AB Webinar Complete
Transcript of AB Webinar Complete
-
8/10/2019 AB Webinar Complete
1/1110
FHWA-HEP-13-002
TMIP Activity Based Model Webinar Series
Instructors Manual
OCTOBER2012
FHWA-HEP-13-002
-
8/10/2019 AB Webinar Complete
2/1110
Notice
This document is disseminated under the sponsorship of the U.S. Department of Transportation
in the interest of information exchange. The U.S. Government assumes no liability for the use of
the information contained in this document. This report does not constitute a standard,
specification, or regulation.
The U.S. Government does not endorse products or manufacturers. Trademarks or
manufacturers names may appear in this report only because they are considered essential to the
objective of the document.
-
8/10/2019 AB Webinar Complete
3/1110
Table of Contents
Introduction ..................................................................................................................................... 1
Webinar Schedule ........................................................................................................................... 2
Acknowledgements ......................................................................................................................... 3
Webinar Content ............................................................................................................................. 4
Session 1: Executive Perspective ................................................................................................... 5
Session 1 Questions and Answers ............................................................................................. 64
Session 2: Institutional Issues for Managers ................................................................................ 69
Session 2 Questions and Answers ........................................................................................... 153
Session 3: Technical Issues for Managers ................................................................................. 156
Session 3 Questions and Answers ........................................................................................... 237
Session 4: Frameworks and Techniques .................................................................................... 240
Session 4 Questions and Answers ........................................................................................... 334
Session 5: Population Synthesis and Household Evolution ....................................................... 336
Session 5 Questions and Answers ........................................................................................... 434
Session 6: Accessibilities & Treatment of Space ...................................................................... 436
Session 6 Questions and Answers ........................................................................................... 561
Session 7: Long-Term and Mobility Choice Models ................................................................. 563
Session 7 Questions and Answers ........................................................................................... 637Session 8: Activity Pattern Generation ...................................................................................... 639
Session 8 Questions and Answers ........................................................................................... 720
Session 9: Scheduling & Time-of-Day Choice .......................................................................... 722
Session 9 Questions and Answers ........................................................................................... 812
Session 10: Tour Mode, Primary Destination, Intermediate Stop Location, and Trip Mode .... 815
Session 10 Questions and Answers ......................................................................................... 904
Session 11: Network Integration ................................................................................................ 907
Session 11 Questions and Answers ....................................................................................... 1009
Session 12: Forecasting and Application ................................................................................. 1011
Section 12 Questions and Answers ....................................................................................... 1109
-
8/10/2019 AB Webinar Complete
4/1110
1
Introduction
This document contains presentation materials from a webinar series on activity-based modeling
held in 2012. The webinar series was sponsored by the Travel Model Improvement Program
(TMIP), which was created to advance the state of the practice of travel modeling by advancingresearch and building the technical capabilities of transport agency staff. The overall goal for the
webinar series was to improve the capacity of Metropolitan Planning Organizations (MPOs) to
evaluate and deploy advanced modeling approaches, primarily focused on activity-based travel
demand modeling. The key objectives of the webinar series were as follows:
Educate staffinvolved in MPO forecasting on advanced modeling principles, theoretical
frameworks, and model components as well as identifying opportunities that activity-basedmodels offer for planning purposes that are difficult to achieve reliably or cost-effectivelywith trip-based models.
Address obstaclesto the deployment of advanced models by describing the costs andbenefits of advanced models, in relation to the costs and benefits of existing models. Costswill include staff time, consultant costs, software and hardware needs, and the time it willtake to deliver results. Benefits will include new and improved performance measures, newplanning policies that can be evaluated and improved understandings of travel behavior toprovide explanations of impacts to decision-makers.
Discuss implementationstrategies for advanced models that address specific applicationneeds, incremental deployment of hybrid models, migration from traditional 4-step planningmodels, and the resources and expectations needed to manage the development of activity-based models.
Motivate adoptionof advanced models to improve performance-based planning byexpanding the set of useful performance measures and improving the accuracy and level ofdetail of existing performance measures.
A series of twelve webinars were held to address these objectives with three different audiences
in mind: one session for the MPO executive to understand the big picture and the motivation,
two sessions for modeling managers to consider the institutional and technical issues of
developing, maintaining and updating activity-based models, and nine sessions to educate staff
on the principles, frameworks, and techniques to deploy advanced models, as well as options for
implementation.
Advanced Models
The term advanced models can include a wide variety of forecasting methods that are
developed to support transportation planning, including activity-based passenger demand
forecasting models, tour-based and supply chain freight demand forecasting models, land use
forecasting models (integrated with travel models), dynamic traffic assignment models
(integrated with travel demand models), emissions models, and cost-benefit models. Although
the focus of the webinars is on activity-based models, the material was presented in the context
-
8/10/2019 AB Webinar Complete
5/1110
2
of the larger modeling system for freight, land use, traffic, emissions, and cost-benefits so that
practitioners could evaluate their own approach within this context. A key aspect of the webinars
was to provide practical examples of the benefits of activity-based models for addressing new
transportation challenges, such as transport pricing, shifts in demographic trends such as aging
population, travel demand management strategies, and greenhouse gas emissions.
Diversity of Activity-based Models
Activity-based models have been developed by over a dozen MPOs, are being considered for
development by another dozen MPOs, and are in active use for planning applications in at least a
handful of places. There are two prominent frameworks in active use around the U.S. (CT-
RAMP and DaySim) and at least two others under development at MPOs (AMOS and
CEMDAP) as well as numerous other academic frameworks in the U.S. and abroad. The purpose
of the webinars was not to dwell on specific platforms but to educate participants on the features
in activity-based models and the differences that exist between approaches. The webinars strived
to represent the different frameworks accurately and fairly to present an objective view of the
possible options. The consultant team selected for the project included representation of the
developers of the two prominent frameworks (Parsons Brinckerhoff and John Bowman/Mark
Bradley) and representatives who have used the other two frameworks (Bhargava Sana of
Resource Systems Group for AMOS and Kostas Goulias for CEMDAP). Nearly every webinar
was instructed by a representative of each primary firm (Resource Systems Group and Parsons
Brinckerhoff) to represent the different frameworks and experiences adequately during each
webinar. Material was reviewed by a set of key technical advisors, including John Bowman,
Mark Bradley, and Kostas Goulias, to ensure that all aspects of the different frameworks are
adequately represented.
Webinar Schedule
The webinars were held over eight months in 2012, as shown inTable 1.Also shown are the date
that each webinar was held and the instructors for the webinar. In general, the first instructor
listed was the lead instructor and primarily responsible for content, though in most cases both
instructors and a number of other consultant staff contributed significantly to content as well. As
noted above, the webinar series was presented in two parts; the first three sessions focused on
agency management contemplating moving to an activity-based model for their region, while the
second nine sessions provided more technical detail on the formulation, theory, and mechanics of
activity-based models and their application to a variety of policy scenarios.
-
8/10/2019 AB Webinar Complete
6/1110
3
Table 1: Activity-Based Modeling Webinars, Dates, and Instructors
Session
Numbe
r
Description Date
Instructors
Executive and Management Sessions
1
Executive Perspective February 2 Maren
Outwater, Joel
Freedman
2
Institutional Topics for Managers February 23 John Gliebe,
Rosella Picado
3
Technical Issues for Managers March 15 Joel Freedman,
Maren Outwater
Technical Sessions
4
Activity-Based Model Framework April 5 John Gliebe,
Joel Freedman
5
Population Synthesis and Household Evolution April 26 John Gliebe,
Peter Vovsha
6
Accessibility and Treatment of Space May 17 Joel Freedman,
Kostas Goulias
7
Long-Term and Medium Term Mobility Models June 7 Maren
Outwater, Peter
Vovsha
8
Activity Pattern Generation June 28 Peter Vovsha,
John Gliebe
9
Scheduling and Time of Day Choice July 19 Peter Vovsha,
Maren Outwater
10
Tour and Trip Mode, Intermediate Stop Location August 9 Joel Freedman,
John Gliebe
11
Network Integration August 30 Joe Castiglione,
Peter Vovsha
12
Forecasting, Performance Measures and Software September
20
John Gliebe,
Peter Vovsha
Acknowledgements
This project was sponsored by the Travel Model Improvement Program (TMIP), which was
created to advance the state of the practice of travel modeling by advancing research and
building the technical capabilities of transport agency staff. The TMIP project manager was
-
8/10/2019 AB Webinar Complete
7/1110
4
Sarah Sun. The webinar series was developed and presented by a consultant team which included
Resource Systems Group (RSG) and Parsons Brinckerhoff (PB). John Gliebe served as RSG
project manager, and Joel Freedman was the PB project manager. Stephen Lawe (RSG) also
provided management support for the project. Content was developed and delivered largely by
the following staff: John Gliebe (RSG), Maren Outwater (RSG), Joel Freedman (PB) and Peter
Vovsha (PB). The following staff also provided content and presented material: Rosella Picado
(PB), Joe Castiglione (RSG), Greg Erhardt (PB), Kostas Goulias (University of California -
Santa Barbara), Bhargava Sana (RSG), Nazneen Ferdous (RSG), and Jason Chen (RSG). John
Bowman, Mark Bradley and Kostas Goulias reviewed and the material and made
recommendations. RSG staff members Bhargava Sana, Brian Grady and Sumit Bindra were
responsible for media production, setting up the webinar software and technical issues.
Webinar Content
The following pages of this document contain the content of each webinar, including the slides
and speaker notes. The questions and answers from the mid-point break and the end of the
webinar are given at the end of each webinar session.
-
8/10/2019 AB Webinar Complete
8/1110
5
Session 1: Executive Perspective
-
8/10/2019 AB Webinar Complete
9/1110
6
Page 1
Activity-Based Modeling
Session 1: Executive Perspective
Speakers: Maren Outwater & Joel Freedman February 2, 2012
TMIP Webinar Series
This is the first of twelve activity-based modeling webinars that we will conduct over the next
nine months. This session is designed as a high-level view of activity-based models, designed for
executives. The next two sessions are designed for modeling managers. The remaining nine
sessions are technical in nature and are designed for modeling staff.
-
8/10/2019 AB Webinar Complete
10/1110
7
Page 2
Activity-Based Modeling: Executive Perspective
AcknowledgmentsThis presentation was prepared through the collaborative efforts
of Resource Systems Group, Inc. and Parsons Brinckerhoff.
Presenters
Maren Outwater
Joel Freedman
Content Development, Review and Editing
Maren Outwater Joel Freedman
John Gliebe, Peter Vovsha, Rosella Picado
Media Production
Bhargava Sana, Brian Grady
2
Resource Systems Group and Parsons Brinckerhoff have developed these webinars
collaboratively, and we will be presenting each webinar together.
Maren Outwater and Joel Freedman are co-presenters. They were also primarily
responsible for preparing the material presented in this session.
Stephen Lawe is the session moderator.
Content development was also provided by John Gliebe, Peter Vovsha, and Rosella
Picado.
Bhargava Sana and Brian Grady were responsible for media production, including setting
up and managing the webinar presentation.
-
8/10/2019 AB Webinar Complete
11/1110
8
Page 3
Activity-Based Modeling: Executive Perspective
Learning Outcomes How travel demand models are used
Benefits and limitations of activity-based models
Why current models cant answer certain policy
questions
Time and resources needed to implement an activity-
based modeling system
3
At the end of this presentation, you should understand the following executive viewpoints on:
Why travel demand models are used in planning;
What activity-based models can do well and what some of the limitations and challenges
in using these models are;
What policy questions are better answered with activity-based models; and
The staff, software and hardware resources needed to implement an activity-based model.
-
8/10/2019 AB Webinar Complete
12/1110
9
Page 4
Activity-Based Modeling: Executive Perspective
Outline Overview of activity-based models and their use
Practical advantages of activity-based models
Limitations of activity-based models
Policy evaluations that benefit from activity-basedmodels
Staff and resource requirements
4
(Maren Outwater) I will cover an overview of activity-based models, including providing some
specific practical advantages of their use. In addition, I will cover some of the challenges and
limitations of using activity-based models to provide a balanced perspective (activity-based
modeling is certainly not appropriate for every agency or every purpose). Then, Joel will cover
examples of policy evaluations where activity-based models have an advantage over traditional
methods. Lastly, Joel will discuss the staff and resource requirements of activity-based models.
-
8/10/2019 AB Webinar Complete
13/1110
10
Page 5
Activity-Based Modeling: Executive Perspective
Terminology A travel demand model that produces tours
with activity stops
Activity-basedmodel
A chain of trips that begin and end at home orwork
Tours
A travel demand model that produces tripsTrip-based model
Applied at a disaggregate level, typically withgreater spatial and temporal detail
Advanced models
Integration of economic, land use, travel, trafficand air quality models
Integrated modelingsystem
5
For discussion purposes, wed like to define the following terms:
Activity-based modelis a travel demand model that produces tours with activity stops, also
called a tour-based travel model
Toursrefers to a chain of trips that begin and end at home or work; these trips are linked so that
travelers, destinations, modes and times are all consistent in the context of the tour
Trip-based modelis a travel demand model that produces trips, also called a 4-step planning
model
Advanced modelsincludes activity-based models, dynamic traffic assignment, land use,
economic and air quality models that are applied at a disaggregate level, typically with greater
spatial and temporal detail than traditional models
-
8/10/2019 AB Webinar Complete
14/1110
-
8/10/2019 AB Webinar Complete
15/1110
12
Page 6
Activity-Based Modeling: Executive Perspective
Key Concepts Activity-based models
provide sensitivities to policies and more intuitive analysisthan existing methods
produce many performance measures that are not possiblewith existing methods
do not necessarily take longer or cost more to develop and
apply than existing methods An all-new activity-based model is a similar level of effort and cost to
developing an all-new trip-based model
An incremental change to an existing activity-based model is similarin effort and cost to an incremental change in a trip-based model
6
One of the most important reasons to move to an activity-based model is to provide sensitivities
to policies that are not possible using existing methods. Pricing policies have been pushing many
MPOs into activity-based models because prior models did not have sensitivity to price on
demand, destination or route choice. Another strong benefit is that many performance measures
that are important for decision-making are now possible. For example, traveler benefits accruing
to different populations can be provided to assess the equity of transportation investments.
Now that the first wave of activity-based models have been developed, the time and cost of
developing a new model does not necessarily take longer or cost more. It is difficult, of course,to make an apples-to-applescomparison of these costs, but some agencies have developed
activity-based models with the same timeframe and costs as a trip-based model.
-
8/10/2019 AB Webinar Complete
16/1110
-
8/10/2019 AB Webinar Complete
17/1110
14
Page 8
Activity-Based Modeling: Executive Perspective
What is an activity-based travel model? Travel is a derived demandit results from the need of
people to engage in activities outside the home
Activity-based travel models are based on behavioraldecision-making theory whether to travel
where to travel to
when to travel how to travel
This makes them better suited to address policies thataffect how people make travel decisions than trip-basedmodels
8
Activity-based models are more intuitively correct than traditional models because they closely
follow an individuals decision-making process, whether to make a trip outside the home (or
engage in activities at home), where this activity will take place, and when and how to get there.
Results of activity-based models tend to be more intuitive than trip-based models also. This is
because the modeled relationships underlying in the outcome behavior are more intuitive.
-
8/10/2019 AB Webinar Complete
18/1110
15
Page 9
Activity-Based Modeling: Executive Perspective
Modeling Daily Activity Schedules
9
5 23
1-Work
7:30 A.M. 5:00 P.M.
1. Schedule Work Tour
2. Calculate residual time windows
< 7:30 > 5:00 P.M.
3. Schedule Discretionary Tour
2-Disc
7 9 P.M.
One concept in activity-based models is to model the full daily activity pattern and set schedules
to fit these activities and the travel associated with them into a single day. Typically mandatory
activities, such as work, are scheduled first and discretionary activities, such as shopping or
eating out, are scheduled into remaining time periods.
-
8/10/2019 AB Webinar Complete
19/1110
16
Page 10
Activity-Based Modeling: Executive Perspective
Modeling Trip Chains and Tours
7 trips
2 tours
4 stops
1 stop
10
Work
Home
Car
Car
Car
Grocery Store
DaycareCenter
Car
GasStation
Car
LunchWalk
Walk
Another concept is that trips are part of a larger tour that may accomplish one or more activities
and that all trips on a tour should be linked. For example, if you take your car in the morning to
work, then you must use your car for running errands on the way home. You may also go out to
lunch during the day, which represents another tour. Changes in this system may prompt you to
go home before running errands, which means more trips and possibly different destinations,
modes, or timing for these trips.
-
8/10/2019 AB Webinar Complete
20/1110
17
Page 11
Activity-Based Modeling: Executive Perspective
Why use an activity-based model? Connects travel throughout the day, similar to how
decisions are made
Is sensitive to cost, time, demographics, and policies
Allows for greater spatial and temporal detail
Allows greater household/person attribute detail.
Tracks individuals travel behavior (not averages)
11
Activity-based models are consistent in their representation of travel behavior, which produces
more consistent responses to changes in the transportation system. So, a change to the
transportation system will affect whether someone will make a trip, where they make that trip,
how and when in the same way. Trip-based models do not have the same level of consistency
throughout the process. The other important aspect about activity-based models is that there are
significantly more details and resolution on travelers, space and time, which provides more
information on transportation impacts for decision-making.
-
8/10/2019 AB Webinar Complete
21/1110
18
Page 12
Modeling Individuals in Households
Household Attributes number of persons
housing tenure
residential building size/type
number of persons age 65+
number of persons under age 18
number of persons that are part
of the family number of children
household income
number of vehicles ownednumber of workers
number of students
Person Attributes relationship to householder
gender
age
grade in school
hours worked per week
worker status
student status
12Activity-Based Modeling: Executive Perspective
For example, activity-based models can take advantage of additional household and person
attributes that are available in trip-based models in a more limited fashion. These include
household attributes and person attributes, which are listed on this slide. Activity-based models
utilize these attributes by synthesizing a population based upon Census data records.
-
8/10/2019 AB Webinar Complete
22/1110
-
8/10/2019 AB Webinar Complete
23/1110
20
Page 14
Activity-Based Modeling: Executive Perspective
Activity Purposes Work
School/College
Personal Business (e.g., Medical)
Shopping
Meals
Social/Recreational
Escort Passenger(s)
Joint Participation
Home (any activity which takes place within the home)
14
Activity-based models typically have many more purposes than trip-based models so that these
can be associated with specific land uses. Often college trips are separated from grade-school
trips, in order to send the right trips, by mode and time-of-day, to the right destination. Escorting
passengers and joint participation in travel provide the means to track the interactions of persons
in a household so that decisions that affect this joint travel are connected. Eating meals is often
modeled as a separate trip purpose from other discretionary travel.
-
8/10/2019 AB Webinar Complete
24/1110
21
Page 15
Contrasting Modeling Approaches
Trip-Based
Trips are generated from zonalaggregations of households
Each trip is independent ofevery other trips generation,distribution, mode and timing
Timing/direction of trips is notan explicit choice (fixed factors)
Travel demand is not affectedby accessibility or the builtenvironment
Market stratification limited byability to maintain trip tablesthroughout model stream
Activity-Based
Simulation of individualhouseholds and persons
Trips are chainedmodeled aspart of tours, sub-tours andlarger daily activity patterns
Starting and ending time ofactivities are modeled choices
Built environment andaccessibility variables affecttravel demand
Market stratification is afunction of individual andhousehold attributes
15Activity-Based Modeling: Executive Perspective
Many of you have employed trip-based (or 4-step) travel demand forecasting models for
planning purposes at your agencies. I am going to talk about some of the benefits and limitations
of activity-based models in a minute, but wanted to start with a simple comparison of the
approaches.
Most activity-based models simulate individual travel, whereas most trip-based models
generate aggregate zonal estimates of travel;
Most activity-based models model trip timing as a choice, whereas most trip-based
models use fixed factors for trip timing;
Most activity-based models show how accessibility and the built environment affect
travel demand, whereas most trip-based models do not; and
Trip-based models have limited market segmentation capabilities, whereas activity-based
models do not.
-
8/10/2019 AB Webinar Complete
25/1110
22
Page 16
Activity-Based Modeling: Executive Perspective
Practical Advantages: Behavioral Models behavior more intuitively and is therefore easier to
explain results
Travel is based on round trips, which is how people makedecisions
All relevant variables can affect decisions, rather than beinglimited to a few (because of disaggregate logit choice models)
This also allows for incorporation of travel time and cost(weighted by mode and destination and time of day) to beincluded in higher level models (like auto ownership and tripgeneration)
Travel behavior is modeled consistently throughout the process(e.g. trip chaining)
16
One of the best features of activity-based models is that travel choices are based on round trips
and daily activity patterns. For example:
If I need to stay late at work and there is no bus home at that hour, I will not choose to
ride transit to work regardless of how good the service is.
If I decide to run errands near work at lunchtime, then I wont need to stop on the way
home.
If I am telecommuting to work or school, then I wont need to travel at all.
If there are new tolls on the system, I may choose to shop somewhere closer to home or
on-line.
All of these factors are modeled consistently by the behavioral processes in an activity-based
model.
-
8/10/2019 AB Webinar Complete
26/1110
-
8/10/2019 AB Webinar Complete
27/1110
24
Page 18
Activity-Based Modeling: Executive Perspective
Ability to Derive Performance MeasuresShopping Trip
FrequencyTime
PeriodDistrict
Work ActivityArrival/Depar
ture TimesDistrict
Mean TripLength
Age GroupTime
Period
Trips Per Tour GenderValue of
Time
Mode Share
Income
Group
Trip
Purpose
Mode Shareof Persons
Within -mile ofTransit
ParcelsWalk
Trips/Person
Tolls paidTrip
PurposeTAZ
18
Can summarize travelbehavior metrics by
various combinationsof the activity-basedmodel dimensions
Some examples are
There are many more examples of performance measures that are possible because activity-based
models are based in individuals, which can be summarized across any number of traveler or trip
characteristics. These measures include time spent in various activities, frequency of travel for
various purposes, and person-type summaries of model outputs.
-
8/10/2019 AB Webinar Complete
28/1110
25
Page 19
Activity-Based Modeling: Executive Perspective
Practical Advantages: Spatial Detail Can be developed at a highly detailed level (parcels),
Census block level (micro-zones) or an aggregate level(zones)
Increased spatial detail (with parcels or micro-zones)provides more precision than is possible with 4-stepmodels
Used to create accessibility buffers for access toemployment, population, transit stops, paid parkingsupply, and surrounding intersection connectivity
Non-motorized and transit trips can be more accuratelyrepresented
19
Spatial detail in activity-based models has been developed at the parcel level, the micro-zone
level, or the traditional analysis zone (TAZ) level. The increased detail of parcels and micro-
zones offers more precision, more information for reporting, and more intuitive results. For
example:
Shopping activities would primarily be located on retail parcels
Each job will be filled by a single worker in that industry
The built environment can be represented by buffers of population and employment within acertain distance of transit stops or parking and by network or urban densities. For example,
transit oriented development can be specifically represented. Non-motorized travel (walk and
bike) and walking to transit also can be explicitly modeled with this additional spatial detail.
-
8/10/2019 AB Webinar Complete
29/1110
26
Page 20
Activity-Based Modeling: Executive Perspective
Practical Advantages: Temporal Detail Models are much more detailed (e.g. 30-min, 5-min, 1-
min)
Time chosen for travel is represented by the complexdemands of household members, work and schoolschedules, etc.
Trip timing is affected by congestion and tolls thatchange by the minute (dynamic) resulting in peakshifting
20
Activity-based models are typically much more detailed temporally as well. Often time is
measured in 30 minute time intervals, if not smaller. This provides benefits for evaluation of
operational strategies at the regional level as well as traffic operations at a local level. With this
additional level of detail, analysis of dynamic pricing strategies is possible.
-
8/10/2019 AB Webinar Complete
30/1110
27
Page 21
Activity-Based Modeling: Executive Perspective
Example: Jacksonville Temporal Resolution
21
1 overnight skim 9 hourly midday & shoulder skims 12 30-min peak peri od skims
EV PM EVAM MD
0%
1%
2%
3%
4%
5%
6%
7%
8%
9%
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
%o
fRegionalTravel
Here is an example of additional temporal resolution in the Jacksonville model. The variations
within a traditional broader time period are significant and may produce misleading results when
an average volume or delay is calculated.
-
8/10/2019 AB Webinar Complete
31/1110
-
8/10/2019 AB Webinar Complete
32/1110
29
Page 23
Activity-Based Modeling: Executive Perspective
Practical Advantage: Visualization of Results There are many new types of
measures that can be reported
Detailed spatial or temporaldata can be visualized quickly
Aggregated results can be
reported across many differentdimensions
23
The visualization of results in activity-based models is possible because of the additional spatial
and temporal detail and market segmentation that are contained in the models. For example, this
plot of change in real estate prices for each parcel in the Seattle region (1.2 million) shows a
positive change in price due to expanded highway capacity.
-
8/10/2019 AB Webinar Complete
33/1110
30
Page 24
Activity-Based Modeling: Executive Perspective
Limitations: Computational Challenges Tradeoffs between
Model features
Optimized software
Hardware
Run time
New, unconventional software platforms
24
One of the bigger challenges for activity-based models in the past has been the development of
new software platforms, which are now more stable than they were in the beginning. The
computational challenge for these software platforms has been the tradeoff between modeling
features, optimization of the programs, more expensive hardware and run times. Each agency
may identify one or more of these as objectives and must tradeoff the others in order to achieve
the objective. For example, if I want to limit run time, then I will need some combination of
fewer model features, more optimized programs, and more expensive hardware.
-
8/10/2019 AB Webinar Complete
34/1110
31
Page 25
Activity-Based Modeling: Executive Perspective
Limitations: Behavioral and Spatial Realism Some activity-based models have intra-household interactions to
show how travel is coordinated among household members,which adds complexity to the calibration effort
Some activity-based models have parcel-level or micro-zone datainputs to show how travel is affected by nearby land uses andaccessibility to transit; some do not because of poor data quality
Inclusion of travel times and costs at different parts of theprocess adds realism, but also adds complexity and time
Some activity-based models model have increased temporalresolutionmodel more time periodsthis adds realism andaids accuracy, but also results in more computational time anddisk storage
25
While more complexity is possible, it is not always desirable, and it should be tailored to the
region's needs. Tradeoffs for behavioral and spatial realism are inevitable. It is also important to
note that activity-based models can be developed in phases to add detail over time.
-
8/10/2019 AB Webinar Complete
35/1110
32
Page 26
Advantage and Limitation: Data
Traditional data that is
generally applicable:
Household travel surveydata
Highway and transitnetworks and zone
systems
On-board surveys
Other data desired includes:
Parking supply and cost
Built environment
Pedestrian/bike
26
Data can be limited to existing sources, but advantages of theactivity-based models will be dependent on level of detail,quality and completeness of the data
Activity-based models offer an advantage in that many new types of data can be utilized and the
models can take advantage of more detailed data. Activity-based models also can be
implemented with primarily traditional data sources, but this will limit its advantages so
incremental improvements should include enhancements to the data. Activity-based models use
traditional data in more rigorous ways, so the quality and completeness of these data are more
important (and also easier to check and correct).
-
8/10/2019 AB Webinar Complete
36/1110
33
Page 27
Questions and Answers
27
-
8/10/2019 AB Webinar Complete
37/1110
-
8/10/2019 AB Webinar Complete
38/1110
35
Page 29
Activity-Based Modeling: Executive Perspective
Example:Manhattan
Congestion
Pricing
Study
29
Central
Business
District
Congestion Pricing
Zone Boundary
Congestion Pricing
Zone Portals
One of the first activity-based model applications for a major pricing project in the United States
was the application of the New York Metropolitan Region (NYMTC) activity-based models to a
congestion pricing policy for Manhattan. The application tested a number of congestion pricing
schemes, including a cordon pricing scheme, where all auto trips crossing the zone boundaries
indicated on the slide were charged a fee.
-
8/10/2019 AB Webinar Complete
39/1110
36
Page 30
Analyzing Who pays? and How much?Type of Driver/ Group
Level of
Discount
Taxi, Transit FREE
Commercial Vehicles, Shuttles FLEET
Rental Cars & Car Sharing FLEET
Toll-payer Fee-bate $1 off
Low-Income (Lifeline Value) 50% off
Disabled Drivers 50% off
Zone Residents 50% off
Low-Emission Vehicles -
HOV/Carpool -
May be accompanied by
investment in Means-Based
Fare Assistance Program
Helps minimize administrative
impacts for businesses, and
keeps industry moving
Would require
documentation of
inability to take transit
30Activity-Based Modeling: Executive Perspective
Another congestion pricing application involved the San Francisco County Transportation
Authority (SFCTA) activity-based model. This shows an example of one of the toll policies
explored in the study. The complexity of the policy, in terms of the types of discounts offered to
different user groups, is difficult to represent efficiently with a trip-based model.
-
8/10/2019 AB Webinar Complete
40/1110
37
Page 31
Activity-Based Modeling: Executive Perspective
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
$- $5 $10 $15 $20 $25 $30
Value o f T ime ($/Hou r)
Prob
ability
Density
Income $0-30k
Income $30-60k
Income $60-100k
Income $100k+
Estimated San Francisco Resident Values of Time
31
A key assumption in any road pricing study is travelers value of time, which determines the tolls
that travelers are willing to pay to achieve certain travel time savings. We know from many
surveys and studies that values of time are situational and that they vary greatly, from person to
person and even for any given person, depending of the situation. The SFCTA model represents
this value of time variability explicitly, and doing so helps to obtain a more logical response to
tolls from the model.
-
8/10/2019 AB Webinar Complete
41/1110
38
Page 32
Activity-Based Modeling: Executive Perspective
Travel Demand Management Strategies to change travel behavior in order to reduce
congestion and improve mobility
Telecommuting\Work-at-home
Flexible work schedules (off-peak)
Rideshare programs
Scenario-based approaches necessary
Model system captures the effects of TDM policy outcomes Cannot identify which policies will affect flexible work
schedules
But can estimate the impact on transportation systemperformance of shift from a 5-day 8-hour work week to a 4-day 9+ hour work week
32
Travel demand management schemes are another policy application that activity-based models
are particularly well-suited for. Travel demand management strategies seek to change travel
behavior in order to reduce congestion and improve mobility, and include strategies such as
telecommuting, flexible work schedules, and rideshare programs. Though it is difficult for any
model to predict participation in such programs, it is possible to use a scenario-based approach in
order to model the programs effects on transport demand, congestion, and air quality. A
scenario-based approach involves making assumptions about participation rates (or borrowing
rates from other existing programs) and adjusting model demand to match those assumptions.
The model is then run to determine the impacts of those assumptions.
-
8/10/2019 AB Webinar Complete
42/1110
39
Page 33
Activity-Based Modeling: Executive Perspective
TDM Analysis: Burlington, VT Flexible Schedule
scenario
Asserted assumptionsabout:
Fewer individual workactivities
Longer individual workdurations
Aggregate workdurations constant
Target: FulltimeWorkers
0
1
2
3
4
5
6
7
8
Duration
1.0
0
2.0
0
3.0
0
4.0
0
5.0
0
6.0
0
7.0
0
8.0
0
9.0
0
10.0
0
11.0
0
12.0
0
13.0
0
14.0
0
15.0
0
%ofTours
Work Tour Duration Distribution
Original
Adjusted
Tours by Purpose (Fulltime Workers)Ori gina l Adj usted Adj /Orig
Work 94,408 78,472 0.83
School 115 140 1.22
Escort 8,070 9,023 1.12
Pers Bus 13,519 16,848 1.25
Shop 10,531 12,938 1.23
Meal 3,817 3,842 1.01
Soc/Rec 13,076 14,360 1.10
Workbased 27,949 23,211 0.83
Total 171,485 158,834 0.93
33
For example, a flexible schedule scenario was run using the Burlington, Vermont activity-based
model. The scenario assumed that there would be approximately 20% fewer work and work-
based tours as a result, but with longer work tour durations. The tour generation and time-of-day
choice models were adjusted according to these assumptions, and the model was run to
determine the impacts on other dimensions of travel.
-
8/10/2019 AB Webinar Complete
43/1110
40
Page 34
Activity-Based Modeling: Executive Perspective
TDM: DemandImpacts
~4% Reduction in overall trips
Reduced peak period andmidday travel
More early AM travel andevening travel
Fewer, and earlier, work trips
More nonwork trips in morningand evening with fewer inmidday
-4000
-3000
-2000
-1000
0
1000
2000
3000
4000
03:00
04:00
05:00
06:00
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
00:00
01:00
02:00
Difference in Trips by Time of Day
TDM
-4000
-3000
-2000
-1000
0
1000
2000
3000
4000
03:00
04:00
05:00
06:00
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
00:00
01:00
02:00
Difference in Trips by Time of Day
TDM-WORK
TDM-NONWORK
34
The results shows a 4% overall reduction in trips, with reduced peak period and midday travel,
but more early AM and evening travel (due to the longer work hours). There were also more
non-work trips in the morning and the evening, as workers seek to fulfill travel needs (such as
shopping and escorting) at other times in the day.
-
8/10/2019 AB Webinar Complete
44/1110
41
Page 35
Activity-Based Modeling: Executive Perspective
TDM: Supply Impacts Total VMT declines slightly
Reduced peak period and midday VMT,increased VMT in evening
Reduced peak period and midday delayacross all facility types, additional delay inthe evening
0
50000
100000
150000
200000
250000
300000
0:00
1:00
2:00
3:00
4:00
5:00
6:00
7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
30-minute time period
VMT by 30 Minute Period
BASE
TDM
0
200
400
600
800
1000
0:00
1:00
2:00
3:00
4:00
5:00
6:00
7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
30-minute time period
Hours of Delay - Major Arterials
BASE
TDM
0
50
100
150
200
250
300
0:00
1:00
2:00
3:00
4:00
5:00
6:00
7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
30-minute time period
Hours of Delay - Minor Arterials
BASE
TDM
0
100
200
300
400
500
0:00
1:00
2:00
3:00
4:00
5:00
6:00
7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
30-minute time period
Hours of Delay -C ollectors
BASE
TDM
35
Only slight declines were observed in vehicle-miles of travel (VMT), with slight increases in the
evening.
-
8/10/2019 AB Webinar Complete
45/1110
42
Page 36
Activity-Based Modeling: Executive Perspective
Policies: Transit Destination and mode choices for round trips (tours) affect
destination and mode choices for individual trips
Tour-level destination and mode choices consider bothoutbound and return availability, travel times and costs
Added detail from home to the transit stop and from thestop to the destination and for local walk and bike travel has
improved accuracy Transit fare passes and drivers licenses can be explicitly
represented
Built environments affect station area ridership
36
Activity-based models have also been successfully used for major transit applications, including
New Starts forecasting. Activity-based models offer a number of advantages over trip-based
models for transit analysis. Because activity-based models consider round-trip levels-of-service,
PM peak and evening transit service can affect transit demand throughout the day. Transit fare
policies can be better modeled by explicitly modeling transit fare pass ownership at a person-
level instead of a trip level. Increased spatial accuracy between the origin\destination and the
transit stop results in a more realistic representation of access and egress time.
-
8/10/2019 AB Webinar Complete
46/1110
43
Page 37
Transit New Starts Application:
Muni Central Subway
1.4 miles connecting South ofMarket to Chinatown
Third Street LRT 7.1 milesurface line (IOS = Baseline)
37Activity-Based Modeling: Executive Perspective
The New Central Subway was the first New Starts project in the United States to be evaluated
with an activity-based model. This project involved the evaluation of a 1.4-mile long
underground extension to the Third Street light-rail line in San Francisco, connecting the South
of Market area to Chinatown.
-
8/10/2019 AB Webinar Complete
47/1110
44
Page 38
Activity-Based Modeling: Executive Perspective
Work Tour Destination-Based User Benefit
38
This map shows User Benefits provided by the Central Subway compared to a baseline
alternative, specifically for work tours by destination zone. The green zones are winners; that
is, zones that see an overall improvement in mobility due to the subway. The red zones are
losers; zones that see an overall decrease in mobility due to the subway. In this particular
alternative, there are losses in mobility along the existing Embarcadero light-rail line, due to re-
routing of trains to the Central Subway corridor, causing an increase in headway and wait time.
-
8/10/2019 AB Webinar Complete
48/1110
45
Page 39
Activity-Based Modeling: Executive Perspective
Another (non-New Starts) Transit
Application: Sacramento State BRT Project
Activity-based model used to simulatecampus arrivals and departures by hour time periods
Parking lots fill up -> park further from
destination
Choice of BRT or walk from lot to
destination
39
The Sacramento Area Council of Governments (SACOG) activity-based model was used to feed
a simulation model developed for Sacramento State University in order to measure demand for a
bus-rapid transit (BRT) project. The activity-based model produces travel demand in 30-minute
intervals. The simulation model disaggregated demand to and from Sacramento State University
to a more refined zone system. Trips driving to and from campus were allocated to one of the
parking lots on campus, and their choice of mode (walk versus transit) between their campus
destination and the parking lot was explicitly modeled.
-
8/10/2019 AB Webinar Complete
49/1110
46
Page 40
BRT Boardings By Time Period
0
100
200
300
400
500
600
5:0
0
6:3
0
8:0
0
9:3
0
11
:00
12
:30
14
:00
15
:30
17
:00
18
:30
20
:00
21
:30
23
:00
Time Period
Boardings
BRT Boardings
Total Available Parking By Time Period
0
2,000
4,000
6,000
8,000
10,000
12,000
14,000
5:00
6:30
8:00
9:30
11:00
12:30
14:00
15:30
17:00
18:30
20:00
21:30
23:00
Total Spaces
The tour-based modeltracks time in hourperiods
Conventional modelsdo not have this levelof detail
Parking constraintsand policies affecttransit ridership
Temporal Analysis of BRT Parking and Boardings
40Activity-Based Modeling: Executive Perspective
The results of the Sacramento State campus area application are shown. The top chart shows how
parking spaces are utilized throughout the day. As parking lots in more desirable locations fill
up, students and faculty must park further from their on-campus destination. As that occurs, BRT
boardings (shown below) increase. BRT boardings are due to the timing of on-campus arrivals
and departures and the use of the BRT line as an intra-campus distribution system (as well as
demand from the nearby light-rail station which the BRT line also serves). Various parking
configurations were tested with the model.
-
8/10/2019 AB Webinar Complete
50/1110
47
Page 41
Activity-Based Modeling: Executive Perspective
Policies: Environment and Climate Change Disaggregate data on travel provides more accurate
estimates of emissions
Trip chaining provides better data on starts/stops
Compact Urban Form and Transit Oriented Developmentrepresented more completely through greater level of detail
Pricing and TDM are important policies for GHG
reduction
Vehicle ownership (type, age) affects emissions
41
Activity-based models have been used to test policies involving the environment and climate
change. One useful aspect of activity-based models is that vehicle-miles of travel and emissions
calculations can be traced back to the household, since non-home-based trips are modeled as part
of tours. This makes it easier to describe the effects of land-use policy on emissions.
-
8/10/2019 AB Webinar Complete
51/1110
48
Page 42
Activity-Based Modeling: Executive Perspective
Combined with Emissions Modeling
42
GHG estimates by residence parcel -- Sacramento Area Council of Governments
Here is a plot that shows greenhouse gas emissions by residential parcel, from the SACOG
activity-based model. Households residing in more urbanized areas generate relatively less
greenhouse gas emissions than households living in more rural areas, due to relatively smaller
household sizes, shorter trip lengths, and increased use of non-motorized and transit modes.
-
8/10/2019 AB Webinar Complete
52/1110
49
Page 43
Activity-Based Modeling: Executive Perspective
Evacuation Modeling:Persons Not at Home by TAZ and Hour
Atlanta Regional Commission
43
Activity-based models can be used to perform evacuation modeling. This animation shows the
height of each zone based upon the number of persons in that zone who do not live in the zone,
by hour of the day. These are persons who are traveling for work, shopping, and other out-of-
home activities, which is possible because the activity-based model tracks how people are
spending their time throughout the day. This provides an opportunity to model evacuation plans;
the simulation can be stopped for a specific time period and the behavior of each person can be
modeled based upon supplementary survey data.
-
8/10/2019 AB Webinar Complete
53/1110
50
Page 44
Activity-Based Modeling: Executive Perspective
Policies: Land Use More direct representation of different land uses
(dwelling unit type, industry categories, parks, etc.) withtypes of travel (recreation, eating out, shopping,etc.) and the households that occupy those units
Use of worker occupation better connects workers withtheir right jobs
Parcel-based and micro-area systems allow for moredetail at businesses/destinations and to aggregate atdifferent level for households
44
There are a number of advantages that activity-based models offer to better address land-use
policy. Activity-based models often use a finer spatial system than the zone, so they are able to
provide a more realistic representation of density, mixed-use land-use, and other pedestrian
environment variables.
-
8/10/2019 AB Webinar Complete
54/1110
51
Page 45
Activity-Based Modeling: Executive Perspective
Effects of Transportation Capacity on Parcel Prices
45
The Puget Sound Regional Council (PSRC) model was a hybrid model where the land use and
activity pattern generator were micro-simulated. These micro-simulation model steps were then
integrated with a trip-based destination and mode choice model. These examples come from the
activity-based part of the model. These graphs show the results from a sensitivity test where core
urban highway capacity was doubled (i.e. the same networks as the baseline with a doubling of
the lane capacities for the core urban highway facilities (I-5, I-405, I-90, and SR-520) for the
first graph and halved for the second graph). The changes in the parcel prices, along with
changes in the accessibility, filter down through the land use, workplace location choice, and
activity generation models to produce shifts in VMT (8% increase for double capacity; 10%
decrease for half capacity). Some of these shifts come from more trips and some from longer trip
lengths.
-
8/10/2019 AB Webinar Complete
55/1110
52
Page 46
Activity-Based Modeling: Executive Perspective
Effects of Transportation Improvements on Land Use
46
This slide shows the changes in population and employment at regional centers in the Puget
Sound Region (Seattle). These are centers for their transportation plan where they have targeted
new growth. Alternatives that support increases in growth in these centers are considered to be
better than alternatives that do not support this growth. MICs are Manufacturing and Industrial
Centers.
The alternatives are combinations of projects with increasing levels of pricing in each (Alt. 1 has
minimal pricing; Alt. 5 is full network system tolling). Alt. 2 has more highway projects than the
others, and Alt. 5 has more transit. The shifts in land use were modest for the alternatives, asexpected.
-
8/10/2019 AB Webinar Complete
56/1110
53
Page 47
Activity-Based Modeling: Executive Perspective
Policies: Induced (Latent) Demand Additional travel demand resulting from a transportation
investment is directly represented
Additional travel demand resulting from a change in growthpatterns due to a new transportation investment can berepresented if the model is integrated with a land use forecastingmodel
Induced demand may be tempered by changes in performanceafter the investment is in place (improved speeds on a facilityinduces more travel in that corridor, which lowers the speed)these interrelationships are important to capture induceddemand
47
Activity-based models represent the effects of transport policy on induced demand through their
inclusion of accessibility variables on tour- and stop-generation components.
-
8/10/2019 AB Webinar Complete
57/1110
54
Page 48
Activity-Based Modeling: Executive Perspective
Effects of Transportation Investments on Demand
48
These graphs show how the effects of transportation improvements on the land use changes that
we just saw also have an impact on induced demand. The activity-based demand model showed
changes to vehicles owned and number of trips made, differentiated by work and non-work
activity types.
-
8/10/2019 AB Webinar Complete
58/1110
55
Page 49
Activity-Based Modeling: Executive Perspective
Requirements: Staff Resources Need to understand discrete choice models
Need to learn activity-based models modeling process
May require more custom scripting and light programming
Helpful to understand database or statistical queries (in additionto working with matrices)
Will require time to maintain and prepare scenario databases, if
parcels or micro-zones represent land use Network codingpotentially more time-of-day networks to
code (PM in addition to AM)
49
There are a number of staff training issues to consider if an agency is contemplating adopting an
activity-based model. Many of the model components have theoretical roots in choice behavior
theory, so knowledge of discrete choice modeling is essential. In addition, the model system
application may require more custom scripting and programming than trip-based models. These
skills are necessary in order to maintain and enhance the system, but may not be necessary to run
the models. Since activity-based models produce databases containing the travel choices of the
synthetic population, it is important to have familiarity with statistical and/or database software.
There are also implications for the development of input data and the maintenance and coding of
networks, depending upon the details of how the system represents space and time.
-
8/10/2019 AB Webinar Complete
59/1110
-
8/10/2019 AB Webinar Complete
60/1110
57
Page 51
Activity-Based Modeling: Executive Perspective
Requirements: Hardware and Software Some activity-based models run on single, multi-core
processor machines, others run on clustered solutions
Hardware and runtime is a function of
Size of region\population
Number of alternatives in models
Number of feedback iterations and constraints
Several software platforms available, none throughtraditional vendors of 4-step models; these are all opensource and freely available
51
Model run times depend on several factors, the most important of which is the number of agents
in the model. Models for larger regions, such as the San Francisco Bay or Atlanta regions
typically distribute computational burden across multiple computers because the simulations are
for millions of people. Other issues that may require more computing power include the number
of alternatives in various models, extent of shadow pricing and feedback loops, type of sampling
used for models with large numbers of alternatives, number of time periods and modes skimmed,
and efficiency of program code. Another option for sharing resources is cloud computing, but
documentation is limited (less extensive than for off-the-shelf software) and support must be
negotiated.
-
8/10/2019 AB Webinar Complete
61/1110
58
Page 52
Activity-Based Modeling: Executive Perspective
Extensions: Travel Markets At their core, activity-based models cover daily person travel
generated by households (similar to existing methods)
May need separate models for other special markets
Visitors
Airports
Universities
Commercial travel
Internal\External and through-travel
Other long-distance travel
Special events
An integrated land use model would be needed to model impactsof travel activity and accessibility on urban development andland values
52
Just as with four-step models, special market models may be required in addition to the core
resident activity-based model. These markets might include visitors, airports, internal-external
travel, and other markets. These models can either be adopted from existing trip-based methods,
or developed specifically to be consistent with the activity-based model. Tour-based treatments
for many of these markets were recently developed specifically for the San Diego activity-based
model system.
-
8/10/2019 AB Webinar Complete
62/1110
59
Page 53
Activity-Based Modeling: Executive Perspective
Interpreting Activity-Based Model Forecasts Models are based on simulation, so there is random
variation across forecasts
A distribution of outcomes is more realistic, but may beuncomfortable for those looking for a single answer
Fixing random numbers can limit result to a single,replicable answer (but only one point on a distribution)
Multiple runs can be averaged
Important to conduct reasonableness checks and
sensitivity tests to gain confidence in model outputs
53
Activity-based models rely upon random number sequences to determine results. Therefore there
is random variation within and across forecasts. In such cases, it is useful to analyze a
distribution of results; particularly for model outputs in which a limited number of decision-
makers are affected (such as a local street volume, or ridership on a low volume transit route).
Such distributions are useful in order to communicate the uncertainty associated with particular
outputs. An alternative would be to fix random number seeds in order to ensure consistent results
across model runs, though it should be recognized that such methods result in only one
realization or outcome from a distribution and could be misleading. A better approach is to
average multiple runs. In all cases, it is important to conduct reasonableness checks and
sensitivity checks on models in order to ensure that models react reasonably to changes to inputs
and are ready to be used for forecasting policies of interest.
-
8/10/2019 AB Webinar Complete
63/1110
-
8/10/2019 AB Webinar Complete
64/1110
61
Page 55
Activity-Based Modeling: Executive Perspective
Further Research Advancements in modeling decisions across multiple
dimensions (destination, mode, tours, trips, schedules)
Testing models with information technology policyparameters
Integration with dynamic traffic assignment models
Transferability of activity-based models Visualizing and communicating model outputs for
decision making
55
There are many advancements being made in activity-based modeling, some of which are listed
on this slide. They include advancements in discrete choice models related to modeling many
alternatives and multiple dimensions simultaneously, integration with dynamic traffic assignment
models, the transferability of activity-based models, and software and techniques to mine and
visualize the data produced by activity-based models.
-
8/10/2019 AB Webinar Complete
65/1110
62
Page 56
Questions and Answers
56
-
8/10/2019 AB Webinar Complete
66/1110
63
Page 57
Activity-Based Modeling: Executive Perspective
2012 Activity-Based Modeling Webinar SeriesExecutive and Management Sessions
Executive Perspective February 2
Institutional Topics for Managers February 23
Technical Issues for Managers March 15
Technical Sessions
Activity-Based Model Framework March 22
Population Synthesis and Household Evolution April 5
Accessibility and Treatment of Space April 26Long-Term and Medium Term Mobility Models May 17
Activity Pattern Generation June 7
Scheduling and Time of Day Choice June 28
Tour and Trip Mode, Intermediate Stop Location July 19
Network Integration August 9
Forecasting, Performance Measures and Software August 30
57
Thank you for joining us this week. The next webinar will be held in three weeks, and will cover
institutional topics for managers.
-
8/10/2019 AB Webinar Complete
67/1110
64
Session 1 Questions and Answers
Joel, you mentioned adjusting parameters for policy tests for demand management, how is that
done?
Joel: Let's say I have a daily pattern activity model that shows if workers go to work, do a non-
work activity, or stay home. If we're looking at policy of increased telecommuting, that's a pretty
complex thing to show direct impact in the model because it's difficult to know who has the
ability to telecommute, which employers allow it, where they are located, etc. We make an
assumption that 10% of the work force is going to participate, or even select by 10% of
downtown workers. We then can make that assumption in the constant for 'stay home' for
workers and then re-run the tour generation component of model. Then we make conclusions
about how working at home affected travel. Workers would still be allowed to make non-work
tours during day (i.e. shopping, getting lunch, picking up kids). You are changing the alternative
specific constant for 'working at home,' not the mode-specific constant on any particular mode.
Have you calibrated to TDM policies and if so what data was available? What were your
experiences?
Maren: Testing TDM policies really has to do with sensitivity. Many of these policies being
tested don't exist, or they exist in some form in the base year. We calibrate the base year to make
sure right number of people are working at home, right number of people go to particular
locations, right number of people work 8-hour days, etc. As Joel said, we have to make
assumptions about how many people participate in a TDM program. Then we use model to test
the impacts of the policy with that assumption of how many participants.
Joel: Models haven't been calibrated to TDM policies exactly, but have been compared to other
research to make sure they are reasonable.
Maren, what kind of built environment data are typically used?
Maren: A wide variety. Data being used has to do with land use, i.e. square footage of different
buildings types, amount of open space, and distance from parcels to open space. Other examples
might be based on 'area type,' i.e., is an area a CBD or suburban. Activity-based models are
flexible in being able to incorporate these types of variables, so a wide variety of variables exist
in different models.
How can we incorporate seasonal variation?
Joel: We are building a model for Phoenix now, which will have seasonal variation since there
are large differences between times of year. For example, residents go on vacation in summer
and schools are out. In winter, people vacation in Phoenix and residents are home. One way to do
this is to change the way synthetic population works. For example, in summer, the synthetic
-
8/10/2019 AB Webinar Complete
68/1110
-
8/10/2019 AB Webinar Complete
69/1110
-
8/10/2019 AB Webinar Complete
70/1110
-
8/10/2019 AB Webinar Complete
71/1110
-
8/10/2019 AB Webinar Complete
72/1110
-
8/10/2019 AB Webinar Complete
73/1110
70
Page 1
Activity-Based Modeling
Session 2: Institutional Issues for Managers
Speakers: John Gliebe and Rosella Picado February 23, 2012
TMIP Webinar Series
-
8/10/2019 AB Webinar Complete
74/1110
71
Page 2
Activity-Based Modeling: Management Institutional
AcknowledgmentsThis presentation was prepared through the collaborative effortsof Resource Systems Group, Inc. and Parsons Brinckerhoff.
2
Resource Systems Group and Parsons Brinckerhoff have developed these webinars
collaboratively and we will be presenting each webinar together.
John Gliebe and Rosella Picado are co-presenters. They were primarily responsible for
content, along with Joel Freedman.
Stephen Lawe is the session moderator.
Content development was also provided by Peter Vovsha.
Bhargava Sana and Brian Grady were responsible for media production, including setting
up and managing the webinar presentation.
-
8/10/2019 AB Webinar Complete
75/1110
72
Page 3
Activity-Based Modeling: Management Institutional
2012 Activity-Based Modeling Webinar SeriesExecutive and Management Sessions
Executive Perspective February 2
Institutional Topics for Managers February 23
Technical Issues for Managers March 15
Technical Sessions
Activity-Based Model Framework April 5
Population Synthesis and Household Evolution April 26
Accessibility and Treatment of Space May 17
Long-Term and Medium Term Mobility Models June 7
Activity Pattern Generation June 28
Scheduling and Time of Day Choice July 19
Tour and Trip Mode, Intermediate Stop Location August 9
Network Integration August 30
Forecasting, Performance Measures and Software September 20
3
For your reference, here is a list of all of the webinars topics and dates that have been planned.
As you can see, we will be presenting a different webinar every three weeks. Three weeks ago,
we attempted to provide a somewhat high-level executive view of activity-based modeling.
Today, we will be covering the second in topic in the seriesInstitutional Topics for Managers.
Our objective is to get into a bit more depth on the issues that we have found to be important to
the people we have talked to in our work in activity-based model development. Today we will be
talking about what it takes to transition between a trip-based model operation and one that relies
primarily on an activity based model. We will be talking about development time and costs,
resource allocation, and issues related to productivity.
So, in this webinar we will try to stay away from the more technical issues surrounding activity-
based modeling. As you can see by the schedule, there will be plenty of technical detail in the
remainder of the series.
-
8/10/2019 AB Webinar Complete
76/1110
73
Page 4
Activity-Based Modeling: Management Institutional
Learning Outcomes Typical motivations and concerns of agencies
considering an activity-based model
Familiarity with the evolution of activity-based modelsin the U.S.
Development options for migrating from 4-step toactivity-based models
Resources needed to implement an activity-basedmodel program
Experience with stakeholder acceptance and use
4
Our audience today is composed of modelers from public agencies, consulting firms and
academic institutions. We also know that there are managers of various levels among you. Our
goal in this webinar is to provide you with more of the institutional context for how travel
demand modeling has evolved to the point where we are today in which there seems to be a
growing demand for more advanced modeling tools. Accordingly, at the end of this webinar you
should have a good understanding of the motivations and concerns that public agencies have
when contemplating moving to an activity based modeling system. To begin to address some of
those concerns, it is helpful to review how activity-based models have evolved over the last
decade or so in different parts of the U.S. To make things a little more concrete, well discuss the
various options that some agencies have followed in developing their activity-based modeling
systems. Resource requirements are always an important issue, and we will share with you some
examples of what some agencies have invested in consulting fees, data development, hardware
and software, and staff resources. Finally, we will discuss some of the experiences to date of
users of these systems, including project use and potential use by stakeholders.
-
8/10/2019 AB Webinar Complete
77/1110
-
8/10/2019 AB Webinar Complete
78/1110
75
and validation using local data. This allows the agency to get started fast. We will discuss these
three strategies in more detail later in the webinar.
Page 6
Activity-Based Modeling: Management Institutional
Universal Transportation Modeling System
(UTMS)
Developed in 1950s
4-step process
Limited by data availability and computing power
Primary applications were planning for highwaycapacity--emphasis on vehicle trips and flows
Reliance on simplified trip-based approach
Aggregate relationships
6
In order to provide context for our discussion, lets step back in time and review how we got
here. Travel demand models were first used in the U.S. during an era in which the Interstate
Highway System was being planned. It was an era of suburban expansion and a post-war baby
boom. Consequently, the focus on modeling efforts in those days was highway capacity
planning. Needless to say, computing power was not nearly what is today, so the process that
was developed, which became the UTMS, was necessarily simple. It was based on the predictionof aggregate trips being generated from zones, composed of aggregations of households and
businesses, distributed between zones, and assigned to a network to determine how well the
network would perform.
Some of the difficult questions that transportation planners face todaygreenhouse gas
emissions, travel demand management, congestion pricing, transit-oriented development and
-
8/10/2019 AB Webinar Complete
79/1110
76
environmental justicehad not yet emerged as important topics in the early days of travel
demand modeling.
-
8/10/2019 AB Webinar Complete
80/1110
77
Page 7
7
Trip-Based Models Today: Advanced UTMS
Transportation
System
Land Use / Activity System
Network Assignment
Trip (End) Generation
Trip Distribution
Mode Choice
Level of
Service
Between
Origins and
Destinations
Travel times and costs
The trip-based models of today are really just advanced version of the UTMS process. Here you
see what many of us know as the familiar 4-step process, consisting of trip generation,
distribution and mode choice. Over time, the profession has added explicit representation of
transit and, in some places, pedestrian and bicycle travel modes. With the introduction of discrete
choice models to the profession, models based on utility theory and estimated from individual
observations were an early improvement, although in the end they are still applied to
aggregations of trips rather than to individual travelers. In addition, trips are assigned to
networks that typically represent peak and off-peak travel periods, which provide some
differentiation between level-of-service conditions during different parts of the day.
Another major improvement is the feedback loop in which travel times and costs are fed back
turned into skims tables and fed back into trip distribution and mode choice. This has long been
standard practice in the U.S. It is interesting and relevant to point out here that feedback loops
were mandated as the result of legal challenges and became a recommended best practice for
consistency for air quality modeling. When a particular interest group opposes a proposed action
-
8/10/2019 AB Webinar Complete
81/1110
78
based on a forecast, they challenge the methods used to produce the forecast. In the case of
feedback loops, critics pointed to the need for consistency between the travel times being
produced by the network assignment process and the representation of travel times and costs
being input to the demand models.
-
8/10/2019 AB Webinar Complete
82/1110
79
Page 8
home
in zone X
work place
in zone Y
(work)
restaurant
in zone W
(lunch)
grocery store
in zone V
(shopping)
7:30 A.M.
8:00 A.M.12:00 P.M.
12:10 P.M.
12:50 P.M.1:00 P.M.5:00 P.M.
5:30 P.M.6:00 P.M.
6:30 P.M.
by auto
on foot
on foot
by autoby auto
home
in zone X
work place
in zone Y
(work)
restaurant
in zone W
(lunch)
grocery store
in zone V
(shopping)
7:30 A.M.
8:00 A.M.12:00 .
12:10.
12:50 P.M.1:00 P.M.5:00 P.M.
5:30 P.M.6:00 P.M.
6:30 P.M.
by auto
on foot
on foot
by autoby auto
Modeling a Day in the Life
8
HB Work
HB Shop
Non-HB
Non-HB
Non-HB
Lets consider how people really travel. Here weve depicted an individual who goes to work at
7:30 a.m., arriving at 8. Around 12 noon, this person walks to lunch and then returns to her work
place at 1. She leaves work at 5 p.m. and stops at the grocery store before going home.
The way this would typically be represented in the trip-based modeling world would be the
following. (Step through HB work, HB shop, and three Non-HB trips). The HB-Work and HB-
Shop trips are in the AM and PM Peak periods. One of the Non-HB trips is in the PM Peak, and
two Non-HB trips are in the off-peak period. We know their modes and trip lengths.
One question that transportation planners typically struggle with is how to explain to stake
holders in your area the impact of particular project, plan or policy on non-home-based trips?
What does a non-home-based trip mean to them? A trip-based model assumes that all of these
trips are independent of one another. It does not account for the fact that all of these trips are
actually part of one large daily activity pattern, anchored around a mandatory work activity. A
trip-based model does not account for the fact that trips are chained into tours and that there is
actually a work-based sub-tour within the larger tour.
-
8/10/2019 AB Webinar Complete
83/1110
80
It also does not account for the fact that, because this person walked to lunch, they do not have
their car available to get back to the office. Further, a trip-based model would not recognize that
this person needed to arrive at work at 8 a.m. and therefore, did not have the time to drive her
son to school since his school is in the opposite direction. So, he has to take the bus. Nor would a
trip-based model recognize that this worker needed the car for work on this particular day
because her planned agenda included a big grocery shop after work. The trip-based model would
also not recognize that persons who work in this location are likely to go out for lunch more
today than ten years ago, because there are now more dining opportunities within walking
distance of this office. An activity-based model would take into account all of this additional
information.
-
8/10/2019 AB Webinar Complete
84/1110
81
Page 9
Activity-Based Modeling: Management Institutional
Why activity-based models? Activity-based models provide more information than
trip-based models
Intuitive models of behavior
Consideration of individuals, not just groups of households
Tour concepts (how trips are actually organized and scheduled)
Spatial, temporal, modal consistency between trips in same day
Motivation for travel in activity participation (substitution betweentravel and other means of meeting personal and household needs)
Interpersonal linkages and obligations
Effects of accessibility (urban form) on travel generation
Long-term and short-term decision perspectives represented
9
All of the additional information that an activity-based model takes into account are important,
because in real life trips are not independent from one another and people do not respond to
changes in transportation system level of service changes or policies as if they were. In real life,
trips are organized into tours that make them interdependent. People plan activities at the end of
the day that cause them to make certain travel decisions at the beginning of the day. Mode
choices may be somewhat constrained by household linkages and obligations, such as taking care
of children. The opportunities presented by surrounding land uses may induce people to make
more or fewer discretionary stops. And in the long-run, people do make choices of where to live,
work, go to school, and whether and what types of vehicles to own that are at least partiallybased on the transportation environment.
From a technical perspective, this comes down to accurately representing the actual alternatives
available to people in their activity-travel choices. What is really in their choice set? What are
their real short- and long-term elasticities? We will cover the finer points of choice sets and
elasticities in future webinars.
-
8/10/2019 AB Webinar Complete
85/1110
82
Page 10
Activity-Based Modeling: Management Institutional
Why activity-based models? Policy questions related to willingness or ability to pay
Fuel prices, mileage taxes and other operating costs
Parking costs
duration-based fees, employer subsidies
Road pricing
Variable time-of-day tolls (congestion/time of day)
Area pricing
HOT/HOV lanes
Transit fare policies (individual discounts, monthly passes)
Environmental justice
Impacts on minority or disadvantaged populations
10
Instead, lets talk about policies. How can we better estimate peoples response to changes in
travel costs? For example, how can we better estimate change in VMT as a function of gasoline
prices? If gas prices this summer reach a new all-time high in the U.S., will people take more
transit? Travel less frequently? Make shorter trips? Car pool? Buy more fuel efficient cars? or
forego family vacations and eating out? If high prices persist, will some people choose to work
closer to their residences? These are all legitimate responses that we observe in data, or at least
anecdotally.
These same set of responses are relevant for other policy examples, too. This slide also lists anumber of policies related to how people value their time when faced with changes in travel
costsroad pricing, transit fares, environmental justice. Trip-based models typically do not do a
good job of capturing the multi-faceted response of real people, because the basic unit of analysis
is the individual trip. Important contextual information is simply not there. In addition, trip-based
model make aggregate-level predictions for households of a certain type, but are unable to
-
8/10/2019 AB Webinar Complete
86/1110
83
distinguish between individuals within households. Consequently, they tend to do a poor job of
portraying how individuals value their time.
Page 11
Activity-Based Modeling: Management Institutional
Why activity-based models? Policies that involve coordination between individuals and time-
sensitive scheduling constraints
Demo