes
2 =
No
42-45 I4 Location of Final Destination (See 17-20)
46-47 I2 Frequency of Trip
48-53 F6.2 Expansion Factor
54-55 I2 Trip Type
1 =
Home-Work
2 =
Home-Shop
3 =
Home-Other
4 =
Work-Other
5 =
Other-Other
1991 EXTERNAL TRAVEL SURVEY
Travel Behavior Survey responses are contained in a file called
"survey" under "/max7/data/ext91." Other files that are derived from
the original survey responses are too numerous to document here. The
survey file is an ASCII file with one record for each surveyed trip.
Columns Variable TypeDescription
1-5 I5 Survey Number
6-8 I2 Survey Station, where:
1 = SR-79
2 = SR-78
3 = I-8
4 = SR-188
5 = Otay
4 = West
9-10 I2 Beginning Hour of Interview
11-11 I1 Vehicle Occupancy
12-12 I2 Vehicle Type
13-13 I1 Residence
14-67 A54 Address at Origin
68-72 I5 Zipcode at Origin
73-74 I2 Purpose at Origin
75-76 I2 Purpose at Destination
77-80 I4 Place Code at Destination
81-84 F4.1 Expansion Factor
CALTRANS TRAFFIC COUNTS
Caltrans counts are contained in a coverage called "ststa" that is
located under the /max7/data/covs directory. The coverage contains
two points for each freeway count location; one for each direction of
travel. Count stations on surface streets are represented by one
point. Additional stations have been placed on future Caltrans
facilities where posted volumes are to be produced. These future
stations have no count data. The following is a list of point
attributes coded in the count station coverage.
AREA Blank.
PERIMETER Blank.
STSTA# Arc/Info assigned unique, sequential ID number for point.
STSTA-ID User assigned unique, fixed ID number for point.
STA Station number at Caltrans count stations. Otherwise
SANDAG assigned station number.
DIR Direction of travel.
VOL Directional 1990 average weekday traffic count.
AMVOL Directional 1990 morning peak hour traffic count.
PMVOL Directional 1990 afternoon peak hour traffic count.
DISTANCE Arc/Info assigned distance in feet to nearest highway
arc.
NM Count station description.
SEQSTA SANDAG assigned sequential station number.
HWYCOV# Arc/Info ID number of nearest highway arc.
CITY OF SAN DIEGO TRAFFIC COUNTS
City of San Diego counts are contained in a coverage called "sdsta"
that is located under the /max7/data/covs directory. The coverage
contains one point for each count station location that is on a street
contained in the master transportation coverage (TCOV). The following
is a list of point attributes coded in the count station coverage.
AREA Blank.
PERIMETER Blank.
SDSTA# Arc/Info assigned unique, sequential ID number for point.
SDSTA-ID User assigned unique, fixed ID number for point.
STA City assigned count station number.
VOL Directional 1990 weekday traffic count.
HWYCOV# Arc/Info ID number of nearest highway arc.
DISTANCE Arc/Info assigned distance in feet to nearest highway
arc.
NM Street name of count station.
LIM Limits covered by station.
SEQSTA SANDAG assigned sequential station number.
COUNTY OF SAN DIEGO TRAFFIC COUNTS
County of San Diego counts are contained in a coverage called "ctsta"
that is located under the /max7/data/covs directory. The coverage
contains one point for each count station location that is on a street
contained in the master transportation coverage (TCOV). The following
is a list of point attributes coded in the count station coverage.
AREA Blank.
PERIMETER Blank.
CTSTA# Arc/Info assigned unique, sequential ID number for point.
CTSTA-ID User assigned unique, fixed ID number for point.
DIR Direction of travel.
YR Year of count.
VOL Directional 1990 weekday traffic count.
WAY One/Two Way
STA County assigned count station number.
NM Street name of count station.
XM Cross street name at count station.
HWYCOV# Arc/Info ID number of nearest highway arc.
DISTANCE Arc/Info assigned distance in feet to nearest highway
arc.
SEQSTA SANDAG assigned sequential station number.
TRAFFIC FLOW MAP COUNTS
Traffic Flow Map counts are contained in two history files called
"hist1" and "hist2." Count files are updated annually. Files are
located in a directory called "/max7/proj/vmt00," where 00 indicates
the count file year. The files are in ASCII format and contain one
record for Traffic Flow Map Link.
The first history file (hist1) covers the years 1978-1989 with the
following format.
Columns Variable TypeDescription
1-7 I7 Link Number
7-14 I7 1977 Weekday Traffic Count
15-16 A2 1977 Count Status, where:
Blank =
Link Counted
"N" =
Link Not Counted
"E" =
Estimated Count
"V" =
Vacated Link
17-23 I7 1978 Count
24-25 A2 1978 Status
26-32 I7 1979 Count
33-34 A2 1979 Status
35-41 I7 1980 Count
42-43 A2 1980 Status
44-50 I7 1981 Count
51-52 A2 1981 Status
53-59 I7 1982 Count
60-61 A2 1982 Status
62-68 I7 1983 Count
69-70 A2 1983 Status
71-77 I7 1984 Count
78-79 A2 1984 Status
80-86 I7 1985 Count
87-88 A2 1985 Status
89-95 I7 1986 Count
96-97 A2 1986 Status
98-104 I7 1987 Count
105-106 A2 1987 Status
107-113 I7 1988 Count
114-115 A2 1988 Status
116-122 I7 1989 Count
123-125 A2 1989 Status
The second history file (hist2) is set-up to cover the years 1990-2002
with the following format.
Columns Variable TypeDescription
1-7 I7 Link Number
7-14 I7 1990 Count
15-16 A2 1990 Status
17-23 I7 1991 Count
24-25 A2 1991 Status
26-32 I7 1992 Count
33-34 A2 1992 Status
35-41 I7 1993 Count
42-43 A2 1993 Status
44-50 I7 1994 Count
51-52 A2 1994 Status
53-59 I7 1995 Count
60-61 A2 1995 Status
62-68 I7 1996 Count
69-70 A2 1996 Status
71-77 I7 1997 Count
78-79 A2 1997 Status
80-86 I7 1998 Count
87-88 A2 1998 Status
89-95 I7 1999 Count
96-97 A2 1999 Status
98-104 I7 2000 Count
105-106 A2 2000 Status
107-113 I7 2001 Count
114-115 A2 2001 Status
116-122 I7 2002 Count
123-125 A2 2002 Status
CHAPTER 4
GROWTH FORECASTS
GROWTH FORECASTS
The Region's growth rate and location of new development within the
Region largely determine future travel patterns. Thus, it is
important to understand the growth forecasting process before
discussing transportation models.
Every three to five years, SANDAG produces a new set of regional
growth forecasts that account for updated existing development,
regional growth trends, and local general plans. Work began on the
current Series 8 forecasts in 1990. SANDAG's Board approved the use
of Interim Series 8 forecasts in February 1994.
A brief overview of the growth forecasting process is provided in this
report. The process is made up of five major steps:
þ Collecting input assumptions
þ Forecasting regionwide growth control totals
þ Allocating employment to 224 subareas called zones for urban
modeling (ZUMs)
þ Allocating dwelling units to ZUMs
þ Allocating growth to 25,929 Master Geographic Reference Areas
(MGRAs)
Annual regional-level forecasts are produced for the years 1990
through 2015. Sub-regional allocation models are run only for
selected years. Standard Series 8 forecast years are 1990, 2000,
2010, and 2015. Sub-regional allocations for other years are produced
as requested.
There is extensive local involvement throughout the growth forecasting
process. SANDAG's Board of Directors represents local jurisdictions
within the region and provides policy direction. At a technical
level, local planners provide input assumptions and work with SANDAG
staff to resolve forecasting issues.
Every effort is made to ensure that regional forecasts agree with
local development plans. MGRA, ZUM and regional holding capacities
are carefully computed based upon adopted general plans. The task of
the growth forecasting models is to phase development between existing
and ultimate conditions.
Most of the data files involved with growth forecasting are maintained
in Arc/Info coverages. Arc/Info simplifies file editing, enables data
from different files to be related, and has the flexibility to produce
almost any kind of map for checking and display purposes. Arc/Info
enables SANDAG to forecast growth at the detailed MGRA level without a
loss of precision.
Forecasts were prepared for two major alternative development
scenarios under Series 8. One scenario termed "Existing Policies" is
based upon adopted local general plans and regional transportation
plans. The other "Quality of Life" alternative modifies general plans
to put more growth in existing urban areas located near transit
facilities.
Additional growth forecasts were developed for evaluating Regional
Transportation Plan (RTP) alternatives. There is assumed to be a
relationship between development and accessibility. RTP growth
forecasts reflect the effects upon land use of alternative assumptions
about transportation facilities that might exist with different
funding levels or with shifts of highway funding to transit.
INPUT ASSUMPTIONS
Developing the databases that feed into the growth models is the most
time consuming part of the process. This effort consists of
collecting data on base-year conditions, local general plans, site-
specific projects, development constraints, and transportation
facilities.
Base-Year Conditions
An accurate description of existing conditions is important because
roughly 3/4 of the urban land that will ultimately be developed in the
Region is already developed. One of the basic assumptions of the
growth forecasting process is that land now developed will retain
existing characteristics unless identified as part of a redevelopment
or in-fill project. Three base-year data files are created:
þ 1990 Regional Land Use Inventory
þ 1990 Census block statistics
þ 1990 Employment Inventory
The Regional Land Use Inventory is an Arc/Info coverage that
delineates 1990 land uses in 80 land use categories described at the
end of the chapter. The inventory is somewhat generalized. The
smallest features identified are usually 2.5 acres. The level of
detail is meant to be commensurate with proposed land use in general
plans.
The 1990 inventory is an update of a 1986 inventory used in Series 7
forecasts. Satellite imagery for 1986 and 1990 was analyzed to detect
areas where land use changes had occurred. Other sources of
information include aerial photography, cross-checks with other SANDAG
files, ground inspection, and local planner review.
Census data is used to obtain residential dwelling unit
characteristics such as structure type, household size, income, and
ethnic composition. In most cases, block data is input directly.
About 10% of blocks are split into smaller areas to form MGRAs. In
these instances, block data is apportioned to each MGRA based upon the
MGRA's proportion of total block area. Most split blocks are in rural
areas with little existing development so this is not a major problem.
The employment inventory is derived from a file that was initially
purchased from Dun and Bradstreet. The file had names, addresses,
employee count, and Standard Industrial Code for every private
employer in the Region. A considerable effort was necessary to make
the file useful for planning purposes. An address matching process
was used to assign coordinates to individual employer records. An
Arc/Info coverage of employer locations was developed that could be
related to other parts of the forecasting process. Address-based
coordinates were cross-checked with land use files and adjusted where
necessary. Other processing included adding in government employers,
factoring in self-employment estimates, and cross-checking with
Commuter Computer's large employer listing.
General Plans
Local general plans are assumed to represent ultimate conditions.
These plans are periodically updated, so that actual development may
differ significantly from what is now depicted in plans. Nonetheless,
general plans provide the only effective method of evaluating land use
policies, implementing policy changes, and forecasting land use for
small areas.
Discrepancies in planned land use can occur in areas that are now
unincorporated but fall into an incorporated city's sphere of
influence. Both plans are digitized so that either plan can be
selected for use. Normally, city plans supersede unincorporated plans
where there is a contradiction.
General plan land use designations differ between jurisdictions.
Proposed land uses are assigned a 1990 land use inventory category
that most closely matches the general plan category. General plans
also have a range of residential densities associated with each
residential land use category. Individual jurisdictions specify where
in the range they expect their area to develop.
Site-Specific Projects
Projects that are in the "pipeline" are important for accurate near-
term growth forecasts. While most projects conform to general plan
land use, more detailed site plans may be available for projects
nearing implementation. Local jurisdictions submit as much
information as possible about these projects in terms of physical
layout, land use make-up, dwelling unit counts, and employment levels.
A site-specific coverage is created that maps each project.
Development Constraints
Environmental constraints are becoming more important in dictating
where growth can occur. SANDAG maintains Arc/Info coverages of steep
slopes, public lands, floodways, and riparian habitat that could
preclude development. Local jurisdictions review maps of these areas
and identify which constraints should be included in the forecasting
process.
Transportation Facilities
Correctly accounting for the relationship between land use and
transportation facilities is important. Transportation conditions are
represented by highway and transit times between ZUMs from the
transportation models. Another input file describes the percent of
work trips by transit and is used to weight transit and auto times.
Assumptions about specific transit and highway facilities vary by
alternative. In all cases, travel times reflect the effects of
congestion as estimated by transportation models. ZUM-to-ZUM travel
times are computed by weighting zone-to-zone travel times by zone-to-
zone trips for zones comprising each ZUM.
Travel times are input in a cyclical manner. Base-year travel times
are input to the first growth forecast increment. Resulting growth
forecasts are input to transportation models, along with corresponding
transportation facility assumptions. Travel times from the
transportation models are then input to the next growth forecast
increment. This process is repeated until the final 2015 growth
forecast increment is complete.
Capacity File
An MGRA level capacity file, described at the end of the chapter,
summarizes data for input to the growth models. Arc/Info is used to
overlay the base-year land use coverage, general plan coverage, site-
specific coverage, and development-constraint coverage. The following
rules are applied to guide creation of the capacity file:
þ Constrained areas and developed areas with matching base-year and
general plan land use designations are assumed to retain their
base-year land use.
þ Vacant, unconstrained areas within site-specific projects are
assumed to develop according to site-specific plans.
þ Vacant, unconstrained areas outside of site-specific projects are
assumed to develop according to general plans.
þ Developed areas with differing base-year and general plan land use
designations are brought to the attention of local planners. Some
areas are put into a redevelopment category.
þ Developed residential areas with densities below allowable general
plan densities are brought to the attention of local planners.
Some areas are put into a residential in-fill category.
A capacity file record is created for each base-year/general plan land
use combination within an MGRA. The resulting capacity file contains
about 58,000 records with 46 fields of data. The following items are
used in transportation modeling:
þ MGRA number
þ Percent of capacity used
þ Area (acres)
þ Base-year land use
þ Employment by 15 SIC groups
þ General Plan Land Use
REGIONAL GROWTH CONTROL TOTALS
SANDAG's Demographic and Economic Forecasting Model (DEFM) produces
year-by-year regional-level forecasts of some 700 demographic and
economic variables out to the year 2015. The model is documented in
DEFM Forecast 1993 to 2015, Volumes 1-5 (SANDAG, 1993). Forecasts are
based upon a history, starting in 1950, of variables at the regional,
state-wide, and national level and some assumptions about future
conditions at the same three levels. It is assumed that regional
growth is independent of local land use assumptions. SANDAG's Board
of Directories reviews regional DEFM forecasts before proceeding with
the rest of the forecast.
San Diego's historical population growth rate has fluctuated
dramatically. High growth years occurred during WWII, the Korean
Conflict, the Vietnam War and the Cold War build-up during the 1980's.
Population growth has averaged 40,000 new residents a year over the
last 20 years. During the mid-1980's, San Diego experienced its
highest recent growth years, with an 80,000 population increase during
the peak year of 1989.
Even though San Diego's economy has been in recession since mid-1989,
population has continued to increase. Population growth has averaged
40,000 over the last three recession years. Population growth is
expected to increase by 51% over the 25 year Series 8 forecast period,
averaging about 50,000 new residents a year.
Employment growth has been even more variable, although it tracks
population growth fairly closely. The San Diego Region added 36,000
new jobs a year during the 1980's. Job growth of only 13,000 jobs a
year is expected over the 1990-2015 forecast period, as poor economic
conditions are expected to persist for some time. There were 2.4
persons per employee in 1980 and 2.2 persons per employee in 1990.
This rate is expected to increase to 2.6 persons per employee in 2015.
Part of the increased unemployment is due to a higher proportion of
the population in both retirement and school age groups, although
somewhat higher unemployment is forecasted for the working age
population of 20-60.
SUB-REGIONAL EMPLOYMENT ALLOCATION
The EMPAL model distributes employment growth from the regional level
to 224 Zones for Urban Modeling (ZUMs) for seven employment categories
and three time intervals of 1990 to 2000, 2000 to 2010, and 2010 to
2015. Employment categories are based upon single-digit Standard
Industrial Classes of:
(1) Manufacturing
(2) Transportation, communication and utilities (TCU)
(3) Wholesale trade
(4) Retail trade
(5) Finance, insurance, and real estate (FIRE)
(6) Services
(7) Government
Inputs to the EMPAL model include:
þ Regional employment growth totals from the DEFM model by category
þ Employees by ZUM and employment category from the previous time
interval
þ Total occupied housing units by ZUM from the previous time
interval
þ Employment holding capacity of each ZUM
þ Peak period highway ZUM-ZUM travel times
þ Site-specific employment
The model allocates employment for each category outside of site-
specific projects to ZUMs as a function of travel times to residential
areas, travel times to other employment areas, and the incremental
employment capacity of the ZUM. SANDAG constrains employment
forecasts to be equal or less than the employment capacity by ZUM and
category computed from general plans. Site-specific employment is
added after the EMPAL allocation is complete.
Table 4-1 summarizes the results of the employment allocation by Major
Statistical Areas (Figure 2-6). As indicated, the highest employment
growth rate occurs in the South Suburban MSA, which has large areas of
vacant, industrial land. The largest absolute employment growth is
expected to occur in the North City area, where the largest share of
the Region's employment is currently located. The Central Area MSA is
largely built-out and thus has the lowest growth rate.
SUB-REGIONAL RESIDENTIAL ALLOCATION
After the EMPAL model has been run to distribute employment growth to
ZUMs for a particular time interval, the Projective Land Use Model
(PLUM) is run. PLUM's primary function is to allocate residential
growth to ZUMs, although other miscellaneous functions are performed.
Inputs to the PLUM model include:
þ Single and multi-family housing stock capacity of each ZUM
þ Peak period highway and transit ZUM-ZUM travel times
þ Fraction of work trips by transit by ZUM
þ Site-specific dwelling units
þ Regional control totals from DEFM
Table 4-1
SERIES 8 EMPLOYMENT BY MAJOR STATISTICAL AREA
Major Statistical Area 1990 2015 Change
Centre City 65,000 90,200 +39%
Central Area 299,700 303,200 + 1%
North City 396,200 493,500 +26%
South Suburban 76,800 131,700 +72%
East Suburban 139,700 181,500 +30%
Northwest County 146,000 196,500 +35%
Northeast County 119,900 180,800 +51%
East County 3,500 5,600 +60%
Region 1,246,800 1,582,100 +27%
PLUM's residential allocation works by first allocating employees from
employment ZUMs to residence ZUMs based upon the travel times between
ZUMS and the dwelling unit capacity in residence ZUMs. Separate
allocations are performed for employees living in single-family
dwelling units and those living in multiple-family units. Employees
are further disaggregated into those that commute by automobile and
those that commute by transit so that highway and transit travel times
can be weighted appropriately.
Once employees have been allocated, housing stock requirements are
estimated by applying employed resident per household rates that vary
by ZUM and vacancy rates that vary by ZUM and structure type. Housing
stock requirements by ZUM and structure type are checked against
housing stock capacity. Dwelling units are reallocated from ZUMs with
too little capacity to the closest ZUM with sufficient capacity to
meet demand. Site-specific dwelling units, base-year "other" dwelling
units, and factored base-year mobile homes are added to single and
multiple-family dwellings from the allocation process to arrive at
total dwelling units for each ZUM.
Total housing stock is converted to occupied units by applying a
vacancy rate that varies by ZUM and structure type. Population
estimates are derived by applying household size factors that vary by
ZUM and forecast year to occupied units and then adding in group
quarters population. Occupied dwelling units are apportioned to seven
income ranges using modified log normal curves that vary by ZUM and
forecast year. PLUM estimates are adjusted to match regional control
totals from DEFM.
Table 4-2 summarizes PLUM results. It should be noted that by 2015
all urban, residential land identified in existing general plans is
used-up so that 2015 dwelling unit forecasts are a function of
available capacity, not the PLUM allocation process.
Table 4-2
SERIES 8 DWELLING UNITS BY MAJOR STATISTICAL AREA
Major Statistical Area 1990 2015 Change
Centre City 5,800 35,300 +509%
Central Area 199,900 260,200 +30%
North City 219,600 315,300 +44%
South Suburban 83,100 160,300 +93%
East Suburban 154,000 225,200 +46%
Northwest County 107,800 170,000 +58%
Northeast County 110,400 199,900 +81%
East County 6,800 18,900 +176%
Region 887,400 1,385,100 +56%
MGRA ALLOCATION
A SANDAG Fortran program (SOAP) distributes dwelling units and
employment produced by PLUM and EMPAL from 224 ZUMs to 25,929 MGRAs.
In addition, SOAP performs land use accounting. SOAP uses "capacity"
files described above that are derived from existing development and
planned land use. PLUM produces "soapbase" files that list ZUM level
residential and employment estimates. Finally, "access weight" files
are input that indicate the order in which MGRAs will be developed.
Access weights are determined through the use of Arc/Info procedures.
Weights reflect the number of dwelling units and employment within 1/2
mile of each MGRA from the previous time interval.
SOAP produces updated capacity files and "soapbase" files, described
at the end of the chapter, that summarize MGRA level dwelling unit,
employment, and land use estimates for a forecast year. These
"capacity" and "soapbase" files are the files used by the
transportation models to generate trips and determine transit access.
DATA FILE
DOCUMENTATION
SERIES 8 LAND USE CODE DEFINITIONS
100 SPACED RURAL RESIDENTIAL - Single-family homes located in rural
areas with lot sizes of approximately 1 to 10 acres. Homes in
areas of lower densities are coded as agricultural or vacant,
not residential. Rural residential estates may have small
orchards, fields or small storage buildings associated with the
residential dwelling unit.
110 SINGLE-FAMILY RESIDENTIAL - Single-family detached housing
units, on lots smaller than 1 acre. This category may also
include smaller mixed uses such as churches, schools, post
offices, libraries, gas stations, small commercial
establishments (7-11, Circle K), etc. Newer developments may
include clubhouses, recreation areas, pools, tennis, etc.,
located within and associated with the residential development.
120 MULTIPLE-FAMILY RESIDENTIAL - Attached housing units, two or
more units per structure - includes duplexes, townhouses,
condominiums, apartments, and SRO's in Centre City. Newer
developments may include clubhouses, recreation areas, pools,
tennis, etc., located within and associated with the residential
development.
130 MOBILE HOME PARKS - Includes mobile home parks with 10 or more
spaces that are primarily for residential use. (RV type parks
are included within the commercial recreation category.)
1401 JAILS/PRISONS
1402 DORMITORIES
1403 MILITARY BARRACKS
1404 MONASTERIES
1409 OTHER GROUP QUARTERS - Convalescent or retirement homes not
associated with or within a health-care facility, rooming
houses, half-way houses, California Conservation Corps, Honor
Camps and other correctional facilities.
1501 HOTEL/MOTEL/RESORTS - Hotels, motels, resorts, and other
transient accommodations. Commonly found along freeways and
prime commercial areas, in downtown areas, and near tourist
attractions. Examples of resorts would be La Costa Health Spa,
Olympic Resort in Carlsbad near the airport, and Lawrence Welk.
2001 HEAVY INDUSTRY - Shipbuilding, airframe, and aircraft
manufacturing. Usually located close to transportation
facilities and commercial areas. Parcels are typically large,
20-50 acres. SANDAG has a list of heavy industrial sites in San
Diego.
2101 INDUSTRIAL PARKS - Office/industrial uses clustered into a
center. The primary uses are industrial but may include high
percentages of other uses in service or retail activities.
SANDAG's 1990 Employment file was used to assess the types of
employment within industrial areas and to classify industrial
areas by type. (See attachment regarding use of 1990 Employment
file to classify industrial by type.)
2103 GENERAL LIGHT INDUSTRY - All other industrial uses and
manufacturing not included in the categories above. These are
not located inside of parks, but are usually along major streets
or clustered in certain areas. Includes manufacturing uses such
as lumber, furniture, paper, rubber, stone, clay, and glass; as
well as light industrial uses as auto repair services and
recycling centers. Mixed commercial and office uses (if not
large enough to be identified separately) are also included.
General industrial areas are comprised of 75 percent or more of
industrial uses (manufacturing, warehousing, and wholesale
trade). SANDAG's 1990 Employment file was used to assess the
types of employment within industrial areas and to classify
industrial areas by type. (See attachment regarding use of 1990
Employment file to classify industrial by type.)
2104 WAREHOUSING/PUBLIC STORAGE - Usually large buildings located
near freeways, industrial or strip commercial areas. Public
self-storage buildings are typically long, rectangular and
closely spaced.
2201 EXTRACTIVE INDUSTRY - Mining, sand and gravel extraction, salt
evaporation.
2301 JUNKYARD/DUMPS/LANDFILLS - Active or in current use. The
landscape should show visible signs of the activity.
4101 COMMERCIAL AIRPORTS - Lindbergh Field only.
4102 MILITARY AIRPORTS - Airports owned and operated by the military.
Found on Military bases.
4103 GENERAL AVIATION AIRPORTS - All general aviation airports.
4104 AIRSTRIPS
4110 OTHER TRANSPORTATION GENERAL
4111 RAIL STATIONS/TRANSIT CENTERS/SEAPORTS - Major transit centers
(i.e., Oceanside Transit Center, El Cajon Transit Center), rail
stations (i.e., Santa Fe Depot, Del Mar Train Station, and major
trolley stations), and seaport terminals (Port of SD). Parking
areas associated with these uses are included. Transit centers
within shopping centers are included within the shopping center
category.
4112 FREEWAYS - Divided roadways with 4 or more lanes, restricted
access, grade separations, and rights-of-way with a width of 200
feet or more. Includes all right-of-way and interchange areas,
but not frontage roads.
4113 COMMUNICATIONS AND UTILITIES - TV and radio broadcasting
stations, relay towers, electrical power generating plants,
water and sewage treatment facilities.
4114 CENTER CITY SURFACE PARKING - All surface parking lots found in
center city. These are identified for use in the Regional
Growth Forecasts. They provide vacant land for future
development.
4115 CENTER CITY STRUCTURE PARKING - All large parking structures
found in center city areas of large cities.
4116 PARK AND RIDE LOTS - Stand alone parking areas that are not
associated with any land use. These are usually located near
freeways.
4119 OTHER TRANSPORTATION - Maintenance yards and their associated
activities, transit yards, railroad yards.
5000 GENERAL COMMERCIAL
5001 WHOLESALE TRADE - Usually located near transportation
facilities. Structures are usually large and cover the majority
of the parcel. Examples are clothing and supply.
5002 REGIONAL SHOPPING CENTERS - Usually contain 1 to 5 major
department stores, have more than 50 tenants, and cover 40 acres
or more.
5003 COMMUNITY SHOPPING CENTERS - Smaller in size than the regional
shopping centers. Usually contain a junior department store or
variety as a major tenant, have 15 to 50 other tenants, and
cover 8 to 20 acres.
5004 NEIGHBORHOOD SHOPPING CENTERS - Usually less than 10 acres in
size, containing a supermarket or major drug store as the major
tenant. May include office uses that are not large enough to
code separately. Only centers with 100,000 square feet or more
are included. These centers are inventoried by the Chamber of
Commerce and The Union Tribune (Copley).
5005 SPECIALTY COMMERCIAL CENTERS - Tourist or specialty commercial
shopping areas such as Seaport Village, Marina Village, Ferry
Landing at Coronado, Bazaar del Mundo, Flower Hill, Glasshouse
Square, The Lumberyard, Park Plaza at the Village, the Vineyard,
Promenade, Belmont Park, Del Mar Plaza.
5009 OTHER RETAIL TRADE AND STRIP COMMERCIAL - Includes commercial
activities found along major streets, not in planned centers.
May include mixed office uses which are not large enough to be
identified as a separate area. Also may include mixed
residential uses; i.e., residential on top of commercial, or
residential units adjacent to commercial establishments.
6100 GENERAL OFFICE
6001 HIGH-RISE OFFICES - Buildings with more than 4 stories
containing banking, offices for business and professional
services (finance, insurance, real estate), some retail
activities and restaurants. Data provided by KPMG (formerly
TURI, Mountain West) is used to classify office buildings.
6002 LOW-RISE OFFICES - Buildings with less than 5 stories containing
banking, offices for business and professional services
(finance, insurance, real estate), some retail activities and
restaurants. Data provided by KPMG (formerly TURI, Mountain
West) is used to classify office buildings.
6003 GOVERNMENT/CIVIC CENTERS - Large government office buildings or
centers (outside of military reservations) and civic centers, or
city halls of local governments.
6100 GENERAL PUBLIC SERVICES
6101 CEMETERIES
6102 CHURCHES
6103 LIBRARIES
6104 POST OFFICES
6105 FIRE/POLICE STATIONS
6108 MISSIONS
6109 OTHER PUBLIC SERVICES - Cultural facilities, museums, art
galleries, social service agencies, humane societies.
6500 GENERAL HOSPITALS
6501 MAJOR HOSPITALS - UCSD, VA, Balboa Naval Hospitals
6502 GENERAL HOSPITALS - Hospitals not included above.
6509 OTHER HEALTH CARE - Medical centers and buildings or offices,
health care services and other health care facilities. Smaller
medical offices and facilities may be included within office,
strip commercial or other surrounding uses.
6701 MILITARY USE - Training grounds, defense installations, storage
depots and maintenance areas found within miliary installations
which do not fall into any other land use category.
6800 GENERAL SCHOOL
6801 MAJOR UNIVERSITY - SDSU or UCSD
6802 OTHER UNIVERSITIES AND COLLEGES
6803 JUNIOR COLLEGES - Includes trade or vocational schools.
6804 SENIOR HIGH SCHOOLS
6805 JUNIOR HIGH SCHOOLS AND MIDDLE SCHOOLS
6806 ELEMENTARY SCHOOLS
6807 SCHOOL DISTRICT OFFICES
7200 GENERAL COMMERCIAL RECREATION
7201 TOURIST ATTRACTIONS - Sea World, Zoo, Wild Animal Park, etc.
7202 STADIUMS/SPORTS ARENAS - Sports Arena, San Diego Stadium.
7203 RACETRACKS - Del Mar, EL Cajon Speedway, Carlsbad Raceway, San
Luis Rey Downs.
7204 GOLF COURSES - Clubhouses, swimming and tennis facilities, and
parking lots associated with the golf course are also included.
7206 CONVENTION CENTERS - Centre City, Embarcadero.
7207 MARINAS
7209 OTHER RECREATION - RV parks, drive-in theaters, campgrounds,
boys/girls clubs, YMCA's, rifle ranges, swim clubs, and stand-
alone movie theaters.
7601 ACTIVE PARKS - Recreation areas and centers containing one or
more of the following activities: tennis or basketball courts,
baseball diamonds, soccer fields, or swings. Examples are Robb
Field, Morley Field, Diamond Street Recreation Center, Presidio
Park. Smaller neighborhood parks with a high level of use are
also included as active parks.
7602 PASSIVE PARKS - State, regional and local parks, National
monuments which allow public access and have some sort of
improvements or developments and facilities. Examples are
Cabrillo National Monument, Sunset Cliffs.
7603 OPEN SPACE RESERVES, PRESERVES - Wildlife and nature preserves,
lands set aside for open space, and parks with limited
development and access. Determined primarily from secondary
source materials. Examples are Torrey Pines State Reserve,
Penasquitos Canyon Reserve, San Elijo Ecological Preserve,
Nature Conservancy Properties.
7604 ACTIVE BEACHES - Accessible sandy areas along the coast or major
water bodies (San Diego and Mission Bay) allowing swimming,
picnicking, and other beach-related recreational activities.
Usually has parking associated with it.
7605 PASSIVE BEACHES - Other sandy areas along the coastline with
limited parking and access (beaches along cliffs, or near
preserves).
8000 GENERAL AGRICULTURE
8001 ORCHARDS AND VINEYARDS
8002 INTENSIVE AGRICULTURE - Nurseries, greenhouses, flower fields,
dairies, livestock, poultry, and equine ranches.
8003 FIELD CROPS - Row crops, grains, pasture, fallow.
9101 VACANT AND UNDEVELOPED LAND
9200 WATER
9201 BAYS, LAGOONS
9202 INLAND WATER - Lakes, reservoirs and large ponds.
9501 RESIDENTIAL UNDER CONSTRUCTION - Usually located near existing
residential developments.
9502 COMMERCIAL UNDER CONSTRUCTION - Usually located near existing
commercial or residential areas.
9503 INDUSTRIAL UNDER CONSTRUCTION - Usually located near existing
industrial or commercial developments.
CAPACITY FILES
Capacity files are produced by growth models and used in trip
generation. A different capacity file exists for each growth
alternative and forecast interval. Capacity files are created on
SANDAG's PRIME computer and transferred to SUN workstations for use in
transportation modeling. Capacity files are usually located under a
temporary workspace created to run a particular alternative. Capacity
files have one record for each base-year/future-year land use
combination within each MGRA. The number of records varies as
databases are modified. Capacity files are ASCII files with the
following format.
Columns Variable Type Description
1-6 I6 MGRA
7-12 I6 Census Tract
13-18 I6 ZUM
19-24 I6 Sphere
25-30 I6 City code
31-36 I6 Forecast Interval, where:
0 = 1990-1995
1995 =
1995-2000
2000 = 2000-2005
2005 = 2005-2010
2010 = 2010-2015
37-39 I3 Development Type Code, where:
1 = Developed to capacity
2 = Unusable
3 = Vacant
4 = Employment Infill
5 = Single Family (SF) Infill
6 = Multiple Family (MF) Infill
7 = SF-EMP Redevelopment
8 = MF-EMP Redevelopment
9 = SF-MF Redevelopment
10 =
Site Specific Project
40-46 F7.5 Proportion of Capacity Used
47-55 F9.2 Acres
56-60 I5 Base Year Land Use Code
61-65 I5 Base Year SF Units
66-70 I5 Base Year MF Units
71-75 I5 Base Year MH Units
Columns Variable Type Description
76-80 I5 Base Year Other Units
81-85 I5 Base Year Agriculture Employment
86-90 I5 Base Year Construction Employment
91-95 I5 Base Year Manufacturing Employment
96-100 I5 Base Year TCU Employment
101-105 I5 Base Year Wholesale Trade Employment
106-110 I5 Base Year Retail Trade Employment
111-115 I5 Base Year FIRE Employment
116-120 I5 Base Year Services Employment
121-125 I5 Base Year Government Employment
126-130 I5 Base Year Basic Military Employment
131-135 I5 Base Year On-Base Military Employment
136-140 I5 Future Year Land Use Code
141-145 I5 SF Units Incremental Capacity
146-150 I5 MF Units Incremental Capacity
151-155 I5 MH Units Incremental Capacity
156-160 I5 Other Units Incremental Capacity
161-165 I5 Agriculture Employment Capacity
166-170 I5 Construction Employment Capacity
171-175 I5 Manufacturing Employment Capacity
176-180 I5 TCU Employment Capacity
181-185 I5 Wholesale Trade Employment Capacity
186-190 I5 Retail Trade Employment Capacity
191-195 I5 FIRE Employment Capacity
196-200 I5 Services Employment Capacity
201-205 I5 Government Employment Capacity
206-210 I5 Basic Military Employment Capacity
211-215 I5 On-Base Military Employment Capacity
216-220 I5 Civilian Group Quarters
221-225 I5 Military Group Quarters
226-231 F6.2 Low Residential Density
232-237 F6.2 High Residential Density
238-242 I5 Sequential Counter
SOAPBASE FILES
Soapbase files are produced by growth models and used in trip
generation. A different soapbase file exists for each growth
alternative and forecast interval. Soapbase files are created on
SANDAG's PRIME computer and transferred to SUN workstations for use in
transportation modeling. Soapbase files are usually located under a
temporary workspace created to run a particular alternative. Soapbase
files have five records for each MGRA. Soapbase files are ASCII files
with the following format.
RECORD 1 - GEOGRAPHIC DATA
Columns Variable Type Description
1-6 I6 MGRA
7-11 I5 City Code
12-16 I5 Sphere
17-21 I5 Community Plan Area
22-26 I5 Zipcode
27-30 I4 Transportation Zone
31-36 A6 Census Block
37-42 I6 Census Tract
43-44 I2 SRA
45-49 I5 ZUM
RECORD 2 - RESIDENTIAL DATA
Columns Variable Type Description
7-12 I6 Total Population
13-18 I6 Household Population
19-24 I6 Civilian Group Quarters
25-30 I6 Military Group Quarters
31-36 I6 Total Housing Units
37-42 I6 Single Family Housing Stock
43-48 I6 Multiple Family Housing Stock
49-54 I6 Mobile Home Housing Stock
55-60 I6 Other Housing Stock
61-66 I6 Occupied Housing Units
67-72 I6 Single Family Occupied Units
73-78 I6 Multiple Family Occupied Units
79-84 I6 Mobile Home Occupied Units
85-90 I6 Other Occupied Units
RECORD 3 - EMPLOYMENT DATA
Columns Variable Type Description
7-11 I6 Total Employment
12-16 I6 Civilian Employment
17-21 I6 Military Employment
22-26 I6 Agricultural/Mining Employment
27-31 I6 Construction Employment
32-36 I6 Manufacturing Employment
37-41 I6 T/C/U Employment
42-46 I6 Wholesale Trade Employment
47-51 I6 Retail Trade Employment
52-56 I6 FIRE Employment
57-61 I6 Services Employment
62-66 I6 Government Employment
67-71 I6 On-Base Military Employment
72-76 I6 Basic Military Employment
RECORD 4 - INCOME DATA
Columns Variable Type Description
7-13 I7 Less than $10,000 Occupied Units
14-20 I7 $10,000 - $14,999 Occupied Units
21-27 I7 $15,000 - $24,999 Occupied Units
28-34 I7 $25,000 - $34,999 Occupied Units
35-41 I7 $35,000 - $49,999 Occupied Units
42-48 I7 $50,000 - $74,999 Occupied Units
49-55 I7 $75,000 or More Occupied Units
56-62 I7 Median Income
RECORD 5 - LAND USE DATA
Columns Variable Type Description
7-13 F7.2 Unusable Acres
14-20 F7.2 Developed Low Density Single Family Acres
21-27 F7.2 Developed Single Family Acres
28-34 F7.2 Developed Multiple Family Acres
35-41 F7.2 Developed Mobile Home Acres
42-48 F7.2 Developed Other Residential Acres
49-55 F7.2 Developed Agricultural/Extractive Acres
56-62 F7.2 Developed Industrial/TCU Acres
63-69 F7.2 Developed Services Acres
70-76 F7.2 Developed Office Acres
77-83 F7.2 Developed School Acres
84-90 F7.2 Developed Park Acres
91-97 F7.2 Developed Road Acres
98-104 F7.2 Vacant Low Density Single Family Acres
105-111 F7.2 Vacant Single Family Acres
112-118 F7.2 Vacant Multiple Family Acres
119-125 F7.2 Vacant Mobile Home Acres
126-132 F7.2 Vacant Other Residential Acres
133-139 F7.2 Vacant Agricultural/Extractive Acres
140-146 F7.2 Vacant Industrial/TCU Acres
147-153 F7.2 Vacant Services Acres
154-160 F7.2 Vacant Office Acres
161-167 F7.2 Vacant School Acres
168-174 F7.2 Vacant Road Acres
175-181 F7.2 Redevelop Single-to-Multiple Family Acres
182-188 F7.2 Redevelop Single-to-Employment Acres
189-195 F7.2 Redevelop Multi Family-to-Employment Acres
196-202 F7.2 Infill Single Family Acres
203-209 F7.2 Infill Multiple Family Acres
210-216 F7.2 Infill Employment Acres
217-223 F7.2 Total Acres
224-230 F7.2 Total Developed Acres
231-237 F7.2 Total Vacant Acres
CHAPTER 5
TRANSPORTATION NETWORKS
TRANSPORTATION NETWORKS
Networks play an important role in travel forecasting, determining
basic patterns of trip making, as well as volumes on specific roads
and transit facilities. SANDAG maintains an extensive network
database that ties together highway modeling, transit modeling, and
other network-related mapping and analysis needs.
A master transportation network is stored and maintained as an
Arc/Info coverage. Arc/Info uses the term "coverage" to refer to all
of the files which together represent a system in digital form. The
network coverage contains the following facilities that either exist
or are planned:
þ Freeways
þ Ramps
þ High Occupancy Vehicle (HOV) lanes
þ General Plan circulation streets
þ Rail lines
þ Streets used by public transit
þ Highway access links
þ Some local streets needed for system connectivity
Having one file with all network elements minimizes coding time,
facilitates error checking, eliminates unintended discrepancies
between alternatives, and allows comparisons between different model
runs. Networks for specific applications are selected from the master
coverage.
Arc/Info breaks down network information into arcs, nodes, vertices,
and routes. Coders enter arcs representing roads and rail lines.
Arc/Info depicts arc curvature by locating vertices along arcs.
Arc/Info software creates nodes at intersections of arcs in the
system. Coders can further subdivide arcs with pseudo-nodes located
at traffic signals and transit stops between major street
intersections. Routes are coded to represent prohibited turn moves.
Finally, coders select the path taken by each transit route. Arc/Info
maintains a sequential list of arcs through which the route passes.
In all, SANDAG's master transportation coverage includes approximately
42,000 arcs and 34,000 nodes.
Care is taken to accurately code the alignment of transportation
facilities so that they register properly with other digitized
features. The location of existing roads and rail lines was initially
obtained from other agencies and cross checked with 1990 aerial
photographs of the Region. Planned freeway alignments are drawn from
route location studies conducted by Caltrans. Longer range proposed
freeways without adopted alignments are located less precisely.
Future street alignments are obtained from each jurisdiction's general
plan circulation element and subsequently reviewed with local
planners. SANDAG works closely with transit agencies so that future
rail line alignments are available from on-going studies.
The master transportation coverage includes highway zone connectors.
These links are required for highway modeling and schematically
represent traffic access from transportation zones to the highway
network. One end of each access link lies at the label point of the
zone coverage, ensuring correct node numbering.
Once the physical locations of transportation facilities are entered,
information about their existing and proposed conditions can be coded.
The 118 arc and 33 node attributes maintained in the transportation
coverage are described at the end of the chapter. Only three
attributes must be coded for every arc: name, functional class, and
initial year. Other attributes provide the flexibility to represent
almost any network condition.
Some data items are maintained by Arc/Info. Unique link and node
numbers are assigned by the software each time the network is edited.
The software computes arc lengths as arcs are added or repositioned.
Arc/Info stores a user coded link identification number that SANDAG
uses to assign a unique number to each arc that only changes if an arc
is split. This arc number enables data to be related between
alternatives.
Coders are required to enter road names for each arc and SANDAG
software assigns cross street names at arc endpoints. Names are
useful for locating arcs in edit sessions as well as generating
reports and plots. Arc/Info determines the city in which the arcs and
nodes are located. A SANDAG program assigns route and link numbers
based upon street names, which enable arcs to be sorted sequentially
by route for reporting purposes.
A number of different classification schemes are coded. Most
important is the functional classification assigned by local
jurisdiction circulation elements. A number of other data items can
be given default values based upon the local functional
classification. In addition, functional classification is the primary
method of summarizing network output. A federal functional
classification scheme is coded for generating network reports.
Finally, a regional classification identifies whether arcs are under
Caltrans jurisdiction, Congestion Management Plan routes, regionally-
significant arterials, or other.
SANDAG monitors level of service for freeways and regionally-
significant arterials using Highway Capacity Manual (HCM) procedures.
The following data are coded for this analysis: existing peak hour
percent, truck percent, arterial class, directional split, peak hour
factor, intersection control type, intersection geometries,
intersection turn moves, approach green-to-cycle time ratio, approach
arrival type, signal cycle length, and signal type. Caltrans provides
data for state routes, and local jurisdictions provide data for other
regionally-significant arterials.
Link capacities are a key input to highway modeling. One or two-way
operation, median type, main lanes, auxiliary lanes, high occupancy
vehicle lanes, intersection control type, and intersection approach
lanes allow detailed capacities to be computed. SANDAG does not yet
have a systematic monitoring system for collecting roadway capacity
data. Existing conditions are obtained from a variety of sources.
Freeways, regional arterials, and freeway ramps were driven in 1990 to
survey some of this information. Characteristics for other facilities
are drawn from previous studies. In all cases, plots of network data
are reviewed by local planners.
Future conditions are also drawn from a number of different sources,
including:
þ Caltrans descriptions and diagrams of proposed projects
þ Regional Transportation Improvement Program project listings
þ Local Capital Improvement Program project listings
þ Local circulation plans
Travel times are another important transportation modeling input.
Posted speeds and intersection control devices are used to compute
travel time inputs to the models. Most jurisdictions conduct speed
surveys in order to establish speed limits. This information is coded
where available. Station-to-station light rail speeds are entered for
existing rail lines and those lines nearing implementation. Existing
posted speeds are assumed to remain the same on roads without proposed
improvements. Otherwise, speeds on proposed facilities are entered
based upon the facility's functional classification.
Nodes are placed at all intersections, traffic signals, existing and
proposed bus stops, and rail stations. The type of intersection
control and traffic signal characteristics are coded for highway
network purposes. Node attributes such as type of stop, park-and-ride
provisions, timed transfer conditions, and unusual station dwell times
are coded at transit stops for transit modeling purposes. Most of
this information is available from SANDAG's bus stop inventories and
transit operators.
Physical attributes of an arc may vary by direction. Arc/Info refers
to "from" nodes and "to" nodes as a way of identifying link direction.
"From-to" direction and "to-from" direction arc attributes are
included so that directional data can be coded.
Because the master transportation coverage contains both existing and
planned facilities, attributes that may change over time are repeated
for initial, upgrade, and final conditions. For example, a freeway
may now have four lanes in each direction that would be coded in the
initial from-to lane item. The Regional Transportation Improvement
Program (RTIP) may propose a near term project to widen the freeway to
five lanes in each direction that would be coded in the upgrade from-
to lane item. Finally, the Regional Transportation Plan (RTP) may
call for a high occupancy vehicle lane to be added in the long term
that would be coded in the final from-to HOV lane item.
Initial, upgrade, final, and delete years are coded to indicate the
phasing of proposed improvements. Using the previous example, the
initial year would be coded as 1990 since the freeway already exists.
The upgrade year would be coded with the year the RTIP shows the new
auxiliary lane being constructed. The final year would reflect the
phasing from the RTP. Some roads will be vacated in the future,
usually in conjunction with the construction of a nearby freeway. The
delete year indicates the year a road will be closed to traffic.
The phasing of transportation projects is difficult to obtain and
subject to frequent changes as new funding estimates become available.
Phasing is tied to the RTIP, RTP, and, where available, local capital
improvement programs. RTIP and RTP projects have been assigned
numbers in order to simplify coding of phasing scenarios. These
project numbers and descriptions are entered into a table along with
years that the projects would be opened to traffic under alternative
phasing plans. These same project numbers are coded in the master
transportation coverage in initial, upgrade, final, and delete project
number items that override corresponding year items. The project
phasing table can be related to the transportation coverage and used
to input new phasing assumptions without actually editing
transportation coverages.
DATA FILE
DOCUMENTATION
TRANSPORTATION COVERAGE
The master transportation network coverage is called "tcov" and is
located under the /max7/data/covs directory.
ARC ATTRIBUTES
TCOV has one record for every arc with the following arc attributes.
FNODE# ARC assigned node number at "from" end of arc.
TNODE# ARC assigned node number at "to" end of arc.
LPOLY# ARC assigned left polygon number.
RPOLY# ARC assigned right polygon number.
LENGTH ARC computed length of arc (feet).
TCOV# ARC assigned unique, sequential ID number.
TCOV-ID User assigned unique, fixed ID number.
FX ARC assigned x-coordinate at FNODE#.
FY ARC assigned y-coordinate at FNODE#.
TX ARC assigned x-coordinate at TNODE#.
TY ARC assigned y-coordinate at TNODE#.
ROADID RUIS ID number.
WHO AML assigned login initials and date of last edit.
PFX Road name prefix from RUIS.
SFX Road name suffix from RUIS.
ID RUIS ID number.
TMP1 Temporary variable.
TMP2 Temporary variable.
RTNO Program assigned route number.
LKNO Program assigned sequential link number.
NM Road name.
FXNM Program assigned cross street name at "from" end of arc.
TXNM Program assigned cross street name at "to" end of arc.
AN Program assigned Tranplan node number at "from" end.
BN Program assigned Tranplan node number at "to" end.
COSTAT Count station number.
ADTLK ADT link number.
ADTVL ADT in hundreds.
PKPCT Peak hour percent.
TRPCT Truck percent.
SECNO Section number for level of service analysis.
DIR Link direction, where:
1 = Southbound,
2 = Eastbound,
3 = Northbound,
4 = Westbound.
FFC Federal functional class, where:
1 = Not Classified,
11 = Urban Interstate,
12 = Urban Freeway or Expressway,
13 = Urban Principal Arterial,
14 = Urban Minor Arterial,
15 = Urban Major Collector,
16 = Urban Minor Collector,
17 = Urban Local,
21 = Rural Interstate,
22 = Rural Principal Arterial,
23 = Rural Minor Arterial,
24 = Rural Major Collector,
25 = Rural Minor Collector,
26 = Rural Local.
CLASS Arterial class for level of service class, where:
1 = Urban Design,
2 = Intermediate Design,
3 = Suburban Design.
OSPD Observed speed (miles per hour).
ASPD Calibrated adjusted speed (miles per hour).
OLOS Observed level of service, where:
1 = LOS A
2 = LOS B
3 = LOS C
4 = LOS D
5 = LOS E
6 = LOS F.
DYR Year (if any) that arc is deleted.
DPROJ Delete project number.
IYR Initial year that the arc is first opened to traffic.
IPROJ Initial project number.
IFC Initial circulation element functional classification,
where:
1 = Freeway,
2 = Prime Arterial,
3 = Major Arterial,
4 = Collector,
5 = Local Collector,
6 = Rural Collector,
7 = Rural Light Collector,
8 = Local Street,
9 = Ramp,
10 = Zone Connector,
11 = Rail Line,
12 = Bus Street,
13 = ADT Link,
14 = HOV Facility.
IJUR Initial jurisdiction controlling arc, where:
1 = Local Facility,
2 = Regional Facility,
3 = Non-State, Congestion Management Facility,
4 = State Facility.
ISPD Initial posted speed (miles per hour).
IWAY Initial one- or two-way operation, where:
1 = One way,
2 = Two way.
IMED Initial median condition, where:
1 = No Median,
2 = Raised Median,
3 = Center Left Turn Lane.
IFTIMP Initial AML-computed impedance in "from-to" direction.
IFTLN Initial mid-block lanes in "from-to" direction.
IFTAU Initial auxiliary lanes in "from-to" direction.
IFTHO Initial high occupancy vehicle lanes in "from-to" direction.
IFTPCT Initial direction split in "from-to" direction.
IFTPHF Initial peak hour factor in "from-to" direction.
IFTCNT Initial intersection control type at "to" end, where:
0 = No Control,
1 = Traffic Signal,
2 = Allway Stop Sign,
3 = Two-way Stop Sign,
4 = Ramp Meter,
5 = LRT Crossing,
6 = Toll Booth,
7 = Prevent control.
IFTTL Initial intersection approach through lanes at "to" end.
IFTRL Initial intersection approach right turn lanes at "to" end.
IFTLL Initial intersection approach left turn lanes at "to" end.
IFTAT Initial intersection arrival type at "to" end.
IFTGC Initial intersection green-to-cycle ratio at "to" end.
IFTTV Initial intersection approach through volume percentage at
"to" end.
IFTRV Initial intersection approach right turn volume percentage
at "to" end.
IFTLV Initial intersection approach left turn volume percentage at
"to" end.
ITFIMP ... ITFLV Initial data in "to-from" direction.
UYR ... UTFLF Data for first time arc is upgraded.
XYR ... XTFLF Data for second time arc is upgraded.
NODE ATTRIBUTES
TCOV has one record per node with the following node attributes.
ARC# ARC assigned TCOV# of an arc at the node.
TCOV# ARC assigned unique, sequential node identification number.
TCOV-ID User assigned unique node identification number.
XNM1 Program assigned cross street name.
XNM2 Program assigned cross street name.
TEMP Temporary variable.
IUCNT Program assigned initial/upgrade control code for plotting.
SPHERE Sphere number, where:
100 = Carlsbad,
200 = Chula Vista,
300 = Coronado,
400 = Del Mar,
500 = El Cajon,
600 = Encinitas,
700 = Escondido,
800 = Imperial Beach,
900 = La Mesa,
1000 = Lemon Grove,
1100 = National City,
1200 = Oceanside,
1300 = Poway,
1400-1499 = City of San Diego Planning Areas,
1500 = San Marcos,
1600 = Santee,
1700 = Solana Beach,
1800 = Vista,
1900-1999 = County of San Diego Planning Areas.
HNODE Unique node number for Tranplan highway models.
IYR Initial year that the node first exists.
IPROJ Initial project number.
IJUR Initial jurisdiction controlling intersection, where:
1 = Local Intersection,
2 = Regional Intersection,
3 = Non-State, Congestion Management Intersection,
4 = State Intersection.
ICNT Initial intersection control at node, where:
0 = No Control,
1 = Traffic Signal,
2 = Allway Stop Sign,
3 = Two-way Stop Sign,
4 = Ramp Meter,
5 = LRT Crossing,
6 = Toll Booth,
7 = Prevent Control.
IHOV Initial HOV access, where:
1 = HOV access,
2 = No HOV access.
ICYCLE Initial signal cycle length in seconds.
UYR ... UHOV Data for first time node is upgraded.
TNODE Unique node number for Tranplan transit models.
TNODE2 Unique transit node number when route uses node a second
time.
TNODE3 Unique transit node number when route uses node a third
time.
ITYR Initial transit stop year.
ISTOP Initial stop type, where:
0 = No Stop,
4 = Commuter Rail,
5 = Light Rail,
6 = Express,
7 = Local.
IPARK Initial park-and-ride availability, where:
1 = Parking not available,
2 = Parking available,
3 = Parking available with 1 minute penalty,
4 = Parking available with 2 minute penalty,
n = Parking available with n-2 minute penalty.
ITT Initial timed-transfer conditions, where:
1 = Routes timed,
2 = Routes not timed,
7 = San Ysidro Border,
8 = Otay Mesa Border,
9 = Forced transit access point.
IDWELL Initial dwell time in seconds
UTYR ... UTT Data about stop for the first time it is upgraded.
PROJECT TABLE
A project table is maintained as an INFO file called HWYPROJ.LUT under
the /max7/proj/sr8 directory. The INFO file has the following items.
IPROJ Project number corresponding to project numbers coded
on transportation coverage links.
DESCRIPTION Short description of location and nature of project.
ALTYR1 Year that the project would be open to traffic under
first phasing alternative, which is the recommended
Regional Transportation Plan.
ALTYR2 Year that the project would be open to traffic under
second phasing alternative, which is the cost-
constrained Regional Transportation Plan.
ALTYR3 Year that the project would be open to traffic under
third phasing alternative, which is the no-build
Regional Transportation Plan.
ALTYR4 ... ALTYRn
Opening years under other phasing scenarios.
CHAPTER 6
HIGHWAY NETWORKS
HIGHWAY NETWORKS
The master transportation coverage contains all existing and proposed
transportation facilities in Arc/Info format. This part of the
transportation modeling process creates inputs to Tranplan's highway
network modeling functions and obtains highway network outputs used
elsewhere in the modeling process. Files differ for each analysis
year and alternative as roadway assumptions change. The files
include:
þ Arc/Info highway coverages
þ Tranplan network input datasets
þ List of turn prohibitors
þ Tranplan internal network datasets
þ Tranplan zone-to-zone travel time and distance files
Figure 6-1 shows roads that are modeled in SANDAG's 1990 and 2015
highway networks for the western part of the Region. The table below
summarizes highway network characteristics for 1990; and two 2015
Regional Transportation Plan alternatives: a recommended system and a
cost-constrained system. As indicated, an ambitious road-building
program is planned. However, even under the most optimistic funding
assumptions reflected in the recommended plan, roadway capacity
increases fail to match expected population growth of 51% over the
same time span.
Table 6-1
HIGHWAY NETWORK SUMMARY
Measure 1990 2015 Cost Constrained 2015 Recommended
Tranplan Nodes 16,702 17,906 (+ 7%) 17,935 (+ 7%)
Tranplan Links 43,023 46,696 (+ 8%) 46,793 (+ 9%)
Freeway Route Miles 294 345 (+17%) 348 (+18%)
Total Route Miles 3,896 4,461 (+14%) 4,490 (+15%)
Freeway Lane Miles 2,006 2,378 (+18%) 2,514 (+25%)
Total Lane Miles 10,837 14,355 (+32%) 14,643 (+35%)
(Insert Map/Figure 6-1 - Highway Network)
NETWORK PROCEDURES
Arc/Info software selects a highway coverage from the master
transportation coverage, given an analysis year and project phasing
alternative. Transit arcs and roads with an initial year higher than
the analysis year are dropped. Initial conditions are replaced with
upgrade or final conditions on arcs with appropriate upgrade or final
years. Some of the master transportation data items are dropped at
this point to reduce disk space requirements. Other items specific to
highway networks are added. Highway coverage attributes are listed at
the end of the chapter. An Arc/Info "unsplit" function is performed
to remove unnecessary pseudo-nodes and thereby reduce the number of
links and nodes that are carried into Tranplan. Arc/Info "near"
functions are performed to find the nearest highway coverage node to
label points from the zone coverage and the nearest highway coverage
arcs to count station points.
A SANDAG written Fortran program provides the connection between
Arc/Info and Tranplan datasets. The program performs edit checks,
assigns Tranplan node numbers, fills in missing data with default
values, computes capacities and times, generates reports, and creates
a condensed dataset for input to Tranplan. These functions are
described in detail below.
Tranplan requires unique numbers at each node, and permanent node
numbers at the interior end of zone connector links to match the
number of the zone that the connector is representing. By convention,
SANDAG codes zone connectors beginning at zone centroids. This
enables an Arc/Info "near" function to create a conversion between
highway coverage nodes and zone numbers. Other nodes are numbered
sequentially as they are encountered by the program.
Missing values are filled with default values based upon functional
classification listed in Table 6-2. This enables coders to enter only
a few data items for most facilities, but also allows detailed
conditions to be coded where necessary.
Table 6-2
DEFAULT ROADWAY ATTRIBUTES
Functional Posted Mid-Block Intersection Lanes
Classification Speed Median Lanes Through Left Right
Freeway 60 Yes 4 N/A N/A N/A
Prime Arterial 45 Yes 3 3 2 0
Major Arterial 40 Yes 2 2 1 0
Collector 35 No 2 2 1 0
Local Collector 30 No 1 2 0 0
Rural Collector 40 No 1 1 0 0
Rural Light 35 No 1 1 0 0
Local Street 25 No 1 1 0 0
Freeway Ramp 30 No 1 1 1 0
N/A = Not Applicable
CAPACITY
Roadway capacity measures the maximum amount of traffic that can pass
through a section of roadway over a period of time. Capacity is used
in the modeling process to compute congestion delays. Comparing
traffic volumes to capacity indicates the need for new facilities.
Techniques for computing capacity are fairly well standardized in the
Highway Capacity Manual (HCM). However, HCM procedures are primarily
geared for site-specific applications. Some simplifying assumptions
are necessary in order to apply the procedures in a modeling context
where many variables affecting capacity are not available on a
regionwide basis or may be unknown for future conditions.
SANDAG computes directional capacities at level of service "E."
Different procedures are used for freeways, surface streets, and
freeway ramps. Surface street links are further categorized into
those with interrupted and uninterrupted flow. Interrupted flow links
are defined as those with a signal, stop sign, or light rail crossing
at the link endpoint. Freeway ramps are also broken down into
metered, unmetered, and tolled facilities.
Freeway capacity is calculated using the following formula:
C = 1800 + 2000*(ML-1) + 1200*AL + 1800*HL + 200 (if AL>0)
where:
C = Hourly Directional Freeway Capacity (Vehicles per Hour)
ML = Number of Main Freeway Lanes
AL = Number of Auxiliary Lanes
HL = Number of High Occupancy Vehicle Lanes
Main freeway lanes are assumed to carry 2000 vehicles per hour.
Merging movements are assumed to lower the capacity of outside freeway
lanes to 1800 vehicles per hour. The capacity of freeway auxiliary
lanes (truck climbing lanes and ramp merging lanes) is assumed to be
1200 vehicles per hour due to the limited nature of these lanes.
Auxiliary lanes also have the effect of removing most merging moves
from the outside freeway lane. Thus, outside freeway lanes are
assumed to carry 2000 vehicles per hour if auxiliary lanes are present
instead of the usual 1800 vehicles per hour.
Diamond lanes reserved for high occupancy vehicles (HOVs) are planned
for most of San Diego's freeways. These lanes need to operate at
higher speeds than main freeway lanes in order to provide a
ridesharing incentive. A level of service "D" speed is assumed to be
the minimum acceptable speed for an HOV lane, which translates into a
capacity of 1800 vehicles per hour.
Most applications add together main lane, auxiliary lane, and HOV lane
capacities to calculate total directional freeway capacity. Some
studies require more detailed consideration of HOV lane operation.
Highway networks for these studies include separate HOV links and
access ramps. Highway assignment concurrently assigns single and
high-occupancy vehicle trips to appropriate links. This technique
doubles highway assignment computer processing time, limiting its use
to HOV-oriented studies.
Surface street capacity calculations are complicated by the effects of
traffic signals and stop signs on traffic flow. Many transportation
modelers group links into categories such as suburban arterials and
then assign a uniform capacity to all links in a category. SANDAG
takes another approach and codes the explicit location of all existing
and proposed traffic control devices. Intersection capacities are
then computed for controlled links and mid-block capacities are
computed for uncontrolled links. This approach creates capacities
that differ by direction and "lumpy" capacities along routes as
traffic control devices are encountered. While these variations in
capacity may differ from standard practice, they better reflect actual
conditions.
The capacity of uninterrupted urban surface streets is not addressed
in the Highway Capacity Manual. The following SANDAG equation
calculates capacities for these conditions.
C = 1800*L - 300 - 200 (IF M<2)
where:
C = Hourly Mid-block Directional Capacity (Vehicles per Hour)
L = Number of Mid-block Lanes
M = Median Code (0 or 1 Indicates no Median)
A per-lane capacity of 1800 vehicles per hour is assumed that is
consistent with intersection capacity assumptions. Interference with
parked cars and right turn moves is assumed to lower outside lane
capacity by 300 vehicles per hour. If no median or center left turn
lane is present, then left turn moves are assumed to reduce capacity
by another 200 vehicles per hour.
SANDAG assigns signalized link capacities using the following formula:
C = 1800*TL*GC + RL*TC + LL*TC
where:
C = Hourly Intersection Approach Directional Capacity (Vehicles per
Hour)
TL = Number of Intersection Approach Through Lanes
RL = Number of Intersection Approach Free Right Turn Lanes
LL = Number of Intersection Approach Left Turn Lanes
GC = Green-to-Cycle Time Ratio
TC = Turn Lane Capacity (Vehicles per Hour)
A through lane capacity of 1800 vehicles per hour of green time
reflects the amount of traffic that could be accommodated by through
lanes if traffic flowed continuously without signal interruption.
This value is multiplied by a green-to-cycle time ratio, indicating
the fraction of a signal cycle that a particular approach is in the
green phase. Left and right turn lane capacity is added to through
lane capacity to arrive at total approach capacity.
Green-to-cycle time ratios (G/Cs) used by the model reflect signal
operation in a generalized manner based upon the functional
classification of intersecting streets, as shown in Table 6-3.
Percentages are calculated by assuming per-lane approach volumes that
vary by functional classification. Green time is apportioned to each
approach based upon the ratio of cross-street approach volumes.
Actual signal operation is much more complex and not suited for model
simulation.
Table 6-3
GREEN-TO-CYCLE TIME RATIOS
Click HERE for graphic.
The effect of turn lanes is to remove turning vehicles from main
approach volumes. This effect is modeled by adding in capacity that
approximates the number of turning vehicles, as shown in Table 6-4.
Capacities are based upon a daily design capacity; 50% direction
split; 10% peak hour percent; and 10% left and right turn move
assumptions.
Table 6-4
TURN LANE CAPACITIES
Functional Classification Turn Lane Capacity
Primary Arterial 250 VPH
Major Arterial 150 VPH
Collector 100 VPH
Local Collector 100 VPH
Rural Collector 100 VPH
Rural Light Collector 100 VPH
Local Street 100 VPH
Ramp 100 VPH
Table 6-5 summarizes one-way hourly capacities that are produced when
capacity procedures described above are applied to default roadway
attribute assumptions. Note that a range of intersection capacities
is possible, depending upon the cross-street functional
classification.
Table 6-5
DEFAULT CAPACITIES
(Vehicles Per Hour)
Classification Mid-Block Intersection
Freeway 10,800 N/A
Prime Arterial 5,100 2,444 - 4,226
Major Arterial 3,300 1,266 - 2,526
Major Collector 3,100 784 - 2,476
Local Collector 1,300 306 - 1,188
Rural Collector 1,300 162 - 1,170
Rural Light 1,300 90 - 882
Local Street 1,300 90 - 882
Freeway Ramp 1,800 784 - 2,476
N/A = Not Applicable
These same procedures are used to calculate capacities for links that
are controlled by two-way or all-way stop signs. Since only about 5%
of all traffic is on stop-controlled links, development of separate
procedures is not warranted.
Unmetered freeway ramps are assumed to carry 1200 vehicle per hour per
lane. Metered ramps are assigned a capacity of 750 vehicles per hour
per lane. HOV by-pass ramp capacity is set at 1000 vehicles per hour
per lane.
Toll booths are assigned a capacity of 500 vehicles per hour per
booth. The Coronado Bridge is San Diego's only existing tolled
facility, although corridor studies are increasingly considering toll-
ways as improvement alternatives.
TRAVEL TIME
Travel times over links determine the relative attractiveness of links
and thus how much traffic will be assigned to individual links.
SANDAG's program computes input travel times as a function of link
distance, posted speed, and intersection delay. Congestion delay is
an important component of overall travel time that is added later in
the assignment model. The following equation computes travel times:
OT = D*60/S*1.10 + OD
PT = D*60/PS(AG) + PD
where:
OT = Off-peak Travel Time (Minutes)
D = Distance (Miles)
S = Posted Speed (Miles/Hour)
OD = Off-peak Signal Delay (10 Seconds or 0.17 Minutes)
PT = Peak Travel Time (Minutes)
PS = Peak Speed (Miles/Hour)
AG = Assignment Group
PD = Peak Signal Delay (21 Seconds or 0.25 Minutes)
Off-peak freeway speeds are set at 60 MPH in urban areas, even though
they are posted at 55 MPH. California Highway Patrol speed checks
show that 60 MPH better approximates actual operating speeds. A
posted 65 MPH freeway speed is used in rural areas. Non-freeway
operating speeds are assumed to be 10% higher than posted speeds to
match freeway speed increases. This assumption was supported by City
of San Diego spot speed checks that are used to establish speed
limits. An additional 1/4 minute is added to freeway ramps to
represent weaving, acceleration, and deceleration time.
A constant signal delay time of 10 seconds per signal is added on
surface streets. This average delay time was determined from a drive-
through survey of the Region's streets. It includes the time spent
idling at signals, acceleration time, and deceleration time. The
likelihood of stopping at signals is also factored in.
Peak period speeds are input that approximate overall congestion
levels within an assignment groups described in Table 6-7. Speeds are
set for assignment groups as a whole so that there are reference
points for adjusting speeds later in the assignment model. Speed
adjustment factors are determined by tabulating the average peak
period volume-to-capacity ratio within an assignment group and looking
up a corresponding speed from the Highway Capacity Manual. Peak
period signal delay time reflects average congestion levels at
signalized intersections.
Peak period ramp meter delays of 1.5 minutes are added where
appropriate based upon a Caltrans survey of existing ramp meter
operation. Note that this is an input delay. Ramp meter delay times
computed during highway assignment increase in line with congestion.
A three minute penalty is added at toll booths to represent toll
charges. The penalty was adjusted to bring model-estimated volumes
into agreement with observed volumes on the Coronado Bridge.
TRANPLAN INPUT FILE
Tranplan makes use of only a few of the arc attributes from the
highway coverage. The final step of the Fortran program is to write
out a file containing the data items listed in Table 6-6 for input to
the Tranplan Build Highway Network function.
Table 6-6
TRANPLAN HIGHWAY INPUTS
Tranplan Field SANDAG Data
A-Node A-Node
B-Node B-Node
Assignment Group Assignment Group
Distance Distance (Miles*100)
Field Option "T"
Field 1 Off-peak Time (Minutes*100)
Field 2 Peak Time (Minutes*100)
Link Group 1 Lanes
Link Group 2 Sphere
Link Group 3 Functional Classification
Capacity Capacity (Vehicles per Hour)
Volume Arc/Info ID number
B-A Field Option "1" or "2"
A-node and B-node numbers are used by Tranplan to identify links.
Assignment group codes are used in the highway assignment function to
adjust input times based upon congestion levels, as described in the
chapter on highway assignment. SANDAG makes use of ten assignment
group codes listed in Table 6-7 that group links according to
congestion effects on link travel times.
Table 6-7
ASSIGNMENT GROUP DEFINITIONS
Assignment Group Definition
0 Uncontrolled Arterial Link, 30 MPH Posted Speed or
Less
1 Uncontrolled Arterial Link, 35 MPH Posted Speed
2 Uncontrolled Arterial Link, 40 MPH Posted Speed
3 Uncontrolled Arterial Link, 45 MPH Posted Speed
4 Uncontrolled Arterial Link, 50 MPH Posted Speed or
More
5 Freeway Link
6 Ramp-Metered Link
7 Controlled Link, 0.36 Minute Off-peak Link Time or
Less
8 Controlled Link, 0.37 to 0.54 Minute Off-peak Link
Time
9 Controlled Link, 0.55 Minute Off-peak Link Time or
More
Arc/Info automatically calculates link distances in feet based upon
vertex coordinates. The Fortran program converts distances in feet to
distances in miles for input to Tranplan. Tranplan allows either
times or speeds to be coded on links. SANDAG inputs times computed
using the method described above.
Tranplan has three "link group" codes that can be filled with any
integer data that the user finds helpful. Link groups are used to
summarize data for Tranplan reports but do not affect model results.
SANDAG codes lanes, sphere number, and functional classification in
these three fields.
Hourly capacities described above are entered in the capacity field.
Tranplan has a final field available that would normally be used for
coding an observed link volume or a second capacity. SANDAG uses the
field to hold highway coverage Arc/Info ID numbers preceded by either
a "1" or "2." The ID number is carried through Tranplan's highway
assignment function and allows Tranplan-estimated link volumes to be
easily matched back to a highway coverage for plotting and reporting
purposes. A prefix of "1" indicates a Tranplan link is in the
Arc/Info's "from-to" direction, while a "2" indicates a "to-from"
direction.
A "2" in the B-A field option indicates that link data is the same in
both directions for the link. In this case, data described above
would be duplicated for the opposite direction. Since SANDAG codes
link attributes by direction, a "1" is coded in the B-A field option
for all roadway links. Zone connector data is the same for both
directions and a "2" is coded in the B-A field for all zone connector
links.
TURN PROHIBITOR FILE
Certain intersection turn moves may be prevented through signing or
intersection design. The most common occurrence of turn prohibitions
is as at freeway interchanges that are designed to minimize turn
conflicts. Turn prohibitors are coded at these locations to prevent
Tranplan from making turns where they are actually not allowed.
Turn prohibitors are input to the Tranplan Highway Selected Summation
and Equilibrium Highway Load functions through a list of "from" node,
"through" node, and "to" node numbers indicating a prohibited move.
SANDAG codes turn prohibitors as two-arc Arc/Info routes. This allows
turn prohibitors to be easily displayed for checking and editing
purposes. SANDAG's program reads the internal arc numbers making up
each turn prohibitor route, finds the Tranplan nodes at the arc
endpoints, and outputs a turn prohibitor record.
TRANPLAN HIGHWAY NETWORK FILE
Tranplan's Build Highway Network function converts the highway network
file into an unformatted, binary Tranplan highway network file for use
by other Tranplan programs. This internal Tranplan file speeds
computer processing time and reduces disk space requirements. No new
data items are generated, however a number of edit checks are
performed.
TRANPLAN ZONE-TO-ZONE TRAVEL TIME FILES
Highway network travel times between zones are used throughout the
modeling process in the land use, trip distribution, and mode choice
models. The Tranplan Highway Selected Summation function generates
these zone-to-zone travel times. A minimum path algorithm finds the
shortest path between each zone pair. The program then totals times
and distances over the minimum path.
The Tranplan highway network file and turn prohibitor file are input
to the path-building function. Several Tranplan parameters are
available to users for controlling the method used to build paths and
output data. Initially, SANDAG builds minimum time paths based upon
off-peak travel times.
The output from the selected summation process is a table with a row
for each origin zone and a column for each destination zone. Cells of
the table contain the time or distance between zones found in the
path-building process. Travel time tables are stored as Tranplan
unformatted, binary files. Zone-to-zone distance and off-peak travel
time tables are initially produced.
The Highway Selected Summation function leaves zeroes in intra-zonal
cells where the origin zone equals the destination zone. Intra-zonal
times and distances can be calculated through a "nearest neighbor"
technique, using Tranplan's Build Intrazonal Impedances function.
This technique halves the time or distance to one or more adjacent
zones. The placement of zone connector links can significantly affect
intra-zonal impedance estimates using this method.
Arc/Info automatically calculates the area and perimeter of each zone.
Basing intra-zonal impedances upon the actual shape of a zone
eliminates network coding biases and produces better impedance
estimates. A SANDAG Fortran program reads an Arc/Info produced file
of zone perimeters. The program assumes that zones are roughly
square-shaped and approximates the length of each side by dividing the
perimeter by four. Intra-zonal trips would travel somewhat less than
the full length of a zone. The program estimates intra-zonal
distances as 3/4 of the length of each side. Intra-zonal times are
found by dividing distance by intra-zonal speeds. These speeds are
assumed to increase from 20 MPH to 35 MPH as intra-zonal distance
increases, on the theory that large zones are typically rural zones
with higher operating speeds.
DATA FILE
DOCUMENTATION
HIGHWAY COVERAGES
A highway network coverage called "hwycov" is re-selected from the
master transportation coverage and located under a temporary workspace
created to store files for a particular network alternative.
ARC ATTRIBUTES
Highway coverages have one record for each arc with the following arc
attributes.
FNODE# Arc/Info assigned node number at "from" end of arc.
TNODE# Arc/Info assigned node number at "to" end of arc.
LPOLY# Arc/Info assigned left polygon number.
RPOLY# Arc/Info assigned right polygon number.
LENGTH Arc/Info computed length of arc (feet).
HWYCOV# Arc/Info assigned unique, sequential ID number.
HWYCOV-ID User assigned unique, fixed ID number.
AVOL Program assigned above line plot volume.
BVOL Program assigned below line plot volume.
FX Arc/Info assigned x-coordinate at FNODE#.
FY Arc/Info assigned y-coordinate at FNODE#.
TX Arc/Info assigned x-coordinate at TNODE#.
TY Arc/Info assigned y-coordinate at TNODE#.
TMP1 Temporary variable.
TMP2 Temporary variable.
RTNO Program assigned route number.
LKNO Program assigned sequential link number.
NM Road name.
FXNM Program assigned cross street name at "from" end of arc.
TXNM Program assigned cross street name at "to" end of arc.
AN Tranplan highway node number at "from" end.
BN Tranplan highway node number at "to" end.
ADTLK ADT link number.
ADTVL ADT in hundreds.
PKPCT Peak hour percent.
TRPCT Truck percent.
SECNO Section number for level of service analysis.
DIR Link direction, where:
1 = Southbound,
2 = Eastbound,
3 = Northbound,
4 = Westbound.
FFC Federal functional class, where:
1 = Not Classified,
11 = Urban Interstate,
12 = Urban Freeway or Expressway,
13 = Urban Principal Arterial,
14 = Urban Minor Arterial,
15 = Urban Major Collector,
16 = Urban Minor Collector,
17 = Urban Local,
21 = Rural Interstate,
22 = Rural Principal Arterial,
23 = Rural Minor Arterial,
24 = Rural Major Collector,
25 = Rural Minor Collector,
26 = Rural Local.
CLASS Arterial class for level of service class, where:
1 = Urban Design,
2 = Intermediate Design,
3 = Suburban Design.
OLOS Observed level of service, where:
1 = LOS A
2 = LOS B
3 = LOS C
4 = LOS D
5 = LOS E
6 = LOS F.
IYR Initial year that the arc is first opened to traffic.
IJUR Initial jurisdiction controlling arc, where:
1 = Local Facility,
2 = Regional Facility,
3 = Non-State, Congestion Management Facility,
4 = State Facility.
IFC Initial circulation element functional classification,
where:
1 = Freeway,
2 = Prime Arterial,
3 = Major Arterial,
4 = Collector,
5 = Local Collector,
6 = Rural Collector,
7 = Rural Light Collector,
8 = Local Street,
9 = Ramp,
10 = Zone Connector,
11 = Rail Line,
12 = Bus Street,
13 = ADT Link,
14 = HOV Facility.
ISPD Initial posted speed (miles per hour).
IWAY Initial one or two way operation, where:
1 = One way,
2 = Two way.
IMED Initial median condition, where:
1 = No Median,
2 = Raised Median,
3 = Center Left Turn Lane.
IFTLN Initial mid-block lanes in "from-to" direction.
IFTAU Initial auxiliary lanes in "from-to" direction.
IFTHO Initial high occupancy vehicle lanes in "from-to"
direction.
IFTPCT Initial direction split in "from-to" direction.
IFTPHF Initial peak hour factor in "from-to" direction.
IFTCNT Initial intersection control type at "to" end, where:
0 = No Control,
1 = Traffic Signal,
2 = Allway Stop Sign,
3 = Two-way Stop Sign,
4 = Ramp Meter,
5 = LRT Crossing,
6 = Toll Booth,
7 = Prevent control.
IFTTL Initial intersection approach through lanes at "to" end.
IFTRL Initial intersection approach right turn lanes at "to"
end.
IFTLL Initial intersection approach left turn lanes at "to" end.
IFTAT Initial intersection arrival type at "to" end.
IFTGC Initial intersection green-to-cycle ratio at "to" end.
IFTTV Initial intersection approach through volume percentage at
"to" end.
IFTRV Initial intersection approach right turn volume percentage
at "to" end.
IFTLV Initial intersection approach left turn volume percentage
at "to" end.
IFTDC Tranplan "from-to" direction code, where:
1 = Southbound Freeway,
2 = Eastbound Freeway,
3 = Northbound Freeway,
4 = Westbound Freeway,
5 = Southbound Arterial,
6 = Eastbound Arterial,
7 = Northbound Arterial,
8 = Westbound Arterial,
13 = Zone Connector.
IFTOPTM Overall arc off-peak travel time in "from-to" direction.
IFTPKTM Overall arc peak travel time in "from-to" direction.
IFTAG Tranplan assignment group in "from-to" direction.
IFTCP Hourly capacity in "from-to" direction.
ITFLN ... ITFCP Data in "to-from" direction.
NODE ATTRIBUTES
Highway Coverages have one record for each node with the following
node attributes.
ARC# Arc/Info assigned HWYCOV# of an arc at the node.
HWYCOV# Arc/Info assigned unique, sequential node identification
number.
HWYCOV-ID User assigned unique node identification number.
XNM1 Program assigned cross street name.
XNM2 Program assigned cross street name.
IUCNT Program assigned initial/upgrade control code for
plotting.
SPHERE Sphere number, where:
100 = Carlsbad,
200 = Chula Vista,
300 = Coronado,
400 = Del Mar,
500 = El Cajon,
600 = Encinitas,
700 = Escondido,
800 = Imperial Beach,
900 = La Mesa,
1000 = Lemon Grove,
1100 = National City,
1200 = Oceanside,
1300 = Poway,
1400-1499=
City of San Diego Planning Areas,
1500 = San Marcos,
1600 = Santee,
1700 = Solana Beach,
1800 = Vista,
1900-1999=
County of San Diego Planning Areas.
HNODE Unique node number for Tranplan highway models.
IYR Initial year that the node first exists.
IJUR Initial jurisdiction controlling intersection, where:
1 = Local Intersection,
2 = Regional Intersection,
3 = Non-State, Congestion Management Intersection,
4 = State Intersection.
ICNT Initial intersection control at node, where:
0 = No Control,
1 = Traffic Signal,
2 = Allway Stop Sign,
3 = Two-way Stop Sign,
4 = Ramp Meter,
5 = LRT Crossing,
6 = Toll Booth,
7 = Prevent Control.
TRANPLAN HIGHWAY NETWORK LINK FILE
Arc/Info highway coverages are processed to produce ASCII highway
network link files called "netdata" files. These files are temporary
files that are input to Tranplan for building a highway network in
Tranplan format. Highway link files are located under temporary
workspaces created to evaluate alternative networks. The files have
one record per highway link with the following format.
Columns Variable Type Description
1-5 I5 A-node number
6-10 I5 B-node number
11-11 I1 Assignment Group Code
12-15 F4.2 Link Distance (Miles)
16-16 A1 "T"
17-20 F4.2 Off-peak Link Travel Time (Minutes)
21-24 F4.2 Peak Link Travel Time (Minutes)
27-28 I2 Number of Lanes
29-30 I2 Sphere number, where:
1 = Carlsbad,
2 = Chula Vista,
3 = Coronado,
4 = Del Mar,
5 = El Cajon,
6 = Encinitas,
7 = Escondido,
8 = Imperial Beach,
9 = La Mesa,
10 = Lemon Grove,
11 = National City,
12 = Oceanside,
13 = Poway,
14 = City of San Diego,
15 = San Marcos,
16 = Santee,
17 = Solana Beach,
18 = Vista,
19 = Unincorporated Area
31-32 I2 Functional Classification
33-38 I6 Link Capacity (Vehicles per Hour)
39-39 I1 From-to Code, where:
1 = A-B node in Arc/Info from-to direction
2 = A-B node in Arc/Info to-from direction
40-44 I5 HWYCOV-ID
TRANPLAN TURN PROHIBITOR FILE
Arc/Info highway coverages are processed to produce ASCII turn
prohibitor files called "turns.lis" files. These files are input to
Tranplan during "Highway Selected Summation" and "Equilibrium Highway
Load" functions. Turn prohibitor files are located under temporary
workspaces created to evaluate alternative networks. Turn prohibitor
files have one record for each prohibited move with the following
format.
Columns Variable Type Description
1-1 A1 "T"
2-6 I5 From Node Number
7-11 I5 Through Node Number
12-16 I5 To Node Number
CHAPTER 7
TRANSIT NETWORKS
TRANSIT NETWORKS
Transit network procedures are more complicated than those for highway
networks because of the need to account for transit route
characteristics in addition to roadway characteristics used in highway
modeling. It is also important to specify the amount of activity
within walking distance of transit in order to get accurate transit
patronage estimates. SANDAG considers 1/2 mile to be the maximum walk
distance throughout the modeling process.
SANDAG has automated as much of the transit network coding process as
possible. This reduces coding biases that could influence model
results. Automated procedures also enable more alternatives to be
considered than would be possible with manual procedures.
SANDAG's transit modeling procedures differ from those used by most
other agencies. Development of these procedures was partially
dictated by the large number of zones in SANDAG models. SANDAG
procedures separate transit access computations from transit network
computations. Transit access points (TAPs) are located every 1/2 mile
along transit routes. Tranplan software builds minimum time paths and
finds transit travel times between TAPs. There are about 2,000 TAPs
in a network, as opposed to 4,545 transportation zones, saving
computer processing time and disk space.
Transit access files are used within the mode choice model to add in
network access times. Walk access files indicate the fraction of each
zone's trips in homogeneous transit access areas and the TAPs within
walking distance of each area. Auto access files contain drive times
from zones to nearby TAPs.
Figure 7-1 shows routes that are modeled in SANDAG's 1990 and 2015
recommended transit networks. Table 7-1 summarizes transit networks
for 1990 and two 2015 Regional Transportation Plan alternatives: a
recommended system and a cost-constrained system. Both alternatives
call for an expanded light rail system with additional express bus
service. Overall transit service is planned to increase at a somewhat
higher rate than population (51%).
Table 7-1
TRANSIT NETWORK SUMMARY
Measure 1990 2015 Cost Constrained 2015 Recommended
Tranplan Links 9,298 9,564 (+ 3%) 10,778 (+ 16%)
Rail Miles 4,374 6,655 (+ 52%) 19,715 (+351%)
Express Bus Miles 10,050 26,913 (+168%) 26,041 (+351%)
Local Bus Miles 55,991 68,683 (+ 23%) 69,470 (+ 24%)
Vehicle Miles 70,415 102,250 (+ 45%) 115,222 (+ 64%)
(Insert Map/Figure 7-1 - Transit Network)
INPUT FILES
Arc/Info is used to maintain most transit network information in the
master transportation coverage. The use of Arc/Info enables transit
network data to be related to other SANDAG geographic data and produce
high-quality plots for editing and display purposes.
Arc/Info Coverages
An Arc/Info "route attribute table" is associated with the
transportation coverage and contains a sequential listing of arcs that
are used by each transit route. Coders update the route attribute
table to add new routes or alter existing routes. Arc/Info updates
the route attribute table when underlying network changes are made
such as splitting arcs. Coders also locate nodes at transit stops in
the transportation coverage. All node attributes are listed at the
end of Chapter 5. The following attributes specifically apply to
transit networks:
þ Stop type
þ Park and Ride Conditions
þ Dwell times
þ Timed transfer points
Route Description File
A route description file contains transit network information that is
not spatially oriented. The file has an entry for every route
configuration to be included in a transit network alternative. Routes
as defined by transit operators need to be partitioned into route
"configurations" or Tranplan "lines" for modeling purposes. Almost
every route has two configurations that differ by direction. In
addition, many routes have some configurations that take alternate
roads or travel only part of the main route. Finally, service
modifications for a route may be proposed for the future that add more
configurations. Route description files contain the following
information:
þ Transit operator name
þ Route number
þ Route configuration direction
þ Route configuration number
þ Transit mode
þ Morning peak period headway
þ Afternoon peak period headway
þ Mid-day headway
þ Night headway
þ Night hours of operation
þ Cross street names at starting point
þ Cross street names at ending point
The transit operator name is used to summarize data and generate
summary reports but does not affect model results. The route number,
direction, and configuration number together uniquely identify each
route configuration. Descriptions of start and end points of routes
are also included to indicate differences between route
configurations.
Transit modes group together routes with similar characteristics in
order to assign service characteristics and summarize model results.
The table below lists modes used in SANDAG transit modeling.
Table 7-2
TRANSIT MODES
Mode Type of Service
4 Commuter Rail
5 Light Rail
6 Express Bus
7 South County Local Bus
8 North County Local Bus
Headways indicate the time between bus arrivals and hence the
frequency of service for route configurations. The model uses
headways to compute the time transit riders would encounter when
initially waiting for a bus arrival or when transferring between
buses.
Transit operators may adjust the frequency of service throughout the
day to match transit capacity with ridership demand. In order to
reflect these different service levels, headways are coded for morning
peak (6:00 to 9:00 AM), mid-day (9:00 AM to 3:00 PM), afternoon peak
(3:00 PM to 6:00 PM), and night (6:00 PM to 6:00 AM). The hours of
evening service are coded to reflect the variation in night service by
route. Morning peak period and mid-day headways are input to the mode
choice model and affect transit patronage estimates. Afternoon peak
period and night headways only go into the calculation of total daily
transit miles, affecting transit cost and emission estimates.
Base-year headways are obtained from published time schedules. Buses
arrive at regular intervals on most routes and the time between bus
arrivals is obvious. The following equation is used to calculate
average headway on other routes with service variations:
H(P) = 60*HR(P) / N(P)
where:
H = Headway for time period (Minutes)
HR = Hours of service in time period
P = Time period (AM, Mid, PM, Night)
N = Number of buses or trains in time period
A few routes have very limited service during a time period. A
typical example would be commuter express bus service, which may have
two runs in the morning peak period, timed to match peak arrival
times. The equation above would yield a headway of 90 minutes for
these routes. Some practitioners would code a shorter headway since
the longer headway overstates the time a transit rider would actually
wait at a transit stop. SANDAG codes the full computed headway for
these routes to reflect the inconvenience of infrequent service.
Later in the mode choice model, long wait times are factored down to
prevent over-emphasis.
Headways for future year alternatives may be carried over from the
base year, adjusted to meet policy service levels, or adjusted to meet
forecasted ridership levels.
NETWORK PROCESSING
Once the master transportation coverage and route description files
have been edited, a combination of Arc/Info routines and SANDAG
Fortran programs generate transit network inputs for Tranplan's Build
Transit Network function. Network inputs consist of "link" and "line"
records. Line records describe line characteristics and the path of
each transit line. Link records describe network characteristics over
which lines operate.
Arc/Info Procedures
Arc/Info selects a transit coverage from the master coverage for a
particular alternative and year containing arcs, routes, and nodes
used by route configurations listed in the route description file for
the alternative. Other roads that exist in the analysis year are put
into a non-transit coverage which is used for reference purposes when
producing plots. Items not needed for transit modeling are dropped
from the transit coverage. Other items specific to transit modeling
are added. Transit coverage arc and node items are listed at the end
of the chapter.
Transfer Links
A SANDAG Fortran program generates transfer links between routes that
come close to each other but do not intersect. The program reads
route description records, node coordinates, stop types at nodes, and
nodes used by route configurations. Straight-line distances are
computed between stops on a route and stops on other routes after
dropping stops within 1/2 mile of any common stops between route
pairs. When distances of 1/2 mile or less are found, Arc/Info node
numbers and the distance between nodes is written out to a temporary
file.
Tranplan Input File
Another SANDAG Fortran program creates Tranplan transit network input
files. The program reads arcs and nodes associated with each route
configuration in the route description file. Arc/Info transit
coverages contain nodes at transit stops, traffic signals, and street
intersections. This is more detail than is necessary to carry into
Tranplan. The first step in the program is to assign Tranplan nodes
at selected transit stops. The program puts these nodes called
transit access points (TAPs) at route intersections, rail stations,
timed transfer points, transfer link nodes, and other stops every 1/2
mile along a route. TAPs are assigned sequential numbers that are
used in Tranplan modeling. Tranplan does not allow a node to be used
more than once on a line. Duplicate nodes are assigned a "dummy"
number. Transfer links with zero time connect dummy nodes to
underlying nodes.
Once TAPs have been located along a route, the program writes out a
Tranplan link record with the following data:
þ "1" indicating the record is a link record
þ A-node
þ B-node
þ Mode number
þ Distance (miles)
þ AM peak time (minutes)
þ PM peak time (minutes)
þ Mid-day time (minutes)
A-node and B-node numbers hold the sequential TAP numbers assigned at
link endpoints. The mode number contains the mode of the route
variation using the link. Distance is computed by summing up Arc/Info
distances for individual arcs making up a transit link. Morning peak
and off-peak bus times are computed using the following formula:
BT(P) = HT(P) + D*N
where:
BT = Bus link time (minutes)
P = Time Period (AM peak or off-peak)
HT = Highway link time (minutes)
D = Stop Delay (18 seconds or 0.30 minutes per stop)
N = Number of transit stops on link
Highway times include the effects of modeled congestion upon speeds
and are drawn from the highway assignment model. Buses incur
additional delay when stopping for passenger boarding and alighting.
Differences in bus stop activity, the type of routes serving a stop,
and the physical characteristics of a stop make simulating bus stop
delays very complicated. SANDAG uses an average stop delay for all
bus stops that is based upon actual system-wide transit speeds.
Rail station-to-station times for existing light rail lines are drawn
from published time tables. Some proposed light rail lines have been
through studies where station-to-station times were estimated through
train simulation. Station-to-station time estimates are coded
directly where they exist. Preliminary studies of rail extensions
lacking detailed time estimates use station-to-station times from the
following equation:
RT = DI*60/MS + MS/60*(AC(MS)*DE) + DW
where:
RT = Rail link time (minutes)
DI = Link distance (miles)
MS = Maximum link speed (MPH)
AC = Acceleration rate (2.48 MPHPS at 25 MPH to 0.83 MPHPS at 50 MPH)
DE = Deceleration rate (3.31 MPHPS)
DW = Station dwell time (0.33 Minutes)
Maximum speeds of 35 MPH to 50 MPH are coded on rail arcs in the
master transportation coverage. These maximum speeds along with
acceleration rate, deceleration rate, and station dwell time
assumptions allow station-to-station rail times to be computed. The
program estimates a maximum speed based upon acceleration and
deceleration rates for stations spaced too closely together to reach
the maximum coded speed.
Parallel walk transfer links are created to Centre City street links.
These walk links allow transit riders to walk between transit lines in
Centre City that do not intersect. Transfer links created in previous
steps only allow transfers between routes that are within 1/2 mile of
each other.
The program generates all the link records making up a line and then
outputs a Tranplan line record with the following data:
þ "2" indicating the record is a line record
þ Company number
þ Mode number
þ Line number
þ Record sequence number
þ "1" indicating the line is a one-way line
þ AM peak period headway
þ PM peak period headway
þ Mid-day headway
þ Transit nodes
þ Route number, route direction, configuration number
A sequential line number and node listing for each route are generated
by the program. Other line record data items come directly from the
route description file. A company number is assigned based upon the
company name coded in the route description file as listed in the
table below.
Table 7-3
TRANSIT COMPANY DESCRIPTIONS
Company Company Name
1 San Diego Transit Corporation (SDTC)
2 North County Transit District (NCTD)
4 San Diego Trolley, Inc. (SDTI)
5 Chula Vista Transit (CVT)
6 National City Transit (NCT)
7 County Transit System (CTS)
8 Metropolitan Transit Development Board (MTDB)
In addition to Tranplan link and line cards, the program produces
reports summarizing route information. Computer-generated plots
showing route alignments, walk access assumptions, and transit network
inputs are also available.
Tranplan Transit Network
Link and line files produced by SANDAG's program are merged, and
Tranplan's Build Transit Network function is run to produce an
unformatted, binary Tranplan file that is used by Tranplan transit
functions. The Build Transit Network function also performs some edit
checks.
Tranplan Zone-to-Zone Travel Time and Fare Files
Tranplan's Build Transit Paths function is run to determine minimum
time paths between transit access points. A number of parameters are
available in the program to control the types of paths that are built.
SANDAG uses the following parameters:
þ Minimum bus wait penalties of 5.0 minutes
þ In-vehicle time factors of 0.5
þ Wait time factors of 1.25
þ Transfer walk time factor of 1.0
þ Added transfer penalties of 6.0 minutes
The minimum wait penalty of 5 minutes reflects the fact that buses
tend to bunch up on streets with very frequent service. The parameter
prevents over-estimating short Centre City transit trips where
frequent transit service is prevalent.
Wait time factors penalize initial wait and transfer time more heavily
than transit in-vehicle time. This reflects surveys that show transit
riders find wait time to be more onerous than in-vehicle time. The
factors coincide with weights attached to transit travel time
components in the mode choice model.
The transfer link factor was increased until transit paths used
transit rather than transfer links for longer trips within Centre City
San Diego where transfer links are prevalent.
Transfer penalties are added to reflect the cost and inconvenience of
transferring. A six-minute base transfer penalty is consistent with
penalties used in the mode choice model. These penalties were
adjusted until model-estimated transfer rates agreed with actual
transfer rates.
Tranplan's Transit Selected Summation function is run after building
paths to obtain transit time components between TAPs based upon the
minimum path. Walk and auto access times are added in the mode choice
model from transit access files so that access times are not saved.
The following travel time tables are output from the selected
summation process:
þ Initial wait time
þ Transfer wait time
þ Transfer link walk time
þ Number of transfers
þ Commuter rail and light rail in-vehicle time
þ Express bus in-vehicle time
þ Local bus in-vehicle time
Tranplan's Load Transit Station To Station function is run to
determine the number of timed transfer points encountered between TAP-
to-TAP interchanges. Transfer penalties are reduced in the mode
choice model at timed transfer locations.
TAP-to-TAP transit fares are obtained from the transit network using
Tranplan's Build Fare Matrix function and a SANDAG program to account
for San Diego's fare policies. Boarding cash and discounted fares by
company are listed in the Table 7-4 for 1990 and 1993 when a fare
increase went into effect. Base-year model calibration uses 1990
fares. Forecasts use 1993 fares adjusted to 1990 based upon the
Consumer Price Index in both years. Cash fares are discounted to
account for the estimated amount of transit pass usage for each type
of service. Discount rates are calculated by dividing total revenue
by estimated linked trips.
Table 7-4
TRANSIT FARES BY COMPANY
----- 1990 Fares ---------- 1993 Fares ---------
Company Cash Discounted Cash Discounted 1990 $'s
SDTC-Urban $1.00 $0.90 $1.50 $1.35 $1.23
SDTC-Express (1 zone) $1.25 $1.01 $1.75 $1.42 $1.29
SDTC-Express (2 zones) $1.50 $1.22 $2.00 $1.63 $1.48
SDTC-Express (3 zones) $1.75 $1.42 $2.25 $1.83 $1.66
CTS-Local $0.75 $0.67 $1.00 $0.89 $0.81
CTS-Express (1 zone) $2.00 $1.60 $2.50 $2.00 $1.82
CTS-Express (2 zones) $2.25 $1.80 $2.75 $2.20 $2.00
CTS-Express (3 zones) $2.50 $2.00 $3.00 $2.40 $2.18
NCT $0.75 $0.67 $1.00 $0.89 $0.81
CVT $0.75 $0.67 $1.00 $0.89 $0.81
MTDB-Local $1.00 $0.90 $1.00 $0.89 $0.81
MTDB-Urban $1.00 $0.90 $1.50 $1.35 $1.23
NCTD-Local $0.80 $0.71 $1.00 $0.89 $0.81
NCTD-Express $1.25 $1.01 $1.75 $1.42 $1.29
NCTD-Rail (1 zone) - - $2.50 $2.00 $1.82
NCTD-Rail (2 zones) - - $2.75 $2.20 $2.00
NCTD-Rail (3 zones) - - $3.00 $2.40 $2.18
NCTD-Rail (4 zones) - - $3.25 $2.60 $2.36
Fares on the light rail system are variable depending upon the number
of stations traversed as indicated in Table 7-5. In order to assess
rail fares, a list of light rail links is input to Tranplan's Build
Fare Matrix function. A SANDAG program looks up a rail fare based
upon the number of light rail links between TAP-to-TAP interchange.
The program sets the final fare equal to the highest fare encountered
over a minimum path. The final fare could be the boarding fare of the
first bus, the boarding fare of a transfer bus, or a light rail fare.
Table 7-5
LIGHT RAIL TRANSIT FARES
---- 1990 Fares ---------------- 1993 Fares ------------
Company Cash Discounted Cash Discounted 1990 $'s
Centre City $0.50 $0.38 $1.00 $0.77 $0.70
1 Station $0.50 $0.38 $1.00 $0.77 $0.70
2 Stations $0.75 $0.58 $1.00 $0.77 $0.70
3 Stations $1.00 $0.77 $1.50 $1.15 $1.05
4-7 Stations $1.25 $0.96 $1.50 $1.15 $1.05
8-12 Stations $1.50 $1.15 $1.75 $1.34 $1.22
13-17 Stations $1.75 $1.34 $1.75 $1.34 $1.22
18-22 Stations $2.00 $1.54 $1.75 $1.34 $1.22
23-27 Stations $2.25 $1.73 $1.75 $1.34 $1.22
28 Stations $2.50 $1.93 $1.75 $1.34 $1.22
NETWORK VALIDATION
Several quality assurance checks were made on the 1990 baseline
transit network before proceeding with transit model runs. One
concern was how closely estimated transit route times matched actual
times. The table below summarizes estimated route time errors.
Table 7-6
DISTRIBUTION OF ROUTE TIME ERRORS
Error Off-peak Peak
-25% or More 4% 2%
-25% to -10% 7% 6%
-10% to 0% 37% 27%
0% to +10% 34% 36%
+10% to +25% 13% 18%
+25% or More 5% 11%
Published time schedules were used to obtain peak and off-peak run
times. These times were compared with estimated times from the
transit network for each route configuration by computing the
percentage difference between estimated and observed times. As
indicated, network time estimates are quite good and about 2/3 of all
route variations have estimated off-peak times that are within 10% of
actual times. Peak-period network times are generally over-estimated
compared to scheduled times, resulting in higher peak-period error
rates. Only about 1/4 of all route schedules show an increase in
peak-period run times over off-peak times; it is possible that
scheduled peak times are understated to some degree.
Another quality check was to assign a TAP-to-TAP transit trip table
obtained from the 1990 Transit Ridership Survey to the Tranplan
transit network. Model-estimated boardings were within 10% of actual
boardings from SANDAG's Passenger Counting Program for two-thirds of
all transit routes. Routes with large discrepancies between estimated
and actual boardings were examined for coding errors.
Table 7-7 summarizes observed and model-estimated boardings for the
ten highest volume transit routes. The table also compares boardings
by mode and system-wide boardings. The modal comparison indicates a
slight express bus bias in the procedures. System-wide estimated
boardings are within 3% of actual boardings, indicating that network
procedures are correctly representing transfers.
Table 7-7
ASSIGNMENT OF SURVEY TRANSIT TRIPS
Route Observed Estimated Absolute Percent
or Mode Boardings Boardings Error Error
510 35,020 37,650 +2,630 +7%
520 16,856 15,662 -1,194 -7%
7 14,327 14,060 -267 -2%
11 9,128 9,460 +332 +4%
34 7,626 7,369 -257 -3%
2 7,158 6,240 -918 -13%
3 7,036 6,962 -74 -1%
29 6,330 6,376 +46 +1%
302 5,122 4,616 -506 -10%
15 4,994 5,091 +97 +2%
Light Rail 51,876 53,312 +1,436 +3%
Express Bus 13,052 14,054 +1,002 +8%
South Local Bus 116,929 118,232 +1,303 +1%
North Local Bus 29,792 31,374 +1,582 +5%
System Total 211,649 216,972 +5,323 +2%
ACCESS PROCEDURES
Transit access files are produced along with the transit network
described above for each network alternative. These files connect
transportation zones with transit access points (TAPs) and are used
later on in the mode choice model. A set of three SANDAG programs
generate access files.
The first program finds all transit stops within walking distance of
each MGRA. The program reads coordinates and elevations of MGRA
centroids from the MGRA coverage; stop coordinates and elevations from
the transit coverage; and coordinates of walk barriers from a walk
barrier coverage. MGRA and transit coverages have already been
described. The walk barrier coverage contains arcs representing
features such as ridge lines, steep slopes, water body boundaries,
freeways, and fenced property lines that could block walk access.
Elevations of MGRA centroids and transit stops are determined by
overlaying coverages on a 100 foot contour coverage.
A file of trip productions and attractions by MGRA from the trip
generation program is read. Straight-line distances are computed
between MGRA centroids and transit stops in the vicinity of the MGRA.
Elevation differences are weighted by a factor of seven and added to
straight-line distances. Connections between MGRAs and stops that
exceed the 1/2 mile maximum walk distance or cross walk barriers are
deleted. The program writes out records for the remaining MGRA-stop
connections with the following items:
þ MGRA number
þ Zone number of MGRA
þ Stop node number
þ MGRA-stop distance (feet)
þ Home-based productions
þ Non-home-based productions
þ Total attractions
If no stops are within walking distance of the MGRA, the record
contains a zero in the stop number field.
The next program aggregates MGRA level output from the first program
to zones, and generates the walk access file used in transit modeling.
Route configurations at each stop are stored and MGRA-stop connection
records are read for a zone.
For zones outside of Centre City San Diego, the first step is to group
a zone's MGRAs into areas with similar transit access opportunities,
defined as having the same route configurations within a short walk
(0-1/4 mile) and a long walk (1/4 to 1/2 mile) of the MGRA. MGRA trip
ends are summed for each access area and the percentage of the zone's
trip ends in each area is computed. Areas with less than 10% of a
zone's trips are grouped with other areas to avoid splintering the
data too much. Once access areas have been defined, one or more TAPs
are identified for each access area such that access is provided to
all routes within 1/2 mile of the access area.
Access opportunities in Centre City are too complicated to represent
with the procedures described above. Instead, the closest TAP to each
Centre City zone is identified. A Centre City walk network is coded
that allows access to other TAPs in Centre City without explicitly
coding each zone-to-TAP connection.
An output file is written containing the following data items:
þ Zone number
þ Sequential access area number
þ Sequential TAP number for zone
þ TAP node number
þ Distance Category (1 = 0 to 1/4 mile or 2 = 1/4 to 1/2 mile)
þ Percent of zonal trip ends allocated to access area
There is one record in the file for each zone-access area-TAP
combination. Some zones have no entries indicating that the zone is
not accessible to transit by walking. Zones with entries may be
partially inaccessible to transit by walking. The part of a zone
inaccessible to transit is the difference between 100% and the sum of
TAP percentages for each access area in the zone.
The final transit access program generates auto access connections.
An Arc/Info "identity" function is run to find the zone in which each
park-and-ride lot is located. A SANDAG program reads the resulting
lot-to-zone conversion table and zone-to-zone peak-period highway
travel time tables from the post-assignment process. The program
finds travel times from each zone to all park-and-ride lots based upon
the zone in which each park-and-ride lot is located. Optional time
penalties can be coded in the "IPARK" node attribute and are added to
network based times. Time penalties are used to prevent parking
demand at a lot from exceeding supply. Times are manually coded by
comparing supply and demand from a previous model run.
An auto access connection is made from every zone to the closest park-
and-ride lot. Additional auto access records are created for other
lots that are within 15 minutes of the closest lot.
An auto access file is produced with each record containing the
following data items:
þ Zone number
þ TAP number of park-and-ride lot
þ Time to TAP (minutes)
DATA FILE
DOCUMENTATION
TRANSIT COVERAGES (INPUT)
A transit network coverage called "trcov" is re-selected from the
master transportation coverage and located under a temporary workspace
created to store files for a particular network alternative.
ARC ATTRIBUTES
Transit coverages have one record for each arc with the following arc
attributes.
FNODE# Arc/Info assigned node number at "from" end of arc.
TNODE# Arc/Info assigned node number at "to" end of arc.
LPOLY# Arc/Info assigned left polygon number.
RPOLY# Arc/Info assigned right polygon number.
LENGTH Arc/Info computed length of arc (feet).
TRCOV# Arc/Info assigned unique, sequential ID number.
TRCOV-ID User assigned unique, fixed ID number.
AVOL Program assigned above line plot volume.
BVOL Program assigned below line plot volume.
FX Arc/Info assigned x-coordinate at FNODE#.
FY Arc/Info assigned y-coordinate at FNODE#.
TX Arc/Info assigned x-coordinate at TNODE#.
TY Arc/Info assigned y-coordinate at TNODE#.
TMP1 Temporary variable.
TMP2 Temporary variable.
RTNO Program assigned route number.
LKNO Program assigned sequential link number.
NM Road name.
FXNM Program assigned cross street name at "from" end of arc.
TXNM Program assigned cross street name at "to" end of arc.
AN Tranplan highway node number at "from" end.
BN Tranplan highway node number at "to" end.
DIR Link direction, where:
1 = Southbound,
2 = Eastbound,
3 = Northbound,
4 = Westbound.
OSPD Observed speed.
IYR Initial year that the arc is first opened to traffic.
IJUR Initial jurisdiction controlling arc, where:
1 = Local Facility,
2 = Regional Facility,
3 = Non-State, Congestion Management Facility,
4 = State Facility.
IFC Initial circulation element functional classification,
where:
1 = Freeway,
2 = Prime Arterial,
3 = Major Arterial,
4 = Collector,
5 = Local Collector,
6 = Rural Collector,
7 = Rural Light Collector,
8 = Local Street,
9 = Ramp,
10 = Zone Connector,
11 = Rail Line,
12 = Bus Street,
13 = ADT Link,
14 = HOV Facility.
ISPD Initial posted speed (miles per hour).
IWAY Initial one or two way operation, where:
1 = One way,
2 = Two way.
IMED Initial median condition, where:
1 = No Median,
2 = Raised Median,
3 = Center Left Turn Lane.
IFTLN Initial mid-block lanes in "from-to" direction.
IFTAU Initial auxiliary lanes in "from-to" direction.
IFTHO Initial high occupancy vehicle lanes in "from-to"
direction.
IFTCNT Initial intersection control type at "to" end, where:
0 = No Control,
1 = Traffic Signal,
2 = Allway Stop Sign,
3 = Two-way Stop Sign,
4 = Ramp Meter,
5 = LRT Crossing,
6 = Toll Booth,
7 = Prevent control.
IFTTL Initial intersection approach through lanes at "to" end.
IFTRL Initial intersection approach right turn lanes at "to"
end.
IFTLL Initial intersection approach left turn lanes at "to" end.
IFTOPTM Overall arc offpeak travel time in "from-to" direction.
IFTPKTM Overall arc peak travel time in "from-to" direction.
ITFLN ... ITFPKTM Data in "to-from" direction.
SRT1 ... SRT20 Transit Routes on arc, sorted on route number.
NODE ATTRIBUTES
Transit coverages have one record for each node with the following
node attributes.
ARC# Arc/Info assigned TRCOV# of an arc at the node.
TRCOV# Arc/Info assigned unique, sequential node identification
number.
TRCOV-ID User assigned unique node identification number.
XNM1 Program assigned cross street name.
XNM2 Program assigned cross street name.
IUCNT Program assigned initial/upgrade control code for
plotting.
SPHERE Sphere number, where:
100 = Carlsbad,
200 = Chula Vista,
300 = Coronado,
400 = Del Mar,
500 = El Cajon,
600 = Encinitas,
700 = Escondido,
800 = Imperial Beach,
900 = La Mesa,
1000 = Lemon Grove,
1100 = National City,
1200 = Oceanside,
1300 = Poway,
1400-1499=
City of San Diego Planning Areas,
1500 = San Marcos,
1600 = Santee,
1700 = Solana Beach,
1800 = Vista,
1900-1999=
County of San Diego Planning Areas.
HNODE Unique node number for Tranplan highway models.
IYR Initial year that the node first exists.
IJUR Initial jurisdiction controlling intersection, where:
1 = Local Intersection,
2 = Regional Intersection,
3 = Non-State, Congestion Management Intersection,
4 = State Intersection.
ICNT Initial intersection control at node, where:
0 = No Control,
1 = Traffic Signal,
2 = Allway Stop Sign,
3 = Two-way Stop Sign,
4 = Ramp Meter,
5 = LRT Crossing,
6 = Toll Booth,
7 = Prevent Control.
TNODE Unique node number for Tranplan transit models.
TNODE2 Unique transit node number when route uses node a second.
TNODE3 Unique transit node number when route uses node a third
time.
ISTOP Initial stop type, where:
0 = No Stop,
4 = Commuter Rail,
5 = Light Rail,
6 = Express,
7 = Local.
IPARK Initial park-and-ride availability, where:
1 = Parking not available,
2 = Parking available,
3 = Parking available with 1 minute penalty,
4 = Parking available with 2 minute penalty,
n = Parking available with n-2 minute penalty.
ITT Initial timed-transfer conditions, where:
1 = Routes timed,
2 = Routes not timed,
7 = San Ysidro Border,
8 = Otay Mesa Border,
9 = Forced transit access point.
IDWELL Initial dwell time in seconds.
USTOP Upgrade stop type.
UPARK Upgrade park-and-ride availability.
TRANSIT HEADWAYS (INPUT)
A master transit headways file is maintained as an INFO file called
"info.mhwf" under the /max8/data/tr directory. The file contains the
following data items that describe transit route configurations.
OP Transit Operator Name.
ROUTE Transit Route Number.
DIR Direction of Route Configuration, where:
sb = Southbound,
eb = Eastbound,
nb = Northbound,
wb = Westbound,
cw = Clockwise,
cc = Counter-clockwise.
CONFIG Transit Route Configuration Number.
MODE Transit Mode, where:
4 = Commuter Rail,
5 = Light Rail,
6 = Express Bus,
7 = South County Local Bus,
8 = North County Local Bus.
AMPK AM Peak Period Headway (Minutes*10).
PMPK PM Peak Period Headway (Minutes*10).
OFPK Off-Peak Period Headway (Minutes*10).
NITE Night Period Headway (Minutes*10).
NH Hours of Night Operation.
ORIG Description of Route Configuration Origin.
DEST Description of Route Configuration Destination.
VIA Description of Route Configuration Path.
PKOBS Peak Period Observed Running Time (Minutes).
OPOBS Off-Peak Period Observed Running Time (Minutes).
SR890 Indicator for 1990 Base Year Network, where:
0 = Route Configuration Not In 1990 Network
1 = Include Route Configuration in 1990 Network
2 = Include Route Configuration in 1990 Network and
Reports.
CC95 Indicator for 1995 Cost Constrained Network.
.... Indicator Items for Other Networks, as Needed.
ASCII headway files called "headways" are selected from the master
headways file. These files contain one record for each route
configuration to be included in an alternative and are located under
the workspace created to evaluate a network alternative.
Columns Variable Type Description
2-6 A5 Transit Operator Name
9-11 I3 Transit Route Number
13-14 A2 Direction of Route Configuration
16-16 I1 Transit Route Configuration Number
18-18 I1 Transit Mode
20-20 A1 Blank or "p" to include configuration in reports
22-25 I4 AM Peak Period Headway (Minutes*10)
27-30 I4 PM Peak Period Headway (Minutes*10)
32-35 I4 Off-Peak Period Headway (Minutes*10)
37-40 I4 Night Period Headway (Minutes*10)
42-42 I1 Hours of Night Operation
TRANPLAN TRANSIT NETWORK LINKS AND LINES (OUTPUT)
Transit coverages are processed to produce ASCII transit network link
and line files called "trdata" files. These files are input to
Tranplan for building a transit network in Tranplan format. Transit
link and line files are located under temporary workspaces created to
evaluate alternative networks. Transit files have one record per
transit link with the following format.
Columns Variable Type Description
1-1 I1 "1" Indicating a Link Record
2-6 I5 A-node number
7-11 I5 B-node number
14-14 I1 Transit Mode, where:
4 = Commuter Rail
5 = Light Rail
6 = Express Bus
7 = South County Local Bus
8 = North County Local Bus
23-26 F4.1 Link Distance (Miles)
30-32 F3.1 AM Peak Link Travel Time (Minutes)
36-38 F3.1 PM Peak Link Travel Time (Minutes)
42-44 F3.1 Off-peak Link Travel Time (Minutes)
68-72 I5 Arc/Info TRCOV FNODE#
73-77 I5 Arc/Info TRCOV TNODE#
Two-way walk transfer links connect nearby transit routes and have a
slightly different format.
Columns Variable Type Description
1-1 I1 "1" Indicating a Link
2-6 I5 A-node number
7-11 I5 B-node number
14-14 I1 Transit Mode ("1")
23-26 F4.1 Link Distance (Miles)
30-32 F3.1 AM Peak Link Travel Time (Minutes)
36-38 F3.1 PM Peak Link Travel Time (Minutes)
42-44 F3.1 Off-peak Link Travel Time (Minutes)
45-45 I1 "2" Indicating Two-Way Link
Transit lines are coded with from one to 23 records per line,
depending upon the length of the transit line. One record with a "9"
in the first column separates link and line records.
Columns Variable Type Description
1-1 I1 "2" Indicating a Line Record
2-3 I2 Transit Company, where:
1 = SDTC
2 = NCTD
4 = SDTI
5 = CVT
6 = NCT
7 = CTS
8 = MTDB
4-5 I2 Transit Mode, where:
4 = Commuter Rail
5 = Light Rail
6 = Express Bus
7 = South County Local Bus
8 = North County Local Bus
6-8 I3 Transit Line Number
9-10 I2 Line Record Number (1-23)
11-11 I1 "1" for One-Way Operation
12-14 F3.1 AM Peak Headway (Minutes)
15-17 F3.1 PM Peak Headway (Minutes)
18-20 F3.1 Off-peak Headway (Minutes)
27-31 I5 First Node on Line
32-36 I5 Next Node on Line
37-41 I5 Next Node on Line
42-46 I5 Next Node on Line
47-51 I5 Next Node on Line
52-56 I5 Next Node on Line
57-61 I5 Next Node on Line
62-66 I5 Next Node on Line
67-71 I5 Next Node on Line
72-72 A1 Blank or "T" if Last Record for Line
73-73 I3 Transit Route Number
76-77 A2 Direction of Transit Line
78-78 I1 Configuration Number of Transit Line
TRANPLAN TRANSIT NETWORK FARE LINKS (OUTPUT)
A list of commuter rail and light rail transit links is output to an
ASCII file called "farelink." This file is input to Tranplan's Build
Fare Matrix function. Transit fare link files are located under
temporary workspaces created to evaluate alternative networks.
Transit fare files have one record per commuter rail link with the
following format.
Columns Variable Type Description
1-12 A12 "AF=1000, FL="
13-16 I4 A-node number
17-17 A1 ","
18-21 I4 B-node number
22-35 A14 ", M=4, L=1-254"
Files have one record per light rail link with the following format.
Columns Variable Type Description
1-10 A10 "AF=20, FL="
11-14 I4 A-node number
15-15 A1 ","
16-19 I4 B-node number
20-33 A14 ", M=5, L=1-254"
TRANPLAN-ARC/INFO CONVERSION FILE (OUTPUT)
An unformatted, binary file is produced that is later used to match
Tranplan transit outputs with the transit Arc/Info coverage for
plotting and reporting purposes. These "arcline" files are located
under temporary workspaces created to evaluate alternative networks.
The files have one record for each transit coverage arc that transit
lines pass through.
Variable Type Description
I*2 Transit Mode, where:
4 = Commuter Rail
5 = Light Rail
6 = Express Bus
7 = South County Local Bus
8 = North County Local Bus
I*2 Tranplan Transit Line Number
I*4 Sequential arc number from transit coverage
I*4 Sequential node number from transit coverage
I*2 Blank or Tranplan Transit Node Number
A*20 Arc name from transit coverage (NM)
A*20 Cross-street name from transit coverage
I*2 Transit Stop Type, where:
4 = Commuter Rail Station
5 = Light Rail Station
6 = Express Bus Stop
7 = South County Local Bus Stop
8 = North County Local Bus Stop
A*1 Blank or "P" to include line in printouts
MGRA-NODE FILES (INTERMEDIATE)
A list of transit coverage nodes within 1/2 of each MGRA centroid is
output to an ASCII file called "mgranode." This file is used to
determine transit walk connections for zones. MGRA-node files are
located under temporary workspaces created to evaluate alternative
networks. The files have one record for each MGRA-node combination
with the following format.
Columns Variable Type Description
1-10 I10 Zone
11-20 I10 MGRA
21-30 I10 Node Number from Transit Coverage (TRCOV#)
31-40 I10 Distance from MGRA to Node (Feet)
41-50 I10 Home-Based Productions in MGRA
51-60 I10 Home-Based Attractions in MGRA
61-70 I10 Non-Home-Based Trip Ends in MGRA
TRANSIT ACCESS FILES (OUTPUT)
A list of zone-to-transit access point (TAP) walk connections is
output to an ASCII file called "wkacc." A similar zone-to-TAP auto
access file is called "auacc." These files are used by the mode split
model to find transit times for zonal interchanges. Access files are
located under temporary workspaces created to evaluate alternative
networks. Walk access files have one record for each homogeneous
transit access area within a zone and TAPs that provide access to all
route configurations within 1/2 mile of the area.
WALK ACCESS FILE:
Columns Variable Type Description
1-5 I5 Zone
6-10 I5 Sequential Transit Access Area
11-15 I5 Sequential TAP Connection for Area
16-20 I5 TAP number
21-25 I5 Distance Category
1 = 1/4 Mile or Less
2 = 1/4 to 1/2 Mile
26-30 I5 Percent of Total Zone Trip Ends In Access
Area
Auto access files have one record for each zone to the closest auto
access TAP. There are additional records to other auto access TAPs
that are within 15 minutes of a zone.
AUTO ACCESS FILE:
Columns Variable Type Description
1-5 I5 Zone
6-10 I5 TAP number
11-15 I5 Peak Period Zone to TAP Time (Minutes*100)
CHAPTER 8
TRIP GENERATION
TRIP GENERATION
The purpose of trip generation is to estimate the number of trip ends
based upon forecasts of demographic variables from the growth
forecasting process. For example, SANDAG surveys show that single-
family dwelling units generate about 20 percent more trips than
multiple-family dwelling units. The model uses this information to
generate fewer trips from high-density areas, as opposed to low-
density areas of the Region. Similarly, as the household mix changes
over time, overall trip generation rates are affected.
The model generates trip ends for ten trip types: home-work, home-
college, home-school, home-shop, home-other, work-other, other-other,
serve passenger, visitor, and regional airport. Trip types are
designed to group together trips with similar travel patterns.
Home-work trips account for commute trips without intermediate stops.
This is an important category of trips since work trips make up a
large proportion of travel during peak periods when most traffic
congestion occurs. Commute trips have a higher propensity to use
transit, a lower vehicle occupancy, and longer trip lengths than
average. Commute trips are also targeted by proposed travel demand
management ordinances. About one-third of all commute trips have an
intermediate stop and are considered to be two trips that fall into
other trip types.
Home-college trips include trips for all forms of higher-education
such as universities and junior colleges. These trips tend to behave
like commute trips since they are also made during peak periods.
College trips may also be the subject of travel demand management
ordinances.
Home-school trips cover elementary, junior high, and senior high
school trips. School trips make up a significant proportion of
transit ridership and have a high percentage of biking and walking.
Schools are usually located close to residential areas so that school
trips are shorter than average.
Home-shop trips are also quite short. These trips often involve
carrying packages so that transit, bike, and walk use is relatively
low.
Serve passenger trips are those trips where the sole purpose is to
pick up or drop off a passenger. These trips are by definition
vehicle driver trips. They tend to focus upon certain land uses such
as schools.
All other trips with one end at home fall into the home-other
category. These trips make up the largest proportion of home-based
trips and are made to a wide range of land uses.
Non-home based trips are split into two trip types: work-other and
other-other. Work-other trips have one end at a work site and include
commute trips with an intermediate stop, lunch-hour trips to shop or
eat, and business-related trips. The other-other category contains
the most trips, made from the widest range of land uses.
While visitor travel is fairly substantial in San Diego, past
transportation models combined visitor trips with other-other trips.
It was decided to model these trips separately since they have fairly
well defined origin and destination patterns. Visitor trips include
trips made by business travelers, tourists, day visitors, and visitors
staying in the homes of San Diego residents.
In order to facilitate ground access studies for existing and proposed
regional airport sites, a separate trip type for these trips was
created, even though they make up a small part of overall regional
travel. Previously, airport trips were included in other trip types
for most model runs. Special estimation procedures were used for
airport studies in order to match survey trip patterns.
The trip generation model estimates trip ends. Each trip has two trip
ends. One end is classified as a trip production and the other end as
a trip attraction. The home end of home-based trips is defined as the
production end and the other end is defined as the attraction end.
The work end of work-other trips is defined as the production end and
the other end as the attraction end. Other-other trip ends are split
evenly into trip productions and trip attractions. Over a 24-hour
period, roughly the same number of trips will originate in a zone as
are destined there. However, residential zones will generate
primarily trip productions while non-residential zones will generate
primarily trip attractions. The production/attraction distinction is
important for the trip distribution model discussed in the next
chapter.
MODEL STRUCTURE
The trip generation model estimates person trip ends. Person trips
account for trips by all forms of transportation, including
automobiles, trucks, taxicabs, motorcycles, public transit, bicycling,
and walking. The model works by applying trip rates to forecasts of
independent variables from the growth models. These trip rates are
listed at the end of this chapter.
Residential Trip Rates
Residential trip rates are expressed as trips per occupied dwelling
unit. Dwelling units can potentially be categorized many different
ways to reflect trip rate differences across the Region and over time
as the demographic makeup of the Region changes. Research suggests
household income, structure type, automobiles per household, household
size, accessibility, residential density, neighborhood type, and
household life cycle as possible ways of classifying households to
improve the accuracy of trip generation estimates.
SANDAG uses a relatively simple stratification of dwelling units by
three structure categories: single family, multiple family, and
mobile home. These categories tie in with local planning studies,
which makes model results easier to understand, troubleshoot, and
update by outside planners.
While adding more variables to the trip generation model would improve
the base-year accuracy of the model, future-year accuracy improvements
would be partially offset by errors in forecasting additional
variables. For example, adding household income to the model may
improve the base-year trip generation estimates. However, the
magnitude of income growth over time is subject to a great deal of
uncertainty. Locating future dwelling units by income level within
the Region is also difficult.
Non-Residential Trip Rates
Non-residential trip rates are expressed as trips per acre and trips
per employee by 80 different land use categories. These land use
categories, described at the end of Chapter 4, are derived from
SANDAG's 1990 Regional Land Use Inventory and local general plans.
Both acre-based and employee-based rates have been calibrated to match
survey data and yield the same number of regionwide trips. Most model
applications use the acre measure of non-residential activity since
acre-based forecasts are easier to understand and are subject to less
error than employment forecasts.
Theoretically, an employee count better measures the intensity of use.
However, employment inventories have a number of problems, making
their accuracy questionable. Employees of employers with multiple
sites are difficult to locate, as are employees with no fixed place of
work, such as construction workers. The location of self-employed
workers is not known. Part-time employment is not always
differentiated from full-time employment. Finally, it is difficult to
place employment sites in the correct zone and land use polygon since
employer addresses may not register well with other geography.
Grouping Standard Industrial Classifications (SICs) is a more commonly
used method of categorizing employment for trip generation than the
land use categorization used by SANDAG. Land uses improve upon SIC
groups since they provide more categories that better reflect actual
trip generation. For example, SIC groups normally treat all
government employees the same, yet a government employee at a school
has different trip generation characteristics than a government
employee at an office building. Furthermore, computing trip rates for
SIC groupings is more difficult than computing land use rates because
most uses have a mix of SIC groupings.
Excessive trip generation sometimes occurs when acre-based trip rates
published by the Institute of Traffic Engineers are applied to
unconstrained land use. SANDAG trip rates are derived from SANDAG's
Travel Behavior Survey and calibrated to match 1990 travel patterns.
SANDAG rates are generally lower than published rates. Furthermore,
SANDAG land use forecasts are constrained to agree with dwelling unit
and employment regional control totals that are converted into land
use consumption using density assumptions.
Centre City San Diego is one area where employee-based rates are used
instead of acre-based rates. Centre City densities are much higher
than regional averages so that acre-based rates underestimate travel.
Centre City also has a greater potential for growth than indicated by
land use redesignations that determine acre-based trips.
Time-of-Day Factors
Some model applications distribute peak period trips separately from
off-peak period trips. The peak period is defined as 6:00 to 9:00 in
the morning and 3:00 to 6:00 in the afternoon. The off-peak period
covers the remaining 18 hours of the day. Factors splitting daily
trip ends into the two time periods are listed at the end of the
chapter. These factors allow subtle peaking characteristics to be
represented by the model. For example, hospitals are staffed
continuously and have a lower work trip peak period split than
offices, which tend to be on an 8-5 schedule.
Factors vary by trip type and land use. Many land use-trip type
combinations have too few survey trip ends to enable factors to be
computed. Regionwide factors for a trip type are used in these cases.
Unique Generators
The model first accounts for trips from unique generators. Unique
generators are major traffic generators where traffic counts indicate
that standard trip rates would misrepresent actual trip making. The
figure on the next page identifies the location of SANDAG's unique
generators. Included are all military bases, major tourist
attractions, major beaches, golf courses, the University of California
San Diego (UCSD), San Diego State University (SDSU), Lindbergh Field,
and Horton Plaza. Approximately 1 percent of all
trip ends are from a unique generator.
Military Bases. The military periodically provides SANDAG with
traffic counts for most military bases. Traffic is assumed to remain
constant over time since there are no firm plans for these bases. The
one exception is the Naval Training Center, which has been slated for
closure and is assumed to be vacated by the year 2000.
Tourist Attractions. The San Diego Convention and Visitors Bureau
tracks attendance at seven tourist attractions. Annual attendance is
divided by 365 and multiplied by two to estimate average daily trip
ends at these sites. Base-year trips are factored by population
growth to estimate future year trips.
Major Beaches. Annual beach attendance estimates are obtained from
the City of San Diego Lifeguard Service and the California State Park
system and converted to average daily trip ends. Future beach trips
are tied to population growth.
(Insert Figure 8-1 - Unique Generators)
UCSD. Base-year trip making was obtained from a 1987 traffic study as
a part of an Environmental Impact Report for a Campus Master Plan.
Future year trips are factored based upon the change in regional
college school enrollment from SANDAG's DEFM model. Recent traffic
counts for SDSU are not available and UCSD trip rates were applied to
SDSU enrollment.
Horton Plaza. This Centre City shopping center is a multi-story
facility that generates too few trips when acre-based rates are
applied. Base-year trip counts were obtained and factored to future
years based upon population growth.
Golf Courses. Golf courses typically meander through several zones.
Although a standard trip rate is applied to all golf courses, they are
treated as unique generators so that traffic can be assigned to the
zone in which the golf course entrance is located.
Trip End Calculations
Once unique generator trips have been accounted for, residential trip
rates are applied to estimates of dwelling units by structure type by
MGRA from SANDAG's SOAP model. Total trip ends are allocated to
productions and attractions by the ten trip purposes using the
percentage splits listed with the rates at the end of the chapter.
Daily trips are apportioned to peak and off-peak periods using factors
by trip type and land use category also listed at the end of the
chapter. Zone level trips are accumulated from MGRA level estimates.
The distribution of home-based trip productions by income level from
each zone is needed by the mode choice model. While residential trips
are being generated, dwelling unit trip rates by income level are
applied and zone level trips by income level are accumulated.
The model next computes trips from non-residential land uses. The
user can elect to apply either acre-based or employee-based trip rates
to SOAP model estimates of acres or employees by land use in each
MGRA. Production/attraction, trip type, and time-of-day factors are
then applied to non-residential trip ends.
Trip ends are multiplied by a factor of ten during the trip generation
process. This factor is included to reduce rounding errors in
Tranplan routines that deal with whole numbers rather than fractional
values. A higher factor would produce trip values that exceed the
maximum value Tranplan can accommodate.
Trip End Balancing
By definition total trip productions must equal trip attractions at
the regional level. However, trip productions are estimated
independently from trip attractions and imbalances between the two
estimates can occur. A balancing routine adjusts trip attractions by
zone so that attractions for home-shop, home-other, and serve
passenger trips equal productions. These adjustments are based upon
the assumption that trip productions tied to households provide a more
accurate estimate of future activity.
Productions and attractions for home-work, home-college, and home-
school, work-other, other-other, visitor and airport trip types are
adjusted to match regional control totals. SANDAG's DEFM model
produces regionwide estimates of total employment and residents by
five-year age groups. Base-year estimates of home-work and work-other
trips are factored by the change in employment over time so that the
ratio of work trips to employees remains constant. Similarly, base-
year college trips are factored by the change in college age residents
and school trips by the change in school age children. Regional
airport trips are factored based upon projections of air passenger
demand from special studies. The other-other and visitor trip types
are not highly correlated to a particular DEFM estimated variable and
are therefore adjusted to match regional population growth.
Table 8-1 illustrates the magnitude of the balancing factors for 1990
base-year and 2015 trip estimates. Base-year balancing factors are
negligible because rates are calibrated to remove imbalances. Future
year balancing factors can be more significant. Work and school trip
productions based upon dwelling unit trip rates exceed control totals
and are factored down. Land use based attractions usually fall
somewhat short of productions and are factored up.
MODEL OUTPUTS
The primary output of the trip generation model is a file of
productions and attractions by the ten trip types for each zone. Zone
level production and attraction files are produced for three time
periods: daily, peak period, and off-peak period. Production and
attraction files are fed directly into the trip distribution model
which follows.
The model also produces an unformatted, binary file of daily
productions and attractions by the ten trips for MGRAs with trip ends.
This file is used in transit modeling to estimate the percent of each
zone accessible to transit. MGRA level trip ends are occasionally
used to show the amount of activity within a certain distance of
selected features.
A zone level unformatted, binary file listing the percent of home
based productions by three income levels is output for subsequent use
by the mode choice model.
Several printed reports summarize trip generation results. A
balancing report shows how trip ends were adjusted by the model during
the balancing process. The number of dwelling units by structure type
and the number of acres or employees by land use category are listed
for each zone, along with the number of trips generated by each
activity within the zone. Finally, a regionwide summary of land uses
and trip ends is produced.
Table 8-1
TRIP END BALANCING FACTORS
---------- 1990 -------------------- 2015 ----------
Trip Type Productions Attractions Productions Attractions
Home-Work 1.000 1.030 0.878 1.088
Home-College 1.000 1.046 0.714 0.962
Home-School 1.000 0.990 0.940 1.281
Home-Shop 1.000 0.990 1.000 1.280
Home-Other 1.000 1.028 1.000 1.206
Work-Other 1.000 0.940 1.064 1.024
Other-Other 1.000 1.002 1.191 1.189
Serve Passenger 1.000 0.995 1.000 1.210
Visitor 1.000 0.973 1.000 0.953
Airport 0.941 1.000 1.366 1.002
Table 8-2 summarizes the expected change in total person trip ends
over time within the Region by Major Statistical Area (MSA). Trip
production growth by MSA tracks dwelling unit growth shown in Table 4-
1, while trip attraction growth tracks employment growth shown in
Table 4-2.
Table 8-3 looks at the change in person trip ends by trip type. Home-
college trips, tied to the growth in college age residents, increase
by the smallest amount. Home-work and work-other trips increase with
employment. Given the sluggish economic projections, work trips also
exhibit relatively low growth rates. The growth in other trip types
reflects the 56% growth in dwelling units forecasted for the region.
Table 8-2
TOTAL PERSON TRIPS BY MAJOR STATISTICAL AREA
------ Productions ------ ------ Attractions ------
MSA 1990 2015 Change 1990 2015 Change
Centre City 277,000 571,000 +106% 387,000 565,000 +46%
Central Area 2,524,000 3,157,000 + 25% 2,257,000 2,975,000 +32%
North City 3,330,000 4,561,000 + 39% 3,317,000 4,804,000 +45%
South Suburban 1,057,000 1,922,000 + 82% 1,108,000 1,873,000 +69%
East Suburban 2,002,000 2,852,000 + 42% 2,008,000 2,823,000 +41%
Northwest
County 1,408,000 2,205,000 + 57% 1,440,000 2,342,000 +63%
Northeast
County 1,488,000 2,523,000 + 70% 1,546,000 2,482,000 +61%
East County 128,000 251,000 + 96% 151,000 238,000 +58%
Region 12,214,000 18,102,000 + 48% 12,214,000 18,102,000 +48%
Table 8-3
PERSON TRIPS BY TRIP TYPE
Trip Type 1990 2015 Change
Home-Work 1,415,000 1,927,000 +36%
Home-College 177,000 203,000 +15%
Home-School 699,000 1,025,000 +47%
Home-Shop 1,591,000 2,491,000 +57%
Home-Other 2,535,000 3,958,000 +56%
Work-Other 1,396,000 1,894,000 +36%
Other-Other 3,084,000 4,669,000 +51%
Serve Passenger 679,000 1,064,000 +57%
Visitor 577,000 755,000 +31%
Airport 61,000 116,000 +90%
Total 12,388,000 18,102,000 +48%
Table 8-4
PERSON TRIP RATE SUMMARY
1990 2015 Change
Home-Work Trips/Employee 1.31 1.30 + 0%
Home-Work Trips/Person 0.58 0.50 -14%
Home-Work Trips/Household 1.64 1.38 -16%
Total Trips/Person 4.92 4.76 - 0%
Total Trips/Household 13.93 13.12 - 6%
MODEL CALIBRATION
The starting point for calibrating the trip generation model was to
obtain trip rates from the 1986 Travel Behavior Survey. New survey
expansion factors were first calculated to bring trip estimates up to
the 1990 base year. These factors were based upon household counts by
structure type, vehicle ownership, and Major Statistical Area from the
1990 Census.
Residential trip rates could be tabulated directly from the Travel
Behavior Survey since surveys were conducted at the household level.
Non-residential trip rates were more difficult to determine. The
survey collected land use type at the start and end of each trip.
With this information, 1990 person trip ends by land use type could be
estimated. SANDAG's 1990 Land Use Inventory provides total acres by
land use type. SANDAG's 1990 employment file was matched with the
land use inventory to determine total employees by land use type.
Land use category definitions differed somewhat between the survey and
land use inventory. Common land uses were found and trips per acre
were computed. Resulting trip rates were cross-checked with SANDAG's
Traffic Generators Manual. The Manual summarizes trip rates at a
number of different land uses based upon traffic counts at selected
sites.
A 1991 Visitor Survey obtained similar data for visitors to the San
Diego area. This survey was tabulated to obtain visitor trip rates by
land use category.
Model calibration also involved running through the entire
transportation model chain and comparing model estimated travel with
ground counts. Selected trip rates were adjusted to bring model
estimates in line with observed data.
Tables 8-5 and 8-6 compare 1990 trip generation model trip end
estimates against trip ends from the 1986 Travel Behavior Survey
expanded to 1990 conditions at the Major Statistical Area level. It
should be noted that the East County MSA is not included in the table
since the Travel Behavior Survey only covered the western part of the
County called the Cordon Area. Centre City and Central Areas are
combined because there is an insufficient Centre City sample size to
allow a meaningful comparison.
Estimated and observed total person trip ends agree closely at the MSA
level. All estimated trip ends are within 5% of observed values.
Home-work trips show a larger percentage difference between estimated
and surveyed values. There was some concern that the larger work trip
differences were indicating a problem with home-work trip generation
procedures that needed to be addressed. On the other hand, errors
could simply reflect survey error due to small sample size.
In order to determine the source of the home-work error, employed
residents living in each MSA were tabulated from the 1990 Census. A
constant home-work trip rate per employed resident was applied and
compared with Travel Behavior data. As shown in Table 8-7, this more
elaborate home-work trip generation method also compares poorly with
Travel Behavior survey data. This points to the survey as being the
source of the problem rather than inadequate trip generation
procedures.
Total person trip ends and home-work trip ends from the two estimation
techniques were also compared with Travel Behavior data by Sub-
Regional Areas shown in Figure 2-5 to further investigate potential
trip generation problems. Root mean square (RMS) errors between
estimated and observed values are presented in Table 8-8. RMS error
is used throughout the report to summarize error rates and is
calculated using the following equation:
RMS = SQRT (SUM [ (OBS(I) - EST(I)) **2 ] / NOBS)
where:
RMS = Root mean square error
SQRT = Square root function
SUM = Summation over all zones
OBS(I) = Observed value for zone "i"
EST(I) = Estimated value for zone "i"
NOBS = Number of observations
Error rates are higher at the Sub-Regional Area level than at the
Major Statistical Area level. However, basic conclusions are the
same.
Table 8-5
OBSERVED AND ESTIMATED TOTAL PERSON TRIPS
-------- Productions ------------ Attractions -----
MSA Observed Estimated Error Observed Estimated Error
Central Area 2,743,000 2,841,000 +4% 2,739,000 2,705,000 -1%
North City 3,497,000 3,364,000 -4% 3,495,000 3,389,000 -3%
South Suburban 1,151,000 1,092,000 -5% 1,150,000 1,123,000 -2%
East Suburban 1,976,000 2,021,000 +2% 1,971,000 2,037,000 +3%
Northwest County 1,376,000 1,440,000 +5% 1,375,000 1,448,000 +5%
Northeast County 1,534,000 1,519,000 -1% 1,530,000 1,555,000 +2%
Cordon Area 12,277,000 12,277,000 +0% 12,260,000 12,260,000 +0%
Table 8-6
OBSERVED AND ESTIMATED HOME-WORK PERSON TRIPS
-------- Productions -------- -------- Attractions ---
MSA Observed Estimated Error Observed Estimated Error
Central Area 362,000 343,000 -5% 376,000 332,000 -12%
North City 449,000 372,000 -17% 464,000 500,000 + 8%
South Suburban 132,000 135,000 + 2% 122,000 118,000 - 3%
East Suburban 213,000 253,000 +19% 202,000 181,000 - 5%
Northwest County 140,000 173,000 +24% 130,000 132,000 + 2%
Northeast County 158,000 178,000 +13% 160,000 181,000 +13%
Cordon Area 1,454,000 1,454,000 + 0% 1,454,000 1,454,000 + 0%
Table 8-7
OBSERVED AND REVISED ESTIMATED HOME-WORK PERSON TRIPS
MSA Observed Estimated Error
Central Area 362,000 311,000 -14%
North City 449,000 389,000 -13%
South Suburban 132,000 138,000 + 4%
East Suburban 213,000 260,000 +22%
Northwest County 140,000 174,000 +24%
Northeast County 158,000 182,000 +15%
Cordon Area 1,454,000 1,454,000 + 0%
Table 8-8
TRIP GENERATION ROOT MEAN SQUARE ERROR
Major Statistical Area Sub-Regional Area
Total Productions 4% 20%
Original Home-Work Productions 16% 47%
Revised Home-Work Productions 17% 48%
Total Attractions 3% 19%
Home-Work Attractions 10% 35%
DATA FILE
DOCUMENTATION
TRIP GENERATION RATES (INPUT)
An ASCII file of person trip generation rates is applied to growth
forecast data in the trip generation model. The trip rate file is
called "rates" and is located under /max7/data/sr8. The file has two
records for each land use category with the following format.
RECORD 1:
Columns Variable Type Description
1-4 I4 Land Use Code
6-33 A28 Land Use Description
34-40 F7.1 Daily Person Trips per Dwelling Unit
41-47 F7.1 Daily Person Trips per Acre
48-56 F9.1 Daily Person Trips per Employee
58-58 A1 "P" for Production End
59-63 F5.3 Production Fraction of Total Trip Ends
64-68 F5.3 Home-Work Fraction of Productions
69-73 F5.3 Home-College Fraction of Productions
74-78 F5.3 Home-School Fraction of Productions
79-83 F5.3 Home-Shop Fraction of Productions
84-88 F5.3 Home-Other Fraction of Productions
89-93 F5.3 Work-Other Fraction of Productions
94-98 F5.3 Other-Other Fraction of Productions
99-103 F5.3 Home-Serve Passenger Fraction of Productions
104-108 F5.3 Visitor Fraction of Productions
109-113 F5.3 Regional Airport Fraction of Productions
RECORD 2:
Columns Variable Type Description
58-58 A1 "A" for Attraction End
59-63 F5.3 Attraction Fraction of Total Trip Ends
64-68 F5.3 Home-Work Fraction of Attractions
69-73 F5.3 Home-College Fraction of Attractions
74-78 F5.3 Home-School Fraction of Attractions
79-83 F5.3 Home-Shop Fraction of Attractions
84-88 F5.3 Home-Other Fraction of Attractions
89-93 F5.3 Work-Other Fraction of Attractions
94-98 F5.3 Other-Other Fraction of Attractions
99-103 F5.3 Home-Serve Passenger Fraction of Attractions
104-108 F5.3 Visitor Fraction of Attractions
109-113 F5.3 Regional Airport Fraction of Attractions
(Insert Land Use Person Trip Generation Rates - page 1)
(Insert Land Use Person Trip Generation Rates - page 2)
(Insert Land Use Person Trip Generation Rates - page 3)
(Insert Land Use Person Trip Generation Rates - page 4)
TIME-OF-DAY FACTORS (INPUT)
An ASCII file of time-of-day factors is applied in the trip generation
model to split daily trip ends into peak and off-peak period trip
ends. The file is called "lupkop" and is located under
/max7/data/sr8. The file has two records for each land use category
listing off-peak fractions. Peak-period fractions are computed within
the trip generation model by taking the complement of off-peak
fractions.
RECORD 1:
Columns Variable Type Description
1-4 I4 Land Use Code
6-33 A28 Land Use Description
35-35 A1 "P" for Production End
37-41 F5.3 Home-Work Off-Peak Fraction of Productions
42-46 F5.3 Home-College Off-Peak Fraction of Productions
47-51 F5.3 Home-School Off-Peak Fraction of Productions
52-56 F5.3 Home-Shop Off-Peak Fraction of Productions
57-61 F5.3 Home-Other Off-Peak Fraction of Productions
62-66 F5.3 Work-Other Off-Peak Fraction of Productions
67-71 F5.3 Other-Other Off-Peak Fraction of Productions
72-76 F5.3 Home-Serve Passenger Off-Peak Fraction of
Productions
77-81 F5.3 Visitor Off-Peak Fraction of Productions
82-87 F5.3 Regional Airport Off-Peak Fraction of
Productions
RECORD 2:
Columns Variable Type Description
35-35 A1 "A" for Attraction End
37-41 F5.3 Home-Work Off-Peak Fraction of Attractions
42-46 F5.3 Home-College Off-Peak Fraction of Attractions
47-51 F5.3 Home-School Off-Peak Fraction of Attractions
52-56 F5.3 Home-Shop Off-Peak Fraction of Attractions
57-61 F5.3 Home-Other Off-Peak Fraction of Attractions
62-66 F5.3 Work-Other Off-Peak Fraction of Attractions
67-71 F5.3 Other-Other Off-Peak Fraction of Attractions
72-76 F5.3 Home-Serve Passenger Off-Peak Fraction of
Attractions
77-81 F5.3 Visitor Off-Peak Fraction of Attractions
82-87 F5.3 Regional Airport Off-Peak Fraction of
Attractions
(Insert Land Use Off-Peak Period Percentages - page 1)
(Insert Land Use Off-Peak Period Percentages - page 2)
(Insert Land Use Off-Peak Period Percentages - page 3)
(Insert Land Use Off-Peak Period Percentages - page 4)
MGRA-ZONE CONVERSION (INPUT)
A file listing the transportation zone for each Master Geographic
Reference Area (MGRA) is used in the trip generation model to total
MGRA trip ends to zones. The file is called "xref.sr8" and is located
under /max7/data/sr8. The conversion file is an ASCII file with one
record for each MGRA. Records in the file have conversions to other
standard SANDAG geography in addition to zones.
Columns Variable Type Description
3-7 I5 MGRA
8-8 I1 Major Statistical Area
10-11 I2 Sub-regional Area
14-19 I6 Census Tract
21-24 A4 Census Block
26-27 I2 Census Block Suffix
29-32 I4 Census Place
34-38 I5 Zipcode
40-44 I5 Sphere
46-50 I5 Community Plan Area
52-56 I5 Zone for Urban Analysis
58-61 I4 Transportation Zone
63-70 F8.2 MGRA Area (Acres)
UNIQUE GENERATOR COVERAGE (INPUT)
A unique generator coverage is used in trip generation to add in
unique generator trip ends. The coverage is called "unique" and is
located under the /max7/data/covs directory. The coverage has a
polygon outlining each unique generator. Label points are positioned
to overlay in the zone where the trips are to be reported. The
following is a list of polygon attributes coded in the unique
coverage.
AREA Arc/Info computed polygon area in square feet.
PERIMETER Arc/Info computed polygon perimeter in feet.
UNIQUE# Arc/Info assigned unique, sequential ID number for
polygon.
UNIQUE-ID User assigned unique, fixed ID number for polygon.
ACRES Polygon area in acres.
MGRANO MGRA at label point.
ZONE Transportation zone at label point.
LU90 1990 Land Use Code.
REPLACE Indicator to add or replace unique generator trips, where:
0 = Add unique generator trip ends to trip generation
model trip ends
1 = Replace trip generation model trip ends with unique
generator trip ends
POINT Indicator to point load trip ends or spread across area,
where:
0 = Add unique generator trip ends in zone overlaying
unique generator label point
1 = Apportion unique generator trip ends based on area
of unique generator in zone
IYR Inial year that unique generator exists.
ITRIPS Initial year unique generator trip ends. (Blank where
trip rate is to be applied to acres.)
UYR Future year for horizon year trip ends. (Blank where
initial trips are to be factored.)
UTRIPS Horizon year trip ends. (Blank where initial trips are to
be factored.)
UNIQUE GENERATOR DATA (INPUT)
An ASCII version of unique generator data is input to the trip
generation model. The file is called "uniqdata" and is located under
the /max7/proj/sr8 directory. The file has one record for each unique
generator with the following format.
Columns Variable Type Description
1-6 I6 MGRA
7-11 I5 Zone
12-24 F14.3 Polygon area in acres
25-29 I5 1990 Land Use Code
30-34 I5 Inial year that unique generator exists
35-41 I7 Initial year unique generator trip ends
42-46 I5 Future year for horizon year trip ends
47-53 I7 Horizon year trip ends
DEFM DATA (INPUT)
An ASCII file of selected variables from the regional DEFM forecast is
created and used to generate regional trip generation control totals.
The DEFM data file is called "totals" and is located under
/max7/data/sr8. The DEFM data file has one record for each year
between 1990 and 2015. The file has the following format.
Columns Variable Type Description
8-11 I4 Year
16-26 F11.3 Population (000's)
27-37 F11.3 Employment (000's)
38-48 F11.3 Retail and Wholesale Trade Employment (000's)
49-59 F11.3 Retail Trade Employment (000's)
60-70 F11.3 Population Age 5-9 (000's)
71-81 F11.3 Population Age 10-14 (000's)
82-92 F11.3 Population Age 15-19 (000's)
93-103 F11.3 Population Age 20-24 (000's)
Another ASCII file used to compute regional controls totals for
airport trips is called "map" and has the forecasted air passenger
demand for the years 1990 to 2015. The file has the following format.
Columns Variable Type Description
1-4 I4 Year
16-26 F10.6 Millions of Annual Air Passengers
REGIONAL CONTROL TOTALS (INTERMEDIATE)
AN ASCII file of regional level trips by forecast year and trip type
are used in the trip generation model to control the number of some
trip types. The file is called "totals" and is located under
/max7/proj/sr8. The file has one record for each year from 1990 to
2015 with the following format.
Columns Variable Type Description
1-4 I4 Year
5-12 I8 Home-Work Trips
13-20 I8 Home-College Trips
21-28 I8 Home-School Trips
29-36 I8 Home-Shop Trips
37-44 I8 Home-Other Trips
45-52 I8 Work-Other Trips
53-60 I8 Other-Other Trips
61-68 I8 Home-Serve Passenger Trips
69-76 I8 Visitor Trips
77-84 I8 Regional Airport Trips
85-92 I8 Total Trips
ZONE PRODUCTIONS AND ATTRACTIONS (OUTPUT)
ASCII files of productions and attractions by zone and trip type are
the primary outputs from the trip generation model. These files are
subsequently input to Tranplan's Gravity Model trip distribution
function. Daily trip ends are called "gmdata.ie," peak-period trip
ends are called "pkgmdata.ie," and off-peak period trip ends are
called "opgmdata.ie." These files are located in temporary workspaces
created to evaluate alternatives. It should be noted that prior to
use in trip distribution, internal-external trip ends are subtracted
during the external trip factoring process and new trip end files are
created with a "noff" suffix, replacing the "ie" suffix. The files
have one trip production record and one trip attraction record for
each zone that has non-zero trip ends. Trip ends are factored by ten.
Columns Variable Type Description
1-2 A2 "GP" or GA" for productions or attractions
4-7 I4 Zone
9-9 I1 "1"
11-17 I7 Home-Work Productions or Attractions (x 10)
18-24 I7 Home-College Productions or Attractions (x 10)
25-31 I7 Home-School Productions or Attractions (x 10)
32-38 I7 Home-Shop Productions or Attractions (x 10)
39-45 I7 Home-Other Productions or Attractions (x 10)
46-52 I7 Work-Other Productions or Attractions (x 10)
53-59 I7 Other-Other Productions or Attractions (x 10)
60-66 I7 Home-Serve Passenger Productions or
Attractions (x 10)
67-73 I7 Visitor Productions or Attractions (x 10)
74-80 I7 Regional Airport Productions or Attractions (x
10)
MGRA PRODUCTIONS AND ATTRACTIONS (OUTPUT)
An unformatted, binary file is produced that has daily productions and
attractions by trip type for each MGRA. MGRA trip ends are used in
transit access procedures and special applications. MGRA files are
located under temporary workspaces created to evaluate alternatives.
The files have one record for each MGRA with the following data.
Variable Type Description
I*2 Zone
I*4 MGRA
I*4 Home-Work Productions
I*4 Home-College Productions
I*4 Home-School Productions
I*4 Home-Shop Productions
I*4 Home-Other Productions
I*4 Work-Other Productions
I*4 Other-Other Productions
I*4 Home-Serve Passenger Productions
I*4 Visitor Productions
I*4 Regional Airport Productions
I*4 Home-Work Attractions
I*4 Home-College Attractions
I*4 Home-School Attractions
I*4 Home-Shop Attractions
I*4 Home-Other Attractions
I*4 Work-Other Attractions
I*4 Other-Other Attractions
I*4 Home-Serve Passenger Attractions
I*4 Visitor Attractions
I*4 Regional Airport Attractions
ZONAL INCOME DATA (OUTPUT)
An unformatted, binary file is produced called "msinc" that summarizes
the fraction of home-based trip ends by income level for each zone.
Income data is used in the mode choice model to apportion trips to
income levels. Income files are located under temporary workspaces
created to evaluate alternatives. The files have one record for each
zone with the following data.
Variable Type Description
I*4 Zone
I*4 Fraction of home-based trip ends in low income level
(*1000)
I*4 Fraction of home-based trip ends in mid income level
(*1000)
CHAPTER 9
TRIP DISTRIBUTION
TRIP DISTRIBUTION
The trip distribution model links together trip ends from the trip
generation model to determine trip interchanges between zones. The
model produces trip tables that contain a row for each production zone
and a column for each attraction zone. Cells in the table contain the
number of trips estimated between zone pairs.
The model is designed to modify trip patterns in response to new
development. For example, the opening of a new shopping center would
shift trips from other nearby shopping areas to the new development.
More distant shopping areas would experience less of an impact.
The model also modifies trip patterns in response to changes in
roadway conditions. Construction of a new road might connect
previously unserved areas. New trip movements would be created
between zones along the new facility at the expense of trips to other
areas.
MODEL STRUCTURE
SANDAG uses Tranplan's Gravity Model function to distribute trips
between zones. The model allocates trip productions from each zone to
other zones in direct proportion to the number of attractions in other
zones and in inverse proportion to the travel time, using the
following formula.
TR(I,J) =
P(I) * [A(J) * FF(TI(I,J)) * BF(J)] / SUM [A(K) * FF(TI(I,K)) * BF(K)]
where:
TR(I,J) = Trip movement between zones "i" and "j"
P(I) = Trip productions in zone "i"
A(J) = Trip attractions in zone "j"
FF = Friction factor for time interval
TI(I,J) = Travel time between zones "i" and "j"
BF(J) = Balancing factor for zone "j"
SUM = Summation over all zone pairs
Friction factors listed at the end of this chapter vary with trip
length and trip type. These factors reflect the importance of travel
time when choosing trip ends. This effect can be seen when home-work
and home-shop friction factors are compared. Shop trips are much
shorter than commute trips so that shop friction factors diminish more
rapidly than work friction factors.
Three gravity model iterations are performed to bring gravity model
attraction estimates in line with trip generation estimates. The
first model iteration typically overestimates trips to highly
accessible areas and underestimates trips to inaccessible areas. The
program computes a balancing factor by dividing estimated attractions
into input attractions. The resulting factor is applied to estimated
attractions in the next cycle.
The gravity model uses three data inputs:
þ Person trip ends by trip type, zone, and time period
þ Zone-to-zone times by time period
þ Friction factors by trip type, interval, and time period
Friction factors remain unchanged over the forecast period. Trip
productions and attractions come directly from the trip generation
model. Highway times are produced from the highway network simulation
described in Chapter 6.
SANDAG applies the transportation models in two stages. Gravity model
inputs differ for each stage. First-stage trip distribution inputs
consist of daily trip ends, off-peak highway times, and daily friction
factors. In the second stage, peak and off-peak trip ends are
distributed separately using peak and off-peak highway times that
reflect congestion delays from the first-stage highway assignment
process. Peak and off-peak friction factors are also input.
As explained in the model overview section, first-stage applications
require fewer inputs, are less time-consuming, and produce results
that are acceptable for most studies. Applications requiring greater
sophistication proceed with the second stage, requiring more
complicated procedures.
In both cases, person trips are distributed based upon highway travel
times. This is a commonly accepted simplification although a more
theoretically correct approach would be to either distribute trips by
mode separately, using times appropriate for each mode, or weight
times for modes by their expected mode share. Since over 90% of all
trips are highway trips, the accuracy improvement from this approach
would be quite minimal.
MODEL OUTPUTS
As described above, the distribution model produces a 4,545 by 4,545
zone trip table for each of the ten trip types, resulting in over 200
million cells of data. One way of summarizing this data is to look at
the average trip lengths computed by the model. Table 9-1 compares
1990 average trip lengths in miles and minutes with 2015 trip lengths.
Comparisons are shown for both first and second-stage applications.
As indicated, first-stage trip lengths increase over time as growth
continues and activities become more dispersed. Second-stage base-
year trip lengths in miles are the same as first-stage trip lengths.
However, future-year second-stage trip lengths in miles increase at a
slightly slower rate than first-stage trip lengths. This is due to
the effects of congestion.
Table 9-1
PERSON TRIP LENGTHS
Click HERE for graphic.
Another way of analyzing model output is to aggregate zones into
larger areas and trip types into broader categories. Tables 9-2 and
9-3 summarize commute trips (home-work and home-college) and total
person trips between Major Statistical Areas (MSAs). As indicated,
the largest interchanges are generally those that stay within MSAs.
In most cases, interchanges with the most trips in 1990 also have the
most trips in 2015. Large percentage increases occur for interchanges
with a Centre City production MSA. Relatively little Centre City
residential development existed in 1990. Between 1990 and 1994, a
large number of Centre City residential units have been built and
occupied, a trend that is expected to continue through the forecast
period. East County interchanges also show large percentage
increases; however, trip volumes are low so the change in the number
of trips is small.
MODEL CALIBRATION
Gravity model calibration involves adjusting friction factors until
model estimated trip length frequency distributions agree with
observed trip length frequencies. A SANDAG Fortran program matches
base-year model estimated travel times and distances between zones
with 1986 Travel Behavior Survey trip records. The program produces
observed average trip lengths and the percent of total trips by one
minute and one mile increments for each of the ten trip types.
A gravity model run is made to distribute base-year trip ends from the
trip generation model using base-year model estimated travel times and
friction factors that are initially set to one for all time periods
and trip types. Another SANDAG Fortran program reads trip tables from
the gravity model, model estimated travel times, observed trip length
distributions, and the initial set of friction factors. The program
adjusts old friction factors using the following equation.
AFF(TI,TT) = OP(TI,TT)/EP(TI,TT) * OFF(TI,TT)
where:
AFF = Adjusted friction factor
TI = Time interval
TT = Trip type
OP = Observed percentage of trips in time interval
EP = Estimated percentage of trips in time interval
OFF = Old friction factor
The program produces a new set of friction factors that are input to
another round of the gravity model. The gravity model and friction
factor adjustment process are repeated several times until model
estimated trip lengths agree with observed trip lengths.
Table 9-2
COMMUTE PERSON TRIPS BETWEEN MAJOR STATISTICAL AREAS
Click HERE for graphic.
Table 9-3
TOTAL PERSON TRIPS BETWEEN MAJOR STATISTICAL AREAS
Click HERE for graphic.
Adjusted friction factors can be irregular for longer trip lengths
where there are few observed trips. Friction factors are smoothed for
time intervals longer than nine minutes to overcome the sparse data
problem. A moving sum of friction factors is computed using time
intervals within 10% of the interval being smoothed. A smoothed
friction factor is computed as the average of the moving sum.
Resulting friction factors are scaled from a maximum value of
"999999." The few irregularities remaining after application of the
smoothing function are manually adjusted.
Second-stage friction factors are calibrated in a similar manner.
Peak and off-peak period observed trip length frequency distributions
are tabulated from the 1986 Travel Behavior Survey by matching either
model estimated peak or off-peak period travel times to survey
records, depending upon the time at the midpoint of surveyed trips.
Friction factors are then calibrated for peak and off-peak time
periods using the friction factor adjustment process described above.
There was initially some question as to whether different friction
factors should be developed for each time period or whether one set
should be applied to both. There is little theoretical basis for two
sets of friction factors since this implies that travelers value time
differently during peak and off-peak periods. However, an examination
of the data showed that peak-period trip distances would be
underestimated and off-peak distances overestimated if only one set of
friction factors was used. This is probably due to a different
character of trip during the two times periods. For example, off-peak
work trips tend to be made up more of part-time and second jobs closer
to home.
There was also a question as to how many gravity model iterations are
appropriate. Each additional gravity model iteration reduces the
error between input and gravity model estimated attractions, but also
adds 1/2 hour of processing time. Base-year gravity model runs were
made with from one to six iterations in order to evaluate the trade-
off between improved accuracy and increased processing time. Table 9-
4 summarizes error rates from each run. Error rates were determined
by first computing the percentage difference between input and model-
estimated attractions for each zone and trip type. Observations with
an error of more than 5% and more than 10% were tabulated. As shown
in the table, error rate improvements grow smaller with each
iteration, and slow considerably after the third iteration.
Screenlines can be used to evaluate trip distribution accuracy. A
screenline totals up traffic counts and estimated traffic volumes
along an imaginary line cutting across a number of facilities. Any
error in the total volume across a screenline can be largely
attributed to trip distribution error, although other steps in the
model contribute some error. Figure 9-1 shows SANDAG's screenlines
that are located along MSA boundaries. Table 9-5 demonstrates that
the distribution model is closely matching screenline crossings.
Table 9-4
ATTRACTION ERROR RATES BY ITERATION
Iterations 10%+ Error 5%+ Error
1 23% 38%
2 14% 27%
3 11% 21%
4 10% 19%
5 8% 18%
6 8% 17%
Table 9-5
OBSERVED AND ESTIMATED SCREENLINE TRAFFIC VOLUMES
Number Location Observed Estimated Error
1 Centre City - Central Area 338,900 318,900 -6%
2 Central Area - North City 957,300 964,700 + 1%
3 Central Area - South Suburban 434,600 457,600 + 5%
4 Central Area - East Suburban 316,600 326,700 + 3%
5 North City - East Suburban 332,800 344,100 + 3%
6 North City - Northwest County 252,400 249,700 - 1%
7 North City - Northeast County 171,600 183,600 + 7%
8 South - East Suburban 3,600 4,200 +17%
9 East Suburban - East County 32,600 35,200 + 8%
10 Northwest - Northeast County 269,600 293,900 + 9%
11 Northeast - East County 16,700 18,700 +12%
Total 3,126,700 3,197,300 + 2%
(Insert Figure 9-1 - Highway Screenlines)
The ability of the model to match observed intra-zonal trips is
another concern when evaluating trip distribution validity. (Intra-
zonal trips are trips that have both trip ends in one zone.) Because
of SANDAG's small zones size, there are relatively few intra-zonal
trips. Table 9-6 shows close agreement between model-estimated and
observed intra-zonal trip percentages.
Table 9-6
OBSERVED AND ESTIMATED
INTRA-ZONAL PERSON TRIP PERCENTAGES
Trip Type Observed Estimated Error
Home-Work 0.3% 0.4% +0.1%
Home-College 0.0% 0.1% +0.1%
Home-School 3.0% 3.8% +0.8%
Home-Shop 1.4% 2.1% +0.7%
Home-Other 2.2% 2.9% +0.7%
Work-Other 2.4% 3.1% +0.7%
Other-Other 3.8% 4.4% +0.6%
Serve Passenger 2.7% 3.6% +0.9%
Visitor 1.6% 2.3% +3.7%
Airport 0% 0% +0%
Total 3.1% 3.6% +0.5%
The extent to which the trip distribution model matches observed
average trip lengths determines the aggregate amount of travel in the
Region. The model calibration process assures close agreement between
observed and estimated average trip lengths as demonstrated in Table
9-7.
Tables 9-8 and 9-9 summarize commute and total trips between MSAs.
The tables compare model-estimated trips with trips from the Travel
Behavior Survey that have been adjusted to remove trip generation
errors. East County was not surveyed in 1986 and, hence, is not
included in the tables. Tables 9-8 and 9-9 have somewhat lower trip
totals than Tables 9-2 and 9-3 that include East County. As
indicated, model estimates agree closely with observed data for most
interchanges with a large number of trips. Small errors in the
estimated number of trips for interchanges with few trips can lead to
large percentage errors that have little bearing on overall model
performance.
Table 9-7
OBSERVED AND ESTIMATED PERSON TRIP LENGTHS
Click HERE for graphic.
Table 9-8
OBSERVED AND ESTIMATED COMMUTE PERSON TRIPS
BETWEEN MAJOR STATISTICAL AREAS
Click HERE for graphic.
Table 9-9
OBSERVED AND ESTIMATED TOTAL PERSON TRIPS
BETWEEN MAJOR STATISTICAL AREAS
Click HERE for graphic.
DATA FILE
DOCUMENTATION
FRICTION FACTORS (INPUT)
ASCII files of friction factors by one-minute time interval and trip
type are created during gravity model calibration and subsequently
input to Tranplan's Gravity Model trip distribution function. Daily
friction factors are called "ff," peak-period friction factors are
called "pkff," and off-peak friction factors are called "opff." These
files are located under /max7/proj/sr8. The files have one record for
each one minute time interval between 1 and 90 minutes.
Columns Variable TypeDescription
1-2 A2 "GF" for friction factors
4-7 I4 Time Interval (Minutes)
9-9 I1 "1"
11-17 I7 Home-Work Friction Factors
18-24 I7 Home-College Friction Factors
25-31 I7 Home-School Friction Factors
32-38 I7 Home-Shop Friction Factors
39-45 I7 Home-Other Friction Factors
46-52 I7 Work-Other Friction Factors
53-59 I7 Other-Other Friction Factors
60-66 I7 Home-Serve Passenger Friction Factors
67-73 I7 Visitor Friction Factors
74-80 I7 Regional Airport Friction Factors
(Insert Trip Distribution Friction Factors - Home-Work)
(Insert Trip Distribution Friction Factors - Home-College)
(Insert Trip Distribution Friction Factors - Home-School)
(Insert Trip Distribution Friction Factors - Home-Shop)
(Insert Trip Distribution Friction Factors - Home-Other)
(Insert Trip Distribution Friction Factors - Work-Other)
(Insert Trip Distribution Friction Factors - Other-Other)
(Insert Trip Distribution Friction Factors - Serve Passenger)
(Insert Trip Distribution Friction Factors - Visitor)
(Insert Trip Distribution Friction Factors - Airport)
CHAPTER 10
VEHICLE TRIP FACTORING
VEHICLE TRIP FACTORING
First-stage transportation model applications use a person trip to
vehicle trip conversion program to obtain vehicle trip tables for
input to the highway assignment process. This process is intended to
account for the most important factors affecting vehicle use while
avoiding the need for transit network modeling and a mode choice
model.
Transit modeling considerably complicates and lengthens the modeling
process. Many model applications deal exclusively with evaluating
roadway improvements where holding transit patronage assumptions
constant across alternatives would not seriously compromise model
validity. Approximately 1.2% of all person trips are now made by
transit. The most ambitious planned transit improvements would
increase transit's mode share to 2% of all trips. Thus, small errors
in transit use would be well within the range of expected model
accuracy.
In addition to reducing the time and cost of SANDAG's in-house highway
model applications, use of a factoring process allows outside modelers
such as consultants and local jurisdictions to share the same
procedures and datasets as SANDAG and still run the models on PCs.
MODEL STRUCTURE
Person trip to vehicle trip factors are applied by two time periods,
three attraction location zones, five distance ranges, and ten trip
types. A set of peak-period home-work trip factors reflecting the
effects of proposed travel demand management (TDM) ordinances can also
be applied for future year analyses.
Time Periods
The two time periods considered are an off-peak period and a combined
morning and afternoon peak period. The peak period includes travel
during the hours of 6:00 to 9:00 in the morning and 3:00 to 6:00 in
the afternoon. The off-peak period covers travel in the remaining 18
hours of the day.
Higher automobile level of service during off-peak hours when highway
congestion is negligible leads to slightly higher rates of off-peak
vehicle use for most trip types. However, total off-peak vehicle use
across all trip types is slightly lower than peak-period vehicle use
since trip purposes with a low vehicle occupancy comprise a smaller
percentage of off-peak trips.
AVR Zones
In the course of developing TDM ordinances for the Region, three
"average vehicle ridership" (AVR) zones, shown at the end of Chapter
2, were identified. The potential for ridesharing and use of other
alternatives to the automobile is believed to differ between these
zones.
One AVR zone is Centre City San Diego. All other incorporated cities
make up the second AVR zone. Finally, the unincorporated areas of the
County fall into the third AVR zone. With good transit service, high
parking costs, and dense development, Centre City has the lowest
vehicle use of the three zones. Conversely, unincorporated areas have
the highest vehicle use rates. Vehicle use in the suburban AVR zone
generally lies between the other two zones.
Trip Length
Vehicle use tends to increase with trip length. The most obvious
factor is that walking and biking options fall off quickly as distance
increases. The use of other modes is more variable with trip length.
The following five distance categories were established to put 1/5 of
total person trips in each category.
þ 1.1 miles or less
þ 1.2 to 2.1 miles
þ 2.2 to 4.1 miles
þ 4.2 to 9.2 miles
þ 9.3 miles or more
Trip Types
A fairly strong relationship exists between trip type and vehicle use.
By definition "serve passenger" trips are all vehicle trips. School
trips have the lowest rate of vehicle use since most students are
below driving age. Family members often accompany drivers on home-
other, other-other, and home-shop trips so that these trip types also
tend to have lower vehicle use.
Fortran Program
A SANDAG Fortran program is run to convert person trips to vehicle
trips. The program processes each cell of the ten trip tables
produced by the trip distribution model. The factoring process uses
the following input files.
þ Person trip tables from trip distribution model
þ External trip tables from external trip model
þ Zone-to-zone distance files from highway network
þ Vehicle use factors from Travel Behavior Survey
þ Assignment factors from Travel Behavior Survey
The program splits daily person trips from the trip distribution model
into peak and off-peak period trips using the factors listed in the
following table. These factors are assumed to remain constant over
time.
Table 10-1
TIME OF DAY FACTORS
Trip Type Off-Peak Peak
Home-Work 35% 65%
Home-College 60% 40%
Home-School 38% 62%
Home-Shop 66% 34%
Home-Other 61% 39%
Work-Other 67% 33%
Other-Other 64% 36%
Serve Passenger 42% 58%
Tourist 73% 27%
Airport 69% 31%
Total 57% 43%
A look-up table determines the AVR zone in which the attraction zone
lies. The production to attraction zone distance is read from a zone-
to-zone distance table produced by the highway network model and used
to find a distance range category. The appropriate vehicle use factor
is applied to the person trips to compute vehicle trips.
Another function of the vehicle trip factoring process is to convert
production-attraction trip tables from the trip distribution model
into origin-destination trip tables for traffic assignment. Vehicle
trips are split into "major" and "minor" flow directions at this point
by applying the factors listed in Table 10-2. These factors are
tabulated from the Travel Behavior Survey and account for the
directional imbalances in traffic by time period.
Peak-period traffic is assigned in the morning peak-period direction.
Thus, peak-period major flow factors represent the fraction of trips
in the morning peak period that are traveling in the production-to-
attraction direction and trips in the afternoon peak-period traveling
in the attraction-to-production direction. Almost all work trips are
in the major flow direction, while home-other trips are more evenly
split.
Off-peak period traffic is assigned so that the attraction-to-
production direction is the major flow direction. Factors are
computed by summing up trips in the heaviest direction of travel
during each off-peak hour and dividing by total off-peak trips. As
one would expect, off-peak factors show a more even directional split
compared to peak-period factors.
Tranplan trip tables are in whole numbers. Before writing out a trip
table record, a "bucket" rounding technique is used to convert
fractional vehicle trips to whole numbers while maintaining overall
trip totals. Output trip tables retain a factor of ten from the trip
generation model.
Table 10-2
DIRECTIONAL VEHICLE TRIP FACTORS
Trip Type Peak Off-Peak
Home-Work 95% 75%
Home-College 82% 74%
Home-School 91% 73%
Home-Shop 69% 61%
Home-Other 67% 61%
Work-Other 33% 57%
Other-Other 42% 40%
Serve Passenger 70% 57%
Visitor 70% 56%
Airport 71% 55%
Total 67% 55%
Resulting origin-destination vehicle trips are accumulated in a zone-
to-zone matrix. When all internal trip tables have been analyzed,
external trip tables are read and added to internal trips.
MODEL OUTPUTS
The following outputs are created:
þ Off-peak period origin-destination vehicle trip table
þ Peak period origin-destination vehicle trip table
þ Reports
MODEL CALIBRATION
Travel Behavior Survey trip records contain the trip type, start time,
end time, mode of travel, production zone, and attraction zone of each
trip. Peak and off-peak factors were tabulated directly from the
Survey.
Trip distances were added to survey records by looking up the
production zone to attraction zone distance from the highway network
model. A look-up table was used to determine the AVR zone of the
attraction zone. Each survey record was added to a matrix of total
person trips by time period, AVR zone, distance range, and trip type.
If the trip was made in an automobile, truck, taxicab, or on a
motorcycle and the person making the trip was the driver of the
vehicle, the trip was also added to a vehicle trip matrix by time
period, AVR zone, distance range, and trip type.
The resulting vehicle trip matrix was divided by the person trip
matrix to obtain vehicle trip usage factors. When fewer than ten
survey trip records occurred in a cell of the person trip matrix,
average vehicle use rates were substituted.
Peak period home-work vehicle use factors were adjusted to reflect the
TDM ordinance targets of increasing average vehicle ridership by 25%
over baseline rates at firms with 100 or more employees. About 1/2 of
all employees work at large firms so an overall 12.5% increase in AVR
rates is assumed.
DATA FILE
DOCUMENTATION
VEHICLE TRIP FACTORS (INPUT)
ASCII files of person trip to vehicle trip factors are applied during
the vehicle trip factoring process. One file called "ptvt.pct,"
located under /max7/proj/sr8 contains vehicle factors for non-TDM
applications. Another file called "ptvt.tdmpct" has modified peak
period home-work factors that reflect the effects of a TDM ordinance.
The files have one record for each combination of time period, trip
type, distance range, and AVR zone.
Columns Variable Type Description
1-2 I2 Time Period, Where:
1 = Off-Peak Period
2 = Peak Period
3-4 I2 Trip Type, Where:
1 = Home-Work
2 = Home-College
3 = Home-School
4 = Home-Shop Friction
5 = Home-Other Friction
6 = Work-Other Friction
7 = Other-Other Friction
8 = Home-Serve Passenger
9 = Visitor
10 = Regional Airport
5-6 I2 Distance Range, Where:
1 = 1.1 Miles or Less
2 = 1.2 to 2.0 Miles
3 = 2.1 to 4.2 Miles
4 = 4.3 to 9.4 Miles
5 = 9.5 Miles or More
7-8 I2 AVR Zone, Where:
1 = Centre City
2 = Suburban
3 = Rural, Unincorporated
9-18 F10.4 Vehicle/Person Trip Ratio
CHAPTER 11
MODE CHOICE
MODE CHOICE
The mode choice model splits total person trips from the trip
distribution model into trips by individual forms of transportation
called modes. Unlike the vehicle trip factoring process described in
the previous chapter, the mode choice model is designed to link mode
use to demographic assumptions, highway network conditions, transit
system configuration, parking costs, transit fares, and auto operating
costs. Estimating transit patronage for light rail extensions and new
bus routes is one of the primary functions of the mode choice model.
The model assumes that travelers make logical and systematic decisions
about which form of transportation to take based upon the time and
cost of completing a trip by alternative modes. The model is
sensitive to a wide range of facility improvements and policies,
however, the model is insensitive to programs designed to alter mode
use without altering times or costs, such as:
þ Advertising campaigns to increase the use of transit, bicycling,
or ridesharing
þ Rideshare matching programs
þ Construction of bicycle lanes
þ Replacing older buses to increase the attractiveness of transit
A common misconception about the mode choice model is that it
underestimates future transit use for alternatives with expanded
transit service because model calibration is based upon current
conditions. This should not be the case. The model estimates transit
use for each zone-to-zone movement based upon the quality of transit
service relative to other modes. Existing light rail corridors
provide a basis for determining potential transit use with high
quality transit service. As more light rail lines are built and bus
service is expanded, the model recognizes the resulting transit
service improvements and shifts travel to transit from other modes.
MODEL STRUCTURE
The model considers six modes of transportation:
þ Drive alone or one person auto
þ Two person auto
þ Three or more person auto
þ Transit-walk
þ Transit-auto
þ Other
Three auto modes differentiate auto trips by number of occupants,
allowing ridesharing impacts to be evaluated. The term "auto" is used
generically to include travel by automobiles, pick-ups, vans, trucks,
taxicabs, and motorcycles.
Two public transit modes separate out transit trips by access mode.
Transit-auto trips are those trips where the rider gets to or from a
transit stop by either driving, carpooling, or being dropped off from
an automobile. Riders who walk to and from transit stops are
classified as transit-walk trips. While auto access trips make up a
very small percentage of overall trips (0.1%), they are an important
transit sub-market and thus warrant special consideration.
Trips by walking, bicycling, and school bus are grouped into an
"other" mode. These trips are included in the modeling process to
provide a complete accounting of trip making. However, estimation
procedures do not allow policy issues for these trips to be addressed.
The model computes mode use separately for two time periods, three
income levels, and eight trip types. The two time periods split
travel into peak and off-peak hours. The peak period extends from
6:00 to 9:00 in the morning and 3:00 to 6:00 in the afternoon; the
off-peak period covers the remaining 18 hours of the day. It is
important to evaluate mode use separately for the two time periods
because the quality of service can vary dramatically by mode. For
example, transit operators often provide more frequent transit service
during peak hours, reducing wait times for transit riders.
Conversely, highway congestion is at its worst during peak hours
making auto modes less attractive relative to transit.
Mode use also varies by income level. People in low-income households
tend to own fewer automobiles and hence make more trips by transit and
carpooling. People in higher-income households tend to value the
convenience and flexibility offered by the automobile and have a lower
use of alternative modes. Households are split into three income
categories designed to put 1/3 of total transit trips in each
category. The categories are based upon annual household income in
1990 dollars and have the following breakpoints:
þ Under $10,000
þ $10,000 to $24,999
þ $25,000 or more
There is also a strong relationship between mode use and trip type.
For example, most students are below driving age so school trips
generate almost no drive alone trips but have a very high rate of
transit and other mode use. Home-other trips tend to be made with
other household members, so that two and three or more person auto
modes are more heavily used than other trip types. In order to reduce
computer processing time, visitor and airport trips are combined with
other-other trips for mode share computations. There are relatively
few visitor and airport trips, and mode use percentages are similar to
other-other trips.
SANDAG's mode choice model is a multi-modal logit model in a form that
has been extensively used throughout the country. Model parameters
have been adjusted to replicate San Diego travel conditions.
The mode choice model makes use of a wide variety of data in order to
estimate mode use. The following datasets are input to the model:
þ Peak and off-peak person trip tables from gravity model
þ Peak and off-peak mixed-flow highway travel time matrices
þ Off-peak period highway distance matrix
þ Peak period HOV travel time matrix
þ Peak and off-peak transit travel time matrices
þ Peak and off-peak transit fare matrices
þ Peak and off-peak transit access/egress station matrices
þ Transit walk access file
þ Transit auto access file for home-based productions
þ Distribution of home-based productions by income level
þ Parking costs by zone
þ Auto terminal times by zone
A SANDAG Fortran program performs mode choice computations.
Accounting for the relationship between transit access points (TAPs)
and zones adds complexity to the mode choice model. Transit TAP-to-
TAP impedances are stored in an internal matrix. The main body of the
program processes each production zone with trips. Zone-related files
are read for the production zone, along with person trips and auto
impedances from the production zone to all other attraction zones.
The program then cycles through each attraction zone with trips.
Impedances are computed for auto and other modes. Finally, trips are
allocated to transit access areas at the production and attraction
zone. For each production-attraction transit access area pair, the
program adds in transit access times to TAPs; finds the TAP-to-TAP
interchange that provides the lowest overall impedance; and computes
mode shares. TAP-to-TAP transit trips for transit assignment are
accumulated in an internal matrix and zone-to-zone auto trips are
written out for highway assignment. A more detailed description of
this process follows.
Transit Impedances
The program first reads in transit travel time and fare matrices,
computes network-related transit impedances, and stores the impedances
in an internal matrix for later use in the mode choice program. The
following equation is used to compute transit impedances:
TI(P,I,J) = (SFW(I,J)+LFW(I,J)/2)*FWC(P) + (XT(I,J)+(NX(I,J)*3-
TT(I,J)))*XTC(P)
+ IV(I,J)*IVC(P) + (XW(I,J)*XWC(P) + FR(I,J)*FRC(P)
where:
TI = Transit impedance between TAPs "i" and "j"
P = Purpose (commute, non-commute)
SFW = Short first wait time between TAPs "i" and "j" (minutes)
LFW = Long first wait time between TAPs "i" and "j" (minutes)
FWC = First wait time coefficient
XT = Transfer wait time between TAPs "i" and "j" (minutes)
NX = Number of transfers between TAPs "i" and "j"
TT = Number of timed transfer points between TAPs "i" and "j"
XTC = Transfer Wait time coefficient
IV = In-vehicle time between TAPs "i" and "j" (minutes)
IVC = In-vehicle time coefficient
XW = Transfer walk time between TAPs "i" and "j" (minutes)
XWC = Transfer walk time coefficient
FR = Transit fare between TAPs "i" and "j" (cents)
FRC = Transit fare coefficient
Transit TAP-to-TAP travel times and fares are obtained from the
Tranplan transit network path building and selected summation
functions. Peak and off-peak period travel times are input,
reflecting the transit service and bus speed variations by time
period.
Transit travel times are broken down into four components so that
different coefficients can be applied to each, as shown in the Table
11-1. Research has shown that transit users perceive wait and walk
time as being more onerous than in-vehicle time. Thus, initial wait
time and walk time coefficients are double in-vehicle coefficients.
Transfer wait time coefficients are 2.5 times higher than in-vehicle
coefficients. The commute trip transit fare coefficient equates cost
and time components at a rate of 4.5 cents per minute, based upon 25%
of the Region's average wage rate of $11.00 per hour.
Coefficients differ between commute trips (home-work and home-college
trips) and non-commute trips. Non-commute coefficients are lower
across the board than commute coefficients. This makes non-commute
mode shares less sensitive to level of service changes than commute
trip mode shares. The non-commute fare coefficient also reflects a
value of time that is 40% of commute trips.
Several adjustments are made to transit times produced by the Transit
Selected Summation function. Initial wait times longer than 10
minutes are weighted half as much as the first 10 minutes of initial
wait time. This is intended to reflect the inconvenience of long
waits without overstating their importance. Transfer times are also
adjusted. During model calibration, it was found that the model over-
estimated transfer rates. In order to bring model-estimated transfers
into better agreement with observed data, a three minute transfer
penalty is added to network-based transfer times. This penalty is
reduced to two minutes where transfers occur at locations with timed-
transfers.
Transit access time is another transit impedance component. These
times are added later in the mode choice process.
Table 11-1
MODE CHOICE COEFFICIENTS
Mode Component Commute Non-Commute
Drive Alone In-Vehicle Time -0.0250 -0.0100
Cost -0.0055 -0.0055
Terminal Time -0.0500 -0.0200
2 Person Auto In-Vehicle Time -0.0250 -0.0100
Cost -0.0028 -0.0028
Terminal Time -0.0500 -0.0200
3 Person Auto In-Vehicle Time -0.0250 -0.0100
Cost -0.0018 -0.0018
Terminal Time -0.0500 -0.0200
Transit Initial Wait Time -0.0500 -0.0200
Transfer Time -0.0625 -0.0250
In-Vehicle Time -0.0250 -0.0100
Walk Time -0.0500 -0.0200
Fare -0.0055 -0.0055
Auto Access Time -0.1250 -0.0500
Other Distance -1.0000 -0.4000
Auto Impedances
After computing and storing transit impedances, the model processes
each production-attraction zone pair. If there are no trips between a
zone pair, the interchange is skipped. Otherwise, auto impedances are
computed using the following equation:
AI(M,P) = T(I,J)*TC(P) + (D(I,J)*C+P(J))*CC(M,P) + W(J)*WC(P)
where:
AI = Auto impedance
P = Purpose (Commute,non-commute)
M = Mode (Drive alone, 2 person auto, 3+ person auto)
T = Travel time between zone "i" and zone "j" (minutes)
TC = Travel time coefficient
D = Distance between zone "i" and zone "j" (miles)
C = Auto cost (15 cents per mile)
P = Parking cost at zone "j" (cents per occupant)
CC = Cost coefficient
W = Walk time at zone "j" (minutes)
WC = Walk coefficient
Auto impedance equations apply coefficients that vary by purpose to
the highway time between a zone pair. Highway times are obtained from
a post-assignment highway network and include the effects of modeled
congestion on travel times. Different travel times are used for peak
and off-peak mode splits that reflect the reduced congestion levels in
off-peak hours. The same times are used for all auto modes except
peak period high occupancy commute trips (2 and 3+ auto modes) where
high occupancy vehicle (HOV) travel times are used. HOV times are
lower than mixed-flow times where HOV lanes or HOV by-pass ramps
exist.
Auto Costs
Auto impedances also include auto costs multiplied by cost
coefficients. Costs are computed by applying an auto operating cost
of 15 cents per mile to zone-to-zone highway distances and adding in
parking charges. Cost coefficients for the two person auto mode are
divided by two and coefficients for the 3+ auto mode are divided by
three to account for each person's share of total costs.
There is some debate as to what auto operating costs are appropriate
for including in a mode choice model. At a minimum, gasoline costs
should be considered. Caltrans estimates that 1990 average vehicle
fuel efficiency was 14.6 miles per gallon and average gasoline costs
were $1.10 per gallon. This works out to a gasoline cost of 7 cents
per mile.
A maximum operating cost would add in vehicle depreciation, insurance,
and maintenance costs, bringing the cost to 33 cents per mile for an
intermediate automobile as estimated by The Federal Highway
Administration. However, many of these non-gasoline costs are fixed
costs that are unlikely to enter into a mode choice decision. Two-
thirds of these costs are assumed to be fixed and one-third variable,
resulting in the 15 cent per mile operating cost assumption. There
appears to be no consensus in the literature as to where auto
operating costs are headed in the future. Auto operating costs are
kept fixed over time.
Parking charges are another cost component. Most trips go to areas
with free parking. The map on the next page highlights areas where
parking costs are assumed in the model.
(Insert Map/Figure 11-1 - Parking Costs)
Some pay parking exists outside of Centre City at universities and
suburban core areas. SANDAG lacks a method of monitoring these rates
and a $1.00 parking charge is assumed.
Centre City San Diego is the area with the highest parking charges in
the Region. SANDAG conducted surveys of posted Centre City parking
rates in 1981, 1984, and 1988. These surveys are used to establish
baseline parking rate assumptions and trends for increasing parking
rates over time.
The most recent survey in 1988 found average posted parking rates in
the Centre City core area to be $7.90 per day, and $3.90 per day in
fringe areas of Centre City. Some motorists receive parking subsidies
or park by the month to get rate reductions. The extent of these
discounts is not known. A 25% reduction is assumed, bringing down the
core area rate to $6.00 per day and the fringe area rate to $3.00 per
day. The fringe area is split into $4.50 and $3.00 areas for use in
the models to better reflect different levels of activity within the
fringe area. Future year parking rates are tied to Centre City
employment increases. A 40% increase in parking rates is expected by
2015.
The parking costs discussed are applied to commute trips (home-work
and home-college) that tend to park for long periods of time. A 50
cent parking charge is assumed for non-commute trips in pay parking
areas, based upon a meter rate of $1.00 per hour and an average
parking duration of 1/2 hour. Commute and non-commute parking costs
are divided by two within the mode choice model to apportion round-
trip costs to each leg of the trip.
Auto Terminal Time
Auto terminal time makes up the final element in auto impedance
calculations. Terminal time accounts for the walk time from a parking
space to ultimate destination and other miscellaneous out-of-vehicle
time. A one minute terminal time is assumed for most auto trips.
Trips to areas with large on-site parking lots or areas with
significant off-site parking are assumed to incur an additional walk
time penalty. The following map shows areas where higher-than-normal
walk times are assumed.
Establishing walk times involves considerable judgement since survey
data is sparse. A SANDAG Centre City Parking Study conducted in 1978
indicated a core area downtown average walk distance of 870 feet,
which translates into a walk time of 3 minutes, assuming a 3 MPH walk
speed. Walk times for other areas are scaled down from this estimate.
Existing terminal times are held constant for future year forecasts.
(Insert Map/Figure 11-2 - Walking Time)
Other Impedances
Impedance calculations for the "other" mode are less complicated, as
indicated by the following equation:
OI(P) = D(I,J)*DC(M,P)
where:
OI = Other impedance
P = Purpose (Commute, non-commute)
D = Distance between zone "i" and zone "j" (miles)
DC = Distance coefficient
No cost component is included since the costs of bicycling and walking
are essentially zero. The other mode includes a number of different
modes, so impedances are based upon distance, not time.
Mode Share Computations
The next part of the program computes mode shares. Walk and auto
transit access points are evaluated at production and attraction ends.
Mode shares are computed for each combination of production and
attraction transit access areas for zones that are at least partially
accessible to transit by walking at the attraction end. Other
interchanges not accessible to transit use simplified whole-zone
procedures involving only auto and other modes.
Transit-Auto Impedances
For transit accessible zones, the program cycles through each
attraction end transit access area. Transit-auto impedances are
computed by looking up the stored transit impedance for the TAP-to-TAP
interchange being evaluated; adding auto access time at the production
zone; and adding walk time at the attraction end. The following
equation is applied:
TAI(P) = TI(P,I,J) + AT(IZ,I)*ATC(P) + WT(J,JZ)*WTC(P)
where:
TAI = Transit-auto impedance
P = Purpose (commute, non-commute)
TI = Transit impedance between TAPs "i" and "j"
AT = Auto access time from zone "iz" to TAP "i" (minutes)
ATC = Auto access time coefficient
WT = Walk time from TAP "j" to zone "jz" (minutes)
WTC = Transfer walk time coefficient
Impedances are calculated for each park-and-ride lot connection at the
production end and TAP at the attraction end transit access area. The
connection offering the lowest impedance is selected.
Transit-Walk Impedances
The program next cycles through each transit access area at the
production end. Transit-walk impedances are calculated by applying
the following equation:
TWI(P) = TI(P,I,J) + WT(IZ,I)*WTC(P) + WT(J,JZ)*WTC(P)
where:
TWI = Transit-walk impedance for purpose "P"
P = Purpose (commute, non-commute)
TI = Transit impedance between TAPs "i" and "j"
WT = Walk time from zone "iz" to TAP "i" (minutes)
WT = Walk time from TAP "j" to zone "jz" (minutes)
WTC = Transfer walk time coefficient
Impedances are calculated for each TAP at the production end transit
access area and TAP at the attraction end transit access area. The
TAP combination with the lowest impedance is selected.
Trip Factoring
The program then cycles through the eight trip types and three income
levels used in the mode choice model. Trips by purpose from the trip
distribution model are apportioned to income level and production-
attraction transit access area combination by applying the following
equation:
PT = T(IZ,JZ)*100*IF(IL,IZ)*PF(IZ,I)*AF(JZ,J)
where:
PT = Partial person trips
T = Person trips between zones "iz" and "jz"
IF = Fraction of productions in income level "il" from zone "iz"
PF = Fraction of productions from zone "iz" in transit access area
"i"
AF = Fraction of attractions to zone "jz" in transit access area "j"
The trip generation model multiplies estimated trip ends by ten to
reduce rounding errors during trip distribution. An additional factor
of 100 is applied during mode share computations to further reduce
rounding problems. It should be noted that income data is not
available within the model for non-home-based trips because the
production zone for these trips is away from the home zone. Thus,
work-other and other-other are not split into income levels.
Impedance Transformation
Prior to computing mode shares, impedances for each mode are
transformed by adding four sets of constants and then exponentiating
the resulting sum. The following equation is used:
XI(M) =
EXP(II(M) + MC(TT,M) + IC(P,IL,M) + LC(P,DR,M) + TDM(M,AVR(JZ)))
where:
XI = Transformed impedance for mode "m"
M = Mode
EXP = Exponential function
II = Initial impedance for mode "m"
MC = Mode constant for mode "m" and trip type "tt"
TT = Trip Type
IC = Income constant for income level "il" and purpose "p"
P = Purpose
IL = Income level
LC = Trip length constant for distance range "dr" and purpose "p"
DR = Distance Range
TDM = Drive alone TDM constant for certain applications
AVR = AVR zone at attraction zone "jz"
The first set of mode constants, listed in Table 11-2, is used to
obtain the correct mode shares by trip type. These constants account
for factors affecting mode use other than time and cost. For example,
park-and-ride is primarily geared toward commuters. Thus, the
transit-auto mode has large negative constants for non-work trips,
which means that regardless of the level of transit-auto service there
will be relatively few non-work transit-auto trips.
The second set of income constants, listed in Table 11-3, reflects the
overall propensity of trips in a particular income level and purpose
to use a mode. For example, survey data shows that higher income
households are somewhat less likely to use transit, regardless of the
quality of service. Thus, the transit-walk mode has larger negative
constants as income increases. Transit-auto use is less income
dependant, leading to constants that are more variable by income
level.
Table 11-2
MODE CONSTANTS
---------------------------- Mode ------------------------------
Trip Drive 2 Person 3 Person Transit Transit
Type Alone Auto Auto Walk Auto Other
Home-Work -0.00 -3.08 -3.69 -0.31 -2.49 -1.05
Home-College -0.00 -2.89 -4.63 -1.28 -4.05 -0.98
Home-School -1.00 -1.76 -1.04 -0.00 -3.25 +1.07
Home-Shop -0.00 -2.25 -2.55 -2.66 -5.86 -2.54
Home-Other -0.00 -1.92 -1.86 -2.27 -5.30 -2.19
Work-Other -0.00 -1.56 -2.72 -4.19 -6.48 -1.20
Other-Other -0.00 -0.32 -0.65 -4.03 -6.60 -1.82
Table 11-3
INCOME CONSTANTS
-------------------------------- Mode ---------------------------
Drive 2 Person 3 Person Transit Transit
Purpose/Income Alone Auto Auto Walk Auto Other
Commute
$0-$9,999 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00
$10,000-$24,999-0.00 +1.17 +1.11 -0.51 -0.16 +0.48
$25,000+ -0.00 +1.23 +0.51 -1.42 -0.18 -0.35
Non-Commute
$0-$9,999 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00
$10,000-$24,999-0.00 +2.37 +2.36 +0.34 +0.28 +2.01
$25,000+ -0.00 +2.01 +1.87 -1.07 -0.70 +1.04
A third set of trip length constants, listed in Table 11-4, ensures
that the model replicates observed trip lengths by mode and purpose.
The three distance range categories put 1/3 off all person trips into
each category. The model tends to over-estimate short transit trips.
The positive transit-walk constants in the over-two-mile distance
ranges offset this tendency.
Table 11-4
TRIP LENGTH CONSTANTS
Drive 2 Person 3 Person Transit Transit
Purpose/Income Alone Auto Auto Walk Auto Other
Commute
0.0-1.9 Miles -0.00 -0.00 -0.00 -0.00 -0.00 -0.00
2.0-8.9 Miles -0.00 -0.35 -0.82 +0.77 +1.46 +0.35
9.0+ Miles -0.00 -1.42 -1.79 +0.51 +1.67 +1.89
Non-Commute
0.0-1.9 Miles -0.00 -0.00 -0.00 -0.00 -0.00 -0.00
2.0-8.9 Miles -0.00 -0.48 -0.71 +0.72 +1.81 -1.05
9.0+ Miles -0.00 -1.26 -1.66 +0.34 +1.80 +0.33
Travel Demand Management (TDM) regulations are represented in the mode
choice model by adding a TDM constant to peak period drive alone mode
impedances. The constant reduces overall drive alone mode use so that
AVR targets are met. No constant is needed for Centre City trips
where AVR targets are met through expected transit service
improvements and parking cost increases. TDM constants of -0.75 for
suburban trips and -1.00 for rural trips are needed to meet AVR
targets.
Proposed Travel Demand Management (TDM) regulations receive special
treatment in mode choice calculations because of their potential for
significantly altering mode use. California and Federal Clean Air
Acts both mandate that programs be enacted to reduce solo-driver
commute trips. It is SANDAG's policy to include TDM effects in all
transportation model forecasts for years beyond 1996, when TDM
ordinances are targeted for full operation.
Locally, work is well underway toward implementing TDM measures.
SANDAG has prepared a model TDM ordinance for adoption by local
jurisdictions. The San Diego County Air Pollution Control District
has a matching regulation awaiting adoption by its Board. Two cities
(San Diego and San Marcos) have early versions of TDM programs already
in place.
These measures are designed to require employers with 100 or more
employees to file plans that show how they will achieve a 25% increase
in their peak period average vehicle ridership (AVR) rate over
baseline conditions. The AVR rate is computed as the number of
employees divided by the number of employees driving to work.
Substantial fines of up to $1000 per day would be imposed upon
employers disregarding TDM requirements.
Representing TDM effects through use of an overall adjustment reflects
TDM ordinances as they are now written. Specific actions to achieve
AVR targets are not prescribed. Instead, employers are required to
come up with their own methods of increasing AVR rates. Procedures
for modeling TDM ordinances will be fine-tuned as more experience is
gained with implementation.
Transit MSA Adjustment Factors
During model calibration, it was found that estimated transit trips
between Major Statistical Areas (MSAs) differed from observed trips
for some MSA pairs. A set of MSA adjustment factors listed in Table
11-5 was developed to compensate for these errors. Adjustment factors
are held constant for all model applications. An additional MSA is
available to ensure that the model accurately estimates transit
ridership for special transit studies. A Mission Valley East area has
been defined to assist with a Mission Valley East Alternatives
Analysis study that is currently underway. The factors are applied to
transit impedances using the following equation:
AXI(M) = XI(M) * AFAC(M,P,IMSA(IZ),JMSA(JZ))
where:
AXI = Adjusted transit impedance
M = Mode (Transit-walk or transit-auto)
XI = Transformed transit impedance
AFAC = MSA adjustment factor
P = Purpose
IMSA = MSA at production zone "iz"
JMSA = MSA at attraction zone "jz"
Trips by Mode
Finally, the program splits partial trips to individual modes in
direct proportion to the mode's transformed impedance relative to the
sum of impedances for all modes as follows:
MT(M) = PT * XI(M) / (XI(1) + XI(2) ... + XI(6))
where:
PT = Partial person trips
MT = Person trips by mode
M = Mode
XI = Transformed impedance
Table 11-5
TRANSIT COMMUTE MAJOR STATISTICAL AREA ADJUSTMENT FACTORS
Click HERE for graphic.
Table 11-6
TRANSIT NON-COMMUTE MAJOR STATISTICAL AREA ADJUSTMENT FACTORS
Click HERE for graphic.
Vehicle Trip Factoring
Resulting trips by mode are accumulated for assignment and reports.
Once all trip purposes and income levels have been processed for a
production-attraction zone pair, vehicle trips for highway assignment
are computed. Two person auto trips are converted to vehicle trips by
dividing by two. Three or more person auto trips use the average
vehicle occupancies listed in the table below to obtain vehicle trips
from person trips.
Table 11-7
VEHICLE OCCUPANCIES FOR 3+ PERSON AUTOS
Trip Type Vehicle Occupancy
Home-Work 3.41
Home-College 3.41
Home-School 3.66
Home-Shop 3.39
Home-Other 3.66
Work-Other 3.49
Other-Other 3.50
Serve Passenger 3.66
Visitor 3.66
Airport 3.66
Production-attraction vehicle trips are split into "major" and "minor"
flow directions at this point by applying the same procedures
described in Chapter 10. Tranplan trip tables are in whole numbers.
Before writing out a trip table record, a "bucket" rounding technique
is used to convert fractional vehicle trips to whole numbers while
maintaining overall trip totals by distance range. The factor of 100
introduced in the mode split process is also removed, leaving output
trips factored by ten.
Transit Trips
Transit trips by walk and auto are accumulated in an internal TAP-to-
TAP matrix after each zone pair has been evaluated. A two-purpose
transit trip table is written out after all zone pairs have been
processed.
MODEL OUTPUTS
The mode choice model produces reports, Tranplan trip tables, and
files of trip end data by mode. The reports summarize trips by mode,
trip type, time period, and AVR zone of the attraction zone. Summary
statistics such as transit mode percentages and auto occupancies are
also reported by trip type, time period, and AVR zone. The model
generates the following Tranplan trip tables to be used later in
highway and transit assignment functions:
þ Off-peak period major flow vehicle trips (total or mixed-flow and
HOV)
þ Off-peak period minor flow vehicle trips (total or mixed-flow and
HOV)
þ Peak period major flow vehicle trips (total or mixed-flow and HOV)
þ Peak period minor flow vehicle trips (total or mixed-flow and HOV)
þ Off-peak period production-attraction walk access transit trips
þ Peak period production-attraction walk access transit trips
þ Off-peak period production-attraction auto access transit trips
þ Peak period production-attraction auto access transit trips
A file of vehicle trip ends by trip type, zone, and
production/attraction end is produced for later use in air quality
emissions modeling. A file of total transit trips by zone is used in
factoring external transit trips.
Table 11-8 summarizes mode choice results for 1990 base-year
conditions; and 2015 with the RTP recommended transit network and TDM
ordinance in place. Drive alone trips increase at a slower rate than
other modes due to the effects of the TDM ordinance. The number of
transit trips is highly dependent upon transit service assumptions.
The extensive light rail expansion recommended in the RTP leads to
transit trip increases that out-pace other modes. Drive alone trips
increase at the same rate as other trips and the increase in transit-
walk trips drops from 79% to 57% when the model is run without TDM
assumptions.
Tables 11-9 and 11-10 show how commute and total transit trips change
by Major Statistical Area. The North City MSA is the focus of many
recommended transit service improvements. When coupled with strong
population and employment growth, North City shows the highest
increase in transit trips. The Central Area that is largely built out
with good base-year transit service is expected to experience the
slowest rate of growth in transit trips.
Table 11-8
TRIPS BY MODE
Click HERE for graphic.
MODEL CALIBRATION
SANDAG's mode choice model is calibrated by borrowing coefficients
from mode choice models in other urban areas and then adjusting
constants until model estimates agree with observed values. As part
of a contract with the consulting firm of Parsons, Brinkerhoff, Quade
and Douglas (PBQD), mode choice models in use across the country were
reviewed. A set of PBQD recommended coefficients based upon that
review was the starting point for model calibration.
If the equations that transform impedances and compute mode shares are
considered, it can be seen that decreasing the constant for a mode
decreases the transformed impedance for the mode and lowers its mode
share. Thus, constants can be increased or decreased to bring
estimated mode shares into agreement with observed mode shares.
Individual coefficients for a mode can also be adjusted to affect
model estimates within a mode. For example:
þ Increasing transit transfer coefficients decreases the transit
transfer rate
þ Increasing auto cost coefficients reduces drive alone trips to
high parking cost areas
þ Increasing transit fare coefficients makes transit trips more fare
sensitive
Several iterations through a special version of the mode choice
program make use of these relationships to calibrate the model. In
the first iteration, all constants except home-school drive alone
constants are set to zero, all MSA adjustment factors are set to one,
and coefficients recommended by PBQD are input. Initial work with
model calibration found that a coefficient of -1.00 for the home-
school drive alone mode reduces calibration time due to the limited
number of drive alone trips.
Table 11-9
COMMUTE TRANSIT TRIPS BETWEEN MAJOR STATISTICAL AREAS
Click HERE for graphic.
NOTE: ++% = Exceeds 999%
Table 11-10
TOTAL TRANSIT TRIPS BETWEEN MAJOR STATISTICAL AREAS
Click HERE for graphic.
NOTE: ++% = Exceeds 999%
During an iteration of mode choice calibration, the procedures
described above are applied. Instead of producing trip tables, trips
are totaled so that model-estimated trips can be compared with
observed values. The following trip totals are accumulated:
þ Person trips by mode and trip type
þ Person trips by mode, income level, and purpose (commute or non-
commute)
þ Person trips by mode, trip length category, and purpose
þ Transit trips by access mode, purpose, production MSA, and
attraction MSA
At the end of an iteration mode, shares by trip type are computed.
Mode constants for modes other than drive alone are updated using the
equations listed below. Table 11-11 summarizes observed trips by mode
from the Transit Ridership Survey and Travel Behavior Survey that are
the basis for mode share calibration. The table also compares
observed trips with final model-estimated trips. As indicated,
calibration procedures produce close agreement between observed and
estimated trips by mode.
IF OMS(TT,M) > EMS(TT,M), Then:
NMC(TT,M) = OMC(TT,M) - (OMS(TT,M) - EMS(TT,M)) / OMS(TT,M)
Else:
NMC(TT,M) = OMC(TT,M) + (EMS(TT,M) - OMS(TT,M)) / OMS(TT,M)
where:
NMC = New mode constant for mode "m"
TT = Trip type
M = Mode
OMC = Old mode constant
OMS = Observed mode share
EMS = Model-estimated mode share
The fraction of trips by income level for each mode and purpose is
also computed; compared with survey data; and used to update income
constants in the same manner that mode constants are updated.
Similarly, trip length constants are updated. Tables 11-12 and 11-13
summarize observed income and trip length distributions used to
calibrate the model. The tables also compare observed income and trip
length distributions with model estimates after calibration is
complete.
Finally, new transit MSA adjustment factors are computed at the end of
each iteration using the equation listed below, until the last two
iterations are reached. Adjustment factors remain unchanged during
the last two iterations so that the mode constants can be updated to
arrive at the correct number of trips by mode. Adjustment factors in
cells with fewer than 25 expanded survey transit trips stay at 1.0 and
are not updated. Adjustment factors are prevented from exceeding 2.0
or going below 0.5. Limiting the range of adjustment factors prevents
over-compensation when applying the model in a forecasting mode.
NFAC(M,P,IMSA,JMSA) = OFAC(M,P,IMSA,JMSA) *
OMSA(M,P,IMSA,JMSA) / EMSA(M,P,IMSA,JMSA)
where:
NFAC = New MSA adjustment factor
M = Mode (Transit-walk or transit-auto)
XI = Transformed transit impedance
P = Purpose (commute, non-commute)
IMSA = MSA at production zone
JMSA = MSA at attraction zone
OFAC = Old MSA adjustment factor
OMSA = Observed fraction of transit trips
EMSA = Estimated fraction of transit trips
Tables 11-14 and 11-15 summarize observed transit trips by MSA used to
calibrate the model. The tables also compare observed trips by MSA
with model estimates after calibration is complete. The model
accurately predicts most interchanges with a significant number of
transit trips. Trips produced in Centre City are somewhat under-
estimated based upon the Transit Ridership Survey. However, this
probably indicates survey error as opposed to model error because
transit riders who transferred in Centre City sometimes mistakenly
indicated a Centre City origin or destination.
Table 11-11
OBSERVED AND ESTIMATED TRIPS BY MODE
Click HERE for graphic.
Table 11-12
OBSERVED AND ESTIMATED INCOME DISTRIBUTION BY MODE
Click HERE for graphic.
Table 11-13
OBSERVED AND ESTIMATED TRIP LENGTH DISTRIBUTION BY MODE
Click HERE for graphic.
Table 11-14
OBSERVED AND ESTIMATED COMMUTE TRANSIT TRIPS BETWEEN MAJOR STATISTICAL
AREAS
Click HERE for graphic.
NOTE: ++% = Exceeds 999%
Table 11-15
OBSERVED AND ESTIMATED TOTAL TRANSIT TRIPS BETWEEN MAJOR STATISTICAL
AREAS
Click HERE for graphic.
NOTE: ++% = Exceeds 999%
DATA FILE
DOCUMENTATION
ZONE TO AVR ZONE CONVERSION FILE (INPUT)
An ASCII file listing the average vehicle ridership zone that each
transportation zone lies in is called "zone.msavr" and is located
under /max7/proj/sr8. The file has one record for each zone with the
following format.
Columns Variable TypeDescription
1-6 I6 Zone
7-12 I6 Average Vehicle Ridership Zone
SUMMARY AREA ZONES CONVERSION FILE (INPUT)
An optional ASCII file lists transportation zones that make up a
special summary area for mode choice reports. Files exist under
temporary workspaces created to evaluate alternatives. If the file is
omitted, mode choice results will only be reported for standard areas.
The first record in the file contains a summary area name that is used
in mode choice reports. The rest of the file has one record for each
zone in the summary area with the following format.
Columns Variable TypeDescription
1-5 I5 Summary Area Zone
ZONAL MODE CHOICE DATA (INPUT)
An ASCII file of zone level data needed for mode choice calculations
is called "msdata" and is located under /max7/proj/sr8. The file has
one record for each zone with the following format.
Columns Variable TypeDescription
1-4 I4 Zone
5-6 I2 Attraction End Walk Terminal Time (Minutes)
7-8 I2 Parking Cost Code, where:
1 =
Free Parking
2 =
Average $1.00 All-day Parking Rate
3 =
Average $3.00 All-day Parking Rate
4 =
Average $4.50 All-day Parking Rate
5 =
Average $6.00 All-day Parking Rate
OBSERVED TRANSIT TRIPS BY MAJOR STATISTICAL AREA (INPUT)
An ASCII file containing observed transit trips between Major
Statistical Areas is called "ms.obsmsa" and is located under
/max7/proj/sr8. The file has one record for each mode, purpose,
production MSA, and attraction MSA combination.
Columns Variable Type Description
1-2 I2 Mode, where:
1 = Transit-Walk
2 = Transit-Auto
3 = Total
3-4 I2 Purpose, where:
1 = Commute
2 = Non-Commute
3 = Total
5-6 I2 Production MSA
7-8 I2 Attraction MSA
9-18 I10 Observed Transit Trips
TRANSIT MAJOR STATISTICAL AREA ADJUSTMENT FACTORS (INPUT)
An ASCII file containing observed transit Major Statistical Areas
adjustment factors is called "ms.xmsa" and is located under
/max7/proj/sr8. The file has one record for each mode, purpose,
production MSA, and attraction MSA combination.
Columns Variable Type Description
1-2 I2 Mode, where:
1 = Transit-Walk
2 = Transit-Auto
3 = Total
3-4 I2 Purpose, where:
1 = Commute
2 = Non-Commute
3 = Total
5-6 I2 Production MSA
7-8 I2 Attraction MSA
9-18 F10.5 Transit Adjustment Factor
MODE CHOICE COEFFICIENTS (INPUT)
An ASCII file of time and cost coefficients needed for mode choice
calculations is called "ms.coef" and is located under /max7/proj/sr8.
The file has one record for each impedance type with the following
format.
Columns Variable Type Description
1-2 I2 Impedance Code, where:
1 = Drive Alone In-Vehicle Time
2 = Drive Alone In-Vehicle Cost
3 = Drive Alone Walk Terminal Time
4 = 2 Person Carpool In-Vehicle Time
5 = 2 Person Carpool In-Vehicle Cost
6 = 2 Person Carpool Walk Terminal Time
7 = 3+ Person Carpool In-Vehicle Time
8 = 3+ Person Carpool In-Vehicle Cost
9 = 3+ Person Carpool Walk Terminal
Time
10 = Transit First Wait Time
11 = Transit Second Wait Time
12 = Transit In-Vehicle Time
13 = Transit Walk Time
14 = Transit Cost
15 = Transit Auto Access Time
16 = Other Distance
17 = Transit Impedance Factor
3-12 F10.6 Commute Purpose Coefficient
13-22 F10.6 Non-Commute Purpose Coefficient
MODE CONSTANTS (INPUT)
An ASCII file of modal constants needed for mode choice calculations
is called "ms.cnst" and is located under /max7/proj/sr8. The file has
one record for each trip type and mode combination with the following
format.
Columns Variable Type Description
1-2 I2 Trip Type, where:
1 = Home-Work
2 = Home-College
3 = Home-School
4 = Home-Shop
5 = Home-Other
6 = Work-Other
7 = Other-Other
8 = Home-Serve Passenger
3-4 I2 Mode, where:
1 = Drive Alone
2 = 2 Person Carpool
3 = 3+ Person Carpool
4 = Transit-Walk
5 = Transit-Auto
6 = Other
5-11 F7.2 Mode Constant
INCOME CONSTANTS (INPUT)
An ASCII file of income constants needed for mode choice calculations
is called "ms.icnst" and is located under /max7/proj/sr8. The file
has one record for each income level, mode, and purpose combination
with the following format.
Columns Variable Type Description
1-2 I2 Income Level, where:
1 = $0-$9,999
2 = $10,000-$24,999
3 = $25,000 or More
3-4 I2 Purpose, where:
1 = Commute
2 = Non-Commute
5-6 I2 Mode, where:
1 = Drive Alone
2 = 2 Person Carpool
3 = 3+ Person Carpool
4 = Transit-Walk
5 = Transit-Auto
6 = Other
7-13 F7.2 Income Constant
TRIP LENGTH CONSTANTS (INPUT)
An ASCII file of trip length constants needed for mode choice
calculations is called "ms.lcnst" and is located under /max7/proj/sr8.
The file has one record for each trip type and mode combination with
the following format.
Columns Variable Type Description
1-2 I2 Distance Range, where:
1 = 2.9 Miles or Less
2 = 3.0 to 8.9 Miles
3 = 9.0 Miles or More
3-4 I2 Trip Type, where:
1 = Commute
2 = Non-Commute
5-6 I2 Mode, where:
1 = Drive Alone
2 = 2 Person Carpool
3 = 3+ Person Carpool
4 = Transit-Walk
5 = Transit-Auto
6 = Other
7-13 F7.2 Trip Length Constant
VEHICLE TRIP ENDS (OUTPUT)
ASCII files of vehicle trip ends, produced by the mode choice model,
are subsequently used to compute motor vehicle emissions. The files
are called "vehpa" and are located under temporary workspaces created
to evaluate alternatives. It should be noted that Series 7 trip type
definitions are used for vehicle trip end summaries. Home-college,
home-school, and home-serve passenger trips are combined with home-
other trips. Visitor and airport trips are combined with other-other
trips. Files have one record for each zone with the following format.
Columns Variable Type Description
1-4 I4 Zone
5-14 F10.0 Home-Work Vehicle Trip Productions
15-24 F10.0 Home-Shop Vehicle Trip Productions
25-34 F10.0 Home-Other Vehicle Trip Productions
35-44 F10.0 Work-Other Vehicle Trip Productions
45-54 F10.0 Other-Other Vehicle Trip Productions
55-64 F10.0 Home-Work Vehicle Trip Attractions
65-74 F10.0 Home-Shop Vehicle Trip Attractions
75-84 F10.0 Home-Other Vehicle Trip Attractions
85-94 F10.0 Work-Other Vehicle Trip Attractions
95-104 F10.0 Other-Other Vehicle Trip Attractions
TRANSIT TRIP ENDS (OUTPUT)
ASCII files of transit trip ends, produced by the mode choice model,
are subsequently used to factor base-year external transit trip
tables. The files are called "trpa" and are located under temporary
workspaces created to evaluate alternatives. Files have one record
for each zone with the following format.
Columns Variable Type Description
1-4 I4 Zone
5-14 F10.0 Transit-Walk Transit Trip Ends
15-24 F10.0 Transit-Auto Transit Trip Ends
CHAPTER 12
EXTERNAL TRIPS
EXTERNAL TRIPS
External trips are those trips that have one or both trip ends outside
of San Diego County. San Diego is fairly isolated so that external
travel makes up a relatively small proportion of all travel. The fact
that San Diego lies on an international border with Mexico means that
the nature of San Diego's external travel differs from most other
urban areas.
About three percent of all vehicle trips are external trips. Since
external trips are longer than internal trips, external trips comprise
about 15 percent of total vehicle miles of travel. External transit
trips are more significant. External transit boardings make up about
ten percent of all transit boardings and thirty percent of the
boardings on the light rail line that goes to the border.
The problem with modeling external trips is largely institutional.
Inputs to the transportation models may not be available from outside
planning agencies or may require considerable adjustment. These
institutional problems are compounded when attempting to model
external trips with an end in Mexico. Work is progressing on building
linkages between San Diego and Mexican transportation models to help
address trans-border transportation issues. However, long-range
transportation model forecasts are not yet available for the Tijuana
area. Cultural differences between the two countries and the effects
of customs and immigration policies make a unified modeling approach
difficult to implement.
MODEL STRUCTURE
The approach used to estimate external travel is to factor base-year
surveyed trip tables to future years. Resulting external travel
forecasts are added to internal travel estimates from the rest of the
modeling process prior to assignment. External highway and transit
trips are estimated in a similar manner, although separate procedures
are applied to each. No attempt is made to estimate external mode
share changes over time.
Highway Forecasts
The following data items from the 1986 and 1991 External Trip Surveys
are used in the external modeling process:
þ External zone of survey
þ Zone of first stop in County for inbound trips
þ Zone of last stop in County for outbound trips
þ External zone used for trips through the County
þ Trip purposes at origin and destination ends of trip
þ Expansion factor
External trips are categorized into the same ten trip types as
internal trips. A production-attraction Tranplan trip table is built
for each trip type from the survey data. A two-step procedure is used
to forecast external trips. The first step determines production and
attraction control totals by trip type at each external zone for the
years 1990 to 2015. This step is run only when new information about
external trips becomes available. The second step produces the actual
external trip tables for a model application. A two-step process is
used so that the trip generation model can correctly account for
differences between external productions and attractions when
balancing trips.
In the first step, a SANDAG Fortran program reads base-year external
trip tables from the surveys and aggregates trips to SRAs. After
aggregation, there is a matrix with a row for each of the nine
external zones and 45 columns for each of the SRAs. Cells in the
matrix contain trips produced at external zones and attracted to SRAs
within the region. A similar matrix contains trips attracted to
external zones and produced at SRAs. A third matrix has a row for
each external zone and a column for each external zone. Cells contain
trips through the County.
The program next reads 1990 and 2015 daily zone level person trip
productions and attractions by ten trip types from the trip generation
model. Trip ends are aggregated to SRAs. Production and attraction
growth factors can then be computed for each SRA and trip type.
These growth factors are applied to the SRA level external trip
tables, resulting in forecast year SRA level trip tables. Factoring
is done at the SRA level rather than the zone level to avoid the
problem with undeveloped base-year zones. These zones would have no
base-year survey trips for expansion and thus would have no future
year trips.
External zone traffic volumes from the growth factoring process are
totaled and adjusted to match traffic volume estimates at each
external zone based upon a trend line analysis of 1977 to 1993 traffic
counts. Table 12-1 lists external vehicle trips for 1990 from traffic
counts and 2015 projected vehicle trips from the trend line analysis.
The program writes out an intermediate dataset used in the next step
of the external trip forecasting process.
The second step repeats much of the process described in the first
step, except productions and attractions for a particular application
are input instead of 2015 trip ends. External trips are adjusted to
match the control totals produced in the first step.
At the end of the process, adjusted SRA level forecast trip tables are
broken down to zones based upon each zone's share of SRA internal
productions and attractions by the ten trip types. Daily external
trips are split into peak and off-peak periods by direction using the
same factors used for internal trips shown in Tables 10-1 and 10-2.
External trip ends are subtracted from total trip ends produced by the
trip generation model.
Table 12-1
EXTERNAL VEHICLE TRIPS
Zone Location 1990 2015 Change
1 I-5 at Mexican Border 77,600 100,900 + 30%
2 SR-905 at Mexican Border 21,300 53,100 +149%
3 SR-188 at Mexican Border 4,200 5,700 + 36%
4 I-8 at Imperial County 9,200 13,200 + 43%
5 SR-78 at Imperial County 900 2,300 +155%
6 SR-79 at Riverside County 1,800 3,300 + 83%
7 Pala-Temecula at
Riverside County 1,900 5,200 +174%
8 I-15 at Riverside County 62,700 103,400 + 25%
9 I-5 at Orange County 109,800 175,100 + 59%
Total 289,400 462,200 + 60%
Transit Forecasts
External transit forecasting is somewhat less complicated because only
one external zone, located at the San Ysidro U.S.-Mexican border
crossing, has a sufficient number of transit trips to warrant
consideration. External origin-destination transit survey data is
obtained from the 1990 Transit Ridership Survey.
Base-year external trips from the survey are compressed to SRAs. SRA
transit growth factors are computed based upon future-year and base-
year transit trip ends from the mode choice model. External transit
crossings from the factoring process are adjusted to match a control
total computed as follows:
ET(Y) = 16,300 + 300 * (Y-1990)
where :
ET = External Transit Trip
Y = Forecast year
The 16,300 transit border crossings in 1990 were obtained from the
passenger counting program. A 300 increase in transit border
crossings is based upon a trend line analysis of a 15-year history of
combined transit and highway border crossings. Resulting transit
trips at the SRA level are apportioned to zones based upon a zone's
fraction of SRA transit trips. External transit trips are added to
internal trips from the mode choice model.
DATA FILE
DOCUMENTATION
EXTERNAL VEHICLE TRIP CONTROL
TOTALS (INTERMEDIATE)
An ASCII file listing external vehicle trip control totals is called
"extotals" and is located under /max7/proj/sr8. The file has a total
of 702 records with three records for each of the nine external zones
and 26 years between 1990 and 2015. The file has the following
format.
RECORD 1. EXTERNAL TRIP ATTRACTIONS.
Trips produced in San Diego and attracted to outside locations.
Columns Variable TypeDescription
1-4 I4 Year
5-6 I2 External Zone Number
7-13 I7 Home-Work Attractions
14-20 I7 Home-College Attractions
21-27 I7 Home-School Attractions
28-34 I7 Home-Shop Attractions
35-41 I7 Home-Other Attractions
42-48 I7 Work-Other Attractions
49-55 I7 Other-Other Attractions
56-62 I7 Serve Passenger Attractions
63-69 I7 Visitor Attractions
70-76 I7 Airport Attractions
RECORD 2. EXTERNAL TRIP PRODUCTIONS.
Trips produced outside San Diego and attracted to inside locations.
Columns Variable TypeDescription
7-13 I7 Home-Work Productions
14-20 I7 Home-College Productions
21-27 I7 Home-School Productions
28-34 I7 Home-Shop Productions
35-41 I7 Home-Other Productions
42-48 I7 Work-Other Productions
49-55 I7 Other-Other Productions
56-62 I7 Serve Passenger Productions
63-69 I7 Visitor Productions
70-76 I7 Airport Productions
RECORD 3. TOTAL EXTERNAL TRIP ENDS
Total trip ends at zone.
Columns Variable TypeDescription
7-13 I7 Home-Work Trip Ends
14-20 I7 Home-College Trip Ends
21-27 I7 Home-School Trip Ends
28-34 I7 Home-Shop Trip Ends
35-41 I7 Home-Other Trip Ends
42-48 I7 Work-Other Trip Ends
49-55 I7 Other-Other Trip Ends
56-62 I7 Serve Passenger Trip Ends
63-69 I7 Visitor Trip Ends
70-76 I7 Airport Trip Ends
77-83 I7 Total Trip Ends
CHAPTER 13
HIGHWAY ASSIGNMENT
HIGHWAY ASSIGNMENT
Highway assignment is the process of loading vehicle trips between
zones onto specific segments of roadway. The resulting traffic
forecasts and related data are some of the most commonly used outputs
from the entire modeling process. Hence, a great deal of effort is
spent to make these forecasts as accurate as possible.
SANDAG loads traffic using Tranplan's Equilibrium assignment model.
This is an iterative technique for balancing estimated link volumes
with available capacity by minimizing an overall congestion index.
The model first finds minimum time paths between zones based upon
input speed assumptions. Trips between zones are accumulated on links
making a minimum time path between each zone pair. Once all trips
have been assigned, link volumes are divided by link capacities,
producing a volume-to-capacity (V/C) ratio which measures link
congestion. Revised link speeds are determined based upon the type of
facility and V/C ratio.
A second assignment iteration is then made using revised link speeds
from the first iteration. The accuracy of estimated volumes increases
with the number of iterations. However, each assignment iteration
takes about two hours of processing time so there is a practical limit
to the number of iterations that can be performed.
SANDAG performs four peak-period and four off-peak period assignment
iterations. Additional iterations were found to have minimal benefit.
Peak and off-peak trip tables are loaded separately since motorists
may choose different travel paths during the two time periods. For
example, heavily-congested facilities and metered ramps would tend to
be avoided during peak periods, but not during off-peak hours. Daily
volumes are obtained by adding together peak and off-peak assignments.
Tranplan has the capability of simultaneously assigning separate
single and high occupancy vehicle trip tables. This option requires
the coding of HOV facilities and about four hours of additional
computer processing time. Because of these factors, separate HOV
assignments are only performed for special applications where HOV
facilities are being evaluated.
MODEL STRUCTURE
The following files are input to the assignment model:
þ Tranplan highway network
þ Turn prohibitor files
þ Peak period origin-destination trip table
þ Off-peak period origin-destination trip table
þ Assignment parameters
Highway network and turn prohibitor coding was described in Chapter 6.
Peak and off-peak period vehicle trip tables are obtained from either
the person-to-vehicle trip factoring process or from the mode choice
model. These trip tables represent directional vehicle trip flows
between zones.
A number of Tranplan parameters are available for controlling the
assignment process. One parameter is a time period conversion factor
that scales trip tables during the assignment process to match hourly
capacities coded in highway networks. Conversion factors of 20% for
the peak period and 10% for the off-peak period are used. These
factors are derived from the Travel Behavior Survey. The peak-period
factor is computed by dividing the highest hourly volume in the peak
period by the total number of trips in the peak period. The median
hourly off-peak volume is divided by total off-peak trips to compute
the 10% off-peak conversion factor.
Other parameters are input that override Tranplan's default method of
adjusting speeds based upon congestion and a standard BPR curve.
Table 13-1 reflects Highway Capacity Manual (HCM) relationships
between congestion and speeds on uncontrolled facilities that are
input to the assignment model. HCM relationships show speeds being
relatively insensitive to congestion until volumes approach capacity,
at which point speeds fall off sharply. Once capacity is reached,
unstable conditions exist as traffic queues up behind the point of
maximum congestion. Minimum freeways speeds of 20 MPH and surface
street speeds of 15 MPH are used within the assignment model.
Table 13-1
SPEEDS BY VOLUME/CAPACITY RATIO
------------------ Posted Speed --------------------
V/C Ratio 60 50 45 40 35 30
0.0 60 50 45 40 35 30
0.1 60 48 44 39 35 30
0.2 60 47 43 39 35 30
0.3 60 47 43 38 35 30
0.4 60 46 42 37 34 30
0.5 60 45 40 36 34 30
0.6 57 43 38 36 34 30
0.7 54 42 37 35 33 30
0.8 52 38 35 31 32 30
0.9 47 35 33 32 31 30
1.0 30 30 30 30 30 30
1.1 25 25 20 20 20 20
1.2 20 20 15 15 15 15
The effect of congestion at signalized intersections and metered ramps
is to increase delay times as shown in the Table 13-2. Again, once
capacity is reached, unpredictable delay times occur and a maximum
delay is set.
Table 13-2
SIGNAL DELAY TIMES BY VOLUME/CAPACITY RATIO
(Minutes)
V/C Ratio Signals Ramp Meters
0.0 0.17 1.17
0.1 0.21 1.21
0.2 0.22 1.22
0.3 0.23 1.23
0.4 0.25 1.25
0.5 0.26 1.26
0.6 0.28 1.28
0.7 0.30 1.30
0.8 0.34 1.34
0.9 0.42 1.42
1.0 0.65 1.65
1.1 1.24 2.24
1.2 1.24 2.24
As described in Chapter 6, assignment groups categorize links into
freeways, uncontrolled surface streets, signalized surface streets,
and metered ramps. Uncontrolled streets are further sub-divided based
upon posted speed. Signalized streets are broken down by total time
on a link. Tables 13-3 and 13-4 show how speed and delay times are
translated into Tranplan assignment model parameters by specifying
ratios of congested speeds to input speeds for each assignment group
and ten percentage point increase in volume-to-capacity ratio.
For assignment groups 0 through 5 that are uncontrolled, a ratio of
congested speed to input speed is input. Off-peak input speeds
reflect freeflow conditions so that speeds decrease or remain the same
at all V/C ratios. Peak-period input speeds reflect average
congestion levels within an assignment group. Thus, adjusted speeds
are higher than input speeds where computed link congestion is less
than average.
For signal-controlled assignment groups 7 through 9, delay times for a
V/C ratio are added to the average time on links within the assignment
group. The ratio of average time to total time, including congestion,
is input. Speed adjustments differ between peak and off-peak
assignments because peak-period input signal delays reflect an average
V/C ratio of 0.4, while off-peak delays reflect no congestion.
Metered ramps have an assignment group code of 6. Meters are assumed
to be turned off in the off-peak period so that there would be
sufficient capacity to prevent delay for all V/C ratios. Peak-period
adjusted speeds reflect the effects of ramp metering delays upon ramp
speeds.
Table 13-3
OFF-PEAK PERIOD ADJUSTED TO INPUT SPEED RATIOS
Click HERE for graphic.
Table 13-4
PEAK PERIOD ADJUSTED TO INPUT SPEED RATIOS
Click HERE for graphic.
POST-ASSIGNMENT PROCESSING
After completing Tranplan highway assignments, additional processing
is needed to produce reports, data files, and plots tailored to SANDAG
needs. A SANDAG program and Arc/Info functions generate these
additional outputs. Arc/Info highway coverages and peak and off-peak
loaded Tranplan highway networks are inputs to the post-assignment
processing. Some of the major functions of this post-assignment
processing are described below.
Freeway Volumes
Most users want to see the total freeway volume at a location, not the
directional volumes that come out of Tranplan. Coders assign freeway
section numbers to all freeway arcs between interchanges. An Arc/Info
count station point coverage contains a point for each freeway section
by direction at the location where volumes are to be displayed.
Unique count station numbers are assigned to each location, but are
the same by direction. The program totals up directional freeway
volumes and assigns them to all arcs with the same section number for
display purposes.
Volume Adjustments
Individual model-estimated link volumes differ from ground counts even
after model calibration is complete. In the past, manual adjustments
were made to model-estimated volumes when traffic volume forecasts
were requested. In order streamline this process, an automated
adjustment procedure was developed.
The procedure keeps track of counts and base-year model-estimated
volumes for all locations with counts. Model-estimated volumes are
adjusted to compensate for base-year error. Over-estimates are
treated differently from under-estimates so that the potential for
negative numbers and large correction factors is eliminated. If the
model-estimated base-year volume exceeds the observed count, then a
count-to-estimate ratio is computed and applied to model-estimated
volumes for all forecasts as follows.
AVOL(L) = MVOL(L) * BYC(S) / BYE(S) (IF BYE > BYC), where:
AVOL = Adjusted Volume for Link "L"
MVOL = Model-Estimated Volume for Link "L"
BYC = Base Year Count at Count Station "S"
BYE = Base Year Model-Estimated Volume at Count Station "S"
Links where the model-estimated base-year volume is lower than the
observed count are adjusted by computing the difference between the
count and estimate. The difference is added to model-estimated
volumes for all forecasts as follows.
AVOL(L) = MVOL(L) + BYC(S) - BYE(S) (IF BYC > BYE), where:
AVOL = Adjusted Volume for Link "L"
MVOL = Model-Estimated Volume for Link "L"
BYC = Base Year Count at Count Station "S"
BYE = Base Year Model-Estimated Volume at Count Station "S"
Application of the adjustment procedure differs somewhat, depending
upon the type of facility. Model-estimated volumes are not adjusted
on facilities that did not exist in 1990. Freeway volumes for all
links with the same section number are adjusted based upon the
estimated and observed counts at the count station for the section.
Model-estimated volumes on surface street links in the City of San
Diego and unincorporated part of the County are adjusted based upon
the nearest count station on the route within 10 miles. A route is
defined as a group of connected links that have the same name.
Volumes on routes without count stations are not adjusted. Volumes on
links in other jurisdictions are adjusted based upon estimated and
observed counts by ADT link. Links not included in SANDAG's ADT
monitoring program are not adjusted.
In all cases, users have the option of reporting adjusted or
unadjusted volumes.
Travel Times
Peak and off-peak period link travel times are re-computed based upon
adjusted link volumes using Highway Capacity Manual Procedures. Peak
and off-peak period volume-to-capacity (V/C) ratios are computed.
Congested speeds are determined for un-controlled links based upon
free-flow speeds and V/C ratios as shown in Table 13-1. Congested
signal delay times are determined for controlled links based upon V/C
ratios, as shown in Table 13-2, and added to free-flow link times.
For most applications, high occupancy vehicle (HOV) facility speeds
are approximated by assuming HOV speeds are 5 MPH faster than adjacent
mixed-flow speeds. Ramp meter delays at ramps with HOV by-pass lanes
are not included. Applications where separate HOV facilities have
been included would estimate HOV speeds based upon HOV lane V/C
ratios.
Level of Service Computations
Procedures from Chapter 11 of the Highway Capacity Manual are used to
compute arterial level of service. Coders group individual highway
coverage arcs into sections and assign a unique number to each
section. As highway links are processed, congested times, distances,
and number of signals are accumulated for each section. Once all
links have been processed, level of service is computed for each
section based upon free-flow speed, average signal spacing, and the
ratio of the section running times with and without signal delay
times.
Caltrans uses special procedures to estimate freeway level of service.
A separate file of freeway data is produced as input to these
procedures.
MODEL OUTPUTS
A number of outputs are produced from the post-assignment process in
addition to Tranplan's loaded highway networks. Adjusted volumes,
unadjusted volumes and level
of service are written to highway coverage arc attribute tables.
Arc/Info plot routines produce various plots for checking and display
purposes.
Off-peak and peak period travel times are also written to highway
coverage arc attribute tables. This enables transit network
procedures to make use of modeled highway travel times when estimating
transit running times.
Inputs to Tranplan's Build Highway Network function are created for
mixed-flow and HOV networks. Table 6-6 lists Tranplan highway network
inputs. Mixed-flow networks contain model-estimated off-peak link
times in "Field1" and peak period link times in "Field2." Other
fields are the same as described in Table 6-6. HOV network inputs
contain HOV facility link times where they exist and mixed-flow times
on other facilities. Tranplan networks are subsequently built for
mixed-flow and HOV conditions. Zone-to-zone travel time files are
produced for peak period mixed-flow conditions, peak period HOV
conditions, and off-peak period mixed-flow conditions. These travel
time files are input to the mode choice program and second stage trip
distribution.
Another file is created that summarizes link speeds, times, and
volumes for input to SANDAG's emission estimating routine.
A number of reports summarize assignment results. Table 13-5 shows
adjusted and unadjusted daily vehicle miles of travel. As indicated,
the adjustment process has little effect upon regionwide VMT, although
the effects upon specific functional classes and individual links are
more significant. VMT increases at a slightly faster rate than
vehicle trips as activities become more dispersed and average trip
lengths increase.
Table 13-5
VEHICLE MILES OF TRAVEL
---------- Unadjusted -------------------- Adjusted ----------
Functional 1990 2015 Change 1990 2015 Change
Class
Freeway 31,391,000 50,073,000 +60% 30,309,000 48,252,000 +59%
Prime 4,947,000 8,966,000 +81% 5,306,000 9,085,000 +71%
Major 11,065,000 15,058,000 +36% 11,657,000 15,323,000 +31%
Collector 4,190,000 7,222,000 +72% 4,367,000 7,080,000 +62%
Local
Collector 3,183,000 4,689,000 +47% 3,356,000 4,697,000 +40%
Rural
Collector 244,000 628,000 +157% 214,000 516,000+141%
Rural Light 325,000 570,000 +75% 298,000 497,000 +67%
Local 550,000 831,000 +51% 625,000 893,000 +43%
Ramp 2,900,000 4,310,000 +49% 2,895,000 4,305,000 +49%
Zone
Connector 3,016,000 4,941,000 +64% 3,016,000 4,941,000 +64%
Total 61,811,000 97,288,000 +57% 62,043,000 95,589,000 +54%
Another important output from the assignment model is the effect of
congestion upon speeds. The model computes off-peak and peak-period
speeds for each highway link based upon input speeds, capacities,
assigned volumes, and relationships between speeds and congestion.
Table 13-6 summarizes resulting model-estimated average speeds by
functional class. Future year speeds depend upon assumptions about
the extent of roadway improvements that will be implemented. Table
13-6 speeds are based upon the recommended highway system in the
Regional Transportation Plan. Roadway capacity in the recommended
system grows at a slightly slower rate than travel so average speeds
drop over time.
Table 13-6
AVERAGE SPEEDS AFTER ASSIGNMENT
Functional Class 1990 2015 Change
Freeway 48.3 46.1 -5%
Prime 27.3 27.4 +0%
Major 24.9 23.9 -4%
Collector 24.6 25.8 +5%
Local Collector 22.2 21.8 -2%
Rural Collector 34.2 32.8 -4%
Rural Light 34.1 33.1 -3%
Local 20.0 19.8 -1%
Ramp 22.3 20.4 -9%
Zone Connector 20.0 20.0 +0%
Total 31.9 31.4 -2%
CALIBRATION
Comparing traffic volumes from the highway assignment model with
observed traffic counts provides one of the few opportunities to check
the accuracy of model output. Computer-generated plots comparing
model-estimated link volumes with traffic counts are one tool for
evaluating model results. Large discrepancies between counts and
model estimates may indicate network coding errors, misrepresentation
of access opportunities, land use coding errors, or inappropriate trip
generation rates. Root mean square (RMS) error, computed using the
equation below, provides a method of summarizing link volume errors.
RMS = SQRT (SUM [ (OBS(I) - EST(I)) **2 ] / NOBS )
where:
RMS = Root mean square error
SQRT = Square root function
SUM = Summation over all links
OBS(I) = Observed traffic count link "i"
EST(I) = Estimated traffic volume for link "i"
NOBS = Number of observations
Comparing screenline totals of traffic counts and model estimates
provides another check of model accuracy. Screenline errors typically
point to trip distribution errors. Figure 9-1 shows SANDAG's
screenlines and Table 9-5 summarizes screenline accuracy.
Comparing observed and estimated daily vehicle miles of travel (VMT)
provides a final model check. VMT is calculated by multiplying the
observed count or estimated volume on a link by the length of the link
and summing over all links. Total VMT is a measure of how well the
entire chain of models is performing. Inaccurate total VMT estimates
could be due to errors in the number of person trips from the trip
generation model, trip lengths from the trip distribution model, or
auto mode shares from the mode choice model. Comparing observed and
estimated VMT by functional class may indicate systematic errors in
the assignment model or network coding.
SANDAG has traditionally performed four iterations of off-peak and
peak-period assignment. One of the issues when setting up the
equilibrium assignment model was determining the optimal number of
assignment iterations that would provide accurate assignment results
while minimizing the number of iterations. Tranplan's "Number of
Iterations" and "EPS" parameters control the number of assignment
iterations.
Model-estimated volumes for 1990 were compared with 1990 traffic
counts after assigning vehicle trips using a low EPS value and two,
four and eight iterations. As indicated in the table below, base-year
model accuracy improves slightly when the number of assignment
iterations is increased from two to four. However, increasing
assignment iterations from four to eight yields no accuracy
improvement.
Table 13-7
ASSIGNMENT ERROR BY NUMBER OF ITERATIONS
Number of Iterations Freeway RMS Error Total RMS Error
2 15.5% 39.6%
4 14.1% 37.5%
8 14.3% 37.6%
The number of iterations is more likely to make a difference in the
future when higher levels of congestion are expected to occur.
Vehicle trips for 2015 were assigned to a 1990 highway network in
order to test a worse-case situation. When Tranplan's default EPS
value of 0.1 is used, off-peak assignment reaches closure after two
iterations and peak-period assignment stops after three iterations.
This would indicate that four iterations is sufficient. Future year
trips were also assigned to the 1990 network using a low EPS value and
two, four and eight iterations. Table 13-8 compares the two and eight
iteration assignment results to the four iteration results by
summarizing the percentage of total roadway mileage where the assigned
volumes are within 10% of each other. Differences between two and
four iterations are more significant than those between four and eight
iterations. Furthermore, links with the largest differences are
located in heavily congested corridors where eight-iteration volumes
appeared to be no more logical than the four-iteration results. It
was decided to continue to perform four off-peak and four peak
assignment iterations as a result of these tests.
Table 13-8
ASSIGNMENT DIFFERENCES BY NUMBER OF ITERATIONS
Functional Class Two vs Four Iterations Four vs Eight Iterations
Freeway 88% 99%
Prime 66% 94%
Major 63% 90%
Collector 59% 84%
Local Collector 52% 72%
Rural Collector 66% 87%
Rural Light 84% 92%
Local 50% 66%
Ramp 49% 70%
Zone Connector 83% 89%
Total 69% 85%
Table 13-9 compares observed and estimated daily vehicle miles of
travel (VMT) by functional class produced by the assignment model.
Note that VMT estimates in Table 13-9 are based upon an expansion of
links with traffic counts. Resulting VMT estimates differ slightly
from those shown previously in Table 13-5 that are based upon all
links. Total estimated VMT agrees with observed VMT, indicating that
models accurately estimate aggregate travel.
Comparing observed and estimated VMT by functional class may indicate
systematic errors in the assignment model or network coding. While
Table 13-9 shows freeway VMT is slightly over-estimated, VMT estimates
for all functional classes with significant VMT are within 10% of
observed values.
A more stringent measure of model accuracy is provided by the root
mean square error between estimated and observed link volumes. This
measure summarizes the error in individual link volumes and eliminates
the tendency of VMT summaries to obscure results due to compensating
errors. For example, the model could be performing poorly by under-
estimating half of all freeway links by 50%, while over-estimating the
other half by 50%. In this case, observed and estimated freeway VMT
comparisons would agree and not indicate a problem. The RMS error
would be 50% and better reflect model performance. The tables that
follow summarize VMT and RMS error by functional class and MSA; and
volume group and MSA.
Table 13-9
OBSERVED AND ESTIMATED VEHICLE MILES OF TRAVEL
Functional Class Observed Estimated Error
Freeway 30,110,000 31,391,000 + 4%
Prime 5,412,000 4,947,000 - 9%
Major 11,508,000 11,065,000 - 4%
Collector 4,372,000 4,190,000 - 4%
Local Collector 3,203,000 3,183,000 - 1%
Rural Collector 217,000 242,000 +12%
Rural Light 291,000 325,000 +12%
Local 678,000 550,000 -19%
Ramp 797,000 979,000 +23%
Total 56,588,000 56,872,000 + 0%
Table 13-10
ROOT MEAN SQUARE ERROR BY FUNCTIONAL CLASS
Click HERE for graphic.
Table 13-11
ROOT MEAN SQUARE ERROR BY VOLUME GROUP
Click HERE for graphic.
Table 13-12
VMT ERROR BY FUNCTIONAL CLASS
Click HERE for graphic.
Table 13-13
VMT ERROR BY VOLUME GROUP
Click HERE for graphic.
DATA FILE
DOCUMENTATION
TRANPLAN HIGHWAY NETWORK LINK FILE
Temporary ASCII highway network link files called "sovdata" and
"hovdata" are created for input to Tranplan for building highway
networks in Tranplan format. They contain mixed-flow and high
occupancy vehicle (HOV) travel times from a first stage model run.
Highway link files are located under temporary workspaces created to
evaluate alternative networks. The files have one record per highway
link with the following format.
Columns Variable Type Description
1-5 I5 A-node number
6-10 I5 B-node number
11-11 I1 Assignment Group Code
12-15 F4.2 Link Distance (Miles)
16-16 A1 "T"
17-20 F4.2 First Stage Off-peak Link Travel Time
(Minutes)
21-24 F4.2 First Stage Peak Link Travel Time (Minutes)
27-28 I2 Number of Lanes
29-30 I2 Sphere number, where:
1 = Carlsbad,
2 = Chula Vista,
3 = Coronado,
4 = Del Mar,
5 = El Cajon,
6 = Encinitas,
7 = Escondido,
8 = Imperial Beach,
9 = La Mesa,
10 = Lemon Grove,
11 = National City,
12 = Oceanside,
13 = Poway,
14 = City of San Diego,
15 = San Marcos,
16 = Santee,
17 = Solana Beach,
18 = Vista,
19 = Unincorporated Area
31-32 I2 Functional Classification
33-38 I6 Link Capacity (Vehicles per Hour)
39-39 I1 From-to Code, where:
1 = A-B node in Arc/Info from-to
direction
2 = A-B node in Arc/Info to-from
direction
40-44 I5 HWYCOV-ID
HIGHWAY EMISSIONS FILE
Temporary ASCII files called "dtimdata" are created for subsequent
emissions analysis. Files are located under temporary workspaces
created to evaluate alternative networks. The files have one record
per highway link with the following format.
Columns Variable Type Description
1-7 F7.2 Link Distance (Miles)
8-14 F7.2 Off-peak Link Travel Time (Minutes)
14-21 F7.2 Peak Link Travel Time (Minutes)
22-28 I7 Off-Peak Speed (Miles per Hour)
29-35 I7 Peak Speed (Miles per Hour)
36-42 I7 Off-Peak Period Volume (Vehicles per
Period)
43-49 I7 Peak Period Volume (Vehicles per Period)
50-56 I7 Off-Peak Period Vehicle Miles of Travel
57-63 I7 Peak Period Vehicle Miles of Travel
64-70 I7 Daily Truck Volume (Vehicles per Day)
71-77 I7 HWYCOV#
78-84 I7 Functional Classification
CHAPTER 14
TRANSIT ASSIGNMENT
TRANSIT ASSIGNMENT
Transit assignment loads transit trips between transit access points
(TAPs) onto specific links and lines. While this function has few
user parameters and is largely dependant upon results from other parts
of the modeling process, it produces results that are important for
evaluating the validity of transit models and the effectiveness of
proposed transit service improvements.
There are two ways of accounting for transit trips that can be
somewhat confusing when presented with transit patronage estimates.
Linked transit trips, estimated by the mode choice model, are defined
as starting prior to boarding the first bus or train and ending at the
ultimate destination after the final transit stop. Linked trips are
sometimes referred to as revenue trips since full fare is usually
collected only once on an overall transit trip.
The transit assignment model estimates unlinked trips where each
transit vehicle boarded in an overall trip is counted separately.
Linked and unlinked trips are the same for trips without a transfer.
Two or more unlinked trips occur on transit trips involving a
transfer. Unlinked trips are sometimes called boardings. The ratio
of unlinked to linked transit trips is a measure of the transfer rate.
Both methods of categorizing transit trips have their uses. Unlinked
trips are easier to monitor. However, linked trips are usually
considered the better indicator of transit system performance. Some
transit service changes force transfers that drive up the number of
unlinked trips but actually degrade system performance as measured by
linked trips.
MODEL STRUCTURE
Tranplan's Load Transit Network function assigns transit trips given
the following inputs:
þ Peak period Tranplan TAP-to-TAP walk access transit trip table
þ Peak period Tranplan TAP-to-TAP auto access transit trip table
þ Peak period Tranplan TAP-to-TAP transit paths
þ Off-peak period Tranplan TAP-to-TAP walk transit trip table
þ Off-peak period Tranplan TAP-to-TAP auto transit trip table
þ Off-peak period Tranplan TAP-to-TAP transit paths
þ Tranplan transit network
Transit trip tables are initially produced by the mode choice model
and then modified to add transit border crossings in the external
transit trip function. The assignment program accumulates trips
between zones on links and lines making up the minimum path between
each zone pair. When two or more lines are on the minimum path, trips
are apportioned to each line based upon the line's frequency of
service.
The model produces a file of non-transit link loads, an unformatted,
binary loaded transit legs file, and summary reports. A SANDAG
program matches Tranplan output files with Arc/Info transit coverages
to produce reports and plots that are easier to work with than
standard Tranplan outputs.
MODEL OUTPUTS
SANDAG's program produces route boarding reports and transit link
volume plots. The program adds together peak and off-peak loaded legs
files to obtain daily totals. An intermediate file created during
transit network processing provides a connection between Tranplan
transit network links and Arc/Info transit coverage arcs.
A summary of transit boardings by route and mode is one of the model
results. Table 14-1 shows the change in model-estimated boardings by
mode between the 1990 transit system and the 2015 transit system
recommended in the Regional Transportation Plan. Light rail boardings
increase the most due to the extensive light rail expansion proposed
in the RTP. Some local bus routes are phased out as new rail service
comes on-line. As a result, local bus boardings show the slowest rate
of increase.
Dividing transit boardings from transit assignment by unlinked transit
trips from mode choice produces a measure of the transit rate. A
slight increase in the transfer rate is expected as light rail service
becomes more prevalent.
Table 14-1
TRANSIT BOARDINGS
Mode 1990 2015 Change
Commuter Rail 0 8,000 +0%
Light Rail 52,000 297,000 +471%
Express Bus 14,000 40,000 +186%
South County Local Bus 129,000 188,000 +46%
North County Local Bus 32,000 75,000 +134%
Systemwide Total 227,000 608,000 +168%
Unlinked Trips 145,000 320,000 +121%
Transfer Rate 1.57 1.90 +21%
Access modes used to arrive at boarding transit stops, and egress
modes used to go from de-boarding transit stops to final destinations
are also available from transit assignment. Table 14-2 summarizes the
change in access/egress mode use from 1990 to 2015. As local bus
routes are restructured to feed new rail lines, the percentage of
total transit trips with a transfer increases and the percentage of
trips that walk at both ends drops.
Table 14-2
ACCESS/EGRESS MODE USE
------- Access/Egress Mode -------
Transit Mode Walk Drive Transfer Total
Commuter Rail 1990 0% 0% 0% 0%
2015 21% 35% 44% 100%
Change +21% +35% +44% +100%
Light Rail 1990 51% 13% 36% +100%
2015 35% 13% 52% 100%
Change -16% +0% +16% +100%
Express Bus 1990 52% 6% 42% +100%
2015 51% 3% 46% 100%
Change -1% -3% +4% +0%
South County Local Bus 1990 69% 1% 30% +100%
2015 63% 1% 36% 100%
Change -6% +0% +6% +0%
North County Local Bus 1990 69% 3% 28% +100%
2015 61% 3% 36% 100%
Change -8% +0% +8% +0%
Systemwide Total 1990 64% 4% 32% 100%
2015 48% 8% 44% 100%
Change -16% +4% +12% +0%
MODEL CALIBRATION
Reports of transit boardings provide one of the most useful methods of
evaluating transit model estimates. Table 14-3 compares estimated and
observed boardings for the ten routes with the highest ridership.
Mode and the system boardings are also summarized. Observed transit
boardings are from SANDAG's Passenger Counting Program described in
Chapter 3. Model estimated system-wide boardings are within 7% of
actual boardings, indicating that the model is correctly estimating
transfers.
Mode-specific boardings show more error, but are still within
acceptable limits. The model over-assigns trips to bus routes at the
expense of the light rail mode. This overestimation may indicate the
need for "mode-specific" mode choice parameters that would account for
the preference of transit riders for light rail above its time and
cost characteristics.
Tranplan uses a relatively unsophisticated transit assignment
technique that assigns all transit trips to the minimum path. As a
result, accurate route-level boardings are difficult to estimate.
Overall, estimated boardings are within 10% of observed boardings on
1/3 of all routes; and within 25% on 2/3 of all routes. Table 14-3
shows that some high volume routes have relatively large errors.
Table 14-3
OBSERVED AND ESTIMATED TRANSIT BOARDINGS
Route Observed Estimated Absolute Percent
or Mode Boardings Boardings Error Error
510 35,020 35,030 +10 +0%
520 16,856 17,583 +727 +4%
7 14,327 11,792 -2,535 -18%
11 9,128 9,291 +163 +2%
34 7,626 5,792 -1,834 -24%
2 7,158 5,784 -1,374 -19%
3 7,036 4,823 -2,213 -31%
29 6,330 7,037 +707 +11%
302 5,122 3,660 1,482 -28%
15 4,994 5,925 +931 +19%
Light Rail 51,876 52,613 +737 +1%
Express Bus 13,052 14,161 +1,109 +8%
South Local Bus 116,929 128,709 +11,780 +10%
North Local Bus 29,792 31,964 +2,172 +7%
System Total 211,649 227,447 +15,798 +7%
Accurate access mode estimates are also important. Table 14-4
compares observed access/egress mode use from the Transit Ridership
Survey with model-estimated mode use. Because the overall transfer
rate is slightly over-estimated by the model, transfer access mode use
is also somewhat over-estimated. The relative use of different
access/egress modes by transit mode is accurately reflected.
Table 14-4
OBSERVED AND ESTIMATED ACCESS/EGRESS MODE USE
------- Access/Egress Mode -------
Transit Mode Walk Drive Transfer Total
Light Rail Observed 61% 16% 23% +100%
Estimated 51% 13% 36% 100%
Error -10% -3% +13% +0%
Express Bus Observed 57% 12% 31% +100%
Estimated 51% 3% 46% 100%
Error -6% -9% +15% +0%
South County Local Bus Observed 70% 2% 28% +100%
Estimated 69% 1% 30% 100%
Error -1% -1% +2% +0%
North County Local Bus Observed 71% 4% 25% +100%
Estimated 69% 3% 28% 100%
Error -2% -1% +3% +0%
Systemwide Total Observed 67% 7% 26% 100%
Estimated 64% 4% 32% 100%
Error -3% -3% +6% +0%
CHAPTER 15
MOTOR VEHICLE
EMISSION MODELING
MOTOR VEHICLE EMISSION MODELING
State and federal legislation requires plans and programs to undergo
rigorous air quality analysis. As a result, the amount of motor
vehicle emissions generated under different transportation system and
land use alternatives is an important measure of effectiveness. Air
quality considerations are even more important because San Diego is
currently designated as a serious non-attainment area by the Federal
Environmental Protection Agency. Non-attainment areas are required to
submit plans that demonstrate how air quality standards will be met.
Emission modeling has two primary tasks. One task is to estimate
emissions totals for a forecast year given land use, demographic, and
transportation system assumptions. The other task is to quantify the
effectiveness of the various proposed transportation control measures
that are designed to reduce vehicle use and resulting emissions.
Three pollutants are of primary concern: reactive organic gases
(ROG), carbon monoxide (CO), and nitrogen oxide (NOx). CO and NOx are
pollutants on their own. ROG and NOx emissions also react to form
photochemical smog, San Diego's most pervasive air quality problem.
Reactive organic gases are sometimes referred to as reactive hydro-
carbons (RHC). Portions of the model deal with total organic gases
(TOG) as opposed to ROG, which is the part of total organic gases that
reacts to form smog.
Air quality standards are based upon the concentration of pollutants
in the atmosphere which is determined by the amount, timing, and
location of emissions as well as meteorological conditions. Air
quality planners typically make use of two levels of emission
estimates. One level of reporting is an average daily emission total
by pollutant for the San Diego Air Basin. The Air Basin coincides
with the County of San Diego and the Region modeled by SANDAG.
Emission totals can be used to evaluate air quality impacts with the
underlying assumption that there is a direct relationship between the
amount of emissions for the entire air basin and the concentration of
pollutants in the atmosphere at particular locations within the air
basin.
In fact, the relationship between emissions and air pollution is
highly complex. Air quality dispersion models have been devised to
better represent the factors affecting air pollution. These models
require much more detailed emission estimates by pollutant, hour of
the day, and location in the Region. Air quality modeling grids shown
in Figure 15-1 are used to summarize emissions by location.
(Insert Figure 15-1 - Air Quality Modeling Grids)
Motor Vehicle Emission Sources
On-road motor vehicle emissions are attributed to several different
processes:
þ start-up
þ hot soak evaporation
þ diurnal evaporation
þ running exhaust
þ running losses
Engines produce start-up emissions during the first few minutes of
engine operation. Start-up emissions are expressed as grams per trip
start and are applied to trip ends. The engine-off time prior to a
trip start can significantly affect start-up emission rates. An
engine off time longer than one hour is assumed to put a vehicle in a
cold-start mode and increase emissions. Vehicles without catalysts
are not assumed to go into a cold-start mode until after four hours.
The shorter time for catalyst vehicles reflects the fact that the
catalyst must be hot to be effective, and after one hour, the catalyst
is essentially cold. Start-up emission rates are not affected by
speed but generally drop as ambient temperature increases. ROG, NOx
and CO emissions are all produced during engine start-up.
Hot soak evaporative emissions occur after an engine is turned-off as
fuel evaporates from the vehicle. More hot soak emissions are
produced the longer a vehicle is parked, up to a two hour maximum.
Hot soak emissions are expressed as grams per stop and are applied to
trip ends. Hot soak emission rates increase as ambient temperature
increases. Only ROG emissions are produced during hot soak. Emission
factors are adjusted as follows:
Table 15-1
HOT SOAK EMISSION RATES
Minutes Parked Percent of One Hour Soak Emissions
0 0%
30 70%
60 100%
120 or More 130%
Diurnal emissions occur when fuel evaporates from gas tanks and
escapes into the atmosphere due to ambient temperature increases.
Diurnal emissions are expressed in terms of grams per day per vehicle
and are applied to trip ends. Again, diurnal evaporation only
produces ROG emissions.
Running exhaust emissions are tailpipe emissions produced by the
combustion of fuel. Exhaust emission factors are expressed as grams
per vehicle hour and are applied to traffic volumes on roads. All
three pollutants (ROG, CO, and NOx) are produced by combustion.
Emission factors vary depending upon traffic speed and temperature.
Running losses account for evaporative emissions while vehicles are in
use, as opposed to hot soak and diurnal emissions that occur while
vehicles are at rest. Running loss emissions are included with
running exhaust emission factors and are expressed as grams per
vehicle hour. Running losses produce only ROG emissions.
MODEL STRUCTURE
Emission modeling is complicated by the wide range of factors
affecting emissions, the number of agencies involved, and the
specificity required of model output. Emission model inputs include:
þ Emission factors
þ Temperature assumptions
þ Air quality modeling grid conversions
þ Time of day distribution of travel
þ Vehicle trip ends
þ Highway link volumes, speeds, and distances
þ Transit link volumes, speeds, and distances
Emission Factors
Emissions are calculated by applying emission factors for each process
to a corresponding measure of transportation activity. Emission
factors are obtained from the California Air Resources Board's (ARB)
"EMFAC" computer program. Emission factors vary by vehicle class
(light-duty auto, light-duty truck, medium-duty truck, urban bus,
heavy-duty diesel truck, heavy-duty gasoline truck, motorcycles);
technology (catalyst-equipped, non-catalyst, diesel); and year.
Emission factors that vary by temperature (start-up, hot soak, and
running exhaust) are further broken down by specified temperature
ranges. Exhaust emission factors are also broken down by speed range.
ARB periodically updates the EMFAC program. EMFAC7F is the program
currently in use. EMFAC is set up to produce emission factors for 25,
three degree temperature ranges between 36 and 105 degrees fahrenheit.
An additional 21, three MPH speed ranges from 3 MPH to 65 MPH are used
for emission rates that vary by speed. Table 15-2 shows how another
program, Impact Rate Summarization (IRS), combines EMFAC output for
the seven vehicle classes into three broader vehicle groupings:
vehicles used for personal, non-commercial use; commercial trucks; and
urban buses. Data files from IRS are obtained for the three vehicle
groups and various analysis years.
The emission model includes emissions produced by Mexican vehicles
while travelling in San Diego. Mexican vehicle emission control and
fueling regulations are less stringent than California regulations;
however, specific Mexican vehicle emission factors are not available.
Emission factors for twenty years prior to the main analysis year are
assumed to reflect Mexican vehicle emission rates.
Table 15-2
VEHICLE CLASSIFICATION GROUPS
---------- Vehicle Group ----------
Vehicle Classification Personal Truck Bus
Light-Duty Vehicle 74.5% 0.0% 0.0%
Light-Duty Truck 19.0% 0.0% 0.0%
Medium-Duty Vehicle 5.7% 0.0% 0.0%
Heavy-Duty Diesel Truck 0.0% 45.0% 0.0%
Heavy-Duty Gasoline Truck 0.0% 55.0% 0.0%
Urban Bus 0.0% 0.0% 100.0%
Motorcycle 0.5% 0.0% 0.0%
Temperature Data
Ambient temperature assumptions are another non-SANDAG emission model
input. As described above, emission rates vary as temperatures
increase. Some model applications use simplified, ARB-supplied
temperature assumptions listed below.
Table 15-3
SIMPLIFIED TEMPERATURE ASSUMPTIONS
(Degrees Fahrenheit)
Time Period Temperature
Midnight to 6:00 AM 66
6:00 AM to 9:00 AM 70
9:00 AM to Noon 88
Noon to 3:00 PM 91
3:00 PM to 6:00 PM 85
6:00 PM to Midnight 75
Actual temperatures vary widely across the air basin, by hour of the
day, and from one day to the next. The Air Pollution Control District
(APCD) supplies files with temperatures for each hour and air quality
modeling grid for various "episode" days that coincide with
meteorological conditions that are to be simulated.
Time-of-Day Distribution of Travel
The distribution of travel throughout the day is needed to match
travel activity with temperature assumptions described above. In
addition, emissions produced during certain times of the day have a
disproportionate affect upon air quality. For example, morning
emissions have more time to react to form photo-chemical smog than
those produced later in the day. Typical vehicle use patterns can
further exacerbate air quality impacts of travel by time of day. For
example, most vehicles start up cold in the morning after sitting
overnight. Thus, early morning emission rates are typically higher
than those later in the day when trips are more likely to be made in a
warmed-up state.
The 1986 Travel Behavior Survey was tabulated to obtain the following
time-of-day distributions of vehicle use. Resulting distributions are
listed at the end of the chapter. Distributions are assumed to remain
constant for all model applications over the forecast period.
þ Percent of daily vehicle trip ends starting-up by hour of day,
trip type, and trip end type (production or attraction)
þ Percent of vehicle starts in an hour with a duration of one hour
or more prior to trip start by hour, trip type, and trip end type
(represents catalyst cold starts)
þ Percent of vehicle starts in an hour with a duration of four hours
or more prior to trip start by hour, trip type, and trip end type
(represents non-catalyst cold starts)
þ Percent of daily vehicle trip ends stopping by hour of day, trip
type, and trip end type
þ Weighted percent of one hour hot soak emissions at the end of a
trip using the relationship shown in Table 15-1 by hour of day,
trip type, and trip end type
þ Average duration at the end of trip by hour of day, trip type, and
trip end type
þ Percent of off-peak and peak period vehicle miles of travel by
hour
Caltrans vehicle classification counts in San Diego were also analyzed
to determine the hourly distribution of heavy-duty truck travel that
was not covered by the Travel Behavior Survey.
Air Quality Modeling Grid Conversions
Transportation models produce vehicle travel activity data at the zone
and link level. These outputs need to be converted to air quality
modeling grids for emission modeling purposes. Arc/Info procedures
overlay zones, highway links and transit links on grids. Data files
are produced that summarize the fraction of each zone's area in a
grid; the fraction of each highway link's length in a grid; and the
fraction of each transit link's length in a grid.
Emission Model
Once transportation models have been run for a particular scenario and
all of the data files described above have been tabulated, an emission
forecast can be produced using SANDAG's emission model. The model
produces reports summarizing daily countywide emissions as well as a
detailed dataset for input to air quality models.
The model works by applying start-up, hot soak, and diurnal emission
factors to daily vehicle trip forecasts. Internal vehicle trip ends
by zone from either the mode choice model or vehicle trip factoring
process are first apportioned to air quality modeling grids. A
similar process is used to assign the San Diego end of external
vehicle trips to zones. External trips are further sub-divided into
Mexican and U.S. trips by applying Mexican residence factors to trips
from external zones along the U.S.-Mexican border.
Daily trip ends are subsequently distributed to each hour of the day.
Emission factors and hourly vehicle use patterns are applied to the
resulting hourly trip ends to estimate emissions. These emissions are
accumulated in an internal matrix by hour, grid and the following
combination of pollutant, process, and technology:
þ TOG cold start non-catalyst
þ TOG cold start catalyst
þ TOG cold start diesel
þ TOG hot start non-catalyst
þ TOG hot start catalyst
þ TOG hot start diesel
þ TOG hot soak non-catalyst
þ TOG hot soak catalyst
þ TOG diurnal non-catalyst
þ TOG diurnal catalyst
þ CO start
þ NOx start
Another part of the model applies running exhaust and running loss
emission factors to traffic volumes on each roadway segment from
Tranplan's highway assignment model, modified by SANDAG's post-
assignment processing. Off-peak and peak link volumes are apportioned
to air quality modeling grids by hour using a link-to-grid conversion
file and a VMT diurnal distribution file.
Running exhaust emission factors vary by speed and temperature. In
most cases peak and off-peak link speeds from the transportation model
and gridded hourly temperatures from APCD are used to find the
appropriate emission factor. The "dtimdata" file described at the end
of Chapter 13 translates highway link data between the transportation
models and the emission model. Emissions are calculated by applying
running exhaust and emission factors to the vehicle hours on the link
based on link distance, speed, and volume. Emissions are accumulated
in an internal matrix by hour, grid and the following combination of
pollutant, process, and technology:
þ TOG stabilized running exhaust non-catalyst
þ TOG stabilized running exhaust catalyst
þ TOG running loss non-catalyst
þ TOG running loss catalyst
þ CO stabilized running exhaust
þ NOx stabilized running exhaust
ARB feels that running exhaust emissions increase dramatically at high
speeds. In order to account for these emissions, traffic volumes on
freeway links that are in a free-flow condition (modeled speeds above
58.5 miles per hour) are re-allocated to three speed ranges using a
speed distribution based upon California Highway patrol spot speed
studies. One-third of link volumes are put in the 58.5 to 61.5 MPH
category, one-third in the 61.5 to 64.5 MPH category, and one-third in
the 64.5 MPH or more category.
Running exhaust emissions from buses are computed in a similar manner
using bus volumes and speeds from modeled transit networks.
MODEL OUTPUT
Table 15-4 summarizes motor vehicle emission forecasts that were
produced by the emission model for the air quality conformity analysis
of the 1994 Regional Transportation Improvement Program. Motor
vehicle emissions are expected to drop significantly over time as the
continued phasing-in of cleaner vehicles offsets projected travel
increases.
Table 15-4
MOTOR VEHICLE EMISSION FORECASTS
(Tons per Day)
Pollutant 1990 2015 Change
ROG 160 82 -49%
CO 1137 348 -69%
NOx 145 81 -44%
Emission forecasts vary somewhat, depending upon transportation
facility assumptions, transportation policy assumptions (such as the
proposed TDM ordinance), and land use assumptions. The main effect of
many transportation improvements is to improve operating speeds on the
highway system. The emissions effects of these speed changes are
variable, depending upon the type of facility improvement.
Transportation facilities and policies also affect the automobile's
share of total trips, which leads to additional emission impacts.
In addition to emission summaries, the Air Pollution Control District
receives gridded emission files for subsequent air pollutant
dispersion modeling.
DATA FILE
DOCUMENTATION
EMISSION FACTORS (INPUT)
Emission factor files are located under "/max7/data/aq." Sub-
directories are created for each new version of EMFAC, and emission
factors associated with that version are located under the sub-
directory. Files containing factors for non-commercial, personal
vehicles have names that begin with "per;" followed by a "y" or "n"
indicating whether the emission factors contain the effects of
inspection and maintenance programs; followed by the final two digits
of the year that the emission factors represent. Files for commercial
truck factors begin with "trk" and files for bus factors begin with
"bus."
Emission factor files are ASCII files with two records for each
temperature and speed category within a process. Formats change
slightly with each new version of EMFAC. The following formats are
for EMFAC7F factors.
RECORD 1
Columns Variable Type Description
1-1 I1 "9"
2-2 I2 Process, where:
C = Indicates Cold Start
H = Indicates Hot Start
I = Indicates Hot Stabilized Transient
(Not Used)
K = Hot Soak
L = Indicates Diurnal and Resting
Losses
R = Running Exhaust
3-3 I1 "1"
4-4 I1 Technology, where:
1 = Catalyst Vehicle
2 = Non-Catalyst Vehicle
5-10 I6 Dewpoint Temperature (30 Degrees
Fahrenheit)
11-15 I5 Mid-Point of Temperature Range (Degrees
Fahrenheit)
16-20 I5 Mid-Point of Speed Range (Miles per Hour)
21-29 F9.5 TOG Non-Catalyst Exhaust Factor
(Grams/Hour)
30-38 F9.5 TOG Catalyst Exhaust Factor (Grams/Hour)
39-47 F9.5 TOG Diesel Exhaust Factor (Grams/Hour)
48-56 F9.5 TOG Non-Catalyst Evaporative Factor
(Grams/Hour)
57-65 F9.5 TOG Catalyst Evaporative Factor
(Grams/Hour)
66-74 F9.5 TOG Diesel Evaporative Factor (Grams/Hour)
75-83 F9.5 TOG Total Factor (Grams/Hour)
Columns Variable Type Description
84-92 F9.5 NOx Non-Catalyst Factor (Grams/Hour)
93-101 F9.5 CO Catalyst Factor (Grams/Hour)
102-110 F9.5 Exhaust PM Factor (0.0 Grams/Hour)
111-119 F9.5 Tire Wear PM Factor (0.0 Grams/Hour)
120-128 F9.5 Fuel Consumption Factor (0.0 Gallons/Hour)
RECORD 2
Columns Variable Type Description
21-29 F9.5 TOG Total Evaporative Emission Factor (0.0
Grams/Hour)
TEMPERATURE FILES (INPUT)
Temperature files are located under "/max7/data/aq" and are called
"temp.nnn," where "nnn" indicates the julian calendar date that the
temperatures represent. Temperature files are ASCII files with one
record for each air quality modeling grid. Temperatures are in
degrees kelvin.
Columns Variable Type Description
1-12 F12.3 State Plane "Y" Coordinate (Feet)
12-21 F9.0 State Plane "X" Coordinate (Feet)
22-25 I4 Grid Number
26-36 G11.0 Temperature for Midnight to 1:00 AM
37-48 G12.0 Temperature for 1:00 AM to 2:00 AM
49-60 G12.0 Temperature for 2:00 AM to 3:00 AM
61-72 G12.0 Temperature for 3:00 AM to 4:00 AM
73-84 G12.0 Temperature for 4:00 AM to 5:00 AM
85-96 G12.0 Temperature for 5:00 AM to 6:00 AM
97-108 G12.0 Temperature for 6:00 AM to 7:00 AM
109-120 G12.0 Temperature for 7:00 AM to 8:00 AM
121-132 G12.0 Temperature for 8:00 AM to 9:00 AM
133-144 G12.0 Temperature for 9:00 AM to 10:00 AM
144-156 G12.0 Temperature for 10:00 AM to 11:00 AM
157-168 G12.0 Temperature for 11:00 AM to Noon
169-180 G12.0 Temperature for Noon to 1:00 PM
181-192 G12.0 Temperature for 1:00 PM to 2:00 PM
193-204 G12.0 Temperature for 2:00 PM to 3:00 PM
205-216 G12.0 Temperature for 3:00 PM to 4:00 PM
217-228 G12.0 Temperature for 4:00 PM to 5:00 PM
229-240 G12.0 Temperature for 5:00 PM to 6:00 PM
241-252 G12.0 Temperature for 6:00 PM to 7:00 PM
253-264 G12.0 Temperature for 7:00 PM to 9:00 PM
265-276 G12.0 Temperature for 8:00 PM to 9:00 PM
277-288 G12.0 Temperature for 9:00 PM to 10:00 PM
289-300 G12.0 Temperature for 10:00 PM to 11:00 PM
301-312 G12.0 Temperature for 11:00 PM to Midnight
START DATA FILE (INPUT)
A file called "start.data" located under "/max7/data/aq" contains
vehicle trip start percentages and cold start percentages by hour of
the day. Start data files are ASCII files with one record for each
trip type, trip end type, and hour combination. Series 7 trip type
definitions are used for emission modeling.
Columns Variable Type Description
1-5 I5 Trip Type, where:
1 = Home-Work
2 = Home-Shop
3 = Home-Other
4 = Work-Other
5 = Other-Other
6-10 I5 Trip End Type, where:
1 = Production End
2 = Attraction End
11-15 I5 Hour of the Day
16-20 F5.1 Percent of Daily Trip Ends Starting in
Hour
21-25 F5.1 Percent of Trip Ends Starting in Hour With
One Hour or More Engine-Off Time
26-30 F5.1 Percent of Trip Ends Starting in Hour With
Four Hours or More Engine-Off Time
PERCENT OF TOTAL TRIP ENDS STARTING IN HOUR
Click HERE for graphic.
PERCENT OF TRIP STARTS WITH ENGINE-OFF TIME OF ONE OR MORE HOURS
Click HERE for graphic.
PERCENT OF TRIP STARTS WITH ENGINE-OFF TIME OF FOUR OR MORE HOURS
Click HERE for graphic.
STOP DATA FILE (INPUT)
A file called "stop.data" located under "/max7/data/aq" contains
vehicle trip stop percentages, weighted percent of one hour hot soak
emissions, and average parking duration by hour of the day. Stop data
files are ASCII files with one record for each trip type, trip end
type, and hour combination. Series 7 trip type definitions are used
for emission modeling.
Columns Variable Type Description
1-5 I5 Trip Type, where:
1 = Home-Work
2 = Home-Shop
3 = Home-Other
4 = Work-Other
5 = Other-Other
6-10 I5 Trip End Type, where:
1 = Production End
2 = Attraction End
11-15 I5 Hour of the Day
16-20 F5.1 Percent of Daily Trip Ends Stopping in
Hour
21-25 F5.1 Percent of One Hour Hot Soak Emissions for
Trip Ends Stopping in Hour
26-30 F5.1 Average Parking Duration of Trip Ends
Stopping in Hour (Minutes)
PERCENT OF TOTAL TRIP ENDS STOPPING IN HOUR
Click HERE for graphic.
WEIGHTED PERCENT OF ONE HOUR HOT SOAK EMISSIONS
Click HERE for graphic.
AVERAGE PARKING DURATION AT TRIP DESTINATION (MINUTES)
Click HERE for graphic.
VMT DATA FILE (INPUT)
A file called "vmt.data" located under "/max7/data/aq" contains non-
truck vehicle miles of travel percentages by hour of the day. Travel
models produce off-peak period and peak period VMT estimates. VMT
data files split VMT for a time period into hourly time slices. Data
files are ASCII files with one record for each hour.
Columns Variable Type Description
1-5 I5 Hour of the Day
16-20 F5.1 Percent of VMT for Time Period in Hour
HOURLY DISTRIBUTION OF VMT BY TIME PERIOD
VMT
Hour Percentage
M-1 1.4%
1-2 1.2%
2-3 1.1%
3-4 1.2%
4-5 1.9%
5-6 5.1%
6-7 15.2%
7-8 16.7%
8-9 11.9%
9-10 8.4%
10-11 8.9%
11-N 10.7%
N-1 9.7%
1-2 10.6%
2-3 12.3%
3-4 18.8%
4-5 21.4%
5-6 16.0%
6-7 8.2%
7-8 5.6%
8-9 4.8%
9-10 3.9%
10-11 2.8%
11-M 2.1%
HOURLY DISTRIBUTION OF DAILY TRUCK VMT
VMT
Hour Percentage
M-1 0.7%
1-2 1.0%
2-3 0.9%
3-4 1.0%
4-5 1.1%
5-6 2.0%
6-7 5.1%
7-8 7.0%
8-9 7.7%
9-10 8.5%
10-11 8.6%
11-N 8.1%
N-1 7.3%
1-2 7.5%
2-3 7.7%
3-4 6.9%
4-5 4.8%
5-6 3.3%
6-7 3.0%
7-8 2.2%
8-9 1.7%
9-10 1.8%
10-11 1.1%
11-M 0.8%
ZONE-GRID FILE (INPUT)
A zone to air quality modeling grid conversion file called "zone.grid"
is located under "/max7/data/aq." The zone-grid file is an ASCII file
with one record for each zone-grid combination.
Columns Variable Type Description
1-6 I6 Zone Number
7-11 I5 Grid Number
12-24 F13.3 Percent of Zone Area in Grid
VEHICLE TRIP END BY GRID FILES (INTERMEDIATE)
Summaries of daily vehicle trip ends by trip type and air quality
modeling grids are produced from mode choice model results. The files
are called "pagrid" and are located under temporary workspaces created
to evaluate alternatives. Trip end files are in ASCII format with one
record for each grid with trips. Another file, "mpagrid," has the
same format but contains San Diego trips ends of trips by Mexican
vehicles from the external trip forecasting process.
Columns Variable Type Description
1-4 I4 Grid Number
5-14 F10.0 Home-Work Productions
15-24 F10.0 Home-Shop Productions
25-34 F10.0 Home-Other Productions
35-44 F10.0 Work-Other Productions
45-54 F10.0 Other-Other Productions
55-64 F10.0 Home-Work Attractions
65-74 F10.0 Home-Shop Attractions
75-84 F10.0 Home-Other Attractions
85-94 F10.0 Work-Other Attractions
95-104 F10.0 Other-Other Attractions
HIGHWAY LINK-GRID FILE (INTERMEDIATE)
Highway link to air quality modeling grid conversion files called
"hwycovlb.grid" are located under temporary workspaces created to
evaluate alternatives. Highway link-grid files are ASCII files with
one record for each link-grid combination.
Columns Variable Type Description
1-6 I6 Arc/Info Internal Highway Arc ID Number
7-11 I5 Grid Number
12-24 F13.3 Link Length in Grid (Feet)
TRANSIT BUS VEHICLE HOURS OF TRAVEL FILE (INTERMEDIATE)
Summaries of transit bus vehicle hours of travel (VHT) by time period
and air quality modeling grid, called "busgrid," are located under
temporary workspaces created to evaluate alternatives. Bus VHT files
are created by totalling information from coded transit networks. The
files are in ASCII format with one record for each grid with transit
service.
Columns Variable Type Description
1-5 I5 Grid Number
6-10 I5 Average Bus Speed (Miles per Hour)
11-20 F10.3 Peak Period Bus VHT in Grid
21-30 F10.3 Off-Peak Period Bus VHT in Grid
EMISSION FILES (OUTPUT)
Summaries of motor vehicle emissions by grid, hour and type of
pollutant are created by the emission modeling process and transferred
to the Air Pollution Control District for subsequent use in air
pollution dispersion models. Files are located under temporary
workspaces created to evaluate alternatives. Files are in ASCII
format with one record for each grid-hour combination with emissions.
Columns Variable Type Description
1-5 I5 Grid Number
6-10 I5 Hour
11-20 E10.3 TOG Cold Start Non-Catalyst Emissions
(Grams)
21-30 E10.3 TOG Cold Start Catalyst Emissions (Grams)
31-40 E10.3 TOG Cold Start Diesel Emissions (Grams)
41-50 E10.3 TOG Hot Start Non-Catalyst Emissions
(Grams)
51-60 E10.3 TOG Hot Start Catalyst Emissions (Grams)
61-70 E10.3 TOG Hot Start Diesel Emissions (Grams)
71-80 E10.3 TOG Hot Soak Non-Catalyst Emissions
(Grams)
81-90 E10.3 TOG Hot Soak Catalyst Emissions (Grams)
91-100 E10.3 TOG Diurnal Non-Catalyst Emissions (Grams)
101-110 E10.3 TOG Diurnal Catalyst Emissions (Grams)
111-120 E10.3 TOG Stabilized Running Non-Catalyst
Emissions (Grams)
121-130 E10.3 TOG Stabilized Running Catalyst Emissions
(Grams)
131-140 E10.3 TOG Stabilized Running Diesel Emissions
(Grams)
141-150 E10.3 TOG Running Loss Non-Catalyst Emissions
(Grams)
151-160 E10.3 TOG Running Loss Catalyst Emissions
(Grams)
161-170 E10.3 CO Total Emissions (Grams)
171-180 E10.3 NOx Total Emissions (Grams)
