Estimating the relationship between measured wind speed and overturning truck crashes using a binary logit model

This paper develops a quantitative model that correlates overturning freight vehicle crash records in Wyoming to measured wind speeds at nearby weather stations. The database consists of 14,700 truck crashes from 1994 to 2003 and wind speed and gust information from 21 weather stations. A binary logit model was estimated from the data to determine if there was significant correlation between weather station wind data and the likelihood that the crash was of the overturning type. While it is reasonably known that local wind speeds at the crash location are critical in predicting overturning truck crash likelihood, it was not known if distant weather station data were an adequate predictor of these crash types. The results from this work indicate that weather station data can be used as a predictor of overturning crashes. This work provides the necessary first step for the development of operational rules for roadway sections that run high risk of overturning truck crashes in high wind conditions.