was 4.7. These ratios are considerably smaller than
from Method 8330 for individual subgrids and
observed for the seven core samples taken in a
grid composites is presented in Figure 10. The
wheel pattern to represent short-range heteroge-
Method 8330 results are shown because these data
neity. Those ratios, as found here and in our pre-
are available for all 16 grids, whereas the on-site
vious study had values as high as 688 with a
results were only obtained for the 10 grids that
mean of about 60 (Table 5). The area-integrated
were jointly sampled by the Canadian and U.S.
subgrid samples representing mid-range hetero-
team. Concentrations ranged from over 1000
geneity include considerably more soil and, there-
mg/kg in several subgrids adjacent to the target
fore, would be expected to show less variability
tanks to values of about 1 mg/kg at a distance of
than the cores. Despite this spatial heterogeneity,
20 m in front of the targets.
the means of the four subgrids and their corre-
Table 4 presents the on-site colorimetric results
sponding grid composites were in good agree-
for HMX in the 10 grids where on-site analysis
ment, never differing by a factor greater than about
was conducted. Data are shown for individual
2. The mean percent difference was 20.2% for all
subgrids, the mean of the four subgrids within a
values, and it was only 14.1% when two values
given grid, and the results for analysis of the grid
with mean concentrations less than 50 mg/kg were
composites. Results from the acetone HPLC analy-
sis and those from Method 8330 are not shown
excluded. Thus, at this location, subgrid-scale het-
here, but they were very similar. Ratios of highest
erogeneity was large, despite taking area inte-
concentrations divided by lowest concentrations
grating samples, but the ability to produce grid-
composite samples that represent the arithmetic
for individual subgrids within a grid (subgrid
mean of the four subgrid samples was quite good.
variability ratio) varied by factors of 2.6 to 18
The question remains, however, how closely
with a mean ratio of 6.6. This ratio is somewhat
does a subgrid analytical result represent the mean
inflated, though, by the variability for two sub-
surface concentration within the subgrid? To ad-
grids (D9 and D10) with concentrations near or
dress this question we randomly selected 9 sub-
below 50 mg/kg, but even when these values
grids from the 40 discussed above to resample.
were removed, the mean subgrid variability ratio
12
82
285
Grid D7
170
405
135
375
1000
300
1250
835
Grid C4
Grid C5
Grid D1
Grid D2
Grid D4
Grid D5
TANK C
TANK D
1600
970
545
114
180
134
445
455
595
280
94
19
1900
955
580
114
84
43
70
380
575
105
12
28
800
350
890
835
165
135
950
810
Grid C6
Grid D8
13
100
350
500
18
236
490
265
Grid C7
Grid D9
5.8
4.4
65
130
18
2.1
7.1
26
Grid C8
Grid D10
<1
3.8
1.5
3.6
15
39
1.4
1.6
2.1
4.9
1.8
2.0
Grid C10
Grid C9
Grid D11
Figure 10. HMX concentrations from commercial laboratory analysis using Method 8330.
14
Previous Page
