Acknowledgments
This project was funded
jointly by Mack Blackwell Transportation Center (MBTC) with matching funds from
the Kansas Department of Transportation (KDOT) through the Kansas
Transportation Research and New Developments (KTRAN) research program. MBTC is
a national center at the
Disclaimer
The contents of this report
reflect the views of the authors who are responsible for the facts and accuracy
of the data presented herein. The contents do not necessarily reflect the views
or the policies of
Quick Response Community Planning
Abstract
The objective of this report
is to present the outcomes of a research project conducted for formulating a
unique and shortcut procedure for travel demand modeling and forecasting future
traffic for small cities or urban areas with population less than 15,000, using
a GIS platform. The TransCAD academic version 3.2 was used as the GIS platform. The major focus has been to reduce the time
and cost for the overall travel demand modeling process. Also, the level of detail of the network
required for building an appropriate travel demand model for such small areas
was determined. As a case study for application
of the method, McPherson, a small city in
Two travel demand models were
developed using a different number of Traffic Analysis Zones (TAZs). One consisted of 21 TAZs within the study
area and was called the low-density-zoning scheme. The second consisted of 63 TAZs and was
called the high-density-zoning scheme.
From the resulting values of a screenline analysis and average daily
vehicle distances of travel estimated from the traffic volumes assigned to the
network by the models, it was concluded that the travel demand model developed
from the high-density level of network provided better results as compared to
the low-density network. However, in
regard to the values of traffic volume assigned to the proposed bypass, the
travel demand models for the two levels essentially performed the same.
It was determined that the GIS/TransCAD platform provides useful tools for data organization and analysis of results through graphical features. It was concluded that there is a need of updating technical resources for applying these modern methods of travel demand modeling.
TABLE OF CONTENTS |
PAGE NO. |
List of Figures
|
ii |
List of Tables
|
iv |
Acknowledgements
|
v |
Chapter 1:
Introduction, Background and Objectives
|
1 |
|
Chapter 2: Selection of Software |
10 |
|
Chapter 3: Information Collection
and Data Preparation |
12 |
Chapter 4:
Developing Traffic Analysis Zones
|
19 |
|
Chapter 5: Building A Travel Demand Model |
31 |
|
Chapter 6: Results |
35 |
|
Chapter 7: Use of QRSII in
Modeling Small Urban Area Travel |
48 |
Chapter 8: Discussion and Conclusions |
58 |
|
References |
61 |
Appendix 1:
Socioeconomic Data for Census Tracts Within Study Area
|
63 |
|
Appendix 2: Graphs from KDOT
Corridor Study |
69 |
|
Appendix 3: Vehicle Kilometers
Travel Data for McPherson |
70 |
LIST OF TABLES |
PAGE NO. |
Table 3.1
Employment Categories Based on SIC Code
|
16 |
Table 3.2
Network Attributes
|
18 |
|
Table
3.3 Capacity for Links Based on Functional Classification |
18 |
|
Table
6.1 Demographic and Socioeconomic Data of TAZs for the Low Density Level of
Zoning |
35 |
|
Table
6.2 Demographic and Socioeconomic Data of TAZs for the High Density Level of
Zoning |
36 |
Table 6.3 Screenline Analysis Results
for Assigned Traffic Volumes From Original TAZ Data
|
40 |
Table 6.4
Screenline Analysis Results after Increasing Trip Productions
|
41 |
Table 6.5 Total
Length of Streets According to Functional Classification
|
42 |
Table 6.6
Average Daily Vehicle Miles of Travel
|
43 |
|
Table 6.7 Average
Daily Vehicle Miles of Travel for Links With Traffic Counts for Low Density
Level of Zoning |
43 |
|
Table
6.8 Average Daily Vehicle Miles of Travel for Links With Traffic Counts for High
Density Level of Zoning |
44 |
|
Table 6.9 Linear
Regression Analysis for Low Density Level of Zoning |
45 |
|
Table 6.10 Linear
Regression Analysis for High Density Level of Zoning |
46 |
Table 6.11 Comparison of Traffic
Assignments on the North Bypass
|
47 |
Table 7.1 Comparison of Traffic Assigned
to North Bypass for Low and High Levels
|
54 |
Table 7.2 Screenline Analysis for the
QRSII Models
|
57 |
LIST
OF FIGURES
|
PAGE NO. |
Figure 1.1 Traditional Travel Demand
Forecasting Process [Source: UTPS (2)]
|
1 |
|
Figure
1.2 Using A Calibrated Travel Demand Model for Forecasting Future Trips [Source: UTPS (2)] |
2 |
|
Figure 1.3 Street Network
Within the City Limits of McPherson and Proposed Bypass |
9 |
|
Figure
3.1 TransCAD Window for Locating Employers by Address Matching |
15 |
|
Figure
3.2 TransCAD Window for Preparing Population and Housing Data for Census
Blocks |
17 |
|
Figure
4.1 TAZs for the Low-Density-Zoning Scheme |
20 |
Figure 4.2 TAZs for the High-Density-Zoning Scheme |
21 |
|
Figure
4.3 TransCAD Window for Preparing Population and Household Data for TAZs for
Low Density Zoning Scheme |
23 |
|
Figure
4.4 Microsoft Excel Window for Calculating the Number of Employees for
Different Employment Types for Each TAZ in Low-Density-Zoning Scheme |
25 |
|
Figure
4.5 TAZs and External Stations for Low Density Zoning |
26 |
|
Figure
4.6 TAZs and External Stations for High Density Zoning |
27 |
|
Figure
4.7 Street Network for Low Density Zoning |
28 |
|
Figure 4.8a Street Network for High Density Zoning
for the Whole Study Area |
29 |
|
Figure 4.8b Street Network for High Density
Zoning; Expanded View for the Central
Area. |
30 |
|
Figure
5.1 Screenline Used for Model Evaluation |
33 |
|
Figure
6.1 Scaled Representation of Links According to Directional Speed |
38 |
|
|
|
|
|
|
LIST
OF FIGURES
|
PAGE NO. |
|
Figure
6.2 Scaled Representation of Links According to Directional Capacity |
39 |
|
Figure
6.3 Plot of Assigned Traffic Volumes vs. Traffic Counts on Streets on the
Screenline and the Respective Regression Lines |
41 |
Figure 6.4 Plot of Assigned Traffic Volumes vs. Traffic Counts on Streets on the Screenline and the Respective Regression Lines for Increased Trip Productions |
42 |
|
Figure
6.5 Plot of Assigned Traffic Volume vs. Traffic Counts and Regression Line
for Low Density Network of Zoning |
45 |
|
Figure
6.6 Plot of Assigned Traffic Volume vs. Traffic Counts and Regression Line
for the High Density Level of Zoning |
46 |
|
Figure 7.1 Street Network with CBD
Bypass |
49 |
|
Figure
7.2 Existing Network with Traffic Volumes for the Low Density Network |
50 |
|
Figure
7.3 Existing Street Network with Traffic Volumes for the High Density Network |
51 |
|
Figure
7.4 Low Density Network with North Bypass Included |
52 |
|
Figure 7.5 High Density Network with North
Bypass Included |
53 |
|
Figure 7.6 Low Density Network with CDB Bypass Included |
55 |
|
Figure 7.7 High Density Network with CDB Bypass Included |
56 |
Chapter 1: Introduction, Background and Objectives
Introduction
Urban travel demand forecasting is the process of
predicting the impact of various changes in a study area on future travel
characteristics and demand. The changes
may be due to various policies and programs for development of the study area,
which results in change in socioeconomic characteristics, improvement and
changes in transportation facilities, and other factors. In general, travel demand forecasting
attempts to quantify the amount of travel on transportation system, where, the
travel demand is created by separation of urban activities and transportation
system is represented by characteristics of highway and transit network. The Traditional process for travel demand
forecasting consists of four steps: Trip Generation, Trip Distribution, Mode
Split and Traffic Assignment. Figure 1.1
below describes the four step modeling process along with information needs and
the feedback process.
Figure 1.1 Traditional Travel Demand Forecasting Process [Source: UTPS
(2)]
As shown in the Figure 1.1, the inputs to the
four-step model are the urban activities and highway and transit network
information for the study area. These
parameters may be modified after evaluating the impact of traffic volumes
assigned to the highway and transit networks making changes to land use
forecasts necessary. The modified
parameters again serve as input to the model and the process is carried out
until no further changes are made to the input parameters. The four-step travel demand model is
calibrated using existing transportation characteristics of the study area and
then is used to forecast and study the transportation characteristics in a
future year. Figure 1.2 below shows the
use of a calibrated model for future trip forecasting.
Figure 1.2 Using A Calibrated Travel Demand Model for Forecasting Future Trips
[Source: UTPS (2)]
It is important to realize that modification in
transportation system policies and street networks have impact on land use
changes and economic growth (1)*.
Therefore the process of feedback of modified input parameters to the
model and evaluating the system is necessary for any transportation planning
process.
Information
Requirements:
The information required as data input to build the traditional travel demand
forecasting model are divided into four broad categories (2) as follows:
· Study area
·Urban activities
· Transportation system
· Travel
* Numbers in Parentheses Indicate References.
The information on the study area should be known
before proceeding with collecting any other data. The boundary of the study must be defined at
the beginning. The boundary must
consider the scope of the project for which the study is being conducted. A general urban study usually includes the
surrounding area that is expected to be urban within twenty years. Project specific studies must include the
points of decision for routing traffic.
The study area has to be divided into several
Traffic Analysis Zones (TAZ s) according to land use, physical barriers and
street classification as well as the scope and limit of projects. Urban activities consist of demographic and
socioeconomic data of the study area.
Some of the specific data include population, dwelling units, automobile
ownership, income, household size and employment locations and
characteristics. The transportation
system consists of the present street classification and characteristics. This information is used to build a transportation
network for the study area. Travel information includes how, when and
where people are currently travelling.
The information is studied to determine the underlying factors causing
people to make certain travel decisions so that models can be calibrated and
used to forecast how people will travel in the future.
It was realized that conducting manual surveys for
collecting this necessary information is expensive and time consuming. Literature reveals the efforts made to
develop different techniques for the travel demand forecasting process (3, 4,
5, 7, 8, 10 and 11) to reduce time and money spent for data collection. The advent of computers with larger storage
capacities and faster processing speeds helped in a proliferation of different
techniques which provide efficient means of data management. The following section discusses some of the
present methods for travel demand forecasting.
Background
Quick Response Methods and Computer Application:
National Cooperative Highway Research Program
(NCHRP) Report 187 (3) provides
different transferable parameters, factors and manual techniques to enable
simplified and quick response urban travel demand estimation without conducting
extensive surveys for data collection.
The parameters provided by the report were based on several research
results and can be used for computerized methods. Various shortcut manual techniques are
suggested in this report.
NCHRP 187 was updated in NCHRP Report 365 (4) , and extensive manual procedures discussed
in the former report were excluded. The
report provided transferable parameters that can be applied in any available
travel demand softwares. Also the use of
geographic information systems (GIS) for building travel demand forecasting
model database are discussed in the report.
Smith and McFarlene (5) evaluated an internal volume
forecasting (IVF) model as a replacement for conventional urban travel demand
models in small and medium sized areas.
The model used interzonal trip probability factors as a measurement of
the magnitude of the trip between zones.
The trip probability factors were calculated as a function of production
and attraction variables of zones, interzonal travel time impedance and related
exponential factors. Instead of
conducting surveys and home interviews for estimating productions and
attractions for each zone, the model used production and attraction variables
such as population, employment and rating for commercial and recreational use
of each zone. The model was applied to
calibrate a travel demand model for the
V =
181 + 0.031 ´ Pi Pj
----------------------------------------------------(Equation 1.1)
Where, Pi is the
population of zone i and Pj is the population of zone j.
The model was used to forecast traffic volumes for
future years using trip probability factors, which were calculated from
forecasted changes in production and attraction variables. The model used linear regression method for
calibration.
Other non-linear models have been proposed by
Hogberg (6). Also, the model did not
consider intrazonal travel time impedance.
The inclusion of origin-zone accessibility was recommended by the
study. The major advantage of this
approach was the elimination of conducting extensive surveys for collecting
socioeconomic data for estimating zonal productions and attractions, and using
ground traffic counts for calibration.
The study concluded that the model was well suited for application in
small urban areas with population less than 50,000.
McPherson, Heimbach and Goode (7) set forth a
computerized method for updating data bases for travel demand forecasting. The study focused on development of a
computerized system capable of monitoring and recording changes in urban
activities that could be used to update transportation planning data bases
periodically. Two major analytical
techniques were used to develop the model, as follows:
i)
Computerized Geocoding:This technique provided the
mechanism for linking together a variety of diverse data sources that may have
only one common link (e.g., street addresses).
The process provided a convenient framework for continuous monitoring
and updating transportation planning data.
ii)
Income Model Construction
Using Harmonic Analysis: Harmonic analysis was used to express income-estimating model as a
Fourier series. Data for calibrating
income models were available from the 1970-1980 census urban transportation
planning packages.
The study expressed the necessity of developing an income model because of confidentiality requirements by law prohibited state and federal revenue agencies from releasing income tax return files to other agencies. Conclusions of this study were supported by the findings from the
system
implemented and tested in
Rutherford and Pennock (8) discussed the features
and application of the software ‘Quick Response Microcomputer System’ (QRS)
developed by Comsis Corporation, to incorporate the techniques of travel demand
forecasting presented in NCHRP 187 (3).
The software could be used for rapid calculation of zonal productions
and attractions. Some of the limitations
of the original software were:
· maximum number of TAZs was limited to 50,
· maximum number of assignment links was
limited to 800, and
· maximum travel time allowed was 40 minutes,
The socioeconomic data, such as number of dwelling
units, income, auto-ownership, retail and non-retail employment were input to
the trip generation procedure. The
travel times were calculated from zone types, coordinates of zone centroids,
map scale and a circuity factor to convert airline distance to over-the-road
distance. Rutherford and Pennock
believed that the mode choice model of the software was not a particularly
useful tool, as it required inputs which might be less than logical.
Goulias and Kitamura (9) presented a model for
forecasting travel demand based on microanalytic simulation and dynamic
analysis of travel characteristics. The
system consisted of two components: a microsimulator of household socioeconomic
and demographic data and a dynamic model system of household auto ownership and
mobility. Each component comprised
interlinked models formulated at the household level. The interactions and causal paths that
underlie life cycle evolution of individual households were replicated in the
socioeconomic and demographic microsimulator.
Simulation units evolved from year to year, experiencing changes in
urban activities. Employment, income,
education level, household size were among the variables that were generated in
the microsimulation. The parameters of
the model for generating the socioeconomic and demographic data were estimated
from observed data from Dutch National Mobility Panel Data. The model system provided a flexible tool for
forecasting travel demand; however, the process was complex and required a
large amount of data for estimating the parameter which triggered the
simulation process.
Application of
Geographic Information Systems (GIS):
Computer based GIS integrates data from diverse
disciplines and various formats to generate useful information about an area of
earth or physical features on earth such as street networks, location of
employment centers, etc. GIS software
packages are capable of capturing, organizing, processing and analyzing
spatially referenced data. The physical
features are associated with related databases and are represented graphically
to scale. In recent years, the Census
bureau has made census data available to be used with GIS, by producing TIGER
(topologically integrated geographic encoding and referencing) files. GIS shows promise as an important tool for
transportation professionals.
Mugler and Quinn (10) explained the use of GIS
technologies for preparation of demographic forecasts in the Denver Regional
Council of Governments (DRCOG). They
subdivided the 1990 census tracts based on land use to create TAZs. Demographic attributes of the TAZs were
extracted from TIGER files. GIS
processing was used to create population and employment estimates from state
income tax roles, census data, electric utility records and privately developed
databases. The data were geocoded on the
TAZ polygon layer and the road base map.
Raley (11) expressed the efforts of Delaware
Department of Transportation (DelDOT) in using GIS based census data and TIGER
files for cost effective transportation planning. DelDOT had developed modified planning grids
in 1960s, which were comprised largely of census geography. The grids were updated to provide current
TIGER geography, and were used as building blocks for TAZs. Census data were used for assignment of population
and employment forecasts to the modified grid level.
Gan (12) formulated a GIS-aided procedure for
creating TAZs from census data and TIGER files.
He expressed the problem concerning data disaggregation when census and
TAZ boundaries are non-coterminous. For
presentation purposes he used the TAZ system of
Slavin (13) proposed a fundamental reformulation of
the travel demand modeling process based upon a richer spatio-temporal
conceptual and empirical approach. The
proposed approach focussed on improvements and extensions to traffic assignment
models, and combined the following four elements:
· modeling joint choices as super-networks,