You are on page 1of 8

Geometallurgy, Geostatistics and Project Value Does Your

Block Model Tell You What You Need to Know?


S Dunham1 and J Vann2
ABSTRACT
The key functions of project evaluation are to assess potential
profitability and to develop an effective operational design. For mining
projects, design aspects include mining methodology, ore treatment
methodology and production rates. All of these characteristics impact on
project economics and overall value.
Resource models, which form the ultimate basis of project evaluation,
typically consist of tonnes above cut-off, grade above cut-off and the
spatial distribution (connectivity) of tonnes/grade above cut-off, for one
or more variables. Dilution, ore loss and metallurgical recovery are all
modifying factors applied to the resource model during evaluation and
determination of profitability. A key issue is: how is the value
determined? In addition to traditional combination of in situ tonnes/grade,
the following merit serious consideration: the concentration of deleterious
elements, throughput rates, mining/processing costs and importantly
metallurgical recovery.
Geometallurgy is an emerging field targeted at integrating these issues
by identifying either direct measures or proxies for throughput (hardness,
grindability), recovery (liberation, mineral shape/texture, etc) and
concentrate quality from easily collected macro-, meso- and microscopic
data. These geometallurgical variables drive project costs and revenues
in a fundamental way. For geometallurgical characterisation to have a real
impact on business design, it must enable improved mine planning. When
creating spatial estimates for such variables, it is preferable to find
proxies that can be measured as close to the intact rock mass as possible.
However, spatial estimation of proxies for geometallurgical properties
and responses is complex and requires special consideration. Unlike
grades, both proxies and absolute measures of geometallurgical variables
are not necessarily linear or additive and therefore require very careful
geostatistical consideration. In some cases, models must predict extreme
values of geometallurgical attributes rather than averages.
Incorrect characterisation of metallurgical recovery/throughput can
(and has) led to misspecification of the scale of projects, and thus can be
seriously value destructive. Aspects of geometallurgy have long been
important in iron, bauxite, manganese and coal deposits and are
increasingly on the radar for base and precious metal miners. However, as
more metallurgically complex deposits are developed and mining of
large-scale, low-grade deposits becomes more commonplace, the
importance of characterising the metallurgical response in order to
generate properly optimised projects will rapidly increase. It is possible
that in many traditional (ie grades and tonnes) reserve models, the
variables that drive the project are missing!

INTRODUCTION
The key objective of any resource industry project evaluation is
to develop an effective, value-creating business case for the
mining and processing of an in-the-ground asset (the orebody).
During the evaluation of mining opportunities, different mining
and treatment methods are combined and evaluated at different
production rates. This process is intended to identify economic
combinations and ideally the best case or combination that
maximises the value of the asset. The complexity of dealing with
multiple, inter-related attributes, parameters and systems that
affect value, creates a risk that the business case analysis will be
incomplete and the evaluation flawed. In a worst-case situation,
the critical success factor(s) may be missed or incorrectly stated.
1.

MAusIMM, Principal Consultant, Director, Quantitative Group, PO


Box 1304, Fremantle WA 6959. Email: sd@qgroup.net.au

2.

FAusIMM, Principal Consultant, Director, Quantitative Group, PO


Box 1304, Fremantle WA 6959. Email: jv@qgroup.net.au

Project Evaluation Conference

In these cases, major project re-engineering, sometimes at


significant capital expense, may be required to prevent project
failure.
Resource project analysis is complex. The effective evaluation
of even a single option requires reduction of the input data using
models. Reduction reduces the number and variability of the
multitude of input data that drive project economics. The
successful evaluation of project options depends on well thought
out approaches to this reduction step. Two of the most common
models used during project evaluation are:
1.

the resource model, and

2.

the financial model.

The selection, design and construction of the models used in


project evaluation must include consideration of the accuracy
of the data the model is based on. Oversimplifying or
overcomplicating the model beyond the capacity of the input
data is a fundamental error.
As with all paradigms, it is important to understand the
strengths and weaknesses of the model, two key questions being:
1.

What was the model designed to evaluate?

2.

Is it being used in context of its original design or, have


project requirements evolved and is the model now being
used for purposes outside the original design intent/scope?

Consider a financial model designed to evaluate the impact of


diesel burn rates on mining costs. Such a model may contain
downstream parameters for mill throughput and recovery to
allow a full operational NPV analysis. What would happen if this
model was used to evaluate different processing and recovery
paths? Is it sufficiently detailed in the appropriate areas to
produce meaningful results or will any study of processing
options based on this model be inherently flawed?
These concepts also hold for resource models. It is essential
that geological, mine planning, metallurgical and financial
personnel fully understand the basis of the resource model and
how it is intended to be used. The individual technical
professions involved in evaluating projects often do not
understand how other functions impact the business. For
example: metallurgists take geological resource estimates as
highly precise and deterministic results; geologists dont
appreciate how their imprecise estimates are used in value or net
smelter return (NSR) calculations, etc. The authors suggest
mining companies need first to identify the critical business
drivers and then, across all disciplines, discuss and quantify the
derivation of the drivers, focusing on the precision and
underlying variability inherent in each estimate and model.
Sharing this understanding should (hopefully) trigger a greater
appreciation of the risk and uncertainty the project or operation
is facing, allowing a more complete consideration of risk and
opportunities.

RESOURCE MODELS
It is impossible to effectively evaluate a resource project without
understanding the in-the-ground asset to be exploited. This
understanding is typically in the form of a 3D or 2D block model
of the mineralisation. Mining engineers evaluate different
volumes of ore and waste using various heuristic models of
mining methods to determine the economic limits of zones of

Melbourne, Vic, 19 - 20 June 2007

S DUNHAM and J VANN

mineralisation (Figure 1). This evaluation creates a range of


options with different values. Depending on the project
evaluation philosophy, the mining sequence for the options
with the highest perceived value and/or lowest risk is created.
This mining sequence, which typically consists of a mining
tonnage and grade scheduled over time, is then used by
metallurgists, civil/mechanical engineers, commercial personnel
and management to determine downstream production rates,
processes and the viability of the project. This analysis of
mining, processing and commercial scenarios is iterated until a
robust case that meets or exceeds company investment
requirements is achieved.
The most critical input into the entire evaluation process is the
spatial estimate (or resource model). This defines the spatial
distribution of tonnes and grade for mineralisation and is the
template or blueprint mining engineers use to create mining
value options. All schedules of tonnes and grade, and thus all
cash flows and downstream financial analysis are derived from
this spatial estimate.
Recently, with the coincident increase in computing power and
emphasis on business risk analysis, geologists are incorporating
new factors in their spatial models. As well as the grade of
the key economic elements, spatial estimates for deleterious
elements, important elemental ratios (for example Cu:S in copper
deposits or Fe:MgO in nickel sulfide projects) and factors such as
metallurgical recovery and throughput are being created (Table 1).
The intention of adding these parameters to the resource model is
to permit a higher level of analysis of the value of the in-theground resource. Logically, the earlier information about the
economic drivers of a project is available, the earlier it can be
incorporated into business analysis.

TABLE 1
Examples of geometallurgical variables and their importance.
Property

Geometallurgical description and effect

Cu:S

Proxy for sulfide mineralogy;


influences copper concentrate grade, eg low Cu:S ratio
indicates copper rich secondary minerals resulting in
high copper concentrate grades; and
important consideration in smelter design.

Fe:MgO

Proxy for talc concentration in ore,


talc can contaminate nickel concentrate and cause froth
handling problems in the concentrator, and
Fe:MgO is also critical in smelter design as low ratio
ores are more refractory.

RQD

Rock quality descriptor;


proxy for rock strength;
component in various geotechnical rock mass quality
classification schemes; and
can be used to predict relative throughput rates in SAG
milling circuit.

In considering these additional factors, mining professionals


are looking in more detail at the definition of ore, waste and, by
implication, the definition of cut-off grade. Conceptually there is
a simple, fairly clear-cut distinction:

ore can be defined as material that it is economic to mine,


process and sell; whereas

waste is uneconomic.

FIG 1 - Classic steps in mine planning evaluation.

Melbourne, Vic, 19 - 20 June 2007

Project Evaluation Conference

GEOMETALLURGY, GEOSTATISTICS AND PROJECT VALUE

In many operations a series of cut-offs defines a complicated


allocation of material, for example: waste, mineralised waste,
low-grade, run-of-mine, high-grade, etc. In addition any or all of
these categories may have further criteria applied to produce
additional material classes eg according to deleterious material
content or other non-grade attributes. Although a simple ore/
waste distinction is discussed here, conceptually any complicated
material allocation scheme is just a flow chart of simple
ore/waste type decisions applied successively and the principles
set out here apply to these more complicated (and realistic)
scenarios. The greater the number of material classes in any
classification scheme, the more important it is to have robust
spatial estimates of the criteria used to define each class.
Similarly, the more complex the classification scheme, the higher
the probability of misclassification errors. Consideration of the
impact of these errors is important in any project evaluation.
Regardless of the ore/waste classification scheme, the
ore/waste distinction is only as robust as the inputs and models
used to evaluate the economics of the material in question at the
scale of the envisaged ore/waste decision. During mining
operations, geologists send parcels of ore to the mill based on
proxies for economic drivers (such as grade). The size of the
parcel (ie the volume of material defined as ore/waste) depends
on the mining method, equipment and production rate.
This basic definition of ore can be improved. A review of the
history of cut-off grades (Lane, 1988) shows a steady progression
from basic break-even analysis to calculations that incorporate
the effect of production rate constraints. More recently (Hall and
Stewart, 2004) cut-off grade analysis has incorporated net
present value (NPV) and annualised throughput as key drivers.
This NPV approach can be further extended by considering the
ore/waste decision in terms of the processing rate of parcel in
question, that is the total metal produced per hour. For example,
consider two parcels with the same tonnes and grade (see example
in Figure 2). The first parcel can be processed at a rate of 100
tonnes per hour and has a metallurgical recovery of 75 per cent.
The second parcel can be processed at a rate of 110 tonnes per
hour and has a metallurgical recovery of 72 per cent. Both
parcels are economic (ie they cover their mining, processing,

administration and marketing costs). In a standard definition,


both parcels are ore and would be treated equally. From a time
value perspective, however, the second parcel will deliver a
higher value than the first as its throughput rate is higher even
though the recovery is lower.
This approach, essentially looking at maximising the
recovered metal per unit of time, is the basis of most mine
schedule optimisation approaches. Therefore, in order for an
operation to deliver the expected financial return, the same
approach must be used in grade control and operational
ore/waste decision-making. This requires a much more detailed
understanding of the key value drivers such as recovery and
throughput; ideally a spatial model. This is the motivation for a
geometallurgical approach to mine design.

GEOMETALLURGY
Geometallurgy is a cross-disciplinary approach with the
objective of addressing some of the complexities associated with
determining the value of a resource and therefore if it is
economic to exploit. By integrating geology, mining planning,
operational design, mineral processing and metallurgy,
geometallurgy aims to improve the fundamental understanding
of resource economics.
Geometallurgy is relevant at both feasibility study and
operational phases. A geometallurgical approach to mine
planning and plant design is based on identifying the various
attributes that contribute to the realised value of a resource.
Collectively these attributes constitute material characterisation
and, in addition to traditional attributes such as the grade of the
economic components, include less traditional attributes, for
example:

concentration of deleterious elements (both those resulting in


penalties for sold concentrates as well as those impacting on
recovery or other processing responses),

hardness,
grindability,

FIG 2 - Value is driven by grade, recovery and throughput rate.

Project Evaluation Conference

Melbourne, Vic, 19 - 20 June 2007

S DUNHAM and J VANN

mineral species and mineral grade,


mineral liberation,
metallurgical recovery,
mining recovery,
drillability,
fragmentation and dustability,
reagent consumption, and
smelting characteristics.

GEOMETALLURGY AND RISK


Improved material characterisation, combined with the spatial
modelling of critical physical characteristics that form the
basis of a geometallurgical approach, provides a much improved
basis for operational and mineral processing plant design during
project evaluation studies. This approach reduces the risk
associated with developing new operations or expanding an
existing business.
By increasing orebody knowledge and then using it to design
the entire operational flow sheet (from in situ rock to end
product) a more reliable and realistic model of the production
capacity of the system can be developed. Key bottlenecks,
product constraints and low-cost/high-value opportunities can be
identified. The result can be increased total metal recovery and
improved asset utilisation. This in turn allows better decisionmaking, in terms of capital expenditure, to optimise project
economics.
Applying geometallurgical modelling techniques can directly
reduce the risks associated with project design and viability.
Geometallurgy has the potential to act on both the consequences
and likelihood axes to decrease risk.
As knowledge of material types, their spatial distributions,
spatial associations and likely performance parameters increases,
the quality of scheduling and mine planning outcomes can be
improved. Production plans can be developed to reduce
variability, take advantage of blending to better achieve
downstream product specification and deliver a more reliable
result. Knowledge of the spatial distribution of deleterious
material may, in some instances, be a deciding factor between
alternative mining methods (due to differences in selectivity, for
example). Thus the likelihood of producing out-of-specification
concentrate can be significantly reduced by choosing an overall
mining strategy based on this knowledge. More knowledge
equals less production uncertainty and less likelihood of a new
operation failing due to an overlooked fundamental performance
driver.
The consequences of an adverse project outcome can also be
minimised by improving the understanding of geometallurgical
parameters. By taking a holistic approach to ore characterisation
and spatial estimation of metallurgical parameters, operations
can be designed to recover from potentially unfavourable events.
For example, if we understand that there is a risk of concentrate
being out-of-specification, but (for other compelling reasons, eg
geotechnical limitations) accept a mining method with low
scheduling flexibility, we can manage the consequences.
Changing the engineering design of the processing stream, for
example by adding a step to clean the final concentrate,
may reduce the risk and maintain value. Geometallurgy allows
increased response speed and opportunities to reduce the
duration of unplanned events.

traditional barriers between professional disciplines. The


geologists, engineers and metallurgists involved in a project must
truly want to work together with a shared goal and recognise the
value they can each bring to the program. No individual
discipline leader must be allowed to believe they have all the
answers. This type of approach requires geologists, engineers
and metallurgists to develop significant and shared understanding
of the mining and ore treatment flow charts. Technical
professionals must become multi-lingual, that is capable of
speaking (and understanding) geological, metallurgical and
business language and concepts. Because it is a crossdisciplinary approach, geometallurgy cannot work if driven by
only one silo.
Secondly, project teams must think explicitly (and constantly)
in terms of uncertainty and risk. Most complete financial models
incorporate factors for metallurgical recovery and throughput.
Usually these factors have been developed by metallurgical and
civil/mechanical engineers after relatively limited test work.
Typical examples of tests used to determine throughput include
drop-weight tests or heuristic throughput-hardness relationships
based on unconfined compressive strength (UCS). Recovery
models can be based on locked-cycle tests and pilot plant results.
The cost, time and sample size required for these types of tests
generally limits the total number available at the project
development stage. Thus, decisions on plant design and
operating strategy incorporate a high degree of uncertainty and
risk. The aim of the geometallurgical approach is to reduce this
uncertainty. This is achieved by identifying suitable proxies for
value drivers such as recovery. Ideally, the performance proxies
should be easy to measure, preferably in drill core or during
mapping. In most circumstances, collecting geometallurgical
proxies will mean changes to the traditional core logging
approach. Instead of qualitative logging by (sometimes junior)
geologists, more expert, advanced quantitative and perhaps
automated logging techniques are required. These techniques
may include:

high-definition digital photography and image analysis,


online non-destructive hardness testing,
detailed geophysical characterisation of drill core, and
scanning electron microscopy of crushed core and assay
pulps.

Developing drill core based proxies and automating the


recording of this proxy data allows a step-change increase in the
volume of data available. This, in turn, can improve the
possibility of spatial modelling of the physical value drivers.
These proxies, when validated by traditional metallurgical or
other tests, are then scaled up to give a more complete picture of
the value characteristics of the deposit.
As testing and modelling progresses, using a geometallurgical
approach allows sampling to be focused on the right materials.
Relevant and representative samples indicative of all materials
(economic and deleterious) can be selected to ensure full
orebody characterisation. This addresses a common failure of
traditional metallurgical test work programs their lack of
representivity.
The use of value proxies is implicit in the resource industry.
The main value proxy is grade. Geometallurgical modelling
extends this basic proxy to incorporate the true economic drivers
of the operation.

MODELLING GEOMETALLURGICAL ATTRIBUTES


PRACTICAL IMPLEMENTATION
There are a number of important factors that need to be in place
if the geometallurgical approach is to be successful. The first and
foremost factor for success is to break down some of the

For geometallurgical characterisation to have a real impact on


business design, it must enable improved mine planning. For this
there needs to be a map of the physical characteristics identified
as impacting the value drivers, as discussed above.

Melbourne, Vic, 19 - 20 June 2007

Project Evaluation Conference

GEOMETALLURGY, GEOSTATISTICS AND PROJECT VALUE

Creating a spatial estimate of the geometallurgical attributes of


a deposit is analogous to creating a grade resource model. The
fundamental difference is that the resultant 3D model contains
additional non-grade variables related to the realised value of
each block. In this process, careful selection of appropriate
spatial modelling tools is essential because many geometallurgical
attributes behave quite differently when considered in a spatial
context, compared to the grade variables most operations are
familiar with. This necessitates a well thought out estimation
approach for these new variables. In some cases the variables
may be non-additive and/or require non-linear estimation
approaches. In other cases it may be preferential to simulate
rather than estimate because the important aspect of the attribute
is not local average values, but the frequency of extrema. Each
deposit will be different and understanding the required business
outcome is essential to adopting the correct geostatistical tool.
One of the most critical considerations when dealing with
geometallurgical variables is that of scale or support 3. For
traditional grade variables, it is relatively easy to explain that the
influence of support (or volume variance effect) means that
different sizes or volumes have different variability, as measured
by dispersion statistics such as range or variance. Consider a
mineral deposit sampled by small diameter drill core compared
to the same mineral deposit divided into larger volume mining
blocks. It is intuitive that, although the average grade of drill
cores and mining blocks will be the same, the variability of
grades will be smaller for larger supports. The drill core will
have more extreme high and low grades, thus higher variance,
than blocks.
The same intuitive statement does not necessarily hold for
some geometallurgical variables. What scale, for example, does a
flotation recovery test result represent? Is it the volume of the
original sample or is it really a grain by grain measurement? Can
the recovery measurement on a 100 kg sample be up-scaled to a
100 tonne mining block? This is a complicated issue and the
practical impact on generating and using modelling of recovery
is potentially material. In any case, one thing we do know about
attributes such as recovery is that they have a different support to
the usual grade samples. This applies also for:

grain size,
size distribution curves,
mineral grade, and
presence/absence indicators.

The question of support and scale is also important when


consolidating or combining variables, for example when
evaluating a block model. Some variables can be linearly
averaged and others cannot. The property that allows the mean of
some variables to be calculated by a simple linear average is
known as additivity. To legitimately average values of an
attribute without generating biases, we must ensure that the
attribute we are dealing with is additive. This is true for simple
arithmetic averaging, or for other linear combinations, such as
weighted averages. Kriging and other common spatial estimators
are all based on weighted averages, which presume additivity of
the attribute being estimated.
When a variable is non-additive, simply taking the average or
the tonnage-weighted average of the variable is not correct. This
is easily demonstrated by considering a typical metallurgical
grade-recovery relationship where the recovery is defined using
a fixed tail model. This model might be represented by a
polynomial function derived from regression (Figure 3).

EXAMPLE OF NON-ADDITIVE VARIABLES


Figure 4 illustrates the potential impact of support and
non-additive variables. In this example, a simple grade-recovery
relationship exists. The recovery is stated as a modified
polynomial function. The recovery for all material with a grade
below 0.75 per cent Cu is set to zero. Similarly the recovery
formula prevents spuriously high results by setting the recovery
for material with a grade above 4.5 per cent Cu to 80 per cent.
Two different ore parcels are shown. These parcels could
represent dig lines in an open pit or a preliminary analysis of
stope blocks for an underground operation. The first parcel
(preliminary block) consists of contiguous cells above 2.0 per
cent Cu. For the second parcel (extended block of 20 cells) the
mining outline is extended to incorporate two isolated high-grade
cells lying above the original shape. This type of iteration of
various parcel sizes is typical of a grade control function.
The recovery for the two different parcels can be calculated
using two different approaches:

strength (eg compressive strength or drop weight tests),


liberation,
3.

texture,

Support refers to size, shape and geometry of volumes upon which


we either estimate or measure a parameter, eg 2 m half NQ core
support; 12.5 m 12.5 m 9 m mining block support, etc. See
Journel and Huijbregts (1978) for definitions.

1.

Calculate the tonnes-weighted average grade of all of the


cells within the ore parcel and use this grade as an input

Fixed Tail Grade-Recovery Relationship


90.0%
y = -0.021x4 + 0.249x3 - 1.116x 2 + 2.355x - 1.321

80.0%
70.0%

Recovery

60.0%
50.0%
40.0%
30.0%
20.0%
10.0%
0.0%
-

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

4.50

Grade (% Cu)

FIG 3 - Example of a polynomial regression recovery curve.

Project Evaluation Conference

Melbourne, Vic, 19 - 20 June 2007

S DUNHAM and J VANN

FIG 4 - Example grade and recovery maps.

into the grade-recovery relationship. This approach results


in a recovery of 69.9 per cent for the preliminary ore parcel
and 72.7 per cent for the extended ore parcel.
2.

Calculate the recovery of each individual cell using the


grade-recovery relationship and then combine these
individual cell recoveries to derive the ore parcel recovery
using weighted averages. This approach results in a
recovery of 69.5 per cent for the preliminary ore parcel and
64.9 per cent for the extended parcel.

These results demonstrate the non-additive and non-linear


nature of the polynomial grade-recovery relationship. For the
preliminary ore parcel, the difference in the approaches is minor.
By contrast however, the two calculated recoveries for the
extended parcel are materially different (72.7 per cent and
64.9 per cent). This difference is driven by the presence of lowgrade material in the extended ore parcel. Two cells have grades
below two per cent Cu. One of these two cells has a grade below
the 0.75 per cent grade-recovery threshold and therefore has a
zero recovery.
Which of the two calculations is correct? That depends on the
scale of the original grade-recovery relationship data and hence
the scale implicit in the polynomial regression. For example, if
the grade-recovery relationship is based on test work on drill
core data, it is likely that neither calculation is correct as, the
support of the data informing the polynomial regression is an
order of magnitude less than the support of either the individual
cells or the combined ore parcels. The variability and proportion
of high grades will be greater for the small support of the drill
core than the larger support of the cells or parcels. Therefore the
test work results will be influenced by this greater variability and
high grades. If, however, the grade-recovery relationship is based
a heuristic analysis of daily recovery results achieved in an
operating ore treatment plant, the scale of the underlying data is
the tonnage treated each day.

In our example, consider the support of the grade-recovery


relationship to be ten cells. This might represent test results
from a series of bulk samples collected during a mining study.
During the course of the study it is determined that the tonnage
equivalent to ten cells is the average shift-by-shift throughput
rate of the proposed ore treatment plant. Using this support, the
recovery for the extended ore parcel (20 cells) is 72.7 per cent.
This worked example represents a simple case of modelling an
important high-level geometallurgical variable. For the more
realistically complex mineral deposits, dealing with nonadditivity and questions of support are much more involved than
this simplistic example suggests. Additionally, as with all spatial
estimation, defining spatial domains (zones of like behaviour)
is critical. While it is likely that the quantity of geometallurgical
data will increase by an order of magnitude over the next few
years, the processes of estimating and/or simulating these
attributes remain poorly understood and needs to be the subject
of research.

MISLEADING MODEL ATTRIBUTES


The properties of each attribute in a spatial model must be well
understood. As well as considering the additivity properties of a
variable, the significance of the variable must also be understood.
In the case of metallurgical attributes for example, is it the
average value of the attribute that is important in the physical
process or is it the presence/absence of extreme values. Failure to
properly understand the meaning of an attribute can result in a
poor choice of modelling algorithm, which in turn may destroy
the very aspect of the variable that is important in the
downstream evaluation. While not a metallurgical attribute, a
good example of this problem is illustrated by considering rock
quality designation (RQD), rock mass rating (RMR), or other
similar geotechnical attributes. RQD is a measure of the

Melbourne, Vic, 19 - 20 June 2007

Project Evaluation Conference

GEOMETALLURGY, GEOSTATISTICS AND PROJECT VALUE

proportion of unbroken drill core greater than 100 mm long. It is


used in various geotechnical rock mass classification systems.
RMR is a geomechanical classification system developed by
Bieniawski (1989). RMR is based on six parameters (UCS,
RQD, groundwater conditions, and the spacing, condition and
orientation of rock mass discontinuities.
The absolute RQD value is not the only important aspect of
this variable. As well as knowing the RQD, the frequency
distribution of RQD values provides additional detail about the
rock mass properties. RQD is a non-additive variable. While
averaging RQD values has been common practice, the
engineering properties for the rock mass are more highly
influenced by areas with very low RQD (ie very fractured
ground). Any spatial modelling approach that fails to properly
reflect the presence of low RQD values may result in poor
downstream decision-making.
Linear estimation approaches such as ordinary kriging do not
preserve the frequency distribution of the initial data. Extreme
values (high or low) are smoothed during estimation, which is
appropriate to produced grade estimates that are not
conditionally biased (see Vann, Jackson and Bertoli, 2003).
However, basing engineering design on smoothed estimates of
RQD is risky because the presence of a small zone of low RQD,
that may control how the rock mass will perform, will be
obscured.
It is easy to identify metallurgical attributes that are similar to
RQD in this regard, ie where the extreme values are more
important than the absolute values. For example, in a flotation
system that is sensitive to talc or fibrous minerals, smoothed
spatial estimates might indicate that a parcel of material is within
specification, whereas in reality a small part of the ore parcel is
highly contaminated. If the small contaminated portion is
sufficient to cause circuit instability, this small, unrecognised
region may contaminate the metallurgical performance of a
much larger volume of ore (Figure 5).

CONCLUSION
Evaluating a resource industry project is a complex task
involving multiple variables and parameters. Recognition of the
most important levers and drivers of project success (and
potentially failure) is vital to delivering sound project evaluation
and business opportunities.
A key area for consideration in many projects is metallurgical
performance. In many instances the relationship between
the underlying characteristics of the mineralisation and the
performance of that mineralisation through various ore treatment
and concentration systems is not properly examined and
modelled. The different nature of geology and the various
engineering disciplines often results in information silos, poor
communication and consequently misunderstanding between
professions. In some cases the level of misunderstanding can be
so great that it causes critical success factors to be missed or
incorrectly stated.
The geometallurgical approach is intended to improve
cross-disciplinary communication and address some of the
underlying complexity involved in project evaluation and
modelling. Geometallurgical technology is immature but holds
great promise should some of the underlying assumptions be
resolved. However, the promise of geometallurgy will only be
realised if the evaluation and planning process is modified to
appropriately integrate the new variables and attributes. This will
require changes to our understanding of ore/waste classification
and value. As well as raw grade estimates, successful
geometallurgical models incorporate spatial estimates of ore
treatment performance parameters such as throughput, recovery
and concentrate quality.
There is no one size fits all solution to creating spatial
estimates and maps of geometallurgical attributes. Each
operation and each material type is unique and needs to be
considered uniquely. Careful consideration must be given to the
design and implementation of sampling and data collection
programs for any value driving attributes included in a spatial
estimate. Most importantly, the support and additivity properties
must be thoroughly thought out. The significance of extreme
values on downstream value processes should also be considered.
Both these considerations should determine the selection of
geostatistical tools.
Failure to properly understand the characteristics of
geometallurgical
attributes
or
recognise
fundamental
geometallurgical value drivers has the potential to seriously
impact on project success. Unless upfront thought is given to
ensure all disciplines understand the basis of the value
proposition, it is possible that the primary input of an evaluation
(the block model) will not only fail to tell you what you need to
know but it may mislead you into believing that crucial aspects
have been adequately analysed or modelled when, in fact, the
data underlying the analysis is flawed.

ACKNOWLEDGEMENTS
FIG 5 - Example impact of above specification talc on a flotation
circuit (courtesy of BHP Billiton Nickel West, Mt Keith Operations,
Wiluna, Western Australia).

It is possible that metallurgical attributes such as liberation are


controlled by the extremes in a population rather than the
averages. If this is the case, simply creating spatial estimates of
liberation proxies using an approach such as ordinary kriging
will fail to properly represent the variability and performance of
the system, leading to suboptimal evaluation. In such cases
several geostatistical tools may warrant careful consideration,
particularly geostatistical simulation (Chiles and Delfiner, 1999).
Another possibility is to use indicator geostatistics to estimate
the probability of exceeding some threshold (Goovaerts, 1999).

Project Evaluation Conference

The authors wish to thank Geoff Booth from BHP Billiton, Chris
Willows of Rio Tinto and an anonymous peer reviewer for their
constructive reviews of earlier versions of this paper and our
colleagues at Quantitative Group for their help in developing
concepts. While all this assistance is acknowledged, the authors,
of course, take full responsibility for the viewpoints expressed
herein.

REFERENCES
Bieniawski, Z T, 1989, Engineering Rock Mass Classifications (John
Wiley and Sons: New York).
Chiles, J and Delfiner, P, 1999, Geostatistics Modelling Spatial
Uncertainty, 659 p (John Wiley and Sons: New York).

Melbourne, Vic, 19 - 20 June 2007

S DUNHAM and J VANN

Goovaerts, P, 1997, Geostatistics for Natural Resources


Characterization, 496 p (Oxford University Press: New York).
Hall, B E and Stewart, C A, 2004. Optimising the strategic mine plan
Methodologies, findings, successes and failures, in Proceedings
Orebody Modelling and Strategic Mine Planning, pp 49-58 (The
Australasian Institute of Mining and Metallurgy: Melbourne).
Journel, A G and Huijbregts, C, 1978. Mining Geostatistics, 599 p
(Academic Press: London).

Lane, K F, 1988. The Economic Definition of Ore, 147 p (Mining Journal


Books Ltd: London).
Vann, J, Jackson, S and Bertoli, O, 2003, Quantitative kriging
neighbourhood analysis for the mining geologist A description of
the method with worked examples, in Proceedings 5th International
Mining Geology Conference, pp 215-223 (The Australasian Institute
of Mining and Metallurgy: Melbourne).

Melbourne, Vic, 19 - 20 June 2007

Project Evaluation Conference

You might also like