United States
Environmental Protoctlen
Agency
Cfflcs of Air Quality
Planning and Standards
Research Trian£le Park, NC 27711
EPA 430/4-92-013
Vdume 1
Decsmbar 9,1991
Evaluation of Two Methods
for the Measurement of Mercury
Emissions in Exhaust Gases from a
Municipal Waste Combustor
-------
DCN: 92-239-026-12-07
EVALUATION OF TWO METHODS
FOR THE MEASUREMENT OF MERCURY EMISSIONS IN
EXHAUST GASES FROM A MUNICIPAL WASTE COMBUSTOR
March 1992
Prepared under:
EPA Contract Nos. 68-D10010
68-D90054
68-D10031
U.S. Enviroftmsnts! Protection Agency
»•''•' ''-..1 : >, : ' :- • " " - •
7; •-.-.'-
Prepared for:
Foston Curtis
Emission Measurement Branch (MD-19)
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
CONTENTS
Figures v
Tables vi
1.0 INTRODUCTION 1
1.1 Background 1
1.2 Test Design 2
1.2.1 Plant Description and Operation ... 2
1.2.2 Test Matrix 2
1.3 Conclusions 2
1.3.1 Method Precision for Mercury .... 4
1.3.2 Method Comparison 4
1.3.3 Method Precision for Cadmium and
Lead 4
1.4 Report Organization 5
2.0 TEST DESIGN 6
2.1 Method Validation Requirements 6
2.2 Facility Description 7
2.2.1 Overview 7
2.2.2 Sampling Locations 10
2.3 Sampling Matrix 10
2.4 Description of Method 101A 12
2.4.1 Background 12
2.4.2 Sampling System 12
2.5 Multiple Metals Sampling 14
2.5.1 Background 14
2.5.2 Sampling System 18
2.6 Parameters Measured 22
2.6.1 Volumetric Flow Rate Determination
by EPA Method 2 22
2.6.2 Oxygen and Carbon Dioxide
Concentrations by EPA Method 3A . . . 22
2.6.3 Average Moisture Determination by
EPA Method 4 23
3.0 METHOD 101A RESULTS 24
3.1 Review of Data 24
3.2 Statistical Analysis 26
3.3 Conclusions 29
epp.053
11
-------
CONTENTS, Continued
4.0 MULTIPLE METALS METHOD RESULTS 30
4.1 Review of Data 30
4.1.1 Mercury 30
4.1.2 Cadmium and Lead 32
4.2 Statistical Analysis 37
4.2.1 Mercury 37
4.2.2 Cadmium and Lead 37
4.3 Conclusions 39
4.3.1 Mercury 39
4.3.2 Cadmium and Lead 39
5.0 INTER-METHOD COMPARISON FOR MERCURY 40
5.1 Statistical Analysis 40
5.1.1 Precision 40
5.1.2 Inter-method Difference in
Measured Mercury Concentrations ... 43
5.2 Conclusions 49
6.0 SAMPLING AND ANALYTICAL PROCEDURES 50
6.1 Method 101A 50
6.1.1 Sampling Equipment 51
6.1.2 Equipment Preparation 51
6.1.3 Reagent Preparation 51
6.1.4 Sample Train Operation 52
6.1.5 Sample Recovery 52
6.1.6 Analytical Preparation 53
6.1.7 Analysis 54
6.2 Multiple Metals Method 54
6.2.1 Sampling Equipment 54
6.2.2 Equipment Preparation 55
6.2.3 Reagent Preparation 55
6.2.4 Sample Train Operation 56
6.2.5 Sample Recovery 57
6.2.6 Metals Analytical Procedures .... 59
6.2.7 Mercury Standards and Quality
Control 60
7.0 QUALITY ASSURANCE/QUALITY CONTROL 61
7.1 QA/QC Summary 61
7.2 QA/QC Definitions and Objectives 63
7.3 Manual Flue Gas Sampling and Recovery
Parameters 65
7.3.1 Mercury by Method 101A Sampling
Quality Assurance 65
7.3.2 Multiple Metals Sampling Quality
Assurance 69
epp.053 111
-------
CONTENTS, Continued
7.4 Analytical Quality Assurance 69
7.4.1 Mercury by Method 101A Analytical
Quality Assurance 69
7.4.2 Multiple Metals Analytical Quality
Assurance 72
7.5 Data Variability 76
7.5.1 Overview 76
7.5.2 Test Program Data Variation 78
APPENDICES
Appendix A:
Appendix B:
Appendix C:
Appendix D:
Appendix E:
Appendix F:
Appendix G:
Appendix H:
Sampling and Analytical Protocols
A.I Method 101A - Mercury
A.2 Draft Method 29 - Multi-Metals
Field Data Sheets
B.I Mercury
B.2 Multi-Metals
Sample Parameter Calculation Sheets
C.1 Mercury
C.2 Multi-Metals
Analytical Data
D.1 Mercury
D.2 Multi-Metals
OMSS Process Data
Sample Equations
Project Participants
Dry Gas Meter Calibration Data
epp.053
IV
-------
FIGURES
Number Page
2-1 General plot plan 8
2-2 Process flow diagram for the Stanislaus County MWC . 9
2-3 Stack gas sampling ports 11
2-4 EPA Method 101A sampling train 13
2-5 Method 101A sample recovery scheme 15
2-6 Method 101A sample preparation and analysis scheme . 16
2-7 Schematic of Multiple Metals sampling train 19
2-8 Metals sample recovery scheme 20
2-9 Metals sample preparation and analysis scheme .... 21
4-1 Multiple Metals Method RSD values versus mercury
concentrations 39
5-1 Comparison of RSD values for both mercury testing
methods 42
epp.053 V
-------
TABLES
Number Page
1-1 Test Conditions Matrix for OMSS Emissions Control
Field Test (1991) 3
2-1 Sampling Times, Minimum Sampling Volumes, and
Detection Limits 17
3-1 Method 101A Fractional Results 25
3-2 Mercury Flue Gas Concentrations and Operating
Parameters Method 101A 27
3-3 Method 101A Precision 28
4-1 Multiple Metals Method Fractional Results 31
4-2 Mercury Concentration and Operating Parameters
Multiple Metals Method 33
4-3 Multiple Metals Method Cadmium and Lead Fractional
Results 34
4-4 Cadmium and Lead Concentrations Based on Nondetects
at Zero Multiple Metals Method 35
4-5 Cadmium and Lead Concentrations Based on Nondetects
at Detection Limit Multiple Metals Method 36
4-6 Multiple Metals Method Precision 38
5-1 Variability of Multiple Metals and Method 101A
Mercury Data 41
5-2 Comparison of Variances for Multiple Metals and
Method 101A Mercury Data 44
5-3 Comparison of Differences in Average Mercury
Concentrations for Multiple Metals Method and
Method 101A 46
5-4 t-Statistic and Correction Factors for Average
Mercury Concentrations 48
7-1 Summary of Estimated Precision, Accuracy, and
Completeness Objectives and Results 62
epp.053 VI
-------
TABLES, Continued
Number Page
7-2 Sampling, Sample Control, and Analytical Errors
with Associated Corrective Actions 64
7-3 Isokinetic Results for the Stack Mercury (101A)
Tests 67
7-4 Dry Gas Meter Calibration Check 68
7-5 Isokinetic Results for the Stack Multiple Metals
Tests 70
7-6 Mercury 101A Method Blank Results 71
7-7 Mercury 101A Matrix Spike Results 73
7-8 Mercury 101A Laboratory Control Sample Results ... 74
7-9 Multiple Metals Method Blank, Matrix Spike, and
Laboratory Control Sample Results 75
7-10 Coefficients of Variation for the Outlet Flue Gas
Concentrations 77
epp.053 VI1
-------
1.0 INTRODUCTION
1.1 BACKGROUND
Section 129 of the 1990 Clean Air Act Amendments requires
the U.S. Environmental Protection Agency (EPA) to promulgate
mercury (Hg) emission limits for municipal waste combustion (MWC)
facilities. Section 129(c)(3) specifies that the test methods
and procedures required as part of these regulations must be
validated on solid waste incineration units. The two existing
EPA test methods for sampling and analysis of Hg emissions are
Method 101A "Determination of Particulate and Gaseous Mercury
Emissions from Sewage Sludge Incinerators" and Draft Method 29
for multiple metals (MM) "Determination of Metals Emissions from
Stationary Sources." Both of these methods are modifications of
EPA Method 5 for determining particulate emissions from
stationary sources, but EPA has not field validated either method
on an MWC.
Radian Corporation was contracted by the EPA's Air and
Energy Engineering Research Laboratory (AEERL) to conduct tests
on one of the two MWC units at the Ogden Martin Systems of
Stanislaus, Inc. (OMSS) facility in Crows Landing, CA. The
original objective of the tests was to determine the effect of
activated carbon injection on Hg emissions.1 The AEERL work
assignment was modified before testing began to include work for
the Emission Measurement Branch (EMB) of EPA's Office of Air
Quality Planning and Standards. The primary objective of the EMB
work was to assess the precision of Method 101A and the MM method
A separate EPA report discussing the carbon injection testing
is available and is entitled "OMSS Field Test Report on
Carbon Injection for Mercury Control."
epp.053
-------
for determining Hg emissions and the MM method for determining
cadmium (Cd) and lead (Pb) emissions from MWC's. A secondary
objective of the testing was to determine whether there is a
difference in the average measured Hg concentrations between the
two methods. This difference in average values between methods
is referred to in this report as inter-method bias. Radian
conducted field tests at OMSS from July 22 through August 10,
1991.
1.2 TEST DESIGN
1.2.1 Plant Description and Operation
The OMSS facility, which began operation in 1988, consists
of two identical Martin GmbH mass burn waterwall (MB/WW)
combustors, each of which is capable of combusting 400 tons per
day (tpd) of municipal solid waste (MSW). Each unit is equipped
with ammonia injection into the furnace, a spray dryer (SD), and
a fabric filter (FF) to control emissions. The facility normally
operates at full capacity 24 hours per day. All testing was
conducted on Unit No. 2.
1.2.2 Test Matrix
All of the Method 101A and the MM method test data contained
in this report were collected at the stack, downstream of all
control devices. Two dual-train systems, each with side-by-side
nozzles, were used to validate Method 101A and the MM method.
The two dual-trains were operated through perpendicular ports,
with inverse traversing, taking simultaneous samples during three
runs at each of five plant operating conditions. Table 1-1
presents the test conditions for these validation tests. Each
test run lasted approximately one hour. Test conditions were
determined by the requirements of the AEERL control system
evaluation effort and are not of specific importance to method
validation, other than to provide a range of emission levels for
sampling.
1.3 CONCLUSIONS
The following conclusions were reached based on the testing
and analysis conducted on the OMSS MWC.
epp.053
-------
rH
CT\
cn
EH
CO
U
EH
Q
U
H
b
O
EH
2
O
CJ
co
2
0
co
H
CO
CO
H
S
CO
co
0
O
X
H
OS
EH
g*
co
2
O
H
EH
M
Q
2
O
U
EH
CO
W
EH
•
rH
1
H
W
i_3
(Q
f^
EH
CO
W
EH
U
|
CU
o
2
OPERATI
u
H
^
O
OS
CJ
ABRIC
fa
CARBON3
2
O
CQ
<
CJ
si
"s
20
O 2
H O
EH H
CJ EH
U
(0
s
3
(Pj
(V
CU
-P
•H
C
CP
•H
i_]
03
C
3
^
rH
(0
3
•H
•H
T3
C
-H
tP
C
xT"T(
S
0 o
in CM
CTl CTl
CM
II 1
in
(0 r-
CU fN
cu
CU CT>
O (0
(0 H
3 (0
CO
01
- CU
^3 5- 1 •
o 3 -P
H 4J 0)
CJ (0 rH
OH k C
CU M
o cu
u e ^
OS CU CU
i^ i j ^^
Q r-l
4J Q
II CU
rH 4-1 (0
(0 3 r4
-H o a
±4 CO
CU i-l
4-> CU II
(0 4J
S rH 4->
•H 0)
3 4H rH
(0 C
OS O M
•H
rH ^ •
(0 X! Q
O (0 •
CJ CK CO
tP
C
-H
^
3
T3
C?
^^
o\°
4->
a
»-i|
^
rH
n
T3
0)
Cr>
(C
0)
^
frt
(U
.
r\
2vels, whic
reduction]
»•*
_J tN
H 0
CM CO
O
co <*>
\Q
| ^ 0^
0) 1
rH CT*
-P CO
3
o o
C
O rH
(0
T3 3
CU &
10 0)
<0 "-^
XI
CO
CU C
4-> 3
l-i
rH
T3 fd
CU 3
Q) T3
^
Q) -H
6 'O
•H C
p-3 **H
xi o T3
-------
1.3.1 Method Precision for Mercury
At measured Hg concentrations above approximately
200 micrograms per dry standard cubic meter (/ig/dscm) , the
relative standard deviation (RSD) of Method 101A was 15.1 percent
and the RSD of the MM method was 9.8 percent. At measured
concentrations less than 200 jug/dscm, the RSD of Method 101A was
40.7 percent, and the RSD of the MM method was 20.7 percent.
Both methods were demonstrated to be sufficiently precise in the
determination of Hg concentrations of MWC stack gas to meet the
EPA criterion of 50 percent RSD in "Protocol for the Field
Validation of Emission Concentrations from Stationary Sources,"
EPA 450/4-90-015, April 1991 (hereafter referred to as the
"validation protocol"). The variance of the MM method was lower
than for Method 101A, but this difference could have been due to
random variation in measurements, rather than real differences.
1.3.2 Method Comparison
The validation protocol allows alternative methods to be
evaluated against validated methods to determine inter-method
bias. Although no decision on the validation status of either
method for MWC's has been made at this time, an assessment of
inter-method bias will yield useful comparison information.
The MM method consistently yielded higher Hg concentration
measurements than Method 101A. The difference in measured
concentration averaged 39 jiig/dscm and is statistically
significant at the 80 percent confidence level specified in the
validation protocol. The bias correction factor (CF) exceeds the
±10 percent criteria allowed by the validation protocol.
1.3.3 Method Precision for Cadmium and Lead
Levels of Cd and Pb measured by the MM method during this
test program were near or below the detection limits in many
cases. Measured concentrations of Cd were especially low.
Statistical analysis using two alternative assumptions regarding
nondetected sample concentrations indicate that the precision of
the method is probably sufficient to meet the EPA protocol of
50 percent RSD. However, a significant uncertainty is associated
epp.053
-------
with any conclusions about the Cd and Pb measurements with the MM
method because of their proximity to the detection limits.
1.4 REPORT ORGANIZATION
Section 2.0 of this report details test design, including a
description of the methodologies used for interpretation of the
test data, the facility tested, and the sampling and analysis
methods used. Section 3.0 details the results of tests of
Method 101A for Hg. Section 4.0 details the MM test results for
Hg, Cd, and Pb. Section 5.0 is a comparison of the results of
the two methods in determining Hg emissions. Section 6.0
provides a detailed review of the Method 101A and MM sampling and
analytical procedures used during the testing. Section 7.0
covers quality assurance/quality control (QA/QC) procedures used
to ensure test data quality. Supporting documentation for the
sampling and analytical data are provided in the appendices.
epp.053
-------
2.0 TEST DESIGN
2.1 METHOD VALIDATION REQUIREMENTS
Both Method 101A and the MM method were evaluated for Hg
measurement precision. The Hg concentrations measured by each
method were also compared to see if there was a significant
difference in their values. The precision of the MM method
for measuring Cd and Pb concentrations was also evaluated.
The validation protocol for quadruplet sampling trains was
used as a source for specific statistical techniques. The
calculations and equations used are presented in Sections 3.0,
4.0, and 5.0 of this report.
Method precision statistics include sampling, recovery,
and analysis variations. The primary statistics used to
evaluate method precision are the RSD and variance. The RSD
is the calculated method standard deviation divided by the
mean value of the test concentrations. The validation
protocol specifies an upper limit of 50 percent for the RSD in
method validation studies. The method variance and standard
deviation were calculated as intermediate values in the
determination of RSD for each sampling method. To eliminate
the effect of process variation on method precision
statistics, estimates of precision were based on the
differences in concentrations measured by sampling train pairs
within single runs.
In addition to statistical comparisons outlined in the
validation protocol, the variances of the two Hg sampling
methods were compared by using a one-tailed F test to
determine whether the variance of the Method 101A measurements
was significantly greater than for the MM method.
epp.053
-------
Finally, Hg concentrations determined by the MM method
were compared to those determined by Method 101A by comparing
the mean values using the t statistic at the 80 percent
confidence level.
Assessment of method accuracy was beyond the scope of
this project, as known concentrations of Hg were not measured
in the field. However, laboratory analysis accuracy was
assessed using matrix spikes and laboratory control samples.
Results of these tests are covered in Section 7.0.
2.2 FACILITY DESCRIPTION
2.2.1 Overview
The Stanislaus County MWC is owned and operated by OMSS.
The plant began operation in 1988 and is located in Crows
Landing, CA. The facility consists of two identical Martin
GmbH mass burn waterwall (MB/WW) combustors, each of which is
capable of combusting 400 tpd of municipal solid waste (MSW).
The MSW burned at this plant is received from the City of
Modesto and from Stanislaus County. Steam produced by the two
boilers is used to generate electricity which is sold to
Pacific Gas & Electric Company. The facility normally
operates at full capacity, 24 hours per day. A plot plan of
the facility is shown in Figure 2-1.
The air pollution control on each combustor consists of
an Exxon Thermal DeNOx® system and a Flakt SD/FF system. A
general schematic of the system is shown in Figure 2-2. The
Thermal DeNOx® system injects NH3 into the upper furnace to
control NOX emissions. Flue gas leaving the combustor and
boiler is routed to the top of the SD through a 10 foot (ft) x
3.5 ft duct that contains two sharp turns. The primary duct
section between the economizer and SD is vertical and is
approximately 40 ft long. Just prior to entering the SD, the
flue gas is equally distributed to three inlet dispersers. At
the exit of each disperser, slaked lime slurry is injected
through dual-fluid nozzles. The slurry feed rate is
controlled according to the stack S02 concentration, and the
epp.053
-------
Figure 2-1. General Plot Plan
Ogden Martin Systems of Stanislaus, Inc. (1991)
8
AC-0077
-------
§•9 o'-
u
S
-P
c
3
o
u
tn
3
(0
rH
in
•H
C
(C
-p
OQ
0)
g
(0
(0
•H
T3
a)
o
o
S-i
CM
I
CM
0)
J-i
-------
dilution water flow is controlled according to the SD outlet
temperature. A flue gas residence time in the SD of roughly
15 seconds is provided to dry the slurry and to neutralize
acid gases. Following the SD, the flue gas enters the pulse-
jet FF at a design flow rate of 94,000 cubic feet per minute
(cfm) at 285°F (2,700 m3/min at 140°C). The FF has six
compartments of Teflon®-coated fiberglass bags (1,596 bags
total) and a net air-to-cloth ratio of 3.2 cfm/ft2. The
cleaning cycle of the bags is approximately 2 minutes per
compartment, equal to 12 minutes for the entire FF.
2.2.2 Sampling Locations
All flue gas samples were taken from the stack of Unit 2.
Flue gas exits the FF on the east side of the unit and is
directed into an induced draft (ID) fan located at the base of
the stack. The in-house continuous emissions monitoring
system (GEMS) stack probes are located on the downstream side
of the ID fan in the stack breeching. The gas then enters the
stack flue and is emitted into the atmosphere approximately
140 feet above ground level. The stack sampling platform is
located approximately halfway up the stack as shown in
Figure 2-3.
The flues for Units 1 and 2 are located in the same stack
shell. Each flue has an inside diameter of 5 ft 3 inches.
The two ports used for sampling were 6-inch flanged pipe ports
with coplanar axes, perpendicular to each other and to the
flue axis. The nipple length of the ports is approximately
20 inches.
The two ports meet the EPA Method 1 criterion of being
perpendicular and of being eight duct diameters downstream and
two duct diameters upstream from the closest flow
disturbances. A total of 12 sample points was used. The
stack sampling location was accessed from the ground by a
ladder located on the outside wall of the stack.
2.3 SAMPLING MATRIX
Mercury emissions testing by Method 101A and the MM
method was performed during Conditions 4, 5, 6, 8, and 9 of
epp.053 10
-------
11
-------
the overall OMSS test program. Operating parameters for each
test condition are shown in Table 1-1. Conditions 4 and 5
were baseline tests run with no carbon injection for control
of Hg emissions. Conditions 6, 8, and 9 were run with carbon
injection at the spray dryer inlet. The carbon feed rate and
carbon type were varied between Conditions 6, 8, and 9.
Condition 5 was conducted with the Thermal DeNOx® system out
of operation; the system was operating during Conditions 4, 6,
8, and 9. As a result of these process changes, a broad range
of Hg levels was expected at the stack.
Three sampling runs were made for each operating
condition. In each sampling run, two samples were collected
using dual Method 101A trains, and two samples were
simultaneously collected from the adjacent sampling port using
dual MM method sampling trains.
2.4 DESCRIPTION OF METHOD 101A
2.4.1 Background
Method 101A was developed specifically for determination
of particulate and gaseous Hg emissions from sewage sludge
incinerators. The method is similar to EPA Method 101, which
was developed for chlor-alkali plants. Method 101A has been
used to measure Hg emissions from other incineration sources,
including MWC's, although its use on MWC's has not been
validated by the EPA.
2.4.2 Sampling System
Method 101A for Hg emissions testing is found in 40 CFR
Part 61, Appendix B. The method calls for isokinetic
extraction of flue gas through a sampling train similar to the
standard EPA Method 5 train. The Method 101A sampling train
is shown in Figure 2-4. Although Method 101A states that the
use of a sampling train filter is optional, filters were
included in all of the sampling trains used during the OMSS
test. After passing through the filter, the sample stream is
bubbled through three impingers containing acidified potassium
permanganate (KMnO4) solution.
epp.053 12
-------
U6ZW19S
c
O)
"a
co
£
LLI
^•'
C
-------
In the Method 101A tests conducted at OMSS, the sample
train was recovered in two fractions: the probe
rinse/impinger catch and the filter. The Method 101A recovery
scheme is shown in Figure 2-5. Following sample recovery, the
rinse/impinger catch and filter fractions were stored in
separate containers and shipped back to Radian's Perimeter
Park (PPK) laboratory for analysis.
The analytical preparation procedures are shown in
Figure 2-6. They consisted of filtering the rinse/impinger
and filter solutions, combining the filtrates, and analyzing
an aliquot by Cold Vapor Atomic Absorption Spectroscopy
(CVAAS). Studies recently conducted by the EPA show that
after a certain sample storage time, the laboratory filtering
procedures may remove a portion of the collected Hg contained
in a manganese dioxide (MnC>2) precipitate. For this test
program, the analytical filters were redigested, the digestion
solution was filtered, and the filtrate was analyzed for Hg by
CVAAS. A graphical output of the CVAAS results is provided
for each analysis.
The approximate analytical and corresponding flue gas
detection limits for Hg using Method 101A are summarized in
Table 2-1. A detailed discussion of Method 101A sampling and
analytical procedures is presented in Section 6.1. A full
description of the method is provided in Appendix A of this
test report.
2.5 MULTIPLE METALS SAMPLING
2.5.1 Background
The MM method, also known as Draft Method 29, was
developed by the EPA to support regulation of toxic metals
emissions from incineration processes. The method was
designed to measure emissions of the following 16 toxic
metals: Hg, Cd, Pb, zinc (Zn), phosphorus (P), chromium (Cr),
copper (Co), nickel (Ni), manganese (Mn), selenium (Se),
arsenic (As), beryllium (Be), thallium (Tl), silver (Ag),
antimony (Sb), and barium (Ba). The method has undergone
epp.053 14
-------
I
CO
I
£
o
0)
g
0)
XI
o
t/1
o
o
0)
"9
2s
iff
jr
0)
rH
a
o
H
T3
o
X!
4J
0)
s
i
in
I
CM
0)
Cn
•H
fa
epp.053
15
-------
Sample Filter in KMnO4
Transfer to beaker
& evaporate to
25 ml
Probe/Front Half
KMnO4 Rinse
Impinger Contents &
KMnO4 Rinses
Add 20 ml concentrated
HNO,& heat for
2 hours @ 70°C
Allow to cool
Combine
Filter & Wash
with 8N HCI
Dilute to known
Volume with DIH20
Analyze for Hg
using CVAAS
8NHCI
Impinger Rinse
(if used)
Digest Filter
wRh 25 mL
8NHCI
Soak for 24 hours
Fitter
Analyze for Hg
using CVAAS
-— Archive
Filters
Figure 2-6. Method 101A sample preparation and analysis scheme.
epp.053
16
-------
iIMITS
_!
5^
DETECTIO]
Q
<
^
W
W
e
j
o
>
o
z
H
J
cu
w
g
S
H
Z
s
..
M
U
S3
H
EH
u
z
H
J
a.
s
IS*
w
•
H
1
CM
W
J
PQ
<
EH
1 1
4-*
•H
g
-H
(J
c
O
Detect!
rH
(0
u
-H
4J
K^
i^
rH
rd
c
<
in
fO
o
CD
3
rH
fc
CD
4J
>i
I
r^
re
C
<
CD
rj
g rH
3 o r?
%* o
.5^
~ *^ w ^
s a
g
rd
w
0)
g
-H
r, ^f^
w
CT> r-l
C 3
-H O
rH X!
a^-
g
(0
w
CP
3-S
^ <°
& M
1^
W
H e e ^
6 -C. <. e
****** ^r\ ^n *"'****•
& * * &
x. ^- ^- 3.
" I i ^
o o o o
e g g g
n ° ° u
SS^S
T> ^ ^_ T3
"t? 21 5* "^
3. =5. a. ^
in * N in
M 0 0 N
Cn T3 X! Cr-
K U CU K
o o
n n
/rt
™L H
f^
<
rH S
o g
H
•
CD
g
•H
4J
t^
•rH
rH
£
(0
en
CD
4J
rrl
'W
^J
rH
3
O
rH
/rt
(VJ
U
o
-p
Tl
w
CD
U)
3
W
rrt
1U
C
•*H
g
m
i j
+J
"W
in
•
0
4H
0
CD
4J
rO
r^
CP
C
•H
rH
a
g
rfl
W
CD
CT
fO
!H
CD
(0
C
<
(0
17
-------
extensive laboratory testing and field evaluation, including
field use at MWC's.
The OMSS test samples were analyzed for As, Cd, Cr, Pb,
Hg, and Ni. Results for Cd, Pb, and Hg are presented in this
report.
2.5.2 Sampling System
The MM method uses the sampling train shown in
Figure 2-7. The seven-impinger train consists of a quartz
nozzle/probe liner followed by a heated filter assembly with a
Pallflex Tissuequartz 2500QAS filter and Teflon® filter
support, a series of impingers, the usual EPA Method 5
meterbox, and a vacuum pump. The sample is not exposed to any
metal surfaces in this train. The contents of the sequential
impingers are: two impingers with a 5 percent nitric acid
(HN03)/10 percent hydrogen peroxide (H2C>2) solution, two
impingers with a 4 percent KMnC>4/10 percent sulfuric acid
(H2SC>4) solution, and one impinger containing silica gel.
Empty knockout impingers are located both before and after the
HN03/H2O2 impingers. The impingers are connected together
with clean glass U-tube connectors. Sampling train components
are recovered in five fractions as shown in Figure 2-8.
The laboratory preparation and analysis scheme for the MM
method samples is shown in Figure 2-9. The sampling train
filter was digested with hydrofluoric acid (HF) and HNC>3. The
front half acetone rinse was dried, dissolved in HNO3, and
combined with the front half HN03 rinse. This solution was
combined with the filter digestion solution and aliquots were
analyzed for target metals by Inductively-Coupled Plasma (TCP)
spectroscopy, for Pb by Graphite Furnace Atomic Absorption
Spectroscopy (GFAAS), and for Hg by CVAAS. The HNO3/H2O2
impingers fraction was digested and aliquots were analyzed by
GFAAS and CVAAS. Aliquots from the KMnC>4 impingers and the
rinse of the two empty impingers were analyzed by CVAAS for
Hg.
The approximate analytical and corresponding flue gas
detection limits for Hg, Cd, and Pb for the MM method are
epp.053 18
-------
c
-H
(0
^
-p
(0
(fl
I— I
(0
-p
(U
0)
rH
a
O
•H
^>
(0
g
Q)
4=
O
I
CN
Q)
epp.053
19
-------
uziwose
2«£
g gjg
lie
If 1|| - Hi!1
c.Q «,iu UJ.Q H
epp.053
20
-------
C CD CD
'
o
O d-
with Acid
manganat
Cfor2hr
d Analyze
by CWAS
o o
*-'
m
§
O
p
>
ca
<
T3
CO
O
1
CO
Q.
_
a
i
(n
8
o
q>
c\i
o
3
O)
21
-------
summarized in Table 2-1. A detailed discussion of the MM
sampling and analytical procedures is presented in
Section 6.2. A full description of the method is provided in
Appendix A.
2.6 Parameters Measured
In addition to Hg, Cd, and Pb emissions, a number of
operating parameters were measured. Stack gas flow rate,
moisture, oxygen (02) concentration, carbon dioxide (CC>2)
concentration, and temperature were all measured at the
sampling location during each run.
2.6.1 Volumetric Flow Rate Determination by EPA Method 2
Volumetric flow rate was measured according to EPA
Method 2. A Type K thermocouple and S-type pitot tube were
used to measure flue gas temperature and velocity,
respectively.
For EPA Method 2, the pitot tubes were calibrated before
use following the directions in the method. Also, the pitots
were leak-checked before and after each run.
The parameters that were measured include the pressure
difference across the pitots, stack temperature, stack static
pressure, and ambient pressure. A computer program was used
to calculate the average velocity during the sampling period.
2.6.2 Oxygen and Carbon Dioxide Concentrations by EPA
Method 3A
The 02 and CC>2 concentrations were determined by GEMS
following EPA Method 3A. Flue gas was extracted from the duct
and delivered to the CEMS through heated Teflon® tubing. The
sample stream was then conditioned (particulate and moisture
removed) and directed to the analyzers that determine 02 and
CC>2 concentrations on a dry basis. Average concentrations
were calculated for each test period.
Only the 02 values were used for evaluating the data
collected for EMB. The 02 concentration was measured using a
Thermox Model WDG III which uses an electrochemical cell.
Zirconium oxide contained in the cell conducts electrons when
it is hot, due to the mobility of 02 ions in its crystal
epp.053 22
-------
structure. Porous platinum electrodes attached to the inside
and outside of the cell provide the instrument voltage
response. A difference in 02 concentration between the sample
side of the cell and the reference (outside) side of the cell
produces a voltage. This response voltage is proportional to
the logarithm of the 02 concentration ratio. A linearizer
circuit board is used to make the response linear. Reference
gas is ambient air at 20.9 percent ©2 by volume.
2.6.3 Average Moisture Determination by EPA Method 4
The average moisture content of the flue gas was
determined according to EPA Method 4. Before sampling, the
initial weight of the impingers was recorded. When sampling
was completed, the final weights of the impingers were
recorded, and the weight gain was calculated. The weight gain
and the volume of gas sampled were used to calculate the
average moisture content (percent) of the flue gas. The
calculations were performed by computer. Method 4 results
were incorporated in all isokinetic sampling methods used
during the test.
epp.053 23
-------
3.0 METHOD 101A RESULTS
3.1 REVIEW OF DATA
Paired samples for Hg analyses were gathered using the
dual Method 101A sampling trains during each of the five
operating conditions shown in Table 1-1. Three sampling runs
were conducted for each condition, except for Condition 6.
During Run 1 of Condition 6, one of the sampling trains was
invalidated because it failed the final leak-check. A fourth
run was added because of the leak-check problem.
The recovered fractions of each sample were combined and
analyzed for Hg content. The results, showing Hg in jug, are
shown in Table 3-1. An "R" after the run number notes the
second sample collected by the dual train. The Hg recovered
from the laboratory filter is also shown in Table 3-1. The Hg
results at the outlet varied by test condition, with the
highest levels measured during Conditions 4 and 5, which were
run without carbon injection. Average Hg levels during
Condition 6, 8, and 9 vary with the carbon feed rate.
There were analytical problems that resulted in the
exclusion of data from two runs of Condition 4. These
problems were caused by illegible labels, smudged in transit.
As a result, the impinger contents for the replicate 101A
train during Run 2 (2R) of Condition 4 were accidentally
switched with the impinger contents of a MM train from the
same run. Also during Condition 4, the HCl rinse for the
replicate 101A train during Run 3 (3R) was accidentally added
to the other Run 3 101A train. As a result, the Hg results
from the two outlet trains were added together and reported as
a single value for Run 3.
epp.053 24
-------
TABLE 3-1. METHOD 101A FRACTIONAL RESULTS
Measured Mercury (^g)
Condition
4
5
6
8
9
Run
1
1R
2
2R
3
3R
1
1R
2
2R
3
3R
1
1R
2
2R
3
3R
4
4R
1
1R
2
2R
3
3R
1
1R
2
2R
3
3R
Front Half
+ Impingers
201
164
235
— a
706
__b
361
318
207
329
351
287
c
22.3
32.9
31.7
7.8
23.4
46.9
37.1
77.8
34.1
60.1
45.0
24.5
50.5
86.0
148
57.8
96.0
57.9
79.1
Analytical
Filters
2.1
4.7
0.3
0.3
24.8
__b
0.5
6.0
3.1
0.3
3.2
5.7
— C
4.4
0.4
1.5
4.0
2.8
1.4
0.7
0.8
2.9
2.9
4.5
3.8
1.9
1.9
1.8
2.6
0.7
4.3
1.2
Total
203
168
235
a
730
__b
361
324
210
329
354
292
— C
26.7
33.3
33.2
11.8
26.2
48.3
37.8
78.6
37.0
63.0
49.5
28.3
52.4
87.9
149
60.4
96.7
62.2
80.3
Percentage
of Total on
Analytical
Filters
1%
3%
0.1%
3%
0.1%
2%
1%
0.1%
0.9%
2%
16%
1%
5%
34%
11%
3%
2%
1%
8%
5%
9%
13%
4%
2%
1%
4%
0.7%
7%
1%
a Impinger fraction mistakenly switched with KMn04 fraction from
MM train for same run.
b Fractions from the paired trains on Condition 4, Run 3 were
accidentally combined. Mercury values are therefore the
average of the two runs.
c Sampling train failed to pass final leak-check.
epp.053
25
-------
The Hg content on the laboratory filters varied from
0.3-6.0 jug on all but one run. During Condition 4, Run 3, the
Hg level was 24.8 nq. The filters averaged 2.4 percent of the
total Hg found in each train, and ranged from less than
1 percent to a high of 34 percent during Condition 6, Run 3.
The Hg content of these filters was less than 5 percent of the
total on all but 7 of the 29 trains with valid data. All 7 of
these trains were from runs during which carbon was being
injected and total Hg levels were less than 62 p,g. The Hg
levels on the laboratory filters during these 7 runs were
2.8-4.5 jug.
Mercury concentrations in the flue gas in /ug/dscm at
7 percent 02 and associated sampling train parameters for each
run are shown in Table 3-2. Measured temperatures and 02
concentrations in the flue gas are also listed in Table 3-2.
3.2 STATISTICAL ANALYSIS
Method 10 1A was taken to be the standard method for Hg
sampling for the purposes of this test. Data from
Condition 4 , Runs 2 and 3 , were not used in the calculation of
method variance because of the problems discussed in
Section 3.1.
Techniques for determining the precision of the sampling
method were taken from the validation protocol. Statistics
relative to the sampling method precision are shown in
Table 3-3. Precision of the Method 101A data is indicated by
the variance of the sample data. Variance is calculated using
the following equation:
(3-1)
2n
where s2 = the variance of the sample set;
di = the difference between the measured Hg
concentrations of the paired sample trains in a
run ; and
n = the number of runs.
The square root of the variance is the standard
deviation, which is another common indicator of the precision
epp.053 26
-------
en
w
EH
W
^5
OH
ft
o
2
H
EH
W
ft
0
Q
2
W
2
O
EH **•
,rf ^
S
EH H
CONCE
METHOE
01
jrf
o
w
3
tn
S
p
u
w
s
*
CM
1
ro
M
CQ
EH
O —
03 <*>
O "7
« a?"""
4->
0)
Z> £ IT
O O
^ iH n
t, ^
^•g-
CT> 5u
K 3 <0
73
O C"
03 ^
H
O "^
se ^"^
*!•*
-P
CD
rH -^
° *t
^"g"
s5|
Nd?
u ^
CM""
0 *
Q) ' ~
£
rH
VO
VD
VO
H
rH
CM
CM
rn
r-
co
^
0
rH
O^
^D
r-
CM
H
"*
cn vo
£ 5
VO CM
CTl CO
H H
CO VD
VO CO
in vo
rH rH
CO 43
1 1
1 1
n n
co r>-
cn cn
cn ^
CO O
rH CM
o cn
CO *t
in vo
rH H
oo r-«
^3* O
in n
n in
CM M
cn cn
C5
cn
•
, cn
rH
r- rH
in vo
CM CM
CM M
in H
s s
0 VO
H O
(N CN
n cn
n o
• CO
(N CN
M H
VO
"? "*
! s
co r>
cn o
H CN
oo r^
cn H
in vo
rH rH
O VO
*t n
H ^
s s
in H
cn o
rH CM
CO *3*
o cn
vo r-
H H
(N VO
00 VO
^* ^H
Q\
* o
rH
in vo
cn cn
co cn
in <•
CO CO
CN CM
CM n
"1 co
(^
rH
cn CM
o o
CM CN
CO CO
^ r^
in 'a-
rH rH
O •*!•
CM •*
in vo
VO CO
• *
H O
o o
H rH
CM in
o cn
CM rH
cn CN
co •ta-
in vo
H rH
rH ^
in ^
vo n
rH
CO CO
cn co
o
cn o
co o
oo r^
CN CN
^* r™H
CO
r- cn
in r-»
00 H
CO O
H CN
o cn
CO O
vo VO
rH H
cn m
? ?
(N "
s s
r-- vo
cn oo
rH rH
VO CM
n o
CO CO
H H
vo in
rH CO
cn n
ro o">
cn cn
rH O
0 0
rH rH
r- r»
VD in
CM (N
CN n
CO ^J1
vo vo
cn cn
rH r-
cn o
H CM
VO VD
VO O
vo r-
rH rH
<* CO
CO CM
CM -^
CN rH
in ^
^ S
vo t-~
co cn
rH rH
in ^
r^ CM
VD t>
rH H
n. ^
M °
rH °"
rH CO
cn cn
• . -H
Q) (0
Q) rH ^
O*1 J2 -P
• (0 (0
•HO) C
m > tn -H
M CO -H rH
-p J3 a
d) -P S
S 43 (0
o
g - C
O T3 C -H
rH 0) g
(II £ rj >yJ
•H rH 03
C 42 O 0)
0 g 0 rH
-H 0
-P O -P O
O 0) -P
CO >irH
4H rH 3 3
CO 0 73
43 -P
•P C Q) Q)
•H rj xl -P
S 73 -P ca
•H C
73 O C -H
Q) U -H g
O 73 Q)
•P Q) Q) -P
-H b -P 0)
^ Q) ^i 73
03 S O C
Q. .^
>i 03 Q)
rH C ^1 0)
CO -H
0) -H T3
4^ -P C C
CO O CO O
-P CO -H
03 M 73 -P
•H <4H 0) fO
g -P ^
QJ CO H-I
CU 01 rH C
rH 3 3 0)
£X CO U O
g U rH C
CO 0) CO O
W CQ U U
cO 43 O
27
-------
0
H
Cfl
H
U
w
&4
<
H
O
rH
Q
O
E
H
W
g
*
n
1
M
W
m
^
rd
Q "CT
w iH.
tf
10 o?
ra'H U
4-) W
*d «* , , j
C i ^^
frt ^ .
i \ ^ u
yj Q) 3.
Q ' — '
0) "->
Varianc
Mg/dscm
N-"
c
en °
E -P "i"
„, rd O
wl rH 01
-P *d
t C — -
m ^ ^
> "5
< 0
U
. -.
^H QJ .«
O rH j"
P^ '^
O rd J?
S W *
C
o
-rH
-P
•H
•d
c
o
u
-P
Q)
EH
^1* VO
• •
•^f in
CN H
o o
• I
VD in
«* vo
in r~
H CN
H CM
CN ^
n co
CO VO
CO H
rH ^*
CO
t~i
c
o
•H
H->
u
CD
c
I—I , — .
in
c -
o ^
X!
rd c
U 0
•H
-H -P -P
rH 3 -H
rd Of}
rH X! C
0) -P O
> -H 0
o s —
in
.
CTi
n
n
.
*3*
n
H
H
00
"
vO
00
0>
c
o
•rH
-P ~
U O
•H U
X --
CN
•
rH
"
CTi
*
CN
CO
in
CO
0
H
C\
(T\
f"*
n
C
O
•rH
-P
O
0)
•I—I
d
H
C —
O co
,Q
rH d
rC O
U -H
g -H
3 T3
•H d
Tf O
CD U
g --
VO
•
in
n
•<*
.
CO
CO
CN
o
VO
H
n
d
O
•rH
•P
U
CD
•n
d - — »
H rj\
d d
O 0
X! "H
iH -P
rd -H
U T3
d
Z O
0 U
d
o
•rH
-P
rd
rH
-p
d
0)
u
d
o
o
iverage
IU
•1-
d
o
•H
-P
•H
^
0)
'd
T3
rH
rd
d
rd
-P
U)
*°~*
d
0
•H
•P
Q)
T3
T3
rH
rd
T!
d
rC
H-J
CO
0)
£>
•H
-P
rd
rH
0)
rH
II
Q
co
P*
rd
28
-------
of a sampling method. The RSD is the ratio of the sample
standard deviation to the mean value of the data, and is
expressed as a percentage. This statistic is often used as an
indicator of precision at a particular concentration level and
reflects the need for more precise measurements at low data
values. As shown in Table 3-3, Hg concentrations in the stack
gas were lower during test conditions in which carbon was
injected into the duct ahead of the SD. The variance and
standard deviation of the data collected during runs with
carbon injection were lower than for the data collected
without carbon injection. The RSD, however, was higher with
carbon injection. This reflects the impact of the lower mean
Hg concentrations when carbon injection occurred. Values
shown in Table 3-3 for Conditions 6, 8, and 9 were each
calculated using paired sample train data from three runs and,
therefore, have less statistical significance than the values
derived from the pooled data for all runs with carbon
injection.
3.3 CONCLUSIONS
Section 2 of the validation protocol states that "The
precision of the method at the level of the emission standard
shall not be greater than 50 percent relative standard
deviation." As shown in Table 3-3, the RSD for Method 101A
was 39.5 percent at an average measured Hg concentration of
86.8 /ig/dscm and 15.6 percent at an average of 416.8 /^g/dscm.
This method therefore meets the RSD criterion for acceptance.
epp.053 29
-------
4.0 MULTIPLE METALS METHOD RESULTS
4.1 REVIEW OF DATA
Paired flue gas samples were gathered using dual MM
sampling trains for the same five operating conditions during
which Method 101A dual trains were used. Three runs were made
for each operating condition.
4.1.1 Mercury
Analytical results, showing the amount of Hg in jug
collected in each fraction of the sampling train, are shown in
Table 4-1. The data show that during most runs, 80-90 percent
of the Hg was found in the HNO3/H202 impingers, with most of
the remainder found in the KMnO4 7^804 impinger and rinse
fraction. The primary exception was Condition 6 (high carbon
feed rate) during which the KMnO4/H2SO4 impingers accounted
for 28-53 percent of the total in three of the six trains.
Very little Hg was associated with the front half (FH) or with
the empty impinger rinse, except during Condition 5, when
Train 3R accounted for roughly 6 percent of the total.
During Condition 4, the empty impinger and rinse samples
from Trains 1 and 1R were apparently combined with another
fraction, and thus no separate data are available for these
two rinses. Also during Condition 4, the probe rinse from
Train 2 was combined with the rinse of another train. Because
very low Hg levels were found in similar fractions collected
during the other runs, the impact of these fractions on total
Hg levels during this run is expected to be small. As a
result, Hg measurements from the rest of the fractions during
these runs were included. Results for Train 2R of Condition 4
epp.053 30
-------
TABLE 4-1. MULTIPLE METALS METHOD FRACTIONAL RESULTS
Front
Condition Train Half
4 1
1R
2
2R
3
3R
5 1
1R
2
2R
3
3R
6 1
1R
2
2R
3
3R
8 1
1R
2
2R
3
3R
9 1
1R
2
2R
3
3R
<0
<0
<0
<0
<0
<0
1
4
0
0
0
19
<0
<0
<0
<0
<0
<0
<0
<0
<0
1
<0
<0
<4
<0
<0
5
<0
<0
.392
.392
.392b
.392
.392
.392
.32d
.14
.392
.392
.392
.4
.392
.392
.392
.392
.392
.392
.392
.392
.392
.7*
.392
.392
.34
.392
.392
.6
.392
.392
Empty
Impinger
Rinse
a
a
<0
<0
0
0
1
<0
<0
1
0
0
<0
<0
<0
<0
<0
<0
<0
<0
<0
<0
<0
<0
<0
<0
<0
<0
<0
<0
.235
.235
.838d
.449d
.078
.221
.247
.036d
.534d
.768d
.216
.216
.071
.216
.227
.204
.180
.255
.196
.204
.196
.243
.267
.270
.212
.286
.227
.172
Hg
(M9>
HN03/H202
Impinger
198
200
208
133
405
380
275
321
226
236
334
264
49.
60.
38.
7.
22.
5.
68.
41.
112
70.
80.
74.
106
114
110
97.
87
95.
1
5
2
9d
1
2d
2
4
7
4
9
3
5
KMnO4/H2SO<,
Impinger
41.
38.
51.
— c
139
125
44.
47.
28.
50.
38.
36.
7.
8.
8.
8.
9.
8.
7.
6.
7.
6.
5.
4.
<1.
8.
3.
10.
2.
2.
6
2
7
8
8
7
3
2
7
ld
6d
Od
9d
Od
2d
8d
ld
5d
7d
7d
5d
8
ld
Od
1
2d
6d
Total
239.
238.
259.
C
544.
505.
322.
372.
255.
287.
373.
320.
56.
69.
46.
16.
31.
13.
76.
47.
119.
79.
86.
79.
106
122.
113
113
89.
98.
6
2
7
8
4
2
9
1
7
1
9
2
1
2
8
1
4
0
5
5
1
1
4
1
6
1
a = No HNO3 rinse sample sent to lab.
b = Does not include probe rinse fraction.
c = Switched with KMnO4 catch from Method 101A train.
d = Less than five times the detection limit.
31
-------
were invalidated because the KMn04 impinger contents and rinse
were mistakenly switched with a 101A train.
All of the KMnO4/H2SC>4 impinger fractions during testing
with carbon injection (Conditions 6, 8, and 9) were less than
five times the detection limit. The HNO3/H202 impingers from
Trains 2R and 3R of Condition 6 were also less than five times
the detection limit.
The Hg concentrations determined from the analytical data
and operating parameters for each run are shown in Table 4-2.
The Hg concentrations are on a dry basis and are corrected to
7 percent oxygen.
4.1.2 Cadmium and Lead
Results from the analyses of the MM samples for Cd and Pb
are shown in Table 4-3. The levels of these metals in the
back half impinger solutions for all of the samples were
either below detection limits or less than 5 times the
detection limit. In addition, all but 2 of the 27
measurements of Cd on the sampling train filter were less than
5 times the detection limit. Because the total amount of
these metals captured was small, the true precision of the
measurements is uncertain. Additionally, in many cases, the
detection limit of the metal in the impinger solution is
relatively large compared to the amount measured on the
filter. As a result, the procedure used for handling
nondetects will influence statistical estimates of precision
based on the dual train measurements.
Concentrations of Cd and Pb in the stack gas were
determined using the analytical data and operating conditions
for each run. Table 4-4 shows the concentrations assuming
that nondetected levels of metals in the sample fractions were
zero. Table 4-5 shows these concentrations assuming that
sample fractions with metals content below the detection limit
contained the metals at the detection limit. These two
assumed concentration levels represent the upper and lower
bounds on the actual values. Tables 4-4 and 4-5 also contain
operating conditions for each run. As opposed to the Hg data,
epp.053 32
-------
CM
I
m
4}
3
O
(0 ^
M ^
o *"*
CMM
33 ^
|S
3
o
00
a
o —
||
Is
u
"1
^^
o «<
u ^
c
« _
&5
W £l|
s ^—
0)
3 &
•-H E
*H
I
TJ
C
0
u
-*•
cn
r-.
rH
CM
to
to
CO
m
cn
CM
cn
m
p^
to
to
to
m
cn
00
PS,
o
cn
to
CM
rH
.fl-
cn
to
cn
to
cn
rH
CO
to
m
i
0)
cn
CO
o
CO
m
CO
CM
O
O
p-
m
CM
CM
to
PS.
cn
CM
CO
rH
to
CO
to
to
0
o
cn
o
to
to
o
rH
cn
cn
cn
•H
to
CM
cn
in
CO
CD
0
rH
CM
CO
cn
rH
CO
m
0)
m
m
to
o
rH
cn
to
0
CM
rH
m
rH
O
rH
to
0
CM
CO
O
tn
cn
-a-
-a-
cn
cn
rH
O
O)
rs.
0
(O
CM
O
rH
O
cn
O)
rH
CO
CM
CO
O)
O)
rH
CM
CM
m
to
PS,
rH
to
o
CM
to
m
>n
o
rH
0
rH
rH
cn
to
CO
CM
«
o>
CO
at
CO
0
N
CM
•n
^
0
O)
ID
cn
N
in
ID
-»•
CO
-a-
O
rH
in
O)
o>
m
tM
rH
to
U1
o
o
t-t
CO
o>
rH
CO
Ol
m
0
rg
O)
in
cn
m
0
to
m
n
to
01
cn
cn
CO
in
N
CM
0)
01
^
0
CM
rH
(0
CO
o
CM
rH
rH
0
-3-
O)
rH
2
0
rH
to
cn
cn
4-
CM
cn
CO
Px
cn
CM
o
CN
CO
-X
rH
£
rH
in
cn
CM
•a-
to
rH
CO
CM
CO
CO
to
o
o
Px
CM
rH
CO
P*.
m
0)
CO
CO
rH
0
CO
to
to
to
CM
rH
O)
cn
to
cn
CO
rn
rH
cn
CD
rH
O
r*
to
CM
CM
cn
px
cn
rH
O
CM
O)
O
CM
CM
rH
to
DO
CM
O
CO
in
o
CO
en
cn
o
o
p-»
in
CM
CO
00
10
cn
rH
cn
rH
(O
to
to
o
o
rH
CM
0)
to
CO
m
fs,
to
to
o
to
rH
O)
o
o
m
CM
rH
CD
-a1
to
cn
Px
0
CM
to
0
0
to
to
rH
rH
cn
CM
CO
to
to
cn
O)
CM
O
O
to
CM
CM
m
o
0
rH
m
cn
rH
m
r-.
m
^
-< H
33
-------
TABLE 4-3. MULTIPLE METALS METHOD
CADMIUM AND LEAD FRACTIONAL RESULTS
Cadmium
Cond . Run
4 1
1R
2
2R
3
3R
5 1
1R
2
2R
3
3R
6 1
1R
2
2R
3
3R
8 1
1R
2
2R
3
3R
9 1
1R
2
2R
3
3R
Front Half
2
0.
0.
0.
0.
0.
-
-
0.
0.
1.
0.
0.
0.
0.
0.
0.
0.
0.
0.
-
0.
0.
.95
84b
73b
95b
97b
88b
-d
-d
78b
74b
27b
65b
75b
79b
81b
73b
85b
87b
91b
97b
-d
78b
72b
1.08
0.
0.
0.
0.
0.
0.
75b
77b
78b
98b
72b
74b
(MS)
Back Half
0.
0.
0.
(0
0.
(0
0.
(0
(0
(0
(0
0.
0.
(0
(0
0.
0.
(0
(0
(0
(0
0.
(0
(0
(0
(0
(0
0.
(0
(0
24b
23b
80b
.22)
40b
.22)
61b
.22)
.22)
.22)
.22)
25b
33b
.23)
.22)
65b
45b
.22)
.22)
.22)
.22)
32b
.22)
.22)
.22)
.22)
.22)
30b
.23)
.22)
TOTAL3
3
1
1
0
1
0
-
-
0
0
1.
0.
1
0
o
1
1
0
0
0
-
1
0
1
0
0
0
1.
0
0
.19
.07
.53
.95
.37
.88
-d
_d
.78
.74
.27
.90
.08
.79
.81
.38
,30
.87
.91
.97
-d
.10
.72
.08
.75
.77
.78
.28
.72
.74
Front Half
3
2
2
2
2
2.
-
-
2.
2,
3
2.
1,
2.
3.
3
3.
2.
4
4 .
-
2.
1.
2,
2.
2.
2.
2.
2.
2,
.51
.55
.20
.55
.82
.82
-d
_d
.00
.27
76
.14
.73
.14
.03
,09
,03
.68
.60
,33
-d
,00
,73
.41
.40
,27
.40
,81
,54
.81
Lead
(MS)
Back Half
(0.
0.
(0
(0
0.
(0
0.
0.
0.
0.
0.
0.
0.
0.
0.
0
(0
(0
0.
(0
0.
0.
0.
(0
0.
1.
0.
0.
0.
0.
33)c
53b
.33)
.33)
78b
.33)
45b
86b
53b
80b
53b
61b
37b
37b
37b
71b
.33)
.33)
36b
.33)
54b
36b
37b
.34)
5Zb
02b
94b
45b
70b
68b
TOTAL3
3.51
3.08
2.20
2.55
3 60
2.82
-d
-d
2.53
3.07
4.29
2.75
2.10
2.51
3.40
3.80
3.03
2.68
4.96
4.33
-d
2.36
2.10
2.41
2.92
3 29
3.34
3.26
3.24
3.49
a Totals are calculated assuming that nondetected values are zero.
Measured value is less than five times the detection limit.
c Numbers in parentheses are detection limits for samples in which the analyte was not detected.
° Filter could not be analyzed for Cd and Pb due to inadvertent addition of KMnO^ to sample container.
34
-------
o
OS
w
N
EH
W
EH
U
W
EH
2
O
2
2
Q
H
W
m
en
2 Q
O O
H ffi
EH EH
<; w
& s
EH
2 cn
M J
U <
2 EH
o w
U S
w w
*+ . T
iJ
J "~i
rH >— '
fe S
Q
^ft
to in ^
m to* rs' co £j
CO Oi CO CD ~
rs to CM o to
PS. CD CO rH O
rH rH rH CM CM
CM 00 tO CO O)
to to co co o
to in to -* m
O CO rH rt fs.
(O 0) O> <**, O
-a- co to ' co
0 (O N n tO
10 -» N •", rH
CM CO CO ^ O
to oo rs. i co
CD CO CO ^ *
m co •» P-. co
rs, CQ O CO rH
rH rH CM «-H CM
to o co to o
ID eo -* m en
(D V) CO -*• -T
-T -
CO
m
CM
CM
-a-
O)
CO
rs.
o
CM
r^.
rH
to
CO
rH
-a-
m
CO
fs.
0
o
i-H
rH
O
CM
-T
CO
rs.
CD
CO
•4-
CO
CO
CO
0
to
CO
CO
*a-
N
CO
CO
rs.
CD
CM
o
CM
CO
rs-
•j-
m
03
to
CO
m
co
0
0
t-H
in
CO
CM
-*
to
to
CO
CO
CO
m
CO
CO
to
0
o
PS.
CM
rH
CO
rs. CO
in rs
CD CD
CO rH
CO 0
rH CM
0 O)
CO O
to to
hx, CO
rs. to
CO CO
m m
rs. to
CM W.
8 S
hs to
CD 00
tO CM
CO O
CO CO
rt CO
• ~.
1 to
*. °
1-1
co to
CO CO
rH 0
O O
rs rs.
to m
CM N
CM CO
CO
to
CO
rH
cn
rH
to
to
to
to
to
in
CO
CO
•n
PS
CO
to
CO
«n
rs
to
CM
<•
.»•
•»
rH
*~*
rH
CO
-»
0
o
m
CM
rH
CO
-a-
to
CD
rs
o
CM
(0
o
fs.
O)
rs.
-a-
CO
CO
rs.
»n
o
rH
fs.
CO
-»•
CM
rs
N
CO
-sT
m
rH
rH
to
CO
CM
o
o
to
N
CM
*n
o
o
rH
m
CD
rH
m
rs
in
CM
CO
m
to
CM
CO
CM
O
rH
rs
o
CM
m
m
m
o
J->
•H
^
(D
fl
N
>^
i-i
(0
5
§
O
4-1
0
3
T3
0)
i-H
A
id
i-H
•H
>
ed
4J
o
t c
u at
•H C
•P 0
-------
EH
to
EH
U
W
EH
W
Q
53
O
Q
H
W
<
CQ
S3
0
H
EH
i
OH
EH
2
M
O
O
U
10
o
M
D
rJ
fa
Q
W
1-3
Q
g
H
S
a
u
•
m
i
^
W
r-3
CQ
EH
Q
O
K
EH
U
S
CO
J_^
<
EH
S
W
j "|
eu
H
EH
^
s
EH
H
S
H
t "]
O
M
EH
U
W
EH
W
Q
*
-P
to m *~I
O) Ol Ol O> ^
r^ to CM o to
rv O) CO rH 0
fH rH rH (N N
CN CD tO « O)
CO tO CO •". O
co in co i O) 0) ^ O)
m 01 -*• r^. OT
r»* co o 01 <-H
rH rH CM rH N
to o o) ro o
tO CO J- W1 O)
to *n oj en
rH rH
O) rv rH tO Ol
to in co o rH
CN CM CM CM CN
rH (N m rH CN
•a- in
en O) t
CO CO* g
Ol 0) 2
O) CO CO
rH 0 Ol
fN CN <-H
eN -* co
-a- N O)
»n m m
to r^ -a-
o r^ -
iH
^
"
O
O
f^
CM
rH
CO
f". Ol CO
m rs. to
Ol Ol O)
CO rH rH
CO 0 01
rH CN rH
0 O) tO
CO O tO
(O tO (O
f^ Ol tO
r^ rH ID
in Ol rH
r*x 01 r*s
S S 5
r*. to to
O) CO CO
to CM m
CO O P^
CO CO tO
n 0) CM
•°, rH «
1 CO -a-
•Q 3 S
1 r-1 rH
en to rH
O) O) O)
rH O •*
000
r^ r*- o
to »o m
CM CM CN
CM CO rH
O)
-a-
CO
01
^
0
CN
to
o
rx
Ol
^
CO
CO
m
0
rH
r--
01
^.
CM
CM
Ol
-a-
CD
•a*
rH
en
01
CN
O
O
to
CM
CM
m
o
o
m
O)
rH
m
CM
Ol
CM
to
CO
CM
O
rH
r*.
o
CM
>n
m
>n
o
rH
m
O)
«a-
rH
O)
o
rH
m
CM
CM
CO
M
CD
•H
t»H
CD
3
N
£
a
g
§
O
0
3
CD
-------
there is no apparent difference in Cd and Pb levels during the
individual test conditions.
4.2 STATISTICAL ANALYSIS
4.2.1 Mercury
The MM data were evaluated for precision using techniques
in the validation protocol. Statistics relative to the
precision of sampling method are shown in Table 4-6. The
precision of the method is indicated by the variance of the
sample data. This was calculated across the range of sampling
conditions using equation 3-1. The standard deviation and RSD
were calculated as described in Section 3.2.
Mercury concentrations in the stack gas were lower
during test conditions in which carbon was injected into the
duct. As shown in Table 4-6, the variance and standard
deviation of the data collected during periods of carbon
injection were lower than for the data collected during the
baseline condition without carbon injection. The RSD,
however, was higher during carbon injection. This reflects
the impact of low Hg concentrations on RSD values. Note also
that the RSD values increased as average Hg concentrations
decreased during individual test conditions, again reflecting
the impact of low Hg concentrations on calculated RSD values.
4.2.2 Cadmium and Lead
The precision of the measurements of Cd and Pb
concentrations in the stack gas was calculated using the same
method as for the Hg data. Statistical indicators of the
precision of Cd and Pb measurements are shown in Table 4-6.
Because carbon injection had no apparent effect on the
concentration of these metals in the stack gas, the data were
pooled to estimate overall precision numbers. The values were
calculated using two alternative assumptions concerning sample
fractions in which either Cd or Pb was not detected. The
first assumption is that the level of nondetected analytes is
zero. The second assumption is that nondetected analytes are
present at the detection limit. Because of the low overall
levels of Cd and Pb measured in the samples, the choice of
epp.053 37
-------
o
H
CO
H
U
w
OS
OH
Q
0
tc
w
g
CO
<
fn
U
g
W
OH
1 — |
E-i
D
S
I
*3*
W
tj
CQ
^
EH
Q —
CO o\°
f* j«-— ^
^ o g
Si (0 T3
C _j -^
£"> ^
c\
Q) -^
U g
c o
(0 W
•H T3
VH "-.
fO 0^
> 3.
C
O
•H ^~.
g, ro |
2 -P T3
ti) ^ "~
r Q) fT1
^ O 3.
O
U
<4-l Q) w
di ^
* g "rrt
o ro $
•z co ft
c
o
•H
-P
•rH
T3
C
O
U
-P
W
0)
H
VO VO
ro 01
H
n ro
** r-!
ro ^j-
in vo
t-> ro
H CM
H CM
in co
H ro
in (0 O 'd
SH rH XI C
3 0) -P O
O > -H U
rH O S ~
0)
s
r.
o
CM
CM
^
CM
in
CO
in
en
vo
H
H
en
C
O
-H
-P -*
U en
QJ -
-o co
C -
H U3
C W
O C
£! 0
^ -H
rrj -p
U -H
T3
X5 C
-P O
-H U
cn
in
ro
in
o
CM
cn
H
o
r^
in
ro
C
O
•H
-P
O
Q)
•f\
C
H *-»
vD
q
0 C
^a o
^H "^
(0 -P
U -H
•d
43 C
tji O
•H U
HH ^ ^
co in ^o
CM ro cn
CM H ro
^ CO 'J
CO CM
!N CM o
r- o P
H CM '
co in .
o
"I ^ S
in co !
CM M3 ^
rH H
ro ro ro
H
O
VH
Q)
C N
O
-H Q)
-PC S-i
O O (0
0) -H
•n -P U)
CO -P
H oj o
•o 0)
C •— C *-** -P
O co H cn 0)
o ^j
rH C C C C
(0 O O O O
U -H 42 -H S
-P M -P
g -H (0 -H
3 T3 U 73 rH
-H C C rH
T3 O S O g (0
Q) U O U 3 >H
g — J — -H Q)
e >
TJ O
(0
o
O H •>*
ro HO
ro H H
CM *t cn
CO O CO
in in ^
O 00
cn ^j1 o
ro in -tf
ro CM CM
O o O
t~> in t^
•
M* ^r
ro ro ro
H H H
O
(-3 Vl tJ
Q)
Q NO
Q) 0) 0)
VH" VH ^
rt3 (0 (0
Ui W W
-P -P -P
0 U U
0) 0) Q)
-P -P -P
0) Q) 0)
T5 T3 T3
C C C
0 O O
Z S3
>. .. -
rH rH rH
r- 1 rH rH
(0 fO ffl
(i \ i t i
OJ Q) (U
^> T3 K* ^
o ro o o
CL)
^
38
-------
assumptions regarding nondetects affects the values of the
precision statistics.
4.3 CONCLUSIONS
Section 2 of the EPA protocol states that "The precision
of the method at the level of the emission standard shall not
be greater than 50 percent relative standard deviation." This
criterion is the basis for evaluating the precision of the MM
train for measuring Hg, Cd, and Pb.
4.3.1 Mercury
As shown in Table 4-6, the RSD for the MM method was
20.7 percent at an average Hg concentration of 117 /ig/dscm
with carbon injection and 9.6 percent at an average Hg
concentration of 494 ^g/dscm without carbon injection. The
use of this proposed method for measurement of Hg therefore
appears to meet the EPA precision criterion for acceptance on
the basis of RSD. For the three test conditions with carbon
injection, the RSD values increase with decreasing average Hg
levels.
4.3.2 Cadmium and Lead
When assuming that Cd and Pb concentrations in
nondetected samples were zero, the RSD of the Cd concentration
measurements was 39.6 percent at an average stack gas
concentration of 1.5 /xg/dscm. The RSD of the Pb concentration
measurements was 11.1 percent at an average stack gas
concentration of 4.6 jug/dscm. When using the alternative
assumption that Cd and Pb are present in nondetected samples
at the detection limit, the RSD of the Cd concentration
measurements was 33.0 percent at an average stack gas
concentration of 1.8 /ug/dscm. The RSD for Pb was 10.4 percent
at an average stack gas concentration of 4.7 /xg/dscm. RSD's
under both assumptions were within the validation protocol
requirement of 50 percent.
epp.053 39
-------
5.0 INTER-METHOD COMPARISON FOR MERCURY
The results of the Method 101A testing and the
simultaneous MM method testing were compared to assess the
relative precision and the difference in measured Hg
concentrations between the two methods. These comparisons
were based on the procedures in the validation protocol.
5.1 STATISTICAL ANALYSIS
5.1.1 Precision
Precision is an indicator of the ability of a measurement
method to achieve similar results under identical operating
conditions (i.e., reproducibility). Table 5-1 summarizes the
statistics comparing the precision of Method 101A and MM
method trains that are presented in Sections 3.0 and 4.0.
Figure 5-1 is a plot of the RSD statistics for the two
sampling methods based on the runs conducted with and without
carbon injection. Based on these data, the MM train appears
to have better precision (i.e., less variability) than
Method 101A and satisfies the requirement in the validation
protocol that requires an alternative method (in this case,
MM) to have less variance than the established method
(Method 101A).
A one-tailed F test was used to determine whether the
lower variance in the MM data is statistically significant.
The F test assesses the likelihood that the difference in
variances of two data sets could be due to random chance by
comparing the F statistic, which is the ratio of the data set
variances, to the known distribution of F. The F statistic
for the MM and Method 101A data sets is:
epp.053 40
-------
TABLE 5-1. VARIABILITY OF MULTIPLE METALS AND
METHOD 101A MERCURY DATA
Statistic Units
No. of Observations
Overall
With Carbon Injection
Without Carbon Injection
Average Hg Concentration Mg/dscm
Overall
With Carbon Injection
Without Carbon Injection
Variance (/xg/dscm) 2
Overall
With Carbon Injection
Without Carbon Injection
Standard Deviation Mg/dscm
Overall
With Carbon Injection
Without Carbon Injection
RSD percent
Overall
With Carbon Injection
Without Carbon Injection
101A
13
9
4
188.3
86.8
416.8
2115
1177
4227
46.0
34.3
65.0
24.4
39.5
15.6
MM
14
9
5
251.5
116.9
493.8
1175
585
2236
34.3
24.2
47.3
13.6
20.7
9.6
41
-------
C
o
'>
0)
Q
O
C
o
CO
(D
>
_D
0)
o:
o
o
o
o
CM
"D
O
SI
£ |
U -i-i
(/I
cn .1
3 "^
C ^
O
O 4.
i_ '
-*~t
C
u
C
o
o
en
I
<
O
0)
n
o
o
o
o
CM
(%) asa
42
-------
(s-i)
where s2p = the variance of the MM data; and
s2v = the variance of the Method 101A data.
The F distribution is dependent on the number of observations
in the two data sets being compared. Values for F
distributions associated with variance observation numbers are
tabulated at specific percentage points in most statistics
texts.
As shown in Table 5-2, the calculated F statistics for
the overall data set and for the runs with and without carbon
varied from 0.508 to 0.555. Due to the differences in the
number of data points (observations) within each data set, the
F statistics represent points on different F distributions.
The F statistics for each data set were compared with critical
values of the appropriate F distribution at different
confidence levels. The results for all three data sets
indicate that the variance of the MM method cannot be said to
be smaller than the variance of Method 101A with 90 percent
confidence. However, the assertion can be made that the
variance of the MM method is smaller with 75 percent
confidence for the overall data set and for the data set based
on carbon injection, but not for the data set without carbon
injection. As a result, there appears to be a reasonable
chance that the lower variability of the MM train measurements
is due to random chance in the measurements, rather than to an
actual difference.
5.1.2 Inter-method Difference in Measured Mercury
Concentrations
The test data were analyzed for systematic difference in
values measured by the two measurement methods by comparing
the mean value of the pair of measurements made by one
sampling method during a single run with the mean value of the
pair of measurements made by the other sampling method in the
epp.053 43
-------
TABLE 5-2. COMPARISON OF VARIANCES FOR MULTIPLE METALS AND
METHOD 101A MERCURY DATA
Confidence Level for
Critical Value of F
All
With
Conditions
Carbon Injection
Without Carbon Injection
Calculated F
0.
0.
0.
555
508
529
75%
0.680
0.610
0.488
90%
0.476
0.386
0.239
95%
0.385
0.291
0.152
44
-------
same run. This comparison was made for all runs when the two
dual trains were run simultaneously. As shown in Table 5-3,
there were a total of 13 data points, 8 with carbon injection
»
and 5 without carbon injection.
The difference, d^, in measured Hg levels between the two
methods during each run was calculated as:
(5-2)
where V]_j_ and V2i = the measured values from the two dual
trains using Method 101A during the
i-th test; and
Pli and ?2i = "the measured values from the dual MM
trains.
The mean difference was calculated for each test condition by
calculating the mean of the d^'s from runs in that test
condition. Test conditions were then grouped into those with
carbon injection, those without carbon injection, and the
overall data set. The mean difference for each of the grouped
data sets was calculated by averaging the dj/s of all runs and
conditions. As shown in Table 5-3, the average measured Hg
levels are higher using the MM method than when using
Method 101A. The average difference in measured Hg
concentration for all of the data was 39 jug/dscm. The average
difference is 54 jug/dscm during tests without carbon
injection, but is strongly influenced by the large d-[
calculated for Run 3 of Condition 4. The other values of d^
obtained during testing without carbon injection appear
similar to the values measured during testing with carbon
injection. Based on the limited number of data points,
particularly for tests conducted without carbon injection, it
is uncertain whether the differences calculated for subgroups
with and without carbon injection are real or due to small
sample size.
The statistical significance of the difference in
measured values was determined using a t test. The t
statistic for the data was calculated using the formula:
epp.053 45
-------
TABLE 5-3. COMPARISON OF DIFFERENCES IN AVERAGE MERCURY
CONCENTRATIONS FOR MULTIPLE METALS METHOD AND METHOD 101A
Cond.
4
5
6
8
9
Average :
Run
1
2
3
1
2
3
1
2
3
4
1
2
3
1
2
3
w/o Carbon Injection
(4,5)
w/ Carbon
(6,8,9)
Overall
Injection
101A
(jiig/dscm)
296.2
177.4
530.7
521.9
390.7
458.2
—
47.6
27.1
58.5
98. 1
82.8
58.9
180.3
111.7
115.9
439.6
90.3
224.6
MM
(jug/dscm)
346.7
—
705.6
521.3
405.1
490.6
94.2
43.8
33.0
—
101.6
149.0
125.9
185.15
166.4
153.6
493.9
119.8
263.7
di
(Mg/dscm)
-50.4
—
-174.9
0.6
-14.4
-32.3
—
3.9
-5.9
—
-3.5
-66.2
-67.0
-4.8
-54.8
-37.7
Avg d-[
-112.7
-15.4
-1.0
-45.6
-32.4
-54.3
-29.5
-39.0
46
-------
(5-3)
where dm = the mean of the dj/s of each run;
Sp = the standard deviation of the measured
concentrations using the MM method; and
n = the total number of paired samples.
As specified in the validation protocol, the calculated t
statistic was compared to the appropriate critical value for t
at an 80 percent confidence level in each test case. In all
cases, as shown in the upper half of Table 5-4, the calculated
t statistic exceeded the critical value for t at a confidence
level of 80 percent. As a result, the difference between
methods is considered statistically significant in all cases.
The table also shows the tabulated t values at more
stringent confidence levels. These additional t values
represent the highest confidence level for which tabulated
critical t values are lower than the calculated t statistics
for the actual test data. For example, these data indicate
that the difference in mean concentrations for all of the runs
are statistically different at the 99 percent confidence
level.
Since the inter-method bias was shown to be statistically
significant, a correction factor (CF) was calculated according
to the validation protocol for the MM method data using the
equation:
CF =
d»
V
m
where Vm = the mean of the Method 101A Hg measurements.
As shown in the lower half of Table 5-4, the CF is 1.14
for the data without carbon injection (during which Hg levels
typically exceeded 200 /xg/dscm) and 1.49 for the data with
carbon injection (during which Hg levels were less than
epp.053 47
-------
TABLE 5-4.
t STATISTICS AND CORRECTION FACTORS FOR AVERAGE
MERCURY CONCENTRATIONS
Value
Critical Confidence
Value Level
t Statistic
Overall
With Carbon Injection
Without Carbon Injection
Correction Factor3
Overall
With Carbon Injection
Without Carbon Injection
4.106
3.448
2.568
1.21
1.49
1.14
1.356
3.055
1.415
2.998
1.638
2.353
1.1
1.1
1.1
80%
99%
80%
98%
80%
90%
—
—
—
a Correction Factor was calculated using only data
from runs which yielded values from all four sampling trains.
48
-------
200 ^g/dscm). The large variation in these two values
suggests that the CF may vary with Hg concentration, rather
than being constant as is normally assumed. Both values
exceed the criteria in the validation protocol that CF values
should be between 0.9 and 1.1.
5.2 CONCLUSIONS
Although not statistically significant at the 90 percent
confidence level, the calculated precision of the MM method
sample values, as indicated by the sample variances, was
greater than precision of the Method 101A values, both when
grouped by control options (with and without carbon injection)
and for the overall data set. This meets the EPA protocol for
method precision when comparing a proposed to a validated
method. There is a reasonable likelihood that the lower
variance of the MM method is due to random variation and is
not real.
The MM method measured higher Hg concentrations than
Method 101A. Comparison of the calculated inter-method bias
t statistics to the critical values of the t distribution
indicates the difference in measured values is statistically
significant at or above the 90 percent confidence level in all
cases. The fact that the MM method measures higher
concentrations of Hg raises the question of whether all the Hg
in the flue gas is being captured and recovered in the
Method 101A train. This, coupled with the indication that the
MM method appears to have better precision than Method 101A,
may be good reason to further investigate the two methods.
epp.053 49
-------
6.0 SAMPLING AND ANALYTICAL PROCEDURES
This section provides additional information on the
equipment preparation, sampling, and analytical procedures
used with Method 101A and the MM method during the OMSS test
program.
6.1 METHOD 101A
Mercury emissions were tested by Method 101A as specified
in 40 CFR Part 61, Appendix B. The method calls for
isokinetic extraction of flue gas using a sampling train
similar to the Method 5 train. Use of a sampling train filter
is optional; however, for this test program, a filter was used
at all sample locations.
In Method 101A, flue gas is extracted, passed through the
filter, and bubbled through acidified KMnO4. There are
two fractions of the sampling train: the probe rinse/impinger
catch and the sampling train filter. Following sample
recovery, the KMnO4 and filter solutions were shipped back to
the laboratory for analysis. The analytical preparation
procedure consisted of filtering the KMnO4 and filter
solutions and analyzing the filtrates by CVAAS. Because of
concern that the laboratory filtering procedure may remove a
portion of the collected Hg, the laboratory filters were re-
extracted and analyzed separately from the KMnO4 solution, and
the results are reported separately.
The following sections briefly describe the Method 101A
testing procedures. A full description is given in the
reference method located in Appendix A.
epp.053 50
-------
6.1.1 Sampling Equipment
The Method 101A sampling train, including the optional
heated filter, is shown in Figure 2-4. The front half of this
train is similar to Method 5 train, incorporating all
isokinetic sampling apparatus. A glass nozzle/probe liner
unit was used so that the sample stream did not touch any
metal surfaces. The filter was a low-metals glass fiber
filter. Four impingers were used in the train. The first
three contained 50 ml, 100 ml, and 100 ml, respectively, of
acidified 4 percent KMnO4. The last impinger was filled with
silica gel to remove water prior to the sampling train meter
and pump. All reagent preparation followed strict QA/QC
guidelines as dictated in the Method 101A protocol.
6.1.2 Equipment Preparation
All sampling equipment was calibrated in accordance with
EPA Method 5 guidelines. This included dry gas meters, pitot
tubes, and nozzle orifices. All glassware was cleaned as
follows:
• Soaked in 10 percent HNO3 acid bath;
• Rinsed three times with 50 percent HNC>3 ;
• Rinsed three times with tap water;
• Rinsed three times with 8N hydrochloric acid (HC1) ;
• Rinsed three times with tap water; and
• Rinsed three times with deionized/distilled (or
equivalent) water.
Glassware was then sealed with Parafilm™, wrapped in bubble
wrap, packed, and shipped to the test site.
All nozzles and probe liners were cleaned on site between
runs by following the above rinsing procedures. Nozzle
calibration was checked on site.
6.1.3 Reagent Preparation
The following reagents were used during sampling
operations:
• 8N HC1 = 67 ml concentrated HC1/100 ml deionized
(DI) water (H20);
epp.053 51
-------
• 4 percent KMnC>4 = 4 percent solution in ^0 and
H2SC>4; specific instructions for preparation can be
found in Section 2.3.2.
• 50 percent HNO3 = equal parts acid and DI H2O.
Blank samples were taken of all reagents used, to determine if
Hg contamination was present.
6.1.4 Sample Train Operation
The Method 101A sampling train was operated similarly to
an EPA Method 5 train. Care was taken to determine the proper
isokinetic sampling rate below 1.0 cfm. Actual rates were
approximately 0.5 cfm. Temperatures of the stack gas, oven
(filter skin), silica gel impinger, and inlet and outlet to
the gas meter were monitored. Additional recordings of dry
gas meter readings, velocity head (AH), orifice pressure (Ap),
and sample vacuum were taken. The above data were collected
at each sample point every 5 minutes. Total sampling times
for each run were approximately 60 minutes.
Leak-checks of the sampling train were performed prior to
the test, following train removal from a port (port change),
and following completion of the test. The maximum acceptable
leak rate is 0.02 cfm or 4 percent of the average sample rate,
whichever is less.
6.1.5 Sample Recovery
The Method 101A flue gas samples were recovered as shown
in Figure 2-5. The first step after completion of the
post-test leak-check was to dismantle and seal the train into
the following components:
• Probe nozzle and liner,
• Filter holder, and
• Impinger train.
These components were transported back to the laboratory
trailer for recovery operations. The impingers were then
weighed to determine flue gas moisture levels.
The contents of the KMnC>4 impingers were poured into a
950-ml sample bottle. The nozzle and probe were then brushed
and rinsed three .times with fresh 4 percent KMnO4 and placed
epp.053 52
-------
in a 200-ml bottle. Although no visible deposits were
observed during any of the runs, a small amount of 8N HC1 was
used to rinse the glassware used during Conditions 4, 5, and
6. The rinse was placed in a separate 125-ml bottle. The
glassware used during Conditions 8 and 9 was not rinsed with
HC1.
The sampling train filter was carefully placed in a
500-ml sample jar and 50 to 100 ml of fresh 4 percent KMn04
was added. Any residual filter pieces left on the filter
holder were guantitatively removed using a sharpened edge
blade and/or nylon bristle brush and added to this container.
A filter and reagent blank were also collected.
Following recovery operations, the samples were fully labeled,
logged in the sample logbook, and chain of custody forms were
filled out.
6.1.6 Analytical Preparation
After the samples were received by the laboratory, the
chain of custody forms were signed and fluid levels checked to
determine if any sample loss occurred during transport. As
stated in the previous section, three to four sample
containers were generated from each train. The preparation
scheme used for each of these fractions is shown in
Figure 2-6.
Prior to analysis, the front half rinse and impinger
fractions were combined and filtered through a Whatman 40
filter. If an HC1 rinse fraction was generated, it was also
filtered through the same filter. Finally, the filter was
washed with the HCl and KMnC>4 and the rinsings were combined
with the filtrate for analysis.
The sampling train filter sample was transferred to a
beaker, placed in a steam bath, and evaporated down to where
most of the liquid had disappeared (not dryness). Then, 20 ml
of concentrated HNO3 was then added and placed on a hot plate
(with watch glass cover) and heated for two hours at 70°C.
This solution was allowed to cool and then was filtered using
the same laboratory filter discussed above. The filtrate was
epp.053 53
-------
then combined with the front half rinse/impinger sample
filtrate prior to analysis.
At this point, the laboratory filter is normally
discarded. For this test program, however, the laboratory
filter from each run was digested with 25 ml of 8N HC1. The
resulting digestion solution was then filtered, the filter was
rinsed with 8N HC1 and KMn04, and the resulting filtrate was
analyzed separately to determine the fraction of Hg captured
on the laboratory filter.
6.1.7 Analysis
The KMnC>4 samples were brought up to a known volume using
DI water. A sample aliquot was removed and placed in 25 ml of
DI H2O already in the aeration bottle. First, 4 ml of 5
percent KMnC>4 was added, then 5 ml of 15 percent HNO3 was
added, followed by addition of 5 ml of 5 percent KMnC>4. The
solution was mixed thoroughly with the exit arm stopcock
closed. The reducing agents, sodium chloride hydroxylamine
and tin (II), were added as specified in the method and
aeration was initiated. Absorbance was then read using CVAAS
at 253.7 nanometers. The same analysis was done on the lab
filter solution. Each analytical reading had a corresponding
printout of the absorbing Hg peaks. These graphs are included
with the data package in Appendix D.
6.2 MULTIPLE METALS METHOD
Sampling for metals was performed according to an EPA
draft protocol entitled "Methodology for the Determination of
Metals Emissions in Exhaust Gases from Incineration and
Similar Combustion Processes." The protocol is presented in
the Appendix A. Analyses of the OMSS test samples were
performed for Hg as well as As, Cd, Cr, Ni, and Pb. Results
for Hg, Cd, and Pb are presented in this report.
6.2.1 Sampling Equipment
The methodology calls for using the sampling train shown
in Figure 2-7. The seven-impinger train consists of a glass
nozzle and glass probe liner followed by a heated filter
assembly with a Teflon® filter support, a series of impingers,
epp.053 54
-------
the usual Method 5 meterbox, and vacuum pump. The sample is
not exposed to any metals surfaces in this train. The
contents of the sequential impingers are: two impingers with
a 5 percent HNO3/10 percent ^2Q2 solution, two impingers with
a 4 percent KMnC>4/10 percent H2SO4 solution, and one impinger
containing silica gel. Empty knockout impingers are added
both before and after the HNO3 impingers. The impingers are
connected together with clean glass U-tube connectors.
Sampling train components were recovered and analyzed in
separate front and back half fractions according to the
described method.
6.2.2 Equipment Preparation
Glassware was washed in hot soapy water, rinsed three
times with tap water and then rinsed three times with
deionized distilled water. The glassware was then subjected
to the following series of soaks and rinses:
• Soaked in a 10 percent HN(>3 solution for a minimum
of 4 hours,
• Rinsed three times with DI distilled water rinse,
and
• Rinsed with acetone rinse.
The cleaned glassware was allowed to air dry in a
contamination-free environment. The ends were then covered
with Parafilm™. All glass components of the sampling train,
and any sample bottles, pipets, Erlenmeyer flasks, petri
dishes, graduated cylinders, and other laboratory glassware
used during sample preparation, recovery, and analysis were
cleaned according to this procedure.
6.2.3 Reagent Preparation
The acids and H2O2 used were Baker "Instra-analyzed"
grade. The H2O2 was purchased specifically for this test site
and was kept cold until it was opened.
The reagent water was Baker "Analyzed HPLC" grade. The
lot number, manufacturer, and grade of each reagent used was
recorded in the laboratory notebook.
epp.053 55
-------
The HN03/H2O2 absorbing solution and the acidic KMnC>4
absorbing solution were prepared daily according to Sections
4.2.1 and 4.2.2 of the reference method. Each reagent had its
own designated transfer and dilution glassware. This
glassware was marked for identification with a felt tip glass
marking pen and used only for the reagent for which it was
designated.
The analyst prepared the acidic KMnC-4 solution using the
following procedure, beginning at least one day before the
reagent was needed:
• Quantitatively remove 400 ml from a 4-liter bottle
of Baker "Analyzed HPLC" water so that 3.6 liters
remain in the bottle. Label this bottle 4.4 percent
KMnO4 in water.
• Quantitatively add 160 g of KMnC>4 crystals to the
bottle; stir with a Teflon® stirring bar and
stirring plate as thoroughly as possible. This
reagent will be stored on the counter in a plastic
tub at all times.
• Each morning the acidic reagent is needed, decant
900-ml of KMnO4 solution into a 1000 ml volumetric
flask. Carefully add 100 ml of concentrated H2SO4
and mix. This reagent is volatile and must be mixed
cautiously. Hold the flask cap on the flask, mix
once, vent quickly. Complete the mixing slowly
until the mixture is homogenous. Allow the solution
to cool and bring the final volume to 1000 ml with
H2o.
• Carefully filter this reagent through Whatman 541
filter paper into another volumetric flask or
2-liter amber bottle. Label this bottle "4 percent
acidic KMnC>4 absorbing solution." Vent the top and
store the reagent in a plastic tub at all times.
The remaining equipment preparation tasks included
calibration and leak checking of all train equipment as
specified in Method 5. Equipment that was calibrated included
probe nozzles, pitot tubes, metering system, probe heater,
temperature gauges, metering system, and barometer.
6.2.4 Sample Train Operation
The sampling operations used for metals testing are
virtually the same as those listed in EPA Method 5. Detailed
epp.053 56
-------
instructions for assembling the metals sampling train are
found beginning on page 14 of the reference method.
6.2.5 Sample Recovery
The recovery procedures began as soon as the probe was
removed from the stack and the post-test leak check was
completed. To facilitate transfer from the sampling location
to the recovery trailer, the sampling train was disassembled
into three sections: the nozzle/probe liner, filter holder,
and impingers in their bucket. Each of these sections was
capped with Teflon® tape or Parafilm™ before transport to the
recovery trailer.
Once in the trailer, the sampling train was recovered as
separate front and back half fractions, as shown in
Figure 2-8. No equipment with exposed metal surfaces was used
in the sample recovery procedure. The weight gain in each of
the impingers was recorded to determine the moisture content
in the flue gas.
Following weighing of the impingers, the front half of
the train was recovered, which included the filter and all
sample-exposed surfaces forward of the filter. The probe
liner was rinsed with acetone by tilting and rotating the
probe while spraying acetone into its upper end so that all
inside surfaces were wetted. The acetone was quantitatively
collected into a tared bottle. This rinse was followed by
additional brush/rinse procedures using a non-metallic brush
to remove any residual particulate matter; the probe was held
in an inclined position and acetone was sprayed into the upper
end as the brush was pushed through with a twisting action.
All of the acetone and particulate were caught in the sample
container. This procedure was repeated until no visible
particulate remained and finished with a final acetone rinse
of the probe and brush. The front half of the filter was also
rinsed with acetone and brushed until all visible particulate
was removed. After all front half acetone washes were
collected, the cap was tightened, the liquid level marked, and
the bottle weighed to determine the acetone rinse volume.
epp.053 57
-------
The nozzle/probe liner and front half of the filter
holder were then rinsed three times with 0.1N HNC>3 and the
rinse solution was placed into a separate bottle. The bottle
was capped tightly, the weight of the combined rinse recorded,
and the liquid level marked. The filter was placed in a
clean, well-marked glass petri dish and sealed with Teflon®
tape.
The contents of the knockout impinger were recovered into
a preweighed, prelabeled bottle with the contents from the
HN03/H2O2 impingers. These impingers and connecting glassware
were rinsed thoroughly with 0.1N HNC>3, the rinse was collected
in the appropriate impinger contents bottle, and a final
weight was taken.
The impingers that contained the KMnC>4/H2S04 solution
were poured together into a preweighed, prelabeled bottle.
The impingers and connecting glassware were rinsed with at
least 100 ml of the KMnO4/H2S04 solution (from the same batch
used for sampling) a minimum of three times. Rinses were
added to the sample recovery bottle. A final 50 ml 8N HC1
rinse was conducted and placed into the sample recovery
bottle. A final weight was recorded and the liquid level was
marked on the bottle. The bottle cap was replaced loosely
enough to allow venting.
A reagent blank was recovered in the field from each of
the following reagents:
• Acetone blank - 100 ml sample size;
• 0.1N HNO3 blank - 1000 ml sample size;
5 percent HN03/10 percent H2C>2 blank - 200 ml sample
size;
4 percent KMnO4/lO percent 1*2804 blank - 1000 ml
sample size; this blank should have a vented cap;
8N HC1 blank - 50 ml sample size;
Dilution water; and
Filter blank - one each.
epp.053 58
-------
Each reagent blank was from the same lot used during the
sampling program. Each lot number and reagent grade was
recorded on the field blank label.
The liguid level of each sample container was marked on
the bottle in order to determine if any sample loss occurred
during shipment. No sample loss was observed for any of the
samples collected during the OMSS test.
6.2.6 Metals Analytical Procedures
A diagram illustrating the sample preparation and
analytical procedure for the target metals is shown in
Figure 2-9. Approximate detection limits for the various
metals of interest are summarized in Table 2-1.
The front half fractions were digested with concentrated
HNC>3 and HF in a microwave pressure vessel. The microwave
digestion takes place over a period of approximately 10 to 12
minutes in intervals of 1 to 2 minutes at 600 Watts. The
digested filter and the digested probe rinses were combined to
yield the front half sample fraction. The fraction was
diluted to a specified volume with water and divided for
analysis by applicable instrumentation.
The absorbing solutions from the HN03/H2O2 impingers were
combined. An aliguot was removed for the analysis of Hg by
CVAAS, and the remainder was acidified and evaporated to near
dryness. The sample was then digested with 50 percent HNC>3
and 3 percent H2O2 by microwave digestion. After the fraction
had cooled, it was filtered and diluted to a known volume with
water.
Each sample fraction was analyzed by ICP using EPA Method
6010. If iron and aluminum were present, the sample was
diluted to reduce their interferences on Pb. Graphite Furnace
Atomic Absorption Spectroscopy was used to analyze for Pb by
EPA Method 7421. Matrix modifiers, such as specific buffering
agents, may be added to these aliquots to react with and tie
up interfering agents. The total volume of the absorbing
solutions and rinses for the various fractions were measured
and recorded in the field notebook.
epp.053 59
-------
To prepare for Hg analysis by CVAAS, aliquots from the
KMnC>4 impingers, HNO3/H2C>2 impingers, filter digestion, and
front half rinses were digested with acidic reagents at 95°C
in capped BOD bottles for approximately 3 hours.
Hydroxylamine hydrochloride solution and stannous chloride
were added immediately before analysis. The AAS analysis for
Hg followed EPA Method 7470.
6.2.7 Mercury Standards and Quality Control
An intermediate Hg standard was prepared weekly; working
standards were prepared daily. The calibration curve was made
with at least six points. Quality control samples were
prepared from a separate 10 micrograms per milliliter (jig/ml)
standard by diluting it into the range of the samples.
A quality control sample must agree within 10 percent of
the calibration, or the calibration will be repeated. A
matrix spike on 1 of every 10 samples from the HN03/H2O2 back
half sample fraction must be within 20 percent or the samples
will be analyzed by the method of standard addition.
epp.053 60
-------
7.0 QUALITY ASSURANCE/QUALITY CONTROL
Specific QA/QC procedures were strictly adhered to during
this test program to ensure the production of useful and valid
data throughout the course of the project. As discussed in
Section 7.2, the system of QC procedures is a collection of
routine "checks" performed to ensure data quality. The QA
parameters presented in this section are a collection of data
quality indicators used to assess actual data quality.
Detailed QC procedures for all manual flue gas sampling,
process sample collection, and CEM operations are presented in
the OMSS Test Plan.
Section 7.1 presents a brief summary of the test program
quality assurance. Section 7.2 presents the QA/QC definitions
and data quality objectives. Section 7.3 presents manual flue
gas sampling and recovery QA parameters for the outlet
sampling location. Section 7.4 presents method-specific
analytical QA parameters. Section 7.5 discusses data
variability.
7.1 QA/QC SUMMARY
The QA/QC objectives and achievements are summarized in
Table 7-1. Precision, accuracy, and completeness objectives
are presented. All objectives were met. Based on test
design, precision and accuracy results for parameters which
are solely based on field measurements (i.e., flue gas flow,
flue gas moisture, etc.) cannot be determined. As an
indicator of data variability within each test condition, a
coefficient of variation (CV) or an RSD value was also
calculated for all test measurements. These values do not
reflect on the precision of the test measurements because they
epp.053 61
-------
Z
fl«
K^
U
§
1— I
I-J
u
u w
.J
rH
^ M
O W
H 05
W
H Q
Ur-T
55
W »4l
Q >
r»i i_i
2STIMATI
OBJECT]
HH
f T . r ft
HH VJ
o w
z
>H W
05 B
1 ^
G n
w o
*
^
w
CQ
(5
"
W
03
QJ
C
QJ
-P
0)
rH
O
g
o
u
CO
-p
CO
Q
•P
rH ^
Q) """
OH
QJ
>
•H
-P —»
O 0\°
0 >—
•o
ja
o
o
CO
3
O
U
.p
rH
3 *
03 ~
0)
C£
l-H
0)
•H
-P — «
0 o\°
Q) — •
•o
X!
O
.sion
•i— i
u
ft
-P
Q) ^
05
QJ
^
-P — -
O -
•o
*-^
AX
O
,—,
73
O
A
0)
g
•— '
i-i
QJ
4J
Q)
g
CO
(H
cO
0)
O
H ^^
^ OJ
Q rH
w a
g e
"-' cO
W
0)
-P rH
CO O
O i-i
-H 4J
rH C
a o
3 0
• 73
QJ QJ in
rH ^ O
Gi'H -P
g a co
cO 01 SH
03 O
X 43
Q) -iH CO
-P >H rH
CO -P
O CO 4-1
•H g 0
rH
Q473 Q)
QJ C 3
cO
C ^ >
QJ W
QJ g 73
S -^ QJ
-P -P
Q) QJ U
ja x o)
U 01 Q)
c
Q) X U3
r4-H 3
QJ >H 01
73
C 73
-P QJ Q)
C QJ . ti
0) 5 01 3
O 4J -P 03
OJ 42 "3 S
a 01 g
C QJ
•• 0 M -
^i-|H >i
rH 4-J Cn rH
C cfl X C
O -H 0
> C
rH QJ -H rH
cO 73 cO CO
O )-4 U
•H Q) 4J -H
>i CO rH >,
CO QJ -P CO
C > O C
^4 *^4 C
c c o
-H -H -H
(fl 1 ^
>H QJ O
-P rj CO
Q) !H
rH ^ . <4H
CO 73 •
4-1 H Q) rH 03
O -P rH 4->
-P C >
03 C C 03 -H
4J O -H 4J -P
rH -H rH CO
3 -P QJ 3 >-(
03 -H ^ 03 O
QJ 73 Q) Q) 43
r* C £ r4 CO
0 rH
^-»U CM tnrH
W X O
U - C O
J n 3 Q
C g ft
Q) 3 co W
rH O5 73
a c c c
g n O CO O
(0 i-l -H
03 4-> W 73
C-H g QJ
rH O 73 03
o -H q c co
JH -P O QJ 43
-P -H U Q)
C 73 S >i
O C 73 -P U
U O C 0) co
>i 3
O O O
•P >^ C -H cO
m H 3 -p
i-l O i-| CO 73
OH -H C
43 ID > CO
cO 73 Q)
• iH O C 73 C
0) 45 O O
OJ QJ 4-> -H QJ-H
I>1 CO g 'H CO -H
rH M 73 M O
CO QJ < C QJ Q)
C > CU O > i-l
CO ^ W O rfl ft
73 Q) i
rH
(4H
Q)
-H
^1
0)
CO
73
C
CO
Q
X
•H
73
C
Q)
a
<
c
•H
O
01
Q)
fO
01
-P
rH
3
01
QJ
^_i
rH
CO
• 0
rH 4.)
CO rH ^
O cO •
•H c r**
rH CO
a c
a<: o
<: CX-H
-P
4-> rH O
O rH 0)
Z rtj CO
II
rtl W
Z EH
O
45 Z
62
-------
also incorporate plant operation variability as well as
variability in the waste feed itself. Each condition's CV and
pooled CV values are given in Section 7.5.
There were 32 Method 101A runs conducted at the FF outlet
duct (stack). All test runs met the QA/QC isokinetic
criterion of ±10 percent of 100 percent isokinetic. One run
was invalidated due to a sampling problem: a post-test leak
check problem occurred at the stack during Condition 6, Run 1.
The invalidated run was repeated.
There were approximately 116 Method 101A sample fractions
sent to Radian's PPK laboratory for Hg analysis. Sample
control problems with a small number of these fractions
occurred, which resulted in invalidated or modified test
results. Most of these problems were due to illegible labels,
smudged during transit. During Condition 4, the impinger
contents from outlet 101A Run 2R and MM Run 2R were switched
and therefore invalidated data from both trains. A similar
problems occurred for Condition 4 Run 3 with the outlet 101A
HC1 rinse for Train 3R being added to the outlet Run 3 sample.
These results were added together and reported as one run
(Run 3). A brief synopsis of these anomalies are presented in
Table 7-2.
The MM tests were conducted at the stack during 15 runs
(3 runs times 5 conditions). After reviewing the sampling
data, all of these test runs were accepted as valid, with one
run having to be leak corrected with a final (2nd half) leak
check at 0.03 cfm. However, after reviewing the analytical
data, two MM test runs (Condition 5, Run 1 and Condition 8,
Run 2) were invalidated for Pb and Cd due to placement of the
sampling train filters in KMnC>4. All MM test runs met the
isokinetic criterion.
7.2 QA/QC DEFINITIONS AND OBJECTIVES
The overall QA/QC objective is to ensure precision,
accuracy, completeness, comparability, and representativeness
for each major measurement parameter in the test program. For
epp.053 63
-------
TABLE 7-2. SAMPLING, SAMPLE CONTROL, AND ANALYTICAL
ERRORS WITH ASSOCIATED CORRECTIVE ACTIONS
OMSS, CROWS LANDING, CA (1991)
Condition
4
Run
2R &
Method
101A &
Sample
Location
Outlet
Error
KMnO,,
Corrective
Action
Results for
impinger
contents
switched.
both trains
invalidated,
4 3 & 101A &
3R 101A
5 1 & MM (Cd
1R & Pb)
6 1 101A
8 2 MM (Cd
& Pb)
Outlet HC1 rinse
from Run 3R
added to
Run 3 .
Outlet Filters
placed in
KMn04.
Outlet Severely
bad leak
check.
Outlet Filters
placed in
KMnO,,.
Results for
Runs 3 and
3R were
added
together.
Run
invalidated
(Cd & Pb
only) .
Run
invalidated
and
repeated.
Run
invalidated
(Cd & Pb
only) .
64
-------
this test program, QC, QA, and data quality are defined as
follows:
• Quality Control; The overall system of activities
whose purpose is to provide a quality product or
service. The QC procedures are routinely followed
to ensure high data quality.
• Quality Assurance; A system of activities whose
purpose is to provide assurance that the overall
quality control is being done effectively.
Assessments can be made from QA parameters on what
degree of data quality was achieved.
« Data Quality; The characteristics of a product
(measurement data) that bear on its ability to
satisfy a given purpose. These characteristics are
defined as follows;
Precision; A measure of mutual agreement among
individual measurements of the same property,
usually under prescribed similar conditions.
Precision is best expressed in terms of the
standard deviation and in this report will be
expressed as the RSD or CV.
Accuracy; The degree of agreement of a
measurement (or an average of measurements of
the same thing), with an accepted reference or
true value.
Completeness; A measure of the amount of valid
data obtained from a measurement system
compared with the amount that was expected to
be obtained under prescribed test conditions.
Comparabi1itv; A measure of the confidence
with which one data set can be compared with
another.
Representativeness; The degree to which data
accurately and precisely represent a
characteristic of a population (actual
condition).
7.3 MANUAL FLUE GAS SAMPLING AND RECOVERY PARAMETERS
The following section reports method-specific sampling QA
parameters used to assess the quality of emissions test data
produced from manual tests during the test program.
7.3.1 Mercury by Method 101A Sampling Quality Assurance
Successful completion of the post-test leak-checks
ensures that no dilution of the sampled stack gas was
epp.053 65
-------
occurring during the test. Leak-checks were completed .after
completion of each change in the flue gas duct. All reported
101A test runs at the outlet met the leak rate criterion,
except Condition 6, Run 1. Data from this run was invalidated
and the test repeated. No results were reported for the
invalidated run. All leak check results are shown on the
field data run sheets shown in Appendix B.
Tables 7-3 presents the isokinetic sampling rates for the
101A outlet sampling trains. The acceptance criterion is that
the average sampling rate must be within 10 percent of
100 percent isokinetic. All test runs met the isokinetic
criterion.
All dry gas meters are fully calibrated every 6 months
against an EPA intermediate standard. The full calibration
factor or meter Y is used to correct the metered sample volume
to true sample volume. To verify the full calibration, a
post-test calibration is performed. The full and post-test
calibration coefficients must be within 5 percent to meet
Radian's internal QA/QC acceptance criterion. As can be seen
from Table 7-4, the post-test calibration factors for meter
boxes used for all manual flue gas testing were well within
5 percent of the full calibration factor.
Field blanks were collected to verify the absence of any
sample contamination. A sample filter field blank in KMnO4
and a KMnC>4 solution field blank were analyzed. Relatively
small amounts of Hg were detected in the filter/KMnO4 and
KMnC>4 blanks; 0.7 and 0.4 total pq, respectively (58.8 and
31.3 total ng/L). These levels are very small compared to
test run values which averaged 124 total pg at the outlet.
(Average outlet FH/BH fractions = 120 total jug; average outlet
filter catch fractions = 3.5 total A*g.) Because of the small
amount of Hg in the sample blanks as compared to the total
sample, and because there is unknown consistency of any Hg
contamination in the reagent blanks, no blank corrections were
employed. Analytical blank results are discussed in
Section 7.4.
app.053 66
-------
TABLE 7-3. ISOKINETIC RESULTS FOR THE STACK MERCURY (101A)
TESTS; OMSS, CROWS LANDING, CA (1991)
Isokinetic Rates
Condition
4
4R
5
5R
6
6R
8
8R
9
9R
Run 1
96.2
101
102
98.5
98.1
98.9
101
97.8
97.5
96.8
Run 2
98.3
96.9
99
102
101
101
99.2
95.7
106
96.4
Run 3
97.3
97.6
102
98.3
102
99.4
107
97.4
103
100
Run 4a
--
—
—
—
--
100
--
—
--
—
a A fourth run was only conducted for Condition 6
67
-------
TABLE 7-4. DRY GAS METER CALIBRATION CHECK;
OMSS, CROWS LANDING, CA (1991)
Meter Box
I.D. No.
N-31
14
7
A-36
8
Sampling Full
Location Calibration
Factor
Stack, FFb
Stack
Stack
Stack
Stack
1.
0.
1.
1.
1.
0060
9973
0022
0254
0065
Post-Test
Calibration
Factor
1
0
0
1
.0385
.9590
.9845
.0307
Deviation3
(%)
3
-3
-1
-0
.2
.8
.8
.5
NCC
Post Test - Full
Full
x 100
b FF = Fabric Filter Inlet
c NC = Not Completed
68
-------
7.3.2 Multiple Metals Sampling Quality Assurance
All MM leak-checks passed the leak-check criterion of
0.02 cfm, except for Condition 8, Run 3 at the stack. A post-
test leak rate of 0.03 cfm was measured. The sample volume
for this run was leak corrected according to method protocols.
The isokinetic sampling rates for the MM trains are listed in
Table 7-5. All isokinetic values were within 10 percent of
100 percent.
The post-test dry gas meter calibration checks for boxes
used for the MM sampling are shown in Table 7-5. The results
are well within the 5 percent acceptance criterion.
The Hg analyses were completed on three MM reagent blank
samples. No Hg was detected on the HNO3 rinse solution,
HN03/H2O2 impinger solution, or the DI with reagent blanks.
The detection limits were <0.6, <0.5, and <1.0 total /xg,
respectively (<2.0 pg/L for each of the three fractions).
7.4 ANALYTICAL QUALITY ASSURANCE
The following section reports QA parameters for the Hg
101A and MM analytical results.
7.4.1 Mercury by Method 101A Analytical Quality Assurance
The analysis of all flue gas impinger and filter samples
was completed at Radian's PPK laboratory. An EPA-approved
modification to the Method 101A sample preparation procedure
was incorporated into the analytical protocol for this test
program. This was based on recent information revealing that
possible removal of sampled Hg could be occurring during the
laboratory filtering process. Therefore, the laboratory
filters were also analyzed and the results included into the
Hg emission calculations. A breakdown of the amounts of Hg
collected in each sample fraction is given in Table 3-1. The
analysis was completed using normal Method 101A protocols
employing CVAAS. More detail on these procedures is presented
in Section 2.2.
Laboratory method blanks were analyzed to verify the
absence of Hg contamination originating in the laboratory.
Table 7-6 presents the results from those analyses. Out of
epp.053 69
-------
TABLE 7-5. ISOKINETIC RESULTS FOR THE STACK MULTIPLE
METALS TESTS; OMSS, CROWS LANDING, CA (1991)
Isokinetic Rates
Condition Run * Run 2 Run 3
4 99.3 101 103
4R 94.8 96.9 98.1
5 105 106 103
5R 99.2 97.6 97.0
6 99.1 103.7 105
6R 99.1 99.3 98.4
8 96.9 106 97.3
8R 102 96.5 96.9
9 104 100 105
9R 92.4 102 99.6
70
-------
TABLE 7-6. MERCURY 101A METHOD BLANK RESULTS
OMSS, CROWS LANDING, CA (1991)
Analytical
Batch
(Condition)
1-4, 6
3, 7-8
7, 8
6, 8
7
1-4, 6
6-8, 3
7, 8
7, 8, 3
7, 8
Sample
Type
101A
Train
101A
Train
101A
Train
101A
Train
101A
Train
Lab
Filter
Lab
Filter
Lab
Filter
Lab
Filter
Lab
Filter
Total
No. of
Samples
Run
35
26
25
25
8
35
22
22
22
18
238
No. of
Method
Blanks
6
4
5
5
2
0
4
4
4
4
44
No. of
Positive
Detections
0
0
0
0
0
0
0
0
0
0
0
Average
Detection
Limit
(total /ug)
<7.3
<4.1
<22.5
<18.6
<16.9
<0.02
<0.03
<0.01
<0.01
<0.01
<6.9
Note: See Analytical Results in Appendix D for Individual Method
Blank Values.
71
-------
the 44 method blanks analyzed (1 for every 10 samples
analyzed), no detections were found. The results for each
blank analysis are given in the analytical results in
Appendix D.
Possible analytical matrix interferences were
investigated by conducting matrix spike analyses. A known
amount of Hg was added to a series of samples and the recovery
calculated. Radian's internal QA criterion for matrix spike
recoveries is ±20 percent of 100 percent. Table 7-7 presents
the Method 101A matrix spike recovery results. All recoveries
were within 20 percent of 100 percent. In addition to matrix
spike (MS) analyses, matrix spike duplicate (MSD) analyses
were also completed. The QA criterion for MSD recoveries
results are within 10 percent of the MS value. The individual
MSD results are listed in Appendix D. All MSD results met the
QA criterion.
Laboratory Control Samples (LCS) were also analyzed to
verify the continuing accuracy of the spectrometer calibration
curve. A known concentration of Hg prepared from a source of
Hg separate from the calibration Hg stock was submitted for
analysis. Acceptable results of LCS analyses were to be
within 10 percent deviation from the actual LCS value.
Table 7-8 presents the Method 101A LCS range of results. All
LCS samples met the acceptance criterion. Individual LCS
results are given in Appendix D.
7.4.2 Multiple Metals Analytical Quality Assurance
Table 7-9 presents the method blank metals results for
the Multi-Metals flue gas samples. Results are given for both
the FH and BH fractions. A small amount of Hg was found in
one of the method blanks analyzed with the Condition 5
samples. This value was 5.9 total jig for a FH fraction.
Typical FH amounts of Hg collected in the test samples were
less than 1 jug. (Total FH/BH Hg collected averaged 181 total
/ig.) Therefore, the level of Hg found in the method blanks
does not appear to be significant. Lead was not detected in
epp.053 72
-------
TABLE 7-7. MERCURY 101A MATRIX SPIKE RESULTS
OMSS, CROWS LANDING, CA (1991)
Analytical
Batch
(Condition)
1-4, 6
3, 7-8
7, 8
6, 8
7
1-4, 6
6-8, 3
7, 8
7, 8, 3
7, 8
a MSD = Matrix
Note: Accer
Sample
Type
101A Train
101A Train
101A Train
101A Train
101A Train
Lab Filter
Lab Filter
Lab Filter
Lab Filter
Lab Filter
Total
No. of
Samples
Run
35
26
25
25
8
35
22
22
22
18
238
No. of
Matrix Spikes
(including
MSD)a
6
4
4
4
2
4
6
4
2
4
40
Range of
Recovery
Values (%)
85.3-104
94.5-102
94.5-102
95.2-103
94-98
96.8-99.9
88.6-107
92.4-103
89.9-96
97.8-102
Spike Duplicates
)tance criterion for
matrix spikes is
±20 percent.
The MSD criterion is ±10 percent of the duplicate
result. (See Analytical Results in Appendix D for
individual MS and MSD values.)
73
-------
TABLE 7-8.
MERCURY 101A LABORATORY CONTROL SAMPLE RESULTS
OMSS, CROWS LANDING, CA (1991)
Analytical
Batch
(Condition)
1-4, 6
3, 7-8
7, 8
6, 8
7
1-4, 6
6-8, 3
7, 8
7, 8, 3
7, 8
Sample Type
101A Train
101A Train
101A Train
101A Train
101A Train
Lab Filter
Lab Filter
Lab Filter
Lab Filter
Lab Filter
Total
No. of
Samples
Run
35
26
25
25
8
35
22
22
22
18
238
No. of
LCS Samples
3
4
7
8
3
3
4
4
4
4
44
Range of
Recovery
Values (%)
94.8-106
96.2-105
95.8-106
93.2-106
97.3-101
93.8-101
99.7-103
92-110
92-101
91.5-107
LCS = Laboratory Control Sample acceptance criterion is
±10 percent of 100.
74
-------
Q
2
<
w"
H
ft
W
X
H
K W
^ £-• ^
Jg
2 ~~"
CQ fX
Q 3 r,T
O CO S
S H "^
W 2 ^
< 8 w
EH S
W ^ §
S U
W EH v
t* S co
L_* /*\ W
M O ^i
EH CQ §
S ^
S
CTi
|
ps.
w
CQ
S
"
^,
tf,
QJ
•rH
QJ
w
X
-H
^1
-p
IS
r-j ^^
/rt 'rt
•H 3
CQ
iH
T3 (0
O 4-1
x; o
-P 4J
0) ^_
!£
Analytical
Q
w
x --.
s
Q
CO
•o s
u -^
s
Q
W
-Q S
ft ~-v
s
X
u
^
ft
to
c
o
-H
-p
o
(0
ft,
Batch
(Condition)
,, •* o n ^ „
""^CTiH^^H^CTiH -^
fasffgf^g ~
H H 2 H 2 CO H H CTN ^
• • o • ^
in H H r- 0
>1
0
-p
no o vo H
H H H CO O
•^ >«». i i ^> o> i I i -Q
vo O O ~-~- ™
OH or- "J
H H H CT*
^•^MOJOCNCOI-
o.JooininHH
v^'vvvvvv
VO H (N (N
• 220022
O O V V
^ " < <; ^ ^ ,< <
00220022
V V V V
££ggggg§g
CO CO CO CO CO
VO ^D VD VO VO
^ *. *» v
IO LO LO LO
CN in CNJ
o^ ^o o^
n
0 1 1
H
n
0 1 1
X S g
CQ ^ ^J
CO
cr\ ^ cr\
vo
in in
d)
(0
O
rH
ft
Q
QJ a;
f*£ >^£
•H -H
a a
o x x
0) Tj -H -H
^^ fTj '*T t| li
f_^ C ty'i f^i J J i J
(13 -H r- 1 t, fl3 fO
x « (0 £ s s
X
-P 0) II
C X3 ^ O II
0 0 U C Q
M )-( (0 S W CO
fe ft CQ W 2 S
II II II II
x K x s aJ
fci ft CQ & -P
0
(0 2
75
-------
any of the method blanks, however, a small amount of Cd
(approximately 0.3 jig) was found in two samples.
Table 7-9 also presents the MS results for the metals
analyses. All spiked recoveries were within the QA allowance
of ±20 percent of 100 percent. All MSD recoveries were within
10 percent of the MS value for Pb, Cd, and Hg except one BH Hg
sample (MS = 106 percent, MSD = 94 percent). With 17 MSD
recoveries meeting the 10 percent criterion, and only one MSD
value not meeting it, the MM analytical QA appears to be
acceptable.
LCS values are also shown in Table 7-9. All results are
within 10 percent of 100 percent thereby meeting the QA
criterion.
7.5. DATA VARIABILITY
Simple CV values are presented in Table 7-10. These
values do not reflect on the precision of the sampling and
analytical method since they do not compare duplicate trains
as was done in previous sections. The following values are
presented as indicators of data variation within each test
condition. Pooled CV values are also presented.
7.5.1 Overview
Coefficients of variation were calculated for all the
final stack gas pollutant concentrations. The CV or RSD is
calculated by dividing the standard deviation by the mean and
expressed as a percentage. The CV values expressed in the
following tables are not intended to represent
sampling/analytical precision. They are more a reflection of
the variability of the data as a whole, including process
caused emission variability, as well as variability in the
waste feed. The CV values presented here should not be
compared to any acceptability criterion. They are only shown
to provide insight into the variability of the data.
The CV values for each test condition are calculated as
follows:
epp.053 76
-------
2
O
EH
s
EH
W
U
«z
o
CJ
CO
o
a
fe~
EH rH
W (7>
EH H
O
W <
W U
EH
K O
0 2
fci H
Q
2 2
EH
H |5
« O
< OH
> U
,
O CO
CO g
EH 0
M
M
o
H
fc.
W
Q
CJ
0
W
OQ
EH
C
O
-p
(0
o
o
I-H
c
&
g
(0
.p
0)
1 I
4-1
3
O
™ tt~
(0 r^oX0
« K ~
0)
r_j
O
•H ^
3 *""
g
a ^
< *~
rH
O
,—)
Kt!.
£! ^
^ — '
O t_x
^^
* — •
N *—<
O oV>
CJ ^
Oj'-
aT^
•^
O d?
fc-
73
C
O
CJ
CN f~- CTi C"1 O VD
VD C*^ ^* VO ^J* O^
^ rH
"* ^ ^ ^ CO ^
•^ H n rH W CN
^ r^- ^* cj\ oo
,J "* o H r-
00 H H n H
en en en vo o ^j*
r** vo v^ ^ L^) [^
CN CN in in CN **
in in t-* o „ »*
^t a\ en ^ .^ CN
CN ^* H r- «*
oo r- VD r- o, o
CN in in o _J H
in n CN ^o •^<
d VD CO rH rH VO
^t o o n ^ ^
in in vo CN co CN
m H CN m H »*
p* ** vo en n co
r- in H CN in in
co co ^t en CN n
CN n co in in in
•P 73
0) 0)
«• in vo co en £ -£
3 O
O &i
en
en
(T
73
0)
C
•H
A
e
o
CJ
*
CN
•
O
1 1
•4-*
73
OJ
4-J
O
0)
O
O
>n
UJ
C
O
•H
-P
(C
4-1
q
CO
q
0
o
q
o
73
0)
J2
CO
V-t
CO
W
CO
3
rH
(0
rH
rH
<
• •
EH
O
2
77
-------
CV = — x 100 (7-1)
M V '
where CV = Coefficient of variation;
S = Standard deviation (using n-1 in the
denominator where n = number of data
points); and
M = mean.
The CV's from several distinct groups of data can be combined
into a "Pooled CV." Pooled CV's presented for all test
conditions are calculated as follows:
cvp =
£«.
(7-2)
where CVp = pooled coefficient of variation;
CVi = Coefficient of variation for a simple
sample set i; and
ni = Number of data points in that sample set.
7.5.2 Test Program Data Variation
Tables 7-10 presents the CV's for all measured flue gas
concentrations at the APCD outlet. Values are presented for
each individual condition which was composed of two to four
runs. Condition CV values are presented for each parameter at
the outlet sample location. Pooled CV values are presented
for each parameter inclusive of all test conditions.
Duplicate trains (side by side nozzles/trains) were
operated during five test conditions for both the Method 101A
Hg train and the MM train. Assessments of precision of the
101A train and the MM train was completed in Sections 3.0 -
5.0. The CV values for Hg concentrations are presented here
without any statistical comparisons made.
epp.053 78
------- |