Stirred composite samples
Table 7. Ratio of high to low analytical replicates
and ratio of high to low means for field replicates
The extreme sampling heterogeneity observed
taken at five nodes of a 1.82-m-square grid.
in the discrete sampling was reduced by compos-
iting (Tables 8 and 9). Relative standard devia-
First field
Second field
Ratio of
tions (%RSD) for each of the three mixing proto-
Analytical replicate
sample
sample
high:low means
cols were all much less than those observed for
Row 2, column 3
the discrete samples. However, none of the mix-
1
0.0049
0.0059
ing protocols consistently produced what might
2
0.0063
0.013
be considered low %RSDs. These results empha-
Mean
0.0056
0.00945
1.69
Ratio of high:low rep.
1.29
2.20
size the extreme difficulty in sampling for this
heterogeneously distributed contaminant.
Row 4, column 3
1
0.0542
0.0636
2
0.0293
0.0601
normality, and the mean %RSDs are 58.2, 32.6,
Mean
0.04175
0.0619
1.48
and 29.9% for mixing protocols 1, 2, and 3, respec-
Ratio of high:low rep
1.85
1.06
tively. Both of the mixing protocols that called for
Row 6, column 4
water being added to the composites had lower
1
0.0128
0.0571
variability, and subsamples from the 1:1 sedi-
2
0.0117
0.0541
ment:water mixtures (mixing protocol 2) yielded
Mean
0.01225
0.0556
4.54
Ratio of high:low rep
1.09
1.06
mean concentrations that were in fairly close
agreement to the stirred sediment (mixing proto-
Row 7, column 3
1
0.0944
0.169
2
0.0985
0.186
concentration estimates obtained by the analysis
Mean
0.09645
0.1775
1.84
of discrete samples and those obtained by com-
Ratio of high:low rep.
1.04
1.10
positing, we see graphically that compositing
Row 8, column 4
aided in normalizing the data (Fig. 7). As a result,
1
1.22
0.671
estimates of the mean concentrations obtained
2
1.27
0.818
from composites may be used for statistical com-
Mean
1.25
0.745
1.67
Ratio of high:low rep
1.04
1.22
parisons that assume Gaussian distribution. Mix-
ing protocol 3 (1:2 sediment:water) did not
ber of particles (102 particles in 117 g of sediment)
appear to yield any added benefits over mixing
was found at the sample site with the highest con-
protocol 2.
centrations, which were 671 and 629 g/g for the
Estimates of mean concentration obtained by
duplicate analyses.
the three mixing protocols were compared within
Table 8. White phosphorus concentration estimates for subsamples of stirred composites taken from four subdivi-
sions of a 7- 20-m area, and comparisons of means within each subdivision for composites with and without water
added.
Comparison of means
Comparison of means
Concentrations (g/g)
Concentrations (g/g)
ANOV F
A
Kruskal-Wallis
ANOV F
A
Kruskal-Wallis
Sediment 1:1 sed:water 1:2 sed:water
ratio
H value
Sediment 1:1 sed:water 1:2 sed:water
ratio
H value
Rows 1 to 3
Rows 7 to 9
0.0242
0.0263
0.0191
4.49
6.41
7.02
0.0357
0.0229
0.0228
2.18
2.73
1.93
0.0259
0.0248
0.0264
4.35
4.75
1.84
0.0250
0.0238
0.0206
6.88
1.96
2.14
0.0247
0.0274
0.0239
19.2
2.27
1.84
Mean
Mean
0.0271
0.0250
0.0226
2.2 (p = 0.15) 3.62 (p = 0.16)
7.43
3.62
2.95
1.6 (p = 0.24) 3.98 (p = 0.14)
Rows 4 to 6
Rows 10 to 12
0.892
1.16
0.895
38.3
19.2
18.2
0.813
1.53
0.627
42.7
21.7
14.5
2.22
0.701
0.671
30.8
37.8
16.9
4.97
0.717
0.737
24.4
36.3
16.5
0.865
0.741
0.822
18.9
42.3
12.0
Mean
Mean
1.95
0.970
0.750
1.8 (p = 0.20) 3.98 (p = 0.14)
31.0
31.5
15.6
5.89 (p = 0.0165) 9.38 (p = 0.0092)
12