Evaluating Pedestrian Safety Using System-Centric and User-Centric Metrics Across Day and Night Scenarios 8 views

Pedestrian fatalities have been rising in the United States since 2009. In Virginia, the number of pedestrian-related crashes increased substantially by 43% between 2020 and 2024. Although the increase in crashes involving pedestrians has been observed across all times of day, the increase is notably higher at nighttime compared to daytime. This dissertation explores the use of system-centric and user-centric approaches in proactively evaluating pedestrian safety across different times of day in Virginia. Efforts to understand pedestrian safety have traditionally relied on real world observations, examining crash data before and after implementation of safety countermeasures. Evaluation of the effectiveness of safety improvements is time-consuming, as crash analysis typically requires 2 to 3 years of crash data before and after implementation. Additionally, crash data tend to be overdispersed and underreported for pedestrians. This dissertation consists of two studies using two distinct sets of data to demonstrate more proactive approaches to pedestrian safety analysis. The first study takes a system-centric approach by utilizing video data and focuses on pedestrian safety at two geometrically-similar signalized intersections, one banning right turn on red (Alderman in Charlottesville), and one allowing permissive right turns on red (Beamer in Blacksburg) in Virginia. Interactions between pedestrians and vehicles were studied using the safety surrogate measure (SSM) post-encroachment time (PET), measured in seconds. The objectives of this research were: (i) to examine the variation of the rate of near-miss events across day and night periods, and (ii) to examine the severity of pedestrian-vehicle interactions across day and night periods. The findings showed that the rate of near-miss events or traffic conflicts between pedestrians and vehicles was lower at nighttime, compared to daytime at both signalized interactions. However, the near-miss events between pedestrians and vehicles at dusk were more severe at the two intersections compared to daytime. At the Beamer intersection, the nighttime near misses were also more severe than daytime near misses. The second part of this dissertation takes a user-centric approach by utilizing pedestrian-specific datasets, i.e., perceived safety data from surveys and gaze data from eye-tracking glasses. 63 participants walked along an urban street in Charlottesville, Virginia, where they could choose to cross anywhere along a four city-block corridor across day and night periods. Using instantaneous and sequential choice models, the goal of this research was to obtain a more informed understanding of how pedestrians make street crossing decisions at intersections in day and night scenarios. The findings showed daytime decisions were affected more by static variables such as pedestrian’s age, household income, and the presence of traffic signal and pedestrian push button at an intersection. The provision of the pedestrian push button at a signalized intersection consistently increased the probability of a pedestrian choosing that crossing location during day and night, but dynamic variables like perceived safety and physiological behavior seemed to heavily influence pedestrians’ crossing choices at nighttime. Together, these two studies demonstrate two types of proactive approaches to evaluate pedestrian safety. Identifying the frequency and severity of near miss events between pedestrians and vehicles may help transportation planners and engineers implement safety improvements effectively, but system-centric metrics like PET cannot account for unobserved heterogeneity in road user behavior, and may misclassify conflict severity. User-centric metrics may be more useful in proactive evaluation of different types of pedestrian infrastructure, e.g. through simulation environments. Alternatively, SSMs may be used to continually monitor and evaluate infrastructure once it has been implemented. In sum, the data-driven approaches to pedestrian safety used in this dissertation examine the interplay between transportation infrastructure and pedestrian preferences in crossing behavior.

PHD (Doctor of Philosophy)

Pedestrian safety; Time-of-day; User-centric metrics; System-centric metrics

English

All rights reserved by the author (no additional license for public reuse)

2026-04-20