CHAPTER 2

State of the Practice Review

This chapter presents key findings from the state of the practice review. It starts with a summary of the conducted literature review related to the development of CMFs for ATSPMs. It then presents agency outreach activities to understand better the way agencies use ATSPMs to make signal timing adjustments, especially for safety-based decision-making and determine potential sites for CMF development during Phase II of the research.

Literature Review Findings

The review of the literature revealed that there is limited research on CMFs for ATSPMs as ATSPMs are typically used to enhance operational conditions at intersections or along corridors. Therefore, the research team expanded the scope of the literature and reviewed the previous research and studies pertaining to the safety effects of other traffic signal technologies (e.g., adaptive traffic signal control) or signal system solutions (e.g., left turn phasing schemes). This is because past studies that have already established CMFs or quantified safety effects of certain signal timing improvements that are related to the ATSPMs can still be utilized during the development of especially case B CMFs even though ATSPM was not used for those signal timing changes. Additionally, the review focused on approaches taken and lessons-learned in finding sites with the needed data, especially for studies that developed CMFs. The research team also reviewed study design (e.g., type of data collected, duration of data, variables considered) as well as the statistical methods used in these studies.

Key Findings Related to Safety Effects of ATSPMs

As discussed above, only a limited number of studies were found in the literature that explored the safety effects of ATSPMs. This is likely because most ATSPM-related studies focus on their operational benefits.

One of the most relevant studies was by Lavrenz et al. (2016), where they used high-resolution signal controller data to estimate red light running (RLR) at signalized intersections. One intersection in Indiana was analyzed in this research to explore the relationship between RLR and level of congestion using split adjustments on a side street that experiences split failures. The study utilized conventional stop bar loop detection from high-resolution signal controller logs with signal state data to identify vehicles that enter the intersection after the start of red. A vehicle was flagged as RLR when the stop bar detector recorded an “on” event and subsequent “off” event after the start of red on the phase. The methodology was validated with on-site video collection at several locations, and the algorithm was refined to reduce the incidence of false RLR indications (Lavrenz et al. (2016)). Results from the study site showed that a change in green split from 20% of cycle length to 24% on the subject side street reduced the RLR rate. This trend is shown in Figure 1.

The study also found that the increase in split duration resulted in a reduction of approximately 34% in daily RLR counts. While this study developed a methodology to estimate RLR events, with the advancements in detection technology, most agencies today can use “count” detectors that are placed just downstream of a stop bar (with a minimum speed threshold) to directly obtain RLR using the Yellow and Red Actuation ATSPM report.

Key Findings Related to Safety Effects of other Signal System Solutions

Most research related to the CMF development for signal system projects was found in the area of left turn phasing strategies (i.e., permitted versus protected) and adaptive traffic signal control (ATSC) deployment. For the left turn phasing strategies, while ATSPMs were not utilized in these studies, the findings can be beneficial during the development of case B CMFs for the left-turn gap analysis.

Studies Related to Left Turn Phasing Strategies

Srinivasan et al., (2012) estimated CMFs from before-and-after evaluations of two treatments targeted at reducing left-turn crashes at signalized intersections: (1) changes from permissive to protected-permissive phasing, and (2) changes from the implementation of a flashing yellow arrow (FYA) for permissive left turns. For the first evaluation (i.e., from permissive to protected-permissive), data was collected from 59 treated and 626 untreated intersections in Toronto and 12 treated and 49 untreated intersections in North Carolina (various unspecified urban areas). For the second evaluation (i.e., FYA), data was collected from 5 treated and 27 untreated intersections from Kennewick, Washington, and 30 treated and 45 untreated intersections from Oregon (Beaverton, Gresham, Oregon City, and Portland), and 16 treated and 49 untreated intersections in North Carolina. Crash data was collected for at least 3 years from these various sites. To determine intersection related crashes, crashes that occurred within 25 meters (about 82 feet) of an intersection are used in Toronto and within 72 meters (about 250 feet) in North Carolina (other sites were not specified in the study). Table 3 provides a summary of the CMFs for different treatments and under varying groupings.

Table 3. CMF Results for protected-permitted and flashing yellow arrow treatments (source: Srinivasan et al., 2012).

Bold values indicate CMFs that are statistically different from 1.0 at the 5% confidence level

FYA ‒ Flashing Yellow Arrow.

Results show that changing from permissive to protected-permissive phasing statistically significant reductions on left turn opposing through crashes with a CMF of 0.862. However, when all crashes were combined or only injury and fatal crashes were considered, CMFs are very close to 1.0 and not statistically significant. Results from the FYA implementation showed that, not surprisingly, when a protected left turn phase was converted to protected-permitted with FYA, crashes increased considerably compared to before conditions. This increase is most likely due to the introduction of the permitted phase rather than the FYA

deployment. Results show that the implementation of FYA provided a benefit at intersections that had some of kind of permissive left turn operation in the before conditions (Srinivasan et al., (2012)).

Goughnour et al., (2021) focused on developing CMFs for improving pedestrian safety for protected or protected/permitted left turn phasing and LPIs. Researchers collected data from four North American cities from over 100 sites both for left turn phasing and LPIs. Data was collected from reference sites as well. Crash data was also collected from these cities with a focus on both vehicle-pedestrian and vehicle-vehicle crashes. Results showed that the provision of protected left-turn phasing reduced vehicle-vehicle injury crashes but did not produce statistically significant results for vehicle-pedestrian crashes.

Studies Related to Adaptive Traffic Signal Control (ATSC) Deployment

A recent study by Avelar et al., (2021) conducted safety effectiveness evaluations of ATSC deployments on urban corridors and developed CMFs for ATSC. Safety data was collected from Florida (87 intersections), Texas (25 intersections), and Virginia (68 intersections). For all three data sets, the research team used several variables to describe each site. These variables are grouped in the following categories:

• Traffic control device and ATSC related variables (e.g., when and where it was installed)
• Roadway design elements (e.g., number of legs, number of through lanes, lane width, number of left turn lanes, presence of bike lanes, median width, etc.)
• Traffic volume (i.e., AADT for major and minor legs), and
• Crash data (e.g., total number of crashes, crash severity)

To select intersection-related crashes, only crashes occurred within a 250-feet buffer around each intersection are selected. A total of 7 to 8 years of crash data (varied by site) were utilized for the before and after analysis. Table 4 provides a summary of the analysis results. For total crashes, CMFs ranged from 0.867 to 1.045 and only for the Virginia site, deployment of ATSC resulted in statistically significant reductions in total crashes. There were no discussions in the study related to why the produced CMF values ranged from one location to another. Additionally, most of the CMFs in the below table are not statistically significant and close to 1.000.

Table 4. Comparison of CMFs for adaptive traffic signal control (Source: Avelar et al., 2021).

* ‒ Statistical significance at a 90-percent confidence level.

** ‒ Statistical significance at a 95-percent confidence level.

Tang et al., (2020) also researched the safety effects of ATSC deployments from Pennsylvania. Data was collected from 342 intersections with ATSC deployments. Similar to previous studies, crashes reported within 250 feet of the intersection were included. The “after” data period (after the ATCS were deployed)

ranged from 1 year to 8 years. CMFs were estimated for varying scenarios including crash severity levels and crash types, intersection locations (all intersections and intersections along corridors only), and intersection configurations (3-leg versus 4-leg intersections). The results suggested that ATSCs are associated with a marginal increase in total and angle crashes and a marginal decrease in fatal plus injury crashes and rear-end crashes. Interestingly, all statistically significant CMFs indicated an increase in crashes with ATSC deployments.

Ma et al., (2016) also explored the safety implications of ATCS by examining 47 urban and suburban intersections along 10 corridors in Virginia. Vehicle crash data was collected between 2006 and 2013 for the before-and-after comparison. The study found that installing ATSC resulted in a CMF of 0.83 for total intersection crashes that are statistically significant at a 95 percent confidence level. However, when only fatal and injury crashes were analyzed, crashes did not change by a statistically significant amount. Similar to previous studies, Ma et al., (2016) found that safety benefits varied from corridor to corridor and by volume levels. However, no model is developed, and no discussion is provided regarding the way the results of this study can be utilized for other agencies or corridors with different characteristics.

Key Findings Related to CMF Development

Most studies related to the CMF development for signalized intersection treatments used before-and-after (i.e., before the treatment and after the treatment) evaluations using crash data. For those studies, crash data was also collected from untreated intersections with similar characteristics to the treatment site (i.e., control site or reference site) for comparison. To determine intersection related crashes, researchers typically used crashes that occur within 250 feet of an intersection. To ensure statistically significant CMFs can be developed, at least three years of crash data was obtained and analyzed from several signalized intersections (sometimes over 100 intersections). In addition to crash data, most researchers (Goughnour et al., (2021)) (Tang et al., (2020)) also collected the following data types to understand the effect of intersection characteristics on safety:

• AADT and pedestrian volumes
• Number of intersection legs
• One- or two-way direction of streets
• Number of through lanes
• Number of turn lanes
• Combination of through and turn lanes
• Presence of a crosswalk
• Presence of a median

Most studies found in the literature developed CMFs for various crash types. These include developing CMFs for all crashes, injury/fatal crashes, and target crash types (e.g., left turn opposing crashes for protected and permitted left turns, rear-end crashes for arrivals on green). One main challenge that was raised by researchers in these studies was the difficulties in finding sites with the needed data. Some of the studies indicated that after requesting data and locations from agencies, additional data collection was needed as the initial sites did not meet the desired data needs to develop CMFs (e.g., limitations in completeness of the data, varying data format). To address the data limitations, researchers sometimes utilized different “study designs” by location.

Another common observation related to CMF development is the various analysis methods in the evaluation. Most researchers used Empirical Bayes and Fully Bayesian methods. Other methods utilized in the CMF development include random parameter negative binomial model and generalized linear regression. While not many studies related to CMFs for signalized intersection treatments compared these different methods, a recent study by Avelar et al., (2021) claims that the Full Bayesian method produced better CMF values than the Empirical Bayes.

One main limitation regarding the application of the CMFs produced in some of the studies is that CMFs were developed for each state and not much discussion is provided related to why the CMF values were different for each location. This makes it challenging for practitioners to use the findings as design and intersection characteristics can substantially influence the safety performance of ATSC.

Agency Outreach Summary

The agency outreach task was completed in two sub-tasks as described below:

1. A survey was distributed to identify agencies and practitioners who regularly utilize ATSPMs for signal systems and gather some general information about these systems.
2. In addition to the survey, the research team reached out directly to targeted agencies and practitioners that are known to use ATSPMs to conduct phone interviews. The objective of these interviews was to obtain information about specific ATSPM systems and the availability of related archival data and potential sites for Phase II of the research.

Agency Survey

The team developed a draft survey for distribution to the agencies to better understand the way agencies use ATSPMs to make signal timing adjustments, especially for safety-based decision making, and whether agencies would be interested in sharing data with us for this project. Additionally, the survey aimed at identifying potential intersections and corridors for use for those agencies that utilize ATSPMs. Based on the Panel comments, the survey is refined and finalized.

The finalized version of the survey was distributed during the monthly ATSPM meeting that is led by the Federal Highway Administration (FHWA). Additionally, the survey was distributed to the members of the AASHTO Committee on Traffic Engineering (CTE) and to the members and friends of the TRB’s Traffic Signal Systems Committee. Finally, the survey was also posted on the ITE’s e-Community.

Following the outreach activities, the research team received 13 responses to the survey. Out of the 13 responses, 11 represent public agencies, one consulting firm, and one data collection vendor. Nine of the respondents indicated that they use ATSPMs. Table 5 summarizes agency type along with their characteristics and the use of ATSPMs for those agencies that utilize ATSPMs.

The highest number of ATSPM-equipped signals are operated by Georgia DOT, followed by Utah DOT, and Minnesota DOT. For cities and counties, agencies operate 100 or fewer ATSPM-equipped signals. Interestingly, when it comes to the type of ATSPM interface used, four agencies use other ATSPM platforms in addition to the Open Source (UDOT) ATSPM software. In terms of the typical detection scheme for ATSPM-equipped signals, each agency provided a different combination of detectors used.

Out of the 11 agencies that operate ATSPM-equipped intersections, 7 agencies reported that they used ATSPM reports explicitly for safety-based decision-making. Each response consists of multiple responses. The frequency of each ATSPM report used for safety-based decision-making is shown in Figure 2.

The trends shown in Figure 2 indicate that the Split Failure, Phase Termination, Timing and Actuation, and Purdue Coordination Diagram are the most common ATSPM reports used during safety-based decision making. However, it may be misleading to extrapolate these observations to other agencies due to the limited number of responses to the survey.

Targeted Agency Outreach and Phone Interviews

In addition to the survey, the research team conducted targeted outreach and phone interviews. Table 6 shows the list and provides a summary of key findings related to the potential corridors and intersections that can be used in Phase II.

Table 6. Selected individuals for targeted outreach and a summary of key findings.