Results of the intercalibration study of laboratories involved in assessing the environmental consequences of the Chernobyl accident

Appl. Radiat. Isot. Vol. 43, No. 1/2, pp. 149-160, 1992 Int. J. Radiat. Appl. Instrum. Part A Printed in Great Britain. All rights reserved

0883-2889/92 $5.00+ 0.00 Copyright © 1991 PergamonPress plc

Results of the Intercalibration Study of Laboratories Involved in Assessing the Environmental Consequences of the Chernobyl Accident E.L. C O O P E R * , V. V A L K O V I C , V. S T R A C H N O V , R. D E K N E R , P.R. D A N E S I

International Atomic Energy Agency, Agency's Laboratories, Seibersdorf, Austria *On attachment from AECL-Research, Chalk River Laboratories, Chalk River, Canada

Within the framework of the International Chernobyl Project, the IAEA's Seibersdorf Laboratories organized an intercalibration exercise among some of the laboratories which were involved in assessing the environmental contamination in the USSR due to the accident. The objective was to assess the reliability of the radioanalytical data for food and environmental samples, which were used to assess the doses. In the initial study reference materials from the stocks of the IAEA's Analytical Quality Control Services (AQCS) were re-labelled and submitted to 71 laboratories as blind samples. These natural matrix materials included samples of milk (containing 2 different levels of radioactivity), soil, air •ters and clover. The concentrations of radionuclides in these samples were known from previous intercalibration exercises. The overall range in performance was broad, which is similar to what has been observed in previous international intercomparisons. The results obtained by gamma-ray spectrometry tended to be somewhat underestimated, on average. On the other hand, the laboratories showed an overall tendency to overestimate 90St and possibly 239pu, which were analysed radiochemically. The intercalibration exercise is continuing with nine materials, including: soil, grass, hay and milk powder contaminated with fallout from the Chernobyl accident. These materials, which were prepared by laboratories in the USSR, are now being tested by AQCS prior to future intercomparison exercises. Work with these materials is expected to continue for several years.

Introduction In 1990 the International A t o m i c E n e r g y A g e n c y ( I A E A ) initiated the International C h e r n o b y l P r o j e c t in o r d e r to assess the r a d i o l o g i c a l c o n s e q u e n c e s in the U S S R due to the C h e r n o b y l r e a c t o r accident. T a s k 2 o f the project i n v o l v e d the c o r r o b o r a t i o n o f the a s s e s s m e n t o f e n v i r o n m e n t a l c o n t a m i n a t i o n , while part o f T a s k 3 included the a s s e s s m e n t o f doses b a s e d on the a v a i l a b l e e n v i r o n m e n t a l data. W i t h i n the f r a m e w o r k o f T a s k 2, an 149

150

Analytical quality control

intercalibration exercise was organized by the Analytical Quality Control Services (AQCS) of the IAEA's Seibersdorf Laboratories. The main objective of this exercise was to assess the reliability of the analytical data which was used for the dose assessment. However, the results of the exercise also provide valuable feed-back to the laboratories, which can review their performance and take remedial action in cases where performance is below par.

Organization of the Intercalibration Exercise The intercalibration exercise was organized in two phases. Phase 1, which involved the analysis of available reference materials, will be completed in 1991o Phase 2 involved the collection and preparation of materials, which were contaminated by radioactive fallout from the Chernobyl reactor accident, by a group of laboratories in the USSR in collaboration with AQCS° Over the next few years these materials will undergo further preparation, testing and intercomparison and they will eventually be certified as new reference materials. The materials used in Phase 1 were selected from the available stocks of AQCS reference materials which had previously been certified for radionuclide content through international intercomparison exercises or which were being intercompared in 1990. These materials, which included samples of milk (two different levels - H and L), soil, air filters and clover, were re-labelled and submitted to the laboratories as blind samples. The concentrations of radionuclides in these reference materials, which present a variety of radioanalytical challenges, are summarized in Table 1. Full descriptions of the materials are given in: Cooper et al., 1988a; Strachnov et al., 1990; Pszonicki et al., 1984; Cooper etal., 1988b; and Strachnov etal., 1991. A new reference date of 1989 January 1 was given for all of the materials and the values in Table 1 have been corrected for decay accordingly. Because of the different reference dates, some of the values in Table 1 differ from those given in the original reports. Sample sets were distributed to 71 institutions during 1990. Samples of milk (H), milk(L), soil and clover were sent to all of the institutions; however, air filters were distributed only to selected institutions because the available stocks were limited. In addition to the samples, copies of the information sheets on the materials and instructions for reporting the results were sent to the institutions,

Results and Discussion Results from 23 institutions (2 sets were received from one of them) have been used in the preparation of this paper. The numbers of results received from the various institutions were quite variable in terms of analytes, replicates and materials analysed. This has imposed some limitations on the analysis of the results° Fewer results were reported for radiochemical analyses than for analyses by gamma spectrometry. Individual means and standard deviations were calculated for each radionuclide in each material from the results reported by the individual laboratories, It was decided to present the means and standard deviations in plots which are similar to those proposed by Youden (presented in Youden and Steiner 1975) for intercomparison exercises. In his plots,

Analytical quality control

151

Table 1: Concentrations of Radionuclides in the Reference Materials Radionuclide

Milk Powder (H):

Recommended Value (Bq/kg)

Confidence Interval* (Bq/kg)

IAEA 152

137C5 134Cs

2065

1991-2143

489

462- 513

9°Sr

7.45

6.78-8.03

9°K

539

510- 574

Milk Powder (L):

IAEA 321

137Cs

72.6

71.1 - 74.2

134Cs

15.5

14.8 - 16.2

9OSr

3.3

3.16 - 3.44

4OK

552

536 - 569

Soil:

Soil 6

137Cs

46.8

44.9 - 50.5

226Ra

79.7

69.4 - 93.2

239pu

1.04

0.96 - 1.11

9°Sr

26.3

21.0 - 27.4

Air •ters:

IAEA 083 (Bq/filter)

(Bq/ftlter)

60Co

1456

1411-1501

9°Sr

215

206 - 224

137Cs

1102

1043-I 163

210pb

138

104- 170

133Ba

712

661 - 763

Clover:

IAEA 156 (Bq/kg)

(Bq/kg)

137Cs

264

253 - 274

134Cs

132

126- 138

40 K

657

637 - 676

9°St

14.8

13.4-16.5

* See original reports for criteria applied.

152

Analytical quality control

Youden graphed the individual results obtained by each laboratory for one material against their results for another, similar material. These plots allow one to visually assess the performance of individual laboratories and they highlight any consistent biases in the results. In this paper such plots have also been used to compare the performance of the laboratories with different matrix materials, different radionuclides and different levels of the same radionuclide. The results for the analysis of 137Cs in the two milk powder samples are shown in Figure 1. The IAEALAQCS confidence limits on the reference value for milk(L) are shown as two dashed vertical lines, while those for milk(H) are shown as dashed horizontal fines. The shaded rectangle where the confidence limits for the two materials intersect is the area where consistently good results would lie. The individual results for each material, plotted against each other, are shown as dots. The horizontal error bars are for milk(L), while the vertical bars are for milk (H~o Error bars are not shown where the standard deviations could not be calculated° ~:rom the variations in lengths of the error bars it is evident that there is a wide range in the precision attained by individual laboratories. In order to put the plots into perspective the upper-left, lower-left and lower-right comers have been labelled with the corresponding percentage deviations from the reference value (i.e. the numbers in brackets are the percentage biases at the edges of the plots). From Figure 1 it can be seen that there are five laboratories which overlap the shaded area. There is a number of laboratories which do well with milk(H), but have problems at lower levels. Perhaps unexpectedly, there are some laboratories which have the opposite problem (i.e. results for milk (L) are good but milk(H) is overestimated). Laboratories in the lower-left quadrant are showing a consistent negative bias, while those in the upper right quadrant have a consistent positive bias° Some results fall outside the edges of the plots and these are shown by arrows. Both results for number 8 fail outside, indicating consistently poor performance, but for numbers 4 and 7 only one result of each pair has a very large bias. The latter combinations can arise from occasional blunders. Points which fall in the upper-left or lower-right quadrants (such as number 15) show inconsistent performance which probably results t¥om more than one type of analytical problem. It is evident from Figure 1, that the range in performance is broad. Some laboratories have done quite well, while others have done very poorly. However, most of them would fall in the category of "not consistently good, but not overly poor". Results for 134Cs in milk powder are shown in Figure 2. In this case only one laboratory falls in the shaded area and there seems to be a more even distribution through the lower left, lower right and upper right quadrants. Laboratory 15 is again showing severely inconsistent performance. The results tbr 4°K, which are shown in Figure 3, are rather interesting, since the concentrations (552 and 539 Bq/kg) are almost the same in the 2 samples. Laboratories number 8 and 15 consistently overestimated 4°K with large biases. The results from laboratories 5 and 12 show good precision for each material and good results for milk(L) but they find significant differences between the two materials, although they should be

Analytical quality control

153

Cs-137 in M i l k P o w d e r

Cs-134 in M i l k P o w d e r

(16%)24001

(13%)

55Ov ......... r -~---,T .......%---

i ilS

!

sod 2200

i

__ .............................................. ~, 1o

,

~i

2ooo....................~ ~ ~, ~_~_ '~

I,

19001 rt' 18001

t~ ~

i i

450i

[,i

4001

-...~-

""-

-

Z~

.~

..... -'h---

0 ¢h

-~ !,6

350i

17ooi ~.

......

. . . . . . . . . . . ; ....

I

1600!

300~

15ooi 1400

2501 i

1300 t

i 200t i. i

(47%)I10~i0 55

60

65

70

75

80

85

)

90

(-69%)

(24%)

15

20

25

Lower Level Milk Powder - Bq/kg

(-33%) (61%) L o w e r L e v e l Milk P o w d e r - B q / k g

Figure I

Figure 2

K-40 in Milk Powder

Sr-90 in M i l k P o w d e r

(39%) 750]

(]69%)

20r

/ /

7O0

i t5i L

650 i I I

@

600]

a

o"

,20

i

t

J q !

; ............

~. . . . . . . . . . .

~J.

._.j

--

,oo ......f .......I V .... ~

!

i

. . . . . . . .

: p

,i'u

450 i *P (-26%)

40~)~ (-2a%)

~-

i

~'--~

!!,.

[ 500

600

700

Lower Level Milk Powder. Figure 3

800 (45%) Bq/kg

5

(doo%)

I0

(-100%)

15 (355%)

L o w e r Level M i l k P o w d e r - B q / k g Figure 4

154

Analytical quality control

virtually identical. Perhaps different analysts measured the two samples or different spectrometers were used° Figure 4 illustrates the results obtained for 9°Sr by radiochemical analysis. The tendency for most laboratories to overestimate 9°Sr in milk powder samples is clearly evident. Laboratories 2 and 3 grossly overestimated 9°Sr by factors ranging up to about nine times. Only one laboratory, number 6 obtained good results for both samples. Laboratory number 18 severely underestimated 9°Sr in both cases° The data for clover and soil provide the opportunity to compare performance with two dissimilar materials. The results for 137Cs in the 2 materials are shown in Figure 5. It is evident that more laboratories obtained good results for soil, inspite of the lower concentration, compared to clover° This may reflect more experience with the analysis of soils. There appears to be a greater tendency to underestimate ~37Cs in clover than in soil. Similar results for 9°St in clover and soil are shown in Figure 6, Only two results overlap with the shaded area and the performance seems to be better with analysis of clover° A significant fraction of the laboratories whose results are shown have overestimated 9°Sr in both materials and the biases range up to more than 100 percent. The data can also be used to compare performance in measuring different analytes. Both 137Cs and 134Cs were measured by gamma spectrometry and the results for clover are compared in Figure 7. In the analysis of L34Cs by this technique there is the potential for counting losses due to coincidence summing and this can lead to low results. The magnitude of this effect depends on the counting geometry and if it is significant a correction should be applied° Coincidence summing is not a problem in the analysis of 137Cs. The results shown in Figure 7 do not appear to indicate any differences in the tendencies to underestimate 137Cs and J34Cs, which probably means that coincidence summing was not significant for the clover samples° Coincidence summing can also be a problem in the analysis of 6°(2o and the effect on the results should be greater for air filters. The data shown in Figure 8 for 137Cs and 6°(2o in ah" filters seem to reflect this, since there are more low results for 6o('.0 and none of the results fall in the shaded area° The results for 137Cs in milk(H) are compared to those for 9°Sr, which was measured radiochemically, in Figure 9. It is evident that some laboratories are able to obtain better results by gamma spectrometry than by radiochemistry and only a few do well with both techniques. Plutonium-239 was also measured radiochemically and the results are compared to those for 9°St in soil in Figure 10. Here again, some laboratories show good performance in one type of analysis and poor performance in the other~ The number of overestimated results exceeds the number of underestimates and only one laboratory did well in both analyses.

Analytical quality control

155

Sr-90 in Clover & Soil

Cs-137

in Clover & Soil ,14%)

300!

(69%)

25[-. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I

J [

1

:oi ,

260

t.

i tqa O

..= t.~

!--6

_~._

240

o

..=

[

~ .... 7~

i~'

I

15

220

I

I

--r-

1

(-: ~ )

2o~i)

(-32%)

45

30

55

5O

(.~s~)

70

50

(167%)

(-62%)

(18%)

C s - 1 3 7 in Soil - B q / k g

S r - 9 0 in Soil - B q / k g

Figure 5

Figure 6

Cs-137 & Cs-134 in Clover

in Simulated Air Filters

Cs-137

& Co-60

(~7~) 14001

(~ 4~t) 300?

P

1300 t Ii

~10 ~e i

'

t_

260 ..........

~i

s

[

7

11oo

•-~

i

--~

i ~"~I000

J

F

F

!..........................................

] I I

.= !~ 240

1 o0i

--~r-~

220 900

(.24~) 20( 50 (.77~t)

70

90

110

I30

150

170 (29%)

(.27~) 80~,

;0

1050

1150

1250

1350

.450

(-3~%)

Cs-134 in Clover - Bq/kg

C o - 6 0 in A i r F i l t e r s - B q / f i l t e r

Figure 7

Figure 8

1550

156

Analytical quality control

Sr-90 & Pu-239 in Soil

Cs-137 & Sr-90 in Clover (52%1

(14~) 3001

40r

.......

7--

............

7 . . . . . . . . . . . . . . . . .

:i

280! i

.

.

c t.

.

.

.

/ .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

3s!

!i

30~

~:

?

!

O

!

--y-

~ol

25t 220 s

I

i 15

!. . . . . . . . . . . . 20 ...

(-32%)

,~.~ ~25 (69%)

(.z4%) 20~. . . . . .

2

3

(-100%)

4

5' (381%)

Sr-90 in Clover-Bq/kg

Pu-239 in Soil - Bq/kg

Figure 9

Figure 10

The figures provide a means to assess the performance of individual laboratories in analyzing individual materials for different radionuclides and they give some idea of general performance. However, the objective of the intercalibration exercise was to assess the overall performance and the effect that it has on the quality of the analytical results. This is not normally required for intercomparisons organized by AQCSo Consequently, different statistical techniques had to be explored in order to find ones that would give a meaningful assessment of performance and data quality. The individual means were rated in order to determine whether there were significant differences in the tendencies to overestimate and underestimate 137Cs, 13'lOs, 4°K, 9°Sr and 239puo Each mean v a l u e was rated "0" if it (including the error bars) overlapped the confidence interval for the reference value° Results that were biased high were rated "+", while low results were rated "-". The overall results are shown in Table 2. Just over half of the results for 137Cs were rated 0 (no bias), while less than half of the results were rated 0 for the rest of the radionuclides in Table 2. In order to have no overall bias the number of + 's would have to be the same as the number of -'s for each radionuclide. A chi-squared test was used to compare the frequencies of + and - biases. The results are also shown in Table 2. Significant (P=0.05) differences were found for ~34Cs, 4°K and 9°St, while the chi-squared value for 137Cs was just under the critical value for P=0.05. There were too few results for 239pu for the chi-squared test to be valid° The results in Table 2 indicate that the tendency to underestimate results by gamma spectrometry is

Analytical quality control

157

Table 2: Frequencies of Biases in Individual Means Radionuclide

Number of Means 0 +

Significant +/- bias*

)37Cs

26

42

13

no

134Cs

19

24

7

yes

4°K

20

25

7

yes

9°Sr

8

19

21

yes

239pu

]

3

4

---

chi-squared test was used to compare the frequencies of + and- results at the significance level P--0.05

* A

greater than the tendency to overestimate. The trend in the radiochemical results for 9°St is in the opposite direction. This also may be the case for z39pu, but this cannot be tested. In order to assess the overall accuracy of the results, means and standard deviations were calculated using all the results for each radionuclide in each material. The percentage differences (biases) between the overall means and the reference values were :alculated and a t-test was used to see if the biases were significant at P=0.05 (i.e. whether :he means differ from the reference values). The results are given in Table 3. Biases ;hown in brackets were not found to be significant by the t-test. Although some of the ~iases are quite large, they were not found to be significant because of the insensitivity of •he t-test. The insensitivity arises from the presence of a small number of heavily biased ,)utliers, which artificially inflate the estimated standard deviatiom as well as introducing a bias into the mean. In order to improve the sensitivity, the outliers have to be identified ~md removed° In the normal intercomparisons organized by AQCS, outlier tests are applied to the data. However, these tests themselves suffer from insensitivity because of the biased means and inflated standard deviations. Because of this, it was decided to ~eject biased results based on the reference values. Specifically, results were accepted if t ley (plus their error bars) intercepted the reference values _+ 33%. This was a somewhat ~rbitrary choice, but it was quite fair to the participants who really ought to be able to do Letter than _+ 33% (on top of their own random uncertainties). Furthermore, results that .ore off by ___33% will not have an undue influence on the estimated means and standard (zeviations unless there are many of them. At any rate, most of the rejected results differed f'om the reference value by much more than ___33%. After censoring the data new means and percentage differences from the reference values were calculated. A second round of t-tests was used to determine whether the censored means differed significantly (P=0.05) flora the reference values. In cases where the mean did differ significantly from the r~:ference value, another t-test was applied to determine whether it differed significantly fiom the confidence interval of the reference value. The results are given in Table 3 along with the percentage of results censored. The biases of the censored means are shown in brackets where the t-test did not indicate a significant difference from the reference value.

158

Analytical quality control

Table 3: Biases of Overall and Censored Means, Radionuclide:

Biases Overall Censored Mean: Mean:

Percent Censored:

Outside Confidence Interval:

9 0

no o-

9.5 67

Yes o~

Milk (L) 137Cs 134Cs

(+15.1) (+ 5.6)

-6.2 --

4°K 9OSr

(+ 13.6) +134

-5.9 (+17.4)

(-4.6) -13.3

(-1.0) -6.4

15 13

-no

(50.7) (+ 135)

(-5.3) (+7.2)

14 33

---

137Cs

(+ 2.2)

(+ 1.1)

10

-°

9°Sr 239ptl

(+26) (+60.6)

(+ 3.8) (+12.5)

50 25

---

~37Cs

(-3.1)

--

0

--

6°Co 9°Sr

(-10.5) (- 2.4)

(-5.8) --

17 0

---

Milk (H) 13"7Cs 134Cs 4°K 9oSr Soil

Air Filters

Clover 137Cs

(- 4.4)

-9.1

8

no

134Cs

(-12.3)

(-7.3)

9

--

4OK

(+42.3)

-6.8

9

--

9°Sr

(+12.2)

--

0

--

Notes: *based on t-test with P=0.5. Biases in brackets were not found to be significant (P=0~05) by the t-test. Confidence Intervals: usually 95% - please see original reports.

E x c e p t for 137Cs in clover, r e m o v a l of the outliers g a v e a s i g n i f i c a n t i m p r o v e m e n t in the m e a n values. F o r 137Cs, 134Cs, 6°C0 and 4°K the p e r c e n t a g e s of outlying results w h i c h had to be rejected tell within the r a n g e o f the p e r c e n t a g e s rejected in the i n t e r c o m p a r i s o n s originally used to certify the materials. H o w e v e r , a surprisingly large p r o p o r t i o n of the results for 9°Sr in m i l k ( L + H ) and soil and 239pu in soil had to be rejected in order to l o w e r the large positive biases in the means. T h i s r e i n f o r c e s earlier o b s e r v a t i o n s o f a t e n d e n c y to o v e r e s t i m a t e 90Sr and p r o b a b l y 239pu as well. Part of the p r o b l e m m a y be due to the low levels o f 9°St in these s a m p l e s , particularly the m i l k samples. T h e c o n c e n t r a t i o n s in both m i l k s a m p l e s are m o r e than an order o f m a g n i t u d e l o w e r than the limits f o r f o o d

Analytical quality control

159

given in the FAO/WHO Codex Alimentarius (FA0/WHO 1989). Results for 9°Sr in air filters and clover were much better and no values were rejected. Only five of the censored means were significantly different from the reference values and all of these, except 4°K in milk (L), were within the % confidence intervals. All of the significant biases for results obtained by gamma spectrometry were negative; however, all of the biases in the censored means were less than 10 percent. No significant biases were found in the radiochemical results, although most of censored means were larger than the reference values and the differences tended to be greater.

Table 4: Results of Homogeneity Tests. Material No:

1 2 3 4 5 6 7 8 9

Material:

Standard Deviation for 8 subsamples:

soil grass milk powder grass milk powder hay soil soil milk powder

1.4% 0.8% 1.8% 1.7% 1.3% 6.4% (6.0%, 5.5%) 1.3% 1.4% 2.3%

In the fall of 1990 nine materials (4.5 tonnes in total) were sent to the Agency's l,aboratories at Seibersdorf as part of Phase 2 of the intercalibration. These materials ~.¢ere tested for homogeneity by measuring 137Cs in 8 subsamples of each one. The results are given in Table 4. Most of the materials are quite homogeneous (standard deviations ,:2°3%). However, the hay (number 6) seemed to show an estimated standard deviation of about 6% and this was verified by repeated tests.

Conclusions The results of the intercalibration exercise carried out in Phase 1 indicate that, on a cerage, biases in results obtained by gamma-ray spectrometry tend to be fairly small, e cen when they are significant. Biases in results from some individual laboratories are qJite high, but these laboratories constitute a relatively small percentage of the total n ~mber participating. This suggests that analytical results for 137Cs and 134Cs in food and environmental samples should be reasonably good (bias < 20%) on average, with a small fraction of individual outlying results. From the number of outlying results it is apparent that a significant proportion of the laboratories have problems with radiochemical analysis of 9°St, at least at lower levels. There is a greater tendency to overestimate 9°St, although it is difficult to show that the

Analytical quality control

160

overall means have a statistically significant bias. The fact that most of the outlying results are high suggests that the overall means should be biased on the high side. This in turn suggests that analytical results for 9°Sr in food and environmental samples may be overestimated. There could be an overall positive bias in the 239pu results, but there are too few results to show it. Finally, the results from homogeneity testing show that at least 8 of the 9 materials prepared for Phase 2 are suitable for intercomparison and eventual certification as reference materials.

Acknowledgements The co-operation of the laboratories which participated in the intercalibration exercise is ~atefuUy acknowledged. The authors would like to thank Mr. D. Bridi for preparing the plots and Ms. Y.W. Huang for typing the manuscript. One of the authors (E.L. Cooper) was partly supported by AECL-Research, Chalk River, Canada.

References Cooper E.L., LaBrecqueJ.J., DeknerR., ReichelF. and SchelenzR. (1988a) Intercomparison Study IAEA-152 on the Determination of Elevated Levels of Fallout Radionuclides in Milk Powder. IAEA report no. IAEA/AL/009 International Atomic Energy Agency, Agency's Laboratories, Analytical Quality Control Services~ Vienna, 1988. Cooper E.L., LaBrequeJ.J., Hanna A.H., DeknerR., Scott T., StrachnovV. and SchelenzR. (1988) Intercomparison Study IAEA/AIR-4(083) on the Determination of Radionucldies of Simulated Air Filters. IAEA report no. IAEA/AL/006 International Atomic Energy Agency, Agency's Laboratories, Analytical Quality control Services, Vienna, November 1988. FAO/WHO (1989) Supplement 1 to CODEX ALIMENTARIUS VOLUME XVII, Food and Agricultural Organization and World Health Organization, Rome. PszonickiL., Hanna A.N. and SuschnyO. (1984) Report on Intercomparison IAEA/SOIL-6 of the Determination of Cs-137, Pu-239, Ra-226 and Sr-90 in soil. IAEA report no. 1AEA/RL/111 International Atomic Energy Agency, Vienna, April 1984. Strachnov V., Burns K., and Dekner R., (1990) Intercomparison Study IAEA-321 on the Determination of Radionuclides in Milk Powder. IAEA report no. IAEA/AL/026 International Atomic Energy Agency, Agency's Laboratories, Analytical Quality Control Services, Vienna, January 1990. Strachnov V., Valkovic V. and Dekner R. (1990) Intercomparison Study IAEA-156 on the Determination of Radionuclides in Clover. (1991) IAEA report no. IAEA/AL/035 International Atomic Energy Agency, Agency's Laboratories, Analytical Quality Control Services, Vienna, January 1990. Youden W.J. and Steiner E.H. (1975) Statistical Manual of the Association of Official Analytical Chemists, Association of Official Analytical Chemists, Arlington, VA.