Great Lakes Regional Transportation Operations Coalition

Mobility Performance

Multistate Mobility Performance Measures

Outreach Materials

The Great Lakes Coalition contracts with the Wisconsin TOPS Lab for administration and project execution. Through the work by the TOPS Lab on performance measures, including utilization of the NPMRDS (see below) for GLRTOC and the Wisconsin Department of Transportation’s performance reporting, a variety of outreach materials have been produced. A selection is posted here. Be in touch with Peter Rafferty with any questions.

NPMRDS Interstates Map

See the NPMRDS 2013 or 2014 links above/left.

The Federal Highway Administration (FHWA) has acquired a National Performance Management Research Data Set (NPMRDS), consisting of five-minute travel times nationwide, for use in performance measurement and related purposes. Since October 2013, this data set is being made freely available to States and Metropolitan Planning Organizations (MPOs) for their own use. This is made available through the Freight Performance Management (FPM) and Urban Congestion Relief (UCR) programs. The data are furnished by Nokia’s HERE business unit (which encompasses NAVTEQ), and consists of the following:

  • Average travel times;
  • Five-minute bins;
  • National Highway System (NHS) coverage, namely interstates and US highways;
  • By traffic message channel (TMC) links;
  • By passenger, freight (5+ axle, class 7-8 per ATRI), and combined;
  • With GIS shapefiles;
  • From October 2011 to present, updated monthly; and
  • On a state by state basis, with a usage agreement required.

For more information on NPMRDS, refer to the FHWA page on performance measurement where you’ll find a link to a FAQ and archived webinars.

Obviously there are many ways to analyze, slice, and visualize this data set. What is depicted on the maps are just one way to paint a picture of what performance looks like from the data. It shows the planning time index (PTI) – a measure of travel time reliability – on all interstates nationwide (not AK or HI), based on all 2013 data, and split by the freight and passenger columns in the NPMRDS. In addition to normal zooming, you can hold the shift key and click and drag a box to zoom in to. Also be sure to try the vertical left-right slider to compare freight vs passenger measures.

We used the same PTI scales between freight and passenger in part to not obscure their inherent differences and in part to illustrate observed differences in the NPMRDS, e.g., while freight travel times are higher on average, PTIs are lower than passenger, or put another way, times in the freight column are more tightly distributed than the passenger column.

The PTI used here is the ratio of the 95th percentile travel time to the free flow travel time, as shown on the following summary graphic from SHRP2 (click the graphic for a larger view).


The free flow in the denominator follows a UCR definition (off peak 15th %ile). The timeframe for purposes of this depiction takes all 2013 lumped together – about 2 billion records for interstates. We recognize this will trouble some who think of reliability being relevant for something like weekday non-holiday peak period travel – which is exactly what we do for other purposes – but that is not the point here.

The link color symbology is not tied to any specific meaning or interpretation of PTI or travel time reliability. It allows the data to speak for itself as those colors are simply quintiles dividing observations into five equal bins by TMC count.

Other comments, given this is a research data set, with intentionally missing and outlier observations:

  • If < 10 percent possible observations present (over the year), the TMC is either not shown on the map if both freight and passenger absent, or will be gray “No Data” if one or the other is absent.
  • A lightweight outlier scrub is in place, per TMC, per fright or passenger, over the year, for times greater than 4.5 standard deviations above the mean (arguably an arbitrary number we landed on for another purpose, but it removes only a small fraction of a percent of very high times (hours long), which equally arguably minimally affects the identification of 95th percentile times).
  • Note the problem posed by missing observations when trying to pinpoint any percentile travel time, i.e., if only half of observations are present, and we take the 95th percentile of that half, we are likely overestimating that value, especially if one makes the case that if there are no observations, it’s often safe to assume travel times are below 95th, not above (not always the case in extreme situations such as a freeway closure); in fact we make that implicit assumption here by counting down from the top as though all missing observations are likely below 95th percentile.

For questions on the map, send a note to or