within the 10 x 10-m grid is presented in Figure 20. Upon close inspection, one set of
higher concentrations of TNT might be present in a line from minigrid #41 (which
had the highest TNT concentration of 289 mg/kg) diagonally to minigrid # 5,
although other minigrids randomly located within the overall 10 x 10-m grid had
similar concentrations. No clear-cut hot spots of high concentrations were
distinguishable.
While we collected four 30-increment composite samples within this 10 x 10-m area,
multi-increment composites with various numbers of increments from the 100
discrete samples can be mathematically simulated. This is valid because multi-
increment composite samples are a physical average of the increments used to create
the composite, and equivalent results have been shown if the increments are
individually analyzed and combined mathematically or composited and the composite
subsampled and analyzed [32].
Sets of 50 multi-increment results for TNT were simulated from the 100 discrete
samples for values of n ranging from 5 to 50. A random number generator was used
to select values with replacement. The minimum, maximum, mean, median, standard
deviation, and tolerance limits (5%) for these distributions are shown in Table XI.
Histograms for the distributions for n equal to 5, 30, and 50 are shown in Figures 21,
22, and 23. As expected, as the number of increments per sample increases, the
difference between the minimum and maximum decreases and the median and mean
become closer together. The trend toward a more Gaussian distribution (as predicted
from the central limit theorem of statistics) is less observable in these data than for
similar data sets from other sites with residues of energetic compounds [31]. The
presence of several high concentrations for individual minigrid samples such as
minigrid # 41 where the TNT concentration was 289 mg/kg and minigrid # 37 where
the TNT concentration was 100 mg/kg may account for this observation or
relationship. Nevertheless, increasing the number of increments per composite
samples does reduce the tolerance range (Table XI) and improve the likelihood of
obtaining a result that is an acceptable estimate of the mean.
It is interesting to compare the results of the four 30-increment composite samples
collected from within this 10-m x 10-m grid with the simulated results for 30-
increment composites. The four measured values for TNT ranged from 10.2 to 11.6
mg/kg with a mean of 10.7 mg/kg. The simulated composites ranged from 7.45 to
41.9 mg/kg with a mean of 15.5 mg/kg. Estimates of the uncertainty from these two
sets of data are quite different. This difference in measured and simulated uncertainty
estimates (minimum to maximum) was unexpected because a similar comparison for
samples collected at the Donnelly Training Area indicated nearly identical results
[33]. The Donnelly data, however, was for 2,4-DNT at a firing point area whereas the
results from the Shaver River Range were for TNT at an impact area. A few of the
minigrid samples from Shaver River Range may have included several small pieces of
explosive that by chance were not included within the 30-increment composite
samples.
4.5 Primerose Lake, Jimmy Lake and the Shaver River
Eight surface water samples were collected in Primerose Lake, in Jimmy Lake and in
the Shaver River to evaluate the quality of surface water. Out of these, 6 were
19
DRDC Valcartier TR 2004-204
Previous Page
