In South Dakota, an average of 15,000 crashes, including 120 fatal crashes, occurr over 83,744 miles of highway every year. The density of these fatal crashes was barely 0.0015 crashes per mile. It is apparent that hot spot analysis may not be applicable for many locations. A system-wide deployment of safety treatments with substantial coverage may be more appropriate. However, without an effective systematic approach to identifying the boundaries of safety improvement projects, it will be cost prohibitive to repair and retrofit tens of thousands of miles of highways. Given the sparsely distributed crashes across various highway systems, this study designed an empirical Bayes (EB) based sliding window technique within a spatial context. By examining roadway safety spatially, the safety analysts are able to account for high-risk locations completely within longer predefined segments and locations, which may include multiple predefined roadway segments. Removing the dependence on predefined segmentation can also bring to the forefront safety issues previously ignored. The robustness of the EB method significantly improves the crash estimation accuracy. In conjunction with several different but complementary safety metrics, a complete view of rural highway safety performance can be presented. To ease the use of such a new technique, the South Dakota GIS Highway Safety Review (GIS-HSR) Tools was developed, which provides a data-driven approach toward identifying high-risk locations. Only very basic user input and interaction is required for the tool, which is implemented on a system-wide basis. SD GIS-HSR is designed to address the rural environment and is tuned specifically for South Dakota. However, the general architecture and design are valid in any location. (Author/publisher)
Abstract