Professional Documents
Culture Documents
基于不确定性的露天采矿边界品位策略选择多准则排序系统
基于不确定性的露天采矿边界品位策略选择多准则排序系统
Resources Policy
journal homepage: www.elsevier.com/locate/resourpol
a r t i c l e i n f o a b s t r a c t
Article history: Cut-off grade strategy (COGS) is a concept that directly influences the financial, technical, economical,
Received 7 December 2011 and environmental issues in relation to the exploitation of a mineral resource. Despite the simple
Received in revised form definition of cut-off grade, the COGS problem is one of the complex and complicated problems in the
26 October 2012
mine planning process. From the optimization point of view, the COGS with an objective of maximizing
Accepted 22 January 2013
the present value of future cash flows is a non-linear and a non-convex problem that even in its
Available online 15 March 2013
deterministic form can be solved using approximate optimization methods. This optimization problem
JEL classifications: will also be more complex and complicated under uncertainty conditions. This paper proposes an
D81 uncertainty based multi-criteria ranking system to investigate the problem of COGS selection
C63
considering metal price and geological uncertainties. The proposed system aims at selection of the
G13
best COGS among technically feasible alternative COGSs under uncertainty circumstances. Our
L72
Q32 developed system is based on integrating metal price and geological uncertainties as well as operating
flexibility to close the mine early. We incorporate this operating flexibility into the proposed system
Keywords: using a Monte Carlo based real options (RO) valuation model. For this purpose, in addition to the
Open pit mining expected value, other risk criteria are considered to rank the alternatives. These risk criteria include
Cut-off grade strategy (COGS) abilities of strategies in producing extra profits, minimizing losses, and achieving the predefined goals
Least-Squares Monte Carlo real options of the production. In this study, the technically possible COGSs are generated using the Lane
valuation comprehensive algorithm. To demonstrate the effectiveness of the proposed system, we utilize data
Metal price uncertainty of an Iranian gold mine. Results show that the proposed system outperforms conventional methods in
Geological uncertainty
the sense that it shows significantly lower average mis-ranking than the other methods and also selects
Multi-criteria ranking system
a strategy with a higher value. The sensitivity analysis of the proposed system relative to the gold price
shows that the system is highly dependent on the parameters of the stochastic process used to model
the evolution of the metal price. Therefore, special consideration should be given in estimating
stochastic process parameters.
& 2013 Elsevier Ltd. All rights reserved.
0301-4207/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.resourpol.2013.01.004
Y. Azimi et al. / Resources Policy 38 (2013) 212–223 213
Describing the solution of the above-mentioned questions, as Recently, according to the new developments in the simula-
discussed thoroughly in the literature (Dagdelen, 2007; Osanloo tion based RO valuation, researchers have attempted to develop
et al., 2008), is out of the scope of this paper. However, the multi-criteria ranking systems with the aims of minimum sub-
solution for the third question carries the same solution for the jective judgments, sound quantification of uncertainties and
COGS optimization problem. Conventionally, planning is done in a valuation of management flexibilities for appraisal of investments
circular manner as shown in Fig. 1. Hence, any wrong COGS will in minerals industry (Abdel Sabour et al., 2008). To select the best
certainly lead to wrong planning, inefficient performance of alternative in such a system, besides the expected value of
different stages of the mining operation, and considerable losses. alternatives other economical and operational risk indicators
Traditionally, a constant price trend and a grade tonnage curve should also be considered in the evaluation process.
estimated from the kriged grades are used in open pit mine One of the earliest studies on incorporating uncertainties in
planning. But the real situation is much more complex, where calculation of the COGS is the work by Dowd (1976). He devel-
mining investments are often associated with diverse sources of oped a stochastic dynamic programming model in which price
both endogenous and exogenous uncertainties. Recently, much uncertainty was modeled as a stochastic Markov process. Despite
research has tried to address the issue of geological and market some operational simplifications considered in regard with con-
related uncertainties in the mining industry where their adverse centrating and refining plant capacities, it seems that Dowd’s
effects could cause substantial losses of millions of dollars model has the potential to be extended to a more general model.
(Osanloo et al., 2008). Krautkraemer (1988) studied the optimal response of the COGS
In this respect, some outstanding work has been done incorpor- based on anticipated and unanticipated price changes using
ating geological uncertainty in long term mine planning using optimal control theory. He provided some useful theoretical
simulation based optimization and stochastic programming meth- discussions, but his model is not general and does not seem
ods (Dimitrakopoulos and Abdel Sabour, 2007; Gholamnejad and applicable in a real life mining conditions. Recently, McIsaac
Osanloo, 2007; Osanloo et al., 2008). While market related uncer- (2008) and Thompson (2010) employed golden search optimiza-
tainties in mining investments are resolved employing the modern tion inside simulation based models considering metal price
stochastic valuation method, known as Real Option (RO) valuation. uncertainty to find a robust fixed cut-off grade and a fixed mining
The RO valuation is an alternative approach to the conventional rate for the whole life of the mining operation in metalliferous
Discounted Cash Flow (DCF) model. Brennan and Schwartz (1985) underground and open pit mining, respectively. However,
applied the RO analysis in the mineral industry for the first time, and employing a fixed cut-off grade throughout the whole mining life
since then it has been extended by researchers considering various time will certainly lead to non-optimal operations and misuse of
types of flexibilities and employing different analytical and numer- mining, milling, and refining capacities.
ical methods. Researchers have confirmed that the RO valuation The contribution of this paper is to propose a multi-criteria
performs better than the DCF approach under uncertainty condi- ranking system to select an optimum COGS, incorporating geolo-
tions. It also provides a basis for developing flexible mine plans by gical and metal price uncertainties while considering the operat-
valuing the strategic and managerial flexibilities in response to ing flexibility to abandon the mine early under adverse economic
uncertain market conditions (Mardones, 1993; Trigeorgis, 1996; conditions. In this regard, the Lane comprehensive algorithm
Frimpong and Whiting, 1997; Samis and Poulin, 1998; Abdel (Lane, 1988) is used to generate technically feasible alternative
Sabour and Poulin, 2006; Samis et al., 2006; Dimitrakopoulos and COGSs. The price and geological uncertainties are quantified using
Abdel Sabour, 2007; Cortazar et al., 2008; Akbari et al., 2009). the stochastic process and geostatistical simulations, respectively.
Despite the usefulness of these methods, they suffer from The Least Squares Monte Carlo (LSM) RO valuation model is
some important shortcomings. These shortcomings include mak- applied to valuate alternative strategies. Also, the usefulness of
ing subjective judgments in the decision making process, con- the proposed methodology and its sensitivity to the estimated
sidering constant market conditions, and the weakness of the parameters of the price stochastic process are investigated. In the
conventional DCF approach in valuing managerial flexibilities next sections, the mechanism of the employed ranking system
inherent in the mineral industry. Although the RO valuation is adapted for selection of the COGS is outlined and applied to an
able to value managerial and operating flexibilities, like conven- Iranian gold mine under price and geological uncertainties.
tional methods, it does not necessarily prescribe the best alter-
native among the available alternatives and the results are often
interpreted subjectively. Methodology
Fig. 2. The proposed multi-criteria ranking system for selection of COGS for open pit mining.
Monte Carlo which is used to incorporate the value of operating naturally respected confidence level according to the probability
flexibility under multiple uncertainties. The method is mainly function at each time interval (Dixit and Pindyck, 1994).
based on simulating multiple realizations of the uncertain vari- As assumed by Schwartz (1997), the gold price P evolves
ables and making the optimal decision at each period using value according to the GBM process. According to the lognormal
expectations; (5) applying a multi-criteria ranking system to distribution, future prices for Pt can be generated as follows:
select the best COGS. The last step employs statistics of cash flow pffiffi
Pt ¼ P 0 expfða0:5s2 Þt þ Nð0,1Þs t g ð3Þ
distribution functions including upside potential, downside
potential, value statistics and probability of completion. The where N(0,1) is the standard normal distribution random variable.
following sections provide the theoretical basis for the proposed
methodology. Modeling the geological uncertainty
Modeling the price uncertainty During the past two decades, conditional simulation has been
used as a complementary tool for modeling of geological uncer-
There are many stochastic processes that can be used in tainty in the strategic mine planning and design (Osanloo et al.,
quantifying market uncertainty of commodity prices. The most 2008). The main idea behind this technique is the use of
commonly used processes are the Geometric Brownian Motion conditioning data to produce equally probable realizations of
(GBM) as shown in Eq. (1) and the Mean Reverting Process (MRP) the in-situ ore-body based on the borehole data. The geostatistical
as shown in Eq. (2) (Dixit and Pindyck, 1994): simulation is a generalization of the concepts of Monte Carlo
simulation to include three-dimensional spatial correlation to
dP ¼ aP dt þ sP dz ð1Þ
ensure the following: (1) the simulated values of each variable
are the same as the measured values of those variables at sampled
dP ¼ ZðlnPln PÞP dt þ sP dz ð2Þ
locations; (2) all simulated values of a given variable have the
where P is the price of a metal, a is the expected trend, s is the same spatial relationships as observed in the data values, i.e.,
standard deviation, dz is an increment in a standard wiener spatial correlation; (3) all simulated values of any pair of variables
process, dt is an increment of time and Z is the speed at which have the same spatial interrelationships as observed in the data
the price reverts to its long term equilibrium level of P. Generally, values, i.e., spatial cross-correlation; and (4) histograms of the
the GBM is suitable for variables that exhibit a constant trend, simulated values of all variables are the same as those observed
such as precious metal prices and security prices while the MRP is for the data (Dowd, 1997).
appropriate for modeling variables that have a long term equili- Several conditional simulation algorithms are available in the
brium level, such as base metal prices (Schwartz, 1997). The literature. This research employs the well published Sequential
advantage of stochastic process modeling is that the selection of Gaussian Simulation (SGS) to generate in-situ ore-body realiza-
appropriate process and calculating corresponding parameters tions. Therefore, Datamine software that benefits the GSLIB codes
are based on analysis of the historical market prices of a metal. for SGS is used to generate equally probable realizations of the
After selecting the appropriate model and its parameters, in the in-situ ore-body (Datamine, 2007). More details about the SGS
simulation process it is important to generate sample paths can be found in Deustech and Journel (1998).
according to the Markov property (Dixit and Pindyck, 1994), in
which only the current information is sufficient to forecast the Generating alternative feasible COGSs
future value of a variable. Simulating price paths based on this
property enhances time connectivity, reduces unrealistic shifts The Lane’s original model based on Bellman dynamic pro-
throughout the single Markov and simultaneously guarantees the gramming was published on 1964, and then it was extended to
Y. Azimi et al. / Resources Policy 38 (2013) 212–223 215
a more comprehensive model in 1989, which considered the valuation technique in real capital investment and LSM is one of
effects of economic parameters variations on the COGS. The Lane’s the promising methods in this regard.
model results in making the best use of capacities of mining Longstaff and Schwartz (2001) assumed that American options
operation to maximize the present value. Many researchers have can be exercised at discrete early exercise dates. The time to maturity
used this model as the basis for their work (Whittle and Wharton, T is divided into K discrete times (0ot1 rt2 ryrtK ¼T) and N
1995; Ataei and Osanloo, 2003, 2004; Osanloo and Ataei, 2003; simulated paths are generated under the risk-neutral measure. Like in
Bascetin, 2007; Rashidinejad et al., 2008), while others employed any American option valuation procedure, the optimal exercise policy
different operational research methods to solve the COGS decision at any point in time tK is obtained as the maximum of the
problem (Akaike and Dagdelen, 1999; Cetin and Dowd, 2002; immediate exercise value of I(o; tK) and the expected continuation
Cairns and Shinkuma, 2003; Azimi and Osanloo, 2011; Azimi value of F(o; tK). Given that the expected continuation value depends
et al., 2011). on future outcomes, the procedure must work its way backwards,
As in Lane (1988), suppose that for a resource with amount of starting from the end of the time horizon T. The put option payoff
R, there is a small decrement of r as extraction, and then the cash C(o; tK) can be interpreted as the Bellman equation of dynamic
flow of c per unit of extracted resource can be computed while the programming as follow:
strategy o is given to exploit r during the time t. According to the Cðo; t k Þ ¼ max½Iðo,t k Þ,Fðo,t k Þ ð5Þ
Lane’s algorithm (1988) Eq. (4) is the fundamental equation for
determining the optimum COGS: The immediate exercise value of I(o; tK) along the path o is
simply the difference between the exercise price of H and the
dV dV stock price of S(o; tK). At the expiration date along the sample
¼ Maxo fctF g and F ¼ dV þ ð4Þ
dR dT path o the expected continuation value is as follows:
where d is the discount rate, t ¼t/r is the time taken to work Cðo; t k ¼ TÞ ¼ max½ðHSðo,TÞÞ,0 ð6Þ
through one unit of the resource, V is the maximum present value
But at any early exercise time tK oT, assuming that the option
of a resource at a time (T), dV/dR is the economic value added by
is not exercised until after tK and the option holder following the
mining and processing of one unit of resource, and F is the
optimal stopping strategy for all x, tK oxrT, under the risk-
opportunity cost. The optimum cut-off grade at each period is
neutral pricing measure Q, the expected continuation value of
determined by maximizing Eq. (4) with respect to o, and
F(o; tK) based on the information b at time tK is as follows:
considering the capacity limiting constraints of the three main 2 3
stages of mining operation, namely mining, milling and refining X
k Z tj
or marketing capacities. Accordingly, the optimization problem in Fðo; t k Þ ¼ EQ 4 exp rðo,xÞdx Cðo,t j ; t k ,T9btk 5 ð7Þ
j ¼ kþ1 tk
Eq. (4) is divided into six sub-problems and the optimum cut-off
grade at each period is the one that provides the greatest change where C(w,tj; tK,T) is the remaining cash flows and r(o,x) is the
in the present value of the mining operation. Finally, the set of risk-free discount rate. The main contribution of the LSM method
selected optimum cut-off grades for the whole mining lifespan is is to estimate the expected conditional continuation value at tK 1,
known as the optimum COGS. As can be seen from Eq. (4), there is tK 2,y, t1 by regressing the discounted future option values on a
a circular dependency between F and PV that is handled by linear combination of M basis functions Lj (S) such as Laguerre
considering two streams of exploitation policies of V and W. It polynomials, trigonometric series, or simple powers of the stock
is supposed that the second stream (W) commences at the end of price S such as
the first period of the stream V. Fig. 3 shows the flowchart of the
Lane’s (1988) algorithm used to develop a Matlab based code to X
M
Fðo; t k Þ ¼ aj Lj ðSÞ ð8Þ
generate feasible COGSs. More details on Lane’s algorithm can be j¼0
found in Lane (1988) and Azimi (2012).
After the regression, the optimal obtained coefficients aj are
used to estimate the expected continuation value Fð ^ o; t k Þ. For all
Simulation based RO valuation in-the-money paths at time tK, the optimal exercise policy is
obtained by choosing the maximum between the immediate
Generally, the value of an option can be estimated using either exercise value and the estimated continuation value at time tK.
analytical or numerical methods. Analytical methods are only Consequently, the cash flows along the path o from time tK to
suitable for valuing simple options as the one developed by Black time T are revised. For out-of-the-money paths, the option payoff
and Scholes (1973). But, numerical methods started to generate is zero. The recursion proceeds backward until time t1. After the
greater interest in assessing complex options such as American exercise decisions are determined at each exercise date along all
type options with multiple uncertain state variables. The most the paths, the value of the option is estimated by discounting the
widely known numerical techniques are finite difference cash flows to time zero at the risk-free rate and averaging over
(Schwartz, 1977), binomial lattice (Cox et al., 1979), and Monte the number of paths, N.
Carlo simulation (Boyle, 1997). Amongst them, the Monte Carlo Abdel Sabour and Poulin (2006) extended the LSM to value
technique has superior performance in dealing with dimension- mining investments under multiple market uncertainties, which
ality difficulties encountered in the problems employing multiple was further developed by Abdel Sabour et al. (2008) to evaluate
and complex uncertain state variables. the mining extraction scheduling under both multiple market and
Monte Carlo simulation was originally proposed by Boyle (1997) geological uncertainties. Cortazar et al. (2008) have also extended
for pricing of European options as a forward intuition technique and the approach to cover the three-factor risk models which is more
extended for pricing of American options by combining the simplicity suitable for long-term commodity real options. The extended LSM
of forward induction with the ability of backward induction in method is mainly based on simulating multiple realizations of the
determining the optimal option exercise policy. In this regard, uncertain variables and making the optimal decision at each
Longstaff and Schwartz (2001) developed a promising method of period using value expectations.
Least Squares Monte Carlo (LSM) simulation for valuing path depen- In this study, the operating flexibility to close the mine early is
dant American options. Recently, the Monte Carlo RO valuation integrated into the value estimate of the COGS taking into account
method has also been introduced as an alternative to the DCF the geological and metal price uncertainties. In order to evaluate
216 Y. Azimi et al. / Resources Policy 38 (2013) 212–223
the mining operations under the flexible operating mode, it is Then, recursively starting from the last time step towards the
assumed that the mining operation status can be changed at some first period at any period t (between zero and T), a status that
certain discrete dates per year with a time step of Dt apart. maximizes the mining expected value is chosen as the optimum
Conditional to N simulated metal price paths according to the operation policy conditional to a sample price path of nAN and
Markov property and M ore-body models, the cash flows at each ore-body mAM. Due to the computational complexities, the price
discrete date are defined based on the amount of production, the realizations, N is much larger than the number of ore-body
metal price and the unit production cost. realizations M. Therefore, to have N possible grades and tonnages
Y. Azimi et al. / Resources Policy 38 (2013) 212–223 217
for each strategy, N independent draws are generated randomly where Ka is the abandonment cost. This optimization process is
from the M simulated ore-body models. carried out for all simulated realizations of metal prices through-
Let CVO denote the continuation value of ‘‘open mining’’ status. out all time periods. After revising the cash flows according to the
As described in Longstaff and Schwartz (2001), CVO is a function of determined optimum operating policy at the discrete exercise
the uncertain state variable and can be estimated as the linear points throughout all simulated metal price paths, the strategy
combination of basis functions of uncertain variables. Given that P value at time zero is determined by discounting the cash flows at
denotes the price of the metal and G denotes ore-body character- the risk-free rate and averaging over the number of simulated
istics (i.e., tonnage and grade) simulated by the ore-body model metal price paths. Besides the expected value, various statistics
m, the expected continuation value of mining is approximated as for the overall value as well as annual cash flows of mining
strategy value can be generated throughout this analysis. These
X
N
EðCV O 9n,m,tÞ ¼ ai Li ðP n,t ,Gm,t Þ ð9Þ statistics will be employed within a multi-criteria ranking system
i¼0 in order to select COGS under geological and price uncertainties.
The flowchart of the LSM algorithm used in this paper is shown
The parameters of these functions are estimated at each time
in Fig. 4.
period by regressing the sum of the discounted value of cash
flows beyond time t onto their basis functions. After estimating
the basis functions parameters, the expected mine value at the Multi-criteria ranking system
time t for each simulated sample path n is approximated as the
sum of the expected continuation value and the cash flow of Multi-criteria ranking system is an effective tool enhancing the
mining operation CF(n,m,t) during the time period t. role of objective interpretations against subjective interpretations
when a decision making problem exists. In this paper, based on
EðV9n,m,tÞ ¼ EðCV O 9n,m,tÞ þ CFðn,m,tÞ ð10Þ
the available information at the initial planning time, the following
At any time t based on the expected mine value (E(V)), the aspects are considered to select the best exploitation strategy.
decision whether to keep the mine open or to close it early is
taken as follows:
Keep the mine open if (1) Upside potential that indicates the ability of alternative
strategies to gain possibly more profit than those expected
EðV9n,m,tÞ 4 K a ð11Þ if outcomes were favorable.
Close the mine if (2) Downside risk that reflects the difference between strategies
in minimizing negative cash flows risk throughout the mine
EðV9n,m,tÞ r K a ð12Þ life time.
Fig. 4. Flowchart of the LSM algorithm considering managerial flexibility to close the mine early under metal price and geological uncertainties.
218 Y. Azimi et al. / Resources Policy 38 (2013) 212–223
(3) Statistics of the estimated operation values which includes the design is above or below that average. The indicator related to the
average, lower and upper limits at a certain confidence level. expected value for a design m, EVIm is calculated as follows:
(4) Probability of completion, which is the probability that the ! !
XM X
M
mine will operate as planned throughout its planned life. EVIm ¼ 100 EV m EV m =M = EV m =M ð17Þ
m¼1 m¼1
For simplicity, these criteria are considered to have equal The upper limit indicator (ULIm) of the expected value at a
weight in this paper while different weight can be acquired based predefined confidence level is
on the mining company policies. The method of calculating every ! !
individual corresponding indicator is explained in details as XM XM
ULIm ¼ 100 ULm ULm =M = EV m =M ð18Þ
follows. m¼1 m¼1
Table 1 Table 2
Economical and technical information for the Iranian gold mine. Economic and stochastic process parameters for copper mine valuation.
Costs Mining cost ($/ton) 2.6 Stochastic process parameters Gold price and valuation parameters
Milling cost ($/ton) 14
Refinery ($/ton) 1,000,000 P0 (ton/$) 18,991,110.33
Fixed cost ($/year) 2,000,000 a (%) 7
s (%) 20
Capacities Mining throughput (ton/year) 250,000
Abandonment cost ($) 0
Milling throughput (ton/year) 150,000
Risk free interest rate (%) 10
Marketing throughput (kg/year) 300
Fig. 5. 3D views of four simulated ore-body models of the considered part of the gold reserve. The simulation models were generated using the Sequential Gaussian
Simulation (SGS).
Table 3
The first 10 high rank strategies using the mentioned four ranking methods.
a
Real options valuation.
b
Expected value of present value ($).
c
Total ranking indicator.
Table 4 study has a higher value than the best strategy proposed by the
The first 10 high rank strategies based on their PV using actual data. other ranking systems.
To have a comparison of overall performance of the ranking
Rank Strategy PV ($) systems, a term of average mis-ranking (amr) was calculated.
1 141 2,893,768.74
The amr is defined as the average value of the absolute differences
2 80 2,892,505.25 of alternative ranks in a specified ranking system and their
3 59 2,892,146.00 corresponding actual rank. The amr can be calculated using the
4 138 2,891,553.46 following formula:
5 4 2,891,553.46
6 5 2,891,513.15
!
X
S
7 10 2,891,191.41 amr ¼ 9Ract i Ri 9 =S ð23Þ
8 69 2,891,191.41 i¼1
9 135 2,890,842.20
10 6 2,890,622.79 where Ract i and Ri are ranks of the ith strategy estimated by the
actual ranking and one of the mentioned ranking systems,
respectively. S is the total number of the possible strategies.
Table 5 A higher amr indicates that the method can result in a misleading
The amr of the four ranking systems. decision while a lower amr suggests an efficient performance for
the method. The amr was calculated for each of the four ranking
Ranking methods E (DCF) TRI (DCF) E (RO) TRI (RO)
systems and the results are shown in Table 5.
Average mis-ranking (amr) 72.15 80.15 45.96 31.82
Evaluating COGSs based on the RO expected values reduced
the amr from 72.15 in the NPV analysis to 45.96. This amr has
been reduced further to 31.82 when using the RO valuation based
10 strategy with the highest actual PV. As shown in Table 4, indicator that ranks COGSs using multiple risk criteria while
strategy 138 is the fourth strategy with the highest PV. But none considering the flexibility to close the mine early. The results of
of the best strategies selected by the other methods are among Tables 4 and 5 reveal that under uncertainty circumstances, the
the 10 actual best strategies among the 300 alternatives. There- COGSs real values using the actual data can be significantly
fore, the strategy selected by the proposed ranking system in this different from those estimated at the planning time. As shown
Y. Azimi et al. / Resources Policy 38 (2013) 212–223 221
in Table 3, the values of the strategies using the DCF model are necessarily have a declining trend as described by Lane (1988).
negative while these strategies all have positive PV. Furthermore, the selected COGS by the proposed ranking system
The estimated distribution function of cash flows is usually has more present value than the three other conventional ranking
assumed to be approximately symmetrical around the expected methods.
value when no operating flexibility is considered in the valuation.
Therefore, integrating multi-criteria ranking systems without
considering the operating flexibility to react to the new informa-
tion will not necessarily assure improvement in the decision Sensitivity analysis of the proposed system on the parameters
making process. As is shown in Table 5, application of the DCF of the GBM process
based multi-criteria ranking system does not improve the selec-
tion process for the case studied here. A sensitivity analysis was performed to measure the depen-
On the contrary, using the RO valuation instead of the dency of the proposed system on the estimated parameters of the
conventional DCF technique can improve the economic based GBM stochastic process. Therefore, 30 COGS alternatives were
decision process. The results show that integrating multiple risk generated and used to analyze the sensitivity of the proposed
criteria with the flexibility to revise the originally taken decisions system to the drift and volatility of the GBM process. In this
regarding the early closure of the mine can significantly change respect, the amr of the RO based multi-criteria ranking system
the decision and improve the selection process. However, in was calculated relative to the actual gold prices from 2006 to
addition to the higher value of the selected alternative and lower 2010 and actual grade tonnage curve. To consider the stochastic
mis-rank of the proposed system overall performance, the RO
valuation used by the proposed system estimates the values of x 107
the strategies better than the DCF model. Some of the results on 4.5
the ranking systems performance are in agreement with the
results of the research published recently by Abdel Sabour et al.
(2008). 4
The COGS 138, as the best strategy selected by our proposed
ranking system, is illustrated graphically in Fig. 6. It is seen that Actual gold price
under the uncertainty conditions the selected COGS does not 3.5
Gold price ($/t)
1.5
Hypothetical gold price 2
1
0 1 2 3 4
Year
Fig. 6. COGS 138, selected as the best strategy using the proposed system. Fig. 8. Hypothetical actual gold price.
9 5.5
Run 1 Run 1
8
Run 2 5 Run 2
Average mis−ranking (amr)
Average mis−ranking (amr)
Run 3 Run 3
7
4.5 Run 4
Run 4
6
Run 5 Run 5
4
5
3.5
4
3
3
2 2.5
0.05 0.1 0.15 0.2 0.25 0.3 0.05 0.1 0.15 0.2 0.25 0.3
Drift Volatility
Fig. 7. Sensitivity of the proposed ranking system to the estimated values of the (a) drift and (b) volatility.
222 Y. Azimi et al. / Resources Policy 38 (2013) 212–223
12
11
10
Average mis−ranking (amr) 10
8 8
7
Hypothetical gold price 1 Hypothetical gold price 1
6
6
2 3
0.05 0.1 0.15 0.2 0.25 0.3 0.05 0.1 0.15 0.2 0.25 0.3
Drift Volatility
Fig. 9. The effect of the actual gold price trend on the response of the proposed system performance to the variation of GBM process parameters (a) drift (b) volatility.
nature of the generated price paths this analysis was repeated five prefer the proposed methods over the other methods, and more
times, and the results are shown in Fig. 7. investigation is needed.
Fig. 7 shows that as the drift and volatility of the GBM process
is increased, the amr of the proposed ranking system will be
decreased. This improvement in the performance of the proposed Conclusion
system could be due to the higher trend of the gold price from
2006 to 2010. The gold price is modeled in a more representative Cut-off grade is a geological/technical concept that embodies
manner by increasing the parameters of the GBM process. There- important economic aspects of the mining venture and plays an
fore, as closely as the stochastically generated price paths resem- influential role in the optimality of mine planning and design.
ble the real condition, the result of the ranking system will be In this paper, a multi-criteria ranking system was proposed for
similar to the real condition. However it is impossible to have a selection of COGS cut-off grade strategy (COGS) for an open pit
ranking system with zero mis-ranking. mining operation under metal price and geological uncertainties.
In justification of the stated reason, two other hypothetical The proposed ranking system is based on the multiple risk and
gold price trends were also considered as depicted in Fig. 8. cash flow characteristics measures of each COGS. In the proposed
Considering the hypothetical gold prices as the actual gold prices, system, alternative COGSs were valuated using a simulation
the performance of the proposed ranking system was assessed based real option (RO) valuation model. The RO valuation benefits
against the variation of drift and volatility of the GBM process, the proposed system with the valuation of operating flexibilities
and the results are shown in Fig. 9. inherent in the mineral industry and consequently offers better
As Fig. 9 shows, for the first hypothetical trend in which gold evaluation of different alternatives. Hence in this study, to avoid
prices had an increasing trend, the variation of the amr is the unwanted losses in unfavorable market conditions, the insuring
same as described for the actual gold prices from 2006 to 2010. operating flexibility to close the mine early was considered in the
But for the second hypothetical gold price showing a decreasing valuation. In this study, the Lane’s comprehensive algorithm was
trend with time, the amr of the proposed ranking system applied to generate the technically feasible COGSs while metal
increases by an increase in the parameters of the GBM process prices and geological uncertainties were resolved using stochastic
(drift and volatility). This result obtained since majority of the processes and geostatistical simulation, respectively.
simulated price paths may not be a good representative of the To explore the advantages of the proposed system, it was
considered decreasing actual price trend. applied to select the COGS under uncertainties for an Iranian gold
Therefore, according to the results, for a gold mining operation mine along with three other ranking methods including the RO
if the market condition is expected to be improved in the future; valuation expected value, the Discounted Cash Flow (DCF) valua-
considering higher drift and volatility for the GBM process may tion expected value and the DCF based multiple criteria ranking
result in improvement of the decision process and vice versa. system. Comparing the performance of the ranking methods with
Nevertheless, it is impossible to exactly know whether the price the actual rank of alternatives calculated using actual values of
will be increased or decreased or both in the future. Hence, uncertain variables revealed that the average mis-rank of the
considering the high sensitivity of the valuation results and proposed system is lower than the other three ranking methods.
proposed ranking method to the estimated parameters of the The results also suggest that selection procedure may be
chosen stochastic process, it is recommended that a professional improved when incorporating uncertainty and operating flexibil-
econometrician be retained to parameterize metal price processes ity. It was seen that under the uncertainty conditions, the selected
for any valuation exercise. COGS would not necessarily have a declining trend as described
Finally it should be mentioned that the results of this study are by Lane (1988). Furthermore, the results of the sensitivity analysis
specific to the case study discussed herein and the data fed into showed that the integrity of the proposed system is highly
the valuation model. Different results might be obtained if other dependent on the parameters of the price stochastic process.
economic condition or other operating flexibilities were consid- It also concluded that when the market prices are expected to
ered in valuation. Therefore, the results cannot be generalized to increase in the future, using higher drift and volatility may
Y. Azimi et al. / Resources Policy 38 (2013) 212–223 223
improve the ranking results and vice versa. Further extensions Dowd, P.A., 1997. Risk in minerals projects analysis, perception and management.
could be included considering more risk analysis criteria as well Mining Technology: IMM Transactions Section A 106, 9–18.
Dowd, P.A., 1976. Application of dynamic and stochastic programming to optimize
as more operating flexibilities. At last, the proposed system cutoff grades and production rates. Mining Technology: IMM Transactions
suggests a framework that would enhance the quality of the Section A 85, 22–31.
decision making process in mineral recourse industry by providing a Dagdelen, K., 2007. Open pit optimization—strategies for improving economics of
mining projects through mine planning. In: Dimitrakopoulos, R. (Ed.), Ore-Body
more informative and certain platform for project evaluation.
Modelling and Strategic Mine Planning. AusIMM, Perth, Western Australia,
pp. 145–148.
References Dagdelen, K., 1992. Cut-off grade optimization. In: Proceedings of the 23th
APCOM. Littleton, Colorado, USA, pp. 157–165.
Deustech, C.V., Journel, A.G., 1998. GSLIB: Geostatistical Software Library and
Abdel Sabour, S.A., Dimitrakopoulos, R., Kumral, M., 2008. Mine plan selection User’s Guide, 2nd edn. Oxford University Press, New York.
under uncertainty. Mining Technology: IMM Transactions Section A 117 (2), Dimitrakopoulos, R., Abdel Sabour, S.A., 2007. Evaluating mine plans under
53–64. uncertainty: can the real options make a difference? Resources Policy 32 (3),
Abdel Sabour, S.A., Poulin, R., 2006. Valuing real capital investments using the 116–125.
least-squares Monte Carlo method. Engineering Economist 51 (2), 141–160. Dixit, A.K., Pindyck, R.S., 1994. Investment Under Uncertainty. Princeton University
Akaike, A., Dagdelen, K., 1999. A strategic production scheduling method for an Press, Princeton.
open pit mine. In: Proceedings of the 28th APCOM. Golden Colorado, USA, Frimpong, S., Whiting, J.M., 1997. Derivative mine valuation: strategic investment
pp. 729–738. decisions in competitive markets. Resources Policy 23 (4), 163–171.
Akbari, A.D., Osanloo, M., Shirazi, M.A., 2009. Minable reserve estimation while Gholamnejad, J., Osanloo, M., 2007. Using chance-constrained binary integer
determining ultimate pit limits (UPL) under price uncertainty by real option programming in optimizing long term production scheduling for open-pit
approach (ROA). Archives of Mining Sciences 54 (2), 321–339.
mine design. Mining Technology: IMM Transactions Section A 117 (2), 58–66.
Ataei, M., Osanloo, M., 2004. Using a combination of genetic algorithm and the grid
Krautkraemer, J.A., 1988. The cut-off grade and the theory of extraction. Canadian
search method to determine optimum cutoff grades of multiple metal
Journal of Economics 21 (1), 146–160.
deposits. International Journal of Surface MiningReclamation and Environ-
Lane, K.F., 1988. The Economic Definition of Ore: Cutoff Grades in Theory and
ment (IJSM) 18 (1), 60–78.
Practice. Mining Journal Books Ltd, London.
Ataei, M., Osanloo, M., 2003. Determination of optimum cutoff grades of multiple
Longstaff, F.A., Schwartz, E.S., 2001. Valuing American options by simulation: a
metal deposits by using golden section search. Journal of South African
simple least-squares approach. Review of Financial Studies 14 (1), 113–147.
Institute of Mining and Metallurgy (SAIMM) 103 (8), 493–500.
Mardones, J., 1993. Option valuation of real assets: application to a copper mine
Azimi, Y., 2012. Modeling of Optimum Cut-off grades in Single Metal Deposits for
with operating flexibility. Resources Policy 19, 51–65.
OpenPit Mines Considering Uncertainty of Price and Grade. Ph.D. Thesis.
McIsaac, G., 2008. Strategic Design of an Underground Mine Under Conditions of
Amirkabir University of Technology, Iran.
Metal Price Uncertainty. Dissertation. Queen’s University, Kingston, Ontario,
Azimi, Y., Osanloo, M., 2011. Determination of open pit mining cut-off grade
Canada.
strategy using combination of nonlinear programming and genetic algorithm.
Osanloo, M., Gholamnejad, J., Karimi, B., 2008. Long-term open pit mine production
Archive of Mines 56 (2), 189–212.
Azimi, Y., Osanloo, M., Esfahanipour, A., 2011. Optimisation of mining policy under planning: a review of models and algorithms. International Journal of Mining,
different economical conditions using a combination of non-linear program- Reclamation and Environment 22 (1), 3–35.
ming and genetic algorithm. In: Proceedings of the 35th APCOM. Wollongong, Osanloo, M., Ataei, M., 2003. Using equivalent grade factors to find the optimum
Australia, pp. 501–516. cutoff grades of multiple metal deposits. Minerals Engineering 16 (8), 771–776.
Bascetin, A., 2007. Determination of optimal cut-off grade policy to optimize NPV Rashidinejad, F., Osanloo, M., Rezai, B., 2008. Incorporating environmental issues
using a new approach with optimization factor. Journal of South African into optimum cut-off grades modeling at porphyry copper deposits. Resources
Institute of Mining and Metallurgy 107 (2), 87–94. Policy 33, 222–229.
Black, F., Scholes, M., 1973. The pricing of options and corporate liabilities. Journal Samis, M., Davis, G.A., Laughton, D., Poulin, R., 2006. Valuing uncertain asset cash
of Political Economy 81 (3), 637–654. flows when there are no options: a real options approach. Resources Policy 30
Boyle, P., 1997. Options: a Monte Carlo approach. Journal of Financial Economics 4, (4), 285–298.
323–338. Samis, M., Poulin, R., 1998. Valuing management flexibility: a basis to compare the
Brennan, M.J., Schwartz, E.S., 1985. Evaluating natural resource investments. standard DCF and MAP frameworks. CIM Bull 91 (1019), 69–74.
Journal of Business 58 (2), 135–157. Schwartz, E.S., 1997. The stochastic behaviour of commodity prices: implications
Cairns, R.D., Shinkuma, T., 2003. The choice of the cutoff grade in mining. for valuation and hedging. Journal of Finance 52 (3), 923–973.
Resources Policy 29, 75–81. Schwartz, E.S., 1977. The valuation of warrants: Implementing a new approach.
Cetin, E., Dowd, P.A., 2002. The use of genetic algorithms for multiple cutoff grade Journal of Financial Economics 4 (1–2), 79–93.
optimization. In: Proceedings of the 30th APCOM. Littleton, Colorado, USA, Taylor, H.K., 1972. General background theory of cutoff grades. Mining Technol-
pp. 769–780. ogy: IMM Transactions Section A 81, 160–179.
Cortazar, G., Gravet, M., Urzua, J., 2008. The valuation of multidimensional Thompson, J., 2010. Test of an Innovative Stochastic Design System on an Open Pit.
American real options using the LSM simulation method. Computers & Dissertation. Queen’s University, Kingston, Ontario, Canada.
Operations Research 35 (1), 113–129. Trigeorgis, L., 1996. Real Options: Managerial Flexibility and Strategy in Resource
Cox, J.C., Ross, S.A., Rubinstein, M., 1979. Option pricing: a simplified approach. Allocation, vol. 28. MIT Press, Cambridge, pp. 1–20.
Journal of Financial Economics 7 (3), 229–263. Whittle, J., Wharton, C., 1995. Optimizing cut-offs over time. In: Proceedings of the
Datamine, 2007. Studio 3 Help. 25th APCOM. Brisbane, Australia, pp. 261–265.