Investigating the effect of spatial and mode correlations on active transportation safety modeling

Abstract

This paper describes the development of macro-level crash models for active modes of transportation incorporating spatial and mode correlation effects. The models are based on data from 134 traffic analysis zones (TAZs) in the City of Vancouver. Five years of cyclist and pedestrian crash data, as well as traffic exposure and large GIS data, were used to establish the macro-level crash models. The GIS data included land use, built environment, socioeconomic, bike network, and pedestrian network indicators. Full Bayesian multivariate models with and without spatial effects were developed and compared to the corresponding univariate models. The multivariate modeling approach allowed for including a different set of covariates for each modeled crash type. The univariate/multivariate crash models incorporating spatial effects consistently outperformed those that did not account for spatial effects. The correlation between pedestrian and cyclist crashes was found significant indicating the importance of accounting for the dependency among active commuters’ crash types. The mode and spatial correlations were affected by the number of the explanatory variables added to the model. Overall, the multivariate models outperformed the univariate ones, and the multivariate model incorporating spatial effects yielded the best fit among all the tested crash models. The associations between cyclist as well as pedestrian safety and various zones’ characteristics were also investigated in this study.