Geoscience Reference
In-Depth Information
Table 14.12
MIK kriging plans by estimation domain, São Francisco
Estimation
domain
Pass
Rotated
Y
search (m)
Rotated
X
search (m)
Rotated
Z
search (m)
Min no. comps
Max no. comps
Rotation angles
(after rotation)
Θ
1
/
Θ
2
/
Θ
3
HAZ-Hi
1
20
10
20
5
8
−30/−60/0
2
40
20
40
4
8
−30/−60/0
3
130
65
130
3
10
−30/−60/0
SAP
1
20
10
20
5
8
−30/0/0
2
40
20
40
4
8
−30/0/0
3
130
65
130
3
10
−30/0/0
HAZ-Lo
1
20
10
20
5
8
−30/−60/0
2
40
20
40
4
8
−30/−60/0
3
150
75
150
3
10
−30/−60/0
4. There is a change in continuity and orientation at or about
the 0.8-1.2 g/t thresholds, particularly for mineralization
within the Haz-Lo envelope. It is likely that it corresponds to
the mixture of populations within the Haz-Lo, which would
include a more disseminated or stockwork-style zone, with
the occasional presence of the narrow veining characteristic
of the higher-grade central portion of the orebody (Haz-Hi).
5. For most thresholds, 70-80 % of the total variability is
reached at or less than 30-40 m. This implies that for any
kind of kriging, the weights assigned to data beyond this
distance will be minor, and approximately the same in all
directions.
Figure
14.34
shows, as an example, the directional fits for
the three main directions of anisotropy for the 0.2 g/t Au in-
dicator variogram, HAZ-Hi domain.
The indicator models (all exponential structures) behave
as expected, confirming general directions of anisotropy ex-
pected from geologic knowledge and also consistent with
those developed in prior resource models. The 0.2 g/t Au in-
dicator shows the general N-W trend of the data, while cross-
cutting features drive the 3.0 g/t indicator model (for the first
variogram model structure), while the much less important
(in terms of contribution to the overall sill) second model
structure following the general N-W trend as well.
fine correction applied to the estimated distribution quantiles.
This is done prior to deriving the e-type point estimates.
The first approach assumes that if the variance of the dis-
tribution of estimates is similar to the predicted variance of
the SMU distribution, then an appropriate amount of internal
dilution has been incorporated into the resource model. If
so, the grade-tonnage curves obtained from the block model
will approximate the expected grade-tonnage curves of the
SMUs. The second approach relies on a direct correction of
the estimated quantiles. This approach was not attempted
here, mostly because of the lack of production data to help in
the calibration of the method.
The Discrete Gaussian method (DG) was used to obtain
the theoretical (target) grade-tonnage curves from the 2 m
composites. Using the appropriate correlogram models, and
considering the 10 × 10 × 5 m SMU, block variances were
found to be 20.9 % for SAP, 25 % for Haz-Hi, and 17.1 %
for Haz-Lo, respectively, of the original composite varianc-
es. These values imply that the variance reduction is very
significant, which makes the accurate prediction of diluted
resources and reserves difficult, particularly for small vol-
umes. For example, in the case of HAZ-Hi, at an Au cutoff of
0.4 g/t the expected internal dilution (from a 2 m composite
to a 10 × 10 × 5 m block) is about 35 % in grade and 17 % in
tonnage, while for HAZ-Lo the expected grade dilution is
about 12 %, with a slight increase in tonnage. The expected
internal dilution is more significant at higher cutoffs.
14.2.5
Volume-Variance Correction
Consideration of a volume-variance correction is necessary
because ore will be mined on volumes different than the
volume of the composites used in grade estimation, and in
general also different than the volume of each block in the
resource model.
The Selective Mining Unit (SMU) for the operation is
expected to be 10 × 10 × 5 m (500 m
3
). This is based on the
equipment size and characteristics of the open pit operation. In
order to achieve the expected volume-variance correction, two
approaches can be considered: (a) implement a more restric-
tive estimation, whereby the smoothing of the block model
grades is controlled through the kriging plan; the e-type esti-
mates are thus used without further corrections; and (b) an af-
14.2.6
Block Model Definition and Multiple
Indicator Kriging
The block size chosen for the São Francisco resource block
model was 10 × 10 × 5 m, intended to reflect the drill hole
spacing available. It is considered adequate for the available
drill hole spacing, and a reasonable compromise between
drilling density and model resolution. It happens to be also
the assumed size for the SMU block, but in fact there need
not be any relationship between the two.
The resource model is defined by the three triangulations
that represent the estimation domains (HAZ-Hi, HAZ-Lo,
