2.6.1 Discrete Sampling
Discrete, or grab, soil sampling has a long history of use within the environmental industry. Analytical results from a number of discrete samples collected from a site are typically used to make environmental decisions regarding the site. For example, they may be used to provide an estimate of the mean in some meaningfully sized volume of soil.
A number of factors influence the ability of such a discrete sampling plan to provide an unbiased estimate of mean concentration. The primary factor is the number of discrete samples collected, but sample location, collection method, sample support, and lab handling are also important. For discussion purposes, two types of discrete sampling plans of different sample numbers are identified below: high and low density. Of course, this is a gross oversimplification, but it is useful here for the purposes of highlighting some important concepts.
When the true mean is well above or below an action level, even a small number of discrete samples usually results in a correct decision. At face value, low numbers of discrete samples are tempting in terms of cost, ease of implementation, and simplicity. However, simulation studies, empirical evidence, and sampling theory suggest that low numbers of discrete samples do not produce very accurate or precise estimates of the mean because such an approach does not account for heterogeneity. When only costs are considered, discrete sampling plans have historically been preferred. However, comparisons between different sampling approaches must be evaluated not only in terms of their costs but also in terms of the total error and resulting decision quality. The two types of discrete sampling plans discussed below result in dramatically different costs, but they also result in dramatically different decision qualities. Collecting the number of discrete samples sufficient to make a defensible decision at a site may at times be precluded by cost considerations.