density was high and kriging standard deviations

locations. Consequently, values of zero were used

were low. Kriging estimates were produced for the

for the nine new measurement locations and only

same grid and the basic univariate kriging estimate

the resultant map of kriging standard deviations

statistics are listed in Table 5-2 (water level B).

(Figure 5-3d) is of interest. The map shows that

The map shown in Figure 5-3c indicates that the

the kriging standard deviations in the lower left

ratio of the original kriging standard deviations and

corner, which formerly had values of about 0.8,

the kriging standard deviations with the nine mea-

have been decreased by a factor of approximately

sured locations removed is always very close to

0.25, which indicates that the kriging estimates,

1.00, which indicates that there is very little dif-

based on the geometry of the network, are more

ference between the two sets of kriging standard

reliable.

deviations and that water levels are oversampled in

the area where the nine measured locations were

removed.

nine locations were added in the southwest corner

elevations. The principal purposes of the examples

where the sampling density was relatively low and

are to familiarize the reader with a kriging exercise

the kriging standard deviation was relatively high.

using bedrock elevations and to describe block

In section 2-4, Equation 2-47 indicates that the

kriging. The data come from an area where bed-

universal kriging variance depends on the vario-

rock consists of a series of intercalated terrestrial

gram, the type of trend, and measurement loca-

deposits that have been weathered somewhat and

tions; in this respect the kriging standard deviation

then covered with alluvium. The opportunity for

does not depend on the values at measurement

measurement error in these types of data is

5-6

Integrated Publishing, Inc. |