Geoscience Reference
In-Depth Information
Fig. 12.9
Resource classification
contours, Bench 2440m, Cerro
Colorado 2003 Resource Model,
Northern Chile.
Red
encloses
measured material,
green
outline
encloses indicated material.
Courtesy of BHP Billiton
12.3.5
Smoothing and Manual Interpretation
of Resource Classes
Figure
12.9
shows an example of smoothing through
hand-contouring done at Cerro Colorado, BHP Billiton's
porphyry copper operation in Northern Chile. The smooth-
ing was done by interpretating on benches and smoothing
out the edges and, in some cases, the intermixing of resource
classes. The red outline defines the measured volume, the
bright green outline the indicated volume, and the remaining
material is classified as inferred. Note how some of the mate-
rial originally classified as indicated is inside the red outline
(central-East portion of the bench), and thus finally classi-
fied as measured. Also, there is a small area in this bench to
the Northeast of the picture where measured runs directly
into inferred, due to a change in the geologic environment.
Since resource classification is usually performed on a block
by block basis, most of the non-probabilisitic methods men-
tioned above will generally require a posterior smoothing of
the resulting volumes, mostly because of the common accepted
idea that the classified material should be fairly homogeneous,
without intermixing of resource classes over short distances.
This is mostly an aesthetic issue, since classification
schemes are meant to provide global indicators of confi-
dence, and not necessarily smooth block-to-block images.
Any of the methods described above will likely produce vol-
umes for each resource class that are consistent with the cri-
teria used to specify them. It is common to see in areas with
heterogeneous drill hole spacings, variable geologic charac-
teristics and abrupt transitions between the resource classes.
If smooth and contiguous volumes are desired, then
manually interpreting the zones, based on the initial definition,
is probably one of the most practical means to achieving this.
Alternatives could include running a smoothing algorithm
that would transform, based on windows of certain sizes,
the resource classification of the blocks within to produce
more homogeneous volumes. In any case, this should be
done with care, not to bias or significantly alter the global
volumes defined by the criteria established. There should
only be minor corrections for consistency and what may be
deemed inconsistent classification classes based on geologic
or geostatistical knowledge. It is good practice to check the
overall grade-tonnage curves by resource class before and
after the smoothing process, to understand the degree of
changes introduced.
12.4
Summary of Minimum, Good and Best
Practices
Minimum practice for the development of uncertainty mod-
els requires the application of simple and more traditional
statistical techniques. The scope of application of these
models is relatively small, and can only be attached to large
volumes. The two most common examples include Resource
Classification (for all the methods described, with the excep-
tion of conditional simulations), and global confidence inter-
vals derived from the variance of averages for large volumes.
Risk assessments are thus limited, and normally qualitative.
Good practice requires, in addition to the above, the de-
velopment of conditional simulation to obtain realizations
of an uncertainty model. This model should be reasonably
comprehensive, in the sense of including as many sources of
uncertainty as possible, but principally geologic and grade
