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

-------