vvEPA
United States
Environmental Protection
Agency
Environmental Monitoring and Support EPA-600/4 81 -025
Laboratory April 1981
Research Tna.igle Park NC 27711
Research and Development
National Performance
Audit Program
1979 Proficiency
Surveys for Sulfur
Dioxide, Nitrogen
Dioxide, Carbon
Monoxide, Sulfate,
Nitrate, Lead and
High Volume Flow
-------
EPA 600/4-81-025
April 1981
NATIONAL PERFORMANCE AUDIT PROGRAM
1979 PROFICIENCY SURVEYS
FOR
SULFUR DIOXIDE, NITROGEN DIOXIDE, CARBON MONOXIDE,
SULFATE, NITRATE, LEAD AND HIGH VOLUME FLOW
by
Berne I. Bennett, Robert L. Lampe, and John C. Puzak
Quality Assurance Division
Environmental Monitoring Systems Laboratory
Research Triangle Park, North Carolina 27711
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
U.S. E--/;:v,r-;. ,,: : ; roi;,c; on Agency
Region V, Lie' -'j
230 South Dearborn Siioat
Chicago, Illinois 60604
-------
DISCLAIMER
This report has been reviewed by the Environmental Monitoring Systems
Laboratory, U.S. Environmental Protection Agency, and approved for publica-
tion. Approval does not signify that the contents necessarily reflect the
views and policies of the U.S. Environmental Protection Agency. Mention of
trade names or commercial products does not constitute endorsement or
recommendation for use.
u,s.
11
-------
CONTENTS
Abstract iv
Figures v
Tables vi
Acknowledgment viii
1. Introduction 1
2. Program Profile 2
3. Statistical Considerations 3
4. Survey Material 7
Sulfur Dioxide Samples 7
Nitrogen Dioxide Samples 7
Carbon Monoxide Samples 9
Sulfate-Nitrate Samples 9
Lead Samples 10
High Volume Reference Flow Device 14
5. Results 16
Sulfur Dioxide 16
Nitrogen Dioxide 21
Carbon Monoxide 27
Sul fate 33
Nitrate 38
Lead 45
Hi-vol flow 48
References 53
-------
ABSTRACT
The Quality Assurance Division of the Environmental Monitoring Systems
Laboratory, Research Triangle Park, North Carolina, administers semiannual
Surveys of Analytical Proficiency for sulfur dioxide, nitrogen dioxide, carbon
monoxide, sulfate, nitrate, and lead. Sample material, simulating ambient air
pollution samples as closely as possible, are furnished to participating
laboratories. Surveys of hi-vol sample flow are conducted annually using a
modified orifice.
The various sample materials are monitored by the Quality Assurance
Division to assure that samples are stable, of uniform composition and are
representative of pollutant concentration levels encountered under field
sampling conditions and that all materials conform to prescribed standards of
accuracy. Sample materials are required to be similar enough to true air
pollution matrices not to introduce unrealistic conditions of sample prepa-
ration or impose handling techniques that are not a part of the normal
monitoring and analytical activity.
A major survey objective is the assessment of routine analytical per-
formance. After results are evaluated by the Quality Assurance Division, an
individual report is promptly returned to each participant. This report
contains a summary of survey results for the year 1979.
IV
-------
FIGURES
Number Page
1 Reference flow devices mounted on hi-volume sampler ... 13
2 Reference flow device with resistance plate 13
3 Mean values of SCL survey 0479 vs. expected values ... 22
4 Mean values of S0? survey 1079 vs. expected values ... 22
5 Mean values of N0? survey 0679 vs. expected values ... 29
6 Mean values of N0? survey 1279 vs. expected values ... 29
7 Mean values of CO survey 0379 vs. expected values ... 32
8 Mean values of CO survey 0979 vs. expected values .... 32
9 Mean values of SOT survey 0279 vs. expected values ... 39
10 Mean values of SOT survey 0879 vs. expected values ... 39
11 Mean values of NO, survey 0279 vs. expected values ... 44
12 Mean values of NO- survey 0879 vs. expected values ... 44
13 Mean values of Pb survey 0179 vs. expected values .... 49
14 Mean values of Pb survey 0779 vs. expected values .... 49
-------
TABLES
Number Pane
————^^ 3
1 Concentration of S0? Survey Samples 8
2 Concentrations of Nitrogen Dioxide Survey Samples .... 8
3 Concentrations of Carbon Monoxide Survey Samples .... 10
4 Concentrations of Sulfate Survey Samples 11
5 Concentrations of Nitrate Survey Samples 11
6 Concentrations of Lead Survey Samples 12
7 Agency Apportionment of Su'lfur Dioxide Surveys 17
8 Summary of Sulfur Dioxide Proficiency Surveys 18
9 Sulfur Dioxide by Analytical Methods 19
10 Sulfur Dioxide, Percent of Measurements Within Indicated
Percent of Expected Value 19
11 Agency Apportionment of Nitrogen Dioxide Surveys .... 23
12 Summary of Nitrogen Dioxide Proficiency Surveys 24
13 Nitrogen Dioxide by Analytical Methods 25
14 Percent of Nitrogen Dioxide Measurements Within
Indicated Percent of Expected Value 26
15 Agency Apportionment of Carbon Monoxide Surveys 28
16 Summary of Carbon Monoxide Proficiency Surveys 30
17 Carbon Monoxide by Analytical Methods 30
18 Percent of Carbon Monoxide Measurements Within Indicated
Percent of Expected Value 31
19 Agency Apportionment of Sulfate Surveys 34
20 Summary of Sulfate Proficiency Surveys 35
21 Sulfate by Analytical Method 36
22 Percent of Sulfate Measurements Within Indicated
Percent of Expected Value 37
23 Agency Apportionment of Nitrate Surveys 40
24 Summary of Nitrate Proficiency Surveys 41
25 Nitrate by Analytical Method 42
-------
Number
26 Percent of Nitrate Measurmtsnts Within Indicated
Percent of Expected Value 43
27 Agency Apportionment of Le
-------
ACKNOWLEDGMENTS
The enthusiastic assistance rendered by colleagues in the Performance
Evaluation Branch, Quality Assurance Division, including Blaine F. Parr for
compiling, interpreting and reducing enormous volumes of raw data, Linda F.
Porter and Avis P. Mines for their analytical services, is gratefully acknowl-
edged. Recognition is due to the staff of the Data Management and Analysis
Division, Environmental Monitoring Systems Laboratory for their invaluable aid.
Gratitude is expressed, too, for the efforts of the staffs of Stewart Labora-
tories, Inc. and Northrop Services, Inc. for their diligent analyses needed
to ascertain the precision and assure the accuracy of the various types of
sample materials used in the surveys.
vm
-------
SECTION I
INTRODUCTION
The 1979 Proficiency Surveys continue the regular surveys by the
Environmental Monitoring Systems Laboratory (EMSL) of agencies which
routinely collect and analyze ambient air samples. Sample materials,
furnished by EPA for this purpose are designed to simulate as closely as
possible several types of collected air pollution samples. It is prudent
to note that these samples treat only the analytical portion of the total
air monitoring capability, and do not deal with errors from sample collec-
tion, transportation, handling, storage, and data processing. Rankings in
the surveys, except as may occasionally be due to unpropitious circum-
stances, reflect the effectiveness of internal quality assurance programs.
The Proficiency Surveys allow EPA to monitor the caliber of air
pollution analyses, and permit the participating agencies to assess their
own performances vis-a-vis their peers.
With the assistance and cooperation of the EPA Regional Offices, the
surveys are conducted by the Quality Assurance Division (QAD)/EMSL,
Environmental Research Center, Research Triangle Park, North Carolina
27711. Inquiries and applications to participate should be directed to
that address. Included in this report is a discussion of the program,
description of the survey materials, statistical summaries and the
results.
-------
SECTION 2
PROGRAM PROFILE
Participants in the surveys are solicited by the Regional Quality
Control Coordinator in each of the ten Regions. Once a laboratory enrolls
in a survey for a particular pollutant, it is automatically notified of
subsequent surveys for that pollutant. Participants are assigned an
identifying code number which remains in effect for all surveys. Included
in the surveys are representatives of federal, state, local, industrial
and foreign air pollution monitoring agencies.
Soon after a roster is established, instructional information and
blind sample materials are mailed. Surveys are presently conducted twice
a year for carbon monoxide (CO), sulfur dioxide (S0?), nitrogen dioxide
(N02), lead (Pb) on filter strips, sulfate (SO^) and nitrate (N0~) on
filter strips and once a year for hi-vol flow. Reports now indicate the
reported value and include an historical frequency distribution of test
results. A comprehensive report is prepared yearly summarizing the survey
results of that year.
Laboratories submitting abnormal measurements are offered an oppor-
tunity to analyze another set of unknown samples, similar to those of the
main survey, but of different concentrations. However, the retest results
are not included in this report.
-------
SECTION 3
STATISTICAL CONSIDERATIONS
Before 1979, the surveys gave target and sample ranges that defined
the sphere of all creditable results. Under the former system, results
falling within the sample range indicated a fine analytical job and
those within the wider target range n respectably good one. This format
enjoyed wide favor, due to the clear, explicit and uncontestable tableau
it presented for scoring any single test result.
Beginning with the 1979 survey year, in an effort to enable scoring
of individual results, a new format is used which presents cumulative
frequency distributions of the results of earlier surveys.
Investigators have long agonized over which measurements are totally
believable and which are wholly discreditable. Judgment of the investi-
gator was the historical criterion for rejection of data. David Bernoulli,
writing in an earlier century about astronomical observations , stated
that he could see no way of drawing a dividing line between data values
that are to be utterly rejected and those that are to be entirely retained.
A fair and dispassionate system for judging outliers is needed in any
survey.
The criterion chosen for the 1979 surveys is one that has been in use
2
for a long time, the Chauvenet's Criterion. This criterion is based on
the normal distribution and advises rejection of an extreme observation
if the probability of occurrence of such deviation from the mean of the n
measurements is less than l/2n. Since inclusion of spurious data vitiates
test results by biasing both the survey mean and precison, but removal of
3
good measurements merely excludes some of the data , it is thought that
it is better to reject some good dat;.» than to include truly anomalous
measurements.
-------
Several of the statistical procedures used to evaluate survey results
are reviewed below. The F test, t test and an extension of the analysis
of variance. The F test was used to test whether the variance of one test
method exceeded that of another. The ratio of the variances was not expected
to exceed a critical value, which is based on the number of observations
(measurements) in each test method, unless differences exist in the pre-
cision of the two methods. If F equals one, there is no difference in vari-
ability between the two methods. Values less than one have no meaning.
The value of the statistic F is given by the expression:
F = sl ^S2 ^Eq- 1^
where:
s.. = standard deviation of method 1
Sp = standard deviation of method 2
Two assumptions that underlie the use of the t test for comparing
the means of two samples are that their population distributions are
normal and that their population variances are equal. When those
conditions are satisfied, the variance estimates can be pooled and
the statistical relations shown belov/ are assumed to apply. The
pooled variance is given by the following equation:
2 2
2 (N - l)s + (N - l)s
s (pooled) = 1122 (Eq. 2)
p N + N + 2
1 2
The standard error of the difference between the mean is given by the
equation:
-------
2 2
/(N - l)s + (N -
s =/ 1 1 2 2 /_!_ + _1_ (Eq. 3)
d i/ N+N-2 i/N N
12 '12
The formula for the standard error of the difference between the means
is substituted in the formula for t, giving:
x - x
t = 1 2 (Eq. 4)
N
2
where:
x, = the average of test method no. 1
x,, = the average of test method no. 2
N-. = no. of measurements by test method no. 1
N = no. of measurements by test method no. 2
To compare the averages of three or more sets of measurements (the
t test can compare only two sets), an extension of the analysis of
o
variance , the statistic w, is used. If the absolute differences
between the averages of all the sets of data are less than w, then
it is considered that no differences exist among the averages. The
critical value of w is computed thus:
w = q g5 S/\/n (Eq. 5)
-------
where:
w = computed critical value from standardized range
q g^ (t,v) = a statistic that is a function of the number
of data sets, t, and degrees of freedom, v
S = pooled standard deviation
C*
n = number of data sets
-------
SECTION 4
SURVEY MATERIALS
SULFUR DIOXIDE SAMPLES
The sample material was composed of freeze-dried mixtures of sodium
sulfite and potassium tetrachloromercurate (TCM) contained in 5-ml glass
ampoules. Sample sets consisted of five ampoules containing 4 to 64 |jg of
SOp equivalent per container. The sample material was stored at -20°C in
the QAD Repository to sustain the integrity of the SOp activity, which was
confirmed by periodic retesting. Analyses were performed by the reference
method for the determination of S09 in the atmosphere (pararosaniline
5 ^
method). The sample forms a dichlorosulfitomercurate complex when solu-
bilized in 0.04N TCM. This complex is reacted with pararosaniline and
formaldehyde to form intensely colored pararosaniline sulfonic acid. The
absorbance of the solution is measured spectrophometrically at 548 nm.
At least 15 samples from each concentration level were analyzed
(Table 1) and the analyzed values are the reference (expected) values of
surveys 0479 and 1079 (surveys are numbered by month and year). In addition,
independent corroborative tests were conducted by another laboratory. The
presumption was that each sample was collected in 50 ml of absorbing
reagent with a total sample air volume of 300.2.
NITROGEN DIOXIDE SAMPLES
The samples consisted of 5 mL of aqueous sodium nitrite contained in
glass ampoules. A set consisted of 5 ampoules. When mixed with caustic
absorbing reagent, the samples simulated ambient samples ranging in concen-
tration from 0.17 to 1.00 ug/mL.
-------
Analysis of 15 samples in each concentration level was performed using
an equivalent method for the determination of NOp in ambient air.
Measurements were made spectrophotometrically at 540 nm. The values
contained in Table 2 were the reference (expected) values for surveys
0679 and 1279. Values are based on the entire sample being diluted to
50 ml with absorbing reagent.
TABLE 1. CONCENTRATIONS OF S0£ SURVEY SAMPLES
Survey 0479
ug S09 ug S09/m3
f. £.
Sample No. x* s x* s Samole
1 4.18 0.69 13.9 2.3 1
3 28.0 0.42 93.3 1.4 2
4 44.5 0.37 148 1.2 3
5 63.5 0.42 212 1.4 4
5
*n = 15
tn = 25
TABLE 2. CONCENTRATIONS OF NITROGEN
Survey 0679
(jg/mL NO-
Sample No. -A - Sample No
A S
1 0.259 0.0018 1
2 0.405 0.0034 2
3 0.514 0.0019 3
4 0.700 0.0018 4
5 0.935 0.0032 5
Survey 1079
M9S
No . xt
4.04
11.6
24.0
37.5
49.4
®2 I-1 9
s xt
0.30 13.5
0.25 38.5
0.92 80.0
1.30 125
0.87 165
S02/m3
s
1.0
0.83
3.1
4.3
2.9
DIOXIDE SURVEY SAMPLES
Survey
ug/mL
x*
0.172
0.342
0.595
0.746
1.00
1279
NO,
s
0.0065
0.0046
0.0076
0.0174
0.0275
-------
CARBON MONOXIDE SAMPLES
Samples consisted of compressed gas mixtures of carbon monoxide (CO)
and artificial air. Also contained in each sample was 2 ppm of methane
(CH.) and approximately 365 ppm of carbon dioxide (C02). Aluminum cylin-
ders were used in the surveys. Sample concentrations ranged from 3 to 44
ppm of CO. Each participant received a set of three cylinders, one from
each of three concentration levels.
Verification testing was accomplished by use of a non-dispersive
infrared analyzer (NDIR). Fifteen samples from each concentration level
were analyzed by QAD and an independent testing laboratory. Table 3 lists
the reference (expected) values for surveys 0379 and 0979.
SULFATE-NITRATE SAMPLES
Samples consisted of 1.9 x 20 cm (0.75 x 8 in.) glass fiber filter
strips with depositions of potassium sulfate (I^SO^) and lead nitrate
(Pb[N03J2). Filter strip samples included concentrations of various S04
and NO- concentrations. Each strip was sealed in a plastic envelope. The
3
concentrations of sulfate ranged from approximately 1.5 to 29 ug/m .
o
Nitrate levels spanned between 1.5 and 12 ug/m . Concentrations were
determined using the requisite filter dimensions of 20 x 25.4 cm (8 x 10
3
in) and a collected air volume of 2000 m . Presuming that gravimetric
preparation and transfer onto the filter strips could be carried out more
accurately than could existing analytical methods, the reference (expected)
values were obtained theoretically from the deduced mass of inorganic
salts deposited on the filters. Verification analyses assured that the
accuracy and precision of the samples were within prescribed limits. The
sulfate concentrations have been given in Table 4 and the nitrate values
are listed in Table 5. These were the values which were applied to surveys
0279 and 0879.
-------
LEAD SAMPLES
Samples were composed of 1.9 x 20 cm (0.75 by 8 in) glass fiber
filter strips with depositions of lead nitrate (Pb[N03]2). Filter strip
sample sets contained combinations of various lead concentrations, each
sealed in a plastic envelope. The lead content ranged from 1.5 to 12.9
3
ug/m . Concentrations were calculated presuming that the samples were
collected on the prescribed 20 x 25.4 cm (8 x 10 in) hi-vol filter with a
3
total air volume of 2000 m .
The precision and accuracy measurements of this sample material were
done by atomic absorption analysis. Table 6 lists the reference (expected)
values used in surveys 1079 and 0779.
TABLE 3. CONCENTRATIONS OF CARBON MONOXIDE SURVEY SAMPLES
Survey 0379
ppm CO
Sample No. x* s
1 6.53 0.06
2 19.8 0.19
3 43.7 0.26
Survey 0979
Sample No.
1
2
3
ppm CO
x*
2.98
14.8
33.8
s
0.02
0.06
0.12
*n = 15
10
-------
TABLE 4. CONCENTRATIONS OF SULFATE SURVEY SAMPLES
Sample
No.
2
6
0
3
1
4
Survey 0279
= — 3
jjg SO./strip ug SO./m /filter
250.0
1000.0
1750.0
2500.0
3250.0
4000.0
TABLE 5.
1.5
6.0
10.5
15.0
19.5
24.0
CONCENTRATIONS OF
Survey 0879
$NoPle (jg s04/stn'P M9 S0^/m3/f liter
5 1200.0 7.2
0 1516.7 9.1
4 2183.3 13.1
3 3950.0 23.7
2 4800.0 28.8
NITRATE SURVEY SAMPLES
Sample
No.
0
1
o
e..
4
5
3
Survey
ug N03/strip ug
250.0
600.0
950.0
1300.0
1650.0
2000.0
0279
N03/m3/f liter
1.5
3.6
5.7
7.8
9.9
12.0
Survey 0879
Sample - - -
No. ug N03/str1p ug N03/m /filter
5 200.0 1.2
0 566.7 3.4
2 1000.0 6.0
4 1383.3 8.3
3 1700.0 10.2
11
-------
TABLE 6. CONCENTRATIONS OF LEAD SURVEY SAMPLES
Survey 0179
c,mnlp MS Pb/test strip
No. x*
3 243.4
4 586.7
5 899.0
6 1908.2
7 1246.2
8 1584.6
s
6.
14.
30.
70.
36.
62
09
89
07
11
83
.17
ug/m3/fi
X*
1.46
3.52
5.39
11.45
7.48
9.51
Hter
s
0.037
0.089
0.180
0.421
0.221
0.373
Survey 0779
MS Pb/test strip
Sample
No.
0
1
2
3
4
5
X*
587.
2143.
974.
1731.
1320.
196.
9
1
7
4
7
8
s
13.
33.
24.
27.
18.
3.
52
30
08
58
93
45
MQ/m3/f ilter
X*
3.53
12.86
5.85
10.39
7.92
1.18
s
0.081
0.200
0.144
0.165
0.114
0.021
*n = 10
12
-------
Figure 1. Reference flow device Mounted on high volume sampler.
Figure 2. Reference flow device with resistance plate.
13
-------
HIGH VOLUME REFERENCE FLOW DEVICE (ReF)
A single ReF was supplied to each participating agency. Organizations
were instructed to check as many hi-vol sampling units as feasible within
the allotted time. The unit received by each laboratory consisted of a
modified orifice, wind deflector, manometer, and resistance plates (to
change flow rates).
During measurement of the air flow of a hi-vol sampler, the ReF was
mounted on top of the sampler replacing the filter face plate as shown in
Figure 1. A wind deflector was necessary to prevent fluctuation in the
readings due to wind blowing across the orifice. The resistance plates,
when inserted into the ReF, simulated various filter loading conditions
as illustrated in Figure 2.
By calibrating each ReF with a positive displacement meter (roots
meter), in conjunction with measurements of pressure drops and tempera-
tures, a calibration curve in the form of an orifice equation was derived.
The equation shown below was used to determine the "K" orifice constant
for each unit.
(Eq. 6)
where:
Q, = volumetric flow at conditions of T, and P, (m /min)
A = area of orifice (in )
Y = expansion factor
C = orifice coefficient
AP = pressure drop across orifice (in H?0)
P, = upstream pressure (barometric pressure,mm Hg)
T, = upstream temperature (ambient temperature, °K)
14
-------
Because A is constant for a given orifice, and Y and C are essentially
constant over the flow range in question, a new orifice constant "K" was
defined as:
K = AYC (Eq. 7)
Thus, the orifice equation becomes:
Ql " K ' i (Eq. 8)
During calibration of the ReF, Q,, AP, T, and P, were also measured. The
constant K was determined by regressing a series of Q, measurements onto
the square root of the values under the radical.
During the survey, operating personnel measured AP, T, and P,. By
knowing K, the "true flow" can be calculated. This flow was compared
with the flow measured by the high volume sensor to determine the accuracy
of flow measurements.
15
-------
SECTION 5
RESULTS
SULFUR DIOXIDE
Test Methods
Approximately 80 percent of the respondents in 0479 and 1079 surveys
used the manual, and 20 percent the automated pararosaniline procedures.
The few remaining laboratories used other methods. The results of the two
principle methods were subjected to the t test at the 95 percent confi-
dence level. The difference between the averages was not statistically
significant. The F test showed that the two methods do not differ with
regard to variability.
Agency Apportionment
Participation in the April 1979 SO, survey decreased by 12 percent
7 *
from the previous year , with the October 1979 survey continuing at about
the same as 1978. The reduction was largely accounted for by decreased
use of the pararosaniline method by state agencies. The number of users
of the pararosam'line method stabilized at around 100. The distribution
of agencies in the surveys is shown in Table 7.
16
-------
TABLE 7. AGENCY APPORTIONMENT OF SULFUR DIOXIDE SURVEYS
Agency Survey 0479, %
Regional (Federal)
State Agencies
Local Agencies
I ndustri al /Contractor
Foreign
1.8
37.8
42.4
18.0
0.0
Survey 1079, %
0.0
39.0
43.8
16.2
1.0
Data Summary
Survey results are summarized in Table 8, with all methods included.
Results according to analytical method are given in Table 9. Table 10
gives the frequency distribution of the results in terms of percent of
expected values. The values termed expected values are the best estimates
of the true concentrations and are derived from the analyses performed by
EPA laboratories, a commercial corroborative laboratory, and the manufac-
turer's analyses. The sample material was evaluated as to stability,
homogeneity, and accuracy before use in the survey.
Anomalous measurements are excluded from the summary tables. In
survey 0479, 7 percent, and in survey 1079, 4.6 percent of the measurements
were excluded by Chauvenet's Criterion.
17
-------
TABLE 8. SUMMARY OF SULFUR DIOXIDE PROFICIENCY SURVEYS
Sample
no.
Respondents*
Expected
value
|jg/m
Survey
mean.,
ug/m
Survey
std. dev.
Survey interval
Survey 0479 (April 1979)
I 100
3 105
4 103
5 105
Survey 1079 (October 1979)
1 101
2 103
3 102
4
5
103
104
13.9
93.3
148.3
212.7
13.5
38.5
80.0
125
165
14.15 4.85 2.17 - 25.76
92.56 7.97 70.74 - 109.10
154.30 10.45 126.51 - 175.68
214.48 15.03 170.80 - 256.99
12.63 5.00 1.08 - 33.30
35.80 6.57 10.13 - 54.10
77.26 8.14 48.14 - 108.40
125.24 11.31 88.00 - 166.50
165.62 14.16 104.22 - 206.97
*With outliers removed.
18
-------
TABLE 9. SULFUR DIOXIDE BY ANALYTICAL METHOD
(ug/m3)
Survey
Sample
1
3
4
5
Survey
0473* (April 1979)
Pararosani line - manual
No. Nf Mean Std. dev
78 14.87 7.19
83 92.19 10.06
81 153.59 13.95
84 214.30 16.37
1079* (October 1979)
.
Pararosani 1 i ne-manual Pararosani 1 i
Sample
I
2
3
4
5
no. N Mean Std. dev. N
79 13.09 9.12 20
81 35.25 9.70 20
80 76.22 13.85 20
81 123.27 16.86 20
82 166.12 27.00 20
Mean
11.75
35.32
75.26
121.26
169.14
Pararosani line
N Mean
22 14.33
22 91.34
22 153.73
21 210.08
ne- automated
Std. dev. N
4.44 2
9.28 2
11.90 2
20.38 2
24.49 2
- automated
Std. dev.
5.06
8.34
12.98
17.68
All others
Mean Std. dev.
15.13 0.10
36.08 1.68
71.20 0.99
127.28 13.04
168.94 9.70
*0utliers removed.
tNumber of respondents.
TABLE 10. SULFUR DIOXIDE, PERCENT OF MEASUREMENTS
WITHIN INDICATED PERCENT OF EXPECTED VALUES
Sample
1
3
4
5
Survey 0479
no. 10% 20% 30% 50%
20 35 57 82
74 97 99 100
72 100 100 100
79 100 100 100
Sample
1
2
3
4
5
Survey 1079
no. 10% 20%
34 57
54 80
72 89
81 92
78 93
30% 50%
66 79
85 94
92 95
96 98
96 97
19
-------
The mean values from survey 0479, plotted against the expected values,
gave a linear relationship, as follows:
y = a + bx (Eq. 7)
where:
y = survey average
x = expected value
a = y intercept = -0.0957
b = slope = 1.0164
2
R = coefficient of linearity squared = 0.9991
y = -0.0957 + 1.0164 (expected value)
The plot of the survey means (y) against the expected value (x)
has been shown in Figure 3.
The means for survey 1079, plotted against the expected values, gave
a linear relationship, as follows:
Survey average = -2.5100 + 1.0168 (expected value)
R2 = 0.9996
The plot of the survey means (y) against the expected values (x)
is shown in Figure 4.
Summary
Proficiency surveys for SO^ were conducted in April and October 1979.
Approximately 100 participants completed each of the surveys. The pararo-
saniline method predominated with 80 percent of the laboratories using
that procedure. No systematic discrepancies or substantial bias were found
in the S09 surveys.
20
-------
NITROGEN DIOXIDE
Test Methods
The predominant analytical method used in surveys 0679 and 1279 was
the manual sodium arsenite colorimetric procedure; 72.7 percent of the
0679 respondents used it. Slightly fewer reported using the manual method
in the 1279 survey. Around 21 percent of respondents used the automated
sodium arsenite method. The manual and automated Saltzman and TGS-ANSA
manual methods made up the balance of the test methods. Several partici-
pants did not indicate the method of analysis.
The averages of all the manual test methods were compared by the
statistic w. It was found that there were no differences in the means of
the several different methods, according to this conservative statistical
test.
The variability between the two dominant methods was tested by the F
test. In the 0679 survey, the variability of samples 4 and 5 was greater
by the manual arsenite method, while in the 1279 survey, samples in the
same concentration range exhibited greater variability by the automated
method. There was no reason to suspoct that there was any inherent
difference in variability between the; two test methods.
Agency Apportionment
Participation in the 1979 Proficiency Survey continued at close to
the 1978 level , with the second of the twice yearly tests having approxi-
mately 11 percent fewer laboratories taking part. The agencies included
in the N0? surveys by type are shown in Table 11.
21
-------
CD
250
200
150
100
,
Ol
= 50
50 100 150
Expected value,
200
250
13.9 14.15
93.3 92.56
148.3 154.3
212.7 214.5
co
c
3.
O)
0)
Figure 3. Means of S0? survey 0479 vs. expected values.
200
Expected values, yg/rrT
X
13.5
38.5
80.0
125.0
165.0
y
12.63
35.80
77.26
125.2
165.6
Figure 4. Mean values of S02 survey 1079 vs. expected values.
22
-------
TABLE 11. AGENCY APPORTIONMENT OF NITROGEN DIOXIDE SURVEYS
Agency
Regional (Federal)
State Agencies
Local Agencies
Industrial /Contractor
Foreign
Survey 0679, %
1.1
42.5
44.7
10.6
1.1
Survey 1279, %
0.0
33.3
52.6
12.8
1.3
Data Summary
Survey results are tabulated in Table 12 with all methods included.
Table 13 lists the results by specific analytical method. Frequency
distributions by percent of the expected value are shown in Table 14.
Table 14 shows the percents of reported measurements which lie within the
indicated percent (10, 20, 30, or 50) of the expected value. The
expected values, or best estimates of the true concentrations, were
derived from analyses performed by the QAD, by a commercial testing
laboratory and by the manufacturer. The sample material was tested for
conformity to established criteria for precision and accuracy.
23
-------
TABLE 12. SUMMARY OF NITROGEN DIOXIDE PROFICIENCY SURVEYS
Sample Respondents* Expected
No. value
|jg/ml_
Survey 0679
1
2
3
4
5
Survey 1279
1
2
3
4
5
(June 1979)
88
88
88
87
86
(December 1979)
76
78
76
77
76
0.259
0.405
0.514
0.700
0.935
0.172
0.342
0.595
0.746
1.000
Survey
mean
ug/mL
0.26
0.40
0.51
0.71
0.95
0.18
0.35
0.59
0.74
0.97
Survey
std. dev.
ug/mL
0.02
0.02
0.03
0.03
0.04
0.04
0.05
0.07
0.09
0.12
Survey interval
ug/mL
0.21 -
0.34 -
0.39 -
0.62 -
0.83 -
0.09 -
0.14 -
0.22 -
0.28 -
0.34 -
0.30
0.45
0.62
0.78
1.07
0.44
0.55
0.73
0.90
1.16
*With outliers removed.
24
-------
TABLE 13. NITROGEN DIOXIDE BY ANALYTICAL METHOD
ug/mL
N)
cn
Survey 0679* (June 1979)
Saltzman-manual
Sample No.
N Mean Std. dev.
1 3 0.26 0.01
2 3 0.42 0.02
3 3 0.52 0.01
4 3 0.72 0.01
5 3 0.97 0.04
Survey 1279* (December 1979)
Saltzman-manual
Sample No. N
1 3
2 3
3 3
4 3
5 2
Mean
0.17
0.36
0.59
0.71
1.02
Std. dev.
0.02
0.01
0.05
0.09
0.02
Sodium arsenite-
manual
N Mean Std. dev.
Sodium arsenite-
automated
N Mean Std. dev.
TGS-ANSA
manual
All others
N Mean Std. dev.
64 0.26 0.02 18 0.26 0.02 1 0.26 0.00
64 0.40 0.03 18 0.40 0.02 1 0.40 0.00
64 0.52 0.04 18 0.50 0.06 1 0.51 0.00
64 0.71 0.05 18 0.70 0.02 1 0.69 0.00
62 0.95 0.07 18 0.94 0.02 1 0.95 0.00
Saltzman-automated Sodium arsem'te- Sodium arsem'te-
manual automated
N
1
1
1
1
1
Mean
0.17
0.32
0.59
0.71
0.96
Std. dev.
0.00
0.00
0.00
0.00
0.00
N
48
51
49
50
50
Mean
0.18
0.36
0.61
0.75
0.99
Std. dev.
0.02
0.03
0.03
0.04
0.05
N
16
16
16
16
16
Mean
0.19
0.36
0.62
0.74
1.01
Std. dev.
0.02
0.02
0.04
0.09
0.07
N
2
2
2
2
2
N Mean Std. dev.
2 0.26 0.01
2 0.39 0.00
2 0.70 0.26
1 0.73 0.00
2 0.99 0.01
TGS-ANSA-
manual
Mean Std. dev.
0.20
0.34
0.58
0.73
0.99
0.02
0.04
0.06
0.02
0.05
N
1
1
1
1
1
All others
Mean
0.18
0.38
0.62
0.76
1.00
Std. dev.
0.00
0.00
0.00
0.00
0.00
*With outliers removed.
-------
TABLE 14. PERCENT OF NITROGEN DIOXIDE MEASUREMENTS
WITHIN INDICATED PERCENT OF EXPECTED VALUES
Survey 0679
Survey 1279
Sample no. 10% 20% 30% 50%
Sample no. 10% 20% 30% 50%
86 97 100 100
90 99 100 100
85 92 97 99
93 97 99 100
90 97 100 100
63 86 92 93
81 91 94 94
88 92 95 95
86 95 95 95
89 93 95 95
The means from survey 0679, plotted against the expected values, gave
a linear relationship, as follows:
Survey average = -0.0121 + 1.0276 (expected value)
R2 = 0.9997
The plot of the survey means (y) against the expected values (x)
is shown in Figure 5.
The means from survey 1279, plotted against the expected values, gave
a linear relationship, as follows:
Survey average = 0.0205 + 0.9553 (expected value)
R2 = 0.9
-------
The plot of the survey means (y) against the expected values (x) has
been shown in Figure 6.
Summary
Proficiency Surveys in 1979 for NCL were conducted in June with
approximately 90 participants and in December with close to 80. The
analytical method used by approximately 70 percent of the respondents was
the manual sodium arsenite procedure. No systematic discrepancies or
substantial bias were apparent in the N0? survey data.
CARBON MONOXIDE
Test Methods
Of the test methods listed on the survey information forms, the NDIR
method dominated, with 91 percent of the laboratories using that method in
the 0379 survey and 87 percent in the 0979 test. Facilities using the GC
method increased from 6 to 12 percent during the biannual testing period.
Other methods were reported being used by approximately 2 percent of the
respondents.
Results by the two principal methods were compared by the t test to
determine whether differences in the averages were significant. In survey
0379 they were not; in 0979 survey, the difference was significant for
sample number 2. Although the difference in the averages of only one of
six samples was statistically significant, the GC survey standard devia-
tions were substantially larger in five of six samples.
The great disparity in of the number of laboratories using the two
methods was a factor in comparing the results. Were the number of users
of the GC method equivalent with tho.se using the NDIR method, it would be
expected that the GC standard deviation would be larger yet. The 1979 and
previous survey results lead to the conclusion that the precision of the
GC method is less than that of the NDIR method. The F test supported the
conclusion. Fvaluatinn the ratio of the variances of the two methods
27
-------
made clear that the variability of the GC method exceeds that of the NDIR
procedure.
Agency Apportionment
Laboratories taking part in the 1979 CO Proficiency Surveys numbered
close to those participating during the previous year . State laboratory
response increased by approximately & percent, while local agencies were
about 9 percent less. The agencies comprised in the CO surveys are shown
in Table 15.
Data Summary
The results of the 1979 CO surveys, with all methods included, are
summarized in Table 16. The results arranged by analytical test method,
are shown in Table 17. Table 18 consists of the percentiles of reported
measurements which lie within the indicated percent (10, 20, 30 and 50)
of the expected value. The concentrations identified as expected values
were confirmed as the "true values" by analyses of the QAD corroborative
tests, and the analysis of the manufacturer.
TABLE 15. AGENCY APPORTIONMENT OF CARBON MONOXIDE SURVEYS
Agency
ERC (Federal)
Regional (Federal)
State Agencies
Local Agencies
I ndustri &1 /Contractor
Foreign
Survey 0379, %
0.3
2.0
46.9
46.2
2.3
2.3
Survey 0979, %
0.9
1.3
48.9
41.7
4.1
3.1
28
-------
cr.
c
(O
1.00
.80
.60
a? .40
>
s-
.20
0
.20 .40 .60 .80
Expected value, yq/ml
1.00
0.259 0.26
0.405 0.40
0.514 0.51
0.700 0.71
0.935 0.95
Figure 5. Mean values of NO^ survey 0679 vs. expected values.
cr>
3.
re
O)
0)
S-
3
C/)
1.00
.80
.60
.40
.20
0
.20 .40 .60 .80
Expected value, yg/mL
1.00
0.172 0.18
0.342 0.35
0.595 0.59
0.746 0.74
1.000 0.97
Figure 6. Mean values of NO^ survey 1279 vs. expected values.
29
-------
TABLE 16. SUMMARY OF CARBON MONOXIDE PROFICIENCY SURVEYS
Sample Respondents* Expected
No. value
ppm
Survey 0379
1
2
3
Survey 0979
1
2
3
(March 1979)
291
295
294
(September 1979)
290
268
297
6.53
19.8
43.7
2.98
14.8
33.8
Survey Survey
mean std. dev.
ppm ppm
6.39
20.08
44.14
2.73
14.75
34.14
0.53
0.89
1.51
0.48
0.75
1.37
Survey interval
ppm
4.92
17.30
38.23
1.25
12.24
29.50
- 8.50
- 23. CO
- 49.70
- 4.75
- 17.30
- 39.00
*With outliers removed.
TABLE 17. CARBON
MONOXIDE BY ANALYTICAL
(ppm)
METHOD
Sample No.
Survey 0379*
1
2
3
Survey 0979*
1
2
3
NDIR
N Mean Std.
(March 1979)
265 6.36 0.
268 20.03 1.
270 44.12 1.
(September 1979)
255 2.75 0.
229 14.77 0.
259 34.07 1.
dev. N
66 16
01 16
41 13
61 31
82 35
48 35
GC
Mean Std.
6.16 1.
19.78 1.
44.16 2.
2.69 0.
14.36 1.
34.31 2.
All other
dev. N
05 10
42 11
90 11
48 4
06 4
03 3
Mean
6.13
19.94
42.07
4.73
14.84
32.28
Std. dev.
0.67
0.62
7.42
3.52
1.01
3.97
*0utliers removed.
30
-------
TABLE 18. PERCENT OF CARBON MONOXIDE MEASUREMENTS
WITHIN INDICATED PERCENT OF EXPECTED VALUE
Survey 0379
Sample No.
1
2
3
10%
81
93
99
20%
94
99+
100
30%
98
100
100
50%
100
100
100
Survey
Sample No.
1
2
3
10%
46
92
97
0979
20%
77
100
100
30%
91
100
100
50%
97
100
100
The mean values from survey 0379, plotted against the expected values,
gave a linear relationship, as follows:
Survey average = -0.1461 + 1.0145 (expected value)
R2 = 1.0000
The plot of the survey means (y) against the expected values (x)
has been shown in Figure 7.
The means from survey 0979, plotted against the expected values,
gave a linear relationship, as follows:
Survey average = -0.3078 + 1.0179 (expected value)
R2 = 1.0000
A plot of the survey mean (y) against the expected values (x)
is shown in Figure 8.
Summary
Proficiency Surveys for CO were conducted in March and September 1979.
Operational assessment of approximately 300 instruments were included in
each of the semiannual surveys. The NDIR method was employed by approximately
31
-------
CXI
Survey mean, ppm
CQ
c
o
co
o
tn
o
ro
o>
=3
o
o>
c
ro
en
CO O
NJ O
c
-5
(D
O
(£>
VO
X
o
fD
o
rt
ro
Q.
a>
fD
v*
T3
ro
o
co
o
o
X
•a
(D
n
en
OJ
c.
ro
co
co
ro
00 00 VD
CO
co i—'
o en co
Survey mean, ppm
(Q
C
ro
ro
01
Q)
C
ro
o
o
c
-5
I
O
CO
ro
x
•o
fD
O
c+
ro
Q.
—• ro
o o
co
O
en
O
x
fD
o
n>
Q.
ro
a
3
(.0
-P.
o
en
o
a>
c
ro
VI
co
CT)
CO cn
co
o
O
oo
co
h
-------
90 percent of the survey respondents. No systematic discrepancies or sub-
stantial bias were identified in the CO surveys. The precision of the NDIR
method proved to be generally superior to that of the GC procedure.
SULFATE
Test Method
Participants in the 0279 and 0879 surveys employed four principal test
methods. They were the automated methyl thymol blue, manual barium chloride,
©
manual Sulfa-Ver and ion chromatograph procedures. There was little
change in the relative use of the methods between the semiannual tests,
with approximately 38 percent using the methyl thymol blue, 30 percent the
©
manual barium chloride, and 16 percent the Sulfa-Ver methods. Ten per-
cent used the sensitive and accurate ion chromatographic method. The
remainder utilized other methods or the methodology was undetermined.
The averages of all the named test methods were compared by an extension
of the analysis of variance at the 5 percent significance level. Since the
absolute differences between the means of the separate methods were less than
the critical value of w, there was no reason to believe that the averages
differed. The manual and automated barium chloride methods were the most
variable.
Agency Apportionment
Agency balance in the 1979 surveys differed little from the distri-
bution during the previous survey period . State and local agencies
accounted for better than 70 percent of the participation. The classi-
fication of agencies involved in the 1979 surveys is shown in Table 19.
33
-------
TABLE 19. AGENCY APPORTIONMENT OF SULFATE SURVEYS
Agency
ERC (Federal)
Regional (Federal)
State Agencies
Local Agencies
Industrial /Contractor
Foreign
Survey 0279, %
1.5
1.5
44.0
27.3
22.7
3.0
Survey 0879, %
1.7
1.7
41.6
26.7
21.6
6.7
Data Summary
The survey results of all methods are tabulated in Table 20. Results
listed by analytical method are shown in Table 21. Frequency distributions
of the percent of measurements falling within the indicated percent of
the expected value are presented in Table 22. The expected values used
in the sulfate surveys were the theoretical concentration of sulfate
ion of an inorganic salt which was deposited on glass fiber filter strips.
The theoretical values were confirmed by a corroborative laboratory and by
the QAD.
Anomalous measurements are excluded from the summary tables. Approxi-
mately 3 percent of the reported measurements in both surveys were rejec-
ted on the basis of Chauvenet's Criterion and the reviewers' judgment.
34
-------
TABLE 20. SUMMARY OF SULFATE PROFICIENCY SURVEYS
Sample Respondents
no.
Survey 0279
0
1
2
3
4
5
Survey 0879
0
2
3
4
5
(February 1979)
65
65
65
65
65
65
(August 1979)
59
58
60
58
59
Expected
value
ug/m
10.50
19.50
1.50
15.00
24.00
6.00
9.10
28.80
23.70
13.10
7.20
Survey
mean-
ug/m
10.61
18.96
1.96
14.74
23.80
5.75
8.38
27.42
22.90
12.16
6.77
Survey
std. dev.
ug/m
2.21
1.73
1.17
1.34
2.22
1.13
1.25
2.57
2.58
1.58
1.18
Survey interval
o
|jg/m
6. 65 ~
14.10 -
0.30 -
11.70 -
16.48 -
3.06 -
3.00 -
20.00 -
14.00 -
6.45 -
3.50 -
21.60
23.64
5.60
18.03
31.80
9.15
11.11
33.60
30.11
15.90
10.13
*With outliers removed.
35
-------
TABLE 21. SULFATE BY
ANA
LYTICAL METHOD
oo
Survey 0279* (February 1979)
Methylthymol blue-
automated
Sample No. N Mean Std. dev. N
0
1
2
3
4
5
26
26
26
26
26
26
10.32
18.42
1.60
14.60
23.22
5.85
1.00
1.95
0.87
1.07
1.20
0.74
Survey 0879* (August 1979)
Methylthymol blue-
manual
Sample No
0
2
3
4
5
N
1
1
1
1
1
Mean Std. dev.
6.80
25.20
21.00
10.10
4.70
0.00
0.00
0.00
0.00
0.00
Barium chloride-
manual
Mean Std. dev.
20 13.05
20 19.20
20 2. 72
20 16.12
20 25.87
20 6. 69
6.71
3.23
2.07
3.74
5.85
2.64
Methylthymol blue-
automated
N
22
21
22
22
22
Mean Std. dev.
8.41
27.12
22.15
12.27
6.80
0.82
2.56
4.06
1.11
0.75
Barium chloride-
automated
N Mean Std. dev.
3
3
3
3
3
3
9.78
21.78
0.93
15.14
28.07
5.10
2.74
1.97
0.90
2.62
7.85
1.97
Barium chloride-
manual
N
18
18
19
17
18
Mean Std. dev.
8.48
27.78
23.02
12.17
6.74
1.96
3.14
3.11
3.20
1.55
Sulfa-Ver-
manual
N Mean Std. dev.
10 10. 14
10 18.50
10 1.99
10 14.04
10 23.71
10 5.27
1.08
2.11
1.07
0.72
2.03
0.78
Ion chromatograph
N Mean Std. dev.
6 9.47
6 19.13
6 1.59
6 12.73
6 23.41
6 5.71
Barium chloride-
automated
N
1
1
1
1
1
Mean Std. dev.
7.30
23.35
20.70
11.07
5.46
1.96
0.00
0.00
0.00
0.00
N
10
10
10
10
10
1.30
1.12
0.60
4.01
1.23
0.95
Sulfa-Ver-
manual
Mean Std. dev.
8.44
28.61
24.12
13.85
7.00
3.16
3.90
2.93
4.67
1.36
Ion chromatograph
N
7
7
7
7
7
Mean Std. dev.
9.31
27.65
22.06
12.05
6.88
2.36
1.20
1.59
0.71
0.88
.
*0utliers removed.
-------
TABLE 22. PERCENT OF SULFATE MEASUREMENTS WITHIN
INDICATED PERCENT OF EXPECTED VALUE
Survey
Sample No.
0
1
2
3
4
5
10%
63
75
25
74
75
52
20%
80
89
40
91
92
74
0279
30%
89
95
43
97
94
82
Survey 0879
50%
94
100
57
97
97
95
Sample No.
0
2
3
4
5
10%
56
76
67
59
54
20%
83
91
92
84
80
30%
88
98
98
93
88
50%
93
100
100
95
98
The means from survey 0279, plotted against the expected values, gave
a linear relationship, as follows:
Survey average = 0.2542 + 0.9712 (expected value)
A plot of the survey 0279 means (y) against the expected values (x)
is shown in Figure 9.
The mean values from survey 0879, plotted against the expected values,
gave a linear relationship, as follows:
Survey average = -0.3604 + 0.09699 (expected value)
R2 = 1.0000
A plot of the survey means (y) against the expected values (x)
is shown in Figure 10.
37
-------
Summary
Proficiency Surveys for SO^ were conducted in February and August 1979.
Approximately 60 laboratories participated. Better than 70 percent of
the roster of participating laboratories comprised state and local agencies.
Six test methods were employed; the methyl thymol blue and the manual barium
chloride procedures dominated. Other major methods used were the Sulfa-
©
Ver and ion chromatography. No systematic discrepancies occurred. The
barium chloride procedures exhibited the greatest variability of the test
methods.
NITRATE
Test Method
Over half of the respondents reported using the automated cadmium
reduction method; the balance used one of eight other methods. Choice
of analytical method did not change notably between the biannual tests,
though participation was reduced in the later survey. The dominant
unlisted method was the ion chromatographic procedure. Others were
©
phenoldisulfonic acid, brucine, specific ion electrode, Szechrome
and ultraviolet spectrophotometric procedures. The averages of all
the methods, including all the unlisted methods, were compared by
an extension of the analysis of variance at the 5 percent significance
level. Since the absolute differences between the averages were less
than the critical value of w, it was concluded that the averages did
not differ. No single method was conspicuously imprecise.
Agency Apportionment
Involvement of state agencies in the nitrate surveys decreased by
approximately 3 percent from the previous survey year with a corres-
ponding increase in local air monitoring boards. State and local offices
embodied 65 percent of the total participation. The divisional composition
of the surveys is shown in Table 23.
38
-------
CO
30
24
£ 12
QJ
I 6
CO
12
18
24
30
Expected value, yg/nf
24.00 23.80
6.00 5.75
Figure 9. Mean value of SO. survey 0279 vs. expected values.
9.10 8.38
28.80 27.42
23.70 22.90
13.10 12.16
7.20 6.77
Expected value, yg/rf
Figure 10. Mean values of SO. survey 0879 vs. expected values.
39
-------
TABLE 23. AGENCY APPORTIONMENT OF NITRATE SURVEYS
Agency
ERC (Federal)
Regional (Federal)
State Agencies
Local Agencies
Indus trial /Contractor
Foreign
Survey 0279, %
1.9
1.8
46.3
18.5
27.8
3.7
Survey 0879, %
2.3
2.3
39.6
20.9
27.9
7.0
Data Summary
The survey results are listed in Table 24 inclusive of all methods.
Results by analytical method are given in Table 25 with the total of
unlisted methods included in the "all other" category. Frequency dis-
tributions by percent are shown in Table 26. The expected values
used in the nitrate surveys were derived from the theoretical amounts of
nitrate ion which were deposited on glass fiber filter strips. Chemical
composition was verified by corroborative tests carried out by an independent
laboratory and by the QAD.
Approximately 7 percent of the measurements were rejected on the
basis of independent judgment and Chauvenet's Criterion.
40
-------
TABLE 24. SUMMARY OF NITRATE PROFICIENCY SURVEYS
Sample Respondents*
no.
Survey 0279
0
I
2
3
4
5
Survey 0879
0
2
3
4
5
(February 1979)
50
50
50
50
50
50
(August 1979)
40
39
40
40
40
Expected
value
Mg/m
1.50
3.60
5.70
12.00
7.80
9.90
3.40
6.00
10.20
8.30
1.20
Survey
mean-
ug/m
1.58
3.52
5.52
11.54
7.60
9.68
3.44
5.90
9.97
7.94
1.21
Survey
std. dev.
ug/m
0.41
0.38
0.52
1.32
0.77
0.98
0.44
0.51
0.95
0.84
0.27
Survey interval
ug/m
0.35 -
2.53 -
4.39 -
7.83 -
5.04 -
7.32 -
2.37 -
4.22 -
7.12 -
4.99 -
0.36 -
3.39
4.60
7.29
14.79
9.16
13.74
4.40
6.88
11.57
9.20
1.82
*With outliers removed.
41
-------
TABLE 25. NITRATE BY ANALYTICAL METHOD
(ug/mJ)
Sample no.
Survey 0279*
0
1
2
3
4
5
Survey 0879*
0
2
3
4
5
Cadmium reduction-
manual
Expected
value N Mean Std. dev.
Cadmium reduction-
automated
N Mean Std. dev.
N
Hydrazine
reduction
Mean Std. dev.
Hydrazine reduction-
automated
N Mean Std. dev.
N
All others
Mean Std. dev.
(February 1979)
1.50
3.60
5.70
12.00
7.80
9.90
(August
3.40
6.00
10.20
8.30
1.20
3
3
3
3
3
3
1979)
4
4
4
4
4
1.67
3.38
5.26
10.84
6.88
8.42
3.48
5.31
9.40
7.33
0.99
0.23
0.32
0.68
1.69
1.70
1.11
0.78
0.80
1.70
1.37
0.42
29
29
29
29
29
29
20
19
20
20
20
1.68
3.58
5.61
11.66
7.55
9.50
3.47
5.87
10.24
8.69
1.65
0.87
1.01
1.16
1.98
1.18
1.57
0.30
0.67
1.30
2.47
1.82
2
2
2
2
2
2
2
2
2
2
2
1.68
3.52
5.09
8.78
6.53
7.83
3.14
6.01
9.18
7.91
1.11
0.02
0.32
0.27
0.26
0.18
0.47
0.91
0.59
0.51
0.61
0.55
5
5
5
5
5
5
3
3
3
3
3
1.71
3.73
5.55
11.84
8.15
8.35
3.40
5.72
10.06
7.93
1.06
0.12
0.12
0.34
0.71
0.65
3.64
0.07
0.24
0.25
0.39
0.11
11
11
11
11
11
11
11
11
11
11
11
1.95
3.44
5.77
11.01
7.37
10.01
3.42
5.81
9.92
7.82
1.31
1.92
0.49
1.58
1.42
0.88
1.43
0.53
0.86
1.42
1.14
0.29
*With outliers removed.
-------
TABLE 26. PERCENT OF NITRATE MEASUREMENTS WITHIN
INDICATED PERCENT OF EXPECTED VALUE
Survey 0279
Sample No.
0
1
2
3
4
5
10%
52
72
74
74
78
78
20%
74
90
88
86
94
90
30%
82
96
94
96
96
94
50%
88
96
94
98
98
96
Sample No.
0
2
3
4
5
Survey 0879
10%
55
79
75
80
55
20%
88
89
90
93
70
30%
98
95
93
93
83
50%
100
100
100
98
93
The mean values from survey 0279, plotted against the expected values,
gave a linear relationship, as follows:
Survey average = 0.1117 + 0.9573 (expected value)
R2 = 0.9998
A plot of the survey means (y) against the expected values (x) is
shown in Figure 11.
The mean values from survey 0879, plotted against the expected values,
gave a linear relationship, as follov/s:
Survey average = 0.1002 + 0.9608 (expected value)
R2 = 0.9994
A plot of the survey means (y) against the expected values (x) is
shown in Figure 12.
43
-------
12.0
°°E 9.6
(13
OJ
7.2
oT 4.8
S-
" 2.4
0
2.4 4.8 7.2
[\xoected value,
9.6
12.0
x
1.50
3.60
5.70
12.00
7.80
9.90
1.58
3.52
5.52
11.54
7.60
9.68
Figure 11. Mean values of NO, survey 0279 vs. expected values.
12.0
9.6
7.2
(C
c' 4.8
CD'
>
2.4
2.4
4.8 7.2
Exoected value,
9.6
12.0
3.40
6.00
10.20
8.30
1.20
Figure 1?. Mean values of NO. survey 0879 vs. expected values
44
3.44
5.90
9.97
7.94
1.21
-------
Summary
Proficiency Surveys for NO- were completed by approximately 45 partici-
pants in February and August 1979. The predominant test method was the
automated cadmium reduction procedure?, used by over half the respondents.
In all, nine analytical methods were reported in use. No method was
particularly bad and no systematic discrepancies existed.
LEAD
Test Method
Virtually all participants in both the 0179 and 0779 surveys employed the
atomic absorption analytical method as they did in previous years. Only one
participant reported using another method, the anodic stripping voltmeter
method.
Agency Apportionment
State and local agencies taking part in the lead survey together
decreased by 17 percent while participation of private industry
increased by 1.2 percent from the previous year. The categories of
agencies which were involved in the lead surveys are listed in Table 27.
TABLE 27. AGENCY APPORTIONMENT OF LEAD SURVEYS
Agency
ERC (Federal)
Regional (Federal)
State Agencies
Local Agencies
Industrial /Contractor
CAMP
Foreign
Survey 0179, %
1.3
7.9
47.4
25.0
17.1
1.3
0.0
Survey 0779, %
0.0
6.9
44.8
25.3
19.5
1.2
2.3
45
-------
Data Summary
The survey results are tabulated in Table 28. Results by analytical
method are summarized in Table 29. Frequency distributions of the percent
of the expected value are shown in Table 30. The expected values used in
the lead survey were designated after considering the results of corroborative
analyses.
Anomolous measurements are not included in the summary tables. Judgment
and Chauvenet's Criterion were used to reject approximately 1 percent of
the measurements in the 0179 survey and 8 percent in the 0779 test.
TABLE 28. SUMMARY OF LEAD PROFICIENCY SURVEYS
Sample Respondents*
no.
Survey 0179
3
4
5
6
7
8
Survey 0779
0
1
2
3
4
5
(January 1979)
76
76
76
76
76
76
(July 1979)
81
81
79
81
80
81
Expected
value
ug/m
1.46
3.52
5.39
11.45
7.48
9.51
3.53
12.86
5.85
10.39
7.93
1.18
Survey
mean-
ug/m
1.45
3.46
5.39
11.31
7.61
9.38
3.53
12.85
5.90
10.42
7.99
1.17
Survey
std. dev.
ug/m
0.16
0.28
0.57
1.15
1.10
0.96
0.29
0.87
0.34
0.91
0.48
0.12
Survey interval
ug/m
0.93 -
2.84 -
2.88 -
7.74 -
4.80 -
6.18 -
2.34 -
10.59 -
4.77 -
7.50 -
6.81 -
0.84 -
1.95
4.32
6.63
14.10
12.00
11.57
4.20
15.30
6.97
13.80
9.97
1.50
*With outliers removed.
46
-------
TABLE 29. LEAD BY ANALYTICAL METHOD
Sample
Survey
3
4
5
6
7
8
Survey
0
1
2
4
5
3
Atomic absorption
no. N
0179* (January 1979)
75
75
75
75
75
75
0779* (July 1979)
81
81
79
80
81
81
Mean Std.
1.
3.
5.
11.
7.
9.
3.
12.
5.
7.
1.
10.
48
49
50
19
71
42
53
85
90
99
17
42
0.
0.
1.
1.
1.
1.
0.
0.
0.
0.
0.
0.
dev.
24
50
15
48
68
72
29
87
34
48
12
91
All others
N
1
1
1
1
1
1
Mean
1.30
3.20
5.10
10.60
7.00
8.70
Std. dev.
0.00
0.00
0.00
0.00
0.00
0.00
*0utliers removed.
TABLE 30. PERCENT OF LEAD
INDICATED PERCENT OF
MEASUREMENTS WITHIN
EXPECTED VALUE
Sample
3
4
5
6
7
8
Survey 0179
No. 10% 20% 30%
74 92 95
79 97 99
76 91 96
78 91 97
76 91 92
70 93 97
50%
99
99
97
99
93
99
Sample
0
1
2
3
4
5
Survey
No. 10%
85
89
92
86
95
74
0779
20%
98
100
100
95
98
100
30% 50%
99 100
100 100
100 100
98 100
99 100
100 100
47
-------
The mean values from survey 017*), plotted against the expected values,
gave a linear relationship, as follows:
Survey average = 0.0328 + 0.9895 (expected values)
R2 = 0.9994
A plot of the survey means (y) against the expected values (x) is
shown in Figure 13.
The mean values from survey 0779, plotted against the expected values,
gave a linear relationship, as follows:
Survey average = 0.0122 + 1.0011 (expected value)
R2 = 1.0000
A plot of the survey 0779 means (y) against the expected values (x)
is shown in Figure 14.
Summary
Proficiency Surveys for Pb were conducted in January and July 1979, with
approximately 80 facilities participating. The atomic absorption method of
analysis was virtually the only method used. No systematic discrepancies
or bias was observed in the Pb surveys.
HI-VOL FLOW
Test Method
Six measurement methods were listed by survey participants. The
pressure transducer and rotameter dominated with 72.6 percent using those
methods. Other minor methods, as noted on the survey information forms,
were flow gauge, manometer, orifice manometer, and magnehelic gauge. Five
percent of the methodologies were not reported. The categories of the
various methods, together with a summary of sites, is given in Table 31.
48
-------
E
-------
TABLE 31. ANALYTICAL METHODS - HI-VOL
Method No. of sites
Pressure transducer 546
Rotameter 496
Flow gauge 72
Manometer 56
Orifice manometer 36
Magnehelic gauge 34
Other 58
Agency Apportionment
The number of air monitoring offices taking part in the 0579 survey
and the distribution of agencies remained close to the same as the previous
year . The division of agencies is shown in Table 32.
TABLE 32. AGENCY APPORTIONMENT OF HI-VOL SURVEY
Agency Survey 0579, %
ERC (Federal) 0.3
Regional (Federal) 0.8
State Agencies 40.6
Local Agencies 53.9
Industrial/Contractor 1.8
Foreign 2.6
50
-------
Data Summary
It was not practical to furnish standardized ReF specimens from
an invariant population of samples. Each sample ReF unit was calibrated
individually and is distinct from all other units, in as much as flow
rates could not be duplicated precisely enough among the test devices to
establish flow values that were applicable to all units. Rather than
comparing collective measurements against a common standard, results
from each ReF are compared to a calibration which is unique for that
unit. The calibrated flow values, or expected values, developed for
each of five flow constrictor plates, are compared to the values recorded
on the survey forms. To facilitate the comparison, the expected values
are termed "x" and the reported values are referred to as "y". The five x,
y pairs are plotted to obtain a linear regression plot, with the expectation
that the coefficient of linearity (r) should be 1, the slope should be 1,
and the y intercept should pass through the x, y origin. Departures from
this norm are due to any single or combination of defects that falsify
the measured flow. Three categories of measurements were evaluated—data
obtained by use of a pressure transducer, those obtained by a rotameter
and those from all methods combined together. A substantial part of the
measurements fall in the "all other" category. Table 31 makes reference
to the secondary methods that were used. The linear regression equations
of the three categories are given in Table 33.
TABLE 33. LINEAR REGRESSION EQUATIONS OF PAIRED VALUES
All Methods: y = 0.925x + 3.560 (9)
Pressure transducer: y = 0.943x + 2.498 (10)
Rotameter: y = 0.907x + 4.696 (11)
where: y = reported value, cu. ft./min.
x = expected value, cu. ft./min.
51
-------
Frequency distributions of the percent differences between the
reported and expected values for each measurement pair are shown in
Table 34.
TABLE 34. PERCENT DIFFERENCE FROM EXPECTED FLOW
Number of
Measurements*
5902
Percent of Measurements
10%
-9.2
20%
-4.8
30%
-2.6
50%
0.3
70%
3.0
90%
7.4
*With outliers removed.
Summary
The Proficiency Survey for hi-vol flow was conducted in May 1979 with
approximately 1,300 sites being tested. Foreign, federal, state and
industrial laboratories participated in the testing program. The pressure
transducer and rotameter were the most common methods used with four other
minor methods. A slight bias existed in the major methods and in the data
from all methods combined.
52
-------
SECTION 6
REFERENCES
1. Barnet, Vic and T. Lewis. Outlines in Statistical Data, John Wiley
and Sons, New York, New York, pp. 18-19 (1978).
2. Chauvenet, William. A Manual of Spherical and Practical Astronomy.
J.B. Lippincott and Co.: Philadelphia, PA (1863).
3. Natrella, Mary. Experimental Statistics. National Bureau of Standards
Handbook 91. U.S. Government Printing Office, Washington, D.C.,
Chapter 3.
4. Harshbarger, Thad. Introductory Statistics. The MacMillan Co.:
New York, New York, pp. 145-146 (1971).
5. Code of Federal Regulatons (40 CFR) 50.11, Appendix A, pp. 5-11.
6. Federal Register, Vol. 38, No. 110, June 8, 1973, pp. 15175-15176.
7. Bromberg S., B. Bennett, and R. Lampe. Summary of Audit Performance
Measurement of SO.,, NO,,, Sulfate, Nitrate, Lead, and Hi-Vol Flow Rate -
1978. EPA-600/4-80-017, 1980.
53 tt U.S. GOVERNMENT PRINTING OFFICE: 1981--757-064/0319
------- |