Detecting the spatial–temporal autocorrelation among crash frequencies in urban areas

Abstract

Reducing the number and severity of crashes has been a primary concern of safety specialists. Identifying the crash patterns is an essential step in safety management since clustering of crash data helps the analysts to prioritize the safety improvements to such areas. The spatial essence of crash data particularly the presence of spatial autocorrelation reveals that crash occurrences are not only inclined to cluster in the same locations but within particular time intervals. This paper aims to detect spatial–temporal dependencies of crash occurrences using bivariate Moran's I and local indicator of spatial association (LISA). Employing the yearly crash frequencies aggregated in 253 traffic analysis zone (TAZ) in Mashhad, Iran over 4 years (2006–2009), indicated that both bivariate Moran's I and LISA yield significant patterns of clustering of crashes. The results of this analysis help the safety officials to sufficiently allocate the limited resources by prioritizing the detected clusters.