Geoscience Reference
In-Depth Information
Fig. 14.46
Location map of reference BH data (left) and sampled BH data (right) for use in comparing model approaches
Fig. 14.47
Profit maps for ore/waste classification from kriging (left) and simulation (right)
of 532 BH samples of Zn, Pb, Fe and Ba. From this dataset,
25 samples separated at a nominal 100 × 100 ft spacing were
chosen to represent exploration data, consistent with the DH
data available for Red Dog. This subset of data was used as
conditioning data for kriging and simulation.
The variograms for kriging were calculated for the original
data, and the variograms for simulation were calculated and
fitted for the stepwise conditionally transformed data. In both
sets of variograms, a trend was apparent from the experimen-
tal points extending beyond the sill of 1.0. This was not sur-
prising given that the area was purposely chosen to be in the
transition zone between ore and waste, hence a trend from low
to high grades was expected. Trend modeling was not per-
formed for this exercise because of the relatively small area.
For kriging, each variable was estimated independently
using ordinary kriging. For simulation, the stepwise condi-
tionally transformed variables were independently simulated
using sequential Gaussian simulation to generate 100 real-
izations of the grades and subsequently back transformed to
the original units of the data.
14.4.6
Model Construction
The model grid was chosen to be 10 × 10 × 25 ft, which is
similar to the 10 × 12 × 25 ft spacing of the BH data. A total
of 1,600 grid points were modeled. Further, variograms for
both approaches were calculated and fitted using the refer-
ence 532 BH data. This filtered out the influence of poor
variogram inference due to scarce data.
