November 20, 2014

About the Presentation

Traditionally, spatial crash analyses are point-, segment-, or zonal-based. The most basic statistical evaluation of crash mapping includes mean center, standard distance, and aggregating crashes based on roadway segments and spatial zones. However, these statistical tools only provide preliminary results. Enhancements over the past 20 years include the utilization of global and local calculations of the Getis-Ord Gi and Moran’s I statistic when evaluating the spatial autocorrelation seen in point-based maps. Segment-based analyses continue to evolve from basic aggregated crash counts to more refined crash densities. And finally, zonal-based analyses continue to evolve by comparing neighboring zones using clustering identification and smoothing crash counts using Bayesian statistics.

This seminar reviewed the proposed future of crash mapping, which uses increasing computer power along with improvements in crash location technologies. The synergy of these efforts will allow for the identification of high-risk crash areas. The improvements in crash areas will occur by the interpolation of clustered crashes, the identification of the spatial relationships of environmental impacts, and the factors that cause movements within the predicted areas.

About the Speaker

William Schneider

William Schneider is an assistant professor of civil engineering at the University of Akron. His research interests include transportation air quality, ITS data management, and the impacts of geometric design on safety and operations. Prior to joining the University of Akron, Schneider spent more than two years at the Texas Transportation Institute, where he worked on the following projects: developing predictive models for estimating speeds of vehicles traveling in urban right-turn lanes, incorporating safety in the highway design process, work-zone speed compliance and safety, and more.