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


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.


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 4: Frameworks and Techniques .................................................................................... 240


Session 5: Population Synthesis and Household Evolution ....................................................... 336


Session 6: Accessibilities & Treatment of Space ...................................................................... 436


Session 7: Long-Term and Mobility Choice Models ................................................................. 563

Session 7 Questions and Answers ........................................................................................... 637Session 8: Activity Pattern Generation ...................................................................................... 639


Session 9: Scheduling & Time-of-Day Choice .......................................................................... 722


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


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


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.


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


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.


8/1110

5

Session 1: Executive Perspective


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.


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.


11/1110

8

Page 3


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.


12/1110

9

Page 4


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.


13/1110

10

Page 5


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


14/1110


15/1110

12

Page 6


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.


16/1110


17/1110

14

Page 8


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.


18/1110

15

Page 9


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.


19/1110

16

Page 10


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.


20/1110

17

Page 11


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.


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.


22/1110


23/1110

20

Page 14


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.


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


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.


25/1110

22

Page 16


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.


26/1110


27/1110

24

Page 18


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.


28/1110

25

Page 19


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.


29/1110

26

Page 20


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.


30/1110

27

Page 21


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.


31/1110


32/1110

29

Page 23


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.


33/1110

30

Page 24


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.


34/1110

31

Page 25


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.


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).


36/1110

33

Page 27

Questions and Answers

27


37/1110


38/1110

35

Page 29


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.


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


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.


40/1110

37

Page 31


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.


41/1110

38

Page 32


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.


42/1110

39

Page 33


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.


43/1110

40

Page 34


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.


44/1110

41

Page 35


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


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


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


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.


45/1110

42

Page 36


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.


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)


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.


47/1110

44

Page 38


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.


48/1110

45

Page 39


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.


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


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.


50/1110

47

Page 41


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.


51/1110

48

Page 42


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.


52/1110

49

Page 43


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.


53/1110

50

Page 44


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.


54/1110

51

Page 45


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.


55/1110

52

Page 46


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.


56/1110

53

Page 47


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.


57/1110

54

Page 48


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.


58/1110

55

Page 49


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.


59/1110


60/1110

57

Page 51


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.


61/1110

58

Page 52


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.


62/1110

59

Page 53


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.


63/1110


64/1110

61

Page 55


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.


65/1110

62

Page 56

Questions and Answers

56


66/1110

63

Page 57


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.


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


68/1110


69/1110


70/1110


71/1110


72/1110


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


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.


75/1110

72

Page 3


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.


76/1110

73

Page 4


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.


77/1110


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


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


79/1110

76

environmental justicehad not yet emerged as important topics in the early days of travel

demand modeling.


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


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.


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.


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.


84/1110

81

Page 9


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.


85/1110

82

Page 10


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


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


Why activity-based models? Policies that involve coordination between individuals and time-

sensitive scheduling constraints

Demo

AB Webinar Complete

Documents

Transcript of AB Webinar Complete