Evaluation of Dynamic Speed Feedback Signs on Curves: A National Demonstration Project

CHAPTER 2. SELECTION OF TREATMENT AND CRASH ANALYSIS CONTROL SITES FOR DSFS SYSTEMS

The intent of the project was to select sites in States that represented geographic diversity across the United States and that were willing to participate in DSFS system installation. Travel considerations were also important because the Iowa-based team made initial visits to potential sites plus subsequent data-collection trips to all selected sites in the participating States.

Seven States participated in this demonstration project: Arizona, Florida, Iowa, Ohio, Oregon, Texas, and Washington. As a result, DSFS systems were tested in the Northwest, Southwest, Midwest, and Southeast regions of the United States, but none were tested in the Northeast.

Each participating State was asked to identify high-crash curve sites and provide initial information about the sites. For the purposes of identifying initial sites, the definition of high-crash was left up to the discretion of each State or agency.

The team then narrowed the initial list to a set of potential sites. Additional information was obtained for the potential sites, and site visits were made to each State. After the site visits, the team selected a set of final treatment and crash analysis control sites in each participating State.

DSFS systems were installed at treatment sites. Control sites, without DSFS systems, were used to conduct crash analyses. The general methodology used to select sites in each State is described in the following sections.

INITIAL REVIEW

A request for initial data was made to each State. The States were requested to provide at least 20 high-crash curve sites on rural two-lane roadways. It was left to the discretion of each agency to determine what it thought were high-crash locations. Rural was defined as 1 or more miles outside an incorporated area. Each curve was required to meet the following criteria:

• No rehabilitation or reconstruction activities that change the geometry of the roadway scheduled during the 2-year project.
• No geometric or cross-section changes made for 3 years prior to the study.
• Posted speed limit on preceding tangent section 50 mph or greater.

Each State was also requested to provide the following information about the potential sites:

• Crash frequency.
• Traffic volume (AADT and percent trucks).
• Geometry (lane width, shoulder width, and type).
• Speed limit (posted/advisory).

Different amounts and levels of detail were provided by the various States. Washington, Oregon, Arizona, and Ohio provided potential sites. Florida, Texas, and Iowa provided roadway and crash data, and the team selected potential sites. The team then followed up with all the States for additional information about the selected sites.

Once an initial list of high-crash curve sites for each State was obtained, the team located each of the curves using Google Earth™ or aerial images provided by the agency to determine whether there was anything about the site that made it inappropriate. A site was considered inappropriate and removed from further consideration if it was close to a major development, railroad, or major access points, including intersections other than low-volume intersections.

After the team removed inappropriate sites from the list, additional information about the remaining sites was requested from each State, if not already available. This included the following: 1) presence of posted speed advisory on curve, 2) information about crashes (speed-related, severity, etc.), 3) expert opinion about safety and speed problems, and 4) the existence of unusual traffic or other conditions.

Once this information was obtained, the team reviewed the list of potential sites. The sites were ranked in terms of number of crashes. A threshold was determined for each State to indicate what constituted high-crash locations. This varied from State to State because the number of years of crash data provided by each State was not consistent. In many cases, the crash information covered more than one curve, and this was taken into account. Sites with the number of crashes above the threshold were retained and included in the list for site visits.

SITE VISITS

Visits were then made to potential sites in each State. Information was recorded about each site, including layout, conditions, presence of speed and advisory speed signs, general conditions, as well as an indication of whether anything was unusual about the site. Images were also taken of various areas throughout the curve. Information about each site was recorded in a database.

Researchers conducted a preliminary radar gun speed study at each site to determine whether a speeding problem existed. The team collected data for both directions of traffic unless they were physically unable to collect data for one or both directions owing to adverse topography.

An attempt was made to collect at least 25 speed samples for each direction of traffic at each site. In several cases, a low number of vehicles were observed, and it was difficult for the team to remain at the site long enough to obtain this sample size. Mean speed, by direction, was calculated for all locations. When sample size was sufficient, 85th percentile speed was also calculated. A site was determined to have a speeding problem if at least one of the following conditions existed:

• Mean speed exceeded the advisory speed limit by 5 mph or more, or exceeded the posted speed limit by 5 mph or more if an advisory speed was not present.
• 85th percentile speed exceeded the advisory speed limit by 5 mph or more, or exceeded the posted speed limit by 5 mph or more if an advisory speed was not present.

SELECTION OF FINAL SITES

After the site visits, the team met and reviewed information about each site. Locations that did not have a speeding problem were removed from further consideration. If any other information from the site visit indicated the site was not feasible, it was also removed. This resulted in a final list of sites that were selected using similar criteria. At this point, sites had been selected without making any determination about whether the site would be a treatment or control site.

In most cases, treatment and control sites were quasi-randomly selected from the final list. It was determined that installation of the DSFS system would be challenging at a few sites so it was determined that it was more feasible to use these sites as control sites. For instance, at several sites, there was limited right-of-way to place the DSFS system. Several sites had sheer embankments that offered limited room to place a DSFS system, and several sites had significant drop-offs bordered by guardrail, which would have made data collection dangerous.

In several instances, curve sites were near each other. If one curve was selected as a treatment site and the team felt that placing a DSFS system at one curve would affect behavior on adjacent curves, the adjacent curves were dropped from the list and not used as either a treatment or control site.

Once final treatment sites were selected, one of the two different DSFS systems was randomly assigned. Table 24 lists treatment and control sites by State, and figure 34 through figure 40 show final locations of treatment and control sites. Control sites were selected for use in the crash analysis.

LOCATION OF DSFS SYSTEM AND SELECTION OF SIGN DIRECTION

Given only one DSFS system was available for each treatment site curve, it was necessary to determine in which direction of travel the system would be installed (i.e., eastbound (EB) versus westbound (WB)). If one direction had a higher percent of speed-related and/or single-vehicle run-off-road crashes than the other direction, the DSFS system was placed in this direction.

It should be noted that directional information was not available for a number of crashes. If no predominant crash direction was noted, the DSFS system was assigned to whichever direction of travel had the highest speeds based on the initial speed study. The DSFS systems were placed as close to the PC as possible. In all cases, the highest crash direction was the outside of the curve.

Table 24. List of final curve sites selected.

*Crashes were over several curves
ADT = Average daily traffic
NB = Northbound
SB = Southbound

Original image: ©2014 Google®; map annotations provided by CTRE. See reference Google 2014a.
Red markers indicate curve test sites.
Yellow markers indicate curve control sites.

Figure 34 Map. Location of test and crash analysis control sites in Arizona.

Original image: ©2014 Google®; map annotations provided by CTRE. See reference Google 2014b.
Red markers indicate curve test sites.
Yellow markers indicate curve control sites.

Figure 35. Map. Location of test and crash analysis control sites in Florida.

Original image: ©2014 Google®; map annotations provided by CTRE. See reference Google 2014c.
Red markers indicate curve test sites.
Yellow markers indicate curve control sites.

Figure 36. Map. Location of test and crash analysis control sites in Iowa.

Original image: ©2014 Google®; map annotations provided by CTRE. See reference Google 2014d.
Red markers indicate curve test sites.
Yellow markers indicate curve control sites.

Figure 37. Map. Location of test and crash analysis control sites in Franklin County, OH.

Original image: ©2014 Google®; map annotations provided by CTRE. See reference Google 2014e.
Red markers indicate curve test sites.
Yellow markers indicate curve control sites.

Figure 38. Map. Location of test and crash analysis control sites in Oregon.

Original image: ©2014 Google®; map annotations provided by CTRE. See reference Google 2014f.
Red markers indicate curve test sites.
Yellow markers indicate curve control sites.

Figure 39. Map. Location of test and crash analysis control sites in Texas.

Original image: ©2014 Google®; map annotations provided by CTRE. See reference Google 2014g.
Red markers indicate curve test sites.
Yellow markers indicate curve control sites.

Figure 40. Map. Location of test and crash analysis control sites in Washington.