Developing Safety Performance Functions for Urban Intersections in Sri Lanka

Abstract

Crash frequency modeling is a good approach in identifying factors that influence on crash frequency at road ways or intersections. This research study aimed at developing Safety Performance Functions (SPFs) for urban intersections in Sri Lanka based on crash frequency modeling. This is a novel approach in the local context. The study accommodated 369 urban intersections. For each intersection, crash data, geometric data and traffic related data were collected. Out of 369 urban intersections, 107 intersections were located in Colombo district. An intersection was the unit of analysis in crash frequency modeling. Data were obtained for the period from 2015 to 2019. Police reported crash data under all severity levels were considered for the study. Geometric variables included width of lanes, shoulder, median, side walk and existence of geometric features at an intersection. Vehicle Kilometers of Travel (VKT) data were considered in terms of traffic data. VKT data were collected through an island-wide paper-based travel survey. As timely VKT data are not available with the relevant authorities in the country, estimating VKT is one of key contributions to the existing literature. Count data modeling methods; Poisson and Negative Binomial regression models were adopted to develop crash frequency models. In the study, two different SPFs were developed for Colombo district and the entire country. Out of 36 independent variables that related to geometry and traffic characteristics of urban intersections, few variables were remained statistically significant in the developed crash frequency models. Existence of left turn lanes on major leg, bus halting places, traffic signal lights, illegal road side parking, roads connected to minor leg and average shoulder width of major road variables were remained statistically significant in the Colombo district crash frequency model. In the model that was developed for urban intersections in Sri Lanka, VKT, existence of left turn lanes in minor leg, right turn lanes in major leg, traffic signal lights, road hand rails, presence of approach lanes in the minor leg, existence of kerbs in the minor leg, existence of roads connected to the minor leg, road signs and width of median in the major road variables were found statistically significant. Those variables were recommended to appear in the SPFs. Identifying influencing traffic and geometric variables on crash frequency at urban intersections in Sri Lanka is another key contribution from this study. SPFs illustrated crash frequency at urban intersections in terms of statistically significant variables with a link function. This is the ultimate output of the research study that allowed to propose and compare the most suitable countermeasures in improving safety at urban intersections. Also, limitations and recommendations of the study were discussed that can effectively be adopted in future research studies for a better approach in determining crash frequency at urban intersections in Sri Lanka.