The Effect of Weather Temporal Instability on the Injury Severity in Single-Vehicle Crashes: A Random Parameters with Heterogeneity in Means Approach

Nowadays, most current studies on weather factors are qualitative analyses and do not consider temporal instability in prediction models. To this end, with road traffic crash data from 2015 to 2017 in Hong Kong and real-time weather data of 1-min accuracy, this study explores the temporal instability of the weather impact on the injury severity of single-vehicle crashes. To use Stata software for the contributing factors selection and multicollinearity test, we matched the Hong Kong crash data with the high-precision weather data from the Hong Kong Observatory. Then, we used NLOGIT to establish fixed parameter logit models, mixed logit models, and mixed logit models with mean heterogeneity, respectively. The results show that the model parameters affecting the injury severity of single-vehicle crashes have significant temporal instability. The injury severity of single-vehicle crashes increases in fall when the humidity is over 80% or the air temperature is between 20 degrees C and 25 degrees C, as the opposite of the winter.