Complete-Linkage Clustering Analysis of Surrogate Measures for Road Safety Assessment in Roundabouts

Abstract

This paper presents the findings of a comparative road safety assessment between an existing two-lane roundabout and proposed basic turboroundabout, both designed for the same intersection, to determine which one is safer, based on traffic conflicts and surrogate safety measures. We performed microsimulation models in VISSIM to replicate the fieldobserved traffic operation, and the SSAM to determinate six surrogate measures. We validated the consistency of values obtained by several statistical analyzes. The number of conflicts was 72% lower at the turboroundabout. Through a complete-linkage clustering analysis and Euclidean distances of the surrogate measures, we found that traffic conflicts at the turbo-roundabout tend to cluster in a group, whereas conflicts at the roundabout are scattered, suggesting better organization of traffic flows at the turbo-roundabout. Three-dimensional graphical analysis of clusters and its centroids allowed verifying that surrogate measures point out a safer operation at the turbo-roundabout, even though it presented higher operating speeds. Reducing the dimensionality by principal components analysis, the cumulative variance for the first two components (87.72%) allowed observing results on a two-dimensional graph and their clusters. To endorse conflicts classification, resulting of clusters, we used discriminant analysis. Results validate the methodology and the safety benefits of the turbo-roundabout.