Statistical Sampling Strategies for Survey of Soil Contamination - Dealing with Contaminated Sites: From Theory Towards Practical Application - page 199

see the final equation in Eq. ( 4.40 ). This extra constraint makes the OK-estimator
unbiased.
It is beyond the scope of this chapter to explain spatial interpolation by kriging
in detail. I refer to Isaaks and Srivastava ( 1989 ), for an introduction in geostatistics,
to Goovaerts ( 1997 ) for an exposé of the many versions of kriging, and to Deutsch
and Journel ( 1997 ) for kriging software.
4.4.3 Required Number of Sampling Locations
The procedure used to determine the maximum grid spacing for kriging the mean
concentrations of blocks, given a constraint on the block kriging variance (Section
4.3.3 ), can also be used to determine the maximum grid spacing for estimating the
concentrations at point locations, given a constraint on the maximum point-kriging
variance, see Fig. 4.14 for an example. Note the large standard deviations com-
pared to those of the stimated block-averages, which is due to the averaging out of
spatial variation within blocks. As we are uncertain about the variogram, I recom-
mend repeating the calculations with several variograms, and choosing the smallest
maximum grid spacing computed with these variograms.
Fig. 4.14 Standard deviation of the estimated log(Pb) concentration at the central point locations
of square grid cells as a function of the grid-spacing
4.5 Detecting and Delineating Hot Spots
A hot spot is defined here as a relatively small area with a concentration of a
contaminant that exceeds a certain threshold concentration. Examples are sites
with soil or groundwater contaminated by some point-source and anomalous sites