United States
Environmental Protection
Agency
Environmental Monitoring
Systems Laboratory
Las Vegas. NV 89193-3478
Research and Development
EPA/600/S2-87/082 July 1988
Project Summary
Two Test Procedures for
Radon in Drinking Water:
Interlaboratory Collaborative
Study
E.L. Whittaker, J.D. Akridge and J. Giovino
Two analytical methods for the
determination of radon concentra-
tions in water were tested in a
multilaboratory study with twenty-
eight participating laboratories.
Eighteen laboratories analyzed pre-
pared samples by the liquid
scintillation (LS) method, and twelve
laboratories analyzed the same
samples by the Lucas Cell (LC)
method. Several laboratories ana-
lyzed the samples by both methods.
Because of the short half life of
radon-222 (3.82 days), special
standard and samples were prepared
in which each standard and each
sample contained Its own sealed
radlum-226 source that emanated
radon-222 Into the standard and
sample water containment. There
was a radon hold-back loss factor
associated with the standards and
samples that were provided to the
participant laboratories. However,
because the standards and samples
were prepared identically, the radon
hold-back loss was common to
standards and samples alike and did
not bias the test results.
A comparison of the grand
averages for the three samples with
the known values for those samples
showed good accuracy for both
methods. The accuracy index was
not less than 94 percent for any of
the three samples when analyzed by
either method. The average accuracy
for the LS method for the three
samples was 95.2 ± 2.0 percent, and
for the LC method it was 100.7 ± 4.6
percent at the 95 percent confidence
level.
Test results for the LS method
showed better precision than test
results for the LC method. The aver-
age repeatability (within-laboratory
precision) for the LS method was 3.6
± 3.0 percent at 95 percent
confidence, and for the LC method it
was 6.4 i 3.8 percent at 95 percent
confidence. The average reproduc-
ibility (combined within- and be-
tween-laboratory precision) for the
LS method was 10.2 ± 4.2 percent at
95 percent confidence, and for the LC
method it was 17.6 ± 4.2 percent at
95 percent confidence.
The importance of the sampling
technique to the analytical accuracy
is discussed in the project report.
The authors and the Project
Officer recommend that the two
analytical methods be considered as
validated and equivalent methods.
This Project Summary was
developed by EPA's Environmental
Monitoring Systems Laboratory, Las
Vegas, NV, to announce key findings
of the research project that is fully
documented in a separate report of
the same title (see Project Report
ordering information at back).
Introduction
This is a summary report of a
multilaboratory test of two analytical
-------
methods for the purpose of validating
those methods. In one method, liquid
scintillation (LS) counting of 10 ml
portions of water samples for the alpha
and beta particle emissions from the
radon and its short lived decay progeny
is used to determine radon/water con-
centrations. In the second method, the
radon gas is emanated from measured
portions of water samples into prer
evacuated Lucas Cells (LC) in which only
the alpha particle emissions from the
radon and its short lived decay progeny
are counted to determine radon/water
concentrations. Twenty-eight laborato-
ries participated in the multilaboratory
study, eighteen laboratories in the LS
method and twelve laboratories in the LC
method.
Standards and samples that con-
tained their own radium-free-radon/
water generating sources were prepared
and used in the study.
This study was conducted by the
principal author as an employee of
Lockheed Engineering and Management
Services Company, Inc., for the EPA
under Contract No. 68-03-3249.
Procedures
1. Analytical Test Procedures
The two analytical methods that
were tested in this study are described in
detail in Appendix B and Appendix D of
the project report. The Appendix B
method is the EPA liquid scintillation
method. The Appendix D method is the
Lucas Cell detector method detailed for
water grab samples.
2. Collaborative Test Procedure
Forty laboratories responded to an
invitation to participate in the two-
method multilaboratory validation study.
Standards, samples and instructions
were sent to the participating labora-
tories. Twenty-eight laboratories sub-
mitted test result data, eighteen
laboratories for the LS method and
twelve laboratories for the LC method.
Precision, accuracy and bias values
were calculated for the three samples
that were analyzed by each of the two
methods. The radon/water concentration
levels in Samples A, B and C were 1
pCi/L, 16,000 pCi/L and 66,000 pCi/L,
respectively.
Results and Discussion
Table 1 lists the test result averages,
individual laboratory standard deviations
from replicate analyses, the number of
replicates averaged, and the factor
(cpm/pCi) for converting counts per
minute to picocuries of radon-222, for
the liquid scintillation (LS) method. Only
two laboratory average outliers were
found in those test results (both for
Sample B). Laboratory 21 analyzed the
samples two times by the LS method,
once using the mineral oil liquid
scintillator cocktail and once using a
detergent-type liquid scintillator cocktail
(data included in Table 1).
Table 2 is a statistical summary of
the accuracy, bias and precision of the
Table 1 test results, as calculated by the
equations given in the Data Processing
Procedures. The 95.2 percent accuracy
index corresponds to a negative bias of
4.8 ± 2.0 percent (95 percent confi-
dence). The 4.8 ± 2.0 percent negative
Table 1. Uquid Scintillation Method
Factor
Laboratory cpm/pCi±S
Radon-222 pd/1000 g±S
Sample A
Sample B
Sample C
. Replicates
1
2
3
4
6
7
10
11
13
15*
17
18
18
20
21
21*
23
26
27
27
X±SX
Known Value (Y)
8.88 ± 0.00
8.25 ± 0.23
7.26 ± 0.19
8.45 ± 9.05
8. 13 ± 0.34
7.50 i 0.23
8.59 * 0.06
7.57 ± 0.45
6.75 ± 0.08
7.70 ± 0.17
8.30 £ 0.49
8.40 ± 0.08
8.12 ± 0.37
8.08 ± 0.10
8.19 ± 0.05
9.12 ± 0.09
6.82 ± 0.003
9.00 ± 0.09
3.64
3.34 + 0.47
1,571 ± 31
1,570 ± 31
1,599 ± 75
1,568 ± 8
1,283 ± 82
1.656 ± 62
1,501 ± 11
1,358 ± 78
1,671 ± 33
1,604 t 152
1,280 ± 20
1,629 ± 18
1,568 t 97
1,828 ± 33
1,614 i 22
1,567 ± 28
1,556 ± 30
1,553 ± 31
1,290
1,117 ± 309
1,520 t 769
7,620 ± 49
16,308 ± 46
16,065 ± 235
15,404 ± 293
16,270 ± 33
16,966 ±419
16,734 ± 237
14,943 ±1112
(10,387 ± 646p
16,916 ± 174
16,548 ± 487
17,060 ± 110
15,158 ± 166
14,722 ± 581
12,968 ± 362
16,660 ± 157
15,300 ± 229
15,281 ± 256
15,920 ± 360
14,200
(9,804 ±5,665/>
75,700 ± 7700
76,300 ± 501
67,076 ± 30
65,797 ±1160
58,934 ± 750
51,916 ± 130
67,666 ± 1014
68,432 ± 1416
61,020 ± 660
68,366 ± 309
55,733 * 7878
48,940 t 2543
52,050 ± 1140
65,862 ± 295
63,341 ± 3220
76,640 ± 132
69,427 ± 250
60,920 ± 627
65,375 ± 1039
66,560 ± 625
60,700
66,141 ± 1380
62,900 ± 6600
66,200 ± 2030
2
6
4
2
3
5
2
3
3
6
2
3
6
3
3
3
2
3
1
2
aDetergent liquid scintillator was used instead of mineral oil liquid scintillator.
bOutlier not used in the grand average.
-------
Table 2.
Parameter9
Yj (pCi/L)
Xj (pdlL)
Aj (%)
Bias (°/op
SXj (pd/L)
Sri (PCi/L)
Sy (pCUL)
Spj (pCi/L)
V/y (%)
VLJ (%)
vRi (%)
T/-1C
r,-3
Liquid Scintillation Method (Accuracy, Bias, and Precision Summary)
Sample A Sample B Sample C Average % Values ± S
1620 ± 49
1520
94.0
-6.0
169
87
168
189
5.7
11.1
12.4
2.56
1.44
16,300 ± 501
15,700
96.5
-3.5
1,097
385
1,090
1,160
2.5
6.9
7.4
2.22
0.35
66,200 ± 2030
62,900
95.0
-5.0
6,653
1,628
6,640
6,840
2.6
10.6
10.9
2.22
2.01
95.2 ± 1.0
-4.8 ± 1.0
3.6 i 1.5
9.5 i 1.9
10.2 ± 2.1
Parameters are described in the test.
bThe sign before the number indicates the direction of bias.
cThe critical value (tc) for significant difference at the 5 percent significance level for 18 labs
(Sample B) is 2.105 and for 20 labs (Samples A and C) is 2.090.
Tj1 values are calculated from Table 1 test results.
T-3 values are calculated from Table 3 test results.
bias is a significant (real) bias as
indicated by the T values for the samples
being greater than the critical values for
T. However, that bias is not a serious
bias and is likely due to a loss of radon
activity in the transfer of successive
aliquots from the same sample bottle
water for replicate analyses.
Table 2 shows that the estimated
average repeatability (within-laboratory
precision) of the liquid scintillation
method over the radon/water
concentration of 1,600 to 66,000 pCi/L is
3.6 ± 3.0 percent at 95 percent
confidence.
Table 2 shows that the estimated
average reproducibility (combined with-
in- and between-laboratory precision)
of the liquid scintillation method over the
radon/water concentration range of 1,600
to 66,000 pCi/L is 10.2 + 4.2 percent at
95 percent confidence.
Tables 3, 4 and 4a are not listed in
this summary report but are included in
the Project Report.
Table 5 lists the laboratory test result
averages, their standard deviations from
replicate analyses, the number of
replicates averaged, and the factor
(cpm/pCi) for converting counts per
minute to picocuries of radon-222, for
the Lucas Cell (LC) method. No outliers
were found in the Table 5 test result
averages.
Table 6 is a statistical summary of
' the accuracy, bias, and precision of the
Table 5 test results, as calculated by the
equations given in the Data Processing
Procedures. A comparison of the known
values for Samples A, B, and C with the
respective grand averages of the Table 5
test results shows an average accuracy
index of 100.7 ± 4.6 percent at 95
percent confidence. The bias values of
+ 1.9 percent, +2.7 percent and -2.6
percent for Samples A, B and C,
respectively, are within the 95 percent
confidence limits of the average
accuracy index. The t-test to show
significant difference applied to known
values and grand averages for the
samples shows that there are no
significant differences (T values for the
samples are less than the critical value).
Table 6 shows that the estimated
average repeatability (within-laboratory
precision) of the Lucas Cell method over
the radon/water concentration range of
1,600 to 66,000 pCi/L is 6.4 + 3.8
percent at 95 percent confidence.
Table 6 shows that the estimated
average reproducibility (combined with-
in- and between-laboratory precision)
over the radon/water concentration range
of 1,600 to 66,000 pCi/L is 17.6 ± 4.2 at
95 percent confidence.
Conclusions
A satisfactory multilaboratory test of
the two analytical methods was demon-
strated by the low number of outlier test
results (2 out of 60 laboratory averages
for the liquid scintillation method and
none for the Lucas Cell method).
Equivalency of the two methods was
demonstrated by the high accuracy of
the test results obtained by both methods
(accuracy index was not less than 94
percent for any of the three samples
when analyzed by either method), and
the lack of a serious bias by either
method.
A comparison of Table 2 and Table 6
shows the liquid scintillation (LS) method
had better precision than the Lucas Cell
(LC) method. The average repeatability
(within-laboratory precision) for the LS
method was 3.6 ± 3.0 percent at 95
percent confidence, and for the LC
method it was 6.4 + 3.8 percent at 95
percent confidence. The average repro-
ducibility (combined within- and
between-laboratory precision) for the LS
method was 10.2 + 4.2 percent at 95
percent confidence, and for the LC
method it was 17.6 ± 4.2 percent at 95
percent confidence.
There was a combined hold-
back/transfer radon loss (HB/TL) as-
sociated with the type of standards and
samples that were provided to the
participants. However, since standards
and samples were prepared identically,
the hold-back/transfer radon loss was
common to standards and samples alike
and did not bias the test results.
The variations in the cpm/pCi factors
in Table 1 are not reflected by cor-
responding differences in the sample
radon-222 concentration test results.
This shows that the differences in the
sample aliquot transfer technique used
by the laboratories did not significantly
affect the test results because standards
-------
Table 5.
Lucas Cell Method
Laboratory
5
8
12
14
16
17
19
21
22
24
25
28
X
s*
Known Value (Y)
Factor
cpm/pd ± S
3.79 ± .08
4.24 i .04
2.50 i .04
4.78 ± .13
4.89 ± .04
4.67 + .21
4.66 ± .15
±
4.57 + .24
1.53 i .13
0.1053
Radon-222 pd/1000 g±
Sample A
1,838 ± 96
1,608 ± 26
1,552
1,702 ± 93
1,510 ± 84
1,495 ± 265
1,636 ± 60
1,586 ± 68
2,059 i 81
1,550 ± 85
2,100 ± 200
1650
261
1620 ± 49
Sample B
17,450 ± 815
14,514 ± 72
17,550
16,493 ± 1452
16,746 ± 935
16,950 i 300
16,680 ± 827
16,466 ± 277
13,537 i 379
19,600 t 580
19,257 ± 792
21,000 ± 2000
16,770
2480
16,300 ± 507
S ;
Sample C
72,280 ± 12,418
45,730 ± 392
60,780
73,921 ± 2604
65,854 ± 2510
49,200 ± 520
67,756 ± 2459
66,197 ± 1445
79,670 ± 5752
77,900 ± 2339
70,000 ± 7000
64,520
12,020
66,200 ± 2030
Replicates
(n)
5
2
1
3
6
2
6
3
5
5
2
-
and samples were transferred by the
same technique within each laboratory.
A comparison of the cost per
analysis between the two methods favors
the LS method significantly when LS
counting capability is available to the
analyst.
Recommendations
The authors recommend that the two
analytical methods tested in this multi-
laboratory test be considered validated
and equivalent for the determination of
radon/water concentrations in potable
water systems.
It is recommended that sampling be
considered as a critical part of the ana-
lytical procedure for analytical methods
that are specified for the determination of
radon/water concentrations. A positive
pressure sampling or transfer technique
should be used, avoiding negative
pressure techniques, aeration, and tur-
bulence whenever it is possible.
For the LS method it is recom-
mended that samples be transferred to
LS vials directly at the sampling site as
described in the EPA method (Appendix
B), but by a positive pressure technique
similar to the one described in the NIRS
Sampling Instruction-Radon (Appendix
C of the Project Report), filling the LS
vial only to the shoulder of the bottle.
Pre-weighing the LS vials containing 10
mL of mineral oil cocktail and weighing
again after water sample has been added
provides for determining sample size.
Poly Seal caps on the LS vials seem to
retain the mineral oil cocktail better than
other caps.
The emanation bubblers for the LC
method are both fragile and expensive.
Therefore, it is recommended that
samples be collected in the field in 4-
ounce or larger glass bottles fitted with
Poly Seal caps and the sample bottles
brought or sent to the laboratory for an
early analysis by the LC method.
Samples should be collected by a
positive pressure sampling technique as
described in the Appendix D procedure
of the Project Report. An alternative
positive pressure sampling technique has
been described by the Sanitation and
Radiation Laboratory of the California
Health Department (Appendix E of the
Project Report).
-------
Table 6.
Parameter8
Yj (pCi/L)
Xj (pCi/L)
Aj (%)
Bias (%/>
SX (pCi/L)
Srj (pCUL)
Sij (pCUL)
Sp, (pCi/L)
vr/ (%)
Vtj (%)
VR/ (%)
Tf
Lucas Cell Method (Accuracy, Bias,
Sample A Sample B
1620 ± 49
1650
101.9
+ 1.9
261
94
259
276
5.7
15.7
16.8
0.411
16,300 ± 501
15,770
102.7
+ 2.7
2,480
776
2,470
2,590
4.6
14.8
15.5
0.624
and Precision Summary)
Sample C Average % Values ± S
66,200 ± 2,030
64,500
97.4
-2.6
12,020
5,792
11,900
13,200
9.0
18.4
20.5
0.498
100.7 ± 2.3
6.4 ± 1.9
16.3 t 1.5
17.6 ± 2.1
Parameters are described in the test.
bThe sign before the number indicates the direction of bias.
cThe critical value (Tc) for significant difference at the 5 percent significance level for 12 labs
(Sample B) is 2.18 and for 11 labs (Samples A and C) is 2.20.
E.L Whittaker is with Lockheed Engineering and Management Services
Company, Inc. Las Vegas NV 89114; J.D. Akridge and J. Giovino, EPA
authors, are with the Environmental Monitoring Systems Laboratory, Las
Vegas, NV 89193-3478.
Chung-King Uu is the EPA Project Officer (see below).
The complete report, entitled "Two Test Procedures for Radon in Drinking
Water: Interlaboratory Collaborative Study," (Order No. PB 88-197
306/AS; Cost: $14.95, subject to change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Environmental Monitoring Systems Laboratory
U.S. Environmental Protection Agency
Las Vegas, NV 89193-3478
------- |