E STIMATION OF U NCERTAINTY OF S AMPLING As it was shown, the average residues and CV of residues in individual crop units in
5.2.3 E STIMATION OF U NCERTAINTY OF S AMPLING As it was shown, the average residues and CV of residues in individual crop units in
samples of size 100 –120 (that is each sample consists of 100–120 individual crop units, e.g., oranges) taken repeatedly from the same parent population (e.g., from a field or a lot) may vary significantly. The best estimate of the uncertainty of sampling is provided by the average of CV values. 24
The sampling uncertainty depends on the size of composite samples and the distribution of residues in the sampled commodity. Based on 174 residue data sets consisted of 22,665 valid residue data derived from specifically designed super-
vised trials 25 and sampling lots from the market, 26 the estimated typical sampling uncertainty for different kinds of plant commodities and sample sizes specified by 22 the Codex Standard for sampling are summarized in Table 5.2.
There were no data for estimation of the uncertainty of sampling cereal grains, eggs, and processed products. The variation of residues in composite samples taken from different fields is much larger. The typical CV values of between fields variation of residues in composite samples ranged between 80% and 120%. The data evaluation revealed that the coefficients of variation of residues within field and between fields are 27 practically independent of the pesticide, the preharvest interval, and dosage rate.
TABLE 5.2 Typical Sampling Uncertainty for Various Fresh Plant Commodities with Lower (LC) and Upper (UC) Confidence Intervals
Confidence Limits of CV Styp Commodity
No. of
SD of
Sample Groups
a CV S
Sample
Sets Values
Size
LC 0.99 LC 0.95 CV Styp UC 0.95 UC 0.99
Small 18 0.31 1 0.57 0.62 0.78 0.93 0.99 commodities
0.10 10 b 0.18 0.20 0.25 0.29 0.31 (unit mass 25 g) Medium-size
76 0.25 1 0.73 0.75 0.81 0.86 0.88 commodities
0.08 10 b 0.23 0.24 0.25 0.27 0.28 (unit mass 25 –250 g) Large
64 0.27 1 0.58 0.60 0.67 0.73 0.76 commodities
0.12 5 b 0.26 0.27 0.30 0.33 0.34 (unit mass
0.09 10 0.18 0.19 0.21 0.23 0.24 >250 g) Brassica leafy
17 0.16 1 0.33 0.37 0.45 0.53 0.56 vegetables
0.07 5 b 0.15 0.16 0.20 0.24 0.25 0.05 10 0.11 0.12 0.14 0.17 0.18
a Primary samples making up the sample sets were taken from different fields=lots. b Minimum number of primary samples to be taken for a composite sample for testing compliance with Codex MRLs. 25
136 Analysis of Pesticides in Food and Environmental Samples
Fully nested experiment Staggered nested experiment 1
4 FIGURE 5.1 Experimental designs for estimation of the uncertainty of sampling steps.
1: Lots; 2: sampling of lots; 3: subsampling; and 4: measurements.
The ISO Standard 11648-1 for sampling bulk materials 28 recommends to apply fully nested or staggered nested experimental design to obtain information on the uncertainty of withdrawing the bulk samples from different lots, reducing the sample size with subsampling (sample preparation) and analysis. The procedures are illus- trated in Figure 5.1. The standard recommends that for obtaining sufficient informa- tion about the variability of the analyte, ~20 lots should be sampled, preferably several pairs of samples taken from each lot.
Sampling of the same residue data population by withdrawing random composite samples with replacement 24 or applying the fully nested experimental design gave very similar results for the average residue and the average CV of the residues. For instance, even if 30 pairs of random composite samples of size 10 were withdrawn 100 times from a data population having a CV of 0.28, the minimum and maximum CV values observed were 0.205 and 0.365, respectively, which is in agreement with the confidence limits shown in Table 5.2.
Concerning the sampling uncertainty, one should always remember that the MRLs refer to the residues in the bulk sample. Hence, for testing compliance with an MRL any amount of material satisfying the minimum sample size is sufficient and the sampling uncertainty need not be taken into account. On the other hand, where the compliance of a lot before shipment has to be verified, then the sampling uncertainty must be included in the combined expanded uncertainty of the measured residue value.