August 2001
Environmental Technology
Verification Report
Nippon Instruments Corporation
Model MS-1/DM-5 Mercury
Continuous Emission Monitor
Prepared by
Baltelle
Putting Technology To Work
Battel le
Under a cooperative agreement with
£EPA U.S. Environmental Protection Agency
ETV EjV ElV
-------
I lll ENVIRONMENTAL TECHNOLOGY VERIFICATION
&EPA
U.S. Environmental Protection Agency
PROGRAM J
ETV
Battelle
.. . Putting Technology To Work
ETV Joint Verification Statement
TECHNOLOGY TYPE: Continuous Emission Monitor
APPLICATION:
MEASURING ELEMENTAL AND OXIDIZED
MERCURY EMISSIONS
TECHNOLOGY
NAME:
Model MS-l/DM-5
COMPANY:
Nippon Instruments Corporation
ADDRESS:
WEB SITE:
E-MAIL:
14-8 Akaoji-cho, Takatuki-shi PHONE: +81-726-94-5195
Osaka 569-1146 Japan FAX: +81-726-94-0663
http://www.smglink.com/nic
nic@rigaku.co.j p
The U.S. Environmental Protection Agency (EPA) has created the Environmental Technology Verification (ETV)
Program to facilitate the deployment of innovative or improved environmental technologies through performance
verification and dissemination of information. The goal of the ETV Program is to further environmental protec-
tion by substantially 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 standards and testing organizations; with stakeholder groups that
consist 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 protocols to ensure that data of known and adequate quality are generated and that the results
are defensible.
The Advanced Monitoring Systems (AMS) Center, one of six technology centers under ETV, is operated by
Battelle in cooperation with EPA's National Exposure Research Laboratory. The AMS Center has recently
evaluated the performance of continuous emission monitors used to measure mercury in flue gases. This
verification statement provides a summary of the test results for the Nippon Instruments Corporation Model
MS-l/DM-5 mercury continuous emission monitor (CEM).
-------
VERIFICATION TEST DESCRIPTION
The verification test was conducted over a three-week period in January 2001 at the Rotary Kiln Incinerator
Simulator (RKIS) facility at EPA's Environmental Research Center, in Research Triangle Park, North Carolina.
This mercury CEM verification test was conducted jointly by Battelle's AMS Center, EPA's Office of Research
and Development, and the Massachusetts Department of Environmental Protection. A week of setup and trial runs
was followed by two weeks of verification testing under different flue gas conditions. The daily test activities
provided data for verification of the following performance parameters of the MS-l/DM-5 relative accuracy in
comparison to reference method results, correlation with the reference method, precision in sampling at stable
flue gas conditions, calibration/zero drift from day to day, sampling system bias in transfer of mercury to the
CEM's analyzer, interference effects of flue gas constituents on CEM response, response time to rising and
falling mercury levels, response to low levels of mercury, data completeness over the course of the test, and setup
and maintenance needs of the CEM. The Ontario Hydro (OH) draft American Society for Testing and Materials
mercury speciation method was used as the reference method in this verification test. Paired OH trains were
sampled at each of two different locations in the RKIS duct to establish the precision of the OH method.
Quality assurance (QA) oversight of verification testing was provided by Battelle and EPA. Battelle QA staff
conducted a data quality audit of 10% of the test data, a series of performance evaluation audits on several
measurements at the RKIS, and both an internal and an external technical systems audit of the procedures used in
this verification. EPA QA staff also conducted an independent technical systems audit at the RKIS.
TECHNOLOGY DESCRIPTION
The MS-l/DM-5 is a continuous monitor for elemental and oxidized vapor-phase mercury. The monitor consists
of one Nippon MS-1 mercury speciation unit and two Nippon DM-5 mercury detectors. The DM-5 units each
detect elemental mercury continuously by cold vapor atomic absorption without a preconcentration step. The
MS-1 unit separates elemental and oxidized mercury by means of a wet scrubbing and chemical reaction system
fed by a peristaltic pump. In the MS-1, filtered flue gas is first contacted with deionized water, which collects all
oxidized mercury while leaving elemental mercury in the vapor phase. The gas containing the elemental mercury
is passed directly to one of the DM-5 units for detection. The collected oxidized mercury is reduced to elemental
mercury in a continuous liquid flow system, using a proprietary reagent. The elemental mercury produced is then
swept from solution into a clean air stream and sent to the second DM-5 monitor. The two DM-5 monitors thus
provide separate and continuous measurements of elemental and oxidized mercury in a parallel two-channel mode
of operation. Oxidized mercury readings have a time lag of about one minute relative to the elemental mercury
readings, due to the delay in the liquid flow system. Each DM-5 unit has a digital display in (ig/m3 of mercury,
along with RS-232 output for recording data by a laptop computer or data logger. No external gas supplies are
required. The MS-1 is 43 cm wide x 23 cm deep x 59 cm high (17 in. W x 9 in. D x 23 in. H) and weighs 16 kg
(35 lbs.). Each DM-5 is 43 cm W x 22 cm D x 55 cm H (17 in. W x 8.7 in. D x 21.7 in. H), and weighs 25 kg
(55 lbs). These instruments operate on 100 to 110 V AC.
VERIFICATION OF PERFORMANCE
Relative accuracy: During the first week of verification testing, the MS-l/DM-5 accuracy relative to the OH
method was 13.2% for total mercury, at total mercury levels of about 7 to 8 (.ig/ni3. Testing showed relative
accuracy of 11.0% for elemental mercury, and 54.9% for oxidized mercury, at elemental mercury levels of
approximately 6 to 7 (ig/m3 and oxidized mercury levels of approximately 1 to 1.5 (ig/m3. In the second week, the
relative accuracy was 39.1% for total mercury, at total mercury levels of about 70 to 120 (.ig/m3. Relative
accuracy of 50.4% for elemental mercury, and 49.1% for oxidized mercury, was found at elemental mercury
levels of about 5 to 25 (ig/m3 and oxidized levels of about 45 to 110 (ig/m3.
-------
Correlation with the reference method: The coefficient of determination (r2) of the MS-l/DM-5 and OH ele-
mental mercury results was 0.417 based on data from both weeks combined. The corresponding r2 value for
oxidized mercury was 0.937, and for total mercury was 0.938.
Precision at stable flue gas conditions: The precision, as percent relative standard deviation (% RSD), of the
MS-l/DM-5 response for elemental mercury was within 10% in 11 of the 15 OH periods and within 15% in 14 of
the periods. For oxidized mercury, precision was never within 10% RSD, but nine of the 15 periods showed
precision within 15% RSD. For total mercury, precision was within 10% RSD in 10 of the 15 OH periods and
within 15% in 14 of the periods. These precision results include both variability in the test facility and in the
MS-l/DM-5.
Calibration/zero drift: Analysis of zero gas and elemental mercury standard gas results for the MS-l/DM-5
showed a 2.9% RSD for each DM-5 detector in repeated analysis of standard gas during the first week. During
the second week, the results showed a 0.5% RSD for one DM-5 detector and 0.6% RSD for the other.
Sampling system bias: The bias in transport of elemental mercury through the Nippon inlet system was
approximately -7%.
Interference effects of flue gas constituents: Elevated levels of sulphur dioxide, nitric oxide, and carbon
monoxide had no effect on MS-l/DM-5 response to elemental or oxidized mercury in flue gas. The presence of
hydrogen chloride reduced elemental mercury readings by about 25%, without a corresponding increase in the
oxidized mercury readings of the MS-l/DM-5. The presence of chlorine reduced elemental mercury readings by
about the same amount as did hydrogen chloride, but also caused a large increase in oxidized mercury readings.
When all these gases were present at once in the flue gas, the MS-l/DM-5 readings were close to those seen with
only mercury in the flue gas, indicating no substantial interference from the combination of these gases.
Response time to changing mercury levels: The rise and fall times of the MS-l/DM-5 were about 50 and about
35 seconds, respectively.
Response to low levels of mercury: The MS-l/DM-5 responded to as little as 0.57 (ig/m3 of mercury in flue gas,
but the response to concentrations of 0.57 to 4.5 (ig/m3 averaged only about 65% of the nominal total mercury
concentration.
Data completeness: Data completeness for the MS-l/DM-5 was 100%.
Setup and maintenance needs: No repair or maintenance of the MS-l/DM-5 was needed during the verification
test. The unit produces 2 to 3 L/day of aqueous waste solutions, in the form of the reagents used to separate
elemental and oxidized mercury.
Gabor J. Kovacs Date Gary J. Foley Date
Vice President Director
Environmental Sector National Exposure Research Laboratory
Battelle Office of Research and Development
U.S. Environmental Protection Agency
Gina McCarthy Date
Assistant Secretary for Pollution Prevention
Environmental Business and Technology
Massachusetts Executive Office of Environmental Affairs
-------
NOTICE: ETV verifications are based on an evaluation of technology performance under specific,
predetermined criteria and the appropriate quality assurance procedures. EPA and Battelle make no expressed or
implied warranties as to the performance of the technology and do not certify that a technology will always
operate as verified. The end user is solely responsible for complying with any and all applicable federal, state,
and local requirements. Mention of commercial product names does not imply endorsement.
-------
August 2001
Environmental Technology Verification
Report
ETV Advanced Monitoring Systems Center
Nippon Instruments Corporation
Model MS-1/DM-5
Mercury Continuous Emission Monitor
by
Thomas Kelly
Charles Lawrie
Karen Riggs
Battelle
Columbus, Ohio 43201
Jeffrey Ryan
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
with support from
Massachusetts Department of Environmental Protection
-------
Notice
The U.S. Environmental Protection Agency (EPA), through its Office of Research and
Development, has financially supported and collaborated in the extramural program described
here. This document has been peer reviewed by the Agency and recommended for public release.
Mention of trade names or commercial products does not constitute endorsement or
recommendation by the EPA for use.
ii
-------
Foreword
The U.S. 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 assess-
ment. In 1997, through a competitive cooperative agreement, Battelle was awarded EPA funding
and support 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/07/07_main.htm.
iii
-------
Acknowledgments
The authors wish to acknowledge the support of all those who helped plan and conduct the
verification test, analyze the data, and prepare this report. In particular we would like to thank
Bea Weaver, Adam Abbgy, and Paul Webb of Battelle. We thank those staff of EPA's
Environmental Research Center who provided assistance in the field tests. We also acknowledge
the support of Chris Winterrowd and other staff of ARCADIS Geraghty & Miller for their work
in operating the test facility, conducting Ontario Hydro sampling, and reporting data for this
report. The participation of Koji Tanida and Munehiro Hoshino of Nippon Instruments is
gratefully acknowledged.
iv
-------
Contents
Notice ii
Foreword iii
Acknowledgments iv
List of Abbreviations ix
1. Background 1
2. Technology Description 2
3. Test Design and Procedures 4
3.1 Introduction 4
3.2 Test Conditions 6
3.2.1 RKIS Conditions 6
3.2.2 MS-l/DM-5 Operation 8
3.2.3 Ontario Hydro Reference Method 11
3.3 Verification Procedures 12
3.3.1 Relative Accuracy 12
3.3.2 Correlation with Reference Method 13
3.3.3 Precision 13
3.3.4 Calibration/Zero Drift 14
3.3.5 Sampling System Bias 14
3.3.6 Interferences 14
3.3.7 Response Time 15
3.3.8 Low-Level Response 15
3.3.9 Data Completeness 16
3.3.10 Setup and Maintenance Needs 16
3.4 Equipment and Materials 16
3.4.1 Commercial Elemental Mercury Standards 16
3.4.2 Performance Evaluation Equipment 18
v
-------
4. Data Quality 20
4.1 Facility Calibrations 21
4.2 Ontario Hydro Sampling and Analysis 21
4.2.1 Ontario Hydro Precison 22
4.2.2 Ontario Hydro Blank Trains 27
4.2.3 Ontario Hydro Reagent Blanks 27
4.3 Mercury Mass Balance 28
4.4 Audits 29
4.4.1 Technical Systems Audit 29
4.4.2 Performance Evaluation Audits 29
4.4.3 Data Quality Audit 32
5. Statistical Methods 33
5.1 Relative Accuracy 33
5.2 Correlation with Reference Method 33
5.3 Precision 34
5.4 Calibration/Zero Drift 34
5.5 Sampling System Bias 34
5.6 Interferences 35
5.7 Response Time 35
5.8 Low-Level Response 35
6. Test Results 36
6.1 Relative Accuracy 36
6.2 Correlation 38
6.3 Precision 40
6.4 Calibration/Zero Drift 42
6.5 Sampling System Bias 42
6.6 Interference Effects 45
6.7 Response Time 47
6.8 Low-Level Response 49
6.9 Data Completeness 50
6.10 Setup and Maintenance 50
6.11 Cost 50
7. Performance Summary 51
8. References 53
vi
-------
Figures
Figure 2-1. Nippon Instruments Corporation Model MS-l/DM-5
Mercury Continuous Emission Monitor 2
Figure 3-1. Side View (top) and End View (bottom) of RKIS Test Facility 9
Figure 6-la. Correlation of MS-l/DM-5 and OH Results from Week One
of Verification Testing 39
Figure 6-lb. Correlation of MS-l/DM-5 and OH Results from Week Two
of Verification Testing 39
Figure 6-lc. Correlation of MS-l/DM-5 and OH Results from Both Weeks
of Verification Testing 40
Figure 6-2. Nippon MS-l/DM-5 Elemental and Oxidized Mercury Response in
Interference Test, January 18, 2001 46
Figure 6-3. Nippon MS-l/DM-5 Total Mercury Response in Interference Test,
January 18, 2001 46
Figure 6-4. Low-Level Mercury Response of the Nippon MS-l/DM-5,
January 19, 2001 49
Tables
Table 3-1. General Schedule of the ETV Mercury CEM Verification Test 5
Table 3-2. Schedule of Daily Activities in the Mercury CEM Verification Test 7
Table 3-3. Summary of RKIS CEMs 10
Table 3-4. Target Flue Gas Constituent Concentrations Used
in Verification Test 10
Table 3-5. Schedule of OH Sampling Runs During Mercury CEM Verification 13
Table 3-6. Schedule of January 18, 2001, Interference Test 15
Table 3-7. Schedule of January 19, 2001, Low-Level Response Test 16
Table 3-8. Mercury Standard Gas Identification and Analysis Results 17
vii
-------
Table 3-9. Performance Evaluation Audit Equipment Used for the Verification Test 19
Table 4-la. Elemental Mercury Results from OH Sampling in the First Week
Verification Testing 23
Table 4-lb. Oxidized Mercury Results from OH Sampling in the First Week
of Verification Testing 23
Table 4-lc. Total Mercury Results from OH Sampling in the First Week
of Verification Testing 24
Table 4-2a. Elemental Mercury Results from OH Sampling in the Second Week
of Verification Testing 24
Table 4-2b. Oxidized Mercury Results from OH Sampling in the Second Week
of Verification Testing 25
Table 4-2c. Total Mercury Results from OH Sampling in the Second Week
of Verification Testing 25
Table 4-3. Equivalent Flue Gas Mercury Concentrations («g/m3) Found in
Blank Ontario Hydro Trains 27
Table 4-4. Percent Recovery of Total Mercury Injected into the RKIS 28
Table 4-5. Summary of PE Audits on Mercury CEM Verification 30
Table 4-6. Results of PE Audit of OH Train Analysis 31
Table 6-1. Average Mercury Results from Nippon MS-l/DM-5
During OH Sampling Runs 37
Table 6-2. Relative Accuracy Results for the Nippon MS-l/DM-5 38
Table 6-3. Correlation of Nippon MS-l/DM-5 Data with OH Results 38
Table 6-4. Precision Results for the Nippon MS-l/DM-5 41
Table 6-5. Zero and Standard Gas Responses of the Nippon MS-l/DM-5 43
Table 6-6. Summary of Calibration/Zero Drift Results for the Nippon MS-l/DM-5 44
Table 6-7. Results of Sampling System Bias Test of the Nippon MS-l/DM-5 45
Table 6-8. Results of Response Time Tests on the Nippon MS-l/DM-5 48
viii
-------
List of Abbreviations
AMS
Advanced Monitoring Systems
ANOVA
analysis of variance
APCS
air pollution control system
CEM
continuous emission monitor
Cl2
chlorine
CO
carbon monoxide
co2
carbon dioxide
EPA
U.S. Environmental Protection Agency
ETV
Environmental Technology Verification
GFCIR
gas filter correlation infrared
h202
hydrogen peroxide
h2so4
sulfuric acid
HC1
hydrogen chloride
Hg
mercury
Hg°
elemental mercury
KC1
potassium chloride
KMn04
potassium permanganate
L/min
liters per minute
m3
cubic meters
MDEP
Massachusetts Department of Environmental Protection
mg/m3
milligrams per cubic meter
mL
milliliter
NDIR
non-dispersive infrared
NIST
National Institute of Standards and Technology
NO
nitric oxide
N0X
nitrogen oxides
o2
oxygen
OD
outside diameter
OH
Ontario Hydro
ORD
Office of Research and Development
PE
performance evaluation
PPb
parts per billion
ppm
parts per million
Psig
pounds per square inch gauge
QA
quality assurance
QC
quality control
QMP
Quality Management Plan
RA
relative accuracy
ix
-------
RKIS Rotary Kiln Incinerator Simulator
RSD relative standard deviation
S02 sulfur dioxide
TSA technical systems audit
x
-------
Chapter 1
Background
The U.S. Environmental Protection Agency (EPA) has created the Environmental Technology
Verification (ETV) Program to facilitate the deployment of innovative environmental tech-
nologies through performance verification and dissemination of information. The goal of the
ETV Program is to further environmental protection by substantially accelerating the acceptance
and use of improved and cost-effective technologies. ETV seeks to achieve this goal by provid-
ing high-quality, peer-reviewed data on technology performance to those involved in the design,
distribution, permitting, purchase, and use of environmental technologies.
ETV works in partnership with recognized standards and testing organizations; with stakeholder
groups that consist of buyers, vendor organizations, and permitters; and with the full participation
of individual technology developers. The program evaluates the performance of innovative tech-
nologies 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
protocols to ensure that data of known and adequate quality are generated and that the results are
defensible.
The EPA's National Exposure Research Laboratory and its verification organization partner,
Battelle, operate the Advanced Monitoring Systems (AMS) Center under ETV. The AMS Center
recently evaluated the performance of continuous emission monitors (CEMs) for mercury
emissions in combustion flue gas. This verification report presents the procedures and results of
the verification test for the Nippon Instruments Corporation MS-l/DM-5 mercury CEM.
1
-------
Chapter 2
Technology Description
The objective of the ETV AMS Center is to verify the performance characteristics of environ-
mental monitoring technologies for air, water, and soil. This verification report provides results
for the verification testing of the Nippon Instruments Corporation MS-l/DM-5 mercury CEM.
The following description of the MS-l/DM-5 is based on information provided by the vendor.
The MS-l/DM-5 mercury CEM is a continuous monitor for elemental and oxidized vapor-phase
mercury. As tested in this verification, the monitor consisted of one Nippon MS-1 mercury
speciation unit and two Nippon DM-5 mercury detectors. The DM-5 units each detect elemental
mercury continuously by cold vapor atomic absorption without a preconcentration step. The
MS-1 unit separates elemental and oxidized mercury by means of a wet scrubbing and chemical
reaction system fed by a peristaltic pump. In the MS-1, filtered flue gas is first contacted with
deionized water, which collects all oxidized
mercury while leaving elemental mercury in the
vapor phase. Since the MS-1 uses a continuous
flow scrubbing approach, collection of acid
gases and consequent oxidation of mercury are
minimized. The gas containing the elemental
mercury is passed directly to one of the DM-5
units for detection. The collected oxidized
mercury is reduced to elemental mercury in a
continuous liquid flow system, using a pro-
prietary reagent. The elemental mercury
produced is then swept from solution into a
clean air stream and sent to the second DM-5
detector. The two DM-5 detectors thus provide
separate and continuous measurements of
elemental and oxidized mercury in a parallel
two-channel mode of operation. Oxidized
mercury readings have a time lag of about one minute relative to the elemental mercury readings,
due to the delay in the liquid flow system.
As operated in this verification test, the DM-5 detectors had a detection limit of about 0.1 (ig/m3
for emission monitoring. Each DM-5 unit has a digital display in (ig/m3 of mercury, along with
RS-232 output for recording of data by a laptop computer or data logger.
Figure 2-1. Nippon Instruments Corporation
Model MS-l/DM-5 Mercury Continuous
Emission Monitor
2
-------
The MS-l/DM-5 mercury CEM requires a sample flow of about 0.5 standard liters per minute
and liquid reagent flows totaling about 2 mL/min. Thus, the total reagent volume consumed, and
the total waste solution produced, is 2 to 3 L/day. No external gas supplies are required. The
MS-1 is 43 cm wide x 23 cm deep x 59 cm high (17 in. W x 9 in. D x 23 in. H) and weighs 16 kg
(35 lbs.). Each DM-5 detector is 43 cm W x 22 cm D x 55 cm H (17 in. W x 8.7 in. D x 21.7 in.
H), and weighs 25 kg (55 lbs.). These instruments operate on 100 to 110 V AC.
Improvements in the MS-l/DM-5 in development at the time of this verification included
adopting a different scrubbing solution and combining the dual DM-5 detectors in a single
package.
3
-------
Chapter 3
Test Design and Procedures
3.1 Introduction
This verification test was conducted according to procedures specified in the Test/QA Plan for
Pilot-Scale Verification of Continuous Emission Monitors for Mercury.(1) The Nippon
Instruments MS-l/DM-5 CEM was verified for its measurements of elemental, oxidized, and
total mercury by comparison to reference method measurements, by challenges with interferant
species, and by repeated sampling of elemental mercury standard gases. The test activities
provided data for verification of the following performance parameters of the MS-l/DM-5
monitor:
¦ Relative accuracy in comparison to reference method results
¦ Correlation with the reference method
¦ Precision in sampling at stable flue gas conditions
¦ Calibration/zero drift from day to day
¦ Sampling system bias in transfer of mercury to the CEM's analyzer
¦ Interference effects of flue gas constituents on CEM response
¦ Response time to rising and falling mercury levels
¦ Response to low levels of mercury
¦ Data completeness over the course of the test
¦ Setup and maintenance needs of the CEM.
All but the last parameter listed were evaluated quantitatively, using data produced by the
planned sequence of tests. The data used to verify these parameters are specified later in this sec-
tion, and the statistical calculations used to quantify these parameters are presented in Section 5
of this report. The last parameter listed, setup and maintenance needs, was evaluated qualitatively
by observing the operation and maintenance of the CEM by vendor staff during the test.
The verification test was conducted over a three-week period in January 2001 at the Rotary Kiln
Incinerator Simulator (RKIS) facility at EPA's Environmental Research Center, in Research
Triangle Park, North Carolina. The RKIS is a gas-fired, two-stage, pilot-scale incinerator that
allows flue gas composition to be manipulated by injecting pollutant gases, particulate matter,
and mercury at various points within and downstream of the combustion zone. The RKIS flue
gases pass through an extended length of duct before entering an air pollution control system
(APCS). When mercury is introduced into the facility flue gas, the facility must operate as a
permitted hazardous waste facility, with limitations on hours of operation and personnel training.
4
-------
In particular, mercury injection into the RKIS flue gas could be performed only between the
hours of 6:00 AM and 6:00 PM.
This mercury CEM verification test was conducted jointly by Battelle's AMS Center, EPA's
Office of Research and Development (ORD), and the Massachusetts Department of Environ-
mental Protection (MDEP). Specifically, ORD research programs on mercury chemistry and
emissions made major contributions to this verification test in the form of planning and organiza-
tion, facility operations, data acquisition, and leadership in conducting the tests at the RKIS. The
MDEP provided financial support and comments on the test plan for this verification. In addition
to these collaborations, important contributions to the study were made by the staff of ARCADIS
Geraghty & Miller, Inc. Under subcontract with EPA/ORD, ARCADIS staff operated the test
facility and associated monitoring equipment and conducted ancillary tasks, such as sampling, to
confirm the concentrations of commercially prepared elemental mercury gas standards. Under
subcontract with Battelle, ARCADIS staff performed reference flue gas mercury sampling,
coordinated the analysis of those samples, and provided the resulting data for this report.
The schedule for the mercury CEM verification followed that stated in the test/QA plan,(1) in that
a week of setup and trial runs was followed by two weeks of verification testing under different
flue gas conditions. The overall schedule of the test is shown in Table 3-1.
Table 3-1. General Schedule of the ETV Mercury CEM Verification Test
Date
Verification Test Activities
January 8-12, 2001
Installation of vendor CEMs; shakedown of facility operations and mercury
injection procedures; trial sampling of flue gas
January 15-19, 2001
Verification testing with flue gas composition approximating that of a coal-
fired power plant
January 22-25, 2001
Verification testing with flue gas composition approximating that of a
municipal waste incinerator
January 26, 2001
Packing of vendor CEMs; end of test
The first and second weeks of verification testing (January 15 to 19 and 22 to 25, respectively)
were similar, in that largely the same daily sequence of tests was conducted in each of these
weeks. However, as noted in Table 3-1, the flue gas composition differed between the two
weeks.
It should be pointed out that, although facility shakedowns and trial sampling of flue gas were
conducted during the week of January 8 to 12, relatively little time was available for the mercury
CEMs to sample the full flue gas matrix. The variety of startup activities made it impossible to
conduct full days of trial flue gas sampling. This situation contrasts with the normal installation
procedures for such CEMs, in which one or two days are often allotted for extensive sampling of
the facility flue gas to tailor the CEM operation for the facility. The verification results reported
here must be considered in light of this mode of operation.
5
-------
The daily schedule of the verification test is illustrated in Table 3-2, which shows the testing
activities conducted on each day and the CEM performance parameters addressed by each
activity. The exact times when specific test procedures were done are noted when appropriate
elsewhere in this report. On each day of the week, the MS-l/DM-5 was challenged with a zero
gas and a commercial compressed gas standard of elemental mercury. These gases were supplied
to the analyzer one at a time as a test of day-to-day stability and speed of response. This test was
done twice a day, except that on the last day of each week it was done only once. Monday
through Wednesday of each week, the flue gas was sampled simultaneously by the MS-l/DM-5
and the Ontario Hydro(2) reference method. On Thursday and Friday of the first week,
interference and low-level mercury response tests were done, respectively.
Throughout the verification test, the MS-l/DM-5 was operated by representatives of Nippon
Instruments Corporation. The intent of the testing was for the MS-l/DM-5 to operate in a manner
simulating operation at a combustion facility. Therefore, once the verification test began, no
recalibration was performed other than the MS-l/DM-5's built-in automated zeroing procedure.
Nippon staff prepared reagents, maintained the monitor's inlet system, and handled recovery of
data from the MS-l/DM-5.
3.2 Test Conditions
The MS-l/DM-5 was one of four mercury CEMs tested in this verification effort. All verification
testing took place simultaneously, so that all CEMs were subjected to exactly the same test
conditions.
3.2.1 RKIS Conditions
The natural gas combustor of the RKIS was operated continuously throughout the test period, to
maintain elevated duct temperatures and thereby minimize the chance of retention and subse-
quent release of mercury by the refractory or other components of the system. The flue gas from
the natural gas combustor was spiked with gases, mercury, and particulate matter to achieve
different representative flue gas compositions. In all cases, once the required injection rates of
mercury, gases, and particulate matter were established, at least 30 minutes of stabilization time
was allowed before the start of any reference method sampling or verification data collection.
Mercury was injected into the RKIS flue gas using a peristaltic pump and aqueous solutions of
mercury(II) chloride (HgCl2). Different injection locations and injection solutions were tried in
trial runs to achieve stable total mercury levels and a reasonable split between elemental and
oxidized mercury in the duct. The final selection was to use mercury(II) chloride solutions, with
different injection locations (i.e., different injection temperatures) to achieve different elemental/
oxidized splits in the two weeks of testing.
6
-------
Table 3-2. Schedule of Daily Activities in the Mercury CEM Verification Test
Test Day AM/PM Activity (Performance Parameters)
Monday 1/15/01 AM Zero/Hg° standard gas (Calibration/Zero Drift, Response Time)
AM/PM Flue gas sampling (Relative Accuracy, Correlation, Precision)
PM Zero/Hg° standard gas (Calibration/Zero Drift, Response Time)
Tuesday 1/16/01 AM Zero/Hg° standard gas (Calibration/Zero Drift, Response Time)
AM/PM Flue gas sampling (Relative Accuracy, Correlation, Precision)
PM Zero/Fig0 standard gas (Calibration/Zero Drift, Response Time)
Wednesday 1/17/01 AM Zero/Fig ° standard gas (Calibration/Zero Drift, Response Time)
AM/PM Flue gas sampling (Relative Accuracy, Correlation, Precision)
PM Zero/Fig0 standard gas (Calibration/Zero Drift, Response Time)
Thursday 1/18/01 AM Zero/Fig ° standard gas (Calibration/Zero Drift, Response Time)
AM/PM Spiking of flue gas with interferant gases (Interferences)
PM Zero/Fig0 standard gas (Calibration/Zero Drift, Response Time)
Friday 1/19/01 AM Zero/Fig0 standard gas (Calibration/Zero Drift, Response Time,
Sampling System Bias)
AM/PM Preparation of low Fig levels in flue gas (Fow-Fevel Response)
Monday 1/22/01 AM Zero/Fig ° standard gas (Calibration/Zero Drift, Response Time)
AM/PM Flue gas sampling (Relative Accuracy, Correlation, Precision)
PM Zero/Fig0 standard gas (Calibration/Zero Drift, Response Time)
Tuesday 1/23/01 AM Zero/Fig ° standard gas (Calibration/Zero Drift, Response Time)
AM/PM Flue gas sampling (Relative Accuracy, Correlation, Precision)
PM Zero/Fig0 standard gas (Calibration/Zero Drift, Response Time)
Wednesday 1/24/01 AM Zero/Fig ° standard gas (Calibration/Zero Drift, Response Time)
AM/PM Flue gas sampling (Relative Accuracy, Correlation, Precision)
PM Zero/Fig0 standard gas (Calibration/Zero Drift, Response Time)
Thursday 1/25/01 AM/PM Repeat of some interference tests (Interferences)
PM Zero/Hg° standard gas (Calibration/Zero Drift, Response Time,
Sampling System Bias)
7
-------
Potential interferant gases (sulfur dioxide, nitrogen oxides, carbon monoxide, hydrogen chloride,
chlorine) were injected into the RKIS duct downstream of the combustion zone, using cylinders
of the pure compressed gases as the source. Nitrogen oxides in the flue gas were prepared by
injection of nitric oxide (NO). The target gas levels were established by monitoring the
concentration in the duct using the RKIS facility CEMs (see Table 3-3), rather than by estimation
based on dilution in the duct flow. The only exception was for chlorine, for which no CEM was
available. The chlorine concentration was established by first injecting chlorine into the flame
zone of the RKIS, where complete conversion of chlorine to hydrogen chloride is thermo-
dynamically assured, and measuring the resulting hydrogen chloride using the hydrogen chloride
CEM. The chlorine injection rate was then maintained, but moved to a lower temperature
injection point, and the concentration of HC1 was measured again. The chlorine concentration
was established from the difference in the two HC1 readings.
Particulate matter was injected into the duct at the downstream end of the RKIS combustion
zone, using a K-Tron Soder Model KCLKT20 screw feeder, which incorporated a strain gauge
measurement of the mass of material injected. The fly ash used was a lignite coal fly ash,
specially chosen for its low reactivity with mercury. The low target particulate loading
(30 mg/m3) required operating the feeder at the extreme low end of its operating range and
resulted in substantial variation in the particulate loading in the duct (see below).
Figure 3-1 shows two views of the RKIS test facility and indicates the locations of the injection
points for particulate matter, mercury, and interferant gases. Also shown are the port locations
RM1 (Port 2) and RM2 (Port 8) where the reference Ontario Hydro (OH) method samples were
collected. That sampling is presented in detail in Section 3.2.3. Between these two locations, at
Ports 6 and 7, were the sampling locations of the mercury CEMs undergoing testing; the MS-1/
DM-5 was located at Port 7. The MS-l/DM-5 test setup is described in more detail in
Section 3.2.2. Figure 3-1 also shows the sampling location for the facility CEMs, downstream of
all the sampling locations. The CEMs in place at the RKIS for this test are shown in Table 3-3.
The test conditions maintained throughout OH sampling in the two weeks of verification testing
were intended to represent a coal-fired power plant flue gas and an incinerator flue gas,
respectively. Table 3-4 summarizes the target and actual average levels of mercury and of the
other constituents in the flue gas in the two weeks of testing. In general, the actual constituent
levels in the flue gas were close to the target levels. The mercury present during the first week of
testing was predominantly in elemental form. During the second week of testing, to challenge the
speciation capabilities of the CEMs, the injected mercury was predominantly in oxidized form.
The flue gas water and oxygen contents were very consistent throughout both weeks of testing, at
about 7% water and 14.6% oxygen, and the flue gas temperature at the CEM sampling ports was
400 to 430°F .
3.2.2 MS-l/DM-5 Operation
The MS-l/DM-5 was installed on a table within about three feet of the duct, at the port location
identified as Port 7 in Figure 3-1. The MS-l/DM-5 shared a sample probe and particle filter with
another Nippon Instruments mercury CEM. The probe consisted of a glass tube of about 8-mm
outside diameter (OD) enclosed in a stainless steel tube that extended to near the center of the
8
-------
Gale
^a've Interference
s Inj.
Port 2
Port 1
Secondary Combustion Chamber
Mercury
Duct 1
injection
Porto —^0
2nd F oor
Flyash
Legend
Injection
4" Port
Afterburner
2" Port
I) Thermocouple
Porte
Main Burner
M T1
Ground F oor
K8n Section Transition Section
Draft Damper
I J
FGCS Manifold
CEM
HQ, S02, H20
o
4" Port
0
3" Port
0
2" Port
Thermocouple
Port 8
RM 2
Figure 3-1. Side View (top) and End View (bottom) of the RKIS Test
Facility
9
-------
Table 3-3. Summary of RKIS CEMs
Measurement
Analyte
CEM
Principle
Range (s)
o2
Rosemount Analytical Model 755R
Paramagnetic
0 - 25%
co2
Fuji Electric ZRH-1 Gas Analyzer
NDIR
0 - 20%
CO
Horiba Model PIR 2000
NDIR
0 - 500 ppm
NO/NOx
Thermo Environmental Model 10
Chemiluminescent
0 - 500 ppm
S02
Bodenzeewerk Model MCS 100
GFCIR
0 - 250, 0 - 2,500 ppm
HC1
Bodenzeewerk Model MCS 100
GFCIR
0- 100,0 - 1,000 ppm
Table 3-4. Target Flue Gas Constituent Concentrations (and Actual Ranges) Used in
Verification Test
Total Hg
so2
NOx
HC1
Particles
Test Week
(pg/m3)
(ppm)
(ppm)
(ppm)
(mg/m3)
One
8a
1,000
250
25
30
(7.3 - 7.8)
(952 - 978)
(226 - 245)
(21.5-24.3)
(0.2-71)
Two
80b
50
150
100
30
(72-119)
(51-63)
(137 - 159)
(77 - 92)
(2 - 54)
a Predominantly elemental mercury.
b Predominantly oxidized mercury.
duct. The probe connected to a heated glass container holding a glass fiber thimble filter. Each of
the two Nippon Instruments CEMs drew sample gas through a separate heated 6-mm OD Teflon
line, connected to a glass "T" fitting at the outlet of the filter housing. The sample gas flow of the
MS-l/DM-5 was 0.5 L/min. The total gas flow through the inlet probe and filter was about
1 L/min, and sampling was not isokinetic.
The nature of the inlet system is an important factor in verifying the performance of a mercury
CEM, because the materials and temperature of the inlet may affect the chemical speciation and
even the quantity of mercury that reaches the CEM. Since the CEMs were verified in this test
partly by comparison to the OH method, each vendor was informed of the characteristics of the
inlet system used with the OH trains. Vendors then made their own decisions about how best to
sample the flue gas to assure consistent comparisons with the OH method. In the inlet system
used with the MS-l/DM-5, the filter was maintained at about 400°F, similar to the duct tempera-
ture; the heated line from the filter holder to the MS-l/DM-5 was maintained at about 300°F. At
the vendor's discretion, due to the low particulate matter loading, the inlet filter was changed
10
-------
only after the first week of testing. The MS-l/DM-5 determines vapor-phase elemental and
oxidized mercury only, so the particulate matter collected on the filter was not analyzed.
The MS-l/DM-5 is a truly continuous mercury monitor. That is, the two DM-5 detectors
continuously provide indications of the elemental and oxidized mercury in the flue gas. The sum
of these two separate channels is a continuous indication of total vapor-phase mercury. The
outputs of the MS-l/DM-5 were recorded on a laptop computer at 10-second intervals
throughout the verification test. During the first week of testing, data were recorded at 10-second
intervals for six and one-half minutes, then data collection was interrupted for 30 seconds for
calculation of the average readings over the period, and then data collection resumed. In the
second week of testing, data were recorded for eleven and one-half minutes before the 30-second
averaging period and resumption of data collection.
3.2.3 Ontario Hydro Reference Method
The OH mercury speciation method® was used as the reference method in this verification test,
because of the need for a recognized approach to verify the elemental and oxidized mercury
measurements made by the CEMs. Since the OH method is not officially designated as a
reference method by EPA or the American Society for Testing and Materials, it was judged
necessary to document the precision of the OH method itself during the verification test. As a
result, in each OH sampling run, two OH trains were operated simultaneously at each of the two
reference sampling points at the RKIS (Port 2, or RM1, and Port 8, or RM2, in Figure 3-1). At
each location, two trains were placed at ports on opposite sides of the duct (designated 2A, 2C,
8A, and 8C, respectively), with their inlet probes positioned close together near the center of the
duct. Thus, four OH trains were used for sampling at the same time in each OH run. Each OH
train used a separate inlet probe and heated filter and sampled isokinetically near the duct center-
line. An EPA Method 2 traverse was conducted at each OH sampling location, both before and
after each OH run, to set the isokinetic sampling rate. The average of pre-and post-run traverses
was used for final calculations.
The particulate filters in the OH trains were maintained at 250°F. No heating was applied to the
sample probe, instead the temperature dropped naturally from the duct temperature to the filter
temperature in passage from the probe to the filter. A sampling flow of about 1 m3/hr (about
16.7 L/min) was used in all OH runs.
None of the CEMs undergoing verification determined particulate mercury. As a result, the par-
ticulate filters from the OH trains were used for two purposes: to determine particulate mercury
and to determine the particulate mass loading in the duct. For each OH run, one of the particulate
filters from each sampling location (RM1 and RM2, Figure 3-1) was used for gravimetric deter-
mination of particle mass loading. The other was analyzed for particulate phase mercury. This
approach provided a check on the particulate mercury level, while avoiding the expense of
additional Method 5 runs to determine the particle loading. As detailed in Section 4.2.1, the
particulate mercury determined by the OH sampling was negligible.
The meter boxes of the paired OH trains at each location were located side by side to allow
monitoring by a single operator during sampling. During the runs, OH train data were recorded
11
-------
by hand on data sheets prepared for this test and also were entered immediately into an Excel
spreadsheet running on a laptop computer at the sampling location. That spreadsheet auto-
matically calculated run parameters, such as percent of isokinetic sampling rate, providing rapid
and accurate control of the sampling process.
The OH trains were prepared by ARCADIS staff in a dedicated laboratory close to the RKIS high
bay. Trains were brought to the RKIS for sampling and returned to the same laboratory for
sample recovery immediately after sampling. Blank trains (one per day of OH sampling) were
also brought to one of the sampling locations at the RKIS and returned along with the sampled
trains for sample recovery. The glassware set used for the blank train was chosen at random each
day from the several sets on hand. In addition, blank samples were taken of all reagents and rinse
solutions used in the OH method. Samples were recovered into uniquely identified pre-labeled
glass containers according to OH method procedures. Samples were delivered by ARCADIS staff
within about 24 hours after sampling to Oxford Laboratories, in Wilmington, North Carolina, for
analysis. Staff of Oxford Laboratories verified sample identities from chain-of-custody sheets
before beginning sample analysis.
OH sampling was conducted under conditions of stable flue gas composition, on Monday,
Tuesday, and Wednesday of each test week (i.e., on January 15 to 17 and 22 to 24, 2001). Two
OH runs per day, of two hours each, were made during the first week. Three OH runs per day, of
one hour each, were made during the second week. Typical sampled gas volumes were about
2 cubic meters and 1 cubic meter in the first and second weeks of testing, respectively. Table 3-5
shows the actual schedule of OH sampling in the verification test. The OH sampling proceeded
smoothly, with the single exception of Run 9, which was interrupted by failure of the mercury
solution injection pump. That run was resumed after repair of the pump, and no data were lost.
3.3 Verification Procedures
This section describes the specific procedures used to verify the CEM performance parameters
that were listed in Section 3.1. The statistical procedures used to calculate the verification results
are described in Section 5 of this report.
3.3.1 Relative A ccuracy
Relative accuracy was assessed by the quantitative comparison of the mercury results from the
OH method to those from the tested CEM. For the MS-l/DM-5, this comparison was made for
elemental, oxidized, and total mercury. The MS-l/DM-5 results during the period of each OH
run were averaged for comparison to the OH result. The accuracy was calculated separately for
the two different test conditions, i.e., the 24 OH results from the first week of testing and the 36
OH results from the second week were the basis for separate calculations of accuracy.
12
-------
3.3.2 Correlation with Reference Method
The correlation of the MS-l/DM-5 with the OH results was assessed using the same data used to
assess accuracy. The average MS-l/DM-5 mercury results over each OH sampling period were
calculated and compared to the corresponding OH results.
Table 3-5. Schedule of OH Sampling Runs During Mercury CEM Verification
Week
Run Number
Date
Start Time
End Time
One
1
1/15/01
12:07
14:08
2
1/15/01
15:36
17:37
3
1/16/01
12:12
14:13
4
1/16/01
15:45
17:45
5
1/17/01
10:58
12:59
6
1/17/01
14:21
16:22
Two
7
1/22/01
11:32
12:33
8
1/22/01
13:25
14:25
9
1/22/01
16:15
16:5 la
17:40
18:07a
10
1/23/01
10:47
11:47
11
1/23/01
12:50
13:51
12
1/23/01
16:00
17:01
13
1/24/01
11:15
12:15
14
1/24/01
14:30
15:31
15
1/24/01
16:15
17:16
a Run 9 was interrupted by a failure of the peristaltic pump used to inject mercury solution into the RKIS; the run
was resumed after fixing the pump and allowing concentrations to stabilize.
3.3.3 Precision
Precision of the MS-l/DM-5 was determined based on the individual MS-l/DM-5 results in each
OH sampling run. The relative standard deviation of the successive MS-l/DM-5 readings was
calculated as the measure of precision. This calculation was intended to assess CEM variability
under conditions of stable mercury levels. Consequently, this calculation was limited to those
time periods in which CEM data, facility data, and the observations of testing staff indicated that
mercury addition and mercury flue gas levels were stable. Occasional spikes in mercury
concentration were excluded from the calculation of precision, provided that the spikes were
attributable to occurrences at the test facility, either by corroboration of multiple CEMs, or by
observations of testing staff.
13
-------
3.3.4 Calibration/Zero Drift
Day-to-day drift in MS-l/DM-5 response to calibration and zero gases was assessed by sampling
zero gas (nitrogen) and a commercial standard of elemental mercury in nitrogen with the two
DM-5 detectors each day. The standard gases used for this procedure are described in
Section 3.4.1. Those gases were not used as absolute mercury standards, but only as sources of
stable mercury levels over the test period. The calibration/zero check was done twice on each day
of testing, before and after all other test activities. However, on the last day of each week, this
test was done once and was combined with a check of the sampling system bias (see
Section 3.3.5). A low concentration mercury standard (-15 (ig/m3) was used in the first week,
and a higher concentration standard (~ 50 (ig/m3) was used in the second week, to parallel the
flue gas mercury levels used (see Table 3-4).
Drift was assessed in terms of the range and relative standard deviation of the repeated zero and
calibration checks. Separate assessments of drift were made for the two weeks of testing.
3.3.5 Sampling System Bias
In nearly all cases, the drift checks described in Section 3.3.4 were conducted by supplying the
zero and standard gases directly at the DM-5 detector inlets of the CEM. However, on the last
day of each week of testing, a drift check conducted in that manner was followed immediately by
a similar check in which each gas was supplied at the inlet of the CEM's flue gas sampling
system. The ratio of the CEM response at the analyzer to that through the entire sampling system
determined the bias caused by the sampling system.
The sampling system bias test was performed by withdrawing the Nippon sampling probe from
the duct and connecting the tip of the inner glass probe to a Teflon line supplying the zero or
calibration gas. A "T" fitting in the line allowed venting of a small excess flow of gas beyond
that drawn by the Nippon CEMs. The gas supplied in this way passed through the probe, filter,
and heated sampling lines before entering the inlet of the MS-l/DM-5 CEM.
3.3.6 Interferences
On Thursday, January 18, the same low mercury level used on previous days (see Table 3-4) was
established in the RKIS flue gas. The addition of mercury alone continued for approximately two
hours, during which time the MS-l/DM-5 sampled the flue gas to establish a baseline level of
response. Then, with the mercury injection continuing as before, a series of potential interferant
gases was injected into the flue gas, first one at a time, and then in combination. Finally, the
gases were turned off, and the mercury injection alone was continued for about another half hour.
The effect of the interferant gases was assessed by the changes in MS-l/DM-5 response
occurring during the introduction of the gases.
Table 3-6 summarizes the schedule of the interference test on January 18. Shown are the sub-
stances injected into the RKIS flue gas in successive time intervals during this test. The concen-
trations shown in Table 3-6 are the target concentrations; actual concentrations were maintained
within ± 10% of the target values.
14
-------
Table 3-6. Schedule of January 18, 2001, Interference Test
Hga
(pg/m3)
Substances Injected in Flue Gas (Target Concentration, ppm)
Time
NO CO S02 HC1 Cl2
11:05 - 13:06
8
13:06-13:39
8
500
13:39-14:16
8
500
14:16-15:05
8
2,000
15:05 - 15:50
8
250
15:50-16:30
8
10
16:30-17:06
8
500 10
17:06-17:36
8
500 500 2,000 250 10
17:36 - 18:02b
8
Approximate total mercury level.
bBrief inadvertent injection of Cl2 to the duct occurred near the start of this time period.
3.3.7 Response Time
Response time of the MS-l/DM-5 was determined in both weeks of the test, as part of the
calibration/zero drift test. The rise time and fall time were determined by recording successive
MS-l/DM-5 readings when the delivery of elemental mercury standard gas was started or
stopped, respectively. The time to reach 95% of the final value was estimated by interpolating
between successive 10-second readings.
3.3.8 Low-Level Response
The mercury levels shown in Table 3-4 were not intended to challenge the detection limit of the
MS-l/DM-5. Instead the ability of the MS-l/DM-5 to detect low mercury levels was tested by a
procedure conducted on January 19. On that day, the MS-l/DM-5 initially sampled RKIS flue
gas with no mercury present, but with sulfur dioxide, nitrogen oxide, and hydrogen chloride
present at the concentrations used in the first week of testing (i.e., 1,000 ppm, 250 ppm, and
25 ppm, respectively; Table 3-4). Injection of mercury then began, starting with a low mercury
concentration, and stepping upwards in concentration at time intervals of about one-half hour.
After the highest mercury concentration, the mercury injection was shut off, and flue gas without
added mercury was again sampled. The mercury solutions used for this test had aqueous con-
centrations of 0.304 to 2.45 |ig/mL, sufficient to produce total flue gas mercury concentrations of
nominally 0.57 to 4.54 |ig/m3. The actual flue gas mercury concentrations were not determined;
as a result, the nominal total mercury concentrations are used for comparison to the CEM results.
The low-level response of the MS-l/DM-5 was determined based on the lowest mercury
15
-------
concentration detected above the flue gas background. Table 3-7 summarizes the schedule for
this test.
Table 3-7. Schedule of January 19, 2001, Low-Level Response Test
Time Period
Nominal Flue Gas Total Hg Concentration
(pg/m3)
10:44-11:39
0
11:39-12:21
0.57
12:21 - 12:56
1.13
12:56-13:33
2.27
13:33-14:15
4.54
14:15-14:55
0
3.3.9 Data Completeness
Data completeness was determined by comparing the data recovered from the MS-l/DM-5 to the
amount of data expected upon completion of all portions of the verification test. Data complete-
ness was evaluated in terms of the percentage of total data recovered.
3.3.10 Setup and Maintenance Needs
Setup and maintenance needs were documented qualitatively, through observation and com-
munication with the vendor during the verification test. Factors noted included the frequency of
scheduled maintenance activities, the extent of any downtime, the number of staff operating or
maintaining the CEM, and the quantity of consumables used and/or waste materials produced.
3.4 Equipment and Materials
3.4.1 Commercial Elemental Mercury Standards
Four commercial compressed gas standards of elemental mercury in nitrogen were used as stable
sources of elemental mercury for verification of day-to-day instrument calibration drift and
sampling system bias. These gases were purchased from Spectra Gases, Inc., Branchburg, New
Jersey. Spectra Gases reported the nominal prepared mercury concentration of each standard and
performed an initial analysis of each cylinder in November 2000, using a commercial CEM of a
different design than those participating in the verification test. In addition, each of these
standards was sampled before and after the verification test, using a miniature impinger train
modeled after the OH method. Collected samples were then submitted for laboratory mercury
analysis, along with the collected samples from OH flue gas sampling, to determine the cylinder
16
-------
gas mercury content. The protocol for conducting this miniature impinger sampling was sub-
jected to review by ARCADIS and EPA staff and was approved by EPA prior to use in this
study.(3) Unfortunately, the pre-test analysis results showed evidence of excessive loss of mercury
in sampling the gas with the mini-impinger train. This possibility was subsequently confirmed by
conducting the sampling with a different pressure regulator on the mercury gas standard. The
regulator originally used was shown to remove some mercury from the gas and to require
extremely long equilibration times to achieve stable delivery of mercury. Consequently, the pre-
test results are not valid. Post-test results from the mini-impinger train, and from an EPA-owned
mercury CEM like the one used by the gas vendor, are shown in Table 3-8.
Table 3-8 lists the mercury standards used, along with the vendor's nominal prepared concen-
tration and initial analysis result, provided by Spectra Gases. Also listed in Table 3-8 are the
cylinder concentrations determined in March 2001 by means of the mini-impinger sampling
method and an EPA-owned mercury CEM.
Table 3-8. Mercury Standard Gas Identification and Analysis Results
Pre-Test Post-Test
Cylinder
Number
Prepared
Hg°Conc.
(pg/m3)a
Gas Vendor
Analysis
(pg/m3)
OH Mini-Train
Analysis
(pg/m3)
EPA CEM
Analysis
(pg/m3)
CC19870
12.5
14
11.7
9.3
CC19931
14.1
16
13.6
13.0
CC20219
44.0
49
50.2
47.0
CC20291
49.8
55
46.8
50.6
a Concentrations in |ig/m ! converted from ppbv concentrations stated by Spectra Gases, using a conversion factor of
1 ppbv mercury = 8.3 (ig/m3 at 20°C and 1 atmosphere pressure.
Table 3-8 shows generally good consistency among the gas vendor's prepared mercury concen-
trations, the gas vendor's own pre-test analysis, and the post-test analyses. In particular, the
agreement between the mini-impinger OH samples obtained post-test and the vendor's prepared
concentration is within 6.4%, 3.5%, 14.1%, and 6.0%, respectively, for cylinders CC19870,
CC19931, CC20219, and CC20291. This agreement is good, considering the novel nature of
these standards and the four-month time interval between their preparation and the post-test
analysis.
Cylinders CC19931 and CC19870 were prepared at full cylinder pressure of 2,000 psig. How-
ever, because of vapor pressure limitations of elemental mercury, cylinders CC20291 and
CC20219 were prepared with an initial pressure of only 900 psig. To assure consistent testing of
day-to-day drift, each CEM undergoing verification was challenged repeatedly with just one gas
17
-------
standard in each week of testing. In addition, to assure sufficient gas to complete the testing in
each week, the gas standards were assigned to participating CEMs in such a way as to balance
the consumption of all standard gases. Thus, the MS-l/DM-5 monitor was challenged in the first
week of testing only with mercury standard cylinder number CC19870 and, in the second week,
only with cylinder number CC20219. In both weeks of testing, the use of those cylinders was
shared among the MS-l/DM-5, another Nippon CEM, and another vendor's CEM. The final gas
pressures in the elemental mercury standards at the conclusion of testing were approximately
900 psig and 1,000 psig in cylinders CC 19870 and CC 19931, respectively, and approximately
300 psig and 550 psig in cylinders CC20219 and CC20291, respectively.
The mercury standard cylinders were used with miniature low-volume stainless steel regulators
(Spectra Gases 04-1D3EATNN-018), which were flushed for about 30 minutes before use with a
0.5 L/min flow of the standard gas. The delivery plumbing from the regulators was originally
made almost entirely of SilcoSteel® fittings, i.e., stainless steel internally coated with a layer of
glass. However, long equilibration times were observed in first supplying the 12.5 and
14.1 (ig/m3 standards to the CEMs in the first week of testing. Consequently, the delivery system
was modified to consist of all Teflon, except for the regulator itself. This change notably
improved the delivery of the mercury gas standards to the CEMs.
3.4.2 Performance Evaluation Equipment
Performance evaluation (PE) audits were conducted on several key measurements at the RKIS.
Each of those audits was conducted using a reference standard or measurement system provided
by Battelle that was independent of that used in the verification test. Table 3-9 lists the PE audit
equipment used and, when appropriate, the date of the calibration of the audit equipment prior to
the verification test. The PE audit of the OH mercury analysis was done by spiking two blank
trains with a dilution of a commercial National Institute of Standards (NIST)-traceable mercury
standard. The results of the PE audits are reported in Section 4.3.2.
18
-------
Table 3-9. Performance Evaluation Audit Equipment Used for the Verification Test
Measurement
Audited
PE Equipment
Date
Calibrated
Source of
Calibration
Flue gas 02 and C02
LandTec GA-90 Gas Analyzer,
Model GA 1.1 (S. No. 693)
12/19/00
Vendor
Flue gas temperature
Fluke Model 52, (S. No. 73970010) with
Type K Thermocouple
11/29/00
Battelle
Instrument Lab
Barometric pressure
Taylor Model 2250M Aneroid Barometer
(Inventory No. LN163610)
12/20/00
Battelle
Instrument Lab
Flue gas pressure
Magnehelic Model 2005
(S. No. R51006LG64)
10/25/00
Battelle
Instrument Lab
Impinger weighing
Cenco Class T Weight Set
(200 g and 500 g weights)
1/3/01
Battelle
Instrument Lab
Ontario Hydro
mercury analysis
EM Science Atomic Absorption Standard
MX0399-2, traceable to NIST SRM
#1333.
NA
Vendor
19
-------
Chapter 4
Data Quality
The quality of the verification data was assured by quality assurance and quality control
(QA/QC) procedures, performed in accordance with the quality management plan (QMP) for the
AMS Center(4) and the test/QA plan(1) for this verification test. Minor deviations from the test/QA
plan were documented in the verification records at the RKIS during testing. Deviations required
the approval of Battelle's AMS Center Manager. A planned deviation form was used for
documentation and approval of each of the following changes:
1. The elemental mercury gas standards were not analyzed by the University of North
Dakota Energy and Environmental Research Center because the procedure to be
used would have consumed a substantial fraction of each standard.
2. The performance evaluation for oxygen and carbon dioxide measurements at the
RKIS used an electrochemical monitor for these gases rather than paramagnetic and
infrared monitors, as stated in the plan.
3. The solution used for the injection of mercury into the RKIS duct was made with
mercuric chloride instead of mercuric nitrate.
4. As a result of the use of mercuric chloride rather than mercuric nitrate for the
injection solution, the stock solution was made up with hydrochloric acid instead of
nitric acid, and no additional acid was added in making up dilutions of the stock
solution.
5. Recovered OH samples were stored at room temperature before delivery to the
analytical laboratory, consistent with OH method requirements, instead of under
refrigeration.
6. To better test the capabilities of the mercury CEMs, the low-level response test was
conducted with flue gas mercury levels of about 0.5, 1, 2, and 4 (ig/m3 instead of 1,
2, 4, and 8 (ig/m3.
7. The RKIS carbon dioxide CEM was changed before the verification test began. A
Fuji Electric Model ZRH-1 was in place at the RKIS, instead of the Horiba VIA 510
stated in the test/QA plan.
20
-------
8. Calculation of interference effects in terms of "relative sensitivity" was not done,
since this calculation was inappropriate in situations where chemical trans-
formations of mercury species took place.
None of these deviations had any significant effect on the quality of the verification data.
4.1 Facility Calibrations
Continuous monitors for oxygen, carbon monoxide, carbon dioxide, nitrogen oxides, sulfur
dioxide, and hydrogen chloride are installed at the RKIS and its associated APCS to monitor
system performance and document flue gas composition. For this verification test, those monitors
were calibrated according to standard facility practice on each test day. Calibration procedures
consisted of a multipoint calibration check at the start of the day and a drift check at the end of
the day. On days when OH sampling was conducted, the drift check was also conducted after
each OH run. All calibration results were within the allowable tolerances for drift and linearity.
All calibration results were documented for inclusion in the verification data files.
Flue gas water content was determined from impinger weights in the OH trains by means of an
electronic balance located in the OH train preparation/recovery laboratory. Copies of the calibra-
tion records for that balance were included in the data file.
Key measurements that factored directly into the verification test results were also the subject of
PE audits, as described in Section 4.3.2. Those measurements included the facility CEMs for
oxygen and carbon dioxide and the balance used for determination of water.
4.2 Ontario Hydro Sampling and Analysis
The preparation, sampling, and recovery of samples from the OH trains followed all aspects of
the QA/QC requirements in the OH method. A daily blank train was prepared, kept at either the
upstream or downstream sampling location during sample runs, and recovered along with the
sampled trains. The impinger glassware used for preparing the blank train was selected at random
each day from among the several sets used, so that blank results reflect the actual state of the
sampling equipment. All required reagent blanks were collected, and additional reagent blanks
(beyond those required in the method) of the acetone rinse reagent and of the 5% w/v potassium
permanganate called for in Section 13.2.8.3 of the OH method were collected. A sample number-
ing system was implemented that provided unique identification of each train and of each
recovered sample from that train. This numbering system was implemented by means of pre-
printed labels applied to sample containers arranged in order of sample recovery for each train.
The recovered samples were delivered to the analytical laboratory within about 24 hours after
collection. Oxford Laboratories, which conducted the OH analyses, similarly adhered to all
requirements of the OH analytical process. Replicate analysis of samples was performed as
required by the OH method, and all results met the 10% acceptance criterion. The analytical
results for each set of analyses were accompanied by data quality documentation that reported the
laboratory calibration procedures and results applicable to those analyses.
21
-------
Because of the importance of the OH data in this verification, the following sections present key
data quality results from the OH data.
4.2.1 Ontario Hydro Precision
The results of the OH flue gas sampling and analysis are shown in Tables 4-la-c and 4-2a-c, for
the first and second weeks of verification testing, respectively. Tables 4-la and 4-2a show the
elemental mercury results, Tables 4-lb and 4-2b the oxidized mercury results, and Tables 4-lc
and 4-2c the total vapor-phase mercury results from the OH runs. Particulate mercury, deter-
mined on one OH particulate filter at each location in each OH run, was not detectable during the
first week of verification testing. During the second week of testing, of the 18 filters analyzed for
particulate mercury, 11 showed mercury of less than 0.5 (ig/m3, five showed levels 0.5 to
1 (ig/m3, and two showed levels of 1.0 to 1.5 (ig/m3. Thus, particulate mercury was a negligible
fraction of the total mercury present in RKIS flue gas.
Inspection of the data in these tables shows that, in nearly all cases, the agreement between
duplicate OH results (i.e., the precision of the OH method) was good at both the upstream and
downstream sampling locations. Furthermore, the mercury levels determined at the upstream
location generally agreed closely with those at the downstream location. This observation indi-
cates that little loss of mercury or change in mercury speciation occurred during the transit of flue
gas from one location to the other. The variability in OH results was larger in the second week of
sampling (Tables 4-2a-c) than in the first week (Tables 4-la-c). This undoubtedly results in part
from the greater proportion of oxidized mercury present in the second week. The oxidized
mercury is more difficult to transport and sample, and the proportion of oxidized mercury may
have varied due to small variations in RKIS conditions (e.g., temperature). Day-to-day variations
in mercury results in the second week are also due in part to changes in delivery conditions. Most
notably, the pump used to inject the mercury solution into the RKIS failed during OH Run 9 on
January 22 and was replaced with a different pump. The higher total mercury levels measured on
January 23 and 24, relative to those on January 22, are probably due to a slightly higher delivery
rate on those days.
To quantify the characteristics of the OH reference data before using them for verification of the
Nippon MS-l/DM-5, the OH data were subjected to a statistical analysis, addressing three issues:
¦ The precision of the duplicate OH results at both the upstream and downstream sampling
locations
¦ The agreement between the mercury levels determined by the OH method at the upstream
and downstream sampling locations
¦ The identification of any outliers in the data.
These issues were addressed separately for the elemental, oxidized, and total mercury data.
22
-------
Table 4-la. Elemental Mercury Results from OH Sampling in the First Week of
Verification Testing
Elemental Hg (p/m3)
Upstream
Downstream
Date
Port Location
2A
2C
8A
8C
1/15/01
Run 1
6.78
6.49
6.28
6.29
1/15/01
Run 2
6.20
6.23
6.18
5.98
1/16/01
Run 3
6.51
6.52
6.09
6.12
1/16/01
Run 4
6.43
7.00
7.55
6.50
1/17/01
Run 5
6.31
6.53
6.03
5.95
1/17/01
Run 6
6.10
6.56
5.29
5.99
Table 4-lb. Oxidized Mercury Results from OH Sampling in the First Week of
Verification Testing
Oxidized Hg (pg/m3)
Upstream
Downstream
Date
Port Location
2A
2C
8A
8C
1/15/01
Run 1
1.11
1.11
1.23
1.07
1/15/01
Run 2
0.93
1.11
1.20
1.36
1/16/01
Run 3
1.05
1.05
1.25
1.12
1/16/01
Run 4
1.09
0.99
1.40
1.26
1/17/01
Run 5
1.32
1.29
1.42
1.30
1/17/01
Run 6
1.70
1.42
1.64
1.68
23
-------
Table 4-lc. Total Mercury Results from OH Sampling in the First Week of Verification
Testing
Total Hg (pg/m3)
Upstream
Downstream
Date
Port Location
2A
2C
8A
8C
1/15/01
Run 1
7.89
7.60
7.51
7.36
1/15/01
Run 2
7.13
7.34
7.38
7.34
1/16/01
Run 3
7.56
7.57
7.34
7.24
1/16/01
Run 4
7.52
7.99
8.95
7.76
1/17/01
Run 5
7.63
7.82
7.45
7.25
1/17/01
Run 6
7.80
7.98
6.93
7.67
a Shaded cells indicate data excluded as outliers.
Table 4-2a. Elemental Mercury Results from OH Sampling in the Second Week of
Verification Testing
Elemental Hg (p/m3)
Upstream Downstream
Date
Port Location
2A
2C
8A
8C
1/22/01
Run 7
11.1
12.4
11.4
11.6
1/22/01
Run 8
14.3
15.1
13.6
11.5
1/22/01
Run 9
25.1
20.9
27.4
27.4
1/23/01
Run 10
7.0
6.0
6.9
1/23/01
Run 11
8.2
9.3
7.2
7.5
1/23/01
Run 12
8.0
8.1
6.7
3.7
1/24/01
Run 13
6.8
7.2
6.4
6.2
1/24/01
Run 14
4.9
5.4
6.3
5.7
1/24/01
Run 15
6.7
6.9
7.6
5.2
a Shaded cells indicate data excluded as outliers.
24
-------
Table 4-2b. Oxidized Mercury Results from OH Sampling in the Second Week of
Verification Testing
Oxidized Hg (p/m3)
Upstream
Downstream
Date
Port Location
2A
2C
8A
8C
1/22/01
Run 7
61.4
67.3
63.3
69.1
1/22/01
Run 8
71.2
75.2
67.2
67.4
1/22/01
Run 9
46.5
2
44.0
46.0
1/23/01
Run 10
60.5
97.0
96.3
97.7
1/23/01
Run 11
111.2
110.2
100.5
98.5
1/23/01
Run 12
113.8
114.3
105.5
70.5
1/24/01
Run 13
88.0
88.0
85.6
84.5
1/24/01
Run 14
77.7
87.7
81.6
67.5
1/24/01
Run 15
86.9
91.4
84.8
81.8
a Shaded cells indicate data excluded as outliers.
Table 4-2c. Total Mercury Results from OH Sampling in the Second Week of Verification
Testing
Total Hg (pg/m3)
Upstream Downstream
Date
Port Location
2A
2C
8A
8C
1/22/01
Run 7
72.5
79.7
74.7
80.7
1/22/01
Run 8
85.4
90.2
80.7
78.9
1/22/01
Run 9
71.5
51.1
71.4
73.4
1/23/01
Run 10
61.0
104.0
102.3
104.6
1/23/01
Run 11
119.4
119.5
107.7
106.1
1/23/01
Run 12
121.8
122.5
112.2
71.1
1/24/01
Run 13
94.9
95.2
92.0
90.7
1/24/01
Run 14
82.6
93.1
87.9
73.2
1/24/01
Run 15
93.6
98.3
92.4
87.0
a Shaded cells indicate data excluded as outliers.
25
-------
The precision of duplicate OH trains at both sampling locations was assessed by Student's t-tests.
Nonparametric Wilcoxon Rank Sum tests were also used to compare the results from the t-tests.
The outcome of this analysis was that, in the first week of verification testing, the average agree-
ment of paired OH trains was within 2.7% for elemental mercury, within 4.4% for oxidized
mercury, and within 1.7% for total mercury. In the second week of testing, agreement of paired
trains was within 14.8% for elemental mercury, within 7.2% for oxidized mercury, and within
8.1% for total mercury. These results are very good considering the low concentrations of
mercury species and the difficulty of preparing and sampling oxidized mercury in the flue gas.
The statistical analysis found no significant differences between the paired OH results (i.e.,
port A vs. port C) for elemental, oxidized, or total mercury at either the upstream or the down-
stream sampling location.
Evaluation of upstream/downstream differences in the OH results used an analysis of variance
(ANOVA) approach. Student's t-tests and Nonparametric Wilcoxon Rank Sum tests were also
used to compare the results of the ANOVA. The statistical analysis found no significant differ-
ences between the upstream and downstream levels of elemental or total mercury in either week
of testing. The upstream and downstream elemental mercury levels agreed on average within
4.9% and within 7.0% in the first and second week of testing, respectively. Those of total
mercury agreed within 2.0% and within 4.3%, respectively. Similarly, in the second week of
testing the upstream and downstream levels of oxidized mercury agreed within 4.2%. On the
other hand, a significant upstream/downstream difference of 11.2% in oxidized mercury was
found in the first week of testing. However, it must be noted that this resulted from an average
absolute difference of only 0.15 (ig/m3 at oxidized mercury levels of 1.2 to 1.3 (ig/m3. This
absolute difference is comparable to the detection limit of the OH method. Although a significant
upstream/downstream difference may be indicated by the statistical tests, the agreement between
upstream and downstream oxidized mercury levels is still very good considering the low mercury
levels present.
The analysis for outliers in the OH data relied on a studentized residual approach as the primary
criterion, with COVRATIO and DFFITS statistics® as secondary criteria. A mercury measure-
ment was considered an outlier if the primary criterion and at least one of the secondary criteria
were met. The aim of this analysis was to identify OH results that were not accurate indications
of the flue gas mercury content, rather than to eliminate data that resulted from facility or other
variations. For example, the statistical analysis identified all the elemental mercury results from
OH Run 9 (20.9 to 27.4 (ig/m3, Table 4-2a) as outliers relative to other data from the second
week of testing. However, those values are thought to arise from the failure of the mercury solu-
tion delivery pump in that run and the interruption and resumption of mercury delivery to the
RKIS. That is, the OH elemental mercury results from Run 9 are thought to reflect actual test
facility variability. As a result, most of the OH results from Run 9 have been retained in the data
set. On the other hand, the analysis did disclose a few outliers in the OH data, which are shown
in Tables 4-la-c and 4-2a-c in shaded cells. Those values were excluded from the calculations
used to verify mercury CEM performance on the basis of the duplicate precision and upstream/
downstream differences for individual results. In this analysis, the precision of paired results was
compared to the average precision results stated above. Individual values showing differences
greater than three times the average precision were excluded. As Tables 4-1 and 4-2 show, this
26
-------
procedure excluded a few individual values differing sharply from the other simultaneous OH
results.
As a result of these analyses of the OH data, it was concluded that each OH result, excluding
outliers, could be taken as a separate and independent measurement of flue gas composition. The
result of excluding the outlying values is that, for the first week of testing, 24 values for
elemental and oxidized mercury were used in the verification and 23 values for total mercury.
For the second week of testing, 34 values for elemental mercury and 33 values for oxidized and
total mercury were used.
4.2.2 Ontario Hydro Blank Trains
Table 4-3 shows the analytical results from the six blank OH trains collected during the verifica-
tion test. This table lists the flue gas concentrations (in (ig/m3) that would be inferred from the
blank train results, assuming gas sample volumes of 2 m3 in the first week of testing and 1 m3 in
the second week. Results are shown for oxidized mercury (from the potassium chloride
impingers of the OH trains) and for elemental mercury (from both the peroxide and potassium
permanganate impingers of the OH trains). The detection limit for mercury in these samples was
0.1 |ig. As Table 4-3 indicates, the great majority of blank train results were below the detection
limit, and the few detectable mercury levels were negligible when compared to the levels actually
found in the flue gas samples (Section 4.2.1). These results indicate that OH samples were not
exposed to contamination sources during sample recovery, handling, and analysis.
Table 4-3. Equivalent Flue Gas Mercury Concentrations (jig/m3) Found in Blank Ontario
Hydro Trains
Date
Train ID
Oxidized Hg
Hg° (HA)
Hg° (KMnOJ
1/15/01
I-GR-OHI-la
< 0.05
< 0.05
0.29
1/16/01
I-GR-OHA-2a
<0.05
<0.05
<0.05
1/17/01
I-GR-OHB-3a
<0.05
<0.05
<0.05
1/22/01
II-GR-OHI-4b
<0.1
<0.1
0.31
1/23/01
II-GR-OHI-5b
<0.1
<0.1
<0.1
1/24/01
II-GR-OHI-6b
0.12
<0.1
<0.1
a Equivalent flue gas concentrations calculated from blank train results using assumed sample volume of 2 m3.
b Equivalent flue gas concentrations calculated from blank train results using assumed sample volume of 1 m3.
4.2.3 Ontario Hydro Reagent Blanks
A total of 39 samples of the various OH reagents were collected for analysis between January 15
and January 24, covering all impinger reagents and train rinse solutions. Mercury was found at
detectable levels in only two of those 39 samples. The levels found in those two samples were
27
-------
negligible in terms of their equivalent flue gas mercury concentration when compared to the
concentrations found in the actual flue gas samples (Section 4.2.1).
4.3 Mercury Mass Balance
Because of the potential for loss of mercury from the flue gas, a valuable QA exercise is to
calculate the mass balance of mercury in the RKIS facility. This was done by comparing the total
mercury result from each OH train (Section 4.2.1) to the expected total mercury level in the
RKIS flue gas, based on the duct flow rates and the concentrations and flow rates of the injection
solutions. The ratio of the OH total mercury to the expected total mercury, expressed as a
percentage, is defined as the percent recovery of the injected mercury. Table 4-4 summarizes this
comparison, showing percent recovery values for both the upstream and downstream OH results
for both weeks of the verification test. In each case, the mean and standard deviation, maximum,
minimum, and median of the percent recovery values are shown. The outliers identified in
Section 4.2.1 were excluded from this calculation.
Table 4-4. Percent Recovery of Total Mercury Injected into the RKIS
Period
Parameter
Upstream Recovery (%)
Downstream Recovery (%)
Week One
Mean (Std. Dev.)
112.2 (13.1)
111.3 (8.7)
Maximum
137.7
122.2
Minimum
100.8
99.5
Median
105.0
113.0
Week Two
Mean (Std. Dev.)
110.7 (14.2)
112.1 (14.4)
Maximum
134.4
136.2
Minimum
91.2
87.3
Median
108.4
110.8
Table 4-4 shows that the mass balance for mercury injected into the RKIS was good. Over the
entire verification test, the minimum mass balance value was 87.3% and the maximum was
137.7%. For both sampling locations in both weeks of testing, the average mass balance was
about 111 to 112%. These results confirm that mercury injected into the RKIS was not lost, and
in particular confirm that there was no significant difference in mercury levels between the
upstream and downstream sampling locations. Table 4-4 does indicate that the OH results were
usually higher than the expected mercury level by about 10 to 12%, on average.
28
-------
4.4 Audits
4.4.1 Technical Systems Audit
Battelle's Quality Manager performed an internal technical systems audit (TSA) of the verifica-
tion test on January 16 and 17, 2001, during the first week of testing. The TSA ensures that the
verification test is conducted according to the test/QA plan(1) and that all activities associated
with the test are in compliance with the AMS Center QMP.(4) On January 16 the Battelle Quality
Manager visited the RKIS test site, where he toured the test area, observed the performance of
OH method sampling, reviewed Battelle notebooks and calibration gas certificates, reviewed
facility CEM calibration records, and met with the ARCADIS Quality Manager. On January 17,
the Battelle Quality Manager visited Oxford Laboratories and reviewed the OH analysis pro-
cedures in use there. All the observations of the Battelle Quality Manager were documented in a
TSA report. There were no findings of any issues that could adversely affect verification data
quality. The records concerning the TSA are permanently stored with the Battelle Quality
Manager.
In addition to the internal TSA performed by Battelle's Quality Manager, an external TSA was
conducted by EPA on January 23, 2001. The EPA external TSA included all the components of
the internal TSA, except that EPA QA staff did not visit Oxford Laboratories. The findings of
this TSA were documented in a report submitted to the Battelle Quality Manager. No adverse
findings were noted in this external TSA.
4.4.2 Performance Evaluation Audits
A series of PE audits was conducted on several different measurements at the RKIS to assess the
quality of the measurements made in the verification test. These audits were performed by
Battelle staff and were carried out with the cooperation of EPA and ARCADIS staff. These
audits addressed only those measurements that factored into the data used for verification. Each
PE audit was performed by analyzing a standard or comparing to a reference independent of
standards used during the testing (see Section 3.4.2). Each PE audit procedure was performed
once during the verification test, with the exception that OH trains were spiked once in each
week of testing. Table 4-5 summarizes the PE audits on several measurement devices at the
RKIS; Table 4-6 summarizes the PE audit results from spiking OH trains.
Table 4-5 shows that all the PE audit results on measurement devices were well within the
required tolerances stated in the test/QA plan.(1) The PE audit for oxygen and carbon dioxide was
conducted by sampling the same gas entering the facility oxygen and carbon dioxide CEMs,
using a portable monitor for those gases. This was done using a "T" fitting at the inlet of each
facility CEM, so that the readings from the portable audit monitor and the CEMs were obtained
in simultaneous sampling of the flue gas. Duct temperature measurement was audited by
inserting a calibrated thermocouple into the same location in the duct as the temperature probe
used for Method 2 velocity measurements. The small temperature difference observed probably
resulted from the inability to place the two probes at exactly the same location in the small duct.
29
-------
Table 4-5. Summary of PE Audits on Mercury CEM Verifications
Measurement
Audited
Date
Audit Method
Observed
Agreement
Required
Agreement
Flue gas 02
1/19/01
Comparison to independent 02
measurement
0.1% 02
1% 02
Flue gas C02
1/19/01
Comparison to independent C02
measurement
5% of C02
reading
10% of C02
reading
Flue gas
temperature
1/18/01
Comparison to independent
temperature measurement
1.0%
absolute T
2%
absolute T
Local barometric
pressure
1/15/01
Comparison to independent
barometric pressure measurement
0.43 " H20
0.5 " H20
Flue gas pressure
1/19/01
Comparison to independent
pressure measurement
0.02 " H20
0.5 " H20
Impinger weights
(electronic
balance)
1/18/01
Weighing certified weights
0.03%
(0.14 g at
500 g)
Larger of 1%
or 0.5 g
Barometric pressure was audited using a calibrated aneroid barometer. Flue gas pressures were
audited using a Magnehelic gauge installed in parallel with a set of similar gauges used for the
Method 2 velocity determinations. Gauges with ranges of 0 to 0.25, 0 to 0.5, and 0 to 2 inches of
water were audited. The PE audit of the electronic balance used certified weights of 200 and
500 grams; the observed agreement shown in Table 4-5 is for the 500-gram weight, which
showed the greater percentage deviation.
The PE audit of the OH train mercury analysis was conducted once in each week of the test, and
the results are summarized in Table 4-6. In the first week, impingers 1 (KC1), 4 (H202/HN03),
and 5 (KMn04/H2S04) of a blank OH train (I-LB-ST-1) were each spiked with 1 mL of a NIST-
traceable solution containing 10 |ig/mL of mercury. In the second week, the same impingers of a
blank train (II-LB-ST-2) were each spiked with 3 mL of the same solution. As Table 4-6 shows,
the recovery of the spiked mercury was well within the 10% tolerance stated in the test/QA
plan(1) for all spiked samples except the potassium permanganate impinger from the first spiked
train. These results indicate that the OH analysis was accurate at spiked mercury levels compar-
able to those collected in the actual OH sampling. The high readings obtained in analysis of the
potassium permaganate impinger from the first spiked train may have been caused by the need to
spike that train twice. That is, train I-LB-ST-1 was originally spiked on January 18 and stored
30
-------
Table 4-6. Results of PE Audit of OH Train Analysis
Analyses
Date
Train
Impinger"
Number
Spikeb
(Pg)
1st
(Pg)
2nd
(Pg)
Average
(Pg)
Observed
Agreement0
(%)
Required
Agreement
(%)
1/19/01
I-LB-ST-1
1
10
9.75
9.53
9.64
-3.6
10
4
10
9.85
9.61
9.73
-2.7
10
5
10
11.6
12.0
11.8
18
10
1/24/01
II-LB-ST-2
1
30
29.6
29.9
29.75
-0.8
10
4
30
29.9
30.0
29.95
-0.2
10
5
30
28.7
28.5
28.6
-4.7
10
* Impinger 1 = KC1, 4 = H202, 5 = KMn04.
b Amount of mercury injected based on dilution of NIST-traceable standard.
c Observed Agreement = (Average analysis - Spike)/Spike x 100.
-------
overnight for recovery. However, inadvertent pressurization caused a spillover between
impingers 4 and 5 overnight. As a result, the train was emptied, rinsed, refilled, and spiked on
January 19, as indicated in Table 4-5. The higher-than-expected mercury level in impinger 5 may
have resulted from contamination during this process.
4.4.3 Data Quality Audit
Battelle's Quality Manager audited at least 10% of the verification data acquired in the verifica-
tion test. The Quality Manager traced the data from initial acquisition, through reduction and
statistical comparisons, to final reporting. All calculations performed on the data undergoing
audit were checked.
32
-------
Chapter 5
Statistical Methods
The following statistical methods were used to reduce and generate results for the performance
factors.
5.1 Relative Accuracy
The relative accuracy (RA) of the MS-l/DM-5 with respect to the elemental and total mercury
results of the reference (OH) method was assessed by the following equation, which is incorpo-
rated in EPA performance specifications for continuous emission monitoring systems:
\dI + Ct ~
R l = X 100% (1,
X
where drefers to the arithmetic difference between corresponding OH and MS-l/DM-5 results,
and x corresponds to the OH result. Sd denotes the sample standard deviation of the differences,
while tan_j is the t value for the 100(1 - a)th percentile of the distribution with n-1 degrees of
freedom. To calculate RA, the OH and corresponding MS-l/DM-5 results were paired, and the
differences between the paired results were calculated. Then, the absolute mean {d) and standard
deviation (Sd) of those differences were calculated. The mean of the OH results ^x^was
calculated, and the value of was taken from appropriate tables for the relevant values of n
and a. The RA was determined for an a value of 0.025 (i.e., 97.5% confidence level, one-tailed).
RA was calculated separately for the first and second week of testing. The OH results used for
the RA calculation were as stated in Section 4.2.1.
5.2 Correlation with Reference Method
The degree of correlation of the MS-l/DM-5 with the reference method results was assessed in
terms of the coefficient of determination (r2), which is the square of the correlation coefficient (r).
This calculation was made separately for the first and second week of testing, and also for the full
data set from both weeks of testing.
33
-------
5.3 Precision
Precision was calculated in terms of the percent relative standard deviation (RSD) of a series of
CEM measurements made during stable operation of the RKIS, with mercury injected at a
constant level into the combustion zone. During each OH sampling run, all readings from the
MS-l/DM-5 were recorded, and the mean and standard deviations of those readings were
calculated. Precision (P) was determined as
S
P = = xl00 (
X {
where S is the standard deviation of the readings and X is the mean of the readings.
5.4 Calibration/Zero Drift
Calibration and zero drift were determined in a relative sense, rather than as deviations from an
absolute standard, using the elemental mercury gas standards and high-purity nitrogen as zero
gas. In the first week, nine elemental mercury standard readings, and nine zero readings, were
used for this calculation. In the second week, seven readings of each type were available for this
calculation. Drift was calculated in terms of the RSD, as
RSD = 4 * 100 (3)
x
where x is the mean, and S the standard deviation, of the daily readings on standard or zero gas.
This calculation, along with the range of the data, indicated the day-to-day variation in zero and
standard readings.
5.5 Sampling System Bias
Sampling system bias was calculated as the difference in MS-l/DM-5 response when sampling
elemental mercury standard gas through the entire sample interface, compared to that when
sampling the same gas directly at the analyzer, expressed as a percentage of the response at the
analyzer. That is,
B = Rn - Ra xJQQ
R
a
where B is the percent bias, Rsj is the reading when the standard gas is supplied at the sampling
inlet, and Ra is the reading when the standard is supplied to the analyzer.
34
-------
5.6 Interferences
Interferences were determined during sampling of combustion flue gas, in terms of the difference
in response to a constant mercury level when potential interferant gases were added or removed.
Interferences were assessed in terms of the effect of the interferant species (N0X, CO, S02, HC1,
Cl2) introduced alone or together into the flue gas. The interferant levels were established by
means of the facility CEM responses for each interferant, as described in Section 3.2.1.
5.7 Response Time
The response time was determined as the time after a step change in mercury concentration when
the MS-l/DM-5 reading reached a level equal to 95% of that step change. Both rise time and fall
time were determined. The MS-l/DM-5 response times were determined in conjunction with a
calibration/zero drift check by starting or stopping delivery of the elemental mercury standard gas
to the DM-5 monitor inlets or the entire inlet system, recording readings until stable readings
were obtained, and then estimating the 95% response time.
5.8 Low-Level Response
The ability of the MS-l/DM-5 to determine low mercury concentrations was assessed by compar-
ing responses at successive nominal total mercury levels of 0, 0.57, 1.13, 2.27, and 4.54 (ig/m3 in
the RKIS flue gas. This test was conducted in a flue gas matrix containing elevated levels of
sulfur dioxide, nitrogen oxides, and hydrogen chloride, as described in Section 3.3.8. The lowest
mercury level producing a response above that with no mercury added is of interest in this test.
35
-------
Chapter 6
Test Results
The results of the verification test of the Nippon AM-2 mercury CEM are presented in this
section, based upon the statistical methods shown in Chapter 5.
6.1 Relative Accuracy
Table 6-1 shows the elemental, oxidized, and total mercury results from the Nippon MS-l/DM-5
during the period of each OH sampling run. The MS-l/DM-5 results were calculated by subtract-
ing from each data point the daily average zero gas reading of the MS-l/DM-5 (from the calibra-
tion/zero drift results in Section 6.4) and then averaging the data over the period of the OH run.
The zero-corrected and averaged results are shown in Table 6-1 and may be compared to the OH
results in Tables 4-1 and 4-2.
In general, the MS-l/DM-5 results were quite similar to those of the OH method in the first week
of testing, but not during the second week. Comparison of Tables 4-1 and 4-2 with Table 6-1
shows that, in the second week, the MS-l/DM-5 indicated elemental mercury levels that were
quite uniform from day to day and that usually exceeded the OH results by a substantial margin.
On the other hand, the oxidized mercury results from the CEM in the second week were always
substantially below those from the OH method. The total mercury results followed the same
pattern because of the predominance of oxidized mercury in the duct in that week. These results
suggest that the MS-l/DM-5 was unable to properly handle the high proportion and large con-
centration of oxidized mercury in the second week of testing. Given the difficulty of sampling
oxidized mercury, it is possible that this disagreement may be due to loss of oxidized mercury in
the inlet system of the MS-l/DM-5.
The results of the calculation of RA for the MS-l/DM-5 are shown in Table 6-2, which provides
separate relative accuracy results for the two weeks of testing for measuring elemental, oxidized,
and total vapor-phase mercury by the MS-l/DM-5. The RA results reflect the observations noted
above. The similar RAs of about 11 and 13% for elemental and total mercury in the first week
reflect the predominance of the elemental fraction in that period. An RA of 11% indicates that
the MS-l/DM-5 agreed with the OH results for elemental mercury within less than 1 (ig/m3, at
the flue gas levels of approximately 6 to 7 (ig/m3. Note that the bias in transfer of elemental
mercury through the MS-l/DM-5 inlet was about -7% (Section 6.5). Thus, a large portion of the
RA in week one can be accounted for by the 7% bias in the sampling system. The RA for
oxidized mercury in that week reflects the fact that the CEM results were generally below
1 (ig/m3, whereas the OH results were generally 1 to 1.7 (ig/m3.
36
-------
The results in Table 6-2 for the second week reflect opposing trends: the MS-l/DM-5 con-
sistently read higher than the OH results for elemental mercury and consistently lower for
oxidized mercury. The RA of about 50% for elemental mercury in week two is surprising, given
the 11% RA in week one and the fact that elemental mercury concentrations in week two were
often higher than those in week one. This RA for elemental mercury may indicate that the MS-1
speciation unit had difficulty separating elemental and oxidized mercury at the relatively high
oxidized concentrations used in the second week. The RA values for oxidized and total mercury
in the second week reflect the consistently low readings obtained from the CEM relative to the
OH results and suggest loss of mercury in the CEM inlet. One possible factor could be the
difference in flow rates over the particle filters, i.e., the OH method used a flow of 16.6 L/min,
and a new filter was used for each run; whereas the Nippon system used a flow of 1 L/min, and
the filter was changed after the first week of testing.
Table 6-1. Average Mercury Results from Nippon MS-l/DM-5 During OH Sampling Runs
MS-l/DM-5 Results (ug/m3)
OH Run ——-
Date Number EMa OMa TMa
January 15
1
6.04
0.68
6.72
2
6.09
0.57
6.66
January 16
3
5.75
0.55
6.30
4
5.29
1.54
6.83
January 17
5
6.13
0.47
6.60
6
6.16
0.69
6.85
January 22
7
14.1
39.1
53.2
8
14.3
42.9
57.2
9
18.8
38.4
57.2
January 23
10
11.2
46.4
57.6
11
12.9
49.1
62.0
12
12.1
53.6
65.7
January 24
13
13.4
49.6
63.0
14
14.5
51.7
66.2
15
13.4
57.0
70.4
a EM = elemental Hg, OM = oxidized vapor-phase Hg, TM = total vapor-phase Hg.
37
-------
Table 6-2. Relative Accuracy Results for the Nippon MS-l/DM-5
Relative Accuracy
(%)
Test Period
Elemental Hg
Oxidized Hg
Total Vapor Hg
Week One
11.0
54.9
13.2
Week Two
50.4
49.1
39.1
6.2 Correlation
The correlation of the MS-l/DM-5 monitor with the OH reference method was calculated using
the elemental, oxidized, and total mercury results shown in Table 6-1. Correlation plots of the
MS-l/DM-5 and OH data from the first and second weeks of testing, and from both weeks
combined, are shown in Figures 6-la to c, respectively. These figures also show the linear
regression results and correlation coefficients (r) for elemental, oxidized, and total vapor-phase
mercury. The resulting coefficients of determination (r2) for each of these types of data for each
week of testing, and for both weeks combined, are shown in Table 6-3.
The results in Table 6-3 show that r2 values for all three mercury fractions were 0.36 or less in
the first week of testing. The r2 values for elemental and oxidized mercury in the first week were
derived from negative values of r (i.e., anticorrelation between the OH and MS-l/DM-5 results).
These low values are undoubtedly due in part to the limited range of mercury concentrations
present in that first week. In the second week, the r2 value for elemental mercury was highest and
that for total mercury the lowest. The data set from both weeks combined is the largest and has
the widest range of data values, and consequently is most suitable for calculation of correlation.
When data from both weeks were combined, r2 values exceeding 0.935 were found for oxidized
and total mercury, but r2 for elemental mercury was 0.417.
Table 6-3. Correlation of Nippon MS-l/DM-5 Data with OH Results
Test Period
Coefficient of Determination (r2)
Elemental Hg
Oxidized Hg
Total Vapor Hg
Week One
0.362a
0.009a
0.083
Week Two
0.735
0.464
0.153
Both Weeks
0.417
0.937
0.938
aThese r2 values are based on negative values of r.
38
-------
8
^ Elemental Hg, y = -0.447x + 8.74, r = -0.601
O Oxidized Hg, y = -0.168x + 0.961, r = -0.096
A Total Hg, y = 0.198x + 5.16, r = 0.287
7
6
CO
E
"5) c
3 5
CO
3
w
<1> „
a: 4
Q
5? 3
-------
90
^Elemental Hg, y = 0.543x + 5.92, r = 0.646
O Oxidized Hg, y = 0.535x + 2.00, r = 0.968
ATotal Hg, y = 0.607x + 3.92, r = 0.968
80
70
AA
60
AJD
TO
U)
50
oo
CD
O
o o
w
40
0
-------
Table 6-4. Precision Results for the Nippon MS-l/DM-5
Elemental Mercury Oxidized Mercury Total Vapor Mercury
Mean
Mean
OH Run
(Std.
%
(Std.
Mean
Date
Number
Dev.)
RSD
Dev.)
% RSD
(Std. Dev.)
% RSD
1/15
1
6.04
(0.46)
7.6
0.68
(0.08)
12.0
6.7
(0.47)
6.9
2
6.09
(0.21)
3.5
0.56
(0.17)
30.3
6.7
(0.28)
4.2
1/16
3
5.75
(0.18)
3.2
0.56
(0.25)
45.3
6.3
(0.28)
4.4
4
5.28
(0.40)
7.7
1.52
(0.19)
12.4
6.8
(0.47)
6.9
1/17
5
6.12
(0.24)
4.0
0.46
(0.08)
16.8
6.6
(0.25)
3.7
6
6.16
(0.24)
3.8
0.69
(0.08)
12.4
6.8
(0.26)
3.8
1/22
7
14.1
(1.59)
11.3
39.0
(6.56)
16.8
53.1
(6.80)
12.8
8
14.2a
(0.88)
6.2
39.9a
(5.43)
13.6
54.0a
(5.51)
10.2
9
14.la
(5.76)
41.0
35.7a
(7.35)
20.6
49.8a
(11.9)
23.9
1/23
10
11.2
(0.76)
6.8
46.3
(5.51)
11.9
57.4
(5.76)
10.0
11
12.3a
(1.23)
10.0
49.0
(6.32)
12.9
61.3a
(6.67)
10.9
12
12.1
(1.22)
10.1
53.5
(8.27)
15.5
65.5
(8.81)
13.5
1/24
13
13.3a
(1.39)
10.4
49.la
(5.38)
11.0
62.4a
(5.88)
9.4
14
14.5
(1.24)
8.6
51.5
(5.89)
11.4
66.0
(6.33)
9.6
15
13.5
(1.21)
9.0
57.2
(6.39)
11.2
70.7
(6.73)
9.5
a Brief spikes in data removed prior to calculation of precision.
41
-------
concentrations with a high proportion of oxidized mercury in the second week. For total mercury,
the MS-l/DM-5 exhibited precision within 15% RSD in 14 of the 15 OH sampling periods and
within 10% in 10 of the periods. The results in Table 6-4 indicate not only that the MS-l/DM-5
exhibited good precision, but that in most cases the RKIS facility maintained highly stable
mercury levels. The precision results for OH Run 9 in Table 6-4 exclude large spikes that were
caused by the failure of the mercury solution delivery pump. The results for that run may still
include some increased variability caused by the impending failure of that pump. The
MS-l/DM-5 showed a gradual increase in oxidized mercury throughout OH Run 3 that resulted
in a relatively large percent RSD for the oxidized mercury that could not be traced to any
occurrence in the test facility.
6.4 Calibration/Zero Drift
The daily MS-l/DM-5 readings on zero gas and elemental mercury standard gas are listed in
Table 6-5. The data from the two weeks of testing are listed separately because of the different
standard gases used. Table 6-6 summarizes the calibration/zero drift results in terms of the mean,
standard deviation, percent RSD, maximum, and minimum of the zero and standard gas
responses in each week. The first analysis of the mercury standard gas in the first week was
excluded from the drift calculation because, as described in Section 3.4.1, the materials and
procedure for delivering that standard did not allow proper equilibration in the first run.
Tables 6-5 and 6-6 show that the MS-l/DM-5 gave a stable response to both zero and standard
gases in the test. In particular, response to the low-level elemental mercury standard gas showed
a 2.9% RSD with both DM-5 detectors in the first week of testing, and response to the higher
level standard gas showed percent RSD values of less than 1% in the second week. These
observations show the stability of the MS-l/DM-5 and also support the utility of the elemental
mercury standards for confirmation of CEM stability. The mercury concentrations in the gas
standards are not known absolutely, but the average mercury concentrations reported by the
MS-l/DM-5 for the two standard gases (15.1 to 15.2 and 55.0 to 55.6 (ig/m3, respectively) are
within about 8% and 13%, respectively, of the concentrations of 14 and 49 (ig/m3, based on
analysis by the gas vendor (Table 3-8, Section 3.4.1).
6.5 Sampling System Bias
Table 6-7 shows the results of providing the elemental mercury gas standards directly to the
DM-5 detectors of the CEM, and at the inlet of the Nippon sampling probe, on January 19 and
25, 2001. Shown are the final stable readings on the zero gas and the mercury standard for both
DM-5 detectors of the CEM. The resulting bias in transport of the elemental mercury standard
42
-------
Table 6-5. Zero and Standard Gas Responses of the Nippon MS-l/DM-5
Elemental Mercury Response"
Oxidized Mercury Responseb
Date
Zero Gas
(pg/m3)
Standard Gas
(pg/m3)
Zero Gas
(pg/m3)
Standard Gas
(pg/m3)
Week One
1/15/01
0.1
11.5C
0.0
11.4C
0.0
15.3
0.0
15.2
1/16/01
0.0
14.9
0.0
14.8
0.0
15.0
0.1
14.8
1/17/01
0.1
16.1
0.1
16.0
0.1
15.0
0.1
14.8
1/18/01
0.0
14.9
0.1
14.8
0.0
15.4
0.1
15.3
1/19/01
-0.1
14.7
0.0
14.7
Week Two
1/22/01
0.0
55.4
0.1
54.8
0.0
55.9
0.1
55.4
1/23/01
0.1
55.3
0.1
54.5
0.0
55.7
0.1
55.0
1/24/01
0.1
55.2
0.1
55.3
0.1
55.7
0.1
55.3
1/25/01
0.0
55.7
0.1
55.0
a DM-1 mercury detector Serial No. W341003.
b DM-1 mercury detector Serial No. W341004.
c This value excluded from calculation of drift; delivery procedure for elemental mercury standard gas not
properly equilibrated.
43
-------
Table 6-6. Summary of Calibration /Zero Drift Results for the Nippon MS-l/DM-5
Elemental Mercury Response® Oxidized Mercury
Responseb
Result
Zero Gas
(pg/m3)
Standard Gas
(pg/m3)
Zero Gas
(pg/m3)
Standard Gas
(pg/m3)
Week One
Mean
0.02
15.2
0.06
15.1
Std. Dev.
0.07
0.44
0.05
0.44
% RSD
NA
2.9
NA
2.9
Maximum
0.1
16.1
0.1
16.0
Minimum
-0.1
14.7
0.0
14.7
Week Two
Mean
0.04
55.6
0.10
55.0
Std. Dev.
0.05
0.26
0.00
0.32
% RSD
NA
0.5
NA
0.6
Maximum
0.1
55.9
0.1
55.4
Minimum
0.0
55.2
0.1
54.5
aDM-l mercury detector Serial No. W341003.
bDM-l mercury detector Serial No. W341004.
through the inlet system is also shown in each case. Note that when elemental mercury passes
through the inlet system of the MS-l/DM-5, no response should result in the oxidized mercury
channel of the CEM (i.e., a bias of -100% shold be observed) since the purpose of the MS-1
speciation module is to separate the elemental and oxidized forms of mercury.
Table 6-7 shows that the zero gas readings of both DM-5 detectors were slightly elevated when
sampling through the entire inlet system, suggesting the retention of some flue gas mercury in the
inlet. When the zero-corrected responses were compared, the bias in transport of elemental
mercury to the MS-1 speciation unit and on to the DM-5 detectors was -7.4% with the lower
level standard and -6.8% with the higher level standard. As expected, the MS-1 unit effectively
separated the elemental mercury from the gas flow supplied to the oxidized mercury detector of
the CEM: the bias in transport of the two standards to the oxidized mercury detector was close to
the expected -100% in both cases.
44
-------
Table 6-7. Results of Sampling System Bias Test of the Nippon MS-l/DM-5
Elemental Mercury® Oxidized Mercuryb
Response Response
Gas Supplied
Date To:
Zero Gas
(pg/m3)
Standard Gas
(pg/m3)
Zero Gas
(pg/m3)
Standard Gas
(pg/m3)
1/19/01 Analyzer
-0.1
14.7
0.0
14.7
Probe Inlet
0.5
14.2
0.3
0.4
Sampling System Bias
for Hg°
-7.4%c
-99.3%
1/25/01 Analyzer
0.0
55.7
0.1
55.0
Probe Inlet
1.1
53.0
0.8
0.9
Sampling System Bias
for Hg°
-6.8%
-99.8%
a DM-1 mercury detector Serial No. W341003.
b DM-1 mercury detector Serial No. W341004.
c Bias calculated as described in Section 5.5, using zero-corrected standard response
through inlet, relative to zero-corrected standard response at analyzer.
6.6 Interference Effects
The effect of interferences was tested by injecting a constant level of mercury into the RKIS and
altering the levels of several interferants, as described in Section 3.3.6. Figures 6-2 and 6-3
illustrate the response of the MS-l/DM-5 during this test on January 18. Figure 6-2 shows the
elemental and oxidized mercury readings of the CEM during this test; Figure 6-3 shows the total
mercury response, which is simply the sum of the elemental and oxidized readings. The dashed
vertical lines indicate the time periods in which each of the indicated interferants or combinations
of interferants was introduced.
Figure 6-2 shows that 500 ppm nitric oxide, 500 ppm carbon monoxide, and 2,000 ppm sulfur
dioxide had no significant effect on the elemental or oxidized mercury readings of the
MS-l/DM-5. Adding 250 ppm hydrogen chloride to the duct caused a decrease in elemental
mercury response of about 25%, from about 7 to 5.5 (ig/m3. Stabilization of the hydrogen
chloride concentration at the target level took several minutes, and this is reflected in the
elemental mercury readings in this period, which decreased gradually throughout this period.
However, the oxidized mercury readings of the MS-l/DM-5 were unaffected by the added
hydrogen chloride, showing no increase corresponding to the decrease in elemental mercury.
Replacing the hydrogen chloride with 10 ppm of chlorine had about the same effect on the
45
-------
Hg only NO co so2 HCI cl2 N
500 ppm 500 ppm 2,000 ppm 250 ppm 10 PPm
3 + Cl2 AN
Hg only
|
I
Elemental Hg a k ji jJ
JL
t r
r,
'W
ilf/lj
lyr w 1
\
f\ 11 1 1
1
VMvvJ" |
I
12:00:00 13:12:00 14:24:00 15:36:00 16:48:00 18:00:00
Time of Day
Figure 6-2. Nippon MS-l/DM-5 Elemental and Oxidized Mercury Response
in Interference Test, January 18, 2001
Hg only NO CO
SO,
HCI
500 ppm 500 ppm 2,000 ppm 250 pnm
Cl2
10 ppm
3 + CI
'2 All
Hg only
E
o>
12:00:00 13:12:00
14:24:00 15:36:00
Time of Day
16:48:00 18:00:00
Figure 6-3. Nippon MS-l/DM-5 Total Mercury Response in Interference
Test, January 18, 2001
46
-------
elemental mercury readings as did the hydrogen chloride, but the presence of the chlorine caused
a sharp short-term increase in the oxidized mercury readings. Adding nitric oxide along with the
chlorine caused the elemental mercury readings to rise toward the level seen with only mercury
present, while the oxidized mercury readings continued a gradual return to the original levels.
When all the interferants were added at once, the MS-l/DM-5 readings were again close to those
with only mercury present. The MS-l/DM-5 responses with all interferants present show that the
CEM is not subject to serious interferences from high levels of several key pollutants in
combustion emissions. Note that the spikes seen around 17:40 in Figures 6-2 and 6-3 result from
an inadvertent large injection of chlorine into the duct for a brief period.
Figure 6-3 shows that the total mercury response was stable throughout most of the interference
test. The only exceptions are the gradual decrease in total mercury seen when hydrogen chloride
was added and the large short-term increase in total mercury when chlorine was added. These
two cases are somewhat contrary, since, in the former case, the oxidized mercury showed no
increase, while elemental mercury decreased significantly, and in the latter case, oxidized
mercury increased greatly even though elemental mercury was only reduced slightly. However,
the portion of the test with all interferants present at once shows that the MS-l/DM-5 measure-
ments of total mercury were essentially unaffected by a mixture of several interferant gases at
elevated levels.
6.7 Response Time
The rise and fall times of the MS-l/DM-5 response were tested as part of the zero and standard
gas checks done to assess calibration and zero drift (Section 6.4). Because of the rapid response
of the MS-l/DM-5, response time could be assessed using data from most of those checks. The
only exceptions were instances in which the rise or fall of the CEM response was interrupted by
the CEM's autozeroing procedure, by incomplete data acquisition, or when problems occurred in
delivery of the standard gas to the DM-5 detectors. Regarding the latter point, the MS-l/DM-5
was sensitive to the pressure at which the standard gas was delivered to the analyzers, so care
was taken to vent a minimal excess (i.e., approximately 100 mL/min) of the standard gas to avoid
overpressurizing the line carrying the standard to the DM-5s. In some span checks, adjustment of
the vent flow was needed, and no response time could be calculated.
Table 6-8 summarizes the results of the response time tests, listing the MS-l/DM-5 rise and fall
times in checks conducted both with standard gas introduced directly at the DM-5 detectors and
through the entire inlet probe. These values were calculated by interpolating between 10-second
readings from the two channels of the CEM to estimate the occurrence of the 95% rise or fall in
response.
Table 6-8 shows that the typical rise time of the DM-5 detectors was about 50 seconds, with
considerable variation about that value. The typical fall time was about 35 seconds, with
relatively less variation about that value. The reason for the longer rise time, and greater
variability in the rise time, may be that the rise time determinations were subject to equilibration
of the standard gas delivery. That is, the rise times are a composite of the response time of the
CEM and the rate at which the elemental mercury standard gas achieved equilibration in the
47
-------
Table 6-8. Results of Response Time Tests on the Nippon MS-l/DM-5
Response Time (seconds)
Rise/Fall Time
Date
Elemental Mercury® Oxidized Mercuryb
Rise Time
Fall time
1/15/01 a.m.
1/17/01 a.m.
1/22/01 a.m.
1/22/01 p.m.
1/23/01 a.m.
1/24/01 a.m.
1/25/01 p.m.
1/25/01 p.m.c
1/15/01 a.m.
1/15/01 p.m.
1/16/01 p.m.
1/17/01 a.m.
1/17/01 p.m.
1/18/01 p.m.
1/19/01 a.m.
1/23/01 a.m.
1/23/01 p.m.
1/24/01 a.m.
1/24/01 p.m.
1/25/01 p.m.
1/25/01 p.m.
50
29
71
64
69
55
42
46c
36
28
37
36
36
30
31
29
39
37
38
37
42
55
29
58
67
53
44
29
30
39
37
37
30
34
28
39
36
38
49
a DM-1 mercury detector Serial No. W341003.
b DM-1 mercury detector Serial No. W341004.
c Standard supplied to CEM inlet probe, all other results are for standard supplied directly to DM-5 analyzers.
48
-------
delivery plumbing. This effect does not apply to the fall times, since the delivery plumbing was
simply disconnected once the zero/span procedure was completed. Overall, Table 6-8 shows that
the rise and fall times of the MS-l/DM-5 were about 50 and 35 seconds, respectively, with no
significant difference between the response times of the two DM-5 detectors.
6.8 Low-Level Response
Figure 6-4 illustrates the response of the MS-l/DM-5 to low levels of mercury injected into the
duct, as described in Section 3.3.8. The dashed vertical lines indicate the time periods during
which each nominal total mercury level was introduced. Both the elemental and oxidized
mercury responses of the MS-l/DM-5 are shown in Figure 6-4. The periods of autozeroing by the
MS-l/DM-5 are apparent in the data.
6.0
5.0
4.0
CO
£
3 3.0
u>
I
2.0
1.0
0.0
11:22:10 12:22:31 13:23:00 14:23:10
Time of Day
Figure 6-4. Low-Level Mercury Response of the Nippon MS-l/DM-5, January 19, 2001
Figure 6-4 shows that the MS-l/DM-5 responded to all mercury concentrations injected, produc-
ing an increase above the flue gas background even with the injection of as little as 0.57 (ig/m3 of
mercury. The added mercury was detected almost exclusively in the form of elemental mercury,
with oxidized mercury readings of the CEM increasing only slightly in the period of highest
mercury addition. The primary observation from these data is that the MS-l/DM-5 is clearly
sensitive enough to detect mercury at levels below 1 (ig/m3. However, the increases in response
were not entirely quantitative. For example, with a nominal added mercury concentration of
4.54 (ig/m3 in the duct, the MS-l/DM-5 gave a total mercury response about 3 (ig/m3 above flue
gas background. When the MS-l/DM-5 CEM responses at each mercury level are averaged, a
regression of average MS-l/DM-5 response vs. nominal mercury level gives the equation
49
4.54 ug/m3
2.27 ug/m3
No Hg
added
No Hg
added
0.57 ug/m3
1.13 ug/m3
-------
CEM = 0.65 x (Hg, |ig/m3) + 0.68 (ig/m3, with r2 = 0.999. These results could mean that some of
the injected mercury was lost in the duct or that the MS-l/DM-5 response is low, perhaps due to
loss of mercury in the MS-l/DM-5 inlet. Mass balance calculations of the mercury injected
during OH runs indicate that mercury was not lost in the duct. Therefore, the data in Figure 6-4
indicate a somewhat low response from the MS-l/DM-5 at the low mercury levels present in this
test.
6.9 Data Completeness
The MS-l/DM-5 operated reliably throughout all of the verification test procedures, and no test
data were lost as a result of any malfunction or down time. Consequently, the data completeness
was 100%.
6.10 Setup and Maintenance
The MS-l/DM-5 and a second Nippon CEM were set up by two Nippon representatives. Both
those representatives were on site for the first week of testing, and one remained for the second
week. The MS-l/DM-5 was set up and ready to sample flue gas within about a half day after it
was placed at the sampling port. The MS-l/DM-5 setup proceeded smoothly, and operation of
the instrument was nearly trouble-free throughout the test. The single exception resulted from a
reagent solution that was incorrectly made, causing precipitation of the reagent and clogging of
the liquid flow system in the MS-1 unit. This problem was readily corrected, and no data were
lost. The instrument required no external gas supplies, and no scheduled maintenance was
required. Consumables consisted of the reagent solutions used in the MS-1, which resulted in 2
to 3 L/day of waste solutions requiring disposal.
6.11 Cost
As tested, the approximate purchase price of the Nippon MS-l/DM-5 monitor is about $40,000.
50
-------
Chapter 7
Performance Summary
During the first week of verification testing, the Nippon MS-l/DM-5 mercury CEM provided
accuracy relative to the OH method of 13.2% for total mercury, at total mercury levels of about
7 to 8 (ig/m3. Testing showed relative accuracy of 11.0% for elemental mercury, and 54.9% for
oxidized mercury, at elemental mercury levels of approximately 6 to 7 (ig/m3 and oxidized
mercury levels of about 1 to 1.5 (ig/m3. In the second week of verification, the MS-l/DM-5
provided relative accuracy of 39.1% for total mercury, at total mercury levels of about 70 to
120 (ig/m3. Relative accuracy of 50.4% for elemental mercury, and 49.1% for oxidized mercury,
was found in that week. These RA values in week two occurred largely because of over-
estimation of elemental mercury, and underestimation of oxidized mercury, by the MS-l/DM-5
relative to the OH results, at elemental mercury levels ranging from about 5 to 25 (ig/m3 and
oxidized mercury levels ranging from about 45 to 110 (ig/m3.
The coefficient of determination (r2) of the MS-l/DM-5 and OH elemental mercury results was
0.417 based on data from both weeks combined. The corresponding r2 value for oxidized
mercury was 0.937, and for total mercury was 0.938.
Precision of the MS-l/DM-5 response was assessed in periods of stable mercury levels in the flue
gas during the 15 OH sampling periods. The precision (as percent RSD) of the MS-l/DM-5
response for elemental mercury was within 10% in 11 of the 15 periods and within 15% in 14 of
the periods. For oxidized mercury, precision was never within 10% RSD, but nine of the
15 periods showed precision within 15% RSD. For total mercury, precision was within 10% RSD
in 10 of the 15 periods, and within 15% in 14 of the periods. These precision results include both
variability in the test facility and in the MS-l/DM-5.
Calibration and zero drift were determined by repeated analysis of zero gas and elemental
mercury standard gases. Nine such analyses in the first week of verification gave zero gas
responses of 0.02 (± 0.07) and 0.06 (± 0.05) (ig/m3, respectively, for the two DM-5 detectors.
The corresponding standard gas responses were 15.2 (± 0.44) and 15.1 (± 0.44) (ig/m3. The
standard gas results equate to a 2.9% RSD. Seven such analyses in the second week of verifica-
tion gave zero gas readings of 0.04 (± 0.05) (ig/m3 and 0.10 (± 0.00) (ig/m3 for the two DM-5s,
respectively. The corresponding standard gas responses were 55.6 (± 0.26) and 55.0 (± 0.32)
|ig/m3. These standard gas results equate to percent RSD values of 0.5 and 0.6, respectively. The
MS-l/DM-5 exhibited rise and fall times of about 50 and 35 seconds, respectively.
51
-------
Sampling system bias of the inlet system used with the MS-l/DM-5 was determined using
elemental mercury gas standards. The bias in transport of elemental mercury through the inlet
system was approximately -7%.
Elevated levels of sulphur dioxide, nitric oxide, and carbon monoxide had no effect on
MS-l/DM-5 response to elemental or oxidized mercury in flue gas. The presence of hydrogen
chloride reduced elemental mercury readings by about 25%, without a corresponding increase in
the oxidized mercury readings of the MS-l/DM-5. The presence of chlorine reduced elemental
mercury readings by about the same amount as did hydrogen chloride, but also caused a large
increase in the oxidized mercury readings of the CEM. However, when these gases were all
present at once in the flue gas, the MS-l/DM-5 readings were close to those seen with only
mercury in the flue gas, indicating no interference from the combination of these gases.
The MS-l/DM-5 responded to as little as 0.57 (ig/m3 of mercury in flue gas, but the response to
concentrations of 0.57 to 4.5 (ig/m3 averaged only about 65% of the nominal mercury
concentration.
Data completeness for the MS-l/DM-5 was 100%, and no repair or maintenance was needed.
The unit uses about 2 to 3 L/day of aqueous reagents to separate elemental and oxidized mercury.
52
-------
Chapter 8
References
1. Test/QA Plan for Pilot-Scale Verification of Continuous Emission Monitors for Mercury,
Battelle, Columbus, Ohio, November 30, 2000.
2. Standard Test Method for Elemental, Oxidized, Particle-Bound, and Total Mercury in Flue
Gas Generatedfrom Coal-Fired Stationary Sources (Ontario Hydro Method), American
Society for Testing and Materials, Draft Method, October 27, 1999.
3. Test Procedure for the Determination of Elemental Mercury (Fig °) Concentration from
Compressed Gas Standards Using Modified Ontario Hydro Mini-Impinger Trains, prepared
by ARCADIS Geraghty & Miller, Research Triangle Park, North Carolina, for Work
Assignment No. 2-18 of EPA Contract No. 68-C-99-201, January 11, 2001.
4. Quality Management Plan (QMP) for the E TV Advanced Monitoring Systems Pilot,
Version 2.0, U.S. EPA Environmental Technology Verification Program, Battelle,
Columbus, Ohio, October 2000.
5. Belsley, D. A., Kuh, E., and Welsch, R. E., Regression Diagnostics, John Wiley and Sons,
New York, 1980.
53
------- |