Preparing for LNG by Rail Tank Car: A Review of a U.S. DOT Safety Research, Testing, and Analysis Initiative (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.

29 This chapter examines the completeness and quality of 6 of the 15 tasks in the Pipeline and Hazardous Materials Safety Administration–Federal Rail- road Administration (PHMSA–FRA) Task Force initiative that have specific relevance to the transportation of liquefied natural gas (LNG) by rail tank car because of their focus on LNG and its properties and/or the DOT-113 specification tank car authorized for LNG transportation. The six tasks are International Experience Transporting LNG by Rail, Full-Scale Impact Test- ing, Punctures and Derailment Simulation Modeling, Portable Tank Fire Testing, Worst-Case Scenarios Model, and Quantitative Risk Assessment. INTERNATIONAL EXPERIENCE TRANSPORTING LNG BY RAIL The Task Force reviewed the available literature on the transportation of LNG by rail in other countries and the regulations and guidance that these countries provide for safe transportation. The Task Force identified seven countries besides the United States that authorize the transportation of LNG by rail, although only four of them—Germany, Japan, Portugal, and Spain1—are actively transporting the product by rail and all use por- table tanks, such as International Organization for Standardization (ISO) containers, rather than tank cars. An important difference between these countries and the United States is that the railroad systems in the former are used primarily for passenger transportation, whereas the U.S. railroad 1 PHMSA states that Canada authorizes a rail tank car equivalent to the DOT-113, the TC-113. However, LNG has only been used as fuel for locomotives. 3 Tasks Specifically Relevant to Transporting LNG by Rail

30 PREPARING FOR LNG BY RAIL TANK CAR system is used mainly for the movement of freight in much larger and heavier cars and on longer trains. The Task Force reviewed the experience in Japan in the most depth, finding that 8 to 10 railcars carry LNG per day. All shipments are on flatcars in inter modal portable tanks having a proprietary design rather than in ISO containers.2 Operational regulations in Japan limit LNG shipments, like all other freight, to nighttime movements in relatively short train lengths com- mensurate with shorter emergency braking distances. All rail and shipper personnel who load and unload LNG and who are responsible for respond- ing to incidents must undergo mandatory safety training on an annual basis. Japan reports no unintended releases of LNG in transport for the past 20 years.3 The Task Force members could not find reports of major incidents in the other three countries with active LNG shipping (Germany, Portugal, and Spain); however, rail safety databases in these countries do not always record incidents by commodity type. Observations About Completeness and Quality The major differences in the freight rail systems among the United States and other countries that authorize the movement of LNG by rail limit the applicability of their experience to the U.S. context, although the committee believes the Task Force was obligated to conduct this international review given its potential to identify relevant safety-related information. In general, it appears that LNG has been moving without major inci- dents in the handful of countries where there is demand for its transporta- tion by rail and approval to do so. The Task Force’s review centered mostly on Japan’s experience and was largely cursory, in part because of the limita- tions of incident data records. However, a more thorough and detailed review of the experience transporting LNG in other countries would be difficult to justify given the rail systems and containers used, none of which are characteristic of the U.S. situation. The Task Force did, however, learn that personnel in Japan who work with the LNG shipments must undergo annual safety training. More information on these training practices specific to LNG could be instructive for U.S. railroad and shipper personnel, a pos- sible supplement to the task on Loading and Unloading Safety Assessment. 2 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Administra- tion, “FRA–PHMSA LNG by Rail Task Force Interim Report,” p. 8. 3 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Administra- tion, “International Empirical Review Task Resource,” August 13, 2020. Response from Japan Petroleum Exploration Company on FRA Questions, http://onlinepubs.trb.org/onlinepubs/ dvb/LNGrail/Intl_Review.pdf. Loss of containment has occurred, for example, because of a faulty valve.

TASKS SPECIFICALLY RELEVANT TO TRANSPORTING LNG BY RAIL 31 FULL-SCALE IMPACT TESTING The Task Force’s Full-Scale Impact Testing of a DOT-113 tank car is part of an ongoing FRA testing program intended to develop standardized testing and simulation methodologies for quantifying the puncture resistance of tank cars. The testing program originated in 2007 when The Dow Chemical Com- pany, Union Pacific Railroad, and Union Tank Car Company collaborated with FRA and the Volpe National Transportation Systems Center on the Next Generation Rail Tank Car Project4 for testing designed to be repeatable and reproducible.5 Impact testing establishes the baseline puncture resistance of a given tank car design reported in relation to speed and impactor size. The results from multiple tests on a range of tank car designs are used to establish the relative puncture resistance of different tank car designs. Test results also provide empirical data for the development and validation of impact and puncture finite element (FE) model capabilities. After validation, these capabilities are used to simulate the puncture resistance associated with various changes in impact conditions and tank design parameters. In November 2019, to support the Task Force, FRA’s research pro- gram conducted a full-scale impact test on a standard DOT-113 tank car built with an outer tank shell thickness of 7/16 inches, as consistent with design specifications effective before the July 2020 rule authorizing the rail transportation of LNG in a modified DOT-113 tank car (i.e., a stan- dard DOT-113C120W). The test entailed propelling a ram car equipped with a 12×12 inch impactor, at a velocity informed by the FE model, into the side of a tank car placed against a support wall. The tank car’s inner tank and outer shell were punctured by the impactor at a closing speed of 16.7 mph, which validated the FE model’s prediction.6 In June 2020, the researchers tested a custom-built tank car as a surrogate for the newly speci- fied DOT-113C120W9. It had an outer tank shell thickness of 9/16 inches and was made of a stronger grade of steel. This time, when traveling at a slightly higher closing speed of 17.3 mph, the impactor failed to puncture either wall of the surrogate tank car.7 Further tank car testing scheduled for 2020 and early 2021 had to be suspended due to travel restrictions during the pandemic. The testing 4 Federal Railroad Administration, “Next Generation Tank Car Project (NGRTC),” Novem- ber 13, 2019, https://railroads.dot.gov/program-areas/hazmat-transportation/next-generation- tank-car-project-ngrtc. 5 F. Gonzalez III, “FRA Hazmat Tank Car Impact Test Research,” Railway Age, March 3, 2021, https://www.railwayage.com/regulatory/fra-hazmat-tank-car-impact-test-research. 6 Federal Railroad Administration Office of Research, Development and Technology, “Full- Scale Shell Impact Test of a DOT-113 Tank Car Surrogate,” July 2020, p. 2, https://railroads. dot.gov/elibrary/full-scale-shell-impact-test-dot-113-tank-car-surrogate. 7 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Administra- tion, “FRA–PHMSA LNG by Rail Task Force Interim Report,” p. 13.

32 PREPARING FOR LNG BY RAIL TANK CAR program, however, is expected to resume in late 2021 with two additional tests with tanks holding a cryogenic liquid. The first test will use another surrogate tank car filled with liquid nitrogen. The second test will use a new DOT-113C120W9 (currently being manufactured) also filled with liquid nitrogen. In discussing this activity with the committee, the Task Force members noted that the impact testing has interdependencies with several other tasks. For instance, performance of the Worst-Case Scenarios Model, Punctures and Derailment Simulation Model, and Train Energy and Dynamics Simula- tor tasks will benefit from the empirical data and increased understanding obtained from the DOT-113 tank car’s structural response to impacts. Observations About Completeness and Quality Because of the pandemic, the most relevant impact test involving a newly specified DOT-113 car loaded with a cryogenic liquid will be conducted after this review is completed. The testing to date, however, supports use of FRA’s FE model to predict puncture resistance of a standard DOT-113 tank car and the enhanced resistance of a surrogate tank built with a stronger outer tank shell as required in the new DOT-113 specification. Indeed, testing and simulation are critical for understanding low-frequency punctures in contrast to the extensive experience with tank car service-life performance, such as failures caused by tank deterioration over time. A few clarifying points would have strengthened the Task Force’s re- porting on this task and its purpose to avoid misinterpretations. The Task Force’s description should explain that the testing is designed to measure baseline puncture resistance under controlled conditions. The measure- ments will allow for comparisons to be made among tank cars for the devel- opment of data-driven tank car design changes. It is important to recognize that the impactor’s closing speed is not intended to represent a train’s opera- tional speed when derailing. The relationship between impactor speed and operational speed is complicated by the chaotic nature of train derailments. A point requiring explanation is the choice of tank impact location. The committee views the outer tank shell circumferential closure seams as a potentially more vulnerable location to impact than the center of an outer tank shell plate or the post-weld heat treated seam of a tank. An explana- tion for not testing the impact at the closure seams would be helpful. PUNCTURES AND DERAILMENT SIMULATION MODELING The Punctures and Derailment Simulation Modeling task had two objec- tives: to estimate the number of tank cars that would derail at a given train speed, and to estimate how many of the derailed tank cars would be

TASKS SPECIFICALLY RELEVANT TO TRANSPORTING LNG BY RAIL 33 punctured given different tank car design specifications. Simulations were run for multiple scenarios where: • The derailment initiates at 30, 40, and 50 mph with the leading truck of the first car; • Terrain varies; • Loads experienced by the tank vary; • Objects impacting the tank vary (i.e., such as a coupler head or broken rail); and • Tank car designs vary. The number of punctures for the standard and newly specified DOT-113 tank cars is simulated relative to the baseline case of how the general pur- pose, single-tank DOT-111 tank car would perform in a derailment under the same conditions.8 The simulation used the modeled forces at the three initiating derail- ment speeds (30, 40, and 50 mph) to predict the likely puncture resistance of tanks having outer shell thicknesses of 7/16 inches and 9/16 inches, corresponding with the standard DOT-113C120W and the newly specified DOT-113C120W9. The model simulated 18 derailment scenarios for each of the three initiating speeds under varying conditions to estimate the col- lision impact forces. For instance, it also simulated varying terrains based on three different values for the friction between steel and soil.9 The results from the scenarios were grouped by speed to represent collision impact forces of a typical derailment at 30, 40, and 50 mph,10 which are used to calculate the performance of different design specifications. This task is part of an ongoing PHMSA-FRA research program and will be completed in winter 2021–2022 after the Full-Scale Impact Testing has concluded and its results inform the model. However, the model was used earlier to inform the requirement for DOT-113 tank cars carrying LNG to have a 9/16-inch-thick outer shell.11 That earlier model simulated the derailment of a unit train consisting of 100 LNG tank cars. The model predicted that DOT-113C120W9 tank cars would sustain 4.2 punctures in a derailment at 40 mph, compared with standard DOT-113 tank cars 8 Federal Railroad Administration Office of Research, Development and Technology, “Eval- uation of Risk Reduction from LNG Tank Car Design Improvements,” July 2020, p. 2, http:// onlinepubs.trb.org/onlinepubs/dvb/LNGrail/Pnctr_Prob_LtrReport_July2020.pdf. 9 Federal Railroad Administration Office of Research, Development and Technology, “Eval- uation of Risk Reduction from LNG Tank Car Design Improvements” (Washington, DC, July 2020), p. 3. 10 Federal Railroad Administration Office of Research, Development and Technology, p. 6. 11 Pipeline and Hazardous Materials Safety Administration, “Hazardous Materials: Lique- fied Natural Gas by Rail—Final Rule.”

34 PREPARING FOR LNG BY RAIL TANK CAR sustaining 5 punctures.12 At 50 mph, the DOT-113C120W9 tank cars would sustain 6 punctures, compared with 7.3 for the standard DOT-113 tank cars.13 It merits noting the model predicted that some of the tank cars could sustain multiple punctures so that the number of punctured tank cars could be lower than the total number of punctures.14 The simulation results relate to other tasks regarding tank car design and derailment dynamics. As discussed below, the Worst-Case Scenarios Model task uses the predicted number of punctures and derailed tank cars as input for assessing the likely consequences of an accident. The Worst-Case Scenarios Model will, in turn, inform the Quantitative Risk Assessment guidance being developed by the Task Force. As noted above, refinement of the Punctures and Derailment Simulation Modeling awaits empirical data from the Full-Scale Impact Testing scheduled for later in 2021. Observations About Completeness and Quality This task provides qualitative and quantitative information for understand- ing the factors that affect a tank’s ability to resist puncture. The results provide support for the proposition that the DOT-113C120W9’s thicker outer tank shell is more robust than the standard DOT-113 design, and thus more likely to resist loss of containment in a derailment. The committee recognizes that the simulation did not intend to include all possible derailments scenarios as there may be diminishing returns when increasing the granularity of a model. However, more discussion of the vari- ous conditions and factors that can be important for predicting the number of derailed cars is warranted. Examples include: • Track type and class; • Track grade and curvature; • Abrupt changes in the track stiffness (e.g., grade crossings, bridges, and other track structures or features); • Effects of buffer car size on the derailment forces; • Location of initiation of derailment; and • Planar (2D) versus space (3D) kinematics. 12 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Adminis- tration, “Punctures and Derailment Modeling Task Resource,” April 6, 2020, pp. 5–7, http:// onlinepubs.trb.org/onlinepubs/dvb/LNGrail/PunctureDerailment_Sim.pdf. 13 Federal Railroad Administration Office of Research, Development and Technology, “Evaluation of Risk Reduction from LNG Tank Car Design Improvements,” p. 13. 14 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Adminis- tration, “Punctures and Derailment Modeling Task Resource,” p. 3.

TASKS SPECIFICALLY RELEVANT TO TRANSPORTING LNG BY RAIL 35 Without knowing these conditions, some of the results presented from the simulation would be difficult to verify or reproduce. Reducing the uncer- tainty about these parameter choices would build confidence in the model’s applicability. More explanation for some of the scenario choices would also be helpful, for instance, by giving the reason that derailment forces are always presumed to be located at the leading truck of the first car, which is not valid for all derailments. PORTABLE TANK FIRE TESTING This two-phased task evaluates how an ISO container (a UN-T75 portable tank) would perform when exposed to a fire in a derailment scenario. The UN-T75 portable tank was selected because its shares the same basic design as the DOT-113C120W9 tank car, as both have an inner tank enclosed by an outer tank shell with a vacuum-insulating space in between. The testing’s purpose is to predict the performance in a fire of the DOT-113’s tank, pres- sure relief valve (PRV) system, and tank insulation in preventing a boiling liquid expanding vapor explosion (BLEVE) that could cause severe harm to emergency responders, bystanders, and property as a result of the fireball and shrapnel from tank fragmentation. The testing under the first phase has been completed, but additional analysis of Phase 1 data remains to be done. The second phase has been delayed because of the pandemic. In the first phase, the portable tank was filled with liquid nitrogen, placed on a rail flatcar, and exposed for 2 hours and 35 minutes to liquefied petroleum gas (LPG) burners configured to simulate a rectangular pool fire. LNG was not used as the fuel for the pool fire because of the difficulty of handling it in the test environment. Instead, LPG was selected because it has combustion characteristics similar to LNG. In addition to the fire testing, this phase entailed modeling activities. The results from the Phase 1 test indicated that the PRV system worked properly to evacuate the vaporizing liquid nitrogen to prevent a tank rup- ture or BLEVE; but the insulation degraded, the tank vacuum was lost, and the steel in the flatcar underneath the tank exceeded its softening tempera- ture and sagged into the burners and piping. The PRV system continued to release jets of nitrogen for nearly 4 hours after the fire test ended.15 Phase 1 modeling results predicted that the failure pressure for a heated tank will be lower than for a tank at ambient temperature. The model also indicated that more testing is needed to characterize the performance of the tank’s annular vacuum when exposed to heat and to understand how heat flux to 15 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Admin- istration, “Portable Tank Fire-Testing Task Resource,” August 13, 2020, p. 67, http:// onlinepubs.trb.org/onlinepubs/dvb/LNGrail/UNT75_Fire_Test.pdf.

36 PREPARING FOR LNG BY RAIL TANK CAR the tank is affected by experimental design features such as wind, shielding from the flatcar, and burner size.16 The report from this testing identified additional data analysis needed to support the second phase of testing, including measurements of inter- nal tank conditions, degradation of the insulation material, the heat flux applied to the tank, and the flow out of the PRVs as a function of time and internal pressure. The plan for the second phase of testing is to fill the portable tank with LNG and place it on a flatcar exposed to an LPG pool fire. The aim of Phase 2 is to more thoroughly characterize the effect of wind on heat flux to the tank and performance of the thermal insulation. Another aim is to characterize the formation of a vapor cloud from leaks and a torch fire from the ignition of vapors released from the PRV system. This second phase will include a tank filled with LNG to various levels and subjected to extreme heat to determine the effects on tank materials, internal pressure, and fluid level. Additional work could include use of data collected in Phase 1 to validate modeling methods (e.g., internal temperature and pressure) and the rapid phase change and expansion of LNG as it relates to causation of a BLEVE, but a detailed analysis plan for this possible work was not presented. Knowledge of the types and extent of fire hazards gained from this testing will inform the level of risk in the task on Quantitative Risk Assess- ment and determinations about the likelihood of certain hazards in the Worst-Case Scenarios Model. The test results will also provide insights about resources needed for emergency response as part of the Educational and Outreach Plan. Observations About Completeness and Quality Characterizing the degree to which the survivability of a portable tank en- gulfed in an LPG pool fire predicts the survivability of a DOT-113C120W9 tank car engulfed in an LNG pool fire is difficult to do for various reasons. One reason is that the two tanks differ with regard to thermal mass, con- struction, and insulation. Another reason is that the heat flux from an LNG pool fire is 2–3 times greater than the heat flux to an engulfed object from an LPG pool fire.17 Acknowledgment of these two major differences and their implications is needed for making valid characterizations. Likewise, although the Task Force stated that the design characteristics are basically 16 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Admin- istration, pp. 67–68. 17 Blanchat et al., “The Phoenix Series Large Scale LNG Pool Fire Experiments”; Pipeline and Hazardous Materials Safety Administration, “UN-T75 Portable Tank Fire-Testing Task Resource,” fig. 42. The committee compared Figure 42 in the section on “Fire Performance of Cryogenic ISO UN-T75” in this resource document with the work by Blanchat et al.

TASKS SPECIFICALLY RELEVANT TO TRANSPORTING LNG BY RAIL 37 the same, details about similarities and differences in the PRV systems of the portable tank and DOT-113 tank car are needed for this purpose. Recognizing that Phase 2 has not commenced, the committee believes that there is time to make changes to the planned testing that could improve the quality of the data collected and their analysis. Changes that should be made include: • Increasing the pool size, changing the pool shape to circular, and moving the pan upwind relative to the portable tank. These changes will enhance full fire engulfment and ensure consistent impinge- ment of the tank under wind conditions;18 • Removing the flatcar from the experimental setup. This will allow for better assessment of the thermal response of the tank because the flatcar offers thermal protection;19,20 • Placing the portable tank in an orientation (approximately 120° rollover angle) where the pressure relief valve will vent liquid, based on National Transportation Safety Board reporting,21,22 which presents a more severe test of the PRV system;23 • Using LNG as the fuel, stored in a protected location in case of a BLEVE, for the pool fire because an LNG pool fire has approxi- mately 2–3 times higher heat flux loads to engulfed objects than LPG;24,25 18 P.A. Croce, K.S. Mudan, and J. Moorhouse, “Thermal Radiation from LNG (Liquefied Natural Gas) Trench Fires. Volume 1. Main Report. Final Report, September 1982–September 1984,” September 1, 1984, https://www.osti.gov/biblio/6192926-thermal-radiation-from-lng- liquefied-natural-gas-trench-fires-volume-main-report-final-report-september-september. 19 The committee notes that FRA stated that the height of the flatcar deck places the portable tank at the point where the heat is greater than if the portable tank were immersed in a pool fire because the temperature is hottest at 1–2 meters above the pool (personal communication). 20 Federal Railroad Administration Office of Research, Development and Technology, “Temperatures, Pressures, and Liquid Levels of Tank Cars Engulfed in Fires: Volume 1, Results of Parametric Analyses,” June 1984, https://railroads.dot.gov/elibrary/temperatures-pressures- and-liquid-levels-tank-cars-engulfed-fires-volume-1-results. 21 National Transportation Safety Board, “Derailment of Norfolk Southern Railway Company Train 68QB119 with Release of Hazardous Materials and Fire New Brighton, Pennsylvania, October 20, 2006,” Accident Report, May 13, 2008, fig. 1, https://www.ntsb. gov/investigations/AccidentReports/Reports/RAR0802.pdf. 22 National Transportation Safety Board, “Derailment of CN Freight Train U70691-18 with Subsequent Hazardous Materials Release and Fire, Cherry Valley, Illinois, June 19, 2009,” Accident Report, Washington, DC, February 14, 2012, fig. 3, https://www.ntsb.gov/ investigations/AccidentReports/Pages/RAR1201.aspx. 23 Federal Railroad Administration Office of Research, Development and Technology, “Tem- peratures, Pressures, and Liquid Levels of Tank Cars Engulfed in Fires: Volume 1, Results of Parametric Analyses.” 24 G.A. Mizner and J.A. Eyre, “Large-Scale LNG and LPG Pool Fires.” 25 Blanchat et al., “The Phoenix Series Large Scale LNG Pool Fire Experiments.”

38 PREPARING FOR LNG BY RAIL TANK CAR • Performing the fire test for 100 minutes and a torch fire for 30 min- utes in conformance with regulation to provide consistency in evaluating fire performance according to tank car specifications;26 • Evaluating an LNG fireball and tank fragmentation in the event of a BLEVE to prepare emergency response personnel for combustion and non-combustion hazards; • Assessing the potential for cryogenic damage cascading to adja- cent tanks by evaluating topography surrounding the rail tracks that could support pool formation because a cryogenically dam- aged tank impacted by a pool fire can potentially alter PRV performance;27,28 and • Performing a non-destructive thermal test of an LNG-laden DOT- 113C120W9 tank car using radiant heat panels to gather data on the internal thermal response of the tank that could be used to assess model performance with regard to predicting heat transfer and multi-phase behavior. WORST-CASE SCENARIOS MODEL The Worst-Case Scenarios Model task posited four hazard scenarios involv- ing a release of LNG simultaneously from five tank cars damaged from a unit train derailment accident. The task simulates the severity of the con- sequences and characteristics of the hazards under high wind conditions to understand how to mitigate harm to emergency response personnel. The hazard scenarios include (a) an unconfined LNG pool spread following a release; (b) dispersion of unignited vapors from the spreading pool, which has the potential to result in a flash fire; (c) thermal radiant heat emitted by pool fire exposing an area to the hazard; and (d) a fireball resulting from a BLEVE exposing an area to the hazard. The simulation results from the scenarios were presented to the committee during meetings, but a final report was not available for the committee’s review. The modeling methodology for the four hazard scenarios uses Monte Carlo simulation to generate probabilities for harm to health and the envi- ronment along segments of an LNG transportation route based on various 26 Pipeline and Hazardous Materials Safety Administration, “Thermal Protection Systems,” 49 CFR § 179.18, n.d., https://www.govinfo.gov/content/pkg/CFR-2013-title49-vol3/pdf/CFR- 2013-title49-vol3-sec179-18.pdf. 27 J. McKinley, “Strength, Creep, and Toughness of Two Tank Car Steels: TC128B and A516-70,” Transport Canada, 2019, https://epe.lac-bac.gc.ca/100/201/301/weekly_ acquisitions_ list-ef/2020/20-39/publications.gc.ca/collections/collection_2020/tc/T86-56-2019-eng.pdf. 28 B.W. Williams et al., “Capturing Variability in the Fracture Response of TC128B Steel Using Damage Mechanics,” vol. 28, 2020, pp. 1024–1038, https://doi.org/10.1016/j. prostr.2020.11.118.

TASKS SPECIFICALLY RELEVANT TO TRANSPORTING LNG BY RAIL 39 factors such as population density and the areas affected by the hazards. The model assumes that a unit train consisting of standard DOT-113 tank cars (with an outer shell thickness of 7/16 inches) that derails will experi- ence five punctures in the bottom of the tanks.29 The tank punctures are created by an object comparable in size to a coupler (i.e., 12×12 inches) and lead to the full contents of the tank being released in less than 1 minute. (Some of these assumptions derive from the results of other tasks discussed above.) Other values in the model (e.g., derailment speed, number of tank cars derailed and punctured, derailment impact area) are based on historical data from derailments of unit trains that were hauling crude oil or ethanol. For modeling the hazard areas, heat flux, and overpressure blast, the Task Force used the explosive TNT as a proxy for LNG.30 The Worst-Case Scenarios Model of the pool spread predicts that the released LNG will cover a circular area having a diameter of 95 meters, which amounts to a pool having a maximum radius of about 10 times the length across a five–tank car pileup. The concentration of LNG vapors suffi- cient to remain a combustion hazard (i.e., the lower flammability limit) was estimated to extend as far as 2,380 meters (1.5 miles) from the source of the release. The radiant heat flux31 emitted by a pool fire and a fireball from a BLEVE were estimated to be 165 kW/m2 and 250 kW/m2, respectively.32 This heat exposure would cause a person to experience second degree burns at a distance of 670 meters (0.4 miles) from the longer burning pool fire (60 seconds) and 230 meters (0.14 miles) from a fireball (15 seconds) measured from the fire’s center.33 Analyses that have not been completed include evaluating the physical and thermal conditions that can result in a BLEVE and unit train derail- ment modeling assuming that the tank cars are the newly specified DOT- 113C120W9 design (i.e., outer-shell thickness of 9/16 inches). The Task Force intends to do more work to understand the BLEVE scenario. How- ever, Task Force members expressed the view that it is highly unlikely that an undamaged DOT-113 tank car involved in a derailment would fail due to a BLEVE. The reason for their confidence is that the DOT-113 tank car is designed to avoid a BLEVE by meeting the standards for loading pressure 29 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Admin- istration, “Worst-Case Scenarios Model Task Resource,” August 13, 2020, p. 3, http:// onlinepubs.trb.org/onlinepubs/dvb/LNGrail/WorstCase_Model.pdf. 30 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Admin- istration, pp. 113–114. 31 Radiant heat flux is the rate of heat transferred as thermal radiation for a unit area. 32 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Adminis- tration, “Worst-Case Scenarios Model Task Resource,” p. 130. 33 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Adminis- tration, “FRA–PHMSA LNG by Rail Task Force Interim Report,” pp. 16–17.

40 PREPARING FOR LNG BY RAIL TANK CAR requirements for cryogenic materials, redundant pressure relief systems (valves and safety vents), and insulation systems (an insulating wrap and annular vacuum space to prevent external heat reaching the inner tank). The Worst-Case Scenarios Model interrelates with three other tasks that are specifically relevant to LNG transport by rail and one more broadly in support of railroad safety. The modeling of punctures and derailments provides the number of punctures given a derailment scenario. That model- ing, in turn, is supported by the portable tank fire testing and DOT-113 impact-testing tasks. In 2022, the Worst-Case Scenarios Model task will inform an analysis of the modal conversion for LNG transportation be- tween truck and rail. Observations About Completeness and Quality The quality of the modeling in this task could not be fully assessed in the absence of a detailed final report and the results of pending tasks that are related. However, based on what could be gleaned from Task Force presen- tations, the committee believes that there are opportunities to improve the design and execution of the task. With regard to the scenario assumptions about the number of DOT-113 tank punctures in a derailment, the Task Force relied on the predictions from the Punctures and Derailment Simulation Modeling task discussed above. It is important to point out, however, that the puncture analysis did not predict absolute values for the number of punctures of DOT-113 tank cars, but rather the number of punctures relative to baseline modeling of punctures sustained by a DOT-111 tank car. The committee, therefore, recommends the following considerations to improve the quality of the punctures and derailment simulation as it pertains to the worst-case sce- narios modeling: • Providing upper bound values of predicted number of punctures to identify the worst-case release rather than using nominal values; and • Using a train speed of 50 mph for the predicted number of 7.3 punctures, rather than the 40 mph used, because it was the upper speed evaluated in the puncture and derailment simulation. While the hazard analyses for pool spread, vapor dispersion, area encompassed by a flash fire, and thermal radiant heat emitted by pool fire have been completed, pending analyses include the conditions that would cause a BLEVE and the modeling of a unit train comprised of the newly specified DOT-113C120W9 tank cars. The worst-case scenarios simula- tion results could be updated on the basis of the results from the pending

TASKS SPECIFICALLY RELEVANT TO TRANSPORTING LNG BY RAIL 41 fire testing and full-scale impact testing of the DOT-113C120W9 tank car. Furthermore, for completeness of the task, the Task Force should make the following enhancements: • Evaluating the heat flux from a jet fire from a punctured tank and impinging on an adjacent tank to assess the potential for cascading damage from this combustion event because it was missing from the hazards initially listed in the model; • Evaluating the potential for valve damage and ruptured lines that could contribute to a release, which was not included in the initial model; • Evaluating the total amount of LNG that potentially could be re- leased from cascading damage to adjacent tank cars from partial submersion in an unignited pool of LNG and/or partial exposure to the heat flux from a pool fire; • Evaluating the potential hazard to emergency responders of a rapid phase transition from an LNG spill onto a body of water, consider- ing that track infrastructure commonly runs along rivers; • Evaluating explosion hazards from an unignited spill of LNG re- sulting in vapor dispersion in an environment with confined or congested spaces because the model only represents the scenario occurring in an open area without any factors that could affect the spread of the pool or vapor cloud; • Discussing fire propagation in a high-density environment and the potential resulting semi-confined, confined, or congested haz- ard areas because the model does not account for the population exposed; • Using the thermal radiant heat flux range of 475–540 kW/m2 in- stead of the 250 kW/m2 currently used in the model for an LNG fireball when determining the hazard area because the higher value is LNG-specific and greater than what is reported for an LPG fireball;34 and • Using an approved code (i.e., LNGFIRE3)35 that meets the re- quirements specified in regulations to determine hazard areas from thermal radiant heat flux emitted by pool fires, fireballs, and jet or torch fires (while the dispersion calculations were performed with an approved code, the models used to evaluate thermal radiant 34 Betteridge and Phillips, “Large Scale Pressurised LNG BLEVE Experiments.” 35 Federal Energy Regulatory Commission, “Recommended Parameters for Solid Flame Models for Land Based Liquefied Natural Gas Spills,” January 2013, https://www.ferc.gov/sites/ default/files/2020-04/RecommendedParametersforSolidFlameModelsforLandBasedLNGSpills. pdf. In addi tion, 49 CFR § 193.2057 (“Thermal radiation protection”) requires the use of the solid flame model LNGFIRE3 for predicting radiant heat from LNG pool fires on land.

42 PREPARING FOR LNG BY RAIL TANK CAR heat flux emitted by pool fire and fireball were not). The Task Force acknowledged the use of TNT as a proxy was a limitation of the model because it does not account for the specific proper- ties of LNG, including its temperature, vaporization, likelihood of ignition, or fire type (i.e., pool fire, flash fire from spreading vapor, and fireball). QUANTITATIVE RISK ASSESSMENT Quantitative Risk Assessment (QRA) is intended to measure risk quanti- tatively to inform the selection of operational and other controls to apply on transportation routes. The methodology entails a systematic accounting for hazards, the potential consequences, and the likelihood of their occur- rence between the origin and destination of a commodity. The Task Force noted that a QRA can provide useful insights to the main contributors to LNG transport risk, allowing mitigation measures to be focused where they are most effective in controlling or reducing risk. For example, a QRA can inform comparative analyses of alternate modes and routes to support selection of the best option for a planned movement of hazardous materials. The inputs for such an analysis include, for instance, historical accident records for derailments and releases of hazardous materials and the population data along the route. Although the Task Force initially intended to perform a QRA as part of this task, it was later clarified that the scope would be limited to the development of a framework for the requisite data and a suggested meth- odology for an effective QRA.36 The rationale given for the revised scope is that because of a lack of a historical record of past shipments of LNG by tank car and few details on planned shipments, there is insufficient data and context to conduct an analysis. With such limited information, a QRA would require unsupportable assumptions about candidate origin– destination pairs. Instead, the Task Force cited the Volpe National Trans- portation Systems Center’s review of a risk assessment submitted for a 2019 special permit granted by PHMSA for the shipment of LNG by DOT-113 tank cars over a rail transportation route between Wyalusing, Pennsylvania, and Gibbstown, New Jersey.37 The findings from this review were used by the Task Force to develop recommendations for actions needed to construct an improved framework for a QRA for LNG transport by rail. 36 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Adminis- tration, “Quantitative Risk Assessment Task Resource,” March 15, 2018, pp. 13–24, http:// onlinepubs.trb.org/onlinepubs/dvb/LNGrail/QRA1.pdf. 37 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Admin- istration, pp. 25–34.

TASKS SPECIFICALLY RELEVANT TO TRANSPORTING LNG BY RAIL 43 The Volpe Center determined that the methodology in the QRA for the special permit was reasonable and achieved several of its stated goals. The review also concluded, however, that the analysis was limited in sev- eral areas, including a comparison with the risks involved in other modes long-permitted to transport LNG (i.e., marine tankers and trucks) to use as a benchmark for an equivalent level of safety and the consideration of hazards and failure modes other than a release caused by a derailment. Drawing on these findings, a white paper prepared by the Task Force for its QRA framework guidance recommended the following38: • Further structural analysis of the puncture resistance of the stan- dard DOT-113 tank car to understand the conditions during a derailment that could cause a puncture of the inner tank; • Enhancement of train dynamics modeling, such as the Train Energy and Dynamics Simulator, to better anticipate the amount of LNG that would be released in the case of a puncture, which would be accomplished by estimating the probabilities of different puncture sizes of the inner tank; • Estimation of the likely frequency of LNG spills at loading and unloading facilities by review of LNG trucking operations and the loading and unloading of tank cars in chemical and petrochemical facilities; • Modeling the amount of the time an LNG release will result in various hazards such as pool spread, vapor dispersion with flash fire, pool fire, fireball, or a BLEVE so as to know when a particular hazard is possible and its likelihood; and • Examination of the factors that could contribute to a BLEVE in a DOT-113 tank car with a vacuum-insulated inner tank enclosed in an outer tank, which, as noted above regarding worst-case sce- narios modeling, is not currently understood. Observations About Completeness and Quality In re-scoping this task, the Task Force identified important areas where additional work is needed to strengthen the applicability of QRA to LNG rail transport, based on an assessment of the QRA submitted with the 2019 special permit for LNG shipments between Pennsylvania and New Jersey. However, the Task Force’s recommendations, as cited above, need to be implemented to improve the effectiveness of a QRA for LNG transportation by rail. A fully executed QRA can serve as a mechanism for pulling together 38 Pipeline and Hazardous Materials Safety Administration and Federal Railroad Admin- istration, p. 24.

44 PREPARING FOR LNG BY RAIL TANK CAR the results of other Task Force evaluations, such as the tasks for puncture and derailment analyses, tank car performance in accidents, worst-case scenario modeling, loading and unloading operations, route analyses, and security evaluations. Moreover, a QRA can serve as a means of continu- ous improvement by integrating information as new results from the Task Force’s program and other safety assurance activities progress. To be sure, the QRA submitted for the special permit would have been enhanced by integrating the results of the interdependent tasks mentioned above. How- ever, there is no evidence to date that the Task Force intends to update that QRA to integrate the results of the other task evaluations.