Table 10. 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
217
265
1,970
2,150
9.1
8.1
Hawthorne location 5
5.3
11.0
121
131
22.8
11.9
Volunteer location
7
--*
7,680
--*
19,800
--*
2.6
Volunteer location 7R
5,120
6,320
24,700
27,600
6.1
4.4
Volunteer location
9
1.0
1.0
10.4
12.4
10.4
12.4
* Data unavailable.
either 87.6% for the discrete samples or 100.5%
tion has been available where the components of
for the composites. Clearly, use of the TNT colori-
error have been quantified for soil characteriza-
metric method, with this degree of accuracy, is
tion at explosives-contaminated sites.
justified with nearly any conceivable data quality
For some of the nine sampling locations stud-
objective, but particularly where we have direct
ied here, analyte distribution exhibited such ex-
evidence of the short range heterogeneity present
tremes that use of classical normal distribution
in soils concentrations at these locations.
statistics to fractionate the error was not possible.
On-site results for 2,4-DNT were only avail-
For locations 4, 5, 7, 7R and 9, however, we were
able for sampling location 2 at the Monite site.
able to fractionate the total error variances be-
These results correlate well with laboratory analy-
cause concentration variations were modest (Table
ses (Fig. 5) but the accuracy is not nearly as good
10). For these four locations, standard deviations
as that for TNT. Results are adequate for map-
attributable to analysis were always much lower
ping analyte distributions, selecting samples for
than the corresponding standard deviations from
more quantitative laboratory analysis and locat-
sampling and, hence, total error was dominated
ing areas of high concentration. On-site results
by sampling error. This was true whether charac-
may not be adequate for making decisions in the
terization was done using field analysis or labo-
field about concentrations necessary for action
ratory analysis. For the other sampling locations,
levels.
sampling error was even greater and so over-
The data obtained for the ammonium picrate
whelmed analytical error that this type of frac-
on-site method for sampling location 6 at
tionation would only be possible using asymmet-
Hawthorne AAP are the first validation results
ric (logarithmic) limits. Clearly, if we want to
for the method developed by Thorne and Jenkins
significantly improve the quality of site charac-
(1995). These results were very encouraging and
terization data, the major effort should be placed
it appears that this method may be as accurate,
on reducing sampling error. Single grab samples
relative to the lab method, as the on-site TNT
are totally inadequate.
method.
To reduce sampling error, samples analyzed
The data from this study can also be used to
must be more representative of average concen-
put in perspective the uncertainty introduced in
trations within the area that the sample is sup-
results by analysis vs. that from sampling. In do-
posed to represent than is possible using discrete
ing so we must keep in mind that the goal of site
grab samples. For the data here, if we assume that
characterization is to provide data that can be
the mean analyte concentration of the seven
used to make decisions on whether the degree of
samples taken from this circle with 122 cm diam-
contamination for a given area warrants a cleanup
eter is the "true" concentration, we can assess the
action. Although random grab sampling is ap-
difficulty in achieving representativeness by look-
pealing cost-wise, it may be totally inadequate
ing at the ratio of highest to lowest values in the
for decisions about remedial procedures. To pro-
group of seven mean determinations. These ra-
vide data that can satisfy this need with a high
tios are presented in Table 11 under the heading
level of confidence, total error associated with
of local heterogeneity. These values range from
site characterization must be understood and re-
3.8 to 243 for the on-site TNT method and 3.0 to
duced to acceptable levels. Little or no informa-
315 for the lab method. Much larger grids than
34
Previous Page
