Developing Zonal Crash Prediction Models with a Focus on Application of Different Exposure Measures

Abstract

Assessing the safety impacts of Travel Demand Management (TDM) policies is essential to be carried out by means of a proactive approach. Since TDM policies are typically implemented at an aggregate level, Crash Prediction Models (CPMs) should also be developed at a similar level of aggregation. These models should match better with the resolution at which TDM evaluations are performed. Therefore Zonal Crash Prediction Models (ZCPM's) are considered to construct the association between observed crashes and a set of predictor variables in each zone. This is carried out by the Generalized Linear Modeling (GLM) procedure with the assumption of Negative Binomial (NB) error distribution. Different exposure, network and socio-demographic variables of 2200 Traffic Analysis Zones (TAZs) are considered as predictors of crashes in the study area, Flanders, Belgium. To this end, an activity-based transportation model framework is applied to produce exposure measurements. Crash data used in this study consist of recorded injury crashes between 2004 and 2007. The network and socio-demographic vaiables are also collected from other resources. In this study, different ZCPMs are developed to predict the Number of Injury Crashes(NOICs); including fatal, severely and slightly injury crashes. These models are classified into three different groups, i.e. 1)flow-based models, 2)trip-based models and 3)a combination of the two. The results show a considerable improvement of the model performance when both trip-based and flow-based exposure variables are used simultaneously in the model's formulation. The main purpose of this study is to provide a predictive tool at the planning-level which can be applied on different TDM policies to evaluate their traffic safety impacts.