December 2015 Inventory of U.S. Greenhouse Gas Emissions and Sinks: Revisions under Consideration for Natural Gas Distribution Emissions Substantial new data from several sources are available on emissions from the natural gas distribution segment. See Table 1 below for a summary of the new data available. The EPA is evaluating approaches for incorporating this new data into its emission estimates for the Inventory of U.S. Greenhouse Gas Emissions and Sinks (GHGI). The EPA is seeking stakeholder feedback on these updates under consideration. Please send any comments or new information or data to ghginventorygepa.gov by January 14, 2016. Background on Distribution Segment in the GHGI The natural gas distribution segment includes pipelines that take high-pressure gas from the transmission system at "city gate" stations, reduce the pressure, and distribute the gas through primarily underground mains and service lines to individual end users. Distribution system emissions, which in the 2015 GHGI account for approximately 20 percent of methane (CH4) emissions from natural gas systems and less than 1 percent of non-combustion carbon dioxide (C02) emissions, result mainly from fugitive emissions from gate stations and pipelines. An increased use of plastic piping, which has lower emissions per unit length than other pipe materials, has reduced both CH4 and C02 emissions from this segment over time. In the 2015 GHGI, distribution segment emission sources are organized as: • Meter/Regulator (M&R) stations o Stratified by station type (metering and regulating versus regulator stations), location (vault versus above ground) and inlet pressure range • Pipeline leaks o Stratified by type (mains versus service lines) and pipeline material • Customer meters o Stratified by customer type (residential versus commercial/industrial) • Routine maintenance, including pressure relief valve releases and pipeline blowdowns • Upsets, including mishaps (dig-ins) Note that the term "M&R stations" as used in the GHGI and this memorandum encompasses city gate stations (i.e., transmission-distribution custody transfer stations) and any above ground and below ground stations that meter and/or regulate natural gas pressure within the distribution system. The 2015 GHGI methodology largely relies on emission factors (EF) generated through a joint Gas Research Institute (GRI)/EPA study published in 1996 which uses 1992 as the base year. Many emission factors in the current GHGI are considered to represent "potential" emissions. The current GHGI accounts for advancement in and increased adoption of emission reduction technologies and practices by subtracting emission reductions reported to the EPA's Gas STAR program from the calculated potential emissions to estimate "net" emissions. Over the 1990-2013 time series, the Gas STAR program data show reductions achieved due to activities including: inserting flexible liners in cast iron and unprotected steel mains; implementing directed inspection and maintenance programs, and replacing high-bleed pneumatic devices with lower-emitting devices. A comparison of the GHGI emissions and Gas STAR reductions is shown in Appendix A. 1 ------- December 2015 Data Sources Available for Potential Updates Petroleum and natural gas system facilities meeting the emissions reporting threshold of 25,000 metric tons of C02 equivalent (MT C02e) report emissions and other information under subpart W of the EPA's greenhouse gas reporting program (GHGRP). The data reported to subpart W include activity data (AD) (e.g., frequency of certain activities, equipment counts) and emissions. Emissions are calculated using differing methodologies depending on the emission source, including the use of emission factors or direct measurements. For the most part, the emission sources included in subpart W are similar to those in the GHGI, but there are differences in coverage and calculation methods. Facilities have been reporting data under subpart W since 2011.1 The GHGRP subpart W data used in the analyses discussed in this memo reflect submissions from facilities as of August 18, 2014. Any emissions estimates in the 2016 Inventory that are based on GHGRP data will reflect updated published data. In 2015, Lamb et al. published findings from direct measurements at local distribution company (LDC) systems in the United States and survey data, the most comprehensive study on distribution systems in the United States since the 1996 GRI/EPA study. Lamb et al. investigated M&R stations, pipeline leaks, pipeline blowdowns, and mishaps (dig-ins), and observed overall lower emissions compared to the GHGI (which is calculated using the GRI/EPA study data). The Gas Technology Institute (GTI) and Innovative Environmental Solutions published a report in 2009 for Operations Technology Development (OTD) that investigated methane emission factors for select distribution sources (GTI 2009).2 The emission sources included M&R stations and customer meters. GTI produced another report for OTD in 2013 that investigated emission factors for plastic pipelines (GTI 2013). Clearstone Engineering published a report in 2011 for Environment Canada that investigated methane emission factors for residential customer meters (Clearstone report).3 The American Gas Associated (AGA) publishes an annual Gas Facts report that provides substantial data on the natural gas industry. Data in these reports are obtained from multiple sources, including the Uniform Statistical Report, the Energy Information Administration, and the Federal Energy Regulatory Commission. The EPA has reviewed data generated in these studies to assess potential improvements to GHGI methodologies. The type of data (i.e., AD or EF) that each of these studies evaluates is shown in Table 1. A summary of study designs is provided in Appendix B. Table 1. Identification of the Type of Data (AD and/or EF) Evaluated by Each Data Source Emission Source GHGRP Lamb et al. Clearstone AGA GTI 2009 GTI 2013 M&R Stations AD, EF EF - - EF - 1 For local distribution companies, reporting under subpart W of the GHGRP includes distribution pipelines and equipment at M&R stations that are "operated by a LDC within a single state that is regulated as a separate operating company by a public utility commission or that is operated as an independent municipally-owned distribution system." 2 Gas Technology Institute and Innovative Environmental Solutions, Field Measurement Program to Improve Uncertainties for Key Greenhouse Gas Emission Factors for Distribution Sources, November 2009. GTI Project Number 20497. OTD Project Number 7.7.b. 3 Clearstone Engineering, Development of Updated Emission Factors for Residential Meters, May 2011. 2 ------- December 2015 Pipeline leaks AD EF - - - EF Customer Meters - - EF AD EF - Pressure Relief Valve Releases - - - - - - Pipeline Blowdowns - EF - - - - Mishaps (Dig-Ins) - EF - - - This memorandum includes detailed evaluations of available data for M&R stations, pipeline leaks, and "other" emission sources (customer meters, pressure relief valve releases, pipeline blowdowns, and mishaps (dig-ins)). For each of these three categories, the following information is summarized: • Activity data; • Emissions data; • National estimates under various options; • Options for developing the time series of emissions estimates from 1990-2014; and • Approach under Consideration. At the end of this memorandum, specific requests for stakeholder feedback are outlined. M&R Stations Table 2 below presents an overview of AD and CH4 EFs used in the 2015 GHGI to develop CH4 emission estimates for M&R stations. Emissions are calculated separately for stations with metering and regulating, versus regulator stations, versus regulator vault (below grade) stations. AD and EFs are also stratified by station inlet pressure. Table 2. Year 2013 M&R Station Data in the 2015 GHGI Station Type & Inlet Pressure (psig) AD (# stations) AD source CH4 EF (scfh/station) CH4 EF source CH4 Emissions (MT CO;e) M&R >300 4,095 GRI/EPA, PHMSA, EIA 179.80 GRI/EPA 3,105,893 M&R 100-300 14,946 GRI/EPA, PHMSA, EIA 95.60 GRI/EPA 6,026,586 M&R <100 7,988 GRI/EPA, PHMSA, EIA 4.31 GRI/EPA 145,225 Reg >300 4,478 GRI/EPA, PHMSA, EIA 161.90 GRI/EPA 3,057,637 Reg-Vault >300 2,630 GRI/EPA, PHMSA, EIA 1.30 GRI/EPA 14,419 Reg 100-300 13,545 GRI/EPA, PHMSA, EIA 40.50 GRI/EPA 2,313,904 Reg-Vault 100-300 6,086 GRI/EPA, PHMSA, EIA 0.18 GRI/EPA 4,620 Reg 40-100 40,648 GRI/EPA, PHMSA, EIA 1.04 GRI/EPA 178,308 Reg-Vault 40-100 36,046 GRI/EPA, PHMSA, EIA 0.09 GRI/EPA 13,152 Reg <40 17,236 GRI/EPA, PHMSA, EIA 0.13 GRI/EPA 9,669 M&R Station Activity Data 3 ------- December 2015 In the current GHGI, M&R station counts in 1992 are calculated by multiplying GRI/EPA study data on station count per mile of main, developed from a survey of 12 companies, with Pipeline and Hazardous Materials Safety Administration's (PHMSA) Office of Pipeline Safety (OPS) data for the total miles of main in 1992. For non-1992 years, the station counts are estimated by scaling the 1992 station count by the total pipeline miles for the given year relative to the pipeline miles in 1992. Total pipeline miles in a given year are estimated by scaling the total pipeline miles in 1992 (from GRI/EPA) by residential gas consumption (from EIA) in the given year relative to 1992. M&R station activity is stratified by station type, location (vault versus above ground) and inlet pressure range. The GRI/EPA study did not focus on below grade transmission-distribution transfer stations (which exist in the GHGRP data set as discussed below) and this station type is not explicitly represented in the current GHGI activity data. LDCs are required to report to the GHGRP if their facility emissions exceed a threshold of 25,000 MT C02e. Comparing reported distribution pipeline main mileage for pipeline types in common between GHGRP and PHMSA for years 2011 through 2013, the approximately 180 GHGRP reporters account for approximately 71 percent of U.S. distribution pipeline mileage, on average across years. It may be reasonably assumed that there is an approximately constant number of M&R station per distribution pipeline mile across the United States—therefore the GHGRP activity data for M&R stations are expected to represent approximately 71 percent of total U.S. M&R stations. GHGRP reporters report activity (i.e., station count) and equipment leak emissions data separately for four categories: below grade transmission-distribution transfer stations; below grade M&R stations (which includes transmission-distribution transfer stations); above grade transmission-distribution transfer stations; and above grade M&R stations (which includes transmission-distribution transfer stations). For purposes of this memorandum, the subpart W station AD are presented as "transfer station" data and "non-transfer station" data, and stratified between above grade and below grade. Non-transfer station data equals the count of M&R stations (including T-D transfer) minus the count of transfer stations. Lamb et al. do not attempt to independently develop a national estimate of M&R station activity data, and rely on current GHGI AD in conjunction with EFs developed in the Lamb et al. study to produce a national emissions estimate. The GTI 2009 study does not estimate M&R station AD; it only evaluated M&R station EFs. The Clearstone report did not evaluate M&R station AD or EFs. Table 3 below presents counts of above grade and below grade stations for years 2011 through 2013 as reported to the GHGRP (as of August 18, 2014) by facilities exceeding the threshold, compared to national counts in the 2015 GHGI. Table 3. Activity Data in the GHGI (national total) and GHGRP (reporter total) for 2011 through 2013 Data Source / Station Type 2011 2012 2013 Above Grade Stations GHGRP/Transfer 14,497 18,372 18,217 GHGRP/Non-Transfer 62,735 61,165 65,832 GHGRP Total 77,232 79,537 84,049 GHGI Total 98,207 86,436 102,936 Below Grade Stations GHGRP/Transfer 2,751 2,142 2,778 GHGRP/Non-Transfer 23,310 25,881 20,573 GHGRP Total 26,061 28,023 23,351 GHGI Total 42,705 37,587 44,761 4 ------- December 2015 All Stations GHGRP Total 103,293 107,560 107,400 GHGI Total 140,912 124,023 147,697 M&R Station Emissions Data In the current GHGI, M&R station emissions are calculated using EFs developed in the 1996 GRI/EPA study. The GRI/EPA study used a tracer measurement approach: a known quantity of tracer gas is released next to a source of methane emissions, and the downwind concentration ratio of methane to tracer gas is measured using real-time instruments and canisters; assuming similar characteristics, the methane emissions can be determined by the ratio of methane to tracer concentration and the release rate of tracer gas. The GRI/EPA study developed emission factors by this approach stratified by station type (M&R versus regulator stations), location (vault versus above ground), and inlet pressure range. Emissions data for M&R stations collected under subpart W of the GHGRP are calculated using EFs. For above grade transmission-distribution transfer stations, reporters are required to conduct leak detection surveys and apply a "leaker" EF to each component (e.g., connectors, control valves, pressure relief valves, regulators, open ended lines) that is found to be leaking; the component leaker EFs provided in subpart W were obtained from the Handbook for Estimating Methane Emissions from Canadian Natural Gas Systems (1998) and the Measurement of Natural Gas Emissions from the Canadian Natural Gas Transmission and Distribution Industry (2007). For above grade meter-regulating stations, reporters must use an EF that is developed from the leak detection surveys of their above grade transfer- distribution stations. For all below grade stations, reporters multiply the count of stations by a station EF that varies by station inlet pressure from the GRI/EPA study. The Lamb et al. study employed a high-flow sampling method as the primary measurement technique to quantify leaks from components at M&R stations; the study also included a tracer measurement approach similar to the 1996 GRI/EPA study to verify the high-flow sampling measurements. Lamb et al. measured emissions from a total of 229 M&R stations (including transmission-distribution transfer stations) across 14 companies. Lamb et al. evaluated several possible distributions (e.g., lognormal distribution, inverse Gaussian distribution, Weibull distribution) and used probabilistic modeling to develop an average leak rate for each station type. Similar to the GRI/EPA study findings, Lamb et al. calculated higher emissions for facilities with higher inlet pressures, and lower emissions for vaulted (below grade) facilities. The Lamb et al. study observed that vented devices (e.g., natural gas-powered pneumatic controllers) contribute significantly to total station emissions, at stations equipped with such devices. The GTI 2009 study evaluated M&R station emissions based on direct measurement of individual components at stations. The study surveyed emissions at over 100 total custody transfer stations and pressure regulating stations operated by six companies. The GTI 2009 study determined that M&R station subcategories segregated by pressure range and above versus below ground were less appropriate and meaningful than a functional segmentation focused on types of stations and components at each. Therefore, the GTI 2009 study breaks out regulating stations into district regulators and pressure limiting stations. The study develops EFs that are weighted average values of the EFs developed by each company, wherein the company average is weighted according to the number of stations it surveyed. The GTI 2009 study notes that regulator stations with the lowest inlet pressures are likely to be district regulators and regulator stations with the highest inlet pressures are likely to be pressure limiting stations with continuous venting pneumatic devices. The GTI 2009 study notes that the EFs do not include additional vented emissions from emergency or maintenance events. 5 ------- December 2015 Table 4 below summarizes the EFs used in the current GHGI compared to findings from the Lamb et al. study, factors derived from GHGRP subpart W data (for reporting year 2013, as of August 18, 2014), and factors from the GTI 2009 study. Table 4. M&R Station CH4 Emission Factors from GRI/EPA, Lamb et al., GHGRP, and GTI 2009 Station Type & Inlet Pressure (psig) GRI/EPA CH4 EF (scfh/station) Lamb CH4 EF (scfh/station) Subpart WCH4 EF (scfh/station)bc GTI 2009 Station Type and CH4 EF (scfh/station)d Above Grade Stations M&R >300 179.8 12.7 Above Grade M&R Stations (Including T-D Transfer Stations) = 3.58 Custody Transfer Station = 26.6e M&R 100-300 95.6 5.9 M&R <100a 4.31 - Reg >300 161.9 5.2 District Regulator = 0.98 District Regulator with No Venting Devices = 0.3 Pressure Limiting = 92.5 Pressure Limiting with No Venting Devices = 30.6 Reg 100-300 40.5 0.85 Reg 40-100 1.04 0.97 Reg <40a 0.13 - Below Grade Stations R-Vault >300 1.3 0.3 Below Grade M&R Stations (including T-D transfer stations) = 0.30 R-Vault 100-300 0.18 0.3 R-Vault 40-100 0.09 0.3 a. Lamb et al. did not develop EFs for these categories. Lamb et al. did not collect data on stations in the M&R 100 psig category, and only surveyed one station in the Reg <40 psig category. b. Under subpart W, facilities report emissions from all M&R stations at their facility (including T-D transfer stations). Inlet pressure data are not reported under subpart W. c. Subpart W EFs presented in this table were developed from verified RY2013 data, and calculated as an average wherein each individual station is weighted equally (i.e., regardless of whether it is the only station within a reporting facility or one of hundreds). Facilities that reported zero emissions for their stations were included in the EF calculations. d. The GTI 2009 study presents their data using different station categories than used in the GHGI; this table presents GTI 2009's station categories aligned with the GHGI categories based on the best assignments possible. For example, not all M&R stations will be custody transfer stations, but "custody transfer station" category is the only M&R station category presented by GTI 2009. GTI 2009 also presents two regulating station types (district regulator and pressure limiting station) and does not distinguish by inlet pressure or whether a station is above ground or vaulted. e. This EF is based on the average equipment counts for a station. Specific EFs were developed for continuous venting devices, odorizers, and catalytic heaters which were used to estimate an average custody transfer station EF. The custody transfer station EF can be recalculated to reflect the equipment at a specific station. National Estimates of M&R Station Emissions Table 5 below summarizes national emissions estimates for years 2011 through 2013 from the GHGI, and estimates developed using Lamb et al. EFs in conjunction with GHGI AD. GHGRP reported emissions (as of August 16, 2014) are also included in the table for comparison; though note that they are not national emissions estimates, they include only the subset of facilities that report to GHGRP. Table 5. M&R Station Methane Emissions from GHGI (national total), Lamb et al. (national total), and GHGRP (reporter total) (MT C02e)a Station Type/ Data Source 2011 2012 2013 Above Grade Stations 6 ------- December 2015 Station Type/ Data Source 2011 2012 2013 2015 GHGI 14,155,567 12,458,941 14,837,221 Lamb et al. 1,009,719 888,699 1,058,342 GHGRP 791,252 770,135 1,270,570 Below Grade Stations 2015 GHGI 30,712 27,031 32,191 Lamb et al. 54,038 47,561 56,640 GHGRP 31,433 108,685 30,650 All Stations 2015 GHGI 14,186,280 12,485,973 14,869,412 Lamb et al. 1,063,757 936,260 1,114,982 GHGRP 822,685 878,820 1,301,220 a. For the 2015 GHGI, these are potential emissions and do not reflect Gas STAR reductions. GHGI Time Series Considerations for M&R Station Emissions Lamb et al. generally found lower average per-station emissions than those found in the GRI/EPA study. Lamb et al. suggest that the lower emissions reported in Lamb et al. illustrate the impact of nearly 20 years of advances in emission reduction technologies and adoption of changes to operational procedures that reduce emissions. Lamb et al. also conducted a survey on facility equipment upgrades and noted the influence of such upgrades on observed emissions in recent years compared to the GRI/EPA study 1992 base year. The GTI 2009 study noted that continuous bleed pneumatic controller replacement has led to reduced M&R station emissions over time; since the GRI/EPA study was conducted, many LDCs have instituted programs to replace continuous bleed devices with intermittent, low bleed or no-bleed devices. Reasons for the replacement of continuous bleed devices, as stated in the GTI 2009 study, include "improved performance of the new devices, reduced emissions of odorized gas to reduce impact on neighbors, lower emissions of natural gas to improve worker safety and conditions, difficulty in finding replacement parts for old pneumatic devices, and high maintenance costs for the old devices." The GTI 2009 study also stated that "Some LDCs have designed and implemented a standard custody transfer station containing no venting equipment, and all new stations use this design." The GHGRP provides four recent years of data on this emission source and shows lower emissions than the GHGI and other data sources. It is difficult to determine precisely what leads to the difference between GHGRP and the GHGI on this source and whether it indicates a change in emissions from the GRI study (e.g., fewer leaking components in recent years), or if is due to different emission calculation approaches (application of a station-level factor for the GHGI and component-level leaker factors in the GHGRP). Over the 1990-2013 time series, the Gas STAR program data show reductions achieved due to activities including directed inspection and maintenance at surface facilities and replacing high-bleed pneumatic devices with lower-emitting devices. These reductions are included within the category of "other" distribution segment emission reduction that is presented in the current GHGI. These reductions are small compared to total emissions from M&R stations, and contribute to the net emissions in the GHGI being much higher than Lamb et al. estimates. Lamb et al. noted the limited impact the reductions and suggested that the GHGI does not reflect the changes that have occurred at M&R stations over time. See Appendix A for additional detail on source-specific and "other" Gas STAR emission reductions. Approach for M&R Station Emissions under Consideration 7 ------- December 2015 For activity data for years 2011 through 2014, the EPA is considering using counts of above grade and below grade stations reported to subpart W, scaled up for national representation. The scaling would be based on an estimated subpart W coverage factor developed from comparing subpart W reporter LDCs to PHMSA company-level data on pipeline mileage. For example, the 2012 PHMSA data show that subpart W LDCs appear to account for approximately 71 percent of U.S. gas distribution pipeline mileage. Subpart W station counts would be divided by the coverage factor (e.g., 0.71) to calculate a national station count estimate. This revised activity data approach for years 2011 through 2014 would assume the same split of station subcategories (e.g., by inlet pressure range) as used in the current GHGI. For 1990-2010, the level of year-to-year variation in the total station counts was assessed and it was determined that it would be relatively consistent across the time series whether the counts are driven from 1992 or derived from subpart W data, so under this approach, activity data for years 1990 through 2010 would not be revised. The current GHGI accounts for emissions reductions from industry practices (which result in effectively lower station EFs) by using Gas STAR reductions data. Based on the results of Lamb et al. and the discussion in Lamb et al., it is possible that the current data set does not include significant reductions that have occurred over time for this activity. Lamb et al. surveyed study partners on upgrades since 1992. The responses indicated that 60% of the 90 sites included in 5 companies responding had undergone some level of equipment changes since 1992. An additional survey sent to AGA showed that half of the 14 respondents had replaced entire facilities, and at least $345 million was spent on facility upgrades by the respondents. Lamb et al. also noted that "It was also clear from our interactions with M&R personnel at different LDCs that maintenance activities and attention to leaks have increased, in part, due to the GHG reporting requirements implemented in the past several years (40 CFR 98 Subpart W)." It is also possible that the Lamb et al. field measurements did not capture enough data to adequately represent superemitters in development of its EFs. The EPA is considering applying GRI/EPA study-based EFs for earlier time series years, and Lamb et al. EFs for later time series years. The EPA will consider recalculating GRI/EPA study-based EFs to take into consideration a heavy tail distribution. The EPA could then develop year-specific EFs assuming a linear correlation for the intermediate years (unless there was a specific year when an industry-wide change is recognized). Regarding potential application of GTI 2009 EFs for purposes of developing a national estimate, as the GTI 2009 study notes, the number and type of components at stations are needed to extrapolate the report's data to develop a national GHGI estimate. Such data are not readily available and therefore the EPA is not considering using the GTI 2009 EFs. National emission estimates according to the approach under consideration for the GHGI—using scaled subpart W activity data and recent EFs from Lamb et al.—are shown in Table 6 below. Table 6. Year 2013 M&R Station Methane Emissions Calculated by Various Approaches 2013 Emissions (MT CO;e) Station Type & Inlet Revision Under Pressure (psig) 2015 GHGIJ Consideration13 Above Grade Stations M&R >300 3,105,893 248,521 M&R 100-300 6,026,586 421,335 M&R <100° 145,225 164,514 Reg >300 3,057,637 110,182 Reg 100-300 2,313,904 55,014 8 ------- December 2015 Station Type & Inlet Pressure (psig) 2013 Emissions (MT CO;e) 2015 GHGIJ Revision Under Consideration13 Reg 40-100 178,308 188,396 Reg <40° 9,669 10,953 Below Grade Stations R-Vault >300 14,419 2,408 R-Vault 100-300 4,620 5,573 R-Vault 40-100 13,152 33,013 All Stations Total 14,869,412 1,239,910 a. For the 2015 GHGI, these are potential emissions and do not reflect Gas STAR reductions. b. For the revision under consideration, these are net emissions. c. Lamb et al. did not develop EFs for these categories. The revision under consideration uses GRI/EPA EFs. Pipeline Leaks Table 7 below presents an overview of AD and CH4 EF data used in the 2015 GHGI to develop CH4 emission estimates for distribution pipeline leaks. Table 7. Year 2013 Distribution Pipeline Data in the 2015 GHGI Emission Source AD AD source CH4 EF CH4 EF source CH4 Emissions (MT CO;e) Mains Cast Iron 30,904 miles PHMSA 238.70 Mscfy/mile GRI/EPA 3,551,922 Unprotected Steel 60,633 miles PHMSA 110.19 Mscfy/mile GRI/EPA 3,216,971 Protected Steel 486,521 miles PHMSA 3.07 Mscfy/mile GRI/EPA 718,453 Plastic 674,808 miles PHMSA 9.91 Mscfy/mile SoCal/GRI 3,219,958 Services Unprotected Steel 3,668,842 services PHMSA 1.70 Mscfy/service GRI/EPA 3,004,487 Protected Steel 14,751,424 services PHMSA 0.18 Mscfy/service GRI/EPA 1,253,616 Plastic 46,153,036 services PHMSA 0.01 Mscfy/service GRI/EPA 206,630 Copper 973,107 services PHMSA 0.25 Mscfy/service GRI/EPA 119,165 Pipeline Leaks Activity Data In the current GHGI, miles of distribution mains and counts of services are obtained directly from the U.S. Department of Transportation's (DOT) Pipeline Hazardous Materials Safety Administration (PHMSA), for each year of the time series. On its website, PHMSA makes available data collected via annual reports that are submitted by operators of natural gas transmission and distribution pipelines. Annual reports include general information such as total pipeline mileage, commodities transported, pipeline miles by material, and installation dates. 9 ------- December 2015 LDCs are required to report to the GHGRP if their facility emissions exceed a threshold of 25,000 MT C02e. Based on GHGRP and PHMSA data on LDCs for years 2011 through 2013, GHGRP reporters account for approximately 12 percent of LDCs and approximately 71 percent of U.S. distribution pipeline mileage, on average across years. Beginning in RY2014, reporters provided activity (i.e., counts or miles) and emissions data separately for distribution mains by material type (unprotected steel, protected steel, plastic, and cast iron) and distribution services by material type (unprotected steel, protected steel, plastic, and copper), including back-reported data for RYs 2011 through 2013. Lamb et al., the GTI 2009 study, and the Clearstone report did not evaluate national pipeline activity data. Lamb et al. relies on current GHGI activity data in conjunction with EFs developed in the Lamb et al. study to produce a national emissions estimate. Pipeline Leaks Emissions Data In the current GHGI, emissions are calculated using EFs developed from the 1996 GRI/EPA study. For plastic mains, in addition to the six plastic pipeline data points from the GRI/EPA study, the EF incorporates seven data points from a 1993 Southern California Gas Company (SoCal) study. The GHGI EFs are in units of thousand standard cubic feet per mile (or service) per year. The GRI/EPA EFs used in the GHGI were developed by first measuring individual leak rates from mains (and total leak rates from services) to develop an average leakage rate in scf CH4 per hour by pipeline material; the averaging method used in the GRI/EPA study is not specified. To measure leak rates, the pipeline was unearthed and measured at the source; therefore, soil oxidation had to be taken into account in developing atmospheric emission rates. For cast iron pipelines, a "segment test" approach was used to develop leak rate, rather than measuring individual leak rates, so the resulting test data represent leakage rate per unit length of cast iron main. The GRI/EPA study also used national-level leak repair records to estimate equivalent leaks per mile of main (or service) and translate average leakage rates to an "equivalent leak" basis (where an equivalent leak represents a leak that exists for one entire year). For plastic mains, an average leak rate was calculated using a weighted average of the individual leak rates of the sample points and the number of leaks in each sample point across the GRI and SoCal study data. Similar to the approach for other pipeline materials, the average leak rate was then adjusted by soil oxidation rate to yield an average leak rate. Emissions data on distribution pipelines collected under subpart W of the GHGRP are calculated using the GRI/EPA study-based EFs that underlie the current GHGI. Reporters are required to apply the appropriate pipeline material-specific EFs to the material-specific lengths of distribution pipeline and counts of services within the reporting LDC. Note that subpart W provides the EFs on an hourly basis so that reporters can calculate annual emissions for mains and services that may not have been operating a full year. Lamb et al. measured leak rates from underground pipelines at the ground surface using a high flow sampler. The high flow sampler included a surface enclosure system to capture leak emissions. The pipeline was not unearthed as it was for GRI/EPA measurements. Several probabilistic models (e.g., lognormal distribution, inverse Gaussian distribution, Weibull distribution) were evaluated to develop an average leak rate for each pipeline type. The study also employed a similar approach as GRI/EPA in translating findings to an equivalent leak basis. This study generally observed both lower leak rates (CH4 emitted per hour) and lower equivalent leaks per mile (or service), compared to the GRI/EPA study; the 10 ------- December 2015 only exception to this is protected steel mains, where the Lamb et al. leak rate was higher than the GRI/EPA leak rate. GTI 2013 analyzed leak rates from polyethylene plastic pipeline using a Hi-Flow Sampler and an enclosure to measure 30 leaks above ground, and also conducted flow rate measurements using a Laminar Flow Elements (LFE) device on isolated below ground segments for a subset (21) of the same leaks. GTI did not take oxidation into account for the below ground measurements. GTI used the Hi-Flow results in its leak factor calculations. GTI observed a relatively small number of records with high leak rates and that leak records are represented by a lognormal distribution; therefore, GTI applied a weighted function to measurements, resulting in a recommended weighted emission factor (3.72 scf/leak/hour) that was higher than the mean of the measurements (3.3 scf/leak/hour). For comparison with Table 8 below, for plastic pipeline mains, GTI calculated a leak rate of 3.72 scf/leak/hour, a leak rate per mile of 0.07 (based on recent DOT leak repair rate data in conjunction with the leak-repair ratio assumed in the GRI/EPA study), and an EF of 2.28 Mscf CH4/mile/year. Using only the GTI measurements made with the LFE device results in a higher unweighted mean (5.7 scf/leak/hour) than use of the Hi- Flow measurements from that subpopulation, which results in an unweighted EF of 5.0 Mscf/mile/year. The GTI 2009 study and the Clearstone report did not evaluate pipeline leak EFs. Table 8 below summarizes the emissions data and EFs used in the current GHGI compared to findings from the Lamb et al. study. Table 8. Distribution Pipeline Leak Emissions Data in the 2015 GHGI and Lamb et al. Emission Source Leak rate (scf CHVleak/hour) Equivalent leaks per mile (or service) EF (mscf CH4 per mile or service per year) 2015 GHGI Lamb et al. 2015 GHGI Lamb et al. 2015 GHGI Lamb et al. Mains Cast Iron 27.3a 2.83 - 2.424 238.7 60.1 Unprotected Steel 5.9 2.40 2.127 2.005 110.2 42.1 Protected Steel 2.3 3.79 0.151 0.113 3.1 3.8 Plastic 5.85 1.04 0.184 0.050 9.4 0.5 Services Unprotected Steel 2.306 1.020 0.084 0.030 1.701 0.267 Protected Steel 1.050 0.400 0.019 0.033 0.176 0.115 Plastic 0.272 0.400 0.004 0.003 0.009 0.011 Copper 0.877 - 0.033 0.021 0.254 - a. This value is scf Cm/mile/hour. As described above, the GRI/EPA study developed the cast iron pipeline emission factor on a unit length basis rather than individual leak basis. National Estimates of Pipeline Leak Emissions Table 9 below summarizes emissions in the 2015 GHGI compared to calculated emissions using EFs from the Lamb et al. study, for years 2011 through 2013. Emissions in the table below are calculated using the EFs from the two right-most columns in Table 8. The activity data set is the same for both sets of emissions presented—miles of main and counts of services, stratified by pipeline material, are obtained from PHMSA for each calendar year. For comparison with Table 9 below, the GTI factors for plastic pipelines would result in 2013 national emissions of approximately 740,000 MT C02e (GTI- recommended factor) and approximately 1,144,000 MT C02e (using a factor calculated with unweighted LFE data only). 11 ------- December 2015 Table 9. Distribution Pipeline Leak Methane Emissions (MT C02e)a Emission Source 2011 2012 2013 2015 GHGI Lamb et al. 2015 GHGI Lamb et al. 2015 GHGI Lamb et al. Mains Cast Iron 3,869,829 974,130 3,724,553 937,561 3,551,922 894,105 Unprotected Steel 3,447,607 1,318,608 3,379,801 1,292,674 3,216,971 1,230,396 Protected Steel 721,698 886,477 719,875 884,237 718,453 882,490 Plastic 3,099,569 141,668 3,157,525 144,317 3,219,958 147,170 Services Unprotected Steel 3,392,655 532,671 3,207,625 503,620 3,004,487 471,726 Protected Steel 1,298,099 844,586 1,270,714 826,768 1,253,616 815,644 Plastic 198,319 224,688 202,240 229,131 206,630 234,104 Copper 129,388 - 123,591 - 119,165 - a. For the 2015 GHGI, these are potential emissions and do not reflect Gas STAR reductions. GHGI Time Series Considerations for Pipeline Leak Emissions Pipeline replacement is captured in the current methodology, since annual AD are obtained directly from PHMSA and stratified by pipeline material. Lamb et al. suggests that pipeline leaks have decreased over the past twenty years due to factors including efforts to seal cast iron joints and enhanced leak detection and repair procedures. The current GHGI accounts for advancement in and increased adoption of emission reduction technologies and practices by subtracting emission reductions reported to the EPA's Gas STAR program from the calculated potential emissions—however, similar to M&R stations, it is difficult to quantify the impact of Gas STAR on all pipeline-related emissions because some activities are categorized as "other" reductions (except controlling cast iron fugitives, and those reductions are very small). See Appendix A for additional detail on source-specific and "other" Gas STAR emission reductions. As discussed above, there are two components of the pipeline leak EFs (emissions per mile) developed by both GRI/EPA and Lamb et al.: (1) leak rate (scf CH4 per hour); and (2) equivalent leaks per mile (or service). Lamb et al. generally observed both lower leak rates and lower equivalent leaks per mile (or service), compared to the GRI/EPA study. In developing the estimate of equivalent leaks per mile (or service), both GRI/EPA and Lamb et al. relied on national LDC leak survey data compiled by the DOT and company survey information to estimate leaks per leak repaired. The Lamb et al. study used data from six companies to calculate a ratio of 1.63 leaks per leak repaired for year 2011, while GRI/EPA also used data from six companies to calculate a ratio of 2.14 leaks per leak repaired for year 1991. Once extrapolated to a national level using national leak repair data, GRI/EPA calculates a higher number of equivalent leaks per mile (or service) than Lamb et al. for most pipeline types. This might imply a higher leak incidence rate and/or a lower leak repair rate throughout the distribution segment in the early years of the time series compared to more recent years. The EPA seeks stakeholder feedback to confirm whether there are known trends in the industry over time that would result in overall lower leak emission rates (scf/leak/hour) and/or lower leak incidence rate (equivalent leaks per mile) throughout the United States in recent years compared to the early 1990's timeframe. Approach for Pipeline Leak Emissions under Consideration for the GHGI 12 ------- December 2015 For pipelines, PHMSA data provide national activity data on an annual basis, stratified by pipeline material. There is no clear advantage to using an alternate data source for activity. In the current GHGI, emissions are calculated using EFs developed from the 1996 GRI/EPA study, for all types of pipelines except plastic mains. Plastic main estimates are based on an updated factor developed in 2005 that incorporates data from the Southern California study in addition to using GRI/EPA data. Comparing the GRI/EPA and Lamb et al. studies, leak incidence rate is lower for the more recent data set (Lamb et al.). For plastic pipelines, the GTI results support the Lamb results of a lower leak frequency in recent years. Leak incidence is one of two aspects factored into the calculation of the GHGI EFs which are in units of emissions per mile (or service) per year. The other component of the EFs (leak emission rate) does not appear to exhibit as much of a trend between GRI/EPA and Lamb et al.—though Lamb et al. do point out that the sample selection methodology and sampling methodology differences between the two studies might contribute to discrepancies in results. The EPA is still assessing these data. Therefore, for the GHGI, the EPA is considering applying Lamb et al. emission factors for recent time series years, current GHGI EFs for earlier years, and linear interpolation between each EF for intermediate years' EFs. In the future and based on stakeholder feedback and other information, the EPA will consider potential approaches such as combining GRI/EPA and Lamb et al. leak emission rate data (scf CH4 per hour), but for early time series years apply an EF that uses GRI/EPA observed leak incidence and for later time series years apply an EF that uses Lamb et al. observed leak incidence. Table 10 below presents national emission estimates for year 2013 according to the approach under consideration for the GHGI—using the same activity data methodology as the 2015 GHGI in conjunction with Lamb et al. EFs. Table 10. Year 2013 Pipeline Leak Methane Emissions Calculated by Various Approaches 2013 Emissions (MT CO;e) Revision Under Emission Source 2015 GHGIJ Consideration13 Mains Cast Iron 3,551,922 894,105 Unprotected Steel 3,216,971 1,230,396 Protected Steel 718,453 882,490 Plastic 3,219,958 147,170 Services Unprotected Steel 3,004,487 471,726 Protected Steel 1,253,616 815,644 Plastic 206,630 234,104 Copper0 119,165 119,165 a. For the 2015 GHGI, these are potential emissions and do not reflect Gas STAR reductions. b. For the approach under consideration, these are net emissions. c. For copper services, Lamb et al. did not develop an EF. The methodology update under consideration would use the GRI EF for all years. Other Distribution Emission Sources—Meters. Pressure Relief Valves. Pipeline Blowdowns. and Mishaps Table 11 below presents an overview of AD and CH4 EF data used in the 2015 GHGI to develop CH4 emission estimates for customer meters (residential and commercial/industrial), pressure relief valve 13 ------- December 2015 releases, pipeline blowdowns, and mishaps (dig-ins). These sources are collectively referred to as "Other Distribution" sources in this memorandum. Table 11. Year 2013 "Other Distribution" Emission Source Data in the 2015 GHGI Category AD AD source CH4 ef CH4 ef source CH4 Emissions (MT CO.e) Customer meters- Residential 42,192,085 meters GRI/EPA, EIA 143.27 scfy/meter GRI/EPA 2,910,615 Customer meters- Commercial/lndustry 4,797,283 meters GRI/EPA, EIA 47.90 scfy/meter GRI/EPA 110,644 Pressure Relief Valve Releases 1,252,866 miles PHMSA 0.05 Mscfy/mile GRI/EPA 30,163 Pipeline Blowdown 1,366,993 miles GRI/EPA, EIA 0.10 Mscfy/mile GRI/EPA 67,137 Mishaps (Dig-ins) 1,366,993 miles GRI/EPA, EIA 1.59 Mscfy/mile GRI/EPA 1,046,550 Other Distribution Sources Activity Data In the current GHGI, other distribution source activity data are obtained from the GRI/EPA study, the U.S. Energy Information Administration (El A), and PHMSA, depending on the emission source. Residential and commercial/industrial customer meter counts for 1992 are provided in the GRI/EPA study. To estimate non-1992 residential and commercial/industrial customer meter counts in the GHGI, the 1992 base meter count is multiplied by the ratio of residential or commercial/industrial gas consumption for a given year to 1992 residential or commercial/industrial gas consumption. Residential and commercial/industrial gas consumption data are obtained from EIA monthly reports. To estimate year 1992 residential and commercial/industrial customer meter counts, GRI/EPA started with year 1992 end user data from AGA's Gas Facts publication and applied two steps to convert the end user AD into relevant customer meter AD. First, GRI/EPA assumed that the number of end users equaled the number of customer meters. Second, for residential meters, GRI/EPA calculated the proportion of residential meters located outdoors versus indoors using data from 22 individual gas companies within different regions of the country (Gas Facts also reports residential end users by region); GRI/EPA assumed indoor meter emissions were negligible because leaks within the confined space of a residence are readily identified and repaired. Table 12 below presents the percent of residential meters that are outdoors, as reported by GRI/EPA. The relevant (outdoor) residential meter AD were thus determined by multiplying the percentages from Table 12 times the number of total residential meters in each region. Table 12. Percent of Residential Customer Meters that are Outdoors, as Reported by GRI/EPA Region Average Percent Residential Outdoor Meters New England 48% Middle Atlantic 39% East North Central 83% West North Central 60% South Atlantic 79% East South Central 100% West South Central 100% Mountain 100% Pacific 95% 14 ------- December 2015 For pressure relief valve releases, the current GHGI activity data are distribution main miles, which are obtained from PHMSAfor each year of the time series. Pipeline blowdowns and mishaps (dig-ins) activity data are the total miles of distribution mains and services, using 1992 data available in the GRI/EPA study as the base year. To estimate the activity data for non-1992 pipeline blowdowns and mishaps, the 1992 mileage is multiplied by the ratio of residential gas consumption from EIA for a given year to 1992 residential gas consumption. Subpart W distribution segment requirements do not include reporting of customer meters (residential and commercial/industrial), pressure relief valve releases, pipeline blowdowns, or mishaps. Therefore, subpart W activity data are not available for the "other" sources. Lamb et al. did not investigate "other" sources activity data for their study. They focused on emissions data, as discussed below. When calculating emissions, Lamb et al. used the same activity data as the GHGI. The GTI 2009 study and Clearstone report also did not investigate "other" sources activity data. Other Distribution Sources Emissions Data In the current GHGI, emission factors for customer meters, pressure relief valve releases, pipeline blowdowns, and mishaps are estimated using data from the GRI/EPA study. Outdoor residential meters at 10 sites across the United States, including a total of approximately 1,600 meters, were screened. An average leak rate of scfy CH4/meter was determined for each of the 10 locations. The GHGI emission factor is calculated as the weighted average of the 10 average location leak rates (using the number of outdoor residential meters screened at each site). The GRI/EPA study also screened 149 commercial/industrial customer meters across four sites. GRI/EPA calculated an average commercial/industrial meter EF for each site, then averaged the four sites' averages together to calculate a commercial/industrial meter emission factor (scfy CH4/meter), which is used in the current GHGI. One of the sites where commercial/industrial meters were screened did not have any leaks, and thus had a site EF of zero scfy CH4/meter; this site was included when the unweighted average commercial/industrial meter emission factor was calculated. Emission factors for pressure relief valve releases, pipeline blowdowns, and mishaps were based upon company studies, and a weighted average emission factor (based on the pipeline length over which the reported emissions occurred for each company) is provided for each emission source in the GRI/EPA study; each of these factors is used in the GHGI. As discussed above regarding activity data, subpart W of the GHGRP does not cover customer meters, pressure relief valve releases, pipeline blowdowns, or mishaps; therefore, subpart W emission data are not available. Lamb et al. did not examine emissions from customer meters or pressure relief valve releases, and instead relied on the GRI/EPA study EFs in developing their national emissions for these sources. Lamb notes that customer meters were not included in their measurement program due to available data from the GTI 2009 study. For blowdowns and mishaps (dig-ins), Lamb et al. mailed surveys to LDCs that requested information on the number of events and the average methane estimated to be emitted per event. Comparing results of the Lamb et al. survey against the GHGI, the Lamb et al. survey resulted in a higher EF for mishaps (dig-ins) and a lower EF for pipeline blowdowns. The surveys conducted for both the GRI/EPA study and the Lamb et al. study had a limited number of respondents, so the Lamb et al. study combines the data sets to determine average emission factors based on the larger pool. Table 13 presents the EFs for mishaps (dig-ins) and pipeline blowdowns based on data collected in GRI/EPA, 15 ------- December 2015 based on new data collected in Lamb et al., and the combined data set EFs developed by Lamb et al. Note that Lamb et al. calculated their EFs for these sources differently than the GRI/EPA study; the GRI/EPA study calculates a weighted average in which company-level average EFs are weighted using the pipeline length over which the reported emissions occurred, while Lamb et al. calculates an unweighted average in which each company's reported average EFs are weighted equally. The GTI 2009 study conducted sampling of customer meters using screening and Hi-Flow Samplers to estimate leak rates; this technique is similar to the GRI/EPA study that is the basis of the GHGI EFs. The GTI 2009 study sampled 2,400 outdoor residential meters during six field tests; 395 commercial meters at six companies; and 46 industrial meters at five companies. An average EF was determined for each field test or company and an overall weighted average EF was then calculated based on the number of meters tested for each field test or company. A comparison of the EFs for each meter type is presented in Table 14. The GTI 2009 has a lower EF for residential meters, but higher EFs for commercial and industrial meters. The GTI 2009 study also identified a significant distinction between commercial and industrial meters, and developed unique EFs for different types of industrial meters, whereas the GRI/EPA study combined all commercial and residential meter data together. The GTI 2009 study determined that industrial meters have much higher emissions than commercial meters, and stated that the largest industrial meters more closely resembled a custody transfer station and had considerable vented emissions which were not identified in the GRI/EPA study. In the Clearstone report, residential meters were screened, and individual components (e.g., connectors, regulators, valves, diaphragm meters, and open-ended lines) of a meter were tested using a Hi-Flow Sampler. An EF for each component was determined, along with the average count of each of the components on a typical residential meter. The residential meter EF was then calculated as the summation of individual component EFs, using the average count of each component. A total of 1,883 residential meters were surveyed for the Clearstone report (it was not specified if the residential meters were outdoors or indoors). The residential meter EF from the Clearstone report is presented in Table 14. Table 13. Emission Factors for Pipeline Blowdowns and Mishaps (Dig-Ins) in the GRI/EPA Study and Lamb et al. Emission Source CH4 Emission Factor (Mscfy/mile) GRI/EPAJ Lamb et al.b Combinedc Pipeline Blowdowns 0.102 0.0061 0.054 Mishaps (Dig-Ins) 1.59 2.43 1.84 a. Calculated as a weighted average. b. Using new data from Lamb et al., calculated as an unweighted average. The EFs equal 0.0042 for pipeline blowdowns and 1.92 for dig-ins if calculated as a weighted average. c. Using all data points from GRI/EPA and Lamb et al., calculated as an unweighted average. The EFs equal and 0.086 for pipeline blowdowns 1.66 for dig-ins if calculated as a weighted average. Table 14. Comparison of Residential and Commercial/Industrial Customer Meter CH4 Emission Factors from the GRI/EPA Study, the GTI 2009 Study, and the Clearstone Report Emission Source CH4 Emission Factor (scfy/meter) GRI/EPA GTI 2009 Clearstone Residential Customer Meter 143.27 48.99 61.86 16 ------- December 2015 Emission Source CH4 Emission Factor (scfy/meter) GRI/EPA GTI 2009 Clearstone Commercial/Industrial Customer Meter 47.9 Commercial Meter = 505.4a Industrial Meter = 202,585b Industrial Using Commercial Meters = 445.1 * # meters0 Industrial Meter with Regulating Equip. = 443,746d - a. GTI noted that commercial meter EF is biased high by one large leak. If this leak is excluded the EF is 328 scfy/meter. b. A default EF is applied if no information is available to determine the type of industrial meter c. Applies if the industrial meter uses standard commercial diaphragm and turbine M&R sets. Assumes the industrial meter is equivalent to a grouping of multiple commercial meters. d. Applies if the industrial meter uses M&R station regulating equipment with continuous pneumatic venting devices National Estimates of Emissions from Other Distribution Sources Table 15 below summarizes emissions in the 2015 GHGI compared to calculated emissions using EFs from the Lamb et al. study and combined EFs (for certain sources), for years 2011 through 2013. The AD from the current GHGI are used for each set of emissions presented. For comparisons for year 2013 using GTI 2009 and Clearstone data on meters, refer to Table 17. Table 15. Methane Emissions for Other Sources (MT C02e) Emission Source 2011 2012 2013 2015 GHGI Lamb1 Combined 2015 GHGI Lamb' Combined-1 2015 GHGI Lamb1 Combined-' Customer meters- Residential 2,776,895 NA NA 2,444,068 NA NA 2,910,615 NA NA Customer meters- Commercial/lndustry 104,419 NA NA 104,111 NA NA 110,644 NA NA Pressure Relief Valve Releases 29,779 NA NA 29,981 NA NA 30,163 NA NA Pipeline Blowdown 64,053 3,848 34,029 56,376 3,387 29,951 67,137 4,034 35,668 Mishaps (Dig-ins) 998,469 1,522,958 1,153,245 878,797 1,340,423 1,015,022 1,046,550 1,596,295 1,208,779 NA - The Lamb et al. study did not determine a revised emission factor for this emission source a. Calculated by Lamb et al. using unweighted average emission factors shown in Table 13 GHGI Time Series Considerations for Emissions from Other Distribution Sources Limited data are available to determine how or if emissions from other distribution sources would be expected to significantly change over the GHGI time series due to industry technological advances. The GTI 2009 and Clearstone EFs for residential meters are both less than half of the GRI/EPA EF value. It is unclear whether this difference is the result of changes over time in average residential customer meter emissions, or an artifact of study design or methods. The EPA seeks feedback on whether these EFs reflect emissions in recent years but not earlier years (i.e., there have been industry advances that would result in lower average meter EFs in recent years) or whether these EFs represent additional available data that may be used in conjunction with the GRI/EPA study data to recalculate EFs for use across all GHGI years. 17 ------- December 2015 Regarding commercial and industrial meter emissions, GTI 2009 provides EFs that are ten times higher for commercial meters and thousands of times higher for industrial meters compared to current GHGI EFs from the GRI/EPA study. GTI 2009 specifically stated that certain high emitting industrial meters were not included in the GRI/EPA data, and as such, a higher EF for industrial meters is appropriate. The EPA seeks feedback on whether there are trends over time in commercial and industrial meter emissions that should be reflected in the time series. There are multiple orders of magnitude difference between the pipeline blowdown EFs from GRI/EPA and Lamb et al. Lamb et al. acknowledges that the pipeline blowdown EFs they developed from a limited voluntary survey have significant uncertainty, as is the case for GRI/EPA that based their pipeline blowdown EF on data from surveying four companies. It is therefore unclear whether the difference between GRI/EPA and Lamb et al. average EFs are the result of a change over time in how facilities implement pipeline blowdowns. The EPA seeks feedback on whether the more recent pipeline blowdown EF is representative of emissions in recent years but not earlier years (i.e., there have been industry advances that would result in lower average pipeline blowdown EFs in recent years) or whether the new EF represents additional available data that may be used in conjunction with the GRI/EPA study data to recalculate an EF for use across all GHGI years. Over the 1990-2013 time series, the Gas STAR program data show reductions achieved for pipeline blowdown and mishap (dig-in) minimization practices; see Appendix A. These were unique instances where facilities implemented practices to reduce pipeline blowdown or mishap emissions and reported reductions to Gas STAR. The Gas STAR data for pipeline blowdown emissions shows varying magnitudes of reduction. In recent years, the pipeline blowdown emission reductions are less than three percent of the GHGI emissions calculated for this source; however, in prior years, Gas STAR reductions equal approximately 35 percent of the GHGI emissions and for one year, 2005, the Gas STAR reductions were 146 percent of the GHGI emissions for pipeline blowdowns. Gas STAR reductions for mishaps in recent years account for just under two percent of the annual emissions, and for one year (2011) there are reductions equal to approximately ten percent of annual mishap emissions. Approach for Other Distribution Source Emissions under Consideration for the GHGI For residential, commercial, and industrial customer meters, data are available that could be used to update both the current GHGI AD and EFs. Customer meter AD are available for each year of the time series in Gas Facts reports. Using annual meter count data would improve accuracy compared to the current GHGI methodology of using 1992 counts driven by gas consumption. When determining the applicable AD for residential meters, GRI/EPA applied the percentage of outdoor meters in each region, as provided in Table 12, to the Gas Facts total count of residential end users; the EPA has not identified a data source to update these percentages. Using Gas Facts data to separate commercial and industrial meter AD would allow the EPA to apply unique EFs to each category, which could increase the accuracy of the GHGI. It should be noted that the Gas Facts methodology used to determine meter counts changed in 1996. Pre-1996 customer meter data were based on industry-reported numbers, but the entire industry did not report data, so the totals are estimates. Post-1996 customer meter data are reported by the entire industry. A comparison of customer meter activity data from Gas Facts and the GHGI is presented in Table 16 for recent years; the EPA is in the process of accessing and compiling data for earlier years. 18 ------- December 2015 Table 16. Customer Meter Counts from the GHGI and Gas Facts, Recent Years Year Customer Meters - GHGI Customer Meters - Gas Facts Residential Commercial/ Industrial Residential0 Commercial/ Industrial13 Commercial Industrial outdoor meters meters outdoor meters meters meters meters 2005 41,216,697 4,280,819 50,189,147 5,382,900 5,178,200 204,700 2006 37,303,114 4,168,356 50,980,751 5,474,700 5,274,900 199,800 2007 40,322,005 4,312,826 51,436,318 5,500,500 5,305,600 194,900 2008 41,773,665 4,381,970 51,756,432 5,501,800 5,307,300 194,500 2009 40,808,738 4,142,418 51,805,248 5,528,600 5,321,200 207,400 2010 40,834,355 4,429,256 51,960,164 5,491,600 5,299,100 192,500 2011 40,253,691 4,527,396 52,302,282 5,512,100 5,319,400 192,700 2012 35,429,055 4,514,014 52,853,737 5,544,900 5,355,600 189,300 2013 42,192,085 4,797,283 52,940,047 5,553,800 5,361,900 191,900 a. These values are not directly from Gas Facts - rather, the outdoor meter regional factors from Table 12 are applied to Gas Facts total residential meter counts to obtain these values, b. Equals the sum of Commercial plus Industrial meter counts. In both the GTI 2009 and Clearstone reports, which investigated residential meter emissions, the calculated EFs are significantly lower than the current GHGI EF. If a decreasing industry trend is supported by stakeholder feedback and other information, the EPA could apply an EF developed from the GTI 2009 and/or Clearstone data for recent years, use the current GRI/EPA EF for earlier years, and develop year-specific EFs assuming a linear correlation for the intermediate years (unless there was a specific year when an industry-wide change is recognized). Alternatively, if an industry trend is not supported by stakeholder feedback, the EPA may implement one of three approaches: (1) apply an EF developed from the GTI 2009 and/or Clearstone data to all years, and not use the older GRI/EPA residential meter EF; (2) apply an EF developed from all the available data (GTI 2009, Clearstone, and GRI/EPA) to all years; or (3) retain the current EF. Regarding commercial and industrial meter emissions, GTI 2009 EFs are ten times higher for commercial meters and thousands of times higher for industrial meters compared to current GHGI EFs from the GRI/EPA study. GTI 2009 specifically stated that certain high emitting industrial meters were not included in the GRI/EPA data, and as such, a higher EF industrial meters is appropriate. The EPA may implement one of three approaches for commercial and industrial meter EFs: (1) Use GTI 2009 factors and segregate commercial and industrial meter emission sources (which would necessitate the use of Gas Facts for updated activity data that break out commercial versus industrial meter counts); (2) develop an updated EF for combined commercial and industrial meters using all available data from both GRI/EPA and GTI 2009; or (3) apply the GTI 2009 commercial meter EF to all commercial and industrial meters, recognizing that there are 395 data points in the commercial data set and only 46 widely varying emissions rates in the industrial data set—and in future Inventories reassess whether data are available for updating the industrial meter factor. Table 17 shows calculated year 2013 emissions for customer meters based on various potential approaches discussed above. The 2015 GHGI year 2013 emissions estimates are provided for reference. 19 ------- December 2015 Table 17. Year 2013 Customer Meter Methane Emissions Ca culated by Various Approaches EF AD 2013 Emissions EF & AD Data Source (scfy/meter) (# meters) (MT CO;e) Residential Meters 2015 GHGI EF& AD 143.27 42,192,085 2,910,615 GHGI EF/Gas Facts AD 143.27 52,940,047 3,652,062 GTI 2009 EF/Gas Facts AD 48.99 52,940,047 1,248,680 Clearstone EF / Gas Facts AD 61.86 52,940,047 1,576,766 Commercial & Industrial Meters GHGI (Commercial & Industrial) EF & AD 47.90 4,797,283 110,644 GTI 2009 (Commercial) EF / Gas Facts AD 505.40 5,361,900 1,304,824 GTI 2009 (Industrial) EF / Gas Facts AD 202,585 191,900 18,718,795 GTI 2009 EF / Gas Facts AD - Total Commercial n/a 5,553,800 20,023,619 & Industrial For pressure relief valve releases, the activity data are directly obtained for each year in the time series from PHMSA; the EFs currently used in the GHGI are the only EFs available based on studies reviewed. Therefore, the EPA is not proposing a revision to the current methodology. For pipeline blowdowns, the GHGI currently uses 1992 distribution main and service miles and scales this value for non-1992 years using relative residential gas consumption. However, scaling mileage based on residential gas consumption has introduced volatility across the time series that does not likely correlate to pipeline mileage trends (as gas consumption is affected by other factors such as equipment efficiency and climate). The EPA is considering revising the AD for this source to use annual data on total distribution main and service miles which are available directly from PHMSA.4 The total distribution miles estimated by PHMSA are higher than current GHGI activity estimates for every year of the time series, so national emissions for each year would increase. A comparison of total distribution main and service miles from PHMSA and the current GHGI is presented in Table 18. Table 18. Total Distribution Main and Service Miles from the GHGI and PHMSA Year 2015 GHGI PHMSA 1990 1,214,918 1,546,955 1991 1,260,384 1,560,633 1992 1,297,569 1,536,382 1993 1,371,267 1,612,973 1994 1,341,181 1,739,152 1995 1,341,905 1,700,449 1996 1,450,107 1,694,925 1997 1,378,827 1,734,443 1998 1,250,595 1,818,184 1999 1,307,420 1,764,724 2000 1,382,259 1,788,100 2001 1,320,055 1,838,359 2002 1,352,557 1,899,845 2003 1,405,270 1,872,748 2004 1,347,018 1,925,748 4 http://www.phmsa.dot.gov/pipeline/library/data-stats/annual-report-mileage-for-gas-distribution-systems 20 ------- December 2015 Year 2015 GHGI PHMSA 1990 1,214,918 1,546,955 1991 1,260,384 1,560,633 2005 1,335,392 1,962,351 2006 1,208,594 2,022,428 2007 1,306,404 2,025,685 2008 1,353,437 2,075,144 2009 1,322,174 2,086,642 2010 1,323,004 2,102,191 2011 1,304,191 2,120,902 2012 1,147,876 2,137,593 2013 1,366,993 2,149,299 Lamb et al. finds a much lower EF than the GHGI for pipeline blowdowns. As discussed above, Lamb et al. acknowledges that the pipeline blowdown EFs they developed from a limited voluntary survey have significant uncertainty, as is the case for GRI/EPA that based their pipeline blowdown EF on data from surveying four companies. The EPA seeks feedback on whether the more recent pipeline blowdown EF is representative of actual emissions in recent years but not earlier years (i.e., there have been industry advances that would result in lower Mscfy/mile average pipeline blowdown emissions in recent years) or whether the EF represents additional available data that may be used in conjunction with the GRI/EPA study data to recalculate EFs for use across all GHGI years. If an industry trend toward decreasing pipeline blowdown emissions over time is supported by stakeholder feedback and other information, the EPA could apply an EF developed from the Lamb et al. study data for recent years, use the current GRI/EPA EF for earlier years, and develop year-specific EFs assuming a linear correlation for the intermediate years (unless there was a specific year when an industry-wide change is recognized). Note that based on Gas STAR data, it appears that more facilities may be controlling blowdowns in post- 2000 years and as such, using Lamb's EF for years 2000 and beyond may be appropriate, while using the GRI/EPA EF for 1992 and assuming a linear correlation for intermediate years. Alternatively, if an industry trend is not supported by stakeholder feedback and other information, the EPA may develop a revised EF using all available data (both Lamb et al. and GRI/EPA), similar to the "combined" EF shown in Table 13, that would be applied across all years. The EPA may consider developing a weighted average in which company-level average EFs are weighted using the pipeline length over which reported emissions occurred over for each company. Table 19 shows 2013 emissions for pipeline blowdowns, based on various potential approaches discussed above. The 2015 GHGI year 2013 emissions estimates are provided for reference. Table 19. Year 2013 Pipeline Blowdown Methane Emissions Calculated by Various Approaches EF & AD Data Source EF (mscfy/mile) AD (miles) 2013 Emissions (MT COze) GHGI EF& AD 0.102 1,366,993 66,951 GHGI EF/PHMSA AD 0.102 2,149,299 105,266 Lamb EFa / PHMSA AD 0.0042 2,149,299 4,315 Combined EFa / PHMSA AD 0.086 2,149,299 88,518 a. These EFs are the calculated weighted average EFs provided in the Table 13 footnotes. 21 ------- December 2015 For mishaps (dig-ins), activity data are identical to pipeline blowdowns. Therefore, the EPA is considering using PHMSA annual data for total distribution main service miles as an improved methodology, as shown in Table 18. Regarding the EF for this source, Lamb et al. data show a higher EF compared to the current GHGI EF. Similar as one approach under consideration for pipeline blowdowns, it may be appropriate to develop a "combined" EF using all available data. The EPA may consider using a weighted average in which company-level average EFs are weighted based on the pipeline length over which reported emissions occurred over for each company. Table 20 shows 2013 emissions for mishaps (dig-ins), based on various potential approaches discussed above. The 2015 GHGI year 2013 emissions estimates are provided for reference. Table 20. Year 2013 Mishaps (Dig-ins) Methane Emissions Calculated by Various Approaches EF & AD Data Source EF (mscfy/mile) AD (miles) 2013 Emissions (MT CO;e) GHGI EF& AD 1.59 1,366,993 1,046,550 GHGI EF/PHMSA AD 1.59 2,149,299 1,645,471 Lamb EFa / PHMSA AD 1.92 2,149,299 1,986,980 Combined EFa / PHMSA AD 1.66 2,149,299 1,716,085 a. These EFs are the calculated weighted average EFs provided in the Table 13 footnotes. For each of the "other" sources, the averaging methodology for calculating EFs can be a weighted average (e.g., studies with more observations or companies with more observations or facilities with more observations carry more weight than those with less observations) or unweighted average calculation. Applying a more complex statistical procedure (e.g., fitting a certain distribution to the data such as Lamb et al. does for M&R stations and pipeline leaks) may not be justified for these sources due to the limited data set sizes. Uncertainty The most recent uncertainty analysis for the natural gas and petroleum systems emissions estimates in the GHGI was conducted for the 1990-2009 GHGI that was released in 2011. Since the analysis was last conducted, several of the methods used in the GHGI have changed, and industry practices and equipment have evolved. In addition, new studies and other data sources such as those discussed in this memorandum offer improvement to understanding and quantifying the uncertainty of some emission source estimates. The distribution segment studies evaluated for this memorandum present information on uncertainty. Lamb et al. calculates a mean and a 95th percentile EF for each emission source using a bootstrap procedure applied to a fitted probabilistic model (eight candidate probabilistic models were evaluated for each emission source). Clearstone reports a 95% confidence interval for each component type measured on residential meters. The GTI 2009 study did not present confidence intervals for the emission source EFs; instead, the GTI 2009 study generally discusses uncertainties present in the emissions data. As updates to the GHGI data and methods are selected, the EPA will review information on uncertainty and consider how the GHGI uncertainty assessment can be updated to reflect the new information. 22 ------- December 2015 Requests for Stakeholder Feedback M&R Stations 1. As noted above, the Lamb et al. study discussed changes in M&R stations that contributed to decreased emissions. The EPA seeks stakeholder feedback on the time frame of upgrades to M&R stations and information on whether the upgrades occurred as a gradual transition. The EPA seeks available data that would allow for activity and/or emission factors to be developed and applied as appropriate across the time series in order to calculate net M&R station emissions in each year. The Lamb et al. EF for two station categories (R-Vault 100-300 psi and R- Vault 40-100 psi) increased compared to the findings of the GRI study. The EPA seeks feedback on changes that took place at these subcategories of stations that resulted in increased emissions and over what time frame they occurred. 2. The EPA seeks feedback on the potential update to the GHGI for this source. The EPA seeks stakeholder feedback on whether the Lamb et al. M&R station EFs can be considered representative of the U.S. population in recent years, in both reflecting station upgrades and reflecting the subpopulation of superemitters. Pipeline Leaks 3. The EPA seeks information on factors that might impact a change in the leak rate and/or leak incidence over time. For example, based on the Lamb et al. study, the EF for two pipeline categories (protected steel mains and plastic services) increased compared to the findings of the GRI study. EPA seeks feedback on changes that took place at these subcategories of pipes that resulted in increased emissions and over what time frame they occurred. 4. Stakeholders have suggested that the EPA treat newer plastic pipeline and vintage plastic pipeline as two distinct categories in the GHGI. The EPA seeks available data that could be used to provide a time series of activity data for each category, and emissions data that could be used to develop emission factors for each category. 5. The EPA seeks information on whether Lamb et al. estimates, from measurements conducted during May through November (no measurements were collected during winter conditions), may over- or under- estimate average annual emissions, which may fluctuate based on temperature and resulting increases or decreases in throughput. Customer Meters 6. Residential customer meters - The EPA seeks stakeholder information on trends in the industry over time that would result in lower customer meter emissions (scfy/meter) in recent years compared to the early 1990's timeframe. 7. Commercial/Industrial customer meters - The EPA seeks stakeholder feedback on potential approaches to incorporate GTI 2009 factors. For example, the EPA could replace the current GHGI EF for commercial and industrial meters. The EPA could apply the GTI 2009 commercial 23 ------- December 2015 meter EF to the combined population of commercial and industrial meters, or could develop separate EFs. Alternatively, GTI 2009 and Clearstone study data could be used in conjunction with the GRI/EPA study data to recalculate EFs for use across all GHGI years). The EPA seeks information on trends over time that should be reflected in EF or AD in the time series. Other Issues and Revisions under Consideration 8. Pipeline blowdowns - The EPA seeks feedback on the Lamb et al. pipeline blowdown EF (which is lower than the GRI/EPA EF currently used in the GHGI). Is the new Lamb et al. EF representative of emissions in recent years but not earlier years (i.e., have there been industry advances that would result in lower Mscfy/mile average pipeline blowdown emissions in recent years)? 9. Mishaps/dig-ins - Lamb et al. data show higher emissions compared to the current GHGI EF. The EPA seeks feedback on whether industry trends have led to a higher EF from this source over time or whether the more recent EF could be applied over all years in the time series. Another option would be to use Lamb et al. data in conjunction with the GRI/EPA study data to recalculate EFs for use across all GHGI years. The EPA seeks feedback on these approaches. 10. Pressure Release Valves - The EPA seeks stakeholder information on available new data for this source. 11. Top down/bottom up discrepancy - The Lamb et al. study generally observed lower emissions than the GRI/EPA study. However, at least one top down study estimated that GRI/EPA factors underestimate emissions in distribution.5 The EPA is seeking stakeholder comment on potential causes for the discrepancy and how this information could be taken into account in the GHGI. 12. Hi-Flow Sampler- Much of the available measurement data on distribution segment emissions were developed using Hi-Flow samplers. A recent study, Howard 2015, highlights potential malfunctions in certain Hi-Flow instruments under certain conditions that can lead to underestimates. The EPA is seeking stakeholder feedback on the impacts of the Hi-Flow sampler issue on the results of studies highlighted here and whether are there methods for recalculating some of the data points to correct for it. In some studies, sources measured with the Hi-Flow sampler were also measured using other methods, such as LFE and tracer methods. The EPA seeks stakeholder input on this issue. 13. Natural gas leaks at point of use - In addition to the sources covered in the current GHGI and discussed in this memorandum, methane emissions also occur downstream of customer meters due to leaks at the point of use (e.g., domestic heating boiler cycling and pre-ignition losses from domestic and commercial gas appliances). Limited data are available on this emission source. At least one country, the United Kingdom, includes an emission estimate for this source in its national greenhouse gas emissions inventory. The 2012 estimate for gas leakage at the point of use for domestic boilers, domestic cooking appliances, and commercial gas appliances in the U.K. is 2.7 kt CH4, or 0.1 MMTC02e. 5 See, for example, McKain et al. Methane emissions from natural gas infrastructure and use in the urban region of Boston, Massachusetts. Proceedings of the National Academy of Sciences 112(7):1941-1946. 24 ------- December 2015 The U.K. calculation is based on U.K. specific data on boiler size, frequency of use, and other data. The EPA has not conducted a detailed analysis of boiler data to determine if U.K. emission factors are appropriate for the United States. The EPA has calculated a rough estimate of U.S. emissions using data on domestic and commercial gas consumption data for the U.S. and the U.K. In 2013, the U.S. residential and commercial gas consumption was around six times higher than that of the U.K. Scaling up the U.K. emissions based on relative consumption (and not factoring in differences between gas use in the two countries), emissions from natural gas leaks at point of customer use in the United States could be around 0.4 MMTC02e. The EPA seeks stakeholder feedback on the addition of this emission source to the GHGI, including available U.S.-specific emissions data for this source. 14. Drive around studies - EDF has conducted a series of leak detection studies in cities across the United States, using measurement technologies mounted on cars.6 While it is not possible to attribute methane leaks to specific sources from these studies (i.e., the leaks would include any methane above the detection limit, not limited to pipelines, and not limited to oil and gas), the EPA seeks stakeholder feedback on whether and how findings from these studies may be used to improve or analyze the GHGI. In the EDF studies, the areas with the highest emissions rate were Boston and Staten Island with 1 leak per mile. The lowest leak rate was in Indianapolis, with 0.005 leaks per mile. Other cities studied (Los Angeles, Burlington, Chicago, and Syracuse) had leak rates ranging from 0.1-0.5 leaks per mile. 6 https://www.edf.org/climate/methanemaps/city-snapshots 25 ------- December 2015 Appendix A Potential Methane Emissions and Gas STAR Emission Reductions in the 2015 GHGI for Distribution Sources 26 ------- December 2015 Table A-l. GHGI Potential CH4 Emissions and Gas STAR Reductions for Each Distribution Source from 1990 - 2001 (MT C02e) Emission Source Data Source 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Pipeline Leaks GHGI 22,864,392 21,992,559 21,143,900 21,272,936 21,773,345 20,356,192 19,616,926 19,671,393 19,045,693 18,989,456 18,860,339 18,784,088 Gas STAR - - - - -626 -963 -1,279 -1,279 -1,279 -1,279 -1,279 -1,279 M&R Stations GHGI 13,215,221 13,709,769 14,114,250 14,915,892 14,588,642 14,596,514 15,773,476 14,998,130 13,603,288 14,221,405 15,035,467 14,358,838 Gas STAR - - - - - - - - - - - - Customer Meters GHGI 2,686,009 2,786,091 2,869,074 3,031,273 2,969,469 2,976,654 3,211,340 3,056,453 2,779,244 2,898,222 3,059,635 2,917,334 Gas STAR - - - - - - - - - - - - Pressure Relief Valve Releases GHGI 22,731 21,388 21,401 22,360 24,084 24,116 23,477 25,732 24,552 24,193 25,242 26,462 Gas STAR - - - - - - - - - - - - Pipeline Blowdowns GHGI 59,668 61,901 63,728 67,347 65,869 65,905 71,219 67,718 61,420 64,211 67,887 64,832 Gas STAR - - - - - - - -71,370 - - - - Mishaps (Dig-Ins) GHGI 930,123 964,931 993,399 1,049,821 1,026,788 1,027,342 1,110,180 1,055,609 957,436 1,000,941 1,058,237 1,010,614 Gas STAR - - - - - - - - - - - - Emissions Not Assigned to a Specific Distribution Source GHGI - - - - - - - - - - - - Gas STAR - - - -513,375 -629,845 -492,326 -598,134 -686,053 -649,555 -823,019 -746,691 -923,383 Total Distribution GHGI 39,778,144 39,536,639 39,205,751 40,359,629 40,448,197 39,046,724 39,806,618 38,875,035 36,471,634 37,198,429 38,106,807 37,162,168 Gas STAR - - - -513,375 -630,471 -493,289 -599,413 -758,702 -650,834 -824,299 -747,970 -924,662 Table A-2. GHGI Potential CH4 Emissions and Gas STAR Reductions for Each Distribution Source from 2002 - 2013 (MT C02e) Emission Source Data Source 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Pipeline Leaks GHGI 18,305,657 17,533,813 17,849,214 18,038,376 18,118,267 17,850,262 17,706,776 17,314,283 16,406,145 16,157,165 15,785,925 15,291,200 Gas STAR -1,279 -1,279 -1,279 -1,279 -1,279 -1,279 -1,279 -1,279 -1,279 -1,279 -1,279 -1,279 M&R Stations GHGI 14,712,376 15,285,765 14,652,125 14,525,664 13,146,432 14,210,355 14,721,952 14,381,890 14,390,918 14,186,280 12,485,973 14,869,412 Gas STAR - - - - - - - - - - - - Customer Meters GHGI 2,989,464 3,098,394 2,974,794 2,942,060 2,669,488 2,881,078 2,982,816 2,910,725 2,919,108 2,881,314 2,548,179 3,021,259 Gas STAR - - - - - - - - - - - - Pressure Relief Valve Releases GHGI 27,292 26,595 27,884 27,989 28,535 28,916 29,073 29,327 29,560 29,779 29,981 30,163 Gas STAR - - - - - - - - - - - - Pipeline Blowdowns GHGI 66,428 69,017 66,156 65,585 59,358 64,161 66,471 64,936 64,977 64,053 56,376 67,137 Gas STAR - -11,653 -28,905 -95,532 0 -24,728 -23,455 -7,345 -13,246 -451 -1,202 -2,027 Mishaps (Dig-Ins) GHGI 1,035,497 1,075,854 1,031,256 1,022,356 925,282 1,000,163 1,036,171 1,012,237 1,012,872 998,469 878,797 1,046,550 Gas STAR -534 -1,749 -5,869 -6,373 -7,695 -9,196 -10,731 -20,009 -19,830 -117,178 -17,771 -21,150 Emissions Not Assigned to a Specific Distribution Source GHGI - - - - - - - - - - - - Gas STAR -3,887,260 -3,067,298 -2,728,606 -1,105,566 -1,560,635 -1,372,521 -1,223,590 -1,569,948 -1,330,404 -1,332,377 -1,109,580 -988,620 Total Distribution GHGI 37,136,714 37,089,437 36,601,429 36,622,030 34,947,361 36,034,937 36,543,259 35,713,398 34,823,580 34,317,060 31,785,230 34,325,720 Gas STAR -3,889,073 -3,081,979 -2,764,659 -1,208,750 -1,569,610 -1,407,724 -1,259,056 -1,598,582 -1,364,759 -1,451,285 -1,129,832 -1,013,076 27 ------- December 2015 Appendix B Study Design Information for New Data Sources 28 ------- December 2015 Emission Source Measurement Type # Sources Location & Representativeness EF Calculation Method Lamb et al. (2015) M&R Stations High Flow Sampler & Tracer Ratio 229 Stations at 14 companies: M&R > 300 = 59 stations M&R 100-300 = 10 stations Reg > 300 = 41 stations Reg 100-300 = 41 stations Reg 40-100 = 13 stations Reg < 40 = 1 station Vaults = 23 stations Spread across 12 states in the U.S. Used stratified random sampling to select locations. Companies accounted for 18% of the distribution pipeline mileage, 23% of the services, and 14% of the total gas delivered to customers in 2011. Lamb et al. determined distribution and applied probabilistic modeling to develop average EF. Pipeline Leaks High Flow Sampler & Tracer Ratio 230 leaks measured Same as M&R. Companies also have a similar distribution of pipeline material as compared to the national distribution. Lamb et al. determined distribution and applied probabilistic modeling to develop average EF. Pipeline Blowdowns Companies estimated emissions 4 LDCs estimated emissions for the survey Location information not provided. Lamb developed an unweighted average EF. Mishaps (Dig- ins) Companies estimated emissions 4 LDCs estimated emissions for the survey Location information not provided. Lamb developed an unweighted average EF. GTI 2009 (2009) M&R Stations High Flow Sampler 125 total stations, at 6 companies: District Regulator = 77 Pressure Limiting = 11 Custody Transfer = 37 Spread across five areas of the U.S. Stations selected based on a mixture of age, throughput, and equipment types. GTI developed a weighted average EF based on number of stations tested. Residential Meters High Flow Sampler 2,400 meters at 6 companies Spread across five areas of the U.S. Randomly selected meters. The meters tested equal approximately 0.05% of the meters in operation at the 6 companies. GTI developed a weighted average EF based on number of meters tested. Commercial Meters High Flow Sampler 836 meters at 6 companies Spread across five areas of the U.S. Randomly selected meters. The meters tested equal approximately 0.11% of the meters in operation at the 6 companies. GTI developed a weighted average EF based on number of meters tested. Industrial Meters High Flow Sampler 46 meters at 5 companies Spread across five areas of the U.S. Randomly selected meters. GTI developed a weighted average EF based on number of meters tested. 29 ------- December 2015 Emission Source Measurement Type # Sources Location & Representativeness EF Calculation Method Clearstone (2011) Residential Meters High Flow Sampler 1,883 meters at 9 companies in Canada Meters located in 9 different Canadian provinces. Clearstone determined individual component EFs and then summed (based on typical component count) to get average per meter EF. GTI (2013) Plastic Pipelines Hi Flow sampler and an enclosure to measure leaks above ground, and flow rate measurements using LFE device of isolated belowground segments Thirty leaks from 5 utilities were measured aboveground, a subset of 21 were also measured from belowground. The 5 utilities were located across the U.S.; sites randomly selected from locations identified in the leak records of the participating utilities Leak records of the PE mains had a small number of records with higher leaks and GTI introduced a weighted function to the measurements of the utility sites and field testing facility to develop an EF. GHGRP (2015) M&R Stations Emission Factors applied based on a subset of monitored stations for the number of leaking components (for above grade) or number of stations by inlet pressure (for below grade) 168 LDC facilities reporting calculated emissions in RY2013 Facilities are spread across the U.S., only those that meet a 25,000 mt CC>2e threshold report. Estimated to account for approximately 71%of all stations. For this memo, the EPA used total reported emissions for M&R stations and developed an average EF wherein each station is weighted equally. 30 ------- |