Geoscience Reference
In-Depth Information
Fig. 2.12 Illustration of the
cell declustering method
23.0
These techniques include trend modeling for debiasing and
debiasing using qualitative data, subjects that are not cov-
ered in this topic.
22.0
2.3.2
Declustering with Multiple Variables
Declustering weights are determined on the basis of the geo-
metric configuration of the data; therefore, only one set of
declustering weights is calculated in presence of multiple
variables that have been equally sampled. However, different
declustering weights will need to be calculated when there is
unequal sampling. For example, there are sometimes different
sets of Copper and Molybdenum samples in a Cu-Mo porphyry
deposit, which would require two sets of declustering weights.
Declustering weights are primarily used to determine a
representative histogram for each variable; however, we also
require the correlation between multiple variables. The same
set of declustering weights can weight each pair contributing
to the correlation coefficient (Deutsch 2002 ).
21.0
20.0
minimum declustered mean
19.0
0.
200.
400.
600.
800.
1000.
Cell Size
Fig. 2.13 Plot of declustered mean vs. cell size
2.3.3
Moving Windows and Proportional Effect
The origin of the cell declustering grid and the number of
cells L must be chosen such that all data are included within
the grid network. Fixing the cell size and changing the ori-
gin often leads to different declustering weights. To avoid
this artifact, a number of different origin locations should be
considered for the same cell size. The declustering weights
are then averaged for each origin offset.
Declustering assumes that the entire range of the true
distribution has been sampled. If this is not the case, then
the data is biased and debiasing techniques may be required.
Moving windows are used to understand the local spatial
behavior of the data, and how it may differ from global statis-
tics. The process is to lay over the volume of interest a grid of
cells, which may or may not be partially overlapping, mov-
ing them over the entire domain or deposit, and obtaining sta-
tistics within them. Overlapping windows are typically used
when there are few data within the window to provide reli-
able statistics (Goovaerts 1997 ; Isaaks and Srivastava 1989 ).
 
Search WWH ::




Custom Search