sorb strongly to soils by hydrophobic interactions.
composite proved to be an efficient way to pro-
duce a representative sample for the grid.
have aqueous solubilities in the 50150 mg/L
A number of on-site methods for TNT and RDX
range and, from a kinetic point of view, dissolve
have been developed. Here we assessed colori-
slowly into aqueous solution. Because of these
metric methods for TNT and HMX, and an en-
factors, high concentrations of these compounds
zyme immunoassay method for TNT. The accu-
can persist in near-surface soils for decades. Once
racy of the colorimetric methods was evaluated
they are dissolved in water, though, they can
by comparison with a laboratory reference method
migrate rapidly through the unsaturated zone to
and the results for both TNT and HMX were quite
the water table and form plumes in underlying
impressive. The slopes of the linear models were
aquifers miles in length.
1.008 and 1.05 for HMX and TNT, respectively,
with correlation coefficients of 0.992 and 0.975.
in near surface soils leads to substantial spatial
Intercepts were also quite low indicating that the
heterogeneity in distribution. Characterization of
the short-range spatial heterogeneity was con-
their detection limits of about 1 mg/kg. The
ducted both here and in an earlier study (Jenkins
results for the colorimetric TNT method agree
et al. 1996). These studies demonstrate that con-
with those presented elsewhere (Jenkins et al. 1996,
centration estimates for discrete samples collected
Myers et al. 1994). The enzyme immunoassay
less than a meter apart can vary by factors rang-
method for TNT produced results in ranges in
ing from 3 to 43,000, with a median value of about
contrast to the colorimetric method that provides
50. Thus, the use of a single discrete sample to
a quantitative result. Correlation analysis between
represent even a small geographical area will lead
the midpoint of the range established by this
to enormous uncertainty. Using on-site homog-
method with laboratory analyses was not quite
enization of several discrete samples followed by
as good as that for the colorimetric method, but
compositing, however, produced a sample that
entirely adequate.
was much more representative of the mean con-
By combining the ability to produce represen-
centration over the area sampled.
tative samples via on-site homogenization and
Midscale heterogeneity was also assessed in
compositing with the ability to obtain accurate
this study. Using an "area-integrated" sampling
analytical estimates using inexpensive on-site
scheme, we were able to reproduce concentration
methods, site investigators can effectively over-
estimates for 3- 3-m sized subgrids quite well
come the problem of spatial heterogeneity for
(Table 7). Even with this scheme, though, the ra-
explosives-contaminated areas.
tio of the highest concentration divided by the
lowest concentration among the four 3- 3-m
on an active firing range. Very little information
subgrids, within a 6- 6-m grid, averaged about 5
of this type is currently available, particularly
(Table 4). Combining the samples from the four
where HMX is the major potential contaminant.
"area integrated" subgrid samples to form a grid-
Firing ranges are found on many military bases
Table 11. Fractionation of total error into analytical and sampling components.
Standard deviation
Ratio
Analytical
Sampling
Sampling/analytical
Sampling location
On-site
Lab
On-site
Lab
On-site
Lab
Hawthorne location 4 (TNT)
217
265
1,970
2,150
9.1
8.1
Hawthorne location 5 (TNT)
5.3
11.0
121
131
22.8
11.9
Volunteer location 7 (TNT)
--*
7,680
--*
19,800
--*
2.6
Volunteer location 7R (TNT)
5,120
6,320
24,700
27,600
6.1
4.4
Volunteer location 9 (TNT)
1.0
1.0
10.4
12.4
10.4
12.4
CFB-V (HMX)
10.9
38.1
123
135
11.3
3.5
* Data unavailable.
26