Static assignment vs. DTA in freeway capacity planning

2 posts / 0 new
Last post
Static assignment vs. DTA in freeway capacity planning


The TMIP Forum discussion on the above topic seems to have ended, so time to stir the pot one more time....
But first, a caveat: "The views expressed here are my own and not necessarily those of US DOT."

So what is the take-away from the discussions on the Forum, plus some of the off-the-Forum discussions that have recently taken place? My personal interpretation:

-- Although there are large DTA implementations in existence (such as in San Francisco County and Phoenix), there does not appear to be a full-region DTA implementation in any large USA region with moderate-to-high traffic congestion levels, that is being used as the "traffic assignment component" of an MPO's actual multimodal regional model that is used to support an Air Quality conformity analysis.

-- There are multiple flavors of DTA packages available (commercial and stand-alone), that range from approaches which don't require any more "network detail" than what's needed for a static assignment implementation, to the "microsimulation-based DTAs" that include significantly more detail; one could write pages on the subject, but perhaps good to at least cryptically note that there are some big "differences of opinion" as to not only how "far" one should go to replace the traditional static assignment approaches used in a big-MPO model with "something else," but that just because someone is using "a DTA" does not necessarily mean it is going to be "better" than what can be done with a well-behaved static assignment approach. An issue brought up in off-the-listserv emails are the difficulties associated with "getting things right" when it comes to managed lanes.

-- At least one thing that seems to be holding DTA back from becoming more mainstream is the longer model run times when compared to static assignments, but another is a lack of clearly (and publicly) presented information that convincingly demonstrates the static-assignment models are not performing well (or not well-enough to support a local decision-making process) in high-congestion corridors, in regards to the how the model-based prediction of directional time-of-day volumes and travel times on tolled and non-tolled lanes compare to the observed.

-- At least some recognition (see Frank Milthorpe's email copied below) that we may be under-estimating what may be the REAL prediction problem here, which is the preparation of realistic OD vehicle trip tables for not just an observable year, but to predict how those tables will change over time if zonal demographics are assumed to grow in already-congested corridors at a rate that is faster than the growth in available road capacity in those corridors.

Regarding the last bullet, Norm Marshall notes "the feedback process" should, at least in theory, prevent the occurrence of over-capacity future-year demand matrices. But feedback to what: to the trip distribution model? Or is this a situation where there is a clear "lack of integration" between the plausibility of predicted zonal growth in population and employment (particularly in those parts of the region that are already experiencing noticeable congestion), if those parts of the region are not predicted to have a noticeable increase in tolled or non-tolled road capacity? The future-year "trip tables" could be modified in some mathematical fashion so that the peak hour volume/capacity ratios never exceed 1.0 on any link, but where is "the evidence" that those vehicle trip tables are likely to come anywhere close to being a plausible response to a higher-congestion future? At some point, one might think, from a technical standpoint, that "the problem" to solve is to not expect people to significantly change their desired time-of-day travel patterns (and modes of travel-including tele-work and tele-shop substitutions) to "fit" the available peak hour road capacity, but that the expected/desired "growth" in population and employment in those already-congested corridors (or specific zones in those corridors) will simply not occur.

From: [] On Behalf Of Frank_Milthorpe
Sent: Monday, April 1, 2019 8:17 AM
Subject: Re: [TMIP] Re[2]: [TMIP] Static assignment vs. DTA in freeway capacity planning

A very interesting discussion which has focussed on the assignment issues.

There has been little discussion of the models which are producing the demand matrices which are being used in the assignment.

In most major cities there is limited available road capacity today. If there are moderate levels of population growth (1% per annum or higher) in 30 years time most models that I am aware of will produce car demand estimates which when assigned with have V/C ratios > 1 for peak periods even after taking into consideration the impacts of additional congestion, predicting substantial increase in PT usage and etc. Are there examples of good practice models where these impacts are taken into consideration and assignment volumes do not result in V/C > 1?


Frank Milthorpe

Manager Major Projects Demand Forecasting | Network and Asset Intelligence
Roads and Maritime Services
Level 21 101 Miller St North Sydney NSW 2060


Projecting the present into the future is easy. If we accept that future traffic growth will be constrained by limited capacity, then it follows that much of the value we can add as modelers is in helping understand how the future will be different than the present.

Some of this is relatively easy to model, some is harder, and some likely is best considered within scenarios.

In previous posts I have suggested 2 areas that I think are practical: DTA assignment and feeding back DTA travel times -particularly to non-work trips. In moderately-congested regions, these steps shift traffic off of congested ramps and freeways in peak periods but do not suppress VMT significantly.

In highly-congested regions, relying on these mechanisms may be insufficient. Ken mentions land use which is challenging to model but also can be analyzed with scenarios. Better time-of-day modeling would be an important tool. Behavioral changes could be explored using scenarios.

The end product could be multiple plausible scenarios rather than a single impossible scenario with traffic volumes exceeding capacity.