Table 11. Comparison of measures of analytical precision, accuracy and discrete sample
representativeness.
Precision
Accuracy
Local heterogeneity
Largest
Slope of
Ratio of highest mean
concentration ratio
0-intercept
concentration vs. lowest
Sample
RSD of duplicates
of duplicates
model
for discrete samples
location
On-site
Lab
On-site
Lab
On-site vs. lab
On-site
Lab
1
3.9
11.1
1.157
1.473
0.815
243
315
2 (DNT)
23.0
10.0
1.655
1.461
0.350
10.6
33.4
3
16.7
6.5
1.822
1.186
1.464
50.0
98.1
4
12.5
13.5
1.696
1.986
0.911
69.0
58.1
5
3.3
4.9
1.126
1.157
0.847
28.9
29.5
6 (Picrate)
11.6
11.9
1.500
1.875
0.967
688
43,000
7R
4.9
5.1
1.265
1.214
0.677
3.8
3.0
8
19.7
4.5
1.731
1.185
1.070
53.1
55.6
9
4.1
5.1
1.131
1.167
1.032
8.2
5.7
Mean
(TNT only)
9.3
7.2
1.418
1.338
0.974
65.1
80.7
Table 12. Comparison of results for discrete and composite soil analysis.
Discrete
Composite
Sampling
Major
On-site
samples
samples
mean SD*
mean SD
Installation
location
analyte
or lab
13,500 16,800
13,100 532
Monite
1
TNT
O
16,300 20,200
14,100 1,420
L
16,100 11,700
23,800 3,140
Monite
2
DNT
O
34,800 42,200
33,600 2,390
L
19.8 42.0
12.6 1.2
Monite
3
TNT
O
12.9 29.0
4.16 0.7
L
1,970 1,980
1,750 178
Hawthorne
4
TNT
O
2,160 2,160
2,000 298
L
156 121
139 16.6
Hawthorne
5
TNT
O
168 131
193 7.7
L
869 1,600
970 32
Hawthorne
6
Ammonium
O
901 1,660
1,010 92
Picrate
L
84,900 33,400
57,000 2,600
Volunteer
7
TNT
O
89,100 20,500
107,000 9,230
L
57,500 25,000
54,800 5,840
Volunteer
7R
TNT
O
86,900 27,900
107,000 7,520
L
9,920 12,000
11,300 2,020
Volunteer
8
TNT
O
8,910 11,600
9,620 409
L
13.7 10.4
16.6 1.5
Volunteer
9
TNT
O
13.0 10.3
11.8 0.3
L
* The discrete sample standard deviations for locations 1, 2, 3, 6 and 8 are all larger than their corresponding
means because the results from these locations are not normally distributed. These results may be log-
normally distributed, in which case the data should be transformed.
deviation (Table 12). It is also useful to note that
the areas we sampled are typically used for site
standard deviations for the on-site analysis of all
characterization and this would only serve to fur-
of the composite samples are low relative to mean
ther increase uncertainties from sampling.
concentration (low RSDs), indicating that in-field
Analysis of composite samples, however, gave
homogenization procedures used were adequate.
results that were good estimates of the mean of
Thus, the number of analyses of the composite
the seven discrete samples, with a low standard
35
Previous Page
