Traffic demand modeling a quasi-science?

[southern_yank]

I deal with a lot of urban design projects which involve traffic studies. Often, if a level of service E or F shows up in an alternative for out years (which, aside from the failing LOS, may be good urban design), this alternative is automatically taken off the table. I have a problem with this.

I know how these traffic projections are derived (also akin to knowing how sausage is made), and there are too many unknowns and flimsy variables involved to give me certainty that for X intersection in the year 2030, Y number of cars will travel through it causing Z delay. Traffic is more self regulating than we give it credit for (see Anthony Downs), and in applying fluid dynamics to human behavior, I think we water down both sciences into something more half-assed.

I'm interested to hear other people's thoughts on this and also rationales that can give me more faith in traffic models.

[JusticeZero]

In all honesty, I have less confidence in modeling than you do. The models are generally self-fulfilling prophecy, and cases of dramatic traffic evaporation are both well documented and apparently impossible under modeling theory. Car technology is on it's way out; oil peaking is coming up rather hard; and I am hesitant to add much more expensive public infrastructure for it that does not meet alternate goals.

[urban burban]

You are half-right: traffic demand modeling is bogus.......when engineers use it. First of all, common travel demand model platforms (i.e. EMME2/3, FORTRAN, Corsim, VISSIM, VISSUM) don't properly account or accomodate land use variables or land use growth - which is largely the parent variable of those so disdained LOS E's and F's.

Second, engineers (and planners to some extent) often fail to question or adjust forecast policy. For example, if the Town of Somewhereville has been experienced population loss over the last five to ten years, and does not expect to see any major industry growth for the next 20 years - why forecast travel demand out 20 years?! I think engineers and planners fail to realize that travel demand modeling is intended to be used as a tool for guiding policy direction, not forecasting the future.

[transguy]

I don't see the issue as being with the model as much as with the user of the information provided by the model. Too many people expect that they will be hard concrete data from the model. People need to understand and accept the limitations and assumptions that go into modeling. This is a process that can, and is very often, abused by people who aren’t sure what they’re doing.

[DetroitPlanner]

Traffic modeling is not a quasi science, but intepreting the results is an art.

[waterloowarrior]

"It is popularly assumed that when universities give science degrees in traffic engineering, as they do, they are recognizing aboveboard expert knowledge. But they aren't. They are perpetrating a fraud upon students and upon the public when they award credentials in this supposed expertise."

[maxxoccupancy]

While it may seem naive to some on this board, I was actually surprised and angry to find that the public sector unions operate in a completely different culture, trying to get as much money as they can out of the taxpayer--rather than being frugal, efficient, and honest. Of course, the incentives are to spend more (especially at the end of the fiscal year), but there's even a tendency, nowadays, to create artificial crises in traffic, schools, and even crime, in order to justify raising taxes.

The interest groups who really get your tax money even try to change the culture in your area in order to panic people into "needing" those really high tax rates.

Welcome to Cyburbia, maxx. A reminder of two of our rules. 2.11 Single issue posters / one-trick ponies The Forums are not intended to furnish people with a venue for single-mindedly promoting their personal agenda. Staff may ask users to limit or refrain from posting on a certain topic. 2.14 Dissenting opinions There are some people that register not to participate, but to disrupt civil discourse and promote a personal agenda. Posts critical of the planning profession phrased in a confrontational manner ("All you ☭commie☭ planners are freedom-hating social engineers who want to take away our cars", "You steam-era planners want to force us back into crowded tenements", etc ) are not permitted. Users with a history of making such posts, even on other message boards or Usenet, may be banned.

[Gotta Speakup]

These traffic models are an attempt to understand and quantify what are really extremely complex interactions. They are based on lots of assumptions and the results are difficult to interpret. But there is little alternative ways to do what they do.

[Masswich]

Travel Demand Models appear very scientific, However, they are based on the gravity model, which is only scientific when applied to large masses such as planets and moons that actually have gravity. The application of the idea of gravity to traffic is clever, and perhaps as good as we can do right now. But there's no reason to assume that travellers behave like planets, hurled towards attractions like meteorites.

Of course, standards TDMS also take land use as an input and traffic as an output. Land use will respond to traffic as much as traffic responds to land use, esp. over a 20 year time frame.

So while I think TDM has a role in big picture things like air quality monitoring, I don't think its theoretical basis is particularly compelling. I personally wouldn't make a land use decision based on what a TDM tells me about LOS 20 years from now.

Years ago I worked in a resort region doing their TDM. The model worked terribly in determining traffic patterns under existing conditions until I added a new "attraction" to the database- parking spaces at beaches. I had no empirical reason to do so, nor did I know how many cars a day would be attracted by a parking space. But it sure got the model to work better!

[Gotta Speakup]

Travel Demand Models appear very scientific, However, they are based on the gravity model, which is only scientific when applied to large masses such as planets and moons that actually have gravity. The application of the idea of gravity to traffic is clever, and perhaps as good as we can do right now. But there's no reason to assume that travellers behave like planets, hurled towards attractions like meteorites.

!

What makes this even more scary is that gravity models can predict the orbit of the moon around the earth, or the earth around the sun, but they cannot function once there are more than two objects. At that point, the models break down. So we try to do on earth that which fails in the heavens.

[Tide]

The problem is with the variables. However, if you read This Article at Wired.com it speaks about scientific theory and models, I found it interesting. Basically it says all models are wrong because they work in a positive or negative results. But in the age of Google and massive cloud computing the computing power is so strong they can simply search for correlations, which don't need to be explained, but simply that they exist. Then science can work on finding out why.

For example, to get a true traffic count ideally you would count 24/7/365 and find the trends, instead we look to a table and the type of road. Also, wouldn't it be great is not only did we record 24/7/365 data but also GPS tracked each trip through an intersection or road to find out if there are some anomalies? Awesome, and it doesn't really matter why but the simple fact that it exists.

Just something to chew on, I never liked traffic models, they are often too conservative or completely miss the mark.

[Masswich]

The problem is with the variables. However, if you read This Article at Wired.com it speaks about scientific theory and models, I found it interesting. Basically it says all models are wrong because they work in a positive or negative results. But in the age of Google and massive cloud computing the computing power is so strong they can simply search for correlations, which don't need to be explained, but simply that they exist. Then science can work on finding out why.

For example, to get a true traffic count ideally you would count 24/7/365 and find the trends, instead we look to a table and the type of road. Also, wouldn't it be great is not only did we record 24/7/365 data but also GPS tracked each trip through an intersection or road to find out if there are some anomalies? Awesome, and it doesn't really matter why but the simple fact that it exists.

Just something to chew on, I never liked traffic models, they are often too conservative or completely miss the mark.

Interesting idea, but isn't that Data Mining? Doesn't the theory have to predate the correlation? Its been a long time since grad school...

[andreplanner]

I know nothing about traffic impact studies. So it doesn't mean much. Ya ya. Useless info.