Estimating Safety Performance Functions for Two-Lane Rural Roads Using an Alternative Functional Form for Traffic Volume
Vikash Gayah, Pennsylvania State UniversityShow Abstract View Presentation
Eric Donnell, Pennsylvania State University
Statistical models of expected crash frequency are referred to as Safety Performance Functions (SPFs) in the first edition of the American Association of State Highway and Transportation Officials’ Highway Safety Manual (HSM). The SPFs in the HSM specify expected annual crash frequencies as a function of various roadway and roadside features, with the most important predictor variable being traffic volume, which serves as a measure of vehicle exposure to crashes. Traffic volumes are typically measured using the average annual daily traffic and are incorporated into the SPFs using a natural logarithm transformation. This specification suggests that the relationship between expected crash frequency and traffic volume increases non-linearly with a constant elasticity over the range of observed values. While researchers concur that the relationship between expected crash frequencies and traffic volume is non-linear, further exploration of the functional form of this relationship may offer additional insights concerning the association between safety performance and vehicle exposure. This paper proposes an alternative functional form for the traffic volume variable in SPFs that allows for different elasticities between traffic volume and expected crash frequency within different traffic volume ranges, while preserving the same general non-linear relationship in existing HSM SPFs. The proposed functional form was applied to SPFs developed for two-lane rural roadways in Pennsylvania. Comparison with SPFs developed using the traditional functional form in the HSM suggests that this proposed functional form offers an improved fit and predictive performance, and thus might be considered for the development of future SPFs.
Safety Evaluation of Directional Interchange with Semi-Direct Ramp Connections and Loops
Khaled Hamad, University of SharjahShow Abstract View Presentation
Abdulkarim Ismail, Freelancer
The purpose of this paper is to evaluate the safety of the Directional Interchange with Semi-Direct Ramp Connection with Loops (DI-SDRL). Towards this end, the FHWA’s Interchange Safety Analysis Tool-Enhanced (ISATe) was utilized to predict the safety performance of this interchange under 30 different scenarios covering a wide range of traffic volumes. The performance of this interchange has been also compared to a conventional one, i.e. Directional with Loops Interchange (DLI) in terms of the percentage difference and using the paired t-Test. The results showed that DI-SDRL ramp segments witnessed higher number of crashes than the DLI by the average of 36% and 37% for fatality-injury (FI) and property-damage-only (PDO) crashes, respectively. The statistical t-test results showed that the differences between the two interchanges were statistically significant, except for the freeway segments. On the other hand, the differences reported by crash type were not as large as ramp segments. For example, the DI-SDRL reported higher number of multiple-vehicle crashes than the DLI by the average of 4% and 2% for FI and PDO severity levels, respectively. As for the total SV crashes, the DI-SDRL reported higher number of crashes that the DLI by average of 2% for both FI and PDO severity levels. Nevertheless, the statistical t-test indicated that the difference between the studied interchanges is statistically significant.
Investigation of Associations Between Multiple Freeway Roadway Characteristics and Freight Safety Performance
Juneyoung Park, Hanyang UniversityShow Abstract
Mohamed Abdel-Aty, University of Central Florida
Ling Wang, Tongji University
Gunwoo Lee, Chung-Ang University
Jungyeol Hong, University of Seoul
Since various freeway design features are simultaneously installed on roadways, it is important to assess their combined safety effects correctly. This study investigated associations between multiple roadway cross-section design features on freeways and traffic safety. In order to consider the interaction impact of multiple design features and nonlinearity of predictors concurrently, multivariate adaptive regression splines (MARS) models were developed for all types and freight vehicle crashes. In MARS models, a series of basis functions is applied to represent the space of predictors and the combined safety effectiveness of multiple design features can be interpreted by the interaction terms. The generalized linear regression models (GLMs) with negative binomial (NB) distribution were also evaluated for comparison purposes. The results determine that the MARS models show better model fitness than the NB models due to its strength to reflect the nonlinearity of crash predictors and interaction impacts among variables under different ranges. Various interaction impacts among parameters under different ranges based on knot values were found from the MARS models whereas two interaction terms were found in the NB models. The results also showed that the combined safety effects of multiple treatments from the NB models over-estimated the real combined safety effects when using the simple multiplication approach suggested by the HSM (Highway Safety Manual). Therefore, it can be recommended that the MARS is applied to evaluate the safety impacts of multiple treatments to consider both the interaction impacts among treatments and nonlinearity issue simultaneously.
Time-Varying Analysis of Traffic Conflicts in Upstream Toll Plaza Diverging Area
Lu Xing, Southeast UniversityShow Abstract
Jie He, Southeast University
Qing Cai, University of Central Florida
Jinghui Yuan, University of Central Florida
This study investigates traffic conflicts in the up-stream toll plaza diverging area during the entire diverging implementation period. Based on the vehicle trajectory data extracted from video using an automated video analysis system and vehicle collision risk computed by the extended TTC, a time-varying logistic regression (TLR) model is reproduced to examine the time varying effects of influencing factors on vehicle collision risk, and models including the standard random effects logistic regression (RELR) model and random parameters logistic regression (RPLR) model are developed for model performance comparison. Results indicate that the TLR model has the highest prediction accuracy. Eight influencing factors are found that have time-varying effects on collision risk, and six factors are found to exhibit heterogeneous effects at different times. One important finding is that the vehicle comes from initial lane1 has an increasing trend to be involved in traffic conflicts, whereas the collision risk of other vehicles decrease as the travel time increases. Moreover, the vehicle with higher speed has a decreasing probability to be involved in crash with its leading vehicle over the travel time. These findings provide more helpful information to the accurate assessments of the potential for vehicle crash, which is a key step toward improving safety performance of the toll plaza diverging area.
Revisiting Expressway Single-Tunnel Crash Characteristics Analysis: A Six-Zone Analytic Approach
Amjad Pervez, Central South UniversityShow Abstract
Ye Li, Central South University
Huang Helai, Central South University
Chunyang Han, Central South University
Wang Jie, Central South University
Considerable research has been conducted to investigate the tunnels’ traffic safety. However, the entrance and exit parts of a tunnel are mostly considered symmetrical in previous studies, and the different lengths (long, medium, and short) of tunnels have also been studied aggregately. This study aims to investigate the characteristics of traffic crashes in expressway single tunnels by separately considering the entrance and exit of the tunnel as well as the different lengths of tunnels. A six-zone approach is proposed, and the data from 156 single tunnels in Hunan province, China, are applied for safety analysis. The crash rate, crash type, and contributing crash factors are compared between the conventional four-zone approach and the proposed method, and the three different lengths of tunnels are also compared for in-depth analysis.
Variation in Crash Modification Factors Due to Specification and Measurement Error
Yemi Adediji, Ryerson UniversityShow Abstract View Presentation
Robert Noland, Rutgers, The State University of New Jersey
For highway safety projects it is common to use deterministic crash modification factors (CMF) to determine the appropriate safety treatment. CMFs are based on estimates of the correlations of various road geometry attributes with crash frequency, which are derived from models specified according to the analyst’s discretion and affected by the analyst’s data processing decisions. how data is processed. This means that estimation results are subject to large variations in coefficient estimates. Specifying crash models incorrectly (introducing specification error) and using problematic data (introducing measurement error) can lead to bad decisions on which safety countermeasures to implement. This problem may be exacerbated by standardization through manuals such as the Highway Safety Manual (HSM). Using data from North Carolina, we examined the specification error problem associated with omission of variables and measurement error associated with missing data. We compared the results of statistical models affected by the specification and measurement error to models where we attempted to rectify the problems to see if there was an improvement in results. Results were mixed, showing no substantial change in the coefficients between models affected by specification and measurement error and models unaffected for certain variables while showing notable change for others.
Accuracy Assessment of Using Application Programming Interface in the Development of Safety Performance Function on Indian Expressway
Leeza Malik, Indian Institute of Technology, DelhiShow Abstract
Amber Gupta, Ramboll
Geetam Tiwari, Indian Institute of Technology, Delhi
Data accessibility and quality form a critical element in the development of multivariate safety performance function. Over the past few years, access to new large-scale data resources through various map Application Programming Interfaces (APIs) have grown. However, limited understanding is available on the accuracy of the archived data present with these mapping platforms. Thus, the primary focus of the present study is to examine the scope of using the information available at mapping services in developing SPF. An Indian expressway is chosen to illustrate the methodology. The automatic bulk acquisition of link-based operating speed and road geometry is demonstrated using Google APIs and Google Earth. API derived link-based average travel speed, horizontal and vertical geometry is compared with actual data available for the study section. Encouraging accuracies indicate the scope of utilizing the expanding databases of the online map APIs in safety applications. Advantages and limitations of using the online-based information are further explored.
Traffic Conflict Prediction Model in Interchange Diverging Area Based on Gap Acceptance Theory
Zhanji Zheng, Southeast UniversityShow Abstract
Yongfeng Ma, Southeast University
Qiaojun Xiang, Southeast University
Han Li, Southeast University
Yangyang Zhao, Traffic Management Research Institute of the Ministry of Public Security
Lane changing behavior occurs frequently on highway interchange diverging area, which leads to the disruption of traffic flow and frequent traffic conflicts. By the use of unmanned aerial vehicle (UAV) photography and video processing techniques, the vehicle headway distribution on multi-lane was analyzed, besides, the conflict zone and lane changing zone were judged. The gap acceptance theory (GAT) was used to assess the critical gap and the gap interval of traffic conflict, and a traffic conflict prediction model was built to forecast the number of traffic conflicts. The results show that: 1) A significant correlation existed between the lane changing zone and the traffic conflict zone, which were located within 600 m from the reference point and the first 60 m of the deceleration lane. 2) Critical gap of traffic conflict was 2.21s and the gap interval of traffic conflict based on gap acceptance theory was (2.06s, 2.31s). 3) No significant difference was found in the number of traffic conflicts between the traffic conflict prediction model and actual statistical, and the error was blew 10%. Length of deceleration lane and diverging rate are the main factors that affect the number of traffic conflicts, it can provide the basis for the improvement of traffic safety on diverging area.
Economic Assessment of Road Infrastructure Safety Schemes in Greece Using Crash Prediction Methodology
George Yannis, National Technical University of Athens (NTUA)Show Abstract View Presentation
Anastasios Dragomanovits, National Technical University of Athens (NTUA)
Julia Roussou, National Technical University of Athens (NTUA)
Dimitrios Nikolaou, National Technical University of Athens (NTUA)
The economic assessment of intervention projects at hazardous locations aims to maximize road safety benefits by exploiting the full potential of limited available funds for road safety. The study presents a case study for the economic assessment of road safety schemes in crash prone locations in rural highways in Viotia and Imathia sub-regions in Greece, using crash prediction models from the AASHTO Highway Safety Manual. The models were suitably adjusted, calibrated and adapted according to data availability, and were used, along with findings from road safety inspections of the locations under consideration, to estimate expected reductions in fatalities and casualties due to the implementation of specific road safety schemes. Road safety benefits were then translated into monetary terms, and, taking also into account construction and maintenance costs for each scheme, the economic rate of return (ERR) of the project was estimated. The economic rates of return were estimated at 27.1% for Viotia sub-region and 18.2% for Imathia sub-region, thus demonstrating the very high cost effectiveness of road safety intervention schemes in hazardous locations.
Application of Geographically and Temporally Weighted Regression Models for Estimating Safety Performance Functions of Multi-Lane Rural Highways in Tennessee
Amin Mohammadnazar, University of Tennessee, KnoxvilleShow Abstract View Presentation
Numan Ahmad, University of Tennessee
Iman Mahdinia, University of Tennessee, Knoxville
Asad Khattak, University of Tennessee
Although the Highway Safety Manual (HSM) provides default SPFs, they recommend that states develop jurisdiction-specific SPFs using local crash data. Accordingly, crash and road inventory data were integrated for multi-lane rural highway segments in Tennessee covering 2013-2017. Besides developing SPFs similar to those contained in HSM, this study applied a new methodology that can capture variation in crashes both in space and over time. Specifically, Geographically and Temporally Weighted Regression (GTWR) models for localization of SPFs were developed. The new aspect is incorporating temporal aspects of crashes in the models as the impact of a specific variable on crash frequency may vary over time due to several reasons. Results indicate that negative binomial models have a better fit with the crash data than Poisson models and that GTWR models remarkably outperform the traditional regression models by capturing spatio-temporal heterogeneity. Moreover, a majority of parameter estimates vary substantially across space and over time. In other words, the association of contributing variables with the number of crashes can vary from one region and period of time to another. This fact weakens the idea of transferring default SPFs to other states and even applying a single localized SPF for all regions of a state. Enabled by growing computational power, the results emphasize the importance of accounting for spatial and temporal heterogeneity and developing highly localized SPFs. The methodology of this study can be used by researchers to follow the temporal trend and location of critical factors and identify sites for safety improvements.
Safety Performance Functions for Rural, Two-Way, Stop-Controlled Intersections
Ming Sun, University of Louisiana, LafayetteShow Abstract View Presentation
Xiaoduan Sun, University of Louisiana at Lafayette
M. Ashifur Rahman, University of Louisiana, Lafayette
Mousumy Akter, University of Louisiana at Lafayette
Subasish Das, Texas A&M Transportation Institute
Rural intersection safety continues to be a crucial issue throughout the United States. More than 20 percent of all traffic fatalities in the United States occur at intersections, and over 80 percent of intersection-related fatalities in rural areas occur at unsignalized intersections. The first edition of Highway Safety Manual (HSM) has already published the crash prediction models based on the intersection data from several states. Considering each state has unique situations, this paper introduces a safety model development for two-way stop-controlled intersections on rural two-lane highway. After a lenthy data verification process, totally 2,658 rural stop-controlled intersections including both three-leg (3T) and four-leg (4T) from all Parishes (counties) in Louisiana were used for the model development. A series of safety performance functions were developed with Zero-inflated Poisson (ZIP) models with the most recent five-year crash data. The results indicate thatgreater curve radiuses of major roads, greater curve lengths of major roads, greater lane widths of minor roads, and higher speed limits of major roads led to significantly smaller expected crash frenquencies for both 3ST and 4ST intersections. However, unlike 4ST intersections, exclusive right-turn lanes increase the likelihood of crashes at 3ST intersections. In addition to traffic volume,intersections located in the middle of curves led to significantly greater crash occurrences. The results of Louisiana specific models are different from that of HSM models and the difference varies by AADT. The data sources, sample size, modeling structure, and the direct variable selection could have contributed to the difference as well.
Improving the Transferability of the Crash Prediction Model Using the TrAdaBoost.R2 Algorithm
Dongjie Tang, Tongji UniversityShow Abstract
Xuesong Wang, Tongji University
Xiaohan Yang, Tongji University
The crash prediction model is a useful tool for traffic administrators to identify significant risk factors, estimate crash frequency, and screen hazardous locations. Since only limited or low-quality data can be collected for some jurisdictions interested in traffic safety analysis, calibration methods should be applied to an available crash prediction model. The problem with current calibration methods is that the aggregate method limits prediction accuracy and the disaggregate method is resource-consuming. Transfer learning is a technique aimed toward learning knowledge from old data domains to solve problems in new data domains. TrAdaBoost.R2, an instance-based transfer learning technique, is adopted in this paper since it meets the requirement of site-based crash prediction model transfer. A comparison was made to examine TrAdaBoost.R2’s efficiency in extracting knowledge from spatially outdated source data domain (old data domain). The target data domain (new data domain) was split into two parts to test the technique’s adaptability to a small sample size. Calibration factor based on a negative binomial model was employed to compare predictive performance of the transfer learning technique. Mean square error was calculated to evaluate the prediction accuracy. Two cities in China, Shanghai and Guangzhou, were taken as source data domain and target data domain mutually. Results show that the models constructed with TrAdaBoost.R2 improve the prediction accuracy compared to the negative binomial model. The TrAdaBoost.R2 is further recommended due to its predictive performance and adaptability to a small sample size.
Examining the Standard Errors of Crash Modification Factors Developed with Empirical Bayes Before-After Studies
Lingtao Wu, Texas A&M Transportation InstituteShow Abstract
Ying Chen, Changsha University of Science and Technology
Zhongxiang Huang, Changsha University of Science and Technology
Estimating crash modification factor (CMF) for safety treatments (i.e., safety effectiveness evaluation) is an important step in the process of roadway safety management. Empirical Bayes (EB) before-after study is the state-of-art approach for developing CMFs amid various methods, and it is always preferred when applicable. The EB method corrects the regression-to-the-mean bias and improves estimation accuracy. However, the performance of the CMFs derived from the EB method has never been fully investigated. The primary objective of this study is to examine the accuracy of CMFs estimated with EB before-after studies. Particularly, the focus is the quality of the estimated CMF standard errors. Artificial realistic data (ARD) and real crash data on rural two-lane roadways are used to evaluate the estimated CMF standard errors. The results indicate that: (1) The CMFs derived with the EB before-after method are very close to the pre-assumed true values. (2) The estimated CMF standard errors do not reflect the true values. The estimation remains at the same level regardless of the pre-assumed CMF standard error. The EB before-after study is not sensitive to the variation of CMF among sites. (3) The analyses on real-world traffic and crash data with dummy treatment indicate that the EB method tends to under-estimate the standard error of the CMF. Safety researchers should recognize that the CMF variance may be biased when using the EB method to evaluate safety effectiveness. It is necessary to revisit the algorithm for estimating CMF variance within the EB method.
Safety Evaluation of Mandatory Lane-Change Behaviors in Lane-Unbalanced Merging Area Using Vehicle Trajectory Data
Ye Chen, Southeast UniversityShow Abstract
Meng Li, Southeast University
Zhibin Li, Southeast University
Yutin Luo, Southeast University
Lane-unbalanced merging area is a common bottleneck which usually leads to traffic jam and increased crash risks. This paper aims at proposing an insight which combines the lane change decision behavior with traffic states to analyze the safety impact of mandatory lane change from the microscopic level. A total of 292 mandatory lane change trajectories were obtained from UAV videos on an urban expressway in Nanjing, China. Based on the vehicle trajectory data, the lane change duration and location were investigated, and the difference of mandatory lane change decision behavior under free flow state and congested state were considered. Then, an indirect lane change evaluation index was developed to estimate the overall crash risk between the merging vehicle and its neighboring vehicles. The findings suggest that the lane change duration and location are affected by the surrounding vehicles and traffic conditions. With respect to the congested state, the merging vehicles tend to have shorter lane change gap, longer lane change duration, and higher risk potential than the free flow state. The result demonstrates that the traffic state could significantly affect the safety of mandatory lane changes in the lane-unbalanced merging area. Findings of this study will be valuable for the safety evaluation and real-time crash risk prediction in lane-unbalanced merge area.
Developing Safety Performance Functions for Rural Multi-Lane Highways in Tennessee: Accounting for Unobserved Heterogeneity
Behram Wali, Massachusetts Institute of Technology (MIT)Show Abstract View Presentation
Numan Ahmad, University of Tennessee
Asad Khattak, University of Tennessee
Amin Mohammadnazar, University of Tennessee, Knoxville
This study investigates the safety performance of divided and undivided multilane rural highways and explores the presence of unobserved heterogeneity on such highways. First, a unique database was created with five years of crash data (2013-2017), traffic (AADT) data, and roadway inventory data, extracted from various sources of Tennessee Department of Transportation. Then, in addition to testing different functional forms and distributional assumptions, specifications for safety performance functions were developed. The results indicate that statewide calibration factors for divided and undivided multilane highways are 2.573 and 2.347 respectively—this implies that the Highway Safety Manual substantially under-predicts crashes in Tennessee, or that the safety performance on such roadways, as indicated by crash frequency is substantially worse than predicted by HSM. That is, the observed crashes on divided and undivided multilane highways are at least 1.573 and 1.347 times greater than predicted by HSM. Further, substantial heterogeneity was uncovered by estimating random parameter crash count models for multilane divided highways, with the effects of lane width (feet) and average five years’ AADT (in 1000s) their varying substantially across different road segments. For undivided multilane highways, the average five years’ AADT, and indicators for commercial and residential land use showed substantial heterogeneity. The study underscores the importance of accounting for unobserved heterogeneity and shows that on similar roadways, the factors that account for heterogeneity can be different. The findings can help practitioners and transportation agencies apply more appropriate and highly localized countermeasures to improve safety performance.
Development of Network Screening Safety Performance Functions for Roadway Departure Safety in Virginia
Young-Jun Kweon, National Highway Traffic Safety Administration (NHTSA)Show Abstract View Presentation
In-Kyu Lim, Virginia Department of Transportation
Roadway departure (RD) crash is recognized as one of eight emphasis areas in Virginia’s 2017-2021 Strategic Highway Safety Plan, and the Virginia Department of Transportation (VDOT) has been using annual counts of RD crashes to identify locations for RD safety improvement. However, identifying locations based on crash counts is subject to bias and inaccuracy. The safety performance functions (SPFs) developed and deployed by VDOT for statewide network screening might be used for RD safety, but this could lead to undesirable outcomes in that those SPFs are intended for all crash types and RD safety issues are believed to be different from other crash types. This study was to develop RD SPFs that should be implemented for statewide network screening in Virginia using existing resources. A total of 93 RD SPFs were successfully developed with three functional forms: (1) SPFs with AADT in the logarithmic form, (2) SPFs with AADT in a customized functional form, and (3) SPFs with AADT and other predictors in customized functional forms. The study found that the RD SPFs vary in their functional forms across site types. The logarithmic form of AADT, regarded as a standard for an SPF, is deemed suitable in general for a typical range of AADT. However, that form could be severely deviated from the underlying relationship. Therefore, a proper functional form of AADT for an RD SPF should be determined for each site type and by severity level separately whenever possible.
Research on Characteristics and Prediction Methods of Expressway Accidents
Jiyuan Tan, North China University of TechnologyShow Abstract View Presentation
Qianqian Qiu, North China University of Technology
Shuofeng Wang, Tsinghua University
Na Xie, Central University of Finance and Economics
Yuelong Su, AutoNavi Software Company
Yujing Wang, AutoNavi Software Co
Weiwei Guo, North China University of Technology
Ke Zhang, North China University of Technology
With the construction of expressways and the increase in car ownership, the traffic safety problems of expressways have become increasingly severe. Based on the data of expressway accident data and vehicle mobile navigation trajectory data in a province, this paper firstly uses the statistical analysis method to study the characteristics of traffic accidents. Then it uses Wilcoxon signed rank test to quantitatively analyze the space-time characteristics of the road speed before and after traffic accidents; Then it uses the average speed of the upstream and downstream sections one hour before the traffic accident and five kilometers from the accident point to train the support vector data description (SVDD) model, and the corresponding hypersphere was obtained. Finally, the accuracy and feasibility of the accident prediction method proposed in this paper were verified by experiments. The research results will help the management department to predict expressway traffic safety, implement preventive measures in a targeted manner, and pre-regulate deployments, and issue early warning messages to reduce accident rates and mitigate accident hazards.
Two-Way, Stop-Controlled Intersection Analysis with Zero-Inflated Models
Ming Sun, University of Louisiana, LafayetteShow Abstract View Presentation
Xiaoduan Sun, University of Louisiana at Lafayette
M. Ashifur Rahman, University of Louisiana, Lafayette
Mousumy Akter, University of Louisiana at Lafayette
Subasish Das, Texas A&M Transportation Institute
Intersection safety continues to be a crucial issue throughout the United States. In 2016, 27 percent of the 37,461 traffic fatalities on U.S. roadways occurred at or near intersections. Nearly 70 percent of intersection-related fatalities occured at unsignalized intersections. At such intersections, vehicles stopping or slowing to turn create speed differentials between vehicles traveling in the same direction. This is particularly problematic on two-lane highways. Research was performed to analyze safety performance for intersections on rural, two-lane roadways, with stop control on the minor roadway. Roadway, traffic, and crash data were collected from 4,148 stop-controlled intersections of all 64 Parishes (counties) statewide in Louisiana, for the period of 2013 to 2017. Four count approaches, Poisson, Negative Binomial (NB), Zero-inflated Poisson (ZIP) and Zero-inflated Negative Binomial (ZINB) were used to model the number of intersection crashes for different severity levels. The results indicate thatZIP models provide a better fit than all other models. In addition to traffic volume, greater curve radius of major and minor roads, greater curve lengths of major roads, and greater lane widths of minor roads led to significantly smaller crash occurrences. However, intersections close to the beginning of curves, or located in the middle of curves, higher speed limits of minor roads, and urban areas led to significantly greater crash occurrences. Four-leg stop-controlled (4ST) intersections have 35 percent greater total crashes, 49 percent greater fatal and injury crashes, and 25 percent greater property damage only (PDO) crashes, relative to three-leg (3ST) intersections.
Safety Analysis of Displaced Left Turn Intersections
Ahmed Abdelrahman, University of Central FloridaShow Abstract View Presentation
Mohamed Abdel-Aty, University of Central Florida
Jaeyoung Lee, Central South University
Lishengsa Yue, University of Central Florida
Ma'en Al-Omari, University of Central Florida
Displaced left-turn intersections (DLTs) are designed to enhance the operational performance of conventional intersections that are congested due to heavy left-turn traffic volumes by excluding the left-turn movements at the main intersection. This results in reducing the number of potential conflict points and increasing the intersection capacity. However, since drivers are not familiar with DLTs’ operation, there is a need to assess the safety and operational efficiency of this type of intersections. This paper evaluates the safety performance of DLTs using two common methods, which are a before-and-after study with comparison group and cross-sectional analysis. Furthermore, it investigates the operational performance of DLTs using a general linear model describing the relationship between a selected measure of performance and other operational and geometric characteristics based on high-resolution traffic data. The safety analysis indicates that DLTs can increase the crash frequency in comparison to conventional intersections. In addition, the operational analysis implies that DLTs have a potential to reduce the delay at intersections. The study concludes that DLTs are more dangerous than conventional intersections for many crash types; but it might be more efficient for operational performance. It is recommended that appropriate safety countermeasures should be developed and implemented to enhance traffic safety at DLTs.
Safety Evaluation of High-Friction Surface Treatments
Craig Lyon, Persaud and Lyon Inc./Traffic Injury Research FoundationShow Abstract View Presentation
Bhagwant Persaud, Ryerson University
David Merritt, The Transtec Group, Inc.
Joseph Cheung, Federal Highway Administration (FHWA)
The intent of the study was to provide high quality crash modification factors (CMFs) and benefit/cost (B/C) ratios for high friction surface treatments (HFSTs) and in so doing to recommend where and under what conditions to use them to effectively reduce roadway crashes. The state-of-the-art empirical Bayes before-after methodology was applied to evaluate the effects of HFST treatments on various crash types – total, injury, wet road, run-off-road, and head-on plus opposite direction sideswipe (for curves only) using data obtained from West Virginia (curve sites), Pennsylvania (curve sites), Kentucky (curve and ramp sites), and Arkansas (ramp sites). The results for curve sites indicate substantial and highly significant safety benefits (low CMFs) for each State and the 3 States combined. This is especially so for the primary crash types targeted by HFST programs: run-off-road, wet road, and head-on plus opposite direction side-swipe crashes. The results for ramp sites for the two States were inconsistent, except for wet weather crashes for which the benefits are quite large and highly significant). The benefits for all crashes and injury crashes are substantial for Kentucky while there are negligible effects for these crashes in Arkansas. A disaggregate analysis of the curve sites suggested that there appears to be a logical and consistent relationship between CMFs and three variables: friction improvement, traffic volume, and expected crash frequency before treatment. These variables were used in developing recommended crash modification functions.
Calibration of Wyoming-Specific Safety Performance Functions for Urban and Suburban Five-Lane Arterial Roadway Corridors
Mohamed Ahmed, Federal Highway Administration (FHWA)Show Abstract
Ahmad Elterkawi, University of Wyoming
Irfan Ahmed, University of Wyoming
Urban arterials play a major role in the roadway network system in urban cities providing accessibility and mobility to major zones of cities. According to the National Highway Traffic Safety Administration, 45% (14,414) of the fatal crashes occurred on urban roadways during 2015 in the US. Wyoming experienced 1,630 fatal and injury crashes on urban roads in 2015 which represented 56% of the total fatal and injury crashes. The role of urban arterials in providing accessibility, and mobility, as well as the increased number of fatalities highlight the importance of evaluating their traffic safety. The objective of this study was to calibrate the Highway Safety Manual’s (HSM) Safety Performance Functions (SPFs) for Wyoming conditions. The study was conducted using crash data from 2003 to 2012 for five-lane arterials (5T) with a center two-way left-turn lane, and urban and suburban arterial signalized four-leg intersections (4SG). This study assessed the transferability of the HSM SPFs to Wyoming conditions and compared the calibrated HSM’s SPFs to newly developed Wyoming-specific SPFs for urban and suburban 5T segments, and 4SG intersections. Furthermore, SPFs were developed for urban and suburban corridors, i.e., 5T and 4SG were combined and compared to individual 5T and 4SG SPFs. SPFs were developed using Negative Binomial regression model for Total, Property Damage Only, and Fatal and Injury crashes. The HSM SPFs were found to over-estimate the total crashes for 5T segments and 4SG intersections in Wyoming. Furthermore, the Wyoming-specific SPFs performed better than the calibrated HSM SPFs in predicting total crashes.
Safety Effectiveness of All-Electronic Toll Collection System
Meghna Chakraborty, Michigan State UniversityShow Abstract View Presentation
Steven Stapleton, Michigan State University
Mehrnaz Ghamami, Michigan State University
Timothy Gates, Michigan State University
Tolling has been recognized as one of the effective ways to reduce traffic congestion. There are three types of tolling systems commonly in practice, namely, Traditional Toll Plaza (TTP), Hybrid Toll Plaza (HTP) and All-Electronic-Toll-Collection (AETC) system that includes Open Road Tolling (ORT) and High Occupancy Toll (HOT) lanes. The upgrade from toll plazas to the AETC system has demonstrated operational benefits, but little is known about the safety impacts of these new tolling systems. Currently, Texas ranks second in the nation in terms of the length of toll roads. This study determines the safety impacts of the conversion from the HTP to ORT on Loop 1, Austin, and from HOV (High Occupancy Vehicle) to HOT lanes on I-635, Dallas. As the conversions on Loop 1 took place in January 2013, crash data were analyzed from 2010 to 2015. Crashes on I-635 were studied from 2013 to 2018, excluding 2016, as the conversions occurred in October 2016. Empirical Bayes before-after analysis including the development of Safety Performance Functions (SPFs) was separately carried out for Loop 1 and I-635. The results reveal that an upgrade to the ORT system on Loop 1 significantly reduced total, fatal and injury, and property damage crashes. Similarly, conversion from HOV lanes to HOT lanes on I-635 also resulted in a significant reduction in total, fatal and injury, and property damage crashes, although to a lesser extent than that for Loop 1. This study provides more evidence for the crash reduction potential of AETC systems.
Safety Performance of Autonomous Vehicles on an Urban Arterial in Proximity of a Driveway
Seyedeh Maryam Mousavi, Texas A&M University, College StationShow Abstract View Presentation
Dominique Lord, Texas A&M University, College Station
Seyed Reza Mousavi, Shiraz University
Maryam Shirinzad, Texas A&M University, College Station
Maryam Shirinzad, Texas A&M University, College Station
Urban traffic network has been growing as an integral part of cities. Urban arterials, as the backbone of the urban traffic network, are characterized by closely spaces driveways and carry a high traffic volume per day. The literature consistently reported that there is a positive relationship between driveway density and crash rate. Therefore, managing driveways, which usually work as three-legged unsignalized intersections, located along urban arterials is crucial, especially under high traffic demand, to improve both safety and operation. However, due to the cost and space limitation, conventional methods are impractical and, therefore, new solutions should be implemented. Autonomous Vehicles (AVs), as a multidisciplinary technology, have been the focus as a replacement for human-driven vehicles to improve both traffic safety and operation. In this study, the effect of AVs on the safety of an urban arterial in the proximity of an unsignalized intersection was evaluated. A microsimulation model was used to develop an urban network with an unsignalized access point under various traffic congestion levels for both conventional vehicles and AVs. Afterward, the frequency and distribution of the conflicts for conventional vehicles and AVs were compared. The results indicated that AVs can enhance safety significantly compared to the conventional vehicles in proximity of an access point, especially under congested traffic situations. However, providing an exclusive lane on the arterial for the driveway vehicles to merge to the arterial promotes safety and operation of the network.
Field Implementation of Directional Rumble Strips to Deter Wrong-Way Driving on Freeways
Chennan Xue, Auburn UniversityShow Abstract View Presentation
Huaguo Zhou, Auburn University
Dan Xu, Auburn University
This paper presents the field implementation results of directional rumle strips (DRS), a low-cost traffic control device (TCD), designed to deter wrong-way driving (WWD) on freeways. Southbound off-ramps at Exits 208 and 284 on I-65 in Alabama were selected for implementation because they were ranked as high-risk locations by a network screening tool. Three patterns (D3, C, and E.1) were recommended for field implementation according to the previous test results. Pattern D3 was installed at the off-ramp terminal near the stop bar or yield line. Pattern C was implemented at the segment between the terminal and ramp curve. Pattern E.1 was placed on the tangent part before the ramp curve. WWD incidents and distances before-and-after the implementation were collected using cameras. Field driving tests were conducted to collect sound and vibration data at various speed categories for both right-way (RW) and WW directions. Before-and-after studies evaluated the effectiveness of DRS patterns in deterring WWD incidents. Sound and vibration analysis quantified the differences between RW and WW drivers’ perceptions. Results showed that the number of WWD incidents and average driving distances significantly reduced after implementing all the DRS. The results also confirmed that WW drivers can perceive elevated sound and vibrations when passing the DRS. A general guideline was developed for implementing three different DRSs on freeway off-ramps to deter WWD.
SPF Real-Time Data Versus AADT
MAURICIO BURGOS, Florida International UniversityShow Abstract View Presentation
Cecilia Kadeha, Florida International University
Priyanka Alluri, Florida International University
Albert Gan, Florida International University
Annual Average Daily Traffic (AADT) has been one of the most fundamental exposure variables used in developing Safety Performance Functions (SPFs). However, since AADT is an aggregate measure of traffic conditions, it does not reflect the real-time traffic variations. In other words, while AADT gives the number of vehicles per day averaged over a year, the actual number of vehicles on a roadway varies drastically with season, day of week, and time of day. It is therefore hypothesized that developing SPFs using disaggregate traffic data instead of aggregated AADT could yield better crash predictions. With the increasing availability of exhaustive real-time traffic data, it is crucial to determine if developing SPFs using real-time data improves the crash predictions. This research sets out to use Level of Service (LOS) as a proxy for real-time data in lieu of AADT to develop SPFs for freeways. The paper attempts to answer the following questions: would simple SPFs developed using LOS alone perform better than the simple SPFs developed using AADT? If so, could the crash predictions be improved by considering other influential variables in addition to LOS? The analysis was based on 2016-2018 crash data on I-75 in Florida. The results indicated that the SPFs based on LOS alone predicted the crashes better than the SPFs developed using AADT. Further, the simple SPFs developed using LOS alone and those developed using additional influential variables showed almost similar prediction performance. Therefore, SPFs developed using LOS alone would yield acceptable crash predictions on freeway facilities.
Safety Performance Functions for Fatal Crashes on National Highways Under Heterogeneous Traffic Flow
Hasan Naqvi, National Highways Authority of IndiaShow Abstract
Geetam Tiwari, Indian Institute of Technology, Delhi
The objective of this paper is to explore the effect of the road elements of two-, four- and six-lane National Highways (NHs) under heterogeneous traffic (including pedestrians) flow on fatal crashes. The generalized linear model technique, i.e., negative Binomial (NB) regression is used for analyzing linear and non-linear effect of continuous and categorical predictor variables on discreet dependent variable (fatal crashes) separately for each NH segments. In India, NHs are not usually access controlled, and heterogeneous vehicles travel on highways. The probable explanatory variables are short-listed after thorough literature review, and availability of data. These variables comprise of vehicular traffic, highway elements, and roadside land use. The fatal crash data for the historic period (2009-2013), traffic and highway inventory data have been collected for NHs having varying lane configuration: two-lane NH-8, four-lane NH-24 and six-lane NH-1. The study results revealed negative binomial regression model fit the data statistically, and also identified number of statistically significant variables (‘segment length’, ‘roadside land use’, ‘presence of service road (SR)’ and ‘terrain type’) to estimate fatal crashes at NHs segments. The results of the safety performance functions (SPFs) showed that out of seven explanatory variables examined for each NH (segments), the significant explanatory variable is found to be ‘segment length’ in km for all three models of NHs (segments). Other significant variable is ‘land use’ along NHs for both two-lane NH-8 and four-lane NH-24. Similarly, the explanatory variables ‘presence of SR’ and ‘terrain type’ are found significant for four-lane NH-24 and two-lane NH-8 respectively.
Factors Influencing the Likelihood of Occurrence of a Wrong-Way Driving Crash and Injury Severity
Sarvani Duvvuri, University of North Carolina, CharlotteShow Abstract View Presentation
Srinivas Pulugurtha, University of North Carolina, Charlotte
Head-on and sideswipe collisions are the possible consequences of a driver traveling in the direction opposite to the mainline flow. Higher fatality rates associated with the wrong-way driving (WWD) crashes calls for an investigation of these crashes along with identification of their corresponding contributing factors. Crash data for the years 2012-15 for the State of North Carolina was gathered. The crashes resulting due to driving in the opposite direction were identified for the analysis and modeling. This research is two-fold. The former examines and identifies risk factors contributing to the occurrence of WWD crashes while the latter identifies the factors associated with various levels of injury severity. Binary logistic regression model and partial proportionality odds model were developed to examine and identify the crash risk factors, significant at a 95% confidence interval. The results indicate that driver-related characteristics, crash location characteristics, weather condition, driver impairment (DUI), time of the day, day of the week and the pavement characteristics are significantly associated with the likelihood of the occurrence of a WWD crash. Similarly, driver characteristics, weather condition, variables corresponding to the crash location, and temporal factors were found to have a significant effect on the injury severity of the crash. This research assists by identifying the characteristics of areas/locations prone to wrong-way entries. The results from the models help the agencies proactively plan and reduce the occurrence of WWD crashes and associated injury severity.
The Variability of Urban Safety Performance Functions for Different Road Elements: An Italian Case Study
Paolo Intini, Politecnico di BariShow Abstract
Nicola Berloco, Politecnico di Bari
Gabriele Cavalluzzi, Politecnico di Bari
Pasquale Colonna, Politecnico di Bari
Dominique Lord, Texas A&M University, College Station
Vittorio Ranieri, Polytechnic University of Bari
Safety performance functions are used to predict crash frequencies based on several possible variables which include at least traffic volumes, geometric, and traffic-control variables. In urban environments, safety predictions are usually differentiated for homogeneous road elements: segments and intersections. Further disaggregations are often considered, such as one-way/two-way, one-lane/multilane segments, three/four-legged, signalized/unsignalized intersections. In the context of a National research project, data about crashes, traffic, geometric, traffic-control and additional variables were collected for the road network of the City of Bari, Italy. 320 homogeneous segments and 120 intersections were included in the sample of sites, on which more than 1,500 fatal+injury crashes have occurred in a 5-years period (2012-2016). The study was conceived for research purposes and for being useful for practitioners. The main research questions concerned: a) finding the best possible subsets for segments and intersections for safety modelling purposes, by discussing the related problems, and inquiring into the variability of predictors within subsets; b) comparing the modelling results with existing literature to highlight common trends and/or main differences; c) assessing the importance of additional crash predictors, besides traditional variables. As a result of the study, six detailed models were developed for: one-way/two-way homogeneous segments, three/four-legged, signalized/unsignalized intersections. Crash predictors greatly vary within the different subsets considered. The effect of vertical signs on minor roads/driveways, critical sight distance, cycle crossings, pavement/markings maintenance was specifically discussed. Some common trends but also notable differences in both types and effect of crash predictors were found by comparing results with relevant literature.
Using No-U-Turn Hamiltonian Monte Carlo Bayesian Method to Investigate the Contributing Factors of Crash Injury Severity in Very Low-Volume Rural Roads of Wyoming
Irfan Ahmed, University of WyomingShow Abstract
Mohamed Ahmed, Federal Highway Administration (FHWA)
haun Wulff, University of Wyoming
In Wyoming, the percentage of traffic fatalities on rural roadways have always surpassed traffic fatalities in urban roadways. Crash count models provided in the Highway Safety Manual (HSM), as a function of traffic volume and segment length might not be adequate to relate the various crash contributing factors to different types of crashes and severity, especially for low-volume rural roadways. Hence, crash injury severity models with discrete severity outcomes are modelled with respect to roadway, driver, environmental, and other crash characteristics. A dataset was prepared using crash records from the years 2007-2016 of 28 different two-way two-lane roadways in Wyoming with average annual daily traffic less than 400 vehicles per day. A binary logistic model was developed to carry out the crash injury severity analysis using a Bayesian approach. Fixed- and random-effects models were developed to investigate the relationship between the crash injury severity and its associated contributing factors. Results showed that the random-effects model is a better fit to the data than the fixed-effects model. Parameter estimates are sampled from the posterior distributions using a No-U-Turn Hamiltonian Monte Carlo sampling technique which is a more efficient method than other Markov chain Monte Carlo methods. The population-averaged estimates included driver impairment, improper use of restraint, speeding, lane departure, and motorcycle involvement and were found to increase the odds of a fatal/injury crash. Furthermore, the combined effect of nightly crashes and improper driving action leads to increased likelihood of fatal/injury crash.
Development of Conflict Severity Index for Safety Evaluation of Right Turn Related Crashes at Unsignalized Intersections on Intercity Highways
Madhumita Paul, Indian Institute of Technology, RoorkeeShow Abstract
Soumyodeep Chatterjee, Indian Institute of Engineering Science and Technology
Indrajit Ghosh, Indian Institute of Technology, Roorkee
Unsignalized intersections in developing countries experience severe crashes due to prevalence of violating driving behavior. Most of the existing traffic conflict indicators do not sufficiently explain the consequences of probable collisions in terms of severity at such traffic facility. This study proposes a conflict severity index (CSI) which combines Post Encroachment Time (PET) and expected loss in kinetic energy after collisions (∆KE). Right turn related collisions (left-hand driving) and related crossing conflicts between right turners and through moving vehicles are observed from 8 unsignalized intersections located on intercity highways. Once critical conflicts are identified, expected ∆KE after collision is calculated utilizing Delta-V, mass ratio and conflict angle. It is observed that by decreasing PET or increasing ∆KE, CSI value increases. Sensitivity analysis results show that for a specific PET value, CSI increases with increasing speed of right of way vehicle and conflict angle. ∆KE and resulting CSI for the vehicle combination increases when mass ratio between involved vehicles increase. However, for same mass ratio, KE and resulting CSI are found to be higher for the vehicle combination in which individual mass is higher. Later, CSI values are clustered in High Severe, Mild Severe and Non-Severe Groups based on severity. Finally, the proposed index is validated by developing a relationship between CSI values observed for severe conflicts and number of serious crashes. This index will be very useful in making an accurate decision for safety improvement at problematic locations along with reducing the number of crash victims.
Effects of Traffic Signal Control Parameters on Vehicular Crash Frequency at Four-Leg Signalized Intersection Approaches
Dusan Jolovic, Garver LLCShow Abstract View Presentation
Milos Pljakic, Univerzitet u Pristini
This paper provides an insight into the correlation between the signal timing parameters and the crash frequency at the 4-leg signalized intersection approaches. Crash data were obtained from 2003 to 2013 for Fort Lauderdale, FL. The authors investigated 162 intersections. Crash frequency was modeled using Poisson model, Negative Binomial model, Poisson Zero Inflated, and Negative Binomial Zero Inflated models. Multicollinearity was tested using Poisson correlation coefficient. The matrix of multicollinearity of all the observed independent variables in the models was developed and there was no high correlation reported. Model comparison was conducted to choose the best performing model based on the Aikake Information Criterion (AIC) and Bayesian Information Criterion (BIC). The results showed that the Negative Binomial model performed best when compared to other models. The results were obtained for the total number of crashes, rear-ends crashes only, and the lane-change crashes only. The results were consistent across the sets. The AADT, cycle length, and the number of phases within a cycle were statistically significant across all tested options.
Effect of Intermittent Shoulder Rumble Strips Application Strategy on the Development of Crash Modification Factors
Md Julfiker Hossain, University of ConnecticutShow Abstract
Mohamed Ahmed, Federal Highway Administration (FHWA)
Lane departure crashes (LDC) are the most common type of crashes in Wyoming, contributing to 72% of all fatal and incapacitating injury crashes. In the US, single vehicle run off the road crashes contribute to a total societal loss of $80 billion annually. Shoulder Rumble Strips (SRS) have been proven to have a significant effect on reducing LDC. The process of quantifying safety effectiveness of SRS becomes challenging when resurfacing and shoulder rehabilitation may result in an intermittent presence of SRS in some locations as opposed to the assumption of continuous presence.
Estimating Baseline Numbers for Safety Measure Target Setting in Virginia
Scott Himes, VHBShow Abstract View Presentation
Vikash Gayah, Pennsylvania State University
Jeffrey Gooch, VHB
Stephen Read, Virginia Department of Transportation
The Federal Highway Administration established the Safety Performance Management program (Safety PM) to support the Highway Safety Improvement Program. The Safety PM Final Rule requires State Departments of Transportation (DOTs) to establish and report safety targets annually. The FHWA does not identify a specific methodology to use when establishing safety targets. Many State DOTs apply annual growth/decline factors to previous year safety measures. However, State DOTs also have flexibility to use a data-driven process. The Virginia Department of Transportation (VDOT) recently pursued the development of a more robust data-driven safety target setting methodology. This paper presents a methodology for establishing safety target baselines for several measures, including: 1) fatalities; 2) serious injuries; and, 3) non-motorized fatalities and serious injuries. Predictive models were developed for establishing a baseline for 2019 targets and were further refined for 2020 targets. The predictive models include macro-level inputs and were developed for monthly, VDOT district-level outcomes. Performance measure data from 2018 were withheld from models for validation purposes and 2018 through 2020 model inputs were forecasted based on recent data. As 2019 data become available, the models should incorporate newer data and new models should be developed for revised 2020 and beyond predictions, as necessary. Refined models should include additional data elements as predictors, include more years of data to increase sample size, and capture moments when unobserved annual factors (i.e., unobserved underlying macro-level trends) begin to change.
Safety Performance Functions for Rural Arterial Roads in Egypt
Sania Elagamy, Mansoura UniversityShow Abstract View Presentation
Usama Shahdah, Mansoura University
Sherif El-Badawy, Mansoura University
Sayed Shwaly, Mansoura University
Zaki Zidan, Mansoura University
This paper presents the development of safety performance functions (SPFs) for total, fatal, injury, and damage only crashes for five rural arterial roads in Egypt using crash data between 2008 and 2011. Four segmentation methods were used for the SPFs development: (1) fixed section length of one kilometer ( S1 ); (2) homogenous sections ( S2 ); (3) variable sections with respect to presence of curvatures ( S3 ); and (4) variable sections with respect to the presence of both curvatures and U-turns ( S4 ). The generalized linear modeling (GLM) technique was used for SPFs development using stepwise procedure, with/without considering time-effect (year-to-year variation). The Akaike information criterion (AIC) along with the cumulative residual (CURE) plots were used to evaluate the prediction accuracy of the proposed models. The segmentation method was found to affect the prediction accuracy of the model. For all crash types, the developed SPFs using the segmentation method S1 and S3 , with multiple variable crash model form, were found to produce the most accurate predictions compared to other SPFs using other segmentation methods. In addition, each road has its own crash pattern, as the results show that the coefficients for the roads are statistically significant for all the developed models. The results also showed that by increasing shoulder, lane, and median widths, the probability of crashes is likely to decrease. Finally, the presences of either horizontal curves and/or U-turns are most likely to reduce the probability of crash occurrence.
Highway Safety Manual Calibration: Variance of the Calibration Factor and the Sample Size
Mahdi Rajabi, JacobsShow Abstract View Presentation
Patrick Gerard, Clemson University
Jennifer Ogle, Clemson University
Crash frequency has been identified by many experts as one of the most important safety measures, and the Highway Safety Manual (HSM) encompasses the most commonly accepted predictive models to predict the crash frequency for specific road segments and intersections. The HSM recommends that the models should be calibrated using data from a jurisdiction where the models will be applied. One of the most common start-up issues with the calibration process is how to estimate the required sample size to achieve a specific level of precision, which can be a function of the variance of the calibration factor. The published research has indicated great variance in sample size requirements, and some of the sample size requirements are so large that they may deter state departments of transportation from conducting calibration studies. In this study, an equation is derived to estimate the sample size based on the coefficient of variation of the calibration factor and the coefficient of variation of the observed crashes. This equation is verified using a regression analysis on a dataset from two recent calibration studies, South Carolina and North Carolina. Whereas, the minimum sample size requirement published in the HSM is based on the summation of the observed crashes, this paper demonstrates that the summation of the observed crashes may result in calibration factors that are less likely to be equally precise and the coefficient of the variation of the observed crashes can be considered instead.
Before-and-After Empirical Bayes Evaluation of Citywide Installation of Driver Feedback Signs
Mingjian Wu, University of AlbertaShow Abstract View Presentation
Karim El-Basyouny, University of Alberta
Tae J. Kwon, University of Alberta
Speeding is a leading factor that contributes to approximately one-third of all fatal collisions. Over the past decades, various passive/active countermeasures have been adopted to improve drivers’ compliance to posted speed limits in order to improve traffic safety. The Driver Feedback Sign (DFS) is considered a low-cost innovative intervention that is being widely used, in growing numbers, in urban cities to provide positive guidance for motorists. Despite their documented effectiveness in reducing speeds, limited literature exists on its impact on reducing collisions. This study addresses this gap by designing a before-and-after study using the Empirical Bayes method for a large sample of urban road segments. Safety performance functions and yearly calibration factors are developed to quantify the sole effectiveness of DFS using large-scale spatial data and a set of reference road segments within the city of Edmonton, Alberta, Canada. Likewise, the study followed a detailed economic analysis based on three collision costing criteria to investigate if DFS was indeed a cost-effective intervention. The results showed significant collision reductions that ranged from 32.5% to 44.9%, with the highest reductions observed for severe speed-related collisions. The results further attested that the benefit-cost ratios, combining severe and Property-Damage-Only collisions, ranged from 8.2 to 20.2 indicating that the DFS can be an extremely economical countermeasure. The findings from this study can provide transportation agencies in need of implementing cost-efficient countermeasures with a tool they need to design a long-term strategic deployment plan to ensure the safety of travelling public.
Probe-Speed Based Safety Performance Metrics in Georgia: A Case Study
David Ederer, Georgia Institute of Technology (Georgia Tech)Show Abstract View Presentation
Michael Rodgers, Georgia Institute of Technology (Georgia Tech)
Michael Hunter, Georgia Institute of Technology (Georgia Tech)
Kari Watkins, Georgia Institute of Technology (Georgia Tech)
Speed is a primary risk factor for road crashes and injuries. Previous research has attempted to ascertain the relationship between individual vehicle speeds, aggregated speeds, and crash frequency on roadways. Although there is a large body of research linking vehicle speeds to safety outcomes, there is not a widely applied performance metric for safety based on regularly reported speeds. With the increasingly widespread availability of probe vehicle speed data, there is an opportunity to develop network level safety performance metrics. This analysis examines the relationship between percentile speeds and crashes on a large arterial in Metropolitan Atlanta. This study uses data from the National Performance Metric Research Database, the Georgia Electronic Accident Reporting System, and the Highway Performance Monitoring System. Negative binomial regression models were used to analyze the relationship between speed percentiles, and speed differences to crash frequency on roadway sections. Results suggest that differences in speed percentiles, a measure of speed dispersion, are related to the frequency of crashes. Based on the models, the difference in the 85th percentile and median speed is proposed as a performance metric. This difference is easily measured using NPMRDS probe vehicle speeds, and provides a practical performance metric for assessing safety on roadways.
Analysis, Modeling, and Simulation Framework for Safety Performance Assessment of the Wyoming Connected Vehicle Pilot Deployment Program
Guangchuan Yang, Institute for Transportation Research and Education (ITRE)Show Abstract
Mohamed Ahmed, Federal Highway Administration (FHWA)
Eric Adomah, University of Wyoming
The 402-mile of Interstate 80 in Wyoming was selected by the U.S. Department of Transportation to develop, test, and deploy a suite of Connected Vehicle (CV) applications (WYDOT CV Pilot). It is expected that after full deployment of the technology, the pilot will improve safety and mobility under adverse weather conditions by creating new ways to communicate road and travel information to both drivers and fleet managers. In this regard, this research developed an Analysis, Modeling, and Simulation framework to assess the safety performance of the WYDOT CV Pilot. A 23-mile representative I-80 corridor was selected for developing the microsimulation models. Traffic flow and driving behavior data under winter snowy weather condition were collected to calibrate the baseline microsimulation model. A driving simulator experiment was conducted to quantitatively investigate the impacts of CV technology on driving behavior; accordingly, the driving behavior data under CV environment were employed to properly update the calibrated CV microsimulation models. The safety effectiveness of the WYDOT CV Pilot were assessed for various demand levels and CV penetration rates. It was concluded that WYDOT CV applications increased drivers’ situational awareness under adverse weather conditions, and thus, have the potential to reduce crash risks. The reductions in conflicts displayed a decreasing trend with the increase of CV penetration rates, although the reduction was not significant when CV penetration was lower than 10 percent. The maximum reduction in conflicts was 85 percent, when all trucks were equipped with CV technology.
Single-Vehicle Roadway Departure Crashes on Rural Curved Segments: Analysis of Injury Severity Using Random Parameters with Heterogeneity in Means and Variances
Mouyid Islam, University of South FloridaShow Abstract View Presentation
Anurag Pande, California Polytechnic State University, San Luis Obispo
Roadway departure crashes are one of the core emphasis areas in Strategic Highway Safety Plans (SHSP). These crashes, especially on rural roads, lead to a disproportionately higher number of fatalities and serious injuries. The focus of this study to identify and quantify the factors affecting injury severity outcomes for single-vehicle roadway departure (SV-RwD) crashes on rural curved segments in Minnesota. The crash data is extracted from the Highway Safety Information System (HSIS) from 2010 to 2014. This study applied a mixed logit with heterogeneity in means and variances approach to model driver injury severity. The approach accounts for possible unobserved heterogeneity in the data resulting from driver, roadway, traffic, and/or environmental conditions. This analysis adds value to the growing body of literature because it uncovers some unobserved heterogeneity in the form the attributes specific to driver injury severities in contrast with the standard mixed logit approach. The model results indicate that there is a complex interaction of driver characteristics and actions (male drivers, age below 30 years, unsafe speed and distraction), roadway and traffic characteristics (two-lane undivided road, county roadways, and low traffic volume), environmental conditions (adverse weather, cloudy weather, dark condition and dry surface condition), crash event (rollover), and vehicle characteristics (vehicle type – sport utility vehicle). The results also provide some evidence of the effectiveness of a highway curve safety improvement program implemented in one of the Minnesota Department of Transportation (MnDOT) districts.
Movement-Based Safety Performance Functions for Signalized Intersections
Taehun Lee, North Carolina State UniversityShow Abstract View Presentation
Thomas Chase, Institute for Transportation Research and Education
Christopher Cunningham, North Carolina State University
Shannon Warchol, Institute for Transportation Research and Education
The traditional safety performance functions (SPFs) for signalized intersections predict the total crash frequency using annual average daily traffic (AADT). By its nature, the traditional SPF has limitations in evaluating the safety of individual movements. To improve on this limitation, this study proposes a novel crash prediction method using movement-based SPFs. Movement-based SPFs consist of two models: the conflict point (CP) SPF that predicts the crashes for a CP using the CP types (crossing, merging and diverging) and conflicting movement volumes; and the non-conflict point (NCP) SPF that predicts the NCP crashes using AADTs at intersection-level. The intersection-level SPF with the traditional SPF model form is developed as a reference model. The 1,689 crashes observed from 2010 to 2017 at 21 conventional signalized intersections were used for model estimation. Three crash frequency models for severities (TOT, FI, and PDO) were estimated for movement-based SPFs and intersection-level SPF using the negative binomial regression model. The results showed most of the estimated coefficients in movement-based SPFs are statistically significant at 95% confidence level. Also, it showed the crossing CP has obviously higher crash risk than merging and diverging CPs. The model validation was conducted by CURE plots and model performance measures. The results showed movement-based SPFs have unbiased prediction for the entire range of exposure variables and outperform the intersection-level SPF. The applicability of movement-based SPFs can be extended for the quantitative safety evaluation for new or highly unrepresentative intersection designs such as alternative intersections or interchanges.
Comparison of Crash Modification Factors for Engineering Treatments Estimated by Before-After Empirical Bayes and Propensity Score Matching Methods
Bo Lan, UNC Highway Safety Research CenterShow Abstract View Presentation
Cross-sectional and the empirical Bayes (EB) before-after are two of the most common methods for estimating crash modification factors (CMFs). The EB before-after method has now been accepted as one way of addressing the potential bias due to regression to the mean (RTM) problem. However, sometimes before-after methods may not feasible due to the lack of data from before and after periods. In those cases, researchers rely on cross-sectional studies to develop CMFs. One of the primary challenges of cross-sectional studies is the issue of confounding. Propensity score (PS) matching methods along with cross-sectional regression models is one of the methods that can be used to address confounding. Though the PS methods are widely used in epidemiology and other studies, there are only a few studies using the PS matching methods in CMF derivations in transportation safety. The intent of this study is to evaluate and compare the performance of cross-sectional regression models with PS matching methods with the results from the EB and traditional cross-sectional methods. These methods were evaluated and compared with the traditional cross-sectional using two carefully selected simulated datasets. The results indicated that a particular type of PS matching method (optimal propensity score distance (PSD) matching with maximum variable ratio of 5) performed quite well compared to the EB and the traditional cross-sectional methods. Keywords: empirical Bayes (EB) before-after, cross-sectional, propensity score