Geoscience Reference
In-Depth Information
SC Zn vs. SC Pb
SC Zn vs. SC Fe
4.0
4.0
3.0
3.0
2.0
2.0
1.0
1.0
ρ =
-.017
ρ =
.013
.0
.0
-1.0
-1.0
-2.0
-2.0
-3.0
-3.0
-4.0
-4.0
-4.0
-3.0
-2.0
-1.0
.0
1.0
2.0
3.0
4.0
-4.0
-3.0
-2.0
-1.0
.0
1.0
2.0
3.0
4.0
SC: Zn
SC: Zn
SC Pb vs. SC Fe
4.0
This feature is
due to correlation
within the first
class of Zn
(Y < -1.28)
3.0
2.0
1.0
ρ =
.016
.0
-1.0
-2.0
-3.0
-4.0
-4.0
-3.0
-2.0
-1.0
.0
1.0
2.0
3.0
4.0
SC: Pb
Fig. 14.42
Crossplot between stepwise conditionally transformed variables for Zn, Pb, and Fe. Zn was transformed first, then Pb was transfor med
conditional to Zn, and finally Fe was transformed conditional to both Zn and Pb (Leuangthong et al.
2006
)
14.4.3
Profit Comparison
Figure
14.44
shows a comparison of the crossplot re-
production from simulation to those crossplots from 25 ft
composites and the existing long term resource model. In
general, the simulated realizations reproduce the trivariate
relations with comparable variability to the 25 ft compos-
ites; the corresponding plots from the existing long term
model shows similar bivariate relations but with notice-
ably reduced variability. Recall that it is this variability
between the multiple elements that was impacting the Zn
recovery, and provided the motivation to undertake such a
case study.
Once all simulated models were generated and vali-
dated by rock type, a single realization for each variable
was obtained by merging the simulated properties from
each rock type. With these multiple realizations (see
Fig.
14.45
), the uncertainty at any location and/or region
can be assessed.
In practice, multiple variables are estimated independently
with ordinary kriging. It is interesting to address the impact
of the multivariate simulation approach using the stepwise
conditional transform relative to the conventional practice
of kriging. Note that this exercise is for illustrative purposes
only, prices and recovery functions have been greatly simpli-
fied for this specific comparison.
The idea is to apply a profit function to obtain a true profit
dataset for Red Dog. A subset of the reference data will be
extracted and used to model grades using both kriging and
simulation. The profit function will be applied to these grade
models. Based on the expected
profit
from each approach,
each location within the model will be classified as either
ore or waste. The true profit at each location is known, so the
profit differential from each model can be calculated.
