Geoscience Reference
In-Depth Information
Table 14.18
Ore/waste classification summary of kriging (left) and simulation (right) relative to true ore/waste classification
pared with the true profit of $ 7.89 M. The results from such
a comparison showed that the simulation approach yielded
$ 7.28 M while kriging yielded $ 7.06 M in profit. Although
these profit values appear high for the relatively small area
of a single bench, the relative percentage increase in profit is
the key result. Multivariate simulation resulted in 92 % of the
true profit relative to the 89 % yielded by kriging. In practice,
this 3 % difference may translate to several millions of dol-
lars in increased profit if a larger area and multiple benches
are considered.
Conventional Gaussian cosimulation approaches are suf-
ficient for straightforward multivariate problems; however,
for the complexity of the Red Dog deposit, these common
approaches are inadequate. The availability of multiple
metal grades within multiple domains warrants some con-
sideration of the relationship between these grades, and how
these relationships change from one rock type to the next.
The approach shown here was designed to explicitly address
this key issue. Consequently, the resulting models not only
reproduce the univariate data and its spatial variability, but
taken together, they also honour the multivariate relations
between the different metals/minerals within the different
domains.
vals that result from the simulation model. Resource clas-
sification is an exercise intended for public disclosure and
thus considers large volumes. It is global in nature. They
should not be used to provide a technical answer on a local
scale, such as risk assessment of mine schedules. This case
study considers two different mining methods (open pit and
underground), which implies that recoverable reserves are
assessed based on different Selective Mining Units (SMUs).
Probability intervals derived for these different SMUs are
contrasted to the uncertainty model developed from the clas-
sification scheme used by the operation to report resources
and reserves. The study assesses the risk of not achieving
predicted tonnages and grades within a mine plan, which is
based on selecting measured and indicated blocks only.
There are several aspects of resource classification
schemes that should be emphasized to better understand the
motivation for this case study:
1. Resource classification is intended to provide some mea-
sure of the degree of confidence in the resource state-
ments. In this sense, it is a global uncertainty model. The
same can be obtained deriving probability intervals from
the conditional simulation models.
2. Internally, mining companies sometimes misuse resource
classifications as risk assessment tools, although the
manner in which this is done varies widely among geolo-
gists, mine planners, and mine management. This stems
from the temptation to use resource classification codes
on a block by block basis, or at a more local scale than
warranted.
3. Despite the existence of codes and guidelines that may
give the appearance of objectivity to the process of re-
source classification, Competent Persons would use dif-
ferent resource classification schemes for any given de-
posit. Management's perception of risk is likely to be dif-
ferent. Technical personnel will generally disagree about
the application of the Resource Classification scheme
used, and how to inject the different levels of confidence
into the mine plan and projected cash flow.
This general lack of understanding of the purpose and mean-
ing of resource classification schemes can be mitigated by
using a geostatistical model of uncertainty. Although they
are not objective, a more detailed description of the pre-
dicted uncertainty for each particular block, phase, zone, or
14.5
Uncertainty Models and Resource
Classification: The Michilla Mine
Case Study
Geostatistical simulation provides a model of uncertainty at
different stages of a mining project and for different types
of risk assessment. Simulation has been used for grade con-
trol in daily operations, to assess the uncertainty of minable
reserves at the project's feasibility stage, and to assess min-
eralization potential in certain settings. Other applications
include assessment of recoverable reserves, resource and re-
serve classification, and drill hole spacing optimization stud-
ies. All large-scale applications of conditional simulations
intend to benefit from a model of uncertainty that describes
the variability observed in the data and its impact on the pro-
cess assessed.
In this case study, a resource classification derived through
more conventional methods is compared to probability inter-
