Professional Documents
Culture Documents
Resources Policy
journal homepage: www.elsevier.com/locate/resourpol
A BS T RAC T
Against a backdrop of decarbonisation in the energy sector and the increasing share of nuclear power in the
global power mix, availability of uranium resources presents a major challenge. Since it will be some time before
technologies of future nuclear reactors which remove the reliance on natural uranium are fully deployed, it is
relevant to analyse the availability conditions for uranium in the 21st century.
The first two conditions - technical accessibility and economic value - are associated with production cost. We
study these by modelling the ultimate uranium resources (including both identified and undiscovered resources)
and their cost. This method is based on a regional breakdown of global uranium resources, current known
deposits and an economic filter. It allows us to establish a long-term supply curve in which the quantities of
technically accessible uranium resources are represented as a function of production cost.
The other uranium availability conditions considered are associated with market dynamics, which created by
the relationship between supply and demand. These are modelled in the form of dynamic constraints in a partial
equilibrium market model. This is a deterministic model in which the market players are represented by
regions. It allows us to take account, for instance, of the short-term correlation between price and exploration
expenditure. In the longer term, the constraints modelled include anticipation of demand from electric utilities
(the consumers) and the increasing scarcity of the least costly ultimate resources.
Using a series of forward looking simulations, we demonstrate that the rate of growth in demand for uranium
in the 21st century and its anticipation have a major impact on the increase in price in the long-term.
Conversely, uncertainties related to the estimation of ultimate resources have a limited effect. Lastly, some
variations in supply (uranium production shutdown in a particular region, for example) or demand (irregular
increase) also have a significant impact on long-term price trends or cycles.
1. Introduction This model was used to quantify the technically accessible uranium
resources on a regional and worldwide scale and to estimate their
Determining the supply of natural uranium is a high priority for the associated production costs. The result is represented by a long-term
future of nuclear energy, in terms of competitiveness and sustain- cumulative supply curve (LTCS).
ability. Current generation of light water reactors (LWRs) benefits from In order to be able to develop a market model, we then identified
a significant competitive advantage over other centralised sources of the dynamic constraints impacting the long-term supply of uranium
electricity production in that their fuel represents a small percentage of (see Section 3). These constraints have a direct impact on either
their overall cost. In the future, if uranium resources become scarcer exploration or production. Analysis of annual exploration activity and
and their price rises sharply, this advantage may be lost, thus the associated discovery costs allowed us to introduce two key relation-
compromising the competitiveness of nuclear power over other cen- ships which were subsequently used in the market model developed.
tralised electricity production sources, or even compared with nuclear Two original uses of the R/P (reserves-to-production) ratio are also
technologies having a lower consumption rate of natural uranium. The introduced to model the constraints associated with anticipation of
study of uranium availability conditions in the 21st century has been demand and security of supply.
the subject of a 3-years research programme (Monnet, 2016) which is The study of the balance between supply and demand over time has
summarised in this article. helped to develop a uranium market model by considering that the
For the basis of this analysis, we first developed a model for long-term dynamic is a succession of short-term (annual) equilibria
estimating the "ultimate resources" of uranium (see Section 2). (see Section 4). This involves a deterministic economic model, in
⁎
Corresponding author.
E-mail address: sophie.gabriel@cea.fr (S. Gabriel).
http://dx.doi.org/10.1016/j.resourpol.2017.07.008
Received 14 March 2017; Received in revised form 31 May 2017; Accepted 27 July 2017
Available online 29 August 2017
0301-4207/ © 2017 Elsevier Ltd. All rights reserved.
A. Monnet et al. Resources Policy 53 (2017) 394–407
Fig. 1. Cumulative global uranium consumption scenarios (Baschwitz et al., 2009). q = q0 ∬ f (g, t )dgdt (1)
partial equilibrium given that it only takes account of the uranium q0 is estimated from the mass of rock M in the geological environment
market. and the mean grade of the crust clarke (q0 = M × clarke). Note that q0
This is the market model that was ultimately used for the forward does not include any consideration of economic value or recovery rate.
looking simulations based on exogenous demand scenarios (see The numerical method used to determine the parameters of the
Section 5). statistical distribution of grade and tonnage f(g,t) is inspired by the
Our analysis was not based on modelling demand scenarios as this work of Drew. It uses a cost-grade-tonnage relationship as an economic
involved too many underlying assumptions, hence we adopted a filter (see below). The form of this relationship is inspired by the work
literature review based approach (IIASA et al., 1998). Two demand of Drew and Harris, however its calibration is a separate procedure and
scenarios were subsequently selected: scenario A3, representing high the cost limit used in the filter is specific to our study. To model the
global demand for nuclear power (5,400 GWe installed capacity in costs for all deposits (known and undiscovered), our study proposes a
2100) with a consequently high demand for natural uranium (810 ktU/ method which is more representative of the industrial reality than that
year); and scenario C2, representing moderate demand (2,100 GWe used by Drew and Harris. The numerical procedure used to calculate
in 2100 and 340 ktU/year). No scenarios representing low growth or a the cumulative resources at less than a given cost is also specific to this
reduction in the world's nuclear fleet were considered based on their study.
limited relevance to the study (the availability of uranium resources is It is worth mentioning here that our model is based on the
not a constraint in either of these scenarios). assumption that uranium is the primary product, i.e. that it incurs all
the investment and production costs associated with mineral extraction.
2. Modelling ultimate resources
2.1.1.1. Lognormal distribution of grade and tonnage. It is common,
Several bivariate and multivariate statistical models can be found in although sometimes criticised, to assume that f follows a bivariate
the literature to estimate the abundance and production costs asso- lognormal distribution (Harris, 1977) which results in the
ciated with a non-ferrous metal such as uranium. The statistical models mathematical form of f described by Eq. (2), where grade g and
give an economic assessment at deposit level then provide a sum of all tonnage t are assumed to be independent variables.
the deposit resources. One advantage of this approach is that models ⎛ ⎛ ⎞2 ⎞
can be specific to each geological environment. Among the models ⎜ (lng − μx )2 ⎜lnt − μy ⎟ ⎟
⎝ ⎠
reviewed in the literature, three have been applied to uranium. They exp⎜ − − ⎟
2σx2 2σy2
⎜ ⎟
were developed by Chavez-Martinez (1982), Harris (1977, 1984, 1988), ⎝ ⎠
f (g , t ) =
Drew (1977) and Brinck, (1967, 1971), De Wolde et al. (1971) at the 2πgtσxσy (2)
end of the 1970s, early 1980s. None of them were used to establish a
complete long-term cumulative supply (LTCS) curve of world re- µx, σx2and µy, σy are the means and variances of x = ln(g) and y =
2
sources, however; they were essentially used to estimate potential ln(t) respectively.
reserves (undiscovered resources which could be produced at a lower In a given geological environment, the statistics for g and t, taken
cost than the market price at the time the studies were conducted) (in from known deposits, can be used to determine the statistical
the case of Drew and Brinck) or to estimate the ultimate resources in distribution of grade and tonnage. Unfortunately, deposits are not
different cost categories, but within a specific region of the world (in the sampled randomly. On the contrary, the richest deposits (high grade,
case of Drew and Harris). high tonnage) tend to attract economic interest first.
In addition to these models, a more advanced method, called
Quantitative Mineral Resource Assessments, has been developed by 2.1.1.2. Economic filter. To correct the sampling bias that affects
the United States Geological Survey (USGS) and applies to all non- known deposits (Harris, 1984), the idea is to model an economic
ferrous metals (Singer, 2007). One additional purpose of this method is filter. This filter is a function which truncates the probability density
to locate undiscovered resources: within a geological environment, function of deposits and separates them into observable and non-
subregions are delimited and data from the Geographic Information observable deposits, according to whether they are above or below a
System (GIS), available in well-explored subregions, are analysed to given cost limit.
find similarities with unexplored or less well known subregions.
This aspect of locating undiscovered resources is not included in With this filter, the empirical data available corresponds mainly to
our approach and our model has been inspired more by the work of observable deposits (known deposits may nevertheless be subeconomic
Harris and Drew than the more recent studies by the USGS. due to the partially random exploration characteristic). Because of
Our methodology allows for an estimation of ultimate uranium truncation, the grade and tonnage of observable deposits no longer
395
A. Monnet et al. Resources Policy 53 (2017) 394–407
follow a lognormal distribution. Instead, they follow a truncated costs) (see (CRU International.) for example). This is the premise on
lognormal distribution, whereby the truncation limits (glim and tlim) which pre-feasibility studies estimate the costs by extrapolating costs
are associated with a cost limit through a cost-grade-tonnage relation- from mines or recent projects2 in proportion to production capacities
ship which characterises the economic filter and which is defined later. and according to exponential relationships (Wellmer et al., 2008).
When glim and tlim are known, the probability density functions for Given that the cost-grade-tonnage relationship explained
truncated distributions have an explicit form which can be related to above does not take account of an optimum processing capacity, a
that of non-truncated distributions (Harris, 1984). (The mathematical separate cost relationship is introduced to estimate the LTCS curve
expressions are given in the Appendix.) in the next step in our model, which provides a more realistic cost
estimate for the industry. The costs of deposits (known or undiscov-
2.1.2. Cost-grade-tonnage relationship ered) are then estimated based on their optimum ore processing
It is assumed that the production cost of a deposit depends on the capacity Kmill (Taylor's rule (Taylor, 1986)) and an exponential cost-
grade and tonnage of its mineral as shown below (Eq. 3): capacity relationship calibrated on the basis of data from recent
mining projects in the same way as for the cost-grade-tonnage
⎛ g ⎞ βg ⎛ t ⎞ βt
c(g , t ) = c0⎜⎜ ⎟⎟ ⎜ ⎟ relationship.
⎝ g0 ⎠ ⎝ t0 ⎠ (3) Kmill (t ) = 0.0143 × t 0.75(in tonnes of ore /year ) (8)
This cost-grade-tonnage relationship can also be written
⎧ ⎛ ⎞ βcc ⎧ ⎛ ⎞ βop
according to Eq. (4), where x = ln(g), y = ln(t), and A is a constant. ⎪ CC (t ) = CC × ⎜ Kmill(t ) ⎟ ⎪ OP(t ) = OP × ⎜ Kmill(t ) ⎟
ln(c(g , t )) − A = βgx + βy ⎨ 0
⎝ Kmill ⎠
0
and ⎨ 0
⎝ Kmill ⎠
0
t (4) ⎪ ⎪
⎪
⎩ ln( CC ) = β × ln( K ) + B ⎩ ln( OP ) = β × ln( Kmill ) + Bop
cc mill cc op (9)
Our model uses this relationship to define the economic filter
required to determine the unbiased function f(g,t).
2.1.4. Plotting the LTCS curve (cumulated resources for each cost
2.1.2.1. Calibration of the cost-grade-tonnage relationship. For a category)
given deposit, the full cost is defined based on investment (capital) The cumulated resources available below a certain production cost,
costs CC, operating costs OP,1 processing capacity Kmill, lifetime LT ( = C1, are calculated using the following intermediate calculations: the
t/Kmill), development time DT and the discount rate a (Eq. 5). numerical calculation of N, the total number of deposits in the
0 LT (t ) environment (total mass of rock M divided by the mean tonnage of
CC (Kmill ) Kmill × OP(Kmill )
Ctot (g , t ) = ∑ + ∑ all deposits), and m(C1), the mean metal content of deposits which are
i =−DT (DT +1)(1+a )i i =1 (1+a )i (5) "cheaper than C1". Eq. (10) and Eq. (11) give the analytical expression
Ctot can be calculated on the basis of data from operating mines or of N and m(C1) in terms of their mathematical expectation.
recent projects, the costs of which are known and available in a public M
N= ∞
database (WISE, 2015).
The parameters βg, βt and the constant A in Eq. (4) are then ∬ tf (g, t )dgdt
0 (10)
obtained through optimisation so that c(g,t) is as close as possible to
LT (t )
Ctot (g, t )/(gt ∑i =1 (1+a )−i ).
m(C1) = ∫ ∫c(g,t )≤C gtf (g, t )dgdt
1 (11)
2.1.2.2. Use of the cost-grade-tonnage relationship. Once calibrated,
Finally, the LTCS curve is created by plotting the function C1 → N ×
the cost-grade-tonnage relationship is used for the economic filter. In
P(C1) × m(C1).
practice, this involves fixing a unit limit cost Clim equal to the current
long-term price of uranium, which corresponds to the cost limit above
which deposits are too costly and non-observable, and then 2.2. Applying the model
determining the grade and tonnage limits glim and tlim below which
deposits are not observed based on the following expressions: 2.2.1. Regional breakdown
The regional breakdown used to estimate ultimate resources
⎛ ⎞ consists of 6 subregions: USA, Canada, Africa, Australia,
glim = exp⎜(ln(C )−A − βt ln (t ))/ β ⎟
⎝ g⎠
(6) Kazakhstan and the rest of the world. These regions were selected
on the basis of the following criteria:
⎛⎛ ⎞ ⎞
tlim = exp⎜⎜⎜ln(C ) − A − βg ln (g)⎟ / β ⎟⎟
⎝ ⎝ ⎠ t⎠ (7) ● Representativeness: the top 5 subregions account for almost 85% of
world production and close to 80% of reasonably assured resources
(RAR) at < USD 130/kgU in 2013 (OECD NEA, 2014).
2.1.3. Cost-capacity relationship ● Availability of data on known deposits and recent mining projects.
When the resources of a deposit are estimated, the lifetime, or the ● Minimal variability in the types of deposit encountered and a certain
equivalent production capacity, is one of the first parameters selected degree of standardisation of mining regulations within the regions to
by industry players when scaling a mining project. It is an economic ensure relative consistency when estimating the costs of ultimate
choice which comes into play very early on in a project and which calls resources.
upon empirical rules, such as Taylor's rule (Taylor, 1986), to
determine the optimum production level for a mine in order to achieve The cost calculation method (cost-capacity relationship) was used
economies of scale (Queen'’s University, 2015; Taylor, 1986; Wellmer for each subregion, taking account of the specific regional character-
et al., 2008). The recent project benchmarks provide evidence of a istics in terms of mining techniques. Hence, for Africa, which is
correlation between production capacity and investment costs (mainly specialised in open pit mining, we decided to use extraction capacity
fixed costs) or, to a lesser extent, operating costs (fixed and variable
2
The greater the similarity between the benchmark mines and the project studied
1
Operating costs are expressed here in units of mineral processed and not in units of (same techniques involved, similar grades, same mineralogy, same legislative and
uranium produced. environmental constraints, etc.), the more relevant the extrapolation.
396
A. Monnet et al. Resources Policy 53 (2017) 394–407
Table 1
Summary of regional estimates of ultimate resources.
Region RUM model (MtU) 2013 Red Book (RAR+IR+PR+SR) (MtU) 2013 Red Book (RAR+IR) (MtU)
< 80 < 130 < 260 < 80 < 130 < 260 < 80 < 130 < 260
USD/kgU USD/kgU USD/kgU USD/kgU USD/kgU USD/kgU USD/kgU USD/kgU USD/kgU
USAa 0.37 0.97 3.56 0.88 2.34 2.60 0.039 0.21 0.47
Canada 0.91 1.53 2.75 0.47 1.34 1.50 0.42 0.49 0.65
Africa 4.61 10.49 31.22 0.28 1.53 1.91 0.24 1.38 1.60
Australia a 3.59 7.25 18.72 NA NA NA 1.35 1.66 1.74
Kazakhstan 3.53 5.71 11.09 0.73 1.35 1.58 0.52 0.68 0.88
Rest of World 0.76 1.64 4.70 1.10 3.63 5.07 0.74 1.44 2.23
Totalb 13.8 27.6 72.0 2.62 9.77 12.23 1.96 5.90 7.64
a
Data from 2011 Red Book (2013 data incomplete)
397
A. Monnet et al. Resources Policy 53 (2017) 394–407
Table 2
Results of the estimated discovery cost-cumulative exploration expenditure correlations.
log(EXPLO(t )) = 1.15 × log(p(t − 1)) + 8.37(R 2 = 0.76) Fig. 4. Correlation between discovery cost and cumulative exploration expenditure.
(0.10) (0.38) (12)
398
A. Monnet et al. Resources Policy 53 (2017) 394–407
There is a strong correlation in all the regions studied (R2 > 0.9 of reserves associated with the financial guarantees they must have in
except in Australia where R2 = 0.8). On a global level, the correlation is order to continue to develop their production facilities.
significant, but less marked (R2 = 0.8). Correlation is weak in the rest of The political choices made by individual governments can also
the world region (R2 = 0.14 with a slight fall in discovery cost as affect the value of the regional R/P ratio. The R/P ratio of a private
exploration continues, which is explained by incomplete historical producer or a producing region nevertheless remains very closely
data) such that it is preferable to apply the relationship established for linked to the world market given that, on the one hand, production P
the whole world in this region. contributes to supply which balances world demand, and, on the other,
market price stimulates world exploration expenditure which leads to
3.3. Security of supply and anticipation of demand the discovery of resources R at a local level.
We have therefore adopted a minimum R/P ratio value of 10 years
The reserves-to-production ratio, normally written as R/P, is an in our model.
indicator of availability of a non-renewable resource used primarily in
the oil and gas industry (British Petroleum, 2015). Its simplest and 3.3.3. Modelling and scarcity rent
most widely used interpretation states that at a rate of production P, In our model, demand is exogenous and entirely satisfied by mining
the industry can continue to produce the resource for R/P years living production; applying a constraint to the R/P ratio means that the level
on the reserves R already identified. This interpretation is often of identified resources R is constrained. Thus, at a time t, the expected
criticised, though, as it can lead to false prospective analyses. In fact, year-to-year R/P ratio is given by Eq. (14), where RES_LTCS(t) is the
it ignores two essential aspects for the long-term availability of the discovered resources.
resource: production fluctuations on the one hand (there is no EXPLO(t ) EXPLO(t +1)
guarantee that production will remain constant in the long-term if it RES _LTCS (t −1) + DISC _COST (t )
− D (t ) + DISC _COST (t +1)
R / P(t +1) =
follows demand), and the discovery of new resources on the other. D(t +1)
We propose instead two original interpretations of the R/P ratio. ≥ R / Pmin (14)
On a global level, R/P is a simple indicator of scarcity, which can
represent a constraint associated with the anticipation of world Exploration in year t+1 (EXPLO(t+1)) depends on the price at time
demand (see section 0). On a regional level, this ratio can be t, p(t). If the constraint in Eq. (14) is not verified, i.e. that the perceived
interpreted further as the result of a producer's technical, budgetary scarcity of the resource exceeds its critical value, a retroactive increase
and financial constraints or a regional strategy (see section 0). of the price is applied until the additional exploration expenditure
EXPLO(t+1) is sufficient to satisfy this constraint. This type of
3.3.1. World R/P ratio: Anticipation of demand and the need for modelling introduces the notion of scarcity rent: we do not apply a
visibility by electric utilities constant growth rate to rent as expounded in Hotelling's rule, rather we
To interpret the R/P ratio on a global level, it is assumed that the use the minimum increase required to maintain supply security
long-term annual production P equals annual demand D. (represented by the world R/P ratio) at its minimum value.
Mining producers are more sensitive to expected demand, sometimes
even decades ahead, whereas consumers, who are essentially electric 4. Models of the uranium market
utilities, are more sensitive to supply. The construction of new reactors
sends a strong signal to the mining industry about future demand and Based on the supply constraints described earlier, which may be
thus about the production capacity needs. Although demand is exogenous short-term (such as the causal link between price and exploration
in our market modelling (see Section 4), it is important to understand the expenditure) or longer term (such as anticipation of demand), we will
signals that consumers send to the market to stimulate the production study the long-term availability of uranium using a succession of short
process. We are therefore interested in their supply strategies. or medium-term economic balances.
Electric utilities secure their supply in the short-term by using The proposed model is a partial equilibrium economic model, i.e.
strategic stocks of several years and in the longer term through restricted to a single market as it does not take account of forces
contracts negotiated with producers. Some utilities also seek to secure outside the uranium market such as the markets for other commod-
their supply into the even longer term by investing directly in mining ities, in other sectors of the economy (inflation, discount rate and other
projects, sometimes as soon as construction of a new plant is under macro-economic factors), or even in the non-economic context.
way. This need for supply security is reflected in a need for visibility on To introduce this model gradually, we present 3 separate market
mining resources and is explained by the capital intensity of reactors mechanisms, gradually incorporating these constraints:
and their long lifetime.
In a hypothetical situation where an extreme increase in demand ● M1: perfect competition model, which does not incorporate any
pushes the R/P ratio close to 0, the risk of a supply shortage should be dynamic constraints applying to supply (see Section 4.1).
considered. Conversely, in a situation of rising demand, a stable R/P ● M2: model incorporating exploration constraints which create short
ratio that exceeds several decades is a sign of managed risk. In practice, or medium-term economic balances (see Section 4.2).
it is less likely that shortages will arise unexpectedly, particularly given ● M3: model incorporating dynamic constraints enabling the long-
that nuclear reactors have a very predictable fuel consumption. term supply-demand balance to be studied using a succession of
These considerations mean that the world R/P ratio can be short and medium-term balances (see Section 4.3).
interpreted as a macro-economic indicator of the scarcity of a resource
as it is perceived at a given moment. Historically, this indicator has The M3 model, which is the most comprehensive, and which we
never fallen below 60 years, hence we have used this value as the believe to be the most relevant, is subjected to a sensitivity analysis to
default lower value in our model. measure the influence of each parameter. The three mechanisms are
then compared (see Section 4.4).
3.3.2. Regional R/P ratio: economic constraints of producers and
policy decisions of individual states 4.1. M1: Perfect competition with no constraints
On a regional level or for a producer, P represents a production
capacity rather than the exogenous global demand or even local demand. The first mechanism (M1) does not incorporate any constraint. All
A producer's R/P ratio is constrained by a minimum production value the ultimate resources are assumed to have been identified and fully
for self-funding their exploration activity, as well as by a minimum value available from the outset (no constraints on the production capacities).
399
A. Monnet et al. Resources Policy 53 (2017) 394–407
Fig. 6. M1: Price trends (A3 and C2 72 MtU). Fig. 7. M2: Price trends (A3 and C2 72 MtU).
400
A. Monnet et al. Resources Policy 53 (2017) 394–407
Each region has its own discovery cost, whereas the exploration
expenditure and anticipation of demand constraints are global. The
global exploration expenditure is deducted from Eq. (12) then dis-
tributed by default in each region in proportion to their profits. Each
region “defends” its market share: if demand increases at t, each
region's production increases in proportion to its market share at t-1. If
a region is short of resources (regional R/P ratio reaching its threshold
value R/Pmin, loc) and therefore unable to maintain its market share, the
regional production is limited by the local constraint and the residual
Fig. 9. M3: Resource rent trends (A3 and C2 72 MtU).
demand is distributed across the remaining regions. In each region,
production (respectively exploration) is carried out by increasing cost
from among the known resources (respectively in proportion to the Two marked trends can be seen in prices for scenario A3: the price
remaining ultimate resources for which the cost is lower than the global remains fairly stable in the short and medium term (around the 2013
marginal production cost). At all times, the price is set at the global levels), then increases significantly after 2035 (Fig. 8). In fact, this
marginal cost, i.e. the highest marginal cost from among the regions change occurs when the global R/P ratio reaches its 60-year threshold.
that are producing, which introduces a differential rent. If the In this situation, the M3 mechanism introduces an increasing scarcity
critical level for security of supply (R/P = R/Pmin) is reached globally, rent to keep the R/P ratio at its minimum level (Fig. 9). It can be seen
a scarcity rent is added to the global marginal cost to establish the by comparing the two types of rent that the scarcity rent appears and
price. becomes dominant at that moment. Previously, there were only
The market is modelled as an oligopoly of 6 regions with explora- differential rents. Over the period of the scenario, the growth rate of
tion and anticipation constraints (the market players are combined and the differential rents is lower than that of the scarcity rent.
modelled together in these regions), but no rent other than the scarcity For scenario C2, the R/P ratio reaches its threshold later and the
rent or the differential rents is included in the market price to represent price increase is lower.
a possible dominant position of any one player. The term oligopoly is The following sections cover the sensitivity analysis carried out in
thus used in the literal sense (a small number of producers) without order to gain a better understanding of the influence of certain
assuming any collusion between the players (we are not planning to parameters and the effect of regionalisation of the players.
model a cartel).
The proposed model removes the main limits we outlined in the 4.3.1. Anticipation of demand constraint
previous section: The global R/P > R/Pmin constraint conveys some anticipation of
demand by the market and a need for visibility on the availability of
● Ultimate resources (which includes undiscovered resources) are uranium for consumers. By default, R/Pmin = 60 years has been
taken into account and exploration activity is endogenous. retained as the most likely assumption.
● Dynamic constraints are applied to the production capacities. Fig. 10 shows the long-term price dynamics obtained with the M3
● Resource rents (scarcity rent or differential rents) are introduced mechanism for scenario A3 – 72 MtU. Different price trends can be
and calculated explicitly in the model. seen according to the level of anticipation of demand.
● The market players are modelled at regional level.
● If there is no anticipation (no R/Pmin constraint), the shortage
Fig. 8 and Fig. 9 show the price and resource rent trends for already observed with the M2 mechanism occurs again, but is
demand scenarios A3 and C2. delayed (2100 as against 2080 with M2). The slower decrease in
Fig. 8. M3: Price trends (A3 and C2 72 MtU). Fig. 10. M3: Effect of the anticipation of demand constraint (A3 72 MtU).
401
A. Monnet et al. Resources Policy 53 (2017) 394–407
It can be seen in Fig. 11 that this constraint does not have any
significant long-term effect. The detailed results show that this
constraint affects the regional production costs and the market shares
of the regions and therefore the differential rents, but not the scarcity
rent which is dominant for setting the long-term price.
Fig. 13. M3: Effect of distribution of the exploration expenditure (A3 72 MtU).
402
A. Monnet et al. Resources Policy 53 (2017) 394–407
Fig. 14. M3: Effect of the demand profile (A3 and C2 72 MtU).
4.3.7. Summary
The sensitivity analysis has shown that anticipation of demand and
the need for visibility for consumers, represented by the constraint on
the global R/P ratio, directly affect price trends, by influencing the
scarcity rent.
Some parameters, such as the regional production constraint, the
parameters representing the dynamics of identifying new resources, or Fig. 15. Comparison of the market mechanisms: price trends (A3 and C2 72 MtU).
403
A. Monnet et al. Resources Policy 53 (2017) 394–407
Two supply scenarios have been chosen here: the first with
estimated ultimate resources of 72 MtU at < USD 260/kgU, and the
second with just half, 36 MtU at < USD 260/kgU.
Fig. 16 shows the price trends with these two supply scenarios, for
the two demand scenarios A3 and C2, and Table 3 shows the dates
when the critical value of R/Pmin is reached.
In the case of increasing demand scenarios, the following observa-
tions can be made:
404
A. Monnet et al. Resources Policy 53 (2017) 394–407
of the grade and tonnage of deposits, on the use of an economic filter and
on cost functions which take account of economies of scale (principally
associated with the tonnage of the deposits and their production capacity)
and the envisaged type of mining. The model incorporates the available
data and the specific economic and geological characteristics of each
region in the best possible way by calibrating these distributions and cost
functions differently.
Despite significant uncertainties, the results obtained have shown
that, due to some regions’ specialisation in particular mining techni-
ques and/or their specific economic and geological characteristics, the
ultimate resources of uranium are distributed unevenly throughout the
world. The discount rate, the depth of the earth's crust taken into
account or the price of uranium representative of current economic
conditions are all particularly decisive parameters in the estimation
carried out. However it has also been shown that this estimation only
has a second-order influence on long-term uranium price trends.
The analysis of the structure of the market and its dynamic
Fig. 18. M3: Effect of doubling Kazakhstan's production in 2040. constraints has enabled us to define a new model for studying the
market. This is a deterministic model which calculates a series of short-
and also the dominant production technique (in-situ leaching) techni- term economic balances in order to carry out a long-term prospective
cally enables new production capacities to be developed quickly and study. This model takes into account the causal link between price and
with relatively low amounts of investment. exploration expenditure, increases in discovery costs and anticipation
According to the results (Fig. 18), doubling the production of a of demand. The regionalisation of the market players (modelled by an
country such as Kazakhstan would introduce price fluctuations in the oligopoly with no collusion) enables differential rents to be introduced
short and medium term, but would have little long-term effect. which have a limited short and medium-term effect in the increasing
These forward looking simulations have enabled us to identify demand scenarios that have been studied; while the demand anticipa-
determining factors for the availability of uranium, by testing some tion constraint introduces a scarcity rent, enabling the margins on
assumptions in possible scenarios. The results show that the model that security of supply to be maintained, but also leading to a significant
has been developed is robust with regard to more or less sudden supply increase in the price of uranium in the long-term.
variations: the response of the model in terms of price agrees on a long- A sensitivity study has revealed the particular importance of the
term trend (identical or offset in relation to the reference situation) in demand scenarios, the demand anticipation constraint and even the
all the situations studied. regionalisation of the discovery costs: they all have a significant effect
However, this model does not enable the uranium demand of the on price trends by influencing the scarcity rent.
different regions to be individualised (the demand is global). The The value of the market model that has been developed lies in its
available input data (global nuclear electricity production scenarios) flexibility and the range of possibilities it offers for carrying out
has led to this choice of model. This is a useful improvement which prospective studies. Firstly, it enables resource rent trends and a large
could be made to the model, as it would enable specific strategies to be number of regional or global indicators (R/P ratio, production costs,
taken into account for regions that are “producers and consumers” or exploration expenditure, etc.) to be monitored, which makes it easier to
“producers but not consumers” or “consumers but not producers” of interpret the results: for example, price rises due to an increase in
uranium. production costs, or price rises due to scarcity of identified resources,
The model that has been developed no longer takes account of the can be clearly seen. Secondly, the structure of the model is flexible
two main forms of contract on which today's uranium exchanges are enough to enable a variety of scenarios to be simulated (supply or
based: spot contracts and long-term contracts. It only produces a single demand scenarios).
price indicator. This limits the interpretation of the results, creating a However there are still some uncertainties in the estimation of the
certain cyclicality of prices as the electric utilities seem to limit the use production costs and the differential rents because mines are not
of spot contracts (for which the price index is the most volatile) when discretised in the model. For the scenarios studied which correspond to
there are price spikes in order to hedge against these fluctuations. a growing demand for uranium in the long term, scarcity rent generally
Finally, the forward looking simulations analysed in this section becomes preponderant over differential rent and this approximation is
have the same limitations as all prospective studies: they cannot be limited. On the other hand, for volatile demand scenarios, the model
exhaustive. It must therefore be remembered that the results of the would need to be improved on this point.
preceding simulations must be used with caution. In particular, they Finally, it is important to remember that there are some limitations
are only valid for nuclear electricity production scenarios and increas- on the interpretation of the results.
ing uranium demand. The results on uranium price trends can provide information for
consideration of the competitiveness of nuclear energy and the
6. Conclusion competitiveness of fast reactors in relation to current nuclear technol-
ogies. However in the absence of an additional analysis of trends in
This analysis of the key parameters for the long-term availability of other costs (cost of other electricity production methods, nuclear
uranium resources is based on modelling the ultimate uranium energy investment and operating costs), it is not possible to draw any
resources (discovered and undiscovered resources and those already conclusions on the changing competitiveness of nuclear energy.
mined) and modelling the market. Besides, the scenarios envisaged are far from exhaustive. For
A new methodology for estimating the ultimate resources and their example, we have implicitly assumed that the current market economy
associated production costs has been developed using available data on will continue in the 21st century and that there will be no technological
known uranium deposits. The model is based on a lognormal distribution breakthroughs in uranium production methods.
405
A. Monnet et al. Resources Policy 53 (2017) 394–407
Appendix
The probability density functions for truncated distributions have an explicit form which can be related to that of non-truncated distributions
(Harris, 1984). After several mathematical conversions, the mathematical expectations (mean values) for grade, tonnage and metal content (γg, γt,
and γm respectively) among the truncated population elements can be expressed as a function of the unknowns µx, σx2 and µy, σy2 (see to ):
⎛ ⎛ 2⎞
Clim 1 c−μ ⎞
exp (μx + σx2 /2) ∫ exp⎜⎜ − 2 ⎜ σ c ⎟ ⎟⎟dc
⎝ ⎝ c ⎠⎠
−∞
γg =
C
⎛ ⎛ c − μ ′ ⎞2 ⎞
∫−∞lim exp⎜⎜− 12 ⎜ σc c ⎟ ⎟⎟dc
⎝ ⎝ ⎠⎠ (15)
⎛ ⎞ Clim ⎛ ⎛ 2⎞
1 c − μ′′′ ⎞
exp ⎜μy + σy2 /2⎟ ∫ exp⎜⎜ − 2 ⎜ σ c ⎟ ⎟⎟dc
⎝ ⎠ −∞
⎝ ⎝ c ⎠⎠
γt =
C
⎛ ⎛ c − μ ′c ⎞2 ⎞
∫−∞lim exp⎜⎜− 12 ⎜ ⎟ ⎟dc
⎝ ⎝ σc ⎠ ⎟⎠ (16)
C⎛ c − μc′′ ⎞2 ⎞
⎛
⎟ ⎟⎟dc
∫−∞lim exp⎜⎜− 12 ⎜
⎝ ⎝ c ⎠⎠
σ
⎛ 2 2
⎞
γm = exp⎜μx + σx /2 + μy + σy /2⎟ ×
⎝ ⎠ ⎛ ⎛ 2⎞
C c− μ′ ⎞
∫−∞lim exp⎜⎜− 12 ⎜ σc c ⎟ ⎟⎟dc
⎝ ⎝ ⎠⎠ (17)
Where:
μc = βμ
t y
+ βg(μx + σx2 )
γg, γt and γm correspond to the theoretical mean values of the variables according to the truncated distributions (i.e. when bias exists and is not
corrected). Their empirical equivalents are thus the estimators g , t , m . If γg, γt and γm are replaced by their empirical estimates in equations to , the
system contains 3 equations and 4 unknowns. It can be solved using the additional constraint of . The quadruplet solution (µx, µy, σx, σy) can be
found numerically using an optimisation routine which minimises the error Δ defined in
⎛ γg ⎞2 ⎛ γ ⎞2 ⎛ γ ⎞2
∆ = ⎜1 − ⎟ + ⎜1 − t ⎟ + ⎜1 − m ⎟
⎝ g⎠ ⎝ t⎠ ⎝ m⎠ (19)
References Drew, M. W.: US uranium deposits: A geostatistical model. In: Resources Policy Bd. 3
(1977), Nr. 1, S. 60–70–.
Harris, DeVerle P., 1977. Quantitative methods for the appraisal of mineral resources
Baschwitz, A., Loaec, C., Fournier, J., Delpech, M., Legée, F., 2009. Long term (Progress report Nr. GJBX-14-77) : University of Arizona, Department of Mining &
prospective on the electronuclear fleet: from GEN II to GEN IV. In: Proceedings of Geological Engineering.
the Global 2009 congress - The Nuclear Fuel Cycle: Sustainable Options & Industrial Harris, De.Verle P., 1984. Mineral Resources Appraisal: Mineral Endowment, Resources,
Perspectives. Paris, France. and Potential Supply : Concepts, Methods and Cases. Oxford University Press, ISBN
Brinck, Johan W.: Calcul des ressources mondiales d’uranium. In: Bulletin de la 978-0-19-854456-2.
communauté européenne de l’énergie atomique Bd. 6 (1967), Nr. 4, S. pp. 109–114. Harris, De.Verle P., 1988. Geostatistical crustal abundance resource models. In: Chung,
Brinck, Johan W., MIMIC - The prediction of mineral resources and long-term price C.F., Fabbri, A.G., Sinding-Larsen, R. (Eds.), Quantitative Analysis of Mineral and
trends in the non-ferrous metal mining industry is no longer Utopian. In: Energy Resources, NATO ASI Series. Springer, Netherlands, 459–488.
Eurospectra Bd. 10 (1971), Nr. 2, S. pp. 46–56. IAEA: World Distribution of Uranium Deposits (UDEPO). URL 〈http://infcis.iaea.org/
British Petroleum: Statistical review of the world energy 2015 : British Petroleum, 2015. UDEPO/UDEPOMain.asp?Order = 1&RPage = 1&Page = 1&RightP = List〉. -
Cameco: Uranium Price. URL 〈https://www.cameco.com/invest/markets/uranium- abgerufen am 2013-11-27. — International Atomic Energy Agency.
price〉. - abgerufen am 2016-04-20. — Cameco Website. IIASA, WEC, McDonald, A., Grübler, A. (Hrsg.): Global Energy Perspectives,
Chavez-Martinez, Mario Luis: A potential supply system for uranium based upon a International Institute for Applied Systems Analysis : Cambridge University Press,
crustal abundance model, 1982. 1998 — ISBN 978-0-521-64569-0.
CRU International: An Overview of the Uranium Mining Industry of Kazakhstan. Monnet, Antoine: L’uranium dans une nouvelle ère : vers un marché très concentré ? In:
De Wolde, H. I. ; Brinck, Johan W.: The estimation of mineral resources by the computer La lettre de l’I-tésé (2014), Nr. 23, S. 7–10.
program „IRIS“ ( Nr. EUR4607e) : Commission of the European Communities, 1971. Monnet, Antoine: Disponibilité à long terme des ressources mondiales d’uranium,
406
A. Monnet et al. Resources Policy 53 (2017) 394–407
407