This paper documents a new method for monitoring traffic collision data from continuous roadway facilities to detect high collision concentration locations. Many existing methods for detecting collision concentration locations require segmentation of roadways and assume traffic collision data are spatially uncorrelated, resulting in both false positives (i.e., identifying sites for safety improvements that should not have been selected) and false negatives (i.e., not identifying sites that should have been selected). The proposed method does not require segmentation of roadways; spatial correlation in the collision data does not affect the results of analysis. This new method has a lower false positive rate than the conventional sliding moving window approach. This paper shows how the proposed method can proactively identify high collision concentration locations and capture the benefit of safety improvements observed in the project location and in neighboring sites.
July 1, 2009