Development of an Operating Speed Prediction Model and Investigation of the Relationship between Speed Variability and Crash Frequency: A Case Study of Residential Urban Streets in the City of Saskatoon

This research focuses on two topics of major interest in highway engineering: predicting operating speeds and investigating the relationship between speed and safety for urban roads. Speed is a foundational parameter in highway safety, highway design and operational analysis. However, appropriate speed levels to meet satisfactory operational targets might negatively affect safety. In fact, higher speed levels can reduce travel time and its costs, but high speeds could potentially lead to higher collision frequency. The study was based on speed and collision data collected for 140 residential streets as part of neighbourhood traffic reviews (NTRs) by the City of Saskatoon between 2016 and 2018. In the first part of the thesis, covering operating speed prediction models, i.e., models used by highway engineers to design and build roads that are expected to generate desired (operating) speed, are developed. In urban residential areas, operating speeds often exceed intended design speeds, as well as posted speeds, creating speeding concerns among residents. One of the reasons for this is a lack of performance-based design procedures that incorporate expected operating speeds during the planning and design stages of urban roadways. Therefore, this study seeks to understand the relationship between operating speeds and various road characteristics, though operating speed prediction models, with the goal of promoting performance-based roadway design procedures. Since 85th percentile speed is commonly used to model operating speeds, 85th percentile free-flow speed prediction models were developed. Multiple linear regression models were developed using fixed and then mixed effects to account for unobserved heterogeneity (unmeasured differences) among neighbourhoods. The results show that segment length, pedestrian crossing density, bus stop density, tree density, traveled-way width, on-street parking, centerline (marking) presence and school zone are highly associated with operating speeds and these effects are subsequently discussed. Moreover, the overall goodness of fit of the mixed-effects model showed the importance of accounting for unmeasured heterogeneity at the neighbourhood level. In the second part of the thesis, the relationship between speed on residential urban streets (in particular, speed variability among vehicles) and road safety in the form of predicted crash frequency is investigated. More research is, in fact, needed to demonstrate whether vehicles travelling slower or faster than mean traffic speed are more often involved in collisions than vehicles travelling close to the mean speed. The coefficient of variation (CV) of speed was computed for this research as a mediator variable, a variable which is predicted by roadway and traffic characteristics to predict crash frequency. This latter modelling was conducted using regression models within a path analysis framework, a methodology that can describe both direct and indirect dependencies among a set of variables. Within path analysis, a relationship between an independent and a dependent variable can be direct or mediated by a third factor. The results showed that CV was positively related to crash frequency (sites with higher speed variation showed lower safety levels) and the relationship was found to be statistically significant. Moreover, the interrelationship among roadway and traffic factors and crash frequency was also analyzed, providing a better understanding of the indirect effect of independent predictors of CV on crash frequency. Overall, the results could be particularly important in the context of using speed-related variables as surrogate measures of safety, which would allow assessment of safety levels for urban residential streets by measuring speeds without waiting for collisions to occur.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.