Analysis of Naturalistic Driving Study Data: Offset Left-Turn Lanes (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

1Executive Summary The purpose of this research project was to demonstrate one application of the SHRP 2 Natural- istic Driving Study (NDS) data, showing how these data can be used to answer a highway safety question and provide guidance for the implementation of safety countermeasures. The specific safety countermeasure evaluated in this project is offset left-turn lanes. The research team evalu- ated the effect of left-turn lane offset on gap-acceptance behavior, which serves as a surrogate safety measure for this study. Specifically, the objective was to evaluate left-turning gap rejection and acceptance by a large number of drivers at a large number of intersections with a broad distribution of left-turn lane offsets (ranging from negative to positive). Negative offsets typically occur at intersection approaches with medians, in which the oppos- ing left-turn lane is shifted to the left in the field of view of a left-turning driver. Left-turn lanes with negative offsets can create a situation in which opposing left-turning vehicles block each other’s view of oncoming through vehicles. Left turn lanes with a zero offset are those that directly oppose each other. Positive offsets are a countermeasure that can be implemented by moving each left-turn lane to the left within the median (creating space between the through lanes and left-turn lanes) so that the opposing left-turn lane moves to the right of the driver’s field of view. Designing the intersection such that opposing left-turn lanes are each located fur- ther to the left, in positions that provide a positive offset, reduces the instances in which a left- turning driver’s view of opposing through vehicles is blocked by an opposing left-turning driver. In this study, data were obtained for situations both when the turning driver’s view was obstructed by the presence of an opposing left-turn vehicle and when it was not. This type of study would not be feasible using a traditional fixed-camera method for two major reasons. First, the cost associated with data collection at so many sites would likely be too great to be practically feasible. Second, from a fixed camera outside the vehicle, it would be difficult to determine if the driver’s view was indeed obstructed. This research was conducted in two phases. Phase 1 was designed as a proof-of-concept study, in which the research team requested and was provided small samples of the earliest data gathered from trips made by study drivers while the NDS was still in progress. The pur- pose of Phase 1 was to determine whether the NDS data could be used to answer the specific proposed safety research question and to demonstrate how the data would be used in a full analysis when more of the NDS data were available for querying. The Phase 1 data set was intentionally limited in size—NDS data were obtained for only six intersections. However, the research team was able to successfully demonstrate a data collection methodology and an analysis plan and was awarded a Phase 2 contract. The purpose of Phase 2, the results of which are presented in this report, was to answer the highway safety question proposed in Phase 1 using a much larger sample of NDS data. Phase 2 was conducted at the time that the NDS was concluding. At the same time, the development of the Roadway Information Database (RID) was nearing completion, and the RID data, in a very

2early form, were made available to the research team to assist with intersection site selection. In Phase 2, the site selection began with a list of nearly 6,500 intersections. Through a series of steps in which the list was filtered and the intersections were reviewed, 44 signalized intersection left-turn offset pairs (at 33 intersections) and 14 two-way stop-controlled intersection left-turn offset pairs (at 14 intersections) were identified for inclusion in the analysis. This reduction from 6,500 candidate intersections to 47 intersections included in the study was due both to selection criteria and to the limited number of left turns made by NDS drivers at each intersection approach available to be queried from the NDS data at the time the research was conducted. The primary source of data for this study consisted of the forward- and rear-facing cameras placed in the vehicles of NDS drivers. Clips of these videos were obtained for instances in which NDS drivers made a left-turning maneuver at one of the study intersections. A video data reduc- tion interface was developed to make data reduction as efficient as possible while minimizing data entry errors and allowing synchronized viewing and control of the forward and rear videos. From the videos, data were collected for a large number of variables, including weather and lighting conditions, signal indications at specific times, the presence of opposing left-turning vehicles, the presence of following vehicles, and—most important—the start and end time of each gap rejected or accepted by the turning driver. In addition, many videos provided views of non-NDS drivers making left turns; to the extent that information could be collected about their gap-acceptance and gap-rejection decisions, this information was recorded as well. Only about 20% of the NDS data set that will ultimately be available was completed in time for inclusion in this research. The primary analysis for the study was a logistic regression to predict the critical gap length (the length of gap equally likely to be accepted and rejected by a driver) for each offset category, both when sight obstructions were present and when they were not. The analysis of all gaps found that, in general, critical gap length increased as left-turn offset became more negative, although the differences between offset categories were not statistically significant. This result was expected since the left-turning vehicle typically has farther to travel and requires more time to clear the intersection in cases with a large negative offset. Positive offsets typically shift left- turning vehicles closer to the opposing through lanes, shortening the distance they must cross to clear the intersection and allowing them to take shorter gaps. The critical gaps in cases when sight distance was restricted by an opposing left-turn vehicle were then compared with cases when it was not restricted. Critical gaps were found to be 2 s longer when sight distance was restricted than when it was not restricted. When all offset categories were combined, this dif- ference was statistically significant. This result indicates that intersection geometries that allow opposing left-turning drivers to block each other’s sight distance can have a negative effect on intersection operation, since longer critical gaps mean that drivers must wait longer at the intersection to find a gap they feel comfortable accepting. When each driver waits longer for an acceptable gap, intersection delay increases for turning drivers and the intersection level of service decreases. The research team requested information about any crashes or near crashes recorded in the NDS data set at any of the study intersections. While a total of six events took place at one of the study intersections, none were related to left-turn maneuvers. In addition, the video data reduc- tion included a variable to record any instance of avoidance maneuver made by a study driver or an opposing driver during the left-turn maneuver. Of the nearly 3,500 events recorded in the database, only six included an avoidance maneuver, and none of these six events demonstrated a specific safety concern. While no safety incidents were observed in the data, an analysis of the shortest postencroachment times (the time between the initiation of the left-turn maneuver and the arrival of the next opposing through vehicle at the intersection) by offset category indicated a potential opportunity for safety improvement at negative-offset left-turn lanes. While, on average, drivers at negative-offset left-turn lanes are more cautious and wait for longer gaps than drivers at other intersections, they are apparently also more likely to leave a short amount of clearance time between their turn and the arrival of the next opposing through vehicle than drivers at other intersections.

3The research team considered a range of other potential analyses, in addition to those described above. However, the data set was found to be too limited to address these other issues, especially when trying to evaluate a combination of variables (such as offset and driver age). In most cases, the researchers determined that a formal analysis would not provide meaningful results; therefore, such analyses were not conducted. However, descriptive statistics are provided in this report for a number of variables that may influence turning behavior, and general observations about trends are made. Several types of analyses related to left-turning behavior could be conducted with a larger data set—that is, if additional NDS videos were reduced for each intersection—or with a data set specifically structured to increase the number of observations under specific condi- tions or by drivers with specific characteristics. The results of this research are most relevant to designers and traffic engineers who have a demonstrated or anticipated safety or operational concern related to permissive left-turn maneu- vers. Differences in critical gap between specific offset categories were not statistically significant; however, the data clearly showed that negative-offset left-turn lanes resulted in longer accepted gaps than positive-offset left-turn lanes because opposing left-turn vehicles were much more likely to cause a sight restriction at negative-offset left-turn lanes than at positive-offset left-turn lanes. Therefore, even intersections with opposing left-turn lanes with a minimally negative offset (-1 ft to -5 ft) have the potential to benefit from shifting the lanes to a positive offset. In addition, the data did not show that offsetting the left-turn lanes to make them less negative (e.g., changing an offset from -16 ft to -6 ft) would have a substantial effect on safety or operational concerns.