2.5.6 Controlling Gy Errors
To correctly collect samples as defined by Pitard (1993), all these errors should be addressed. Table 2-2 provides a summary of the various errors described by Gy together with measures that might be taken to control each.
In practice, the focus is usually on FE and GSE; however, the other errors can be important if correct sampling procedures are not used. As illustrated above, the FE can be minimized by collecting a sufficient mass of sample, and the GSE can be minimized by collecting numerous increments.
|Factor leading to error||Sampling error||Error results from||How to control|
|Compositional heterogeneity (CH)||Fundamental error (FE)||Size and compositional distribution of the particles||Increase the sample mass and/or reduce the size of the particles|
|Distributional heterogeneity (DH)||Grouping and segregation error (GSE)||Heterogeneous distribution of particles within the population||Increase the mass of the sample or increase the number of increments|
|Large-scale heterogeneity||Long-range heterogeneity fluctuation error (CE2)||Changes in concentration across space or over time||Reduce the spatial interval between samples|
|Periodic heterogeneity||Periodic heterogeneity fluctuation error (CE3)||Periodic changes in concentration over time||Change the spatial and/or temporal interval between samples|
|Identifying the correct increment geometry||Increment delimitation error (DE)||Incorrect shape (in all three dimensions) of the sample or increment selected for extraction from the population||Use correct sampling plan design and correct sampling equipment that can sample the entire thickness of the population|
|Shape of the sample extraction device and nature of the soil||Increment extraction error (EE)||Incorrect extraction of the sample or increment because the sampling device is too small||Use correct sampling equipment that does not push larger particles aside, and use correct sampling protocols|
|Loss or gain of contaminants during sample handling||Preparation error (PE)||Contamination loss or gain due to alteration, evaporation, degradation, cross-contamination, mistake, or fraud||Use appropriate sample handling, preservation, transport, and preparation measures|
The mass of a sample necessary to minimize FE is primarily related to the largest particle size of the population being sampled. Hyperlink 14 provides more information concerning the calculations for determining sample mass to minimize the FE.
The number of increments needed in the field depends on a number of factors, including heterogeneity within the DU, the difference between the mean concentration and the level of interest (e.g., action level), and project DQOs. It is theoretically possible to determine the number of increments necessary. For example, at a large site or a site with many DUs, a pilot study could be conducted on a portion of the site to provide initial estimates of heterogeneity and mean concentration. That data could then be used to determine the number of increments needed to manage decision error sufficiently. However, this process is often not practical due to cost and time constraints. Then, how many increments are sufficient?
When no prior data are available to estimate heterogeneity within a DU, a default range of 30–50 increments is recommended. One approach is to use a sufficiently conservative default number of increments—one that is high enough to result in a representative sample for the majority of cases even when the DU is heterogeneous. Based on simulation studies discussed in Section 4 and empirical evidence gathered from using ISM at a variety of sites, a default range of 30–50 increments is adequate for most sites. However, as many as 100 increments may be necessary for larger DUs where the CSM indicates that high heterogeneity is anticipated. One indication of how well the increment density is capturing the heterogeneity within the DU is variation between ISM replicates. If all other sources of error are held constant, the degree to which the number of increments collected are capable of capturing the heterogeneity present in the DU is reflected in how well replicate ISM sample results agree. One should use caution, however, when interpreting results between ISM replicates since this measure of variability integrates all of the sampling errors described above.