Environmental Protection Research and Development
Agency Washington, DC 20460
EPA-600/R-95-167
November 1995
EPA Evaluation of the High
Volume Collection System
(HVCS) for Quantifying
Fugitive Organic Vapor Leaks
-------
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-95-167
November 1995
Evaluation of the High Volume Collection System (HVCS)
for Quantifying Fugitive Organic Vapor Leaks
By:
Eric S. Ringler
Southern Research Institute
6320 Quadrangle Drive, Suite 100
Chapel Hill, NC 27514
EPA Contract 68-D2-0062, Work Assignments 1/029 and 2/041
EPA Project Officer:
Charles C. Masser
National Risk Management Research Laboratory
Research Triangle Park, NC 27711
Prepared for:
U.S. Environgmental Protection Agency
Office of Research and Development
Washington, DC 20460
U.S.Environmental Protection Agency
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
-------
FOREWORD
The North American Research Strategy for Tropospheric Ozone (NARSTO) evolved out of a
call by the Clean Air Act Amendments of 1990 [Sec. 185B], in conjunction with the National
Academy of Sciences desire to reexamine the role of ozone (O3) precursors in tropospheric O3
formation and control. Therefore, the NARSTO study is examining the roles of nitrogen oxide
and volatile organic compound emission reductions, and the extent to which these reductions
contribute (or are counterproductive) to the achievement of O3 attainment in different
nonattainment areas.
ABSTRACT
Fugitive VOC emissions associated with gas and/or petroleum processing facilities have
historically been difficult and expensive to measure accurately. A measurements technique
has recently been developed that offers the potential for providing an easy-to-use and cost-
effective means to directly measure organic vapor leaks. The method is called the High
Volume Collection System (HVCS). The HVCS uses a high volume sampling device and a
portable flame ionization detector (FID) for field analysis. The HVCS can obtain direct
measurements of mass emission rates without the need for tenting and bagging. This study
of HVCS method performance included both field and laboratory testing. Laboratory
evaluation of HVCS results closely matched EPA method results with a difference in total
measured emissions of only about 3 percent. In one field test, the HVCS matched the EPA
estimate of total facility emissions within about 4 percent. In the second field test, the HVCS
measured an overall average of 18 percent more emissions than the EPA method. However,
the bias was present only in the early part of the test. In the latter part of the test, after efforts
were made to identify and correct the source of the bias, HVCS bias was essentially zero.
With some physical and procedural enhancements, the HVCS may be offered to EPA for
approval as an acceptable alternative to the EPA protocol bagging method with gas
chromatographic analysis.
Acknowledgements
The authors would like to acknowledge the participation of several individuals and
organizations who worked to make this study possible. Mike Webb, of STAR Environmental,
was a participant in the field study, completing the Method 21 screening and HVCS
measurements, and assisting in the data analysis and interpretation. The Gas Research
Institute (GRI) sponsored the initial development of the HVCS with STAR Environmental and
assisted in the study planning. Robert Lott represented the GRI. The American Petroleum
Institute (API) provided guidance and assistance in site selection. The API was represented
by Paul Martino and Charles Tixier.
-------
TABLE OF CONTENTS
Page
1.0 Introduction 1
1.1 Project Objectives and Scope 3
1.2 Site Descriptions and Leak Identification 4
1.3 Notes on the Data Analysis and Presentation 5
2.0 Laboratory Study Results 8
2.1 EPA Method Results 9
2.2 HVCS Method Results 13
2.3 EPA/HVCS Comparisons 16
2.4 Summary of Laboratory Study Findings 18
3.0 Field Study Results 20
3.1 South Texas Results 21
3.2 West Texas Results 24
3.3 Sources of HVCS Bias 27
3.3.1 High Background Levels 29
3.3.2 Analytical Bias 30
3.4 Gas Composition 31
3.5 Comparison of Direct Measurement Results with EPA Emission Factors ... 32
4.0 Data Quality 34
5.0 Conclusions 37
6.0 References 40
7.0 Data Tables : 41
Appendix A. Quality Assurance Project Plan 47
-------
FIGURES
Fig. No. Page
1-1 HVCS Diagram and Schematic 2
2-1 EPA Vacuum Method Laboratory Performance - Data Validation 10
2-2 EPA Vacuum Method Laboratory Performance - Regression Analysis 11
2-3 EPA Vacuum Method Laboratory Performance - Box Plot Comparisons 12
2-4 HVCS Laboratory Performance - Data Validation 13
2-5 HVCS Laboratory Performance - Regression Analysis 14
2-6 HVCS Laboratory Performance - Box Plot Comparisons 17
3-1 South Texas Results - Data Validation 21
3-2 South Texas Results - Regression Analysis 22
3-3 South Texas - Bias Dependence on Leak Rate 23
3-4 West Texas Results - Data Validation 24
3-5 West Texas Results - Regression Analysis 25
3-6 West Texas - Bias Dependence on Leak Rate 26
TABLES
Table No. Page
1-1 Screened Components and Emitters 5
2-1 Laboratory Results Summary 9
3-1 Field Study Results Summary 20
3-2 Summary of Gas Composition (percent) 32
3-3 Comparison of Emissions Estimates by Different Methods 33
4-1 Concentration Measurements - Data Quality 34
IV
-------
SECTION 1
INTRODUCTION
Fugitive emissions of methane and other organic vapors from leaking pipelines, valves, flanges,
and seals associated with natural gas, petroleum and chemical production and processing
facilities are an important source of methane and other organic emissions to the atmosphere.
Such emissions have historically been difficult to measure accurately. EPA Reference Method
21 "Determination of Volatile Organic Compound Leaks" describes instruments and procedures
that can be used to locate and assess the magnitude of such leaks. This consists of screening
components with a portable hydrocarbon analyzer and recording concentration values obtained
at the component interface where leaks are detected. Method 21 does not provide a direct
measure of the mass emission rate. Mass emission rates for leaking components have
traditionally been determined by "tenting and bagging" the leaks. This entails constructing a
sampling enclosure around the leaking component, introducing a known flow of diluent gas
through the enclosure, and determining the concentration of leaking gas in a sample captured
from the enclosure (usually done off-site in a laboratory). The concentration times the total flow
rate (diluent and leaking gas) gives the leak rate. This is clearly a laborious and expensive
method. A more practical, and less expensive method is desirable to inspect and repair leaking
components at a facility.
According to the current EPA protocol (EPA 1993), mass emission rates may be estimated
indirectly by one of four methods. (1) Emissions may be estimated by applying published
emission factors (EPA 1993) to an inventory of components by type (valves, flanges, etc.) and
service (e.g., gas, liquid). (2) More refined emissions estimates may be obtained by identifying
leaking components (per Method 21) and applying separate emission factors to leaking and
non-leaking components. (3) Still more refined estimates are determined by applying published
correlation equations (EPA 1993) to screening values for each component (obtained per Method
21). These emission factors and correlations were developed over the last 15 years based on
field studies at petroleum refineries, gas plants, and Synthetic Organic Chemical Manufacturing
Industry (SOCMI) plants. (4) Finally, the EPA protocol specifies procedures for developing unit
specific correlation equations that may more accurately estimate emissions for a specific facility.
This entails obtaining pairs of screening values and direct measurements of mass emissions (by
tenting and bagging) at a sufficient number of components so that representative, site specific
correlation equations can be developed.
A measurements technique has recently been developed that offers the potential for providing a
faster, easy-to-use, and cost-effective means to directly measure organic vapor leaks from gas,
oil, and chemical industry sources. The method is known as the High Volume Collection System
(HVCS). The HVCS was designed to obtain direct measurements of mass emission rates without
the need for tenting and bagging. The HVCS uses a 6 volt battery powered pump to draw
ambient air across a leaking component at controlled and metered rates between 10 and 500
standard cubic feet per hour. Figure 1-1 shows a diagram of the HVCS front panel and plumbing
(rear view). The flow may be directed through one of three rotameters, which are used to meter
the flow. The flow rate is controlled using the needle valves on the smaller rotameters (2-10 and
10-50 scfh), and the ball valve downstream from the larger rotameter (100 to 1000 scfh). A
portable Flame lonization Detector (FID) (Foxboro OVA Model 128) is used to measure the
hydrocarbon concentration in the HVCS exhaust.
1
-------
Exhaust
Control Valve
100 to 1000 scfh
Rotameters
100 to 1000 scfh
10 to 50 scfh...
2 to 10 scfh
Vacuum Pump (6V)
Inlet
-.
1
"^
(
-1
~
--
)
f
~
(
c
c
i
1
:
;
""-
:
i
i
>N
n
1
E
i
)
h
>
L
>
I i
Front Panel
Rear View
f
HVCS Inlet
ik^f
V
Sample (to OVA)
HVCS Schematic
Figure 1-1. HVCS Diagram and Schematic
-------
The nominal range of the OVA is from 1 to 10,000 ppm. The range can be extended somewhat
by directly accessing the voltage output of the OVA. The OVA voltage output is proportional to
the logarithm of concentration (in ppm) up to about 5.5 volts, which corresponds to roughly 15,000
ppm (this is somewhat instrument dependent). The OVA was calibrated specific to methane for
this study so that the hydrocarbon concentration is reported as methane.
The HVCS mass emission rate is determined as the product of the sample flow rate and the
hydrocarbon content in the flow. The success of the method depends on capturing all of the
leaking gas from a component in the flow entering the sample inlet. The inlet is constructed to
enhance this capture (the inlet is shaped like the mouthpiece of a snorkel). Diffuse leaks from
larger components (such as a large flange) are captured by wrapping the component in
polyethylene sheeting so that the air flow passes over the entire leaking surface. Prior to this
study, limited laboratory and field testing of the HVCS indicated that the system had the potential
to provide a practical method for quantifying fugitive hydrocarbon emissions.
1.1 PROJECT OBJECTIVES AND SCOPE
The purpose of this project was to complete a detailed evaluation of HVCS method performance
over the wide range of leak sizes, component types and operating conditions characteristic of
natural gas production in the United States. The focus of this evaluation is direct comparisons
of HVCS results versus controlled leak rates (laboratory) and EPA protocol "tent and bag" method
results (field). Consideration is also given to the broader issue of whether the HVCS can be used
to accurately determine the total emissions from a facility. In the report, the results are presented
from each of these points of view.
The study included both field and laboratory testing. The field testing assessed the accuracy of
the HVCS method relative to an EPA protocol emissions measurements method (tenting and
bagging). The EPA protocol identifies two methods that can be used to determine mass emission
rates: the vacuum method, and the blow through method. The vacuum method was selected for
the study based on preliminary laboratory testing (see Appendix A). The goal was to challenge
HVCS performance over the range of leak rates, component types and sizes, and operating
conditions characteristic of U.S. natural gas production. A major focus of the study was to
develop performance criteria for field use of the HVCS method. This included identifying
strengths and weaknesses of the prototype HVCS system and making recommendations for
improvement, identifying conditions under which best and worst HVCS performance is achieved,
and recommending procedures for obtaining optimum results. The laboratory testing was
conducted to establish the accuracy and precision of the EPA protocol (bagging) and HVCS
methods compared to controlled leak rates. This testing provided necessary support to the field
test results by examining the performance of both methods under controlled conditions. In the
laboratory, many of the sources of uncertainty associated with field testing were eliminated; the
most important of which is that the actual leak rate in the field is not known.
Laboratory testing was completed before the field testing and consisted of EPA and HVCS
measurements on constructed leaks representing a range of leak rates and component types
typical of natural gas production. Field testing was conducted at two gas production fields; one
in south Texas, and one in west Texas. The sites were selected to represent typical facilities
where one would expect to find some leaking components; that is, average age of approximately
-------
15 or more years, moderate operating pressure (< 1000 psi), low hydrogen sulfide levels, and with
no active leak detection and repair program. The sites also contained a sufficient number of
wells, compressor stations, and other installations, in a small enough area, that screening and
quantification could be accomplished in a cost effective manner. Leaks were identified with soap
solution, and selected leaks were quantified by both the EPA (bagging) and HVCS methods.
Gas composition was determined in the field by gas chromatography for each of the bag samples.
A detailed Quality Assurance Project Plan (QAPjP) was prepared, reviewed, and approved prior
to beginning any actual testing. This plan was conscientiously adhered to and served as a guide
throughout the field testing and final data analysis. Detailed sampling, analytical and QA/QC
procedures are described in the QAPjP, which appears as Appendix A to this report.
1.2 SITE DESCRIPTIONS AND LEAK IDENTIFICATION
The south Texas site is a gas production operation consisting of wells, tank batteries, and
compressor stations in three separate fields. Testing was completed in two of the three fields.
Field number 1 contained 48 wells, 3 tank batteries, and a compressor station with two units.
This was an older field operating at relatively low pressure (<100 psi). Field number 2 contained
15 wells, 1 tank battery, and 1 compressor station with a single unit. This was a relatively new
field, and was operating at higher pressure (100 to 1000 psi). All sampling in both fields was
conducted at the wells.
All components at a total of 44 wells were screened using a Foxboro Organic Vapor Analyzer
(OVA) (35 wells in field number 1 and 9 wells in field number 2), for a total of 7628 components.
A total of 59 leaks (0.77 percent of screened components) were located using soap solution.
Observable soap bubbles are generally not evident with very small leaks (screening values less
than 500 ppm). The majority of the leaks identified in this study had screening values of 10,000
ppm or greater. It should be noted, however, that a small leak will produce a screening value in
excess of 10,000 ppm. For example, a leak as small as 0.04 slpm (0.08 Ib/day as methane), and
perhaps smaller (this was the smallest leak constructed in the laboratory studies), produces a
Method 21 screening value in excess of 10,000 ppm with calm wind conditions. Most of the
emitters were valves (25/59), followed by various types of connectors (15,759), open ended lines
(8/59), and miscellaneous components (11/59). These were generally pressure regulators and
gauges.
The west Texas site is a significantly larger operation divided into three separate gas fields
feeding a common processing plant. Hydrogen sulfide levels are somewhat higher at this site
ranging from a few ppm to over 100 ppm at some wells. H2S scrubbers are present throughout
the field, and H2S levels at the plant are less than 10 ppm. There are also carbon dioxide
stripping facilities located in the fields. Recovered CO2 is compressed and re-injected into the
field at a number of wellheads. Oil wells are interspersed with gas wells at this site. Most of the
screening and leak quantification was conducted within the gas plant (51/76 samples). The
remaining samples were collected at wellheads (15) and at a propane storage area (10).
Operating pressures in the plant are about 890 (inlet) psi down to 30 psi (lowest). Pressure is
typically low at the wells (15 to 20 psi). The initial plant stages are about 35 years old. The plant
is now in its third phase of development. A total of 13,443 components were screened at 63
locations (plant operating units or well heads). Of these, 98 were identified as leaks (0.73
-------
percent). Most of the leaking components were valves (72/98), followed by threaded connectors
(11/98). Table 1-1 summarizes the screening and leak identification results.
Table 1-1. Screened Components and Emitters
Component
Flanges
Threaded Connectors
Tube Connectors
Valves
Open End Lines
Pressure Relief Valves
Miscellaneous
Total
South Texas
Screened
889
3733
931
1764
216
21
74
7628
Leaks
0
12
3
25
8
0
11
59
West Texas
Screened
1401
8010
982
2890
55
11
94
13443
Leaks
1
11
0
72
0
7
7
98
1.3 NOTES ON THE DATA ANALYSIS AND PRESENTATION
These notes are provided to clarify and explain the treatment and presentation of the data in the
analysis. The data presentation in sections 2 and 3 is consistent with this reasoning.
The data were validated prior to computing final summary statistics. Validation was necessary
since both the EPA and HVCS methods can give spurious results under some circumstances
(discussed in some detail throughout this report). Such results are not representative of method
performance. Two types of validation were performed: operational and statistical. Operational
data validation refers to excluding data where operational problems were noted during sampling
or analysis that would compromise data quality. For the EPA method, data were excluded if the
total hydrocarbon concentration in the two samples collected from each leak differed by more than
10 percent. EPA method data were also excluded when there were very large leaks. The
sampling train used for bag sampling had a maximum flow capacity of about 20 slpm. It was
found that if the leak rate exceeded about 8 slpm, the sampling train was unable to capture the
leak because the volume of diluent gas was insufficient to promote complete mixing in the bag.
HVCS method data were marked suspect if leak capture could not be verified by reproducing the
leak rate (within 20 %) at more than one HVCS flow rate. Data were also excluded if there was
a known or strongly suspected outside source of bias. In the south Texas results, there were
several instances where there was evidence that the leak rate had changed substantially between
the EPA method and HVCS quantifications. In the west Texas data, several tests were made in
a propane storage area, where the gas composition was primarily propane. Since there is a
-------
known difference in OVA and GC response to propane, these data were also excluded. A
statistical validation based on examination of scatter plots with regression analysis was also
performed. If a point representing an EPA/HVCS data pair was substantially outside of a 95
percent confidence interval for the regression, it was removed. It is recognized that this type of
validation could not be performed for routine HVCS field measurements; however, it is considered
that removal of outlying values that significantly skew the results improves the overall
representativeness of the HVCS method evaluation.
In this study, the fractional (or percent) difference is used as a measure of overall method
performance (independent of leak size). That is, differences between EPA and HVCS results are
normalized for the size of the leak (taken as the EPA method result). This measure was chosen
because it allows comparison of results over a range of leak sizes, and is simple to grasp
intuitively. By its nature, the distribution of the fractional bias is not symmetric. It takes on values
from -1 to +00, with negative biases represented by values between -1 and 0, and positive biases
spread from 0 upwards. The logarithm of these values will produce a symmetric distribution.
Thus, the underlying distribution for the fractional bias should be described by a 3 parameter
lognormal distribution with an offset, T, of one.
One unfortunate characteristic of log-normally distributed measures, is that the mean and variance
of the log transformed data are not the logarithms of the mean and variance of the original,
untransformed data. This affects how summary statistics for method performance should be
calculated and presented. Since a logarithmic distribution is skewed right (high biases have
greater magnitude than low biases), the arithmetic mean computed from the raw data may be
biased high. Also, since a logarithmic distribution is not symmetric, confidence intervals do not
have the same width on either side of the mean. Unbiased estimators of the mean and variance
of log-normally distributed data are available in the literature (e.g., Gilbert 1987). Minimum
variance unbiased (MVU) estimators for the mean are used in this report. Summary results are
reported as the MVU estimate for the mean, followed by the lower and upper limits for a 95
percent confidence interval in parenthesis. The confidence interval for log-normally distributed
data is determined as recommended by Gilbert (1987), and using the tables provided. As it
happens, the sampled biases are roughly symmetrically distributed and do not differ greatly from
a normal distribution (based on probability plotting). Therefore, the arithmetic mean and
confidence interval computed from the "t" distribution is very similar to that computed for the
lognormal distribution. The lognormal statistics, however, lend a somewhat greater degree of
statistical power to the results. That is, the lognormal confidence intervals are "tighter"
Significance tests for differences between means are conducted on the log transformed data
using the standard "t" statistic.
Similar to the fractional difference, the leak rates themselves tend to be log-normally distributed.
That is, there tend to be a greater number of small leaks, and relatively few very large leaks.
This takes on importance in the regression analysis where the results from the two methods are
again compared. Again, there is a problem with direct transformation from log space to arithmetic
space. To correct for scale bias introduced during the transformation, a scale bias correction
factor (SBCF) must be applied. This gives a MVU estimate of the slope of the regression line.
The form of the linear regression in log space is given in equation 1-1, where and b are constants
determined from the regression.
-------
\n(HVCS) = a + b \r\(EPA) '
HVCS = (SBCF) exp(a) EPA" 1'2
In arithmetic space, the regression equation should be expressed as in equation 1-2, where the
SBCF depends on the mean square error (average of the squared residuals) and the number of
data points. It is obtained as the sum of an infinite series (EPA 1993).
Grouped box plots (see Figure 2-3 for an example) are used in this report to illustrate
comparisons of segregated data sets. The box plot succinctly illustrates both the central tendency
and the distribution of a set of data. The center line in the box represents the median of the data.
The upper and lower boundaries of the box represent the 75th and 25th percentiles respectively.
The "whiskers" extend vertically above and below the box a distance from the median to the
furthest data point within 1.5 times the inter-quartile range (the difference between the 75th and
25th percentiles), or the height of the box. For normal distributions, the whiskers encompass
approximately 95 percent of the data. Data lying outside the whiskers are plotted as points.
These are often considered outliers. Side by side comparisons of box plots provide a means to
make meaningful comparisons that account for how the data are distributed. This helps, for
example, to avoid erroneous comparisons of "average" values that may be greatly skewed by
outliers, or by non-symmetrical distributions.
-------
SECTION 2
LABORATORY STUDY RESULTS
The EPA vacuum bagging method, as well as the HVCS method, was evaluated in the laboratory
studies. Since the EPA method, like any measurement, is subject to imprecision and bias,
quantification of these data quality indicators was essential before the EPA method could reliably
be used in the field as a measure of HVCS performance. Only limited controlled testing of EPA
protocol method performance has been conducted previously (Radian 1980), and this did not
include treatment of errors associated with the total sampling system, including the "bag" or
component enclosure. The laboratory tests conducted for this study were devised to represent
"real world" components and leak types so that overall errors (including total sampling errors) are
represented. Actual pipeline components were assembled in such a manner that induced leak
rates could be precisely controlled and accurately metered against a primary flow standard.
Components tested included a 2 inch gate valve, a 4 inch threaded coupling, a 6 inch pipe flange,
and a 1/2 inch pump shaft. These represent component types and sizes that are typically
encountered at natural gas production and processing facilities. Details of the laboratory test
bench set-up and test matrix are given in the QAPj'P (Appendix A). Laboratory test procedures
were identical to the field test procedures.
The EPA protocol vacuum bagging method was selected for the study based on preliminary
laboratory testing. Initially, the blow-through method was selected because the required
apparatus is less complicated, and because a non-combustible gas such as N2 can be used as
the diluent, eliminating the risk of obtaining explosive mixtures in the sample bags. However,
during initial testing, the blow-through method gave unsatisfactory results. Errors were large, and
results were inconsistent. One problem is that the blow-through method does not provide for
direct determination of the total flow (including the leak flow) through the system, since only the
flow of carrier gas is measured. This introduces a direct and significant error when leak flows are
large relative to the carrier gas flow. An attempt was made to measure the total flow downstream
from the enclosure (bag); however, this requires a completely leak-free enclosure, which is
generally impractical to execute. Initial trials with the vacuum method gave good results, so it
was adopted for the study.
Leak rates induced in the lab study span 4 orders of magnitude and are representative of the
range of leaks likely to be encountered at actual gas and oil production facilities. Induced leak
rates ranged from 0.02 to 20 slpm (0.4 to 40 Ibs/day as CH4). This range is representative of
essentially all of the emissions and more than half of the emitters (including all of the larger
emitters) characterized in recent measurements at on-shore gas and oil production operations
(API 1993) at 16 sites where nearly 140,000 components were screened and 4200 emitters
identified. A summary of the laboratory test results is given in Table 2-1. Laboratory test results
are discussed in detail in the following paragraphs.
-------
Table 2-1. Laboratory Results Summary
Method Bias
EPA Method Bias vs. Induced
HVCS Bias
HVCS Relative Bias
HVCS "True" Bias
Difference in Total Emissions
EPA Method vs. Induced
HVCS Method vs. Induced
EPA vs. HVCS
Mean
(MVU)
-7.4%
-8.3%
0.3%
-7.1%
Leak
(slpm)
96.1
175.1
25.1
(EPA)
Lower
95% Limit
-9.7%
-12.0%
-9.8%
-15.7%
Result
(slpm)
90.4
164.5
24.4
(HVCS)
Upper
95 % Limit
-5.0%
-4.3%
8.5%
0.5%
Diff (%)
-5.9%
-6.1%
-2.8%
No.
55
32/551
9/221
9/221
No.
55
55
22
2.1
Summary results are calculated using only data that are unaffected by the probe position bias (see
Section 2.2).
EPA METHOD RESULTS
A total of 59 constructed leaks were measured by the EPA vacuum method. Fifty-five of these
were valid measurements. Three tests were invalidated because the concentration analyses for
the two bag samples differed by more than 10 percent. One additional test was invalidated based
on a regression outlier test (point was outside of a 95 percent confidence interval for the
regression in log space, and removal of the point improved the regression fit). The QAPjP data
quality objective for the EPA laboratory measurements of 45 valid quantifications was met. Figure
2-1 shows EPA method bias (in chronological order) for the raw and validated data sets.
The total induced leak rate from all components tested was 96.1 slpm (198.9 Ib/day). The total
leak rate measured by the EPA method was 90.4 slpm (187.1 Ib/day) , for an overall difference
of -5.9 percent. If the totality of the test results is viewed as a facility, this result can be viewed
as the error in the total facility emissions estimate.
The MVU (see Section 1.3) estimator of the mean bias in the EPA method is -7.4 (-9.7 to -5.0)
percent, where the range given in parenthesis represents the lower and upper limits for a 95
percent confidence interval for a lognormal distribution (Gilbert 1987). These results are
summarized in Table 2-1. Clearly, the EPA method exhibited a negative bias for the laboratory
study. This was most likely due to incomplete mixing in the bag. That is, dilution air
(atmospheric) may have been taken up by the sampling system directly, without having mixed
completely with the leaking gas. This would explain a negative bias. Every effort was made to
minimize this artifact. The normal procedure included leak testing sampling enclosures prior to
-------
c
o
U
f-J
w
CO
to
(U
c
.2
CO
c
(D
Ģ
o
CO
D
m
TD
O
(
^J
cu
-^
E
13
U
D
~^
1.2
1 .0
0.8
0.6
0.4
0.2
r\ o
U. U
-0.2
-0.4
-0.5
-0.8
1 n
* \Ja\\d Data
: Raw Data
^
/\
A / \
ŧ> \
A A
AA
-
_
.
-
.
, I , I , I , I , I , I , I , I , I , I , I , I , I . I , I , I , I , I , I , I , I , I , I , I i 1 , 1 , 1 , 1 , 1 . 1 . ! , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , | , ] , | , | , | , | , | , | 1,1,1,1,1.1,1,1,1.
<
D_
OOF 01C 04b 07b 1 7b 18b 1 9a 20b 22a 22f 24e 25d
Sample ID (Chronological Order)
Figure 2-1. EPA Vacuum Method Laboratory Performance - Data Validation
sampling (the bag should collapse over the enclosure when vacuum is applied), and then cutting
small air inlets into the bag opposite the leak from the sampling inlet. There is a strong, linear
relationship between the EPA results and the induced leak (r2 = 0.991). The regression is shown
in Figure 2-2.
The bias in the EPA method bias is not strongly dependent on leak rate, component type, leak
type, or internal pressure. Figure 2-3 shows box plot comparisons. At first, it may appear that
the method is more strongly biased for the smaller leaks (< 0.1 slpm); however, this is an artifact
of the use of the fractional bias to measure method performance. By its nature, the fractional bias
will tend to be larger with smaller leaks simply because a small difference may be a large fraction
of a small number. Since the absolute magnitude of these errors is small, it was decided that
rigorous statistical tests to infer whether the apparently larger bias for small leaks is significant
were unnecessary. From the box plots, there is no observable dependence of method
performance on component type, pressure, or leak character.
10
-------
1 01
3
2
00
3
2
10-1
3
2
10-2
EPA = 0.931 * 1 .003 * InducedO °'3), r2 = Q.991
345 1 Q-1 2345 ] QO 2
Induced Leak Rate (slpm)
345
101
Figure 2-2. EPA Vacuum Method Laboratory Performance - Regression Analysis
11
-------
31
(Q
c
co
m
TJ
>
u
o
c
c
3
S
(D
-*
3^
O
a
u
cr
o
(D
3,
O
0)
O
(D
CD
O
X
O
O
3
TJ
0)
2.
w'
o
0 3
0 2
10
? 0 1
CD
O
c 00
u
o
^ 0 1
-0 2
0 3
0 2
10
9 0 1
4 slpm)
Induced Leak Rate
Lab Test - Vacuum Method Results
Data Distribution by Leak Character
Lą[] [D
""1""
[ J 1
Diffuse jŧt
Li'ok Tharoc ter
-------
2.2 HVCS METHOD RESULTS
A total of 75 constructed leaks were quantified by the HVCS method. Sixty of these were
operationally valid, meaning that the coefficient of variation (ratio of the standard deviation to the
mean) for three leak rate determinations at different flows was within 20 percent. Most of the 15
invalid tests were invalidated due to incomplete leak capture at lower HVCS flows. These could
have been recovered by simply monitoring the results during testing, and repeating tests until
three "good" measurements were obtained. For the field studies, a procedure was developed to
identify instances where incomplete leak capture may have occurred and, if possible, recover
valid data by omitting these occurrences (see Section 4). Five additional samples were excluded
based on regression outlier tests, so that a total of 55 valid tests remained. The QAPjP data
quality objective for the HVCS measurements of 45 valid quantifications was met. Figure 2-4
shows HVCS method bias for the raw and validated data sets.
Valid Data
Raw Data
c
o
o
D
(ft
_0)
C
o
'on
C
a>
E
O
00
O
CD
00
CJ
002a 01 B 07b 08f 1 3c 09a 1Oe 12a_1 1 4c 17e P6
Component ID
1 Probe Position Effect
Figure 2-4. HVCS Laboratory Performance - Data Validation
The total induced leak rate from all components tested was 175.1 slpm (84.6 Ib/day). The total
leak rate measured by the HVCS method was 164.5 slpm (79.5 Ib/day) , for an overall difference
of -6.1 percent. If the totality of the test results is viewed as a facility, this result can be viewed
as the error in the total facility emissions estimate.
13
-------
Figure 2-4 also indicates a change in sampling methodology that was initiated when a sampling
bias related to the position of the OVA probe in the HVCS exhaust stream was discovered. It was
found that when the OVA probe is positioned directly into the HVCS exhaust (probe at 0 degrees
to the exhaust flow), a higher concentration reading is obtained than when the probe is positioned
at 90 degrees to the exhaust flow. This is due to the increased pressure on the OVA inlet, which
increases the sample flow to the detector, and produces a bias. It should be noted, that in earlier
work with the HVCS, the OVA probe was positioned directly into the HVCS exhaust. In order
to determine the magnitude of the probe position bias, a series of measurements were made with
the probe alternately positioned directly into the flow and at 90 degrees to the flow. Results with
the probe positioned directly into the flow were consistently higher than those where the probe
was positioned at a right angle to the flow. The bias over four series of tests (at leak rates of
0.76, 0.305, 0.045, and 0.0135 slpm, covering the operational concentration range for the OVA,
and spanning the effective HVCS flow range) averaged +11.6 ą1.1 percent. It was concluded
that the probe should be positioned at a right angle to the HVCS exhaust flow and "T" adapter
was constructed that was used in all further laboratory and field testing.
E
CL
3
2
O1
3
2
1 0°
3
2
10-1
3 -
HVCS = 0.890 * 1 .004 * InducedO 029)_ r2 = 0.975
3 4 1 Q-1 234 1 QO 2345
Induced Leak Rate (slpm)
Figure 2-5. HVCS Laboratory Performance - Regression Analysis
A total of 23 of the valid HVCS leak rate measurements were obtained with the OVA before the
effect of probe position was noted. While the probe position bias was clearly apparent in the
14
-------
special testing described above, it was not possible to distinguish the bias in the overall data set.
In many cases, the effect is probably overwhelmed by other sources of bias, such as incomplete
leak capture. On this basis, it was decided that applying a "correction" for the bias was not
advisable. Conclusions and recommendations made in this report are based on the 32 HVCS
measurements that are unaffected by the probe position bias.
For the data unaffected by the probe position bias, the MVU estimate of the mean is -8.3 (-12.0
to -4.3) percent. The HVCS determinations correlate well with the induced leak rates (r2=0.992).
The regression is shown in Figure 2-5. The magnitude and direction of the HVCS bias is similar
to that for the EPA method; however, there is considerably more variability in the HVCS data.
(e.g., compare Figures 2-1 and 2-4). Nonetheless, it is clear that the overall bias is negative.
The negative bias is probably due to incomplete leak capture. In the laboratory studies, the
HVCS "snorkel" inlet (about 2 inches across) was used alone on the smaller components (e.g.,
the 2 inch valve, and pump shaft). On larger components (e.g., the 4 inch threaded coupling),
a larger snorkel shaped inlet (about 4 inches across) was used, and capture was enhanced by
shielding the side of the component opposite the HVCS inlet with Mylar film. On the 6 inch
flange, a Mylar shield was wrapped around the component to form a partial enclosure. It is
believed that some of the leaking gas was not caught up in the HVCS slip stream, even with the
aid of the shields and partial enclosure. Partial enclosures were used much more extensively in
the field. In the field, the HVCS inlet was typically attached near the leak interface and the entire
interface and inlet were wrapped in light weight polyethylene. A spring coil (toy "Slinky") was
used, when necessary, to prevent the wrapping from obstructing the leak and allow unrestricted
airflow around the component. This additional wrapping appears to significantly improve leak
capture, especially in windy conditions.
The bias in the HVCS method bias is not strongly dependent on leak rate, component type, leak
type, or internal pressure. Figure 2-6 shows box plot comparisons for the entire (uncorrected)
data set. As with the EPA data, it appears at first that the method is more strongly biased for
smaller leaks (< 0.1 slpm), however, this is probably an artifact of the measure used for bias.
Once again, since the absolute magnitude of these errors is small, it was decided that rigorous
statistical tests to infer whether the apparently larger bias for small leaks is significant were
unnecessary. From the box plots, there is no observable dependence of method performance
on component type, pressure, or leak character.
The prototype HVCS demonstrated in the lab tests uses a 6 volt DC vacuum fan motor to draw
ambient air at up to 400 scfh across a leaking component. The gas flow is measured by one of
3 rotameters, covering a range from 2 to 1000 scfh. The concentration of hydrocarbon in the gas
mixture (air and leaking gas) is measured by a Foxboro OVA 108 with an output scale from 1 to
10,000 ppm and a voltage output that is proportional to the log of concentration from 0.0 up to
about 5.5 VDC (for a maximum concentration of about 14,500 ppm with the unit used in the lab
study). Above 5.5 VDC, the OVA output is highly non-linear.
HVCS leak measuring capacity is limited by the HVCS flow capacity and/or the concentration
range measurable by the OVA. At maximum flow (about 400 scfh), leaks of up to 1.9 slpm (3.8
Ibs/day methane) can be measured on the conventional OVA scale (up to 10,000 ppm) and leaks
up to about 2.7 slpm (5.4 Ibs/day methane) can be measured by extending the OVA scale (up
to 5.5 V or about 15,000 ppm) using the voltage output. However, the leak rate should be
determined at more than one HVCS flow to verify leak capture. If measurements are to be made
15
-------
at each of the three HVCS flow scales (100 to 1,000 scfh, 20 to 100 scfh, and 2 to 20 scfh), the
maximum measurable leak rate is around 0.07 slpm (assuming a lower flow of 10 scfh). This is
among the smallest leak rates measured in the lab studies. As a practical matter, HVCS
precision can be determined using several flows on the upper HVCS flow scale (100 to 400 scfm).
This gives a maximum directly measurable leak rate of about 0.7 slpm, or 1.4 Ib/day (with a lower
flow of 100 scfh), a moderate sized leak. To quantify larger leaks, a dilution probe must be used
on the OVA, or a hydrocarbon analyzer with greater range must be used.
To cover the range of leaks in the lab tests, HVCS leak measuring capacity was extended using
a Geotechnical Instruments GA 90 Infrared (IR) gas analyzer (Model* 601003), which measures
the methane concentration from 0.1 to 100 percent. Concentrations below 1 percent (10,000
ppm) were measured by the OVA, and above 1 percent were measured by the IR analyzer. This
worked well in the laboratory studies since the leaking gas was nearly 100 percent methane. The
response of the IR analyzer to hydrocarbons other than methane was not tested directly. In the
field, however, the response of the IR analyzer to the gas mixtures encountered was
unpredictable, indicating that the device is not well suited for gas field measurements. Of the 55
valid measurements, 35 were made using the OVA, and 20 were made with the IR analyzer. It
should be noted that the IR analyzer did not exhibit the probe position bias that was observed for
the OVA. There was no distinct difference in HVCS method bias depending on which analyzer
was used. Foxboro's dilution probe was used extensively in the field, with apparently good
results. Most dilution factors were on the order of 30 times. Higher dilution ratios should
probably be avoided.
2.3 EPA/HVCS COMPARISONS
For 22 tests, both EPA and HVCS measurements were made on exactly the same constructed
leak, allowing for direct comparisons. Only 9 of these tests, however, were unaffected by the
OVA probe position bias. Summary results for these 9 data pairs are presented here. For this
subset, HVCS bias was -8.6 (-17.3 to -1.2) percent. EPA bias was -8.7 (-13.4 to -3.2) percent.
These results are consistent with those for the overall data set. HVCS bias relative to the EPA
method was +0.3 (-9.0 to +8.5) percent, indicating that the two methods get essentially the same
results, though both exhibit small, negative biases. The HVCS and EPA methods correlate very
well with each other (r2 = 0.999).
The total induced leak rate from all components tested by both methods was 26.1 slpm (54.0
Ib/day). The total leak rate measured by the EPA method was 25.1 slpm (52.0 Ib/day), and the
total leak rate measured by the HVCS method was 24.4 slpm (50.5 Ib/day). The overall
difference between the EPA and HVCS methods was of -2.5 percent. If the totality of the test
results is viewed as a facility, this result can be viewed as the difference in total facility emissions
estimates that would be obtained by the two methods.
The "true" HVCS bias (i.e., relative to the actual leak rate as opposed to the EPA method result)
is calculated from the HVCS and EPA measurements, and the measured bias in the EPA method
(EPAE, as determined in the laboratory) using equation 2-1, which is derived in the QAPjP For
the laboratory comparisons, "true" HVCS bias is found to be -7 1 (-15.7 to +0.5) percent, using
the -7.4 percent bias determined for the EPA method. This calculated "true" HVCS bias of -7 1
percent is reasonably consistent with the directly measured HVCS bias of -8.3 percent. The
16
-------
(O
ro
p>
I
O
0)
u
tr
o
a
(D
u
3
O
(D
i
CD
O
X
O
o
3
o
ĢU
2.
(A
0 5
0 4
0 3
°2
o
tD 0 1
"o
.1 °'°
§ -01
-02
-03
-0 4
0 5
0.4
0.3
ŧ °2
D
CD 0 1
~D
0 ° °
u -o,
U -0.2
0 3
-0 4
Lab Test - HVCS Results
Data Distribution by Component Type
| 18 | [T] | 16 | [T] | 11 |
- T T .
: I r-1-!
t
1 1 I
' 1 1
-
1 i
Valve Open Coupling Flange Pump
Component Type
Lab Test - HVCS Results
Data Distribution by Internal Pressure
r | 22 | | 19 | | 14 |
1 T -L
T
1 ' i '
1
i
Lev. (< J.Opsi Med(<27.7psi) High (>27.7 psi)
Component Internal Pressure
0.5
0 4
0 3
°'Z
o
m 0 i
o
o °°
g -01
-02
-0 3
-0 4
0.5
0.4
0.3
0-2
a
CD 0.1
o
c no
g -0.1
^ -0.2
-0 3
-0.4
0 5
Lab Test - HVCS Results
Data Distribution by Induced Leak Rate
[7] | 16| | 19 | | 13 |
T
t T ^-L_
I i
' r1-! -r- 1
-
1
i ....
5 (< O.t slpm) M (<1 slpm L(<4slpm) XL()4slpm)
Induced Leak Rate
Lab Test - HVCS Results
Data Distribution by Leak Character
r \JK\ [23]
1 T
1
1
Diffuse Jet
Leak Character
-------
"true" bias is an important measure for the field performance of the HVCS, since the true leak rate
will not be known absolutely, but must be inferred from the EPA method results.
HVCS - EPA + EPAE HVCS
True Bias - E-
where, 2-1
HVCS is the leak rate determined by the HVCS (slpm),
EPA is the leak rate determined by the EPA method,
EPAE is the laboratory measured bias for the EPA method
2.4 SUMMARY OF LABORATORY STUDY FINDINGS
The primary goal of the laboratory study was to determine the precision and accuracy of the EPA
method under controlled conditions. This was essential to the study since the EPA method is
used in the field as the point of comparison for determining HVCS method performance, and
representative test data were not available from other sources. The laboratory study yielded
reliable estimates of EPA vacuum method precision and accuracy representative of a wide range
of leak rates, and component types and sizes. The EPA method exhibited a small, but significant
negative bias, averaging about -7.4 percent, which was quite consistent with a 95 percent
confidence interval of about ą2.4 percent. The cause of the bias is most likely incomplete mixing
within the sampling enclosure, which occurs even when reasonable measures are taken to
promote mixing. Perhaps some modification to the method would improve performance; however,
such a determination was beyond the scope of this project.
The second goal of the laboratory study was to examine HVCS performance under controlled
conditions to determine its operating characteristics and optimum operating conditions. Several
important findings were made.
The HVCS compared well with the EPA method in direct comparisons. Overall, the HVCS
exhibited a negative bias of nearly the same magnitude as the bias in the EPA method.
The cause of the HVCS bias was incomplete leak capture. The cause of the EPA method
bias was incomplete mixing in the bag.
Both methods gave accurate determinations of "total facility emissions", viewed as the
sum of emissions from all tested components.
The analyzer probe should be positioned perpendicular to the HVCS exhaust flow to avoid
a positive sampling bias on the order of 10%.
The prototype HVCS unit used with the Foxboro OVA 108 is not adequate for measuring
larger leaks (more than about 1 slpm) without the use of a dilution probe or an alternate
analyzer with greater range. In some tests, the HVCS exhaust contained methane
concentrations as high as 15 percent by volume.
18
-------
In general, the flow capacity of the prototype HVCS is not generally sufficient to effect
complete leak capture unless an enclosure is constructed around the leaking component.
Finally, "true" HVCS bias based on comparisons of HVCS results with bias corrected EPA method
results closely matches the HVCS bias determined from actual comparisons to known leak rates.
19
-------
SECTION 3
FIELD STUDY RESULTS
The overall goal for the field study was to obtain 125 valid pairs of HVCS and EPA method leak
quantifications. A total of 135 pairs of operationally valid measurements were obtained. Of these,
114 pairs were used in the final data analysis. To locate leaks for the study, over 21,000
components were screened at two gas production facilities located in separate areas (south
Texas and west Texas). A more detailed description of the two facilities and of the numbers and
types of components screened and emitters found is given in Section 1.2. Table 3-1 gives a
summary of the field study results. Note that, for the west Texas study, results are presented for
the overall study and as segregated before and after September 26 (1994). This was done
because the apparent HVCS bias changed dramatically after this date. Sources of bias, and
possible reasons for the improvement in results are discussed in section 3.3.
Table 3-1. Field Study Results Summary
Method Bias
South Texas HVCS Relative Bias
South Texas HVCS "True" Bias
West Texas HVCS Relative Bias
Overall
West Texas HVCS "True" Bias -
Overall
West Texas HVCS Relative Bias
On and Before September 26
West Texas HVCS "True" Bias
On and Before September 26
West Texas HVCS Relative Bias
After September 26
West Texas HVCS "True" Bias
After September 26
Difference in Total Emissions
South Texas
West Texas - Overall
West Texas -
On and Before September 26
West Texas - After September 26
Mean
(MVU)
15.2%
6.6%
44.5%
33.8%
67.0%
54.7%
5.7%
0.0%
EPA
(slpm)
55.8
79.2
32.7
46.5
Lower
95% Limit
4.1%
-3.6%
34.3%
24.4%
56.5%
44.9%
-0.4%
-7.3%
HVCS
(slpm)
58.2
93.9
48.1
45.8
Upper
95 % Limit
29.8%
20.2%
56.7%
45.1%
84.8%
71.1%
18.6%
9.8%
Diff. (%)
+4.3%
+ 18.6%
+47.1%
-1.5%
No.
48
48
66
66
40
40
26
26
No.
48
66
40
26
20
-------
3.1 SOUTH TEXAS RESULTS
At the south Texas site, 56 operationally valid data pairs were obtained, and 48 of these were
retained in the final data analysis. Four pairs were eliminated as outliers (regression test). Two
pairs were eliminated because there were very large absolute differences in the leak rates
determined by the two methods. These "leaks" were probably, in actual fact, pressure releases
due to a compressor failure that occurred downstream, so that the emission rate changed
substantially in the interval between the two measurements. Two additional pairs were eliminated
due to unusually high percentage differences in the two measurements. Figure 3-1 shows a time
plot of the raw and validated data from the south Texas site.
c
o
u
o
CO
LO
_0>
c
o
w
c
0)
E
Q
0_
LJ
O
0)
tr
o
bo
00
o
X
9
8
7
6
5
4
3
2
1
0
Raw Data
Valid Data
RP1 1 1 RP121 TW212 TW2110 TW343 TW42 TW321 CNG1 1 2
Component ID (Chronological Order)
Figure 3-1. South Texas Results - Data Validation
Leak rates measured at the south Texas site ranged from less than 0.01 to more than 9 slpm as
measured by the EPA method. Some larger leaks were not measurable by the EPA method.
The HVCS method measured leak rates up to 13 slpm. The average leak rate was about 1.2
slpm, with a median of 0.25 slpm. Total measured emissions (EPA method) were about 70 slpm,
or about 140 Ib/day, representing most of the leaks in two of three gas fields served by the
facility.
21
-------
1 Q1 ; HVCS = 0.951 * 1 .04 * EPA(o.945), r2 = o 961
a
GJ
10-2
10-3
1 O-4 L
1 0-42 10-32 1 0-22 1 Q-1 2
EPA Leak Rate (slpm)
Figure 3-2. South Texas Results - Regression Analysis
u
HVCS bias was +15.2 (4.1 to 29.8) percent relative to the EPA method, and "true" HVCS bias
was +6.6 (-3.6 to 20.2) percent. There is a significant degree of imprecision in these results due
to the high degree of variability in the data (the standard error of the mean is about 45 percent
of the mean value). This yields wide confidence intervals. There appears to be a positive bias
in the field data compared to the laboratory results; however a statistical comparison gives a
probability that the means are different of. only 65 to 75 percent. That is, given the variability in
the data, the average laboratory study and south Texas field study results cannot be distinguished
with a very high level of statistical significance. Nonetheless, it is reasonable to suspect that
there is some positive HVCS bias in the south Texas results. The negative bias observed in the
laboratory studies was probably eliminated by the additional component wrapping (see section
2.2) that was routinely performed in the field. The source of the additional positive bias has not
been definitively determined, but may be an analytical bias caused by differences in OVA and GC
response to the same samples. This is discussed in detail below (see Section 3.3).
Despite the differences, the HVCS and EPA measurements correlate very well (r2 = 0.96). A
regression plot and equation is shown in Figure 3-2. Segregation of the data by leak size (see
Figure 3-3) shows that the HVCS bias is larger for smaller leaks. As previously noted, this is
typical of the use of a relative bias, or percent difference as a measure of method performance.
22
-------
In terms of an overall inventory, the relative bias overstates the difference between the two
methods. Overall measured emissions (final validated data only), are 58.2 slpm (about 120
Ib/day) for the HVCS method and 55.8 slpm (about 115 Ibs/day) for the EPA method, an overall
difference of only about 4 percent. SRI also investigated possible dependence of the HVCS bias
on the delay time between the EPA and HVCS measurements, component type, and use of the
dilution probe (6 tests). None of these factors had an observable influence on HVCS bias. For
south Texas, delay times were often significant (up to several days); however, this did not appear
to have a significant impact on the data. The delay time problems in South Texas were caused
by excessive time required to locate leaks, since the gas well locations were widely scattered.
There were no problems with delay time in the West Texas study, where most of the leaks were
located within a gas plant.
c
o
L)
O
c
o
'en
tz
0)
E
Q
D_
Ld
O
CD
2.0
.5
1 .0
0.5
0.0
Qi -0.5
in
o
CD
C/)
O
I
1 .0
S(< 0.1 slpm) M (<1 slpm) L(<4slpm) XL (>4 slpm)
EPA Leak Rate
Figure 3-3. South Texas - Bias Dependence on Leak Rate
23
-------
3.2 WEST TEXAS RESULTS
At the west Texas site, 79 operationally valid data pairs were obtained, and 66 pairs were
retained in the final data analysis. Two pairs were eliminated as outliers (regression test). Five
pairs were eliminated due to very high absolute differences in the leak rate. These were very
large leaks (> 8 slpm) where the EPA method may have been ineffective since the maximum
vacuum flow that could be attained with the apparatus used in the study was about 20 slpm. In
addition, 6 pairs were excluded because the gas composition was primarily propane, and the OVA
response to propane is less than half of the GC response. Figure 3-4 shows a time plot of the
raw and validated data from the west Texas site. The improvement in the HVCS bias after
September 26 (just before ID71) is readily apparent in the Figure. This improvement is discussed
in section 3.3.
o
o
in
in
_QJ
C
O
' in
L
CD
E
Q
CL
LJ
O
0)
rr
IT,
g
m
CO
o
I
2.0 r
1 5
1 0
0 5
0 0
-0.5
A Raw Data
-A Valid Data
.°1\ . ' ..At
ID4 ID9 ID25 ID35 ID45 ID52 TEST2 ID71 1D77 IDB IDI ID91
Component ID (Chronological Order)
Figure 3-4. West Texas Results - Data Validation
24
-------
Leak rates measured at the west Texas site ranged from less than 0.01 to more than 20 slpm.
The average leak rate was about 1.2 slpm, with a median of 0.7 slpm. Total measured emissions
(EPA method) amounted to about 130 slpm (about 260 Ib/day). The majority of the
measurements were obtained in the gas plant (51/79). The remainder were collected at well
heads, and in a propane storage area.
Average HVCS bias relative to the EPA method was +44.5 (34.2 to 56.7) percent. HVCS "true"
bias was +33.8 (24.4 to 45.1) percent. As in south Texas, there is a significant degree of
variability in the results; however, it is clear that there is a very significant positive HVCS bias in
the overall results. The probability that the mean bias in the west Texas results is the same as
in the laboratory studies is very small (0.4 percent). The probability that the mean west Texas
bias is the same as in the south Texas data is also small (0.9 percent). The most highly biased
results are concentrated in the early part of the study (before September 26). This is significant,
CD
O)
O
o:
-*.
o
0)
_)
in
o
:r
3
2
00
3
2
0-1
3 -
2 -
0-2 :
HVCS - 1 .278 * 1 .03 * EPA(0-945)_ r2 = o.897
2 3 1 Q-2 23 1Q-123 ] QO 2 3
EPA Leak Rate (slpm)
O1
Figure 3-5. West Texas Results - Regression Analysis
as there were some operational and procedural changes that were made in the field after that
date in an attempt to improve results (see section 3.3). Data collected on and before September
26 show an average HVCS bias of +67.0 (56.5 to 84.8) percent, while data collected after that
date show an average bias of only +5.7 (-0.4 to 18.6) percent.
25
-------
For the validated data, the HVCS method measured a total of 93.9 slpm (194 Ib/day), while the
EPA method came up with only 79.2 slpm (164 Ib/day). As seen before, the HVCS bias is larger
for small leaks, but the difference in the total emissions for the overall study is still significant at
18.6 percent. Most of this bias was seen in the early part of the study. Before September 26, the
difference in total emissions was 47 1 percent, and after this date the difference was essentially
zero (-1.5 percent).
Despite the differences, the overall HVCS and EPA results correlate well (r2 = 0.897). A
regression plot and equation are shown in Figure 3-5. Bias as a function of leak rate is illustrated
in Figure 3-6. Aside from the relationship with leak size, no other operational relationships were
observed in the west Texas data. Delay times were minimal (most less than 2 hours), and there
was no dependence on component type or the use of a dilution probe. In west Texas, the dilution
probe was used for 27 of the 66 valid measurements. A more detailed discussion of the sources
of the HVCS bias in the field study data is given below (Section 3.3).
u
D
O
in
c
OJ
D_
LJ
O
1)
o:
in
O
CD
(D
1 .5
0.
-0.5
* r~- I
> 6 i
c f
(< 0.1 slpm) M (< 1 slpm) L (< 4 s I p m) X L v > 4 s I p rn)
EPA LeaK Rate
x
Figure 3-6. West Texas - Bias Dependence on Leak Rate
26
-------
3.3 SOURCES OF HVCS BIAS
Preliminary analysis of the south Texas data suggested the presence of a positive HVCS bias.
(6.6 percent "true" bias); however, the magnitude of the bias was not considered strongly
significant given the variability in the results, and the fact that the difference in total emissions
determined by the two methods was small. At the west Texas site, the HVCS bias was much
larger (33.8 percent "true" bias). This was noted immediately in the field. The field crew
conducted numerous quality control and operational checks, to determine the source of the bias.
The HVCS and bag sampling apparatus were carefully leak checked, and additional flow
calibrations were performed. The OVA was calibrated before and after each HVCS quantification
using the same methane standards as were used to calibrate the GC. Controlled leak tests, and
other special tests were also conducted in an attempt to isolate the source of the bias.
After efforts in the early part of the study failed to eliminate the bias, a concentrated effort was
made on September 26 to isolate and eliminate the bias, if possible. This included controlled leak
tests, equipment checks and cleaning, and minor changes in operating procedures. Early in the
day, a set of two controlled leak tests were conducted using the same equipment used in the
laboratory studies. These were "blind" tests, i.e., the HVCS and EPA method operators did not
know the actual leak rate produced. The known leak rates were 0.168 slpm (test 1) and 3.27
slpm (test 2). The HVCS results were 0.163 slpm (-3.0 percent) and 3.026 slpm (-7.5 percent),
respectively. The EPA method results were 0.095 slpm (-43.2 percent) and 1.879 slpm (-42.5
percent), respectively. These results focused attention on a possible negative bias in the EPA
method as the source of the apparent positive bias in the HVCS system.
In the EPA sampling system, samples were withdrawn from the main flow using a small
secondary sampling pump connected at a "tee" located about 3.5 inches upstream from the main
pump exhaust. The main pump was typically operated at 15 slpm and the smaller pump was
operated at about 1.5 to 2 slpm. Both pumps were diaphragm type. A hypothesis to explain a
negative bias in the EPA system was that the secondary pump may have been able to pull in
ambient air over a short distance against the main flow; especially if the exhaust pressure from
the main pump was weakened due to a defective or obstructed reed valve.
To test this hypothesis, sample concentrations were measured with and without a length of tubing
added to the exhaust. This test was based on the assumption that dilution air could not enter the
sample over the longer distance. An initial test was conducted in the field at a propane leak using
the OVA to obtain concentration measurements at the outlet of the EPA sample line. This test
showed an approximate twofold increase in sample concentration when the extension tube was
added. This effect was magnified at lower sample flows in the EPA system. After this test, it was
decided to clean the main pump to clear any possible obstruction. This was accomplished by
flushing a small quantity of water through the pump. The team then conducted similar tests in
the field laboratory using a conjrolled leak rate of pure methane. Sample concentrations were
measured directly with the OVA. In consecutive tests at the same leak rate conducted with and
without the extension tube, sample concentrations (read by the OVA) were consistent (5404 and
5143 ppm, respectively). This suggests that the extension tube alone was not responsible for the
results seen in the field. However, as a precautionary measure, an extension tube was used
during the remainder of the study. A third controlled leak test was then conducted. In this test,
the HVCS and EPA results agreed very well. The HVCS result was 0.156 slpm, and the EPA
result (without the extension tube) was 0.151 slpm, for an HVCS bias of +3.3 percent. A second
27
-------
EPA method quantification using the extension tube on the same controlled leak gave a slightly
lower result of 0.13 slpm.
The proper conclusion to be reached from this set of tests is confounded somewhat by the
differences in sample gas (propane in the field vs. methane in the laboratory) and by the fact that
the pump was cleaned before the laboratory tests were conducted. The extension tube had no
effect in the laboratory tests using methane and after the pump was cleaned, but appeared to
have some effect before the pump was cleaned and when the sample consisted primarily of
propane. The agreement between the HVCS and EPA results in the controlled leak tests
improved in the test conducted after the pump was cleaned compared to the tests conducted
earlier on the same day. This appears to suggest that cleaning the pump improved EPA method
performance; however, the mechanism for this improvement was not determined.
The effect of contamination in a diaphragm pump would be to prevent the reed valve from sealing
completely on the pump downstroke. Significant pump contamination or a leaking reed valve
would prevent the pump from functioning properly, and there would be no flow, or significantly
reduced flow. This has been noted in other studies using the same model of pump. However,
flows of 20 Ipm (which was the maximum effective system flow) were achieved before the pump
was cleaned. In addition, at normal flows of 15 slpm, it is difficult to conceive how the secondary
pump, operating at less than 2 slpm could pull in sufficient dilution air against the main flow to
produce the magnitude of bias observed. In many cases, this would require that one half or more
of the total sample consisted of dilution air. Thus, the explanation that cleaning the pump
eliminated a negative bias in the EPA method is not completely satisfactory.
Two additional possibilities that could explain the apparent HVCS bias were identified: (1) high
background hydrocarbon levels, and (2) analytical bias of the OVA versus the GC. Background
hydrocarbon concentrations can produce a positive HVCS bias since the background
concentration is multiplied by the higher HVCS flow rate to obtain the emission rate. That is, the
HVCS result is more strongly influenced by background levels than the EPA vacuum method,
even though both use the air surrounding the leaking component as dilution gas. Although both
the GC and the OVA use flame ionization detection (FID), an analytical bias could result from
differences in instrument design that make the OVA response sensitive to sample contaminants,
sample gas composition, and possibly environmental effects such as pressure. These
possibilities are discussed in more detail below (see sections 3.3.1, and 3.3.2, respectively).
In addition to the possible effect of operational changes that were implemented after September
26 (the pump was cleaned, and an extension tube was added to the EPA apparatus), the
improvement in HVCS results after September 26 may be related to a combination of two other
factors. First, as part of efforts to determine and eliminate the source of the bias, measuring in
areas with high background levels was avoided after that date. Earlier in the study, most of the
measurements were obtained from dense clusters of leaking components where there were
potentially high background levels. At the south Texas site, background levels were minimal
(samples were obtained in remote, open areas and wind speeds were very high). Second, the
average leak rate after September 26 (1.8 slpm) is larger than the average leak rate before that
date (1.2). This would reduce an HVCS bias related to background hydrocarbon since such a
bias is less significant for larger leaks.
28
-------
3.3.1 High Background Levels
Measured background levels averaged about 30 ppm in the plant area and levels as high as 1500
to 1750 ppm were detected in several instances. Near the well heads, in the gas fields,
background levels averaged 5 to 10 ppm. One can calculate the background level required to
produce the biases observed in the west Texas data. At the high end of HVCS flows typically
used in the study (about 125 slpm) a background concentration of 3,600 ppm would be required
to produce an HVCS bias of +45 percent at a leak rate of about 1.0 slpm (using an EPA flow of
15 slpm). These are typical values for the early part of the study. At lower HVCS flows,
significantly higher background levels would be required. Only for the smaller leaks (less than
0.1 slpm) were measured background levels high enough to produce the biases observed. In
addition, it can be demonstrated that results obtained at different HVCS flow rates (at the same
leak rate) do not generally differ by as much as would be expected if background interferences
were significantly affecting the results. Based on this, it is unlikely that high background levels
contributed significantly to the overall bias observed in the data for this study.
However, the issue of background interferences with the HVCS quantification is important.
Sufficiently high background hydrocarbon concentrations will interfere with HVCS leak rate
quantifications, producing a positive bias relative to "low flow" methods, or methods that use
hydrocarbon free dilution gas. This situation may not be avoidable in some circumstances, such
as when leaks are densely clustered, or are located near an emissions source. Methods should
be adopted for use in accounting and correcting for such interferences.
The simplest approach is to directly measure background levels using the OVA, and correct for
these concentrations in the HVCS sample flow. However, in this study it was found that this is
not always straightforward. Background measurements taken using the OVA during HVCS
quantifications often fluctuated wildly (especially in windy conditions, which predominated), and
it was not possible to obtain a representative reading. An alternative means for determining the
background level from the HVCS results themselves has also been proposed. The method
depends on the fact that the HVCS can conveniently measure the leak rate at more than one
flow. As the HVCS flow rate increases, the fraction of total hydrocarbon in the HVCS exhaust due
to the fugitive leak decreases. If the HVCS flow could be increased infinitely, the contribution of
the fugitive leak would be zero, but the contribution of the background concentration would remain
constant. The background concentration can be found by plotting a line from several HVCS leak
rate determinations where the x-axis is the reciprocal of HVCS flow and the y-axis is the HVCS
outlet concentration (in ppm). If this line is extended to zero, the intercept represents
concentration at infinite flow. A positive intercept is correctly interpreted as the background
concentration.
This method could produce a background level that is representative of actual background
hydrocarbon taken in during HVCS sampling, even under conditions where ambient
concentrations could not be reliably quantified by direct measurement. However, for the method
to give true results, one must be certain that changes in HVCS outlet concentration are due solely
to changes in HVCS flow, that is, that total leak capture is attained at all HVCS flows. When
there is incomplete leak capture, increases in the HVCS outlet concentration do not keep pace
with the decreases in HVCS flow. If this method is applied to such data, there will be an apparent
background concentration (positive intercept) that is really just an artifact of incomplete leak
capture.
29
-------
The prototype HVCS device used in this study was not generally able to maintain complete leak
capture at any but the highest flows (see laboratory study results). If this correction is applied
to data where there was not complete leak capture at all flows, a negative bias will be introduced.
For example, the method appears to correct for much of the positive bias in the west Texas data;
however, if the same method is applied to the south Texas data, a negative bias appears.
Because of such inconsistency, it was decided that this method could not be consistently applied
to correct the evaluation study results. However, the method might be successfully applied in
future work with an improved HVCS with greater flow capacity.
2.3.2 Analytical Bias
During the west Texas study, the field crew conducted a number of special tests designed to
identify the source of the HVCS bias. One set of special tests (conducted on September 24)
provided evidence of an analytical bias between the OVA response and the GC response to the
same samples. First, the concentration in an EPA method bag sample was measured directly
by the OVA and by GC. The OVA recorded 8828 ppm, and the GC recorded 6433 ppm, a +37
percent O /A bias. Second, a bag sample was collected from the HVCS exhaust. The OVA
recorded 10,372 ppm and the GC recorded 8229 ppm, a +26 percent OVA bias. Although both
the OVA and the GC use Flame lonization Detection (FID), there are two major differences in
instrument design that might be responsible for the observed bias.
First, the OVA uses sample air to supply oxygen to the FID flame, whereas the GC FID has an
independent air supply that provides a steadier, more controlled flame. FID response is very
sensitive to flame characteristics (e.g., temperature, fuel/air ratio). Contaminants in the sample
air or differences in sample gas composition could produce a bias in the OVA results. Second,
all of the components in the sample gas mixture are presented to the OVA FID simultaneously,
and the OVA response is to the total hydrocarbon present in the mixture. In the GC, the sample
is separated into its components by the GC column before reaching the detector, and the total
hydrocarbon concentration i-c arrived at by summing the FID response for each gas. If a
compound were present in the sample that was not detected in the GC run (e.g., because the run
was not long enough for the compound to elute from the column), a bias would be introduced.
However, since natural gas is made up primarily of light hydrocarbons (C1-C5), and these were
detected in the GC runs (see Section 3.4 - Gas Composition), the "excess" compound could be
present only at low concentration, and would have to elicit a very large OVA FID response
(relative to methane) in order to produce a significant bias.
After the field study, it was determined that there was a need for additional laboratory testing to
further investigate the source of bias in the west Texas results. Five gas samples were obtained
in pressurized stainless steel sample canisters from the west Texas plant. The samples were
collected progressively through the gas processing stages and represent the areas in the gas
plant where leak quantifications were made, in addition, the same OVA used in the study was
obtained for comparative tests with the GC. The laboratory testing was designed to identify a
contaminant in the gas samples, or an unidentified compound that could produce the observed
bias.
One possible contaminant that could produce a positive OVA bias is hydrogen gas (H2). There
is potential for excess H2 in the west Texas sample gas occurring as a byproduct of H^S removal.
A simple laboratory test was conducted to determine wnether this could explain the bias in the
30
-------
west Texas results. Hydrogen gas was introduced incrementally into steady flows containing
approximately 100 and 1000 ppm methane. This test showed that it would take an H2
concentration of more than 5 percent (50,000 ppm) to cause a significant change in OVA
response (more than a few percent). GC analysis of pure gas samples from the field showed that
H2 levels do not begin to approach such high values, so this explanation can be eliminated.
While it is possible that some other contaminant might act to alter OVA FID response, none has
been identified.
The second set of tests consisted of conducting long GC runs on each of the 5 pure gas samples
to allow time for any heavier hydrocarbons to elute. Also, since these tests were conducted on
undiluted samples, there is a greater likelihood of detecting "trace" compounds that could go
undetected in the diluted field samples collected using the EPA bag sampling method. The pure
gas samples showed roughly the same proportions of light hydrocarbons as the field samples.
While a few unidentified compounds were noted on the chromatograms for the pure gas samples,
FID response to these compounds was not significant relative to the response to the known light
hydrocarbons.
Finally, the pure gas samples were diluted to approximately 1000 ppm in room air, and the diluted
samples were analyzed directly by GC and by the OVA. The relative proportions of light
hydrocarbons in the diluted samples were nearly the same as in the pure gas samples, indicating
that the diluted samples were thoroughly mixed. These tests failed to reproduce the positive bias
noted in similar tests conducted in the field. In fact, the OVA gave negatively biased results in
the laboratory tests (this was traced to poor OVA response to higher molecular weight
hydrocarbons).
In summary, no evidence was found in the laboratory tests to confirm a positive analytical bias
of the OVA versus the GC. In addition, no contaminants or excess compounds were identified
that could have produced such an analytical bias. The only identified factors that were not
investigated in these tests are environmental factors, such as the difference in atmospheric
pressure due to the high altitude of the west Texas site (about 3500 feet above mean sea level
compared to near sea level elevations for the laboratory and south Texas studies). While the
OVA response is known to be sensitive to sample inlet pressure, this should not affect the results
since the OVA was calibrated at the pressure at which it was used.
3.4 GAS COMPOSITION
Gas composition (C1 through C6 alkanes and alkenes, carbon dioxide, oxygen, and nitrogen) was
determined for each bag sample. Two samples were analyzed for each component tested. The
percentage difference between the two bags averaged 0.24 ą0.07 percent for the TCD, and 0.33
ą0.03 percent for the FID (west and south Texas combined). Only methane, ethane, propane,
butene, butane, pentene, and pentene were detected in more than trace amounts. Results are
summarized in Table 3-2. In the Table, results are given as percentages of total hydrocarbon,
where the total hydrocarbon is simply the sum of GC response to each compound, calibrated as
methane. In most cases, methane makes up more than 85 percent of total hydrocarbon. Ethane
is second, and propane is third. A few samples collected in a propane storage area at the west
Texas site are predominately propane. In most cases only trace amounts of C4 and heavier
31
-------
hydrocarbons were present in leaking gas. This is consistent with the species composition of
emissions from the gas and oil industry as identified by EPA (EPA 1988).
Table 3-2. Summary of Gas Composition (percent)
Methane
Ethane
Propane
Butane
Butene
Pentane
Pentene
South Texas
Average
Max
Min
88.1
97.7
7.6
5.7
8.5
2
4.2
87
0
0.7
2
0
0.7
2
0
0.2
0.8
0
0.3
2.3
0
West Texas
Average
Max
Min
84.2
97.8
0.3
5
15
1.7
10.4
92.2
0
0.2
2.3
0
0.2
1.3
0
0
1.2
0
0
0.6
0
3.5 COMPARISON OF DIRECT MEASUREMENT RESULTS WITH EPA EMISSION
FACTORS
It is interesting to compare the directly measured leak rates at each field site with the emissions
computed using the EPA protocol emission factors. The EPA protocol gives several methods for
estimating emissions from equipment leaks including average emission factors that are applied
to each component inventoried, and leak/no leak emission factors that are applied separately to
non-leaking and leaking components as designated using Method 21. In the field studies, most
of the components identified as leaking (using soap solution) had screening values of 10,000 ppm
or more. For the purposes of this exercise, all leaking components are considered to have
screening values of at least 10,000 ppm, which is the definition of a leak for purposes of applying
the leak/no leak emission factors. Table 3-3 summarizes total emissions for directly measured
(EPA and HVCS methods), EPA average emissions factors, and EPA leak/no leak emission
factors. The EPA protocol correlation approach is not applicable here since direct screening
values were not obtained for small leaks that did not produce observable soap bubbles, and those
leaks with observable bubbles generally "pegged" the OVA at 10,000 ppmv. EPA has recently
published interim emission factors and correlation equations for gas and oil production (EPA
1995). These new values are acceptable for immediate use for estimating emissions from leaking
equipment; but may change based on additional input from industry and state and local air
pollution control agencies. Table 3-4 also includes emissions calculated using these new factors.
It is apparent that the use of the original EPA protocol average emission factors can yield a
dramatic overestimate of actual emissions. The 1993 protocol leak/no leak emission factors
improve the estimate, but are still quite conservative. The new interim leak/no leak emission
factors appear to yield very reasonable emissions estimates, compared to measured values.
32
-------
Table 3-4. Comparison of Emissions Estimates by Different Methods
Screening Information
Component Type
Flange
Connector
Valve
Open Line
Pressure Relief
Misc.
Total
South Texas
Screenec
889
4,664
1,764
216
21
74
7.628
No Leak
889
4,647
1,741
208
21
66
7,572
Leak
0
17
23
8
0
8
56
/Vest Texas
Screenea
1,401
8,992
2,897
55
4
94
13,443
No Leaf
1,400
8,985
2,833
55
3
88
13,364
Leal
7
64
0
6
79
Direct Measured Emissions
Component Type
Flange
Connector
Valve
Open Line
Pressure Relief
Misc.
Total (kg/hr)
Total (Ib/day)
South Texas
EPA Em. (kg/hr)
0.00
0.83
1.11
0.06
0.00
0.75
2.75
145.05
HVCS Em.(kg/hr;
0.00
0.88
1.22
0.13
0.00
1.05
3.28
173.37
/Vest Texas
EPA Em. (kg/hr)
0.02
0.09
4.28
0.00
0.17
0.64
5.20
27432
HVCS Em.(kg/hr
0.02
0.13
579
0.00
0.12
0.81
6.87
362.72
1993 Protocol Average Emissions Factors
Component Type
Flange
Connector
Valve
Open Line
Pressure Relief
Misc. (as Valve)
Total (kg/hr)
Total (Ib/day)
South Texas
Factor (kg/hr)
0.0011
0.0011
0.0200
0.0220
0.1880
0.0200
Em. (kg/hr;
0.98
5.13
35.28
4.75
3.95
1.48
51.57
2722.81
/Vest Texas
Factor (kg/hr)
0.0011
0.0011
0.0200
0.0220
0.1880
0.0200
Em. (kg/hr
1.54
9.89
57.94
1.21
0.75
1.88
73.21
3865.72
1993 Protocol Leak/No Leak Emissions
Component Type
Flange
Connector
Valve
Open Line
Pressure Relief
Misc. (as Valve)
Total (kg/hr)
Totalllb/day)
South Texas
No Leak EF (kg/hr)
6.0E-05
6.0E-05
2.9E-03
1 .5E-03
4.5E-02
2.9E-03
Leak EF (kg/hr)
0.03360
0.03360
0.09800
0.17400
0.86300
0.09800
Em. (kg/hr)
0.05
0.85
7.30
1.70
0.94
0.98
11.82
62433
/Vest Texas
No Leak EF (kg/hr)
6.0E-05
6.0E-05
2.9E-03
1 .5E-03
4.5E-02
2.9E-03
Leak EF (kg/hr)
0.03360
0.03360
0.09800
0.17400
0.86300
0.09800
Em. (kg/hr;
0.12
0.77
14.49
0.08
1.00
0.84
17.30
913.57
1995 Interim Leak/No Leak Emission Factors
Component Type
Flange
Connector
Valve
Open Line
Pressure Relief
Misc.
Total (kg/hr)
Total (Ib/day)
South Texas
No Leak EF (kg/hr)
3.1E-07
7.5E-06
7.8E-06
2.0E-06
4.0E-06
4.0E-06
Leak (EF (kg/hr)
0.08500
0.02800
0.06400
0.03000
0.07300
0.07300
Em. (kg/hr;
0.00
0.51
1.49
0.24
0.00
0.58
2.82
148.97
/Vest Texas
No Leak EF (kg/hr)
3.1E-07
7.5E-06
7 8E-06
2.0E-06
4.0E-06
4.0E-06
Leak (EF (kg/hr)
0.08500
0.02800
0.06400
0.03000
0 07300
0.07300
Em. (kg/hr;
0.09
0.26
4.12
0.00
0.07
0.44
4.98
262.86
1995 Interim Correlation Emission Factors
Component Type
Flange
Connector
Valve
Open Line
Pressure Relief
Misc.
Total (kg/hr)
"otal (Ib/dav)
South Texas
Default Zero (kg/hr)
3.1E-07
7.5E-06
7.8E-06
2.0E-06
4.0E-06
4.0E-06
Corr. EF (kg/hr)'
0.00290
0.00130
0.00220
0.00130
0.00300
0.00300
Em. (kg/hr)
0.00
0.06
0.06
0.01
0.00
0.02
0.16
8.45
/Vest Texas
Default Zero (kg/hr)
3.1E-07
7.5E-06
7.8E-06
2.0E-06
4.0E-06
4.0E-06
Corr. EF (kg/hr)*
0.00290
0.00130
0.00220
0.00130
0.00300
0.00300
Em. (kg/hr
000
0.08
0.16
0.00
0.00
0.02
0.26
13.95
Notes:
'All Screening Values Taken as 10,000 ppmv.
Assume Wt. Fraction VOC is 100 percent ( ncl. methane)
Direct Measured Emissions Given for Operationally Valid Measurements Only
33
-------
SECTION 4
DATA QUALITY
Sampling, analytical, and quality control procedures were carried out as specified in the QAPjP
that was prepared and approved for this AEERL QA Category III project. Overall data quality was
very high with precision and accuracy for all primary measurements remaining within data quality
objectives throughout the study. Data capture and representativeness are discussed in detail in
Sections 2 and 3 of this report. Overall precision and accuracy for the EPA and HVCS methods
was determined from the laboratory studies and is discussed in detail in Section 2. Controlled
leak tests were also conducted in the field; however, the laboratory results give a much better
representation of overall data quality for the two methods since a much larger number of
measurements are available. In the west Texas study, replicate HVCS quantifications were made
for 16 leaks. The average difference in the repeat measurements was -4.5 ą7.7 percent.
Table 4-1 summarizes precision and accuracy for the concentration measurements (OVA and GC)
as determined from calibration data. Precision and accuracy for the OVA are calculated as the
arithmetic mean and standard deviation of the percentage differences between consecutive
calibrations. The OVA was calibrated before and after each HVCS quantification, so drift was
minimized. Precision and accuracy for the GC is presented for each detector, per Method 18,
as the average and standard deviation of percent differences between upscale calibration points
and linear fit. Injection precision was also calculated as the standard deviation of percentage
differences between duplicate injections for each sample. For the laboratory study, injection
precision was 0.81% (TCD) and 0.26% (FID). For the field studies (combined), injection precision
was 0.21% (TCD), and 0.08% (FID).
Table 4-1. Concentration Measurements - Data Quality
Measurement
Portable HC Analyzer OVA - Lab
Portable HC Analyzer OVA -
South Texas
Portable HC Analyzer OVA -
West Texas
Gas Chromatography - Lab
Gas Chromatography South
Texas
Gas Chromatography West
Texas
Precision (%)
DQO
15
15
15
5
5
5
Actual
5.6
10.3
11.6
1.78 (TCD)
1.30 (FID)
1.80 (TCD)
3.60 (FID)
3.60 (TCD)
2.20 (FID)
Accuracy (%)
DQO
15
15
15
10
10
10
Actual
0.1
0.2
-1.2
+0.11 (TCD)
-0.21 (FID)
0.22 (TCD)
-0.19 (FID)
-0.71 (TCD)
-0.62 (FID)
34
-------
The rotameters used to determine flows for the EPA vacuum bag method and the HVCS were
calibrated at the start of the laboratory study and checked during the south and west Texas field
studies. The large HVCS rotameters were calibrated during the laboratory study against an EPA
primary standard spirometer. In the field, the large HVCS rotameters were checked against a dry
gas meter traceable to the EPA spirometer. The HVCS small rotameter and the EPA vacuum bag
rotameter were calibrated against a primary standard bubble meter. Good linearity (r2 > 0.95) was
obtained for all rotameter calibrations. Actual calibration slopes and intercepts were used to
determine the correct flow rates from the direct rotameter readings. The average and standard
deviation of differences from the regression value (curve fit) at calibration set points can be used
as an indicator of the accuracy and precision of the rotameter flows. For the bag cart rotameter
(2-20 slpm), the accuracy was +0.38 percent and the precision was 5.28 percent. The QAPjP
data quality objectives were 10 percent for precision and 10 percent for accuracy. For the large
HVCS rotameter (50 to 500 slpm), the accuracy was -0.02 percent, and the precision was 1.40
percent. For the medium HVCS rotameter (5 to 25 slpm), the accuracy was +0.53 percent, and
the precision was 0.09 percent. For the small HVCS rotameter (1 to 10 slpm), the accuracy was
+0.09 percent, and the precision was 2.80 percent. HVCS flows were always read from a single
rotameter. The data quality objectives for the HVCS flows were 5 percent for precision and 5
percent for accuracy. On-site temperature and pressure obtained at the time of the
measurements were used to correct flows to standard conditions. This was important in south
Texas as ambient temperatures were very high (as high as 114 °F, or 46 °C), and in west Texas
which was at a high elevation (ambient pressures of about 680 mm Hg, or 90.65 kPa).
No external audits were planned or performed for this study.
As discussed earlier, the prototype HVCS was generally unable to maintain complete leak
capture, except at the highest flows. Typically, the measured leak rate would begin to fall off after
the first or second determination at decreasing HVCS flow rates. Ideally, the average of 3 HVCS
quantifications at different flows should be used to determine the leak rate. Since this was not
always possible, it was necessary to adopt a procedure to recover useful data in instances where
marginal leak capture occurred at lower flows. A decision procedure was developed and applied
to the field study results in cases where incomplete leak capture may have occurred. The
procedure used is as follows.
At least three HVCS leak quantifications were made for each component. The three
measurements corresponding to the highest HVCS flows were used as initial inputs to a
spreadsheet that was used to automate repetitive calculations. At this point, one of two tests was
applied depending on the apparent size of the leak (average of the three measurements). For
leaks larger than 0.15 slpm (about 0.3 Ib/day), the maximum percentage difference relative to the
mean of the 3 measurements had to be within ą20 percent. For leaks smaller than 0.15 slpm,
the maximum absolute difference from the mean had to be smaller than 0.03 slpm. If the test
was passed, the average of the three determinations was taken as the valid HVCS result.
If the test was failed, the leak determination corresponding to the highest flow was substituted for
the determination that gave results furthest from the mean (generally the lowest HVCS flow).
That is, the leak determination corresponding to the highest HVCS flow was used twice in the
average, weighting the average to the highest flow and greatest leak capture. HVCS results
where this occurred were flagged, but considered valid. If the test was failed again, the highest
HVCS flow alone was used to obtain the HVCS result. Such results were flagged as suspect
35
-------
because, since only a single point was used, there was no verification of leak capture. The dual
test was necessary to avoid excluding data for small leaks where absolute differences were small,
but percentage differences were large.
It should be emphasized that this procedure was used only as a means to recover data in
instances where the flow capacity of the prototype HVCS was insufficient to fully capture the leak.
The flow capacity of the HVCS system should be improved before the system is used in future
work. The cut-off value of 0.03 slpm was chosen as a minimum leak rate of interest, since
previous testing has shown that essentially 100 percent of emissions are due to much larger leaks
(API 1993). The 20 percent test criteria was adopted, based on the laboratory study findings, as
a practical value that could generally be attained by the prototype HVCS, but would still yield
useful results. While these test criteria were reasonable for the test data set with the prototype
HVCS, more stringent criteria should be applied in future work with an improved system.
36
-------
SECTION 5
CONCLUSIONS
As demonstrated in the laboratory study and field study results, the HVCS is capable of
accurately quantifying fugitive leaks over a wide range of leak sizes, and component types and
sizes.
Laboratory evaluation of HVCS performance was very favorable. The HVCS results
closely matched EPA method results with a difference in total measured emissions of only
about 3 percent. The HVCS also reproduced a wide range of known leak rates with an
average bias of -8.3 percent. The negative bias is probably due to incomplete leak
capture. In the laboratory tests, HVCS leak capture depended solely on the ability of the
HVCS to capture all of the leaking gas in the slip stream of dilution air entering the HVCS
inlet. No enclosures were constructed to shield components and direct gas into the HVCS
inlet.
The HVCS also performed very well in the south Texas field study. The HVCS matched
the EPA estimate of total facility emissions within about 4 percent, similar performance to
that obtained in the laboratory studies. In the field, enclosures were constructed to shield
components from wind and assist in directing leaking gas into the HVCS inlet.
During the early part of the west Texas study, an apparent positive bias was observed in
the HVCS results. On and before September 26 (about midway through the study), the
HVCS measured 47.1 percent more total emissions than the EPA method. After this date,
no appreciable bias was observed. Over the entire study, the HVCS measured 18.6
percent more emissions from the facility than the EPA method. The source of the bias
in the early part of the study is unclear; however, the results suggest that there may have
been some operational problem that was overcome as a result of efforts undertaken in the
field. Other factors that may have contributed to the change in results include; (1) efforts
that were made to avoid sampling in areas with potentially high background concentrations
that could cause a positive bias in the HVCS results, and (2) the fact that average leak
rates were higher in the later part of the study, which would tend to lessen the effect of
background interference on the HVCS quantifications (see section 3.3).
Overall, these results are within acceptable limits for field emissions measurements. With some
physical and procedural enhancements, the HVCS should offer an acceptable alternative to the
EPA protocol bagging method with GC analysis.
Special precautions must be taken to obtain accurate HVCS quantifications in areas where
they may be elevated background concentrations, such as in confined areas, or where
there are dense clusters of leaking components or very large leaks. The simplest
approach is to attempt to quantify background levels with the OVA, and apply an
appropriate correction to the results. This must be done very carefully since background
levels in such areas have been observed to range widely in small areas and change very
rapidly. An alternative method for determining the background level has been suggested
that, in some instances, could provide a useable correction, even when background levels
cannot be practically measured (see Section 3.3). The limitation of this method is that one
37
-------
must be certain that changes in HVCS outlet concentration are due solely to changes in
HVCS flow. That is, that total leak capture must be attained at all HVCS flows. A third,
though more difficult to execute, alternative would be to supply the HVCS with
hydrocarbon free dilution air.
Improved HVCS flow capacity, control, and metering are needed to enhance leak capture
and provide greater reliability and ease of use in the field. With the current rotameter set-
up, the capacity could be doubled by simply increasing pump capacity. Power
requirements would also be increased, but the unit could still be battery operated (a 12
V pump could reach near 1000 scfh, or about 470 slpm). Much larger flows would require
more power, decreasing portability, and the metering system would also have to be
modified substantially to handle the higher flows. Increased flow capacity would also
increase the size of leaks that could be quantified without the need for a dilution probe,
or other alternative to extend the range of the portable hydrocarbon monitor. Enhanced
leak capture might also make it possible to measure leaks from larger components without
the need for auxiliary bagging. This could decrease the time required for each
measurement.
Increased range and enhanced stability of the portable hydrocarbon monitoring device
used with the HVCS are also needed. The portable hydrocarbon monitor used with the
HVCS needs greater range and reliability than the Foxboro OVA Model 108 that is
currently used. The OVA's upper range is at 10,000 ppm, or 1 percent. This can be
extended to perhaps 15,000 ppm using the direct voltage output from the OVA; however,
precision rapidly deteriorates at this upper end. In the field, Foxboro's dilution probe was
used to extend the quantification range, with generally good results; however, the use of
the dilution probe adds a degree of complexity. The OVA is also very sensitive to
sampling conditions, contaminants, battery levels, and other factors. The OVA requires
frequent calibration, which adds significantly to the expense and level of uncertainty
associated with its use in quantitative applications. The OVA is very sensitive to sample
gas composition since the detector is exposed to the entire sample stream at once and
uses sample stream air as combustion air for the FID. The OVA exhibits varying
responses to different hydrocarbons, and sometimes radical responses to "contaminants"
(water, dust, and presence of gases, e.g., excess hydrogen, that affect the response of
the FID). Some research is needed to identify and test alternative analyzers with greater
range and stability than the OVA. This might include infrared devices and electrochemical
sensors. In addition, there has been some preliminary development of a catalytic
combustorthat would determine HC concentration by stoichiometry, using measurements
of O2 and CO2 at the entrance and exit from the combustor.
The HVCS may have a very significant role to play in applications where rapid, cost effective, on-
site leak quantifications are important.
With the HVCS, a single operator can quantify approximately 30 leaks per day. With the
EPA bagging method, approximately 10 leaks can be sampled per day with additional time
and expense required for GC analysis.
38
-------
The HVCS could be very useful in evaluating the effectiveness of different inspection and
maintenance programs, and determining the most cost effective approach for maintaining
emissions below a given level.
The HVCS would also be useful for emissions inventory and compliance testing activities
related to Federal and State air permit requirements.
In addition, the HVCS may be valuable in evaluating the performance of optical sensing
based techniques for determining fugitive emission rates under real world conditions.
Such methods are currently under development at EPA, and may soon be tested under
actual site conditions. For these tests, there will be a need to independently determine
leak rates from multiple fugitive sources as a basis for evaluating the performance of the
optical sensing based methods.
39
-------
SECTION 6
REFERENCES
American Petroleum Institute, Fugitive Hydrocarbon Emissions from Oil and Gas Production
Operations, API Publication Number 4589, December 1993.
Gilbert, Richard O., Statistical Methods for Environmental Pollution Monitoring, Van Nostrand
Reinhold, New York, 1987
Hausle, K.J., Protocol for Equipment Leak Emission Estimates, EPA-453/R-93-026 (NTIS PB93-
229219), June 1993.
Provost, L.P., et al. /Assessment of Atmospheric Emissions from Petroleum Refining, Vol. 4,
EPA-600/2-80-075d (NTIS PB81-103830), 1980.
Shareef, G.S., et al., Air Emissions Species Manual - Vol. I, Volatile Organic Compound (VOC)
Species Profiles, EPA-450/2-88-003a (NTIS PB88-225792), 1988.
U.S. Environmental Protection Agency, New Equipment Leak Emissions Factors for Petroleum
Refineries, Gasoline Marketing, and Oil and Gas Production Operations, Available on EPA
Technology Transfer Network, CHIEF Electronic Bulletin Board System. February 1995.
40
-------
SECTION 7
DATA TABLES
41
-------
Laboratory Study Data
Method
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
ID pate Component Leak Confiquration Leak Character Pressure (osi) LeakRate Result Flaq Pias
01B
002a
003a
003b
003c
04a
04b
04c
05a
07a
07c
07d
07e
08a
08b
08c
08e
08h
09a
09b
09c
10a
10b
10c
10d
10e
10f
11a
11b
11c
11d
12a 1
14a
14a 1
14b
14c
14c 1
13a
13b
13c
13c 1
13d
13e
13e 1
13f
13g
07/01
07/05
07/05
07/06
07/06
07/06
07/06
07/06
07/07
07/08
07/08
07/08
07/08
07/11
07/11
07/11
07/11
07/12
07/12
07/12
07/12
07/12
07/12
07/12
07/12
07/12
07/12
07/13
07/13
07/13
07/13
07/13
07/13
07/13
07/13
07/13
07/13
07/14
07/14
07/14
07/14
07/14
07/14
07/14
07/14
07/14
2" Gate Valve
2" Gate Valve
3ump Body
Pump Body
Dump Body
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
3ump Body
2" Open Ended Line
2" Open Ended Line
5" Pipe Flange
5" Pipe Flange
5" Pipe Flange
5" Pipe Flange
3" Pipe Flange
Pump Body
Pump Body
Dump Body
4" Threaded coupling
4" Threaded coupling
4" Threaded coupling
4" Threaded coupling
4" Threaded coupling
4" Threaded coupling
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
Pump Body
2" Open Ended Line
2" Open Ended Line
2" Open Ended Line
Dump Body
oose body nut
oose body nut
Dack plat loose
jack plat loose
sack plat loose
no sealant on threads
no sealant on threads
no sealant on threads
no sealant on threads
hand tight
oose back plate
NA
NA
but gasket
cut gasket
cut gasket
gasket damaged - groove
oose bolts
damaged shaft seal
damaged shaft seal
damaged shaft seal
10 sealant
10 sealant
10 sealant
10 sealant
10 sealant
10 sealant
oose body nut
oose body nut
oose body nut
loose body nut
loose body nut
oose body nut
oose body nut
oose body nut
oose body nut
oose body nut
oose body nut
oose body nut
oose body nut
oose bodyjiut
damaged shaft seal
open line beyond gate valv
open line beyond gate valv
open line beyond gate valv
damaged shaft seal
2 pin leaks
diffuse
NA
diffuse
diffuse
diffuse
diffuse
diffuse
multiple jets - high pressur
ow flow, low pressure
ow flow, low pressure
ow flow, low pressure
ow flow
et
et
et
et
Droad, diffuse
Droad leak around shaft
Droad leak around shaft
Droad jets -
several point sources
nultiple point sources
several jet leaks
nultiple pin leaks
nultiple pinhole leaks
nultiple pin leaks
Droad & diffuse
Droad & diffuse
Droad & diffuse
narrow jet, high velocity
narrow jet, high velocity
NA
NA
NA
NA
NA
broad, low velocity
broad, low velocity
broad, low velocity
jroad, low velocity
broad, low velocity
ow velocity
ow velocity
ow velocity
broad
10.90
2320
830
595
1.60
2980
11.75
4.00
94.15
0.00
0.00
0.00
0.00
000
790
26.75
51.70
000
2035
24.55
42.25
3900
10.45
19.30
45.20
2340
1200
40.40
47 10
4320
4675
4705
0.00
000
000
000
000
000
3685
0.10
010
650
0.00
000
010
695
0.25
2.772
1.667
1.007
0.186
1.317
0.599
0.242
6.068
0.04
0.041
0.042
0.751
0.042
0.748
2.012
3.864
0.107
6.628
8.579
20.856
18.633
5.065
9.166
8.016
3.79
1.79
15.833
11.777
5.221
2.514
2.761
0.761
0.761
0.305
0.045
0.045
0.73
0.466
4.557
4.557
1.183
3.023
3.023
0.953
1.185
0.26
3.04
1.87
1.10
0.16
1.25
0.52
0.18
5.18
0.03
0.03
0.02
0.78
0.03
0.69
1.89
4.18
0.07
7.00
8.67
21.71
16.27
5.52
9.06
6.83
3.10
1.58
11.92
10.86
3.77
2.00
2.47
0.63
0.57
0.22
0.03
0.03
0.70
0.45
5.69
3.72
1.11
3.08
3.18
1.11
1.26
Suspect
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Suspect
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
4.83%
9.59%
12.00%
9 40%
-11 25%
-5.32%
-12.47%
-26 88%
-14.57%
-16.16%
-14.09%
-47 38%
3.83%
-22.98%
-7.77%
-5.91%
8.27%
-31.38%
5.56%
1 .02%
4 12%
-12.68%
8 92%
-1.14%
-14.79%
-18.27%
-11.57%
-24.72%
-7.77%
-2781%
-20.59%
-10.45%
-17.37%
-25.72%
-28.38%
-33.65%
-38.90%
-363%
-3.60%
24.90%
-1832%
-6.08%
2.02%
5 06%
16.01%
6.56%
-------
00
Laboratory Study Data
Method
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
HVCS
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
ID
17a
17b
17c
17d
17e
18a
18b
18c 1
18d 1
OOC
OOF
OOG
01A
002b
01C
003a
003b
003c
04a
04b
04c
05a
06a
07a
07b
07c
07e
17a
17b
17c
17d
17e
18a
18b
18c
18d
18e
18f
19c
19d
20a
20b
20c
21b
20d
)ate
07/19
07/19
07/19
07/19
07/19
07/20
07/20
07/20
07/20
06/28
06/28
06/29
07/01
07/05
07/05
07/05
07/06
07/06
07/06
07/06
07/06
07/07
07/07
07/08
07/08
07/08
07/08
07/19
07/19
07/19
07/19
07/19
07/20
07/20
07/20
07/20
07/21
07/21
07/22
07/22
07/26
07/26
07/26
07/27
07/27
Component
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
2" Gate Valve
2" Gate Valve
Pump Body
Pump Body
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
2" Gate Valve
Pump Body
Pump Body
3ump Body
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
2" Gate Valve
Pump Body
2" Open Ended Line
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
4" Threaded Coupling
2" Gate Valve
2" Gate Valve
3ump Body
3ump Body
Pump BOdy
3ump Body
2" open end line
2" open end line
3" Flange __,
5" Flange
3" Flange
2" open end line
3" Flange
.eak Configuration
10 thread sealant
no thread sealant
10 thread sealant
10 thread sealant
10 thread sealant
oose body nut
oose body nut
damaged shaft seal
damaged shaft seal
Loose Body Nut
Loose Body Nut
Loose Body Nut
Loose Body Nut
Loose Body Nut
Loose Body Nut
Back Plat Loose
Back Plat Loose
Back Plat Loose
No sealant on threads
No sealant on threads
No sealant on threads
No sealant on threads
Hand Tight
Hand Tight
Loose Body Nut
Loose Back Plate
NA
Mo Thread Sealant
Mo Thread Sealant
Mo Thread Sealant
Mo Thread Sealant
No Thread Sealant
Loose Body Nut
.oose Body Nut
Damaged shaft seal
Damaged shaft seal
Damaged shaft seal
Damaged shaft seal
:ut gasket
:ut gasket
:u( gasket
;ut gasket
.eak Character
diffuse
diffuse
diffuse
diffuse
diffuse
et
et
et
et
Diffuse
Diffuse
2 small jets on opposite sid
2 pin leaks at 10 & 2 o'cloc
2 pin holes, 10 and 2 o:clo
Diffuse
NA
Diffuse
Diffuse
Diffuse
Diffuse
Diffuse
Multiple Jets - High Pressu
NA
ow flow, low pressure
ow flow, low pressure
ow flow, low pressure
owflow
Diffuse
Diffuse
Diffuse
Diffuse
Diffuse
Jet
Jet
Jet
Jet
Jet
Diffuse
Diffuse
Diffuse
Diffuse
Diffuse
Diffuse
diffuse
Diffuse
'ressure (psi)
8.80
1.00
0.40
0.75
0.80
60.05
39.05
17.20
15.05
0.00
0.00
98.60
10.90
32.00
23.20
8.30
5.95
1.60
29.80
11.75
4.00
94.15
17.20
0.00
0.00
0.00
0.00
8.80
1.00
0.40
0.75
0.80
60.05
39.05
17.20
1505
11.40
0.95
0.00
0.00
000
0.00
0.00
000
0.00
.eakRate
1.304
0.158
0.076
0.112
0.147
2.275
1.336
3.634
2.054
0.687
3.048
1.044
0.25
4.807
2.772
1.667
1.007
0.186
1.317
0.599
0.242
6.068
10.278
0.04
0.04
0.041
0.751
1.304
0.158
0.076
0.112
0.147
2.275
1.336
3.634
2.054
6.254
0.149
2.122
4.692
4.756
2.959
1.161
3.044
0.14
Result
1.37
0.14
0.07
0.09
0.12
2.08
1.15
3.62
2.03
0.64
2.89
1.12
0.27
4.01
2.27
1.60
1.06
0.18
1.29
0.52
0.21
6.19
10.47
0.04
0.04
0.04
0.71
1.31
0.16
0.07
0.09
0.14
1.67
1.04
3.46
2.03
5.64
0.14
2.10
4.29
4.82
2.67
1.08
2.92
0.13
:lag
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Suspect
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Suspect
Valid
Valid
Valid
Valid
Valid
Suspect
Valid
Valid
Valid
Valid
Valid
Valid
Suspect
Valid
Valid
Valid
Valid ,
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Suspect
Bias
4.92%
-9.43%
-13.61%
-17.04%
-17.77%
-8.69%
-13.91%
-0.33%
-1.33%
-6.84%
-5.16%
6.91%
8.13%
-16.48%
-17.98%
-3.78%
5.37%
-3.23%
-1 .88%
-14.02%
-11.91%
2.09%
1.89%
0.85%
9.51%
3.00%
-5.82%
0.76%
-0.92%
-4.08%
-16.07%
-3.16%
-26.62%
-22.39%
-4.89%
-1.21%
-9.74%
-4.99%
-1.16%
-8 52%
1.34%
-9.68%
-7.11%
-4.09%
-9.29%
-------
Laboratory Study Data
Method
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
EPA
ID
21a
22a
|22c
^2b
22d
22e
22f
24a
24b
24c
24d
24e
24f
25b
25c
25d
26a
26b
26c
Date
07/27
07/27
07/27
07/27
07/29
08/01
08/01
08/09
08/09
08/09
08/09
08/09!
08/09
08/10
08/10
08/10
08/10
08/10
08/10
Component
2" open end line
2" inch gate valve
2" inch gate valve
2" inch gate valve
2" inch gate valve
2" inch gate valve
2" inch gate valve
5" Flange
5" Flange
3" Flange
5" Flange
5" Flange
5" Flange
2" Open Line
2" Open Line
2" Open Line
4" threaded coupling
4" threaded coupling
4" threaded coupling
Leak Configuration
oose body nut
oose body nut
oose body nut
oose body nut
oose body nut
oose body nut
Cut Gasket
Cut Gasket
Cut Gasket
Cut Gasket
Cut Gasket
Cut Gasket
10 sealant
10 sealant
no sealant
Leak Character
Diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
diffuse
'ressure (psi)^
000
0.00
000
000
040
030
2420
000
000
0.00
000
000
0.00
000
000
000
000
000
000
ueakRate
1.24
0.143
2.909
1.341
8.641
9.031
1.092
0.049
0.05
0.03
0.029
0.031
0.03
0.033
0.044
0.05
0.05
0.044
0.044
Result
1.22
0.14
2.54
1.26
7.28
8.97
1.31
0.04
0.04
0.03
0.03
0.03
0.03
0.03
0.04
0.04
0.04
0.03
0.03
Flaq
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Valid
Suspect
Valid
Valid
Valid
Valid
Suspect
Valid
Valid
Valid
Valid
Bias
-1 84%
-2.85%
-1270%
-5 76%
-15.81%
-0.71%
19.72%
-17.52%
-1026%
-1227%
-12.93%
-1729%
-11.84%
-1231%
-5 96%
-1282%
-27.51%
-3307%
-32.54%
-------
South Texas Data
ID
RP111
RP111A
RP112
RP112A
RP113
RP113A
TEST1
RP121A
RP161
EB161
EB162
EB163
TW211
TW212
TW213
TW214
TW215
TW216
TW217
TW218
TW2110
TW31
TW32
TW341
TW341A
TW342
TW343
TW344
TW345
TW346
TW162
TW41
TW42
TW43
TW81
TW271
TW301
EC22
CNG471
TW321
CNG641
CNG621
CNG622
CNG611
CNG461
RP122
TEST2
TW321A
Date Component Tvoe HVCS (slom) EPA (slpm) Absolute Bias (slpm) Relative Bias (%) True Bias (%)
08/23
08/23
08/23
08/23
08/23
08/23
08/23
08/23
08/23
08/24
08/24
08/24
08/24
08/24
08/24
08/25
08/25
08/25
08/25
08/25
08/25
08/25
08/25
08/26
08/26
08/26
08/26
08/26
08/26
08/26
08/26
08/26
08/26
08/27
08/27
08/27
08/27
08/27
08/29
08/29
08/29
08/29
08/29
08/29
08/30
08/30
08/31
08/31
ZN
3N
VL
VL
3N
ZN
DL
:T
CT
CN
VL
VL
VL
VL
OL
OL
OL
CN
VL
VL
VL
DL
DL
VL
VL
VL
VL
VL
VL
VL
VIS
VL
ST
VIS
3N
VL
VL
VL
VIS
CT
CT
MS
CB
MS
MS
CN
OL
;T
0.0944
0.1035
0.1279
0.1397
0.3854
0.6711
0.0870
2.0862
0.2920
5.2570
10.2516
0.9309
0.2136
10.4233
0.0341
0.7030
0.3456
0.0113
0.0176
1.3658
0.4611
0.0020
0.0031
0.6631
0.6926
1.4156
0.9648
0.2606
0.0194
0.3408
0.7242
0.0320
0.0028
0.0216
0.0177
0.0002
0.1820
0.0806
8.8271
1.5959
2.6064
1.0553
0.0038
0.0593
2.9543
0.0004
0.0923
1.5959
0.1116
0.1116
0.1010
0.1010
1.2167
1.2167
0.0581
3.6759
0.2606
4.9158
9.1666
0.7982
0.2513
9.0375
0.0370
0.5651
0.2670
0.0168
0.0292
2.2238
0.6171
0.0010
0.0017
0.7672
0.7672
1.5955
1.0518
0.2777
0.0416
0.5682
0.6542
0.0198
0.0021
0.0119
0.0108
0.0001
0.0951
0.0688
7.2788
1.0677
1.5027
0.7145
0.0047
0.1100
3.4697
0.0003
0.0512
0.9236
-0 0172
-0 0081
0.0269
0.0387
-08313
-0.5456
0.0289
-1 5897
00313
0.3412
1.0850
0.1326
-0.0377
1.3857
-0.0029
0 1378
0.0786
-0.0055
-0.0116
-0.8580
-0.1560
0.0010
0.0013
-0.1041
-0.0746
-0.1799
-0.0870
-0.0171
-0.0223
-0.2274
0.0700
0.0122
0.0007
0.0096
0.0069
0.0001
00869
0.0119
1.5483
0.5283
1.1037
0.3408
-0 0008
-0.0507
-0.5154
0.0000
0.0411
06723
-15.4%
-7.3%
26.6%
38.3%
-68.3%
-44.8%
49 8%
-43.2%
12.0%
6.9%
1 1 8%
16.6%
-15.0%
153%
-7 8%
24 4%
29 4%
-32.6%
-39.8%
-38 6%
-25.3%
97.6%
76.9%
-13.6%
-9.7%
-11.3%
-8.3%
-62%
-53.5%
-40.0%
10.7%
61.9%
36.0%
80.7%
63.9%
130.8%
91.4%
17.2%
21.3%
49.5%
73.4%
47.7%
-179%
-46.1%
-14.9%
10.1%
80.4%
72.8%
-0.22
-0 14
0.17
028
-0.71
-0.49
0.39
-047
0.04
-0.01
0.04
0.08
-0.21
0.07
-0 15
0.15
0.20
-038
-0.44
-0.43
-0.31
0.83
0.64
-0.20
-0.16
-0 18
-0.15
-0.13
-0.57
-0.44
0.03
0.50
0.26
0.67
0.52
1.14
0.77
0.09
0.12
0.38
0.61
0.37
-0.24
-0.50
-0.21
0.02
0.67
0.60
45
-------
/Vest Texas Data
0 Pate Component Type HVCS (slpm) EPA (slpm) Absolute Bias (slpm)
IID4
IDS
D6
D7
D3
ID1
ID2
ID9
ID10
ID12
ID13
ID15
ID21
ID24
ID26
ID30
ID31
ID34
ID35
ID36
ID42
ID37
ID38
ID43
ID44
ID45
ID40
ID46
ID49
ID50
ID51
ID52
ID56S
ID57
ID60
TEST1
TEST2
ID61
ID62
TESTS
ID70
ID71
ID73
ID72
ID47
ID75
ID76
ID77
ID78
ID79
ID1 R1
ID12R1
ID46 R1
IDF
ID84
ID85
ID87
IDI
ID88
ID89
ID95
ID96
ID92
ID93
ID91
ID90
09/21
09/21
09/21
09/21
09/21
09/21
09/21
09/21
09/21
09/22
09/22
09/22
09/22
09/22
09/22
09/23
09/23
09/23
09/23
09/23
09/23
09/23
09/23
09/23
09/23
09/23
09/24
09/24
09/24
09/24
09/24
09/24
09/24
09/24
09/24
09/26
09/26
09/26
09/26
09/27
09/27
09/27
09/27
09/27
09/27
09/27
09/27
09/27
09/27
09/27
09/28
09/28
09/28
09/28
09/28
09/28
09/28
09/28
09/28
09/28
09/29
09/29
09/29
09/29
09/29
CN
VL
CN
CN
VL
VL
VL
VL
VL
CN
VL
VL
VL
VL
*JL
VL
VL
VL
MS
CN
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
VL
"N
VL
VL
VL
N
VL
VL
VL
;N
us
CF
VL
VL
VL
VL
VL
VL
VL
VL
09/29 M-
0.0298
0.0798
05777
0.3524
00138
3.9422
1 2705
0.3604
1 7620
1.3587
1.6213
0.0691
43116
2.0488
7.8876
08039
0.0992
00172
1 3897
0.0869
0.0069
0.4509
1 8613
0.0077
0.0274
00269
0.3831
1.9290
0.2647
0.1168
0 1241
0.8201
0.1442
0.2204
1.0875
0.1633
30263
2.8341
63681
0.1562
2.0472
2.8108
1 1530
1.9447
38225
0 1718
0.7859
0.0915
0.0204
0.0079
2.9474
09535
1 5865
1.3279
0 1828
1 3955
0.5777
6.9937
69424
0.8320
04644
0.6434
0.8397
0.3208
0.5719
63306
0.0200
00845
0.2985
0.2038
00077
1.6919
0.6363
0.1536
0.8151
0.7767
0.9676
0.0405
3.1760
1.3544
6.7354
0.5014
0.0777
00130
0.5803
0.1032
00038
0.3099
0.9521
0.0048
0.0138
00099
0.1702
1.0186
0.1566
0.0693
0.0660
0.3448
0.1447
0.1542
0.6568
0.0954
1 8791
4.5508
3.7301
0.1505
1.9435
2.6418
1 .3602
1.7184
38435
0.1501
06883
0.0627
0.0209
0.0089
4.3063
0.9016
07210
0.9726
0.2703
1.5214
0.5449
65334
76512
0.6747
05827
0.5062
07532
05350
04740
7.1402
0.0099
-0 0047
02792
0 1486
00061
2.2503
06342
02068
0.9469
0 5820
0.6537
0.0286
1 1356
06944
1 1521
0.3024
00216
00042
0.8094
-00163
0.0031
0.1410
0.9092
0.0029
0.0136
0.0170
0.2130
0.9105
0.1080
0.0475
00581
0.4752
-0.0005
00663
04308
0.0679
1.1472
-1 7167
2.6380
0.0057
0.1037
0.1690
-0 2072
0.2263
-0.0210
0.0216
00977
00288
-0.0005
-00010
-1 3588
00518
08655
03553
-0 0875
-0 1259
00329
04604
-07089
0 1574
-0 1184
0 1372
00865
-0 2142
0 0979
-0 8096
Relativp Bias (%) [True Bias ("/ŧ)
49 5%
-5 6%
93 5%
72.9%
79 9%
1330%
99.7%
1 34 6%
116.2%
74 9%
67.6%
70 7%
35 8%
51 3%
17 1%
60.3%
27 8%
32.8%
1395%
-158%
80 0%
45 5%
95.5%
61 7%
99.0%
170.7%
125.2%
89.4%
69 0%
68.6%
88.0%
1378%
-0 3%
43 0%
65 6%
71 1%
61.0%
-37 7%
70 7%
38%
53%
64%
-15.2%
132%
-0.5%
144%
14 2%
45 8%
-2 4%
-11 0%
-31 6%
57%
1200%
36 5%
-32 4%
-8 3%
60%
70%
-9 3%
23 3%
-20 3%
27 1%
11.5%
-40.0%
20 7%
-11 3%
38.5%
-126%
79.2%
60 1%
66 6%
1158%
84 9%
1173%
1002%
62 0%
55 2%
58 0%
25.7%
40 1%
84%
48 4%
18 3%
23 0%
121 8%
-22 0%
66 6%
34 7%
81.0%
49 7%
84 3%
1507%
108.5%
75 4%
56.5%
56.1%
74 0%
120.2%
-7 7%
32 4%
53 3%
58.5%
49 1%
-42 3%
58 1%
-3 9%
-2 5%
-1 5%
-21.5%
48%
-7 9%
59%
57%
35 0%
-9 6%
-176%
-36 6%
-2 1%
1038%
26 4%
-37 4%
-15 1% |
-1 8%
-0 9%
-16 0%
14 2%
-26.2%
17 7%
32%
-44 5%
1 1 7%
-179%i
46
-------
APPENDIX A
MEASUREMENT OF FUGITIVE ORGANIC VAPOR LEAKS
Quality Assurance Project Plan
EPA Contract No. 68-D2-0062
Prepared for:
Charles Masser
National Risk Management Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Prepared by:
Southern Research Institute
Environmental Studies Division
Research Triangle Park, NC 27709-3825
May 18, 1994
47
-------
Southern Research Institute
Environmental Studies Division
MEMORANDUM
To: Charles C. Masser
From: Eric Ringler
Date: June 29, 1994
Subject: Response to Comments on QA Plan for the HVCS Evaluation Study
This memorandum constitutes SRI's response to the comments of the reviewers of the
Quality Assurance Project Plan for the HVCS Evaluation Study. We appreciate the
careful and constructive review of the plan, and especially the very favorable response
to the section on Data Quality Objectives. The responses follow the order in which the
comments are given in the review dated June 3, 1994. We feel that these responses
adequately address the concerns of the reviewers. We understand that the plan is
approved with consideration of the comments and have no plans to revise the plan itself
at this time. A revised version of the plan can be prepared at your request. The
comments and these responses should be provided to those who may have received
copies of the plan and should accompany any future distribution.
Since the draft plan was submitted, we have come up with some criticisms of our own
that the reviewers did not address. Changes to procedures that address these issues are
included after the responses to the reviewers comments. Please let us know if you have
any questions or further comments.
cc: Ashley Williamson
Stephen Piccot
48
-------
Response to Reviewers Comments
GENERAL
As recommended, text in section 1 should be amended to read: EPA Method 21
is not included in the laboratory testing because it is intended to locate and classify
leaks only, and is not intended to be used as a direct measure of mass emission
rates (40 CFR 60, Appendix A, Method 21, Paragraph 1.2).
It should be noted; however, that method 21 procedures will be used for obtaining
screening values before and after each measurement of mass emissions on the
laboratory test components, just as will be done in the field. However, the
laboratory testing is not intended to evaluate the performance of method 21.
The unexplained asterisk in Table 2-2 should refer to the following comment:
Estimated precision and accuracy objectives for HVCS. Determining the true
precision and accuracy of the HVCS measurements is the purpose of the study.
LABORATORY MEASUREMENTS TEST MATRIX
The precision of the flow measurements required for the laboratory testing will be
established based on comparisons with a primary standard bubble flow meter.
These comparisons will be performed immediately before and after each
measurement. It is not necessary to repeat the HVCS or bagging procedures to
establish the precision of the flow measurements.
DATA QUALITY OBJECTIVES
This comment describes the careful treatment given to DQO's in the plan. We are
grateful for the reviewers appreciation of this section.
It would be desirable to randomize the order in which HVCS and EPA protocol
measurements are performed in the field to minimize the impact of possible
changes in leak rate with time. It is expected that the two types of measurements
will not always be performed in the same order; however, as a practical matter, it
is likely that efficient use of time and manpower will be the deciding factor. As the
reviewers stated, screening values will be obtained before and after each
emissions measurement, and these will be used to determine if a change has
occurred. A data quality objective of 20 percent is given for the repeatability of the
screening measurements. Thus, a change in screening value of less than 20
percent would not be considered significant. If there is a significant change in
screening value, the effect would be to invalidate the comparison. The simplest
remedy is to repeat either the HVCS or EPA protocol measurement (as
appropriate) as soon as possible.
49
-------
SCREENING MEASUREMENTS
Each OVA will be leak checked daily and the sample flow rate will be checked
during the daily calibrations - 3 times per day. Screening values will be obtained
per EPA protocol procedures; with the probe at the surface of the leak interface
when there are no moving parts, and 1 cm from the interface when there are
moving parts. The 1 cm distance will be gauged by eye. No spacer will be used.
No glass wool will be used in the probe. The test plan specifies multipoint
calibrations 3 times per day using a centrally located certified gas supply. A Mylar
bag containing 1,000 or 10,000 ppm methane (from the certified gas supply) can
be obtained and carried with the operator if needed due to difficulty in returning to
the central gas supply or if frequent flameouts are expected (the calibration should
be checked after each flameout).
SCREENING QA/QC PROCEDURES
Certified calibration gases that have been quantitatively analyzed after mixing will
be obtained from a commercial specialty gas vendor. The analytical accuracy of
the gas certification is guaranteed by the vendor to be within ą2 percent. In
addition, SRI will cross check the gas concentrations against other certified gas
standards and verify the gas composition by FTIR spectroscopy.
After analysis, each bag will be purged with room air 3 times. Before sampling,
each bag will be purged with sample gas 3 times. SRI has direct experience to
show it is not necessary to take extraordinary measures in order to minimize
carryover when reusing sample bags that have been used to sample natural gas
(primarily methane and other light hydrocarbons). Purging, as described above,
is sufficient. It is not necessary to segregate high concentration bags from low
concentration bags.
QA/QC FOR BAG SAMPLING
For natural gas sampling, purging as described above, is sufficient to minimize
carryover. It is not necessary to check hydrocarbon concentrations in the purge
air in every case. However, it is a reasonable precaution to check purge air for
several bags collected from different areas at the start of the study. If carryover
is found to be a problem, then purge air would be checked on every bag.
Alternatively, it may be more cost effective to simply destroy the bags and make
new ones, reusing only the valve. SRI will have the capability to make heat sealed
sample bags in the field.
CONTROLLED LEAK TEST QA/QC
For the controlled leak tests, the flow will be metered as in the laboratory studies.
SRI has modified this approach since the draft Plan was submitted. The new
approach to controlling and metering the leak rate in a simulated component is
described below.
50
-------
AUDITS
SRI has adopted careful measures to ensure the quality of the calibration
standards that will be used (see above). Use of independent audit gases is not
necessary to meet Category III requirements.
PROJECT ORGANIZATION
The controlled leak rates will not be revealed to the operators (STAR and SRI) of
the HVCS and bagging apparatus during the blind leak tests. The apparatus that
will be used (see below) will be arranged so that the" leak rate cannot be read by
the operator. The leak rates will be set and controlled by the SRI field coordinator.
STAR will follow all QA/QC procedures in the QAPP. Field review of STAR
activities and data will be conducted by the SRI field coordinator.
Further Modifications to Procedures
SRI has adopted changes to the EPA protocol emissions sampling and laboratory test
procedures since the draft QAPP was submitted for review.
First, SRI noted that the "Blow-Through" tent/bag protocol does not adequately account
for the total gas flow when testing large leaks. Specifically, the volume of the leaking gas
is not directly measured. If this volume is large compared the carrier gas volume, then
a significant underestimate of the total flow (and the leak rate) will result. This error is
equal to the ratio of the leak flow to the carrier flow. SRI initially decided to modify the
"Blow-Through" method so that the total flow exiting the bag is measured. However,
during initial testing, it proved to be impractical to consistently construct a sufficiently leak
tight enclosure so that the exit flow could be accurately measured. Therefore, it was
decided to evaluate the vacuum bag method, especially for use with larger leak rates.
The vacuum method is not as sensitive to enclosure leaks since the total gas flow is
measured directly, including any air that may infiltrate the bag. The vacuum method is
slightly more complex to execute in the field and there is a small potential that an
explosive gas mixture could be sampled. However, the performance advantages of the
vacuum method outweigh these disadvantages. Based on this evaluation, SRI has now
decided to use the vacuum bag protocol in the field. The laboratory testing will include
a limited number of comparisons of the vacuum and blow through method; however, it is
beyond the scope of the current work to fully evaluate differences in performance between
the two methods.
Second, since the draft QAPP was submitted, SRI has developed an improved method
for controlling and metering the leak rates in the laboratory test components (and in the
field leak rate tests). The method specified in the QAPP called for developing a "family"
of calibration curves for each rotameter over a range of back pressures. In practice, it
turned out to be difficult to maintain constant pressure over a range of flow rates so that
these curves could be generated. In addition, the effect of back pressure increases on
rotameter readings was such that a large number of calibration curves would be needed
to adequately cover the pressure range from 0 to 100 psig. Each 1 psi increase in back-
51
-------
pressure decreases the rotameter reading by about 2.5 to 4 percent (the relationship is
not linear). This means that a family of 10 to 20 calibration curves would be needed to
bracket the 0 to 100 psi pressure range so that no flow measurement in this range was
more than 10 percent away from a calibration curve. This was not anticipated when the
draft QAPP was prepared.
In initial testing, it was also shown that the usual formula used to correct for back
pressure on a rotameter based on the square root of the ratio of the pressures leads to
significant underestimates (as much as 30 percent) of the true flow rate with pressures
as high as 90 psig. In addition, this formula underpredicts systematically by different
amounts at different flow rates (the underpredicition is more severe at lower flows). This
is a consequence of the fact that the ratio of the pressures does not fully account for
rotameter behavior, especially at elevated back pressures. A full treatment of rotameter
mechanics is given in the student manual for EPA's APTI Course 435, Atmospheric
Sampling.
The modified leak metering and control procedure is not subject to these difficulties and
is simple to execute. In this procedure, the leak rate is induced at a desired back
pressure by manipulating the downstream component and a flow metering valve
(upstream from the rotameter). Pressure is metered directly upstream from the
component using a sensitive pressure transducer (0.5 torr resolution). The rotameter
reading and back pressure are recorded. A 3-way valve is then actuated so that the flow
from the rotameter is directed through a needle valve and then to a bubble flowmeter.
The needle valve is used to re-create the exact back-pressure condition on the rotameter
that was used in establishing the component leak. The actual flow (at room conditions)
is then measured by the bubble meter. The flow is then returned to the component and
pressure and rotameter readings are checked to ensure that they have not changed. The
flow measured by the bubble meter is the leak rate. The leak rate is verified after each
mass emissions measurement by the same method. If a change of more than 5 percent
has occurred, the measurement is repeated.
52
-------
Appendix A
TABLE OF CONTENTS
Section Title
1 PROJECT DESCRIPTION 56
1.1 Introduction 56
1.2 Project Objectives 57
1.3 Test Site Selection 57
1.4 Study Design Overview 58
1.4.1 Field Measurements 58
1.4.2 Laboratory Measurements 60
1.4.3 Data Analysis and Reporting 61
2 DATA QUALITY OBJECTIVES 65
2.1 Determining Data Quality Requirements 65
2.2 Data Quality Requirements for HVCS and Bagging Methods 67
2.3 Data Quality Objectives for Direct Measurements 69
2.3.1 HVCS and Bagging Emissions Measurements 70
2.3.2 Screening Measurements 71
2.3.3 Gas Chromatography 72
3 DATA QUALITY INDICATORS , 73
3.1 Completeness 73
3.2 Precision . 73
3.3 Accuracy 74
3.4 Representativeness 74
3.5 Comparability 75
4 SAMPLING AND ANALYTICAL PROCEDURES 77
4.1 Component Screening 77
4.1.1 Screening Method Description 77
4.1.2 Screening Test Procedures 78
4.1.3 Screening QA/QC Procedures 79
4.2 HVCS Measurements 80
4.2.1 HVCS Description 80
4.2.2 HVCS Test Procedures 81
4.2.3 HVCS QA/QC Procedures 81
4.3 EPA Protocol Mass Emissions (Bagging) Measurements 82
4.3.1 Bagging Measurement Description 82
4.3.2 Bag Sampling Procedures 83
4.3.3 QA/QC for Bag Sampling 83
4.4 Gas Chromatography 84
4.4.1 Gas Chromatography Description 84
4.4.2 Method 18 Test Procedures 85
4.4.3 Method 18 QA/QC 86
53
-------
TABLE OF CONTENTS (Cont.)
Section Title
4.5 Controlled Leak Testing 87
4.5.1 Controlled Leak Test Bed Description 87
4.5.2 Controlled Leak Test Bed Procedures 90
4.5.3 Controlled Leak Test QA/QC 90
5 DATA REDUCTION, VALIDATION, AND REPORTING 91
5.1 Data Reduction 91
5.2 Data Validation 91
5.3 Data Reporting 93
6 AUDITS 96
7 CORRECTIVE ACTION 97
8 PROJECT ORGANIZATION 98
9 TEST PROGRAM HEALTH AND SAFETY 99
9.1 General Safety Issues 99
9.2 Project Specific Safety Issues 100
10 REFERENCES 102
APPENDIX A - HVCS Measurement Procedures 103
APPENDIX B - Field Data Forms 107
54
-------
LIST OF TABLES
Table 1-1 Field Measurements Test Matrix 63
Table 1-2 Laboratory Measurements Test Matrix 64
Table 2-1 Precision Requirements 69
Table 2-2 Data Quality Objectives 72
Table 5-1 QA/QC Test Conditions, Data Validity and Corrective Actions 94
Table B-1 Field Data Forms 107
LIST OF FIGURES
Figure 4-1 Flow Metering Schematic for Controlled Leak Test Bed 89
Figure 5-1 Generalized Data Flow 92
55
-------
Section No. 1
Revision No. 0
Date: 5/18/94
SECTION 1
PROJECT DESCRIPTION
1.1 INTRODUCTION
Fugitive emissions of methane and other organic vapors from leaking pipelines, valves, flanges,
and seals associated with natural gas, petroleum and chemical production and processing
facilities are an important source of methane and other organic emissions to the atmosphere.
Such emissions have historically been difficult to measure accurately. EPA Reference Method
21 "Determination of Volatile Organic Compound Leaks" describes instruments and procedures
that can be used to locate and assess the magnitude of such leaks. This consists of screening
components with a portable hydrocarbon analyzer and recording concentration values obtained
at the component interface where leaks are detected. Method 21 does not provide a direct
measure of the mass emission rate. Mass emission rates for leaking components have
traditionally been determined by "tenting and bagging" the leaks. This entails constructing a
sampling enclosure around the leaking component, introducing a known flow of diluent gas
through the enclosure, and determining the concentration of leaking gas in a sample captured
from the enclosure (usually done off-site in a laboratory). The concentration times the total flow
rate (diluent and leaking gas) gives the leak rate. This is clearly, a laborious and expensive
method that is not generally practical to apply for each leaking component at a facility.
According to the current EPA protocol (EPA 1993), mass emission rates may be estimated
indirectly by one of four methods. (1) Emissions may be estimated by applying published
emission factors (EPA 1993) to an inventory of components by type (valves, flanges, etc.) and
service (e.g., gas, liquid). (2) More refined emissions estimates may be obtained by identifying
leaking components (per Method 21) and applying separate emission factors to leaking and
non-leaking components. (3) Still more refined estimates are determined by applyling published
correlation equations (EPA 1993) to screening values for each component (obtained per Method
21). These emission factors and correlations were developed over the last 15 years based on
field studies at petroleum refineries, gas plants, and Synthetic Organic Chemical Manufacturing
Industry (SOCMI) plants. (4) Finally, the EPA protocol specifies procedures for developing unit
specific correlation equations that may more accurately estimate emissions for a specific facility.
This entails obtaining pairs of screening values and direct measurements of mass emissions (by
tenting and bagging) at a sufficient number of components so that representative, site specific
correlation equations can be developed.
A new measurements technique has recently been developed as a result of work sponsored by
the Gas Research Institute (GRI) (GRI 1994). The method is known as the High Volume
Collection System (HVCS). The HVCS was designed to obtain direct measurements of mass
emission rates without the need for tenting and bagging, and offers the potential for providing an
easy to use, and cost effective means to measure organic vapor leaks from gas, oil, and chemical
industry sources. The HVCS uses a battery powered pump to draw ambient air across a leaking
component at controlled and metered rates between 10 and 500 standard cubic feet per hour.
A portable Flame lonization Detector (FID) is used to measure the hydrocarbon concentration in
56
-------
Section No. 1
Revision No. 0
Date: 5/18/94
the HVCS exhaust. The mass emission rate is determined as the product of the sample flow rate
and the hydrocarbon content in the flow. The success of the method depends on capturing all
of the leaking gas from a component in the flow entering the sample inlet. The inlet is
constructed to enhance this capture (the inlet is shaped like the mouthpiece of a snorkel). Diffuse
leaks from larger components (such as a large flange) are captured by wrapping the component
in polyethylene sheeting so that the air flow passes over the entire leaking surface. Prior to this
study, limited laboratory and field testing of the HVCS indicated that the system had the potential
to provide a practical method for quantifying fugitive hydrocarbon emissions (GRI 1994).
1.2 PROJECT OBJECTIVES
The purpose of this project is to complete a study of HVCS method performance. The study will
consist of both field and laboratory testing. The field testing will assess the accuracy of the HVCS
method relative to an EPA protocol emissions measurements method. The performance of the
HVCS method will be evaluated specific to the range of component types, operating conditions
and leak rates encountered during the field testing. The goal is to challenge system performance
under the range of conditions representative of natural gas production. A major focus of the study
is to develop performance criteria for field use of the HVCS method. These criteria will identify
conditions under which best and worst HVCS performance is achieved and recommend
procedures for obtaining optimum results.
Laboratory testing is needed to establish the accuracy and precision of the EPA protocol
(bagging) and HVCS methods compared to controlled leak rates. This testing will strengthen and
support the field test results by associating the methods with test results under controlled
conditions. The laboratory tests are devised to represent "real world" components and leak types
so that overall errors (including sampling errors) are represented. Laboratory test procedures will
mirror the field test procedures. Laboratory testing will be completed prior to the field tests. Field
testing will be completed during the summer of 1994.
1.3 TEST SITE SELECTION
Field testing for this study will be conducted at natural gas production fields. The sites will be
selected by EPA in cooperation with GRI and API. The sites should be representative of a
geographic region associated with significant gas production. The sites should provide a
reasonably large number and high density of components so that a sufficient number and type
of leaking components can be identified without undue effort. The sites should be of "average"
age for U.S gas production sites (15 ą 5 years). The sites should be operating at moderate
pressure (200 to 600 psi). Few leaks are expected at very high pressure (2000 to 3000 psi) sites
since these sites must be highly maintained. More leaks might be expected at low pressure sites
(30 to 100 psi); however, expected leak rates would be small. Sites where the gas contains large
amounts of hydrogen sulfide (H2S) would not be considered because of the obvious safety
hazards and because very few leaks would be expected at these sites.
57
-------
Section No. 1
Revision No. 0
Date: 5/18/94
Gas production areas meeting these criteria can be found in south and west Texas. In these
regions, it is expected that a sufficient number of components can be found among 4 to 8 gas
production fields grouped around one or two gas processing plants. Sites in these areas are
being surveyed by API to identify industry participants for the study.
1.4 STUDY DESIGN OVERVIEW
This section presents an overview of the test plan for all field and laboratory measurements. The
test matrix proposed for field and laboratory measurements is presented in this section and
identifies the type and number of measurements to be completed. This includes primary
measurements that will be used directly to assess the performance of the HVCS method, and
additional measurements and calibrations necessary for quality assurance and quality control
purposes. This section also discusses how field activities will be coordinated to achieve the
project objectives.
1.4.1 Field Measurements
The first step in conducting the field study will be to identify and classify leaking components. SRI
will screen a sufficient number of components (up to 25,000) to allow selection of a
representative set of components for comparative analysis of the HVCS and bagging
measurements. This screening will be conducted according to EPA Reference Method 21 using
a Foxboro Model 108 organic vapor analyzer (OVA) calibrated specific to methane to obtain
instrument screening values (ISV's). The screening value, component identification number, type,
location, and operating parameters will be recorded for all components where readings over 10
ppmv (above background) are obtained. Leaking components will be further classified by
equipment type, type of service, and process location. Non-emitting components (ISV less than
10 ppmv) will simply be inventoried.
Experience from recent measurements at gas production fields in the Eastern United States
(conducted by STAR Environmental for GRI) indicates that screening 25,000 components can be
expected to yield 400 to 600 small leaks (ISV less than 1000 ppm), 300 to 400 medium leaks
(ISV less than 10,000 ppm), 200 to 300 large leaks (ISV greater than 10,000 ppm) and perhaps
20 very large leaks (ISV greater than 100,000 ppm) (GRI 1994). Recent measurements at
Western gas production fields indicates that 25,000 components can be expected to yield about
250 small leaks, 200 medium leaks, 200 large leaks and as many as 150 very large leaks (API,
1993). It appears that Eastern fields contain an overall larger number of leaking components,
while Western fields contain more larger leaks. This is probably related to the relative age of
Eastern and Western fields, as well as the average operating pressure. Western fields tend to
be newer and operate at higher pressure than Eastern fields.
A subset of at least 200 of the leaking components will be selected from the screened
components for emissions measurements using the HVCS (see section 4.2). The total number
will depend on how much screening is necessary to find leaking components. HVCS
measurements will be made at three flow settings for each component. If the HVCS is functioning
properly and captures all of the gas leaking from the component, the leak rates at each flow
58
-------
Section No. 1
Revision No. 0
Date: 5/18/94
setting should agree. The selected components should include most of the largest leaks, and
enough medium and small leaks to make up the balance. The goal is to fully characterize the
largest leaks and select enough medium and small leaks that a representative range of leak rates,
component types, and operating conditions is encountered and the system performance is
challenged under a wide range of conditions. SRI will review the screening data on a continuous
basis in the field, and coordinate measurement activities to achieve this goal.
The emission rates for at least 125 of the leaking components measured by the HVCS will also
be measured using a mass emissions sampling protocol published by EPA (see sections 4.3 and
4.4). The EPA protocol methods involve isolating a component leak, introducing a known flow
rate of a carrier gas, and measuring the concentration of the leaking gas to determine the
emission rate. The 125 components will characterize the largest leaks and should also represent
the range of leak types, sizes and operating conditions encountered. The bagged samples for
all of the 125 components will be analyzed for methane and total hydrocarbon content by gas
chromatography (GC) following EPA's Reference Method 18.
Further analysis will be conducted on about 50 of the 125 bag samples to determine speciation.
This will also be done following EPA Reference Method 18 (see section 4.4). The speciation will
be specific to methane, ethane, propane, and n-butane. According to EPA's SPECIATE database
for fugitive emissions from gas and oil production, these four compounds should make up more
than 80 percent of total fugitive hydrocarbon released from valves and fittings in gas service, and
more than 60 percent of fugitive emissions from valves and fittings in liquid service. Methane
made up more than 90 percent of the gas composition in a small number of samples (8) recently
obtained from gas production facilities (GRI, 1994).
In this study, the speciation information will be used to assess the response of the portable FID
to gases other than the calibration gas (methane) that are present. Ethane, propane and n-
butane all have FID response factors for the OVA of less than 1 (0.57 for ethane, 0.88 for
propane, and 0.38 for n-butane). That is, more of these gases will be present than indicated by
the OVA. Speciation will be performed on samples selected to represent groups of tested
components where differences in chemical composition may exist.
SRI anticipates that about 24 days of field testing will be needed to complete the field evaluation.
This time will be distributed over 2 to 3 trips, depending on the number and location of sites
selected. A brief pre-survey trip (2 to 3 days) is also planned. During the pre-survey, SRI will
survey the site(s), obtain equipment inventories and process diagrams, consult with plant
engineers and management, and make scheduling and logistical arrangements. Field activities
will consist of component screening, HVCS measurements, tenting and bagging, GC operation,
and field coordination. The field crew will consist of a total of up to 6 personnel: 2 screening, 1
HVCS, 1 tent/bag, 1 GC operator and 1 coordinator. The activities of these personnel will be
coordinated so that the required number of measurements of each type are completed in an
efficient manner. The role of the coordinator is to monitor screening and other data collection
activities in the field, select an appropriate number and type of leaking components for HVCS,
bagging, and speciation analysis, and direct field activities. The coordinator will also ensure that
QA/QC procedures are followed, and initiate corrective actions as necessary.
59
-------
Section No. 1
Revision No. 0
Date: 5/18/94
SRI anticipates that the full crew of 6 will not be needed at all times in the field. For example,
the screening crew might arrive on site a few days before the rest of the crew to get a head start
on identifying leaking components and then leave a few days early. In addition, the screening
crew could be supplemented with HVCS and tent/bag personnel until a sufficient number of
leaking components are identified. Measurement activities will be sequenced so that initial
screening will be followed as closely as possible by HVCS and bagging measurements. In
accordance with the EPA protocol, all components measured by HVCS and bagging will be
re-screened immediately before and after the measurement to verify screening values.
Table 1-1 presents the test matrix for the field measurements. In the table, the 4 primary
measurement types: (1) Method 21 screening, (2) HVCS emissions measurements, (3) EPA
protocol emissions measurements, and (4) speciated measurements are grouped together with
the associated calibration and quality control measurements that will be used to assess data
quality. Measurement frequencies are based on study requirements and quality control needs.
Numbers of measurements are based on program and quality control requirements and on the
anticipated number of trips and days in the field needed to complete the measurements.
1.4.2 Laboratory Measurements
In this study, the primary means of establishing the performance of the HVCS will be through
comparisons of HVCS measurements with results from the EPA protocol emissions sampling
method. Thus, it is important that the precision and accuracy of the bagging method is carefully
determined.
Quality assurance testing of the bagging method was conducted in a previous study (EPA 1980c).
In this study, the overall sampling and analytical precision of the bagging method (vacuum bag
protocol) was found to be about 17 percent based on tests involving introduction of a known
amount of gas into the sampling apparatus. It should be noted, however, that these tests did not
account for any sampling errors associated with the leak enclosure. For example, errors resulting
from constructing an imperfect enclosure around a component that leaks excessively or prevents
adequate mixing of the dilution air with the leaking gas would not have been reflected in the tests.
Other tests involving repeat measurements of the leak rate from the same component yielded an
overall precision of about 40 percent. The researchers concluded that this apparent imprecision
was due to actual changes in the leak rate of the component over time.
The EPA protocol describes two mass emission sampling methods, referred to as the vacuum-
bag and blow-through methods, and allows for variation in sampling and analytical technique.
Because of this variation, it is important that the accuracy of the method,, as actually applied, be
established on a case by case basis. The protocol recommends that the accuracy of the
sampling/analytical method be verified by performing leak rate checks. According to the protocol,
this can be accomplished simply by introducing a known flow of a known gas into a sampling bag,
measuring the gas concentration, and calculating the leak rate. The calculated leak rate should
agree with the induced leak rate. EPA recommends that these checks be conducted in the
laboratory to verify the performance of equipment and procedures, and repeated periodically in
the field. These leak rate checks do not test the accuracy of the bagging procedures as they
60
-------
Section No. 1
Revision No. 0
Date: 5/18/94
would actually be applied in the field since they do not address the total sampling error (i.e., the
enclosure is left out of the sampling system).
The tenting and bagging procedure is fairly complex and there are numerous ways in which
sampling and analytical errors might occur. Since, in this study, the bag sampling data will be
used as the reference against which the field performance of the HVCS method will be measured,
the precision and accuracy of the data acquired with the bagging method must be evaluated
under controlled conditions that closely simulate field conditions. To accomplish this, SRI will
construct controllable leaks in a number of representative components (see section 4.5). The
leak rate will be controlled by varying the pressure delivered to the component, and/or by
physically altering the size of the leak (See Section 4.5). A representative range of leak rates will
be established and measured for each component. In the EPA emission factors for gas plants
(EPA Protocol, June 1993), leak rates (as methane) range from 0.002 to 1.1 liters per minute
(Ipm) for small leaks (defined as yielding an ISV less than 10,000 ppm) and 0.8 to 20 Ipm for
large leaks (ISV more than 10,000 ppm), depending on the type of component. Leak rates
between 0.0002 and 1.3 Ipm (with one leak as high as 25 Ipm) have recently been measured at
gas production operations in the Gulf and Appalachian regions (GRI, 1994). Based on this, and
considering that larger leaks are of greater importance for this study, SRI will induce leak rates
in the test components in the range of 0.1 to 20 l/min. This range covers most of the largest
leaks and all but the smallest of the smaller leaks.
In these laboratory studies, emissions from the controlled leaks will also be measured using the
HVCS method. These tests will establish the accuracy and comparability of the HVCS and
bagging techniques under controlled conditions before testing under field conditions. The tests
will also serve to identify and evaluate any problems with HVCS performance before field testing.
One of the key areas of uncertainty in the HVCS method is the ability of the sampler to capture
all of the leaking gas, especially for larger leak rates and larger component sizes. Table 1-2
presents a test matrix for the laboratory measurements. The actual test matrix executed may
differ somewhat from Table 1-2 depending on the range of leak rates and pressures that can
actually be achieved in the test components.
Controlled leak rate tests will be repeated periodically in the field using a suitable component (see
section 4.5). The performance of both the HVCS and bagging methods will be assessed in the
field leak rate tests. The field leak rate tests will also be used as "blind" audits to test operator
proficiency.
1.4.3 Data Analysis and Reporting
Upon the conclusion of field measurements, data screening, and data analysis activities, a report
will be prepared that describes the study findings and documents how these findings were arrived
at. The report will assess the overall performance of the HVCS and identify factors that affect
its performance. HVCS performance will also be assessed for different component types, leak
rates, and operating conditions (e.g., pressure). HVCS performance will also be evaluated under
varying operating conditions of the HVCS (e.g., flow rate, and concentration range). Appropriate
descriptive statistics and significance tests will be used to represent the data and interpret the
results (see sections 2, 3 and 5).
61
-------
Section No. 1
Revision No. 0
Date: 5/18/94
If sufficient data are available, the report will also present a comparison of emissions estimates
based on HVCS, tenting/bagging, and EPA protocol emission factor and correlation approaches.
It is important that such comparisons are carried out over representative groups of components
as the EPA emission factor and correlation approaches are properly applied only to relatively
large sets of components. Finally, the report will include an assessment of data quality based
on data quality indicators (as described in section 3), and on consideration of any operational or
methodological problems that may occur. The report will also contain summary tables of data
collected for the study.
62
-------
Table 1-1 Field Measurements Test Matrix
Measurement Type
Method 21 Screening
OVA Calibrations
HVCS Emissions
Measurements
Flow
Concentration
Pre/Post Screening
Field Leak Rate Test
OVA Calibrations
Flow Calibrations
EPA Protocol
Emissions
Measurements
Flow
Concentration
Pre/Post Screening
Field Leak Rate Test
GC Calibrations
Speciated
Measurements
GC Calibrations
Method or Device
FoxboroOVA108
Cylinder Gases
STAR HVCS
Rotameters
Foxboro OVA 108
FoxboroOVA 108
Controlled leak
Cylinder Gases
Dry Gas Meter
Tent/Bag
Rotameter
GC/FID/TCD
HP5890
Foxboro OVA 108
Bubble Flowmeter
& Cylinder Gas
Cylinder Gases
GC/FID/TCD
HP5890
Cylinder Gases
Quantity
Screening Value
Concentration
Leak Rate
Flow
Concentration
Screening Value
Flow
Concentration
Flow
Leak Rate
Flow
THC Concentration
Screening Value
Flow
Methane
Concentration
Methane, Ethane,
Propane, n-Butane
Concentration
Frequency
900 to 1800 per day
Daily Pre/Post Test
approx. 15/day
approx. 15/day
approx. 15/day
Each measurement
Three per field test
Daily Pre/Post Test
Pre/post field test
5 to 10/day
5 to 10/day
5 to 10/day
Each measurement
Three per field test
Daily Pre/Post test
2 to 4 per day
Daily Pre/Post test
Number
up to 25,000
42 (per OVA)
200
200
200
400
9
36
6
125
125
125
250
9
48
50
36
Data Use
(1) Identify leaking components, (2) provide
screening values for emissions estimates
Analytical P&A
Evaluation of HVCS Performance
Input to leak rate calculation
Input to leak rate calculation
Verification of screening value
Field P&A , operator bias
Analytical P&A
Verify HVCS flow system accuracy
Evaluation of HVCS Performance
Input to leak rate calculation
Input to leak rate calculation
Verification of screening value
Flow/Concentration P&A
Analytical P&A
Evaluation of FID response to gas mixture
Analytical P&A
-------
Table 1-2 Laboratory Measurements Test Matrix
Measurement Type
Method or Device
Quantity
Frequency
Number
Data Use
Laboratory Controlled Leak Testing
Release Rate
Flow
Concentration
Method 21 Screening
OVA Calibrations
EPA Protocol
Emissions
Measurements
Flow
Concentration
GC Calibrations
HVCS Emissions
Measurements
Flow
Concentration
OVA Calibrations
Flow Calibrations
Calculated
Bubble Flowmeter
Cylinder Gas
Foxboro OVA 108
Cylinder Gases
Tent/Bag
Bubble Flowmeter
GC/FIDrrCD
HP5890
Cylinder Gas
STAR HVCS
Rotameters
Foxboro OVA 108
Cylinder Gas (3)
Dry Gas Meter
Leak Rate
Flow
Concentration
Screening Value
Concentration
Leak Rate
Flow
Concentration
Methane
Concentration
Leak Rate
Flow
Concentration
Concentration
Flow
5 components, 3 leak
rates, 3 pressures
5 components, 3 leak
rates, 3 pressures
5 components, 3 leak
rates, 3 pressures
5 components, 3 leak
rates, 3 pressures
Daily Pre/Post Test
5 components, 3 leak
rates, 3 pressures
5 components, 3 leak
rates, 3 pressures
5 components, 3 leak
rates, 3 pressures
Daily Pre and Post Test
5 components at 3 flow
rates, 3 leak rates, 3
pressures
5 components at 3 flow
rates, 3 leak rates, 3
pressures
5 components at 3 flow
rates, 3 leak rates, 3
pressures
Daily Pre and Post Test
Pre and Post Test
45
45
45
45
10
45
45
45
10
135
135
135
10
2
Reference leak rate
Input to leak rate calculation
Input to leak rate calculation
Associate Method 21 screening values with
laboratory tests
Analytical P&A
Overall Sampling/Analytical P&A for Method
Input to leak rate calculation
Input to leak rate calculation
Analytical P&A
Overall Sampling/Analytical P&A for Method
Input to leak rate calculation
Input to leak rate calculation
Analytical P&A
HVCS Flow System Accuracy
-------
Section No. 2
Revision No. 0
Date: 5/18/94
SECTION 2
DATA QUALITY OBJECTIVES
In this section, data quality objectives are developed which state the values of key data quality
indicators that must be achieved to draw conclusions from the study with the desired level of
confidence. This section lays out the criteria for the overall experimental design. First, overall
data quality requirements are determined by starting with the study objective and applying
appropriate statistical procedures to determine the level of data quality needed to meet that
objective at a pre-determined level of significance. Then, these errors are propagated to the level
of the direct measurements required to conduct the study. This process yields practical and
defensible data quality objectives. Data meeting these objectives can be used to draw
conclusions from the study with a known level of confidence. The following section (Section 3)
describes how values for the data quality indicators are established for each measurement.
2.1 DETERMINING DATA QUALITY REQUIREMENTS
As stated earlier, the primary goal of the study is to assess the field performance of the HVCS
emissions measurement method. Ideally, this assessment would be based on comparisons of
the HVCS leak rate measurements with the true leak rate for a wide variety of components, leak
rates and operating parameters. These comparisons could be expressed as
D = ->L_I (1)
where, H is the HVCS measured leak rate, T is the true leak rate, and D is the difference relative
to the true leak rate. If the HVCS measurement is unbiased, this difference will be zero.
However, after a number of such comparisons have been made, one would expect some
variability in D due to systematic and random errors in measurement. Data quality objectives can
be developed by considering how much variability in D can be tolerated while detecting a given
bias in the HVCS measurement at a given level of significance. This is accomplished using an
appropriate statistical test of significance. This "allowable" error is then propagated down to the
individual measurements that make up the difference.
The first step is to consider the frequency distribution of D. This will be done using normal
probability plots (Gilbert, 1987). If the distribution is approximately normal (i.e., symmetric about
the mean and relatively free from outliers), then standard "t" procedures for paired comparisons
can be used. However, the value of D as defined in equation 1 is restricted between -1 and +00.
Thus, if H is positively biased as often as negatively biased, the distribution of D will be right
skewed and may be lognormal. In this case, the "t" procedures would be applied to the log
transformed differences (In D) (Griffiths, 1986). In either case, the form of the test is as follows.
The null hypothesis is that the mean difference is zero (H0:u0 = 0). The alternative hypothesis
is that the mean difference is not zero (Ha:u * 0). The t statistic is then given by:
65
-------
Section No. 2
Revision No. 0
Date: 5/18/94
t - -- (2)
sDl \fn
where D is the average of the measured differences, SD is the standard deviation of the
differences, n is the number of data pairs, and the true mean (u) difference is assumed to be
zero. This is a 2-sided test, so when the absolute value of t exceeds a critical value for the given
level of significance, H0 must be rejected, meaning that the measurement is biased. Alternatively
if |t| does not exceed the critical value, H0 cannot be rejected, meaning that the measurement is
unbiased. That is,
if | t\ > ti(n-l), must reject H0, \i * 0
(3)
// ] t \ $ ^'(/7-1), cannot reject H0, \i = 0
where t0' (n-1) is the critical value for the t distribution with n-1 degrees of freedom. For a 95
percent significance level, a = 0.025.
In analyzing the data from this study, we will draw conclusions from sets of comparisons
representing different component types, leak rates and operating parameters. These sets may
be represented by different numbers of data pairs. In general, increasing the number of data
pairs, will increase the level of variability that can be tolerated in detecting a given level of bias.
As an example, a "worst case" of 15 data pairs will be considered. This is likely to be the
minimum number of pairs from which conclusions will be drawn. Critical values are found using
standard statistical tables. For example,
rB" (15-1) = 2.145 (4)
To arrive at criteria for data quality, equations 2 and 3 are combined and the critical values in (4)
are substituted as follows.
I t\ = i "_ > f.'(15-1) 2.145 (5)
s0//15
This expression is then evaluated to find the coefficient of variation (CV).
66
-------
Section No. 2
Revision No. 0
Date: 5/18/94
CV = -=Ģ-< 1.8 (6)
\D\
Equations 5 and 6 state that, for 15 data pairs, when the CV is less than 1.8, the critical t-value
is exceeded: meaning that the HVCS measurements are biased at the 95 percent significance
level. Similarly, when the CV is greater than or equal to 1.8, the HVCS measurements are
unbiased.
The level of precision needed to detect a given level of bias can easily be determined from
equation 6. For example, a 10 percent average difference between the HVCS measurement and
the true value (D = 0.1) can be detected with precision as high as 18 percent (s = 0.18).
Similarly, precision of 8.5 percent would be required to detect a bias of 5 percent. These results
are consistent with the logic that more precise measurements are needed to distinguish results
that are not far apart, and less precision can be tolerated when looking for larger differences.
2.2 DATA QUALITY REQUIREMENTS FOR HVCS AND BAGGING METHODS
The discussion above compares the HVCS measurements to a true, and unbiased, leak rate.
In the field, however, the true leak rate will not be known. Instead, an EPA protocol bagging
method will be used as the reference against which the quality of the HVCS data are to be
assessed. The EPA protocol methods have wide acceptance; however, they are subject to some
bias and variability. Error in the bagging measurements must be addressed explicitly. Laboratory
testing using controlled representative leaks (see section 4.5) will be used to determine the
magnitude and variability of the error in the bagging technique. The fractional error (bias) in the
bagging data, BE, can be expressed as:
Be = 2-j1 (7)
where, B is the bagging measurement and T is the true leak rate. Equations 1 and 7 can then
be combined so that the difference between the HVCS measurement and the true value is
expressed in terms of the value of the bagging measurement and its associated error, BE.
D-H-^I^
B
The precision of the overall difference, SD, is then propagated downward to obtain the precision
required for the HVCS (H) and bagging (B) measurements. This is done by applying the standard
formula for the propagation of errors (equation 9) where the result R is a function of N variables
Xi (Clifford et. al., 1973) to equation 8.
67
-------
OX,-
Section No. 2
Revision No. 0
Date: 5/18/94
(9)
This yields the following expression
1/2
1)°Ŧ+ -4 "°Ŧ + "°a/l C"0)
D
D s
relating the precision of the difference between the HVCS and the true leak rate, GD to the
precision of the HVCS and bagging measurements, and the error in the bagging measurements
(aH, aB, and a8E). It can be shown, however, that the relatively simple expression given by
°D =
1/2
(11)
closely approximates equation 10 so long as the variance of BE is small, and H is of about the
same magnitude as B. The approximation is valid even when the bias in the bagging
measurements is fairly large (e.g., 20%).
Table 2-1 gives values for the precision of the HVCS and bagging measurements that must be
attained to detect a 5, 10, and 15 percent bias in the HVCS measurement (with 95 percent
significance). Values are given for comparisons with 125, 30, and 15 data pairs. The Table also
gives the corresponding values for aD. The Table assumes that the bagging and HVCS
measurements have the same precision (
-------
Section No. 2
Revision No. 0
Date: 5/18/94
Table 2-1. Precision Requirements
To Detect 5 Percent HVCS Bias
Number of Data Pairs
125
30
15
Required HVCS/Bagging
Precision (aH and aB)
19.9 %
9.5 %
6.4 %
Precision of Difference (aD)
28.2 %
13.4 %
9.0 %
To Detect 10 Percent HVCS Bias
Number of Data Pairs
125
30
15
Required HVCS/Bagging
Precision (aH and aB)
39.8 %
19.0 %
12.8%
Precision of Difference (aD)
56.5 %
26.8 %
18.1 %
To Detect 15 Percent HVCS Bias
Number of Data Pairs
125
30
15
HVCS/Bagging Precision
(aH and aB)
59.7 %
28.5 %
19.2 %
Precision of Difference (aD)
84.8 %
40.2 %
27.1 %
2.3 DATA QUALITY OBJECTIVES FOR DIRECT MEASUREMENTS
This section completes the discussion of data quality objectives by specifying objectives for each
of the direct measurements that will be made in the field study. The reasons for the selection of
the data quality objectives are discussed in each case. In the case of the flow and concentration
measurements that make up the HVCS and bagging results, data quality objectives are developed
using a formal propagation of errors. In other cases, the justification relies simply on
consideration of data quality requirements in light of study goals. Data quality objectives for the
overall sampling/analytical methods and for the direct measurements are summarized in Table
2-2.
69
-------
Section No. 2
Revision No. 0
Date: 5/18/94
2.3.1 HVCS and Bagging Emissions Measurements
The HVCS and bagging results are calculated from flow rate and concentration measurements.
For the HVCS measurement, flow is measured by rotameters and concentration is measured by
a portable FID (Foxboro OVA 108). For the bagging measurements, flow will be measured by
a bubble flowmeter and concentration will be measured by GC analysis. The leak rate is equal
to the product of flow rate and concentration. By applying the standard formula for propagating
errors (see equation 9), the standard error in the leak rate can be expressed in terms of the error
in the flow and concentration measurements as follows:
L = FC, -^ =
L
1/2
(12)
C2
where, L is the leak rate, F is the flow rate, and C is the concentration. That is, the fractional
error in the leak rate is proportional to the square root of the sum of the squares of the fractional
errors in the flow and concentration measurements. If bias in the flow and concentration
measurements is expressed simply as the ratio of the measured value to the true value, then the
fractional bias in the overall leak rate LE is given simply as the product of the bias in the flow and
concentration measurements, FE and CE minus one.
*FE T' - t,-
(13)
L - L LT L 1 FC 1 FJC 1
LE 7 - * -- 1 h^ '
For the bagging measurements, the bubble meter flow determinations are accurate and
repeatable within at least 1.0 percent (according to manufacturer specifications). Allowing for flow
measurement error due to differences in the sampling set up and potential leakage, flow
measurements are still expected to be accurate and repeatable within 10 percent. Total
hydrocarbon concentrations (as methane) measured with the GC should be accurate to within 10
percent and repeatable within 5 percent (according to Method 18). Applying equation 12,
precision for the overall bagging measurement is expected to be within about 12 percent. This
is well within the level of precision required for the study (see section 2.2). Applying equation 13,
bias in the bagging measurements should be within 20 percent. If proper sampling technique is
used, the bagging method is expected to capture all of the leaking gas. As discussed above
(section 2.2), this amount of bias in the bagging measurements can be easily tolerated so long
as it is quantified and remains fairly consistent. The actual precision and bias for the bagging
measurements will be determined in the laboratory studies.
In a key study assessing the emissions from petroleum refining conducted in the late 1970's,
(EPA, 1980c) a quality assurance program was implemented to assess the analytical and overall
precision and accuracy of the bagging method. In this case, the vacuum bag apparatus was
70
-------
Section No. 2
Revision No. 0
Date: 5/18/94
used. The apparatus was challenged with standard gases in a well defined QA protocol of
sample recovery tests. The precision of the sampling and analytical systems (standard deviation
for sampling and analysis of standard gases) was about 17 percent with no significant bias. A
precision of about 2.4 percent was achieved by the analytical system (GC) alone. It is important
to note that these tests involved only introducing standard gases into the sampling apparatus and
they do not reflect any variability or bias associated with the sampling enclosure. It is also
noteworthy that the 17 percent precision achieved in these studies is comparable to (but higher
than) the 12 percent expected precision based on the calculations presented above. In the same
study, repeat analyses of the same components were also conducted. This yielded a precision
of only about 40 percent. Much of the difference was attributed to short term variations in leak
rates for the sampled components with some of the variation being attributed to operator effects.
Since evaluating the precision and accuracy of the HVCS measurement is the purpose of the
study it may seem out of place to specify data quality objectives for the HVCS. However, it is
possible to estimate part of the expected error in the HVCS measurement by the same means
used above. For the HVCS measurements, the rotameter flows are accurate and repeatable to
within at least 5 percent according to manufacturer's specifications (2 percent or better) and
including expected human reading error. The OVA concentration measurements should be
accurate and repeatable within at least 15 percent. While the portable analyzers are known to
drift by more than 15 percent under some conditions, accuracy and precision of 15 percent or
better can reliably be maintained in the field if frequent calibration checks are implemented. For
the HVCS measurements, calibrations will be repeated before and after each set of
measurements. Applying equation 12, precision for the HVCS leak rate determinations should
be within 16 percent. Similarly, the bias in the HVCS measurements should remain within about
20 percent (equation 13). Note that this does not account for sampling error that would occur
when the HVCS fails to capture the entire leak.
A total of about 125 bagging measurements are planned for the study. Since it is important that
a representative sample of HVCS/bagging data pairs are collected, a completeness criteria of 95
percent will be the objective. A total of 200 HVCS measurements are planned, but only 125 of
these same components will also be selected for bagging. The excess HVCS measurements are
to allow flexibility in selecting a representative sample of components for the HVCS/bagging
comparisons. A data quality objective of 90 percent valid HVCS measurements (for a total of
180) should be sufficient to meet the study objectives.
2.3.2 Screening Measurements
The screening measurements will serve two purposes for the study. The first, and primary
purpose of the screening data is to identify leaking components and classify the components in
terms of their leak rate (based on the instrument screening value, or ISV). The secondary use
of the screening data is to develop emission rates based on EPA protocol emission factors and
correlation equations. The EPA protocol emission rate will later be compared to the emission rate
measured directly by the HVCS and bagging methods. These purposes will be served if the
precision and accuracy of the screening measurements are maintained within ą20 percent.
Based on experience, it is reasonable to expect these objectives will be met in the field. Actual
71
-------
Section No. 2
Revision No. 0
Date: 5/18/94
precision and accuracy during field measurements will be documented based on frequent
calibrations and quality checks (see section 3 and 4.1). If proper procedures are followed, it
should be possible to obtain valid screening values for more than 90 percent of screened
components.
2.3.3 Gas Chromatography
Measurement of the concentrations of the major constituents of the gas samples will be used to
characterize the gas streams encountered during the study. EPA's Reference Method 18,
Measurement of Gaseous Organic Compound Emissions by Gas Chromatography will be the
guideline for conducting the speciated measurements. According to Reference Method 18,
precision of 5 percent and accuracy of 10 percent should be achievable by a skilled GC operator.
A total of 50 valid speciation profiles will be obtained for 100 percent data capture. Additional
samples will be analyzed to make up for any samples that are invalidated.
Table 2-2. Data Quality Objectives
Measurement Type
Method 21 Screening
HVCS Emissions
Measurements
Flow
Concentration
EPA Protocol
Emissions
Measurements
Flow
Concentration
Speciated
Measurements
Method or Device
Foxboro OVA 108
STAR HVCS
Rotameters
Foxboro OVA 108
Tent/Bag
Rotameter
GC/FIDH-CD
HP5890
GC/FIDH-CD
HP5890
.. . I Completenes
Number II
up to 25,000
200
200
90%
90%
90%
200 || 90 %
125
125
125
50
95%
95 %
95%
100 %
Precision
20 %
(16 %)'
5%
15%
12 %
10 %
5%
5%
Accuracy
20%
(20 %)'
5%
15 %
20%
10%
10 %
10%
72
-------
Section No. 3
Revision No. 0
Date: 5/18/94
SECTION 3
DATA QUALITY INDICATORS
Data quality objectives for each measurement were developed and presented in the previous
section. Data quality objectives specify values of key data quality indicators that must be attained
in order to draw conclusions at a given level of confidence. This section describes how values
for each of the data quality indicators will be established.
3.1 COMPLETENESS
Completeness is a measure of the amount of valid data obtained compared to the amount
required to satisfy the study objectives. Completeness is established based on a count of valid
data remaining after data validation is completed (see section 5.2).
3.2 PRECISION
Precision is a measure of the repeatability of a measurement and is generally expressed as an
average deviation from a mean value (or standard deviation) when repeated measurements of
the same quantity are performed. Analytical precision is normally determined based on repeat
measurement of a known standard. Overall precision is generally comprised of sampling and
analytical components. Overall precision is more difficult to determine and is generally addressed
in a manner specific to the sampling/analytical system used.
For the bagging measurements, analytical precision will be determined as the standard deviation
of the difference (normalized for changes in standard concentration) between GC response and
gas standard concentration for the calibrations and quality control checks associated with a
sample run. Each sample run will be bracketed by multipoint calibrations (3 upscale points) and
will include the introduction of at least 1 QC standard during the analyses for a total of at least
7 data points to be used in determining the precision for the run. In general, a sample run will
consist of one day's measurements and approximately 10 samples. Overall analytical precision
for the study will also be calculated based on all calibrations and QC checks that were performed.
Overall precision for the bagging measurements will be determined based on controlled leak
studies and field controlled leak tests (see section 4.5).
For the HVCS measurements, analytical precision will be determined as the standard deviation
of the difference (normalized for changes in standard concentration) between the portable
analyzer response (Foxboro OVA 108) and certified standard gas concentrations. Each HVCS
measurement will be bracketed by multipoint calibrations of the portable analyzer. It is expected
that at least 12 HVCS determinations will be completed in the field per day. Precision will be
determined on a daily basis and reported separately for each portable analyzer used for HVCS
leak rate determinations. Since a sufficient number of comparisons will be available, precision
will be reported at each of the 3 upscale standard levels. Overall precision for the HVCS
measurements will be determined based on controlled leak studies and field controlled leak tests
73
-------
Section No. 3
Revision No. 0
Date: 5/18/94
(see section 4.5). Precision of the screening measurements will be based on daily pre-test and
post-test multipoint calibrations (for each analyzer) and calculated in the same manner.
Calibration precision for each portable analyzer based on Method 21 performance tests will also
be reported.
3.3 ACCURACY
Accuracy is a measure of the ability of a measurement to obtain the true value for the quantity
being measured and is also known as bias. Accuracy determinations must be based on
comparisons with known standards. Overall accuracy is comprised of sampling and analytical
components. Accuracy is generally expressed as the average difference from the standard value
expressed as a fraction (or percentage) of the standard value. This is also the calculation that
will be used in this study.
For the bagging measurements, analytical accuracy will be determined on a daily basis as the
average difference between GC response and standard gas concentration values based on the
7 comparisons with standard gases (see section 3.2). The overall accuracy of the bagging
measurements will be determined based on the controlled leak studies and field controlled leak
tests. The overall fractional bias in the bagging measurements is expressed in equation 7
(section 2.2).
For the HVCS measurements, analytical accuracy will be determined on a daily basis for each
analyzer based on frequent multipoint calibrations. The overall accuracy of the HVCS
measurement will be determined in the controlled leak testing and in the field testing. In the
controlled leak testing, HVCS bias will be assessed in the conventional manner (against a known
leak rate). The overall field bias in the HVCS measurements is expressed in equations 1 and 8
(section 2), and depends on the bias in the bagging measurements (BE in equation 8) since, in
the field, the bagging measurements are the reference against which the HVCS performance will
be assessed.
The accuracy of the screening measurements will be determined based on daily pre-test and
post-test multipoint calibrations for each portable analyzer and reported as the average difference
between the analyzer response and certified gas concentration.
3.4 REPRESENTATIVENESS
Representativeness is a measure of the extent to which a measurement or measurements
program represents the true quantities that are intended to be measured. Representativeness
is generally addressed in terms of the study design and sampling plan. In this study, the
objective is to evaluate the performance of the HVCS measurement systems and methodology
under field conditions at gas production facilities. Thus, it is important to challenge HVCS
performance for a range of component types, leak rates, and operating conditions that
74
-------
Section No. 3
Revision No. 0
Date: 5/18/94
satisfactorily characterizes leaking components at these types of facilities. The overall study
design, sampling plan, and field operations have been conceived with this goal in mind.
The expected population of leaking components at gas production fields is described in recent
screening work conducted by GRI (GRI, 1994) and API (API, 1993) at gas production fields in the
Eastern and Western United States. This work was conducted at 14 gas production fields in the
Louisiana and Appalachian regions of the Eastern United States, and 4 additional sites in the
West. In this work, over 65,000 components were screened and nearly 3,000 leaking
components were identified. The majority of leaking components (about 58 percent overall, 80
percent in the east, and 51 percent in the west) were among connectors. These were primarily
threaded couplings 2 to 4 inches in diameter. Most of the remaining leaking components were
valves (about 32 percent overall, 18 percent in the east, and 35 percent in the west). These were
primarily smaller valves (2 inches and below) with 1/4 inch to 1/2 inch valve stems. There were
also a fairly large number of leaking open ended lines pressure relief valves, and other
miscellaneous components. Operating pressures associated with the majority of leaking
components are 30 to 50 psi in the East and 200 to 400 psi in the West. The Louisiana sites are
high pressure fields (2,000 to 3,000 psi) where much larger components are used (8 to 20
inches). However, these sites contained a very small number of leaking components. In general,
valves and open ended lines were associated with the highest emissions (based primarily on
screening values).
The controlled leak test bed will be constructed to represent this general population of leaking
components (see section 4.5.1). The equipment inventory and screening values obtained in this
study will provide information that will be used to assess the representativeness of the HVCS and
bagging measurements in terms of the number and type of components, leak rates, and operating
parameters (e.g., pressure). The study report will include an analysis of representativeness
based on these data.
3.5 COMPARABILITY
Comparability is a measure of the confidence with which one set of data can be compared with
another. Comparability is generally established by ensuring that the data collected are of known
quality through the use of QA planning, calibrations and quality control checks, audits and the
determination of data quality indicators. Only validated data sets should be compared.
Comparability may also be established through the use of equivalent sampling and analysis
methods and procedures.
Many of the conclusions that will be drawn from this study will be based on comparisons between
HVCS and bagging data pairs for the same component. It is important that the comparability of
these data pairs is carefully established. For example, a change in leak rate between the two
measurements would introduce an unwanted (and unqualified) bias into the comparison. To
help ensure that the leak rate remains constant, the HVCS and bagging measurements will be
made as close together in time as possible (the goal will be within 2 hours). In addition,
screening values will be obtained before and after both bagging and HVCS measurements. Data
75
-------
Section No. 3
Revision No. 0
Date: 5/18/94
screening procedures will include examination of the associated screening values and sampling
times before the data are compared.
It is also important for this study that the conventional bagging and screening measurements be
conducted in a manner that is equivalent to similar measurements that have been conducted in
several previous studies involving many facilities, and hundreds of thousands of components.
To ensure that this is accomplished, sampling and analytical procedures will be consistent with
EPA standard reference methods (Method 21 and Method 18) and the EPA Protocol For
Equipment Leak Emissions Estimates where the bagging methods are set forth.
76
-------
Section No. 4
Revision No. 0
Date: 5/18/94
SECTION 4
SAMPLING AND ANALYTICAL PROCEDURES
This section describes the sampling and analytical procedures to be used in the performance
evaluation of the HVCS device and emissions measurement methodology. Detailed operating
procedures, where necessary, are provided in the appendices or incorporated by reference to
EPA Standard Methods and Protocol Documents. This section is subdivided according to the
different measurements that will be made for this study.
4.1 COMPONENT SCREENING
The primary purpose of the screening measurements in this study is to identify leaking
components and categorize them according to leak rate as indicated by concentration
measurements in the vicinity of the component with a portable hydrocarbon analyzer. These
measurements are known as instrument screening values or ISV's. Screening values will also
be obtained before and after each HVCS and bagging measurement. These data will be used
as an indication of whether the leak rate for that component changed during the measurement,
or between the HVCS and bagging measurements.
Up to 25,000 components will be screened at two to three sites selected by EPA in conjunction
with the American Petroleum Institute (API) and the Gas Research Institute (GRI). Experience
from recent studies at Gulf and Appalachian gas production fields conducted by STAR
Environmental on behalf of GRI indicate that screening 25,000 components can be expected to
yield 400 to 600 small leaks (screening value less than 1000 ppm), 300 to 400 medium leaks
(screening value less than 10,000 ppm), 200 to 300 large leaks (screening value greater than
10,000 ppm) and perhaps 20 very large leaks (screening value greater than 100,000 ppm).
Leaking components will be classified by equipment type, type of service, and process location.
4.1.1 Screening Method Description
EPA Reference Method 21 specifies requirements for the use of portable analyzers to identify and
categorize leaking components. Method 21 also gives procedures for operating a portable
analyzer and for obtaining representative screening values. The description of Method 21 in the
40 Code of Federal Regulations (CFR) Part 60, Subpart A is short (2 pages). A companion
document titled "Protocol for Equipment Leak Emission Estimates" (EPA-453/R-93-026) provides
more detailed instructions for how to screen specific component types. The instrument screening
value (ISV) is the maximum calibrated response (specific to a given calibration gas) of a portable
hydrocarbon analyzer when the inlet is placed in the vicinity of a VOC leak. The ISV can be
generally correlated with a mags emission rate as averaged over a relatively large number of
components; however, an ISV is not necessarily closely related to the leak rate for any given
component. Initially, a leaking component will be defined as yielding an ISV of more than 10
parts-per-million (ppm) above the ambient air response. This threshold may be revised if large
numbers of leaking components are identified at this level.
Method 21 allows that any portable instrument that meets general performance criteria can be
used. The Foxboro Model 108 Organic Vapor Analyzer (OVA) is commonly used for Method 21
77
-------
Section No. 4
Revision No. 0
Date: 5/18/94
screening and will also be used for this project. The Model 108 OVA was designed to be a
portable instrument. The main component which houses the controls, pump, detector, battery
and electronics, are worn on a strap over the operators shoulder. The sampling probe and
analog meter output are located on a remote sampling head which is tethered to the main
component by an umbilical. This permits the operator to view the response of the instrument
quickly as the screening work is performed. The range of the Model 108 is from 1 to 10,000 parts
ppm when the analyzer is calibrated for methane gas. The electrical signal output of the Foxboro
Model 108 is proportional to the log of the gas concentration in the range of 1 to 10,000 ppm.
The sweep of the analyzer's analog meter is 270 degrees with 100 ppm set at midscale in the
twelve o'clock position. This analyzer uses the flame ionization detection (FID) principle, which
is highly sensitive to trace organics in ambient air. In the FID, hydrocarbon containing molecules
are combusted in a hydrogen flame. This combustion of the hydrocarbon generates ions which
conduct a small (picoampere) electrical current. The current generated between the high voltage
positively charged burner tip and the negatively charged collector located above the burner tip
is responsive to the number of ions present.
Method 21 lists specific types of components such as valves, flanges and pressure relief vents
etc. that are to be checked for fugitive emissions. The method describes the particular
mechanical features of each component and highlights the specific areas to check with the
screening analyzer. In most instances, the VOC escapes from a mechanical seal within the
component. These would be the packing materials used around a valve stem, the seals around
a pump or compressor, or an open ended line downstream of a valve or pressure relief device.
Method 21 also describes proper measurement technique, such as the positioning of the analyzer
relative to the component.
4.1.2 Screening Test Procedures
The daily operation of the portable OVA 108 analyzers is described in this section. In accordance
with Method 21, the analyzers will be prepared for use by following the manufacturer's operational
instructions. The important features of these instructions include replenishment of the hydrogen
fuel for the analyzer's detector and positioning of the control switches for the power-up of the
pump and electronic components. The analyzer's detector is ignited by depressing the ignite
switch and watching the signal displayed on the analog meter for a positive short duration
response. After successful ignition of the detector, the analyzers can be disconnected from their
battery chargers. After ignition the analyzers should be allowed approximately ten minutes to
stabilize.
The daily calibration of the OVA will be performed by introducing a zero gas followed by a span
gas, with adjustment of the electronic gain of the amplifier so that the analyzer's display meter
matches the concentration of the span gas. Additional mid-point span gases at lower
concentrations than the first span gas can be used to verify the linearity and calibration of the
analyzer. No adjustment of the analyzers response is made to these mid-point span gases. The
first calibration of the day is a full multi-point calibration. Additional zero and single-point span
checks will be performed at the midpoint of the days activities, or whenever deemed necessary
by the operator. All data are recorded on standardized data sheets (see Appendix B).
78
-------
Section No. 4
Revision No. 0
Date: 5/18/94
At the start of each day, each operator will be assigned groups of components by the field
coordinator. A component inventory, process flow diagram, or production field map will be used
to assign the components and direct the operators to the correct locations. Approximately 50
components are expected to be screened per hour. The operator will record the data generated
during the screening activities on standardized field data sheets. The list of items on the field
data sheets will include the facility name, date, time, component and process ID from the process
map or inventory list, type of component and initial screening value. An example data sheet is
provided in Appendix B. At the completion of daily screening activities, the analyzers will be
recalibrated, serviced and returned to battery chargers for the night. The data sheets will be
reviewed by the field supervisor on a daily basis for completeness and accuracy.
4.1.3 Screening QA/QC Procedures
Prior to the start of screening activity, each analyzer will be tested to show compliance with the
performance criteria specified in Method 21. These criteria include a calibration precision test
with alternating zero and hi-span gases, measurement of sample demand rate and a response
time test. The precision test, sample demand and response time tests need not be repeated on
a daily basis, unless a mechanical change such as replacement of a pump has occurred and that
would require these tests to be repeated. Other checks such as battery condition, fuel supply and
condition of required safety equipment will be noted daily for project records as part of the
morning calibration. Problems will be noted on the log sheets, reported immediately to the field
coordinator, and appropriate corrective actions will be taken, including repair or replacement of
the analyzer.
Calibration gases of ą2 percent certified by the manufacturer are sufficient to meet the
requirements of this study as well as Method 21 and will be used. Copies of the gas certifications
will be maintained in project files, available for inspection. The calibration gas concentrations will
be approximately 100, 1,000, and 9000 parts per million of methane in hydrocarbon free air.
These calibration gases will be contained in high pressure cylinders and delivered at
approximately 1.1 times the total analyzer demand rate at atmospheric pressure using a flow-
through manifold. The analyzers will draw off their own supply of calibration gas and the excess
will be vented through a rotameter. The rotameter will be used to monitor and maintain
calibration gas flow rates through the manifold system. If an analyzer has drifted by more than
20 percent beyond its precalibration response, the data collected will be considered suspect.
Components will be flagged for re-screening if the drift was sufficient that leaking components
would have been incorrectly identified (at the 10 ppm level). In most cases, the drift will be
downward so, components with screening values close to 12 ppm would be re-screened. In other
cases, a downwardly biased screening value would result only in some components being
miscategorized, for example, from large to medium, or medium to small. This should not have
a serious adverse impact on the study unless large numbers of components are miscategorized.
The OVA 108 analyzers use the same sample air that is being analyzed by the FID detector to
combust the hydrogen flame in the detector. If a very high concentration source is suddenly
discovered during screening, the lack of oxygen in the sample stream will extinguish the flame.
If flame out of the FID occurs during screening activities because of momentary oxygen
starvation, the operators will perform only a single point calibration check using the 1000 ppm
79
-------
Section No. 4
Revision No. 0
Date: 5/18/94
span gas (this concentration is reliably read on the logarithmic analyzer readout). If the response
of the analyzer is within 10 percent of the span gas concentration, screening operations will
continue. If the response differs by more than 10 percent, the operator will return the analyzer
for a complete multipoint recalibration. Flameouts caused by moisture or contamination entering
the probe will require servicing the OVA analyzer. This situation will be apparent if the flame
cannot be re-ignited or if a major baseline shift occurs.
4.2 HVCS MEASUREMENTS
The High Volume Collection System (HVCS) has been developed by Star Environmental to
provide a rapid, simple, and cost-effective approach to getting direct measurements of mass
emission rates without the need for tenting and bagging. The field performance of the HVCS is
being tested in this project. This performance evaluation is based on direct comparisons of HVCS
measurements against EPA Protocol bagging measurements. (The tent and bag procedures are
described in the section 4.3) Emissions from a total of 200 leaking components will be measured
using the HVCS. 125 of these 200 components for will be selected for direct checks against EPA
protocol tent and bag procedures.
4.2.1 HVCS Description
In both the HVCS and bagging measurements, emissions from a leaking component are
determined by capturing the leaking gas in a known volume of carrier gas and measuring the
pollutant concentration in the carrier gas. The emission rate is the product of the flow rate and
the concentration. The HVCS generates a fast-moving slip stream of ambient air around a
leaking component. Fugitive emissions are entrained in this slip stream for measurement. The
HVCS is a portable dynamic dilution device that is fitted to the regular sample probe of the Model
108 OVA. It uses a battery operated pump coupled with three calibrated rotameters (2 to 20, 20
to 100 and 100 to 1000 scfh) to draw a known volume of ambient air across a leaking component
at flow rates between 10 and 500 standard cubic feet per hour (SCFH). Sample flow rates
through the rotameters are metered by means of valves on the three rotameters. The FID
measures the VOC concentration in the diluted air. The output of the FID is measured with a
small digital voltmeter attached to the recorder output jack of the Model 108. This voltage is
proportional to the log of the concentration. Obtaining the OVA output as a voltage from the
recorder jack allows for concentration measurements somewhat in excess of full scale (10,000
ppm) up to about 17,000 ppm to be obtained.
Measurements at more than one HVCS flow rate are performed to verify the capture efficiency
of the system. VOC concentration will increase as the airflow across the leaking component is
reduced. If all of the leaking gas is captured, the same leak rate should be obtained for different
air flows. To aid in the capture of the fugitives, the sample inlet of the HVCS probe is shaped
like mouthpiece of a snorkel and it is held in close proximity to the leaking component. The
capture efficiency for large (>4"pipe flanges) of the Star HVCS may be enhanced by shielding the
component with mylar film so that the airflow is directed across the entire leaking surface. Further
information on the HVCS is available in the Protocol For Using HVCS by Star Environmental (see
Appendix A). Star has also produced a draft report on preliminary laboratory and field testing of
the HVCS that contains some details of HVCS design and operation.
80
-------
Section No. 4
Revision No. 0
Date: 5/18/94
4.2.2 HVCS Test Procedures
This section describes the daily operation of the HVCS in the field. Detailed operating procedures
are presented in Star Environmental's Protocol for Using HVCS (Appendix A). Initial preparations
for using the HVCS include calibration of the rotameters against traceable flow standards (dry gas
meters) and Method 21 performance checks on the OVA. The HVCS testing will begin when
screening work has identified a sufficient number of leaking components. Daily Preparation of
the HVCS includes a leak check, and checking battery condition.
Before and after each HVCS measurement, a screening value for the component is obtained (see
section 4.1). Screening also serves to identify the location of the leak as indicated by the point
at which the maximum OVA reading is obtained. For HVCS measurements, multipoint
calibrations are performed before and after each emissions determination. This is to ensure that
the optimum accuracy of the portable analyzer is attained. Prior to making HVCS measurements,
background readings are taken in the vicinity of the leaking component. HVCS measurements
are then made at 3 flow rate settings. The first setting uses the highest possible flow rate,
yielding a concentration close to the lowest calibration standard. The other flow settings are
selected so that the concentration reading is close to the mid and low calibration standard values.
For very high leak rates, the range of the OVA (10,000 ppm) may be exceeded, even at the
highest flow setting. In such cases, the OVA 108 dilution probe adaptor supplied by the
manufacturer can be used in making the HVCS measurements. This provides dilution ratios
between 10 and 150 and extends the OVA range up to 1,000,000 ppm (100 percent). It is
important that the dilution air is drawn from a point away from sources of hydrocarbon. This
should be verified with a background reading near the dilution air inlet. Finally, the HVCS is again
leak checked at the conclusion of daily activities. Data will be recorded on standardized field data
sheets (see Appendix B).
4.2.3 HVCS QA/QC Procedures
The same QA/QC requirements that apply to the operation of the OVA 108 FID during the
screening work will still apply during the HVCS phase.
The rotameters used in the HVCS will be subjected to three point calibrations against a flow
standard (dry gas meter) traceable to a primary standard (spirometer) before and after the test
program to verify the engraved markings of the manufacturer. Care should be taken to avoid
particulate contamination of the HVCS inlet as any debris that clings to the rotameter ball or
inside walls affects the accuracy of the rotameter. Other visual qualitative checks of the HVCS
should be concerned with battery condition, sample line leaks and residual organic contamination.
If results of post measurement OVA calibrations indicate a change of more than 15 percent for
any point from pre-measurement results, the measurement will be repeated. If acceptable results
cannot be obtained, a problem is indicated. This will be documented and corrective actions will
be initiated.
During the field studies, at least 3 controlled leak rate tests will be conducted at each site (before,
during, and after testing). If there is more than one HVCS operator, each operator will complete
81
-------
Section No. 4
Revision No. 0
Date: 5/18/94
the same number of tests (3). These will be blind tests. The results will be used to assess
operator bias and to identify problems with the performance of the overall HVCS system.
The HVCS results will be compared against the bagging results for the same components as the
study is underway. It is expected that emission rates for the two systems should agree to within
about 20 percent. If the two methods do not agree to within 20 percent, every effort will be made
to independently verify the procedures and measurement accuracy of each individual method.
The results from the laboratory study will provide performance baselines for the two methods.
4.3 EPA PROTOCOL MASS EMISSIONS (BAGGING) MEASUREMENTS
The bagging measurements are the reference against which the field performance of the HVCS
system will be assessed in this study. Emissions from a total of 125 components will be
measured by bagging for direct comparison against HVCS measurement of emissions from the
same components. The "Blow-Through" bagging method recommended in the EPA protocol will
be used in this study. The "Blow-Through" method was selected primarily because of the simpler
sampling train. Nitrogen will be used as the carrier gas for the bagging measurements. This will
prevent the possibility of explosive atmospheres in the sampling bags. The EPA protocol
Document (Section 4.2) provides a detailed discussion of the comparative advantages of the
"Blow-Through", and the alternative "Vacuum bag" methods.
4.3.1 Bagging Measurement Description
Like the HVCS measurement, the bagging measurement is based on measuring a pollutant
concentration in a known flow of carrier gas. Essentially, the product of the flow rate and the
concentration is the leak rate. A detailed description of the "Blow-Through" sampling method is
given in section 4.2 of the EPA Protocol. In general, the leaking component and leak location are
first identified based on method 21 screening. The leak is enclosed by constructing a tent of a
non-porous film that is also chemically inert relative to the leaking compounds to be measured.
The tent is secured in place using adhesive tape, rope or clamps. The diluent gas is introduced
at one or more openings in the tent as necessary to ensure good mixing. Another opening is
used as the sampling port. The oxygen concentration in the effluent from the tent is measured
as an indicator of when the initial air in the enclosure has been displaced by the carrier gas.
Since it can take a long time to completely displace the air in the tent, the protocol provides for
correcting the effluent flow rate based on the remaining oxygen level at the time of sampling. The
tent should be purged long enough for oxygen levels to fall below 5 percent (by volume). Percent
oxygen levels will be monitored using a portable infrared analyzer (Geo-Group Infrared Gas
Analyzer). Analysis of the sampled hydrocarbon is performed by gas chromatography (see
section 4.4).
82
-------
Section No. 4
Revision No. 0
Date: 5/18/94
4.3.2 Bag Sampling Procedures
Detailed sampling procedures for the "Blow-Through" method are given in section 4.2 of the EPA
Protocol. These will be followed closely. Analytical procedures for the bagging measurements
are discussed in section 4.4. The following is a general description of the sequence of events
in the bag sampling procedure. First, the selected component is re-screened following Method
21 to obtain a new screening value and identify (as closely as possible) the exact location of the
leak. A tent is then constructed over the area of the leak. Carrier gas is introduced at a steady
flow rate. The flow rate should be sufficient to establish an equilibrium with the leaking gas in the
bag, but not so large that the hydrocarbon concentration in the samples is below the detection
limit of the analyzer. The portable OVA used for screening can be used to monitor the
hydrocarbon concentration in the effluent from the bag while selecting a suitable flow.
Concentrations should be within the calibration range selected for the analyzer (see section 4.4).
In general, flow rates of less than 60 Ipm will be used. Sufficient time is allowed for mixing of the
carrier gas with the pollutant in the bag and for purging of ambient air remaining in the bag.
Mixing can be verified by probing at various locations in the bag with the portable analyzer.
Oxygen levels are monitored to determine when sufficient purging has taken place. Once the tent
is equilibrated, two Mylar sample bags (1-2 liters) are filled from the effluent stream using a
portable sampling pump and transported to the field laboratory for analysis. These bags will be
nitrogen purged before use. Analysis should take place within 2 hour of sample collection. All
readings, sampling conditions, and notes are recorded on standardized data forms. One form
is used for each sample.
4.3.3 QA/QC for Bag Sampling
Quality assurance considerations and quality control checks for the bag sampling protocols are
addressed in detail in the EPA Protocol. Some additional general considerations are mentioned
here. Leak enclosures do not have to be leak tight, but should be tight enough to ensure that
equilibration is maintained. In situations where leaks are suspected, the portable analyzer can
be used to screen the edges of the enclosure. Enclosures and sampling bags will be constructed
of Mylar. In most cases, enclosure materials will not be re-used; however sample bags will be
re-used. Sample bags will be purged after each use at least twice with clean nitrogen gas.
Purge gas from a portion of the bags will be analyzed by GC to verify the integrity of the purging
procedure. During purging, the sample bags will be checked for leaks, and leaking bags repaired
or discarded. The integrity of the diluent gas in each bottle purchased will also be verified by GC
analysis. Sample bag preparation activities will be logged.
Field leak rate checks will be conducted at least 3 times per field trip by each bagging technician.
Field leak rate checks will be conducted by performing the complete bagging operation on a
component with a controlled leak rate as described in section 4.5. These will be blind checks.
These checks will test the overall sampling, analytical, and field operation components of the
measurement. Corrective action will be called for when the field leak rate test is unable to
reproduce the true leak rate within 20 percent.
Quality assurance procedures also include calibrations of all measuring devices used in the
procedure. Flows will be measured using a calibrated rotameter (4 to 40 Ipm). The rotameter
will be calibrated at the beginning and end of the field program against a dry gas meter traceable
83
-------
Section No. 4
Revision No. 0
Date: 5/18/94
to a primary standard (spirometer). The response of the oxygen meter will be checked against
a zero (pure nitrogen gas) and standard atmospheric oxygen levels (21 percent). Factory
calibrations will be relied on for the temperature probe. Calibration of the GC is addressed in
section 4.4.
4.4 GAS CHROMATOGRAPHY
The 125 bagged samples collected during the mass emission sampling will be analyzed by gas
chromatography to determine methane and total hydrocarbon concentration. Fifty of these
samples will be further speciated to determine light hydrocarbons that comprise the bulk of the
material found in gas production facilities (ethane, propane, and n-butane). This will allow for a
more accurate determination of the actual emission rate as opposed to determining the emissions
solely as methane. EPA Reference Method 18 is frequently used for source emission and
ambient air testing to speciate and quantify selected organics using Gas Chromatography (GC).
The description and procedures for Method 18 in the Federal register are basic procedures
considering the wide variety of techniques available to skilled chromatographers. Reference
Method 18 will be used in this project as a basis for all analyses by gas chromatography.
4.4.1 Gas Chromatography Description
In gas chromatography individual compounds from a sample are separated according to selected
physical properties and introduced sequentially to one or more detectors. A small quantity of a
sample mixture is injected onto the inlet of a chromatographic column. An inert carrier gas flows
through the column and the interaction of a carrier gas working against the effect of the column
causes the compounds to separate and elute at discrete intervals over a given time period. The
sequential elutions of compounds are sensed by a calibrated detector and the output is directed
to a recording device or integrator. The integrator displays the chromatogram as a series of
spikes rising above a baseline and calculates an area for each of the peaks. The retention times
of the compounds are given by the integrator and these are used to identify the compounds.
Commonly used detectors include the Flame lonization Detector (FID) which is highly specific and
linear for part-per-million level organics and the Thermal Conductivity Detector (TCD) which is
sensitive to percent level changes in gas compositions. Quantifications are based on compound
specific calibrations of detector response.
EPA Reference Method 18 lays out general guidelines for developing specific sampling, analysis
and quality assurance procedures for a given measurement program. The procedures for Method
18 are given in 40 CFR Part 60, Subpart A. The quality assurance guide for Method 18 is part
of the EPA Quality Assurance Handbook for Air Pollution Measurement Systems, Volume III for
Stationary Source Specific Methods (EPA-600/4-77-027c). These documents will be used
directly as a guide for chromatographic measurements conducted for this project. Method 18
permits different analytical components such as detectors and chromatographic columns, as well
as different sampling techniques/interfaces to be used. The method allows for expert judgement
on the part of the GC operator in selecting the most appropriate equipment and procedures.
Details of the Method 18 configurations intended for this project are presented below as part of
the test procedures.
84
-------
Section No. 4
Revision No. 0
Date: 5/18/94
4.4.2 Method 18 Test Procedures
Method 18 describes a start-to-finish approach for conducting gas chromatographic analysis of
samples starting with a pre-survey sampling phase and an analytical development phase, before
the actual measurement phase. The compounds of interest in this project are methane and other
light hydrocarbons. Chromatographic methods for methane analysis are well established with
multiple published references. Thus, there is no need to conduct presampling and analytical
developmental phases for this project.
The GC to be used in this project is a Hewlett Packard Model 5890 Series II GC. The data
output from the GC will be directed to a Hewlett Packard Model 3396B dual channel integrator.
The integrator accepts data simultaneously on two channels, displaying real-time data of one
channel and electronically buffering the second channel for immediate printing after the
completion of the sample run. The FID and TCD detectors intended for use in this project will
permit the analysis of gas species over the widest of anticipated concentrations, from high part-
per-billion concentrations of organics to percent concentrations with the fixed gases.
Chromatographic columns have been selected according to published guidelines and vendor
literature with prime consideration given to the resolvability of the expected compounds and
durability under adverse conditions. The packed type chromatographic columns selected for this
project are Haysep D porous carbosieves in 10' by 1/8" and 30' x 1/8" metal tubing. The carrier
gas will be high purity helium. Column flows and programmed oven temperatures will be decided
by initial test experiments to give the best compound separations with rapid sample turn-around-
time.
A Valco brand ten port gas sampling valve with two independent fixed size sample loops will be
used on the GC to introduce the sample simultaneously to different packed columns and both
detectors. The advantage in using the gas valve and sample loop is the elimination of manual
syringe injections and associated imprecision. The zero plus three upscale calibration gases used
for the GC/FID will be the same 100, 1000 and 9000 ppm methane in air cylinders used for the
Method 21 OVA analyzers. Higher concentrations of methane in air, if needed for the FID, will
be obtained using dynamic blending of pure standards and diluent gases as described in Method
18. Method 18 does not specify a detection range of target calibration gas concentrations, but
only requires that the concentrations of calibration standards bracket the expected sample
concentrations. The linearity of the FID, over a wide range of concentrations, will be established
using multiple methane concentrations. Factory specifications list FID midscale error at less than
10 percent over a range of 107 in concentration. Since the target test facilities for this study are
natural gas production facilities, methane is the component of most importance. The FID will be
calibrated for other light volatile organics (ethane, propane and n-butane) using two additional
standards. The calibration of the GC will be accomplished by performing a least squares linear
regression of the GC response or area counts (AC) against the concentrations of the calibration
standards in parts per million (ppm). Estimation of the minimum detectable limits of the GC will
be the intercept generated by the least squares linear regression divided by the slope of the
regression. Calibration of the TCD will be accomplished using dynamic dilution of air or nitrogen
or methane with helium to generate three different percent concentration standards. Factory
specifications list TCD midscale error at less than 5 percent over a range of 105 in concentration.
Multipoint calibration of the GC/FID/TCD will be performed daily, before and after the sample
analyses. A midday zero and span check will also be performed each day.
85
-------
Section No. 4
Revision No. 0
Date: 5/18/94
Field samples will be introduced to the GC using Mylar bags because methane exhibits virtually
no affinity or reactivity with the Mylar surfaces. Proper operation of the GC requires that the
samples be introduced to the system using the same technique in which the calibration gases are
introduced. Calibration gases or samples will be drawn or pushed through the valve sample loops
using a downstream pump. This will be done for approximately 60 seconds while the GC is idling
between runs. Sample flow through the sample valve will be monitored with a small exit
rotameter. The pump will be powered-off and the pressures in the sample loop allowed to
equalize with atmospheric. Ten seconds will be allowed to elapse after the rotameter ball has
returned to zero, thus indicating zero flow through the sample valve. The GC operator will
manually and simultaneously activate the sampling valve from Load to Inject positions and
depress the GC's start-run button. The GC will output the chromatogram to the integrator. The
sample identification and any other pertinent data will be noted on each chromatogram. The GC
will stop automatically at the end of a preset time interval. Following completion of the GC run,
the gas valve will be returned to the load position to repeat the process. Run times will only be
long enough to elute all of the compounds of interest to permit a rapid analytical turn-around.
In all bagged samples, Method 18 will be used to report methane concentrations in ppm. This
will be accomplished by subtracting the least squares regression intercept from the methane area
counts and then dividing the remaining methane area counts by the least squares regression
slope. The total hydrocarbon (as methane) concentrations for the 125 bagged samples will be
reported using the same procedure except that the total area counts for all other peaks plus
methane will be used. For the samples that require speciation, the area counts for each
individual compound, as identified by the retention times, will be converted to ppm using the
appropriate least squares intercept and slope.
4.4.3 Method 18 QA/QC
QA/QC guidelines for the Method 18 procedures used in this project can be found in the
previously mentioned QA Handbook. Method 18 spells out the number of calibration points (3
upscale that bracket sample concentrations) and the allowances for analyzer drift (10 percent)
and calibration error (7 percent curve fit on individual points). GC drift checks are easily
performed by re-analyzing selected calibration standards. The normal procedure when using
bagged samples is to analyze the samples within two hours of collection. A degradation study
will be conducted at the beginning of the project to verify the non-reactivity of the target
compounds with the Mylar bag surfaces. The use of external audit cylinder standards for this
project is neither anticipated or needed. In addition to the pre- and post-calibrations, periodic
calibration checks of the GC/FID/TCD will be interspersed with the samples as deemed necessary
by the operator to ensure calibration accuracy (at least once per day). If the GC/FID/TCD shows
more than a 10 percent relative standard difference between a calibration standard and the
calculated value through the calibration curve, then the GC will be subjected to a complete
multipoint recalibration.
The setup of the GC and its operating parameters such as column flow and programmed
temperatures will be determined before field testing so that baseline separation of eluting
compounds on the chromatogram exhibit as nearly complete baseline separation as possible
(maximum 5 percent peak overlap at baseline). Identification of peaks will be accomplished by
86
-------
Section No. 4
Revision No. 0
Date: 5/18/94
injecting analytes singly to establish retention times and the sequential order of elution.
Calibration of the GC will be required if the retention times shift by more than 0.5 minutes.
Copies of certification sheets for compressed gas calibration cylinders will be included with the
report.
4.5 CONTROLLED LEAK TESTING
A careful, controlled leak study is an important element of the overall experimental design. The
primary function of the controlled leak study is to assess the precision and accuracy of the overall
bagging method under controlled laboratory conditions. This is necessary because the bagging
measurements will be used in this study as the standard against which the performance of the
HVCS method will be assessed. Since this is a field evaluation, it is also important that the
laboratory testing closely simulate "real world" sampling conditions and include a representative
range of component types, leak rates, and operating conditions. The controlled leak study will
be performed on a test bed containing a representative range of component types and sizes, with
apparatus for precisely metering and controlling the leak rate of a known concentration (99.9
percent) of methane. The test bed will also be capable of simulating different types of leaks (e.g.,
jet or diffuse leaks). The test components will be modified to create a leak, for example, by
loosening the packing nut on a valve stem, or scoring the gasket on a flange. Components will
be tented and samples bagged and analyzed according to the same procedures to be used in
the field. The test bed will also be used to evaluate the performance of the HVCS under
controlled conditions.
This testing will be more representative of field conditions than the leak rate tests discussed in
the EPA protocol or conducted in earlier studies for EPA (EPA, 1980c). Sections 2 and 3 of this
plan show exactly how the data from the controlled leak test will be applied to the overall analysis
and quality assessment of the data for this study. The test bed will contain a total of 5
components which will each be tested at up to 5 different leak rates in the range from 0.1 to 20
Ipm, and over a range of pressures (up to 100 psig).
4.5.1 Controlled Leak Test Bed Description
Past extensive screening and mass emission measurements at gas and oil production fields
provides a rich data set from which representative populations of the leaking components (leak
configurations) can be characterized in terms of the component type and size, leak rate, and
operating pressure. The population of leaking components at U.S. gas production fields is
characterized in section 3.4. Based on this characterization, the following components and leak
types were selected for the test bed:
4" threaded connector, (1) not tight, (2) threads scored
2" gate valve (1/4 to 1/2 inch shaft), (1) packing not tight, (2) packing damaged
2" pressure relief valve, (1) seal damaged
4" open ended line
4" flange, (1) not tight, (2) gasket scored
87
-------
Section No. 4
Revision No. 0
Date: 5/18/94
Where feasible, larger components were selected in order to present more of a challenge to the
methods. The cost of components is also a consideration in their selection.
The challenge in constructing the test bed is to accurately meter the flow rate while leaving the
test components under simulated field conditions. It is not generally possible to meter the leak
rate downstream of the test component as this would limit the types of leaks that could be
constructed and could interfere with the bagging and HVCS measurements. Use of common flow
metering devices such as bubble meters and rotameters for upstream flow metering is
complicated by the need to achieve pressures internal to the components as high as 100 psi.
A bubble flowmeter measures the true volume of the gas at the given internal component
pressure (e.g, 30 psi). A simple pressure correction would give the actual volume of the leak at
ambient (or standard) conditions. However, commercially available bubble flowmeters are
designed to be operated at ambient pressures and might not withstand elevated pressures. A
rotameter reading depends on the density of the gas passing the float, and should therefore be
calibrated specific to the gas that will be metered (methane) and the pressure at which it will be
used. Pressures necessary to achieve a desired leak rate (over the range of 0.05 to 20 Ipm) in
a given component are expected to vary widely (over the range from 1 to 50 psi).
The approach to metering test bed flows is based on the use of rotameters calibrated over a
range of pressures. Before the test, the rotameters will be calibrated over a range of pressures
using the apparatus illustrated in Figure 4-1. A valve will be used to control the pressure
downstream of the rotameter and a bubble flow meter (primary standard) will be used (at ambient
outlet conditions) as the flow standard. Rotameter flow as a function of pressure and temperature
is expressed by:
<14'
where, 1 is initial conditions, 2 is final conditions, Q is flow, P is pressure, and T is temperature.
However, it is generally recommended that rotameters be calibrated under the exact conditions
under which they will be used. To accomplish this a family of calibration curves will be generated
for various flows over a range of pressures. A sufficient number of curves will be developed so
that there is no more than a 10 percent difference in flow between pressure curves. During the
tests, pressure will be monitored and rotameter flow will be determined from the appropriate
calibration curve for the pressure during that test to obtain the true flow under standard
conditions. The test bed flow metering scheme is also illustrated in Figure 4-1 .
Two parallel series of tests will be conducted on each test component; one for the bagging
method, and one for the HVCS method. Each series will consist of tests at at least 3 leak rates
and up to 3 internal pressures. The controlled leak test matrix is given in Table 1-2. Once a leak
rate is established, it will be measured by both methods. Selection of leak rates and pressures
for each component will depend on the component and leak type. The goal is to test leak rates
over the range of 0.1 to 20 Ipm under pressures up to 100 psig. However, as a practical matter
it may not be possible to achieve this range for each component. It is expected, however, that
the overall range will be achieved for the test bed as a whole. All components will be connected
-------
Section No. 4
Revision No. 0
Date: 5/18/94
CH4
Rotameter
01-1 1pm
02-4 1pm
3.0-20 1pm
(100psi
rated)
Electronic Bubble Flow Meter
100ml to 30 Ipm or
Bubble Cylinder 10ml to 1 Ipm
CALIBRATION SET-UP
CH4
Pressure Gauge
;/N
Rotameter
0.1 -1 1pm
0.2-4 1pm
3.0-20 1pm
(100psi
rated)
Leak to Ambient
J
4" Threaded
Connector
(Diffuse Leak)
_ 1
\
1
_
4" Pipe
Cap
LEAK TEST SET-UP
Figure 4-1. Flow Metering Schematic For Controlled Leak Test Bed
89
-------
Section No. 4
Revision No. 0
Date: 5/18/94
to the flow metering apparatus using caps, reducers, and other fittings as needed. The entire
system should be leak tight except for the intended area of the leak. This will be verified by
screening with soap solution.
For the field leak rate tests, only one component from the test bench will be used. The selected
component will be capable of achieving a wide range of leak rates and should be a challenging
(but not too difficult) test for both the bagging and HVCS methods. The threaded coupling is
probably the best candidate for use in the field leak rate tests since the majority of leaks identified
in recent studies have been associated with this type of component.
4.5.2 Controlled Leak Test Bed Procedures
Measurements of emissions from the simulated leaking components will be conducted in the
same manner as the field measurements. As in the field, screening values will be obtained
before and after each HVCS and bagging measurement. HVCS and bagging measurements in
the controlled leak study will be conducted according to the identical procedures described for the
field study in sections 4.2 and 4.3 above. The standard field data forms will be used. The only
additional procedures are metering the controlled flow rates and calibrating the flow metering
apparatus. These procedures were described in the test bed description above. The calibration
curves generated for the test will be included in the study report.
4.5.3 Controlled Leak Test QA/QC
QA/QC for the OVA screening, the HVCS and the bagging measurements will be the same as
described above (see sections 4.1-4.4). The methods to operate and calibrate the flow
measurement devices (rotameters and bubble flow meters) are listed in EPA stationary source
and ambient air method descriptions. Details may be found in found in several sections of the
EPA QA for Air Pollution Measurement Systems Handbooks (the Redbooks). One such write-up
is included in Volume III Section 3.4 for EPA Reference Method 5-Determination of Particulate
Emissions from Stationary Sources.
90
-------
Section No. 5
Revision No. 0
Date: 5/18/94
SECTION 5
DATA REDUCTION, VALIDATION, AND REPORTING
This section presents a brief description of methods and procedures for data reduction, validation
and reporting as applicable to key measurement parameters. These key data elements are (1)
the instrument screening values and component descriptions gathered during the screening, (2)
flow rates and total hydrocarbon concentration values for leak rate determinations by HVCS and
bag sampling, and (3) concentrations of speciated compounds. Other supporting information
includes the component inventory or process description, and results from calibrations and quality
control checks. Figure 5-1 shows a generalized data flow diagram illustrating how key data
elements feed into the comparative analyses which will be the focus of the data analysis. This
Figure follows the study design overview presented in section 1.3.
5.1 DATA REDUCTION
Descriptions of necessary calculations to determine leak rates are given in section 4 of this plan
or are referred to in the detailed procedures found in the appendices to this plan and in
referenced EPA documents. The primary measurement quantities and units for this study are
hydrocarbon concentration (volume ratio, e.g., percent, ppm), flow (Ipm), temperature (Kelvin),
and pressure (Torr). Some measurement devices that will be used provide direct readings in
English units (e.g., HVCS rotameters read in standard cubic feet per hour). All final data will be
reported at standard conditions (760 torr, 296 °K) in metric units. Standard unit conversions will
be used as necessary. Computerized spreadsheets will be used to automate calculations and
provide a high level of consistency.
Descriptions of calculations to produce summary results that will be used as data quality
indicators are given in section 3 for each measurement. Other summary calculations will include
totals and standard descriptive statistics for measures of central tendency (mean, median, and
mode), and variability (standard deviation, variance, range).
5.2 DATA VALIDATION
The QA/QC procedures for each measurement presented in section 4 in this plan describe tests
and checks designed to verify that all sampling and analytical procedures, apparatus, and
instrumentation are functioning properly. These procedures also spell out conditions that call for
corrective action and that may indicate that collected data should be considered as suspect or
invalid. These conditions are summarized for each measurement in Table 5-1. Suspect data can
not be completely relied upon, but may contain useful information. For invalid data, the quality
of the data has been so severely compromised that it is of no use to the study. In general, a non-
valid (invalid or suspect) status would apply to all data collected subsequent to the last successful
calibration, check or test. In all cases, failure to meet such conditions will be documented on
standardized field data forms, calibration sheets, logs, and problem reports. All data collected
for this study will be examined to determine its validity.
91
-------
Section No. 5
Revision No. 0
Date: 5/18/94
Component Inventory
or Process Desc.
Screening
ISVs, Component Desc.
Screening Data Review
Select Leaking
Components For Testing
HVCS
Flow
Bag Sampling
Flow
OVA
THC Concentration
GC
THC/Speciation
*ŧ- Leak Rate HVCS
Leak Rate Bagging
(Speciated Data)
Final Data Review
Overall Comparative
Analyses
HVCS Performance
Criteria
Data Quality Reporting
Leak Type
Leak Rate
Component Type
Operating Conditions
Indicates field data review and corrective action decision point.
Figure 5-1. Generalized Data Flow
92
-------
Section No. 5
Revision No. 0
Date: 5/18/94
Data validity will largely be determined by association with operational problems that are known
to have occurred during sampling or analysis based on checks and tests as described above and
on any additional observations made by the operators. Operational data validation will be based
on examination of field data sheets, calibrations, logs and problem reports. An exploratory or
statistical data validation approach will also be used. This consists of examining the end data to
look for impossible values (e.g., negative concentrations), extraneous values (outliers) or unusual
patterns. In general, an outlier is no more than an observation that does not conform to the
pattern of other observations in the data set. When an outlier is suspected, operational evidence
explaining the presence of the outlier will be sought. If no operational evidence is available,
appropriate statistical outlier tests will be used as necessary to investigate the significance of the
outlier. If the apparent outlier cannot be explained based on operational evidence, or clearly
shown to be an outlier in a statistical sense (e.g., at a 95 percent significance level), then the
datum must be retained. However, subsequent data analysis may treat the data with and without
the "outliers" as separate cases.
5.3 Data Reporting
The study report will describe the findings of the field and laboratory studies and document the
means by which these findings were obtained. SRI will use the data collected to assess the
performance of the HVCS and identify factors that affect its performance. The 125 HVCS/bagging
data pairs are key data elements that will be analyzed to assess the overall performance of the
HVCS system. The screening data and component inventory will be used to demonstrate the
overall representativeness of these comparisons in terms of component type, leak rate, and
operating conditions encountered in the field. The methane/non-methane and speciated data will
provide information that will be used to estimate the actual leak rates in terms of the compound
specific FID response for the compounds actually present in the gas stream.
In addition to an assessment of overall HVCS performance, the data analysis will include a break
down of HVCS performance for different component types, leak rates, and operating conditions
(e.g., gas or liquid service, and pressure). The operating conditions of the HVCS (e.g., flow rate,
measured concentration range) will also be considered. Based on these analyses, SRI will
assess the accuracy of the HVCS under field conditions and establish performance criteria for use
of the HVCS in the field.
The data collected for the study will also be used to compare overall emissions estimates based
on HVCS, tenting and bagging, and EPA protocol emission factor and correlation approaches.
In so doing, it is important to note that the EPA correlations and emission factors should be
applied only to groups of similar components and not to individual components. This is because
emission rates determined using the emission factors and correlations in the EPA protocol
represent the average emissions expected if a large number of similar components were
measured. According to the EPA protocol, the actual mass emission rate for a given component
may be more than an order of magnitude greater or less than the estimated emissions, depending
on the type of equipment and the operating parameters.
93
-------
Section No. 5
Revision No. 0
Date: 5/18/94
Table 5-1. QA/QC Test Conditions, Data Validity and Corrective Actions
Condition
Validation Status
Corrective Action
Screening Measurements
OVA fails Method 21 performance
tests.
OVA fails daily operation checks.
Repeatability check failure.
OVA fails drift check.
OVA flame out.
Test conducted prior to measurements.
Data not affected.
Test conducted prior to measurements.
Data not affected.
Collected data suspect.
Collected data suspect.
Data may be suspect or invalid if flame
out is due to moisture or contamination.
Repair or replace analyzer. Repeat test.
Repair or replace analyzer. Repeat test.
Recalibrate OVA. Examine screening
method.
Flag components for re-screening as
needed. Recalibrate OVA.
Perform single point check. Continue
screening or return for multipoint as
indicated.
HVCS Measurements
HVCS fails leak check.
OVA fails drift check.
Field leak rate test failure.
Test conducted prior to measurements.
Data not affected.
Data invalid.
Data may be suspect or invalid
depending on cause of failure.
Repair. Repeat check.
Recalibrate. Repeat measurement.
Determine and correct cause of failure.
Bagging Measurements
Field leak rate test failure.
Data may be suspect or invalid
depending on cause of failure.
Gas Chromatography
GC fails drift check.
Sample hold time exceeded.
Data invalid.
Data suspect.
Determine and correct cause of failure.
Perform multipoint calibration. Re-
analyze samples.
Collect new samples as needed.
The report will also include an assessment of overall data quality based on the calibration,
laboratory test results, and other QA/QC data collected for the study. Measurement precision and
accuracy will be reported for all portable FID's, and the GC based on frequent comparisons with
known standards. In addition, an assessment will be made of overall operational precision and
accuracy for the HVCS and tent/bag sampling and analytical methods based on the controlled
leak testing in the laboratory. All reported data will be carefully screened based on a study of
field logs, calibration records, and problems and corrective actions. The results of this screening
will be included in the report. Screened data will be clearly identified and the justification for
removing any data will be stated.
94
-------
Section No. 5
Revision No. 0
Date: 5/18/94
Finally, the report will include summary tables of data collected for the study. These tables will
include (1) component identification/description and screening values for all components with
screening values over 10 ppmv, (2) calibration results for all portable FID's, (3) calibration results
for the GC, (3) laboratory test results for the HVCS and bagging measurements. Raw data will
be recorded and maintained on standardized field data forms (see Appendix B). These forms will
be kept on file and made available for inspection upon request. Other information that will be
maintained on file will include calibration gas certifications, and problem logs. To expedite data
analysis and preparation of summary tables and graphs for data presentation, much of the field
data will also be entered and stored in machine readable data tables in standard PC compatible
spreadsheet and database formats (Lotus 123, Paradox).
95
-------
Section No. 6
Revision No. 0
Date: 5/18/94
SECTION 6
AUDITS
There is no plan to make use of an independent auditor beyond the staff principally involved in
the study. However, quality assurance tests built into the study as part of this plan do perform
an auditing function. The blind field leak rate tests serve as an internal performance audit on the
HVCS and bagging measurements procedures. The field leak rate tests will also serve as a
"blind" check on total hydrocarbon determinations by GC/FID/TCD. The repeatability checks
serve as an audit of the screening measurements. Assurance that these procedures have been
followed will be provided by documentation of the results of these tests. While there is no
provision for a technical systems audit, this quality assurance plan has received internal and
external review and approval before its implementation in the field. Assurance that the plan is
followed is provided by the "paper trail" of documentation for the study and the content of the
study report.
96
-------
Section No. 7
Revision No. 0
Date: 5/18/94
SECTION 7
CORRECTIVE ACTION
This section presents procedures that will be followed to ensure that appropriate corrective
actions are implemented in response to any problems encountered in the laboratory or field
testing. The QA/QC procedures given for each measurement in section 4 of this QAPjP also
state conditions requiring corrective action. These conditions are summarized in Table 5-1
(presented in section 5 above). Corrective action procedures will be initiated when any of these
conditions are encountered. In addition, the field technicians who will be employed in the testing
are trained to recognize unanticipated problems. Corrective action procedures can be initiated
by any field or laboratory personnel if they feel that a problem that could compromise data quality
has occurred or is likely to occur.
Corrective action procedures begin when a problem is reported to the field coordinator. Problems
should be reported by the person directly responsible for the measurement. The field coordinator
will then initiate corrective action procedures. Corrective action procedures consist of (1) clearly
identifying the problem, (2) identifying and implementing a solution, (3) verifying that the problem
is solved, (4) assessing the impact of the problem on data quality (5) implementing steps to flag
and/or recover data, and (6) documenting all steps taken in the problems log. For example, a
portable OVA being used for screening fails a drift check. As called for in the QA/QC procedures,
the OVA is returned for checkout and a multipoint calibration. During checkout, the drift is verified
and it is discovered that the drift was probably caused by a low battery charge. It is verified that
the battery had been properly charged the night before. The battery is replaced with a new
battery and the OVA is subjected to a calibration precision test over a period of several hours.
Drift is now within acceptable limits and the OVA is put back into service. Screening data
collected prior to the failed drift check and subsequent to the last successful drift check is flagged
as suspect and examined. Components with low screening values near the leak criteria (10 ppm)
are flagged for re-screening since there is the potential that one or more of these is leaking, but
was not identified as such during the initial screening.
97
-------
Section No. 8
Revision No. 0
Date: 5/18/94
SECTION 8
PROJECT ORGANIZATION
Mr. Stephen Piccot will serve as project manager and, as such, will have overall responsibility for
SRI's technical and budget performance. Mr. Eric Ringler will serve as principal investigator. Mr.
Ringler will oversee all routine project activities and will provide direct support for preparation of
the QAPjP, field measurements, data analysis, and report preparation. In the field, Mr. Ringler
will serve as the field coordinator. Mr. John Sokash will coordinate mobilization for the field
measurements, conduct laboratory testing of the HVCS and bagging methods, perform GC
analyses, and support data analysis and preparation of the QAPjP. In addition, SRI will employ
a field technician trained in Method 21 screening and EPA protocol tenting/bagging for field and
laboratory measurements activities. Internal quality assurance will be provided by Larry Felix in
SRI's Birmingham office. Mr. Felix will review the QAPjP and will also review QA/QC data
collected during each measurements trip. Mr. Felix will have no direct project involvement other
than for internal QA, and does not report within the direct management of the project. SRI will
employ a subcontractor to conduct screening and HVCS measurements and provide support for
evaluation of the HVCS. SRI has concluded that the developers of the HVCS method, STAR
Environmental, would be the ideal subcontractor. STAR'S experience in developing the method
and in the initial field and laboratory testing are essential for the successful completion of the
project in a timely and cost effect manner. STAR also has extensive experience with
measurement of fugitive VOC leaks at gas and oil production facilities
98
-------
Section No. 9
Revision No. 0
Date: 5/18/94
SECTION 9
TEST PROGRAM HEALTH AND SAFETY
This section discusses Health and Safety practices for the laboratory and field components of this
project. All safety issues start with the recognition of potentially dangerous conditions and require
the best possible avoidance of them. This section cannot cover every conceivable issue that
may be encountered. Personnel will be trained in hazard recognition and avoidance.
9.1 GENERAL SAFETY ISSUES
All personnel on this project will be given a copy of this safety section for reference, and be
required to read and sign-off on this Section 9 of the QAPjP before being permitted to work on
this project. Serious safety hazards such as exposure to dangerous or carcinogenic chemicals,
radioactive substances and biological agents are not expected. Material Safety Data Sheets
(MSDS) for those materials supplied by vendors will be kept in a project file for ready reference
by test personnel. All personnel are ultimately responsible for their own safety and the safety of
others. Any unusual incidents or potential safety problems should be brought to the attention of
the field coordinator in a timely manner. All personnel will report injuries to the project
supervisors no matter how insignificant.
One universal hazard in this project involves the use of and potential exposure to flammable
gases, such as methane. The upper and lower explosive limits of methane are 15.4 and 5
percent by volume at normal atmospheric pressures and oxygen concentration. The explosive
limits of methane decrease as the pressures of the storage vessel increases. As a safety
concern, no blends of calibration gases or collection of samples will be permitted to have methane
concentrations below 25 percent, or above 1.25 percent. The volumes of samples collected in
bags for later analysis will be limited to a maximum of two liters. No open flames or sources of
ignition will be permitted in any area where methane vapors may be present. No spark producing
devices will be used on this project. Spark producing devices include any steel-on-steel striking
or contacting devices. Other common spark producing devices which field personnel must
recognize include series wound motors which are identifiable by the use of carbon brushes
between the field and armature windings; and shaded pole motors with mechanically activated
starter windings which are identifiable by a clicking sound as the motor starts or decelerate Any
motors used in the vicinity of potentially flammable methane atmospheres must be labeled as
intrinsically safe. If there is any doubt as to the intrinsic safety rating of a device, it will be kept
distant from sources of flammable gas, or sealed in a container and kept under a nitrogen purge.
Compressed gas cylinders will be used in this project. Compressed Gas Manufacturers
Association (CGA) Guidelines will govern the storage, use and handling of compressed gas
cylinders. A complete copy of the Handbook of Compressed Gases will be available, and the
section detailing use and handling will be given to all test personnel. Cylinders will be kept
secure at all times and not be permitted to strike each other. Regulators attached to cylinders
will be leak-checked upon initial installation. A secondary risk of compressed gas cylinders
involves the possibility of simple asphyxiation within confined spaces as normal oxygen levels are
displaced. No uncontrolled release of a compressed gas will be permitted.
99
-------
Section No. 9
Revision No. 0
Date: 5/18/94
The use of any electrical devices will be in accordance with National Electric Code (NEC)
guidelines. The use of Ground Fault Circuit Interrupting (GFCI) devices will be required whenever
moist surfaces or water may be encountered, and at all times when outside of permanently
constructed buildings. All AC-powered equipment must have working three-wire grounding plugs.
The use of safety glasses (eye protection) with side shields is required at all times in the
laboratory and field. Long hair must be tied back whenever in the laboratory or in the field. The
use of steel toed safety shoes in the field is required. Footwear that does not provide for
adequate traction will not be permitted. Conditions leading to slips, trips or falls, especially on
wet surfaces must be avoided. All walkways will be kept free of equipment, tools or debris.
Access to fire extinguishers or emergency exits will not be blocked. Each work area will have two
exits, remotely located. No personnel are to lift any materials over 45 pounds without assistance.
All lifting shall be done by keeping a straight back and bending at the knees. Objects will be
brought straight up without bending or twisting at the back.
9.2 PROJECT SPECIFIC SAFETY ISSUES
Initial site surveys will be conducted by SRI before final arrangements are made for field
measurements. During these surveys, information on site specific safety concerns will be
obtained from the site's safety officer. All SRI and subcontractor personnel will be briefed on site
specific hazard recognition and avoidance before being allowed on site. The following general
site safety issues are based on previous experience at similar facilities.
The first responsibility of all project personnel upon entering a site will be to attend a safety
orientation at the facility. Project supervisors will obtain information on and directions to the
nearest emergency treatment center or hospital. Project supervisors will maintain a limited first
aid supply cabinet at the field site for use with minor cuts and scratches. Personnel will be
expected to have a basic knowledge of facility emergency procedures, sirens, the location of
emergency exits, safety showers and eye wash stations, and emergency assembly points. In the
event of an facility evacuation, field personnel will determine the prevailing wind direction and
evacuate the facility at right angles to the wind direction and away from the danger zone. Project
supervisors will conduct a head-count or test personnel at the emergency assembly point. No
SRI personnel will be permitted to return to an evacuated area until clearance for re-entry from
plant personnel is obtained.
All field personnel will remain within visual sight of each other or within radio communication.
Field personnel will not climb on portable extension ladders or any equipment ladders more than
12 feet above ground level. Field personnel will not venture into any areas that would require a
safety harness and tie-off, or leave the confines of a standard guard-rail equipped walkway. Field
personnel will not make any entry into confined spaces. A confined space is any area where the
potential lack of free air movement exists, or any area where vapors may accumulate causing
normal oxygen levels to be reduced below 19 percent. Examples of confined spaces include
manways into tanks, closed rooms on two or more sides, trenches, and containment berms
around tanks. Field personnel will not enter any areas where respiratory protection or self-
contained breathing apparatus would be required. Hard hats will be required whenever there is
100
-------
Section No. 9
Revision No. 0
Date: 5/18/94
a risk of injury from falling or stationary overhead equipment. Hearing protection will be required
whenever elevated noise levels make it difficult to understand a person speaking from 3 feet
away. The use of close fitting, organic fibre long sleeve shirts in the field will be required
whenever there is a potential risk of thermal (hot and cold) burns. Cryogenic temperatures may
be encountered around liquid natural gas storage vessels or piping. Sandals or cut-off clothing
will not be permitted.
Working in the vicinity of high pressure gas production and pipelines can present a special
hazard. High pressure leaks can cause serious injury to body parts, and can accelerate dust/dirt
particles to supersonic velocities. Field personnel will not be permitted to check components
using their bare hands. Field personnel will be required to wear leather gloves at all times when
checking components with the portable analyzers.
Field work will be suspended whenever thunderstorms are in the vicinity of the field site. This is
defined as audible thunder overhead or lightning visible within 5 miles of the field site. Field
personnel will keep abreast of weather forecasts for the field site by monitoring a portable weather
radio or local radio station. If a possibility of severe weather exists, field personnel will determine
the location of a suitable severe-weather shelter and be prepared to relocate to that shelter
immediately when conditions warrant. Field personnel should be aware of the hazards of heat
prostration and maintain fluid intakes accordingly. Fluids will be available to field personnel
whenever outside air temperatures are above 75 °F Work schedules may be adjusted
accordingly to permit work in the cooler morning and early evening hours.
101
-------
Section No. 10
Revision No. 0
Date: 5/18/94
SECTION 10
REFERENCES
Clifford, A.A., Multivariate Error Analysis, John Wiley and Sons, NY, 1973.
Gilbert, P.O., Statistical Methods for Environmental Pollution Monitoring, Van Nostrand Reinhold,
NY 1987.
Hausle, K.J., Protocol for Equipment Leak Emission Estimates, EPA-453/R-93-026, (NTIS PB93-
229219), 1993.
Wetherold, R.G., et. al., Assessment of Atmospheric Emissions from Petroleum Refining, Volumes
1-4 including Appendices A-E. EPA-600/2-80-075a-d (NTIS PB80-225253, -225261, -225279, and
PB81-103830), 1980.
American Petroleum Institute, Fugitive Hydrocarbon Emissions from Oil and Gas Production
Operations, API publication number 4589, 1993.
Gas Research Institute, Fugitive Hydrocarbon Emissions - Eastern Gas Wells, GRI Publication
No. GRI-950117, 1995.
102
-------
Section No. App. A
Revision No. 0
Date: 5/18/94
APPENDIX A
HVCS MEASUREMENT PROCEDURE
The following procedure was prepared by STAR Environmental and incorporated into this QAPjP
by SRI.
Screening of the Component
The component is screened using a portable monitoring instrument to determine the point of
highest emission and the maximum instrument screening value (ISV). Screening is conducted
in accordance with EPA Method 21 "Determination of Volatile Organic Compounds Leaks", and
the highest instantaneous reading is recorded.
Collection of fugitive emissions using HVCS
The main principle of the HVCS is the drawing of ambient air around a fugitive hydrocarbon
source at a sufficiently high rate so as to effect complete capture of the hydrocarbon. The
resulting mixture of ambient air and hydrocarbon is measure volumetrically using a series of
rotameters. The hydrocarbon fraction of the mixture is measured at the outlet of the HVCS with
a separate flame ionization detector (FID) instrument. The range of measurement for FID
instruments is typically 10 ppmv to 10,000 ppmv hydrocarbon (0.00001% to 1%); therefore, large
leaks can only be quantified by attaching a dilution tip to the FID probe to dilute the HVCS outlet
stream to a level that is within the FID range.
Calibration of the HVCS
The three flow meters on the HVCS have been calibrated at the STAR laboratory. No field
calibration is necessary, but attention should be given to the condition of each flow meter before
using the HVCS. Dirt, moisture, or other foreign objects on the ball inside the flow meters or on
the walls of the flow meters will result in incorrect readings. The speed of the air pump in the
HVCS is a function of battery condition. Air flow through the air pump is controlled solely by the
valves on the three flow meters. If field calibration of the flow meters is deemed necessary, it can
be done using a dry gas meter, bubble tube, or other type of meter of known calibration.
Checking the HVCS for leaks
Every morning before use the HVCS inlet should be checked by opening the smallest flow meter
to its maximum rate (above 20 scfh) and holding the palm of the hand across the inlet tube to
stop all flow. The flow meter must drop to zero. The outlet is checked by placing the palm of the
hand across the outlet tube to stop all flow. The flow meter must drop to zero. If the meter fails
to drop to zero during either the check of the inlet or outlet tube, the HVCS cannot be used until
the condition is corrected. At the end of each day the inlet and outlet tube are again checked.
If the meter fails to drop to zero during either check, the meter reading is recorded on the data
sheet.
103
-------
Section No. App. A
Revision No. 0
Date: 5/18/94
Calibration of FID
The response of the FID is affected by temperature, battery condition, and other parameters that
are subject to change within a few minutes time, therefore the FID must be calibrate before
quantification of each sample using zero air, a low calibration gas (20 to 100 ppmv methane in
air), a medium calibration gas (3,000 to 5,000 ppmv in air), and a high calibration gas
(approximately 10,000 ppmv methane in air). The FID output voltages are recorded on the field
sheet. The following equations are used to relate FID output voltage to part per million by volume
of methane.
Concentration(ppm) =
a (slope) , '
n I
b (intercept) = log k - al
where, k = low standard (ppm), I = low standard (volts)
m = high standard (ppm), n = high standard (volts)
If a dilution tip is to be used, the dilution tip is immediately attached to the instrument probe and
a new reading of the high calibration gas is taken. The dilution ratio of the tip is the reading of
the high calibration gas without the dilution tip in place divided by the reading of the same gas
with the dilution tip in place. The dilution tip must be calibrated in the same configuration as it
will be used; i.e., the side-arm supply line must attached with the fresh air supply end located at
the position it will be in during the testing. [NOTE: Only the high gas will give a sufficiently high
response with the dilution tip in place to allow calculation of the dilution ratio. The reading of the
low calibration gas through the dilution tip will be near the lower detection limit of the instrument.]
The dilution ratio should be set between 10 and 150. The accuracy of dilution ratios above 150
have not been verified in the laboratory. Also the flow rate through the dilution tip decreases with
increasing ratio. At a ratio of 150, flow thorough the tip is approximately 7 cubic centimeters per
minute. Higher dilution ratios and the resulting lower flows would greatly increase the time
required to collect a representative sample of the HVCS exhaust.
Measurement of Background Readings
Background readings must be taken in the area of the component to be tested and in the area
used to supply fresh air to the dilution tip side-arm to assure that no other high leaks interfere with
the collection and quantification of the target component.
104
-------
Section No. App. A
Revision No. 0
Date: 5/18/94
Attaching the HVCS Inlet Tube to the Leak
The HVCS sample line inlet is secured on the component being tested at the point of highest ISV
(as determined during screening). Thin plastic film (such as Saran Warp) or duct tape is applied
to the leak to channel all hydrocarbon emissions to the HVCS sample line and to deflect wind.
Care must be taken to avoid over-sealing the area and thus preventing flow of ambient air across
the leak.
Location of HVCS unit relative to leak
The high intake rate of the HVCS (and the corresponding high outlet rate) creates the potential
for recirculation of air from the outlet tube to the inlet tube of the HVCS. Recirculation would
result in erroneously high readings. Therefore, the outlet of the HVCS must be kept at least 6
feet (2 meters) from the inlet tube to prevent recirculation.
Location of the Ambient Air Line to the Dilution Tip Sidearm
Fresh air for the dilution tip side arm must be obtained at least 16 feet (5 meters) from the HVCS
outlet. The hydrocarbon concentration in the ambient air coming through the tube must be
checked and recorded each time the HVCS is relocated.
Operation of the Flow Meters
The HVCS contains three flow meters with the following ranges; 100 to 1000 scfh, 20 to 100 scfh,
2 to 20 scfh. Only one flow meter should be opened during each measurement, the other two
must be closed to assure accurate flow measurements.
Collection of Quantification Data
Collection is begun at a high sampling rate. The FID reading of the HVCS outlet is allowed to
stabilize and then it is recorded. The sampling flow rate is decreased (and recorded) in a number
of steps while the corresponding increases in FID output voltage are recorded until the minimum
flow of the HVCS is reached (approximately 3 cfh) or the hydrocarbon concentration becomes too
high for the FID (approximately 10,000 ppmv with or without the dilution tip). NOTE: The
accuracy of readings above 10,000 ppmv with the dilution tip in place have not been verified in
the laboratory.
A minimum of three different flow rates are used for each component with the flow rates selected
so that the FID voltages are approximately the same as those obtained from the high, medium,
and low calibration gases. [NOTE: Readings within a factor of 100 times the background
concentrations should be avoided.]
As each HVCS flow rate and corresponding voltage are recorded, the leak rate is calculated as
follows:
Volume of leak = Concentration (ppm) * HVCS Flow
105
-------
Section No. App. A
Revision No. 0
Date: 5/18/94
Where the concentration is determined from the OVA output voltage using the equation presented
earlier. If the three predicted rates vary by more than 20% over-all, the possible causes should
be investigated. Possible causes include: (1) another high leak in immediate proximity (2) hoses
accidentally uncoupled, (3) battery failure, (4) actual change in process at facility, (5) incorrect
recording of reading. After the possible causes have been investigated and remedied, if possible,
the quantification must be repeated.
Verification of FID response
After each HVCS sampling is completed, the FID is immediately recalibrated using low and high
concentrations of calibration gas to determine if the FID output has drifted. The voltage are
recorded on the field sheet. If the dilution tip has been used, its response to the high
concentration of calibration gas is immediately measured and recorded.
Calculation of hydrocarbon mass emission rate
The calculation of parts-per-million from FID voltages and hydrocarbon flow rates are calculated
a second time at the STAR offices using the original field sheets as a quality assurance measure.
An exponential equation that relates the calibration gas concentration to FID output voltage is
calculated for each calibration session. The voltage readings obtained during the sample
collection period are then converted to "parts-per-million volume as methane" using this equation.
If the dilution tip has been used the "ppmv" reading thus obtained are corrected to true "ppmv"
readings by multiplying by the dilution ratio.
Total volumetric flow rates through the HVCS as recorded from the meters are multiplied by the
calculated parts-per-millions of hydrocarbon to determine volumetric hydrocarbon flow rates (as
methane). These in turn are converted to mass hydrocarbon emission rates by assuming a
molecular weight of 16 pounds of hydrocarbon (methane) per 370 cubic feet of hydrocarbon.
The average mass emission rate of hydrocarbon is calculated as the mean of at least three
measurements made at different volumetric flow rates; the standard deviation of the mea-
surements around the mean is expressed as percent of mean.
106
-------
Section No. App. B
Revision No. 0
Date: 5/18/94
APPENDIX B
FIELD DATA FORMS
Table B-1 lists and describes the data forms that will be used in the field and laboratory studies
for the HVCS evaluation. The laboratory testing and field leak rate tests will be performed in the
same manner as the field tests and documented on the same forms. An example of each form
follows.
Table B-1. Field Data Forms
Form Name
Form Usage
Screening Measurements
Component Screening Field Data Form
Portable OVA Calibration and QC Check Form.
Method 21 performance check form.
One form is completed for each wellhead or other functional unit.
Form records descriptive information and screening values for
components with screening values more than 10 ppm above the
local background. Form also inventories (checklist) components
screened. Component ID'S will be assigned to leakers.
One form is completed per OVA per day.
One form is completed per OVA per measurements trip.
HVCS Measurements
HVCS Field Data Form
HVCS Flow (Rotameter) Calibration
Form includes all information for an HVCS measurement (OVA
pre/post calibrations, QC checks, component ID, pre/post ISV's,
flows, concentrations). OVA Calibration and Method 21 performance
check forms should be completed for the OVA.
Complete one form for each HVCS for each field test.
Bagging Measurements
Bagging Field Data Form
Gas Chromatography
GC Calibration Spreadsheet - EPA Method 18
GC Data Reduction Spreadsheet - EPA Method 18
Form includes all sampling and analytical information for Blow-
Through bagging protocol (QC checks, component ID, pre/post
ISVs, flows, O2 levels, THC concentration, speciated data (if
applicable). Attach chromatograms.
Documents daily pre/post multipoints. Spreadsheet automates
calculations. One sheet completed each day. File name
MMDD.CAL
Documents GC data reduction. One sheet completed each sample.
Filename COMPID.SMP
107
-------
00
Bag Sampling Field Data Form
Location
Date Time:
Operator
Instructions: Complete one form for each bag sample. Sketch details of bag construction, component configuration. Show bag inlet/outlet.
Bags should be filled to about 75 % of capacity to avoid breakage. Fill/purge bags 3X with sample gas before collecting sample. All flows
are as indicated.
Wellhead or process unit ID:
Ambient Conditions
Temperature (deg C)
Pressure (torr)
Screening Values
Initial Screening ISV (ppm)
Pre Measurement ISV (ppm)
Post Measurement ISV (ppm)
Leak Character:
QC Checks
Leak Check Sampling Train
Leak Check Enclosure
Bag Sampling Info Method
Induced Flow (Ipm)
Return Flow (Ipm)
Stabilized exhaust cone. CH4/O2
Bag 1 THC (ppm)
Bag 2 THC (ppm)
D
D
Component Type: Component ID:
:rom Screening Form
DVA
DVA
soap and/or OVA
'lug inlet, zero flow.
Jag Should Collapse
Minimum 5 Ipm.
)VA or IR
Sketch
Comments
O 3) C/>
Qj CD CD
Ģ < °
"ft
*?
CJ1
CO
O CO
-------
HVCS Field Data Form
Location
Date
Operator
Instructions: Complete one form for each HVCS measurement. Sketch construction of partial enclosures. OVA voltages must not exceed 5.5 VDC.
Wellhead or process unit ID:
Component Type: (Component ID
HVCS ID
Battery OK
Analyzer Calibrations
Pre Cat
Point
Zero
High
Mid
Time:
Cert, ppm
0
VDC
Slope
>ost Cat
Point
Zero
High
Mid
Low
Time:
Cert, ppm
0
VDC
Slope
(LOG(HIPPM)-@LOG(LOWPPM))/(HIVOLTS-LOWVOLTS)
Intercept
ntercept
LOG(LOWPPM)-SLOPE*LOWVOLTS
HVCS Measurements
Flow Range
High
Med
Low
Other
Other
.eak Capture
OVA
Volts
-low Rate
[scfh)
Leak Rate Sigma
Cone (ppm)
Direct
Cone (ppm)
Nl Dilution
Leak Rate Avg.
Leak Rate (Ipm)
Calculated
Sigma/Avg.
OVA Flow OK
.eak Check
H2 Pressure
Screening Values
Dilution Probe
5td. Cone.
/DC
2alc Cone.
Dil. Ratio
\mbient Conditions
Temperature
'ressure
Additional Instructions:
nitial Screening ISV (ppm)
're Measurement ISV (ppm)
Jost Measurement ISV (ppm)
.eak Character:
Calculate Concentration as 10A(SLOPE+INTERCEPT) from PreCal
Sketch
Comments
D JD cn
Q) CD CD
2 < o
CJI
_i
ss
CD
O DO
-------
Component Screening Field Data Form
Location
Date Paae of
Operator
Instructions: Complete one form for each well head or process unit. Multiple sheets may be used. Complete emitter screening data
for components with ISV > 10 ppm. Complete total component inventory for all components. Return completed sheets to field coordinator ASAP.
Wellhead or process unit identification.
Emitter Screening Data
No.
1
;
^
t
i
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Component ID
abr999
Cmp. Type - CN, VL,
OL, PR, CS, PS, Ms
_ine Size
inches)
ISV
Bckgrnd
3 Character unit abbreviation:
Total Component Inventory
ressure Relief (PR)
Compressor Seals (CS)
>ump Seals (PS)
Miscellaneous (MS)
Additional Instructions:
\ each wellhead or process unit, conduct screening from start to end in terms
if gas flow. For example, a wellhead would be screened starting where the
jas leaves the ground. Component ID'S are assigned by the screener using a
1 character abbreviation for the wellhead or process unit followed by
sequential numbering. Complete total component inventory using hatch marks.
O 33 C/)
0) CD CD
2 < °
"ft
5?
CJ1
S2
CO
O 00
-------
Method 21 Performance Check Form
Date:
.ocation:
Operator:
Instructions: Complete one form per analyzer (OVA and IR) per measurements trip. This form kept at field office.
Analyzer Suitability Checklist
Analyzer ID:
Portable Analyzer Make/Model Foxboro OVA 108 Organic Vapor Analyzer
Intrinsically Safe Rated..
All Components Present (list missing components if any)..
Detector Type:
Analyte: Matural Gas - Primarily Methane
Analyzer Flow Rate (0.1 to 3.0 Ipm).,
Response Time (<30 seconds).,
Certified Intrinsically Safe
Flame lonization Detector (FID)
Ipm
seconds
Method 21 Calibration Precision Test
Species
Air
Zerol
Methane
Low
Methane
Mid
Methane
Hig.n1
Air
Methane
Air
Methane
Air
arget Analyte
Point
Cert, ppm
Zero 2
High 2
Zero 3
High 3
Zero 4
Response Factor
OVA ppm
101.1
1011
9900
9900
9900
Diff. (%)
Calculated Precision (Avg. Diff. of 3 Span Gases)
alculated Response Factor (Cert. ppm/OVA ppm)
Comments
D 3J C/5
0) CD CD
Z < O
?. W =?.
§3
cn
^ >
Ŧ 3
CO
^ O CD
-------
Daily OVA Calibration and QC Check Form
Location
Date
Operator
Instructions: Complete one form for each analyzer each day. Form travels with analyzer when in use, then filed in field office at end
of each day. Recalibrate if span check differs from std. by more than 10 %. Record component ID'S assocaited with initial and final Cats.
Daily QC Checks
Battery Level
Hydrogen Pressure (Hi/Low)
Flow Rate (0.1 to 3.0 Ipm) - Unit vertical
Response Time (<30 seconds)
Initial Calibration
Cert. PPM
0
9900 (adj)
1011
101.1 (adj)
Read PPM
Component ID (1st)
Additional Calibration
Cert. PPM
0
9900
1011
101.1
Cert, ppm
0
Time:
VDC
Read/Std
Time:
OVA ppm
Drift
Component ID (Last)
Start
Midday Span
Cert. PPM
End
Check
Read PPM
Analyzer ID:
Electronic Low Span
Electronic High Span
Excess flow in CAL manifold (COB)
Other:
Time: Final Calibration
VDC(opt) Drift Cert. PPM Read PPM
0
9900
1011
101.1
Component ID (Last)
Component ID (Last)
Start
End
Time:
VDC
Read/Std
Comments: Record all problems, flameouts. Note times.
Calculate drift as 1 minus the ratio of the PPM value read on the OVA scale and the standard value. Span check standard should be
at or near the same concentration as the high standard used in the multipoint calibration.
a 3 c/>
Q) (D CD
Ul
^ >
S2 ?
CD
-P> O CO
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverie before completing)
. REPORT NO.
EPA - 600/R-9 5-167
2.
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
Evaluation of the High Volume Collection System
(HVCS) for Quantifying Fugitive Organic Vapor Leaks
5. REPORT DATE
November 1995
6. PERFORMING ORGANIZATION CODE
. AUTHOR(S)
Eric S. Ringler
B. PERFORMING ORGANIZATION REPORT NO.
0. PERFORMING ORGANIZATION NAME AND ADDRESS
Southern Research Institute
6320 Quadrangle Drive. Suite 100
Chapel Hill, North Carolina 27514
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-D2-0062, Tasks 1/029
and 2/041
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
Task Final; 3/94 - 6/95
14. SPONSORING AGENCY CODE
EPA/600/13
is.SUPPLEMENTARY NOTES APPCD project officer is Charles
541-7586.
C. Masser, Mail Drop 62, 919/
16. ABSTRACT -j;ne repOrt discusses a recently developed measurements technique that
offers the potential for providing an easy-to-use and cost-effective means to directly
measure organic vapor leaks. The method, called the High Volume Collection System
(HVCS), uses a high volume sampling device and a portable flame ionization detector
(FID) for field analysis. The HVCS can obtain direct measurements of mass emission
rates without the need for tenting and bagging. This study of HVCS method perform-
ance included both field and laboratory testing. Laboratory evaluation of HVCS results
closely matched EPA method results with a difference in total measured emissions of
only about 3%. In one field test, the HVCS matched the EPA estimate of total facility
emissions within about 4%. In the other field test, the HVCS measured approximately
18% more emissions than the EPA method. However, the bias was present only early
in the test. Later in the test, after efforts were made to identify and correct the
source of the bias, HVCS bias was essentially zero. With some physical and proce-
dural enhancements, the HVCS may be offered to EPA for approval as an acceptable
alternative to the EPA protocol bagging method with gas chromatographic analysis.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Pollution
Measurement
Organic Compounds
Emission
Vapors
Flames
Ionization
Pollution Control
Stationary Sources
Volatile Organic Com-
pounds (VOCs)
Flame Ionization
Fugitive Emissions
13 B
14G
07C
07D
2 IB
07B
18. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (ThisReport)
Unclassified
21. NO. OF PAGES
117
20. SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
113
------- | |