| Today in the U.S., our common landscape is shaped in large part
by public consensus. This approach is in contrast to past years,
when infrastructure facilities that were deemed necessary by power
brokers could be built without much public input, and in many cases
without significant analysis. Today, however, the public consensus
required for building infrastructure is increasingly difficult to
achieve.
Many people have an improved understanding that our built environment
affects, if not governs their choices about how to organize their
lives. They also understand that, with the prevalence of computers
in our professional and personal lives, increasingly powerful analyses
and simulations are now possible. Their lives are complicated, requiring
tradeoffs in where, when, and how to spend their time, and they
have a right to expect our analytic methods to reflect that complexity.
Also, given tight public budgets, only financially justified infrastructure
projects can compete successfully for funds. Fund providers, typically
representatives of national, state or local government, expect appropriate
proof of a project’s impacts. These impacts extend well beyond
satisfying an identified need to include financial, environmental,
and other societal impacts.
The discipline of transportation planning attempts to achieve this
public and governmental consensus. Within that discipline lies the
half-art, half-science of travel demand modeling and forecasting.
This report summarizes new approaches to modeling travel behavior
that promise to aid planners in building and maintaining consensus
about our future transportation environment.
Project Planning and Travel Demand Modeling
Project planning begins with identification of a current or future
unmet need. Once a need is identified, the next step involves developing
alternatives that might satisfy that need. To evaluate these alternatives
and choose a preferred or best one, planners identify and evaluate
impacts, both positive—such as improving travel time or decreasing
emissions—and negative—such as cost, wetlands loss,
and disruption from construction.
Travel demand analysts construct models of the way people make travel-related
decisions and, thus, what future travel patterns might be. These
models can assist in the evaluation of alternatives, the construction
of alternatives, and even in the identification of unmet needs.
A proposed project’s travel-related effects on individuals
can be basic, such as changing the daily route to work or forgoing
additional activities; or more complicated, such as purchasing a
new car or shifting the home or work location to take advantage
of newly accessible opportunities.
Modern models attempt to address the range of effects outlined above.
They have evolved to include trips by most travel modes—riding
the bus or subway, walking, biking, driving alone or with others—and
to cover the full range of decisions from home and work locations
to route choice.
Travel as Tours, Journeys, and Trips
Early travel demand models focused on point-to-point trips as the
primary unit of analysis because of limitations on computer power
and modeling techniques, and the relatively simple decisions the
models were intended to support. The most promising travel model
improvement on the near horizon recognizes explicitly interdependencies
among trips. As household members become more pressed for time and
congestion worsens, the phenomena of trip-chaining has become more
prevalent. Trip chaining occurs when one or more additional destinations
are visited during a sojourn away from home. By incorporating trip
chaining in travel demand models, we can realistically address many
more reasons for traveling and better predict future travel patterns.
A person’s average daily routine consists of one or more tours,
defined as a round-trip beginning and ending at home. The main tour
of the average weekday is to the workplace and perhaps includes
lunch or a meeting away from the office during the midday. This
tour may comprise more stops on the way to and from the workplace
also, including dropping off and picking up passengers and stopping
at the store. After returning home from work, one might take another
tour, for example out to the movies and to dinner.
Think of these tours as being comprised of journeys, which are defined
as one-way sections of a tour. The example work tour above would
include a journey to work, two journeys at work (out and back) and
a journey from work. The outing for dinner and a movie would be
comprised of two journeys; the breakpoint between the two could
be defined at either of the two stops. For students, school might
form an anchor as significant as work to workers, so journeys to,
from and at school could also be defined.
These journeys are of course comprised of individual non-stop trips
of the conventional definition. However, the linkages among these
trips are preserved. Thus, models can consider the probability that
if a stop is made on the journey to work, the traveler will probably
not choose to ride the bus. Conversely, taking the bus will lead
to a smaller chance of making stops on the way to or from work.
Figure 1: Definition of Journay-Based Trip Purposes |
Recall that the goal of travel models is to predict travel volumes
and speeds on particular facilities. That requires knowing the point-to-point
movement of travelers (i.e. trips). Journeys are easily decomposed
into constituent trips, as illustrated in Figure 1.
Earlier travel demand models treated three classes of trips separately:
- Home-based work (HBW)
- Home-based for other purposes (HBO)
- Non-home-based (NHB)
In this segmentation, only the direct JTW:HBW trip would be counted
as a work trip (HBW). Figure 1 also shows how many more categories
of trips can be related to the work trip. All other trips in Figure
1 with one end at home would, in the conventional three-purposes,
be labeled as HBO, including the JTW:HB, HBSch, HBCol, HBShp, and
HBOth trips. All other trip purposes shown would have been NHB category,
including four work-related purposes (JTW:WB, JTW:NB, JAW:WB, JAW:NB),
and the trip between dinner and the store (NWR:NB). This decomposition
illustrates the futility of attempting to model as the same all the
various trip types that fall into the three classical purposes. It
also shows how recognizing that many conventional HBO and NHB trips
are related to the journey-to-work can lead to better, more intuitive
and explainable results.
Implementation
of Journey-Based Models
Current work explores at least two methods to take advantage of this
expanded classification of trips into journeys. Model development
in Honolulu attempts to model journeys directly, while model work
in New York City uses the expanded trip purpose set in a more conventional
trip-based framework. This type of modeling makes large demands on
both computer storage and model estimation techniques, and thus should
probably not be attempted in every setting.
Honolulu.
The Honolulu environment is unique. Honolulu residents have adjusted
to congestion and lack of alternate routes by forming extremely complex
travel patterns. That, along with the limited size of the island and
its transportation network, make Honolulu an inviting setting for
this state-of-the-art work. The Honolulu models predict first linkages
between home locations and workplaces. In other words, knowing where
people in general live and work, where will people who live in a specific
location work? Given these predictions for each small analysis area,
the model predicts in an integrated way the time of day of the journey,
the number and location of stops on the journey and the mode used.
This model form ensures that all workers leaving for work arrive at
the correct workplace, a desirable property not guaranteed under the
traditional trip purpose stratification.
New York
City. The modeling approach remains trip-based largely in
response to the overwhelming size of the region which, by any measure,
is the largest in the U.S. and has the most complex transportation
system. Current work uses the full set of trip purposes defined in
Figure 1. One additional purpose recognizes that JTW:HBW trips paired
on a work-tour with another direct JTW:HBW have different characteristics
than JTW:HBW trips paired with stopped trips. By adopting this expanded
purpose stratification, each trip purpose category becomes more homogeneous,
and modelers have a better chance of fitting models that explain observed
variation.
For example, the JTW-related trips that would have been generic NHB
trips previously can now be modeled knowing the characteristics of
the household making the trip. In addition, explicit recognition of
the conservation relationships among work-related trips will yield
more accurate trip distribution results. As in Honolulu, every worker
leaving for work is guaranteed to arrive at his or her correct workplace.
Conclusions
Although differing local conditions require different modeling approaches,
both methods described above make use of new understandings of the
ways people organize travel decisions. In each of these settings the
new modeling approach will more understandably replicate the way real
people make real travel decisions. Thus, they can be expected to produce
better estimates of the effects of planned projects on individuals,
and their aggregate impacts. With better information from more understandable
and believable techniques, planners will be equipped to better serve
their role in building and maintaining public consensus. |
