NCHRP RPT 365

NeTIoNAL CooPERATIVB HIGHWAY RsEARCH PnoGRAM
Report 365
Travel Estimation Techniques

for Urban Planning
WILLIAM A. MARTIN
NANCY A. MCGUCKIN
Barton-Aschman Associates, lnc.
Washington, DC
Subject Areas
Plannng and Admnistration
Research Sponsored by the American Association of State

Highway and Transportation Officials in Cooperation with the
Federal Highway Administration
TRnsponrATtoN ResEnRcn Boano

Nalor'ru Resencx Cour,cl-
NATIONAL ACADEMY PRESS

Washington, D.C. 7998
NATIONAL COOPERATIVE HIGHWAY RESEARCH NCHRP REPORT 365
PROGRAM
Systematic, well-designed research provides the most effective Project 8-29(2) FY '95
approach to the solution of many problems facing highway
tssN 0077-5614
.l
'i administrators and engineers. Often, highway problems are of local
interest and can best be studied by highway departments rsBN 0-309-05365-X
individually or in cooperation with their state universities and
L. C. Catalog Card No.94-78144
others. However, the accelerating growth of highway transportation
develops increasingly complex problems of wide interest to O 1998 Transportation Research Board
highway authorities. These problems are best studied through a

coordinated program of cooperative esearch. Price $41.00
In recognition of these needs, the highway administrators of the
American Association of State Highway and Transportation
Officials initiated in 1962 an objective national highway research
program employing modern scientific techniques. This program is
supported on a continuing basis by funds from participating
member states of the Association and it receives the full cooperation
and support of the Federal Highway Administration, United States NOTICE
Department of Transportation.
The project that is the subject of tbis report was a part of the National Cooperative
The Transportat-ion Research Board of the National Research
Highway Research Program conducted by the Trnsportation Research Board with the
Council was requested by the Association to administer the reseach approval of the Governing Board of the National Reseach Council. Such approval
program because of the Board's recognized objectivity and
--- :|
understanding of modern research practices. The Board is uniquely
reflects the Goveming Board's judgment that the program concerned is of national
importance and appropriate with respect to both the purposes and resources of the
suited for this purpose as it maintains an extensive committee National Research Council.
structure from which authorities on any highway transportation
The members of the technical committee selected to monitor this prcject and to review
subject may be drawn; it possesses avenues of communications and this report were chosen for recognized scholarly competence and with due
cooperation with federal, state and local governmental agencies, consideration for the balance ofdisciplines appropriate to the project. The opinions and
universities, and industry; its relationship to the National Research conclusions expressed or implied are those of the research agency that pertbrmed the
Council is an insurance of objectivity; it maintains a full-time research, and, while they have been accepted as pproprite by the technical committee,
research correlation staff of specialists in highway transportation they are not necessarily those of the Transportation Research Board, the National
matters to bring the findings of research directly to those who are in Reseach Council, the American Associaton of State Highway and Transportation
a position to use them. Officials, o the Federal Highway Administration, U.S. DePartment ofTransportation.
The program is developed on the basis of research needs Each report is eviewed and accepted for publication by the technical committee
identifled by chief administrators of the highway and transportation according to procedures estabished and monitored by the Transportation Research
departments and by committeei of AASHTO. Each year, specifrc Board Executive Committee and the Goveming Board of the National Research
areas of resech needs to be included in the program are proposed Council.
to the National Research Council and the Board by the American

Association of
State Highway and Transportation Officials.
fulfill these needs are defined by the Board, and
Research projects to
qualified research agencies are selected from those that have
submitted proposals. Administration and surveillance of research
contracts are the responsibilities of the National Research Council
and the Transportation Research Board. Published reports of the
The needs for highway research are many, and the National
NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM
Cooperative Highway Research Program can make signiflcant
contributions to the solution of highway transportation problems of are available from:
mutual concern to many responsible groups. The program,
Transportation Research Board
however, is intended to complement rather than to substitute for or
National Research Council
duplicate other highway research programs.
2101 Constitution Avenue, N.W.
Washington, D.C. 20418
and can be ordered through the Internet at:

Note: The Transportation Research Board, the National Research Counc|,
the Federal Highway Administration, lhe American Association of State http://www.nas.edu/trb/index,html
Highway and Transportaton Officials, and the Individual states participating n
the National Cooperative Highway Research Program do not endorse products
or manufacturers. Trade or manufacturers' names appear herein solely
because they are considered essential to the object of this report. Printed in the United States of America
FOREWORD This report updates NCHRP Report 187, "Quick-Response urban Travel Estima-
tion Techniques and Transferable Parameters." It provides a thorough review of the
By Staff four-step travel demand process and transferable parameters that can be used in simple
Transportation Res earch planning analyses. It will be particularly useful to planners in smaller urban areas that
Board cannot afford to develop area-specific parameters.
In 1978, the Transportation Research Board published NCHRP Report tB7,

"Quick-Response Urban Travel Estimation Techniques and Transferable Parameters."
That report described transferable parameters, factors, and manual techniques for a
simple planning analysis. This report and its default data have been used widely, in one
form or another, in many transportation studies. The report has been an invaluable
travel-data source. However, the manual techniques have been largely supplanted by
microcomputer planning models, and the parameters and factors are based on data from
the 1960s and early 1970s.
Under NCHRP Project 8-29, Barton-Aschman Associates, Inc., updated the travel
demand estimation techniques and parameters presented in NCHRP Report 187 using
more current travel survey procedures and data. To provide the most reliable informa-
tion to practitioners, the Federal Highway Administration provided funds for a follow-
on effort, NCHRP Project 8-29(2),In this project, Barton-Aschman Associates, Inc.,
collected additional data to validate the trip-generation rates and trip-distribution
friction factors developed in the initial project.
In addition to a thorough review of the four-step travel demand process with com-
mon extensions, the report provides transferable parameters for use when area-specific
data are not available or need to be checked for reasonableness. The material focuses
primarily on the needs of smaller urban areas, but some material will be useful to other
areas. In general, more complex procedures will be needed for large urban areas, grow-
ing medium-sized urban areas, and severe air quality nonattainment areas. Area-
specific parameters will almost always be preferable to transferred parameters, though
it may not be cost-effective to develop them for smaller urban areas. The techniques
and parameters are organized to be easy to use in many of the widely available travel
demand forecasting programs. A case study illustrates how the techniques and param-
eters can be applied in a typical study.
Those interested in looking more deeply into transferable parameters should visit
the 1995 Nationwide Personal rransportation survey world wide website at
http://www-cta.ornl.gov/npts. This website allows anyone to develop parameters like
those in this report based on the 1995 NPTS data.
- t-
;''
:'tt
,j' :
't,
CONTENTS 1 SUMMARY
4 CHAPTER 1 Introduction
Purpose and Organization of Manual, 4
Travel-Demand Forecasting: Trends and Issues, 4
Areas of Development, 5
The Four-Step Travel-Demand Forecasting Process, 5
Model Iteration and Equilibrium, 6
Summary of Techniques and Parameters, 7
Trip Generation, 7
Trip Distribution, 7
External Trip Estimation, 8
Mode-Choice Analysis, 8
Automobile-Occupancy Characteistics, 8
Time-of-Day Characteristics, 8
Traffic Assignment, 8
Capacity Analysis, 9
Development Density/Highway Spacing Relationships, 9
Data Sources, 9
Case Study, 9
Summary,9
10 CHAPTER 2 Building a Transportation Database
Introduction, 10
Network Data, 10
Base Network, 10
Source for Network Data, 10
Digitized Map Files, 10
Scaled Maps, 11
Network Coverage, 11
Network Attributes, 12
Link Distance, 12
Link Speeds, 12
Link Capactty,12
Area Type Considerations, 13
Final Network Database, 13
Transit Networks, 13
Socioeconomic Data, 13
Structure of the TAZs, 13
Relationship to Census Geography, 14
TAZ and Highway Network Comparability, 14
Sources for SocioeconomicData, 14
Data Source Deficiencies, 15
Control Totals for the Database, 15
Impact and Uses of GIS on Databases, 15
Geocoding, l5
Cartography, l6
Case Study, 16
Network Description, 18
Traffic Count Data, 18
Summary, 19
20 CHAPTER 3 Trip Generation
Introduction, 20
Basis for Development, 20
Site-Specific Vehicle Trip Rates, 20
Model Trip Rates,22
Data Required for Application, 24
Site-Specifrc Vehicle Trip Rates, 24
Model Trip Rates, 24
Trip Generation Data and Examples of Use, 27
Site-Specifi c Vehicle T rips, 27
Model Trip Rates, 28
Balancing Productions and Attractions, 32
Limitation of Data, 33
Case Study, 34
Trip Productions, 34
Trip Attractions, 34
36 CHAPTER 4 Trip Distribution

Introduction, 36
- .1 .. Basis for Development, 36
Theory of the Gravity Model, 36
Trip Distribution Process, 38
Travel Impedances, 38
Intrazonal Travel Times, 39
Terminal Times, 39
Choice ofFriction Factors, 39
Average Trip Lengths, 40
Trip-Length Frequency Distribution, 41
Look-Up Tables Versus Formulas, 41
Creation of Trip Tables, 43
Method for Splitting Trip Tables for Corridor and Subaea Applications, 44
Case Study,45
Estimation of Travel Times, 45
Intrazonal Times, 45
i
Terminal Times, 45
:
Choice of Friction Factors, 45

Creation and Assignment of Free-Flow Trip Tables, 46
i
Creation and Assignment of Congested Trip Tables, 46
48 CHAPTER 5 External Travel Estimation

Inoduction,48
Estimation of Through Trips at External Stations, 49
Distribution of Through Trips Between Stations,50
Example of Through-Trip Table Estimation, 51
Estimation of Extenal-Internal Trip Productions ad Attractions, 53
Distribution of Internal-External and External-Internal Trips, 56
Case Study, 56
Classification of External Stations, 57
Estimation of Through-Trip Percentages, 57
Distribution of Through Trips to External-External Trip Table, 57
Conversion of Internal-External Trips to Person-Trip Productions
and Attractions, 59
Summary,60
62 CHAPTER 6 Mode-Choice Analysis

Introduction, 62
Logit Formulation,62
Incremental Logit Formulation, 63
Model Coefficients, 64
Incremental Model Application, 64
Example Application of Incremental Mode-Choice Model, 66
Problem,67
Summary of Example Problem Results, 73
Analysis of Travel Demand Management (TDM) Strategies, 73
Case Study,75
76 CHAPTERT Automobile-OccupancyCharacteristics
Introduction, 76
Features and Limitations, 77
Variation in Auto Occupancy by Urbanized Area Population and by Trip
Purpose, TT
Variation in Auto Occupancy by Time of Day,77
Variation in Auto Occupancy by Income Level,'17
Variation in Auto Occupancy by Facility Type,79
Data Requirements and Example Problems, 80
Usefulness of Regionally Deveoped Models, 80
Case Study, 80
81 CHAPTER 8 Time-of-Day Characteristics

Introduction,8l
Vehicle Travel, 82
Diunal Distribution, 82
Use of Time-of-Day Tables, 82
Transit Hourly Distributions, 84
Trip Matrix Convesion Factors, 85
Creation of an Origin-Desrination Trip Table, 87
Case Study, 9l
93 CHAPTER 9 Traffic Assignment Procedures

Introduction, 93
Traffic Assignment Model Parameters, 94
Application of Volume-Delay Curves in Highway Assignment, 94
Node Characteristics, 96
Validation of Highway Assignment and Network, 96
Distribution of Assigned Volumes Among Available Facilities, 96
Traffic Shift Methodology for Corridors, 100
Case Study, 103
104 CHAPTER 10 Capacity Analysis

Introduction, 104
Initial Settings for Capacities for Use in Travel Forecasting Models, 105
Assumptions and Extensions for Initial Capacity, 105
Determination of Intersection Capacity, 107
Input Data Requirements, 107
Worksheet Applications, I 08
Computational Requirements, I 08
Instructions for the Lane Volume Worksheet, 109
Signal Operations Worksheet, I l4
Limitations of the Planning Method, 117
Illustrated Example, I 17
123 CHAPTER 11 Development DensityIighway Spacing Relationships
Introduction, 123
Data Required for Application, 124
Features and Limitations, 124
Limitations of the Methodology and Substitutability of Local Data, 124
Applying the Developmenr Densiry/Highway Spacing Methodology, 125
Steps in Application, 126
Feedback,131
An Example Application, 132
Input Information, l32
Methodology, 133
Output Information, 135
136 CHPTER 12 Case Study Application of Default Parameters

Introduction, 1 36
Transportation Database, I 36
Network Description, 138
Traffic Count Data, 138
Trip Generation, 138
Trip Productions, 138
j' '
+j
Trip Attractions, 139
External Travel, 140
Classification of External Stations, 140
Estimation of Through-Trip Percentages, 140
Distribution of Through Trips to External-Extemal Trip Table' l4l
Conversion of Internal-External Trips to Person-Trip Productions
and Attractions, 142
Trip Distribution, 144
Estimation of Free-Flow Travel Times, 144
Intrazonal Times, 145
Terminal Times, 145
Choice ofFriction Factors, 146
Creation of Free-Flow Trip Tables, 146
Mode-Choice AnalYsis, 147
Automobile-Occupancy Characteristics, 147
Time-of-Day Characteristics, 1 47
Traffic Assignment of Free-Flow Trip Table' 148
Creation and Assignment of Congested Trip Tables' 148
Model Validation and Screenline Comparisons, 148
Conclusions, 150
152 BIBLIOGRAPHY
'-l 154 APPENDIX A NTPS and Home Interview Survey Data

i
',
:I
157 APPENDIX B Asheville, North Carolina, Case Study
,]
I
'r: -_ l
_l
AUTHOF ACKNOWLEDGMENTS
The esearch reported herein was performed under NCHRP nel who contributed significantly to this report were David Kurth,
Project 8-29 by Barton-Aschman Associates, Inc. Principal Associate; Richard Donnelly, Principal Associate; and
William A. Martin, Principal Associate and John R. Hamburg, Eric Bierce, Senior Associate.
Vice President, were the principal investigators. The other authors Sincerest thanks are extended to all the agencies and indiv-
of this report are Nancy A. McGuckin, Senior Associate; Hugh iduals who contributed time and effort in responding to interviews
Morris, Associate; Shawn Sabanayagam, Senior Associate; Thomas and contributing local data for the parameters set forth in this
Harrington, Associate; and Matthew Lawor, Intern. Other person- report.
..
."i .1
CHAPTER 1
INTRODUCTON
PURPOSE AND ORGANIZATION OF MANUAL TRAVEL-DEMAND FORECASTING :

TRENDS AND ISSUES
This report updates NCIL/RP Report 187, "Quick-Response
Urban Travel Estimation Techniques and Transferable Pa- The practice of travel-demand forecasting is roughly 35
rameters," published by the Transportation Research Board years old. Travel-demand forecasting started in the United
in 1978. Like that guide, this report is organized to follow the States with areawide transportation studies in Chicago and
traditional travel-demand forecasting steps of trip generation, Detroit. Since then, the practice has progressed through vari-
trip distribution, mode choice, and traffic assignment. Unlike ous schools of thought, while the advent of the microcomputer
the earlier report, this report does not give manual techniques has dramatically changed the environment in which such
for applying the travel procedures significant attention. The analyses are carried out.
parameters presented are organized in a format that allows for The early impetus for the development of travel-demand
application in many of the widely available travel-demand models was to provide an objective tool for evaluating major
forecasting pro&rams that run on microcomputers. infrastructure investments and preparing long-range, regional
The report provides transferable parameters for use when transportation plans. These travel forecasts were produced
area-specific data are not available or need to be checked with mainframe software underwritten by the federal gov-
for reasonableness. The material focuses primarily on the emment. The shift from large, cumbersome computer models
needs of smaller urban areas, but some material will be use- to quick, manual techniques was advocated by many profes-
ful to other areas. In general, more complex.procedures will sionals, especially for planning in smaller urban areas. The
be needed for large urban areas, growing medium-sized best-known approach for such quick-response techniques
urban areas, and severe air quality nonattainment areas. was embodied in NCHRP Report 187.
Area-specific parameters will almost always be preferable to Perhaps the most significant development in the field of
transferred parameters, though it may not be cost-effective travel-demand forecasting has been the microcomputer revo-
to develop them for smaller urban areas. lution, which has brought desktop computer power to all
In addition to the chapters devoted to the steps in travel- transportation agencies. With few exceptions, these packages
demand forecasting, a chapter is presented that identifies data rival their mainframe and minicomputer counterparts in fea-
requirements and sources of data for building the travel- tures, quality, and performance. Because of the wide extent of
demand database. The data requirements and data sources microcomputer applications in travel-demand forecasting, the
include both the transportation supply system (highway and focus of this manual is on travel parameters and their uses,
transit networks) and socioeconomic data. The remaining rather than on packaged techniques for applying the models.
chapters of this report are as follows: Just as the availability of travel-demand forecasting tech-
niques has been dramatically increased by the microcomputer,
. Chapter 2, Building a Transportation Database; the demands placed on the results of travel-demand forecast-
. Chapter 3, Trip Generation; ing also have increased significantly. No longer is the regional
. Chapter 4, Trip Distribution; travel-demand model run every few years to update the
. Chapter 5, External Travel Estimation; regional transportation plan. Today, daily demands placed on
. Chapter 6, Mode-Choice Analysis; travel-demand forecasting include
. Chapter 7, Automobile-Occupancy Characteristics;
. Chapter 8, Time-of-Day Characteristics; . Project-level studies requiring hourly volumes used in
. Chapter 9, Traffic Assignment Procedures; geometric design;
. Chapter 10, Capacity Analysis; . Subarea traffic circulation studies requiring peak-hour
. Chapter 11, Development Density/Highway Spacing (period) tuming movements;
Relationships; and . Feasibility analysis of public transportation investments
. Chapter 12, Case Study Application of Default Param- (e.g., sketch-planning ridership estimates of light rail,
eters. busway, and commuter rail systems);
3
procedure, the assumption is made that the traffic assignment is done by a travel
demand package and that the default parameters required are the relationship of travel
time to volume and capacity. Different functions are presented for different facility
types, including freeways and multilane arterials. The corridor diversion and screen-
Iine smoothing techniques presented in NC/1RP Report 187 arepresented in this report.
Capacity analysis is presented from two views-the analysis of intersection level of
service and the development of link capacities that can be used as input to the building
of the highway network in the modeling process. The revised r9g5 Highway capacity
Marunl was the source for the intersection procedures and the user is referred to that
document as the primary source for applications.
A case study has been developed in order to illustrate the application of the parame-
ters and techniques described in this report. The data included in this case study were
provided by the State of North Carolina for the City of Asheville, North Carolina. The
applications of the study parameters and techniques to this case study are presented at
the conclusions of Chapters I through 9. This case study allows the user to follow the
development and application of the travel forecasting model beginning with the data
collection phase, where the highway networks and the socioeconomic data are pre-
sented. Subsequent chapters follow the model development through trip generation,
where standardized trip generation rates are applied, and trip distribution, where stan-
dardized friction factors are applied. Ultimately, the final traffic assignment is pre-
sented, along with screenline comparisons of the existing traffic counts to the model
results.
The final chapter of this report presents the case study in its entirety, from data col-
lection through traffic assignment. As can be seen in this report, the results of this
demonstration are quite reasonable when compared with the observed traffic volumes.
2
For trip generation analysis, two sets of parameters arc presented. The first set rep-
resents vehicle trips generated by specific site activities. The data for these rates were
extracted from the Institute of Traffic Engineers' Trip Generation ManuaL,5th Edition.
The second set of rates are typical for trip production and attraction models. These rates
represent average daily person trips and were arrived at by using both the data from the
recent National Personal Transportation Survey and several home interview surveys
taken since 1985. An interesting finding in this study is that, although the trip rates are
divided by urban areas with different populations, the variation between small and large
urban areas was not as great as presented in the NC/1R P Report 187. The rates are more
closely grouped around an average of 9.0 daily person trips per household. Different
rares are presented for the population ranges of 50,000 to 200,000; 200,000 to 500,000;
500,000 to 1,000,000; and greater than 1,000,000.
The trip distribution section presents the standard gravity model formulation. The
report assumes that the user will be developing the zone-to-zone travel times from a
network-based travel demand package and the default data required are the travel
impedance friction factors. The friction factors are presented as both a gamma function
and a lookup table. Presented in this section is a discussion of how the gravity model
can be calibrated to match observed trip length distributions. Unlike the trip generation
section, the default friction factors are not grouped by urban area size. The trip distri-
bution within an urban area depends heavily on both the local highway (and transit net-
work for areas with significant transit shares) network and the geographic location of
the households and employment.
External travel estimation has been the least documented component of the travel
demand models. A chapter has been included that presents a procedure for estimating
through, intemal-external, and external-internal trips for small urban areas. The research
concluded that, although the procedure works adequately for small urban areas, it is not
applicable for larger areas. Research into external travel revealed that very little has been
done in the advancement of external travel estimation, yet for many reas, external travel
is not easily transferable between urban areas-particularly through-trip estimation,
which depends heavily on the urban area location (in relation to other urban areas) and
size. It is recommended that local external travel data be collected to the extent possible
and that further research is needed into the collection and estimation of external travel.
Mode choice is a step in the modeling process that has been largely ignored in small-
to medium-sized urban areas. These sizes of areas have transit systems that carry a
small percentage of total person trips and the data have not been collected from which
a locally calibrated model can be derived. Advances in mode choice modeling have
largely been tied to the analysis of major investments in fixed-guideway transit systems
such as new light rail starts. The chapter on mode choice presents a discussion of the
logit formulation of the mode choice model as well as a presentation of the incremen-
tal logit model structure that can be used for the analysis of transit and HOV alterna-
tives in corridors and subareas. It is in these analysis areas that mode choice analysis
using transferable parameters is most applicable.
Auto occupancy and time-of-day pa-rameters are presented in separate chapters. The
research resulted in a conclusion contradictory to the NCHRP Report 187 assumption
that auto occupancy would increase. During the last decade, auto occupancy has actu-
ally been declining. This is reflected in an average occupancy rate for home-based work
trips of about 1.11 versus the NCHRP Report 187 rate of about 1.35. The time-of-day
factors can be used to construct trip tables for peak and off-peak periods. Also included
in the time-of-day chapter is a discussion of procedures for converting production-
attraction formatted tables to origin-destination tables.
The traffic assignment chapter presents refinements to the standard BPR formula-
tion for travel times as a function of volume and capacity. As with the trip distribution
TRAVEL ESTIMATION TCI-INIQI.J ES
FOR URBANI PLANNING
SUMMARY This project was conducted in two phases. The first phase was to identify the critical
travel estimation areas that would require updating or adding to the earlier NCHRp
Report 187.The second phase was to collect the necessary data, update the travel esti-
mation parameters and techniques, and prepare the revised report. During the fust phase
of this project a survey was conducted of the metropolitan planning organizations
(MPOs) and state DOTs with the objective of identifying
. If and how NCI1RP Report 187 was used,

' What issues the transpofation planners are facing that place added demands on the
travel demand model, and
. Any travel surveys that the agency had conducted in recent years.
The survey found that the great majority of applications of the quick response tech-
niques and parameters were for trip generation, either the site-specific vehicle rates or
the general household-based trip production models. The mode choice procedures con-
tained in NCHRP Report 187 hadnever been used by almost 90 percent of the respon-
dents and only 3 percent of the respondents are still using the model choice technique.
With the rapid growth in the capacity and deployment of microcomputers, the use of
manual application techniques has been minimized. This report concentrates on travel
parameters that can be applied in any of the available travel demand programs. The
extensive, non-network-based, manual procedures (e.g., trip distribution and traffic
assignment) contained in the earlier report are not included in this report. The travel
parameters and techniques presented in this report follow the basic four-step process:
trip generation, trip distribution, mode choice, and traffic assignment.
A chapter has been included that discusses the databases required to build a travel
demand model. These include supply-side data (e.g., highway and transit networks) and
demand-side data (e.g., zonal socioeconomic data on population and employment). A
description of the data requirements is presented along with sources for building the
database. Also in this chapter is a brief discussion of the use of geographic information
systems (GIS) and the opportunities for using GIS in the building of the travel demand
database and in model application. The survey of MPOs and state DOTs revealed that
more than one-half of the agencies have GIS available.
5
. Evaluation of the impacts of transportation investments THE FOUR-STEP TRAVEL-DEMAND

on development levels (that is, the iterative relationship FORECASTING PROCESS
of land use patterns and transportation systems);
. Air quality analysis for both regional conformity analy- Travel-demand forecasting is often referred to as the
sis and localized non-attainment areas (such as intersec- "four-step" process. The steps are: trip generation, trip dis-
tions in non-attainment for CO concentrations); and tribution, mode choice, and assignment. These are the four
. Analysis of travel reduction programs, travel demand major model components of the travel-demand forecasting
management (TDM) strategies, and Congestion Man- process. Other submodels that compose the complete model
agement System plans (as required by the Intermodal set are illustrated in Figure 1.
Surface Transportation Efficiency Act IISTEA] of l99l). The purpose of trip generation estimation (Chapter 3) is to
determine the number of person or vehicle trips to and from
These and other analyses require more detailed results than activities in an analysis area. Trip generation is functionally
the 24-hour volume estimates for major facilities tradition- related to the use of land, which is described in terms of the
ally associated with travel-demand forecasting. Factors that character, intensity, and location of activities. Specific fac-
must now be considered in the travel-demand forecasting tors that influence the number of trips in a region include
process include time-of-day analysis, peak-period spreading, automobile ownersirip, income, household size, density and
automobile-occupancy rates, and feedback mechanisms for type of development, availability of public transportation,
congested speeds and land-use changes. Unfortunately, few and the quality of the transportation system. The best trip-
areas have current, locally generated travel behavior data. generation techniques use disaggregate socioeconomic data,
For these reasons, the parameters in this report give added such as households classified by vehicle ownership, family
attention to trip generation rates, treatment ofexternal travel, size, or income group. This step produces estimates of the
time-of-day parameters, and automobile-occupancy rates. trip productions and trip attractions for each zone in the
analysis area.
Trip distribution (Chapter 4) links rhe rrip producrions to
the trip attractions for each pair of zones in the analysis area.
AREAS OF DEVELOPMENT
The critical factor in trip distribution is the ease of travel
between the two zones being analyzed. This is influenced by
At least two recent, significant developments have af-
fected the travel modeling approach and process. First is the
the distance between the zones and the efficiency of the
use of geographic information systems (GIS) in the forecast-
transportation system linking them.
ing process. GIS allows the user to digest and display data Mode-choice analysis (Chapter 6) is the rhird srep in rhe
relevant to the task at hand. Before the modeling process, traditional four-step travel-demand forecasting process.
GIS may be used to Mode-choice modeling splits the total zone-to-zone person
trips resulting from the trip-distribution model into trips
using each available mode between the zone pair. Mode-
. Map a study area network to determine the level of road- choice modeling is also used to evaluate improvements in
way detail needed for the model, bus systems and analyze high-occupancy vehicle (HOV)
. Batch out the designated network in a format accepted strategies.
by the model, Several simple submodels can be used to refine the esti-
. Map demographic data at census blocVtract level, mates from the first three steps on the basis of analysis needs.
. Convert census blocks/tracts to traffic analysis zones Auto-occupancy estimates (Chapter 7) are used to convert per-
(TAZs), and
son trips to vehicle trips. Time-of-day modeling (Chapter 8)
. Export TAZ structure to form a demarcation file for use produces hourly estimates of travel. External travel estima-
in the model.
tion (Chapter 5) captures those trips that originate or end out-
side the analysis area.
Post-forecasting uses for a GIS include display of model The last of the major steps in the traditional four-step
outputs, such as link volumes, and display of trip ends by TAZ. process is traffic assignment, both for highways and transit
The other significant development has been the changing (Chapter 9). The assignment of trips to the network is the
urban form. Suburban sprawl has changed the travel direc- final output of the modeling process and becomes the basis
tion from simply suburb to central business district (CBD) for validating the model set's ability to replicate observed
travel to suburb to suburb as well. This change has mani- travel in the base year as well as to evaluate transportation
fested itself in increases in automobile ownership as well as improvements in future years.
vehicle miles traveled (VMT). Trip generation rates, how- The simple four-step modeling process has undergone some
ever, have remained relatively stable. It is not necessarily the refinements in an effort to create a process that more accu-
number of trips that have changed, but rather the way those rately reflects the interdependency ofits components. Specif-
trips are made. ically, the introduction of feedback loops in the modeling
'i
Socioeconomie
Data
- Households -
Networks
- Highway -
r, = Auto oocupancy may bo part of mode choice

z = For p6ak perod assgnments
--- = Feedback loop
fEi = Traditional four-step procedure
T---'] = Othr sumodels
V = lnput datra
+ = Common iteration
>> = Desrable iteration
Figure 1. Travel-demand forecasting process.
process allows for dynamic interaction and adjustments to Model lteration and Equilibrium
take place. Instead of simply stepping through the process,
feedback loops allow interaction between virtually every The components of travel-demand forecasting models are
step. For instance, the final step in the modeling process is highly intenelated and require iterations back through the
traffic assignment. Congestion can affect a person's mode of model chain to reach a stable or equilibrium state. The best
travel, so a feedback loop of travel times is inserted into the example of the iterative nature of the models is the use of
mode-choice model that may predict that more people will travel times in the trip-distribution model.
take a transit mode. This loop can be repeated until the two In the initial application of a model, interzonal travel times
steps reach a steady state. Likewise, travel times can affect are not known until the highway and transit network imped-
trip distribution and the base socioeconomic makeup of the ances are calculated (skimmed). However, congested zone-
study area. Land use and household characteristics often to-zone travel times are not known until trip tables are
reflect the condition of the transportation facilities that serve assigned to the networks and a reasonable approximation of
them. The model iteration process is described in the next travel times that reflect traffic volumes is obtained. The
section. accepted procedure is to make an initial estimate of zone-to-
'l
zone travel times (often taken as uncongested or free-flow lishing regional forecasts of population and employment
travel times in the initial estimation of the trip tables) and make the use of a model subordinate to regionally adopted
then iterate the revised travel times produced after the trip forecasts.
tables are assigned to the networks. This procedure is This discussion of the evolution of travel-demand fore-
repeated until there are few or no observed changes in the casting is presented with the intention that the transportation
resultant trip tables. At this point the tables are considered to planner using the parameters contained in this report has
be in equilibrium with the travel times. A possible deficiency some appreciation of the dynamic nature of travel-demand
in the process is the assumption that actual trip ends (trip gen- forecasting and understands that the process is as much a
eration) do not change as travel times change. craft as it is a science.
In a paper, prepared for the National Association of
Regional Councils by Harvey and Deakinr, the lack of itera-
SUMMARY OF TECHNIQUES
tion of various travel-demand model components is identified
AND PARAMETERS
as the most significant weakness in the application of tradi-
tional travel-demand models. The primary purpose of the As noted earlier, the parameters and techniques in this
paper was to evaluate the effects of improved travel-demand manual are presented in chapters that follow the traditional
models on the estimation of environmental impacts of trans- four-step travel-demand forecasting process. Supporting
portation systems, particularly on air quality evaluation. techniques (e.g., the treatment of external travel, time-of-day
There are several opportunities for iterations within the characteristics, and automobile-occupancy rates) are pre-
traditional travel-demand forecasting process. These fall into sented in the order in which they would typically be
two groups: those that are commonly done today and those addressed in the process. For each chapter, an example ofthe
that are desirable and subject to further research. application of the parameters and techniques accompanies a
discussion of the basis for development of the parameters.
Common iteration procedures are as follows:
Trip Generation
. Congested highway travel times and costs resulting from
traffic assignment iterated back into the trip-distribution Trip generation parameters are presented in two formats.
and mode-choice models, The first format presents the vehicle-trip generation rates that
. Transit times (where transit is on the highway network) are commonly used for site-impact analysis and for estima-
and costs iterated back into the mode-choice model, and tion of vehicle-trips from special generators. The source of
. Transit times and costs iterated back to the trip-distribution these rates is the Institute of Transportation Engineers (ITE)
model in the case where travel times are a composite of Trip Generation report,5th Edition (1991). Only a subset of
highway and transit interzonal times and costs (compos- the trip generation rates contained in the ITE manual are
ite impedance). extracted for this report. The user should refer to the com-
plete and most current manual for more detailed categories
of vehicle-trip generation rates. These rates also include
Desirable iteration procedures are as follows:
information on trip rates during peak periods of both the gen-
erator and the background traffic.
. Highway and transit interzonal travel times and costs iter- The second set oftrip generation parameters is presented in
ated back to the zonal socioeconomic data-residential the format of cross-classification trip production and attrac-
and employment location models, tion rates typically used in travel-demand models. These
. Automobile occupancy iterated back into time of day, rates are daily person trips. As with NCIRP Report 187, trip
and purpose parameters are presented. The standard trip purposes
. Highway and transit interzonal travel times and costs used in this manual are: Home-Based-Work (HBW), Home-
iterated back to the automobile-ownership models used Based Non-Work or Home-Based Other (HBO), and Non-
to predict trip-generation rates. Home-Based (NHB).
Included in the trip generation chapter are the submodel
parameters of automobile ownership and household income
Travel time linked to trip-generation and automobile-
distributions. In addition to the trip generation parameters,
ownership models is the least understood. Residential and
balancing regional productions and attractions is discussed.
employment location models have been the subject of con-
siderable research; however, the political realities of estab-
Trip Distribution
lBarton-Aschman Associates, lnc., Development anr) Calbration ofTravel-Denwnd
Models for the Nev Orleans Area, prepared for the Regional Planning Commission,
The trip distribution pa-rameters presented are consistent
Jeffeson, Oreans, St. Bernard, ad St. Bemard Parishes, Louisiana ( I 98 I ). with the standard gravity model input requirements. The
8
manual calculation techniques presented in NCHRP Report sit demand resulting from level-of-service changes to exist-
187 are not included in this report, since microcomputer- ing systems. For medium and small urban area bus systems
based procedures are assumed to be widely available. The that wish to evaluate such changes, the techniques contained
new parameters may be used with the manual techniques in this report are adequate. They are particularly appropriate
when appropriate. for estimating changes in automobile demand resulting from
Average free-flow speeds for freeways, arterials, and col- HOV strategies.
lectors are presented for use in building travel times between
zones. A more detailed discussion of speed, volume, and
Automobile-Occupancy Characteristics
capacity relationships is included in the traffic assignment
chapter. While highway travel time often is presented as the
Automobile-occupancy is usually a post-mode-choice
only method for measuring interzonal impedances, the inclu-
process where average occupancy rates are applied for each
sion of cost in the equation is becoming more widely
trip purpose. In more complex model applications, occu-
accepted. Travel time factor (often referred to as friction fac-
pancy may be incorporated into the mode-choice model. In
tor) curves, look-up tables, and formulas are included.
either case, the result is the conversion of automobile person
trips to vehicle trips. Automobile-occupancy rates are pro-
vided by urban area size, trip purpose, income, and facility
External Trip Estimation
type. Adjustments for occupancy by time of day and purpose
External trips are those trips that have at least one end out- also are presented.
side a study area defined by an encircling cordon line. When
both ends of a trip (origin and destination) are outside the Time-of -Day Characteristics
cordon line, the trip is termed a through trip. The estimation
of external travel requires a count of average daily traffic Time-of-day parameters allow the conversion of daily
(ADT) at each of the highway crossings of the cordon line travel to hourly travel by direction. Two methodologies are
and classication of the ADT into percent automobiles and presented for producing peak-hour traffic volumes:
trucks. The technique first estimates the number of through
tss*+l trips. After these trips are removed from the total external . Post-assignment factoring of daily link volumes, using
.l trips, the remaining trips are distributed to the internal zones hourly and directional factors, and
I
using a gravity model. The percent of through trips at exter- . Post-distribution, preassignment factoring of trip tables
nal cordon stations has been found to be related to the func- by purpose to produce peak-hour, directional origin-
tional classification of the roadway, the connectivity of each destination trip tables.
pair of external stations, the average daily volume at the sta-
tion, the population of the study area, and the classification This report outlines factors and procedures for both methods.
ofvehicle at the external station. Procedures are included for
estimating productions and attractions for internal zones
before the distribution step. Traffic Assignment
The chapter on traffic assignment has three sections. The

Mode-Choice Analysis rst section presents various traffic assignment techniques
that can be applied within most travel-demand software
The mode-choice procedure is based on the incremental packages. These techniques include "all or nothing" iterative
application of a logit model. This is commonly referred to as capacity restraint, incremental capacity restraint, and equi-
the pivot point technique. A discussion ofthe application of librium assignment. In addition to the assignment method-
various mode-choice model structures is presented, along ologies, the relationships ofspeed, volume, and capacity are
with a discussion of HOV analysis and the estimation of discussed, and various equations and relationships are pre-
travel changes resulting from TDM strategies. sented. These include the commonly used Bureau of Public
Mode-choice modeling has been the most researched and Roads (BPR) formula and variations of the formula based on
advanced component of the travel-demand forecasting pro- the 1985 Highway Capacity Manual. Also included is a dis-
cess in recent years. This has been largely driven by the cussion of incorporating intersection delay into the assign-
analysis of major capital investments in mass transit, such as ment process.
fixed-guideway systems. These include busways, HOV facil- The second section presents a procedure for balancing or
ities, light rail, and heavy rail systems. The analysis of these smoothing assigned traffic across a cutline or screenline. The
systems requires the development of locally calibrated third procedure is a sketch-planning process for estimating
travel-demand models, a subject outside the scope of this probable shifts in volumes between competing facilities in a
manual. However, transferable parameters and techniques corridor. This procedure is based on the stochastic assignment
can be used to estimate marginal changes in HOV and tran- process but does not require the use of an origin-destination
9
table. Rather, a diversion function is developed, based on Transportation (US DOT), Federal Highway Administration
known volumes and operating speeds. This function is used (FHWA). More information on the NPTS can be found in US
to estimate shifts between facilities based on improvements DOT report FHWA-PL-92-007, published December 1991.
to one facility. The NPTS uses household interview techniques to collect
data from each household over the telephone. The sample is
drawn from listed and unlisted telephone numbers. Each
Capacity Analysis sampled household is assigned a 24-hour travel day for
which data on all travel by household members is collected.
Correct designation of link capacity is critical to produc-
Each person older than 13 years is asked to report all trips
ing a model that accurately reflects real-world situations. A
taken during the designated travel day. The household mem-
link capacity regulates the volume of travel that can be car-
bers are interviewed within a 6-day period immediately fol-
ried over that link. Thus, for measuring current congestion
lowing the travel day.
levels as well as those forecast, it is necessary to reflect accu-
Data from travel surveys conducted in 1 I cities around the
rate capacities in the network database. Default link capacities
country using home-interview techniques were also used.
are presented, as is the planning procedure for intersection
The trip rates derived from these surveys were used as a
capacity analysis.
cross-check to the NPTS data. For more information on these
data, see Appendix A.
Development Density/Highway
Spacing Relationships
CASE STUDY
The basic purpose of including this manual technique is to
permit the rapid development of a "first-cut" estimate of A case study for the City of Asheville, North Carolina, has
future highway needs, based on a desired level of highway been developed in order to illustrate the application of the
service. Given a distribution within an area of land use, either parameters and techniques described in this report. The appli-
in activities (people, households,jobs) or in acres by type of cation of the study parameters and techniques to this case
use, and given the presence of an existing highway system, study are presented throughout this report, at the conclusion
future vehicle-trip ends are computed and adjusted for ofeach chapter. This case study allows the user to follow the
improved transit service. The average trip distance is then development and application of the travel forecasting model
computed from counts or from trip-length frequency curves beginning with the data collection phase, where the highway
and adjusted for the future. networks and the socioeconomic data are presented. Subse-
Average arterial volumes for a given spacing of freeways quent chapters follow the model development through trip
and arterials can be determined from the computation of generation, where standardized trip generation rates are
vehicle miles of travel (VMT), and the level of service pro- applied, and trip distribution, where standardized friction
vided by each subarea can be computed. A comparison of the factors are applied. Ultimately, in Chapter 9, the final traffic
cor.nputed level of service with a desired level of service indi- assignment is presented, along with screenline comparisons
cates a measure of highway needs for the study area. It is also of the existing traffic counts to the model results.
possible to adjust land use inputs to revise the level of ser-
vice or to compute the amount of additional land use that can
be developed while still maintaining a given level of service.
SUMMARY
Data Sources Although the core of the modeling process has remained
unchanged, many refinements to the process and technolog-
Several sources of data were used to develop the trip gen- ical advances have made the four-step traffic forecasting
eration, trip distribution, automobile occupancy, time of day, process a more intricate and comprehensive process. The
and other parameters in this volume. The primary data source computational power now available to virtually all planning
was the 1990 Nationwide Personal Transportation Survey departments is far superior to that in existence 20 yearc
(NPTS). The NPTS is a periodic survey conducted to obtain ago. The next chapter, Building a Transportation Database,
travel behavior data from a large number of households describes some of the data sources available to the modeler
(21,869). The project is sponsored by the U.S. Department of and tells how to make use of them in modeling.
t0
j
CHAPTER 2
BUI LDING A TRANSPORTATION DATABASE
INTRODUCTION Accurate transportation model calibration and validation

requires that the transportation network represent the same
The basic components of the forecasting model are the time period (year) as the land-use data that will be used to
highway network, the transit network, and socioeconomic estimate travel demand. For instance, if 1990 was specified
data. This chapter describes the network and socioeconomic as the base year to ensure consistency with the 1990 census
data needed and some sources for such data but does not give data, roadway improvements completed after about April
significant detail on ways to build the needed databases. It 1990 should be excluded from the base year network.
reviews the types of information that will be necessary and The process of translating the highway system into a dig-
the ways of obtaining it. The computer programs used to cre- ital format is called network coding. The various segments of
ate the model will give the specific data formats needed. the highway system are represented in transportation models
using two basic data descriptors called links and nodes. Cod-
I
NETWORK DATAI
ing a network requires decisions regarding level of detail
required, the type and amount of information to assemble,
Base Network and the format and limitations of the link and node files in the
software being used.
One of the most important aspects of travel-demand mod-
eling is the method used to represent the transportation sys-
:
tem. The estimation of travel demand requires an accurate Source for Network Data
I
representation ofthe transportation system serving the region.
i
The most direct method is to develop an abstract model of Selection and mapping of roadway links is the first major
the system elements. This is called a network. A network is step in developing a network because links represent those
developed for each travel mode (i.e., automobile and transit). facilities (highways, roads, and streets) that actually consti-
The representation of the automobile system is called a high- tute the highway system. The two nodes that mark the end
way network and includes those streets, roads, thoroughfares, points define the link in the transportation network, and
and freeways that make up the regional highway system. nodes are typically given x and y coordinates to permit the
The network is basically a map of these routes, defined in a plotting of graphic displays. Nodes are locations in the high-
manner that can be read, stored, and manipulated by standard way system where vehicles are able to change direction
transportation planning computer programs. (intersections, interchanges) or where level of service alters
The highway network serves several purposes in trans- significantly (e.9., a road narrows from four lanes to two
portation systems analysis. First, it is an inventory of the lanes). Base data for mapping the roadway network can be
existing road system. It is an official record, for present and compiled from either census or commercially digitized map
future years, of the physical status of the highway system. files or can be digitized from a good scaled map of the plan-
Second, the network is used in traffic modeling to estimate ning region.
the highway travel impedance between zones in the region,
which is the driving time and highway distance between dif-
ferent areas of the planning region. This information is criti-
Digitized Map Files
cal in the trip-distribution and mode-choice portions of the
analysis. Third, the roadway network is used to simulate
The Census Bureau compiles digital data for all 1990 cen-
automobile travel and estimate associated impacts, such as
sus map features (such as roads, railroads, andrivers) and the
pollution, energy use, and accidents.
associated collection geography (such as census tracts and
blocks), political areas (such as cities and townships), and,
IBalon-AschmanAssocites,lnc.,".f. DunkinAssociates,lnc.Manual,preparedfor within metropolitan areas, address ranges and zip codes
the Central Texas Council ofGovemments (October l99l). for streets. Users can order a single-county file, a group of
ll
county files that make up a Metropolitan Statistical Area cdcffi
(MSA), or all files for a state. IH
If a highway network has not been prepared for a plan- a lto
ning region, and an application software (GIS) is available, Octu
the basic data record (record type 1) of the Topologically
Integrated Geographic Encoding and Referencing database
(TIGER)/line can be used to create a roadway network.
Each segment of the basic data record contains the geo-
graphic area code, latitude and longitude coordinates for all
line segments, the name of the feature (such as highway
name), and a class code (functional class). The geographic
and cartographic data of a TIGER/line file can be combined
with other statistical information (such as population, hous-
ing, or income) for planning purposes.
The census TIGER files are a comprehensive source of
data. They can be used to produce area maps, such as census
blocks or county boundaries, as well as a street system net-
work. The TIGER files are available on CD-ROM from the
Census Bureau, and most disks contain the data for two or
three states, while some of the larger states are split onto two
disks. The amount of detail available in these databases is
more than is necessary to build the model network. Conse-
quently, the user must take care to filter out unwanted detail,
such as local streets. Even arterials and highways are Figure 2. Example of network node and link plot with
described in significant detail in these databases. Great care centroid connectors.
was taken in their construction to create links that accurately
reflect the true shape of the highway feature. The number of
shape points used for this purpose, while useful in creating
roughly midlink, as shown. Care should be taken to suppress
maps in GIS, is too much data for the models. Commercial
"backchannel" traffic on centroid links, although most mod-
digitized map files are also available from many vendors.
eling packages allow centroid control of this.
The quality of census TIGER files may not be adequate for
some study area's needs, and commercial vendors can be
used as a source of digital maps.
Network Coverage
With the use of GIS, many commercially produced net-
works are now available. Many of these are simply enhanced The process used to select the links that will be included
TIGER files, which save the user the time and effort of edit- in the coded highway network requires the official functional
ing census TIGER files.
classification of the roadways within the region, the average
traffic volumes, street capacities, and a general knowledge of
the area. Generally, all streets that carry a substantial volume
Scaled Maps
of traffic should be included. The magnitude of volume con-
The base data to construct a computerized highway net- sidered to be substantial will vary by the population size of
work file (location of links and nodes) can be digitized from the area to be modeled, but it is reasonable to expect that all
scaled maps. That is, each facility selected for use in the net- arterial streets and some collector streets will need to be rep-
work can be copied from the base map into digital form with resented. Local residential streets do not need to be included
an application software and a digitizer tablet. The digital in the network, because they will be simulated using direct
input is scaled to either the latifude and longitude or a user- connections between traffic zone centroid nodes and the arte-
defined y coordinate system. There should be no dead-end rial street system via centroid connectors.
links in the system. The links should be mapped with a node The kind of analysis for which the network will be used
at every intersection and at each location where the number determines the level of detail required. For example, if a
of lanes changes. regional network is under analysis, all freeways and major
Figure 2 shows an example of a node and link plot with a arterials in the study area should be included as links. Addi-
zone centroid and connectors. Links connecting a centroid to tional links should be added when they: (1) create a con-
the network (called centroid connectors) should be made tiguous network, connecting arterials and freeways together;
12
(2) represent several parallel streets that collectively carry in turn is used in the trip
areas of the planning region. This
arterial levels of traffic; or (3) interchange with the freeway distribution and mode-choice phases of the travel-demand
system. process. A typical urban transportation model uses both peak-
The boundary of the study area is usually drawn along a period travel speeds and off-peak travel speeds, depending on
political boundary (county, census tract). A cordon line is the analysis period being considered.
drawn across streets where they enter or exit the study area.
Such data as roadside surveys and traffic volume counts
Link speeds obviously vary due to numerous factors, including
taken at the cordon line are useful in calibrating the model.
. Prevailing traffic volume on the link,

Network Attributes2 . Posted speed limits,
. Adjacent land use activity and its access control,
Highway links are assigned attributes representing level of . Functional classification ofthe street, and
service afforded by the segment and intersections ofthe high- . Type of intersection control and spacing of intersection
way system. Travel time, speed, and any delays attributable
controls.
to travel time must therefore be assigned to the link, which
may also designate energy consumption and air and noise
pollutant emissions. The trend is toward greater detail of Transportation models can use any of several approaches
link-based data to allow intersection delay to be incorporated to simulate appropriate speeds for the links included in the
into the travel-demand model. However, three basic items network.
are needed by a transportation model to determine imped- A relatively simple approach uses the posted speed limit
ance for the appropriate assignment of trips to the network: for each link included in the network. Because the trans-
distance, speed, and capacity. portation model will lower the initially coded link speeds in
Because all three of these attributes can be estimated in response to the traffic volumes assigned to the link, this sim-
various ways, it is useful to understand the alternative plified method works reasonably well in most cases. This
approaches and the possible implications ofeach approach. approach requires a relatively complete knowledge of the
snlh}-jl
posted speeds throughout the entire modeled region and may
'I
l
not be acceptable for air quality analysis.
Link Distance An alternative approach is to use a free-flow speed look-
up table like the one shown in Chapter 4. Such a table lists
Most software packages measure link length automati- default speeds by area and facility type.
cally. The length of each link also can be estimated from a
properly scaled map. The accuracy of this approach is ade-
quate, but the method is fairly labor intensive. The distances
estimated in this manner also must be entered directly into Link Capacity
the link records, and this provides an additional opportunity
for errors to occur. Link capacities are a function of the number of lanes on a
An alternative to hand measuring scaled distances is to use link. However, lane capacities can also be specified by facil-
the x and y coordinates ofthe two nodes that define each link ity and area type. Several factors are typically used to
and computerize the process of estimating link distances. The account for the variation in per-lane capacity in a highway
accuracy of this method is related directly to the quality of network. These include
the digitizing process; for instance, whether curved links are
represented by straight lines. A significant advantage ofthis
. Peak-hour factors,
approach is the consistency of the estimates throughout the
. Type ofintersection control,
network.
. Percent trucks, and
. The green to cycle-length ratio at signalized intersec-
tions.
Link Speeds
Link speeds are a major input within the modeling process Determination of actual capacity for each link in a
because they are used to determine the fastest paths between
network is a time- and data-intensive process. If resources
are not available, Chapter 10, Capacity Analysis, provides
2Pasons Brinckerhoff per-lane default capacities by area type, facility type, and
Quade & Douglas, evizw of Best Practices, prepared for the
Metropolitan \Vashington Council of Governments (December 1992). default speed.
13
Area Type Considerations . Route name/number, and

. An average speed, although this can be derived from the
Area type refers to a method of classifying small geo- assignment process and fed back into the transit model.
graphic areas by a rough measure of land use intensity, pri-
marily based on population and employment density. A In addition to the routes, the transit network will need to
higher intensity of land use generally means more inter- include walk links if they are not adequately covered by the
sections, driveways, traffic signals, turning movements, and centroid connectors. Ifpark-and-ride lots are used, highway
pedestrians, and, therefore, slower speeds. links are used to connect the zones to the park-and-ride lots.
Typically, roadway link speeds and capacities are adjusted
slightly based on the area type where they are located. Many
models use approximately five area type codes representing: SOCIOECONOMIC DATA
CBD, CBD fringe (or outlying business district), urban, sub-
urban, and rural. Socioeconomic data used in the planning process include
Roadway link records include arr rrea type field that com- household and employment data. Socioeconomic and land-
puter software can use, in conjunction with facility type use data are compiled and coded to units of geography to give
information and the number of lanes, to determine appropri- the transportation planner an understanding of the way land
ate link capacities and speeds. is used in each section of the planning area. Most urban areas
already have developed TAZs and collected land-use and
socioeconomic base data within the zone geography. Vy'here
Final Network Database this is not the case or where the level of detail, comparabil-
ity with census tract or block boundaries, or a specific proj-
This section describes procedures for establishing the net- ect requires it, zones will need to be developed or modified.
work database. Among the attributes that should be included
for each link are
Structure of the TAZs
. "4" and "8" node numbers and their associated x, y
Zones are geographic areas dividing the planning region
coordinates;
. into relatively similar areas of land use and land activity.
Links defined by "4" and "B" nodes;
. Zones represent the origins and destinations of travel activ-
Link length;
. ity within the region. As it is not computationally feasible to
Functional classification, including the divided or un-
represent every household, place of employment, shopping
divided status of the link's cross-section;
center, and other activity as a separate origin and destination,
. Number of lanes;
these entities are first aggregated into zones and then further
. Controlled or uncontrolled access;
simplifred into a single node called a centroid.
. One-way versus two-way status;
A centroid is a point that represents all travel origins and
. Area type;
destinations in a zone. Zone centroids can be placed in the
. Speed; and
center of activity of the zone, using land use maps, aerial
. Capacity. photographs, and local knowledge. Unique special genera-
tors, especially when surrounded by a very different land use
In addition, the coding of the traffic analysis zone in which (such as a college campus in a predominantly residential
each link is located on the link record facilitates the identifr- area) should be isolated as separate zones. The center of
cation of area type for the link and is useful in producing activity is not necessarily the geographic center-it is the
assignment summaries. Base-year ground counts also may be midpoint of activity. Often judgment is more useful than
coded for automatic screenline analysis. measurement technique in determining where the zone cen-
troid should be located.
Each centroid, or loading point, must be connected to the
Transit Networks highway and transit systems. Typically in the highway net-
work, these centroids are connected to the highway system at
The transit network is developed after the highway net- several points to represent the many paths over which each
work. Essentially, aroute name is designated and the node-to- ofthe discrete origins and destinations within a zone accesses
node path that makes up the transit line is coded. Additional the highway system. Multiple connections tend to smooth the
information about the transit line includes traffic on the adjacent links-if only one connection is given,
the point at which the centroid connects to the street system
. Headway (typically, three different headways can be will show abrupt changes in traffic volume at that point. Spe-
coded: eu, Midday, and rvr), cial generators should have the actual load points coded,
t4
even ifit is only one, for a realistic distribution oftraffic on suitability for collection and input to the master database for
the network. the transportation planning model. Another selection crite-
Once the zone system is developed and mapped, and a cen- rion is the accuracy ofthe data and ability to update the data-
sus equivalency table is constructed, zonal socioeconomic base periodically. Using these two criteria, the following data
and land-use data can be assembled for the transportation sources should be evaluated.
planning process.
1. U.S. Census-The decennial census offers the best
source for population and the required companion
Relationship to Census Geography
characteristics data. These data are available at block
Zone systems should follow available census data bound- level and can be aggregated to traffic zones. Although
the census documents employment in the form of labor
aries (either tracts, block groups, or blocks), so that data
collected in the decennial census can be used with minimal force, it is a household-based count rather than a count
manipulation. To implement a robust database collection and of employment by location. Labor force statistics pro-
maintenance methodology, an equivalency table correlating vide counts of employment in various areas where
census tracts and blocks to TAZs should be developed. The residential population is located, but do not specify em-
table below gives an example. An equivalency table will ployment by employer or location. The Census Trans-
enable immediate cross reference and database aggregation portation Planning Package (CTPP) provides sums of
to the TAZs and various planning or other study areas. resident employment attheTAZ; therefore, the census
can be used for some general cross-checks for employ-
ment, but cannot be used to compile employment data
for small geographic areas.
TAZ Census Block 2. State Employment Cornmission (Employment Com-
! 101 540393291 04320
mission)-The state employment commissions gener-
101 540393291 04321 ally document all employees for tax purposes. Each
101 540393291 04322 employer is identified by a federal identification num-
102 540393291 04s23 ber, number of employees, and an address usually
102 540393291 04324
keyed to where the payroll is prepared for the specifred
number of employees. The state employment commis-
sion data are significant because the address can tie the
employer, and therefore the number of employees, to a
Zones should also be relatively homogeneous in chaacter
specific geographical area. One consideration with the
and consider major physical and transportation boundaries.
Where possible, a zone should not contain various land uses,
employment commission data is that often up to half
nor should it
cross a river, freeway, or other major topo- the employers list Post Office (P.O.) boxes as the
graphical barrier. address of record. In addition, employers with more
than one location are not always disaggregated; that is,
the headquarters of a firm may be listed with the total
TAZ and Highway Network Comparability employment combined for all franchises.
1 Market Research Listings-Many market research
The size of zones should reflect the purpose of the firms offer commercial listings of all (or major)
intended analyses. For systemwide planning, where volumes
employers and number of employees by county and
on individual transportation links are not of concern, zones
city. All the listings show business locations by street
can be larger than the arterial grid. If traffic volumes on the
addresses, as well as P.O. boxes. Commercial listings
arterials are the subject of analysis, zones should not be so
offer these data on a subscriber basis with a range of
large that two or more arterials traverse the zone in any direc-
access and purchase options.
tion. Vy'hen this happens, travel generated by the zone tends
4. Local Area Population and Employment Data-Many
to be lumped together during the assignment process, to pro-
areas collect and record some type of population data.
duce unrealistic travel patterns. If corridor analysis is the pur-
pose of zone development or modification, zone sizes in the
But few areas record employment data other than a
broad listing of the employers with the highest number
corridor or subarea should be more fine-grained.
of employees locally. Chambers of Commerce often
publish lists of member businesses.
Sources for Socioeconomic Data 5. Aerial Photography and Existing Land Use-Often
aerial photographs can be used to update existing land
The availability of various data sources, among other fac- use. The resolution of the photography is good enough
tors, will be one of the primary criteria in determining the to differentiate many residential and nonresidential
15
areas. When compared with the aerial photographs, most efficient and accurate method by which employment
each land use can be associated with a particular land can be collected and organized into the database file.
use type (e.g., retail and industrial) for each building.
6. Telephone Directory-The telephone directory is a rich
database that lists residential units and businesses by Control Totals for the Database
street address for any planning area; however, the rate
of unlisted residential numbers is rising and may influ- The control totals for the database should be determined
ence the use of telephone directories. before the compilation of the data. The source of the control
1. Polk Directory--he Polk Directory is a comprehen- totals for population should be the U.S. census.
sive list of household and employment data sorted by
name and address. For households, such information
IMPACT AND USES OF GIS ON DATABASES
as occupation and employer can be ascertained. For
business establishments, type of business-including
Geographic information systems are designed to capture,
associations, libraries, and organizations that may not
store, retrieve, analyze, and display data files referenced to
be on the tax file----can be determined.
detailed geographic location; for example, latitude and longi-
tude, state plane coordinates, census tract or block, or a
locally developed geographic scheme. GIS software enables
Data Source Def iciencies
the planner to process geographic-based information from a
l. Populatioz. The only data source acceptable to initially multitude of sources. For instance, the tax assessor's office
establish a residential database is the U.S. census. All collects and maintains a wealth of information organized by
of the other data sources identified above do not pro- tax parcel and indexed by a complicated map reference sys-
vide comparable population statistics by specific area tem. Likewise, public utilities, school districts, police, and
(i.e., block level). fire departments all collect information on the populations
2. Employment. Each of the above data sources has some they serve and often have their own system of geographic
deficiency in accurately specifying employment for reference. State departments of transportation have data on
small geographic areas. The census provides total labor traffic accident locations, pavement condition, bridge loca-
force by TAZ; however, this represents only employ- tions, signs, and rights-of-way.
ment location of residents and not total employment. The A GIS organizes and provides access to these data, allow-
employment commission data provide accurate employing the user to overlay and analyze it using a common frame
ment for each business but only partially list street of reference (either address- or block-specific), and display
addresses (with the remaining having P.O. box listings). it in an easily understood format. This allows otherwise
Commercial listings have all employers by street ad- unavailable or cumbersome data to be used in transportation
dress. Although these listings are extensive, the accu- planning. Increasingly, the major differences between tradi-
racy is controlled intemally and often camot be consid- tional planning and GIS are becoming blurred.
ered comprehensive (because of the lack of information Three typical applications of GIS in transportation plan-
regarding collection methodology), but it offers a check ning are geocoding, processing socioeconomic data, and
for other data sources. The aerial photography assists cartography.
only in determining potential nonresidential areas. The
land-use data obtained from aerial photography provide
Geocoding
a geographic location of businesses but do not provide
, employment numbers. A similar condition exists with
One of the reasons that the compilation of transportation
the telephone directory. The directory covers most busi-
information is so expensive is that data gathered from sur-
nesses and provides addresses for most of them. Several
veys on trip origins and destinations must be related to spe-
multi-use areas, such as shopping centers, do not provide
cific physical location. This process, commonly referred to
numbered street addresses.
as geocoding, is usually a tedious and time-consuming
manual process. Many GIS applications include a geocod-
In summary, none of the data sources offers a complete ing capability that automates this process, allowing a street
inventory of employment by geographic location. address, place name, or intersection to be geographically
Employment data are the most difficult data component to referenced to latitude and longitude, census tract, or TAZ.
collect. The census data show labor force statistics by indus- Good geocoding software includes the ability to handle
try group but do not compile this by employer and specific misspelled names; ambiguities caused by similar names
geographic area (i.e., block). Therefore, the methodology for given to avenues, streets, and boulevards; and incomplete
developing the employment database should be based on the addresses.
' :.: ...11
r. ;
t6
Socioeconomic Data I-40, running east to west, and I-26, which connects
Asheville to Atlanta. A belrline, I-240, makes a half-circle
Socioeconomic data used in transportation planning around the City to its northern side. As of 1991, City Coach
include information on household characteristics, employ- carried fewer than 4,000 daily passengers, approximately I
ment, and land development. These data are the common percent of the daily person trips made in the Asheville area.
..1
characteristics of each zone that affect the generation and The remainder of the mechanized person trips on the trans-
..-.1
,i attraction oftrips (e.g., trips are generated by households and portation network were made in private vehicles.
..1
attracted by work or retail locations). Information can be
obtained from various sources in addition to those listed
above, such as tax assessment records, zoning maps, school Socioeconomic Data
enrollment files, and public utility records.
The trip generation equations introduced in this chapter
use socioeconomic and land-use data to describe the quantity
and type of travel activity in the region. Required data
Cartography include the following: number of households by size; house-
' -: 1..
hold income or automobile ownership; and employment by
A GIS in conjunction with digital cartographic informa- type. These data should be allocated throughout the region
tion (such as U.S. Geological Survey IUSGS] digital line according to aTAZ structure that is appropriate for the level
graphs or the U.S. Bureau of the Census' Topologically Inte- of analysis and the detail in the selected network.
grated Geographic Encoding and Referencing database The North Carolina Department of Transportation
[TIGER]/line files) and geocoded trip and socioeconomic (NCDOT) originally developed the Asheville TAZs. Their
data can be used to prepare maps to support technical analy- zone system contains 353 internal zones and 36 external sta-
sis. The USGS and census TIGER maps show streets and tions. NCDOT compiled household and employment data for
railways; features such as schools, parks, and hospitals; cen- the MSA at the 353 TAZ level of detail. A review of the
sus blocks and tracts; political boundaries; hydrographic fea- NCDOT zone structure revealed that it provided more detail
tures; and topographic relief. The difference that GIS makes than was needed for the case study example, which is
rffi*4 to cartography is the ability to relate items spatially within intended to forecast traffic volumes only on the major road-
the database. ways. The 353 zones were aggregatedinto 107 internal zones
Geographic systems co-relate point, line, and area data and 16 external stations for the case study.
that describe the mapped region and attributes about those The socioeconomic data for this case study were extracted
data. For example, a point-on-line analysis might be to iden- from the 1990 U.S. Census. Serial Tape File (STF) 3-A was
tify the schools within 5 miles of a hazardous material route. downloaded from the North Carolina State Library. These
.l
.j
A point-on-area inquiry could display building permits data included the following summary information:
,t approved within the last 2 years in aTAZ, A line-on-area
query could display all four-lane roadways within a metro- . Household income by household size,
politan area. An area-on-line analysis could identify high- . Median household income,
employment zones served by a proposed transit route. . Number of persons in household,
. Mode of travel to work,
. Time of departure from work, and
CASE STUDY
. Private vehicle occupancy for work trips.
At this point we will introduce the Asheville, North Car- The census data revealed that the study area contains a
..t.' .
'.. ...i olina, test case. The Asheville MSA lies in the western quar- populationof approximately 110,000 persons in 46,492
.. :.
ter of the State of North Carolina, roughly 230 miles west of households. A listing of households by size for each of the
the state capital in Raleigh, and 1 10 miles east of Knoxville, 107 TAZs is provided as Appendix Table B-1.
Tennessee. As defined by the U.S. Census Bureau, the The household and income data will be used in the trip
region consists of the City of Asheville and surrounding generation equations to calculate the person-trip produc-
Buncombe County. In 1990, the entire MSA had a popula- tjozs. Person-trip attractions, on the other hand, are based
tion of l'7 4,821, and the City of Asheville had a population on employment data, stratified by type of employment. The
of I,429. NCDOT data provided for this study were divided into sev-
Figure 3 displays a base map for the Asheville region. eral different sectors according to the Standard Industrial
Asheville's transportation network consists primarily of its Classification land use code. Total employment for the
roadway system and the City Coach bus service, a l2-rotte MSA is estimated at 59,037 by NCDOT's surveys. A list-
I
system operated by the Asheville Transit Authority. Two ing of the employment in each of the 107 TAZs is provided
U.S. Interstate routes meet just south of the City of Asheville: as Appendix Table B-2.
t7
F
o
!
o

Yrlrl tfi4r
".\ (
ld
L.:{nlbl
Figure 3. Asheville, North Carolina, highway map (courtesy of North Carolina Department of
Transportation).
18
TABLE 1 Base network summary
Nodes Quantity
Zone Centroids (lnternal) 107

Zone Centroids (External) 16
Regular Nodes 1,156
Total Nodes 1,279
Free-Flow Capacity Veh.

Link Type Description Quantity Speed (mph) per Hr.
I Freeway 168 55 1 350

2 Major Arlerial 922 45 825
3 Minor Arterial 1,550 35 825
4 Centroid Collector 632 20
Total Links 3,272
Network Description
The highway network for the region was acquired from

NCDOT as a downloaded file on disk, which was then input
directly into the transportation modeling software package.
Each link in the network was then coded with a facility clas-
siflcation (freeway, major arterial, or minor arterial), number
of exclusive lanes in the direction of travel, the free-flow
speed, and the hourly per-lane capacity of the roadway.
As summarized in Table 1, the base highway network con-
sists of 1,156 regular nodes connected by 168 freeway links,
922major aterial links, and 1,550 minor aterial links. After
the regular nodes were connected by links defined to repre-
sent actual highways, centroid connectors were added to
allow the 107 internal centroids and the 16 external stations
access to the highway network. (If we were planning to
model the transit volumes on the local bus routes, we would
have used the highway network as the base for building a
transit network. However, given that only 1 percent of the
person trips in the Asheville MSA use transit, that compo-
nent of the transportation network was ignored.)
Figure 4 is a plot of the coded base highway network,
including all link types except centroid connectors.
Traffic Count Data
NCDOT also provided a selected count map, which dis-

played total daily vehicle volumes for various intersections
and cordon points. These counts, summarized in Table 2,
were used to provide the traffic volumes at the 16 external
stations (see Chapter 5) and to perform screenline compar-
isons of the actual versus modeled traffic counts within the
network (see Chapter 9). Figure 4. Asheville, North Carolina, TAZs.
l9
TABLE 2 External sttions
Stalion No. Description 1989 ADT Classification
108 Route 251 1,800 Minor

109 Routes 19 & 23 Bypass 27,700 Principal
110 Routes 19 & 23 Business 7,000 Minor
111 BRP (N) 2,850 Minor
112 Snope Creek Road 2,000 Minor
113 Route 70 1 6,1 00 Principal
114 r-40 (E) 24,700 lnterstate
11s Route 74 11,000 Minor
117 l-26 33,'100 lnterstate
1't8 Routes 191 & 280 7,400 Minor
119 BRP (S) 970 Minor
'121 r-40 (vv) 27,500 lnterstale
'122 Leicesler Highway 14,000 Principal
123 Bear Creek Road 3,940 Minor
SUMMARY amount of data available and to use only what is truly neces-
sary to create the desired output.
The range of electronic data now available to the modeler With the use of GIS, many commercially produced net-
is substantial. These data are most useful in the creation of works are now available. Many of these are simply enhanced
the socioeconomic inputs to the model as well as the devel- TIGER files.
opment of the model network itself. This allows finer detail The next chapter, Trip Generation, begins the four-step
and greater accuracy than was available even a few years modeling process with a discussion of data needs and the
ago. While these advances are on the whole good for the process used to create a basic trip table that will be used as
modeler, one must be careful not to be overwhelmed by the an input to the model.
20
CHAPTER 3
TRIP GENERATION
INTRODUCTION ending point. However, the trip ends at the household are
called productions and the trip ends at nonresidential land
The first step in travel forecasting is determination of the uses are called attractions. The production-attraction format
trips currently undertaken in a planning region. Trip genera- does not indicate the true direction of travel as opposed to
tion, or the amount and type of travel in a region, is func- the origin-destination format. The procedures for converting
tionally related to the use of land. To estimate trip generation, productions and attractions is discussed in the last section of
land use is described in terms ofthe character, intensity, and Chapter 8.
location of activities. Factors influencing the amount of Trip-generation models consist of two submodels includ-
travel in a region include automobile ownership, income, ing trip-production models and trip-attraction models. Trip
household size, and density and type of development. The productions are the trip ends associated with the traveler's
availability of public transportation and the quality of the home. Trip attractions are the trip ends associated with the
transportation system also influence trip generation. The best non-home end of the trip, such as a workplace, shopping cen-
travel-demand techniques make use of disaggregate socio- ter, or school. For example, if a person went from home to
economic data such as households classified by vehicle work in the moming, and then went from work to home in
ownership, households classified by family size, or house- the evening, the traveler would have generated two produc-
holds classied by income group to determine the amount of tions at the home zone and two attractions at the work zone.
travel generated in the region. The procedures presented in The household and employment data are used to estimate the
this report are not well suited to generating trips for non- total travel generated (produced and attracted) by each zone
motorized modes, bicycle or pedestrian. in the planning region.
The purpose of trip generation estimation is to determine In this chapter, trip-generation models are presented as
the number of person or vehicle trips to and from activities person-trip productions based on household size, income and
in an analysis area. Trip generation is important in a number auto ownership (Tables 4 through 7), and auto ownership and
of phases of transportation planning and traffic engineering income (Table 7). Table 8 presents trip-generation models
studies. It is necessary for for person-trip attractions.
. Regional studies that consider a range of land uses and

related social and economic characteristics, BASIS FOR DEVELOPMENT
. Regional transportation altematives studies and devel-
opment of long-range transportation plans, Site-Specific Vehicle Trip Rates
. Short- and long-range plans that evaluate transportation
In Table 3, site-specific vehicle-trip generation rates were
needs in a corridor or specific subarea, and
. extracted from the Institute ofTransportation Engineers' Ir
Impact studies that assess the effect of new development
Generation,5th Edition. While the stated trip rates for many
such as a shopping center, residential development, or
common land uses are provided here, the ITE Trip Genera-
industrial park (site impact analysis).
tion report actually provides other methods for determining
the trip generation rates for these land uses. The manual
These different requirements for trip generation informa- should be reviewed to determine the proper trip generation
tion have resulted in various levels of trip generation data. rate procedure to use and appropriate adjustments for factors
Site-specific vehicle trip rates (Table 3) are presented for such as transit use and multi-use projects. Consistent with the
impact studies where land use is known for a small area, and format of the ITE manual, the rates are presented as average
vehicle rates are needed to assess the impact of develop- daily vehicle trips and percent of vehicle trips during the a.m.
ment. For most other types of studies, the use of person trips and p.m. peak hours of the generator. Only a subset of the
is preferred. most commonly used rates are presented, and the user is
In describing the direction for a trip, the term originrcfers referred to the most current version of the ITE manual for
to the starting point and the term destinallo refers to the rates on generators not included or those which require more
21
TABLE 3 Site-specific vehicle trip generation rates
Percent ot
Total Daily
Vehicle Trips
Daily Vehicle- A.M. P,M. ITE
Land Use Trip Rate per Peak Peak Code
Resldential
Single-Family 9.s5 DU 8.Olo 10.7"/" 21O
Apartment 6.47 DU 8.6 10.7 220
Condo/Townhouse 5.86 DU 7.5 9.2 230
Mobile Home Park 4.81 Occupied DU 8.9 12.1 240
Planned Unit Development 7.44 DU 7.8 9.7 270
Retalll
Shopping Center
Under 100,000 sq. ft. 70.7 1,000 sq. ft. GFA 2.3"/o 9.2% 820
100,000 to 1,000,000 sq. ft. 38.7 1,000 sq. ft. GFA 2.'t 9.5 820
500,000 to 1,000,000 sq. ft. 32.1 1,000 sq. ft. cFA 2.0 9.3 820
More than 1,000,000 sq. tt. 28.6 1,000 sq. ft. cFA 1.8 9.1 820
Offlce
Genera12 11.85 1,000 sq. ft. GFA 13.8"/" 13.10/o 710
Medical 34.17 1,000 sq. ft. GFA 10.0 13.0 720
Otfice Park 11.42 1,000 sq. ft. GFA 16.1 13.2 750
Research and Development Center 7.70 1,000 sq. ft. GFA 16.0 13.9 760
Business Park 14.37 1,000 sq. tt. GFA 'l 1.3 10.3 770
Restaurant3
Quality Restaurant 96.51 1,000 sq. ft. GFA 6.6o/o 10.1o/o 831
High Turnover (Sit Down) 205.36 1,000 sq. ft. GFA 8.7 15.5 832
Fast Food without Drive-Through 786.22 1,000 sq. ft. GFA 9.7 13.7 833
Fast Food with Drive-Through 632.12 1,000 sq. ft. GFA 9.5 7.3 834
Bank
Walk-ln 140.61 1,000 sq. ft. GFA 13.7% 0.4o/o 91 1
Drive-Through 265.21 1,000 sq. ft. GFA 13.3 19.3 912
HoteUMotel
Hotel 8.7 Occ. Room 8.7"h
7.5o/o 310
Motel 10.9 Occ. Room 6.7 7.0 320
Parks and Recreatlon

Marina 2.96 Berth 5.7"/" 7.1"/o 420
Golf Course 37.59 Hole 8.6 8.9 430
City Park 2.23 Acre NA NA 411
County Park 2.99 Acre NA NA 412
State Park 0.50 Acre NA NA 413
Hospital
General 11.77 Bed 11.67" 610
10.0"/"
Nursing Home 2.6 Occupied Bed 7.7 10.0 620
Clinic (one data pont) 23.79 1,000 sq. tt. GFA NA NA 630
(continued on next page)
22
TABLE 3 (Continued)
Percent of
Total Daily
Vehicle Trips
Daily Vehicle- A.M. P.M. ITE
Land Use Trip Rate per Peak Peak Code
Educatlonal
ElementarySchool '10.72 1,000 sq. GFA 25.6"/o 23'20/"
ft' 520
High School 10.90 1,000 sq. ft. GFA 21.5 17.8 530
Junior/Commun College 12.57 1,000 sq' fi. GFA 17.2 8'2 540
University/College 2.37 Student 8.4 10'1 550
Alrport
Commercal 104.73 Average 7.8"/o 6.6/" 021
Flights/Day
General Aviation 2.59 Avorag 10.4 12.7 O22
Flights/DaY
lndustrlal
General Light lndustry 6.97 1,000 sq. ft. GFA 14.5o/" 15.57" 110
General Heavy lndustry 1.5 1,000 sq. ft. GFA 34.0 4s.3 120
Warehouse 4.88 1,000 sq. ft. GFA 11.7 12.3 1s0
Manufacturing 3.85 1,000 sq. ft. GFA 20.3 19.s 140
lndustrial Park 6.97 1,000 sq. ft. GFA 11.8 12.3 130
Not: Rates are often given for other factors, such as acres, employees, or sq. ft. of gross floor area.
.-r--,'.,."r Weekend rats are also given for many uses. For some, like shopping centers, the weekend rates
: are higher than weekdaY rates.
*-1
. t Rates given are for high end of indicated range. ITE's Ir Generation details rates for 15 sizes.
..1
:i _ Weekend rates for shopping cnters are significantly different from the weekday rates given hre. Trip
' 2 ITE details rate for 11 size categories, from 10,000 sq. tt. to 800,000 sq. ft. The rate given here is for a
';
t Rates are given for the number of seats in the restaurant.
,:. - specific site characteristics. When appropriate, local data ified into the population subgroups listed. These subgroups
,', ' ' should be collected and rates developed as in Table 3. are different from those used in NCHRP Report 187 becawe
'' ' utilize of federally mandated planning requirements that specify
Please note that the site-specific trip generation rates
,.' I the number of dwelling units as a key data input. This is be- population stratification groups. The small urban size is now
s--- cause site-specific planning relies on the size and number of consistent with requirements specified in ISTEA.
i'. ' physical structures in a study area. Transportation modeling
.: ' ' ! techniques, on the other hand, rely heavily on census informa- Population stratification groups are as follows:
,- : '. tion as a data source. Therefore, regional trip generation models
,: . , . , use households, or the number of family units, as the primary . 50,000 to 199,999 persons,
. data source for estimating home-based ffip generation. . 200,000 to 499,999 persons,
;l:. - *
i.;: .r..:..: ' 500,000 to999,999 persons, and
. More than 1,000,000 persons
.:t,:, .,
.,', - , Model TriP Rates
,..:... , Original home-interview surveys for the 50,000 to 199,999
,.," ... The data presented in Tables 4 through 9 were derived category were limited. In addition, variations in survey
,,, - . - from available home interview surveys taken since 1985 and design, definitions, collection methodologies, and the expan-
, ', ,l from the 1990 NPTS. The data from both sources were strat- sion of the data made it difficult to develop conclusive tnd-
23
TABLE 4 Percent ofhouseholds by autos owned and income
Autos Owned
lncome 0123+
Urbanlzed Area Slze = 50,(XX) - 199,999
Low' 1755235
Medium 3314818
High 012s236
Weighted Average 8324020
urbanlzed Area slze = 200,(xx) - 499,999
Low 1751248
Medium 2325313
High 0135334
Urbanlzed Area Slze = 500,fl)0 - 999,999
Low 215s204
Medium 3394414
High 110s930
Urbanized Area Slze = 1,(X)0,(X)0 +
Low 1952227
Medium 2404216
High 1105s34
' ln actual 1990 dollars: Low= lessthan 520,000, Mediurn = 520,000to 39,999, and
High = $4g,ggg and uP.
ings about variations in trip rates across urban area sizes and
sented in NCHRP Report 187. The findings of the study
characteristics. The NPTS data provided the only consistent resulted in trip generation rates closely grouped around an
source oftravel data throughout all ranges of urban area size
average of 9.0 daily person trips per household. A summary
and locationl however, the NPTS rates were low compared of the comparison of the average trip production rates con-
with the rates collected by local areas. The lower rates can be
tained in this report and those contained in NCFIRP Report
attributed to the collection methodology of the NPTS. For 18Zis presented below.
generation data, the NPTS rates were factored in a normative The impact of the revised trip rates should be reviewed by
distributiontomatchmorecloselytheratesdeterminedbythe those who developed models using the earlier NCHRP l.
urban areas used as source data. actual Report 187 rates.In many cases, it was acknowledged that
A comparison of the ,
rateSiSshowninAppendixTableA-2.theratesforsmalIurbanareaSwerehighandtheactual
A major finding after reviewing the home-interview data applied rates were adjusted downward. If the complete .
and the NPTS data was that the difference between trip gen- model set was calibrated and validated to match observed .
eration rates across urban areas of different sizes was not traffic count data, then any adjustments to the model should
large, particularly when compared with the difference pre- be reviewed in light of the revised rates.
24
' 1.1
'.
' :;
i
Urbanized Person Trips/ Person Trips/Household

j Area Population Household NCHBP Repoft 187*
-l
.-
,'
.
.:.1 50,000 lo 200,000 9.2 14.1
...i'.
i 200,000 to 500,000 9.0 1'1.8
..:, r
.l
500,000 to 1,000,000 8.6 7.6
. ^.1: > 1,000,000 8.5 7.6
., l
.,
''!
* Note: Because of different urban area size calegories belween NCHRP Reporl 187 andlhis
,1
report, the rales shown were chosen from the closest matching category.
....,.1.,1.
.'l
DATA REOUIRED FOR APPLICATION
Site-Specific Vehicle Trip Rates

..i
For site-specific estimates, the following input is required:
.i
l
'i
Generator lnput Required
Residential locations Type of residence and number of dwelling unils or acres of

development
I ndustrial/manufacturin g, Gross floor area (GFA), or employees, or acres of
offices development
Restaurants Gross floor area or acres of development
Banks Gross floor area or employees
Parks and recreation sites Acres (or employees for a few types)
I
Hospitals Statf or beds
Educational locations Students or staff
Airports Take-otfs and landings or employees or acres
Hotels/motels Rooms or employees
Retail sites Gross floor area or employees or acres
Military bases Military personnel and civilian employees or total employees
Race tracks, stadiums Seats or attendees
Service stations Number of pumps
ModelTrip Rates a region.r For the basic cross-classification model, the second
independent variable is a measure of the wealth of the house-
The data presented in this section provide useful informa- hold, typically either directly as income or indirectly as auto
tion for estimating trip rates based on differences in income ownership.
and/or auto ownership. The data have been summarized into The data requirements for application of the information
four urbanized area population groups. Income is in three provided for trip generation include land-use and socio-
income groups (low, medium, and high in 1990 tertiles). economic characteristics generally used for areawide planning
To predict the number of trips produced in an area, the and site-specific characteristics used in land development
planner usually employs a set of household trip production analysis. The material provided allows some variation in data
rates stratified by at least two of the relevant variables that requirements based on data availability, level of analysis
describe households, such as the number of persons in the required, and time available.
I
household and income, or number of persons in the house- To use the material in Table 4 as input data for trip gener-
I
hold and auto ownership. Most of these cross-classification ation, estimates are required of the number of households by
models use household size as the independent variable to
predict trips per household. The variation in trips between
household sizes is so large that models without this variable lParsons, Brinckerhoff
Quade & Dou glas, Review of Best Practces, prepared for the
ae inferior in approximating the observed travel patterns in Metropolitan Washington Concil of Govemments (December 1992).
25
TABLE 5 Average daily person trips per household by persons per household and income
Persons per Household

Weighted
5+ Average
Urbanlzed Area Slze = 50,0(X) - 199,999
Low* 3.6 6.5 9.1 11.5 13.8 6.0

Medium 3.9 7.3 10.0 13.1 Ir o oe
High 4.5 9.2 12.2 14.8 18.2 12.7
Weighted Average 3.7 7 .6 I 0.6 13.6 16.6 9.2
Urbanlzed Area Slze = 2(X),0(X) - 499,999
Low 3.1 6.3 9.4 12.5 14.7 6.0

Medium 4.8 7.2 10.1 13.3 15.5 9.4
High 4.9 7.7 12.5 13.8 16.7 11.8
Weighted Average 3.7 7 .1 1 0.8 13.4 15.9 9.0
urbanlzed Area slze = 500,000 - 999,999
Low 3.6 7.1 9.0 12.O 14.0 6.0

Medium 4.8 7.1 9.8 12.7 14.6 8.9
High 4.8 7.8 11.5 13.6 16.6 11.5
Weighted Average 4.O 7.3 10.2 13.0 15.4 8.7
Urbanlzed Area Slze = 1,0(X),0(X)+
Low 3.7 6.3 8.1 10.0 11.8 5.7

Medium 4.9 7.6 9.1 12.3 15.1 9.0
High 5.4 7.9 10.3 12.4 15.3 10.8
Weighted Average 4.2 7.3 9.3 12.0 14.8 8.5
ln actual 1990 dollars: Low = less than 520,000, Medium = 520,000 to 39,999, and
High = S40,000 and up.
income or auto ownership category. Such estimates can be shown. Table'l presents estimates of average daily person
developed from census data. Table 4 shows an estimate of the trips given an income distribution and the number of auto-
percent of households by autos owned in each income cate- mobiles owned.
gory. An estimate of the percent of households in each cell Estimation of trip attractions (Table 8) requires-for each
may be obtained by using either an income distribution or an analysis area-total employment, retail, service and other
auto-ownership distribution. Table 5 allows an estimate of employment, and number of households.
average daily person trips given an income distribution and After these trip attractions for CBD and non-CBD are
the number of persons per household. Table 6 estimates aver- computed, they are added together before proceeding to trip
age daily person trips, given the number of autos owned and distribution.
the number of persons in the household. In cases where the Table 9 presents household data and can be used when
number of autos owned is greater than the number of persons detailed information is not available. Only the number of
in the household, the data are inconsistent; these cells were households is required, if the distribution by family size or
combined and weighted averages for combined cells are income is not known, to use the weighted average by urban
26
'
TABLE 6 Average daily person trips per household by persons per household and auto ownership
Persons per Household
Autos Weighted
Owned Average
urbanlzed Area slze = 50,(x)0 - 199,999
Zo 2.6 4.8 7.4 9.2 11.2 3.9

One 4.0 6.7 9.2 11.5 13.7 6.3
Two 4.0 8.1 10.6 13,3 16.7 10.6
Three Plus 4,0 8.4 11.9 15.1 18.0 13.2
Urbanlzed Area Slze = 2(X1,0(Xl - 499,999
Zero 2.1 4.0 6.0 7.0 8.0 3.1

One 4.3 6.3 8.8 1',1 .2 13.2 6.2
Two 4.3 7.5 10.6 13.0 1s.4 10.1
Three Plus 4.3 7 .5 13.0 15.3 18.3 13.5
I
Weighted Average 3.7 7 .1 1 0.8 1 3.4 15.9 9.0
Urbanlzed Area Slze = S(X),(XX) - 999,999

I
i
-"--
Zero 2.5 4.4 5,6 6.9 8.2 3.4

One 4.6 6.7 8.8 11.0 12.8 6.4
:,,.-'] Two 4.6 7.8 10.4 13.0 15.4 10.3
Three Plus 4.6 7.8 12.1 14.6 17.2 12.9
-t
Urbanlzed Area Slze = 1,(XX),(XX)+
Zro 3.1 4.9 6.6 7.8 9.4 4.1

One 4.6 6.7 8.2 10.5 12.5 6.3
Two 4.6 7.8 9.3 11.8 14.7 9.7
Three Plus 4.6 7.8 10.5 13.3 16.2 11.8
Weighted Average 4.2 7.3 12.O 14.8 8.5
.,..1
area size to estimate autos owned, average daily person trips cation of employment data. Rather than dividing employ-
and vehicle trips and the percent of person trips that are ment into simply retail and non-retail, the rates presented
home-based-work (HBW), home-based-other (HBO), or here are based on retail, service, and other. These would cor-
non-home-based (NHB). respond to the following Standard Industrial Classification
The trip-attraction model in NCHRP Report 187 wasbased (SIC) codes:
on rates included in the old Urban Transportation Planning
System (UTPS) standard model set and presented as a func- . Retail: Major Groups 52 through 59 (5200-5999)
tion of employment by type (retail and non-retail) and the . Service: Major Groups 60 through 90 (6000-9000)
number of households in a zone. The Report 187 rates were . Other: Major Groups 1 through 51 and 91 through
compared with rates from similar employment types used in
99 (100-5100 and 9100-9999)
other urban areas. For the most part, the Report 187 rates are
consistent but they may be a little high. This is particularly
true for HBW and HBO trips per retail employee. It is also "Other" includes both basic industries and government of
important to consider the employment types. The question the office type (rather than license bureaus, for example). The
of what constitutes retail trips has often led to a misclassifi- user should refer to the S/C Manual, Office of Management
2'l
TABLE 7 Average daily person trips by income and autos owned
Autos Owned
Weighted
3+ Average
Urbanlzed Area Slze = 50,(Xlo . 199,999
Low' 3.4 5.3 8.7 10.6 6.0

Medium s.3 7.0 10.1 12.1 9.3
High 7.1 8.9 12.4 14.6 12.7
Weighted Average 3.9 6.3 10.6 13.2 9.2
urbanlzed Area slze = 2(x),(xxl - 499,9fr9
Low 2.8 4.9 8.6 11.5 6.0

Medium 4.0 7.1 10.0 13.4 9.4
High 5.2 8.4 1',t.2 14.0 11.8
Weighted Average 3.1 6.2 10.1 13,s 9.0
Urbanlzed Area Slze = S(X),(XX - 999,999
Low 3.2 5.5 9.2 11.8 6.0

Medium 4.0 7.0 10.0 11.9 8.9
High 4.9 8.1 11.0 13.8 11.5
Weighted Average 3.4 6.4 10.3 12.9 8.7
Urbanlzed Area Slze = 1,o{Xl,(XlO+
Low 3.7 5.0 7.9 9.8 5.7

Medium 5.8 7.1 9.8 12.O 9.0
High 6.8 8.3 f 0.4 12.1 10.8
Weighted Average 4.1 6.3 9.7 f 1.8 8.5
ln actual 1990 dollars: Low = less than S20,000, Medium = S20,000 to 39,999, and
High = 540,000 and up.
and Budget (OMB) 1987 for detailed descriptions of the different generators. The basic information given for each
industry classifications. The variation in HBO trip rates per generator is as follows:
retail employee was found to be significant between CBD
and suburban zones; therefore, the rates presented are sepa- I. Daily Vehicle-Trip Rare. This provides the trip rate
rated by CBD versus non-CBD zones. based on the most appropriate land use measures for
each type of site, such as trips per employee, acre, or
household. A trip is defined as a one-way vehicle
TRIP GENERATION DATA movement with either the origin or the destination in
AND EXAMPLES OF USE
the study area. Therefore, the trip rates shown represent
Site-Specific Vehicle Trips the sum of the vehicular trips to and from a site (or trip
ends) divided by a measure of the land use such as
Table 3, Site-Specific Vehicle Trip Generation Rates, number of households, acres, employees, and the like.
provides information on vehicle trip rates for a number of Vehicles include automobiles, trucks, taxis, and buses.
28
TABLE 8 Person-trip attraction estimating relationships for all population groups
To eslimate trip attractions for an analysis area, uset
I
HBW Attractions = 1.45 x Total Employment
HBO Attractions CBD = 2.00 x CBD RE + 1.7 x SE = 0.5 x OE + Q.9 x HH
HBO Attractions NBD = 9.00 x NCBD BE + 1.7 x SE + 0.5 x OE + o.9 x HH
NHB Attractions CBD = 1.40 x DBD BE + 1.2 x SE + 0.5 x OE + 0.5 x HH
NHB Attractions NCBD = 4.1 0 x NCBD RE + 1.2 x SE + 0.5 x OE + Q.5 x HH
where
CBD RE = Retail Employment in Central Business District Zones,

NCBD RE = Retail Employment in Non-Central Business District Zones,
SE = Service Employment,
OE = Other Employment (Basic and Governmenl), and
HH = Households.
.t
1
Note: The coefficients for these equations were derived from a variety of trip attraction models
for urban area studies and represent a consensus of these models.
2. Percent of Total Daily Vehicle Trips in Peak Hour. developed or available. Where the number of households by
This provides percentages that can be applied to daily income range is known, an estimate can be obtained of
--: trip (or trip end) estimates based on the trip rates to
r obtain the traffic generated in the a.m. peak hour and . Percent of 0, l, 2, and 3l auto households (Table 4),
the p.m. peak hour on the surrounding street system. . Average daily person trips per household for low,
medium, and high income (Table 5),
For example, a general hospital can be expected to gener- . Average daily person trips per household for 0, 1, 2, and
afe 11.77 trips per bed (Table 3). For a hospital with 100 3* auto households (Table 6),
beds, one can expect 1,171 trips (or trip ends) per day . Percent oftrips by purpose (Table 9), and
(100 x 11.7'7).In the a.m. peak hour (normally occuning . Average autos per household (Table 9).
sometime between 7:00 a.m. and 9:00 a.m.), approximately
10 percent of total daily trips, or 1 I 8 vehicles (0. I 0 x 7,17'7), Some of these data are available from the decennial cen-
can be expected to enter or leave the facility. In the p.m. sus, and the census tracts/blocks/urbanized areas (UAs) that
peak hour (normally occurring sometime between 4:00 p.m. make up the study area can be aggregated (a GIS makes this
and 6:00 p.m.), approximately 12percent of total daily trips simple) to obtain control totals for the region. For instance,
or 141 vehicles (0.12 x l,lll) can be expected to enter or the 100 percent data will furnish the number of households
leave the facility. by persons per household; the sample data will give number
of households in each auto ownership or income category.
Total trips for the region can be estimated from these aggre-
ModelTrip Rates gate data, thereby providing a basis for controlling estimates
derived at the zone or household level.
The data in Tables 4 through 9 can be used in several ways. To illustrate the use of the material presented here,
Ifan estimate of zonal average autos per household is avail- assume an urbanized area with 150,000 population and
able (for instance, from the census sample data), an estimate 65,000 households. With 9.2 average daily person-trips per
of average daily trips per household can be made directly household (Table 9), a total of 598,000 internal-internal plus
using the data provided in Table 7. For example, in an urban- internal-external trips can be expected. A total of20 percent
ized auea of 50,000 to 199,999 population, 9.2 av erage daily of these (Table 9) would be home-based work (HBW) trips,
person trips per household can be expected to be produced or 119,600 daily person trips; 57 percent home-based non-
where the average car ownership is 1.8 per household. This work (HBO), or 340,860; and 23 percent non-home-based
is an appropriate use where zonal averages are most easily (NHB), or 131,540. To obtain auto vehicle trips, the total
29
TABLE 9 Trip estimtion variables by urban size
% Average Daily Person

Trips by Purpose
Average Daily Average Daily
Average Autos Person Trips Vehicle Trips
per Household per Household per Household HBW HBO NHB
Urban Area = 50,000 to 199,999
lncome
Low' 1.2 6.0 4.8 16 60 24

Medium 1.9 9.3 8.I 21 56 23
High 2.4 12.7 11.7 20 55 25
Weighted Average 1.8 9.2 8.1 20 57 23
Household Size
One Person 0.9 3.7 3.2 20 54 26

Two Person 1.8 7.6 6.5 22 54 24
Three Person 2.1 10.6 9.4 19 56 25
Four Person 2.4 13.6 1 1.8 19 58 23
Five Person Plus 2.4 16.6 14.0 17 62 21
Weighted Average 1.8 9.2 8.1 20 23
Urban Areae = 200,000 to 499,999
lncome
Low' 1.3 6.0 4.8 17 60 23

Medium 1.8 9.4 8.2 20 56 24
High 2.4 11.8 10.7 23 52 25
Weighted Average 1.8 9.0 7.8 2'-1 56 23
Household Slze
One Person 1.0 3.7 3.3 20 56 24

Two Person 9.9 7.1 6.4 23 53 24
Three Person 2.1 10.8 9.8 22 54 24
Four Person 2.2 13.4 11.2 18 61 21
Five Person Plus 2.4 15.9 12.9 19 59 22
(continued on next page)

30
TABLE 9 (Contnued)
7o Average Daily Person

Trips by Purpose
Average Daily Average Daly
Average Autos Person Trips
Vehicle Trips
per Household per Household per Household HBW HBO NHB
...' :
Urban Area = 500,000 to 999,999
lncome
';l
'.j
..1
.,i Low' 1.1 6.0 4.8 18 59 23
Medium 1.7 8.9 7.5 23 ss 22
High 2.3 11.5 10.3 22 s4 24

-::: I
Household Size
''l One Person 0.9 4.0 3.5 23 s4 23

Two Person 1.8 7.3 6.7 24 s3 23
i
Three Person 2.O 10.2 8.8 23 54 23
.l
-'--
Four Person 2.3 13.0 10.6 21 57 22
'-.., Five Person plus 2.4 15.4 12.5 18 62 20
Weighted Average 1.8 8.7 7.5 22 s6 22

:.*.'J
Urban Area = 1,000,000 +

-l
lncome
Low' 1.2 5.7 3.8 16 62 22

Medium 1.8 9.0 6.9 21 56 23
High 2.4 10.8 8.9 24 51 2s
Household Size
One Person 0.9 4.2 3.1 23 s0 27

Two Person 1.7 7.3 5.9 25 52 23
Three Person 1.9 9.3 7.7 25 52 23
Four Person 2.2 12.O 9.9 21 59 20
Five Person plus 2.3 14.8 11.2 19 62 19
Weighted Average 1.7 8.5 6.9 22 s6 22
ln actual 1990 dollars: Low = less than 520,000, Medium = 520,000 to 39,999, and High = 540,000 and up.
3l
trips would be multiplied by 0.88 (average daily vehicle area. With a generalized land-use map providing land allo-
trips/average daily person trips, or 8.I + 9.2 : 0.88). The cated to various uses and with more specific details on cer-
results would be 526,240 vehicle trips for the region tain generators (e.g., educational facilities and hospitals),
(598,000 x 0.88 :526,240). total trip ends by analysis unit (e.g., block, zone, and district)
Table 8, Person-Trip Attraction Estimating Relationships can be developed for residential and nonresidential uses. The
for All Population Groups, allows estimation of trip attractions data in Table 4 or Table 5 can be used to develop a control
by purpose and by two area groupings, CBD and non-CBD. total for the entire study area, as well as to adjust the total trip
For the example described previously, total area production ends developed from Table 3.
controls were as follows: To calculate trips produced in a study area (e.g., region,
zone, district, and corridor), an estimate is required of the
Trip Purpose Trip Production Controls number of households by household size and one of the two
HBW 119,600 "wealth" vaiables-either income range (tertiles in 1990
HBO 340,860 dollar base) or autos owned. With this estimate, the user can
NHB t3'7,540 obtain from the tables for a particular urbanized area size the
Total 598,000
. Average autos per household,
Assume also that the analysis area for the example has the . Average daily person trips per household,
following employment/residential mix: . Percent households by autos owned,
. Average daily person trips per household by number of
Employment Type CBD Non-CBD autos per household, and
Total Employment 5s,000 20,000 . Percent average daily person trips by purpose.
Retail Employment 16,000 10,000
Service Employment 18,000 7,000 For example, if 1,000 households are in the low-income
Other Employment 21,000 3,000 range in an urbanized aea of 200,000 to 499,999 population,
Dwelling Units 10,000 55,000 two methods can be used to compute the trip generation for
these households. The first method is the aggregate method,
Trip attractions for the analysis area can be developed using which shows that 1,000 low-income households in an urban-
the relationships presented in Table 8 as follows: ized area of 200,000 to 499,999 population should produce
6.0 trips per household (Table 7). Therefore, a total of 6,000
Trip Attractions for CBD trips would be produced by these households.
The second method for computing the trip generation for
HBW Attractions = 1.45 x 55000 = 79J50 these 1,000 households is the disaggregate method. In this
HBO Attractions = (2.0 x 16000) + (l .7 x 18,000) method, the households are first disaggregated into the
+ (0.5 x 21000) + (0.9 x 10000) expected number of households by auto ownership level. The
= 82,100 trip generation is calculated for these households and then
NHB Auractions = (1.4x 16000) + (1.2 x 18,000) aggregated to estimate the total trip generation. This method
+ (0.5 x 21,000) + (0.5 x 10,000) is summarized as follows:
= 59,500
Trip Generation Data for Low-Income Households-
Trip Attractions for Non-CBD Disaggregate Method (example)
HBW Attractions =1.45x 20000 = 29000 . Households with 0 cars lTVo or 170 HHs
HBO Attractions = (9.0 x 10,000) + (1.7 x 7000) (Table 4)
+ (0.5 x 3000) + (0.9 x 55,000) . Households with I car 5l%o or 510 HHs
= 152,900 . Households with 2 cars 247o or 240 HHs
NHB Attractions = (4.1 x 10,000) + (1.2 x 7000)
. Households with 3 * cars 87o or 80 HHs
+ (0.5 x 3000) + (0.5 x 55000) . Trips by 0 car households 2.8lHHor 476
= 78,400 (Table 7)
. Trips by 1 car households 4.9lHHor 2,499
Note that the calculated attractions do not equal the produc- . Trips by 2car households 8.6/HH or 2,062
tion control totals. A procedure for balancing these numbers . Trips by 3+ car households 11.5/HH or920
will be discussed later in this chapter. . Total trips by all households: 5,957
. Trips generated: HBry' lTVo or l,Ol3
It is possible to use the material provided in Table 3 for (Table 9)
specific activity sites to develop trip-end estimates through . Trips generated: HBO 607o or 3,574
aggregation for an entire urbanized area, corridor, or small . Trips generated: NHB 237o or 1,370
'''i
..t : l
,..t 32
i
'i
.l
.'':
'.-'t.l These two methods result in trip generation rates for the productions and attractions consist ofproductions and attrac-
study area that vary by less than 1 percent. tions from TAZs, and productions and attractions from exter-
,,] ,'.1
i
The HBW and HBO trips are generated at the household, nal stations.
whereas the NHB trips are generated elsewhere. Productions and attractions at external stations are a func-
'-
.-.i
'i
i
At the end of the application of the trip generation data tion of observed and forecast traffic counts at the stations
presented here, the user will have a matrix of zones with the and, therefore, are fixed and not factored. Productions and
necessary information-that is, the total productions (person attractions at external stations, however, will not be in bal-
! '.! trips by purpose) and the attractions for each zone-to move ance, and the difference must be balanced across zonal trips.
'; on to the next step, balancing production and attractions. As noted earlier, zonal trips include I-I, I-8, and E-I trip ends.
i
The following equation is used to obtain the control total of

'
. '1
..t
trip productions:
a ..
BALANCING PRODUCTIONS
AND ATTRACTIONS
cTP=>P+>n->A" (3-1)
The last step in trip generation modeling is the balancing
of regional trip productions and attractions. Because of the where
nature of the survey data used in this chapter to derive trip
generation rates, the resulting productions and attractions CT.o: the control total of productions,
include both internal-internal (I-I) and internal-external (I-E) P. : trip productions for each zone,
.l
trips by residents of the study area. In Chapter 5, definitions n : fip productions at each external station, and
i of I-E, external-internal (E-I), and through trips made by both A" : tnp attractions at each external station.
..1
i
residents and nonresidents of the study area are presented.
!
That chapter also presents parameters and a technique for Control totals are computed for each trip purpose. The next
.--*. estimating external travel. Because the distribution of I-E and step is to compute the balancing factor for each trip purpose,
l
E-I trips is accomplished by using the gravity model struc- using the following equation:
ture presented in the next chapter, the regional total of trip
CT.-
productions (including trips produced at the external cordon) Factor = (3-2)
ffi must be equal to the total of trip attractions (again, including ti
the trips attracted at the external cordon).
The estimated total nips produced at the household level where
should be equal to the total trips attracted at the activity cen-
ters. Each trip must have two ends-a production and an Factor: balancing factor (computed for each trip pur-
attraction. In reality, the estimation of trip productions and pose), and
attractions will not be exactly equal. While trip production and A,: Irip attractions at each zone (by purpose).
attraction rates may contribute to the imbalance, the majorify
of the difference can be explained by the estimation of the Each zone's trip attractions are then multiplied by the bal-
number of households, the socioeconomic characteristics of ancing factor (by purpose) to arrive at the final balanced trip
the households, and the estimation of the number of employ- attractions. The equation to use is
ees by type. To bring the regional totals in balance, either the
zonal productions or the attractions are scaled to equal regional A'.=FactorxA- (3-3)
control totals. In the majority of cases, the control totals of trips
are the regional totals of trip productions by purpose. This is where
because we generally have a greater degree of confidence in
household data than we do in employment data. This is partic- Ai: balanced trip attractions for each zone.
ularly true when the 100 percent decennial census data are
used to develop the number of households by zone. The External station attractions are not factored. A final summa-
employment data from which the attractions are computed are tion of productions and attractions by puqpose for all zones
less certain, not only on a regional basis but, more critically, at and external stations should be made, and they should be
theTAZlevel of geography. There are cases where trip attrac- equal. Any special generators should be handled in the same
tions are used as the control total and productions are scaled to manner as external stations in the matrix balancing process.
match attraction totals. Special generators and subarea studies The last step is to replace the zonal NHB productions with
with detailed information about employment and trip attrac- zonal NHB attractions.
tions are examples of instances where attractions would be the By definition, home-based trips (HBW and HBO) have
control for the balancing process. one end of the trip in the zone in which the household is
Regional control totals must be calculated to balance pro- located. For non-home-based trips, we know how many
ductions and attractions. This is done for each purpose- NHB trips are made by the household; however, we do not
HBW, HBO, and NHB. In this procedure, zonal attractions know where they take place. The regional total of NHB trips
will be balanced to match regional productions. Regional produced by the households is judged to be the best estimate
of the control total of NHB trips, but the NHB attractions are may vary by location in an urbanized area, by size of urban-
judged to be the best estimate of where these trips take place. ized area, and by location within the United States (e.g.,
Therefore, the zonal NHB productions are replaced by the East Coast versus rWest Coast). However, the material
balanced NHB attractions. While home-based trips do not does provide estimates that are useful for many applica-
have to be equal by zone, NHB productions will be equal to tions.
NHB attractions for each zone. The trip data and car ownership data in Tables 4 through
The following example illustrates the balancing process. 7 and Table t have been summarized from a number of
Note that there are fewer trips for the external stations than urbanized areas and grouped by population. Consideration
would be typically found in order to have the proper ratio of must be given to the fact that trip rates for a specific area
zonal and external station trips for this example. Any special may vary significantly based on special characteristics
generators would be treated the same as an external station (e.g., high proportion of retired persons and high tourism).
in the balancing process. Local data will provide more specific information for
Unbalanced Work Trips Balanced Work Trips

Zone Productons Attracons ffi
1 100 4,000 100 4,540
2 300 3,000 300 3,405
3 500 5,000 500 s,675
4 1,000 1,000 1,000 1,135
5 1,500 1,200 1,500 1,362
6 1,000 1,500 1,000 1,703
7 5,000 500 5,000 568
8 7,500 100 7,500 114
9 3,000 1,500 3,000 1,703
10 1,000 2,000 1,000 2,270
Subtotal 20,900 19,800 20,900 22,475
Extemal Station
11 500 50 500 50
12 1,000 100 1,000 100
13 250 25 250 25
Subtotal 1,750 1,750 175
Total 22.650 19.975 22.650 22,650
*Calculated by multiplying unbalanced attractions by f .1351 (see below).
The balancing factor for the above example is computed as the area in question than the material presented in this
chapter.
20,900+rJso-175
Factor = Although the data used for developing the parameters pre-
19,800
sented here showed little difference across regions or
Factor = l.l35l between urban sizes, it is clear that the urban form has an
immense impact on the transportation system and that the
The balanced trip productions and attractions are now ready transportation system has an immense impact on urban
to be used as input to the trip distribution model presented in development. The shape individual cities assume as they
the next chapter. grow affects the transpofation system by influencing travel
demand and the provision of transportation infrastructure
LIMITATION OF DATA and services. In turn, the provision of services and infra-
structure determines the relative accessibility of various land
In using the information in Table 3, the user must keep parcels and thereby affects land use decisions and urban
in mind that the values given are averages and that they form.
34
Disaggregate analysis, such as the cross-classification at where

the household level reflected by the material presented here,
HH(n) : the number of households with n occupants.
produces results that can be applied at any level for which
land use and related characteristics can be developed. Like-
These equations were used to calculate the trip produc-
wise, at the nonresidential end, sufficient disaggregation is
tions for each of the 107 internal zones in the Asheville
desirable to allow a detailed accounting for any specialized
MSA. The total number of trip productions for the region is
land uses in the area of study. 3 8 3,006, which includ es 7 6,033 HBW produc tions, 21 5,407
HBO productions, and 91,566 NHB productions.
CASE STUDY
Trip Attractions
The Asheville, North Carolina, case study was introduced
in Chapter 2 along with a presentation of the transportation Trip attractions were also calculated on a spreadsheet
networks and the socioeconomic data for that region. The using the parameters from Table 8. The input data for these
techniques described in this chapter have been followed to calculations include the employment by type-specifically
obtain trip productions and attractions and to balance pro- retail, service, and other employment-and the total house-
ductions and attractions. holds for each of the lO7 TAZs.
The home-based-work trip attractions for all 107 internal
zones are calculated using the following equation:
Trip Productions
Home-Based Work Attractions = 1.45 x Total Employment
The estimation of trip productions using disaggregate
travel-demand models typically uses a cross-classification of For the other two trip purposes, home-based other and
household size data with a measure of wealth, such as income non-home based, two different equations are used for calcu-
or the number of automobiles available to the household. For lating the number of attractions for each TAZ, depending on
-,..'l this case study, however, no cross-tabulations were yet avail- whether the zone is located within the CBD. In the City of
able from the Census Transportation Planning Package for Asheville, zones I through 15 are considered to be within the
--l the Asheville region. As a result, trip production rates were CBD, and the remaining zones, 16 through 107, are consid-
calculated by using the average values for the region, strati- ered to be in the non-CBD category. The trip attraction rates
fied only by household size. The average daily vehicle trips for CBD zones 1 through l5 are calculated using the follow-
ing equations:
per household are taken from Tables 5, 6, and 9 .
Since the trip productions are classified according to trip
Home-Based Other Attractions =
purpose, the information regarding trip purpose by house-
2.0 x Retail Employment + 1.7 x Service Employment
hold size was also used. These data can be found in Table 9.
+ 0.5 x Other Employment + 0.9 x Total Households
The calculation of trip productions is performed easily
N on- H ome - B as e d Attractions
with a computer spreadsheet. The input data can be arranged
1.4 x Retail Employment + I.2 x Service Employment
in five columns reflecting the household size groupings from + 0.5 x Other Employment + 0.5 x Total Households
one-person households through five-plus-person households.
The output data are the three columns reflecting the trip pur- The trip attraction rates for non-CBD zones 16 through
poses-home-based work, home-based other, and non-home- 107 are calculated using the following equations:
based. The person trip calculations are expressed by the
following formulas: Home-Based Other Attractions =
9.0 x Retail Employment + l.l x Service Employment
Home-Based Work Productions = + 0.5 x Other Employment + 0.9 x Total Households
0.20 x 3.7 x HH(I) + 0.22 x 7.6 x HH(z) Non- Home- Based Attractions =
+ 0.19 x 10.6 x HH(3) + 0.19 x 13.6 x HH(4) 4.1 x Retail Employment + 1.2 x Service Employment
+ O.l7 x 16.6 x HH(5) + 0.5 x Other Employment + 0.5 x Total Households
Home-Based Other Productions =
0.54 x 3;7 x HH(I) + 0.54 x1.6x HH(2) For all internal zones, the trip attractions in the region
+ 0.56 x 10.6 x HH(3) + 0.58 x 13.6 x HH(4) totaled 383/41, of which 85,604 were HBW trips, 188,806
+ 0.62 x 16.6 x HH(5) were HBO trips, and 109,331 were NHB trips. These totals
Non-Home-Based Productions = reflect the unbalanced attractions before they are matched to
0.26 x 3.7 x HH(I) + 0.24 x 7.6 x HH(2) the productions in the region. Appendix B-3 lists the pro-
+ 0.25 x 10.6 x HH(3) + 0.23 x 13.6 x HH(4) ductions and attractions for the three trip purposes for each
+ 0.21x 16.6 x 11(5+) of the 107 internal TAZs.
35
TABLE 10 Unbalanced trips
Productions Attractions
HBW HBO NHB NBW HBO NHB
lnternal 76,033 215,407 91,566 85,604 188,806 109,331

External 48,U2 62,986 26,087 20,932 41,990 26,087
Total 124,875 278,393 117,652 106,536 230,796 135,418
TABLE ll Balanced trips
lnlernal 76,033 215,407 91 ,566 236,402

103,9,13 91,566
External 48,U2 62,986 26,087 20,932 41,990 26,087
Total 124,875 278,393 117,6s2 124,875 278,393 117,652
Balancing Productions and Attractons Similarly, balancing factors are calculated for the other trip
purposes as follows:
The final step in the trip generation phase of travel-demand
forecasting is the balancing ofregional trip productions and
HBO Factor = (278,393 - 41,990) + 188306 = 1.2521
attractions. The next step of the model, trip distribution, NH B F actor = (l l'7,652 - 26,087 ) + 0933 I = 0.8375
1
requires that the total number of regional trip productions After the balancing factors are applied, the total numbers
equal the total number of regional trip attractions for each of
of productions are as summarized in Table 11. The total num-
the trip purposes. Table l0 summarizes the internal trip and
ber of attractions for the internal and external zones is then
external trip totals (see Chapter 5) before balancing.
520,920, which matches the total productions for the region.
The regional control totals for all three trip purposes are The final step in trip balancing is updating the non-home-
set to equal the combined internal plus external trip produc-
based trips. Although the zonal estimates for NHB trip pro-
tions. For example, the control total for home-based work duction are useful for determining the total number of NHB
trips is set at 124,815 trips. The balancing process is accom-
trips, they are not useful for determining where the trip is
plished by applying a balancing factor to the attraction trips
located because neither trip end for an NHB trip is at the
for all internalTAZs. The balancing factor is designed to household. After the NHB trip attractions are scaled so that
change the total number of internal attractions so that the
the total attractions equal the total productions, the NHB
total number of attractions, including external stations, trip productions in each zone are set equal to the NHB trip
equals the total number of productions. Following the same
attractions.
example for home-based work trips, the goal is to factor the
The balanced productions and attractions for the three
85,604 internal HBW attractions so that the total number of
trip purposes are listed in Appendix B-5. These values for
attractions equals the total number of productions. In order
the productions and attractions are ready to be used in the
to factor the 106,536 total HBW attractions to equal the
trip distribution phase of model development in order to
124,875 productions, the internal HBW attractions must be
prepare the person trip tables. Therefore, the production
factored to equal the total number of attractions, minus the
and attraction data have been imported into the travel-
number of external HBV/ trips (since external trips ae based
demand forecasting software and saved in three origin vec-
on existing traffic volumes, they are not factored). The bal-
tors for the production data for the three trip purposes and
ancing factor for HBW trips is therefore calculated as
three destination vectors for the attraction data for those
HBW F acto r = (124,8' l 5 - 20,932) + 85,604 = 1.2142 same trip purposes.
36
CHAPTER 4
(..1
"_'t
..1'i
i.
.i
TRIP DISTRIBUTION
..1
.x
,:
i.l
.l
INTRODUCTION The most common form of model used for trip distribution
is the gravity model. The gravity model is used to distribute
Trip distribution is the second major step in the travel the trip productions and attractions estimated by the trip-
modeling process. The first major step, Trip Generation generation model component. Gravity models are imple-
..i
''I (Chapter 3) provides a methodology for estimating trip pro- mented as mathematical procedures designed to preserve the
.,l
ductions and trip attractions or, in other words, how many observed frequency distribution of trip lengths for each mod-
trips have their start in each zone and how many trips have eled trip purpose.
their end in each zone. The gravity model structure can be calibrated and imple-
':l Trip distribution is the step that links the trip productions
't mented using conventional transportation planning pro-
I
to the trip attractions for each zonal pair. Productions and grams. Although the basic model structure is standard from
I
i
I
attractions are not the same as origins and destinations. The area to area, there are differences in the methods of estimat-
-_,j"
,..'l I tables that are produced by the gravity model must be con- ing calibration parameters and in the definition of the mea-
verted to origin-destination (O-D) format, after mode split, sure of separation between zones. The procedure to calibrate
before being assigned to the network. A method for convert- the gravity model is an iterative process in which travel time
ing these tables is presented in Chapter 8. or impedance factors are developed for each trip purpose
i
Trip distribution is a vital part of the planning process and a mathematical function, such as an inverse exponential
because it is the trip interchanges between each zone pair that function or a gamma function, is used to describe spatial
eventually have to be accommodated by the transportation separation.
system. Figure 5 provides schematic representation of the Gravity model procedures produce a trip table for each trip
trip generation and trip distribution process. In the schematic, purpose. The trip tables produced by these procedures can be
the upper lefrhand sphere represents a zone with 300 pro-
factored to represent the proportion of travel projected to
ductions and 700 attractions. The 300 trips produced in the
occur over an entire day, or any specific time period that
zone are distributed based on attractions in all three zones
needs to be investigated (see Chapter 8).
and distance between the zones. At the same time, the zone
In some areas, growth factor techniques are used to
is attracting 700 trips from all three zones.
develop future values for a known trip distribution by multi-
The critical factors of trip distribution are trip length and
plying interchange values by adjustment factors calculated at
travel orientation (e.g., suburb to CBD and CBD to suburb)
the origins and destinations of the trips. Such adjustments are
and the resulting magnitude of traffic and passenger vol-
based on changes in land-use and socioeconomic character-
umes. In summary, the trip distribution models are designed
istics, such as new development or a decline in family size in
to convert the trip production and attraction input data into
the zones of origin and destination.
trip tables that represent movements between the TAZs that
In this chapter, the gravity model concept is reviewed,
constitute the region being modeled. The results of trip dis-
along with its mathematical definition. This overview is fol-
tribution are assigned to the highway andlor public trans-
lowed by a discussion of the parameters required for trip
portation systems to determine the travel demand as related
distribution, such as freeway and arterial travel speeds and
to the carrying capacity of the facilities in question.
terminal times, and default friction factor curves, look-up
tables, and gamma functions by trip purpose. A method for
BASIS FOR DEVELOPMENT splitting trip tables for corridor and subarea applications and
an example application is included.
Trip-distribution models estimate trip interchanges between
zones based on characteristics ofthe land-use pattem and the
transportation system. Most distribution models have similar THEORY OF THE GRAVITY MODEL
characteristics; trip interchange between areas is a function
of the amount and type of land development and the spatial The most widely used trip distribution procedure is the
separation between zones. gravity model. As its name suggests, the gravity model for
5t
R F
v, fJ
n
,4,
4rdt
t.!
c,

*
I
\y eroAce
qv
/ P-550 \
Tdp Generalion Tdp Dlsibution
Figure 5. Diagrammatic representation of the trip generation and

trip distribution proce ss.
transportation planning is based on the gravitational theory and attractions in the attraction zone. In addition, the number
of Newtonian physics. The Newtonian theory predicts that of interchanges will be inversely proportional to the spatial
the force or attraction between bodies is directly proportional separation between the zones. For example, suppose that
to the mass of the bodies and inversely proportional to the from all of the zones in a region, four zones with the follow-
square of the distance between the bodies. Similarly, the ing characteristics were considered:
gravity model of transportation planning predicts that the rel-
ative number of trips made between two geographical areas Distance from
or TAZs, is directly proportional to the number of trip ends Zone Productions Attractions Zone A
(productions or attractions) in each TAZ and inversely pro- A 1,000 0
portional to a function of the spatial separation (or travel BO 200 10
time) between those two areas. CO 200 20
Therefore, zones with large amounts of activity tend DO 400 l0
to exchange more trips, and zones father from each other
tend to exchange fewer trips. Most modeling processes In the above example, the interchange between zones A
develop trip-distribution models for three to six trip pur- and D would be greater than the interchange between zones
poses. Typically, the number of purposes is defined by A and B because the number of attractions in zone D is
resource constraints (i.e., each additional purpose requires greater than the number of attractions in zone B and the dis-
added computer time and disk space to model) and the abil- tance between the two pairs of zones are equal. This
ity to project trip ends. One common set of purposes is HBW, demonstrates the direct relationship of trip attractions in the
HBO, and NHB. gravity model. Likewise, even though the attractions in
Trip distribution can be improved by additional stratifi- zones B and C ae the same, the interchange between zones
cation. For example, HBO includes all home-based and A and B would be greater than the interchange between
non-work trip purposes including shopping, school, social- zones A and C because the distance between zones A and
recreational, personal, and other trips. Shopping trips are B is less than the distance between zones A and C. This
attracted to retail locations and school trips are atftacted to edu- demonstrates the inverse relationship of spatial separation
cational institutions. If retail employment and school locations in the gravity model.
can be forecast, the HBO distribution can be improved by Mathematically, the gravity model for trip distribution is
stratifying the purpose into FIBShop, FIBSchool, and HBOther defined as follows:
(including home-based personal and social-recreational trips).
The gravity model trip distribution technique is an adapta- ll
tion of the basic theory of gravitational force.r As applied in T,, = P,l rr!^F'Ku I (4-r)
transportation planning, the gravity model theory states that
the number of trips between two TAZs will be directly pro- [ '-o-u* J
portional to the number of productions in the production zone
where
lBarton-Aschman Associates, lnc., Plan Regioml De Trapoarion San Jrcn, Sm

Juan, Puerto Rico (October 1992). P : the number of trip productions in zone i,
38
A; : the number of trip attractions in zone i, result of the trip-generation analysis performed in
F: the friction factor relating the spatial separation Chapter 3.
between zone i and zonej,and 2. A coded network. The basic network coding procedure
K,: an optional trip-distribution adjustment factor for is to first digitize the node coordinates, then to record
interchanges between zone i and zone i. the a-node and b-node combinations that are used to
define each roadway network link. Network coding is
The friction factor is the primary independent variable and outlined in Chapter 2.
quantifies the impedance or measure of separation between 3. Coded-link attributes such as link distance, functional
two zones. Friction factors are inversely related to spatial classification, and posted speeds for each link in the
separation of the zones-as the travel time increases, the fric- network.
tion factor decreases. It attempts to represent the behavior of
the traveler in terms of the perception of distance. The will- The development of travel impedance represented by friction
ingness of the traveler to spend time or distance on a trip factors and the implementation of the gravity model to cre-
varies by trip purpose. ate trip tables is discussed in the following sections.
A number of different functional forms have been used for
friction factors. In fact, early gravity models used hand-fitted
TRAVEL IMPEDANCES
friction factors. More recently, however, it has been discov-
ered that the gamma function does a very good job for trip One of the major inputs to a gravity model trip distribution
distribution. Using such a function produces a smooth, con- method is the creation of highway impedances. In areas with
tinuous curve. The gamma function can be stated as follows: extensive transit systems, some modelers use composite
impedances that incorporate both highway and transit times.
F,,=oxtbuxe"""; (4-2)
Composite impedances should be calculated with care
because the relationships of the composite's components are
where specific to the transit system being modeled. The weighing
F : the friction factor between zones i and j, of each component must be carefully developed. In deter-
a, b, and c : model coefficients; both and c should, in mining the travel impedance (path of least resistance
most cases, be negative; is a scaling factor between each pair of zones), travel times, distances, and/or
and can be varied without changing the dis- tolls are summed for the links between each zone pair and the
-l
tribution, results are stored in azone-to-zone travel impedance matrix.
:
t the travel time between zones i and j, and In order to calculate zone-to-zone travel impedance, the com-
e: fhe base of the natural logarithms. ponents that define impedance for the path-building must be
defined. Time and distance estimates require link length and
Some distribution models use k-factors, or socioeconomic link speeds to be identifled.
factors, to modify the results of the gravity model to more Link length may be part of the network definition
closely match real trip characteristics. For example, river (TIGEMine files carry link length, for instance) or it may be
crossings tend to require a "barrier" effect-that is, the calculated using the x and y coordinates of the a-node and
movement between zones separated by a bridge may not be b-node that define the link (see Chapter 2, Building a Trans-
as great as would be expected from using only quantifiable portation Database). The value of including distance in the
measures. The planner can use either k-factors or artificial impedance calculation is that, if faced with two paths of
times on the bridge links to match the actual interchange of equal or near-equal time, travelers will choose the path of
travel. lesser distance. In many urban areas free-flow travel times
are used for creating the travel impedances, although an
emerging trend is toward the use of congested speeds to build
TRIP DISTRIBUTION PROCESS shortest-time paths for both distribution and mode-choice
modeling.2
The trip-distribution process is usually implemented in Link-specific speeds can be first estimated by simply car-
two steps. The first step is the estimation of friction factors
rying posted speeds on the links or using Table 12 to estimate
based on estimated or existing travel times. The second step link speeds by facility type and area type.
uses the gravity model to distribute the trip productions and Once each link is defined with a length and a speed in the
attractions estimated by trip generation. attribute file, shortest travel time paths can be built, and zone-
The following information is necessary to proceed with to-zone travel impedance can be calculated.
the trip-distribution process :
1. Production and attraction trip ends by analysis area or ?Parsons Brinkerhoff

Quade & Douglas,lnc., Review of Best Practices, prepared for
TAZ for each trip purpose. These estimates are the Metropolitn Washington Council of Governments, (December 1992) p.2-\5
39
TABLE 12 Link speeds by facility type and area type
Area Type
Facility CBD Suburban Rural
Type Free FIow Free Flow Free Flow
Freeway 60 60 60
Expressway 45 45 55
Princpal Arterial Divided 35 45 50
Undivided 35 35 45
Major Arterial Divded 35 45 40
Undivided 25 35 35
Minor Arterial 30 35 35
Collector 15 30 30
Source: Various urban transportation studies.
INTRAZONAL TRAVEL TIMES miles per hour, and the intrazonal speed for a rural zone
could be set at 30 miles per hour.
The matrix of travel times produced by most software
packages will result in times for each zone pair but will not
include the travel time within the zone. This travel time TERMNAL TIMES
within the zone is called the intrazonal travel time. The trip-
distribution model will not only distribute trips between
Building the matrices of zone-to-zone travel times or
zones but will also determine the number of trips that stay
travel impedances (skims) also involves adding terminal
times to the over-the-network travel times. The nearest
within the zone. Therefore, the average travel time for trips
neighbor technique should be applied to the travel times
that stay within a given zone must be estimated. The intra-
before the terminal times are added. This is due to the fact
zonal trips largely take place on the local street network that
that the terminal times are a function of area type and the
are not coded. An approximation of the intrazonal time can
neighboring zones may have different area types. The termi-
be made by a number of techniques. One popular technique
nal times for the zones should be added after the nearest
is the neaest neighbor technique.
neighbor procedure is applied.
The nearest neighbor technique assumes that the travel
Terminal times represent impedances at both ends of a trip
time within a zone is equal to one-half the average travel time
such as the amount of time and the time value of money
to the nearest adjacent zones. Many software packages have
required to walk to and from a transit mode, to park or access
specific modules that apply this technique. The user may be
a parked car, to pay parking cost, and so forth. As such, ter-
asked to input either a list of neighboring zones for each
minal impedances vary by area type. Terminal times are
zone, or parameters for the program to find the nearest neigh-
added to the initial impedance times at both the origin and the
bors (using the centroid coordinates) and compute the aver-
destination end of a trip. Terminal times are typically esti-
age travel time.
mated as a function of population and employment density
Another technique assumes that intrazonal travel times
within a traffic zone or district. If terminal impedances have
can be expressed as a function of the aea of the zone and the
not been estimated for an analysis region, the default values
intrazonal speed. When this method is employed, the intra-
presented in Table 13 may be used.
zonal time lor aTAZ may be calculated using the following
equation:
CHOICE OF FRICTION FACTORS
Intrazonal Time = O.S x erea x AO
^[1Zo"at
* Intrazonal Speed (Area Type) Once the zone-to-zone travel times have been estimated,
the gamma function can be applied to calculate friction fac-
where the intrazonal time is expressed in minutes, the zonal tors for each zone pair. Friction factors from several cali-
area is expressed in square miles, and the intrazonal speed in brated models for smaller urban areas were used to produce
miles per hour varies by the area type of the zone. For exam- Figure 6. These curves correlate to the gamma function coef-
ple, the intrazonal speed for a CBD zone could be set at 15 ficients presented in Table 14. If these coefficients are used
40
i
TABLE 13 Terminal times for different area types 15 to 20 minutes in smaller communities, up to as much as
...
.l
1
25 to 30 minutes in large metropolitan areas.

'.1
..1
The closest correlation that we have found between aver-
',1 Area Type Terminal Time (minutes)
age trip length and urban area size relates the average trip
'-.-.i
length to the land area of the urbanized area. This correlation
CBD 5
between trip length and land area is slightly stronger than the
CBD Fringe 4 conelation between trip length and population. The average
Urban 3 trip length for home-based work trips can be estimated using
Suburban 2 the following equation:
Rural 1
Average HBW Trip Length (minutes)

: 5.0 + 0.10 x ^lLtrdAr*
in the gamma function presented in Equation 4-2, withtravel
Average trip length data for other trip purposes are not as
times (impedances) of 1 to 60 minutes, the friction factors
readily available as for home-based work trips. One problem
shown in Table 15 result.
-_i - -t that we encountered with respect to home interview data is
The friction factors in Table 15 should cover the range of
that other trip purposes are subject to a great deal of varia-
L travel times for most small urban areas. If travel times exceed
tion. Specifrcally, the number of home-based non-work trip
60 minutes, the table should be expanded. It is not wise sim-
purposes varies from one to five categories including school,
-l ply to assume the 60-minute friction factor for all times
college, shopping, personal, social-recreational, and other
greater than 60 minutes. This would imply that travelers are
trip purposes. Also, analysis of home interview survey data
insensitive to increases in travel times greater than 60 min-
collected for various urban areas has shown that the average
.i utes. Friction factors should not be rounded to zero, as this
trip lengths for other trip purposes are much less predictable
would imply that no trips are interchanged for the given time
than for work trips. For example, in some communities,
factor.
school or college trips are relatively short, indicating the
Two goals should be considered when choosing or calibrat-
presence of a centralized campus, while in other communi-
wi ing friction factors for use in a travel-demand model: (1) the
ties school or college trips are relatively long, resulting either
average trip length should be reasonable; and (2) the trip-
_-; from a commuter type campus or a school busing program.
length frequency distribution should be reasonable.
In fact, the correlation between the different data sources is
I
I so shaky that it is difficult to say whether home-based non-

I
Average Trip Lengths work or non-home-based trips are longer or shorter than the
other, in general.
The most desirable source with which to calibrate trip- The best correlation that we were able to find regarding
length data is derived from a home interview survey con- trip lengths for other trip purposes relates the average trip
ducted in the study area. If no home interview survey data length for home-based work trips to the average trip length
are available, census journey-to-work data can be used as a for all other trip purposes combined, which we will refer to
source of average trip-length data for work trips in every as non-HBW trips. For urbanized areas with populations of
Metropolitan Statistical Area (MSA) in the country. Average less than 500,000, the average trip length for the non-HBW
trip lengths for home-based work trips generally range from trip purposes is typically 75 percent to 85 percent of the
30
\\\
'l
!zo
G
c
'i
l
o
ro \\"
i
.l I
_i
i
fl 13 15 17 $21?32521 n 3l r s
Minuteo
+HBW )+HBO +NHB

Figure 6. Syntheticfrictionfactor curves by purpose.
4t
TABLE 14 Gamma function coefficients for friction factors
Trip Purpose
HBW 28,507 -0.020 -0.123

HBO '139,173 -1 .285 -0.094
NHB 21s,1 13 -1 .332 -0.'100
Note: for use in the following formula, the "a' is a constant scaling factor which may be omitted.
Fu=axtlxe""u
average trip length for HBW trips. (For larger metropolitan time increment (usually 1 rrinute). The data in Table l5 can
areas of over 1,000,000 population, the ratio ranges from 60 be used for these cases. Other programs allow for the input
to 70 percent of the HBW trip lengrh.) of a friction factor equation. In these cases, the gamma func-
tion can be input using the coefficients presented in Table 14.
If trip-length frequency distribution data are available
Trip-Length Frequency Distribution (from a home interview survey, an existing model, or census
data), an iterative process can be used to calibrate the friction
Along with the average trip length, another indicator of the
factors to match the observed trip-length frequency. Each
distribution oftrips in trip tables is the trip-length frequency
iteration of the calibration consists of the following steps:
distribution (TLFD), which describes the shape of the curve
that is summarized by the average trip length. For example,
1. Determine the number of observed trips in each time
Figure 7 displays the TLFDs for HBW and non-HBW trips
period for each purpose: HBW, HBO, and NHB. This
made in the San Francisco MSA. These curves display the
is called the observed trip-length frequency.
distribution of trip lengths that, in combination, average 24
2. Calculate an initial friction factor for each time period.
minutes per home-based work trip and 14 minutes per trip for
A spreadsheet software package makes this relatively
other purposes.
easy by filling one column with numbers from 1 to 60
Figure 8 displays a simila set of trip distribution curves for
(minutes) and applying the loglinear form of the
a much smaller urbanized area-Santa Barbara, California.
gamma function to each time increment. Initially the
Comparison of the two figures shows that the relationship
coefficients presented in Table 15 may be used.
between the HBW and non-HBW curves for the two regions
is remarkably similar. For both aeas, the distribution of non- The transformation of the gamma function to a log-
HBW trips peaks at between 5 and 10 minutes per trip and linear function is:
then falls sharply, with very few trips over 30 minutes in
length. The home-based work curves, on the other hand, peak ln(,f) : ln(a) + x ln(r) + c x t (4-3\
between 10 and 15 minutes and fall more gradually, with tails
approximately twice as thick as in the non-HBW curves.
The most striking difference between the HBW and non- 3. Current iteration friction factors for each time interval
(usually 1 minute) included in the trip-length frequency
HBW curves is the presence of very short trips in the non-
HBW trip purposes and the lack of these very short trips in distribution are factored by the ratio of the observed
number of trips in the time interval (from the observed
the HBW curves. Notice that the non-HBW curve for San
Francisco is quite different from the HBVy' curve for Santa
trip-length frequency distribution) divided by the cur-
rent number of trips in the time interval:
Barbara, even though they both represent an average trip
length of 14 minutes per trip. T:bs
F,'*t = F,' x (4-4)
7,,
Look-Up Tables Versus Formulas where
The application ofthe friction factors in the trip-distribution F,i+r - the friction factor for time interval t for iter-
model depends on the required format of the software pack- ation I * I,
age. Many packages require the friction factors to be input as F,' : the friction factor for time interval I for iter-
a look-up table with a corresponding factor for each travel ation i,
...1
'I
42
i
TABLE 15 Synthetic friction factors
Minutes
(impedance) HBW HBO NHB
1 25,214 126,652 198,293

'..': 2 21,990 47,295 71,303
3 19,291 25,2 37,607
4 16,963 16,O72 23,203
5 14,936 10,979 1s,601
6 13,161 7,W4 11,075
7 11,605 5,900 8,163
I 10,236 4,522 6,184
9 9,032 3,537 4,7A4
10 7,972 2,811 3,763
11 7,O37 2,263 2,999
12 6,213 1,841 2,417
13 5,486 1,511 1,966
14 4,845 1,2s0 1,612
15 4,2 1,041 1,33'l
16 3,780 872 1,105
I
17 3,339 734 923
18 2,950 620 774
l9 2,607 527 652
20 2,303 449 ss1
:--' 21
22
2,035
1,798
38l
329
467
397
-., 282 339
23 1,589
24 1,404 245 290
_, --,i.,- ",
1
25 1,241 210 248
26 1,097 182 213
' --l 27 969 1s8 184
28 857 137 158
29 757 119 137
30 669 104 118
31 592 90 102
32 523 79 89
33 462 69 77
g4 409 60 67
35 361 53 58
36 319 47 51
37 282 41 44
38 249 36 39
39 221 32 34
40 195 28 "30
41 172 25 26
42 152 22 23
43 135 19 20
44 119 't7 18
45 105 1s 15
46 93 13 14
47 82 12 12
48 7S 10 10
49 64 I I
50 57 I I
51 50 7 7
52 44 6 6
53 39 6 6
54 35 5 5
55 31 4 4
56 27 4 4
57 24 4 3
58 21 s 3
59 19 3 3
60 't7 3 2
43
E20
t-o
+ HBW Ttips (24 Mn.)
+ NooHBW Trip6 (14 Mh.)
o
o
0,
10
G5 11-15 21-25 31-35 11-15 51-55 615 7l-75 E15

6-10 fS20 230 3G40 ,l50 560 6&70 760 E6+
me (Minutes)
Figure 7. Trip-length frequency distribution for San F rancisc o, Calfo rnia.
Tfb' : the observed number of trips in time interval iterations, depending on the choice of values for the ini-
t, and tial iteration.
T,' : the estimated number of trips in time inter- The friction factors directly influence trip length (in min-
val for iteration i. utes). Therefore, the reasonableness of the friction factor is
reflected in the forecasted average trip lengths. The faster the
4. The revised friction factors are used as independent friction factor decreases (see Figure 4), the shorter the aver-
variables in a linear regression to estimate revised age trip length. The average trip length for home-to-work
gamma function model coefficients for the next itera- trips is available for urbanized areas as part of the census
tion. (The natural log of the revised friction factor is the
journey-to-work data.
dependent variable. The time increment and the natural
log of the time are the independent variables.) The CREATION OF TRIP TABLES
coefficients of the converted log-linear regression
model can be taken directly from the regression output With an estimate of the impedance (friction factors)
or estimated with a statistical analysis program. between zones, and the attractions and the productions within
each zone, the trips can be distributed from all zones to all
The iterative calibration process stops when the model other zones. This process includes iterations to balance the
coefficients become stable. This normally takes four or five matrices to ensure that the total attractions and productions
(!
0
l-
3zo r HBW Trip (14 Mh.)
c,
a)
o
+ NontlBWTri(11Mh.)
o
fL
G5 11-15 21-25 3l-35 41-45 51-55 615 71-75 815

&10 1&20 230 340 l&50 560 6&70 76{0 E6+
ime (Minutes)
Figure 8. Trip-Iength frequency distribution for Santa Barbara, Calirnia.

44
-l
(row and column totals) match the total productions and The splitting of a zone into smaller zones is the splitting of
attractions in each zone. The most common way to balance trip productions and attractions of the old zone. The splitting
the zonal productions and attractions is the Fratar method. of the trip table should always be done by purpose. The fol-
The trip tables that are produced by the gravity model are lowing steps can be followed to split a trip table:
in the production-attraction format. Before these trips can
be assigned to the network they normally are converted to
origin-destination format. Chapter 8 outlines a procedure 1. Compute productions and attractions for new zones, by
to create an origin-destination format from a production- purpose, using the new zones' socioeconomic data. The
attraction trip table. Production-attraction format trip tables total productions and attractions for the new zones
are used as an input to the mode-choice model, primarily must match those for the old zone.
because for non-auto modes the production and attraction 2. Compute percent of old zone's productions and attrac-
ends often have different access modes. For example: at the tions for each new zone.
home end of a work trip, the traveler might drive to the bus
3. Compute trip interchange percentages for the new
stop or rail station, but at the work end, the traveler must
zones by multiplying the percent productions times the
walk because he or she does not have access to a car.
percent attractions for each ofthe zone pairs. The sum
If a subarea analysis is being performed, then the follow-
ofall percentages should equal 100 percent.
ing section will be useful.
4. Multiply the percentages obtained in Step 3 by the old
zone trips to obtain new zone trip interchanges. Check
METHOD FOR SPLITTING TRIP TABLES
FOR CORRIDOR AND SUBAREA the sum of new zones' trips to see if it matches the old
APPLICATIONS zone's total.
A smaller sized zone structure is required in corridor or

i
subarea studies that use the travel-demand model. This pro- The following example illustrates this procedure
=a Trips from
S Zone 1 lo Zone 2 Old Zone New Zones
*.---.]I
Prr = 700 70%

Pr = 1,ooo Prz = 3oo 30%
200 Azr = 500 25%

Az = 2,000 75/"
422 = 1,500
New Zone Percentages New Zone Trips

Attraction Zone Atlraction Zone
Production Zone 2'l 22 21
11 17.5% 52.5"/" 35 105

12 7.5% 22.5/o 15 45
duces traffic assignment for a finer level of roadway analy- where

sis. The first step in this type of analysis is to modify the trip
P : Productions in old zone l,
table for the desired new zone system. This requires the split-
A2 : Attractions in old zone 2,
ting of zones within the study area and aggregating zones
Py : Productions in new zonej of old zone 1, and
outside the study area to districts. The aggregation of zones
Azj : Attractions in new zone j of old zone 2.
to districts is done by simply adding the zones' trips within
the new district together. Historically this aggregation was Using a spreadsheet or database program, the user can cre-
done to reduce the run time on the mainframe computer. ate a table of equivalency of new zones to old zones and the
With the power of the microcomputer application of the percentage of old zone productions and attractions in each
travel-demand model, this aggregation may not be necessary. new zone. Then the new zone structure trip table can be cre-
The splitting of the trip table to smaller zones and a proce- ated. Some transportation planning packages include mod-
dure for splitting the zones are presented below. ules to split trip tables.
45
CASE STUDY travel time matrix, which is essentially the diagonal of the
107 by 107 matrix of internal zones, was produced using
The distribution process for the Asheville, North Caolina, the nearest neighbor method. This consisted of identifying
case study was implemented in two parts. First, a matrix of the zones adjacent to eachofthe 107 internal zones, taking
zone-to-zone travel times was estimated, and then the grav- the free-flow travel time from the zone of interest to all
ity model was used to produce a full trip table for each trip adjacent zones, calculating a mean for that set of times, and
purpose. halving that value to arrive at the assumed intrazonal travel
This distribution process was repeated in order to use con- time. These steps resulted in intrazonal times ranging from
gested travel times as the measure of impedance for the ulti- 0.54 minute to 4.33 minutes, with an average value of 1.85
mate traffic assignments. The free-flow times were used to minutes.
perform the first distribution of trips for each trip purpose.
These trip tables were converted into a daily trip table and
assigned to the highway network with an equilibrium assign-
Terminal Times
ment. The result of this assignment was a set of traffic vol-
umes and congested speeds for each link in the highway
Terminal times represent impedances at both ends of a
network. These congested speeds were used to produce a sec-
trip, such as the time required to park or access a car, park-
ond matrix of congested travel times, which were used to per-
ing cost, etc. For the Asheville case study, the study area was
form a second application of the trip-distribution model, assumed to have three kinds of area types: CBD, suburban,
which resulted in the ultimate trip distribution.
and rural. Zones I through 15 are designated as CBD zones
and all trip-ends at those zones have a terminal time of 5 min-
utes. Zones 16-59,6I,62,65-1I,14,77-80,82, and 83 are
Estimation of Travel Times designated as suburban zones and have terminal times of 2
minutes for each trip-end. The remainder of the internal
The free-flow zone-to-zone travel matrix was constructed zones are designated as rural zones and have a 1-minute ter-
using the default speeds posted on each link in the Asheville minal time associated with all trip-ends. The average termi-
highway network. Speeds were set at 55 mph on freeways, nal time for the 107 internal TAZs is 2.2 minutes.
45 mph on major arterials, and 35 mph on minor arterials. The total travel time was calculated for each zone-to-zone
Travel time was calculated using the simple relationship: pair by adding the terminal times at both the origin and des-
tination ends of the trip to the free-flow travel time or to the
Link Length (in Miles) x 60
Travel Tme (in minutes) = intrazonal travel time in the case of intrazonal zone pairs.
Speed (in mph)
The transportation planning software was used to produce

the matrix of travel times, or travel skims, based on the min-
Choice of Friction Factors
imum time path between each pair of zones. These free-flow
travel times were based on speed and distance only; no vol- The gamma function was used to estimate the travel
ume delay was included. impedances between zones in the Asheville region. The cal-
culation of friction factors for each zone pair was performed
within the travel-demand model software by using a matrix
lntrazonal Times calculator. The preliminary friction factors used in this case
study were produced by using the gamma function coeffi-
The free-flow matrix did not contain any intrazonal cients listed previously in this chapter.
travel times, which represents the travel time required to The coefficients listed below for the three trip purposes
make a trip wholly within a single TAZ. An intrazonal were used for the first application of the gamma function:
Trip Purpose
Home-Based Work 100 -0.020 -0.125

Home-Based Non-Work 100 -1.300 -0.100
Non-Home-Based 100 -1.350 -0.100
..1
46
For example, the home-based work friction factors were tend to experience congested traffic during their home-based
calculated using the equation: work trips. The average modeled travel time of 16.9 minutes,
achieved with the use of the default parameters, passes this
HBW Friction Factoti,, = 100 x Timeo,ozo x erinci'ix({t251 reasonableness check.
The second reasonableness check suggests that the average
for all interchanges between origin zone i (from I to 123) and trip length for home-based other trips and non-home-based
destination zonej (from I ro 123). trips should be approximately 80 percent of the average
The final step in the calculation of the friction factors was home-based work trip length. For the Asheville MSA, this
to set the values for the external-external zone pairs to zero. corresponds to 15.0 for home-based non-work and non-
This adjustment was performed in order to prevent the grav- home-based trips. The modeled results, using the default
ity model from distributing any trips to external-external coefficients for the gamma function, produce average trip
zone pairs. (Otherwise, the distribution model and the lengths of 14.4 minutes and 14.8 minutes for home-based
through traffic model would combine to overestimate the non-work and non-home-based trips, respectively. Since
number of through trips made in the region.) A friction fac- these values are based on free-flow speeds, they are well
tor value of zero was used to replace the calculated value within the range of reasonableness.
for trips with both origin and destination zones between 108 Aside from trip lengths, another way to check the reason-
and I23. ableness of the trip-distribution results is by comparing the
trip table data to any data regarding observed travel patterns.
Such data, if they existed, would usually come in the form of
Creation and Assignment trip movements between districts or groups of zones, which
of Free-Flow Trip Tables
could be compared to the model estimated interchanges.
After the friction factor matrices were created for the three Unfortunately, such data do not exist for the Asheville MSA.
trip purposes, the trips were distributed using the gravity
model component of the travel-demand modeling software. Creation and Assignment
The balanced productions and attractions by trip purpose of Congested Trip Tables
were set as the row and column control totals. The Fratar
method was then applied to the trip ends so that the row and The estimation of congested travel times is similar to the
column totals matched the total productions and attractions process used to estimate the free-flow travel times. The major
in each zone. difference between the two processes is that the travel time
The output of the distribution process was a set of three function used to build the congested travel times assumes
person-trip tables. These matrices contain the same number of that the link speeds are subject to volume delay. The person
trips as the trip-generation control totals: 124,815 home-based trips resulting from the free-flow trip distribution were con-
work trips, 278,393 home-based other trips, and ll7 ,652non- verted to vehicle trips, and the production-attraction format
home-based trips. However, unlike the production and attrac- was changed to origin-destination format. An equilibrium
tion vectors, the person-trip tables are two-dimensional and assignment was performed and the congested travel times
reflect the movement of trips between zones. between zones were saved in a matrix. The congested trip
Since the distribution oftrips is calibrated to the trip length time matrices were completed by calculating the congested
in minutes for each trip purpose, it is useful to review the trip intrazonal times and adding the terminal times to the congested
lengths after applying the gravity model. Average trip length travel times. The revised friction factors were calculated and
was obtained by weighting the free-flow travel time matrix, the gravity model was applied to create new person-trip tables
including intrazonal times and terminal times, with the person- based on the congested times.
.':r trip tables. The trip lengths for the first application of the Average trip lengths were also calculated for the revised
gravity model are as follows: home-based work-l6.9 min- person-trip tables based on the congested travel time. Once
utes; home-based other-14.4 minutes; and non-home- again the travel-time matrices were weighted by the person
based-14.8 minutes. trip tables. Model-produced congested trip lengths were:
Two reasonableness checks should be performed on these home-based work-Il .l minutes; home-based other-14.9
results. First, the average trip length for home-based work minutes; and non-home-based-15.4 minutes.
trips resulting from the gravity model should be compared to The estimated travel time increased slightly for all three
the average home-based work trip length derived from the trip purposes as the result of the congested travel times. The
1990 census data. The l990journey-to-work statistics show Asheville region probably does not experience a great deal of
that residents of the Asheville MSA reported an average congestion, which can help explain the relatively small
home-based work trips length of 18.7 minutes. The average degree by which the average trip lengths changed.
trip length produced by the gravity model, using free-flow Since the average trip length for home-based work trips
speeds to build the impedances, should be slightly less than was approximately I minute less than the target value,
the average trip length reported by actual commuters, who the researchers performed two more iterations of the trip-
47
distribution model to achieve a more acceptable average The resulting trip tables had average trip lengths of 18.7 min-
travel time for this trip purpose. The coefficients listed below utes for HBV/, 15.0 minutes for HBNW, and 15.4 minutes
were the ultimate coefficients used to apply the gamma func- for NHB trips. The trip-length frequency distribution curves
tion for the three trip purposes: for these three trip purposes are displayed in Figure 9.
Trip Purpose
Home-Based Work 100 -0.300 -0.070

Non-Home-Based 100 -1 .350 -o.100
g. 30

tE
o
r Home-Based Work
Ezo
o
+ Home'Based Other
E
o
+ Nonflom+.Based
o
o
10
G5 11-15 21-25 3't-35 4145 5r-55

10 120 2&30 36-40 l&50 5660
Mnutes
Figure 9. Model-generated trip length distribution-by trip purpose.

48
CHAPTER 5
EXTERNAL TRAVEL ESTI MATION
INTRODUCTION The first method provides data only on through travel and
does not allow for the estimation of observed extemal-intemal
External trips are trips that have at least one end outside or intemal-extemal travel. The second method, although pro-
the study area defined by an encircling cordon line. When viding data on all extemal travel, has the disadvantage of a
both the origin and destination of a trip are outside the cor- definite time lag between the time the trip is actually made
don line, the trip is termed a through trip or external-extemal and the time the survey form is received by the driver. Even
trip. When one trip end is outside the study area, the trip is with direct entry of the plate number into a computer and
classified as an extemal-internal or internal-extemal trip. The overnight matching of numbers to registrant, it is at least 3
point on the roadway where the area cordon is crossed is days (and more likely 4 to 5) before the registrant receives
referred to as an external station. Figure 10 displays the var- the survey forms. The registrant may not recall exactly where
ious types of external travel. the trip was made or in some cases was not the driver of the
" "---1
-"1 I
Because of the small proportion of external travel relative vehicle. For these reasons, the roadside intercept is still the
to total travel, the effort on measuring and modeling extemal most cost-effective method for obtaining external travel data.
I
travel has been less intensive than for internal travel. How- Techniques for estimating the number of trips generated
:=- ever, while the percentage of total travel that is external may
:ffi1 i
within an area were discussed in Chapter 3. Depending on the
i be small, decisions regarding improvements to facilities that size and geography of the study area, a majority of these trips
.5-: cany high percentages of external trips must be made with
I
will take place completely within the study area. The larger
some degree of confidence in the estimate of extemal travel the study area's geographic limits, the less impact that exter-
behavior. Very little is known about the population and nal travel has on total travel.
employment characteristics at the end of the trip that is out- This chapter presents a method for estimating external
side the internal study area. Travel is measured in vehicle travel in a study area where an external survey is not avail-
trips instead ofperson trips, and transit trips from outside the able or possible. This step is typically done before trip dis-
region are often ignored. Future-year external travel is typi- tribution because the external-intemal trips are distributed
cally growth factored, using an average annual growth rate. using the same procedure as intemal trips. Through trips are
Historically, the most popular method for collecting exter- needed before a traffic assignment can be performed. As will
nal travel data is to conduct a roadside intercept survey at the be noted in the next section, the procedure for estimating
regional cordon. Very few roadside surveys have been con- extemal travel is applicable only to smaller sized urban areas.
ducted in recent years,Trimfily because of the concern that
stopping vehicles on the highway would be perceived as an
unacceptable intrusion on the motorist. Poorly conducted
BASIS FOR DEVELOPMENT
roadside surveys have resulted in unnecessary delays and
extended queues of vehicles. Alternative, nonintrusive sur- In most regional or large-area studies, an external cordon
vey methods have been used to collect extemal survey data. survey is a required input to the travel modeling process. An
These include the following: external survey can provide accurate information on trip
interchanges, particularly for through trips. In addition to the
. The recording of license plate numbers (either through trip origin and destination, a number of other variables are
the use of video tape, direct reading of the plates into a needed to model external travel. The following information
tape recorder, or direct entry into a notebook computer is typically asked during a roadside survey of vehicles enter-
by a survey recorder) and matching plate numbers at the ing the study area:
cordon to obtain through trip tables; or
. The recording of license plate numbers (using one of the l. Vehicle C/ss. Vehicle class is important from several
above methods), matching the number with Department points of view. The vehicle's impact on the highway
of Motor Vehicle registration, and mailing out a survey varies by size and weight, as does its impact on capac-
form to the registered owner of the vehicle. ity and air quality. The minimum number of categories
49
StudyAre Cofdon The trip rates presented in Chapter 3 represent all trips
::::::::: Centscil Connec*q made by residents, including trips in which one end of the
thk trip is outside the study area. These internal-external trips are
o Node
part ofthe total productions for azorre. To create a trip table
o Conbc*l
Elernal SEtlon of internal-external movements, the relative attractiveness of
@
each exit route or external station is needed.
The estimation of external trips assumes that counts of the
average daily traffic (ADT) on each of the major highways
entering the study area af the cordon line are available. The
sum of the counts for all stations, representing total cordon
crossings, is greater than the total number of external trips
because through trips cross the cordon twice. If possible,
classification counts should be conducted to determine the
split between autos and trucks.
The following steps are required for developing internal-
external, external-internal, and external-external volumes:
Estimation of through trips at each station,

Distribution of through trips between stations,
a Estimation of external-internal trip productions and
attractions, and
Distribution of internal-external and external-internal
trips between internal zones and external stations.
EE = Extemal-External (Through) Tdp The procedure presented below produces reasonable re-
El = Extemal-lntemd Tdp sults for small urban areas, particularly those with popula-
]E = lntemal-'Extemal Trip tions of 50,000 or less. For interstates and principal arterials,
the rates appear to be reasonable for areas with a population
Figure 10. External travel diagram.
up to about 100,000. For areas with populations greater than
100,000, the method produces through trip percentages that
are less than zero, an illogical conclusion. The research con-
would seem to be cars, vans, and pickups as a group ducted in this project yielded very little in the treatment of
and trucks as a group. Some argument might be made external travel behavior. The characteristics of external
for dividing trucks into light, medium, and heavy, and travel are much more a function of the unique geographic
combining light trucks with automobiles, vans, and location and character of each urban area and, as such, the
pickups to yield three strata. Of course, each added opportunity for transferring external travel characteristics
stratum imposes additional base year data requirements between urban areas is limited. The procedures presented
and methodological requirements. below should be applied with extreme caution and the rea-
2. Trip Purpose. The major person-trip purposes are sonableness of the results must be thoroughly reviewed.
work, shop, and school. The work trips typically have
a longer trip length than do the shop trips. A minimum
stratification probably should include work and other.
ESTIMATION OF THROUGH TRIPS
No stratification of truck trips by trip purposes seems AT EXTERNAL STATIONS
necessary.
3. Resident Status. The resident status for persons is sim- The first step in the process will be to estimate through
ply whether they reside in the region, and for trucks, trips at the external stations. Previous research has shown
where they are garaged; t.e., f a truck is garaged in the that the percent of through trips at and between stations is
study area normally, it is considered a resident. related to the functional classification of the external high-
way, the connectivity ofeach external station pair, the average
The smaller area- and sketch-planning studies for which daily volume at the station, the relative size of the station, the
this report has been designed may not have the resources to size of the population of the study area, and the vehicle com-
conduct a survey of external travel. An alternative method position at the external station.
for estimating external travel is required and presented in this Through trips as a percent of all external trips vary from
chapter. place to place. Data for selected cities ae shown in Table 16.
50
TABLE 16 Through trips as a percent of external trips
1 990 External - lnternal External - External

Place Population lnternal - External (Through) Total
.
....1i..,
Chicago 6,070,000 95% 5o/o 100%
.:. . i
,..:,1 Twin Cities 2,464,OO0 93 7 100
San Diego 2,498,000 88 12 100
Phoenix 2,122,OOO 86 14 100
Reno 255,000 87 13 100
37,000 80 20 '100
Wausau
Through trips as apercent oftotal external trips range from For illustration, given a route with ADT of 7,000, 6 per-
5 percent in the largest region, Chicago, to 20 percent in the cent heavy and medium trucks (excluding vans and pickups),
smallest region, Wausau. and 10 percent vans and pickups, the following through trips
D.G. Modlin, Jr., working with the State of North Car- percentages shown in Table 17 would be predicted by func-
olina,r'2 developed a model for estimating through trip ends tional class using equation 5-1.
at a station on the cordon of a study area. The model used Because total through-trip percentages can vary substal-
functional classification of the highway, the ADT at the tially, it is important that the overall through trips be reason-
I external station, the percentage oftrucks (excluding vans and able and the total should be checked after application of the
pickups), the percentage ofvans and pickups, and the popu- equation. Regression models are particularly susceptible to
lation ofthe study area. error when used outside of the range of data used for the ini-
The equation for estimating the percent through trips at an tial fitting or calibration.
external station is If classification counts are not available at the cordon, the
_:l percentage of trucks at the extemal stations must be esti-
Y ="16.76+11.22x I -25.74x PA mated. InNCHRP Report 187, total areawide trucktrips were
- 042.18x MA+ 0.00012 x ADT, + 0.59 (5-1) presented as a percent of areawide vehicle trips. At the time
x PTKS,- 0.48 x PP8 - 0.000417 x POP that report was released, truck traffic represented anywhere
from2l percent of total trips in areas with less than 100,000
where population to I 6 percent of total trips in the largest urb anized
i
:- areas. Recent studies suggest that trucks are a smaller portion
:
l
Y percentage of the ADT at external station i, that of the total vehicles on the road now, because of the increase
':... .1 .
I :- are through trips, in personal nonworktrips. A truckpercentage between 5 per-
1: interstate (0 or 1), cent and 15 percent of the total trips might be more realistic.
P : principal arterial (0 or 1), Once the percent of through trips crossing the cordon is
MA : minor arterial (0 or 1), estimated, the number of through trips can be calculated by
ADT1 : average daily traffic at external station l, station.
PTKS: percentage of trucks excluding vans and pickups Using the example problem from Table 17, assume that an
at external station i,
area with a population of 25,000 had a minor arterial with
PPS : percentage of vans and pickups at external sta- counts of 3,600 inbound and 3,400 outbound for a total of
tion i, and 7,000 ADT. The total through trips at the station would equal
POP : population inside the cordon area. 24 percent of 7,000 or 1,680 crossings. This would be split
into 864 through trips entering the area and 816 through trips
In equation 5-1, an external station can be only one ofthe leaving the area. The remaining 5,320 crossings have a trip
three functional classifications. For that classification, the end in the study area.
value of the vaiable is 1; for the other two, the value will be
0 (i.e., functional class is a dummy variable).
DISTRIBUTION OF THROUGH TRIPS
BETWEEN STATIONS
lDavid G. Modlin, Jr. Synthesis of Through Trip Patterns in Small Urban Areas,
Department of Civil Engineering, North Carolina Stare University, Raleigh ( 197 l). The distribution of the estimated through-trip ends from an
'?David G. Modlin, Jr., "Synthesized Through-Trip Table for Small Urban Areas," external station to each of the other external stations is the
Transportaton Research Record 842, Transportation Research Board, National
Research Council, Washington, DC (1982). next step in obtaining a matrix of through trips among sta-
51
TABLE 17 Alternative through-trippercentages
Population
Functional Class 25,000 50,000 1 00,000
lnterstate 77% 67% 46%

Principal Arterial 40 30 I
Minor Arterial 24 13 o2
t Example problem assumes the following:
ADT = 7,000
Heavy and Medium Trucks = 6%
Vans and Pickup Trucks = 107"
2
Computed value less than O%, therefore use O%
tions. If an alea had l0 external stations, then the resulting RTECON'j: route continuity between stations i j:
and
vehicle trip table would be a matrix with 10 origins and 10 1 : Yes,0 : No, and
destinations.
Modlin developed equations, one for each functional ADT : average daily traffic at the destination sta-
class, to estimate the distribution of through trips that enter tion j.
the analysis area at an origin external station (l) to each of the
destination stations (). For estimation of each interchange, Station-to-station trip movements also can be estimated
the functional class of the destination station dictates which using a simple factoring procedure which uses an external
equation is to be used. station's portion of the total through trips. However, because
the geographic characteristics of the study aea often deter-
Interstate: mine the likely connections between stations, some effort
should be made to ascertain the existing through movement
Y = -2.70 + 0.21x PTTDES j patterns either by reference to ealier studies of the area or by
(s-2) general observations. The likely movements can be set using
+67.86x RTECON,,
control totals.
Principal Arterial:
Y = -7.40 + 0.55 x PTTDES j
Example of Through-Trip Table Estimation
+24.68x RTECON,, + 45.62* ,!O'' (s-3)
\enr,
j=l
To illustrate the application of through-trip procedures, a
simple five-station external example is presented. Assume
that the data in Table l8 have been observed at the external
Minor Arterial: stations.
In this example, stations 101 and 103 are two points on a
Y = -0.63+ 86.68 * ,ro"' continuous route, and stations lO2 and 104 are two points on
another continuous route.
\ tnr, (s-4)
The estimated through trips for each station are computed
+ 30.04 x RTECON,
using the equation:
where Y = 76.76 +ll.22x I -25.74x PA - 42.18x MA

: + 0.00012 x ADT, + 0.59 x PTKS. - 0.48 x PP.i
1 percentage distribution ofthrough-trip ends
from origin station I to destination stationj, - 0.000417 x POP
PTTDESj: percentage through-trip ends at destination For example, the percent through trips for station 101
station j, would be:
52
'i
TABLE 18 Example data for through-trip estimation
Functional Percent Percent Vans and

Station Classification ADT Trucks Pickups
101 Principal Arterial 15,000 5 10

102 lnterstate 25,000 10 10
103 Prncipal Arterial 10,000 7 10
104 lnterstate 20,000 10 10
10s Minor Arterial 5,000 3 10
Total 7s,000
previously are used and the results are normalized in order

Ytot = 76.76 + ll.22x0 -25.74 x 1- 42.18 x 0
+ 0.00012 x 15000 + 0.59 x 5 - 0.48 x 10 for the sum of the distribution percentages to be equal to 100
- 0.000417 x 50000 = 30 percent. For example, the distribution of trips from station
101 to the other four stations is presented in Table 20.
The resulting through trips are presented in Table 19. The The through-trip distributions are computed for each of the
trips have been rounded to the nearest 100 trips. four remaining external stations. Table?I contains the normal-
The next step is to estimate the distribution of the through ized percentages of through trip distribuons among the five sta-
_',] I
trips between the external stations. The equations presented tions. The percentages sum to 100 percent down each column.
"sl TABLE 19 Through trips
--l
Percent Through E-land l-E
Station ADT Through Trips Trips
101 15,000 30 4,500 10,500

102 25,000 71 17,800 7,200
103 10,000 31 3,100 6,900
'104 20,000 71 14,100 5,900
105 5,000 11 600 4,400
Total 75,000 40,100 34,900
TABLE 20 Distribution of through trips for external station 101
Origin Destination Calculated Normalized

Slalion Station Percent Percent
102 12% 18o/o
103 40 58
101 104 12 17
10s 5 7
Total 100
53
TABLE2l Through-trip distributionpercentages
Origin Station
Destination
Station 101 102 103 104 105
101 15 59 16 31
102 18 17 67 21
103 58 13 13 27
104 17 67 17 21
105 7 4 7 4
Total 100 100 100 100
The percentages presented in Table 21 are applied to the tables that are growth factored. Table 25 contains the bal-
through trips presented in Table l9 for each external station. anced or "Fratared" external through-trip table.
'lable22 contains the initial through-trip table. The resulting through-trip table is saved for later use in
Note that the row totals of trips do not equal the desired traffic assignment. The station-to-station vehicles are added
number of trips for each external station and that the table to the total vehicle trips and assigned using the standard high-
is not symmetrical about the intrastation diagonal. For way assignment procedures. Although the through trips are a
example, the trips from 101 to 102 equal2,736 trips while minor portion of total vehicle trips in a region, the external-
the trips from 102 to 101 equal 790. Because the trips rep- external volumes have a significant impact on facilities
resent average daily trips, the table should be symmetrical. crossing the cordon line and passing entirely through the
The trip table is averaged to produce a table symmetrical study area.
about the diagonal. This symmetrical trip table is presented
in Table 23.
ESTIMATION OF EXTERNAL.INTERNAL TRIP
At this step in the process, the row and column totals are PRODUCTIONS AND ATTRACTIONS
equal; however, they are not equal to the desired number of
through trips at each external station. This difference is pre- The estimation of external-internal trip productions and
sented in Table 24. attractions is needed as part of the trip generation process. In
The most common procedure for adjusting a trip table to Chapter 3, the section on balancing productions and attrac-
match desired row and column totals is the matrix balancing tions specified the need for external travel information in
or Fratar technique. Many of the travel demand software developing regional control totals by purpose. In fact, the
packages have programs for applying this technique. The approach for developing external productions and attractions
major use of the technique is to produce future-year trip is determined by whether or not the external trips made by
T^BLE 22 Initial through-trip table
Origin Station
Destination
Station 101 102 103 104 105 Total
101 ,,r: 1,837 2,165 188 6,926

102 790 524 9,483 126 10,924
103 2,595 2,329 1,843 160 6,927
104 782 11,965 519 't2s 13,391
105 332 770 220 609 1,932
Total 4,500 17,800 3,100 14,100 600 40,100

54
TABLE 23 Averaged through-trip table
Origin Station
Destination
Stalion 101 102 103 104 105 Total
101 1,763 2,216 1,474 260 5,713

102 1,763 '1,426 10,724 448 14,362
103 2,2'16 'l,426 1,181 190 5,013
104 1,474 10,724 1,181 17,746
105 260 448 190 367 "y 'l
,266
Total 5,713 14,362 5,013 13,746 1,266 40,1 00
residents of the region are included in the trip generation truck trips could be treated separately in determining the res-
rates by trip purpose. If external trips are not included with idency status based on the garage location. However, when
the home-based work, home-based other, or non-home-based basic information on external travel is not available, it will
trips, the external trips could be treated as a separate purpose. be necessary to apply typical factors by station. These factors
The approach outlined in this report assumes that external are applied to the two-way ADT by station.
trips are included in the trip rates by purpose. While the external travel characteristics of cities and met-
The first part of this step will be to summarize the through- ropolitan areas can vary significantly, a few common vari-
trip matrix by direction and station and subtract these totals ables exist. Earlier it was stated that the size of the study area
from the station counts. The remainders represent the over- affects the percentage ofthrough trips. The size of a region
all control totals by station for external-internal trips. V/hile (in area), its socioeconomic characteristics, and proximity to
the counts conducted at the external stations might show dif- other urbanized and suburban areas are other factors that
ferences in the number of vehicles traveling into and out of affect the purpose and residency status of external-internal
the study area on a particular route, the directional differ- trips. The existence of a strong employment center within
ences are ignored. This assumption, that the total trips enter- the study area will tend to pull more nonresidents into the
ing the study area equals the total trips leaving the study region to seek employment. In areas where trip attractions
area on a typical day, simplifies the process used to estimate such as employment and shopping are distributed more
external-internal travel. evenly between the areas inside and outside the cordon, the
The next step involves separating the external trips by pur- split between resident and nonresident trips at the cordon
pose and resident status. The resident totals by purpose becomes relatively equal. Alternatively, a region that is
become the attractions at each station. Nonresident totals by mostly suburban may have a shortage of overall employ-
purpose become the productions at each station. ment opportunities and a surplus of service and retail
If an external survey has been conducted, information on employment. In such a community, the flow of trips across
the purpose and residency status of trips could be used the cordon could reflect a net export of work trips and a net
directly to estimate productions and attractions. External import of other trips.
TABLE 24 Difference between calculated and desired external station

through trips
External Calculaled Desired Ratio Desired/

Station Trips Trips Calculated
101 5,713 4,500 0.79

102 14,U2 17,800 1.24
103 5,013 3,'100 .63
104 13,746 14,100 1.03
105 'l 600
,266 .47
Total 40,100 40,100

55
TABLE 25 Balanced through-trip table
Origin Station
Destination
Station 't01 102 103 f04 105 Total
101 2,781 974 662 83 4,500

102 2,781 t'u1 12,952 383 17,800
103 974 1,684 397 45 3,100
104 662 12,952 397 89 14,100
105 83 383 ,15 89 600
Total 4,500 17,800 3,100 14,100 40,100
Tables 26 and 27 show the split of internal-external trip based-l5 to 25 percent. These default values for external
totals by purpose and resident status for an area with a cen- travel should be used cautiously, however, and external sur-
tralized activity center-San Juan, Puerto Rico-and aarea veys are highly recommended.
with a more even distribution of activities on either side of Returning to the example in Table 19 in which through
the cordon-San Diego, California. These are presented to trips were estimated, the external-internal trip productions
show the variation possible when planning regions have and attractions also can be estimated. It was noted that 34,900
strong central activity centers or more diffuse employment. of the crossings represented external-internal or internal-
These examples also demonstrate another characteristic of external trips. The 10,500 external-internal trips for station
traffic across cordons: the number of work trips across 101 could be split as follows: 3,675 home-based work
the cordons is greater than the regional share of work trips. (35Vo);4,200 home-based other (407o); and2,625 non-home
Two obvious reasons can be used to explain this phenome- based (257o).
non: (1) auto occupancy for work trips is generally lower The next step in the process is to translate the vehicle
than for all other purposes, so the share of vehicle trips that trips into productions and attractions. External station pro-
are carrying work trips is higher than the share of person ductions are trips whose home base is outside of the region,
trips that are work trips; and(Z) average trip lengths for work and external station attractions are trips whose home is
trips are generally longer than average trip lengths for other within the region. The task of splitting the vehicle trips is
trip purposes so work trips tend to be more likely to pass therefore dependent on a basic level of knowledge of the
between regions than trips made for other purposes. If no general land use and travel patterns in and around the study
data are available to estimate the distribution of inter- area. Suppose that the study area in the preceding example
regional vehicle trips, local knowledge should be used to is primarily a suburban area 20 miles away from a major
estimate the values from the following ranges: home-based urban area. Suppose also that the primary attractions in
work-25 to 50 percent of total vehicle trips across the the study area ate a major university and several regional
cordon; home-based other-30 to 50 percent; non-home- shopping centers. If local knowledge tells us that the
TABLE 26 External trip purpose/residency factors for centralized areasr
Trip Purpose Resident Non-Resident Total
Home-Based Work 12"/" 34o/o 46"/"

Home-Based other I 23 32
Non-Home-Based 11 11 22
Total 32 68
I San Juan, Puerto Rico 1990 External Cordon Survey.

56
TABLE 27 External trip purpose/residency factors for evenly distributed areasr
Trip Purpose Resident Non-Resident Total
Home-Based Work 15% 1O"/" 257o

Home-Based other 27 23 50
Non-Home-Based I 17 25
Total 50 50
ISan Diego Region.
predominant flow of traffic during the a.m. peak period tion and purpose, they are used in the standard modeling
flows from out of the study area at the ratio of 3 to 1, we process to reflect trips between internal zones and external
could make the following assumptions: 75 percent of exter- stations. In trip generation, control totals oftrip productions
nal-internal home-based work trips have productions within can be calculated using the external totals to balance pro-
the study area: 4O percent of the home-based other trips ductions and attractions. While attractions should be nor-
have productions within the study area; and (by definition) malized to productions for internal trips, the external sta-
50 percent of the non-home based trips have productions tion attractions should be held constant, because they
within the study area. The following table would summa- represent actual counts of the base year. This was shown in
i
rize the 10,500 external-internal trips associated with sta- Equation 3-1 in Chapter 3. Trip distribution of internal-
lffis
tion 101: external and external-internal trips follows the conven-
i-
Trip Purpose Productions Attractions Total
Home-Based Work 919 2,7s6 3,675

Home-Based Other 2,520 1,680 4,200
Non-Home Based 1,312 1,313 2,625
Total 4,751 5,749 10,500
The trip attractions would equal 5,749 vehicles and the trip tional gravity model approach described in Chapter 4, Trip
productions would equal 4,751. vehicles at external station Distribution.
number 101. These vehicle trips need to be converted to per-
son trips using the automobile occupancy rates presented in CASE STUDY
Chapter 7 before they can be included in the final steps ofthe
trip generation process. As has been discussed previously, external trips can be
divided into two categories: external-external trips, which
pass completely through the region without having a trip-end
DISTRIBUTION OF INTERNAL-EXTERNAL within the region, and external-internal trips, which have one
AND EXTERNAL-INTERNAL TRIPS trip-end within the region and one trip-end outside of the
region. The external-internal trips are converted to person
After the external productions (nonresident trips) and trip-ends and incorporated into the regional trip generation
attractions (resident trips) have been estimated for each sta- model, while the external-external trips are expressed as a
57
separate trip table that is added to the other vehicle-trip tables aeas with less than 50,000 in population. Therefore, local
before assignment. experience must be relied upon to estimate the through-trip-
The procedures used to estimate external travel for the making potential for the Asheville region. This experience
Asheville case study are listed below. All of the calcula- was used to classify four facilities, those carrying ADT vol-
tions were performed with the aid of a computer spread- umes of greater than 20,000, as interstates, each of which
sheet program. was estimated to contribute 30 percent of its traffic to the
external-external trip table. Another two facilities, desig-
Classification of External Stations nated as principal afierials, were estimated to have a 10 per-
cent through-trip share each. The remainder of the external
Average daily traffic (ADT) counts were collected for l6 stations were designated as minor arterials and were assumed
facilities crossing the external cordon around the Asheville to contribute a negligible share of their ADT volumes to the
region. These ADT counts were collected at each location through-trip table.
where significant traffic volumes flow into or out of the Table 28 displays the external station volumes including
Asheville region. Each of these external stations was clas- the estimated number of through trips and internal-external
sified as either a minor arterial, a principal arterial, or an trips. AII of the data in this table reflect vehicle trips, because
interstate facility. In addition, external station pairs that the data are based upon existing traffic count data.
were linked by a continuous facility were noted because
they would be expected to carry a statistically significant
Distribution of Through Trips
share of external-external traffic. The most notable of these
to External-External Trip Table
pairs in the Asheville region are stations 109 and 117,
which are connected by the Route l9l23 bypass and Inter- The distribution of through trips between stations is esti-
state 26, and stations I 14 and 1 2 I , which are connected by mated using Equations 5-2,5-3, and 5-4. The relative shaes
Interstate 40. were first calculated as in the following example for the inter-
change between external stations 109 and 117, which repre-
Estimation of Through-Trip Percentages sent the eastern and western extremities of I-40 within the
study area:
The synthetic procedures outlined previously in this chap-
ter for estimating the share of external cordon trips that are Yu = -2.10 + 0.21x PTTDES j + 67.86x RTECONI
likely to be through trips are only appropriate for urbanized = -0.70 + 0.21 x 30 + 67.86 xl = 71.46
TABLE 28 External station through-trip summary
Station 1 989 Percent External- lnternal-

Number Description ADT Classification Through External External
108 Rout 251 1,800 Minor 0 0 1,800

109 Routes '19 & 23 Bypass 27 ,7o0 lnlerstate 30 8,310 19,390
110 Roules 19 & 23 Business 7,000 Minor 0 0 7,000
111 BRP (N) 2,8s0 Minor 0 0 2,850
112 Snope Creek Road 2,000 Minor 0 0 2,000
113 Route 70 16,100 Principal 10 1 ,61 0 14,490
114 l-40 (E) 24,7OO lnterstate 30 7,410 17,290
115 Route 74 1'1,000 Minor 0 0 11,000
116 Route 25 12,450 Minor 0 o 12,450
117 l-26 33,100 lntrstate 30 9,930 23,170
118 280
Rottes 191 & 7,4oo Minor 0 o 7,400
119 BRP (S) e7o Minor 0 0 970
120 Route 151 f ,550 Minor 0 0 1,550
121 f-40 (w) 27,500 lnterstate 30 8,250 19,250
122 Leicester Highway 14,000 Principal 10 1,400 12,600
123 Bear Creek Road 3,940 Minor 0 0 3,940
"l
...1 58
.i. 'i
,'...1
where external stations expected to contribute a significant number
--lr'.
'l
1, : percentage distribution of through-trip ends
ofthrough trips to the external-external trip table. For station
109, the adjusted shares were used to distribute the 8,310
from origin station i to destination stationj,
j: 109,
through trips originating at that station to the other five sta-
j:tr1, tions. The same procedure was used to distribute the through
PTTDESj: percentage through trip ends at destination trips associated with the other five interstate and principal
station j, and external stations. The results ofthis process are displayed in
RTECONj: I : route continuity flag for stations i and j. Table 30. Note that, for intuitive reasons, there are no intra-
zonal trips within the external zones, and that there are no
',.',.1
j:
The calculations for the other external station pairs ae dis- trips allowed between stations 113 and 114, which are prox-
played in Appendix Table B-4. imate, parallel facilities unlikely to attract trips from one
'':
-:.1
The relative shares for each of the possible destinations another.
from a cordon station are added together and the result is Given that the values arrived at in Table 26 are not sym-
used to normalize the raw data. Table 29 displays the raw metrical (i.e., the number of trips from station I to j is not
shares and the normalized shares for each of the potential equal to the number of trips from j to l) the next step is to
destinations for through trips entering the region at each of average the U and jl values to produce a symmetrical trip
the six external stations. table. For example, given that the estimated value from sta-
Next, the normalized shares were used to distribute the tion 109 to station Ill is7,031, and the value from station
through trips originating at those stations to the other tve 117 to 109 is 8,402, the average value between stations 109
TABLE 29 Through-trip distribution-raw and normalized percentages
Origin Station
Destination
Staton 109 113 '121 '122
Raw Percentages
109 3.60 3.60 71.46 3.60 3.60

113 3.23 3.23 3.23 3.23
114 3.60 3.60 71.46 3.60
117 71.46 *o *o 3.60 3.60
12'l 3.60 3.60 7',1 .46 3.60 3.60
122 2.56 2.56 2.56 2.56 2.56
Total 84.46 13.36 81.22 84.46 84.46 17.63
Norm. Factor 1.184 1.184 1_184 5.671
Normalized Percentages
109 26.94 4.43 84.61 4.26 20.42

113 trt 3.83 3.83 18.33
114 4.26 4.26 84.6'l 20.42
117 84.61 za,s *t 4.26 20.42
121 4.26 26.94 87.98 4.26 20.42
122 3.03 19.1 I 3.16 3.03 3.03
Total 100 100 100 100

59
TABLE 30 Through-trip table-asymmetrical
Origin Station
Destination
Station 109 1f3 114 117 121 122 Total
109
113
114
318
354
434
'! 8,402
380
423
3s2
3f6
6,981
286
257
286
9,802
1,271
8,O44
117 7,O31 434 328 352 286 8,431
121 354 434 6,s19 423 8,016
't22 252 309 234 301 250 'T 1,347
Total 8,310 1,610 7,410 9,930 8,2s0 1,400 36,910
and ll'7 isl ,111 . The results of this exercise are displayed in tion totals as shown in Table 28. Next the external trip pur-
Table 31, the symmetrical trip table. pose factors were applied to the external-internal totals.
The result of this latest maneuver, however, is a trip table Local knowledge of the region is used to, estimate that the
in which the row totals and column totals are not equal to the traffic crossing the external cordon is composed of 40 per-
through volumes estimated in Table 28. The recommended cent home-based work trips, 40 percent home-based other
solution to this problem is to apply the Fratar technique to the trips, and 20 percent non-home-based trips. Local experience
symmetric trip table, using the through trip volumes in Table is then used to further estimate that the Asheville area is a net
28 as the row and column targets. The ultimate result of the importer of work trips, by a ratio of 70 to 30, and that the
Fratar process is the final external-external vehicle trip table, region is a net importer of other home-based trips by a ratio of
as displayed in Table 32. 60 to 40. The non-home-based trips are assumed to be
balanced between productions and attractions. Finally, auto-
occupancy factors (from Chapter 7) of 1. 1 1 persons per vehi-
cle for home-based work trips, 1.67 persons per vehicle for
Conversion of lnternal-External Trips
to Person-Trip Productions and Attractions home-based other trips, and 1.66 persons per vehicle for non-
home-based trips were used to convert the vehicle trips to per-
In order to estimate the internal-external vehicle trip totals, son trips. The resulting estimates of trip productions and
the through-trip totals were subtracted from the external sta- attractions for external stations in the Asheville region are
TABLE3l Through-triptable-symmetrical
Origin Stalion
Destination
Station 't09 113 114 117 121 122 Total
109 7,717 353 269 9,056

113 376 "2 '1 407 375 283 't,440
114 341 376 6,750 260 7,727
117 7,717 407 376 387 294 9,180
121 353 375 6,750 *t 8,133
122 269 283 260 294 268 "y 1,373
Total 9,0s6 1,440 7,727 9,180 8,133 1,373 36,910
Target 8,310 1,610 7,410 9,930 8,250 1,400 36,910
Adj. Factor 0.918 1.118 0.9s9 1.082 1.014 1.020

60
TABLE 32 Through-trip table-after the Fratar adjustment
Origin Station
Destination
Station 109 113 114 117 121 122 Total
109 222 167 7,526 243 '152 8,310

113 222 676 439 273 1 ,6'10
l
't14 167 515 6,521 207 7,4',1O

117 7,526 6; 5'15 746 467 9,930
'121 243 139 6,521 746 301 8,250
'122 152 273 207 467 301 1,400
Total 8,3f 0 1 ,610 7,410 9,930 8,250 1,400 36,910
summaized in Table 33. This table shows that the estimated SUMMARY
157,150 extemal-internal vehicle trips crossing the cordon
*"-j
l
around the Asheville region carried 226,925 person trips, This is a review of steps required to estimate external
'-'- -. -
including 137,915 productions (from locations outside the travel. The key is knowing the ADT by direction for trucks
_1
region) and 89,010 attractions (to locations outside the region). and autos at each external station.
TABLE 33 External-internal person-trip productions and attractions
Station Number
HBW HBO NHB HBW HBO NHB
108 559 721 298 239 480 298

109 6,026 7,771 3,218 2,582 5,181 3,218
110 2,175 2,805 1,162 932 1,870 1,162
111 885 1,142 473 379 761 473
1'12 62',1 801 332 266 534 332
113 4,503 5,807 2,405 1,930 3,871 2,405
114 5,373 6,929 2,870 2,303 4,619 2,870
115 3,418 4,408 1,826 1,465 2,939 1,826
't16 3,869 4,989 2,066 1,658 3,326 2,066
117 7,201 9,286 3,846 3,086 6,191 3,846
118 2,299 2,965 1,228 985 1,977 1,228
119 301 388 161 129 259 161
120 481 621 2s7 206 414 257
121 5,982 7,715 3,195 2,5 5,14| 3,tres
122 3,916 5,050 2,091 1,678 3,366 2,091
123 1,224 1,579 654 524 1,0s2 654
Tolal Person Trips 48,842 62,985 26,086 20,932 41,990 26,086

61
1. Collect classif,cation counts at each cordon station. correspond to attractions at each external station. Non-
2. Estimate percentage of through trips (E-E) at each cor- resident totals by purpose correspond to productions at
don station. each external station.
3. Take through trips and distribute to create the E-E trip 6. Convert external-internal vehicle-trip productions
table. This vehicle trip table will be used in the traffic and attractions to person trips (Chapter 7) and com-
assignment step. plete balancing of Ps and As by purpose as shown in
4. Subtract through trips from total ADT at each station to Chapter 3.
ger E-I/I-Erotals. 7. Distribute the E-I and I-E trips using the gravity model
5. Apply the trip purpose and residence (direction) factors by trip purpose (i.e., HBW, HBO, and NHB) as shown
to two-way ADT by station. Resident totals by purpose in Chapter 4.
'I
62
CHAPTER 6
MODE-CHOICE ANALYSIS
INTRODUCTION description of all modes of available or proposed highway,

HOV, and transit and are extremely data-intensive. The
Mode-choice analysis is the third step in the traditional incremental logit or pivot-point formulation allows for
four-step travel-demand forecasting process. It is the most analysis of transit improvement strategies or policies without
complex of the modeling steps and in the last decade most of the complete simulation of the entire transit system and its
the research and advancement in travel-demand models has altematives. A limitation of this formulation is that it cannot
related to this step. The focus on mode-choice models has be used to estimate transit use in an area that does have exist-
been generated by the analysis of major new proposed and ing transit service and patronage. This incremental logit
constructed fixed-guideway systems throughout the United structure is, however, the most transferable of the three
States. Mode-choice modeling is also used to evaluate among different urban areas and will therefore receive the
improvements in bus systems and for analysis of HOV strate- most discussion.
gies. In mode-choice analysis, the total zone-to-zone person The use of incremental logit often is used for the eval-
trips resulting from the trip-distribution model are split into uation of Travel Demand Management (TDM) strategies
trips using each available mode between each zone pair. By directed at reducing vehicle travel during peak periods. The
incorporating various levels of auto occupancy into the mode- application of the incremental logit model to evaluate exam-
choice model, the vehicle-trip tables are produced directly by ple TDM strategies is presented in this chapter.
the model and the need for further auto-occupancy factors is
eliminated. Similarly, transit trips by submode (e.g., local
bus, express bus, and rail) and access mode (e.g., walk to BASIS FOB DEVELOPMENT
transit or drive to transit) are produced and ready for the last
A brief discussion of the multinomial and incremental
step-assignment.
formulations of the logit model is presented below. The fol-
Most mode-choice models are based on the logit formula-
lowing discussion is presented as background for the devel-
tion. The following mode-choice model formulations will be
opment of the incremental logit model; full application of
discussed:
either a multinomial or nested logit model requires, as a
minimum, a calibration and validation of the mode-specific
. Simple multinomial logit, model constants to reflect localized transit and mode-choice
. Incremental logit (pivot point), and characteristics.
. Nested logit.
The third form of the logit model, the nested logit model, LOGIT FORMULATION
is gaining use by larger urban areas where there are compet-
ing modes of public transpofiation (e.g., local bus, express The generalized logit model formulation is a mathemati-
cal relationship that estimates the probability of choosing a
bus, rail, and HOV) and multiple access modes. For the
specific mode by using the following equation
small- to medium-sized urban areas that are primarily evalu-
ating local bus service as a competition for the auto mode, the
multinomial logit model is usually adequate. However, if tD-e"'
- --I-- (6- 1)
mode of access and/or auto occupancy (HOV) is desired in 2,"'

the mode-choice model, the use of a nested model should be
considered.
where
The multinomial logit and nested logit formulations are
used to estimate nlode shares for most transit strategies, P : the probability of a traveler choosing mode i,
including the introduction of a new transit mode (e.g., rail) u : alinear function of the attributes of mode i that
or for introduction of transit service into an area that cur- describe its attractiveness, also known as the
rently has no service. These formulations require a complete utility of mode i, and
63
Figure I I presents the multinomial logit model with three

ILI ,"' : the summation of the linear functions of the modes: drive-alone, shared-ride, and transit. This structure
,=r attributes of all the alternatives, k, for which a allows for the analysis of network and policy alternatives that
choice is available. are designed to increase the shared-ride, or HOV, mode.
These alternatives can include dedicated HOV lanes and
The linear function of the attributes, or utility function a, is preferential parking for HOV that result in a quantifiable dif-
composed of ference in the level of service for each mode. In many urban
areas, there are no significant level-of-service differences for
drive-alone and shaed-ride modes (other than the sharing of
Lt = d +b,x IVTT, +c,xOVTT, +d xCOST. 6-2) operating costs) that can by measured by a model. In these
cases, a simple binary or two-mode model is appropriate. The
where modes would be transit and auto. The resulting auto person
trips are converted to vehicle trips using auto occupancy
MTi : the in-vehicle travel times for mode i,
ratios presented in Chapter 7. The level-of-service coeffi-
OVTT1: set of variables measuring the out-of-vehicle
cients, in-vehicle travel time, out-of-vehicle travel time, and
travel times for mode i-walk, wait, and trans-
costs, that are presented in ihis chapier ae the same for the
fer times-may all be kept separate or com-
total auto mode as they are for the drive-alone and shared-ride.
bined, depending on the calibrated structure of
the model,
COSTi : the cost of mode l,
a :
INCREMENTAL LOGIT FORMULATION
mode-specific coefficient (constant) to account
for mode bias not measurable with the level-of-
The multinomial mode-choice model is estimated on the
service vaiables,
basis of the complete characteristics of the transit system and
: coefficient for the IVTT variables of mode l,
the potential users. An alternative method of applying these
c : Set of coefficients for OVTT v ariables of mode
models is in an incremental formulation that begins with
i, and
: existing mode shares and modifies these baseline values
d, coefficient for CO,9Z variable of mode i.
based on changes in the characteristics of the transit or auto
networks.r The following are the principal advantages of the
The level-of-service variables may be aggregated to total incremental formulation:
MT, OVTT, and COST, or they may be kept separate with
specific coefficients. For example, out-of-vehicle variables . Uses observed, measured mode shares
such as walk time, first wait, and second wait times are often . Requires a description only of the changes to the transit
kept separate with different estimated coefficients. In most service
simple applications, a single total is computed each for IVTT, . Highly transferable among urban areas because mode-
OVTT, andCOST. specific and socioeconomic bias is accounted for in the
Examples of the structures of the simple multinomial and observed, measured shares of travel by each mode.
nested logit models are presented in Figure 1L
The incremental form is a derivation of the standard multi-
Simple Multinonial Logit
nomial logit formulation presented earlier. The formulation is
D,- --f-Exrou'
r (6-3)
-
) {4 * "o'n)
Nested Logit
where
P : the baseline probability (share) of using mode i,

P : the revised probability of using mode i, and
Au: the change in utility for mode l.
The change in utility expression can be derived from the util-

ity formula presented earlier as
Barton-Aschman Associates, Inc., and Parsons BrinckerhoffQuade & Douglas, Inc.,

Task 3.03 SeNice and Pqh'onage Forecasring Metlndology-Honolulu Rapid Transit
Figure I l. Mode-choice model structures. Program, Honolulu, Hawaii (March 1992).
64
Lu = (a, + bi x NTIi + ci x OVTIT' + d, x COST,) Orleans and Seattle models use coefficients for drive-to-
- (a, + b, x IVTT, + c, transit times approximately 7 times as great as the "normal"
x OVTT, + d, xCOST,) in-vehicle times, the Dallas model uses a drive-to-transit
coefficient that is the same as used for out-of-vehicle time.3
Lu, = 6 yn-'- Mlt) + c(OVTT - OvTTt') (6-4) This variation in coefficients is related to network coding
+dt(cosri- cosrt) conventions and market segmentation.
A default coefficient for in-vehicle travel time of -0.025
Lu - b x NVTT, + ci x LOVTfi is reommended for use in the incremental logit model.
+ d, x ,COST, The out-of-vehicle time coefficients, some of which are
stratified by walk, initial wait, and transfer time, are larger
where than the in-vehicle time coefficients by a factor of I.5 to 2.3.
: The coefficients range from -0.114 to -0.028, with the
IVTT' i, OVTT' i, COST' i level-of-service variables
majority being between -0.077 and -0.030. A default coef-
after proposed change for
ficient of out-of-vehicle time equal to 2 times the in-vehicle
mode i, and
coefficient, -0.050, is recommended. The IVTT and OVTT
LNTfi, LOVTT, ACOST:: the change in level-of-ser-
default coefficients are summarized in Table 36.
vice variables for mode i.
The coefficient for cost is a function of the average income
of the urban area and the relative value the trip maker places
The mode-specific constants, , fall out of the computa-
on time and cost. As a general rule, for higher income trip
tions. The only terms entering the equation are those that makers, cost has less importance and minimizing time has
change. The lack of mode-specific constants is a particularly
more importance in the decision-making process. Table 35
attractive feature of the incremental logit structure. These presents the values of time in dollars per hour and as a per-
constants capture the effects of such unmeasurable attributes centage of income for a sample of urban area models. Most
I as transit reliability, image, and other characteristics that can
of the cost coefficients range between 20 and 30 percent of
vary substantially from one urban area to another. In fact, the regional average income. Table 36 presents a set of default
level-of-service coefficients of a mode-choice model are cost coefficients. Users should enter the table with the aver-
often transferred from one urban area to another, and the age income for their region and the desired value of time as
model is calibrated to fit observed transit shares by adjusting a percentage of income. If this percentage cannot be deter-
the mode-specific constants or coefficients. Because they mined, then an average value of 25 percent may be used. A
I
drop out of the incremental form, differences in these unmea- plot of the cost coefficients as a function of income and value
surable attributes also drop out, to produce a model that is of time is presented in Figure 12.
more likely to be transferable between urban areas.2 The cost coefficients in Table 36 can be derived using the
formula
MODEL COEFFICENTS .
*i
1.248x b,
(6-s)
d--
- IncomexTYP
Table 34 summarizes a review of several mode-choice
models used around the country and forms the basis for where
establishing suggested default coefficient values to be used d : coefficient for COSI,
in the application of the incremental model discussed in the : coefficient for IVTT (-0.025 used for Table
Basis of Development section. Most of the models presented 36),
are of the multinomial logit formulation, but there is some Income: average regional household income,
experience with the nested logit structure. Table 35 summa- TVP : value of time percentage (expressed as decimal),
rizes the values of time coefficients for home-based work and
mode-choice models. 1,248 : factor to convert income in $/year to @/minute.
A review of Table 34 reveals that although there are dif-
ferences in the coefficients, the major independent variables Table 36, Part B, shows the results of this derivation for var-
of in-vehicle time, out-of-vehicle time, and cost have coeffi- ious income levels.
cients within a similar range. The in-vehicle time coefficients
range from -0.015 to -0.040, with most being in the
INCREMENTAL MODEL APPLICATION
-0.017 to -0.028 range. The New Orleans, Seattle, and Dal-
las models have separate coefficients for normal in-vehicle
The application of the incremental mode-choice model
time and drive-to-transit in-vehicle time. Although the New
can be done in any spreadsheet program or by using simple
?Parsons Brinckerhoff & Douglas, Review of Best Practces, Prepared for the
Quade
Metropolitan'Washington council of Govemments (December 1992).
65
TABLE 34 Review of home-based work mode-choice coeflicients
Transit Out-of- Auto Translt lnitial Transit Auto

ln-Vhcl Driv-Accsss Vehicle ferminal Walk Transt Transfsr Total oporatng Transit Parking
Cty Time Tlme m Time Time Wait me 'llme Cost Cost Fare Cost
Coefflclenls on Se|lce-Levsl Varlablos lrom a Samplo of l{om-Based-Work Modeholco Modls
Nw Orlsans -.015 -.033 -.o77 -.032 -.008
Minn/St. Paul -.031 -.o44 -.030 -.044 -.014
Chlcago -.o28 -.030 -.1 14 -.o23 -.114 --0121
Los Angeles -.o20 -.112 -.o144

Ssata -.040 -.o41 -.030 -.o44 -.O'14
Cincinnal -.0t9 -.o28 -.004s
Washngton -.o17 -.058 -.004 -.004 -.009
San Francisco -.o25 -.o58 -.0039
Dallas -.030 -.0s5 -.059 -.005 -.005 -.o12
Shirlsy (low) -.o22 -.05s -.035 -.0037
Shirlsy (high) -.034 -.044 -.0046
Salt Lako Cty -.019 -.037 -.0059
Porand -.034 -.o72 -.o1384
Sourcas: Parsons Brincksrhoff Ouade & Douglas, lnc., Revew ol BeEt Practces, Washnglon, D.C. (1992).
KPMG Psat Marwlck, Compendlum o Tnvel Dmand Focasting Methodologas, Wepard lor Fdoral Transit Administraton, Washington, D.C. (February 1992)
TABLE 35 Review of mode-choice coefficients-values of time
Values of Tme as Pgrcent of Msdlan lncome lol Rslative lmporlance of Travel Time coeffcisnts
2
Values of Tlme for Hom6-Ba8ed Work
I
Home-Bassd work Moclels ol Home-Based Work Modsls
Cllv cllvtl Cllvt CIyl) Crlvt cllvt I! Cllvll Clavl) Clwalkl Clwaltl C(xferl
C(cost) C(fare) C(perfi) C(cost) C(oper) C(fare) C(Park) C(tvt) c(tu0 c(lv0 c0v0
New Orleans 2.76 30.3 2.200 5.133 2.133
Mlnn/St. Paul 2.48 20.8 1.419 .968 1.4t9
Chlcago 2.s6 21.9 4.127 .844 4.127
Los Angsls 1.12 10.6
Seattls 2.09 20.3 't.f 00 .750
Clnclnnatl 2,98 27.4
Washlngton 19.s
San Franclsco 25.4
Dallas 25.3 25.3 10.1 1.862
Shlrley (low) 2.29 16.S 1 .018
Shirlsy (high) 3.74 27.4 2.00
Salt Lake cY 1.90 na 2.00
Portands 1.48 na 2.'112
ivt In-Vehicle Time

Not: lncoms valugs for 1979.
cost Total Cost
t Expressed as S/hour. opcr uto Operating Cost
?
Expressed as percentg of hourly incoms. fare Transit Fae
3
lncome for years othsr han 1979. park Parking Cost
ovt Ouof-Vehicle Time
walk Transit Walk Time
wait Initial Transit Wait Time
xfer Transit Transfer Time
66
TABLE 36 Default home-based lvork mode-choice model coefficients
Part A: ln-Vehicle and Oul-of-Vehicle Coefficients
'
' \;": Level-of-Service Variable Coefficient
':t
r,l
ln-Vehicle Travel Time (b,) -0.025
",'I
'I Out-of-Vehicle Travel Time (c,) -0.050
.:- i
Part B: Cost Coefficients (d,)
Value of Time as Percentage of lncome

Average
lncome 10.0Yo 15.0% 20.Oo/" 25.0o/" 30.0% 35.0%
s10,000 -0.031% -o.o21% -0.016% -o.012% -0.010% -0.00970

12,500 -0.025 -0.017 -0.012 -0.010 -0.008 -0.007
15,000 -o.o21 -0.014 -0.010 -0.008 -0.007 -0.006
.l
17,500 -0.018 -0.012 -0.009 -0.007 -0.006 -0.005
-- I 20,000 -0.016 -0.010 -0.008 -0.006 -0.005 -0.004
I
22,500 -0.014 -0.009 -0.007 -0.006 -0.005 -0.004
-l 25,000 -0.012 -0.008 -0.006 -0.005 -0.004 -0.004
27,500 -0.011 -0.008 -0.006 -0.005 -0.004 -0.003
30,000 -0.010 -0.007 -0.005 -0.004 -0.003 -0.003
. t' ':
32,500 -0.010 -0.006 -0.005 -0.004 -0.003 -0.003
- --:1 35,000 -o.009 -0.006 -0.004 -0.004 -0.003 -0.003
37,500 -0.008 -0.006 -0.004 -0.003 -0.003 -0.002
40,000 -0.008 -0.005 -0.004 -0.003 -0.003 -0.002
Note: Assumes 2,080 working hourVyear for conversion from annual income to g/hr. and /min.
Source: Derived from formula 6-5.
worksheets. Included in this section are sample worksheets (for work trips only) and travel survey data (if the sample size
that can be used to apply the model for a specific corridor or is large enough to represent the use of transit adequately).
interchange. The major steps to applying the model are as The best way to illustrate the application of the incremen-
follows: tal mode-choice model is through the use of an example.
1. Estimate initial mode shares (Figure 13, Worksheet 1),

EXAMPLE APPLICATION OF INCREMENTAL
2. Estimate the incremental change in level of service MODE-CHOICE MODEL
(Figure 14, Worksheet 2),
3. Apply model to compute revised mode shares (Figure One of the most common applications of the incremental
15, \Vorksheet 3), and mode-choice model is the analysis of HOV strategies. Of
4. Compute vehicle trips (Figure 16, Worksheet 4). particular concern is the analysis of HOV lanes on existing
freeways. HOV lanes can be created either by adding one or
The method for estimating existing or base mode shaes more lanes or by taking an existing lane and converting it
depends on the transit or HOV strategy being analyzed. The to HOV only. The other consideration in the analysis is the
two basic sources for the base mode shares are (1) observed number of persons per vehicle that defines an allowed HOV
traffic counts and occupancy level for the corridor being stud- on the lane. The process of analysis becomes iterative with
ied, or (2) base shares for trip interchange from existing the objective being stable shifts in mode shares and travel
regional travel-demand model. Information on base mode times. The following example illustrates this iterative analy-
shares may also be available from Census joumey to work data sis procedure.
6'7
Mode Cho i ce Mode I Cost Coeff c ents

-o. oo
-0.0
-o.06
-o. ool
-o. o'l
- o. 012
-o.or
!g
-o.01
-o.0r!
E -o.*
ot
-o.oa
-o.04.
-o.026
-o. oal
-o. 03
- o.032
-o.0.
ToloooTarslooo I aojom | asjooo | ojooo I slooo I golooo
ra-soo r7,5@ r22.tOO 327.500 332,500 337.500
lElcl l
^Ytga
o 1Ot + 1! o 2 a 25 x 30 I 3t
Figure 12. Cost cofficients as afunction of income.
Problem the PM peak hour. The following steps will be used to ana-
lyze this alternative:
An urban area has a major sixlane freeway oriented radi-
ally to the CBD. A major section with a length of l0 miles is
operating under congested conditions both now and into the Step I: Compute the base modal shares (Worksheet 1).
future. An alternative to the standard widening of the free-
way is the use of HOV lanes. An alternative that would be Using the existing count data, Worksheet 1 (displayed as Fig-
examined is the taking away of one of the lanes in each direc- ure 17) can be used to compute the base shares for each level
tion during the peak periods and converting these lanes to of vehicle occupancy and transit.
HOV only. Allowable vehicles for the HOV lane would be
buses and vehicles with two or more persons per vehicle.
Observed PM peak-hour counts revealed the following mix Step 2: Compute the existing HOV and non-HOV travel times.
of vehicles throughout the 1O-mile corridor:
From Chapter 10, the freeway capacity per hour, per lane
is 1,800 vehicles. Directional capacity is therefore 5,400
vehicles/hr. Free-flow speed would be 55 mph. This re-
Vehicle P.M. Peak-Hour sults in an uncongested travel time of 1l minutes (10 miles
Classification Count (Vehicles)
at 55 mph) in the corridor. From Chapter 9, the follow-
Drive-Alone 4,182 ing formula is used to compute existing congested travel
2 persons/vehicle 356 time:
3 persons/vehicle 231
4+-persons/vehicle
Vanpools
126
25
r:r,"[t*083x(:)"]
Trucks 540
The existing volume-to-capacity ratio is 1.01 (5,46015,400)

Total 5,460 and the resulting congested travel time for the corridor is
Tr = lt x [t + 0.83 x (1.01)s5] = 21
In addition to the above vehicles, the count program For the HOV lane, the initial assumption will be made that it
counted 100 person trips using buses in the corridor during operates at free-flow conditions and the travel time will be 11
68
Vehicle Factor Person Base

Count Trips Shares (%)
Auto
Drive-Alone x1=
2-person carpool x2=
3-person carpool x3=
4+-person carpool* x4=
Shared-Ride x2.5=
Vanpool x7=
"-,
I
Bus
'

'
i
Total
lToo% I
* Note: This assumes
;
i\'"''t*'4 that 4+-person carpools have 4 people in them. lf there are more than 4
Persons, this process will underestimate the number of person trips.
'I
Figure I 3. Worksheet I : computation of base shares.
Cost
ACOST Coef ut
ÎVTT ÔVTT
Auto
Drive-Alone
2.personcarpool(l-lx(-o'o25))-(l-lx(.o.o5o))-(|-l
3.personcarpool(l-lx(.o.o25))-{l-lx(.o.o5o))-([-l
4+.personcarpool(l-lx(-o.o25))-(|-lx(-o.o5o))-(l-l
Shared-Ride
Vanpool
Bus
Total
69
Revised
Lu, ut Put Shares (%)
Auto
Drive-Alone
2-person carpool
3-person carpool
4+-person crpool
Shared-Ride
Vanpool
Bus
Total
Figure 15. Worksheet 3: computation of revised shares.
Revised Total Revised Revised

Shares Person Person Vehicle
P,'(7") Trips Trips Trips
Drive-Alone +1=
2-person carpool +2=
s-person carpool +9=
4+-Pefson carpoor +4=

[|
Shared-Ride :)E-
Vanpool
Bus
rotar |;;l
Figure 16. Worksheet 4: revised vehicle trips.
ji .l
'?.. ]
. ._ i
,...: i
,= '
70
Vehicle Person Base

Count Factor Trips Shares (%)
F-
Auto
Drive-Alone x1= frr- I |-*"1

2-person carPool t.*l x2= |'."-l
3-person carpool i*l X3= l.';-l f *' I
4+-person carpool f"._l x4= f.''_l |^"-l
Vanpool
r-l x7= l-''.-l r;,-l
Bus f ''.l t*'l
Total l'3..-l t *.-" l
Figure 17. Worksheet I: computation of base shares-example HOV problem.
minutes. Therefore, the change in in-vehicle travel time for the The average vehicle occupancy excluding the buses is com-
HOV vehicles and buses will be -10 minutes. The non-HOV puted as
'I vehicles will have to use the remaining two unrestricted
lanes. For this initial iteration, the time for these vehicles will Averase vehicte occupancy =6'322;" = t.to
be assumed not to change. After the new shares and number
ofdrive-alone vehicles are estimated, the congested speed on Step 4: Check volume-to-capacity ratios for HOV and regu-
the unrestricted lanes will be computed. lar lanes.
Step 3: Compute change in modal utilities, shares, and vehi- The volume-to-capacity (v/c) raio for the HOV lane is equal
cle trips. to the two-person and above vehicles divided by the single-
lane capacity of 1,800. The result is
The revised vehicle trips are calculated using Worksheets 2
through 4 (displayed as Figures 18 through 20). A summary Vro, 4I1 +2'70+147+29
cro,
- = 0.48
of the revised shares and vehicle trips follows: 1,800
Revised Revised Revised

Mode Shares Person Trips Vehicle Trips
ÎVTT
Drive-Alone 59.84'/" 3,809 3,809
2-person carpool 13.08% 833 417 -10
3-person carpool 12.75"/" 812 27'l -10
4+-person carpool 9.26% 590 148 -'10
Vanpool 3.23% 205 -10

Bus 1.84"/" 117 : -10
. '.: i
...1 Total 6,366 4,674
'..i
71
Cost
LCOST Coef LUt
ÎVTT ÔVTT
Auto
Drive-Alone (F_l x(-o.o2s)+( F_lx(-o..so)).(l-;_l .F_l)=l oI

2-person carPool ( Tl;l x(-0.02s)).(l-l-l x(-o.oso)).( [;_l ' l-o-l )= f;;l
3-person carPool ( [_*-lx(-0.02s)).(T;-lx(-o.oso)).( l- o_l.l-o-l )= l-.rs-l
([ -l [;l )= l- 25
4+-person carPool -1o-l x(-o.ozs)).( f-o I x(-o.oso)).( I o ,. I
Vanpool ( l-.101 xG0.02s)).( [;l x(-0'0so)).( [;l . |ol )= I 25 I

_l
Bus ( I -10-l x(-0.025)).(l- o xGo'oso)).( [;_l " l-;l )= l--l
Total
Figure 18. Worksheet 2: computation of change in utilities-example HOV problem.
The v/c ratio is low enough that it can be determined that the Step 5: Compare AIVTT and iterate as necessary'
HOV lane will operate at free-flow conditions. The v/c tatio
of the non-HOV or regular lanes is equal to the drive-alone This congested travel time is 16 minutes longer than the
vehicles plus the trucks divided by the remaining two-lane existing conditions. Therefore, the drive-alone AIVZZ should
capacity. The result is reflect the increased travel times for the drive-alone mode'
This increase in congestion for the drive-alone mode makes
3,811+540 the HOV modes even more attractive. However, if the entire
Vnon-Hov _ =1.21 16 minutes is added to the drive-alone mode then the result-
Cuon-Hov 1,800 x 2
ing vlc ratio will probably be less than 1.21 and the travel
time will be less than37 minutes. The process can be iterated
The vlc ratio of the regular lanes is higher than the initial until the resulting shares for drive-alone produce avlc ratio
value of 1.01 for all three lanes. The new congested travel that is in balance with the time used for input to the change
time for the regular lanes is computed as in IVTT for drive-alone. For this example, a L'IVTT of +8
minutes for the drive-alone mode will be tested. The follow-
tt = ll x [t + o.a: x (1.21)ss] = 37 minutes ing is a summary of those results:
Revised Revised Revised

Mode Shares (%) Trips
Person Vehicle TriPs
ÎVTT
Drive-Alone 54.97% 3,500 3,500 +8
14.68% 934 467 -10

2-person carpool
3-person carPool 14.29"/" 910 303 -10
4+-person carpool 10.39% bot 165 -10
3.61o/" 230 JJ -10

Vanpool
2.06% 131 -10
Bus
Total 6,366 4,468

72
Revised
Lu, u Prxelut Shares (%)
Auto
Drive-Arone F-l t*l t.r*l f-"*"1

2-personcarpoor
|ol |G;l lG-l lT*"1
3-personcarpoor
t *l |G;l f r.- I @
4+-personcarpoor
I *l |G t r*rl t '*"1
vanpoor I *l t t"-l fs*l Et
Busl*]l*--lf---l l'",.
rotar t-r.r;l f;-"1
Figure 19. Worksheet 3: computation of revised shares-example HOV problem.
Revised Total Revised Revised

Shares Person person Vehicle
Pi&) Trips Trips Trips
Drive.Arone l*.ro*l tr--l-E=F;;;-l

2-personcarpoor
lr..orl trg3-l.E=|;-
3-personcarpoor
ltr.r. l f *l-Fl =f;l
x I 6.366 |
4+.personcarpoor
I r.ru I lG_l -E -|i-l
Vanpoor r*] [r*-l _E =[l_l
Bus l'* | |7l
rotar Iroo.oor.l E-roo-l li-r-l
Figure 20. Worksheet 4: revised vehicle trips-exampte HOV problem.
Average vehicle occupancy is now 1.40 versus the existing demand model or by factoring growth into the existing count
value of 1.27, or an increase of more than 9 percent. The data. To complete the analysis, a second alternative in which
resulting v/c ratios are the HOV lane is added to the existing regular lanes could be
examined. The same analysis process would be used.
467+303+165+33
HOV Lane vlc = = 0.53
1,800
ANALYSIS OF TRAVEL DEMAND
3500 + 545 MANAGEMENT (TDM) STRATEGIES
Regular Lanes vlc = = 1.I2
300
Travel-demand management is a set of strategies designed
to encourage the use of altematives to driving alone, particu-
The resulting travel time for the regular lanes is larly during the peak periods. The analysis of supply-side
TDM strategies that can be expressed as time and cost changes
t = ll x [t + o.sr x (1.12)55] = 28 minutes can be accomplished through the application of the incremen-
tal mode-choice model presented earlier in this chapter. This
The AÎVTT for drive-alone mode is *7 minutes as com- includes transit improvements, priority treatments such as
pared with the input value of *8 minutes. An additional iter- HOV lanes, and financial incentives/disincentives. The finan-
ation could be made and the acfial L,IVTT would be about cial strategies can include increased parking costs and tran-
*7.5 minutes; however, a difference of I minute is acceptable. sit cost reduction through subsidized or reduced fares. In a
report prepared for the Federal Highway Administration,
Summary of Example Problem Results Employer-Based Travel Demand Management Programs-
Guidance Manu\, a good overview of TDM strategies and
The following table summarizes the results of the take-a- a procedure for evaluating the employer-based strategies are
lane HOV alternative for the existing traffic conditions: presented.a The discussion of TDM analysis contained in this
Existing Take-a-Lane HOV

Conditions Alternative Change
Vehicle Trips - 4,929 4,668 -252

Person Trips * 6,366 6,366 0
Vehicle-Miles of Travel (VMT) 49,200 miles 46,680 miles -2,520
Vehicle-Hours of Travel (VH) 1,722 hours 1,81'l hours 89
Person-Miles of Travel (PMT) 63,660 miles 63,660 miles 0
Person-Hours of Travel (PH 2,228 hours 2,159 hours -69
HOV Lane Travel Time 21 minutes 11 minutes -10
Reg. Lanes Travel Time 21 minutes 28 minutes +7
HOV v/c n/a 0.53 nla
Regular v/c 1.01 1.12 +0.'11
Average Vehicle Occupancy 1.27 1.40 +0.1 3
' Excludes truck volumes
The results of this analysis show that although the HOV report uses that manual as a source and only a brief summary
lane will reduce total vehicle traffic in the corridor, it will of the analysis procedures is presented here. The user is
increase congestion in the remaining two regular lanes, referred to that publication from the FHV/A for a more
thereby increasing overall corridor vehicle hours of travel. detailed description of the analysis procedures.
There is, however, some saving in corridor person-hours of
travel. The example is for existing conditions; a more com-
rComsis Corporation, Ernployer- Based Travel Dernand Management Programs-
plete analysis should be done for future years. Future-year Gudance Manual, Prepared for US DOT, ETA, and FHA, Washington, D.C. (June
volumes can be obtained either from a regional travel- I 993).
74
The term TDM encompasses both alternative modes to driv- The AVR for a site is computed using a worksheet. The
ing alone and the techniques, or strategies, that encourage use data are collected from observed count data and through
of these modes. TDM modal alternatives include employee surveys. The number of employees arriving by
t'
each mode is converted to total vehicle trips to the site and
I
1' .l-l
.i
. Carpools and vanpools, can be compared with observed traffic counts.
. Public and private transit (including buspools and shut-
tles), and Step 3:Define Site's Modal "Bias."
. Bicycling, walking, and other nonmotorized travel.
A basic input to the procedure is the determination of
Alternative work hours are also a TDM strategy, such as pro-
whether the site is "transit favorable, rideshare favorable, or
grams that are designed to reduce the number of trips made
mode neutral." A worksheet is provided to make this deter-
during the peak periods, either by reducing the number of
mination. If the number of arrivals made by transit is more
days the commuters need to travel to the worksite or by shift-
than 50 percent of all other modes, then the site is considered
ing commuting travel to non-peak period times of the day.
..: i to be transit favorable. If the rideshare number of arrivals is
:.*,i Examples include
greater than 50 percent of all other modes, then the site is
. rideshare favorable. If neither of these conditions is true, then
Compressed work weeks-working 40 hours in less
the site is considered to be mode neutral. This mode bias is
than 5 days,
. used as input to the estimation of the percent vehicle-trip
Flexible work schedules-shifting start and stop times
reduction.
to less-congested times of day, and
. Telecommuting-working 1 or more days at home or at
a satellite work center closer to the home. Step 4: Calculate Peak Vehicle Trips (Optional).
Employer-based TDM programs often are the most effec- In some situations, the trip reduction may be required or
tive in reducing peak-period trips. TDM strategies can be cho- desired for a specific peak hour or peak period. A worksheet
E5'!1 sen to meet a relatively naffow set of worksite, operational, is provided for computing peak-period trips from total per-
and commuters' demographic and travel characteristics. There son trips traveling to site.
are many reasons for implementing an employer-based TDM
program, with the most likely being one of the following: Step 5: Set TDM Goal.
. Response to a trip-reduction regulation, The TDM goal is used to identify the appropriate type
. Solution to a transportation-related problem at the work and intensity of the TDM program. The goal is typically
site, expressed as a percent reduction in vehicle-trips. The proce-
. Expansion of employee (or tenant) benefits package, and dure contained in this manual uses this measure for the eval-
. Reduction in company expenses. uation and development of TDM programs and strategies.
Employ e r- B as e d Trav e I D emand M anag ement P ro grams-

Step 6: Develop TDM Program Options.
Guidance Manual presents a manual procedure for develop-
ing and evaluating a TDM program for a specific employer
In this step, the process of identifying TDM program pack-
site. The basic procedure requires the following steps:
ages that will meet the goal is begun. More than one package
of strategies will allow the achievement of the trip reduction
Step l: Define Site's Employment Type.
goal. A worksheet is provided that is used to evaluate modal
shift strategies, alternative work arrangements, and time-
The first step is to classify the site as either office or non-
shift actions. Tables are provided that are used to look up the
office. Office sites are substantially professional/white collar
jobs, with work schedules that fall within the daily peak percent vehicle-trip reduction based on employer type, start-
ing AVR, mode bias, and levels of transit, carpool, and van-
travel periods. Non-office sites include those that are sub-
pool support.
stantially blue collar or crafts/nonprofessional employment,
with work schedules that may or may not fall within tradi-
tional peak periods. Step 7: Estimate Trip Reduction Impacts.
Step 2: Define Site's Baseline Traffic Conditions (Starting A worksheet is provided that allows for the comparison
Average Vehicle Occupancy or Average Vehicle Ridership, and summation of several trial TDM strategies that can be
AVR). combined into the TDM program.
75
The final TDM program will be a result of the trial testing CASE STUDY
of multiple TDM strategies and the nal packaging of multi-
ple actions. The report contains all of the necessary work- Because transit use in the Asheville region is so small, rep-
sheets, trip reduction factors, and an example application. resenting less than I percent of the average daily person trips
This example is applicable only to a single site. Often it is made in the region, we have chosen not to use a mode-choice
necessary to make a similar assessment for an entire area. model in this case study.
Implementing Effective Travel Demand Management Mea- If we had wished to do so, however, we would need to
sures: Inventory of Measures and Synthesis of Experience, build a transit network and transit travel time matrices (skim
which was published by the FTA, discusses techniques that tables) and apply one of the mode-choice models described
may be applied to an area rather than a site. in this chapter.
76
' ':
I
,l
.j.i

CHAPTER 7
AUTO MO BI LE.OCCU PANCY CHARACTE RISTICS
INTRODUCTION smaller carpools-four- and five-person carpools declined

by more than 50 percent, and two-person carpools now
As the mode-split process outlines, the most commonly account for 85 percent of all carpools.a
used modeling approach employed by urban transportation Factors that may influence auto occupancy are: (l) the
planners generates total person trips and then distributes and journeys that are made for different purposes at different
splits these trips into auto and transit modes. Auto-occupancy times of day, (2) conditions that exist where the trip begins
factors are then applied to the auto person-trip dataset to and ends, and (3) characteristics of the travelers and the
produce a vehicle-trip table for use in the traffic-assignment households in which they live. These elements describe the
process. The importance of auto occupancy in this process of
same kinds characteristics represented by the typical
becomes apparent when it is considered that a slight error in mode-choice model.
the auto-occupancy rates (e.g., 1.36 versus 1.50, or about l0 Available data from various urban areas were analyzed to
I percent) translates into a difference of more than 10,000 develop some insight into the variation of auto occupancy with
vehicles per day on a high-volume facility carrying 150,000 these factors, especially travel for different purposes through-
person trips.
out the day. Tables for estimating auto occupancy by urban-
Contrary to the assumption made in NCHRP Report 187, ized area population, trip purpose, and time of day have been
the overall trend in vehicle-occupancy rates has been a provided to assist in responding to auto-occupancy questions.
decline during the last few decades. Factors that may have Analysis of local or work-site-based TDM efforts can
contributed to this decline include the increase in auto owner-
often provide better information on vehicle occupancies than
ship and the decrease in household size. Since 1960, the are presented in the tables, which are based on national data.
number of households has increasedby 73 percent while the
TDM includes a variety of techniques to cope with escalat-
number of persons per household has declined 21 percent.
ing traffic. Metropolitan areas of all sizes are looking to a
The percent of households with no vehicles available has
mix of transportation modes and are initiating programs such
dropped 46percent in the same period.r
as ridesharing, developer requirements, and the exclusive use
Decreases in the number of persons per household mean
or preferential treatment of transportation facilities to serve
that there are fewer persons in each household that could be
high-occupancy vehicles. Work-site-based strategies include
traveling together for any home-based trip purpose. As the
carpool matching, subsidized transit passes, free and priority
number of employed persons per household has risen, people
parking for carpools, and flex-time. Programs like these are
are chaining trips to accomplish different purposes on their
meant to reduce vehicle trips, either regionwide or to a spe-
way to and from work, such as trips to schools or day care
cific site, to provide additional capacity, conserve energy,
centers, which makes it more difficult for carpools to form
and improve air quality. In many areas, vehicle occupancies
for work-related travel.
should be viewed as a policy input to the planning process
In the last decade, the increase in drive-alone trips to work
rather than as an output. The section, Usefulness of Region-
exceeded the number of new workers.2 Persons who travel in
ally Developed Models, discusses how TDM or other local
carpools, especially carpools involving members from dif-
data can be used to augment the tables and produce better
ferent households, constitute a very small portion of all trav-
estimates of auto occupancy.
elers. Only 13 percent of persons traveling to work will share
a ride on any day, and only 7.5 percent of all vehicles trans-
porting people to work will be a carpool or a vanpool.3 In BASIS FOR DEVELOPMENT
addition, over the last decade the shift has been to ever-
Investigation indicates that, although many reports have
been written to describe the results of counts or surveys of
tJourney-ro-Work Trends in the IJnited States and its Mjor Metrcpolitan Areas,
1960-1990,lorUS DOT, FHWA, and Office of Highway Administration (1994) p.2-2.
?Alan Pisarski, New Pe rspectites in Commuting for IJS DOT, FHWA,
and Otfice of
Highway Management (1992) p.5. aAlan Pisarski, Travel Beltavior Issues in the 90s, US DOT and
Federal Highway
1J
o ur ney- to-W ork Tt e nds, page 2-6, 5- 16, 5 - 17 . Administration for Office of Highway Infomation and Management ( 1992).
77
vehicle occupancy, relatively little attention has been given will usually travel together for other trip purposes because
to the development of procedures to determine auto occu- they want to be together during travel and when they get to a
pancy. It appears to be common practice to develop auto- common destination.
occupancy factors by trip purpose from base year data and to Table 37 presents average auto-occupancy values by
use this one set of factors for all subsequent planning efforts. urban area size and trip purpose for 1990. These values
The major data sources used in the development of the reflect average daily auto-occupancy rates. Compared with
auto-occupancy rates presented here were the NCllrRP Report 187 rates, the 1990 rates are significantly
lower for all purposes except home-based other and non-
. The Nationwide Personal Transportation Study (NPTS, home-based. This reflects the rise in auto ownership and
1990), drive-alone trips, as well as the increase in suburb-to-suburb
. Travel Behavior Issues in the 90s (Alan Pisarski, US travel discussed earlier.
DOT, and Federal Highway Administration for Office
of Highway Information and Management),
. VARIATION IN AUTO OCCUPANCY
Vehicle Occupancy Determinants (Barton-Aschman BY TIME OF DAY
Associates, Inc., Arizona Department of Transportation.
US DOT, and FIIWA, Reporf Number FHWA-4289- As percentage of trips by purpose shifts throughout the
252, August 1989), and day, vehicle-occupancy rates also vary by time of day. A
. Various urban transportation study reports. larger portion of trips during the AM peak are home-based
work trips, which exhibit the lowest occupancy rates; there-
fore, average vehicle occupancy for peak periods is often
FEATURES AND LIMITATIONS
lower than for non-peak periods. As noted above, shopping
and social trips generally exhibit higher vehicle occupancies;
The modeler needs to evaluate the output of the mode-
these purposes account for a greater portion of all trips made
choice model. Some mode-choice models produce trip tables
that are given in person trips while others produce vehicle during off-peak periods.
trips. Ifvehicle trips were produced, then the auto-occupancy
In addition to the overall occupancy varying throughout
the day, the auto occupancy for individual trip purposes also
calculation has already been made and does not need to be
varies by time of day. Table 38 presents adjustment factors
repeated here. The information in this chapter is useful, how-
ever, as a test of the reasonableness of the auto occupancies
for time-of-day variation in the average auto occupancy by
generated by the mode-choice model. trip purpose. Because of data limitations, this table was cre-
ated without regard to urban size and thus should be used for
In an effort to develop auto-occupancy estimates for use
in urbanized areas of varying population size, investigation all urban sizes.
was undertaken to determine what factors influenced auto-
occupancy rates and what differences are observed in vari- VARIATION IN AUTO OCCUPANCY
ous urban areas, and by trip purpose, time of day, trip length, BY INCOME LEVEL
and household income.
The data presented here reflect average auto-occupancy In addition to the variables discussed thus far, auto occu-
rates as of approximately 1990 (transportation studies con- pancy is known to be a function of the income level of the
ducted after 1985 and the NPTS conducted in 1990). trip-maker and of parking cost at the destination of a trip; that
is, the auto occupancy of low-income trip-makers is higher
than for similar trips by high-income trip-makers, and the
VARIATION IN AUTO OCCUPANCY auto occupancy for trips to high-parking-cost areas is higher
BY URBANIZED AREA POPULATION than for comparable trips to low-parking-cost areas.
AND BY TRIP PURPOSE
This basic relationship between auto occupancy and the
economics of travel is extremely important but is often
Trip purpose is the most significant factor influencing auto
neglected in the planning process. In particular, the use of
occupancy, where other factors such as household income
average auto-occupancy rates by trip purpose will tend to
and trip distance are less important determinants of vehicle
overestimate vehicular trips to areas of high parking cost and
occupancy.s Going shopping or to different forms of enter-
underestimate vehicular trips to areas where parking costs
tainment are the most likely trip purposes to represent higher
are either low or nonexistent.
vehicle occupancies. While persons may travel together in
As part of this user's guide, generalized relationships
carpools to get to work in order to save money or because
between auto occupancy and income level are provided to
they have no other form of transportation available, people
assist the user in assessing the effects of such variables. These
rates are presented in Table 39. This table should be used for
5Alan Pisarski, Nle Perspeclives in Conunutirtg for US DOT, FHWA, and Office of all urban sizes. More variation is indicated by demographics
Highway Management. p.8 (1992). and income than by city size characteristics. The data did not
. l.-
:1 .:
fi
78
TABLE 37 Average daily auto-occupancy rates by urbanized area population and purpose
Trip Purpose
Urban Area Size HBW HBShop HBSoc HBother NHB AI
Updated Parameters
.1
50,000 to 199,999 1.11 1.44 1.66 1.67 1.66 1.49

200,000 to 499,999 1.12 1.48 1.72 1.6s 1.68 1.s1
500,000 to 999,999 1.13 1.45 1.66 1.65 1.66 1.48
1,000,000+ 1.1'l 1.48 1.69 1.66 1. 1.49
Source: NPTS, 1990
Trip Purpose
.t Urbanized Area
Population HBW HBShop HBSoc HBOther HBNW NHB Ail
.'.
','1
''_l
Parameters From NCHRP 187
'I
50,000 to 100,000 1.38 1.57 2.31 'l .52 1.82 1.43 1.50
s.*]*r 100,000 to 250,000 1.37 1.57 2.31 'l .52 1.81 1.43 1.50
250,000 to 750,000 1.35 1.57 2.30 1.s2 1.77 1.43 1.50
750,000 to 2,000,000 1.33 1.58 2.29 1.51 1.74 1.43 1.51
TABLE 38 Auto-occupancy rate adjustment factors by time of day
HBShop HBSoc HBO NHB
12:00 Midnight to 5:00 a.m.

5:00 a.m. to 6:00 a.m. 0.08 {.30 -0.63 0.09 -o.24
6:00 a.m. to 7:00 a.m. 0.03 -0.23 -0.29 0.11 -0.13
7:00 a.m. to 8:00 a.m. 0.06 0.11 -0.03 o.27 0.19
8:00 a.m. to 9:00 a.m. 0.03 -o.17 -0.03 o.2'l -o.o2
9:00 a.m. to 10:00 a.m. -0.07 -0.08 -0.09 -0.30 -o.24
10:00 a.m. to 1 1 :00 a.m- -0.04 4.15 -0.01 -0.09 -0.06
1 1 :00 a.m. to 12:00 Noon -0.11 -0.01 -0.14 :0.15 -0.07
.uJ
.l r . 12:00 p.m. to 1:00 p.m. 4.O7 0.00 {.11 -0.16 -o.12
1:00 p.m. to 2:00 p.m. 0.04 -0.03 0.13 -0.10 -0.18
2:00 p.m. to 3:00 p.m. 0.09 -o.10 0.00 o.22 0.03
3:00 p.m. to 4:00 p.m. 0.07 -0.06 0.09 o.22 0.06
4:00 p.m. to 5:00 p.m. 0.02 0.01 0.07 0.06 -0.05
5:00 p.m. to 6:00 p.m. -0.04 0.09 0.0s -0.05 -0.08
6:00 p.m. to 7:00 p.m. -0.01 o.22 0.03 o.14 o.37
7:00 p.m. lo 8:00 p.m. -0.01 0.34 0.28 0.06 0.41
8:00 p.m. to 9:00 p.m. 0.07 0.25 o.o2 o.17 0.31
9:00 p.m. to 10:00 p.m. -o.o7 0.19 0.18 -0.20 o.17
1o:00 p.m. lo 11:00 p.m. -0.10 -0.03 -0.01 -0.16 0.01
11:00 p.m. to 12:00 Midnight -0.03 o.02 -0.20 -o.22 0.08
Source: NPTS,1990.
'79
TABLE 39 Auto-occupancy rates by income ctegory and purpose for urban areas
rip PurPose
lncome
't.19 1.49 1.77 1.66 1.69

Low
1.47 1.67 1.65 't.57
Medium 1.12
1.1 1.43 1.s6 1.58 1.s0
High 1
1.12 1.44 1.63 1.62 1.56

All
Source: NPTS,1990.
allow a direct categorization of parking costs; it is assumed recreational trips, which exhibit higher occupancy rates' are
for this table that these costs are related to trip type. more likely to be made on arterials and to be of shorter trip
length.
The result of these differences in auto occupancy by trip
VARIATION IN AUTO OCCUPANCY purpose and length is that the highest vehicle occupancies on
BY FACILITY TYPE weekdays occur on lower-volume roadways, and during off-
peak hours, especially on arterials in suburban areas in the
Thus far, it has been shown that auto-occupancy rates vary
evening. The lowest vehicle occupancies occur on higher-
by trip purpose and by time of day. Auto occupancy is also a
volume roadways (particularly freeways in core and urban
function of trip length in time or distance, and the length of
areas) during the AM peak when work trips predominate'
the trip often influences the type of roadway used for travel'
As home-based work trips are generally the longest dis- Vehicles traveling on freeways have lower occupancy than
those on aferials and collectors because of the different types
tance trips made by a household, they represent a higher por-
tion of all trips made on freeways in a planning region, espe- of trips these separate roadways serve. Table 40 presents
cially during the peak commute hours. Other types of travel, auto-occupancy rates by facility type and area type for peri-
such as parents driving their children to school or social- ods during a l2-hour daY'
TABLE 40 Auto occupancy by roadway type and area type
A.M. P.M.
Area and RoadwaY TYPe Peak Midday Peak Evening All DaY
All Facilities in the Region 1.226 1.335 1.361 1;385 1.s04 1.337
All Freeways in the Region 1.204 1.308 1.332 1.297 1.396 1.291
All Arterials n the Region 1.233 1.343 1.369 1.408 1.s33 1.350
All Facilities in lhe Core 1 .191 1.289 1.331 1.293 1.403 1.282
Freeways in the Core 1.185 1.270 1.252 1.278 't.370 1.262
Arterials in the Core 1 .194 1.297 1.347 1.299 1.417 1.291
All Facilities in Urban 1 .195 1.315 1.324 1.361 1.494 1.312

Freeways in Urban 1.165 1.284 1 .316 1.292 1.389 1.270
Arterials in Urban 1.204 1.324 't.327 1.380 1.523 1.324
All Facilities in Suburban 't.293 1.395 1.430 1.483 1 .590 1 .410
Freeways in Suburban 1.304 1.408 1.42',1 1.334 1.446 1.375
Arterials in Suburban 1.291 'I .393 1.432 1.s08 1.61s 1.416
A.M. Peak 6:00 a.m. to 9:00 a.m.

Midday 9:OO a.m. to 12:00 p.m. and 2:00 p'm. to 4:00 p.m'
Noon 12:00 p.m.to 2:00 P.m.
P.M. Peak 4:00 p.m. to 6:00 P.m.
Evening 6:00 p.m. to 7:00 P.m.
Source: Limited urban transportation studies'

80
DATA REQUIREMENTS demonstrate how the user can quickly calculate reasonable
AND EXAMPLE PROBLEMS
values for auto occupancy under varying conditions, know-
The goal of this chapter is to provide simplified proce-

ing little more than the population of the study area. It
should be noted, however, that many problems facing urban
dures to assist the user in developing answers related to auto
planners are complex and require more sophisticated esti-
occupancy in relatively rapid order. In the development of
mating procedures than those outlined here. One example
auto-occupancy parameters, considerable care was taken to
would be the impact of an exclusive carpool and bus lane on
minimize necessary data acquisition for application and yet
produce reasonable estimates.
air quality-a problem that requires an estimate of the shift
Following are several example problems designed to illus- of persons from the automobile to public transit, and from
trate how the procedures developed can be employed to pro-
low- to high-occupancy vehicles in a corridor as a result of
a change in travel time. In cases such as this, the usefulness
vided reasonable estimates of auto occupancy given limited
information with which to work. of regionally calibrated policy-sensitive models cannot be
overstated.
Problem 1; What are reasonable values for all-purpose trip In addition, regionally developed information can play
auto occupancy and HBW-trip auto occupancy in an urban_ an important role in the refinement of the average nation-
ized area of 275,000 population? wide information presented here. For example, a simple
table of auto-occupancy rates by trip purpose reflecting
Solution 1: Table 37 indicates that for an urbanized area of conditions specific to the user's study area could be used in
200,000 to 499,999 population, the auto-occupancy rates for place of the rates presented in Table 3J as a more accurate
total trips and HBW trips in 1990 were 1 .5 1 and I . l2 persons starting point in the estimation process. Other local infor-
per vehicle, respectively. mation (if available), such as auto occupancy by time of
'i
day, by land-use caregory, and the like could also be used
Problem 2.' A new shopping center is undet construction in to refine the tables. In summary, if similar information is
an urbanized area of 225,000 population. It is estimated that available for the user's study area from the local regional
the center will generate 10,000 auto person-trips per day, planning organization, it should be used. Lacking such
and that 1 8 percent of such trips will occur during the period
information, however, the procedures documented here can
8:00 PM to 9:00 PM. How many vehicle trips can be antici-
be applied to develop, quickly and economically, reason-
pated during this peak hour?
able answers to a diverse set of questions relative to auto
occupancy.
Solution 2: Table 37 indicates that the average auto occu-
pancy for the purpose of HBO shopping in an urbanized
area of 225,000 population is about 1.48. Given that most
trips during this hour wilt be shopping trips, it is reasonable CASE STUDY
to use this occupancy for all trips if more detailed informa-
tion on trip purpose is not available. Table 38 indicates that The estimation of vehicle trips for the purposes of the
an adjustment factor of 0.25 should be used during the period Asheville, North Carolina, case study is based on the auto-
8:00 PM to 9:00 PM. The number of vehicle trips to be antic- occupancy factors found in Table3i. The average values by
ipated is, therefore, 1,040; that is, trip purpose for urban areas with under 200,000 population
were applied to convert the 520,92I total person trips into
10900 Person Trips x 0.18
1.48+0.25 - I'U40 350,017 vehicle trips, as shown in the following equations:
-
HBW Vehicle Trips = 124,87 5 HBW Person Trips + I .ll
= 112,500 HBW Vehicle Trips
USEFULNESS OF REGIONALLY
DEVELOPED MODELS HBO Vehicle Trps = 278,393 HBO Person Trips + 1.61
= 166,i02 HBO Vehicle Trips
The application of the auto-occupancy estimating proce- NHB Vehicle Trips = ll7 ,652 NHB Person Trips + 1.66
dures developed through these example problems attempts to
= 70,875 NHB Vehicle Trps
8r
CHAPTER 8
TI M E.O F.DAY CHARACTE R ISTICS
INTRODUCTION Transit use is highly oriented to the AM and PM peak peri-

ods with lesser amounts of travel at other times. The differ-
Through the mode-choice step in the travel forecasting ences between auto and transit use are most visible during
process, the forecasting procedures usually consider total the evening hours when many trips are made for social and
travel over the full 24-hour day. For many applications, shopping purposes. Compared with auto trips, few evening
travel must be estimated for specific periods or hours of the trips are made via transit. Also, the peak hour for transit
day. These applications may include the analysis of highway travel usually occurs during the AM peak period-caused
facilities, transit services, and emissions. Peak-period speeds by a concentration of work trips-whereas the peak hour
and volumes are critical for assessing the level of service pro- for auto-related trips usually occurs during the PM peak
vided by the transportation system, the competitiveness of period.
transit with autos on the highway network, and the size of the The AM peak is most critical for air quality analysis, as
transit fleet. morning emissions of volatile organic compounds (VOCs)
Analyses of special times are generally required to judge and nitrous oxides (NO-) have a longer time to react fo light
transportation system requirements. During limited periods than do pollutants emitted in the PM peak. As a resulf, ozone
during the day-the peak time period-the transportation concentrations typically peak during the late morning or
system is loaded, and sometimes overloaded, with travelers. early aftemoon hours. Aftemoon winds also tend to disperse
For general highway traffic, the critical peak hour most often pollutants more than in the early morning.
occurs during the afternoon when people are returning from On the other hand, the PM peak is critical for system
work, going shopping, completing recreational trips, and analysis because areawide traffic volumes and congestion
being picked up at school. However, critical traffic move- are typically higher during the aftemoon peak. Ultimately,
ments may occur at other times, particularly during the morn- the choice of which peak period to model must take into
ing commute hours. account such considerations as the availability ofcount data,
In 1990, for the first time, the census provided start-time previous modeling efforts, local conditions, and the applica-
information for work trips. In aggregate, these data show tion for which the model is intended.r
that there is no such thing anymore as a peak hour. The Time-of-day factors are applied to the mode-specific trip
most heavily loaded hour in trip starts is from 7:00 AM to tables produced by the mode-choice models. The most
8:00 AM, and this hour accounts for only 32percent of com- straightforward applications stratify the factors only by trip
muting trip start times. purpose; however, time-of-day factors can be stratified by
Geographic location can add another dimension to the both trip purpose and mode if mode-specific surveys and
time-of-day stratification to account for unique peaking char- counts have been obtained. Otherwise, it is assumed that the
acteristics of individual corridors or subareas. This is partic- same mix of purposes is uniform across modes.
ularly applicable where a subarea contains a major generator The purpose of this chapter is to provide tables to allow
such as a hospital or university that has significantly differ- determination of hourly travel from estimates of total daily
ent peaking times than conventional commuter trips. travel. Material is provided in this chapter for both automo-
A final issue with regard to peaking occurs after the bile travel and transit travel. The data are also extremely use-
assignment and aggregation of link volumes by period. ful in converting daily work trips from census information to
Several issues with regard to capacity and facility design re- peak-hour all-purpose trips. The techniques used for vehicle
quire peak-hour estimates of volumes on highways. transit travel are different from those used for transit time-of-day
lines, transit stations, and park-and-ride lots. Mode-specific analysis; therefore, a separate section is provided for each.
factors capture more fully the unique peaking characteris-
tics of each mode. For example, carpools on HOV facil-
rGreig Hrvey and Elizabeth Deakin, A Manual of Transportotion Air QualiD' Mod-
ities tend to have peaks that are sharper than general high- eling for Metropolitan Plottning Organizatiorts, National Assocation of Regional
way traffic. Councils (1992).
'..'i
82
.l
.l l
BASIS FOR DEVELOPMENT analysis. The general organization of the charts provided in
l. this chapter is by the four urbanized area population groups.
Time-of-day analyses are used for several types of studies, Each set of charts (by urbanized area population) is further
''..t
l
'I
and, since the introduction of Transportation System Man- stratified to present data to
agement (TSM) requirements, are becoming a more critical
part ofthe overall transportation planning process. Examples . Analyze auto driver travel by trip purpose,
of time-sensitive studies are as follows: . Analyze total vehicle travel,
. Determine total vehicle travel, by time period or in
t. Traffic impact studies are analyses to determine the aggregate, from internal auto driver trips,
impact a specific residential, commercial, or other type . Determine trip volume by route type, subregion, and ori-
of development has on the area transportation network. entation to study area core, and
2, Trip accumulation studies are usually done to deter- . Determine directional split of travel by route type, sub-
mine the peak accumulation of vehicles for parking region, and orientation to the study area core.
studies, taking into account the mix of trip purposes
involved. The tables were developed using the NPTS with confu-
3. Highway v/c studies are evaluations using peak factors mation from analysis of travel data of home interview data
(essentially the type of information provided later in sets from around the country.
this chapter) to determine peak-loading conditions in
vehicles per hour (VPH) for highway traffic assign-
ment and determination of capacity requirements. DIURNAL DISTRIBUTION
4. TSM studies specifically address transportation solu-
tions for the critical peak period, generally in the form The use of diurnal factors in time-of-day analysis allows
of traffic engineering or operations improvements. peak-hour assignments that are representative of the peak-
5. TDM has found a significant place in transportation hour direction of trips and the percent by hour. These factors
planning in the past decade. As roadway capacities are used to produce peak-hour directional volumes.
have filled, ways to fit more travelers in the same road Twenty-four-hour production/attraction trip tables are
tff<stl
space have become an alternative to widening or build- converted to time-of-day-specific, origin/destination trip
ing new roadways. TDM strategies are used not only to tables by applying time-of-day and directional split factors.
lessen the number of vehicles on the network during the The creation of time-of-day origin/destination trip tables
peak time periods, but also to shift some trips to non- from 24-hour origin/destination trip tables is somewhat eas-
peak time periods. ier as the tables need to be factored only by time-of-day fac-
tors, not by time-of-day and directional split factors.
This shifting can seriously affect the time-of-day trip char-
acteristics of a region. Adjustments in the trip table must
Use of Time-of-Day Tables
be made to compensate for TDM efforts, either in place or
planned, that affect the time when trips are made. This section presents tables of travel by time of day and by
The procedures presented here are based on observed purpose for the different population groups. An example is
vehicle-miles of travel (VMT). VMT is the product deter- presented offactoring a daily vehicle trip table by purpose to
mined when a given trip is multiplied by its trip length an AM peak-hour trip table.
(distance). As such, it is truly a measure of travel and not a There are two basic approaches to developing estimates of
measure of the distribution of trips during a Z4-hour period. directional peak-hour vehicle volume: (1) post-processing of
The trip-length distribution may vary by trip purpose over a daily highway assignments using link-based peak-hour and
24-hotn period. However, within the context of acceptable directional percentages and (2) preassignment factoring of
transportation planning procedures, such as trip-distribution daily trip tables by purpose, using factors for AM peak, PM
modeling, the VMT distribution can be used to approximate peak, and off-peak periods.
the distribution of trips by purpose. The first approach historically has been used in conjunc-
tion with the assignment of a daily vehicle-trip table. The
peak percentages for a link may be based on Z4-hour
VEHICLE TRAVEL machine counts of traffic, but most commonly the assigned
ADT is multiplied by a single factor ranging between 8 and
Time-of-day analysis is usually undertaken at one of two 12 percent of daily traffic to achieve an estimate of total bi-
points: (1) just after application of the auto-occupancy proce- directional peak-hour travel. A directional split (e.g., 60/40)
dures to isolate a time period for further analysis or (2) after based on observations of traffic conditions is then applied.
assignment of 24-hour travel in preparation for capacity This procedure yields a rough approximation of peak traffic
83
that may be appropriate for smaller urban areas where the this section. Using this reassignment factoring process, aver-
duration and intensity of congestion is limited. age daily traffic assignments can be produced by summing the
The preassignment approach uses time-of-day factors to results of the AM peak, PM peak, and off-peak assignments.
create the AM peak, PM peak, and off-peak trip tables by pur- Table 4l presents the percent of vehicle trips by hour by trip
pose that are then used in the assignment of vehicle trips to the purpose for different urban population groups. Table 42 shows
network. An example of this method is included at the end of the diurnal distribution of trips by time and purpose. The data
TABLE 41 Percent ofvehicle trips by hour by trip purpose
Urban Size = 50,000 to 199,999

Hour Beginning Hbto NI.tB All Purposes
Midnight 0.33 0.40 0.49 0.41

1:00 a.m. o.o7 0.17 0.12 o.12
2:00 a.m. 0.50 0.23 0.27 0.33
3:00 a.m. 0.61 0.07 0.12 0.27
4:00 a.m. 1.00 0.08 0,00 0.36
5:00 a.m. 2.79 0,18 0.06 1.01
6:00 a.m. 8.34 1 .10 0.46 3.30
7:00 a.m. 13.57 5.53 2.07 7.06
8:00 a.m. 7.84 s.64 2.27 5.25
9:00 a.m. 3.36 4.27 3.76 3.80
10:00 a.m. 2.79 s.86 s.40 4.68
11:00 a.m. 2.65 6.44 7.22 5.44
Noon 3.72 6.40 11.26 7.13
1:00 p.m. 3.26 6.34 8.77 6.12
2:00 p.m. 4.12 7.70 8.31 6.71
3:00 p.m. 8.30 8.06 9.74 8.70
4:00 p.m. 10.31 7.25 9.28 8.9s
5:00 p.m. 10.66 7.32 8.56 8.85
6:00 p.m. 5.01 7.44 7.19 6.55
7:00 p.m. 2.79 6.71 5.52 5.01
8:00 p.m. 1.72 5.24 3.46 3.47
9:00 p.m. 2.29 3.95 3.06 3.10
10:00 p.m. 2.26 2.25 1.55 2.02
11:00 p.m. 1.69 1.37 1.06 1.37
source: 1990 NP[S.
Percent of Vehicle Trips by Hour by Trip Purpose

Urban Size = 50,000 - 199,999
9.20
l-
rs
o
Eto
o
F
o
E
o
Io
o-0
Mldnlght O4:( AM O8:q) AM Nqr (X:O Ptf O8:d) Pt
2 AM 06:00 AM f 0:00 AM m:(x) PM OE:(x) PM r0: PM
Trip StaftTime
*HBW +HBO *NHB

84
TABLE 41 (Continued)
Urban Size = 200,000 to 499,999

Hour Beginning Ail HUTpOSeS
Midnight 0.35 0.29 0.48 0.37

1:00 a,m. 0.22 0.26 0.16 0.21
-,.- 't. 2:00 a.m. 0.35 0.15 0.38 0.29
3:00 a,m. 0.06 0.22 0.10 0.13
4:00 a.m. 1.03 0.17 0.16 0.45
5:00 a.m. 2.s7 0.29 0.00 0.95
6:00 a.m. 8.58 1.20 0.48 3.42
7:00 a.m. 14.46 5.28 1.33 7.02
8:00 a.m. 8.06 5.43 2.45 5.31
9:00 a.m. 3.03 4.72 3.08 3.61
10:00 a.m. 2.63 5.15 4.62 4.13
11:00 a.m. 2.29 5.09 8.39 s.26
Noon 2.86 6.43 10.04 6.44
1:00 p.m. 2.86 6.19 9.08 6.04
2:00 p,m. 4.40 7.50 9.20 7.03
3:00 p,m. 6.58 8.25 10.36 8.40
4:00 p.m. 9.78 7.45 10.25 9.16
5:00 p.m. 12.24 7.23 9.20 9.56
6:00 p.m. 6.86 8.47 s.84 7.06
7:00 p.m. 2.63 6.72 4.3',1 4.55
8:00 p.m. 1.94 5.36 3.67 .bo
9:00 p.m. 2.29 3.96 3.14 3.13
10:00 p.m. 2.0s 2.47 2.02 2.18
1l:00 p.m. 1.89 1.76 1.28 1.64
Source:1990 NPTS.

Urban Size = 200,000 - 499,999
9.20
l-
rs
E
6'
Eto
o
t-
o

o
o
o
40
Mldnlght O4:q)AM 08:q)AM Noon (X:q) PM 08:00 PM
0A:00 AM (F:(x, AM l0: AM 2:00 PM 06:( PM l0: PM
Tdp StartTime
+HBW +HBO *NHB
are presented separately for the three trip purposes-HBW, peak period. It is the peak-period volume that dictates the
HBO, NHB, and for all purposes combined. required size of the transit fleet; therefore, the following
Table 42 breaks down the frequency distribution of start approach for transit time-of-day analysis has been taken.
time and trip purpose into a diurnal format. From this table, Table 43 gives factors for deriving ridership estimates for
either peak hour, peak period, or, for air quality modeling, any of the following time periods if patronage for any one
8-hour period can be determined. of the time periods is known: (1) total transit patronage
expected to occur on an average weekday; (2) peak-period
Transit H o urly Distributions
trips where the peak period is the sum of transit patronage
expected for the 2 morning peak hours and the 2 evening
Most analyses of time-of-day distribution of transit vol- peak hours; (3) the peak hour of transit patronage for the
umes center about a peak period or specific segments of the day, usually occurring between 8:00 AM and 9:00 AM; and
85
TABLE 4l (Continued)
Urban Size - 500,000 to 999,999

Hour Beginning ffi All Purposes
Midnight 0.43 0.24 0.25 0.31

1:00 a.m, 0.43 0.27 0.12 0.27
2:00 a.m. 0.34 0.33 0.23 0.30
3:00 a.m. 0.37 o.12 0.03 o.17
4:00 a.m. 0.98 0.1s o.25 0.46
5:00 a.m. 2.46 0.22 0.08 0.92
6:00 a.m. 9.31 1.14 0.29 3.58
7:00 a.m. 15.21 5.22 1.83 7.42
8:00 a.m. 8.07 5.28 2.92 s.42
9:00 a.m. 3.04 5.21 3.55 3.93
10:00 a.m. 1.64 4.89 4.69 3.74
11:00 a.m. 1.95 5.19 7.71 4.95
Noon 3.44 6.16 10.32 6.64
1:00 p.m. 2.67 5.63 8.88 5.73
2:00 p.m. 3.70 7.93 8.06 6.56
3:00 p.m. 7.16 8.25 9.43 8.28
4:00 p.m. 11.14 7.15 9.63 9.31
5:00 p.m. 11 .17 7.21 10.17 9.52
6:00 p.m. 5.93 7.83 6.39 6.72
7:00 p.m. 2.83 8.30 5.36 5.50
8:00 p.m. 1.92 5.21 4.44 3.86
9:00 p.m. 2.20 4.08 2.83 3.04
10:00 p.m. 1.80 2.s0 1.32 1.87
11:00 p.m. 1.80 'l .49 1.21 1.50

Urban Size = 500,000 - 999,999
g- 20
t-
rs
E
o
Eto
o
l-
o
,5
o
I
o
o-0
Mldnlght
-
(X:OO AM 08:00 AM Nn O4:) PM 08: PM
O2:OO AM 06: AM 10:00 AM e PM 0:00 PM to:m PM
Tdp Startme
riHBW +HBO *.NHB
(4) peak-hour direction where transit patronage is estimated all trip purposes (HBV/, HBO, NHB) is not practical. For
in the direction of peak flow during the peak hour' example, the user may wish to know the consequences of
The user should keep in mind that the factors are to be total trip movement within a major travel corridor and only
applied to all-purpose transit trips and not to trips by purpose. have time to generate a trip matrix describing the HBW trip
purpose. For this reason, Tables 44 and 45 have been pro-
vided to permit a quick application of generalized factors by
Trip Matrix Conversion Factors urban area population to the work trips to produce total daily
trips or total trips for a designated time period.
It is conceivable the user may be required to perform an The tables cross-relate total travel and HBW travel by the
analysis within a period of time so brief that a full analysis of time periods most used for travel and air quality analysis' The
J
'rj
86
TABLE 41 (Continued)
Urban Size = 1,000,000+

Hour Beginning All Purposes
Midnight 0.35 0.32 0.34 0.34

1:00 a.m. o.21 0.19 0.2s 0.22
'r i
2:00 a.m. 0.36 0.26 0.32 0.31
3:00 a.m. 0.37 0.19 0.12 0.23
-.i'' ':- ' 4:00 a.m. 0.88 0.06 0.06 0.33
5:00 a.m. 2.94 0.24 0.07 1.08
.i
'.,.i
',,. 6:00 a.m. 7.90 1.08 0.31 3.10
. I
7:00 a.m. 14.06 4.79 1.05 6.63
'." i 8:00 a.m. 9.63 6.18 2.25 6.02
9:00 a.m. 4.30 4.88 3.32 4.'t7
10:00 a.m. 2.26 5.55 5.39 4.40
11:00 a.m. 1.86 5.61 7.47 4.98
Noon 2.92 6.06 11.37 6.78
'--"--'r
1:00 p.m, 2.68 5.72 8.92 5.77
2:00 p.m. 3.80 7.63 9.'t5 6.86
..'..:l 3:00 p.m. 6.78 9.10 9.51 8.46
4:00 p.m. 9.31 6.90 8.64 8.28
, 5:00 p.m. 12.04 7.37 9.01 9.47
6:00 p.m. 6.61 7.04 6.82 6.82
7:00 p.m. 3.26 6.92 5.61 5.26
l
''.'' 8:00 p.m. 2.20 5.38 3,89 3.82

-,:.:, 9:00 p.m. 1.91 4.25 3.04
I
3.07
10:00 p.m. 1.75 2.48 1.67 1.97
*--=-*j
11:00 p.m. 1.61 1.79 1.42 1.61
sourc: 199O NPTS.
[email protected]
-l
Percent of Vehicle Trips by Hour by Trip purpose

Urban Size = 1,000,000 +
g- 20
F
rs
o
Elo
o
t-
o

cv
Q'
g
o-0
Q'
Mldnight 04:( AM 08:00 AM Noo.t (X:( pM 08:00 pM

02:( AM 06:00 AM t0:( AM OA(x pM O6:(x) pM 1O:OO pM
Trip Start Time
+HBW +HBO *NHB
analyst should be aware of the definition of each time period . 8-Hour Peak Period-the 8-hour period during the
and of inherent subtleties contained in the tables. The defin- analysis day during which the greatestpercentage ofthe
itions of the time periods contained in Tables 44 and 45 are total daily trips occur.
as follows:
. Total Travel-the sum of travel for all purposes (i.e., As the user becomes familiar with the conversion tables,
HBW, HBO, and NHB) occurring during rhe 24-hour it will be discovered that the conversion factors are useful
analysis day, from the standpoint of gaining a perspective about the dis-
. Total Work Travel-the total travel for the HBW purpose tribution of travel as well as helpful for detailed analysis
estimated to occur during the 24-hour analysis day, and purposes.
87
Table 42 Diurnal distribution by purpose and direction
Home-Based Work Home-Based Other
Start From To From To Non-Home

Time Home Home Home Home Based Total
Midnight 0.000 0.004 0.000 0.002 0.003 0.003

1:00 a.m. 0.000 0.002 0.000 0.001 0.002 0.002
2:00 a.m. 0.000 0.003 0.000 0.002 0.002 0.003
3:00 a.m. 0.002 0.002 0.000 0.001 0.001 0.002
4:00 a.m. 0.009 0.001 0.001 0.000 0.001 0.003
5:00 a.m. 0.032 0.00'l 0.002 0.000 0.001 0.008
6:00 a.m. 0.093 0.002 0.010 0.001 0.004 0.026
7:00 a.m. 0.1 36 0.006 0.050 0.004 0.015 0.062
8:00 a.m. 0.076 0.004 0.048 0.008 0.026 0.054
9:00 a.m. 0.030 0.003 0.038 0.01 1 0.035 0.043
l v.vv d.ilt. 0.014 U.UU4 U-UJY U.UI/ U.U.' v.vlt
11:00 a.m. 0.010 0.006 0.029 o.024 0.078 0.052
Noon 0.011 0.013 0.029 0.032 0.110 0.066
1:00 p.m. 0.011 0.01 1 0.029 0.027 0.087 0.057
2:00 p.m. 0.014 o.021 0.026 0.045 0.087 0.069
3:00 p.m. o.o12 0.062 0.028 0.060 0.1 00 0.089
4:00 p.m. 0.01 1 0.092 0.029 0.040 0.093 0.082
5:00 p.m. 0.009 0.114 0.035 0.039 0.091 0.088
6:00 p.m. 0.008 0.057 0.o42 0.032 0.065 o.o71
7:00 p.m. 0.006 0.026 0.036 0.033 0.053 0.058
8:00 p.m. 0.004 0.017 0.016 0.036 0.037 0.043
9:00 p.m. 0.003 0.018 0.009 0.031 0.029 0.034
10:00 p.m. 0.005 0.015 0.004 0.018 0.015 0.021
11:00 p.m. 0.003 0.015 0.002 0.013 0.012 0.015
Total 0.500 0.500 0.500 0.500 1.000 1-OO0
Source: NPTS, 1990.
CREATION OF AN ORIGIN.DESTINATION indicate the location of the home end of the trip and attrac-
TRIP TABLE tions indicate the work, school, or shop end ofthe trip, but the
trip movements in a production-attraction trip table are not in
Before a traffic assignment can be performed, the trip the correct direction. For this example, it is assumed that this
tables must represent actual trips between TAZs (though is a daily trip table. The procedure for obtaining the origin-
assignment of trips in a production-attraction format can be destination trip table is to add one-halfofthe table to one-half
useful in identifying atypical directionalities). Productions of the transposed trip table. The equation for HBW trips is
TABLE 43 Conversion factors for critical periods of transit patronage
AnnualAverage Combined
Weekday Peak Period Peak-Hour Peak-Hour
Volumes (4-Hour) Volumes Volumes Peak-Direction
Annual Average 0.41 0.14 o.12
Weekday Volumes
Combined Peak-Period 2.40 0.33 0.28
(4-Hour) Volumes
Peak-Hour Volumes 7.37 3.07 0.88
Peak-Hour Peak- 8.35 3.47 1.13
Direction
.l
-i
88
TABLE 44 Conversion factors for critical periods of internal person travel
Combined
Peak
Total Daily Daily Work Total A.M. Total P.M. Period
Travel Travel Peak Hour Peak Hour (8-Hour)
Urbanized Area Populaton 50,000 to 199,000
Total Daily Travel 0.24 0.08 0.09 0.57
Daily Work Travel 4.08 0.33 0.37 2.35

Total A.M. Peak Hour 12.53 3.07 1.',14 7.20
Total P.M. Peak Hour 11.03 2.70 0.88 6.34

Combined Peak Period (8-Hour) 1.74 0.43 0.14 0.16
Urbanized Area Population 200,000 to 499,999
.:_+ I

Total A.M. Peak Hour 12.49 3.25 1.20 7.23
Total P.M. Peak Hour 10.43 2.7',! 0.84 6.04

_,,,:"--l
Combined Peak Period (8-Hour) 1.73 0.45 0.14 o.17
'i
Urbanlzed Area Populatlon 500,000 to 999'999

Combined Peak Period (8-Hour) 1.78 0.48 0.15 o.t6
Urbanized Area Populatlon 1,000,000+
Daily Work Travel 3.60 0.27 0.34 2.O2

89
TABLE 45 conversion factors for critical periods ofinternal auto driver travel
Combined
Peak
Total Daily Daily Work Total A.M. Total P.M Period
Travel Travel Peak Hour Peak Xour (8-Hour)
Ubanized Area Population 50,000 to l99,OO0
otal Daily Travel o.32 0.08 0.08 0.57

Daily Work Travel 3.'12 o.23 o.26 1.78
Total P.M. Peak Hour 12.10 3.88 0.91 6.9r
Combined Peak Period (8-Hour) 1.75 0.s6 0.13 0.14
Urbanized Area Population 200,000 to 499,999

Daily Work Travel 3.0s 0.25 0.31 1.75
Total A.M. Peak Hour 12.28 4.03 1.23 7.O4
Combined Peak Period (8-Hour) 1.74 o.57 0.14 0.18
Ubanized Area Population 500,000 to 999,999

Total A.M. Peak Hour 12.35 4.33 1.18 6.8s
Combined Peak Period (8-Hour) 1.80 0.63 0.15 0.17
Urbanized Area Population 1,000,000+

Daily Work Travel 2.90 o.21 o.28 1.60
Total P.M. Peak Hour 10.38 3.s8 0.7s 5.73
HBW, = 0.5x HBWt,n +0.5x HBW;, (8-1) HBW;": transposed HBW trip table, production-attrac-
tion format.
where
HBW", : HBW trip table in origin-destination format, Table 46 shows the production-attraction trip table before
HBW,": HBW trip table in production-attraction for- application of the equation.
mat, and Table 47 is the resulting origin-destination trip table.
90
TABLE 46 Example production-attraction table
Attraction Zone
1 50 30 20 100
2 100 70 30 200
3 250 200 50 500
'II otal 400 300
TABLE 47 Example origin-destination table
Destination Zone
Origin Zone Total
1 50 65 135 250
2 65 70 115 250
I
3 135 115 50 300
*,-:1
i
Total 250 250 300 800

rqF
This example is for the conversion of a 24-hour trip table' AMPK(O-D)

To prepare the trip table for a time of day (peak hour or = HBW-FAC (From Home)x HBW-TR4:
peak period), assignment factors presented previously can + HBW - FAC (To Home) x HBW -TRPi
be applied using the same basic equation as shown above' + HBO-FAC (From Home)x HBO-TRP
The major difference is the different factors used for the + HBO-FAC (To Home)x HBO'TRPi
production-to-attraction direction versus the attraction-to- + NHB-FACx NHB-TRPi
production direction (the transpose of the trip table).
The conversion equation needs to be applied only to the where
HBV/ and HBO trip tables. The NHB production-attraction
trip table is by definition the same as the origin-destination AMPK (O-D),: AM peak-hour trips from origin i to des-
table and therefore the daily table does not have to be con- tination j,
verted to origin-destination format. However, if a peak- HBW-FAC: time-of-day/direction split factor for the
period NHB is to be produced, then the time-of-day factors selected time period in the direction
must be applied. from home to work,
The diurnal factors in Table 42 can be applied to convert HBW-TRPij: home-based work trips in production-
the trip tables by purpose in production-attraction format to attraction format from production i to
an origin-destination trip table for a selected time period' attractionj, and
Note that the non-home-based trip table is already in origin- HBW-TRPj1 : Transpose matrix of HBW-TRP .
destination format as the trips have no home end'
The AM peak-hour trip table from 7:00 AM to 8:00 AM
can be created as follows: If the daily HBW trip table looks like the following:
j=1 j=2 l= 3 Total
-- 1 10 40 20 70
i=2 15 25 35 75
30 50 5 8s
Total 55 115 60 230

9t
and the daily HBO trip table looks like the following:
j= 1 j=2 j=s
,= l 80 70 50 200
i=2 75 85 55 215
l= 3 60 45 35 140
Total 215 200 140 555
and the daily NHB trip table looks like the following:
l=1 j=2 i=3 Total

i= 1 20 10 35 65
i=2 15 25 45 85
t=o 5 30 f5 50
Total 40 65 95 200
then the O-D trip table is created as follows:
AMPK (O-D) =
j=1 2 3
(0.136 x 10 + 0.006 x 10) + (0.136 x 40 + 0.006 x 15) + (0.136 x 20 + 0.006 x 30) +
(0.050 x 80 + 0.004 x 80) + (0.050 x 70 + 0.004 x 75) + (0.050 x 50 + 0.004 x 60) +
(0.015 x 20) (0.015 x 10) (0.015 x 35)
(0.136 x 15 + 0.006 x 40) + (0.136 x 25 + 0.006 x 25) + (0.136x35+0,006x50) +
(0.050 x 75 + 0.004 x 70) + (0.050 x 85 + 0.004 x 85) + (0.050x55+0.004x45) +
(0.015 x 1s) (0.01s x 25) (0.015 x 45)
(0,136 x 30 + 0.006 x 20) + (0.136xs0+0.006x35) + (0.136 x 5 + 0.006 x 5) + (0.050

(0.050 x 60 + 0.004 x 50) + (0.050x4s+0.004xs5) + x 35 + 0.004 x 35) +
3
(0.01s x 5) (0.015 x 30) (0.015 x 1s)
j=1 j=2 j=3 Total

i=1 6.04 9,48 6.165 21.69
i=2 6.535 8.515 8.665 23.72
i=3 7.475 9.93 2.825 20.23
Total 20.05 27.93 17.66 65.64
With an origin-destination table, vehicle trips made during interchanges did not reflect the true direction of the trips from
the AM peak hour can be assigned to the highway network. one zone to another. Before the desired O-D matrix was for
The table shown above lists fractional trips for all of the whole day, the split from home to activity was assumed
theinterchanges.Althoughfractionaltripsdonotexistinreal-toequaloutoVeranentiredayforHBWandHBotrips.
ity, many of the procedures used in the travel model process The conversion from production-attraction format to origin-
produce noninteger values. Rounding of the trips tables destination format used the following equation:
before assignment could result in a loss of accuracy. A better
approach is to carry the fractional trips through assignment,
Daily Vehicle Trips (O_D) =
but round the assigned volumes on individual links. 0.5 x HBWro O.S x HBW,, + 0.5 x HBO*A
+ 0.5 x HBOA. + NHB + ThroughTrips :
CASE STUDY
where HBWp is the transpose of HBWpo. The non-home-
The vehicle trip tables built in Chapter 7 were still in based and through trips are not factored because they are
production-attraction (P-A) format. Therefore, the trip already balanced in origin-destination format.
92
The time-of-day characteristics presented in this chapter tions is minimal. As a result, assigning the trips by time
provide the ability to factor a daily trip table to create peak- period would not necessarily produce a better assignment.
period and off-peak period trip tables. Assigning traffic by For the purposes of this case study, the daily trip table was
time-of-day considers the relative levels of congestion and converted to a peak-hour trip table by factoring the entire trip
the alternate optimal travel paths between zone pairs that table by a 10 percent factor. The trip table assigned to the net-
vary by time period. By adding traffic volumes from each of work was calculated by applying the 10 percent factor to all
the time periods together, an estimate of the daily volume on trips such that
a link is produced.
The results of trip distribution for Asheville have indicated Hourly Vehicle Trips (O-D) =
that the difference between free-flow and congested condi- 0.10 x Daily Vehicle Trips (O-D)
'|'' ':
---.-.1
1
''.i
'' ''. -'l
j
.1
93
CHAPTER 9
TRAFFIC ASSIGNMENT PROCEDURES
INTRODUCTION In addition to the more detailed coding of the link and zone
system, standard procedures have come to include assign-
Assignment is the fourth and last major step of the tradi- ment by time of day. The outmoded procedure for obtaining
tional four-step process. This includes both highway and aZ4-hour assiqnment was either to factor the dailv trin table
transit assignment of vehicle and person trips, respectively. by 10 percent to obtain a peak-hour trip table or to multiply
The assignment of trips to the network is the final output of the hourly capacity by 10 to get daily capacities. These were
the modeling process and becomes the basis for validating then used as input to the capacity-constrained assignment
the model set's ability to replicate observed travel in the base algorithm. For example, the UTPS software package
year as well as to evaluate transportation improvements in included a parameter called "CONFAC" that was defined as
future years. Depending on the level of analysis being done, the percentage of the daily trip table that represented peak
the assignment can be to a regional highway and transit sys- hour. This factoring procedure was used to produce a 24-
tem for systemwide planning or to a detailed network for a hour capacity-constrained assignment. In fact, there is no
subarea or corridor study. such thing as daily capacity, but this was a close approxima-
The level of precision of the assignment procedure is a tion of daily assignments that reflected congestion on the
function of the detail of the coding on the networks and the highway networks. The more current and accepted procedure
size of the associated zone system. Traditionally, highway for obtaining daily highway volumes is to sum the results of
and transit assignment procedures were used primarily for three separate assignments: AM peak period, PM peak
systems analysis of large-scale transportation improvements. period, and off-peak.
In recent years, the necessity for peak-hour or peak-period The capacity-constrained assignment procedures were all
forecasts of vehicle demand on the highway system has designed to produce assigned traffic that approximated the
required refinement of the traffic assignment procedures and equilibrium of congested travel paths in the network. This was
parameters. Associated with the better assignment algo- done through some combination of incremental and iterative
rithms and parameters, the level of detail in the highway net- assignment of the vehicle trip tables to the network. Now
work increased dramatically (both in the amount of highway most of the available software packages include the equilib-
system actually coded into the network and more specific rium assignment algorithm in which iterations of assignments
definition of link attributes such as link capacity and inter- are made until the available travel paths all have the same
section delay). travel times. The state of equilibrium is now computed math-
Historically, only lane capacities by facility type were ematically rather than being treated as a goal by simple capac-
coded on the network and a single volume-delay function, ity constraint. The same volume-delay function can be used in
the Bureau of Public Roads (BPR) curve, was used to esti- equilibrium assignment as was used in the previous capacity-
mate link travel times resulting from the assigned volumes. constraint procedures. Equilibrium assignment is the recom-
This chapter will present recommended modifications to the mended procedure to be used in highway assignment.
coefficients for the basic structure of the BPR curve that are The input for highway and transit assignments include the
consistent with the 1985 Highway Capacity Manual. As with coded networks and the trip tables produced by the mode-
other techniques in this report, it is assumed that the modeler choice model. If the mode-choice model produces person
will use one of the popular travel-demand forecasting soft- trips instead of vehicle trips, auto occupancy rates can be
ware packages that can implement highway assignment from used to convert the person trips to vehicle trips (Chapter 7).
a coded network. Also consistent with the needs of small- to Time-of-day factors can be used to convert the daily trip
medium-sized urban areas, the focus will be on the highway tables to peak period, peak direction (Chapter 8).
assignment procedures. Most of the available software pack- This chapter contains the following sections:
ages have transit assignment procedures. However, the tran-
sif procedures are not usually capacity constrained, and . Parameters for traffic assignment models,
therefore the key to transit assignment is the assignment of . A post-assignment traffic-smoothing technique, and
transit trips to the proper path and modes of access. . A corridor traffic diversion/traffic shift technique.
94
The last two techniques are extracted directly from The basic BPR formula used the values of 0.15 and 4.0 for
NCHRP Report 187 and can be applied using any of the cr and B, respectively. This formula continues to be used in
available spreadsheet programs. The smoothing technique is many urban areas. In Horowitz's work for the FHWA, coef-
a post-assignment procedure that can be used to obtain more ficients were calibrated for the BPR formulation that better
precise link-specific volumes from a regional travel-demand replicated delay as computed using the 1985 Highway
forecast. The corridor traffic diversion technique can be used Capacity Manual procedures. In that work the BPR coeffi-
to evaluate quickly, on a sketch level, the impact of major cients presented in Table 48 were developed.
capacity changes to a facility within a defined corridor on These functions are depicted in Figure 21. The speeds
both the facility itself and on competing roadways within the shown in the above table are design speeds ofthe facility, not
corridor. the free-flow speeds. Capacities used in the vlc ratio are ulti-
mate capacity, not a design capacity as used in the standad
BPR curve. The curves based on the HCM exhibit a speed of
BASIS FOR DEVELOPMENT about 35 mph at a vlc ratio of 1.0. This is consistent with stan-
dard capacity rules that the denser traffic flows occur at this
The source for the parameters for the volume-delay rela- speed. The ultimate capacity used for these curves was 1,800
tionships used in the traffic assignment algorithms is Alan vehicles per hour, per lane for a 1-mile section. This value is
Horowitz's report for the FHWA, Delay-Volume Relations the ultimate capacity for typical prevailing conditions, not
forTravel Forecasting, based on the 1985 Hghway Capac- those under ideal conditions, which would have a capacity of
ity ManuaLt The parameters were derived using the basic 2,000 vehicles per hour, per lane.
BPR formulation of volume-delay. The parameters were fit For each curve the BPR standard curve with coefficients
to the speed/volume relationships contained in the Highway of a : 0.15 and I : 4.0 is plotted to illustrate the change
Capacity Software, Version 1.5, which closely approximate that the HCM curves represent. The BPR curve has a much
those in the HCM. The coefficient, a, of the BPR function higher speed atavlc equal to 1.0 than do the HCM curves. It
was determined by forcing the curve to fit the speed/volume can also be observed that the multi-lane curves have a steeper
data at zero volumes (free-flow speed) and at capacity (level decline with the vlc < I.0 than do the freeway sections for the
of service ILOSI E). The second coefficient, B, was found by same design speed. Another characteristic of both the BPR
nonlinear regression. and HCM curves is that they extend beyond the point where
The traffic-smoothing and corridor traffic diversion/ traffic thevlc ratio is equal to 1.0, or where the flow has reached
shift techniques are the same as contained in the original capacity. In capacity analysis, this portion of the curve is
NCHRP Report 187.Both of these techniques continue to be considered to be unstable and curves in the 1985 Highway
useful post-assignment analysis and sketch-planning tools. Capacity Manual end at this point. For travel-demand mod-
No changes to either technique were required. eling, however, the curve must extend beyond 1.0 to account
for the theoretical assignment of the traffic.
TBAFFIC ASSIGNMENT MODEL PARAMETERS Application of Volume-Delay Curves

in Highway Assignment
The traffic assignment process is driven by the relation-
ship of assigned volume and the resulting delay caused by The many highway assignment software packages all vary
congestion. As traffic volumes increase, travel speeds in how the volume-delay function is determined. Many sim-
decrease because of increased congestion. The following ply default to the standard BPR formulation and require spe-
BPR formulation has been used to estimate link travel times cial input to vary the curve. They may or may not allow for
as a function of the volume-to-capacity ratio multiple curves to be used for different facility types. Some
require the curves to be input as data points in the form of a
(e-l) look-up table. The user will have to refer to the appropriate
T,=Tf
"('*""[:]') documentation of how many curves can be used in the soft-
where ware and how these are input.
Three possible levels of volume-delay formulations could
[ : congested link travel time, be applied to highway assignment algorithms. These are
4: link free-flow travel time,
v: assigned link traffic volume (vehicles), . A single formula used for all facility types,
c: link capacity, and . Multiple formulas that vary by facility type, and
o,9 : volume/delay coefficients. . Multiple formulas that vary by facility type combined
with estimation of delay at controlled intersections.
lAlan J. Horowitz, Delay-Volutne RelationsforTravel Forecasting, based on the 1985
The first level is the most rypical and the second is becom-
Highway Capacity Manual, prepared for the Federal Highway Administration, U.S.
Department of Transportation, Washington, D.C. (1991). ing more common. The third level is based on the recognition
95
TABLE 48 BPR coefficients
Freeways Multi-lane
Coefficient 70 mph 60 mph 50 mph 70 mph 60 mph 50 mph
0.88 0.83 0.56 1.00 0.83 0.71

9.8 5.5 3.6 5.4 2.7 2.1
that the major source ofdelay on urban streets is at the con- delay at a signalized intersection. Inputs to the model (for
trolled intersection. This control includes both traffic signals each intersection) are the cycle length, green time, saturation
and stop signs. flow rates, and arrival type. Arrival type is a general catego-
The volLrme-delay rela-tionship for interrr:pted flow (inter- rization of the quality of progression on the approach and
rupted by a traffic control device) comprises includes dense platoons arriving at the beginning or middle
of the red./green phase, and totally random arrivals. For most
. The delay on the link using the above formulas, and assignments of long-range future trip tables, standard green
. The delay at the intersection caused by the probability times can be developed as a function of the facility types of
of being stopped and the time stopped at the control the intersecting links. A uniform cycle length of90 could be
device. used, and it could be assumed that the arrival type is random.
Even under these average circumstances, if the network is
The 1985 Highway Capacity Manual includes procedures composed of freeways, expressways, major and minor arte-
for estimating delay for each approach to the intersection. As rials, and collectors (tve facility types), there would have to
with the unintemrpted highways described earlier, the delay be 25 different volume-delay functions for each combination
function is well behaved up to a vlc ratio of 1.0, and then it of intersecting links. Only a few of the currently available
becomes unstable. In Horowitz's report, coefficients were travel-demand software packages can process this many dif-
developed for the BPR formulation that fit a curve to the ferent functions-
TRAVEL SPEEDS
llrtrScoldr
E
E

6
0.5c 0.97 0.7t 0.s0 r.00 f.ff lz l.3g
v/o Rrto
O OtTtBPROv! + HCH-HPH O HOH.GO HPI{ HCr-HPft
Figure 21. Multi-lane volume delay.

96
Recognizing the complexity of incorporating intersection Screenlines (checks trip distribution as well as assign-
delay into the volume-delay function, it is recommended that ment),
this level of detail be considered only if the local urban area Cordon lines (CBD, for example, checks both trip gen-
has both the software and the experience to implement this eration and trip distribution),
function in the travel-demand software. Software packages Cutlines for major corridors (checks assignment func-
that incorporate intersection delay into the assignment algo- tions and link attributes),
rithm contain specific applications of those procedures that Link-specific volumes, and
should be followed rather than nationally developed default Regional statistics (such as root mean square of error
functions. Inclusion of intersection delay in the assignment IRMSI that produces statistics on assigned versus ob-
models is emerging as the most significant improvement in served traffic by facility type and volume groups).
the algorithm.
A small urban area that is using the default parameters dis- Once the cordon lines and screenlines are validated and the
cussed and wishes only to use a single volume-delay func- trip distribution model is judged to be producing acceptable
tion, should use the following coefficients: results, the assignment volume-delay functions can be mod-
;
ified systematically to produce the desired assignments. It
-l
I
: 0.84
o has been the practice in some urban areas to adjust individ-
ual link attributes to get an assignment that matches the link
:5.5 counts. In many cases these adjustments have produced unre-
alistic values of link speeds and capacities (free-flow speeds
An urban area that wishes to use a unique function for each
of 5 mph, for example) that worked only to get the desired
facility type can use a combination of the coefficients for
assignment results. The adjustment of link attributes should
freeways and multi-lane highways. Judgments will have to
-r be limited to minor systematic adjustments to speeds and
i
be made relating major/minor arterials and collectors to the

capacities for groups of links that have the same facility and
i
multi-lane highway coefficients. The design speed can be
area type.
used as a surrogate for facility type. Most freeways or inter-
state highways are constructed for a 70-mph design speed,
and that should be the value used for freeways unless a free-
DISTRIBUTION OF ASSIGNED VOLUMES
I
way is older and was built to lower design standards. Many AMONG AVALABLE FACILITIES
urban major arterials are multi-lane, and one of the multi-
Iane sets of coefficients can be chosen. The selection of the In any assignment of travel to a highway network, whether
volume-delay functions is not absolute, and the final set of by manual methods or through the use of a computerized
functions chosen will be those that best produce highway technique, the link-assigned volumes may require some
assignments that reflect observed traffic volumes. For exam- redistribution between available facilities to more closely
ple, if the major arterial system has consistently high vol- reflect actual operating conditions. Historically, transporta-
umes, then the next lower set of coefficients may be used. tion planning procedures have used screenlines and auxiliary
This will "slow" down the congested speeds and shift traffic cutlines to validate atd analyze assignment results, and the
to other facility types. redistribution technique described follows the same approach.
The technique described to reallocate travel among com-
peting facilities after traffic assignment is based on screen-
Node Characteristics line theory and was developed by R. H. Pratt Associates.2
This technique requires analysis of multiple overlapping cut-
Some modeling software accommodates the input of node lines of major screenlines within an analysis area. It may
characteristics. Specifically, the type of intersection control appear to the user that the procedure is difficult and time-
device and cycle characteristics may be input. consuming, but it will be found that an analysis area con-
taining 10 vertical and 10 horizontal major screenlines can be
processed and summarized in 2 person-days. The analysis for
Validation of Highway Assignment and Network most areas will not be as extensive.
The underlying assumption of the redistribution proce-
Although this manual is not intended how-to report for
as a
dure is that forecast-year volumes on parallel facilities
calibrating and validating the highway networks and the
should tend to be distributed proportionally to the volumes
associated assignment, some basic steps in the process are
as observed on the facilities in the base year. Further stated,
helpful in applying the volume-delay functions. The valida-
if no capacity changes (e.g., widenings and new facilities)
tion of the highway assignment is the final validation of the
complete travel-demand model set. The check of assigned
volumes with observed traffic counts is done at the follow- Trafic
']R.H. Pratt Assoc- A Method for Distributitlg Volumes Among Competing
ing levels: Fac il it i e s, Kensington, Maryland ( I 976).
97
occur between the year observations are made and the fore- facilities (Figure 23). The second cutline should do the same,
cast year, the forecast-year volumes on the links intercepted and overlap the first cutline such that the overlap extends
by the screenline are inclined to be proportional to the base- across approximately half of each individual cutline. Prefer-
year system. All capacity changes to the forecast year system ably, more than one facility should be intercepted within the
are interpreted as new facilities, including widening to exist- overlap. The third cutline should be similarly laid out and
ing facilities. should start where the first cutline terminates. Additional cut-
Figure 22 shows the beginning point for applying the Iines as needed should be similarly established. Unless irreg-
volume-redistribution technique. It is assumed the user will ulaities in the street system dictate otherwise, the cutlines in
employ these techniques after the appropriate vehicle trips parallel screenlines should be opposite each other so as to
have been assigned to the highway network via the all-or- intercept the same sets of highway facilities.
nothing assignment procedure. The major screenlines to be As an example, Figure 23 shows the subdivision of Screen-
used in balancing the trips between competing and available line A-A into three overlapping cutlines (i.e., p-p, q-q, and
facilities are shown, along with the facilities under study. The r-r) to be used in the redistribution of forecast-yeal assign-
following points should be kept under consideration while ment volumes. Screenline A-A will be analyzed using the
constructing the analysis lines. Screenlines need be Cefined hypothetical traffic data given in Table 49. Note that the
only across facilities within the directional analysis area. forecast-year assignment volumes are supposed to have been
That is, if only north/south highways are under investigation, obtained from all-or-nothing assignment procedures. Note
only screenlines A-4, B-8, C-C, and D-D would be required. also that link 50-5 1 is a proposed facility for the forecast year
Major screenlines should be constructed midway between and is expected to add capacity across screenline A-4.
major intersections or every 2 miles, whichever is less' The work sheet used for redistribution of assigned vol-
Except in special cases, screenlines should cut a minimum of umes is given in Table 40. Link description, plus traffic data
three facilities. for columns a, c, and e are filled in Table 40 using the data
The manner in which each screenline is subdivided into given in Table 49. Such information is recorded for each
cutlines is as follows. Starting at one end of the screenline, of the three cutlines of Screenline A-A shown on Figure 23.
the frrst cutline should normally extend across at least three The cutlines are processed one at a time and the total
I
LEGEND:
Network
UnUNode
hlghway fadllty
Screenlinc raVsls tuea boundary
Figure 22. Definition of maior screenlines.

..1 98
?,ed haht4u
ft n\ "# te a*, yeo.n
c
-i scrUne A-.
':
cutlme p -
-F
7
fr,nay
Figure 23. Cutlines of screenline A-Afor redistribution analysis.
assignment-adjustment volumes (column h, Table 40) are counts do not exist and therefore columns a and b are
input, when appropriate, into column e of the subsequent left blank.
cutline analysis. The order in which the cutlines are processed 2. Because link 50-51 contributes additional capacity in
is arbitrary, but such computations should proceed in an the forecast year, columns c and d are filled in a man-
orderly fashion from one end of the screenline to the other ner similar to Step 1.
(e.g., from left to right). 3. Column e is now completed using the forecast-year
The calculations necessaly for completing Table 50 are as assignment volumes in Table 49 (from the all-or-nothing
follows: assignment).
4. As a capacity change is expected to occur across cut-
1. Sum the base-year volumes; that is, traffic counts (col- line p-p, column f is completed for link 50-5i. Thus,
umn a), and determine the percent volume contribution the capacity-assignment adjustment for link 50-51 is
(column b) for each link of cutline p-p. Note that 23.|Vo X 12,200 : 2,818 (i.e., this volume of traffic
because link 50-51 is a new facility, base-year traffic can be expected for the new facility). The remaining
TABLE 49 Traffic data for highway links crossing screenline A-A
Link Descriptor
Forecast Year
Base Year Assignment
A Node B Node Traffic Count Capacity Volume
1 15 1,850 2,200
50 51 3,000 4,000
2 13 5,000 7,800 8,200
4 I 2,500 2,750 2,500
5 7 2,650 3,500 4,800
Total 12,000 19,250 19,500

99
TABLE 50 Worksheet for balancing forecast-year assignment volumes

(a) (b) (c) (d) (e) (r) (s) (h)
Volume Total
Link Descriptor Percnt o/" ol Forecast Year Capacity Assgnmnt Assignment
Base Year Base Yar TotalCapacity Assignmont Assignment Adjustmsnt Adjustment
Cutlinef ANode BNode Volume Volume on Cutlne Capacity on Cutline Volume Adjustment (s)=(b)x(E(e)-E(e)) (h)=(f)+(g)
p-p f 15 r'r1 27.O 2,200 16.9 0 2,s33 2,533 2
50 51 3,000 23.1 4,000 2,818 2,818

tJ 5,000 73.O 7,800 60.0 8,200 6.849 6,849
2,8'18
q-q 50 51 3,000 22.1 2,818 t,ttt 2,689 '?
213 5,000 66.7 7,800 57.6 6,849 6,322 6.322

4A 2,500 33.3 2,750 20.3 2,s00 3,1 56 3,156
100.0 12.167 12,167
r-r 2 13 5,000 49.3 no nw facilitis proposd 6,322 se note in 7,039 7,O39 2
4 I 2,s00 24.6 across this cutln: thus 3,156 columns (c) 3,512 3,512 2
5 7 2,6s0 26.1 calculalions not ncessary 4,800 3,727 3,727

2
for these columns and (d)
10,150
t All traffic data are two-drectional and measured in vhicles per hour
2
Final, balanced volumes as a rosult of tratfc redistrbution.
forecast-year assignment volume in column f (i.e., ments in column g are now derived by proportioning the sum
12,200 - 2,818 : 9,382) is distributed to the other of traffic in column e using the percentages in column b.
links of cutline p-p. Thus, for link 4-8, the proportioned traffic equals 24.6Vo X
5. Hence the volume-assignment adjustments (column g) 14,218 : 3,512. The asterisks in column h of Table 50 indi-
for links l-15 and 2-13 can be computed in the propor- cate the final balanced volumes resulting from the redistrib-
tion given in column b. Hence, for the former link, this ution technique. To refine these volumes, Screenline A-A
adjustment is 27 .}Vo x 9,382: 2,533 (.e.,this volume could be reprocessed through the six steps outlined previ-
oftraffic can be expected for link 1-15). ously. This second iteration might result in a small gain
6. Finally, the total-assignment adjustment for each link in accuracy of the balanced volumes; iterations beyond the
crossing cutline p-p is computed by adding the volumes second one are not recommended. Figure 24 shows the
in columns f and g. Note that the totals for columns e capacity, the base-year volumes, the forecast-year assign-
and h are the same for cutline p-p; only the traffic ment volumes, and the balanced volumes for links crossing
within the cutline has been redistributed among the Screenline A-4. The user can observe the effect of the re-
three links. distribution of volumes among the facilities.
The user is cautioned that this technique does not keep
The six steps are repeated for cutlines q-q and r-r. For q-q, track of turning movement volumes and does, in fact, negate
the volumes for links 50-51 and 2-13 in column e are the the turning movement volumes from the all-or-nothing
assignment adjustments from column h of the previous cal- assignment procedure. Reestablishing a table of turning
culations for cutline p-p. A similar transformation is made movements is possible, but the redistribution procedure
for r-r (the volumes for links 2-13 and 4-8 in column e a-re requires many iterations to reach convergence and is not
adjustments from column h for cutline q-q). For r-r, however, practical as a manual tool. If the user is interested in analyz-
no new facilities cross the cutline. Therefore, the computa- ing turning movements, the trips from the original assign-
tions in columns c, d, and f are not necessary. The adjustment application should be used.
. :.i 100
,l
l .' .i
1
LEGEND: where
"'-
,.'i
.- 22/oo
tt"l-t""e V,,,,,. : volume on minimum time route,
:. ; -Ceoecibt
yearvolume 0 : diversion parameter,
33: ror*".t year assgnmert volume t,,, : time on alternate route fNote: (t,, - t,) is always
't ,,
'+.
\ B"l"rt"d after redsibution negativel,
"olume Vr : total volume on two facilities (Vr: V.,, f Vt), and
Rd h'trat,| V : volume on alternate route.
eury hfoeeeast yr
The process assumes that current traffic volumes and oper-
ating characteristics for a base condition are known. These
data may be obtained from traffic volume counts and
speed/delay travel-time studies or from the results of the
I planning process. To describe application ofthe process, the

following simplifled example is offered. The example of two
competing facilities is shown at the top of Figure 25.
l,* To calculate the diversion parameter (based on existing

conditions and volumes), the function given in Equation 9-4
rc would be used as follows:
r.240
ln
---
^ 7.500 -l .8
e----1r"::- =0.36,1
.-.1
\ 1,1 - t2.0 -4.9
This parameter describes the diversion of traffic between

\ {c the two routes being considered. Assume an improvement
is to be made in route A by adding another lane in each
tq^Jarl
direction. A speed of 50 mph for the improved facility is
''': l estimated based on a capacity calculation using the original
Figure 24. Comparison of capaciry and base year, volume of 7,500 vehicles. The v/c ratio would be developed
forecast yer, and bqlanced volumes. from
. L-^1< 1,500 vph

TRAFFIC SHIFT METHODOLOGY 5 lanes x2,000 vph capacity
FOR CORRIDORS
The travel time for the 5-mile route A section would then be
For corridor analysis, often a sketch-planning technique is
calculated as:
desirable to evaluate the effects of an improvement in one
of the facilities in the corridor. Such a process, if it is to 5 mi 60 min
.-X-=O.Vrlull
be applied quickly, should not consider origin-destination 50milhr hr
movements but rather a general shift of traffic between facil-
ities. Such a process has been suggested by the multi-route Based on this improvement, a new estimate of the average
probabilistic process developed by Dial.3 The required equa- volumes can be calculated using Equations 9-2 and 9-3 as
tions for the usual two competing-route problem areas are follows:
v,,, = v^,, x (7,500 + 1,240) = 7,869 vph

t* r#4^(vr) (e-2)
|+ e0.361(6.o-r2.0)
v,=#ffi,u,) vi- e0.36'7(6.O-12.0)
(e-3) rrre, -0.367(6.0-12.0)

(8,740) = Sllvph
.v, The v/c ratio for route A would now be 11,869 + (5 X

, ='n'; (e-4) 2,000)1, or 0.19, resulting in a speed of about 48 mph as cal-
tu, - t culated from capacity curves. Route B would carry about 871
vehicles per hour. Another iteration of the process could be
rDial, R.B. A Probabiltstic MultipathTraffic Assignment Motlel which Obviats Ettu. carried out to try to effect a closer relationship between
nrcrdtiott, prepared for the US DOT, \r'ashington, DC (1970). volume and speed, but for sketch-planning purposes and
t01
l'*t. V.7,soov
Conidor
rridltF
2 miles
ti - 12,0 mlT. S - 25 mptr.
Coffidor _
- vE rrir^
rrr
Lgngt
n-7.1 mins
S'42 mPh.
Gonidor
widtlF
2 miles
r..tr,ortnr. nl3r'
t.14.0 rrins. !'21 mPh.
tr'' ' 5 Mir
Figure 25. Example corridors for trffic shifi analyses.
because of the inaccuracies of volume/capacity/speed rela- To calculate the effects of the improvement in route A to
tionships, especially for arterials, it is not expected to im- 50 mph, the volume on routes A and B would be calculated
prove the results. Only if a large imbalance exists in result- at7,869 and 871, respectively, as shown previously. To cal-
ing volumes and speeds should additional iterations be tried culate the effect relative to routes B and C, the following
(i.e., greater than 5 mph difference between input and calcu- computations would be made:
lated speeds).
To demonstrate how the foregoing process handles varia- l/v,u, = x (871 + 800) = L0I6 vph
tions in operating conditions, Table 51 is provided for the ll7l,ltr*
example case described previously. oO.219\12-14)
tfv (1,671)
Where three competing facilities exist in a corridor, the -- ;--:-fJi:r
t+e ^ = 655 vph
process must be applied twice, with the calculation of two 0
values. Assume routes A, B, and C as shown in the lower half These calculations may be iterated a few times to bring the
of Figure 25. As shown previously, the 0 for route pair A and results to a more stable condition. For example, now consid-
B would be calculated as shown previously and equals 0.367. ering the volumes of 7,869,1,016, and 655, a new calcula-
Using Equation9-4, the 0 value for route pair B and C would tion between routes A and B would result in volumes of
be calculated as follows: 8,000 and 885 for A and B, respectively. The three-route case
for a corridor is unusual; generally, only two competing
r- 800
"' 1.240 routes will be handled.
^ -0.438 Capacity analysis should be considered as part of this
12 - 14 -2.0 traffic-shift analysis. After the process is applied, volume/
t02
TABLE 51 Variations in traffic volumes with changes in speed
Base Conditions: Five-Mile Section

7" Route A Volume 86%
Route A Speed = 42 mph; Route B Speed = 25 mph
Calculated 0= 0.367
Speed Difference in 7" Volume
Travel Time = on
Route A Route B Route B-Route A Route B
30 25 -2.0 67
35 25 -3.4 7A
40 25 -4.5 84
45 25 -5.3 87
50 25 -6.0 90
55 25 -6.5 83
30 35 +1 .4 37
35 35 0.0 50
40 35 -1 .1 60
45 35 -1.9 67
50 35 -2.6 72
55 35 -3.1 76
30 45 +3.3 23
35 45 +1 .9 33
40 45 +0.8 43
45 45 0.0 50
50 45 -0.7 56
-,-.-..'l 55 45 -1.2 61
:,1
capacity/speed calculations should be performed to deter- Usually, the diversion pa-rameter is frrst determined for a cor-
mine if the resulting speed is in balance with the speed used ridor, given travel volumes and travel times on the two routes.
in the preceding described process. If not, the new speed Then to study the effects of a travel time change on any one
should be used to redo the calculations. route, 0 is held constant and the new volumes are determined.
Usually, a number of sections will exist along each facil-
ity in a corridor in which volumes and speeds may vary. The To illustrate the use of the graph shown in Figure 26, consider
approximate speeds and section distances should be used to the example illustrated in the upper portion of Figure 25. In the
calculate section times and added to obtain the total time condition shown, the user knows the following variables:
through the corridor. An average volume should be used
based on the calculation v .., =V,,,,v""''
+
x roo
V,
2 (Volume in Sectionx Section Length) 7,500
Average Volume =
2 Section Lengths = 7,500 + 1,240 x 100 = 85.87o
Traffic shifts also can be determined graphically by using

a simple set of curves as shown on Figure 26. To use the therefore
graph, the user has to know at least two variables:
Vo V, = 100 - 85.8 = I4.2Vo
1. If the diversion parameter, 0, for routes within a corri-

Lt = t,n - t = 7.1 - 12 = -4.9 min
dor is to be determined, then the user must input the
percent volume on the minimum time route; that is, By entering the curves in Figure 26 atV,n,,: 86 percent and
V,,,,,, and the travel time difference, Lt, between the
At: -4.9 min,0 is interpolated ar0.37, which checks with
that calculated mathematically in the example described ear-
faster and slower routes; that is, t^ - ti.
lier.
2. If the percent volume on any route is to be determined,
then the user must input the diversion parameter 0 and
the travel time difference Ar. Note that in all cases, the Now suppose, as before, route A is improved so that the
following relationships hold: travel time on this route is reduced to 6 min from the origi-
r'al7 .l min. Thus,
Vo V,,n + V = l00Vo and
Lt = t,, - /, ( 0 (alwaysnegative) At=6-L2=-6min
103
o80
5
o o
.
o o
E .
F o

5eo
.E
45
.c
c
o
=co o
o
E :E
=4
o 9o
o E

co p
Io o
o-
o-?
0
-, -1S -18 17 -16 -15,-1,{ -13 -12 -ll -10 -7 { -S 1 3 -Z -l
Tavel Tme Dierence between Rout
Figure 26. Graphfor determining traffic shifis betweenfacilities in a corridor.
Entering the graph in Figure 26 at Lt: -6 min and 0 : 0.37, 48. Freeway links used values of 0.83 for o and 5.5 for B that
the corresponding %oV,,,,,. is read off at 90 percent and VoV at correspond to a design speed of 60 miles per hour. Arterial
10 percent. Because the total volume entering the corridor is links used values of 0.71 for a and 2.1 for B that correspond
known to be 8,740 vph, then to a design speed of 50 miles per hour on multi-lane roads.
Congested travel time is calculated using the following
s0
V.,, = l*Ox 8J40 = 1,866 vph formulas:
Freeway Travel Time =

V=+xBJ40=874vph
' 100 ((Eogl)x votume
ool l, * o.*, *( * ,,,,.,"1
\\speed) )" ( \capacity ) )
These results check with the values of V,,,,,and 14 obtained in
the previous example.
Arterial Travel Time =
Thus, the traffic diversion method is accurate enough for (( ItoEl)* ool * ( , * o.r, ,( ,otume
, ,o,".,1''l
sketch-planning and quite simple to use, and it is recom- \\speed) ) \, \capacity ) )
mended if time is not available to complete modeled alterna-
tives analysis. The equilibrium traffic assignment produced traffic volumes
for each link in the network. Volumes were factored by a
value of 10 to reflect total daily conditions. Traffic volumes
CASE STUDY have been summarized at a number of screenlines. A com-
parison of the estimated and observed daily traffic volumes
Delay on the roads caused by congestion is calculated provides an indication of the accuracy of the travel models.
using the Bureau of Public Roads curve shown in this chap- A summary of estimated and observed volumes can be found
ter. Coefficients for the formula were obtained from Table in Chapter 12.
1.04
CHAPTER 10
CAPACITY ANALYSIS
INTRODUCTION section would operate over, at, near, or under capacity under
the future travel demand.The 1994 Highway Capacity Man-
This chapter addresses capacity analysis as used in the al provides a procedure that is presented in this chapter. The
planning of transportation facilities. Capacity analysis is user is directed to the manual for information on applying a
used at two stages in the planning process: more detailed analysis of signalized intersections.
This chapter contains two sets of capacity parameters and
. As input to the link attributes required in the network- procedures. The first is the provision of initial values of link
based travel-demand models and capacities that can be used by travel forecasting models. The
. As a post-modeling analysis tool to evaluate the ability second is the planning procedure for determining the capac-
of the transportation system to serve the future traffic ity of a signalized intersection.
..1 demand adequately or to determine how much addi-
tional demand the existing transportation system can
I accommodate before improvements are necessary.
- .j
BASIS FOR DEVELOPMENT
As noted in Chapter 2, link capacity is usually input as the
number of vehicles per hour (vph) per lane or as directional The techniques and procedures in this chapter have been
capacity per hour, depending on the requirements ofthe travel selected to address the types of problems a user of this report
demand software. The preferred method for deriving link is likely to encounter. The initial capacities of facilities that
capacity is to use the procedures contained inthe 1994 High- are needed as input parameters in the building of a highway
way Capaciry Manual (HCM) and compute capacities spe- network and the procedure for capacity analysis of signalized
cific to the physical limitations of each link. However, this is intersections for use once the volume projections from the
often not feasible, and the altemative method is to use link model are available are two such procedures.
capacities that reflect average conditions for various link The first part of the chapter concentrates on initial esti-
types. Tables 52 through 59 provide initial link capacities that mates of capacities for different facilities. These estimates
can be used in the building of the highway network. These are provided for use in the initial building of a travel fore-
capacities are based on the ultimate, or LOS E, capacity, casting model and are based on information contained in
The post-modeling analysis is typically done for both DelaylVolume Relations for Travel Forecasting which is
roadway segments and critical intersections within the urban based on the 1985 Highway Capacity Manual.t Estimates
area. The link-based or roadway-segment analysis is accom- based on the 1994 HCM are provided in NCHR P Report 387 ,
plished by comparing the assigned volume with the link's "Planning Techniques to Estimate Speeds and Service Vol-
capacity. This is done by computing the link volume/capacity umes for Planning Applications."
atio (vlc) and posting the result on network plots. The chapter ends by outlining procedures for signalized
The intersection analysis is based on the use of one of intersection capacity analysis based on the planning method-
the two procedures outlined inthe 1994 Highway Capacity ology in Chapter 9 of the 1994 Highway Capacity Manual.z
Manual. These procedures are closely linked to the operational analy-
The more rigorous of the two procedures is the operations sis of signalized intersections contained in the same chapter
'.ti: or design analysis of signalized intersections. This analysis of the 1994 HCM.
requires detailed information about the geometrics of the The operating condition of a highway facility is generally
intersection, the proposed signal plan, the volumes for each
measured using the concept of level of service. Level of ser-
tuming movement, the mix of vehicles in the traffic flow, and
vice has been stratified into six classes defined as follows:
:: j..
the arrival type of the traffic flow. This level of detail is not
available for long-range transportation planning and, more
important, is not appropriate for use in such planning. It is tAlan J. Horowitz, D elaylVolume Relations
forTravel Forecastin.g, based on the I 985
useful, however, to apply a generalized procedure for capac- Highway Capacity Manual,prepared fbr the Fedeal Highway Administration, U.S.
Department of Transportation (1991).
ity analysis of intersections in the plan development process. ?Transportation Research Board, Highwoy Capaciry Manual Chapter 9 (Signalized
The objective of such a procedure is to determine if an inter- Intesections) (1994).
105
Level of Service Operating Conditions
Free flow, low volume, high operating speed, high maneuverability

B Stable flow, moderate volume, speed somewhat restricted by tratfic
conditions, high maneuverability
c Slable flow, high volume, speed and maneuverability determined by
trattic conditions
Unstable flow, high volumes, tolerable but fluctuating operaton speed
and maneuverabilily
Unstable flow, high volumes approaching roadway capacity, limited
speed (= 30 mph), intermittent vehicle queuing
Forced flow, volumes lower lhan capacity because of very low speeds,
heavy queuing of vehicles, frequent stops
Historically, LOS C has been used as the goal for evalua- Iane volume by the number of lanes. Other software pack-
tion of the performance of the highway system. Recent lim- ages allow capacities to be set for individual links, thereby
itations of financial and physical resources have pushed this providing the user with more flexibility during calibration.
goal lower, to LOS D in larger, congested urban areas. The The capacities provided in Tables 52 through 58 ae rec-
user should be aware that many traditional travel demand ommended for starting values. These capacities have been
models use a capacity value that is equated to a given LOS, determined in accordance with the guidelines provided in the
usually C. In these cases, a v/c of 1.0 indicates that the link 1985 Highway Capacity Manual. Where the capacities are
will operate at LOS C. The initial capacities provided in this given as total directional capacities, they can be divided by
chapter are for LOS E, or ultimate capacity. the number of through lanes to obtain maximum lane vol-
There are two basic and independent indicators oflevel of umes. These values should not be varied by more than -20
service-the v/c rutio and the operating speed. Where applic- percent unless justified by abnormal deviation from ideal
able, the procedures in this chapter use only the v/c ratio to conditions.
assess service levels. For a detailed discussion of the con-
cepts discussed above, the user is directed to the 1994 High-
way Capacity Manual.3 Assumptions and Extensions
for lnitialCapacity
INITIAL SETTINGS FOR CAPACITIES FOR USE The initial capacities for uncontrolled road segments
IN TRAVEL FORECASTING MODELS assume 14 percent trucks, 4 percent recreational vehicles,
and 0 percent buses, as suggested for default by the 1994
Ideally, capacities should be set according to those obtained HCM for two-lane roads. The forecast period is I hour. Other-
from the 1994 HCM orfrom the Highway Capacity Software wise, ideal conditions are assumed.
(HCS) or similar programs. However, setting capacities sep- Priority ofsignalized intersections in Tables 55 through 57
arately on every link or on every intersection approach can relates to percent of available green time for the approach as
be quite tedious, especially considering that many of the val- follows: 33 percent : low priority; 50 percent = medium;
ues may change during network calibration. One approach is and67 percent : high. Turns in those tables relate to the per-
to start with rough estimates of capacities and then refine centage of non-through movements: 0 percent : low turns;
these estimates during calibration. 25 percenf : high turns. Initial capacities for a medium num-
Depending upon the forecasting software, the capacities ber of turns may be interpolated from the values for low and
can be entered in various ways. For example, Urban Trans- high turns.
portation Planning System (UTPS) and similar packages Consistency of priority must be maintained for all ap-
require that capacities be computed as a function of area type, proaches at any given intersection. For example, it would be in-
facility class, and number of lanes. A look-up table must be appropriate to have more than two high-priority approaches
prepared giving the maximum lane volume as a function of at an intersection.
area type and facility class. The software determines the The ultimate capacity of an intersection will be greater if
capacity of the link by multiplying the looked-up maximum the intersection has exclusive right-turn lanes. Ultimate
capacity for an exclusive right-turn lane can be added as fol-
sTransportation Research Board, 1iglrwat' Capaci rt Manuul,
Special Repon 209, 3rt1
lows for each through lane: 0 vph for low turns; 75 vph for
Ed. (r994). medium turns; and 150 vph for high turns. Additional design
106
TABLE 52 Initial capacities for multi-lane highways, each lane: ultimate

capacity
60, 70 mph 50 mph
Rural Divided Level Terrain 1,800 1,700

Rolling Tenain 1,350 1,250
Undivided Level Terrain 1,700 1,600

Rolling Tenain 1,250 1,200
Suburban Divided Level Terrain 1,600 1,500

Rolling Terrain 1,150 1 ,100
Undivided Level Terrain 1,450 1,350

'I Rolling Terrain 1,050 1,000
*t
I
----*.-t i
capacity for an exclusive right-turn lane should be provided . Additional capacity for each lane beyond
I
as follows for each through lane: 0 vph for low turns; 50 for the first 475
S: l medium turns; and 100 for high turns. For example, the ini- . Total capacity ofthree-lane approach 7,175
,J

tial ultimate capacity for an approach with two through lanes,
I
both exclusive left- and right-turn lanes, high priority, and Two-way stops are seldom included in regionwide net-
high turns shouldbe2,300 (2,000 + [2 x 150]). works. Capacity varies greatly with the amount of conflict-
For signalized approaches with three or more lanes, it is ing traffic for signed approaches at a two-way stop. Ultimate
necessary to extrapolate from the data for one and two lanes. capacity for each lane should not exceed 1,000 vph. See
For example, the initial ultimate capacity for a three-lane Chapter 10 of the 1994 HC]jNf for more information about
approach with high turns, medium priority, and an exclusive two-way stops.
left-turn lane may be computed as follows: For travel forecasting software packages that explicitly
allow signs and signals in the network, consult the software
. Two lanes, exclusive left, medium priority, reference manual. For example, QRS II requires that the
high turns 1,300 capacity be set to the total saturation flow rate of the through
. One lane, exclusive left, medium priority, lanes at the approach, without adjusting for signalization pri-
high turns 825 ority (amount of green) or amount of turning. For roadway
TABLE 53 Initial capacities for freeways, each lane: ultimate capacity
60, 70 mph 50 mph
Level Terrain 1,800 1,700

Rolling Terrainl 1,350 1,250
For planning purposes, grades of two percenl or higher may be considered rolling. For
more detailed evaluation of terrain, refer to Chapter 3 of the 1985 Highway Capacity
Manual.
t07
TABLE 54 Initial capacities for two-lane roads: ultimate capacity
Level Rolling
Peak Little No-Passing 1,500 1,050

Extensive No-Passingl 1,500 950
Off Peak Little No-Passing 1,200 800

Extensive No-Passingt 1,200 750
when no-passing zones exceed 50 percent of the length of roadway being evaluated,
extensive no-passing may be assumed.
sections containing multiple intersections, choose the small- described in this section, on the other hand, makes use of
est capacity. carefully determined default values for most of the data
required and is, therefore, much less data intensive. For a
more detailed explanation of the procedures explained in the
DETERMINATION OF INTERSECTION following sections, the user is directed to Chapter 9 of the
CAPACITY 1994 Highway Capacity Manual.
Once the volume projections from a travel forecasting

model are available, it is often necessary to perform capacity Input Data Requirements
analysis at signalized intersections in the network to test the
adequacy of these intersections and to identify improvements It is possible to perform an approximate capacity analysis
if necessary. The methodology outlined in this section at a traffic signal through the use of assumed values for most
enables the user to perform this task. ofthe data that are required. For planning purposes, the only
The intersection capacity analysis methodology described site-specific data required are the traffic volumes and the
here is based on the new methodology for planning applica- number of lanes on each approach, with a minimal descrip-
tion contained in Chapter 9 of the 1994 Highway Capacity tion of the signal design and other operating pa-rameters'
Manual. The operational analysis method provides an Tables 59 and 60 contain recommended default values for
extremely detailed treatment of the operation of a traffic sig- other data items to be used in the planning analysis.
nal. The level of precision inherent in this analysis often The planning analysis described here is intended for use in
exceeds the accuracy of available data. The requirement for sizing the overall geometrics of a signalized intersection or
a complete description of the signal timing plan is data inten- in identifying the general capacity sufficiency ofan intersec-
sive, especially when the method is being applied in trans- tion for planning purposes. This procedure is based on the
portation planning situations. The planning analysis method sum of critical lane volumes and requires minimum input
TABLE 55 Initial capacities for single-lane, signalized intersection approaches: ultimate capacity
Low Tums High Tums
No Exclusive Left Low Priority' 550 350

Medium Priority2 825 550
High Prioritya 1,100 900
Exclusive Left Low Priority' 550 550

Medium Priority'? 825 825
High Prioritf 1,100 1,100
I j ,''.]
l
:: 108
,i
':
TABLE 56 Initial capacities for twolane, signalized intersection approaches: ultimate capacity
''.i
.;
Low Tums
.l High Tums
,..1
.-..1,!. :l No Exclusive Left Low Priorityr 1,100 650
t.* ',i
Medum Priorityz 1,650 900
'," j
'-':T'" High Priorityo 2,200 1,400
.,'''t'.i
,:,]
l Exclusive Left Low Priorityt 1 ,100 850
..: J
Medium Priority'z 1,650 1,300
''.: - .l
''..: ..] High Prioritf 2,200 2,O00
'-,:i
information. Three worksheets are provided for this analysis; Note that one Lane Volume Vorksheet is required for each
they include the basic worksheet shown on Figure 27, the of the four approaches. This will determine the equivalent
lane volume worksheet shown on Figure 28, which is used to hourly lane volume for each approach. The hourly volumes
establish individual lane volumes on each approach, and the are then combined on the Signal Operations Worksheet to
signal operations worksheet shown on Figure 29, which is determine the critical movement sum and the intersection
used to synthesize the signal-timing plan and to determine status. Optionally, the cycle length and phase times may also
the operational status of the intersection. The relationship be determined.
between these worksheets is illustrated schematically on Fig-
ure 30. The objective of using these worksheets is to deter-
mine the critical movement y/c fatio,X",,,, which is an approx-
imate indicator of the overall sufficiency of the intersection
Computational Requirements
geometrics. Although it is not possible to assign a level of
The capacity analysis design parameters must be based on
service to the intersection based onXc,,,, it is possible to eval-
il', the traffic volumes and lane confrguration of each approach to
i uate the operational status of the intersection for planning
--l the intersection. The steps in performing the analysis follow:
.i purposes. Table 6l expresses the status using descriptive
terms "over," "at," "teat," or "under" capacity.
1. Determine the lane volumes for each movement. The
detailed instructions for the lane volume worksheet
WORKSHEET APPLICATIONS describe this process.
2. Determine the type of left-turn protection for each
The relationship between the Lane Volume Vy'orksheet direction. For planning applications, the actual left-
and the Signal Operations Worksheet is shown on Figure 30. turn protection should be used if known. A left turn is
TABLE 57 Initial capacities for each lane beyond two, signalized intersection approaches: ultimate capacity
Low Tums High Tums
No Exclusive Left Low Priorityr 550 300

Medium Priorityz 825 350
High Priorityo 1,100 500
Exclusive Left Low Priorityr 550 300

Medium Priority'? 825 475
High Priority3 1,100 900
:. .. ,
When the green time for the cross slreet at a signalized intersection exceeds the green time for the approach being
,
'.1
:,.,,. j 1
evaluated, then the approach being evaluated has low priority.
When the green time for the approach being evaluated and the cross street at a signalized intersection are
approximately equal, then the approach beng evaluated has medium priority.
When the green time for the approach being evaluated exceeds that of re cross street at a signalized intersection,
the approach being evaluated has high priority.
l0g
TABLE 58 Initial capacities for all-way stops: ultimate capacity
Low Gonflicting Volume High Conflicting Volume
One Lane 1,000 500

Two or More Lanes 2,000 600
considered to be protected if it is able to proceed at lnstructions for the Lane Volume Worksheet
some point in the cycle while the oncoming through
movement is stopped. If the actual left-turn protection The following instructions cover the step-by-step proce-
is unknown, a simple method will be presented later for dure for completing all of the items on the lane volume work-
- -- l --
determining an appropriate choice. snget. Eacn step ls ltulltucteu [u uurrcsPullu
-^,--^^-^-l wll.ll .L^
--.:L
luw url
trl -^.., ^-
3. Select the phase plan from a choice of six alternative the worksheet.
plans that will provide the desired degree of left-turn
protection and will accommodate the observed lefrturn L LefrTurn Volume: The first item is the left-turn volume
volume balance. (in vehicles per hour) on the approach. In the case of
4. Determine the sum of the critical volumes for each protected-plus-permitted phasing with an exclusive left-
phase and the intersection status (under, near, at, or turn lane, two vehicles per cycle should be removed
over capacity). from the left-turn volume to account for the effect of
sneakers. If the cycle length has not been established,
This completes the planning analysis. If an estimate of the the maximum cycle length should be used. To prevent
level of service based on stopped delay is required, two addi- uffeasonably short protected left-tum phase durations,
tional steps are involved. In this case, the user is directed to this volume adjustment step should not reduce the lefr
Chapter 9 of the 1994 Highway Capacity Manual. turn volume to a value below four vehicles per cycle'
TABLE 59 Default values for use in planning analysis
Trafflc Characterlstlcs
ldeal saturation flow rate 1,900 pcphgpl.
Pedestrian crossing volume Low 50 peds/hr
Moderate 200 peds/hr
High 400 peds/hr
Percent heavy vehicles 2%
Grade o%
Number of buses 0
Parking maneuvers 20lhr. where parking exists
Arrival type 4 if coordinated
3 if isolated
Peak-hour factor 0.90
Lane use factor See Table 60
Facillty and Trafflc Slgnal Characterlstlcs

Signaltype Pre-timed
Cycle length range 60 sec. to 120 sec.
Lost time 3.0 sec,/phase
Yellow plus all-red 4.0 sec./phase
Area type Non-CBD
Lane width 12ft.
* Passenger cars per hour of green lime per lane.

ll0
TABLE 60 Lane use factors
Lane Group Number of Lanes Percent of Traffic Lane Use

Movements in Lane Group in Heaviest Lane Factor, U
Through or shard 1 100.0 1.00

2 52.5 1.05
3 36.7 1 .10
Exclusive lefl turn 1 100.0 1.00

2 s1.5 1.03
Exclusive right turn 1 100.0 1.00

2 56.5 1 .13
2. Opposing Mainline Volume: Opposing mainline volume If there is no exclusive righrtum lane, a value of 1.0
is defined as the total approach volume minus the left-tum should be used for Step 7. The result is entered as Step
volume from exclusive lanes or from a single lane (in vehi- 9 if one or more exclusive right-turn lanes exist or as
cles per hour). The cross product (t2l X tll) may now be Step 10 if right turns must share the lane.
computed by multiplying the opposing mainline volume 11. Through Volume: Total through volume for the
by the left-tum volume. This gives a value for comparison approach excluding left and right turns should be
to determine if a protected phase should be assumed. placed in the appropriate column to correspond with
3. Number of Exclusive LeftTurn Lanes: This would be the applicable treatment for left tums (permitted, pro-
the number of lanes exclusively designated to accom- tected, or not opposed).
modate the left-turn volumes. 12. Parking Adjustment Factor: The parking adjustment
4. LefrTurn Adjustment Factor: The left-turn adjustment factor should be placed in the appropriate column, as
factor applies only to protected left turns from exclu- explained in Step 11. This factor corresponds to the
sive left-turn lanes or to left turns that are not opposed. assumed value of 20 parking maneuvers per hour and
This factor is given as 0.95 for single lanes and is fur- depends on the number of through lanes available.
'I
I
ther reduced to 0.92 for dual lanes. If the left-turn The values are 0.800, 0.900, and 0.933 for one, two,
I
movement is not opposed because of a one-way street and three lanes, respectively. Ifno parking exists, the
:
or T-intersection, pedestrian interference must be con- factor equals 1.0.

,] sidered. The corresponding value of 0.85 for one lane
i 13. Number of Through Lanes Including Shared Lanes:
I
and 0.75 for two lanes should be used. This step is self-explanatory. Exclusive turn lane or
5. Left-Turn Lane volume (tll/l3l x [4]): The roral lefr- lanes should be excluded.
turn volume from Step 1 should be divided by the prod-
uct of the number of exclusive left-turn lanes (Step 3) At this point it is necessary to distinguish between
and the left-turn adjustment factor (Step 4). The left- exclusive left-turn lanes and shared left-turn lanes.
turn volume should be entered directly if there is no The procedure for exclusive left-turn lanes will be
exclusive left-turn lane. The result is expressed in vehi- described first. Note that Steps 15 and l7 do not apply
cles per hour per lane. Zero should always be entered if to exclusive left-turn lanes.
the left turns are permitted.
6. Right-Turn Volume: Right-turn volumes (in vehicles 14. Total Approach Volume ((tlOl + [l])/[2]): The
per hour) from either a shared through and right-turn total approach volume is the total of the shared lane
lane or from an exclusive turn lane or lanes should be right-turn volumes plus the through volumes. Note that
entered. If available, the right-turn-on-red volume the through volumes are adjusted (increased) by the
should be subtracted. parking adjustment factor to account for the effect of
7. Exclusive Lanes: This is the number of lanes assigned parking on through volumes, for example, momentary
exclusively for right turns, if any. lane blockage. Note also that left-turn volumes are
8. RighrTurn Adjustment Factor: The right-turn adjust- excluded because they are not a part ofthe lane group.
ment factor is given as 0.85 for a single lane or a shared 15. Not applicable to exclusive left-turn lanes.
lane and reduced to 0.75 for two lanes. 16. LefrTurn Equivalence: Left-turn equivalence, deter-
9,10. Right-Turn Lane Volume (l6l/(Ul x [8])): The total mined from Table 62, is not used in lane volume cal-
right-turn volume from Step 6 should be divided by culations when exclusive left-turn lanes exist. This step
the product ofthe number ofexclusive right-turn lanes is, however, required for permitted left turns to assess
(Step 7) and the right-turn adjustment factor (Step 8). the adequacy of the left-turn treatment in Step 20.
llr
PLANNING METTIOD INPLTT WORKSTIEET
J l\* A
WBTOTAL
EB TOTAL
-r 1r- NB TOTT
APPROACHDATA NB SB EB WB AnType
cBD D
Puking Allowcd !t!
Cmdinaion
Ifr-Tum Trsent
tr PIIF
Other C
Pemitted
Prctectd
trtrtr!
D Clcle -Langth
Not Opposed tr!tr Min
Max
-
-
Figure 27. Planning method input worksheet.
17. Not applicable to exclusive left-turn lanes. lanes for through volumes. Left-turning vehicles
18. Through-Lane Volume ([a]/[13]): The total approach blocking the shared left-turn and though lane will pre-
volume should be divided by the number of lanes to vent through vehicles from proceeding until the hrrn-
obtain volume per lane, which is the basis for com- ing vehicles have been able to make the turn.
puting critical lane volumes.
19. Critical Lane Volume: Step 19 is normally the same as 14. Total Approach Volume: The total approach volume
Step 18 except when the right turn has an exclusive is computed in nearly the same manner as in Step 14
lane or the left turn is not opposed and either of these for exclusive left-turn lanes, that is, ([10] +
movements is more critical than the through move- tl1l)/t121. The difference is that the volume from
ment. If both conditions apply, the critical lae volume Step 5 must be added to the through volume in Step
will be Max (t5l,tgl,tl8l). If a shared lane exists for the 11 if the left turn is not opposed.
right turn, Step 9 should be eliminated. If the left tum 15. Proportion of Left Turns in Lane Group: Step 15 is
is permitted or protected, Step 5 should be eliminated. self-explanatory. This data item is required for the fol-
low-up computations.
The case of shared left-turn lanes is more compli- t6. Left-Turn Equivalence: Determined from Table 62, this
cated and therefore requires a more detailed procedure. is one of the factors needed to compute the applicable
Steps 14 through 18 are used to approximate the effect formulas from Table 63 for sharedJane permitted left
that left-turning vehicles have in reducing available tums. It is not used at all when the left turn is protected.
PIINING UETIIOD IJ{E VOIrItllE WORKSHEET
Locat i on: Direction

Left Turn llovsrt Ridrt Turn lovsrt Excltsive Shared
R [rF nf [a
1. LT votune . RT volne
2. 0pposing mainline votrne 7. RT Lanes
3. llo of exctusive LT tanes 8. RT adjustrnent factor

| 4. LT adjustnrent fsctor RI tane vot: t9l t10l
(See instructions)
Cross product: t2l * t1l

5. LT tane votrme: 11 I( t3 r t4l ) 0
Thradr Xovent
11. Through votume
12. Parking adjustnnnt factor

13. No. of through tanes inctrding shared lanes
Lf tr cqtatid
Exctrsive
14. Iotat approach votum: (t10 + tlll) /t121

1. Left turn equivatence: (Figure 9-7) xxxxxxxxx xxxxxxxxx
lE. Through lane vottme: l14l I t13l
19. Criticat Iane volup: (see instructions)
Shared LT ls cdrtatics
14. Totat approach votup: (See instructions)
15. Proportion of left turns in the lane group xxxxxxxxx xxxxxxxxx

16. Left tun eguivatence: (Figure 9-7) xxxxxxxxx xxxxxxxxx
17. Left turn adjustment factor: (Tabte 9-15) 1.0
18. lhrough lane votup: l14l I (t131 * tln )

19. criticat tane votrm: llax( f9l , tlEl )
.i,. .,..
Left Turn [heck (if tl > 8)

20. Permitted left turn sneaker capacity: 7200 I C^_ xxxxxxxxx xxxxxxxxx
:. i: ".
Figure 28. Planning method lane volume worksheet.
113
PIJANING I,ETHOD SIGNL OPEnTIONS WORKSHEET
phase Plan selectiqr fro La Yotr. uorksheets E sE(lt lfisTBflm xnTnBflm STTTHB(IIL
CriticaI Through-RT Iane votune: [19]
LT lane vottme: [5]
Left turn protection: (Perm, Prot, N/0)

Dominant left turn: (lndicate by r*r
Selectiqr Criteria bssed m the Plr 1: Per Perr Pelr PerE

specified left tur'n treGtt: Perr ll0 Perr )lO
I/o Per I,l0 Pert
Plsr 2a: Perr Prot Petr Prot
Pll 2b: Prot Per. Pot Peil
t lrdicates the dcinsrt teft turn Pl 3a: rProt Prot tProt Plot
for each oposirg pair Plgr 3b: Prot tPlot Prot rProt
Ptsr 6: t,lo )lO tllJ ll0
Phase ptan setected (1 to 4)
l,lin, cycte [C.h] _ l4ax cycle [C,-] _ IPHF] (trom nput l,orksheet)
Phasir Ptan Frc Table 9-l EAST-ttr$ Ielll-S(UTll -----

Iote Vat.c Phase I Plrase 2 Phase 3 Phase I Phase 2 Phase 5
l,lovement codes
Criticat Phase Volme [CV]
Criticat sun tcsl 1
Lost time/phase [PL]
Lost time/cycte tTLl 2
CBD adjustment ICBDI 3
Critical v/c ratio [X-l 4
tntersection status 5
(tifit Ti.irF Ptan cGtatiqr

Reference Sun tRSl 6
Cycte tensth ICYCI 7
Green time
Iotcs
1. Criticat sun = stm of critical phase volunes [cv's] for atI phases.
2. Lost time/cycte = Srm of att lost times/phase, [PL's].
3. CBD adjustment = .9 rithin cBD, 1.0 etsexhere.
4. critical v/c ratio = CS /((1'[TL]/C.-) * 1900 i ICBDI * IPHFI).
5. Status: (See instructions).
. Reference Su = 1710 * [PHt] * tcBDl.
7. Cycte length = tTL] / (1-(l{in(tcsl,tRsl) / tRsl)), subject to [c.h] and [C,ol.
8. Green time = (tcYcl-tTLl) * (tcvl/tcsl) + tPL.
Figure 29. Planning method signal operaions worksheet.
114
left-turn volume is greater than two vehicles per cycle

Signal Operations (i.e., [] > l,200lC^^*), it is most likely that the sub-
Worksheet ject left turn will not have adequate capacity without
a protected phase.
,.'.1 Signal Operations Worksheet

J
Of the six steps involved in the planning method, only the

first two are carried out by the lane volume worksheet. The
last four steps are included in the signal operations worksheet,
Lane Volume Worksheets which is shown in Figure 29.'1o facilitate the use of the sig-
(1 per Approach) nal operations worksheet, the lane volumes are transferred
from the lane volume worksheet before the computations
Figure 30. Planning method worksheet relationships. begin. Note that the through-movement lane volume is taken
as the heavier of the through or right-turning movement when
17. Left-Turn Adjustment Factor for Through Traffic /: an exclusive right-turn lane is present. In other words, if the
The appropriate formula in Table 63 should be used. volume of a right turn from an exclusive lane is heavier than
This is a reduction factor applied to the through vol- that of the through movemenq the right-turn lane volume will
umes to account for the effect of left-turn vehicles be considered as the through volume for design purposes.
waiting for a gap in the opposing traffic to make the
turn. Note that for lanes that are not opposed, the fac- 1.Transcribed Data ltems: The peak hour factor (PHF) was
--l tor must be 1.0 because these vehicles will have gaps entered on the Planning Method Input Worksheet. The
i in which to turn. appropriate value is discussed in connection with the
18. Through-Lane Volume: Total through volume in the description of that worksheet. The left-tum treatment is
approach should be divided by the number of through also transcribed to the signal operations worksheet from
:)95i!q lanes. Note that the number of lanes is reduced by the the input worksheet. It is not necessa-ry to specify
factor obtained in Step 17 to account for the effect of whether the treatment includes a permitted phase for the
the left-turning vehicles. left turn in addition to a protected phase. The synthesis
19. Critical Lane Volume: The critical lane volume is the of the signal timing plan does not consider protected-
maximum of either the value computed by Step 18 or plus-permitted operation. That, of course, does not pre-
the right-turn volume from an exclusive rightturn clude specification of this type of operation in the analy-
lane as computed in Step 9. sis. At this time, only determination of reasonable values
20. Left-Turn Check: If one or more left turns have been for the cycle length and phase times is of interest.
designated as permitted (i.e., no protected phase has 2. Phase Plan Selection: The phase plan is selected from
been assigned), the need for a protected phase should six alternatives that cover the full range ofleft-turn pro-
be reexamined at this point. If the cross product ([2] x tection requirements. A phase plan deals with only one
[1]) exceeds the adopted thresholds, a protected left- street at a time. The complete signal sequence will
turn phase should be assigned for planning purposes involve two phase plans: one for the east-west street
unless existing traffic volumes have been used and it and one for the north-south street. The choice between
is known that such a phase does not exist. phase plans is made by examining the left-turn protec-
It was indicated in Table 62 (left-turn equivalence) tion for both pairs of opposing left turns. The altema-
that values above 8.0 indicate that left-turn capacity is tives include the following:
derived substantially from sneakers. Therefore, if the
left-turn equivalence [16] is greater than 8 and the . Plan 1: No left-turn protection in either direction.
In this case, the phase plan includes only one
TABLE 61 Intersection status criteria for signalized phase, in which all through and left-turn move-
intersection planning analysis
ments may proceed, with the left turns yielding to
the opposing through traffic.
Critical v/c Ratio Relationship to . Plans 2a and 2b: These two plans involve left-tum
X"t Probable Capacity protection for only one of the two opposing left
tums. Two phases will be involved in this case. In
x-_ < 0.85 Under Capacity the first phase, the protected left turn will proceed
.85<X",<0.95 Near Capacity with the through movement in the same direction.
.95<X".<1.00 At Capacity In the second phase, the two through movements
x", > 1.00 Over Capacity
will proceed. Plans 2a and 2b differ only in terms
of which of the two opposing left turns is protected.
115
TABLE 62 Through-car equivalents, E,, for permitted left turnsa
Totat No. Type of llo. of

of Signat Left Turn 0ppos i ng Opposing Flor, V"
Phases Lane Lanes
0 200 400 00 800 1 000 >1200
1 1 .05 2.0 3.3 6.5 1.0* 1. 0* 1.0*

Shared ? 1 .05 1.9 2.6 3.6 6.0 16. 0* 1.0*
:3 1 .05 1.8 2-5 3.4 4.5 .0 1.0*
2
1 1 .05 1.7 2.6 1.7 10.4* 10.4* 10.4*

Exctusive 2 1 .05 1.6 2.2 2.9 4.1 6.2 10.4*
3 .05 i. 2-1 . J- 4.8 tu..r'
1 1 .05 z.z 4.5 11.0* 1 1.0* 11.0* 11.0*

Shared 2 1 .05 2.0 5.1 4.7 1 1.0* 11 .0* 11.0*
l,lore than >5 1 .05 2.0 2.9 4.2 .0 11.0* 11-0*
z
1 1 .05 1.8 3.3 8.2t 9.2* 8.2* 8.2*

Exctusive 2 1 .05 1.7 ?.4 3. 5.9 8.2* 8.2*
>_3 1 .05 1.7 2.4 3.3 4.6 .8 8.2*
* Generatty indicates turning capacity onty avaitabte at end of phas+rtsneakers,,onty.

4
Mnsspn, C. J., and Frrrno, D. 8., "Critical Lane Analysis for Intersection Design." Transportation Research Record 644,
Transportation Research Board, Vy'ashington. D.C. (1977').
. Plans 3a and 3b: Both opposing left turns are umes (CVs), and lost time per phase may be entered on
protected here. In the first phase, the two oppos- the worksheet. The appropriate choice for critical lane
ing left turns will proceed. In the second, the volumes is given in the phase plan summary shown in
dominant left turn will continue with the through Table 64 along with a code that identifies the move-
movement in the same direction. In the third, the ments allowed to proceed on each phase. The movement
two through movements will proceed. Plans 3a codes ae defined in a note to Table 64. For example,
and 3b differ only in terms of the dominant left "NST" indicates that the northbound and southbound
turn that governs the display in the second phase. through movements have the right-of-way on the spec-
. Plan 4: This is generally known as "split-phase" ified phase. The corresponding code for the two oppos-
operation. Two phases are involved, with the ing left turns moving concurrently is "NSL." If the
through and left-turn movements from one of the northbound through and left turns are moving together,
two opposing directions proceeding on each the code is "NTL." Table 64 also indicates the lost time
full directional sep-
phase. This has the effect of to be assigned to each phase.
aration between the two approaches. From a Thus, the movement codes and CVs must be deter-
capacity analysis point of view, it is equivalent to mined for each phase from Table 64 and entered on the
two one-way streets that meet at a common point. signal operations worksheet. When all phases have
been completed, the critical sum (CS) of the CVs must
The selection criteria are presented in a table on the be entered on the next line.
signal operations worksheet. Note that the selection is 4. Lost Time Determination: For planning purposes, it is
made on the basis of the user-specified left-turn protec- assumed that there is a lost time value of 3 sec per phase
tion and the dominant left-turn movement identified from in which any movement is both started and stopped.
the lane volume worksheet. For one- and two-phase plans, there is a lost time asso-
3. Critical Phase Volume, CV: When the phase plan has ciated with each phase. For three-phase plans (Plans 3a
been selected, the movement codes, critical phase vol- and 3b), the second phase requires no lost time because
116
TABLE 63 SharedJane left-turn adjustment knowledge of the intersection status is sufficient. The
computations for planningJevel analysis timing plan is required only if the planning analysis is
PERMITED LEFT TURN to be extended to estimate the level of service.
The cycle length may be determined from the fol-
Iane groups with two or more lanes:
lowing formula:
',.,: llTl = {[13] - I + ({r3l'r('6v6m)}/[13]
; .::,r
*'-:Ti TL
Subject to a minimum value that applies at very low left-tuming c= 1- / S]
(10-1)
volumes when some cycles \ilill have no left-turn arivals: [Min(CSrRS)
U7l = {U3l - I + (-rl'c/36l}/u3l

,': Lane groups with only one lane for all movements:
where rRS is the reference sum of phase volumes rep-
'..:;: )' resenting the theoretical maximum value that the
[17] = -(o'0r'tttc'10'{15))'tlct600}
intersection could accommodate at an infinite cycle
PROTECTED-PLUS-PERMITTED LEFT TURN length.
(oNE Dn.ecrtoN oNLY)
The recommended value for the reference sum is
tf 21 < t22o (I,7I0 x PHF). This value should be reduced by 10
ltTl = 111 + [(235 + 0.435*[2])*0511/(1400 - t2l)] percent in CBD locations. The value of 1,710 is 90
tf 121> t220
percent of the ideal saturation flow rate of 1,900
pcphgpl. It will attempt to produce a 90 percent v/c
llll = 111+ 4.525*51)
ratio for all critical movements. The cycle length
determined from this equation should be checked
none of the movements are both started and stopped. against reasonable minimum and maximum values.
Thus, as a simple rule, phase Plan I involves 3 sec of The determination of appropriate values is discussed
lost time per cycle, and all other plans require 6 sec.
in connection with the Planning Method Input Work-
When the lost times have been determined for each
sheet.
phase, the total lost time per cycle (TL) may be com-
The lost time per cycle must be subtracted from the
puted and entered on the worksheet.
total cycle time to determine the effective green time
5. Critical v/c Ratio, X.,,: The planningJevel critical v/c
per cycle, which must then be apportioned among all
ratio, Xa,, is the ratio of the critical sum, CS, to the sum
the phases. This is based on the proportion of the criti-
of the critical lane volumes that could be accommo-
cal phase volume sum for each phase determined in a
dated at the maximum cycle length, computed as
previous step. The phase time should be entered on the
(1 TL/C^^) X 1,900 X CBD X PHF worksheet.
-
As a final step, the lost time must be added to the
The intersection status is determined directly from X"^ effective green time for each phase to determine the
using the threshold values given in Table 61. total phase time per cycle. The phase times for all of
6. Timing Plan Development: The development of a tim- the phases should be equal to the cycle length and
ing plan is optional. For many planning applications, a should be entered on the last line of the worksheet.
TABLE 64 Phase plan summary for planning analysis
EAST-WEST NORTH-SOUH
PHASE PLAN PHASE NO. LOST TIE MOVEMENT CODE CRTICAL STIM MOVEMENT CODE CRTICAL SUM
I I EWT Max(ET,EL,WT,WL) NST Max(NT,NL,ST,SL)

2a I 3 WTL WL STL SL
2 3 EWT Max(WT-WL, ET) NST Max(ST-SL, NT)
2b I 3 ETL EL NTL NL
) 3 EWT Max(ET-EL, W NST Max(NT-NL, ST)
3a I 3 EWL WL NSL SL
,,
0 ETL EI-WL NTL NL-SL
3 3 EWT Max(WT,ET-(EL-'fi-)) NST Max(ST,NT-(NL-SL))
3b t 3 EWL EL NSL NL
2 0 WTL ItrL-EL STL SL.NL
3 3 EWT Max(ET,tIlT-(wL-EL)) NST Max(NT,ST-(SL-NL))
I J ETL
Max(ET,EL) NTL Max(NT,NL)
) 3 WTL
Max(WT,1VL) STL Max(ST,SL)
Norz: E\VT = castboud and wcstboud though; ETL = castbound through and lcft; ttrTL = wcstbound through and lcft: NST = nothbound and southboud
thorgh STL = southbound through and lcfr; l,lll- = norrhbord through and lc: ET = eastbound through: EL = casttround left: WT = wcstboud though; WL =
wcstbord lcft: NT = norrbboud through; NL = northbound lcfr: ST = southbound thmugh: SL = southbound left.
tl't
PLANNING METHOD INPUT WORKSHEET
Elden Street arid Park -venue
Sharn Sabanayagam TmcPcriodalyzedt A!{ Peak

Herndon/r.rirqinia
Park .Aveaue
SB OTAL
FBr I 2 A
l*r tra
g\-
WB TOTAI
Elden Street
ffi---1
/-
EBrorAL \
l
80
rz3 "rIru
134
F-lNB TOTAL
fYpc
APPROACHDATA NB SB EB WB Ara
cBD tr
Parting Allowcd
Coordinstion
ntrtrtr
trtrtrtr Othcr E
pmo--90
Ift-Tlm TrtEct
Pc|lrntd 6n
Dtr8 Cyclc Lcngth
Prcctcd
Not Opposcd ttrtr $l
'ti
Figure 3 L Planning method input worksheet-illustrated example'
LIMITATONS OF THE PLANNING METHOD is strong confidence in the validity of the traffic data, this
method should not be taken beyond the worksheet stage.
The planning analysis technique described in this chapter
offers a method for synthesizing a reasonable and effective lllustrated Example
signal-timing plan based on the traffic volumes and lane uti-
lization at an intersection. Using the worksheets included
The following pages illustrate an example using the
here, it is possible to determine the approximate status of the
of a signalized intersection with respect to its methodology for capacity analysis of signalized intersec-
operation
tions described in this chapter. The volumes shown on Fig-
capacity.
It is also possible to take the analysis considerably further ure 31 are existing volumes. Typically, the volumes used
are future-year projections from a travel forecasting model,
and obtain the level of service on each approach by the
operational analysis method. Software has already been though existing volumes may be used. The purpose of the
developed that witl implement the worksheets and invoke the analysis is to determine whether this intersection is operat-
operational analysis method. This introduces a very powerful ing below, near, at, or over capacity in the year for which vol-
capability. However, the numerical precision of the results umes are available.
may greatly exceed the accuracy of the original data. In par- Figures 32 through 36 present an illustrated example of the
ticular, great caution should be employed when using traffic planning application methodology for capacity analysis of
volume projections to some point in the future. Unless there signalized intersections.
118
1
PIIWING IIBIBOD LTTE VOT.Ire TNORSSE8T
Location: Elden Street and Park Ave. Dtrectron

. Eastbound
Lcft furn lovrrnrt Ridtt lurn lovr-rt Ercleiv: Shred
R Le RT L
1. LI voture 80 . RT volp t34

2. Opposing maintine votre 9 7. RT Lanes 1
3. llo of exctusive [T lncs

I 8. RI 6djustment fctor 0.85
NA (permitted)
4. LT sdjustnt factor R lane vot: t9r - flot 1 5g
(Sec instructions)
Cross product: t2l r tll 720 Pcrittcd Protcctcd lot @rd

5. LT tane voture: t1l I ( t3l r f4t ) O
lh'aCl fovcnt
I
1,l. Through votune I23

l
12. Porking adjustlurt foctor 1.0
13. llo. of thror,gh tancs inclrding shorcd lncs
r - Ecl,ic LT l caFttiq
14. lotal approach votp: (tlOl + f1t) /tt2 28t
i
1. Left turn equivalence: (Figure 9-7) 1'1 xxxxxxxxx xxxxxxxxx
r lE. Through tane votrrp: fl4l / fl3 _281

19. Criticat [ne votrp: (Sec instructions) 28r
Shr.d tT l cqttic --------
14. Totat approach votue: (See instructions)
15. Proportion of teft turns in the tane grotp xxxxxxxxx xxxxxxxxx
1. Left turn eqrivatcnce: (Figure 9-7) xxxxxxxxx xxxxxxxxx
17. Left turn adjusttrpnt factor: (Tabte 9-15) 1.0
18. Ihrough tane votrrp: l14l t (fl5l t ti7t)
19. criticat lane votule: Hax( t9l , flEl )
Lcft fur Ghccl (if t16t > 8)

20. Prnitted left turn sncaker capacity: 7Z0O I C^- xxxxxxxxx xxxxxxxxx
Figure 32. Planning method lane volume worksheet-illustrated example: eastbound.

n9
PIÎUING !EI8OD IE VOIIS I|ORRSBEET
o."on, Elden Street and Park Ave. Direction trilestbound

Lcft frrrr lov-rt liCrt lurn lowGrt Eclrivc slurcd
RT LT T Ld
10 2
1. LT votum . RT votup
2. Opposing nintine volrrp r23 7. RI Lanes 1
3. of exctusive 0. 85
lo LT tencs 8. RT djustment factor
4. LT adjustt factor lI A(permitted) R1 tane vot : 19 - t10
(See instructions)
cross prod.rct: l2l r f 'll 1' 230 ---' Prtlittld Petrctrd *et aleld
5. LI tane votne: 11 / ( l r f4 ) o-
Ihrcdl lov-rt
11. Through votue
12. Paking Edjuctrnt factor 1.0
13. llo. of through tancs including sharcd tars I
Erctuirr Lf t c.Ttrtiq
14. lotot approach votrp: (tlol + fll) /tlz 1l
1. Left turn eivatence: (Figurc 9-7) 2.o xxxxxxxxx xxxxxxxxx
lE. Thrqgh tane volrrp: tl / t13 11
19. Criticat tane votrp: (S instructions) t1

Shrrcd LT t cqtticc
14. Totat approach volum: (See instructions)
15. Propoition of teft turns in the lane groLp xxxxxxxxx xxxxxxxxx
16. teft turn eqrivatence: (tigure 9-7) xxxxxxxxx xxxxxxxxx
17. left turn sdjustnt factor: (Tabte 9-15) t.0

18. Through lane votrr: 1tl I (tlSl * tln)
19. Critical, lane votme: l{ex( t9!, t18l )
Et!-Iurft_ltrcc (if t16l > 8)

20. Permitted teft turn sneaker capecity: 72OO I c^- xxxxxxxxx xxxxxxxxx
Figure 33. Planning method lane volume worksheet-illustrated example: westbound.

120
LTTWIIIG HEMOD ftNE VOIIDG WORSSEET
Elden Street and l?ark Ave. Northbound

Location: Direction
LGft lu lorcft Ridrt ltra tovr-rt Ecllivc shrd
RT RT LT
83 :d 27
1. LT volum . RT volum
2. Opposing ninline volurp 374 7. RI Lanes 1
3. llo of exctusive LT lencs E. RT adjustmnt factor 0.85
4. LT adjustmnt fector l{ A(oermtted) RT tanc vot: t9 - n0 32
(sc instructions)
cross Product: t2l r Il! 3IrO42 "'> Pcfittcd Protcctcd lot ltGd
5. LT tane voture: 111 I( f3 t t4l ) 0
thl.d toYnt
11. Through votue 529
12. Prking adjustr*nt factor 1.0

13. ilo. of through lancs inctr.ding shared lancs I
Ectiw tT t cqtrtio
14. lotal spproach votrrp (tlol + tl) /tlzl s61
'16. Lcft turn cquivtcnce: (tigurc 9-7) 2.5 xxxxxxxxx xxxxxxxxx
18. Through lanc volrrp: t14t / tll 561
19. Criticat tnc votrrp: (Sc instructims) 561
Sr.cd Lf t caFt.tiq --------

14. lotot epproach votrrp: (See instructions)
15. Proportion of teft turns in thG tnc aroup xxxxxxxxx xxxxxxxxx
1. Lcft turn eqrivotcnce: (Figurc 9-7) xxxxxxxxx xxxxxxxxx
17. Left turn adjustmnt fsctor3 (Tebte 9-15) t.0

t8. Ihrough lanc votue: a14l I I t13 r ttn )
19. Cri ticel, tanc votu: l{ax( t9l , tl8l )
Icft Tur fhcck (if fl > 8)

20. Permitted teft turn snclker cepacity: 7200 I C^_ xxxxxxxxx xxxxxxxxx
Figure 34. Planning method lane volume worksheet-ilLustrated example: northbound.
:.i
.'
121
PIINNING llEllOD IttE VOIIDIE !|ORSEEET
Elden Street and Park ,{ve Southbound

Locat i on: Direction
Lcft Tlr loulcrt tEI!E!-!sl!E Echivr srcd
RT LT RT Lr
27
1. LT votne . RT votune
2. opposing mEintine volr s56 7. RI Lancs 1
0.85
3. tlo of exctusive LT tarcs E. RT djustnt factor
N A(oermtted) Rf lane vot: t9 i2
4. LI adjust[Ent factor t10
(see instructions)
cross proct:tzl I tll 3'982 "-' rllJlrc.! llglg! Iot @rd

5. LI tane votue: t1l / ( 15 r t4l ) 0
ThFd loYtrt
347
11. Ihrough voltp
12. Perking adjustrnt fsctor
1.0
13. llo. of through lacs inctding shercd trt*s
Erctuiru Lf tc cqttiqc
379
14. Totat approoch votrr:: (t10 + tlll) /t121
1. Left turn cqrivatence: (Figurc 9-7)
4
'2 xxxxxxxxx xxxxxxxxx
379
18. Through lanc volrrp: t14l / ttll
379
19. criticat Iane votunc: (sc instructions)
s!rcd LT t caFtttia i
14. Totat epproach votue: (See instructions)

15. Proportion of teft turns in the [6ne group xxxxxxxxx xxxxxxxxx
1. Left turn eqtrivatence: (Figurc 9'7) xxxxxxxxx xxxxxxxxx
17. Left turn adjrstmnt factor: (Iabte 9-15) 1.0
18. Ihrough tene volune: a14'l I ( t13l r tin )
19. Critic6l tane votrne: llax( t91 , llEl )
Left T!rn dtccl (if t16l > E)

20. Permitted teft turn sncoker capocity: 7200 I c^-
Figure 35. Planning method lane volume worksheet-illustrated example: southbound.
122
PITIINING ITIETITOD SIG}II OPERTIONS WORKSHEET

Phase Ptr Setectiqr frc Lr Yot.- llorksleets EtSIglD lfSTB(t l(nfB[fx, s[ftf[BtD
Criticat Through-RT lane votume: [191 281 11 561 379
LT tane votume: [51 0 0
pern perm Derl:t oerm
Left turn protection: (Perm, Prot, til/O)
r Dominant teft turn: (Indicate by t*r) * :t -

Selectim Criteria boeed qr the Ptsr l: Pcrr Pcrr Perr Peil
spccified left tura treaGrt: Perr r'to Perr rlo
vo Perr )lo Perr
Ptsr 2a: Perr Prot Per Prot
Ptm 2b: Prot Perr Prot Pe!
r lrdicates the daint teft turn Ptrr lla: tProt Prct .Prot Prot
for eadr opoeil peir Plsr 5: Prot rProt Pot tProt
Ptsr 4: I,l0 rlo rto tto
I
Phase ptan setected (1 to 4)
I
llin. cycte tc,h lq ttax cycte tc,-l .L4l IPHFI (From Inpr.rt ltorksheetl ![
i
Phirp Pts Frc Table 9-l - E Sr-tfsl xnTil-slrrTH -----

I
Iote Val.n Phasc I Phese 2 Phe 3 Plrosc t Plrasc 2 Phase 5
l4oveilent codes EI.N NST

I
Criticat Phase Votrme tcvl 281 561

Criticat Sum tCSl 842
Lost tine/phase tPLl 3
Lost time/cycte tTLt 6
cBD ICBDI 3
edjustnent 1.0
Criticat v,/c ratio [X-l 4 .052
tntersection status 5 under capacty
(tiql8t fiirr Ptsr Cutatim
:. Reference Sum IRSI 6
Cycte tength [CYC] 7
Green time
Iotes
1. Critical, sutt = Srm of criticat phase volurs [CVrsl for ail fliases.
2. Lost tinn/cycl,e = Srn of atI tost times/phase, tpLrs.
3. CBD adjustnent = .9 xithin CBD, 1.0 etserhere.
4. Criticat v/c ratio = CS /((1-[TL],/C._) * 1900 t TCBDI * tpHFl).
5. Status: (See instructions).
. Reference Sun = 1710 * tpHF * tCBDl.
7. Cycte length = tTL / (-(l,tin(tCSl,IRS) / tRS])), Subject to [C,.J and [C,ol.
8. Green time = (tCyC-tTLl) * ([CVI/[CS) + tplt.
Figure 36. Planning method signal operations worksheet-illustrated example.
123
CHAPTER 11
D EVE LO PM E NT D ENSITY/H IG HWAY S PACI NG RE LATIO NS H I PS
INTRODUCTION The purpose of this chapter is to present a methodology

designed to relate suburban development to estimates of high-
The trend toward lower population densities in and around way levels of service so that the planner and policy maker can
major metropolitan areas has been underway for decades and
rapidly assess the highway transportation needs of land-use
nnrentl\/ ic nnnfinrinc 'Fhe I OOf li.^l^.^l
uvlvJvu +L^+
lll4t growth anci change. T'he method developed in NCHR P Report
almost 60 percent of the population of all metropolitan areas
187 is restated. This method interrelates land development
lived outside the central city. A large percentage of people
and its subsequent transportation demands with highway sys-
seem to prefer low-density living. In addition, employment
tem supply and the level of highway transportation service to
opportunities have followed the increase in households in
be provided.
suburban areas. This has resulted in a growing tendency for
The following sections describe the methodology and pro-
people to both live and work in relatively low-density sub-
vide examples to illustrate the various steps involved. An
urban areas. The 1990 census reports that nearly 60 percent
example application is presented at the end of the chapter to
of the work trips made by people who live in the suburbs are
enable the user to execute and become acquainted with the
to the suburbs.
entire methodology. This example provides the specifics of
This basic change in the structure of urban areas has been
accompanied by increased demands for travel by automobile.
computation, definitions of analysis areas, and the like.
Almost three-quarters of work travel is made by people driv-
ing alone in metropolitan areas, and the trends (despite
improvements to transit systems) indicate that auto travel BASIS FOR DEVELOPMENT
will continue to increase in suburban areas. Almost all travel
for purposes other than work in suburban areas is made by The methodology described here is designed ro provide a
automobile. In turn, people are becoming increasingly simple, straightforward means of computing the need for
dependent on the automobile. A major difficulty is that the improved highways based on increasing land-use activities
auto, especially under low-density conditions that force in suburban areas.
lengthy travel, generates the need for substantial investments NCHRP Report IST listed several criteria considered desir-
in the highway system. able in developing such a method, as well as simplifying
New or widened freeways and arterials will be required in assumptions that must be made. The criteria and assumptions
growing suburban areas ifthe level oftransportation service are as follows:
is to remain at acceptable levels. Transportation facilities,
however, are not now and never have been ends in them- 1. Desirable criteria:
selves. It is becoming obvious to decision makers that it will a. An absolute minimum amount of information
no longer be possible to provide an unlimited supply of new would be required.
transportation facilities to meet these travel demands, and b. The terms and concepts would be understandable to
that other altematives must be pursued. Such altematives can citizens and politicians, as well as planners.
include mixed public transportation systems, including taxi, c. The method could be applied quickty and easily so
dial-a-ride, or some other form of flexible-route systems that many altematives could be evaluated.
interfacing with line-haul transit modes. Other alternatives d. No computer would be required.
include ride-sharing modes such as van pooling. These sys- 2. Simplifying assumptions:
tems are being planned and made operational in many urban a. The levels of transportation service being exam-
areas and show promise in reducing the need for new and ined would not so radically depart from today's
improved highway systems. Another method of reducing service levels that travel demand would be altered
travel demands is to locate new development in a manner to significantly.
more fully use available capacity or to place development b. The pricing of transportation service would not so
where capacity can be provided, rather than permitting such radically depart from today's costs that travel
development to overload existing facilities. demands would be altered significantly.
: .l
....t'ii
: . j
124
.,-
i
i
I DATA REQUIRED FOR APPLICATION highways required in an analysis area given a level of land-
i use activity, a freeway system, and a desired level of arterial
,.i
..1 The basic data required consist of two parts: (1) land-use traffic service for that analysis area.
ativity data and (2) data about the highway transportation An estimate of the number of miles of freeway to be pro-
system. If a major investment in transit is to be considered, vided is made outside the procedure, but the method does
some information is needed about that system as well. indicate where such additional facilities would be desirable
The land-use activity data needed are used as the means to to improve the level of transportation service provided.
--l ...
, i generate the amount of highway travel by analysis areas (dis-
...1
tricts). Some experimentation may be required to determine
'ri Limitations of the MethodologY
the size and number of analysis areas to be used. The devel-
'i and Substitutability of Local Data
.;-i opers of the Community Aggregate Planning Model (CAPM)
,. recommend that the size of the basic analysis units range in
"l-l The development density/highway spacing method de-
area from 8 to 30 square miles. scribed is quite similar to the Community Aggregate Plan-
Land use activity data required include as a minimum ning Model (CAPM).I (CAPM is a computer-based model,
. ,.';
not a manual procedure, that is used to generate regional
. Number of households and system-sensitive travel demand, distribute the demand to
. Number of jobs (at-place employment). the freeway and arterial system in each community, and com-
pute a full range of useful evaluation measures') The densityi
As an option, slightly better (more accurate) results may spacing methodology does not, however (as CAPM does)'
be obtained if the household information is subdivided fur- output economic, social, and environmental measures, being
ther into limited in scope to the land use/highway spacing area. But
because the methodology does contain performance mea-
. Number of apartment units, sures (the amount of VMT on freeways, and the arterial level
. Number of townhouse units, and of service distribution), it is possible to produce travel speed
. Number of single-familY units. measures on an areawide basis if the user so desires.
For such a case, it may be useful to express level of ser-
wi Also, the employment information may be divided into vice as a speed, as well as a percentage of VMT, over a spec-
ified level of service. Figure 37 expresses the relationship
. Office employment, between these variables. The curve was constructed for arte-
. Manufacturing employment, rial routes by assuming the level-of-service speeds given in
. Retail employment, and Table 65 and weighting those speeds by the amount of travel
. Other employment. at different levels of service. The daily curve reflects an
assumption of no congestion in the off-peak period.
The existing highway transportation system data needed Estimates of the average speed of travel, along with VMT,
include the number of miles of highway by type by analysis can be used in conjunction with emission rates by speed of
area. Types of highways include travel to provide first-cut estimates of changes in air quality.
Speed of travel may also be used in estimating changes in
. Two-lane arterials and major collectors, operating, accident, and travel-time costs in an area. This
. Four-lane arterials, information can be used in evaluating the cost-effectiveness
. Six-lane arterials, and of alternative program proposals. Because most social, eco-
. Freeways. nomic, and environmental measures require vehicle-miles of
travel and speed as inputs to subsequent calculations of
The method does not deal explicitly with non-line-haul accessibility, mobility, value of travel time, and air quality
.j
transit improvements such as jitneys or dial-a-ride systems. computations, it would be possible to add such output capa-
Existing levels ofconventional bus service resulting in typi- bilities to the density/spacing methodology.
cal levels of suburban transit use are assumed by the method As volumes increase on a facility (the new volumes being
used. Corrections may be made, if desired, to account for output from a traffic assignment), speed declines, and oper-
variations from the typical "mode split" percentage assumed. ating, accident, and time costs (i.e., user costs) increase. At
Corrections for auto-occupancy levels above or below those some point, a new or widened facility, HOV, or greater tran-
assumed may also be made, if desired' sit investment is warranted because the costs of improving
the system are exceeded by the costs in allowing congestion
to continue.
FEATURES AND LIMITATIONS
The development density/highway spacing methodology H. Schleirer, S.L. Zimmerman, and D.S. Gendell, "CAPM-The Community Aggre-
is designed to produce the number of lane-miles of arterial gate Plannng Mode|," Transportatiott Research Record 582' pp I27 (1976).
r.l
'., 'i
..,I
125
Exande;
lf on an artorial, 75% of the V[,T b over Level of Service C, lhen:
- average peak hour speed - 16.8 mph
- average daily speed = 25.5 mph.

g
B,
o
r
E
o
--g{eq--
0102030lO5060700OSt
Percert ofVMT over Levd d Servica C
Figure 37. Arterial level of service speeds by analysis area.
Other limitations revolve about the assumptions made and APPLYING THE DEVELOPMENT
the use of average trip rates and trip lengths. These can be DENSITY/HIGHWAY SPACING
overridden, however, and locally supplied data substituted. METHODOLOGY .,
There probably is no adequate substitute for a complete set
of traffic counts in this regard. Many problems of limitations At least three distinct, potential applications of the density /
in accuracy owing to generalization can be overcome based spacing method exist. The method is an attempt to fill a
on traffic counts and with the use of common sense. critical void in transportation planning-that is, the rapid I ,
TABLE 65 Arterial level of service volumes
Traffic Volumes All Lanes

Approximate
Peak-Hour Two-Lane Four-Lane Six-Lane
Operatng
(mph)
Speed A5- TF-- -F6-- Levet of
Hour" Dailt' Houru Dailyb Houf Dailyb Service
35 <250 <4,150 <800 <8,330 <1,300 21 ,500 A

30 250 4,1s0 800 8,300 1,300 21 ,s00 B
25 375 6,2s0 1,200 10,000 1,9s0 32,s00 c
20 450 7,500 1,440 14,000 2,340 39,000 D
1s s00 8,333 ,600 26,600
1 2,600 43,300 E
10 >500 >8,333 >1 ,600 >26,600 >2,600 >43,300 F
a. one way
b. two-way (peak hour factor (K) = 0.10 and directional factor (D) = 0.60)
126
estimation of the effects of alternative land-use and trans- Step I: Computation of Vehicle Trip Ends
portation plans on the level oftransportation service.
The first application would be Vehicle trip generation rates are based on those given in
Chapter 3, Trip Generation. Two methods (I and II) can be
Given existing land development and existing transportation used, one assuming that only the number of households and
the number of at-placejobs are known by analysis area, and
facilities, what level of service is being provided by the trans-
portation system? the other assuming further breakdowns into type of dwelling
unit and kinds of employment as outlined in the preceding
The density/spacing method would indicate the following section, Data Required for Application. Note that because
key items: trips have both an origin and destination trip end, and
because the procedure involves calculation of trip-end gen-
. Percentage of vehicle-miles of travel accommodated on eration for both residential and nonresidential activities, the
freeways and arterials, sum derived for trips to and from all trip generators will be
. Average volume per lane on the freeway system, and twice the area total number of one-way trips. Accordingly,
. Percentage of arterial vehicle-miles of travel over level the number of trip ends estimated must be cut in half for use
of service C. in computing vehicle-miles of travel.
Method I uses just total households and total employment
A second application would be for each analysis area. Rates for this method are derived from
data given in Table 3, Chapter 3. For example, a trip rate per
Given a future land use plan, what increases in transporta- household of 4.8 one-way vehicle trips daily (9.55 + 2)
tion facilities are required in order to maintain an existing would be used for single-family dwellings, 2.9 for medium-
I
(or desired) level of transportation service? density dwellings, and 3.2 for apartments. Based on the
approximate proportion of areawide single-, medium- and
The density/spacing method would indicate the following high-density dwelling units expected, a single overall rate per
key items: dwelling unit can be computed and used.
Method I uses (for nonresidential activities) an average
. Computation of vehicle-miles of travel on freeways and trip rate per employee derived from a weighted average of
arterials, given a fixed or revised freeway system; rates for individual employment categories. For example, if
. Number of equivalent lane-miles of arterials that need to the proportion oftotaljobs in a study area were 21.5 percent
be added, either by widening existing routes or by adding for office employment, 18.5 percent for retail, 10.0 percent
new construction, to achieve an existing (or desired) for manufacturing, 23.0 percent for military, and 27.0 per-
level of service; and cent for other, and the trip rates were, respectively, 1.15, L0,
. Construction of enlarged or new freeways or improved 1.5, 1.25, and 5 one-way vehicle trips daily, the weighted
transit service to reduce the need for arterial improve- average daily vehicle trip rate per job would be 4.0. This
ments. average trip rate is applied to all analysis areas.
Method II uses these rates directly by type of residential
A third possible application would be unit for each analysis area rather than develops the single
overall rate previously described. In this case, a breakdown
Given an existing orfuture transportation plan, what amount by type ofunit is needed for each analysis area.
of land development can be addedwithout allowing the level For nonresidential activities, trip-generation rates can be
of trffic service to deteriorate below a specified level? expressed as functions of at-place employment, floor space,
or acres as given in the ITE Trip Generation manual. Again,
This third application is best accommodated through a trial- vehicle trips per day are used, but reduced by half to reflect
and-error process, successively increasing (or reducing) the the one-way nature of travel. The best measure, if available,
amount of land development by analysis area until the level- is employment, as this can be summed to a control total for
of-service limitation is reached. Because the technique can the area as a check.
be applied rapidly, many iterations can be made in a reason- Method II applies individual rates to each land-use or
able time. The effect of freeways and additional transit ser- employment category for each analysis area, thereby requir-
vice can also be taken into account. ing more detailed input information than Method I.
Steps in Application Step 2: Computation of Transit Use

and Auto - O c cup ancy Adj ustment
Applying the density/spacing method requires undertak-
ing the following steps for each analysis area in the study In some urban areas, particularly larger ones, transit
area of interest. improvements may be planned which could have significant
12-7
impacts on future vehicle-miles of travel within the area. Enmola:

- @ I .4 aub occuparrcy, eub driver tips = 7196
Figure 38 shows the effect of changes in the percentage use - @ 1.5 aub occupency, stb driver tipe = 5795
of transit on the percentage of auto driver trips (of total per- - refore, the percent charo = (67 -711 I 71 x f( = 6.7% reduct-on in
son trips), and hence on VMT. For example, if an analysis aub driver ipe.
area had a percentage transit use of 6 percent, and this could
be increased to I 5 percent, the percentage of auto-driver trips
would drop (given an auto-occupancy rate of 1.33) from 70
to 64 percent. This represents a change of 8.6 percent, assum-
ing that total travel would remain constant. Where changes
.$
in the relative use of transit are contemplated, this curve can F

estimate the effect on auto use and VMT. In addition to this 2

curve, local relationships can be used (or derived) to estimate

E
changes in transit use. <g
I
Changes in auto occupancy also affect vehicular travel and E
VI\.4T Fiorrre i9 shorx.s fhe nercentace chnnse in auto-driver
"_-^^- __
I

trips as vehicle occupancy increases. Again, reductions in a
o
vehicular miles of travel can be computed for various increases tt
in auto occupancy using a method similar to that illustrated F
for transit increases. This is particularly applicable where I

o
l
Enmple:
Given an alb Gupency rato of 1.33 persons/aub, then
- @ 6% transt use, auto driverips = 7l%
- @ 15% Fansit use, aub driwr tips = e[% 'r.0 r.t 1.2 r.3 1.7 r.8 r.0 20
- therefore, the percent chang = (64 - 7l ) / 71 x 1(X) = 9.6% li.#r,l;i
reductir in ar.rto driver ips.
Figure 39. Effect of change in auto occupancy on auto
lf the aub rpncy reb wr6 b concuneny ncreas ftom 1 .3Xl b 1.5O
persons/auto, then: driver trips.
- @ 6% t'ansit uss and 1.3l persons/ao, aub drivar bips = 719
- @ 15% tanit use and 1.5() Person6/aub, euto drivor ips = 5795
- therefore, the percontchang= (57 -711 171 x 1l = 19.8% reduc{ion vigorous carpooling and vanpooling programs are planned.
in aub driver bipe.
Local area experience can be used as required. Present daily
car occupancies range from about L 1 for HBV/ trips to 1.7
for home-based social and recreation trips.
Adjustments for transit use and auto occupancy are
optional, but the use of high-occupancy vehicles can affect
freeway and arterial highway requirements and should be
.$ eo
l- considered, if warranted.
E
o

E*
E
Step 3: Computation of Vehicle-Miles of Travel

o
o
f The third step is to compute vehicle-miles of travel for

t each analysis area by multiplying the results of Step 1 (i.e.,
o
t l.) Person/Arb vehicle-trip ends) by the areawide, average over-the-road
t- 1.33 Persons/Aub vehicular trip length. The average work-trip distance can be
E \\-\- 1.50 Persons/Atb
obtained from the census sample data for individual MSAs.
t.eo Person/Ab
o
Alternatively, the airline trip distance can be estimated from
t Figure 40. The data points shown were collected in the 1960s
and the average trip lengths adjusted upward to account for
increased speeds and lower densities of development in
0 urbanized areas since these data were obtained. For future
lo 20 cr (} 50 gl m E fll t(x,
Trrlt tla * Pmnl of Tolil PolH Trlpc years, an estimate of such corections to be applied can be
obtained from Figure 41 for home-based work trips.
Figure 38. Effect ofchange in transit use on auto driver To illustrate the use of Figure 41, suppose that the aver-
trips. age network speed change is * l0 percent over the base-year
.....i
r1
.'
,l
-, :'l.'] 128
'';
'' .i .i
Example: Example:
:l
.'i For a clty of 3,000,000 population, the average vehicle trip distance Assuming that the average network speed change is +10 percent for a
approximately 4.8 airline miles. region. (Revised/Base Speed ='1.10.)
....1
Then, the average auto HBW trlp distance change is glven by (Revised/
Base Distance) = 1.15. That is, the change in average trip distance is
I +15 percent.
.l
:* -j
,.'.1
...i 1.1
,l
'
':',|
],.J
'1 'I

I
E

$
'ji _
..
''' .: ...r i
;0

!E t't

a
6
c
'E
""""""""'-";l E
o
F
t I e
.l
6
1
'-/'' =l
!
T

.
t-
n.
o

g
T
Ur{ntsd Arr. Polron (Mllllon)
0.t L
Figure 40.
Average vehicle trip distance (airline) vs. 0.t 0.0 I 1.r 1
Arrg Nrt Spd Crng (RfllEd/Bq)
urbanized area population
Figure 41. Adjustment to average auto home-based work
trip distance (arline) for average network speed change.
conditions, that is, S2/S1 :
1.10. Entering Figure 41, the
average auto HBW trip distance (airline) change would be
given by LJL1 : 1.15, that is, a change in trip distance of vehicle-trip ends) to compute vehicle-miles of travel by
* 15 percent. analysis area.
If the average trip distance for a study area is known, it Step 3A: Computation of external vehicle-miles of travel
should be used. If the average HBO trip distance is estimated adjustment,In addition to the vehicle-miles of travel generated
using Figure 42 (which relates the length of HBW and HBO by the residential and nonresidential activities within each
trips), then by weighting the trip lengths by the amount of analysis area, an adjustment has to be made for traffic gener-
HBV/ and HBO travel, one can obtain an estimate of total trip atedbeyondthe boundaries of the study area. This adjustment,
length. For example, in Washington, D.C., the average HBW however, need only be applied to those analysis areas located
airline trip length is then 8.0 mi. Table 9 (Chapter 3) indicates on the periphery of the study area; here, the "external" traffic
that about 2l percent of all trips are for work purposes. contributes significantly to the VMT calculated from Step 2.
Therefore, the daily weighted average is [(8.0 X 0.21) + This correction was deemed necessary through empirical test-
(4.0 x 0.79)1, or 4.84 airline miles. Note that Figure 40 ing of the density/spacing methodology.
shows that the result for a city of just over 3 million popula- This extemal traffic is obtained from counts located at the
tion is approximately 4.8 airline miles, thus confirming the circumference (cordon) of the analysis area in question. Note
aforementioned results. that the count must first be adjusted to account for the double-
As these figures represent airline distance travel, they need counting ofthrough trips. Should any ofthe count stations be
to be expanded to over-the-road trip distances by multiply- located at a freeway, such counts must be excluded altogether
ing by a circuity factor. This factor can range from 1.2 to I.4 from the adjustment process. This is based also on empirical
(or even higher) depending on the configuration of the high- evidence gathered through testing of the density/spacing
way network in the urbanized arca. The presence of rivers method.
or topographic barriers cause higher values. Thus, having Because some of the external trips at non-freeway cordon
obtained the areawide average over-the-road trip distance, locations are already reflected at one end in the peripheral
this figure is then multiplied by the results of Step I (i.e., analysis areas, they should be reduced by half. The result is
129
_e
t
=
=
oo
I
G
HBO Distance = 1.7O + 0.30 x HBW Distance

// :"'e"-='
o e
tl
6
^ '---'/
5
o e

Affig.Ao HBVlrTdp Dkar (Arth. M{-)
Figure 42. Relationship between auto home-based work nd auto home-based non-work
trip distances (airline ).2
then multiplied by the average trip length computed from where

Step 3 to arrive at the external VMT. Then, this VMT owing
to the external trips is added to the VMT calculated from Step
V1: average daily traffic on freeway;
P: average daily vehicle trip origins/square mile,
3 for the peripheral analysis area to obtain the total VMT.
7: average vehicle trip distance (mi),
Z1 : freeway spacing (mi),
Step 4: Computation of Vehicle-Miles of Travel 4: arterial spacing (mi), and
on Freeways and Arterials Zt : local street spacing (mi).
The vehicle-miles of travel computed through application This relationship can be used to solve for freeway spacing
of the previous steps must be accommodated by three levels if desirable freeway traffic volumes are known. Then in
of highway transportation systems; that is, freeways, arteri- solving for 21, and approximating 23 at 0 to simplify the
als, and major collectors. The amount of travel that will need solution
to be accommodated on the freeway system is a function of
the spacing between freeways, the spacing of arterial and t. :
2U(7 + 22) (tt-2)
local routes, the average trip length, and the average vehicle- Pr2 - 2v,
trip density.3
The relationship can be expressed as follows Figure 43 shows desirable freeway spacing based on this
relationship for a 6-mile average trip length. Thus, for exam-
,t-M
r/- P7
(l r-1)
-
ple, for a daily vehicle trip origin density of 14,000 trip ends/
square mile, a six-lane freeway must be spaced at 4.6 mi,
LZ, 7 7(Zt - Zr) ) and an eight-lane freeway at 7.8 miles.
Given the information requireda for the relationships pre-
viously described, either freeway volumes or spacing can
rWilbur Smith and Associates, Transportatiotl and Parking
fitr Totnorrow's Ciries,
prepared under commission from the Automobile Manufactuers Association, New
Haven, Connecticut ( 1969).
3M. Schneider, "A Direct Approach to Traftc Assgnment," Highway Research rSpacing, Z can easily be computed from the formula Z
= 2A + L, where L is the
Record 6, pp7 l-75 (1963). number of mies of route within an area A in square miles.
r30
Freeway Volumes:
4-Lanes 4-Lane 66,700 VPD
6-Lane 100,000 VPD
8-Lane 133,000 VPD
15
6-Lanes Peak hour factor

(K) = 0.10
3
=o Directional factor
(D) = 0.60
E
8. 10 _
U'
o
G I
I
8-Lanes
o
\/l
I
\,/ l
Average Daily Vehicle Trip Origins/Square Mile
l
Figure 43. Freeway spacing vs. average daily vehicle trip origins per
square mile.
be computed. If a specific level of service is desired, either Step 5: Computation of Average Arterial Volumes
freeway volumes, spacing, or miles of freeway can be com- Per Lane and Level of Service
puted. Similarly, if spacing is set (i.e., no new routes are con-
templated), then the traffic volume on freeway facilities (and For uniform trip distributions and arterial loadings, traffic
hence VMT) can be computed. demand can be expressed as a function of arterial grid spac-
Subtracting this freeway VMT from the total gives the ing and traffic volume. The relationship between traffic
residual VMT that must be accommodated on arterial and demand, arterial grid spacing, and traffic volume is given by
local streets. After subtracting a percentage of the total VMT the equation:
for local streets, the residual is the VMT on arterial routes.
Thus, for any analysis area other than a peripheral analy-
sis area (i.e., not at the boundary of the metropolitan region),
P=?v
s (1 1-3)
the arterial VMT is given by
where
Arteral VMT = (residential + nonresidential) VMT
- freeway VMT - local VMT D: the arterial vehicle-miles of travel per square mile,
V: the average daily traffic volume (VMT per mile of
route), and
For a peripheral analysis area (i.e., at the boundary of the S = the distance between adjacent arterials in miles
metropolitan region), the arterial VMT is given by (spacing).
AtterialVMT = (residential + nonresidential) VMT Volume (VMT per mile of route), although a useful indica-
* external VMT - freeway VMT tor, is not as useful as volume per lane, because urban and
- localVMT suburban areas have a mix of two-, four-, and six-lane arter-
ial facilities.
To compute freeway volumes or spacings, areas larger Table 65 gives the level-of-service volumes of different
than the analysis areas (such as a subarea; i.e., a group of dis- arterial facilities. It should be noted that two-lane arterials
tricts) used for arterials should be described. For example, have a significantly lower service volume per lane than a
areas on both sides of a freeway should be included. multi-lane arterial at LOS C.
131
A better method, and one that is used in the example pro- Determination of the number of equivalent arterial lane-
vided at the end of this chapter, is as follows: miles of travel by district through the method illustrated can
be used with these guidelines to aid in the design of an arte-
. The Equivalent-lnne Concept. To relate traffic demands rial highway system for an urbanized area. The next step is
on different size arterial routes on an equal basis, each to convert the number of equivalent lanes to miles of 2-, 4-,
lane of a 4-lane arterial is set equal to 1.6 lanes of a or 6-lane facilities and space them as desired. Widening
2-lane arterial, and each lane of a 6-lane afterial is set existing routes, providing HOV, or increasing transit should
equal to 1.73 lanes of a 2-lane arterial. Average volumes also be considered as appropriate.
per equivalent lane are then computed by analysis area
(i.e., arterial VMT divided by equivalent lane-miles) and
Feedback
related to the level of service provided to the analysis area.
. Arterial Level of Servlc. Because it is not possible to The process described also can be used to modify the lane-
calculate the traffic volume on each segment of each miles of arterial routes needed by subareas. The planner can
arterial (only an average volume can be calculated), a reverse the process or "feedback" to prior steps by modify-
relationship was developed from traffic count data ina innrrrc ac lpcirprl The follnr-rino nntions are aVailable:
^l/u!o
between the average equivalent-lane volume and the
percentage of all VMT in the analysis area operating 1. Revise the level of service desired. By accepting a
above levels of service C, D, and E. This relationship higher percentage of vehicle-miles of travel over a set
was derived from complete count data for Fairfax level of service, the number of equivalent lanes can be
County, Virginia, and is displayed on Figure 44. In addi- reduced because a higher average volume per equiva-
tion, another relationship was derived relating the per- lent lane can be accommodated.
centage of route-miles over specified levels of service to 2. Add capacity on freeways. Adding high+ype limited-
the percentage of VMT over such levels of service as access facilities or increasing capacity on such facili-
displayed on Figure 45. These relationships may be used ties in areas of high travel demands can reduce the
to measure the level of arterial service provided. volume on arterials.
3. Increase the use of transit or increase auto occupancy for
the analysis area. See the discussion on TDM measures.
Example:
lf for an anaia area, the arterial VMT has been computed at 75,000 and
he equlvalent arterlal lane-miles = 15, then equivalent lane volume =
Example:
75,000/15 - 5,000 VPD. Hence:
From Flgure 44, percent VMT over Level of SeMce C - 93o/o.
Percent VMT over Level of SeMce C . 93% Thereforc, percent roue milee over Level of SeMce C 86%. '
Pecent VMT over Level of SeMce D . 80%
Percent VMT over Level of Sevlce E - 66%
I
o
oo
'
b
th
o
!
3
T
=
fo
.
E
e
r
b

m
o 20 & t0 lcx
Equlvant Lne Volume (Vehktor por Day) Pcrc.nt VlrtT ovor Lcll of serv. C
Figure 44. Arterial VMT level of service vs. equivalent Figure 45. ReLationship between level of service and
lane volumes. route miles.
,.. --l. ri
.l
'' ,J
,'
:l
.j 132
'i'i
:i
'I
4. Reduce or reallocate land use. Reductions or realloca- transportation supply in the subarea (and districts within the
r.'. .li.l tions of land use will reduce travel demands in areas subarea), the objective is to determine the level of service at
that have low levels of service. which the transportation system will operate for that future
year. It is anticipated that improved transit and carpool pro-
Testing of various options, singularly or in combination, grams are to be put in effect. Concurrently, the highway net-
is most useful in designing or evaluating a land use or trans- work itself will undergo traffic-flow improvements resulting
portation plan. This can be accomplished in a very short time in increased average speeds.
using the relationships developed. This example describes the use of the development-density/
The following section provides an illustrative example to highway-spacing methodology. Computation steps similar to
enable the user to apply the development densityrighway those outlined here must be executed for all districts of inter-
spacing methodology described. est in a real application.
AN EXAMPLE APPLICATION lnput lnformation
Suppose the transportation service in a subarea in the Assume that the following input data are available for the
northeast quadrant of a hypothetical metropolitan region is to metropolitan region, the study subarea, and the peripherally
be analyzed for some future year. More specifically, given located study district 21. Except where noted (and where
the projected land development density and the projected inappropriate), these input data represent the future condition.
1. Existing population of the metropolitan region 1,200,000
2. Area of the study subarea (sq mi) 55
3. Area of study district 21 (sq mi) I

Residential development in district 21 :
Single{amily units (at 1 DU/acre) 5,100
ry"lT Townhouse units 2,000
Apartment units 500
-'i
Total number of dwelling units 7,600
.
'l
'-.^;,-i Nonresidential development in district 21 :
General office (sq. tt. cFA) 100,000
lndustrial park (sq. ft. GFA) 100,000
Transit use in district 21:

Existing (transit as a percent of total person trips) 10
Future (transit as a percent of total person trips) 20
Auto-occupancy rates in district 21 :

Existing (persons/auto) 1.5
Future (persons/auto) 1.6
Average network speeds in melropolitan region;

Existing (mph) 26
Future (mph) 29
Daily through and external traffic volumes at the e)ternal counl

stations (excluding freeway volumes) 20,000 (vehicles)
10. Facility mileage:
For district 21
2-lane arterials (mi) I
4-lane arterials (mi) 3
6-lane arterials (mi) 12
All arterials (mi) 23

Freeways (mi) 6
For study subarea

All arterials (including district 21) (mi) 86
Freeways (including district 21) (mi) 18
133
Methodology of the average airline trip distance with respect to urbanized

area population; that is, if such a measure is not available
The density/spacing methodology is applied in a step-by- from local information. For an urbanized area of 1,200,000
step manner as discussed in the preceding sections. existing population, Figure 40 shows that an average vehicle-
trip distance of approximately 4.0 airline miles is the current
measure of trip length.
Step I: Compute Vehicle-Trip Ends
Because it is expected that average network speeds will
Using the average daily vehicle-trip generation rates given change in the future (in the ratio 29/26; i.e., I.l2), Figure 41
in Table 3 (Chapter 3) and Method II described previously, provides the means for estimating the corresponding changes
the future one-way vehicle-trip ends for the residential and in HBW and HBO average airline trip distance.
nonresidential development in district 2l are computed as From Figure 41:
follows for
Adjustment factor for HBW auto trips : 1.18
Single-family unirs = yr(9.55 x 5,100) = 24,353 trips
Townhouse units = %(5.86 x 2,000) = 5,860 trips Therefore,
Apartment units = %(6.47 x 500) = 1,618 trips
Adjusted HBW average auto airline-trip distance : 4X
General offices : %(11.85 x 100) = 593 trips
1.18: 4.l2mi
Industrial park = yr(6.97 x 100) = 349 trips
Adjusted HBO average auto airline-trip distance : 3.1 1 mi
Hence, the total one-way vehicle trips generated daily by the
(from Figure 42)
development in district 2l is given by 32,771vehicle trips.
Note that Table 3 provides vehicle trips to and from the gen-
erators (i.e., vehicle-trip ends); consequently, such trips must Assuming that work trips will constitute 25 percent of all
be halved as shown previously to obtain the one-way trips. trips, then:
V/eighted auto airline trip distance : l(4.12 X 0.25) +

Step 2: Compute Transit-Use (3.11 x 0.75)l mi : 3.51 mi
and Auto - O ccup ancy Adj ustme nt s
Assuming a circuity factor of 1.22:
Figure 38 can be used to adjust the daily vehicle trips out-
put from Step 1 for future improvements in transit and car-
pooling programs. Thus Average auto over-the-road trip distance for the metropoli-
tan region : 3.51 x 1.22 : 4.28 mi
For the existing condition, at 10 percent transit use and 1.5
persons/auto, auto-driver trips as a percent of total person- Average daily internal-internal vehicle miles of travel for
trips : 607o district 2l :27,309 x 4.29: 117,100 VMT
For the future condition, at 20 percent transit use and 1.6 Step 3A: Compute External Vehicle-Miles of Travel
persons/auto, auto driver trips as a percent of total person- Adjustment. District 21 is located at the periphery of the
trips : 507o metropolitan region, and, therefore, a significant amount of
traffic within its boundary can be attributed to the traffic
Therefore, percent reduction in auto-driver trips = that has origins or destinations external to that district. This
external traffic contribution must be added to the VMT
50-60 x from Step 3. Note that this addition of external traffic is
60
100 = -l6.ivo
made to traffic in districts that are peripheral to the metro-
politan region. It is assumed that, for the internally located
Therefore, adjusted daily vehicle-trips : 32,111 (l - 0.167) districts, the external traffic contribution is small, because
: 21,309 most of these trips will have "dropped off" in the peripheral
districts.
These trips represent the future internal-internal daily vehi- For district 21, daily volume of traffic at the non-freeway
cle trips in district 21. external count stations is 20,000 vehicles. This volume
includes through trips, which must first be accounted for.
Step 3: Compute Vehicle-Miles of Travel Table 16, Chapter 5, shows that for an urban area with a pop-
ulation of 1,200,000, approximately l3 percent of external
Before computing VMT, the average over-the-road trip vehicle trips are through trips. Reference to the example for
distance must be calculated. Figure 40 enables the estimation the conversion ofcordon counts to external trips shows that
134
Cordon Count 2
External trips = Arterial spacing Z, = :86 -55 = 1'3 n-ti
(1 + Proportion of Through Trips)
Local spacin Zz =
Hence 0 mi (assumptionfor ease of calculation)
So, using equation 1l-1, average daily traffic on the free-

External trips = l9'9rq- = (t,'Toovehicles ways, V1, in the subarea is given by
ll+
\ 100/
I
945,600
Becausemanyofthesetripsareaccountedforbythepopu-I1=#=I9,553vehicles
lation and employment trip estimates made for the districts, |
: + - + -*
L6'l 4'4 4'4(6'I) )
I
the external trips are reduced by one-half.

Therefore Note that 953,700 + 55 (i.e., average daily total VMT for
subarea + area of subarea) is the numerator Pr in Equation
Average daily external vehicle-miles of travel : ,/r(I7
,700) x I 1- l . Also, 4.4 is the average auto over-the-road trip distance
4.29 mi: 37,900 VMT (miles) for the metropolitan region.
Average daily total VMT in district 21 : 117,100 4 37 ,900 : For the study subarea' then:
155'000 vMT
Freeway vMT = v, x freeway mileage = 19,553 x 18
= 352,000 VMT
At this point, the planner must be reminded that Steps I
--l ^,- all
through 3 must be accomplished for
L^ ^:-L
^li the ^..-l-.
eight study dis-
J:^
Therefore, since
tricts in the study subarea. For illustrative purposes, assume
the following daily total VMTs have been calculated for the Arterial VMT = subarea total VMT - freeway VMT
*-,..1 eight districts. - locql VMT
CalculatedAverage Percent Subarea

District Number Daily Total VMT Total VMT
18- 120,200 12.7%

* 135,700 14.4
19
20 96,800 10.2
21 * 155,000 16.4
22 1 10,900 11.7
23 80,1 00 8.5
24 95,000 10.0
25' 151 ,900 16.1
Subarea total VMT 945,600 100%
* Peripheral districts requiring addition of external station VMT.
, Arterials and if it is assumed that 10 percent of all subarea VMT is on

-i ,
Step 4: Compute VMT on Freeways and
: local streets, then
To calculate arterial VMT in district 21, first obtain the
1, ,' freeway VMT in the entire study subarea by employing AfierialVMT = 945,600 - 352,000
..:r
-. Equation 11-2. This equation requires measures for freeway - (0.10 x 94500 VMT) = 499,000
spacing (Zt mr), arterial spacing (Zzmi), and local spacing
- - :, as ^--r--'
(Zt mi) *" input. Thus, for the whole subarea:
This subarea arterial VMT can then be distributed to each of
,., , Freeway spacing, the eight districts within the study subarea in proportion to
_ 2A 2 (study subarea\ 2 x 55 the distribution of the total VMT (derived earlier). Then, the
' ' L ' l'''
Z'l=-' ''') freeway mileage 18 ='.'' arterial VMT by district is as follows:
). . '
;
r35
District % Subarea Proportioned

Number Total VMT Arterial VM
18 12.7"/o 63,400
19 14.4 71 ,600
20 10.2 51 ,100
21 16.4 81 ,800
22 11.7 s8,500
23 8.5 42,300
24 10.0 50,100
25 16.1 80,200
Subarea Arterial VMT 100.0% 499,000
Thus for study district 21,the average daily arterial VMT that tt:t=oro
can be expected is 82,100 VMT. = 33.56
= 2137 vehicles/daY
For 2,437 average daily vehicles per equivalent lane, the

Step 5: Compute Average Arterial Volumes following level-of-service results can be obtained (for dis-
Per Lane and Level of Service
trict 2L):
To determine the level of service provided by the arterial . VMT over Level-of-Service C : 5IVo
network in district 2L,the equivalent lane-miles must first be . VMT over Level-of-Service D : 4270
calculated. (See previous section in this chapter, The Equiv-
. VMT over Level-of-Service E : 36Vo
alent Lane Concept.) Now,
Next, Figure 45 can be entered to determine the percent of arte-

Equivalent lane-miles = 1.00 (Z-lane arterial mileage)
+ 1.60 (4-lane arteriolmileage) rial route-miles operating over level-of-service C. Hence, 37
+ 1.73 (6-lane arterial mileage) percent of the aterial mileage in District 2l can be expected
to operate over level-of-service C.
Therefore
Equivatent tane- mite s t

y?%: :ffSl . t.i 3(r2) Output lnformation
= It can be seen that the development density/fiighway spac-

Therefore ing methodology, as applied previously, yields the freeway
VMT, the arterial VMT, and the level of service provided by
Equivalent arteriql lane volume = the transportation network in district 21 for some future year.
Average daily arterial VMT The remaining seven districts in this study subarea can be
Equiv alent art e rial lane - mile s similarly analyzed.
136
_..r j I
.-, : i
'..t .t-i
CHAPTER 12
CASE STUDY APPLICATION OF DEFAULT PARAMETERS
INTRODUCTION Socioeconomic Data
Throughout this report we have used a case study to illus- The trip generation equations require socioeconomic and
trate the application of the parameters and techniques de- land-use data to describe the quantity and type oftravel activ-
scribed in various chapters of the report. The data included ity in the region. Required land-use data include the follow-
in this case study were provided by the State of North Car- ing: number of households by size; household income or auto
olina for the City of Asheville, North Carolina. The applica- ownership; and employment by type. These data should be
tions of the study parameters and techniques introduced in allocated throughout the region according to a traffic analy-
't Chapters I through t have been applied to this case study and sis zone (TAZ) structure that is appropriate for the level of
presented at the conclusions of the chapters. The case study analysis and the detail in the selected network.
'--.'-l has allowed the user to follow the development and applica- The North Carolina Department of Transportation
tion of the travel forecasting model beginning with the data (NCDOT) originally developed the Asheville TAZs. Their
collection phase. Subsequent chapters followed the model zone system contains 353 intemal zones and 36 external sta-
tions. NCDOT compiled household and employment data for
''
development process through trip generation, trip distribu-

lgMi tion, and, ultimately, the final traffic assignment. the MSA at the 353 TAZ level of detail. A review of the
This final chapter of the report presents the case study in NCDOT zone structure revealed that it provided more de-
its entirety, from data collection through traffic assignment. tail than was needed for the case study example, which is
The material in this chapter is presented in the order in which intended to forecast traffic volumes only on the major road-
it should be collected, processed, and analyzed. ways. The 353 zones were aggregated into 107 intemal zones
and 16 external stations for the case study. Figure 47 displays
the revised zone structure with 107 internal TAZs. This zone
system is somewhat coarse but should be sufficient for fore-
TRANSPORTATION DATABASE casting traffic volumes on the major roadways.
The socioeconomic data for this case study were extracted
In Chapter 2, we introduced the transportation database for from the 1990 U.S. Census. Serial Tape File (STF) 3-A was
the Asheville, North Carolina, test case. The Asheville Metro- downloaded from the North Carolina State Library. These
politan Statistical Area (MSA) lies in the westem quarter of data included the following summary information:
the state of North Carolina, roughly 230 miles west of the
state capital in Raleigh and 1 l0 miles east of Knoxville, Ten- . Household income by household size,
nessee. As defined by the U.S. Census Bureau, the region con- . Median household income,
sists of the City of Asheville and surrounding Buncombe . Number of persons in household,
County. In 1990, the entire MSA had a population of 174,821, . Mode of travel to work,
and the City of Asheville had a population of 170,429. . Time of departure from work, and
Figure 46 displays a base map for the Asheville region. . Private vehicle occupancy for work trips.
Asheville's transportation network consists primarily of its
roadway system and the City Coach bus service, a twelve These census data revealed that the study area contains
route transit system operated by the Asheville Transit Author- a population of approximately 110,000 persons in 46,492
ity. Two U.S. Interstate routes meet just south of the City of households. The household data for the 107 internal zones of
Asheville: I-40, running east to west, andl-26, which connects the Asheville study area are tabulated in Appendix B-1.
Asheville to Atlanta. A belt-line, I-240, makes a half-circle The household and income data were collected for use in
around the City to its northern side. As of 1991, City Coach the trip generation equations to calculate the person-trip
carried fewer than 4,000 daily passengers, approximately I productions. Person-trip attractions, on the other hand, are
percent of the daily person trips made in the Asheville area. based on employment data, stratified by type of employ-
The remainder of the daily person trips on the transportation ment. The NCDOT data provided for this study were div-
network were made in private vehicles. ided into several different sectors according to the Standard
137
z
o
ru Bingham 9l\ t c
3 xeights ti 4
u Y
t u
U
q

U
o
a
.o
o
-9 7-,:::::-:- ts
I
t
:<
o
ts
c Fomers'.-\ I
q ttr. \-."<Ce-

E
I '--'- Siltator Hltc F
6
()
Fo
6 ts
8ilroc '.i
li Blttmoro t
Forest ;
Yrlhy Spd4r
Figure 46. Asheville, North Carolina, map.

138
model the transit volumes on the local bus routes, we would

have used the highway network as the base for building a
transit network. However, given the fact that only 1 percent
of the person trips in the Asheville MSA use transit, that
component of the transportation network was ignored.)
..- 1.
Figure 48 is a plot of the coded base highway network,
including all link types except centroid connectors.
Traffic Count Data
NCDOT also provided a selected count map, which dis-

plays total daily vehicle volumes for a variety of intersections
and cordon points. These counts, summarized in Table 67,
were used to provide average daily traffic (ADT) at the 16
facilities crossing the external cordon around the Asheville
region. These ADT counts were collected at all locations
where significant traffic volumes flow into or out of the
region.
'l
TRIP GENERATION
i
: ---i
The techniques described in Chapter 3 were followed to
:_. obtain trip productions and attractions.
.--..=
Trip Productions
The estimation of trip productions using disaggregate

travel demand models typically uses a cross-classification of
Figure 47. Asheville, North Carolina, TAZs. household size data with a measure of wealth, such as income
or the number of automobiles available to the household. For
this case study, however, no cross-tabulations were yet avail-
able from the Census Transportation Planning Package for the
Industrial Classification land use code. Total employment for
Asheville region. As a result, trip production rates were cal-
the MSA is estimated at 59,031 by NCDOT's surveys. The
culated by using the average values for the region, stratified
employment data for the 107 internal zones of the Asheville
only by household size. The average daily vehicle trips per
study area are tabulated in Appendix B-2.
household were calculated using the rates shown in Table 68.
Since the trip productions are classified according to trip
Network Description purpose, the information regarding trip purpose by house-
hold size was also used. Table 69 lists the percent of trips by
The highway network for the region was acquired from purpose and household size.
NCDOT as a downloaded file on disk, which was then input The calculation of trip productions was performed using a
directly into the transportation modeling software package. computer spreadsheet. The input data were arranged in flve
Each link in the network was then coded with a facility clas- columns reflecting the household size groupings from one-
sification (freeway, major arterial, or minor arterial), number person households through five-plus-person households. The
of exclusive lanes in the direction of travel, the free-flow output data are the three columns reflecting the trip purposes-
speed, and the hourly per-lane capacity ofthe roadway. home-based work, home-based other, and non-home-based.
.!i
As summarized in Table 66, the base highway network The person trip calculations are expressed by the following
consists of 1,156 regular nodes connected by 168 freeway formulas:
links,922 major arterial links, and 1,550 minor arterial links.
After the regular nodes were connected by links defined to Home-Based Work Productions =
represent actual highways, centroid connectors were added 0.20 x 3.7 x HH(I)+ 0.22x7.6x HH(z)
to allow the 107 internal centroids and the 16 external sta- + 0.19 x 10.6 x HH(3) + 0.19 x 13.6 x HH(4)
tions access to the highway network. (If we were planning to + 0.17 x 16.6 x HH(5)
t39
TABLE 6 Base network summary
Nodes Quantty
Zone Centroids (lnternal) 107

Zone Centroids (Elrternal) 16
Regular Nodes 1 ,156
Total Nodes 1,279
Free-Flow Capacity Veh.

Link Type Descripton Quantity Speed (mph) per Hr.
1 Freeway 168 55 1 350

2 Major Arterial 922 45 825
3 Minor Arterial 1,550 35
4 Cntroid Collector 632 20 ',.u
Total Links 3,272
Home-Based Other Productions = culating the number of attractions for each TAZ, depending
0.54 x 3.7 x HH(I) + 0.54 x1.6 x HH(z) on whether the zone is within the CBD. In the City of
+ 0.56 x 0.6 x HH(3) + 0.58 x 13.6 x HH(4) Asheville, zones I through 15 are considered to be within the
+ 0.62 x 16.6 x HH(5) CBD, and the remaining zones (from 16 through 107) are
Non-Home Based Work Productions = considered to be in the non-CBD category.
0.26 x 3.7 x HH(I) + 0.24 x 7.6 x HH(z) The trip attraction rates for CBD zones I through 15 were
+ 0.25 x 10.6 x HH(3) + 0.23 x 13.6 x HH(4) calculated using the following equations:
+ O.2l x 6.6 x HH(s+)
Home-Based Other Attractions = 2.0 x RE + 1.7 x SE
where
+0.5xOE+0.9xTH
Non-Home Based Attractions = t.4 x RE + 1.2 x SE
HH(n) : the number of households with n occupants. +0.5xOE+0.5xTH
These equations were used to calculate the trip produc- The trip attraction rates for non-CBD zones 16 through
tions for each of the 107 internal zones in the Asheville 107 were calculated using the following equations:
MSA. The total number of trip productions estimated for the
region is 383,006, which includes 76,033 HBW productions, Home-Based Other Attractions = 9.0 x RE + 1.7 x SE
215,407 HBO productions, and 9I,566 NHB productions. +0.5xOE+0.9xTH
Non-Home BasedAttractions = 4.1 x RE + I.2 x SE
+0.5xOE+0.5xTH
Trip Attractions
where
Trip attractions were also calculated on a spreadsheet
using the parameters from Table 8. RE: retail employment;
The input data for these calculations include the employ- SE : service employment;
ment by type-specifically retail, service, and other employ- OE: other employment; and
ment-and the total households for each of the 107 TAZs TH: tol households.
within the study area.
The home-based work trip attractions for all 107 internal For all internal zones, the trip attractions in the region
zones were calculated using the following equation: totaled 383,741, of which 85,604 were HBry' trips, 188,806
were HBO trips, and 109,331 were NHB trips. These totals
Home- Based Work Attractions = L45 x Total Employment reflect the unbalanced attractions before they are matched to
the productions in the region. Appendix Table B-3 lists the
For the other two trip purposes, home-based other and productions and attractions for the three trip purposes for
non-home-based, two different equations were used for cal- each of the 107 internal TAZs.
_. : I
.'
140
trips, which have one trip-end within the region and one trip-
end outside of the region. The external-internal trips were
converted to person trip-ends and incorporated into the
regional trip generation model, while the external-external
trips were expressed as a separate vehicle-trip table that was
added to the other vehicle-trip tables before assignment.
r ':. i
The procedures used to estimate external travel for the
1i Asheville case study are listed below. All of the calcula-
'tl tions were performed with the aid of a computer spread-
.l sheet program.
:, i
Classification of External Stations
ADT counts were collected at all locations where signifi-

cant traffic volumes flow into or out of the Asheville region.
Each of the 16 external stations selected were classified as
either a minor arterial, a principal arterial, or an interstate
facility. In addition, continuous facilities were noted forpairs
of external stations that would be expected to carry a statis-
tically significant share of external-external traffic. The most
notable continuous pairs in the Asheville region are between
stations 109 and 117, which connect the Route 19/23 bypass
in the north to Interstate 26 in the south, and stations ll4 and
121, which connect the eastern and western extremes of
Interstate 40.
ffil
Estimation of Through-Trip Percentages

Figure 48. Base highway network.
The synthetic procedures outlined in Chapter 5 for esti-
EXTERNAL TRAVEL mating the share of external cordon trips that are likely to be
through trips are appropriate only for urbanized areas with
As was discussed in Chapter 5, external trips can be less than 50,000 in population. Therefore, local experience
divided into two categories: (1) external-external or through was relied upon to estimate the through-trip making poten-
trips, which pass completely through the region without hav- tial for the Asheville region. This experience was used to
ing a trip-end within the region; a (2) external-internal classify four facilities, each carrying ADT volumes of greater
TABLE 67 External stations
Station No. Description 1989 ADT Classification

109 Routes 19 & 23 Bypass 27,700 Principal
110 Routes 19 & 23 Business 7,000 Minor
111 BRP (N) 2,850 Minor
112 Snope Creek Road 2,000 Minor
113 Route 70 16,100 Principal
114 r-40 (E) 24,700 lnterstate
117 126 33,100 lnterstate
118 Routes 191 & 280 7,400 Minor
119 BRP (S) 970 Minor
121 r-40 (w) 27,500 lnterstate
122 Leicester Highway 14,000 Principal
123 Bear Creek Road 3,940 Minor
141
TABLE 68 Trip productions by household size where i : 109 and j : 111 . The calculations for the other
external station pairs are displayed in Appendix B-4.
Average Person-Trips
Household Size per Household
The relative shares for each of the possible destinations
One Person 3.7 from a cordon station were added together, and the result
Two Person 7.6 was used to adjust, or normalize, the raw shares. Table 71
Three Person 10.6 displays the raw shares and the normalized shares for each of
Four Person 13.6
the potential destinations for through trips entering the region
Five+ Person 16.6
at each of the six interstate and principal external stations.
Next, the normalized shares were used to distribute the
Weighted Average 9.2 through trips entering the study arca at these stations to the
other five external stations that contribute a significant num-
ber ofthrough trips to the external-external trip table. For sta-
tion 109, the adjusted shares were used to distribute the 8,310
than 20,000, a.s intersta-te fa,cilities which were estimated to - -L-L:
mrougn trlps onglnattng at tllat slatlull -,- .- .l--
LU Lll utrrr
-L^- llvs st4-
C.,^ ^+^
contribute 30 percent of their traffic to the external-external tions. The same procedure was used to distribute the through
trip table. Two other facilities, designated as principal arteri- trips originating at those other five interstate and principal
als, were estimated to have a 10 percent through-trip share.
external stations. The results ofthis process are displayed in
The remainder of the external stations were designated as
Table72. Note that, for intuitive reasons, there are no intra'
minor arterials and were assumed to contribute a negligible
zonal trips within the external zones. Similarly' there are no
share of their ADT volumes to the through-trip table.
trips allowed between stations 113 and 114, which are prox-
Table 70 displays the external station volumes, including
imate parallel facilities unlikely to attract trips from one
the estimated number of through trips and internal-external
another.
trips. All of the data in this table reflect vehicle trips, since
Since the values arrived at in Table 72 are not symmetri-
they are based upon existing traffic count data.
cal (i.e., the number of trips from station I to station j is not
equal to the number of trips from j to l) the next step was to
average the ry and ji values to produce a symmetrical trip
Distribution of Through TriPs
to External-External TriP Table table. For example, since the estimated value from station
109 to station I 17 is 7,03 1, and the value from station 1 17 to
The distribution of through trips between stations was esti- 109 is 8,402, the averagevaluebetween stations 109 and 117
mated using Equations 5-2,5'3, and5-4. The relative shares is 7 ,711 . The result of this exercise is displayed in Table 73'
were flrst calculated as in the following example for the inter- the symmetrical trip table.
change between external stations 109 and 117, which repre- Unfortunately, the symmetrical through{rip table in
sent the eastern and western extremities of I-40 within the Table 73 is a trip table in which the row totals and column
study area: totals are not equal to the through volumes estimated in
Table 70. The recommended solution to this problem was to
Y, = -2.70 + 0.21x PTTDEST + 67.86 x RTECON,t apply the Fratar technique to the symmetric trip table' using
= -2.'lO + 0.21 x 30 + 67.86 xI= 11.46 the through-trip volumes in Table 70 as the row and column
TABLE 69 Percentage oftrips by purpose and household size
Percent of Average Daily Person Trips

Household Size HBW HBO NHB
One Person 20 54 26
Two Person 22 54 24
Three Person 19 56 25
Four Person 19 58 23
Five+ Person 17 62 21
Weighted Average 20 57 23
r42
TABLE 70 External station through-trip summary
Station 1 989 Percent External- lnlernal-

Number Description ADT Classification Through External External
108 Route 251 1,800 Minor 0 0 1,800

109 Routes 19 & 23 Bypass 27,70O lnlerstate 30 8,310 1 9,390
110 Routes 19 & 23 Business 7,000 Minor 0 0 7,000
111 BRP (N) 2,850 Minor 0 0 2,850
112 Snope Creek Road 2,000 Minor 0 0 2,000
113 Route 70 16,100 Principal 10 ,610
1 14,490
114 l-40 (E) 24J00 lnterstate 30 7,410 17,290
115 Route 74 11,000 Minor 0 0 11,000
116 Route 25 12,450 Minor 0 0 12,450
117 l-26 33,100 lnlerstate 30 9,930 23,170
118 Routes 191 & 280 7,400 Minor 0 0 7,400
119 BRP (S) e7o Minor 0 0 970
120 Route 151 1 ,550 Minor 0 0 1,550
121 l-40 (w) 27,5oo lnterstate 30 8,250 19,2s0
122 Leicester Highway 14,000 Principal 10 1,400 12,600
123 Bear Creek Road 3,940 Minor 0 0 3,940
targets. The ultimate result of the FRATAR process is the BALANCING PRODUCTIONS
'--_ 1
..
final external-external vehicle-trip table, as displayed in AND ATTRACTIONS
Table74.
The final step in the trip generation phase of travel demand
forecasting is the balancing ofregional trip productions and
Conversion of Internal-External Trips attractions. The trip distribution phase of the travel demand
to Person-Trip Productions and Attractions forecasting process requires that the total number ofregional
tripproductions equals the total number of regional trip attrac-
In order to estimate the internal-external vehicle-trip tions for each of the trip purposes. Table 76 summarizes the
totals, the through-trip totals were subtracted from the exter- internal-trip and external-trip totals before balancing.
nal station totals as shown previously in Table 70. Next, the The regional control totals forproductions and attractions of
external trip purpose factors were applied to the external- all three trip purposes were set to equal the combined intemal
internal totals. Local experience in the region was used to plus external trip productions. For example, the control total
estimate that the traffic crossing the external cordon is com- for home-based work trips was set at 124,875 trips. The bal-
posed of 40 percent home-based work trips, 40 percent ancing process was accomplished by applying a balancing fac-
home-based other trips, and 20 percent non-home-based tor to the attraction trips for all internal TAZs. The balancing
trips. Local experience was then used to further estimate that factor is intended to change the total number ofinternal attrac-
the Asheville area is a net importer of work trips, by a ratio tions so that the total number of attractions, including external
of 70 to 30 and that the region is a net importer of other stations, equals the total number of productions. Following the
home-based trips by a ratio of 60 to 40. As usual, non-home- example for home-based work trips, the goal was to factor the
based trips were assumed to be balanced between produc- 85,604 internal HBW trips so that the total number of attrac-
tions and attractions. tions equalled the total number ofproductions. In order to fac-
Finally, auto-occupancy factors (from Chapter 7) of 1. 1 1 tor the 106,536 total HBW attracrions to equal the 124,875
persons per vehicle for home-based work trips, 1.67 persons productions, the intemal HBW attractions had to be factored
per vehicle for home-based other trips, and 1.66 persons per to equal 124,875 minus the number of external HBW attrac-
vehicle for non-home-based trips were used to convert the tions (since external trips are based on existing traffic volumes,
vehicle trips into person trips. The resulting estimates of trip they were not factored). The balancing factor for HBW trips
productions and attractions for external stations in the was therefore calculated as
Asheville region are summarized in Table 15. This table
shows that the estimated 157,150 external-internal vehicle
HBW Factor =
124,875 - 20932
= 1.2142
trips crossing the cordon around the Asheville region carried 85,00
226,925 person trips, including I37,915 productions (trips
from locations outside the region) and 89,010 attractions Similarly, balancing factors were calculated for the other trip
(trips to locations outside the region). purposes as follows:
143
TABLE 71 Through-trip distribution-raw and normalized percentages
Origin Station
Destination
Slation 114 117 121
Raw Percentages
109 3.60 3.60 71.46 3.60 3.60

113 3.23 3.23 3.23 .zo
1't4 3.60 3.60 71 .46 3.60
'117 7',! .46 3.60 3.60 3.60 3.60
121 3.60 3.60 71.46 3.60 3.60
'122 2.56 2.56 2.56 2.56 2.56
Tolal 81.22 17.63
Norm. Factor 1.184 7.483 1.231 1 .184 1.184
Normalized Percentages
109 ,u.y 4.43 84.61 4.26 20.42

113 3.; 3.83 3.83 18.33
114 4.26 4.26 84.61 20.42
1't7 84.61 26.94 4.43 4.26 20.42
121 4.26 26.94 87.98 4.26 20.42
122 3.03 19.18 3.16 3.03 3.03
.l00.00
Total 100.00 100.00 100.00 100.00 100.00
HBo Factor= Z#ry =r.2521

The frnal step in trip balancing involved updating the non-
home-based productions. Remembering that the NHB trips
that are made by the residents of a household do not have
NHB Factor =l!ffiY= 0.8375 either trip-end at the household (that is why they are called
non-home-based trips), the non-home-based trip productions
were replaced by the distribution associated with the non-
home-based attractions. In other words, the calculation of the
After the balancing factors were applied, the total numbers number of non-home-based productions was performed for
of productions were calculated as summarized in Table 77. the sole purpose of calculating the total number of non-
The total number of attractions calculated for the internal and home-based trips for the region. The distribution of trip-end
external zones is 520,920, which matches the total produc- locations is best estimated for both productions and attrac-
tions for the region. tions using the NHB attraction equations.
T^BLE7? Through-trip table-asymmetrical
Origin Station
Destination
Station 109 113 117 121 122 Total
109
113
114
318
354
v "?
8,402
380
423 6,981
352
316
286
257
286
9,802
1,271
I,OM
117 7,031 434 328 352 286 8,431
121 354 434 6,519 423 286 8,016
122 252 309 2U 301 250 1,U7
Total 8,310 1 ,610 7,410 9,930 8,250 1 ,400 36,910

144
TABLET3 Through-triptable-symmetrical
Origin Station
Destination
Station 109 'l 13 114 117 121 122 Total
'109 7,717 353 269 9,056

113 376 '1 '1 407 375 283 1,440
114 341 376 6,750 260 7,727
117 7,717 407 .r/t) 387 294 9,180
121 353 375 6,750 347 268 8,1 33
122 269 283 260 294 268 1,373
Total 9,056 1,440 7,727 9,180 8,133 1 ,373 36,910
Target 8,310 1,610 7,410 9,930 8,250 1 ,400 36,9'10
Adj. Factor 0.918 1.118 1.082 1 .014 1 .01 I
The balanced productions and attractions for the three trip Free-flow times, or impedances, were used to perform the
purposes are listed in Appendix B-5. At this point, these val- first distribution of trips for the three trip purposes. The
ues for the productions and attractions were ready to be used resulting person-trip tables were then converted to vehicle
in the trip distribution phase of model development in order trips, which in turn were combined to build a daily vehicle-
to prepare the person-trip tables. Therefore, the production trip table. This table was then assigned to the highway net-
and attraction data were imported into the travel demand fore- work using an equilibrium assignment. The result of this
casting software and saved in three origin vectors for the pro- assignment was a set of traffic volumes and congested speeds
duction data for the three trip purposes and three destination for each link in the highway network. These congested speeds
vectors for the attraction data for those same trip purposes. were subsequently used to produce a second matrix of con-
gested travel times, which were used to perform a second
application of the trip distribution model, which resulted in
TRIP DISTRBUTION the ultimate trip distribution and traffic assignment.
The trip distribution process and subsequent steps were

performed twice during the course of the Asheville case Estimation of Free-Flow Travel Times
study. The first time through the process was intended to esti-
mate baseline traffic volumes that result from the assignment The free-flow zone-to-zone travel time matrix was con-
of a trip table based on free-flow speeds on the highway net- structed using the default speeds posted on each link in the
work. The second run was designed to use congested travel Asheville highway network. Speeds were set at 55 mph on
times resulting from the first run as the measure of imped- freeways,45 mph on major arterials, and 35 mph on minor arte-
ance for the second application of the trip distribution model. rials. Travel time was calculated using the simple relationship:
T 74 Through-trip table-Fratared
^BLE
Origin Station
Destination
Station '109 113 114 117 121 122 Total
109 ,- ,y 7,526 243 152 8,310

113 2; 676 439 273 1 ,610
114 167 515 6,521 207 7,410

117 7,526 676 515 746 467 9,930
121 243 439 6,521 746 8,250
122 152 273 207 467 301 '1 1,400
Total 8,310 1 ,6'10 7,410 9,930 8,250 1.400 36.910

145
TABLE 75 External-internal person-trip productions and attractions
Station
Number HBW HBO NHB HBW HBO NHB
108 559 721 299 240 481 299

109 6,026 7,772 3,219 2,583 5,181 3,219
110 2,176 2,806 1j62 932 1,870 1j62
111 886 1,142 473 380 762 473
112 622 802 332 266 534 332
113 4,503 5,808 2,405 1,930 3,872 2,405
114 5,374 6,930 2,870 2,303 4,620 2,870
115 3,419 4,409 1,826 1,465 2,939 1,826
116 3,869 4,990 2,067 1,658 3,327 2,067
117 7,201 9,287 3,846 3,086 6,191 3,846
118 2,300 2,966 1,228 986 1,577 1,228
ii9 30 389 tot tv cY tot
120 482 621 257 206 414 257
12'l 5,983 7,715 3,196 2,564 5,1M 3,196
122 3,916 5,050 2,O92 1,678 3,367 2,O92
123 1,225 1,579 654 525 1,053 654
Total Person Trips 48,842 62,986 26,087 20,932 41 ,990 26,087
by the diagonal of the 107 by 107 matrix of internal zones,

Travel Time (in minutes) = Link Length (in Miles)
was produced using the nearest-neighbor method. This con-
60
sisted of identifying the zones adjacent to each of the 107
Speed (in mph)
internal zones, taking the free-flow travel time from the zone
of interest to all adjacent zones, calculating a mean for that
The transportation planning software was used to produce
set of times, and halving that value to arrive at the assumed
the matrix of travel times, or highway skims, based on the
intrazonal travel time. These steps resulted in intrazonal
minimum time path between each pair of zones. These free-
times ranging from 0.54 minute to 4.33 minutes, with an
flow travel times were based on speed and distance only; no
average value of 1.85 minutes.
volume delay was included.
lntrazonal Times Terminal Times
The free-flow matrix, as produced in the previous step, did Terminal times represent impedances at both the origin
not contain any intrazonal travel times, which represent the and destination ends of a trip, such as the time required to
travel time required to make a trip wholly within a single park or access a car, parking cost, and so forth. For the
TAZ. An intrazonal travel time matrix, which is represented Asheville case study, the study area was assumed to have
TBLE 76 Unbalanced trips
lnternal 76,033 215,407 91,s66 85,604 88,8061 109,331

External 48,842 62,986 26,O87 20,932 41 ,990 26,087
Total 124,875 278,393 117,652 106,536 230,796 135,418

146
T^BLE77 Balanced trips
lnternal 76,033 215,407 91,566 103,943 236,402 91,566

External 48,842 62,986 26,087 20,932 41,990 26,087
Total 124,875 278,393 117,652 '124,875 278,393 '117,652
three distinct area types: CBD, suburban, and rural. Zones The final step in the calculation of the friction factors was
I through 15 were designated as CBD zones and all trip- to set the friction factors for the external-external zone pairs
ends at those zones have a terminal time of 5 minutes. to zeo. This adjustment was performed in order to prevent
Zones 16 through 59,61,62, 65 through7l,74,77 through the gravity model from distributing any trips to external-
80, 82, and 83 were designated as suburban zones and have external zone pairs. (Otherwise, the distribution model and
terminal times of 2 minutes for each trip-end. The remain- the through traffic model would combine to overestimate
der of the internal zones were designated as rural zones and the number of through trips made in the region.) A friction
have a 1-minute terminal time associated with all trip-ends. factor value of zero was used to replace the calculated value
I
The average terminal time for the 107 internal TAZs is2.2 for trips with both origin and destination zones between 108
minutes. and 123.
t' -_ --1
' The total travel time for each zone-to-zone pair was cal-
culated by adding the terminal times at both the origin and
destination ends of the trip to the free-flow travel time (or to Creation of Free-Flow Trip Tables
the intrazonal travel time in the case of intrazonal zone pairs).
After the friction factor matrices were created for the three
This process was used within the matrix calculator to pro-
I
duce a matrix of zone-to-zone impedances between all I23

trip puqposes, the trips were distributed using the gravity
model component of the travel demand modeling software.
zones in the Asheville region.
The balanced productions and attractions by trip purpose
were set as the row and column control totals. The Fratar
Choice of Friction Factors method was then applied to the trip ends so that the row and
column totals matched the total productions and attractions
The gamma function was used to calculate the friction fac- in each zone.
tors that represent the travel impedances between zones in The output of the distribution process was a set of three
the trip distribution gravity model. The calculation of friction person-trip tables. These matrices contain the same number
factors for each zone pair was performed within the travel of trips as the trip generation control totals: L24,8'75 home-
demand model software by using a matrix calculator. The based work trips, 27 8,393 home-based other trips, and I 17,652
preliminary friction factors used in this case study were cal- non-home based trips. However, unlike the production and
culated using the gamm function coefficients listed below attraction vectors, the person-trip tables are two-dimensional
for the three trip purposes: and reflect the movement of trips between zones.
Trip Purpose
Home-Based Work 100 -0.020 -0.125

Non-Home-Based 100 -1.350 -0.100
For example, the home-based work friction factors were Since the distribution of trips is calibrated to the trip length
calculated using the equation: in minutes for each trip purpose, it is useful at this point to
review the trip lengths after applying the gravity model.
o'o
HBW Friction Factor(i,j) = 100 x tf * etx(4tzs\ Average trip length was obtained by weighting the free-flow
travel time matrix, including intrazonal times and terminal
where fu : the impedance for all interchanges between origin times, with the person-trip tables. The average trip lengths
zone i (from I to 123) and destination zonej (from I to 123). for the initial application of the gravity model are as follows:
147
home-based work-l 6.9 minutes; home-based other-14.4 occupancy factors found in Table 31 of Chapter 7. The aver-
minutes; and non-home based-14.8 minutes. age values by trip pu{pose for urban areas with under
Two reasonableness checks were performed on these 200,000 population were applied to convert the 520,921 total
results. First, the average trip length for home-based work person trips into 350,071 vehicle trips, as shown in the fol-
trips resulting from the gravity model was compared with the lowing equations:
average home-based work trip length derived from the 1990
census data. The 1990 Journey-To-Work statistics show that
124,87 5 HBW Person Trips
residents of the Asheville MSA reported an average home- HBW Vehicle Trips =
1.11
based work trip length of 18.7 minutes. The average trip
length produced by the gravity model, using free-flow speeds 112,500 HBW Vehicle Trips
to build the impedances, should be slightly less than the 278,393 HBO Person Trips
average trip length reported by actual commuters, who tend HBO Vehicle Trips =
t.67
to experience congested traffic during their home-based 166,102 HBO Vehicle Trips
work trips. The average modeled free-flow travel time of
16.9 minutes, achieved with the use of the default prame- 117,652 NHB Person Trips
NHB Vehicle Trps =
L:^ -^^^^-^Ll^-^^^ ^L^^1. L.66
tr, P4 tlrr rc4vll4urgrl lr.
The second reasonableness check suggests that the average 10,815 NHB Vehicle Trips
trip length for home-based other trips and non-home-based
trips should be approximately 80 percent of the home-based
work trip length. For the Asheville MSA, this corresponds to
TIME.OF.DAY CHARACTERISTICS
15.0 minutes for home-based non-work and non-home-based
trips. The modeled results, using the default coefficients for
This is the point during the travel-demand modeling
the gamma function, produced average trip lengths of 14.4
process at which the daily trip tables, which had been main-
minutes and 14.8 minutes for home-based non-work and
tained in production-attraction format (P-A) were converted
non-home-based trips respectively. Since these values are
to origin-destination format (O-D) for the time periods to be
based on free-flow speeds, they are well within the range of
reasonableness. analyzed. For the Asheville case study, we wished to use the
Later, after the assignment of the vehicle-trip table to the model only to estimate ADT volumes. Conversion of the
congested highway network, the average travel time was three P-A trip tables and the through-trip table into one daily
recalculated to ensure that the average travel time remained vehicle-trip table was accomplished using the following
reasonably close to the average travel times reported in the equation:
J o urney -To -Work statistics.
Aside from trip lengths, another way to check the reason- Daily Vehicle Trips (O-D) = 0.5 x (HBW'. + HBIU.
ableness of the trip distribution results is by comparing the + HBOro + HBOA.)
trip table data with any data regarding observed travel pat- + NHB + ThroughTrips
terns. These data, if they existed, would usually come in the
form of trip movements between districts or groups of where HBWp is the transpose of HBWp. The non-home-
zones, which could be compared with the model estimated
based and through trips were not factored since they were
interchanges. Unfortunately, such data do not exist for the
already balanced in origin-destination format.
Asheville MSA.
The time-of-day characteristics presented in Chapter 8
provide the ability to factor a daily trip table to create peak
MODE.CHOICE ANALYSIS period and off-peak period trip tables. Assigning traffic by
time-of-day considers the relative levels of congestion and
Because transit usage in the Asheville region is so small,
the alternate optimal travel paths between zone pairs that
representing less than 1 percent of the average daily person
vary by time period. By adding traffic volumes from each of
trips made in the region, we did not use a mode-choice model
the time periods together, an estimate of the daily volume on
component in this case study.
a link was produced.
If we had done so, however, we would have needed to build
a transit network and transit travel time matrices (transit The results of trip distribution for Asheville indicated that
skims) and to apply one of the mode-choice models described the difference between free-flow and congested conditions
in Chapter 6. is minimal. As a result, assigning the trips by time period
would not necessarily produce a better assignment. For the
purposes of this example problem, the daily trip table was
AUTOMOBILE-OCCUPANCY
CHARACTERISTICS converted to a peak-hour trip table by factoring the daily trip
table by a 10 percent factor. The trip table assigned to the
The estimation of vehicle trips for the purposes of the network was calculated by applying the 10 percent factor to
Asheville, North Carolina, case study was based on the auto- all trips such that:
: . -7-: l
::.: i
:
'l:
'..',1 j 148
. .,-',,
-i
,.1
Hourly Vehicle Trips (O-D) = travel times. The revised friction factors were calculated
..;. .l 0.10 x Daily Vehicle Trips (O-D) using the same gamma function coefficients as before, and
the gravity model was applied to create new person-trip
'':.:.. ..."i tables based on the congested times.
t _
TRAFFIC ASSIGNMENT OF FREE-FLOW In order to check the validity of the trip distribution model,
'-l
TRIP TABLE
the average trip lengths were calculated for the revised
person-trip tables based on the congested travel time. Aver-
The hourly vehicle-trip table from the previous step was
age trip lengths produced using congested travel times were
assigned to the base highway network using an equilibrium
home-based work-I7 .7 minutes, home-based other-14.9
assignment. The travel-demand modeling software package
minutes, and non-home-based-15.4 minutes.
used for the case study included algorithms for an equilib-
The estimated travel time increased slightly for all three
rium assignment using parameters from Chapter 9. trip purposes as the result of the congested travel times. The
Delay on the roads caused by congestion was calculated Asheville region does not experience a great deal of traffic
using the Bureau of Public Roads (BPR) curve shown in congestion, which can help explain the relatively small
Chapter 9. Coefficients for the formula were also obtained degree to which the average trip lengths changed.
from that chapter. Freeway links used values of 0.83 for a Since the average ffip length for home-based work trips was
and 5.5 for B that correspond to a design speed of 60 mph. approximately I minute less than the target value, we per-
Arterial links used values of 0.71 for cr and 2.1 for B that cor- formed two more iterations of the trip distribution model to
respond to a design speed of 50 mph on multi-lane roads. achieve a more acceptable average travel time for this trip pur-
Congested travel time was calculated using the following pose. The coefficients listed below were the ultimate coeffi-
formulas: cients used in the gamma function for the three trip purposes:
I
Trip Purpose
Work
Home-Based 100 -0.300 -0.070
Non-Work
Home-Based 100 -1.250 -0.100
Non-Home-Based 100 -1 .350 -0.100
qswYil
Freeway Travel Time = The trip tables resulting from these model coefficients had
/ r / / volume,
. ls.s\ average trip lengths of 18.7 minutes for HBW, 15.0 minutes
llensth x oo x l t * o.*r r[ nn"rl"'" I for HBNW, and 15.4 minutes for NHB trips. The trip length
\ speed / l. \capacity ) ) frequency distribution curves for these three trip purposes are
Arterial Travel Time = displayed in Figure 49.
/ \ I \z.l\ After the calibration of the trip distribution model was com-
llensth x ao x L * n.r, *f volume , mn".rl''' I pleted and the congested person-trip tables were produced, the
\ speed / l. \capacity ) ) next step was to repeat the auto-occupancy and time-of-day
steps to build a congested vehicle-trip table. These steps were
The equilibrium traffic assignment used several iterations
followed by a flnal equilibrium traffic assignment of the vehi-
to assign the trip table to the shortest paths that would result
cle-trip table. As with the assignment of the free-flow trip
in the least amount of overall congestion on the highway net-
table, this assignment produced traffic volumes for each link
work. The end product of the equilibrium traffic assignment
in the highway network. These volumes were then factored by
was a network of link volumes and congested speeds.
a value of 10.0 to reflect total daily traffic conditions. Figure
50 is a bandwidth plot of the volumes on the network.
CREATION AND ASSGNMNT
OF CONGESTED TRIP TABLES MODEL VALIDATION AND SCREENLINE
COMPARISONS
The estimation of a congested travel-time matrix is simi-
lar to the process used to estimate the free-flow travel-time Traffic volumes were summarizedatfrve screenlines in the
matrix. The major difference between the two processes is Asheville region. Table 78 presents a comparison of the esti-
that the travel times use the link speeds from the preceding mated and observed daily traffic volumes across these screen-
equilibrium traffic assignment, which are subject to volume lines. These comparisons provide an indication ofthe accuracy
delay. The equilibrium assignment was used to estimate con- of the travel models. Table 78 shows that the simulated vol-
gested travel times between zones, which were then saved in umes across the five screenlines range from 19 percent below
a congested travel-time matrix. The congested travel-time the count volumes to 5 percent above. The overall trend seems
matrix was completed by calculating the congested intra- to be that travel across these screenlines is underestimated by
zonal times and adding the terminal times to the congested approximately 10 percent.
'; - j
-i"r
.',1
149
g. 30
t-
I Home-Based Work
o
!zo
o + Home.Based Other
o * Non-Home.Based
g
o
fL 10
o-5 11-15 21-25 3't-35 4145 51-55

r0 16-20 2:.:0 . 36-40 450 s660
MINUIES
Figure 49. M odel- generate d trip -length distribution-by trip purpo s e
The screenline comparisons have shown that the estimated As displayed in Table 19, the model-generated traffic
traffic volumes match the observed traffic counts reasonably assignments were very reasonable for the freeway facilities,
well. Another check of the reasonableness of the model where the average link volume was 1.016 times the count
results can be obtained by comparing the observed and esti- volume, with a standard deviation of 0.222. The average traf-
mated traffic counts for all facilities for which count data are fic assignment on major arterials was also reasonably close
available. The NCDOT provided a selected count map that to the count data, with an average ratio of 0.945; however,
displayed ADT count data for 386 links on the Asheville net- the standard deviation for these counts was a rather large
work. This database included counts on 48 freeway links, 0.969. Average traffic assignments on minor arterials and
176 major merials, and 162 minor arterials or collector roads. collector facilities were far less respectable, with an average
The Federal Highway Administration' s manual, Calibration ratio of 0.567 and a standard deviation of 0.635. These results
and Adjustment of System Planning Models (1990) listed are consistent with the original intent of the case study, which
the following suggested limits by functional classification: was intended to forecast traffic volumes on the major road-
freeways-less than 7 percent; principal arterials-less than ways. The traffic assignments on the minor facilities could
10 percent; minor arterials-less than 15 percent; and col- be improved by disaggregating the zone structure. Disaggre-
lector-less than 25 percent. gation would improve the ability to forecast traffic on minor
facilities by increasing the accuracy with which trips from
eachTAZ access the coded highway network via centroid
DAILY fFFFIC VOLUI4ES
connectors and by decreasing the number oftrips that are dis-
tributed to intrazonal cells in the trip tables given that these
trips are not assigned to the highway network.
Model validation is usually performed at different levels.
First, systemwide performance is reviewed to determine if
regional inputs or parameters should be changed. For exam-
ple, given that the assignment volumes appear to be slightly
Iower than the observed volumes, changes in the socio-
economic data, trip generation rates, auto-occupancy factors,
or trip length could be used to increase volumes throughout
the region. Second, if the assigned volumes on different facil-
ity types are estimated less accurately than on others, the
default speeds or capacities on the various facilities could be
modified in order to balance the results. Third, problems on
specific links confined to a small area of the network could
indicate network coding errors.
For this case study, no furlher actions were performed. The
initial results provide an indication of the ability of the para-
Figure 50. Traffic assignment bandwidths. meters to match observed volumes without extensive adjust-
150
i
TABLE 78 Traffic volumes at major screenlines

-"j
Obserued Estimated Difference
East Screenline
u.s. 70 26,000 19,850 -23.7"/"
Swannanoa Road 8,500 2,380 -72.O
1I l-240 27,800 24,630 -11.4
u.s.74 11,500 9,290 -19.2
l-40 22,400 26,970 20.4
Total 96,200 83,120 -13.67"
South Screenline
u.s. 25-A 10,200 7,530 -26-2"/"
u.s. 2s 19,600 14,700 -25.0
l-26 36,000 36,200 0.6
NC 191 8.400 4.810 -42.7
Total 74,200 63,240 -14.8"/"
West Screenline
l-40 24,000 25,490 6.2"h
1240 40,200 42,710 6.2
U.S. 19-23 Haywood 17,600 9,770 -44.5
NC 63 Leicester Highway 18,800 12,200 -35.1
Total 100,600 90,170 -10.4"/"
North Screenline
NC 251 Riverside Drive 6,700 3,980 -40.6"/"
l-70 30,300 31,060 2.5
Broadway Skeet 7,1 00 13,660 92.3
U.S. 25 Merrimon Avenue 17,000 13,110 -22.5
NC 694 own Mountain 500 2,820 464.O
Total 61 ,600 64,630 4.97"
lnside Loop
Swannanoa River 8,500 7,380 -13.2"/"
Baltimore Avenue 19,500 16,270 - lb.
McDowell Street 14,600 9,130 -37.5
Meadow Road 8,000 8,090 1.1
Total 50,600 40.870 -19.2"h
ments. If observed data other than traffic counts, such as occu- example of an urbanized area with less than 200,000 in pop-
pancy rates, were available, these could have been used in con- ulation. The ability to replicate observed traffic volumes in a
junction with the default parameters to improve model results. small region with minimal transit ridership was seen as a
likely use for the default parameters.
CONCLUSIONS Few resources were used to complete the case study.
Inputs to the process included socioeconomic and network
This case study has presented one application of the pro- data which were already available. Software requirements
cedures in the manual. The Asheville region provides a good were limited to a standard computer spreadsheet for the trip
151
Table 79 Assignment validation by facility type
AverageRatio Standard
Facility Type Observations (Est./Obs.) Deviation
Freeway 48 1.016 0.222

Major Arterial 176 0.945 0.969
Minor Arterial and Collector 162 0.567 0.635
Total
generation and external data calculations and a PC-based Additional validation adjustments could have improved the
travel-demand modeling software for the distribution, matrix model results
calculation, and assignment processes. Daily traffic counts Application of the calibrated travel models to produce
had already been collected in the region. forecasts of future travel would now be relatively easy. Pro-
Although the assignment results were not final, the default jections of socioeconomic changes and network updates '.
parameters were able to replicate observed traffic volumes couldbe analyzed toplanthetransportation systemrequired : r'
onthemajorfacilitieswithinacceptablelevelsoferror.tomeefthedemandsofthefuture.
152
't :,, i
BIBLIOGRAPHY
I'.1
Atken, S., "Transportation Planning-Is There a Road Ahead?" sentation at the 71st Annual Meeting of the Transportation
Trffic Engineering and Conn"o/, Vol. 18., No. 2(1977)pp.5842. Research Board, National Research Council, Washington, D.C.
Babin, 4., et a1., "EMME/2: Interactive Graphic Method for Road (1992).
and Transit Planning." Transportation Research Record 866, Garreau, J., Edge City: Life on the New Frontier. New York:
Transportation Research Board, National Research Council, Doubleday (1991).
'Washington,
D.C. (1982) pp. 1-9. Goodwin, P., "Some Principles of Dynamic Analysis of Travel
Barton-Aschman Associates, Inc. "Vehicle Occupancy Determina- Behavior." ln Developments in Dynamic and Activity-Based
tors: Report FHWA-4289-252." Phoenix, Ariz. (1989). Approaches to Travel Analysis. P. Jones (ed.). Transport Study
Bates, J., and Dasgupta, M., "Review of Techniques of Travel Unit, Oxford University (1990).
Demand Analysis: The Policy Context." Contractor Report 282. Hanson, 5., The Geography of Urban Transportation New York:
Transport and Road Research Laboratory. Crowthorne, Berk- Guilford Press (1986).
shire, U.K. (1992). Hartgen, D., "Travel Analysis Methods for the 1980s." Special
Benson, J., et al., "Development of a Peak Period Traffic Assign- Report 201, Transportation Research Board, National Research
ment Capability." Research Report 454-lF, Texas Transporta- Council, Washington, D.C. (1983) pp. 3-4.
rion Institute, Texas A&M University (1988). Harvey, G., and Deakin, E., "Toward Improved Regional Trans-
Band, D., "Reseach Needs for Analyzing the Impacts of Trans- portation Modelling Practice" Contractor Report, National Asso-
portation Options on Urban Form and the Environment." In ciation of Regional Councils, Washington, D.C. (1991).
"Transportation, Urban Form, and the Environment." Special Hauer, E., and B-Y T. Shin, "Origin-Destination Matrices from
Report 231, Transportation Research Board, National Research Traffic Counts: Application and Validation on Simple Systems."
Council, Washington, D.C. (1991). Trffic Engineering and Control, YoI. 22, No. 3 (1981) pp.
Carlson, J., et al., "Small-Area Population and Employment Fore- 1t8-204.
casts and Trip Generation for Subregional Planning." Technical Holmes, R., "Panel Discussion Notes." In Proceedings of the ASCE
Report 4 , North Central Texas Council of Govemments, Arling- Transportation P lanning and Ait' Quality C onferenc e,SantaBar-
ton, Tex. (1977). bara, Calif. (1991) (Forthcoming).
Cervero, R., "Land Use and Travel at Suburban Activity Centers." Howe, S., and Gur, Y., "Trip Distribution in Subregional Analysis."
Transportation Quarterly, Vol. 45, No. 4 (1991) pp.419491. Technical Report 7, North Central Texas Council of Govem-
Cook, P., et al., "Comprehensive Transportation Models: Current ments, Arlington, Tex. (1978).
Developments and Future Tends." ITE Journal, Vol. 59, No. 6 Johnston, R., "Peak Spreading." Trffic Engineering and Control,
(1989) pp. 33-38. Yol.32, No. 1 (1991) pp.20-24.
Deakin Harvey Skabardonis, et al., "A Manual of Transporta- Jones, P., et al., Understanding Travel Behavior. London: Gower
tion-Air Quality Modeling for Metropolitan Planning Organiza- Publishing. (1983).
tion" (1992). Lee Engineering,"1992 Study of Occupancy and Vehicle Classifi-
Dueker, K., et al., "The Impact of EMME/2 on Urban Transporta- cation in the Metropolitan Phoenix Area." Phoenix, Ariz. (1993).
tion Planning: A Portland Case Study," Intemal Report. Metro- Lomas, T.J., and Morris, D.E., "Guidelines for Estimating the Cost
politan Service District (METRO) of Portland (1985). Effectiveness of Highway-Occupancy VehiclesLanes." Research
Dunn, R., and Winder, P., "Strategic Planning Modelling Software: Report 339-5, Texas Transportation Institute (November 1985).
Introducing EMMEp to Auckland." Proceedings of the 15th Loudon, W., et al., "Predicting Peak-Spreading Under Congested
Australian Road Research Board Conference, Darwin, Northem Conditions." Presentation at the 67th Annual Meeting of the
Territory, Australia ( I 990). Transpofation Research Board, National Research Council,
Finance, Planning, Programming, Economic Analysis, and Land Washington, D.C. (1988).
Development 1991. Transportation Research Record 1305, Mahmassani, H., et al., "Characteristics of Urban Commuter
Transportation Research Board, National Research Council, Behavior: Switching Propensity and Use of Information." Trans-
Washington, D.C. (1991). portation Research Record 1285, Transportation Research
Florian, M., "Nonlinea Cost Network Models in Transportation Board, National Research Council, Washington, D.C. (1990) pp.
Analysis." Mathematical Programming Study, Vol. 26 (1986) 5749.
pp.167-196. Meyer, M., and Sarasua, W., "AGIS-Based Transportation Pro-
Gallimore, W., et al., "Applications of GlS-Transportation Analy- gram Management System for County Transportation Agency."
a
sis Packages in Super Regional Transportation Modelling." Pre- Presentation at the Tlst Annual Meeting of the Transportation
153
Research Board, National Research Council, Washington, D.C. Roden, D., "The Development of a Regional Information System
(1992). and Subarea Analysis Process." ITE Journal, Vol. 58, No. 12
Murakami, E., and Watterson, W., "Developing a Household Travel ( 1988) pp. 27-31 .
Panel Survey for the Puget Sound Region." Transportation Rogers, K. "Congested Assignment and Capacity Capping-
ResearchRecordl2S,TransportationResearchBoard,National Constraining the Trip Matrix to Reflect Network Capacity."
Research Council, Washington, D.C. (1990) pp.4046. Trffic Engineerng and Control, Vol. 32, No. 7/8 (1991) pp.
Nihan, N., and Miller D., "Development and Demonstration of the 342-346.
Subarea Focusing Concept for Trip Distribution in the Puget Rossetti,M.A.,andEversole,B.S.,"Journey-to-WorkTrendsinthe
Sound Region." Paper presented to the Transportation Research United States and Its Major Metropolitan Areas l96G-1990."
Board, National Research Council., Washington, D.C. (1976). FHWA-PL-94-02 (November 1993).
1990 Nationwide Personal Transportation Survey: Summary of Sharaf-Eldien, O., "Hierarchial Traffic Flow Optimization for Con-
Travel Trends, FHWA-PL-92-027,Federal Highway Adminis- gestedNetworks."PhDThesis.Universityof Maryland,College
tration (March 1992). Park (1988).
l990NationwidePersonalTransportationSurvey: TravelBehavior Sosslau, A., et al., "Quick Response Urban Travel Estimation
Issues in the 90's. FHWA-PL-93-012,Federal Highway Admin- Techniques and Transferable Parameters: User Guide." NCI1RP
istration (July 1992). Report 187, Transportation Research Board, National Research
Parsons Brinckerhof Quade & Douglas, Inc. First Phase of Model Council. Washington, D.C., (1978).
Development: Review of BestPractices. Washington, D.C. (1992). Supernak, J., "Transportation Modelling: Lessons from the Past and
Pas, 8., "Is Travel Demand Analysis and Modelling in the Dol- Tasks for the Future." Transportation, Vol. 12, No. 1 (1983) pp. :
drums?" ln Developments in Dynamic and Activity-Based 79-90.

Approaches to Travel Analysis. P. Jones (ed.). Transport Studies Thomas, R., Trffic Assignment Techniques. London: Avebury
Unit, Oxford University, U.K. (1990). Press (1991).
Pisarski, 4.E., "New Perspectives in Community.- FHWA-PL-72- "Transportation, Urban Form, and the Environment." Special
026 (July 1992). Report 231, Transportation Research Board, National Research
Prashker, J., et al., "Equilibrium Assignment Method for Pointwise
Council, Washington, D.C., (December 9-12, 1990).
Flow Delay Relationships." Working Paper UCI-ITS-WP-91-2. Ulberg, C., "An Evaluation of the Cost Effectiveness of HOV .
University of California, Irvine(1991). Lanes: Research Project Y3399 Task 7" (March 26, 1987).
Purvis, C., "Survey of Travel Surveys lI." Transportation Research U.S. Department of Transportation. "A Summary of the Inter-
Recordl2Tl,TransportationResearchBoard,National Research modalSurfaceTransportationEfficiencyActof 1991." FHWA-
Council, Washington, D.C. (1990) pp.23-32. PL-92-008. Federal Highway Administration, Washington, D.C.
Purvis, C., "Enhancing the Credibility of Travel Demand Forecasts." (1992).
Presentation at the 71st Annual Meeting of the Transportation Van Vliet, D., and Willumsen, L., "The Most Likely Trip Matrix
Research Board, National Research Council, Washington, D.C. Estimated from Traffic Counts." Transportation Research, Yol.
(t992). 138, No.3 (1980) pp.28l-93.
Replogle, M.4., "Improving Transportation Modeling for Air Qual- Van Vliet, D., "SATURN: A Modern Assignment Model." Trffic
ity and Long-Range Planning." Paper No. 930342. Transporta- Engineering andControl,Yol.23, No. 12 (1982) pp.578-81.
tion Research Board, National Research Council, Washington, Weiner, E., "Urban Transportation Planning in the United States:
D.C. (January 10-14, 1993). An Historical Overview." Technology Sharing Report DOT-T- '
Resource Division Consultants, Inc. "Redesigning the Travel 88-26,U.5. Departmentof Transportation (1988).
Demand Forecasting Process-No. DTFH61-92-Q-00099." Wong, S., "TRAF-NETSIM: How It Works, What It Does." IIE '
Washington, D.C. (1992). Journal, Vol. 60, No. a (1990) pp.22-27.

154
1
APPENDIX A
:-..
-**
NTPS AND HOME INTERVIEW SURVEY DATA
.l
I
- --"1
''i
.;
. .l
l{umbr ol Auto md tunbr of Srmpld HourHold h Vrrhu Sdl
Mcn Number of Tdpr
Housohold Number of
Size Auto ::i:i,tii'.,: A' lftr;,:
',,,Glrlor i:,
,s '.iVrntiha. 'Dlre .: qrql' NPNI
a tt t9 ta 7 l 22 llt a:l tt t2 lt t.t5
(h.P.rr
(ho C.r tt 2tt t6 lo tgt x2 ta 2t1) l5t 211 nt 2,W
Tro Crd : ,, 6 2t t1 a lt t I lt t, 7t to l9 ,t
nrcCn 0 5 I t t3 t9 o 7 : 6 tt 2 ilt
Zo C i . ta t2 t t g 7 6t2 t6 t7 It t5 {o3
Os C-rr 55 tt l(5 c, ?o tfl 22 t tu ln il2 mt 2,O3t

Tro Pro
TgoCrr. 2 zaa lvt 26 t2t m5 11 ttz tztt 57? t.9 2to 3,9r5
th C.t ?lui tt ao t at 76 ztt a ll9 a0 t7t t76 ll l,t',s

Zco C t 7 3 6 I 2a t 7 a) ll I t2 2tt
ODo C.3 33 lt 3t EN to t 3t att 3$ a2 at a t75
TGC F.rro
lwo Cr ,r n? t2l ta t6 ?6 tt9 5t tn t?lt 2t2 2t t6 t.w2
b Ci Pl It t9 6t u? a7 t@ a7 tm t4 t9t xn l0 a,l1t
Zo Ct t a o 2 2a 5 52 2n t2 I 9 2t9
Oc C.r::ri: 5t ?2 ta a6 62 t r a70 10 tot 2l tt t95
T*o C aaz t7l t22 6 Jtz ,x2 t t29 M t2t 242 210 2,t31
Tb Cn llu t u? tot 9 79 tta at a7a tolr 29'!t ln It2 2,21t

Tord l.tot t,tzz l.o5a I,O7 t96 2,9r' tr {.t70 l,4{i 2,754 2.att a,t27 22,3t7
Appendlx A-1
ltit t', I1i I
,ta ' lffiri

I t[i ']
-t ._.. _
o\
Avrf 'cron Trlp pcf Houchold by of Aulo md Poron pcr
bv ilumbcr ot 9cl Hdtchold h Vrlou Trmtgortalhn Sdlc
Mean Number of Tripe
Household
l.\:r:vL,.tT !'wv.+:+i
tr.".S.*if.sgtft
Size iltffiBi iqI*.j irut tlr f,ffift :::::!f!f[!iii
:::::l:::::::::.::l:::::.::: j
t.(x) r.55 t.79 2.t2 t .29 .56 t.t2 a. t.) 3 .27 0 .7t l .3t L17
5.(X 1.16 1 .a2 a .at 3 . a .(I) 5.27 4 .7t a. t0 5.9t 3.92 3.51 3. r
Ono Pcrm
4.m 1.13 1.97 s.lt 3 .s2 5.22 5.(x a .t1 1 ,t6 5 .79 t .62 t .ll 3 .51
o. t.@ 3.t9 a.(x a.2t 1 .tt o.(n a .tt 6 ,n t .3t 1 .72 2. 3 .t2
o.m 2.29 2.75 2.I) 2 .(x, 3.03 1.4t a.37 t. 7 .3t 2.17 t.tt I .29
7.5r 7. 6 .n 5. 3 t .50 .5a t.t6 a.f 6 .at l0 .21 6.11 t.t { .3
Two Prnoo
7.26 7.9 7.22 7. rt 1 .72 7.37 7.6 a.19 7 .&t l0 .tt 7 .lo 6.62 5 .53
7.@ 9.a3 E.2t 7.% I .t7 7.39 6.75 a.77 t.v2 t2.3t t .6t t.29 6 .m
2.67 t .lt t .67 7 .25 t. 5.75 6.ao t.0 .01) I l.a0 s.62 a.J() a .t2
?. tr t,29 t.29 6.7t 5 .,10 7.2t t0. to t .27 9 . t1.21 t0.33 .q 6.6
Thrcc Pero
0.0: l0 .lt t.3t 9. r9 t .t2 to.2t to ,tt a .76 9 .52 t5.t0 to.ot 7.76 7 .st
to.2t lr.t5 lt.ot lo.0f) 7. t0.73 il .a3 I .il 12 .05 t.4 to.?a to.(n t .
o.) 7.75 o.m .r0 ta . 4.79 7 .ao 3.t t2 ,t t2.25 to.?r 4.5 6.t5
10.20 t2 .t2 o.2 to.5t 7. 19 to.3E t2 .t6 a.9a lO.lO .12 la.2t 7 .m 9. lo
Four Pcrm Pfu
t2.27 16.26 11.32 tt.12 7 .70 t3 .26 ta .75 a.7s la .9E 21;1t t 5.76 l .3t I l.o7
il.il t't.62 15.26 t3.9 0. t6 l5 .tt l6 .25 1.vt t .63 25.' t.2r t1 .t6 t.37
Avcregc Retc to.ot 9.29 t ,5t 7.99 5.t2 t.9t t0 .72 a.g) 9 .05 1a.30 t.22 7.15 6.70
Appendix A-2
157
APPENDIX B
ASHEVILLE, NORTH CAROLINA, GASE STUDY
158
Appendix B-l
Household Size Data - Asheville, NC MSA fSource: 1990 U.S. Gensusl
Size of Household in Persons Total
TP, 5+ Households
I ." '1.
.T'-lt-."
1 71 I 0 0 0 79
2 61 33 4 o 0 104
::.:::: ii, iiiiii:i:::::::i:
3
i: ii :
0 0 0 0 0 0
.lriljl:
't,", i .,:.t,i::,!,.l[.,',i .,.,.,i.:,.:!,i .,.,t'i,i
::i:::::::::ltlt:::t:::ji:j::::::::::::::::::::::t: i:i:i:;;iili:l:i:r:::.::l:ii8a2.i:::ii:iiiiiili:;:i:l::lii:iiiili:::::2siiii:::::::::::i;iiiiiiiiiiil:;::::::iliill;:li:i:,0:i:::iiiriill
':: : t:11:itt;ti::i:i:i:i:::::
'r-." . .:.:..i.i...r':.: :.:.t,:.:.:.:.:.:::: : :.: :: ::
, j:::: j:
: :::.::::.::.::: :
.
:
: :
:
:..:: :: ::::: ::: :
:
:: :
:
:
:
:
:
:
:
:
:
:
:
:
::
' 'r'r: r .r 'rlil rrrlrrlr::rrrrl

'r .: :r 'liii:v1:r11rrr!r!M:
:r, _: 1::i :: iii::iiii:i:ii]]ii]i
7 131 110 45 24 4 314

I 114 119 59 29 26 347
I 324 80 39 o I 457
13 302 88 20 13 14 437
14 62 82 66 45 60 315
15 117 123 42 43 24 349
19 195 38 34
20 228 162 51 60 45 546
21 132 105 39 41 I 325
28 35 13
111 115 19
31 228 192 86 113 78 697

32 167 184 52 66 23 492
33 57 54 19 11 20 161
i.:i:::.:::.:::.:.:::::::2'o:i:iii..l.i.i::.:::.:::;:i::l:::::i:::t::::t:t:.tt:t:i:it:::::i:.:::t22::::iii:i.:::::::.::ii::
.,,,i.....'.':.t.i.ffi .:'..;.r,'i.i
,,'i.',;j.1j;:,i;35,'i:.li:i,'lli
i:::i:::;:::ijilrl:::itiT2iiitii:i.i:iiiiii.lili:li:iiiii:i:iiii::;::i:it::::::::i:::ii:rs8::::i::i:::ii:i:i::i
.:..;,:.,.,::.,.t:;.;.i3tt.::r,,t,:
': i:: J.i I rr, : ::rrlj:: :lillli:;:iil::illi:;:i:i:!',81i;:ii!;:i:,:,:iiiiiiiiliiill::i:ii:;r::;i::iiiililii:,::::,:,i:::34: il:::iiiiiri:iii:ili:l
37 20 39 13 18 797
38 90 170 56 49 29 394
39 7 36
159
Appendix B-l
Data . Ashevllle. NG
TAZ 5+ Households
40 111 168 90 55 23 447

41 3 3 2 3 0 11
42 36 o 28 10 083
::i:l::::ili::::::::.i:.::':ii|:::::::l:::g,i,,:i'.., ','.,::.:.:i:i:i:.:.:.:.:.:i.:.i.ii.l:.,.i.:iii:286:'l::,1:,,.
,,,,,,,,,,,,',tii,,.,,i:,:iii,ll6ii,l.'','l.ll,l.i.l.il'.i.ii.i.l.liilli:..;i.ii...:':...,.:.:i:298..:.:::l:l....l.;:..
.i.:..'..-.ii:......:...ii:i.......L::::.5::'..,.,,...'.'..:..'.ii.................i:iii.........:............63.i.ii.....,.:i......l
46 84 93 74 52 42 345
qAa
I
q
rrrlt 4eA n
vv A^
rv o ,aA
48 93 82 39 19 44 277
,49.,,.,,,,, ,,,,,, ',, " , ,,,:?/J02.,t,,,.'
.,,,,'1:40,''::,.
:50.1:,,,,',,'.,',,,i,:.:::,,,,,,,,,.' ..78:l:...:,.l ii:,,:.l;.87:,,,::::,.
t$l,:,::::::.:''l:,::.,t.',.,.,:.,:,.'::,;;:,:.:,.::,:,:,!,,:,,:t,t,:,1,8. .,,:, .40.;ti.

23 0
;i;:1,.l
52 58 90 31 202
275 132 73
104
149 132 37
59 144 170 101 16 35 466
60 52 255 106 88 0 501
64 107 120 83 112 43 465

65 97 120 105 30 18 370
66 136 124 103 74 51 488
i67..r,..
i.g',, 1,a0:', ...166, ' 136',.;.,i.',........'l..l.,.,.:.'''.'|l,04..,1,'i;,:.,,.
,.69...
70 86 217 92 66 23 484
71 98 83 90 29 49 349
72 11 168 114 93 48 434
: : ... ::::: r. ::: ::r. .i.. . .'.:r: .... ...:...: :::i::: :.. r::
"'.;'6' 4i ro' 77 4s o 333

77 65 70 15 27 18 195
78 185 290 154 74 14 717
160
Appendix B-l
Household Size Data - Asheville. NC MSA lSource: lgg0 U.S. Census
TAZ Households
l
,1
I
.:_-*,-..]
i
"_-***
l
+HjE#l
Total 13,787 16,313 7,996 5,592 2,914 46,492

t6r
Appendix B-2
StuOv nrea gmpovme
Employment Type Total

Employment
390 791 226 1,407

1,171 120 170 1,461
52 421 224 697
fl,'.f',l::::::l::::li:ii:::l:::lil:iiili:l:l:::l:l::::::,:,l:'il:i409
::iiii:ir:r:::::iii::iiiiiii:iiiiiiiiiiii:iiiiiiiiiiiiiii::iiii:iiiiiiiiiii::=2sl:l
iiji:::7.6liitti::t:::::l::::iii:iiii::i:::ji::::l:j:::::i::i::l:.i:::i:i::i:::i:iiiir8i:i;ii
358 208 106 672

.l I O.) cla têt
.+oo
9\,
t t()9 C,v
69 43 28 140
or,+...........,..:...;..,.............:...:.....i.i.'.lii::':i:..i.:.S
2.7.1:,1',1::-i:::',,'.,',:.:,,::1:,:::,,ii.'::.ii.l:.::li.il:::i:lili::l;i505
388::i:t.'l.tl.i'..:.t.:::.!,::tt,:..,.,':.,,,,,,,,,,,.,,,,i',,,,,.62d
807 535 246 1,599

19 85 33 137
98
19 48 33
20 82 126 82
21 20 7
25 50 37 83 170
26 38 13 103 154
27 493 153 455 1,101
:...'.;:...li.l....l......,.i...24..:..........iiii.,;.t.:.i.r,iit...i..:i,..,'i.,'',....,,.l;..l...:
''llll
:. :,: .: ,,.: .:,,,:iii29:i:i,:.,::,: ::,',: ,.,.:' ,::
:;.'', :' ,.
: ,,..30,,.,..,,,''.: .,:, , ,' "''

31 485 89 81 655
32 113 216 578 907
33 00 0 0
,'i..iii.,.::.....ji.il'......;i'34..i....i.', ...,..i.,'.i',.i.i,i.i. ...,,,,..',f.'',07''a...,......'.,.,..i...,.:,,,',,,.,...'.'........1,;::43'
...'...,,,,,.....':::'...,.:::::::,.:.:. . .'.,,.'i ;204
;:;:;:::::: ;:::::::::::::i:::::::::::::l:::::::lt::;:::::::::l::::::::j: ii:ii:iiii:l.i:iiiii.ireliiii:i:i:i
''.'l.;ii,'......,,. '
'.',:,':'l'.:l::l.:,:.,,:..:l.::3,5t:.:.:l.,,..':.':..,..,..i:l.:..':.,...l::,:.i.i.,:'i,i....',,i: ,:,::..: '.'.,..i. '',:815,.,,:,,:.::,,.::.,,1.:.':;::':.:':,::i.,,:.',.,;':.,.,,:,::,.98,.:
1 1.:::::j 1:: i::::: f;::::
::t.::,.:':.:.t.t.t'tl:,|:::il:t:t:;'ll:.:|:l:.::i.ti::::87|
::::.: :1::::::.:..::::.::.:.:: :.
t:r,.i1ii
';.:;:'i,:.ij,.,.::36.il..,:,.,.,,,,,,,.,,;.,;,:.,1 :,,,,;:,,,,;',,,,.,.,,'''.:,: ,,. .ti!it.!.ti].!.,.,itii.,.iti:i,.i::.,.!,,:trtri.iii.64t.
:::',; ,.'...;.,i,.i::,::,',''.....'..:'79'. .86ii:;':iiif ::lt
37 20
'.,,',,,,.,',;,.,',:43,',,,',,,.,.,
24 0
':i:.;::::;:::.:.:.:.:i::!
44
38 44 162 7 213
39 0264 30
162
Appendix B-2
Studv Area Emplovment bv Zone - Asheville. NC MSA.(9ource: NGDOT. 19901
Retail Employment
40 29
41 29
42 275
46 365 273 89 727

47 27 19 21 67
48 123 52 58 233
52 28 26 56 110
53 M 121 19 184
228
I
54 60 58 110
::.:::::::::::::::::::::::::::::::l:i:::i:::::::,:::.:::::::::::::::::::::::::::::::::::r::r1::::.::
i:i:::i::i]i- 1ii!:h hif r:i:::rir ji:i:::i]:irrl:: : r.ii:
::...::::::::::.:::.:iir:::::::::::::::::::::::::.::::::::i:::::i:i:::::':
::::::::::::::::::.:::.:r':r':.::.:::.::.::,:r':
ji:r:ji:ji:i1l!liii::
r:1:i :i
,.,,,..,,,,,.:.i','..,.;i::,9Q,,,,,,,,.,,..:,.',,,,,,.,,',,,,,,,i:,,.,',,,.",,',.., ,.
.:.:i:........r..:r::,,,......J, ., ::i.: : ::i: i:i::::t;:it: i :a:i ::.::::
.:.:: ar:. 1.:rr:?-!t!. ..::::i.i ii.: .i: :,,.,tt::1,,.:
.t. :'::::::::::j:::':::::::::::::::f::::::::::::::::::::::::::::::::::ji::::.:::i::i'::::i'::'l:
,,,3,'.,i ;,,,,,,,,.,.,',,,,,i
I
58 77 71 181
o
vv 215 158 338 711
60 241 178 198 617
64 15 28 81 124
65 59 59 58 176
66 255 37 136 428
:.',,,, l,.,,,,, '.:.,5_l::,.;..., .',',,'.,.,,,
,, ..,t,.67.,....,'.'
.. . ..,,6.8.,1;..'.,.,,.'.. '..',.,t ti:.:;'..,'' .. .:,:l.

',,,.,,1,,,.'.
.::',,,,,,:,:,i.69,iri,'i.:,,',,:,,::i ,,,',,:233:::;,,,.:,,','',
.,,'......,......,,,,,,,',,,',1'20
70 257 349 341 947

71 1,026 4 0 1,030
72 617 111 230 958
,:,iii .r:i:,:l 28rr:r:r::::::.:.:i:ii;::i:::::::::
:::::::::::;i:::::::it::::::::::::i:i:::i:::lt::::::i::::::::::::::::
iii':ii:'ii.ii,zzs i::::::::rii:ii:iil;ii..iii:
i :i,i.::i::i:il,,58iiiiiiiii;;:::r::i ii:!.iiii,i,i
76 492 31 30 553
77 345 89 559 993
78 134 36 25 195
163
Appendix B-2
Studv nrea gmpbvme
TM Employment
79 88 58 258 404
809 6 10 25
81 25 51 104 180
.:.:.:.:.:'::t::,:':,::,:,::::::i:82.li.:::::::,i.:.:::::.:.:.:.:.:::.:.:.:.::,,:,:.,.::.:,:':,:,:,::i::','::.':tt..:..l.,:,.,':.;'.:,1;,,i3-0
::ll:t:::::li:;::l:;'l::::;i:::83ii::jiiil::i:iliiiiiii:::::i:ii::i:,:li',:iiiii:i:iiilr;i:r:i;::::::::rt::::::::,:::::::::::56
l..li....:..i..tl:::i::i::;i::ti:;gf;iii:ji'::i:ti:i.:iitit:i;:li::i:i:::i:::i;:.llil:::::::::ll::i:i:::ii,.:ij:,ji,i::iljlr,,iiiiiiii:ii;iii2:ii;
85 32 3l
EA , e
87 54 35
l::4El.:.:.,.it:..:..t..::..i:...t.,:,:::t:.
::j i:: ::: r:::::::::::::::::::::i::::: r
r::: r::::::r::::
:!]46:i:i:tit,i:i:i:]:!iiiilIiiit:t i:i:i:;:
S'3,i::,i :,;':'::i:::r::::,;:::i:::,::
334 55 28 417
732 346 406 1,484
219 129 471
ri.i:.:,:1,80:,:,ii:,:,:,:
..:,l:ll::.llili:::i:i:i:ii::tii::itl::;::li:llllliilllllllll::liliiiii:llii:;S:;;H..:iil :: :::
:
:
:
:
:
:
:
:
:: :.: .;.: : i t::.:.:.: i
t:
-:.:.:
iiii;iiiiiir67i:!:!:!iii:ii
:,::::::ii;:i:iii:i23.iiiiii.il:ii:.::ji:::::::iiiiiiiiii::::iiiil:;ii:i:i:iiiiiiiiiiti:ifu:::::in:i:ril+::iii:
::::::::l:l::ii59f iiii::iit:i:i:i::::::::::::i:i::::::i::ri:i:::::::iii..r,i.iiii:iii58iii:iiiiiiiiiiiiixi:::t: iiiiiit:l-03iiiiii jtii:
97 798 104 989

98 79 124 163 366
99 306 64 261 631
ll.:::.::.t:.1:::.,iiii:0.ii.l:il::::,;:ll::,:.:'....l....
,':
- ,, ,,':il:Oili.:. .,, ,,',:,,
,:
'r
.,'.'.,,,''...',..,..1,t}j..i' .:"',
103 25 78
104 106 1,694
105 36 62
Total 27,343 16,327 15,367 59,037

164
Appendix B-3
Unbalanced Productions and Attractions - Asheville. NC MSA
Tdp Produclions Tdp Attractions
TAZ HBW HBO NHB TOTAL NHB TOTAL
1 66 175 83 321 2,U0 2,3EE 1,613 6,042

2 124 328 14E 601 2,118 1,677 1,010 4,804
30000 1,011 1,270 884 3,164
.:i:,,:::.:i!:i:i;;f ii:i:i:tiiliii:.itii.i.ii.:l:itliiii2:l:8i:1.
.:]]1:i:.:::1:i
:.i::.::::::::::::::+::::::::::::::::::::l:::::::::::::::lijiZflJ:::i:i:
. ii::1r:i:itiliii:::ri:i::=r:ir:iiii!iii..:iii:
::i:;irjiii,ii:ii0i:i:::i:i:ii:iiii:i;:ti:i::::i:::iiit;Afl9:iiiliiii:i:iii:,.,:i:i,.':jiii628iiii:iiii:iii:i:iil::
7 445 1,211 535 2,191 974 1,201 754 2,930

8 550 1,563 664 2,77E 679 1,006 635 2,319
9 490 1,337 608 2,435 203 607 357 1,167
13 484 1,330 594 2,407 2,303 2,608 1,666 6,577

14 602 1,825 734 3,160 199 527 326 1,051
15 556 1,574 666 2,796 142 436 252 830
'l5Sl'l:.::.:::;..l:l.lll,ll,e=eS
4fl;78,,,:.,,.,,,.;;,;...',;..9it.#,
,::::3Sl:iii;i:.:::.r.: :::i:::::i:::::f .:;596i

19 700 1,871 E34 3,405 1E1 797 52E 1,506
20 824 2,360 995 4,178 421 1,620 929 2,970
l 21 4E0 1,332 578 2,390 102 393 274 769
37 172 491 206 869 64 283 157 503

3E 672 1,895 799 3,366 309 1,707 892 2,907
39 67 196 80 343 44 259 129 432
43 414 1,142 495 2,052 255 1,480 763 2,498

44 426 1,1E5 517 2,127 1,143 6,336 3,010 10,489
45 465 1,39E 570 2,432 2,506 13,485 6,2E3 22,273
165
Appendix B-3
Unbalanced Productions and Attractions - Asheville. NC MSA
Tdp Productions Trip Attractions
46 620 1,831 756 3,206 1,054 3,109 1,581 5,745

47 500 1,381 594 2,475 97 379 265 741
48 458 1,357 555 2,370 338 816 483 1,636
:l::l':,::l:;:::ls?j:l:iil'.ii j.ii...i25:l::::l:iii j
.i.:i:.:.:ii:.:36liii::::l::::::.::''.:.:::.::;;1.ft.,:..,'',..'
:i:.:::::::::,:928::i:::.iiiii,iii:iiiiiiii:i:i:zi9ES::iiii:i:::i I
52 315 E51 374 1,540 160 455 289 904

53 1,234 3,513 1,484 6,231 267 't,53'! g't't 2,709
54 616 1,757 745 3,11E 331 952 5E8 1,870
:ij:ii:i.'::::.:;i'83.;l:i:::::;j::::iii:iiii;iiil:l:504:;:i:i:;l;,;::;:.ii
,;::::i;iiiiii:l5Ef ......li............illi.liiiit.,i0rg'liiii.:...,
iii:i::i:ijl:::,:799iiii:iii:iiii:t::i:,i,r;iiiiii4596,,::r::::::i:i
58 503 1,373 601 2,477 262 1,021 595 1,878

59 734 2,071 E8E 3,694 1,031 2,423 1,394 4,848
60 906 2,474 1,071 4,451 895 2,406 1,338 4,639
64 85E 2,525 1,042 4,425 636 452 1,268

65 612 1,731 747 3,091 255 86E 526 1,649
66 851 2,501 1,039 4,390 621 1,038 686 2,345
:ti:98.::::i.:.t:::,.:.:.:::::::li:::i:;.Zg:.ii:i:i:i::::::ti:::.iii::::iZillE9,,,,,,i:i,,,.
70 847 2,366 1,009 4,?22 1,373 4,194 2,211 7,778

71 606 1,E04 746 3,155 1,494 1,134 704 3,331
72 894 2,616 1,077 4,588 1,389
E02 939 576 2,317

77 316 904 377 1,598 1,440 2,159 1,306 4,905
78 1,163 3,202 1,395 5,760 283 846 603 1,731
82 424 1,192 499 2,115 413 1,552 836 2,E02

83 345 954 423 1,722 309 1,024 590 1,912
84 760 2,107 895 3,762 91 775 476 1,342
88 655 1,845 786 3,287 380 1,467 817 2,665

89 1,459 4,102 1,730 7,291 107 571 490 1,168
90 670 1,878 806 3,354 1,030 2,300 1,290 4,620
t66
Appendix B-3
Unbalanced Productions and Attractions - Asheville, NC MSA
Tdp Productions Tdp Attractions
TAZ HBW NHB TOTAL HBW HBO NHB TOTAL
91 1,221 3,400 1,442 6,063 605 1,225 EOE 2,63E

92 1,129 3,002 1,355 5,486 2,152 4,915 2,723 9,790
93 686 1,912 797 3,395 683 1,762 981 3,426
' rr:i..,:-_
97 501 1,410 603 2,514 1,4U 1,957 1,084 4,475

98 1,108 3,128 1,333 5,570 531 1,778 1,057 3,365
99 1,291 3,712 1,565 6,568 915 1,681 1,114 3,710
103 596 1,6E5 3,020 113 362 287 762

104 776 2,185 3,910 2,456 2,922 1,644 7,022
I
105 446 1,31E 2,310 90 293 224 607
o6:r.i.i.:
..j:i.:r:
ili!.ri:f g7rii,i..r:l
gNr:r:a'
.. fi
lnternal 76,033 215,407 91,566 3E3,006 85,604 188,806 109,331 3E3,740
109 6,026 7,772 3,219 17,017 2,583 5,1El 3,219 10,982

110 2,176 2,E06 1,162 6,143 932 1,870 1,162 3,965
111 886 1,142 473 2,501 380 762 473 1,614
',,1,;l ;2i::,r,:.,:.,,,',; :, :.,,i :,,;$p1t1,:||1
:i : :t :r. :tl:.. t :,i.:i.., ..i,.;.:t1:: t: : r:: i :jj:: j :.:.
1::_
;.;.,.'f;,1'3.;.1',.;.',''..: ,;,'.4n5,u03:.,,
....r.il.'l:4ii'....;'.i.:::,':.r.:i .,,i5i374l.,.
115 3,419 4,409 1,826 9,654 1,465 2,939 1,826 6,230
116 3,E69 4,990 2,067 10,926 1,658 3,327 2,067 7,052
117 7,201 9,287 3,846 20,334 3,086 6,191 3,846
ii:::,:*
13J.23
2966.'',,,..',.,..;.,,,l.28ii.iii:.;..i...i.ri:..:;:;6i494...iiji.:j.::::::ii.jii::i::....ii.jiii::986::i.:iii.i...j.;..ij.'l
li.l:f iii:si::lr:ii.ii ii:iii.i.i.::ii:iiruo.:iiiiii
'.:if ,.1 9::::.:,,.,,,:,i,,.,,'....,'. 30l'.ii:.:..i'i,
,t:t:l 20,t:': :tr'.:: ;:;:,::, : ...t 82.::.,i:,::: i:i.l..:.li::87.'8i
121 5,983 7,715 3,196 16,894 2,564 5,144 3,196 10,903

122 3,916 5,050 2,092 11,058 1,678 3,367 2,092 7,137
.
123 1,225 1,579 654 3,458 s25 1,053 654 2,232
l '. J'':;
External 48,842 62,9E6 26,0E7 137,915 20,932 41,990 26,O87 89,010
Total 124,875 278,393 117,653 520,921 106,536 230,797 135,417 472,750

Appendix 84
- Asheville. NG MSA
From External To External
Station Nurnber Station Number
109 113 Y = -7.4O + 0.55 x 10 + 24.68x 0+ 45.6i2x 16,100/ 143,100= 3.23
114 Y = -2.7O + O.21 x 30 + 67.86x 0 = 3.60
117 Y = -2.7O + O.21 x 30 + 67.86x I =71.46
121 Y = -2.7O + O.21 x 30 + 67.86x 0 = 3.60
122 Y=-7.4O+ 0.55x 10+ 24.68x 0+ 45.62x14,000 1143,1OO= 2.56
113 109 Y = -2.7O + O.21 x +30 67.86 x 0 = 3.60
117 Y = -2.7O + O.21 x +30 67.86 x 0 = 3.60
121 Y = -2.7O + O.21 x +30 67.86 x 0 = 3.60
122 Y = -7.4O + 0.55x 10+ 24.68x 0+ 45.612x 14.000 I 14 100 = 2.56
114 109 != -2.7O + O.21 x 30 + 67.86 x 0 =
3.60
117 l= -2.74 + O.21 x 30 + 67.86 x 0 =
3.60
121 l= -2.7O + O.21 x 30 + 67.86 x 1 = 71.46
122 f= -7.4O + 0.55 x 10 + 24.68 x 0 + 45.6i2x 14,000/ 143,100= 2.56
'417 109 f= -2.7O + O.21 x 30 + 67.86 x 1 = 71.46
113 l= -7.4O + 0.55 x 10 + 24.68 x 0 + 45.612 x 16,100 / 143,100 = 3.23
1'44 f= -2.7O + O.21 x 30 + 67.86 x 0 = 3.60
121 l= -2.7O + O.21 x 30 + 67.86 x 0 = 3.60
122 Y = -7.4O + 0.55x 10+ 24.68x 0+ 45.6,2x 14.OOO I 1 100 = 2.56
121 109 f= -2.7O + O.21 x 30+ 67.86 x 0 = 3.60
113 f= -7.4O + 0.55 x 10+ 24.68 x 0 + 45.612 x 16,100 / 143,100 = 3.23
114 f= -2.7O + O.21 x 30+ 67.86 x 1 = 71.46
117 f= -2.7O + O.21 x 30+ 67.86 x 0 = 3.60
122 f= -7.4O + 0.55 x 10+ 24.68 x 0 + 45.6i2 x 14.000 I 143.1OO = 2.56
122 109 f= -2.7O + 0.21 x 30 + 67.86 x 0 = 3.60
113 f= -7.4O + 0.55 x 10 + 24.68 x 0 + 45.6i2 x 16,100 I 143,1OO = 3.23
114 f= -2.7O + O.21 x 30 + 67.86 x 0 = 3.60
117 f= -2.7O + O.2'l x 30 + 67.86 x 0 = 3.60
{
168
Appendix B-5
Balanced Productions and Attractions - Asheville, NG MSA
Trip Productions Trip Attractions
HBO NHB TOTAL HBW HBO NHB TOTAL
66 175 1,351 1,592 2,477 2,990 1,351 6,818
124 328 845 1,297 2,572 2,099 845 5,516
0 0 741 741 1,227 1,590 741 3,558
445 1,211 632 2,288 1,183 1,504 632 3,319

550 1,563 531 2,644 824 1,260 531 2,615
490 '1,337 299 2j26 246 760 299 1,305
iss#i
iff:4i
i:iiiiisg
13 484 1,330 1,395 3,209 2,796 3,265 1,395 7,456
14 602 1,825 273 2,700 241 659 273 1,173
15 556 1,574 211 2,341 173 546 211 930
if ifjliHiiiiiii*iiiiiif #isil:ii
iI*iiffiiiiiii# fi l.fi fi .i:i:!:!i':t:!:!:i:iil:fr ii#::
iiiiiiiitlt:l llirllii!i111i1i1i:iiilliii:
:::l:i:::::ii::::::r:iIii:;:::::::iili
!1ii1i11:111:a i1111::il i:i:iii::fi rgii:i:!:!:::tii:!:i:i:$,il:2ui:
19 700 1,871 442 3,013 220 998 442 1,660

20 824 2,360 778 3,962 511 2,029 778 3,319
21 480 1,332 230 2,042 123 492 230 845
ig6rfi
l:i:::::::i:::::i:I:lalii:i:i:il:::::
li:iiiiiliiiii,siiliiiiiiiiiiiiiiiiii:iiiii:i:Hf:iiiiiiiiiiiiiiiilii:i:iii::"'-e',,fr,ii
iltia$6ii:::;;:i:::;iiLr:iiiiifi sh*fi ::::i;:i:::::iiii*$'illitri:

::::::::::::::: i:::::::::i:::::::::lt:::::::::::::::i:::::::::l::.::::::::::::::::::::::::::i:::i . . .
i:i:ii:i:ifiltir,aiiti::i:;:::ii:i:iiitffiiiiiiiiilii:iii:iiiiiil#iiii
25 443 1,239 362 2,044 299 846 362 1,507
26 1,191 3,307 463 4,961 271 856 463 1,590
27 797 2,133 1,437 4,367 1,939 3,618 1,437 6,993

;-
31 1,175 3,448 882 5,505 1,153 2,158 882 4,193

32 771 2,155 1,576 4,502 1,597 4,1 00 1,576 7,273
33 256 741 67 1,064 0 101 67 168
eiiil:i:!:!:iiilt::::i:i
liftiiffi-q;i
j;ifl $9i!iii:i:i:iirr ::i:ii$i
iitiiittllttt*llllli
iiiiii:!ilHFiiii:i:iii:i:i:l ri#il:f,Hii
172 491 131 77 354 131 562
672 1,895 747 375 2,137 747 3,259
67 196 108 53 324 108 485
fiS#l
43 414 1,142 639 2,195 310 1,853 639 2,802
44 426 1,185 2,521 4,132 1,387 7,933 2,521 11,841
45 465 1,398 5,262 7,125 3,042 16,884 5,262 25,1 88
t69
Appendix B-5
Balanced Productions and Attractions - Asheville, NC MSA
TM H BW HBO NHB TOTAL HBO NHB TOTAL
46 6n 1,831 1,324 3,775 1,280 3,893 1,324 6,497
47 500 1,381 222 2,103 118 475 222 815
48 458 1,357 404 2,219 410 1,021 404 1,835
52 315 851 242 1,408 194 570 242 1,006

53 1,234 3,513 763 5,510 324 1,917 763 3,004
54 616 1,757 493 2,866 401 1,191 493 2,085
:i::::::::::::::::::::::::::lr
:ii;::'::i:::::::.it
:::::i:::::::::::::sl::
.iiiilriflliiliri
58 503 1,373 498 2,374 319 1,278 498 2,095

59 734 2,071 1,167 3,972 1,252 3,034 1,167 5,453
60 906 2,474 1,121 4,501 1,086 3,013 1,121 5,220
lii .H$ilj
64 858 2,525 379 3,762 218 796 379 1,393
65 612 1,731 441 2,784 310 1,086 441 1,837
66 851 2,501 575 3,927 754 1,299 575 2,628
il;:::i:::::::::!:{tl:i
r*o$*
:;i!:i' ::ii:iifitt
:r##$
i:il:::;iffir i:iSi*g
70 847 2,366 1,851 5,064 1,667 5,251 1,851 8,769
71 606 1,804 590 3,000 1,913 1,420 590 3,923
72 894 2,616 1,052 4,562 1,687 2,707 1,052 5,446
l::::::::::::::::::i:::::::+::::::ffi V::::: :::
i::i:+to+0,*+i:ii:iiiiiiif Iiiiiiiiin*8fi iiiiniii#i.fi*ii
i*i:'liiii::t+iii::"i:#+iiiii
iiiii:ii+i11i:nlitri*iii:i
riiii:,riiii,:#Hiliiii:i:::iiii:iiiTi*Sj:iiliiiiI!::i#Sg i
116l .iii:i.iiit+li:{.tti}}i:+L}iinliittiii: i;:::i#:ii::i#fr H::iiliii::;:ii:iiiliffi :iiiii:ii"', s$fij
76 584 1,594 482 2,660 974 1,176 482 2,632
77 316 904 1,094 2,314 1,748 2,704 1,094 5,546
78 1,163 3,202 505 4,870 343 1,059 505 1,907
i:2iiil:*8
424 1,192 700 2,316 502 1,944 700 3,146
345 954 485 1,784 375 1,282 485 2,142
760 2,107 398 3,265 111 971 398 1,480
ijffiHiii iiiiii:i$i!di:ifr iii:ii*!i*.ii1!tu
i$ffi .::::i.,iI:::iti:$#.i
r:ii#-fgliriiiiril.rr:,iffi i fr
655 1,845 685 3,185 461 1,837 685 2,983
1,459 4,102 410 5,971 130 715 410 1,255
670 1,878 1,081 3,629 1,250 2,880 1,081 5.211
Y11
t70
Appendix B-5
Balanced Productions and Attractions - Asheville, NC MSA
HBO NHB TOTAL HBO NHB TOTAL
91 1,221 3,400 677 5,298 734 1,534 677 2,945
.r,. :----*.,-_j 92 1,129 3,002 2,281 6,412 2,613 6,153 2,281 11,047
93 686 1,912 822 3,420 829 2,206 822 3,857
iiiiiiiiiiiffiliiiiii iiiiii iiiiiiit#:niiiii$i4o,r'*iiiiii.liiiiiiil4iiiiiiiiiii4,1{$fiiir;iiiiriiiiii iiiiiii#fiEiiiiiii :ii*istrniiiiiiiiiiii-iiii:iiii#iliii:iiii-:iiiiiftisii
iiitii:i:i::iiiiii*i:iiii iiiiiiiiiiiili,ffiiiiiiiiiriiiiii$is#-qiiiiiiiiiliiiiiiiii'8f,.iiii:iii::...'i r'rrlriiiiiriiiiiiir: i:iif:il$i:ii:i:i :::::il$.::iiii:i:;:::iifr f,'ii:;:i::;;::iii:i,rtii[#0ji
fi;igpi;::ii $iiig#siiiiii:::rliiiiiiiiiitir,n4ji;:iiiiiriiiiiilsi$.3nj
97 501 1,410 908 2,819 1,741 2,450 908 5,099
98 1,108 3,128 885 5,121 644 2,226 885 3,755
99 1,291 3,712 933 5,936 1,111 2,104 933 4,149
103 596 1,695 241 2,522 137 453 241 831

104 776 2,185 1,377 4,338 2,983 3,658 1,377 8,018
105 446 1,318 188 1,952 109 367 188 664
lnternal 76,034 215,409 91,566 383,009 103,940 236,401 91,566 431,907
s$s"s4Pl
108 559 721 299 1,579 240 481 299 1,020

109 6,026 7,772 3,219 17,017 2,583 5,181 3,219 10,983
110 2,176 2,806 1,162 6,144 932 1,870 1,162 3,964
i!:'l::::i:::i#,1:iljii:ii!::i'li:i::::ili:ii::iiii:S$#:iii:::i:i:i',,.J:::i:iitrii:il;#
t:::l:::::::::::::::il:::::::::::i::::::::::*:::::::*::::::::::*::::::H:::::::...
ii'iiiiiif ir,Firriiii{i!il*:i. wiiiiiiiiiEii:iiiiiiiiiiifiH

l:iri::iililiiiiiil::ri:i:::i'.+:i:iiiiiiiii:i#rffi i::$i::iiiii:.si$Hn
114 5,374 6,930 2,870 15,174 2,303 4,620 2,870 9,793
115 3,419 4,409 1,826 9,654 1,465 2,939 1,826 6,230
116 3,869 4,990 2.067 10,926 1,658 3,327 2,067 7,052
iiiiiiiiiiiiiiliiiiiii
"4#iiiiiili:iii#..ffi
$,F.'9*iii:tii:i:iiiiiiiiiH#*:ri:iI:ii::.tli:::lii$il
f,"ar:iil:ri;;:r;;:r::*i:*S.il::i:::rrii::l:::i::::::ii:ii.
120 482 621 257 1,360 206 414 257 877
121 5,983 7,715 3,196 16,994 2,564 5,144 3,196 10,904
122 3,916 5,050 2,092 11,058 1,678 3,367 2,092 7,137
External 48,842 62,987 26,087 137,916 20,931 41,991 26,097 gg,00g
Total 124,876 278,396 117,653 520,925 124,871 278,392 117,653 520,916

:
L.*
,...,- :. .:. -

NCHRP RPT 365

Uploaded by

Copyright:

Available Formats

NCHRP RPT 365

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

NCHRP RPT 365

Uploaded by

Copyright:

Available Formats

NeTIoNAL CooPERATIVB HIGHWAY RsEARCH PnoGRAM

Travel Estimation Techniques

Plannng and Admnistration

Research Sponsored by the American Association of State

TRnsponrATtoN ResEnRcn Boano

NATIONAL ACADEMY PRESS

highway authorities. These problems are best studied through a

to the National Research Council and the Board by the American

and can be ordered through the Internet at:

In 1978, the Transportation Research Board published NCHRP Report tB7,

36 CHAPTER 4 Trip Distribution

Choice of Friction Factors, 45

48 CHAPTER 5 External Travel Estimation

62 CHAPTER 6 Mode-Choice Analysis

81 CHAPTER 8 Time-of-Day Characteristics

93 CHAPTER 9 Traffic Assignment Procedures

104 CHAPTER 10 Capacity Analysis

136 CHPTER 12 Case Study Application of Default Parameters

'-l 154 APPENDIX A NTPS and Home Interview Survey Data

PURPOSE AND ORGANIZATION OF MANUAL TRAVEL-DEMAND FORECASTING :

. If and how NCI1RP Report 187 was used,

. Evaluation of the impacts of transportation investments THE FOUR-STEP TRAVEL-DEMAND

r, = Auto oocupancy may bo part of mode choice

Figure 1. Travel-demand forecasting process.

The chapter on traffic assignment has three sections. The

BUI LDING A TRANSPORTATION DATABASE

INTRODUCTION Accurate transportation model calibration and validation

. Prevailing traffic volume on the link,

Area Type Considerations . Route name/number, and

TABLE 1 Base network summary

Zone Centroids (lnternal) 107

Total Nodes 1,279

Free-Flow Capacity Veh.

I Freeway 168 55 1 350

Total Links 3,272

The highway network for the region was acquired from

Traffic Count Data

NCDOT also provided a selected count map, which dis-

TABLE 2 External sttions

Stalion No. Description 1989 ADT Classification

108 Route 251 1,800 Minor

. Regional studies that consider a range of land uses and

TABLE 3 Site-specific vehicle trip generation rates

Drive-Through 265.21 1,000 sq. ft. GFA 13.3 19.3 912

Parks and Recreatlon

TABLE 4 Percent ofhouseholds by autos owned and income

urbanlzed Area slze = 200,(xx) - 499,999

Urbanlzed Area Slze = 500,fl)0 - 999,999

Urbanized Area Slze = 1,(X)0,(X)0 +

Urbanized Person Trips/ Person Trips/Household

DATA REOUIRED FOR APPLICATION

Site-Specific Vehicle Trip Rates

Residential locations Type of residence and number of dwelling unils or acres of

Persons per Household

Urbanlzed Area Slze = 50,0(X) - 199,999

Low* 3.6 6.5 9.1 11.5 13.8 6.0

Weighted Average 3.7 7 .6 I 0.6 13.6 16.6 9.2

Urbanlzed Area Slze = 2(X),0(X) - 499,999