Professional Documents
Culture Documents
Theory of Regionalized
Variables
Introduction
It has been known two methods for statistical analysis of mineral
deposit characterization: classical statistics and spatial statistics.
Classical statistical is used to define the properties of sample
values with assumption that sample values is realization of random
variables.
Samples composition/support is relatively ignored, then assumed
that all sample values in the mineral deposit have the same
probability to be picked up.
The presence of trends and ore shoots in mineralization zones can
be ignored.
The fact in earth sciences shows that two samples taken in vicinity
gives the similar value compared to the others in further distance.
In contrary, spatial statistics is used if the sample values are
realizations of random function.
In this hypothesis, sample values is function of their locations in
deposit and their relative position is considered in analysis.
The similarity of sample values which is function of the samples
distance is the basics theory in spatial statistics.
In order to define how closely the spatial correlation among points
in deposit, we must know the structural function which is
represented by variogram model.
Defining variogram model is the first step in spatial statistics
(geostatistics) analysis.
Principally, the calculation in geostatistics needs computer
assistance.
GAMV is a program packet for variogram calculation, while
KB2D and KT3D are program packets for 2D and 3D kriging
application. Both packets are included in GSLIB (Geostatistical
Software Library, by Deutsch and Journel, 1998).
SGeMS (Stanford Geostatistical Earth Modeling Software, by
Remy, 2004) is public domain or open source software packet for
3D geostatistical modeling.
What is geostatistics?
2. Covariance
Covariance of z(xi) and z(xi+h) are:
or practically is:
N
2 (h ) = [z(x i ) z(x i + h )] / N (h ) (4b)
2
i =1
stationary
d. Second Order Stationarity
and