| Freeways typically have a design speed of 80-120 km/h and an annual average daily design traffic volume of more than 15,000 passenger cars,which undertakes medium-and long-distance travel within and between cities.China’s freeways are developing rapidly,with a total mileage of more than 160,000 kilometers in 2020,ranking the top in the world.However,the large number of existing freeways are facing prominent traffic safety issues: the annual number of fatalities due to traffic crashes on China’s freeways has remained at around 6,000 since 2007.It is urgent to carry out research on freeway traffic safety assessment and improve freeway traffic safety management.Freeway safety assessment in China lacks the basic statistical models,making the impact of factors such as road design and traffic control on crashes unclear.In addition,the identification of crash hotspots in China is based on the number of crashes or fatalities,which has ignored the random fluctuation of crash data and made it impossible to evaluate the potential for safety improvement.While looking at the United States(U.S.),its Highway Safety Manual(HSM)has established the crash contributing factors analysis and hotspot identification methods based on safety analysis models.Investigating the transferability of U.S.freeway safety analysis models to China’s freeways has great significance for the rapid establishment of China’s freeway safety assessment methods.On the other hand,China’s freeways have a significantly higher proportion of single-vehicle crashes than other road facilities,and the freeway crashes also exhibit site correlation and spatial correlations,which have not been studied sufficiently in U.S.and European countries,making it necessary to conduct further safety assessment research based on the complex characteristics of China’s freeway crash data.Traffic law enforcement is an important means of traffic safety management.The implementation of traffic enforcement involves key elements such as enforcement patterns,enforcement locations,enforcement time,vehicles to be focused on,and illegal behaviors to be captured.To optimize freeway traffic safety enforcement,it is necessary to clarify the impact of traffic enforcement on crashes and determine the focus of enforcement measures.Traffic enforcement acts on road users and reduces traffic crashes by reducing traffic violations.Therefore,traffic enforcement,traffic violations and traffic crashes are endogenous to each other,which brings challenges to revealing the safety impact of traffic enforcement.Present freeway safety enforcement in China mainly relies on experience,resulting in a biased focus of traffic enforcement,which is manifested in the mismatch between the time,location,and illegal driving behaviors captured and the characteristics of traffic crashes.To improve the initiative and pertinence of traffic enforcement,it is urgent to optimize the above key elements of traffic enforcement based on crash information.Focusing on freeways,this study has firstly evaluated the transferability of U.S.safety analysis models to China,and conducted the safety assessment research based on the complex crash characteristics of Chinese freeways.Then the impact of traffic safety enforcement on traffic crashes on Chinese freeways was investigated,and the key elements of freeway traffic enforcement were excavated from crash information.The main research contents and conclusions of this study are as follows:Firstly,investigating the transferability of U.S.freeway safety analysis models to China.Based on the freeway crash data of U.S.Austin,New York,Orlando,and Chinese Shanghai and Suzhou,the local safety analysis models were developed for each city according to the suggestion of HSM.The models indicated that freeway crashes in the U.S.were more sensitive to traffic volume than in China.The direct transfer of safety analysis models showed poor model transferability from the U.S.to China.Then the calibration factor given by HSM was employed to calibrate the models being transferred,but both the transfer index and the cumulative residual curve showed a limited improvement in the model transferability after calibration.Consequently,it is necessary to carry out safety assessment research according to the crash data characteristics of Chinese freeways.Secondly,establishing the local safety analysis model according to Chinese freeways’ crash data characteristics,analyzing crash influencing factors,and identifying hotspots.Based on Shenhai freeway’s crash data,the discrepancies in the spatial distribution of single-vehicle(SV)and multi-vehicle(MV)crashes,site correlation between SV and MV crashes,and spatial correlation have been found.To address the three data features,the bivariate conditional autoregressive negative binomial model(Bi-CAR-NB)was developed,and potential for safety improvement was used to identify crash hotspots.The results showed that SV and MV crashes differ in both influencing factors and hotspots,furthermore,the safety analysis model based on total crashes was proved to be inadequate for safety analysis and hotspot identification.Based on the modeling results,engineering countermeasures were developed for freeway segments near the entrance and exit ramps.To summarize,differentiating between SV and MV crashes has advantages over total crashes when developing safety analysis models,as it can improve the model performance as well as the pertinence of engineering countermeasures.Thirdly,analyzing the impacts of traffic enforcement on freeway traffic safety through multivariate time series technologies.Focusing on the entire freeway system in Shanghai,police patrolling time,technology-detected violations,police-detected violations,and crash data were collected in the time series structure at the daily level.Taking the four time-series variables as endogenous variables and holiday and weather as exogenous variables,a vector autoregressive model,a structural vector autoregressive model,the granger causality test,and a BEKK-GARCH model were developed to examine the contemporaneous effects,dynamic interactions,and potential interactions between enforcement,violation,and crashes.The results indicated that traffic crashes and violations had weekly variation and were significantly impacted by holiday and weather conditions,while police patrol time was not impacted.A contemporaneous negative impact of police patrol time was found in traffic crashes,while technology-detected violations and police-detected violations showed opposite effects on contemporaneous crashes because police officers played a vital role in enforcement activities.In terms of dynamic interactions,the lag of traffic crashes had significant impacts on current police patrol time,but not vice versa.Finally,significant potential interactions in conditional variances of traffic enforcement,violations and crashes have been identified.By employing multivariate time series technologies,this study has revealed the complex relationships between enforcement,violations,and crashes while dealing with the endogeneity among them.The findings are helpful to assist the optimization and adjustment of traffic enforcement intensity.Fourthly,identifying traffic enforcement targets for the freeway network and crash hotspots.To improve the implementation of traffic enforcement activities,the crash location,crash occur time,weather conditions,vehicle types,traffic violations,and crash types were collected to construct the traffic crash transaction data set.Focusing on the entire freeway network of shanghai,the condition itemsets that trigger SV crashes and MV crashes were mined through the Apriori algorithm.Items that have a strong correlation with SV crashes include {Raocheng freeway},{passenger cars},and{other improper driving operation},while items that have a strong correlation with MV crashes are {Waihuan freeway} and {changing lanes illegally}.The clarified differences in the conditional itemsets that trigger SV and MV crashes echo the necessity of distinguishing SV and MV crashes when building safety analysis models.Focusing on the top 3 crash hotspots,the target patrol time,weather condition,vehicle type,and traffic violations of traffic enforcement were identified through the Apriori algorithm,which can be used to directly guide the implementation of traffic enforcement activities.This study has investigated the key technologies of freeway traffic safety management from the aspects of safety assessment and enforcement optimization.In terms of freeway safety assessment,the differences between the U.S.and China’s freeway safety analysis models were revealed,and the evaluation found that U.S.models were not transferable in China.Further,the Bi-CAR-NB model was developed for Chinese freeway crash data to deal with the complex crash characteristics,and the differences in influencing factors and hotspots between SV and MV crashes were clarified.In terms of enforcement,the complicated interactions including contemporaneous effects,dynamic interactions,and potential interactions among traffic enforcement,traffic violations and traffic crashes have been revealed using a series of multivariate time series models.The traffic enforcement targets including patrol time,weather condition,vehicle type and traffic violations were optimized for freeway network and crash hotspots based on association rules.The research findings have important theoretical and practical significance for China’s freeway traffic safety management. |
