Safety Regulation for Small LPG Distribution Systems (2018)

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.

30 This chapter describes liquefied petroleum gas’s (LPG’s) physical proper- ties and hazard characteristics that must be controlled, examines available data on LPG pipeline incidents, and reviews several incidents that stand out from the data. The incident records reported by pipeline operators and maintained by the U.S. Department of Transportation’s Pipeline and Hazardous Materials Safety Administration (PHMSA) do not differentiate between smaller and larger systems subject to federal regulation and thus cannot be used to identify only those incidents involving systems serving 100 or fewer customers. The U.S. Fire Administration’s (USFA’s) National Fire Incident Reporting System (NFIRS) provides a secondary source of records on LPG incidents, but it also lacks identifiers for incidents involv- ing pipeline distribution systems, and thus may include incidents involving single-user systems that are non-jurisdictional (i.e., not regulated by the federal government). An annex to the chapter provides more information on these two databases and their limitations for the purposes of this study. While these limitations complicate assessments of the safety performance of small LPG pipeline systems specifically, the small number of reported LPG transportation incidents generally suggests that industry and government measures are working to assure the safety of LPG pipeline systems. Never- theless, the events and circumstances that gave rise to the few consequential LPG pipeline incidents that have occurred over the past 30 years show the importance of these measures. 3 Hazard Characteristics and Safety Performance

HAZARD CHARACTERISTICS AND SAFETY PERFORMANCE 31 LPG PROPERTIES AND HAZARD CHARACTERISTICS LPG may contain butane, butylene, isobutane, isobutylene, propane, pro- pylene, or mixes of each compound. Like natural gas, propane and butane are colorless and odorless, and therefore chemical odorants such as ethyl mercaptan are added to LPG to aid in the detection of escaping gas. LPG suppliers may mix propane with other LPGs depending on whether the distribution system is located in warmer or colder climates. While each compound has distinct properties that can lead to somewhat different haz- ard characteristics, the standardized techniques and practices used for LPG storage and transportation are intended to accommodate the full range of LPGs. Because the small LPG pipeline systems that are of interest to this study are used predominantly for propane with varying concentrations of butane, this chapter focuses on the hazard characteristics and safety per- formance of propane and propane–butane mixtures when transported by pipeline. Physical properties of propane and butane relevant to their hazard characteristics are summarized in Table 3-1. As discussed in Chapter 2, end users consume LPG in a gaseous state, but suppliers transport and store it as a liquid through the use of pres- sure, cooling, or a combination of both. Propane naturally occurs as a gas because it boils at –44°F. However, because butane boils at 31°F, propane mixed with butane can create system performance and safety issues in northern states where low winter temperatures can cause gas to condense in pipelines. Condensation in a gas pipeline can interrupt the fuel flow and feed liquid to gas appliances.1 In addition, the integrity of a plastic piping system may be compromised by the presence of LPG in a liquid state. If LPG liquid is contained in a plastic line with closed valves, warmer tem- peratures can cause the liquid to expand and rupture the pipe. The force exerted by a gas on its container when it transitions from a liquid phase is a function of its vapor pressure. The higher the vapor pres- sure, the more readily a liquid will evaporate, indicative of volatility. Exert- ing about tenfold and double the force of atmospheric pressure at sea level, respectively, propane and butane are highly volatile compounds that present flammability hazards in an uncontrolled release. Because of the expansion ratio of liquid propane, 1 ft3 released from a container will volatilize to cre- ate 270 ft3 of gas.2 When mixed with air, this release can produce more than 1 Propane can also condense in pipelines under certain low-temperature and moderate- pressure conditions. For instance, a service line exposed to approximately 20 pounds per square inch gauge (psig) (34.7 psi absolute) of pressure and an atmospheric temperature of –5°F or lower can lead to propane gas condensation. National Fire Protection Association, LP-Gas Code Handbook (Quincy, MA: National Fire Protection Association, 2017), 233. 2 Gregory G. Noll and Michael S. Hildebrand, Pipeline Emergencies, 3rd edition (Chester, MD: National Association of State Fire Marshals, 2011), 12.

32 SAFETY REGULATION FOR SMALL LPG DISTRIBUTION SYSTEMS TABLE 3-1 Select Physical Properties of Propane and Butane Propane Butane Chemical formula C3H8 C4H10 Initial boiling point –44°F 31°F Vapor pressure, pounds per square inch absolute (psia) at 70°F at 100°F 145 218 32 52 Cubic feet of vapor per gallon (ft3/gal) at 60°F 36.38 31.26 Relative vapor density (air = 1) 1.50 2.01 Flash point (closed cup method) –156°F –76°F Auto-ignition temperature 871°F 761°F Flammability limits, lower 2.1% 1.55% Flammability limits, upper 9.6% 8.6% SOURCES: National Fire Protection Association, LP-Gas Code Handbook, 549–550; National Oceanic and Atmospheric Administration, “CAMEO Chemicals: Liquefied Petro leum Gas,” accessed March 20, 2018, https://cameochemicals.noaa.gov/chris/LPG.pdf. 12,000 ft3 of fuel–air mixture that will extend well beyond the immediate vicinity of the release.3 As it spreads, this mixture can form an explosive, low-hanging vapor cloud that is visible near the release area, an invisible but ignitable vapor cloud farther way, and flash fire areas just beyond the invisible vapor cloud.4 The diffusion of LPG differs from that of natural gas because of differ- ences in their vapor density. Propane and butane, unlike natural gas (mainly methane), are heavier than air. As a result, the two LPGs will sink when released, whereas natural gas, which has about half the relative vapor den- sity of air, will diffuse upward and dissipate if not contained. LPG’s higher vapor density can cause it to creep along the ground and concentrate in low-lying areas, as well as migrate through soil in the case of an under- ground leak. Highly saturated or frozen soil can create a barrier to the dissipation of gas as it travels through the ground.5 Instead of venting to the atmosphere, the gas may travel along other underground infrastructure 3 O. John Jacobus, “Odorization of Propane” (Meeting 3, Washington, DC, December 7, 2017), http://onlinepubs.trb.org/onlinepubs/Propane/Jacobus120717.pdf. 4 Hildebrand, Noll, and National Propane Gas Association, Propane Emergencies, 98. 5 Pipeline and Hazardous Materials Safety Administration, “Operations and Maintenance Enforcement Guidance: Part 192 Subparts L and M,” July 21, 2017, 64, https://www.phmsa. dot.gov/sites/phmsa.dot.gov/files/docs/regulatory-compliance/pipeline/enforcement/5776/o-m- enforcement-guidance-part-192-7-21-2017.pdf.

HAZARD CHARACTERISTICS AND SAFETY PERFORMANCE 33 into buildings, such as basements. Because LPG disperses from the spill site relatively slowly and pools, it can retain its potential as a source of fuel for a fire or explosion longer than a similar release of natural gas. Although LPG is not toxic, concentrations of the gas that displace air can also create an asphyxiation risk. The flammability of LPG arises from multiple properties, including its flammability limits, auto-ignition temperature, burning velocity, and flash point. A vapor with a wider range of flammability limits and lower temperatures associated with its flash point and auto-ignition is considered more flammable.6 The flammability limits, which describe the range of va- por concentration in air that is supportive of combustion, are low enough for even a 2 percent concentration of LPG to combust. The burning velocity of a fuel describes how quickly a fuel–air mixture will burn and flash back toward the source of ignition, which typically occurs by means of heat, flames, spark, or discharge of static electricity. Because of the high burning velocity of LPG, the flame can travel back to its source of ignition.7 Also, the flash point for propane and butane occurs at such low temperatures that these fuels would already be present as a mix of vapor and air at a concen- tration corresponding at least to the lower flammability limit at standard temperature and pressure. Knowledge of these particular properties and hazards of LPG have informed measures intended to ensure its safety as a common consumer fuel. However, major incidents—including the consequential ones discussed next—do occur at times and reveal the importance of controlling these hazards. NOTABLE LPG PIPELINE INCIDENTS Before examining the incident data for evidence of the safety performance of LPG pipelines, a brief review of a few major incidents that have occurred over the past 30 years is helpful because investigations of their causes and consequences—more than analyses of incident statistics—are often the impetus for changes in safety practices, techniques, and standards. While major incidents are rare, the following ones illustrate some of LPG’s hazard characteristics, particularly those associated with its diffusion behavior. All but one of the notable LPG incidents occurred on a jurisdictional LPG dis- tribution system; it was included because the circumstances demonstrate the 6 David Lord et al., “Literature Survey of Crude Oil Properties Relevant to Handling and Fire Safety in Transport” (Sandia National Laboratories, March 2015), 83, http://www.osti. gov/scitech/biblio/1177758. 7 National Oceanic and Atmospheric Administration, “CAMEO Chemicals: Liquefied Petro- leum Gas.”

34 SAFETY REGULATION FOR SMALL LPG DISTRIBUTION SYSTEMS significance of LPG hazard characteristics in a system configuration similar to commercial jurisdictional systems. Parkers Prairie, Minnesota On August 5, 1991, an LPG pipeline distribution system ruptured and killed an employee at a delicatessen in Parkers Prairie, Minnesota. Accord- ing to the incident narrative in a report submitted to PHMSA, the state fire marshal determined that the LPG leaked from a service line outside the store and migrated into the basement where it concentrated before igniting. This downward path of the escaping gas illustrates a hazard of LPG’s high density relative to air. San Juan, Puerto Rico On November 21, 1996, an LPG pipeline distribution system in a com- mercial district in San Juan, Puerto Rico, exploded, killing 33 people and injuring at least 69 others.8 Although the explosion involved a large juris- dictional LPG system, it provides an example of how LPG’s high vapor density (relative to air) can create a serious flammability hazard. As of July 2018, this incident was the largest cause of fatalities and injuries resulting from a gas distribution facility in the United States. The National Transportation Safety Board (NTSB) determined that the propane gas explosion was fueled by an excavation-caused leak, after backfilling and compacting soil over a water line 4 years earlier imposed excessive stresses on the plastic gas service pipe, which later caused the service pipe to fail.9 The escaping propane migrated downhill along piping through voids in the ground and under a sidewalk until it reached the basement of a six-story building.10 The gas exploded when sparked by a heating, ventilation, and air conditioning fan motor. In the days preceding the explosion, several individuals reported smelling an odor in the vicinity and inside buildings. Barholes drilled by gas company technicians to detect leaking gas were too shallow to detect the propane that had migrated to lower depths. NTSB concluded that the gas company had inaccurate maps of buried facilities, insufficiently trained employees to test for and respond to reports of potential leaks, and did not have an excavation-damage pre- 8 National Transportation Safety Board, “Pipeline Accident Report: San Juan Gas Company, Inc./Enron Corp. Propane Gas Explosion in San Juan, Puerto Rico, on November 21, 1996,” December 23, 1997, vii, https://ntsb.gov/investigations/AccidentReports/Reports/PAR9701.pdf. 9 National Transportation Safety Board, 41. 10 Stephen Barlas, “NTSB Report on San Juan Raises Broader Questions,” 39–40, accessed February 1, 2017, http://connection.ebscohost.com/c/articles/325723/ntsb-report-san-juan- raises-broader-questions.

HAZARD CHARACTERISTICS AND SAFETY PERFORMANCE 35 vention program such as a one-call notification system. One-call notifica- tion systems are intended to facilitate safe excavation through the use of communications centers that field inquiries from excavators and notify all underground infrastructure operators (who are members of the system) of impending digging.11 Snow Hill, Maryland On September 1, 2002, an explosion killed one person and injured 17 others in a residential neighborhood in Snow Hill, Maryland. A corroded LPG service line is presumed to have leaked the gas that fueled the explo- sion, though the source of ignition is unknown.12 Lengthy and heavy rain is believed to have contributed to a ground disturbance that compromised the service line.13 The operator of the jurisdictional LPG system was a natural gas utility serving the rural Eastern Shore of Maryland. Two gas utility company employees and several volunteer firefighters responded to a call about the smell of propane at a house where gas had accumulated in the basement after traveling from the ruptured service line. The incident demolished the customer’s house, where one of the gas utility employees died. The explosion necessitated evacuation of neighbors from their homes and led to the detection of LPG at three other nearby homes. Door County, Wisconsin On July 10, 2006, gas leaking from an underground LPG gas line exploded, killing two people and injuring four others.14 The explosion, which oc- curred in a resort community in Ellison Bay, Wisconsin, was fueled by gas leaking from a pipe that had been damaged by excavation 3 days earlier.15 The utility worker who caused the rupture in the pipe had been installing electrical cables and was unware of the LPG pipeline, which was not located by the local one-call system even though the piping reportedly was installed 11 “Mandatory Participation in Qualified One-Call Systems by Pipeline Operators,” 62 Federal Register 61695, https://www.gpo.gov/fdsys/browse/collection.action?collectionCode=FR. 12 Chris Guy, “Town’s Gas Leak Concern Grows,” The Baltimore Sun, September 9, 2002, http://articles.baltimoresun.com/2002-09-09/news/0209090002_1_snow-hill-propane-gas. 13 Chris Guy and Jennifer McMenamin, “Snow Hill Residents Cope, Care after Blast,” The Baltimore Sun, September 3, 2002, http://articles.baltimoresun.com/2002-09-03/news/ 0209030118_1_propane-explosion-snow-hill-ruth-young. 14 Pipeline and Hazardous Materials Safety Administration, “Pipeline Incident Flagged Files,” accessed April 30, 2018, https://www.phmsa.dot.gov/data-and-statistics/pipeline/ pipeline-incident-flagged-files. 15 Pipeline and Hazardous Materials Safety Administration, “Pipeline Incident Flagged Files.”

36 SAFETY REGULATION FOR SMALL LPG DISTRIBUTION SYSTEMS with tracer wire to enable detection.16 The LPG migrated underground through porous rock until it reached the crawl spaces of several buildings, where it remained undetected until it ignited in the middle of the night. Ghent, West Virginia On September 25, 2008, a propane technician was exchanging an older tank with a replacement tank on a non-jurisdictional system serving a con- venience store in Ghent, West Virginia, when a defective liquid withdrawal valve caused propane to escape. After sustaining frostbite from the escaping liquid, the technician sought guidance from another technician by phone but waited 15 minutes before calling emergency responders.17 During that time, the technician did not evacuate the store. Shortly after firefighters and the second propane technician arrived, the escaping vapor ignited, killing the two propane technicians and two emergency responders, seriously injur- ing six others, and destroying the store and nearby vehicles. In its investigation, the Chemical Safety and Hazard Investigation Board concluded that the explosion resulted from several factors, includ- ing insufficient hazardous materials training for the technician and the improper installation of the tank adjacent to the outside wall of the store contrary to federal and state regulation.18 Because of the tank’s location, LPG was able to enter the store in large quantities through exhaust vents. The board found that the volunteer firefighters lacked LPG-specific training to know that they should have promptly evacuated the store. REVIEW OF INCIDENT STATISTICS The main database containing reports of pipeline incidents in the United States is PHMSA’s Pipeline Incident Flagged Files. This database, which dates back to 1986, can be used to track the history of LPG incidents meet- ing certain consequence thresholds and their reported causes. To provide additional insight into potential safety issues for LPG pipelines, a second PHMSA database is consulted that tracks the annual system condition reports submitted by large LPG operators (those serving 100 or more cus- tomers) of their experience with leaks and excavation damage. 16 Administrator, “Families, Victims Sue over Door County Explosion,” The Daily Reporter– WI Construction News and Bids, September 25, 2006, http://dailyreporter.com/2006/09/25/ families-victims-sue-over-door-county-explosion. 17 U.S. Chemical Safety and Hazard Investigation Board, “Investigation Report: Little General Store—Propane Explosion,” September 2008, 1–2, 24, https://www.csb.gov/assets/1/20/csbfinal reportlittlegeneral.pdf?13741. 18 U.S. Chemical Safety and Hazard Investigation Board, “Investigation Report: Little General Store—Propane Explosion.”

HAZARD CHARACTERISTICS AND SAFETY PERFORMANCE 37 A concern of the committee is that some incidents involving small LPG systems may not have been reported to PHMSA, and therefore fire records in USFA’s NFIRS were also examined.19 The results of analyses of these databases, which are presented next, suggest that incidents involving LPG pipeline systems are rare. The causal information in the databases, however, is insufficient for drawing conclusions about specific regulatory require- ments and their safety contribution. PHMSA Incident Records PHMSA’s incident records are derived from reports by pipeline operators, which are required to submit reports of jurisdictional system incidents to the National Response Center (NRC) within 1 hour of confirming an incident has occurred.20 After contacting NRC, operators must submit a detailed report to PHMSA within 30 days. Reporting is required for inci- dents that involve a fatality or personal injury, estimated property damage of $50,000 or more, or a significant event as determined by the operator. Because PHMSA has revised its reporting forms and criteria over the years, the records for three periods—2010 to the present, March 2004 through 2009, and 1986 through February 2004—contain some variability in reporting information.21 The most important difference for the purposes of this study is the lack of a ready means for identifying the type of gas re- leased in incidents reported before 2010. There are also differences among the three reporting periods in the availability of data for the volume of gas released and in the amount of detail on incident causes. Despite the differ- ences, some of this information can be gleaned from a review of the nar- ratives in the individual reports. Incomplete records, however, would resist this type of analysis, as in the case of an incident in 2002.22 A line-by-line review of the pre-2010 records also allows for discarding records that fall well outside the study scope, such as incidents involving propane torches. 19 U.S. Fire Administration, “About the National Fire Incident Reporting System,” accessed January 25, 2018, https://www.usfa.fema.gov/data/nfirs/about/index.html. 20 The National Response Center (NRC) is staffed around the clock by U.S. Coast Guard (USCG) personnel and is a part of the federal government’s National Response System. USCG jointly leads federal response efforts with the U.S. Environmental Protection Agency if an incident reported to the NRC triggers mobilization of federal resources, which includes approximately a dozen other federal agencies. The NRC website is http://www.nrc.uscg.mil. 21 Pipeline and Hazardous Materials Safety Administration, “Incident Report Criteria History,” May 27, 2014, https://hip.phmsa.dot.gov/Hip_Help/pdmpublic_incident_page_allrpt. pdf. 22 PHMSA’s record for the 2002 incident in Snow Hill, Maryland, contained empty data fields for the narrative and others that initially prevented identification of the type of gas released (Pipeline and Hazardous Materials Safety Administration, “Pipeline Incident Flagged Files;” Chris Guy, “Town’s Gas Leak Concern Grows”).

38 SAFETY REGULATION FOR SMALL LPG DISTRIBUTION SYSTEMS Useful information was extracted from the pre-2010 and more recent re- cords; however, PHMSA does not collect incident data regarding certain LPG-specific factors, such as the number of customers served by a system or the size and configuration (that is, aboveground or underground) of the storage tank. It is, therefore, often impossible to determine the size of the system involved in an incident. Tables 3-2, 3-3, and 3-4 show the number of LPG pipeline incidents reported to PHMSA by operators during each of the three periods, includ- ing reports of fatalities and injuries for jurisdictional systems of all sizes. During 1986 to February 2004 (see Table 3-2), 12 incidents were reported in 7 years, including the 1996 San Juan explosion, which accounted for 33 of the 35 fatalities and 42 of the 64 injuries. (The NTSB report indicates that there were as many as 69 injuries.) From March 2004 through 2009 (see Table 3-3), seven incidents were reported, including the Door County explosion that accounted for both fatalities and four of the six injuries re- ported. During this reporting period, PHMSA also began collecting data on the number of persons evacuated in response to gas pipeline incidents. The Door County incident was responsible for half of the individuals evacuated. For the reporting period 2010 to 2017, zero fatalities and 10 injuries from LPG pipeline incidents were reported. Figure 3-1 shows the causes of the LPG pipeline incidents reported during these three periods combined. Third-party excavation damage was reported for eight incidents; unknown or “miscellaneous” for seven; incor- rect operation and other outside force damage for four each; corrosion damage for three; natural force damage for two; and material, welding, or joint failure for one. The 10 incidents reported from 2010 to 2017 included three caused by third-party excavation damage, three by other outside force damage (damage to the meter, vandalism, and fire), three by incorrect op- eration, and one by natural force damage. PHMSA Leading Indicator Data In addition to collecting reports of consequential incidents, PHMSA col- lects data from operators of large LPG systems (those serving 100 or more customers) on certain leading indicators of safety performance, includ- ing reports of leaks, hazardous leaks, and excavation damage.23 PHMSA defines a “leak” as an unintended release of gas from a pipeline facility, excluding leaks that can be repaired by basic maintenance activity, such as by lubrication or tightening. A more severe “hazardous leak” is defined as 23 Pipeline and Hazardous Materials Safety Administration, “Gas Distribution, Gas Gather- ing, Gas Transmission, Hazardous Liquids, Liquefied Natural Gas (LNG), and Underground Natural Gas Storage (UNGS) Annual Report Data.”

39 T A B L E 3 -2 L PG I nc id en ts R ep or te d to P H M SA , 19 86 –F eb ru ar y 20 04 Y ea r R ep or ts Fa ta lit ie s In ju ri es Ig ni ti on E xp lo si on T ot al C os t, 20 16 D ol la rs 19 86 1 0 2 1 1 $0 19 87 0 0 0 0 0 $0 19 88 1 0 2 1 0 $0 19 89 2 0 0 2 1 $2 08 ,7 80 19 90 2 0 0 2 2 $2 92 ,1 35 19 91 3 1 1 2 2 $1 4, 59 3 19 92 0 0 0 0 0 $0 19 93 0 0 0 0 0 $0 19 94 0 0 0 0 0 $0 19 95 0 0 0 0 0 $0 19 96 1 33 42 1 1 $7 ,2 72 ,2 52 19 97 0 0 0 0 0 $0 19 98 0 0 0 0 0 $0 19 99 0 0 0 0 0 $0 20 00 0 0 0 0 0 $0 20 01 0 0 0 0 0 $0 20 02 1 1 17 1 1 U nk no w n 20 03 1 0 0 1 1 $5 79 ,5 83 20 04 0 0 0 0 0 $0 To ta l 12 35 64 11 9 $8 ,3 67 ,3 44 SO U R C E : P ip el in e an d H az ar do us M at er ia ls S af et y A dm in is tr at io n, “ Pi pe lin e In ci de nt F la gg ed F ile s, ” A pr il 30 , 2 01 8, h tt ps :// w w w .p hm sa .d ot .g ov / da ta -a nd -s ta ti st ic s/ pi pe lin e/ pi pe lin e- in ci de nt -fl ag ge d- fil es .

40 T A B L E 3 -3 L PG I nc id en ts R ep or te d to P H M SA , M ar ch 2 00 4– 20 09 Y ea r R ep or ts Fa ta lit ie s In ju ri es Ig ni ti on E xp lo si on Pe rs on s E va cu at ed T ot al C os t, 20 16 D ol la rs 20 04 0 0 0 0 0 0 $0 20 05 1 0 0 1 1 0 $0 20 06 2 2 4 2 2 47 0 $1 ,6 62 ,8 56 20 07 0 0 0 0 0 0 $0 20 08 1 0 0 1 1 5 $7 89 ,4 30 20 09 3 0 2 3 3 10 $2 89 ,4 79 To ta l 7 2 6 7 7 48 5 $2 ,7 41 ,7 65 SO U R C E : Pi pe lin e an d H az ar do us M at er ia ls S af et y A dm in is tr at io n, “ Pi pe lin e In ci de nt F la gg ed F ile s. ”

41 T A B L E 3 -4 L PG I nc id en ts R ep or te d to P H M SA , 20 10 –2 01 7 Y ea r R ep or ts Fa ta lit ie s In ju ri es Ig ni ti on E xp lo si on G as R el ea se d, T ho us an ds o f C ub ic F ee t (f t3 ) Pe rs on s E va cu at ed T ot al C os t, 20 16 D ol la rs a 20 10 2 0 0 1 1 35 4 10 $1 ,2 36 ,4 11 20 11 1 0 1 1 1 18 0 $2 6, 58 4 20 12 1 0 2 1 1 10 0 $1 4, 75 6 20 13 3 0 0 3 3 18 .1 7 0 $4 8, 03 4 20 14 1 0 2 1 1 0. 24 0 $1 1, 80 6 20 15 0 0 0 0 0 0 0 $0 20 16 1 0 1 1 1 32 .2 10 $7 02 ,0 00 20 17 1 0 1 1 1 88 0 $1 3, 16 8 To ta l 10 0 7 9 9 52 0. 61 20 $2 ,0 52 ,7 60 a D ue t o a m od ifi ca ti on in r ep or ti ng m et ho do lo gy s in ce 2 01 0, p ro pe rt y da m ag e an d th e co st o f ga s re le as ed m us t be a dd ed t o co m pa re t o th e to ta l co st f ro m p re vi ou s re po rt in g pe ri od s. SO U R C E : Pi pe lin e an d H az ar do us M at er ia ls S af et y A dm in is tr at io n, “ Pi pe lin e In ci de nt F la gg ed F ile s. ”

42 SAFETY REGULATION FOR SMALL LPG DISTRIBUTION SYSTEMS FIGURE 3-1 Causes of incidents in LPG pipeline systems reported to PHMSA, 1986–2017. SOURCE: Pipeline and Hazardous Materials Safety Administration, “Pipeline Incident Flagged Files.” an uncontrolled release of gas that demands an immediate response to avoid a hazard to people or property. Figures 3-2 and 3-3 contain data from 2016 on both types of leaks reported in LPG mains and service lines. For both categories of pipeline, excavation damage caused about one-third of leaks, but it figured more prominently in hazardous leaks (accounting for more than three-quarters). Because leaks caused by excavation damage can involve sparks and human exposure, they will often require immediate action, thus requiring classifica- tion as a “hazardous leak.” Other causes of leaks tend to be time-dependent mechanisms such as corrosion or the failure of a pipe, weld, or joint. These leaks are usually discovered before they present a hazardous situation and are thus more likely to be reported simply as “leaks.” To better understand the causes of excavation damage, PHMSA collects more granular data when operators report excavation-caused leaks. Most reports of excavation damage, as shown in Figure 3-4, are attributed to in- sufficient one-call practices. Problems relating to one-call systems can stem from an inaccurate registry of pipeline locations, inaccurate marking of 3-1 Third-Party Excavation Damage 29% Material/Weld/Equipment Failure 4% Miscellaneous or Unknown 25% Incorrect Operation 14% Natural Force Damage 7% Other Outside Force Damage 14% Corrosion Internal 3% Corrosion External 4%

HAZARD CHARACTERISTICS AND SAFETY PERFORMANCE 43 FIGURE 3-2 The share of and count of leaks and hazardous leaks in LPG mains by cause reported to PHMSA for systems serving 100 or more customers, 2016. SOURCE: Pipeline and Hazardous Materials Safety Administration, “Gas Distribution, Gas Gath- ering, Gas Transmission, Hazardous Liquids, Liquefied Natural Gas (LNG), and Underground Nat- ural Gas Storage (UNGS) Annual Report Data,” accessed January 2, 2018, https://www.phmsa.dot. gov/data-and-statistics/pipeline/gas-distribution-gas-gathering-gas-transmission-hazardous- liquids. Leaks Other Outside Force Damage Equipment Failure Natural Force Damage Corrosion Failure Pipe, Weld, or Joint Failure Other Cause Excavation Damage 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Hazardous Leaks 2 14 1 1 1 3-2 7 2 1 8 9 16 22

44 SAFETY REGULATION FOR SMALL LPG DISTRIBUTION SYSTEMS FIGURE 3-3 The share of and count of leaks and hazardous leaks in LPG service lines by cause reported to PHMSA for systems serving 100 or more customers, 2016. SOURCE: Pipeline and Hazardous Materials Safety Administration, “Gas Distribution, Gas Gathering, Gas Transmission, Hazardous Liquids, Liquefied Natural Gas (LNG), and Under- ground Natural Gas Storage (UNGS) Annual Report Data.” 3-3 Leaks 13 8 2 33 59 13 98 130 Incorrect Operation Natural Force Damage Other Outside Force Damage Corrosion Failure Other Cause Equipment Failure Pipe, Weld, or Joint Failure Excavation Damage Hazardous Leaks 8 91 5 3 2 2 1 1100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

HAZARD CHARACTERISTICS AND SAFETY PERFORMANCE 45 FIGURE 3-4 The share of and count of excavation damage by cause in LPG sys- tems reported to PHMSA for systems serving 100 or more customers, 2016. SOURCE: Pipeline and Hazardous Materials Safety Administration, “Gas Distribution, Gas Gathering, Gas Transmission, Hazardous Liquids, Liquefied Natural Gas (LNG), and Under- ground Natural Gas Storage (UNGS) Annual Report Data.” 3-4 Excavation Damage Other Locating Practices Not Sufficient Excavation Practices Not Sufficient One-call Notification Practices Not Sufficient 21 23 35 91 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

46 SAFETY REGULATION FOR SMALL LPG DISTRIBUTION SYSTEMS pipelines, inadequate information from an excavator, insufficient notifica- tion time, and communication errors. Instances of these problems can lead to poorly informed excavation practices. The Door County LPG pipeline explosion, as noted above, originated with a failure of the one-call system to locate the existing propane pipeline. National Fire Incident Reporting System Incident data are reported to the NFIRS, which is maintained by USFA, a division of the Federal Emergency Management Agency. Unlike the incident reports collected by PHMSA from pipeline operators, NFIRS reports originate from fire departments across the United States. The NFIRS reports are also not limited to incidents involving a fatality, injury, property damage, or evacuation. The reports contain information on the nature of the incident, including whether the fire department responded to a fire, medical emergency, hazardous material release, or other type of emergency. In the case of fires, USFA estimates that the NFIRS reporting accounts for about three-quarters of all incidents,24 because incident re- porting is voluntary in some states.25 The committee’s review of the 2010 through 2016 NFIRS data (which is described further in the chapter’s annex) found 49 reports involving fires at LPG distribution pipelines (see Table 3-5).26 Because USFA estimates that NFIRS records only 75 percent of fire incidents, the 49 reports may be indicative of about 65 incidents nationwide during the 7-year period.27 However, because determination of whether a pipeline system is jurisdic- tional is immaterial to the response effort, NFIRS is likely to include reports from jurisdictional pipeline systems and from small non-jurisdictional facili- ties. At most, the NFIRS data suggest there may be about 10 incidents per year that involve LPG pipeline facilities of all types. Synopsis of Incident and Leak Data Available incident data contain too few details to examine the safety per- formance of small LPG systems specifically. Nevertheless, during more than 24 U.S. Fire Administration, “Review and Assessment of Data Quality in the National Fire Inci dent Reporting System,” May 2017, 4, https://www.usfa.fema.gov/downloads/pdf/ publications/nfirs_data_quality_report.pdf. 25 U.S. Fire Administration, National Fire Operations Reporting System, “NFIRS Requirements by State Law,” 10–11, accessed January 25, 2018, http://www.nfors.org/ assets/ StateFireData_Requirements.pdf. 26 Only fire incident data could be reliably extracted from NFIRS because gas releases cannot be differentiated by type of gas in the database. 27 U.S. Fire Administration, “About the National Fire Incident Reporting System.”

HAZARD CHARACTERISTICS AND SAFETY PERFORMANCE 47 TABLE 3-5 NFIRS Data on LPG Pipeline Fires in the United States, 2010–2016 Year Number of LPG Pipeline Fires Property Loss Incident States 2010 7 $320,000 IA, IN, KS, SC, TN 2011 15 $1,270,200 AL, CO, GA, IN, KS, MO, ND, NV, OH, TX, WA 2012 2 $0 MA, WI 2013 4 $0 KY, MI, MN, TX 2014 9 $30,000 AZ, CA, FL, IA, IL, MD, MN, OK, TX 2015 6 $300 CO, FL, GA, ID, PA, WI 2016 6 $0 AL, NC, TX Total 49 $1,620,500 Median 6 $300 Mean 7 $33,071 SOURCE: U.S. Fire Administration, “Download Fire Data and Data Analysis Tools,” June 5, 2018, https://www.usfa.fema.gov/data/statistics/order_download_data.html. 30 years of incident reporting to PHMSA, consequential incidents involving LPG distribution systems of all sizes have been infrequent events, averaging about one incident report per year. No fatalities have been reported since 2006. Supplemental data on LPG incidents involving distribution pipelines, as reported by fire departments to NFIRS, suggest that PHMSA may not be receiving reports of some LPG incidents that do not meet thresholds for consequences or because some pipeline operators are not submitting reports. Here again, it is difficult to know how many of these incidents in- volve jurisdictional systems having 100 or fewer customers, but it is likely not more than 10 per year. With so few incidents, it is difficult to identify causal patterns for inci- dents. The NFIRS data cannot be used to identify incident causes. The most common cause in PHMSA incident data (again, for all LPG systems) is ex- cavation damage. This cause is consistent with information from PHMSA’s leading indicators database, which shows that excavation damage is the most common cause of leaks that are reported by operators of larger LPG systems (those having 100 or more customers).

48 SAFETY REGULATION FOR SMALL LPG DISTRIBUTION SYSTEMS SUMMARY ASSESSMENT When released inadvertently, LPG behaves differently than natural gas. The high relative vapor density of LPG, which makes it heavier than air, can cause it to creep along the ground or seep underground, which can lead to pooling in concentrations that are within its flammability limits, rather than disperse into the atmosphere like natural gas. Also, because it is stored as a liquid, LPG can expand to 270 times its size as it vaporizes, potentially swelling into a flammable vapor cloud that adds to the risk of handling LPG liquid during transfers, such as from delivery trucks or from one tank to another. Despite these hazards, LPG pipeline distribution system incidents are rare. PHMSA records of federally regulated LPG distribution systems for more than the past 30 years suggest an average of less than one incident with a fatality or serious injury per year. Incidents reported by fire depart- ments suggest this total could be as high as 10 incidents per year, but with an uncertain portion involving smaller LPG distribution systems. Because consequential incidents are rare, and because of their relatively simple con- struction and operation, the pipeline systems that distribute LPG are gener- ally viewed as safe. However, the consequences of incidents can be severe. While the incident statistics offer some insight, the available incident reporting omits basic data such as the size of the storage tank, the number of customers served on a system, and whether the system configuration is aboveground or underground. The absence of such information limits the range of inferences that can be drawn about the safety performance of these systems, especially as it pertains to the effectiveness of any specific regula- tory requirements discussed in the next chapter. Three pipeline incidents illustrate how LPG can behave and present hazards when transported. In Parkers Prairie, Minnesota, in 1991; San Juan, Puerto Rico, in 1996; and Snow Hill, Maryland, in 2002, LPG re- leased from pipeline distribution systems migrated underground, pooled in low-lying areas, and ignited to cause explosions and fires. The number of fatalities in these two incidents was one and 33, respectively. A ruptured LPG pipeline that led to gas traveling through soil before igniting killed two people at a resort in Door County, Wisconsin, in 2006. While the specific factors contributing to these incidents differed, they each involved a lack of familiarity with LPG properties and behavior (the tendency to sink to low areas) and leaks that were undetected while the gas accumulated. In response to the kinds of hazards demonstrated by these and other serious incidents, the federal government has regulated the safety of LPG pipeline systems for nearly 50 years. Many states also regulate their safety, and LPG-specific safety codes are developed and maintained by the National Fire Protection Association (NFPA). The content and enforcement of these

HAZARD CHARACTERISTICS AND SAFETY PERFORMANCE 49 regulations and standards are discussed in the next chapter, including a discussion of concerns raised by industry about the applicability of some of the federal requirements to the smaller LPG systems given their safety performance and the coverage of the NFPA codes.

50 SAFETY REGULATION FOR SMALL LPG DISTRIBUTION SYSTEMS ANNEX With the extensive presentation of data in Chapter 3, this annex provides an additional level of detail for context about the data sources and limitations. SAFETY DATA SOURCES AND THEIR LIMITATIONS The data in Chapter 3 provide an overview of incident frequency, types of risks, prevalence of leading indicators for safety performance such as leaks and excavation damage, and fire reporting for liquefied petroleum gas (LPG) pipeline systems. This annex further explains the basis for the qualifications that the committee has offered elsewhere in the report about the three main safety data sources previously introduced: incident data from the Pipeline and Hazardous Materials Safety Administration’s (PHMSA’s) Pipeline Incident Flagged Files,28 leading indicators from operators’ self- reported annual system condition reporting from PHMSA’s Annual Report Data,29 and fire incident reporting from the U.S. Fire Administration’s (USFA’s) National Fire Incident Reporting System (NFIRS).30 Although these data sources contribute to an understanding of LPG pipeline safety performance, they also place limitations on the analysis in Chapter 3. PHMSA Pipeline Incident and Leading Indicator Data The PHMSA data for incidents and leading indicators share a few features in common. PHMSA makes both sets of data accessible through its website, which can be reviewed for varying periods of time.31 Both types of records also use an operator identification number (OPID), which is a unique identi- fier assigned to pipeline operators by PHMSA, as the key record identifier for incidents and leading indicators. The two datasets are also similar in their absence of commodity-specific records for pipeline distribution sys- tems until after 2009.32 28 Pipeline and Hazardous Materials Safety Administration, “Pipeline Incident Flagged Files.” 29 Pipeline and Hazardous Materials Safety Administration, “Gas Distribution, Gas Gather- ing, Gas Transmission, Hazardous Liquids, Liquefied Natural Gas (LNG), and Underground Natural Gas Storage (UNGS) Annual Report Data.” 30 U.S. Fire Administration, “About the National Fire Incident Reporting System.” 31 PHMSA provides the tabular data formatted as spreadsheet files. 32 Tables 3-2 through 3-4 summarize LPG incidents for three time periods. These periods reflect changes in PHMSA incident reporting forms that progressed to a more granular format. The historical data consequently have limited value for comparison from one period to an- other. For example, the dataset for 1986–February 2004 omits fields for the type of commodity released, the volume of gas released, number of evacuees, and added details on the number of commercial or residential properties affected in an incident.

HAZARD CHARACTERISTICS AND SAFETY PERFORMANCE 51 For large LPG pipeline systems (i.e., service to 100 or more custom- ers) that submit annual system condition reporting, the use of an OPID for leading indicators obscures important details. The use of an OPID as a key record identifier means that incident and system condition reporting is connected to the operator rather than the LPG system. This feature of the datasets is problematic, especially for leading indicator data. Because one operator generally reports multiple pipeline systems under one OPID and separate reporting for each system is not required,33 disaggregated data are unavailable to determine system-level characteristics. For instance, the reliance on OPID for annual report data on leading indicators results in aggregation of all systems operated by an LPG operator, so that the record could note thousands of feet of LPG pipeline when each system may only consist of a few hundred feet. Aggregated data also conceal the number of service lines per system, which would otherwise be useful as an approxima- tion for the number of customers per system. Reporting by OPID is less of an issue for incidents because each record describes a specific event and includes information on the incident location. Another challenge entails the availability of commodity-specific records. PHMSA began recording incidents by the type of commodity involved in 2010, while system condition reporting used for leading indicators became available for LPG in 2015. For incident data on jurisdictional LPG pipeline systems prior to 2010, the dataset lacks a formal marker to identify the type of gas involved in an incident. The committee overcame this issue by reviewing incidents that include LPG-related terms in the “narrative” field of the data.34 A line-by-line review of the narrative description for each record in the pre-2010 incident dataset yielded the tabulated results in Tables 3-2 and 3-3. Even when applying a filter for these terms, numerous incidents fell beyond the scope of the study. For instance, incident reporting for the period from March 2004 through 2009 was sufficient to determine that seven of the 13 incidents are relevant, while the rest involve propane accessories, such as a propane-fueled turkey fryer. Importantly, across all reporting periods, when an incident is reported to PHMSA, the record does not reflect certain details relevant to LPG systems, such as the number of customers, size of the storage tanks, or ag- gregate tank volume. Leading indicator data began to indicate the type of gas transported by a pipeline system in 2015. However, anomalies appeared in the LPG pipeline system data in that first year. For example, operators reported 44.7 miles of LPG mains with an unknown decade of construction in 2015; 33 Piyali Talukdar, “Cost-Benefit Analysis at the Office of Pipeline Safety” (August 24, 2017), 17, http://onlinepubs.trb.org/onlinepubs/Propane/Talukdar82417.pdf. 34 Search terms included “propane,” “butane,” “LP [gas],” and “LPG.”

52 SAFETY REGULATION FOR SMALL LPG DISTRIBUTION SYSTEMS the equivalent figure for 2016 was 7.9 miles. Likewise, for service lines, 2,057 service lines of unknown decade of construction were reported in 2015, which was followed by 614 service lines reported in 2016. Although it may be reasonable to expect an adjustment period for newly collected information that is self-reported by operators, the reporting irregularities and short timeframe for the collected data warranted caution. Therefore, the committee refrained from drawing many strong conclusions from the leading indicators data. In addition to the two data sources noted above, the committee sought additional perspectives on LPG pipeline distribution systems. Consultations with the LPG industry yielded extensive information from subject-matter experts for review by the committee. However, the committee was unable to obtain quantitative or statistically meaningful data to analyze that are particular to small LPG systems (i.e., those with fewer than 100 custom- ers). An industry representative indicated that the National Propane Gas Association does not maintain its own incident data on jurisdictional or non-jurisdictional systems. The committee also contacted insurers of LPG operators. Nonetheless, specific information on loss experience or even broad measures of safety were not available from the insurance industry. Because of the lack of supplemental data on incidents and the commit- tee’s interest in ascertaining whether LPG pipeline system incidents may be underreported in the PHMSA database, the committee chose to seek more information on incident rates. The committee’s attention turned to incident reporting by firefighters because of the potential flammability of LPG dur- ing an uncontrolled release. NFIRS The committee reviewed the data available from NFIRS to confirm whether LPG pipeline distribution system incidents were occurring without being reported to PHMSA. However, because NFIRS records no distinction on the size or configuration of an LPG pipeline system, the committee believes that an expansive search query for LPG system incidents would capture too many incidents that would likely fall outside of the study charge (i.e., whether a system is jurisdictional and its size). The narrow focus of the approach used in the search query to explore the NFIRS records prevents the inclusion of extraneous results, such as incidents involving small LPG cylinders used for outdoor gas grills and forklifts. However, the approach used also leaves open the possibility that some number of LPG pipeline sys- tem incidents are not included in the statistics in this report. Therefore, the information in Table 3-5 comprises a conservative estimate of LPG pipeline system fire incidents. Moreover, the NFIRS database was queried so as to draw only on data that belong to the set of mandatory reporting fields.

HAZARD CHARACTERISTICS AND SAFETY PERFORMANCE 53 The analysis of NFIRS to establish a count of LPG fires uses the code for “LPG” found in the data field for the type of material contributing most to the spread of the fire coupled with the code for “pipeline distribution system” in the data field for the specific use of the property at an incident site. Selection for these data fields is based on the characteristics used for identifying LPG pipeline distribution systems and on their status as man- datory reporting fields for firefighters when filing reports to NFIRS.35 By contrast, data fields such as factors that contributed to the growth, spread, or suppression of the fire; on-site materials or products; and stored material are not required items and could introduce uncertainty into the analysis because of inconsistent reporting. Other nonmandatory data fields were considered for the analysis, but were also set aside because, for instance, the item first ignited is vulnerable to confusion because the data field includes four codes that appear nearly identical. By not selecting nonmandatory data fields and avoiding codes that could be easily confused by those submit- ting incident reports, it is thus possible that more LPG incidents occurred from 2010 to 2016 than are captured in the narrow, though more reliable, criteria used in this analysis. To potentially supplement PHMSA’s leading indicator data, the com- mittee considered pipeline leak data also accessible from NFIRS. However, the relevant code for this data field references an indicator that includes leaks of both natural gas and LPG from distribution pipelines; no method to isolate LPG from natural gas leaks was found. As noted earlier in Chapter 3, NFIRS incident data represent a subset of all fire incidents. Not only is it probable that the raw data count of 49 LPG fire incidents that occurred from 2010 to 2016 would be scaled up to 65 fires using a standard national estimates technique,36 inclusion of even more incidents may be possible with refinements to an analysis of NFIRS. Therefore, the committee believes that inclusion of the NFIRS data is reasonable when considering LPG pipeline system incidents, though it is unknown how many of these incidents involve jurisdictional LPG systems. 35 U.S. Fire Administration, “National Fire Incident Reporting System: Complete Refer- ence Guide,” January 2015, A1–A17, https://www.usfa.fema.gov/downloads/pdf/nfirs/NFIRS_ Complete_Reference_Guide_2015.pdf. The NFIRS “Complete Reference Guide” includes a sample of its reporting form showing which fields are mandatory. 36 U.S. Fire Administration, “Review and Assessment of Data Quality in the National Fire Incident Reporting System,” 10–11.