Geoscience Reference
In-Depth Information
1.3.7
Validation and Reconciliation
Production information should be used with great care.
Oftentimes, tonnages and grades reported by the process-
ing plant do not adequately represent true mill feed (head)
tonnages and grades, that is, the material delivered by the
mine. Rather, they may be influenced by plant performance
parameters, which will bias the comparisons with the head
grades and tonnages reported by the mine. The implication
is that reliable head tonnages and grade information are best
obtained from direct sampling of the material delivered at
the entrance of the plant. In some cases these comparisons
may not be possible due to the characteristics of the opera-
tion such as extensive stockpiling or lack of reliable mill
feed information. Often, only very general statements can be
made about the quality of the reconciliation data.
Checking resource models involves several steps and re-
quires a significant amount of time and effort. There are two
basic types of checks to be done: graphical and statistical.
Graphical checks involve 3-D visualization and plotting
the estimated values on sections and plans. Every estimated
block grade should be explained by the data surrounding it and
the modeling parameters and method used. Although these
graphical checks can be performed on computer screens, it is
often worthwhile to have a hardcopy set of maps because of
the level of detail required and the important record-keeping
and audit trails. Unfortunately, this practice is disappearing,
as some operations do not take the time to produce sets of
geological sections and plans views on paper.
Statistical checks are both global (large scale or depos-
it-wide) and local (block-wise or by smaller volumes, such
as monthly production volumes). The checking, valida-
tion, and reconciliation procedures should ensure the in-
ternal consistency of the model, as well as reproduction
of past production if available. Some of the more basic
checks are:
• The global average of the model should match the aver-
age of the declustered data distribution. This check needs
to be performed for each estimation domain.
• The smoothing of the distribution of the block model
grades: the comparison with respect to the predicted
(SMU) grade distributions should be reasonable. If the
predicted SMU and block model grade-tonnage curves
are very different, it is likely that the block model has
incorporated too much or too little dilution.
• The spatial and statistical relationships between the
modeled variables must correspond to the relationships
observed in the original data set.
• A resource model should be constructed using an alter-
native method. The results and differences should be as
expected, given the characteristics of each method.
• The estimates should be compared to previous estimates.
This should be done cautiously and considering the dif-
ferences in data quantity and quality, as well as the meth-
odology used for the different resource estimate.
• The estimates should be compared to all available his-
torical production data. Ideally, resource models should
predict past production. This provides some indication
that the block model may also predict future mining.
Reconciliation against past production should be done based
on pre-defined volumes of interest and according to speci-
fied error acceptance criteria. Additionally, production can
provide an initial indication of the expected uncertainty of
the resource model. This expected uncertainty should be ex-
pressed in the classical form of
within x% confidence limit
p% of the time
.
1.3.8
Resource Classification
The purpose of classifying resources is to provide a global
confidence assessment to the project's stakeholders includ-
ing mining partners, stockholders, and financial institutions
investing in the project. There are several resource and re-
serve classification systems used by different government
agencies around the world. Most of them share in their main
characteristics and objectives.
The assessment of confidence is critical for project de-
velopment since sufficient resources and reserves must be
known with enough confidence to be considered assets. For
operating mines, continued confidence in future long-term
production is also important in providing shareholder value
and supporting long-term planning.
The terminology used in most guidelines for classifica-
tion is purposefully vague. They must be applicable to many
different types of deposits, locations and mining methods.
The guidelines do not prescribe specific methodology for
quantifying uncertainty or risk. Rather, there is increased
reliance on the judgment of the resource estimator, formal-
ized through the concept of a competent or qualified per-
son. A common basis for comparison is therefore difficult
to achieve, since the wording may have different meaning
under different circumstances, and depends on the individu-
als involved. A possible solution is to attempt to describe
confidence in traditional statistical terms, and as a function
of production units. There is an industry trend towards using
a statistical description of uncertainty to supplement tradi-
tional classification criteria.
The confidence assessment required by the sharehold-
ers of a mining project is generally global, and mostly con-
cerned with long-term performance. This is different from
the shorter-term mining risk assessment that engineers need
in the day-to-day operation of the mine. Unfortunately, a
global confidence assessment is frequently also used as a
