GEORGIA DOT RESEARCH PROJECT 19-14 Final Report EVALUATION OF GUARDRAIL PERFORMANCE IN HIGH-RISK ACCIDENT ZONES ON GEORGIA ROADWAYS AND IDENTIFICATION OF ALTERNATIVE BARRIERS Office of Performance-based Management and Research 600 West Peachtree Street NW | Atlanta, GA 30308 November 2021 TECHNICAL REPORT DOCUMENTATION PAGE 1. Report No.: 2. Government Accession No.: 3. Recipient's Catalog No.: FHWA-GA-21-1914 N/A N/A 4. Title and Subtitle: 5. Report Date: Evaluation of Guardrail Performance in High-Risk Accident Zones on November 2021 Georgia Roadways and Identification of Alternative Barriers 6. Performing Organization Code: N/A 7. Authors: 8. Performing Organization Report No.: Xiaoming Yang (PI), Ph.D., P.E. (https://orcid.org/0000-0002-6177-3491); 19-14 Russel Krenek (co-PI), Ph.D. (https://orcid.org/0000-0003-1119-0041); David Scott (co-PI), Ph.D. (https://orcid.org/0000-0002-0228-7768) 9. Performing Organization Name and Address: 10. Work Unit No.: Georgia Southern University N/A PO Box 8077 11. Contract or Grant No.: Statesboro, GA 30460 PI#0016867 Phone: (912) 478-1894 Email: xyang@georgiasouthern.edu 12. Sponsoring Agency Name and Address: 13. Type of Report and Period Covered: Georgia Department of Transportation (SPR) Final Report (August 2019 Office of Performance-based November 2021) Management and Research 14. Sponsoring Agency Code: 600 West Peachtree St. NW N/A Atlanta, GA 30308 15. Supplementary Notes: Prepared in cooperation with the U.S. Department of Transportation, Federal Highway Administration. 16. Abstract: W-beam guardrail systems are the predominant roadside safety barrier used on Georgia highways. These systems are usually installed in accordance with guidelines for the Midwest Guardrail System and generally perform very well across the state. However, in certain areas of high traffic volume in Georgia, repetitive accident locations may benefit from the installation of alternative systems rather than the traditional guardrail system. The objective of this research project is to identify representative high-accident-rate (or "high-risk") zones in Georgia and evaluate the type and effectiveness of the barrier system deployed in these areas. Twenty-eight (28) freeway and freeway-ramp sections with frequent roadside-barrier collisions were identified in Georgia. Road design, traffic, and crash records pertaining to the selected road sections were collected. Based on the collected information, a barrier crash-frequency model and a crash-severity model was developed through statistical regression. The regression models were used in the benefit cost analysis to determine whether a concrete barrier is a more economical alternative to the guardrail. Simple decisionmaking tool were developed for selecting cost-effective roadside barrier types. 17. Keywords: Roadside Barrier, Safety, BenefitCost Analysis, Concrete Barrier, Guardrail 18. Distribution Statement: No Restriction 19. Security Classification (of this report): Unclassified Form DOT 1700.7 (8-72) 20. Security Classification (of this page): Unclassified 21. No. of Pages: 22. Price: 55 Free Reproduction of completed page authorized GDOT Research Project No. 19-14 Final Report EVALUATION OF GUARDRAIL PERFORMANCE IN HIGH-RISK ACCIDENT ZONES ON GEORGIA ROADWAYS AND IDENTIFICATION OF ALTERNATIVE BARRIERS By Xiaoming Yang Assistant Professor Russel Krenek Assistant Professor David Scott Professor Georgia Southern University Research Foundation, Inc. Contract with Georgia Department of Transportation In cooperation with U.S. Department of Transportation Federal Highway Administration November 2021 The contents of this report reflect the views of the authors, who are responsible for the facts and accuracy of the data presented herein. The contents do not necessarily reflect the official views or policies of the Georgia Department of Transportation or the Federal Highway Administration. This report does not constitute a standard, specification, or regulation. ii Symbol in ft yd mi in2 ft2 yd2 ac mi2 fl oz gal ft3 yd3 oz lb T oF fc fl lbf lbf/in2 SI* (MODERN METRIC) CONVERSION FACTORS APPROXIMATE CONVERSIONS TO SI UNITS When You Know Multiply By To Find LENGTH inches 25.4 millimeters feet 0.305 meters yards 0.914 meters miles 1.61 kilometers AREA square inches 645.2 square millimeters square feet 0.093 square meters square yard 0.836 square meters acres 0.405 hectares square miles 2.59 square kilometers VOLUME fluid ounces 29.57 milliliters gallons 3.785 liters cubic feet 0.028 cubic meters cubic yards 0.765 cubic meters NOTE: volumes greater than 1000 L shall be shown in m3 MASS ounces 28.35 grams pounds 0.454 kilograms short tons (2000 lb) 0.907 megagrams (or "metric ton") TEMPERATURE (exact degrees) Fahrenheit 5 (F-32)/9 Celsius or (F-32)/1.8 ILLUMINATION foot-candles foot-Lamberts 10.76 3.426 lux candela/m2 FORCE and PRESSURE or STRESS poundforce 4.45 newtons poundforce per square inch 6.89 kilopascals Symbol mm m m km mm2 m2 m2 ha km2 mL L m3 m3 g kg Mg (or "t") oC lx cd/m2 N kPa Symbol mm m m km mm2 m2 m2 ha km2 mL L m3 m3 g kg Mg (or "t") oC lx cd/m2 N kPa APPROXIMATE CONVERSIONS FROM SI UNITS When You Know Multiply By To Find LENGTH millimeters 0.039 inches meters 3.28 feet meters 1.09 yards kilometers 0.621 miles AREA square millimeters 0.0016 square inches square meters 10.764 square feet square meters 1.195 square yards hectares 2.47 acres square kilometers 0.386 square miles VOLUME milliliters 0.034 fluid ounces liters 0.264 gallons cubic meters 35.314 cubic feet cubic meters 1.307 cubic yards MASS grams 0.035 ounces kilograms 2.202 pounds megagrams (or "metric ton") 1.103 short tons (2000 lb) TEMPERATURE (exact degrees) Celsius 1.8C+32 Fahrenheit ILLUMINATION lux candela/m2 0.0929 0.2919 foot-candles foot-Lamberts FORCE and PRESSURE or STRESS newtons 0.225 poundforce kilopascals 0.145 poundforce per square inch Symbol in ft yd mi in2 ft2 yd2 ac mi2 fl oz gal ft3 yd3 oz lb T oF fc fl lbf lbf/in2 **SISisI tihsetshyemsbyoml fborotlhfeoIrnttehrenaItniotenarnl SatyisotnemaloSfyUsntietsm. AopfpUronpirtisa.teArpoupnrdoipnrgiastheouroldubnedminagdeshtooucoldmbpley mwitahdSeetcotiocnom4 opflyASwTiMthES3e8c0t.ion 4 of (Revised March 2003) ASTM E380. (Revised March 2003) iii TABLE OF CONTENTS LIST OF FIGURES ........................................................................................................ VI LIST OF TABLES ........................................................................................................ VII EXECUTIVE SUMMARY .............................................................................................. 1 CONCLUSIONS........................................................................................................... 1 RECOMMENDATIONS ............................................................................................. 2 CHAPTER 1. INTRODUCTION .................................................................................... 4 BACKGROUND........................................................................................................... 4 OBJECTIVE ................................................................................................................. 5 SCOPE........................................................................................................................... 5 LITERATURE REVIEW ............................................................................................ 5 CHAPTER 2. HIGH-RISK ROADSIDE BARRIER SECTIONS IN GEORGIA ..... 9 CRASH REPORT DATA ............................................................................................ 9 MAINTANANCE DATA........................................................................................... 12 HIGH-RISK BARRIER SECTIONS ....................................................................... 12 ROAD AND TRAFFIC DATA COLLECTION ..................................................... 14 CHAPTER 3. FREQUENCY AND SEVERITY OF ROADSIDE-BARRIER COLLISIONS.................................................................................................................. 16 UNREPORTED CRASHES ...................................................................................... 16 Estimation Based on Maintenance Record.......................................................... 16 Estimation Based on Crash Severity .................................................................... 20 CRASH FREQUENCY AND CRASH RATE ......................................................... 22 CRASH SEVERITY................................................................................................... 27 CHAPTER 4. BENEFITCOST ANALYSIS .............................................................. 31 INTRODUCTION TO BENEFITCOST ANALYSIS .......................................... 31 BENEFITCOST ANALYSIS FOR HIGH-RISK SECTIONS............................. 34 BENEFITCOST ANALYSIS FOR ROAD SECTIONS WITH KNOWN CRASH FREQUENCY.............................................................................................. 35 BENEFITCOST ANALYSIS FOR ROAD SECTIONS WITH UNKNOWN CRASH FREQUENCY.............................................................................................. 39 iv CHAPTER 5. CONCLUSIONS AND RECOMMENDATIONS ............................... 41 SUMMARY................................................................................................................. 41 CONCLUSIONS......................................................................................................... 41 RECOMMENDATIONS ........................................................................................... 42 APPENDIX A. SELECTED ROADWAY SECTIONS DATA .................................. 44 ACKNOWLEDGMENTS .............................................................................................. 46 REFERENCES................................................................................................................ 47 v LIST OF FIGURES Figure 1. Photos. Standard roadside barriers deployed in Georgia in addition to the W-beam system................................................................................................................... 4 Figure 2. Map. Vehicle collisions with roadside barriers 20172020. ............................. 10 Figure 3. Chart. KABCO severity of single-vehicle barrier-collision crashes in Georgia.............................................................................................................................. 11 Figure 4. Chart. KABCO severity of single-vehicle barrier-collision crashes at different posted speeds...................................................................................................... 11 Figure 5. Map. Selected high-risk barrier sections. .......................................................... 13 Figure 6. Map. Crash and maintenance records at the Exit 32 ramp on I-285EB. ........... 17 Figure 7. Map. Crash and maintenance records on I-85SB near Exit 77.......................... 18 Figure 8. Map. Crash and maintenance records at the Exit 237 ramp on I-75SB. ........... 19 Figure 9. Graph. Unreported crashes at different speeds.................................................. 22 Figure 10. Graph. RSAP predicted vs. actual number of crashes. .................................... 25 Figure 11. Chart. Crash frequency vs. degree of curve. ................................................... 26 Figure 12. Graph. Crash severity in high-risk guardrail sections. .................................... 28 Figure 13. Graph. Crash severity in high-risk concrete-barrier sections. ......................... 29 Figure 14 Graph. Expected comprehensive crash cost per accident at different speeds .. 33 Figure 15. Graph. Barrier Selection Chart for Sections without Crash Record................ 40 vi LIST OF TABLES Table 1. General information of the selected high-risk barrier sections........................... 14 Table 2. Estimated unreported crashes for collisions on guardrails. ................................ 21 Table 3. Estimated unreported crashes for collisions on concrete barriers....................... 21 Table 4. Frequency and Rate Single Vehicle Barrier Collisions of High-Risk Sections ............................................................................................................................. 24 Table 5. ANOVA on crash frequency............................................................................... 26 Table 6. Ordinal logistic regression model for crash severity. ......................................... 30 Table 7. Unit life-cycle cost for BCA. .............................................................................. 32 Table 8. Weighted comprehensive crash cost. .................................................................. 32 Table 9. BCA for the selected guardrail sections. ............................................................ 35 Table 10. BCA Example 1 freeway segment................................................................. 36 Table 11. BCA Example 2 freeway ramp. ..................................................................... 37 Table 12. BCR result with known crash frequency. ......................................................... 38 Table 13. Recommended barrier type for sections with known crash frequency. ............ 38 Table 14. BCR result with estimated crash frequency...................................................... 40 Table 15. Road and traffic characteristics......................................................................... 44 Table 16. Crashes and severity. ........................................................................................ 45 vii EXECUTIVE SUMMARY The objective of this research project is to identify representative high-accident-rate (or "high-risk") zones of roadside barrier collisions in Georgia and to evaluate the types and the effectiveness of the barrier systems deployed in these areas. Based on this study, alternative barrier approaches will be considered for recommendation to reduce potential injury, crash severity, and repair costs in these high-risk zones. Twenty-eight (28) freeway sections with frequent roadside-barrier collisions were identified in Georgia. Road design, traffic, and crash records pertaining to the selected road sections have been collected. Based on the collected information, a barrier crash-frequency model and a crash-severity model were developed through statistical regression. The regression models were used in the benefitcost analysis to determine whether a concrete barrier is a more economical alternative to the guardrail. A simple chart was developed as a quick decision-making tool for future roadway design projects. CONCLUSIONS The following conclusions can be drawn from this research: 1. The frequency of roadside-barrier collisions is mostly affected by the traffic volume and the degree of horizontal curve of the road. 2. The existence of unreported crashes poses a challenge to the barrier safety research. It leads to a mismatch between the crash data and the maintenance record. The percentage of unreported crashes reduces nonlinearly with the posted speed of the 1 road. Concrete barrier shows a slightly lower percentage of unreported crashes than guardrail. 3. The severity of barrier collisions is mostly affected by the posted speed of the road and the barrier type. The crash severity in general increases exponentially with the posted speed of the road. When the posted speed of the road is less than 55 mph, concrete barriers produce more severe crashes than guardrail due to the rigidity of the barrier. When the posted speed of the road is more than 55 mph, guardrails produce more fatal and severe (K/A) crashes due to the increased odds of penetrating/ vaulting. 4. The BCA result showed that concrete barrier is more cost-effective for road sections with a posted speed of 55mph or higher. Guardrail barrier is more costeffective for road sections with a posted speed of 35 mph or less for the range of of crash frequencies analyzed. For roads with a posted speed of 40 to 50 mph, concrete barrier should be considered for higher-risk road sections. 5. When the past crash record is unavailable, the cost-effective barrier type can also be determined based on an estimated barrier collision rate. In general, conditions that favor a concrete barrier over guardrail are straight (or slightly curved) road sections or sections with a one-way AADT of 40,000 or more. RECOMMENDATIONS 1. The Roadside Safety Analysis Program (RSAP) developed from the National Cooperative Highway Research Program (NCHRP) Project 22-27 is the most sophisticated tool available for benefitcost analysis of roadside safety features. 2 However, the RSAP baseline model often underestimates the actual crash frequency of "high-risk" barrier sections, especially for sharp horizontal curves (e.g., ramps). Although the program allows use of a modification factor to boost the predicted crash frequency, it still requires a judgment of the total number of crashes (including the unreported crashes) of the road. It is recommended that the Georgia Department of Transportation (GDOT) wait for the next updated version of the program. 2. The roadside barrier selection tools developed in this research can be used as a quick decision-making tool for barrier-upgrade projects. When a more accurate benefitcost analysis is needed, the barrier crash-frequency model and the crashseverity model developed in this research can be applied in a spreadsheet or other computation program. Guardrail maintenance record should be used when available to determine the past average crash frequency of the road. When crash records are used, unreported crashes must be considered using the method presented in this study. 3. It is recommended GDOT build an inventory database for roadside barriers and use an asset management system for tracking the history and condition of roadside barriers. 3 CHAPTER 1. INTRODUCTION BACKGROUND Roadside barriers play an important role in highway traffic safety. The purpose for installing roadside barriers is to redirect and protect off-road vehicles from more harmful obstacles behind the barrier, such as a steep slope, a river, trees, or the opposing direction of traffic. W-beam guardrail systems are the predominant roadside safety barrier used on Georgia highways; other roadside barrier types in Georgia include thrie-beam (T-beam), concrete barrier, and cable barrier (figure 1). These systems are usually installed in accordance with guidelines for the Midwest Guardrail System (MGS)[1] and generally perform very well across the state. However, in certain areas of high traffic volume in Georgia, repetitive accident locations may benefit from the installation of alternative systems, rather than the traditional guardrail system. The most common alternative barrier system in Georgia is the single-slope concrete barrier system. Figure 1. Photos. Standard roadside barriers deployed in Georgia in addition to the W-beam system. 4 OBJECTIVE The objective of this research project is to identify representative high-accident-rate (or "high-risk") zones in Georgia and evaluate the type and effectiveness of the barrier system currently deployed in these areas. Based on this analysis, alternative barrier approaches will be considered for recommendation to reduce potential injury, crash severity, and repair costs in these high-risk accident zones. SCOPE This research focuses mainly on selection between two types of roadside barrier systems: the W-beam guardrail and the concrete barrier. The research goal is to: (1) determine whether concrete barrier is a more cost-effective alternative than guardrail on some of the high-risk sections in Georgia, and (2) develop simple selection criteria that are applicable to other high-risk road sections. The road and traffic characteristics considered in this research include roadway alignment, cross-section geometry, and traffic volume. Although the frequency and severity of barrier collisions can be affected by many other factors, such as weather and pavement surface conditions, these factors are not considered in this research. Also, all high-risk sections studied in this project are from either a simple freeway section or a freeway ramp; therefore, the research findings are not applicable to undivided highways or intersections. LITERATURE REVIEW The selection of barrier system is a comprehensive judgment based on many factors. The American Association of State Highway and Transportation Officials (AASHTO) 5 Roadside Design Guide (RDG)[2] provided a list of eight selection criteria for roadside barriers. These selection criteria include performance capacity, deflection, site conditions, compatibility, cost, maintenance, aesthetics, and field experience. The RDG also suggested that, at high-volume and high-collision-frequency sections, the repair cost may become the overriding consideration. A benefitcost analysis (BCA) can be performed to compare the life-cycle cost and benefit of different barrier systems. A general guideline is provided by the U.S. Federal Highway Administration (FHWA) Highway Safety BenefitCost Analysis Guide.[3] Elvik used meta-analysis to summarize evidence from 32 previous studies that evaluated the safety effects of median barriers, guardrails along the edge of the road, and crash cushions (i.e., impact attenuators).[4] Two hundred and thirty-two (232) estimates of safety effects were included in the meta-analysis. Based in part on this work, the Virginia DOT (VDOT) developed a risk-based management tool to use as a decision aid for allocation decisions for roadside safety hardware.[5] The decision tool comprises three parts: database, screening, and evaluation. A similar risk management study was performed by the Indiana Department of Transportation (INDOT).[6] The INDOT study examined the use of roadside guardrails on state roadways, conducted field visits to fatal crash sites, analyzed two-year crash data, and investigated the characteristics of crashes and main contribution factors. In addition, the study developed the probabilities for crash predictions and identified the costs associated with guardrail crash repairs and maintenance for guardrail benefitcost analysis. Fewer studies, however, specifically investigate the relative effectiveness of highway safety hardware alternatives, particularly in high-crash zones. Zou et al. investigated the safety performance of road barriers in Indiana in reducing the risk of injury.[7] The authors 6 compared the risk of injury among different hazardous events faced by an occupant in a single-vehicle crash. The hazardous events included rolling over, striking three types of barriers (i.e., guardrails, concrete barrier walls, and cable barriers) with different barrier offsets to the edge of the traveled way, and striking various roadside objects. A total of 2,124 single-vehicle crashes (3,257 occupants) that occurred between the years 2008 and 2012 on 517 pair-matched homogeneous barrier and nonbarrier segments were analyzed. A binary logistic regression model with mixed effects was estimated for vehicle occupants. The modeling results revealed that hitting a barrier was associated with lower risk of injury than a high-hazard event (e.g., hitting a pole, rollover, etc.). This study found that the odds of injury were 43 percent lower when striking a guardrail instead of a median concrete barrier that was offset 1518 ft, and 65 percent lower when striking a median concrete barrier offset 714 ft. The odds of injury when striking a near-side median cable barrier were 57 percent lower than the odds for a guardrail face. This reduction for a far-side median cable barrier was 37 percent. Thus, the authors concluded that a guardrail should be preferred over a concrete wall, and a cable barrier should be preferred over a guardrail where the road and traffic conditions allow. Zou and Tarko studied the probabilities of various types of crash events possible under various road and barrier scenarios.[8] Seven barrier-relevant crash events possible after a vehicle departs the road were identified based on existing crash data, and their probabilities were estimated given the presence and location of three types of barriers: median concrete barriers, median and roadside W-beam steel guardrails, and high-tension median cable barriers. A multinomial logit model with variable outcomes was estimated based on 2,049 barrier-relevant crashes occurring between 2003 and 2012 on 1,258 unidirectional traveled 7 ways in Indiana. The results of this study indicated that road departures lead to less frequent crossings of unprotected (no barriers) medians 5080 ft wide than for narrower medians 3050 ft wide. More recently, Russo and Savolainen investigated barrier performance using an analysis of crash frequency and severity data from freeway segments where high-tension cable, thrie-beam, and concrete median barriers were installed.[9] They conducted a manual review of crash reports to identify crashes in which a vehicle left the roadway and encroached into the median. This review also involved an examination of crash outcomes when a barrier strike occurred, which included vehicle containment, penetration, or redirection onto the travel lanes. Statistical models were developed to identify factors that affect the frequency, severity, and outcomes of median-related crashes, with particular emphasis on differences between segments with varying median barrier types. Several roadway-, traffic-, and environmental-related characteristics were found to affect these metrics, with results varying across the different barrier types. The Florida DOT (FDOT) designates three factors to assess when considering barrier upgrades: (1) nature and extent of barrier deficiencies, (2) past crash history, and (3) costeffectiveness of the recommended improvement.[10] However, limited specific information on the use of these assessment factors is presented. 8 CHAPTER 2. HIGH-RISK ROADSIDE BARRIER SECTIONS IN GEORGIA Road sections with more frequent barrier collisions were identified in this research. Two sources of information were considered during the process: crash report data and maintenance data. CRASH REPORT DATA Crash report data are available in the Georgia Electronic Accident Reporting System (GEARS). In this database, crashes involving roadside-barrier collisions can be identified by filtering the "first hazardous event" with keywords of "guard rail face," "guard rail end," "median barrier," and "cable barrier." In this study, the crash data from 2017 to 2020 were used. Earlier crash data were excluded because of the potential change of the road characteristics. In the four-year period, a total of 18,108 crashes were reported where a single vehicle first struck on the roadside barrier. As shown in figure 2, most of these roadside-barrier collisions occurred in the Atlanta metropolitan area or on the high-volume national highway. The crash-severity data in GEARS showed that, overall, 70 percent of the single-vehicle roadside-barrier collisions are rated as property damage only (PDO), and about one percent are fatal (figure 3). In general, the crash severity increases with the posted speed of the road up to 6065 mph. Interestingly, the crashes that occurred on roads with a posted speed of 70 mph (interstate routes) showed less crash severity (figure 4). This trend indicated that there may be other factors affecting the severity of roadside-barrier collisions. 9 National Highway System Figure 2. Map. Vehicle collisions with roadside barriers 20172020. (Data source: GEARS) 10 K - Fatal, 114, 1% A - Serious Injury, 385, 2% PDO - Property Damage Only, 12955, 70% B - Visual Injury, 1935, 11% C - Complaint Injury, 3021, 16% K - Fatal A - Serious Injury B - Visual Injury C - Complaint Injury PDO - Property Damage Only Percentage Figure 3. Chart. KABCO severity of single-vehicle barrier-collision crashes in Georgia.(Data source: GEARS, 20172020) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 15 25 30 35 40 45 50 55 60 65 70 Posted Speed (mph) K A B C PDO Figure 4. Chart. KABCO severity of single-vehicle barrier-collision crashes at different posted speeds.(Data source: GEARS, 20172020) 11 MAINTANANCE DATA Another source of information about roadside-barrier collisions is the barrier-maintenance data. In the past, the guardrail and cable barriermaintenance records have been kept by each district in different table formats. There is an ongoing effort to transfer all these maintenance data into a central GIS database. However, at the time of this research project, only a small amount of the maintenance data has been transferred into the new database. Another drawback of the barrier-maintenance data is that it does not show collisions with concrete barriers because concrete barriers are rarely damaged by a traffic crash. Therefore, the barrier-maintenance data are used only as a supplemental data source in this research. HIGH-RISK BARRIER SECTIONS Based on the crash-report and roadside barriermaintenance data, 28 high-risk sections were selected for further analysis, including 15 concrete-barrier sections and 13 guardrail sections. General information about the selected barrier sections is presented in table 1. The locations of the selected barrier sections are shown in figure 5. Of the 28 high-risk sections selected, 14 sections are located in the Atlanta metropolitan area (District 7). All the selected road sections in this research showed more frequent barrier collisions compared to nearby road sections on the same route. The 20172020 crash report information on each barrier section was collected from GEARS. Note that when counting barrier collisions at each section, only collisions on one side (either median or shoulder) of the barrier were counted. In the case where the two sides of the road have the same barrier type, the side with more barrier collisions was selected. A significant effort was made to read the descriptions in police crash reports to 12 determine the exact location and nature of the accident. From 2017 to 2020, the singlevehicle barrier collision accidents on these high-risk sections totaled 590. On average, each section has 5.3 reported single-vehicle barrier collisions every year. The actual collision frequency is expected to be higher due to unreported crashes, a well-known issue with crash report data. DISTRICT 7 Figure 5. Map. Selected high-risk barrier sections. 13 Table 1. General information of the selected high-risk barrier sections. Section Barrier No. Type* Route 1 CB I-520 2 CB SR-104 3 CB I-520 4 CB SR-204 5 CB I-75 6 CB I-75 7 CB I-285/I-75 8 CB I-20 9 CB I-285 10 CB I-185 11 CB I-20 12 CB I-20 13 CB I-20/I-75 14 CB I-85 15 GR I-85 16 GR I-285 17 GR I-285/I-85 18 GR I-20 19 GR I-75 20 GR I-75 21 GR SM Fwy 22 GR I-20 23 GR I-75 24 GR US80/I-185 25 GR I-20 26 GR I-75 27 GR I-75 28 GR I-85 *CB = Concrete Barrier, GR = Guardrail Route Type Freeway Freeway Freeway Ramp Freeway Freeway Ramp Ramp Freeway Freeway Freeway Ramp Ramp Freeway Freeway Ramp Ramp Freeway Freeway Freeway Ramp Ramp Freeway Ramp Freeway Ramp Freeway Freeway Speed Length (mph) (ft) 60 4,330 45 1,848 65 1,267 25 1,056 70 1,214 65 686 40 106 65 1,267 65 1,109 70 475 70 2,165 35 1,426 25 370 55 1,901 55 1,056 15 158 25 264 70 686 70 2,218 70 1,109 30 317 25 264 65 1,056 30 581 70 581 25 106 70 2,640 55 1,000 AADT (OneWay) 30,000 11,950 45,700 22,100 22,400 62,500 12,510 70,000 84,000 37,050 23,550 54,550 48,900 85,750 85,750 8,160 12,510 18,200 38,100 46,350 4,980 7,930 64,000 6,940 38,800 1,790 44,750 85,750 Reported SingleVehicle Barrier Crashes (20172020) 11 9 11 19 23 23 18 53 21 16 14 43 54 54 21 16 8 13 9 13 23 13 17 12 25 24 12 15 ROAD AND TRAFFIC DATA COLLECTION The road and traffic information was collected from several different sources, including Georgia DOT (GDOT) Office of Transportation data, the GDOT Traffic Analysis & Data 14 Application (TADA), Google Earth, and Google satellite image. The collected information is presented in the appendix of this report. 15 CHAPTER 3. FREQUENCY AND SEVERITY OF ROADSIDE-BARRIER COLLISIONS In this chapter, the frequency and severity of the roadside-barrier collisions are evaluated at the 28 high-risk barrier sections. The goal is to identify sensitive parameters from road, traffic, and barrier characteristics and develop prediction models for the frequency and severity of barrier crashes for high-risk sections. These prediction models are the basis of the subsequent benefitcost analysis for comparing different roadside barriers. UNREPORTED CRASHES Unreported crashes are traffic crashes that did not generate a police report record. A national telephone survey conducted in 2010 estimated that about 30 percent of traffic crashes went unreported.[11] Although many of the unreported crashes can be assumed to be PDO, the repair cost of the roadside barrier should be considered in the benefitcost analysis. The number of unreported crashes can be estimated from the maintenance records or from the crash severity. Estimation Based on Maintenance Record One way to estimate the number of unreported crashes is to investigate the guardrail maintenance record of the site. For example, figure 6 shows the crash and maintenance records for the guardrail on the Exit 32 ramp of I-285EB between May 2019 and March 2021. During that period, the W-beam guardrail has been repaired 10 times, of which 6 repairs do not have a clear corresponding reported crash. Knowing that not all damages 16 3/9/2021 11/21/2020 10/4/2020 7/6/2020 5/5/2020 4/19/2020 3/17/2020 10/29/2019 9/22/2019 5/2/2019 to the guardrail need a repair, it can be estimated that this guardrail section has at least 60 percent unreported crashes. Repair Reported Crashes Figure 6. Map. Crash and maintenance records at the Exit 32 ramp on I-285EB. Figure 7 shows the crash and maintenance records of another guardrail section on I-85SB near Exit 77 in Atlanta. Between June 2019 and March 2021, the guardrail section has been repaired 11 times, of which 5 repairs do not have a clear corresponding reported crash. 17 1/16/2021 8/29/2020 3/29/2020 12/23/2019 11/21/2019 9/3/2019 2/22/2021 1/14/2021 11/18/2020 11/3/2020 10/4/2020 7/13/2020 3/24/2020 1/26/2020 1/9/2020 12/2/2019 6/24/2019 Knowing not all damages to the guardrail need a repair, it can be estimated that this guardrail section has at least 45 percent unreported crashes. Repair Reported Crashes Figure 7. Map. Crash and maintenance records on I-85SB near Exit 77. Figure 8 shows the crash and maintenance records of another guardrail section on the I-75SB Exit 237 ramp south of Atlanta. Between April 2019 and June 2021, the guardrail section has been repaired 13 times, of which 9 repairs do not have a clear corresponding 18 2/12/2021 12/24/2020 8/25/2020 3/10/2020 1/14/2020 10/30/2019 6/7/2021 4/20/2021 3/14/2021 10/15/2020 10/8/2020 6/29/2020 5/27/2020 4/26/2020 1/26/2020 7/28/2019 7/19/2019 5/22/2019 4/28/2019 reported crash. Knowing not all damages to the guardrail need a repair, it can be estimated that this guardrail section has at least 70 percent unreported crashes. Repair Reported Crashes Figure 8. Map. Crash and maintenance records at the Exit 237 ramp on I-75SB. At the time of this research, only District 7 has about two years of relatively complete guardrail maintenance records in the GDOT 411 database. Thus, the above analysis cannot be performed for all selected guardrail sections. Furthermore, as discussed in chapter 2, 19 5/15/2021 2/12/2021 11/17/2019 9/3/2019 7/21/2019 6/30/2019 1/17/2019 1/12/2019 concrete barrier sections are rarely damaged by vehicle collisions; therefore, this method cannot be applied to determine the percentage of unreported crashes in concrete-barrier sections. Estimation Based on Crash Severity The National Cooperative Highway Research Program (NCHRP) Project 22-27 proposed a method to estimate the number of unreported crashes based on the crash severity data.[12] The basic assumption is that the percentage of non-PDO crashes increases with the posted speed of the road with a square-power relationship. Then, the percentage of unreported crashes (assumed to be PDO) at different posted speeds can be estimated by fitting the square-power relationship. The strength of this method is that it considers the effect of the posted speed. Intuitively, lower-speed roads should have more unreported crashes on roadside barriers. This method can also be applied to all barrier types. However, the GEARS data do not clearly differentiate between concrete barriers and guardrails in the "First Harmful Event" field. A collision on a double-face W-beam median can be classified as "median barrier," and a collision on a concrete shoulder barrier can be classified as "guardrail face." To identify crashes on a particular type of barrier, we filtered the "Crash Narrative" field in the GEARS data and looked for entries in which the keyword "guardrail" or "concrete" closely followed (within 40 characters) the keyword "struck." With the 20172020 crash data, we identified 4,174 guardrail collisions and 1,408 concrete-barrier collisions. Table 2 and table 3 show the statistics of KABCO severity ratings on roads with different posted speeds for collisions on guardrails and concrete barriers, respectively. 20 Table 2. Estimated unreported crashes for collisions on guardrails. Speed K A B C 30 0 0 5 2 35 2 6 27 40 40 0 2 9 5 45 5 17 67 61 50 2 1 10 10 55 9 20 111 115 60 1 5 12 15 65 14 19 114 123 70 20 42 186 246 *Injury Rate = (K+A+B+C)/Total PDO 21 195 42 371 44 542 58 539 1,039 Total Total 28 270 58 521 67 797 91 809 1,533 4,174 Observed Injury Rate* 0.28 0.28 0.29 0.34 0.32 0.36 0.33 0.32 Estimated Injury Rate 0.08 0.11 0.13 0.16 0.20 0.24 0.28 0.32 Est. Unreported Crashes 662 94 606 73 486 48 163 1 Est. % Unreported Crashes 71.03 61.84 53.77 52.14 37.88 34.53 16.77 0.07 Table 3. Estimated unreported crashes for collisions on concrete barriers. Speed K A B C 30 0 0 1 0 35 0 1 6 6 40 0 0 1 6 45 0 2 7 17 50 0 1 5 7 55 0 3 37 68 60 0 1 6 10 65 0 6 32 67 70 1 4 62 59 *Injury Rate = (K+A+B+C)/Total PDO 3 33 11 92 22 249 41 216 325 Total Total 4 46 18 118 35 357 58 321 451 1,408 Observed Injury Rate* 0.22 0.30 0.29 0.33 0.28 Estimated Injury Rate 0.12 0.17 0.20 0.24 0.28 Est. Unreported Crashes 606 486 48 163 1 Est. % Unreported Crashes 53.77 37.88 34.53 16.77 0.07 Table 2 and table 3 also show the estimated percentages of unreported crashes based on the observed injury rates. Note that the calculation was only performed when the total observed crashes exceeded 50 to obtain a reliable estimation. Figure 4 presents the estimated percentages of unreported crashes for guardrails and concrete barriers. Percentage of unreported crashes decreases with the posted speed of the road. Concrete-barrier sections 21 showed lower percentage of unreported crashes than guardrail sections, especially for lower-speed roads. This result is expected because a concrete barrier is more likely to disable a vehicle and result in a police report. Two smoothed curves were drawn in figure 9 to fit the data. These curves were used to estimate the actual frequency of crashes for each of the high-risk sections selected in this research. Meanwhile, the observed minimum unreported crashes from the three road sections in District 7 were also plotted on figure 9. The field observation matched reasonably well with the fitted curves. Estimated % of Unreported Crashes 80 Guardrail 70 Concrete Barrier 60 50 40 30 20 10 0 0 Observed % of Unreported Crashes 10 20 30 40 50 60 Speed (mph) 70 80 Figure 9. Graph. Unreported crashes at different speeds. CRASH FREQUENCY AND CRASH RATE Crash frequency is defined as the number of single-vehicle barrier collisions per year. The This parameter does not consider the traffic volume or the length of the road. The crash 22 rate of the barrier section in this study is defined as the number of single vehicle barrier collisions per unit 1000 ft of road per 1,000,000 traffic exposures, as shown in equation 1. 1,000,000,000 = 365 ()()() (1) where is the number of single vehicle barrier collisions, is the length of the barrier section in ft, is the average daily traffic (one-way), and is the number of years of record. Table 4 shows the calculated frequency and rate of single vehicle barrier collisions for the 28 high-risk sections. The Roadside Safety Analysis Program (RSAP) developed by the NCHRP Project 22-27 includes a crash-frequency prediction model.[13] The prediction is based on the road, traffic, barrier, and roadside slope features. A comparison was made between the predicted and the estimated actual number of crashes for the 28 selected barrier sections (figure 10). The RSAP underpredicts number of crashes for most of the barrier sections. This is reasonable because the baseline encroachment model in the program is not designed to represent highrisk road sections. Further evaluation showed that the RSAP is not sensitive to the radius of the horizontal curve. Therefore, the program significantly underpredicted the number of barrier crashes for highly curved highway ramps. 23 Table 4. Frequency and Rate Single Vehicle Barrier Collisions of High-Risk Sections Reported ID Crashes 1 11 2 9 3 11 4 19 5 23 6 23 7 18 8 53 9 21 10 16 11 14 12 43 13 54 14 54 15 21 16 16 17 8 18 13 19 9 20 13 21 23 22 13 23 17 24 12 25 25 26 24 27 12 28 15 Estimated Unreported Crashes 3 8 3 31 0 5 20 12 5 0 0 57 88 33 13 54 23 0 0 0 56 37 4 29 0 19 0 12 Total Crashes 14 17 14 50 23 28 38 65 26 16 14 100 142 87 34 70 31 13 9 13 79 50 21 41 25 43 12 27 Crash Frequency (Reported) 2.75 2.25 2.75 4.75 5.75 5.75 4.5 13.25 5.25 4 3.5 10.75 13.5 13.5 5.25 4 2 3.25 2.25 3.25 5.75 3.25 4.25 3 6.25 6 3 3.75 Crash Rate (Reported) 0.082 0.279 0.130 1.115 0.351 0.298 9.333 0.409 0.154 0.509 0.188 0.465 1.592 0.227 0.159 8.479 3.318 0.579 0.073 0.173 9.985 4.253 0.172 2.039 0.760 86.964 0.070 0.120 Crash Frequency (Total) 3.5 4.25 3.5 12.5 5.75 7 9.5 16.25 6.5 4 3.5 25 35.5 21.75 8.5 17.5 7.75 3.25 2.25 3.25 19.75 12.5 5.25 10.25 6.25 10.75 3 6.75 Crash Rate (Total) 0.104 0.527 0.166 2.935 0.351 0.363 19.702 0.502 0.191 0.509 0.188 1.081 4.186 0.366 0.257 37.094 12.858 0.579 0.073 0.173 34.297 16.358 0.213 6.967 0.760 155.811 0.070 0.216 The Roadside Safety Analysis Program (RSAP) developed by the NCHRP Project 22-27 includes a crash-frequency prediction model.[13] The prediction is based on the road, traffic, barrier, and roadside slope features. A comparison was made between the predicted and the estimated number of crashes for the 28 selected barrier sections (figure 10). The RSAP underpredicts number of crashes for most of the barrier sections. This is reasonable because the baseline encroachment model in the program is not designed to represent high-risk road 24 Number of Barrier Crashes per Year sections. Further evaluation showed that the RSAP is not sensitive to the radius of the horizontal curve. Therefore, the program significantly underpredicted the number of barrier crashes for highly curved highway ramps. 40 35 30 25 20 15 10 5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 Section ID RSAP Predicted Estimated Actual Figure 10. Graph. RSAP predicted vs. actual number of crashes. To develop a crash-prediction model for high-risk barrier sections, an analysis of variance (ANOVA) was first conducted to evaluate the relationship between the estimated actual crash frequency and different road characteristics. For the model development purpose, the crash frequency was normalized based on an average daily traffic (ADT) of 1,000 and a length of 1,000 ft. The result of the ANOVA is presented in table 5. Out of the six factors evaluated, the two factors that showed significant effects on the crash rate are the degree of horizontal curve and posted speed. The most significant factor to the frequency of roadside-barrier collisions is the degree of horizontal curve (=5729.6/radius). 25 A sharper horizontal curve (such as that in a ramp) significantly increases the frequency of roadside-barrier collisions (figure 11). Table 5. ANOVA on crash frequency. Df Sum Sq Degree of curve 1 7955 Lane width 1 1134 Number of lanes 1 65 Vertical grade 1 18 Lateral clearance 1 764 Posted speed 1 6147 Residuals 21 8509 *Significant factor (p-value<0.05) Mean Sq 7955 1134 65 18 764 6147 405 F value 26.658 4.717 1.324 0.044 3.139 19.169 P-value 0.0002* 0.1091 0.6926 0.8371 0.1841 0.0008* Figure 11. Chart. Crash frequency vs. degree of curve. 26 Based on the ANOVA, two multilinear regression models were developed to describe the crash rate of barrier collisions in high-risk sections. The first regression model (equation 2) considers both degree of the horizontal curve and the posted speed. The second regression model (equation 3) uses only the degree of the horizontal curve. The coefficients of determination (2) of the two models are 0.759 and 0.758, respectively. It appears that dropping the posted speed from the equation does not reduce the 2 value of the regression model much. This is result understandable because the degree of the curve and the speed of the road are correlated variables in roadway design. = (0.1372-0.0069-0.9661) 2 = 0.759 (2) = (0.1454-1.4053) 2 = 0.758 (3) where, = barrier crash rate (/1,000,000 exposure/1,000 ft/year) of single vehicle barrier collisions, = degree of curve ( = 5729.6 / radius), and = posted speed of the road. It should be noted that equations 2 and 3 represent only the high-risk sections. In fact, there are many ramp sections showed lower numbers of crash rate. Therefore, these equations are not supposed to replace the performance functions in the RSAP programs. CRASH SEVERITY The KABCO severity information was collected from the GEARS database, including 200 crashes on guardrails and 390 crashes on concrete barriers. These crash data were combined with the 513 unreported crashes (all assumed to be PDO crashes) estimated based on figure 9. The combined KABCO severity distribution at different posted speeds is shown in figure 12 for guardrail sections and figure 13 for concrete-barrier sections, 27 respectively. Both types of barriers showed increased crash severity with the increase of posted speed. This increasing trend is more obvious in guardrail sections, as low-speed sections showed very small injury rates. As for concrete barriers, collisions at all speeds can produce more than a 10 percent injury rate, and the increase of injury rate increases more gently with speed. At 70 mph, the injury rates of the two barriers become similar. 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 15 20 25 30 35 40 45 50 55 60 65 70 K A B C PDO K+A B+C Figure 12. Graph. Crash severity in high-risk guardrail sections. 28 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 15 20 25 30 35 40 45 50 55 60 65 70 K A B C PDO K+A B+C Figure 13. Graph. Crash severity in high-risk concrete-barrier sections. A prediction model for the barrier crash severity was developed by performing an ordinal logistic regression (OLR) analysis. Ordinal logistic regression is a statistical analysis tool suitable when the dependent (e.g., crash severity) falls into several ordered classes (e.g., K, A, B, C, and O). Because fatal-injury (K), severe-injury (A), and visible-injury (B) crashes are rare events, it is difficult to develop a reliable regression for each class with a small amount of crash data. In this study, the five severity classes were combined into three: K+A, B+C, and PDO. The OLR analysis was conducted using the statistical analysis program R. The resultant regression models for guardrail and concrete barriers are provided in table 6. For comparison, the predicted cumulative probabilities of K+A, B+C, and PDO crashes are plotted in figure 12 for guardrail barriers and figure 13 for concrete barriers, respectively. 29 The OLR model provides a reasonable match to the crash-severity distribution for both types of barriers. K+A B+C PDO Table 6. Ordinal logistic regression model for crash severity. Guardrail 1 + = 1 - (-0.03796+6.1415) 1 + = 1 - (-0.03796+3.4935) - + 1 = 1 - 1 - (-0.03796+3.4935) Concrete Barrier 1 + = 1 - (-0.01785+5.2731) 1 + = 1 - (-0.01785+2.1701) - + 1 = 1 - 1 - (-0.01785+2.1701) 30 CHAPTER 4. BENEFITCOST ANALYSIS INTRODUCTION TO BENEFITCOST ANALYSIS In this study, the benefitcost analysis considers two roadside-barrier alternatives: Alternative A: 31 Guardrail Alternative B: Single-Slope Concrete Barrier The decision to switch from Alternative A to Alternative B is made by determining the benefitcost ratio (BCR). A BCR greater than 1 indicates that concrete barrier is more effective; otherwise, guardrail is more effective. According to the Highway Safety Benefit Cost Analysis Guide, BCR can be calculated as the ratio between the present value benefit (PVB) and the present value cost (PVC) (equation 4). Both present values are calculated from the life-cycle cost of the barrier with an annual discount rate. = (4) The life-cycle cost of a roadside barrier includes the installation cost and the continuous cost. The continuous cost also includes two parts: the comprehensive crash cost and the repair cost. When switching from a guardrail to a concrete barrier, any reduction in the lifecycle cost is counted into the PVB, and any increase in the life-cycle cost is counted into the PVC. Table 7 lists the unit life-cycle costs used in the BCA of this study. For a simple analysis, the BCA compares the two barrier options for a period of 40 years. 31 Table 7. Unit life-cycle cost for BCA. Initial Construction Cost (/ft) Repair Cost (/crash) Service Life (year) Alternative 1: Guardrail $25 $1,000 20 Alternative 2: Concrete Barrier $375 $20 40 Since the crash-severity model in this study was developed based on three combined severity levels (K/A, B/C, and PDO), the weighed comprehensive crash costs (table 8) were calculated following the FHWA guideline.[3] Considering the crash-severity model, the expected comprehensive crash cost per accident can be determined for different speeds. Figure 14 shows that the when the posted speed of the road is 55 mph, the expected comprehensive crash costs per barrier-collision accident for guardrail and concrete barrier are approximately the same. At higher speeds, crashes on guardrail barriers costs more than crashes on concrete barriers. At lower speeds, the trend is the opposite. Table 8. Weighted comprehensive crash cost. K/A B/C PDO $3,085,873 $154,063 $11,900 For high-risk guardrail sections, the time period when the guardrail is damaged and not functioning should be considered. In this study, a crash is assumed to make a 100-ft-long guardrail section nonfunctioning for two weeks. The crash cost during this period is assumed to be doubled. Note that these numbers were selected arbitrarily. In reality, the effect of a nonfunctioning guardrail depends on the more hazardous obstacles behind the 32 guardrail. The adjustment factor to the crash cost of guardrail sections is shown in equation (5) 100 2 = N 52 + 1 < 2 (5) where, = crash cost adjustment factor for the damaged guardrails = number of predicted crashes in a year = length of the section Expected comprehensive crash cost per accident (in thousand dollars) 160 140 120 100 80 60 40 20 0 20 30 40 50 60 70 80 Posted Speed (mph) Guardrail Concrete Barrier Figure 14 Graph. Expected comprehensive crash cost per accident at different speeds 33 BENEFITCOST ANALYSIS FOR HIGH-RISK SECTIONS The thirteen high-risk guardrail sections (sections 1628) were analyzed to determine whether concrete barrier is a more economical solution. In the BCA, the number of the barrier collisions was estimated based on the crash data from GEARS instead of the crashfrequency model. Table 9 shows the calculated BCRs for the selected high-risk sections. All sections with greater than 55 mph posted speed showed that concrete barrier is the more economic barrier type compared to guardrail. This result can be explained from figure 12 to figure 14. Crashes at this speed into a guardrail have a higher chance to result in a K/A injury than those collisions into a concrete barrier. By switching to a concrete barrier, the reduction in crash cost (as well as repair cost) outweighs the increased installation cost. For sections with less than 55 mph posted speed, although the frequent crashes produced a higher repair cost for a guardrail barrier, the reduced crash cost still made the guardrail barrier more economic compared to a concrete barrier. It was also observed from the BCA that the change in the construction or repair cost was usually insignificant compared to the change in the crash cost. 34 Table 9. BCA for the selected guardrail sections. ID Route Speed Degree of Curve 16 I-285 15 42 17 I-285/I-85 25 33.9 18 I-20 70 0 19 I-75 70 1 20 I-75 70 0 21 SM Fwy 35 26 22 I-20 25 33 23 I-75 65 1 24 US80/I-185 30 17.4 25 I-20 70 0 26 I-75 25 32 27 I-75 70 2 28 I-85 55 4.6 * GR = Guardrail; BC = Concrete Barrier Current Barrier Type* GR GR GR GR GR GR GR GR GR GR GR GR GR Crash per Year 17.5 7.8 3.3 2.3 3.3 19.8 12.5 5.3 10.3 6.3 10.8 3.0 6.8 BCR 0.1 0.1 37.9 26.1 37.9 0.1 0.1 42.2 0.1 75.4 0.1 34.3 16.3 Economic Barrier Type* GR GR CB CB CB GR GR CB GR CB GR CB CB BENEFITCOST ANALYSIS FOR ROAD SECTIONS WITH KNOWN CRASH FREQUENCY To understand the cost-effectiveness of the two barrier types, the BCA was repeated for a range of speeds and crash frequencies. Two examples are provided below. The first example (see table 10) is a 1000-ft slightly curved freeway segment with a radius of 1,500 ft and a speed limit of 60 mph. Given there are 10 crashes per year observed in this section. The concrete barrier reduces crash cost and repair cost with a total PVB of $17,197,000, which outweighs the increase in installation cost of $339,000. Therefore, the concrete barrier is a more economical alternative compared to guardrail. 35 Table 10. BCA Example 1 freeway segment. Input: Interest Rate = 4% Posted Speed = 60 mph Radius = 1,500 ft Length = 1,000 ft Number of Crashes = 10/year Output: Estimated Crashes in each KABCO category 1. Guardrail 2: Concrete Barrier K + A = 0.2 K + A = 0.1 B + C = 2.1 B + C = 2.4 PDO = 7.7 PDO = 7.5 Adjustment Factor = 1.04 Life-cycle Costs NPV1 NPV2 ($1000) ($1000) Construction Cost 36 375 Crash Cost 103,300 86,140 Repair Cost 950 19 Benefit ($1000) 17,160 37 BCR Cost ($1000) 339 50.7 The second example (see table 11) is a 200-ft curved freeway ramp section with a radius of 250 ft and a speed limit of 30 mph. Given there are 10 crashes per year observed in this section. Compared to guardrail, the concrete-barrier option increased the crash cost due to rigid barrier with a total PVC of $7,998,000. Although the repair cost reduced with a PVB of $931,000, the benefit is insignificant compared to the cost. Therefore, the guardrail is still a more economical barrier type compared to the concrete barrier. 36 Table 11. BCA Example 2 freeway ramp. Input: Interest Rate = 4% Posted Speed = 30 mph Radius = 250 ft Length = 200 ft Number of Crashes = 10/year Output: Estimated Crashes in each KABCO category 1. Guardrail 2: Concrete Barrier K + A = 0.1 K + A = 0.1 B + C = 0.8 B + C = 1.5 PDO = 9.1 PDO = 8.4 Adjustment Factor = 2 Life-cycle Costs NPV1 NPV2 ($1000) ($1000) Construction Cost 7 75 Crash Cost 49,620 57,550 Repair Cost 950 19 Benefit ($1000) 931 BCR Cost ($1000) 68 7,930 0.12 The above analysis was repeated for a range of posted speeds and crash frequencies. The resulted BCR values are presented in table 12. The shaded cells in the table indicate BCR values greater than 1. The BCA result showed that concrete barrier is more cost-effective for road sections with a posted speed of 55mph or higher. Guardrail barrier is more costeffective for road sections with a posted speed of 35 mph or less for the range of of crash frequencies analyzed. For roads with a posted speed of 40 to 50 mph, concrete barrier should be considered for higher-risk road sections. Table 14 can be used as a guide in selecting roadside barriers for high-risk road sections. It should noted that the maintenance record is a preferred way to estimate the crash frequency when using table 14. If the crash record is used, the number of unreported crashes should be estimated based on figure 9. 37 Table 12. BCR result with known crash frequency. Crash Posted Speed Frequency (/1000-ft /year) 20 25 30 35 40 45 50 55 60 65 70 50 0.1 0.1 0.1 0.2 2.5 92.3 197.8 329.8 492.9 692.3 933.6 45 0.1 0.1 0.1 0.2 0.6 66.9 159.2 275.0 418.2 593.4 805.7 40 0.1 0.1 0.1 0.1 0.4 45.0 124.8 225.0 349.1 501.2 685.5 35 0.1 0.1 0.1 0.1 0.3 26.8 94.6 179.9 285.7 415.5 573.0 30 0.1 0.1 0.1 0.1 0.2 12.2 68.6 139.6 227.9 336.4 468.1 25 0.1 0.1 0.1 0.1 0.2 1.0 46.7 104.2 175.8 263.9 370.9 20 0.1 0.1 0.1 0.1 0.1 0.4 29.0 73.7 129.4 197.9 281.4 15 0.1 0.1 0.1 0.1 0.1 0.3 15.5 48.0 88.6 138.6 199.5 10 0.1 0.1 0.1 0.1 0.1 0.2 6.1 27.1 53.4 85.8 125.4 5 0.1 0.1 0.1 0.1 0.1 0.1 1.0 11.1 23.9 39.6 58.8 Table 13. Recommended barrier type for sections with known crash frequency. Crash Frequency* Posted Speed (/1000-ft /year) 20 25 30 35 40 45 50 55 60 65 70 50 GR GR GR GR CB CB CB CB CB CB CB 45 GR GR GR GR GR CB CB CB CB CB CB 40 GR GR GR GR GR CB CB CB CB CB CB 35 GR GR GR GR GR CB CB CB CB CB CB 30 GR GR GR GR GR CB CB CB CB CB CB 25 GR GR GR GR GR CB CB CB CB CB CB 20 GR GR GR GR GR GR CB CB CB CB CB 15 GR GR GR GR GR GR CB CB CB CB CB 10 GR GR GR GR GR GR CB CB CB CB CB 5 GR GR GR GR GR GR CB CB CB CB CB *GR=Guardrail, CB=Concrete Barrier 38 BENEFITCOST ANALYSIS FOR ROAD SECTIONS WITH UNKNOWN CRASH FREQUENCY If the past crash record is unavailable, equation 3 and figure 11 can be used to estimate the crash rate. The BCA was repeated for a range of degrees of curve and traffic volumes. The resulted BCR values are listed in table 14. The shaded cells in the table indicate BCR values greater than 1. Conditions that favor a concrete barrier over guardrail are road sections that are straight (or slightly curved) or with a one-way AADT of 40,000 or more. Sharply curved road sections (e.g., a freeway exit ramp) with a high traffic count also justifies a concrete barrier according to table 14. However, The crash data analyzed in this study did not cover concrete barriers on roadways with a degree of curve > 20 (or radius < 300ft). Therefore, at this moment, concrete barriers are not recommended in these sharp ramps because a collision from a speeding vehicle into a concrete barrier may cause a severe accident. The results in table 14 can be converted to graph form (see figure 15) for roadside-barrier selection when past the crash record is unavailable. It should be noted that figure 15 is based on the BCA result on high-risk freeway barrier sections. In practice, other factors should also be considered in the selection of roadside barriers, such as the deflection limitation. 39 Table 14. BCR result with estimated crash frequency. Degree of Curve () AADT 0 3 6 9 12 15 18 21 24 27 30 100,000 106.3 176.9 222.8 140.6 133.7 224.8 548.6 1465.0 2914.2 4137.8 5959.1 50,119 50.4 81.6 96.6 44.6 15.5 1.2 29.8 175.1 645.2 1966.3 2986.6 25,119 24.6 39.2 44.6 15.9 0.4 0.2 0.2 0.2 0.9 221.8 925.2 12,589 12.1 19.2 21.4 6.3 0.2 0.1 0.1 0.1 0.1 0.2 1.5 6,310 6.0 9.5 10.5 2.8 0.2 0.1 0.1 0.1 0.1 0.1 0.1 3,162 3.0 4.7 5.2 1.3 0.1 0.1 0.1 0.1 0.1 0.1 0.1 1,585 1.5 2.4 2.6 0.6 0.1 0.1 0.1 0.1 0.1 0.1 0.1 794 0.8 1.2 1.3 0.3 0.1 0.1 0.1 0.1 0.1 0.1 0.1 398 0.4 0.6 0.6 0.1 0.0 0.0 0.0 0.1 0.1 0.1 0.1 200 0.2 0.3 0.3 0.1 0.0 0.0 0.0 0.0 0.0 0.0 0.1 100000 Concrete Barrier 10000 One-Way AADT W-Beam Guardrail 1000 Note: This chart should not override the deflection requirement on the barrier. 100 0 5 10 15 20 25 30 Degree of Curve () Figure 15. Graph. Barrier Selection Chart for Sections without Crash Record. 40 CHAPTER 5. CONCLUSIONS AND RECOMMENDATIONS SUMMARY Twenty-eight (28) freeway sections with frequent roadside-barrier collisions were identified in Georgia. Road design, traffic, and crash records pertaining to the selected road sections have been collected. Statistical analysis was performed on the barrier crash rate and the crash severity. The regression models were used in the benefitcost analysis (BCA) to determine whether a concrete barrier is a more cost-effective alternative to a guardrail barrier. The results of the BCA were converted to simple decision-making tools for selecting the cost-effective barrier type for different road sections. CONCLUSIONS The following conclusions can be drawn from this research: 4. The frequency of roadside-barrier collisions is mostly affected by the traffic volume and the degree of horizontal curve of the road. 5. The existence of unreported crashes poses a challenge to the barrier safety research. It leads to a mismatch between the crash data and the maintenance record. The percentage of unreported crashes reduces nonlinearly with the posted speed of the road. Concrete barrier shows a slightly lower percentage of unreported crashes than guardrail. 6. The severity of barrier collisions is mostly affected by the posted speed of the road and the barrier type. The crash severity in general increases exponentially with the posted speed of the road. When the posted speed of the road is less than 55 mph, concrete barriers 41 produce more severe crashes than guardrail due to the rigidity of the barrier. When the posted speed of the road is more than 55 mph, guardrails produce more fatal and severe (K/A) crashes due to the increased odds of penetrating/ vaulting. 7. The BCA result showed that concrete barrier is more cost-effective for road sections with a posted speed of 55mph or higher. Guardrail barrier is more cost-effective for road sections with a posted speed of 35 mph or less for the range of of crash frequencies analyzed. For roads with a posted speed of 40 to 50 mph, concrete barrier should be considered for higher-risk road sections. 8. When the past crash record is unavailable, the cost-effective barrier type can also be determined based on an estimated barrier collision rate. In general, conditions that favor a concrete barrier over guardrail are straight (or slightly curved) road sections or sections with a one-way AADT of 40,000 or more. RECOMMENDATIONS 1. The Roadside Safety Analysis Program (RSAP) developed from the National Cooperative Highway Research Program (NCHRP) Project 22-27 is the most sophisticated tool available for benefitcost analysis of roadside safety features. However, the RSAP baseline model often underestimates the actual crash frequency of "high-risk" barrier sections, especially for sharp horizontal curves (e.g., ramps). Although the program allows use of a modification factor to boost the predicted crash frequency, it still requires a judgment of the total number of crashes (including the unreported crashes) of the road. It is recommended that the Georgia Department of Transportation (GDOT) wait for the next updated version of the program. 42 2. The roadside barrier selection tools (table 14 and figure 15) developed in this research can be used as a quick decision-making tool for barrier-upgrade projects. When a more accurate benefitcost analysis is needed, the barrier crash-frequency model and the crashseverity model developed in this research can be applied in a spreadsheet or other computation program. Guardrail maintenance record should be used when available to determine the past average crash frequency of the road. When crash records are used, unreported crashes must be considered using the method presented in this study. 3. It is recommended GDOT build an inventory database for roadside barriers and use an asset management system for tracking the history and condition of roadside barriers. 43 APPENDIX A. SELECTED ROADWAY SECTIONS DATA Table 15. Road and traffic characteristics. ID Route Dir. LAT LONG 1 I-520 EB 33.40788 2 SR-104 WB 33.51786 3 I-520 NB 33.47510 4 SR-204 Ramp 31.98267 5 I-75 SB 31.06391 6 I-75 NB 34.03271 7 I-285/I-75 Ramp 33.89180 8 I-20 WB 33.71616 9 I-285 SB 33.84371 10 I-185 NB 32.49149 11 I-20 EB 33.61288 12 I-20 Ramp 33.71283 13 I-20/I-75 Ramp 33.74393 14 I-85 NB 33.66611 15 I-85 NB 33.68327 16 I-285 Ramp 33.90165 17 I-516 EB 32.04268 18 I-20 EB 33.68162 19 I-75 SB 33.12454 20 I-75 NB 32.96043 21 SM Fwy Ramp 33.82236 22 I-20 Ramp 33.71640 23 I-75 NB 34.02303 24 US-80/I-185 Ramp 32.54461 25 I-20 WB 33.40788 26 I-75 Ramp 34.89399 27 I-75 NB 34.08167 28 I-85 SB 33.68655 *CB = Concrete Barrier, GR = Guardrail -82.0347 -82.0053 -82.0862 -81.2004 -83.4016 -84.5766 -84.4592 -84.2599 -84.4876 -84.9431 -83.8266 -84.2433 -84.3899 -84.4169 -84.4101 -84.2728 -81.1469 -85.2816 -84.0062 -83.8143 -84.1647 -84.2591 -84.5709 -84.9583 -82.0347 -85.0744 -84.6297 -84.4018 Length (ft) 4330 1848 1267 1056 1214 686 106 1267 1109 475 2165 1426 370 1901 1056 158 1162 686 2218 1109 317 264 1056 581 581 106 2640 1000 AADT OneWay 30,000 11,950 45,700 22,100 22,400 62,500 12,510 70,000 84,000 37,050 23,550 54,550 48,900 85,750 85,750 8,160 12510 18,200 38,100 46,350 4,980 7,930 64,000 6,940 38,800 1,790 44,750 85,750 Post_ Speed 65 45 65 25 70 65 40 65 65 70 70 35 25 55 55 15 55 70 70 70 30 25 65 30 70 25 70 55 Degree of Curve 0 5 2 19 0 0 10 2 1 3 0 13 12 3 5 42 4 0 1 0 26 33 1 17.4 0 32 2 4.6 Vertical Grade (%) -3.5 1.1 1 -1 -1.3 0.8 0 -1.6 0 4.3 -1 0 3 -1.8 0 -1 -1.7 -2.2 -2.4 2.6 -1 1 0 0 0 1 -2.2 0 Barrier Location Median Median Median Median Median Median Shoulder R Median Median Median Median Shoulder L Shoulder R Median Median Shoulder Shoulder Shoulder L Median Shoulder R Shoulder Shoulder L Median Shoulder L Shoulder R Shoulder R Shoulder Shoulder L Barrier Type* CB CB CB CB CB CB CB CB CB CB CB CB CB CB CB GR GR GR GR GR GR GR GR GR GR GR GR GR Barrier Clear. 16 10 12 12 12 9 20 12 9 8 18 12 12 8 8 16 12 12 12 12 15 35 13 7 16 12 12 8 No. of Lanes 2 2 4 2 3 3 2 5 4 3 2 2 2 3 3 1 2 2 3 2 1 1 3 1 3 1 3 4 Lane Width 12 12 12 10 12 12 12 12 12 12 12 12 12 12 12 12 12 11 12 12 10 12 12 12 12 14 12 12 44 Table 16. Crashes and severity. ID K A B C O U Total Crash Posted Degree of Vertical Speed Curve Grade (%) Barrier Location Barrier Type Barrier Clear. No. of Lanes Lane Widt h 1 0 0 0 2 9 0 11 65 0 -3.5 Median CB 16 2 12 2 0 0 2 2 5 0 9 45 5 1.1 Median CB 10 2 12 3 0 0 1 4 6 0 11 65 2 1 Median CB 12 4 12 4 0 0 1 5 13 0 19 25 19 -1 Median CB 12 2 10 5 0 0 4 3 16 0 23 70 0 -1.3 Median CB 12 3 12 6 0 0 2 6 15 0 23 65 0 0.8 Median CB 9 3 12 7 0 0 2 4 12 0 18 40 10 0 Shoulder R CB 20 2 12 8 0 1 5 9 38 0 53 65 2 -1.6 Median CB 12 5 12 9 0 1 3 2 14 1 21 65 1 0 Median CB 9 4 12 10 0 0 1 4 11 0 16 70 3 4.3 Median CB 8 3 12 11 0 0 2 2 10 0 14 70 0 -1 Median CB 18 2 12 12 0 0 4 13 26 0 43 35 13 0 Shoulder L CB 12 2 12 13 0 3 10 10 29 2 54 25 12 3 Shoulder R CB 12 2 12 14 0 1 5 14 33 1 54 55 3 -1.8 Median CB 8 3 12 15 0 2 4 6 9 0 21 55 5 0 Median CB 8 3 12 16 0 0 2 2 12 0 16 15 42 -1 Shoulder GR 16 1 12 17 0 0 1 2 5 0 8 55 4 -1.7 Shoulder GR 12 2 12 18 0 1 2 1 9 0 13 70 0 -2.2 Shoulder L GR 12 2 11 19 0 0 2 2 5 0 9 70 1 -2.4 Median GR 12 3 12 20 0 0 0 5 8 0 13 70 0 2.6 Shoulder R GR 12 2 12 21 0 0 6 1 16 0 23 30 26 -1 Shoulder GR 15 1 10 22 0 1 1 2 7 2 13 25 33 1 Shoulder L GR 35 1 12 23 0 0 2 3 12 0 17 65 1 0 Median GR 13 3 12 24 0 0 2 1 9 0 12 30 17.4 0 Shoulder L GR 7 1 12 25 0 0 1 2 22 0 25 70 0 0 Shoulder R GR 16 3 12 26 0 1 1 1 21 0 24 25 32 1 Shoulder R GR 12 1 14 27 1 0 3 3 5 0 12 70 2 -2.2 Shoulder GR 12 3 12 28 0 1 0 4 10 0 15 55 4.6 0 Shoulder L GR 8 4 12 45 ACKNOWLEDGMENTS This research project received tremendous support from many GDOT offices and personnel. We especially thank Mr. Frack Flanders and Mr. Christopher Rudd from the Office of Design Policy and Support for providing many valuable inputs on the research subject throughout the project period. Mr. Brennan Roney from the Research Office helped us connect to different offices and resolve many issues during the project progression. Mr. Teague Buchanan and Ms. Sindhura Surapaneni from the IT office provided access to the GDOT data system. Dr. Hong Liang and Mr. Doug Hall from the GIS team spent a lot of time preparing the road map data and the GeoPI data for our use. Mr. Joseph Monti and Mr. Eric Conklin from the State Transportation Data Office provided the state route, shoulder, median, and lane data. Mr. Ed Adams from the Traffic Safety Program helped provide access to the GEARS and the NUMETRIC websites. We also thank all the district engineers and maintenance engineers for compiling and sending us all the barriermaintenance records. The research goal could not be accomplished without all the quality data provided. Finally, we thank all the undergraduate and graduate students who worked on this project, including Ms. Maria Ordonez (former GDOT engineer), Ms. Caitlyn Stephens, Mr. Garrett Dean, and Mr. Saliu Babatunde. 46 REFERENCES 1. Sicking, D.L., Reid, J.D., and Rohde, J.R. (2002). "Development of the Midwest Guardrail System." Transportation Research Record: Journal of the Transportation Research Board, 1797(1), pp. 4452. Available online: https://doi.org/10.3141%2F1797-06. 2. American Association of State Highway and Transportation Officials (AASHTO). (2011). Roadside Design Guide. 4th edition, RSDG-4, AASHTO, Washington, DC. Available online: https://trust.dot.state.wi.us/ftp/dtsd/bts/environment/library/PE/AASHTO-RSDG-4- 2011-with-2012-&-2015-errata.pdf. 3. Lawrence, M., Hachey, A., Bahar, G., and Gross, F. (2018). Highway Safety Benefit Cost Analysis Guide. FHWA-SA-18-001, Federal Highway Administration, Office of Safety, Washington, DC. Available online: https://safety.fhwa.dot.gov/hsip/docs/fhwasa18001.pdf. 4. Elvik, R. (1995). "The Safety Value of Guardrails and Crash Cushions: A MetaAnalysis of Evidence from Evaluation Studies." Accident Analysis and Prevention, 27(4), pp. 523529. Available online: https://doi.org/10.1016/0001-4575(95)00003i. 5. Center for Risk Management of Engineering Systems, University of Virginia. (2001). Risk-Based Management of Guardrails: Site Selection and Upgrade. Report VTRC 02-CR1, Virginia Transportation Research Council, Virginia Department of Transportation, Richmond, VA. Available online: https://www.virginiadot.org/vtrc/main/online_reports/pdf/02-cr1.pdf. 6. Zhu, K. and Li, S. (2009). Risk Management and Assessment of Upgrading and Standardizing Guardrail. Report FHWA/IN/JTRP-2009/7, Joint Transportation Research Program, Purdue University, West Lafayette, IN. 7. Zou, Y., Tarko, A.P., Chen, E., and Romero, M.A. (2014). "Effectiveness of Cable Barriers, Guardrails, and Concrete Barrier Walls in Reducing the Risk of Injury." Accident Analysis and Prevention, 72, pp. 5565. Available online: http://dx.doi.org/10.1016/j.aap.2014.06.013. 8. Zou, Y. and Tarko, A.P. (2016). "An Insight into the Performance of Road Barriers Redistribution of Barrier-Relevant Crashes." Accident Analysis and Prevention, 96, pp. 152161. Available online: https://doi.org/10.1016/j.aap.2016.07.022. 47 9. Russo, B.J. and Savolainen, P.T. (2018). "A Comparison of Freeway Median Crash Frequency, Severity, and Barrier Strike Outcomes by Median Barrier Type." Accident Analysis and Prevention, 117, pp. 216224. Available online: http://dx.doi.org/10.1016/j.aap.2018.04.023. 10. Florida Department of Transportation (FDOT). (2017). "Roadside Safety" Plans Preparation Manual, Chapter 4, Office of Roadway Design, Florida Department of Transportation, Tallahassee, FL, pp. 5152. Available online: https://www.fdot.gov/roadway/ppmmanual/2017ppm.shtm. 11. M. Davis and Company, Inc. (2015). National Telephone Survey of Reported and Unreported Motor Crashes. DOT HS 812 183, National Highway Traffic Safety Administration, Washington, DC. Available online: https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/812183. 12. Ray, M.H. and Weir, J. (2003). In-Service Performance Evaluation Procedures Manual. NCHRP Report 490, National Cooperative Highway Research Program, Washington, DC. Available online: https://onlinepubs.trb.org/onlinepubs/nchrp/nchrp_rpt_490aD.pdf. 13. Mak, K.K. and Sicking, D.L. (2003). Roadside Safety Analysis Program (RSAP) Engineer's Manual. NCHRP Report 492, National Cooperative Highway Research Program, Washington, DC. Available online: https://onlinepubs.trb.org/onlinepubs/nchrp/nchrp_rpt_492.pdf. 48