Changes in Screenline Traffic Volumes Over Time

4 posts / 0 new
Last post
KenCervenka
Changes in Screenline Traffic Volumes Over Time

Hello,

I hope that a few of you might be able to send me web links (via a public TMIP reply, for I think others might be interested) that contain good examples in which the following has been done as part of a "validation over time" test/check:

1. For a model calibration year forecast: the results of screenline and/or cordon line predicted-to-actual checks for that same year.

2. For a "future year" forecast that has become an observable year: the same test for that "future" year, but also a "change in screenline volumes" test.

There are lots of examples that include just #1 (so please don't send those), but finding examples that include both #1 and #2 are harder to find than hen's teeth.

Best to you,
Ken Cervenka

p.s., by "screenline" I am referring to a group of east-west roadway links that cross an imaginary north-south line (or a group of north-south links that cross an east-west line), and by "cordon line" I am referring to an even-larger group of links that cross a circle/polygon around a specific geographic area of local interest. This would, ideally, be by direction for each time period of a weekday that is represented in the traffic assignment model. Even better would be to have summaries for each vehicle class in the assignment process (e.g., autos separated from trucks), but I will be pleasantly shocked if anyone comes up with an example where this has been done for more than just a few screenlines in a region.

rob.schiffer

Ken,

This topic has interested me for years and, you're right, there's little available on such comparisons. As an archivist of old planning reports, I decided to take a stab at this by rummaging through a series of reports I helped author in 1988, at the ripe age of 28. The figures listed below are from the Brevard Area Transportation Study 2010 Plan Update... this is the "Space Coast" area of Florida (NASA, "I Dream of Jeannie", etc.). The model base year was for 1980 and the forecast year was for 2010. UTPS was the software used.

As indicated below, the model generally over-estimated traffic volumes by a significant amount on most screenlines, in spite of the base year numbers validating reasonably well. One possible cause is that land use forecasts in Florida were very optimistic at the time and did not foresee the Great Recession, which had a tremendous impact on the development industry in Florida in the latter part of that planning period. Thankfully, we update our LRTPs every five years or so... I know that Florida MPOs went through a period of using significantly reduced land use forecasts for their LRTPs, though it seems that growth is largely back to pre-recession levels now.

[cid:image002.png@01D3D8C9.9EAE5E50]

A few caveats on these numbers:

1. I have not looked at the consistency of the 2010 traffic counts in detail... count histories should be reviewed for anomalies
2. A few locations did not have 2010 AADTs available and other years were substituted
3. Some of the count locations were not ideal, in the sense that more recent street patterns have intercepted the screenlines
4. 2010 forecasted volumes are only available at the aggregate screenline level, whereas counts are available for each segment
5. I am assuming I reported these 2010 forecasts correctly back in 1988, as I don't have any loaded networks to do QC with
6. I have not yet looked at regional population forecasts and compared these to actual Census 2010 numbers but will do so

I hope to do more of this rummaging around other reports over time... thoughts and comments are welcome!

-Rob

Robert G. Schiffer, AICP
President, FuturePlan Consulting, LLC
1256 Walden Road
Tallahassee, FL 32317
850-570-8958 | 850-877-1995
rob.schiffer@futureplan.us
https://futureplan.us/
[cid:image003.png@01D3D8C9.23912FE0]

From: Kenneth.Cervenka=dot.gov@mg.tmip.org On Behalf Of KenCervenka
Sent: Thursday, April 19, 2018 12:11 PM
To: TMIP
Subject: [TMIP] Changes in Screenline Traffic Volumes Over Time

Hello,

I hope that a few of you might be able to send me web links (via a public TMIP reply, for I think others might be interested) that contain good examples in which the following has been done as part of a "validation over time" test/check:

1. For a model calibration year forecast: the results of screenline and/or cordon line predicted-to-actual checks for that same year.

2. For a "future year" forecast that has become an observable year: the same test for that "future" year, but also a "change in screenline volumes" test.

There are lots of examples that include just #1 (so please don't send those), but finding examples that include both #1 and #2 are harder to find than hen's teeth.

Best to you,
Ken Cervenka

p.s., by "screenline" I am referring to a group of east-west roadway links that cross an imaginary north-south line (or a group of north-south links that cross an east-west line), and by "cordon line" I am referring to an even-larger group of links that cross a circle/polygon around a specific geographic area of local interest. This would, ideally, be by direction for each time period of a weekday that is represented in the traffic assignment model. Even better would be to have summaries for each vehicle class in the assignment process (e.g., autos separated from trucks), but I will be pleasantly shocked if anyone comes up with an example where this has been done for more than just a few screenlines in a region.
--
Full post: https://tmip.org/content/changes-screenline-traffic-volumes-over-time
Manage my subscriptions: https://tmip.org/mailinglist
Stop emails for this post: https://tmip.org/mailinglist/unsubscribe/12678

rob.schiffer

Apparently the image of my screenline summary did not display so I am attaching a PDF.

KenCervenka

Hello Robert Schiffer,

Thanks for your (April 20) reply on the "changes in screenline traffic volumes over time" topic.  As we (all) ponder what might happen over the next 3-40 years in the real-world, and whether our models are "up to the task" of at least reasonably predicting short-range (or mid-change) changes in travel, it is interesting to note that once (or if) public agencies start to rely on "Big Data" sources (presumably fused/integrated with appropriate amounts of real-world count data) for preparation of region-wide but link-specific "traffic flow maps" for specific time periods of an average weekday, and how those flow maps change from one year to the next, this will at least make it easier to perform the screen/cordon/other "travel model validation over time" checks that are still only rarely found in practice.  Or to think more holistically about this, perhaps two validation-related checks will start to become common practice:

1) Use of "independently collected" traffic counts as a way to spot-check confirm the ground truth reliability of the "Big Data" approaches that are (essentially) estimating time-of-day traffic volumes for thousands (upon thousands) of directional roadway links;

2) Comparison of the traffic assignment model outputs of link volumes to the data-driven traffic flow maps (plus, for that matter, comparisons of predicted to actual VMT and VHT for different parts of a region).