Spatial analysis of weather crash patterns

Spatial statistical techniques can be an effective tool for analyzing patterns and autocorrelation in crash data, especially weather-related crashes. Since weather is a geographic phenomenon, it tends to show distinct geographic patterns affecting certain locations more than others. Accordingly, "weather-related" crashes may also display similar distinct patterns or clustering. The objective of this research was to use spatial statistical techniques to identify significant patterns of weather-related crashes. Weather-related crashes, defined as those crashes which occurred in adverse weather conditions, were analyzed using the Getis-Ord G(i)*(d) statistic. The statistic reveals spatial patterns for weather-related crashes which are clustered at different locations depending upon weather conditions (snow, rain, and fog). The results also show geographic areas (counties) of statistically significant high and low relative crash rates for each weather condition. Furthermore, the resulting patterns of crashes were validated by comparing counties of high and low crash rates with areas of varying weather data. The establishment of this relationship between weather and crashes is imperative in identifying the variables contributing to these crash types and the implementation of effective countermeasures for road weather safety audit purposes.