Evaluation of Pavement Safety Performance

CHAPTER 7. HIGH FRICTION SURFACING TREATMENT

INTRODUCTION

The HFS treatment strategy was analyzed separately from the conventional pavements because of the nature of HFS treatments. Unlike the conventional treatments discussed in previous chapters, HFS is used almost exclusively for safety improvement (friction restoration or enhancement) purposes and not for pavement preservation or rehabilitation. In addition, with few exceptions, HFS treatments are used primarily for spot treatments of ramps or individual curves, rather than over longer sections of a roadway. As such, the data collection and analysis procedures differ somewhat from the conventional treatments.

DATA COLLECTION

HFS is a relatively new pavement treatment in the United States, at least in terms of systemic use. A limited number of States (including ELCSI-PFS States) have HFS treatments, and there are generally only a few treatments in those States.

States with HFS treatments were identified by the project team, FHWA, and PFS contacts, and also through related efforts such as the FHWA Surface Enhancements at Horizontal Curves (SEAHC) study. Data requested for HFS treatments were the same as that summarized in Table 4, previously. However, challenges with data collection for these sites included the following:

• Obtaining traffic data for ramps-Many States do not routinely collect traffic data on ramps, but rather just the mainline highway leading into or away from a ramp.
  
  
• Obtaining accurate crash data for ramps-Crash data for ramps can be difficult to obtain owing to inconsistencies in how the crash data are coded when recorded. They may be coded for the roadway (and associated milepoints) leading into the ramp or for the roadway leading away from the ramp.
  
  
• HFS treatment information-Knowing what material was used (specifically the aggregate type), the exact limits of the treatment, and the dates of installation is needed. Many HFS treatments were installed as demonstration or trial projects, and detailed records were not available. Some may have been removed prematurely or overlaid with another treatment.
  
  
• Identifying reference sites-HFS treatments are most commonly applied to curves or ramps with high crash rates that may be unique in geometry, location, traffic, etc. Finding similar sites to use as reference sites can be very difficult, if not impossible. Invariably, reference sites will likely have lower crash rates because the main criterion for selecting treatment locations was higher crash rates.
  
  
• Collecting roadway data for HFS sites-It was often necessary to use satellite imagery (Google Earth™) to verify the limits of an HFS treatment, lane and shoulder widths, and radius of curvature.
  
  
• Friction data-Before and after friction data were available for the SEAHC sites, including data from 1 and 3 years after installation. Unfortunately, friction data were not available for the remaining sites or for any of the reference sites, precluding the use of friction data in the analysis.

SUMMARY OF HFS TREATMENT DATA COLLECTION

Below are summaries of the data collection process for each of the volunteer States that provided candidate sites. The complete list of HFS treatment sites is provided in Appendix B.

Colorado

HFS treatment sites were on curves and were part of the FHWA SEAHC demonstration program. Treatment sites were originally selected by the Colorado Department of Transportation (CDOT) based on high crash rates at those curves. CDOT provided before and after crash data for the treatment sites. Roadway data (traffic volume, number of lanes, lane width, shoulder type and width, median type and width) were collected from the SEAHC project information and from an online CDOT roadway information database.

Reference sites were selected as segments of the roadway upstream and downstream from the treatment sites. Segments were selected based on similar traffic volume, number and width of through lanes, shoulder type and width, and median type. CDOT provided crash data for the same before and after periods as the treatment sites for these segments of roadway.

Kansas

HFS treatment sites were part of the FHWA SEAHC demonstration program and included two curves and two ramps. Treatment sites were originally selected by the Kansas Department of Transportation (KDOT) based on high crash rates at those locations. KDOT provided before and after crash data for the treatment sites and traffic information. Roadway information (pavement type, number and width of lanes, shoulder type and width) was collected from the SEAHC demonstration project information.

Reference sites were identified by KDOT based on similar roadway characteristics to the treatment sites (traffic and roadway geometry). KDOT provided crash data for the reference sites for the same before and after periods as the treatment sites.

Kentucky

HFS treatment site data were provided by the Kentucky Transportation Cabinet (KTC). KTC provided the list of treatment locations and summary before and after crash data. Roadway information (traffic volumes, number and width of lanes, shoulder type and width, and median information) were collected from KTC’s online Highway Information System (HIS) database, and crash data were obtained from the Kentucky State Police Collision Analysis online database.

Reference sites were identified using roadway data from the HIS database. For treatments on curves, segments of the roadway upstream and downstream from the treatment sites with similar characteristics (traffic volume, number and width of through lanes, shoulder type and width, and median type) were identified as reference sites. For treatments on ramps, ramps with similar geometry in the vicinity of the treatment sites were selected. Crash data were collected through the Kentucky State Police Collision Analysis database.

Michigan

HFS treatment sites were part of the FHWA SEAHC demonstration program and from various safety improvement projects by the Michigan Department of Transportation (MDOT). Treatment sites were originally selected by MDOT based on high crash rates at the curve and ramp locations identified. MDOT provided before and after crash data for the treatment sites as well as roadway information (traffic volume, underlying pavement type, and treatment length). Additional information for lane and shoulder widths was estimated using satellite imagery (Google Earth™).

Reference sites were identified by MDOT based on similarity in roadway characteristics to the treatment sites. MDOT provided before and after crash data for the reference sites for the same time periods as the treatment sites.

Montana

HFS treatment sites were part of the FHWA SEAHC demonstration program. Treatment sites were originally selected by Montana Department of Transportation (MDT) based on high crash rates at the two locations. MDT provided before and after crash data for the treatment sites, while roadway data (traffic volume, number of lanes, lane width, shoulder type and width, median type and width) were collected from the SEAHC demonstration project information.

Reference sites were identified by MDT based on similar roadway characteristics to the treatment sites. MDT provided crash data for the reference sites for the same before and after periods as the treatment sites.

South Carolina

HFS treatment sites were provided by the South Carolina Department of Transportation (SCDOT) from various safety improvement projects using HFS treatments. SCDOT provided the treatment site locations, before and after crash data, traffic, and underlying pavement information.

Reference sites were identified by SCDOT based on similar roadway characteristics to the treatment sites. SCDOT provided reference site locations and before and after crash data for the selected reference sites.

Tennessee

HFS treatment sites were identified by Tennessee Department of Transportation (TDOT) from safety improvement projects completed by TDOT. Six treatment locations were originally provided, but two were intersection approaches and not considered in the analysis. TDOT provided the treatment locations, and roadway information (traffic volumes, lane width, shoulder type and width, and median information) was collected from the online Tennessee Roadway Information Management System (TRIMS) maintained by TDOT. Detailed before and after crash data were also obtained from the TRIMS database.

Reference sites were identified using the roadway data from the TRIMS database. Segments of the same highway upstream and downstream from each of the treatment sites with similar characteristics (traffic volume, number and width of through lanes, shoulder type and width, and median type) were identified as reference sites. Before and after crash data were collected through the TRIMS database.

Wisconsin

One HFS treatment installed under the FHWA SEAHC demonstration program in 2011 was provided by the Wisconsin Department of Transportation (WisDOT). WisDOT provided before and after crash data for the treatment site, and roadway data (traffic volume, number of lanes, lane width, shoulder type and width, median type and width) were collected from the SEAHC demonstration project information.

Other States

Various other HFS treatments were provided by several States, but were not included in the final analysis because of insufficient crash data, information on the treatment site itself, or a lack of reference sites.

California-Caltrans provided a list of 48 completed and planned HFS treatments in the State. Of those, seven were selected as potential candidates for analysis. Unfortunately, a lack of crash data for each of these sites (due in part to most being less than 2 years old) and a lack of reference sites precluded their use in the analysis.

Iowa-Four sites that were installed as part of safety improvement projects in 2012 by the Iowa Department of Transportation were identified. A lack of raw crash data for these sites and reference sites for the analysis precluded their use.

Louisiana-One HFS site, installed as a safety improvement project in 2010, was provided by the Louisiana Department of Transportation. Crash data were provided for the site, as well as traffic and roadway information. However, because the treatment location was an elevated structure (bridge deck), and reference sites could not be identified, it was not included in the analysis.

Mississippi-One HFS site, installed as a safety improvement project in 2008, was identified by the Mississippi Department of Transportation, and before and after crash data were provided. However, a lack of reference sites (due to the unique characteristics of the treatment site) precluded its inclusion in the analysis.

Texas-Two HFS sites installed as safety improvement projects, and tested under the FHWA SEAHC program, were identified for Texas. A lack of crash data and reference sites precluded the use of these sites in the analysis.

West Virginia-West Virginia Department of Transportation provided a list of 24 horizontal curve HFS treatment sites installed as safety improvement projects. Before and after crash data and reference sites could not be obtained for these sites, and many of them were just over 1 year old, precluding their use in the analysis.

SUMMARY OF HFS TREATMENT SITES

Table 36 and table 37 provide a summary of the treatment and reference site data that were collected and used in the study.

Table 36 . Summary statistics of HFS treatment site data collected.

PCC = Portland cement concrete

Table 37 . Summary statistics of HFS comparison site data collected.

PCC = Portland cement concrete

ANALYSIS OBJECTIVES

Similar to the conventional treatments, the objective of this analysis was to estimate CMFs for the safety effect of HFS treatments using data from the States listed above. Treatment sites identified were either freeway ramps or individual curves. The treatments were generally applied because of a perceived problem with friction-related crashes.

The basic objective of the crash data analysis was to estimate the change in target crashes. Only nonintersection, nonanimal related crashes, and crashes not involving snow or ice were considered. Crash types examined total and wet-road crashes. Because of the limited sample sizes, other crash types were not investigated. Even for wet-road crashes, the sample size becomes small, but because these constituted the primary target crash type, they were analyzed separately. It should be noted that because HFS treatments are installed in a shorter period of time than conventional treatments, only the month in which the treatment was applied was masked off from the before and after periods, rather than the entire year.

ANALYSIS METHODOLOGY

The data collection for HFS treatments proved more difficult than for conventional treatments, particularly for treatments on ramps. For States that could provide untreated reference site data, the number of such sites was often limited, and, in many cases, traffic (i.e., AADT) information was missing. For South Carolina, fewer years of crash data were provided for reference sites than for treatment sites.

The lack of available data prohibited the application of the robust EB before-after methodology at this time. In the interim, both naïve before-after and comparison group (C-G) before-after studies were conducted with the limited data available, and guarded conclusions made on the basis of the results, given the methodological issues with these studies. Even so, not all of the treatment sites with before and after data had comparison sites for a C-G study. Below is a description of the methodology for the naïve and C-G studies, taken from Gross et al.⁽⁴⁴⁾

Naïve Study Approach

The simple before-after study, also referred to as the naïve before-after study, is a comparison of the number of crashes before and after treatment. The CMF for a given crash type at a treated site is estimated by first summing the observed crashes for the treatment site for the two time periods (assumed equal). The notation for these summations is summarized as follows:

K = the observed number of crashes in the before period for the treatment group.
L = the observed number of crashes in the after period for the treatment group.

The expected number of crashes for the treatment group that would have occurred in the after period without treatment is estimated using the equation in figure 27:

Figure 27. Equation.Estimated number of crashes that would have occurred in the after period with no treatment in the naïve study.

The variance of B is estimated using the equation in figure 28 :

Figure 28. Equation. Estimated variance of B in the naïve study.

The CMF and its variance are estimated using the equations in figure 29 and figure 30.

Figure 29. Equation.Estimated CMF in the naïve study.

Figure 30. Equation. Estimated CMF variance in the naïve study.

This method assumes that the number of crashes before the treatment is a good estimate of the expected crashes that would have occurred without the treatment. This assumption is in fact problematic because it does not take into account any other factors that can affect this estimate, such as changes in traffic volume and external causal factors. Most critically, sites that receive treatments such as HFS are typically selected on the basis of a high crash count, which introduces a regression to the mean (RTM) error whereby, without any treatment, the total number of crashes would have naturally declined in the after period. Thus, the results of a naïve before-after study can be biased toward overestimating the benefit of HFS treatment (i.e., underestimating the CMF).

C-G Study Approach

The before–and–after study using the C-G method is similar to the simple before–and–after study. It uses a comparison group of untreated sites to compensate for the external causal factors that could affect the change in the number of collisions. It does this by assuming that the ratio of crashes between the before and after period of the untreated sites would have been the same for the treated sites. Therefore, any external changes that would have changed the number of crashes in the after period throughout the area would be accounted for.

The CMF for a given crash type at a treated site is estimated by first summing the observed crashes for both the treatment and comparison groups for the two time periods (assumed equal). The notation for these summations is summarized as follows:

K = the observed number of crashes in the before period for the treatment group.
L = the observed number of crashes in the after period for the treatment group.
M = the observed number of crashes in the before period in the comparison group.
N = the observed number of crashes in the after period in the comparison group.

The comparison ratio (N/M) indicates how crash counts are expected to change in the absence of treatment (i.e., owing to factors other than the treatment of interest). This is estimated from the comparison group as the number of crashes in the after period divided by the number of crashes in the before period. The expected number of crashes for the treatment group that would have occurred in the after period without treatment is estimated using the equation in figure 31:

Figure 31. Equation. Estimated number of crashes that would have occurred in the after period with no treatment in the C-G study.

If the comparison group is ideal, the variance of B is estimated using the equation in figure 32:

Figure 32. Equation. Estimated variance of B in the C-G study.

The CMF and its variance are estimated as using the equations in figure 33 and figure 34:

Figure 33. Equation. Estimated CMF in the C-G study.

Figure 34. Equation. Estimated CMF variance in the C-G study.

This method, like the naïve method, does not account for RTM because it does not account for the natural reduction in crashes in the after period that would occur for the sites with abnormally high numbers of crashes, which would characterize the sites typically selected for HFS treatments. Thus, again, the results would likely be biased toward overestimating the benefit of HFS treatment (i.e., underestimating the CMF).

RESULTS

Results are provided in table 38 and table 39 for both the naïve and C-G studies. As mentioned earlier, not all treatment sites could be analyzed using the C-G method because reference sites were either unavailable or lacked the required data.

As noted, the results from applying these two methods are likely biased toward underestimating the CMF, and thereby exaggerating crash reductions, because RTM is likely at play and is not accounted for. An approximate method for resolving this problem has been suggested in the process of developing CMFs for the Highway Safety Manual.⁽⁴⁵⁾ That report suggests (on page 7) that "for a large RTM bias, where only a few sites with the highest crash frequency were treated out of the total population and few years of before-crash data were included in the evaluation study," the biased CMF should be corrected by multiplying it by a factor of 1.25. That recommendation seems appropriate for this evaluation, so this correction of 1.25 was applied to the biased CMFs. The CMFs with this RTM correction are shown in addition to the biased ones in table 38 and table 39 . These indicate that HFS treatments have a substantial beneficial impact on safety, especially for wet-road crashes.

Table 38 . Results for the naïve before-after study based on all sites.

CMF = Crash modification function
HSM = Highway Safety Manual
RTM = Regression to the mean

Table 39 . Results for the before-after C-G study for treatment sites for which comparison sites were available.

CMF = Crash modification function
HSM = Highway Safety Manual
RTM = Regression to the mean

CONCLUSIONS FOR HFS TREATMENT

This analysis was limited because there were insufficient treatment and reference group data to conduct a state-of-the-art EB analysis. Naïve before-after study results for all treatment sites, and those of a C-G study of treatment sites for which comparison sites were available, were obtained. These results are likely biased because the HFS treatments sites were likely selected on the basis of high crash counts, resulting in RTM that was not accounted for with the less rigorous methods that could be applied. A correction based on a method used to obtain Highway Safety Manual CMFs from similarly biased studies was applied as an approximation. The corrected results suggest that HFS can be a highly effective safety treatment whose implementation should continue.

Deployment of HFS as a safety countermeasure for curves and ramps is continuing in many States. It is strongly recommended that additional data be collected to conduct a robust EB before study to derive a CMF that could be recommended to practitioners and for which a BC ratio could be confidently estimated. The future data collection should, where possible, focus on those States with available traffic counts in both the before and after periods and that can identify appropriate reference sites.