arv
United States
Environmental Protection
Agency
GRI-94 /0257.26
EPA - 600/R-96-080i
June 1996
Research and
Development
METHANE EMISSIONS FROM THE
NATURAL GAS INDUSTRY
Volume 9: Underground Pipelines
Prepared for
Energy Information Administration (U. S. DOE)
Prepared by
National Risk Management
Research Laboratory
Research Triangle Park, NC 27711
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before comp
1. REPORT NO. 2.
EPA-600/R~96-080i
4. TITLE AND SUBTITLE
Methane Emissions from the Natural Gas Industry,
Volumes 1-15 (Volume 9: Underground Pipelines)
5. REPORT DATE
June 1996
6. PERFORMING ORGANIZATION CODE
7. authoris) l. Campbell, M. Campbell, M. Cowgill, D. Ep-
person, M. Hall, M. Harrison, K. Hummel, D, Myers,
T. Shires, B. Stapper, C. Stapper, J. Wessels, and *
8. PERFORMING ORGANIZATION REPORT NO.
DCN 96-263-081-17
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Radian International LLC
P. C. Box 201088
Austin, Texas 78720-1088
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
5091-251-2171 (GRI)
68-D1-0031 (EPA)
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Air Pollution Prevention and Control Division
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final; 3/91-4/96
14. SPONSORING AGENCY CODE
EPA/600/13
15. supplementary notes EPA project officer is D. A. Kirchgessner, MD-63, 919/541-4021.
Cosponsor GRI project officer is R. A. Lott, Gas Research Institute, 8600 West Bryn
Mawr Ave., Chicago, IL 60631. (*)H. Williamson (Block 7).
16. ABSTRACTXhe 15_volume report summarizes the results of a comprehensive program
to quantify methane (CH4) emissions from the U.S. natural gas industry for the base
year. The objective was to determine CH4 emissions from the wellhead and ending
downstream at the customer's meter. The accuracy goal was to determine these
amissions within +/-0. 5% of natural gas production for a 90% confidence interval. For
the 1992 base year, total CH4 emissions for the U. S. natural gas industry was 314
+/- 105 Bscf (6.04 +/- 2.01 Tg). This is equivalent to 1.4 +/- 0.5% of gross natural
gas production, and reflects neither emissions reductions (per the voluntary Ameri-
Gas Association/EPA Star Program) nor incremental increases (due to increased
gas usage) since 1992. Results from this program were used to compare greenhouse
gas emissions from the fuel cycle for natural gas, oil, and coal using the global war-
ming potentials (GWPs) recently published by the Intergovernmental Panel on Climate
Change (IPCC). The analysis showed that natural gas contributes less to potential
globed warming than coal or oil, which supports the fuel switching strategy suggested
by the IPCC and others. In addition, study results are being used by the natural gas
industry to reduce operating costs while reducing emissions.
•)7. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COS at l Field/Group
Pollution
Emission
Greenhouse Effect
Natural Gas
Gas Pipelines
Methane
Pollution Prevention
Stationary Sources
Global Warming
13 B
14G
04A
21D
15 E
07C
18. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (ThisReport)
Unclassified
21. NO. OF PAGES
95
20. SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 <9-73)
-------
FOREWORD
The U.S. Environmental Protection Agency is charged by Congress with pro-
tecting the Nation's land, air, and water resources. Under a mandate of national
environmental laws, the Agency strives to formulate and implement actions lead-
ing to a compatible balance between human activities and the ability of natural
systems to support and nurture life. To meet this mandate, EPA's research
program is providing data and technical support for solving environmental pro-
blems today and building a science knowledge base necessary to manage our eco-
logical resources wisely, understand how pollutants affect our health, and pre-
vent or reduce environmental risks in the future.
The National Risk Management Research Laboratory is the Agency's center for
investigation of technological and management approaches for reducing risks
from threats to human health and the environment. The focus of the Laboratory's
research program is on methods for the prevention and control of pollution to air,
land, water, and subsurface resources; protection of water quality in public water
systems; remediation of contaminated sites and groundwater; and prevention and
control of indoor air pollution. The goal of this research effort is to catalyze
development and implementation of innovative, cost-effective environmental
technologies; develop scientific and engineering information needed by EPA to
support regulatory and policy decisions; and provide technical support and infor-
mation transfer to ensure effective implementation of environmental regulations
and strategies.
This publication has been produced as part of the Laboratory's strategic long-
term research plan. It is published and made available by EPA's Office of Re-
search and Development to assist the user community and to link researchers
with their clients.
E. Timothy Oppelt, Director
National Risk Management Research Laboratory
EPA REVIEW NOTICE
This report has been peer and administratively reviewed by the U.S. Environmental
Protection Agency, and approved for publication. Mention of trade names or
commercial products does not constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Information
Service, Springfield, Virginia 22161.
-------
EPA-600/R-96-080i
June 1896
METHANE EMISSIONS FROM
THE NATURAL GAS INDUSTRY,
VOLUME 9: UNDERGROUND PIPELINES
FINAL REPORT
Prepared by:
Lisa M. Campbell
Michael V. Campbell
David L. Epperson
Radian International LLC
8501 N. Mopac Blvd.
P.O. Box 201088
Austin, TX 78720-1088
DCN: 95-263-081-16
For
GRI Project Manager: Robert A. Lott
GAS RESEARCH INSTITUTE
Contract No. 5091-251-2171
8600 West Bryn Mawr Ave.
Chicago, IL 60631
and
EPA Project Manager: David A. Kirchgessner
U.S. ENVIRONMENTAL PROTECTION AGENCY
Contract No. 68-D1-0031
National Risk Management Research Laboratory
Research Triangle Park, NC 27711
-------
DISCLAIMER
LEGAL NOTICE: This report was prepared by Radian International LLC as an account
of work sponsored by Gas Research Institute (GRI) and the U.S. Environmental Protection
Agency (EPA). Neither EPA, GRI, members of GRI, nor any person acting on behalf of
either:
a. Makes any warranty or representation, express or implied, with respect to the
accuracy, completeness, or usefulness of the information contained in this report, or
that the use of any apparatus, method, or process disclosed in this report may not
infringe privately owned rights; or
b. Assumes any liability with respect to the use of, or for damages resulting from the
use of, any information, apparatus, method, or process disclosed in this report.
NOTE: EPA's Office of Research and Development quality assurance/quality control
(QA/QC) requirements are applicable to some of the count data generated by this project.
Emission data and additional count data are from industry or literature sources, and are not
subject to EPA/ORD's QA/QC policies. In all cases, data and results were reviewed by the
panel of experts listed in Appendix D of Volume 2.
ii
-------
RESEARCH SUMMARY
Title
Contractor
Methane Emissions from the Natural Gas Industry,
Volume 9: Underground Pipelines
Final Report
Radian International LLC
GRI Contract Number 5091-251-2171
EPA Contract Number 68-D1-0031
Principal
Investigators
Report Period
Objective
Technical
Perspective
Results
Lisa M. Campbell
Michael V. Campbell
David L. Epperson
March 1991 - June 1996
Final Report
This report describes a study to quantify the annual methane emissions
from underground pipelines in natural gas production, transmission, and
distribution.
The increased use of natural gas has been suggested as a strategy for
reducing the potential for global warming. During combustion, natural
gas generates less carbon dioxide (C02) per unit of energy produced than
either coal or oil. On the basis of the amount of C02 emitted, the
potential for global warming could be reduced by substituting natural gas
for coal or oil. However, since natural gas is primarily methane, a potent
greenhouse gas, losses of natural gas during production, processing,
transmission, and distribution could reduce the inherent advantage of its
lower C02 emissions.
To investigate this, Gas Research Institute (GRI) and the U.S.
Environmental Protection Agency's Office of Research and Development
(EPA/ORD) cofunded a major study to quantify methane emissions from
U.S. natural gas operations for the 1992 base year. The results of this
study can be used to construct global methane budgets and to determine
the relative impact on global warming of natural gas versus coal and oil.
The national annual emissions from underground pipelines, taking into
account soil oxidation, are: 41.6 ± 65% Bscf for distribution; 0.2 ± 89%
Bscf for transmission; and 6.6 ± 108% Bscf for production. Following is
a comparison of the methane emissions from underground pipelines to
the total methane emissions from all sources in each industry segment.
As shown, the total methane emissions from underground pipelines
HI
-------
represents about 15% of the total national methane emissions from the
gas industry.
COMPARISON OF METHANE EMISSIONS FROM
UNDERGROUND PIPELINES TO INDUSTRY-WIDE EMISSIONS
Segment
Total Industry
Emissions, Bscfy
Underground Pipeline
Emissions, Bscfy
Production
84.4
6.6
Processing
36.4
-
Transmission/Storage
116.5
0.2
Distribution
77.0
41.6
TOTAL
314
48
Based on data from the entire program, methane emissions from natural
gas operations are estimated to be 314 ± 105 Bscf for the 1992 base
year. This is about 1.4 + 0.5% of gross natural gas production. The
overall program showed that the percentage of methane emitted for an
incremental increase in natural gas sales would be significantly lower
than the baseline case.
The program reached its accuracy goal and provides an accurate estimate
of methane emissions that can be used to construct U.S. methane
inventories and analyze fuel switching strategies.
Technical A leak measurement technique was developed and implemented as a
Approach method to quantify methane emissions from underground pipelines in the
natural gas industry. A cooperative program was developed between
distribution companies volunteering to provide leakage measurements
and GRI/EPA. A total of 146 leak measurements have been collected by
the participating companies. These data were used to derive the emission
factors for estimating methane leakage from distribution, transmission,
and production underground pipelines. The leakage rate data were
adjusted for soil oxidation of methane based on the results- of a separate
study conducted at Washington State University and the University of
New Hampshire. The total emissions are a product of the emission
factor and activity factor, and are stratified by pipe use (mains versus
services) and pipe material categories to improve the precision of the
estimate.
iv
-------
In the distribution segment, activity factors were based on the national
database of leak repairs, allocated into pipe material categories based on
data provided by ten companies. These data were combined with
historical leak records provided by six companies. The activity factors
represent the number of equivalent leaks that are leaking year round,
with repaired leaks being accounted for as fractional equivalent leaks.
In the transmission and production segments, the emission factors were
based on the leak measurement data collected from distribution mains as
part of the cooperative leak measurement program. The activity factors
were derived from a nationally tracked database of pipe mileage/leak
repair records.
Project For the 1992 base year, the annual methane emissions estimate for the
Implications U.S. natural gas industry is 314 Bscf ± 105 Bscf (± 33%). This is
equivalent to 1.4% ± 0.5% of gross natural gas production. Results from
this program were used to compare greenhouse gas emissions from the
fuel cycle for natural gas, oil, and coal using the global warming
potentials (GWPs) recently published by the Intergovernmental Panel on
Climate Change (IPCC). The analysis showed that natural gas
contributes less to potential global warming than coal or oil, which
supports the fuel switching strategy suggested by IPCC and others.
In addition, results from this study are being used by the natural gas
industry to reduce operating costs while reducing emissions. Some
companies are also participating in the Natural Gas-Star program, a
voluntary program sponsored by EPA's Office of Air and Radiation in
cooperation with the American Gas Association to implement cost-
effective emission reductions and to report reductions to the EPA. Since
this program was begun after the 1992 baseline year, any reductions in
methane emissions from this program are not reflected in this study's
total emissions.
Robert A. Lott
Senior Project Manager, Environment and Safety
v
-------
TABLE OF CONTENTS
Page
1.0 SUMMARY 1
2.0 INTRODUCTION 3
3.0 PROGRAM OVERVIEW 4
4.0 TEST DESIGN 7
4.1 Population Stratification 7
4.2 Factorial Design and Implementation 10
5.0 SAMPLE SELECTION 13
6.0 LEAK RATE MEASUREMENT METHOD 17
6.1 Steps of Measurement Procedure 17
6.2 Soil Sampling/Analysis 21
6.3 Quality Assurance/Quality Control 23
7.0 DATA ANALYSES 24
8.0 EXTRAPOLATION METHOD 36
8.1 Emission Factor Development 36
8.1.1 Distribution Emission Factors 36
8.1.2 Transmission Emission Factors 38
8.1.3 Production Emission Factors 40
8.2 Activity Factor Development 40
8.2.1 Distribution Activity Factor 41
8.2.2 Transmission Activity Factors 46
8.2.3 Production Activity Factors 49
9.0 RESULTS AND CONCLUSIONS 54
9.1 Distribution Underground Pipeline Emissions 54
9.2 Transmission Underground Pipeline Emissions 55
9.3 Production Underground Pipeline Emissions 57
10.0 REFERENCES 59
APPENDIX A - Results of Outlier Tests for Plastic Pipe Leakage Data . A-l
APPENDIX B - Source Sheets for Underground Pipeline Leakage .... B-l
vi
-------
LIST OF FIGURES
Page
4-1 Primary Variable Stratification 9
5-1 Comparison of Miles of Main by Pipe Material-
Top 100 Distribution Systems versus 9 Program Participants 14
5-2 Comparison of Number of Services by Pipe Material-
Top 100 Distribution Systems versus 9 Program Participants 15
6-1 Schematic of the Test Procedure 18
7-1 Scatter Plot of Leakage Rate versus Soil Silt Content for
Distribution Mains 31
7-2 Scatter Plot of Leakage Rate versus Soil Clay Content for
Distribution Services 32
7-3 Scatter Plot of Leakage Rate versus Pipe Age for Cast Iron Mains 33
7-4 Scatter Plot of Leakage Rate versus Operating Pressure for
Protected Steel Services 35
A-l Frequency Histogram for Plastic Pipe Flow Rate Data A-5
A-2 Frequency Histogram for the Natural Logarithms of the Plastic Pipe
Flow Rate Data A-6
A-3 Depiction of the Fourths (FL and Fy), Fourth Spread (df), and
Boundaries (Fl-1.5x
-------
LIST OF TABLES
Page
4-1 Test Matrix for Leak Measurement Program 12
7-1 Summary of the North American Leak Measurement Data 25
7-2 Correlation Coefficients for Combined Main and Service Leak Data 27
7-3 Correlation Coefficients for the Main Leak Data 28
7-4 Correlation Coefficients for the Service Leak Data 29
8-1 Methane Leakage Rates for Underground Distribution Pipelines 37
8-2 Methane Emission Factors for Underground Distribution Pipelines 39
8-3 Methane Emission Factors for Underground Gathering Pipelines
in the Production Segment 40
8-4 National Leak Repairs Allocated by Pipe Material Category 42
8-5 Summary of Leak Record Data from Participating Companies 44
8-6 Summary of Activity Factors for Distribution Underground Pipelines 47
8-7 Summary of Activity Factors for Transmission Underground Pipelines 49
8-8 Site Specific Data on Gathering Line Mileage Per Gas Well 51
8-9 Summary of Activity Factors for Gathering Pipelines
in the Production Segment 53
9-1 Summary of Methane Emissions Estimate from Underground Distribution
Pipelines 54
9-2 Summary of Methane Leakage Estimate from Underground Distribution
Pipelines 56
9-3 Summary of Methane Emissions Estimate from Underground Transmission
Pipelines 57
9-4 Summary of Methane Emissions Estimate from Underground Production Pipelines 57
viii
-------
LIST OF TABLES (Cont'd.)
Page
A-l Results of the Outlier Tests A-3
A-2 Plastic Pipe Flow Rate Data and Natural Logarithms of the Flow Rates A-4
ix
-------
1.0 SUMMARY
This report is one of several volumes that provide background information
supporting the Gas Research Institute (GRI) and U.S. Environmental Protection Agency
Office of Research and Development (GRI-EPA/ORD) methane emissions project. The
objective of this comprehensive program is to quantify the methane emissions from the gas
industry for the 1992 base year to within + 0.5% of natural gas production starting at the
wellhead and ending immediately downstream of the customer's meter.
This report documents the estimation of methane emissions from underground
pipelines in natural gas production, transmission, and distribution. A leak measurement
technique was developed and implemented as a method to quantify methane emissions from
underground pipelines in the natural gas industry. A cooperative program was developed
between distribution companies volunteering to provide leakage measurements and GRI/EPA.
A total of 146 leak measurements have been collected by ten participating companies. These
data were used to derive the emission factors for estimating methane leakage from
distribution, transmission, and production underground pipelines. The leakage rate data were
adjusted for soil oxidation of methane based on the results of a separate study conducted at
Washington State University and the University of New Hampshire. The total emissions are
a product of the emission factor and activity factor, and are stratified by pipe use (mains
versus services) and pipe material categories to improve the precision of the estimate.
In the distribution segment, activity factors were based on the national
database of leak repairs allocated by pipe material categories based on data provided by ten
companies. These data were combined with historical leak records provided by six
companies. The activity factors represent the number of equivalent leaks that are leaking
year round, with repaired leaks being accounted for as fractional equivalent leaks. The
activity factors combined with the emission factors derived from the leak measurement data
were used to produce an annual methane emissions estimate. Annual methane emissions to
the atmosphere are 41.6 billion standard cubic feet (Bscf) accounting for soil oxidation, with
1
-------
a 90% confidence interval of ± 65%. Soil oxidation of methane reduces emissions from
distribution underground pipeline leaks by about 18%. The largest contributor to the overall
emissions estimate was cast iron mains, followed by unprotected steel mains and services.
In the transmission and production segments, the methane emissions estimate
was based on the emission factors derived from the leak rates measured on distribution mains
and on activity factors derived from a nationally tracked database of pipe mileage/leak
repairs. For transmission pipeline leakage, the annual methane emissions were 0.2 Bscf
accounting for soil oxidation, with a 90% confidence interval of ± 89%. For gathering
pipeline in gas production, the estimated annual methane emissions were 6.6 Bscf, with a
90% confidence interval of ± 108%.
2
-------
2.0
INTRODUCTION
Early in the Gas Research Institute (GRI) and U.S. Environmental Protection
Agency (EPA) methane emissions project, preliminary estimates were developed for each
source of methane emissions in the natural gas industry. These preliminary estimates were
used to prioritize sources of methane emissions in the natural gas industry for further
research. One source of methane emissions that was identified was leakage from
underground distribution mains and services. Leakage from underground piping systems is
caused by corrosion, material defects, and joint and fitting defects/failures. Based on limited
leak measurement data from two distribution companies, leakage from underground
distribution mains and services was targeted as a potentially significant source of methane
emissions from the gas industry.
A comprehensive program was developed and implemented by GRI and EPA
to expand the database of leakage measurements from underground pipelines in the gas
industry. This program was designed as a cooperative effort between participating
distribution companies and the program sponsors. The data collected from this cooperative
effort was not only used to develop an estimate of emissions in the distribution segment of
the gas industry, but was also used to extrapolate emissions to the underground pipelines
used in the production and transmission segments of the industry.
This report documents the overall approach used to estimate leakage from
underground pipelines in the U.S. natural gas industry. An overview of how the program
was developed and implemented is provided in Section 3. Test design and sample selection
are described in Sections 4 and 5, respectively. Section 6 documents the leak measurement
protocol used during the testing efforts. The data analyses and the extrapolation
methodology used to derive a national estimate of methane emissions from underground
pipelines are discussed in Sections 7 and 8, respectively. The results and conclusions are
presented in Section 9. This report is one of several volumes under the GRI/EPA methane
emissions project.
3
-------
3.0 PROGRAM OVERVIEW
A cooperative program was developed by GRI and EPA to improve the
precision of the leakage estimate from underground distribution mains and services in the
natural gas industry. Two companies had published the results from limited studies of the
magnitude of natural gas leakage from underground mains and services in their distribution
networks.1-2 These data were evaluated to project the overall magnitude of methane
emissions from the entire U.S. distribution segment of the industry. Based on the limited
data available, the overall estimates of leakage from the distribution segment were significant
and, therefore, targeted as a high priority for further research.
To devise a measurement program for leakage from underground pipelines, the
number of measurements, or samples, required needed to be determined. The determination
of the appropriate sample size was based on a number of considerations, including:
• Size of the population of mains and services in the distribution segment
of the gas industry;
• Nature and distribution of the leak rate (dependent variable) and any
influences on leak rate, such as pipe age and pipe material (independent
variables);
• Expected mean and variance of the dependent and independent
variables;
• Target accuracy for the final estimates;
• Anticipated actual accuracy of the final estimates; and
• Costs associated with collecting the required information for each
individual leak measurement.
The target accuracy was defined as an overall leakage estimate for an individual company
that would be within ± 25 % of the true value based on a 90% level of confidence. The
mean and associated variance from the available preliminary leak test data were used to
4
-------
calculate an estimated accuracy. Initial calculations suggested a required sample size of
nearly 500 leak tests for a simple random sampling scheme. To reduce the sample size, the
effect of various experimental designs and sampling schemes on total sample size was also
examined. Based on the assumptions surrounding a stratified, random sampling scheme, the
required number of leak tests could be significantly reduced by dividing or stratifying the
sample population into categories or strata that reduce the variability within strata. The final
calculation determined that a total sample size of at least 200 tests could potentially achieve
the defined target accuracy if a stratified sampling approach were adopted.
To quantify the methane emissions from distribution mains and services by
performing 200 tests, a cooperative program was developed between GRI/EPA and
distribution companies volunteering to participate in the study. The cooperative program was
developed to share the cost of performing extensive testing. The concept that sufficient
parameters could be identified to distinguish differences between leakage characteristics from
location-to-location and company-to-company was the underlying basis of the cooperative
program. By pooling data from each contributing company, a higher overall accuracy could
be achieved compared to the single contribution from an individual company.
To identify and select potential participants in the program, GRI invited a
representative cross-section of gas companies in the U.S. to participate. Over 30 companies
were contacted and invited to participate, which represented different geographical areas in
the U.S. The companies volunteering to participate were asked to perform a total of
20 tests. Of the original invitees, only nine U.S. companies elected to participate by
providing either data previously collected or the sites and resources required to complete the
testing. In addition to the nine U.S. companies, two Canadian companies and two European
companies volunteered to participate in the study. (Note: To date, only six U.S. companies
have provided data for this study.)
The cooperative program includes program planning, design, coordination, and
data analysis provided by GRI/EPA. GRI/EPA developed a standardized testing protocol to
5
-------
guide companies in making measurements and help ensure consistent testing and quality
assurance/quality control (QA/QC) procedures. A test plan for the cooperative program was
developed and issued to each company participating in the study.3 Training sessions were
held at several host sites to provide classroom and hands-on training to participants. To
date, about 146 leak tests have been performed by the participating companies out of the
targeted 200 tests.
6
-------
4.0 TEST DESIGN
As part of the test plan for the cooperative leak measurement program, a test
matrix was developed to meet the objectives of the program. The test matrix is based upon
the concept of a stratified sampling scheme.
4.1 Population Stratification
Stratification refers to division of the population in categories, or strata, which
are expected to have significantly different leak characteristics. The goal of stratifying the
population was to decrease the variability in the leak rate data within a given strata.
Controlling variability by stratification has the following advantages:
• Reduction in the total sample size required for the test program as
compared to a non-stratified sampling approach;
• Increase in the overall precision by segregating the population into
homogenous subsets;
• Reduction in the estimated error terms within the homogenous strata as
compared to the total error associated with the heterogenous, pooled
population;
• Improved ability to identify specific factors that influence gas leak
rates; and
• Provide a more detailed assessment of overall estimate accuracy
through the assessment of accuracy within each stratum.
Stratification was necessary to meet the target accuracy defined for the leakage estimates of
each participating company while minimizing the number of tests required. However,
employing an experimental design with too many stratifying variables, while potentially
identifying and partitioning the error terms with even greater detail, would likely result in a
prohibitively large sample size.
7
-------
Many factors that may influence below-ground natural gas leak rates were
identified by industry experts and engineering judgement. Several meetings were held with
an industry review group to develop the testing protocol and sample matrix. With all
potential influences identified that may impact leak rate, the test matrix required would have
been excessively large, defeating the purpose of defining parameters to reduce the overall
number of tests required. Therefore, industry expertise was used to rule out parameters that
were likely not important factors in influencing the leak rate. In this manner, the test matrix
was reduced to a manageable size. The prioritization of the influencing factors is a function
of both the suspected level of impact on leak rates and the availability of each factor in terms
of characterizing both the sample and target populations. The ability to extrapolate the final
sample leak rate estimates to the entire population was a vital component in the overall
program.
Stratification for this study was limited to three primary variables:
• Pipe use (i.e., mains versus services);
• Pipe material; and
• Pipe age.
The proposed classes or levels assigned to each of the primary stratifying variables were
selected using the available industry characterization data and engineering judgement from
distribution industry experts.
Using 4 material types and 3 age intervals, 12 possible strata were identified
both for mains and services, for a total of 24 possible strata. However, several age/material
combinations are not common in the population and were consequently omitted from the test
matrix, as shown in Figure 4-1. Therefore, a total of 16 strata, 8 for mains and 8 for
services, were identified for the test design.
8
-------
MAINS
SERVICES
l're-1940
1940-1969
1970-1990
Material
Unprotected
Steel
Protected
Steel
Plastic
Pre-1940
1940-1969
1970-1990
Material
Unprotected
Steel
Protected
Steel
Plastic
Copper
= Stratum included in design.
= Stratum omitted from design.
Figure 4-1. Primary Variable Stratification
-------
4.2
Factorial Design and Implementation
The remaining factors that may potentially influence below-ground leak rates
were also evaluated in the test design, and included:
• System leak detection and repair programs;
• Pipe operating pressure;
• Distribution system soil characteristics; and
• Pipe diameter.
To detect the impact of these secondary factors, leak tests were allocated within each stratum
through an embedded factorial test matrix. The factorial design assigns individual
observations to all combinations of the secondary parameters. The analysis of variance
(ANOVA) used in a factorial experiment determines which independent variables, both
singularly and in combination, significantly influence the dependent variable. The efficiency
afforded by factorials lies in the development of the factor/level combinations.
The factorial design defined two levels for the four secondary factors. The
two levels presented for each of the four factors were defined based on industry expertise.
Soil type defines coarse, sandy soils as porous and heavy, clay soils as nonporous. Ideally,
soil type would be a criteria for selecting suitable measurement sites. However, because of
practical constraints of the program, soil type selection was not a variable that could be
controlled.
The leak detection and repair classes were defined in an attempt to account for
company-specific differences in the standard leak survey and repair practices. The leak
survey, detection, monitoring, and repair practices for each participating company were rated
relative to one another based upon the stated company practices, with an appropriate level
assigned. A more definitive measure of the relative status of a company's leak detection and
10
-------
repair practices was also included as part of the overall protocol for the leak measurement
program. This included a side-by-side comparison of the company's standard leak survey
procedures with a standardized, rigorous procedure outlined in the program test plan. By
comparing the number of leaks identified using both survey procedures on the same sections
of pipe, along with the average number of leak indications per mile, a relative ranking of
leak detection and repair practices could be assigned to each company. (Note: To date, the
survey comparison evaluation has not been completed by all companies participating in the
program.)
Gas operating pressure was classified as high and low and is dependent on the
pipe use (i.e., main versus service) and pipe material type. Pipe diameter was also a factor
used in the factorial design because diameter was believed to possibly influence the leak rate
from cast iron pipe. (The leak rate of a single joint in a cast iron main may be a function of
the circumference of the joint.) Again, the large and small categories of pipe diameter
depend upon the pipe use and pipe material type.
To maximize the amount of information gained from the factorial design,
individual leak tests were assigned to the participating companies in two stages.
Relationships defined by the analysis of the initial or stage one leak tests were used to
determine the assignment of the stage two leak tests. With 16 strata (8 for mains and 8 for
services), the full factorial design would require 256 tests (16 tests in each strata) to observe
each of the 4 secondary factor/level combinations. The efficiency gained by employing a
two-stage sampling scheme permitted the test matrix used for the program to be reduced to a
half-factorial design. Table 4-1 shows the eight factor level combinations for each of the
primary strata (i.e., pipe service, material, and age).
11
-------
TABLE 4-1. TEST MATRIX FOR LEAK MEASUREMENT PROGRAM
Test Number
LDAR* Practices
Gas Operating
Pressure
Soil Type
Pipe Diameter
1
Good
High
Porous
Large
2
Good
High
Nonporous
Small
3
Good
Low
Porous
Small
4
Good
Low
Nonporous
Large
5
Fair
High
Porous
Large
6
Fair
High
Nonporous
Small
7
Fair
Low
Porous
Small
8
Fair
Low
Nonporous
Large
"LDAR = Leak detection and repair.
The stage one allocation included assigning roughly two-thirds of the total
number of tests each company committed to utilize this half-factorial design. A target of
128 tests were to be conducted within stage one of the testing program. The remaining
72 stage two tests would then be assigned to specific factor/level combinations, within
specific strata where additional data are needed to detect significant influences. The stage
two leak test allocations were consequently made in the strata with the greatest total
emissions and strata with highly variable leak rates. (Note: To date, most participating
companies have not completed all their tests. A total of 146 samples have been collected
with a resulting accuracy of the national emissions estimate from underground mains and
services of ± 65% based on a 90% level of confidence. This level of accuracy is well
within the target accuracy for the national emissions estimate from underground mains and
services and meets the 80% completeness criteria for the program.)
12
-------
5.0
SAMPLE SELECTION
To prevent bias in the final extrapolated emissions estimate from underground
distribution mains and services, the sample population should be representative of the
national population and the actual measurement sites chosen should be selected as randomly
as possible. The assumption that the underground sample population (i.e., program
participants) is representative of the target population (i.e., national distribution industry) was
an important consideration in developing the experimental design, defining the appropriate
sampling scheme, and assessing the accuracy of the leakage estimates. A comprehensive
industry characterization analysis suggests that the nine U.S. program participants are very
representative of the national industry with respect to pipe material and pipe age.
Figures 5-1 and 5-2 compare the nine participants to the top 100 U. S. gas distribution
systems. The ranking of the top 100 distribution systems is based on total miles of
underground mains. These 100 distribution systems account for roughly 80% of the total
national gas throughput. Figure 5-1 shows that the relative proportions of distribution main
pipeline materials for the nine companies are nearly equal to the proportions for the top 100
companies. Figure 5-2 shows that the program participants are representative of the national
population in terms of number of services broken down by pipe material.
Theoretically, the actual measurement sites chosen should be completely
random to eliminate bias in the selection. The factorial approach requires that leaks be
selected for testing that meet specific constraints; however, it was important that random test
sites be selected within the specifications of the test matrix. Therefore, criteria were
established to guide selection of leak measurement sites.
For test site selection associated with the segment testing method (entire
segment potentially containing multiple leaks is tested), the candidate selections for testing
were generally pipe sections that were being taken out of service. To prevent bias in the
segment selection for testing, the acceptable criteria for these segments included: 1) intact
pipe relocated due to proposed construction activities, and 2) pipe scheduled for an across the
13
-------
(Thousands)
350
300
250
200
150
100
Miles of Main
by Pipe Material
(Thousands)
5.3%
33.4%
B.0%
Bare Coated Bare Coaled Plastic Iron
Unprotected Cathodlcally
Steel Protected
Steel
46.(3%
23.7%
11.8%
Dare Coated Bare Coated Plastic Iron
Unprotected Cflthodlcally
Steel Protected
Steel
n
100
n. = 9
Figure 5-1. Comparison of Miles of Main by Pipe Material—Top 100 Distribution
Systems Versus 9 Program Participants
-------
(millions)
16
Number of Services
by Pipe Material
40.1%
36.1%
11.0%
5.7%
4.2%
2.7%
Rare Coated Rare Coated Plastic Iron Copper
Unprotected Cathodically
Steel Protected
Steel
n
100
(thousands)
3500
3000
2500
2000
1500
1000
500
8.5% 0.6%
30.3%
1.4%
39.5%
3.7%
0.0%
1/
Bare Coated Bare Coated Plastic Iron Copper
Unprotected Cathodically
Steel Protected
Steel
n
9
Figure 5-2. Comparison of Number of Services by Pipe Material-Top 100
Distribution Systems Versus 9 Program Participants
-------
board replacement program. To prevent biasing the test results, pipe was not chosen if it
was being replaced due to excessive leakage (for the pipe category tested).
Leaks due to corrosion in mains and services are believed to increase in size
(and leakage rate) with time. Therefore, for a distribution system with a multi-year survey
cycle, the leaks detected in the current year's survey should represent the largest leaks in the
system, and the leaks which have occurred in the segment of the system surveyed and
repaired in the previous year should represent the smallest leaks in the system. Therefore, to
accurately represent an average leakage rate for the entire network, leaks from both sections
(the section currently scheduled for survey and the section surveyed in the previous year)
should be selected for testing. Randomly selected areas within both of these sections of the
network should be surveyed (or resurveyed for the areas surveyed in the previous year) using
the standardized leak survey protocol (outlined in the test plan) to identify all detectable
leaks. The final selection of leaks for testing should be randomly drawn from the resulting
pool of detectable leaks in both sections surveyed.
16
-------
6.0
LEAK RATE MEASUREMENT METHOD
The basic technique used to measure the leakage rate from underground pipe is
based on a metered gas measurement procedure which has been used by several gas
companies to quantify their leakage contribution to unaccounted-for (UAF) gas. The general
procedure entails selecting a suitable test site; centering the leak(s); isolating the segment of
pipe containing the leak(s) (without disturbing the soil surrounding them); and measuring the
gas flow rate required to maintain the segment at normal operating pressure. Figure 6-1 is a
general schematic representation of the piping and associated gas routing system needed to
allow emission rate testing of an individual leak in a main pipeline. The steps of the
measurement procedure are described below.
6.1 Steps of Measurement Procedure
Identifying/Centering the Leak
Gas distribution operators use leak detection procedures to locate and classify
leaks for repair. To identify a leak in a section of pipe, a portable hydrocarbon analyzer or
flame ionization detector (FID) was used to screen immediately above the ground while
walking the pipeline. Any excursions above the background level (typically 2-3 ppm) may
indicate a nearby leak. The leak was centered by boring holes on each side of the pipe for a
short distance and determining the point of maximum leak concentration. To avoid
disturbing the soil immediately surrounding the leak, the depth of the barholes were specified
not to exceed approximately 12 inches above the level of the pipe. The gas concentration in
each barhole was measured, with the point of highest concentration typically being the most
probable location for the leak. Once the leak was centered, the site was prepared for testing.
Isolating the Pipe Segment
Once the most probable leak location was determined, a careful excavation
was made to expose the pipe at least 10 feet on each side of the leak without disturbing the
17
-------
Known Leak
to be Repaired
Test
Measurement
Apparatus
Regulator
set @ NOP
of Main
O
o
Gas Supply
(surrounding distribution
system, bottle gas,
DFM, etc.)
Figure 6-1. Schematic of the Test Procedure
18
-------
soil surrounding the leak. Using appropriate methods, the section of pipe containing the leak
was isolated by first routing the gas flow through a bypass line around the segment
containing the leak. The segment of pipe containing the leak was then physically severed
and capped off at each end of the live main and each end of the segment to be isolated. The
excavation and isolation procedures were performed without disturbing the soil or surface
covering immediately surrounding the leak.
Measuring the Leak Rate
After the segment of pipe containing the leak was isolated, it was equipped to
receive gas from the live main or pressurized gas cylinder through the leak measurement
apparatus. The isolated segment was first returned to its normal operating pressure and, then
after the pressure had stabilized, the gas flow rate required to maintain operating pressure
was measured. (For low pressure lines, less than 30 inches water column, the pressure in
the isolated segment would not reach the pressure of the live main due to the pressure drop
across the flow measurement equipment. This was overcome by using a pressurized gas
cylinder as the gas supply, or installing a U-tube manometer on the gas supply and isolated
segment to monitor the differential pressure. Although not deemed necessary, the gas flow
rate measurements could be adjusted by the differential pressure to more accurately represent
the leakage rate associated with the supply line pressure.)
Leak Measurement Apparatus — The leak measurement apparatus used in the
program consists of a series of progressively larger laminar flow elements (LFEs), which can
rapidly and very accurately measure a pressure differential across the LFE. The LFE causes
the flow to be laminar by passing through a series of capillaries within the device. The flow
rate is a linear function of the differential pressure within the range of the LFE, as specified
on a vendor supplied calibration curve specific to each instrument. Four LFEs were
specified for the test apparatus, to accurately cover a flow rate range between 0 and 450
scfh, with individual capacity ranges as follows:
19
-------
• 0-0.8 scfh;
• 0.8-10 scfh;
• 10 - 170 scfh; and
• 170 - 450 scfh.
For leakage measurements exceeding 450 scfh, a dry gas meter was used to
quantify the leakage rate. The leak measurement apparatus contains a temperature and
pressure gauge in order to convert the actual flow rates to standard conditions, along with an
inclined manometer to measure differential pressure.
Test Approach — The segment to be tested was either: 1) a service which
was isolated (capped-off) at the service-to-main connection and the customer's meter, 2) a
short segment of main (at least 20 feet long) containing the detectable leak which was
isolated by capping off both ends of the isolated segment and both ends of the live main, or
3) a long segment of main containing multiple leaks which was isolated by capping off each
end of the segment to be tested and each end of the live main.
For all pipe materials except cast iron, an individual leak test approach was
used. As previously described, the general procedure for testing individual leaks entailed
selecting and centering the leak, isolating the short segment of pipe containing the leak, and
measuring the gas rate required to maintain the isolated segment at normal operating
pressure. For services, the procedure was identical except that the isolated segment included
the entire service line from the main-to-service connection to the customer's meter. This
technique was based on testing leaks which are detected using leak survey procedures (i.e.,
detected leaks), and may exclude smaller or more diffuse leaks that are not detected at the
soil surface.
For cast iron mains, a segment test approach was used since many undetected
leaks are known to exist in cast iron. Cast iron pipe fittings are prone to leak because of
20
-------
their bell and spigot design, and the frequency of the fittings (every 10-14 feet). The general
procedure for testing entire segments of main included selecting the segment to be tested,
isolating the segment, and measuring the gas rate required to maintain the segment at normal
operating pressure. The resulting test data represent a leakage rate per unit length of main
which includes all sources of leakage in the segment, even leaks that may not be detected at
the soil surface. Based on a separate study of the oxidation of methane in the soil,4 many
small leaks from cast iron are oxidized before they reach the soil surface. The segment of
pipe tested was also surveyed to determine the number of detected leaks and the
corresponding concentration of methane detected for each leak in the segment.
6.2 Soil Sampling/Analysis
The key soil characteristics which were expected to affect leakage of gas from
distribution systems were divided into two groups.
Soil Characteristics Influencing Vapor Transport
The first group of parameters which likely influence vapor transport through
the soil and, therefore, leakage rate include:
• Porosity/bulk density;
• Moisture content; and
• Particle size distribution.
Porosity is the percentage of the total soil volume occupied by open pore space. Soil gas
diffusion rates are controlled to some extent by porosity. Bulk density is the mass of dry soil
per unit bulk volume, including the air space. The higher the bulk density, the lower the
porosity of the soil and, therefore, the lower the expected diffusion rate of gas through the
soil.
21
-------
The soil moisture content affects the gas diffusion rate since water can occupy
the pore space. Therefore, the higher the moisture content of the soil, the lower the
expected diffusion rate of gas through the soil. Particle size distribution also affects the soil
gas diffusion rates. Very fine soil particles, such as clays, tend to increase the compaction
capacity, thereby decreasing the diffusion rate of gas through the soil.
Soil Characteristics Influencing Corrosivity
The second group of key soil characteristics influence the corrosivity of the
soil. Although the corrosive nature of the soil does not impact the gas diffusion rate, it may
affect the frequency of leak occurrence and possibly the mean leak size. The soil parameters
which determine the corrosivity of the soil include:
pH;
• Resistivity or conductivity; and
• Moisture content.
The lower the soil pH, the higher the hydrogen ion content of the soil which promotes
greater corrosion potential. The lower the resistivity, the greater the potential for current
flow causing corrosion. The moisture content of the soil tends to lower resistance to current
flow and increases the corrosive conditions.
Soil samples were collected at each test site using a core sampling device and
sent to a laboratory for measurements of bulk density, particle size distribution, moisture
content, pH, and resistivity. Soil samples were collected at three different soil depths
relative to the pipe location for each test site: at the soil horizon, halfway between the soil
horizon and pipe location, and at the pipe location. Because of the possibility of damage to a
soil sample while in transit, bulk density (using a push penetrometer) and relative moisture
content (using a soil moisture meter) were also measured at each test site when the soil
samples were collected.
22
-------
6.3
Quality Assurance/Quality Control
Quality assurance specifications were developed as part of the testing protocol
to ensure that the data collected from various companies would be acceptable. In the design
of the test equipment, an in-line filter was specified to assure that any particulates or
moisture in the gas stream were removed before flowing through the laminar flow elements.
As long as the laminar flow elements are not contaminated by particulates or moisture, the
calibration should remain unchanged over the life of the equipment. Additionally, each
company was instructed to leak check the test apparatus before each measurement was made.
According to the manufacturer's specifications, the laminar flow elements have
an accuracy in the range of 1/2% of the reading. They are calibrated by the manufacturer
and include calibration curves specific to each instrument to convert from differential
pressure to standard flow rate. In addition to the manufacturer's calibration, the prototype
test assembly used for training purposes was calibrated using a bubbleometer standard to
confirm the manufacturer's calibration curves.
A scheduled audit of each company's test procedures was performed to
identify and correct any deviations from the specified testing protocol. During the audit,
calibrated orifices were used to check the calibration of each company's test apparatus. The
calibration checks were conducted using critical orifices as calibration standards as described
in 40 CFR Part 60, Appendix A, Method 5, Section 7.2.5
The leak measurement data obtained from the participating companies were
reviewed for consistency and proper interpretation and conversion. All measurement data
were checked for accuracy and reasonableness. The data were entered into a database,
which was thoroughly checked to ensure accurate entry and interpretation. Data validation
was performed on the measurement data including calculations for precision and accuracy
and comparison with the program objectives. Data potentially identified as outliers or
otherwise suspect were investigated.
23
-------
7.0 DATA ANALYSES
The leak measurement data collected to date include 108 data points from the
North American companies and 38 data points from the European companies participating in
the cooperative program. (Although the original program design specified that a total of 200
tests should be collected, the program still achieved an overall accuracy of ± 65% based on
a 90% level of confidence, which is within the target accuracy guidelines for the national
estimate.) These data were statistically analyzed to determine primary and secondary
influences on leakage. The data were first analyzed to establish whether they represent a
normal distribution and were concluded to be lognormally distributed.
Preliminary statistical test results indicate that the influence of pipe use and
pipe material is statistically significant. Therefore, the data were disaggregated by mains
versus services and by material types. Table 7-1 shows an overview of the North American
leak test data with a summary of the data disaggregated by mains versus services and by pipe
material. Table 7-1 presents the sample size or number of tests performed, the mean
emission rate, and the 90% confidence limits around the estimated mean emission rate. As
shown, there is a large variance in the mean leak rates for the data disaggregated by pipe
material, ranging from 2.6 to 12.5 scf/leak-hour for mains (excluding cast iron, with an
average leak rate of 0.009 scf/foot-hr) and from 0.4 to 2.5 scf/leak-hour for services.
The relatively high leak rate for plastic mains is due to a very small sample
size (six data points) with one large data point that skews the average emission rate.
Companies participating in the measurement program were encouraged to collect additional
test data on plastic mains to reduce the overall uncertainty. However, according to the
participants, leaks in plastic mains are relatively infrequent and suitable plastic main test sites
were not identified. Therefore, the large data point is likely not representative of the average
leakage rate from plastic mains, but no technical reasons to omit the data were identified.
24
-------
Statistical outlier tests were performed to determine whether the large plastic
main data point could justifiably be omitted from the data set. The results of the statistical
outlier tests conclude that the large data point cannot be excluded based on a statistical
evaluation. (The statistical outlier test results are discussed in Appendix A.) However, as
indicated in Sections 8.1.1 and 9.1, even though the data suggest that plastic mains may be
subject to relatively high leak rates on a per leak basis, plastic mains experience significantly
fewer leaks than the other pipe materials; therefore, the overall contribution to methane
emissions from plastic mains is small. (Note: If the large data point were excluded from
the dataset for plastic mains, the average leak rate would be 2.7 scf/leak-hour, a factor of
nearly 5 lower.)
TABLE 7-1. SUMMARY OF THE NORTH AMERICAN LEAK
MEASUREMENT DATA
Pipe Use
.;;; Pipe Material;:: i;;H£
Sample
Size
Average Leak Rate,
(scf/leak-honr)*
90% Confidence
Interval,
(scf/leak-hour)"1
Mains
Cast Iron
21
0.00935
0.0053c
Unprotected Steel
20
6.45
5.61
Protected Steel
17
2.55
2.01
Plastic
6
12.45
19.81
Services
Unprotected Steel
13
2.50
2.46
Protected Steel
24
1.15
0.62
Plastic
4
0.37
0.51
Copper
5
0.94
0.62
a Leak rate of natural gas (not adjusted for methane content or soil oxidation).
b 90% confidence interval around the mean value (upper bound minus the mean).
c scf/foot-hour.
A statistical analysis was performed on the leak measurement database to determine if any of
the secondary data parameters are influencing the leak rates and could potentially be used to
help predict the leak rates. First, the SAS CORR6 procedure was used to produce correlation
matrices for the parameters of interest. A strong correlation would indicate that the
25
-------
parameter might be correlated with the leakage rate (dependent variable) and, therefore,
useful in a predictive model. Next, scatter plots that corresponded to the correlations were
constructed to provide visual confirmation of the correlations. Finally, the SAS stepwise
regression option in the REG7 procedure was applied to the data parameters to help uncover
possible regression models.
Because the leak measurement data were collected for different pipe use (i.e..
mains versus services) and pipe materials, the data were divided into subsets of homogeneous
categories to remove confounding effects in the statistical analysis. Table 7-2 shows the
correlations of key parameters with the leak rate for mains versus services. The cast iron
mains were analyzed separately because the leak rate is expressed in terms of scf/leak-hour.
As shown by the low correlation coefficients, none of the correlations were statistically
significant at any of the common significance levels (i.e., 99%, 95%, or 90%). This is not
surprising for the combined main or service groups, because these groups contain different
pipe materials that may confound the statistical results.
Table 7-3 shows the correlations of the key parameters with the leak rate
broken down by pipe material for mains, and Table 7-4 shows the correlations for services.
For mains broken down by pipe material, none of the correlations were statistically
significant and only a few correlations were statistically significant for plastic or protected
steel services. Even for cast iron mains which have a sufficient sample size, no statistically
significant correlations were found with soil type, age, diameter, or operating pressure.
As footnoted in the tables, caution should be used for sample sizes of 10 or
less. This is because such small sample sizes do not contain enough information to
determine if the data really are correlated or not. Thus, the statistically significant
correlation coefficients for plastic and copper services should be discounted based on an
insufficient sample size. The correlation coefficients for protected steel services associated
with soil content should also be discounted because of insufficient sample size. And though
statistically significant with an adequate sample size, the leak rates from protected steel
26
-------
TABLE 7-2. CORRELATION COEFFICIENTS FOR COMBINED MAIN AND
SERVICE LEAK DATA
(sample size is indicated in parentheses)3
Leak Parameter1'
Mains0 (scf/leak-hr)
Services (scf/service-hr)
Sand % (top)
-0.10
(16)
0.41
(15)
Sand % (middle)
0.06
(16)
0.34
(15)
Sand % (bottom)
-0.00
(16)
0.38
(15)
Silt % (top)
0.22
(16)
-0.31
(15)
Silt % (middle)
0.11
(16)
-0.24
(15)
Silt % (bottom)
0.09
(16)
-0.27
(15)
Clay % (top)
-0.14
(16)
-0.39
(15)
Clay % (middle)
-0.21
(16)
-0.45
(15)
Clay % (bottom)
-0.12
(16)
-0.40
(15)
Operating Pressure (psig)
-0.03
(43)
-0.13
(46)
Pipe Age (year)
-0.00
(39)
-0.17
(40)
Pipe Diameter (inches)
-0.08
(43)
0.16
(46)
' None of the correlation coefficients are statistically significant at either the 99%, 95%, or 90%
significance levels.
b Top refers to near the soil surface; middle refers to halfway between the soil surface and the pipe
location; bottom refers to the pipe location.
c Excludes cast iron mains.
27
-------
TABLE 7-3. CORRELATION COEFFICIENTS FOR THE MAIN LEAK DATA
(sample size is indicated in parentheses)"
Leak Parameter'
Cast Iron
Unprotected Steel
Plastic
: Protected Steel
Sand % (top)
0.11
(41)
-0.14
(11)
-
-0.17b
(4)
Sand % (middle)
0.09
(41)
0.05
(ID
-
-0.04"
(4)
Sand % (bottom)
0.06
(41)
-0.04
(11)
-
0.40b
(4)
Silt % (top)
-0.03
(41)
0.33
(11)
-
0.05b
(4)
Silt % (middle)
-0.10
(41)
0.21
(11)
-
-0.2 lb
(4)
Silt % (bottom)
-0.02
(41)
0.17
(11)
-
-0.60b
(4)
Clay % (top)
-0.04
(41)
-0.21
(11)
-
0.27b
(4)
Clay % (middle)
-0.03
(41)
-0.28
(11)
-
0.46"
(4)
Clay % (bottom)
-0.08
(41)
-0.15
(11)
-
-0.00b
(4)
Operating Pressure
(psig)
0.06
(46)
-0.19
(20)
0.36b
(6)
0.19
(17)
Pipe Age (year)
0.22
(56)
0.14
(20)
0.82b
(4)
0.17
(15)
Pipe Diameter (inches)
-0.06
(57)
-0.21
(20)
-0.04b
(6)
-0.17
(17)
* None of the correlation coefficients are statistically significant at either the 99%, 95%, or 90%
significance levels.
b Caution should be used for sample sizes of 10 or less.
c Top refers to the near soil surface; middle refers to halfway between the soil surface and the pipe
location; bottom refers to the pipe location.
28
-------
TABLE 7-4. CORRELATION COEFFICIENTS FOR THE SERVICE
LEAK DATA
(sample size is indicated in parentheses)
Leak Parameter1
Unprotected
Steel
Copper
Plastic
Protected Steel
Sand % (top)
0.45"
(6)
-0.28*
(4)
-
-0.68*
(4)
Sand % (middle)
0.49"
(6)
0.08s
(4)
-
-0.53*
(4)
Sand % (bottom)
0.33'
(6)
-0.49"
(4)
-
0.42'
(4)
Silt % (top)
-0.38'
(6)
0.19*
(4)
-
0.79'
(4)
Silt % (middle)
-0.36'
(6)
-0.25s
(4)
-
0.81"
(4)
Silt % (bottom)
-0.31"
(6)
0.93'
(4)
-
0.40'
(4)
Clay % (top)
-0.56*
(6)
0.21a
(4)
-
-0.14'
(4)
Clay % (middle)
-0.61*
(6)
0.52®
(4)
-
-0.06'
(4)
Clay % (bottom)
-0.34"
(6)
-0.13s
(4)
-
-0.90'
(4)
Operating Pressure
(psig)
-0.18
(13)
0.25'
(5)
o.9 rb
(4)
0.46"
(24)
Pipe Age (year)
0.05*
(10)
-0.27'
(5)
0.20'
(3)
0.14
(22)
Pipe Diameter (inches)
0.12
(13)
-0.31'
(5)
-0.06'
(4)
-0.37b
(24)
* Caution should be used for sample sizes of 10 or less.
b Indicates that the correlation coefficient is statistically significant at the 90% significance level.
c Top refers to near the soil surface; middle refers to halfway between the soil surface and the pipe
location; bottom refers to the pipe location.
29
-------
services were only marginally correlated with the operating pressure or pipe diameter.
However, since no other pipe use/material combinations showed a similar trend in correlation
between leak rate and operating pressure or pipe diameter, the correlation for protected steel
services is suspect. Thus, the correlation analysis indicated either inconclusive results for the
data from small samples or no correlations when sample sizes were sufficient, with the
exception of marginal correlations for operating pressure and diameter for protected steel
services.
Scatter plots were constructed to visually confirm the results of the correlation
analysis. First, a separate scatter plot was constructed for the main and service categories
shown in Table 7-2. Figure 7-1 shows the scatter plot for the mains leak data (including
protected steel, unprotected steel, and plastic mains) versus the soil silt content (%), which
gave the largest, but still insignificant, correlation coefficient for this group shown in
Table 7-2. For each of the pipe materials separately, the sample size was not large enough
to provide dependable information for the results of the correlation analysis; namely, the leak
data were not correlated with the soil silt content. The scatter plot confirms this by showing
the large overall variability in the data and the variability within the few data points for each
pipe material.
Figure 7-2 shows the scatter plot for the service (including copper, plastic,
protected steel, and unprotected steel services) leak data versus the soil clay content (%),
which gave the largest, but still insignificant, correlation coefficient for this group shown in
Table 7-2. Again, in this case the sample size was not large enough to provide dependable
information for the results of the correlation analysis for any of the pipe materials; namely,
the leak data were not correlated with the soil clay content. The scatter plot confirms this by
showing the large spread among the few data points for each pipe material.
Figure 7-3 shows the scatter plot for the cast iron mains leak data versus the
pipe age. As shown in Table 7-3, although the sample size was large enough to draw a
conclusion, the leak data were not correlated with the pipe age.
30
-------
Leakage Rate Versus Soil Silt Content
o
0
O
-O-
10
20 30
Soil Silt Content % (top)
40
50
60
© Unprotected Steel ~ Plastic
o Protected Steel
Figure 7-1. Scatter Plot of Leakage Rate Versus Soil Silt Content for Distribution Mains
-------
20
Leakage Rate Versus Soil Clay Content
•
0
u
E
01
¦12
u_
O
w
<->
£
*
o
U_
"S
IB
T3
C
S
M
15
10
0
0
o
o
o
—A-
O
10 15 20
Soil Clay Content % (middle)
25
30
o Unprotected Steel © Copper
~ Plastic
m Protected Steel
Figure 7-2. Scatter Plot of Leakage Rate Versus Soil Clay Content for Distribution Services
-------
0.16
0.14
5 0.12
u.
0
6 0.10
1
* 0.08
0
ijj 0.06
<5
1 0 04
55
0.02
0.00
Leakage Rate Versus Pipe Age
1890 1900 1910 1920 1930 1940 1950 1960
Pipe Age (year)
Figure 7-3. Scatter Plot of Leakage Rate Versus Pipe Age for Cast Iron Mains
-------
Figure 7-4 shows the pipe material-specific scatter plot for the protected steel
service leak data versus the operating pressure of the pipe, which gave a statistically
significant correlation coefficient of 0.46. This coefficient was considered marginal, and
Figure 7-4 shows that the data are very scattered, which supports classifying this relationship
as weak or inconclusive. Since the relationship between leak rate and operating pressure for
protected steel services was the strongest, but still considered inconclusive, the overall data
suggest that there is not a correlation between leak rate and normal operating pressure.
[Note: For a given pipeline system, when the pressure is increased above the normal
operating pressure, the leak rate does increase. (For a given hole size in the pipe, the leak
rate is a function of pressure.) However, when comparing one system operating at a lower
pressure to another system at a higher pressure, the data suggest that the leak rates are not
dependent on the normal operating pressure of the systems.]
Even though the correlation analyses and scatter plots gave no promise of
developing a predictive model from the leak database, a preliminary stepwise regression
analysis was performed to confirm the correlation analyses results. The stepwise regression
analysis selects the single parameter that gives the best model R2 (i.e., the most variation is
explained by this single parameter.) Then, the parameter that increases the R2 the most is
added to the model, and so on. This procedure was applied using all of the data parameters
for which correlations were examined (see the listings in Table 7-2 through Table 7-4) for
each of the service/pipe material groups. In many cases, no parameter met the input criteria
for entry into the model. That is, a specified small amount of variance must be explained.
Models could only be generated from this procedure for two groups: copper services and
protected steel services. However, only four data points were available for the soil
composition data for both groups, which makes the results inconclusive. Thus, even when
forcing a model, no useful results were obtained.
34
-------
Leakage Rate Versus Line Pressure
© ©
© @ ®
0 a 1 t I
10 20 30 40 50 60
Line Pressure (pslg)
Figure 7-4. Scatter Plot of Leakage Rate Versus Operating Pressure for Protected
Steel Services
-------
8.0 EXTRAPOLATION METHOD
The calculated emissions estimate from underground pipeline leaks is the
product of an emission factor and an activity factor. The emission factor is derived from the
leak measurements provided by participating companies. As previously discussed, the
participating companies measure gas leakage from underground mains by testing either
individual leaks or leaking services (units = standard cubic feet/leak-hour) or pipe segments
(units = standard cubic feet/mile-hour).
The activity factor is derived by combining national leak repair records for
underground pipelines with leak history data provided by the participating companies, to
determine the total number of leaks. (For cast iron mains that are tested using a segment
method, the activity factor is the total mileage of cast iron pipe in the United States.) Leak
estimates were derived from historical leak records provided by the participating companies
in combination with nationally tracked statistics of leak repairs. A detailed discussion of the
approach used to estimate the total number of leaks for each pipe material category is
included in Section 8.2.
8.1 Emission Factor Development
The emission factors were derived from the leak measurement database,
segregated into the pipe use (mains versus services) and pipe material categories. The
emission factors for distribution pipelines were also applied to underground transmission and
production pipelines as discussed in Sections 8.1.2 and 8.1.3, respectively.
8.1.1 Distribution Emission Factors
The emission factors for underground distribution pipelines represent the
average leak rates for each pipe use/material category. The leakage rates previously
presented in Table 7-1 were adjusted for the average methane content of pipeline quality gas.
36
-------
93.4 volume percent. The methane emission rates on an hourly basis were then converted to
an annual basis, assuming that each leak or leaking segment is continuously leaking year
round. The fact that many leaks are repaired throughout the year and, therefore, are not
leaking year round is accounted for in the activity factor. (Leaks that are repaired during the
year are counted as partial leaks.)
Table 8-1 presents the average methane leakage rates from underground
distribution mains and services, stratified by pipe material. As shown, the leakage rate for
unprotected steel is significantly higher than for protected steel for both mains and services.
As previously mentioned, the leakage rate for plastic mains is even higher than for
unprotected steel mains, because of a small sample size with one very large leak
measurement. As discussed in Appendix A, the single large leak measurement cannot be
justifiably excluded from the data set based upon the results of statistical outlier tests.
However, since plastic mains have significantly fewer leaks than other pipe material
categories, their overall contribution to emissions is small.
TABLE 8-1. METHANE LEAKAGE RATES FOR UNDERGROUND
DISTRIBUTION PIPELINES
Pipe Use
Pipe Material
Average Methane
Leakage Rate'
(scf/leak-yr)
90%
Confidence Interval,**
(scf/leak-yr)
Mains
Cast Iron
399,867c
227,256c
Unprotected Steel
52,748
45,876
Protected Steel
20,891
16,479
Plastic
101,897
162,102
Services
Unprotected Steel
20,433
20,130
Protected Steel
9,438
5,064
Plastic
3,026
4,134
Copper
7,684
5,061
' Methane leakage rate, not adjusted for soil oxidation.
b 90% confidence interval around the mean value (upper bound minus the mean).
c scf/mile-year.
37
-------
The leakage rate for cast iron mains was derived from the data collected in
North America, which represents a sample size of 21. Since the leakage characteristics of
cast iron mains from the European participants are different than the North American
companies, only data from the North American companies were used to derive the cast iron
emission factor for the United States.
The methane leakage rates were adjusted for soil oxidation based on data
presented in Soil Consumption of Methane from Natural Gas Pipeline Leaks.4 The soil
oxidation rates of methane were experimentally determined to be a function of the methane
leakage rate, depth of pipe, soil moisture content, and soil temperature. In general, the
larger the leakage rate per leak, the lower the soil oxidation rate. Because of the variation
in leakage rates among the pipe material categories, the average soil oxidation rates are
different for the various pipe materials. Table 8-2 shows the methane emission factors for
distribution pipelines adjusted for soil oxidation. The precision of the soil oxidation
adjustment was assumed to be ± 25% based on engineering judgement. Therefore, the
overall 90% confidence interval was calculated by propagating the errors for the leakage
rate estimate and the soil oxidation rates.
8.1.2 Transmission Emission Factors
Leak survey practices for transmission lines are generally more stringent than
for distribution mains. Transmission lines are required to be surveyed annually, and more
frequently in populated areas. In addition, many transmission companies perform additional
routine aerial surveys to monitor the transmission lines for leakage. Based on conversations
with several transmission companies, any leaks found in the pipe wall are extremely small
and are repaired immediately for safety reasons. Based on the rigorous leak survey and
repair practices of transmission companies (i.e., leaks are discovered and repaired earlier in
transmission lines), the average leak rate from a transmission line is believed to be of the
same order of magnitude as a leak found in a distribution main, even though there may be a
substantial difference in the operating pressure of the pipelines. (Note: As discussed in
38
-------
TABLE 8-2. METHANE EMISSION FACTORS FOR UNDERGROUND
DISTRIBUTION PIPELINES*
Pipe Use
Pipe Material
Average
Methane
Leakage Rate
(sef/leak-yr)
Soil
Oxidation
(%)
Average
Emission
Factor"
{scf/leak-yr)
90% Confidence
Interval*"1'
(scf/leak-yr)
Mains
Cast Iron
399,867°
40.3
238,736'
152,059°
Unprotected
Steel
52,748
1.8
51,802
48,212
Protected Steel
20,891
3.0
20,270
17,243
Plastic
101,897
2.0
99,845
165,617
Services
Unprotected
Steel
20,433
1.1
20,204
21,129
Protected Steel
9,438
2.6
9,196
5,581
Plastic
3,026
21.2
2,386
3,412
Copper
7,684
0
7,684
5,559
4 Adjusted for soil oxidation of methane.
b 90% confidence interval around the mean value (upper bound minus the mean).
c scf/mile-year.
Section 7, data from the underground distribution program suggest that the leakage rate is not
a function of the pipeline operating pressure. When comparing one pipeline system at a
lower operating pressure to another system at a higher operating pressure, no discernible
difference in leakage rate is observed, based on the current data available. One possible
explanation for this observation is that most leaks are detected and subsequently repaired at
around the same surface threshold concentration, regardless of system operating pressure.)
Therefore, the emission factors for leakage from transmission pipelines are
based on the average leakage rates for main pipelines from the cooperative distribution
leakage measurement program. A mean value of the estimated leak rate per leak was
calculated using the test data, for all pipe materials except cast iron. For cast iron mains, a
segment test approach was used which quantifies the leakage rate for a long isolated segment
of pipe; therefore, the mean leakage rate for cast iron is in terms of leakage per unit length
of pipe. The natural gas leak rate was adjusted for methane content by multiplying by the
39
-------
volume percent of methane for transmission (93.4 volume percent). The methane emission
factor was also adjusted for soil oxidation of methane. The value of the emission factor and
90% confidence interval for each pipe material category is identical to that shown in
Table 8-2 for distribution mains.
8.1.3 Production Emission Factors
The emission factors for leakage from gathering pipelines are based on the
average leakage rates for main pipelines from the cooperative distribution leakage
measurement program. The natural gas leakage rates were adjusted for methane content by
multiplying by the average volume percent of methane in the production segment
(78.8 volume percent), and adjusted for soil oxidation of methane. The resulting emission
factors for gathering pipelines in the production segment of the industry are shown in
Table 8-3.
TABLE 8-3. METHANE EMISSION FACTORS FOR UNDERGROUND
GATHERING PIPELINES IN THE PRODUCTION SEGMENT
Pipe Material
Average Methane
Emission Factor,*
(sd/teak-yr)
90% Confidence
Interval,*"
(sc(/leak-yr)
Protected Steel
17,102
14,548
Unprotected Steel
43,705
40,675
Plastic
84,237
139,729
Cast Iron
201,418°
128,290c
J Adjusted for soil oxidation of methane.
b 90% confidence interval around the mean value (upper bound minus the mean).
c scf/mile-year.
8.2 Activity Factor Development
The methodology used to derive the total leaks in underground distribution
mains and services is presented in Section 8.2.1. Sections 8.2.2 and 8.2.3 present the
activity factors for the transmission and production sectors of the gas industry, respectively.
40
-------
8.2.1
Distribution Activity Factor
Since the emission factor for quantifying emissions from underground
distribution mains and services was stratified by pipe use (mains versus sendees) and by
pipe material (i.e., cast iron, cathodically protected steel, unprotected steel, plastic, and
copper), the activity factor was also stratified to extrapolate emissions.
With the exception of cast iron main pipeline, the activity factor used to
extrapolate the leakage estimate for underground distribution mains and services was the
number of annual equivalent leaks. (For cast iron pipeline, the activity factor was the total
mileage of cast iron mains in the United States, which is a nationally tracked statistic.8)
Annual equivalent leaks are defined as the number of leaks that leak continuously year
round. For example, if leaks that are repaired during the year are leaking for half the year,
on average, then each repaired leak would be counted as half an annual equivalent leak.
The number of annual equivalent leaks was derived from the national
database of leak repair records broken down by mains and services [U.S. Department of
Transportation (DOT), Research and Special Programs Administration (RSPA)].8 To
allocate leak repairs into pipe material categories, data were collected from ten local
distribution companies representing different regions within North America. The average
leak repairs per mile or per service based on the company data was multiplied by the
national miles/services,8 to provide the percentage of leak repairs in each material category.
The total number of nationally tracked leak repairs for mains and services, respectively, was
used to estimate the national leak repairs in each category. An estimate of the national leak
repairs allocated by pipe material type is shown in Table 8-4.
To derive annual equivalent leaks from the national leak repair records,
additional information was needed including the number of leaks found during the year
(leak indications) and the unrepaired leaks at the beginning of the year (outstanding leaks).
41
-------
TABLE 8-4. NATIONAL LEAK REPAIRS ALLOCATED BY PIPE MATERIAL CATEGORY
Pipe Use
Pipe Material
Average Leak
Repairs/Mile
or Service*
National
Miles/Servicesb
Extrapolated
Leak Repairs
Percent Leak
Repairs
Estimated
Total Leak
Repairs
90%
Confidence
Interval0
(Leak Repairs)
Mains
Cast Iron
1.38
55,288
76,400
33.8
69,776
42,382
Protected Steel
0.08
451,466
34,954
15.5
31,924
14,982
Unprotected
Steel
1.09
82,109
89,377
39.6
81,627
34,359
Plastic
0.08
299,421
25,189
11.2
23,006
24,134
Subtotal
225,920
100
206,333b
Services
Protected Steel
0.006
20,352,983
126,799
42.2
182,562
221,755
Unprotected
Steel
0.027
5,446,393
148,823
49.5
214,271
205,990
Plastic
0.001
17,681,238
22,367
7.5
32,202
27,067
Copper
0.011
233,246
2,507
0.8
3,608
3,517
Subtotal
300,496
100
432,643"
a Based on data provided by ten companies.
b Based on nationally tracked database, U.S. Department of Transportation, Research and Special Programs Administration.6
c 90% confidence interval around the mean value (upper bound minus the mean).
-------
Since leak indications and outstanding leaks are not tracked nationally, this information was
requested from individual companies.
Data were collected from the companies participating in the cooperative leak
measurement program on the annual number of leak repairs, number of leak indications, and
outstanding leaks at the beginning of the year (reference year in most cases was 1991). The
data were requested to be disaggregated by mains versus services and by pipe material.
Complete data were provided by only four companies, coupled with a breakdown of the total
mileage of mains and number of services by pipe material. Two additional companies
provided the data requested, although with no breakdown by pipe material or use. Table 8-5
shows the data from the North American companies that provided a complete set of data
required to estimate the total number of leaks in their distribution system. (Note: The leak
data disaggregated by pipe use and material type have been difficult to obtain from many
companies, since leak records are often not maintained in this manner.)
An estimate of the total annual equivalent leaks for each of the six companies
was developed for each pipe material category except cast iron, based on the following
methodology:
TEL = OL + LI + UDL + URL - (0.5 x RL)
where
TEL = Total annual equivalent leaks
OL = Outstanding leaks at the beginning of the year
LI = Leak indications recorded during the year, including call-ins
UDL = Undetected leaks which cannot be found using an industry standard
survey procedure
URL = Unreported leaks that have developed in parts of the network not
surveyed during the current year
RL = Repaired leaks — estimated to be leaking half the year, on average
Undetected leaks which cannot be found using an industry standard survey procedure were
quantified based on information provided by Southern Cross.9 According to their experience
43
-------
TABLE 8-5. SUMMARY OF LEAK RECORD DATA FROM PARTICIPATING COMPANIES
Company
Annual Leak
Indications
Annua! Leak
Repairs
Annual Outstanding
Estimated Total
Equiva lent Leaks
Ratio of Equivalent
Leaks to Leak Rcpairs
A
3,747'
2,06 r
0
3,378
1.64
B
9,249
17,003
11,701
18,796
1.11
C
2,115
2,443
0
2,832
1.16
D
b
14,681
__b
41,286
2.81
E
1,999
2,287
2,396
6,250
2.73
F
5,992
3,421
1,558
11,597
3.39
Average
2.14
90%
Confidence
Interval
0.79
* Mains only.
b Data not available.
-------
in performing leak surveys and survey audits, Southern Cross predicts that a standard
industry survey procedure using a flame ionization detector (FID) instrument finds 85 % of
the leaks. (Note: The standard industry survey procedure involves using either a walking or
mobile survey, as appropriate for the area being surveyed, using an FID instrument. Any
potential leak that is found with the FID instrument, registering a concentration above
background, is investigated using bar holing procedures.) Therefore, the number of
undetected leaks is estimated by:
UDL = [(1/0.85) - 1] x LI
The total annual equivalent leaks are derived using the estimated leak duration for each type
of leak, based on the following:
• Repaired leaks are assumed to be leaking half the year, on average.
• Outstanding leaks, leak indications, and undetected leaks are estimated
to be leaking the entire year (i.e., 8,760 hours per year).
The leak duration of unreported leaks is factored into the estimation methodology for these
leaks. Unreported leaks are those leaks which occur in parts of the network not surveyed
during the year (i.e., multi-year survey cycle). The number of unreported leaks is based on
the annual leak indications and the undetected leaks as well as the frequency of the leak
survey. The number of unreported leaks in the system that is surveyed every "n" number of
years is calculated based on the following:
• For the first year in the cycle — 1/n X (LI + UDL) leaks are leaking
half the year; (n-l)/n x (LI + UDL) leaks are not yet leaking.
• For the second year in the cycle — 1/n X (LI + UDL) leaks are
leaking the entire year; 1/n X (LI + UDL) leaks are leaking half the
year. (n-2)/n x (LI + UDL) leaks are not yet leaking.
45
-------
• For the third year in the cycle ~ 2/n x (LI + UDL) leaks are leaking
the entire year; 1/n x (LI 4- UDL) leaks are leaking half the year;
(n-3)/n x (LI + UDL) leaks are not yet leaking.
• For the fourth year in the cycle — 3/n X (LI + UDL) leaks are leaking
the entire year; 1/n x (LI + UDL) leaks are leaking half the year;
(n-4)/n x (LI + UDL) leaks are not yet leaking.
Based on the methodology described above, the number of equivalent leaks was estimated for
each of the six companies providing detailed data. The ratio of equivalent leaks to leak
repairs was then calculated for each of the companies. The average ratio of equivalent leaks
to leak repairs was used to extrapolate the national database of leak repairs. Table 8-5
presents a summary of the leak record data provided by the six companies, the estimated
equivalent leaks, and the corresponding ratio of equivalent leaks to leak repairs. As shown,
the average ratio of equivalent leaks to leak repairs is 2.14.
The national estimate of annual equivalent leaks, broken down by pipe use and
material type, is shown in Table 8-6. As shown, the activity factor for cast iron mains is
miles of pipeline, to correspond to the emission factor in units of scf/mile-year. The
estimate of annual equivalent leaks is highest for unprotected steel services, followed by
protected steel services. For mains, unprotected steel is the category with the highest
estimated annual equivalent leaks. The precision of the estimate is based on the variability in
leak repair data allocated by material type from ten companies and the variability in the ratio
of equivalent leaks per leak repair from six companies.
8.2.2 Transmission Activity Factors
The activity factors for the transmission segment were derived from the total
number of transmission pipeline leaks (excluding pipeline incidents) reported to the U.S.
DOT, RSPA.8 The leaks reported to RSPA include both repaired leaks (6,120 leaks) and
46
-------
TABLE 8-6. SUMMARY OF ACTIVITY FACTORS FOR DISTRIBUTION
UNDERGROUND PIPELINES
Average Activity
90% Confidence
Interval
Pipe Use
Material
Category
Estimated Total
Leak Repairs*
Factor
; (equivalent leaks)b
Activity Factor,'
(equivalent leaks)
Mains
Cast Iron
69,776
55,288d
2,764d
Unprotected Steel
81,627
174,657
101,685
Protected Steel
31,924
68,308
42,545
Plastic
23,006
49,226
58,018
Subtotal
206,333
Services
Unprotected Steel
214,271
458,476
499,850
Protected Steel
182,562
390,628
526,354
Plastic
32,202
68,903
66,840
Copper
3,608
7,720
8,521
432,643
1 Based on national leak repair database' and data provided by six companies (see Table 8-4).
b Based on estimated ratio of annual equivalent leaks to leak repairs of 2.14 (see Table 8-5).
c 90% confidence interval around the mean value (upper bound minus the mean).
d Miles.
outstanding leaks at the end of the year (1,369 leaks). Therefore, the total number of leak
indications is the summation of the repaired leaks and outstanding leaks at the end of the
year, or 7,489 leak indications:
Leak Indications = Leak Repairs + Outstanding Leaks
Because transmission lines are surveyed at least once per year using a
walking survey method, the number of undetected leaks is estimated based on the
effectiveness of the walking survey. According to one company specializing in distribution
surveys,9 roughly 85% of the leaks are found using a walking survey. This estimated
survey effectiveness was applied to transmission surveys, resulting in roughly 1,320
undetected leaks:
Undetected Leaks = [(Leak Indications / 0.85) - Leak Indications]
47
-------
The leak duration for outstanding leaks and undetected leaks is estimated to
be 8,760 hours per year, and the leak duration for repaired leaks is half a year
(4,380 hours/year), on average. Because transmission lines are surveyed at least once per
year, there are no unreported leaks from multiple year survey cycles. The resulting
estimate of equivalent leaks represents the number of leaks with a year round leak duration.
(Each leak repair is counted as half an equivalent leak to compensate for leak duration.)
Therefore, the equation used to estimate equivalent leaks is:
Equivalent Leaks = Leak Indications + Undetected Leaks
= (0.5 x Repaired Leaks) + (Outstanding Leaks + Undetected Leaks
= (0.5 x Repaired Leaks) +
[((Repaired Leaks + Outstanding Leaks)/0.85)-Repaired Leaks]
The total number of equivalent transmission pipeline leaks, 5,750. was
allocated on a pipeline material category basis in the same proportion (adjusted for the
fraction of mileage as well as the different leak frequency in each material category) as in
the distribution sector. (The ratio of percent leaks to percent miles in the transmission
segment is the same as the ratio in the distribution segment.)
The precision of the estimated total leaks was calculated based on the
estimated 90% confidence interval associated with each parameter in the activity factor
equation:
• Repaired leaks: ±100%;
• Outstanding leaks: ±100%;
• Leak duration: ±25%; and
• Leak survey effectiveness: ±15%.
48
-------
A statistical software program (@ RISK)10 was used to determine the overall 90%
confidence interval of the activity factor, ± 76%.
For cast iron transmission lines, the activity factor is the total mileage based
on the RSPA database for transmission and gathering lines. The precision of the estimate is
assumed to be ± 10%. Table 8-7 presents the transmission pipeline activity factors.
TABLE 8-7. SUMMARY OF ACTIVITY FACTORS FOR TRANSMISSION
UNDERGROUND PIPELINES
Pipe Material
Total Miles
Average Activity Factor,
(equivalent leaks)
90% Confidence interval
of Activity Factor,1
(equivalent leaks)
Protected Steel
287,155
5,077b
3,859b
Unprotected Steel
5,233
659
501
Plastic
2,621
14
11
Cast Iron
96
96*
10*
a 90% confidence interval around the mean value (upper bound minus the mean).
b Miles.
8.2.3 Production Activity Factors
The estimated number of leaks in field gathering pipelines is based on a leak
repair frequency for gathering lines owned and operated by transmission companies
reported in the RSPA database.8 This database reports an estimated 8,153 repaired leaks
and 270 outstanding leaks in 31,918 miles of gathering pipeline. The leak frequency is
derived by compensating for leaks that are repaired during the year and, therefore, not
contributing to leakage year round. On average, the repaired leaks are assumed to be
leaking for half the year, and each leak repair is counted as half an equivalent leak.
Outstanding and undetected leaks are assumed to be leaking the entire year.
Most production lines owned and operated by production companies are not
regulated by the U.S. DOT and many are not monitored for leaks in the rigorous fashion
employed by distribution and transmission companies. Therefore, undetected leaks are
49
-------
accounted for based on the effectiveness of the survey method performed, which is
estimated to find 35 and 85% of the total leaks for a vegetation and walking survey,
respectively. This is based on the experience of Southern Cross, a company specializing in
leak surveys.9 Based on limited data provided by production companies, production
company owned gathering lines are only surveyed using a vegetation method. However,
transmission company owned gathering lines are surveyed annually using a FID instrument
while walking the lines.
Based on an analysis of equivalent leaks (similar to the analysis presented for
transmission activity factor development in Section 8.2.2). the leak frequency is 0.18 leaks
per mile for a walking survey and 0.63 leaks per mile for a vegetation survey. This leak
frequency was used to ratio the number of leaks to the total estimated population of
gathering pipeline.
Gathering Pipeline Mileage
Total gathering pipeline mileage is not reported or tracked nationally and,
therefore, was determined from the value of the number of miles of gathering line per well
supplied by various production companies. The "gathering pipeline" designation includes
three categories of pipeline: 1) production company-owned gathering pipeline for gas wells
not associated with oil production (i.e., non-associated gas wells); 2) production company-
owned gathering pipeline for oil wells that produce marketed gas (i.e., associated gas
wells); and 3) transmission company-owned gathering pipeline. The third category of
transmission-owned pipelines are assumed to be in addition to the production pipeline miles
associated with wells. This is consistent with the site visit data since gathering lines owned
by transmission companies were intentionally excluded from the site mileage totals. (The
production companies did not report pipeline miles beyond their custody transfer meters.)
Total miles of gathering pipeline for non-associated gas wells were estimated
using site visit data from the thirteen production companies shown in Table 8-8. Seven of
50
-------
the thirteen sites provided estimates of their total miles of pipeline. The fifth site's mileage
was estimated from a map of its pipelines.
TABLE 8-8. SITE SPECIFIC DATA ON GATHERING LINE MILEAGE
PER GAS WELL
Site
Gathering Miles
Number of Wells
Miles per Total Wells
Site 1
46.3
80
Site 2
8
26
Site 3
40
130
Site 4
15.4
12
Site 5
11
6
Site 6
5.2
193a
Site 7
600
1000
Site 8
441.3
425
Site 9
0.7
1
Site 10
27.7
24
Site 11
2.1
3
Site 12
7.1
7
Site 13
154.2
126
TOTAL
1359.0
2033
0.67 ± 28%"
' Includes 55 oil wells.
b 90% confidence interval.
The estimate of total gathering miles per non-associated gas well was derived
as the weighted average total miles divided by total wells (0.67 ± 28%). The average mile
per well ratio was extrapolated by the nationally tracked number of non-associated gas
wells (276,000).11 The resulting estimate of national gathering pipeline miles associated
with gas wells is 184,000.
51
-------
For the gathering pipeline mileage associated with oil wells that market gas,
the same ratio of gathering miles per well was applied. However, it was assumed that only
half of the gathering pipeline mileage was attributed to the gas industry; the other half was
attributed to the oil industry. Therefore, the average ratio of pipeline miles to oil wells
marketing gas was estimated to be 0.33. This average ratio was extrapolated by the
estimated number of oil wells marketing gas in the United States (209.000).12 The resulting
estimate of gathering pipeline mileage associated with oil wells that market gas is 70,000.
The third category of gathering pipeline owned by transmission companies is
reported by the American Gas Association to be 86,200 miles.11 Utility-owned pipelines
were assumed to be included in the total production owned gathering pipeline miles and are
not included in the transmission company owned gathering line mileage.
The resulting total national gathering pipeline mileage from gas wells, oil
wells marketing gas, and transmission companies was estimated to be 340,200 miles. A
rigorous determination of the 90% confidence interval gave an error less than ±4%, which
was considered low based on the quality of the data used to generate the activity factor.
Thus, a 90% confidence interval of ± 10% was assumed based on engineering judgement.
Equivalent Leaks
Based on the analysis resulting in a leak frequency of 0.18 leaks per mile for
transmission-owned gathering lines employing a walking survey and 0.63 leaks per mile for
production-owned gathering lines employing a vegetation survey, the activity factor can be
calculated as follows:
[(86,200 x 0.18)] + [(340,200 - 86,200) x 0.63] = 174,779 equivalent leaks/year
The breakdown of total equivalent leaks by pipe material category is based
on the breakdown of pipe mileage reported in the 1991 DOT RSPA8 database for
52
-------
transmission-owned gathering lines. It was estimated that the breakdown of production-
owned gathering line mileage into material categories is equivalent to the transmission-
owned pipelines, with the exception of cast iron. It was assumed that no additional cast
iron gathering lines are in service. (The cast iron gathering line mileage reported in the
RSPA database for transmission-owned gathering lines accounts for the total in the United
States.)
The total number of estimated gathering line leaks was allocated on a
pipeline material category basis in the same proportion (adjusted for the fraction of mileage
in each material category) as in the distribution sector. The 90% confidence interval of the
estimated total leaks was calculated using a statistical program (@ RISK)10 to be ± 76%.
For cast iron gathering lines, the mileage is based on the RSPA database for transmission
and gathering lines.8 The 90% confidence interval of the cast iron mileage estimate is
assumed to be ± 10%. Table 8-9 summarizes the estimated average activity factor and the
90% confidence interval for gathering pipeline.
TABLE 8-9. SUMMARY OF ACTIVITY FACTORS FOR GATHERING
PIPELINES IN THE PRODUCTION SEGMENT
Pipe Material
Total Miles ;
Average Activity Factor,
(equivalent leaks)
90% Confidence Interval
of Activity Factor,1
(equivalent leaks)
Protected Steel
268,082
53,657
40,779
Unprotected Steel
41,400
114,655
87,138
Plastic
29,862
6,467
4,915
Cast Iron
856
856"
86"
a 90% confidence interval around the mean value (upper bound minus the mean).
b Miles.
53
-------
9.0 RESULTS AND CONCLUSIONS
9.1 Distribution Underground Pipeline Emissions
The overall methane emission estimate for underground mains and services in
the distribution sector extrapolated to a national average is presented in Table 9-1, which has
been adjusted for soil oxidation of methane. As shown, the mean methane emission factors
for each pipe use/material type category are combined with an estimate of the number of
equivalent leaks (or miles of main for cast iron) in the nation to predict the total emissions.
As shown in Table 9-1, the annual estimated methane emissions to the atmosphere from
underground distribution mains and services in the U.S. natural gas industry is 41.6 Bscf,
with an estimated 90% confidence interval of ± 27.1 Bscf (± 65%).
TABLE 9-1. SUMMARY OF METHANE EMISSIONS ESTIMATE FROM
UNDERGROUND DISTRIBUTION PIPELINES*
Pipe Use
Pipe Material
Average Methane
Em ission Factor,
(scffleak-yr)
Average Activity
Factor,
(equivalent teaks)
Methane
Emissions
Estimate,
(Bscf)
90%
Confidence
Interval of
Emissions
Estiraateb(B$cf)
Mains
Cast Iron
238,736c
55,288"
13.2
8.4
Unprotected Steel
51,802
174,657
9.1
11.1
Protected Steel
20,270
68,308
1.4
1.6
Plastic
99,845
49,226
4.9
13.9
Services
Unprotected Steel
20,204
458,476
9.3
17.5
Protected Steel
9,196
390,628
3.6
6.1
Plastic
2,3S6
68,903
0.2
0.4
Copper
7,684
7,720
0.1
0.1
Total
41.6
27.1
* Adjusted for soil oxidation of methane.
b 90% confidence interval around the mean value (upper bound minus the mean).
c Scf/mile-year.
d Miles.
As shown, the contribution to the total emissions from cast iron mains (over
30%) is higher than any other single category, even though the total mileage of cast iron
54
-------
mains represents the smallest material category. Unprotected steel services and mains
represent the second and third largest contributor to emissions, respectively, at around 22%
of the total emissions for each. All other categories combined represent only 25% of the
total emissions. Therefore, even though the emission factor for plastic mains is relatively
high, the overall contribution to total emissions is relatively small (around 12% of the
total).
Table 9-2 presents the total estimate of methane leakage from distribution
mains and services, which has not been adjusted for the soil oxidation of methane. (Note:
The methane leakage rate has been adjusted for an average 93.4 volume percent methane in
natural gas.) The total estimated methane leakage, 51.1 Bscf, is about 9.5 Bscf higher than
the total annual methane emissions to the atmosphere. The major difference between the
estimated methane leakage rate shown in Table 9-2 and the methane emissions rate shown
in Table 9-1 is the soil oxidation of methane from leaks in cast iron mains. The methane
leakage rate from cast iron mains is 22.1 Bscf, while the methane emissions rate is 13.2
Bscf. The reason that the emissions rate is significantly lower than the leakage rate is that
cast iron mains have a lower leakage rate per individual leak than other pipe materials,
which results in a higher soil oxidation rate.
9.2 Transmission Underground Pipeline Emissions
Table 9-3 summarizes the estimated methane emissions from transmission
pipeline leaks in the U.S. natural gas industry. As shown, the annual methane emissions to
the atmosphere are 0.16 Bscf with a 90% confidence interval of ± 0.14 Bscf (± 89%).
As shown in Table 9-3, the largest contributor to the overall emissions
estimate from transmission underground pipelines is protected steel, representing about 67%
of the total leakage. This is expected since the total mileage of transmission pipeline is
made up of around 97% protected steel pipe, according to nationally tracked statistics.
55
-------
TABLE 9-2. SUMMARY OF METHANE LEAKAGE ESTIMATE FROM
UNDERGROUND DISTRIBUTION PIPELINESa
Pipe
Use
Pipe .
Material
Average
Methane
Leakage
Rate,
(scf/Ieak-yr)
Average
Activity
Factor,
(equivalent
leaks)
Methane
Leakage
Estimate,
(Bsef)
Confidence
Interval of
Leakage
Estimate,6
(Bscf)
Mains
Cast Iron
399,867°
55,288d
22.1
12.6
Unprotected
Steel
52,748
174,657
9.2
10.7
Protected
Steel
20,891
68,308
1.4
1.6
Plastic
101,897
49,226
5.0
13.7
Services
Unprotected
Steel
20,433
458,476
9.4
17.1
Protected
Steel
9,438
390,628
3.7
6.0
Plastic
3,026
68,903
0.2
0.5
Copper
7,684
7,720
0.1
0.1
Total
51.1
28.1
a Methane leakage rate not adjusted for soil oxidation. The leakage rate for methane has
been adjusted for an average 93.4 volume percent methane in natural gas.
b 90% confidence interval around the mean value (upper bound minus the mean).
c Scf/mile-year.
d Miles.
56
-------
TABLE 9-3. SUMMARY OF METHANE EMISSIONS ESTIMATE FROM
UNDERGROUND TRANSMISSION PIPELINES'
Pipe Material
Average Emission
Factor
(scf/leak-yr)
Average Activity
Factor
(equivalent leaks)
Methane
Emissions
Estimate (Bscf)
90% Confidence
Interval of
Emissions
Estimate1* (Bscf)
Protected Steel
20,270
5,077
0.10
0.14
Unprotected Steel
51,802
659
0.03
0.05
Plastic
99,845
14
0.001
0.003
Cast Iron
238,736c
96"
0.02
0.02
Total
0.16
0.14
1 Adjusted for soil oxidation of methane.
b 90% confidence interval and the mean value (upper bound minus the mean).
c Scf/mile-year.
d Miles.
9.3 Production Underground Pipeline Emissions
Table 9-4 summarizes the estimated methane emissions from gathering pipeline
leaks in the production segment of the gas industry. As shown, the annual methane
emissions to the atmosphere are 6.6 Bscf with a 90% confidence interval of ± 7.2 Bscf
(± 108%).
TABLE 9-4. SUMMARY OF METHANE EMISSIONS ESTIMATE FROM
UNDERGROUND PRODUCTION PIPELINES"
Pipe Material
Average Emission
Factor,
(scf/leak-yr)
Average Activity
Factor,
(equivalent leaks)
Methane
Emissions
Estimate,
(Bscf)
90%
Confidence
Interval of
Emissions
Estimate,1'
(Bscf)
Protected Steel
17,102
53,657
0.9
1.2
Unprotected Steel
43,705
114,655
5.0
7.0
Plastic
84,237
6,467
0.6
1.2
Cast Iron
201,418C
856d
0.2
0.1
Total
6.6
7.2
* Adjusted for soil oxidation of methane.
b 90% confidence interval around the mean value (upper bound minus the mean).
c ScC'mile-year.
d Miles.
57
-------
Although the majority of gathering pipeline was estimated to be protected steel
(around 19% of the total mileage), the largest contributor to total emissions from gathering
lines is unprotected steel pipelines. Not only is the total number of equivalent leaks from
unprotected steel gathering lines greater than from protected steel lines, but the emission
factor for unprotected steel gathering lines was estimated to be significantly higher than for
protected steel lines.
58
-------
10.0 REFERENCES
1. Pacific Gas & Electric, Unaccounted-for Gas Project, GRI-90/0067.1, June 7,
1990.
2. Southern California Gas Company, A Study of the 1991 Unaccounted-for Gas
Volume of the Southern California Gas Company, prepared for Gas Research
Institute, GRI-93/0115, April 1993.
3. Radian Corporation, Program Plan for the Cooperative Leak Test Program,
prepared for Gas Research Institute, March 1993.
4. Washington State University, University of New Hampshire, and Aerodyne
Research, Inc., Soil Consumption of Methane from Natural Gas Pipeline
Leaks, Environmental Science & Technology. December 1994.
5. Code of Regulations. Title 40, Part 60, Appendix A, Method 5, Section 7.2.
National Archives and Records Administration, Office of the Federal Register,
July 1, 1995.
6. SAS Institute Inc., SAS® Procedures Guide, Version 6, Third Edition, Cary,
NC: SAS Institute Inc., 1990.
7. SAS Institute Inc., SAS/STAT® User's Guide, Version 6, Fourth Edition,
Volume 2, Cary, NC: SAS Institute Inc., 1989.
8. U.S. Department of Transportation. Research and Special Programs
Administration, 1991.
9. Southern Cross Corporation. Comments on Docket PS-12S Notice 1, Leakage
Surveys, 49 CFR Part 192, Department of Transportation, Research and
Special Programs Administration, Materials Transportation Bureau, Office of
Pipeline Safety Regulations, December 19, 1991.
10. Palisade Corporation, @Risk, Risk Analysis and Simulation Add-in for Lotus 1-
2-3. Version 1.5, March 1989.
11. American Gas Association, Gas Facts, 1992.
12. Stapper, B.E. Methane Emissions from the Natural Gas Industry, Volume 5:
Activity Factors, Final Report, GRI-94/0257.22 and EPA-600/R-96-080e. Gas
Research Institute and U.S. Environmental Protection Agency, June 1996.
59
-------
APPENDIX A
Results of Outlier Tests for Plastic Pipe Leakage Data
A-l
-------
APPENDIX A
RESULTS OF OUTLIER TESTS FOR
PLASTIC PIPE LEAKAGE DATA
Overview
The GRI gas data for plastic pipes were screened for potential outliers. The
Grubbs test (Grubbs, 1969), the Dixon test (Grubbs, 1969), the Fourth-Spread test
(Hoaglin et al.. 1983), and a conservative approach (NSI, 1989) were used to identify
potential outliers in the plastic pipe data. The Grubbs and Dixon tests require that the data
being screened are normally distributed. The Fourth-Spread test does not strictly require
normality, but it could produce spurious results if the data distribution were markedly
asymmetric. The conservative approach addresses cases of normality and non-normality.
The largest value and the smallest value in the plastic pipe dataset were tested
separately. Table A-l lists the results of the four outlier tests for both the largest and
smallest plastic pipe data values. The smallest value is identified as a potential outlier only
in the Fourth-Spread test; all other tests indicate no outliers. However, the test criteria from
both the Grubbs and Dixon tests suggest that the smallest value is closer to being a
potential outlier than the largest value.
Data
The plastic pipe flow rate data and the natural logarithms of these data, as
well as the means and standard deviations, are shown in Table A-2. The data in Table A-2
are arranged so that the smallest value appears in the first row and the largest value appears
in the last row of the table. Only six data points comprise the plastic pipe data and these
six points span five orders of magnitude, ranging from 0.008 SCF/leak-hour to 61.000
SCF/leak-hour.
The Shapiro-Wilk W statistic, generated by the SAS UNIVARIATE (SAS,
1990) procedure, was used to determine whether the nontransformed and natural
A-2
-------
TABLE A-l. RESULTS OF THE OUTLIER TESTS
Outlier Test
Data Value Tested
Criteria3
Result
(natural logarithm)
Grubbs
Minimum:
-4.8283 (ID 2014)
1.7K1.82
not an outlier
Maximum:
4.1109 (ID 2002)
1.26<1.82
not an outlier
Dixon
Minimum:
-4.8283 (ID 2014)
0.50<0.56
not an outlier
Maximum:
4.1109 (ID 2002)
0.20<0.56
not an outlier
Minimum:
outside bounds:
OUTLIER
F-Spread
-4.8283 (ID 2014)
-4.3850 to 6.3571
Maximum:
inside bounds:
not an outlier
4.1109 (ID 2002)
-4.3850 to 6.3571
Conservative
Minimum:
-4.8283 (ID 2014)
inside bounds:
-8.7334 to 9.3532
not an outlier
Approach
Maximum:
inside bounds:
not an outlier
4.1109 (ID 2002)
-8.7334 to 9.3532
a The criteria are based on the 5% significance level for the Grubbs and Dixon tests
A-3
-------
TABLE A-2. PLASTIC PIPE FLOW RATE DATA AND NATURAL
LOGARITHMS OF THE FLOW RATES
Test ID Number
Standard Flow Rate
(SCF/leak-hour)
Natural Log of Standard
Flow Rate
2014
0.008
-4.8283
3020
0.700
-0.3567
3019
1.130
0.1222
3039
1.620
0.4824
11002
10.266
2.3288
2002
61.000
4.1109
Mean
12.454
0.309894
Standard Deviation
24.084
3.014434
log-transformed plastic pipe data were normally distributed. For the nontransformed data,
the W-statistic was 0.6068 and the associated p-value was 0.0001, indicating that the
nontransformed data were not normally distributed. However, for the natural
log-transformed data, the W-statistic was 0.9396 and the associated p-value was 0.6747,
indicating that the natural log-transformed data were normally distributed, within random
variability. Because of the small sample size (6), however, this test is not highly sensitive.
Small or moderate deviations from normality might not be detected based on a hypothesis
test with this sample size. Figure A-l Shows the frequency histogram for the
nontransformed data and Figure A-2 shows the frequency histogram for the natural log-
transformed data to illustrate the results suggested by the W-statistics. The nontransformed
data are obviously skewed and not normally distributed, while the natural log-transformed
data are much more symmetric and appear to be closer to the normal distribution.
A-4
-------
0 5
M'rfpoint "for Ftcw Rat^s CSCF/;
-------
Many different tests exist to screen for outliers, some of which have certain
limitations that prevent them from being applied to all datasets. Some tests require that the
data be distributed normally because statistical parameters are used in the outlier test, while
other tests rely on other types of information from the data to perform the outlier test.
Because of the variety and number of different outlier tests, it is therefore important that no
datum be discarded solely on the basis of a single statistical test. There should always be
some plausible explanation other than the test result that warrants the exclusion or the
replacement of an outlier (Gilbert, 1987). If possible, several different types of tests should
be applied to validate the results of the outlier screening process.
The four different tests applied to the GRI plastic pipe data represent some of
the different types of outlier tests. The Grubbs test (Grubbs, 1969) relies on statistical
parameters (mean and standard deviation), the Dixon test relies on ratios of values in the
-4 -2 0 2 A
Midooirt for Nature* Loos of F ow Rate
Figure A-2. Frequency Histogram for the Natural Logarithms of the Plastic
Pipe Flow Rate Data
A-6
-------
tails, the Fourth-Spread test relies on the spread of the central half of the data, and the
conservative approach is capable of handling any data distribution. Following are specific
details regarding how each of these tests were applied to the plastic pipe data.
Grubbs Test (Grubbs. 1969)
The hypothesis tested in the Grubbs test is that all observations in the sample
come from the same normal population. Thus, transformation of skewed data, such as
taking the natural logarithms, may be necessary. The data are ordered from smallest to
largest for the Grubbs test, such that:
The Grubbs test is then applied to a single suspect value-either the largest value (XJ or the
smallest value (X,). For the largest value (Xn), the test statistic (Tn) is calculated as
follows:
{X, < X2 < X3 < ... < XJ
(1)
X„-X
T,
(2)
s
where:
Xn = the largest data value.
X = the arithmetic average of all n values, and
s = the sample standard deviation, with n — J degrees of freedom.
For the smallest value (X,), the test statistic (Tj) is calculated as follows:
X-X,
T,«
(3)
where:
the smallest data value, and
the same as for Equation (2).
A-7
-------
The test statistic (T, or Tn) is compared to the appropriate critical value for the statistic.
When the test statistic is larger than the critical value, then the suspect data point is deemed
a potential outlier.
Using the mean and standard deviation shown in Table A-2 for the plastic
pipe data, T,=1.71 and Tn=1.26 for the natural logarithms of the flow rates. The critical
value for a one-sided test using the 5% significance level for a sample size of six is 1.82,
and the critical value using the recommended 1% significance level (Grubbs, 1969) is 1.94.
Therefore, neither the largest nor the smallest of the natural logarithms of the plastic pipe
flow rates were considered outliers by the Grubbs test.
Dixon Test ("Grubbs, 1969)
The Dixon test is an alternative system that does not rely on the calculation
of statistical parameters (e.g., the mean or standard deviation), and is based entirely on
ratios of differences between some of the observations. As with the Grubbs test, the Dixon
test requires a normal data distribution because the ratios of differences are calculated from
both tails. One drawback to the Dixon test is that not all of the data are utilized-only data
from the tails are used. Similarly to the Grubbs test, the data are ordered from smallest to
largest for the Dixon test, as shown in Equation (1). The Dixon test is then applied to a
single suspect value, either the largest or smallest of all of the data values. A test statistic
(r^) that depends on sample size is calculated. The formula for the largest value (XJ from
a sample size of 6 (the plastic pipe data sample size) is:
Xn-X,.,
rio = (4)
Xn-X,
where:
the
largest data value,
the
second largest data value, and
x,=
the
smallest data value.
A-8
-------
The corresponding formula for the smallest value (X,) from a sample size of 6 (the plastic
pipe data sample size) is:
x2-x,
xn-x,
where:
X2 = the second smallest data value,
X, = the smallest data value, and
Xn = the largest data value.
(5)
The test statistic (r10) is compared to the appropriate critical value for the statistic. When
the test statistic is larger than the critical value, then the suspect data point is deemed a
potential outlier.
Using the data shown in Table A-2 for the plastic pipe data, r10=0.20 for the
largest value and rlo=0.50 for the smallest value of the natural logarithms of the flow rates.
The critical value for a one-sided test using the 5% significance level for a sample size of
six is 0.560, and the critical value using the recommended 1% significance level (Grubbs,
1969) is 0.698. Therefore, neither the largest nor the smallest of the natural logarithms of
the plastic pipe flow rates were considered outliers by the Dixon test.
Fourth-Spread Test (Hoaalin et al.. 1983)
The Fourth-Spread (F-Spread) test does not rely on the calculation of the
mean or standard deviation, rather it relies on information from the center half of the data
mass to define the distance, beyond which, data points should be considered potential
outliers. The center half of the distribution is relatively insensitive to outliers and,
therefore, provides a reasonable basis for characterizing the distribution under the
hypothesis that no outliers are present. As with the Grubbs and Dixon tests, the data must
be arranged from smallest to largest, as shown in Equation (1). The data need not be
normally distributed, but the distribution should be symmetric. First, the lower and upper
fourths (also called the 25th and 75th percentiles, respectively) for the data distribution are
A-9
-------
calculated. The F-Spread (df) is then calculated by subtracting the lower-fourth (FL) from
the upper-fourth (F,j). Any data points larger than Fl--(1.5xJf) and any data points smaller
than FL—(1.5x
Fl- (1. 5xdF) Fl F-j F0+ (1. 5xdF)
25th 75th
percentiles
Figure A-3. Depiction of the Fourths (FL and Ftl), Fourth-Spread (dT), and
Boundaries (FL-1.5xrfF; FL,+1.5xrfF) for the F-Spread Outlier Detection Method
The F-Spread for the plastic pipe flow rate data was 2.6E55 (F1;=2.3288 and
FL=—0.3567). Thus, data values smaller than —4.3850 or larger than 6.3571 should be
considered potential outliers. One of the six plastic pipe data points, the smallest (ID=2014,
value=0.008 SCF/leak-hour, In value=—4.8283), was therefore considered a potential
outlier.
Conservative Approach fNSI. 1989)
This approach is conservative because it screens for only the most blatant
outliers. Thus, data points that may be considered outliers in other methods, may not be
considered outliers by this approach, unless they are separated by a rather large distance
from the main data mass. The histogram for the nontransformed data and the histogram
from the natural log-transform of the data are used as visual aids in this method. Some
A-10
-------
measure of the normality (e.g., the Shapiro-Wilk W-statistic) for the data distributions
shown in the two histograms is also used in this method. The following four steps are
applied in sequence until the conditions are met and the criteria are defined for identifying
outliers:
(1) The untransformed data distribution is normal. Values more than 3 standard
deviations from the mean (mean±3xstandard deviation) are considered
potential outliers.
(2) The natural log-transformed data distribution is normal. Values more than 3
standard deviations from the mean of the natural logarithms
(mean±3xstandard deviation) are considered potential outliers.
(3) The untransformed data distribution is visually symmetric, but not normal.
Values more than 3 standard deviations from the mean (mean±3xstandard
deviation) are considered potential outliers.
(4) The untransformed data distribution is not normal and not visually symmetric.
Values more than 6 standard deviations from the mean (mean±6xstandard
deviation) are considered potential outliers.
For the plastic pipe data, this method produced the following results for the
first two steps (at which point the conditions were met and outlier criteria were
established):
(1) The untransformed data distribution is not normal. Go to step 2.
(2) The natural log-transformed data distribution is normal. Therefore, values
more than 3 standard deviations from the mean are considered potential
outliers. Thus, using the mean and standard deviation shown in Table 2 for
the natural logarithms of the flow rates, values more than 9.3532 or less than
— 8.7334 should be considered potential outliers. None of the data points met
these criteria and therefore there were no outliers.
References
Gilbert, Richard O., 1987: Statistical Methods for Environmental Pollution Monitoring.
Van Nostrand Reinhold, New York, 320 pp.
Grubbs, Frank E., 1969: Procedures for detecting outlying observations in samples.
Technometrics, 11. No. 1, 1—21.
A-ll
-------
Hoaglin, D.C., F. Mosteller, and J.W. Tukey, 1983: Understanding Robust and Exploratory
Data Analysis. John Wiley & Sons, Inc., New York, 447 pp.
Hunt, W.F., Jr.. G. Akland, W. Cox, T. Curran. N. Frank, S. Goranson, P. Ross, H. Sauls,
and J. Suggs, 1981: U.S. Environmental Protection Agency Intra-Agency Task Force on
Air Quality Indicators, EPA-450/4-81-015 (NTIS PB81-177982). Office of Air Quality
Planning and Standards, Research Triangle Park, NC.
NSI Technology Services Corporation, 1989: WHO and WMO Program Documentation.
Prepared for Environ. Monitoring Systems Laboratory, U.S. EPA, Research Triangle Park,
NC, under contract No. 68-02-4444. NSI Publication Number SP-4420-89-28.
SAS Institute, Inc., 1990: SAS Procedures Guide, Version 6, Third Edition. Carv, NC:
SAS Institute, Inc., 705 pp.
A-12
-------
APPENDIX B
Source Sheets for Underground Pipeline Leakage
B-l
-------
D-2
DISTRIBUTION SEGMENT SOURCE SHEET
SOURCES:
OPERATING MODE:
EMISSION TYPE:
ANNUAL EMISSIONS:
Main and Service Pipeline
Normal Operations
Steady, Fugitives (Leakage)
41.6 Bscf +/- 65%
BACKGROUND:
Distribution mains are the pipelines that serve as a common source of natural gas supply for more than one
customer. Services are the branch connection lines from the mains to the customer meters. Leakage from the
underground distribution network occurs from corrosion pits, joint and fitting failures, and pipe wall fractures.
Gas distribution operators use leak detection procedures to locate and classify leaks. The leak is classified
and prioritized for repair based on the concentration of gas detected and the proximity of the leak to existing
structures.
EMISSION FACTOR: (scf/leak-year)
The value of the emission factor and standard deviation for each pipe material category is given below:
Material
Category
Pipe Use
Number
of
Samples
Average
Emission
Factor1
Units of
Emission
Factor
90%
Confidence
Interval of
Emission
Factor
Cast Iron
Main
21
238,736
scf/mi-yr
152,059
Unprotected
Steel
Main
20
51,802
scf/lk-yr
48,212
Protected Steel
Main
17
20,270
scf/lk-yr
17,243
Plastic
Main
6
99,845
scf/lk-yr
165,617
Unprotected
Steel
Service
13
20,204
scflk-yr
21,129
Protected Steel
Service
24
9,196
scf/lk-yr
5,581
Plastic
Service
4
2,386
scf/lk-yr
3,412
Cooper
Service
5
7,684
scf/lk-yr
5,559
2 Adjusted for the soil oxidation of methane.
A cooperative leak measurement program has been de%'eloped to measure a representative sample of
underground leaks to estimate the average leak intensity, which is combined with company leak records to
estimate leak frequency. Leak measurements were performed at five U.S. companies and two Canadian
distribution companies in accordance with the testing protocol developed as part of the program. The test
data were disaggregated by material type and mains versus services, based on a combination of statistical
analyses and engineering judgement. A mean value of the estimated leak rate per leak was calculated using
the test data, for all pipe materials except cast iron. In these tests, an individual leak was randomly selected
B-2
-------
for testing based on criteria outlined in the program plan. For cast iron, long segments of pipe were tested to
measure the leak rate per mile rather than the leak rate per leak. Cast iron was tested in long segments since
it tends to have a very high frequency of leaks (due to the joint spacing of 10 to 16 feet) and the relatively
high occurrence of undetectable leaks in cast iron. The measured natural gas leak rates were adjusted for the
average volume percent of methane in pipeline-quality gas (93.4 vol. %), and the soil oxidation rates of
methane.
ACTIVITY FACTOR:
The mean activity factor and standard deviation for each pipe material category is given below:
Material
Category
Pipe Use
Estimated
Total Leak
Repairs
Average
Activity
Factor
(Equivalent
Leaks)
Units of Activity
Factor
90%
Confidence
Interval of
Activity Factor
Cast Iron
Main
69,776
55,288
miles
2,764
Unprotected
Steel
Main
81,627
174,657
equivalent leaks
101,685
Protected Steel
Main
31,924
68,308
equivalent leaks
42,545
Plastic
Main
23,006
49,226
equivalent leaks
58,018
Unprotected
Steel
Service
214,271
458,476
equivalent leaks
499,850
Protected Steel
Service
182,562
390,628
equivalent leaks
526,354
Plastic
Service
32,202
68,903
equivalent leaks
66,840
Copper
Service
3,608
7,720
equivalent leaks
8,521
The national database of leak repairs was used to extrapolate data provided by individual companies. Data
were requested from each company participating in the underground leak test program, based on their
historical leak records. To allocate leak repairs into pipe material categories, data were collected from ten
local distribution companies representing different regions within North America.
Data on the total number of annual leak repairs, leak indications, and outstanding leaks within the distribution
system were provided by six companies. An estimate of the number of annual equivalent leaks for each of
the six companies was developed based on the following methodology:
Total Equivalent Leaks = Outstanding Leaks + New Leaks - Leak Repairs
The total number of annual equivalent leaks represents the equivalent leaks which are leaking all year. (That
is, for leaks with a leak duration of half year, these leaks are counted as half an equivalent annual leak.)
The total number of leaks in the system are quantified by incorporating the leak duration into the estimated
equivalent leaks. For example, if a leak is only leaking half the year, it is counted as 0.5 equivalent leaks.
The assumptions made in deriving the estimated number of equivalent leaks for each company include:
B-3
-------
Approximately 85 percent of leaks are found during a leak survey when an organic
vapor analyzer (OVA) instrument is used along with bar holing.
Leaks that are repaired during the year are leaking half of the year, on average.
Outstanding leaks are leaking at the beginning of the year.
The number of new leaks in the system is estimated based on the annual leak
indications and the frequency of the leak survey.
The number of new leaks in a system that is surveyed every n years is calculated based on the following:
For the first year in the cycle — l/n leaks are leaking half the yean (n-l)/n leaks are
not yet leaking.
For the second year in the cycle - l/n leaks are leaking the entire year; l/n leaks
are leaking half the year; and (n-2)/n leaks are not yet leaking.
For the third year in the cycle ~ 2/n leaks are leaking the entire year; l/n leaks are
leaking half the year; and (n-3)/n leaks are not yet leaking.
For the fourth year in the cycle - 3/n leaks are leaking the entire year; l/n leaks are
leaking half the year; and (n-4)/n leaks are not yet leaking.
Based on the data provided by each of the six companies, a ratio of the annual equivalent leaks to leak repairs
was calculated. The average ratio (2.14) was multiplied by the estimated number of leak repairs in each pipe
material category to extrapolate the national database of leak repairs to represent annual equivalent leaks. The
precision of the estimate is based on the variability in the leak repair disaggregation provided by ten
companies and the variability in the calculated ratio of annual equivalent leaks to leak repairs provided by six
companies.
The activity factor for cast iron mains is the total estimated mileage of cast iron mains in the U.S., as reported
by the U.S.DOT RSPA.1 The standard deviation was assumed to be 5% of the estimated mileage, based on
engineering judgement.
B-4
-------
EMISSIONS ESTIMATE: (41.6 +/- 65 %)
The emissions estimate for each category of pipe material/use was dervied multiplying the respective emission
factor (scf'leak-yr or scCmile-yr) by the activity factor (total number of leaks or miles).
Material
Category
Pipe
Use
Average
Emission
Factor
(scf/lk-yr)
Average
Activity
Factor
(equivalent
leaks)
Annual
Emissions
Estimate
(Bscf)
90% Confidence
Interval of
Emission Estimate
(Bscf)
Cast Iron
Main
238,736"
55,288"
13.2
8.4
Unprotected Steel
Main
51,802
174,657
9.1
11.1
Protected Steel
Main
20,270
68,308
1.4
1.6
Plastic
Main
99,845
49,226
4.9
13.9
Unprotected Steel
Service
20,204
458,476
9.3
17.5
Protected Steel
Service
9,196
390,628
3.6
6.1
Plastic
Service
2,386
68,903
0.2
0.4
Copper
Service
7,684
7,720
0.1
0.1
Total
41.6
27.1
ascf7mile-yr
bmiles
REFERENCES
1. U.S. Department of Transportation. Research and Special Programs Administration. 1991.
B-5
-------
T-3
TRANSMISSION SOURCE SHEET
SOURCES: Transmission Pipelines
OPERATING MODE: Normal Operations
EMISSION TYPE: Unsteady, Fugitives (Pipeline Leaks)
ANNUAL EMISSIONS: 0.16 Bscf +/- 89%
BACKGROUND:
Transmission pipelines are the inter- and intrastate high pressure underground pipelines that transport natural
gas from the production/processing operations to the end user or distribution network. Leakage from
underground transmission lines occurs from corrosion pits, joint and fitting failures, pipe wall fractures, and
external damage.
9
EMISSION FACTOR: (scf/leak-year)
Leak survey practices for transmission lines are generally more stringent than for distribution mains.
Transmission lines are required to be surveyed annually, and more frequently in populated areas. In addition,
many transmission companies perform additional routine aerial surveys to monitor the transmission lines for
leakage. Based on conversations with several transmission companies, any leaks found in the pipewall are
extremely small and are repaired immediately for safety reasons. Based on the rigorous leak survey and
repair practices of transmission companies (i.e., leaks are discovered and repaired earlier in transmission
lines), the average leak rate from a transmission leak is believed to be of the same order of magnitude as a
leak found in a distribution main, even though there may be a substantial difference in the operating pressure
of the pipelines.
Therefore, the emission factors for leakage from transmission pipelines are based on the arithmetic average
leakage rates for main pipelines from the cooperative underground distribution leakage measurement program.
A mean value of the estimated leak rate per leak was calculated using the test data, for all pipe materials
except cast iron. For cast iron mains, a segment test approach was used which quantifies the leakage rate for
a long isolated segment of pipe; therefore, the mean leakage rate for cast iron is in terms of leakage per unit
length of pipe. The natural gas leak rate is adjusted for methane by multiplying by the volume percent of
methane for transmission (93.4 vol. %), and is adjusted for the soil oxidation of methane. The value of the
emission factor and standard deviation for each pipe material category is given below:
Pipe Material
Number of
Samples
Average
Emission
Factor
Units of
Emission
Factor
90%
Confidence
Interval of
Emission
Factor
Protected Steel
17
20,270
scf'leak-vr
17,243
Unprotected Steel
19
51,802
scf/leak-yr
48,212
Plastic
6
99,845
scf71eak-yr
165,617
Cast Iron
21
238,736
scf/mile-yr
152,059
Preliminary data from the underground distribution program indicate that the leakage rate is not a function of
the pipeline pressure. Therefore, the leakage rates for transmission pipelines have not been adjusted based on
the difference in average operating pressure of the transmission lines versus distribution lines.
B-6
-------
EMISSION FACTOR DATA SOURCES:
1. Leakage rate data on a rate per leak basis for cathodically protected steel mains,
unprotected steel mains, and plastic mains from the cooperative leak measurement
program.
2. Leakage rate data on a rate per unit length basis for cast iron mains from the
cooperative leak measurement program for distribution mains.
3. Assumes that the leak rates from transmission pipelines are identical to leak rates
from distribution mains, based on the more rigorous leak survey and repair practices
of transmission companies.
4. Assumes that the leak rates from underground pipelines are independent of pressure
and pipe diameter, based on preliminary results from the underground distribution
leak measurement program.
ACTIVITY FACTOR: (equivalent leaks)
The mean activity factor and precision for each pipe material category is given below:
Pipe Material
Total Miles
Average Activity
Factor
Units of Activity
Factor
90% Confidence
Interval of
Activity Factor
Protected Steel
287,155
5,077
equi%ralent leaks
3.859
Unprotected Steel
5,233
659
equivalent leaks
501
Plastic
2,621
14
equivalent leaks
11
Cast Iron
96
96
miles
10
The number of total leaks (excluding pipeline incidents) in transmission pipelines is based on the 1991 DOT
RSPA database1 for transmission pipelines, including both repaired leaks (6,120 leaks) and outstanding leaks
(1,369 leaks). Because transmission lines are surveyed at least once per year using a walking survey method,
the number of unreported leaks is estimated based on the effectiveness of the walking survey. According to
one contract company specializing in distribution surveys, roughly 85 percent of the leaks are found using a
walking survey. This estimated survey efficiency was applied to transmission surveys, resulting in roughly
1,320 unreported leaks.
The leak duration for outstanding leaks and unreported leaks is estimated to be 8,760 hours per year, and the
leak duration for repaired leaks is half a year (4,380 hours/year), on average. The resulting estimate of
equivalent leaks represents the number of leaks with a year round leak duration. (That is, each leak repair is
counted as half an equivalent leak to compensate for the leak duration.) Therefore, the equation used to
estimate equivalent leaks is:
0.5 x (repaired leaks) + {[(repaired leaks + outstanding leaks)/0.85] - repaired leaks}
The total number of estimated transmission pipeline leaks, 5,750, was allocated on a pipeline material
category basis in the same proportion (adjusted for the fraction of mileage in each material category) as in the
distribution sector. (That is, the ratio of percent leaks to percent miles in the transmission segment is the
same as the ratio in the distribution segment.) The precision of the estimated total leaks was calculated based
on the estimated 90% confidence interval associated with each parameter in the activity factor equation:
repaired leaks; outstanding leaks: ± 100%
B-7
-------
leak duration: ± 25%
leak survey effectiveness: ± 15%
A statistical software program (@RISK2) was used to determine the overall 90% confidence interval of the
activity factor: + 76%.
For cast iron transmission lines, the mileage is based on the 1991 DOT RSPA database for transmission and
gathering lines. The precision of the estimate is assumed to be ± 10%.
ACTIVITY FACTOR DATA SOURCES:
1.
2.
J.
4.
5.
1991 DOT RSPA database' for transmission and gathering pipelines.
Total number of leaks is assumed equal to the total number of leak repairs plus the
outstanding (unrepaired leaks) and unreported leaks.
Leak survey effectiveness estimation provided by Southern Cross Company.3
The allocation of estimated leaks per pipe material categoiy is based on the leak
frequency for underground distribution main pipelines, adjusted for the fraction of
total mileage per pipe material category.
@RISK statistical software program2 used to estimate the 90% confidence interval.
ANNUAL EMISSIONS: (0.16 Bscf ± 89%)
Pipe Material
Average Emission
Factor
(scfrleak-yr)
Average Activity
Factor
(equivalent leaks)
Annual Emissions
Estimate,
(Bscf)
90% Confidence
Interval of
Emissions
Estimate,
(Bscf)
Protected Steel
20,270
5,077
0.10
0.14
Unprotected Steel
51,802
659
0.03
0.05
Plastic
99,845
14
0.001
0.003
Cast Iron
238,736"
96b
0.02
0.02
Total
0.16
0.14
scf/mile-yr
bmiles
The total leakage was determined by multiplying an emission factor for each type of pipeline material by the
estimated number of leaks in each respective pipe material category.
REFERENCES
1. U.S. Department of Transportation. Research and Special Programs Administration. 1991.
2. Palisade Corporation. @ Risk, Risk Analysis and Simulation Add-in for Lotus 1-2-3, Version 1.5, March
1989.
B-8
-------
3. Southern Cross Corporation. Comments on Docket PS-123 Notice 1, Leakage Surveys, 49 CFR Part 192,
Department of Transportation, Research and Special Programs Administration, Materials Transportation
Bureau, Office of Pipeline Safety Regulations, December 19, 1991.
B-9
-------
P-3
PRODUCTION SEGMENT SOURCE SHEET
SOURCES:
OPERATING MODE:
EMISSION TYPE:
ANNUAL EMISSIONS:
Gathering Pipelines
Normal Operations
Steady, Fugitives (Pipeline Leaks)
6.6 Bscf ± 108%
BACKGROUND:
Gathering field pipelines transport the gas from the production well to gas conditioning or processing
facilities. Leakage from gathering pipelines occurs from corrosion, joint and fitting failures, pipe wall
fractures, and external damage.
EMISSION FACTOR: (scf/leak-year)
The emission factors for leakage from gathering pipelines are based on the arithmetic average leakage rates
for main pipelines from the cooperative underground distribution leakage measurement program. A mean
value of the estimated leak rate per leak was calculated using the test data for all pipe materials except cast
iron. For cast iron mains, a segment test approach was used which quantifies the leakage rate for a long
isolated segment of pipe; therefore, the mean leakage rate for cast iron is in terms of leakage per unit length
of pipe. The natural gas leak rate is adjusted for methane by multiplying by the volume percent of methane
for production (78.8 vol. %), and is adjusted for the soil oxidation of methane. The value of the emission
factor and standard deviation for each pipe material category is given below:
Pipe Material
Number of
Samples
Average
Emission
Factor
Units of
Emission
Factor
90%
Confidence
Interval of
Emission
Factor
Protected Steel
17
17,102
scfi'leak-yr
14,548
Unprotected
Steel
20
43,705
scf'leak-yr
40,675
Plastic
6
84,237
scf71eak-yr
139,729
Cast Iron
21
201,418
scf/mile-yr
128,290
EMISSION FACTOR DATA SOURCES:
1. Leakage rate data on a rate per leak basis for cathodically protected steel mains,
unprotected steel mains, and plastic mains from the cooperative leak measurement
program.
2. Leakage rate data on a rate per unit length basis for cast iron mains from the
cooperative leak measurement program for distribution mains.
3. Assumes that the leak rates from gathering lines are identical to leak rates from
distribution mains.
B-10
-------
ACTIVITY FACTOR:
The estimated number of leaks in field gathering pipelines is based on a leak repair frequency for gathering
lines owned and operated by transmission companies reported in the 1991 DOT RSPA database.' This
database reports an estimated 8,153 repaired leaks and 270 outstanding leaks in 31,918 miles of gathering
pipeline. The leak frequency is derived by compensating for leaks that are repaired during the year and,
therefore, not contributing to leakage year round. On average, the repaired leaks are assumed to be leaking
for half the year, and each leak repair is counted as half an equivalent leak. Outstanding and unreported leaks
are assumed to be leaking the entire year.
Most production lines owned and operated by production companies are not regulated by DOT and many are
not monitored for leaks in the rigorous fashion employed by distribution and transmission companies.
Therefore, unreported leaks are accounted for based on the effectiveness of the survey method performed,
which is estimated to find 35% and 85% of the total leaks for a vegetation and walking survey, respectively,
based on one contract company specializing in distribution surveys. It is estimated that production company
owned gathering lines are only surveyed using a vegetation method. However, transmission company owned
gathering lines are estimated to be surveyed annually using a walking method, based on conversations with
several transmission companies.
Based on this analysis of equivalent leaks, the leak frequency is 0.18 leaks per mile for a walking survey and
0.63 leaks per mile for a vegetation survey. This leak frequency was used to ratio the number of leaks to the
total estimated population of gathering pipeline.
Total gathering pipeline mileage is not reported or tracked nationally and must be estimated. The "gathering
pipeline" designation includes three categories of pipeline: 1) production company owned gathering pipeline
for gas wells not associated with oil production (i.e., non-associated gas wells); 2) production company owned
gathering pipeline for oil wells that produce marketed gas (i.e., associated gas wells); and 3) transmission
company owned gathering pipeline. The third category of utiiity-owned pipelines are assumed to be in
addition to the production pipeline miles associated with wells. This is consistent with the site visit data
since gathering lines owned by transmission companies were intentionally excluded from the site mileage
totals. (The production companies did not report pipeline miles beyond their custody transfer meters.)
Total miles of gathering pipeline for non-associated gas wells were estimated using site visit data from the
thirteen production sites shown in the following table. Seven of the thirteen sites provided estimates of their
total miles of pipeline. The fifth site's mileage was estimated from a map of its pipelines.
B-ll
-------
Site
Gathering Miles
Number of Wells
Miles per Total
Wells
Site 1
46.3
80
Site 2
8
26
Site 3
40
130
Site 4
15.4
12
Site 5
11
6
Site 6
5.2
193'
Site 7
600
1000
Site 8
441.3
425
Site 9
0.7
1
Site 10
27.7
24
Site 11
2.1
3
Site 12
7.1
7
Site 13
154.2
126
TOTAL
1359.0
2033
0.67 +/- 28%
"Includes 55 oil wells.
The estimate of total gathering miles per non-associated gas well was derived as the weighted average total
miles divided by total wells (0.67 + 28%). The average mile per well ratio was extrapolated by the nationally
tracked number of non-associated gas wells (276,000). The resulting estimate of national gathering pipeline
miles associated with gas wells is 184,000.
For the gathering pipeline mileage associated with oil wells that market gas, the same ratio of gathering miles
per well was applied. However, it was assumed that only half of the gathering pipeline mileage was
attributed to the gas industry; the other half was attributed to the oil industry. Therefore, the average ratio of
pipeline miles to oil wells marketing gas was estimated to be 0.33. This average ratio was extrapolated by
the estimated number of oil wells marketing gas in the U.S. (209,000). The resulting estimate of gathering
pipeline mileage associated with oil wells that market gas is 70.000.
The third category of gathering pipeline owned by transmission companies is reported by the American Gas
Association (A.G.A.)2 to be 86,200 miles. Utility-owned pipelines were assumed to be included in the total
production owned gathering pipeline miles and are not included in the transmission company owned gathering
line mileage.
The resulting total national gathering pipeline mileage from gas wells, oil wells marketing gas, and
transmission companies was estimated to be 340.200 miles. A rigorous determination of the 90% confidence
interval gave an error less that 4%, which was considered low based on the quality of the data used to
generate the activity factor. Thus, a 90% confidence interval of ± 10% was assumed based on engineering
judgement.
B-12
-------
Based on the analysis resulting in a leak frequency of 0.18 leaks per mile for transmission-owned gathering
lines employing a walking survey, and 0.63 leaks per mile for production-owned gathering lines employing a
vegetation survey, the activity factor can be calculated as follows:
[(86,200 x 0.18)] + [(340,200 - 86,200) x 0.63] = 174,779 equivalent leaks/year
The breakdown of total equivalent leaks by pipe material category is based on the breakdown of pipe mileage
reported in the 1991 DOT RSPA database1 for transmission-owned gathering lines. It was estimated that
production-owned gathering line mileage is equivalent to the transmission-owned pipelines, with the exception
of cast iron. It was assumed that no additional cast iron gathering lines are in service. (That is, the cast iron
gathering line mileage reported in the RSPA database accounts for the total in the United States.)
The total number of estimated gathering line leaks was allocated on a pipeline material category basis in the
same proportion (adjusted for the fraction of mileage in each material category) as in the distribution sector.
The precision of the estimated total leaks was calculated based on the estimated 90% confidence interval
associated with each parameter in the activity factor equation:
repaired leaks; outstanding leaks: ± 100%
leak duration: ± 25%
leak survey effectiveness: ± 15%
A statistical software program (@ RISK3) was used to determine the overall 90% confidence interval of the
activity factor: ± 76%.
For cast iron gathering lines, the mileage is based on the 1991 DOT RSPA database for transmission and
gathering lines. The precision of the cast iron mileage estimate is assumed to be ± 10%. The following table
summarizes the estimated average activity factor and the precision:
Pipe Material
Total Miles
Average Activity
Factor
Units of Activity
Factor
90% Confidence
Interval of
Activity Factor
Protected Steel
268,082
53,657
equivalent leaks
40,779
Unprotected Steel
41,400
114,655
equivalent leaks
87,138
Plastic
29,862
6,467
equivalent leaks
4,915
Cast Iron
856
856
miles
86
ACTIVITY FACTOR DATA SOURCES:
1. Leak repair frequency from (DOT RSPA1) gathering line data.
2. Leak survey effectiveness provided by Southern Cross Company.4
3. The gathering miles for gas and oil wells marketing gas was estimated using Phase
3 site visit data for seven production companies. The number of gas and oil wells
for these companies was also used to extrapolate out to the national estimate.
4. The number of producing gas wells in the United States was taken from A.G.A. Gas
Facts1 for 1992.
5. The number of oil wells producing marketed gas in the United States was estimated
by Radian.' See activity factor section and sheet P-2.
B-13
-------
6.
The field and gathering miles owned by transmission companies was taken from
A.G.A. Gas Facts1 for 1992.
ANNUAL EMISSIONS: (6.6 Bscf ± 108%)
The activity factor was multiplied by the emission factor to derive this total leakage rate. The 90%
confidence intervals were propagated through this multiplication.
Pipe Material
Average Emission
Factor
(scf/leak-yr)
Average Activity
Factor
(equivalent leaks)
Annual Emissions
Estimate,
(Bscf)
90% Confidence
Interval of
Leakage Estimate
(Bscf)
Protected Steel
17,102
53,657
0.9
1.2
Unprotected Steel
43,705
114,655
5.0
7.0
Plastic
84,237
6,467
0.6
1.2
Cast Iron
201,418s
856b
0.2
0.1
Total
6.6
7.2
ascf;mile-yr.
bmiles.
REFERENCES
1. U.S. Department of Transportation, Research and Special Programs Administration. 1991.
2. American Gas Association, Gas Facts, 1992.
3. Palisade Corporation. @ Risk, Risk Analysis and Simulation Add-In for Lotus 1-2-3, Version 1.5, March
1989.
4. Southern Cross Corporation. Comments on Docket PS-123, Notice 1, Leakage Surveys, 49 CFR Part 192,
Department of Transportation, Research and Special Programs Administration, Materials Transportation
Bureau, Office of Pipeline Safety Regulations, December 19, 1991.
5. Stapper, B.E. Methane Emissions from the Natural Gas Industry, Volume 5, Activity Factors, Final
Report, GRI-94/0257.22 and EPA-600/R-96-Q80e. Gas Research Institute and U.S. Environmental
Protection Agency, June 1996.
B-14
------- |