not possible. However, if 20- 20-cm samples are taken from the 0- to 5-cm depth
(via a scoop or shovel), the larger volume compared to a 5.6-cm-diameter core sample
might cancel the effect of imperfect mixing.
Let's divide the unit into eight 25- 25-m grids and then divide each grid into
sixteen 6.25- 6.25-m subgrids. One 20- 20-cm aliquot from each subgrid would
be formed into a single pile, homogenized as described earlier (section 5), and a
representative 200- to 400-g area-integrated sample would be carefully collected.
Duplicate 20-g subsamples would be analyzed on-site to yield eight mean concen-
tration estimates. If greater precision and accuracy are required, duplicate samples
could be collected from the pile.
When the objective is to produce a mean and upper confidence limit for an entire
remediation unit, compositing could go as follows. Two 20- 20-cm aliquots would
be collected from each of the eight grids, and sampled as described above. Repli-
cate composites, each containing 16 aliquots, could be prepared and 20-g subsamples
analyzed. The number of composites required will vary with the precision speci-
fied for the mean. Aliquot locations within subgrids could be made randomly or
From the above discussion, it should be clear that each site will require decisions
that are based on the preliminary study and coupled to data quality objectives. As
results accumulate, the transition from preliminary study to full-scale sampling
should become more efficient and reliable.