Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers (2022)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

3   Literature Review The following literature review summarizes the impetus for improving median barrier guid- ance, the history of median barrier guidance through 2015, and the evolution of crash testing specifications for median barriers. Recent additions to the literature on the placement of median barriers are also included. This review captures the explosion of interest in highway safety modeling. Specifically, the study of crash data to model median-related crashes (Carrigan 2018; Graham 2014; Harwood 2014) and the recent study of vehicle behavior during encroachments on various slopes (Bligh 2020b) have been captured in this review. Available analytical tools are discussed and the ability to build on past research is assessed. 3.1 Interest in Developing Median Barrier Guidance NTSB is an independent federal agency charged by Congress with investigating significant highway accidents. “The NTSB determines the probable cause of the accidents and issues safety recommendations aimed at preventing future accidents.” (NTSB 2016) The following sections contain summaries of MREs that have been investigated by the NTSB in the last 20 years and resulted in a recommendation by the NTSB for the development of median barrier guidance. 3.1.1 1998 On February 12, 1997, a truck-tractor with double trailers lost control while traveling north- bound on U.S. Route 41 near Slinger, Wisconsin, and fully crossed the median into the south- bound lanes where a flatbed truck traveling southbound struck the tractor. After the collision, the flatbed truck lost control, may have run off the road to the right, then crossed the median, and entered the northbound lanes. A passenger van traveling northbound struck and under- rode the flatbed truck and a refrigerator truck also collided with the flatbed truck as shown in the scene diagram in Figure 1. (NTSB 1998) The double and flatbed truck drivers received minor injuries. The refrigerator truck driver received no injuries. There were nine occupants of the van; the driver and seven passengers were fatally injured and one passenger was seriously injured. The crash history for this site is summarized in Table 1 and the highway and median char- acteristics are summarized in Table 2. As a result of this investigation, the NTSB issued several recommendations related to median barrier selection, including: To the Federal Highway Administration: Review, with the American Association of State Highway and Transportation Officials, the median barrier warrants and revise them as necessary to reflect changes in the factors affecting the probability of C H A P T E R 3

4 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers Figure 1. Crash scene diagram: U.S. Route 41 near Slinger, Wisconsin. (NTSB 1998)

Literature Review 5   Crash location U.S. Route 41 near Slinger, Wisconsin Interstate 65 near Munfordville, Kentucky Interstate 5 in Orland, California Statewide fatal accidents • 0.6% of total head-on collisions with median barriers • 1.9% of total without median barriers • 1.74 per 100 million vehicle miles traveled (2008) • Unknown Site crash statistics • No cross-median crashes within 6 miles of the site within 3 years • 0.49 cross-median crashes per mile per year within 10 miles • 0.37 fatal cross-median crashes per mile per year within 10 miles • No cross-median collision history at the crash site Table 1. Summary of crash statistics for NTSB median crossover investigations. Crash location U.S. Route 41 near Slinger, Wisconsin I-65 near Munfordville, Kentucky Interstate 5 in Orland, California Crash date February 12, 1997 March 26, 2010 April 10, 2014 Highway type Principal rural arterial Principal rural arterial Principal urban arterial Vehicle crossing Truck-tractor in combination with 2 empty trailers Truck-tractor in combination with a 53- foot-long van semitrailer Truck-tractor in combination with two 28- foot trailers Lanes Four 12-foot lanes Four 12-foot lanes 4 lanes Shoulders Left: 6 feet (3-foot paved) Right: 10 feet (6-foot paved) Left: 4 feet NB and 3.5 feet SB; Right: 11 feet Left: 4 feet Right: 12 feet Rumble strips None present Located on all 4 shoulders Located on all 4 shoulders Median width 50 feet 60 feet 58 feet Median characteristics Depressed grassed median with 1:8 slope Depressed earthen median with 1:4 slope Gravel earthen median with oleander bushes 3–5 feet tall located near the centerline Median barrier None present 4-strand high-tension cable median barrier with cables mounted at 20, 25, 30, and 39 inches None present Test level NA NCHRP Report 350 TL3 NA Barrier placement NA 8 feet to the left of the NB traveled-way NA Grade 0.6% NB –2.6% SB Unknown Horizontal Tangent Tangent Tangent ADT (veh/day) 24,050 (1996) 36,800 (2008) 23,400 (2012) %Trucks 21% (1993) 35% (2008) 25% (2012) Design speed 70 mph 70 mph Unknown Posted speed 65 mph 70 mph 70 mph for cars and buses; 55 mph for trucks with 3 or more axles 85th percentile speed for cars 64 mph for all vehicles on the morning of the crash 76 mph 76 mph 85th percentile speed for trucks 70 mph 61 mph Table 2. Site summaries for NTSB median crossover investigations.

6 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers cross-median accidents, including changes in the vehicle fleet and the percentage of heavy trucks using the roadways. (H-98-12) Include a data element for cross-median accidents in the Guideline for Minimum Uniform Crash Criteria, which you are developing with the National Highway Traffic Safety Administration and the National Association of Governors’ Highway Safety Representatives. (H-98-13) To the National Highway Traffic Safety Administration: Include a data element for cross-median accidents in the Guideline for Minimum Uniform Crash Criteria, which you are developing with the Federal Highway Administration and the National Association of Governors’ Highway Safety Representatives. (H-98-17) To the National Association of Governors’ Highway Safety Representatives: Include a data element for cross-median accidents in the Guideline for Minimum Uniform Crash Criteria, which you are developing with the National Highway Traffic Safety Administration and the Federal Highway Administration. (H-98-18) To the American Association of State Highway and Transportation Officials: Review, with the Federal Highway Administration, the median barrier warrants and revise them as necessary to reflect changes in the factors affecting the probability of cross-median accidents, including changes in the vehicle fleet and the percentage of heavy trucks using the roadways. (H-98-24) Safety Recommendation H-98-24 was reclassified “Closed—Superseded” when it was super- seded by Safety Recommendation H-11-31, as discussed below. 3.1.2 2011 A tractor-semitrailer truck was traveling south and departed the left lane, crossing a 60-foot- wide depressed median of Interstate 65 on March 26, 2010, near Munfordville, Kentucky. After crossing the median, the truck overrode the high-tension cable median barrier adjacent to the left northbound shoulder. The truck entered the northbound travel lanes where it crossed in front of a 15-passenger van (containing 12 people) traveling northbound on Interstate 65. The passenger van struck the tractor. The truck continued across the northbound lanes to hit a cut rock wall where the truck caught fire. After impact with the truck, the van also hit the cut rock wall. The scene diagram is shown in Figure 2. The truck driver, van driver, and nine van pas- sengers died. Two van passengers sustained minor injuries. (NTSB 2011) The crash history for this site is summarized in Table 1 and the highway and median char- acteristics are summarized in Table 2. As a result of this investigation, the NTSB issued several recommendations related to median barrier selection, including: To the Federal Highway Administration: Work with the American Association of State Highway and Transportation Officials to establish war- rants and implementation criteria for the selection and installation of Test Level Four and Test Level Five median barriers on the National Highway System. (H-11-21) Work with the American Association of State Highway and Transportation Officials to identify cross- median crash rates that call for special consideration when selecting median barriers. (H-11-22) Work with the American Association of State Highway and Transportation Officials to define the criteria for median barrier selection, including heavy vehicle traffic volume. (H-11-23) To the National Highway Traffic Safety Administration: Work with the Governors Highway Safety Association to add a standard definition of “cross-median crash” and a data element for cross-median crash accidents to the Model Minimum Uniform Crash Criteria. (H-11-28) To the American Association of State Highway and Transportation Officials: Work with the Federal Highway Administration to establish warrants and implementation criteria for the selection and installation of Test Level Four and Test Level Five median barriers on the National Highway System, and publish those warrants and criteria in the Roadside Design Guide. (H-11-31) [This recommendation supersedes Safety Recommendation H-98-24.]

Literature Review 7   Work with the Federal Highway Administration to identify cross-median crash rates that call for special consideration when selecting median barriers, and publish the rates in the Roadside Design Guide. (H-11-32) Work with the Federal Highway Administration to define the criteria for median barrier selection, including heavy vehicle traffic volume, and publish the criteria in the Roadside Design Guide. (H-11-33) To the Governors Highway Safety Association: Work with the National Highway Traffic Safety Administration to add a standard definition of “cross- median crash” and a data element for cross-median crash accidents to the Model Minimum Uniform Crash Criteria. (H-11-34) 3.1.3 2015 On April 10, 2014, a truck-tractor with double trailers was southbound on Interstate 5 in Orland, California when it crossed a 58-foot-wide median. The truck-tractor first struck a northbound Nissan Altima which subsequently ran off the road. The truck-tractor then struck a northbound motorcoach, the two vehicles ran off the road, and a post-crash fire followed, as shown in Figure 3. The truck and motorcoach drivers as well as eight motorcoach passengers were fatally injured. Thirty-seven motorcoach passengers were injured. The two occupants of the Nissan Altima received minor injuries. (NTSB 2015) The crash history for this site is summarized in Table 1 and the highway and median char- acteristics are summarized in Table 2. The NTSB, as part of the investigation, assessed the need for median barriers at this location against current practice. The NTSB found that median Ce nt er m ed ia nFurrow and plow marks I-65 Southbound I-65 Northbound Curved tire marks Gouge and scrape marks High tension cable barrier (HTCB) Dodge 15-passenger van Freightliner tractor-trailer Cut rock slope Cut rock slope N 0 50 100 Scale in feet Figure 2. Crash scene diagram: Interstate 65 near Munfordville, Kentucky. (NTSB 2011)

8 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers Figure 3. Crash scene diagram: Interstate 5 in Orland, California. (NTSB 2015)

Literature Review 9   barriers were not warranted at this location, noting “. . . because of the severity of cross-median crashes, some states, including California, have stronger median barrier application policies than the RDG. . . . Even with the more robust barrier application policies in the State of Cali- fornia, the Orland crash site did not exceed the Caltrans minimum ADT of 45,000 vehicles for the 58-foot-wide median, and the fatal cross-median crash rate had not been exceeded in the preceding 5 years.” (NTSB 2015) Although the NTSB made no new recommendations to the FHWA or AASHTO as a result of this investigation, “[t]he NTSB is encouraged by the recent TRB announcement of an NCHRP research project (no. 22-31) to develop guidelines on median barrier placement location criteria and selection of median barriers types. It is anticipated that these guidelines will be integrated into an updated edition of the RDG and, therefore, will augment the criteria used by California and other states.” (NTSB 2015) 3.1.4 Summary of Crashes Investigated by NTSB The crash site characteristics and crash history of these crashes investigated by NTSB are summarized in Table 1 and Table 2. While these crashes only provide anecdotal evidence, these summaries illustrate the types of events the NTSB is called to investigate and try to prevent, through recommendations, in the future. 3.2 Evolution of Median Barrier Guidance 3.2.1 Historic Guidance Guidelines for installing barriers began with Highway Research Board Special Report 81 in 1964. In 1967, in Highway Design and Operation Practices Related to Highway Safety, (AASHO 1967) the AASHO Traffic Safety Committee published median barrier warrants stating that: “Effective median barriers should be installed on all existing and proposed high volume, high speed divided highways with narrow medians where traffic engineering studies establish the need. On multi- lane undivided highways, where there are similar traffic conditions, median barriers may also be a con- tribution to safety.” In the 1967 Highway Design and Operations Practices Related to Highway Safety, the AASHO Traffic Safety Committee also observed: “Throughout the nation there have been many serious cross-median accidents resulting in multiple deaths where the traffic volume has been much less than 40,000 vehicles per day. Some States are pro- ceeding on the basis that on any highway with a median width less than 20 feet a barrier rail should be installed regardless of traffic volume. . . . On a heavily travelled section, say with volumes over 20,000 per day, barriers should be considered on medians up to about 30 feet in width.” NCHRP Report 54: Location, Selection, and Maintenance of Highway Guardrails and Median Barriers was published in 1968. NCHRP Report 54 presented median barrier warrants based on median width and a 2-year projection of traffic volumes. When the median exceeded 40 feet, median barriers were said to not be warranted, as shown in Figure 4. NCHRP Report 54 con- sidered median barriers to have the sole purpose of reducing across-the-median, head-on collisions between vehicles in the opposing direction of travel. Consideration of the need for a longitudinal barrier due to the median terrain and/or obstacles was based on roadside warrants. (Michie 1968) NCHRP Report 118: Location, Selection, and Maintenance of Highway Traffic Barriers, pub- lished in 1971, superseded NCHRP Report 54. NCHRP Report 118 reiterated the NCHRP Report 54 statement about the purpose of median barriers and continued to base the warrant on a 2-year projection of traffic volumes and median width. The minimum width for a median to not have a barrier, however, was extended from 40 feet to 50 feet. A provision was also added that medians that exceeded 50 feet with adverse accident experience may also be considered for

10 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers barriers. (Michie 1971) The median barrier requirements from NCHRP Report 118 are shown in Figure 5. The second edition of Highway Design and Operations Practices Related to Highway Safety was published in 1974, at which time the guidance for the introduction of median barriers was changed, removing any reference to traffic volumes and only referring to median width as follows: “For narrow medians, 30  feet or less in width, a flush paved median with internal drainage and a median barrier should be considered. The median should be kept free of abrupt slopes and obstacles. . . .” (AASHTO 1974) In 1977, AASHTO published the Guide for Selecting, Locating, and Designing Traffic Barriers (1977 Barrier Guide), and median barriers were addressed in Chapter 4. (AASHTO 1974) Warrants were suggested for median barriers on high-speed, controlled-access roadways as shown in Figure 6. These warrants were suggested for use “. . . in the absence of cross median acci- dent data for a specific site.” It was further suggested that “[a]n evaluation of the number of [median Figure 4. NCHRP Report 54 median barrier requirements. (Michie 1968)

Literature Review 11   openings], accident history, alignment, sight distance, design speed, traffic volume, and median width should be made prior to non-freeway installations.” (AASHTO 1974) These guidelines were based on previous research findings reported by the Traffic Department of the State of California in 1968, by the Texas Transportation Institute (TTI) in 1974 (Research Report 140-8), and the judgment of the AASHTO Task Force for Traffic Barrier Systems, the predecessor to today’s Technical Committee on Roadside Safety (TCRS). AASHTO published the first edition of the Roadside Design Guide (RDG) in 1989 and Chapter 6 explicitly addressed median barriers. (AASHTO 1989a) The 1989 RDG reiterated the 1977 Barrier Guide warrant. No changes were made to the warrant; however, it was noted Figure 5. NCHRP Report 118 median barrier requirements. (Michie 1971)

12 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers that the warrant was for freeways and expressways rather than “high-speed, controlled access roadways.” (AASHTO 1977; AASHTO 1989a) It was also stated that the guidance was “rela- tively subjective and does not specifically address the cost-effectiveness issue.” It was further noted that guidelines that account for speed, median, slope, vehicle mix, and ADT were under development. (AASHTO 1989a) The 1989 RDG provided factors that would indicate the need to consider higher containment barriers. Specifically, the factors listed include high percentages of heavy vehicles, horizontal curvature, or severe consequences of penetration. (AASHTO 1989a) Chapter 6 of the AASHTO 1996 RDG addressed median barriers. The 1996 RDG was pub- lished using SI units and the two axes of the warrant were flipped. The “optional” portion of the warrant was extended to include all medians with widths between 30 and 50 ft (10 and 15 m) as shown in Figure 7. Additionally, the word “warranted” was changed to “evaluate need for barrier.” The warrant continued to apply to “freeways and expressways,” as did the 1989 RDG. No changes were made to the language used in the 1989 RDG suggesting factors that would indicate the need to consider higher containment barriers. The ADT por- tion of the warrant continued to be based on a 5-year projection of traffic. (AASHTO 1989a; AASHTO 1996) Chapter 6 of the AASHTO 2002 RDG reiterated the 1996 guidance in dual units. No substan- tive changes were made to the warrant; however, while the 1996 guidelines applied to freeways and expressways, the 2002 guidelines applied to high-speed, fully controlled-access roadways. Figure 6. 1977 AASHTO Barrier Guide and 1989 Roadside Design Guide median barrier warrants. (AASHTO 1977; AASHTO 1989a)

Literature Review 13   The 2002 RDG referenced an ongoing study on median barrier warrants, indicating that changes to the warrant would be forthcoming. (AASHTO 1996; AASHTO 2002) The ongoing study was presumably NCHRP Project 17-14, “Improved Guidelines for Median Safety,” which was delayed due to data collection difficulties. When completed, the analysis results were incon- clusive. The research effort concluded in 2004, but the results were not incorporated into sub- sequent editions of the RDG. (Hughes 2004) 3.2.2 Current Guidance In 2006, AASHTO published an updated Chapter 6 of the RDG. The updated Chapter 6 referenced a 2004 survey of cross-median crashes conducted by the FHWA. The FHWA received responses from 25 states indicating “. . . a significant percentage of fatal cross-median crashes occurring where median widths exceed 10 m [30 ft]. While the survey found that some cross- median crashes occurred in medians in excess of 60 m [200 ft] wide, approximately two-thirds of crashes occurred where the median was less than 15 m [50 ft].” (AASHTO 2006) Unfortunately, the only existing documentation of this survey is a PowerPoint presentation given by Mr. Richard Powers in December 2004 at the Transportation Engineering and Safety Conference held in University Park, Pennsylvania. Figure 7. 1996 and 2002 AASHTO Roadside Design Guide median barrier warrants. (AASHTO 1996; AASHTO 2002)

14 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers The current guidance (i.e., issued in 2006 and reiterated in 2011) for the use of median barriers is shown in Figure 8. These guidelines apply to high-speed, fully controlled-access roadways, as did the 2002 guidelines. Notice that where the guidance was once “evaluate the need for barriers,” it is now “barrier recommended.” Where barriers were once “optional,” barriers are now “considered.” “Barrier not normally considered” was changed to “barrier optional.” Additionally, the small triangle cutout in the 1996 and 2002 guidelines at 20,000 to 30,000 vehicles per day and 6 to 10 m has been removed. The 2006 guidelines essentially justified barriers for any median less than 50 feet wide at any ADT rate over 20,000 vehicles per day, whereas the 1996 and 2002 guidelines suggested 30 feet wide medians or less at any ADT rate over 20,000 vehicles per day. There have been two additional subtle changes including highway types and ADT. The 1977, 2002, and 2006 warrants apply to high-speed, controlled-access highways, whereas the 1989 and 1996 warrants applied to freeways and expressways. (AASHTO 1977; AASHTO 1989a; AASHTO 1996; AASHTO 2002; AASHTO 2006) The ADT portion of the warrant has histori- cally been based on a 5-year projection of traffic. This 5-year projection, however, was not carried over to the 2006 RDG. A roadway with a median width of 20 feet that experiences 18,500 vehicles per day with 2% traffic growth per year would fall in the “barrier optional” group according to the 2006 guidance. The same roadway characteristics that “warranted” a barrier according to the 1977 Barrier Guide would have been classified as “evaluate the need” under the 1996 and 2002 guidance. The barrier test level selection factors first listed in the 1989 RDG and repeated in the 1996 and 2002 versions of the RDG were again repeated in the 2006 RDG. Recall those test level selection Figure 8. 2006 AASHTO Roadside Design Guide median barrier warrants. (AASHTO 2006)

Literature Review 15   factors include: high percentages of heavy vehicles, horizontal curvature, or severe consequence of penetration. Many states adopted the recommendations of the RDG and have incorporated them into their standards. Sometimes, to provide a higher level of safety and address specific local con- cerns, states adopt policies more stringent than the recommendations in the RDG. The 2006 RDG observed that some states have experienced increases in CMCs and have developed their own guidelines. The following is a summary of some of the state guidelines for the use of median barriers. 3.2.2.1 Arizona. In 1977 the Arizona Department of Transportation (ADOT) pub- lished the Manual of Highway Geometric Design that stated that “a barrier is used on any freeway median, or portion thereof, less than 36 feet in width.” (ADOT 1977) This general policy was modified in 1996 when median barriers were to be used for rural highways or any controlled-access highway with a median width less than 9 meters (i.e., 30 feet), which conformed to the AASHTO RDG guidance at the time, as shown earlier. Arizona, like sev- eral other states, began to re-examine its median barrier warrants in the 1990s in response to the perception of more cross-median crashes. ADOT personnel surveyed numerous states to determine their experience and policy concerning median barriers. After carefully studying the research and experience of other states, ADOT issued a report with new guid- ance for urban divided highways in 1999. (ADOT 1999) The new median barrier guidelines stated that: A. “Median barriers will be installed on urban freeway sections having median widths of 50ʹ and less. B. Median barriers will be considered for urban freeway sections having median widths of up to 75ʹ wide when there are three or more through traffic lanes in each direction.” (ADOT 1999) The 1999 guidance was specifically intended for urban roadways. Since 2009 the ADOT guid- ance has been to use median barriers on urban and rural high-speed, fully controlled-access highways with median widths of 50 ft or less when there are fewer than three lanes in one direc- tion and 75 ft or less when there are three or more lanes in one direction. The 2012 Arizona Roadway Design Guidelines include guidelines applicable for placement of median barriers with new construction. “Median barrier shall be installed on high-speed fully controlled-access highways having traversable medians under the following conditions: a) Median widths 50 ft and less. b) Median widths 75 ft and less when there are three or more through lanes in each direction.” (ADOT 2016) 3.2.2.2 California. The 2006 RDG states that “each transportation agency has the flex- ibility to develop its particular median barrier guidelines.” (AASHTO 2006) The RDG further states that for “locations with median widths equal to or greater than 15 m [50 ft], a barrier is not normally considered. The RDG goes on to mention the State of California as an example of a state that developed an accident history-based median barrier warrant, without necessarily endorsing it. The Caltrans Traffic Manual was published on January 5, 2012, and Chapter 7 addresses the subject of median barriers. Caltrans states “[t]he purpose of median barriers is to reduce the risk of an errant vehicle crossing the median and colliding with opposing traffic.” (Caltrans 2012) Caltrans specifically defines a cross-median collision as “. . . one in which an errant vehicle crosses the median of a highway with four or more lanes and strikes, or is struck, by a vehicle

16 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers from the opposite direction.” (Caltrans 2012) Either a collision or volume/width study warrant is suggested to identify locations for study (i.e., this is not a barrier installation warrant but a study warrant). The collision study warrant applies to freeways, expressways, and conventional highways. The collision study warrant is met if a location with four or more lanes satisfies either of these criteria: • A location has three or more cross-median collisions of any severity and a total cross-median collision rate of at least 0.5 collisions per mile per year in 5 years, or • A location has three fatal collisions or more and a fatal cross-median collision rate of at least 0.12 collisions per mile per year in 5 years. If a highway has two or three lanes, the collision study warrant is met based only on the fatal collision criteria. Caltrans provides a quantitative definition for each highway type. A brief extract of each defi- nition is shown here to illustrate that these study warrants apply to both divided and undivided highways as well as both controlled- and uncontrolled-access highways. • “A freeway is defined as a divided arterial highway with full control of access.” • “An expressway is defined as an arterial highway with at least partial control of access, and which may or may not be divided.” • “A multilane conventional highway (two or more lanes in each direction) is defined as a high- way without control of access . . . These highways may or may not be divided.” • “Two- and three-lane conventional highways are defined as highways without control of access, and . . . there are at-grade intersections.” Caltrans provides the volume/width study warrant shown in Figure 9, which applies to free- way medians only. “The need for a median barrier should be considered on freeways whenever the volume and median width plot in the gray area.” (Caltrans 2012) When either of these study warrants is met, Caltrans suggests that the location be further studied. “All studies must document the decision to install or not to install a median barrier on the freeway system, and the District Traffic Safety Engineer must approve the decision to install or not install median barrier, and the decision must be documented in the project files.” (Caltrans 2012) Notice the Caltrans policy does not mandate the use of a barrier if these crash rates are exceeded; it recommends only that the site be studied for possible installation of a median barrier. As the Caltrans policy explicitly states, these are study warrants, not installation warrants. More detailed instructions on how to implement the study warrants are provided in Chapter 7 of the Caltrans Traffic Manual. Accompanying the study warrants are placement guidelines to be used after the decision to install permanent median barriers has been made. These placement guidelines are shown in Table 3. 3.2.2.3 Connecticut. Chapter 13 of the December 2003 Connecticut Highway Design Manual calls for NCHRP Report 350 TL3 median barriers on all freeway medians of 66 feet or less regardless of traffic volumes. “On non-freeways, the designer should evaluate the crash his- tory, traffic volumes, travel speeds, median width, alignment, sight distance and construction costs to determine an appropriate median barrier.” (CTDOT 2003) 3.2.2.4 Kentucky. Kentucky’s “Guidelines for Median Barrier Application on Depressed Medians of Fully Controlled-Access Highways” (Kentucky 2006) dated March 6, 2006, were developed while AASHTO was developing the 2006 RDG with the updated Chapter 6. Kentucky

Literature Review 17   Figure 9. Caltrans freeway median barrier study warrant. (Caltrans 2012) Table 3. Caltrans median barrier placement guidelines. (Caltrans 2012)

18 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers was aware of the pending 2006 revisions of the RDG. The purpose of the guidelines was to provide “. . . direction to designers, maintenance engineers, and others on the use of crossover protection on depressed medians where the installation of median barrier has not been previ- ously warranted by AASHTO guidance.” (Kentucky 2006) The Kentucky guidelines note that median barrier may be beneficial in cases other than those specified, but specifically, the Kentucky guidance instructs the consideration of median barriers as follows: • All fully controlled-access highways with traversable, depressed medians up to 30 feet wide; • Highways with speeds 55 mph or greater, median widths of 30 to 72 feet, and ADT counts exceeding 40,000; or • For medians of any width or ADT, when at least three CMCs in five years have been observed on the highway section and meeting or exceeding one of these two criteria: – CMC rate exceeding 0.50 CMCs per mile per year; or – Fatal CMC rate exceeding 0.12 CMCs per mile per year. The Kentucky guidelines provide designers “. . . wide latitude when selecting the type(s) of barrier and its location within the median.” The designers are instructed to “. . . select a median barrier type and location which will reach an optimal balance in minimizing the num- ber and severity of collisions, life-cycle and installation costs, and environmental impacts.” (Kentucky 2006) An update to the Kentucky guidelines was published on April 16, 2008. The change between 2006 and 2008 included an update to the CMC warrant portion of the guidelines such that they would “. . . be more representative of Kentucky crash data. . . .” (Kentucky 2008) The crash warrant portion was updated as follows: • 0.35 CMCs of any severity per mile per year. • 0.25 injury or fatal crashes involving CMCs per mile per year. • 0.20 fatal crashes involving CMCs per mile per year. 3.2.2.5 Maryland. The March 2006 “State Highway Administration Guidelines for Traffic Barrier Placement and End Treatment Design” warrants median barriers along expressways and fully controlled-access highways when the criteria shown in Figure 10 are satisfied. When the criteria are not satisfied, “. . . barrier may be warranted due to accident history or by recom- mendation of the [State Highway Administration].” (Maryland 2006) 3.2.2.6 New Jersey. The New Jersey Department of Transportation 2015 Roadway Design median barrier warrants are shown in Figure 11. (NJDOT 2015) These warrants apply to high- speed, access-controlled highways with traversable slopes of 10H:1V or flatter. If consultation of the figure indicates a median barrier is warranted, a barrier should only be installed if one of the following conditions are met: 1. 0.50 CMCs per mile per year of any crash severity, or 2. 0.12 fatal CMCs per mile per year. These study warrants are the same as those used by Caltrans. New Jersey notes that “. . . calcu- lation of conditions 1 and 2 above requires a minimum of three crashes occurring within a five (5) year period.” (NJDOT 2015) The gray shaded optional area warrants a median barrier if there has been a history of cross-median crashes. When a median barrier is warranted, the type used is determined using Table 4. Modified thrie beam median barrier is suggested in place of beam guide rail when there are 12% or more trucks, 12,000 vehicles per lane, and other conditions. (NJDOT 2015) New Jersey adds that if the study location is within 1 mile of an interchange, the median barrier may be warranted at lower traffic volumes, as shown in the cross-hatched section of Figure 11.

Literature Review 19   3.2.2.7 South Dakota. Chapter 10 of the South Dakota Road Design Manual recom- mends a study to determine whether median barriers are warranted when the criteria shown in Figure 12 are met. The study should include a cost/benefit analysis, a review of crashes, or both. (SDDOT 2016) 3.2.2.8 Texas. Bligh et al. developed median barrier recommendations for the State of Texas in 2006. (Bligh 2006) The guidelines developed by Bligh include both a benefit–cost approach as well as a crash history component. The Texas researchers defined a cross-median crash in the same way it was defined by Caltrans: a vehicle had to completely cross the median and strike another vehicle or be struck by another vehicle in the opposing lanes of travel. The warrants were presented on a volume–width graph, and the graph is divided into four zones as shown in Figure 13. When implementing the study, TXDOT simplified the results of the research to those shown in Section 8 of Appendix A of the Texas Roadway Design Manual effective October 1, 2014. (TXDOT 2014b) Section 8 addresses the topic of median barriers. Texas considers concrete barriers or high-tension cable barrier systems appropriate median barriers. “The utilization of other median barriers, such as metal beam guard fence, may be appropriate based on the need to protect point obstacles. . . .” (TXDOT 2014b) Texas recommended guidelines are shown in Figure 14. When “Evaluate Need for Barrier” is indicated by the recommended guidelines, the Texas Roadway Design Manual suggests an engineering analysis be performed that considers the following (TXDOT 2014b): • Type of median (flush, depressed V-ditch or flat-bottom); • Width of the median; Figure 10. Maryland median barrier warrants. (Maryland 2006)

20 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers Figure 11. New Jersey warrants for median barrier for freeways and expressways. (NJDOT 2015) Median Width Median Barrier Type Up to 12 feet Concrete barrier curb (New Jersey-shape) 13 feet to 26 feet Concrete barrier curb (preferred treatment) or beam guide rail, dual-faced or modified thrie beam, dual-faced Above 26 feet Beam guide rail, duel-faced or modified thrie beam, dual-faced Table 4. New Jersey median barrier type selection guidelines. (NJDOT 2015)

Literature Review 21   Figure 12. South Dakota median barrier study warrant. (SDDOT 2016) Figure 13. Median barrier warrants developed for TXDOT. (Bligh 2006) AADT, annual average daily traffic.

22 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers • Traffic volumes, including estimated traffic growth and percentage of trucks; • Types and severity of crashes; • Posted speed limit; • Type of facility, including controlled-access or non-controlled-access with crossovers; • Roadway alignment; • Ramp locations; and • Elimination of barrier gaps. 3.2.2.9 Washington. Before 2001, the Washington State DOT (WSDOT) median barrier policy essentially followed the AASHTO 1977 Barrier Guide recommendations. (AASHTO 1977) In 2001, the WSDOT Design Manual revised its guidance in Chapter 700, Figure 700-7, shown below in Figure 15, such that median barriers were generally warranted at traffic volumes of 20,000 vehicles per day and above when the median was less than 10 m wide. That same year, WSDOT initiated a cost–benefit study to provide guidance on the use of median barriers, particularly cable median barriers, on wider medians. (Olson 2013) That study, in combination with crash test results and pending revisions to the 2006 AASHTO RDG, motivated changes to the 2006 WSDOT Design Manual. (WSDOT 2015) In response to several dramatic crashes, WSDOT initiated a complete review of their median barrier policy and crash history experience in 2007. The review results are summarized in Fig- ure 16 and are structured quite differently than those of the Caltrans crossover crash history warrant since they are a function of the vehicle miles traveled rather than rate per mile per year. (Olson 2013; WSDOT 2007; WSDOT 2009) Section 1600.05 of the WSDOT Design Manual now states: “Provide median barrier on full access control multilane highways with median widths of 50 feet or less and posted speeds of 45 mph or higher. Consider median barrier on highways with wider medians or lower posted speeds when there is a history of cross-median crashes.” (WSDOT 2015) Median barrier war- rants have undergone significant study and updating through the last decade in Washington. Figure 14. Texas recommended guidelines for installing median barriers on high-speed highways. (TXDOT 2014b)

Literature Review 23   Figure 15. 2001 WSDOT median barrier warrants. (WSDOT 2001) Figure 16. WSDOT median crash history study warrant. (WSDOT 2007)

24 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers 3.3 Crash Testing Specifications for Median Barriers NCHRP Report 153: Recommended Procedures for Vehicle Crash Testing of Highway Appurte- nances (Bronstad 1974) was published in 1974 to provide uniform barrier testing procedures and criteria. NCHRP Report 230: Recommended Procedures for the Safety Performance Evaluation of Highway Safety Appurtenances, (Michie 1981) published in 1981, updated NCHRP Report 153 and provided more detailed guidelines for performing and evaluating full-scale vehicle crash tests. Neither NCHRP Report 153 nor NCHRP Report 230 explicitly included performance or test levels. The so-called minimum crash test matrix included small, medium, and large passenger cars. Supplemental tests for heavier vehicles such as utility buses (i.e., school buses), small and large intercity buses, tractor-trailer trucks, and tanker trailer trucks were included in NCHRP Report 230. (Bronstad 1974; Michie 1981) NCHRP Report 239: Multiple Service-Level Highway Bridge Railing Selection Procedures (Bronstad 1981) included four service levels for bridge railings and attempted to establish the service levels from the capacity of the bridge railings based on the NCHRP Report 230 supple- mental tests. (Bronstad 1981) The AASHTO Guide Specification introduced the concept of multiple performance levels for bridge railings. (AASHTO 1989b) NCHRP Report 350 was published in 1993 and expanded the concept of performance levels to the other longitudinal barriers, specifying six different test levels (TLs) for roadside hardware. (Ross 1993) Changes in vehicle fleet characteristics prompted NCHRP Project 22-14(02), “Improved Procedures for Safety-Performance Evaluation of Roadside Features.” (Sicking 2008) NCHRP Project 22-14(02) led to the development of the AASHTO Manual for Assessing Safety Hard- ware (MASH), published in 2009. (AASHTO 2009) MASH includes essentially the same TL approach as NCHRP Report 350 with some changes to vehicle types and impact angles. While NCHRP Report 350 used a small car, a supplemental small car, and a pickup truck to represent the passenger vehicle fleet (i.e., 820C, 700C, 2000P), MASH eliminated one of the small cars and increased the weight of the remaining small car as well as the weight of the pickup truck (i.e., 1100C and 2270P). The Single Unit Truck (SUT) increased in weight (i.e., 8000S to 10000S). The weight of the tractor van-trailer and tractor tank-trailer did not change between NCHRP Report 350 and MASH. (FHWA 2009; Ross 1993) The impact speeds and angles for the length of need (LON) minimum test matrix for longi- tudinal barriers tested under NCHRP Report 350 and MASH are shown in Table 5. The changes between NCHRP Report 350 and MASH are highlighted in a bold-italic font. Notice that the impact speed and angle did not change for the pickup truck, tractor van-trailer, or tractor tank- trailer tests. The impact speed did not change for the small car, but the angle was increased from 20 degrees to 25 degrees to match that of the pickup truck. The impact angle did not change for the SUT, but the speed was increased from 50 mph to 56 mph. (FHWA 2009; Ross 1993) Updates to the MASH longitudinal barrier test matrix for median barriers in a v-ditch were included in the 2016 update to MASH. These updates included crash tests for barriers placed anywhere in the v-ditch and at specific offsets of the v-ditch. An example of the single median barrier placed anywhere in a 4H:1V v-ditch test matrix is shown in Table 6 with placement ref- erenced to the slope breakpoint (SBP) 3.4 Median Barrier Placement 3.4.1 NCHRP Report 711 For NCHRP Report 711: Guidance for the Selection, Use, and Maintenance of Cable Barrier Systems, Marzougui et al. performed vehicle dynamics analyses to develop placement guidelines for cable barrier systems where the top cable is 33 inches or higher and the bottom cable is at

Literature Review 25   TL1 TL2 TL3 TL4 TL5 TL6 NCHRP Report 350 (Ross 1993) Small Car 820C 31 mph/20° 44 mph/20° 62 mph/20° Small Car 700C Pickup 2000P 31 mph/25° 44 mph/25° 62 mph/25° SUT 8000S 50 mph/15° Tractor Van-Trailer 36000V 50 mph/15° Tractor Tank- Trailer 36000T 50 mph/15° MASH Small Car 1100C 31 mph/25° 44 mph/25° 62 mph/25° Pickup Truck 2270P 31 mph/25° 44 mph/25° 62 mph/25° SUT 10000S 56 mph/15° Tractor Van-Trailer 36000V 50 mph/15° Tractor Tank- Trailer 36000T 50 mph/15° NOTE: Content in bold-italic font highlights changes from NCHRP Report 350 to MASH. Table 5. Recommended longitudinal barrier LON impact speed and angle. Vehicle Type Impact Conditions V- Ditch Width (ft) Barrier Position Barrier Location Critical Impact Point Speed (mph) Angle (deg) 2270P 62 25 46 Front slope 12 ft from front SBP 1 ft upstream from post 1100C 62 25 46 Front slope 12 ft from front SBP Midspan location 1100C 62 25 46 Back slope 4 ft from ditch bottom Midspan location 1100C 62 25 46 Back slope 4 ft from back SBP Midspan location 1500A 62 25 46 Front slope Variable Midspan location 2270P 62 25 46 Back slope 8 ft from back SBP 1 ft upstream from post Table 6. Median barrier placed in 4H:1V V-ditch. 21 inches or lower. Marzougui et al. presented the following general conclusions about cable median barrier placement: • “Cable barrier systems should not be placed on slopes steeper than 4H:1V (unless the system has been designed for and successfully crash-tested under these conditions). • Cable barrier systems can be used on 4H:1V or shallower sloped medians or roadsides (6H:1V or shallower sloped medians or roadsides are preferable), provided the placement guidelines listed below are followed.” (Marzougui 2012a)

26 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers A series of placement schematics were included in NCHRP Report 711 showing shaded areas that should be avoided. The symmetric V-shaped median placement schematics are shown in Figure 17; however, Figure 18 should be used if the slopes of the median exceed 6H:1V. The flat-bottom median schematic is shown in Figure 19; however, Figure 20 should be referenced for median slopes exceeding 6H:1V. For non-symmetrical medians, the authors recommended placing the cable system on the shallower slope and following the placement criteria shown for the appropriate median cross section (e.g., v-ditch, flat-bottom). Note that in Figure 17 and Figure 19, the distances are measured from the median centerline out toward the roadway, while in Figure 18 and Figure 20 the distances are measured from the slope breakpoint in toward the median centerline. These placement guidelines were devel- oped to minimize vehicle overrides. It is possible that, for some median widths, the placement guidelines cannot be satisfied, and cable barrier will not be an option when minimizing vehicle overrides is a concern. 3.4.2 NCHRP Project 22-22(02) The draft final report for NCHRP Project 22-22(02), “Effectiveness of Traffic Barriers on Non- Level Terrain” has been submitted. This draft report provides comprehensive guidelines for the placement of common barriers on median slopes to provide adequate safety performance for impacting vehicles. (Bligh 2020a) The preliminary results of NCHRP Project 22-22(02) were obtained and used in this effort. While the results are not shown explicitly within this report, Figure 17. V-shaped and rounded-bottom medians. (Marzougui 2012a) Figure 18. V-shaped medians with slopes steeper than 6H:1V. (Marzougui 2012b)

Literature Review 27   coordination with the NCHRP Project 22-22(02) research team provided the preliminary limits for placement. These results are reflected in the guidance presented herein. 3.5 Run-off-road Crash Modeling Historically, two methods have been used to model run-off-road (ROR) crashes: crash-based methods and encroachment-based methods. Both methods typically use a regression model with either a crash rate or crash frequency as the dependent variable and highway characteristics such as traffic volume, geometrics, and roadside design as the explanatory variables. 3.5.1 Encroachment-Based Probability Models The encroachment-based approach models a series of events from when the vehicle “encroaches” onto the roadside through any subsequent events including a crash. This approach allows for accounting for each event and how many vehicles encroach on the roadside as compared with how many events result in a police-reported crash. This approach has been used extensively in roadside safety policy development, in part due to the ability to capture roadside safety successes (i.e., low-severity, non-reported crashes) and the ability to model design alternatives where crash data have not or cannot be collected (e.g., new barrier designs, different roadside alternatives). Figure 19. Flat-bottom medians. (Marzougui 2012a; Marzougui 2012b) Figure 20. Flat-bottom medians with slopes steeper than 6H:1V. (Marzougui 2012a)

28 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers The 1977 Barrier Guide presented a hand-calculation encroachment-based method based on work by Glennon. (AASHTO 1977) The 1989 RDG expanded the 1977 Barrier Guide and included a computer program called Roadside based on the encroachment probability model. (AASHTO 1989a) As computer applications became more sophisticated and additional research was performed to refine and improve encroachment models, the Roadside Safety Analysis Pro- gram was completed in 2003 and documented in NCHRP Report 492: Roadside Safety Analysis Program (RSAP)—Engineer’s Manual by Mak and Sicking. (Mak 2003) Additional research on measured vehicle trajectories during encroachments and the replacement of severity indices with the equivalent fatal crash cost ratio (EFCCR), as well as continued advancements in com- puters, culminated in the third update of RSAP in 2012, RSAPv3. (Ray 2012a) The need to sepa- rate and document the risk of a crash from the cost–benefit analysis became apparent during the recent economic downturn when crash costs were still increasing while construction costs were decreasing. This update was incorporated into RSAPv3 under NCHRP Project 22-12(03) such that RSAPv3 can be used to assess both risk and benefit–cost. (Ray 2021) 3.5.1.1 RSAPv3 RSAPv3 is a computer program for modeling encroachment-based ROR crashes and using the results of the modeling effort to evaluate alternative design scenarios through either cost– benefit analysis or risk analysis. The encroachment-based model estimates the frequency and severity of roadside crashes for each particular roadside design alternative. For example, vehicles will leave the roadway (i.e., encroach) at various speeds, angles, and orientations; vehicles will leave the road at various points along the road segment and the path taken by the vehicle off the road will depend on driver steering and braking input. Not all vehicles that leave the road, however, will strike an object, so there is a probability distribution associated with the likelihood of striking an object once the vehicle leaves the road. Even when a vehicle does strike an object like a median barrier, the severity of the crash can vary from no injuries to multiple fatalities. Since estimating the frequency and severity of roadside crashes involves several conditional probabilities, the encroachment probability model within RSAPv3 is built on a series of condi- tional probabilities. The conditional probability model takes this form (Ray 2012a): ADT LN,M NE CC P Encr P Cr Encr P Sev Cr E CC Sevs s s∑ ( )( ) ( ) ( ) ( )= •• • • • where E(CC)N,M = Expected annual crash cost on segment N for alternative M. ADT = Average daily traffic in vehicles/day. LN = Length of segment N in miles. P(Encr) = The probability a vehicle will encroach on the segment. P(Cr|Encr) = The probability a crash will occur on the segment given that an encroachment has occurred. P(Sevs|Cr) = The probability that a crash of severity s occurs given that a crash has occurred. E(CCs|Sevs) = The expected crash cost of a crash of severity s in dollars. First, given an encroachment, the crash prediction module assesses if the encroachment would result in a crash, P(Cr|Encr). If a crash is predicted, the severity prediction module esti- mates the severity of the crash, P(Sev|Cr). The severity estimate of each crash is calculated using crash cost values, so the output is in units of dollars. The original version of RSAP estimated the crash costs using a Monte Carlo simulation tech- nique that simulates tens of thousands of encroachments based on a probability distribution of

Literature Review 29   encroachment speed and angle. The probability distribution was calculated using data collected by Cooper in Canada during the late 1970s (i.e., the Cooper data). (Cooper 1980) The frequency and severity of each simulated encroachment are then predicted. Straight-line vehicle trajecto- ries were assumed. RSAPv3 compares field-collected vehicle paths (i.e., trajectories) that include driver inputs such as braking and steering to the location of roadside features for a possible crash. RSAPv3 proceeds by overlaying field-collected encroachment trajectories on the roadside and examining which trajectories strike objects, the probability of penetration or rolling over the object, and the likely severity of those collisions. The passenger vehicle trajectories used in RSAPv3 were gathered from reconstructed ROR crashes under NCHRP Project 17-22. (Sicking 2009b) RSAPv3 provides, for the first time, the ability to explicitly examine cross-median crashes using the encroachment probability model. RSAPv3, however, does not model the probability of observing a crash after the errant vehicle has completely traversed the median; rather, a constant probability is assumed. Adding this module to the encroachment probability model was neces- sary for the development of median barrier guidance under this research effort. Mathematically, the encroachment probability model used in RSAPv3 is conditional. There- fore, a new condition was added. Given a vehicle has encroached into the opposing traffic lanes, the new model assesses the probability of a head-on collision P(HCr|OppEncr) and the subse- quent severity. 3.5.2 Crash-Based Models for Median-Related or Crossover Crashes Considerable effort has been expended on the development of crash-based models to rep- resent median-related and cross-median crashes. Graham et al. recently summarized the literature on this subject in NCHRP Report 790: Factors Contributing to Median Encroach- ments and Cross-Median Crashes. (Graham 2014) Harwood et al. also summarized the literature in NCHRP Report 794: Median Cross-Section Design for Rural Divided Highways. (Harwood 2014) There are research projects nearing completion or recently completed that assembled available crash data on crossover crashes and developed crash-based prediction models from these data. A few of the more germane projects have been highlighted here, including NCHRP Project 17-54, “Consideration of Roadside Features in the Highway Safety Manual.” (Carrigan 2018) 3.5.2.1 NCHRP Report 790 The objectives of NCHRP Project 17-44 were to (1) identify design and operational factors and combinations of factors that contribute to the frequency of median encroachments and cross-median crashes and (2) identify potential countermeasures suitable for addressing these contributing factors. The research culminated in the publication of NCHRP Report 790: Factors Contributing to Median Encroachments and Cross-Median Crashes. (Harwood 2014) Hardwood et al. documented extensive literature on both median encroachments and median crashes dating back to Hutchinson and Kennedy’s study of median encroachments circa 1962. (Hutchinson 1962) The review also included the data collected by Cooper in 1980 on roadside encroachments in Canada. (Cooper 1980) It is important to note that NCHRP Report 790 often uses the terms encroachment and crash interchangeably, which could have profound implica- tions since they define two different, though related, events. Much of the published literature on crash modeling related to medians is, unfortunately, limited to single-vehicle ROR crashes. This assumption has important repercussions. A CMC,

30 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers by definition, must include multiple vehicles. Recall the cross-median crash definition provided in Section 2. Harwood et al. concluded, after a review of crash data for sites with high frequencies of median-related crashes, “. . . that 73 percent of median-related crashes began with a single vehi- cle losing control, while 27 percent resulted from vehicle–vehicle interactions.” (Harwood 2014) Neither the review of past efforts summarized in NCHRP Report 790 nor the past efforts them- selves were neglectful, but are simply a reflection of each authors’ reliance on older, less detailed crash report coding such as single-vehicle crashes as a surrogate for ROR crashes and a different mindset that only crashes where one vehicle is involved can result in an ROR crash. A review of anecdotal crash reports shows that many median encroachments are initiated by vehicle-to- vehicle interactions like avoidance maneuvers, cutting off vehicles during lane changes, braking due to suddenly backed up traffic, etc. When the data analyses are limited to single-vehicle ROR crashes, crashes that started as multi-vehicle crashes are lost. Many of these multi-vehicle-initiated crashes result in vehicles that leave the road at high angles or while yawing, and these less stable trajectories have important implications on the probability of crossing the median. Harwood et al. ultimately recommended a slight variation of the forgiving roadside approach long employed by the RDG to improve median safety, specifically: • “Remove, relocate, or use breakaway design for fixed objects in medians; • Provide barrier to shield objects in medians; • Provide wide medians; • Provide continuous median barrier; • Flatten median slopes; • Provide U-shaped (rather than V-shaped) median cross sections; and • Provide barrier to shield steep slopes in median.” (Harwood 2014) 3.5.2.2 NCHRP Report 794 The objective of NCHRP Project 22-21 was to develop improved guidelines for designing typ- ical median cross sections on new and existing rural divided highways, particularly rural freeways. The research included a review of current literature on median design guidelines and a survey of state practices in median design. The survey was a replica of a survey conducted only a few years earlier under NCHRP Report 790. The additional research culminated in the publication of NCHRP Report 794: Median Cross-Section Design for Rural Divided Highways. (Graham 2014) Graham et al. documented considerable literature on CMC-based models. Graham et al. developed new crash-based prediction models as part of this study. The crash-based modeling effort showed the performance of traversable medians without longitudinal barriers and non- traversable medians with longitudinal barriers (non-traversable medians without longitudinal barriers were not considered). Crash-based prediction models were developed to represent the following, among other crash types: cross-median crashes (CMCs) and cross-median events (CMEs). The models took this form with the coefficients shown in Table 7: = + + +0 1 2 3N eb b lnADT b MW b SR where N = Predicted crash frequency per mile per year. ADT = Average daily traffic volume (vehicle per day). b0, . . ., bn = Regression coefficients determined by model fitting. MW = Median width (feet). SR = Slope ratio of the median (i.e., the horizontal component of the median foreslope). Theoretically, the CMC model divided by the CMC+CME model should approximate the proportion of vehicles that cross the median and are involved in a crash with a vehicle in the

Literature Review 31   opposing lanes as compared with the total number that cross the median. Ideally, the four-lane non-freeway model could be used to develop an encroachment adjustment factor for RSAPv3 because this model most closely resembles the base conditions of the encroachment probability model. Unfortunately, there appears to be a typographical error in the model printed in NCHRP Report 794, as both the CMC+CME and CMC models shown are identical. While it is feasible that the difference between the models is small, developing identical models to represent different outcomes from two different data sets is not likely. 3.5.2.3 Texas Bligh et al. developed crash-based models for medians with and without barriers using Texas crash data from 1998 through 1999. (Bligh 2006) CMCs are not explicitly coded in the Texas crash data; therefore, the analysts searched the data set by vehicle movement and manner of collision to isolate crashes that were apparent CMCs. Using this filtering strategy, models were developed for both CMCs and other median-related crashes. Vehicles that crossed the median but did not have a vehicle-to-vehicle collision were not identified. The authors found the median width, number of lanes, and posted speed limit were significant predictors of a cross-median or median-related crash. The authors also found that ADT was not a significant predictor of crashes for the medians without a barrier; however, ADT was significant for medians with a barrier. These findings could be explained by unreported encroachments and/or low-severity crashes where vehicles enter the median and recover. As noted by Bligh et al. regarding barrier place- ment, “[t]he more lateral offset afforded a driver, the better the opportunity for the driver to regain control of the vehicle in the traversable median and avoid a barrier crash.” (Bligh 2006) 3.5.2.4 Kansas Sicking et al. examined 8,233 crashes occurring between 2002 and 2006 that involved a vehicle entering a median on 761 miles of Kansas freeways without median barrier and widths vary- ing from zero to greater than 90 feet. “An accident involving a vehicle traveling completely across the median and entering opposing lanes was identified as a cross-median event (CME). When a CME resulted in a multiple vehicle collision in the opposing travel way, the accident was classified as a cross-median crash (CMC).” The reviewed cases included 525 CMEs and 115 CMCs. The authors concluded that for 60-foot-wide medians, the relationship of median width to CMEs per 100 million vehicle miles (MVM) is more or less constant at 2.2 CME per 100 MVM for a single-direction traffic volume. The authors found that CMCs per 100 MVM is 6.26 ×10-5 × single-direction ADT. (Sicking 2009a) 3.5.2.5 NCHRP Project 17-54 Carrigan and Ray developed, under NCHRP Project 17-54, crash-based models to represent ROR crash frequency for divided and undivided roadways by the edge. This crash-based modeling Model Highway type Intercept ADT MW SR CMC + CME 4-lane freeway -24.0562 1.9119 0.0000 0.1100 CMC + CME 4-lane non-freeway -21.7518 2.4317 0.0000 -0.5406 CMC + CME 6-lane freeway -20.0770 1.5599 -0.0160 0.1810 CMC 4-lane freeway -29.5036 2.0385 0.0000 0.5523 CMC 4-lane non-freeway -21.7518 2.4317 0.0000 -0.5406 CMC 6-lane freeway -23.1034 1.8886 -0.0226 0.1474 Table 7. Coefficients for CMC and CME models. (Graham 2014)

32 Selection and Placement Guidelines for Test Level 2 Through Test Level 5 Median Barriers effort also developed encroachment adjustment factors for both curves and grades to be used in the encroachment probability model to represent the influence of the variety of horizontal curves or vertical grades on encroachment frequency. These models can be used to represent how many vehicles enter the median and subsequently have a crash on divided roadways and how many vehicles cross the centerline of an undivided roadway and crash on the opposite edge of the roadway. Unfortunately, this effort did not include specifically capturing how many vehicles engaged in a cross-median or head-on collision. (Carrigan 2015b) This large modeling effort can be used to develop encroachment adjustment factors for use with the encroachment probability model. It is common to use crash-based modeling to develop encroachment adjust- ment factors due to the availability of crash data and the lack of new encroachment data. 3.6 Summary Guidelines for installing barriers in the median first appeared in 1964. Changes to the warrant- ing of the barrier based on width and traffic volumes have been accompanied by an evolution in the language used to describe the different regions of each successive warrant (e.g., optional, considered). Subtle changes have occurred to the traffic volumes used and the applicable highway types. Some states adopted the AASHTO guidance directly, but some states have adopted more stringent guidance including “study warrants” based on crash data. There is a good deal of national variety in median barrier warranting that should be considered as this research progresses. The NTSB recommendations to AASHTO and FHWA, following investigations of MRE, can be summarized as: • Define CMCs. • Identify the factors affecting the probability of CMCs. • Establish warrants for median barrier selection that include consideration of heavy vehicles. • Identify CMC rates that call for special consideration when selecting median barriers. The literature review provided some insights into the factors affecting the probability of CMCs. Median width, median slope, the presence and placement of barriers, and highway geo- metrics were found to be significant predictors of MREs. (Bligh 2006; Carrigan 2015a; Graham 2014; Harwood 2014) Many studies have found, however, that traffic volume is not a signifi- cant predictor of MREs. (Bligh 2006; Sicking 2009a) Despite the long-held tradition of relying upon traffic volume when warranting median barriers, there is a well-recognized complicated relationship between encroachment probability and traffic volume. While much-needed new research is underway to gather new encroachment data, this research will rely upon available research, including the Cooper encroachment data. (Cooper 1980; Gabauer, forthcoming-b) There is a great variety of past and existing national and regional guidance for median barriers. There is detailed available literature on both barrier crash testing and in-field evaluations of median performance. The past and present median barrier guidance, the performance of medians and median barriers in the field and the crash testing laboratory, and the detailed NTSB investigations of individual crashes will be valuable contributions toward the success of this research.