Recommended Guidelines for the Selection of Test Levels 2 Through 5 Bridge Railings (2021)

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 According to a Federal Highway Administration (FHWA) sponsored study in the 1980’s, there are about 500,000 bridges in the U.S., about half of them on the National Highway System. [Mak83] Mak also found that the fatal crash rate was about three times higher on bridges than on similar road segments. [Mak83] One of the consequences of Mak’s findings was a steady evolution in the guidelines for the design and testing of bridge railings. In the 1980’s, bridge railings did not have to be crash tested and many bridge railings were found to be structurally inadequate. Persistent research and testing in the past several decades has provided many improved bridge railings with crash-test demonstrated impact performance. The increase in miles of public road and miles of bridges have increased at about the same pace over the last 10 years with bridges consistently remaining approximately 0.40 percent of the total mileage as shown in Figure 1.[FHWA12a] The mileage of urban bridges, however, is increasing at a faster pace than rural bridges (Figure 2) indicating that more bridge rail penetrations in more sensitive urban areas may become more common in the future. [FHWA12a] Figure 1. Mileage of Public Roads and Bridges. 3,940,000 3,960,000 3,980,000 4,000,000 4,020,000 4,040,000 4,060,000 4,080,000 4,100,000 15,000 15,200 15,400 15,600 15,800 16,000 16,200 16,400 16,600 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 M ile s o f P ub lic R oa dw ay s M ile s o f B ri dg es Year Bridges Public Road

4 Figure 2. Miles of Urban and Rural Bridges. Exemplar Crashes The rare occasions when a bridge railing fails to restrain an errant vehicle often results in dramatic crashes. Such crashes have the potential to involve loss of life, the involvement of multiple vehicles, extensive property damage and significant traffic delays. While they do not occur often, when they do occur they nearly always are reported in the news media and demand public attention. The purpose of this next section is to review some bridge railing crashes that have appeared in the media and that have been investigated by the National Transportation Safety Board (NTSB) in order to gain a perspective on both the causes and consequences of bridge railing failures. An appreciation of the causes and consequences will be vital to properly selecting appropriate test levels based on the traffic, and the operational and site conditions of a particular bridge. Crashes in the Media St. Petersburg, Florida, 2001 On January 1st of 2001 a single-unit truck was traveling south on I-275 across a bridge over 54th Avenue South in St. Petersburg, Florida when the truck struck the bridge railing. The impact fractured the concrete railing and the front axle was separated from the truck allowing the axle and pieces of the concrete railing to fall into the bed of a pickup traveling on the roadway below. The axle-less vehicle continued downstream where it straddled the concrete barrier and vaulted over the top, coming to rest in the lanes of 54th Avenue South below where it struck a bus. The driver of the bus was killed and two other people were injured in the crash. Concrete and vehicle debris struck at least three vehicles in addition to the bus. Information from the scene 5,000 5,500 6,000 6,500 7,000 7,500 8,000 8,500 9,000 9,500 10,000 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 M ile s o f B ri dg es Year Rural miles Urban miles

5 suggest the single-unit truck left the roadway on the right and then crossed two lanes before striking the bridge railing on the left. [Alberson04] Wiehlthal Bridge, Germany, 2004 There have even been a few instances where a truck crash caused major structural damage to the bridge requiring the replacement of the bridge itself. Figure 3, for example, shows the result of a truck crash in 2004 on the Wiehlthal Bridge in Germany. On August 26, 2004 a passenger car collided with a fuel tanker truck on the Wiehlthal Bridge on the A4 motorway between Cologne and Olpe, Germany. The fuel tanker truck, which was carrying 8,500 gallons of fuel, penetrated the bridge railing, fell 100 ft. and then burst into flames, killing the driver. The flames burning under the bridge structure caused the steel to deform and lose its load- bearing capacity resulting in the closure of the bridge and the need to completely replace it. Temporary repairs to restore traffic cost the equivalent of $42 million. The total crash cost has been estimated at nearly $400 million; certainly one of the most expensive traffic crashes in history. Figure 3. Tractor-trailer truck which penetrated the Wiehlthal Bridge, Germany. [Wiehltal04] San Francisco, California, 2009 A tractor-trailer truck penetrated the bridge railing on the Bay Bridge between Oakland and San Francisco, California in November 2009.[Zimbio09] The truck fell 200 ft. onto an island, killing the driver. According to news reports, traffic was stopped on this bridge, which

6 carries 250,000 vehicles/day, for over 9 hours. Such long delays affecting such large numbers of people create a significant travel delay cost. These costs are not captured in the usual crash cost data. Boston, Massachusetts, 2007 On Tuesday April 2nd, 2007 a tractor-trailer truck was traveling on the entrance ramp to I-93 south in Charleston Massachusetts. [WBZ07; Globe07; Boston07] The truck struck a 30- inch tall concrete safety shape barrier on a horizontally curved elevated ramp and rolled over the barrier resulting in a 70 foot free fall onto another elevated on-ramp below. The tractor-trailer struck a luminaire and fell on to a sport utility vehicle traveling on the ramp below. While the driver of the sport utility vehicle and the truck driver were both hospitalized with non-life threatening injuries, the situation could have been much worse. The crash happened at off-peak travel hours so there were relatively few vehicles traveling on the ramp below. If the crash had occurred during the peak travel hours, the truck would have fallen onto many more vehicles. Also, the ramp the tractor-trailer truck fell onto was also an elevated ramp that crossed over the passenger rail lines to North Station. With only slightly different impact conditions the truck could have easily struck vehicles on the lower ramp and continued to fall onto the rail line where it may have been struck by a passenger train. The consequences of such a crash could have jeopardized the health and safety of hundreds of people in this highly congested heavily urbanized area especially if the truck had been carrying fuel or a hazardous cargo. This particular crash, unfortunately, was not an isolated event since a very similar crash occurred at essentially the same location only a few months before. The bridge railing at this site was essentially a TL3 F-shape concrete barrier. According to the project design documentation, the railing was originally intended to have a metal top rail which would have made it a TL4 bridge railing. The TL4 railing appears to have been “value engineered” out of the project to save funds. The crash history at this site suggests that a TL3 railing was perhaps not the best choice for this heavily traveled roadway with such potential for catastrophic crashes. Amesbury, Massachusetts, 2011 A fatal bridge railing crash occurred on I-95 in Amesbury, Massachusetts involving a small passenger car in 2011. [Eagle-Tribunne11] While driving northbound on I-95 on the approach to the Whittier Bridge over the Merrimack River in Amesbury, Massachusetts, a 64- year old man lost control of his 2002 Toyota Camry and struck the bridge railing. The vehicle vaulted over the bridge railing and fell 100 feet into the Merrimack River where it sank another 20 feet to the river bottom. An extensive search and rescue effort was required to locate the vehicle and driver but, unfortunately, the driver died as a result of the crash. The bridge railing at this site appears to be a 32-inch tall concrete safety shape. There appear to have been tall snow banks in front of the barrier that may have contributed to the vehicle vaulting over the railing. This crash points out that while heavy vehicles have more energy and are often associated with

7 bridge rail penetrations and rollovers, passenger vehicles can also vault or roll over bridge railings. Avon, Colorado, 2012 Sunday, May 13, 2012 at 9:45 AM in Avon, Colorado, a tandem trailer truck lost control in the I-70 westbound lane and penetrated the Avon Road overpass. One of the trailers landed on a Honda CRV traveling south on Avon Road, crushing the driver's side of the vehicle. Both of the occupants of the Honda CRV were unharmed. However, the truck driver was fatally injured. A witness said “…the rear trailer had become detached and was traveling at a high rate of speed, both airborne and facing backwards when it flew off the highway and landed on the CRV. Moments later, the cab and trailer to which it was attached also tumbled off the overpass” leading to an explosion and fire. Figure 4 shows the W-beam rail in place at the time of the crash. [Avon12] Figure 4. Damaged W-beam bridge rail and final position of truck on local road under bridge, Avon, CO. [Avon12] Syracuse, New York, 2012 On July 22, 2012, the front cab of a tractor-trailer penetrated a concrete bridge rail in Syracuse, New York and dropped 25 feet. The two occupants of the cab were injured, but responsive after the crash. The tractor-trailer was traveling north on Interstate 81 and penetrated the rail between Erie Boulevard East and Water Street. The frame, motor and the axle remained hanging from the bridge. About 100 gallons of diesel fuel was spilled.[Syracuse12]

8 Montreal, Quebec, 2011 On Wednesday, December 28th, 2011 a pickup truck traveling west on the Sainte Anne de Bellevue Road near the entrance to Highway 20 lost control and left the roadway. The vehicle flipped off an overpass and fell onto railway tracks, where it was then hit by a train. Both occupants of the pickup truck died in the crash. Police speculated poor visibility and slippery roads may have played a role. [CBC01] The section of roadway that runs over the railroad tracks had a W-beam guardrail as the only barrier, however, after the crash occurred, Transport Quebec installed a concrete barrier at the location. Avellino, Italy, 2013 On July 29th, 2013, a bus carrying around 50 people hit several vehicles before penetrating a bridge rail and falling 98 ft. into a gorge near Avellino, Italy. At least 38 people were killed in the crash, with an additional 10 people injured. [BBC01] The rail type is unknown. Beaverton, Oregon, 2012 On Saturday, November 24, 2012 in Beaverton, Oregon a crash resulted in a pickup truck hanging from the SW Denney bridge railing over HWY 217. The driver was the only occupant . Fire and rescue personnel responded and secured the truck to a fire engine to keep the truck from falling off the bridge. A fire engine equipped with a basket was used to get the driver out of the truck safely. [KATU01] The bridge railing appears to be a PL2 system named the “Foothills Parkway Bridge Railing,” approved under AASHTO Guide Specification for Bridge Railings. The railing as it appeared before the impact is shown in Figure 5. Figure 5. Beaverton Bridge Rail Before Crash, Beaverton, OR. [Google Earth]

9 Bronx, New York, 2012 On April 29th, 2012, a white 2004 a white Honda Pilot minivan “…was traveling southbound at about 70 mph in a 50-mph zone of the Bronx River Parkway when the 12:30 p.m accident occurred.” [NYDN01] The minivan lost control while driving southbound on the Bronx River Parkway and hit the single-slope median barrier. After striking the barrier, the minivan veered to the right, crossed three lanes of traffic, struck the curb on the right side of the road and vaulted over an older style metal post and tube bridge rail. “The car never even touched the 4- foot-high iron guardrail, though it left traces of motor oil on the rail as it sailed over it.” There were no skid marks found on the road before the first impact, however, there were skid marks leading up to where the van left the bridge. [NYDN01] The vehicle fell 60ft off of the bridge, landing on its roof in a wooded area. All seven passengers, including three children, were killed instantly. [CNN01] Less than a year earlier, in the same section of the Bronx River Parkway, a second crash occurred. On June 4, 2011, two people were uninjured after the driver lost control of the vehicle, “… flew off a parkway overpass, and landed 22 feet below in the parking lot of a police station. The media report the 2005 Acura was speeding while traveling north in the left lane before it slammed into a concrete median barrier. “Then, it flew across both lanes of traffic, crashing through a guard rail and a chain link fence before partially landing on an officer's GMC pickup truck in the parking lot of the NYPD 12th district station in Morris Park.” [WPIX01] A picture of the approximate location of both crashes can be seen in Figure 6. Figure 6. Bronx River Parkway, near crash site, Bronx, NY. [GoogleEarth].

10 Grand Prairie, Texas, 2013 On August 3rd, 2013 a tractor-trailer was traveling westbound on Interstate 30 in Grand Prairie, Texas when the truck veered off the right shoulder. The truck penetrated the guardrail, then struck the bridge abutment and continued down the embankment in between the guardrail and abutment until it crashed down onto State Highway 161 below. [KVUE01] The truck landed on State Highway 161 and hit a single-slope concrete barrier head on, penetrating the barrier when it burst into flames, narrowly missing the I-30 westbound bridge pier. The driver died in the crash. There were no other injuries. While this crash appears not to be technically a bridge railing crash since the vehicle penetrated the approach guardrail prior to the bridge, it does point out the importance of guardrail-bridge railing transitions. In particular, typical W-beam guardrails are TL3 devices whereas bridge railings may be TL3, TL4 or TL5. This crash raises the question of how far in advance of the bridge railing a higher test level barrier may be needed to prevent this type of catastrophic heavy vehicle crash. Boston, Massachusetts, 2013 On August 6, 2013, a beer delivery truck partially penetrated the bridge rail on I-93 North resulting in the cab dangling 3 feet beyond the bridge rail. The driver and the passenger were uninjured. Traffic on I-93 was significantly slowed, with one commuter who was stuck in traffic noting: “I’ve gone about 100 feet in an hour.” [WBZ01] This crash also disrupted commuter traffic on the Orange Massachusetts Bay Transportation Authority (MBTA) subway line since the commuter rail parallels the highway. Less than a week later and only a few feet away, on August 9, 2013 a Cadillac struck a single-unit truck on Interstate 93 northbound in Boston, Massachusetts sending the single-unit truck into the bridge rail. The truck vaulted over the rail and landed on the Exit 26 southbound ramp. The rail type can be seen in Figure 7. Despite falling approximately 40 ft. before crashing onto the roadway below, the driver of the truck sustained only minor injuries. [Patriot01, WCVB01]

11 Figure 7. Close-Up of Bridge Rail Type (upper left) and Exit 26 Ramp Truck, Boston, MA. [Google Earth] Buellton, California, 2012 On January 12, 2012 on Highway 101 in Buellton, California, a trailer truck drifted out of its lane and sideswiped a car in the lane next to it. [NBC01] After repeatedly pushing the passenger car into the bridge rail, the wheels of the trailer ran over the car and broke through the rail. The truck and trailer fell down into the creek below and burst into flames, killing the driver. The passenger car was left hanging off the bridge, still containing the driver and two children. Fortunately, a unit of Navy Seabees was on the adjacent bridge transporting a heavy-duty forklift which they used to stabilize the vehicle while emergency personnel extracted the passengers. The driver and one child suffered serious injuries while the second child was unharmed.[MOUK01] Galesburg, Illinois, 2013 On September 7, 2013, an SUV was involved in a single-vehicle run-off road crash. The SUV failed to negotiate a curve on U.S. Route 150 on the Gates Bridge. The vehicle went over the guardrail, landed on the railroad tracks below the bridge, and came to rest in a wooded area on the south embankment. One person who was partially ejected from the vehicle was killed. Another passenger was injured in the crash, but the extent of the injuries was not released. [PJSTAR01] A Google Earth image of the pre-crash area is located in Figure 8.

12 Figure 8. Gates Bridge, Galesburg, IL. [Google Earth] Williamsburg, Kansas, 2012 A 17-year-old boy with a provisional license was driving a Freightliner motor home that was pulling a trailer on Interstate 35 when it went off the road, struck a guardrail and crashed through a bridge rail and fell into a ravine on April 1, 2012. It is not clear from news reports if the guardrail was penetrated first or if the bridge railing was penetrated after the guardrail redirected the vehicle. Of the 18 passengers, five were killed and the remaining 13 were all injured, two critically. [MOUK02] On March 23, 2013, it was announced that the NTSB would be taking over the investigation of this crash, including looking into laws that allowed the 17-year-old boy with a provisional license to drive the 57,000 lbs vehicle. Crashes Investigated by the NTSB The NTSB investigates and determines the probable cause of “significant crashes” on highways and other modes of transportation with the goal of promoting transportation safety and preventing future similar crashes. In total, NTSB investigates approximately six highway crashes per year, each investigation lasting approximately 20 months. NTSB crash investigation teams vary in size from three or four to more than 12 specialists who routinely handle investigations within their specialized field (i.e., rail, highway, marine and pipeline). Highway crash teams include specialists with backgrounds such as a truck or bus mechanical expert, a highway engineer, a weather specialist, a human performance specialist, and survival factors specialist. The team is led by an Investigator-in-Charge. The following sections contain brief summaries of crashes involving bridge railings which have been investigated by the NTSB since the mid-1970’s. While from a statistical point of view, these crashes are anecdotal, they do serve to point out important features of catastrophic crashes involving bridge railings.

13 Fort Sumner, New Mexico, 1972 On December 26, 1972, a school bus transporting 34 people was traveling westbound while a tractor-trailer truck transporting cattle was eastbound on US-60 near Fort Sumner, New Mexico. [NTSB74] As the truck approached a narrow bridge, the driver swerved to the right after seeing approaching headlights that appeared to be on his side of the road. The truck struck a crash cushion at the entrance to the bridge and the right-rear wheel of the trailer mounted the curb on the bridge. The tractor “snagged” the bridge railing and rotated, mounting the curb and causing the trailer to jackknife. The bus collided with the jackknifed trailer in the westbound lanes. Nineteen people in the bus were killed and 15 others sustained a variety of injuries. As a result of this crash, the NTSB recommended that the FHWA “expedite a program to improve, where feasible, substandard bridge rail systems on existing bridges to increase resistance to pocketing or penetration by impacting vehicles of all classes and redirect those vehicles. Research, including crash testing, should also be expedited to develop criteria for mandatory standards for bridge rail and guardrail designs for new bridge construction (H-74-7).” [NTSB74] Nashville, Tennessee, 1973 On July 27, 1973 while traveling through a morning fog a car carrying nine people penetrated the bridge railing on the Silliman Evans Bridge in Nashville, Tennessee and fell 65 feet to the ground below. Seven passengers and the driver were killed. The barrier on the bridge was apparently a type of box-beam barrier mounted on a nine-inch curb. As a result of this crash the NTSB recommended that the FHWA “establish national performance standards, including dynamic testing procedures, for bridge rail systems. Such standards should extend performance criteria to include impacts by heavy vehicles and should improve performance characteristics for impacts by all classes of vehicles. The establishment of these standards should be of high priority and compliance should be mandatory for all new bridge rail systems used on public roadways (H-74-18).” [NTSB74] Siloam, North Carolina, 1975 In the morning of February 23, 1975, an automobile was traveling in a heavy fog when it penetrated a timber bridge railing and struck a structural member on the Yadkin River Bridge near Siloam, North Carolina. [NTSB76] The bridge was a through-truss bridge so the penetration of the bridge railing allowed the vehicle to damage a vital structural component of the bridge. As a result of the collision, the bridge collapsed and fell into the river. Six additional vehicles drove into the river in the next 17 minutes resulting in four people being killed and 16 others being injured. As a result of this crash the NTSB recommended that the FHWA develop and publish “guidelines for the structural retrofit of bridge railings on existing bridge structures to protect vital structural members from impact by vehicles.” [NTSB76] Martinez, California, 1976 On May 21st, 1976 a school bus was traveling on I-680 near Martinez, California with 52 people onboard when it struck a bridge railing on an off-ramp. The bus rolled over the bridge

14 railing of the curved bridge, landing on its roof. Twenty nine people were fatally injured in the crash. [NTSB77a] The bridge railing and the integrated curb were cited by the NTSB as one of the contributing factors to the crash. As a result of this crash, the NTSB made three recommendations that dealt with various aspects of bridge railing design and placement. The three recommendations were: H-77-12: “Develop bridge railing designs that will meet performance standards to be established by the FHWA for various classes of vehicles and that will be sufficient in number to meet the various state requirements with regard to climatic and other physical conditions that affect the operation and maintenance of a roadway system. Such bridge barrier railing designs should be available to states that do not desire to develop their own designs in accordance with mandatory performance standards issued by the FHWA.” [NTSB77b] H-77-13: “Investigate through dynamic crash testing and analytical procedures the effects of various geometric configurations and adjacent roadway surfaces on the performance of traffic barrier rail systems. The investigation should also consider how maintenance practices or the lack of maintenance affects the performance of the barrier rail systems.” [NTSB77c] H-77-14: “In cooperation with the states, establish priority guidelines for improving, through modification or retrofit, the performance of existing traffic barrier rail systems at bridges. Consideration should be given in the priority guidelines to the potential for multi- fatality accidents involving high occupancy vehicles such as buses. [NTSB77d] Houston, Texas 1976 A tractor-trailer truck hauling over 7,500 gallons of anhydrous ammonia was traveling on an elevated ramp between I-610 and US-59 in Houston, Texas on May 11, 1976. [NTSB77e] At this location, US-59 is at the ground level and I-610 is elevated over it. The ramp between I-610 and US-59 passes between I-610 above and US-59 on the ground below. While negotiating the ramp, the tractor-trailer truck began to roll due to the horizontal curvature of the ramp and the truck’s speed. The truck penetrated the bridge railing and fell about 15 feet onto US-59. As the truck fell it also struck and sheared off a column supporting the elevated portion of I-610 above. The truck and trailer became detached during the crash and the trailer broke into several parts allowing the rapid escape of the ammonia into the atmosphere. Twelve automobiles were damaged by flying debris from the tractor and trailer as it crashed onto US-59. The driver of the truck was fatally injured in the crash and five people were killed and 178 people injured due to breathing the ammonia gas that escaped from the ruptured trailer. I-610 was designated in 1970 as a hazardous materials route by the City of Houston and all vehicles transporting hazardous materials through the city were restricted to this route. I-610 is an elevated five-lane highway near the crash site. The ramp onto US-59 where the crash occurred consists of two lanes arranged in an interconnected 3, six and 12 degree compound curve; the crash occurred on the third curved section (i.e., the 12 degree curve). The bridge railing was an oval pipe section 33.5-inches high, mounted on five-inch wide, ¾-inch thick steel bars. The support bars were bolted behind a 14-inch wide, 12-inch tall curb. The crash destroyed

15 94 feet of bridge railing, caused damage to the bridge deck, a column supporting the I-610 overpass was sheared off and guardrails on US-59 were damaged. As a result of this crash, the NTSB recommended that the FHWA “in consultation with state and local governments, establish highway design criteria for the selection, location and placement of traffic barrier systems that will redirect and prevent penetration when struck by heavy vehicles. The criteria for preventing vehicle penetration should consider the human exposure to injury and the effects of hazardous cargo that could result from barrier penetration (H-77-5).” [NTSB77e] Elkridge, Maryland, 2004 On January 13, 2004 a tractor tanker-trailer truck was hauling 8,800 gallons of gasoline southbound on I-895 near Elkridge, Maryland. [NTSB09a] As the tanker truck approached the curved and elevated I-95 overpass it entered the right shoulder and struck the guardrail and continued on to strike the attached bridge railing. The truck and trailer mounted and vaulted over the bridge railing falling 30 feet onto the northbound lanes and median of I-95. The NTSB estimated the vehicle’s speed did not exceed 49 mi/hr. An explosion and large fire resulted from the tanker truck striking the ground and four vehicles traveling northbound on I-95 drove into the conflagration. Four of the five vehicle operators were fatally injured in the crash. Figure 9. Hazardous material truck crash near Houston, Texas in 1976. [NTSB77e]

16 The barrier on the I-95 overpass, shown in Figure 10, was a 32-inch tall concrete safety shape bridge railing installed adjacent to a four-ft. shoulder on the overpass. The overpass was curved to the left which promoted the vehicle rolling over the barrier while the driver was trying to regain control by steering to the left. The guardrail transition to the bridge rail may also have been a factor in the crash. While the NTSB was not critical of the choice of the bridge railing at this location, the crash site did involve some of the risk factors cited by the RDG for higher test level bridge railings. The horizontal curve had a radius of about 954 feet. I-95, the road the bridge crossed over, had an average daily traffic (ADT) of 189,750 in 2004. I-895 had an ADT of 13,350 in 2004 with 5.5 percent single-unit truck classes 4 through 7 and 3.2 percent combination trucks classes 8 through 13 for a total percent trucks of 8.7. Figure 10. Crash site in Elkridge, MD where a fuel truck penetrated a concrete bridge railing. [NTSB09a] The 32-inch tall concrete safety shape is probably the most common Report 350 TL4 bridge railing in use today. The curved alignment of the roadway, the hazardous material being transported and the fact that the overpass was crossing a very heavily traveled and important interstate magnified the importance of the bridge railing at this particular location. Huntsville, Alabama, 2006 On November 20, 2006 at about 10 a.m. a school bus with 40 students onboard was traveling westbound in the left lane of an elevated ramp portion of I-565 in Huntsville, Alabama. [NTSB09b] A 1990 Toyota Celica was following the bus and apparently moved into the right- hand lane and accelerated in order to pass the bus on the right. As the Toyota was abreast of the bus it began to “fishtail” and the driver lost control, veering to the left and striking the right-front tire of the school bus. Both vehicles swerved to the left and struck a 32-inch tall concrete bridge railing on the left side of the ramp. The school bus climbed up onto the bridge railing and traveled about 117 feet before completely rolling over the railing and falling about 30 feet below onto a dirt and grass area underneath the ramp shown in Figure 11. The crash resulted in four

17 fatalities, 17 serious injuries, 17 minor injuries and three bus occupants were uninjured. The bus driver was ejected in the initial crash and four passengers were either fully or partially ejected when the bus struck the ground below. The Toyota did not penetrate the bridge railing and came to rest against the bridge railing. The driver and passengers of the Toyota were not injured in the crash. Figure 11. Final rest position of a school bus that penetrated a concrete bridge railing near Huntsville, AL in 2006. [NTSB09b] With respect to highway design issues, the NTSB noted that the bridge railing as a 32- inch high Report 350 TL4 concrete safety shape installed adjacent to a four-foot left shoulder. The bus was traveling no more than 55 mi/hr, it struck the railing at 9-10 degrees and its gross empty weight was 17,700 lbs so the impact conditions were not extraordinary in comparison to the standard Report 350 TL4 test (i.e., 18,000-lbs single unit truck striking the barrier at 15 degrees and 50 mi/hr). The NTSB concluded that the Toyota restricted the bus from moving back into its lane and essentially held the front of the bus to the railing until it eventually rolled over it. Sherman, Texas, 2008 On August 8th, 2008 at about 12:45 a.m. a motorcoach with 55 passengers and a driver were traveling at about 68 mi/hr northbound in the right-hand lane of the four-lane US 75 near Sherman, Texas. [NTSB09c] As the motorcoach approached Post Oak Creek its right steer axle failed and the motorcoach struck a seven-inch high curb at about a four-degree impact angle which it overrode and then struck a steel bridge railing. The motorcoach struck the railing at about 44 mi/hr and then slid along the railing for about 120 feet until it penetrated the bridge railing and fell about 8 feet onto the creek embankment below. Seventeen passengers were fatally injured, the driver was seriously injured and 38 passengers received minor to serious injuries in the crash.

18 Figure 12. Site of a motorcoach bus crash in Sherman, TX, 2008. [NTSB09c] US-75 in the area of the crash had a traffic volume of about 47,000 vehicles/day in 2006 and commercial vehicles accounted for 16 percent of the total traffic volume. The bridge railing at the crash site, shown in Figure 12, was a 27-inch tall steel beam and post system side-mounted on an 18-inch wide, seven-inch tall curb adjacent to a 22-inch wide shoulder. The bridge railing was 279-ft long. The bridge railing was a Texas Type II railing which was originally designed in 1954 in accordance with the AASHTO Bridge Design Specifications in effect at the time. Apparently, this bridge railing had been struck previously in 2001 by a tractor-trailer truck. It had penetrated the bridge railing causing some damage to the railing anchorages in the deck. Based on its height alone this bridge railing would be classified today as no more than a TL3 railing but it is likely that it was never crash tested so its impact performance is doubtful. As a result of the Sherman, Texas motorcoach crash, NTSB issued three safety recommendations dealing with the design and warranting of bridge railings. H-09-17: “Establish, in conjunction with the American Association of State Highway and Transportation Officials, performance and selection guidelines for bridge owners to use to develop objective warrants for high-performance Test Level Four, Five, and Six bridge railings applicable to new construction and rehabilitation projects where railing replacement is determined to be appropriate.”

19 H-09-25: “Work with the FHWA to establish performance and selection guidelines for bridge owners to use to develop objective warrants for high-performance Test Level Four, Five, and Six bridge railings applicable to new construction and rehabilitation projects where railing replacement is determined to be appropriate, and include the guidelines in the Load and Resistance Factor Design (LRFD) Bridge Design Specifications.” H-09-26: Revise Section 13 of the LRFD Bridge Design Specifications to state that bridge owners shall develop objective warrants for the selection and use of high-performance Test Level Four, Five, and Six bridge railings applicable to new construction and rehabilitation projects where railing replacement is determined to be appropriate. NTSB Recommendations The NTSB first made recommendations on the design and selection of bridge railings in 1977 in its recommendations H-77-12 through 14 as a result of the Martinez, California crash discussed earlier. In 1980, the NTSB issued SEE-80-5 to, in part, assess FHWA’s efforts in implementing the H-77-12 through 14 recommendations. SEE-80-5 contained an additional recommendation (i.e., H-80-64) which stated: H-80-64: “Establish mandatory performance standards, and associated test procedures to be used in determining compliance, for all traffic barriers constructed on Federal-aid roads after January 1, 1982. The performance standards should first address automobiles and should be expanded for heavier passenger vehicles and trucks as research is completed to provide needed information.” [NTSB80] The result of these recommendations was an era of vigorous re-design and crash testing of bridge railings by the FHWA. In the 10 year period following H-80-64, 74 bridge railings were crash tested and accepted including W-Beam bridge railing, Thrie-Beam bridge railing, Metal Tube bridge railing, Vertical Concrete parapet, New Jersey Barrier bridge railing, Tall Wall type, F-Shape Concrete Barrier bridge railing and timber bridge railings. The FHWA published three Technical Memoranda where the states were advised of the requirements to crash test bridge railings, provided a list of the bridge railing designs that had been crash tested, and provided a list of all FHWA-accepted bridge railings. [FHWA86; FHWA90; FHWA97a] Based on the results of this decade of research, the FHWA and AASHTO developed bridge railing design and crash test criteria which appeared as the 1989 Guide Specification to Bridge Railings which will be discussed in a later section.[AASHTO89] The recommendation was “closed” by the NTSB and considered superseded by AASHTO’s adoption of the Guide Specification. [AASHTO89] Unfortunately, while the FHWA’s design and crash testing research undertaken in the 1980’s and the AASHTO Guide Specification made great advances in the design and testing of bridge railings, the warranting or selection of which types of bridge railings are best suited to which conditions was still very subjective. Recently, the FHWA added “Identification of Potentially Deficient Systems” and “Bridge Rail Retrofits” to their website. This website suggests that bridge railings designed prior to 1964 may not meet current specifications and includes a list of items to evaluate (e.g., base plate connections, anchor bolts, material brittleness,

20 welding details, and reinforcement development). Additionally, the FHWA suggests that open- faced railings present a snagging hazard and that curbs or walkways in front of the bridge railings also present hazards. Suggestions for retrofitting any outdated bridge railings are included and discussed (e.g., concrete retrofit is economical if the structure can carry the added load; W-beam/thrie-beam retrofits provide an inexpensive, short-term solution; and metal post and beam retrofits for structures with walkways).[FHWA10] This website does not address the techniques one may use to conduct the investigation, how one may prioritize the need to retrofit deficient bridge railings or the identification of deficient bridge rails built after 1964. Crash Testing Crash testing is the most direct means of assessing barrier impact performance. Containment capacity is a function of the strength and height of a barrier. If a barrier is not strong enough, an impacting vehicle can penetrate through it. If a barrier is not tall enough, an impacting vehicle can override or roll over it. Full-scale crash testing is typically used to verify containment capacity of a barrier for a selected test level but testing guidelines do not indicate what traffic, geometric and operational characteristics should be used to assess the risk of a particular crash type occurring. Hundreds of crash tests have been performed in the past several decades, many involving bridge railing designs for heavy vehicles. The purpose of this section is not to list or catalog all the various types of bridge railings that have been crash tested but, rather, to point out the different performance and test levels that have been used over the years to evaluate crash test performance. NCHRP Report 230 NCHRP Report 230 was published in 1981 to provide guidelines for performing and evaluating full-scale vehicle crash tests of a variety of safety appurtenances including bridge railings. [Michie81] Report 230 did not explicitly include a multiple performance level or test level approach although it did include a number of 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. The so-called “minimum” crash test matrix included small, medium and large passenger cars. The supplemental tests recommended in Report 230 were developed in conjunction with another NCHRP project (i.e., NCHRP Report 239) which presented recommendations for a multiple service level approach to bridge railing design. The supplemental tests were intended to satisfy the recommendations of Report 239 with respect to both lower containment bridge railings used on lower volume, low speed bridges and higher containment bridge railings intended for use in situations where bus and truck impacts would be more likely and more serious. Report 239 included four service levels and attempted to establish the service levels based on the capacity of the bridge railings. The Report 230 supplemental tests were the means used to determine the bridge railing capacity.

21 1989 AASHTO Guide Specification for Bridge Railings The 1989 AASHTO Guide Specification for Bridge Railings (GSBR) recommended that all bridge railings should be evaluated in full-scale crash tests to verify that a given bridge rail design meets the desired impact performance criteria. [AASHTO89] Three bridge rail performance levels and associated crash tests were recommended to access the performance of the bridge railings. The crash test matrices for each performance level were described by crash test conditions defined in terms of vehicle type, vehicle weight, impact speed, and impact angle. Two passenger vehicles – a small 1800-lb passenger car and a 5400-lb pickup truck - were common to all three performance levels. Test conditions associated with Performance Level 1 (PL1) included the 5,400-lb pickup truck impacting at a speed of 45 mph and an angle of 20 degrees. For PL2, the speed of the pickup truck test was increased to 60 mph and a test with an 18,000 lb single-unit truck impacting the barrier at a speed of 50 mph and an angle of 15 degrees was added to the test matrix. The highest performance level, PL3, incorporated a test with a 50,000 lb van-type tractor-trailer impacting the barrier at a speed of 50 mph and an angle of 15 degrees. While crash testing procedures for bridge rails would quickly be wrapped into NCHRP Report 350 in 1993, only 4 years after the AASHTO GSBR was published, the concept of multiple performance levels was introduced for the first time in an AASHTO guide by the GSBR and this concept was retained and even expanded in Report 350. Prior to 1989, there were a few higher performance bridge railings available but the publication of the 1989 GSBR and the later inclusion of bridge railings in Report 350 has resulted in a large number of bridge railings being designed and crash tested. The AASHTO- ARTBA-AGC Task Force 13 on-line bridge railing guide today (i.e., 2013) shows 117 crash- tested bridge railings; 59 of which are essentially PL1 bridge railings; 48 PL2 bridge railings and 10 PL3 bridge railings. Today there are many crash tested FHWA-accepted bridge railings with demonstrated performance across a wide spectrum of test levels that were not available in 1989. Although the 1989 GSBR recommended crash testing as the basis for bridge rail evaluation and acceptance, it did provide the bridge engineer with suggested design information including the magnitude, distribution, and vertical location of railing design loads for each performance level. The transverse loads were derived from two related research studies in which vehicle impact forces were measured using instrumented concrete walls. [Noel81; Beason89] In the first study, an instrumented concrete wall (shown in Figure 13) was designed to, for the first time, measure the magnitude and location of vehicle impact forces. [Noel81] The wall consisted of four 10-ft long concrete panels, each laterally supported by four load cells. Each of the 42-in tall, 24-in thick panels was also instrumented with an accelerometer to account for inertia effects. Surfaces in contact with the supporting foundation and adjacent panels were Teflon coated to minimize friction. Eight full-scale crash tests were conducted using various sizes of passenger cars and buses ranging from an 1,800-lb sedan to a 32,020-lb intercity bus. In the second such study, a new wall with a height of 90 inches was constructed using similar design details. [Beason89] Three full-scale vehicle crash tests were performed with tractor-trailer vehicles ranging in weight from 50,000 lbs to 80,000 lbs.

22 Figure 13. Instrumented crash wall. [Noel81] The data from the instrumented wall tests were analyzed to determine the resultant magnitudes, locations, and distributions of the contact forces. Maximum forces were obtained by averaging the data over 50 msec intervals to reduce the effect of force “spikes” in the data that were believed to have little consequence to the required structural integrity of the bridge railings due to their short duration. The force measurements were obtained from a nearly rigid barrier and, therefore, were considered to represent the upper bound of forces that would be expected on an actual bridge railing. Any deformation of the bridge rail during impact would tend to reduce the magnitude of the impact forces below those obtained on the “nearly rigid” instrumented concrete wall. NCHRP Report 350 In 1993, NCHRP Report 350 was published superseding the previous crash testing guidelines contained NCHRP Report 230. [Ross93] One major change in Report 350 is that six different test levels for roadside hardware were added for roadside hardware in general and bridge railings in particular. The intent was to provide test guidelines for developing a range of bridge railings that could be used in different situations. Test levels 1 through 3 are focused on the impact performance of passenger vehicles (e.g., small passenger cars and pickup trucks) and vary by impact speed, with increasing impact speeds defined for increasing test levels. The base test level for longitudinal barriers, including bridge railings, to be used on the National Highway System (NHS) is Test Level 3 (TL3). The structural adequacy test for this test level consists of a 2,000 kg (4,409 lb) pickup truck impacting a barrier at 100 km/h (62 mph) and 25 degrees. Test levels 4 through 6 also include consideration of passenger vehicles, but additionally incorporate consideration of various sizes of trucks. Many state transportation departments require that their bridge railings meet Report 350 TL4. The impact conditions for NCHRP Report 350 TL4 involve an 8,000 kg (17,637 lb) single-unit truck impacting the barrier at 80

23 km/h (50 mph) and 15 degrees. These impact conditions are similar to those associated with Performance Limit 2 (PL2) in the 1989 AASHTO “Guide Specification for Bridge Rails.” [AASHTO89] A TL5 test involves an 80,000-lb (36,000-kg) van-type tractor-trailer impacting the barrier at a speed of 50 mph (80 km/hr) and an angle of 15 degrees. Test Level 6 (TL6) uses the same impact conditions, but incorporates an 80,000-lb (36,000-kg) tractor-tank trailer. Barriers meeting these higher containment levels are used when the owner-agency considers that site conditions warrant the added expense. Site specific factors that might justify use of a high- containment barrier include a high percentage of heavy truck traffic or truck related crashes and/or an unusually high risk associated with barrier penetration. Such barriers are necessarily taller, stronger, and more expensive to construct. The higher test levels were intended for locations where there were a high percentage of trucks and where the consequences of trucks penetrating or rolling over the bridge railing would be severe. While Report 350 provided the testing recommendations, only general guidance was provided about what field conditions would indicate the need for a higher test level bridge railing. FHWA has established approximate equivalences between the Report 239 multiple service levels, the three AASHTO GSBR performance levels and the Report 350 test levels in a memorandum to FHWA Regional Administrators in 1997. [Horne97] The equivalencies set out by the FHWA are summarized in Table 1, but the guidance from FHWA simply established crash test equivalencies without providing any guidance on when and where to use each type of bridge railing. Table 1. Crash Test Acceptance Equivalencies from the FHWA. [Horne97] Bridge Railing Testing Criteria Acceptance Equivalencies Report 350 TL1 TL2 TL3 TL4 TL5 TL6 Report 230 MSL-1 MSL-2† AASHTO Guide Spec PL1 PL2 PL3 AASHTO LRFD Bridge Spec PL1 PL2 PL3 † This is the performance level usually cited when describing a barrier tested under NCHRP Report 230. It is close to TL3 but adequate TL3 performance cannot be assured without a pickup truck test.

24 Manual for Assessing Safety Hardware (MASH) Since the publication of NCHRP Report 350 in 1993, changes have occurred in vehicle fleet characteristics, operating conditions, technology, etc. NCHRP Project 22-14(2), "Improvement of Procedures for the Safety-Performance Evaluation of Roadside Features," was initiated to take the next step in the continued advancement and evolution of roadside safety testing and evaluation. The results of that research effort culminated in the 2009 AASHTO MASH which superseded Report 350. [AASHTO09] MASH includes essentially the same test level approach with some changes to the impact conditions for the higher longitudinal barrier test levels. Barrier performance levels identified in MASH were modified from its predecessor, NCHRP Report 350. [AASHTO09; Ross93] These modifications were primarily related to the size of the test vehicles. Impact conditions associated six test levels tend to be calibrated off of impact conditions associated with Test Level 3 (TL3). TL3 is intended to represent barrier applications on typical high speed high-volume roadways since TL3 is the “default” test level used on the NHS. Impact speeds and angles for TL3 have traditionally been selected to be equal to the presumed 85th percentile impact speed and 85th percentile impact angle from ran off road crashes. Further, vehicle masses are normally selected to be equal to the 95th and 5th percentile values from the passenger car fleet. However, in recognition of the recent increase in the size of passenger vehicles and the expectation that high gasoline prices might push vehicle masses down, the light truck vehicle mass was reduced to the 90th percentile and the small car mass was reduced to the 2nd percentile of the 2002 new vehicle fleet. Even with these adjustments, the severity of the TL3 test condition was increased significantly. The weight and body style of the pickup truck test vehicle changed from a 2,000 kg (4,409 lb), ¾-ton, standard cab pickup to a 2,270 kg (5,000 lb), ½-ton, 4-door pickup. This change in vehicle mass of approximately 15 percent was deemed to produce an impact condition that was similar to, and possibly more severe than the TL4 single-unit truck (SUT) test from NCHRP Report 350. The primary concern was that if TL3 and TL4 converged, highway agencies would lose one of the longitudinal barrier options. The impact energy associated with the TL4 crash test conditions was increased by changing the SUT mass from 8,000 kg (17,637 lb) to 10,000 kg (22,000 lb) and the speed of the test vehicle from 80.47 km/h (50 mph) to 90.12 km/h (56 mph). This is particularly important to this study because the 57 percent increase in impact severity for MASH TL4 has resulted in higher design impact loads which will require stronger barriers and increased overturning moment which will require increased barrier height to prevent the heavier SUTs from rolling over the top of the barrier. In short, MASH TL4 barriers will likely have much higher capacities than Report 350 TL4 barriers. While a barrier height of 32 inches satisfied NCHRP Report 350 TL4 impact conditions, recent crash testing conducted at the Midwest Roadside Safety Facility (MwRSF) at the University of Nebraska (UNL) under NCHRP Project 22-14(2) and the Texas Transportation Institute (TTI) under NCHRP Project 22-14(3) has demonstrated that taller barriers will be required to accommodate MASH TL4 (i.e., Report 350 and MASH TL4 are not equivalent). The

25 Texas Department of Transportation is currently sponsoring research to determine the minimum barrier height and design impact load for MASH TL4. [Sheikh11] The vehicle specifications and impact conditions for TL5 and TL6 have not changed. Several user agencies have already begun applying the MASH criteria in their crash test programs. While no official crash test equivalencies have been released to compare Report 350 and MASH test level, Table 1 can be expanded to add the first line to represent the approximately equivalencies for MASH test levels. Table 2. Approximate Crash Test Acceptance Equivalencies. [after Horne97] Bridge Railing Testing Criteria Acceptance Equivalencies MASH TL1 TL2 TL3 TL5 TL6 Report 350 TL1 TL2 TL3 TL4 TL5 TL6 Report 230 MSL-1 MSL-2† AASHTO Guide Spec PL1 PL2 PL3 AASHTO LRFD Bridge Spec PL1 PL2 PL3 † This is the performance level usually cited when describing a barrier tested under NCHRP Report 230. It is close to TL3 but adequate TL3 performance cannot be assured without a pickup truck test. High-Containment Barriers Although fewer in number than TL3 and TL4 barriers, several barrier systems have been successfully designed and crash tested to TL5 criteria with an 80,000-lb tractor-van trailer. Only one barrier system is known to have been designed and successfully tested to TL6 with an 80,000-lb tractor-tank trailer (see Figure 14). [Hirsch85] TTI researchers conducted a study using a 90-in tall rigid instrumented concrete wall to quantify the magnitude and location of impact loads for a variety of trucks up to and including an 80,000 lb tractor-trailer with both a van-type and tank-type trailer. [Beason89] Speeds in these tests ranged from 50 mph to 60 mph and the impact angles ranged from 15 degrees to 25 degrees. Eleven additional tractor-trailer crash tests have been performed by TTI. In these tests, the gross vehicle weights ranged from 50,000 lbs to 80,000 lbs. Figure 15 shows a photograph from a test of a 42-in tall vertical wall bridge parapet being impacted by a 50,000 lb tractor-van trailer at a speed of 50 mph and an angle of 15 degrees. [Menges95] This test conforms to the impact conditions of Performance Level 3 (PL3) of the 1989 AASHTO “Guide Specification for Bridge Rails.” Table 3 shows a summary of the test information and parameters for the tractor- trailer barrier impacts run at TTI. Additionally, UNL researchers successfully developed both a 42-in. and a 51-in. tall, F- shape, half-section concrete barrier for TL5 impact criteria.[MWRSF10] For the 42-in. height, several configurations were provided with top barrier widths ranging from 10 in. to 12 in. and having barrier capacities ranging from 211 kips to 224 kips. Two preferred configurations were

26 recommended for the 51-in. tall barrier having top barrier widths of 11 in. and 12 in. The size, quantity, and spacing of longitudinal and vertical steel reinforcing bars were selected to prevent concrete blowouts as well as prevent vehicle penetrations through or vaulting over the top of the barriers. Figure 14. The only crash-tested TL6 bridge railing. [Hirsch85] Figure 15. 50,000-lbs tractor-trailer impacting a 42-inch vertical wall bridge railing. [Menges95]

27 Table 3. Summary of test information for selected heavy tractor-trailer crash tests. Test Date Test No. Barrier Type Barrier Description Van Trailer Type Model Yr Vehicle Make Gross Weight (lb) Impact Speed (mph) Impact Angle (deg) Pass/ Fail 1/16/2005 475680-1 Median Barrier CA Type 60 36000V 1993 International 59,928 49.2 Pass 2/27/2004 475150-1 Median Barrier CA Type 50 w/ Glare Screen 36000V 1989 Freightliner 45,236 41.3 34.0 Fail 12/12/1995 405511-02 Bridge Rail 1.07 m Vertical Wall 36000V 1983 Freightliner 79,200 49.8 14.5 Pass 8/9/1990 7162-01 Median Barrier Ontario "Tall Wall" 80000A 1980 International 80,000 49.6 15.1 7/11/1988 7069-13 Bridge Rail 42 in Vertical Wall 80000A 1979 International 50,050 51.4 16.2 Fail 5/27/1988 7046-09 Wall Instrumented Wall 80000A 1979 International 50,000 50.4 14.6 3/3/1988 7069-10 Bridge Rail 42 inch F-Shape 80000A 1979 International 50,000 52.2 14.0 Pass 5/8/1987 7046-04 Wall Instrumented Wall 80000A 1971 Peterbilt 79,900 54.8 16.0 4/7/1987 7046-03 Wall Instrumented Wall 80000A 1973 White 80,080 55.0 15.3 9/18/1984 2416-01 Bridge Rail Mod. T5 w/ Metal Rail 80000A 1981 Kenworth 80,080 48.4 14.5 Pass 10/25/1983 2911-01 Bridge Rail Modified T5 80000A 1980 Kenworth 80,120 51.4 15.0 Pass 5/26/1983 4798-13 Median Barrier New Jersey Safety 80000A 1974 International 80,180 52.1 16.5 7/14/1982 4348-02 Median Barrier Modified Safety 80000A 1978 Autocar 80,420 52.8 16.0 8/21/1981 2230-06 Bridge Rail Modified C202 80000A 1978 AutoCar 79,770 49.1 15.0 Pass

28 Guidelines and Specifications FHWA and AASHTO AASHTO Standard Specifications for Highway Bridges Historically, design of bridge rails has followed guidance contained in the AASHTO “Standard Specifications for Highway Bridges.” Prior to 1965, the AASHTO specification required very simply that “substantial railings along each side of the bridge shall be provided for the protection of traffic.” It was specified that the top members of bridge railings be designed to simultaneously resist a lateral horizontal force of 150 lb/ft and a vertical force of 100 lb/ft applied at the top of the railing. The design load on lower rail members varied inversely with curb height, ranging from 500 lb/ft for no curb to 300 lb/ft for curb heights of 9 in. or greater. It was further specified that the railing have a minimum height of 27 inches and a maximum height of 42 inches above the roadway surface. These loads are only a fraction of what are used today. Based on poor accident history, accentuated by increased exposure due to dramatically increasing travel volumes, the engineering community came to realize that these criteria were inadequate. There was an urgent necessity for a railing specification that established loading requirements more in line with the weights and increased speeds of vehicles of that day. Olson was perhaps the first to systematically examine the performance requirements for bridge railings in 1970. His results, documented in NCHRP Report 86, suggested using appropriate transitions, evaluation through crash testing, ability to minimize bridge rail penetration and many other things that are considered standard objectives of bridge railing design today. [Olson70] Bronstad built upon Olson’s work in NCHRP Report 239 where he presented a multiple service level approach to selecting bridge railings. [Bronstad81] Bronstad’s approach was a benefit-cost approach where the number of crashes exceeding the presumed capacity of the bridge railing was estimated. Four service levels were identified where the capacity was based on crash test results. Unfortunately, there was very little data available for Bronstad to use in developing his collision frequency and severity models so the resulting method was not definitive and was never widely adopted. Revised bridge railing specifications were subsequently published in 1965 in the 9th edition of the AASHO “Standard Specifications for Highway Bridges.” [AASHTO65] It required that rails and parapets be designed for a transverse load of 10,000 lbs divided among the various rail members using an elastic analysis. The force was applied as a concentrated load at the mid- span of a rail panel with the height and distribution of the load based on rail type and geometry as provided in an accompanying figure. Posts were designed for the transverse loading applied to each rail element plus a longitudinal load of half the transverse load. The transverse force on concrete parapet walls was distributed over a longitudinal length of 5 ft. The height of the railing was required to be no less than 27 inches and railing configurations successfully crash tested were exempt from the design provisions.

29 These bridge rail design procedures were retained through numerous editions of the specifications. In fact, the provisions in the 17th edition of the AASHTO “Standard Specifications for Highway Bridges” published in 2002 are essentially the same as the specification adopted in 1965. [AASHTO02] These requirements were intended to produce bridge rails that function adequately for passenger cars for a reasonable range of impact conditions. The reserve load capacity of the rail, beyond its elastic strength offers some factor of safety to accommodate more severe impact conditions or heavier vehicles. However, several catastrophic accidents involving large vehicles (e.g., buses and trucks) increased awareness of design requirements for bridge rails and the need to extend protection beyond passenger cars. 1989 AASHTO Guide Specification for Bridge Railings In 1989, AASHTO published the “Guide Specifications for Bridge Railings” (GSBR) to provide a more comprehensive approach for the design, testing, and selection of bridge rails than that contained in the AASHTO “Standard Specifications for Highway Bridges.” [AASHTO89] The GSBR introduced several new and important concepts to the practice of selecting bridge railings including: • Bridge railing performance should be demonstrated in full-scale crash tests. • There should be multiple performance levels (i.e., three were recommended) to address the different risks and costs associated with specific traffic and bridge characteristics. • A cost-benefit encroachment modeling software program, benefit-cost analysis program (BCAP), was introduced to help designers make bridge railing selection decisions. • Generic selection guidelines were presented that recommended the appropriate test level based on traffic volume, percent trucks, speed limit, horizontal curvature and grade. The crash testing aspects of the 1989 AASHTO GSBR were discussed previously in the section on Crash Testing so this section will focus on the selection guidelines in the 1989 GSBR. The 1989 AASHTO GSBR provided bridge engineers guidance for determining the appropriate railing performance level for a given bridge site. Selection guidelines were provided that estimated the appropriate railing performance level for a given bridge site based on highway and site characteristics. The highway and site characteristics used in the selection guidelines included highway type (e.g., divided, undivided, etc.), design speed, traffic volume, percent trucks, and bridge rail offset. The tables applied to bridges that were on tangent, level roadways with deck surfaces approximately 35 feet above the underlying ground or water surface. It was further assumed that there was low occupancy land use or shallow water under the bridge structure. Correction factors were provided to permit the engineer to adjust the traffic volume for horizontal curvature, vertical grade, different deck heights, and different densities of land use beneath the bridge. Table 4 shows a portion of the 1989 GSBR Table G2.7.1.3B which illustrates the selection recommendations. For example, if the design speed of a four-lane divided highway is 60 mi/hr (100 km/hr) and the percent of trucks is about 15 percent and the bridge railing is offset

30 from the edge of travel 8 feet, a PL1 railing is appropriate for traffic volumes up to 3,700 vpd; a PL2 bridge railing is recommended for traffic volumes between 3,700 and 31,900 vpd and a PL3 bridge railing is recommended for traffic volumes greater than 31,900 vpd. These recommendations presume that the highway section is straight with no grade, a deck height above the surface of 35 ft or less and the bridge does not pass over a sensitive or occupied area. If it does have horizontal curvature, grade or passes over a sensitive or occupied area, adjustment factors are presented. These adjustments are multiplied by the traffic volume in Table G2.7.1.3B. Table 4. 60 mi/hr portion of the 1989 GSRB selection table.[AASHTO89] Returning to the example, if the highway being considered has an ADT of 28,000 vpd, 60 mi/hr design speed, 15 percent trucks and an 8-ft shoulder; Table 4 suggests a PL2 bridge railing is appropriate. If we assume that the bridge is on a vertical down grade of minus 6 percent, a horizontal curve of six degrees and is actually 50 ft above a high-occupancy land use surface

31 below the adjustment factors would be 2.0, 2.0 and 1.8. The ADT would be adjust to 18,000 ∙ 2.0 ∙ 2.0 ∙ 1.8 = 129,600 vpd which would place the railing into the PL3 category. The tables, therefore, allow the designer to select an appropriate rail based on the traffic volume, design speed, percent trucks, railing offset, horizontal curvature, grade, height of the structure and land use. These selection procedures were developed using a benefit-cost analysis combined with engineering judgment. The BCAP estimated roadside encroachments, the consequences of these encroachments, and the cost of the consequences. An incremental benefit-cost ratio was computed to facilitate comparison of the relative merits or benefits of one design alternative to another. Table G2.7.1.3B was to be used for bridge railing selection unless the designer used the BCAP program. The BCAP program will be discussed in more detail in a later section but obviously since Table G2.7.1.3B was based on the predictions of BCAP, the accuracy and validity of BCAP were fundamental to the validity of the recommendations. NCHRP Project 22-08 was initiated in order to assess BCAP and validate the 1989 AASHTO GSBR recommendations. [Mak94] Unfortunately, Mak and Sicking, the principal investigators for NCHRP 22-08, found some serious short comings of BCAP itself and the assumptions that were built into the selection tables. Mak and Sicking found that BCAP seriously over predicted bridge railing penetrations and seriously under predicted rollovers; the opposite of what would normally be expected. Based on crash test experience and anecdotal information, most bridge railings “fail” due to a heavy vehicle rolling over the barrier rather than penetrating after a structural failure so the BCAP results were counter intuitive. When a series of base-line simulations were performed with BCAP mimicking the GSBR recommendations, the researchers found that BCAP predicted 32.7 percent of tractor-trailer trucks striking a PL2 bridge railing would penetrate the bridge railing yet there were no predictions of rollover even though the center of gravity of a typical tractor- trailer truck is 64 inches high and the typical PL2 barrier height was 32 inches high (i.e., the c.g. of the vehicle is 32 inches higher than the top of the barrier). [Mak94] Mak and Sicking discovered several reasons for this. One reason was the algorithm used to predict rollovers resulted in unreasonably high critical velocities. A new rollover algorithm was proposed and implemented as will be discussed in the later section devoted to BCAP. Another reason involved barrier capacity. BCAP estimates the forces on the barrier using an algorithm first developed by Olson. The algorithm, as will be described later, is a simple derivation of the force based on the overall mechanics of the impact. After the impact force imparted by the vehicle is calculated, it is compared to the assumed bridge rail capacity. If the impact force is greater than the capacity, the bridge rail is considered failed. Estimating the actual capacity of bridge railings is more difficult than it might first seem. Materials are routinely assumed to be less strong and loads are routinely over estimated in design so even if the theoretical capacity is calculated it is likely a very conservative value. For example, in designing concrete structures a resistance factor 0.85 is usually used for bending which essentially takes advantage of only 85 percent of the strength of concrete. Likewise, if an allowable stress design

32 method for steel were used, 67 percent of the strength of the steel is assumed. In both cases, the designer is neglecting a significant portion of the capacity of the structure. While this makes excellent design sense, it makes it difficult to estimate the real failure conditions of the structure. BCAP assumed that PL1 bridge railings have a capacity of 15 kips, PL2 railings have 35 kips and PL3 railings have 55 kips. While there are relatively few crash tests where structural failure of the bridge railing was observed, Mak and Sicking were able to find some cases where the bridge railing experienced some degree of structural failure (i.e., hairline cracking, spalling, etc.). When they compared the limited crash test results to the BCAP assumptions they found that the BCAP assumptions were about half what could be supported by crash tests as shown in Table 5. Table 5. Bridge railing capacity recommendations in BCAP and NCHRP 22-08.[after Mak94] Performance Level BCAP Assumption (kips) Mak/Sicking Recommendation (kips) PL1 15 30 PL2 35 64 PL3 55 108 Adding to the difficulty is the basic assumption in BCAP that when capacity is reached, the bridge railing will totally fail and allow the vehicle to penetrate. In fact, this does not generally happen. Bridge railings can experience structural failure and sometimes will still redirect the vehicle. The “failure” may be cracks or spalls that are considered serious damage to the bridge rail, but the bridge rail may still have enough structural integrity to prevent penetration by the vehicle. Recently, Alberson and others evaluated a 32-inch high PL2 concrete safety shaped barrier that had experienced structural failure problems in the field.[Alberson11; Alberson04] A yield-line structural analysis was performed on the bridge railing which resulted in an estimate of the barrier capacity of 33.6 kips when loaded near a construction joint and 47.7 kips when loaded at the mid-span. The same design was then constructed and statically tested to failure resulting in a near-the-joint capacity of 35.1 kips and a mid-span capacity of 45.1 kips. The bridge railing was also subjected to full-scale Report 350 TL4 crash tests which were passed successfully and which caused relatively minor concrete damage (e.g., hairline cracks and some gouging). As shown by Alberson’s research, the capacity values suggested by the 1989 AASHTO GSBR were grossly over conservative and those proposed by Mak and Sicking were more appropriate although it should be noted that this particular railing was chosen for investigation precisely because there had been some observed field structural failures so this particular railing probably represents the lower end of the capacity of PL2 railings. Since BCAP first assesses the capacity and then the rollover potential, the overly conservative values for capacity tended to predict too many penetrations. Since the higher

33 velocity truck impacts would tend to reach the capacity too early and the rollover algorithm was under conservative, penetrations were over predicted and rollovers under predicted. Mak and Sicking revised the rollover algorithm and adjusted the bridge railing capacities upward as shown in Table 5 and re-ran their analysis. For example, 32.7 percent of tractor-trailer truck crashes penetrated the railing and none rolled over in the initial BCAP runs whereas after the improvements implemented by Mak and Sicking 3.4 percent penetrated which seemed more reasonable. Mak and Sicking also evaluated bridge railing crash data from Texas as will be discussed in more detail in a later section. [Mak94] Mak and Sicking found that the Texas data indicated that 2.2 percent of bridge railing crashes result in the vehicle going through (i.e., penetration) or over (i.e., roll over the barrier) and they believed that even this value was a high-side estimate due to coding errors on the police crash reports. The improved BCAP with the higher capacity limits and improved rollover algorithm resulted in an overall estimate of 10 percent going through (i.e., 1.2 percent penetrating and 8.9 percent rolling over) for the typical Texas conditions so even the improved BCAP appeared to over predict penetrations/rollover by an order of magnitude although the proportion of penetrations to rollovers appears much more reasonable. In short, then, BCAP and the 1989 AASHTO GSBR appear to over predict bridge railing penetrations and under predict rollovers. The improvements from NCHRP 22-08 appeared to improve the results although even the improved BCAP over predicts the incidence of vehicles going through or over the bridge railing. Mak and Sicking developed new versions of the selection tables based on the improved version of BCAP. These new selection tables were structured in an identical way to the prior tables but the traffic volume cutoffs were higher. Table 6 shows the portion of the revised recommendations from NCHRP Project 22-08 that corresponds to Table 4 shown earlier. In the example presented earlier, a PL2 bridge railing would be recommended for a four-lane divided highway with a 60 mi/hr (100 km/hr) design speed, 15 percent trucks and eight-foot offset from the travelled way and no adjustments for traffic volumes between 3,700 and 31,900 vpd whereas Table 6 would suggest a PL2 bridge railing under the same conditions is appropriate for traffic volumes of 17,000 to 51,000 vpd; much higher than the 1989 GSBR. The final report for NCHRP 22-08 was never published since NCHRP Report 350 appeared about the same time and the Roadside Safety Analysis Program (RSAP) program was also released. It was thought that bridge railing selection guidelines would work themselves out in the process of replacing the crash testing recommendations of the 1989 GSBR with the new recommendations of Report 350 and replacing BCAP with RSAP. Unfortunately, such was not the case and the 1989 GSBR was never updated and its recommendations were never superseded.

34 Table 6. Revised selection guidelines for bridge railings based on NCHRP 22-08. [Mak94] Roadside Design Guide 1989 also was the year that AASHTO first published the Roadside Design Guide (RDG). The RDG is a comprehensive guide to designing many aspects of the roadside but Chapter 7 deals exclusively with bridge railings.[AASHTO89] The subject of test level selection procedures is addressed briefly in section 7.3 of the RDG but the reader is referred back to section 5.3 for general guidance on traffic and operational characteristics that should be used in selecting the appropriate test level barrier.

35 RDG section 5.3 lists the following three subjective criteria that should be used in choosing an appropriate test level barrier: 1. Percentage of heavy vehicles, 2. Adverse geometrics (e.g., small-radius horizontal curves), 3. Severe consequences of a penetration by a heavy vehicle. In essence, the RDG restates the generally philosophy of the 1989 GSBR without providing any additional specific information. 2004 AASHTO A Policy on the Geometric Design of Highways and Streets The AASHTO document “A Policy on the Geometric Design of Highways and Streets” (i.e., the Green Book) discusses the subject of bridge railings while addressing interchanges, underpasses, and overpasses. Specifically, the Green Book recommends that “the design vehicle should be safely redirected, without penetration or vaulting over the railing … the railings should not pocket or snag the design vehicle, causing abrupt deceleration or spinout; and it should not cause the design vehicle to roll over.” [AASHTO04] Bridge railings may limit the sight distance at interchanges, intersections, on ramps and along the road. The Green Book acknowledges this concern and suggests the bridge railing “should provide a freedom of view … insofar as practical; however, capability to redirect errant vehicles should have precedence over preserving the motorist’s view.” [AASHTO04] Adjustments to the horizontal alignment are suggested to improve sight distance, when feasible. When pedestrians or bicycles are accommodated on the bridge, the Green Book suggests “a barrier-type bridge rail of adequate height should be installed between the pedestrian walkway and the roadway. Also, a pedestrian rail or screen should be provided on the outer edge of the walkway.” [AASHTO04] FHWA Supplemental Guidance on Accommodating Heavy Vehicles on US Highways The FHWA Office of Safety issued supplemental guidance on accommodating heavy vehicles on US Highways in a 2004 report. [FHWA04] The report addresses both geometric design barrier design, and placement issues. According to the report, there were 302 fatal single vehicle truck crashes in 2002 involving van, cargo, flat-bed or dump type trailers. Of these 302 fatal crashes, the first harmful event in 26 (8.6 percent) of these cases was listed as guardrail, concrete traffic barrier or bridge railing. The FHWA report goes on to note that there are no specific warrants for the use of higher performance or test level barriers because heavy vehicle impacts are generally rare events. The report repeats the general guidance found in the RDG for subjective factors including (1) a high percentage of trucks, (2) adverse geometrics and/or poor sight distance and (3) potentially severe consequences associated with the truck penetrating the barrier and offers some additional specific guidance [FHWA04].

36 This 2004 FHWA report cites a 1997 FHWA policy memorandum which formally adopted NCHRP Report 350 as the guideline for testing bridge railings. This 1997 memorandum summarized over a decade of crash tests conducted on bridge railings, providing a complete list of all the crash-tested bridge railings, the guidelines used to test the bridge railings (e.g., NCHRP 230, NCHRP 350, etc.), “equivalency” listings for other test standards, and sketches for construction of the railings. This memorandum also established that the minimum acceptable bridge railing acceptable on the national network will be Report 350 TL3, “unless supported by a rational selection procedure.” The States were not, as a result of this memorandum, required to upgrade existing bridge railings, beyond normal improvements. [FHWA97b] The FHWA issued another policy memorandum in 2010, “Design Considerations for Prevention of Cargo Tank Rollovers” in response to another fatal tanker truck crash investigated by NTSB. [FHWA10] This memorandum reiterated the guidance in the 2004 FHWA report for accommodating heavy vehicles, while adding some additional guidance. Some geometric factors the FHWA suggests when selecting a bridge railing are: • Conflict points, • Dramatic horizontal and/or vertical alignments, • Lowering of the design speed, and • Super-elevation which may increase large vehicle instability. Some highway characteristics the FHWA suggests considering in the selection of bridge railings: • High-volume highways or other such facilities (i.e., transit, commuter rail, etc.) located beneath a bridge, • Facilities where an impact could lead to catastrophic loss of life (i.e., chemical plants, nuclear facilities, etc.), • Sensitive environmental areas (i.e., public water supplies), or • Regionally or nationally significant bridges and tunnels. In summary, several decades of recommendations by NTSB, crash testing by FHWA, and multiple national research projects have resulted in general guidance from the FHWA for the selection of bridge rails based on geometric factors and highway characteristics but there is still relatively little specific guidance on the selection of bridge railings to fit specific local conditions. The States Many states refer designers to Chapter 13 of the AASHTO LRFD Bridge Design Specifications for strength and geometric requirements, NCHRP 350 for crash test criteria, the 1989 AASHTO Guide Specification for warrants based on ADT, design speed, percentage truck traffic and horizontal and vertical geometry while noting that there is ongoing research to evaluate the warrants in the 1989 Specification. These national documents are supplemented in many States with a Bridge or Structures Design Manual in which the States detail the use of

37 specific railings under certain situations, specify particular test levels for certain roadways and discuss retrofit polices. The Bridge or Structures Design Manuals from many different States have been reviewed and are discussed in this section. Florida DOT The Florida Department of Transportation’s Structures Design Guidelines provide guidance to designers on the selection of bridge railings. [FDOT11] This document provides the following guidance for the installation of bridge railings, which extends to all construction, including new, temporary, 3R (e.g., Resurfacing, Restoration and Rehabilitation) and widening projects: • Permanent installations must install a successfully crash-tested TL4, TL5 or TL6 bridge railing. • Temporary installations must install a successfully crash-tested TL3 (minimum) when shielding drop-offs. TL2 (minimum) may be used when shielding work zones without drop-offs and a design speed of 45 mph or less. • Upgrade both sides of a structure “when widening work is proposed for only one side and the existing traffic railing on the non-widened side does not meet the criteria for new traffic railings.” [FDOT11] Designers should provide a TL5 or TL6 bridge railing “when any of the following conditions exist: • The volume of truck traffic is unusually high. • A vehicle penetrating or overtopping the traffic railing would cause high risk to the public or surrounding facilities. • The alignment is sharply curved with moderate to heavy truck traffic.”[FDOT11] Standard bridge railing designs are suggested by FDOT, however, the use of non-FDOT standard railings is permitted provided the railings meet the requirements listed above and following the review and approval of the FDOT Structures Design Office. When rehabilitation or renovation work is proposed on an existing structure and the bridge railing does not meet the criteria detailed above, the existing railing should be replaced or retrofitted to meet the TL4 minimum performance standards. When selecting a replacement or retrofit bridge railing, FDOT suggests that designers evaluate the following aspects of the project: 1. “Elements of the structure. • Width, alignment and grade of roadway along structure. • Type, aesthetics, and strength of existing railing. • Structure length. • Potential for posting speed limits in the vicinity of the structure. • Potential for establishing no-passing zones in the vicinity of the structure.

38 • Approach and trailing end treatments (guardrail, crash cushion or rigid shoulder barrier). • Strength of supporting bridge deck or wall. • Load rating of existing bridge. 2. Characteristics of the structure location. • Position of adjacent streets and their average daily traffic. • Structure height above lower terrain or waterway. • Approach roadways width, alignment and grade. • Design speed, posted speed, average daily traffic and percentage of truck traffic. • Accident history on the structure. • Traffic control required for initial construction of retrofit and for potential future repairs. • Locations and characteristics of pedestrian facilities / features (if present). 3. Features of the retrofit designs. • Placement or spacing of anchor bolts or dowels. • Reinforcement anchorage and potential conflicts with existing reinforcement, voids, conduits, etc. • Self-weight of retrofit railing. • End treatments. • Effects on pedestrian facilities. • Evaluation of existing supporting structure strength for traffic railing retrofits.”[FDOT11] FDOT suggests the use of the modified thrie beam guardrail or vertical face traffic railing retrofits which are based on successfully crash-tested TL4 designs. Modifications to the designs are offered to designers which should work with the various existing Florida bridges.[FDOT11] Minnesota DOT The Minnesota Department of Transportation publishes the Minnesota Bridge Design Manual where bridge railing application is discussed.[MNDOT06] The Manual requires that “railing designs shall include consideration of safety, cost, aesthetics and maintenance.” [MNDOT06] The Bridge Design Manual details the use of TL2 through TL5 bridge railings for uses with sidewalks, bicycles, various design speeds, and different geometrics. Different bridge railings are also specified for specific routes. These specifications have been summarized and are shown in Table 7.

39 Table 7 Summary of Minnesota “TABLE 13.2.1: Standard Rail Applications” [MNDOT06] Description Test Level Speed Limit Application Comment Concrete Barrier (Type P-1) and Metal Railing TL2 ≤ 40 mph Outside edge of walk on highway bridges with sidewalks where bicycle traffic on the walk is expected and protective screening is not required. 2'-4" parapet with 2'-2" metal rail Cloquet Bridge Railing Bridge No. 09008 and 09009 2'-2 3/4" metal rail on 2'-4" parapet Concrete Barrier (Type P-1) and Wire Fence TL2 ≤ 40 mph Highway bridges with walks. This rail is used on the outside edge of walk and meets bicycle and protective screening requirements. 2'-4" parapet and 6' metal rail with chain link fabric. Concrete Barrier (Type P-1) and Tube Railing with Fence 2'-4" parapet and 5'-8 ½" metal rail with chain link fabric Concrete Barrier (Type P-3) Ornamental 3'-9" metal rail on 2'-4" parapet St. Peter Railing 4'-6" metal rail on 2'-4" parapet TH 100 Corridor Standard 3'-9" metal rail on 2'-4" parapet TH 212 Corridor Standard 5'-8' to 9'-2" metal rail on 2'-4" parapet TH 610 Corridor Standard 5'-51/2" metal rail on 2'-4"parapet Victoria Street Railing 5'-8" metal rail on 2'-4" parapet Concrete Barrier (Type F or P-4) TL4 All Traffic Only 2'-8" tall Concrete Barrier (Type P-2) and Structural Tube Railing TL4 All Traffic Only, where an aesthetic railing is desired. 1'-3" metal railing on 1'-9" parapet Concrete Barrier (Type F) TL5 > 40 mph High Protection Area where Dc > 5° and Speed > 40 mph. 3'-6" tall Concrete Barrier (Type F) TL5 All Between sidewalk and roadway where the shoulder is < 6'. Concrete Barrier (Type F) TL5 All Bridges with designated bike path or where glare screen is required. 4'-8" tall (The additional height is to protect a bicycle rider.)

40 New York State DOT The New York State DOT uses the following recommendations for each performance level, as outlined in the New York State DOT Bridge Manual: [NYSDOT10] • “TL2 (PL1)–Taken to be generally acceptable for most local and collector roads with favorable site conditions, work zones and where a small number of heavy vehicles are expected and posted speeds are reduced. • TL4 (PL2)–Taken to be generally acceptable for the majority of applications on high-speed highways, expressways and interstate highways with a mixture of trucks and heavy vehicles. • TL5 (PL3)–Taken to be generally acceptable for applications on high-speed, high traffic volume and high ratio of heavy vehicles for expressways and interstate highways with unfavorable site conditions.” The railing functional and geometric characteristics are considered. These criteria include the under-crossing features, pedestrian accommodations, and bicycle accommodations. The formal accommodation of pedestrian and bicycle traffic requires the use of 3.5 foot tall railings. Table 6-1 of the NYSDOT Bridge Manual outlines bridge railing designs by various AADTs for TL2 through TL5 which are appropriate for various scenarios including pedestrian and bicycle accommodations and different under-crossing features. Interstate and other controlled-access, high-speed highways are mandated to have concrete bridge railings although parkways without truck traffic and culvert structures are excluded. Steel railings are currently not permitted on interstates or for other TL5 uses since, as the document in-correctly states, there were “no known steel railing systems” crash tested for TL5. In fact, concrete is the first choice of material for most of the bridge railing design categories described above. “This preference is based on the concrete barrier’s strength, durability and low initial and maintenance costs compared to metal railing systems. Factors that may cause an alternative selection to be made are:”[NYSDOT10] • Bridge Deck Drainage • Aesthetics • Visibility • Snow Accumulation “Railing treatments on rehabilitation projects is a complex subject with many project specific considerations….engineering judgment will be required.” Safety is the first considered when deciding on whether or not to upgrade bridge railing on a rehabilitation project. After determining the long-term planning strategy for the area and the structure, the existing railings are examined to determine if the railings meet NCHRP 230 crash test criteria. If the railing does meet those criteria, the existing railings are often

41 retrofitted with guardrail if there is a safety walk or allowed to remain in place as is. If the bridge railings do not meet the NCHRP230 or later criteria, the railings are upgraded. [NYSDOT10] New Jersey DOT The New Jersey DOT specifies, through the New Jersey Design Manual for Bridges and Structures, 5th Edition that TL5 systems are considered the minimally accepted system for bridges which carry interstate traffic. [NJDOT10] TL4 systems are used for other NHS classified roadways. On non-NHS/non-State owned roadways, the use of TL1, TL2, and TL3 bridge railings are permitted. NJDOT suggests that the use of crash test specifications outlined in NCHRP 350 be used to determine a test level which best corresponds with the roadway when choosing a test level less than TL4, specifically considering design speed, truck traffic and how the roadway characteristics relate to the test level specifications to determine an appropriate test level for non-NHS roadways. There are two TL5 systems used on interstate bridges in New Jersey: the 42-inch tall “F” shape concrete barrier is recommended for heavy vehicle containment and for horizontal curves of less than 1000 feet or on exit ramps and the 50-inch high Texas Type HT railing is used in conjunction with noise barriers where heavy vehicle over-tipping and potentially damaging the noise barrier is a concern.[NJDOT10] Rhode Island DOT The Rhode Island LRFD Bridge Design Manual states that “all railings systems shall meet the full-scale crash test criteria as established in the NCHRP Report 350.” [RIDOT11] For new construction, TL4 bridge railings are the minimum test level barrier installed, except on interstate highways where TL5 bridge railing shall be installed. TL5 bridge railings are also considered when the bridge is expected to experience high traffic volumes, high speeds with high truck percentages, or unfavorable site conditions. Unfavorable conditions may include: • “High-occupancy land uses below the bridge, • Deep water below the bridge, • Steep profile grades on or approaching the bridge, • High curvature along the alignment of the bridge, • Anticipated excessive number of van-type tractor-trailers, or • Any other set of conditions which, through sound engineering judgment, may justify a higher level of railing resistance.” [RIDOT11] The installation of barriers with a test level less than TL4 may be considered when the ADT is less than 500 vehicles per day, the percentage of trucks is less than or equal to 5%, the design speed is less than 40 mph, the bridge is on a tangent section and the bridge deck height is less than 28 feet above ground or water surface elevation. Minor detail changes to existing, crash-tested railing systems are permitted, provided

42 “engineering judgment and/or analysis” is used to determine the need for additional crash testing. [RIDOT11] Michigan DOT In contrast to some other states, Michigan does not specify the minimum test level for bridge railings but it does require upgrading all railings when the bridge deck is replaced. [MDOT09] Regarding the installation of bridge railing, the MDOT Bridge Design Manual states simply that the railing “shall be of a type that has passed full-scale impact (crash) tests” and provides a reference to five standard MDOT railings: • Type 4 Barrier - Standard Plan B-17-Series, • Type 5 Barrier - Standard Plan B-20-Series, • 2 Tube railing - Standard PlanB-21-Series, • 4 Tube railing – Standard Plan B-26-series and • Aesthetic Parapet Tube railing - Standard Plan B-25-Series.[MDOT09] Massachusetts DOT The Massachusetts Department of Transportation specifies the use of TL2, TL4 and TL5 barriers under different situations in the MassHighway Bridge Manual.[MassDOT10] The following circumstances dictate the use of these different test levels: • TL2 bridge railings may only be used on non-NHS roadways with speeds not exceeding 45mph. • TL4 bridge railings may be used on NHS and Non-NHS highways, except limited access highways and their ramps • TL5 bridge railings must be used on limited access highways and the ramps. This includes interstate highways, NHS and Non-NHS highways. Details for bridge railing designs for use in specific situations (e.g.., where pedestrian are permitted or forbidden, bridges over electrified rail road tracks, municipally owned bridges, etc.) are provided. Bridge railings other than the standard railings are permitted provided that the use of non-standard railings “can be justified and that they have been crash tested.”[MassDOT10] Ohio DOT An Ohio DOT inter-office memorandum restates the 2003 Design Manual policy which established the minimum acceptable bridge railing shall be TL3, however, now Ohio will allow the existing TL2 Deep Beam Bridge Railings to be maintained provided they are in good condition. [ODOT02] North Carolina DOT The North Carolina Department of Transportation supplements national standards with its Structures Design Manual where a minimum bridge railing of TL3 is established. [NCDOT10] TL2 or aesthetic railings are permitted under the following situations:

43 • “Non-NHS routes, • Design speeds less than or equal to 45 mph, or • In conjunction with a sidewalk.” [NCDOT10] NCDOT suggests the use of vertical concrete barrier rail for bridges on NHS and non-NHS routes. Bridges which accommodate pedestrians can add height to the railing through an added metal railing. When conducting an overlay on the bridge, a minimum rail height is established and noted on the plans to be maintained during construction, however, guidance on establishing the minimum height is not provided in the Structures Design Manual.[NCDOT10] North Dakota DOT North Dakota publishes a Design Manual which includes specifications for all facets of highway design within the state of North Dakota. [NDDOT09] The bridge chapter has specifications for the installation of new and retrofit bridge railings based on height and refers designers to AASHTO for the latest specifications. NDDOT requires all new bridge railing to be 32 inches tall (i.e., essentially Report 350 TL4) while existing railing may be retrofitted with two-tubes where applicable. Bridge railings on sidewalks are to be a minimum of 42 inches tall on the outer edge. Bridge railings on shared-use paths (i.e., bicycle and pedestrian) are to be a minimum of 54 inches tall on the outer edge. Both are to be crash tested. [NDDOT09] Illinois DOT The Illinois DOT provides guidance on the selection of the appropriate bridge railing test level in its Bridge Manual where IDOT states that the owner of a structure is “responsible for determining the test level necessary for each application” and that railings on all new or rehabilitated bridges on Federal and State routes shall be TL4. [IDOT09] The preferred bridge railing is the 34-inch TL4 F-Shape. A 42-inch high TL5 F-shape bridge railing “should only be used in the following scenarios: 1. Structures with a future DHV (one way) × % trucks greater than 250. 2. Structures located in areas with high incidences of truck rollover accidents. 3. Structures with a radius of 1000 ft. or less with truck traffic.”[IDOT09] Following these guidelines, structures carrying 10% or more truck traffic and an ADT of 5,000 vpd or more should install a TL5 bridge railing. Indiana DOT Indiana DOT allows the use of bridge railing designs ranging in performance from TL2 through TL5 but does not include specifications for the specific installation of any particular test level under any particular situation. Designers are referred to the AASHTO LRFD Bridge Design Specifications.[INDOT10]

44 Nevada DOT Nevada DOT provides guidance to designers on the general application of TL3 through TL6 bridge railings in its Structures Manual.[NDOT08] TL1 and TL2 bridge rails “have no application in Nevada.” Specific warrants for TL3 and higher bridge railings are not provided but the following general application guidance is offered for each test level: • TL3 bridge railing is the minimum acceptable performance level. It may be used for roadways with “very low mixtures of heavy vehicles and with favorable site conditions.” • TL4 bridge railing is the minimum performance level for bridges on the NHS system. It may be used on high-speed highways, freeways, expressways and Interstates with a mixture of trucks and other heavy vehicles. • TL5 bridge railing is “for a special case where large trucks make up a significant portion of the vehicular mix” and may only be used when approved by the Chief Structures Engineer. • TL6 bridge railing is “for a special case where alignment geometry may require the use of an extra height rail” and may only be used when approved by the Chief Structures Engineer. NDOT typically uses the 42-inch high F-shaped concrete barrier but the 32-inch high version may also be used when applicable.[NDOT08] Washington DOT In its Bridge Design Manual Washington DOT requires the use of at least a TL4 bridge railing on all new bridges with some exceptions. [WSDOT08] TL5 bridge railings are required under the following conditions: • “T intersections on a structure. • Barriers on structures with a radius of curvature less than 500 ft, greater than 10% truck traffic, and where approach speeds are 50 mph or greater. Particular systems identified as acceptable include the F-shaped and vertical face concrete bridge railings. Washington DOT systematically improves or replaces existing deficient bridge railings within the limits of roadway resurfacing projects by “utilizing an approved crash-tested rail system that is appropriate for the site” or designing a new system. Approved systems are detailed in the Bridge Design Manual and include TL2 through TL4 retrofit designs. [WSDOT08] International Specifications A number of European countries have developed bridge railing selection criteria and these are based on the crash testing standards and containment levels defined in European Normative 1317 (EN 1317). The basic containment levels are described herein in order to provide some basis of comparison between the US AASHTO GSBR/Report 350/MASH test levels and the EN 1317 containment levels. EN 1317 includes four

45 containment levels; containment “T” for low-angle containment consistent with many temporary applications; containment “N” for the normal level on most roads; containment H for high containment levels and the H4 level for very high containment. Containment level T is not appropriate for selecting bridge railings but the other three containment levels are shown below in Table 8 with the nearest MASH test level in terms of the target energy. The MASH and EN 1317 testing requirements are different so the EN 1317 containment levels do not correspond exactly to the MASH test levels but for selection guideline comparison purposes the equivalences shown in Table 8 should be adequate. There is no MASH energy level similar to EN 1317 containment level H3 so in later tables and in this discussion H3 barriers are referred to as TL4+ (i.e., between TL4 and TL5). Table 8. EN 1317 Containment Levels Pertaining to Bridge Railings with the nearest MASH Test Level. Containment Level Acceptance Tests Containment Energy Level Nearest MASH Test Level Minimum MASH Energy Level (kJ) (ft-kips) (kJ) (ft-kips) N1 TB31 370 273 N2 TB32/TB11 700 516 TL2 429 316 H1 TB42/TB11 1,890 1,393 TL3 876 645 H2 TB51/TB11 2,458 1,811 TL4 3,125 2,305 H3 TB61/TB11 3,951 1,087 H4a TB71/TB11 4,890 1,656 H4b TB81/TB11 6,194 4,565 TL5 8,889 6,556 Austria Specifications for choosing bridge railings in Austria are contained in RVS 15.04.71(15.47). The EN1317 containment levels are specified based on the highway type and certain characteristics of the roadway as summarized in Table 9. As shown in Table 9, the basic or default bridge railing containment level for freeways (i.e., divided high-speed highways) is EN1317 containment level H2 which is more or less equivalent to Report 350 TL4. For certain geometric conditions like grades on long bridges, small- radius horizontal curves, bridges with no emergency lanes and bridges that cross over high-density populated areas or other transportation facilities, the containment level is increased. The highest containment level specified is EN1317 H4b, which is roughly equivalent to Report 350/MASH TL5. The guidelines for bridges on secondary roads (i.e., lower speed undivided roadways) have a similar pattern, although the basic containment level is EN1317 N1 which is broadly similar to Report 350 TL2. The containment level can be increased

46 based on the geometry of the bridge and land use up to a containment level of H2 (i.e., roughly TL4). Table 9. Austrian containment level selection guidelines. Roadway Characteristics EN1317 Containment Level Freeways Normal case H2 Upgrade > 4% for a lengths > 400m H3 Tight horizontal curves H3 Roads with no emergency lane H3 Bridges over important or protected areas H3 High-density populated areas H3 Bridges over railroads with train speed >70km/h H4 Bridges over railroads with train speed <70km/h H2 Secondary Roads Normal case N1 Upgrade > 6% for a length > 250m N2 Tight horizontal curves N2 Bridges over important or protected areas H1 High-density populated areas H1 Bridges over railroads with train speed >70km/h H4 Bridges over railroads with train speed <70km/h H2 Canada Within Canada, each Province largely determines its own road design policy. The Alberta province Infrastructure and Transportation Roadside Design Guide , for example, references the 1989 AASHTO Guide Specification and the FHWA Heavy Vehicle Guidance discussed above. [Alberta11, AASHTO89; FHWA04] Using these documents as references, the Alberta Province guides designers as summarized in Table 10 for new and retrofit bridge railing installations.

47 Table 10 Alberta Canada Roadside Design Guide Bridge Rail Specifications Test Level Application TL2 For use on local roads. TL4 Preferred bridge rail for most applications. Urban bridges with cyclists. Urban areas where aesthetics are important. Short bridges (i.e., length < 20 m). TL5 Use when high AADT with heavy truck volumes and/or high structure dictates higher test level. Appendix C1 Section HC1.1 of the Alberta Roadside Design Guide specifically addresses upgrading existing bridge railings. [Alberta11a] Alberta employs a cost-benefit analysis procedure, outlined in Appendix C1, to determine when to upgrade existing bridge railings. Using encroachment rates and lateral extent probability, adjustments for horizontal and vertical alignment, and adjustments for bridge height and occupancy (i.e., Table 11), retrofit alternatives are considered for the remaining design life of the bridge. The null alternative (i.e., doing nothing) is also considered. The most cost-beneficial alternative is then implemented. The Alberta procedure appears to be largely based on the 1989 AASHTO GSBR and BCAP with alterations made specific to the circumstances in Alberta. The basic equation used is: PWCC = R ∙ kc∙ kg∙ P ∙ km ∙ ks ∙ AC ∙ L ∙ KC/100 where: PWCC = Present worth of the collision costs for one side of the bridge, R = Base encroachment rate in encroachments/km/yr/side, kc = Highway curvature adjustment factor (i.e., unitless), kg = Highway grade adjustment factor(i.e., unitless), P = Lateral encroachment probability, km = Multi-lane adjustment factor (i.e., unitless), ks = Bridge height and occupancy factor (i.e., unitless), AC = Cost per collision for severity index, L = Length of bridge railing and KC = Present worth and traffic growth factor. The base encroachment rate, horizontal curve and grade factors, which are provided in tables in the Alberta specification, are all the same as used in BCAP with the exception they are reformulated into severity index (SI) units. The factor P, the lateral extent of encroachment, is similar to the approach used in BCAP and RSAP. The table is organized by shoulder width and design speed. Since bridge railings are continuous and shoulders are usually not much more than a lane-width, the lateral extent of

48 encroachment calculation is easily converted into a simple factor. The multi-lane factor accounts for encroachments from other lanes. The bridge height and occupancy adjustment factor (i.e., ks) is shown in Table 11. The adjustment can be as high as 2.85 for bridges over 75 feet high that span over high-occupancy land use areas. The KC factor lumps together both the present worth factor and traffic growth factors based on an assumed 4 percent discount rate and 2 percent traffic growth and the target project life. One of the more interesting features of the Alberta selection guidelines involves the selection of the accident cost. Appendix C2 provides a table of all the bridge railings accepted for use in Alberta and there is a table with the severity index for each bridge railing by the speed limit of the road the bridge is located on. A portion of the table is shown in Figure 16. The user of these specifications would select a particular bridge railing, find the severity index for the intended speed limit and then use that severity index to look up the accident costs. Once the present worth of the accident costs and the present worth of the upgrading costs are known for each alternative, the alternative with the smallest total present worth is selected. TPW = PWCC+PWUC where: PWCC = Present worth of the crash costs (i.e., see equation above), PWUC = Present worth of the upgrade costs and TPW = Total present worth.

49 Table 11. Bridge Height and Occupancy Factors Bridge Height Above Ground (m) Low Occupancy Land use High- Occupancy Land Use <5 0.7 0.7 6 0.7 0.8 7 0.7 0.9 8 0.7 1 9 0.8 1.15 10 0.95 1.25 11 1.05 1.35 12 1.2 1.5 13 1.3 1.6 14 1.45 1.7 15 1.55 1.85 16 1.7 1.95 17 1.8 2.05 18 1.95 2.2 19 2.05 2.3 20 2.2 2.4 ≥24 2.7 2.85

50 Figure 16. Portion of the Alberta bridge rail severity index selection table. [Alberta11b]

51 Germany The bridge railing selection guidelines for Germany are provided in RPS 2008 in section 3.5.1.1 and are summarized below in Table 12. The German selection guidelines generally segregate bridges into those that pass over sensitive areas, populated areas or other transportation infrastructure. When the bridge crosses over a more sensitive area, the containment level is increased. The basic containment level for low speed roads is H1 which is essentially like Report 350 TL3. The highest containment level specified is H4b which is well in excess of Report 350 TL5 or, for that matter, TL6. The selection is based on the area crossed over and the speed and traffic volume of the roadway. Table 12. German bridge railing selection guidelines. Traffic Characteristics Bridge Characteristics Speed > 100 km/h Speed ≤ 100 km/h AADT > 500 vpd Speed ≤ 100 km/h AADT ≤ 500 vpd Speed ≤ 50 km/h Bridges with dangerous areas beneath like: • Explosive chemical plants , • High-density areas, • High-speed rail tracks with speeds > 160 km/h • Two-lane roads H4b H2 H2 H1 Other cases H2 H2 H1 Italy The Italian Ministry of Infrastructures and Transportation instructs designers on the selection of roadside barriers and bridge railings in a decree titled “Update of technical instructions for the design, approval and use of road safety barriers, and technical regulations for testing road safety barriers.”[Italy11] The decree references the containment levels specified in EN 1317, traffic is subdivided into four categories by volume and heavy vehicle percentage, as shown in Table 13. Table 14 is then used to determine the appropriate application of bridge railing.

52 Table 13. Composition of Traffic by Category. [Italy11] Type of traffic ADT* % of vehicles weighing > 3.5t I 1000 or fewer Any I >1000 5 or lower II >1000 5 to 15 III >1000 >15 *ADT: Average Daily Traffic annually in both directions. Table 14. Bridge Railings and Other Roadside Barriers by Traffic Category. [Italy11] Type of barrier Type of traffic Medians Roadside Barriers Bridge Railings (1) Freeways (A) and main state and local highways (B) I H2 H1 H2 II H3 H2 H3 III H3-H4 (2) H2-H3 (2) H3-H4 (2) Extra-urban and secondary highways (C) and ring roads (D) I H1 N2 H2 II H2 H1 H2 III H2 H2 H3 Urban city street (E) and local roads (F) I N2 N1 H2 II H1 N2 H2 III H2 H1 H2 (1) Bridges or viaducts are defined as structures crossing a space of more than 10 meters; structures crossing less space are considered equivalent to roadsides. (2) The choice between the two classes is decided by the project engineer. A review of Table 14 indicates that freeways generally have a higher performance level bridge railing for any given traffic category then the other roadway types. Generally speaking, EN1317 H2 barriers are more or less equivalent to TL4. The H4 barriers are approximately equivalent to TL5 barriers in the U.S. Therefore, the basic default condition in Italy is a TL4 bridge railing with TL5 bridge railings used on roadways with AADTs greater than 1000 vehicles/day and a percent of trucks greater than 15. United Kingdom Unfortunately, the UK has experienced a number of dramatic and catastrophic crashes involving vehicles leaving bridges and falling onto railway tracks. The 2001 Selby rail crash, for example, occurred on 28 February 2001 in Great Heck near Selby in North Yorkshire. A Land Rover towing a trailer struck an approach guardrail and came to rest below the bridge on the East Coast Main Line. The vehicle was struck by the Newcastle to London King’s Cross train at over 120 mi/hr resulting in the derailment of the train; 10 people were killed and 82 were injured in the crash. [Wainwright02]

53 While the Selby crash involved the approach guardrails to the bridge, there have been other very similar cases involving penetrating bridge rails over rail lines in the UK. For example, a concrete mixer truck penetrated a brick bridge parapet near the town of Oxshott in Surey, England at about 3:30 pm on November 5th, 2010. [RAIB11] The truck landed on top of the sixth car of the Guildford to London Waterloo train. The truck driver was seriously injured; one train passenger was seriously injured and five other sustained minor injuries in the crash. Fortunately, no one was killed in the crash. As a result of the Selby and Oxshott crashes, the UK Ministry of Transport re- evaluated its design guidelines for barriers. The intent was to develop more explicit detailed risk-based design guidelines. The United Kingdom has implemented a performance-based design process, supported by software coded in Excel and based on the Design Manual for Roads and Bridges. [UK06] This process applies to the selection of bridge railings with some pre-established minimum containment levels which reference EN1317. These minimum Containment Levels are provided for bridge railings over or adjacent to roads unless the Road Restraint Risk Assessment Process (RRRAP) (i.e., the UK cost/benefit analysis process similar in some senses to RSAP) dictates a higher containment level: • On roads with a speed limit of 50 mph or more: 1. Normal Containment Level = N2 2. Higher Containment Level = H2 3. Very High Containment Level = H4a • On roads with a speed limit of less than 50 mph: 1. Normal Containment Level = N1 2. Normal Containment Level = N2 3. Higher Containment Level = H2 4. Very High Containment Level = H4a Other than in Northern Ireland, at minimum a very high containment Level (H4a) bridge railing is used on new bridges and structures carrying a road over or adjacent to a railway. The “highest practicable containment level that can be achieved without undue cost,” as determined by the UK RRRAP, is provided for retrofit bridge railings over or adjacent to a railway. Within Northern Ireland, the minimum bridge railing containment level is normal containment level (N2) when the road is over or adjacent to a railway. When a higher containment level is justified through the use of the performance-based design process (i.e., RRRAP), “the level of provision must be confirmed with the Overseeing Organization and the Railway Authority.” On an existing bridge that is not over or adjacent to a railway, the Containment Level requirements are as follows: • Where the existing bridge or structure can support a bridge railing with a containment level derived from the RRRAP, this level of containment must be provided.

54 • Where the existing bridge or structure cannot meet the containment level derived from RRRAP, further assessment is conducted to determine the level of containment possible without strengthening. • If the risks associated with the provision of the lower level of containment determined through the RRRAP, then the lower containment level is provided. If not, the bridge is improved to provide a higher containment bridge railing. Summary Table 15 through Table 18 shows a summary of all the bridge selection guidelines discussed in the previous sections. These tables summarize the guidelines by highway function, heavy vehicle accommodation, a combination of highway design selections and by geometric design factors. The European standards have been included in these summary tables using the following approximate equivalency between EN 1317 containment levels and Report 350 test levels. • EN1317 containment level N1 is approximately equivalent to TL3; • EN1317 containment level H2 is approximately equivalent to TL4; and • EN1317 containment level H4a is approximately equivalent to TL5. Most States allow TL2 railings for non-NHS roadways while TL4 is generally the preferred minimum for NHS roadways. TL5 railings are generally specified for roads with heavy truck traffic, however the different states define heavy truck traffic differently and a number of States recommend TL5 bridge railings on all Interstate highway applications. TL5 railing are also recommended for some sharp horizontal curves with varying definitions of sharp, sometimes with no definition of sharp at all.

55 Table 15. Summary of R350 Bridge Railing Selection by Highway Function. St at e/ Co un try Ba se d on p re se nt w or th o f c ra sh an d up gr ad in g co sts (m in ). M in u se d fo r r et ro fit d es ig ns . M in th at e xi sti ng m ay re m ai n. M in fo r N on -N H S ro ad w ay s M in fo r N H S ro ad w ay s M in fo r p er m an en t in sta lla tio ns . U se d fo r a ll ne w b rid ge s M in fo r a ll Se co nd ar y hi gh w ay s Lo ca l a nd c ol le ct or ro ad s w ith fa vo ra bl e sit e co nd iti on s. H ig h- sp ee d hi gh w ay s, ex pr es sw ay s a nd in te rs ta te s Fo r e xi t r am ps . W or k zo ne s. FL DOT TL4 TL2 NY DOT TL2 TL4 TL2 NJ DOT TL1 TL4 TL5 RI DOT TL4 TL5 <TL4 MA DOT TL2 TL4 TL5 TL5 OH DOT TL2 TL3 NC DOT TL2 TL3 IL DOT TL4 NV DOT TL4 TL3 TL4 WA DOT TL2 Austria TL4 TL2 Alberta TL3 Italy TL4 TL4 UK TL2

56 Table 16. Summary of R350 Bridge Rail Selection Guidelines for Heavy Vehicle Accommodation. St at e/ Co un try Sm al l n um be r o f h ea vy ve hi cl es is e xp ec te d. Th e vo lu m e of tr uc k tra ffi c is un us ua lly h ig h. H ea vy v eh ic le c on ta in m en t H ig h- sp ee d hi gh w ay s w / h ig h tra ffi c vo lu m e an d % T G re at er th an 1 0% T A v eh ic le p en et ra tin g w ou ld ca us e hi gh ri sk . Br id ge c ro ss in g ov er p ro te ct ed ar ea s Fr ee w ay s w ith b rid ge s cr os sin g ov er ra il lin es w ith tra in sp ee d ≥ 50 m i/h r FL DOT TL5 TL6 NY DOT TL2 TL5 NJ DOT TL5 RI DOT <TL4 TL5 TL5 IL DOT TL5 NV DOT TL3 TL5 WA DOT TL5 Austria TL4+ TL5 Table 17. Summary of R350 Bridge Rail Selection Guidelines by Combination of Selectors. St at e/ Co un try H ig h Pr ot ec tio n A re a w he re D c > 5° a nd S pe ed > 4 0 m ph . S pe ed li m it ≤ 30 m i/h r a nd da ng er ou s a re a un de r b rid ge 3 0 m i/h r ≤ S pe ed li m it ≤ 60 m i/h r, A D T ≤ 50 0 vp d an d no sp ec ia l sit ua tio n un de r b rid ge 3 0 m i/h r ≤ S pe ed li m it ≤ 60 m i/h r, an y A D T, a nd a d an ge ro us a re a un de rn ea th th e br id ge 30 m i/h r ≤ S pe ed li m it ≤ 60 m i/h r, A D T > 50 0 vp d an d N o sp ec ia l sit ua tio n un de rb rid ge 60 m i/h r ≤ S pe ed li m it an d a da ng er ou s a re a un de r b rid ge A D T <1 00 0 vp d or T ru ck s ≤ 5 % A D T > 10 00 v pd a nd T ru ck > 5 % MN DOT TL5 RI DOT TL5 Germany TL3 TL3 TL4 TL4 TL5 Italy TL4 TL4+

57 Table 18. Summary of R350 Bridge Rail Selection Guidelines by Geometric Design Considerations. St at e/ Co un try A lig nm en t h as sh ar p ho riz on ta l c ur ve . Fo r h or iz on ta l c ur ve s o f le ss th an 1 00 0 fe et . Ra di us o f c ur va tu re le ss th an 5 00 ft W he re a lig nm en t m ay re qu ire e xt ra h ei gh t. U pg ra de s > 4 % la sti ng m or e th an 4 00 m Ro ad s w ith n o em er ge nc y la ne s Tr af fic O nl y (i. e. , n o bi ke s o r p ed s) . Br id ge s w ith d es ig na te d bi ke p at h ‘T ’ i nt er se ct io ns o n br id ge . Fo r s pe ed s ≤ 4 0 m ph . Sp ee ds ≥ 5 0 m ph FL DOT TL5 MN DOT TL4 TL5 TL2 NJ DOT TL5 RI DOT TL5 NC DOT TL2 IL DOT TL5 NV DOT TL6 WA DOT TL5 TL5 TL5 Austria TL4+ TL4+ TL4+ Crash Data Studies FHWA Narrow Bridge Study The FHWA sponsored research at Southwest Research Institute in the early 1980’s to examine crash characteristics at narrow bridge sites. [FHWA83] The research, which was performed by Mak and Calcote, was intended to determine the extent of the crash problem associated with narrow bridge sites and collect statistics on the frequency, severity and site characteristics. Crashes on 11,880 bridges were collected from five states and a subfile of 1,989 bridge cases were identified for more detailed analysis. Another subfile of 124 bridge crashes were selected for in-depth analysis. The data included not only bridge railings but also approach guardrail, transitions and approach guardrail terminals so it is sometimes difficult to isolate just the bridge railing effects. Since the study was performed in the early 1980’s based on data that had been collected primarily in the 1970’s, most of the structures were built prior to the 1965 AASHTO Bridge Design Specifications. The average construction date in the population file was 1954. The results of this study, therefore, do not represent more modern crash- tested bridge railings. There was relatively little in the narrow bridge data to indicate the proportion of rollovers and penetrations and, in any case, that data would not be reflective of the types of bridge railings that are commonly installed today. Table 19 shows the distribution of barrier performance in terms of the number and percent of vehicles that were redirected, overrode, vaulted or penetrated the barrier.

58 Unfortunately, the authors did not separate out different barrier types so it is believed that the data in Table 19 include bridge railings, transitions and guardrails. Similarly, it is not clear if there is a distinction between rollover back onto the roadway or rolling over the bridge railing and off the bridge. In any case, nearly 75 percent of vehicle collisions resulted in redirection and the remaining 25 percent were a combination of penetrations, vaults and rollovers. If it is assumed that Table 19 represents mostly bridge railing impacts, this would suggest that pre-1965 bridge railings resulted in about 4 percent penetrations and 18 percent rollovers and vaults. Table 19. Barrier Performance in Narrow Bridge Crashes. [FHWA83] Barrier Performance 1st Impact 2nd Impact 3rd Impact Total No. % No. % No. % No. % Redirected 87 73.1 53 79.1 16 76.2 156 75.4 Overrode 12 10.1 9 13.4 5 23.8 26 12.6 Vaulted 10 8.4 1 1.5 0 0.0 11 5.3 Penetrated 5 4.2 3 4.5 0 0.0 8 3.9 Other 5 4.2 1 1.5 0 0.0 6 2.9 Unknown 5 -- 1 -- 1 -- 7 -- Total 124 100.0 68 100.0 22 100.0 214 100.0 Interestingly, 77 percent of the crashes involved multiple impacts where the vehicle struck and re-struck the bridge railing. While the percent redirected stays more or less around 75 percent, the percent of overrides increases from 10 to 23 percent and the percentage of penetrations decreases from 4 percent to zero. This is probably reasonable since there would be less energy available for creating a structural failure in each subsequent crash (i.e., less chance of penetration) but the impact angles and yaw rates probably increase for subsequent impacts which might promote overrides. Mak and Calcote found that the crash severity increased as the bridge length, percent of shoulder reduction and speed limit increased. The departure angle (i.e., encroachment angle) was 15 degrees or less for more than 61 percent of the cases and, as would be expected due to the small distance between the edge of travel and the bridge railing, the distance from departure to impact was less than 50 feet in 78 percent of the cases. Mak and Calcote provided a great deal of information about the encroachment conditions at narrow bridge sites including the encroachment speed, angle, lateral extent and other impact conditions. NCHRP 22-08 NCHRP Project 22-08 was the most comprehensive analysis of bridge rail performance available to-date in the literature. [Mak94] The purpose of NCHRP 22-08 was to evaluate the appropriateness of AASHTO’s 1989 GSBR. Project 22-08 attempted to determine the safety performance of existing bridge rails by examining more than 4,500 bridge rail crashes across the state of Texas. Hardcopies of the accident reports

59 were examined for all reported bridge departures (i.e., penetrations and rollovers). Stratified random samples of accident reports from all other crashes were used as a quality control check to identify the frequency of coding errors. Likewise, the age of bridge rails were examined for all bridge departures and random samples were used to identify characteristics of railings that retained impacting vehicles. This study found remarkably high bridge rail crash severities for impacts involving vehicles retained on the bridge. In fact, a total of 365 (8.1%) serious injury and fatal (A+K) crashes were associated with vehicles retained on the bridge compared to 78 (A+K) (1.7%) crashes arising from a vehicle penetrating through or going over a bridge rail. In other words, more than 4.5 times more serious injury and fatal crashes occurred when a vehicle was contained on a bridge then when the vehicle penetrated through or vaulted over the railing. This ratio of serious injury and fatal crashes when the vehicle is retained versus departing from the bridge was virtually unchanged when the analysis was limited to interstate freeways which would presumably have more modern bridge rails. These findings may indicate that societal costs of bridge rail accidents are more strongly related to bridge rail performance during redirection crashes than to the number of vehicles leaving the bridge. On the other hand, this also may be more a reflection of the generally rural character of many Texas roads. This may actually make sense since if the consequences of penetrating the railing are limited to the truck and its occupants, there is much more potential for harm when the vehicle actually stays on the road where it will interact with other vehicles. This demonstrates that the potential for harm from a redirection or a penetration/vault is very sensitive to the land use around the bridge structure. If the bridge does not pass over a transportation facility or urbanized area, the consequences of leaving the bridge for the general public would be less serious than remaining on the bridge. The importance of serious injury and fatal crashes associated with a vehicle being retained on the bridge was further reinforced when the age of the bridge rail was taken into consideration. Bridge railings designed to more modern standards, AASHTO’s 1965 or later Bridge Specifications, were found to have bridge departure rates (i.e., both rolling over the bridge railing and penetration of it) of approximately 2.9% compared to 5.9% for all vehicle types in the database as shown in Table 20. The results for trucks were even more dramatic as shown in Table 20; single-unit truck rollovers and penetrations decreased from 5.4 percent to 2.3 percent; a 57 percent reduction. Tractor-trailer truck rollovers and penetrations experienced a dramatic decrease from 24.5 percent for bridge railings designed before 1965 to 7.8 percent for those designed after 1965. Clearly, the requirements of the 1965 Bridge Specifications had a dramatic effect on reducing rollovers and penetrations of bridge railings.

60 Table 20. Penetration and rollover percentage in Texas bridge railing crashes. [Mak94] Bridge Railing Design Year All Vehicle Types (%) Single-Unit Trucks (%) Tractor- Trailer Trucks (%) Before 1965 5.9 5.4 24.5 After 1965 2.9 2.3 7.8 Reduction 51 57 68 The bridge departure rate was further reduced when hardcopies of accident reports were carefully reviewed. This hard-copy analysis found that only a third of the reported bridge departures actually involved a vehicle striking a bridge rail. The remaining crashes were found to involve vehicles penetrating through or going over a bridge approach transition, a guardrail or guardrail end. Similarly, when a subset of the data were visually inspected, many of the cases coded as single-unit trucks were, in fact, pickup trucks, utility vehicles and vans. In fact, of the 53 cases where hard-copy were reviewed only 15 actually involved trucks going through or over bridge railings. When improper coding and age of the bridge rail were taken into consideration, it was found that modern bridge rails contained approximately 99 percent of the trucks striking the bridge railing; or conversely, the rollover and penetration percentage for trucks was around one percent. Kansas Bridge Rail Study A more recent study of bridge rail crashes in Kansas found much lower severities than reported in NCHRP 22-08. [Sicking09] This study examined all bridge rail crashes on controlled-access freeways in the state of Kansas for the years 2002 through 2006. A total of 705 bridge rail crashes were identified. The combined A+K rate for bridge rail crashes was found to be 3.43% compared to 8.1% found in the Texas study. The lower crash severities observed in Kansas are believed to be related to this state’s lower accident reporting threshold compared to Texas. When hardcopies of accident reports from all serious injury and fatal crashes were examined, it was found that only one of the 24 reports improperly coded a guardrail crash as bridge rail impact. Of the 23 remaining serious injury or fatal crashes involving a bridge rail, only one involved a vehicle going through or over the railing. In fact, this crash involved a tractor-trailer breaking through a bridge rail and surprisingly, the driver was not killed when his truck fell 50 feet to land on railroad tracks below the overpass. The three reported fatal bridge rail crashes included a passenger car rollover, a light-truck that lost its driver side door, and a passenger’s head extending out of the window to strike a concrete barrier. Based on the more than 20:1 ratio between serious injury and fatal crashes involving containment versus penetration, it appears that vehicles leaving a bridge are not a major source of bridge rail crash costs in Kansas. Kansas’ controlled-access freeways primarily utilize

61 open concrete, New Jersey, and F-shape concrete bridge rails. With the exception of bridges over railroad tracks, almost all existing Kansas bridge rails are 32 inches high and fall into the TL3 category under the new MASH criteria. The relatively low capacity associated with most Kansas bridge railings (i.e., Kansas generally uses MASH TL3 bridge railings) and the infrequency of serious injury and fatal crashes associated with vehicles departing these bridges makes it very clear that bridge rail selection guidelines should not be based solely upon barrier capacity. Analysis Methods for Bridge Railing Selection There is a surprisingly long history of using benefit-cost encroachment-based computer programs in roadside safety. The 1977 Barrier Guide presented a hand- calculation method based on work by Glennon but it was not particularly practical for roadside design practitioners since there was a lot of tedious hand calculation required. In 1989, AASHTO revised, updated and expanded the 1977 Barrier Guide transforming it into the Roadside Design Guide. [AASHTO89] Appendix A of the Roadside Design Guide included a revision of the cost-effectiveness procedures and provided a computer program called Roadside to ease the calculation burden on designers and policy makers. BCAP was largely based on the Roadside method with a number of enhancement and improvements intended for use in selecting bridge railings using a cost-benefit encroachment estimation procedure. [AASHTO89] In their day, Roadside and BCAP were innovative implementations of risk-based probabilistic roadside cost-benefit design. Of course, as computer applications became more sophisticated and additional research was performed to refine and improve encroachment models, severity indices and other aspects of the procedures, it became apparent that a new computer program was needed. The resulting program, the RSAP, was completed in 2003 and documented in NCHRP Report 492 by Mak and Sicking. [Mak03] 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 computers culminated in the most recent update to RSAP in 2012, RSAPv3. [Ray12] This current research is based on simulations performed using RSAPv3, however, the predecessors to RSAPv3are discussed here because each of these software tools represent individual steps forward in the development of bridge rail selection guidelines. BCAP Like Roadside, BCAP assumes an encroachment rate of 0.0005 encroachments/mi/yr per edge of roadway and could be modified to account for grade and curvature of the roadway with adjustment factors. The encroachment model investigated encroachments by 13 vehicle types leaving the roadway at 10 different speeds and up to 12 different angles. The total crash cost associated with a design alternative was calculated by summing the crash costs for each encroachment condition

62 multiplied by their associated probabilities of occurrence. The crash costs for each encroachment were estimated based on the severity of the encroachment (i.e., the consequences of impacting the bridge rail at the prescribed encroachment conditions). BCAP used a SI scale of zero to 10 to define the severity of a predicted barrier collision. Each SI had an assumed distribution of accident outcomes ranging from property damage only (PDO) to fatality. For redirection impacts, the SI was assumed to be linearly related to the lateral acceleration of the vehicle. BCAP also assumed that any bridge railing with acceptable crash test performance would have the same severity index for redirectional collisions. For a barrier penetration, the SI was assigned a value of 7.0. For a rollover on the traffic side of the barrier, the severity was linearly correlated to impact speed. The probability density function (PDF) for encroachment speed ranged from zero to 15 mph above a reference speed which was taken as 0.9 times the highway design speed. For a given encroachment speed, the encroachment angles were varied in three degree increments from zero up to a maximum of 36 degrees. The PDF was assumed to vary linearly between these two points. A model was used to determine the maximum angle a vehicle can leave the roadway without skidding or overturning. The model included consideration of barrier offset distance, encroachment speed, tire-pavement friction, vehicle stability, and minimum turning radius. In cases in which the model precludes higher angle encroachments, the encroachment angle PDF was truncated and adjusted. BCAP assumed a straight line encroachment trajectory. The maximum extent of lateral encroachment was estimated using a constant deceleration rate of 13 ft/sec2, which is equivalent to a braking friction of 0.4. Since BCAP was intended only for bridge railing applications, there is no representation of the roadside (i.e., no slopes or other off- road hazards) and the lateral extent of encroachment is relatively unimportant since bridge rails are by definition placed relatively close to the edge of the roadway. As discussed earlier, BCAP estimates the force imposed on the bridge railing by each collision using the speed, angle and mass of the encroaching vehicle. This force estimate is then compared to the assumed capacity of the bridge railing. If the capacity is less than the impact force, the bridge rail is assumed to be completely penetrated and the vehicle is assumed to fall off the bridge. If the impact force is less than the capacity, redirection is assumed and the vehicle conditions are checked to see if rollover is likely. As discussed earlier with respect to the 1989 AASHTO GSBR recommendations, BCAP was found by Mak and Sicking to over predict barrier penetrations and under predict rollovers by a considerable margin. The penetration model used in BCAP was based on work by Olsen in NCHRP Report 149. [Olsen74] Olsen suggested that the lateral force imparted by the vehicle to the barrier could be approximated as:

63 𝐹 = 𝑊 𝑉 𝑠𝑖𝑛 𝜃2𝑔(𝐴 sin𝜃 − 𝐵2 (1 − cos 𝜃) + 𝐷) where: Flat = The average lateral deceleration of the vehicle, W = The weight of the vehicle in lbs, V = The vehicle impact velocity in ft/sec, Θ = The impact angle, A = The distance from the front of the vehicle to the center of mass in ft, B = Vehicle width in feet and D = Lateral deflection of the barrier in feet. BCAP generates a set of encroachment conditions (i.e., speed, angle and vehicle type) and this lateral force can then be calculated based on those assumed impact conditions. If the lateral impact force is greater than the capacity of a barrier, the barrier is assumed to have failed structurally. While Olsen’s model is a good simple estimator it certainly has its limits. First, it is based on estimating the impact force when damage is more properly related to strain energy. Unfortunately, while impact energy is easy to calculate (i.e., ½ mv2), the strain energy capacity of a barrier is quite difficult to calculate at least in some simplified form. Also, in developing the 1989 GSBR recommendations, it was assumed that the barrier deflection would always be zero. This is probably reasonable for rigid concrete barriers but it has the effect of under estimating the capacity of post-and-beam types of bridge railings. Another flaw with this penetration model, at least with respect to its use in BCAP, is that once capacity has been reached it is assumed the barrier is totally compromised when in fact the capacity load is really just the beginning of the failure process. The barrier may often contain and redirect the vehicle even though there are structural failures; in other words, reaching capacity does not necessarily mean the vehicle will penetrate the barrier. The rollover algorithm only is activated if the bridge railing is not penetrated. BCAP first checks to see if the capacity has been reached. If capacity has been exceeded, the vehicle penetrates the railing. If capacity has not been exceeded, the vehicle is assumed to be redirected and the rollover algorithm is activated. The rollover condition in the original BCAP is: 𝑉 = ( ) ( ) ( )( ) where: Vcr= The velocity in ft/sec that the vehicle would rollover, g = The acceleration due to gravity (i.e., 32.2 ft/s2), Hcg = The height of the vehicle center of gravity in ft,

64 Hb = The height of the barrier in feet and Θ = The impact angle. This formulation assumes that the vehicle forces act at the center of gravity of the vehicle and that the barrier forces act at the very top of the barrier. Mak and Sicking found that this equation yields critical velocity estimates that are too high so BCAP seldom predicted a rollover. Mak and Sicking modified the model by assuming the barrier forces act at the vehicle axle rather than top of the barrier and that the truck would rotate about the top of the barrier when the truck deck settled onto it during the rollover. The improved impulse- momentum model is given by: 𝑉 = ( ) ( ) where the terms are as before in addition to: d = Distance from the vehicle c.g. to the bottom edge of the truck frame in inches, Hf = Height of the center of the truck axle in inches, R= The radius of gyration of the truck and its load about the bottom corner of the truck frame. This model was validated to some extent with HVOSM and NARD and resulted in lower critical velocities and more rollovers in the BCAP analyses. As discussed earlier, the improved rollover algorithm and adjustments to the barrier capacity performed by Mak and Sicking improved the estimates of BCAP but BCAP still predicted many more crashes than comparison to the real-world data available at the time indicated. Both of these rollover models completely ignore the effect of barrier shape on vaulting and rollover by vehicles with c.g. heights lower than the barrier height. For example, many passenger cars vault over safety shaped barriers even though the height of the passenger car c.g. is lower than the barrier height. The reason is that the shape of the barrier in some shallow impact angles has the effect of launching the vehicle over the barrier. This is not accounted for in either model. BCAP used the crash costs shown in Table 21 which, by today’s standards, are very low. There is no explicit provision in the 1989 AASHTO GSBR for updating these costs or adjusting them for inflation.

65 Table 21. Crash Costs used in BCAP. [AASHTO89] Crash Severity Average Cost Fatal $500,000 Severe $110,000 Moderate $10,000 Slight $3000 Property damage only (level 2) $2500 Property damage only (level 1) $500 In summary, the BCAP was innovative and ground-breaking in many ways in its time. It used a benefit-cost approach to develop the guidelines and presented a systematic method for selecting the appropriate bridge railing. Unfortunately, some of the data in BCAP was flawed and some of the algorithms were overly conservative. The general approach was a reasonable way to select bridge railings but the assumptions, lack of data and lack of validation resulted in unrealistic recommendations. RSAP The Roadside Safety Analysis Program (RSAP) is a computer program for performing encroachment-based cost-benefit analyses on roadside designs. A key step in performing such analyses is to estimate the frequency and severity of roadside crashes for a particular roadside design where the design encompasses highway geometric features like the horizontal curvature and grade as well as the roadside features like the location and type of guardrails, the shoulder widths and the slope of the roadside. Once the frequency and severity of crashes has been estimated, the cost can be found by mapping the frequency and severity into units of dollars given the average societal cost of each expected crash. A roadside design that results in a smaller societal cost is, therefore, a safer and better design. If the reduction in crash costs over the design life of the improvement are greater than the construction and maintenance costs of the improvement the design is cost-beneficial and should be constructed. On the other hand, if the reductions in crash costs are less than the construction/maintenance cost of the improvement the project probably is not worth pursuing. Estimating the frequency and severity of crashes for a given roadside design can be challenging since all the variables are probabilistic in nature and many are not well known or understood. For example, vehicles will leave the roadway (i.e., encroach) at a variety of 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, . Likewise, not all vehicles that leave the road 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 guardrail, sign support or tree on the roadside, the severity of the crash can vary from no-injury to one with multiple fatalities. Since estimating the frequency and severity of roadside crashes involves several conditional probabilities, methods like

66 RSAP are really risk-based probabilistic analysis tools where mathematical models of probabilities and risk are manipulated in order to estimate the frequency and severity of crashes. RSAP uses these four modules to assess the cost-effectiveness of a design: • Encroachment Module, • Crash Prediction Module, • Severity Prediction Module, and • Benefit/Cost Analysis Module. The encroachment probability model is built on a series of conditional probabilities. First, given an encroachment, the crash prediction module then assesses if the encroachment would result in a crash, P(C|E). If a crash is predicted, the severity prediction module estimates the severity of the crash, P(I|C). The severity estimate of each crash is calculated using crash cost figures so the output is in units of dollars. The original version of RSAP estimated the crash costs using a Monte Carlo simulation technique that simulates tens of thousands of encroachments and predicts the frequency and severity of each simulated encroachment. For each alternative, an average annual crash cost was calculated by summing the crash costs for all the simulated crashes on each segment. These crash costs were then normalized to an annual basis. Any direct costs, (i.e., initial installation and maintenance) were also normalized using the project life and the discount rate. Similar to its predecessor, the original version of RSAP used straight line vehicle trajectories and a stored table of possible angles and speeds. The third version of RSAP (RSAPv3) was used in this research. [Ray12] Using a series of conditional probabilities, RSAPv3 first predicts the number of encroachments expected on a segment. Given an encroachment has occurred, the likelihood of a crash is assessed by examining the location of roadside features and comparing those locations to a wide variety of possible field collected vehicle paths across the roadside. If a crash is predicted (i.e., one of the possible trajectories intersects with the location of a roadside hazard), the severity is estimated and converted to units of dollars. RSAPv3 proceeds by simulating tens of thousands of encroachment trajectories 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 run-off-road crashes under NCHRP 17-22. [Mak10] After the total crash costs and the direct costs are calculated for each alternative, the concept of incremental benefit/cost (B/C) is used to determine the cost-effectiveness of the design. The B/C ratio used is as follows: B/C Ratio2-1=

67 Where: B/C Ratio 2-1= Incremental B/C ratio of alternative 2 to Alternative 1 CC1, CC2, = Annualized crash cost for Alternatives 1 and 2 DC1, DC2, = Annualized direct cost for Alternatives 1 and 2 RRRAP The RRRAP is a software program developed in the UK to aid in the implementation of the “Design Manual for Roads and Bridges,” TD 19/06. [UK06] RRRAP is a risk assessment software tool to allow engineers to explicitly assess the risks associated with crashes and compare and evaluate different alternatives. Alternatives are compared using a benefit-cost procedure much like has been used in RSAP, BCAP and Roadside. RRRAP was coded in MS Excel using extensive macros to perform the bulk of the numerical calculations. Broadly speaking, RRRAP is limited to “trunk” roadways (i.e., roadways that are not local roads and streets) with posted speed limits of 50 mi/hr and above and AADTs of 5,000 vehicles/day and greater. By default, RRRAP assesses the “null” conditions (i.e., no roadside safety features) with the basic EN 1317 containment level of N2. The user can then elect to explore other containment levels to determine if they are or are not cost- beneficial. Three vehicle classes are considered – “light” vehicles are all vehicles less than 3,000 lbs, “medium vehicles” are those weighing between 3,000 and 7,000 lbs and large goods vehicles are all vehicles weighing over 7,000 lbs. RRRAP treats bridge railings a little differently than most other roadside hazards in that the only risk that is considered is the risk of breeching (i.e., penetrating or rolling over) the bridge railing. In other words, crashes where the vehicle is contained on the bridge are not considered in the risk assessment. The assumption is that the major risk for a bridge railing is the risk to third parties (i.e., either non-road users or users of other transportation facilities beneath the bridge). The basic procedure for bridge railings in RRRAP is to start with the basic EN 1317 N2 containment level and determine the likely number of breeches for each class of vehicles. The bridge parapet breech rate is then computed by dividing the number of breeches in each vehicle category by the length of the bridge and then summing those rates to arrive at the total number breachings per year per foot of bridge. Some modification factors are applied to account for the shoulder width and type. Next the average cost of a crash is multiplied by the number of breachings considering first only the road users and then third parties. If the risk (i.e., total penetrations/mi/yr) is less than a predetermined threshold, then the alternative is acceptable. If the risk is above the minimum threshold, the process is repeated for the next containment level and the alternatives are compared by calculating the benefit-cost ratios between the alternatives until a cost-beneficial solution is achieved.

68 Risk Analysis Another approach not often used explicitly in roadside safety but common in many other types of engineering fields is risk analysis. In risk analysis the risk of experiencing a particular type of event is quantified using probabilistic models. An acceptable level of risk is established over the project life and then the system is engineered to ensure that the risk in-service is below the targeted acceptable risk. For example, a transportation agency might decide that if the risk of a severe or fatal injury over the 30-year life of the project is less than 0.05 it is acceptable. The benefit-cost method used in roadside safety is actually a risk assessment method to estimate the reduction in anticipated crash costs (i.e., the benefits) then a standard benefit-cost analysis that includes the calculated crash costs and agency costs such as construction, maintenance and repair over the life of the project. Roadside safety analysis programs like Roadside, BCAP and RSAP have always calculated the average expected cost of crashes by simulating tens of thousands of possible encroachments and then multiplying by the expected number of encroachments each year. The average crash cost is calculated as follows: 𝐶𝐶 = 1𝑁𝑀 𝑤 𝐶𝐶 where: 𝐶𝐶 = The average annual crash cost, CCij = The crash cost of encroachment i with vehicle type j, Wj = The proportion of the traffic volume accounted for by vehicle type j, N = The total number of encroachments simulated and M = The total number of different vehicle types in the traffic mix. Earlier roadside safety benefit-cost programs simply calculate the average annual crash cost “on the fly” without saving the crash cost of each encroachment but RSAPv3 saves individual terms of the summation so that the distribution of crash costs over the life of the project can be examined. Since the probability distribution of crash costs is saved, the risk of exceeding any particular crash cost (i.e., severity level) can be easily calculated. Conclusions While anecdotal from a statistical point of view, the crashes discussed earlier illustrate several interesting points. Many of the crashes involved horizontal curvature, curved on/off ramps, bridges and overpasses over other highways, hazardous materials routes, heavy vehicles, or highways with large numbers of trucks and buses. In some cases these crashes have occurred on highways specifically designated as “truck routes” or “hazardous material routes” which would seem to suggest that these highways should use barriers capable of restraining such vehicles. On the other hand, some of the crashes

69 reported by the media or investigated by NTSB occurred at sites that would likely not have been considered to be particularly susceptible to a heavy vehicle crash. While some of the crashes certainly occurred at sites with the three risk factors noted in the RDG and the 1989 AASHTO GSBR (i.e., adverse geometry, percent of trucks and adverse consequences of penetration), others do not. For example, the bridge railing involved in the Sherman, Texas crash was probably a non-crash-tested bridge railing that had been constructed long before FHWA made crash testing of bridge railings mandatory. The Sherman, Texas crash site did not have adverse geometric characteristics, did not pass over a sensitive facility or area and was not on a highway with a particularly large percentage of trucks so it is unlikely it would have warranted a higher performance railing than most other highway applications. What was really required at that particular site was a way to identify substandard and un-crash-tested bridge railings and replace them. On the other hand, the Sherman, Texas crash does point out the fact that bridges generally have a design life on the order of 50 or more years so there are likely still many bridges with pre-1965 designed bridge railings. There are likely many un-crash-tested bridge railings still on the National Highway System and there are certainly many on the State and local roadway networks that need to be identified and possibly upgraded. The NTSB has recommended the development of selection criteria for bridge railings and improved bridge railing design for larger vehicles for nearly 30 years. While there has been a great deal of research, design and testing to develop high containment bridge railings the development of selection and location criteria have lagged considerably behind so, while there are now many more higher containment bridge railings available States and designers are still largely left to their own judgment on where and when to use them. While the basic approach set out in the 1989 AASHTO GSBR was a good step forward, its guidance was still largely general and intuitive. The BCAP computer program suggested for use in assessing the cost-effectiveness of different performance level bridge railings was found to have problems with its underlying data and it was soon replaced by the more general RSAP program. Some States are satisfied with TL3 bridge railings in most situations but many other States, particularly more urbanized States, have established TL4 as the minimum acceptable test level bridge railing installed on NHS and/or interstate highways. A few States, generally with more high-volume urban highways, even require that TL5 bridge railings be installed on interstate highways (e.g., New Jersey). Only a few states have established guidelines which define adverse geometry or the percentage of trucks which would require the installation of a higher test level bridge railing. Even fewer states have a policy in place for the systematic retrofitting or replacement of substandard bridge railings, with cost often cited as a major concern when deciding to upgrade to a higher test level for new and retrofit designs. Pedestrian and bicycle accommodations, however, dictate in all reviewed state policies the use of 42 inch bridge railing regardless of costs to reduce the possibility of pedestrians or bicyclists from going over the railing.

70 There is, therefore, a need to develop more specific recommendations on where different test level bridge railings should be used. While the general guidance provided by the 1989 AASHTO GSBR and the Roadside Design Guide are sound, States and designers need more specific characteristics like what percentage of trucks constitute “large truck traffic,” or what traffic volume might be considered “high volume,” or what degree of curvature constitutes a “sharp curve.” Providing these more specific answers is one of the objectives of this research project.