U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
| REPORT |
| This report is an archived publication and may contain dated technical, contact, and link information |
|
 |
| Publication Number: FHWA-HRT-17-070 Date: August 2017 |
Publication Number: FHWA-HRT-17-070 Date: August 2017 |
As mentioned earlier, Illinois, Kentucky, and Missouri provided data for this study. These States also provided data on roadway geometry, traffic volumes, and crashes for both installation and reference sites. They also provided crash injury severities according to the KABCO scale. This chapter summarizes the data assembled for the analysis.
This section describes the installation data, reference sites, roadway data, traffic data, crash data, and treatment cost data for Illinois sites used in this evaluation.
The Illinois Department of Transportation (IDOT) provided a list of interstate highway sections where it had installed cable median barriers, along with the date of installation, brand of barrier, and cost of contract. IDOT installed cable median barriers at these locations to reduce cross-median crashes. The agency based its decisions on an examination of 5 years of crash data. Illinois used Fatality Analysis Reporting System (FARS) crashes on interstate highways and looked at head-on and opposite-direction sideswipe crashes. IDOT used this information to develop a warrant chart that could be used for cable median barrier installation. The final set of cable median barrier treatment sites covered approximately 100 mi.
IDOT did not install rumble strips at the same time as the cable median barriers. Thus, the research team wanted to know the exact time period when rumble strips were installed on the treatment sites. To obtain this information, IDOT ran a query of its contracts database to assemble a list of all contracts involving the installation of rumble strips. Owing to a lack of complete data from previous years, the research team was not able to determine a date at which IDOT first installed rumble strips on the treatment sites. However, the query did enable the research team to document any work involving rumble strips that IDOT performed on the treatment sites during the study period. If any record of work appeared for a site, the research team documented it along with the year it was done.
Although IDOT did not know the exact date of rumble strip installation at these sites, the documentation provided an indication of the earliest known year of rumble strip presence based on a visual inspection of 2004 photolog images. This provided the starting point for the analysis period.
Reference sites for the cable median barrier treatment needed to be segments of road that matched the treatment sites in terms of the type of roadway, the potential for cable median barriers (appropriateness of median), and existence of shoulder rumble strips during the before period of study. The research team identified reference sites for Illinois from an IDOT list of all interstate highway segments in the State known to have shoulder rumble strips based on a visual survey of 2004 photolog images. From this list, the research team selected interstate highway segments if they were on the same route as other segments that had been treated with cable median barriers but that had not received cable median barriers. This ensured that the reference sites would be similar to the treatment sites in terms of driver population, area type, and other factors. The research team selected only segments with a median width of less than 100 ft and no positive barriers. Once the team had assembled this list, they compared it with the query of rumble strip work that IDOT had developed based on a pay item query in its contracts database. The research team documented any work done during the study period involving rumble strips, along with the year the work was done.
This yielded a group of interstate highway segments with rumble strips, medians no wider than 100 ft, and no positive barriers in the median and that were located along the same routes as the treatment sites. The final batch of reference sites comprised approximately 400 mi of road segments.
The research team obtained roadway data for all treatment and reference sites from the FHWA Highway Safety Information System (HSIS) database. HSIS provided these data on an annual basis for each site from 2001 to 2010. The HSIS Guidebook for Illinois provided coding for these variables. (14)
The HSIS database yielded the following variables:
The research team obtained traffic data for all treatment and reference sites as part of the roadway data file from the FHWA HSIS database. HSIS provided an AADT value for each site annually from 2001 to 2010.
The research team obtained crash data for all treatment and reference sites from the FHWA HSIS database. The crash data obtained covered the years 2001 to 2010. The HSIS crash data provided location of crash, date of crash, severity, light condition at time of crash, crash type, and weather. The HSIS Guidebook for Illinois provides coding for the crash variables. The crash data can be linked to the treatment and reference sites by county and route (coded as CNTYRTE in HSIS data) and milepost.(14)
The research team obtained the cost of installing and maintaining cable median barriers and rumble strips from IDOT and summarized these data in table 6.
This section describes the installation data, reference sites, roadway data, traffic data, crash data, and treatment cost data for Kentucky sites used in this evaluation.
The Kentucky Transportation Cabinet (KYTC) provided a list of roadway sections where it had installed cable median barriers, along with the date of installation, brand of barrier, and cost of contract. KYTC generally selected sites for cable median barrier installations based on an annual evaluation process that considered crash experience, median width, median slope profile, traffic volume, traffic composition, and speed. From these variables, KYTC prioritized a list of recommended cable barrier projects. KYTC quantified crashes by examining cross-median fatal crashes per mile and total number of cross-median crashes per mile. The initial treatment group added up to 140 mi.
KYTC revealed problems with crash reporting in Jefferson County during the earlier years of the study period. Because the change in the reporting practices could have negatively affected any before–after analysis using Jefferson County crash data, the research team removed these data. Because Jefferson County is one of Kentucky’s largest counties and had a significant number of cable barrier installations, omitting it from consideration reduced the sample size to 69 mi.
The State did not install rumble strips at the same time as the cable median barriers, but the prior condition of all treated sites included existing rumble strips on both the outside and inside shoulders. Thus, the research team chose these sites for an analysis of the effect of cable median barrier installation on facilities already equipped with rumble strips.
KYTC provided a list of interstate highway segments from the entire State. Each segment had an indication of median type and median barrier presence. The research team selected reference sites from this list based on their having a depressed median, no median barriers, and a median width less than or equal to 200 ft. (Very wide medians are at low risk of cross-median crashes even without barriers.) The research team also cross-referenced this list against the cable barrier treatment list to remove treated sections. The final list of reference sites for the cable barrier treatment was approximately 323 mi. KYTC indicated that all reference sites could be assumed to have had median shoulder rumble strips by 2000.
The authors obtained roadway data in geographic information system (GIS) shapefile format from Kentucky staff. Separate shapefiles, each segmented differently, contained various road characteristics (e.g., shoulder width and traffic volume). The research team obtained GIS files from the KYTC website.(15) The research team obtained characteristics of the treatment and reference sites by matching each study site to the appropriate inventory segment by county, route, and milepost. The team also used the KYTC photolog and Google® aerial and Google Streetview™ imagery to confirm presence, type, and approximate installation dates of rumble strips at cable barrier treatment sites.(16)
Because the sites in Kentucky included freeway sections, and crashes on freeways tend to concentrate near interchanges, the research team requested and obtained data on locations of interchanges from KYTC.
KYTC maintained traffic volume data in the GIS inventory files, namely the traffic flow (TF) shapefile. The authors obtained traffic data for the treatment and reference sites by matching each study site to the appropriate inventory segment by county, route, and milepost. Specifically, the research team used the TF file from 2010 because it provided two datapoints: current (i.e., 2010) AADT and the prior AADT (with an indication of the year taken). The research team used these volume points to extrapolate yearly AADT for the before period.
KYTC supplied crash data for the routes and counties indicated in the treatment and reference site lists. KYTC also provided a data dictionary that described the fields in the crash data. The crash data contained all crashes for the individual routes and could be linked to the sites with a match based on county, route, and beginning and ending milepost. The field labeled “RDWYIDTXT” was present in both the crash and road files to indicate the route. KYTC indicated that crash location quality improved significantly in 2008. This was because officers used a mapping application, which allowed officers to select the crash location on a screen, which applied latitude/longitude coordinates to the crash record.
KYTC provided estimates of the costs and services lives of the treatments (see table 7). These data can be used to conduct a B/C analysis of the treatment.
This section describes the installation data, reference sites, roadway data, traffic data, crash data, and treatment cost data for Missouri sites used in this evaluation.
The State applied cable median barriers with inside shoulder rumble strip treatments statewide on roadways identified as tier 1 or tier 2. (Tier 1 was the highest order of routes followed by tier 2.) MoDOT supplied the list of projects where center-line and shoulder rumble strips or cable median barriers were recently installed or planned to be installed. Among the data the reports provided were the location—including district, State route number, and milepost—and the construction dates. MoDOT reported no other construction activities on these road segments.
The research team identified treatment sites by looking for sites for which both before- and after-period data would be available. The dates of cable median barrier installation did not always match those for one or both sides of the rumble strip installation, but they were close to each other. The team considered before periods to be prior to the installation of both cable barriers and rumble strips and after periods to start after both treatments had been installed.
Because Missouri had installed cable median barriers on a systemwide basis, it was almost impossible to identify comparable roadways without cable barriers for this road type at the time or in the near future. For this reason, the research team did not identify a separate reference group of sites without rumble strips and applied an alternate approach to the standard method for estimating and applying SPFs in the EB before–after methodology. Chapter 4 of this report describes this approach in more detail.
The research team obtained roadway data for the treatment sites from MoDOT, and the following variables were included in the data:
The roadway, traffic, and crash data were stored in a bidirectional manner (i.e., there was a separate record for each direction of travel). MoDOT staff were able to match opposing directions of travel for each site. The research team limited the constructed database to one record per site and the geometric information taken from the primary direction of travel.
The research team obtained traffic data in the form of AADT from 1999 to 2011 from MoDOT in electronic files for all treatment sites.
The compiled crash data contain many variables related to the location, time, and characteristics of each crash. The crash types of interest included non-intersection- and non-animal-related crashes.
MoDOT provided estimates of the costs and service lives of the treatments (see table 8).
Previous research indicated that cross-median crashes could be identified in many ways. In the crash reports from Kentucky and Missouri, there was a variable to indicate whether the crash was a cross-median crash. Illinois did not have such a variable. Staff from KYTC and the Kentucky Transportation Center indicated that the cross-median indicator in the Kentucky crash report was not reliable and suggested not using that variable. As a result, for Illinois and Kentucky, the research team used head-on plus opposite-direction sideswipe crashes as a proxy for cross-median crashes. The research team removed crashes related to wrong-way driving based on the crash reports. For Missouri, the team used two proxies for cross-median crashes: one based only on the cross-median indicator and the other based on the total number of crashes where either the cross-median indicator was included or the crash was designated as head-on. (Missouri did not indicate whether a sideswipe crash was opposite-direction or same-direction.)
Table 9 and table 10 provide summary information for the data collected for the treatment sites. The information in these tables should not be used to make simple before–after comparisons of crashes per mile-year because they do not account for factors, other than the strategy, that might cause a change in safety between the before and after periods. The research team made such comparisons properly with the EB analysis, as presented in chapter 7. Table 11 provides summary information for the reference site data. As discussed in chapter 4, the team used a different approach in Missouri where they could not find an appropriate reference group.
