EPA-600/4-77-050
December 1977
Environmental Monitoring Series
COLLABORATIVE STUDY
OF EPA METHOD ISA
AND METHOD 13B
ironmental Monitoring and Support Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL MONITORING series.
This series describes research conducted to develop new or improved methods
and instrumentation for the identification and quantification of environmental
pollutants at the lowest conceivably significant concentrations. It also includes
studies to determine the ambient concentrations of pollutants in the environment
and/or the variance of pollutants as a function of time or meteorological factors.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/4-77-050
COLLABORATIVE STUDY
OF EPA METHOD ISA AND METHOD 13B
by
William J. Mitchell
Quality Assurance Branch
Jack C. Suggs
Statistical and Technical Analysis Branch
Environmental Monitoring and Support Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
and
Fred J. Bergman
Midwest Research Institute
425 Volker Boulevard
Kansas City, Missouri 64110
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Environmental Monitoring and Support Laboratory
Research Triangle Park, North Carolina 27711
September 1977
-------
ABSTRACT
Described are the results from a collaborative test of U.S. Environmental
Protection Agency Method 13 at a primary aluminum plant. This test method is
used to measure the fluoride emissions from primary aluminum plants and phos-
phate fertilizer plants. In the collaborative test, six laboratories simultane-
ously sample the same stack using two Method 13 sampling trains for a total of
twelve Method 13 samples per sampling run. Ten such sampling runs were accomp-
lished for a total of 120 samples. Each source sample was analyzed for fluoride
content using Method 13A (SPANDNS spectrophotometric procedure) and Method 13B
(ion selective electrode procedure).
The collaborative test results showed that the two methods gave similar
results. The within-laboratory standard deviation in milligrams fluoride per
cubic meter for Method 13A was 0.044 and for Method 13B was 0.037. Similarly,
the between-laboratory standard deviation for Method 13A was 0.064 and for
Method 13B was 0.056. These estimates include sampling and analysis error.
Based on the analysis of aqueous fluoride samples by each collaborator,
the bias in milligrams of fluoride per liter was -0.08 for Method 13B and
-0.10 for Method 13A.
Also presented in the report are precision estimates for the two methods
in terms of repeatability and reproducibility. A discussion of how to use
these collaborative test results for evaluating source testing results is also
given.
111
-------
IV
-------
CONTENTS
Abstract i i i
Figures vi
Tabl es vi i
1. Introduction 1
2. Summary and Conclusions 5
3. Design of the Collaborative Test 8
4. Collaborative Test Site 12
5. Selection of Collaborators 17
6. Test Equipment 19
7. Conduct of the Collaborative Test 32
8. Statistical Analysis of the Results 36
References 46
Appendices
A. Bid Package Sent to Collaborators A-l
B. Statistical Methodology B-l
C. Using Collaborative Test Data to Evaluate Test Results. C-l
-------
FIGURES
Number Page
1 Method 13 Sampling Train 2
2 Sampling Port Locations 11
3 Stack Extension 13
4 Velocity Profile in Stack (m/sec) 14
5 Control .Consoles on Catwalk 4.6 Meters Below Roof 15
6 Paired Train Sampling Box 20
7 Upper and Lower Sampl ing Probes 21
8 Probe Clamp. (2 per Cluster) 22
9 Double Pitot Sampling Arrangement: (Top) Side View;
(Bottom) Upstream View 23
10 Scaffolding Used to Support Sampling Assembly 24
11 Roller Assembly for Moving Sampling Assembly Into Stack 25
12 Nozzle and Pitot Orientation During Sampling 26
13 Impinger Transport Box with Two Paired Train Boxes 27
14 Probe Transport Box 28
A-l Alcoa's Badin, North Carolina, Baghouse A-5
A-2 Double Pitot Sampling Arrangement: (Top) Side View;
(Bottom) Upstream View A-7
A-3 Probes A-l 2
B-l Distribution Curves Corresponding to Differences B-6
-------
TABLES
Number Page
1 Sampling Location Assignments by Collaborator No 10
2 Meter Box Calibration Check Results 30
3 Sampling Results for Method 13A in mg F/Dry Std. M3 37
4 Sampling Results for Method 13B in mg F/Dry Std. M3 38
5 Analysis of Unknown Standard Solution (2 mg F/£) 39
6 Analysis of Variance and Variance Component Estimation
(Method 13A) 41
7 Analysis of Variance and Variance Component Estimation
(Method 13B) 42
8 Precision Estimates of Methods 13A and 13B 43
9 Analysis of Variance of Analytical Results from Standard Sample . 45
I Tentative Schedule for Collaborative Test of Method 13 A-8
vii
-------
SECTION 1
INTRODUCTION
The Environmental Protection Agency (EPA) requires that the fluoride
emissions from phosphate fertilizer plants and aluminum reduction plants be
123
sampled using the Method 13 sampling train which is shown in Figure 1. ' '
Further, unless it can be shown to the satisfaction of the Administrator that
the fluoride is completely water soluble and can be accurately analyzed with-
out distillation, these EPA regulations require that the samples collected be
fused with sodium hydroxide and distilled from sulfuric acid prior to analysis.
Im :effect, this means that most phosphate fertilizer plants can analyze their
samples without fusion and distillation, but that the aluminum plants must
subject their samples to the fusion and distillation procedures before analyzing
them for fluoride. The regulations permit analysis by either the SPADNS spectro-
photometric analytical procedure (Method 13A) or the ion selective electrode
analytical procedure (Method 13B).
Previous work performed on aluminum plant and phosphate fertilizer plant
4
samples by Mitchell and Midgett determined that the fusion and distillation
procedures were the predominant source of imprecision in the Method 13
analytical procedures. Thus, the challenge that the Method 13 analytical pro-
cedures would receive from aluminum plant samples would be much more severe
than the one it would receive from phosphate fertilizer plants. In addition,
the fluoride emissions from aluminum plants are also much lower in concentration
-------
SAMPLING
PROBE
100ml
DISTILLED, DEIONIZEO
WATER
Figure 1. Method 13 sampling train
-------
than those from fertilizer plants. Thus, the sample recovery technique should
receive a more severe challenge at an aluminum plant than at a phosphate
fertilizer plant. For these reasons, we decided that the best place to
adequately evaluate Method 13 through a collaborative test would be at a
primary aluminum reduction plant.
However, the cyclic nature of the primary aluminum reduction process,
Mtself, causes the fluoride concentration in the stack to fluctuate with time.
Thus, the Environmental Protection Agency (EPA) regulations applicable to
2
primary aluminum reduction plants require that the potroom emissions be
sampled for a minimum of eight hours per sampling run using the Method 13
sampling train (Figure 1). These regulations also require that.the stack
cross section be divided into equal concentric areas and that each area be
sampled for an equal period of time during the eight hours of sampling. This
latter requirement, which is termed traversing the stack, is employed because
the fluoride concentration also may not be homogeneously distributed across
the stack cross section.
Now, the objective in conducting collaborative testing of a source test
method is to determine the performance of the method when it is used by
qualified laboratories. Further, a primary requisite for a collaborative
test is that all laboratories sample and analyze essentially identical samples.
This is because the precision of the method is frequently determined from a
statistical comparison of the results obtained by each of the collaborators to
the mean obtained on that sample by all the collaborators. Also, to ensure
that the estimates of precision obtained are reliable, it is usually required
that the precision estimate obtained have at least 30 degrees of freedom
associated with it.
-------
Because of: (1) the above requirements for a collaborative test; (2) the
non-homogeneity in the fluoride concentration across the stack; (3) the random
fluctuations in the fluoride concentration over an eight hour time interval;
and (4) the unusually long sampling time required by the regulation, it was
evident that it would not be cost-efficient to use the standard collaborative
test techniques previously developed for collaboratively testing stationary
678
source test methods ' ' to conduct a collaborative test of Method 13. The
eight hour sampling time alone would require about three weeks of field test-
ing to obtain a statistically adequate number of samples. And of course, the
unstable nature of the fluoride concentration in the stack with time and with
position would make it tenuous to assume that samples taken at the same point
in the stack but at different times would be statistical replicates.
Thus, we decided to employ the four train sampling arrangement previously
4 9
developed and field tested by Mitchell and Midgett. ' This technique, which
employs fixed-point sampling has been shown to yield essentially statistical
replicates when applied to stacks in which a fluctuating or non-homogeneous
pollutant concentration exists. For the collaborative test of Method 13, we
employed six laboratories and three, four train sampling arrangements. Ten,
3-hour sampling runs were accomplished with each laboratory operating two
sampling trains in each run, for a total of 12 samples per run. Two runs
were accomplished each day.
-------
SECTION 2
SUMMARY AND CONCLUSIONS
This report presents the results of a collaborative test of the U.S.
Environmental Protection Agency (EPA) Method 13 that was conducted at a primary
aluminum plant. The collaborative test employed six laboratories with each
laboratory simultaneously operating two identical Method 13 sampling trains in
each of ten sampling runs. Since each laboratory simultaneously sampled the
stack, a total of twelve replicate samples per run were obtained.
At the conclusion of the collaborative test each test participant return-
ed his samples to his own laboratory for analysis for fluoride. In the labo-
ratory each sample was fused with sodium hydroxide and distilled from sulfuric
acid. Then aliquots from each distillate were analyzed by Method 13A (SPADNS
spectrophotometric procedure) and by Method 13B (ion selective electrode pro-
cedure). In this manner, estimates of the within-laboratory and the between-
laboratory precision of Methods 13A and 13B (including sampling error) were
obtained.
Estimates of the accuracy of Methods 13A and 13B were obtained by having
each collaborating laboratory perform duplicate analyses on an aqueous sodium
fluoride standard solution. These samples were fused and distilled as de-
scribed in the methods.
The precision and accuracy estimates obtained are delineated below. Un-
less otherwise noted, these precision estimates include sampling error and
pertain to the determination of a single sampling run result and not to the
5
-------
average of three results that is specified in the performance test for compli-
ance section of the Federal Register:
Method 13A
The within-laboratory standard deviation — 0= 0.044 mg fluoride per stand-
ard cubic meter with 60 degrees of freedom —was estimated from the differ-
ence between the two trains that were operated by the same laboratory and was
determined to be independent of fluoride concentration in the range measured.
The between-laboratory standard deviation—a. = 0.064 mg fluoride per standard
cubic meter with 5 degrees of freedom ——was estimated from the within-run
differences between the results of the six participating laboratories. The
bias in the analytical procedure —based on the analysis of the standard
fluoride sample — was determined to be -0.10 mg fluoride per liter.
Method 13B
The within-laboratory standard deviation —-CT= 0.037 mg fluoride per stand-
ard cubic meter with 60 degrees of freedom—was estimated from the differences
between the two trains that were operated by the same laboratory and was
determined to be independent of fluoride concentration in the range measured.
The between-laboratory standard deviation-^oi = 0.056 mg fluoride per stand-
ard cubic meter with 5 degrees of freedom, was estimated from the within-run.
differences between the results of the six participating laboratories. The
bias in the analytical procedure — based on the analysis of the standard
fluoride sample —— was determined to be -0.08 mg fluoride per liter.
An alternate way to report precision of a test method is in terms of
Mandels' repeatability (a measure of the within-laboratory precision) and
reproducibility (a measure of the between-laboratory precision). For our
purposes, repeatability will be defined as a quantity that will be exceeded
-------
only about 5 percent of the time by the difference, taken in absolute value,
of two randomly selected test results obtained by the same laboratory on
replicate samples. Reproducibility will be defined as a quantity that will
be exceeded only about 5 percent of the time by the difference, taken in
absolute value, of two single test results made on identical samples by two
different laboratories, that is, two labs sampling at the same port.
Thus, if we define a test result as a single sampling run, the repeat-
ability of Methods 13A and 13B is 0.123 mg fluoride per dry standard cubic
meter and 0.102 mg fluoride per standard cubic meter, respectively. The
analogous reproducibility estimates are 0.259 mg fluoride per standard cubic
meter and 0.241 mg fluoride per standard cubic meter.
-------
SECTION 3
DESIGN OF THE COLLABORATIVE TEST
The collaborative test was designed so that the six laboratories each
operated two Method 13 sampling trains identical to the one shown in Figure 1.
All laboratories sampled for fluoride at an isokinetic sampling rate at a
point very near the center of the stack.
The sampling arrangement used in the test involved three sets of
clustered sampling trains. Each cluster consisted of four Method 13 sampling
trains, two "S" type pitot tubes, and two laboratories. The pairing of the
laboratories was switched between runs, so that each laboratory was paired-up
with every other laboratory twice during the collaborative test. Further, each
sampling train was operated independently of the other trains with the exception
that the two trains operated by the same laboratory used the same pitot tube for
setting and maintaining an Isokinetic rate.
Table T shows the location of each laboratory in each of the ten runs and
Figure 2 shows the orientation of the sampling ports. By employing this test
design, it is possible to obtain reliable estimates of the within-laboratory
and between-laboratory precision of Method 13.
Because the sampling, fusion and distillation procedures of the Methods
13A and 13B analytical procedures are identical, it was possible to obtain
separate estimates of the precision of Method 13A and of Method 13B using the
same source sample. This was accomplished by having each laboratory take
-------
every source sample through the fusion/distillation procedure and then analyze
separate aliquots from the distillate using the SPADNS (Method 13A) and the
ton selective electrode (Method 13B) procedures.
Estimates of the accuracy of the two analytical! methods were obtained by
giving each laboratory two identical aqueous samples of sodium fluoride to
analyze using the complete Method ISA and Method 13B analytical procedures.
-------
TABLE 1. SAMPLING LOCATION ASSIGNMENTS BY COLLABORATOR NO.
Run No.
1
2
3
4
5
6
7
8
9
10
SOUTH
A
104
104
102
102
103
106
105
102
105
102
PORT
B
105
102
105
106
102
102
102
104
104
103
WEST
C
102
105
101
101
106
105
103
103
101
105
PORT
D
101
103
103
105
105
101
101
105
102
106
NORTH
E
106
106
104
104
104
103
106
101
103
101
PORT
F
103
101
106
103
101
104
104
106
106
104
10
-------
WEST PORT
SOUTH
PORT
NOT
EAST PORT
USED
NORTH
Figure 2. Sampline Port Locations,
11
-------
SECTION 4
COLLABORATIVE TEST SITE
The collaborative test of EPA Method 13 was conducted at the Aluminum
Company of America (Alcoa) primary aluminum reduction plant in Badin, North
Carolina. The plant operates 24-hours per day, seven days a week. The
fluoride emissions from its reduction pots are controlled by passing these
emissions through an alumina, fluidized-bed scrubber and then through a bag-
house. After leaving the baghouse, the emissions exit to the atmosphere
through a 1.67 meter in diameter, 1.7 meter high stack on the roof of the
baghouse. The baghouse units are shaken on a 3-hour cycle.
The collaborative test was done on one of the fourteen such scrubber/bag-
house units at the plant. However, prior to the actual test, a 2.6 meter long
by 1.67 meter in diameter stack extension was placed on the stack that was to
be sampled during the collaborative test (Figure 3). When the stack extension
was installed, the four, 15 cm in diameter sampling ports were approximately
1.9 meters above the roof or approximately 4.5 meters above the top of the bags
in the baghouse.
Figure 4 shows a velocity profile across the stack at the sampling ports.
This profile was obtained just prior to the first sampling run in the colla-
borative test. During the collaborative test the twelve sampling nozzles
were located very near the center of the stack. The twelve meter boxes were
located on the catwalk that is attached to the side of the baghouse building
12
-------
15.2cm
I.D. PORTS
GUIDEWIRE
ALL DIMENSIONS IN METERS UNLESS NOTED
OTHERWISE
Figure 3. Stack Extension.
13
-------
7
6
5
3
2
NORTH PORT SOUTH PORT
I I i I I
r\
0.3 0.6 0.9 1.2 1.5
STACK DTAMETER (meters)
* «
1
~>- EAST PORT
r
WEST PORT
I I
~\
0 0.3 0.6 0.9 1.2 1.5
STACK DIAMETER (meters)
Figure 4, Velocity Profile 1n Stack.
14
-------
Figure 5. Control Consoles on Catwalk 4.6 Meters Below Roof
15
-------
about 4.6 meters below the roof line (Figure 5). The roof of the baghouse
building extends over the entire width and length of the catwalk. The catwalk
itself is accessible by means of a 21.5 meter aluminum rung, cage enclosed
ladder. Access to the roof from the catwalk is accomplished by means of a
4.6 meter aluminum rung ladder that extends through a hole cut in the roof.
During the collaborative test the equipment was transported from the
ground to the catwalk using an electric-powered hoist. The equipment was then
manually moved from the catwalk to the roof through the hole cut in the roof
for the 4.6 meter ladder.
16
-------
SECTION 5
SELECTION OF COLLABORATORS
The six laboratories that participated in the collaborative test were
selected from a list of 25 laboratories who had expressed interest in partic-
ipating in the collaborative test. To qualify for consideration as a partic-
ipant each of these 25 laboratories was required to analyze four fluoride -
containing samples using at their option either Method 13A (SPADNS spectro-
photometric procedure) or Method 13B (ion selective electrode) and submit the
analytical results with their estimate of what it would cost for them to
participate in the collaborative test. Each fluoride sample was fused with
sodium hydroxide and distilled from sulfuric acid prior to analysis by the
above analytical procedures. The thirteen laboratories that actually analyzed
the samples were compensated for the cost of the analysis to a maximum of
$125.00.
The analytical results submitted by these thirteen laboratories were
subjected to a comparative statistical analysis and those laboratories that
demonstrated competence with the Method 13 analytical procedure were ranked on
the basis of cost. The six lowest bidders were then selected to participate.
17
-------
COLLABORATORS
The collaborators who participated in the sampling part of the collab-
orative test of Method 13 were:
Name Organization
John Haslbeck T.R.W. Environmental Engineering
Billy McCoy Vienna, VA
David Huckabee Entropy Environmentals Inc.
Frank Phoenix Raleigh, NC
Fred Lucree Scott Environmental Technology
Joseph Wilson Plumsteadville, Pa.
Dale Huddleston Aluminum Company of America
Jeff Bishop Pittsburgh, Pa.
Larry Meyers Kaiser Aluminum and Chemical Company
Glenn Wai den Pleasanton, Ca.
Pete Watson Commonwealth Laboratories, Inc.
Kim Thompson Richmond, Va.
NOTE: Throughout the remainder of this report, the collaborating laboratories
are referenced by randomly assigned code numbers as Lab 101 through Lab 106.
These code numbers do not necessarily correspond to the above ordered listing
of collaborators.
The collaborative test was conducted under the general supervision of
Mr. Fred Bergman of Midwest Research Institute, Kansas City, Missouri and
Dr. William J. Mitchell, U.S. Environmental Protection Agency, Environmental
Monitoring and Support Laboratory, Quality Assurance Branch, Research Triangle
Park, North Carolina 27711. They had the overall responsibility for assuring
that the test was conducted in accordance with the collaborative test plan and
that the collaborators adhered to the Method 13 sampling procedure. They were
assisted by Mr. Nick Stich of Midwest Research Institute.
18
-------
SECTION 6
TEST EQUIPMENT
EQUIPMENT SPECIFICATIONS
The impinger and filter holder portion of the two trains that were operated
by the same laboratory were placed in a paired-train sampling box similar to
that shown in Figure 6. Each box was 35.5 cm long by 45.7 cm wide by 33.0 cm high.
The dimensions of the sampling probes used in the collaborative test are
shown in Figure 7. The four probes and two pitot tubes used in each sampling
cluster were held in place by the probe clamp assembly shown in Figure 8.
Figure 9 shows the top and side view of an assembled sampling cluster.
The slotted angle scaffolding and the roller assembly on which the sampling cluster
was moved into and out of the stack are shown in Figures 10 and 11. Figure 12
shows the location and orientation of the twelve sampling nozzles in the stack
itself.
Figures 13 and 14 show views of the impinger transport boxes and the
probe transport boxes, respectively. Each probe transport box could hold 16
sampling probes and each impinger transport box could hold two paired-train boxes.
The other equipment used in the collaborative test is delineated in the bid
package and follow-up letter sent to each participant. The bid package is reproduced
in Appendix A.
PRETEST CALIBRATION REQUIREMENTS
Improper calibration of test equipment and changes in the calibration
factor after laboratory calibration and prior to the test can be sources of
error in stack testing programs. To minimize the chance for mi sealibrated or
out-of-calibration equipment to affect the test results, the collaborators
were required to provide Midwest Research Institute with the calibration
19
-------
"lt
Figure 6. Paired Train Sampling Box
20
-------
ro
T
i
8.6
1
0.89 cm I.D.*-
^^^^H
6
*-
¥
r
r
.0
9.0 1 ft>
7.2 >
[,
h
M
GLASS LINER— >.
™i ( ^**K
UPPER PROBE A \
GLASS LINER — ^
^jk *^ m.
L X LOWER PROBE
^v.
^^2.66 O.D. PROBE SHEATH
•*-6.0~*
••••^^•k
]
L-
w
^:
v
in
u.
I
f
2.5-5.0
j
J
t
2.5-5.0
J
NOTE - ALL DIMENSIONS IN CENTIMETERS
SIZE 28/15 OR r
28/12 BALL (GLASS)
Figure 7. Upper and Lower Sampling Probes.
-------
0.95
CM
CM
NOTE - ALL DIMENSIONS IN CENTIMETERS
Figure «. Probe Clamp. (2 per Cluster).
-------
8.6
• 0.0-
NOZZLE
K
•7.2-
6.0
2.B
Z2>
N°ZZLE
ROBE SHEATH
PITOT TUBE
1
3.2 M-2.4
"S" TYPE PITOT TUBE A
'it
i ' y^^^
TRAIN
3.5
TRAIN
2
TRAIN
"3
4.8—J
"S" TYPE PITOT TUBE B
NOTE - ALL DIMENSIONS IN CENTIMETERS
Figure 9. Double P1tot Sampling Arrangement: (Top) Side View;
(Bottom) Upstream View.
23
-------
.
Figure 10. Scaffolding Used to Support Sampling Assembly.
24
-------
Figure 11. Roller Assembly for Moving Sampling Assembly Into Stack
25
-------
no
en
CENTERLINE OF STACK CROSS SECTION
4.8
f°
in
PITOT
O
( )
PITOT
NOZZLE. 0.89 cm I.D.
CENTERLINE OF STACK
CROSS SECTION
NOTE - ALL DIMENSIONS IN CENTIMETERS
Figure 12. Nozzle and Pi tot Orientation During Sampling.
-------
Figure 13. Impinger Transport Box with Two Paired Train Boxes
-------
Figure 14. Probe Transport Box
28
-------
factors for their "S" type pi tot tube, nomographs and control consoles prior to
the start of the test. The console calibrations were carried out as specified
by Rom , the pi tot tubes were calibrated in the absence of the sampling probe
as described in the proposed Revised Method 2 ; and the accuracy of the nomo-
12
graphs was checked using the method of Shigehara .
As a further check on the console calibration and to determine if any
changes in calibration had occurred during shipment, each console was sub-
jected to a two-point calibration check at the U.S. Environmental Protection
Agency, Environmental Research Center, Research Triangle Park, North Carolina.
The calibration check was done using either a 600 liter spirometer or a dry gas
meter that had been calibrated against the spirometer. The results are presented
in Table 2.
The control consoles were checked at orifice pressure drops of 1.8 and 3.0
inches of water, which was the expected range to be used during the collaborative
test.
Only the two consoles operated by Collaborator 105 were found to be signifi-
cantly out of calibration for both the orifice calibration and the dry gas meter
calibration (Table 2). These two meter boxes were adjusted and recalibrated
against the spirometer. All the other control consoles were found to be within
the presently allowed specifications for the dry gas meter coefficient (y = 1.00 +_
0.02), but three of these consoles had stated orifice coefficients that differed
by more than + 5% from the value determined in the EPA laboratory. The respective
collaborators recalibrated these three consoles against the spirometer and used
this recalibrated value in the collaborative test.
Prior experience had shown that commercially available nomographs are
not always reliable, due to misalignment of the various scales on the nomo-
29
-------
TABLE 2. METER BOX CALIBRATION CHECK RESULTS
Collaborator
Number
101 .
102
103
104
105
106
Box
Number
1
2
1
2
1
2
1
2
1
2
1
2
Orifice
Found
1.65
1.65
2.09
2.12
2.05
2.01
1.72
1.73
1.95
1.53
1.73
1.63
Coefficient
Stated
1.63
1.73
1.93
1.93
1.86
1.92
1.73
1.74
2.27
1.85
1.64
1.70
Gas Meter
Found
0.991
0.980
0.989
0.980
1.005
0.995
1.01
0.999
1.07
1.09
1.00
0.993
Coefficient
Stated
0.982
0.995
0.99
i.oq
1.01
1.00
0.995
0.992
0.991
0.992
0.994
0.99
130
-------
graphs. Thus, prior to shipping their equipment each collaborator was in-
12
structed to check the accuracy of his nomographs by the method of Shigehara .
Further, to preclude an inaccurate nomograph from accidently being used in the
collaborative test, each laboratory was assigned a problem to be worked using
his nomographs at the EPA laboratory. The results of this nomograph check
showed that all nomographs were within acceptable accuracy.
The pi tot tubes to be used in the collaborative test were checked in the
EPA laboratory to be sure they were not physically deformed or dented and to
ensure that the thermocouple was not within 1.9 cm (0.75 inches) of the
pressure sensing ports.
After the calibration and equipment checks were completed the equipment
was placed in a truck and transported to Badin, North Carolina —- a distance
of approximately 183 km (110 miles) from Research Triangle Park, North Carolina.
Equipment Leak Check
Fully-assembled sampling trains were leak-checked by plugging the tip of
the probe nozzle with a septum and pulling a 370 torr vacuum on the entire train.
To pass the leak check, the leak rate under these conditions had to be less
o
than 0.0006 m /min.
In this manner each train was leak-checked before insertion into the stack
and immediately after removal from the stack.
31
-------
SECTION 7
CONDUCT OF THE COLLABORATIVE TEST
Fixed-point stack sampling was employed in the collaborative test. In
each run, the sampling tip of each nozzle was located no further than 8 cm
from the center of the stack. Each sampling cluster consisted of four Method
13 sampling trains, two "S" type pi tot tubes and two laboratories (Figure 9).
Within each cluster each laboratory monitored the stack velocity using their
own "S" type pitot tube and sampled at an isokinetic rate with two of the
four trains in that cluster. For example, if Laboratories 101 and 102 were
using the sampling arrangement shown in Figure 9, Laboratory 101 would use
trains 1 and 2 and pitot tube A and Laboratory 102 would use trains 3 and 4
and pitot tube B.
All twelve sampling trains (4 trains per cluster, 3 clusters) simultan-
eously sampled for fluoride. This was accomplished by inserting the first
cluster through the north port, the second cluster through the south port,
and the third cluster through the west port of the stack (Figure 2). The
actual pairing of laboratories within a cluster varied from run to run to
ensure that every laboratory was paired-up with every other laboratory twice
during the collaborative test.
Of course, with twelve sampling trains sampling within a rectangular
area 16 cm by 13 cm, physical blockage of the stack or misalignment of the
sampling clusters could adversely affect the velocity determinations by one
or more of the pi tot tubes.
32
-------
To determine if physical blockage of the stack area would affect the
velocity measurement, the following study was done before the actual collab-
orative test was initiated:
(1) An isolated "S" type pitot tube was inserted through the east port
and positioned so that its pressure sensing ports were at the center of the
stack.
(2) The velocity head measured by the isolated pitot tube was monitored
as the three sampling clusters were inserted into the stack and located in the
position they would have during the collaborative test.
(3) The velocity head measured by the isolated pitot tube was also
monitored as the sampling clusters were rapidly withdrawn from the stack.
No blockage effect was found, that is, the velocity head measured by the
isolated "S" type pitot tube was not visually affected by the presence of the
three sampling clusters.
To preclude misalignment of the sampling clusters from occurring during
the test, the following procedure was followed before each sampling run:
(1) The nozzle/pitot orientations were visually checked before the
sampling clusters were inserted into the stack.
(2) The first cluster was inserted through the north port and positioned
near the center of the stack.
(3) Then as the second cluster was brought into position through the
south port, the velocity head measured by the two pi tots in the other cluster
was monitored.
(4) Finally, as the third cluster was brought into position through the
west port, the velocity head measured by the two pitot tubes nearest the west
port was compared to the appropriate velocity head measured by the two pitot
33
-------
tubes nearest the east port.
(5) After all clusters were inserted, the test supervisor made a visual
inspection of the orientation of the clusters in the stack by looking through
the unused east port. (It should be noted that in no case was the velocity
head measured within a cluster affected as the other clusters were inserted.
This demonstrates that physical blockage of the stack was not significant
enough to affect the velocity.)
At the conclusion of the first run of that day, the twelve sampling probes
were placed in the probe transport box (Figure 14). Similarly, two, paired-train
sampling boxes were placed in each impinger transport box (Figure 13). After
the sampling trains for the second run were assembled, leak-checked and in-
serted into the stack, the probes and impinger boxes from the first run of
that day were moved in their respective transport boxes to the clean-up
trailers on the ground. At the end of the second run for that day, the
impinger boxes and the probes from this run were placed in transport boxes and
moved to the ground. All sample train assembly, disassembly and sample recovery
operations were done in trailers located near the base of the stack.
By utilizing the above sampling scheme the work scheduled for each day was
achieved in about 10-1/2 hours.
At the conclusion of the collaborative test, the recovered samples were
returned to the participating laboratories for analysis by Methods ISA and 13B.
That is, each source sample was fused and distilled using the Method 13 pro-
cedure and aliquots from the same distillate were analyzed by both the SPADNS
spectrophotometric procedure (Method 13A) and the fluoride ion selective
electrode procedure (Method 13B).
34
-------
As a means to check the accuracy of the analysis, each collaborating
laboratory was also given two aqueous sodium fluoride standard solutions that
contained two mg fluoride per liter and were instructed to analyze these sam-
ples using the complete Methods 13A and 13B. Since both samples were identical
in concentration, the results of these analyses are actually a duplicate anal-
ysis on identical samples. However, the collaborators were not aware that the
two samples were identical, nor were they aware that the samples contained
only sodium fluoride.
35
-------
SECTION 8
STATISTICAL ANALYSIS OF THE RESULTS
Precision
For the statistical analyses, the six laboratories were randomly designated
as Labs 101, 102, 103, 104, 105 and 106. The collaborative test consisted of
ten sampling runs with twelve samples collected in each run for a total of 120
samples. Each source sample was fused and distilled and then separate aliquots
from each sample were analyzed for fluoride using Methods ISA and 13B. In
addition, each laboratory analysed two aqueous sodium fluoride standard solutions
that each contained 2 mg fluoride per liter.
The actual sampling results reported by each laboratory are presented in
Table 3 (Method 13A) and Table 4 (Method 13B). The analytical results on the
fluoride standard solution are reported in Table 5. The results in Tables 3, 4
and 5 were checked for calculation errors and no significant errors were found.
The results in Tables 3 and 4 were subjected to an analysis of variance
(ANOVA) to determine the precision of the two methods. Precision of the methods
deals with the differences between observations made under similar conditions
as determined by the experimental design. For the purpose of the ANOVA, the
results in Tables 3 and 4 were considered to have been gathered using a balanced,
incomplete block design. The design was considered incomplete in the sense that
a sampling port could accommodate only two laboratories in a sampling run. With 15
possible laboratory pairings from six laboratories and with three ports avail-
36
-------
TABLE 3. SAMPLING RESULTS FOR METHOD 13A IN mg F/DRY STD. M3.
Run No.
1
2
3
4
5
6
7
8
9
10
Train
Position
front
rear
front
rear
front
rear
front
rear
front
rear
front
rear
front
rear
front
rear
front
rear
front
rear
Pair
Port A
Lab 104
0.19
0.21
Lab 104
0.18
0.19
Lab 102
0.81
0.84
Lab 102
0.77
0.76
Lab 103
0.47
0-.47
Lab 106
0.45
0.40
Lab 105
0.46
0.41
Lab 102
0.32
0.35
Lab 105
0.35
0.34
Lab 102
0.26
0.27
1
Port B
Lab 105
0.32
0.29
Lab 102
0.23
0.23
Lab 105
1.18
1.16
Lab 106
1.07
1.02
Lab 102
0.43
0.44
Lab 102
0.34
0.20
Lab 102
0.39
0.32
Lab 104
0.39
0.44
Lab 104
0.33
0.25
Lab 103
0.41
0.45
Pair
Port C
Lab 102
0.22
0.24
Lab 105
0.33
0.35
Lab 101
0.87
1.00
Lab 101
1.06
1.11
Lab 106
0.46
O.b7
Lab 105
0.60
0.50
Lab 103
0.52
0.48
Lab 103
0.59
0.68
Lab 101
0.41
0.41
Lab 105
0.49
0.49
2
Port D
Lab 101
0.27
0.30
Lab 103
0.36
0.35
Lab 103
1.18
1.12
Lab 105
1.40
1.29
Lab 105
0.64
0.67
Lab 101
0.45
0.56
Lab 101
0.38
0.51
Lab 105
0.61
0.59
Lab 102
0.32
0.34
Lab 106
0.41
0.44
Port E
Lab 106
0.24
0.23
Lab 106
0.33
0.26
Lab 104
0.88
0.80
Lab 104
1.10
1.14
Lab 104
0.56
0.56
Lab 103
0.40
0.43
Lab 106
0.32
0.37
Lab 101
0.33
0.36
Lab 103
0.33
0.42
Lab 101
0.34
0.37
Pair 3
Port F
Lab 103
0.23
0.20
Lab 101
0.18
0.29
Lab 106
0.87
0.90
Lab 103
1.04
1.23
Lab 101
0.52
0.51
Lab 104
0.49
0.38
Lab 104
0.22
0.33
Lab 106
0.36
0.35
Lab 106
0.30
0.27
Lab 104
0.33
0.32
37
-------
TABLE 4. SAMPLING RESULTS FOR METHOD 13B IN mg F/DRY STD. M3.
Run No.
1
2
3
4
5
6
7
8
9
10
Train
Position
front
rear
front
rear
front
rear
front
rear
front
rear
front
rear
front
rear
front
rear
front
rear
front
rear
Pair
Port A
Lab 104
0.25
0.21
Lab 104
0.24
0.23
Lab 102
0.96
0.86
Lab 102
0.77
0.92
Lab 103
0.46
0.47
Lab 106
0.40
0.38
Lab 105
0.46
0.42
Lab 102
0.38
0.39
Lab 105
0.36
0.34
Lab 102
0.32
0.31
1
Port B
Lab 105
0.32
0.29
Lab 102
0.25
0.23
Lab 105
1.18
1.16
Lab 106
0.97
1.00
Lab 102
0.45
0.46
Lab 102
0.37
0.32
Lab 102
0.39
0.33
Lab 104
0.41
0.44
Lab 104
0.33
0.23
Lab 103
0.40
0.45
Pair
Port C
Lab 102
0.24
0.24
Lab 105
0.33
0.36
Lab 101
0.81
0.92
Lab 101
1.00
1.02
Lab 106
0.50
0.55
Lab 105
0.60
0.50
Lab 103
0.51
0.49
Lab 103
0.60
0.67
Lab 101
0.41
0.37
Lab 105
0.50
0.49
2
Port D
Lab 101
0.27
0.30
Lab 103
0.36
0.35
Lab 103
1.18
1.14
Lab 105
1.40
1.29
Lab 105
0.64
0.67
Lab 101
0.48
0.52
Lab 101
0.40
0.47
Lab 105
0.64
0.59
Lab 102
0.35
0.35
Lab 106
0.42
0.43
Port E
Lab 106
0.21
0.23
Lab 106
0.29
0.25
Lab 104
0.90
0.90
Lab 104
1.10
1.16
Lab 104
0.57
0.56
Lab 103
0.39
0.45
Lab 106
0.35
0.36
Lab 101
0.32
0.34
Lab 103
0.33
0.42
Lab 101
0.32
0.35
Pair 3
Port F
Lab 103
0.23
0.20
Lab 101
0.18
0.19
Lab 106
0.80
0.87
Lab 103
1.01
1.23
Lab 101
0.50
0.51
Lab 104
0.47
0.38
Lab 104
0.24
0.39
Lab 106
0.36
0.33
Lab 106
0.27
0.26
Lab 104
0.33
0.38
38
-------
TABLE 5. ANALYSIS OF UNKNOWN STANDARD SOLUTION (2 mgF/£).
Collaborative No. Method ISA
101 1.80
1.64
102 1.90
1.90
103 2.01
2.01
104 - 2.00
1.80
105 1.93
1.85
106 2.08
1.87
Method 13B
1.89
1.73
1.85
1.80
1.97
1.97
2.20
2.00
1.90
1.80
1.98
1.90
39
-------
able for each run, five sampling runs were needed to do the experiment. The
entire experiment was then repeated.
A linear model for this experiment is:
Yijkl = y+ RUNi + PORT (RUN)./.) + LABk +
where
u = a constant,
RUNi = runs effect (random) i = 1,2 ---- , 10,
PORT (RUNh/^x » sampling port effect (random) j = 1,2,3,
Even though the ports remain stationary, the effect may change
randomly with each run,
LABk = laboratory effect (random) k = 1,2,3,4,5,6
ERROR-i/-ji.> = residual error (random). In the analysis of variance for this
model, the error is split into two parts: intra-port error and
sub-sampling error. Intra-port error is a measure of the differ-
ence between laboratories sampling through the same port. Sub-
sampling error is the difference in the two measurements made
by the two sampling trains operated by the same laboratory. Of
course, the analytical portion also contributes to the error, but
there is no way of separating this from sampling error.
In an experiment of this type, primary interest centers on estimating variance
components. The analysis of variance results and the expected mean squares for
Methods 13A and 13B are presented in Tables 6 and 7, respectively. (The actual
statistical analysis and an explanation of the terms is presented in Appendix B.)
40
-------
TABLE 6. ANALYSIS OF VARIANCE AND VARIANCE COMPONENT ESTIMATION (METHOD 13A)
Source
Run
Port (Run)
Lab (adj. for port effects)
Intra-port Error
Subsampling Error
Total :
D.F.
9
20
5
25
60
119
S.S. M.S. EMS
8.7930 .97770 o 2 + 2oj2 + 4^Rj + 12a2R
.6860 .03430 a 2 + 2aj2 + 4ap2(R)
.2825 .05650 a 2 + 2aj2 + 20(.6)a|_2
.1786 .00714 a 2 + 20j2
.1182 .00197 a2
10.0583
41
-------
TABLE 7. ANALYSIS OF VARIANCE AND VARIANCE COMPONENT ESTIMATION (METHOD 13B)
Source P.P. S.S.
Run 9 8.6501
Port (Run) 20 .6039
Lab (adj. for port effects) 5 .2279
M.S. EMS
.9611 a2 + 2
.0302 a2 + 2 ^ + 4
2
+ 4 2p(R) + 12 2R
(R)
.0456
+ 2 + 20(.6)
Intra-port Error
25
.1836
.0073
Subsampling Error
60
.0819
.00137 a
Total:
119
9.7474
42
-------
TABLE 8. PRECISION ESTIMATES OF METHODS 13A and 13B.
Variance Components
Subsampling (a2)
Intra-Port (aj2)
o
Laboratory (a)
Port UP(R))
Method ISA
.00197
.00259
.00411
.00679
Method 13B
.00137
.00297
.00319
.00573
Standard Deviations of Differences Between Single Observations
Within-Laboratory: ad =\/2^ ' .063
Between-laboratory;
r— /o 2 2
T n V T (.. .1 Ot
1. Same port
JdL
2. Different ports adpL = >/2va2 + oj2 + aL2 + ap
P(R)
.176
.052
.123
.162
43
-------
The precision of a test method deals with the closeness of observations
repeated under similar circumstances and is measured by the standard deviation
of the differences. These standard deviations are reported in Table 8 and
are defined in Appendix B.
Accuracy
To determine the accuracy of the analytical procedures in Methods 13A
and 13B, the true value of 2.0 mg fluoride per liter was substracted from
the values reported in Table 5. Then a one-way ANOVA was performed, on the
results to determine the accuracy.
The results of this one-way ANOVA, which are summarized in Table 9,
shows that the accuracy did not change significantly from laboratory to
laboratory. Although, the overall average for each method is slightly less
than 2.0 mg per liter, the 95 percent confidence limit contains 2.0 mg per
liter in each case. Thus, we cannot conclude that either analytical method
has a significant bias.
44
-------
TABLE 9. ANALYSIS OF VARIANCE OF ANALYTICAL RESULTS FROM STANDARD SAMPLE.
Method 13A
Source
Laboratories
Error
D.F.
5
6
S.S. F
.1005 2.062 N.S.
.0585
Method 13B
Source
Laboratories
Error
D.F.
5
6
S.S. F
.1224 3.478 N.S.
.0423
Total:
11
.1590
11
,1647
N.S. = not significant at the a = .05 level
overall mean = 1.90mg/£
overall std. dev. = .12 mg/a
95% conf. limit for mean [1.80, 2.00 mg/Ji]
overall mean = 1.92 mg/a,
overall std. dev. = .12 mg/SL
95% conf. limit for mean [1.83, 2.03 mg/si]
45
-------
SECTION 9
LIST OF REFERENCES
1. U.S. Environmental Protection Agency. "Standards of Performance for
New Stationary Sources (Amendments to Reference Methods ISA and 13B)",
Federal Register, 41., 52229-52230, November 29, 1976.
2. U.S. Environmental Protection Agency. "Standards of Performance for
New Stationary Sources (Primary Aluminum Industry)", Federal Register,
41, 3825-3830, January 26, 1976.
3. U.S. Environmental Protection Agency. "Standards of Performance for New
Stationary Sources (Phosphate Fertilizer Plants). Federal Register. 40,
33152-33166, August 6, 1975.
4. Mitchell, W. J. and Midgett, M.R. "Adequacy of Sampling Trains and
Analytical Procedures Used for Fluoride", Atm. Envir., 10, 865-872, 1976.
5. Mandle, J. "Repeatability and ReprodudbiHty," Materials Research and
Standards, 1]_, 8-16» 1971-
6 Hamil, H. F. and R. E. Thomas. Collaborative Study of Particulate
Emissions Measurements by EPA Methods 2, 3, and 5 Using Paired
Particulate Sampling Trains (Municipal Incinerators). U.S. Environ-
mental Protection Agency, Research Triangle Park, North Carolina.
Report No. EPA-600/4-76-014. 1976.
7. Hamil, H. F. and D. E. Camann. Collaborative Study of Method for the
Determination of Particulate Matter Emissions from Stationary Sources
(Portland Cement Plants). U.S. Environmental Protection Agency, Research
Triangle Park, North Carolina. Report No. EPA-650/4-74-029. 1974.
8. Hamil, H. F. and R. E. Thomas. Collaborative Study of Method for the
Determination of Particulate Matter Emissions from Stationary Sources
(Municipal Incinerators). U.S. Environmental Protection Agency, Research
Triangle Park, North Carolina. Report No. EPA-650/4-74-022. 1974.
9. Mitchell, W. J. and M. R. Midgett. Method for Obtaining Replicate
Particulate Samples from Stationary Sources. U.S. Environmental Pro-
tection Agency, Research Triangle Park, North Carolina. Report No.
EPA-650/4-75-025. 1975.
10. Rom, J. S. "Maintenance, Calibration and Operation of Isokinetic Source
Sampling Equipment", U.S. Environmental Protection Agency, APTD-0576,
March 1972.
46
-------
11. U.S. Environmental Protection Agency, "Standards of Performance for
New Stationary Sources (Amendments to Reference Methods)", Federal
Register, 41_, 23060-23090, June 8, 1976.
12. Shigehara, R. T. "Adjustments in the EPA Nomograph for Different Pitot
Tube Coefficients and Dry Molecular Weights." Stack Sampling News, 2,
4-11, October 1974.
47
-------
APPENDIX A
BID PACKAGE SENT TO COLLABORATORS
-------
August 16, 1976
This letter is in response to your reply to our letter dated April 2, 1976.
The plans for conducting the collaborative test of EPA Methods 13A (SPADNS)
and 13B (Selective Ion Electrode) have been finalized.
The test is tentatively scheduled to begin on October 12, 1976. Complete
details are given in the Work Plan which is enclosed.
You have been selected as a potential participant from those responding to
the original contact. If you cannot or choose not to participate in the
collaborative test, please let MRI know by August 20, 1976, by calling
Mr. Fred Bergman at 816-561-0202, extension 261.
The following information is addressed to those who wish to participate:
Submit to MRI no later than September 7, 1976 a firm fixed-price pro-
posal for supplying the equipment and manpower to perform the test outlined
in the enclosed Work Statement. This will include (1) fixed-price bid for
the collaborative test, (2) bid for cost per day extra, (3) copies of cali-
bration curves, and (4) analytical results.
You will be sent four (4) fluoride samples and a supply of SPADNS under
separate cover. Two samples will be labeled as solid. These samples must
be fused and distilled with duplicate analysis conducted on the distillate.
Two samples will be labeled liquid which will require distillation and du-
plicate analysis. The report of the analytical results should contain the
fluoride values obtained during aliquot determination (Methods 13A and 13B,
Section 7.3.4) as well as the final fluoride concentration after distillation.
Analysis may be conducted by utilizing either 13A or 13B. Those laboratories
electing to demonstrate their analytical capabilities using both 13A and 13B
will receive preference during collaborator selection over bidders using only
one method. Regardless of the methods used during bidding, the use of both
13A and 13B during the collaborative test will be required. For those bidders
planning on using Method 13B (SIE) it should be pointed'out that the lifetime
of the fluoride electrode has been found to be less than one year. SIEs older
than this should be replaced before conducting the analysis.
A-2
-------
In summary, the analysis conducted during bidding will require a total
of two fusions, four distillations and eight (one method) or sixteen (two
method) analysis.
To partially compensate bidders for the time spent on analysis, all
bidders who submit the required analytical results will receive a fixed
sum of $125.00.
If you have any questions concerning the proposed test, you may contact
either Fred Bergman or Paul C. Constant at 816-561-0202.
Sincerely,
MIDWEST RESEARCH INSTITUTE
Fred Bergman
Senior Chemist
Enclosures: 5
A-.3
-------
WORK STATEMENT FOR COLLABORATIVE TEST OF EPA METHOD 13
I. Background Information
Midwest Research Institute is conducting a program on the evaluation
and collaborative testing of those U.S. Environmental Protection Agency (EPA)
Test Methods that apply to stationary sources. We are now making preparation
to conduct the collaborative test of EPA Method 13. You are invited to submit
a bid that delineates what it will cost Midwest Research Institute for you to
participate in the collaborative test.
II. Test Site
The selected test site is the Alcoa primary aluminum plant in
Badin, N.C. Badin is located approximately 40 miles east of Charlotte, N.C.
The Alcoa plant, which operates 24-hours per day, seven days a week, controls
its fluoride emissions with an alumina, dry-bed scrubber. After leaving the
scrubber, the emissions pass through a baghouse before exiting to the atmos-
phere through a 66-in. diameter stack on the roof of the baghouse. Our col-
laborative test will be conducted on one of the fourteen such scrubber/bag/house
units on the plant.
There is a catwalk located on the side of the baghouse at a point
fifteen feet below the roof. The roof of the building extends over the entire
width of the catwalk. The catwalk is accessible by means of a 70-ft aluminum
rung, cage-enclosed ladder. In addition, an electric-powered hoist is available
to transport equipment from the ground to the catwalk. Access to the roof from
the catwalk is by means of a 15-ft aluminum ladder.
Figure A-l. Alcoa's Badin, North Carolina, Baghouse.
III. Test Plan
Fixed-point, isokinetic sampling will be employed during the col-
laborative test. Presently we plan to have six participating laboratories
each sample simultaneously with two sampling trains for a total of twelve
samples per run. Ten such sampling runs, each 3-1/2 hr in length, will be
made. Two sampling runs per day will be accomplished for a total of five
actual days of testing.
The sampling arrangement used in the test will involve three sets
of clustered sampling trains. Each cluster will consist of four Method 13
A-4
-------
Rain
Cover
Roof Slopes Approx._
3/8" per 4 Ft. Length
t
15'
TI i i i i i * ' » '
70'
Figure A-l.. Alcoa's Badin, North Carolina, Baghouse,
A-S
-------
sampling trains, two "S" type pitot tubes and two laboratories (Figure 2).
Within a cluster each laboratory will monitor the stack velocity using one
of the "S" type pitot tubes and will sample isokinetically with two of the
four trains in that cluster. For example, if two laboratories, A and B, were
using the sampling arrangement shown in Figure 2, Laboratory "A" would use
trains 1 and 2 and the pitot tube nearest to these trains and Laboratory "B"
would use trains 3 and 4 and the other pitot tube.
During the test all meter boxes will be located on the catwalk
fifteen feet (15 ft) below the roof. The sample boxes for the impingers
will be located on the roof. These sample boxes will be supplied by Mid-
west Research Institute.
At the conclusion of each sampling run, the sample boxes and the
probes will be capped and moved to a clean room for recovery of fluoride.
At the conclusion of the collaborative test, the samples will be returned to
the laboratories of the participants for analysis by Methods 13A and 13B
(including the fusion and distillation steps of this procedure). A copy of
each analytical procedure is attached to this bid package. Only one fusion
distillation will be required per sample, because aliquots from the same
distillate are to be analyzed using both the SPADNS and the SIE procedures.
IV. Time Period
As mentioned above, the test will consist of 10, three and one-
half (3-1/2) hours sampling runs made over a five-day period. This scheme
will yield four samples per laboratory per day for a total of 20 samples
for analysis. Testing is tentatively scheduled to start on Monday, October 18
and conclude on Friday, October 22, 1976.
However, the complexity of both the test scheme and the sampling
equipment make it necessary to conduct a pre-test orientation session and
a pre-test dry run. This orientation session will be held on Wednesday
and Thursday, October 13 and 14, 1976, in an EPA facility in Research Triangle
Park, North Carolina. At this time, the accuracy of each laboratory cali-
bration of their meter boxes will be spot-checked against a 600-liter spiro-
meter located in this same EPA facility.
Thursday, October 14, will be used to transport the sampling equip-
ment to Badin, N.C. (approximately 110 miles from Research Triangle Park).
It is anticipated that only one truck will be necessary to transport the equip-
ment to Badin. Midwest Research Institute (MRI) will supply this truck. Col-
laborators will be expected to supply their own transportation to Badin, N.C.
Friday, October 15, will be used for equipment set-up. The dry run
will also be accomplished on this day.
Table A-l summarizes the above schedule.
A-6
-------
•9.C—
8.6
ROBE SHEATH
NOZZLE
-7.2-
n
6.0
v:
2.5
a & NOZZLE I A
dJ
I PITOT TUBE
"S" TYPE PITOT TUBE A
3.2 U 2.4
^tx
02=.
2.4
3.S
TRAIN
1
4
TRAIN
2
TRAIN
3
F- - l^J X^i^X Lj
4.8 J
I f "S"
()
TYPE PITOT TUBE B
NOTE - ALL DIMENSIONS IN CENTIMETERS
Figure A-2. Double pitot sampling arrangement: (Top) Side view;
(Bottom) Upstream view.
A-7
-------
Table A-0.. Tentative Schedule for Collaborative Test of Method 13
Date Day
10/12/76 Tuesday
10/13/76
10/14/76
Wednesday
Thursday
Location
Research Triangle
Park, N.C.
Research Triangle
Park, N.C.
Badin, N.C.
Activity
Collaborators arrive and
carry equipment to the EPA
facility.
Pre-test calibration check
and orientation session.
Equipment transported to
Badin, N.C.
10/15/76
10/16/76
10/17/76
10/18/76
10/19/76
10/20/76
10/21/76
10/22/76
10/23/76
10/25/76
11/76
Friday
Saturday
Sunday
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
Monday
Complete analysis
Badin, N.C.
Badin, N.C.
Badin, N.C.
Badin, N.C.
Badin, N.C.
Badin, N.C.
Badin, N.C.
Badin, N.C.
Badin, N.C.
Home laboratory
and prepare final report
Equipment set-up and d
Open.
Open.
Testing.
Testing.
Testing.
Testing .
Testing.
Equipment disassembly.
Start analysis.
(results of analysis and
final report must be mailed to arrive at MRI on or before November 30,
1976).
A-8
-------
V. Final Report
The final report may consist of one copy submitted in an informal
letter format. It shall include as a minimum:
(1) Copies of the field test data
(2) Copies of the SPADNS and SIE calibration curves
(3) Copies of the meter box calibrations
(4) Volume and concentration of fluoride in each sample reported
in ug/ml.
(5) Results of all calculation required for Method 13A and 13B
(Sections 9.2, 9.3, 9.4, 9.5, 9.6, and 9.7)
(6) Names of personnel performing the sampling and conducting the
analysis
(7) Description of sampling equipment (manufacturer), spectro-
photometer, SIE system and any other special equipment
used during the program
VI. Bid Requirements
A. Selection of Participants
Respondees to this solicitation for bids will be expected to
analyze the enclosed samples for fluoride using either of the enclosed ver-
sions of EPA Method 13A (Sulfuric Acid Distillation/SPADNS) or Method 13B
(Sulfuric Acid Distillation/Specific Ion Electrode). The volume of each
sample will be 800 ml when shipped. The results of these analysis (in mgF/
liter) and the appropriate SPADNS and/or SIE calibration curves used in the
analysis must accompany the respondees bid. These bids must be received by
Midwest Research Institute (MRI) not later than September 7, 1976.
Only those respondees who successfully analyze the enclosed fluoride
containing samples will be considered qualified to participate in the collabor-
ative test.
Those respondees who successfully analyze the samples will be ranked
on the basis of cost and the six lowest cost bids will be selected.
A-9'
-------
B. Cost
You are requested to submit your total cost on a firm fixed-price
basis for the scope of work described. This bid should include any overtime
you believe will be required to maintain the test schedule. In the event
that additional sampling days are required due to inclement weather, etc.,
you are requested to supply a cost per day for additional sampling and analysis
beyond that described in the scope of work. Costs should be prepared on the
basis of a two-man sampling team (2 professionals each to operate one meter box
and to assist each other in sample recovery).
Resumes giving the background and qualifications of the personnel
who will perform the sampling and analysis shall be included in your proposal.
Since experience in the use of EPA Method 13 (which is very close to EPA
Method 5) may be an important evaluation in judging the qualifications of the
specified participants, their experience in using Method 13 should be included
in their resumes.
VII. Equipment Required for the Test
A. Equipment to be Supplied by MRI
1. The dual train sampling box to be used by each laboratory and
the necessary supporting clamps and platforms. These items will
be delivered to the test site by MRI prior to the actual test.
2. Umbilical cords (1/2-in. ID).
3. Sample heads for connecting the umbilical cord to a 28/15 ball
socket on the last impinger.
4. Pitot lines (3/8-in ID).
5. Nozzles - 3/8-in ID with tapered edge, 5/8 OD adapter for
connecting to probe sheath.
6. Orsat analyzer (MRI will perform the Orsat analysis).
7. Cage for moving glassware up to and down from the roof.
8. Clean-up area (Two medium sized trucks).
9. Tables to set the meter boxes on.
A-,10
-------
B. Equipment to be Supplied by Collaborators
1. 2 meter box units, suitable for running Method 13, meeting
the specifications described by Martin (copy attached and
calibrated as described by Rom (copy attached).
2. Three nomographs meeting the specifications described by
Shigahara (copy attached). Electronic calculators may be
used in place of the nomographs, but if calculators are
used, back-up nomographs will be required.
3. Thirty (30) Whatman No. 1, 3 in. or 4 in. diameter paper
filters (as required to fit collaborators filter holder).
4. Five matched, glass liners for sampling probes having the
dimensions shovn in Figure 3 with 28/12 or 28/15 male ball
fitting on exit of probe and a suitable fitting on the
nozzle end to accommodate a 5/8 OD nozzle. Two matched
metal probe sheaths having the dimensions shown in Figure 3.
Probes need not be heatable, because stack conditions do not
require that the sampled gas be heated.
5. One "S" type pitot tube without thermocouple and not attached to
probe sheath calibrated as described in EPA Revised Method 2
(copy attached). The pitot tube should be 42 in. long and
pitot tubes with coefficients outside 0.83 to 0.87 must be
calibrated as described by Shigahara.
6. The fittings necessary to connect the umbilical cords (1/2 inch
ID) to the meter boxes and to connect the pitot tube line to
both the pitot tube and to the meter box.
7. The usual equipment required for source testing and sample re-
covery, e.g., tools, extension cords, distilled water, probe
brush, extra fittings, fuses, caps to fit the probes and
impinger ball joints, etc.
8. Communication equipment to allow verbal communication between
the team member on the roof and the team member on the catwalk
during the leak-check.
9. Four complete Method 13 impinger trains each cleaned with hot
soapy water and rinsed with tap water and distilled water prior
to test.
A-/11
-------
.48"-
Glass Liner
l_
1" Probe Sheath
Union for Attachment of
5/8" O.D. Nozzle to Probe
2.5"
•* »
1-1/2-2"
I
Size 28/15 or
28/12 Ball (Glass)
Figure A-3. Probes
A-12
-------
10. All equipment and chemicals necessary for 13A and 13B sampling
and analysis of collaborative test samples. (Analytical
supplies are not required at the test site).
11. Special Conditions
The calibration data for the meter boxes and the pitot tube
and also the check of the accuracy of the nomograph are
to be supplied to MRI during the orientation session to be
held at the EPA facility in Research Triangle Park, N.C.
A-l-3
-------
APPENDIX B
STATISTICAL METHODOLOGY
-------
APPENDIX B
STATISTICAL METHODOLOGY
Terms and Definitions
Since the effects in the linear model used for analysis of variance in
Tables 6 and 7 are random, the total variance of an observation is made up of
several components:
where:
2
2 22 22
aTot = °R + °P(R) + °L * a E
is the variance of the runs effects
2
aP(R) 1>s the van>ance of tne P°rts within runs effect
2
is the variance of the laboratory effect
and,
2
is the experimental error variance component.
2
The last term, o£, may be further separated into two components
4 - ,\ + o2
where:
2
0j is the intra-port component of variance describing the variation
contributed to an observation by its being taken in a particular port.
and,
2
a 1s the subsampUng variation attributed to making repeat determina-
tions on the same sample.
B-2
-------
An estimate of the subsampling variation was provided without use of
analysis of variance by pooling all the within-laboratory estimates of vari-
ation. The validity of pooling in this manner was checked by Bartlett's test
for homogeneity of variances. (Homogeneity is a very important assumption in
the analysis of variance that is summarized in Tables 6 and 7.) All the other
estimates of variance components were made by equating the expected mean
square columns to the mean square columns in each of Tables 6 and 7 and then
solving the two systems of equations.
Combinations of variance components are useful in measuring differences
between two observations. The standard deviation of the difference between
two observations (analyses) by the same laboratory on the same sample will be
denoted by <*. =/2<*. The standard deviation between two observations each
made by a different laboratory at the same port during the same run will be
_ Xo "~~* ' ' rt ft
represented by adL = IT |/a + aj + aL • Tne standard deviation of the
difference between two observations that were each made by a different labora-
tory regardless of the port, but during the same run, is denoted by CT .„. =
To add a measure of probability to statements concerning differences between
observations, the assumption of normality of the distribution of random
effects is made.
The terms repeatability and reproducibility have become common, ways of
describing differences between observations with the added assumption of
normality. These terms are defined for our purposes as follows:
Repeatability - (od* = 1.96ad)
-- 95 percent of the differences between two subsamples taken within
a laboratory will not exceed this value in absolute value.
B-3
-------
Reproducibility (same port) - (°M* = l-
95 percent of the differences between two laboratories sampling from
the same port will not exceed this value in absolute value.
Reproducibility (between ports) - (o^ = 1.96 cjdp|_)
95 percent of the differences between two laboratories sampling from
different ports will not exceed this value in absolute value.
The relationship of the distribution of differences between samples taken
by two different laboratories at the same port, and by two different labor-
atories regardless of the port but during the same run, are represented
schematically in Figure B-l. The horizontal bars indicate the 95 percent
confidence limits for each of the above cases. The difference between two
observations by the same laboratory but at different ports is not discussed,
since this involves observations made during different runs. These differences
are not of interest, because different concentration levels were encountered in
different runs.
The information given in Table 8 and Figure B-l may be interpreted as
follows:
1. Two observations from the same laboratory on the same sample must differ by
more than ± 0.123 mg/m3 (Method 13A) or + 0.102 mg/m3 (Method 13B) to be declared
significantly different (repeatability).
2. Two observations from different laboratories sampling from the same port
must differ by more than + 0.259 mg/m3 (Method 13A) or +_ .241 mg/m3 (Method 13B)
to be declared significantly different (reproducibility within ports).
3. Two observations from different laboratories sampling at different ports
should differ by more + 0.345 mg/m3 (Method 13A) or + 0.318 mg/m3 (Method 13B)
to be declared significantly different (reproducibility between ports).
B-4
-------
Accuracy of the Method
Accuracy refers to the size of the deviations of estimates from the actual
value (true value). Accuracy 1s often quantified by a term called bias. Bias
is defined as the signed difference between the average of the individual
estimates and the true value, that 1s,
Bias = Average - True
An analysis of the bias 1n the methodology of Methods 13A and 13B is
summarized in Table 9. The chief effect of bias is to distort the probability
of an estimate being 1n error by more than some predetermined amount. For
example, when there is no bias, an estimate is expected in the long-run to
exceed 1.96a by chance alone only 5 times in a hundred, that is, 0.05. In the
case of Method 13A, which has a bias of -0.10 mg/Ji, the probability of an
estimate being in error by more than 1.96a is 0.13 and for Method 13B, which
has a bias of -0.08 mg/£s the probability is 0.09. These probabilities are
both about twice that expected when no bias is present.
B-5
-------
METHOD 13-A
0d = 0.063
I--95 PERCENT CL-H
-0.123 +0.123
METHOD 13-B
od = 0.052
k95 PERCENT CL-»
0.102 +0.102
odL=0.110
-0.259
95 PERCENTCL-
+0.259
odL = 0.105
—95 PERCENT CL
-0.241 +0.241
adPL = 0.129
adPL = 0.121
-0.345
+0.345
-0.718
318
Figure B-l. Distribution Curves Corresponding to Differences,
B-6
-------
APPENDIX C
USING COLLABORATIVE TEST DATA TO EVALUATE TEST RESULTS
-------
APPENDIX C
USING COLLABORATIVE TEST DATA TO EVALUATE TEST RESULTS
In source testing, 1t 1s common for one laboratory to make repeat measure-
ments (sampling runs) on the same stack while the plant operates at constant
process conditions. Under the above conditions, these repeat samples are
as aimed to be replicates, that 1s, they are assumed to belong to the same
statistical population. Collaborative test results offer a means to determine
the likelihood that these repeat measurements are truly replicates. That is,
the difference between any two true replicates should lie within the limits
specified by the repeatability of the measurement method as determined from
the collaborative test results.
Described below are two sample techniques that can be used to determine
if two repeat measurements are replicates. These two methods are illustrated
using the collaborative test results from Method 13A:
Method 1;
Determine the difference between the individual sampling results and the
average and compare these differences to the repeatability estimate for the
method. For example, suppose Method 13A was used to sample an aluminum plant
o
stack and the following results 1n mg F/m were obtained:
Sampling
Run No.
1
2
3
Average:
Sampling
Result
0.259
1.024
0.480
0.588
C-2
Difference
from the Average
0.329
-0.437
0.108
-------
Since the repeatability of Method 13A is +0.123 mg F/m , the results from
runs 1 and 2 He outside the value expected for Method 13A. Thus, the results
do not belong to the same population and we must suspect that the stack concen-
tration was not constant over runs or else that a significant error was made in
one of the measurements.
Method 2
An alternate and quicker means to judge the homogeneity or validity of data
is to employ the relative range approach. In this approach, the difference
(range) between the highest and lowest values is determined and this difference
(R) is then divided by the within-laboratory standard deviation (a). Then the
probability that this quotient (W) exceeds the predicted value is obtained from
a Percentage Point of the Distribution of the Relative Range Table (1). For
example, suppose Method 13A was used to sample a stack for fluoride and the
following concentrations in mg F/m were obtained:
Run 1 0.361
Run 2 0.421
Run 3 0.480
In this example, R is 0.119 (0.480-0.361) and a is + 0.044. Thus:
W = -a = S4H = 2.70
Since W does not exceed the predicted value for three replicates, 4.12, the
test shows that the results all belong to the same statistical population.
(1) Duncan, A.J. "Quality Control and Industrial Statistics," R. D. Irwin Co., Inc.,
p. 908 (1965).
C-3
-------
TECHNICAL REPORT DATA
(Please read InUmcttons on the reverse before completing)
i. REPORT NO.
2.
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
COLLABORATIVE STUDY OF EPA METHOD 13A AND
METHOD 13B
5. REPORT DATE
June 1977
6. PERFORMING ORGANIZATION CODE
371360
7. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NO.
William J. Mitchell, Jack C. Suggs and Fred J. Bergman
9, PERFORMING ORGANIZATION NAME AND ADDRESS
U.S. Environmental Protection Agency
Office of Research and Development
Environmental Monitoring and Support Laboratory
Research Triangle Park, North Carolina 27711
10. PROGRAM ELEMENT NO.
1HD621
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
13. TYPE OF REPORT AND PERIOD COVERED
Environmental Monitoring and Support Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
14. SPONSORING AGENCY CODE
EPA/600-08
IS. SUPPLEMENTARY NOTES
16. ABSTRACT
The results from a collaborative test of U.S. EPA Method 13A and 13B are presented.
The collaborative test was conducted at a primary aluminum reduction plant. In the
collaborative test, six laboratories each operated two Method 13 sampling trains
and all laboratories simultaneously sampled the same stack. Ten such sampling runs
were done for a total of 120 samples. Each source sample was analyzed for fluoride
using both Method 13A and Method 13B. The results of the collaborative test showed
that both test methods gave similar results.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS C. COSATI Field/Group
Fluoride Sampling Analysis, Method 13,
Aluminum Plant, Collaborative Test,
Novel Sampling Arrangement
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report)
21. NO. OF PAGES
75
20. SECURITY CLASS (Thispage)
22. PRICE
EPA Perm 2230-1 (9-73)
------- |