Geoscience Reference
In-Depth Information
Fig. 6.2
Variogram values
from a normal score
transformed variable. The
data points going into the
calculation of three lags are
shown as
h
-scatterplots
outliers and data transformations should be considered. In
general, the variogram is computed in the coordinate sys-
tem and in the units that will be used for modeling, that is,
composite grades are considered and normal scores are con-
sidered if Gaussian uncertainty or simulation techniques are
being considered.
The first prerequisite is to understand the data. Although
experimental variograms may help understand the geol-
ogy and spatial variability of a particular variable, misun-
derstanding and errors could be propagated by calculating
variograms without a reasonable understanding of the data,
trends, and data configuration. The histogram and univariate
statistics of the data should be investigated. Odd data and ex-
tremely high and low data values should be questioned. The
data should be visualized in many different ways. The com-
plexity of the geologic continuity should be understood rela-
tive to the data spacing. The choice of estimation domains
and variables should seem reasonable.
A second prerequisite is to perform variogram calcula-
tions in an appropriate coordinate system. Standard X/Y/Z
coordinates based on elevation and a UTM or a local mine
system may be reasonable for a disseminated deposit. Tabu-
lar or stratigraphic type orebodies may require the calcula-
tion and usage of stratigraphic coordinates (see Sect. 3.4).
Weathered deposits may require the use of depth below topo-
graphic surface as the vertical Z coordinate.
The variogram is a two-point statistic; curvilinear con-
tinuity cannot be represented by a two-point statistic. The
