Impact of vehicle speeds and changes in mean speeds on per vehicle-kilometer traffic accident rates in Japan

•Five min mean speed data is used for real time traffic accident prediction.•A two dimensional additive Poisson model is applied.•In addition to mean speeds, changes in mean speeds affect traffic accident rates.•The highest accident rates occur when speed is reduced from about 110 to 85 km/h.•Accident rates are higher when it is sunny weather, than when it is cloudy.

Speed and speed variation are widely believed to be key issues in the understanding of traffic accidents. However, there has not been a substantial amount of research that focuses on the interaction between the mean speed and the change in the mean speeds. In this paper we use a five-minute continuous monitoring data of the mean speed on an expressway in Japan. Applying a two dimensional additive Poisson model, we show that not only mean speeds but also changes in mean speeds affect per vehicle-kilometer traffic accident rates. The highest probability of an accident occurs when speed reduces from 110 to 85 km/h. Another area of high accident probability occurs when the average speed increases from 65 to 90 km/h. In addition, we found that accident rates are higher when there is sunny weather, rather than when it is cloudy.