83
Afterwards, Sequential Gaussian Simulation was independently performed for each Principal Component. The simulation is performed to create 5 realizations over the grid of nodes
shown in Table 2.
The simulated principal components are back transformed to log-ratios, and these logratios are back-transformed to grades. The plots of pairs of the back-transformed grades are
calculated and compared with the corresponding sample scatter plots (Figure 6).
Results show that statistics of the distribution of simulated grades are reasonably close
to the original sample grades, and even some complex relationships are preserved. This is
mainly controlled by conditioning, and not directly imposed by the proposed method which
only captures linear relationships. Care should be taken to ensure that statistics are preserved
at every step of the methodology, to avoid biasing the results. More realizations would likely
lead to a better characterization of the univariate and bivariate distributions.
Figure 7. Plan view and cross section of a single realization of back-transformed simulated Nickel
and Magnesia.
Figure 8. Histograms and basic statistics of the back-transformed simulated grades.
Bsck~transformed Nl
--·& <
--·
Back·transformed Nl
-·--· ... "" -
.
.
• ~J::tt- w • , a_ n • •
· J ·-· · r::.- •• ~
68.71)0 1621.10
_ _ .. $0
--·· ·
-. ...... .,
==t:s
East
14 .000
10.000
8.000
2.000
., . .,
"' · "'
20.000
0.0
-- ~ . . .. .
~er!