#10: ND
#9: ND
Grid C
#11: 0.06
(0.09)
#12: 0.24
10 m
#13: ND
(ND)
#14: 0.03
#6: 1.13
#5: 0.06
Grid B
Grid D
Tank
Target
5m
15 m
Grid A
#2: 0.05
#7: 0.53
#15: ND
#3: 0.23
(ND)
#16: ND
#8: 0.03
#1: 0.10
(ND)
#4: 0.06
(0.59)
c. TNT.
Figure 8 (cont'd).
estimates was obtained from different samples
found for various areas were related to the amount
analyzed in different batches and at different
of munitions debris and the apparent length of
times.
time since the area was used.
First to be assessed is the long-range spatial het-
erogeneity by comparing concentration differences
Spatial heterogeneity
among samples from different grids. The previ-
In this discussion, we consider spatial hetero-
geneity on four levels (long range, mid-range,
ous section described a large systematic effect of
short range, and very short range). These discus-
decreasing contaminant concentrations as distance
sions will be based on concentration differences
from the targets increased (Table 2). This pattern,
which is to be expected, has implications for any
obtained using RP-HPLC Method 8330. The dif-
comprehensive sampling plan.
ferences we will be discussing are true differences
The mid-range spatial heterogeneity is assessed
among subsamples analyzed, and are not caused
from differences among the four-point samples at
by imprecision or inaccuracy from analytical
determinations, i.e., analysis of the extracts. The
a given depth from within a grid. Short-range vari-
analytical precision for Method 8330 in our labo-
ability is estimated from differences between pairs
ratory has been documented many times and rela-
of samples that are separated by 1.0 m. For illus-
tive standard deviations (RSDs) range from 2 to
tration, the HMX results for grid A, area 1-44, are
3% (Jenkins and Walsh 1987). An assessment across
used here to estimate mid- and short-range vari-
a number of different laboratories demonstrated
ability. These data appear typical for the site and
that RSDs for Method 8330 were always less
they are convenient to use because all results
than 5% (Bauer et al. 1989, 1990). Triplicate deter-
exceed detection limits. The RSDs for the four
minations on the extract from sample 1-44-8
samples at each of five depths range from 48.1 to
yielded an RSD of 1.9% for HMX determination.
122%, with a pooled RSD of 83.5%.
Accuracy was assessed using spiked samples and
This estimate for mid-scale heterogeneity
the results are presented in Table B6. For HMX,
includes the spatial effects of samples where pairs
recoveries ranged from 88 to 116%, with a mean
are 2.4 m apart and the two samples of a pair are
of 97%. Even better recoveries were obtained for
1.0 m apart (Fig. 8a). With RSDs near 100%, it
RDX, where the mean was 101% with a range from
seems that the HMX distribution is unlikely to be
92 to 110%. Note that each of the nine recovery
Gaussian (normal).
15
Previous Page
