Re: [TMIP] RE: [TMIP] Re: [TMIP] Showing how a Travel Forecasting Model Works

1 post / 0 new
curykecom
Re: [TMIP] RE: [TMIP] Re: [TMIP] Showing how a Travel Forecasting Model Works

Guy,

Thanks for responding with the various reports and presentations.

It seems that ARC has a substantial amount of post “bridge-closure” transportation field data. However it is disappointing that, after more than a year, ARC has not prepared a report checking the ABM run results against the field data. I will “stick my neck out” and daresay that the results from the ARC ABM run were not or would not be acceptable when compared to the field data. The “proof is in the ABM pudding” Guy !

Cheers.

Paul Balmacund

'Perfect is the enemy of good'

________________________________
From: grousseau=atlantaregional.org@mg.tmip.org <grousseau=atlantaregional.org@mg.tmip.org> on behalf of Guy Rousseau <grousseau@atlantaregional.org> Sent: Friday, September 28, 2018 3:53 PM
To: TMIP
Subject: [TMIP] RE: [TMIP] Re: [TMIP] Showing how a Travel Forecasting Model Works

Regarding the Atlanta Regional Commission Activity-Based Travel Demand Model, and the I-85 bridge collapse that took place in Atlanta, ARC modeling staff prepared various informational reports & presentations at the time of the incident:

* https://cdn.atlantaregional.org/wp-content/uploads/tp-mug-travelimpactof... * May 2017 Regional Snapshot: I-85 Bridge Collapse
* https://33n.atlantaregional.com/special-features/corridors-speeds
* https://33n.atlantaregional.com/transportation/traffic-patterns-followin... * http://33n.atlantaregional.com/monday-mapday/marta-usage-85-collapse
* http://33n.atlantaregional.com/friday-factday/i-85-reopening
* http://www.ampo.org/wp-content/uploads/2018/02/ARC-AMPO.pdf

Furthermore, ARC staff is just now starting to develop and document a case study of the I-85 bridge collapse & associated regional datasets for travel model development research, as a way to eventually provide data resources suitable for testing and applying new modeling tools & methods. As such, the I-85 bridge collapse offers a unique opportunity to provide travel demand and system performance empirical observations for key time periods that cover before, during and after the bridge collapse incident, including multimodal impacts, mitigation responses, traffic counts, origin-destination, speeds and travel times data catalogs. So stay tuned, more to come in the forthcoming months, into 2019.

Thanks,

Guy

Guy Rousseau
Travel Surveys & Transportation Model Development Manager
Atlanta Regional Commission
229 Peachtree St NE, Suite 100
Atlanta, Georgia 30303
P | 470.378.1565
M| 678.986.4344
grousseau@atlantaregional.com<mailto:grousseau@atlantaregional.com> atlantaregional.com
________________________
CONFIDENTIALITY NOTICE: This e-mail and any attachments are intended solely for the use of the named recipient or recipients. Any dissemination of this e-mail by anyone other than an intended recipient is strictly prohibited. If you are not a named recipient, you are prohibited from any further viewing of the e-mail or any attachments or from making any use of the e-mail or attachments. If you believe you have received this e-mail in error, notify the sender immediately and permanently delete the e-mail and any attachments, and all copies.

From: paulsush=hotmail.com@mg.tmip.org<mailto:hotmail.com@mg.tmip.org> On Behalf Of curykecom
Sent: Wednesday, September 26, 2018 3:10 PM
To: TMIP
Subject: [TMIP] Re: [TMIP] Showing how a Travel Forecasting Model Works

Hello TMIP Forumers,

In his Oct 2017 post to the forum, Bernie Alpern challenged members to “provide the sizable amount of empirical evidence (with open, readily publicly-available references to thorough documentation of such evidence) that clearly demonstrates that all of the complexity and opaqueness of the new-generation models is worth it in terms of producing demonstrably better forecasts.”

A recent survey showed that the following MPOs have developed and are using ABMs:

- New York Metropolitan Transportation Council (NYMTC; New York City)

- Metropolitan Transportation Commission (MTC; San Francisco)

- Atlanta Regional Commission (ARC; Atlanta, Georgia)

- Denver Regional COG (DRCOG; Denver, Colorado)

- Sacramento Area COG (SACOG; Sacramento, California)*

- Mid-Ohio Regional Planning Commission (MORPC; Columbus, Ohio)*

However, it seems that no one has responded to Bernie’s challenge for empirical evidence.

Many at the MPOs who are using ABMs may believe that Bernie may be asking for too much empirical evidence. However, on the other hand, I am asking for far less evidence from MPOs from real and measurable applications, not just abstract forecasts.

Let us take an example:

On March 30, 2017 a bridge on I-85 north of Atlanta collapsed due to a fire. All lanes in each direction of the highway were closed, but were reopened on May 12, 2017. Atlanta Regional Commission (ARC) ran ABM simulations to determine travel demand impacts of this closure on the highway and transit networks. This was an ideal case to check whether the ABM was “forecasting” the closure impacts such as volume changes on the network, travel-time changes on corridors, demand peak-spreading, demand mode shifts, transit volume changes on transit stations/lines/corridors, etc. Most of these data on shifts and changes were available from the ‘after closure’ data which were collected by the highway and transit agencies and were made available for comparison to the ABM outputs.

This type of comparison should be able to show the accuracy and usefulness of the ARC ABM to a real and understandable application.

Remember, general sensitivity shifts from highway to transit, and overly aggregate results on large scales should not be presented to the forum, as such shifts and results would highly likely be obtained even from sketch planning and rudimentary 4-step modeling systems.

Maybe, the other MPOs which have such case studies could also share their experiences and results.

Sincerely,

Paul Balmacund

Unemployed Consultant

"If you are not living on the edge, you are taking too much space !"

________________________________
From: Bernard.Alpern=gmail.com@mg.tmip.org<mailto:gmail.com@mg.tmip.org> > on behalf of balpern2 > Sent: Thursday, October 26, 2017 2:42 PM
To: TMIP
Subject: Re: [TMIP] Showing how a Travel Forecasting Model Works

Hi Ken,

Ken,

While it is with some trepidation that I write against the "groupthink"
that has developed over the past decade or two that behaviorally-elegant
travel models are the only "correct" thing for decent-sized metropolitan
areas to be using, I think that your cry for help speaks volumes about how
the profession has gone too far in advocating for extremely complex model
structures, the "partial explanations" of which give someone like yourself
a headache. North American metropolitan areas have become littered with
extremely complexly-structured travel models, which almost no one
understands or can explain, let alone figure out what is going on and what
needs to be fixed and how if the outputs are obviously wrong.

While the vast amounts of time and money that have gone into developing
such models may have resulted in a better understanding of travel behavior,
it is not at all clear (not to me, at least) that the new generation of
models are doing so much of a better job at generating information that is
useful to decision-makers to make them worth all of the complexity and
opaqueness that they come with. Someone please correct me if you know
about something that I don't, but after all these years and all the
resources that have been poured into the development of more
behaviorally-elegant models, I have yet to see a single study (let alone a
meaningful number of such studies) that clearly demonstrates that a
new-generation model did a better job of predicting the effects of a major
change in transportation facilities, services, and/or policies than did an
enhanced 4-step model such as we commonly used to develop in the 80's and
90's.

I know that I am not alone in feeling this way.

I also know that this is not the type of response that you were seeking.
But this highly-respected (according to the assessment of others, not
myself) and very experienced (over 35 years of professional work) travel
modeler feels that the time has come to finally ask the question "Does the
Emperor have clothes or not?", and seek answers that are well-grounded in
empirical evidence and not just theory.

I throw open the Pandora's Box in the hope that perhaps members of this
forum can provide the sizable amount of empirical evidence (with open,
readily publicly-available references to thorough documentation of such
evidence) that clearly demonstrates that all of the complexity and
opaqueness of the new-generation models is worth it in terms of producing
demonstrably better forecasts.

I apologize for having "gone off course". People who wish to are free to
provide you with the type of responses that you are looking for.

Best regards, and with tremendous respect,

Bernard (Bernie) Alpern
Jerusalem, Israel

On Thu, Oct 26, 2017 at 8:14 PM, KenCervenka
wrote:

> Many of the TMIP Forum subscribers who are close to retirement (or mostly
> already in retirement) may recall the old UTOWN Case Study examples that
> provided a way to demonstrate how a model works in just a few seconds of
> computer processing time, for a made-up city with just a few zones.
> Regardless of how one might "feel" about the usefulness of a UTOWN-type
> application in helping someone understand how a model application responds
> to changes in the model inputs, I doubt I am the only person who gets a
> headache when reading 400 pages of "model documentation" that contains
> tables showing hundreds (upon hundreds) of coefficients and constants, with
> the text between the tables offering, at best, "partial explanations" of
> what the model is actually calculating.
> > While there is value to preparation of detailed tables (and to use a
> somewhat-standard format that makes it easy to compare and contrast the
> properties of one model application to another), as well as value in
> plain-English conceptual explanations, are there any good examples of other
> approaches to help someone (other than software programmers) quickly
> understand "the details" of how a model actually works? For example, to
> prepare an easy-to-use spreadsheet that contains the different traveler
> markets for just one zone pair (e.g., each socio-economic group and travel
> purpose that is separately-represented in the mode choice model), and how
> the information that shows up in mode-specific skims is calculated, plus
> how the mode shares are calculated. Plus similar "simple spreadsheet"
> examples that show the creation of the person travel tables and conduct of
> traffic/transit assignment, but that gets somewhat complicated without a
> UTOWN-type approach because of the need to show the interactions of zone
> pairs (e.g., a spreadsheet with just one zone pair won't be enough to show
> how a distribution model works, or how traffic assignment equilibration or
> assignment-to-distribution feedback is performed).
> > So here is the request: if you are aware of some good/recent examples for
> showing via simple spreadsheets (or maybe a UTOWN-type approach for
> distribution and traffic assignment) how a travel model works, where the
> spreadsheet user can easily "play around" with simple sensitivity tests, I
> would be happy to hear from you. Or even better, please reply directly to
> this Forum post with your examples or insights, since others might also be
> interested in this topic.
> > Ken.Cervenka@dot.gov<mailto:ken.cervenka@dot.gov> > > > > > > > > > > > > > --
> Full post: https://tmip.org/content/showing-how-travel-forecasting-
> model-works
> Manage my subscriptions: https://tmip.org/mailinglist
> Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/3464
>

--
Full post: https://tmip.org/content/showing-how-travel-forecasting-model-works
Manage my subscriptions: https://tmip.org/mailinglist
Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/3464
--
Full post: https://tmip.org/content/re-tmip-showing-how-travel-forecasting-model-wo... Manage my subscriptions: https://tmip.org/mailinglist
Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/12909

--
Full post: https://tmip.org/content/re-tmip-re-tmip-showing-how-travel-forecasting-...
Manage my subscriptions: https://tmip.org/mailinglist
Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/12913