188.8.131.52 Interpreting results of discrete sampling
Action levels are usually derived from risk assessment models that are based on average exposures over time. Use of mean soil concentrations to estimate exposure within a given area of contamination assumes that (a) the estimated mean soil concentration represents the true mean concentration in the exposure area, (b) the receptor is equally likely to be exposed to the soil at any location in the exposure area, and (c) soil concentrations will not change significantly over time. Based on these assumptions, risk assessments and risk management decisions often focus on estimates of the mean soil concentration in each exposure area.
Concentration data obtained from discrete soil samples typically fit frequency distributions that are skewed to the right (i.e., lognormal, gamma, and some nonparametric distributions). Figure 2-15 provides a graphical display of a normal Distribution (A) and a right-skewed distribution. Notice that the "long tail" extending to the right in Distribution (B) reflects the higher concentration results that occur at lower frequencies.
Discrete sample data tend to be clustered around the most frequently observed concentration, which is called the "mode." Because Distribution B in Figure 2-15 is skewed to the right, the mode is less than the mean concentration of the distribution. The tail of such distributions can easily contain concentrations one to two orders of magnitude greater than the value at the mode. In contrast, ISM samples can be expected to fit a distribution closer in shape to Distribution A in Figure 2-15, with less tailing and a mode closer to the mean. This fact can have important implications for making decisions based on discrete sampling data.
Discussion of an idealized spill area scenario is provided in Hyperlink 15 to illustrate the important implications of making decisions based on discrete sampling data for volumes of soil with various levels of contamination.