Field Assessment of Variable Left-Turn Mode by Time-of-Day for Intersections Being Upgraded With Flashing Yellow Arrow Signal Heads and Offset Left-Turn Lanes

Abstract

The present study conducted a field assessment of variable left-turn (VLT) mode by time-of-day (TOD) plan for signalized intersections being upgraded with offset left-turn lanes and four-section vertical flashing yellow arrow (FYA) signal heads in Alabama. The VLT-TOD plan involved changing left-turn phasing from protected-permissive left-turn to permissive by TOD, based on left-turn and corresponding opposing traffic volume criteria at selected intersections. Consequently, the field assessment involved comparing surrogate safety and traffic operational measures for permissive left-turn and through traffic movements before and after the VLT-TOD implementation. Video and signal event data were collected in before and after conditions of the VLT-TOD implementation for the field assessment. The before–after comparison of safety and operational measures revealed that the VLT-TOD plan coupled with the FYA signal heads facilitated improved operational efficiency for the through traffic without compromising the safety and traffic operations of left-turning drivers at the study sites during the TOD hours. The results showed that surrogate measures, such as critical gap, postencroachment time, and follow-up time for permissive left-turns, remained unchanged before and after implementing the VLT-TOD plan, whereas a significant increase in the percent of green arrivals for through traffic was observed. The results are promising as they indicate that at signalized intersections with significant traffic volume fluctuation, the VLT by TOD plan utilizing the four-section FYA signal heads has the potential to significantly enhance overall traffic operations, particularly in situations in which drivers do not encounter sight-distance issues during the permissive left-turn phases.

Keywords

variable left-turn traffic operations flashing yellow arrow intersection sight distance intersection safety traffic signals

Turning left at signalized intersections is one of the most challenging tasks for road users ( 1 ). A study showed that around 30% of intersection crashes are associated with left-turning movement with roughly 70% of these crashes occurring at signalized intersections ( 2 ). Therefore, it is very important for traffic engineers to offer appropriate left-turn treatments that enhance the safety and operational efficiency of left-turning traffic and the overall capacity of a signalized intersection ( 3 ). In the United States, the protected-permissive left-turn (PPLT) operation is widely adopted because of its operational advantage when compared with the protected only (PO) operation. The PPLT operation combines the protected and permissive left-turn phases, providing a shorter green interval to the left-turning vehicles and a longer green interval to the corresponding through traffic, where permissive left-turns are allowed ( 4 ). Such a left-turn phasing keeps the signalized intersections functional and ensures a balance between intersection safety and efficiency ( 5 ).

In general, two signal head configurations are used in the United States to facilitate PPLT operations at signalized intersections where exclusive left-turn lanes are in place ( 4 ). One is the traditional five-section “doghouse” signal head and the other is the four-section flashing yellow arrow (FYA) signal head. Besides these two signal arrangements, some U.S. states also utilize a three-section FYA signal head with a bimodal lens to deal with PPLT operations ( 6 ). NCHRP Report 493 was the first major study that proved the value of utilizing the FYA permissive left-turn indication ( 6 ). Compared with the steady circular green indication for permissive phasing in the traditional five-section signal head, the FYA indication was found to be better understood by the left-turning drivers. Therefore, the FYA indication has been implemented at signalized intersections across the United States, which has led to operational and safety improvements. Furthermore, unlike the traditional five-section signal head, the four-section FYA signal head has the technological feasibility to implement a variable left-turn (VLT) mode ( 7 ). VLT mode involves changing the left-turn phasing from PO to PPLT, or even to permissive only, by time-of-day (TOD) to accommodate varying traffic demands at signalized intersections. Because of such technical feasibility, several state and municipal governments across the United States have adopted the FYA signal heads for upgrading existing or new signalized intersections that allow permissive left-turns, PPLT, or VLT mode by TOD (VLT-TOD hereafter) plan, based on intersection traffic and geometric conditions ( 8 , 9 ). It is important to note that besides traffic demand there are other critical factors that play an important role in determining an appropriate mode of left-turn traffic operation using the four-section FYA signal head ( 3 ). One of the most important criteria among these factors is the sight distance for left-turning drivers, which is further described in the next section.

Completing a permissive left-turn from an exclusive left-turn lane could be a risky behavior, particularly for drivers turning left at intersections situated on multilane divided highways with wide grass medians (see Figure 1) ( 3 ). Typically, at intersections with a wide negative left-turn offset (e.g., −6 ft or longer), opposing left-turning lanes are geometrically located on each other’s left-hand side. In such intersections, opposing left-turning vehicles obstruct each other’s view of the oncoming through traffic when they are positioned in their exclusive left-turn lanes, attempting to identify suitable gaps for making permissive left-turns simultaneously under PPLT or permissive left-turn phasing. Such a sight obstruction jeopardizes intersection safety and increases intersection delay ( 10 , 11 ).

Figure 1.

A signalized intersection with a wide negative left-turn offset ( 3 ).

The intersection safety and operational concerns associated with left-turning drivers’ permissive sight-distance can be truncated by realigning the left-turn offset along with the adjacent median, as shown in Figure 2 from Federal Highway Administration (FHWA) ( 3 ). Such an intersection enhancement with positive left-turn offsets improves the visibility of left-turning drivers by bringing the left-turn lanes laterally closer to the corresponding opposing traffic. The better visibility of oncoming traffic can reduce the likelihood of crashes between left-turn and opposing traffic at the intersections; this has already been corroborated by prior studies ( 10 , 12 , 13 ). Moreover, eliminating the sight-distance issue of left-turning drivers could also significantly increase the capacity of such intersections. This improvement provides traffic engineers with the opportunity to implement the PPLT or permissive left-turn phasing by TOD plan using FYA signal heads. By doing so, a higher number of vehicles are allowed to progress during a green interval for every signal cycle.

Figure 2.

Example of geometric improvements for left-turn traffic: (a) parallel and (b) tapered offset left-turn lanes ( 3 , 15 ).

Study Motivation and Objective

Alabama Department of Transportation (ALDOT) developed its third strategic highway safety plan (SHSP) in 2010 with the aim of achieving a 50% reduction in road fatalities by the year 2035 ( 14 ). One of the road-safety strategies under the SHSP was to implement proven countermeasures at signalized intersections to reduce fatalities and severe injuries resulting from left-turn-related crashes. As a result, the state agency initiated left-turn offset improvement at signalized intersections located on multilane divided highways with a wide grass median (see Figure 1 as an example) to improve the sight distance of left-turning drivers. In addition, the four-section vertical FYA signal head was also installed concurrently in many of these reconfigured intersections to assist in left-turning drivers’ decision making during the permissive phase as well as to increase the intersection capacity. It is important to note that the integration of the VLT-TOD plan, combined with the four-section FYA signal head, would improve traffic operations at signalized intersections under the condition that there is no sight-distance issue for left-turning drivers at these intersections. Such a left-turn treatment enables permissive-, PO-, and even PPLT phasing, depending on the intersection geometry and prevailing traffic conditions ( 4 ). The literature review presents a substantial body of studies that focused on selecting appropriate left-turn phasing options for use in the conventional five-section signal head, taking into account intersection site-specific characteristics ( 3 ). Furthermore, previous studies have examined the safety impact of FYA signal head implementation at signalized intersections through a before–after analysis of historical crash data ( 5 , 6 , 8 ). However, to the best of the authors’ knowledge, the safety and operational benefit of implementing the VLT-TOD plan in conjunction with the four-section FYA signal head has not been explored to date.

The research questions underlined in the present study are as follows: Is it feasible to implement the VLT-TOD plan for intersections being upgraded with four-section FYA-left-turn signal heads and offset left-turn lanes if critical parameters such as sight distance and traffic volume permit? If so, what are the safety and operational benefits associated with adopting this approach? To answer these questions, various surrogate safety and operational measures were employed, facilitating a comprehensive evaluation.

The remainder of this article is organized as follows: the next section reviews the literature to see what has been done in relation to the topic. The subsequent section describes the site, data collection, and surrogate measures selected in this study to achieve the research objectives. The results from the analysis are then discussed, and finally, we draw conclusions from our findings, and highlight study limitations and future works.

Literature Review

The exploration of effective left-turn phasing options that address both driver safety and traffic operations at intersections has been a subject of study since the 1960s. Literature shows that researchers have used variables such as crash history, sight distance, intersection geometry, left-turn and opposing traffic volume, left-turn conflicts, and traffic signal information to justify decision boundaries for advocating the use of PO, PPLT or permissive left-turn operations at signalized intersections ( 1 , 15 –17). The use of such explanatory variables is also prevalent among researchers who adopted computer-based traffic simulation software to establish the ideal left-turn phasing guidelines for signalized intersections ( 18 – 23 ). FHWA’s Signalized Intersection Informational Guide ( 24 ) uses such intersection geometric and traffic volume criteria to promote appropriate left-turn phasing and recommends less restrictive phasing (i.e., permissive or PPLT) at intersections whenever the existing site and traffic conditions allow traffic engineers to do so.

Several studies have examined the safety impact of implementing FYA signal heads for left turns at intersections. Among these studies is the seminal work conducted by Brehmer et al. ( 6 ). This study highlighted a significant improvement in drivers’ perception of permissive left-turn phasing following the installation of the FYA signal head at intersections. The findings from NCHRP Report 493 led to a series of studies exploring the impact of permissive left-turn indications on driver behavior at signalized intersections ( 25 – 29 ). Noyce et al. further investigated the utilization of the FYA signal heads for left turns by analyzing crashes at selected intersections ( 30 ). The study concluded that there was no improvement in safety at intersections that were transitioned from PO- to PPLT phasing with FYA signal heads as long as there were no sight-distance issues at the intersections. This statement was further corroborated by a study conducted by Srinivasan et al. who developed crash modification factors using empirical Bayes methods to represent the safety effect of changing left-turn phasing options from permissive to PPLT and PO to PPLT with the use of FYA signal heads ( 31 ). The study found an increase in left-turn crashes stemming from changing the left-turn phasing from PO to PPLT. Similar results were found in research by Simpson and Troy, which evaluated the safety of 222 FYA signals in North Carolina ( 32 ). A few studies ( 5 , 33 ) have established frameworks or guidelines for determining the eligibility of implementing the VLT-TOD plan using FYA signal heads based on site-specific characteristics. These studies recognized the sight distance available to left-turning drivers as a critical factor in determining the VLT-TOD plan with FYA signal heads. In addition, a study conducted in Texas ( 34 ) advised against using the FYA signals at intersections where heavy left-turn and heavy opposing traffic are present. Equally important to mention here is a recently completed nationwide NCHRP project, 03-125: “Evaluation of Change and Clearance Intervals Prior to the Flashing Yellow Arrow Permissive Left-Turn Indication,” the findings of which are yet to be published.

The literature review revealed that most previous studies relied on crash data analysis or subjective crash risk modeling based on site-specific geometric and traffic parameters to determine appropriate left-turn phasing options for intersections. Furthermore, when examining the safety impact of implementing FYA signals at intersections for permissive or PPLT operations, the site-selection process failed to consider whether such left-turn phasing would be permissible based on the matter of left-turning drivers’ sight distance. This concern is consistently emphasized in AASHTO- ( 35 ) and FHWA guidelines ( 24 ). Nonetheless, a knowledge gap exists between simulation approaches and actual field study for the safety and operational benefits of the VLT-TOD plan at signalized intersections where there is no sight-distance issue for left-turning drivers. The present study sought to overcome this gap in knowledge by evaluating the safety and operational effects of the VLT-TOD plan at intersections where existing geometric conditions allow permissive left-turns after implementation of the offset left-turn lanes and a four-section vertical FYA signal head.

Methodology

Site Selection

The following criteria were used in the present study for site selection. Intersections that met all these criteria were selected:

Signalized intersections with one exclusive left-turn lane and two corresponding opposing lanes for each of major road approach.

Signalized intersections with a posted speed limit of 45 mph or lower along major road approaches.

Signalized intersections with offset left-turn lanes for major road approaches, and thus no sight-distance issue for left-turning traffic.

Signalized intersections where PPLT phasing is utilized to facilitate left-turn traffic from the exclusive left-turn lane of major road approaches onto minor road approaches.

Signalized intersections where the four-section vertical FYA signal head is utilized for PPLT operations of left-turn traffic on major road approaches.

Signalized intersections where the fluctuation of hourly left-turn and the corresponding opposing traffic volume allow permissive and PPLT phasing following FHWA guidelines ( 24 ).

Signalized intersections where the number of crashes involving permissive left-turns is fewer than five within a 12-month period.

Two signalized intersections in Alabama (i.e., US-98 & McCrary Rd and US-98 & Johnson Rd) were selected based on the site-selection criteria listed above (see Figure 3). Note that the selected intersections were improved between 2016 and 2017 as part of Alabama’s SHSP program. These improvements included the installation of the four-section vertical signal head with FYA indication and the implementation of offset left-turns, collectively aimed at offering effective PPLT operations from the major road approaches (i.e., US-98). US-98 is the major road of these intersections, classified as a multilane divided highway by the state highway agency. The two intersections are isolated and uncoordinated with adjacent intersections. In addition, there is no pedestrian activity at the intersections. The lead-lead PPLT phasing with the four-section FYA signal head is used to manage the left-turn traffic on the exclusive left-turn lanes of the major road approaches along the US-98. Following ALDOT’s traffic signal standards, the four-section FYA signal head at these intersections was accompanied with a supplemental sign with the text “LEFT TURN YIELD ON FLASHING YELLOW ARROW.” Note that drivers understand the FYA indication without the supplemental sign ( 29 ), thus, it was not required.

Figure 3.

Two selected study sites in the present study: (a) US-98 & McCrary Rd, (b) US-98 & Johnson Rd, and (c) four-section vertical signal head with FYA indication.

A standard 8-phase dual ring signal phasing was utilized at the US-98 & McCrary Rd intersection, whereas a 6-phase dual ring phasing is used at the US-98 & Johnson Rd intersection. Five seconds of minimum green, 4 s of yellow, and 2 s of all-red intervals were offered for the left-turn traffic along the major road approaches at both intersections. Note that 15 s of the maximum protected green interval was set for the left-turn traffic. The annual average daily traffic of the major road approaches (i.e., US-98) for the US-98 & McCrary and US-98 & Johnson Rd intersections was 28,966 and 28,236, respectively, in 2022 ( 36 ). On weekdays, these two intersections experience significant traffic volume fluctuations by TOD. The average left-turn hourly traffic volume typically varies between 6 and 202 depending on the TOD, whereas the respective opposing hourly traffic ranges between 100 and 987. To ensure that there is no sight-distance issue for left-turning traffic from the major road approaches, a field measurement with a detailed sight-distance analysis was conducted. Readers may refer to previous studies ( 3 , 13 , 15 ) for details of the sight-distance analysis adopted in the present study.

Implementation of VLT by TOD Plan With FYA

Recall that the left-turn traffic from the major road (i.e., US-98) approaches at the two selected sites had been managed using PPLT phasing, in conjunction with FYA signal heads throughout the entire 24-h period. To see whether the FYA VLT-TOD plan would be feasible to deal with the left-turn traffic of these two intersections, the traffic volume criteria provided in FHWA’s Signalized Intersections Informational Guide ( 24 ) were reviewed. The guideline provides recommended left-turn phasing options based on the product of the hourly left-turn- and corresponding opposing traffic volume, which is illustrated in Figure 4. Hourly left-turn-, along with the corresponding opposing traffic volume for the selected study sites were collected and averaged across three typical weekdays in 2022. Afterwards, the hourly 24-h traffic volume data were plotted against FHWA criteria to identify which left-turn phasing option might be suitable for varying traffic demands at the selected intersections. An example of such a plot for the case of the US-98 & McCrary Rd intersection is presented in Figure 4.

Figure 4.

Illustration of FHWA volume criteria ( 19 ) for the selection of appropriate left-turn phasing options for the intersection approach with one exclusive left-turn and two opposing lanes with traffic data collected from US-98 & McCrary Rd.

Taking into account the hourly traffic volume fluctuation over the 24-h period and following FHWA’s volume criteria shown in Figure 4, a schedule for implementing an FYA VLT-TOD plan at the two selected sites was developed. On further consultation with the respective area’s traffic engineers, the VLT-TOD plan was implemented at the selected sites, as described in Table 1. As the table depicts, two analysis scenarios were applied to both selected sites referred to as existing and test conditions. The existing condition represents a signal operation under FYA-PPLT phasing for left turns from the major road to the minor road over the 24-h period. The test condition represents a TOD signal plan during which an FYA-VLT mode was implemented based on the varying traffic demand, subject to FHWA traffic volume criteria. Note that the two selected sites in the present study are currently operated with Centracs Advanced Traffic Management Software, which facilitates real-time traffic operations and signal controller management using a virtual environment ( 37 ).

Table 1.

Implementation Schedule of FYA VLT-TOD Plan for the Selected Intersections

Analysis scenarios	Left-turn signal phasing(from the major road to the minor road)	Research emphasis
Existing (before) condition	FYA-PPLT for 24 h	Left-turning driver behavior during the permissive portion of the PPLT phase
Test (after) condition	FYA VLT-TOD plan as below: •FYA-PPLT: 6 to 9 a.m. •FYA-Permissive-only: 9 to 10 a.m. •FYA-PPLT: 10 a.m. to 7 p.m. •FYA-permissive-only: 7 p.m. to 6 a.m.	Left-turning driver behavior during the permissive-only phases

Note: FYA = flashing yellow arrow; VLT = variable left-turn; TOD = time-of-day; PPLT = protected-permissive left-turn.

In cooperation with ALDOT traffic engineering division, the Centracs software was utilized to implement the test condition (i.e., FYA VLT-TOD plan) at the selected sites during the months of September and October 2022. For each analysis scenario, 3 weeks of video data were collected on normal weekdays. Multiple video cameras were installed at the selected sites to collect left-turning drivers’ behavior data during the permissive phases. The associated signal phase information, such as the length and timestamp information of the green, yellow, and red intervals was collected using Centracs for the corresponding hours of observations. An example of a Centracs signal system interface for intersection US-98 & McCrary Rd is shown in Figure 5.

Figure 5.

An example of Centracs’ signal system interface for intersection US-98 & McCrary Rd.

The signal and video data for the referenced TOD hours (as described in Table 1) were synchronized to extract the desired driver behavior information. Afterwards, the extracted information was utilized to calculate surrogate measures for the evaluation of the safety and operational benefits of the FYA VLT-TOD plan. Note that, to evaluate the impact of the FYA VLT-TOD plan on road users, only the permissive left-turn data during the permissive portion of the PPLT phase in the existing condition were extracted and analyzed to compare with the driver behavior data in the test condition.

In the data processing stage, the focus was on extracting driver behavior information from recorded video- and corresponding signal event data. Permissive left-turning drivers’ behavior during the period when the FYA VLT-TOD plan was implemented was extracted from the data sources to compare with that collected during the before condition (i.e., 24-h FYA-PPLT). The surrogate measures selected for this study involved careful extraction of timestamp data from the video recording (further described in the relevant section ahead). A detailed description of the surrogate measures utilized in this study follows.

Selected Surrogate Measures

Typically, evaluation of transportation infrastructure or system improvements relies heavily on historical crash data analyses before and after implementation ( 38 ). However, such crash-data-based evaluations may not be feasible when countermeasures require immediate implementation from a safety aspect, operational aspect, or both. In addition, conducting a crash-data-based evaluation is highly sensitive to the reliability, quality, and proper management of police-reported crash data, which is often unfeasible for transportation agencies. Because of such challenges, researchers commonly employ surrogate measures that tend to be proactive rather than reactive (i.e., crash-data-based evaluation). Several prior studies have used the behavior of left-turning drivers as evidence instead of relying on crash data. In these studies, driver behavior data were collected by placing video cameras at vantage points around the intersections ( 38 – 47 ). A comprehensive review of the literature revealed that critical gap and postencroachment time (PET) are the two most commonly employed surrogate safety measures by researchers in the field of transportation engineering. For operational evaluations of transportation infrastructure improvements, measures such as follow-up time (FUT) and percent green arrival (PGA) are commonly utilized ( 43 , 48 ). A detailed description of these safety and operational surrogate measures is given below.

Critical Gap and Accepted/Rejected Gaps

Critical gap is defined as the length of a time gap that left-turning drivers are likely to accept or reject during permissive left-turn operations ( 38 ). It can be further described as the gap length in seconds that has an equal probability of being accepted or rejected based on the observed distribution of time gaps at an intersection. Critical gaps estimated from field observations are often compared with those recommended in AASHTO ( 35 ). For permissive left-turns from major roads at an intersection similar to those selected in the present study (i.e., left-turning vehicles crossing two opposing lanes from an offset left-turn lane), the time gaps recommended in AASHTO ( 35 ) are 6.0 s for passenger cars and 7 s for single-unit trucks. If the estimated critical gap falls significantly below these recommended values, it is regarded as posing an increased risk.

An available gap is defined as the interarrival time between two opposing vehicles when a left-turning driver is waiting to make a permissive turn. Depending on the left-turning driver’s decision, a gap length can either be accepted or rejected. An accepted gap represents the duration of time or interarrival time between two oncoming opposing vehicles that is considered reasonable by a left-turning driver to safely complete the permissive turn ( 38 ). Therefore, a left-turning driver occasionally accepting a less desirable gap as a result of peer pressure from a closely following vehicle is also considered an accepted gap. Conversely, a rejected gap indicates the driver’s perception of an inadequate gap duration that is deemed unreasonable or too risky to complete the permissive turn. The accepted and rejected gaps for a left-turning driver were estimated manually from the timestamp data of the video recordings. An example of estimating accepted and rejected gaps from the timestamp event data is illustrated in Figure 6. The method of estimating the critical gap from the gap acceptance probability distribution is further described in the results and discussion sections ahead.

Figure 6.

Illustration of accepted and rejected gap for a left-turning vehicle at the intersection.

Postencroachment Time

PET is defined as the time difference between the moment a left-turning vehicle passes the encroachment area within the intersection-conflict area during a permissive turn, and the moment the next opposing vehicle reaches that same encroachment area ( 49 ). Figure 7 provides an illustration of PET estimation. The significance of the PET value lies in the value of zero indicating a potential crash between a left-turning vehicle and an opposing vehicle at the intersection. Therefore, a lower PET value indicates a higher chance of a crash occurring.

Figure 7.

Illustration of postencroachment time at an intersection.

Follow-up Time

FUT is defined as the time-headway between continuously moving vehicles in the left-turn lane that utilize the same gap for permissive left-turns after waiting a certain amount of time in a queue ( 39 ). In the present study, FUT was measured when queued left-turning vehicles were crossing the nearest opposing through lanes by closely following a lead vehicle. It is important to note that FUT was calculated only for queued left-turning vehicles waiting to complete permissive left-turns amid opposing oncoming traffic using the same gap. Therefore, FUT was not calculated for a platoon of left-turning vehicles that arrived and completed the turning maneuver without needing to wait in the left-turn lane for a potential gap amid opposing traffic.

Percent Green Arrivals

In the test scenario (i.e., FYA VLT-TOD plan), permissive-only left-turn phasing was provided from 9 to 10 a.m. and 7 p.m. to 6 a.m. at the selected intersections. During these periods, the protected green portion of PPLT phasing was removed, as compared to the existing scenario. Thus, the test scenario provided additional green time for corresponding through traffic in the same signal cycle, which would potentially help the progression of through traffic. To evaluate this hypothesis, PGA was employed to compare the operational efficiency of the through traffic between the existing and test conditions ( 48 ). In the present study, PGA was computed using Equation 1, where $V_{G}$ represents the total number of through vehicles that arrived only during the green intervals provided in an hour, whereas $V_{T}$ is the total number of through vehicles that arrived during the entire cycles. A detailed description of separating the green, yellow, and red arrivals for an approach in an hour can be found in the authors’ previous publication ( 48 ).

PGA = \frac{V_{G}}{V_{T}}

(1)

Data Extraction and Descriptions

Once the surrogate measures were identified, the focus was given to manually extracting driver behavior information from the video recording and signal event data collected during the existing and test scenarios as described in Table 1. The extracted driver behavior data were then entered into a spreadsheet, providing detailed information on accepted and rejected gaps, PET, and FUT values for permissive left-turning drivers during the existing and test scenarios. To calculate the selected surrogate measures, three weekdays of video recordings were analyzed for each scenario. As discussed earlier, driver behavior data collected only during the permissive left-turn period of PPLT phasing in the existing scenario were extracted for further analysis. Similarly, the driver behavior data collected during permissive-only phasing were extracted for the case of the test scenario.

In relation to the sample size, the existing scenario had 5,361 rejected gaps and 538 accepted gaps by left-turning drivers, whereas the test scenario had 5,199 rejected gaps and 602 accepted gaps. It is worth noting that the number of samples for PET calculation was equal to the number of accepted gaps in both analysis scenarios because PET was calculated only when a left-turning driver accepted a gap. Note that the calculation of critical gaps for both the existing and test scenarios, as well as the discussion of the selected surrogate measures, will be elaborated in the subsequent sections. Note that this research did not encompass the study of driver behavior associated with U-turn movement at intersections.

Results and Discussions

Calculation of Critical Gaps

Recall that the critical gap is the length of a time gap in seconds that has the equal chance of being accepted or rejected by permissive left-turning drivers based on field observations ( 43 ). The technique selected for estimating the critical gap in this study was logistic regression (LOGIT) in which the left-turning driver’s gap acceptance probability was predicted based on the distribution of the accepted and rejected gap durations ( 38 ). According to the LOGIT model, a driver’s decision to accept or reject a gap can be defined using Equation 2. In the model, the logit function, $f (x)$ , can be defined as shown in Equation 3.

P = \frac{e^{f (x)}}{1 + e^{f (x)}}

(2)

f (x) = \ln (\frac{P}{1 - P}) = β_{0} + β_{1} \times g

(3)

where

P represents the left-turning driver’s gap acceptance probability (i.e., the probability whether a driver accepts or rejects) for a given gap length;

$g$ is length of a gap in seconds; and

$β_{0}$ and $β_{1}$ are regression coefficients for the intercept and the length of a gap $(g)$ , respectively, typically estimated from the maximum likelihood estimates during the process of LOGIT model fitting.

The LOGIT model was utilized to fit the left-turning driver’s gap acceptance behavior/probability based on the explanatory variable (i.e., gap length $[g]$ ) for both the existing and test conditions. The summary of the fitted LOGIT model is shown in Table 2. It can be seen from the table that all the coefficients were statistically significant with p-values less than 0.05 (Pr > ChiSq). The model fit assessment statistic (max-rescaled R²) indicated that approximately 85% of the variability in left-turning drivers’ gap acceptance probability could be explained by the explanatory variable in the existing scenario. For the test scenario, the corresponding value was 0.86. The area under curve value, which indicates the discriminatory power of the model across various threshold limits ( 50 ), was found to be as high as 0.98 and 0.96 for the existing and test scenarios respectively, indicating the strong statistical significance of the models.

Table 2.

Summary of the Fitted LOGIT Model

Analysis scenarios	Estimate of the LOGIT model						Model fit assessment
Analysis scenarios	Regression coefficient	df	Estimate of coefficient	SE	Wald chi-square	Pr > ChiSq	Max-rescaled R²	Area under curve
Existing condition	$β_{0}$	1	−10.245	0.414	610.84	< 2.2e-16	0.85	0.98
Existing condition	$β_{1}$	1	1.667	0.077	484.31	<2.2e-16	0.85	0.98
Test condition	$β_{0}$	1	−10.276	0.416	609.89	<2.2e-16	0.86	0.96
Test condition	$β_{1}$	1	1.697	0.078	476.39	<2.2e-16	0.86	0.96

Note: LOGIT = logistic regression.

Figure 8 illustrates left-turning drivers’ gap acceptance probabilities, predicted based on the LOGIT model presented in Table 2, with varying gap lengths $(g)$ . As the figure indicates, the longer the gap length, the higher the gap acceptance probability by the left-turning drivers. The critical gap, which is either accepted or rejected by left-turning drivers with a 50% chance, was estimated from the corresponding values on the x-axis that represent the gap length in seconds ( 43 ). Based on the LOGIT model, the critical gap was found to be 6.20 s for the existing scenario and 6.10 s for the test scenario, which are very close in value. The slight reduction in the critical gap in the test scenario could be attributed to the omission of the protected green arrow indication (during the permissive-only phase from 9 to 10 a.m. and 7 p.m. to 6 a.m.), which was previously provided in PPLT phasing in the existing scenario.

Figure 8.

Predicted gap acceptance probability for existing and test scenarios.

Note that under PPLT phasing, drivers have the option to complete the left turn in the next cycle if they cannot find a gap in the oncoming opposing traffic during the current cycle. However, in the test scenario, where only the permissive phase was available between 9 and 10 a.m. as well as between 7 p.m. and 6 a.m., drivers may experience relatively longer wait times in the queue if they are unable to find acceptable gaps in consecutive cycles. If this situation persists, drivers may opt for relatively shorter gaps compared with what they would typically take under PPLT phasing. This assumption was considered in the present study. However, it is important to note, based on the results shown in Figures 8 and 9, that the impact of changing signal phasing from PPLT to the permissive-only left-turn phasing was negligible as traffic conditions and other critical factors (such as sight distance) warranted the permissive-only phasing during the time selected in the test scenario. It is also important to note that the critical gap estimated in both the existing (6.2 s) and the test scenarios (6.1 s) are remarkably close to AASHTO’s ( 35 ) recommended values for permissive left-turns at intersections that have similar geometric configurations to those selected for the present study.

To confirm whether the two critical gaps found under the different analysis scenarios were statistically different or not, the present study adopted the visual hypothesis testing method by Smith ( 51 ). Note that Smith’s method was adopted by past researchers to compare the statistical significance of estimated critical gaps across different left-turn offset categories ( 31 ). Smith’s method involves estimating the 95% lower and upper confidence intervals of critical gaps for the two groups (i.e., existing and test scenarios) and then evaluating whether the mean of one group is within the 95% confidence interval of the other group mean. The 95% confidence intervals of the critical gap for both the existing and test conditions were estimated from the gap acceptance probability curve and are illustrated in Figure 9. As shown in the figure, the critical gaps for both scenarios fell within each other’s 95% confidence intervals and were nearly identical. This indicated a failure to reject the null hypothesis according to Smith’s method ( 51 ). Therefore, it was concluded that there was insufficient evidence to suggest a significant difference between the two estimated critical gaps of 6.2 and 6.1 s, respectively, for the permissive left-turning drivers in the existing and test scenarios during the selected time periods at the selected sites.

Figure 9.

Illustration of Smith’s visual hypothesis testing for critical gaps.

Calculation of Postencroachment Time

Recall that the PET value becomes zero in the event of a crash between left-turning and oncoming opposing vehicles. There were no cases where the PET value was less than zero. Thus, the higher the PET values above zero, the safer it is for left-turning drivers during the permissive left-turns. To assess safety between the existing and test conditions, the PET values observed in the dataset within the range of 0 to 5 s were compared. The comparison of these PET values is presented in Figure 10. As depicted in the figure, no PET values of less than 1 s were observed for permissive left-turns in either the existing or test scenarios. Over 50% of permissive left-turning drivers in the total dataset had a PET value less than 5 s. PET values less than 2 s were observed only in 6% and 5% of permissive left-turns, and nearly one-third of the permissive left-turning drivers had PET values less than 3 s in the existing and test conditions, respectively. Overall, there was no significant difference in PET values between the existing and test scenarios when comparing PET distributions between 0 and 5 s.

Figure 10.

Comparison of postencroachment time (PET) values between the existing and test conditions.

Calculation of Follow-Up Time

As discussed earlier, FUT was measured only for continuously moving vehicles that were in the queue and eventually made permissive left-turns utilizing the same gap. In general, FUT is used to measure the sensitivity of left-turn capacity to any change in the traffic conditions at intersections. The box plot in Figure 11 shows the minimum, maximum, and three quartile values of FUT estimated in both the existing and test scenarios. In the existing scenario, 50% of permissive left-turning drivers who met the FUT calculation criteria were observed within the range of the 1st quartile value (2.2 s) and 3rd quartile value (3.0 s) for FUT. Similar results were found for the permissive left-turning drivers in the test scenario with the corresponding FUT values of 2.1 and 3.0 s. The mean FUT values for the existing and test scenarios were 2.6 and 2.5 s, respectively.

Figure 11.

Box plot of FUT values for the existing and test conditions.

Note that from the recorded video data the removal of the protected left-turn phase from PPLT phasing for the test scenario of the permissive-only phase (from 9 to 10 a.m. and 7 p.m. to 6 a.m.) was observed to introduce a slight delay for the queued vehicles waiting to find acceptable gaps for permissive left-turns. When a gap appeared to be acceptable, the queued vehicles quickly followed the leading left-turning vehicle to complete the turning maneuver. Such a change in behavior among left-turning drivers resulted in a slight reduction in the mean FUT value of 2.5 s, which was 0.1 s less than the respective mean FUT value of 2.6 s in the existing condition for the comparable hours. According to Yan and Radwan, a decrease in the FUT value implies an increase in left-turn capacity ( 43 ). However, to gather additional evidence, a two-sample t-test assuming unequal variances was conducted to see whether there was a significant difference between the FUT means for the two groups. At the 5% significance level, the p-value for the two-tailed t-test, t(217) = 1.632, was found to be 0.104 (>0.05). Therefore, the result indicated that there was insufficient evidence to conclude a significant difference between the means of FUT for the existing and test scenarios.

Calculation of Percent Green Arrivals for Through Traffic

The three surrogate measures discussed in the previous sections were to evaluate the safety and operational efficiency of the left-turning drivers as a result of field implementation of the FYA VLT-TOD plan. PGA was selected to assess the impact of the FYA VLT-TOD plan implementation on through traffic before and after its implementation. The westbound approach of US-98 at McCrary Rd and the southbound approach of US-98 at Johnson Rd were selected for the comparison of PGA values in the existing and test scenarios. PGA values were calculated and compared based on the average of three typical weekdays of vehicle count data collected during the existing and test scenarios at each intersection approach for comparable TOD hours. Recall that the existing and test scenarios represented the 3-week before–after period during which the FYA VLT-TOD plan was implemented at the selected sites. The comparison of PGA values was conducted for five representative TOD hours (i.e., 9 to 10 a.m., 7 to 8 p.m., 8 to 9 p.m., 9 to 10 p.m., and 10 to 11 p.m.) between the two analysis scenarios, as shown in Figure 12.

Figure 12.

Comparison of percent green arrival for analysis scenarios at two selected study sites: (a) US-98 westbound at McCrary Rd, and (b) US-98 southbound at Johnson Rd.

As shown in the figure, there was a slight increase in PGA values for all five representative TOD hours of the test conditions. On average, about a 4% increase in PGA values was observed as a result of implementing the FYA VLT-TOD plan at US-98 & McCrary Rd during the study period. There was about a 3% increase in PGA values for the US-98 & Johnson Rd intersection with the FYA VLT-TOD plan implementation in the same period. Overall, both intersections experienced improved operational efficiency for through traffic owing to implementation of the FYA VLT-TOD plan. Such a result can be attributed to removing the protected left-turn phase from PPLT phasing during the study period of the test scenario. The removal of the protected phase resulted in an additional allocation of green time for the through traffic in the same signal cycle along the major road approaches.

Conclusions

The objective of the present study was to evaluate the safety and operational benefit of implementing the VLT-TOD plan at signalized intersections which had recently been improved with the FYA signal head and offset left-turn lanes. Two signalized intersections were selected according to the site-selection criteria. FHWA traffic volume criteria were then utilized to determine the VLT-TOD schedule for traffic fluctuations over time at the selected signalized intersections. Surrogate measures such as critical gap, PET, FUT, and PGA were employed for the before–after evaluation of the VLT-TOD plan at the selected sites. To gather information on the surrogate measures, driver behavior data before and after the implementation of the VLT-TOD plan were collected by setting up video cameras at multiple locations. In addition, the corresponding signal phase information associated with the VLT-TOD implementation was also collected using Centracs ATMS, which facilitates real-time traffic operations and management. The timestamp data needed to compare the surrogate measures were subsequently extracted by combining the video recording and the relevant signal phase information. The surrogate measures were then calculated and compared between the existing and test conditions for the hours during which the VLT-TOD plan was implemented in the field. The critical gaps were found to be 6.2 and 6.1 s for the before and after scenarios, respectively, and were not statistically different. No significant difference in PET values was observed between the before and after conditions when comparing the PET distribution between 0 and 5 s. In the case of left-turning drivers’ operational performance, even though the mean FUT was smaller in the test (2.5 s) than the existing condition (2.6 s), the t-test results showed that the difference was not statistically significant. Furthermore, when the operational efficiency for the through traffic was compared between the two analysis scenarios for five representative TOD hours, an approximate 3% to 4% increase in PGA was observed in the test condition, indicating improved operational efficiency as a result of FYA VLT-TOD field implementation. Overall, this study showed that the field implementation of the VLT-TOD plan improved operational efficiency for through traffic without compromising the safety and operational efficiency of the left-turning drivers at the selected sites where there were no sight-distance issues.

One of the limitations in this study was the number of sample sizes. The before–after evaluation of the VLT-TOD implementation was conducted based on driver behavior data collected at only two study sites. Moreover, the VLT-TOD plan was implemented during selected hours by adhering to FHWA volume criteria ( 24 ) at intersections, which were upgraded with the FYA signal heads and the offset left-turn lanes. One potential area for future research might involve an evaluation of the VLT-TOD plan implemented across various geometric and traffic volume conditions during different TOD periods, which might provide more information on the effectiveness of the VLT-TOD plan implementation on the overall capacity at signalized intersections.

Footnotes

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: M.-W. Kang, P. Biswas; data collection: P. Biswas, R. Hossain, M.-W. Kang; analysis and interpretation of results: P. Biswas, M.-W. Kang; draft manuscript preparation: P. Biswas, M.-W. Kang, R. Hossain, M. Rahman. All authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was partially supported by Alabama Department of Transportation (grant no. 930-980).

ORCID iDs

Pranesh Biswas

Min-Wook Kang

Moynur Rahman

References

Stamatiadis

Agent

K. R.

Bizakis

Guidelines for Left-Turn Phasing Treatment. Transportation Research Record: Journal of the Transportation Research Board, 1997. 1605: 1–7.

Schattler

K. L.

Gulla

C. J.

Wallenfang

T. J.

Burdett

B. A.

Lund

J. A.

Safety Effects of Traffic Signing for Left Turn Flashing Yellow Arrow Signals. Accident Analysis and Prevention, Vol. 75, 2015, pp. 252–263.

Kang

Biswas

Jones

Flashing Yellow Arrow Left Turn Phasing by Time-of-Day: A Study of Sight Distance and Traffic Volume Warrants. Prepared for Alabama Department of Transportation, University of South Alabama, Mobile, AL, 2022.

Sullivan, A., S. Jones, E. Tedla, and E. Doustmohammadi. Traffic Signal Design Guide & Timing Manual. Prepared for Alabama Department of Transportation, The University Transportation Center for Alabama, Montgomery, AL, 2015.

Abou-Senna

Radwan

Harb

R. C.

Navarro

Chalise

Interactive Decision Support System for Predicting Flashing Yellow Arrow Left-Turn Mode by Time of Day. Transportation Research Record: Journal of the Transportation Research Board, 2014. 2463: 16–25.

Brehmer

C. L.

Kacir

K. C.

Noyce

D. A.

Manser

M. P.

Evaluation of Traffic Signal Displays for Protected/Permissive Left-Turn Control. Report No. 493, National Cooperative Highway Research Program, Transportation Research Board, Washington, D.C., 2003.

Abou-Senna

Radwan

Eldeeb

A Peer-to-Peer Logic Environment to Validate Flashing Yellow Arrow Decision Support System. Journal of Traffic and Transportation Engineering (English Edition), Vol. 8, No. 5, 2021, pp. 735–750.

Appiah

Cottrell

B. H.

Safety and Operational Impacts of Optional Flashing Yellow Arrow Delay. Transportation Research Record: Journal of the Transportation Research Board, 2014. 2463: 26–34.

Davis

G. A.

Moshtagh

Hourdos

Safety-Related Guidelines for Time-of-Day Changes in Left-Turn Phasing. Transportation Research Record: Journal of the Transportation Research Board, 2016. 2557: 100–107.

10.

Persaud

Lyon

Gross

Eccles

Safety Evaluation of Offset Improvements for Left-Turn Lanes. Transportation Research Record: Journal of the Transportation Research Board, 2010. 2171: 44–51.

11.

Yan

Radwan

Effect of Restricted Sight Distances on Driver Behaviors during Unprotected Left-Turn Phase at Signalized Intersections. Transportation Research Part F: Traffic Psychology and Behaviour, Vol. 10, No. 4, 2007, pp. 330–344.

12.

Naik

Appiah

Khattak

A. J.

Rilett

L. R.

Safety Effectiveness of Offsetting Opposing Left-Turn Lanes: A Case Study. Journal of the Transportation Research Forum, Vol. 48, No. 2, 2012, pp. 71–82.

13.

McCoy

P. T.

Pesti

Byrd

P. S.

Singh

V. A.

Guidelines for Opposing Left-Turn Lane-Line Widths. Transportation Research Record: Journal of the Transportation Research Board, 2001. 1751: 26–34.

14.

ALDOT. Alabama Strategic Highway Safety Plan: Second Edition. Alabama Department of Transportation, Montgomery, AL, 2012.

15.

Agent

Deen

Warrants for Left-Turn Signal Phasing. Transportation Research Record, National Research Council, Vol. 737, 1979, pp. 1–10.

16.

Agent

Guidelines for the Use of Protected/Permissive Left-Turn Phasing. Prepared for Kentucky Transportation Cabinet, University of Kentucky, Lexington, KY, 1985.

17.

Amiridis

Stamatiadis

Kirk

Safety-Based Left-Turn Phasing Decisions for Signalized Intersections. Transportation Research Record: Journal of the Transportation Research Board, 2017. 2619: 13–19.

18.

Chang

G.-L.

Chen

C.-Y.

Perez

Hybrid Model for Estimating Permitted Left-Turn Saturation Flow Rate. Transportation Research Record: Journal of the Transportation Research Board, 1996. 1566: 54–63.

19.

Al-kaisy

A. F.

Stewart

J. A.

New Approach for Developing Warrants of Protected Left-Turn Phase at Signalized Intersections. Transportation Research Part A: Policy and Practice, Vol. 35, No. 6, 2001, pp. 561–574.

20.

Stamatiadis

Hedges

Kirk

A Simulation-Based Approach in Determining Permitted Left-Turn Capacities. Transportation Research Part C: Emerging Technologies, Vol. 55, 2015, pp. 486–495.

21.

Stamatiadis

Tate

Kirk

Left-Turn Phasing Decisions Based on Conflict Analysis. Transportation Research Procedia, Vol. 14, 2016, pp. 3390–3398.

22.

Appiah

King

F. A.

Fontaine

M. D.

Cottrell

B. H.

Left Turn Crash Risk Analysis: Development of a Microsimulation Modeling Approach. Accident Analysis and Prevention, Vol. 144, 2020, p. 105591.

23.

Adamson

Schultz

G. G.

Saito

Stevens

M. D.

Developing Decision Boundaries for Left-Turn Treatments. Transportation Research Record: Journal of the Transportation Research Board, 2020. 2674: 315–326.

24.

Chandler

B. E.

Myers

M. C.

Atkinson

J. E.

Bryer

T. E.

Retting

Smithline

Trim

, et al. Signalized Intersections Informational Guide: Second Edition. Prepared for Federal Highway Administration: Report No. FHWA-SA-13-027, SAIC, Raston, VA, 2013.

25.

Knodler

M. A.

Noyce

D. A.

Kacir

K. C.

Brehmer

C. L.

Evaluation of Traffic Signal Displays for Protected-Permissive Left-Turn Control Using Driving Simulator Technology. Journal of Transportation Engineering, Vol. 131, No. 4, 2005, pp. 270–278.

26.

Knodler

M. A.

Noyce

D. A.

Fisher

D. L.

Evaluating Effect of Two Allowable Permissive Left-Turn Indications. Transportation Research Record: Journal of the Transportation Research Board, 2007. 2018: 53–62.

27.

Knodler

M. A.

Noyce

D. A.

Kacir

K. C.

Brehmer

C. L.

Analysis of Driver and Pedestrian Comprehension of Requirements for Permissive Left-Turn Applications. Transportation Research Record: Journal of the Transportation Research Board, 2006. 1982: 65–75.

28.

Knodler

M. A.

Noyce

D. A.

Kacir

K. C.

Brehmer

C. L.

Potential Application of Flashing Yellow Arrow Permissive Indication in Separated Left-Turn Lanes. Transportation Research Record: Journal of the Transportation Research Board, 2006. 1973: 10–17.

29.

Knodler

M. A.

Noyce

D. A.

Kacir

K. C.

Brehmer

C. L.

Evaluation of Flashing Yellow Arrow in Traffic Signal Displays with Simultaneous Permissive Indications. Transportation Research Record: Journal of the Transportation Research Board, 2005. 1918: 46–55.

30.

Noyce

D. A.

Bergh

C. R.

Chapman

J. R.

Evaluation of the Flashing Yellow Arrow Permissive-Only Left-Turn Indication Field Implementation. Prepared for National Cooperative Highway Research Program: NCHRP Project 20-7, University of Wisconsin, Madison, WI, 2007.

31.

Srinivasan

Lyon

Persaud

Baek

Gross

Smith

Sundstrom

Crash Modification Factors for Changes to Left-Turn Phasing. Transporation Research Record: Journal of the Transportation Research Board, 2012. 2279: 108–117.

32.

Simpson

C. L.

Troy

S. A.

Safety Effectiveness of Flashing Yellow Arrow: Evaluation of 222 Signalized Intersections in North Carolina. Transportation Research Record: Journal of the Transportation Research Board, 2015. 2492: 46–56.

33.

Davis

G. A.

Hourdos

Moshtagh

Development of Guidelines for Permitted Left-Turn Phasing Using Flashing Yellow Arrows. Prepared for Minnesota Department of Transportation, University of Minnestoa, Minneapolis, MN, 2015.

34.

Chen

Wang

Zhang

Yuan

Persad

K. R.

Use of Flashing Yellow Operations to Improve Safety at Signals with Protected-Permissive Left Turn (PPLT) Operations. Prepared for Texas Department of Transportation, Texas Southern University, Houston, TX, 2012.

35.

American Association of State Highway Transportation Officials. A Policy on Geometric Design of Highways and Streets. AASHTO, Washington, D.C., 2011.

36.

Alabama Department of Transportation. Alabama Traffic Data. Alabama Department of Transportation. https://aldotgis.dot.state.al.us/TDMPublic/.

37.

Chia

Dhaliwal

S. S.

Thai

Jia

Evaluation of Actuated, Coordinated, and Adaptive Signal Control Systems: A Case Study. Journal of Transportation Engineering, Part A: Systems, Vol. 143, No. 9, 2017, p. 05017007.

38.

Hutton

J. M.

Bauer

K. M.

Fees

C. A.

Smiley

Evaluation of Left-Turn Lane Offset Using the Naturalistic Driving Study Data. Journal of Safety Research, Vol. 54, 2015, pp. 5.e1–15.

39.

Claros

Chitturi

Bill

Noyce

D. A.

Impact of Geometry and Operations on Left-Turn Gap Acceptance at Signalized Intersections With Permissive Indication. Transporation Research Record: Journal of the Transportation Research Board, 2021. 2675: 367–380.

40.

Zohdy

Rakha

H. A.

Alfelor

Yang

C. Y. D.

Krechmer

Impact of Inclement Weather on Left-Turn Gap Acceptance Behavior of Drivers. Transporation Research Record: Journal of the Transportation Research Board, 2011. 2257: 51–61.

41.

Zohdy

Sadek

Rakha

H. A.

Empirical Analysis of Effects of Wait Time and Rain Intensity on Driver Left-Turn Gap Acceptance Behavior. Transporation Research Record: Journal of the Transportation Research Board, 2010. 2173: 1–10.

42.

Hamdar

S. H.

Mahmassani

H. S.

Chen

R. B.

Aggressiveness Propensity Index for Driving Behavior at Signalized Intersections. Accident Analysis and Prevention, Vol. 40, No. 1, 2008, pp. 315–326.

43.

Yan

Radwan

Influence of Restricted Sight Distances on Permitted Left-Turn Operation at Signalized Intersections. Journal of Transportation Engineering, Vol. 134, No. 2, 2008, pp. 68–76.

44.

Davis

G. A.

Swenson

Field Study of Gap Acceptance by Left-Turning Drivers. Transporation Research Record: Journal of the Transportation Research Board, 2004. 1899: 71–75.

45.

Gattis

B. J. L.

Low

S. T.

Gap Acceptance at Atypical Stop-Controlled Intersections. Journal of Transportation Engineering, Vol. 2, 1999, pp. 201–207.

46.

Polus

Lazar

S. S.

Livneh

Critical Gap as a Function of Waiting Time in Determining Roundabout Capacity. Journal of Transportation Engineering, Vol. 129, No. 5, 2003, pp. 504–509.

47.

Tarawneh

M. S.

Mccoy

P. T.

Effect of Offset Between Opposing Left-Turn Lanes on Driver Performance. Transportation Research Record: Journal of the Transportation Research Board, 1996. 1523: 61–72.

48.

Rahman

Kang

Biswas

Dynamic Dilemma Zone Protection System for High-Speed Signalized Intersections: A Comprehensive Safety -Operational Assessment. Transportation Engineering, Vol. 12, 2023, p. 100174.

49.

Peesapati

L. N.

Hunter

M. P.

Rodgers

M. O.

Can Post Encroachment Time Substitute Intersection Characteristics in Crash Prediction Models?

Journal of Safety Research, Vol. 66, 2018, pp. 205–211.

50.

Biswas

Kang

M.-W.

Rahman

Machine Learning Based Automated Left-Turn Vehicle Counts With Conventional Presence Mode Long-Loop Detectors: Alabama Case Studies. Transportation Research Record: Journal of the Transportation Research Board, 2022. 2676: 570–587.

51.

Smith

Visual Hypothesis Testing With Confidence Intervals. EcoAnalysis Inc., Ojai, CA, 1997.