Safety Screening of Road Networks with Limited Exposure Data

The first and most important step in road safety management is the identification of roads with safety needs. One possible road screening technique is based on the quality control method. Exposure data such as annual average daily traffic and vehicle miles traveled are used to estimate the expected numbers of crashes, which are then compared with the actual numbers of crashes to identify locations with safety needs. This approach must be modified in applications to county and city roads because of the limited availability of exposure data. Local transportation agencies are not expected to establish traffic counting programs to cover all their roads in the foreseeable future. This paper proposes estimation of the expected number of crashes on county roads on the basis of surrogates of exposure data including (a) segment connectivity with busier state-administered roads and (b) characteristics of the land development in the area. Land development data are included in the census database and are widely used in transportation planning. A method of estimating the expected crashes on the basis of the classification tree is proposed, and application of the method to screen Indiana county roads is presented. The developed classification tree is compared with the benchmark model, negative binomial regression further improved by adding random effects. The proposed classification tree performs better than the benchmark model. The research demonstrates a promising method of screening local roads in a defensible manner.