0037im.jpg)
30
ERDC/CRREL TR-02-1
a couple of cases for TNT was two orders of magnitude. It is believed that soil
samples collected in impact ranges are very susceptible to segregation error due
to the random nature of contaminant dispersion size and shape (i.e., explosions).
When portions of these same bulk soil samples from impact ranges were ground
to a fine powder (200 mesh) using a ring and puck mill and taking precautionary
measures not to generate high temperatures (mill runtimes not exceeding 1
minute), the %RSD for sets of 12 subsamples were frequently below 5%, a level
of variation that is not easily separated from the random variation associated with
the method of instrumental analysis (i.e., Method 8330 [Walsh and Ramsey, in
press]).
Sampling Strategy
The major objective of any sampling plan is to obtain representative samples.
This implies that the concentration determined for the sample be representative,
i.e., provide a valid estimate of the average concentration for the specified area of
concern. Therefore, it is imperative that the area of concern be defined prior to
designing a sampling plan. Two typical scenarios are the identification of a sus-
pected surface hot spot and the average surface concentration of EM over a spe-
cified area, e.g., a given area within a training range. In the past, sampling plans
were written for the collection of discrete samples at a specified number of sam-
pling locations, for each of these cases. However, several studies have shown the
futility of this practice, due to extreme short-range spatial variability that often
exists for explosives in surface soils (Walsh et al. 1993; Jenkins et al. 1996a,
1997a, 1997b, 1997c; Thiboutot et al. 1997a, b).
Therefore, composite sampling is strongly recommended when characteriz-
ing the ground surface at a potentially explosive-contaminated site. In a small
area (1 m 1 m) multiple units (30 or more, each of the same approximate
amount) should be randomly collected and placed into a single container. For
large-scale areas systematic gridding is useful for establishing sampling nodes, at
which, an area of between 3 and 10 m square, should be randomly sampled, by
obtaining 50 or more individual increments. Consult Statistical Methods for
Environmental Pollution Monitoring (Gilbert 1987) for selecting the appropriate
sampling design, i.e., grid spacing, to meet various the projects data quality
objectives.
With the exception of coupling soil profile sampling with the installation of
groundwater sampling points, subsurface (vadose zone) sampling should only
occur after the identification of a surface hot spot. Beneath a hot spot it is rec-
ommended that a continuous vertical profile be collected over the first meter,
with the profile broken into several (five or more) increments, and each incre-
Previous Page
