Environment International, Vol. 14, pp. 271-275, 1988 0160-4120/88 $3.00 + .00 Printed in the USA. All rights reserved. Copyright © 1989 Pergamon Press plc

SOURCES OF VARIATION IN ENVIRONMENTAL RADIOCHEMICAL ANALYSIS

E R. Livens and C. Quarmby Institute of Terrestrial Ecology, Merlewood Research Station, Grange-over-Sands, Cumbria, LA11, 6JU, Great Britain

(Received 4 April 1988; Accepted 3 August 1988)

Both collection and laboratory treatment of samples contribute to the uncertainty in the final analytical results. A detailed study of a pasture field showed that, for realistic sample numbers, field variation was in the range 20-40%. Routine analysis of a reference soil, coupled with the use of some simple criteria for the acceptance of analytical results, showed that laboratory variation could be maintained at well below 10%, generally much less than the field variation.

Introduction

Every step in a sampling and analytical programme is subject to a degree of uncertainty and contributes to the error in an analytical result. Without an assessment of the magnitude of the errors introduced in this way, it is impossible to determine the true significance of a set of analytical results, and hence draw correct con- clusions from them. Thus, a consideration of the errors arising in the sampling and analytical processes and their magnitude is an essential component of any pro- gramme. A very informative treatment of errors in an- alytical chemistry is given by Hunt and Wilson (1986).

Analytical results can be characterised by their 'ac- curacy' (i.e., their closeness to the true value) and their 'precision' (i.e., their reproducibility). Both accuracy and precision are affected by experimental uncertain- ties.

Two types of error can usefully be identified: random errors, arising from uncontrolled and upredictable vari- ations in experimental conditions which influence the end result; and systematic errors, which manifest them- selves as a consistent, and often reproducible tendency for the analytical results to differ from the true value. If random error is dominant, the results will be accurate, but imprecise; thus, the mean of a large number of determinations will be correct, but there will be vari- ation between individual results. On the other hand, systematic error gives results which are inaccurate but precise, so that individual determinations are very sim- ilar, but all equally biased. Both types of error can arise in an environmental measurement programme before

271

samples ever arrive in the laboratory. Frequently, the analyst has no control over the sampling technique used in the field, yet the uncertainties introduced at this stage will frequently dwarf the errors in subsequent analysis (Parkinson & Horrill, 1983).

Uncertainties arising from sampling These can be well illustrated by considering a par-

ticular example--the determination of gross levels of radionuclides in a single pasture field of 3.7 hectares area. This particular field had, at the time of sampling, received inputs of artificial radionuclides from three sources: (a) weapons test fallout, (b) marine discharges of low level waste returned to land via sea spray and (c) aerial discharges of radionuclides from nearby Sel- lafield. The objectives of the programme are, at first sight, quite simple--representative samples of soil and vegetation need to be collected and analysed.

Consider first the sampling of vegetation. The plants in this particular field were a uniform mixture of grassy species, so there was no need to take into account the possible effects of differing vegetation structures and plant growth strategies. These are known to be impor- tant in areas where there are diverse mixtures of species (Horrill, 1983). A common technique used for sampling vegetation is the collection of the standing crop from quadrats, either randomly sited or in a regular sampling grid. This can provide an adequately representative sample of the vegetation within an area. However, this may not be appropriate in some circumstances; for ex- ample, if the intake of grazing animals is to assessed.

272 E R. Livens and C. Quarmby

Table 1. Radionuclide concentrat ions in vegetation and soil from 10 sample points (Bq kg-~).

B2 B5 B8 B l l E2 E5 E8 E l l H5 H14 ~ ± S D

~3vCs V 89 56 67 59 115 70 74 67 122 78 80 ± 22 (27%) T 148 152 167 148 244 122 133 252 174 163 170 ± 44 (26%) B 41 59 78 44 41 52 48 85 56 81 58 ± 17 (29%)

239pu + 24Opu

2,~Am

V 7.0 7.0 14.1 8.1 15.9 7.4 12.9 11.1 10.4 9.6 10.4 -+ 3.1 (30%) T 11.5 10.2 11.3 13.5 21.6 13.5 l l .1 18.5 11.1 10.4 13.3 -+ 3.8 (29%) B 7.4 6.7 8.1 6.3 2.3 5.6 4.1 10.7 6.7 4.8 6.3 -+ 2.3 (36%) V 2.9 2.6 2.8 2.2 4.6 2.2 3.1 2.9 3.0 2.2 2.8 -+ 0.7 (25%) T 1.5 1.8 2.6 2.4 2.5 2.4 1.5 2.8 1.5 1.5 2.0 -+ 0.5 (25%) B 0.8 0.7 1.4 0.9 I).3 0.8 1.4 1.3 0.9 0.8 0.9 + 0.3 (35%)

V = Vegetation. T = Top 5 cm of soil core. B = Lower 5 cm of soil core.

Since some animals are very selective, a different pro- cedure, which mimics grazing behaviour, is likely to be more suitable in this situation (Howard, 1985).

At each of ten sampling points on the 60 m grid, vegetation and soil samples were collected. The vege- tation from a 3 m 2 area was removed and four soil cores (5 cm diameter) collected from within the area. The soils were sectioned at 5 cm depth to separate top and bottom sections and bulked. All the samples were an- alysed for Cs-137 by gamma ray spectrometry and by chemical separation and alpha spectrometry for amer- icium and alpha emitting plutonium isotopes. The re- sults are presented in Table 1.

In all cases, the overall precision, representing the sum of sampling and analytical errors, lies between 25 and 36%; it is in the range 35-50% if the results are calculated in terms of activity per unit area. This low precision is the sum total of all the sources of uncer- tainty in the sampling and measurement procedures and has disturbing implications for the numbers of samples which would need to be collected to ensure a given degree of precision. This has been calculated as part of a detailed statistical analysis of the data relating to this site (Quarmby, 1983) and are shown in Table 2. It is apparent that the sampling intensity required to achieve

Table 2. Sample numbers required

a high degree of precision is generally quite prohibitive. Furthermore, the sample numbers which are often, of necessity, collected in the practical field situation are so small that there will be considerable uncertainties, often of the order of 30-50%, in the results. Given such uncertainty, it can become very difficult to draw firm conclusions from field data, particularly if small changes are to be detected. Clearly, where large uncertainties in data arise it is important to make some attempt to assess the magnitude of the different contributions and, if possible, reduce them. To this end, a study of the precision of the analytical methodology employed is very useful.

Uncertainties arising during analysis During routine analysis for the actinides, 'in-house'

reference materials have been routinely used over the last three years and an extensive data set has gradually been built up. The results for a soil reference material are presented in Table 3 and illustrated in Fig. 1. These reference analyses can be used in two ways. First, they provide an estimate of the overall precision of the meth- ods used, which can be calculated from the entire data set; this is 10.1% for 239.24°pu (Table 4). Second, they can be used to estimate the precision of the laboratory's

for a given degree of precision.

is7Cs 239,241)pu 24tAm

Precision required V T B V T B V T B

50% 1.3 1.1 1.4 2.1 20% 8.0 6.6 9.0 13 10% 32 26 34 53

5% 130 110 140 2111 2% 800 660 900 1300 1% 3200 2600 3400 5300

1.5 2.2 1.3 1.1 2.1 9.6 14 7.8 6.6 13

39 55 31 26 54 150 220 130 110 210 960 1400 780 660 1300

3900 5500 3100 2600 5400

V = Vegetation. T = Top 5 cm of soil core. B = Lower 5 cm of soil core.

Variation in environmental radiochemical analysis

Table 3. Analyses of reference soil (Bq kg J).

Analysis 239,2a0pu 23~pu

1 50.7 12.2 2 49.3 11.1 3 44.7 11.3 4 48.8 10.4 5 37.7" 8.8 6 44.1 10.0 7 43.1 9.7 8 50.6 11.5 9 44.4 9.9

10 50.2 11.1 11 55.2" 10.7 12 46.3 10.1 13 46.3 10.0 14 43.2 9.5 15 43.7 8.7 16 43.5 10.2 17 40.6 8.7 18 44.3 8.8 19 47.2 10.8 20 45.1 10.0 21 43.5 10.4 22 45.3 10.2 23 43.4 9.7 24 47.8 11.0 25 45.6 10.5 26 45.3 10.0 27 43.6 9.7 28 41.4 9.0 29 41.3 10.5 30 42.1 8.5 31 48.1 10.9 32 49.5 14.8 ~ 33 49.9 11.5 34 48.4 11.2 35 40.7 9.5 36 44.2 8.9 37 41.1 9.7 38 57.6" 11.3 39 42.6 9.2 40 40.8 8.8 41 41.9 9.1 42 30.2" 6.2 43 37.9 8.2 44 42.6 8.8 45 41.8 9.1 46 42.5 9.1 47 43.4 9.2 48 42.0 8.9 49 37.4" 9.0 50 42.0 8.9 51 42.9 9.2 52 43.1 9.6

Method A--Results 1-10. Method B--Results 11-34. Method C--Results 35-52. "--Result outside limits of acceptability. ~ R e c o v e r y below 50%.

273

Table 4. Mean values and precision of reference analyses (Bq kg ~).

Coefficient Coefficient of of

23'~24°pu Variation (%) -'3~pu Variation (%)

All Data: Method A 46.4 7.5 10.6 9.6 Method B 45.4 7.3 10.2 12.7 Method C 42.0 11.9 9.1 9.9

Over All Methods 44.4 10.1 9.9 13.1 Accepted Data:

Method A 47.3 6.8 10.6 9.4 Method B 45.0 5.8 10.0 8.0 Method C 42.0 3.6 9.1 3.3

Over All Methods 44.6 6.7 9.8 9.2

routine analyses by applying, to the reference results, the same criteria for acceptance which are applied to ordinary samples. These criteria are that the chemical recovery should be greater than 50% and that any ref- erence analyses should lie within two standard devia- tions of the accepted value. Any batch in which a ref- erence analysis lies outside these limits is rejected. Thus, reference analyses outside the limits should not be used in calculating the precision of routinely accepted anal- yses. The precision of these acceptable analyses is 6.7% for 239-24°pu (Table 4). Furthermore, the data have been generated using the laboratory's routine methods of sample preparation and analysis so that the reference material mirrors as closely as possible our treatment of 'real' samples. Altogether, three different analytical techniques (summarized in Table 5) and numerous dif- ferent batches of both 236pu and 242pu yield monitors have been used. The 'acceptable' reference analyses will therefore be a very fair representation of the precision of the laboratory's output. It is obvious from Table 4 that the use of simple, easily applied criteria, such as those outlined above, can greatly improve the precision of routine analyses. Since the reference materials have been checked by inter-laboratory comparison (see Ta- ble 6), these analyses also allow a continuous check of accuracy to be made.

A comparison of the results from the three different methods shows that, whilst methods A and B are in very good agreement, method C gives rather low re-

Table 5. Analytical techniques employed.

Method Procedure

A Ash, HF/HNO3 digestion, HC1 dissolution, solvent extraction of Fe, 2 × anion enchange stages, elec- trodeposition.

B Ash, HF/HNO3 digestion, HCI dissolution, rare earth fluoride coprecipitation, 1 × anion ex- change step, electrodeposition.

C Ash, HNO3/HC1 digestion, solvent extraction of Fe, 1 × anion exchange step, electro-deposition.

25

20

>. O t- O ;7 (7" O L

EL

15

I0

5

0 I 30 33 3(5 39

274 E R. Livens and C. Quarmby

I 1 1 42 45 48

I 51 54 57 60

P I u b o n t u r n - 2 3 9 , 2 4 . 0 I B q / k g J

Fig. 1. His togram of soil reference analyses for 239, 2~)pu"

suits. This bias is 8% and probably arises from the omission of HF in the extraction stage of the method. This is known to sometimes result in poorer extraction of artificial actinides from environmental materials de- pending on their chemical form in the sample (Meadows et al. , 1975).

Table 6. Intercomparison results from soil reference (Bq kg t _+ I o-). tr denotes s tandard deviation of

individual determinat ions.

239.240pu 238pu

Mean of 3 laboratories 47.7 -+ 6.1 l l .1 - 2.0 This study a 46.4 +- 3.5 10.6 -+ 1.0

~Mean of s.d. of 10 determinat ions by method A.

Conclusions

From the discussion above it is clear that both field procedures and subsequent laboratory treatment of samples can introduce substantial uncertainties into en- vironmental radiochemical analyses. Nevertheless, ap- propriate quality control procedures can reduce the var- iability arising from the analytical procedures used to well below 10% and maintain it at this level over a considerable period. In the overall uncertainty in a mea- surement programme, that is, the resultant of sampling and analytical errors, the analytical contribution is gen- erally much smaller than that from sampling uncertain- ties, even in a reasonably uniform and easily sampled field situation in which sampling problems might be expected to be minimal. The experimental design should

Variation in environmental radiochemical analysis 275

ensure that sampling intensity is sufficient to detect the changes or contrasts sought with a reasonable degree of confidence. In addition, although analytical uncer- tainty is easily assessed and is often used as a measure of total uncertainty, it is unlikely to be the dominant source of error and its use may well therefore lead to an optimistic estimate of the precision of a data set.

References

Horrill, A. D. (1983) Concentrations and spatial distribution of ra- dioactivity in an ungrazed saltmarsh, in Ecological Aspects of Radionuclide Release, P. J. Coughtrey, J. N. B. Bell, and T. M. Roberts, eds., pp. 199-215. British Ecological Soc. Special Pub. No. 3, Blackwell, Oxford.

Howard, B. J. (1985) Aspects of the uptake of radionuclides by sheep grazing on an estuarine saltmarsh. I. The influence of grazing behaviour and environmental variability on daily intake, J. En- viron. Radioactivity 2, 183-198.

Hunt, D. T. E. & Wilson, A. L. (1986) The Chemical Analysis of Water, 2nd ed., 127-299. Royal Society of Chemistry, London.

Meadows, J. W. T., Schweiger, J. S., Mendoza, B., and Stone, R. (1975) Procedure for plutonium analysis of large soil and sediment samples, in Reference Methods for Marine Radioactivity Studies 11. pp. 89-90. IAEA, Vienna.

Parkinson, J. A. & Horrill, A. D. (1983) An assessment of variation due to laboratory and field conditions in the measurement of radionuclides. Nucl. Inst. Meth. Phys. Res. 223, 598-601.

Quarmby, C. (1983). Variability in the distribution of alpha and gamma emitters in a typical pasture field in the vicinity of Sellafield works, West Cumbria. Institute of Terrestrial Ecology report to MAFE Institute of Terrestrial Ecology, Grange-over-Sands, England.