A Practical Approach for Site Investigation - Dealing with Contaminated Sites: From Theory Towards Practical Application - page 161

towards parts of the site that are of more interest and consequently should be sam-
pled more intensively. The big advantage of random sampling is the fact that it
provides a sound basis for statistical conclusions, for example on the mean quality
of the site. Random sampling might have a field of application when determining the
quality of a certain predefined quantity of soil, either being a soil lot on the surface
after excavation, or still in-situ.
Stratified random sampling is far more applied than random sampling as, due
to the stratification, it is impossible that all samples are by chance taken in a very
small part of the site. In stratified random sampling the site is first stratified. In this
stratification, one can take account of the expectation on the level of contamination.
Within the strata, one or more samples are taken at randomly defined locations. The
statistical value is still comparable to true random sampling as long as the strata all
have the same size and in each stratum the same number of samples is taken. When
locations that are of special interest have smaller strata, sampling is intensified in
these areas, even when the number of sampling points within all strata is kept equal.
From a statistical point of view this will result in over-sampling of certain parts of
the site and consequently the determination of the overall quality of the site will
than be biased. At the same time, in these investigations it is most often not the
mean value of a site that is of interest.
The data obtained through sampling can be used to predict where additional sam-
pling is necessary. This can be done in a judgemental way, where the consultant,
based on his expert opinion and in light of all information available (numerical as
well as non-numerical), defines new sampling locations. This is the currently most
used approach. The disadvantage of this approach is that the bias cannot be defined.
Sampling, and consequently the results of the investigation, is as good or bad as the
assumptions made by the investigator. In practice we will be confronted with the
bias when, during a remediation, the actual quantity of soil that is to be excavated is
far larger than was predicted. However, there is no guarantee that a full systematic
or statistical approach will not provide that risk. As long as we cannot look into
the soil and contamination is distributed heterogeneously in that soil, any form of
non-intensive sampling will result in bias.
Alternatively, (geo)statistical sampling techniques can also be applied, where
additional sampling is to be performed where the uncertainty in predicted con-
centrations is too large. There have been numerous attempts to use geostatistical
techniques like kriging for optimization of sampling in the Main Investigation, but
in most situations the spatial relation between data is only limited when compared
to the random variation in contaminated soils.
3.10 Sampling Techniques
Previous Sections discussed the strategic aspects of the investigation of a (poten-
tially) contaminated site with most emphasis on the sampling of soil. Sampling of
other media is of importance as well. During the Exploratory Investigation sampling
will in most cases be limited to soil, while groundwater will be sampled when