Geoscience Reference
In-Depth Information
Fig. 12.8
Schematic example
of resource classification through
kriging variances for reference
drill hole configurations
the corresponding kriging variances for the 4-composite con-
figuration (Case A) define the limit between the indicated and
the inferred categories. Note that the kriging variances are al-
ways used as relative thresholds, since the values themselves
do not have any physical or geological meaning.
Other alternatives for defining resource categories can in-
clude visual inspection of the kriging variances, although rare-
ly there will be a clear break or indication of kriging variances
that can be related to resource classes. Therefore, it is highly
dependent on the subjective criteria to define the thresholds
for each category. Because of this, the method can be consid-
ered equivalent to the distance to the drill hole-based methods,
just developed with in a more formal geostatistical framework.
mine production risk analysis; however, the use of realiza-
tions from which probability intervals can be obtained and
used for resource classification is not yet widespread. The
resource classification codes, beginning with the JORC code,
encourage quantification of uncertainty whenever possible,
but they do not mandate it, nor do the corresponding Guide-
lines suggest specific methodology for such quantification.
Deutsch et al. (
2006
) argue that the uncertainty models
derived from conditional simulations should only be used as
a backup to other more simple, geometric methods, such as
drill hole distance. Several reasons are given in the paper for
this recommendation mostly because the probability inter-
vals are shown to be sensitive to the definition of some of
the parameters used to obtain them, as well as the overall
model dependency. The uncertainty model is dependent on
the specifics of the implementation parameters used in the
simulations (Rossi
2003
).
Probabilities can be checked using actual proportions,
and, whenever possible, this check should be made. Operat-
ing mines will generally maintain sufficiently good produc-
tion records to be able to check actual production tonnages
and grades. If the modeled uncertainty can be verified by
actual production, then there are several good reasons to rely
on the uncertainty model for resource classification: (1) the
magnitude of the grades and the local configuration of data
are accounted for, (2) the mining volume is explicitly ac-
counted for, and (3) uncertainty is perceived as more objec-
tive and transportable to different deposits.
The probability used to define measured, indicated, and
inferred resources depends on the mining company's prac-
tice. Many will simplistically translate the kind of precision
required of other engineering studies and cost estimates
during pre-feasibility or feasibility studies into resource
classification. Typically, a measured resource would be a
quarter known within ± 15 %, 90 % of the time; an indicat-
ed resource, within ± 30 %, 90 % of the time; and inferred,
within ± 30 % and ± 100 %, 90 % of the time. Material known
within more than ± 100 % will not qualify as resource, and
may be flagged (but not publicly reported) as blue sky or
potential mineralization.
12.3.3
Resource Classification Based
on Multiple-Pass Kriging Plans
Another option is to derive the resource classification from
multiple kriging passes. Several kriging iterations are done
to estimate the model grades using different levels of restric-
tions, that is, from a more to a less constrained kriging.
The constraints are defined in terms of requisites for an
estimate to occur; in the more constrained case, a higher min-
imum number of samples combined with a larger minimum
number of drill holes, and shorter search radii may be used. A
smaller number of blocks will be estimated in the more con-
strained pass, but they will be better informed than blocks
estimated in later estimation passes. If the estimation passes
are set based on geologic and geostatistical criteria, a flag for
each block indicating in which pass it was estimated could be
used as an initial indicator for resource classification.
12.3.4
Resource Classification Based
on Uncertainty Models
Conditional simulation provides realizations that provide
models of uncertainty in a global as well as local sense. These
realizations are applicable to both resource classification and
