required to produce data with a high degree of
were generally in good agreement but more re-
confidence is low. Characterization using a com-
sults from other sites are needed.
bination of composite sampling, adequate in-field
Perhaps the most surprising finding was the
sample homogenization and on-site colorimetric
consistency of the overall precision of results for
analysis, is an efficient method of producing data
TNT. For the seven locations where TNT was the
that are not only accurate and precise, but are also
primary contaminant, average RSDs for dupli-
representative of the area.
cate subsamples using the on-site method with
the discrete samples ranged from 3.9 to 19.7%,
with a mean value of 9.3%. Comparable labora-
tory results yielded RSDs from 4.6 to 13.5% with
APPLICATION OF RESULTS
a mean value of 7.2%. Replicate analyses of com-
The results presented here have several unify-
posites produced RSDs ranging from 4.1 to 17.9%
ing themes that can be applied in designing fu-
(pooled = 10.6%) for on-site results and 2.8 to
ture investigations of munitions-contaminated
15.9% (pooled = 9.6%) for laboratory analyses.
sites. First, it is clear that there was extreme het-
The estimates are approximately equal for com-
erogeneity at all sampling locations. A single
posites despite the extra mixing step, probably
sample from any of the 122-cm-diameter circles
because the wide concentration variations of dis-
could differ by orders of magnitude from the mean
crete samples required large differences in dilu-
concentration of the small area sampled. Relative
tions and the ten-times larger sample size used in
standard deviations (RSDs) for the seven discrete
the on-site analysis. Nonetheless, the consistency
samples were often greater than 100%.
of the pooled estimates is both surprising and
A second consistent finding was that compos-
reassuring. We believe that it is fair to claim that
ite samples of the seven discrete samples could be
subsampling and analysis (SA) typically yields
reliably homogenized and subsampled in the field.
RSDs of about 10% for both field and laboratory
This also opens the possibility of compositing dis-
methods and that extremes of 5 to 20% are to be
crete samples representing a larger area if concen-
expected. Compared to the RSDs for sampling,
tration variations suggest that this approach
these precision estimates represent very accept-
would be desirable. Most important, it permits
able levels.
field processing without elaborate apparatus.
Compositing is an effective way to reduce in-
Another major finding was that the specificity
tersample variance caused by the heterogeneous
and accuracy of the TNT on-site method was quite
distribution of contaminants. The total variance
adequate. The two locations where TNT was not
for the formation and analysis of composites can
be expressed as
the major contaminant were readily identified and
the seven locations where TNT appeared to be
CS2
2
the primary contaminant were confirmed by the
CA
=
+
2
CT
reference HPLC method. The on-site concentra-
n
k
tion estimates agreed very well with laboratory
estimates, except for location 7, where major bias
where CT = total percent relative standard de-
was introduced by removing small stones during
viation
the grinding operation. For the other six TNT
CS = percent relative standard deviations
locations, the agreement shown in Figures 9 and
of sampling
10 was excellent. Admittedly, there were small
CA = percent relative standard deviations
but statistically significant differences in concen-
of analysis
tration estimates at some locations, but their mag-
n = number of discrete samples formed
nitude was insufficient to impart meaningful dif-
into a composite
ferences in conclusions. Of course, each site should
k = number of replicate analyses done
include some reference laboratory analyses to vali-
on the composite.
date the on-site analyses.
For location 2, where DNT was the major con-
In Table 13 we show values of CT for various
taminant, there was a rather large bias between
on-site and laboratory results. This was not unex-
sen for CS and CA are typical of those found here
pected since the on-site DNT method is not as
for field or laboratory analyses of TNT. There
reliable as the TNT method. On-site and labora-
would be nothing to prevent using larger values
tory results for ammonium picrate at location 6
of n, but there is no benefit in using larger values
36
Previous Page
