The majority of fatal and serious crashes in New Zealand occur on rural two-lane roads. Data on historic crash patterns is not always sufficient to enable a suitable diagnosis of the safety deficiencies of various sections of this rural road network. It also cannot readily identify safety issues on low-volume roads and shorter sections of highway, where the relative scarcity of crashes may mask the considerable potential for safety improvements.
This pilot study covers the second stage of a three-stage research project that aims to quantify the impact of all key road features on the safety of two-lane rural roads. This stage of the study involved the collection of road alignment, roadside environment, traffic flow, and crash data for 200 sections of rural road, each one 400m long, throughout the Waikato region of New Zealand. The data was used to develop preliminary crash prediction models for two-lane rural roads, using generalised linear regression model techniques developed by Beca.
The data collection exercise covered a total of 28 predictor variables used for developing the preliminary model. The preferred model showed that the crash rate was most influenced by five predictor variables – namely, traffic volume, absolute gradient, distance to non-traversable hazards, skid resistance (SCRIM), and percentage reduction in the curve-negotiation speed of the section as compared with the preceding 500m section.