Intersection Conflict Warning System Human Factors: Final Report

Figure 1. Graph. The number of crashes at each of 1,306 intersections shown as a function of AADT.

Of the 937 crashes represented in figure 1, complete data on crash type and driver injuries were available for 771 crashes. Table 1 characterizes the extent of driver injuries. A total of 19 percent of the crashes resulted in visible or more serious injuries to a driver, and 6 percent resulted in severe or fatal driver injuries. For the same 771 crashes, the type of collision is shown in table 2. Almost half of the crashes were classified as broadside, which is the type of crash ICWSs are intended to reduce. Some of the other crash types might also be mitigated by ICWSs when those crashes result from drivers’ attempts to avoid broadside collisions.

Table 1. Reported frequency of driver injuries.

Table 2. Reported frequency of crashes by crash type.

It has been suggested that ICWSs may be an appropriate crash mitigation strategy for the most problematic stop-controlled intersections. If the 2010 California data are representative, then only a small number of intersections might need ICWS treatment because only a few intersections are responsible for most of the crashes. Table 3, which includes all 1,306 intersections from the California query, shows the percentage of intersections as a function of the number of crashes they experienced. As can be seen in the table, 7.6 percent of the intersections had three or more crashes and accounted for 46.6 percent of all crashes.

Table 3. Crash frequency of 1,306 stop-controlled intersections on California State highways in 2010.

The ENTERPRISE Pooled Fund Study (PFS) undertook a project to collect information and assemble guidance on the use of ICWSs at stop-controlled intersections.⁽⁴⁾ The following three activated warning concepts were suggested:

• Warn drivers on the through approach to an intersection where vehicles have been detected on the stop-controlled minor road approach and that may be entering or crossing the highway.
• Warn drivers at the stop-controlled minor road approach to an intersection where vehicles are about to arrive on the through road.
• Warn drivers on both the minor and major road approaches of vehicles on the other approach.

The ENTERPRISE PFS concluded that ICWSs are effective mitigation strategies for intersections that either have a demonstrated high incidence of crashes or that have characteristics (such as limited sight distance) that place them at higher risk for crashes involving vehicles whose drivers are unaware of the presence of a potentially conflicting vehicle.⁽⁴⁾

The ENTERPRISE PFS elicited input from its members, other States that have deployed ICWS, the National Committee on Uniform Traffic Control Devices, the American Association of State Highway and Transportation Officials, and the National Association of County Engineers.⁽⁴⁾ In addition to several interim products, the PFS produced the following two final reports: Design and Evaluation Guidance for Intersection Conflict Warning Systems (ICWS) and System Requirements for Intersection Conflict Warning Systems (ICWS).^(5,6)

The design and evaluation guidance final report describes the following four ICWS conceptual designs:⁽⁵⁾

• Minor road warnings at the intersection of two two-lane roads.
• Minor road warnings at the intersection of a two-lane minor road with a multilane divided highway.
• Two-lane or multilane major road warnings at the intersection with stop-controlled two-lane roads.
• Combined major and minor road ICWSs.

The most complete ICWS safety evaluation for which a final report is available was conducted by the North Carolina Department of Transportation.⁽⁷⁾ That evaluation included 74 rural stop- controlled intersections, each of which was paired with 5 similar reference or control intersections that did not receive ICWS treatments. The evaluation employed the Empirical Bayes method for computing predicted crash rates with and without the treatment. The results were computed for ⁽¹⁾ all crashes within 150 ft of an intersection, ⁽²⁾ injury crashes, ⁽³⁾ severe (killed or disabled) injury crashes, and ⁽⁴⁾ frontal impact crashes. All treatment intersections had at least 1 year of posttreatment data; most had more than that. All intersections had at least 3 years of pretreatment data. Where more than 3 years of after data were available, the same number of before and after years were used. Use of the Empirical Bayes method to adjust for possible regression to the mean was vital in this study because the treatment sites were selected at least in part because of their below average safety performance.

All of the treatment sites employed loop detector sensors. The location of the detectors upstream of the minor road or in relation to the stop line on the minor road varied greatly between sites; the sample size, though impressive, was not large enough to analyze the effects of detector placement, particularly because sensor placement can be influenced by unique characteristics of individual treatment sites.

Treatment sites were grouped into the following four categories:

• Category 1: Sites with an overhead sign and flasher assembly placed at the intersection to warn drivers on the major (through) road of the presence of vehicles that might be entering from the stop-controlled road.
• Category 2: Sites with an overhead sign and flasher assembly placed at the intersection to provide drivers on the minor (stop-controlled) road with a warning of the approach of potentially conflicting traffic on the major road.
• Category 3: Sites with post-mounted signs and flashing beacons located 350 to 975 ft in advance of the intersection to warn drivers on the major (through) road of the presence of vehicles that might be entering from the stop-controlled road.
• Category 4: Sites that combine the treatments in categories 1 and 3 (i.e., both overhead assemblies and advance warnings directed to drivers on the major road).

The wording on signs varied across sites. Because of sample size considerations, the wording on signs was not evaluated. The North Carolina findings are summarized in table 4.

Table 4. ICWS treatment crash reduction factors from Empirical Bayes analysis of 67 North Carolina intersections.

Note: Blank cells indicate results not statistically significant.

Overall, the Simpson and Troy study yielded a small but statistically significant reduction in crashes of 6.8 percent.⁽⁷⁾ However, the estimated crash reduction factors varied by type of ICWS treatment. Overhead warnings at the intersection itself and intended for drivers on the major road resulted in a statistically reliable increase of 9.1 percent in the crash estimate, whereas warnings to drivers on the major road given in advance of the intersection on roadside mounted ICWSs resulted in a substantial estimated crash reduction of 19.3 percent. The combination of the advance and overhead ICWS on the major road yielded an even greater crash reduction estimate than for the advance sign alone. However, the number of intersections with the combined treatment was small, and the crash reduction estimate in that case was not statistically significant. Further complicating the evaluation of the overhead mounted ICWS was the fact that while the data suggested a reliable increase in all crashes with this treatment, estimates for severe and fatal injury crashes showed substantial reductions.

The Simpson and Troy study yielded a significant reduction in the estimated percent of crashes as a result of installing ICWS warnings on the stop-controlled (minor) road.⁽⁷⁾ The overall estimated crash reduction with this treatment was 3.5 percent; however, the estimated reduction in severe injury crashes was more substantial (14.1 percent) but not statistically reliable.

From a low-cost safety improvement perspective, the North Carolina study suggests that advance ICWS warnings on the major road will provide the greatest safety benefit.⁽⁷⁾ The data also suggest that treating all legs of an intersection and reinforcing advance warnings with a warning at the intersection may yield additional crash reduction benefit over and above the benefit of the major road advance warning.

A more extensive Empirical Bayes evaluation of ICWS was conducted for the Evaluation of Low Cost Safety Improvement (ELCSI) PFS.⁽⁸⁾ That study included 51 intersections that were included in the North Carolina study as well as data from 14 treated intersections in Missouri and 13 treated intersections in Minnesota.^(8,7) The results for two-lane by two-lane intersections are summarized in table 5. The results for two-lane by four-lane intersections are summarized in table 6. A further breakdown of the results suggests that the ICWS treatments are more effective on the major (through) road than on the minor (stop-controlled) road and more effective at intersections with lighting than without lighting. On the major road, post-mounted signs tended to be more effective than overhead signs. It should be noted that the treatments varied greatly both within and between States. In Missouri, the minor road signs were placed on the side of the major road but facing the driver at the stop sign on the minor road. An example of this approach is shown in figure 2. In Minnesota, the minor road warnings at four-lane divided highways were sometimes placed on the far side of the divided highway rather than in the median or overhead. In other cases, the minor road warning was placed upstream of the stop sign. Thus, although it appears that the minor road warnings of the detection of cross traffic on the major road were of limited or no benefit, this finding could be the result of factors that were not considered in the study such as legibility distance or intersection skew angles. The crash reduction benefit of warning major road drivers of the presence of vehicles at cross streets is clear. A crash reduction benefit for warning drivers on the minor road of the approach of major road traffic has yet to be demonstrated.

Table 5. ELCSI PFS Empirical Bayes results as a function of crash type for ICWS treatments at intersections of two-lane stop-controlled roads with two-lane through highways.

Table 6. ELCSI PFS Empirical Bayes results as a function of crash type for ICWS treatments at intersections of two-lane stop-controlled roads with four-lane through highways.

Figure 2. Illustration. ICWS warning installation at stop sign at intersection with four-lane divided highway.