| The freeway crash rate and fatality rate in China are relatively high,so how to improve freeway traffic safety is an important issue to be solved urgently.Research based on crash data is helpful to discover the risk factors that affect the occurrence and severity of freeway accidents,so as to explore the crash mechanism,and put forward corresponding safety improvement suggestions based on the research results.In recent years,weather conditions have been proved by many studies to have an important impact on freeway traffic safety.Clarifying the specific impact mechanism of weather conditions on freeway accidents and the characteristics of accidents under different weather types could provide effective theoretical support for improving the level of freeway safety.Therefore,it is necessary to incorporate meteorological information based on freeway crash data to carry out research.The current domestic and foreign research on freeway crash that incorporate meteorological information has limitations such as low consistency between research hypotheses and objective reality,and low matching of research methods and data features,resulting in insufficient accuracy and reliability of research conclusions.This article takes the Kaiyang Freeway in Guangdong Province as an example.Based on the crash,roadway and traffic data,the meteorological data is incorporated.Based on the econometric models and unsupervised data mining methods,this paper improves the limitations of the above research,and explores the mechanism of the impact of meteorological elements on the severity and frequency of freeway accidents,excavate accident characteristics and regular patterns under different weather categories,so as to put forward reasonable freeway safety improvement suggestions.Then,identify crash hotspot based on the research results.The above measures provide a basis for decision-making for the traffic management department.The main research contents of this article include:(1)Analysis of crash injury severity level incorporating real-time meteorological data.To reduce the proportion of casualties in freeway crash,firstly,this paper matches the corresponding real-time meteorological data for each accident according to the time and location of the accident,and establishes crash severity model to explore the impact of the change of meteorological elements on the crash injury severity.Then,considering that there might exist spatial correlation effect in the modeling dataset,this paper establishes a spatial multinomial logit model,and compares its fitting ability with other discrete outcome models.The results show that wind speed and humidity have a significant impact on the injury severity level of freeway crash.Specifically,the increase of wind speed and humidity reduces and increases the probability of serious injury and above crash,respectively.According to the research results,this paper puts forward safety improvement suggestions that can reduce accident fatalities.(2)Crash frequency analysis considering interaction and nonlinear effect.To reduce the number of freeway crash,this paper establishes a crash frequency model to explore the impact of meteorological factors on freeway crash count.To investigates the complicated influence mechanism,this paper analyzes it from the two angles of considering interaction and non-linear effects.Firstly,considering that the impact of the same road characteristics on crash risk may be different when the weather condition differs,this paper considers the interaction between meteorological elements and roadway characteristics in the crash frequency model,and establishes a Bayesian spatio-temporal model to fit the data.The results show that when the wind speed is high,the accident risk of steep slope is high;When the precipitation is large,the accident risk of curved road segment is high;and the increase of temperature is accompanied by the increase of accident risk.In addition,considering that the relationship between meteorological elements and accident risk may be nonlinear,the generalized additive model based on Poisson distribution is used to fit the data.The results show that the increase of average wind speed and the decrease of visibility will be accompanied by the increase of accident frequency.The increase of average precipitation is accompanied by the increase of accident risk,but when the average daily precipitation reaches 9.5 mm,the increase of precipitation is accompanied by the decrease of accident count;High temperature is accompanied by high accident risk.According to the research results,this paper puts forward safety improvement suggestions that can reduce crash counts.(3)Crash pattern analysis considering differences in weather categories: Due to the different accident regular patterns corresponding to various weather conditions,to further explore the crash mechanism,based on the weather and climate characteristics of Guangdong Province,this paper divides the weather conditions into three categories: good weather,low visibility weather and rainy weather,and uses the multiple correspondence analysis method to explore the correlation relationship between the corresponding variables of crash under various weather conditions(also called "crash pattern").The results show that humidity and visibility have important impacts on the occurrence of low visibility-related crash;Rainfall amount has an important impact on the occurrence of crash in rainy weather,and there is a significant difference between rainstorms and other levels of rainfall;In rainy weather,precipitation amount is closely related to the occurrence of casualty crash,single-vehicle crash and multivehicle crash;In low visibility weather,high humidity is closely related to single-vehicle accidents.According to the research results,this paper puts forward safety improvement suggestions that can reduce accident fatalities and crash counts.(4)Crash hotspots identification based on Bayesian spatio-temporal model.Based on the above research content,at the application level,in addition to providing corresponding safety improvement suggestions,the safety level of freeway segments can also be ranked based on the prediction results of the crash frequency model,which is related to crash hotspot identification.Since researchers have proved that the full Bayesian method has high reliability in the identification of crash hotspot,and the Bayesian spatio-temporal model is superior to other Bayesian models in(2)in terms of model fitting ability,the prediction result of Bayesian spatiotemporal model in(2)would be adopted to identify crash hotspot.By comparing the ranking and crash hotspot identification result between Bayesian spatio-temporal model and other methods,it could be concluded that considering the uncertainty and analyzing the spatiotemporal correlation effect has an important influence on it.Then,the results and distribution of crash hotspot on Kaiyang Freeway in Guangdong Province are displayed in the form of figures and graph to provide a basis for decision-making by traffic management departments. |
PDF Full Text Request