Statistical Sampling Strategies for Survey of Soil Contamination - Dealing with Contaminated Sites: From Theory Towards Practical Application - page 173

x1
Fig. 4.2 Stratified simple random sample of 30 locations in peat soils near Mijdrecht, south of
Amsterdam, polluted by Pb, Cu and Zn. The strata are compact blocks of equal surface area,
formed by k-means clustering of the cells of a fine grid. The geographical stratification enhances
the spatial coverage of the study area by the sample. Compact geographical strata were computed
by R-package spcosa (Walvoort et al. 2009 , 2010 )
with random grid sampling is not fixed, but varies between randomly drawn sam-
ples. We may choose the grid spacing such that, on average, the number of sampling
locations equals the required (allowed) number of sampling locations, but for the
actually drawn sample, this number can be a few locations smaller or larger. A ran-
dom number of sampling locations may be undesirable, for instance, when this size
is prescribed in regulations.
4.2.1.4 Advanced Sampling Designs
There are various, more advanced sampling designs described in handbooks on
sampling theory that have potentials for survey of soil contamination. I will describe
two of these sampling designs, double sampling and adaptive cluster sampling.