May 2011
Environmental Technology
Verification Report
ISOTOPIC CARBON DIOXIDE ANALYZERS
FOR CARBON SEQUESTRATION MONITORING
PICARRO CAVITY RING-DOWN SPECTROSCOPY
ANALYZER FOR ISOTOPIC CO2 - MODEL G1101-/
Prepared by
Battelle
Batteiie
Ilie Business of Innovation
Under a cooperative agreement with
Q"lr\ U.S. Environmental Protection Agency
ET1/ET1/ET1/
-------
May 2011
Environmental Technology Verification
Report
ETV Advanced Monitoring Systems Center
ISOTOPIC CARBON DIOXIDE ANALYZERS FOR
CARBON SEQUESTRATION MONITORING
PICARRO CAVITY RING-DOWN SPECTROSCOPY
ANALYZER FOR ISOTOPIC CO2 - MODEL G1101-/
by
Ann Louise Sumner, Elizabeth Hanft, and Amy Dindal, Battelle
John McKernan, U.S. EPA
-------
Notice
The U.S. Environmental Protection Agency, through its Office of Research and Development,
funded and managed, or partially funded and collaborated in, the research described herein. It
has been subjected to the Agency's peer and administrative review. Any opinions expressed in
this report are those of the author(s) and do not necessarily reflect the views of the Agency,
therefore, no official endorsement should be inferred. Any mention of trade names or
commercial products does not constitute endorsement or recommendation for use.
11
-------
Foreword
The EPA is charged by Congress with protecting the nation's air, water, and land resources.
Under a mandate of national environmental laws, the Agency strives to formulate and implement
actions leading to a compatible balance between human activities and the ability of natural
systems to support and nurture life. To meet this mandate, the EPA's Office of Research and
Development provides data and science support that can be used to solve environmental
problems and to build the scientific knowledge base needed to manage our ecological resources
wisely, to understand how pollutants affect our health, and to prevent or reduce environmental
risks.
The Environmental Technology Verification (ETV) Program has been established by the EPA to
verify the performance characteristics of innovative environmental technology across all media
and to report this objective information to permitters, buyers, and users of the technology, thus
substantially accelerating the entrance of new environmental technologies into the marketplace.
Verification organizations oversee and report verification activities based on testing and quality
assurance protocols developed with input from major stakeholders and customer groups
associated with the technology area. ETV consists of six environmental technology centers.
Information about each of these centers can be found on the Internet at http://www.epa.gov/etv/.
Effective verifications of monitoring technologies are needed to assess environmental quality
and to supply cost and performance data to select the most appropriate technology for that
assessment. Under a cooperative agreement, Battelle has received EPA funding to plan,
coordinate, and conduct such verification tests for "Advanced Monitoring Systems for Air,
Water, and Soil" and report the results to the community at large. Information concerning this
specific environmental technology area can be found on the Internet at
http ://www. epa.gov/etv/centers/centerl .html.
in
-------
Acknowledgments
The authors wish to acknowledge the contribution of the many individuals, without whom, this
verification testing would not have been possible. Quality assurance (QA) oversight was
provided by Michelle Henderson, U.S. EPA and Rosanna Buhl and Betsy Cutie, Battelle. We
gratefully acknowledge Mr. Thomas J. Conway of the National Oceanic and Atmospheric
Administration and Mr. Bruce H. Vaughn of the Stable Isotope Laboratory at INSTAAR,
University of Colorado as collaborators to this verification test who provided analytical services.
Finally, we want thank Professor Sally M. Benson and Dr. Sam Krevor of Stanford University,
Chuck Dene of the Electric Power Research Institute, and Dr. Dominic C. DiGiulio, Dr. Eben
Thoma, and Mr. Bruce J. Kobelski of the U.S. EPA for their review of the test/QA plan and/or
this verification report.
IV
-------
Contents
Page
Foreword iii
Acknowledgments iv
List of Abbreviations ix
Chapter 1 Background 1
Chapter 2 Technology Description 2
Chapters Test Design and Procedures 5
3.1 Test Overview 5
3.2 Test Site Descriptions 7
3.2.1 Ambient Breeze Tunnel 7
3.2.2 GS Test Site 8
3.3 Experimental Design 9
3.3.1 Accuracy, Bias, Precision, and Linearity 12
3.3.2 Isotope Ratio Bias 13
3.3.3 Response Time 13
3.3.4 Temperature and RHBias 14
3.3.5 Minimum Detectable Leak Rate 15
3.3.6 Ambient Air Monitoring 16
3.3.7 Mobile Surveys 16
3.3.8 Comparability to Reference Method 17
3.3.9 Data Completeness 17
3.3.10 Operational Factors 17
Chapter 4 Quality Assurance/Quality Control 19
4.1 Reference Method Quality Control 19
4.2 Phase 2 Leak Flow Rate Quality Control 22
4.3 Dynamic Dilution System Quality Control 23
4.4 Audits 23
4.4.1 Performance Evaluation Audit 23
4.4.2 Technical Systems Audit 24
4.4.3 Data Quality Audit 25
Chapters Statistical Methods 27
5.1 Model G1101-/' Factor Calibration Correction 27
5.2 Water Vapor Correction 27
5.2.1 Delta Water Vapor Correction 28
5.2.2 CO2 Concentration Water Vapor Correction 28
-------
5.3 Accuracy 29
5.4 Bias 29
5.5 Precision 30
5.6 Linearity 30
5.7 Minimum Detectable Leak Rate 31
5.8 Response Time 31
5.9 Comparability 32
5.10 Data Completeness 32
Chapter 6 Test Results 33
6.1 Concentration Accuracy and Bias 33
6.2 Concentration Precision 36
6.3 Concentration Linearity 36
6.4 Response Time 38
6.5 Isotope Ratio Accuracy and Bias 39
6.6 Isotope Ratio Linearity 41
6.7 Temperature and RH Bias 42
6.8 Minimum Detectable Leak Rate 43
6.9 Ambient Air Monitoring 48
6.10 Mobile Surveys 52
6.11 Reference Method Comparability 56
6.12 Data Completeness 57
6.13 Operational Factors 58
Chapter 7 Performance Summary 59
Chapter 8 References 62
Appendix A Daily Checklist 64
VI
-------
Tables
Table 1. Summary of Performance Parameters and Testing Frequency 11
Table 2. CC>2 Concentrations and Order for Multi-point Challenges 12
Table 3. CO2 Concentrations and Isotope Ratios for Bias Tests 13
Table 4. Chamber and Sample Conditions for Temperature and RH Bias Tests 14
Table 5. Mobile Survey Features 17
Table 6. Test Flask Results on Days when References Samples were Analyzed 20
Table 7. PE Sample Results for CO2 Concentration and Isotope Ratio 24
Table 8. Concentration Accuracy Results 35
Table 9. Concentration Precision Results 36
Table 10. Response Time Results 38
Table 11. Isotope Ratio Accuracy Results 40
Table 12. Temperature and RH Bias Results 43
Table 13. Minimum Detectable Leak Rate Results 44
Table 14. Minimum Detectable Leak Rate Keeling Plot Results 47
Table 15. Extrapolated Minimum Detectable Leak Rate Results 48
Table 16. Mobile Survey Features and Identifiers 52
Table 17. Reference Method Comparability Results 57
vn
-------
Figures
Figure 1. The Picarro Model Gl 101-7 carbon dioxide isotope analyzer 2
Figure 2. Block diagram of the Picarro Model Gl 101-7 carbon dioxide isotope analyzer 3
Figures. Photographs of the Model Gl 101-7 installed in the ABT 8
Figure 4. Photographs of the GS test site 9
Figures. CO2 Test Flask Results for 2010 20
Figure 6. CO2 Test Results for Portable Sampling Unit 84M 21
Figure 7. Isotope Ratio Precision Check Results 21
Figure 8. Model G1101-7 Concentration Accuracy Results 34
Figure 9. Model G1101-7 Concentration Linearity Results 37
Figure 10. Model G1101-7 Concentration Linearity Results 37
Figure 11. Model G1101-/' Isotope Ratio Accuracy Results 41
Figure 12. Model G1101-7 Isotope Ratio Linearity Results 42
Figure 13. Model Gl 101-/' Minimum Detectable Leak Rate Results for Day 1 (7/28/2010) 45
Figure 14. Model Gl 101-7 Minimum Detectable Leak Rate Results for Day 2 (7/29/2010) 45
Figure 15. Model Gl 101-7 Minimum Detectable Leak Rate Results for Day 3 (7/30/2010) 46
Figure 16. Model Gl 101-7 Minimum Detectable Leak Rate Results for Day 3 (7/30/2010) 47
Figure 17. Meteorological Conditions and Model Gl 101-7 Ambient Air CO2 Measurements at
theGS Site 49
Figure 18. Keeling Plot of Model Gl 101-7 Ambient Air Data 50
Figure 19. Leak Response Time Results 51
Figure 20. Mobile Survey Results 54
Figure 21. Mobile Survey GPS Trace of Selected Features 54
Figure 22. Mobile Survey Results 55
Figure 23. Mobile Survey GPS Trace of Features Shown in Figure 22 55
Vlll
-------
List of Abbreviations
%D percent difference
%R percent recovery
%RSD percent relative standard
deviation
ABT ambient breeze tunnel
ADQ audit of data quality
AMS Advanced Monitoring
Systems
CCL Central Calibration
Laboratory
cfm cubic feet per minute
CC>2 carbon dioxide
CRDS cavity ring-down
spectroscopy
DFB distributed feedback
DQIs data quality indicators
EPA U.S. Environmental
Protection Agency
ETV Environmental Technology
Verification
GS geologic sequestration
GHz gigahertz
Hz hertz
INSTAAR Institute for Arctic and
Alpine Research
IRMS isotope ratio mass
spectrometry
ISO International Organization for
Standardization
LPM liter per minute
LRB laboratory record book
MHz megahertz
m meters
m/s meters per second
NDIR nondispersive infrared
NOAA National Oceanic and
Atmospheric Administration
ppm parts per million (mole
fraction)
PDB Pee Dee Belemnite
PE performance evaluation
%o per mil (part per thousand)
PO project officer
QA quality assurance
QC quality control
QCL quantum cascade laser
QMP Quality Management Plan
RMO Records Management Office
513C relative difference in stable
carbon isotope ratio from
PDB standard
RPD relative percent difference
SIL Stable Isotope Lab
seem standard cubic centimeters
per minute
s standard devi ati on (n-1)
G standard deviation (n)
TQAP test/quality assurance plan
TSA technical systems audit
WMO World Meteorological
Organization
IX
-------
Chapter 1
Background
The U.S. Environmental Protection Agency (EPA) supports the Environmental Technology
Verification (ETV) Program to facilitate the deployment of innovative environmental
technologies through performance verification and dissemination of information. The goal of the
ETV Program is to further environmental protection by accelerating the acceptance and use of
improved and cost-effective technologies. ETV seeks to achieve this goal by providing high-
quality, peer-reviewed data on technology performance to those involved in the design,
distribution, financing, permitting, purchase, and use of environmental technologies.
ETV works in partnership with recognized testing organizations; with stakeholder groups
consisting of buyers, vendor organizations, and permitters; and with the full participation of
individual technology developers. The program evaluates the performance of innovative
technologies by developing test plans that are responsive to the needs of stakeholders,
conducting field or laboratory tests (as appropriate), collecting and analyzing data, and preparing
peer-reviewed reports. All evaluations are conducted in accordance with rigorous quality
assurance (QA) protocols to ensure that data of known and adequate quality are generated and
that the results are defensible. The definition of ETV verification is to establish or prove the
truth of the performance of a technology under specific, pre-determined criteria or protocols and
a strong quality management system. The high quality data are assured through implementation
of the ETV Quality Management Plan. ETV does not endorse, certify, or approve technologies.
The EPA's National Risk Management Research Laboratory (NRMRL) and its verification
organization partner, Battelle, operate the Advanced Monitoring Systems (AMS) Center under
ETV. The AMS Center recently evaluated the performance of the Picarro Cavity Ring-Down
Spectroscopy Analyzer for Isotopic Carbon Dioxide (CO2) - Model Gl 101-/.
-------
Chapter 2
Technology Description
This report provides results for the verification testing of the Picarro, Inc., Model Gl 101-/'. The
following is a description of the Model Gl 101-/' carbon dioxide isotope analyzer based on
information provided by the vendor. The information provided below was not verified in this
test.
The Model Gl 101-/', shown in Figure 1, is a low-drift, high precision analyzer designed to
measure the stable isotope ratio of carbon (513C) in carbon dioxide (CO2). This analyzer is based
on cavity ring-down spectroscopy (CRDS), which is a technique in which a gas sample is
introduced into a high finesse optical cavity and the optical absorbance of the sample is
determined, thus providing concentration or isotopic ratio measurements of a particular gas
species of interest(1'2).
Figure 1. The Picarro Model G1101-/ carbon dioxide isotope analyzer
-------
Figure 2 shows a block diagram of the Picarro CRDS analyzer. The components which make up
a basic CRDS instrument are a laser, a high finesse optical cavity consisting of two or more
mirrors, and a photo-detector. Operationally, light from a laser is injected into the cavity through
one partially reflecting mirror. The light intensity inside the cavity then builds up over time and
is monitored through a second partially reflecting mirror using a photo-detector located outside
the cavity. The "ring-down" measurement is made by rapidly turning off the laser and measuring
the light intensity in the cavity as it decays exponentially with a time constant, r, that depends on
the losses due to the cavity mirrors and the absorption and scattering of the sample being
measured. After shutting off the laser, most of the light remains trapped within the cavity for a
relatively long period of time (i.e., microseconds [u.sec]), producing an effective path length of
tens of kilometers through the sample. Much like a multi-pass cell, this long effective path length
gives CRDS its high sensitivity.
Patented High Finesse Cavity
Cell Volume = 35cc
xji Sample Gas Inlet
Tunable
Diode Laser
Pressure Gauge
Outlet Gas Flow
(to pump)
Photodetector
i
Laser Control
Electronics
Data Collection &
Analysis Electronics
Figure 2. Block diagram of the Picarro Model G1101-/ carbon dioxide isotope analyzer
The Model Gl 101-/ utilizes a telecom-grade distributed feedback (DFB) laser. Light from the
DFB laser is transported to a wavelength monitor via a polarization maintaining optical fiber.
The analyzer is designed to simultaneously measure optical absorption using a proprietary
traveling wave cavity and the optical frequency at which the absorption occurs using a
proprietary wavelength monitor. The temperature and pressure of the ambient air sample
continuously flowing through the optical cavity are regulated at all times. A typical empty cavity
decay constant, r, is 40 u,sec for this instrument. The normalized reproducibility of the measured
ring-down time constant (Ar/r) is better than 0.02%. With a ring-down acquisition rate of 100
hertz (Hz), the typical sensitivity of the instrument is l^xlO^c
1 2
The analyzer continuously scans the laser over two individual CC>2 rovibrational (i.e., rotational
and vibrational excitation of the CO2 molecule) resonant absorption lines, one for 12CC>2 and one
for 13CC>2. Each spectrum is comprised of absorption loss as a function of optical frequency. The
-------
concentration is proportional to the area under each measured spectral feature. Concentration
measurements are provided approximately every second, corresponding to a total of 100 ring-
down and wavelength monitor measurements, and the isotope ratio is derived from the ratio of
the concentrations.
The wavelength monitor used in the analyzer is solid-state in design and has no moving parts. It
is designed to provide wavelength measurements over a frequency range corresponding to
greater than 100 nm. The wavelength precision (defined as the repeatability of the wavelength
measurement at a single spectral point and calculated as one standard deviation (er)) is
approximately 1 MHz (la). The relative accuracy, defined as the repeatability of the difference
of the wavelength measurement between two spectral points separated by approximately 1 GHz
(the width of a typical absorption line at a typical operating pressure of 140 Torr) during a
spectral scan is designed to be approximately 0.3 MHz. The size and shape of the CO2 spectral
lines are sensitive to temperature and pressure of the sample, but typically are 6250
wavenumbers (i.e., wave property proportional to the reciprocal of the wavelength). Therefore,
the analyzer is designed to control the sample gas temperature to a precision (la) of a few
hundredths of a K over ambient temperatures ranging from 10 to 35°C and the sample pressure to
a precision (la) of 0.05 Torr. In the analyzer, a combination of proportional valves (for flow
control) is used to maintain the cavity at a known constant pressure.
The Model Gl 101-7 weighs 26.3 kg (58 Ibs), has dimensions of 43 x 25 x 59 cm (17" x 9.75" x
23") including the feet, and can be rack mounted or operated on a benchtop. The approximate
purchase price of the Model Gl 101-7 is US $60,500.
-------
Chapter 3
Test Design and Procedures
3.1 Test Overview
This verification test was conducted according to procedures specified in the Test/QA Plan for
Verification oflsotopic Carbon Dioxide Analyzers for Carbon Sequestration Monitoring^
(TQAP) and adhered to the quality system defined in the ETV AMS Center Quality Management
Plan (QMP)(4). As indicated in the test/QA plan, the testing conducted satisfied EPA QA
Category III requirements. The test/QA plan and/or this verification report were reviewed by:
• Chuck Dene, Electric Power Research Institute
• Sam Krevor, Stanford University
• EbenThoma, U.S. EPA
• Dominic DiGiulio, U.S. EPA
• Bruce Kobelski, U.S. EPA.
Battelle conducted this verification test with funding support from the EPA's Forum for
Environmental Monitoring and with in-kind support from the National Oceanic and Atmospheric
Administration (NOAA).
Research on carbon storage in geologic reservoirs such as saline formations, coal seams, and
depleted oil and gas fields, has gained momentum in recent years as interest in mitigation of
greenhouse gases, such as CO2, has increased and a number of pilot-studies have recently been
brought online. Capture and geologic sequestration (GS) of CO2 involves capturing emissions at
a power plant or other large source, separating the emissions to isolate CO2, and compressing the
gas. The compressed CO2 is injected into a deep underground rock formation. Potential sites are
carefully evaluated for adequacy of containment layers, seismic stability, and other factors. As
pilot and full-scale geologic sequestration programs continue to be implemented, so do the needs
to monitor leakage.
Stable isotope analysis can be used in environmental forensics, for example to aid in determining
1^191^17
the source of carbon dioxide. Deviations in the ratio of C to C ( C/ C) in atmospheric CO2
relative to that in ambient air can be used to identify input from other carbon sources, such as
fossil fuel combustion, since atmospheric, carbonate, and plant-derived carbon differ in their
13C/12C relative to the Pee Dee Belemnite (PDB) standard. The relative difference in stable
carbon isotope from the PDB standard, referred to as 513C, is calculated as shown in Equation 1
and expressed in per mil (%o), or part per thousand.
-------
r.l3/-> ( ^1 ^Sample
d ^Sample ~ ( i3C/i2C J-j ^ AUUU ,^x
Since the PDB standard was highly enriched in 13C, most naturally occurring carbon sources
have a negative 513C value. For example, ambient air CO2 has a global average 513C close to
-8%o (cf., ref 5) and the global mean value from a 1991 inventory of fossil fuel types was
-28.5%o.(6>7) Stable isotope measurements are traditionally conducted on discrete samples, such
as air collected in canisters, in the laboratory using isotope ratio mass spectrometry (IRMS), but
recent advances in spectroscopic monitoring technology have made it possible to conduct in situ
measurements of stable isotope ratios with high frequency and precision/8'9-*
The use of isotopic CC>2 analyzers for ambient air monitoring in areas near GS sites, for example,
could be used to identify intrusion of non-ambient CC>2 and provide information about its source.
Large-scale leaks in high risk areas where the source is well-understood can be detected by
conventional CC>2 analyzers. Fast-response, portable analyzers, including infrared "cameras,"
could be useful as a survey tool to quickly assess larger geographic areas for large-scale leaks.
The high sensitivity and fast response of isotopic CC>2 analyzers have the potential to detect
smaller leaks and identify larger subsurface leaks before exceeding the detection limits of less
sensitive techniques. Spectroscopic isotopic CO2 analyzers have been proposed as a potentially
viable technology for monitoring GCS sites, nearby communities, and sensitive ecosystems for
CO2 leaks, where analyzers would need to have sufficient accuracy and precision to detect
background ambient air concentrations (-350 ppm) and 513C values (—8%o) and capture
daily/seasonal variability.
This verification test evaluated the performance of the Model Gl 101-/ while conducting
measurements of CO2 concentration and 513C in synthetic gas mixtures and in ambient air. One
of the goals of this verification test was to provide information on the potential use of the Model
Gl 101-/' for monitoring at or near facilities utilizing GS for captured CC>2. To accomplish this
goal, the experimental design included a combination of controlled gas challenges in an indoor
laboratory environment and a sheltered ambient breeze tunnel, survey measurements for above-
ground leak detection, and continuous ambient monitoring to provide performance data under a
variety of simulated and real-world conditions.
Phase 1 of this verification test was conducted in Battelle laboratories in Columbus, OH to
evaluate the analytical performance of the Model Gl 101-/' under controlled laboratory conditions
from July 9 through July 23, and August 11 through August 17, 2010. The Model Gl 101-/' was
challenged with gas standards of known isotopic composition and concentration to generate test
samples over a range of CO2 concentrations and isotopic compositions. The resulting
concentration and 513C data were used to calculate accuracy, bias, linearity, precision, and
response time, where appropriate. Bias with respect to ambient temperature and relative
humidity (RH) was also assessed.
The ability of the Model Gl 101-/' to detect CO2 leaks was evaluated during Phase 2 of this
verification test, which was conducted at Battelle's Ambient Breeze Tunnel (ABT) facility in
West Jefferson, OH. The ABT was used to simulate leaks of 13C-depleted CO2 in ambient air
under simulated field conditions. The Model Gl 101-/' was installed inside the ABT and ambient
-------
air was drawn through the tunnel at approximately 1.8 meters per second (m/s) and a stream of
pure 12CC>2 at a fixed flow rate was periodically introduced into the ABT. By varying the 12CC>2
flow rate, the minimum detectable CO2 leak rate was determined for several 513C values. In
addition, ambient air reference samples were collected to determine the comparability of the
Model Gl 101-/ to CO2 concentration and 613C i
conducted from July 28 through July 30, 2010.
Model Gl 101-/ to CO2 concentration and 513C reference methods. Testing for Phase 2 was
The utility of the Model Gl 101-/' for monitoring at GS sites was evaluated during Phase 3, which
was conducted at a coal-fired power plant in West Virginia. The analyzer was installed in a shed
near the sequestration wells and sampled ambient air drawn from near the main well head over a
one-week period from August 2 through August 6, 2010. During that period, ambient air
reference samples were collected to determine the comparability of the Model Gl 101-/ to CC>2
concentration and 513C reference methods. The Model Gl 101-/' was also installed in a hybrid
sedan vehicle and operated using battery power to conduct mobile surveys of GS site
transmission lines and infrastructure. Finally, the leak rate response was determined from an
intentional release to simulate an above ground leak.
3.2 Test Site Descriptions
3.2.1 Ambient Breeze Tunnel
Details of the ABT are provided in the TQAP. Briefly, the ABT was designed to conduct
controlled releases in ambient air with down-wind measurements and has dimensions of
approximately 45m x 6m x 6m. A large blower was used to constantly draw ambient air through
the ABT facility at 3400 m3/min, equivalent to wind speeds of approximately 1.8 m/s.
Controlled leaks were generated near the inlet to the ABT and the Model Gl 101-/' was installed
near the exit of the ABT, downstream of the mixing baffle. The enclosed nature of the ABT
provided for controlled unidirectional flows with no confounding cross-winds. Photographs of
the Model Gl 101-/ as installed in the ABT are shown in Figure 3. Panel A shows the Model
Gl 101-/' installed on a scaffolding platform, with the inlet positioned 1.4 m above the concrete
floor. The opening for ambient air is visible in Panel A. The blowers are located at the end of
the tunnel that is visible in Panel B (behind the analyzer). Pure 12CC>2 was released through the
Teflon tube, shown in Panel B, which was positioned approximately 1.5m above the floor. The
mixing baffles are visible in both panels. The blower was turned on each morning and shut off
in the evening.
-------
Figure 3. Photographs of the Model GllOl-i installed in the ABT
3.2.2 GS Test Site
Details of the GS test site are provided in the TQAP. The GS test site was a West Virginia coal-
fired power plant, where CC>2 from the flue gas is being captured, separated, compressed, and
stored in a geologic formation over 7,000 feet below the surface. Due to an unplanned outage,
the plant was not actively capturing or sequestering CC>2 during this verification test; however,
captured CO2 was present at elevated pressures in the transmission lines at the site. CO2 in these
lines was used for testing.
The Model Gl 101-/' was installed in a shed near the above-ground CC>2 transmission lines, as
shown in Figure 4 (Panels A and B), approximately 70 feet from the main injection well. Air
was drawn into the shed to the Model Gl 101-/' inlet through a length of Teflon tubing; the tubing
inlet was positioned near the main injection well, as shown by the yellow circle in Figure 4,
Panel C. The shed had a small window for ventilation and a wood floor in the front section
where the analyzer was installed. Ambient temperatures during Phase 3 reached 37 °C and it
became necessary to install an air conditioner inside the shed to reduce the indoor temperature.
A MetOne meteorological station was installed approximately 15 feet from the transmission lines
(Figure 4, Panel A).
-------
Figure 4. Photographs of the GS test site
3.3 Experimental Design
Per direction from the vendor, the Model Gl 101-/' was installed at each testing location by ETV
testing staff using the instrument manual and without any specific training from the vendor
(executed as designed). No on-site calibrations were performed. The vendor representative
provided a list of parameters to be checked by verification testing staff on a daily basis to verify
the operation of the Model Gl 101-/' and identify signs of malfunction. The checklist, provided
as Appendix A, was completed daily (Monday through Friday) by Battelle staff. In general,
Battelle staff checked the status window for status messages, recorded several instrument
parameter values, checked the analyzer flow rate, and backed up analyzer data.
For this verification test, all Model Gl 101-/' readings were logged. The analyzer generated daily
data files, which contained the raw readings as well as 30-second, 2-minute, and 5-minute
running averages for 12CC>2, 13CC>2, and "delta" (513C). The measurement unit for CO2 mole ratio
(referred to in this report as concentration for convenience) is parts per million (ppm). The
measurement unit for 513C is per mil (%o), or part per thousand.
-------
The Model Gl 101-/ also reported water vapor concentration and a number of instrument
parameters. Each daily data file was 32.4 MB. In addition, select data were saved in smaller (~7
MB) files. Spectral data also generated by the analyzer were downloaded daily in case they
would be needed for troubleshooting purposes. Spectral data files were generated hourly, with
approximately 209 MB of zipped files per day. Data were downloaded daily to an external
expansion drive, which was connected directly to the Model Gl 101-/' by USB port.
Approximately 13 GB of data were generated during this verification test, including spectral and
regular data files.
During Phase 1, gas standard dilutions for each test condition were supplied to the Model
Gl 101-/' for a minimum of 20 minutes, unless otherwise noted. This allowed sufficient time for
the flow from the dilution system to stabilize and the analyzer to record data record data under
stable conditions to use for calculations. For the isotope bias tests, standards were delivered for a
shorter period of time to conserve materials. For gas standard challenges, the average Model
Gl 101-/ value at each test condition was calculated from the last five minutes of raw data. The
last five minutes were selected because the Model Gl 101-/' response appeared to be stable during
that period (i.e., a general increase or decrease in the response was not apparent). The average
Model Gl 101-/' response values were used in the calculations described in Chapter 5 of this
report.
The Model Gl 101-/' CO2 concentration and 513C readings when sampling ambient air were
compared to concurrent measurements using nondispersive infrared (NDIR) analysis and IRMS,
respectively. Ten duplicate whole air samples were collected in glass flasks during Phases 2 and
3 of this verification test. Ambient air samples were collected by flushing the glass flasks for a
period of five minutes and then pressurizing the flasks for one minute. For the comparisons, the
Model Gl 101-/' raw readings were averaged over the five minutes prior to the end of the
pressurization period. For example, for a sample that was flushed from 4:49 to 4:54 PM and
pressurized from 4:54 to 4:55 PM, the Model Gl 101-/' raw readings were averaged from 4:50 to
4:55 PM and that average used for comparisons.
The Model Gl 101-/' was verified by evaluating the parameters listed in Table 1. During all
phases of this verification test, the Model Gl 101-/ was operated by verification testing staff, who
had read and who followed the instrument manual. No additional training was deemed necessary
by the vendor. The ambient temperature and relative humidity in which the Model Gl 101-/ was
operated during all phases of this verification test were recorded using a Hobo data logger with
temperature/RH probe. The performance of the Model Gl 101-/' during this verification test are
presented in Chapter 6 of this report and summarized in Chapter 7.
10
-------
Table 1. Summary of Performance Parameters and Testing Frequency
Phase
Performance
Parameter
Objective
Comparison Based On
Testing Frequency
Accuracy and Bias
Determine degree of
quantitative agreement with
compressed gas standard
Challenges with CO2 gas
standards of known 513C at 3
RH levels and 3 temperatures
-3 runs at each of 12 nominal concentrations
(one 513C value)
-1 run at each of 15 combinations of RH,
temperature, and CO2 concentration (one
513C value)
-2 runs at each of 9 combinations of CO2
concentration and 513C)
Linearity
Determine linearity of
response over a range of CO2
concentrations
Dynamic spiking with gas
standards
-3 runs at each of 12 nominal concentrations
(one 513C value)
-2 runs at each of 9 combinations of CO2
concentration and 513C)
Precision
Determine repeatability of
successive measurements at
fixed CO2 levels
Repetitive measurements under
constant facility conditions
measured
-3 runs at each of 12 nominal concentrations
(one 513C value)
Response Time
Determine 95% rise and fall
time
Recording successive readings
at start and end of sampling CO2
gas standard
Once during each day of dynamic spiking
testing
Minimum
Detectable Leak
Rate
Determine the minimum
detectable CO2 leak rate under
controlled and ambient
conditions
Repetitive measurements of a
low-level 12CO2 leak
Once
Leak Response
Rate
Determine the amount of time
between an intentional release
of captured CO2 and detection
of the leak by the CO2
analyzers
Recording the elapsed time
between start of release and
positive detection
Once
Comparability to
Reference Method
Determine degree of
quantitative agreement with
reference method results
Concentration and 513C results
for ambient air reference
samples
2 sample pairs collected during ambient air
sampling
8 sample pairs collected during ambient air
sampling
-------
3.3.1 Accuracy, Bias, Precision, and Linearity
During Phase 1, the Model Gl 101-/' was challenged with a series of dilutions from a compressed
CO2 gas standard (in CCVfree zero air—ambient air filtered to contain less than 0.1 ppm of total
hydrocarbons) to achieve measurements in the range of expected ambient air concentrations (i.e.,
350 ppm) and also at higher concentrations (up to 5000 ppm CC>2) to simulate concentrations that
could be observed in high hazard areas. Three non-consecutive measurements were recorded at
each of twelve different nominal concentration levels at one 513C value. Each concentration was
supplied to the analyzers for at least twenty minutes. Table 2 shows the CC>2 concentration values
that were supplied to the analyzer, and the order in which the concentrations were supplied. As
Table 2 indicates, the CC>2 concentrations were first supplied to the analyzers in increasing order,
then in random order, and finally in decreasing order. Dilutions were prepared from a certified
compressed mixture of 11% CC>2 in air (Air Liquide Acublend Master Class, 11.0% ±1%) using
an Environics Model 6100 Multi-Gas Calibrator. These tests were conducted at room
temperature without added humidity.
Table 2. COi Concentrations and Order for Multi-point Challenges
Nominal CO2
Concentration (ppm) Measurement Number
0
100
200
300
400
500
750
1600
2450
3300
4150
5000
1
2
3
4
5
6
7
8
9
10
11
12
16
22
18
14
23
20
15
17
21
19
24
13
36
35
34
33
32
31
30
29
28
27
26
25
The Model Gl 101-/ response to the series of CO2 gas standards was used to evaluate accuracy,
bias, precision, linearity, and response time. The statistical procedures used are presented in
Chapter 5. Accuracy was calculated at each concentration and for each replicate relative to the
nominal CC>2 concentration. Bias was calculated once for the series of multi-point CC>2
challenges. The Model Gl 101-/' precision was demonstrated by the reproducibility of the
average Model Gl 101-/' response at each nominal CC>2 concentration. Linearity was assessed by
establishing a multi-point calibration curve from the Model Gl 101-/' response and was
determined once for the full range (0 to 5000 ppm) and once for the range of concentrations
expected in ambient air (0 to 500 ppm).
12
-------
3. 3. 2 Isotope Ratio Bias
Analyzer bias with respect to the 513C value was assessed by challenging the analyzers with
dilutions from three CC>2 isotope mixtures [SMU Stable Isotope Laboratory, through Oztech
Trading Corporation, -3.61 %0; -10.41 %0; and -40.80 %o (0.01 standard deviation)], each at
three CC>2 concentrations (see Table 3). Dilutions were prepared using an Environics Model
6100 Multi-Gas Calibrator, which was calibrated for air. Since the CO2 isotope mixtures were
pure CC>2 and the dilution system's mass flow controllers are calibrated for air/nitrogen, a
correction factor of 0.737 was applied to the input concentrations to account for differences in
specific heat and density of CC>2 versus nitrogen. The resulting actual nominal CC>2
concentrations delivered to the Model Gl 101-/ are shown in Table 3. Accuracy was calculated
at each concentration and 513C value for each replicate relative to the nominal CC>2
concentration. Bias was calculated for each 513C value. Each CO2 concentration/513C pair was
delivered to the analyzers twice for a total of 18 data points. Due to limitations imposed by the
quantity of CC>2 isotope mixtures available from the vendor, two replicates were conducted
instead of the three replicates prescribed in the TQAP for this verification test. This deviation
from the TQAP resulted in a slightly smaller data set than originally planned, with only two
replicates for each test condition instead of three. The average relative percent difference (RPD)
for the two replicates was 0.8% for CO2 concentration, with a range from 0.03% to 5.2%. For
513C, the average RPD was 1.8%, with individual values ranging from 0.03% to 10.1%. Having
more replicates could make it possible to see smaller differences in the Model Gl 101-/' response
due isotope bias, if any, over the analyzer's inherent reproducibility. In an effort to conserve the
gas standard and maximize the data return, each dilution was delivered to the CO2 analyzer for
10 minutes or the time required for the signal to stabilize plus 5 minutes, whichever was longer.
Table 3. COi Concentrations and Isotope Ratios for Bias Tests
Input Nominal COi Actual Nominal
*
Approximate d C (%o) _ Concentrations (ppm) _ Concentrations (ppm)
-3.61±0.01(a)
-10.41
-40.8
±0.01
±0.01
350
500
1000
350
500
1000
350
500
1000
259
370
740
259
370
740
259
370
740
(a) Uncertainties are standard deviations reported on certificates of analysis for each gas standard.
3.3.3 Response Time
The data collected for the multi-point CC>2 challenges (Section 3.3.1) during Phase 1 were also
used to determine the analyzer response time. The 95% rise time and 95% fall times were
13
-------
calculated for the consecutive concentration steps (Table 2, measurements 1-12 and 25-36).
Calculations for response time are described in Chapter 5.
3.3.4 Temperature and RH Bias
Bias due to the ambient and sample temperature and RH was assessed during Phase 1. The
Model Gl 101-/ was tested in a Webber temperature and RH-controlled chamber, which was used
to vary the temperature and RH of the air surrounding the Model Gl 101-/'. Dilutions of a CC>2
gas standard were delivered to the Model Gl 101-/' (inside the chamber) at three concentrations
got each of six temperature/RH condition. Dilutions were prepared from a certified compressed
mixture of 11% CC>2 in air (Air Liquide Acublend Master Class, 11.0% ±1%) using an
Environics Model 6100 Multi-Gas Calibrator. Humidified zero air was added to the output of
the Environics calibrator to achieve the desired sample RH. The resulting mixture passed
through a coil placed within the environmental chamber to assist with temperature equilibration
upstream of the Gl 101-/' inlet. The temperature and relative humidity of the sample stream was
monitored using the Hobo data logger with temperature/RH probe, positioned downstream of the
Gl 101-i inlet. The specific conditions are listed in Table 4. The Model Gl 101-i was subjected
to each test condition once for a minimum of twenty minutes. One 513C value was used for these
tests. Bias was calculated as described in Chapter 5.
Table 4. Chamber and Sample Conditions for Temperature and RH Bias Tests
Nominal
Temperature
(°C)
20 ± 2°C
32 ± 2°C
4±2°C
RH (%)
0±10%
50 ±10%
90 ±10%
50 ±10%
90 ±10%
50 ±10%
Sample
Temperature
(°C)
20.1
19.7
19.7
20.5
20.5
20.5
21.3
20.4
20.4
32.8
32.8
32.8
32.8
32.8
32.8
4.2
4.2
4.2
RH
(%)
0
0
0
50
54
54
91
86
88
55
52
50
91
93
90
49
50
52
Chamber
Temperature
(°C)
19.8
19.9
20.1
20.1
20.1
20.1
20.5
20.4
20.1
32.2
32.0
32.2
32.2
32.1
32.1
4.2
4.1
4.0
RH
(%)
10
n(a)
17(a)
49
49
48
84
88
86
47
48
49
87
86
87
51
50
47
Nominal COi
Concentration
(ppm)
350
500
1000
350
500
1000
350
500
1000
350
500
1000
350
500
1000
350
500
1000
(a) Chamber conditions that were outside the target range specified in the TQAP.
14
-------
During two temperature/RH bias runs at 20°C, the chamber RH conditions were outside the
target range specified in the TQAP for this verification test (11 and 17% RH versus 0% ±10%).
However, the gas supplied to the analyzer was within the target range for these runs (0% RH
actual). The difference in RH for the "ambient" conditions within the chamber, which was a
deviation from the TQAP, is not expected to impact the analyzer's response given that it was
within the vendor-reported operating range.
3.3.5 Minimum Detectable Leak Rate
The ability of a monitoring technology to detect a CC>2 leak under real-world conditions will
depend on the isotopic signature of the leaking CO2, isotopic signature of the ambient air,
meteorological conditions, sampling proximity, local CC>2 sources, and the monitoring
technology performance. The minimum leak rate that can be detected above ambient variability
and the precision of the Model Gl 101-/ was determined under semi-controlled conditions during
Phase 2. The Model Gl 101-/' was installed in the Reference Sampling and Test Section of the
ABT, which is shown in Figure 3, Panel A. During this test, most of the parameters described
above were controlled or accounted for in the experimental design so the performance of the
Model Gl 101-/ could be evaluated under a well-defined set of conditions. A leak was
considered to be successfully identified if an increase or decrease in the measured 513C, greater
than 2 times the variability in ambient 513C, was measured by the Model Gl 101-/' for the last 15
minutes of each leak simulation period. (The 15-minute period was selected to reduce the
potential for impacts due to short-term disturbances, such as nearby vehicle traffic.) The ambient
air 513C variability was determined from one hour of ambient air data measured by the Model
Gl 101-/' on the day of testing. Pure (99.95%) 12CC>2 was added to the ambient air diluent being
drawn into the ABT to simulate a low-level leak of 13C-depleted CC>2. This approach assumed
that a low-level leak would be well-mixed in the ambient air diluent before reaching the Model
Gl 101-/', which is expected for the flow conditions utilized during the testing (approximately 1.8
m/s velocity).
19
The initial CC>2 leak rate was set at a nominally detectable level that was twice the standard
deviation (s) in 513C measured by the analyzer for a period of at least one hour on the day of
testing. A leak at that rate was introduced for approximately 20 minutes with at least 15 minutes
of ambient air flow between simulated leaks. An iterative process was used to slowly approach
the minimum detectable leak rate, starting with a leak flow rate of 0.156 LPM 12CC>2, and
increasing the flow rate until three leak simulation replicates were successfully identified. This
leak rate was defined as the minimum detectable limit for the Gl 101-/' under the test conditions.
A conventional CC>2 analyzer used to monitor the CC>2 concentration in the ambient air diluent to
assist in identifying changes in air mass or nearby CC>2 sources that could impact the CC>2
concentration or 513C. Keeling plots were also investigated as a tool to evaluate the Model
Gl 101-/'s ability to detect CC>2 leaks. Keeling regression analysis can be used to determine the
513C value for CO2 source that periodically impacts the measurement location and was conducted
by plotting the measured isotopic delta (513C) versus the inverse of the CC>2 concentration and
conducting a linear regression analysis. The 513C value for CO2 source(s) was given by the
intercept of the regression line. Regression lines were calculated for data collected during the
background ambient air measurements and during the leak simulations; the resulting intercept
and 95% confidence interval of the intercept were compared to determine whether the 12CC>2
source could be detected.
15
-------
The equivalent leak rate as a function of source 513C was also back-calculated for several
relevant 513C values: -3.5%o, -20%o, and -35%o. This calculation assumed that the magnitude of
the measured changes in 513C values from ambient levels were the same for the pure (99.95%)
12CC>2 and for the calculated leak levels for 13C-depleted CC>2. This assumption allows the
calculation of CO2 release rate necessary to achieve the overall 513C change for a given 513C
source value, knowing the ambient 513C value and the flow rate of air through the ABT. It should
1 9
be noted that the use of a pure C source for leakage (i.e., -1000 %o) maximizes instrument
sensitivity to 513C, although not to total CC>2 concentration; thus, conditions during an actual leak
would be different with respect to both 513C and total CC>2 concentration.
3.3.6 A mbient A ir Monitoring
In the evenings and when the Model Gl 101-/' was not undergoing testing during Phase 3, it was
installed in the shelter near the GS wellhead and monitored ambient air. The purpose of this
activity was to evaluate data completeness and operational factors during deployment for
ambient monitoring. The ambient air measurements (CC>2 concentration and 513C) and
meteorological conditions are reported in Chapter 6 with summary statistics (average and
standard deviation). The meteorological conditions could not be monitored continuously as
stated in TQAP because the meteorological station stopped working overnight on August 4, 2010
for approximately 12 hours. The meteorological data provided supporting information, for
example to plot CC>2 concentration data as a function of wind direction, but was not used
specifically for evaluation of the performance parameters identified in the TQAP for this
verification test. As a result, the missing data did not impact the evaluation of the analyzer's
performance, although fewer data points were included in Figure 15 as a result of the
malfunction.
While the Model Gl 101-/' was installed near the wellhead, captured CC>2 that was present in the
transmission lines, was intentionally released for a brief period. The leak rate response time, the
time between initiation of the release and when the leak is detected by the Model Gl 101-/'
response, as described in Chapter 5, was determined.
3.3.7 Mobile Surveys
During Phase 3, the Model Gl 101-/ was transported to road-accessible features of the GS, such
as transmission lines and monitoring wells. The purpose of these tests was to evaluate the ease
of use and operational factors of the analyzers during use in a mobile survey mode. The Model
Gl 101-/ was installed in the back seat of a Nissan Altima hybrid sedan and operated using power
from a marine deep cycle/RV battery and power inverter. A Teflon inlet was extended out the
rear window and held in place approximately one foot from the ground and one foot from the
vehicle. The hybrid vehicle was operated in electric mode to the extent possible to avoid
contamination of the inlet line with CC>2 from the vehicle exhaust and reduce ambiguity in the
CC>2 sources being monitored. The vehicle was then driven to the features of interest listed in
Table 5, below, with an effort to approach each feature from the downwind side while the
vehicle was operating in electric mode. A Polar RS800CX with GPS capability was used to
track vehicle speed and location during the mobile surveys.
16
-------
Table 5. Mobile Survey Features
Feature Type Number of Each Type
Above-ground transmission lines 1
Deep monitoring well 2
Injection well 2
Shallow aquifer monitoring well 7
Soil gas monitoring well 4
3.3.8 Comparability to Reference Method
The comparability of the Model Gl 101-/' response was evaluated by comparing the analyzer
response to the results of reference analyses for CC>2 concentration and 5 3C, which were carried
out by NOAA and by SIL-INSTAAR, respectively. CC>2 concentration measurements were
conducted using NDIR analysis. The NDIR analysis method is described by Conway et al.(10)
and Komhyr et al.,(11) and references therein. Calibration procedures and gases for the NDIR
CC>2 analysis are described by Komhyr and coworkers/11' 2) The methods used for the carbon
isotope ratio analysis are described in Trolier et al. ^ and Vaughn et al/13-* A detailed
investigation of the calibrations, corrections, and overall uncertainties of the 513C analysis is
presented in Masarie et al./14-* in the context of an international intercomparison study of
atmospheric measurements. The sampling and analytical procedures used for these reference
measurements have been developed and documented over decades of research by NOAA and
SIL-INSTAAR. Complete details of the reference method procedures and associated QA efforts
are detailed in the references cited above and other references therein. A summary of the
reference sample collection procedure is provided in the TQAP for this verification test
(Deviation 3).
Two pairs of duplicate grab samples of ambient air were collected for reference analyses during
Phase 2, and eight pairs of duplicate samples were collected during Phase 3. Samples were sent
first to NOAA for the CO2 concentration analysis and then to SIL-INSTAAR for carbon stable
isotope analysis. The results of the reference method measurements on those 10 duplicate
sample pairs were compared to the average CO2 analyzer response recorded at the same time the
samples were collected to assess the comparability of CO2 analyzer, as described in Chapter 5.
3.3.9 Data Completeness
No additional test procedures were carried out specifically to address data completeness. This
parameter was assessed based on the overall data return achieved by the Model Gl 101-/' and was
evaluated separately for mobile survey testing.
3.3.10 Operational Factors
Operational factors such as maintenance needs, calibration frequency, data output, sustainability
factors such as consumables used, ease of use, repair requirements, and sample throughput were
evaluated based on operator observations. Battelle testing staff documented observations in a
laboratory record book (LRB) and data sheets. Examples of information to be recorded include
17
-------
the daily status of diagnostic indicators for the technology; use or replacement of any
consumables; the effort or cost associated with any maintenance or repair; vendor effort (e.g.,
time on site) for any repair or maintenance; the duration and causes of any technology down time
or data acquisition failure; operator observations about technology startup, ease of use, clarity of
the vendor's instruction manual, user-friendliness of any needed software, overall convenience
of the technologies and accessories/consumables, and the number of samples that could be
processed per hour or per day. These observations were summarized to aid in describing the
technology performance in this report.
18
-------
Chapter 4
Quality Assurance/Quality Control
QA/quality control (QC) procedures were performed in accordance with the QMP for the AMS
Center and the TQAP for this verification test. As noted throughout Chapter 3, there were four
deviations from the TQAP. These deviations and their impact are discussed in the previous
sections. QA/QC procedures and results are described in the following subchapters.
4.1 Reference Method Quality Control
The National Oceanic and Atmospheric Administration (NOAA) Earth System Research
Laboratory (ESRL) provided reference method CC>2 concentration and isotope ratio analyses in
coordination with the Stable Isotope Laboratory (SIL) at the University of Colorado's Institute
for Arctic and Alpine Research (INSTAAR) for this verification test during Phases 2 and 3.
NOAA ESRL Global Monitoring Division is the World Meteorological Organization (WMO),
Global Atmospheric Watch Central Calibration Laboratory (CCL) for CO2. The quality of the
reference measurements was assured by adherence to the requirements of the data quality
indicators (DQIs) and criteria for the reference method critical measurements, including
requirements to perform tank gas calibrations. Gas tank calibrations included participating in
periodic round-robin analyses of external standard gas cylinders, monthly analysis of tanks
spiked at known levels, and daily analysis of an internal standard consisting of dry atmospheric
air obtained from a clean air site on Niwot Ridge in the Rocky Mountains, Colorado. Table 6
shows the results of test flask analyses conducted on the days when ambient samples for this
verification test were analyzed for CO2 concentration. Figure 5 shows test flask results for 2010,
with the difference between the actual and expected concentration plotted on the y-axis. Figure 6
shows CO2 results for tests of the portable sampling unit (PSU84M) used to collect the ambient
air samples. A different nominal CO2 concentration was supplied on each of the three dates
shown. The red symbols indicate flasks filled directly from a cylinder of known mixing ratio
(controls). The blue symbols indicate flasks filled from the same cylinder, but through the
portable sampling unit. The heavy black line represents the known CO2 value and the dashed
lines indicate ±0.1 ppm.
Estimates of uncertainty associated with the reference analyses for isotopic measurements of
CO2 are based on the standard deviation of the last 30 measurements, run over ten days, for the
cylinder that is run as a known unknown. This method of determining the uncertainty provides
an estimate of the short-term precision of the isotope ratio measurement. Figure 7 shows the
IRMS precision check results for 161 runs; the reference analyses for this verification test were
included in runs 3152 and 3156. Unflagged data are shown in blue, flagged data are white, and
19
-------
results for runs for this verification test are shown in red. Error bars are one standard deviation
for triplicate measurements. The solid line shows the mean for 142 unflagged runs and the
dashed lines are one standard deviation from the mean. Based on the results of these precision
checks, the estimated uncertainty for runs 3152 and 3156 is 0.017%o and 0.015%o, respectively.
Table 6. Test Flask Results on Days when References Samples were Analyzed
Flask Number
Expected value
T26-99
T4345-99
T4453-99
T4461-99
T53-99
T91-99
Average
Average Percent Difference
COi Concentration (ppm)
380.65
380.65
380.63
380.67
380.68
380.64
380.72
380.67
0.004%
Percent Recovery
Not applicable
100.00%
99.99%
100.01%
100.01%
100.00%
100.02%
100.00%
E
Q-
CL
O
O
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
-0.3
-0.4
Test Flasks (normalized) L8
i V
mean = 0.03
sigma = 0.07
N = 281
1 1
1
1
1
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
2010
Figure 5. CO2 Test Flask Results for 2010
20
-------
PSU 84M test
E
Q.
Q.
387
386
385
o
<-> 384
383
05 Mor 07
382L_i_L
I I I I
I I 1 I
28 Sep 07
i i i
T I 1 I I I
06 Jon 09
i MI i i i
J L
012345678 910111213141516171819202122232425
Figure 6. CO2 Test Results for Portable Sampling Unit 84M
-8.40
-8.45-
• Unflagged
O Flagged
• ETV Reference Samples
-8.70
T
3000
T
3020
r
3040
T
3060
T
3080
3100
3120
3140
3160
Run Number
Figure 7. Isotope Ratio Precision Check Results
21
-------
NOAA and SIL-INSTAAR followed all standard QC procedures established for their respective
ongoing programs of CC>2 and carbon isotope measurements, respectively. The results from
those procedures, described above, are consistent with published results describing the long-term
performance of these laboratories:
• The CO? concentration reference method should have an analytical precision better than 0.1
ppm: Analytical precision of approximately 0.05 ppm is reported from extensive data sets by
Masarie et al.(14)
• The difference in CO? concentration for duplicate samples should be less than 0.5 ppm:
Masarie et al. ^ report that the average difference in results for CC>2 between duplicate
ambient air flask samples is 0.05 ± 0.12 ppm. The average difference between duplicate
ambient air flask samples for this verification test was 0.38 ppm with a range of 0.02 to 1.95
ppm.
• Accuracy for CO? concentration measurements should be better than 1%: A 7-year
intercomparison study of NOAA CC>2 measurements with those conducted by the
Commonwealth Scientific and Industrial Research Organization (CSIRO) of Australia
showed agreement within 0.21 ± 0.26 ppm/14-* This result is equivalent to agreement within
approximately 0.06%. Analyses of synthetic mixtures will be less accurate due to pressure
broadening and other effects/12'15'16'17)
• Stable carbon isotope measurements should have analytical precision better than ±0.02%o:
The dual inlet IRMS method is reported to have a precision of ± 0.01%o, based on several
hundred QC analysis runs conducted during routine analysis of ambient sample sets/13-*
• Stable carbon isotope analyses on duplicate samples should differ by less than 0.05%o: The
replicate precision for carbon isotope ratios in duplicate samples is comparable to that in
replicate analyses (i.e., ± 0.01%o). The average difference between duplicate ambient air flask
samples for this verification test was 0.03%o with a range of 0 to 0.072%o.
Although some of the duplicate sample results varied more than expected for CO2 concentration
and carbon isotope ratio (up to 0.072%o as compared to the expected <0.05%o), the results are
considered to have sufficient accuracy and precision to be used as a standard against which to
evaluate the Model Gl 101-/' responses in ambient air.
4.2 Phase 2 Leak Flow Rate Quality Control
A data quality indicator (DQI) was established for leak flow rate accuracy in Phase 2 to ensure
that data used to support the quantitative performance evaluations of the CO2 analyzers were of
sufficient quality. During Phase 2, the accuracy of the leak flow rate was verified during each
simulated leak test using an independent flow transfer standard. The acceptance criteria for the
measured value to be within ± 10% of the independent flow transfer standard. Actual results
were for the leak flow rate determined to be the minimum detectable by the Model GllOl-i was
22
-------
0.423 LPM. The flow rate was measured by an independent calibrated flow meter (DryCal DC-
2) was 0.4526 LPM. The relative percent difference (RPD) between the two values was 1.7%.
4.3 Dynamic Dilution System Quality Control
Many of the testing activities utilized gas standards of known CC>2 concentration and/or isotopic
composition to prepare multiple mixtures that were then delivered to the Model Gl 101-/'. The
Model Gl 101-/' was then evaluated against the calculated composition of the resulting mixture.
For example, a 500 ppm CC>2 mixture was prepared by diluting 9.1 standard cubic centimeters
per minute (seem) of an 11% ±1% CC>2 standard to 2000 seem using an Environics 6100 dilution
system. The concentration accuracy of the dilution system is reported at ±1.0% of the setpoint
when the mass flow controllers are operating between 10% and 100% of full scale flow/1 ^ The
calibration of the Environics 6100 dilution system was checked prior to the verification test
against a DH Instruments Moblox Flow Terminal. The calibration was conducted by the Battelle
Instrument Services Laboratory, which is accredited to the International Organization for
Standardization (ISO) 17025 standard. The uncertainty of the flow standard used for the
calibration was 0.2% of the reading. All flows tested (4 per flow controller) were within 1% of
the set point; therefore, the calibration was not adjusted. Given these values, one would expect
that the uncertainty of the concentrations delivered from the Environics 6100 would be ±3% of
the reading, which accounts for uncertainty in flows from two flow controllers used to prepare
each dilution and in the gas standard accuracy. As discussed in Section 4.4.1, the reference
method analyses of dilutions prepared by this system had an uncertainty (95% confidence level,
or two times the standard deviation) that is 3-4 times larger than this estimate. The overall
uncertainty in the prepared concentrations is therefore estimated to be ±7% (95% confidence
level).
The accuracy of the isotope ratio as a result of dilution is known only for samples that were
analyzed by the SIL as dilution may unintentionally fractionate isotopes. Dilutions from one
isotope mixture were analyzed by the reference method and found to be within 1.12% of the
expected value (see Section 4.4.1). These dilutions were prepared at total flow rates (16.5 LPM)
higher than used to evaluate the Model Gl 101-/ (approximately 5 LPM) and only at
concentrations near ambient levels (see Table 7); it is not known whether fractionation affected
dilutions used to evaluate the Model Gl 101-/.
4.4 Audits
Three types of audits were performed during the verification test: a performance evaluation (PE)
audit of the CO2 concentration and isotope ratio reference methods, a technical systems audit
(TSA) of the verification test procedures, and a data quality audit. Audit procedures are
described further below.
4.4.1 Performance Evaluation Audit
The PE audit of the CO2 reference methods (concentration and isotope ratio) was performed by
supplying to each reference method four independent CO2 standards (in duplicate) provided by
Battelle. The four PE samples are summarized in Table 7, below. PE samples were prepared by
23
-------
diluting certified gas standards [-10.41%o pure CC>2 compressed gas standard (SMU Stable
Isotope Laboratory, through Oztech Trading Corporation) and 11.0% ±1% CC>2 in air (Air
Liquide Acublend Master Class)] with Alphagaz Zero Air using a calibrated dilution system.
The diluent was supplied to the reference method sampling inlet, with excess flow vented to
atmospheric pressure, and collected in glass flasks in the same manner as ambient air samples.
The PE samples were analyzed for CO2 in the same manner as for all other ambient air samples
and the analytical results for the PE samples were compared to the nominal concentration/isotope
ratio. The actual results are presented in Table 7. The root mean square of the differences
between the nominal value and the reference method is 3.4% and all were within the target
criterion (30%) for this verification test. The audit samples were supplied to the reference
laboratories with the ambient air samples from the verification test. Since the reference methods
were carried out by laboratories that hold the WMO CCL for CO2, reference method results are
considered to be the true value.
Table 7. PE Sample Results for COi Concentration and Isotope Ratio
Sample
1
2
O
4
Concentration (ppm)
Nominal
329.9
379.8
349.8
392.8
Reference
Method
Result
334.61
334.11
378.74
378.91
329.51
329.34
379.93
379.80
Percent
Difference
1.4%
1.3%
-0.3%
-0.2%
-5.8%
-5.8%
-3.3%
-3.3%
Carbon Isotope Ratio (%o)
Nominal
-10.41
-10.41
Not certified;
evaluated for
concentration
only
Reference
Method
Result
-10.293
-10.295
-10.304
-10.340
-37.009
-37.014
-36.980
-37.032
Percent
Difference
-1.12%
-1.10%
-1.02%
-0.67%
NA
NA
4.4.2 Technical Systems Audit
The Battelle AMS Center Quality Assurance Officer for this verification test performed a TSA
during the both the laboratory and field testing portions of this verification test to ensure that the
verification test was performed in accordance with the QMP for the AMS Center and the TQAP.
The EPA Quality Manager also observed field testing.
The TSA of the laboratory portion of the verification test was performed on July 9, 12, 15, and
22, 2010 at Battelle's Air Quality and Temperature/Humidity Chamber Laboratories in
Columbus, OH. During this TSA, the Battelle AMS Center Quality Assurance Officer observed
verification testing staff conducting tests for concentration accuracy, isotope bias, and
temperature/RH bias.
24
-------
The TSA of the field testing portion of the verification test was performed on August 5, 2010 at
the GS site. During this TSA, the Battelle AMS Center Quality Assurance Officer observed
verification testing staff complete the analyzer daily checklist, troubleshoot the meteorological
station, collect ambient air samples, and load the Model Gl 101-/ analyzer into the Nissan Altima
hybrid sedan in preparation for mobile survey tests.
The TSA of both the laboratory and field testing portions resulted in 3 findings and 5
observations. The first finding was that quantity of gas standards available on July 15, 2010 for
the isotope ratio bias tests at -3.61 per mil CO2 may have been insufficient to conduct triplicate
runs. To address this finding, which also applied to other tests, additional materials were
purchased, the time at each test condition was shortened, and the third replicate at each isotope
ratio/concentration combination was removed from the test. This finding was also addressed in
Deviation 1 as described in Section 3.3.2. The second finding was that the RH values in the
environmental chamber exceeded the limits described in the TQAP for two test conditions. This
finding was addressed in Deviation 2 and is described in Section 3.3.4 of this report. The third
finding was that the meteorological conditions could not be monitored continuously as stated in
TQAP because the meteorological station stopped working overnight on August 4, 2010 for
approximately 12 hours. The cause of the problem was investigated and the meteorological
station started working on August 5, 2010. This finding was addressed in Deviation 4 and
described in Section 3.3.6 of this report. The observations were related to vendor training,
instability in the dilution system flow rates, inconsistencies in the Model Gl 101-/' response to
zero air, apparent slow analyzer response to varying test conditions, and calibration of the
independent flow meter. In response to these observations, the following actions were taken:
• Documentation of the vendor's instructions regarding training of verification testing staff
was provided to the Battelle AMS Center Quality Assurance Officer and included in the
project files
• Performance of the Environics 6100 was investigated and a faulty pressure regulator
discovered and replaced
• Troubleshooting revealed that CO2 present in the dilution system's internal components
diffused into the zero air stream; an alternate method for delivering zero air by bypassing the
dilution system was identified and utilized; affected testing was repeated
• Care was taken to ensure that internal components of the Environics dilution system were
thoroughly flushed prior to starting tests after switching the CO2 source (compressed gas
cylinder)
• The flow meter in use was sent out to be calibrated and a different flow meter was used for
the remainder of the verification test.
TSA reports were prepared and copies were distributed to the EPA.
4.4.3 Data Quality Audit
Records generated in the verification test were reviewed before these records were used to
calculate, evaluate, or report verification results. Data were reviewed by a Battelle technical
staff member involved in the verification test. The person performing the review added his/her
initials and the date to a hard copy of the record being reviewed.
25
-------
100% of the verification test data was reviewed for quality by the Verification Test Coordinator,
and at least 10% of the data acquired during the verification test and 100% of the calibration and
QC data were audited. The data were traced from the initial acquisition, through reduction and
statistical analysis, to final reporting to ensure the integrity of the reported results. All
calculations performed on the data undergoing the audit were checked.
A data audit report was prepared and a copy was distributed to the EPA.
26
-------
Chapter 5
Statistical Methods
The statistical methods used to evaluate the quantitative performance factors listed in Section 3.3
are presented in this chapter. The average Model Gl 101-/' response values (Y) used in the
calculations presented in this section were calculated from the last 5 minutes of each testing
condition (e.g., CC>2 gas standard challenge) from the analyzer's raw data (-0.5 Hz) unless
otherwise indicated. Qualitative observations were also used to evaluate verification test data.
5.1 Model GllOl-i Factor Calibration Correction
The dependence of the delta value measured by Model Gl 101-/' on CO2 concentration is defined
by a constant that is determined during factory calibration tests. For the Model Gl 101-/' that was
evaluated in this verification test, an incorrect value (-0.01686) was input into the software.
This error was reported to the Verification Test Coordinator by Picarro. The error was allowed to
be corrected because the notification of correction was distributed to all users of Picarro
instruments that had been calibrated during that factory calibration test and was not something
that was identified based on the verification test data. To address this error, the values for 513C
were corrected using Equation 2, which was provided by Picarro:
B
"corrected ~ "reported + ^ + 12rn (2)
C t/2
where A is equal to -4.86%o and B is 2156 ppm.
5.2 Water Vapor Correction
Water vapor can interfere with the measurement of the carbon dioxide concentration and isotope
ratios in the following ways:
• Dilution - The dilution effect is simply the change in mixing ratio of CC>2 caused by
variability in the humidity. For example, a dry air mass traveling over warm water will
accumulate humidity, and this additional water vapor will dilute the concentration of the
other gases. Conversely, a humid air mass that becomes drier (as through precipitation) will
cause an inverse dilution effect, increasing the mixing ratios of the other gases. Because it
19 1 ^
affects C and C equally, dilution affects only the concentration, not the reported isotope
ratio. The magnitude of the effect is a 1% decrease in the reported fractional concentration
27
-------
for every 1% increase in water vapor concentration. The dilution effect is largely due to the
most abundant isotopolog of water (1H216O), which is 99.8% of all the water under most
conditions.
• Spectral broadening - The Lorentzian broadening of the spectral lines are affected by the
presence or absence of water vapor. The magnitude of the effect on the reported
concentrations is of the order of the dilution effect (though generally somewhat smaller). The
effect on each of the two lines (i.e., the 12C and 13C spectral lines) is not necessarily identical,
leading to a systematic error in the reported isotope ratio as a function of water vapor
concentration. The effect on delta is proportional to water vapor concentration and
independent of CO2 concentration. As with dilution, this effect is largely due to 1H216O.
• Direct spectral interference - Direct spectral interferences are caused by any water vapor
spectral lines that are in the immediate vicinity of either the 12C or 13C spectral lines. These
can cause offsets to these two gas species that affect both the concentration and reported
isotope ratio. The effect on delta is proportional to the product of the water vapor
concentration and inversely proportional to the carbon dioxide concentration. Unlike
dilution, this effect can depend on whichever isotopolog or isotopologs are interfering with
the 12C and 13C measurements.
Consequently, corrections for water vapor were made to the data reported by the Gl 101-/' before
the data were used in calculations described here or otherwise reported.
5.2.1 Delta Water Vapor Correction
Reported Model Gl 101-/' values for delta were post-corrected for water vapor interferences using
Equation 3 :
( .850 \
Ucorrected_wv = "corrected ~~ "2 ^reoj-ted X I ~~~"0.67 +
where 5correcte^ is the corrected delta value (Equation 1), and uCO2Wet and H2Oreported are the
corresponding 12CC>2 concentration and water vapor molar ratio (%v) reported by the Model
Gl 101-/' , respectively. The resulting delta value, ^corrected >v was used for the calculations
described in the following sections.
5.2.2 CO 2 Concentration Water Vapor Correction
1 9
Reported Model Gl 101-/' values for CC>2 concentration ( CO2Wet) were corrected for water vapor
dilution using Equation 4:
where // equal to -0.01527, m' is 0.358, and H2Oreported is the corresponding water vapor molar
ratio reported by the Model Gl 101-/.
Equation 5 was used to calculate total (12 + 13)
28
-------
12rn + 13ro
2dry -I" 2dry (5)
where, as derived from Equation 1 :
13ro _ 12rn y I 13r/12r y /tfcorrected_wp , -, \]
2 gas
standard dilutions was assessed as the percent recovery (%R), using Equation 6:
%R = [l + (-^)] x 100 (6)
where Y is the average measured CO2 analyzer value (CO^) and X is the nominal CO2 gas
standard concentration. The average, minimum, and maximum %R values are reported for each
series of multi-level CO2 challenges. The accuracy of the analytical standards, as certified by the
manufacturer, is also reported.
The calculation for %R implies that the error would scale with the magnitude of the measured
value, which is true for concentration measurements. However, error in the isotope ratio does
not scale with the magnitude of the delta value, which is arbitrarily set based on a recognized
standard. Therefore, the accuracy of the Model Gl 101-/ with respect to isotope ratio was
determined as the difference between the Model Gl 101-/' measured value and the certified
standard (nominal 513C) value.
5.4 Bias
Bias of the Model Gl 101-/' is defined as a systematic error in measurement that results in
measured error that is consistently positive or negative compared to the true value. The bias was
calculated as the average percent difference (%D) of the Model Gl 101-/ response compared to
the nominal CC>2 gas standard value (with respect to concentration and isotope ratio) and was
calculated for each series of multi-point CO2 challenges and isotope ratio bias tests, using
Equation 7:
X 100 (T
where k is the number of valid comparisons, and Y and X are the same as stated in 5.3. For
temperature and RH bias, the comparison for concentration and isotope ratio utilized the values
measured by the CO2 analyzer at 20°C without added water vapor (dry conditions) as the
"known" value.
29
-------
The calculation for %D implies that the error would scale with the magnitude of the measured
value, which is true for concentration measurements. However, error in the isotope ratio does
not scale with the magnitude of the delta value, which is arbitrarily set based on a recognized
standard. Therefore, the accuracy of the Model Gl 101-/' with respect to isotope ratio was
determined as the difference between the Model Gl 101-/' measured value and the certified
standard (nominal 513C) value.
5.5 Precision
The precision of the Model Gl 101-/' was evaluated from the triplicate responses to each CC>2 gas
standard supplied during the multi-point challenges (summarized in Table 2 and Table 3). The
precision is defined as the percent relative standard deviation (%RSD) of the triplicate
measurements and calculated for each CO2 concentration and isotope ratio listed in Table 2 and
Table 3, respectively, using Equation 8:
%RSDf = = x 100 (8)
YI
where Y, is the average Model Gl 101-/' response at CO2 concentration or isotope ratio /', and s
the standard deviation of the analyzer responses at that concentration. The overall average
%RSD was also calculated for each series of multi-point CO2 challenges (with respect to CO2
concentration) and for each gas standard (with respect to CC>2 513C) included the %RSD for all
CO2 concentrations tested for each gas source.
The calculation for %RSD implies that the error would scale with the magnitude of the measured
value, which is true for concentration measurements. However, error in the isotope ratio does
not scale with the magnitude of the delta value, which is arbitrarily set based on a recognized
standard. Therefore, the accuracy of the Model Gl 101-/ with respect to isotope ratio was
determined as the difference between the Model Gl 101-/' measured value and the certified
standard (nominal 513C) value.
5.6 Linearity
Linearity with respect to concentration and isotopic ratio was assessed by a linear regression
analysis of the gas challenge data using the calculated CO2 concentrations or 513C as the
independent variable and the CC>2 analyzer results as the dependent variable. The results of the
gas challenge tests were plotted and linearity was expressed in terms of slope, intercept, and
coefficient of determination (R2).
30
-------
5.7 Minimum Detectable Leak Rate
The minimum detectable leak rate that represents the minimum level successfully identified by
each of three trials was determined experimentally and all trial results are reported in Chapter 6.
The equivalent leak rate at several 513C values of interest were calculated based on the flow rate
of CC>2 that would be needed to give the same change in measured 513C with respect to ambient
air (approximately -8%o).
The relative difference in stable carbon isotope from the PDB standard, referred to as 513C, is
calculated as shown in Equation 1 and expressed in %o, or part per thousand.
- 0*iooo 0)
Using Equation 4 and inputs for ambient air concentration (391.62 ppm), ambient air 513C
(-7.13), 13C/12CPDB (0.011237), a leak source CO2 concentration (100% or IxlO6 ppm), a leak
source 613C value (-35%o, -20%o, and -3.5%o,), a leak flow rate (e.g., 0.423 LPM), and a total
flow rate, the 513C value for ambient air with a prescribed leak were calculated. Microsoft Excel
was used to create a spreadsheet that calculated the interim variables needed to determine the
final predicted 513C for ambient air with a specific CC>2 leak. In the spreadsheet, the total flow
rate was adjusted from the ABT value (3.40xl06 LPM) to the total flowrate needed to accurately
predict the difference between ambient and ambient air with the 0.423 LPM 12CC>2 leak
simulated during the verification test (613C = 0.808%o). The resulting flow rate, 1.28xl06 LPM,
was used for the subsequent calculations. Using the same spreadsheet, Microsoft Excel's Goal
Seek function was used to set the difference in detected 513C to ± 0.808 by varying the leak flow
10 1 '*
rate. This was tested first for 99.95% CC>2 and repeated for three leak source 5 C values
(-35%o, -20%o, and -3.5%o).
In addition, the 513C values measured by the Model Gl 101-/' analyzer was plotted versus the
inverse of the CO2 concentration (i.e., in Keeling plots) and the uncertainty (95% confidence
interval) in the intercept determined. Separate regressions were calculated for the background
measurement period and leak simulations. Theoretically, the intercept will represent the isotopic
ratio of the leak source. The intercept and uncertainty for the background measurement period
and leak simulations were compared for each day of testing to determine whether the 12CC>2
source could be detected (i.e., the intercept for the leak simulation regression fell outside the
background measurement period's intercept confidence interval.
5.8 Response Time
Response time was assessed in terms of both the rise and fall times (with respect to CO2
concentration) of the Model Gl 101-/' when sampling CC>2 gas standards. The rise time (i.e., 0%
to 95% response time) was determined by recording all CC>2 analyzer readings as the gas
supplied to the analyzers is switched between CC>2 standards of increasing concentration. Once a
stable response has been achieved with the gas standard, the fall time (i.e., the 100% to 5%
-------
response time) was determined in a similar way, by recording all CC>2 analyzer readings as the
CC>2 concentration in the gas supplied is reduced in concentration. Rise and fall times were
determined once during multi-point gas challenges and reported in units of seconds and minutes.
The leak response time was calculated as the time elapsed between the start of the intentional
CC>2 release and two different outcomes: 1) when a visible increase in CC>2 concentration was
observed in the data and 2) when an alarm limit of 1,000 ppm had been reached. In addition, the
leak response time with respect to isotope ratio was determined as the time between the start of
the intentional release and when a value for 513C that was at least 2 standard deviations different
from the average ambient value was reached.
5.9 Comparability
Comparability between the Model Gl 101-/' results and the reference method results for both CC>2
concentration and carbon isotope ratio were calculated in terms of accuracy and bias using
Equation 1 and Equation 2, respectively, for these performance parameters.
5.10 Data Completeness
Data completeness was assessed based on the overall data return achieved by the Model Gl 101-/
during the testing period. The calculation will use the total number of hours during which
apparently valid data was reported by the Model Gl 101-/ divided by the total number of hours
when data could potentially have been collected during the entire field period. The causes of any
incompleteness of data return were established from operator observations or vendor records,
and noted in the discussion of data completeness results in Chapter 6.
32
-------
Chapter 6
Test Results
The results of the verification test of the Model Gl 101-/' are presented in this section. The
analyzer logged raw data (-0.5 Hz) as well as 30-second, 2-minute, and 5-minute running
averages of the instantaneous CO2 concentration and 513C readings. The Model Gl 101-/' was
operated using the factory calibration. Any differences between the factory calibration and the
certified gas standards would be manifested in the accuracy, bias, and potential comparability
performance parameters evaluated during this verification test; other performance parameters
such as the linearity coefficient of determination, precision, and response time would not be
impacted by differences in gas standards or calibration methods because of the nature of the
calculations. All Model Gl 101-/' measurement data were corrected for an error in the factory-set
calibration (affecting the delta value only) and water vapor as described in Chapter 5. CC>2
19 1 ^
concentrations are the sum of CC>2 and CC>2. Negative CO2 concentration values should be
considered the result of differences in gas standards used to calibrate the Model Gl 101-/' and the
gas standards used to evaluate the analyzer during this verification test. Negative CO2
concentration values can also be caused by drift in the Model Gl 101-/' response. Error bars
shown on the x-axes represent flow rate accuracy of the Environics system that was used to
prepare the gas standard dilutions. Errors due to isotope fractionation or other artifacts have not
been quantified.
6.1 Concentration Accuracy and Bias
Accuracy checks were conducted during Phase 1 of the verification test. The Model Gl 101-/'
was challenged with dilutions from a compressed CO2 standard at concentrations between 100
and 5,000 ppm. The standard was diluted with zero air and delivered to the Model Gl 101-/'
through a Teflon tube at a flow rate of 2 LPM; excess flow was vented to atmospheric pressure.
Figure 8 presents the CC>2 concentrations recorded by the Model Gl 101-/' during the
concentration accuracy check gas challenges, along with the nominal CC>2 concentration levels
supplied to the analyzer. The averages of the last 5 minutes (approximately 250 data points) of
measurements at each test condition and the calculated %R values are presented in Table 8. The
SD for each measured concentration is also reported for reference purposes. As shown in
Table 8, %R values ranged from 90.0 to 113%. Except at concentrations less than 300 ppm, the
Model Gl 101-/' response was lower than the nominal value. Bias in the Model Gl 101-/
concentration response to CC>2 gas standards was assessed for the accuracy checks and was
-3.98%. Given the uncertainty estimate for the nominal CC>2 concentrations of ±7%, it is not
33
-------
possible to determine from these measurements alone whether the observed inaccuracies and
biases are due to errors in the instrument response or the gas preparation.
5000-
4000-
3000-
2000-
1000-
5000
4000 -I
3000
c
§ 2000 H
o
0 1000
CM
o
O 0
o Measurement data
Nominal CO2 concentration
Nominal CO2 95% Confidence Level
-_-_^
fi
ri
u
00:00 01:00 02:00 03:00
Elapsed Time (hh:mm)
Figure 8. Model G1101-/Concentration Accuracy Results
04:00
05:00
34
-------
Table 8. Concentration Accuracy Results
Measurement
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
Average
Minimum
Maximum
Bias (%D)
CO2 Gas Standard
Concentration (ppm)
0
100
200
300
400
500
750
1600
2450
3300
4150
5000
5000
300
750
0
1600
200
3300
500
2450
100
400
4150
5000
4150
3300
2450
1600
750
500
400
300
200
100
0
Average G1101-/ (12+13)CO2
Response (ppm)
0.1(a)
110.6
205.8
295.9
374.5
452.8
675.0
1491.8
2304.0
3019.2
3901.2
4735.2
4725.1
297.9
675.7
0.9
1491.2
207.6
3017.7
455.1
2301.5
113.1
375.5
3897.3
4730.2
3894.2
3012.9
2304.4
1494.9
677.9
454.7
376.1
297.1
207.2
111.3
0.6
SD
(ppm)
0.1
0.2
0.2
0.2
0.2
0.2
0.2
0.3
0.4
0.4
0.6
0.5
0.6
0.2
0.2
0.1
0.3
0.3
0.3
0.2
0.5
0.2
0.2
0.6
0.6
0.5
0.5
0.3
0.4
0.2
0.2
0.2
0.2
0.2
0.2
0.1
%R
NA(b)
110
102
98.6
93.6
90.6
90.0
93.2
94.0
91.5
94.0
94.7
94.5
99.3
90.1
NA
93.2
104
91.4
91.0
93.9
113
93.9
93.9
94.6
93.8
91.3
94.1
93.4
90.4
90.9
94.0
99.0
104
111
NA
96.0
90.0
113
-3.98
(a) Data in this table are reported to the number of significant digits appropriate for the analyzer's reported
accuracy and precision
(b) NA = Not applicable
35
-------
6.2 Concentration Precision
Table 9 presents the calculated precision of the Model Gl 101-/' determined from the average
responses to the triplicate challenges at each CO2 concentration level during the concentration
accuracy checks. The precision of the Model Gl 101-/' response varied from 0.1% to 1.2%. The
highest %RSD value was observed for the lowest concentration standard (100 ppm). The
average precision was 0.3%. The precision of the Environics 6100 gas dilution system is not
specified by the vendor. It is therefore not possible to determine from these measurements alone
whether the observed precision was limited by the Gl 101-/' instrument or the gas dilution system.
Table 9. Concentration Precision Results
Gas Standard Overall Average
Concentration G1101-i(12+13)CO2
(ppm) Response (ppm) %RSD
0
100
200
300
400
500
750
1600
2450
3300
4150
5000
Average %RSD
0.5(a)
111.7
206.9
297.0
375.4
454.2
676.2
1493
2303
3017
3898
4730
NA(b)
1.2%
0.5%
0.3%
0.2%
0.3%
0.2%
0.1%
0.1%
0.1%
0.1%
0.1%
0.3%
(a) Data in this table are reported to the number of significant digits appropriate for the analyzer's reported
accuracy and precision
(b) NA = Not applicable
6.3 Concentration Linearity
Figures 9 and 10 show the linearity results for the concentration accuracy checks. A linear
regression was calculated from the results presented in Table 8 (average Model Gl 101-/'
response versus nominal CC>2 gas standard concentration) over the range of 0 to 5,000 ppm and
from 0 to 400 ppm. The 95% confidence interval for the slope and the intercept of each line was
also calculated (shown in the following text within parenthesis). For 0 to 5,000 ppm, the slope of
the regression line was 0.938 (±0.006), with an intercept of-1.32 (±13.6) and R2 value of
0.9997. For the 0 to 400 ppm data set, the slope of the regression line was 0.935 (±0.036), with
an intercept of 11.3 (±8.9) and R2 value of 0.9958. As shown in Figures 9 and 10, the 95% CI for
the regression line includes the 1:1 line for concentrations less than approximately 950 ppm.
36
-------
5000-
•j? 4000-
CL
a.
%
o 3000-1
Q-
CSf
<
2000-
1000-
0-
o Measurement Data
Linear Regression
95% Confidence Interval
— 1 to 1 Line
I I I
0 1000 2000 3000
Nominal CO2 Concentration (ppm)
I I
4000 5000
Figure 9. Model G1101-/Concentration Linearity Results
400-
E
&300
o
CL
(0
CD
200 -
(D
>>
(0
100 -
0 -
o Measurement Data
Linear Regression
95% Confidence Interval
— 1 to 1 Line
I
0 100 200 300
Nominal CO2 Concentration (ppm)
Figure 10. Model G1101-/Concentration Linearity Results
400
37
-------
6.4 Response Time
Response time was determined during the concentration accuracy checks from the amount of
time required for the Model Gl 101-/' to reach 95% of the change in response for the sequentially
increasing and decreasing test conditions shown in Figure 7. The 95% rise and fall times are
presented in Table 10. The average 95% rise time was 2.43 minutes and the 95% fall time was
2.53 minutes. Response times were calculated from raw (unaveraged) data. It is not possible to
determine from these measurements alone whether the observed response time is limited by the
response of the Model Gl 101-/ or the gas dilution system.
Table 10. Response Time Results
Measurement
Number
1
2
3
4
5
6
7
8
9
10
11
12
25
26
27
28
29
30
31
32
33
34
35
36
Average
Minimum
Maximum
CO2 Gas Standard
Concentration (ppm)
0
100
200
300
400
500
750
1600
2450
3300
4150
5000
5000
4150
3300
2450
1600
750
500
400
300
200
100
0
95% Rise
(seconds)
NA(a)
(b)
134.0(c)
133.2
159.8
161.8
160.7
163.8
174.4
96.6
123.4
112.2
142.0
96.6
174.4
Time
(minutes)
NA
(b)
2.23
2.22
2.66
2.70
2.68
2.73
2.91
1.61
2.06
1.87
2.37
1.61
2.91
95% Fall Time
(seconds)
130.6
141.7
90.0
159.8
179.9
174.0
158.4
178.6
168.9
217.5
67.3
130.6
151.5
67.3
217.5
(minutes)
2.18
2.36
1.50
2.66
3.00
2.90
2.64
2.98
2.82
3.63
1.12
2.18
2.53
1.12
3.63
(a) NA = Not applicable
(b) Could not be calculated due to interruption in gas flow to connect analyzer to dilution system.
(c) Data in this table are reported to the appropriate number of significant digits based on precision of the
analyzer's reported data.
38
-------
6.5 Isotope Ratio Accuracy and Bias
Isotope ratio accuracy was evaluated during Phase 1. The Model Gl 101-/' was challenged with
dilutions from three certified CO2 isotope mixtures: -3.60%o, -10.41%o, and -40.80%o. The
standards were diluted with zero air using the Environics 6100 dilution system and delivered to
the Model Gl 101-/' through a Teflon tube at a flow rate of 5 LPM; excess flow was vented to
atmospheric pressure. Higher total flow rates were used for this test to accommodate flow rate
limits of the dilution system.
Table 11 presents the average 513C response recorded by the Model Gl 101-/' during the isotope
ratio accuracy check gas challenges, along with the nominal 513C and CO2 concentration levels
supplied to the analyzer. The averages of the last 5 minutes (approximately 250 data points) of
measurements at each test condition and the calculated absolute differences from the expected
value are presented. The SD for each measured concentration is also reported for reference
purposes. As shown in Table 11, the differences ranged from 1.1 to 2.7%o. The lowest absolute
differences were observed for the -40.80%o standard and at the higher CC>2 concentrations.
Figure 11 shows the isotope ratio absolute differences plotted as a function of CO2 concentration
for each 513C value. The strongest correlation between concentration and measured isotope ratio
r\
was observed for the -3.60%o standard, with an R value of 0.9468, a slope of-0.0025 and
intercept of 3.24.
39
-------
Table 11. Isotope Ratio Accuracy Results
Isotope Ratio (%o)
Gas Standard Gas Standard
Isotope Ratio Concentration
(%o) (ppm)
259
370
37:
740
259
259
370
37:
740
259
259
370
%
740
259
Average
Minimum
Maximum
Average
GllOl-i
Response (%o)
-0.9(a)
-1.2
-2.2
-1.6
-2.3
-1.0
-8.6
-9.1
-9.2
-9.0
-9.3
-8.5
-39.0
-39.7
-39.7
-39.1
-39.7
-39.0
Difference
from
SD Standard (%o)
2.2
1.6
0.8
1.8
0.8
2.2
2.0
1.9
0.9
1.6
0.9
2.3
2.0
1.6
0.8
1.5
0.9
2.0
1.5
0.8
2.3
2.7
2.4
1.4
2.0
1.3
2.6
1.8
1.3
1.2
1.4
1.1
1.9
1.8
1.2
1.1
1.7
1.1
1.8
1.7
1.1
2.7
Average
Difference
(%o)
2.1
1.5
1.4
(a) Data in this table are reported to the number of significant digits appropriate for the analyzer's reported
accuracy and precision.
40
-------
»-3.60permil "-10.41 permil -40.80permil
0) A
a. 1
o
+*
o
I/)
-3.60%o
y = -0.0025x + 3.24
R2 = 0.9468
-10.41%o
y =-0.0013x +2.04
R2= 0.7093
-40.80%o
y =-0.0013x +2.05
R2= 0.6742
0 100 200 300 400 500 600 700 800
Concentration
Figure 11. Model G1101-/Isotope Ratio Accuracy Results
6.6 Isotope Ratio Linearity
Figure 12 shows the linearity results for the isotope ratio accuracy checks. A linear regression
was calculated from the results presented in Table 12 (average Model Gl 101-/' response versus
nominal CO2 gas standard 613C) for -3.60%o, -10.41%o, and -40.80%o standards. The 95%
confidence interval for the slope and the intercept of each line was also calculated (shown in the
following text within parenthesis). The slope of the regression line was 1.01 (± 0.02) with an
intercept of 1.88 (± 0.37) and R2 of 0.9992. As shown in Figure 12, the 95% confidence interval
of the regression line falls just above the 1 to 1 line.
41
-------
-30
-10-
ro
01
2. -20
o
(0
OJ
CD
-30 H
I
-20
I
-10
I
0
o Measurement Data
Linear Regression
95% Confidence Interval
— - 1 to 1 Line
y = 1.01(±0.02)x + 1.88(±0.37)
R2 = 0.9992
T
T
-10
\
0
-30 -20
Nominal Isotope Ratio (%o)
Figure 12. Model GllOWIsotope Ratio Linearity Results
6.7 Temperature and RH Bias
Results for the temperature and RH bias tests are summarized in Table 12. To focus only on the
impact of temperature and RH on the measurements, both concentration and 135C values at the
various temperature and RH combinations were compared to the Model Gl 101-/' results at 20°C
and 0% RH. Bias was calculated for each temperature/RH combination and included in
Table 12. In general, variability in ambient temperature and RH conditions resulted in bias
values of 3% or less for Model Gl 101-/' concentration measurements; isotope ratio values were
within 0.7%o of the value observed at 20°C and 0% RH. The maximum concentration bias value,
3.0%, was observed for CO2 concentration at 4°C/50% RH. The largest isotope ratio average
difference of 0.7%o was observed for 32°C/90% RH. Given the uncertainty estimate for the
nominal CC>2 concentrations of ±7%, it is not possible to determine from these measurements
alone whether the observed non-linearity is due to errors in the instrument response or the gas
preparation.
42
-------
Table 12. Temperature and RH Bias Results
Nominal
Temp-
erature
(°C)
20 ± 2°C
32 ± 2°C
4±2°C
RH
(%)
0
±10%
50
±10%
90
±10%
50
±10%
90
±10%
50
±10%
Cone.
(ppm)
350
500
1000
350
500
1000
350
500
1000
350
500
1000
350
500
1000
350
500
1000
Concentration
Response
(ppm)
324.5
447.6
935.6
327.8
452.5
946.0
333.4
456.4
952.0
333.9
457.6
957.2
334.1
459.8
959.4
337.6
460.2
957.3
%R
100.0
100.0
100.0
101.0
101.1
101.1
102.7
102.0
101.7
102.9
102.2
102.3
102.9
102.7
102.5
104.0
102.8
102.3
Bias
(%D)
0.0%
1.1%
2.1%
2.5%
2.7%
3.0%
Isotope Ratio (%o)
Response
-35.0
-35.8
-36.1
-34.9
-35.6
-36.0
-34.5
-35.2
-36.0
-34.4
-35.1
-36.0
-34.2
-34.8
-35.6
-35.3
-35.8
-36.3
Differ-
ence
0.6
-0.2
-0.5
0.7
0.0
-0.4
1.1
0.4
-0.4
1.1
0.5
-0.4
1.3
0.8
-0.0
0.3
-0.2
-0.7
Avg
Differ-
ence
0
0.1
0.4
0.4
0.7
-0.2
(a) Data in this table are reported to the number of significant digits appropriate for the analyzer's reported
accuracy and precision.
6.8 Minimum Detectable Leak Rate
Testing for the minimum detectable leak rate took place during Phase 2 at the ABT over the
course of three days. The testing involved simulation of a pure 12CC>2 leak. The following
factors that impact leak detestability were controlled or characterized: isotopic signature of the
leaking CC>2(99.95% 12CC>2), isotopic signature of the ambient air (measured), meteorological
conditions (constant 1.8 m/s velocity in ABT), and sampling proximity to leak (fixed). The
impact from CC>2 sources was limited by observing activity, primarily vehicular traffic, near the
ABT. Control of those factors allowed for evaluation of the Model Gl 101-/' performance for
19
leak detection under a well-defined set of conditions. As described in Section 3.3.5, CC>2 was
added to the ambient air diluent being drawn into the ABT to simulate a low-level leak of 13C-
depleted CC>2. The results of the testing are summarized in Table 13. The 12CC>2 flow rate was
increased gradually from 0.156 LPM to 0.423 LPM, where it was detected successfully three
times. Thus, the minimum detectable leak rate under the conditions of this verification test was
0.423 LPM. Time series plots for each day of testing are included in Figures 13, 14, and 15. In
each figure, the upper panel shows CC>2 concentration measurements reported by the Model
Gl 101-7 (blue line) and by the LI-COR LI-820 NDIR CO2 Analyzer (green line). The Model
Gl 101-/' 513C are shown (black trace) on the lower panel (primary axis). The gray shaded areas
show periods when 12CC>2 was released and the leak flow rate (lower panel, secondary axis). The
43
-------
red lines show the Model Gl 101-/' average 513C response. Visual comparison of the Li-Cor CO2
measurements, which were conducted upstream of the 12CC>2 leak, and the Model Gl 101-/ data
suggests that there was no detectable CO2 concentration change with the addition of the
relatively low flow rate of 12CC>2. This was an expected result based on calculations of the
expected CO2 concentration change.
Table 13. Minimum Detectable Leak Rate Results
2SD
Ambient Air
513C
Variability
Date (%o)
7/28/2010 0.549
7/29/2010 0.545
7/30/2010 0.632
12CO2 Flow
Rate
(LPM)
0
0.156
0
0.311
0
0.349
0
0.349
0
0.349
0
0
0.423
0
0.423
0
0.423
0
0.423
0
0
0.623
0
0.623
0
0.623
0
Average
Model G1101-
i513C
Response (%o)
-6.1(a)
-5.8
-5.4
-5.9
-5.5
-5.7
-5.3
-5.8
-5.2
-5.8
-4.9
-7.9
-8.4
-6.8
-7.4
-6.5
-7.0
-5.8
-6.2
-5.0
-5.5
-6.0
-4.9
-5.7
-4.8
-6.0
-4.8
Ambient
Air
Average
513C (%o)
-5.7
-5.4
-5.4
-5.2
-5.1
-7.4
-6.6
-6.2
-5.4
-5.2
-4.8
-4.8
O ^ leak ~
el3r
O ^ambient
-0.1
-0.5
-0.3
-0.6
-0.7
-1.1
-0.8
-0.8
-0.8
-0.8
-0.8
-1.2
Leak
Detected?
No
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
(a) Data in this table are reported to the number of significant digits appropriate for the analyzer's reported
accuracy and precision.
44
-------
460
Model G1101H
LICOR NDIR
-7-
i i i
10:00 AM 11:00 AM 12:OOPM 1:00 PM 2:00 PM 3:00 PM 4:00 PM
7/28/2010
Local Time
Figure 13. Model G1101-/Minimum Detectable Leak Rate Results for Day 1 (7/28/2010)
Model G1101 -i
LICOR NDIR
6:00 AM
7/29/2010
r
7:00 AM
!
8:00 AM
r
9:00 AM
10:00 AM
I
11:00 AM
0.0
Local Time
Figure 14. Model G1101-/Minimum Detectable Leak Rate Results for Day 2 (7/29/2010)
45
-------
£
o
o
O
o1
O
O
O
Model G1101-i
LICOR NDIR
10:00 AM
7/30/2010
11:00 AM
r
12:00 PM
r
1:00 PM
0.0
Local Time
Figure 15. Model G1101-/Minimum Detectable Leak Rate Results for Day 3 (7/30/2010)
For each day of testing, Keeling plots (513C versus inverse CO2) were generated for background
monitoring and leak simulation test data. Linear regressions were calculated separately for each
day's background and test data, as shown in Figures 16. Background data from all three days of
testing are shown together; leak simulation data for each day are plotted separately. The
intercept, which represents the isotopic signature of the CC>2 source, and 95% confidence interval
for each day of testing are summarized in Table 14. While the intercepts for leak simulation test
data did not fall within the confidence interval for the same day's background data, suggesting
that a different CC>2 source was impacting the site during leak simulation testing, the absolute
value of the intercepts were closer to that for the background data (-22.8%o) than for a pure
12,
CO2 source (-955.5%o).
46
-------
-6-
-8-
O
o
-10-
-12-
-14-1
A Background Data
Fit to Background Data
o July 28 Leak Simulation
- - - July 28 Fit
July 29 Leak Simulation
— - July 29 Fit
n July 30 Leak Simulation
July 30 Fit
I I I I I
0.0018 0.0020 0.0022 0.0024 0.0026
1/CO, (1/ppm)
Figure 16. Model G1101-/Minimum Detectable Leak Rate Results for Day 3 (7/30/2010)
Table 14. Minimum Detectable Leak Rate Keeling Plot Results
Test Date
July 28, 2010
July 29, 2010
July 3 0,2010
Background
Intercept
-23.35
-22.99
-23.22
95%
CI R2
0.59 0.2330
0.08 0.8432
0.03 0.9858
Leak Test
95%
Intercept CI
-18.67 0.41
-21.99 0.12
-13.45 0.21
R2
0.0594
0.6427
0.1184
Distinct
Mixing
Curve
Yes
Yes
Yes
The equivalent leak rate as a function of source 513C was also back-calculated for several
relevant 513C values: -3.5%o, -20%o, and -35%o. The results of those calculations are
summarized in Table 15. The predicted CC>2 concentration for the various CC>2 sources begin to
show increases over the ambient concentration for the leaks of 15 LPM or greater.
47
-------
Table 15. Extrapolated Minimum Detectable Leak Rate Results
COi Source
None
99.95% 12CO2
Fossil fuels
Fossil fuels
Heavy
Leak
Source 513C
(%o)
-6.41
-955.5
-35
-20
-3.5
Total Flow
Rate
(LPM)
1.28 x 106
1.28 x 106
1.28 x 106
1.28 x 106
1.28 x 106
Predicted CO2
Leak Flow Concentration Predicted
Rate (LPM)
0
0.423
14.62
31.84
19818
(ppm)
408.6
409.0
421.3
436.3
580.7
513C (%o)
-6.4
-7.3
-7.3
-7.3
-5.6
(a) Data in this table are reported to the number of significant digits appropriate for the respective analyzer's
reported accuracy and precision.
6.9 Ambient Air Monitoring
The Model Gl 101-/ was set up to monitor ambient air when not engaged in other testing
activities during Phase 2 and Phase 3. For Phase 3, a meteorological station was positioned near
the injection well, as shown in Figure 4; meteorological data are reported with the Model Gl 101-
/' ambient air CC>2 concentration and 513C readings in Figure 17. The ambient data set collected
by the Model Gl 101-/ is shown in the bottom panel for CC>2 concentration and 513C. The upper
panels show ambient temperature, RH, wind speed, and wind direction. The average ambient
CC>2 concentration was 411 ppm, with a range of 365 to 488 ppm. The average measured stable
isotope ratio was -6.4%o and values ranged from -9.5 to -4.3%o. The GS site experienced hot
and humid conditions with light to moderate winds except during thunderstorms, which passed
through the area on August 4 and August 5, 2010. The relationship between CC>2 concentration
and 513C was investigated by producing a Keeling Plot, as shown in Figure 18. The value of the
intercept, which represents the 513C of the CO2 source, is -23.15%o, is consistent with the value
for captured CC>2 at this site.
48
-------
360-L
12:00 AM
8/3/2010
12:00 AM
8/4/2010
12:00 AM
8/5/2010
T
12:00 AM
8/6/2010
Local Time
Figure 17. Meteorological Conditions and Model G1101-/ Ambient Air COi Measurements
at the GS Site
49
-------
-5-
-6-
O
o
-8-
-9-
O
813C02(%o)
Linear Regression
y = 6843 (±4)x -23.14 (+0.01)
R2 = 0.926
I I I I I
0.0021 0.0022 0.0023 0.0024 0.0025
1/C02 (1/ppm)
Figure 18. Keeling Plot of Model G1101-/ Ambient Air Data
I
0.0026
I
0.0027
During the ambient air monitoring period, a valve near the main sequestration well was opened
for approximately one minute. Prior to the intentional release of captured CC>2, the inlet to the
Model Gl 101-/' was positioned 0.86 m downwind of the valve. Figure 19 shows the raw data
and 30-second averages for CC>2 concentration and isotope ratio data reported by the Model
Gl 101-/' during the release. The period during which the valve was open is shaded in gray and
the points at which the leak was detected in the Model Gl 101-/' data are marked by open symbols
(O,D,O). The leak response time was determined relative to CC>2 concentration and isotope
ratio, as described in Section 5.8. A change in CO2 concentration was qualitatively identified 29
seconds after the valve was opened; the CC>2 concentration reached 1000 ppm 42 seconds after
the valve was opened. A decrease in measured isotope ratio at least 2 standard deviations from
the average value was observed after 39 seconds. Using all three methods, the leak was detected
by the Model Gl 101-/' less than 60 seconds after the leak was initiated.
A Keeling plot was generated from the data shown in Figure 19 (using 30-second averages) and
a linear regression calculated for the data. Data above the regression line are from the beginning
of the leak (beginning in the shaded area of the lower panel); data fell below the regression line
as values returned to background levels. The R2 value was 0.939 and the intercept was -24.0 (±
0.2)%o, which is similar to the intercept found for ambient measurement data (Figure 17) and is
consistent with the isotope ratio of the CO2 being injected at this site.
50
-------
0
-5-
-10-
-15-
-20
2000
1800-
0.0010
I
1/C02(1/ppm)
0.0015
I
0.0020
0.0025
A 5 CO2 (30 sec averages)
Linear Regression
Valve Open
O 1000 ppm alarm
n Visible concentration change
o 2s isotope ratio change
!! : S li 5
CO2 Concentration (raw)
CO2 Concentration (30 sec averages)
813C02 (raw)
5 CO2 (30 sec averages)
- 0
--5
5:10 PM
8/5/2010
!
5:12 PM
I
5:14 PM
[
5:16 PM
I
5:18 PM
--15
--20
5:20 PM
Local Time
Figure 19. Leak Response Time Results
co
O
o
10
01
3
Q>
-------
6.10 Mobile Surveys
Twice during Phase 3 the Model Gl 101-/' was installed in the Nissan Altima hybrid to conduct
mobile surveys as described in Chapter 3. Results from the mobile surveys have been analyzed
in a qualitative manner to evaluate whether the Model Gl 101-/' could reasonably be operated in
mobile survey mode. Once the first installation in the vehicle had been completed, it generally
took approximately 15 minutes with two testing staff to shut down the analyzer, move all
components into the vehicle, and begin collecting data under battery power. Some additional
time was then needed for the analyzer response to stabilize. At least one hour of monitoring data
could be collected on a single battery. A summary of the features, their brief descriptions, and
other identifying features is provided in Table 16.
Table 16. Mobile Survey Features and Identifiers
Survey
Date
August 3,
2010
August 6,
2010
Feature
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Identifier Description
Shallow aquifer monitoring well
Shallow aquifer monitoring well
A Approach by additional vehicle
B No feature
Deep monitoring well
Shallow aquifer monitoring well
Soil gas monitoring well
Soil gas monitoring well
Shallow aquifer monitoring well
Shallow aquifer monitoring well
Shallow aquifer monitoring well
Shallow aquifer monitoring well
M Meteorological Station
Above-ground transmission lines
Injection well
Injection well
Soil gas monitoring well
Soil gas monitoring well
Deep monitoring well
Environment
Near evergreen trees
Near evergreen trees
Evergreen trees
Gravel lot
Gravel lot
Gravel lot
Gravel lot
Lagoon berm
Roadside
Roadside
Roadside
Gravel lot
Gravel lot
Gravel lot
Edge of gravel lot
Edge of gravel lot
Figure 20 shows time series data for the Model Gl 101-/' concentration and 513C response during
the mobile survey conducted on August 3, 2010. The vertical black lines labeled 1 through 10
show the various features that were surveyed. The line labeled "A" shows the point where
another vehicle approached the hybrid sedan during the survey and shows that exhaust from the
other vehicle was sampled by the Model Gl 101-/'. Water vapor concentration data measured by
the Model Gl 101-/' are included in the top panel (red trace) to assist in identification of vehicle
exhaust, which is expected to have increased water vapor. Features 1 and 2 were surrounded by
52
-------
evergreen trees; at point "B," the vehicle approached trees without a well feature as a
comparison. The vehicle slowed dramatically or came to a complete stop at each of the features
and some were sampled twice. Figure 21 shows the GPS for the surveys of Features 1 and 2 that
have been colored according to the Model Gl 101-/' 513C value at each point. The impact of the
second vehicle is apparent around point A in Figure 21. Other spikes in the CC>2 concentration
and isotope ratio, for example between Features 3 and 4, appeared to be caused by the Nissan
Altima emissions when in non-electric vehicle mode.
Data from the second set of mobile surveys are shown in Figure 22 and Figure 23. Each feature
was passed twice. During this survey, the vehicle was driven out to the plant to charge the
battery before and during the survey. The location of the meteorological station is marked as
"M" in Figure 22. On the second approach to Feature 16, the hybrid sedan exited electric vehicle
mode; it is likely that emissions from the vehicle were being sampled at that time. Water vapor
measurements during this survey showed very little variability and were not helpful in
identifying vehicle exhaust.
53
-------
5:55 PM
8/3/2010
Local Time
Figure 20. Mobile Survey Results
O 0000 00 000000
»000000 Augu«3.20iO
Figure 21. Mobile Survey GPS Trace of Selected Features
54
-------
£400-,
Q.
Q.
395-
g385
§380
0~375
°370
-5-
(O
(0
<5
§
-7-
O -8-
O
-9-
11 12 13 14, 15
A/
16 11 12
9:48 AM
8/6/2010
i
9:52 AM
1 I
9:56 AM
Local Time
i i
10:00 AM
Figure 22. Mobile Survey Results
6mph
1EL
14 P
^ o
o o
Legend:
August 6. 2010
• <-7.0
O -70 -60
O -6.0- -5.5
• -5 5- -5.0
• >-50
o o
Figure 23. Mobile Survey GPS Trace of Features Shown in Figure 22
55
-------
6.11 Reference Method Comparability
The results of 10 pairs of duplicate ambient air reference method measurements were compared
to the average Model Gl 101-/ response over the five minutes preceding the sample
pressurization period to determine the reference method comparability. Two sample pairs were
collected during Phase 2 in the ABT with the blowers running and 8 pairs were collected during
Phase 3 while the Model Gl 101-/' was monitoring ambient air near the injection well. The inlet
for the ambient air sampling system was collocated with the Model Gl 101-/' inlet and the
sampler positioned downwind of the inlet while collecting samples. Care was taken to follow the
sampling and handling instructions provided by the reference laboratory. The reference method
measurements were compared to the Model Gl 101-/ response by calculating the accuracy and
bias for both CO2 concentration and 513C and the results summarized in Table 17. The average
accuracy for CO2 concentration was 97.9% with a range of 93.8% to 100.5%. For 613C, the
average difference was -3.0%o and values ranged from -3.5%o to -2. l%o. Bias was calculated
separately for each site. At the ABT and GS site, concentration bias was -0.2% and -2.5%,
respectively.
56
-------
Table 17. Reference Method Comparability Results
Date
(2010)
1 7/29
2 7/29
3 7/30
4 7/30
5 8/3
6 8/3
7 8/4
8 8/4
9 8/4
10 8/4
11 8/4
12 8/4
13 8/5
14 8/5
15 8/5
16 8/5
17 8/5
18 8/5
19 8/6
20 8/6
Average
Minimum
Maximum
Bias (%D)
Site Time
16:55
ABT 16:55
10:11
10:11
10:54
10:54
10:34
10:34
12:34
12:34
13:36
^c 13:36
GS
10:05
10:05
16:03
16:03
16:44
16:44
8:32
8:32
ABT
GS
CO2
Reference
Method
(ppm)
372.24(a)
372.22
391.81
391.51
402.72
402.51
395.10
395.04
394.52
394.66
395.55
395.16
413.87
414.04
401.69
401.84
395.45
395.09
415.71
417.66
Concentration
Model
G1101-/
(ppm)
368.7
368.7
393.6
393.6
397.0
397.0
388.2
388.2
389.0
389.0
391.2
391.2
408.8
408.8
376.7
376.7
371.1
371.1
412.6
412.6
%R
99.0
99.0
100.4
100.5
98.6
98.6
98.3
98.3
98.6
98.6
98.9
99.0
98.8
98.7
93.8
93.8
93.8
93.9
99.2
98.8
97.9%
100.5%
93.8%
-0.2%
-2.5%
CO2 513C (%o)
Reference
Method
-7.653
-7.597
-8.133
-8.139
-8.625
-8.636
-8.412
-8.447
-8.462
-8.439
-8.543
-8.471
-9.272
-9.324
-8.715
-8.715
-8.438
-8.472
-9.411
-9.455
Model
G1101-/
-4.8
-4.8
-6.1
-6.1
-5.8
-5.8
-5.4
-5.4
-5.1
-5.1
-5.0
-5.0
-6.1
-6.1
-5.8
-5.8
-4.9
-4.9
-6.7
-6.7
Difference
-2.9
-2.8
-2.1
-2.1
-2.9
-2.9
-3.0
-3.0
-3.4
-3.4
-3.5
-3.5
-3.2
-3.2
-2.9
-2.9
-3.5
-3.5
-2.7
-2.8
-3.0
-2.1
-3.5
accuracy and precision.
igits appropriate for the analyzer's reported
6.12 Data Completeness
The Model Gl 101-/' operated for 100% of the available time during Phase 1 and Phase 2 of the
verification test. During Phase 3, internal calibrations took place during 7% of the available
testing time (6 hours). The calibrations were initiated after the analyzer had been powered by a
deep cycle marine battery that was not sufficiently charged to power the analyzer. Due to the
behavior of the inverter, power to the Model Gl 101-/' cycled on and off every few minutes for
approximately 30 minutes. Inspection of the log files by the vendor revealed that an internal
error occurred during that period and this triggered the internal calibrations at the next startup.
Upon completion of the internal calibrations, the analyzer resumed measurements. When
supplied with the necessary power (i.e., a fully charged battery), the Model Gl 101-/' data return
during mobile survey testing was 100%. During a period of 2.6 hours, the temperature in the
57
-------
shed at the GS site exceeded the operating limits identified by the vendor and testing staff elected
to power off the analyzer until an air conditioner could be installed in the shed. The internal
calibrations and temperature-related downtime resulted in a 91% data return during Phase 3 of
this verification test.
6.13 Operational Factors
The Model Gl 101-/' was installed in the laboratory and at both field sites by Battelle testing staff;
the installation was completed in less than one hour. The vendor was readily available to answer
questions and provide support, but no formal training was provided on the instrument.
Instructions in the user manual for the installation were clear and easy to follow. A checklist was
provided by the vendor representative to establish whether the analyzer was in proper working
order during the test. The checklist, shown in Appendix A, was completed by Battelle staff
during the daily checks of the Model Gl 101-/' operating status. No maintenance was performed
on the analyzer. Data were downloaded on a daily basis to a USB memory stick or expansion
drive. The Model Gl 101-/' did not generate any waste or use consumable supplies. Batteries
used to operate the Model 1101-/' during Mobile Surveys were reusable and rechargeable.
In general, the Model Gl 101-/ software was easy to use. Battelle staff found the zoom and other
features on the graphical display to be somewhat cumbersome and not especially intuitive. Ease
of use of the software improved with practice.
58
-------
Chapter 7
Performance Summary
The performance of the Picarro Model Gl 101-/' was evaluated for its accuracy, bias, precision,
linearity, response time, and temperature/RH bias by evaluating the Model Gl 101-/' response
while sampling dilutions of known concentration from certified CC>2 gas standards and ambient
air. When possible, performance parameters were calculated for the Model Gl 101-/ CO2
concentration and 51 C response. The ability of the Model Gl 101-/' to detect CC>2 leaks was
evaluated under the simulated field conditions of the ABT by releasing varying amounts of pure
12CC>2 into ambient air stream. Use of the Model Gl 101-/' to conduct mobile surveys was
evaluated by operating the analyzer in a hybrid vehicle and transporting it to road-accessible
features at the GS site, such as CC>2 transmission lines and monitoring wells. The Model
Gl 101-/' was operated with the factory calibration. All gas standard dilutions were prepared
using the same calibrated dynamic dilution system. Given the uncertainty estimate for the
nominal CC>2 concentrations of ±7%, it is not possible to determine from these measurements
alone whether the observed inaccuracies and biases are due to errors in the instrument response
or the gas preparation. The results of this evaluation are described below.
Concentration Accuracy, Bias, Precision, and Response Time
The accuracy of the Model Gl 101-/' was assessed over the range of 100 ppm to 5,000 ppm in
terms of %R, which ranged from 90.0 to 113.1%, with an average of 96.0%. Bias, or the average
percent difference between the Model Gl 101-/ response and the known value, was -3.98%.
Precision of the Model Gl 101-/ was determined from the average responses to triplicate
challenges at each of 11 CO2 concentrations. The relative standard deviation values ranged from
0.1% to 1.2%, with an average of 0.3%. The average 95% response time was 2.43 minutes (142
seconds) for rise time and 2.53 minutes (152 seconds) for fall time. It is not possible to
determine from these measurements alone whether the observed response time is limited by the
response of the gas dilution system.
Concentration Linearity
Linearity was evaluated in terms of slope, intercept, and R . Over the 0 to 400 ppm range, the
slope of the regression line was 0.935 (±0.036), with an intercept of 11.3 (±8.9) and R2 value of
0.9958. Over 0 to 5,000 ppm, the slope of the regression line was 0.938 (±0.006), with an
intercept of-1.32 (±13.6) and R2 value of 0.9997. (The 95% confidence interval for the slope
and the intercept of each line is shown in parenthesis).
59
-------
Isotope Ratio Accuracy, Bias, and Linearity
The accuracy of the Model Gl 101-/' 513C response was assessed at -3.60%o, -10.41%o, and
-40.80%o at three concentration levels: 259 ppm, 370 ppm, and 740 ppm. Values for 513C
differed from the expected value by between 1.1 to 2.7%o, with an average of 1.7%o. The lowest
absolute differences were observed for the -40.80%o standard and at the higher CO2
concentrations. Isotope ratio linearity was assessed in terms of slope, intercept, and R2 for the
full dataset. The slope, including all isotope ratios and concentrations, was 1.01 (±0.02) with an
intercept of 1.88 (±0.37) and an R value of 0.9992. (The 95% confidence interval for the slope
and the intercept of each line is shown in parenthesis.) The strongest correlation between
concentration and measured isotope ratio, which was evaluated at -3.60%o, -10.41%o, and
-40.80%o, was observed for the -3.60%o standard, with an R2 value of 0.9468, a slope of
-0.0025 and intercept of 3.24.
Temperature and Relative Humidity Bias
Temperature and RH bias were assessed by comparing the Model Gl 101-/' CO2 concentration
and o C response to dilutions from a certified CC>2 standard at 5 temperature/RH conditions to
its response at 20°C and 0% RH. During this evaluation, the Model Gl 101-/' was installed in a
temperature/RH-controlled chamber; humidified zero air was added to the CC>2 gas standard
dilution to achieve the desired RH. The following test conditions were evaluated: 20°C/50%
RH; 20°C/90% RH; 32°C/50% RH; 32°C/90% RH; 4°C/50% RH. In general, variability in
ambient temperature and RH conditions resulted in bias values of 3% or less for the Model
Gl 101-/' concentration measurements and 0.7%o or less for isotope ratio. The maximum
concentration bias value, 3.0%, was observed for CC>2 concentration at 4°C/50% RH. The
largest isotope ratio average difference of 0.7%o was observed for 32°C/90% RH.
Minimum Detectable Leak Rate
The ability of the Model Gl 101-/' to identify CO2 leaks above ambient air variability was
evaluated by simulating leaks under controlled field conditions in the Ambient Breeze Tunnel.
Pure 12CC>2 was periodically released into a constant flow of ambient air and the flow rate
adjusted until the Model Gl 101-/' difference in the 513C response during the "leak" compared to
ambient air was greater than 2 times the ambient air 513C variability. Under conditions that
simulated 1.8 m/s winds, the minimum detectable leak rate was 0.423 LPM 12CC>2, which
resulted in a 0.9%o decrease, on average, in the Model Gl 101-/ 513C readings compared to
ambient air (approximately -6.4%o). This result was extrapolated to determine the equivalent
leak rates for CC>2 sources of with 513C values -35%o, -20%o, and -3.5%o. The equivalent leak
rates were 14.62 LPM, 31.84 LPM, and 198.18 LPM, respectively.
Ambient Air Monitoring
The Model Gl 101-/ monitored ambient air at the GS site between August 2 and August 6, 2010.
During this period, the average ambient CC>2 concentration was 411 ppm, with a range of 365 to
488 ppm. The average measured stable isotope ratio was -6.42%o and values ranged from -9.50
to -4.28%o. The relationship between CC>2 concentration and 513C was investigated by
producing a Keeling Plot. The value of the intercept, which represents the 513C of the CO2
source, is -23. l%o and is consistent with the value for captured CO2 at this site. An intentional
release of captured CC>2 was detected by the Model Gl 101-/' in less than 60 seconds. A Keeling
Plot of the data from the intentional release period had an R2 value of 0.939 and an intercept of
60
-------
-24.0 (± 0.2)%o, which is similar to the intercept found for ambient measurement data and is
consistent with the isotope ratio of the CO2 being injected at this site.
Mobile Surveys
During Phase 3, the Model Gl 101-/ was transported to road-accessible features of the GS, such
as transmission lines and monitoring wells, to evaluate the ease of use and operational factors of
the analyzers during use in a mobile survey mode. The Model Gl 101-/' surveyed 16 features at
the GS while installed in the back seat of a Nissan Altima hybrid sedan and operating on power
from a marine deep cycle/RV battery and power inverter. Once the first installation in the
vehicle had been completed, it generally took approximately 15 minutes with two testing staff to
shut down the analyzer, move all the components into the vehicle, and begin collecting data
under battery power. Some additional time was then needed for the analyzer response to
stabilize. At least one hour of monitoring data could be collected on a single battery.
Comparability to Reference Methods
Comparability was determined as the accuracy (%R) and bias (average percent difference) of the
Model Gl 101-/' response compared to CC>2 concentration (NDIR) and 5 3C (IRMS) reference
method results for 10 duplicate grab samples of ambient air. The average accuracy for CO2
concentration was 97.9% with a range of 93.8% to 100.5%. For 513C, the average difference was
-3.0%o and values ranged from -3.5%o to -2.1%o. Bias was calculated separately for each site.
At the ABT and GS site, concentration bias was -0.2% and -2.5%, respectively.
Data Completeness
The Model Gl 101-/' operated for 100% of the available time during Phase 1 and Phase 2 of the
verification test. During Phase 3, internal calibrations took place during 7% of the available
testing time (6 hours) and the analyzer was shut down for 2.6 hours because ambient
temperatures in the shed where the analyzer was operated exceeded operating limits identified by
the vendor. The internal calibrations and temperature-related downtime resulted in a 91% data
return during Phase 3 of this verification test.
Operational Factors
The Model Gl 101-/' was installed in the laboratory and at both field sites by Battelle testing staff;
the installation was completed in less than one hour; no formal training by the vendor was
necessary. Instructions in the user manual for the installation were clear and easy to follow. A
checklist was provided by the vendor representative to establish whether the analyzer was in
proper working order during the test. No maintenance was performed on the analyzer. Data
were downloaded on a daily basis to a USB memory stick or expansion drive. The Model
Gl 101-/' did not generate any waste or use consumable supplies. In general, the Model Gl 101-/'
software was easy to use. Battelle staff found the zoom and other features on the graphical
display to be somewhat cumbersome and not especially intuitive. Ease of use of the software
improved with practice.
Vendor-Supplied Specifications
The Model Gl 101-7 weighs 26.3 kg (58 Ibs), has dimensions of 43 x 25 x 59 cm (17" x 9.75" x
23") including the feet, and can be rack mounted or operated on a bench top. The approximate
purchase price of the Model Gl 101-i is US $60,500.
61
-------
Chapter 8
References
1. Busch KW, Busch MA, Cavity Ring-down Spectroscopy: An Ultratrace Absorption
Measurement Technique. ACS Symposium Series 720, Oxford (1997).
2. Atkinson, D. B., "Solving chemical problems of environmental importance using cavity ring-
down spectroscopy," The Analyst 128, 117-125 (2003).
3. Test/QA Plan for Test/QA Plan for Verification of Isotopic Carbon Dioxide Analyzers for
Carbon Sequestration Monitoring, Battelle, Columbus, Ohio, July 1, 2010.
4. Quality Management Plan for the ETV Advanced Monitoring Systems Center, Version 7.0,
U.S. EPA Environmental Technology Verification Program, Battelle, Columbus, Ohio,
November, 2008
5. Trolier, M., J. W. C. White, P. P. Tans, K. A. Masarie, and P. A. Gemery, Monitoring the
isotopic composition of atmospheric CO2: Measurements from the NOAA Global Air
Sampling Network, J. Geophys. Res., 707(D20), 25, 897-25, 916. (1996)
6. Andres, R. J., G. Marland, T. Boden, and S. Bischof, Carbon dioxide emissions from fossil
fuel consumption and cement manufacture, 1751-1991, and an estimate of their isotopic
composition and latitudinal distribution, in 1993. Global Change Institute, edited by T.
Wigley and D. Schimel, Cambridge Univ. Press, New York, 1996.
7. Andres, R. J., G. Marland, I. Fung, and E. Matthews (1996), A 1° x 1° distribution of carbon
dioxide emissions from fossil fuel consumption and cement manufacture, 1950-1990, Global
Biogeochem. Cycles, 10, 419-429.[AGU1 http://cdiac.ornl.gov/ftp/db 1013/db 1013 .txt
8. Esler, M.B., Griffith, D.W.T., Wilson, S.R., and Steele, L.P. Precision Trace Gas Analysis by
FT-IRSp<
216-221.
FT-IR Spectroscopy. 2. The 13C/12C Isotope Ratio of CO2. Anal. Chem., 2000, 72 (1), pp
9. Becker, J. F.; Sauke, T. B.; Loewnstein, M. Stable Isotope Ratio Measurements Using
Tunable Diode Laser Spectroscopy. Opt. 1992, 31, 1921-1927.
62
-------
10. Conway, T., P. Tans, L. Waterman, K. Thoning, D. Kitzis, K. Masarie, andN. Zhang (1994),
Evidence for interannual variability of the carbon cycle from the National Oceanic and
Atmospheric Administration/Climate Monitoring and Diagnostics Laboratory Global Air
Sampling Network, J. Geophys. Res., 99(D11), 22831-22855.
11. Komhyr, W., L. Waterman, and W. Taylor (1983), Semiautomatic Nondispersive Infrared
Analyzer Apparatus for CC>2 Air Sample Analyses, J. Geophys. Res., 88(C2), 1315-1322.
12. Komhyr, W. D.; Harris, T. B.; Waterman, L. S. (1985). Calibration of Nondispersive
Infrared CC>2 Analyzers with CO2-in-Air Reference Gases. J. Atmos. & Oceanic Tech. 2, 82-
13. Vaughn, B., D. Ferretti, J. Miller, and J. White (2004), Stable isotope measurements of
atmospheric CO2 and CH/i, in Handbook of Stable Isotope Analytical Techniques, vol 1,
chapter. 14, Elsevier, New York.
14. Masarie, K., C. Allison, T. Conway, E. Dlugokencky, R. Francey, R. Langenfeld, P. Novelli,
L. Steele, P. Tans, B. Vaughn, and J. White (2001), The NOAA/CSIRO flask-air
intercomparison program: a strategy for directly assessing consistency among measurements
derived from independent laboratories, J. Geophys. Res., 106 (D17), 20,445-20,464.
15. Keeling, et al., (1976) Atmospheric carbon dioxide measurements at Mauna Loa
Observatory, Hawaii, Tellus 28.
16. Griffith (1982) Calculations of carrier gas effects in nondispersive infrared analyzers I.
Theory, Tellus, 34.
17. Griffith, et al., (1982) Calculations of carrier gas effects in nondispersive infrared analyzers
II. Comparison with experiment, Tellus, 34.
18. Environics Series 6100 Computerized Multi-Gas Calibration System Specifications.
(http://www.environics.com/Product/ambient-monitor-calibrator-with-ozone-
generator/series-61OO/). Accessed March 5, 2011.
63
-------
Appendix A
Daily Checklist
64
-------
DAILY CHECKLIST
Picarro G1101-/Analyzerfor Isotopic CO2
ETV Verification Test of Isotopic Carbon Dioxide Analyzers
for Carbon Sequestration Monitoring
Instrument Performance:
D Check the 'status' window (scroll up to see all messages) and record any messages:
o Contact vendor for any questions about warnings. Vendor contact:
Chris Rella 408.962.3941
Picarro, Inc. rella@picarro.com
D Put the GUI into 'service mode'. Service mode is identified by an 's' in the LED in the upper right.
To select service mode, go to Settings/Change GUI mode, and then type the password 'picarro'
(no quotes) and click ok.
o Verify that the following sensors are within normal operating parameters, by selecting the
appropriate item from the dropdown list and recording the min / max data from the graph.
Actual Value
Parameter
CAVITY_TEMP
WB_TEMP
CAVITY_PRESSURE
DAS_TEMP
Normal Range
45°C ± 0.02°C
45°C ± 0.02°C
140Torr±0.2Torr
1 0 - 40 °C (this is an
indication of the ambient
temperature)
Within Range? (Y/N)
These same activities can be performed by analyzing the xxx_data.dat files, found in the
C:\Userdata\YYYYMMDD directory structure (see below). These values (along with many more diagnostic
values) can be found in a simple space-delimited ASCII data file.
D Record instrument flow rate with a calibrated flow meter. (Value should be ~20 seem)
Flow meter SN: Flow rate:
Daily Data backup:
D Standard Data files: User data is available in C:\UserData\YYYYMMDD. Download (copy) all
.dat files from the daily directories into a flash drive or via an ethernet connection.
D Spectral data: All raw spectral data are located in C:\Picarro\Archive\YY\MM\DD\HH directories
(these dates in times are in universal time, not in instrument local time). Download (copy) all the
.zip files collected in these directory structures every day. IMPORTANT: the instrument saves
only about 3-4 days of raw spectral data in a first-in first-out buffer in the archive. Be sure to
download these files daily, or the information will be lost.
Operator Name:
Operator Signature:,
Comments:
Date:
65
------- |