Measuring the error caused by within-sample heterogeneity

The amount of error caused by within-sample heterogeneity can be measured using replicate subsamples in the field and/or in the laboratory. When each of the subsamples is analyzed, the difference between their respective results is calculated as indicated in Figure 2-8.

Figure 2-8. Variability among results of laboratory subsample duplicates measures within-sample heterogeneity.

A large difference between results indicates that within-sample heterogeneity is present and is causing sampling error. Field splits and laboratory duplicates for soils are common QC checks that often fail to meet QC acceptance criteria. Unfortunately, nothing is typically done to correct the problem(s) indicated by the failed QC. The data may be qualified as estimated, but in practice they are simply used "as is." Laboratory duplicate results should not be ignored, for they provide very important information about the quality of sample handling and the magnitude of sampling error.

Variability between duplicate subsample results measures heterogeneity within the sample jar.

Duplicate results may vary so widely that a different decision about "clean" or "dirty" may be indicated, depending on which result is used. The question is often asked, "Which result is right?" The answer is that they are probably both right and both wrong. Both are right in the sense that the analysis of both subsamples was probably correct unless other QC samples indicate otherwise. It is just that the laboratory subsamples are fundamentally different. Both may be wrong in the sense that neither result adequately represents the true concentration for the jar of soil, and by extrapolation, for the concentration in the DU. Highly variable field and/or laboratory duplicates should be an indication to decision makers that the data generation process is excessively imprecise and could lead to decision errors. Hyperlink 13 provides a discussion of approaches for dealing with within-sample heterogeneity.