Evaluating the Combined Effects of Weather and Real-Time Traffic Conditions on Freeway Crash Risks

Abstract

Abstract The study presented in this paper investigated the combined effects of environmental factors and real-time traffic conditions on freeway crash risks. Traffic and weather data were collected from a 35-km freeway segment in the state of California, United States. The weather conditions were classified into five categories: clear, light rain, moderate/heavy rain, haze, and mist/fog. Logistic regression models using unmatched case-control data were developed to link the likelihood of crash occurrences to various traffic and environmental variables. The sample size requirements for case-control studies and the interaction between traffic and environmental variables were considered. The model estimation results showed that the light rain, moderate/heavy rain, and mist/fog significantly increase freeway crash risks. The interaction between light rain and upstream occupancy was also found to be statistically significant. Bootstrap analyses were conducted to quantify the interaction effect between these two variables. The crash risk model was compared to a reduced model in which environmental information was not included. It was found that the inclusion of environmental information improved both goodness of fit and prediction performance of the crash risk prediction model. The inclusion of environmental information in crash risk models improved the prediction accuracy of crash occurrences by 6.8% and reduced the false alarm rate by 1.3%. It was also found that the inclusion of environmental information had minor impacts on the prediction performance of the crash risk model in clear weather conditions.