Classifying Typical Hazard Scenarios And Analyzing Injury Severity Of Automobile-Two-Wheeler Accidents Considering Individual Heterogeneity

Two-wheeler riders are more likely to suffer more serious injuries in car and two-wheeler accidents than other road users.Therefore,studying the injury severity of two-wheeled cyclists and the differences under different crash scenarios is the key to proposing relevant accident prevention countermeasures and improving road safety.Previous studies have been fruitful in analyzing the injury severity of two-wheeled cyclists,but have not taken into account the inherent heterogeneity in crash data and the impact of crash scenario differences,which resulting in the obscuring of some important influences that are critical to the development of effective safety countermeasures.To further investigate the differences in injury severity of two-wheeled cyclists in different crash scenarios,the paper first collates a sample of traffic accident in-depth study(CIDAS)data from eight cities in China for the period 2010 to 2020 and selects variables such as two-wheeler rider characteristics,interactive object driver characteristics,crash characteristics,and road and environment characteristics as independent variables to analyze two-wheeler accident sample characteristics and perform descriptive statistics.Secondly,identifying typical hazard scenarios of car-two-wheeler collisions based on latent class clustering models and scene element variables,and determining the optimal number of clusters.Then,three types of fixed-parameter logit models namely,ordered logit models,generalized ordered logit models,and partial proportional odds models and random-parameter logit models,were developed to analyze and compare the goodness of fit and the explanatory power of different models for the injury severity of two-wheeled cyclists.Finally,in order to compare and analyze the differences in injury severity of two-wheeled cyclists in different collision scenarios and the effects of intragroup heterogeneity,the random parameter logit model with the best fit was selected to model and compare the injury severity of cyclists in intersection collision scenarios and general roadway collision scenarios,respectively,and to propose corresponding preventive countermeasures according to the research results.The research results show that a total of seven types of typical hazard scenarios of cartwo-wheeler collisions were identified based on the latent class clustering model,including four types of intersection collision scenarios and three types of general roadway collision scenarios,with significant differences in road speed limits,major two-wheeler collision locations and major car collision points under different collision scenarios.Second,compared to the fixed parameter logit model,the random parameter logit model fits best for the injury severity of twowheeler cyclists and is able to identify the effect of inherent heterogeneity in the accident data,e.g.,in the overall model of two-wheeler accidents,variables such as time of night,bus versus truck,adverse weather conditions,road linearity,and age of the cyclist have heterogeneity in the injury severity of different individual cyclists.In addition,the analysis of injury severity of cyclists in different collision scenarios revealed that the same variable had different or even opposite effects on injury severity of cyclists,also revealing the effect of data heterogeneity within different collision scenarios.The study further reveals the differences in injury severity of cyclists during car and two-wheeler collisions and provides a basis for developing targeted prevention countermeasures.