Geoscience Reference
In-Depth Information
15.2
Assumptions and Limitations
of the Models Used
is in fact the most important asset on which the mining com-
pany bases its value. By extension, the resource model from
which the reserves are obtained is the single most important
asset a mining company has. As such, they are always sub-
ject to scrutiny.
Auditors are not the only ones that benefit from good doc-
umentation. The project or mine owner does as well, since
the lessons learned from one modeling exercise can be better
applied in future resource modeling iterations. But above all,
ethics and transparency, as required by the current Reporting
Standards, necessitate that all relevant aspects of the work
be laid down in a comprehensive resource model report. The
report must discuss and document what was done, the limita-
tions of the work, and the degree of detail achieved.
Any auditor will strictly follow basic steps in checking
a resource model. The resources estimator is well advised
to be aware of them, and anticipate the documentation that
will be required at a later date. The auditor's duty is to find
flaws, errors, and inadequacies, which is the reason for an
often intense level of scrutiny. The checks may include from
the simplest graphical checks to verification of the database
against original data collection/compilation documents, run-
ning an alternative check model, and bench-marking the
software used in modeling against other software, often the
auditor's own programs.
To cover (and pass) all possible checks, no assumptions
can be made. H.M. Parker's basic auditing axioms simply
and eloquently state the concept: (1) Trust no one; (2) As-
sume nothing; (3) Check everything.
A characteristic of resource estimation work is that it
can be organized in modular form, since it a serial process.
Therefore, documentation can also be arranged in such a
way, and developed as work progresses. Assuming that the
database has already been validated in its original reposi-
tory, the process may begin with (1) loading the data into the
modeling software; then, (2) data checks to ensure that the
loading process was correct; then (3) geological interpreta-
tion and modeling takes place; (4) exploratory data analysis
follows; (5) domain definition; (6) construction of the block
model; (7) variography; (8) grade estimation; (9) resource
classification; (10) resource validation; and finally (11) re-
source reporting.
A good practice is to lay down the procedure on a flow-
chart, which specifies the inputs and outputs of each step,
and also the documentation required. The data files ma-
nipulation should be detailed, including the fields, and the
associated scripts, run files, and programs that are used in
each procedure. In those instances where the runs are for
checking or validation purpose, both runs and run specifica-
tions (parameters and other input files) need to be preserved
and archived.
All relevant variables need to be stored. If the kriging
variance is used for resource classification, or the krig-
Healthy skepticism is encouraged. One illogical extreme
viewpoint would be to accept resource models at face value
since best practices were followed and significant cost was
incurred for professionals and software. Another illogical
extreme viewpoint would be to dismiss the resource models
because of the large number of assumptions required. The fa-
mous statistician George E.P. Box wrote that “essentially all
models are wrong, but some are useful.” Healthy skepticism
must be maintained while constructing the best resource
models possible, and then using them for engineering design
as required.
An important assumption relates to the reasonableness
and correctness of the available data. There will be a variety
of QA/QC procedures in place, but there are many possible
sources of bias and error that may not be fully considered
or accounted for. Moreover, we would assume that there is
some geological continuity of those data values. We also
assume that our geologic models are reasonable represen-
tations of the in-situ geology. And that based on the geo-
logic model, the (stationary) domains defined are adequate
for grade estimation. The geologic continuity is related to
grade continuity, which we assume it is adequately captured
with our variogram models. Most of our estimation/kriging
techniques smooth the data and create models with relatively
large areas of high and low grades. We assume that the data
used to predict the degree of variability in the mineralization
is adequate, and realistically represents for each domain the
local and global variances. The engineers will assume that
these models reasonably represent the reality and plan the
details of a mining method; we assume that the final achieved
ore/waste limits are similar in character to those predicted by
early resource models. Of course, the local details are not as
important as the overall assessment of dilution, lost ore and
continuity.
Also, more detailed descriptions of the degree of check-
ing should be made. Also, mining engineers and manage-
ment should pre-define expected uncertainty (errors) in the
predictions. For example, it is appropriate to define an ac-
ceptable error margin for the ore resource model for speci-
fied volumes, for example yearly or quarterly. All validation,
checking, and reconciliation of the model can then refer back
to the expected error for those volumes.
15.3
Documentation and Audit Trail Required
A major typical shortcoming in most resource models is lack
of, or poor, documentation. It is significant because resource
modeling can be a long and complex process, with many
subjective decisions made along the way. The reserve model
